Alumina Preparia

SHREIR’S CORROSION
This page intentionally left blank

SHREIR’S CORROSION
FOURTH EDITION
Volume 1
BASIC CONCEPTS, HIGH TEMPERATURE CORROSION
Editors
R A Cottis
M J Graham
R Lindsay
S B Lyon
J A Richardson
J D Scantlebury
F H Stott
CorrosionandProtectionCentre,,UK
Amsterdam • Boston • Heidelberg • London • New York • Oxford

Paris • San Diego • San Francisco • Singapore • Sydney • Tokyo
Academic Press is an imprint of Elsevier
Elsevier B.V.
Radarweg 29, 1043 NX Amsterdam, The Netherlands
# 2010 Elsevier Ltd. All rights reserved
First Edition 1963
No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by
any means electronic, mechanical, photocopying, recording or otherwise without the prior written
permission of the publisher
Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in
Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333; email: permissions@elsevier.com.
Alternatively you can submit your request online by visiting the Elsevier web site at http://elsevier.com/
locate/permissions, and selecting obtaining permission to use Elsevier material
British Library Cataloguing in Publication Data

A catalogue record fro this book is available from the British Library
Library of Congress Catalogue Number: 2009940135
ISBN: 978-0-444-52788-2
For information on all Elsevier publications

visit out website at books.elsevier.com
Printed and bound in Spain
10 11 12 13 14 10 9 8 7 6 5 4 3 2 1
Contents
Contributors vii CHAPTER 1.08

Stress Effects in High Temperature Oxidation 153
Editorial Board ix M. Schütze
Contents of all volumes xiii CHAPTER 1.09
Foreword xxi Thermodynamics and Theory of
External and Internal Oxidation of Alloys 180
Preface xxiii B. Gleeson
Tribute to L. L. Shreir xxv PART III
TYPES OF HIGH TEMPERATURE
CORROSION
PART I
CHAPTER 1.10
BASIC CONCEPTS
Oxidation of Metals and Alloys 195
CHAPTER 1.01 P. Y. Hou
Chemical Thermodynamics 1
CHAPTER 1.11
R. A. Cottis
Sulfidation and Mixed Gas Corrosion of
CHAPTER 1.02 Alloys 240
Electrochemistry 13 R. John
C. Lefrou, R. P. Nogueira, F. Huet, and H. Takenouti
CHAPTER 1.12
CHAPTER 1.03 Carburization and Metal Dusting 272
Outline of Structural Metallurgy D. J. Young
Relevant to Corrosion 52 CHAPTER 1.13
R. P. M. Procter Nitridation of Alloys 304
U. Krupp
CHAPTER 1.04
Mechanical Properties and Fracture of CHAPTER 1.14
Materials 77 Corrosion in Molten Salts 316
A. H. Sherry and T. J. Marrow M. Spiegel
CHAPTER 1.05 CHAPTER 1.15
Basic Concepts of Corrosion 89 High Temperature Tribocorrosion 331
L. L. Shreir I. A. Inman, P. S. Datta, H. L. Du, C. Kübel, and
P. D. Wood
PART II
PART IV
PRINCIPLES OF HIGH HIGH TEMPERATURE CORROSION
TEMPERATURE CORROSION
ENVIRONMENTS
CHAPTER 1.06
CHAPTER 1.16
Defects and Transport in Oxides and
Types of Environments 399
Oxide Scales 101 B. A. Baker
B. Pieraggi
CHAPTER 1.17
CHAPTER 1.07 Oxidation in Steam and Steam/Hydrogen
Mechanisms and Kinetics of Oxidation 132 Environments 407
S. Chevalier W. J. Quadakkers and J. Żurek
v
vi Contents

Fireside Corrosion 457 High Temperature Corrosion of
N. Otsuka Chromia-forming Iron, Nickel and
Cobalt-base Alloys 583
CHAPTER 1.19
A. Galerie
High Temperature Corrosion Issues for
Metallic Materials in Solid Oxide Fuel CHAPTER 1.24
Cells 482 High Temperature Corrosion of
L. Singheiser, P. Huczkowski, T. Markus, Alumina-forming Iron, Nickel and
and W. J. Quadakkers Cobalt-base Alloys 606
B. A. Pint
CHAPTER 1.20
Gas Turbine Oxidation and Corrosion 518 CHAPTER 1.25
N. J. Simms and J. R. Nicholls High Temperature Corrosion of
Intermetallic Alloys 646
PART V J. W. Fergus
HIGH TEMPERATURE MATERIALS
CHAPTER 1.26
CHAPTER 1.21 High Temperature Corrosion of
Design of High Temperature Alloys 541 Ceramics and Refractory Materials 668
P. F. Tortorelli and M. P. Brady K. G. Nickel, P. Quirmbach, and J. Pötschke

High Temperature Corrosion of Low High Temperature Coatings: Protection
Alloy Steels 558 and Breakdown 691
L. W. Pinder, K. Dawson, and G. J. Tatlock H. E. Evans
Contributors
VOLUME I
B. A. Baker B. Gleeson
Special Metals Corporation, 3200 Riverside Drive, Department of Mechanical Engineering and Materials
Huntington, WV 25705, USA Science, The University of Pittsburgh, 647 Benedum Hall,
Pittsburgh, PA 15261, USA
M. P. Brady
Oak Ridge National Laboratory, Oak Ridge, P. Y. Hou
TN 37831-6115, USA Lawrence Berkeley National Laboratory, Materials Sciences
G. T. Burstein Division, 1 Cyclotron Rd., Berkeley, CA 94720, USA
Department of Materials Science and Metallurgy, University of
P. Huczkowski
Cambridge, Pembroke Street, Cambridge CB2 3QZ, UK
Forschungszentrum Jülich, IEF2, Jülich, Germany
S. Chevalier
Institut Carnot de Bourgogne, UMR 5209 CNRS, F. Huet
University of Bourgogne, 9 Avenue Savary, BP 47870, Laboratoire Interfaces et Systèmes Electrochimiques,
21078 Dijon cedex, France Université Pierre et Marie Curie – Paris 6, CNRS,
UPR15-LISE, Paris, France
R. A. Cottis
Corrosion and Protection Centre, School of Materials, I. A. Inman
,P.O.Box88,SackvilleStreet, Ellison Building, Northumbria University, Newcastle upon
Manchester, M60 1QD, UK Tyne, NE1 8ST, UK
P. S. Datta R. John
Ellison Building, Northumbria University, Newcastle upon Shell Global Solutions (US) Inc., Westhollow Technology
Tyne, NE1 8ST, UK Center, P.O. Box 4327, Houston, TX 77210, USA
K. Dawson
U. Krupp
Department of Engineering, University of Liverpool,
Faculty of Engineering and Computer Sciences, University of
Liverpool, L69 3GH
Applied Sciences Osnabrück, Albrechtstraße 30, 49076
H. L. Du Osnabrück, Germany
Ellison Building, Northumbria University, Newcastle upon
Tyne, NE1 8ST, UK C. Kübel
Group Leader Electron Microscopy, Forschungszentrum
H. E. Evans Karlsruhe, Institute for Nanotechnology, Postfach 3640,
School of Metallurgy and Materials, The University of 76021 Karlsruhe, Germany
Birmingham, Birmingham B15 2TT, UK
J. W. Fergus C. Lefrou
Materials Research and Education Center, 275 Wilmore Grenoble – INP – Grenoble Institute of Technology, UMR
Laboratories, Auburn University, Auburn, AL 36849, USA 5631 and 5266 of CNRS ‘‘LEPMI’’ and ‘‘SIMAP’’,
Grenoble, France
A. Galerie
SIMaP, Grenoble INP/CNRS/UJF, Domaine T. Markus
Universitaire, BP 75, 39402 Saint Martin d’Héres, France Forschungszentrum Jülich, IEF2, Jülich, Germany
vii
viii Contributors
T. J. Marrow P. Quirmbach
SchoolofMaterials,, Deutsches Institut für Feuerfest und Keramik GmbH, An der
Manchester, UK Elisabethenkirche 27, D-53113 Bonn, Germany
J. R. Nicholls M. Schütze
Cranfield University, Cranfield, Bedfordshire MK43 0AL, Karl Winnacker Institut der DECHEMA e.V.,
UK Theodor-Heuss-Allee 25, D-60486 Frankfurt am Main,
Germany
K. G. Nickel
Applied Mineralogy, Institute for Geosciences, University A. H. Sherry
Tübingen, Wilhelmstr. 56, D-72074 Tübingen, Germany ,Manchester,UK
R. P. Nogueira L. L. Shreir{
Grenoble – INP – Grenoble Institute of Technology, UMR
N. J. Simms
5631 and 5266 of CNRS ‘‘LEPMI’’ and ‘‘SIMAP’’,
Cranfield University, Cranfield, Bedfordshire
Grenoble, France
MK43 0AL, UK
N. Otsuka
L. Singheiser
Corporate R&D Laboratories, Sumitomo Metal Industries,
Ltd., 1-8 Fusocho, Amagasaki, Japan
M. Spiegel
B. Pieraggi
Salzgitter-Mannesmann-Forschung GmbH, Ehinger Strasse
Ecole Nationale Supérieure des Ingénieurs en Arts
200, 47259 Duisburg, Germany
Chimiques et Technologique de Toulouse, Institut National
Polytechnique de Toulouse, F-31077 Toulouse Cedex 04, H. Takenouti
France Laboratoire Interfaces et Systèmes Electrochimiques,
L. W. Pinder Université Pierre et Marie Curie – Paris 6, CNRS,
E.ON Engineering Limited Technology Centre, Ratcliffe on UPR15-LISE, Paris, France
Soar, Nottingham, NG11 OEE G. J. Tatlock
B. A. Pint Department of Engineering, University of Liverpool,
Materials Science and Technology Division, Oak Ridge Liverpool, L69 3GH
National Laboratory, Oak Ridge, TN 37831-6156, USA P. F. Tortorelli
J. Pötschke Oak Ridge National Laboratory, Oak Ridge,
Deutsches Institut für Feuerfest und Keramik GmbH, An der TN 37831-6115, USA
Elisabethenkirche 27, D-53113 Bonn, Germany P. D. Wood
R. P. M. Procter Ricardo Plc, Shoreham Technical Centre, Shoreham-by-Sea,
Corrosion and Protection Centre, School of Materials, West Sussex, BN43 5FG, UK
,ManchesterM601QD,UK
D. J. Young
W. J. Quadakkers School of Materials Science and Engineering, University
Forschungszentrum Jülich, IEF2, Jülich, Germany of New South Wales, UNSW Sydney, NSW 2033, Australia
W. J. Quadakkers J. Żurek
Forschungszentrum Jülich, Institute of Energy Research Forschungszentrum Jülich, Institute of Energy Research
(IEF-2), 52428 Jülich, Germany (IEF-2), 52428 Jülich, Germany
{
Deceased
Editorial Board
Bob Cottis graduated from the University of Cambridge in the 1960s/70s
with MA (Metallurgy) and PhD (Metallurgy) degrees. He then worked in
contract research in the corrosion group at the Fulmer Research Institute,
where he was involved in major research projects in the general area of
corrosion, corrosion fatigue and other aspects of environmentally-assisted
fracture. In addition he undertook many corrosion- and materials-related
failure investigations. In 1979 he moved to the Corrosion and Protection
CentreatUMIST,later,whereheisnow
Professor of Corrosion Science and Engineering. His research interests
are in the areas of environmentally-assisted fracture, localized corrosion,
the electrochemistry of corrosion and the application of computing tech-
niques to the control of corrosion.
Bob initiated the regular triennial EMCR (Electrochemical Methods
in Corrosion Research) conference series, and has been on the organising
or scientific committees of numerous other conferences. He is founding
editor of the online Journal of Corrosion Science and Engineering, and
co-founder of the Corros-L corrosion mailing list. He was awarded the
T. J. Hull Award of NACE International in 2005 for contributions in the
field of publication, and elected Fellow of NACE International in 2009.
He has been active in the application of new technology to corrosion
education, co-ordinating the Ecorr (Engineering Corrosion) project that
developed computer-assisted learning materials for the support of the
study of corrosion by engineering students, and leading the development
of an online distance learning route for the established MSc in Corrosion
ControlEngineeringofferedbyUMIST/.
Michael Graham graduated from the University of Liverpool with BSc

(Chemistry) and PhD (Surface Science), before accepting in 1965 a
Post-Doctoral Fellowship at the National Research Council of Canada
(NRC). He was also a Research Officer at Berkeley Nuclear Laboratories,
before returning to NRC, becoming in 1977 Head of Metallic Corrosion
and Oxidation, and later Group Leader of Surfaces and Interfaces. He is
currently an NRC Researcher Emeritus and for many years has been a
Visiting Professor at the Corrosion and Protection Centre.
Professor Graham is a Past Chairman of the International Corrosion
Council, recipient of the W. R. Whitney Award of NACE International,
Queen Elizabeth’s Jubilee Medal, H. H. Uhlig Award of the Electrochemi-
cal Society, U. R. Evans Award of the Institute of Corrosion and the
T. P. Hoar Prize of the Institute of Corrosion. His research interests include
ix
x Editorial Board
thin film formation on metals and semiconductors, high temperature

oxidation, and surface passivation and film breakdown. He has over 260
publications.
Robert Lindsay received a BSc in Chemistry from the University

of Bristol, and a PhD from the University of Liverpool. He has had
appointments at a number of research institutions, including the Fritz-
Haber-Institut (Berlin), Cambridge University, and the CSIC Institute
of Materials (ICMAB) in Barcelona. Currently, he is a lecturer in the
CorrosionandProtectionCentre,.
Rob’s research is concerned with studying surfaces under controlled
conditions to elucidate nanoscale details, focusing primarily on geometric
structure determination. Currently, he is employing such an approach to
understand corrosion inhibition at the atomic/molecular level. He has
published over 60 articles, including several reviews.
Stuart Lyon is Professor of Corrosion Control Technology and Head of

the Corrosion and Protection Centre in the School of Materials, University
of Manchester. He is a Trustee and former Executive President of the
Institute of Corrosion, and Managing Director of its trading subsidiary
(Correx Ltd.) as well as a Member of the Board of Administrators of the
World Corrosion Organisation. Research interests, which were recognised
by the award of the DSc degree by the former UMIST in 2002, include:
atmospheric corrosion and degradation, corrosion protection using organic
coatings, corrosion inhibition, corrosion monitoring and localised corro-
sion. Stuart is Editor-in-Chief of Corrosion Engineering Science and
Technology, a publication of Institute of Materials.
Tony Richardson graduated from UMIST in the 1960s with BSc

(Chemistry) and PhD (Corrosion) degrees. He spent most of his career
in industry, working as a materials/corrosion engineer, initially for
Unilever, and subsequently for ICI plc, where he led the Company’s
materials engineering group. For 5 years in the 1970s Tony was a full-
time academic in the Corrosion and Protection Centre at UMIST. When
he returned to industry, he retained his association with UMIST/Univer-
sity of Manchester, contributing to the Corrosion and Protection Centre
as a visiting professor, and also more widely as a Royal Academy of
Engineering Visiting Professor in the Principles of Engineering Design.
He has also served on the Research Boards of the Welding Institute in the
UK, and the Materials Technology Institute in the USA. He retired from
ICI plc in 2000 to form Anticorrosion Consulting, which offers indepen-
dent materials/corrosion consultancy to the chemical process industries.
Tony has been a regular organiser and contributor to international
corrosion conferences, served on the editorial board of Corrosion Engineer-
ing, Science & Technology for many years, and has contributed regularly to
the research and practice literature. In 1979, he was a first recipient of the
Guy Bengough medal of the Institute of Materials, Minerals and Mining.
Editorial Board xi
David Scantlebury
My research interests and my major teaching are closely interrelated.
For the whole of my academic career (nearly forty years!), I have been
interested in the strange situation between a potentially corroding metal
and the organic coating that is placed on that metal with the intention
of preventing corrosion of that metal. There is a fascinating mixture of
knowledge and disciplines overlapping, that I find attractive, including
metallurgy, surface science, adhesion, electrochemistry, and polymer sci-
ence. And all this arises from real problems with real solutions. My other
related research interests include marine corrosion, cathodic protection,
and the corrosion and corrosion control of rebar steel in concrete. I teach
all these subjects mainly in Module 6 in the MSc in Corrosion Control
Engineering but as well I give most of the year two undergraduate lectures
in the Materials Science degree course. Since 1989, I have organised a
five yearly international conference on corrosion protection by organic
coatings, the next and fifth is in September 2009 at Christ’s College
Cambridge. My next and second Cathodic Protection Conference is here
in Manchester 6/7th February 2006.
My administrative duties include being the course director for the MSc
course in Corrosion Engineering and Control and the one week annual
Short Course in Corrosion Engineering, now in its 35th year.
I have supervised some sixty successful PhD students from every part of
the world and this social, intellectual and cultural interaction is one of the
most satisfying features of my work. I currently hold a visiting Chair in the
Department of Chemistry, University of Xiamen, China.
Link to Original Source: http://www.materials.manchester.ac.uk/
aboutus/staff/davidscantlebury/
Howard Stott graduated from the University of Cambridge with a BA

(Metallurgy) degree in 1967 and from UMIST with a PhD degree in High
Temperature Oxidation in 1970. He was a Post-Doctoral Research Associ-
ate before being appointed Lecturer in the Corrosion and Protection
Centre, UMIST in 1972. He was promoted to Senior Lecturer followed
by Reader then Professor in 1990. He was administrative Head of the
Corrosion and Protection Centre for three 3-year periods from 1987
until 2003 and was Operations Director of The Materials Performance
Centre for the following two years. He was awarded a DSc degree by the
in1984.
During his career at UMIST, he led a research group in the field
of high temperature degradation and protection of materials, with over
50 research students and 30 PDRAs. He served on the Editorial Advisory
Boards of Oxidation of Metals, Corrosion Science, Materials and Corro-
sion and Materials at High Temperature for many years. He was elected
Fellow of the Royal Academy of Engineering in 2001 and was awarded the
UR Evans Award by the Institute of Corrosion in the same year.
Heretiredfromthe(formerlyUMIST)in
2006 and now lives in Vancouver, Canada.
Contents of All Volumes
CHAPTER 1.09
Contents of Volume One Thermodynamics and Theory of
External and Internal Oxidation of Alloys 180
Basic Concepts, High Temperature B. Gleeson
Corrosion
PART III
TYPES OF HIGH TEMPERATURE
PART I
CORROSION
BASIC CONCEPTS
CHAPTER 1.10
CHAPTER 1.01
Oxidation of Metals and
R. A. Cottis
Alloys 195
P. Y. Hou
CHAPTER 1.02
CHAPTER 1.11
Electrochemistry 13
Sulfidation and Mixed Gas
C. Lefrou, R. P. Nogueira, F. Huet, and H. Takenouti
Corrosion of Alloys 240
CHAPTER 1.03 R. John
Outline of Structural Metallurgy
CHAPTER 1.12
Relevant to Corrosion 52
Carburization and Metal
R. P. M. Procter
Dusting 272
CHAPTER 1.04 D. J. Young
Mechanical Properties and Fracture of CHAPTER 1.13
Materials 77 Nitridation of Alloys 304
A. H. Sherry and T. J. Marrow
U. Krupp
Basic Concepts of Corrosion 89 Corrosion in Molten Salts 316
L. L. Shreir M. Spiegel
PART II CHAPTER 1.15

PRINCIPLES OF HIGH TEMPERATURE High Temperature Tribocorrosion 331
CORROSION I. A. Inman, P. S. Datta, H. L. Du, C. Kübel, and
P. D. Wood
CHAPTER 1.06
Defects and Transport in Oxides and PART IV
Oxide Scales 101 HIGH TEMPERATURE CORROSION
B. Pieraggi ENVIRONMENTS

Mechanisms and Kinetics of Oxidation 132 Types of Environments 399
S. Chevalier B. A. Baker

Stress Effects in High Temperature Oxidation in Steam and Steam/Hydrogen
Oxidation 153 Environments 407
M. Schütze W. J. Quadakkers and J. Żurek
xiii
xiv Contents of All Volumes
CHAPTER 1.18
Fireside Corrosion 457 Contents of Volume Two
N. Otsuka
Corrosion in Liquids, Corrosion
CHAPTER 1.19 Evaluation
High Temperature Corrosion Issues for
Metallic Materials in Solid Oxide Fuel PART I
Cells 482 PRINCIPLES OF CORROSION IN LIQUIDS
L. Singheiser, P. Huczkowski, T. Markus,
and W. J. Quadakkers CHAPTER 2.01
Principles of Corrosion in Liquids 725
CHAPTER 1.20 G. S. Frankel and R. A. Cottis
Gas Turbine Oxidation and Corrosion 518
J. R. Nicholls and N. J. Simms CHAPTER 2.02
Passivity and Localized Corrosion 731
G. T. Burstein
PART V
HIGH TEMPERATURE MATERIALS
PART II
TYPES OF CORROSION IN LIQUIDS
CHAPTER 1.21
Design of High Temperature Alloys 541 CHAPTER 2.03
P. F. Tortorelli and M. P. Brady Crevice Corrosion 753
N. Corlett, L. E. Eiselstein, and N. Budiansky
CHAPTER 1.22
High Temperature Corrosion of Low CHAPTER 2.04
Alloy Steels 558 Pitting Corrosion 772
L. W. Pinder, K. Dawson, and G. J. Tatlock M. G. Alvarez and J. R. Galvele

High Temperature Corrosion of Dealloying 801
Chromia-forming Iron, Nickel and R. C. Newman
Cobalt-base Alloys 583 CHAPTER 2.06
A. Galerie Intergranular Corrosion 810
D. L. Engelberg
CHAPTER 1.24
High Temperature Corrosion of CHAPTER 2.07
Alumina-forming Iron, Nickel and Galvanic Corrosion 828
Cobalt-base Alloys 606 H. P. Hack
B. A. Pint
CHAPTER 2.08
CHAPTER 1.25 Environmentally Assisted Cracking
High Temperature Corrosion of Overview 857
Intermetallic Alloys 646 R. C. Newman
J. W. Fergus
CHAPTER 2.09
Stress Corrosion Cracking 864
CHAPTER 1.26
R. C. Newman
High Temperature Corrosion of
Ceramics and Refractory Materials 668 CHAPTER 2.10
K. G. Nickel, P. Quirmbach, and J. Pötschke Hydrogen Embrittlement 902
R. A. Cottis
CHAPTER 1.27
High Temperature Coatings: Protection CHAPTER 2.11
and Breakdown 691 Cracking Stimulated by Hydrogen 923
H. E. Evans C. M. Fowler
Contents of All Volumes xv

Corrosion Fatigue 928 Corrosion in Acid Gas Solutions 1270
R. Akid S. Nesic and W. Sun

Flow Assisted Corrosion 954 Corrosion in Lubricants/Fuels 1299
H. G. Schmitt and M. Bakalli R. W. Wilson and S. B. Lyon
CHAPTER 2.27
CHAPTER 2.14
Corrosion in Body Fluids 1308
Under Film/Coating Corrosion 988
D. J. Blackwood
H. N. McMurray and G. Williams
CHAPTER 2.28
CHAPTER 2.15 Corrosion by Wood 1323
Tribocorrosion 1005 M. J. Schofield
R. J. K. Wood
CHAPTER 2.29
PART III Corrosion in Radiolysis Induced
LIQUID CORROSION Environments 1330
ENVIRONMENTS G. O. H. Whillock
CHAPTER 2.16 PART IV
Atmospheric Corrosion 1051 EXPERIMENTAL TECHNIQUES FOR

I. S. Cole EVALUATING CORROSION

Corrosion in Natural Waters 1094 Electrochemical Methods 1341
S. B. Lyon R. A. Cottis

Marine Corrosion 1107 Spectroscopies, Scattering and
B. Phull Diffraction Techniques 1374
H.-H. Strehblow and D. Lützenkirchen-Hecht
CHAPTER 2.19
CHAPTER 2.32
Corrosion in Soils 1149 Electron and Photon Based Spatially
J. F. D. Stott and G. John
Resolved Techniques 1405
CHAPTER 2.20 X. Zhou and G. E. Thompson
Corrosion in Microbial Environments 1169 CHAPTER 2.33
J. F. D. Stott Scanning Probe Microscopies 1430
P. Marcus and V. Maurice
CHAPTER 2.21
Corrosion in Alkalis 1191 CHAPTER 2.34
J. A. Richardson Corrosion Testing and Determination of
Corrosion Rates 1443
CHAPTER 2.22 P. J. McIntyre and A. D. Mercer
Corrosion in Hydrogen Halides and
Hydrohalic Acids 1207 CHAPTER 2.35
J. A. Richardson Environmentally Assisted Cracking Test
Methods 1527
CHAPTER 2.23 R. N. Parkins
Corrosion in Sulfuric Acid 1226
J. A. Richardson CHAPTER 2.36
Applications of Statistical Analysis
CHAPTER 2.24 Techniques in Corrosion Experimentation,
Corrosion in Nitric Acid 1250 Testing, Inspection and Monitoring 1547
G. O. H. Whillock and S. E. Worthington N. Laycock, P. Laycock, P. Scarf, and D. Krouse
xvi Contents of All Volumes
PART V CHAPTER 3.06

MODELING CORROSION Aqueous Corrosion of Cobalt and its
Alloys 1916
CHAPTER 2.37 A. Neville and U. Malayoglu
Introduction to the Modeling of CHAPTER 3.07
Corrosion 1581 Corrosion of Copper and its Alloys 1937
R. A. Cottis C. D. S. Tuck, C. A. Powell, and J. Nuttall
Modeling of Aqueous Corrosion 1585 Corrosion of Aluminum and its Alloys 1974
A. Anderko G. M. Scamans, N. Birbilis, and R. G. Buchheit
Predictive Modeling of Corrosion 1630 Corrosion of Magnesium and its Alloys 2011
D. D. Macdonald and G. R. Engelhardt K. U. Kainer, P. Bala Srinivasan, C. Blawert, and
W. Dietzel
CHAPTER 2.40
Neural Network Methods for Corrosion CHAPTER 3.10
Data Reduction 1680 Corrosion of Titanium and its Alloys 2042
R. A. Cottis D. W. Shoesmith and J. J. Noël
CHAPTER 3.11
Corrosion of Lead and its Alloys 2053
Contents of Volume Three S. B. Lyon
CHAPTER 3.12
Corrosion and Degradation of Corrosion of Tin and its Alloys 2068
Engineering Materials S. B. Lyon
PART I CHAPTER 3.13

FERROUS METALS AND ALLOYS Corrosion of Zinc and its Alloys 2078
F. E. Goodwin
CHAPTER 3.01
CHAPTER 3.14
Corrosion of Carbon and Low Alloy
Steels 1693 Corrosion of Zirconium and its Alloys 2094
T.-L. Yau
S. B. Lyon
CHAPTER 3.15
CHAPTER 3.02
Corrosion of Cast Irons 1737 Corrosion of Tantalum and Niobium and
A. Reynaud their Alloys 2135
S. B. Lyon
CHAPTER 3.03
Corrosion of Iron Nickel Alloys and CHAPTER 3.16
Maraging Steel 1789 Corrosion of Tungsten and its Alloys 2151
G. N. Flint, J. W. Oldfield, and D. P. Dautovich S. B. Lyon

Aqueous Corrosion of Stainless Steels 1802 Corrosion of Molybdenum and its Alloys 2157
A. Iversen and B. Leffler S. B. Lyon
CHAPTER 3.18
PART II Corrosion of Beryllium and its Alloys 2168
NON-FERROUS METALS AND ALLOYS R. S. Lillard

Aqueous Corrosion of Nickel and its Alloys 1879 Corrosion of Uranium and its Alloys 2181
H. Alves and U. Heubner S. B. Lyon
Contents of All Volumes xvii

Corrosion of Amorphous and Degradation of Polymer Matrix
Nanograined Alloys 2192 Composites 2387
H. Habazaki R. H. Martin

Corrosion of Noble Metals 2205 Degradation of Natural Rubber and
S. B. Lyon Synthetic Elastomers 2407
A. bin Samsuri
CHAPTER 3.22
Corrosion of Passive Alloys: The Effect of CHAPTER 3.34
Noble Metal Additions 2224 Degradation of Wood 2439
J. H. Potgieter L. C. Pinion and E. Ajuong
CHAPTER 3.23 PART V

Corrosion of Metal Matrix Composites 2250 JOINTS
L. H. Hihara
CHAPTER 3.35
PART III Corrosion of Metal Joints 2447
CERAMIC MATERIALS G. Pimenta and R. A. Jarman

Degradation of Carbon and Graphite 2271 Role of Corrosion in the Failure of
S. B. Lyon Adhesive Joints 2463
J. F. Watts
CHAPTER 3.25
Degradation of Engineering Ceramics 2282
R. Morrell
Contents of Volume Four
CHAPTER 3.26
Degradation of Glass and Glass Ceramics 2306 Management and Control of Corrosion
S. Oliver, B. A. Proctor, and C. A. May
CHAPTER 3.27 PART I

Degradation of Glass Linings and SURFACE TREATMENT AND
Coatings 2319 MODIFICATION
G. Schäfer
CHAPTER 4.01
CHAPTER 3.28 Surface Pretreatment 2483
Degradation of Vitreous Enamel G. L. Higgins, R. S. Hullcoop, S. Turgoose, and
Coatings 2330 W. Bullough
T. Curtis
CHAPTER 4.02
CHAPTER 3.29 Phosphate Coatings 2494
Degradation of Ceramic Masonry Linings 2337 M. O. W. Richardson and R. E. Shaw
D. I. Hughes
CHAPTER 4.03
CHAPTER 3.30 Coatings Produced by Anodic Oxidation 2503
Degradation of Cement and Concrete 2348 X. Zhou, P. G. Sheasby, and B. A. Scott
P. Lambert, R. Brueckner, and C. Atkins
PART II
PART IV METALLIC AND INORGANIC COATINGS
POLYMERIC MATERIALS
CHAPTER 4.04
CHAPTER 3.31 General Principles of Protection by
Degradation of Plastics and Polymers 2369 Coatings 2519
D. J. Hourston D. R. Gabe, G. D. Wilcox, and V. E. Carter
xviii Contents of All Volumes
CHAPTER 4.05 PART IV

Diffusion Coatings 2532 ELECTROCHEMICAL PROTECTION
J. R. Nicholls, K. A. Long, and N. J. Simms
CHAPTER 4.18
CHAPTER 4.06 Principles of Cathodic Protection 2747
Hot Dipped Coatings 2556 V. Ashworth
W. J. Smith and F. E. Goodwin
CHAPTER 4.19
CHAPTER 4.07 Sacrificial Anodes 2763
Plated Coatings 2577 R. F. Crundwell
D. R. Gabe and M. Clarke
CHAPTER 4.20
CHAPTER 4.08 Impressed-current Anodes 2781
Sprayed Coatings 2610 W. R. Jacob
K. T. Voisey CHAPTER 4.21
Practical Application of Cathodic
CHAPTER 4.09
Protection 2801
Laser Applied Coatings 2622
B. S. Wyatt
Z. Liu
CHAPTER 4.22
PART III Interaction and Stray-current Corrosion 2833
PAINTS AND ORGANIC COATINGS L. Di Biase
CHAPTER 4.23
CHAPTER 4.10
Cathodic Protection Instrumentation 2839
Paint Application 2637
B. S. Wyatt
N. R. Whitehouse
CHAPTER 4.24
CHAPTER 4.11 Anodic Protection 2857
Paint Formulation 2643 P. Novák
G. P. Bierwagen and A. M. Huovinen
PART V
CHAPTER 4.12
ENVIRONMENTAL MODIFICATION
Mechanisms of Protection by Paints 2666
J. E. O. Mayne
CHAPTER 4.25
CHAPTER 4.13 Introduction to Control of Corrosion by
Temporary Protectives 2678 Environmental Modification 2891
T. N. Tate and E. W. Beale R. Lindsay and S. B. Lyon
CHAPTER 4.26
CHAPTER 4.14
Chemical Treating in Oil and Gas
Marine Paints 2683
Production 2900
P. Jackson
J. L. Dawson
Coatings for Structures in Contact With Environmental Modification for Cooling,
the Ground 2702 Heating and Potable Water Systems 2930
D. Fairhurst C. Frayne
Inspection of Paints and Painting Boiler and Feedwater Treatment 2971
Operations 2720 A. Banweg
N. R. Whitehouse
CHAPTER 4.29
CHAPTER 4.17 Corrosion Inhibitors: Other Important
Paint and Coating Failures and Defects 2728 Applications 2990
B. Fitzsimons and T. Parry V. S. Sastri
Contents of All Volumes xix
PART VI CHAPTER 4.38

CORROSION MANAGEMENT Management of Corrosion of Aircraft 3175
C. J. E. Smith
CHAPTER 4.30
Corrosion Management Overview 3001 CHAPTER 4.39
J. L. Dawson Management of Corrosion of
Infrastructure 3198
CHAPTER 4.31 C. Atkins, R. Brueckner, and P. Lambert
Economic Aspects of Corrosion 3040
J. A. Richardson and J. L. Dawson CHAPTER 4.40
Management of Corrosion in
CHAPTER 4.32
the Petrochemical and Chemical
Materials Selection for Corrosion Resistance 3052
T. A. Richardson and S. B. Lyon Industries 3207
J. A. Richardson
CHAPTER 4.33
Mitigation of Corrosion Risks by Design 3065 CHAPTER 4.41
J. A. Richardson and J. L. Dawson Management of Corrosion in the Oil and
Gas Industry 3230
CHAPTER 4.34
J. Dawson, G. John, and K. Oliver
Risk Based Inspection 3084
P. Horrocks and S. Adair CHAPTER 4.42
CHAPTER 4.35 Management of Corrosion of Onshore
Assessment of Fitness for Service 3102 Pipelines 3270
A. H. Sherry J. M. Race

Corrosion Monitoring and Inspection 3117 Preservation of Metallic Cultural
C. F. Britton Heritage 3307
CHAPTER 4.37
D. Watkinson
Management of Corrosion of Automobiles 3167
N. R. Whitehouse Index 3369
Foreword
In keeping with most of the colleagues who have edited this book, I had the good fortune to meet and know
Lionel Shreir. We knew him first through the 1st edition of Corrosion, published in 1963, that was our
introduction to the subject as we embarked on our postgraduate research. In the UK, Shreir’s Corrosion was
the one-stop-reference work that provided an introduction to most aspects of corrosion and its control,
authored by the great and good of corrosion and heavily laced with scientific fundamentals. Later, we got to
know Lionel personally as a very active educator, researcher and consultant during his tenure as Head of
Metallurgy at Sir John Cass College in London. After Lionel’s retirement, my predecessor, Graham Wood,
invited Lionel to join us in the Corrosion and Protection Centre that had been established in Manchester in the
early 1970s and we were delighted when he accepted an honorary position. He provided valuable advice and
support and, in addition to his considerable professional talents that spawned an impressive collection of awards,
he was a great encourager of the young and a very affable, charming colleague.
We are well placed in Manchester to take on the task of editing a 4th edition of Corrosion in having world
class academics in most areas of corrosion science and engineering. Even so, I know that my colleagues thought
long and hard before taking on such a challenging task and our affection and respect for Lionel were
undoubtedly key influences on their decision to proceed. I applaud their efforts in delivering a 4th Edition of
Corrosion, not least those of our retired and visiting professors, Howard Stott, Mike Graham and Tony
Richardson, and I am delighted to see contributions from Manchester alumni from around the world.
The Corrosion and Protection Centre is proud to be associated with this 4th Edition of Corrosion, which
we dedicate to the memory of one of corrosion’s godfathers, Lionel Shreir.
G E Thompson, OBE FREng,

Head, Corrosion and Protection Centre,
School of Materials,
xxi
Preface to the Fourth Edition
Over the 46 years since it was first published, Shreir’s Corrosion has sustained its reputation as a major
educational and reference work on all aspects of corrosion science and engineering. For the 3rd Edition, Lionel
enlisted the assistance of Tim Burstein and Ray Jarman to edit the work and his good judgement was reflected in
its continuing quality and popularity. So, in agreeing to take on the task of editing a 4th Edition of Shreir, we
were keenly aware of our responsibility to maintain the reputation that the work has earned amongst our peers
in the corrosion science and engineering communities, not to mention the wider engineering community,
worldwide.
In approaching a 4th Edition, we decided to depart from the structure of previous editions and rebalance the
content of the work to reflect the major advances that have been made since the work was first published in the
techniques available for corrosion research, in our understanding of the mechanisms of aqueous and high
temperature corrosion, and in the development of practices for the control and management of corrosion. Thus,
we have commissioned new contributions on experimental techniques for the study of corrosion and the
modeling and prediction of corrosion behaviour, and in an era of intense societal pressures to prevent major
incidents that threaten safety, health and the environment, we have commissioned contributions on the
management of corrosion in mature industries.
In making significant changes, we have been concerned to retain the essential character of Shreir’s
Corrosion, which has always doubled as part textbook and part reference work. As such, we hope that it will
retain its wide appeal to all with a professional interest or involvement in corrosion, including graduate and post
graduate students, academics who teach corrosion courses, or with active corrosion research interests and
scientists and engineers of any discipline working as corrosion professionals in research or consulting, in
supplying corrosion control services or corrosion resistant materials, or in organisations that operate physical
assets of any type that are vulnerable to corrosion.
We would like to record our gratitude to our many authors from around the world for their willingness to
produce chapters in times when both effort and time are scarcer commodities than used to be the case in
professional life. We decided to retain some ‘old master’ contributions from Lionel himself, Tim Burstein,
Robin Proctor, Jack Mayne and Redvers Parkins on fundamental topics that are essentially timeless. We have
contributed some chapters ourselves and, as anyone previously involved in editing an undertaking such as this
will understand, have attempted to update one or two existing chapters in haste where prospective authors were
unable to deliver. While the outcome may not be perfect, we hope sincerely that it will prove as valuable to the
corrosion science and engineering communities worldwide as its predecessors and, in so doing, will help repay
our collective dues to Lionel Shreir, who started it all.
Finally, we record our gratitude to the Elsevier development and production staff who kept us on task
throughout the project and worked so hard for a successful outcome, in particular Arnout Jacob, Adrian Shell,
Natalia Kennedy, Nicola Lally, Simon Wood, Hilary Broadribb, Melinda Debreczeni, Beckie Brand, Fiona
Geraghty, Hazel Harris, Gareth Steed, Bob Donaldson and last, but very definitely not least, Kostas Marinakis.
RAC/MJG/RL/SBL/JAR/JDS/FHS
Corrosion & Protection Centre
School of Materials
xxiii
L. L. Shreir, OBE, 1914–1992
In keeping with most of the colleagues who have edited this book, I had the good fortune to meet and know
Lionel Shreir. We knew him first through the 1st edition of Corrosion, published in 1963, that was our
introduction to the subject as we embarked on our postgraduate research. In the UK, Shreir’s Corrosion was
the one-stop-reference work that provided an introduction to most aspects of corrosion and its control,
authored by the great and good of corrosion and heavily laced with scientific fundamentals. Later, we got to
know Lionel personally as a very active educator, researcher and consultant during his tenure as Head of
Metallurgy at Sir John Cass College in London. After Lionel’s retirement, my predecessor, Graham Wood,
invited Lionel to join us in the Corrosion and Protection Centre that had been established in Manchester in the
early 1970s and we were delighted when he accepted an honorary position. He provided valuable advice and
support and, in addition to his considerable professional talents that spawned an impressive collection of awards,
he was a great encourager of the young and a very affable, charming colleague.
We are well placed in Manchester to take on the task of editing a 4th edition of Corrosion in having world
class academics in most areas of corrosion science and engineering. Even so, I know that my colleagues thought
long and hard before taking on such a challenging task and our affection and respect for Lionel were
undoubtedly key influences on their decision to proceed. I applaud their efforts in delivering a 4th Edition of
Corrosion, not least those of our retired and visiting professors, Howard Stott, Mike Graham and Tony
Richardson, and I am delighted to see contributions from Manchester alumni from around the world. The
Corrosion and Protection Centre is proud to be associated with this 4th Edition of Corrosion, which we
dedicate to the memory of one of corrosion’s godfathers, Lionel Shreir.
G E Thompson, OBE FREng,

Head, Corrosion and Protection Centre,
School of Materials,
xxv
1.01 Chemical Thermodynamics
R. A. Cottis
CorrosionandProtectionCentre,SchoolofMaterials,,P.O.Box88,SackvilleStreet,Manchester,
M60 1QD, UK
L. L. Shreir{ and G. T. Burstein
Department of Materials Science and Metallurgy, University of Cambridge, Pembroke Street, Cambridge CB2 3QZ, UK
This article is a revision of the Third Edition article 20.2 by L. L. Shreir and G. T. Burstein, volume 2, pp 20:57–20:75,
ß 2010 Elsevier B.V.
1.01.1 Scope 1
1.01.2 Some Definitions 2
1.01.2.1 First Law of Thermodynamics 2
1.01.2.2 Second Law of Thermodynamics 3
1.01.2.3 Entropy 4
1.01.2.4 Gibbs Free Energy or Free Enthalpy 5
1.01.2.5 The Chemical Potential 5
1.01.2.6 Activity 6
1.01.2.7 Gibbs–Duhem Equation 6
1.01.3 Spontaneity of a Reaction 7
1.01.3.1 Reversible Cells 7
1.01.4 Chemical Potentials and Equilibrium 8
1.01.4.1 Sign Convention for Equilibrium emfs and Potentials 12
R The gas constant (8.314 J K1 mol1)

Symbols
S Entropy (J)
a Activity (strictly dimensionless – see section
T Temperature (K)
1.01.2.6)
U Internal energy (J)
c Concentration (mol kg1 or mol m3 or mol dm3)
V Volume (m3)
c Concentration in standard state (mol kg1 or
w Work (J)
mol m3 or mol dm3)
z Number of electrons or equivalent positive charge
Cp Molar heat capacity at constant pressure
g Activity coefficient
(J mol1 K1)
m Chemical potential (J mol1)
E Equilibrium potential (V)
m Standard chemical potential (J mol1)
E Standard equilibrium potential (V)
y Number of ions or molecules
F Faraday constant (96 485 C mol1)
G Gibbs free energy or free enthalpy (J)
G Standard Gibbs free energy or free
enthalpy (J)
1.01.1 Scope
H Enthalpy (J)
I Ionic strength (mol kg1 or
Thermodynamics is concerned with the relationship
mol m3 or mol dm3)
between heat energy and work and is based on two
K Equilibrium constant
general laws, the first and second laws of thermody-
n Number of moles
namics, both of which deal with the interconversion
P Pressure (Pa, atm)
of different forms of energy. The third law states that
p Standard pressure (usually 1 atm)
at the absolute zero of temperature, the entropy of a
Q Activity quotient
perfect crystal is zero, and thus provides a method for
q Heat (J)
determining absolute entropies.
{
Deceased
1
2 Basic Concepts
The following features of the thermodynamic If pressure changes from p to p þ dp and temperature
approach to chemical reactions should be noted: from T to T þ dT, then the change in V is given by

1. It does not involve time as a variable, and does not @V @V
dV ¼ dp þ dT ½2
provide information on rates. @p T @T p
2. It does not normally involve a knowledge of the
detailed structure of the system or of the detailed where ð@V =@pÞT is the rate of change of volume with
mechanism of the reaction. pressure at constant temperature, and ð@V =@T Þp is
3. The second law is true only statistically and does the rate of change of volume with temperature at
not apply to individual particles or to a small constant pressure. They are referred to as partial
number of particles, that is, thermodynamics is derivatives ; the subscripts T and p denoting the vari-
concerned with bulk properties of systems. Thus, ables that are maintained constant.
thermodynamics has many limitations, but it is A reversible system is one that is so perfectly
particularly valuable in defining the nature and balanced that only infinitesimal change in an external
structure of phases when equilibrium (a state that conditions is required to reverse the direction of
does not vary with time) has been attained; ther- the process; for reversibility, the change must either
modynamics provides no information on the rate at be infinitesimally small or, if large, it must be carried
which the reaction proceeds to equilibrium, which out at an infinitely slow rate. Although these concepts
belongs to the realm of chemical kinetics. are hypothetical, it is possible to approximate to
reversibility by balancing the electromotive force
(emf) of a reversible cell with the emf of a potentiom-
1.01.2 Some Definitions eter circuit so that only a minute change (<106 V) in
the emf produced by the latter will result in a reversal
A system is any part of external reality that can be of the direction of the cell reaction. An ideal revers-
subjected to thermodynamic treatment; the material ible cell must consist of two reversible electrodes,
with which the system is in contact forms the sur- which are so completely nonpolarizable that their
roundings, for example, an electrochemical cell could potentials are unaffected by charge transfer even at
be the system and the external atmosphere the high rates. This too is a hypothetical concept, but a
surroundings. number of electrodes will approximate to reversibil-
A homogeneous system is one in which the properties ity, provided the rate of charge transfer is small, and
are uniform or vary uniformly throughout the system, cells composed of these electrodes may be treated
whereas a heterogeneous system is one consisting of thermodynamically.
two or more homogeneous systems. Pure copper
and a pure solution of CuSO4 are both homoge-
neous systems, but copper immersed in a solution 1.01.2.1 First Law of Thermodynamics
of CuSO4 is a heterogeneous system characterized The first law is essentially a statement of the experi-
by an interface. The state of a system is described mental fact that during a chemical reaction in an
by its properties, which may be intensive or extensive; isolated system, energy is conserved, although it may
intensive properties are independent of quantity, for be converted from one form to another.
example, temperature, pressure, electrical potential, If a system changes from state A to state B and
activation energy, etc., whereas extensive properties if DU is the change in the internal energy of the
depend upon the amount of the material, for example, system, then it follows from the first law that the
mass, volume, internal energy, enthalpy, entropy, free surroundings must gain or lose an equivalent amount
energy, chemical potential, etc. of energy. A convention of signs is necessary to take
The state of a system containing a constant into account the direction of heat transfer q and work
amount of material depends on a few variables, for performed w: q is taken as positive if heat is trans-
example, pressure p, volume V, and temperature T. ferred to the system from the surroundings and nega-
For a given mass of pure substance, the volume can tive if the transfer is in the reverse direction; in many
be expressed solely as a function of pressure and textbooks, w is taken as positive if work is done by
temperature the system on the surroundings and negative if it
is done on the system by the surroundings, but it is
V ¼ f ð p; T Þ ½1 preferable to allow the sign of w to develop from the
mathematics. Thus, since energy must be conserved Carnot cycle (see below) that there is a theoretical
limit to the extent of this interconversion.
DU ¼ q w ½3
Oxidation of hydrogen to water
where DU is the change in the internal energy (H2 þ 12 O2 ! H2 O) is thermodynamically spontane-
of the system, q is the heat transferred between ous, and the energy released as a result of the
the system and the surroundings, and w is the work chemical reaction appears as heat energy, but the
done. Equation [3] may be regarded as an algebraic decomposition of water into its elements is a non-
expression of the conservation of energy, and it spontaneous process and can be achieved only by
should be noted that DU is the difference in the supplying energy from an external source, for exam-
internal energy of the system between its initial ple, a source of emf that decomposes the water elec-
and final stages and that the absolute value of U is trolytically. Furthermore, although the heat energy
not known. produced by the spontaneous reaction could be con-
If a change from state A to state B occurs in a verted into electrical energy; the electrical energy
system at constant pressure (isobaric) so that only thus obtained would not be sufficient to restore the
pV work is done, then w ¼ p DV and original state of chemical energy of the system, that
is, to regenerate 1 mol of H2 and 12 mol of O2.
DU ¼ qp pðVB VA Þ ½4 The second law, which can be stated in a number
which can be written in the form of different ways, expresses the fact that heat can only
be transformed into work when it flows from a higher
qp ¼ UB UA þ pðVB VA Þ to a lower temperature. Clausius stated the second
½5 law as follows: ‘‘It is impossible for a self-acting
¼ ðUB þ pVB Þ ðUA þ pVA Þ
machine, unaided by an external agency, to convey
Thus, it is convenient to define another function heat from a lower to a higher temperature.’’
In principle, all spontaneous processes may be
H ¼ U þ pV ½6 harnessed and made to do work, but owing to the
wastage of a proportion of the energy as heat energy
where H is the enthalpy or heat content of the system,
there is a theoretical limit to the amount of useful
so that
work that can be obtained, and this can never be
qp ¼ HB HA ¼ DH ½7
exceeded in practice. A high-efficiency power station
where qp is the heat exchanged between the system converting chemical energy (solid or liquid fuel) into
and the surroundings at constant pressure, and DH, electrical energy achieves only 30–40% efficiency.
like DU, depends only upon the initial and final states The maximum theoretical work wmax obtainable
of the system. from a system was derived by Carnot who considered
the transformation of heat energy into work when a
perfect gas in a cylinder with a piston was taken
1.01.2.2 Second Law of Thermodynamics
through a reversible cycle (the Carnot cycle), in
The first law implies that energy is interconvertible which the system was almost at equilibrium during
and that when one form of energy disappears another each step of the cycle. It was shown that
form of energy must appear in its place. However, it
provides no information on (a) the transformations of T2 T1
wmax ¼ q2 ½8
energy that are possible and (b) the extent to which T2
one form of energy can be converted into another.
Some energy transformations are possible and occur where q2 is the heat absorbed from a reservoir at the
spontaneously, while transformations in the reverse higher temperature T2 (K), and T1 is the temperature
direction are nonspontaneous and can be made to of a second heat reservoir at a temperature lower than
proceed only by supplying energy from an external T2 which receives the heat that has not been con-
source. Thus, heat will flow spontaneously from a verted into work. Equation [8] is applicable to any
source at a higher temperature to one at a lower reversible system that undergoes a cyclic process in
temperature, but it cannot pass spontaneously in which the original and final states of the system and
the reverse direction. The heat energy flowing spon- the surroundings are the same. Under these circum-
taneously can be converted by means of suitable stances, all that has changed permanently is that a
machinery to work energy, but it is evident from the quantity of heat q2 at T2 has been transformed into
4 Basic Concepts
work wmax together with a quantity of heat q1 ¼ q2 process is conducted reversibly. If the process is con-
wmax that is returned to the reservoir at the lower ducted irreversibly and the heat absorbed is q, then q
temperature T1. It can be shown that the efficiencies will be less than qrev and q/T will be less than the
of all reversible cyclic processes are the same entropy change DS eqn [12]. It follows that if
provided T1 and T2 are the same, that is an irreversible process takes place between the
wmax q2 q1 temperatures T2 and T1 and has the same heat intake
Thermodynamic efficiency ¼ ¼ q at the higher temperature T2 as the correspond-
q2 q2
ing reversible process, then the efficiency of the
T2 T1 ½9 former must be less than that of the latter, that is
¼
T2
q1 q2 q1 q2
It is evident from eqn [9] that (a) the efficiency > ½15
T1 T2 irrev: T1 T2 rev:
increases as (T2 T1)/T2 increases and (b) the
efficiency is zero when T1 ¼ T2 so that no heat can Thus, from eqns. [11] and [15] it follows that for an
be converted into work. irreversible cycle
Xq
>0 ½15a
T
1.01.2.3 Entropy
and that more positive (q/T) is the more irreversible
The concept of entropy S is implicit in the second is the cycle. In the Carnot cycle, each stage is carried
law, and it may be regarded as a thermodynamic out reversibly and the entropy change is zero, and it
property of a system that is associated with its capac- can be shown that any complete cycle can be reduced
ity for spontaneous change. As the system undergoes to the sum of a large number of infinitesimal Carnot
spontaneous change, the entropy increases and the cycles so that the entropy change for any complete
capacity for further spontaneous change decreases. reversible cycle must always be zero. Thus, for any
Rearranging the q and T terms in eqn [9] and apply- reversible process, the sum of all entropy changes for
ing the convention of signs all the participating systems is given by
q1 q2 X
¼0 ½10 dS ¼ 0 ½16
T1 T2
and generalizing for any reversible cyclic change which provides a criterion for equilibrium.
that proceeds isothermally at a temperature T On the other hand, in any irreversible process,
X qrev although the system may gain (or lose) entropy and
¼0 ½11
T the surroundings lose (or gain) entropy, the system
where qrev is the sum of the heat absorbed or plus surrounding will always gain in entropy eqn [16].
evolved reversibly. The function qrev/T is a measure Thus, for a real process proceeding spontaneously at
of the change in entropy DS, and this change for an a finite rate
isothermal process is defined as X
dS > 0 ½17
qrev
DS ¼ ½12
T For example, if a hot body (the system) is in contact
with the surroundings at an infinitesimally lower
and for an infinitesimal reversible process
temperature, then there will be an isothermal and
dq reversible transfer of heat at T2, and the loss in
dS ¼ ½13
T entropy of the hot body will be DSb ¼ qrev/T2, the
gain in entropy of the surroundings will be DSs ¼
If the process is nonisothermal but finite
þ qrev/T2, and DSb þ DSs ¼ 0. On the other hand, if
ðS2 the surroundings are at a lower temperature T1 than
dq
DS ¼ S2 S1 ¼ ½14 the body, then the transfer of heat will be irrevers-
T ible and spontaneous, and the entropy changes will be
S1
DSb ¼ qrev/T2 and DSs ¼ þ qrev/T1. Since T2 > T1,
It must be emphasized that the heat qrev, which the total change in entropy will be greater than zero,
appears in the definition of entropy (eqn [12]), is and the greater the difference between T2 and T1 the
always that which is absorbed (or evolved) when the greater will be the gain in the total entropy of the
system as a whole, that is, the more spontaneous or or, for a finite change
irreversible is the transfer of heat.
When a substance is heated at constant pressure DG ¼ DH T DS ½22
without change of phase through a temperature rise It can be shown that for a system in which the total
dT, the heat absorbed is Cp dT, where Cp is the molar mass, or the concentrations of the components, remains
heat capacity at constant pressure, and the entropy constant
increase is
@G
ð S2 ð T2 ð T2 ¼V ½23
Cp @P T ;ni ...
dS ¼ ¼ Cp d lnT ½18
S1 T1 T T1 and

Thus, the entropy increase when the substance is @G
¼ S ½24
heated from T1 to T2 is @T p;ni ...
ð T2 ð T2
Combining eqns [21] and [24]
DS ¼ Cp d lnT ¼ 2:303 Cp d logT ½19
T1 T1 @G
GH ¼T ½25
@T p;ni ...
DS may be evaluated by determining the heat capac-
ity Cp at various temperatures between T2 and T1, which is the important Gibbs–Helmholtz equation.
and plotting the values against log T; the area Since in thermodynamics we are concerned mainly
enclosed by the curve multiplied by 2.303 gives the with changes in free energy and enthalpy, eqn [25] can
entropy increase. When there is a change of phase be written as
such as the evaporation of a liquid to saturated vapor,
@DG
melting of a solid to a liquid, or transition from one DG DH ¼ T ½26
crystalline form to another, the process is isothermal @T p;ni ...
and reversible, and Units: Free energy is an extensive quantity, but

q the standard free energy for 1 mol DG298K (in units of
DS ¼ ½20
T J mol1) defined above is an intensive quantity.
where q is the heat associated with the phase change at T.
For example, in the case of the reversible isother-
mal transformation of ice to water at the melting 1.01.2.5 The Chemical Potential
point (273 K), the heat gained by the ice will be The energy of a system can be changed by means of
the latent heat of fusion (DHf = 6006 J mol1) and a thermal energy or work energy, but a further possi-
corresponding quantity of heat will be lost by the bility is to add or subtract moles of various substances
surrounding, and to or from the system. The free energy of a pure
6:006 6:006 substance depends upon its chemical nature, its
DSsystem ¼ DSice þ DSsurr ¼ ¼0 quantity (DG is an extensive property), its state
273 273
(solid, liquid, or gas), and temperature and pressure.
Gibbs called the partial molar free heat content (free
energy) of the ith component of a system its chemical
1.01.2.4 Gibbs Free Energy or Free Enthalpy potential

Gibbs free energy G provides a means of defining equi- @G
mi ¼ ½27
librium or of the tendency of a reaction to proceed in a @ni T ;p;nj
given direction. It is similar to the other thermodynamic
functions described above in which the change in free that is, the change in free enthalpy of the system
energy DG is determined solely by the initial and final when 1 mol of component i is added to the system,
states of the system. The maximum work, or maximum the T, p and concentration of all other substances (nj)
available energy, defined in terms of the Gibbs free being maintained constant. The partial molar free
energy G, also called the free enthalpy, is energy of a pure compound is synonymous with the

standard chemical potential m , which is the free
G ¼ H TS ½21 enthalpy change when 1 mol of the substance is
6 Basic Concepts
formed from its elements under isothermal isobaric For a 1:1 electrolyte (NaCl), a similar procedure
conditions. In order to evaluate the chemical poten- may be adopted, but in this case not only do the
tials of compounds, and of ions in solution, it is constituent ions (Na+ and Cl) become more ideal
necessary to define the following standard states: in behavior as the concentration decreases, but also
the electrolyte becomes more completely dissociated
1. The standard chemical potentials of the elements
into ions. It is not possible to determine individual
in the form that they are normally stable at the
ionic activity coefficient since anions and cations
temperature and pressure under consideration are
cannot be studied in the absence of one another.
given an arbitrary value zero
Thus, for a 1:1 electrolyte eqn [30] is written as

m ðelementÞ ¼ 0 ½28
m ¼ m þ 2RT ln a ¼ m þ 2RT ln ðg c=c Þ ½34
2. In the case of ions in solution and of gases, the where a is the mean ionic activity, g is the mean
chemical potential will depend upon concentra- activity coefficient, and g ! 1 when c ! 0. More
tion and pressure, respectively. An arbitrary value complex relationships are required for unsymmetrical
of zero at a specified concentration is given for electrolytes, such as Na2SO4, La2(SO4)3, TiCl4, etc.
ions in solution, the standard chemical potential of At sufficiently low ionic strengths, the activity
the hydrogen ion, at the temperature and pressure coefficient of each electrolyte in a mixture is given
under consideration by the Debye–Hückel limiting law
pffiffiffi
mHþ ¼ 0; when aHþ ¼ 1 ½29 log g ¼ j zþ z j A I ½35
where aHþ is the activity of the hydrogen ion. where I the ionic strength is defined as
1X 2
1.01.2.6 Activity I ¼ ci zi ½36
2
The chemical potential of the hydrogen ion, or of any and z+ and z are the valencies of the cation and
other ion, will vary with activity anion (NaCl is a 1:1 electrolyte, Na2SO4 a 2:1 elec-

mi ¼ mi þ RT ln ai ½30 trolyte, La2(SO4)3 a 2:3 electrolyte, etc.), and A is
a constant which is 0.51 mol1/2 kg1/2 for water at
where the activity ai is a thermodynamic quantity 25 C (ci is in mol kg1).
that is related to the concentration of the ion and Units: It should be noted that in the SI the activity
takes into account the fact that a solute in solution of a solute is defined with reference to a standard
does not behave ideally at finite concentrations; only state, that is, an ideal solution of molality 1 mol kg1.
at infinite dilution, will ai be equal to ci numerically, Thus, the relative activity of a metal ion in solution is
where ci is the concentration (mol kg1, mol m3, or given by
mMzþ
mol dm3). aMzþ ¼ gMzþ ½37
Substituting for R ¼ 8.314 J K1 mol1 and T ¼ m
298 K, and using the identity ln x ¼ 2.303 log x, eqn
where m ¼ 1 mol kg1, mMz+ is in mol kg1, and gMz+
[30] can be expressed as is dimensionless. It follows that the relative activity is

mi ¼ mi þ 5710 log ai ½31 dimensionless, although units, such as mol kg1, g ion
per litre, etc., are still being used (see below).
A similar equation is applicable to a gas, but the
pressure of the gas pg (or more strictly its fugacity) 1.01.2.7 Gibbs–Duhem Equation
replaces the activity of the ion

The Gibbs free energy change of a system will
mg ¼ mg þ 5710 log ðpg =p Þ ½32 depend not only on temperature and pressure, but
also on the chemical potentials of the species
where p refers to a standard pressure of 1 atm or
involved, and this statement may be expressed in
101 325 105 Pa. For a nondissociating solute (e.g.,
the form of the partial differential
urea)

a ¼ gc=c

½33 @G @G
dG þ dT þ dp
@T p;n1 ;n2 ... @p T1 ;n1 ;n2 ...
where g is the activity coefficient (g ! 1 as c ! 0) and

c is the concentration in the standard state. þ m1 dn1 þ m2 dn2 ½38
and combining with eqns [23] and [24] reverse direction, and if DG ¼ 0, then the system will be
X at equilibrium and will not have any tendency to pro-
dG ¼ S dT þ V dp þ mi dni ½39
ceed in either direction.
For a process carried out under isothermal and iso- Thus, for a spontaneous reaction
baric conditions
X DG ¼ ðcmC þ d mD Þ ðamA þ bmB Þ < 0 ½50
ðdGÞT ;p ¼ mi dni ½40
and if all the substances are in their standard states
and integration of this equation for a system of given
chemical composition gives DG ¼ ðcmC þ d mD Þ ðamA þ bmB Þ < 0 ½51

X
GT ;p;n ¼ ni mi ½41
General differentiation of this equation gives 1.01.3.1 Reversible Cells
X
dGT ;p ¼ ðni dmi þ mi dni Þ ½42 By definition
Since at equilibrium (dG)T,p,n = zero, it follows from DG ¼ zFE ½52
eqn [40] that
X where E (the reversible emf) is the intensity factor in
ni dmi ¼ 0 ½43 the energy term and zF (the total charge per mole of
ions involved in the reaction) is the capacity factor. If
which is referred to as the Gibbs–Duhem equation. the activities of the reactants and products are fixed,
For a binary solution of say A and B E will be a constant, and if aA = aB = aC = aD = 1 is
nA dmA ¼ nB dmB ½44 taken as the standard state, then

Since DG ¼ zFE ½53

mi ¼ mi þ RT ln ai ½45 where DG is the standard free enthalpy change and

E is the standard emf which is a constant for a given
dmi ¼ RT d ln ai ½46 equilibrium at constant T.
Substituting zFE for DG eqn [53] in eqn [26]
Hence
gives
nA d ln aA ¼ nB d ln aB ½47
@E
The Gibbs–Duhem equation is extremely impor- DH ¼ zFE þ zFT ½54
@T p
tant in solution chemistry, and it can be seen from
eqn [47] that it provides a means of determining the which provides an accurate method of evaluating DH
activity of one component in a binary solution and DS in eqn [54] from the emf of a reversible cell
provided the activity of the other is known. and the coefficient of the change of emf with temper-
ature at constant p, since

1.01.3 Spontaneity of a Reaction @ðDGÞ @E
T ¼ zFT ¼ T DS ½55
@T p @T p
An isothermal reaction can proceed spontaneously only
if the total Gibbs free energy of the system decreases; From eqn [49]
that is, the free energy of the reactants must be greater
X
DG ¼ ui mi ½56
than the free energy of the products. For a reaction
aA þ bB ¼ cC þ d D ½48 where ui is the number of each of the ions or mole-
cules involved in the equilibrium; combining eqns
X X
DG ¼ mproducts mreactants ½49 [53] and [54]

where DG is the change in free enthalpy. It follows that S ui mi

S ui mi
E ¼ ¼ ½57
if DG < 0, then the reaction will tend to proceed zF 96 500z
spontaneously in the direction shown; if DG > 0, then Equations [53] and [57] emphasize the essentially
the reaction will tend to proceed spontaneously in the thermodynamic nature of the standard equilibrium
8 Basic Concepts
emf of a cell or the standard equilibrium potential

Since DG ¼ DH TS

of a half-reaction E , which may be evaluated directly

from the emf measurements of a reversible cell or

DG ¼ 461 200 ð298 43:3Þ
indirectly from DG , which must in turn be evaluated
from the enthalpy of the reaction and the entropies of ¼ 448300 J mol1 at 298 K

the species involved (see eqn [22]). Thus, for the DG 448 300
equilibrium Cu2þ þ 2e Ð Cu the standard electrode Ereaction ¼ ¼ ¼ 2:3V
zF 2 96 500
e
potential ECu 2þ
=Cu
, and hence meCu2þ , can be deter-
mined by an emf method by harnessing the reaction and since aHþ ¼ pH2 ¼ 1; EMg2þ =Mg ¼ 2:3V:
Cu2þ þ 2Hg ! Hg2þ

2 þ Cu 1.01.4 Chemical Potentials and
which can be represented by a cell without a liquid Equilibrium
junction (or liquid junction potential)
The relationship between the change in free energy
Hg j Hg2 SO4 ðaq:Þ; CuSO4 ðaq:Þ j Cu DG and the equilibrium constant K is given by the
van’t Hoff isotherm
in which ‘j’ represents the phase boundaries. When a
cell has a liquid junction, the latter is represented by j DG ¼ RT ln K þ RT ln Q ½58
as in the cell and from eqn [48]
M1 j 1 ðaq:Þ
M2þ j 2 ðaq:Þ
M2þ j M2 aCc aDd
K ¼ ½59
aAa aBb
Although in certain cells the liquid junction can be
eliminated by appropriate choice of electrolyte solu- where a denotes the equilibrium activity of the com-
tion, this is not always possible. However, the liquid ponent and Q is the activity quotient. Note that the
junction potential can be minimized by the use of activities of the products are in the numerator and
a salt bridge (a saturated solution of KCl of about activities of reactants in the denominator.
4.2 M) and the liquid junction potential is then Equation [58] shows that DG is dependent on K
only 1–2 mV; this elimination of the liquid junction and on any arbitrary activity of reactants and pro-
potential is indicated by k. ducts that may be introduced into the equation.
The standard electrode potential of Mz+/M can When the reactants and products are in their stan-
be evaluated accurately and conveniently provided it dard states
forms a reversible electrode and can be coupled with aA ¼ aB ¼ aC ¼ aD ¼ 1 and ln Q ¼ 0 ½60
a reference electrode to form a reversible cell. How-
ever, a number of Mz+/M systems do not conform to Hence, from eqn [53]

this criterion, and calorimetric methods must be used DG ¼ DG ¼ RT ln K ½61

to evaluate DH and DS , and hence DG from which

E can be calculated. Thus, in the case of the ∴ DG ¼ RT ln K ¼ zFE ½62
Mgz+(aq)/Mg equilibrium, the enthalpy change dur- and
ing the dissolution of the metal in dilute acid at 25 C
is determined from the reaction RT 0:0591
E ¼ ln K ¼ log K ½63
zF z
MgðsÞ þ 2Hþ ðaqÞ ¼ Mg2þ ðaqÞ þ H2 ðgÞ;
since at 25 C, the term 2.303RT/F = 2.303 8.314

DH ¼ 461 200 Jmol1 298/96 500 = 0.0591 V. Combining eqns [52] and [58]
RT RT
and the standard entropy change DS is evaluated E¼ ln K ln Q ½64
from the standard partial molar entropies of the reac- zF zF
tants and products and from eqn [63]

DS ¼ S ðMg2þ Þ þ S ðH2 Þ S ðMgÞ
0:0591
E¼E log Q

z
þ 2S ðHþ Þ
0:0591 ac ad ½65
1 1 ¼E log Ca Db
¼ 43:3 J K mol z aA aB
which shows how the equilibrium potential E is so that

dependent on the standard electrode potential E , a

ð
1
constant for a given equilibrium, and the activities of dG ¼ nRT dp ½74
p
reactants and products. Equation [65] is termed the
Nernst equation and is applicable only when the sys- Integration of this requires a limit to be defined.
tem is reversible. Since The limit is taken simply as follows. We define a

X standard pressure p at which the Gibbs free energy

DG ¼ ui mi ¼ RT ln K ½66 has a standard value G . We have thereby defined
a standard state for this component of the system: a
P
ui mi standard temperature too is implicit in this since the
log K ¼ at 25 C ½67 above equations are treated for constant temperature.
5710
Integration of eqn [74] now gives
Thus, the equilibrium constant K can be evaluated
p
from standard electrode potential E or from the stan-

G ¼ G þ nRT ln ½75
dard chemical potentials m . p
The importance of the Gibbs free energy and the If only one mole of the gas is considered, the quantity
chemical potential is very great in chemical thermo- G is called the chemical potential, m, or the partial
dynamics. Any thermodynamic discussion of chemi- molar free energy. In this case, n ¼ 1; and eqn [75]
cal equilibria involves the properties of these becomes
quantities. It is therefore worthwhile considering the p
derivation of eqn [58] in some detail since it forms a m ¼ m þ RT ln ½76
p
prime link between the thermodynamics of a reaction

(DG and DG ) and its chemistry. Note that the standard state has simply to be defined;
From the defining equations of H (eqn [6]) and G there is no a priori reason why it should have any
(eqn [21]), differentiation gives particular value, save for the fact that it might as well
be a convenient value. A pressure of 1 atm is commonly
dH ¼ dU þ p dV þ V dp ½68 adopted. We can thus abbreviate the equation to

and m ¼ m þ RT ln p ½77
dG ¼ dH T dS S dT ½69 provided the pressure of the gas, p, is expressed
in atmospheres if the standard pressure has a numeri-
In addition, application of the first and second laws cal value of 1 atm. But note that the term inside
(eqns [3] and [13]) gives the logarithm is actually dimensionless because of the

implicit p .
dU ¼ T dS p dV ½70
If the gas is not ideal, so that the ideal gas equation
cannot be used, we replace the pressure p in eqns [76]
Combination of these gives
and [77] by the fugacity, f, such that the ideal gas
dG ¼ V dp S dT ½71 equation still holds if the pressure p is replaced
by the fugacity, an effective pressure, when the real
This equation is particularly important because by
pressure is p. This form is most convenient because of
carrying out a process isothermally (dT ¼ 0) the
the numerous ways in which nonideality can be
change in the value of G can be related to the volume
expressed and we note that the fugacity is related
and pressure change alone; both of these quantities
to, but not necessarily proportional to, the pressure.
are controllable and measurable. Thus, for isothermal
We can express the fugacity as a function of the
processes
pressure by introducing the fugacity coefficient, gp,
dG ¼ V dp ½72 as f ¼ gpp, which then replaces p in eqn [77] for the
nonideal case. The value of gp tends to unity as the gas
It remains only to integrate this expression. To do behaves more ideally, which means that the pressure
this, we apply initially the equation for n moles of an decreases.
ideal gas Consider now, a simple gas phase (with ideal
nRT gases) reaction
V ¼ ½73
p AþB!C ½78
10 Basic Concepts
For each component, we may write eqn [77]. a ratio of pressures as defined above) is really a
The free energy change, DG for the reaction, is then method of telling one’s audience what the standard
given by states really are. Thus, we could, for example, con-
sider 1 Pa of pressure as the standard state instead of
DG ¼ mC mB mA 1 atm. It would equally be possible to use the concen-
pC pB tration of the gas, since the ideal gas equation con-
¼ mC mB mA þ RT ln RT ln
pC pB tains the term n/V which is the concentration (in, say,
mol dm3), for which the standard state could be
pA ½79
RT ln 1 mol dm3. This does not of course change the reac-
pA
tion, but it does change the numerical value of K and

ðpC =pC Þ correspondingly the numerical value of DG . It is
¼ DG þ RT ln
ðpB =pB ÞðpA =pA Þ not common to do this in gas phase chemistry – the
usual standard state for gas is 1 atm. However, it is
Again, we define the standard state for a gas as 1 atm, important to understand the principle of it, particu-

that is, pA ¼ pB ¼ pC ¼ 1 atm so that the eqn [79] larly in corrosion and electrochemistry, since electro-
becomes chemical systems incorporate not only gases but also
pC solids and solutions, the latter comprising solutes and
DG ¼ DG þ RT ln ½80 a solvent. We can now express
pB pA
pC
Now, since DG is the driving force for the reaction to DG ¼ RT ln K þ RT ln ½83
pA pB
proceed, equilibrium is represented by the condition

that DG ¼ 0. Thus, the value of DG is given as Note that the term inside the second logarithm is not

ðpC" =pC Þ the equilibrium, unless the reaction is in equilibrium
DG ¼ RT (for which DG ¼ 0). These pressures are the actual
ðpB" =pB ÞðpA" =pA Þ

½81 pressures used for the reaction, and the equation

p" essentially describes the fact that the driving force
¼ RT ln " C"
pB pC for the reaction (as written) is represented by the
excess pressure quotient of the reaction over and
where the superscript e refers to the pressure of the above the equilibrium.
particular component when the reaction is at equilibrium. All the above deal with gases and gas phase pro-
It will be recognized that the term inside the loga- cesses. We can now turn to nongaseous components
rithm in eqn [81] is the equilibrium constant for the of the system. There are many ways of expressing
reaction. This relationship (together with eqn [80] is this. Probably, the simplest is to consider an ideal
extremely important because it links the thermody- solution of a solute in a solvent. If the solution is
namics with the chemistry. We have derived the fact ideal, the vapor pressure of the solute is proportional

that DG is simply another way of expressing the to its concentration, that is, p ¼ kc, where c is the
equilibrium constant for the reaction. It is important concentration and k is the proportionality constant.

once again to bear in mind that the term inside the Similarly, p ¼ kc which expresses the fact that the
logarithm is actually dimensionless: each pressure is standard pressure is related to a standard concen-
the ratio of the pressure to the standard pressure tration. Thus, for a particular component we have
(expressed in the same units). Thus, the equilibrium from eqn [76]
constant is also strictly a dimensionless quantity.
Although units are often quoted for the equilibrium c
m ¼ m þ RT ln ½84
constant K for the reaction (they would be atm1 in c
the above example), units are of course inconsistent
with the exponentiation of eqn [81], for which This formulation now entirely eliminates the need for
a pressure and replaces it by the concentration and a

pC" DG standard concentration. As defined above, the stan-
K ¼ " " ¼ exp ½82 dard concentration is that concentration which gives
pB pA RT
rise to the standard pressure p . However, we have
and the exponential term is always dimensionless. defined the original standard pressure arbitrarily as
Citing of units for K (sometimes called Kp since it is 1 atm. We can equally and validly redefine the
standard concentration of the solute as 1 mol dm3 of Exactly analogous, then, to the gas phase reaction

solution (unit molar or 1 M), or 1 mol kg1 of solvent above, the value of DG and DG can be formulated.
(unit molal or 1 m). This, in essence, redefines the Thus, for the following reaction, for example
standard pressure, but that is of little consequence,
since in corrosion and electrochemistry, we are usu- M2þ þ H2 O ! MO þ 2Hþ ½87
ally concerned with concentrations in solution (as an
electrolyte in water, for example) and with concentra- DG is given by
tions of metal in an alloy. Moreover, because the value 2
of k may be different for different components of the ðgMO cMO =cMO Þ gH cH =cH
DG ¼ DG þ RT ln ½88
system, from the point of view of the gas pressures, we ðgM cM =cM Þðgw cw =cw Þ
may be assigning different standard states for different
gases. But that does not matter either, since we are free where the subscripts MO, H, M, and w refer to the
to define the standard states in any way we choose, MO, H+, M2+, and H2O components, respectively.

and we are now interested in solution components and DG is derived from the equilibrium activities as in
not gases. Thus, for example, unit concentration of eqn [81], but by using the appropriate concentration
hydrochloric acid (say 1 m) is a far more convenient terms instead of pressure terms. Note that the equi-
standard state than that concentration which would librium constant and the activity quotient in the
equilibrate with a vapor pressure of 1 atm. However, above equation are again strictly dimensionless.
there is an added complication, already met with gases, Although dimensions are often quoted for K (some-
that many solutions are not ideal. So far, the derivation times called Kc) as described above (for reaction [82]
has still been in terms of ideal solutions because we this would be mol kg1 or mol dm3), these are sim-
assumed that the pressure is proportional to the con- ply an expression of the standard states involved.
centration; hence, the constant k is used above. Now, the standard state of the solid component
Handling of this is analogous to the nonideal gas. We (MO) and the solvent (H2O) are defined by conven-
define a new term, the activity a, of the component tion as the pure components (unit mole fraction). If

which replaces the concentration in eqn [84]. Thus the solution is dilute, cw cw and gw ¼ 1. Thus the
c components of the equation above involving MO and
a¼g ½85 H2O are both unity. In addition, if we define the
c
standard states of the dissolved ions as 1 M, then the
and thence for a nonideal component relationship becomes the more familiar
gC
m ¼ m þ RT ln ½86 g2H cH2
c DG ¼ DG þ RT ln ½89

gM c M
where c and c are the concentration and standard con-
centration, respectively, expressed as molalities, and g is Note that if the water in the reaction is not of unit
the activity coefficient. We are here defining the stan- mole fraction (or nearly so), then its activity would
dard concentration as 1 mole of solute per kilogram of have to be incorporated into eqn [89]. This could,
solvent. This simply means that we are fixing the value for example, be as in a mixed water/methanol

of the chemical potential of the component as m when solution. The water concentration would then be
the concentration is the standard value. Note that the expressed as unit mole fraction, and for accuracy, its
activity and the activity coefficient are dimensionless activity coefficient relative to pure water (the stan-
and a defined standard state is implied. dard state) would need to be known. The same is true
But that is not all. For dilute solutions, the solvent if one of the reactants is a metal (whose standard
concentration is high (55 mol kg1) for pure water state is the pure metal of unit mole fraction), but the
and does not vary significantly unless the solute is reactant is in alloy form. In summary, it is common in
fairly concentrated. It is therefore a common practice electrochemical reactions to adopt the following stan-
and is fully justified to use unit mole fraction as the dard states: for solid components and for the solvent,
standard state for the solvent. The standard state of a the standard state is unit mole fraction; for dissolved
close up pure solid in an electrochemical reaction is components (such as ions), the standard state is 1 mol
similarly treated as unit mole fraction (sometimes solute per kilogram of solvent, or 1 mol solute per
referred to as the pure component); this includes cubic decimeter of solution, and for gas phase compo-
metals, solid oxides, etc. nents, the standard state is 1 atm pressure.
12 Basic Concepts
1.01.4.1 Sign Convention for Equilibrium or

emfs and Potentials Zn ¼ Zn2þ þ 2e ½92a
For ions in solution, the standard reference state is since both are correct. Substituting in the Nernst
the hydrogen ion whose standard chemical potential equation eqn [65] for eqn [91a]
at aHþ ¼ 1 is given an arbitrary value of zero. Simi-
1
larly, for pure hydrogen at pH2 ¼ 1 atm, mH2 ¼ 0. E ¼ E 0:030 log ¼ E þ 0:030 log aZn2þ ½91b
þ
Thus, for the H =‰H2 equilibrium (eqn [57]), aZn2þ
P and for eqn [92a]
ui mi
E ¼ ¼ 0:00 V ½90
zF
E ¼ E 0:030 log aZn2þ ½92b
and the standard hydrogen electrode (SHE) is taken
as the reference electrode for all other equilibria. Thus, if the magnitudes of E are to be the same

For a cell, the equilibrium must be written in the the signs of E must be different, that is,
direction in which it proceeds spontaneously, and E ⊝ ¼ 0.76 V for eqn [91b] and E ⊝ ¼ þ0.76 V for
under these circumstances eqn [92b]. If this convention is adopted, the respec-
tive electrode potentials become
DG < 0 and E > 0
E ¼ 0:76 þ 0:030 log aZn2þ ½91c
that is, the emf of the cell will be positive. Thus, for
the equilibrium between Zn and H+, the reaction is E ¼ þ0:76 0:030 log aZn2þ ½92c
written as
and for any given activity of Zn2þ, the Es will be the
ZnðsÞ þ 2Hþ ðaqÞ ¼ Zn2þ ðaqÞ þ H2 ðgÞ same in magnitude but opposite in sign.
It is now internationally accepted (Stockholm
and if aH þ
¼ aZn
2þ
¼ aH2 ¼ 1, E ¼ E ⊝ ¼ þ0.76 V.
Convention) that the half-reaction must be written
However, in the case of a half-reaction, it is quite
with the electrons on the RHS:
arbitrary as to whether the equilibrium is written as a

reduction (electrons on the LHS) or as an oxidation Cu2þ ðaqÞ þ 2e ¼ Cu E ¼ þ0:34 V ½93
(electrons on the RHS)

Zn2þ þ 2e ¼ Zn ½91a Zn2þ ðaqÞ þ 2e ¼ Zn E ¼ 0:76 V ½94
1.02 Electrochemistry
C. Lefrou and R. P. Nogueira
Grenoble – INP – Grenoble Institute of Technology, UMR 5631 and 5266 of CNRS ‘‘LEPMI’’ and ‘‘SIMAP’’, Grenoble, France
F. Huet and H. Takenouti
Laboratoire Interfaces et Systèmes Electrochimiques, Université Pierre et Marie Curie – Paris 6, CNRS, UPR15-LISE,
Paris, France
ß 2010 Elsevier B.V. All rights reserved.
1.02.1 Introduction 14
1.02.2 Current, Potential, and Interfaces 15
1.02.2.1 Electronic and Electrolytic Materials: The Two Sides of an Electrified Interface 15
1.02.2.2 Electrochemical Reactions 16
1.02.2.3 Electrochemical Reactions and Corrosion: Some Preliminary Remarks 18
1.02.2.4 Basic Description of Metal–Electrolyte Interfaces 19
1.02.2.4.1 Elementary steps and intermediate interfacial species 19
1.02.2.4.2 The electrical double layer 20
1.02.2.4.3 Current and mass transport mechanisms 21
1.02.2.4.4 Potential and concentration profiles across the interface 22
1.02.3 Electrochemical Equilibrium 24
1.02.3.1 Definition and Introductory Remarks 24
1.02.3.2 Electrochemical Equilibrium as a Dynamic State: the Exchange Current Density 24
1.02.3.3 Formal Expression of Eeq: the Nernst Law for the Mz1 /M Couple 25
1.02.3.4 Some More Complex Applications of the Nernst Law 27
1.02.3.5 Applications of the Nernst Law to Pourbaix (E–pH) Diagrams 28
1.02.3.5.1 Purely potential-dependent equilibrium 28
1.02.3.5.2 Purely pH-dependent equilibrium 29
1.02.3.5.3 pH and potential-dependent equilibrium 29
1.02.3.6 Pourbaix Diagram of Gold, Iron, and Aluminum 30
1.02.4 Electrochemical Kinetics in the Case of a Single Forward–Backward Reaction 32
1.02.4.1 Qualitative Approach 32
1.02.4.2 Quantitative Approach 33
1.02.4.2.1 Activation-controlled systems: Butler–Volmer equation 33
1.02.4.2.2 Mass-transport limited systems: Fick’s law and limiting current density 38
1.02.4.2.3 General case: Mixed controlled systems 39
1.02.5 Corrosion: Far from Equilibrium Zero Net Current Systems 41
1.02.5.1 Qualitative Description of the Free Corrosion Scenario 41
1.02.5.2 Electrochemical Kinetics Around the Corrosion Potential 43
1.02.6 The Electrochemical Cell 45
1.02.6.1 Design of the Electrochemical Cell 45
1.02.6.1.1 The working electrode 46
1.02.6.1.2 The counter-electrode 46
1.02.6.1.3 The reference electrode 46
1.02.6.2 Potential Control and Current Control 49
References 50
IUPAC International union of pure and applied

Abbreviations chemistry
CE Counter electrode
NHE Normal hydrogen electrode
DC Direct current
OCP Open-circuit potential
13
14 Basic Concepts
OHP Outer Helmholtz plane Rp Polarization resistance (O)

OPA Operational amplifier S Electrode surface area (m2)
RE Reference electrode t Time (s)
SHE Standard hydrogen electrode T Temperature (K)
WE Working electrode Vset Set potential (V)
ZRA Zero-resistance ammeter y Axial distance to a disk electrode (m)
z Number of electrons exchanged in a reaction
‘ Related to the bulk of the electrolyte
[X] Concentration of species X (mol l1)
Symbols 0
Related to the standard state
a Activity of a species a Charge-transfer coefficient
a Related to the anodic reaction b Tafel slope (V per current decade)
act Related to activation control d Thickness of the diffusion layer (m)
A1 Constant in eqn [36] (V) DX Difference in X values (X unit)
b Tafel coefficient (V1) Dr G G variation for reaction r (J mol1)
C Concentration of a species (mol m3) « Thickness loss (m)
c Related to the cathodic reaction f Flux of species (mol m2 s1)
d Distance between two electrodes (m) F Electrical potential (V)
D Diffusion coefficient (m2 s1) g Electrolyte conductivity (O1 m1)
e Charge on the electron ¼ 1.602 1019 C h Overpotential (V)
E Electrical potential (V) n Kinematic viscosity (m2 s1)
Ecorr Corrosion potential (V) p Polarization (V)
Eeq Equilibrium potential (V) u Coverage ratio of adsorbed species
F Faraday constant ¼ 96 486 C mol1 r Density (kg m3)
g Constant in eqn [32] v Electrode angular rotation rate (rad s1)
G Chemical Gibbs free energy (J mol1)
G Electrochemical Gibbs free energy (J mol1)
GI Gain of a current amplifier
GV Gain of a voltage amplifier 1.02.1 Introduction
int Related to the interface
I Current (A) The starting point of a chapter devoted to the intro-
I0 Exchange current (A) duction of basic aspects of electrochemistry that are
Icorr Corrosion current (A) absolutely vital for understanding corrosion pro-
Iset Set current (A) cesses can take advantage of the meaning of corrosion
j Current density (A m2) itself. In fact, corrosion processes encompass reactive
j0 Exchange current density (A m2) phenomena of metal oxidation, hence yielding a pro-
jcorr Corrosion current density (A m2) gressive and more or less fast chemical deterioration
k Reaction rate constant (m s1) of the material exposed to an aggressive environment
K Constant in eqns [31] and [32] (mol l1) by oxide formation or free dissolution in the form of
lim Related to diffusion limitation dissolved cations. This implicates the existence of
mix Related to mixed control oxidation–reduction reactions, hereafter referred to
mt Related to mass transport control as redox reactions, in which the role of the metallic
M Molar mass (kg mol1) surface and its interactions with the surrounding
n Number of moles (mol) medium are particularly crucial. The basic concepts
N Avogadro’s number ¼ 6.022 1023 issuing from electrochemistry (roughly speaking,
q Electrical charge (C) the branch of chemical sciences devoted to the
r Disk electrode radius (m) study and investigation of reactive systems involv-
ref Related to the reference electrode ing electrical charge transfer) are well adapted and
R Gas constant ¼ 8.31 J mol1 K1 necessary for a convenient description. These basic
Rcontrol Current-control resistance (O) concepts will be helpful for the subsequent under-
Rct Charge-transfer resistance (O) standing of complex corrosion phenomena. The main
Re Electrolyte resistance (O) goal of this chapter is, therefore, to provide an overview
Rm Current-measuring resistance (O) of basic electrochemical concepts, with a special focus
Electrochemistry 15
on electrochemical reaction near the equilibrium different materials with two different metals at the
potential. Then, these concepts will be extended to battery terminals. Since they involve electrical energy
those directly related to corrosion processes. For the exchange, the electrochemical systems only employ
sake of readability for nonexpert readers in electrochem- conductive materials. Electrical conductivity exists as
istry, simplified representations of reactive interactions, soon as the materials have mobile charges. Without
actually much more complex, will be introduced in some going through an exhaustive survey, these key materials
cases. This is a conscious option taken by the authors can be classified into three main categories according
to provide a self-sustained text conveying the main to the nature of the charge carriers: electronic conduc-
concepts and tools to understand the physicochemical tors like metals or semiconductors, ionic conductors
processes associated with corrosion phenomena. (or electrolytes), and finally mixed conductors (such as
some conducting polymers) when the two types of
electrical conduction coexist. To illustrate corrosion
1.02.2 Current, Potential, and phenomena in the following, the text will concentrate
Interfaces only on metals, which are electronic conductors under-
going corrosion, and on electrolytes, which can be
1.02.2.1 Electronic and Electrolytic
aqueous or organic solutions, molten salts, ionic solids,
Materials: The Two Sides of an Electrified
polymers, or even gases.
Interface
Besides the electrical conductivity of these mate-
Electrochemistry is a science, as well as an industrial rials, the specificity of the electrochemical systems
domain, based on physicochemical phenomena tak- comes from the way they are assembled. Regions in
ing place whenever electrical and chemical energy which the different materials (the metal and the
exchanges intervene in a reactive scheme. In this electrolyte, for instance) are in contact and in which
sense, most of the time electrochemistry investigates the electrical continuity between the electronic and
systems in which electrical currents flow, as in the ionic conductions is ensured are generally called
case of electrochemical generators or batteries pres- interfaces. (The term ‘interphase’ would probably
ent in numerous everyday applications such as cars, be more appropriate for bringing up the idea that
mobile telephones, and so on. From this point of view, this is not a single surface in a purely mathematical
the archetypical corrosion process, that is a metallic point of view, but a small physical volume usually a
piece freely exposed to a given aggressive environ- layer of some nanometers as roughly described here-
ment is, to some extent, a bit marginal in the sense after in Section 1.02.2.4.2) The simplest electro-
that the overall current flow observed from the out- chemical assembly is thus composed of two metals
side of the system is zero; the system spontaneously having an interface with the same electrolyte. This
evolves and is said to be at rest or at open-circuit. kind of system is referred to as a galvanic chain,
This is clearly the actual practical corrosion scenario, where each interface is schematically represented
unless, an external source is applied as in the case of by a vertical stroke as in the following examples:
corrosion characterization or measurements and in
some corrosion protection techniques such as Ptjaqueous solutionjAu
cathodic or anodic protection. In spite of this feature, PtjFe2þ ; Fe3þ aqueous solutionjAu ½1
the concepts and tools involved in general electro-
chemistry remain valid and constitute the corner- In the first example, where no corrosion process is
stone of corrosion description and understanding. expected at least in normal experimental conditions
It is possible to find redox reactions in a single (concentration, temperature, etc.), both interfaces are
phase, as in a bulk solution. Nevertheless, when talk- said to be impermeable and nonreactive. In the second
ing about electrochemistry and more particularly cor- example, where the electrolyte is composed of an
rosion, only heterogeneous systems are concerned, aqueous solution containing Fe2þ and Fe3þ cations,
which means that the reactive system is composed of one of the most reactive couples and one often
two or more distinct phases. As an example, let us employed in analytical electrochemistry, the interfaces
consider a classical electrochemical system implying would still be impermeable but could then be reactive –
electrical energy exchanges such as a battery con- even if corrosion of platinum and gold is considered as
nected either to an electrical load (the battery hence being negligible – since, under certain conditions,
discharges) or to an external electrical energy source Fe2þ oxidation at one interface and Fe3þ reduction
(the battery hence charges). This is usually achieved at the other should be observed. The term imperme-
by a multiphase ‘sandwich’ of at least three able implies that the mobile species (the charge
16 Basic Concepts
carriers) at each side of the interfaces are different. heterogeneous chemical reactions taking place at the
Free electrons are mobile and can hence ensure the interface, which allow electrons to be released by
current flow across the metallic phases but are not mobile species in the electrolyte and incorporated
present in the electrolyte phase, in which ions are into the electron ‘gas’ in the metallic phase (anodic
mobile and ensure the electrical charge transport reaction) or, conversely, allow electrons leaving the
between the two metallic phases. In some more com- electronic gas at the metal side to be trapped by an
plex situations, such as certain corroding systems electrolyte species (cathodic reaction). (Remember
where a more or less ionically conductive oxide may that one of the most important characteristics of metals
partially or totally cover the metallic surface, hence is that the valence electrons are free in a state that can
creating an intermediate layer between the metal and be likened to a gas.) These heterogeneous (since they
the electrolyte, an interface between two ionic conduc- concern both the electrolyte and metallic phases) inter-
tors is established and referred to as an ionic junction. facial reactions entail the loss or the gain of electrons
In contrast to that double interface configuration, through a charge-transfer process (either oxidation or
from a macroscopic point of view, most corrosion reduction). A single oxidation or reduction reaction is
processes take place at a single interface. This is the known as a half-cell reaction (since another reaction of
case, for instance, of the simplest corrosion situation: the opposite type is needed to complete the full current
a piece of metal immersed in an electrolyte. It is path and create an electrochemical cell).
obviously not possible for a net electrical current to Whenever the current flow is straightforwardly
flow in or out of this closed system. As discussed in related to one or several heterogeneous reactions,
more depth later in this chapter, this does not mean one refers to Faradaic current or, more generally, Far-
that no electrochemical reaction is taking place at the adaic processes because these phenomena can be
interface. It is also important to notice that this mac- described by a quantitative law called Faraday’s law.
roscopic single interface can be considered either as The current flow across the interface can also be
uniform, as for the corrosion of an almost homoge- related to other types of electrochemical processes,
neous metallic surface (uniform or generalized corro- namely capacitive processes, which correspond mostly
sion) or nonuniform, as for the corrosion of a markedly to transient phenomena and will briefly be discussed in
heterogeneous metallic surface where inclusions or Section 1.02.2.4.2. Of much concern in corrosion
precipitates are preferentially corroded rather than processes, Faraday’s law mathematically expresses the
the metallic matrix (an example of localized corrosion). relationship between the electrical charge transferred
It is then easy to understand that in the latter case two during a heterogeneous reaction and the amount
interfaces are actually concerned (metal|electrolyte| of mass transformed at the same time. This is the
inclusion) in a microscopic approach, even if there is a Lavoisier principle of conservation of mass extended
single macroscopic interface. to electrons. As shown in the following example,
Faraday’s law is the very origin of the relationship
(extremely important in corrosion engineering)
1.02.2.2 Electrochemical Reactions
between the corrosion current (experimentally mea-
As indicated above, a common example of an electro- sured or estimated) and the corrosion rate (in terms of
chemical system is that of an impermeable interface mass or thickness loss), in the case of uniform corro-
between a metal and a solution containing ions where sion. Let us consider the case of a uniformly corroding
the species (charge carriers) ensuring the current flow iron plate. For a divalent corrosion reaction
at each side of the interface cannot traverse it. Indeed,
Fe ! Fe2þ þ 2e ½2
on the one hand electrons are ‘free’ in the metallic
lattice but trapped in the liquid phase by different the corrosion current density (the current density is the
cationic configurations and, on the other hand, ions corrosion current normalized by the geometrical sur-
can move in a liquid but are macroscopically frozen face area and is a simple and intuitive way of comparing
in the metallic phase by the lattice arrangement. corrosion rates independently of the size of the metallic
This means that there is no straightforward continuity surface considered) corresponding to the consumption
of the charge transport across the interface. Conse- of a certain quantity of Fe moles per unit time and per
quently, a steady-state current flow, for example, from unit surface can be expressed by Faraday’s law, as
the electrolyte to the metal requires the transforma-
tion at the interface of the ionic flow in the electrolyte 1 Dn j
¼ ½3
into an electronic flow in the metal. This is done by S Dt 2F
Electrochemistry 17
where Dn/Dt is the number of moles of iron consumed In classical electrochemical situations, the system
per unit time, S is the corroding surface area, j is comprises two interfaces and the existence of a
the corrosion current density, and F is the Faraday steady-state current flow implies that one of the inter-
constant. This constant represents the electrical charge faces behaves as a cathode whilst the other plays the
of one mole of electrons, that is, F ¼ Ne, where N is role of an anode; that is, one of them consumes the
Avogadro’s number and e the charge on the electron electrons produced by the other. Two typical scenar-
(F ¼ 6.0221023 1.6021019 ¼ 96 486 C mol1). ios are depicted in Figures 2 and 3. They illustrate,
For engineering applications, the expression of the for instance, what happens in a mobile telephone
corrosion rate in terms of thickness loss per year, for when it is being used (its battery is discharging,
instance, is more useful: Figure 2, or being recharged, Figure 3).
e j M Figure 2 presents the situation in which there is a
¼ ½4 spontaneous flow of electrons produced by an inter-
t 2F
facial oxidation reaction at the anode. These elec-
where M is the molar mass of the corroding compound trons go through a resistor to be consumed at the
and r is its density. In the case of iron corrosion
(M ¼ 55.84 g mol1 and r ¼ 7.85 g cm3), a uniform
corrosion current density of 5 mA cm2, corresponds
to a thickness loss of 58 mm year1. A convenient rule M E
of thumb is that for most metals 1 A m2 corresponds e l
t e
to a penetration rate of approximately 1 mm year1. a c
Let us come back to the more general description l t
r
of the processes related to a current flow. One of the o
Fe3+
simplest examples in electrochemistry concerns the l
interface between a platinum electrode and an elec- y
e–
t
trolyte containing Fe3þ and Fe2þ ions. If an external Fe2+
e
power source imposes an electron flux from the metal
towards the interface, a heterogeneous chemical reac-
tion takes place at the interface. In this example, the Figure 1 Schematic representation of an impermeable
metal–electrolyte interface at which a reduction reaction
transformation of ferric ions into ferrous ions takes takes place; the interface acts as a cathode.
place according to the reaction:
Fe3þ þ e ! Fe2þ ½5
This reaction indicates that a certain amount of Fe3þ e– Resistor I
cations, proportional to the current, is transformed
into Fe2þ ions, capturing the electrons coming from
the metal and ensuring the continuity of current flow.
This interface, on which a reduction reaction occurs,
is called a cathode. In the case of a current flow in the A C
opposite direction, electrons are produced by the n a
redox reaction. Thus, the interface is the site of an o t
d h
oxidation reaction and is called an anode. Figure 1 e o
schematically illustrates this phenomenon for reac- d
tion [5]. The arrows represent the direction in which e
each reacting species moves (more rigorously, they I
indicate the direction of the corresponding mass
flow vector). –
+
–
It can be seen in Figure 1 that, even in this Electrolyte
very simple example, the current flow entails at least
three important phenomena: electron transport inside
Figure 2 Electrochemical battery during discharge:
the metal, ionic transport in the electrolyte, and the electrons are delivered to the external circuit by oxidation at
heterogeneous redox reaction, that is, charge-transfer the anode (Red ! Ox þ ne ) and are consumed by
reaction. reduction at the cathode (Ox þ ne ! Red).
18 Basic Concepts
cathode by a reduction reaction. In that case, the 1.02.2.3 Electrochemical Reactions and
chemical energy of the reactants is converted to Corrosion: Some Preliminary Remarks
electrical energy, which is then converted into ther-
In most corrosion processes, a single macroscopic
mal energy by the Joule effect in the resistor. For
interface is concerned. However, a system at rest,
mobile telephones, the energy delivered by the elec-
which is the most common corrosion situation, is a
trochemical system is converted into light, sound,
particular case of an electrochemical system under
radio waves, etc. As soon as the discharge process
discharge analogous to that in Figure 2 in the sense
leads to a chemical transformation, the recharge pro-
that the oxidation and reduction reactions are spon-
cess requires the reverse reaction to occur. While the
taneous. In contrast with Figure 2, no external circuit
discharge reaction in Figure 2 takes place spontane-
allowing the current to flow is available and the
ously, the recharge reaction can only occur if an
overall current flow is null since both the anodic
external power source is connected to the system, as
and cathodic reactions take place at the same metallic
illustrated in Figure 3. It is exactly like in mechanics
surface. This idea can be easily handled by consider-
when a ball that will spontaneously roll down a slope
ing the example of galvanic coupling depicted in
needs an external force to go up the slope. The
Figure 4, in which two different metals (iron and
electrode that was spontaneously a cathode is forced
zinc) assembled together to form a single piece are
to work as an anode and the electron flow is reversed.
exposed to an electrolyte rich in dissolved oxygen.
It is worth noticing that the current and electron
In this situation, the zinc is an anode and hence
flows are identified by arrows in opposite directions,
corrodes whilst the iron tends to be protected since
as illustrated in Figures 2 and 3, since by convention
it plays the role of a cathode on which oxygen is
the current flow direction is given by the positive
reduced. It is important to notice that the fact that
charge carriers.
iron is a cathode does not mean that iron will neces-
In another international convention (IUPAC),
sarily be deposited on the metallic surface according
largely accepted but not always respected, it is con-
to the reverse of reaction [2]. It means that the iron
sidered that the anodic and cathodic currents are
surface is the preferential site for cathodic (i.e.,
positive and negative, respectively. As a consequence,
reduction) reactions. The exact nature of these reac-
when anodic and cathodic processes occur on the
tions depends on the electrolyte and other important
same electrode, as in corrosion processes, the current
characteristics of the electrochemical systems that
is then positive when it flows from the anodic areas to
will be better discussed in Section 1.02.2.4.1.
the electrolyte and negative when it flows from the
For the sake of clarity, the cathodic reaction is
electrolyte to the cathodic areas.
e– − Power source + I
Fe
O2 E
C A l
a n Ic e
t o c
h d t
o e r
d I e– I o
e l
Zn y
Zn2+
I t
e
+ Ia
–
Electrolyte
Figure 3 Electrochemical battery in charge: electrons are Figure 4 Null net current balance in a freely corroding
consumed by reduction at the cathode (Ox þ ne ! Red) system: galvanic coupling between iron and zinc in contact
and produced by oxidation at the anode (Red ! Ox þ ne ). with an electrolyte rich in dissolved oxygen.
Electrochemistry 19
represented in Figure 4 by oxygen reduction, one of cathodic zones in any given electrode are different,
the most important reactions in corrosion processes. eqn [8] clearly points out that the system may have
Many similar situations with anodic and cathodic very high partial current densities because of the
areas on the same electrode are encountered in corro- surface ratio. This is extremely important in corro-
sion, as when a metallic surface is not perfectly homo- sion applications. As an example, let us go back to the
geneous, which is often the case. For instance, there are iron–zinc example and consider the possibility of
regions that are preferentially anodic, such as grain using zinc rivets to assemble large iron plates. Equa-
boundaries, intermetallic inclusions, zinc-rich phases tion [8] immediately shows that the zinc rivets in
in brasses, and so on. The existence of cathodic and contact with an electrolyte such as rain water corrode
anodic areas along the same surface can also be the very rapidly because of the very high anodic (corro-
result of inhomogeneous electrolytes, such as those in sion) current density related to the huge cathodic
contact with partially submersed structures. Indeed, (iron)/anodic (zinc) surface ratio.
dissolved oxygen is easily available near the air–liquid
interface, inducing a preferentially cathodic behavior 1.02.2.4 Basic Description of Metal–
of the metallic surface. For a completely uniform Electrolyte Interfaces
metallic surface immersed in a perfectly homogeneous
electrolyte, the idea of a zero net current balance 1.02.2.4.1 Elementary steps and intermediate
related to simultaneous anodic and cathodic reactions interfacial species
still holds, even if no preferential anodic or cathodic The simplified and common aspects of electrochemi-
locations appear at the interface. Indeed, each surface cal reactions taking place at metal–electrolyte inter-
site dynamically and randomly switch between anodic faces considered up to now deserve a better although
and cathodic behavior so that the overall net current still simple description. The first important aspect is
is permanently zero and general corrosion occurs. that the overall charge-transfer reactions involved in
Thus, for the bimetallic surface in Figure 4 as well corrosion processes are often the consequence of
as for any system left at rest, the anodic, Ia, and complex interfacial phenomena that can be described
cathodic, Ic, currents must perfectly and permanently by a more or less intricate reaction mechanism com-
balance out each other, so that no overall current prising several intermediate elementary steps. The
flows to or from this closed system, so that: reactive species implied in this mechanism are not
only the ensemble of reactants and products of the
IZn ¼ Ia ¼ Ic ¼ IFe > 0 ½6 overall reaction such as electrons, the metal itself, ions
(The equality represented in eqn [6] is only true for or dissolved gas as in the previous examples, but also
the average behavior; short-term differences can be other species produced as reaction intermediates.
accommodated by charging or discharging of the Intermediate species can be very important in the
double-layer capacitance, and this leads to the phe- kinetics of heterogeneous phenomena. In corrosion,
nomenon of electrochemical noise.) It is important to these species are surface adsorbates that physico-
note that in this equation, the anodic and cathodic chemically adhere to the metallic surface, such as
currents are identical in absolute values. Also, this water itself, neutral species or anionic species. Let
equation holds for currents, Ia and Ic, and not for us consider two examples of complex mechanisms
current densities, ja and jc, defined as follows: concerning corrosion processes. The first is the
mechanism of nickel corrosion (to a first approxima-
Ia ¼ ja Sa ¼ jc Sc ¼ Ic ½7 tion, this is also the mechanism of iron dissolution)
where Sa and Sc are the surface of the anodic (zinc) proposed in the literature1,2:
and cathodic (iron) areas, respectively. Thus Ni þ H2 O ! þ
NiOHads þ H þ e

Sc NiOHads !Ni2þ þ OH þ e ½10

j a ¼ j jc j ½8
Sa
The simple algebraic addition of both elementary (sin-
or gle electron) reactions gives the global reaction [11]:
Sa Ni!Ni2þ þ 2e ½11
j jc j ¼ ja ½9
Sc (or reaction [2] in the case of iron). This multielectro-
If the surface areas of iron and zinc in Figure 4, and nic reaction is the overall and simplified chemical
more generally the surfaces of preferential anodic or representation of a more complex process. It is also
20 Basic Concepts
important to notice that the intermediate adsorbed reducing the metal dissolution rate. Since the corro-
species (identified by the subscript ‘ads’) plays an sion takes place at a zero overall current, the decrease
important role in the kinetics of the overall reaction. in the rates of either of these processes (anodic or
If the first (NiOHads adsorption) or the second cathodic) slows down the net corrosion rate.
(NiOHads desorption giving Ni2þ) elementary step is Reactions [10]–[14] are simple mechanisms
hindered, the overall Ni dissolution is slower since involving single-adsorbate reactions. Other impor-
Ni dissolution occurs through two consecutive steps. tant reactions, such as iron dissolution9–14 or oxygen
Adsorbates can also be produced by cathodic reac- reduction,15,16 can involve several adsorbates, the
tions, an important example coming from the proton identification of which is often a difficult task. Besides
reduction on metallic surfaces, roughly represented as these complex features of redox reactions with the
presence or the absence of intermediate species, the
2Hþ þ 2e !H2 ½12
current flow across the interface is also dependent on
but actually composed of two (or more) intermediate the transport of reactive species in the electrolyte.
elementary steps, as those of the so-called Volmer– Hence, a complete or at least better description of the
Heyrovski route3: kinetics of the metal–electrolyte interface must
encompass the behavior of these charge carriers in
Hþ þ e !
Hads ½13
the interfacial region.
Hads þ Hþ þ e !H2 ½14
1.02.2.4.2 The electrical double layer
It is easily seen that the addition of the Volmer (reac- In a first approach, this description can be made
tion [13]) and Heyrovski (reaction [14]) elementary simpler by considering a metal–electrolyte interface
steps gives the global representation of reaction [12]. at which no redox reaction takes place so that no
(An alternative to the Heyrovski reaction, known as Faradaic current flows across it. This interface is
the Tafel reaction, involves the direct combination of thus impermeable and nonreactive like the first one
two adsorbed hydrogen atoms to form a hydrogen in scheme [1] and the only observable phenomenon is
molecule – this dominates at low overpotentials.) the accumulation of species at both sides of the inter-
Adsorbates can also play a critical role in the overall face. When these species are electrically charged,
kinetics of corrosion since they constitute some of the important electrical constraints appear. Indeed, since
elementary building blocks of corrosion mechan- both metallic and electrolytic phases are conductive,
isms.4 It is interesting to note that, under certain the attractive and repulsive electrical forces exerted
conditions, the adsorption site can be considered as on the charge carriers imposes electroneutrality (that
temporarily blocked by the Hads adsorbate from the means a net charge equal to zero) of any significant
point of view of corrosion,5,6 which means that it is not volume of metal and electrolyte. (Strictly there can be
able to dissolve as long as this site is occupied by Hads. a net positive or negative charge at a point. This is
This is obviously not an evidence of an effective described by the Poisson equation, which states that
protective role of Hads against corrosion, otherwise the local charge density ¼ee0 d2’/dx2, where ’ is
acidic solutions would not be aggressive, which is the local potential, e the dielectric constant and e0 the
evidently far from true. It is worth recalling that the permittivity of free space. However, ee0 is so small that
hydrogen evolution reaction in acidic media is rather the net charge is negligible except in exceptional
fast, so that Hads is transformed into H2 at a significant circumstances (such as at the metal–solution inter-
rate. In addition, the active dissolution rate, for face).) Significant deviations from electroneutrality
instance of iron, is also fast. Even if only 1% of the can only exist at very short distances, as in a water
surface area is devoted to this process, the corrosion molecule composed of two Hþ and one O2. Never-
current density is far from zero. This brief discus- theless, due to the intrinsic differences in the conduct-
sion illustrates the complexity of corrosion kinetics. ing mechanism in each phase (electronic conduction
In other situations, the protective role of adsorbates as in the metal and ionic conduction in the electrolyte), a
blocking agents is extremely important and is fully local deviation from electroneutrality occurs at the
employed in the development of inhibitors for corro- interface. For example, as schematically represented
sion protection.7,8 Very often, in the presence of by a solid line in Figure 5, an excess of electrons at the
organic inhibitors, the cathodic reaction rate is slowed metallic side can permanently exist provided that
down markedly, whereas inorganic substances, so- positive charges (for instance, an excess of cations)
called anodic inhibitors, are more effective at are in excess at the other side.
Electrochemistry 21
E OHP – E
− +
l l
M + e
e
e M − + + c
ΔF int c e
t t t
t – r
a −
a r + + o
l o –
l l
l − + y
y + t
+
t − e
e
–
Compact Diffuse layer

Figure 5 Schematic representation of the electrical
layer
charge density (solid line) and of the induced potential F
(dashed line) at both sides of the interface. Figure 6 Simplified scheme of the electrolyte double
layer: compact and diffuse layers with solvated cations and
anions in the electrolyte.
Even in these conditions of local charge separa-
tion, the overall electroneutrality must be respected,
which means that these face-to-face opposite charges the closest centers of the ions and the metallic sur-
in excess at the interface must be perfectly balanced face defines the so-called outer Helmholtz plane
out. These spatial and electrical configurations are (OHP),18 which is the boundary of the compact
essentially the same as those of an electrical capacitor layer (thickness of about 1 nm). Under certain condi-
composed of two parallel plates electrically charged tions, ions can show specific adsorption behavior at
with opposite signs. However, this electrochemical the interface (called electrosorption) so that they can
capacitor typically has a distance between the plates penetrate into the compact layer. As a consequence,
of only a few nanometres. The charge distribution they establish a new lower minimal approach distance
across the interface entails a potential shift DFint of the closest adsorbed ion centers that defines the
across the interface, as illustrated by the dashed line inner Helmholtz plane (IHP), which logically modi-
in Figure 5. In steady-state conditions, the electro- fies the potential profiles in this region.
neutrality being respected, no current flows across the Beyond the Helmholtz plane, the double layer
electrochemical capacitor. Nevertheless, any poten- spreads into the diffuse layer. The thickness of this
tial or current perturbation of the system may induce layer depends on the ionic composition of the elec-
a local rearrangement of the charge distribution at trolyte. It is about 1 mm in pure water, whereas it may
each side of the interface, which leads to a transient vary in an electrolytic medium from several tens to a
flow of ‘capacitive current.’ As a consequence, the fraction of nm depending on the salt concentration.
measured current in an electrochemical experiment For instance, the typical thickness of a diffuse layer
generally includes the effects of this capacitive pro- varies from 6 to 20 nm when the concentration of the
cess, in addition to the Faradaic response. ionic species in the electrolyte decreases from 0.01 to
As shown schematically in Figure 5, the distribu- 0.001 mol l1. Thus, in the presence of a supporting
tion of the excess charge at each side of the interface electrolyte (auxiliary electrolyte constituted of non-
is asymmetric. It is sharply confined in a virtually reactive ions at high concentration to increase the
zero volume at the metallic side, but spread out over a electrolyte conductivity) the diffuse layer becomes
nonnegligible distance at the other side. This elec- much thinner.
trolyte zone, in which the electroneutrality is not
respected, is called double layer17,18 as it is frequently 1.02.2.4.3 Current and mass transport
divided into a first layer, close to the metallic surface, mechanisms
called the ‘compact layer’ and a second, called the As mentioned in the preceding sections, electrochem-
‘diffuse layer,’ as represented in Figure 6.19 ical kinetics encompasses not only interfacial reac-
In order to better understand how the double tions but also the transport of mobile species inside
layer is configured, let us remember that ions in an the conductive phases to and from the interface. With-
electrolyte are solvated (roughly speaking, for an out going into mathematical details, the current in a
aqueous electrolyte, they are wrapped by water mole- given infinitesimal volume is the resultant of macro-
cules). As shown in Figure 6, the distance between scopic movements of all the charged species in this
22 Basic Concepts
volume. A complete discussion on mass transport and concentration generated by chemical perturbations at
its influence on electrochemical processes can be the interface, move progressively towards the bulk
found elsewhere.20–22 In this chapter, it is sufficient solution as the diffusion front progresses. However,
to consider that the mass transport phenomena can be convection often occurs in practice, both natural
macroscopically described with three types of driving convection and especially forced convection, as in
force, roughly portrayed as follows: water distribution networks, pipelines, and many
industrial systems. An accurate and quantitative anal-
Migration is the movement of charged species
ysis of convection phenomena is covered in another
under the effect of an electrical field (which is an
section of this book but the general result of Nernst’s
electrical potential gradient). It may be neglected
model illustrated in Figure 7 can be retained. This
for the electroactive species in the presence of a
model predicts that in spite of the presence of con-
supporting electrolyte.
vection, there is a region close to the interface (called
Diffusion is a process analogous to that of migration
Nernst layer) in which the convective forces vanish
in which species are transported under the effect of
since the electrolyte velocity at the interface is actu-
a concentration gradient towards the regions of
ally zero. This is the consequence of the existence of
low concentration, which tends to homogenize
viscoelastic forces at the interface that prevent the
the solution.
liquid phase from sliding along the solid phase and
Convection is the relative movement of a fluid
lead to a continuous decrease of the convection effects
medium, liquid or gaseous, which takes place
from the bulk to the interface. The model approxi-
even for macroscopically stationary electrolytes.
mates this monotonic evolution by assuming a discon-
This is the case of natural convection caused by
tinuity between the bulk and the Nernst layer. Mass
mechanical vibrations or by density gradients, such
transport is controlled by diffusion and migration
as those induced by local differences in tempera-
inside this layer and the reactants exhibit concentra-
ture between the wall of a reservoir and the bulk
tion gradients, while outside this layer the electrolyte
solution, by gas evolution or by the release of dense
is stirred by convection and, therefore, the reactant
ions into solution. Convection can also be forced
concentrations are considered to be homogeneous.
by mechanical stirring or, very often in electro-
The thickness of the Nernst layer, also called diffusion
chemistry, by using rotating disk electrodes or
layer, depends on the stirring conditions. Typical
impinging jet cells.
thicknesses are 10 and 300 mm for forced and natural
In electrochemistry, under certain experimental con- convection, respectively.
ditions (see examples below), diffusion phenomena
are negligible compared to migration and convec- 1.02.2.4.4 Potential and concentration
tion at a distance sufficiently far from the interface. profiles across the interface
Hence, the region in which diffusion cannot be The spatial variations of the reactant concentrations
neglected, compared to other mass transport mechan- and of the electrical potential in the region close to
isms, is defined as the diffusion layer. This zone close
to the interface plays a crucial role in the kinetics of
electrochemical phenomena. It can spontaneously
appear because of the interfacial depletion or abun- E
l
dance of reactants induced by the (electro) chemical e
Diffusion Convection
reactions, since only the species present at the inter- M c
face are able to react. Depending on the kinetics of the e t
t + +
r
reactions, the interfacial concentrations can be mark- a o
edly different from those in the bulk solution (the l Migration Migration l
region where the concentration is homogeneous) and y
are a key parameter for the description of electro- t
e
chemical and hence corrosion phenomena.
In the absence of convection, the current flow or
simply the natural evolution of a metal–electrolyte Nernst layer
interface at rest, as in corrosion processes, induces Figure 7 Simplified schematic representation of the
the formation of a diffusion layer that grows con- metal–electrolyte interface for the Nernst mass-transport
tinuously with time. The changes in the electrolyte model.
Electrochemistry 23
the interface, play an important role in the accurate Indeed, the double layer being so thin at this scale,
modeling of the kinetics of electrochemical phenom- the interfacial concentration Cint of a reactant, that is,
ena and hence of the corrosion processes. First of all, the concentration at the Helmholtz plane (see the
it is important to point out the differences in the zoom in Figure 8), which is the key parameter
thickness of the double layer (typically a few nano- concerning the kinetics of the mass transport and
metres) and that of the diffusion layer (typically a few redox reactions, may be considered as constant in
tens of micrometres). Figure 8 presents the sche- the double layer. In the same way, even if the poten-
matic concentration and potential profiles across the tial varies strongly in the double layer, it is sufficient
double and diffusion layers. In most cases, the diffu- to consider the driving force of the redox reactions as
sion layer is actually the most pertinent scale for a being controlled by the interfacial potential differ-
convenient analytical approach, so that it is sufficient ence DFint, that is, the potential difference across
to define the electrical potential and the reactant the double layer. In the diffusion layer and in the
concentrations at the boundary of the double layer. electrolyte bulk, the potential will vary according to
Cbulk
E
Cint l
e
c
M
t
e
ΔFint r
t
o
a
l
l
y
t
e– e
I
Diffusion layer
Zoom
Cint E
l
e
c
M
ΔFint t
e
r
t
o
a
l
l
y
t
e– e
I
Double layer
Figure 8 Schematic representation of the metal–electrolyte interface: concentration (solid line) and potential (dashed line)
profiles.
24 Basic Concepts
the solution conductivity and current density, as in vibrates, or even moves due to thermal agitation.
Figure 8 (ohmic drop effect, see more details in A static object is hence an intrinsically dynamic sys-
Section 1.02.6.1.2). tem at the molecular level and this is not inconsistent
with its macroscopic motionless state. The same
holds for an electrochemical system: reaching and
1.02.3 Electrochemical Equilibrium observing the electrochemical equilibrium absolutely
does not indicate that, besides the thermal agitation,
1.02.3.1 Definition and Introductory
the interface is frozen without any charge transfer
Remarks
going on with time. For example, a copper bar
Electrochemical equilibrium is one of the major con- immersed in a weak acid solution (no dissolved oxy-
cepts that needs to be correctly handled to really gen present and copper oxides are not stable) con-
understand corrosion processes. A proper definition taining a certain amount of copper Cu2þ cations
should take into account the fact that the equilibrium undergoes the forward–backward redox reaction at
of an electrochemical system involves the equilib- equilibrium:
rium of all materials and all interfaces. In a true
Cu !
Cu þ 2e
2þ
½15
equilibrium, no macroscopic gradient of any physical
variable (temperature, pressure, concentration, etc.) In any given time interval the same number n of moles
exists in the bulk of any phase and there is no net of Cu2þ and 2n of electrons is produced by metal
macroscopic flux across any interface. Thus, in the dissolution (corresponding to an anodic current Ia)
framework of the present work, the electrochemical and concomitantly consumed by metal deposition on
equilibrium will be frequently related to an interface the metallic surface (corresponding to a cathodic cur-
state in which no overall current flows to or from rent Ic). Dissolution and deposition are the forward
the interface. and backward elementary steps of the same overall
Before going further with the analysis of electro- reaction [15] and take place with the same rate, thus
chemical equilibrium, two critical remarks must be yielding Ia þ Ic ¼ 0 or |Ia|¼ |Ic| ¼ I0, I0 being called
kept in mind. Firstly, some systems can show a very the exchange current. Provided that the ensemble of
slow evolution constituting a quasisteady behavior physicochemical parameters (temperature, pressure,
that is not a true equilibrium. As seen below, passive etc.) is kept constant, the dynamic zero current bal-
stainless steels can be considered as nonevolving ance between the forward and backward reactions
systems under certain conditions, but they are actu- implies that the system does not evolve with time:
ally far from being at thermodynamic equilibrium. the Cu2þ concentration, as well as the mass of metallic
The knowledge of electrochemical equilibrium char- copper, remains unchanged ad aeternum or, at least, as
acteristics is, however, still indispensable for the com- long as the equilibrium holds (as seen later, the equi-
prehension and the macroscopic global description of librium can be broken as a consequence of the polari-
the quasisteady state of such systems and most corro- zation of the system under the action of a parallel
sion processes. Another major point is that the equi- reaction, such as oxygen reduction that leads to cop-
librium state necessarily means a net zero current, per corrosion: the system then evolves in nonequilib-
but the converse is far from being true. As already rium conditions as for every corrosion process). It can
pointed out in Section 1.02.2.3, for any corroding then be concluded that the exchange current is an
system freely evolving in any aggressive media, the important attribute of the dynamics of the system.
net current flowing to or from the interface is also Indeed, the constant Cu2þ (or Mzþ for a more general
zero, but the system will not be in a state of true Mzþ/M couple) concentration, characteristic of the
equilibrium. electrochemical equilibrium, may result from very
different interfacial behaviors. High values of I0 (for
instance, 101 A cm2 for aqueous systems at room
1.02.3.2 Electrochemical Equilibrium as a
temperature) indicate that the system is very active
Dynamic State: the Exchange
with many reactive interactions per unit time in each
Current Density
direction (anodic and cathodic), whilst low I0 values
Let us consider a static body in equilibrium. No (typically 105 A cm2 for aqueous systems at room
net macroscopic movement is expected (the body temperature) indicate slow systems. This is a key
does not move provided the ensemble of variables concept in reaction kinetics as discussed later in
remains constant), but each constituent particle spins, this chapter.
Electrochemistry 25
absence of change in the reaction free energy, in

Fmetal Fsolution
this case, electrochemical, Dr G ¼ 0, where the sub-
Ia script r stands for any possible reaction. The electro-
Cu Cu2+ Ia : Cu → Cu2+ + 2e–
chemical reaction free energy decomposition in
Ic : Cu ← Cu2+ + 2e– chemical and electric terms as in eqn [17], entails
the important conclusion that the equilibrium of a
Ic
Cu Cu2+ [Cu2+]eq = Constant ⇒ Ia = Ic
redox reaction involving z electrons sets up an inter-
facial potential difference DFint expressed as a func-
tion of the chemical reaction free energy. It can be
demonstrated that, for reaction [18]23:
Cu Interface Solution
ΔFint = Fmetal – Fsolution

M !
M þ ze
zþ
½18
one has
Figure 9 Schematic representation of the potential
difference and current exchanges across the Dr G
metal–electrolyte interface for a Cu2+/Cu equilibrium. Dr G ¼ 0 ) DF ¼ ½19
zF
The change in the chemical free energy can be
mathematically expressed as24,25
The asymmetric distribution of electrical charges
at both sides of the interface, as depicted in Figure 9 aMzþ
Dr G ¼ Dr G 0 þ RT ln ¼ Dr G 0 þ RT ln½Mzþ
(charged cations at one side, electrons in metallic sites ametal
at the other), entails a well-defined potential differ- ½20
ence, DFint ¼ Fmetal Fsolution. The DFint value thus
depends on the thermodynamic characteristics of where the superscript indicates that all reactants
each phase – in particular their composition – and is are in a standard state, and R is the gas constant
related to the equilibrium potential, Eeq, as will be (8.31 J mol1 K1). The symbol a denotes the species
formalized in the next section. activity, which is equal to 1 in the standard state and
is considered as unity in the right-hand side of
eqn [20] for the metal. Besides, as an approximation
1.02.3.3 Formal Expression of Eeq: the that is sufficient for dilute solutions, it is assumed that
Nernst Law for the Mz1 /M Couple the cation activity is equal to its concentration (rep-
resented by square brackets) in moles per liter.26
Classical thermochemistry deals with the transforma- According to eqn [19], the potential difference DFint
tion of a given species at constant temperature and associated with reaction [18] can be expressed as
pressure leading to the concept of Gibbs free energy.
RT
In the case of charged species intervening in electro- DFint ¼ DF0int þ ln½Mzþ ½21
chemistry, an electrochemical Gibbs free energy G zF
can then be defined. As for any electrical charge q, a Equation [21] is however experimentally useless,
part of this electrochemical energy is electric and is since the DFint values are not directly measurable.
related to the work, qF, supplied for moving this Indeed, even if the metallic phase can be connected to
particle inside an electrical field associated with the one of the terminals of a voltmeter with a conducting
potential F. For one mole of the Mzþ cation, the wire, the other terminal of the voltmeter must be
electrical contribution to the total electrochemical connected to a probe dipped in the aqueous phase.
energyG can be written as The potential difference, DFprobe, between the probe
(necessarily conductive to allow potential measure-
qF ¼ zje jN F ¼ zF F ½16
ments) and the aqueous phase must be taken into
zþ
Hence, the electrochemical free energy of M usu- account in addition to the potential difference,
ally expressed as the addition of a chemical term, G, DFMzþ =M , of the metal–solution interface in the pres-
to the electrical term, zF F: ence of the Mzþ/M couple under investigation.
A voltage DFMzþ =M DFprobe is then actually
G ¼ G þ zF F ½17
measured. To be able to give a value to the DFMzþ =M
For any reactive system, it can be demonstrated that term, the solution is to work with a reference system
the thermochemical equilibrium is related to the to which all electrochemical potentials are referred.
26 Basic Concepts
The reference system chosen is the standard hydrogen applied to corrosion studies (where [Mzþ] is given
electrode (SHE) related to the Hþ/H2 equilibrium, in in mol l1) giving the equilibrium potential, Eeq, of
which the reactants are in the standard state. By con- the redox reaction [18]:
vention, a 0 V value is attributed to DFprobe in this case
RT
at all temperatures. The SHE is difficult to realize Eeq ¼ E 0 þ ln ½Mzþ ½23
experimentally, since it requires that the acid solution zF
has a hydrogen ion activity of 1 (which requires a or, in the more convenient decimal logarithmic scale
molarity greater than 1, of about 1.19 mol l1 for (RT ln(10)/F ¼ 0.059 V at 25 C):
HCl), and a hydrogen activity of 1 (which requires a
partial pressure of hydrogen of 1 bar plus an allowance 0:059
Eeq ¼ E 0 þ log ½Mzþ ½24
for the vapor pressure of the solution). A very similar z
electrode (with a very similar potential) is obtained by It is worth noting that the charge transfer process
bubbling pure hydrogen gas at the surface of a Pt physically takes place at the metal–electrolyte inter-
electrode in a normal (molar) acidic solution, which face. This means that [Mzþ] in the Nernst equation
constitutes the so-called normal hydrogen electrode corresponds to the interfacial concentration of Mzþ.
(NHE). Figure 10 illustrates the experimental assem- Since there is no concentration gradient in solution
bly of a NHE and the complete apparatus for the at the electrochemical equilibrium, [Mzþ] is also the
measurement of the Agþ/Ag potential. Mzþ concentration in the solution bulk. However, in
It is clear that the experimental set-up depicted nonequilibrium conditions, necessarily the case when
in Figure 10 is impractical for daily experiments. corrosion occurs, the interfacial concentration, Cint, is
This is why several other very stable redox couples generally different from the bulk concentration, C1,
are used for reference electrodes as presented in and the electrode potential is no longer related to Cint
Section 1.02.6. In every potential measurement, the or C1 in a simple way. Nevertheless, for some redox
potential determined includes both the desired couples, termed Nernstian, which are almost always
metallic couple and the reference one. In such con- fast and reversible in a kinetic sense, a local near-
ditions, the usual notation introduces the electrode equilibrium can be considered at the interface so that
potential E as the electrode potential obeys a Nernst-like equation:
E ¼ DFMzþ =M DFref ½22 0:059
EðI 6¼ 0Þ ¼ E 0 þ log Cint 6¼ Eeq ½25
z
This means that the potential value read on the
The Nernst equation is an extremely rich tool for a
voltmeter in Figure 10 is not the actual potential
first approach of the analysis and comprehension of
value of the metallic couple, it is the potential differ-
corrosion processes, as shown below. Let us first note
ence between two redox couples.
that E0, the equilibrium potential for the couple
Finally, adding DFref at both sides of eqn [21]
Mzþ/M in the well-defined standard state, is a ther-
leads to the usual form of the Nernst equation
modynamic constant that can be easily found in
tables and reference books.27 Table 1 gives some
important examples in the context of corrosion.
Conducting
H2 V wire
Insulating
Table 1 Standard electrochemical equilibrium potential
sheaths
Salt bridge values of redox couples
Couple E0 (VSHE)
Pt electrode Ag electrode Au3+/Au 1.43

Ag+/Ag 0.79
H2 bubbles Cu2+/Cu 0.34
HCl Ag+
H+/H2 0
NHE Ni2+/Ni 0.26
Fe2+/Fe 0.44
Figure 10 Experimental set-up for measuring the Ag+/Ag Zn2+/Zn 0.76
potential difference, DFAgþ =Ag , versus a normal hydrogen Al3+/Al 1.33
electrode.
Electrochemistry 27
A first glance at Table 1 shows that noble metals is independent of the reference employed and is
such as gold and silver for which corrosion is not called overpotential. The above discussion can be
expected, have high positive values (hence called summarized as follows:
noble potentials), whilst those which are promptly
E > Eeq ; > 0 possibility of metal dissolution
recognized as corroding metals (Fe, Zn) have low
ði:e:; corrosionÞ
(less noble) E0 values. It then clearly appears that E0
E < Eeq ; < 0 possibility of electrodeposition
gives relevant information on the tendency of each
metal to corrode. An apparent contradiction appears
in the case of Al, which nonexpert common experi- 1.02.3.4 Some More Complex Applications
ence would describe as corrosion resistant. In reality, of the Nernst Law
from a thermodynamic point of view, aluminum is
actually extremely reactive and easily transformed Up to now, only the archetypical corrosion redox cou-
into Al3þ. However, this reaction leads to the forma- ple Mzþ/M has been dealt with. Nevertheless, it is
tion of a very homogeneous and corrosion-resistant important to consider the more general approach of
film on the metallic surface, in a self-protective pro- the Nernst law used for other redox couples. Some
cess called passivation, largely discussed elsewhere in important examples concerning acid–base properties
this book. For the purpose of this chapter, it is impor- will be treated in Section 1.02.3.5. As a first example,
tant to retain this false paradox as the consequence of let us consider here a solution containing complexing
the fact that the Nernst equation is the result of a ions, for instance, an aqueous electrolyte with a high
thermodynamic treatment concerning only the first concentration of chloride ions (Cl) in contact with an
formation of Al3þ and not the solid protective oxide iron plate, thus undergoing corrosion. In that case, the
formation, though thermodynamically predicted, as equilibrium potential of the complex couple FeClþ/Fe:
shown in the next section. Fe þ Cl ! þ
FeCl þ 2e

½27
It is also easy to see from the Nernst eqn [23], that
changes in cation concentration yield changes in the can be written according to the Nernst law:
equilibrium potential. The equilibrium potential of
0:059 ½FeClþ
any metal in contact with a solution containing the Eeq ¼ E 0 þ log ½28
corresponding cation can be easily calculated for a 2 ½Cl
given cation concentration. As an example, the equi- Another application of the Nernst law concerns
librium potential of Zn with [Zn2þ] ¼ 102 mol l1 is reactions with adsorbed species. As explained in the
Eeq ¼ 0.76 þ 0.0295 (2) ¼ 0.82 VSHE accord- preceding sections, even if the balance of reactive phe-
ing to eqn [24] and Table 1. If the potential is nomena could be written according to the overall
shifted, for instance, to 0.7 VSHE, the Nernst equa- Mzþ/M redox reaction, intermediate species are fre-
tion predicts that Zn is no longer at equilibrium with quently adsorbed on the metallic surface and play an
[Zn2þ] ¼ 102 mol l1, but with [Zn2þ] ¼ 102 mol l1, important role in corrosion processes. Without going
that is a 10 000-fold increase in concentration. This through the mathematical details of the exhaustive
means that Zn will corrode, liberating Zn2þ cations to theoretical treatment of the interface, a very brief ther-
the solution until the cation concentration predicted modynamic description of the simplest adsorption
by the Nernst equation is reached and a new equilib- models can be presented. In the case of ideal systems,
rium is established. It is then clear that during a those in which no interaction between adsorbates
certain time needed for electrolyte homogenization occurs, the activity of an adsorbed species is given by
at this new Zn2þ concentration, a positive (anodic) its fractional covering ratio y, defined as the ratio
current appears: the system is transiently out of equilib- between the number of adsorbed species and the max-
rium. Alternatively, if the new potential value is lower imal number of adsorbed species when the surface
than that of the initial equilibrium, it is easy to deduce is completely covered. In the previous example of Ni
that a certain amount of cations have to be reduced and (or Fe) corrosion illustrated by reaction [10], the
deposited on the metallic Zn surface to diminish the Nernst equation can be derived from the reaction
cation concentration in solution. The value of the explicitly indicating the free metallic sites available
potential shift from the equilibrium potential for adsorption:
¼ E Eeq ½26 Ni þ H2 O þ s ! þ
NiOHads þ H þ e

½29
28 Basic Concepts
Since there is a single adsorbed species, the number and potential (highly oxidizing atmospheres, rich in
of free sites is proportional to 1 – y so that the oxygen or nitrates, or reducing atmospheres as in
Nernst law takes the following form (remember that deaerated closed water circuits).
pH ¼ log [Hþ]): In this section, the main principles of Pourbaix
y diagrams and rough guidelines for their calculation
Eeq ¼ E 0 0:059pH þ 0:059 log ½30 are introduced. With the help of three typical and
1y
important cases (gold, iron, and aluminum), the large
Equation [30] is the mathematical representation of amount of information contained in those diagrams is
the so-called Langmuir isotherm, often represented as shown. Their key objective is, for a given metal–
y K electrolyte system at known constant temperature
¼ ½31
1 y ½Hþ and pressure, the establishment of domains of ther-
modynamic stability or predominance of any poten-
where K is a potential-dependent thermodynamic tial species for the entire range of pH and electrode
constant. potential. As a result, for practical purposes, the
In contrast with the assumption of the Langmuir- whole inventory of service-life conditions of an inter-
type isotherm, it is often necessary to take into account face can be represented. It then allows the global
the possibility of interaction – attraction, repulsion – thermodynamic predictions of the behavior of a
between adsorbates, mainly for high covering rates y. given metal–electrolyte system (corrosion, immunity,
This is achieved by introducing activity coefficients in oxide formation, etc.) to be accessed readily.
the Nernst law (eqn [31]). The most usual model While more complex diagrams, including the
dealing with these nonideal electrochemical systems formation of complexes in corrosive media, can of
is the Frumkin isotherm that gathers the nonideal course be established, only pure metal–water systems
attributes in the form of an exponential function of will be considered in this chapter. The starting point
the covering rate: of Pourbaix diagram is based on the fact that there
y K are three types of electrochemical equilibrium for
exp ðgyÞ ¼ þ ½32
1y ½H a metal–water system, those depending only on the
electrochemical potential, those depending only on
where g is a constant. g > 0 and g < 0 correspond, the pH, and those depending on both potential and
respectively, to repulsive and attractive interactions pH. The general guidelines of the diagram construc-
between the adsorbates. With g ¼ 0, no interaction is tion are given below for a simple example of a metal
expected and the Langmuir isotherm is retrieved. M that can dissolve into cations Mzþ or form a solid
oxide M2Oz.
1.02.3.5 Applications of the Nernst Law to
1.02.3.5.1 Purely potential-dependent
Pourbaix (E–pH) Diagrams
equilibrium
An important application of the Nernst law for cor- This is the case of the simple redox reaction [18]
rosion applications is the potential–pH diagram, also between the pure metal and its cation for which the
known as the Pourbaix diagram in honor of Marcel equilibrium is described by the Nernst equation [24],
Pourbaix’s exhaustive work devoted to the calcula- as discussed above. For a given Mzþ concentration
tion, interpretation, and diffusion of those diagrams.28 (different Pourbaix diagrams have to be drawn if this
The importance and high usefulness of Pourbaix value is changed), Eeq is independent of pH and is
diagrams come from the fact that corrosion is not represented in an E/pH diagram as a horizontal line,
simply a matter of redox reactions like the basic reac- as shown in Figure 11(a). Hence, regardless of the
tion [2], but very frequently involves species with electrolyte pH, the metal M is in thermodynamic
acid–base properties, such as water itself (remember equilibrium with its cation at the given Mzþ concen-
that most corrosion phenomena are in fact related to tration only on this line. For E < Eeq, Mzþ cations
aqueous corrosion). Also important for its wide appli- reduce to metal, while for E > Eeq the metal under-
cability, is the very wide range of in-service condi- goes dissolution to form cations in solution. This indi-
tions of metals that may be susceptible to corrosion. cates that, for this kind of equilibrium and a given Mzþ
Indeed, corrosion occurs in widely differing environ- concentration, the thermodynamic domain of metallic
ments in terms of both pH (from acidic environments stability is the whole region below the horizontal line,
to alkali industry, or very basic mortars and cements) whilst the domain of cation predominance is the upper
Electrochemistry 29
region. It is important to appreciate that the position of E E 2

the line is dependent on the Mzþ concentration con-
sidered: a modification in the concentration would
therefore change the stability or predominance domain Mz+ Mz+ M2Oz
of the Mzþ species.
1.02.3.5.2 Purely pH-dependent equilibrium

1 1
When dissolved cations precipitate in the form of M M ?
solid compounds in which the metal keeps the same
oxidation number as the cation, for instance, for the (a) pH (b) pH
Mzþ/M2Oz equilibrium, the corresponding forward
E 2 E 2
and backward reactions: O2
2Mzþ þ zH2 O !
M2 Oz þ 2zH
þ
½33 b
Mz+ M2Oz Mz+ M2Oz
convey no charge transfer. Hence, no redox reaction
a H2O
is present (each reactant keeps its valence on both
H2
sides of the reaction) and the equilibrium is indepen-
dent of the potential, but it is a function of the pH 1 3 1 3
M M
because of the protons on the right-hand side of the
equation. The corresponding line of the Pourbaix (c) pH (d) pH
diagram can be determined with the help of the Figure 11 Building-up a generic Pourbaix diagram: (a) the
classical chemical equilibrium theory by introducing metal/cation equilibrium, line 1; (b) the cation–metal oxide
the equilibrium constant K of reaction [33]29: equilibrium, line 2; (c) the metal–metal oxide equilibrium,
line 3; (d) the complete schematic Pourbaix diagram,
½Hþ 2z including the water stability domain, lines a and b.
K¼ ) log K ¼ 2z log½Hþ 2log½Mzþ
½Mzþ 2
pK M2Oz/M equilibrium. It is also clear that reaction
pK ¼ log K ) log ½Mzþ ¼ zpH ½34
2 requires a source of oxygen, which leads to the very
It can be seen from reaction [33] that for low pH values important observation concerning corrosion that
(high proton concentration), the reaction equilibrium oxygen comes mainly from the dissociation of water
is shifted to the left side, the oxide combining with molecules and not from the atmosphere, contrary
protons and producing cations. Low pH values hence to an extremely widespread idea. The overall reac-
entail high dissolved cation concentration. As a conse- tion is then
quence, for a fixed Mzþ concentration, the Mzþ/M2Oz 2M þ zH2 O ! þ
M2 Oz þ 2zH þ 2ze

½35
equilibrium is characterized by a vertical line in the
E/pH diagram (pH ¼ constant, line 2 in Figure 11(b)), for which the equilibrium potential can be written
the left side being the predominance domain of the Eeq ¼ A1 0:059 pH ½36
cations and the right side the oxide stability region. As
shown in Figure 11(b), lines 1 and 2 must in reality be where A1 is a constant. The equilibrium, described by
mutually truncated at their intersection to take into a linear relationship between the pH and the poten-
account the existence of the metal oxide. One should tial is represented by the straight line 3 of slope
then look for the equilibrium between M and M2Oz, the 0.059 V pH1 in Figure 11(c). This line, which
third kind of equilibrium, which is dependent on the pH separates M and M2Oz, must start at the intersection
and on the potential simultaneously, to complete the of lines 1 and 2 and must be plotted towards lower
Pourbaix diagram. potentials and more alkaline pH according to the
negative slope of line 3. Indeed, in the other direction
1.02.3.5.3 pH and potential-dependent the line would be in the Mzþ domain where neither
equilibrium M nor M2Oz can exist at a steady state.
As indicated above, the oxidized M species takes the Figure 11(c) depicts the stability domains of the
form of a solid oxide M2Oz for high pH values (Mzþ three species considered. Since this is a metal–water
can actually exist but its concentration is low). It is system, the diagram must also incorporate the water
then intuitively more appropriate to consider the stability domain, according to the equilibria H2O/H2
30 Basic Concepts
and O2/H2O, characterized at 25 C by the following 2.8

equations: AuO2
2.4
Eeq ¼ 0:00 0:059pH for H2 O=H2
2.0
Eeq ¼ 1:23 0:059pH for O2 =H2 O ½37 Au3+
1.6 Au(OH)3
The limits of the water stability domain are thus repre- HAuO32–
1.2
sented by two parallel lines, identified in Figure 11(d) b
0.8
E (VSHE)
as a (for the H2O/H2 equilibrium) and b (for the
O2/H2O one), indicating that water is thermodynami- 0.4
cally stable between these two lines: below line a a
0.0 Au
hydrogen evolves and above line b, oxygen evolves, as
depicted in Figure 11(d), which is the final form of the –0.4
schematic Pourbaix diagram for the M–H2O system –0.8
involving the three species, M, Mzþ, and M2Oz.
–1.2
Figure 11(d) respects the usual graphic notation
used for the sake of readability of the diagrams. –1.6
Dashed lines indicate an equilibrium between two –2 0 2 4 6 8 10 12 14 16
dissolved species (not illustrated in this simple exam- pH
ple), such as two cations (Fe3þ/Fe2þ, for instance). Figure 12 Simplified Pourbaix diagram for the
Thin solid lines indicate an equilibrium between a Au–H2O system at 25 C. Concentration of dissolved
dissolved and a solid species (lines 1 and 2), and thick species taken as [X] ¼ 106 mol l1.
solid lines indicate an equilibrium between two solid
species (line 3).
Even if schematic, Figure 11(d) illustrates a com- of Au3þ, Au(OH)3, AuO2, and HAuO2 3 stability
mon feature of metal–water Pourbaix diagrams, that is, domains indicate, at first glance, that gold can
metal immunity at more negative potential values, film- undergo corrosion at very oxidizing conditions (very
free metal dissolution at more positive potentials in high potential values). A deeper look, however, shows
acids and oxide formation in the neutral and alkaline that the gold immunity domain comprises and over-
pH ranges (though many metals will also corrode as hangs that of water. This means that gold is always
oxy-anions in alkali). (The terms ‘more positive’ and thermodynamically stable whenever stable water is
‘more negative’ are used here in place of ‘higher’, ‘lower’ present, which obviously includes all natural environ-
and similar terms when referring to electrochemical ments on earth and gives a sound idea of the condi-
potentials to avoid any possible confusion. Because tions to be achieved for gold corrosion. This is also
metal potentials are often negative, there is an unfortu- why gold is always found in its metallic form and not
nate tendency (particularly in the field of cathodic in mineral compounds, unlike most of the metals.
protection, where potentials are always negative) to In the case of iron, certainly the most common
say, for example, that 1.2 V is ‘higher’ than 1 V constructional metal, the Pourbaix diagram at 25 C
(and in some cases even to omit the negative sign).) and 1 atm pressure (Figure 13) shows a very different
behavior that explains why corrosion has always been
a matter of much concern. In contrast with gold, the
1.02.3.6 Pourbaix Diagram of Gold, Iron,
existence of metallic iron is thermodynamically
and Aluminum
impossible in the whole water stability domain: cor-
Let us now analyze some typical examples of Pourbaix rosion is hence a spontaneous process bringing iron
diagrams to illustrate their usefulness. (For the sake of back to its stable form in nature. Thus, iron is found
clarity, these are simplified diagrams that consider in different mineral forms such as hematite (Fe2O3)
only some of the possible dissolved or solid species. and magnetite (Fe3O4) and the industrial transforma-
The reader is invited to refer to the Pourbaix’s Atlas tion into metallic iron requires huge amounts of
of electrochemical equilibrium for more detailed energy (supplied by the oxidation of carbon inside
pictures.28) Figure 12 shows the Au–H2O diagram furnaces during the production of iron and steel) to
at 25 C and 1 atm. The first remarkable feature is break the oxygen–iron bonds. This spontaneous
that gold has a wide immunity domain occupying reverse route (iron towards oxide) is clearly demon-
a large extent of the E/pH space. The existence strated by the fact that the Pourbaix diagram in
Electrochemistry 31
are the hydrogen evolution (in acidic media) and the

1.6 –2 –6
reduction of dissolved oxygen (in neutral or alkaline
electrolytes in contact with air),
1.2 b
Fe3+
0.8 2Hþ þ 2e !H2
O2 þ 2H2 O þ 4e ! 4OH ½38
0.4 Fe2O3
E (VSHE)
a Note that both cathodic reactions yield an alkaliniza-

Fe2+
0.0 tion of the interface, the first by direct consumption of
Hþ and the second by production of OH. As a conse-
–0.4 –2 quence, the system can locally shift from a given point
Fe3O4
–0.8 –6 inside the Fe2þ domain to, for instance, the hematite
domain, which gives a precipitate in the form of what is
–1.2 Fe generically called rust. Only ferrous species can exist
at the interface, as ferric ions are stable only at much
–1.6 more positive potentials than normal corrosion poten-
–2 0 2 4 6 8 10 12 14 16
pH tials. However, in aerated solutions the ferrous ions can
be oxidized in solution to ferric by dissolved oxygen,
Figure 13 Simplified Pourbaix diagram for the
Fe–H2O system at 25 C. Concentration of dissolved
and the ferric ions can then precipitate, usually as
species taken as [X] ¼ 102 or 106 mol l1. an oxyhydroxide (a, b, g, or d-FeOOH) which are,
roughly speaking, the different hydrated forms of
hematite (Fe2O3H2O ¼ 2 FeOOH). (The hematite
Figure 13 predicts iron transformation into magne- is the thermodynamically stable species, the other
tite (less oxidized species) at more negative potentials forms are metastable but their transformation is so
or hematite at more positive potentials. slow that they can be identified in corrosion products
Figure 13 also shows a large active dissolution or even in minerals.) Note that the precipitation cannot
region (the predominance domain of the ferrous cat- occur at the metal surface because the ferric species are
ion Fe2þ, followed by that of the ferric cation Fe3þ for not stable there, so the corrosion product is necessarily
potential values more positive than E ¼ 0.77 VSHE), porous and nonprotective.
often referred to as the corrosion triangle, lying in The last illustrative example concerns aluminum,
the neutral–acid pH range, illustrating and justifying the Pourbaix diagram of which in Figure 14 shows a
the high sensitivity of iron to corrosion in acidic wide oxide domain and an extremely narrow immu-
environments. nity one, well below the domain of water stability.
As mentioned in Section 1.02.3.5.1, the stability Aluminum is thus a particularly reactive metal, as
domain of the species depends on the concentration already deduced from the very negative standard
of the dissolved species considered to be at equilibrium equilibrium potential (Table 1). It is important to
with the solid compounds. The concentrations used note that its high corrosion resistance does not
in building the diagram in Figure 13 (102 and 106 come from a marked immunity and stability, as for
mol l1), for example, for the Fe2þ/Fe equilibrium, gold or other noble metals, but from the formation of
indicate the range of concentrations expected when a homogeneous and impermeable oxide layer in a
iron corrosion is taking place. In other words, beyond very large domain of potential and pH, as shown in
these values, the stable species is no longer metallic Figure 14. This very adherent oxide layer protects
iron, but the ferrous cation, thus leading to corrosion. the metal from further corrosion. This process, called
The line separating the metallic iron and ferrous cation passivation, is also at the origin of the high corrosion
domains move up when the concentration threshold resistance of several metals and alloys such as tita-
for corrosion is increased from 106 to 102 mol l1, nium and stainless steels. In the latter case, passiv-
leaving a larger iron stability domain below it. ation of chromium, one of the main alloying elements
As already discussed, overall corrosion processes in those steels, provides the corrosion protection.
at free potential are the result of coupled anodic and The difference between the corrosion resistance of
cathodic reactions taking place simultaneously on the iron and aluminum in practical life illustrates an
metallic surface. The most common cathodic reac- important aspect that must be taken into account
tions occurring during iron dissolution (reaction [2]) when using Pourbaix diagrams. In fact, both diagrams
32 Basic Concepts
when an overpotential was imposed to the interface

1.2 b –2 –6 –6 –2
(see Section 1.02.3.3 with the Zn2þ/Zn couple as
0.8 illustration), the discussion concerned the descrip-
tion of the new equilibrium state with a new zero
0.4
a Al2O3.H2O overall current towards which the system was going.
0.0 Let us now consider that a constant overpotential
AlO–2 is applied to the interface by means of an external
E (VSHE)
–0.4
device so that the system is kept out of equilibrium.
–0.8 Al3+
This idea can be easily handled with the same exam-
–1.2 ple of Zn in contact with a solution of Zn2þ concen-
tration C1 ¼ 102 mol l1 (Eeq ¼ 0.82 VSHE). If this
–1.6 –2
redox couple is supposed to be Nernstian (cf. Section
–6
–2.0 1.02.3.3), the overpotential at the interface is linked
–2.4 Al –2 to the interfacial concentration, Cint, as (see eqns [24]
–6 and [25]):
–2 0 2 4 6 8 10 12 14 16
pH
ðI 6¼ 0Þ ¼ EðI 6¼ 0Þ Eeq
Figure 14 Simplified Pourbaix diagram for the
Al–H2O system at 25 C. Concentration of dissolved species 0:059 0:059
¼E þ 0
log Cint E þ
0
log C1
taken as [X] = 102 or 106 mol l1. The hatched region 2 2
identifies the large passivation domain of aluminum.
0:059 Cint ½39
¼ log
predict the existence of oxide stability domains, but, 2 C1
in contrast with aluminum, iron is not protected by
these oxides, which are usually nonadherent, friable, When an overvoltage Z of 0.12 V (E ¼ 0.7 VSHE),
and easily hydrated as mentioned before. Some excep- for instance, is applied to the interface, eqn [39]
tions are nonetheless of great technological impor- points out that Zn must corrode to supply Zn2þ
tance. In specific conditions, such as highly alkaline cations in such amount that the interfacial concentra-
environments, a thin and highly protective oxide is tion is 10 000 times higher than the bulk concentra-
formed on steel, which explains, for instance, the very tion. This new equilibrium will be reached after
low corrosion rate (at least initially) of reinforcing a very long time for typical volumes of electrolyte.
bars of low carbon steel embedded in concrete. In the transient period, a nonzero current balance
It is important to stress that Pourbaix diagrams are (I > 0 corresponding to a net oxidation) is established
based on thermodynamic equilibria. The exchange and, since the interfacial concentration is high, the
current density is outside the scope of this approach, diffusion flux of the cations is directed from the
which gives no indication about electrode kinetics. metal to the electrolyte. In the other direction, < 0
Furthermore, corrosion is not an equilibrium process, entails metallic deposition (I < 0 corresponding to a
and the anodic and cathodic reactions are completely net reduction) so that the interfacial concentration is
different in nature, as explained above. In spite of lower than the bulk concentration. The cation flux
these restrictions, Pourbaix diagrams constitute a towards the interface can last a long time since the
very useful tool to predict if the metal is protected electrolyte reservoir is hardly changed.
by immunity or passivity when the solution pH and In this system, the potentiostatic out-of-equilibrium
the open-circuit potential are known. conditions involve a transient behavior. In particular,
the absolute value of the current decreases as a result of
the thickening of the diffusion layer. Forced convection
1.02.4 Electrochemical Kinetics in is very often used to control the interfacial concentra-
the Case of a Single Forward– tion of the cations. As explained in Section 1.02.2.4.3,
Backward Reaction convection allows the thickness of the diffusion layer to
be fixed and the bulk concentration to be maintained
1.02.4.1 Qualitative Approach
constant, provided that large volumes of electrolyte are
Up to now in this chapter, only reactions at equilib- used. In that case, the transient behavior is considerably
rium, for which zero overall current characterizes the accelerated and a steady state, different from that at
electrochemical system, have been discussed. Even equilibrium, is rapidly achieved. (In practice, a steady
Electrochemistry 33
Net oxidation
Current I
Current I
0
Eeq, I = 0
Eeq, I = 0
Net reduction 0
0
Overpotential h 0
Overpotential h
Figure 15 Schematic drawing of the
current–overpotential curve involving the cathodic (net Figure 16 Schematic representation of a
reduction) and the anodic (net oxidation) domains. current–overpotential curve with a mass-transport
controlled plateau in the cathodic domain.
state is impossible to obtain but the evolution of the

system is so slow that it can be considered in a steady electrode potentials in this region, the corresponding
state during the experiment.) The value of the net cation concentration at the interface is essentially
steady-state current, at each potential or overpotential zero, as discussed in the following sections.
then provides important information about the kinetics
of the electrochemical reaction. The ensemble of (E, I) 1.02.4.2 Quantitative Approach
or (, I) couples is often plotted in the form of a The qualitative description of Figure 16 shows that
current–potential or current–overpotential curve, as the actual current flow is the final result of combined
illustrated in Figure 15. This curve is called a polari- kinetic phenomena, charge transfer of the redox reac-
zation curve and its slope near the equilibrium depends tion and mass transport of the cations. The following
very often on the exchange current density of the redox sections discuss in a more detailed and quantitative
reaction kinetics, as shown in Section 1.02.4.2. way the two limiting cases obtained when the kinetics
Since the product I is always positive, as illu- are controlled by a single phenomenon and finally
strated in the Zn dissolution example given above, describe the general case of mixed control. Let us first
the current–potential curve increases monotonically restate, however, that the following development is
around the equilibrium: the higher the overpotential, only valid for a single forward–backward redox reac-
the higher the net current (both in absolute values). tion with two reacting species present at the interface,
It is important to notice that nonmonotonic curves as in the case of the elementary redox reaction:
also exist, mainly in the corrosion domain, as in the
Red k! ½40
ka
case of stainless steels or passivating metals. Such a Ox þ ze
c
feature is, however, closely linked to systems in
such as, for example, the widely employed redox
far-from-equilibrium conditions.
couple Fe3+/Fe2+ taking place at an inert Pt elec-
The property that the polarization curve in
trode. The current–potential relationships are
Figure 15 is monotonically increasing requires two
described below for steady-state conditions and the
conditions: first, the metal is always available at the
electrolyte is assumed to be ideal: the activities of the
interface so that, whatever the overpotential in the
species are equal to their corresponding concentra-
anodic part, the metal can keep supplying cations.
tions, expressed in mol l1.
Second, in the cathodic region, the cations must be
available to ensure the increase in |I| with increasing 1.02.4.2.1 Activation-controlled systems:
||. However, beyond a certain overpotential nega- Butler–Volmer equation
tive value, the amount of cations arriving at the inter- In the activation limited case, the interfacial kinetics
face per unit time is limited by mass transport. The are the controlling factor. The influence of mass trans-
consequence is that the current reaches a diffusion port in the electrolyte, in particular diffusion, is negli-
limiting plateau, as illustrated in Figure 16. For all gible because it is much faster than the charge-transfer
34 Basic Concepts
process. Therefore, the concentrations are identical at

ð1 aÞzF
the interface and in the electrolyte bulk: kc ¼ k 0 exp ðE E 0 Þ ½46
RT
½Oxint ½Ox1 and ½Redint ½Red1 ½41
(Hereafter the usual mathematical treatment that admits
Any chemical reaction evolves according to an intrinsic global multielectron reactions (z > 1) is preferred. It
speed called reaction rate. Obviously, this is also valid must, however, be kept in mind that, as mentioned in
in the case of electrochemical reactions. As already Section 1.02.2.3, a rigorous analysis of the reaction
mentioned, the zero overall current, characterizes the mechanisms encompasses the existence of intermediate
electrochemical equilibrium, at which the forward and single electron steps with z ¼ 1.) The a parameter in
backward reaction currents are equal, irrespective of the exponentials, which is called charge-transfer coeffi-
the exchange current density. cient (0 a 1), indicates how the electrical energy
The knowledge of the mechanism of the overall related to a potential shift from the standard state,
reaction allows the rate law of each elementary step zF(E – E0) is shared between the anodic and cathodic
to be derived (see Section 1.02.2.4.1). In the case reactions, as illustrated below in this section.
considered here, the stoichiometry being very simple The overall current density related to this electro-
for the forward and backward heterogeneous reac- chemical reaction can be written as the algebraic sum
tions, their rate laws can be easily written in terms of of the anodic and cathodic currents:
the cathodic and anodic fluxes, fcathodic and fanodic,
j ¼ ja þ jc ¼ zF ðka ½Red1 kc ½Ox1 Þ
expressed in moles produced or consumed per unit
time and unit surface (mol cm2 s1): azF
¼ zFk ½Red1 exp
0
ðE E Þ
0
RT
fcathodic ¼ kc ½Oxint kc ½Ox1 ½42
ð1 aÞzF
½Ox1 exp ðE E 0 Þ ½47
fanodic ¼ ka ½Redint ka ½Red1 ½43 RT
1
where both rate constants k are expressed in m s , or It must be noticed that this equation remains valid
more usually in cm s1. (From a rigorous point of even if there is only one species in the solution. For
view, eqn [42] should also involve the electron con- example, if there is no Ox species introduced in the
centration at the interface. Nevertheless, since elec- electrolyte, the cathodic terms in eqn [47] vanish and
trons are fully available in a metallic surface, their the expression of the current density is reduced to
concentration is considered as being constant and is the anodic branch. As a consequence, there is no
implicitly incorporated in the constant kc.) (It should equilibrium or zero-current polarization point sim-
be noted that a rate constant does not necessarily ply because there is no cathodic reaction counterpart:
have units of meters per second, in particular when the steady-state current density tends to zero only at
the order of the reaction is not 1 or when k depends very high negative overpotentials. Steady-states in
on the concentration of adsorbates in moles per which only one species is present are very particular.
square centimeters.) Each electrochemical reaction Indeed, the development of the oxidation reaction
rate depends on the potential, according to the itself produces Ox species that progressively accumu-
Arrhenius-type eqns [45] and [46] presented below.4 late at the interface and can in turn be reduced, hence
In the standard state, which is a particular case of creating a redox process.
electrochemical equilibrium, as explained before, the Let us now come back to the previous example in
fluxes of the anodic and cathodic reactions are iden- which both Ox and Red species are initially present
tical with the same concentrations of the redox spe- in the electrolyte. Instead of working with the poten-
cies so that the reaction rates are also identical and tial difference relative to the standard state, E – E0, as
denoted k0: in eqn [47], it is more interesting to refer to the
equilibrium potential and deal straightforwardly
kc ¼ ka ¼ k 0 ½44 with the overpotential = E Eeq to write:
The following expressions of electrochemical rate
zF zF zF
constants are derived: ðE E 0 Þ ¼ ðE Eeq Þ þ ðEeq E 0 Þ
RT RT RT

azF zF zF
ka ¼ k exp
0
ðE E Þ
0
½45 ¼ þ ðEeq E 0 Þ ½48
RT RT RT
Electrochemistry 35
According to the Nernst equation (similar to eqn

4
[23]), Eeq can be written as ja
½Ox1
Current density (mA cm–2)

RT
Eeq ¼ E 0 þ ln ½49 2
zF ½Red1
j
so that j0
0
zF zF ½Ox1
ðE E 0 Þ ¼ þ ln ½50
RT RT ½Red1
–2
By combining eqns [48] and [50], the current density jc
can be expressed as:
–4

½Oxa1 azF
j ¼ zFk ½Red1
0
exp –0.10 –0.05 0.00 0.05 0.10
½Reda1 RT Overpotential h (V)

½Ox1þa
Figure 17 Current–overpotential curve of a purely
1 ð1 aÞzF
½Ox1 exp charge-transfer controlled system illustrating ja and jc
½Red1þa
1
RT (eqns [54] and [55]) in the upper and lower halves,
respectively (dashed lines) and the overall Butler–Volmer
a 1a azF equation [52] (solid line): a ¼ 0.5, z ¼ 1 and j0 ¼ 1 mA cm2.
¼ zFk ½Ox1 ½Red1 exp
0

RT

ð1 aÞzF
exp ½51 As already pointed out, the upper curve in
RT
Figure 17 is obtained when only Red species are
and finally, the so-called Butler–Volmer equation is present in the solution (eqn [54]). In the same way,
derived: the lower curve is obtained when only Ox species are
present in the solution (eqn [55]). The Butler–
azF ð1 aÞzF
j ¼ j0 exp exp ½52 Volmer curve, obtained when the two species are
RT RT present in the electrolyte, appears then as the addi-
where the exchange current density j0 is tion of the two curves at each overpotential. Any
current value depicted in the curve is the net result
j0 ¼ zFk0 ½Oxa1 ½Red1a
1 ½53 of a dynamic balance between the anodic and
As discussed above, at the equilibrium ( ¼ 0), the cathodic reactions, giving for example a zero current
current density is equal to 0 so that ja ¼ |jc|¼ j0. at equilibrium. In other words, except for very high
In eqn [53], the idea of exchange current density overpotentials (see Section 1.02.4.2.2), the cathodic
expressed as a function of two kinetic parameters, k0 (anodic) reaction, although markedly slowed down, is
and a, of the redox system investigated is retrieved. still taking place on the metallic surface even at
The two terms in the right-hand side of eqn [52] positive (negative) overpotentials, which is an impor-
stand for the purely anodic and cathodic branches of tant feature of corrosion processes, as seen hereafter.
the current–overpotential curve, respectively: The Butler–Volmer equation illustrates an impor-
tant difference between purely chemical reactions
azF and electrochemical reactions. In the first case, for
ja ¼ j0 exp ½54
RT given concentrations of reactants and hydrodynamic
conditions, the only way of modifying the reaction
ð1 aÞzF rate is to change the electrolyte temperature. In the
jc ¼ j0 exp ½55
RT case of electrochemical reactions, however, besides
the temperature (explicitly present in the argument
Figure 17 presents the anodic and cathodic of the exponentials but also implicitly present in the
branches (eqns [54] and [55]) and the sum of both k0 value), the electrode potential also strongly affects
currents at each potential related to the overall But- the reaction rate and, therefore, the resultant current.
ler–Volmer equation [52], which fully describes the As an example, for a single electron reaction with
kinetic behavior of an electrochemical interface in a ¼ 0.5 and T ¼ 25 C (298 K), a simple calculation
the case of pure activation control. from the Butler–Volmer equation indicates that an
36 Basic Concepts
overpotential of only 0.5 V entails a exp(aF/RT) ¼ longer negligible and the Butler–Volmer equation no
17 000-fold increase of the anodic reaction, while the longer valid (see Sections 1.02.4.2.1 and 1.02.4.2.2).
cathodic one is depressed by the same magnitude. On the contrary, for small j0, that is, for slow redox
This example can be used to better discuss the couples, high overpotentials are needed to get a sig-
physical meaning of the charge transfer coefficient as nificant net current, as shown in Figure 19. The
the kinetic parameter. This coefficient conveys infor- system is then said to be irreversible, not in the
mation on how the electric energy supplied to the thermodynamic sense, but to express that it is prone
electrochemical system is shared between the anodic to be polarized in one of the parts of the current–
and cathodic reactions. The value a ¼ 0.5 indicates overpotential curve where only one of the forward
that both reactions are boosted or hindered by the and backward reactions completely governs the over-
same factor. The arguments of both exponentials are all reaction. In that case, once a given species has
identical (in absolute values) for any overpotential been oxidized (reduced), the overvoltage is too large
so that the curves are symmetric in Figure 17, for it to be reduced (oxidized) back, so that the
but this is no longer the case in Figure 18 with species has much more chances to move away from
a ¼ 0.8: the anodic branch increases more steeply the interface than to react at the interface.
than the cathodic one. This slight bias relative to a The following paragraphs deal with two approx-
symmetric behavior obtained for a ¼ 0.5 illustrates imations of the Butler–Volmer equation depending
the fact that the response of the forward and back- on the overpotential value. They are valid for slow
ward reactions to a same electric energy shift (|zF|) redox couples and allow the kinetic parameters k0
is different from the mutual mirrored kinetic behav- and/or a to be determined.
ior, depending on the a value.
It is also worth stressing that the final value of the
Small overpotential limit of the Butler–Volmer
current density at any overpotential is directly
equation
dependent on the exchange current density, as illu-
As highlighted in Figure 20, in the region close to the
strated in Figure 19.
equilibrium potential (typically || < 10 mV), the
When j0 is large, the redox couple is said to be fast
current–potential curve can be considered as linear.
and can be considered as reversible, at least for
Indeed, by using the Taylor expansion at the first
applied overpotentials not too large. High j0 values
order
P1 (small Z) of the exponentials ðexpðxÞ ¼
mean that both cathodic and anodic components
k¼0 ðx k
=k!Þ 1 þ xÞ in the Butler–Volmer equa-
strongly contribute to the final net j value and small
tion, the current density is shown to be proportional
applied overpotentials are enough to produce a sig-
to the overvoltage:
nificant current in either direction. However, since
high current values mean a high flux of reactants at zF
j ¼ j0 ½56
the interface, the mass-transport kinetics are no RT
0.4
ja
Current density (A cm–2)
2
j 0.2
j0
0 0.0
–0.2
jc
–2
–0.4
–0.2 –0.1 0.0 0.1 0.2 –0.4 –0.2 0.0 0.2 0.4
Overpotential h (V) Overpotential h (V)
Figure 18 Current–overpotential curve as in Figure 17 Figure 19 Effect of the exchange current density on the
but with a biased charge-transfer coefficient a ¼ 0.8, shape of the current–overpotential curves for a ¼ 0.5, z ¼ 1,
z ¼ 1, and j0 ¼ 1 mA cm2. and j0 ¼ 1 mA cm2 (▪), 10 mA cm2 (), 100 mA cm2 (▲).
Electrochemistry 37
4
Consequently, the following equations, similar to
eqns [54] and [55], can be readily derived:
ja
azF
2 j ja ¼ j0 exp
RT
j
for positive large overpotential ½58
0

ð1 aÞzF
j jc ¼ j0 exp
–2 RT
jc for negative large overpotential ½59
–4 The system is said to be in the irreversible domain
–0.05 0.00 0.05 where one of the reactions is substantially hindered
Overpotential η (V) and can be neglected. In that case, the global current–
Figure 20 Small-overpotential limiting case: the linear overpotential curve tends asymptotically to the
behavior of the Butler–Volmer equation close to the purely anodic curve or the purely cathodic curve
equilibrium potential (a ¼ 0.5, z ¼ 1, and j0 ¼ 1 mA cm2). in Figures 17 and 18. This happens when the ratio
|jc/ja| ¼ exp(zF/RT) for > 0 or its reciprocal
|ja/jc| ¼ exp(zF/RT) for < 0 become small. It
Equation [56] does not give any information about must be realized that these ratios are lower than 5%
the charge-transfer coefficient a since it depends only as soon as || is greater than about 80 mV/z and lower
on the exchange current density j0. However, the deter- than 1% as soon as || is greater than about 120 mV/z.
mination of j0, through the slope of the linearized It is more convenient to express eqns [58] and [59]
Butler–Volmer equation around the equilibrium, is an in a semilogarithmic scale (the ln 10 ¼ 2.3 factor
important kinetic attribute, since it is easily related to stands for the conversion of the natural logarithm
k0 with eqn [53]. It is also important to mention that the into the more easily handled decimal logarithm), as
slope of the linearized Butler–Volmer equation [56] in the following relationships, often called Tafel laws:
around the equilibrium can be rewritten with the easily azF
experimentally measurable ‘polarization resistance’ log j log ja ¼ log j0 þ
2:3RT
(Rp ¼ q/qI ) at the equilibrium (I ¼ 0), and expressed 0:08
in ohms. In the particular case presented here, this for > ð ja
j j c j Þ ½60
z
resistance is equal to the charge-transfer resistance,
Rct, also defined as q/qI at I ¼ 0, but under conditions ð1 aÞzF
where only activation polarization contributes to (i.e., logðj Þ log ðjc Þ ¼ log j0
2:3RT
Rp includes the effects of mass transport and solution
0:08
resistance, but Rct does not): for < ðjjc j
ja Þ ½61
z
1 RT 1 The terms azF/2.3RT and (1 a)zF/2.3RT are
jS ¼ ) Rp ¼ Rct ¼ ½57
Rp zFS j0 known as the anodic and cathodic Tafel coefficients,
ba and bc respectively.
In practice, the SRp value in O cm2 is more commonly Thus, in the irreversible domains (high Z) the
used (and usually also termed Rp, but with units of Butler–Volmer equation is represented by two
O cm2) because it does not depend on the surface straight lines in a semilogarithmic scale, as depicted
area of the electrode. in Figure 21 (same data as in Figure 18) showing a
log |j| vs. plot generally called a Tafel plot. In this
Large overpotential limit of the Butler–Volmer case, the system is therefore said to exhibit Tafel
equation behavior. The Tafel plot is then a powerful tool to
When a large overpotential is applied to the interface, derive kinetic parameters: the equilibrium potential
one of the exponential terms (for instance, the and the exchange current density are obtained at the
cathodic one for large positive ) of the Butler– intersection of the extrapolations of the Tafel lines at
Volmer equation [52] vanishes and the corresponding ¼ 0, while the charge-transfer coefficient a is
reaction (cathodic in this case) can be neglected. derived from the slope of the cathodic or anodic
38 Basic Concepts
|j| in Figure 21, is not as simple as it may appear.

1 Indeed, if an unambiguous linear behavior is not
present, the graphic determination of j0 and a can
0.1 be seriously biased. Hence, it is widely accepted that
a system can be considered as exhibiting Tafel behav-
0.01 αzF ior and, therefore, the kinetic parameters can be
Anodic slope =
2.3 RT graphically estimated, only if the linear behavior
(1–α) zF extends over at least one decade of current.
Cathodic slope = –
2.3 RT
log j0
1.02.4.2.2 Mass-transport limited systems:
–0.2 0.0 0.2 η (V)
Fick’s law and limiting current density
Figure 21 High-overpotential limiting case: the Tafel As already mentioned, mass transport plays a domi-
behavior of the Butler–Volmer equation over a wide nant role in the overall kinetics of the electrochemi-
potential domain (z ¼ 1, j0 ¼ 1 mA cm2, and a ¼ 0.8). cal system in some important situations. This section
is then devoted to systems for which the kinetics of
lines (see Figure 21). It is important to keep in mind the redox reaction is so fast that it is completely
that up to now the cathodic and anodic reactions are controlled by mass transport. Only Nernstian sys-
the forward and backward reactions of the same redox tems, already discussed in Section 1.02.3.3, for
process and, therefore, they have the same kinetic which the potential and the interfacial concentrations
parameters (charge-transfer coefficient a and standard are directly related, are now considered. This is par-
rate constant k0). The slopes of the two straight lines are ticularly the case of fast redox systems that are
then mutually linked according to reversible in a large range of overpotentials.
zF The qualitative description of the kinetics of the
jðAnodic slopeÞ ðCathodic slopeÞj ¼ ½62 redox couple Zn2þ/Zn, illustrated in Figure 16 and
2:3RT
discussed in Section 1.02.2.4.4, allowed the idea of
Thus, the slope of a given branch automatically deter-
concentration profile across the interface related to a
mines the slope of the other one.
current flow to be introduced. This issue is now
In Figure 21, the steep increase of the anodic
treated in a more quantitative way with special atten-
current for positive potentials, due to a high value
tion to interfaces in steady-state conditions. Let us
a ¼ 0.8, is then necessarily coupled with a slower
first remember that, according to Nernst’s model (see
evolution of the cathodic current (in absolute value)
Section 1.02.2.4.3), convection controls the thick-
with respect to the potential.
ness, d, of the diffusion layer at the interface, and
Another crucial aspect that deserves to be men-
allows the bulk concentration to be kept constant not
tioned again is that Tafel behavior assumes the
only by electrolyte homogenization but also because
absence of mass transport effects (see the next sec-
the bulk is voluminous. In these conditions, the con-
tion). This assumption is valid for systems requir-
centration profiles of the reactive species across the
ing high overpotentials to reach significant current
diffusion layer are assumed to be linear.
values, as for slow redox couples characterized by a
As an example, in the anodic branch ( > 0) of the
small exchange current density. Indeed, as already
Zn2þ/Zn couple, metallic Zn is transformed into
discussed, mass transport has no influence on the
Zn2þ cations. Therefore, the interfacial concentration
current in that case. In contrast, for fast redox pro-
of Zn2þ, Cint, is higher than that in the electrolyte
cesses, that is, for redox couples with a high exchange
bulk, C1, so that the concentration profile decreases,
current density, the consumption of the reacting spe-
as depicted in Figure 22. Zn2þ cations in excess at
cies at the interface is faster than the supply of species
the interface diffuse towards the bulk electrolyte.
by mass transport and the system no longer exhibits
Since the concentration profile is assumed to be
Tafel behavior. As a consequence, the Tafel laws
linear inside the diffusion layer, Faraday’s and Fick’s
(eqns [60] and [61]) cannot be experimentally
laws30,31 yield the following relationship between the
observed for fast redox processes, which leads to the
current density and the concentration gradient for
idea that Tafel behavior is an indication of slow
any Mzþ/M couple:
kinetics. It must also be emphasized that the deter-
mination of the kinetic parameters from the linear C1 Cint
j ¼ zDF ½63
part of the current–overpotential curves, as described d
Electrochemistry 39
where D is the diffusion coefficient of the cation, often Zn can be regarded as an infinite reservoir, a limiting
expressed in cm2 s1 (about 105 to 106 cm2 s1 in situation is reached when the interfacial concentration
aqueous solutions). (Fick’s first law states that a diffu- becomes equal to zero. In that case, the charge-transfer
sive flux is proportional to the diffusion coefficient D activation process (redox kinetics) is so fast, relative to
of the diffusing species in a given electrolyte and mass transport, that any cation reaching the interface is
to the concentration gradient across the diffusion immediately reduced. At large negative overpotentials,
layer,31 that is j ¼ DðqCðxÞ=qxÞ.) this leads to the appearance of a plateau defining a
In the cathodic branch ( < 0), the concentration limiting current in the current–potential curve, as illu-
increases with the distance from the electrode because strated in Figure 16. Introducing zero interfacial con-
of the depletion of cations at the interface, as repre- centration in eqn [63] yields:
sented in Figure 23. The reduction reaction that entails C1
metallic Zn deposition on the surface, now depends on jlim;c ¼ zDF ½64
d
the supply of cations diffusing from the bulk towards
the interface, so that eqn [63] remains valid. However, The effective value of the limiting current
in contrast to the anodic branch where the metallic depends on the hydrodynamics (roughly speaking,
the stirring conditions) of the system, which directly
defines the thickness of the diffusion layer d. Just to
Cint give an idea, for a 106 cm2 s1 diffusion coefficient,
0.1 mol l1 concentration, 10 mm thick diffusion layer,
E
C∞ l and z ¼ 1, the limiting current density is about 10 mA
Z e cm2 in absolute value, according to eqn [64].
i c
n t Another example concerns the rotating disk elec-
c r trode, for which the Levich equation predicts the
o
l limiting current density, taking into account both
e− y the diffusion rate across the diffusion layer and the
t
I e complex solution flow pattern32:
Zn2+
Diffusion C = [Zn2+] jlim;c ¼ 0:620zFD2=3 o1=2 n1=6 C1 ½65
Diffusion layer where o is the angular rotation rate of the electrode

(in rad s1) and n is the kinematic viscosity of the
Figure 22 Zn2+ concentration profile across the Nernst solution (about 102 cm2 s1 in common aqueous
diffusion layer for an anodic reaction according to Nernst’s solutions). For the parameter values given above
model.
and a rotation rate of 1000 rpm (104.7 rad s1), the
current density is about 13 mA cm2.
In addition to the quantitative expression of the
Z
limiting current given by eqn [64], the mathematical
E
i l expression for the whole current–potential curve can
n
Zn2+ e be derived by combining eqns [39], [63], and [64].
c C∞ c
Diffusion t For example, the current–potential curve of a fast
e− r redox couple Mzþ/M, schematically represented in
o
I l Figure 16, has the following form:
y
t RT jlim;c j
Cint ≠ 0 e ¼ ln ½66
zF jlim;c
Cint = 0
C = [Zn2+]
1.02.4.2.3 General case: Mixed controlled
Diffusion layer systems
In the general case of an electrochemical system,
Figure 23 Zn2+ steady-state concentration profile across
the diffusion layer for a cathodic reaction. Dashed line:
interfacial charge-transfer and mass transport simul-
limiting case with Cint ¼ 0, corresponding to the diffusion taneously intervene in the kinetics of the reactions at
limited current plateau in Figure 16. the interface. The mathematical expression of the
40 Basic Concepts
current is similar to that used in the current– 2

potential equation [47] with interfacial concentra- jmix,a
tions instead of bulk concentrations. As an example, jact,a

for the model corrosion reaction [18] in which mass 1
transport is assumed to intervene for the cation Mzþ j0 jmix
only when j0,mix
0
j ¼ ja þ jc ¼ zF ðka kc ½Mzþ int Þ jmix,c

azF
¼ zFk exp
0
ðE E Þ
0 –1 jlim,c jact
RT jact,c

ð1 aÞzF
½Mzþ int exp ðE E 0 Þ ½67 –2
RT –0.3 –0.2 –0.1 0 0.1
Overpotential h (V)
2 1
where ka is now expressed in mol m s while kc is
still expressed in m s1. With reasoning similar to that Figure 24 Comparison of the current–overpotential
curves for pure activation (thin solid line and dotted lines for
used in eqns [48]–[51], the following relationship, the anodic and cathodic currents) and mixed (activation and
similar to the usual Butler–Volmer equation, can be mass transport: thick solid line and dashed lines for the
derived: anodic and cathodic currents) kinetics with a cathodic
limiting diffusion current (z ¼ 1, j0 ¼ 1 mA cm2, a ¼ 0.8,
azF ½Mzþ int ð1 aÞzF jlim,c ¼ 1 mA cm2).
j ¼ j0 exp zþ exp
RT ½M 1 RT
½68
where the mass-transport current density, jmt, corre-
where j0 is now given by: sponds to the current density for a very fast redox
reaction ( j0 ! 1 in eqn [70]), that is, for a reaction
j0 ¼ zFk ½Mzþ a1 ½69
under complete mass transport control. This simple
z+
The interfacial concentration of M can be evalu- expression highlights the cornerstone of the idea
ated with the Fick’s first law (eqn [63]), which gives of interface control. If jact is much larger than jmt,
the following expression of the current–potential eqn [71] predicts that the mixed current is very close
curve: to jmt. From a physical point of view the kinetic
1 behavior can be seen as a series connection of two
j ¼ jact ½70 steps: the slowest step determines the overall reaction
d ð1 aÞzF
1 þ j0 exp rate. If the activation kinetics is fast, then the mass-
zFD½Mzþ 1 RT transport process is the limiting step. Conversely, if
where jact stands for the ‘activation current density’ the activation kinetics is slow, the diffusion process
for which the concentration of the cations at the does not hinder the overall reaction rate.
interface is equal to that in the solution bulk (jact ¼ j An important consequence of the mixed control
in eqn [52]). Equation [70] reduces to the diffusion- depicted in Figure 24 is that the exchange current
limited current density in the cathodic region, density, j0,mix, is smaller than the kinetic exchange
already given by eqn [64] with C1 ¼ [Mz+]1, when current density j0 derived from the k0 and a para-
tends to 1. meters. The expression of j0,mix can be obtained from
Figure 24 compares the current–potential curves eqn [70] with Z ¼ 0. In the same way, the polarization
of a purely-activated system and a mixed controlled resistance (reciprocal of the slope of the current–
system with the same kinetic parameters (eqns [52] potential curve in the vicinity of Eeq, see eqn [57])
and [70]). It can be seen that at any potential the depends on mass transport.33 Its expression, derived
current (in absolute value) is always smaller under from the linear approximation of the polarization
mixed control. Another general expression for any curve at small overpotentials, is:
redox reaction may be derived from eqn [70]:

1 1 1 1 q RT 1 1
¼ þ ½71 Rp ¼ ¼ ½72
j jact jmt S qj j ¼0 zFS j0 jlim;c
Electrochemistry 41
This illustrates the fact that the overall polarization and not to coupled kinetic parameters k0 and a as
resistance is always higher under mixed control than in the case of a single forward–backward reaction
under activation control (for the same j0 and a). (with a and 1a, respectively), as discussed before.)
As described below, both Ecorr and Icorr are time-
evolving variables that define the intrinsic thermody-
namic and kinetic characteristics of each redox
1.02.5 Corrosion: Far from
couple taking part in the global reaction. This is
Equilibrium Zero Net Current Systems
why the corrosion potential is often referred to as a
1.02.5.1 Qualitative Description of the Free mixed potential. A simple qualitative description of a
Corrosion Scenario coupled system can be proposed with the example of
Everybody knows or easily guesses the consequence the free corrosion of iron in acidic media: the zero net
current balance is then a result of the following
of abandoning a metallic piece in an aggressive
coupled equations:
medium without specific protection, for instance,
iron in seawater: the metal will be annihilated by
corrosion. Regardless of the rate of the process, Fe!Fe2þ þ 2e
which depends on many factors such as the nature 2Hþ þ 2e !H2 ½73
of the metal and of the electrolyte, temperature,
pressure, and so on, this primary experimental evi- The two individual equilibrium potentials,
dence leads to an important conclusion: any corrosion Eeq;Fe2þ =Fe and Eeq;Hþ =H2 , can be defined for reactions
process evolves under long-term out-of-equilibrium [73] if the four species are present at the metal–
conditions. In spite of being permanently in non- electrolyte interface. Their exact values depend on
equilibrium conditions, the system is, in some sen- the experimental conditions, as described by the
ses, at rest in the free corrosion scenario. In other Nernst law (see Section 1.02.3). The fact that each
words, no net current flux is crossing the metal– couple behaves in a stationary way allows the steady
electrolyte interface. state of the corrosion system to be defined from their
As an obvious consequence, if the metal is being individual polarization curves (see Section 1.02.4) and
consumed by corrosion processes, oxidation reactions its corrosion rate to be calculated. Since iron is oxi-
inevitably take place at its surface although the net dized when corroding, its polarization point must
current flow is zero. Therefore, at every instant the be somewhere along the anodic branch of the Fe2þ/
amount of current generated by these oxidation reac- Fe -I curve, the shape of which is a function of the
tions (for instance, reaction [2]) must be balanced by thermodynamic and kinetic attributes of the couple, as
cathodic counterparts. This implies that anodic and illustrated in Figure 25. The behavior of the second
cathodic reactions must be occurring simultaneously. couple is quite evident, since it must ensure the zero
In contrast with the previous sections, this scenario overall current by consuming the electrons supplied
cannot involve a single redox couple with a simple by the anodic reaction. Hþ is then reduced and the
forward–backward reaction, otherwise the zero net polarization point of the couple is located on the
current would correspond to an equilibrium condi- cathodic branch of the Hþ/H2 -I curve. The corro-
tion in which the iron dissolution and deposition sion process is then iron dissolution with hydrogen
would perfectly balance out and no corrosion would evolution according to reactions [73]. The exact situa-
occur. Corrosion phenomena, therefore, involve at tion is graphically determined by the potential for
least two different redox couples, one (or more) sup- which |Ic| ¼ Ia, as shown in Figure 25.
plying the anodic current and the other one (or more) Since, very often corrosion processes can be con-
supplying the cathodic current and ensuring the zero sidered as activation-controlled systems (at least for
current balance. This kind of self-regulated coupled low current values), another useful representation for
system yields the two fundamental parameters of free corrosion conditions is the so-called Evans dia-
corrosion processes, the corrosion current, Icorr, and gram, which gives the potential evolution against the
the corrosion potential, Ecorr, also called rest potential logarithm of the current absolute value, as depicted
or open-circuit potential (OCP). (The term coupled in Figure 26, strictly Evans originally plotted E
here refers to the fact that the anodic and cathodic against |I|, i.e., using linear rather than log scales.
reactions must deliver the same average current However, the name is now widely used for diagrams
(in absolute value) to maintain a zero current balance, plotting E against log |I|.
42 Basic Concepts
I Fe → Fe2+ + 2e– out of equilibrium and follow the behavior predic-

H2 → 2H+ + 2e− ted by their own polarization curves, as shown in
Figure 25. In spite of this complex behavior, Ecorr
naturally lies between the two individual equilibrium
Ia potentials. The two (or more) reactions do not
Ecorr usually take place with the same overpotential
Ia = |Ic| E
Eeq,Fe2+/Fe Fe2þ =Fe ¼ Ecorr Eeq; Fe2þ =Fe and |Hþ =H2 j ¼ jEcorr
Eeq,H+/H2
Ic Eeq; Hþ =H2 j (even if the equality Fe2þ =Fe ¼ jHþ =H2 j is
of course not forbidden), since what they must ensure is
an identical electron flux, identified to the corrosion
current, Icorr , as depicted in Figure 26. Both redox
H2 ← 2H+ + 2e− couples are displaced from their respective equilibrium
Fe ← Fe2+ + 2e– potential, which means that the Fe2þ/Fe couple shifts
Figure 25 Schematic representation of the polarization the proton reaction in the cathodic direction while the
curves of the two redox couples around their respective Hþ/H2 couple shifts the iron reaction in the anodic
equilibrium potential allowing the corrosion potential Ecorr, direction.
at which |Ic| ¼ Ia, to be determined. Note that the
This idea of mixed potential is extremely impor-
ordinates are plotted in terms of net currents and not
current densities, since the equality between the anodic tant as it indicates that the reaction with the lower
and cathodic counterparts concerns the electron flux equilibrium potential will be anodic and that with the
(see Section 1.02.2.3). higher equilibrium potential will be cathodic. It then
defines in principle which metal will corrode in a
E
given situation of galvanic corrosion. (This is only
necessarily true when both metals are corroding in an
Ianod,H
active state; in practical situations (which often
involve one or both of the metals being passive) the
2H+ + 2e– H2 standard equilibrium potentials are not reliable indi-
Eeq,H+/H cators of the behavior of galvanic couples.) Let us
2 Icath,H
take the example of iron and copper in the presence
Ecorr
of an acidic electrolyte, that is, in the presence of the
Hþ/H2 couple, but in the absence of dissolved oxy-
Ianod,Fe gen. The respective standard equilibrium potentials
are 0.44, 0.34, 0 VSHE for iron, copper, and hydro-
gen, respectively. It clearly appears that iron and
Eeq,Fe2+/Fe Fe Fe2+ + 2e– copper behave differently in contact with an acidic
Icath,Fe solution. The iron will corrode, driven by the proton
reduction with concomitant H2 evolution, whilst the
I0,Fe2+/Fe I0,H+/H2 Icorr log I copper will remain intact since its potential is higher
than that of the Hþ/H2 couple. The importance of
Figure 26 The Evans diagram: a simplified representation specifying the absence of oxygen in this example is
of the mixed potential and the mutually balancing cathodic due to the high potential of the O2/H2O couple
and anodic electron fluxes that determine the corrosion
(1.23 VSHE) that can induce the corrosion of copper
potential and exchange current values. The figure is arbitrarily
asymmetric with Ecorr Eeq, Fe2þ=Fe > |Ecorr Eeq; Hþ=H2 |. as well as that of iron.
The values of the exchange currents are also arbitrary. As already mentioned, it is assumed in reactions
[73] that the species produced by the forward reac-
The main advantage of the Evans diagram is to tions are available at the interface, in particular, the
clearly illustrate the fact that when two different Fe2þ cations are not all trapped in indissoluble cor-
redox couples coexist, the electrode potential always rosion products. Reactions [73] were intentionally
takes a single value, the corrosion potential, which is represented with forward and backward arrows of
easily determined by the intersection of two bran- different size instead of purely anodic and cathodic
ches. The exact Ecorr value (as well as the corrosion reactions [2] and [12]. This is of high importance
current value) depends on the kinetic and ther- because, depending on the overpotential of each
modynamic characteristics of both couples that are reaction relative to the actual corrosion potential,
Electrochemistry 43
the reverse reaction (iron deposition or anodic where the subscripts a and c stand for anodic and
hydrogen oxidation) can also take place on the metal- cathodic respectively and where a ¼ E Eeq,a and
lic surface even though its rate will be less than the c ¼ E Eeq,c. Equation [74] is implicitly derived for
anodic and cathodic counterparts (iron dissolution conditions where both processes are completely inde-
and proton reduction, respectively). When the pendent and in a steady state. This means, for example,
mixed potential Ecorr entails an overpotential, high that the sharing of the surface area in anodic and
enough to bring the reaction inside the irreversible cathodic zones does not change with the potential,
domain, the reverse reaction can be neglected, as and is constant during the experiment.
discussed in Section 1.02.3.2. This is often the case Let us now define the polarization p ¼ E Ecorr ,
in corrosion processes and this constitutes the basis of analogous to the overpotential when dealing with
the simplified kinetic treatment applied to corrosion equilibrium systems. Assuming, as discussed above,
studies, which is developed in the next section. that both couples are sufficiently far from equilibrium
to be in the irreversible domain, the current–potential
1.02.5.2 Electrochemical Kinetics Around curve of the overall system is composed of a single
the Corrosion Potential anodic branch (corresponding to one of the oxidation
In spite of the enormous physical difference between process of the redox couples) and a single cathodic
the equilibrium and the corrosion potentials, both branch (corresponding to the reduction process of the
potentials concern systems that are released at rest other redox couple). The currentpotential relation-
and thus show a permanent zero-net current balance. ship can hence be described by the following equation,
In Section 1.02.4, the kinetic behavior was investigated which is derived from eqn [74] and is very similar to the
when imposing a potential shift with respect to the Butler–Volmer equation34:

equilibrium potential. It is interesting now to consider aa z a F ð1 ac Þzc F
the kinetic consequences of imposing a potential shift j ¼ jcorr exp p exp p
RT RT
with respect to the corrosion potential. This raises a
strongly intuitive analogy between the kinetic beha- ½75
viors around Eeq and Ecorr that is fully verified experi- with
mentally. The first aspect in this analogy is that, as
seen in Section 1.02.4.2.2, beyond a certain deviation aa z a F
jcorr ¼ j0;a exp
from equilibrium, generally about a 100 mV overpo- RT a;corr

tential, the system can be considered to be in the ð1 ac Þzc F
irreversible Tafel domain. Considering the Pourbaix ¼ j0;c exp c;corr ½76
RT
diagrams, or simply a standard equilibrium potential
table, it can be seen that the differences between The parameters aa and ac are no longer mutually
the equilibrium potentials of most of the typical related as for a single transfer coefficient (aa ¼
6 ac),
redox couples inducing metallic corrosion are much and eqn [75] is often represented in a simpler way:
larger than this value of 100 mV (e.g., 1.23 – j ¼ jcorr ðexpðba pÞ expðbc pÞÞ ½77
(0.62) ¼ 1.85 V for Fe–O2, and 0.62 V for Fe–Hþ
according to Figure 13 for [Fe2þ] ¼ 106 mol l1 and where ba (ba > 0) and bc (bc < 0) are the anodic and
pH ¼ 0, just to mention the most common corrosion cathodic Tafel coefficients, respectively, expressed
couples involving iron). This indicates that most of the in V1. It is easily seen that for potentials sufficiently
common corrosion processes take place in the irrevers- far from Ecorr and in the absence of mass-transport
ible domain. In such conditions, and provided that the effects, the anodic or cathodic branches yield a linear
mass transport effects are negligible, the Butler–Volmer behavior in a semilogarithmic scale, as depicted in
equation related to each reaction is reduced to a single Figure 27.
anodic or cathodic branch so that the polarization curve The experimental Tafel slopes, ba and bc, exp-
of the overall coupled system is defined as follows: ressed in volt per current decade, can be determined
from Figure 27 with the following equations:
aa z a F
j ¼ j0;a exp 2:303 2:303
RT a ba ¼ and bc ¼ ½78
ba bc
ð1 ac Þzc F The corrosion current density can also be estimated
j0;c exp c ½74
RT by extrapolation of the Tafel lines at the corrosion
44 Basic Concepts
10 m |j| (A cm–2) 10–4

2
1
1m
jcorr,1
10–5
j (A cm–2)
100 μ
jcorr,2
jcorr
10 μ
10–6
1μ Ecorr = –0.11 mVSHE
–0.15 –0.10 –0.05 0.00 0.05 0.10 0.15 0.20

E (VSHE)
–0.2 0.0 0.2 π (V)
Figure 28 Anodic polarization curve of Zn in a 0.01 mol l1
Figure 27 Kinetic behavior within a large potential NaHCO3 solution under O2 bubbling at 60 C, pH ¼ 8.1:
domain: Tafel (linear) region and estimation of the corrosion misestimation of the corrosion current density from a
current density. Arbitrary values of jcorr ¼ 2 mA cm2, nonlinear or small-range Tafel behavior.
aa ¼ 0.75, ac ¼ 0.5, za ¼ zc ¼ 1.
potential (p ¼ 0), similar to the estimation of the Tafel line as in Figure 28. If the cathodic branch
exchange current density for a redox couple exhibits a Tafel domain, which is often the case for
(Figure 21). It is worth emphasizing that both char- hydrogen evolution but rare for oxygen reduction,
acteristics correspond to far-from-equilibrium con- the corrosion current can be determined by extrapo-
ditions, each branch following its intrinsic kinetic lating the cathodic linear branch up to the corrosion
behavior independently of the other. potential. (A common error in applying Tafel extrap-
From a practical point of view, the accurate deter- olation is to extrapolate the Tafel line for hydrogen
mination of the corrosion current, a sort of Holy evolution, even though the dominant cathodic reac-
Grail for corrosion engineers, deserves two important tion is oxygen reduction.) This is really useful, as the
remarks. The first, extremely relevant and often dis- anodic and cathodic branches are independent with-
regarded, is that the polarization curve should show a out any expected symmetry or even equivalent shape.
large and unambiguous region of Tafel behavior; the Let us come back to the previous figure but in its
linear domain should encompass at least one decade complete form, showing the anodic and cathodic
of current to allow the graphical determination of branches (Figure 29). Even if the anodic reaction
the corrosion current to be carried out. Despite does not present a marked linear log(jcorr) – E char-
that, several commercial corrosion analysis programs acteristic, as seen before, the corrosion current den-
allow manual Tafel fits from small truncated straight sity can be correctly estimated by means of
lines or even when there is no linear behavior at all. the unambiguous and wide Tafel behavior of the
Figure 28 illustrates this issue in the case of a non- cathodic branch.
linear anodic curve measured during corrosion of Zn In the example presented in Figures 28 and 29, the
in a highly aerated carbonate solution at 60 C after jcorr values badly estimated with the anodic branch are
several hours of immersion (for illustration, the higher than the actual jcorr value. This overestimation
cathodic branch is ignored for the moment). The is certainly less harmful than an underestimation, but
magnitude of the possible error in jcorr due to the reverse situation cannot be simply excluded.
the arbitrary choice of two linear fits (represented Indeed, due to the logarithmic scale, this loose
by straight lines 1 and 2), corresponds to a factor of approach to the Tafel laws can yield significant errors,
about 3 between the values of jcorr,1 and jcorr,2. the impact of which in the nuclear domain, for
The second remark is more positive: since jcorr is instance, is easy to imagine and hard to mitigate.
determined at the corrosion potential, the existence It is worth noting, however, that it is possible to
of a single Tafel region (anodic or cathodic) is suffi- evaluate the corrosion current density with the
cient. This is particularly useful when the anodic Stern–Geary relationship in a narrow potential
branch, for example, does not present a well defined domain. In addition, the four unknown variables,
Electrochemistry 45
consequence, the polarization resistance, which is an

10–3 important parameter for the assessment of corrosion
current in field applications, is higher in that case.
10–4
j (A cm–2)
jcorr 1.02.6 The Electrochemical Cell

10–5 This section is devoted to a brief description of an
electrochemical cell and of the basic instrumenta-
tion used for controlling the electrode potential
10–6
Ecorr = –0.11 mVSHE (potentiostat) or the current flowing across the elec-
trode (galvanostat). Numerous techniques, electro-
–0.4 –0.3 –0.2 –0.1 0.0 0.1 0.2 chemical or in situ spectroscopic, may be used
E (VSHE) for studying electrode reactions at equilibrium or
Figure 29 Complete polarization curve obtained in the out-of-equilibrium, such as potentiometry, ampero-
same conditions as in Figure 28: good estimation of the metry, voltammetry, electrochemical impedance,
corrosion current density by means of a single wide Tafel electrochemical noise, scanning electrochemical micro-
region. scopy, atomic force microscopy, Raman spectroscopy,
etc. These techniques will not be described in this
section, the reader may consult other chapters in this
jcorr, Ecorr, aa, and ac, can be evaluated with a param-
book, or Marcus and Mansfeld,35 Greef et al.,36 Bard
eter regression calculation, provided that the poten-
and Faulkner37 for more information.
tial domain is sufficiently large (about 50 mV from
Ecorr) to give a nonlinear current–potential curve
1.02.6.1 Design of the Electrochemical Cell
(four unknowns can then be determined instead of
two for a linear curve) and, at the same time, suffi- Although the electrochemical phenomena occur on a
ciently narrow in order that the system behavior single electrode in most corrosion processes, as
obeys eqn [75]. described in Section 1.02.2, the understanding and
One reason that the anodic branch exhibits a quantitative analysis of these phenomena are usually
poorly defined Tafel domain is the dissolution mech- performed with the help of three electrodes in con-
anism itself. The metal is dissolving with multiple tact with the same electrolyte, to measure and/or
elementary steps, and the rate-limiting reaction step control the electrode potential and current precisely,
may change with the potential, as illustrated by the as described briefly in the following.
extensive work of Keddam et al. On the other hand, at Basically, the electrochemical cell comprises three
the corrosion potential of iron in sulfuric acid, the electrodes immersed in an electrolyte, the working
coverage of the electrode surface by hydrogen atoms electrode (WE), which is the electrode under study,
is close to unity,9 and, as can be conceived readily, the the counter-electrode (CE), the role of which is to
surface area available for iron dissolution increases ensure the current flow across the cell, and the refer-
with increasing potential. This is another reason ence electrode (RE) versus which the potential of the
explaining the narrow linear domain of Tafel behav- WE is measured. The electrodes may be located in
ior of the anodic process. In contrast, the surface area single-compartment containers or in containers with
at which hydrogen evolution is taking place does not two compartments separated by a membrane or a
change significantly so the polarization curve exhibits sintered glass, for example, to prevent the species
Tafel behavior in the cathodic domain. produced at the counter electrode from reaching
It is worth mentioning that when mass-transport the working electrode.
phenomena influence the electrochemical response, Depending on the system under investigation, the
leading to a non-Tafel behavior of the two couples, reaction kinetics may be studied in stagnant solutions
the determination of the corrosion current cannot be or in forced convection conditions. Mass-transport
achieved as straightforwardly as for Tafel behavior. control in the cell may be realized by agitating the
Indeed, for a redox couple under mixed control at solution with gas or a stirrer, by rotating the electrode
equilibrium (Section 1.02.4.2.3), it can be shown that (a disk or a cylinder), or by flowing the solution in
the corrosion current is lower when mass transport is front of the WE (immersed jet cell or channel flow).
controlling the kinetics of the corrosion system. As a Whenever feasible, controlling mass transport effectively
46 Basic Concepts
allows a better control and definition of the flow solution. In practice, the reaction occurring on the
conditions. This helps to ensure steady state condi- CE is that of presenting the lower CE overvoltage. In
tions, improve reproducibility, and make quantitative corrosion studies, they usually consist in the oxida-
analysis easier with the help of the Fick and Levich tion or reduction of a component of the electrolytic
equations. solution, such as hydrogen evolution when the WE is
polarized anodically and oxygen evolution when it is
1.02.6.1.1 The working electrode polarized cathodically (or chlorine evolution in chlo-
The design, size, and materials of the WEs are ride solutions). Another oxidation reaction may take
diverse. Very often, especially in corrosion, the WE place when using a noninert CE, such as stainless
is a metallic disk or square, the lateral surface of steel (special care must be taken in that case to
which is embedded in an epoxy resin or covered avoid contamination of the working electrolyte). As
with an electrical insulator. Care must be taken for mentioned above, the CE must be located in a spe-
mounting the WE since crevice effects between the cific compartment separated with an ion-exchange
sample and the insulator are possible. A thin layer of membrane or a sintered glass when using a stainless
cataphoretic paint may be deposited on the lateral steel CE or when the reaction product, such as dis-
surface of the sample before mounting in the insula- solved oxygen, may alter the behavior of the WE.
tor to avoid this. The size of the WE may be very
large in industrial applications but in most laboratory 1.02.6.1.3 The reference electrode
investigations, the surface area is of the order of A RE is used to control and/or measure the electrode
1 cm2. To obtain a reproducible surface state, the potential in almost all electrochemical investigations.
WE is usually polished before the experiment with However, in early work, before the common use of
abrasive paper, alumina, or diamond paste, rinsed and RE, the potential was sometimes measured with
placed in an ultrasonic bath to remove the polishing respect to a CE of large surface area. This is also
particles. In most cases, the WE is fully immersed in the case at times for experiments performed with
the electrolyte but in marine corrosion applications, microelectrodes. This section is aimed at discussing
the sample may pass through the air–electrolyte the main principles of the RE and its positioning in
interface, which provokes water-line attack due to the electrochemical cell.
enhanced oxygen availability (see Section 1.02.2.3).
The main reference electrodes
1.02.6.1.2 The counter-electrode
As explained in Section 1.02.3.3, the electrode
The CE is often a cylindrical grid of large surface
potentials in the electrochemical systems are referred
area compared to the WE for several reasons. First,
to a standard system called SHE, which, strictly
this minimizes the polarization of the CE (because
speaking, cannot be realized, and the derived experi-
of the low current density on a large electrode) and,
mental RE, the NHE, is not easy to handle. A number of
therefore, the overall voltage required from the
practical REs were, therefore, developed and are now
power source. Second, the current lines between the
commonly used.18 There are two basic classes of prac-
WE and the CE spread inside a larger electrolyte
tical RE, sometimes called secondary REs since they
volume with a large CE, so that the electrolyte resis-
are calibrated against the primary reference of SHE:
tance is located near the WE, as shown below, which
gives a lower influence of the RE position on the REs of the first kind consist of a metal in equilib-
measured potential. When possible, the CE is posi- rium with a controlled concentration of its salt. An
tioned far from the WE so that the cylindrical shape example of this type is copper metal in a solution
ensures that the CE is an equipotential surface. (The of copper sulfate. Then the copper ion concentra-
ready availability of numerical modeling tools that tion controls the potential as defined by the Nernst
can determine the current distribution now makes it equation.
possible to improve the design of electrochemical REs of the second kind consist of a metal in contact
cells for specified situations.) Inert electrodes, metal- with a sparingly soluble salt of the metal; this is
lic grids (often in platinum or titanium) or carbon placed in a solution with a controlled concentra-
sheets may be used. When a single compartment is tion of a soluble salt with the same anion. An
used, the reactions taking place on the CE have to be example of this type is silver in contact with silver
considered with great care since they may cause chloride in a solution of potassium chloride. In this
changes in the concentration of the reactants in case the chloride concentration in solution
Electrochemistry 47
controls the silver ion concentration by way of the A generic consequence of the Nernst equation for the
solubility product of silver chloride, and the silver reversible redox couple involved in these REs, is that
ion concentration controls the potential as defined in order to obtain a RE showing a constant potential
by the Nernst equation. with time, the electrolyte inside the RE must contain
a salt at fixed concentration. One of the advantages of
Commercially-available REs include:
using a saturated salt solution (as in the first two
Hg|Hg2Cl2 (calomel) in a saturated KCl solution, examples) is that the presence of the solid salt main-
named SCE for saturated calomel electrode. This tains a fixed ionic concentration, given by the solu-
system is a very common commercial product bility of the salt, even if small perturbations may
(though tending to lose favor because of toxicity lead to a slight consumption of this ion. However,
issues and a ban on the transportation of products the saturation concentration is temperature sensitive
containing mercury by air): a platinum wire is and this kind of RE has a rather high tempera-
dipped into a mercury pool (Figure 30), which is ture coefficient (e.g., 0.65 mV C1 for the SCE
in contact with a paste of mercury and Hg2Cl2, compared to 0.06 mV C1 for the calomel RE in
itself in contact with a saturated KCl solution 0.1 mol l1KCl).
(4 mol l1 at room temperature). To avoid contaminating the working solution with
Hg|Hg2SO4 in a saturated K2SO4 solution. This the salt used in the RE, a salt bridge filled with the
system may be used for general purposes where working solution is often used, thereby increasing the
the absence of chloride ions in the electrolyte is distance that ions leaking out of the RE have to
crucial. diffuse. Additionally, the use of the Hg|Hg2SO4 RE
Hg|HgO in a KOH solution of fixed concentra- filled with saturated K2SO4 in sulfate media, or the
tion, for alkaline electrolytes. use of the Hg|HgO RE filled with KOH in strongly
Ag|AgCl in a KCl or NaCl solution of fixed con- alkaline media, may minimize solution contamina-
centration. This RE is quite easy to prepare: for tion. Except for the Ag|AgCl RE that can be directly
instance, an Ag wire can be dipped into a concen- dipped in seawater, the use of these solid REs always
trated KCl solution and anodically electrolyzed or introduces one or more liquid–liquid interfaces
it can be dipped a few minutes in a concentrated (e.g., in the porous material used to separate the RE
FeCl3 aqueous solution in which it is spontane- solution from the studied solution) in the potential
ously oxidized. The resulting wire is covered with measurement chain. This will inevitably produce
an AgCl deposit and may be used in any solution liquid junction potentials that introduce some errors
of fixed chloride concentration. It is very com- in the measurement and/or control of the WE inter-
monly used in marine applications when the Ag| face potential. The liquid junction potentials may
AgCl wire can be directly used in the seawater reach a few tens of millivolts in specific conditions.
without a specific electrolyte compartment, in However, when the solution in the RE is highly con-
contrast with the above REs. centrated with a cation and an anion of similar ionic
Cu in a saturated CuSO4 solution – this is the only molar conductivity as, for example, in saturated KCl
widely used RE of the first kind. This electrode is solution, the junction potential can be neglected.38
commonly used in the field, for instance, to evalu- In most corrosion applications, the error introduced
ate the efficiency of cathodic protection of buried by the junction potentials is usually of no consequence
carbon steel pipelines. It consists of a copper rod since the electrode potential does not have to be
dipped in a saturated CuSO4 solution, as shown in measured with an accuracy better than 1 mV.
Figure 31. Table 2 indicates the measured potential of the
REs presented above.
Plug (glass wool)

Hg Saturated KCl
Connector
Ceramic Cu-rod
fiber
Sat. CuSO4
junction
Cu-wire Pt-wire KCl crystals CuSO4 Porous
Hg/Hg2Cl2 crystals plug
Figure 30 Scheme of the saturated calomel reference Figure 31 Scheme of the Cu|saturated CuSO4 reference
electrode. electrode.
48 Basic Concepts
Position of the reference electrode Figure 33 shows that the normalized potential F0/
For a simple electrochemical cell, consisting of two Fint along the disk symmetry axis decreases almost
planar electrodes of the same size and positioned face linearly with the distance y from the electrode when y
to face, similar to a conductivity measurement cell, is smaller than the electrode radius. This potential pro-
the potential gradient between the two electrodes is file justifies the use of a Luggin–Haber probe (the tip of
essentially uniform and the solution resistance Re can the RE is a capillary usually filled with the studied
then be calculated as follows: solution and positioned close to the WE) to reduce
1d
Re ¼ ½79
gS
where g, d, and S are the conductivity of the electro-
lyte, the distance between the two electrodes, and
their surface area, respectively.
For more general cell geometries, the potential is
no longer uniform in the electrolyte. Figure 32 illus-
trates the potential and current distributions around a
Disk
disk electrode when no polarization occurs at the WE Guard ring
and CE and in a cell with an ideal geometry, that is, the
counter electrode of hemispherical shape is located at 0.8 0.6
an infinite distance, the disk electrode (radius r) is 1 h = 0.2
0.4
mounted in an infinite insulating plane, and the elec-
2 0.3
trolyte between the disk and the CE is homogeneous.42 h = 0.4
In these conditions, the equipotential surfaces, which 3 0.2
can be calculated by solving the Laplace equation in
elliptic coordinates, are ellipsoids. The potential and 4 h = 0.6
current distribution with no polarization at the WE
5 h=1 h = 0.9
and CE electrodes is called primary. h = 0.8
Provided the diffusion layers at the WE and CE 6
F
are two very thin parts of the electrolyte volume, = 0.1
x=7 F0
which is often the case with forced convection, and
assuming that the polarization of the electrode can be Figure 32 Primary potential (dotted lines) and current
neglected (this may not be true because of the non- (dashed lines) distributions around a disk electrode for an
uniform current density), the homogeneous part of ideal geometry cell. Adapted from Newman, J. J.
Electrochem. Soc.1966, 113, 501–502.
the electrolyte contributes almost the whole electro-
lyte resistance and the corresponding potential dis-
tribution in the electrolyte is close to that given by
Figure 32. The electrolyte resistance can then be
1.0
calculated from the potential distribution42:
11
Re ¼ ½80 0.8
g 4r
F 2 y
= 1 – tan–1 ( )
Potential F/Fint
Fint π r
Table 2 Potential of the main reference electrodes used 0.6
in corrosion39–41
Electrode Couple Potential (VSHE) 0.4
NHE Pt|H2|1 M HCl þ0

SCE Hg|Hg2Cl2|sat. KCl þ0.24 0.2
Mercurous Hg|Hg2SO4|sat. þ0.64
sulfate K2SO4
0.0
Silver chloride Ag|AgCl|3 M NaCl þ0.21 0 2 4 6 8 10
Silver chloride Ag|AgCl|0.1 M NaCl þ0.29 Distance y/r
Mercuric oxide Hg|HgO|20% KOH þ0.10
Copper sulfate Cu|sat. CuSO4 þ0.32 Figure 33 Potential profile along the symmetry axis of the
disk electrode.
Electrochemistry 49
the distance between the WE and RE as much as of the disc) and because the potential drop decreases as
possible and, therefore, decrease the corresponding 1/r (the current is proportional to r2, whereas the re-
ohmic drop, ReI.31 However, it is important to remark sistance is proportional to 1/r), an alternative approach
that the presence of the capillary close to the electrode to the minimization of solution resistance effects is to
surface induces a screening effect that may disturb the use a very small electrode (a microelectrode), in which
electrode potential markedly. In practice, it is then not case the RE is situated remotely from the WE.
essential to use a Luggin–Haber probe in conductive
solutions since the electrolyte resistivity is low, espe-
1.02.6.2 Potential Control and Current
cially in corrosion applications where the current is
Control
low as well. Because the majority of the solution re-
sistance is concentrated close to the electrode (note in In most recent commercial potentiostats and galva-
Figure 32 that 80% of the resistance occurs within 3r nostats, the potential or current control is performed
DC –Vset
potential
control –1 + I
Vset
OPA
– +
Rm GI
RmGI I
–
CE
RE
Cell
Vset
WE
+
–Vset
I GV
GVVset
–
(a)
DC –Vset
potential
–1 +
control I
Vset OPA
–
CE
RE Cell
Vset
–Vset WE
I
I +
–
GVVset GV – Rm I GI RmGI I
+
OPA
+
(b) Current-to-voltage converter (gain RmGI)
Figure 34 Schematic circuit of a potentiostat: the current is measured in the CE branch with a differential amplifier of
gain GI (a) or in the WE branch with a current-to-voltage converter of gain RmGI (b).
50 Basic Concepts
by using one or several operational amplifiers (OPA) DC

according to the electrical schemes described below. potential Iset
control +
For simplicity, only ideal OPA will be considered Vset OPA
here, which means that no current can flow in the +
–
positive and negative inputs of the OPA and that both Rm GI
inputs have the same potential at any time (assuming –
RmGIIset
that negative feedback is used to keep the OPA in the
linear region of its behavior). Deviations from this Cell
CE
ideal case, including nonzero input currents, limited RE
frequency bandwidth, etc., and their consequences on –
V GV
the potential or current control, have been investi- WE GVV
+
gated thoroughly elsewhere.43,44 Iset
Iset
Figure 34 shows the experimental set-up used for
controlling the electrochemical potential of a WE
Vset Rcontrol Iset = Vset/Rcontrol
connected to the ground. Usually, this ground is
floating, the chassis being connected to the earth
ground for safety reasons. The voltage Vset (with Figure 35 Schematic circuit of a simple galvanostat.
respect to the ground) set by the user on the direct
current (DC) potential control is first inverted
with an amplifier of gain 1 (this is provided only
in order to give the controlled potential the same sign Rcontrol is positioned between the WE and the ground
as Vset, and it may be omitted in some designs). The and a voltage Vset ¼ RcontrolIset (with respect to the
voltage Vset on the positive input of the OPA is then ground) is applied at the positive input of the OPA. In
imposed by the OPA on its negative input so that the these conditions, the voltage at the negative input of
potential difference between the RE and the WE the OPA is equal to Vset as well, so that a current
is Vset as well and, therefore, the potential of the Iset ¼ Vset/Rcontrol crosses the resistor. Since no cur-
WE with respect to the RE is actually equal to Vset. rent can flow at the negative input of the OPA, this
In Figure 34(a), the current I, which necessarily flows current Iset actually flows through the WE. Note that
to the WE since no current can enter the negative the WE is no longer grounded, which may be a
input of the OPA through the RE, is measured from drawback in some applications. To work correctly
the voltage drop RmI across the resistor Rm in the CE with a grounded WE, the potentiostat power supply
branch. This voltage is amplified with a differential must be isolated from ground or a galvanostat with a
amplifier of high input impedance and gain GI so that more complex electrical circuit must be used.
the gain in V/A of the current measurement is RmGI. In
some equipment, the actual potential of the WE is
measured with another differential amplifier of high References
input impedance (of gain GV in Figure 34).
The current flowing to the WE can also be 1. Itagaki, M.; Nakasawa, H.; Watanabe, K.; Noda, K. Corros.
Sci. 1997, 39, 901–911.
measured in the WE branch by using a current-to- 2. Muñoz, A. G.; Vel, M. E.; Salvarezza, R. C. Langmuir 2005,
voltage converter, also called zero-resistance amme- 21, 9238–9245.
ter (ZRA), of gain RmGI, as shown in Figure 34(b). In 3. O’M. Bockris, J.; Khan, U. M. Surface Electrochemistry:
A Molecular Level Approach; Plenum Press: New York,
that case, the WE is held at a virtual ground (the same 1993; Chapter 3, pp 310–318.
potential as ground, but not connected directly to 4. Huet, F.; Nogueira, R. P.; Normand, B.; Takenouti, H.
ground) via the OPA of the ZRA. This arrangement In ASM Handbook Corrosion: Fundamentals, Testing and
Protection Cramer, S.D., Covino, B.S., Jr., Eds.;
has the advantage of measuring the current actually ASM International: Materials Park, OH, 2003; Vol. 13A,
flowing across the WE. As a consequence, in specific pp 52–60.
applications involving several WEs, it is possible to 5. Plonski, I. H. Int. J. Hydrogen Energy 1996, 21, 837–851.
6. Epelboin, I.; Morel, Ph.; Takenouti, H. J. Electrochem.
associate a ZRA with each WE to measure the cur- Soc. 1971, 118, 1282–1287.
rent flowing into it, all WEs being at the same poten- 7. Sastri, V. S. Corrosion Inhibitors: Principles and
tial (virtual ground). Applications; Wiley: West Sussex, 1998.
8. Kaesche, H. Corrosion of Metals: Physicochemical
Figure 35 shows a simple way for imposing a Principles and Current Problems; Springer: Berlin, 2003;
constant current Iset with a potentiostat. A resistor Chapter 7, pp 144–158.
Electrochemistry 51
9. Keddam, M.; Mattos, O. R.; Takenouti, H. J. Electrochem. 29. Atkins, P. W. Physical Chemistry, 4th ed.; Oxford
Soc. 1981, 128, 257–265, 266–274. University Press: Oxford, 1990; Chapter 9, pp 225–237.
10. Heusler, K. E. Z. Elektrochem. 1958, 62, 582–587. 30. Landolt, D. In Corrosion Mechanisms in Theory and
11. O’M Bockris, J.; Drazic, D.; Despic, A. R. Electrochim. Practice; Marcus, P., Oudar, J., Eds.; Marcel Dekker:
Acta 1961, 4, 325–361. New York, 1995; pp 13–16.
12. Kelly, E. J. J. Electrochem. Soc. 1965, 112, 124–131. 31. Greef, R.; Peat, R.; Peter, L. M.; Pletcher, D.; Robinson, J.
13. Hilbert, F.; Miyoshi, Y.; Eichkorn, G.; Lorenz, W. J. In Electrochemistry: Southampton Electrochemistry
J. Electrochem. Soc. 1971, 118, 1919–1926, 1927–1935. Group; Ellis Horwood: West Sussex, 1985; Chapter 1,
14. Epelboin, I.; Keddam, M. J. Electrochem. Soc. 1970, 117, pp 26–32.
1052–1056. 32. Bard, A. J.; Faulkner, L. R. Electrochemical Methods:
15. Greef, R.; Peat, R.; Peter, L. M.; Pletcher, D.; Robinson, J. Fundamentals and Applications, Wiley & Sons: Hoboken,
In Electrochemistry: Southampton Electrochemistry 2001; Chapter 9, p 339.
Group; Ellis Horwood: West Sussex, 1985; Chapter 4, 33. Scully, J. R. Electrochemical Methods in Corrosion
pp 139–143. Science and Engineering; Marcel Dekker: New York, 2002;
16. O’M. Bockris, J.; Khan, U. M. Surface Electrochemistry: Chapter 4, pp 125–150.
A molecular Level Approach, Plenum Press: New York, 34. Stern, M.; Geary, A. L. J. Electrochem. Soc. 1957, 104,
1993; Chapter 3, pp 319–348. 56–63.
17. Crow, D. R. Principles and Applications of 35. Marcus, P.; Mansfeld, F. Analytical Methods in Corrosion
Electrochemistry, 4th ed.; Blackie Academic and Science and Engineering; Taylor & Francis/CRC:
Professional: Glasgow, 1994; Chapter 5, pp 68–71. Boca Raton, FL, 2006; Vol. 22: Corrosion Technology.
18. Bard, A. J.; Faulkner, L. R. Electrochemical Methods: 36. Greef, R.; Peat, R.; Peter, L. M.; Pletcher, D.; Robinson, J.
Fundamentals and Applications, 2nd ed.; Wiley & Sons: In Electrochemistry: Southampton Electrochemistry
Hoboken, 2001; Chapter 1, pp 12–14. Group; Ellis Horwood: West Sussex, 1985.
19. Stern, O. Z. Elektrochem. 1924, 30, 508–516. 37. Bard, A. J.; Faulkner, L. R. Electrochemical Methods:
20. Bagotsky, V. S. Fundamentals of Electrochemistry, 2nd ed.; Fundamentals and Applications, 2nd ed.; Wiley & Sons:
Wiley & Sons: Hoboken, 2006; Chapter 4, pp 51–68. Hoboken 2001.
21. Atkins, P. W. Physical Chemistry, 4th ed.; Oxford 38. Bard, A. J.; Faulkner, L. R. Electrochemical Methods:
University Press: Oxford, 1990; Chapter 25, pp 763–769. Fundamentals and Applications, 2nd ed.; Wiley & Sons:
22. Fisher, A. C. Electrode Dynamics, 4th ed.; Oxford Hoboken, 2001; Chapter 2, pp 69–74.
University Press: Oxford 1996; Chapter 2, pp 13–25. 39. Bard, A. J.; Faulkner, L. R. Electrochemical Methods:
23. Bard, A. J.; Faulkner, L. R. Electrochemical Methods: Fundamentals and Applications, 2nd ed.; Wiley & Sons:
Fundamentals and Applications, 2nd ed.; John Wiley and Hoboken, 2001; Appendix C, pp 808–810.
Sons: Hoboken, 2001; Chapter 2, pp 60–62. 40. Bernard, M.; Busnot, F. Usuel de chimie générale
24. Oldham, K. B.; Myland, J. C. Fundamentals of et minérale; Dunod – Bordas: Paris, 1984; Chapter 3,
Electrochemical Science; Academic Press: San Diego, p 205.
1994; Chapter 4, pp 117–120. 41. http://www.consultrsr.com/resources/ref/refpotls.htm.
25. Bagotsky, V. S. Fundamentals of Electrochemistry, 2nd ed.; 42. Newman, J. J. Electrochem. Soc. 1966, 113, 501–502.
Wiley & Sons: Hoboken, 2006; Chapter 3, pp 39–45. 43. Schiller, C.-A. In Analytical Methods in Corrosion Science
26. Compton, R. G.; Sanders, G. H. W. Electrode Potentials; and Engineering; Marcus, P., Mansfeld, F., Eds.;
Oxford University Press: Oxford, 1996; Chapter 1, Taylor & Francis/CRC Press: Boca Raton, FL, 2006;
pp 12–13. Vol. 22: Corrosion Technology, pp. 361–434.
27. Handbook of Chemistry and Physics, 78th ed.; 44. Gabrielli, C. Identification of electrochemical processes by
Chemical Rubber Publishing Company: Boca Raton, frequency response analysis Solartron, Farnborough:
1997; 8-20–8-30. England, 1980; Technical Report no. 4/83; http://www.
28. Pourbaix, M. Atlas of Electrochemical Equilibrium in solartronanalytical.com/downloads/technotes/
Aqueous Solutions; NACE: Houston, 1974. technote04.pdf.
1.03 Outline of Structural Metallurgy Relevant to Corrosion
R. P. M. Procter
CorrosionandProtectionCentre,SchoolofMaterials,,ManchesterM601QD,UK
This article is a revision of the Third Edition article 20.4 by R. P. M. Procter, volume 2, pp 20:94–20:130, ß 2010 Elsevier B.V.
1.03.1 Structure of Pure Metals 53

1.03.1.1 Point Defects in Crystals 54
1.03.1.2 Stacking Faults and Twins 55
1.03.1.3 Line Defects 56
1.03.1.4 Movement and Interaction of Dislocations 57
1.03.1.5 Polycrystalline Metals: Grains and Grain Boundaries 58
1.03.1.6 Annealing: Recovery and Recrystallization 60
1.03.1.7 Macroscopic Defects in Metals 60
1.03.1.8 Structure of Metal Surfaces and Surface Defects 60
1.03.2 Structure of Alloys 61
1.03.2.1 Components and Phases 62
1.03.2.2 Solid Solutions 62
1.03.2.3 Segregation at Dislocations and Grain Boundaries 63
1.03.2.4 Limited and Complete Solid Solubility 63
1.03.2.5 Intermediate Phases and Intermetallic Compounds 63
1.03.3 Equilibrium Phase Diagrams 63
1.03.3.1 Binary Isomorphous Phase Diagrams 64
1.03.3.2 Coring 65
1.03.3.3 Eutectic Phase Diagrams 65
1.03.3.4 More Complex Binary Phase Diagrams 66
1.03.4 The Fe–Fe3C Phase Diagram 66
1.03.4.1 Decomposition of Austenite 66
1.03.4.2 Formation of Pearlite 69
1.03.4.3 Formation of Bainite 70
1.03.4.4 Formation of Martensite 70
1.03.4.5 Isothermal Transformation Diagrams 71
1.03.4.6 Transformation of Hypo-Eutectoid Steels 71
1.03.4.7 Tempering of Martensite 72
1.03.4.8 Spheroidized Structures in Steels 73
1.03.5 Strengthening Mechanisms in Metals 73
1.03.5.1 Precipitation Hardening 74
1.03.5.2 Complex Alloy Systems 75
1.03.5.3 Inclusions in Metals 76
References 76
Glossary Hypo-eutectoid With more solute than the

Binary A mixture of two elements (referring to eutectoid composition.
phase diagrams). Intergranular Between the grains (i.e., at the grain
Hyper-eutectic With more solute than the eutectic boundary).
composition. Intragranular Within the grains (i.e., in the bulk of
Hyper-eutectoid With more solute than the the grain).
eutectoid composition. Pro-eutectoid Formed at above the eutectoid
Hypo-eutectic With more solute than the eutectic temperature.
composition.
52
Outline of Structural Metallurgy Relevant to Corrosion 53
cubic (b.c.c.). The unit cell (i.e., the smallest group of

Quaternary A mixture of four elements (referring to
atoms possessing the symmetry of the crystal and
phase diagrams).
which, when repeated in all directions, develops the
Ternary A mixture of three elements (referring to
crystal lattice) of each of these metallurgically com-
phase diagrams).
mon and important crystal structures is illustrated in
Widmanstätten A lamellar microstructure first
Figure 1; the crystal structure of each of the more
observed in iron meteorites, now used
important metals is given in Table 1. It is clear from
generally to describe a well defined lamellar
this table that the crystal structures of some metals
microstructure.
change with temperature, that is, these metals are
a-ferrite The low temperature b.c.c. allotrope of iron.
polymorphic, and this phenomenon is of very great
d-ferrite The high temperature b.c.c. allotrope of iron.
practical significance, particularly in the case of iron,
as will become apparent later.
The atoms in a metal crystal can often be satisfac-
torily regarded as hard spheres of about 1010 m
Abbreviations radius. A sheet of such hard balls can be most densely
b.c.c. body centered cubic and closely packed together in the configuration
c.p.h. close packed hexagonal (sometimes also shown in Figure 2(a); such sheets or layers of atoms
known as hexagonal close packed, h.c.p.) are therefore known as close-packed planes. Simi-
f.c.c. face centered cubic larly, the directions arrowed in Figure 2(a) are
referred to as close-packed directions. Close-packed
hexagonal and face-centered cubic crystal structures
can be built up by stacking close-packed planes on
The corrosion of a metal invariably involves some
top of each other in the correct sequence. Consider a
kind of interaction between the metal and its envi-
sheet of eight close-packed atoms, designated by the
ronment, and in many cases, the corrosion (location,
full circles a in Figure 2(b); as shown, a second layer
form, and rate) is significantly affected or even caused
of four close-packed atoms, designated by the dotted
by some structural feature of the metal. It is essential,
circles b, may be laid on top of the first plane. Each of
therefore, for the corrosion engineer to have some
the b atoms ‘sits’ in one of the natural ‘valleys’ which
appreciation of the structure of metals; an elementary
occur between any three contiguous a atoms. There
survey is provided in this chapter, which is repro-
are now two such low-energy valleys in the plane of
duced from the 3rd edition.
b atoms, one of which is immediately above an a atom
The structure of pure metals is discussed in this
and one of which is not (Figure 2(b)). An atom
chapter and this is followed by a description of the
placed in the former results in an aba stacking of
structure of alloys. In general, structural features are
three close-packed planes, while an atom placed in
discussed in order of increasing size, that is, starting
the latter would result in abc stacking. Stacking ababa-
with atomic features and continuing through micro-
bab, etc., develops the c.p.h. crystal structure while
scopic features to macroscopic features.
stacking abcabcabc, etc., develops the f.c.c. crystal struc-
ture. The b.c.c. crystal structure cannot be built up by
1.03.1 Structure of Pure Metals stacking close-packed planes on top of each other.
In the f.c.c. and c.p.h. crystal structures, each atom
A crystal may be defined as an orderly three- has 12 nearest neighbors and is therefore said to have
dimensional array of atoms, and all metals are aggre- a coordination number of 12. In the non-close-packed
gates of more or less imperfect crystals. In considering b.c.c. structure, on the other hand, the coordination
the structure of metals, therefore, it is convenient to number is only 8. In the f.c.c. structure, the four
start with the arrangement of atoms in a perfect metal octahedral {111} planes are close packed and so are
crystal and then to proceed to the imperfections that the h110i directions. (An account of the use of Miller
are always present in the crystal structure. indices to describe crystal planes and lattice direc-
Although there are many different crystal struc- tions is beyond the scope of this article; a very ade-
tures, some of which are rather complicated, most quate treatment of this topic is, however, given in
common metals fortunately have one of the three Reed-Hill)1 In the c.p.h. structure, on the other
relatively simple structures: face-centered cubic (f.c.c.), hand, there is only one close-packed plane, the
close-packed hexagonal (c.p.h.), and body-centered (0001) basal plane; the h1120i directions are close
54 Basic Concepts
(a) (b)
(c)
= Atomic sites
Figure 1 Unit cells of (a) the face-centered cubic (f.c.c.), (b) the close-packed hexagonal (c.p.h.), and (c) the body-centered
cubic (b.c.c.) crystal structures.
packed. The different number of close-packed planes and interstitial atoms. As illustrated in Figure 3(a), a
in f.c.c. and c.p.h. crystals results in the very different vacancy occurs where an atom is missing from the
plastic deformation characteristics of metals with the crystal structure while an interstitial occurs when an
two structures. In the b.c.c. structure, the h111i direc- extra atom is inserted in an interstitial site (an inter-
tions are close packed, but there are no truly close- stice). Vacancies and interstitials are equilibrium
packed planes of the type shown in Figure 2(a); the defects because even though vacancies (and intersti-
planes of closest packing, however, are {110}. tials) increase the internal energy of a crystal, they also
In any crystal structure, the close-packed or closest- increase the entropy, and therefore, the free energy
packed planes are the lowest energy planes. On all of a perfect crystal is reduced by the presence of a
other planes, the density of atoms is lower, and the number of point defects. The equilibrium vacancy
interatomic distance and the energy of the plane are concentration increases exponentially with tempera-
greater. Contrary to intuitive expectations, the diame- ture and typically, in copper at room temperature, one
ter of the largest holes or interstices between atoms in site in 1014–15 may be vacant, while at 1300 K, one site
the close-packed f.c.c. structure is considerably greater in 103–4 might be vacant (equivalent to 1017–18 vacan-
than the diameter of the largest interstices between cies per cubic millimeter). The internal energy asso-
atoms in the non-close-packed b.c.c. structure. ciated with an interstitial is much greater than that
associated with a vacancy; the concentration of intersti-
tial atoms at any temperature is correspondingly very
1.03.1.1 Point Defects in Crystals
much lower than that of vacant atomic sites. Radiation
The smallest imperfections in metal crystals are point damage, however, introduces many interstitials. Often,
defects, in particular, vacant lattice sites (vacancies) a vacancy and an interstitial are associated as a pair.
The practical importance of vacancies is that As illustrated in Figure 3(b), this is accompanied by
they are mobile and, at elevated temperatures, can movement of an atom in the opposite direction;
move relatively easily through the crystal lattice. indeed, the existence of vacancies was originally pos-
tulated to explain solid-state diffusion in metals. In
Table 1 Crystal structure of the more important metals order to ‘jump’ into a vacancy, an adjacent atom must
overcome an energy barrier. The energy required for
Metal Temperature Crystal
this is supplied by thermal vibrations. Thus, the dif-
(K) structure
fusion rate in metals increases exponentially with
Alkali metals Ambient to m.p b.c.c. temperature, not only because the vacancy concen-
Aluminum Below m.p. f.c.c. tration increases with temperature, but also because
Beryllium Below m.p. c.p.h.
Cadmium Below m.p. c.p.h.
there is more thermal energy available to overcome
Chromium Below m.p. b.c.c. the activation energy required for each ‘jump’ in the
Cobalt Below 690 c.p.h. diffusion process.
690 to m.p. f.c.c.
Copper Below m.p. f.c.c.
Gold Below m.p. f.c.c. 1.03.1.2 Stacking Faults and Twins
Iron Below 1180 b.c.c.
In certain close-packed metals and alloys there are
1180–1665 f.c.c.
1665 to m.p. b.c.c. errors in the sequence of stacking of the close-packed
Lead Below m.p. f.c.c. planes. Thus, for example, the sequence of planes in a
Magnesium Below m.p. c.p.h. c.p.h. metal might be . . . abababcaba . . ., in which the
Manganese Below m.p. Four complex region in italics is termed a ‘stacking fault’ and repre-
structures
sents a small volume of f.c.c. material in the c.p.h.
Molybdenum Below m.p. b.c.c.
Nickel Below m.p. f.c.c. structure. Stacking faults have an energy that is higher
Platinum Below m.p. f.c.c. than that of the perfect crystal structure; however, the
Silver Below m.p. f.c.c. value of the stacking-fault energy varies very widely
Tin Below m.p. Two complex between different metals and alloys. Stacking faults
structures
are readily visible in thin-foil transmission electron
Titanium Below 1155 c.p.h.
1155 to m.p. b.c.c. micrographs, as illustrated in Figure 4 (top).
Tungsten Below m.p. b.c.c. Twins are a somewhat similar form of defect.
Vanadium Below m.p. b.c.c. In certain metals, adjacent areas of crystal bear a
Zinc Below m.p. c.p.h. mirror-image orientation relationship to each other
Zirconium Below 1135 c.p.h.
across the twin boundary, as illustrated schematically
1135 to m.p. b.c.c.
in Figure 5. The twin boundaries, like stacking faults,
b
a a
b b
a
a a
b
a a
(a) (b)
Figure 2 (a) Single close-packed sheet of atoms, with the close-packed directions arrowed and (b) stacking of three
close-packed planes of atoms, showing aba and abc positions.
56 Basic Concepts
Vacancy Interstitial atom

(a)
Figure 4 (top) Transmission electronmicrograph showing

a stacking fault and isolation dislocations in a stainless steel
(70 000, courtesy I. Brough) and (bottom) light micrograph
showing an equiaxed grain structure and twins in an
annealed 70/30 brass (200, courtesy of M. Islam).
(b)
Figure 3 (a) Schematic illustration of two types of point
defect and (b) three stages in diffusion by the motion of a
vacancy through a crystal.
have a somewhat higher energy than the surrounding

perfect crystal structure. Twins arise during the anneal-
ing of cold-worked f.c.c. material, during the room-
temperature deformation of c.p.h. materials and during
the low-temperature (or shock) deformation of b.c.c.
materials. Twins are often readily apparent on metal-
lographic sections at relatively low magnifications, as
shown in Figure 4 (bottom).
Twinned crystal Twin boundary
1.03.1.3 Line Defects Figure 5 Twinned crystal.
Point defects, stacking faults, and twins are crystalline

imperfections that basically involve misalignment of Dislocations are extremely important defects, in that
single atoms or of a single plane of atoms. Line defects, the dislocation density and dislocation interactions con-
or dislocations, are somewhat larger-scale imperfec- trol the mechanical properties to a very large extent, in
tions involving misalignment along a line of atoms. particular, the strength and ductility of metals.
The simplest type of line defect is the edge disloca- Figures 7(a)–7(c) illustrate the slip or glide of an
tion, which consists of an extra-half plane of atoms in edge dislocation through a crystal under the influ-
the crystal, as illustrated schematically in Figure 6(a); ence of a shear stress. The atom a in Figure 7(a)
edge dislocations are often denoted by ? if the extra- moves, under the influence of the applied stress, to
half plane ab is above the plane sp or by > if it is below. position a0 in Figure 7(b), and the extra-half plane of
The second type of line defect is the screw dis- atoms shifts one atomic distance to the left, from
location, which is somewhat less easy to visualize. plane y to x. As this process continues, the dislocation
Consider, however, a block of material, half of which moves right through the crystal, the final result,
is sheared one interatomic distance with respect to shown in Figure 7(c), being that the crystal is
the other half, as shown in Figure 6(b). The line cd sheared or slipped by one interatomic distance across
then constitutes a screw dislocation; the arrangement the slip plane sp. Each step in the movement of the
of atoms around a screw dislocation is shown in dislocation (i.e., from Figures 7(a) and 7(b)) clearly
Figure 6(c). requires only a slight rearrangement of the atoms
To a good approximation, only atoms within the around the bottom of the extra-half plane, and is
dotted circles in Figures 6(a) and 6(b) are dis- therefore a low-energy process; dislocations are,
placed from their equilibrium position; in a real three- therefore, readily mobile under low applied stresses.
dimensional crystal, the diameter d of these circles However, producing the slip, shown in Figure 7(c),
would be very much less than the length l of the dislo- from a perfect dislocation-free crystal would be a
cation, that is, the length, perpendicular to the page, of very high-energy process requiring very high stres-
the extra-half plane of atoms ab in Figure 6(a), or of the ses. The difference between these two modes of slip is
line cd in Figure 6(b). Strictly therefore, dislocations that in the latter, all the bonds on the slip plane are
are cylindrical defects of diameter d and length l; how- broken at the same time, while in the former, they
ever, since l d, they are referred to as line defects. are broken one at a time. The existence of disloca-
tions was in fact postulated, long before they were
observed, to account for the observation that the
1.03.1.4 Movement and Interaction of
yield stress of metals is very much lower than that
Dislocations
predicted by theoretical calculations. It is important
As with vacancies, the importance of dislocations to note that the crystal structure, after passage of a
derives from the fact that they are readily mobile, dislocation, is perfect and in no way misaligned. From
in this case under the influence of applied stresses. Figure 8, it is clear that the dislocation line is merely
s b p d
c
(a) (b)
(c)
Figure 6 (a) Atomic arrangement around an edge dislocation, (b) shear required in a homogeneous solid to produce a
screw dislocation, and (c) atomic arrangement around a screw dislocation.
58 Basic Concepts
x y x y vector and the line of the dislocation. In practice, it

is observed that metals slip most easily on close-
packed planes and in close-packed directions, that
is, {111} and <110> in f.c.c. metals and usually
{0001} and <1120> in c.p.h. metals. Since f.c.c. crys-
a a′ tals have four close-packed {111} planes, they are
generally much more ductile than c.p.h. crystals,
which have only one {0001} close-packed plane. In
b.c.c. metals, slip occurs in <111> directions on a
(a) (b)
variety of planes. The movement of dislocations by
x y other mechanisms, in particular by cross-slip and
climb, is beyond the scope of this section.
All real metals contain dislocations; even a
well-annealed metal would typically contain 104–6
s p dislocations per square millimeter, while a heavily
cold-worked metal could contain up to 1010 mm2.
At first sight, this is an anomaly; dislocations were
postulated to account for the low yield strength of
(c) metals, and whereas an annealed material with a low
Figure 7 Three stages in plastic deformation by the dislocation density is weak, a cold-worked metal with
motion of an edge dislocation through a crystal. a high dislocation density is strong. The answer lies in
the fact that when the dislocation density is low, the
dislocations are generally too far apart to interact
with each other very often and are more free to
move under the influence of a low applied stress.
On the other hand, when the dislocation density is
high, the dislocations do interact, and they become
tangled up with each other; dislocation motion is
then difficult and the material is therefore strong.
The interaction of dislocations with each other and
with other structural features in metals is a very
Pure edge complex field; it is also, however, extremely impor-
dislocation tant, as it greatly affects the strength of the metals.
Pure screw The atomic bonds in a cylinder of material around
dislocation dislocations are elastically stretched; dislocations,
like other crystalline defects are, therefore, high-
Slipped crystal
Unslipped crystal energy regions.
Dislocations are readily visible in thin-film transmis-
Figure 8 Schematic illustration of a mixed dislocation
as the boundary between slipped and unslipped crystal.
sion electron micrographs, as shown in Figures 4 (top)
The arrow shows the Burgers vector. and 9 (top). The slip step (Figure 7(c)) produced by the
passage of a single dislocation is not readily apparent.
However, for a variety of reasons, a large number of
the boundary between slipped and unslipped crystal. dislocations often move on the same slip plane or on
The magnitude and direction of the slip is given by bands of closely adjacent slip planes; this results in slip
what is known as the Burgers vector; in a screw steps that are very easily seen in the light microscope, as
dislocation, the Burgers vector is parallel to the line shown by the slip lines in Figure 9 (bottom).
of the dislocation, while in an edge dislocation, it is
perpendicular to it. In real crystals, dislocation lines
1.03.1.5 Polycrystalline Metals: Grains and
are not straight, but are usually bowed or curved. As
Grain Boundaries
illustrated schematically in Figure 8, such disloca-
tions are mixed, being part screw and part edge. Slip So far, we have regarded metals as single crystals; in
occurs on the plane containing both the Burgers fact, most metals, in practice, are polycrystalline and
Figure 10 Light micrograph showing the highly

elongated grain structure of a commercial wrought
high-strength precipitation-hardening Al–Zn–Mg–Cu alloy
(40, section perpendicular to the long transverse direction).
deformed and elongated in the work direction, giving

the highly directional, elongated, grain structure
shown in Figure 10. Although the crystallographic
orientation of all the grains in a polycrystalline metal
is usually different and random, occasionally the
orientation of most of the grains in a piece of metal
is similar. Such preferred orientation can arise in
Figure 9 (top) Transmission electronmicrograph showing
dislocation tangles associated with precipitates in an a number of ways; it is important to note, however,
Al–Cu–Mg–Si alloy (24 000, courtesy S. Blain) and (bottom) that metals with elongated, columnar grain struc-
light micrograph showing slip lines in pure lead (100). tures (Figure 10) do not necessarily have preferred
orientation.
However perfect the crystal structure within a
consist of a number of crystals or grains. In a pure grain, there will inevitably be crystallographic mis-
metal, each grain has the same crystal structure and match and imperfection where two adjacent grains
contains vacancies, dislocations, etc., but the crystal- meet, that is, at grain boundaries; this region of mis-
lographic orientation of each grain is different; in match is, however, only a few atomic diameters wide,
other words, in a polycrystalline metal, the [111] as shown schematically in Figure 11, and on metal-
crystallographic direction of each grain points in a lographic sections, grain boundaries only become
different spatial direction. Typically, the diameter of apparent on etching. Although grain boundaries are
the grains in a coarse-grained material might be regions of imperfection and of crystallographic mis-
about 0.5 mm, while that in a fine-grained metal match, they are nevertheless quite strong, and metals
might be about 0.005 mm. However, in certain elec- do not normally fail intergranularly. Indeed, grain
trodeposits, the grain size is too fine to be resolved boundaries act as barriers to the movement of dis-
with the light microscope, while at the other extreme, locations, and a fine-grained pure metal will nor-
jet-engine turbine blades about 150 mm long, of mally be considerably stronger than a single-crystal
nickel-base superalloy, may sometimes be single or a coarse-grained material.
crystals. The grain size of metals may be controlled A grain boundary is a planar defect in a crystal and
by heat treatment. All the grains may be of more as such has a higher energy than the surrounding
or less the same size and shape as in the equiaxed crystal. By way of comparison, the free surface energy
grain structure shown in Figure 4 (bottom). Alterna- of copper is about 1200–1800 nJ mm2, the grain
tively, if the metal has been heavily cold-worked, for boundary energy is about 300–500 nJ mm2, while
example by rolling or extrusion, the grains may be the stacking-fault energy is about 80–100 nJ mm2.
60 Basic Concepts
handling or operating. Examples of these would be

sharp re-entrant corners, crevices, casting defects,
rolled-in mill-scale, rough-sheared edges, poorly
machined surfaces, weld defects, surface scratches, etc.
However, by far the most important macroscopic
defect, which is inherent in the uses to which metals
are put, and the ways in which they are formed, is
the presence of stresses, particularly tensile stresses.
These stresses may be either applied, resulting from
the fact that the metal is being used in a load-bearing
Grain A Grain B capacity, or residual. Residual stresses may either be
microstresses arising, for example, from dislocations
Grain boundary and piled up against a grain boundary or, much more
adjacent crystallographic mismatch importantly, macrostresses. Residual macrostresses
Figure 11 High-angle grain boundary. arise from a very wide range of effects, for example,
forming, heat treating, machining, welding, and assem-
bling operations can all give rise to high residual tensile
stresses over relatively large regions.
1.03.1.6 Annealing: Recovery and
Recrystallization
1.03.1.8 Structure of Metal Surfaces and
It must be emphasized that many of the crystalline
Surface Defects
defects already discussed are metastable and tend to
be eliminated on annealing at elevated temperatures. So far the structure of pure metals has been discussed
Thus, when a cold-worked metal is heated, the with reference to bulk characteristics and continuous
energy stored in the metal during the cold-working crystals. However, corrosion is essentially a surface
process is released; the dislocation tangles rearrange phenomenon, and it is necessary to consider how the
themselves, and the internal stresses, hardness, and structure and defects already described interact with
strength of the metal are progressively reduced. The free surfaces. At this stage, it is convenient to consider
changes occurring during this process (which is only a film-free metal surface, although of course, in
known as recovery) do not affect the optical micro- most corrosion phenomena the presence of surface
structure of the metal. Recovery is followed at higher films is of the utmost importance. Furthermore, it is
temperatures by recrystallization. New grains are at free surfaces that the hard sphere model of metals
nucleated throughout the metal and grow; the heavily first begins to become inadequate, particularly when
deformed high-dislocation-density cold-worked grains surface films are considered, or in aqueous environ-
are replaced by new equiaxed low-dislocation-density ments, where an electrical double layer exists.
grains. On prolonged heating at even higher tem- Figure 12(a) shows schematically a smooth metal
peratures, some of these grains grow at the expense surface, although it should always be borne in mind
of others, and a coarse-grained material (or even a that even a high-quality electropolished surface will
single crystal) replaces the fine-grained recrystallized not be flat but will have a roughness of about 5 nm or
structure. The changes occurring during recrystal- so high, while the roughness on a high-quality
lization are readily apparent in the optical micro- machined surface may be about 1 mm or so high. In
scope and are accompanied by a drastic softening of most real materials, there will be grain boundaries
the metal. meeting the surface as shown in Figure 12(a). Fur-
thermore, it is extremely unlikely that the plane of
the metal surface will coincide with a low-index low-
1.03.1.7 Macroscopic Defects in Metals
energy high-atomic-density close-packed plane, as in
While vacancies, dislocations, etc. may be regarded grain B. In most cases, the plane of the metal surface
as atomic defects, grains may be regarded as micro- will be a high-index high-energy low-atomic-density
scopic defects. There are also, however, what may con- plane, as in grains A and C.
veniently be termed macroscopic defects in metals. On an atomic scale, metal surfaces are usually
Too often these are the result of poor design, poor described in terms of the terrace–ledge–kink model
processing or production, poor welding, careless shown in Figure 12(b). Essentially, it is suggested that
A C
B
Grain boundaries
(a)
Ledge
Ad - atom
Terrace
Kink Emergent edge

dislocation
Surface vacancy
Emergent
screw dislocation
(b)
Figure 12 (a) Grains of different orientation intersecting a free surface and (b) the terrace–ledge–kink model of a free
metal surface.
clean metal surfaces consist of low-energy low-index As shown in Figure 12(b), an edge dislocation
terraces separated by ledges of monatomic height, intersecting a free surface does not cause a step.
which occasionally contain monatomic kinks. However, the atoms adjacent to the core of the dislo-
By their nature, dislocations cannot end suddenly in cation, being in nonequilibrium positions, have a
the interior of a crystal; a dislocation line can only end higher energy than other atoms in the surface. The
at a free surface or a grain boundary (or form a closed emergent ends of both edge and screw dislocations
loop). Where a screw dislocation intersects a free sur- thus represent potential sites for preferential dissolu-
face, there is inevitably a step or ledge in the surface, tion. Finally, there may be vacancies in the surface
one atomic layer high, as shown in Figure 6(c). Fur- and ad-atoms on the surface (Figure 12(b)).
thermore, the step need not necessarily be straight and
will, in fact, almost certainly contain kinks.
The coordination number of an atom lying in a 1.03.2 Structure of Alloys
(111) surface of an f.c.c. crystal is 9, compared with
the bulk value of 12; an atom at a step in the surface The discussion so far has been limited to the struc-
has a co-ordination number of 7 while that of an atom ture of pure metals and to the defects which exist in
at a kink in a step is only 6. The latter atom, therefore, crystals composed of atoms of one element only. Of
is the most weakly bound to the metal crystal and is course, pure metals are comparatively rare, and all
likely to be the first to be removed from the lattice in commercial materials contain impurities, including
any active dissolution process. Clearly, however, this in many cases, deliberate alloying additions. In the
removal does not eliminate the kink and, in fact, it is production of commercially pure metals and alloys,
impossible, even by repeated removal of atoms, to impurities are inevitably introduced into the metal,
eliminate the step caused by a screw dislocation for example, manganese, silicon, and phosphorus in
intersecting a free surface. There are, however, in mild steel and iron and silicon in aluminum alloys.
addition to screw dislocations, many other causes of However, most commercial materials are not even
steps or ledges on metal surfaces, for example, low- nominally pure metals but are alloys in which delib-
index planes intersecting the surface at an acute erate additions of one or more elements have been
angle (see grain A in Figure 12(a)). made, usually to improve some property of the metal;
62 Basic Concepts
examples are the addition of carbon or nickel and mixture of the two atomic species A and B, which is
chromium to iron to give, respectively, carbon and known as a solid solution of B in A. The solute B atoms
stainless steels and the addition of copper to alumi- may take up either interstitial or substitutional posi-
num to give a high-strength age-hardenable alloy. tions with respect to the solvent atoms A, as illustrated
in Figures 13(a) and 13(b), respectively. Interstitial
solid solutions are only formed with solute atoms that
1.03.2.1 Components and Phases
are much smaller than the solvent atoms, as is obvious
Impure metals and alloys exhibit all the structural from Figure 13(a); for the purpose of this section, only
features and crystal defects of the pure metals already three interstitial solid solutions are of importance, that
discussed. In addition, however, impure metals and is, Fe–C, Fe–N, and Fe–H. On the other hand, the
alloys exhibit many structures that are not observed solid solutions formed between two metals, as for
in pure metals and which, in many instances, have an example in Cu–Ag and Cu–Ni alloys, are nearly
extremely important effect on the properties, partic- always substitutional (Figure 13(b)). Occasionally,
ularly the corrosion resistance. However, before deal- substitutional solid solutions are formed in which the
ing with the structure of impure metals and alloys, it distribution of B in A is not random (Figure 13(c)),
is necessary to consider the concept of metallurgical and such solid solutions are said to be ordered and
components, phases, constituents, and equilibrium exhibit a superlattice. Order–disorder transitions are
phase diagrams. observed with temperature changes. Superlattices are
The fundamental difference between pure metals observed in the Cu–Au, Cu–Zn, and Fe–Ni systems.
and impure metals and alloys arises from the fact that There are a number of differences between inter-
there is only one atomic species present in the former, stitial and substitutional solid solutions, one of the
while there are two or more present in the latter; most important of which is the mechanism by which
thus, a pure metal is a single-component system, a diffusion occurs. In substitutional solid solutions, dif-
pure binary alloy is a two-component system, while fusion occurs by the vacancy mechanism already dis-
impure metals and alloys, strictly speaking, are multi- cussed. Since the vacancy concentration and the
component systems. It is, however, usually valid to frequency of vacancy jumps are very low at ambient
neglect the impurities and, for example, to consider a temperatures, diffusion in substitutional solid solu-
commercial brass as a binary Cu–Zn alloy, that is, as a tions is usually negligible at room temperature and
two-component system. only becomes appreciable at temperatures above
Gibbs’ phase rule, as applied to metallurgical sys- about 0.5TM, where TM is the melting point of the
tems (i.e., under conditions of constant pressure) solvent metal (K). In interstitial solid solutions, how-
indicates that pure metals, being single-component ever, diffusion of the solute atoms occurs by jumps
systems, can only coexist as two phases in equilibrium between adjacent interstitial positions. This is a
at a single temperature. Thus pure copper, for exam- much lower energy process that does not involve
ple, exhibits three phases, namely solid f.c.c. copper,
liquid copper, and gaseous copper. Any two of these
phases, say solid and liquid copper, can coexist in
equilibrium at only one temperature, that is, the
melting point of copper. Pure iron exhibits four
phases, namely two solid phases (f.c.c. and b.c.c. iron
(see Table 1)) and a liquid and a gaseous phase.
Alloys, on the other hand, being multicomponent (a) (b)
systems, can exhibit two or more phases in equilib-
rium over a range of temperatures. The phases which
are present in an alloy and their composition and
distribution markedly affect many of the properties
of the alloy, including its corrosion resistance.
(c)
1.03.2.2 Solid Solutions
Figure 13 (a) Interstitial solid solution, (b) random
When a pure metal A is alloyed with a small amount of substitutional solution, and (c) an ordered substitutional
element B, the result is ideally a homogeneous random solid solution forming a superlattice.
vacancies, and it therefore occurs at much lower element to attract electrons to itself when present
temperatures. Thus, hydrogen is mobile in steel at in a molecule or in an aggregate of unlike atoms;
room temperature, while carbon diffuses quite rap- it is a different property from the electrode
idly in steel at temperatures above about 370 K. potential, which depends on the free energy dif-
ference between an element in its standard state
and a compound or ion in solution.) In addition, a
1.03.2.3 Segregation at Dislocations and metal of a lower valency tends to dissolve a metal
Grain Boundaries of a higher valency more readily than vice versa.
In practice, the distribution of solute atoms in a solid However, just as two liquids may be completely mis-
solution is not ideally homogeneous. There is almost cible and form a complete range of solutions from
invariably a stress field around any solute atom in a one pure liquid to the other, so certain metals, for
solid solution, and there is therefore, a tendency example, copper and nickel, exhibit complete solid
for these atoms to migrate to regions of the crystal solubility over the whole range of compositions from
lattice that are already strained, in particular to grain pure copper to pure nickel. Clearly, for two metals to
boundaries and to dislocations. Thus, in impure be soluble in each other over the whole composi-
metals and alloys, there is an inherent tendency for tional range, they must have the same crystal struc-
impurities and alloying additions to segregate at grain ture, that is, they must be isomorphous.
boundaries and around the cores of dislocations.
Segregation of carbon and nitrogen to dislocations in
1.03.2.5 Intermediate Phases and
steels has a very significant effect on their mechanical
Intermetallic Compounds
properties and is particularly responsible for their
sharp yield point; similar segregation may also affect When metal A is alloyed with enough element B
the corrosion behavior of cold-worked steels. The to exceed the solid solubility of B in A, a new phase
segregation of impurities and alloying elements to is formed. This may be the other terminal solid
grain boundaries is clearly of relevance in any con- solution of A in B. Alternatively, an intermediate
sideration of the strength and reactivity of grain solid-solution phase, with a crystal structure different
boundaries and, in particular, of intergranular corro- from that of either of the terminal solid solutions,
sion, temper embrittlement, etc. may be formed over a range of compositions. An
In addition, it has fairly recently been recognized example of this is the so-called b-phase in brasses,
that impurities and alloying elements will also tend to which is the stable phase at room temperature over the
segregate to free surfaces. The implications of this for composition range 47–50% Zn. In some alloys, inter-
corrosion resistance and particularly for passive-film mediate phases are formed that are best regarded as
formation have received relatively little attention. compounds. Examples are Fe3C (cementite) in the
Fe–C system and the intermetallic compounds MgNi2
and Mg2Ni formed in the Mg–Ni system. While solid
1.03.2.4 Limited and Complete Solid solutions are stable over a range of compositions, com-
Solubility pounds have a unique composition; cementite, for
Just as the saturated solubility of sugar in water is example, corresponds to Fe–25 at.%C (Fe–6.7 wt%C).
limited, so the solid solubility of element B in metal A
may also be limited, or may even be so low as to be
1.03.3 Equilibrium Phase Diagrams
negligible, as for example with lead in iron or carbon
in aluminum. There is extensive interstitial solid
In alloys, as in pure metals, the stable phase or phases
solubility only when the solvent metal is a transition
change with changing temperature. Furthermore, in
element and when the diameter of the solute atoms
alloys, the stable phase or phases, and the composi-
is <0.6 of the diameter of the solvent atom. The
tion of these phases, changes with the composition of
Hume-Rothery rules state that there is extensive
the alloys. These changes are best represented by
substitutional solid solubility of B in A only if:
equilibrium phase diagrams. These define the stabil-
(a) The atomic diameters of A and B do not differ by ity and composition of the phases that can occur in
more than 15%. an alloy system as a function of temperature and
(b) The two elements have similar electronegativity. alloy composition, under conditions of constant pres-
(Note: electronegativity is the power of an sure. Phase diagrams are in many ways analogous to
64 Basic Concepts
Pourbaix diagrams, which define the stability of deal with the various types of binary phase diagrams
phases in a metal/water system as a function of by referring to some specific alloys. The simplest
potential and pH, and like E–pH diagrams, phase phase diagram refers to an isomorphous alloy system
diagrams have certain limitations. First, it must be in which there is complete solid solubility. By way
emphasized that they are equilibrium phase diagrams of example, the Cu–Ni phase diagram is shown in
and that they only predict the stability and composi- Figure 14. This phase diagram indicates that at
tion of phases under equilibrium conditions. They 1600 K, an alloy containing 70% Cu is a liquid
neither give any indication of how rapidly equilib- (point a in Figure 14). On slow cooling, this alloy
rium will be achieved in practice, nor do they give starts to solidify when the liquidus line is reached,
any indication of the distribution of the phases at that is, point b at about 1500 K. At this point, grains of
equilibrium. Somewhat analogously, E–pH diagrams a solid solution containing about 53% Cu are nucle-
give no information about the rate at which a partic- ated from the liquid containing 70% Cu and start
ular metal will corrode or the protectiveness of a to grow. As the temperature continues to fall, the
particular passive film. Second, equilibrium phase composition of the liquid follows the liquidus line
diagrams give no information whatsoever about the while that of the solid falls along the solidus line,
existence of metastable phases, some of which are of until at 1470 K the alloy consists of a mixture of a
the utmost practical significance. These limitations solid solution containing 62% Cu and a liquid con-
become particularly important under conditions of taining 78% Cu (Figure 14). Furthermore, the frac-
rapid temperature change. Phase diagrams, in contrast tion of solid in the two-phase mixture is given by the
to E–pH diagrams, are always determined experimen- ratio of the lengths yz/xz, while the fraction of liquid
tally, although they can be explained, and to a certain is given by xy/xz. This is known as the lever rule.
extent, qualitatively predicted from thermodynamic On further slow cooling, the solidus line is reached,
data and considerations. Nevertheless, in spite of that is, point c at about 1430 K, when solidifica-
these limitations, phase diagrams are an extremely tion of the alloy is complete; the last liquid to
important and useful metallurgical tool. freeze contains about 85% Cu, while the last solid
to solidify contains 70% Cu. At 1300 K, for exam-
ple, and indeed at all temperatures below the soli-
1.03.3.1 Binary Isomorphous Phase
dus, an alloy containing 70% Cu is a solid (point d in
Diagrams
Figure 14). Clearly the phase change in this trans-
Although ternary and quaternary phase diagrams are formation is liquid ! solid. The microstructure of a
beyond the scope of this section, it is appropriate to homogeneous single-phase solid-solution alloy is
Liquid phase
m.p. of stable
pure Ni
1700 Liquidus line
Temperature (K)
1600 a
Solidus
line
1500 Solid + liquid b

phases stable y
x z
c
1400 62 %
53 %
Solid phase m.p. of
78 %
stable pure Cu
85%
1300 d
0 20 40 60 80 100
Cu (wt.%)
Figure 14 Nickel–copper phase diagram.
indistinguishable from that of a pure metal, as shown TE the eutectic temperature, and CE the eutectic
in Figure 4 (bottom). composition, in this case approximately 520 K and
13% Sb, respectively. Like a pure metal, an alloy of
eutectic composition freezes at a single temperature
1.03.3.2 Coring
TE, instead of over a temperature range, as in the case
If the alloy is cooled sufficiently slowly, the alloy of the Cu–30Ni alloy already discussed. Unlike a
itself and all the grains in it will be homogeneous pure metal, however, a eutectic alloy freezes to give
after solidification. If, on the other hand, the alloy a mixture of two phases, in this case Pb and Sb; in
is cooled more rapidly, the grains will be cored, general terms, therefore, a eutectic reaction involves
not homogeneous. During cooling, the composition liquid ! solid a þ solid b.
changes in the solid phase involve solid-state diffu- A hypereutectic alloy containing, say, 50% Sb
sion; if the temperature falls too rapidly, therefore, starts to freeze when the temperature reaches the
there is insufficient time for diffusion. In the Cu– liquidus line (point a in Figure 15). At this tempera-
30Ni alloy referred to earlier, clearly the center of the ture, pure proeutectic Sb nucleates; as the tempera-
grains, the first metal to have solidified, would be ture continues to fall, more antimony is deposited
nickel rich (i.e., low in copper) while the grain bound- from the melt, and the composition of the liquid
ary regions, the last metal to have solidified, would be phase moves down the liquidus line to the eutectic
copper rich. Coring is eliminated by annealing at point. When this is reached, the remainder of the
high temperatures to permit diffusion and homoge- melt solidifies. The microstructure of alloys of eutec-
nization to take place. tic composition varies somewhat with the alloy sys-
tem, but generally consists of an aggregate of small
particles, often platelets, of one of the phases com-
1.03.3.3 Eutectic Phase Diagrams
prising the eutectic in a continuous matrix of the
The equilibrium phase diagram for Pb–Sb alloys other phase. Finally, the microstructure of the hyper-
represents a simple eutectic system and is shown in eutectic 50% Sb alloy already mentioned consists
Figure 15. This type of phase diagram is often of primary (proeutectic) antimony surrounded by
observed when there is negligible solid solubility of the eutectic constituent, as illustrated in Figure 16.
the two metals in each other, and no intermediate This microstructure is clearly to be expected from
phases are formed. The point E is the eutectic point, the solidification sequence just described.
900 m.p of
Liquidus line pure Sb
Liquid
800
Temperature (K)
a
700
Liquid + Pb
m.p of Liquid + Sb
600
pure Pb
E Solidus line
TE 520 k
500
Pb + Sb
CE
400 13%
0 20 40 60 80 100
Sb (wt.%)
Figure 15 Lead–antimony phase diagram.
66 Basic Concepts
equilibrium diagram, in which graphite appears in place

of cementite, is of use only when cast irons are being
considered, as opposed to steels. Other phase diagrams
that are of particular significance are the iron-rich ends
of the Fe–Ni and Fe–Cr systems, which are shown in
Figures 21(a) and 21(b), respectively. It follows from
the diagrams that Ni and C are austenite (g-phase)
stabilizers while Cr is a ferrite (a-phase) stabilizer.
The fact that the interstices in an f.c.c. structure are
bigger than those in a b.c.c. structure accounts for the
fact that C is much more soluble in f.c.c. iron (austen-
ite) than in b.c.c. iron (ferrite) (Figure 20).
Figure 16 Light micrograph showing the microstructure 1.03.4 The Fe–Fe3C Phase Diagram
of a hyper-eutectic Pb–50Sb alloy (50).
Because of the great practical importance of steels, it
is necessary to discuss the Fe–C system in rather
1.03.3.4 More Complex Binary Phase
more detail. The Fe–Fe3C phase diagram consists of
Diagrams
the following three reactions
Most phase diagrams, however, are more complex than
1. At 1766 K and 0.16% C the peritectic reaction:
those shown in Figures 14 and 15. Thus, in most
liquid Fe þ d Fe (b.c.c.) ! g Fe (f.c.c.).
eutectic systems, there is some appreciable solid
2. At 1420 K and 4.3% C the eutectic reaction: liquid
solubility, as in the Ag–Cu phase diagram shown in
Fe ! g Fe (f.c.c.) þ Fe3C.
Figure 17(a). At ambient temperatures, therefore,
3. At 996 K and 0.80% C the eutectoid reaction:
an Ag–50Cu alloy would contain two components
g Fe (f.c.c.) ! a Fe3C.
(Ag and Cu), two phases (the a and b solid solutions),
and two constituents (primary proeutectic b and the The peritectic transformation generally has little
eutectic mixture of a þ b). Furthermore, in many alloy effect on the structure, properties, or corrosion resis-
systems intermediate phases and/or intermetallic tance of steels at room temperature; an exception
compounds are formed. Figure 17(b), for example, to this occurs in the welding of certain steels, when
illustrates the Mg–Sn phase diagram; this exhibits two d-ferrite can be retained at room temperature and
eutectic reactions and the intermetallic compound, can affect corrosion resistance. Furthermore, since
Mg2Sn. Finally, a number of other reactions, in addi- most steels contain less than about 1.0%C (and by
tion to the eutectic reaction, are observed; of particular far the greatest tonnage contains less than about 0.3%
importance are (on cooling in all cases): C), the eutectic reaction is of relevance only in rela-
tion to the structure and properties of cast irons,
1. The eutectoid reaction: solid a ! solid b þ solid g.
which generally contain 2–4%C. This discussion,
2. The peritectic reaction: liquid þ solid a ! solid b.
therefore, will be limited to the eutectoid reaction
3. The peritectoid reaction: solid a þ solid b ! solid g.
that occurs when homogeneous austenite is cooled.
By way of example, the Cu–Zn phase diagram shown
in Figure 18 exhibits a number of different interme-
1.03.4.1 Decomposition of Austenite
diate phases (b, g, d, etc.) and a number of peritectic
reactions and a eutectoid reaction. In many instances, When austenite is cooled, the structures that result
it is not necessary to consider a complete phase from the eutectoid reaction depend on the tempera-
diagram. Thus Figure 19 illustrates the Al-rich end ture at which the transformation takes place. In prac-
of the Al–Cu phase diagram and is used below in a dis- tice, of course, steels are generally continuously
cussion of precipitation hardening. Again, Figure 20 cooled at a rate that may vary from very rapid
shows the Fe–Fe3C phase diagram, the great prac- (quenching) to very slow (furnace cooling). However,
tical importance of which is in no way reduced by it is useful and valid to simplify the situation by
the fact that one of the components in it, namely assuming that the transformation occurs isothermally.
Fe3C (cementite), is a metastable compound. The true In other words, therefore, we shall consider initially,
1400
m.p. of
Pure Cu
1300
Liquid
m.p. of Liquidus
Pure Ag line
1200
Temperature (K)
Liquid + α
Solidus
line
1100 Liquid + β
E
TE 1052K β
α
1000 CE
28 %
Solidus line
α+β
900 Primary β + eutectic
(a) 0 20 40 60 80 100
Cu (wt.%)
m.p. of
Mg2Sn
1100
1000 Liquid Liquid
m.p. of
Pure Mg 900
α + Liq. Mg2Sn + liquid
α E
Temperature (K)
800
700
Sn + liquid
600
Mg2Sn + liquid
Mg2Sn m.p. of
500 Pure Sn
E2
α + Mg2Sn
400
Sn + Mg2 Sn
(b) 0 20 40 60 80 100
Sn (wt.%)
Figure 17 (a) Silver–copper phase diagram and (b) magnesium–tin phase diagram.
68 Basic Concepts
1400
Liquid
α + liquid
β + liquid
1200
γ + liquid
Temperature (K)
1000 β δ + liquid
γ
α δ ε + liquid
800
η + liquid
β
α +
+ γ ε
γ +
β ε
ε + η
600 η
0 20 40 60 80 100
Zn (wt.%)
Figure 18 Copper–zinc phase diagram.
1000 Liquid
m.p. of
pure Al 900
α + liquid 821 K

800 •
α 5.7%
Temperature (K)
700
600
500
α + CuAl2
400
300
0 2 4 6 8 10
Cu (Wt.%)
Figure 19 Aluminum-rich end of the aluminum–copper phase diagram.
the reactions that occur when a steel containing 0.8% takes place. In general, the structures produced
C (i.e., of eutectoid composition) is very rapidly on slow continuous cooling approximate to those
quenched from, say, 1400 K to some temperature resulting from high temperature isothermal transfor-
T < 996 K, at which decomposition of the austenite mation, while rapid continuous cooling results in
m.p. of
pure Fe δ + liquid
1807 K Peritectic point
1800 (1766 K , 0.16%)

0.1%
1700 Liquid
1663 δ + γ 0.51%
1600
γ + liquid
1500
Liquid + Fe3C
1420 K
1400 γ
Temperature (K)
2.06%
Eutectic
1300 point (4.3%)
1200
1183
γ + Fe3C
1100
α+γ
996 K
1000
0.02%
900 α Eutectoid
point (0.8%)
800 α + Fe3C
700
0 1 2 3 4 5
C (wt.%)
Figure 20 Iron-rich end of the metastable Fe–Fe3C phase diagram.
structures that approximate to those obtained by Figure 22(a). At some later time, nucleation of a num-
intermediate-temperature isothermal transformation. ber of colonies of pearlite will have occurred at the
austenite grain boundaries, as shown in Figure 22(b);
at this stage, about 10% of the austenite has trans-
1.03.4.2 Formation of Pearlite
formed. At some still later time, growth of the pearlite
If austenite is allowed to transform just below 996 K, will have occurred by nucleation of new a-Fe and
the reaction product, as predicted by the Fe–Fe3C Fe3C platelets and by extensive edge-wise growth of
phase diagram, is the a Fe þ Fe3C eutectoid. existing platelets as shown in Figure 22(c); at this
This eutectoid constituent is known as pearlite and stage, about 90% of the austenite has transformed
consists of alternate platelets of ferrite (a-Fe) and and eventually, transformation will be complete.
cementite (Fe3C), the former being the continuous In fundamental terms, the transformation temper-
phase. The decomposition of austenite to pearlite ature affects both the driving force for the decompo-
involves nucleation of pearlite, which occurs almost sition of austenite and the diffusion rate of carbon.
exclusively at austenite grain boundaries, followed by In effect, therefore, the transformation tempera-
growth of pearlite; both processes involve the solid- ture alters both the rate of nucleation and the rate
state diffusion of carbon in iron. Immediately after of growth. This in turn manifests itself in differ-
quenching, therefore, the structure of the steel consists ent transformation times at different temperatures.
of untransformed austenite, as shown schematically in Furthermore, the spacing of the pearlite platelets is
70 Basic Concepts
Liquid
δ + liquid
Liquid 1800 α + liquid
1800
δ
δ+γ
1600
1600 γ + liquid
Temperature (K)
Temperature (K)
α+γ
1400
1400
γ
α
γ 1200
1200
1000
1000
α α+γ
0 10 20 30 0 10 20 30
Ni (wt.%) Cr (wt.%)
(a) (b)
Figure 21 (a) Iron-rich end of the iron–nickel phase diagram and (b) iron-rich end of the iron–chromium phase diagram.
characteristic of the temperature at which transfor- visible. The microstructure of a bainitic steel is shown
mation occurred; the lower the transformation tem- schematically in Figure 22(f ); the appearance of bai-
perature the finer are the platelets, and the greater the nite, however, and the structure and composition of
hardness and the strength of the steel. The distinction the carbides in it, vary somewhat with transformation
between coarse and fine pearlite structures is shown temperature, bainite being much harder than pearlite.
in Figures 22(d) and 22(e), respectively; in practice,
the platelets in very fine pearlite may not be resolv-
1.03.4.4 Formation of Martensite
able with a light microscope. The rate of transforma-
tion of carbon steels is also markedly affected by the Finally, at even lower transformation temperatures, a
presence of other alloying elements. However, a detailed completely new reaction occurs. Austenite trans-
consideration of the transformation and structure of forms to a new metastable phase called martensite,
low-alloy steels is beyond the scope of this section. which is a supersaturated solid solution of carbon in
iron and which has a body-centered tetragonal crystal
structure. Furthermore, the mechanism of the trans-
1.03.4.3 Formation of Bainite
formation of austenite to martensite is fundamentally
At lower transformation temperatures (<770 K different from that of the formation of pearlite or
approximately), a second reaction, the formation bainite; in particular, martensitic transformations do
of bainite, intervenes. Like pearlite, the bainite con- not involve diffusion and are accordingly said to be
stituent in steels consists of a mixture of ferrite and an diffusionless. Martensite is formed from austenite by
iron carbide and is formed by a time-dependent the slight rearrangement of iron atoms required to
nucleation and growth process involving diffusion transform the f.c.c. crystal structure into the body-
of carbon. Bainite, however, grows as needle-like centered tetragonal structure; the distances involved
plates, which consist of very fine particles of carbide are considerably less than the interatomic distances.
embedded in a ferrite matrix. The carbide particles A further characteristic of the martensitic transfor-
can only be resolved in an electron microscope; in the mation is that it is predominantly athermal, as
light microscope, only the acicular bainite needles are opposed to the isothermal transformation of austenite
γ
γ
(a) (b) Pearlite

(α – Fe + Fe3C)
Figure 23 Light micrograph showing the microstructure

of a martensitic steel ( 550).
Fe3C
α
(c) (d) after isothermal transformation at different tempera-
Pearlite
tures for varying times. The T–T–T diagram for a
commercial eutectoid steel is shown in Figure 24(a).
Also shown on the curves are the points at which the
microstructures illustrated in Figures 22 and 23 are
observed, and the thermal treatments producing
these structures. When a steel partially transfor-
med to, say, pearlite, is quenched from point a in
(e) (f) Figure 24(a) to below mf, the untransformed austen-
Figure 22 Some of the pearlitic and bainitic ite transforms to martensite.
microstructures observed in eutectoid steels after various
heat treatments.
1.03.4.6 Transformation of Hypo-Eutectoid
to pearlite or bainite. In other words, at a temperature Steels
midway between ms (the temperature at which mar-
As noted above, most steels are, in practice, hypo-
tensite starts to form), and mf (the temperature at
eutectoid. Consider, by way of example, a low-alloy
which martensite formation is apparently complete),
steel containing 0.3%C. The Fe–C phase diagram
the steel will consist of about 50% austenite and about
(Figure 20) indicates that as a steel of this composi-
50% martensite; this transformation of 50% of the
tion is cooled below about 1080 K, the equilibrium
austenite to martensite is effectively instantaneous,
state is a two-phase structure containing primary, pro-
and no matter how long the steel is held at this temper-
eutectoid ferrite and austenite. On further cooling to
ature, no further transformation will occur. Figure 23
below 996 K, the equilibrium structure consists of
illustrates the microstructure of a fully martensitic
pro-eutectoid ferrite and pearlite. These differences
steel. The martensite platelets are heavily twinned,
are reflected in the rather more complicated T–T–T
although this is only apparent in an electron micro-
diagram used for this steel, shown in Figure 24(b).
scope. In fact, in a eutectoid steel, the mf temperature is
Depending on the transformation temperature,
below room temperature although in a more common
the steel may transform to ferrite plus austenite (T1
hypoeutectoid steel containing about 0.2% C, ms
in Figure 24(b)), to ferrite plus pearlite via ferrite
720 K and mf 550 K. Other alloying elements also
plus austenite (T2), directly to ferrite plus pearlite
affect the temperature at which the martensite trans-
(T3) or directly to ferrite plus bainite (T4). The
formation takes place and usually decrease it.
pro-eutectoid ferrite morphology, like the pearlite
morphology, varies with the isothermal transforma-
1.03.4.5 Isothermal Transformation
tion temperature and also with the carbon content
Diagrams
of the steel. The ferrite nucleates heterogeneously
Time–temperature–transformation (T–T–T) dia- on austenite grain boundaries. Any pro-eutectoid
grams are used to present the structure of steels ferrite in slightly hypoeutectoid steels (about 0.6%),
72 Basic Concepts
Eutectoid temperature
(996 K)
1100
9.43 a
9.43 b α
900 9.43 c
Pearlite 9.43 d
Temperature (K)
a Pearlite (a) (b)

9.43 e
Pearlite + bainite
700 10% 90%
Austenite 9.43 f
Bainite
500 Ms
Martensite
Mμ
(c)
300 9.44
Figure 25 Schematic illustration of some of the ferritic/
pearlitic microstructures observed in hypo-eutectoid steels
10–1 10 103 105 after various heat treatments.
(a) Time (s)
cooling rate; they can also be altered by subsequent
heat treatment and working. Grain-boundary and
Widmanstätten ferrite are generally undesirable and
1100 result in poor mechanical properties. Figure 25 illus-
T1 γ
γ + ferrite trates some of the more important structures gener-
T2 ally observed in two-phase alloys.
900 γ + α + pearlite The higher the carbon content of a hypo-eutectoid
Temperature (K)
Ferrite + pearlite steel, the more pearlite there will be in a ferritic/

T3
pearlitic structure and the greater will be the strength
700 γ + α + bainite of the steel, other factors (grain-size, pearlite spacing,
T4 etc.) being equal.
γ Ferrite + bainite
Ms
10% 1.03.4.7 Tempering of Martensite
500 Martensite 90%
Mμ Transformation
Transformation As-quenched martensite is not only extremely hard,
but also extremely brittle. The hardness of as-
300 quenched martensite, which also increases with
increasing carbon content, is due partly to the strain
10–1 10 103 105 involved in retaining the carbon in solid solution in
(b) Time (s) body-centered tetragonal iron, and partly, in low-
Figure 24 Isothermal time temperature transformation carbon martensites, to their very high dislocation
curves for (a) a eutectoid steel and (b) a hypo-eutectoid steel. density. The martensite reaction in steels is not
reversible and on reheating as-quenched martensite
therefore, tends to be distributed along the austenite (Figure 26(a)) to various temperatures below 996 K,
grain boundaries, with the grain centers trans- the carbon tends to precipitate out of the solid solu-
forming later to pearlite, as shown schematically in tion; this process is known as tempering. On temper-
Figure 25(a). At lower carbon contents, Widmanstätten ing at about 370–520 K, the carbon precipitates out as
pro-eutectoid ferrite is observed, as shown schemati- the so-called e-carbide, which may be written as
cally in Figure 25(b), while in low-carbon steels, Fe2.4C and has a different structure from cementite.
massive ferrite is observed (Figure 25(c)). It is These carbides precipitate on certain crystallographic
important to reemphasize that these pro-eutectoid planes and on the twin boundaries in the martensite
ferrite structures also vary with isothermal transfor- plates, forming what is known as a Widmanstätten
mation temperature and therefore, in practice, with structure, as shown schematically in Figure 26(b).
Widmanstätten
Martensite Twins ε-carbide
platelets (a) (b)
Fe3C
precipitate (c) (d)
Figure 26 Some of the microstructures observed during the tempering of martensite.
On tempering at 520–770 K, cementite particles are

precipitated on twin boundaries and on boundaries
between martensite platelets, leaving a matrix of fer-
rite as shown in Figure 26(c). Eventually, the cement-
ite particles grow and become spheroidal in shape 9.47 b
Strength
while the twinned structure breaks down and fine 9.47 a

equiaxed ferrite grains are observed, as illustrated
in Figure 26(d). On tempering, as the carbon preci-
pitates out of the solid solution, the strength and 9.47 c
hardness of the steel falls progressively, as shown in
Figure 27, while the toughness increases. Quenched
and tempered low-alloy steels are widely used where 9.47 d
high-strength high-toughness materials are required. 200 400 600 800
Tempering temperature, (K)
1.03.4.8 Spheroidized Structures in Steels
Figure 27 Effect of tempering temperature on the
On tempering or annealing martensite, bainite, or even strength and microstructure of martensite.
pearlite at even higher temperatures (about 970 K),
a structure consisting of coarse cementite spher-
oids (readily visible in a light microscope) in a fer- 1.03.5 Strengthening Mechanisms
rite matrix is obtained. This is the most stable of all in Metals
ferrite/cementite aggregates, and it is also one of
the softest. Many of the ways of increasing the strength of metals
The structures and phase transformations observed have already been dealt with in passing. For example,
in steels have been dealt with in some detail not only cold working a metal increases its dislocation density
because of the great practical importance of steels, and therefore, as already discussed, increases its
but also because reactions similar to those occurring strength. (The processes involved in the nucleation
in steels are also observed in many other alloy sys- and multiplication of dislocations during cold work-
tems, too. In particular, diffusionless transformations ing are very complex and beyond the scope of this
(austenite ! martensite), continuous precipitation section.) Substitutional solid solutions (e.g., austenitic
(austenite ! pearlite), and discontinuous precipita- stainless steels and brass) are normally stronger than
tion (austenite ! bainite and tempering of martens- the equivalent pure metals (i.e., iron and copper), as
ite) are fairly common in other alloy systems. the lattice distortion involved hinders dislocation
74 Basic Concepts
movement. This is known as solid–solution hardening. readily achieved. At room temperature, because the
Metals with fine grain structures, produced by grain- limit of solid solubility decreases with decreasing
refining treatments, such as cold working followed by temperature, the equilibrium structure consists of
annealing to give recrystallization, are stronger than the a-solid solution plus the CuAl2 intermetallic
coarse-grained materials and single crystals, since compound. This equilibrium state may be achieved
grain boundaries act as barriers to dislocation move- by very slowly cooling the alloy from 820 K to room
ment. Steels may be quenched and tempered to pro- temperature and consists of very coarse discontinu-
duce high-strength and high-toughness materials. ous precipitates of CuAl2 in an a matrix, as shown in
Clearly, all these involve altering the structure of the Figure 28(b). However, on quenching the alloy to
metal in some way. Another very important strength- room temperature, diffusion (and therefore nucle-
ening mechanism, namely precipitation hardening, ation and growth of CuAl2) is prevented, and the
also involves microstructural changes. copper is retained in supersaturated solid solution,
as shown in Figure 28(c). This supersaturated solid
solution is relatively soft, weak, and ductile. If this
1.03.5.1 Precipitation Hardening
metastable supersaturated solid solution is then aged,
Precipitation hardening (also called age hardening that is, heat treated at a relatively low temperature
and dispersion hardening) is particularly important (370–520 K), the excess copper again precipitates out
in high-strength aluminum alloys, but it is also used of the solid solution as CuAl2. In this case, however,
to strengthen other alloys, notably certain steels the CuAl2 particles are extremely fine, certainly
and nickel-base alloys. Consider, by way of example, submicroscopic, and under certain conditions, only
the Al–Cu phase diagram shown in Figure 19. At just resolvable in the electron microscope. The pre-
about 820 K, the equilibrium structure of an Al–4Cu cipitation sequence varies from alloy to alloy but is
alloy (point X in Figure 19) is the single-phase a generally quite complex and involves a number of
solid solution of copper in aluminum, as shown in intermediate metastable precipitates. The initial pre-
Figure 28(a). On annealing at this temperature, that cipitates or coherent zones formed during the early
is, on solution heat treating, this equilibrium state is stages of aging (i.e., after relatively short times)
Slow
Cool
(a) (b)
Quench
+
age
Under-age Over-age
(d) (c) (g)
Peak-age
(e) (f)
Figure 28 Some of the microstructures observed during the heat treatment of precipitation-hardening aluminum alloys.
interact with dislocations and tend to hinder their solute depletion, there is always a precipitate-free zone
movement. This underaged condition is therefore adjacent to the grain boundary. The grain-boundary
stronger than the as-quenched condition, but since structure of a peak-aged high-strength precipitation-
the zones are extremely fine (Figure 28(d)) and are hardening aluminum alloy is therefore as shown
readily sheared by moving dislocations (Figure 29(a)), schematically in Figure 28(f ) and in the electron
it is only moderately so. On further aging, somewhat micrograph in Figure 31.
coarser, semicoherent intermediate precipitates are
formed (Figure 28(e)). These interact strongly with
dislocations and, being too large to be sheared, form 1.03.5.2 Complex Alloy Systems
very effective barriers to dislocation motion. Disloca-
Detailed consideration of the structure of many of the
tions can only move through the matrix by bowing
advanced and complex alloys that are of considerable
round the precipitates and leaving behind disloca-
technological importance (high-strength titanium
tion loops, as shown in Figure 29(b), and in this
alloys, nickel-base superalloys, etc.) is beyond the
peak-aged condition, the alloy is strongest. If aging is
continued, a relatively coarse dispersion of the equilib-
rium precipitate is formed. In this overaged condition
(Figure 28(g)), the precipitates are too coarse and
T1 <T2 < T3 <T4
widely spaced to interact strongly with dislocation;
they are relatively ineffective barriers to dislocation T1
motion, and the alloy again has only intermediate
strength. It should be emphasized, however, that even
in the over-aged condition, the precipitates are still
Strength
T2
submicroscopic.
The rate at which aging occurs depends on the
aging temperature. At very low aging temperatures, T3
over-aging is not observed in a realistic timescale T4
while at high aging temperatures, the strength falls off
almost immediately. In general, the lower the aging
Log ageing time
temperature, the greater the peak hardness. These
effects are illustrated schematically in Figure 30. Figure 30 Curves of strength against the logarithm of the
During aging, nucleation always occurs first at grain ageing time for a precipitation-hardening aluminum alloy
aged at different temperatures.
boundaries; the intergranular precipitates are therefore
always coarser and more advanced than the intragra-
nular precipitates. Furthermore, owing to vacancy and
Precipitate
(a)
Precipitate
Dislocation 0.5 μm
Dislocation loop Figure 31 Transmission electronmicrograph showing
(b)
intergranular and intragranular precipitation and the
Figure 29 (a) Dislocation shearing a precipitate precipitate-free zone adjacent to the grain boundary in a
particle and (b) a dislocation bowing round a precipitate high-strength precipitation-hardening Al–Zn–Mg alloy
particle. (24 000, courtesy of G. Lorimer).
76 Basic Concepts
scope of this section, other than to point out that no on mechanical properties, they can, in certain
new principles are involved. Certain titanium alloys, instances, markedly affect corrosion behavior.
for example, exhibit a martensitic transformation, The concepts discussed in this section are dealt with
while many nickel-base superalloys are age hardening. in greater depth in References Reed-Hill,1 Rollason,2
Similarly, cast irons, although by no means advanced Bailey,3 Smallman,4 Cottrell,5 Cahn,6 Chalmers,7
materials, are relatively complex. McLean,8 Shewman,9 and Peckner,10 inclusive.
References
1.03.5.3 Inclusions in Metals
In conclusion, it must be emphasized again that com- 1. Reed-Hill, R. E. Physical Metallurgy Principles; van
Nostrand Reinhold: New York, 1970.
mercial materials are never pure metals or pure 2. Rollason, E. C. Metallurgy for Engineers; Edward Arnold:
binary or ternary alloys. They always contain impu- London, 1973.
rities that are not particularly beneficial to the prop- 3. Bailey, F. W. J. Fundamentals of Engineering Metallurgy
and Materials; Cassell: London, 1972.
erties of the material; however, so long as they are not 4. Smallman, R. E. Modern Physical Metallurgy;
positively detrimental (as in the case of carbon in Butterworths: London, 1970.
certain austenitic stainless steels), their presence is 5. Cottrell, A. H. An Introduction to Metallurgy; Edward
Arnold: London, 1967.
tolerated as it would be uneconomical to remove 6. Cahn, R. W., Ed. Physical Metallurgy; North-Holland:
them. Some of these impurities are in solid solution. London, 1970.
Others, however, exist as inclusions that are quite 7. Chalmers, B. Physical Metallurgy; John Wiley: New York,
1959.
unaffected by heat treatment, although they may be 8. McLean, D. Mechanical Properties of Metals; John Wiley:
broken up by working. Examples are the Fe–Si-rich New York, 1962.
inclusions in aluminum alloys (Figure 10), manga- 9. Shewman, P. G. Transformations in Metal; McGraw-Hill:
New York, 1969.
nese sulfide inclusions in steel, etc. However, 10. Peckner, D., Ed. The Strengthening of Metals; Reinhold:
although these impurities may have a negligible effect New York, 1967.
1.04 Mechanical Properties and Fracture of Materials
A. H. Sherry
,Manchester,UK
T. J. Marrow
SchoolofMaterials,,Manchester,UK
1.04.1 Mechanical Properties 78

1.04.1.1 Elastic Properties 78
1.04.1.2 Elastic–Plastic Properties 80
1.04.2 Fracture Properties 81
1.04.2.1 The Elastic Stress Intensity Factor 81
1.04.2.2 The J-Integral 84
1.04.2.3 Fracture Toughness 85
1.04.3 Application of Fracture Mechanics to SCC 86
References 88
KIc Mode I fracture toughness under linear-elastic

Abbreviations conditions
BS British Standard
KISCC Threshold stress intensity factor for stress
EN Euro Norm
corrosion cracking
ASTM ASTM International
L0 Initial gauge length
HRR Hutchinson, Rice and Rosegren
L Gauge length
SCC Stress Corrosion Cracking
n Ramberg–Osgood work hardening exponent
UTS Ultimate Tensile Stress (also Ultimate Tensile
P Pressure
Strength)
P Load
R Polar coordinate centered on the crack tip
Ri Inner radius of cylinder
Symbols Ro Outer radius of cylinder
a Crack length rp Radius of crack-tip plastic zone
A Cross-sectional area T Wall thickness of cylinder
A0 Initial cross-sectional area Ti Components of traction vector
B Thickness U Strain energy
de Increment of true strain Ue Elastic strain energy
dL Change in gauge length under an applied load P ui Components of displacement vector
ds Increment of length along contour used to derive Up Plastic strain energy
the J-integral W Strain energy
D Diameter W Width
E Young’s modulus Y Geometry parameter
E0 Young’s modulus E under plane stress and G Anticlockwise contour around crack-tip used to
E/(1n2) under plane strain calculate J-integral
FQ Load at fracture in a fracture mechanics test a Material parameter in Ramberg–Osgood
G0–G5 Geometry coefficients equation
In Integration parameter e True strain
J J-integral e0 Reference strain in Ramberg–Osgood equation
Jc Fracture toughness under elastic–plastic ee Engineering strain
conditions eel True elastic strain
KI Mode I stress intensity factor epl True plastic strain
77
78 Basic Concepts
hp Nondimensional function of crack length specimen of smaller diameter will result in a higher
n Poisson’s ratio applied stress.
u Angular polar coordinate, centered on the crack The true stress applied to the specimen s is
tip defined as
s True stress P
s0 Reference stress in Ramberg–Osgood equation s¼ ½2
A
s0.2 Proof stress at 0.2% plastic true strain
where A is the actual cross-sectional area of the specimen
se Engineering stress
under the load P. The true stress applied to the specimen
sSCC Threshold stress for stress corrosion cracking
is not directly proportional to the applied load, since the
sxx Stress in x-direction
cross-sectional area of the specimen reduces progres-
sy Yield stress
sively under an increasing tensile load. This is because
syy Stress in y-direction
the total volume of the specimen remains practically
szz Stress in z-direction
unchanged as the specimen deforms. Equation [1] there-
s~ij(n,u) Dimensionless function of r and y
fore becomes an increasingly inaccurate measure of the
txy Shear stress (in the x–y plane)
true stress with increasing extension.
Engineering strain ee at a given applied load is
defined by the increase in the gauge length of the
specimen, dL, normalized by the initial gauge length
1.04.1 Mechanical Properties of the specimen, L0, that is
dL
The mechanical properties of materials are conve- ee ¼ ½3
L0
niently measured and described with respect to the
standard uniaxial tensile testing of round-bar or Since the gauge length increases under load, the
flat-plate specimens. Standard testing procedures, true strain e at a given instant in the test is more
such as BS EN 10002–1:2001 and ASTM E8-04,1,2 accurately defined as
provide guidance on the measurement and interpreta- ðL
dL
tion of tensile load-versus-displacement (i.e., gauge e¼ ¼ lnð1 þ ee Þ ½4
L0 L
length extension) data to obtain stress-versus-strain
curves. Typical curves are illustrated in Figures 1(a) where L is the actual gauge length of the specimen at
and 1(b) for a high-strength aluminum alloy and a the load of interest.
ferritic pressure vessel steel respectively. The key The true stress can be related to the engineering
regions of the stress-versus-strain curve are indicated stress and strain through the following relation,
in Figure 2 and described in Sections 1.04.2.1 and which assumes constant volume3:
1.04.2.2. In order to interpret these, it is important to
s ¼ se ð1 þ ee Þ ½5
first provide a definition of stress and strain.
Consider the round tensile bar specimen illu-
strated in Figure 3, which has an initial gauge length 1.04.1.1 Elastic Properties
L0, a diameter D and is under an applied load P. The
The stress-versus-strain curves illustrated in
engineering stress within the specimen se is defined
Figures 1(a) and 1(b) exhibit a number of distinct
by the applied load normalized by the initial cross-
regions. The first of these is characterized by a rela-
sectional area of the specimen, A0, that is
tionship between the applied stress and strain that is
P linear for most materials. Within this region, the
se ¼ ½1
A0 strain accumulated under an applied tensile stress is
fully recoverable upon unloading, that is, the material
For a round bar tensile specimen, A0 is equal to
is ‘elastic’ and its shape is not permanently changed
pD2/4, where D is the initial diameter of the bar. For
by the applied stress. The stress–strain relationship in
a specimen of rectangular cross-section, A0 is equal to
this elastic (linear) region is defined by Hooke’s law
WB, where W is the initial specimen width and B, the
in which Young’s modulus E given by
intial thickness. Equation [1] shows that the engi-
neering stress is directly proportional to the load s
E¼ ½6
applied to the specimen. The same load applied to a e
Mechanical Properties and Fracture of Materials 79
900
800
700
600
True stress (MPa)
500
400
300
200
100
0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
(a) True strain
900
800
700
Engineering stress (MPa)
600
500
400
300
200
100
0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
(b) Engineering strain
Figure 1 Typical stress–strain curves for (a) high strength aluminum alloy 2024-T351 tested at room temperature and
(b) Euro-material – a quenched and tempered ferritic pressure vessel steel DIN 22Ni–MoCr37 (similar to steel type ASTM
A508 Cl.3) tested at 90 C. (a) Adapted from Heerens, J.; Hellmann, D. Eng. Fract. Mech. 2002, 69, 421–449 and (b) Adapted
from Bernauer, G.; Brocks, W. Fat. Fract. Eng. Mater. Struct. 2002, 25, 363–384.
Engineering stress and strain are normally suffi- result in significant errors in the derived value of E. For
ciently accurate to define Young’s modulus for most this reason, other approaches have been developed for
materials with a reasonable degree of accuracy for the accurate determination of Young’s modulus. These
engineering calculations. Within the linear elastic include the measurement of the natural frequency of a
regime, eqn [6] can be used for loading in either vibrating loaded bar by spectroscopic techniques or
tension or compression. derivation from the velocity of sound in the material.3
It is worth noting that the Young’s modulus of many Although not measurable from the data shown in
engineering materials is extremely high. For example, Figures 1(a) and 1(b), Poisson’s ratio n is an impor-
mild steel has a Young’s modulus of 210 GPa at room tant materials parameter, since it defines the ratio of
temperature. Consequently, any small inaccuracies in the tensile strain accumulated in the loading direction
the measurement of displacement in this region will to the compressive (or negative) strain accumulated in
80 Basic Concepts
Fracture
500
450
ning
400
kh arde
Wor Ultimate tensile stress (su)
350
Stress (MPa)
300 0.2% proof stress (s0.2)
250 Yield stress (sy)
200
150
100
Young’s modulus (E) = s/e Aluminum alloy AL2024 T351
50 tested at room temperature
0.2% strain
0
0 5 10 15 20
Strain (%)
Figure 2 Key regions of a typical engineering stress–strain curve.
P specimen during testing, but accurate measurement

requires other methods, similar to those used for
Young’s modulus.
P
The strain energy density or work per unit vol-
ume, U, that is expended by deforming a material is
defined by the area under the true stress–true strain
dL curve, that is,
ðe
U ¼ s de ½8
0
L0 + dL Under linear elastic conditions, the elastic strain

Gauge L0
length energy Ue is defined by
s2
Ue ¼ ½9
2E 0
where elastic theory shows that E 0 = E under plane
Threaded stress conditions and E/(1n2) under plane strain
head conditions. (Plane stress is defined as the condition
where the out-of-plane stress is zero or negligible,
i.e., in thin sheets or on the material surface. Plane
Figure 3 Schematic illustration of a round tensile
strain is defined here as the condition where the out-
specimen under an applied load P.
of-plane strain is zero or negligible compared to the
in-plane strains, i.e. in thick sections.)
the perpendicular directions. Poisson’s ratio is defined
in the elastic regime and is given by
1.04.1.2 Elastic–Plastic Properties
tensile strain
n¼ ½7
perpendicular strain As the applied stress is increased in the tensile test,
there comes a point where the resulting displacement
For many engineering materials, n is close to 0.3 in can no longer be taken up by the elastic deformation
the elastic regime. Poisson’s ratio can be obtained by of the material, and the material ‘yields’ plastically.
measurement of the cross-section dimensions of the Yielding, which occurs in metals by the movement of
crystallographic imperfections called ‘dislocations,’ is 2

n=3
characterized by an increase in strain over and above 1.8
that predicted by Young’s modulus. The total strain of
Normalized stress (s/s 0)

1.6
5
the specimen is thus a combination of elastic and 1.4
plastic strain, that is, 1.2 10
20
1 ∞
e ¼ eel þ epl ½10 0.8
where e is the total true strain, eel is the elastic strain 0.6
defined by s/E, and epl is the plastic strain. 0.4

0.2
The applied stress at the limit of proportionality is
0
defined as the yield stress sy and is a characteristic 0 2 4 6 8 10
property of the material. Loading the specimen up to Normalized strain (e/e 0)
a stress higher than the yield stress and subsequently Figure 4 Normalized stress–strain curves plotted using
unloading will result in the permanent deformation the Ramberg–Osgood equation.
of the specimen equal to epl. Thus, in contrast to
elastic strain, plastic strain is not recovered.
In ductile metals, at a critical strain, the tensile
The limit of proportionality is often difficult to
specimen begins to ‘neck.’ Up to this point, the plastic
discern accurately, and therefore, the yield stress is
deformation occurs uniformly and stably throughout
commonly defined by a proof stress. This is typically
the gauge length as the rate of increase in stress due
the 0.2% proof stress, s0.2, which is defined using a
to the geometric reduction in cross-section is bal-
construction-line of gradient E drawn from an initial
anced by the rate of increase in yield stress due to
strain of 0.2%. The point at which this construction-
work hardening. At the onset of necking, the plastic-
line intersects the stress–strain curve is defined as
ity becomes unstable and is concentrated within a
s0.2. This is shown in Figure 2. This is also done in
localized region of the gauge length, when the geomet-
the case of materials, particularly ferritic steels that
ric and work hardening factors become unbalanced.
show a yield point (e.g., Figure 1(b)).
This is shown in Figure 5(a). This critical strain for
At applied stresses above the yield stress, the resis-
necking occurs at the maximum load, and thus corre-
tance to yielding increases, because the material work
sponds to the maximum in the engineering stress,
hardens as plastic deformation is accumulated. This is
which is termed the ultimate tensile stress (UTS) su
also known as cold work or strain hardening and is a
(Figure 2). It can be shown by Considère’s construc-
consequence of an increase in the number of disloca-
tion that, for a material following the Ramberg–Osgood
tions and their interactions with each other. The gra-
equation, necking begins at a strain e ¼ 1/n.5
dient of the stress–strain curve reduces in the work
For ductile materials, fracture follows necking
hardening region. This region may be characterized by
and normally shows a characteristic ‘cup-and-cone’
a power-law relationship between stress and strain. In
fracture. Ductile void growth and coalescence occurs
this way, eqn [10] may be formulated in terms of true
within the neck of the specimen, and then extends by
stress and true strain to provide a full description of
shear fracture, leading to the characteristic appear-
the stress–strain properties of materials.4 For many
ance of ‘ductile dimples’ from this ‘microvoid coales-
materials, a suitable expression is
cence’ process, Figures 5(b) and 5(c).
n
e s s
¼ þ ½11
e0 s0 s0 1.04.2 Fracture Properties
1.04.2.1 The Elastic Stress Intensity Factor
where a is a material parameter, often close to unity,
s0 is a reference stress usually taken as the yield or The linear elastic stress intensity factor, KI, is a param-
0.2% proof stress, e0 is the reference strain (¼s0/E), eter that quantifies the conditions of stress local to the
and n is the Ramberg–Osgood work hardening tip of an infinitely sharp crack in an elastic material. It
exponent. This work hardening exponent is typically has the units of MPa m1/2. The subscript ‘I’ denotes
between 5 and 10 for ferritic steels. Equation [11] is that the stress intensity factor relates to Mode I load-
known as the Ramberg–Osgood equation, and is illu- ing, in which the applied loading is perpendicular to
strated in Figure 4 with normalized stress-versus- the plane of the crack, Figure 6. Whilst there are other
strain curves plotted for a = 1 and 3 n 1. modes of loading including in-plane and out-of-plane
82 Basic Concepts
Neck
5 μm
Low carbon steel: 20C
(a) (b) (c)
Figure 5 Schematic illustration of: (a) neck formed in tensile specimen tested beyond UTS, (b) development of cup-and-
cone fracture, and (c) Scanning electron micrograph of ductile microvoid coalescence in low carbon steel.
r Point ‘P’
θ
z
x
Mode I Mode II Mode III
Figure 6 Definition of coordinate system and three modes of loading on a crack: Mode I opening, Mode II in-plane shear,
Mode III out-of-plane shear.

shear and combinations thereof, Mode I loading is KI y y 3y
normally considered the most severe and discussion sxx ¼ 1=2
cos 1 sin sin
ð2pr Þ 2 2 2
within this section is restricted to this mode.
KI y y 3y
The equations that relate KI to the stress at a point syy ¼ 1=2
cos 1 þ sin sin
‘P ’ in the vicinity of a sharp crack-tip in an elastic ð2pr Þ 2 2 2

material (Figure 6) are given by series expressions szz ¼ n sxx syy
known as the Westergaard functions. The first and
KI y y 3y
dominant term in the series expansion are given txy ¼ 1=2
sin cos cos ½12
below for plane strain conditions: ð2pr Þ 2 2 2
Within eqn [12], the subscripts indicate the Carte- magnitude of the applied stress, (2) the size of the
sian coordinate system that defines the plane of the flaw, and (3) the geometry of the component. The
crack and the loading direction, r and y are polar general form of the equation that relates the applied
coordinates of the point of interest (Figure 6), and n is stress s and the flaw size a to KI is as follows:
Poisson’s ratio. Figure 7 illustrates the distribution of pffiffiffiffiffi
KI ¼ sY pa ½13
opening stress syy directly ahead of a crack in an elastic
material under Mode I loading, that is, increasing r with where Y is a geometry parameter. Values for Y for
y ¼ 0 . The stresses are proportional to KI and hence common component geometries are tabulated in
the magnitude of the stress field within elastic material standards such as,6 and are typically of the order of
in the vicinity of a sharp crack under Mode I loading unity. As shown by eqn [13], the magnitude of KI is
may thus be deduced from a knowledge of KI. directly proportional to the applied stress acting on
A similar set of Westergaard functions relates the the component. Thus, increasing the stress applied to
local strain in elastic material in the vicinity of a a cracked component by a factor of two will double
crack tip to KI. Thus, as was the case for local stress, the magnitude of KI. The stress intensity factor KI
the magnitude of the strain field within elastic mate- is also dependent on the square root of the crack
rial in the vicinity of a sharp crack under Mode I depth a. Y is also dependent on crack length in
loading may also be deduced from knowledge of KI. some geometries, but the effect is generally less
The stress intensity factor is therefore a direct mea- strong. Consequently, under a fixed applied stress,
sure of the elastic strain energy in a stressed material the magnitude of KI increases as the crack depth
that contains a crack. increases. Hence, for a crack extending gradually by
Since failure by fracture is controlled by critical a mechanism such as SCC, in which the rate of crack
combinations of stress and strain in material close to growth can increase with KI, there can be an acceler-
the crack tip, the stress intensity factor, KI, provides a ation of crack growth.
means for quantifying the ‘crack driving force’ acting An example of equations defining the stress, s,
on a crack-like flaw in a stressed component – albeit and the geometry parameter, Y, is given in Figure 8
strictly only within the elastic regime. The higher the for an axially cracked pipe under internal pressure,
value of KI, the higher the crack driving force. In in which the crack (e.g., a stress corrosion crack) is on
terms of thermodynamics, crack growth requires suf- the inner surface7:
ficient elastic strain energy to balance the energy
pR2
needed to create the surfaces of the crack. This ¼ 2 o 2
includes the work associated with the deformation R o Ri
2 3
and fracture mechanisms of the material. A critical a a a
stress intensity factor, which is sufficient for crack Y ¼ 2G0 2G1 þ 3G2 4G3
Ri 1 Ri Ri
propagation, can therefore be determined. 4
a
The magnitude of KI acting on a given crack-like þ 5G5 ½14
flaw is dependent upon three factors: (1) the Ri
syy
y 1000
900
z
x 800
700
600
syy (MPa)
500
400
300
200
100
0
0 5 10 15 20 25 30 35 40
r (mm)
r
Figure 7 Elastic opening stress field ahead of a crack in an elastic material under Mode I loading with KI = 100 MPa m1/2.
84 Basic Concepts
where s is the circumferential stress acting on the

1 KI 2
plane of the crack due to the internal pressure, a is rp ¼ ½15
6p sy
the crack depth, Ri is the internal radius of the cylin-
der, and Ro is the outside radius of the cylinder.
Fracture toughness testing standards (see Section
The coefficients G0–G5 are tabulated in API 579.7
1.04.3) specify that for KI to be a valid, the crack-tip
Equation [14] is valid within the following geometric
parameter rp must be less than 1/50th of the char-
limits: 0 a/t 0.8 and 2 Ri/t 1000 where t is
acteristic dimensions of the specimen, that is
the wall thickness of the pipe, t ¼ (Ro – Ri). 2
The stress intensity factor is a valid parameter to KI
a; B; ðW aÞ 2:5 ½16
characterize fracture properties of materials under sy
linear-elastic conditions. However, simple inspection
where B is the specimen thickness, W is the specimen
of eqn [12] and Figure 7 shows that in a metal the
width, and (W – a) is the uncracked ligament asso-
stresses close to the crack tip will readily exceed the
ciated with the fracture mechanics test specimen
yield stress. Plastic deformation therefore occurs at
geometry, see Figure 9. Where this condition is not
the crack tip, in a region called the plastic zone, and
satisfied for any of these three measures of specimen
the assumption of elastic conditions would strictly not
dimension, a nonlinear parameter is required to
be valid. However, when the plastic zone size at the tip
account for the plastic deformation and to thereby
of the crack-like flaw is small compared with the
characterize the crack-driving force. The J-integral
dimensions of the component, the deformation of the
provides this parameter and is briefly described in the
component may be regarded as elastic and the stress
following section.
intensity factor can be valid also. Under plane strain
conditions the radius of the plastic zone, rp, may be
calculated according to the following expression8:
1.04.2.2 The J-Integral
For a material that follows the Ramberg–Osgood
stress–strain law, eqn [11], the crack-tip deformation
fields may be characterized by the Hutchinson,9 Rice,
and Rosegren (HRR) fields10 in a manner similar to
that of the Westergaard functions given previously as
eqn [12]. The HRR field equation for the crack-tip
stress field is defined as follows:
1=nþ1
EJ
ij ¼ s0 ~ij ðn; yÞ ½17
p as20 In r
where In is an integration parameter dependent

on the work hardening exponent, n, ~ij ðn; yÞ is a
dimensionless function of n and y, and J is the so-
Figure 8 Schematic illustration of axially-cracked pipe. called J-integral; a contour integral following an
B a
W
a
B
W
(a) (b)
Figure 9 Schematic illustration of fracture mechanics test specimens (a) compact-tension specimen and (b) four-point
bend specimen. Note arrows indicate direction of loading.
containing cracks such as compact-tension (CT) spe-

cimens or Single Edge-Notched Bend (SENB) speci-
mens, Figure 9. Standard sized specimens tend to be
y
used, with B between 12.5 and 50 mm. Much larger
and smaller specimens are possible, however.
x Standard testing procedures provide guidance on
the derivation of fracture toughness properties from
tests performed on fracture mechanics specimens.
ds
These standards describe stringent test conditions to
⌫
ensure that valid measurements of KIc or Jc are
obtained. For example:
Figure 10 Anticlockwise contour around crack tip used to
define the J-integral, eqn [18]. BS7448: ‘Fracture Mechanics Toughness Tests.’
Part 1:1991, ‘Method for determination of KIc,
Critical CTOD, and Critical J values of metallic
anticlockwise path around the tip of a crack, Figure 10,
material.’ British Standards Institute, London.
which defined as follows:
ð ASTM E399-90, ‘Standard Test Method for Plane
@ui Strain Fracture Toughness of Metallic Materials,’
J¼ w dy Ti ds ½18
@x American Society for Testing and Materials,
G Philadelphia, 1990.
where w is the strain energy density, Ti are compo-
Within these procedures, the fracture toughness of
nents of the traction vector, ui are the components of
the material is determined from the measured load
the displacement vector, and ds is an increment of
versus displacement behavior of the specimen, tested
length along the contour G.8
at the temperature of interest. The displacement is
The J-integral is widely used to define the crack-
measured as the crack mouth opening displacement
tip stress and strain fields for materials in the elastic-
(CMOD) along the load line, which is the line
plastic regime, that is, in circumstances where
between the points of loading. Loading is generally
eqn [16] does not hold. Under small-scale yielding
applied in a quasi-static manner (i.e., slow displace-
conditions (i.e., in the elastic regime where the plastic
ment rates of the order of mm/min) so that dynamic
zone size is small compared with the component and
loading effects do not influence the test results.
the assumption of elastic deformation implicit in eqn
Dynamic effects arise because the mechanism of
[16] is adequate), J is related to KI via eqn [19]8
yielding (i.e., movement of dislocations) is strain
KI2 ¼ E 0 J ½19 rate sensitive. This can cause the yield stress to rise
and fracture toughness to be lowered by high strain
rates, particularly in ferritic steels.
1.04.2.3 Fracture Toughness
Under linear elastic conditions, the fracture tough-
The fracture toughness of a material is a measure of ness is derived from expressions that relate the
its capacity to resist failure by the onset of crack applied load at fracture to the stress intensity factor.
extension. It can therefore be used to determine For a CT specimen, the expression is as follows11:
whether a crack in a component will grow by fracture pffiffiffiffiffi
FQ
processes under certain conditions of loading. Under KIc ¼ 1=2
Y pa ½20
linear elastic conditions, the critical stress intensity BW
factor KI measured at the onset of fracture is used where FQ is the load at fracture, B is the specimen
to define the fracture toughness of materials and is thickness, W is the specimen width, and a is the crack
given the subscript ‘c,’ that is, KIc. The unit of KIc is length (Figure 11). The geometry term Y is a func-
MPa m1/2. Under conditions of more widespread tion of the crack length-to-specimen width ratio,
plasticity, the elastic–plastic J-integral is widely a/W. Linear elastic conditions can be obtained in
used to define the critical crack-tip conditions and standard sized specimens in high-strength materials
hence the fracture toughness in the elastic–plastic such as age-hardened aluminum alloys, or structural
regime, Jc. The unit of Jc is kJ m2. ferritic steels tested at low temperatures.
Fracture toughness properties of materials are Under elastic–plastic conditions, the fracture tough-
conveniently measured using standard specimens ness is derived from expressions that depend on the
86 Basic Concepts
J-integral
Experimental data
Power law regression
CMOD 0.2 mm offset line
A
Force F
Jc
Up
Parallel to OA
Δa = 0.2 mm
Crack extension, Δa
Figure 12 Schematic illustration of definition of Jc from a

O Displacement CMOD J R-curve.
Figure 11 Schematic illustration of load F versus crack
mouth opening displacement (CMOD) trace and definition of behavior of structural components that are much larger
area under curve U. than the standard test specimen and in which linear
elastic conditions would apply.
applied load at fracture, but which also account for In most materials, the fracture toughness is not
the additional work done by deformation in the crack strongly affected by temperature. However, the frac-
tip plastic zone. For a CT specimen, the expression ture toughness of ferritic steels is sensitive to temper-
for J is as follows11 ature. This is due to an effect of their body-centered
h pffiffiffiffiffii2 ð1 nÞ p Up cubic crystal structure on the mechanism of yielding.
F
J¼ Y pa þ ½21 For example, the fracture toughness of ferritic pres-
BW 1=2 E B ðW a Þ
sure vessel steels at low temperatures can be as low as
where n is Poisson’s ratio, E is Young’s modulus, a is 20 MPa m1/2. This is the brittle fracture regime. In
the current crack length, and Up is the plastic strain the ductile fracture regime, observed at higher tem-
energy. This is defined by the plastic area under the peratures, the fracture toughness of ferritic steel at
measured load-versus-displacement curve, which is the onset of ductile tearing is typically of the order of
the shaded region in Figure 11. The parameter p is a 200 MPa m1/2. The ‘brittle–ductile transition tem-
function of crack length. perature’ describes fracture behavior between these
The crack length, a, at which J is measured can be two regimes and can be strongly sensitive to the
larger than the initial crack length a0, since stable microstructure of the steel. Neutron irradiation also
crack extension △a can occur during test, for exam- affects the yield behavior and toughness of ferritic
ple, by ductile tearing. To obtain the initiation frac- steels. The fracture toughness test is therefore a very
ture toughness, Jc, it is necessary to plot the variation important tool to predict the effect of temperature
of J as a function of △a, either using data obtained and other factors such as neutron irradiation, on the
from a number of interrupted tests, or using a single structural integrity of ferritic steel components con-
specimen tested using the unloading compliance taining crack-like flaws.
technique.8 A power-law curve is fitted to the valid
data and the initiation toughness defined by the
intersection of a 0.2 mm offset line and the J versus 1.04.3 Application of Fracture
Da power-law curve (known as the J R-curve) as Mechanics to SCC
illustrated in Figure 12.
Elastic–plastic conditions are typically obtained in Under a static loading condition, the growth of stress
standard sized specimens in tough materials, such as corrosion cracks is often characterized using a fracture
structural ferritic steels at ambient or elevated tempera- mechanics-based approach to define the crack growth
tures, or austenitic stainless steels. The measured frac- rate as a function of the stress intensity factor KI. This
ture toughness, Jc, can be used to calculate the fracture approach assumes that the crack length relative to the
toughness, K1c, using eqn [19], in order to predict the length scale of both the material microstructure and
component is sufficient such that the crack driving KI = KISCC

force can be defined by KI. Linear elastic conditions
therefore apply, and such cracks may be described as
Unacceptable
‘long cracks.’ Under such circumstances the required
condition for crack growth by SCC is
Stress s = s SCC
log (stress)
KI KISCC ½22 Acceptable
where KISCC defines the threshold stress intensity fac- Initiation Shallow Deep
tor for sustained crack growth. of SCC cracks cracks
For KI KISCC, an initiated crack will continue to

propagate, whereas for KI < KISCC, an initiated crack
will arrest. (Strictly, the condition for arrest is that log (flaw depth)
crack propagation is not measurable. It may therefore Figure 13 Schematic illustration of the two-parameter
be necessary to assume that crack growth may occur approach to stress corrosion cracking.
at a rate below the resolution of the measurements.)
The stresses in a component can change during the
lifetime, sometimes as a consequence of crack may either be too low and SCC does not occur (long
growth, and this approach can be used to determine cracks) or may be stress-controlled (short cracks). For
whether a growing stress corrosion crack will arrest. short cracks, the crack growth rate tends to increase
Cracks initiated at stress concentrating features with stress. For long cracks, as the applied KI increases
(e.g., corrosion pits) can be too short to be character- above KISCC, the crack growth rate rises rapidly to a
ized by KI, due to either their length relative to the plateau, Figure 14(a). It is important to note that the
microstructure scale or the development of elastic- magnitude of both KISCC and the crack growth rate
plastic deformation at the stresses needed to achieve plateau are both material and environment dependent.
K1 values approaching K1SCC. A threshold stress sSCC For example, Figure 14(b), reproduced from Re-
may then be used to define the conditions required to ference 12, shows the influence of copper content on
propagate such cracks. Under these circumstances, the long SCC crack growth rates for Al–Cu–Mg alloys.
required condition for ‘short crack’ growth by SCC is Reference 13 provides guidance regarding the genera-
tion of KISCC data using precracked laboratory speci-
s sSCC ½23 mens loaded under fixed load or fixed displacement
conditions.
where s is the applied tensile stress. For a component
In practice, most crack-like flaws in components
where s sSCC, an initiated short crack will continue
in which SCC is of concern are found using NDE
to propagate, whereas for s < sSCC, an initiated short
techniques. Cracks nucleating from weld defects are
crack will arrest. The critical stress, sSCC, is measured
typical of this. Such cracks are often of a size that can
experimentally by observation of the behavior of
be treated using the principles of fracture mechanics,
cracks nucleated in statically stressed specimens.
and the KI parameter is appropriate. However, there
This transition between short and long SCC
may be some circumstances when a stress-based
cracks, and s- or KI-controlled crack growth may be
approach is required – particularly when the crack
considered using a two-parameter approach to SCC.
initiates at a surface from pitting and significant crack
This is illustrated schematically in Figure 13.
growth occurs under stress-control. Under these cir-
The solid line defines the conditions above which
cumstances, it is important to take full account of the
SCC will be sustained (the unacceptable region).
stress-controlled extent of crack extension.
Below this line, any crack that has formed will arrest
There are a number of further important issues
(the acceptable region). It can be seen that incorrectly
that should be considered in the measurement and
applying long crack KI-controlled behavior to short
application of KISCC data. These include:
cracks can be nonconservative, that is, shallow cracks
may propagate even when KI < KISCC. KISCC will depend on the environment, material
The prediction of component lifetime requires data condition, and loading configuration used. For
for the effect of the crack-driving force on the crack example, the loading configuration may affect the
growth rate. As shown by Figure 13, depending on the transport of chemical species to the crack tip. The
crack length, the driving force for SCC crack growth conditions used to generate the data should
88 Basic Concepts
10−7
2219-T37 2124-T851 2048-T851
da Fracture (6.3% Cu) (4.4% Cu) (3.3% Cu)
log
dt
10−8
Crack velocity (m s–1)

10−9
10−10
10−11
(a) KISCC KIc KI (b) 0 10 20 30 40
Stess intensity (MN m–3/2)
Figure 14 (a) Schematic illustration of SCC crack growth rate as a function of applied KI, (b) SCC data for Al–Cu–Mg alloys
of varying copper content. Reproduced from Cottis, R. A. Guides to good practice in corrosion control: stress corrosion
cracking, NPL Report, 2000.
therefore reflect as closely as possible the service 3. Ashby, M. F.; Jones, D. R. H. Engineering Materials: An
Introduction to their Properties and Applications;
conditions of interest. Pergamon Press: New York, 1980.
The timescales used to generate KISCC are often 4. Patel, S. A.; Venkatraman, B.; Bentson, J. J. Franklin Inst.
shorter than typical timescales for components. 1963, 275, 98–106.
5. Weil, N. A. J. Franklin Inst. 1958, 265, 97–116.
For example, components in power generation 6. Murakami, Y. Stress Intensity Factors Handbook;
may require lifetimes in excess of 40 years. As Pergamon Press: New York, 1987; Vol. 1 and Vol. 2;
stress corrosion involves time-dependent chemical 1992; Vol. 3.
7. API 579, Fitness for Service; American Petroleum Institute
reactions, rate effects should be considered. Recommended Practice, 2000.
While KISCC testing is conventionally carried out 8. Anderson, T. L. Fracture Mechanics: Fundamentals
under static loading, service loading often includes and Applications, 2nd ed.; CRC Press: New York,
1995.
periodic unloads and reloads (e.g., due to start-up 9. Hutchinson, J. W. J. Mech. Phys. Solids, 1968, pp 13–31,
and shutdown), in which the loading rates can be Vol. 16.
quite slow or very fast relative to the KISCC test. 10. Rice, J. R.; Rosengren, G. F. J. Mech. Phys. Solids 1968,
16, 1–12.
Under such circumstances, KISCC may be affected. 11. BS7448, Fracture Mechanics Toughness Tests. Part 1,
Method for Determination of KIc , Critical CTOD and Critical
References J Values of Metallic Material; British Standards Institute:
London, 1991.
12. Cottis, R. A. Guides to good practice in corrosion control:
1. BS EN 10002-1:2001, Tensile testing of metallic materials. SCC, NPL Report 2000.
Method of test at ambient temperature; British Standards 13. ISO 7539-6:2003, Corrosion of metals and alloys –
Institution, 2001. stress corrosion testing – part 6: preparation of
2. ASTM E8-04, Standard Test Methods for Tension Testing pre-cracked specimens for tests under constant load
of Metallic Materials American Society for Testing and or constant displacement, International Standards
Materials (ASTM) International, 2004. Organisation, 2003.
1.05 Basic Concepts of Corrosion
L. L. Shreir{
This article is a revision of the Third Edition article 1.1 by L. L. Shreir, volume 1, pp 1:3–1:15, ß 2010 Elsevier B.V.
1.05.2 Definitions of Corrosion 90
1.05.3 Methods of Approach to Corrosion Phenomena 91
1.05.3.1 Corrosion as a Chemical Reaction at a Metal–Environment Interface 92
1.05.3.2 Environment 92
1.05.3.3 Metal 93
1.05.4 Types of Corrosion 93
1.05.5 Principles of Corrosion 96
1.05.5.1 Terminology 96
Appendix A – Classification of Corrosion Processes 96
Existing Classifications 96
‘Dry’ Corrosion 97
‘Wet’ Corrosion 98
Corrosion in Organic Solvents 98
Suggested Classification and Nomenclature 99
References 100
1.05.1 Introduction strength, hardness, and abrasion resistance are of major

importance, whereas electrical conductivity is of pri-
Modern technology has at its disposal a wide range of mary significance in electrical transmission. In the case
constructional materials – metals and alloys, plastics, of heat-exchanger tubes, good thermal conductivity
rubber, ceramics, composites, wood, etc., and the selec- is necessary, but this may be outweighed in certain
tion of an appropriate material for a given application is environments by chemical properties in relation to
the important responsibility of the design engineer. No the aggressiveness of the two fluids involved – thus,
general rule governs the choice of a particular material although the thermal conductivity of copper is superior
for a specific purpose, and a logical decision involves a to that of aluminum brass or the cupronickels, the
consideration of the relevant properties, ease of fabri- alloys are preferred when high-velocity seawater is
cation, availability, relative costs, etc. of a variety of used as the coolant, as copper has very poor chemical
materials; often, the ultimate decision is determined by properties under these conditions.
economics rather than by properties, and ideally, the While a metal or alloy may be selected largely on
material selected should be the cheapest possible that the basis of its mechanical or physical properties, the
has adequate properties to fulfill the specific function. fact remains that there are very few applications where
Where metals are involved, mechanical, physical, the effect of the interaction of a metal with its envi-
and chemical properties must be considered, and in this ronment can be completely ignored, although the
connection, it should be observed that while mechani- importance of this interaction is of varying signifi-
cal and physical properties can be expressed in terms of cance according to circumstances; for example, the
constants, the chemical properties of a given metal are slow uniform wastage of a steel structure of massive
dependent entirely on the precise environmental con- cross-section (such as railway lines or sleepers) is of far
ditions prevailing during service. The relative impor- less importance than the rapid perforation of a buried
tance of mechanical, physical, and chemical properties steel pipe or the sudden failure of a vital stressed steel
depends, in any given case, on the application of the component in sodium hydroxide solution.
metal. For example, for railway lines, elasticity, tensile The effect of the metal–environment interaction on
the environment itself is often more important than the
actual deterioration of the metal. For instance, lead
{
Deceased. pipes cannot be used for conveying plumbo-solvent
89
90 Basic Concepts
waters, since a lead level of >0.1 ppm is toxic; similarly, applicable to metals. Many authorities1 consider
galvanized steel may not be used for certain foodstuffs that the term metallic corrosion embraces all interac-
owing to the toxicity of zinc salts. In many chemical tions of a metal or alloy (solid or liquid) with its
processes, the selection of a particular metal may be environment irrespective of whether this is deliberate
determined by the need to avoid the contamination of and beneficial or adventitious and deleterious. Thus,
the environment with traces of metallic impurities that this definition of corrosion, which, for convenience, is
would affect the color or taste of products or catalyze referred to as the transformation definition, includes,
undesirable reactions; therefore, copper and copper for example, the deliberate anodic dissolution of zinc
alloys cannot be used in soap manufacture, since traces in cathodic protection and electroplating as well as
of copper ions result in the coloration and rancidifica- the spontaneous gradual wastage of zinc roofing
tion of the soap. In these circumstances, it is essential to sheet, resulting from atmospheric exposure.
use unreactive and relatively expensive metals, even On the other hand, corrosion has been defined2 as
though the environment would not result in the rapid ‘the undesirable deterioration’ of a metal or alloy, that
deterioration of cheaper metals such as mild steel. is, an interaction of the metal with its environment
A further possibility is that the contamination of the that adversely affects those properties of the metal
environment by metal ions due to the corrosion of one that are to be preserved. This definition – which is
metal can result in an enhanced corrosion of another referred to as the deterioration definition – is also
when the two are in contact with the same environ- applicable to nonmetallic materials such as glass,
ment. Thus, the slow uniform corrosion of copper by a concrete, etc. and embodies the concept that corro-
cuprosolvent domestic water may not be particularly sion is always deleterious. However, the restriction of
deleterious to copper plumbing, but it can result in the definition to undesirable chemical reactions of a
the rapid pitting and consequent perforation of galva- metal results in anomalies that become apparent from
nized steel and aluminum that subsequently comes a consideration of the following examples.
into contact with the copper-containing water. Steel, when exposed to an industrial atmosphere,
Finally, it is necessary to point out that for a number reacts to form the reaction product rust of approximate
of applications, metals are selected in preference to composition Fe2O3H2O, which, being loosely adher-
other materials because of their visual appearance, and ent, does not form a protective barrier that isolates the
for this reason, it is essential that brightness and reflec- metal from the environment; the reaction thus proceeds
tivity are retained during exposure to the atmosphere; at an approximately linear rate until the metal is com-
stainless steel is now widely used for architectural pur- pletely consumed. Copper, on the other hand, forms an
poses, and for outdoor exposure, the surface must adherent green patina corresponding approximately to
remain bright and rust-free without periodic cleaning. bronchantite, CuSO43Cu(OH)2, which is protective
On the other hand, the slow-weathering steels, which and isolates the metal from the atmosphere. Copper
react with the constituents of the atmosphere to form an roofs installed 200 years ago are still performing satis-
adherent uniform coating of rust, are now being used for factorily, and it is apparent that the formation of
cladding buildings, in spite of the fact that a rusty bronchantite is not deleterious to the function of copper
surface is usually regarded as aesthetically unpleasant. as a roofing material – indeed, in this particular appli-
The interaction of a metal or alloy (or a nonmetallic cation, it is considered to enhance the appearance of the
material) with its environment is clearly of vital impor- roof, although a similar patina formed on copper water
tance in the performance of materials of construction, pipes would be aesthetically objectionable.
and the fact that the present work is largely confined to The rapid dissolution of a vessel constructed of
a detailed consideration of such interactions could titanium in hot 40% H2SO4, with the formation
create the impression that this was the sole factor of of Ti4+ aquo cations, conforms to both definitions
importance in material selection. This, of course, is not of corrosion, but if the potential of the metal is raised
the case, although it is probably true that this factor is (anodic protection), a thin adherent protective film of
most neglected by the design engineer. anatase, TiO2, is formed, which isolates the metal
from the acid so that the rate of corrosion is enor-
mously decreased. The formation of this very thin
1.05.2 Definitions of Corrosion oxide film on titanium, like that of the relatively thick
bronchantite film on copper, clearly conforms to the
In the case of nonmetallic materials, the term corrosion transformation definition of corrosion, but not with
invariably refers to their deterioration from chemical the deterioration definition, since in these examples,
causes, but a similar concept is not necessarily the rate and extent of the reaction are not significantly
Basic Concepts of Corrosion 91
detrimental to the metal concerned. Again, magne- system, at the other, the environmental conditions pre-
sium, zinc, or aluminum is deliberately sacrificed vailing may necessitate the use of platinum.
when these metals are used for the cathodic protection The scope of the term ‘corrosion’ is continually
of steel structures, but as these metals are clearly not being extended, and Fontana and Staehle have stated3
required to be maintained as such, their consumption that ‘corrosion’ includes the reaction of metals, glasses,
in this particular application cannot, according to the ionic solids, polymeric solids, and composites with
deterioration, be regarded as corrosion. Furthermore, environments that embrace liquid metals, gases, non-
corrosion reactions are used to advantage in techno- aqueous electrolytes, and other nonaqueous solutions.
logical processes such as pickling, etching, chemical, Vermilyea, who has defined corrosion as a process in
and electrochemical polishing and machining, etc. which atoms or molecules are removed one at a time,
The examples already discussed lead to the conclu- considers that the evaporation of a metal into vacuum
sion that any reaction of a metal with its environment should come within the scope of the term, since atom-
must be regarded as a corrosion process irrespective of ically it is similar to other corrosion processes.4
the extent of the reaction or of the rates of the initial and Evans5 considers that corrosion may be regarded as
subsequent stages of the reaction. It is not illogical, a branch of chemical thermodynamics or kinetics, as
therefore, to regard passivity, in which the reaction prod- the outcome of electron affinities of metals and non-
uct forms a very thin protective film that controls rate of metals, as short-circuited electrochemical cells, or as
the reaction at an acceptable level, as a limiting case of a the demolition of the crystal structure of a metal.
corrosion reaction. Thus, both the rapid dissolution of These considerations lead to the conclusion that
active titanium in 40% H2SO4 and the slow dissolution of there is probably a need for two definitions of corro-
passive titanium in that acid must be regarded as corrosion that depend upon the approach adopted:
sion processes, even though the latter is not detrimental
1. Definition of corrosion in the context of Corro-
to the metal during the anticipated life of the vessel.
sion Science: the reaction of a solid with its
It follows that in deciding whether the corrosion
environment.
reaction is detrimental to a metal in a given application,
2. Definition of corrosion in the context of Corrosion
the precise form of attack on the metal (general, inter-
Engineering: the reaction of an engineering con-
granular, etc.), the nature of the reaction products (pro-
structional metal (material) with its environment
tective or nonprotective), the velocity and extent of the
with a consequent deterioration in properties of
reaction, and the location of the corrosion reaction must
the metal (material).
all be taken into account. In addition, due consideration
must be given to the effect of the corrosion reaction on
1.05.3 Methods of Approach to
the environment itself. Thus, corrosion reactions are
Corrosion Phenomena
not always detrimental, and our ability to use highly
reactive metals such as aluminum, titanium, etc. in
The effective use of metals as materials of construc-
aggressive environments is due to a limited initial cor-
tion must be based on an understanding of their
rosion reaction, which results in the formation of a rate-
physical, mechanical, and chemical properties. The
controlling corrosion product. Expressions such as ‘pre-
last, as pointed out earlier, cannot be divorced from
venting corrosion,’ ‘combating corrosion,’ or even
the environmental conditions prevailing. Any funda-
‘fighting corrosion’ are misleading; with the majority
mental approach to the phenomena of corrosion must
of metals, corrosion cannot be avoided and ‘corrosion
therefore involve a consideration of the structural
control’ rather than ‘prevention’ is the desired goal. The
features of the metal, the nature of the environ-
implication of ‘control’ in this context is that (1) neither
ment, and the reactions that occur at the metal–
the form nor the extent, nor the rate of the corrosion
environment interface. The more important factors
reaction must be detrimental to the metal used as a
involved may be summarized as follows:
constructional material for a specific purpose and
(2) for certain applications, the corrosion reaction 1. Metal – composition, detailed atomic structure,
must not result in the contamination of the environ- microscopic and macroscopic heterogeneities,
ment. Corrosion control must involve a consideration of stress (tensile, compressive, cyclic), etc.
materials, availability, fabrication, protective methods, 2. Environment – chemical nature, concentrations of
and economics in relation to the specific function of the reactive species and deleterious impurities, pres-
metal and its anticipated life. At one extreme, corrosion sure, temperature, velocity, impingement, etc.
control in certain environments may be effected by the 3. Metal–environment interface – kinetics of metal
use of thick sections of mild steel without any protective oxidation and dissolution, kinetics of the reduction
92 Basic Concepts
of species in solution, nature and location of corro- form of the nonmetal – C is usually referred to as the
sion products, film growth and film dissolution, etc. corrosion product, although the term could apply equally
to D. In its simplest form, reaction [1] becomes
From these considerations, it is evident that the
detailed mechanism of metallic corrosion is highly aA þ bB ¼ cC e:g:; 4Fe þ 3O2 ¼ 2Fe2 O3 ½2
complex and that an understanding of the various
where the reaction product can be regarded either as
phenomena involves many branches of the pure and
an oxidized form of the metal or as the reduced form of
applied sciences, for example, metal physics, physical
the nonmetal. Reactions of this type, which do not
metallurgy, the various branches of chemistry, bacte-
involve water or aqueous solutions, are referred to as
riology, etc., although the emphasis may vary with
‘dry’ corrosion reactions.
the particular system under consideration. Thus, in
The corresponding reaction in aqueous solution is
stress-corrosion cracking, emphasis may be placed on
referred to as a ‘wet’ corrosion reaction, and the
the detailed metallurgical structure in relation to
overall reaction (which actually occurs by a series of
crack propagation resulting from the conjoint action
intermediate steps) can be expressed as
of corrosion at localized areas and mechanical tearing,
while in underground corrosion, the emphasis may be 4Fe þ 2H2 O þ 3O2 ¼ 2Fe2 O3 H2 O ½3
on the mechanism of bacterial action in relation to the
Thus, in all corrosion reactions, one (or more)
kinetics of the overall corrosion reaction.
of the reaction products will be an oxidized form of
Although the mechanism of corrosion is highly
the metal, aquo cations (e.g., Fe2+ (aq.), Fe3+ (aq.)),
complex, the actual control of the majority of corrosion
aquo anions (e.g., HFeO 2 ðaq:Þ, FeO4 ðaq:Þ), or solid
2
reactions can be effected by the application of relatively
compounds (e.g., Fe(OH)2, Fe3O4, Fe3O4H2O,
simple concepts. Indeed, the Committee on Corrosion
Fe2O3H2O), while the other reaction product
and Protection6 concluded that ‘better dissemination of
(or products) will be the reduced form of the non-
existing knowledge’ was the most important single
metal. Corrosion may be regarded, therefore, as a
factor that would be instrumental in decreasing the
heterogeneous redox reaction at a metal–nonmetal interface
enormous cost of corrosion in the United Kingdom.
in which the metal is oxidized and the nonmetal is reduced. In
the interaction of a metal with a specific nonmetal (or
1.05.3.1 Corrosion as a Chemical Reaction
nonmetals) under specific environmental conditions,
at a Metal–Environment Interface
the chemical nature of the nonmetal, the chemical
As the first approach to the principles which govern the and physical properties of the reaction products, and
behavior of metals in specific environments, it is pref- the environmental conditions (temperature, pressure,
erable, for simplicity, to disregard the detailed structure velocity, viscosity, etc.) are clearly important in deter-
of the metal and to consider corrosion as a heteroge- mining the form, extent, and rate of the reaction.
neous chemical reaction that occurs at a metal–
nonmetal interface and that involves the metal itself
1.05.3.2 Environment
as one of the reactants (cf. catalysis). Corrosion can be
expressed, therefore, by the simple chemical reaction: Some examples of the behavior of normally reactive
and nonreactive metals in simple chemical solutions
aA þ bB ¼ cC þ d D ½1
are considered here to illustrate the fact that corrosion
where A is the metal and B the nonmetal reactant (or is dependent on the nature of the environment; the
reactants), and C and D, the products of the reaction. thermodynamics of the systems and the kinetic factors
The nonmetallic reactants are often referred to as the involved are considered in subsequent chapters.
environment, although it should be observed that in a Gold is stable in most strong reducing acids,
complex environment, the major constituents may whereas iron corrodes rapidly, yet finely divided gold
play a very subsidiary role in the reaction. Hence, in can be quickly dissolved in oxygenated cyanide solu-
the ‘atmospheric’ corrosion of steel, although nitrogen tions that may be contained in steel tanks. A mixture of
constitutes 75% of the atmosphere, its effect, com- caustic soda and sodium nitrate can be fused in an iron
pared with that of moisture, oxygen, sulfur dioxide, or nickel crucible, whereas this melt would have a
solid particles, etc., can be disregarded (in the high disastrous effect on a platinum crucible.
temperature reaction of titanium with the atmosphere, Copper is relatively resistant to dilute sulfuric
on the other hand, nitrogen is a significant factor). acid, but corrodes if oxygen or oxidizing agents are
One of the reaction products (say, C) will be an present in the acid, whereas austenitic stainless steels
oxidized form of the metal, and D will be a reduced are stable in this acid only if oxygen or other
oxidizing agents are present. Iron corrodes rapidly in have a pronounced effect on the location and the kinet-
oxygenated water, but extremely slowly if all oxygen ics of the corrosion reaction and are considered in
is removed; if, however, oxygen is brought rapidly and various sections throughout this work. Practical envir-
simultaneously to all parts of the metal surface, the onments are shown schematically in Figure 2, which
rate will become very low, owing to the formation of a also serves to emphasize the relationship between the
protective oxide film. Lead dissolves rapidly in nitric detailed structure of the metal, the environment, and
acid, more slowly in hydrochloric acid, and very external factors such as stress, fatigue, velocity, impin-
slowly in sulfuric acid. These examples show that the gement, etc.
corrosion behavior of a metal cannot be divorced
from the specific environmental conditions prevailing,
1.05.4 Types of Corrosion
which determine the rate, extent (after a given period
of time), and form of the corrosion process.
Corrosion can affect the metal in a variety of ways,
which depend on its nature and the precise environ-
1.05.3.3 Metal
mental conditions prevailing, and a broad classification
Heterogeneities associated with a metal have been of the various forms of corrosion, in which five major
classified in Table 1 as atomic (see Figure 1), micro- types have been identified, is presented in Table 2.
scopic (visible under an optical microscope), and Thus, an 18Cr–8Ni stainless steel corrodes uniformly
macroscopic, and their effects are considered in vari- during polishing, active dissolution, or passivation, but
ous sections of the present work. It is relevant to locally during intergranular attack, crevice corrosion,
observe, however, that the detailed mechanism of all or pitting; in certain circumstances, selective attack
aspects of corrosion, for example, the passage of a along an ‘active path’ in conjunction with a tensile
metallic cation from the lattice to the solution, specific stress may lead to a transgranular fracture. Types of
effects of ions and species in solution in accelerating or corrosion are dealt with in more detail in Appendix A.
inhibiting corrosion or causing stress-corrosion crack- Ideally, the metal selected, or the protective system
ing, etc. must involve a consideration of the detailed applied to the metal, should be such that no corrosion
atomic structure of the metal or alloy. occurs at all, but this is seldom technologically or
The corrosion behavior of different constituents economically feasible. It is necessary, therefore, to tol-
of an alloy is well known, since the etching techni- erate a rate and a form of corrosion that will not be
ques used in metallography are essentially corrosion significantly detrimental to the properties of the metal
processes that take advantage of the different corro- during its anticipated life. Thus, provided the corrosion
sion rates of phases as a means of identification, for rate is known, the slow uniform corrosion of a metal
example, the grain boundaries are usually etched can often be allowed for in the design of the structure;
more rapidly than the rest of the grain owing to for example, in the case of a metal that shows an active–
the greater reactivity of the disarrayed metal. passive transition, the rate of corrosion in the passive
Macroscopic heterogeneities, for example, crevices, region is usually acceptable, whereas the rate in the
discontinuities in surface films, bimetallic contacts, etc. active region is not. It follows that certain forms of
Table 1 Heterogeneities in metal
1. Atomic (as classified by Ehrlich and Turnbull,7 see Figure 1).

a. Sites within a given surface layer (‘normal’ sites); these vary according to the particular crystal plane (Figure 3).
b. Sites at edges of partially complete layers.
c. Point defects in the surface layer: vacancies (molecules missing in surface layer), kink sites (molecules missing at edge of
layer), molecules adsorbed on top of complete layer.
d. Disordered molecules at point of emergence of dislocations (screw or edge) in metal surface.
2. Microscopic
a. Grain boundaries – usually, but not invariably, more reactive than grain interior.
b. Phases – metallic (single metals, solid solutions, intermetallic compounds), nonmetallic, metal compounds, impurities,
etc. – heterogeneities due to thermal or mechanical causes.
3. Macroscopic
a. Grain boundaries.
b. Discontinuities on metal surface – cut edges, scratches, discontinuities in oxide films (or other chemical films) or in
applied metallic or nonmetallic coatings.
c. Bimetallic couples of dissimilar metals.
d. Geometrical factors – general design, crevices, contact with nonmetallic materials, etc.
94 Basic Concepts
Kinked
step Step
Adsorbed
ion
Impurity
in surface
Mono-molecular
pit
Screw
Edge
dislocation
dislocation
Figure 1 Surface imperfections in a crystal. Reproduced from Ehrlich, G.; Turnbull, D. Physical Metallurgy of Stress
Corrosion Fracture; Interscience: New York and London, 1959; p 47, with permission from Interscience.
Ambient
Wet Damp
High temp.
.
temp
ric
Oxyge
ca
he
Nit n dio
rbo
Atmosp
r
ate
rog xid
aw
en e
Co
,
H
Se
2 S, nta
mi rs
Hc SO na te
2, nd wa
so NH SO tion
l ,
ra e
lid 4 C 3 , N r ic ve arin
pa l,e H p he i
R tu
rtic tc. 3 , os s es
les tm tion
s ondi
‘Wet’ or ‘dry’ acc d vap A
r
u c
Dry
ing o
o
Na
t
Aq
tu r
ters
Steam le wa
ord
ue
an
Potab
al
Wet
ous
Gases
Metals and alloys

or ‘wet’ corros
chemical properties
Submicroscopic Underground
es
Chemical fum Microscopic Heterogeneities
Macroscopic
ion
Red
s Bio
ase che
eg
ucin
Flu mic
al
g
Non
aque -
Ox
ous
idis
Ch
ing
em
id
ica
utio ids
Ac
Foodstuf
ls
ns
u
l
and nic liq
ra
ls
ut
meta
Ne
Fused salts
sol
ne Temperature
a
ali
fs
Org
Liquid
k
lex Al Pressure
mp
Stress Co ming Velocity
fatigue for Agitation
fretting Cavitation
Figure 2 Environments in corrosion.

corrosion can be tolerated and that corrosion control is time are invariably pitted, although the pits seldom
possible, provided that the rate and the form of the penetrate the metal, that is, the saucepan remains
corrosion reaction are predictable and can be allowed functional and the pitted appearance is of no signifi-
for in the design of the structure. cance in that particular application.
Pitting is regarded as one of the most insidious These considerations lead to the conclusion that
forms of corrosion, since it often leads to perforation the relationship between corrosion and the deteriora-
and to a consequent corrosion failure. In other cases, tion of properties of a metal is highly complex and
pitting may result in a loss of appearance, which is of involves a consideration of a variety of factors such as
major importance when the metal concerned is used the rate and the form of corrosion and the specific
for decorative architectural purposes. However, alu- function of the metal concerned; certain forms of cor-
minum saucepans that have been in service for some rosion such as uniform attack can be tolerated, whereas
others such as pitting and stress corrosion cracking that
ultimately lead to complete loss of function cannot be.
The implications of the terms predictable and unpre-
dictable used in the context of corrosion require fur-
ther consideration, since they are clearly dependent
on the knowledge and expertise of the engineer,
designer, or corrosion designer who takes the deci-
sion on the metal or alloy to be used, or the proce-
dure to be adopted, to control corrosion in a specific
environmental situation. On this basis, a corrosion
failure (i.e., failure of the function of the metal due
to corrosion within a period that is significantly less
Figure 3 Hard-sphere model of face-centered cubic (fcc) than the anticipated life of the structure) may be the
lattice showing various types of sites. Numbers denote result of one or more of the following possibilities:
Miller indices of atom places and the different shadings
correspond to differences in the number of nearest neighbors.
Reproduced from Ehrlich, G.; Turnbull, D. Physical Metallurgy 1. Predictable. (1) The knowledge and technology are
of Stress Corrosion Fracture; Interscience: New York and available, but have not been utilized by the
London, 1959; p 47, with permission from Interscience. designer; this category includes a wide variety of
Table 2 Types of corrosion
Type Characteristic Examples
1. Uniform (or almost All areas of the metal corrode at the same Oxidation and tarnishing; active dissolution in
uniform) (or similar) rate acids; anodic oxidation and passivity;
chemical and electrochemical polishing;
atmospheric and immersed corrosion in
certain cases
2. Localized Certain areas of the metal surface corrode at Crevice corrosion; filiform corrosion; deposit
higher rates than others due to attack; bimetallic corrosion; intergranular
‘heterogeneities’ in the metal, the environment corrosion; weld decay
or in the geometry of the structure as a whole.
Attack can range from being slightly localized
to pitting
3. Pitting Highly localized attack at specific areas Pitting of passive metals such as the stainless
resulting in small pits that penetrate into the steels, aluminum alloys, etc., in the presence
metal and may lead to perforation of specific ions, for example, Cl ions
4. Selective One component of an alloy (usually the most Dezincification; dealuminification; graphitization
dissolution active) is selectively removed from an alloy
5. Conjoint action of Localized attack or fracture due to the Erosion – corrosion, fretting corrosion,
corrosion and a synergistic action of a mechanical factor and impingement attack, cavitation damage;
mechanical factor corrosion stress corrosion cracking, hydrogen cracking,
corrosion fatigue
96 Basic Concepts
design features such as a wrong choice of materials, science, and many of the phenomena of corrosion are
introduction of crevices and bimetallic contacts, not fully understood. Thus, the phenomenon of passiv-
etc., and is the most common cause of corrosion ity, which was first observed by Faraday in 1836, is still a
failures. (2) The knowledge and technology are subject of controversy; the specific effect of certain
available, but have not been applied for economic anions in causing stress-corrosion cracking of cer-
reasons; for example, inadequate pretreatment of tain alloy systems is not fully understood; and the
steel prior to painting and the use of unprotected dezincification of brasses can be prevented by additions
mild steel for silencers and exhaust systems of cars. of arsenic (or other elements such as antimony or
2. Unpredictable. (1) The design has been based on phosphorus), but no adequate theory has been submit-
specific environmental conditions, which have sub- ted to explain the action of these elements. (Editor’s
sequently changed during the operation of the pro- note: Aruguably this phenomenon is now somewhat
cess; in this connection, it should be noted that small better understand, see Chapter 2.05, Dealloying.)
changes in the chemical nature of the environment, An understanding of the basic principles of the
temperature, pressure, and velocity may lead to science of metallic corrosion is clearly vital for cor-
significant changes in the corrosion rate and form: rosion control, and as the knowledge of the subject
the catastrophic oxidation and the failure of steel advances, the application of scientific principle rather
bolts in nuclear reactors in the United Kingdom, than an empirical approach may be used for such
resulting from an increase in the temperature of the purposes as the selection of corrosion inhibitors, the
carbon dioxide, is an example of an unpredictable formulation of corrosion-resisting alloys, etc.
failure due to a change in environmental conditions.
(2) The knowledge of and the experience with the 1.05.5.1 Terminology
metal, alloy, or the environment are insufficient to
predict with certainty that failure will not occur; The classification given in Table 2 is based on the
examples could be quoted of new alloys that have various forms that corrosion may take, but the terminol-
been subjected to an extensive series of carefully ogy used in describing corrosion phenomena often
planned corrosion tests, but have failed in service. places emphasis on the environment or the cause of
attack rather than on the form of attack. Thus, the
Professor M. Fontana8 has made the statement broad classification of corrosion reactions into ‘wet’ or
that ‘‘Virtually all premature corrosion failures these ‘dry’ is now generally accepted, and the nature of the
days occur for reasons which were already well known process is often made more specific by the use of an
and these failures can be prevented.’’ It is apparent adjective that indicates the type or the environment,
from this statement and from the conclusions reached for example, concentration-cell corrosion, crevice cor-
by the Committee on Corrosion and Protection that rosion, bimetallic corrosion and atmospheric corrosion,
category 1 is responsible for the majority of incidents high temperature corrosion, seawater corrosion, etc.
of corrosion failure that could have been avoided if Alternatively, the phenomenon is described in terms of
those responsible were better informed on the the corrosion product itself – tarnishing, rusting, and
hazards of corrosion and on the methods that should green rot. The terminology used in corrosion is given in
have been used to control it. Table 3 and is considered in more detail in Appendix A.
1.05.5 Principles of Corrosion

Appendix A – Classification of
Corrosion Processes
It has been stated that metallic corrosion is an art
rather than a science and that, at present, the available
knowledge is insufficient to predict with any certainty Existing Classifications
how a particular metal or alloy will behave in a spe-
cific environment.4 It should be appreciated that the A logical and scientific classification of corrosion
decision to use a particular metal or alloy in prefer- processes, although desirable, is by no means simple
ence to others in a given environment or to employ a owing to the enormous variety of corrosive environ-
particular protective system is based usually on previ- ments and the diversity of corrosion reactions, but the
ous experience and empirical testing rather than on the broad classification of corrosion reactions into ‘wet’
application of scientific knowledge – the technology of or ‘dry’ is now generally accepted, and the terms are
corrosion is, without doubt, in advance of corrosion in common use. The term ‘wet’ includes all reactions
Table 3 Terminology in corrosion
Type of attack Environmental Cause of attack Mechanical factors Corrosion product

a
General (uniform) Wet Concentration cell Stress Rusting
Localized Dry Bimetallic cell Fretting Tarnishing
Pitting (or intense) Atmospheric Active–passive cell Fatigue Scaling
Intergranular Immersed Stray current (electrolysis) Cavitation Green rot
Transgranular Underground Hydrogen evolution Erosion Tin pest
Selective Seawater Oxygen absorption Impingement
Parting Chemical Impingement
Catastrophic Fused-salt Hydrogen embrittlement
Layer Flue-gas Caustic embrittlement
Filiform Biochemical
Bacterial
High temperature
Liquid–metal
a
See Appendix to this section.
in which an aqueous solution is involved in the reac- Metal Gas Metal ze O2 Metal
tion mechanism; implicit in the term ‘dry’ is the
absence of water or an aqueous solution. Mz+
These terms are evidently ambiguous; for exam-
O2−
ple, it is not always clear whether ‘wet’ is confined to O2 Mz+aq. OH−
aqueous solutions – the ‘wetting’ of solids by mercury ze
indicates that liquid–metal corrosion should be clas-
Oxide Electrolyte
sified as ‘wet’. Even if the term is restricted to aque- film solution
ous solutions, the difficulty arises that the mechanism Anode Cathode Anode Cathode
of the growth of magnetite scale during the reaction (a) (b)
of the interior of a boiler drum with dilute caustic
Figure A1 Anodes and cathodes in corrosion processes:
soda at high temperatures and pressures is best inter- (a) ‘dry’ corrosion and (b) ‘wet’ corrosion.
preted in terms of a ‘dry’ corrosion process. Similar
considerations apply to the reactions of aluminum The corresponding ‘wet’ corrosion half-reactions
and zirconium with high temperature water. (Figure A1(b)) are
Considering oxidation as a typical ‘dry’ reaction,
it follows, as shown in Figure A1(a), that at the ½Mzþ þ zelattice þ H2 O ! Mzþ aq: ½7
interfaces and
M Ð ðM ○=OÞ þ zðe○=OÞ
zþ
½4 O2 þ 2H2 O þ 4e ¼ 4OH ½8
where M ○ is an interstitial metal ion, e○ an inter-
z+
or
stitial electron, and /O indicates the metal–oxide
interface. O2 þ 4Hþ þ 4e ¼ 2H2 O ½9
If the metal dissolves to enter a vacant site, then
M Ð ðMzþ h=O þ zeh=OÞ ½5 ‘Dry’ Corrosion
where M □ represents a cation vacancy and e□ a
z+
These are generally metal–gas or metal–vapor reac-
positive hole. tions involving nonmetals such as oxygen, halogens,
At the gas–oxide interface, the O2 gas ionizes hydrogen sulfide, sulfur vapor, etc., and oxidation,
scaling, and tarnishing are the more important
1
O2 =ads: þ 2ðe=XÞ Ð ðO2 =ads:Þ ½6 forms. A characteristic of these reactions is that the
2
initial oxidation of the metal, the reduction of the
where /X indicates the gas–oxide interface. nonmetal, and the formation of the compound occur
By definition, these interfaces can be considered as at one and the same place at the metal–nonmetal
anodes and cathodes respectively. interface. Should the compound be volatile or
98 Basic Concepts
discontinuous, further interaction at the interface (or solid phase, but as this will form away from the inter-
through a thin film of constant thickness) is possible, face, it will not be protective – the thermodynamically
and in most cases, the reaction rate tends to remain stable oxide can affect the kinetics of the reaction only
constant with time (linear law). If the film is continu- if it forms a film or precipitates on the metal surface.
ous, it will present a barrier to the reactants, and Further points that distinguish ‘wet’ from ‘dry’
further interaction will necessitate the passage of corrosion are
the reactants through the film by (a) diffusion of the
1. In ‘wet’ corrosion, the metal ions are hydrated –
nonmetal or (b) diffusion and migration of ions of
the hydration energy of most metal ions is very
the reactants. The detailed mechanisms of these reac-
high and thus facilitates ionization.
tions are considered in subsequent chapters, but it is
2. In ‘wet’ corrosion, the ionization of oxygen to
appropriate to observe that the formation of a con-
hydroxyl must involve the hydronium ion or water.
tinuous film of reactant product at a metal–nonmetal
3. In ‘dry’ corrosion, the direct ionization of oxygen
interface results in a growth rate that, when the film
occurs.
becomes sufficiently thick to be rate determining,
decreases as the film thickens, that is, parabolic, loga-
rithmic, asymptotic, cubic, etc. Corrosion in Organic Solvents
Corrosion reactions in aggressive organic solvents are

‘Wet’ Corrosion
becoming a more common occurrence owing to the
In ‘wet’ corrosion, the oxidation of the metal and the developments in the chemical and petrochemical
reduction of a species in solution (electron acceptor or industries, and these reactions can lead to the deterio-
oxidizing agent) occur at different areas on the metal ration of the metal and to undesirable changes in the
surface with consequent electron transfer through the solvent. This aspect of corrosion has recently been the
metal from the anode (metal oxidized) to the cathode subject of an extensive review by Heitz,9 who has
(electron acceptor reduced); the thermodynamically considered the mechanisms of the reactions, the simi-
stable phases formed at the metal–solution interface larities between corrosion in organic solvents and in
may be solid compounds or hydrated ions (cations or aqueous solutions, the methods of study, and the
anions), which may be transported away from the inter- occurrence of the phenomenon in industrial processes.
face by processes such as migration, diffusion, and con- Figure A2 shows the weight loss against time
vection (natural or forced). Under these circumstances, curve for nickel in various solvents containing 0.05
the reactants will not be separated by a barrier, and the wt% H2SO4 at various temperatures, and illustrates
rate law will tend to be linear. Subsequent reaction with the unpredictable nature of corrosion in organic sol-
the solution may result in the formation of a stable vents. Thus, the corrosion rates in ethanol are far
Ethanol Acetone Acetic acid Water

40 ⬚C
600 78.5 ⬚C
21 ⬚C
500
21 ⬚C
Weight loss (g m-2)
400 no 100 ⬚C
118⬚ H2SO4
300
40⬚
40⬚
200
21⬚
100
40⬚ 40 ⬚C 21⬚
56.5 ⬚C 21⬚ 118⬚
0 20 40 0 20 40 0 20 40 0 20 40
Days
Figure A2 Corrosion of nickel in different solvents containing 0.05 wt.% H2SO4 at various temperatures. Reproduced from
Heitz, E. In Advances in Corrosion Science and Technology; Fontana, M. G., Staehle, R. W., Eds.; Plenum Press, 1974;
Vol. 4, p 149, with permission from Plenum Press.
greater than those in the aqueous acid, whereas in 2M þ 2RSH ! 2MS þ H2 þ R2 ½15
acetone, the rate is practically zero; even more
surprising is the fact that in acetic acid, the addition Heitz quotes a number of case studies of corrosion
of 0.05% H2SO4 actually decreases the corrosion rate. of metals in organic solvents and concludes that the
Heitz classifies corrosion reactions in organic sol- phenomenology indicates no specific differences from
vents into that experienced in aqueous corrosion. Thus, general
1. Electrochemical reactions, which follow a similar corrosion, pitting, crevice corrosion, intergranular cor-
mechanism to those in aqueous solution. rosion, erosion–corrosion cracking, hydrogen embrit-
2. Chemical reactions, which involve direct charge tlement, etc. can all occur in organic solvents. The
transfer between the metal atom in the lattice of methods of control also follow those used for corrosion
the metal and the oxidizing species. in aqueous solutions, although there are certain differ-
ences. Thus, cathodic and anodic protection are seri-
In the case of electrochemical reactions, the par- ously limited by the resistivity of the solvent, and paint
tial anodic reaction results in the formation of a coatings deteriorate rapidly in contact with the solvent.
solvated metal cation Mzþ solv: , a charged or uncharged
metal complex MX, or a solid compound MXz,
Suggested Classification and
where X is a halogen ion, organic acid anion, etc.
Nomenclature
The cathodic partial reactions are as follows:
(a) Reduction of a solvated proton to H2 gas On a basis of the preceding discussion, the classifi-
1 cation and nomenclature outlined in Table A1 are
Hþ
solv: þ e ! H2 ½10 suggested as a possible alternative to the accepted
2
classification of corrosion reactions into ‘wet’ and ‘dry.’
(b) Reduction of acidic hydrogen of a proton donor It is considered that the main types of corrosion
reactions can be classified as follows:
1
HA þ e ! H2 þ A ½11 1. Film-free chemical interaction in which there is
2
direct chemical reaction of a metal with its envi-
where A is a carboxylic acid anion, alcoholate ronment. The metal remains film-free, and there is
ion, etc. no transport of charge.
(c) Reduction of an oxidizing gas Y 2. Electrochemical reactions that involve transfer of
Yz þ zme ! zYm ½12 charge across an interface. These electrochemical
reactions can be further subdivided into:
where Y can be O2, Cl2, F2, Br2, O3, N2O4, etc. a. Inseparable anode–cathode type (insep. A/C).
(d) Reduction of oxidizing ions such as Fe3+, Cu2+, The anodes and cathodes cannot be distin-
MnO
4 , ClO3 , etc. guished by experimental methods, although
It is evident from these equations that in many sys- their presence is postulated by theory, that is,
tems, the reaction of a metal with an organic solvent the uniform dissolution of metals in acid, alkaline,
follows a mechanism that is similar to the electro- or neutral aqueous solutions, in nonaqueous solu-
chemical mechanism of corrosion in aqueous solution. tion, or in fused salts. (Dr. Pryor considers that in
Nonelectrochemical processes may be repre- certain cases of uniform dissolution of metals in
sented by the general equation acids (e.g., A1 in hydrochloric or sulfuric acid)
or alkalis a thin film of oxide is present on the
M þ 2Cx Hy Xz ! MX2 þ C2x H2y X2x2 ½13 metal surface – the film is not rate-determining
but its presence would indicate that reactions
where X is a halogen and M is a divalent metal, for
of this type should be classified under 2(c)).
example, the Grignard reaction
b. Separable anode–cathode type (sep. A/C).
Mg þ CH3 Cl ! CH3 Mg ½14 Certain areas of the metal can be distinguished
experimentally as predominantly anodic or
A further type of chemical process, which is anal- cathodic, although the distances of separation
ogous to high temperature corrosion, is the reaction of these areas may be as small as fractions of a
of metals with organic sulfur compounds, which millimeter. In these reactions, there will be
follow the equation a macroscopic flow of charge through the metal.
100 Basic Concepts
Table A1 Classification of spontaneous corrosion reactions
1. Film-free Chemical Interaction

(a) Metal/gas–oxide or compound volatile (e.g., reaction of molybdenum with oxygen, reaction of iron or aluminum with
chlorine).
(b) Metal/liquid
Reactions of solid metals with liquid metals (e.g., dissolution of aluminum in mercury)
Dissolution of metal in their fused halides (e.g., lead in lead chloride).
Dissolution of metals in nonaqueous solutions (e.g., reaction of aluminum with carbon tetrachloride).
2. Electrochemical
(a) Inseparable anode/cathode type (insep. A/C)
Reactions with aqueous solutions. Uniform dissolution or corrosion of metals in acid, alkaline, or neutral solutions (e.g.,
dissolution of zinc in hydrochloric acid or in caustic soda solution; general corrosion of zinc in water or during
atmospheric exposure). Reactions with nonaqueous solution (e.g., dissolution of copper in a solution of ammonium
acetate and bromine in alcohol). Reactions with fused salts.
(b) Separable anode/cathode type (sep. A/C)
All reactions of metals in aqueous or nonaqueous solutions or in fused salts where one area of the metal surface is
predominantly anodic and the other is predominantly cathodic so that the sites are physically identifiable.
(c) Interfacial anode/cathode type in which the metal surface is filmed
i. Metal/gas and metal/vapor reactions
All reactions in which charge is transported through a film of reaction product on the metal surface – the film may or
may not be rate determining (e.g., parabolic, logarithmic, asymptotic, etc., or linear growth laws, respectively).
ii. Metal/solution reactions
All reactions involving the uniform formation and growth of a film of reaction product (e.g., reaction of metals with
high temperature water, reaction of copper with sulfur dissolved in carbon disulfide).
c. Interfacial anode–cathode type (interfacial Acknowledgements

A/C). One entire interface will be the anode
and the other will be the cathode. Thus, as Grateful thanks are due to Dr W.B. Jepson,
shown in Figure A1(a), the metal–metal oxide Dr M. Pryor, and Mr J.N. Wanklyn for the helpful
interface might be regarded as the anode and discussions during the preparation of this Appendix.
the metal–oxygen interface as the cathode.
It is apparent that, in general, 2(a) and 2(b) include References
corrosion reactions that are normally classified as
‘wet,’ while 2(c) includes those that are normally 1. Hoar, T. P. J. Appl. Chem. 1961, 11, 121; Vernon, W. H. J.
classified as ‘dry.’ The Conservation of Natural Resources; Institution of
Civil Engineers: London, 1957; Vol. 105; Potter, E. C.
The terminology suggested can be illustrated by Electrochemistry; Cleaver-Hume: London, 1956; 231.
reference to the corrosion behavior of iron: 2. Uhlig, H. H. Ed. The Corrosion Handbook; Wiley: New York/
Chapman and Hall: London, 1948; Uhlig, H. H.
1. Reaction of iron with oxygen at room temperature Corrosion and Corrosion Control; Wiley: New York, 1971;
or with oxygen or water at high temperatures – Fontana, M. G.; Greene, N. D. Corrosion Engineering;
McGraw-Hill, 1967.
interfacial A/C type. 3. Fontana, M. G.; Staehle, R. W. Advances in Corrosion
2. Reaction of iron with oxygenated water or with Science and Technology; Plenum Press: New York, 1990.
reducing acids – inseparable A/C type. 4. Vermilyea, D. A. In Proceedings of the 1st International
Congress on Metallic Corrosion, London, 1961;
3. Reaction of iron containing a discontinuous magne- Butterworths: London, 1962; p 62.
tite scale with oxygenated water, crevice corrosion, 5. Evans, U. R. The Corrosion and Oxidation of Metals; Arnold:
water-line attack, ‘long-line’ corrosion of buried London, 1960; 12.
6. Report of the Committee on Corrosion and Protection,
iron pipes, etc. – separable A/C type. Department of Trade and Industry, HMSO, 1971.
7. Ehrlich, G.; Turnbull, D. Physical Metallurgy of Stress
Although it is realized that this classification and Corrosion Fracture; Interscience: New York and London,
terminology has certain limitations, it represents a 1959; 47.
preliminary attempt to provide a more rational clas- 8. Fontana, M. G. Corrosion 1971, 27, 129.
9. Heitz, E. In Advances in Corrosion Science and Technology;
sification of corrosion processes than that based on Fontana, M. G., Staehle, R. W., Eds.; Plenum Press:
‘wet’ and ‘dry.’ New York, 1974; Vol. 4, 149.
1.06 Defects and Transport in Oxides and Oxide Scales
B. Pieraggi
Ecole Nationale Supérieure des Ingénieurs en Arts Chimiques et Technologique de Toulouse, Institut National Polytechnique de
Toulouse, F-31077 Toulouse Cedex 04, France

1.06.2 Crystal Defects 102
1.06.2.1 Point Defects: Thermal Defects and Constitutive Defects 102
1.06.2.2 Dislocations 104
1.06.2.3 Surfaces 107
1.06.2.4 Grain Boundaries and Interfaces 108
1.06.3 Point Defects in Ionic Crystals 110
1.06.3.1 Structure and Building Elements: Kröger–Vink Notation 110
1.06.3.2 Use of Kröger–Vink Notation: Formation Equilibriums of Frenkel and
Schottky Disorders 111
1.06.3.3 Electronic Defects and Ionized Point Defects 112
1.06.3.4 Point Defect Equilibriums in Nonstoichiometric Oxides 113
1.06.3.4.1 p-Type oxides 113
1.06.3.4.2 n-Type oxides 114
1.06.3.5 Intrinsic Defect Equilibriums Close to Stoichiometry 114
1.06.3.6 Effect of Impurities on Defect Equilibriums 115
1.06.3.7 Extended Defects 115
1.06.4 Diffusion Laws and Mass Transport in Solid State 116
1.06.4.1 Diffusion Laws 116
1.06.4.2 Lattice Diffusion and Diffusion Mechanisms 117
1.06.4.3 Interdiffusion, Intrinsic Diffusion, and Kirkendall Effect in Binary Solid Solutions 118
1.06.4.4 Diffusion in Multiphase Systems 119
1.06.4.5 Interdiffusion in Ternary Systems 120
1.06.4.6 Short-Circuit Diffusion 121
1.06.5 Diffusion and Transport in Oxides 122
1.06.5.1 Chemical Diffusion in Nonstoichiometric Oxides 122
1.06.5.2 Diffusion Processes in Oxide Scales 123
1.06.5.3 Oxide Scale Microstructure and Diffusion Processes 124
1.06.5.4 Point Defects and Diffusion in Common Oxides 127
1.06.5.4.1 Nickel and cobalt monoxides 127
1.06.5.4.2 Iron oxides 128
1.06.5.4.3 Chromia 129
1.06.5.4.4 a-Alumina (a-Al2O3) 129
1.06.6 Concluding Remarks 130
References 131
Abbreviations Symbols
DIGM Diffusion-induced grain-boundary motion b Burgers vector of a dislocation
PSZ Partially-stabilized zirconia n Molar fraction of free electron (mol1)
SZ Stabilized zirconia p Molar fraction of positive hole (mol1)
TLK Terrace-ledge-kink t Time (s)
101
102 Principles of High Temperature Corrosion
xi Mole fraction of species i (mole1) to different aspects of diffusion and mass transport,
Ci Molar concentration of i (mol m3) including diffusion laws, diffusion in two-phase sys-
D Diffusion coefficient (m2 s1) tems and in ternary systems, and short-circuit diffu-
J Diffusion flux (mol m2 s1) sion. Specific aspects of diffusion in oxide scales and
Kr Equilibrium constant for the reaction r defect structure, and diffusion in the most common
P Pressure (bar) oxides are reported in Section 1.06.5 before some
Q Activation energy (J mol1) concluding remarks (Section 1.06.6).
R Universal gas constant (8.314 J mol1 K1)
T Temperature (K) 1.06.2 Crystal Defects
Tm Temperature of melting (K)
Dfg Molar free enthalpy of formation (J mol1) The crystal structure of any solid chemical com-
Dfgi Molar free enthalpy of formation of component pound defines the perfect periodical arrangement of
i (defect or compound) (J mol1) its chemical constituents. However, many physical
DVr Molar relaxation volume (m3 mol1) and chemical processes would not be possible with-
d Departure from ideal stoichiometry (mol1) out the existence of local discontinuities in this per-
f Pilling–Bedworth ratio fect periodic arrangement; these local discontinuities
j Thickness (m) form more or less extended defective zone or crystal
V Molar volume (m3 mol1) defects. Different kinds of crystal defects are distin-
guished from their geometrical dimension D: 0D for
point defects, 1D for linear defects, and 2D for planar
defects. These different defects are briefly described
1.06.1 Introduction
in the following subsections.
High temperature oxidation of many materials often
1.06.2.1 Point Defects: Thermal Defects
results in the formation and growth of a continuous
and Constitutive Defects
external oxide scale and/or in the internal oxidation
of the less-noble constituents of the oxidized materi- Point defects are the simplest defects that can be
als. Therefore, matter transport usually plays a very found in any crystal phase. They are localized on
important role in scale growth and internal oxidation single sites of the crystal structure; these sites can
processes. Indeed, a growing external oxide scale be regularly occupied by some chemical species or
separates its own substrate from the oxidizing chemi- else regularly unoccupied sites of the vacant intersti-
cal environment; thus, oxidant species and/or mate- tial sublattice. Figure 1 defines the simplest and
rial constituents must be transported across the oxide common point defects that can be present in pure
scale during its growth. Similarly, internal oxidation solid chemical elements; vacancies and interstitials
involves the dissolution and diffusion of oxidant spe- can also be present in disordered solid solutions.
cies within the substrate. Furthermore, heat-resisting More complex defects such as divacancies or disso-
metallic materials are protected against oxidation by ciated interstitials can also be formed; they are not
the selective oxidation of their less-noble constitu- considered here.
ents, usually Cr and/or Al, leading to the growth of Interstitial sites are defined from the crystal struc-
chromia or alumina scales that, in addition to matter ture. The simplest case corresponds to the close-
transport across the scale, requires the diffusion of Al packed stacking of hard spheres leading to the
or Cr within the oxidized substrate toward the scale–
substrate interface.
Therefore, processes and mechanisms of diffusion + + + + + + + + + + +
Is
and mass transport in oxide scales and oxidized sub- + + + + + + + + + + +
V
strates are of prime importance in the understanding + + + + + + + + SI + +
and modeling of the high temperature oxidation of
+ + + Ii + + + + + + +
materials. As diffusion and mass transport in solids
require the presence of crystal defects, Section
Figure 1 Schematic representation of point defects in a
1.06.2 is devoted to a general description of common pure solid chemical element: n, vacancy; Is, substitutional
crystal defects. Point defects of ionic crystal are impurity; SI, self-interstitial; Ii, interstitial impurity; the
described in Section 1.06.3. Section 1.06.4 is devoted crosses mark the vacant interstitial sites.
Defects and Transport in Oxides and Oxide Scales 103
face-centered cubic (fcc) lattice or to the hexagonal Although point defects are a characteristic of atomic
close-packed (hcp) lattice. Two kinds of interstitial level, their usual thermodynamic quantities are com-
site, octahedral and tetrahedral, are usually distin- monly expressed in molar quantities ( J mol1).
guished in such close-packed structure. For an fcc Therefore, in eqn [1], Df g d is the molar Gibbs energy
lattice, there is one octahedral site and two tetrahe- of formation of intrinsic defects d. As shown by eqn
dral sites per lattice site. For fcc pure metals such as [1], the mole fraction xd increases with T and depends
Ni or g-Fe, small impurity atoms such as C, H, or O, only on that external independent variable, which
for example, would occupy the octahedral sites that justifies that such intrinsic defects are also called ther-
are the larger ones. For MO oxides of rock salt mal defects. Vacancies are the dominant point defect in
structure such as NiO and FeO and for oxides common metals and alloys; enthalpy of formation and
M3O4 (Fe3O4) or AB2O4 (NiCr2O4) of spinel struc- entropy of vacancies in pure metals have been reviewed
ture, oxygen anions also form a close-packed fcc by ‘Kraftmaker.’3 Equation [1] also means that the
lattice. In rock salt structure, metal cations are only chemical potential of vacancy is null at equilibrium.
localized over octahedral sites of the fcc anionic sub- For disordered solid solution, enthalpy of formation
lattice and the tetrahedral sites always remain empty. and entropy of point defects, particularly for vacancies,
Thus, point defects of such MO oxides would be would depend on solid-solution composition. How-
mainly vacant octahedral sites (cation vacancies) ever, for ideal solid solutions, there is no interaction
and oxygen vacancies or self-interstitials. For spinel between solution constituents and therefore no local
structure, only one-half of the octahedral sites and modification of the composition in the vicinity of point
one-eighth of the tetrahedral sites are occupied by defects; the chemical potential of vacancy at equilib-
metal cations, which permits more complex combi- rium is thus independent of solid-solution composition
nations of cationic point defects as briefly described and remains null. Therefore, the equilibrium vacancy
in Section 1.06.5.4.2. molar fraction varies monotonically as a function of
Vacancies and self-interstitial defects are also composition. The situation is more complex for non-
called intrinsic defects, while substitutional or inter- ideal solid solutions where point defects would interact
stitial impurities are extrinsic defects. The amount of differently with solution constituents. For example, a
point defects in a crystal is usually defined by their constituent of larger atomic radius could trap vacancies
concentration (defects per cubic meters), their site to accommodate more easily the local volume change it
fraction or their molar fraction (defects per mole), introduces. Therefore, in nonideal solid solution, the
which is also, the number of defects per mole of the dependence of defect molar fraction as a function of
considered crystal phase. The thermodynamic argu- solution composition could be complex and the pres-
ment for the stability of defective crystals is well ence and rearrangement of point defects could favor
known and does not need to be repeated here. localized composition change.
A detailed treatment of the thermodynamic of point Under very high pressure or very large stress, the
defects can be found elsewhere.1,2 equilibrium mole fraction of intrinsic defects also
From the Gibbs phase rule, in unstressed state, depends on pressure and stress state.4 For the simple
pure solid chemical element, for example, pure Ni case of a crystal submitted to a hydrostatic pressure P,
or Si, and disordered solid solutions of constant the mole fraction of intrinsic defects varies as
composition have no degree of freedom at constant
PðO þ Dvr Þ
temperature T and pressure P. The corresponding xd ðPÞ ¼ xd ð0Þ exp ½2
RT
thermodynamic systems are invariant so that intrin-
sic point defects cannot be considered as being inde- where O is the crystal molar volume (the sign is ‘þ’
pendent constituents of these systems; these defects for vacancy formation and ‘–’ for interstitial forma-
are nonconservative species that need to be created/ tion) and Dvr is the molar relaxation volume, that is,
annihilated at sources or sinks. 1 mol of the local volume variation resulting from the
By assuming that pressure P remains in a range relaxation of the crystal lattice around a defect d.
such that its influence can be neglected, the mole Equation [2] can be also applied to stressed crystals
fraction xd of the intrinsic point defects d then only by replacing the pressure P by a stress tensor s and
depends on temperature T: the term (O + Dvr) by a formation volume tensor vf .
Like pressure, a compressive stress has a positive sign,
D f gd
xd ¼ exp ½1 and hence, from eqn [2], hydrostatic pressure or
RT compressive stress would decrease the equilibrium
molar fraction of vacancy while a tensile stress would Ni1dO where d is the mole fraction of cation
increase it. An opposite influence would be observed vacancies per mole of NiO. At constant T, the depar-
with interstitial defects. ture d varies as a function of one compositional
The situation is more complex for binary or multi- variable that, in this case, is usually the activity (par-
component compounds. Indeed, an n-ary compound, tial pressure) of oxygen. Therefore, these vacancies
in an unstressed state and at constant T, has (n 1) are not thermal defects; they are structurally linked
degree of freedom, that is, its composition depends on to the crystal structure of NiO, which justifies their
(n 1) composition variables that can vary indepen- denomination of structural or constitutive defects.
dently without affecting its crystal structure. The b-NiAl, contrary to NiO, is stable over a large
chemical formula or ideal stoichiometric composition range of chemical composition, including its stoichio-
of such a compound is fixed by its crystallographic metric composition. In that case, the type and con-
structure, but, as there is (n 1)-independent com- centration of structural point defects compatible with
position variables, the real chemical composition can the conservation of crystal structure vary with the
differ from the ideal stoichiometric composition. mole fraction of Al or Ni. For example, complex
Several types of point defects can explain such a point defects such as triple defects associating one
departure from ideal stoichiometric composition. antisite Ni atom and two Ni vacancies can be found
Figure 2 defines the main types of single intrinsic in stoichiometric b-NiAl, while Ni vacancies and
point defects that can be encountered in binary com- antisite Ni atoms are the predominant defects in
pounds. The point defects of such compounds are Al-rich and Ni-rich b-NiAl, respectively.
usually described by affecting a sublattice for each At constant temperature, the composition range
independent compound constituent. But, in addition over which a binary compound is stable, that is, the
to vacancies and interstitials associated to each sub- concentration limits of point defects, depends on the
lattice, antisite defects can be also present in multi- relative stability of this compound with respect to
component compounds. For a simple binary AB the closest adjacent compounds. A very stable com-
compound, an antisite defect corresponds to the pound, such as alumina, would have a very narrow
wrong occupation of one crystal site as, for example, stability range corresponding to very small depar-
an atom A occupying a lattice site regularly occupied tures from stoichiometry and very low defect con-
by an atom B, or inversely an atom B on a site of centrations. A less stable oxide, such as Ni1dO,
A. Such antisite defects are mainly observed in inter- would have larger stability range and thus larger
metallic and covalent compounds; they cannot be defect concentration, in which case it is equal to d/O.
stable in ionic compounds because of the large and
repulsive electrostatic interactions they would induce.
1.06.2.2 Dislocations
Two simple examples are provided by the crystal
structure of nickel oxide NiO (rock salt structure) Dislocations are linear defects involved in mechani-
and the simple cubic crystal structure of nickel alu- cal behavior, mass transport, and reactivity of materi-
minide b-NiAl. For NiO, its departure from the ideal als.5 In a crystal, a dislocation is a line that, on one slip
stoichiometry is only due to a deficit in Ni2+ cations, plane, forms the boundary between a crystal region
its real chemical composition can be expressed as that has slipped and an adjacent unslipped region so
that the glide of the dislocation line results in plastic
deformation (Figure 3(a)). Such a definition has sim-
ple but important consequences. One of them is that
A B any dislocation cannot simply end in the bulk of a
VA crystal; a dislocation must have its extremities on a
IB surface, an interface or a grain boundary or intersect
VB two other dislocations to form a dislocation node or
IA
BA AB else form a dislocation loop. Dislocations are char-
acterized by the direction of the dislocation line and
the slip vector or Burgers vector b defining the
relative displacement of slipped and unslipped crys-
Figure 2 Intrinsic point defects in a compound AB: VA, VB
and IA, IB are, respectively vacancies and interstitials of
tal regions. Figures 3(b) and 3(c) illustrates sche-
sublattices A and B. AB and BA correspond to antisite matically the two limiting types or characters of
defects. dislocations:
Symbolic representation of dislocations
Dislocation line
Local plastic deformation
(a)
b // dislocation line
b ⊥ dislocation line Screw-dislocation
Edge-dislocation
(b) (c)
Figure 3 Schematic definition of dislocations: (a) mixed dislocation, (b) edge dislocation, and (c) screw dislocation.
1. edge dislocations for which b is normal to the translation vectors, which justify their denomination
dislocation line (Figure 3(b)), of partial dislocations. Such a dislocation dissociation
2. screw dislocations for which b is parallel to the is energetically favorable, only if b21 > b22 þ b23 . The
dislocation line (Figure 3(c)). example of fcc metals is well known and described in
many basic textbooks.5
A dislocation loop has a mixed character when its The definitions of perfect edge and screw disloca-
Burgers vector lies in the plane of the loop, some tions emphasize their role in plastic deformation by
parts are edge and some parts are screw. Such a loop dislocation glide. But, dislocation motion is not
may glide in its own plane. On the contrary, when the restricted to gliding and dislocation lines are usually
Burgers vector is normal to the loop plane, the dislo- not straight. Indeed, many irregularities are com-
cation loop, named prismatic loop, is a pure edge monly observed along dislocation lines resulting
dislocation and cannot glide. from dislocation intersections and also from thermal
Dislocations induce local displacements of lattice activation.5 These irregularities can be described as
sites from their equilibrium position. These displace- more or less sharp steps distributed along a disloca-
ments result in strain and stress fields around each tion line. One step lying in the dislocation slip plane
dislocation. The related elastic strain energy per unit forms a kink, but such a step forms a jog when it is
length of dislocation line is proportional to b2 so that normal to the slip line (Figure 4). Kinks and jogs can
the dislocation lines tend to be straight to minimize move by glide, the slip plane is then stepped for
the energy added to the crystal. One other conse- jogged dislocations. On the contrary, the displacement
quence is that the Burgers vector b of stable disloca- of a jog along one dislocation moves the dislocation line
tions is the shortest translation vector of the crystal in a direction normal to its slip plane; such a movement
lattice, which permits to infer easily the possible is called climb. Climb occurs at jogs by the creation/
Burgers vectors of dislocations for simple crystal annihilation of vacancies and interstitials. Jog moving
lattices – as, for example, the fcc lattice of pure along an edge dislocation line acts as a source or sink for
metals. These dislocations are called perfect disloca- vacancies and interstitials, as illustrated by Figure 5.
tions. One perfect dislocation of Burgers vector b1 The interactions among the jog, dislocation climb, and
can dissociate into two partial dislocations of Burgers intrinsic point defects play an important role in the
vectors b2 and b3 separated by a bidimensional defect recovery and creep as well as in point defect equilib-
called stacking fault. The Burgers vectors b2 and b3 of rium – and thus on transport processes – at high tem-
such partial dislocations are a fraction of lattice perature, that is, at T > 0.4Tm for metallic materials.
Edge dislocation
Jog Kink
b b
Edge segment
Screw segment
Jog = dislocation segment normal to glide plane Kink = dislocation segment parallel to glide plane
Screw dislocation
Jog
Kink
b Edge segment b
Edge segment
Figure 4 Kink and jog on edge and screw dislocations.
Side view of the arrowed plane
Dislocation
line
Vacancy source or
Jog displacement ≡
interstitial sink
Dislocation
line
Edge dislocation
Vacancy sink or
Jog displacement ≡
interstitial source
Figure 5 Schematic representation of the sink/source action of an edge dislocation from climb by jog displacement
along the dislocation line.
All the above definitions are valid for disordered breaking and reformation of chemical bonds. For
metallic materials. The situation is often more com- ionic compounds, electrical neutrality and electro-
plex for intermetallic, covalent, and ionic compounds. static interactions between ions of opposite charges
The ordered structure of intermetallic compounds results in quite complicated electrical effects for dis-
like b-NiAl leads to perfect superdislocations that locations. Figure 6 shows that a dislocation in a
dissociate to form antiphase boundaries bounded by simple 2D ionic crystal induces a local charge at the
dissociated dislocations. In covalent compound, most dislocation emergence. Furthermore, for the disloca-
of the binding energy is localized in the nearest- tion shown in Figure 6, the glide along direction x
neighbor bonds; the structure of glide dislocations is will be easier than the glide along direction z. Indeed,
thus complex and involves dangling bonds so that the the glide along x maintains a separation between ions
glide of such dislocations results in the successive of same sign, which is not the case of glide along z
Z direction. Kinks and jogs are also observed for dis-

locations present in ionic materials; dislocation climb
by the interaction of dislocations and point defects
remain the rule, which is of great importance for high
X temperature deformation (creep) and mass-transport
processes in ionic compounds.
1.06.2.3 Surfaces
Mass and energy exchanges between a solid phase
(a)
and its environment always proceed from its surface.
Many processes such as, for example, adsorption and
segregation, surface chemical reactions or else nucle-
ation and crystal growth depend on the structure and
properties of surfaces.6 Atoms, molecules, or ions on a
X crystal surface have not the same number of nearest
neighbors than in the bulk of the crystal; the surface
of a solid crystal phase is thus a discontinuity, and
consequently, a bidimensional crystal defect. More-
over, such a difference in surface and bulk environ-
(b) ment is at the origin of surface energy g, expressed in
Joules per square meter, or surface tension (then
expressed in the equivalent unit of N m1). For a
solid, surface energy depends on the surface orienta-
tion; this dependence determines the equilibrium
Z shape of the crystal by minimizing its total surface
energy.
The most common model of ideal and unrelaxed
surfaces for solid crystal phases is the terrace–ledge–
kink (TLK) model schematized in Figure 7. Accord-
(c) ing to this model, surfaces are classified into three
categories: singular, vicinal, and rough. A singular
Figure 6 Dislocation glide in an ionic compound: (a)
possible compact glide directions, (b) dislocation formed by surface corresponds to a surface orientation perfectly
glide along z direction, and (c) dislocation formed by glide parallel to a crystal plane of low Miller index. Such a
along x direction. singular surface is formed by one single terrace, that is,
Ledge
Terrace
(a)
L Kink
q = dhkl / L
(b)
dhkl (hkl)
Advacancy Screw dislocation emergence

Adatom
(c)
Figure 7 TKL model of crystal surface: (a) definition of terraces, ledges, and kinks, (b) relation between surface
disorientation and ledge spacing, and (c) surface point defects.
a surface essentially smooth on an atomic scale. Vici-

b
nal surfaces have an orientation close to a singular
surface. For disorientations along only one direction,
the departure from the exact orientation is accom-
b
modated by one set of ledges, that is, steps parallel to d=
q
a low-index crystal direction and delimiting succes-
sive terraces. The height of ledges is usually equal to
the interplanar spacing of reticular planes parallel to
the terraces; thus, the height and spacing of ledges q
can be easily related to the angular disorientation of
the vicinal surface relatively to the singular orienta-
tion of terraces. For vicinal surfaces disoriented in
two directions, the departure from the ideal orienta-
tion of terraces is accommodated by ledges and kinks Figure 8 Formation of a parallel array of edge dislocations
forming single steps along the ledges. The spacing at low-angle tilt boundary.
between successive kinks is similarly related to the
surface disorientation.
Ledges and kinks have an important role in sur- simplest grain boundary structure is obtained by
face reactivity; as schematized in Figure 7, adsorp- joining two similar vicinal surfaces disoriented in
tion or segregation on a terrace or at a ledge or a kink one direction; that is, only formed of terraces and
are not equivalent. Similarly, ledges and kinks are ledges, which, as shown in Figure 8(a), is similar to a
important in crystal growth processes and, more gen- symmetric tilt rotation of axis parallel to the ledge
erally, surface defects play an equally important direction. The resulting grain boundary is a low-
role in surface reactivity. The main surface defects angle tilt boundary that is formed by a set of parallel
are also described in Figure 7. The formation of and equidistant edge dislocations of Burgers vector
these defects is thermally activated; more defects normal to the boundary plane (Figure 8(b)). One
are formed as temperature increases, which results other simple structure is obtained by joining two
in the roughening of the surface becoming disordered similar singular surfaces slightly disoriented in a
and more reactive. rotation of axis normal to these surfaces, which lead
to a low-angle twist boundary formed by an array of
screw dislocations of Burgers vectors parallel to the
1.06.2.4 Grain Boundaries and Interfaces
boundary. From these two limiting cases, any low-
Grain boundaries and interfaces are surfaces where angle grain boundary, also called subboundary, may
two single crystals of same nature and different ori- be described by a more or less complex array of edge
entation (grain boundary) or different nature and and screw dislocations. For these subboundary mod-
orientation (interface) join in a manner permitting els, dislocation spacing is inversely proportional
some kind of matter continuity across these sur- to disorientation angle y so that they cannot hold
faces.7,8 Such a surface introduces local discontinu- when dislocation spacing is too small as that arise
ities and thus must be considered as a bidimensional for y > 0.10 rad.
defect. By their shape, size, arrangement, and combi- However, the variation of grain-boundary energy
nation, grain boundaries and interfaces are important as a function of a given disorientation angle shows
elements of the microstructure of materials and thus minima for specific large disorientations. The sim-
influence their properties. Microstructure is a non- plest cases of such stable high-angle boundaries are
equilibrium property of materials and as such is very twins similar to those usually observed for fcc or hcp
sensitive to temperature, chemical composition, metals. Several models were proposed to describe the
defect nature and concentration, mass transport pro- structure of more general high-angle grain bound-
cesses, time, etc. aries.7,8 Dislocation models based on coincidence
Similar to surface energy, grain-boundary energy site lattice (CSL) are well adapted to high-angle
results from the difference in local chemical environ- boundaries near a twin boundary, but the description
ment between grain boundary and crystal, which in of grain boundary structure by polyhedral structural
turn depends on the mutual orientation of the two units is more realistic. The important point here is
adjacent crystals. For small disorientation, the that, in addition to the orientation dependence
of grain-boundary structure and energy, there is a heterogeneous interface can be modeled by a combi-
corresponding dependence of grain-boundary diffu- nation of misfit and disorientation dislocations,
sivity upon disorientation. ledges and kink, and also a specific interfacial defect
The modeling of the structure of heterogeneous named disconnection9 and symbolized by a dissym-
interfaces separating two different crystal phases is metric dislocation symbol. Disconnections are
more complex. These interfaces are commonly clas- defects specific to heterogeneous interface (Figure
sified into the three categories of coherent, semico- 10(b)), they have partly dislocation character and
herent and incoherent interfaces. The coherent and partly ledge character and are formed at the junction
semicoherent interfaces correspond to specific mu- of two half planes of the same orientation; disconnec-
tual orientations of the two adjacent crystal phases, tions participate in lattice misorientations.
which implies the parallelism of lattice planes and Surfaces, grain boundaries, and interfaces can also
crystal directions. Independent of their structure, the act as sources and sinks for vacancies and intersti-
mutual orientation expresses that two lattice planes tials.10,11 Like for edge dislocations, the displacement
and two crystal directions belonging to the specified of kinks along a ledge or of jog along grain boundary
lattice planes are both respectively parallel. For two or interface dislocations or else the displacement of
crystals noted 1 and 2, such an orientation relation- interfacial ledges and kinks would result in the crea-
ship is expressed in the following form: (h1k1l1) tion or annihilation of vacancies or interstitials.
[u1v1w1]//(h2k2l2)[u2v2w2], where the plane and di- Remarks: The intersection of three grain boundaries,
rection indexes for crystal i must be such that hiui þ named a triple junction, is a complex line defect. The
kivi þ liwi ¼ 0 (i ¼ 1,2) to express their parallelism. structure and properties of these line defects are
Figure 9 shows that, for the simplest case of two neither well known nor understood and appreciated.
sets of parallel lattice planes of differing spacing, But, these defects are more open and disordered than
the disregistry of lattice planes can be accommodated dislocation or grain boundaries. Therefore, their role
by a parallel set of interfacial edge dislocations of and contribution to mass transport is likely less negli-
Burgers vector parallel to the interface plane. These gible than considered till now, particularly for fine-
interfacial dislocations are named misfit dislocations. grained materials such as oxide scales grown at
But, as lattice plane spacings are usually incommen- intermediate temperatures where bulk diffusion is
surable, a small elastic deformation of both lattices is not predominant. Similar observations apply to triple
often required to permit their accommodation in the junction line defects formed at the intersection of one
region of good fit, as exaggeratedly shown in Figure 9. grain boundary and one solid–solid interface.
The resulting stress field can influence the interface
behavior and properties.
A relative twist rotation of the two sets of lattice
planes can be accommodated by disorientation dis-
locations of Burgers vectors lying in the interface 5 2
plane, that is, of screw character. More generally, 1
3 4
as schematically shown in Figure 10(a), any
(a)
d1
d2
L(1)
Interface (b)
L(2) Figure 10 Interfacial defects and disconnection: (a)
interfacial line defects at a semi-coherent metal–oxide
interface: 1, misfit dislocation; 2, misorientation dislocation
with an extra half-plane in metal lattice; 3, misorientation
dislocation with an extra half-plane in oxide lattice; 4, ledge;
Figure 9 Misfit dislocations formed at the interface 5, disconnection. (b) Disconnection equivalent to a
separating two parallel lattices of different interplanar misorientation dislocation of Burgers vector modulus equal
spacings. to d2 d1.
1.06.3 Point Defects in Ionic Crystals 1.06.3.1 Structure and Building Elements:
Kröger–Vink Notation
Binary ionic compounds are formed by interpene-
The commonly recommended and well-accepted
trating anionic and cationic sublattices that contain
notation for point defects has been suggested by
point defects. Several textbooks provide a detailed
Kröger and Vink and is, therefore, called Kröger–
treatment and analysis of point defect equilibriums
Vink notation. This notation is a symbolic represen-
and defect-dependent properties of ionic compounds
tation of structure elements of defective crystals.
and metal oxides.12–15
A structure element is an atom, ion, molecule, or
Two kinds of point defect association permit to
else vacancy or interstitial located on a specified
conserve the stoichiometry of a compound as illu-
lattice site of the perfect crystal lattice. The Kröger–
strated in Figure 11 for an MX compound where
Vink notation provides the following information on
M is a metal and X an oxidant. One association is the
any kind of structure element:
Frenkel defect or disorder consisting of the associa-
tion of one cation vacancy and one interstitial cation. 1. the nature of the chemical species occupying one
The other one is the Schottky disorder associating given lattice site in the real crystal,
one cation vacancy and n anion vacancies for an MXn 2. the nature of the regular occupant of this given
compound. These two disorders are intrinsic defects. lattice site in the perfect crystal,
Indeed, an MXn compound maintained at its ideal 3. the effective electrical charge of the structure
stoichiometry contains only Frenkel and/or Schottky element.
disorders, but loses one degree of freedom so that
The effective charge of a structure element is the
the disorder concentration or its mole fraction only
difference between its charge in the real crystal
depends on temperature. As for pure elements, these
and its charge in the neutral perfect crystal lattice.
thermal equilibriums require operative sources or
Positive effective charges are noted by dots and neg-
sinks of point defects. Departure of MXn compounds
ative effective charges by commas or dashes; the
from ideal stoichiometry arises from reactions with
sign ‘’ is used for a null effective charge. A vacancy
any phase – solid, liquid, or gaseous – containing
is represented by the symbol V and an interstitial
M or X in a form different to MXn; these reactions
is designed by the index i affected to any species
obviously occur at the interface between MXn and
occupying an insertion site. The sites of interstitial
one of these other phases.
sublattice are all noted Vi as they are always empty
The thermal equilibrium of stoichiometric MXn
in the perfect crystal. All these notations are illu-
compound as well as the creation/annihilation of
strated and explained in Figure 12 while the notation
point defects from interfacial reactions can be treated
of all possible neutral and charged single point
with the usual concept of chemical reactions and the
defects of ionic compound MXn is reported in
associated thermodynamic tools, that is, law of mass
Table 1. The use of Kröger–Vink notation is not
action, chemical potentials, and activities, but such a
restricted to ionic compounds; it can be extended
treatment requires a specific notation for point
to any kind of compounds; for nonionic crystal,
defects. This notation must permit to express reac-
the sign ‘’ for null effective charge is then usually
tions balanced with respect to conservation of matter,
omitted.
electrical neutrality, and lattice site ratio fixed by the
The reaction of formation of intrinsic defects or
exact stoichiometry of the considered compound.
the interfacial reactions between an ionic crystal and
its environment often result in the formation or
O2– destruction of one or more building elements of the
M2+ VM considered compound. A building element is a com-
IM bination of structure elements that can be added or
VM IO removed without affecting the ideal stoichiometry
IM or, in other words, conserves the crystal structure by
VM keeping the ratio between the different lattice sites
VO VO
constant. For example, M
M þ OOi or VM þ OO are
additive building elements of oxide MO. Building
elements relative to only one sublattice can also be
Figure 11 Intrinsic point defects of a MO oxide. defined; they must keep constant the total site
Form of Kröger-Vink notation : Sqpeff
S = chemical symbol of the species localized on the considered lattice site in the real crystal
P = chemical symbol of the species localized on the considered lattice site in the perfect crystal
Additional symbols : V = vacancy
i = interstitial
Vi = empty interstitial site
Effective charge
qeff
Species occupying the S
considered crystal site
P Site occupancy in a
perfect crystal
Effective charge = Site charge in the real crystal – Charge of the same site in the perfect crystal
Notation of effective charge : Positive effective charge : dot

Negative effective charge : comma
Neutral effective charge : cross
Additional symbols : h· = electron hole
e⬘= free electron
Figure 12 Kröger–Vink notation of point defects.
Table 1 Kröger–Vink notation of point defects in MOn 1.06.3.2 Use of Kröger–Vink Notation:
oxide Formation Equilibriums of Frenkel and
Defect type Kröger–Vink notation Schottky Disorders
Single point defect The aim of this section is to illustrate the use of
Free electron e0 Kröger–Vink notation on simple examples of Frenkel
Electron hole h_ and Schottky disorders formed from only neutral
ðq2nÞ
Cationic impurity (charge q>2n) IM defects and to underline some difficulties and limits
ð2nqÞ0
Cationic impurity (charge q<2n) IM of this notation.
Neutral cation vacancy VM The formation of cationic Frenkel disorders in
0
Ionized cation vacancy VnM (n 2n) MXn only affects the cation sublattice. By using
Self-interstitial cation Mn i Kröger–Vink notation, the formation reaction of
Interstitial cationic impurity Iqi
Anionic impurity (charge>2) IO
ðq2Þ0 this disorder is
Anionic impurity (charge<2) IO
ð2qÞ
M
M þ Vi ¼ M i þ VM ½3
Neutral anion vacancy V O
Ionized anion vacancy VnO (n2)
A similar disorder may also affect the anion sublattice
0
Self-interstitial anion Oni (anionic Frenkel or anti-Frenkel disorder). Equation
0
Interstitial anionic impurity Iqi [3] fulfills the conditions of matter and lattice site
Combined defects (n ¼ integer number) balances. However, the law of mass action cannot be
0
V2nM þ nVO 0
2
Schottky
Frenkel (or Frenkel-cation) Mi þ V2n
2n directly applied to eqn [3]. Indeed, like for an ionic
M 0
Anti-Schottky M2n i þ nO2n
i
solution for which it is not possible to vary the amount
20
Anti-Frenkel (or Frenkel-anion) Oi þ VO 2
of only one ionic species while keeping all the other
ones constant, it is not possible to maintain constant the
number of regularly occupied cationic sites or the
number of empty interstitial sites of the cation sublat-
number of the considered sublattice to maintain the tice and to simultaneously create a cation vacancy or an
crystal structure and its stoichiometry; for the cation interstitial cation. Equation [3] shows that the chemical
sublattice of MO, subtractive building elements such potential, and thus the activity, of individual structure
as V
M MM or Mi Vi are compatible with this elements cannot be defined and expressed in their
constraint. usual form. However, eqn [3] can be transformed to
0 ¼ ðM
i Vi Þ þ ðVM MM Þ ½4 KSV ¼ xVM :ðxVX Þn ½8
where 0 represents the perfect lattice. The chemical KSV is the equilibrium constant for Schottky disorders
potential and activity of subtractive building ele- involving vacancies (KSi would be used for anti-
ments appearing in eqn [4] can then be defined and Schottky disorders).
thus the law of mass action can be applied to this Therefore, for the sake of simplicity, but keeping
equation leading to in mind that such a procedure is not thermodynami-
cally correct, the law of mass action will be applied to
KFC ¼ aðMi Vi Þ aðVM MM Þ equilibrium reactions, written by using the Kröger–
where KFC ¼ exp(Dfg/RT) is the equilibrium con- Vink notation with the restriction that the defect
stant for the formation of cationic Frenkel disorder and mole fractions remain sufficiently small to allow the
Dfg the molar free energy of formation of this disorder assumption of ideal behavior.
at temperature T. The number of each subtractive
building element is equal to the number of each asso-
1.06.3.3 Electronic Defects and Ionized
ciated point defects. Therefore, by assuming that the
Point Defects
mole fractions of cation vacancies and interstitials are
very small, an ideal behavior may be assumed and the In semiconductors and ionic compounds, the energy
activity of the building elements can be taken as equal levels of the outer electrons form two energy bands: the
to their mole fraction and therefore to the mole frac- valence and conduction bands separated by a forbidden
tion of the associated point defects, so that band or gap. At 0 K, the valence band contains all the
outer electrons while the conduction band is empty.
KFC ¼ xMi xVM ½5
Similar to the formation of a Frenkel defect, as the
Although incorrect, the application of mass action temperature increases, the energy of electrons
law to eqn [3] leads to eqn [5] by assuming that increases up to a limit to permit the jump of electrons
defective and regular structure elements form a from the valence band to the conduction band. Such a
dilute ideal solid solution, the activity of regular jump creates a free or excess electron occupying
structure elements (solvent) is then equal to 1 while a regular empty level of the conduction band and a
the activity of defective structure elements (solutes) vacant electronic level, or electron hole, in the valence
is equal to their mole fraction. band. Free electrons and electron holes are mobile
Both cationic and anionic sublattices are involved electronic defects of, respectively, negative and positive
in the formation equilibrium of Schottky disorders. effective charges. Thus, according to Kröger–Vink
The following formation equilibrium expresses that n notation, free electrons are noted as e0 and electron
independent anionic structure elements and one cat- holes – also called positive holes – are noted as h_. The
ionic structure element combine to form a new MXn equilibrium of electronic defects is then:
unit and equivalent number of vacancies
0 ¼ e0 þ h_ ½9
M
M þ nX
X ¼ V
M þ nV
X þ MXn ½6
The equilibrium constant Ke is:
or, in the form of subtractive building elements
Eg
Ke ¼ np ¼ exp
0 ¼ ðV
M MM Þ þ nðVX XX Þ þ MXn ½60 RT
But, eqn [6] can be simplified by considering where n and p are, respectively, the molar fraction of
that there is no formal difference between the MXn electronic defects e0 and h_, and Eg the energy gap.
unit and the additive building element M
M þ nXX , Point defects and electronic defects can interact to
which leads to the following equilibrium reaction for form ionized point defects. Compared with perfect
the formation of Schottky disorders: crystal, point defects are in an excited state of higher
energy: their outer electrons occupy energy levels
0 ¼ V
M þ nVX ½7
localized in the forbidden bands. There are two
A similar equilibrium, usually named anti-Schottky kinds of such levels: the donor levels close to the
disorder, involves only interstitial defects. Under the conduction band and the acceptor ones close to the
same restrictive conditions as for Frenkel disorders, valence band; the donor levels easily release electrons
the mass action law applied to eqn [60 ] or to eqn [7] to the conduction band while acceptor levels easily
leads to accept electrons from the valence band.
For example, a neutral cationic interstitial Mi semiconductors if the dominant electronic defects are
induce a local excess of electrons and then acts as a positive holes and inversely n-type semiconductors if
donor. For a monovalent metal, the ionization equi- free electrons are dominant.
librium is
1.06.3.4.1 p-Type oxides
M
i ¼ Mi þ e
0
½10
Common p-type oxides present a metal deficit or an
For a divalent metal, there are two ionization equili- oxygen excess. Nickel oxide Ni1–dO is a typical
briums, one similar to eqn [10] and the other example of a metal-deficient oxide. Over a large
Mi ¼ M 0
½100 range of oxygen partial pressure, numerous experi-
i þe
mental evidences have revealed that doubly charged
00
On the contrary, a neutral cation vacancy induces a Ni cation vacancies VNi are the predominant point
local electron deficit and acts as an acceptor defects in nonstoichiometric Ni1–dO. The formation
= of these vacancies can be represented by the follow-
V
M ¼ VM þ h_ ½11
ing equilibrium:
Similarly, a neutral anion vacancy acts as a donor and 1
a neutral anionic interstitial as an acceptor O2 ðgÞ ¼ O 00
O þ VNi þ 2h_ ½14
2
V 0 The dependence of molar fraction, ðxV00 Ni ¼ dÞ, of
X ¼ VX þ e ½12
V00Ni and h_ on PO2 is obtained from the mass action
=
X
i ¼ Xi þ h_ ½13 law applied to this formation equilibrium
xV00 Ni p2
Other ionization equilibriums are possible depending KVC2 ¼ 1=2
½15
on the oxidation state of the metallic and/or oxidant PO2
constituents of MXn. The mass action law is applicable where KVC2 is the equilibrium constant. The condi-
to all these ionization equilibriums, which permits, tion of electrical neutrality implies
in combination with the equilibrium of electronic
defects and the condition of electrical neutrality, to 2xV00 Ni ¼ p
determine the equilibrium molar fractions of ionized These two relations give
and electronic defects.
KVC2 1=3 1=6
xV00 Ni ¼ PO2
1.06.3.4 Point Defect Equilibriums in 22
Nonstoichiometric Oxides This result can be extended to cation vacancies of any
For any oxide MOn, departure from ideal stoichiom- charge q 2n that can appear in nonstoichiometric
etry results in an O/M ratio not equal to n. This ratio M1 dOn oxides:
is greater than n for metal deficit (M1dOn) or oxygen
KVCq 1=ðqþ1Þ n=2ðqþ1Þ
excess (MOn+d) and smaller than n for metal excess xVq= ¼ P O2 ½16
(M1+dOn) or oxygen deficit (MOnd). The departure
M qq
from stoichiometry d, and thus the equilibrium molar and
fractions of ionic point defects and electronic defects, p ¼ qxVq=
depends on temperature and oxygen or metal activ- M
ity. But, d is usually expressed as a function of T and The charge of predominant defects may vary with PO2
oxygen partial pressure PO2 . depending on the free energy of formation of these
A specific type of point defect is commonly pre- defects or, which is equivalent, the equilibrium con-
dominant in nonstoichiometric oxides such as nickel stants associated to their ionization equilibrium. For
oxide Ni1–dO or zirconia ZrO2–d. Equations [10]–[13] example, the cation vacancies of p-oxide M1dO may
show that the ionization of point defects results in the be neutral, singly charged, or doubly charged, so that
formation of electronic defects; therefore, the pre- d ¼ xVXM þ xV0M þ xV00M . The molar fractions can be
dominance of one specific ionized point defect calculated from the three simultaneous equilibriums
induces the predominance of one specific type of 1
electronic defect. Thus, as the mobility of electronic O2 ðgÞ ¼ OXO þ V M
2
defects is much higher than that of ionic defects or 0
VM ¼ VM þ h_
species, most of nonstoichiometric oxides behave like
semiconductors. Nonstoichiometric oxides are p-type V0M ¼ V00M þ h_
and the condition of electrical neutrality interstitial cations is expressed by a relation similar to
eqn [18], with an equilibrium constant KiCn. Equations
p ¼ xV0 M þ 2xV00 M
[16] and [18] show that the mole fractions of ionic and
This last equation determines the two domains where electronic defects increase in p-type oxides and
one charged defect dominates decrease in n-type oxides, at increasing PO2 .
xV0M 2xV00M ) xV00M ¼ p=2
xV00M xV0M ) xV0M ¼ p 1.06.3.5 Intrinsic Defect Equilibriums Close
to Stoichiometry
This procedure, known as Brouwer’s approximation,
can be applied to any combination of ionized points The nature of the dominant defect may vary depend-
defects; it simplifies greatly the calculation of defect ing on PO2 . Some amphoteric p-type oxides can pres-
equilibriums and their representation. ent a metal excess at low PO2 and a metal deficit at
Oxides with oxygen excess caused by interstitial higher PO2 . Reverse situation may be encountered for
anions are also p-type semiconductors. However, n-type oxides with an oxygen deficit at low PO2 and
because of the large size of oxygen anions compared an oxygen excess at higher PO2 .
to metal cations, such p-type oxides are rather Considering the case of an oxide MO with pre-
uncommon in usual high temperature materials. dominant fully ionized cationic defects, the defect
equilibriums to be considered are:
1.06.3.4.2 n-Type oxides
formation of doubly charged cation vacancies (equi-
Zirconium oxide ZrO2–d and zinc oxide Zn1+dO
librium constant KVC2)
are typical of n-type oxides. For ZrO2–d, the forma-
tion of doubly charged oxygen vacancies is expressed 1
O2 ðgÞ ¼ O 00
O þ VM þ 2h_
by the equilibrium 2
1 formation of doubly charged interstitial cations
O 0
O ¼ VO þ 2e þ O2 ðgÞ ½17 (equilibrium constant KiC2)
2
The dependence of molar fraction of V 0
O and e on 1
M 0
M þ OO þ Vi ¼ Mi þ 2e þ O2 ðgÞ
PO2 is obtained from the law of mass action 2
1=2 and the electronic equilibrium (eqn [9]) as both
KV2a ¼ nxVO PO2
e0 and h_ are involved in ionic defect equilibriums.
and the condition of electrical neutrality The combination of these three equilibriums gives
n ¼ 2V
O
the equilibrium formation of a cationic Frenkel dis-
order (equilibrium constant KFC)
These two relations give
M 00
M þ Vi ¼ M i þ VM ½20
KVA2 1=3 1=6
xVO ¼
P O2 The condition of electrical neutrality is
22
where KVA2 is the equilibrium constant. 2xM
i
þ h_ ¼ 2xV00M þe 0 ½21
Again, this result can be extended to oxygen
At the stoichiometric composition, molar fractions
vacancies of charge q (0 q 2)
p and n of electronic defects are equal, p ¼ n ¼ Ke1=2 ,

KVAq 1=ðqþ1Þ 1=2ðqþ1Þ and similarly the molar fractions of ionic defects,
xVqO ¼ PO2 ½18 xM
1=2
¼ xV00 M ¼ KFC , which from eqns [16] and [18]
qq i
corresponds to (PO2 )stoi ¼ KiC2/KVC2. In the vicinity of
Oxides like Zn1+dO with charged interstitial cations their stoichiometric composition, such oxides could
are also n-type semiconductors. The formation equi- have different behavior and properties depending on
librium of doubly charged interstitials Zni is the relative magnitude of the equilibrium constants Ke
1 and KFC. Two different situations are possible:
Zn 0
Zn þ OO þ Vi ¼ Zni þ 2e þ O2 ðgÞ ½19
2 1. Ke
KFC ; in that case, xMi h_ and xV00M e0
where Vi is a vacant site of the interstitial cation and therefore the equality p ¼ n ¼ Ke1=2 hold on a
sublattice. Comparison of eqns [17] and [19] permits more or less large range of PO2 while xV00M and xMi
to easily understand that the mole fraction of charged remain small but vary as a function of PO2 ,
2. Ke KFC, which implies xMi

h_ and xV00M
e0 ; !1=2
1=2 xCr 2KVC2
then, the equality xMi ¼ xV00M ¼ KFC is verified xVNi ¼ Ni and p ¼
1=4
PO2
while p and n vary with PO2 . 2 xCrNi
In case (1), the electronic defects predominate; such Depending on xCrNi and PO2 , the molar fraction p can
oxides are intrinsic electronic oxides; their electrical reach very small values so that the intrinsic electronic
conductivity is purely electronic and independent of equilibrium should be taken into account. Free elec-
PO2 . For the opposite case (2), the ionic defects are trons may become predominant, resulting in an
predominant; these oxides are intrinsic ionic oxides. inversion from p-type to n-type oxide. A similar
Their electrical conductivity can present a significant analysis can be performed for the influence of mono-
ionic component but the electronic component is valent impurities such as Li or Na. A sufficiently
usually not negligible because of the higher mobility large amount of such impurities may induce an inver-
of electronic defects. sion of the type of major ionic defect from cation
Out of the vicinity of stoichiometric composition, vacancies to interstitial cations.
all the ionic and electronic defects coexist but their This analysis can also be extended to nonstoichio-
molar fraction strongly depends on PO2 . From eqns metric n-type oxides. A typical example is given by
[19] and [16], cation vacancies and electron holes, of zirconia ZrO2–d. Its tetragonal and cubic crystal
molar fractions h_ ¼ 2xV00M , are predominant at PO2 structures can be stabilized (SZ) or partially stabi-
significantly higher than (PO2 )stoi. Similarly, from lized (PSZ) by the addition of CaO or Y2O3. A large
eqns [20] and [18], interstitial cations and free increase in anion vacancies and an inversion from
electrons, of molar fractions 2xMi ¼ e0 , are predom- n-type to p-type are induced by these additions;
inant at PO2 lower than (PO2 )stoi. ionic defects are predominant, and SZ and PSZ are
extrinsic ionic oxides.
For all the possible situations, the effect of impu-
1.06.3.6 Effect of Impurities on Defect rities can be summarized as follows:
Equilibriums a. For p-type oxides, the addition of cations of hig-
The above simplified treatment of ionic and elec- her oxidation degree induces a decrease of positive
tronic defects in oxide only considers perfectly pure holes and an increase of cation vacancies for oxides
oxides, which is a rather uncommon case, particularly M1dOn, or of oxygen interstitials for MOn+d; cations
for oxide scale grown on a metallic alloy. Indeed, the of lower oxidation degree have the reverse effect.
molar fractions of ionic and electronic defects are b. For n-type oxides M1 + dO, the addition of cations
strongly affected by the presence of small amounts of higher oxidation degree induces an increase in
of impurities of different oxidation state. In some free electrons and a decrease in anion vacancies
instances, the effect of impurities may be so large for MOnd oxides or cation interstitials for M1+dO
that it can induce an inversion of the type of predom- oxides, cations of lower oxidation degree has the
inant defects. reverse effect.
The influence of a trivalent impurity, like chro-
mium, on the molar fraction of ionic defects in a
1.06.3.7 Extended Defects
nonstoichiometric p-type oxide such as nickel oxide
Ni1dO is a typical and well-known example. Chro- At sufficiently high molar fractions, the interactions
mium cations Cr3+ are in substitution in the cationic between point defects may lead to the formation of
sublattice and constitutes a new structure element more complex defects. The most common example is
CrNi . The condition of electrical neutrality is then wüstite, Fe1dO, where cation vacancies form
ordered defect clusters. However, the formation of
xCrNi þ p ¼ 2xV00 Ni ½22
these defect clusters is directly linked to the large
For Ni1–dO, the departure to stoichiometry d, and departure d (up to 0.15) of Fe1dO from stoichiome-
thus the mole fraction of electron holes, is lower than try. Such defect clusters are not observed in oxides,
103 at temperatures lower than 1300 C. Therefore, a such as CoO and NiO, of smaller stoichiometry
molar fraction of Cr higher than 102 is sufficient to departure; for CoO, d remains smaller than 102 up
induce a significant increase of vacancy and electron to about 1400 C and is 10–100 smaller for NiO.
hole molar fractions. Indeed, for a mole fraction Other complex defect structures are known to be
xCrNi
p, eqns [16] and [22] leads to present in oxides like TiO2 and derived suboxides
TinO2n1, and also in many other oxides of transition Ci

metals. However, up to now, these complex defects xt = <v> (t2 – t1)
and/or oxide structures have not been observed in
oxide scales grown from high temperature oxidation
of metallic materials.
1.06.4 Diffusion Laws and Mass t1 > 0 t2 > t1

Transport in Solid State
x
The growth of a single oxide scale or a complex corro-
sion scale both require the diffusion of molecular Figure 13 Diffusive and displacement components of flux
and/or ionic chemical species through the growing relative to the fixed frame (x,C).
scale and/or the diffusion within the oxidized or cor-
roded substrate of their chemical components toward
the scale–substrate interface as usually required for the broadening of a drop of colored solute in a slowly
selective oxidation of the less-noble substrate com- flowing fluid. The concentration profile of the colored
ponent or deeper into the substrate for the more solute at time t2 > t1 is the sum of the two contribu-
noble components. Diffusion is commonly driven by tions of eqn [23]: broadening of the concentration
the concentration gradients of diffusing species, as profile due to the diffusion under the only influence
expressed by diffusion laws. However, transport phe- of concentration gradient and displacement xt of this
nomena induced by internal or external forces can concentration profile with the fluid flow under a force
interact with concentration gradients to accelerate or such as, for example, the gravitation force.
slow down the motion of mobile chemical species. Use- The displacement rate n depends on the force f
ful definitions of terms specific to diffusion and matter acting on the moving species as expressed by the
transport in solid state are given in Kizikyalli et al.16 Nernst–Einstein relation
f
n ¼ D ½24
1.06.4.1 Diffusion Laws RT
Matter movement by diffusion and/or transport in a However, in many practical situations, there is no
phase is quantified by the flux of moving species, that is, external or internal force acting on the moving spe-
a number of species per time unit and per unit area cies. Equation [23] reduces to the common form of
normal to the movement direction. Referring this move- Fick’s first law
ment relatively to a defined reference frame, the total @Ck
Jk ¼ Dk ½25a
flux Jk of any species k along a direction x is given by @x
@Ck Equations [23] and [25a] are only useful in conditions
Jk ¼ Dk þ nCk ½23
@x such that @Ck =@x remains constant (stationary state)
Fluxes and concentrations are functions of time t and or can be easily assessed as in the growth of an oxide
abscissa x. Equation [23] is a general form of the well- scale on a metal when both metal–scale and scale–gas
known Fick’s first law evidencing the two components interfaces are assumed to remain in equilibrium
of total flux Jk (in mol m2 s1): (pseudo stationary state) as the oxidation proceeds.
For all other situations, the resolution of diffusion
1. the diffusion component is expressed by the term problems required to solve the Fick’s second law, the
Dk ð@Ck =@xÞ, where Dk (in m2 s1) is the diffu- general diffusion equation expressing the conserva-
sion coefficient of species k and Ck (in mol m3) its tion of matter:
local concentration;
@Ck @Jk
2. the transport component, induced by internal or ¼ ½25b
external forces, corresponds to the term nCk where @t @x
n is the average displacement rate of all species For a pure diffusion process and constant diffusion
moving relative to the reference frame. coefficient Dk, this equation transforms into
The relative influence of these two terms is illustrated @Ck @ 2 Ck
¼ Dk ½26
in Figure 13 that corresponds, for example, to the @t @x 2
In eqns [23]–[26], the use of molar concentration in such as NiO. These marked species can also exchange
the expression of the local composition is essentially a their position with adjacent vacancies of Ni lattice or
practical choice. A more formal, and physically more cation sublattice of NiO, thus leading to random
correct, expression links the diffusion coefficient Dk displacements and dispersion of marked species
to the chemical potential of species k. Indeed, Fick’s under the only influence of thermal activation. Such
laws apply only to systems assumed in local equilib- a diffusion process is named self-diffusion. Therefore,
rium, that is, such that the thermodynamic properties for a system remaining in local equilibrium, the self-
of all system constituents are determined from all the diffusion coefficient Dsd can be expressed as the
state variables required for a complete description of product of vacancy molar fraction xn and vacancy
the system equilibrium states, independent of the diffusivity Dn
local gradient of these state variables.
Dsd ¼ xn Dn ½27
Equation [26] can be solved analytically. The
solution depends on the initial and limit conditions; However, vacancy diffusivity Dn is not formally equiva-
many textbooks describe the most common solu- lent to a diffusion coefficient because vacancies are
tions.17–20 For the general case of concentration- nonconservative species of concentration remaining
dependent diffusion coefficients, the solving of the constant and equal to its equilibrium value everywhere
general diffusion equation [25b] required the use of as imposed by the assumption of local equilibrium.
numerical methods that are now implanted in many Equation [27] remains valid for the self-diffusion in a
types of mathematical software. given sublattice as, for example, the self-diffusion of
cation in the cationic sublattice of a ionic compound.
However, for the diffusion of marked species or dilute
1.06.4.2 Lattice Diffusion and Diffusion
impurities, the analysis is more complex than reflected
Mechanisms
by eqn [27]. Indeed, the environment of marked or
In many cases of diffusion in a crystal phase, one impurity species changes after their jump so that all
constituent forms an immobile sublattice that deter- jump directions are not equivalent for such species,
mines a fixed frame of reference where eqns [25] and which leads to the introduction of a correlation
[26] can be applied to diffusing species. For a given factor taking into account this effect that is neglected
crystal phase, site-to-site jumps of diffusing species in eqn [27].
are the most common elementary steps of diffusion The interstitial and vacancy exchange mechanisms
processes, but such jumps must preserve the major are the two basic mechanisms of lattice or bulk diffu-
characteristics of the crystal structure of the phase. sion. They are valid for many kinds of materials;
Therefore, the diffusion of atoms, molecules, or ions however, other elementary mechanisms involving
within a crystal phase is only possible in the pres- more complex defects or jump sequences may be
ence of crystal defects. The simplest situation is the encountered in specific situations such as, for exam-
diffusion of small interstitial impurities within a fixed ple, diffusion in intermetallic compounds, or diffusion
host lattice. Interstitial species then jump from one under high energetic fluxes as occurring in materials
interstitial site to one of the closest adjacent intersti- used in nuclear power plants. Such complex diffusion
tial sites. A common example of simple interstitial processes are out of the scope of this chapter.
diffusion is the diffusion of interstitial oxygen atoms Both interstitial and substitutional diffusion
dissolved in a-Ti or a-Zr during their oxidation at mechanisms involve atomic jumps requiring the
high temperature. The host lattice can be also overcoming of an energy barrier. Diffusion is thus a
assumed immobile for the diffusion of highly diluted thermally activated process and the temperature
substitutional impurities. One impurity atom, ion, or dependence of diffusion coefficient is then expressed
molecule occupying a lattice site having one vacancy by the well-known Arrhenius relation
for nearest neighbor can jump to exchange its posi-
QD
tion with this vacancy. Such a vacancy exchange D ¼ D0 exp ½28
RT
results in the motion of vacancies in a direction
opposite to the diffusion of substitutional impurities. where QD is the activation energy for the diffusion of
The dilute substitutional impurities can be replaced a given species. The order of magnitude of diffusion
by some specific marked species of one constituent of coefficients for solids varies in the range 1023 to
the host lattice. For example, radioisotope 63Ni can 108 m2 s1 depending on T. In metals, the diffusion
be introduced in pure nickel or nickel compounds of interstitial species, such as C, O, or N, is faster than
the diffusion of substitutional species. Diffusion in Initial

covalent or ionic solids is of several orders of magni- contact plane
tude slower than the diffusion in metallic materials. A-rich end Fixed inert
Diffusion coefficients are important and useful prop- markers
erties of materials, particularly at high temperature. B-rich end
Mobile inert
For self-diffusion in pure metals and some classes of markers
materials, empirical correlations permit a rough esti-
mate of the activation energy and the diffusion coeffi- (a) M
cient at melting point as a function of crystal structure
and chemical bond.21
The knowledge of diffusion coefficients allows Kirkendall shift
CA
the calculation of the nominal diffusion length,
JA
lD ¼ ðDt Þ1=2 , which provides an estimation of the JB
size of the domain affected by the considered diffu- Ci
ξ
sion process. Solids or diffusion couples of size much
(b)
greater than lD along the diffusion direction are K
usually qualified as infinite solids for diffusion along JV

two opposite directions, or semi-infinite solids for
diffusion along only one direction. JA
dJV /dx
1.06.4.3 Interdiffusion, Intrinsic Diffusion, JB

and Kirkendall Effect in Binary Solid x
Solutions (c)
In an inhomogeneous, disordered, substitutional bi-
nary solid solution, both constituents i (i ¼ A or B)
are usually able to diffuse, via the vacancy exchange
mechanism, within the unique crystal lattice of such a Pores
Ci
solid solution. This diffusion process is called inter- ξ
diffusion. The well-known experiments of Smigelkas
and Kirkendall for the diffusion in a-brass have (d)
demonstrated that Zn and Cu diffuse simultaneously Figure 14 Kirkendall effect in a homogeneous binary A–B
at different rates and that the faster diffusivity of Zn diffusion couple: (a) initial diffusion couple before diffusion
must be compensated by a displacement of the host annealing, (b) concentration profile and markers position
after diffusion annealing, (c) variation of diffusion fluxes and
lattice evidenced by the displacement of inert markers vacancy source/sink density (dJn/dx), and (d) pore formation
within the diffusion zone. This phenomenon, named in the absence of vacancy sink.
Kirkendall effect, is schematized in Figure 14 for
two constituents A and B of constant partial molar of the initial contact surface after diffusion annealing
volumes. In Figure 14(a), the initial position of the (Figure 14(b)) and, therefore, visualize the drift of
contact surface is marked by fixed inert markers. As lattice planes within the diffusion zone; they mark the
the partial molar volume of both constituents is origin K, also named Kirkendall plane, of the lattice-
assumed constant, the total volume of the diffusion fixed frame of reference, denoted K-frame. N- and
couple remains constant; then, the fixed markers K-frames are the more commonly considered frames
determine also the origin of the frame of reference of reference. The form of Fick’s first law and the
centered on the number of moles or number-fixed meaning of the corresponding diffusion coefficient
frame denoted N-frame, also defined as the Matano depend on the chosen reference frame; a detailed
plane M. Under this assumption of constant couple analysis can be found in diffusion textbooks.17,19
volume, this frame is immobile and its origin can also When using N-frame, only one diffusion coeffi-
be taken at one extremity of the diffusion couple, cient, denoted D e and named interdiffusion coeffi-
which justify the denomination of laboratory-fixed cient, is needed for the analysis of concentration
frame of reference sometimes used for the N-frame. profiles of both constituents and the calculation of
Moving inert markers determine the actual position interdiffusion fluxes Jei from Fick’s first law; D e is
a function of composition. Thus, the interdiffusion But local equilibrium implies that the vacancy concen-
flux of constituents i (i ¼ A or B), Jei must be tration remains equal to its equilibrium value within all
expressed by the diffusion couple. To achieve this requirement, the
density of vacancy sources or sinks must be greater or
Jei ¼ Ji þ vK Ci ½29 equal to the divergence of the equivalent vacancy flux
(Figure 14(c)). This action of vacancy sources and
Ji is the diffusion flux of i relative to the moving
sinks result in the lattice drift characteristic of the
lattice (K-frame), it is thus named intrinsic diffusion
Kirkendall effect in stress-free homogeneous diffusion
flux; Fick’s first law leads to Ji ¼ Di ð@Ci =@xÞ,
couples. For species such as chemical elements, a Fick-
where Di is the intrinsic diffusion coefficient of
ian diffusion flux is always the consequence of a gradi-
species i; and
ent of chemical potential or concentration but such a
vKCi is the transport term, vK is the displacement
gradient is not required for the apparent migration of
rate of the moving lattice relative to the N-frame.
vacancies that are annihilated at sinks and created
The conservation of matter in the N-frame implies elsewhere at sources and vice versa.
that JeA þ JeB ¼ 0. This relation combined with rela- On the side of the faster diffusing species, when
tion [28], applied to A and B constituents, lead to the the density of vacancy sinks is not sufficient, the
Darken equations difference in the diffusion fluxes of substitutional
@xA chemical species would induce local vacancy super-
vK ¼ ðDA DB Þ ½30a saturation and associated build-up of local stress
@x
states within the diffusion zone. Return to local equi-
e ¼ xB DA þ xA DB
D ½30b librium in a stress-free state would be achieved by the
where xA is the molar fraction of A. nucleation of pores leading to the well-known Kir-
The intrinsic diffusion coefficient Di is equivalent kendall porosity (Figure 14(d)). All the intermediate
to the self-diffusion coefficient Di of i in the crystal situations between complete or null Kirkendall shift
lattice of the considered solid solution. The Kirken- are possible depending on local stress states and
dall effect is one consequence of the difference in density, distribution and efficiency of vacancy sinks.
intrinsic diffusivities of chemical constituents of sub- Therefore, local stress states, composition-dependent
stitutional solid solutions (nonreciprocal diffusion). volume changes, geometry, and size of the diffusion
The relative displacement of N- and K-frames is zone are among the parameters and phenomena that
caused by the internal chemical forces resulting can interfere with nonreciprocal diffusion and must
from the interactions between the constituents of be considered in a detailed analysis and understand-
nonideal solid solutions, which lead to the following ing of the behavior of systems when such a diffusion
relation between Di and Di process occurs. Furthermore, when the partial molar
volume of solution constituents are not constant and
@ ln gi varies as a function of local composition, the total
D i ¼ Di 1 þ ½31
@xi volume of the diffusion couple varies continuously as
where gi and xi are the activity coefficient and molar the diffusion proceeds. In addition to the increased
fraction of constituent i. Therefore, when the two difficulty in the analysis of diffusion profiles and
constituents form an ideal solid solution, there is fluxes, such a volume variation is also a source of
no relative displacement of N- and K-frames, that local stresses that can interfere with diffusion pro-
is, no Kirkendall effect. cesses and system behavior.
This classical treatment of Kirkendall effect in
binary homogeneous systems assumes that the differ- 1.06.4.4 Diffusion in Multiphase Systems
ence between the intrinsic diffusion fluxes of the two
substitutional constituents is compensated by the Diffusion in multiphase systems is often only focused
action of local vacancy sinks and sources that main- on the growth kinetics of intermediate phases.22
tain the system in local equilibrium. This local action But, the Kirkendall effect is not restricted to
is formally equivalent to an intrinsic vacancy flux Jn diffusion couples constituted of one single phase;
permitting the conservation of lattice sites, which is it may also occur in diffusion couples associating
expressed by the relation several phases.
For a two-phase diffusion couple constituted of
JA þ JB þ Jn ¼ 0 two adjacent a and b phases of an A–B binary system,
the initial contact plane can be evidenced, as for a constraints related to interface reactions, diffusion
homogenous couple, by the position of a fixed inert fluxes and interface movement.
marker. This situation is more complex for mobile
inert markers that are linked to the displacement of
1.06.4.5 Interdiffusion in Ternary Systems
lattice planes. In a two-phase diffusion couple, the
shift of lattice planes induced by nonreciprocal diffu- Interdiffusion in ternary, and higher systems, is more
sion is different within the two contacting phases. complex than in binary systems. Fick’s laws remain
Therefore, the displacement of lattice planes in valid for the diffusion in such systems. But, for a
both phases cannot be referred to only one frame of nonideal solid solution, the flux and gradient of
reference, which implies that each phase requires one each component are coupled with the fluxes and
specific lattice-fixed frame of reference.23 gradients of all other components. Therefore, the
The difference in intrinsic diffusivity within both diffusion in ternary systems cannot be treated as a
a and b phases requires, as for a homogeneous diffu- simple extension of binary diffusion. The interactions
sion couple, the action of vacancy sink/source between the constituents of a nonideal solid solution
distributed within these two phases. But, the differ- imply that the diffusion flux of a given constituent
ences in the required vacancy sources and sinks in cannot be expressed from only one diffusion coeffi-
each phase implies that the same a/b interface must cient.18,19 A (2 2) diffusion matrix must replace the
be able to either create or annihilate vacancies single diffusion coefficient of binary systems. Fur-
depending upon the intrinsic diffusion coefficients thermore, the terms of this diffusion matrix depends
and the initial composition of contacting a and b on the choice of one of the constituents as solvent, but
phases.23 Vacancy creation or annihilation must be the concentration profiles of each constituent are
considered as an inherent interfacial activity in any obviously independent of this choice. The interac-
diffusion-driven phase transformation involving sub- tions between the diffusing species are expressed by
stitutional solid solutions or ordered compounds with the off-diagonal elements of the diffusion matrix and
diffusion via vacancies. Interfacial defects such as are the cause of the complex shape of concentration
misfit and misorientation dislocations would provide profiles and diffusion paths that are commonly
the vacancy sources or sinks required by such an observed in ternary diffusion experiments.
interface action.10,11 One additional complication in ternary diffusion
But, for a phase transformation totally or partially arises from the extra degree of freedom of ternary
driven by solid-state diffusion, the role of the inter- systems compared to binary systems. This extra
face is not limited to the creation or annihilation of degree of freedom permits the occurrence of two-
vacancies involved in the diffusion process, the interphase fields in diffusion couples which also results in
face has several other roles to play. For one phase large variation in the morphology and microstructure
growing at the expense of one other phase, the inter- of diffusion zones. Therefore, for multiphase diffusion
face needs to satisfy several interrelated requirements in ternary systems, the correlation between the phase
to accommodate: diagram, the diffusion path, the morphology and
microstructure of a diffusion zone are much more
the change in chemical composition and eventu-
complicated than for multiphase binary diffusion.22
ally in chemical bonding,
Equation [28] remains valid for diffusion in a
the mismatch in crystal structure and molar vol-
ternary, disordered, substitutional solid solution.19
ume, and
Like for binary diffusion, the difference in intrinsic
the difference in phase growth or recession rate
diffusivities of the constituents of a ternary solid
and interface displacement rate by annihilating or
solution requires the action of vacancy sources or
creating the point defects involved in the diffusion
sinks for maintaining the systems in local equilib-
processes.
rium, which means that the Kirkendall effect also
If one of these requirements is not realized, the inter- occurs in ternary diffusion. For diffusion in multi-
face and the reacting phases would not be in a state of phase ternary systems, interfaces would have to likely
local equilibrium. Additional constraints must then act as vacancy sources or sinks, but the situation,
be considered and, therefore, additional state vari- particularly the microstructure of diffusion zone, is
ables must be introduced to describe and quantify more complex than for binary systems and would
the properties of contacting phases and the system require a specific and detailed analysis that still
evolution. These constraints can be local stresses or need to be carried out.
1.06.4.6 Short-Circuit Diffusion flux Jt through such a slab is the sum of the solute
flux for each diffusion path
For most common materials, self-diffusion and dif-
fusion of solute elements cannot be limited to diffu- Jt ¼ Jb þ Jd þ Jgb þ Jtj
sion in bulk crystals for which point defects are the
For cross-sectional fraction areas fb, fd, fgb, and ftj of
only defects involved in diffusion processes. Internal
each diffusion path and constant diffusion coeffi-
defects like dislocations, grain boundaries or triple
cients, Fick’s first law lead to the following expression
junctions and interfaces are always present in common
of the total solute flux
materials and are characterized by a disturbed and
more open structure along which diffusion is easier. DC
Jt ¼ fb Db þ fd Dd þ fgb Dgb þ ftj Dtj
Thus, these internal defects are preferred diffusion w
paths leading to short-circuit diffusion. Local strains This relation shows that the total flux can be expressed
and stresses associated with these defects can be par- by using an apparent diffusion coefficient Dapp averag-
tially minimized or relaxed by the positive (attractive) ing the contribution of all the diffusion paths
or negative (repulsive) segregation of diffusing solutes.
Segregation of diffusing species often interferes with Dapp ¼ fb Db þ fd Dd þ fgb Dgb þ ftj Dtj ½32
diffusion processes and must not be neglected. This
The dislocation density rd and the average grain size
influence is well documented in the case of grain-
dgb must be known for the evaluation of fraction
boundary diffusion.24 Diffusion along free surfaces is
areas fd, fgb, and ftj and of the contribution of each
not considered here because surface diffusion pro-
diffusion path. Therefore, no quantitative analysis of
cesses and mechanisms are rather different compared
short-circuit diffusion can be performed without an
to internal diffusion. The influence of short circuits on
accurate analysis of these microstructural character-
the diffusion flux of a given solute is schematized in
istics. For the calculation of these fraction areas, a
Figure 15 for a slab of material such that the common
value of 0.5 nm is usually taken for the radius of
operating short circuits, that is, dislocations, grain
dislocation cores and the thickness of grain bound-
boundaries, and triple junctions, denoted by their
aries. The relative magnitude of these diffusion coef-
respective index d, gb, and tj, are parallel to the diffu-
ficients is usually in the following order
sion direction. Each short circuit is then characterized
by diffusion coefficients Dd, Dgb, and Dtj higher than D b < Dd < Dgb < Dtj
Db, the solute diffusion coefficient in the bulk phase.
Only bulk and grain-boundary diffusion are usually
Assuming that the boundary conditions are differ-
considered in the common situation encountered in
ent but constant on each slab side and not affected by
oxidation of pure metals or selective oxidation of alloys.
the emergence of internal defects, the total solute
Equation [30] is a generalization of the Hart’s equa-
tion. However, for non-steady-state conditions, the use
of Dapp for the calculation of diffusion fluxes or concen-
CI CII
tration profiles from Fick’s second law is only valid when
the nominal diffusion length of bulk diffusion greatly
exceeds the average grain size or the average dislocation
spacing, which is a common situation for the diffusion in
Grain boundaries fgbJgb
a growing oxide scale and/or its substrate. Triple junc-
tions are very efficient short circuits but their influence
is only noticeable for grain size smaller than about
Dislocations fd Jd
100 nm25; therefore, one can wonder if this contribu-
tion can be neglected for fine-grained oxide scales
Bulk
(1–fd –fgb–ftj)Jb often formed at low or intermediate temperatures.
The diffusion mechanisms along dislocations, grain
Triple junctions
boundaries, or triple junctions are neither completely
ftjJtj
known nor understood. These mechanisms are likely
based on atomic jumps involving vacancies and/or
interstitials although the meaning of point defects asso-
Figure 15 Schematic representation of total diffusion flux ciated with dislocations, grain boundaries, or triple
through a slab with different kinds of short circuit. junctions need to be clarified. Furthermore, the
experimental determinations of solute diffusion permit by moving internal defects can result in a large increase
only the measurement of a diffusivity parameter that of apparent diffusion flux. In addition, the analysis can
includes the diffusion coefficient, the segregation fac- be complicated by the occurrence of specific phenom-
tor and the thickness or cross-section area of the con- ena such as diffusion-induced grain-boundary motion
sidered short circuit. All these factors may interfere (DIGM) that can be observed in alloys or materials
and caution must be taken in using diffusion coeffi- experiencing large concentration changes across
cients calculated by taking arbitrary values of segrega- moving gain boundaries.26
tion factors or defect width or section. The situation is
even more complex for ionic compounds for which 1.06.5 Diffusion and Transport in
local space charges may also interfere with the diffu- Oxides
sion processes. Experimental determinations have
shown that diffusion along dislocations and grain 1.06.5.1 Chemical Diffusion in
boundaries are thermally activated and their activation Nonstoichiometric Oxides
energies are usually lower than that for bulk diffusion The diffusion process induced in nonstoichiometric
because short-circuit diffusion is easier than bulk oxides and ionic compounds by a chemical potential
diffusion. Therefore, as schematically illustrated in gradient is named chemical diffusion. In common
Figure 16 for grain-boundary diffusion and bulk nonstoichiometric oxides, there is usually only one
diffusion, short-circuit diffusion is predominant at type of dominant point defect localized on one spe-
low temperatures. It is often quoted that the transition cific sublattice. Therefore, the diffusion process is
temperature for the predominance of grain-boundary restricted to the cationic sublattice for dominant
diffusion is about 0.5Tm (K), where Tm is the melting defects such as cation vacancies or interstitials, and
temperature of bulk phase. However, it can be deduced to the anionic sublattice for dominant anionic defects.
from eqn [30] that this transition temperature would, For an oxide like M1dOn that contains d cation
in fact, depend on the value of fraction area fgb, and vacancy per mole of oxide, the cation concentration
thus of scale microstructure. CM is proportional to the departure from stoichiom-
Finally, it must be emphasized that experimental etry d : CM ¼ ð1dÞ=OMOn, where OMOn is the
measurements and usual treatments of dislocation and molar volume of oxide MOn. Thus, Fick’s first law,
grain-boundary diffusivities are valid for stationary applied in the laboratory frame of reference, implies
dislocations and grain boundaries. In many practical that the cation flux is proportional to the d gradient
situations encountered in high temperature oxidation
e @d
D
or corrosion, these defects would move and their size JeM ¼ ½33
and density would vary, sometimes strongly, as a func- OMOn @x
tion of time and temperature. In that case, solute drag By taking care of the expression for cation and anion
molar fractions in a nonstoichiometric oxide and
neglecting the diffusion coefficient of anionic species,
Log (D)
eqns [16], [30], and [31] lead to the following expres-

e
sion of D
Dgb
e ¼ ð1 þ qÞ DM
D ½34
d
Db where DM is the self-diffusion coefficient of cation and
q the absolute value of the effective charge of cation
vacancies. By introducing the cation vacancy diffusiv-
Dapp
ity DVM through eqn [27], the chemical diffusion coef-
ficient is then related to the cation vacancy mobility,
which does not depend on the vacancy molar fraction
e ¼ ð1 þ qÞD VM
D ½35
1/T(k)
The form of this relation is general and applies to any
Figure 16 Arrhenius representation of bulk, grain
kind of point defect by simply replacing DVM by the
boundary and apparent diffusion coefficients (the gb area diffusivity Dd of the predominant defect involved in
fraction is assumed constant). the diffusion process.
Therefore, from these relations between chemical where F is the Faraday constant (F ¼ eNA). The total
diffusion and defect diffusivity, transport in an ionic conductivity is the sum of the contribution of all the
compound can be expressed only in terms of fluxes of charged species. The mobility of electronic defects is
point defects that are the primary mobile species. largely higher than the mobility of ionic defects and
Concentration gradients and transport of charged thus the total conductibility is mainly electronic except
species (ions, points and electronic defects) induce in the case of some intrinsic ionic oxides. However, for
internal electric fields within ionic compounds nonstoichiometric oxides, the variations of electric
and oxides. For an internal electric field strength E, conductivity with oxygen partial pressure PO2 permit
the force f acting on transported species of charge q the determination of the type and charge of the pre-
is f ¼ qE. The flux of each species i can then be dominant ionic defect as the molar fractions of elec-
expressed from eqns [23] and [24] tronic and ionic defects vary similarly as a function of
PO2 (cf. Section 1.06.2.4). A detailed treatment of dif-
@Ci qi Ei
Jei ¼ Di þ Di C i ½36 fusion in oxides and its relation to electric conductivity
@x RT can be found in textbooks.1,12–15
But, the condition of electrical neutrality implies Predominant defects are involved in mass transport.
that there is no net current flow withinPthe oxide But minority point defects can be involved in other
scale, which lead to the equation: qi Jei ¼ 0. important properties of compounds. This is, for exam-
However, solving this equation by using the expression ple, the case of deformation by creep at high tempera-
[36] for the fluxes of all charged species leads again to ture of ionic compounds that involves dislocation climb.
eqn [35]. Dislocation climb in an oxide requires the simultaneous
When an external electric field is applied to an ionic movement of cationic and anionic species and is thus
compound or an oxide initially at equilibrium, the only controlled by the transport rate of the slowest species,
driving force for the diffusion/transport of charged that is, the one associated to the minority defect.
species is that induced by the electric field. The
Nernst–Einstein relation [24] leads to the expression 1.06.5.2 Diffusion Processes in Oxide
of the flux of each charged species that is related to Scales
their electrical mobility and to the expression of the
The diffusion and transport process occurring in an
contribution of the species i to the electric conductivity
oxide scale can be illustrated by considering the
Ci Di behavior of such a scale when the chemical potential
si ¼ q2 F ½37
RT of one scale constituent differs on both its faces. This
1 O + V + 2e⬘ 1 ?
OXO OXO + νVnM + νnh
2 2 Ö O
2 2
(PO2 )1 MOv (PO2)2 < (PO2)1 (PO2)1 MOv (PO2)2 < (PO2)1
OXO V nM⬘
O h X
e⬘ MM
1 O + V + 2e⬘ ? 1
OXO OxO + νVnM9 + νnh O
2 2 Ö 2 2
(a) (b)
Figure 17 Diffusion fluxes and defect creation/annihilation for an MOn oxide scale maintained in a constant oxygen
gradient: (a) n-type oxide with anion vacancies and (b) p-type oxide with cation vacancies.
situation is schematized in Figure 17 for an MOn scale and cation vacancies are continuously created
scale separating two environments of differing oxy- at the high PO2 side and annihilated at the low PO2
gen partial pressure. Both sides of the oxide scale are side, which result s in a constant flux of cation vacancies
assumed to be in equilibrium so that a gradient in the and an opposite constant flux of cations that causes
concentration of predominant defects is established the observed shift of the oxide scale toward the high
within the scale. Corresponding flux of point defects PO2 side.
can be easily calculated from eqns [33], [34], and [16] This analysis can be extended to the case of two
or [18] when only one type of point defect is pre- oxides, AO and BO, forming a solid solution (A,B)O
dominant. The calculation is more complex for oxi- where A and B cations are randomly distributed over
des in which anionic and cationic species are able to the cationic sites of their common crystal structure;
diffuse simultaneously. charged cation vacancies are assumed to be the predom-
In the case of predominant anion vacancies, as inant point defects in pure oxides and their solid solu-
shown in Figure 15(a), the cation sublattice is not tions. In an oxygen potential gradient, the same causes
affected by the oxygen gradient so that there is no producing the same effect, a shift of the crystal lattice
transport of cationic species and the oxide scale toward the high PO2 side is again observed. But, the
remains stationary. The reactions occurring on the mobility of A and B cations are usually not equal so
faces of the oxide scale can be represented by the that the faster cations, for example A cations, move
oxidation/reduction equilibrium [17] but the oxida- faster toward the high PO2 side, which results in an
tion reaction occurs only on the high PO2 and results enrichment in A at this side, regardless of the relative
in the annihilation of anion vacancies while the oppo- thermodynamic stability of AO and BO; this process is
site reduction reaction occurs on the low PO2 side and named kinetic demixing and results in a steady-state
expresses the formation of anion vacancies. A constant demixing profile established through the oxide scale.1
concentration gradient of oxygen vacancies (and free
electrons) is then established so that anion vacancies
1.06.5.3 Oxide Scale Microstructure and
are continuously created at the low PO2 side and
Diffusion Processes
annihilated at the high PO2 side, which results in a
constant flux of anion vacancy and a constant counter The growth of an oxide scale on a pure metal or an
flux of oxygen anions. Therefore, in such a situation, alloy fits into the general category of diffusion-driven
the oxide scale acts simply as a permeable membrane phase transformation. This general statement does
that controls the oxygen flux from the high PO2 com- not mean that diffusion is the only process governing
partment toward the low PO2 one. or controlling the growth kinetics. Indeed, many
This situation is different for an oxide with predom- other processes, linked to the role and movement of
inant cation vacancies (Figure 17(b)). Indeed, the oxi- gas–oxide and/or oxide–metal interfaces, can be
dation/reduction equilibrium expressed by eqn [14] involved in oxide scale growth, as briefly described
involves species belonging to both anionic and cationic in Section 1.06.3.5 for a general a–b transformation.
sublattices and results not only in the creation/annihi- The understanding of growth mechanisms and
lation of cation vacancies but also in the creation (oxi- kinetics of an oxide scale always requires the knowl-
dation reaction) of one new structural unit of the oxide edge of diffusion and transport modes within this
lattice on the high PO2 side and the destruction of one oxide scale. Such information can be obtained by the
such unit on the low PO2 side (reduction reaction). use of inert markers and/or diffusion tracers, but the
These simultaneous reactions of creation/destruction morphological and microstructural features of oxide
of structural unit of oxide lattice result in a net shift of scales can also provide quite useful information.
the oxide scale toward the high PO2 side.1 This transla- As for multiphase diffusion in a binary system, a
tion of the oxide scale can be seen as the movement of Kirkendall frame of reference (K-frame) can be
ledges operating the advance of oxide lattice planes on defined for the metal and oxide lattices. For pure
the oxidation side and their recession on the reduction metals or ideal solid solutions, the initial unoxidized
side, as schematically shown in Figure 17(b). This surface determines the location of metal K-frame
steady-state shift of cationic oxide scales under a con- (denoted KM) and the Matano plane (denoted M).
stant potential gradient of oxygen is an illustration of The location of oxide K-frame (denoted KOx)
the Kirkendall effect in ionic compounds. Again, a depends on the growth process.
constant concentration gradient of cation vacancies One difficulty for the oxidation of metals or alloys
(and electron holes) is established through the oxide is linked to the large volume change usually induced
by the transformation of a metal into its oxide. This oxidation of pure nickel (p-type oxide) or pure zirco-
volume change is characterized by the well-known Pil- nium (n-type oxide) in a temperature range such that
ling and Bedworth ratio (PBR) f. For the growth of the the bulk diffusion of cations or anions is the only limit-
oxide MOn, this ratio is given by f ¼ ðvMOn =nM Þ, ing growth process. In that case, the oxide scales are
where vMOn is the molar volume of the growing oxide, characterized by a homogeneous and regular micro-
assumed pure, and nM the partial molar volume of metal structure consisting of columnar grains for both NiO
M at the oxide–metal interface. and ZrO2 scales but with a smaller grain size, in a section
In many cases, this volume variation is totally parallel to the oxide–scale interface, for ZrO2 scales.
accommodated or assumed to be accommodated by an The difference in grain size is likely due to the location
elongation parallel to the growth direction. Figure 18 of lattice growth reaction. Indeed, NiO grains grew at
schematizes, for the oxidation of a pure metal, the three the oxide–gas interface where they are free to expand,
different situations that can be observed in such an ideal which explains also their facetted morphology. On the
case depending on the location of the reactions asso- contrary, for ZrO2 scales, the lattice growth reaction is
ciated with the growth of the oxide lattice; M defines localized at the oxide–scale interface, the expansion of
the position of Matano plane and KM and KOx the oxide grains is thus limited by the previously formed
position of Kirkendall planes, respectively associated oxide; furthermore, the first formed oxide is not affected
to the metal and oxide lattices. The thickness of oxi- by its subsequent growth, which explains that the exter-
dized metal is equal to xM and the oxide scale thickness nal morphology of such oxide scales reproduces the
xOx is then such that the ratio xOx/xM is equal to the initial microstructure and/or morphology of its metal
PBR f. For the three situations of Figure 18, the substrate. The smaller grain size of these scales may
recession of metal lattice obviously occurs at the be also due to the continuous nucleation of new oxide
oxide–metal interface. Figure 18(a) shows the relative grains at the scale–substrate interface and/or to local
location of M, KM, and KOx planes for a scale of pure stresses and strain associated with the large value of
p-type oxide for which the oxide lattice growth occurs at PBR f.
the gas–oxide interface. Figure 18(b) defines the same Figure 19 can help for a better understanding of
location for a scale of pure n-type oxide for which the the processes leading to the situation of Figure 18(c).
oxide lattice growth and the metal lattice recession Indeed, Figure 19 reveals the similar microstructures
occur simultaneously along the oxide–metal interface. of a NiO scale (Figure 19(a)), grown at 800 C,27
The typical examples corresponding respectively to the and of a CoSi layer formed in a Co2Si/CoSi2 diffu-
situations depicted by Figures 18(a) and 18(b) are the sion couple (Figure 19(b)).28 Both NiO scale and
KOx
M KM M KM
XOx XOxe
Oxide
XMe
KOx
XOxi
KOx
Metal
(a) (b) (c)
Mobile inert markers

Figure 18 Relative positions of M, KM, and KOx reference planes during the growth of an oxide scale by the oxidation of
a pure metal: (a) scale growth occurring only at the external gas–oxide interface, (b) scale growth occurring only at the
internal oxide–metal interface, and (c) scale growth occurring at both interfaces.
Co2Si 100 μm the location of KOx. Figure 18(c) then corresponds to

O2 a mixed case such that oxide lattice growth occurs
simultaneously at both gas–oxide and oxide–metal
interfaces. The oxide scale fraction growing at the
external gas–oxide interface has a thickness xoxe and
Columnar subscale
the fraction growing at the internal oxide–metal inter-
Markers CoSi
face a thickness xoxi. For diffusion-controlled scale
NiO
Equiaxed subscale
growth, the relative magnitude of xoxe and xoxi is directly
proportional to the relative magnitude of intrinsic dif-
5 μm Ni
fusivity of species involved in the growth process, as
observed in many examples of solid-state reactions.22
CoSi
For a pure metal, the position of M and KM planes
Figure 19 Similarity of the microstructure of a NiO scale always coincides with the position of the initially
grown on a high-purity nickel27 and of a CoSi layer grown nonoxidized metallic surface. Therefore, after the
between CoSi2 and Co2Si. Reproduced from Paul, A.; van growth of an oxide scale, the position of these refer-
Dal, M. J. H.; Kodentsov, A. A.; van Loo, F. J. J. Acta Mater.
ence planes can be accurately determined relative to
2004, 52, 623–630.
a fixed reference plane such as, for example, the
specimen mid-plane or any marked plane far enough
CoSi layer are characterized by a duplex microstruc- from the scale–metal interface supposed to be fixed
ture formed by two sublayers of, respectively, and not affected by the lattice movements induced by
equiaxed and columnar microstructure. scale growth. These displacements can be rationa-
The growth of the columnar CoSi sublayer com- lized by a factor G that defines the displacement x
bines the diffusion of Si from CoSi2 toward Co2Si and of any reference plane of the oxide scale as a function
its reaction at the Co2Si/CoSi. Simultaneously, the of the thickness of oxidized metal xm: x = Gxm30; the
counter diffusion of Co and its reaction at the CoSi– factor G is defined as positive for outward displace-
CoSi2 interface leads to the growth of the equiaxed ment. For the general case of simultaneous growth at
CoSi sublayer. Inert markers, initially located at the the external gas–scale interface and at the internal
Co2Si–CoSi2 interface, are always located along the scale–substrate interface, the factor G is given by
internal interface separating the columnar and
G ¼ fð1 fC Þ 1 ½37a
equiaxed sublayers and they define the location of
the Kirkendall plane of CoSi phase.28 The term fC in relation [37a] is the fraction of cat-
Such a duplex microstructure is commonly ionic scale growth; fC is equal to the ratio xoxe/xox.
observed for NiO scales grown at a temperature Relation [37a] is easily checked from the displace-
lower than 1000 C.27,29 For these duplex scales, ment of KOx plane for pure cationic (Figure 18(a)) or
inert marker location and 18O experiments have pure anionic (Figure 18(b)) growth.
shown that the growth of the external columnar sub- Figure 18 and relation [37a] remain valid for the
scale is associated to the outward diffusion of Ni selective oxidation of ideal disordered solid solutions;
cations and occurs at the scale–gas interface while the value of f is then that of the selectively oxidized
the inward diffusion of oxygen is involved in the constituent. Indeed, the total volume of an ideal solid
growth of the inner equiaxed subscale. Therefore, solution does not depend on the spatial distribution
the internal interface between equiaxed and colum- of its constituents because the partial molar volume
nar sublayers marks also the initial location of the Ni of each constituent is constant and equal to its own
surface before the formation and growth of NiO molar volume. Therefore, there is no difference
scales. These two examples clearly show that similar between pure metal and ideal solid solution regard-
growth processes lead to similar microstructures. The ing the definition and location of planes M and KM.
microstructure of a growing phase directly depends This situation is more complex for nonideal solid
on the nature of diffusing species and on the locali- solutions as the partial molar volume of each constit-
zation of growth reactions as observed in many exam- uent varies as a function of composition. Selective
ples of solid-state and solid–gas reactions and its oxidation of one constituent of a solid solution results
microstructure differs on both sides of its Kirdendall in concentration gradients of all constituents in the
plane.22,28 Therefore, metallographic examinations of alloy underlying the oxide scale and the resulting
oxide scales provide also an easy way to determine local variations of partial molar volumes is the
cause of local volume changes, which in turn results commonly found in high temperature oxidation or
in local deformation and stress fields. Unfortunately, corrosion of metallic materials.12–14 But, only few
an accurate evaluation of such volume changes and very common oxides such as NiO monoxide, iron
related deformation field requires data that are often oxides, chromia, and alumina are briefly considered
not available. Many oxide scales present a duplex in this section. By its importance in the manu-
microstructure similar to NiO scale of Figure 19. facturing of integrated circuits and its protective
This is the case of a-alumina scales grown on role in the oxidation behavior of some iron-based
FeCrAlY alloys oxidized at high temperature (T > materials, silica is also a very common oxide. How-
1000–1100 C); they are formed of an equiaxed ever, silica scales are amorphous and diffusion pro-
outer subscale and a columnar inner subscale.31 cesses in amorphous solids are not considered here.
Like for NiO scales, this duplex microstructure is Most of the studies devoted to point defects and
representative of the simultaneous occurrence of diffusion processes in oxides are often based on mea-
two growth processes, the surface separating these surements of electrical conductivity, scale growth
two subscales defines the location of Kox-plane, that kinetics or else concentration profiles of marked spe-
is, the initial location of the metal surface. But, the cies. For bulk diffusion, the equations reported in
microstructure of alumina scales grown on high tem- Section 1.06.5.1 permit the interpretation of experi-
perature alloys depends strongly on many parameters mental data regarding the diffusion mechanisms and
such as oxidation temperature, oxygen partial pres- the nature and ionization degree of defects involved
sure or alloy composition and more particularly of in diffusion processes. But, it must be kept in mind
the presence of reactive element such as Y, Zr, Hf, etc. that diffusion data obtained from oxide single crys-
Furthermore, the influence of these parameters on tals, very pure oxides and/or oxide samples of con-
the nature and concentration of predominant point trolled, homogeneous and regular microstructure are
defects and electronic defects, diffusion mechanisms, not always useful for the understanding and analysis
and processes is yet neither well known nor under- of diffusion and transport processes in oxide scales
stood. For many oxide scales, a detailed analysis and grown on complex metallic materials.
understanding of correlations between scale growth
microstructure, growth mechanisms and diffusion 1.06.5.4.1 Nickel and cobalt monoxides
processes is needed and is still far away to be possible. Both Ni1–dO and Co1–dO are p-type oxides of rock
Several experimental methods now commonly used salt structure. Cation vacancies and positive holes are
in solid-state electrochemistry can offer new insights the predominant point and electronic defects; the
in oxide-scale properties and scale-growth mecha- departure from stoichiometry d is significantly smal-
nism. This is notably the case of impedance spectros- ler for NiO. The self-diffusion coefficients of cations
copy32 and photoelectrochemical spectroscopy.35 were determined from tracer experiments and from
Impedance spectroscopy is a powerful tool that can the analysis of scale-growth kinetics at temperatures
permit to analyze the influence of many parameters higher than 1000 C. The agreement between these
such as scale composition, scale thickness, scale independent determinations is excellent. The growth
microstructure, and grain size. Photoelectrochemical of NiO and CoO scales at high temperatures is one of
spectroscopy permits the determination of the local the best examples of scale growth controlled by bulk
electronic properties of oxide scales, which is of diffusion according to the Wagner theory as demon-
particular importance for a better understanding of strated by the recent work of Mrowec and Gresik36 on
growth mechanisms. These techniques are particu- the oxidation of high-purity nickel at temperatures
larly interesting at temperatures where bulk diffusion higher than 1100 C. In this temperature range, NiO
is not the limiting process as shown by the results scales are characterized by a homogeneous micro-
obtained on the oxidation of zirconium alloys33 or structure consisting of large columnar grains such
pure nickel.34 that the contribution of grain-boundary diffusion
can be readily neglected. The following equations
express the departure from stoichiometry and the
1.06.5.4 Point Defects and Diffusion in
variation of self-diffusion coefficient of Ni cations as
Common Oxides
a function of T and PO2
Many works have been devoted to the study of bulk 1=6
PO2 9620
and grain-boundary self-diffusion in ionic com- d ¼ 0:153 exp ½38
pounds, particularly in oxides and sulfides that are P0 T
1=6
PO2 nearly independent of d; the formation of highly
DNi ðcm2 s1 Þ ¼ 9:3 103 stable defect clusters combining cationic vacancies
P0
and interstitials interferes with the diffusion pro-
27 900
exp ½39 cesses that are therefore more complex than in NiO
T
or CoO. The main point defect of anionic sublattice
Chemical diffusion coefficient D e in NiO can be are oxygen interstitials but the variation of oxygen
deduced from eqns [34], [38], and [39]; D e and cation self-diffusion as a function of temperature and PO2 is
vacancy diffusivity, calculated from eqn [35], are complex and not yet understood. In wüstite, oxygen
independent of PO2 and thus of the vacancy molar self-diffusion is several orders of magnitude slower
fraction. Diffusion data in NiO have been recently than iron self-diffusion.
reviewed by Monceau.37 Few measurements of self- Magnetite Fe3O4 has a spinel crystal structure
diffusion coefficient of oxygen anions were per- where the octahedral and tetrahedral sites of the fcc
formed at temperatures higher than 1100 C; their sublattice of oxygen anions can be occupied by Fe2+
values are more than five orders of magnitude smaller and Fe3+ cations. The distribution of Fe2+ and Fe3+
than self-diffusion coefficients of Ni. Other data were cations over octahedral and tetrahedral sites is a
obtained from high temperature creep behavior of function of temperature.
NiO, which is controlled by the diffusion of oxygen. Magnetite is stoichiometric at room temperature
At temperatures lower than 1000 C, the influence and Fe3+ cations are equally distributed over tetrahe-
of grain-boundary diffusion cannot be neglected.38 dral and octahedral sites while Fe2+ cations are only
The grain-boundary self-diffusion of cations and in octahedral sites. This distribution of Fe2+ and Fe3+
anions in NiO scales has been extensively investi- cations corresponds to the inverse spinel structure.
gated and discussed.27,29,37,38 However, the morphol- At temperatures higher than about 900 C, the dis-
ogy and microstructure of NiO scales vary strongly as tribution of Fe2+ and Fe3+ cations is randomized
a function of time and temperature27; therefore, the over tetrahedral and octahedral sites and the depar-
only influence of grain-boundary diffusion is likely ture from stoichiometry increases as temperature
not sufficient for a detailed understanding of experi- increases.
mental scale-growth kinetics and complex evolutions The predominant point defects are in the cationic
of oxide-scale morphology and microstructure at sublattice; their nature is thus defined relatively to
T < 1000 C. the normal site occupancy of one-half of octahedral
sites and one-eighth of tetrahedral sites but defects
1.06.5.4.2 Iron oxides localized on octahedral and tetrahedral sites are not
Three oxides can be found in iron–oxygen systems: equivalent and must be distinguished. Cation vacan-
wüstite Fe1dO, magnetite Fe3O4, and hematite cies and interstitials are localized on the same sites
Fe2O3. Pure wüstite is only stable at temperatures and are formed when the proportions of occupied
higher than 570 C and reduced PO2 . Therefore, sites are respectively smaller (vacancy) or higher
oxide scales grown on pure Fe above 570 C is, from (interstitial) than 1/2 and 1/8. The majority defects
pure iron substrate up to oxygen gas phase, formed of are neutral cation interstitials (Fe3+dO4) at low PO2
successive three sublayers of wüstite, magnetite, and close to the wüstite–magnetite equilibrium and neu-
hematite. The relative thicknesses of these sublayers tral cation vacancies (Fe3dO4) at higher PO2 close to
are almost independent of temperature and show the magnetite–hematite equilibrium. The formation
very clearly that diffusion is largely faster in wüstite equilibriums of these defects lead to defects concen-
2=3 2=3
(95% of the total scale thickness) than in magnetite tration varying as PO2 for vacancies and PO2 for
39
(4%) and hematite (1%). interstitials.
Like CoO and NiO, wüstite Fe1dO is a p-type The variation of iron self-diffusion coefficient as a
oxide of rock salt structure but it differs by its larger function of PO2 is similar and expressed by the fol-
departure from stoichiometry d that can reach a lowing relation40
value as high as about 0.15. The defects responsible
2=3
for Fe self-diffusion in Fe1dO are, like for NiO PO2 16 800
and CoO, cation vacancies but Fe self-diffusion in DFe ðcm2 s1 Þ ¼ 4 1011 exp
P0 T
wüstite is about four orders of magnitude faster 2=3
than Ni self-diffusion in NiO. Furthermore, contrary 7 PO2 73 800
þ 8 10 exp
to NiO and CoO, Fe self-diffusion coefficients are P0 T
The minimum of iron self-diffusion coefficients is not kinetics as a function of PO2 are complex and difficult
observed for the stoichiometric composition where to interpret and to compare, particularly because of
the molar fraction of cation vacancies and interstitials their rather wide scattering. However, many data are
are equal. This minimum corresponds to a PO2 value consistent with a defect structure based on the pre-
such that vacancies molar fraction is higher than dominance of V000 Cr cation vacancies at PO2 close to
interstitials molar fraction, which means that inter- atmospheric pressure and on the predominance of
stitials are more mobile than vacancies. Oxygen self- interstitial cations Cr3
i at very lowPO2, close to
diffusion is two orders of magnitude slower than iron Cr2O3–Cr equilibrium, although anion vacancies
self-diffusion and varies also as a function of PO2 may be important minority defects at these low PO2
indicating that the nature of predominant anionic in agreement with experimental data showing that
point defects varies also withPO2 ; the interpretation bulk chromium and oxygen self-diffusion coefficients
of the observed variations involve the formation of are of the same order of magnitude.41 At intermediate
complex point defects at high PO2 . PO2 , free electrons and electron holes are the major
The crystal structure of hematite Fe2O3 is, like defects; Cr2O3 then behaves as an intrinsic electronic
Cr2O3 and a-Al2O3 of corundum type. Corundum semiconductor. These differences in defect structure
structure can be described as a close packing of as a function of PO2 must have a marked influence on
anions, cations occupying two-thirds of the octahe- the microstructure of Cr2O3 scales but there is much
dral sites. Fe2O3 is known to be an oxygen-deficient less work on the characterization of Cr2O3 scales than
oxide with very small departure from stoichiometry. on NiO or a-Al2O3 scales.
Its defect structure is very sensitive to metallic impu- Grain-boundary diffusion play also an important
rities; it changes from oxygen vacancies and n-type role in the growth of Cr2O3 scales, particularly at
conductivity to interstitial cations and p-type conduc- temperatures lower than 1000 C. Grain-boundary
tivity at about 800 C. Iron and oxygen self-diffusion diffusion of cations appears usually faster than
are apparently of the same order of magnitude but the anion diffusion. But more data on the influence of
works devoted to the influence of PO2 on defect struc- T, PO2 and RE on grain-boundary self-diffusion of
ture and self-diffusion in Fe2O3 cannot be interpreted chromium and oxygen are needed for a better under-
on the basis of simple defect structure and lead to standing of diffusion process involved in the growth
rather different interpretations. of Cr2O3 scales.
1.06.5.4.3 Chromia 1.06.5.4.4 a-Alumina (a-Al2O3)

Heat-resisting alloys are commonly classified as Alumina of corundum structure, a-Al2O3, is the most
chromia-forming or alumina-forming alloys accord- protective and stable oxide among all the oxides
ing to the nature of the protective oxide scale formed commonly encountered in high temperature oxida-
upon their high temperature oxidation. However, at tion. Therefore, a-Al2O3 is also the most important
PO2 close to atmospheric pressure, the oxidative oxide from the technological view point of oxidation
evaporation of Cr2O3 in volatile CrO3 chromic resistance and behavior of heat-resisting alloys.
oxide limits the uses of chromia-forming alloys to Unfortunately, despite all the work conducted over
about 1000 C. Cr2O3, like hematite, has the corun- its point defect structure and transport properties,
dum structure but a much more narrow range of there remain major problems in the understanding
nonstoichiometry because of its greater thermody- of a-Al2O3 properties and behavior that are still the
namic stability. Cr2O3 is an electronic semiconductor subject of discussion and controversy as illustrated by
that can be either a p-type or an n-type semiconduc- a recent review.42
tor depending on T, PO2 , and doping impurities. The The great thermodynamic stability of a-Al2O3
influence of these parameters on the nature of pre- results in a so narrow nonstoichiometry range that
dominant defect is complex and not yet fully under- there is no accurate determination of its extent. As a
stood. It is still the subject of many controversies, consequence, the transport properties of a-Al2O3, as
particularly for the role of reactive elements (RE) determined from electrical conductivity, tracer diffu-
such as Y, Ce, etc. that modify the transport properties sion or scale-growth kinetics, are controlled by extrin-
of Cr2O3, and therefore, the growth mode, the growth sic defects associated to cationic impurities. This
kinetics, and the microstructure of Cr2O3 scales. strong influence of impurities likely explains the
The variations of electrical conductivity, chro- wide scattering of experimental data usually observed
mium and oxygen self-diffusion and scale-growth among the numerous studies devoted to a-Al2O3.
One other characteristic of the transport proper- a-Al2O3, the development of such calculations
ties of a-Al2O3 is the large value of the experimental would be very useful and would lead to more consis-
activation energy reported for the bulk self-diffusion tent results and thus to a better understanding of
of O and Al. For oxygen diffusion, mainly studied by defect structure and transport properties of a-Al2O3.
means of 18O tracer, the reported activation energies
range between 570 and 640 kJ mol1. The absence of
natural Al isotope explains why there is only one 1.06.6 Concluding Remarks
study which used26 Al markers, an artificial isotope,
to measure the self-diffusion coefficient of Al.43 The Mass-transport processes are of fundamental impor-
reported activation energy of Al self-diffusion is also tance for the understanding and modeling of high
high and found to be equal to 510 kJ mol1, a value temperature oxidation and corrosion behavior of
close to the activation energy of O self-diffusion. materials but, as roughly shown in Section 1.06.5.4,
Surprisingly, the activation energy for the diffusion the defect structure and transport properties of most
of many foreign metallic cations is smaller, usually common and important oxides are far from being
comprised between 260 and 300 kJ mol1, a value well known and understood. The increased capability
close to the mean activation energy of a-Al2O3 of theoretical treatments would likely permit a more
scale-growth kinetics reported by Hou44 from a com- accurate modeling of chemical bonding, point defect
pilation of growth kinetics of numerous Ni-based and structure and intrinsic and extrinsic behavior of com-
Fe-based alumina-forming alloys. mon oxides in the coming years.
To add to the complexity of transport processes in However, very few scale-growth kinetics can be
a-Al2O3, the experimental activation energies of perfectly described from the only knowledge of defect
grain boundary self-diffusion of O and Al are signifi- structure and bulk transport properties of the growing
cantly higher than the activation energies for bulk scale. For example, the oxidation of high-purity nickel
self-diffusion, a rather uncommon situation. Indeed, in pure and dry oxygen is certainly the simplest situa-
the reported activation energies for grain-boundary tion that can be encountered in the field of high
diffusion are higher than 800 kJ mol1 for both O and temperature oxidation of metallic materials but, at
Al. But, despite all the studies devoted to the oxida- temperature lower than 1000 C, NiO growth kinetics
tion behavior of alumina-forming alloys, there is no are several orders of magnitude faster than expected
convincing explanation of the different activation from the extrapolation of high temperature data.
energies for oxidation and diffusion. The complexity Grain-boundary diffusion is currently invoked to
of a-Al2O3 growth is also reflected in the diversity of explain faster growth kinetics at intermediate tem-
microstructures observed for a-Al2O3 scales that can peratures; there are indeed many experimental evi-
be equiaxed, columnar, or duplex depending on T, dences confirming the major role of short-circuit
PO2 , and alloying elements such as RE or Pt. These diffusion in scale growth. But our understanding of
different microstructures must correspond to differ- the structure and properties of oxide grain bound-
ent growth mechanisms and diffusion processes. But, aries, particularly their defect structure and the influ-
a more detailed analysis of the correlation between ence of intergranular segregations, needs to be
scale microstructure, growth mechanisms and diffu- considered and greatly improved without ignoring
sion processes remains to be done. the influence of triple junctions. Furthermore, this
A better understanding of defect structure and knowledge would be useful for the analysis and mod-
transport properties of a-Al2O3 could be obtained eling of the complex time-dependent evolutions of
from theoretical calculations permitting the determi- oxide-scale morphology and microstructure that
nation of the enthalpy of formation of different point interfere with short-circuit diffusion.
defects. But, the various calculation methods and Surface and grain boundary are common planar
interaction potentials used in these calculations lead defects; the interface separating two crystal phases is
to somewhat differing results. The most recent cal- also a planar defect that, like surfaces and grain
culation has shown that Schottky defects, formed by boundaries, can play a role in the creation or annihi-
0
the combination 2V3Al þ 3V2 O , are the dominant lation of point defects involved in scale growth or in
intrinsic point defects in stoichiometric a-Al2O345 diffusion processes within the oxidized substrate.
but this result is not in agreement with previous There are many examples of solid-state reactions for
results leading to a predominance of Frenkel pairs.46 which solid–solid interfaces and interfacial reactions
However, because of the extrinsic behavior of play a major role.47 In the field of high temperature
oxidation, one well-known example is provided by 14. Smyth, D. M. The Defect Chemistry of Metal Oxides;
Oxford University Press: Oxford, 2000.
the oxidation of silicon. Indeed, molecular oxygen 15. Maier, J. Physical Chemistry of Ionic Materials: Ions and
diffuses through the thin silica layer and reacts with Electrons in Solids; John Wiley & Sons: New York, 2004.
the silicon substrate at the silica–silicon interface; 16. Kizikyalli, N.; Corish, J.; Metselaar, R. Pure Appl. Chem.
1999, 71, 1307–1325.
this interfacial reaction is the controlling step for the 17. Philibert, J. Atom Movements, Diffusion and Mass
growth of the very thin silica layers now required in Transport in Solids; Les Editions de Physique/Les Ulys
the manufacturing of integrated circuits. Therefore, 1991.
18. Kirkaldy, J. S.; Young, D. J. Diffusion in the Condensed
the roles of scale–substrate interfaces need to be State; The Institute of Metals: London, 1987.
more particularly analyzed in relation with some 19. Glicksman, M. E. Diffusion in Solids; John Wiley & Sons:
specific topics related to oxide-scale processes such New York, 2000.
20. Ghez, R. Diffusion Phenomena; Kluwer Academic/Plenum
as interface displacement, growth stresses and injec- Press: New York, 2001.
tion of point defects (vacancy or interstitial). In addi- 21. Brown, A. M.; Ashby, M. F. Acta Metal 1980, 28,
tion, the development of experimental techniques 1085–1101.
22. van Loo, F. J. J. Prog. Solid State Chem. 1990, 20, 47–99.
and methods permitting an accurate determination 23. van Loo, F. J. J.; Pieraggi, B.; Rapp, R. A. Acta Metal
of the relative movement of M, KM, and KOx planes Mater. 1990, 38, 1769–1779.
would help in the modeling of scale-growth mechan- 24. Mishin, Y.; Herzig, C.; Bernardini, J.; Gust, W. Int. Mater.
Rev. 1997, 42, 155–178.
isms. The development of such models combining 25. Wang, H.; Yang, W.; Ngan, A. H. W. Scripta Mater. 2005,
diffusion processes and interface action could lead 52, 69–73.
to a new approach to questions linked to the effect of 26. Shewmon, P. Diffusion in Solids, 2nd ed.; The Minerals,
Metals and Materials Society: Warrendale, 1989.
impurities and of their segregation, the effect of reac- 27. Peraldi, R.; Monceau, D.; Pieraggi, B. Oxid Metel 2002, 58,
tive element and thus a better understanding of the 249–273.
dynamical behavior of scale–substrate interface dur- 28. Paul, A.; van Dal, M. J. H.; Kodentsov, A. A.; van Loo, F. J.
J. Acta Mater. 2004, 52, 623–630.
ing scale growth, which could open up new opportu- 29. Haugsrud, R. Corr. Sci. 2003, 45, 211–235.
nities for the control of oxidation behavior of high 30. Manning, M. I. Corr. Sci. 1979, 21, 301–316.
temperature materials and the improvement of their 31. Nychka, J. A.; Clarke, D. R. Oxid Metel. 2005, 63, 323–352.
32. Maier, J. Solid State Electrochemistry: Thermodynamics
oxidation resistance. and Kinetics of Charge Carriers in Solids. In Modern
Aspects of Electrochemistry N 38; B.E., Kluwer
Academic/Plenum Publishers: New York, 2005.
References 33. Vermoyal, J. J.; Frichet, A.; Dessemond, L. J. Nuclear Mat.
2004, 328, 31–45.
34. Liu, H.; Mojica-Calderon, C.; Lyon, S. B.; Stack, M. M.
1. Schmalzried, H. Chemical Kinetics of Solids; VCH: Solid State Ionics 1999, 126, 363–372.
Weinheim, 1995. 35. Wouters, Y.; Marchetti, L.; Galerie, A.; Bouvier, P.; Petit, J. P.
2. Hillert, M. In Phase Equilibria, Phase Diagrams and Phase Analysing Thermal Oxidation using
Transformations; Cambridge University Press: MicroPhotoElectrochemistry (MPEC). In Local Techniques
Cambridge, 1998; Chapter 16. pp 427–457. for Corrosion Research (EFC45); Woodhead Publishing:
3. Kraftmaker, Y. Phys. Rep. 1998, 299, 79–188. Cambridge, 2007.
4. Aziz, M. Appl. Phys. Lett. 1997, 70, 2810–2812. 36. Mrowec, S.; Grzesik, Z. J. Phys. Chem. Sol. 2004, 65,
5. Hull, D.; Bacon, D. J. Introduction to Dislocations, 4th ed.; 1651–1657.
Butterworth-Heineman: Oxford, 2001. 37. Monceau, D.; Peraldi, R.; Pieraggi, B. Defect Diffusion
6. Hudson, J. B. Surface Science: An introduction; John Forum 2001, 194–199, 1675–1682.
Wiley: New York, 1998. 38. Atkinson, A. Rev. Modern Phys. 1985, 87, 437–470.
7. Howe, J. M. Interface in Materials: Atomic Structure, 39. Martin, M. Solid State Phenomena 1992, 21–22, 1–56.
Thermodynamics and Kinetics of Solid–Vapor, Solid–Liquid 40. Amani, B.; Addou, M.; Monty, C. Defect Diffusion Forum
and Solid–Solid Interfaces; John Wiley: New York, 1997. 2001, 194–199, 1051–1056.
8. Sutton, A. P.; Balluffi, R. W. Interfaces in Crystalline 41. Tsai, S. C.; Huntz, A. M.; Dolin, C. Mater. Sci. Eng. A 1996,
Materials; Oxford University Press: Oxford, 2006. A212, 6–13.
9. Pond, R. C.; Hirth, J. P. Solid State Phys. 1997, 47, 42. Doremus, R. H. J. Appl. Phys., 2006, 100, 1013011/17.
287–365. 43. Le Gall, M.; Lesage, B.; Bernardini, J. Phil. Mag. A 1994,
10. Hirth, J. P. Metel Trans. A 1991, 22A, 1331–1339. 70, 761–763.
11. Pieraggi, B.; Rapp, R. A.; van Loo, F. J. J.; Hirth, J. P. Acta 44. Hou, P. Y. J. Am. Ceram. Soc. 2003, 86, 660–668.
Metall. Mater. 1990, 38, 1780–1790. 45. Sun, J.; Stimer, T.; Matthews, A. Surf. Sci. Coatings 2006,
12. Kofstad, P. Non Stoichiometry, Diffusion and Electrical 201, 4201–4204.
Conductivity in Binary Metal Oxides; Wiley Interscience: 46. Lagerlöf, K. D. P.; Grimes, R. W. Acta Mater. 1998, 46,
New York, 1972. 5689–5700.
13. Mrowec, S. Defects and Diffusion in Solids; Elsevier: 47. Dybkov, V. I. Reaction-diffusion and solid-state chemical
Amsterdam, 1980. kinetics; IPMS Publications: Kyiv, 2002.
1.07 Mechanisms and Kinetics of Oxidation
S. Chevalier
Institut Carnot de Bourgogne, UMR 5209 CNRS, University of Bourgogne, 9 Avenue Savary, BP 47870, 21078 Dijon cedex, France

1.07.2 Thermodynamics 134
1.07.3 Kinetics 135
1.07.4 Diffusion 137
1.07.5 How to Investigate the High Temperature Corrosion Process 139
1.07.5.1 Surface Preparation 139
1.07.5.2 High Temperature Corrosion Investigations 139
1.07.5.3 High Temperature Corrosion Products Characterization 140
1.07.5.4 Prediction and Modeling of High Temperature Oxidation 142
1.07.6 High Temperature Oxidation Behavior of Metals and Alloys 144
1.07.6.1 Description of Chromia- and Alumina-Forming Materials 144
1.07.6.2 Oxide Scale Spallation: Mechanical Properties and Stress 144
1.07.6.3 Role of Minor Elements 146
1.07.6.4 Reactive Element Effect (REE) 146
1.07.6.4.1 Influence of RE on oxide scale growth kinetics 148
1.07.6.4.2 Influence of RE on oxide scale adherence 148
1.07.6.4.3 Location of RE in oxide scales 148
1.07.6.4.4 Influence of RE on oxide scale growth mechanism 148
References 148
Glossary coefficient, Dgb is grain boundary diffusion

Alumina (Al2O3) Aluminum oxide. coefficient.
Alumina (transient) Different phases of alumina Isothermal oxidation Continuous oxidation at a
(g-Al2O3, d-Al2O3, y-Al2O3) able to transform fixed temperature.
into the stable alumina phase, a-Al2O3. Marker (isotopic marker) Two stage oxidation
Alumina-forming (alloy) Alloy able to form an experiment consisting of oxidizing a metallic
alumina scale. substrate at high temperature in 16O2, then in
18
Aluminide Intermetallic compound containing O2, and locating oxygen isotopes across
aluminum (examples FeAl, TiAl, NbAl3,. . .). the thermally grown oxide scale.
Aluminization Surface treatment consisting of Pilling–Bedworth (ratio) Oxide volume relative to
forming aluminide diffusion coating via a native metal volume.
cementation process. Residual stress Stress taking into account growth
Arrhenius plot Relationship between reaction rate stress as well as thermal stress.
and temperature, k ¼ k0 exp ðE=RTÞ with E Silica (SiO2) Silicon oxide.
activation energy. Spallation Detachment of oxide scales
Cyclic oxidation Discontinuous oxidation during cooling from oxidation to room
corresponding to periods at temperature temperature.
followed by rapid cooling to room Thermogravimetry Continuous weight change
temperature. measurement as a function of oxidation time.
Diffusion Transport or migration of atoms, Wagner’s theory Theory which relates kinetic
ions or molecules in a solid, liquid or gaseous parameters to diffusion of species across a
media. growing scale.
Diffusion (coefficients of) Dapp is apparent Zirconia (ZrO2) Zirconium oxide.
diffusion coefficient, DV is volume diffusion
132
Mechanisms and Kinetics of Oxidation 133
oxidation process can be very different depending

Abbreviations on whether the materials are subjected to acids or
AES Auger electron spectroscopy salts in wet conditions, to dry conditions but at high
AFM Atomic force microscopy temperature (over 500 C) or to different aggressive
AISI 304 Austenitic stainless steel mainly atmospheres (O2, hot corrosion, sulfur, chloride,. . .).
composed of (in wt%) Fe–18Cr–10Ni Whatever the exposed conditions, the materials
EDX Energy dispersive X-ray analysis degrade and disintegrate more or less, following dif-
EPMA Electron probe micro analysis ferent mechanisms.1,2
FIB Focused ion beam The understanding of the mechanisms which con-
MOCVD Metal-organic chemical vapor trol these different types of corrosion is fundamental
deposition to try to improve the corrosion resistance of materials
PIPS Precision ion polishing system, ion milling and to extend their lifetime under such severe con-
using low angle Ar guns ditions. One rule is clearly established about the
PLS Photoluminescence spectroscopy corrosion process; one cannot suppress corrosion.
RE Reactive element Thermodynamic data show that the formation of
REE Reactive element effect corrosion products is energetically favored, following
SEM Scanning electron microscopy the chemical reaction:
SIMS Secondary ion mass spectrometry
SNMS Secondary neutral mass spectrometry M ! Mnþ þ ne ½1
TEC Thermal expansion coefficient in C1 or K1 Moreover, these chemical reactions are also
TEM Transmission electron microscopy largely dependent upon the corrosion rates. Thus,
XPS X-ray photoelectron spectroscopy both thermodynamics and kinetics need to be con-
XRD X-ray diffraction sidered while studying the corrosion process; they
have to be carefully controlled to try to protect the
base material. That is the reason why the best protec-
Symbols tion consists of decreasing the corrosion rates.
mproducts Chemical potential of corrosion products
This chapter focuses on the description of funda-
mreactants Chemical potential of reactants
mentals for high temperature oxidation. Specific
DrG0 Variation of standard free enthalpy of reaction
attention will be paid to the high temperature oxida-
DrH 0 Variation of standard enthalpy of reaction
tion mechanisms (from thermodynamics, kinetics,
Dm/A0 Weight gain per unit area
and diffusion points of view) and mainly how to
DrS0 Variation of standard entropy of reaction
relate these mechanisms to the improvement in oxi-
F17Ti Ferritic stainless steel mainly composed of
dation resistance. The better the knowledge of the
(in wt%) Fe–17Cr–0.4Ti
high temperature oxidation mechanisms is, the better
Gr Free enthalpy of reaction
is the improvement in the high temperature perfor-
kp Parabolic rate constant
mance of metallic material.
kl Linear rate constant
High temperature oxidation performances are
based on the formation of a protective oxide scale
Y3Al5O12 Garnet (called YAG for yttrium aluminum
which acts as a diffusion barrier, separating the base
garnet)
materials from the aggressive atmospheres.3–7 It is
most of the time a huge challenge, because an effec-
tive barrier has to be dense and homogeneous, to
cover the entire alloy surface, to possess mechanical
1.07.1 Introduction properties matching those of the base materials, to
consume the least possible metal,8 to remain stable, to
As is evident, humans need oxygen to survive. This is be able to resist complex atmospheres (presence of
a paradox because this fundamental molecule can be water vapor for example9–11) and above all to keep
very aggressive against most metallic materials. Even adherent to the substrate, even after thermal shocks.
if air contains only 20% of oxygen, it is enough to Materials able to form protective chromia, Cr2O3,
provoke large-scale degradation of materials. alumina, Al2O3, or silica, SiO2 layers are the best
This degradation is directly connected to their candidates to be used at high temperatures in oxidiz-
conditions of use. The failures caused by the ing atmospheres. These refractory oxides grown on
Fe-, Co-, or Ni-based alloys possess high melting grain boundaries, sometimes within the oxide grains,
points. Cr (12 wt%) and Al (5 wt%) are necessary to or form mixed oxides. Some of them, called reactive
form chromia on chromia-forming alloys and to form elements or rare earths (RE), have considerable
alumina on alumina-forming alloys, respectively. The effects on the high temperature oxidation behavior
growth rate of chromia scales is much faster than that of alloys. Their addition in small quantities is enough
of alumina scales12 and, at temperatures over not only to greatly decrease the oxidation rates, but
1000 C, chromium oxides (especially CrO3) start to above all to drastically increase the oxide scale
become volatile. adherence to the base material.
As a matter of fact, chromia and alumina scales The present chapter helps define adequate ways to
should only be employed in adequate temperature study high temperature corrosion processes. Ther-
ranges, T 1000 C for Cr2O3 and T 1300 C for modynamics, metallurgy, kinetics and diffusion,
Al2O3, under isothermal exposures, that is, continu- experimental tools, and the effect of complex atmo-
ous maintenance at temperature, or under thermal spheres will be described to answer the still open
cycling conditions, that is, successive exposure at question: how to improve high temperature corrosion
oxidizing temperature followed by rapid cooling. resistance of metals in aggressive atmospheres? Some
During cycling, the oxide scales tend to spall, losing solutions will be proposed in order to limit the deg-
any protective effect. Thermal stresses generated radation of metallic materials when exposed in such
during cooling and due to differences between the atmospheres at high temperature.
thermal expansion coefficient of the base metal and
that of the oxide scale are mainly responsible for
oxide scale spallation.13 The thermal and/or residual
stresses are known to play a huge role during the 1.07.2 Thermodynamics
oxidation of metallic alloys.4,14–20
A recent study showed that the amplitude of oxide Thermodynamics can be considered the driving
scale spallation could not be directly related to the force for the degradation of metals and alloys in
residual stresses in the oxide layers.21 It is also pro- corrosive atmospheres. Effectively, pure metals or
posed that vacancy coalescence at the metal–oxide metallic elements that constitute alloys are not stable
interface, as a result of the oxide scale growth, can in a reactive gaseous atmosphere. Then, if one con-
create pores and cavities and therefore weaken scale siders that the gas is oxygen, a pure metal (M) trans-
adhesion.22 forms to oxide (MaOb), following the chemical
It appears clearly that one has to consider not reaction:
only the chemical evolution of the oxide scale dur- 2a 2
ing its growth, but also its mechanical evolution, as MðsÞ þ O2 ðgÞ $ Ma Ob ðsÞ ½2
b b
well as the close relationship it establishes with the
In case of chromia formation for example, the
base metallic material. All these considerations are
chemical reaction becomes:
usually named ‘physical-chemistry’ of the oxide
scales. 4 2
The proposed situation is the simplest one, as CrðsÞ þ O2 ðgÞ $ Cr2 O3 ðsÞ ½3
3 3
pure oxide scales grown on model materials are
being considered. In fact, commercial alloys contain The free enthalpy of the reaction is expressed by:
a lot of elements added to increase the mechanical X X
properties, to stabilize the austenitic phase instead Gr ¼ mproducts mreactants ½4
of the ferritic phase, to avoid or fix carbide for- where mproducts and mreactants are the chemical poten-
mation (formation of s phases for example) or to tial of the corrosion products and the reactants,
increase their corrosion resistance at ambient or high respectively.
temperature. The minor element additions perturb By detailing each chemical potential, this equation
the oxide scale formation because they may form can be written:
mixed oxides, such as spinel or perovskite oxides;
2 2
these oxides are known to possess poor resistance to Gr ¼ m0Ma Ob þ RT lnðaMa Ob Þ2=b
high temperature oxidizing atmospheres compared b b
to chromia or alumina scales. Most of the added 2a 0 2a PO
mM RT lnðaM Þ2a=b m0O2 RT ln 02 ½5
elements segregate at the interfaces, at the oxide b b P
If one assumes that: different minor elements), in order to know which

oxides are the most thermodynamically stable.
2 2a However, this approach needs the system to be at
r G 0 ¼ m0Ma Ob m0M m0O2 ½6
b b equilibrium; most of the time, this situation is not
aMa Ob ¼ 1 (oxide is pure), aM = 1 (metal is pure), and obtained in real systems. Thermodynamics has some
P 0 = 1 atm. limits in the understanding of the corrosion behavior
The classical expression of the free enthalpy of of metals and alloys at high temperatures. Further
reaction is: details on the thermodynamics of gas–metal reactions
1 can be found in Chapter 1.09, Thermodynamics and
Gr ¼ r G 0 þ RT ln ¼ r G 0 þ RT ln K ½7 Theory of External and Internal Oxidation of
PO2
Alloys.
with K the equilibrium constant.
When the equilibrium is reached, Gr = 0 and then,
r G 0 ¼ RT ln K ¼ RT ln PO2 ½8
1.07.3 Kinetics
A kinetics approach is complementary to a thermody-

Then,
namics approach, because it gives information about
r G 0 the rate of oxidation of a metal or an alloy. One can
lnPO2 ¼ ½9
RT talk about heterogeneous kinetics because a metallic
surface reacts with a gas, oxygen for example.25 The
This expression means that for a partial pressure mechanisms occurring during high temperature oxi-
of oxygen below this critical value (value obtained dation are complex. Figure 1 describes some aspects
when the equilibrium is reached), the metal does not of the reaction between a metallic surface and oxygen
oxidize. For a partial pressure of oxygen over this
value, the metal can be totally oxidized. O2 (g)
Another possibility for expressing the variation of
O O O O O
the reaction standard free enthalpy is:
Metal Adsorption
r G 0 ¼ r H 0 T r S 0 ½10
O2 (g)
with DrH 0 being the variation of the standard Oxide nucleation and growth
enthalpy of the reaction and DrS 0 being the variation O
Oxygen dissolution
Metal
of the standard entropy of the reaction.
These two expressions [8] and [10] allow a com-
bination of the oxidation temperature together with O2– e– Mn+
Oxide scale growth
the partial pressure of oxygen. Then, two diagrams O O
Metal Internal oxidation
will indicate the evolution of a metal or an alloy in an
Time
oxidizing atmosphere at high temperature: phase dia- O2 (g)

grams and Ellingham diagrams.6,7,23 –
O2 e– Mn+
Phase diagrams give the phases that can be formed Cavities, porosities, cracks
depending on the temperature or the pressure. Then, O O
Metal
if for one component (an oxide in our case), several
crystallographic phases can be formed, the use of phase
MeOX (g) O2 (g)
diagrams allows the identification of the one that is
thermodynamically stable for a fixed temperature. As Macrocracks
an example, zirconia (ZrO2) has three allotropic struc- (molten oxide, oxide
O O evaporation)
tures: cubic, tetragonal, and monoclinic, which trans- Metal
form to each of them on decreasing the temperature, as
shown in the Zr–O phase diagram.24 Figure 1 Schematic representation of the reaction
Ellingham diagrams are useful for comparing the between a metallic surface and oxygen at high temperature.
Adapted from Kofstad, P. High Temperature Corrosion;
thermodynamic stability of two oxides formed from Elsevier Applied Science, 1988; Sarrazin, P.; Galerie, A.;
two different metals. Their use is crucial in the case Fouletier, J. Les mécanismes de la corrosion sèche. Une
of oxidation of alloys (i.e., those which contain approche cinétique; EDP Sciences, 2000.
at high temperature.6,7,26 The main steps of the oxida- the relationship between the weight gain per unit
tion process can be described as follows: area (Dm/A) and the exposure time (t) is:
2
oxygen adsorption at the metal surface, m
oxide nucleation with formation of a bidimen- ¼ kp t ½11
A
sional layer
tridimensional growth of oxide with formation of a with kp the parabolic rate law constant (g2 cm4 s1).
thin oxide layer, The parabolic rate law constant can be also expressed
diffusion of species (from metal or from gas) across in square centimeter per second, by using:
the growing scale, 2
apparition of cracks or microcracks in the base kc ¼
M M a Ob
kp ½12
metal and/or in the oxide scale, bMO rMa Ob
development of cracks at the metal–oxide inter-
face, which can lead to scale spallation and loss of where MMa Ob and MO are the molar mass of MaOb and
protectiveness. O, respectively, and rMa Ob is the oxide density.
The plot of Dm/A ¼ f(t1/2) or (Dm/A)2 ¼ f(t) allows
Two limiting cases can be observed during the us to determine the domain where the parabolic rate
growth of an oxide layer: law is verified, and also to calculate the value of kp. The
growth of a dense oxide layer, for which the trans- first plot is preferable because it improves the accuracy
port of species takes place by solid state diffusion, by minimizing the errors in Dm/A:
growth of a porous oxide scale, for which the trans- a linear regime, for which a chemical reaction at
port of species takes place by gaseous diffusion. one of the interfaces (gas-oxide and/or metal-
One can associate two kinetic regimes to these two oxide) is the limiting step in the oxidation process.
limiting cases (Figure 2)27: In that case, the relation becomes:

a parabolic regime, for which in planar symmetry, Dm
¼ kl t ½13
diffusion of species across a growing scale is the A
limiting step in the oxidation process. In that case, with kl the linear rate law constant (g cm2).
When a good protection against high temperature
Not protective scale
oxidation is observed, the parabolic rate law is ver-
Protective scale ified. Conversely, poor oxidation resistance corre-
sponds to a linear rate law (Figure 2).
Other kinetics laws can be found during the
Temporary growth of oxide scales at high temperature:
protection
Cubic law, for which
3
m
¼ kc t ½14
Weight gain
A
Time
Power law, for which
n
Protective scale m
followed by evaporation ¼ kn t ½15
A
or spallation
with 1 < n < 3.
Evaporation
This last law corresponds to an intermediate case
between the three previous cases. When 2 < n < 3, the
kinetic law is generally called subparabolic, which
means deviation from a real parabolic law. It can be
Figure 2 Schematic representation of kinetics observed
during high temperature oxidation. Adapted from Antoni, L.;
caused by the evolution of grain size during the high
Galerie, A. Corrosion sèche des métaux, Techniques de oxidation process, by scale cracking, oxidation of reac-
l’Ingénieur, M4 228, p 9. tive element, or incorporation of metal precipitates.28
Other laws were identified in specific cases (thin with kc, the parabolic rate law constant (cm2 s1),
films, beginning of the oxidation process, or low a ¼ 2 and b ¼ 3 (for Cr2O3 for example), pO2 is the
oxidation temperatures). They are: partial pressure of oxygen for the oxide decomposi-
tion at the metal–oxide interface, and pO2 is the
Logarithmic law, for which
partial pressure of oxygen at the gas–oxide interface.
m This theory can be adjusted in order to fit realistic
¼ A þ B ln t ½16
A situations and to consider short-circuit diffusion
paths for example. In the case of diffusion across
Inverse-logarithmic law, for which chromia and alumina scales in the temperature
range 800–1200 C, these short-circuit paths are usu-
m
¼ C D ln t ½17 ally oxide grain boundaries.
A
Some specific cases are also obtained when mixed 1.07.4 Diffusion
kinetics laws are observed, especially when chemical
reaction at interface(s) takes place together with dif- The diffusion short-circuits for diffusion within oxide
fusion of species across the growing scale. In that scales in the temperature range from 800 to 1000 C
case, the kinetics regime can be described by: are considered to be oxide grain boundaries. Accord-
2 ing to Harrison,37 three diffusion regimes pffiffiffiffifficould be
m m
t ¼AþB þC ½18 identified. The A regime is defined by Dt d, d
A A being the oxide grain size; in that case, the volume
where A, B, and C are constants; B corresponds to 1/kl penetration is superior to the short-circuit size, espe-
and C to 1/kp. cially the grain boundaries. The C regime claims that
A particular case corresponds to successive rate the penetration is continuous at the grain boundaries
pffiffiffiffiffi
laws: parabolic then linear, which signifies a protec- ( Dt d, d being the grain boundary width). Finally,
tive oxidation behavior followed by a loss of protec- the B regime is an intermediate regime where three
tiveness. Both regimes are separated by a transient processes participate in diffusion: pvolume,
ffiffiffiffiffi grain
period. This case, illustrated in Figure 2, is typical of boundary, and lateral diffusion (d Dt d =2).
what happens during the growth of zirconia on Zr or Marker experiments made during oxidation in
18
zircaloy.29–31 O2 lead to the determination of oxygen diffusion
The plot of instantaneous kp can also give crucial coefficients via oxide grain boundaries and via the
information, especially on the evolution of phases in oxide bulk. Taking into consideration the plots of ln
the oxide scale.32,33 This is particularly true for alu- [18O] = f(x), the first part of the curve corresponds to
mina-forming alloys, because of the presence of sev- the apparent diffusion of oxygen (Figure 3). From the
eral alumina phases (y, g, d, a-Al2O334,35). The plot of three diffusion regimes, the B regime is usually cho-
instantaneous kp versus time indicates the evolution sen for the study of intergranular diffusion.38,39 In this
of kp values; a constant kp value is indicative of a
stationary oxidation regime, which can be associated 5.0
with the formation of a stable oxide phase, whereas 4.5
an evolution of kp with time suggests the presence of a Dapp
4.0
transient phase (y-Al2O3 for example). The so-plot- 3.5
ted curves are then helpful in determining the time
ln[18O]
3.0
necessary to form a protective stable oxide phase (a-
Al2O3 for example). 2.5 Dgb
Wagner developed a theory relating the parabolic 2.0
rate law constants to the diffusion coefficient of 1.5
species: anions (O2, S2,. . .) and cations (Cr3+,
1.0
Al3+,. . .).36 Then, for an oxide, MaOb, Wagner’s oxi-
dation theory gives: 1.0 ⫻ 10–5 2.0 ⫻ 10–5 3.0 ⫻ 10–5
ð (cm)
1 pO2 b
kc ¼ Danion þ Dcation d ln PO2 ½19 Figure 3 Plot of ln[18O] = f(x) showing the zones where
2 pO a
2 apparent and oxide grain boundary diffusion predominate.
pffiffiffiffiffi
regime, d Dt d =2 with d being the grain This relation allows the determination of the oxy-
boundary width and d being the oxide grain size. gen bulk diffusion coefficient, Db. Using Dapp and Db,
The apparent diffusion of oxygen corresponds to the relation [20] leads to the calculation of Dgb.
diffusion in the bulk and the oxide grain boundaries. This calculation takes into consideration the oxide
It is expressed in the case of dislocation and bulk scale roughness.50 The principle for determining the
diffusion40 but can be used in the case of grain oxygen diffusion coefficients is the same, but the
boundary and bulk diffusion,41,42 if we assume that fraction of sites associated with grain boundaries is
the beginning of the oxygen distribution profile cor- given by:
responds to Dapp (A regime)38:
f=2 3d
Dapp ¼ ð1 f ÞDb þ f Dgb ½20 fr ¼
½25
f=2 þ 2Ra f
where f is the fraction of sites associated to the grain where Ra is the average roughness of the oxide scale.
boundaries; f can be expressed as: Cation diffusion has also to be taken into account
3d during oxide scale growth. Chromium 54 can be
f ¼ ½21
f used as an isotopic marker in chromia (made of
56
Cr216O3), and also in alumina (made of 27Al216O3).
where d is the grain boundary width (generally
As a matter of fact and taking into account the fact that
assumed to be 1 nm) and f is the average oxide
both chromium and aluminum cations diffuse in the
crystallite size.
same way, the results obtained for chromium diffusion
Dapp is determined from the second Fick’s law
coefficients in alumina scales are applied for alumi-
solution:
! num cation diffusion coefficients in the same scale.
Cðx; t Þ CS x Indeed, the use of a chromium isotope as a diffusive
¼ erf pffiffiffiffiffiffiffiffiffiffiffi ½22
C0 CS 2 Dapp t species is easier than the use of aluminum, which has
no natural isotope; one can only use an artificial alu-
CS is the oxygen 18 concentration at the oxide scale minum isotope, 26Al, which is radioactive and very
surface, C0 is the natural concentration of oxygen 18 expensive. Very few studies have used this radioactive
within the oxide layer (0.2 at.%43), and t is the diffu- tracer.51,52
sion time. In the case of 54Cr or 26Al, the solution of the
This model which associates the beginning of the second Fick’s law corresponding to a thin deposited
distribution profile to an apparent diffusion coeffi- layer is:
cient has been recently discussed by Fielitz et al.44
who proposed that the first part of the oxygen diffu- x2
CðxÞ ¼ C0 exp ½26
sion profile in polycrystalline mullite corresponded 4Dapp t
to the bulk diffusion coefficient in the B regime and
not to the apparent diffusion coefficient. The procedure is the same as that applied for
The second part of the diffusion curve, ln[18O] = the determination of oxygen diffusion coefficients.
f (x), led to determination of the oxygen grain bound- The first part of the plot ln[Cisotope] ¼ f (x) corre-
ary diffusion coefficient (Figure 3). By applying the sponds to Dapp and the second part to Dgb.53 Dapp is
model developed by Suzuoka45,46 and Whipple–Le calculated from the slope of ln[Cisotope] ¼ f (x2),
Claire47–49 Dgb is expressed by: whereas Dgb is determined from the Whipple–Le
rffiffiffiffiffiffiffiffi Claire model.
4Db @ lnC 5=3
Dgb d ¼ 0:661 6=5 Dapp ½23 Figure 450,54–60 is an illustration of oxygen and
t @x
chromium diffusion coefficients determined in
where ð@ lnC=@x 6=5 Þ is the slope of the curve ln[18O] ¼ thermally grown Cr2O3 scales formed on chromia-
f (x6/5). forming alloys. It appears clearly that, in thermally
By combining eqns [22] and [25], one obtains the grown Cr2O3, the grain boundary diffusion coeffi-
following relation: Cr O
cients (Dgb and Dgb ) are higher than the bulk diffu-
rffiffiffiffiffiffiffiffi sion coefficients (Figure 4), in the considered
0:661f 4Db
ð1 f ÞDb þ temperature range. The large discrepancy in the dif-
d t
fusion coefficient values depends on the nature of the
@ lnC 5=3 chromia-forming materials and on the presence or
6=5 Dapp ¼ 0 ½24
@x not of a reactive element.
DCr
app [Hagel]
DOapp [Hagel]
DbCr [Lobnig1]
T (°C) DCr
gb [Lobnig1]
1200 1100 1000 900
DOb [Graham]
DOgb [Graham]
1E–8
DCr
b [Lobnig2]
DCr
gb [Lobnig2]
1E–10 DO
mono [Huntz]
DO
app [Huntz]
DOb [Huntz]
1E–12 DOgb [Huntz]
D (cm2 s–1)
DOapp [Tsai]
DOb [Tsai]
1E–14
DOgb [Tsai]
DOapp [Chevalier]
1E–16 DOb [Chevalier]
DOgb [Chevalier]
Cr
Dapp [Tsai]
1E–18
DCr
b [Tsai]
DCr
gb [Tsai]
1E–20 DOapp Y2O3 [Chevalier]
6.5 ⫻ 10–4 7.0 ⫻ 10–4 7.5 ⫻ 10–4 8.0 ⫻ 10–4 8.5 ⫻ 10–4 9.0 ⫻ 10–4 9.5 ⫻ 10–4 DOb Y2O3 [Chevalier]
DOgb Y2O3 [Chevalier]
T–1 (K–1)
DOapp Nd2O3 [Chevalier]
DOb Nd2O3 [Chevalier]
DOgb Nd2O3 [Chevalier]
DCr
app Y [Tsai]
DCr
b Y [Tsai]
DCr
gb Y [Tsai]
Figure 4 Comparison of oxygen or chromium diffusion coefficients in thermally grown chromia scales doped or not by RE (DO app
is the apparent oxygen diffusion coefficient, DO O
V is the volume oxygen diffusion coefficient, Dgb is the grain boundary oxygen
diffusion coefficient, DCr Cr Cr
app is the apparent chromium diffusion coefficient, DV is the volume chromium diffusion coefficient, Dgb is
the grain boundary chromium diffusion coefficient). The samples doped with RE are indicated by Y2O3, Nd2O3, or Y.
1.07.5 How to Investigate the High 1.07.5.2 High Temperature Corrosion

Temperature Corrosion Process Investigations
1.07.5.1 Surface Preparation The tools used to investigate the high temperature
reactivity of metallic materials are numerous: they
Particular attention has to be paid to the preparation
can be basic (tubular or muffle furnaces, thermogra-
of samples. Depending on the study, one has to ask if
vimetric devices,61,62 . . .) or more elaborate (corro-
it is necessary to polish the surfaces (if the answer is
sion under stress application,63 oxidation with 18
positive, then one has to ask: ‘up to which finishing
oxygen,64–68 . . .).
grade? Is it necessary to achieve a mirror-like finish?’)
For isothermal oxidation tests, the use of a ther-
or to use sand blasting to prepare the surfaces or to
mobalance is of great interest, because it gives infor-
degrease in ethanol or acetone (is it necessary to use
mation about kinetics for more or less long exposures
an ultrasonic bath, deionized water?).
(see Figure 2). However, for most isothermal oxida-
At this step in the preparation, the sample surface
tion tests the duration in a thermobalance is relatively
area has to be precisely determined, because all mass
short (from 24 to 200–300 h typically) compared
gain or loss will be divided by the sample surface area
to real time in service conditions. Discontinuous
in order to make comparisons between tests.
experiments can be used for long oxidation times; morphology at the nanometric scale, in association
these allow the introduction of many samples with local chemical analyses and microdiffraction.88
(10–50 for example) all together in the same furnace Unfortunately, two main difficulties limit the use of
(a muffle furnace is the most used) at a fixed temper- this technique:
ature and removal of specimens after various times.
sample preparation is very delicate and needs long
By weighing each sample after such exposures,
and difficult manipulations,88–90 even though
kinetic curves can be built for long periods (several
the use of a focused ion beam (FIB) drastically
thousands hours).
increases the chance of success for sample prepa-
However, failure of oxide scales mainly occurs
ration91–97),
when the materials are subjected to thermal cycl-
the correspondence between the nanometric and
ing conditions. The stresses, induced by temperature
the macrometric scale is often difficult and needs
drops and rises, promote cracks and spalling, followed
repeated observations.
by drastic enhancement of the damage caused by the
corrosive atmospheres.69–74 Since thermal cycling These parameters are the reasons why TEM
treatments are usually more severe than isothermal observations are usually the last step in the charac-
oxidation, cyclic oxidation tests are widely performed terization process, which logically evolves from
in order to test the scale adherence under such severe macro- to nanoscale, and not the opposite!
conditions.75–85 Cyclic oxidation experiments can be A second very promising tool is the use of two
performed in a thermobalance (this means that the stage oxidation experiments in 16O2/18O2. This
furnace or the sample is moved to produce cycling, method consists of using an oxygen isotope, 18O, as
but the sample is continuously weighed86), although, a marker in a preformed 16O containing layer. The
usually, net and gross weights are collected after principle is to oxidize the metallic substrate first
automatic or manual cycles. The net weight gain under a 16O2 atmosphere; the reaction chamber is
corresponds to the specimen weight excluding any evacuated and 18O2 is introduced for the second
spalled scale. The gross weight gain corresponds to step of the test, without temperature modification.
the specimen weight including any spalled scale col- The technique has been largely used not only to study
lected in the crucible. The difference between the net oxide scale growth mechanisms, but also to determine
and the gross weight gains indicates the proportion of oxygen diffusion coefficients in many chemical sys-
spalled oxide scale. The results in Figure 5 give an tems.44,98–102 The method can be extended to marker
illustration of plots of net and gross weights for alu- experiments in water vapor by making successive
minide diffusion coatings on Fe–Cr alloys, which exposure to H2 16O/H2 18O.103 The oxygen isotope
indicate cyclic behavior with and without spallation.1 distributions are followed using secondary ion mass
spectrometry (SIMS) or secondary neutral mass spec-
trometry (SNMS). Very careful analyses of the oxygen
1.07.5.3 High Temperature Corrosion
isotope distribution profiles allow determination of
Products Characterization
the oxygen diffusion coefficients in growing scales.
The complete understanding of oxide scale growth Figure 7 shows three limiting cases corresponding to
mechanisms needs a very careful analysis of the cor- the diffusion of cations (Figure 7(a)), to the diffusion
rosion products. Figure 6 illustrates most of the of anions (Figure 7(b)) and to mixed diffusion (Figure
possibilities for characterizing thermally grown 7(c)). Bulk and grain boundary diffusion coefficients
oxide scales.1,87 Scanning electron microscopy for oxygen were calculated using the models proposed
(SEM) coupled with energy dispersive X-ray analyses by Suzuoka45,46 and Whipple–Le Claire.47,48
(EDX) is probably the most useful analytical tool for More recently, other characterization tools have
observing oxide scales at the microscopic level. Cou- been used in the field of high temperature corro-
pled with XRD, it gives much necessary information sion. These give complementary information and,
about thermally grown oxide scales. However, one together with ‘more classical’ analyses, they allow
has also to consider optical microscopy; this gives the determination and understanding of oxide scale
more macroscopic information, which is often suffi- growth mechanisms. Among these techniques, Raman
cient for evaluating the oxidation behavior of metals spectroscopy,104–108 photoluminescence spectroscopy
and alloys. Another convenient tool for studying (PLS),35,109–114 X-ray photoelectron spectrometry
oxide scales is transmission electron microscopy (XPS),34,35,115–117 and photoelectrochemistry118–121
(TEM), which allows observations of the scale are probably the most interesting tools.
Net weight
7
Gross weight
5
Δm/A (mg cm–2)
0
0 20 40 60 80 100
(a) Cycles
8
6 Net weight
Gross weight
5
Δm/A (mg cm–2)
0
0 20 40 60 80 100
(b) Cycles
Figure 5 Cyclic oxidation kinetics (24 h cycle at 1000 C); case of aluminized diffusion coatings: (a) the deviation between
gross and net weights means oxide spallation and (b) when doped with Y2O3, there is no spallation. Reproduced from
Chevalier, S. Traitements de surface et nouveaux matériaux: quelles solutions pour lutter contre la dégradation des matériaux
à haute température?; Les Editions Universitaires de Dijon, 2007.
Generally, the study of the corrosion products is X-ray diffraction devices equipped with high temper-
performed ‘ex situ,’ which means after the specimen ature furnaces were developed in order to follow the
has cooled down to room temperature. How can one evolution of crystallographic phases during the oxi-
be sure that what is observed at room temperature is dation process of metallic substrates and of the
representative of what happened at the oxidation underlying substrate microstructure under oxidizing
temperature? In situ analyses are able to follow the conditions.122 The technique is also able to follow
oxidation reactions at the oxidation temperature. growth stress evolution during the establishment of
XRD
SEM
16O 18O
2/ 2
0
TEM
0 500 1000 1500 2000
SIMS/SNMS
(a)
Morphology 16O/18O
Crystal structure Elemental analysis
Segregation Depth profiles

TEM/
SIMS
X-ray
Surface Elemental
structure RHEED/ Metals analysis
Auger
and X-ray Semiconductors
compo-
Depth
sition
profiles
EELS XPS
Chemical Chemical
bonding bonding
Elemental Elemental
maps analysis
(b)
Figure 6 Illustration of characterization tools to analyze thermally grown oxide scale. Part (a) adapted from Chevalier,
S. Traitements de surface et nouveaux matériaux: quelles solutions pour lutter contre la dégradation des matériaux à haute
température?; Les Editions Universitaires de Dijon, 2007 and part (b) adapted from Graham, M. J. Mater. High Temp. 2000, 17, 1.
the growing scales.123,124 More recently, texture ana- during oxidation at 950 C of a Y2O3 coated-Fe–30Cr
lyses of the oxide scale as well as of the metallic alloy. It shows that the peak corresponding to
substrate has been followed during the formation the substrate (Fe–30Cr) disappears after 60 min of
and growth of the corrosion products. These ‘in situ’ oxidation, which corresponds to the formation of
characterizations are very useful for understanding YCrO3.
the oxide scale growth mechanisms, for understand-
ing the role played by minor elements on their for-
1.07.5.4 Prediction and Modeling of High
mation and for linking the chemical properties of the Temperature Oxidation
scale to its mechanical properties. Figure 8 gives
an illustration of the use of in situ XRD; it allows In order to prevent high temperature degradation
one to follow the phase formation and transformation due to oxidation, predictive modeling has been
18O 16 18O
O2 2
2 Mn+
16O
Counts 2
Substrate
(a) Thickness or sputtering time

O2–
16O 18O 16O

2 2 2
Counts
18O
2
Substrate
(b) Thickness or sputtering time
O2–
18O
18O 16O 18O 2
2 2 2 16O Mn+
2
Counts
18O
2
Substrate
(c) Thickness or sputtering time

Figure 7 Three main cases encountered during two stage oxidation experiments. The location of 18O compared to
the16O containing scale leads to the understanding of thermally grown oxide scales, (a) diffusion of cation, (b) diffusion of
anion, and (c) mixed diffusion.
800 Cr2O3 +YCrO3

Y2O3 Y2O3 + YCrO3
700 Cr2O3 Cr2O3 YCrO3
T = 20 °C
600
T = 450 °C
500 330 min
300 min
Counts
270 min
400 240 min
210 min
300 180 min
150 min
120 min T = 970 °C
200 90 min
60 min
100 30 min
0 min
0
30 35 40 45 50 55 60
2θ Fe–30Cr
Figure 8 In situ XRD following phase formation and transformation during oxidation at 950 C of a Y2O3 coated-Fe-30Cr
alloy. Adapted from Chevalier, S. Traitements de surface et nouveaux matériaux: quelles solutions pour lutter contre la
dégradation des matériaux à haute température?; Les Editions Universitaires de Dijon, 2007.
developed. This aspect is of particular interest for below which an alumina scale was no more able to
industrialists who need to target several thousands grow and form a protective oxide scale. For a binary
of hours in service condition. For alumina-forming alloy, this critical value was established from Wagner’s
alloys, it was shown that there was a critical Al level theory as:125
rffiffiffiffiffiffi!
kc In the field of high temperature oxidation, the
cB ¼ f with f ðuÞ ¼ p1=2 uð1 erf uÞ expu2 ½27 classical alloys are chromia-forming (i.e., being able
2D ~
to form a chromia, Cr2O3, layer) or alumina-forming
with kc (cm2 s1) the parabolic rate constant and D~ (i.e., being able to form an alumina, Al2O3, layer).
(cm2 s1) the interdiffusion coefficient of aluminum These oxide scales, together with silica, SiO2, layers,
and iron. are the most protective scales for protection against
When the Al concentration reaches this critical high temperature degradation. These refractory oxi-
value, a catastrophic oxidation phenomenon occurs, des have high melting points (2330 C for Cr2O3,
because of the formation of iron and chromium 2054 C for Al2O3 and 1713 C for SiO2132,133) and
oxides, which soon leads to complete destruction grow on Fe-, Ni-, or Co-base alloys, which contain at
of the samples. Recently, Strehl et al.126 also pro- least 12 wt% Cr for the chromia-forming alloys,134
posed a method, based on Wagner’s theory, to deter- associated with at least 5 wt% Al for the alumina-
mine the critical Al concentration leading to forming alloys.135 One has also to distinguish ferritics
catastrophic oxidation. They found that this critical (containing elements able to stabilize the a-Fe phase,
value was dependent on temperature and varies such as Cr) from austenitic steels (containing ele-
from 0 wt% at 1000 C to 3 wt% at 1300 C. Qua- ments able to stabilize the g-Fe phase, such as
dakkers and Bongartz127 established predictive dia- Ni).1,136 Some steels can be also martensitic or duplex
grams for the life time of metallic materials. The in nature. The properties of the different categories
proposed diagrams were obtained from kinetic con- of steel depend on minor element additions. As an
siderations: example, Figure 9 gives a comparison of thermal
expansion coefficients, TEC, for different materials
tB ¼ 4:4
103 ðC0 CB Þrdk1=n ðm Þ1=ðn1Þ ½28 and oxides.137 The formation of a continuous silica
where tB corresponds to the start of catastrophic layer is guaranteed in the case of silicides, MoSi2 or
oxidation, k and n are kinetic constants and expo- FeSi2, but their physical and chemical properties are
nents established from Dm* ¼ Dm/S ¼ ktn, C0 is the very different from those of chromia- and alumina-
initial aluminum concentration, CB is the critical Al forming alloys. In many cases, Si is added as a minor
concentration, r is the alloy density, d is the sample element to the latter alloys and contributes to the
thickness, and Dm* is the total weight gain measured formation of a discontinuous scale below the main
during the thermal treatment. This modeling was oxide scales.138
in agreement with experimental data in order to As indicated previously, chromia- and alumina-
predict the beginning of oxide scale spallation in forming materials usually guarantee the establishment
Fe–Cr–Al foils at 1100 C.128 of parabolic regimes, which are the most protective.
Recent models developed by Monceau129,130 and One can note that the parabolic rate law constants
Smialek131 predict the life time for alloys exposed for alumina-forming alloys are lower than those for
to cyclic tests. The correlation between modeling chromia-forming alloys (Figure 10).12 This means
and experimental data is clear, despite simple that the oxidation rate is lower when an alumina
hypotheses. scale grows on an alloy surface. Growth of SiO2 scales
result in similar parabolic rate constants as those of
Al2O3 scales.
1.07.6 High Temperature Oxidation

1.07.6.2 Oxide Scale Spallation:
Behavior of Metals and Alloys
Mechanical Properties and Stress
1.07.6.1 Description of Chromia- and
When they are used under appropriate conditions
Alumina-Forming Materials
(maximum temperatures of 1000 and 1300 C, for
In the real world, it is rare to find pure metals; most of M–Cr and M–Cr–Al, respectively), Cr2O3 and Al2O3
the time, alloys, which contain minor elements, such scales possess the main protective properties needed
as Cr, Al, Si, Ti, etc., are used. Generally speaking, for high temperature service conditions, at least under
steels are iron-based and contain between 0.02 and isothermal conditions (for a fixed temperature without
2.06 wt% carbon. Cast-irons correspond to a level of thermal shocks). When exposed to thermal cycling
C over 2.06 wt%. Other alloys used at high tempera- (repeated heating and cooling), the adherence of the
tures are usually Ni- or Co-base. oxide scales is often poor and limits the life time of
Ferritics and martensitics Austenitics
0 5 10 15
SiO2 (amorphous) α
HPSN
Sic 10–6/K
AIN
(max. values, 0...1000 ⬚C)
Si
TiN
MoSi2
Cr2O3, SiO2 (cryst.)
Al2O3, Ti
Pt
Y2O3
Cr
IMI 834, NiAl
TiO2
Ti3Al (α2)
Pd
Ni3Al (adv.)
TiSi2
TiAl (γ)
Co
CoO
738
Alloy S, 713
617, C-4
CMSX 6, 100
C-276, 230
C-22
X
600
X-750
Nio, 556, 718
625
Ni
188
800 H
Figure 9 Comparison of thermal expansion coefficients of different materials and oxides. Reproduced from Schuetze, M.
In Proceedings of the Euroconference 2000 ProEnMach, Clausthal-Zellerfeld, Germany, 20–22 Juillet 2000.
alloys considerably.139–142 The more the amplitude of and the native metal; this change is characterized
the thermal shocks, the more is the tendency for the by the Pilling–Bedworth ratio (PBR) and corresponds
oxide scale to spall off or crack, making the underlying to the growth stress within the scale.151,152 Table 1
substrate free of any protection. The reasons why the gives values of PBR for different oxides.152 When PBR
oxide layers detach from the substrates are mainly due < 1, the volume of oxide is lower than the volume of
to the difference in TEC between the oxides and the metal and tensile stress appears in the oxide scale.
substrates (Figure 9), which generates thermal stress. When PBR > 1, the volume of oxide is higher than
Various stresses are able to develop in the oxide scales: the volume of metal and compressive stress appears
growth stress, thermal stress, and residual stress (as a in the oxide scale. Several complementary models
first approximation, residual stress can be ascribed to deal with growth stresses and attribute them to micro-
the addition of both growing and thermal stress). They structural and/or diffusion considerations,153–157
play a considerable role in the high temperature cor- but are based on the general models established in
rosion mechanisms and their determination gives fun- the past.151,158
damental information on the way to increase the Geometry is also a parameter that influences the
corrosion resistance.4,143–150 Another reason for stress mechanical properties of the scale and has a huge
generation in thermally grown oxide scales is asso- impact on the oxide scale adherence.152 The stress
ciated with the volume change between the oxide generated in an oxide scale as is thermally grown on
It appears then clearly that it is necessary to

consider the chemical evolution of the oxide scale
–6
together with its mechanical evolution, as well as its
TiO2 relationships with the underlying substrate.
Parabolic rate constant (g2 cm–4 s–1)
–8 1.07.6.3 Role of Minor Elements

NiO As indicated by their name, alloys contain minor
element additions which improve their mechanical
Cr2O3 properties, their corrosion resistance, etc. As a matter
–10
of fact, thermally grown oxide scales are never pure
BeO
chromia or alumina. By reacting with them, mixed
oxides are formed; these include the spinel phase,
α-Al2O3 (Mn,Cr)3O4,162,163 perovskite, MCrO3164–168 or
–12 MAlO3,169 or garnet, Y3Al5O12.170–173 Nevertheless,
SiO2
these phases are usually less protective than chromia
or alumina. Some additions have a particularly large
influence on the high temperature performance of
–14
alloys; so-called reactive elements (RE), decrease the
1300 1100 900 oxidation rate and strongly increase the oxide scale
Temperature (°C) adherence.1,6,174
Figure 10 Parabolic rate constants in oxide scales.
Adapted from Hindam, H.; Whittle, D. P. Oxid. Met. 1982,
18, 245. 1.07.6.4 Reactive Element Effect (REE)
The beneficial effect of the RE on the high tempera-
Table 1 Pilling–Bedworth ratio for different oxides ture resistance of alloys has been known for more than
50 years.175,176 Even if the reasons why REs are so
Oxides Pilling–Bedworth ratio
efficient are not fully understood, several hypotheses
K2O 0.45 have been proposed in order to explain their beneficial
MgO 0.81 effects. According to the large numbers of published
Na2O 0.97 papers on this problem, there is probably no single
Al2O3 1.28
ThO2 1.3
theory that is able to explain their role, irrespective of
ZrO2 1.56 the tested materials, the native oxide scale, or the
Cu2O 1.64 nature of the oxidizing atmosphere.177 As an example,
NiO 1.65 some RE additions give the well-known beneficial
TiO2 1.76 effect, whereas others can have a detrimental influence
CoO 1.86
Cr2O3 2.07
on the oxidation performance of the alloy.
Ta2O5 2.5 A decrease in the oxidation rate and an improve-
Nb2O5 2.68 ment in the oxide layer adherence are usually the final
V2O5 3.19 result arising from RE additions.1,178 Generally, REs
WO3 3.3 favor the nucleation and the growth of scales.179–183
They decrease the oxide grain size, and consequently
a tube is not the same as that in an oxide grown on a change scale plasticity184,185 and creep,186,187 which
flat coupon.159 can modify the growth and/or thermal stresses gener-
However, the amplitude of oxide spallation cannot ated within the scale during its growth or on cooling to
be always ascribed to residual stress in the scale.160 room temperature.188–190 They prevent the detrimen-
Oxide spallation is also strongly connected to the tal sulfur effect191–193 and suppress dislocation climb
process of growth of the oxide scale. During its ther- in the metal and, then, limit cation transport (poisoned
mal growth, cation vacancies are generated at the interface model).194,195 REs can also influence the
metal–oxide interface; if they coalesce together, cav- formation and/or transformation of transient phases,
ities are formed, weakening the interface.161 especially in alumina-forming alloys.196,197
T (⬚C)
1100 1000 900 800 700 600
1E-8
Fe–37.5Cr [Felten]
Fe–25Cr [Felten]
1E-9
Ni–15Cr-Co [Chen]
Ni–19Cr-Co [Chen]
1E-10 Ni–18Cr-Co [Chen]
Fe–25Cr [Przybylski]
Fe–16Cr [Przybylski]
Fe–25Cr [Pieraggi]
kp (g2 cm–4 s–1)
1E-11
Fe–17Cr [Perez]
Fe–30Cr [Cabouro]
1E-12 F17Ti [Chevalier]
1E-13
1E-14
1E-15
1E-16
6.0⫻10–4 7.0⫻10–4 8.0⫻10–4 9.0⫻10–4 1.0⫻10–3 1.1⫻10–3 1.2⫻10–3
(a) T –1 (K–1)
T (°C)
1100 1000 900 800 700 Fe-25Cr-Y [Felten]
Fe-37Cr-Y [Felten]
Ni-Cr-ThO2 [Giggins]
Co-45Cr impl. Y [Przybylski]
1E-9 Co-25Cr-Y [Hou]
Co-25Cr impl. La [Hou]
Co-25Cr impl. Y [Hou]
1E-10 Fe-16Cr + CeO2 [Mitra]
Fe-25Cr + CeO2 [Mitra]
AISI 304 + CeO2 [Aguilar]
kp (g2 cm–4 s–1)
AISI 304 + Nd2O3 [Bonnet]

1E-11
Fe-25Cr + CeO2 [Hussey]
Ni-30Cr-Y [Zhang]
Fe-20Cr-La-Ce* [Sigler]
1E-12 F17Ti + Nd2O3 [Chevalier]
F17Ti + Nd2O3-Cr2O3 [Chevalier]
FeCrNi impl. Y [Pérez]
1E-13 FeCrNi impl. Er [Pérez]
AISI 304 impl. Y [Riffard]
AISI 304 + Y2O3 [Riffard]
1E-14 Fe-30Cr + Nd2O3 [Chevalier]
Fe-30Cr + Y2O3 [Cabouro]
Fe-17Cr-Y [Pérez]
1E-15
6.0⫻10–4 7.0⫻10–4 8.0⫻10–4 9.0⫻10–4 1.0⫻10–3 1.1⫻10–3

(b) T –1 (K–1)
Figure 11 Parabolic rate constants established for the growth of chromia scales: (a) undoped and (b) RE doped alloys.
(*Atmosphere N2+H2O+CO2).
1.07.6.4.1 Influence of RE on oxide scale efficiency in increasing the oxidation resistance.

growth kinetics However, the results are mainly contradictory. For
The RE influence on oxide scale growth rate is shown example, no relationship could be found between the
clearly on the kinetic curves for chromia-forming kinetics (kp and mass gain) of REO coated-ferritic
materials. The oxidation rate can be divided by a factor steels and the corresponding RE ionic radius.1,251
of 100 in the presence of the RE. The comparison of
kp values shown in Figures 11(a) and 11(b)198–214 1.07.6.4.4 Influence of RE on oxide scale
illustrates this decrease. The influence of RE on alu- growth mechanism
mina growth rates is much more limited.215 The main consequence of the RE segregation at the
The quantity of RE introduced in the alloys plays a oxide grain boundaries is a change in the oxide scale
considerable role on the oxidation rates. When added growth mechanism. The transport of oxygen becomes
in amounts above a critical content, the RE leads to predominant in RE doped chromia layers,252–255 as
negative effects, including a large increase in the oxi- well as in alumina scales.256–260
dation rate.216,217 This ‘overdoping’ effect is probably Isotopic marker experiments in 16O2/18O2 dem-
due to an increase in oxygen diffusion due to the onstrate the change in the oxide growth mechanism
presence of oxygen vacancies in the RE oxides (REO). of RE doped oxides. The presence of the 18O peak
below the 16O on SIMS profiles shows that oxygen
1.07.6.4.2 Influence of RE on oxide scale transport is the main diffusion process during the
adherence scale growth.261
The most spectacular effect of RE introduction on As a consequence, the morphology and the size of
oxide scale behavior is their much improved adher- the oxide grains forming the scale change. Indeed, the
ence to the underlying substrate.213,218–220 Cyclic annihilation of cation diffusion promotes the forma-
oxidation tests clearly show the resistance to spall- tion of equiaxed chromia grains,262,263 instead of
ation, since the oxide layers formed on the RE doped columnar grain for the undoped steels, and thin
steels remain adherent to the alloys, even after many columnar alumina grains for alumina scales, instead
cycles (see example in Figure 5). of coarse oxide grains.264,265
The segregation of RE at oxide grain boundaries
1.07.6.4.3 Location of RE in oxide scales leads also to a decrease in the oxygen diffusion coef-
The RE is usually very difficult to detect within the ficient1,215 even if the results appear somewhat
oxide scale. Their presence at the surface of oxide contradictory.
layers as perovskite phases, MCrO3164,221–224 or The change in the oxide scale growth mechanism
MAlO3,225 garnets217,226–228 or as not well defined associated with the decrease in the oxygen diffusion
phases (CeTi21O28,229 NdTi21O38,213 YAlxOy230,231) are the main reasons why REs lead to improved
has been observed after long exposures in oxidizing adhesion and slower growth rates for chromia scales
atmospheres at high temperature. and improved adhesion for alumina scales.
The precise location of the RE within the oxide scale
needs to be observed in TEM cross-sections.232–247
Most of the TEM analyses have revealed the RE at
the oxide grain boundaries within the thermally grown References
scale as well as at the metal–oxide interface. The RE
solubility in Cr2O3 and in Al2O3 is very low (around 1. Chevalier, S. Traitements de surface et nouveaux
matériaux: quelles solutions pour lutter contre la
10 wt ppm for Y in Al2O3248). dégradation des matériaux à haute température? Les
Previous results have shown that the efficient dop- Editions Universitaires de Dijon, 2007.
ant in thermally grown oxides (those for which the 2. Huntz, A. M.; Pieraggi, B. Oxydation des métaux
et alliages; Lavoisier: Paris, 2003.
oxidation rate is decreased, the oxide scale adherence 3. Bénard, J. L’oxydation des métaux; Gauthier-Villars et
is improved and the oxide grain morphology is Cie: Paris, 1962.
changed) usually segregates to the oxide grain bound- 4. Béranger, G.; Colson, J. C.; Dabosi, F. Corrosion des
Matériaux à Haute Température; Les Editions de
aries. Nevertheless, other dopants, such as Ti, Nb, Ta, Physique, Ecole d’Hiver du CNRS: Piau-Engaly, 1985.
Ca, are found to segregate at the oxide grain bound- 5. Caillet, M.; Galerie, A.; Audisio, S.; Mazille, H. La
aries but without giving any beneficial improvement in protection contre la corrosion; Les Editions de Physique:
Paris, 1987.
the oxide layer performances at high tempera- 6. Kofstad, P. High Temperature Corrosion; Elsevier
ture.249,250 The dopant size could play a role in its Applied Science, 1988.
7. Sarrazin, P.; Galerie, A.; Fouletier, J. Les mécanismes de 43. Lederer, C. M.; Hollander, J. M.; Perlman, I. Table of
la corrosion sèche. Une approche cinétique; EDP Isotopes, 6th ed.; Wiley: New-York, 1967.
Sciences, 2000. 44. Fielitz, P.; Borchardt, G.; Schmücker, M.; Schneider, H.
8. Green, A.; Bastow, B. In Proceedings of the 2nd Solid State Ionics 2003, 160, 75.
International Conference on the Microscopy of Oxidation; 45. Suzuoka, T. Trans. Jpn. Inst. Met. 1961, 2, 25.
Newcomb, S. B., Bennett, M. J., Eds.; Maney Publishing, 46. Suzuoka, T. J. Phys. Soc. Jpn. 1964, 19, 839.
1993; p 137. 47. Whipple, R. T. P. Philos. Mag. 1954, 45, 1225.
9. Jianian, S.; Longjiang, Z.; Tiefan, L. Oxid. Met. 1997, 48, 48. Le Claire, A. D. Philos. Mag. 1962, 7, 141.
347. 49. Nowick, A. S.; Burton, J. J. Eds. Diffusion in Solids;
10. Yurek, Z. Solid State Phenom. 1995, 41, 185. Academic Press, 1975.
11. Henry, S. PhD Thesis, INPG, Grenoble, 2000. 50. Tsaı̈, S. C.; Huntz, A. M.; Dolin, C. Oxid. Met. 1995, 43,
12. Hindam, H.; Whittle, D. P. Oxid. Met. 1982, 18, 245. 581.
13. Schütze, M. Proceedings of the Euroconference 2000 51. Le Gall, M.; Lesage, B.; Bernardini, J. Philos. Mag. A
ProEnMach Clausthal-Zellerfeld: Germany, 20–22 Juillet, 1994, 70, 761.
2000. 52. Fielitz, P.; Borchardt, G.; Schücker, M.; Schneider, H.
14. Tien, J. K.; Davidson, J. M. In Stress Effects and the Solid State Ionics 2006, 177, 493.
Oxidation of Metals; Cathcart, J. V., Ed.; AIME: New 53. Tsai¨, S. C.; Huntz, A. M.; Dolin, C. Mater. Sci. Eng. A
York, 1975; p 200. 1996, 212, 6.
15. Delaunay, D.; Huntz, A. M.; Lacombe, P. Corros. Sci. 54. Chevalier, S.; Bonnet, G.; Borchardt, G.; Colson, J. C.;
1980, 20, 1109. Larpin, J. P. Ceram. Polish Ceram. Bull. 2000, 61, 177.
16. Huntz, A. M. Mater. Sci. Technol. 1988, 4, 1079. 55. Hagel, W. C.; Seybolt, A. U. J. Electrochem. Soc. 1961,
17. Huntz, A. M.; Schuetze, M. Mater. High Temp. 1994, 12, 108, 1146.
151. 56. Hagel, W. C. J. Am. Ceram. Soc. 1965, 48, 70.
18. Evans, H. E. Mater. High Temp. 1994, 12, 219. 57. Lobnig, R. E.; Schmidt, H. P.; Hennesen, K.; Grabke, H.
19. Evans, H. E. Int. Mater. Rev. 1995, 40, 1. J. Oxid. Met. 1992, 37, 81.
20. Schuetze, M. J. Corros. Sci. Eng. 2003, 6; http://www. 58. Graham, M. J.; Eldridge, J. I.; Mitchell, D. F.; Hussey, R.
umist.ac.uk/corrosion/ JCSE/ J. Mater. Sci. Forum 1989, 43, 207.
21. Tolpygo, V. K.; Clarke, D. R. Mater. Sci. Eng. A 2000, 278, 59. Huntz, A. M. J. Phys. III France 1995, 5, 1729.
142. 60. Tsaı̈, S. C.; Huntz, A. M.; Philibert, J. Defects Diffusion
22. Brumm, M. W.; Grabke, H. J. Corros. Sci. 1993, 34, 547. Forum 1997, 143–147, 1195.
23. Gaboriaud, R. Thermodynamique: Eléments 61. Nicholls, J. R. In Guidelines for Methods of Testing and
fondamentaux de l’énergétique et de la cinétique Research in High Temperature Corrosion; Grabke, H. J.,
chimique; Ellipses: Paris, 1998. Meadowcroft, D. B., Eds.; 1995, p 11.
24. Levin, E. M.; McMurdie, H. F. Phase Diagrams for 62. Grabke, H. J. In Guidelines for Methods of Testing and
Ceramists; The American Ceramic Society, 1975; p 17. Research in High Temperature Corrosion; Grabke, H. J.,
25. Barret, P. Cinétique hétérogène; Gauthier-Villars: Paris, Meadowcroft, D. B., Eds.; 1995, p 52.
1973. 63. Gosmain, L. Ph.D. Thesis, Université de Bourgogne,
26. Soustelle, M. Modélisation macroscopique des Dijon, 2001.
transformations physico-chimiques; Masson: Paris, 64. Jedlinski, J. Solid State Phenom. 1992, 21–22, 335.
1990. 65. Göbel, M. Ph.D. Thesis. Technische Universität
27. Antoni, L.; Galerie, A. Corrosion sèche des métaux, Clausthal, Clausthal-Zellerfeld, 1997.
Techniques de l’Ingénieur, 2003, M4 228, p 9. 66. Åkermark, T. Oxid. Met. 1998, 50, 167.
28. Quadakkers, W. J.; Naumenko, D.; Wessel, E.; 67. Berger, P.; Gaillet, L.; El Tahhann, R.; Moulin, G.;
Kochubey, V.; Singheiser, L. Oxid. Met. 2004, 61, 17–37. Viennot, M. Nucl. Instr. Meth. Phys. Res. B 2001, 181,
29. Cox, B. J. Nucl. Mater. 2005, 336, 331. 382.
30. Pétigny, N.; Barberis, P.; Lemaignan, C.; Valot, C. H.; 68. Alibhai, A. A.; Chater, R. J.; McPhail, D. S.; Shollock, B. A.
Lallemant, M. J. J. Nucl. Mater. 2000, 280, 318. Appl. Surf. Sci. 2003, 203–204, 630.
31. Tupin, M.; Pijolat, M.; Valdivieso, F.; Soustelle, M.; 69. Delaunay, D.; Huntz, A. M.; Lacombe, P. Corros. Sci.
Frichet, A.; Barberis, P. J. Nucl. Mater. 2003, 317, 130. 1980, 20, 1109.
32. Monceau, D.; Pieraggi, B. Oxid. Met. 1999, 50, 477. 70. Huntz, A. M. Mater. Sci. Technol. 1988, 4, 1079.
33. Cadoret, Y.; Monceau, D.; Bacos, M. P.; Josso, P.; 71. Huntz, A. M.; Schütze, M. Mater. High Temp. 1994, 12,
Maurice, V.; Marcus, P. Oxid. Met. 2005, 64, 185. 151.
34. Chevalier, S.; Molins, R.; Heintz, O.; Larpin, J. P. In 72. Evans, H. E. Mater. High Temp. 1994, 12, 219.
Proceedings of the 6th International Conference on the 73. Evans, H. E. Int. Mater. Rev. 1995, 40, 1.
Microscopy of Oxidation, Birmingham, England; 74. Evans, H. E. In Proceedings of the Workshop on Cyclic of
Tatlock, G. J., Evans, H. E., Eds.; 2005; p 365. High Temperature Materials; Schütze, M.,
35. Chevalier, S.; Galerie, A.; Heintz, O.; Chassagnon, R.; Quadakkers, W. J., Eds.; IOM Communications, 1999;
Crisci, A. Mater. Sci. Forum 2008, 595–598, 915. Vol. 27, p 3.
36. Wagner, C. J. Electrochem. Soc. 1952, 99, 369. 75. Moon, C. O.; Lee, S. B. Oxid. Met. 1993, 39, 1.
37. Harrison, L. G. Trans. Faraday Soc. 1961, 57, 1191. 76. Smialek, J. L.; Jayne, D. T.; Schaeffe, J. C.; Murphy, W. H.
38. Philibert, J. Diffusion et transport de matière dans les Thin Solid Films 1994, 253–285.
solides; Les Editions de Physique, 1990; p 227. 77. Jedlinski, J.; Bennett, M. J.; Evans, H. E. Mater. High
39. Philibert, J.; Vignes, A.; Bréchet, Y.; Combrade, P. Temp. 1994, 12, 169.
Métallurgie: Du Minerai au Matériau; Masson, 1998, p 389. 78. Pennefather, R. C.; Boone, D. H. Surf. Coat. Technol.
40. Hart, E. W. Acta. Metall. 1957, 5, 597. 1995, 76–77, 47.
41. Smeltzer, W. W.; Haering, R. R.; Kirkaldy, J. S. Acta 79. Chan, K. W. Metal. Mater. Trans. 1997, 28A, 411.
Metall. 1961, 9, 880. 80. Gleeson, B.; Harper, M. A. Oxid. Met. 1998, 49, 373.
42. Atkinson, A.; Taylor, R. I.; Hughes, A. Philos. Mag. A 81. Strehl, G.; Al-Badairy, H.; Rodriguez, L. M.; Klöwer, J.;
1982, 45, 823. Borchardt, G.; Tatlock, G.; Criado, A. J. In Proceedings of
the Workshop on Cyclic of High Temperature Materials; 105. Bouvier, P.; Godlewski, J.; Lucazeau, G. J. Nucl. Mater.
Schütze, M., Quadakkers, W. J., Eds.; IOM 2002, 300, 118.
Communications, 1999; Vol. 27, p 82. 106. Kemdehoundja, M.; Dinhut, J. F.; Grosseau-Poussard, J. L.;
82. Chevalier, S.; Bonnet, G.; Valot, C.; Colson, J. C.; Jeannin, M. Mater. Sci. Eng. A 2006, 435–436, 666.
Larpin, J. P. In Proceedings of the Workshop on Cyclic of 107. Kadleikova, M.; Breza, J.; Vesely, M. Microelectr. J.
High Temperature Materials; Schütze, M., 2001, 32, 955.
Quadakkers, W. J., Eds.; IOM Communications, 1999; 108. Krishnan, R.; Kesamoorthy, R.; Dash, S.; Tyagi, A. K.;
Vol. 27, p 421. Raj, B. Scr. Mater. 2003, 48, 1099.
83. Amano, T.; Watanabe, T.; Michiyama, K. Oxid. Met. 109. Tolpygo, V. K.; Clarke, D. R. Mater. High Temp. 2000,
2000, 53, 451. 17, 59.
84. Baleix, S.; Bernhart, G.; Lours, P. Mater. Sci. Eng. 2002, 110. Peng, X.; Clarke, D. R.; Wang, F. Oxid. Met. 2003, 60,
A327, 155. 225.
85. Kulinska, A.; Choux, C.; Chevalier, S. Intermetallics 2008, 111. Laxman, S.; Franke, B.; Kempshall, B. W.; Sohn, Y. H.;
16, 1. Giannuzzi, L. A.; Murphy, K. S. Surf. Coat. Technol. 2004,
86. Vangeli, P. In ; Proceedings of the Workshop on Cyclic of 177–178, 121.
High Temperature Materials; Schütze, M., 112. Murphy, K. S.; More, K. L.; Lance, M. J. Surf. Coat.
Quadakkers, W. J., Eds.; IOM Communications, 1999; Technol. 2001, 146–147, 152.
Vol. 27, p 198. 113. Sohn, Y. H.; Vaidyanathan, K.; Ronski, M.; Jordan, E. H.;
87. Graham, M. J. Mater. High Temp. 2000, 17, 1. Gell, M. Surf. Coat. Technol. 2001, 146–147, 102.
88. Rhüle, M.; Schumann, E. In Guidelines for Methods 114. Mu, N.; Izumi, T.; Zhang, L.; Gleeson, B. Mater. Sci.
of Testing and Research in High Temperature Forum 2008, 595–598, 239.
Corrosion; Grabke, H. J., Meadowcroft, D. B., Eds.; EFC 115. Kuiry, S. C.; Seal, S.; Fei, W.; Quick, N. Oxid. Met. 2003,
Publication 14, Maney Publishing, 1995, p 177. 59, 543.
89. Amelinckx, S. van Dyck, D. van Landuyt, J.; van 116. Berthomé, G.; N’Dah, E.; Wouters, Y.; Galerie, A. Mater.
Tendeloo, G. Eds. Handbook of Microscopy: Corros. 2005, 56, 389.
Applications in Materials Science, Solid-State Physics 117. Maurice, V.; Despert, G.; Zanna, S.; Josso, P.; Bacos, M.
and Chemistry Wiley-VCH, 1997, p 751. P.; Marcus, P. Acta Mater. 2007, 55, 3315.
90. Baunier, L. Préparation des échantillons, Ecole 118. Petit, J. P.; Mermoux, M.; Wouters, Y.; Galerie, A.;
Thématique: Microscopie des Défauts Cristallins; St Chemarin, C. Mater. Sci. Forum 2004, 461–464, 681.
Pierre d’Oléron, 2001; p 393. 119. Wouters, Y.; Bamba, G.; Galerie, A.; Mermoux, M.;
91. Langford, R. M.; Rogers, M. Micron 2008, 39, 1325. Petit, J. P. Mater. Sci. Forum 2004, 461–464, 830.
92. Przybylski, K.; Chevalier, S.; Juzon, P.; Galerie, A.; 120. Wouters, Y.; Marchetti, L.; Galerie, A.; Petit, J. P. Corros.
Borchardt, G.; Heintz, O.; Larpin, J. P. Mater. Sci. Forum Sci. 2008, 50, 1122.
2008, 595–598, 1103. 121. Wouters, Y.; Galerie, A.; Petit, J. P. Mater. Sci. Forum
93. Tortorelli, P. F.; More, K. L.; Specht, E. D.; Pint, B. A.; 2008, 595–598, 1181.
Zschack, P. Mater. High Temp. 2003, 20, 57. 122. Chevalier, S.; Strehl, G.; Borchardt, G.; Larpin, J. P.; Le
94. Fukuda, K.; Takao, K.; Hoshi, T.; Usui, Y.; Furukimi, O. Coze, J. Matér. Technol. 2003, 7–8–9, 71.
Mater. High Temp. 2003, 20, 73. 123. Bernard, F. PhD Thesis, Université de Bourgogne, Dijon,
95. Mayer, J.; Penkalla, H. J.; Dimyati, A.; Dani, M.; 1993.
Untoro, P.; Naumenko, D.; Quadakkers, W. J. Mater. 124. Valot, C. PhD Thesis, Université de Bourgogne, Dijon,
High Temp. 2003, 20, 167. 1995.
96. Ford, S.; Young, D. J.; McGrouther, D.; Munroe, P. R. 125. Moulin, G.; Huntz, A. M.; Lacombe, P. Acta Metall. 1980,
In Proceedings of the 6th International Conference 28, 745.
on The Microscopy of Oxidation, Birmingham, 126. Strehl, G.; Beaven, P.; Lesage, B.; Borchardt, G. Mater.
England,; Tatlock, G. J., Evans, H. E., Eds.; 2005; Corros. 2006, 56, 778.
p 183. 127. Quadakkers, W. J.; Bongartz, K. Mater. Corros. 1994, 45,
97. Svensson, H.; Stiller, K. In Proceedings. of the 6th 232.
International Conference on The Microscopy of 128. Klöwer, J. Mater. Corros. 1998, 49, 758.
Oxidation, Birmingham, England; Tatlock, G. J., 129. Poquillon, D.; Monceau, D. Oxid. Met. 2003, 59, 409.
Evans, H. E., Eds.; 2005; p 237. 130. Kartono, R.; Monceau, D.; Young, D. J. Scr. Mater. 2007,
98. Kilo, M.; Borchardt, G.; Lesage, B.; Kaitasov, O.; 57, 647.
Weber, S.; Scherrer, S. J. Eur. Ceram. Soc. 2000, 20, 131. Smialek, J. L. Acta Mater. 20004, 52, 2111.
2069. 132. www.matweb.com
99. Schmidt, H.; Borchardt, G.; Weber, S.; Scherrer, S.; 133. Materials database from NIMS, www.NIMS.jp
Baumann, H.; Müller, A.; Bill, J. J. Appl. Phys. 2000, 88, 134. Barralis, J.; Maeder, G. Précis métallurgie: Elaboration,
1827. Structures-Propriétés, Normalisation; Nathan: Paris, 1997.
100. Kilo, M.; Borchardt, G.; Lesage, B.; Weber, S.; 135. Lacombe, P.; Barroux, B.; Béranger, G. Les aciers
Scherrer, S.; Schroeder, M.; Schulz, O.; Martin, M. inoxydables; Les Editions de Physique: Paris, 1990.
Electrochem. Soc. Proc. 2001, 16, 275. 136. Philibert, J.; Vignes, A.; Bréchet, Y.; Combrade, P.
101. Kilo, M.; Weller, M.; Borchardt, G.; Damson, B.; Métallurgie: du minerai au matériau; Masson:
Weber, S.; Scherrer, S. Defects Diffusion Forum 2001, Paris, 1998.
194–199, 1039. 137. Schuetze, M. In Proceedings of the Euroconference 2000
102. Legros, C.; Lesage, B.; Kilo, M.; Strehl, G.; Borchardt, G.; ProEnMach Clausthal-Zellerfeld: Germany, 20–22 Juillet,
Carry, C. Key Eng. Mater. 2002, 206–213, 401. 2000.
103. Bamba, G.; Wouters, Y.; Galerie, A.; Borchardt, G.; 138. Mikkelsen, L.; Linderoth, S.; Bilde-Sorensen, J. B. Mater.
Shimada, S.; Heintz, O.; Chevalier, S. Scr. Mater. 2007, Sci. Forum 2004, 461–464, 117.
57, 671. 139. Schütze, M.; Quadakkers, W. J. Cyclic Oxidation of High
104. Mougin, J.; Le Bihan, T.; Lucazeau, G. J. Phys. Chem. Temperature Materials; IOM Communications: London,
Solids 2001, 62, 553. 1999.
140. Schütze, M.; Malessa, M. Mater. Corros. 2006, 57, 5. 180. Stringer, J.; Wilcox, B. A.; Jaffee, R. I. Oxid. Met. 1972,
141. Osgerby, S.; Pettersson, R. Mater. Corros. 2006, 5, 11.
57, 14. 181. Whittle, D. P.; Stringer, J. Philos. Trans. R. Soc. Lond. A
142. Niewolak, L.; Malessa, M.; Coleman, S. Y.; 1980, 295, 309.
Quadakkers, W. J.; Schütze, M. Mater. Corros. 182. Yedong, H.; Stott, F. H. Corros. Sci. 1994, 36, 1869.
2006, 57, 31. 183. Ramanarayanan, T. A.; Ayer, R.; Petkovic-Luton, R.;
143. Tien, J. K.; Davidson, J. M. In Stress Effects and the Leta, D. P. Oxid. Met. 1988, 29, 445.
Oxidation of Metals; Cathart, J. V., Ed.; AIME: New York, 184. Zhang, Y.; Gerberich, W. W.; Shores, D. A. J. Mater. Res.
1975; p 200. 1997, 12, 697.
144. Delaunay, D.; Huntz, A. M.; Lacombe, P. Corros. Sci. 185. Zhu, D.; Stout, J. H.; Nelson, J. C.; Shores, D. A. Mater.
1980, 20, 1109. Sci. Forum 1997, 251–254, 437.
145. Huntz, A. M. Mater. Sci. Technol. 1988, 4, 1079. 186. Tolpygo, V. K.; Dryden, J. R.; Clarke, D. R. Acta Mater.
146. Huntz, A. M.; Schütze, M. Mater. High Temp. 1998, 46, 927.
1994, 12, 151. 187. Tolpygo, V. K.; Clarke, D. R. Oxid. Met. 1998, 49, 187.
147. Evans, H. E. Mater. High Temp. 1994, 12, 219. 188. Huntz, A. M. Mater. Sci. Eng. A 1987, 87, 251.
148. Evans, H. E. Int. Mater. Rev. 1995, 40, 1. 189. Forest, C.; Davidson, J. H. Oxid. Met. 1995, 43, 479.
149. Schütze, M. J. Corros. Sci. Eng.; 2003, 6 http://www. 190. Sarioglu, C.; Stiger, M. J.; Blachere, J. R.;
umist.ac.uk/corrosion/JCSE/. Janakiraman, R.; Schumann, E.; Ashary, A.; Pettit, F. S.;
150. Mougin, J.; Lucazeau, G.; Galerie, A.; Dupeux, M. Mater. Meier, G. H. Mater. Corros. 2000, 51, 358.
Sci. Eng. A 2001, 308, 118. 191. Funkenbusch, A. W.; Smeggil, J. G.; Bornstein, N. S.
151. Pilling, N. B.; Bedworth, R. E. J. Inst. Met. 1923, Metal. Trans. A 1985, 16, 1164.
1, 529. 192. Lees, D. G. Oxid. Met. 1987, 27, 75.
152. Béranger, G.; Huntz, A. M.; Pieraggi, B. In Corrosion des 193. Smialek, J. L. In Proceedings of the International
Matériaux à Haute Température; Béranger, G., Colson, J. Conference on the Microscopy of Oxidation; Bennett, M.
C., Dabosi, F., Eds.; Les Editions de Physique, Ecole J., Lorimer, G. W., Eds.; 1990; p 258.
d’Hiver du CNRS: Piau-Engaly, 1985; p 227. 194. Pieraggi, B.; Rapp, R. A. J. Electrochem. Soc. 1993, 140,
153. Bernstein, H. L. Metall. Trans. A 1987, 18, 975. 2844.
154. Evans, H. E. Intern. Mater. Rev. 1995, 40, 1. 195. Pieraggi, B.; Rapp, R. A.; Hirth, J. P. Oxid. Met. 1995,
155. Bull, S. J. Oxid. Met. 1998, 49, 1. 44, 63.
156. Tolpygo, V. K.; Dryden, J. R.; Clarke, D. R. Acta Mater. 196. Rommerskirchen, I.; Kolarik, V. Mater. Corros. 1996,
1998, 46, 927. 47, 625.
157. Clarke, D. R. Acta Mater. 2003, 51, 1393. 197. Montealegre, M. A.; Gonzalez-Carrasco, J. L.
158. Rhines, F. N.; Wolf, J. S. Metall. Trans. 1970, 1, 1701. Intermetallics 2003, 11, 169.
159. Rosenband, V.; Gany, A. Corros. Sci. 1995, 37, 1991. 198. Chen, J. H.; Rogers, P. M.; Little, J. A. Oxid. Met. 1997,
160. Tolpygo, V. K.; Clarke, D. R. Mater. Sci. Eng. A 2000, 278, 47, 381.
142. 199. Felten, A. J. J. Electrochem. Soc. 1961, 108, 490.
161. Brumm, J. M. W.; Grabke, H. J. Corros. Sci. 1993, 34, 200. Pérez, J.; Cristobal, M. J.; Hierro, M. P.; Arnau, G.;
547. Botella, J. Oxid. Met. 2000, 54, 87.
162. Fergus, W. Mater. Sci. Eng. A 2005, 397, 271. 201. Brylewski, T.; Prazuch, J.; Przybylski, K. Ceram. Polish
163. Quadakkers, W. J.; Greiner, H.; Hänsel, M.; Pattanaik, A.; Ceram. Bull. 2000, 61, 221.
Khanna, A. S.; Malléner, W. Solid State Ionics 1996, 202. Brylewski, T.; Nanko, M.; Maruyama, T.; Przybylski, K.
91, 55. Solid State Ionics 2001, 143, 131.
164. Riffard, F.; Buscail, H.; Caudron, E.; Cueff, R.; Issartel, C.; 203. Pieraggi, B.; Rapp, R. A. J. Phys. IV 1993, 3, 275.
Perrier, S. Appl. Surf. Sci. 2002, 199, 107. 204. Cabouro, G.; Caboche, G.; Chevalier, S.; Piccardo, P.
165. Seybolt, A. U. Corros. Sci. 1966, 6, 263. J. Power Sources 2006, 156, 39.
166. Downham, D. A.; Shendye, S. B. Oxid. Met. 1995, 43, 205. Giggins, C. S.; Pettit, F. S. Metall. Trans. 1971,
411. 2, 1071.
167. Cristobal, M. J.; Gibson, P. N.; Stroosnijder, M. F. Corros. 206. Przybylski, K.; Garratt-Reed, A. J.; Yurek, G. J.
Sci. 1996, 38, 805. J. Electrochem. Soc. 1988, 135, 509.
168. Paul, A.; Odriozola, J. A. Mater. Sci. Eng. A 2001, 300, 22. 207. Hou, P. Y.; Stringer, J. Oxid. Met. 1988, 29, 45.
169. Tien, J. K.; Pettit, F. S. Metall. Trans. 1972, 3, 1587. 208. Mitra, S. K.; Roy, S. K.; Bose, S. K. Oxid. Met. 1990,
170. Cueff, R.; Buscail, H.; Caudron, E.; Riffard, F. Mater. Sci. 34, 101.
Forum 2001, 369–372, 311. 209. Bonnet, G.; Larpin, J. P.; Colson, J. C. Solid State Ionics
171. Kim, K. Y.; Kim, S. H.; Kwon, K. W.; Kim, I. H. Oxid. Met. 1992, 51, 11.
1994, 41, 179. 210. Hussey, R. J.; Papaiacovou, P.; Mitchell, D. F.;
172. Pint, B. A.; Hobbs, L. W. J. Electrochem. Soc. 1994, 141, Graham, M. J. Mater. Sci. Eng. A 1989, 120, 147.
2443. 211. Zhang, Y.; Zhu, D.; Shores, D. A. Acta Metall. Mater.
173. Pint, B. A. Mater. Sci. Forum 1997, 251–254, 397. 1995, 43, 4015.
174. Lang, E. Ed. The Role of Active Elements in the Oxidation 212. Sigler, D. R. Oxid. Met. 1996, 46, 335.
Behaviour of High Temperature Metals and Alloys; 213. Chevalier, S.; Bonnet, G.; Larpin, J. P.; Colson, J. C.
Elsevier Applied Science, 1989; p 33. Oxid. Met. 1997, 47, 53.
175. Pfeil, L. B. U.K. Patent 459848, 1937. 214. Riffard, F.; Buscail, H.; Caudron, E.; Cueff, R.; Issartel, C.;
176. Pfeil, L. B. U.K. Patent 574088, 1945. Perrier, S. Appl. Surf. Sci. 2002, 199, 107.
177. Quadakkers, J.; Singheiser, L. Mater. Sci. Forum 2001, 215. Chevalier, S. In Developments in High-Temperature
369–372, 77. Corrosion and Protection of Materials; Gao, W., Li, Z.,
178. Lang, E. Ed. The Role of Active Elements in the Oxidation Eds.; Woodhead Publishing: Cambridge, England, 2008;
Behaviour of High Temperature Metals and Alloys; Chapter 10.
Elsevier Applied Science, 1989; p 111. 216. Klöwer, J.; Li, J. G. Mater. Corros. 1996, 47, 545.
179. Stringer, J.; Wright, I. G. Oxid. Met. 1972, 5, 59. 217. Klöwer, J. Mater. Corros. 2000, 51, 373.
218. Chevalier, S.; Bonnet, G.; Larpin, J. P.; Colson, J. C. Quadakkers, W. J. In ; Proceedings of the 3rd International
Corros. Sci. 2003, 45, 1661. Conference on the Microscopy of Oxidation;
219. Chevalier, S.; Larpin, J. P. Mater. Sci. Eng. A 2003, 363, Newcomb, S. B., Little, J. A., Eds.; 1996; p 166.
116. 246. Kohno, M.; Ishikawa, S.; Ishi, K.; Satoh, I. In
220. Chevalier, S.; Dawah Tankeu, A. P.; Buscail, H.; Proceedings of the 3rd International Conference on the
Issartel, C.; Borchardt, G.; Larpin, J. P. Mater. Corros. Microscopy of Oxidation; Newcomb, S. B., Little, J. A.,
2004, 55, 610. Eds.; 1996; p 55.
221. Seybolt, A. U. Corros. Sci. 1966, 6, 263. 247. Molins, R.; Germidis, A.; Andrieu, E. In Proceedings of
222. Downham, D. A.; Shendye, S. B. Oxid. Met. 1995, the 3rd International Conference on the Microscopy of
43, 411. Oxidation; Newcomb, S. B., Little, J. A., Eds.; 1996; p 3.
223. Cristobal, M. J.; Gibson, P. N.; Stroosnijder, M. F. Corros. 248. Sato, E.; Carry, C. J. Am. Ceram. Soc. 1996, 79, 2156.
Sci. 1996, 38, 805. 249. Pint, B. A.; Alexander, K. B. In Proceedings of the 3rd
224. Paul, A.; Odriozola, J. A. Mater. Sci. Eng. A 2001, 300, 22. International Conference on the Microscopy of Oxidation;
225. Tien, J. K.; Pettit, F. S. Metall. Trans. 1972, 3, 1587. Newcomb, S. B., Little, J. A., Eds.; 1996; p 153.
226. Kim, K. Y.; Kim, S. H.; Kwon, K. W.; Kim, I. H. Oxid. Met. 250. Pint, B. A.; Alexander, K. B. J. Electrochem. Soc. 1998,
1994, 41, 179. 145, 1819.
227. Pint, B. A.; Hobbs, L. W. J. Electrochem. Soc. 1994, 141, 251. Chevalier, S.; Larpin, J. P. Acta Mater. 2002, 50,
2443. 3105–3114.
228. Pint, B. A. Mater. Sci. Forum 1997, 251–254, 397. 252. Quadakkers, W. J.; Speier, W.; Holzbrecher, H.;
229. Aguilard, G. Ph.D. Thesis Université de Bourgogne, Nickel, H. In Proceedings of the International Conference
Dijon, 1992. on the Microscopy of Oxidation; Bennet, M. J.,
230. Pint, B. A.; Garratt-Reed, A. J.; Hobbs, L. W. J. Am. Lorimer, G. W., Eds.; 1990; p 149.
Ceram. Soc. 1998, 81, 305. 253. Papaiacovou, P.; Hussey, R. J.; Mitchell, D. F.;
231. Wu, Y.; Hagihara, K.; Umakoshi, Y. Intermetallics 2004, Graham, M. J. Corros. Sci. 1990, 30, 451.
12, 519. 254. Hussey, R. J.; Sproule, G. I.; Graham, M. J. J. Phys. IV
232. Yurek, G. J.; Przybylski, K.; Garratt-Reed, A. J. 1993, 3, 241.
J. Electrochem. Soc. 1987, 134, 2643. 255. Quadakkers, W. J.; Norton, J. F.; Penkalla, J.; Breuer, U.;
233. Przybylski, K. Mater. Sci. Eng. A 1989, 121, 509. Gil, A.; Rieck, T.; Hänsel, M. In Proceedings of the 3rd
234. Przybylski, K.; Yurek, G. J. Mater. Sci. Forum 1989, 43, 1. International Conference on the Microscopy of
235. Chevalier, S.; Bonnet, G.; Przybylski, K.; Colson, J. C.; Oxidation; Newcomb, S. B., Little, J. A., Eds.; 1996;
Larpin, J. P. Oxid. Met. 2000, 54, 527. p 221.
236. Ul-Hamid, A. Oxid. Met. 2002, 58, 23. 256. Quadakkers, W. J.; Speier, W.; Holzbrecher, H.;
237. Przybylski, K.; Garratt-Reed, A. J.; Pint, B. A.; Katz, E. P.; Nickel, H. In Proceedings of the International Conference
Yurek, G. J. J. Electrochem. Soc. 1987, 134, 3207. on the Microscopy of Oxidation; Bennet, M. J.,
238. Pint, B. A.; Garratt-Reed, A. J.; Hobbs, L. W. In Lorimer, G. W., Eds.; 1990; p 149.
Proceedings of the 2nd International Conference on the 257. Jedlinski, J.; Borchardt, G. Oxid. Met. 1991,
Microscopy of Oxidation; Newcomb, S. B., Bennett, 36, 317.
M. J., Eds.; 1993; p 463. 258. Czyrska-Filemonowicz, A.; Szot, K.; Wasilkowska, A.;
239. Schumann, E.; Yang, J. C.; Graham, M. J.; Rhüle, M. Ennis, P. J.; Breuer, U.; Gil, A.; Quadakkers, W. J. In
Mater. Corros. 1995, 46, 218. Proceedings of the 3rd International Conference on the
240. Schumann, E.; Yang, J. C.; Graham, M. J.; Rühle, M. Microscopy of Oxidation; Newcomb, S. B., Little, J. A.,
Mater. Corros. 1996, 47, 631. Eds.; 1996; p 185.
241. Pint, B. A.; Garratt-Ree, A. J.; Hobbs, L. W. J. Phys. IV 259. Quadakkers, W. J.; Clemens, D.; Bennett, M. J. In
1993, 3, 247. Proceedings of the 3rd International Conference on the
242. Pint, B. A.; Alexander, K. B. In Proceedings of the 3rd Microscopy of Oxidation; Newcomb, S. B., Little, J. A.,
International Conference on the Microscopy of Oxidation; Eds.; 1996; p 195.
Newcomb, S. B., Little, J. A., Eds.; 1996; p 153. 260. Chevalier, S.; Przybylski, K.; Borchardt, G.; Larpin, J. P.
243. Pint, B. A.; Alexander, K. B. J. Electrochem. Soc. 1998, Mater. Sci. Forum 2004, 461–464, 53.
145, 1819. 261. Chevalier, S.; Bonnet, G.; Borchardt, G.; Colson, J. C.;
244. Mennicke, C.; Schumann, E.; Le Coze, J.; Smialek, J. L.; Larpin, J. P. Mater. Sci. Forum 2001, 369–372, 327.
Meier, G. H.; Rhüle, M. In Proceedings of the 3rd 262. Pint, B. A.; Garratt-Reed, A. J.; Hobbs, L. W. Mater. High
International Conference on the Microscopy of Temp. 1995, 13, 3.
Oxidation; Newcomb, S. B., Little, J. A., Eds.; 1996; p 95. 263. Turker, M.; Hugues, T. A. Oxid. Met. 1995, 44, 505.
245. Newcomb, S. B.; Dunin-Borkowski, R. E.; 264. Turker, M.; Hugues, T. A. Oxid. Met. 1995, 44, 505.
Boothroyd, C. B.; Czyrska-Filemonowicz, A.; Clemens, D.; 265. Chevalier, S.; Przybylski, K. Unpublished results.
1.08 Stress Effects in High Temperature Oxidation
M. Schütze
Karl Winnacker Institut der DECHEMA e.V., Theodor-Heuss-Allee 25, D-60486 Frankfurt am Main, Germany

1.08.2 Scale Stresses Arising from Oxide Growth 155
1.08.2.1 Intrinsic Growth Stresses in Oxide Scales 155
1.08.2.1.1 Experimental observations 155
1.08.2.1.2 Modeling 160
1.08.2.2 Geometrically Induced Growth Stresses 161
1.08.2.3 Other Types of Growth Stresses 161
1.08.3 Thermally Induced Scale Stresses Resulting from Temperature Changes 162
1.08.3.1 Experimental Observations 162
1.08.3.2 Modeling 164
1.08.4 The Effect of Stresses on the Integrity of the Oxide Scale 164
1.08.4.1 Measured Critical Strain Values to Scale Failure 164
1.08.4.2 Model Considerations 167
1.08.5 Maintaining the Protective Effect of Oxide Scales Even After Mechanical
Failure under Stresses 171
1.08.5.1 Healing of Cracks or other types of Damage in the Protective Oxide Scale 171
1.08.5.2 The Influence of Subsurface Zone Depletion on Maintaining the Protective
Effect of Oxide Scales Under Stresses 173
References 177
Glossary deformation, (2) resistance of a material

Acoustic emission Acoustic signals (‘noise’) against elastic deformation.
arising from the generation of transient Healing Closure of cracks or other defective sites
elastic waves produced by sudden in a layer or oxide scale by diffusion or scale
redistribution of stress in a material (e.g., growth processes.
cracking, micro-cracking, and dislocation Microcracking Formation of cracks or crack
movement). networks where the single cracks are in the
Buckling Failure by elastic instability under the range of microns, or less.
effect of compressive stresses where the Growth stresses Mechanical stresses originating
structural member (e.g., scale) bends from scale or layer growth directly due to
sidewards in the center of an unattached volume changes.
area. Protrusions Local extension of the surface scale
Coefficient of thermal expansion (CTE) Degree of into the metal substrate.
expansion/contraction divided by the Poisson ratio Ratio of transverse strain
change in temperature. (contraction strain) divided by the extension
Creep Time-dependent plastic deformation at high strain (axial strain).
temperatures (‘flow’). Physical defect Defects in a physical structure
Depletion Drop in element concentration in an alloy (material), for example, pores, voids,
during high temperature exposure. For crevices, cracks, etc.
example, by the consumption of this element Residual stress Stresses remaining in a structure
for surface oxide formation. even after removing external load.
Elastic modulus/Young’s modulus (1) Slope of Subsurface zone Area in a material directly
the stress–strain curve in elastic underneath the surface.
153
Thermogravimetry Test procedure in which the nox Poisson ratio of the oxide
mass change of a specimen due to am Mean coefficient of thermal expansion (CTE) of
high temperature corrosion is recorded by a a multi-scale system
balance. amet Coefficient of thermal expansion (CTE) of the
Thermocycling Test procedure in which a metal
specimen is subjected to regularly repeated aox Coefficient of thermal expansion (CTE) of the
temperature cycles. oxide
«c Critical strain to failure
del Elastic strain
«ext External strains resulting from component
operation
Abbreviations «geo Geometrically induced oxide growth strains
COSIM Computer code for predicting oxide
«int Intrinsic oxide growth strains
spallation
«ox Strain in the oxide
CTE Coefficient of thermal expansion
«pl Plastic strain
DTMO Differential test in monofacial oxidation
«therm Thermally induced oxide strains from
NASA National Aeronautics and Space
temperature changes
Administration
«˙ Strain rate
ODIN Computer code for predicting alloy depletion
«˙ c Critical strain rate for scale crack healing
PBR Pilling-Bedworth-ratio
l ‘Wave length’ of an interface
RT Room temperature (ambient temperature)
s Elastic stress
XRD X-ray diffraction
sc Critical stress for failure
sext External stresses resulting from component
operation
sgeo Geometrically induced oxide growth stresses
Symbols
sint Intrinsic oxide growth stresses
c Physical defect size
sox Stress in the oxide
dmet Thickness of the metal substrate
stherm Thermally induced oxide stresses from
dox Oxide scale thickness
temperature changes
r Interfacial amplitude
v Stress relaxation factor
t Time
DT Temperature change
tB Time to the beginning of breakaway oxidation
g i Intrinsic surface energy
Ao Length of a defect free interface
g o Fracture surface energy
Asep Separated length of an interface
Cc Minimum alloy concentration to form a
protective oxide scale (critical concentration
limit)
1.08.1 Introduction
D (inter-)Diffusion coefficient
E* Effective elastic (Young’s) modulus
The role of stresses in oxidation is of significance
Emet Elastic (Young’s) modulus of the metal
because these stresses can determine the lifetime of a
Eox Elastic (Young’s) modulus of the oxide
high temperature component from a corrosion point of
Gc Critical surface energy
view.1 High temperature corrosion resistance is based on
K Stress intensity factor
the protective effect of dense, slowly growing, and
Kc Fracture toughness
adherent oxide scales, that is, if any chemical or mechan-
ki Oxidation rate constant for the recession of the
ical damage occurs to these scales, the corrosion resis-
oxide/metal interface
tance of the materials will be endangered. This chapter
KIc Fracture toughness under mode I load
deals with the key aspects describing scale integrity
NA,b Atom fraction of element B in an alloy
under the effect of stresses, also taking into account
PBR Value of the Pilling-Bedworth-Ratio
the role of the superimposed oxidation processes. The
R Radius of a scale delamination
latter is important as, at elevated temperature, stresses
T Temperature
and oxidation should not be looked at separately as they
nmet Poisson ratio of the metal
are active simultaneously and interact with each other.
Therefore, not only the topics of mechanical scale Equation [1] summarizes this situation in terms of
failure (cracking, spalling, delamination, etc.), but also the strain values resulting from the different types
the mechanisms of scale damage healing and metal of stresses:
subsurface zone depletion are addressed. eint þ egeo þ etherm þ eext ¼ eel þ epl ½1
An example where the whole complexity of this
interaction becomes evident is the so-called ‘break- Failure occurs when the elastically stored energy in
away mechanism’ by which the protective period in the form of (‘elastic’) stress s reaches or exceeds a
oxidation ends and the nonprotective period starts.2 critical value sc:
From a technical point of view this is the end of the
s ¼ Eox eel sc ½2
lifetime of a component, at least under the aspect of
high temperature corrosion. Although the breakaway where Eox is the elastic modulus of the oxide scale.
effect can also be described solely based on chemical Failure occurs in terms of strains when
mechanisms (loss of chemical stability of the protective
oxide),3 under industrial conditions mechanical stres- eint þ egeo þ etherm þ eext ¼ eel þ epl ¼ ec ½3
ses are always present and the breakaway effect is where ec is the critical strain for the different failure
considerably determined from a joint interaction of mechanisms (see Section 1.08.4). As will be shown
chemical and mechanical mechanisms.3,4 In the sim- later, oxide scales usually behave in a brittle manner,
plest case the latter may be initiated from (intermedi- even at service temperatures of high temperature
ate) cooling, but also at constant operation temperature components, so that at mechanical scale failure
oxide growth stresses can reach critical limits where the epl eel or ec eel.
integrity of the scales is endangered.5 In addition to the failure mechanism, ec is also a
Generally, the following stresses in oxidation can function of the oxide/metal system and the oxidation
be distinguished1: parameters. High values of ec automatically mean
1. Intrinsic growth stresses sint resulting from the high tolerance of the system against failure by mech-
chemical nature of the oxidation process. These anical stresses.
can arise when new oxide is formed inside the The following sections deal in some detail with
existing oxide, for example, by counter diffusion the key issues of the role of stresses in oxidation. For
of oxide forming species. Other sources of such an even more detailed description of this topic, the
stresses can be epitaxial relationships between reader is referred to a textbook which is devoted
metal surface and oxide in particular at the begin- solely to this field.1
ning of oxidation or phase transformations in the
oxide scale, the metal, or coating subsurface zone.
1.08.2 Scale Stresses Arising from
2. Geometrically induced growth stresses sgeo
Oxide Growth
resulting from particular narrow surface curva-
tures where effects of the differences in volume 1.08.2.1 Intrinsic Growth Stresses in
of oxide formed and metal consumed come Oxide Scales
into play.
1.08.2.1.1 Experimental observations
3. Thermally induced stresses stherm resulting from
Most standard investigations of the oxidation behav-
temperature changes and the different values of
ior of materials do not take into account that the
the coefficients of thermal expansion (CTE) for
formation of growth stresses in the scales may take
metal/alloys and oxides.
place during the oxidation process.6 The existence of
4. External stresses sext resulting from operation, for
such stresses has, however, already become evident in
example, centrifugal stresses, gravity, constraints
an optical observation of the oxide scale surfaces as
in structures, etc.
shown as an example in Figure 1.7 In this example
All these stresses have to be accommodated by elastic the scale shows a perpendicular arrangement of
or plastic/creep deformation in the oxide/metal sys- waves that have partially detached from the substrate.
tem. Otherwise, if critical stress values are exceeded, The formation of such waves is an indication that
mechanical failure of the scale results as a stress relief additional volume had been created in the existing
mechanism. In this case, however, the protective oxide which had to be accommodated by a stress
effect of the scale may be lost unless it can be restored relaxation mechanism like detachment. This situa-
by scale healing mechanisms. tion has been observed by a number of authors.8–15
30
Oxidation of Cr
1 atm. O2
Weight gain (mg cm–2)

20
1200 ⬚C
1075 ⬚C
10
25 μm
(a)
1000 ⬚C
Mount 0
0 10 20
Detached Time (h)
scale Figure 2 Repeated cracking of scales leading to a
sequence of ‘parabolic’ stages in the overall oxidation of
Metal pure Cr at 1000–1200 C in O2.8,9
50 μm 1. Measurement of the bending of a thin metal

strip which is protected on one side and oxidized
(b) on the other side.16–22 This test is sometimes
Figure 1 ‘Wave’ formation in the oxide scale due to called differential test in monofacial oxidation
lateral growth stresses on a heat resistant 18 Cr (DTMO).21,22
steel after 100 h oxidation at 1000 C in air.7 (a) SEM 2. Measurement of the bulging of a thin metal plate
image of the surface and (b) optical micrograph of a
cross-section.
protected on one side and oxidized on the other
side.23
3. Measurement of the change in diameter of a
helix or a spiral made of a thin metal strip oxi-
As the example in Figure 1 shows, the formation of dized either on one side24,25 or on both sides.26
waves was possible without damage to the existing On the spiral, the length change can also be
oxide so that its protective effect remained. measured.
Thermogravimetric investigations of the oxidation 4. Measurement of the increase in length of thin
of pure chromium revealed that such (isothermal) metal strips or tubes resulting from creep and
growth stresses could even be sufficiently high to elastic deformation of the substrate due to the
crack the existing oxide and initiate the formation of stresses from the oxide growth.27–30
new oxide on the fresh metal surface exposed under- 5. Measurement of the curvature or a radius change
neath the waves.8,9 This becomes evident in the mass of a U-shaped thin foil specimen.31 A somewhat
change versus time curve shown in Figure 2. After each similar method is the catenary method which
cracking process, the start of a new parabolic partial makes use of a wire sample suspended as a cate-
curve is observed and in the extreme case of very high nary between two supports and where a change
growth stresses at the highest test temperature, in these in the shape of the catenary curve is recorded
investigations the cracking frequency has become so during oxidation.32
high that a quasilinear mass change curve is found. 6. Measurement of the bending of an oxidized strip
Besides the phenomenological observations, a as one of the sides is polished away stepwise and
number of measurements have been performed in the thickness is progressively reduced.33
order to quantify the level of these growth stresses. 7. Measurement of the frequency shift due to scale
The measurement techniques reported in the litera- stresses in Raman spectroscopy/photo lumines-
ture are the following: cence.34–39
8. Using changes in the photocurrent due to stresses and the oxide scale was adherent. This effect was later
by applying a microphotoelectrochemistry tech- also confirmed by Huntz et al.48 The maximum
nique.39 growth stress values which were observed in these
9. Measurement of the distortion of the oxide lat- tests were around 550 MPa at 900 C.
tice using X-rays.40–45 Much higher stresses are reported, for example,
10. Measurement by using synchrotron radiation for chromia layers growing at 750 C in pure oxy-
instead of standard X-ray sources which allows gen.38 In this case, values of up to 2.4 GPa were
one to study thinner oxides and with greater time reached as measured by Raman spectroscopy. Similar
resolution.45,46 values were measured under the same conditions for
steels of the composition Fe–18Cr and Fe–18Cr–
In particular, the foil bending measurements can TiNb. For the oxide scales on pure iron or mild
be combined with acoustic emission recording, which steel, respectively, different values were observed
allows the detection of any cracking or delamination for magnetite and hematite in the same oxide
processes resulting from the build-up of scale stres- scale.45 In magnetite, these stresses reached values
ses.5,21,22 All these measurements have been described of tensile nature up to 135 MPa at 400 C while
in great detail in the respective publications where the those in the same scale for hematite were compres-
drawbacks and the advantages of the different methods sive and went up to 270 MPa. Interestingly when
are also discussed. In the following section, only the oxidizing gray cast iron the stresses in both partial
results from measurements are addressed. layers were of compressive nature at 400 C. For
One of the systems that has been investigated in magnetite, they reached 60 MPa, while for hema-
some detail is NiO on nickel or Ni–Cr.20,21,47–49 The tite the maximum was around 570 MPa. In all cases,
reason for this is that a single phase oxide of signifi- there was first an increase up to a maximum in the
cant thickness is formed, which allows nickel oxide value of the stresses while during further oxidation,
scales to be looked at as a model system. A typical a decrease or an oscillation around a certain level
example of the course of the stress curves as a func- occurred. This was similar to the observations
tion of oxidation time is given in Figure 3(a). In described for NiO on polycrystalline Ni in Figure 3.
many cases, such curves start with a very short tensile The largest number of investigations on oxide
period before they change into a steeply increasing growth stresses focused on the behavior of alumina
compressive slope, which reaches a maximum and scales on iron or nickel based substrates. In most
then falls off slightly. Often after longer times, the cases, surprisingly high compressive stress values
level of stress starts to oscillate around a certain value were found, which make it hard to believe that the
as is also indicated in this example for the two poly- oxide scale could survive such stresses without any
crystalline nickel versions. Interestingly the oxide mechanical damage. As an example, a tensile stress
formed on the nickel single crystal shows a further value of 600 MPa on b-NiAl alloys that had been
increase of the growth stresses which falls together oxidized at 1100 C can be quoted.46 It should, how-
with the effect that no acoustic emission occurs ever, be mentioned that large contradictions also
for the single crystal specimen while on the two occur between the measured values of different
polycrystalline specimens significant acoustic emis- investigations. As an example, the measurements by
sion takes place, starting after 10 h of exposure, Hou et al.50 can be quoted with much lower values,
Figure 3(b). Acoustic emission is an indication that, which were explained by stress relaxation processes.
evidently, micro cracking and oxide delamination has Hou et al.50 also reveals that in the early stages the
taken place because of the formation of growth stres- growth stress in y-Al2O3 is compressive and is con-
ses allowing direct access of oxygen through the scale verted into tensile stresses when the formation of a-
to the oxide/metal interface and the formation of Al2O3 begins. Only when a complete a-Al2O3 layer
new fine-grained oxide at these positions. As a con- had formed at the metal/oxide interface did the
sequence, the typical two layered structure on nickel stresses in a-Al2O3 become compressive, reaching a
is formed as an oxide scale while on the single crystal steady state level around 75 MPa at 1100 C. Other
nickel only a single oxide layer appears. In Przybilla sources report growth stresses on NiAl oxidized at
and Schütze,21 this could be explained in a model by 1100 C in the range of around zero.44
the presence of impurities in the polycrystalline For FeCrAl(Y) alloys and iron-based aluminides
nickel materials which facilitates interfacial delami- extremely high stresses had been found, for example,
nation while the nickel single crystal was ultra pure up to 1.6 GPa.51 Surprisingly there is, however, a big
20 Ni, 900 ⬚C in air
–20
–40
s OX (MPa)
–60
–80
–100
Single crystal
–120 Ni, coarse grained
Ni, fine grained
–140
0 20 40 60 80 100 120
(a) t (h)
Isothermal oxidation of Ni99.99 in synth. air at 900 ⬚C

1200
1000
AE energy (nVs)
120
100
80
60
40
20
0
0 20 40 60 80 100
(b) t (h)
Figure 3 (a) Growth stress in NiO scales on nickel as a function of oxidation time at 900 C in air as determined by XRD
measurements21,49 (single crystal = 99.9999% Ni, Ni coarse or fine grained = 99.99% Ni). (b) Acoustic emission activity
in a plain oxidation test with Ni99.99 at 900 C in air. Reproduced from Przybilla, W.; Schütze, M. Oxid. Met. 2002, 58,
103–145.
scatter between very high compressive and very high during isothermal, and in particular, thermocycling
tensile stresses, depending also on the measurement oxidation of the materials. Interestingly Mennicke
technique. While in high temperature X-ray measure- et al.52 also detected that there is no significant differ-
ments compressive values as high as 1.7 GPa were ence in the growth stresses as well as in the residual
observed,44 similar investigations at the same temper- stresses after cooling or with regard to stress relaxation
ature by synchrotron X-ray radiation led to tensile during thermocycling, between oxides formed on Y-
stresses of the same order of magnitude.46 Tensile doped and undoped FeCrAl materials. Evidently, the
growth stresses of 1 GPa were also observed by presence of a reactive element in this case did not lead
Mennicke et al.52 which addresses, however, also the to a change of the oxidation mechanism in a way that
point that significant stress relaxation may take place influences the intrinsic growth stress situation.
Limited information is available on TiO2 as well of intrinsic growth stresses in oxides is attempted,
as TiO2/Al2O3 mixtures on titanium and titanium which indicates that there is still a significant scatter
aluminides, respectively. While for titania the values between the data and definitely more work is needed
are compressive and exceed 1000 MPa,39 those of to get a clearer view of the role and the level of
the mixed oxide scale on the titanium aluminides can intrinsic growth stresses in oxide scales.
be compressive as well as tensile, both in the range Finally an interesting aspect, which was observed
of 100 MPa.22 for ferritic steels forming Cr-rich oxide scales, is
In Table 1 a rather incomplete summary of the addressed. Here it turned out that water vapor-
existing information in the literature on the level containing environments lead to higher compressive
Table 1 Results from growth stress measurements of oxide scales
Oxide Substrate Temperature ( C) Max. abs. valuea (MPa) Technique Ref.
NiO Ni 627 90 53

Ni 1027 þ12/þ110 53
Ni 940 þ1365 53
Ni Grade A 900 þ20 54
Ni SX 99.999 900 130/þ15 X-ray 49
Ni 99.99 900 55/þ15 X-ray 49
Ni 99.0 900 65/þ15 X-ray 49
Ni 99.99 900 550/þ250 DTMO 21
Ni 99.0 900 400/þ300 DTMO 21
Ni 99.99 800 1500/þ700 DTMO 21
Fe2O3 Mild Steel 400 270 X-ray 45
Grey Cast Iron 400 5770 X-ray 45
Fe3O4 Mild Steel 400 þ135 X-ray 45
Grey Cast Iron 400 60 X-ray 45
Cr2O3 Cr 99.7 750 2400 Raman 38
Fe–18Cr 750 2800 Raman 55
Fe–18Cr–TiNb 750 2200 Raman 55
Cr 875 þ200 19
Cr 940 350 1
y-Al2O3 b-NiAl 1000/1100 480 Synchrotron 50
a-Al2O3 b-NiAl 1000/1100 75/þ500 Synchrotron 50
Al2O3 NiAl 1100 0 X-ray 44
FeCrAlY 1100 1300/200 X-ray 44
FeCrAl 1100 400 X-ray 44
FeCrAlY 1000 1700/500 X-ray 44
FeCrAl 1000 1000 X-ray 44
FeCrAlY 1100 þ1100/þ500 Synchrotron 46
FeCrAlY 1000 þ500 Synchrotron 46
FeCrAlY 1200 þ800/þ100 Synchrotron 46
Kanthal 1000 þ260 Synchrotron 46
Kanthal 1200 150 Synchrotron 46
NiAl 1100 þ600/300 Synchrotron 46
NiAl þ Hf 1100 þ500 46
FeNiCrAl 1000 400 DTMO 18
FeNiCrAlY 1000 80 DTMO 18
FeCrAl 1300 340 DTMO 56
TiO2 Ti99.6 1000 Raman 39
TiO2/Al2O3 TiAl2Cr 800 þ100/100 DTMO 22
TiAl 900 130 DTMO 22
Cr-(Fe)-oxide P91 steel 650/dry air 200 DTMO 5
P91 steel 650/10% H2O 1200 DTMO 5
Nf 616 650/dry air 120 DTMO 5
Nf 616 650/10% H2O 900 DTMO 5
a
Max. abs. value means that this was the highest value measured independent of the stress direction. The latter is indicated by the (þ) or
() sign: (þ), tensile; (), compressive.
stresses than dry air.5 While the stresses in dry air in the oxide can be calculated according to the fol-
oscillate around a constant value over oxidation time, lowing equation60:
there is a significant stress relaxation in the humid
environment, which leads after a while to a drop of eox ¼ o½ðPBRÞ1=3 1 ½4
the growth stresses to values close to those in dry air. where the factor o is introduced to take account of
possible stress-reducing processes. Assuming elastic
1.08.2.1.2 Modeling behavior, the corresponding growth stress in the
The earliest attempt to model the formation of oxide would be:
growth stresses in oxide scales during the oxidation
Eox
process, dates back to 1923 and was performed by sox ¼ eox ½5
Pilling and Bedworth.57 In their model, they des- 1 nox
cribed the volume change that is involved in the where Eox is the elastic modulus of the oxide, and nox
transition from the metal lattice to the cation lattice is the Poisson ratio.
of the oxide when only the oxygen anions are diffus- The PBR values have been used for many years,
ing during the oxide growth process. The ratio of the and are sometimes still used, to explain the formation
volume of the oxide formed by the oxidation process of growth stresses in oxide scales. Nowadays, however,
to the metal consumed in this reaction is termed the it is generally agreed that this model is too simplistic
Pilling–Bedworth ratio (PBR). Values of the PBR are an explanation, especially because it does not take
given in Table 2 for several oxides and metal sub- into account the outward growth of oxides by metal
strates. The key assumption in this model is that the cation transport, which is a dominant mechanism for
oxide scales grow inwardly and they would be com- many technical alloys. In the meantime a number of
pressed if the PBR were greater than one. Tensile more sophisticated models have been developed61
stresses would develop in the oxide if the PBR were describing, for example, the formation of growth
less than one. Consequently, if oxide scales grow stresses based on the formation of new oxide in the
outward by the diffusion of metal cations, the oxide existing oxide scale, as shown in Figure 4.2 The first
scale must remain free of growth stresses as the new model of this type was developed by Rhines and Wolf,
oxide grows freely on the scale surface. Assuming based on the assumption that counter diffusion takes
inward scale growth will occur, the resulting strain place in the oxide scale with the inward movement of
oxygen along grain boundaries and the outward
movement of metal cations within the oxide grains.142
Table 2 PBR values for some technically important oxi- The reactions of the two species at the oxide grain
des and metal substrates2,17,57–59 boundaries would lead to the formation of new
System PBR oxide at these positions with a volume increase
that also generates strains in a lateral direction. As
a-Al2O3/Al 1.28
counter diffusion of the oxide-forming species dur-
CoO/Co 1.86
Cr2O3/Cr 2.07 ing scale growth has also been shown in several
Cr2O3/Fe–25Cr–20Ni 2.1
CuO/Cu 1.70
FeCr2O4/Fe–18Cr–8Ni 2.1
Fe(FeCr)2O4spinel/Fe–Cr 2.1
FeO/a-Fe 1.68 Cr2O3
Fe2O3/a-Fe 2.14
Fe2O3/Fe3O4 1.02 Cr
Fe3O4/a-Fe 2.1 Cr (g) Inward O-transport
Fe3O4/FeO 1.2 M-Cr
MgO/Mg 0.84
Outward Cr-transport
MnO/Mn 1.77
MoO3/Mo 3.27
NiO/Ni 1.65 Region of oxide growth
SiO2/Si 2.15
TiO2/Ti 1.73 Figure 4 Schematic representation of the growth of
ZrO2/Zr 1.57 chromia scales by counter transport of oxygen and
chromium in the scale. Reproduced from Kofstad, P.
Reproduced from Kofstad, P. High Temperature Corrosion; High Temperature Corrosion; Elsevier Applied Science:
Elsevier Applied Science: London, 1988. London, 1988.
investigations by isotope and SIMS measurements,62 In the case of anion diffusion, the new oxide is formed
this approach has also become the basis for many of at the interface between oxide scale and metal. As
the more advanced and more recent models in the most of the technical oxides have a PBR > 1, it means
literature.63–65 Clarke was able to express the growth that additional volume is created at the interface,
strain rate in terms of a number of parameters includ- which shifts the existing oxide scale outward. As a
ing the parabolic rate constant, the scale thickness, consequence, tensile stresses will arise in circumfer-
and the oxide grain size.63 This lateral growth strain ential direction on convex surfaces, while compres-
rate increases with the parabolic rate constant and sive stresses will be present perpendicular to the
decreases with increasing oxide grain size and scale interface between oxide and metal. The circumferen-
thickness. While these models in several cases can tial stresses are the highest in the ‘oldest’ part of the
predict the measured order of magnitude of scale oxide, which is the one on the scale surface. The part
stresses correctly, they still have the problem that that was formed last is almost stress-free with regard
any stress relaxation process, which will inevitably to the circumferential direction. On concave surfaces
take place under practical conditions, is not reflected. the same situation leads to compressive circumferen-
Nevertheless, all these models can help a better tial and perpendicular stresses, as shown in the lower
understanding of the complex situation in the oxida- left part of Figure 5. For cation diffusion, new oxide
tion of particular technical alloys, which is in most is formed on top of the existing oxide and is, thus,
cases away from the simple parabolic growth mecha- the stress-free part of the system. The oldest part is at
nism usually described in textbooks. the oxide/metal interface and, thus, shows the high-
est circumferential stresses. These are compressive
for convex surfaces and tensile for concave surfaces.
1.08.2.2 Geometrically Induced Growth The stresses perpendicular to the oxide/metal inter-
Stresses face are tensile in both cases.
As described by Schütze,1 these scale stresses can
On curved substrate surfaces, the growth of oxide
be calculated quantitatively, based on the oxidation
scales can lead to growth stresses originating from
rate data and the surface curvature. The general
the surface geometry. The reason is that metal is
tendency is that the higher the oxide growth rate
consumed by the oxidation process so that the origi-
and the smaller the surface curvature radius, the
nal metal surface recedes and at the same time there
more marked will be the oxide scale stresses resulting
is a difference between the original metal volume
from the geometrical situation. As a result, fast-grow-
consumed by the oxidation process and the volume
ing oxide scales, especially at sharp edges, are greatly
of the resulting oxide, see PBR. Figure 5 illustrates
endangered from high geometrical growth stresses.
the potential situations that can exist, based on the
As will be shown later, tensile stresses especially can
extreme cases of scale growth by cation diffusion
lead to delamination of the scale or scale cracking, so
alone and by anion diffusion alone for PBR > 1.66
that low curvature radii such as edges can always be
sites of preferred attack by high temperature corro-
sion as the protective oxide scale may suffer damage
Old oxide at these spots first.
Newly formed oxide
Metal
1.08.2.3 Other Types of Growth Stresses
Anion diffusion Cation diffusion
Among the stresses arising from oxide growth, there
are many other sources of stress that can be consid-
M Convex ered intrinsic growth stresses, but are of significance
M
R R
only in very specific cases. In the literature, the role
R R of epitaxial stresses has been discussed often, where
M stresses develop as a consequence of the difference in
Concave M
lattice parameters of the oxide and the substrate.
Such stresses should be maximum at the oxide/
Figure 5 Growth stresses in the oxidation of curved
metal phase boundary and will fall off toward the
surfaces. Reproduced from Christl, W.; Rahmel, A.; oxide surface.67 It can be expected that with the
Schütze, M. Oxid. Met. 1989, 31, 1. increasing thickness of the oxide scale, such epitaxial
stresses will have less significance, which means that Dissolution and precipitation processes in the
they play a role only in extremely thin films and at oxide have also been speculated to create growth
low temperatures.58 stresses that can lead to cracking and by this to a
Another cause for growth stresses in oxide scales change in the oxidation kinetics. An example for
can be changes in the oxide composition.68 For exam- this is the mixed TiO2/Al2O3-oxide scale on inter-
ple, it has been observed that for various Fe–Cr alloys, metallic titanium aluminides, where the initially
the transition from initially formed (Cr, Fe)2O3 to a formed Al2O3 barrier in the scale becomes dissolved
scale with increasing Cr and decreasing Fe content, after a while and reprecipitated as discrete Al2O3
leads to the formation of tensile stresses in course of particles in a TiO2 matrix. This process seems to
time.69 The explanation for this effect was a decrease lead to precipitation stresses and microcracking in
in atomic volume associated with this transition. the scale after an incubation period, and initiates
Another example is the change from the transient the so-called breakaway effect, where the oxidation
alumina scales consisting of y-Al2O3, at the beginning rates increase significantly compared to the former
of oxidation to the later a-Al2O3, which, due to the protective period.76
different crystal structures, also induces stresses in the
oxide, which usually lower the compressive stress
level in the scale.50 Further effects can be present in 1.08.3 Thermally Induced Scale
the case of phase changes in the metal subsurface zone Stresses Resulting from Temperature
where for example, the consumption of aluminum by Changes
oxidation and, thus, a phase change from b-NiAl to
1.08.3.1 Experimental Observations
g-NiAl leads to a volume change.70 The situation of
stresses induced by selective internal oxidation is dis- The CTE of substrate material and oxide scale are
cussed by Hänse et al.71 also. Although such stresses usually different from each other and during temper-
are present in the metal subsurface zone they will ature changes these differences can lead to significant
have an effect on the oxide scale itself. Such internal stresses in the oxide/metal system. Table 3 shows a
oxides or oxide protrusions of the scale/metal inter- compilation of measured and theoretically calculated
face can introduce stresses not only by the growth of CTE data. This compilation should serve as orienta-
the oxide under isothermal conditions but also during tion as the CTEs are a function of temperature with,
cooling, leading to detrimental local stress concentra- in most cases, a more or less linear dependency where
tions.72 Another source of local differences in the higher CTE values are at the higher temperature.
(growth) stress levels can be multiphase substrates Table 3 reports the maximum values found in the
on which different types of surface oxides are formed, literature. For a more exact quantitative consideration
with locally different oxide growth rates.72 of the stress situation in oxide/metal systems during
Finally, it should be mentioned that the involve- temperature changes, it is recommended that the reader
ment of volatile species in the oxidation process can either refers to the original literature or measures the
also have a dramatic effect on the stress situation in CTE data in laboratory tests. As high temperature
the oxide scale. This is particularly known as the so- components have to be cooled down from their opera-
called active oxidation, which takes place in oxidizing tion temperature during inspection intervals or if they
halogen-containing environments.73–75 In this case, are subjected to thermocyclic operation, it is necessary
volatile metal halides, which are formed at the to regard the role of temperature induced stresses in the
oxide/metal interface by the reaction of halogen oxide/metal system in some depth. As general experi-
from the gas phase with the metal migrate in an ence shows, these stresses are the most critical for the
outward direction through the oxide scale and, with protective effect of the oxide scales and can reach values
increasing oxygen partial pressure, are deposited as as high as 3 GPa and even more.78
solid oxides within the scale. As a result, additional For thermally grown oxide scales, the tempera-
volume is created within the existing oxide scale, ture-induced stress cannot be measured directly, as
leading to high compressive stresses followed by the discussed in Section 1.08.1. After a temperature
cracking and opening up of the oxide scale for further change new stresses develop in the scale; however,
ingress of halogens to the metal. These very high these new thermal stresses are added on top of the
growth stresses are the reason why a protective effect preexisting intrinsic growth stresses, geometrically
of the oxide scale cannot be maintained under halo- induced growth stresses, and external stresses (see
genizing conditions. eqns [1]–[3]). Moreover, stress relaxation phenomena
Table 3 Orientational (‘maximum’) values of the coeffi-

cients of thermal expansion (CTE) for several technically 0.008
relevant substrates, coatings, and oxide phases
Out of plane strain

CTE a CTE a 0.006
(106 K) (106 K)
SiO2 (amorphous) 0.9 Incoloy 956 15.0 0.004

HPSN 3.0 738 15.3
Si 4.6 Alloy S, 713 15.4 0.002
SiC 4.9 NiPt39Al 15.6
AlN 6.2 617, Hast C-4 15.6
TiN 7.8 CMSX 6, IN 15.8 0.000
100
MoSi2 7.9 Hast C-276, 15.9 700 800 900 1000 1100 1200
230 Temperature (°C)
Cr2O3, SiO2 8.1 Hast C-22 16.1
Figure 6 A comparison of out-of-plane Al2O3
(cryst.)
residual scale strains after cooling as a function of
Al2O3 8.4 NiCrAlY 16.4
oxidation temperature, as determined by XRD (open
Ti 8.6 Hast X 16.5
circles and open squares) and Raman spectroscopy
Pt 8.8 600 16.6
(full circles and full squares). The materials
Y2O3 8.9 X-750 16.7
investigated were FeCrAl (circles) and FeCrAl + 0.1Zr
Cr 9.5 NiO, 556, 718 17.1
(squares).
IMI 834, NiAl 9.9 625 17.2
TiO2 10.5 Ni 17.5
Ti3Al (a2) 10.9 188 17.8
RuAl 11.0 Austenitic 17.0– was also verified by XRD measurements. At exposure
steels 19.0 temperatures above 1000 C the alloy without the
Pd 11.6 800H 18.3 reactive element Zr spalls, after cooling to RT, and the
Ni3Al 11.9 measured strains go to zero. The alloy with Zr does
NiCo17Cr14AlY 18.5
not spall after cooling and the scale strain remains
TiSi2 12.1 Fe3Al 19.5
TiAl 12.9 FeAl2 19.5 high. At temperatures below 1000 C there is no sig-
Ferritic steels 12.0–14.0 FeAl 21.8 nificant difference in the strain levels between the two
Co 13.9 Fe–28Al–2Cr 22.5 alloys. This aspect of the reactive element effect has
CoAl 14.0 Al 23.8 been experimentally confirmed for the reactive ele-
FeCrAlY 14.5 Fe–28Al– 24.0
ments Zr, Y, and Hf in a variety of FeCrAl, NiCrAl,
5Cr–0.1Zr
CoO 14.9 Fe2Al5 24.2 NiCoCrAl, and NiAl alloys.79–81 The transition tem-
perature (1000 C in Figure 6) above which the reac-
Reproduced from Schütze, M.; Malessa, M.; Rohr, V.; Weber, T. tive element effect is observed can differ from alloy
Surf. Coat. Technol. 2006, 201, 3872–3879. to alloy, for example on NiCoCrAl and even some
FeCrAl. The reactive element effect is observed at
temperatures above 1100 C.79–81 This is caused by
such as scale microcracking, detachment, and plastic the variation in thermal strain due to the different
deformation of the substrate also change the scale CTEs of the respective alloys as shown in Table 3.
stress state. The superposition of all these stress Generally speaking, the higher the alloy CTE the
generators and relaxors are sometimes called the lower the exposure temperature required to produce
‘residual’ scale stress, which is often measured at an influence of the reactive element effect.
room temperature (RT) after high temperature expo- The Raman technique has also been used to inves-
sure by Raman spectroscopy, optical fluorescence tigate the role that sample geometry plays in residual
spectroscopy, and X-ray diffraction (XRD).79–86 The stresses.82–84 These investigations show that the
measured RT residual stresses, even though they are higher compressive residual stresses are found in
not a direct measure of temperature-induced stress, the sample center and decrease as the sample edge
provide insight into the scale stress situation, which or corner is approached.82 The stress analysis per-
has been demonstrated for the reactive element effect formed by finite element modeling shows that the
79–81
and some sample geometry effects.82–84 stress component perpendicular to the sample sur-
An example for the case of the reactive element face is near zero at the sample center and becomes
effect is given in Figure 6 where the Raman technique increasingly more tensile as the sample edge or
corner is approached. This tensile stress, if large Oxide

enough, is expected to cause scale spallation at the Metal
edges and corners of samples.
Convex
1.08.3.2 Modeling = + = Tensile stress

= – = Compressive stress
A quantitative assessment of the level of stresses
induced by temperature changes can be performed Concave
using the following equation87:
Eox DT ðamet aox Þ Figure 7 Direction of the cooling stresses on curved
sox ¼ ½6 surfaces.1 Left part: amet > aox. Right part: amet < aox.
ðEox =Emet Þðdox =dmet Þð1 nmet Þ þ ð1 nox Þ
where Emet, Eox are the elastic moduli of the metal
and the oxide, respectively, nmet , nox are the Poisson perpendicular direction also become compressive.
ratios of the metal and oxide, respectively, dmet, dox The case where the CTE of the oxide is higher than
are the thickness of the metal substrate and the oxide that of the metal is shown in the right part of Figure 7.
scale, respectively, and amet, aox are the thermal The situation where the CTE of the oxide is higher
expansion coefficients of metal and oxide. The elastic than that of the substrate metal occurs only in some
moduli and, as already mentioned, in most cases also specific situations. One of these situations can be found
the CTE, are not independent of temperature, so in the unalloyed or low alloy steels where, due to the
eqn [6] is accurately applicable only for limited tem- change from the normal to the inverse spinel structure
perature intervals. For larger temperature changes in magnetite or chromium–iron spinels, a large varia-
sox has to be calculated by making a summation of tion of the value of the CTE takes place (see Figure 8).
the incremental values over small temperature inter- Such cases require a more sophisticated quantitative
vals. Equation [6] can be simplified if thin oxide treatment, which is described elsewhere.1,66 As a result,
scales on thick substrates are considered: a rather complex stress distribution in the multilayer
Eox DT ðamet aox Þ oxide scale can be observed, leading to a locally differ-
sox ¼ ½7 ent sensitivity to cracking or spalling of the oxide layer
1 nox
system.1
In the case of multilayer oxide scales, where the dif-
ferent layers have different CTE, the stress si of each
individual layer i of n layers can be calculated by88 1.08.4 The Effect of Stresses on the
Ei am ai Integrity of the Oxide Scale
si ¼ DT ½8
1 ni 1 ai DT
1.08.4.1 Measured Critical Strain Values
where am is the mean thermal expansion coefficient of to Scale Failure
the combined scale:
Since the protective effect of a dense surface oxide
P
n
scale guarantees the corrosion resistance of the tech-
di Ei ai =½ð1 ni Þð1 ai DT Þ
i nical high temperature materials under service con-
am ¼ P n ½9
ditions, one of the key issues is to maintain the
di Ei =½ð1 ni Þð1 ai DT Þ
i integrity of the scales or at least to know the mechan-
ical limits before failure of the scales occurs. This
Also for thermally induced stresses, the surface
failure can consist of various forms of damage
curvature may play a role for the sign and the level of
depending on the prevailing stress situation in the
the thermally induced stresses (Figure 7). In most
scale/metal composite system, the mechanical prop-
cases, the CTE of the oxide scale will be lower than
erties and structure of the scale, and the temperature
that of the metal substrate (left part of Figure 7) so that
and deformation rates:
on convex surfaces compressive stresses will be found
in the circumferential direction and tensile stresses 1. fracture of a scale or a component of the scale
perpendicular to the oxide/metal interface. On con- perpendicular to the direction of tensile stress,
cave surfaces, the circumferential stresses are compres- 2. fracture of the scale or a component of the scale by
sive and those at the oxide/metal interface in the the action of shear or buckling stresses in the scale,
30
Expansion coeff. a ⫻ 106 (⬚C)–1

(Fe, Cr)3O4
25
Fe3O4
Fe3O4
20
15 2¼ Cr/1Mo steel
Fe2O3
10
0 50 100 150 200 250 300 350 400 450 500 550 600
(a) Temperature T (⬚C)
0.2 (Fe, Cr)3O4

Fe3O4
0.1
Oxide strain eox (%)
Fe3O4
0
-0.1
-0.2 Fe2O3
Oxidation at 600 ⬚C
2¼Cr 1Mo steel
-0.3
0 50 100 150 200 250 300 350 400 450 500 550 600
(b) Cooling DT (⬚C)
100
50
Oxide stress sox (%)
0 Fe3O4
(Fe, Cr)3O4 Fe3O4
-50
-100 Fe3O3
-150
-200
Oxidation at 600 ⬚C
-250
2¼Cr 1Mo steel
-300
0 50 100 150 200 250 300 350 400 450 500 550 600
(c) Cooling DT (⬚C)
Figure 8 Role of differences in CTE values for the example of the oxide scale formed on a 2.25 Cr1Mo steel at 600 C in
air.1,66 (a) Variation with temperature of CTE,89,90 (b) strains in the oxide on cooling from 600 C, and (c) stresses in the
oxide on cooling from 600 C.
3. formation of microcracks in the scale,

4. disbonding of the scale from the metal or separa-
tion along the interfaces between the components
Preamplifier
of the scale, 4
40 dB
5. spalling of parts of the scale under compressive
stresses.
Before these failure situations are addressed by Amplifier
model considerations in Section 1.08.4.2, it will be 0---60 dB
helpful to look at experimental data on the strains to
scale failure.
The measurement of critical strains to scale failure Counter
requires some experimental effort. The most critical
5
part in these measurements is to record the moment
of scale failure either by optical or acoustic means.
x-t-Recorder
The most elegant technique has been proved to be
the acoustic emission analysis, which has been
applied to the measurement of scale failure in a
•
number of investigations.1 The simplest way would 2
1 eo
3
be to attach an acoustic waveguide to the specimen
by clamping or spot welding in order to lead the
signals from the hot zone of the test environment to
the colder part outside the test rig and a piezoelectric
1 Tensile test specimen
transducer. In equipment that is more sophisticated,
2 Corrosion test specimen
large size specimens that reach out into the colder 3 Specimen grips
part where the piezoelectric transducer can be 4 AE-transducer
attached directly.66 If the respective parameters are 5 AE-waveguide
known, especially the CTE, even the cooling step
after oxidation can be used to assess critical strains
for scale failure together with temperature and acous- Figure 9 Schematic of a test arrangement for measuring
tic signal recording.1 More accurate methods, how- critical strains to oxide scale failure at oxidation
ever, are based on tensile or compressive testing temperature in a furnace using a tensile test machine
equipped with an acoustic emission measurement device
machines in which simple uniaxial tension and com-
and a waveguide. Reproduced from Schütze, M. Oxid. Met.
pression or more sophisticated four point bending 1985, 24, 199–232.
tests can be performed. Both types of tests have been
used successfully and are described in detail in the
literature.1,91–93 An example of the test arrangement
for a tensile test is shown in Figure 9.7 confirm the brittleness of oxide scales. These values
The critical strain values to scale failure existing in can be used for model-based assessments of critical
the literature from such measurements have been sum- failure strains (see Section 1.08.4.2).
marized elsewhere92 and are shown in Table 4. It turns Besides the standard tensile or compression test, or
out that, with the exception of nickel oxide on nickel, the rather universal four point bending test, a number
under compressive strains where a specific stress relief of special tests for scale adhesion measurements
mechanism of simultaneous microcracking and crack have been developed. These have been described
healing in the scale (‘pseudoplasticity’) comes into elsewhere.1,106,107 Most of these techniques are only
play,93 almost all other values lie in the range between applicable at RT since they rely on the strength of an
0 and 1%. This confirms the brittleness of the pro- organic glue. Recently a new type of test has been
tective oxide scales indicating that oxides, even if they developed, which is called the ‘inverted blister test’.107
are grown as metal oxide scales on metallic substrates, In this case, the energy that is necessary to enlarge an
show ceramic nature. Also, the fracture toughness area of detached scale by increasing the hydrostatic
values which can be derived from these critical strain pressure in a complex arrangement through a central
values (see Section 1.08.4.2) as shown in Table 592 hole so that the substrate metal (which is a thin foil)
Table 4 Experimental data on scale failure strains given in the literature
Oxide Substrate Temperature ( C) «c (%) Mode Ref.
Al2O3 18CrAl steel 800 0.08–0.16 Tensile 7

Al2O3 FeCrAlY RT 0.4–1.3 Tensile 94
Al2O3 FeCr alloy RT 0.05 Tensile 95
Al2O3 MA 956 600 0.5–0.6 Tensile 96
Al2O3 FeCr alloy RT 1.4 Compressive 95
Cr2O3 AISI 310 RT 0.78–0.95 Tensile 97
Cr2O3 Alloy 800 800 0.09–0.50 Tensile 7
Cr2O3 Nimonic 75 700 0.002–0.005 Tensile 98
Cr2O3 AISI 310 RT (0.29–0.79) Compressive 97
Cr2O3 Alloy 800 RT (0.6–0.8) Compressive 1, 97
NiO Pure Ni 800 0.2–0.7 Tensile 1
NiO Pure Ni RT (0.9–2.2) Compressive 1
NiO Pure Ni RT (0.8–2.7) Compressive 93
NiO Pure Ni 400 (0.8–2.0) Compressive 93
FeO Low alloy steel 600–900 0.01–0.02 Tensile 89
Fe3O4 Mild steel RT 0.05–0.08 Tensile 89
Fe3O4 Pure iron RT 0.1–0.33 Tensile 89
Fe3O4 Pure iron RT 0.1–0.13 Tensile 89
Fe3O4 Pure iron 300 0.05–0.08 Tensile 89
Fe3O4 AISI 316 RT 0.11–0.20 Tensile 89
Fe3O4 AISI 316 800 0.17–0.31 Tensile 89
Fe3O4/Fe2O3 Mild steel RT 0.15–0.45 Tensile 100
Fe3O4/Fe2O3 Mild steel 550 0.05–0.42 Tensile 100
Fe3O4/Fe2O3 2.25Cr1Mo steel 600 0.1 Tensile 89
Fe3O4/Fe2O3 AISI 316 RT 0.13–0.22 Tensile 89
Fe3O4/Fe2O3 AISI 316 600 0.27–0.35 Tensile 89
Fe3O4/Fe2O3 Mild steel 500 0.01–0.02 Tensile 98
FeO/Fe3O4/Fe2O3 Mild steel 600 0.01–0.02 Tensile 98
Fe3O4/Fe2O3 Mild steel RT (0.18–0.55) Compressive 101
Fe2O3 2.25Cr1Mo steel 600–300 0.6 Compressive 66
TiO2/Al2O3 Ti50Al 900 0.15–0.50 Tensile 102
TiO2/Al2O3 Ti50Al2Nb 900 0.15–1.30 Tensile 102
TiO2/Al2O3 Ti47Al1Cr Si 900 0.6–1.7 Tensile 99
Reproduced from Schütze, M.; Ito, S.; Przybilla, W.; Echsler, H.; Bruns, C. Mater. High Temp. 2001, 18, 39–50.
deforms plastically (forming a growing blister at the a detrimental role on scale adhesion as it weakens
detached area) and separates from the oxide, is the interface and enhances interfacial pore for-
measured. Although this test looks intriguing, it is not mation.109–111 For example, in NiAl, increasing the
easy to use. interface sulfur content from zero to 2.5% reduced
Hou et al.108 contains a number of adhesion energy the interfacial strength by a factor of 10.
data with a scatter even within one metal/oxide
system of up to three orders of magnitude. This
1.08.4.2 Model Considerations
shows the difficulties of such types of measurements.
Much of the scatter will certainly also arise from the Because oxide scales are brittle phases even at high
presence and the level of impurities at the oxide/ temperatures, as is obvious from the experimental
metal interface. Sulfur especially is known to play data, a model description of scale failure is possible,
Table 5 Oxide fracture toughness data calculated from critical strain and defect data given in the literature
Oxide Substrate Temperature ( C) KIc (MPa m1/2) Mode Ref.
Al2O3 Bulk RT 3.0–6.0 Bending test 99

Al2O3 Bulk RT 2.5 Calculated from g0 103, 104
Al2O3 FeCrAl RT 0.4 Through scale 95
Al2O3 Kanthal APM 700 1.0 Through scale 105
Al2O3 18CrAl steel 800 0.3–1.0 Through scale 104
Al2O3 MA 956 1100 0.4–1.7 Through scale, initiation at 105
macro pores
Al2O3 MA 956 1100 0.6–2.1 Through scale, initiation at 105
macro pores
Cr2O3 Bulk RT 1.8 Calculated from g0 103, 104
Cr2O3 Nimonic 75 700 1.8–2.5 Through scale 98
Cr2O3 Alloy 800 800 1.2 Through scale 104
NiO Bulk RT 1.2 Calculated from g0 103, 104
NiO Nickel RT 0.5 Through scale 98
NiO Nickel RT 0.3–0.4 Through scale, initiation at 99
pores
NiO Nickel RT 0.75–1.3 Through scale, initiation at 99
microcracks
NiO Nickel RT 3.6 Scale detachment, parallel 99
to interface
NiO Nickel 800 2.6–6.0 Through scale 104
NiO Nickel 900 1.2 Through scale 98
FeO Bulk RT 0.9 Calculated from g0 103, 104
Fe3O4 Bulk RT 1.4 Calculated from g0 103, 104
Fe3O4 2.25Cr1Mo steel 600 0.2 Through scale 104
Fe3O4 Bulk RT 1.6 Calculated from g0 103, 104
FeO/Fe3O4 Pure iron 570 2.0 Through scale 98
FeO/Fe3O4 Pure iron 670 2.2–4.0 Through scale 98
FeCr2O4 Bulk RT 1.5 Calculated from g0 103, 104
Fe2MnO4 Bulk RT 0.8 Calculated from g0 103, 104
TiO2 Titanium RT 1.3 Through scale 103, 104
TiO2/Al2O3 Ti47Al1CrSi RT 0.2 Scale detachment, parallel 99
to interface
TiO2/Al2O3 Ti47Al1CrSi 700 1.2 Scale detachment, parallel 99
to interface
TiO2/Al2O3 Ti50Al 900 1.5–3.0 Through scale 102
TiO2/Al2O3 Ti50Al2Nb 900 1.0–3.3 Through scale 102
TiO2/Al2O3 Ti47Al1CrSi 900 2.7–7.0 Through scale 99
TiO2/Al2O3 Ti47Al1CrSi 900 1.4–4.0 Scale detachment, parallel 99
to interface
SiO2 Bulk RT 0.9 Calculated from g0 103, 104
MnO Bulk RT 0.9 Calculated from g0 103, 104
CoO Bulk RT 1.0 Calculated from g0 103, 104
Reproduced from Schütze, M.; Ito, S.; Przybilla, W.; Echsler, H.; Bruns, C. Mater. High Temp. 2001, 18, 39–50.
based on fracture mechanics approaches. Such E* which is the effective Young’s modulus (calculated
approaches always assume that so-called physical from the values of different phases if the crack passes
defects (pores, voids, micro cracks, etc.) are present along the phases112,113) and c, which characterizes the
in a structure and are responsible for its failure. From size of the physical defect:
this assumption, a critical stress sc can be calculated 1=2
from the critical surface energy Gc which is released E Gc
sc ¼ ½10
by the growth of the existing physical defect, pc
Assuming linear elastic behavior, the numerator in 2. Beginning of interfacial crack growth/delamina-
eqn [10] can be replaced by the fracture toughness Kc. tion (compressive)115
1=2
Kc ¼ Eox Gc ½11 KIc 1 þ dr ð1 þ nox Þ
ec ¼ pffiffiffiffiffi
i
½13
f pc 2Eox
which is valid for the plain stress situation that usually
predominates in thin scales. In the case of plain strain, 3. Buckling (compressive)116
the right hand side of eqn [11] must be multiplied
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
by the factor 1= ð1 n2 Þ. The fracture toughness 1:22 d 2
ebc ¼ ½14
Kc or KIc is a material property characterizing the 1 n2ox R
resistance of the oxide to crack growth. Crack growth
starts to occur once the stress intensity factor K at the 4. Crack deflection at scale buckles toward surface
tip of the defect in the scale reaches or exceeds the (compressive)10
value of this fracture toughness. Values of the fracture 2
d
toughness are given in Table 5. If the fracture tough- ec ¼ 3:6
bf
½15
ness and the size of the physical defects in the oxide R
scale or the scale/metal system are known, critical
strain values can be assessed by model equations 5. Spalling (compressive)116
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
at which damage to the scale occurs. The respective 2go
equations for the different damage situations are ec ¼
s
½16
dEox ð1 nÞ
summarized here and represented schematically in
Figure 10.
ec is the critical strain in the oxide (i.e., the strain
1. Through scale cracking (tensile): parallel to the oxide/metal interface) at which the
respective type of failure occurs. KIc is the mode I (i.e.,
KIc plain tensile straining) fracture toughness of the
etc ¼ pffiffiffiffiffi ½12
fEox pc plane in which failure starts, Eox is the Young’s
Tensile stresses
Defect 2c d Oxide
(a) Metal
Compressive stresses
Defect
d Oxide
Strong oxide
r
2c Weak interface
l Metal 2R
(b) Interfacial crack growth (c) Buckling
Weak oxide
Strong interface
(d) Crack deflection towards surface (e) Spalling
(f) Shear failure

Figure 10 Schematic of the different potential modes of scale failure. Reproduced from Schütze, M. In High Temperature
Corrosion of Advanced Materials and Protective Coatings; Saito, Y., et al. Ed.; North Holland: Amsterdam, 1992; 29.
modulus of the oxide, and 2c is the size of an A number of Young’s moduli and Poisson ratios
embedded physical defect (pore, void, microcrack, are given in Tables 6 and 7. Further values of the
etc.) which is responsible for failure. The size of a elastic constants, especially for elevated tempera-
surface defect would be c. f is a geometrical factor tures, can be found elsewhere.1
(for embedded defects f = 1) and d is the scale Figures 11 and 12 show how the model equations
thickness. The remaining parameters are the inter- can be used to calculate the dependency of the
facial roughness (amplitude r), the Poisson ratio nox, critical strain to failure on the respective scale para-
and the radius of an already existing delamination R. meters. Assuming a physical defect size of 1 mm or
Finally, g0 is the surface energy needed for the less, this results in critical strain values of a few
creation of two new surfaces by cracking or detach- tenths of a percent under tensile strain for Al2O3,
ment. This energy is an effective energy, which is Cr2O3, and Fe3O4 (see Figure 11). A comparison
influenced by the geometrical parameters of the with actually measured critical strain values in
failure plane in the following way103,117: Table 4 shows that this is very much in the same
order of magnitude and confirms that the model
0:1Eox r A0 Asep
g0 ¼ gi 1 þ ½17 approaches are suitable for a first assessment of the
2gi l A0
sensitivity to mechanical failure of such oxide scales.
where the additional parameter l represents the
‘wavelength’ of a rough interface, A0 the surface with- Table 6 Surface fracture energies, elastic moduli, and
out separations and Asep that part of the surface where fracture toughness values at room temperature (RT) (calcu-
delaminations/separations already exist (i.e., Asep is an lated from atomistic data)
equivalent to c). gi is the intrinsic surface energy,
Oxide gi (J m2) Eox (MPa) KIc (MPa m1/2)
which is a material’s parameter and, thus, similar to
KIc independent of the microgeometry: FeO 3.0 130 000 0.9
pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi CoO 3.0 156 000 1.0
KIc ¼ 2gi Eox ¼ Gc Eox ½18 NiO 3.6 191 000 1.2
MnO 2.7 148 000 0.9
Equation [16] actually corresponds to the original Fe3O4 4.5 208 000 1.4
equation of U.R. Evans.118 However, as shown by FeCr2O4 5.0 233 000 1.5
eqns [17] and [18] this rather simple equation should Fe2MnO4 3.0 108 000 0.8
Al2O3 7.7 419 000 2.5
be extended by micromechanics parameters.
Cr2O3 5.8 283 000 1.8
It should be mentioned here that for the size of Fe2O3 6.0 219 000 1.6
physical defects also, especially where size depends Y2O3 4.8 123 000 1.1
on the oxidation process, some kinetic data exist SiO2 4.4 85 600 0.9
in the literature, at least to a small extent.21,99,102
Reproduced from Robertson, J.; Manning, M. I. Mater. Sci.
Furthermore, these values can also be determined Technol. 1990, 6, 81–91.
from metallographic examinations of oxidized speci-
mens. The same is valid for the other geometrical
Table 7 Values of Poisson ratios nox at RT
parameters such as scale thickness and waviness of
the oxide/metal interface. Oxide v
Sometimes there is a criticism that the physical
Al2O3 0.24
defect sizes needed for scale failure, particularly for Cr2O3 0.29
the buckling situation, are much larger than is com- FeO 0.36
monly observed on scales in metallographic investi- Fe2O3 0.19
gations, that is, the scales fail before the appearance of Fe3O4 0.29
critical physical defect sizes. As the literature shows, NiO 0.32
SiO2 0.07
critical buckling constellations can grow from many TiO2 0.28
small cavities to the necessary size.119 Furthermore, CoO 0.36
small defects can interact with each other in a MnO 0.34
mechanical sense if they are lying close enough to FeCr2O4 0.31
each other so that a much larger ‘effective’ defect size MnFe2O4 0.39
Y2O3 0.38
determines the failure behavior of the scale.1 Rules
on how to calculate the effective physical defect size Reproduced from Robertson, J.; Manning, M. I. Mater. Sci.
are given elsewhere.1,120 Technol. 1990, 6, 81–91.
10-2 10
Al2O3
NiO
Cr2O3
TiO2 9
c = 1 μm
10-3
SiO2
e Cel Al2O3
Fe3O4
8
r = 2 μm
10-4
7
c g = 1 μm
6
10-5
−ec (10−4)
10-1 100 101 102 103 1
c (μm) c = 3 μm
5
Figure 11 Strain values to scale cracking ec under tensile
stress for different oxides as a function of the physical 0.5
defect size c calculated using eqn [12]. Reproduced from 0.3
c = 5 μm
Schütze, M. Oxid. Met. 1995, 44, 29–61. 4
0.2
0.1
Similar values are observed for the beginning of 3

crack growth (scale detachment) along the metal/
scale interface, as represented in Figure 12. These
2
values are slightly lower than those observed in
experimental investigations for global scale failure
under compression, which is, however, sensible, as 1
the beginning of crack growth or delamination along 1 2 3 4 5 6 7 8 9 10
d (μm)
the scale/metal interface is not yet the complete
failure by spalling or detachment of large parts of Figure 12 Strain values to the beginning of scale
the scale. Therefore, the situation in Figure 12 delamination ec under compressive stress for Al2O3 and a
physical defect size cg ¼ 1 mm at the scale/metal interface
represents rather a conservative approach to scale
calculated using eqn [13]. r denotes the ‘interfacial
failure under compressive stresses. See Schütze1,104 amplitude’ (see Figure 10(b)). Reproduced from Schütze,
for an in-depth discussion of the models available for M.; Ito, S.; Przybilla, W.; Echsler, H.; Bruns, C. Mater. High
the theoritical description of scale failure occurs. Temp. 2001, 18, 39–50.
Schütze1 also contains a discussion of the influence
of oxide creep, if existing, on the critical strains to
failure. It can, however, be stated here that this influ-
even more important to discuss the mechanism of
ence is limited and only becomes evident at very
oxide damage healing. It can be stated that all current
low defect or grain sizes of the oxide. For most
technical metallic alloys would not survive for a long
technical situations, the oxide scale will behave in a
time in operation, from a high temperature corrosion
brittle manner as described by the model equations
point of view, if healing of oxide scale damage did
shown, even at the operation temperature of a high
not take place. The reason is that in almost all cases
temperature component.
of high temperature, service strains or stresses of
the types described so far (be it external or internal
1.08.5 Maintaining the Protective stresses of the system) will arise and endanger the
Effect of Oxide Scales Even After integrity of the brittle oxide scale. What distinguishes
Mechanical Failure under Stresses a growing ‘ceramic’ oxide scale from a ‘dead’ artifi-
cially applied ceramic layer is that, during high
1.08.5.1 Healing of Cracks or other types of
temperature exposure, the oxidation process and,
Damage in the Protective Oxide Scale
thus, the growth of the oxide scale, continues. At the
As the discussion on the cracking behavior of oxide same time, this growth process represents the healing
scales shows that brittle failure is to be expected, it is potential of the oxide/metal system, which allows it
O2 O2 to protect itself even after intermediate mechanical

Further growth damage.
Me+ e– O2− e− The principles of the crack healing process are
Further growth illustrated schematically in Figure 13. As the diffu-
Metal Metal sion length at a crack is shorter than in the undam-
aged part of the oxide scale, this means that the local
The growth of gas-tight scales on metals in oxide growth rates will also be higher than in the rest
hot gases of the scale. With increasing time and continuously
O2 O2 increased thickening at these damaged spots, the over-
all scale becomes smooth again so that with time the
Me−+ Me−+ O2− − crack is closed. Figure 14 shows that this is not a
e e 2e
Me+, e− O2− e− theoretical consideration alone but that cracks can
Metal Metal
be closed almost completely by this healing process,
leaving only scars in the oxide surface, which are
Transport by crack formation in gas-tight scales hard to detect by optical means, (Figure 14(c)). As
observed in experiments, healing of oxide cracks
Figure 13 The two extreme cases of material transport
requires that a certain critical deformation rate of
through oxide scales (inward growth and outward growth)
and their effect on scale crack healing. Reproduced from the substrate is not exceeded.122 Otherwise the crack
Bruch, U.; Döhle, K.; Pütz, J.; Rahmel, A.; Schütze, M.; faces of the oxide scale would move apart at a higher
Schuhmacher, K. D. Proceedings of the International Congress rate than can be accommodated by the oxidation
on Metallic Corrosion; NRC, Ottawa, 1984; Vol. 3, 325–329. process from the faces of the scale crack. Figure 15
2 μm
(a)
A (b)
5 μm ε
5 μm
(c)
Figure 14 Examples of healed scale cracks in SEM surface investigations. (a) NiO scale on Ni99.2 after oxidation and
straining at 800 C, (b) oxide scale on Alloy 800H after oxidation and straining at 800 C, and (c) same scale as (b) but later
stage in the healing process.
Oxide
s s
e Metal e
ec
ec Corrosion-deformation
interaction
Oxide Oxide
Metal Metal Oxide
Metal
e e
t t
Figure 15 Schematic illustration of the role of a critical deformation rate e_ c of the substrate for scale crack healing.
Left: below E_c instant healing of the scale crack is possible and in the creep curve e–t no acceleration of creep due to corrosion
damage occurs. Right: above ec healing is impeded and internal corrosive attack at the unprotected site reduces the load
bearing metal cross-section ending up in accelerated creep.
illustrates this schematically, indicating that above of the protective scale-forming alloy takes place in
this critical strain rate, crack healing would be the metal subsurface zone directly underneath the
impeded and internal corrosion in particular along oxide scale. Evidently, in most technical cases, an
grain boundaries could facilitate crack formation in equilibrium situation is not achieved and the oxida-
the metal subsurface zone as observed by Wagner.123 tion process requires a faster supply of scale forming
At lower strain rates, the healing process would be able species from the substrate than can be achieved by
to close the cracks again, restoring the protective effect the alloy. As shown in Figure 16 this is at least true
of the scale. This situation has been modeled by super- for many of the technical alloys in their normal
imposing the oxidation process on the movement of operating temperature range (further examples are
crack phases by continued deformation and the details given elsewhere1,4). This may be different at very
of this model development as described by Barbehön high temperatures and for materials with high self-
et al.124 The model calculations provided quantitative diffusion coefficients of the scale forming elements,
values of the critical strain rates, which were in the where the level of this element is equal over the
range of 106 to 108 s1 and in good agreement with whole material thickness (i.e., no gradient), for exam-
the experimentally determined values.1,102 ple, FeCrAl alloys at 1200 C.127
The drop in the alloy concentration directly
underneath the oxide scale means that it is no longer
the original alloy that was in contact with the oxide
1.08.5.2 The Influence of Subsurface Zone
scale or the environment, if the oxide scale fails. In
Depletion on Maintaining the Protective
order to make this aspect clearer one should take a
Effect of Oxide Scales Under Stresses
look at Figure 17, which shows the parabolic rate
Contrary to what is required by the theoretical mod- constant as a function of the chromium content in
els of oxidation125 it is usual that significant depletion iron chromium steels at 1000 C.128 From this figure
P91 - synth.air + 4%H2O

9.0 650⬚C
Cr-concentration (mass%)
8.5
8.0
7.5 Thin scale

Thick scale
7.0
Cc = 7.3%
6.5
tB = 1000 h
6.0
(a) 0 2000 4000 6000 8000 10000
Time (h)
21
20 Matrix 30
19
18
17 A
20
Cr (%)
x (μm)
16
15
14
B 10
13
12 A = Cr concentration
B = scale thickness
11
0
0 100 200 300 400 500 600 700 800 900
(b) Time (h)
Figure 16 Cr depletion during oxidation as a function of oxidation time measured by EPMA directly underneath the
oxide scale in the metal subsurface zone. (a) P91 in air with 4% water vapor at 650 C,4 (b) alloy 800 in laboratory air at
800 C,126 A ¼ Cr concentration, B ¼ oxide scale thickness.
it becomes clear that steel, containing originally 20% thin chromium-rich oxide scale on the original sur-
of chromium which forms a more or less pure chro- face.129 As can be seen from Figure 16 and also
mia scale, would behave much worse after significant Schütze4 for this kind of material a steep drop in
depletion of chromium in the metal subsurface zone the chromium content takes place during oxidation.
if the scale should crack or spall. At these spots, a Generally, four different cases can be distin-
material with significantly lower chromium concen- guished with regard to the influence of depletion in
tration would be exposed to the oxidizing environ- the metal subsurface zone on the healing capability of
ment and, consequently, fast-growing iron-rich damaged oxide scales. This is summarized schemati-
oxides would grow at these spots strongly deteriorat- cally in Figure 19 for the example of chromia for-
ing the protective effect of the scale. For this, it is not mers. In case (a) subsurface zone depletion takes
necessary that the scale is cracked by external stres- place at a measurable rate but the values do not fall
ses; even intrinsic growth stresses or geometrically below the critical Cr-limit, which is necessary to
induced growth stresses can lead to scale damage form the protective scale. This situation guarantees
as shown in Figure 18, where nodule formation the long-term healing capability of the oxide/metal
of fast-growing iron-rich oxides has taken place at system under operation. In case (b) there is a steep
machining marks on an undeformed specimen of drop in the concentration of the protective scale
9% chromium steel, which protruded through the forming the alloy at the beginning of oxidation but
Fe2O3
Fe3O4
FeO
Fe
10-6
Fe2O3
Fe3O4
FeO
10-7 Fe – 2Cr Fe/Cr oxide
Fe2O3
(Fe, Cr)2O3
10-8
Fe – 9Cr
Fe3O4 Fe2O3
10-9
Cr2O3
Fe – 16Cr Fe Fe(2 – x)CrxO4
10-10
Cr2O3
Fe – 28Cr
10-11
0 10 20 30 40 50 60 70 80 90 100
Alloy chromium content (wt%)
Figure 17 Schematic showing the effect of Cr in Fe–Cr alloys on oxidation rate and oxide scale structure at
1000 C in oxygen. Reproduced from Whittle, D. P.; Wood, G. C.; Evans, D. J.; Scully, D. B. Acta Metall. 1967, 15,
1747.
rediffusion of this element from the metal interior by a number of authors and for different material
fills the reservoir directly underneath the oxide scale systems.130–138 Two of these models have been con-
again so that the critical limit is undercut only tem- verted into an open source code software tool.
porarily. If oxide scale cracking can be avoided in this The first is called COSIM and is available from
first period of exposure time, then the healing capa- NASA.139 The second is called ODIN140 and allows
bility would be restored and would last for the rest of the treatment of the situation of two-dimensional
the lifetime of the component. Case (c) represents the structures, which is of importance at edges and cor-
situation where rediffusion from the metal interior ners. As discussed in Section 1.08.2, these sites are
is not able to fill the alloying reservoir in the metal most endangered from the viewpoint of the stress
subsurface zone to the extent which allows protective situation in the oxide scale. A pragmatic diagram
healing of the oxide scale, and case (d) shows the even has been developed by Whittle using Wagner’s con-
worse situation where the composition of the subsur- siderations on the formation of a continuous protec-
face zone falls below the thermodynamic stability tive oxide scale,141 (Figure 20).142 Further details are
limit of the oxide scale in contact with the metal given elsewhere.1 The diagram in Figure 20 marks
substrate. In the latter case, therefore, even without four regions with increased healing potential. In
any mechanical damage to the oxide scale, the scale region (1), the chromium content is not sufficient to
would no longer be stable from a chemical point of establish a protective chromia scale. In region (2), a
view and would be converted into a nonprotective protective scale can be established but after the first
type of oxide. cracking or spalling the amount of chromium would
The situation of subsurface zone depletion and not be sufficient for protective healing. This region
rediffusion from the metal interior has been modeled would correspond to cases (c) and (d) in Figure 19.
0.6
0.5
4
0.4
NA,b
0.3 3
2
0.2
1
0.1
NiCr
FeCr
0
10-4 10-3 10-2 10-1
ki/(D)
15 μm
Figure 20 Critical values for the atom fraction NA,b
of the protective scale-forming element A as a function
Figure 18 Oxide nodule formation by faster growing of the ki/D.142 This figure is an example for chromia
Fe-rich Cr-(Fe)-oxide at machining marks on an undeformed formers. ki ¼ oxidation rate constant for the recession of
specimen of a 9% Cr steel after 850 h in isothermal the oxide/metal interface, D ¼ interdiffusion coefficient
oxidation at 650 C in air. Reproduced from Wulf, G. L.; of A in the alloy, also shown are the bandwidths of the
McGirr, M. B.; Wallwork, G. R. Corros. Sci. 1969, 9, 739. ki/D ratio for Fe–Cr and Ni–Cr alloys (see text for
explanation).
Region (3) shows the area in which healing can take

Original Cr-level place at least once after cracking or spalling and only
in the alloy region (4) guarantees repeated protective healing for
(a) longer operation times. The latter region would be
(b) equivalent to case (a) in Figure 19. Similar diagrams
Cr-limit for protective can of course be established also for alumina formers
scale formation
Cr-level
and help in the assessment of the choice of the right

material for long-term operation. All these models
can be used for a first theoretical assessment of the
(c) healing potential of materials. In most cases, very
reasonable results can be expected. There is still,
Thermodynamic
(d) instability of however, a need for a broader experimental basis to
0 protective scale
0 Oxidation time
verify these models.
Figure 19 Schematic of the four cases of subsurface

zone depletion of the protective scale forming element (in
this example Cr for chromia formers): (a) Drop of the alloy 1.08.6 Concluding Remarks
concentration at a slow rate. The concentration does not fall
below the Cr-limit for protective scale formation. (b)
Temporary drop below the protective limit but rediffusion
It is the aim of this chapter to illustrate the important
from the alloy interior ¼ temporary jeopardizing of the role of the mechanical aspects of oxidation behavior,
protective effect. (c) Final drop below the protective limit but particularly oxidation resistance of metallic materials
not below the thermodynamic stability limit of the oxide on at high temperatures. In a number of cases where the
the alloy; in case of scale damage the protective effect observed oxidation behavior could not be explained
would be finally lost. (d) Same situation as (c) but drop below
the thermodynamic stability limit ¼ ‘chemical destruction’ of
by the classical models of oxidation theory, the pres-
the protective scale without the need of superimposed ence of stresses in the oxide/metal system was the
stresses. reason for the discrepancies. As shown in the
relatively concise treatment of this topic in this chap- 19. Zhao, J. G.; Huntz, A. M. J. Mater. Sci. 1984, 19, 3166.
20. Huntz, A. M.; Calvarin Amiri, G.; Evans, H. E.;
ter, a number of well-elaborated models describing Cailletaud, G. Oxid. Met. 2002, 57, 499.
the role of stresses, and also the healing of scale 21. Przybilla, W.; Schütze, M. Oxid. Met. 2002, 58, 103–145.
damage and subsurface zone depletion for the pro- 22. Przybilla, W.; Schütze, M. Oxid. Met. 2002, 58, 337–359.
23. Tylecote, R. F. Mem. Sci. Rev. Metall. 1965, 62, 241.
tective effect of the oxide scales, exist. Furthermore, a 24. Jaenicke, W.; Leistikow, S. Z. Phys. Chem. 1958, 15,
limited number of experimental data are available to 175.
characterize at least the more common situations, but 25. Jaenicke, W.; Leistikow, S.; Städler, A. J. Electrochem.
Soc. 1964, 111, 1031.
at this point more work is needed. The model basis is 26. Engell, H.; Wever, F. Acta Metall. 1957, 5, 695.
now sufficiently broad to develop comprehensive 27. Buresch, F. E.; Bollenrath, F. J. Nucl. Mater. 1967, 24,
computer codes regarding the complex synergistic 270.
28. Cathcart, J. V.; Liu, C. T. Oxid. Met. 1973, 6, 123.
interplay of the different chemical and mechanical 29. Roy, C.; Burgess, B. Oxid. Met. 1970, 2, 235.
mechanisms. This has indeed been started141 and may 30. Donaldson, A. T.; Evans, H. E. J. Nucl. Mater. 1981,
in the near future allow a better quantification of the 99, 38.
31. Kang, K. J.; Hutchinson, J. W.; Evans, A. G. Acta Mater.
role of stress effects in oxidation, especially under 2003, 51, 1283–1291.
industrial conditions. At the same time, the basis of 32. Zhao, J. G.; Zhang, Y. T. Mater. Sci. Eng. A 1989, 120,
experimentally determined data that can be fed into 245.
33. Huntz, A. M.; Zhao, J. G.; Boumaza, A.; Moulin, G. Mater.
computer calculations has to be broadened. This Sci. Technol. 1988, 4, 470.
chapter provides a general understanding of this 34. Vosberg, V. R.; Fischer, W.; Berger, M. G.; Clemens, D.;
rather complex situation in which stresses and oxida- Quadakkers, W. J.; Nickel, H. Fresenius J. Anal. Chem.
1996, 355, 745–747.
tion interact with each other, leading sometimes to 35. Zouboulis, E.; Renusch, D.; Grimsditch, M. Appl. Phys.
unexpected oxidation behavior. For a more detailed Lett. 1998, 72, 1.
treatment of the topic the reader is referred to the 36. Birnie, J.; Graggs, C.; Gardiner, D. J.; Graves, P. R.
Corros. Sci. 1992, 33, 1.
textbook literature.1 37. Tolpygo, V. K.; Clarke, D. R. Mater. High Temp. 2000, 17,
59–70.
38. Mougin, J.; Rosman, N.; Lucazeau, G.; Galerie, A. J.
Raman Spectrosc. 2001, 32, 739–744.
References 39. Wouters, Y.; Galerie, A.; Bouvier, P.; Mermoux, M.;
Petit, J. P. Microscopy of Oxidation; Science Reviews:
1. Schütze, M. Protective Oxide Scales and Their Strixton, UK, 2005; Vol. 6, pp 145.
Breakdown; John Wiley: Chichester, 1997. 40. Lawless, K. R.; Gwathmey, A. T. Acta Metall. 1956, 4,
2. Kofstad, P. High Temperature Corrosion; Elsevier 153.
Applied Science: London, 1988. 41. Pyun, Y. J.; Homma, T.; Takakuwa, K. J. Jpn. Int. Metall.
3. Evans, H. E.; Donaldson, A. T.; Gilmour, T. C. Oxid. Met. 1983, 47, 663.
1999, 52, 379–402. 42. Liu, C.; Lebrun, J. L.; Huntz, A. M.; Sibieude, F. Z.
4. Schütze, M.; Renusch, D.; Schorr, M. Mater. High Temp. Metallk. 1993, 84, 140.
2005, 22, 113–120. 43. Goedjen, J. G.; Stout, J. H.; Guo, Q.; Shores, D. A. Mater.
5. Donchev, A.; Fietzek, H.; Kolarik, V.; Renusch, D.; Sci. Eng. A 1994, 177, 115.
Schütze, M. Mater. High Temp. 2005, 22, 139–146. 44. Sarioglu, C.; Schumann, E.; Blachere, J. R.; Pettit, F. S.;
6. Isecke, B.; Schütze, M.; Strehblow, H. H. In Springer Meier, G. H. Mater. High Temp. 2000, 17, 109–115.
Handbook of Materials Measurement Methods; 45. Corkovic, S.; Pyzalla, A. R. Mater. Corr. 2004, 55, 341.
Czichos, H., Saito, T., Smith, L., Eds.; Springer Verlag: 46. Tortorelli, P. F.; More, K. L.; Specht, E. D.; Pint, B. A.;
Berlin, 2006: Chapter 12: Corrosion. Zschack, P. Mater. High Temp. 2003, 20, 303–309.
7. Schütze, M. Oxid. Met. 1985, 24, 199–232. 47. Calvarin-Amiri, G.; Huntz, A. M. R. Molins Mater. High
8. Kofstad, P.; Lillerud, K. P. J. Electrochem. Soc. 1980, Temp. 2001, 18, 91–99.
127, 240. 48. Huntz, A. M.; Lefeuvre, A.; Andrieux, M.; Severac, C.;
9. Lillerud, K. P.; Kofstad, P. J. Electrochem. Soc. 1980, Moulin, G.; Molins, R.; Jomard, F. Microscopy of
127, 2397. Oxidation; Science Reviews: Strixton, UK, 2003; Vol. 5,
10. Evans, A. G.; Crumley, G. B.; Demaray, R. E. Oxid. Met. pp 393.
1983, 20, 193. 49. Gnäupel-Herold, T.; Reimers, W. In Proceedings of the
11. Caplan, D.; Sproule, G. I. Oxid. Met. 1975, 9, 459. International Conference on Residual Stresses;
12. Ecer, G. M.; Meier, G. H. Oxid. Met. 1979, 13, 119. Frisson, T., Odin, M., Anderson, A., Eds.; Linköping
13. Golightly, F. A.; Stott, F. H.; Wood, G. C. Oxid. Met. 1976, University: Linköping, Sweden, 1997.
10, 163. 50. Hou, P. Y.; Paulikas, A. P.; Veal, B. W. Microscopy of
14. Golightly, F. A.; Wood, G. C.; Stott, F. H. Oxid. Met. 1978, Oxidation; Science Reviews: Strixton, UK, 2005; Vol. 6,
14, 217. pp 373.
15. Miner, R. G.; Nagarajan, V. Oxid. Met. 1981, 16, 313. 51. Echsler, H.; Alija Martinez, E.; Singheiser, L.;
16. Pawel, R. E.; Cathcart, J. V.; Campbell, J. J. Quadakkers, W. J. Mater. Sci. Eng. A 2004, 384, 1–11.
J. Electrochem. Soc. 1963, 110, 551. 52. Mennicke, C.; Clarke, D. R.; Rühle, M. Oxid. Met. 2001,
17. Stringer, J. Corros. Sci. 1970, 10, 513. 55, 551–569.
18. Delaunay, D.; Huntz, A. M.; Lacombe, P. Corros. Sci. 53. Homma, T.; Pyun, P. J. High Temperature Corrosion;
1980, 20, 1109. Jap. Int. Metals: Tokyo, 1983; 161.
54. Liu, C.; Huntz, A. M.; Lebrun, J. Mater. Sci. Eng. A 1993, 87. Tien, J. K.; Davidson, J. M. In Stress Effects and the
160, 113. Oxidation of Metals Proceedings; Cathcart, J. V., Ed.;
55. Galerie, A.; Toscan, F.; Dupeux, M.; Mougin, J.; TMS-AIME: New York, 1975; p 200.
Lucazeau, G.; Valot, C.; Huntz, A. M.; Antoni, L. Mater. 88. Metcalfe, E.; Manning, M. I. The spalling of steam grown
Res. 2004, 7, 83–88. oxide from austenitic and ferritic alloys. Report RD/L/R/
56. Dionnet, B. PhD Thesis, University of Limoges, 1993. 1966 Central Electricity Generating Board, CERL,
57. Pilling, N. B.; Bedworth, R. E. J. Inst. Metals 1923, 29, 529. Leatherhead, 1977.
58. Hancock, P.; Hurst, R. C. Adv. Corros. Sci. Technol. 89. Armitt, J.; Holmes, D. R.; Manning, M. I.; Meadowcroft, D.
1974, 4, 1. B.; Metcalfe, E. The spalling of steam grown oxide from
59. Weast, R. C. Physical Constants of Inorganic superheater and reheater tube steels. Report FP686 EPRI,
Compounds Handbook of Chemistry and Physics, 6th Palo Alto, 1978.
ed.; CRC, 1984; B68, B161. 90. Manning, M. I.; Metcalfe, E. Steamside spalling from
60. Bernstein, H. L. Metall. Trans. 1987, A18, 975. Type 316 superheater and reheater tubes. Report RD/L/
61. Stott, F. H.; Atkinson, A. Mater. High Temp. 1994, 12, N15/75 Central Electricity Generating Board, CERL,
195–207. Leatherhead, 1975.
62. Chevalier, S.; Strehl, G.; Favergeon, J.; Desserrey, F.; 91. Saunders, S. R. J.; Nagl, M. M.; Schütze, M. Mater. High
Weber, S.; Heintz, O.; Borchardt, G.; Larpin, J. P. Mater. Temp. 1994, 12, 103–109.
High Temp. 2003, 20, 253–259. 92. Schütze, M.; Ito, S.; Przybilla, W.; Echsler, H.; Bruns, C.
63. Clarke, D. R. Acta Mater. 2003, 51, 1393–1407. Mater. High Temp. 2001, 18, 39–50.
64. Krishnamurthy, R.; Srolovitz, D. J. Acta Mater. 2003, 51, 93. Küppenbender, I.; Schütze, M. Oxid. Met. 1994, 42,
2171–2190. 109–144.
65. Limarga, A.; Wilkinson, D. S.; Weatherly, G. L. Scripta 94. Golightly, F. A.; Stott, F. H.; Wood, G. C. Werkstoffe Korr.
Mater. 2004, 50, 1475–1479. 1979, 30, 487–491.
66. Christl, W.; Rahmel, A.; Schütze, M. Oxid. Met. 1989, 31, 1. 95. Banks, J. P.; Gohil, D. D.; Evans, H. E.; Hall, D. J.;
67. Taniguchi, S. Trans IDIJ 1985, 25, 3. Saunders, S. R. J. In Materials for Advanced Power
68. Appleby, W. K.; Tylecote, R. F. Corros. Sci. 1970, 10, 325. Engineering; Contsouradis, D., et al. Ed.; Kluwer
69. Howes, V. R.; Richardson, C. N. Corros. Sci. 1969, 9, Academic: Dordrecht, 1994; pp 1543–1552.
385. 96. Guttmann, V.; Mediavilla, A.; Ruano, O. Mater. High
70. Singheiser, L. Forschungszentrum Jülich, Private Temp. 1993, 11, 42–50.
communication. 97. Nagl, M. M.; Saunders, S. R. J.; Guttmann, V. Mater.
71. Hänsel, M.; Boddington, C. A.; Young, D. J. Corros. Sci. High Temp. 1994, 12, 163–168.
2003, 45, 967–981. 98. Hancock, P.; Nicholls, J. R. Mater. High Temp. 1994, 12,
72. Quadakkers, W. J.; Tyagi, A. K.; Clemens, D.; Anton, R.; 209–218.
Singheiser, L. In Elevated Temperature Coatings: Science 99. Bruns, C.; Schütze, M. Oxid. Met. 2001, 55, 35–68.
and Technology III; Hampikian, J. M., Dahotre, N. B., Eds.; 100. Nagl, M. M.; Evans, W. T.; Hall, D. J.; Saunders, S. R. J.
The Minerals, Metals and Materials Society: Warrendale, J. Phys. III 1993, 3, 933–941.
1999; pp 119–138. 101. Nagl, M. M.; Evans, W. T.; Saunders, S. R. J.; Hall, D. J.
73. Lee, Y. Y.; McNallan, M. J. Metall. Trans. 1987, 18A, Mater. Sci. Technol. 1992, 8, 1043–1049.
1099. 102. Schmitz-Niederau, M.; Schütze, M. Oxid. Met. 1999, 52,
74. Bramhoff, D.; Grabke, H. J.; Schmidt, H. P. Werkst. Korr. 241–276.
1989, 40, 642. 103. Robertson, J.; Manning, M. I. Mater. Sci. Technol. 1990,
75. Schwalm, C.; Schütze, M. Mater. Corros. 2000, 51, 34, 6, 81–91.
73, 161. 104. Schütze, M. Oxid. Met. 1995, 44, 29–61.
76. Schmitz-Niederau, M.; Schütze, M. Oxid. Met. 1999, 52, 105. Hukelmann, F. Doctoral Thesis, Technical University of
225–240. Clausthal, 2000.
77. Schütze, M.; Malessa, M.; Rohr, V.; Weber, T. Surf. Coat. 106. Hou, P. Y.; Atkinson, A. Mater. High Temp. 1994, 12,
Technol. 2006, 201, 3872–3879. 119–125.
78. Huntz, A. M.; Schütze, M. Mater. High Temp. 1994, 12, 107. Mougin, J.; Dupeux, M.; Galerie, A.; Antoni, L. Mater. Sci.
151–161. Technol. 2002, 18, 1217–1220.
79. Renusch, D.; Grimsditch, M.; Koshelev, I.; Veal, B. W. 108. Hou, P. Y.; Saunders, S. R. J. Mater. High Temp. 2005,
Oxid. Met. 1997, 48, 471. 22, 121–129.
80. Uran, S.; Grimsditch, M.; Veal, B.; Paulikas, P. Oxid. Met. 109. Hou, P. Y.; Priimak, K. Oxid. Met. 2005, 63, 113–129.
2001, 56, 551–569. 110. Funkenbusch, A. W.; Smeggil, J. G.; Bornstein, N. S.
81. Koshelev, I. K.; Paulikas, A. P.; Uran, S.; Beno, M. B.; Metall. Trans. 1985, 16A, 1164.
Jennings, G.; Linton, J.; Veal, B. W. Oxid. Met. 2003, 59, 111. Smialek, J. L. Metall. Trans. 1991, 22A, 739.
469–481. 112. Suga, T. Doctoral Thesis University of Stuttgart 1983.
82. Renusch, D.; Muralidharan, G.; Uran, S.; Grimsditch, M.; 113. Suga, T.; Kvernes, I.; Elssner, G. Z. Werkstofftechn. 1984,
Veal, B. W.; Wright, J. K.; Williamson, R. L. Oxid. Met. 15, 371.
2000, 53, 171. 114. Schütze, M. In High Temperature Corrosion of Advanced
83. Uran, S.; Veal, B. W.; Grismditch, M.; Pearson, J.; Materials and Protective Coatings; Saito, Y., et al. Ed.;
Berger, A. Oxid. Met. 2000, 54, 73–85. North Holland: Amsterdam, 1992; 29.
84. Wright, J. K.; Williamson, R. L.; Renusch, D.; 115. Evans, H. E.; Mitchell, G. P.; Lobb, R. C.; Owen, D. R. J.
Grimsditch, M.; Veal, B. W.; Hou, P. Y.; Cannon, R. M. Proc. R. Soc. London A. 1993, 440, 1.
Mater. Sci. Eng. A 1999, 262, 246. 116. Evans, A. G.; Hutchinson, J. W. Acta Metall. 1989, 37, 909.
85. Lipkin, D. M.; Clarke, D. R.; Hollatz, M.; Bobeth, M.; 117. Evans, U. R. An Introduction to Metallic Corrosion; Arnold
Pompe, W. Corros. Sci. 1997, 39, 231. Publishers: London, 1948.
86. Tolpygo, V. K.; Dryden, J. R.; Clarke, D. R. Acta Mater. 118. Zimmermann, D.; Tolpygo, V. K.; Rühle, M.; Clarke, D. R.
1998, 46, 927. Z. Metallk. 2003, 94, 157–162.
119. Hancock, P.; Nicholls, J. R. Mater. Sci. Technol. 1988, 4, 132. Wright, I. G.; Pint, B. A.; Hall, L. M.; Tortorelli, P. F. In
398. Lifetime Modelling of High Temperature Corrosion
120. Schütze, M. Oxid. Met. 1986, 25, 409. Processes; Schütze, M., Quadakkers, W. J., Nicholls, J.
121. Bruch, U.; Döhle, K.; Pütz, J.; Rahmel, A.; Schütze, M.; R., Eds.; EFC-Publication No. 34: Maney, Leeds, 2001;
Schuhmacher, K. D. Proceedings of the International p 339–358.
Congress on Metallic Corrosion; NRC, Ottawa, 1984; 133. Nesbitt, J. A. In Lifetime Modelling of High Temperature
Vol. 3, 325–329. Corrosion Processes; Schütze, M., Quadakkers, W. J.,
122. Schütze, M. Werkstoffe Korr. 1987, 38, 597. Nicholls, J. R., Eds.; EFC-Publication No. 34: Maney,
123. Wagner, C. Z. Elektrochemie 1959, 63, 772. Leeds, 2001; pp 359–378.
124. Barbehön, J.; Rahmel, A.; Schütze, M. In High 134. Evans, H. E.; Taylor, M. E. Oxid. Met. 2001, 55, 17–34.
Temperature Alloys for Gas Turbines and Other 135. Pragnell, W. M.; Evans, H. E. Oxid. Met. 2006, 66,
Applications; Betz, W., et al. Ed.; D. Reidel Publishers: 209–230.
Dordrecht, 1986; 1267. 136. Renusch, D.; Echsler, H.; Schütze, M. Mater. Sci. Forum
125. Quadakkers, W. J.; Bongartz, K. Mater. Corr. 1994, 45, 2004, 461–464, 729–736.
232–241. 137. COSIM – A finite difference computer model to
126. Wright, I. G. In Metals Handbook 9th ed., ASM predict tenary concentration profiles associated
International: Metals Park, 1987; Vol. 13, Corrosion, 97. with oxidation and interdiffusion of overlay coated
127. Barbehön, J.; Rahmel, A.; Schütze, M. Oxid. Met. 1988, substrates. Funding No. WU-714-04-20-00. Contact:
30, 85. J.A. Nesbitt, NASA John Glenn Research Centre,
128. Whittle, D. P.; Wood, G. C.; Evans, D. J.; Scully, D. B. Cleveland.
Acta Metall. 1967, 15, 1747. 138. Pragnell, W. M. PhD thesis, University of Birmingham,
129. Wulf, G. L.; McGirr, M. B.; Wallwork, G. R. Corros. Sci. 2005.
1969, 9, 739. 139. Wagner, C. J. Electrochem. Soc. 1952, 99, 369.
130. Evans, H. E.; Lobb, R. C. Corros. Sci. 1984, 24, 223. 140. Whittle, D. P. Oxid. Met. 1972, 4, 171.
131. Nicholls, J. R., et al. In Lifetime Modelling of High 141. Schütze, M.; Malessa, M.; Renusch, D.; Tortorelli, P. F.;
Temperature Corrosion Processes; Schütze, M., Wright, I. G.; Dooley, R. B. Mater. Sci. Forum 2006,
Quadakkers, W. J., Nicholls, J. R., Eds.; EFC-Publication 522–523, 393–400.
No. 34: Maney, Leeds, 2001. 142. Rhines, F. N.; Wolf, J. S. . Metall. Trans. 1970, 1, 1701.
1.09 Thermodynamics and Theory of External and Internal
Oxidation of Alloys
B. Gleeson
Department of Mechanical Engineering and Materials Science, The University of Pittsburgh, 647 Benedum Hall, Pittsburgh,
PA 15261, USA

1.09.2 Pure Metal Reactions 181
1.09.2.1 Thermodynamics of a Single-Oxidant Reaction 181
1.09.2.2 Thermodynamics of Dual Oxidant Reactions 184
1.09.2.3 Kinetics of Scale Formation 186
1.09.2.3.1 Parabolic rate law 186
1.09.2.3.2 Linear rate law 187
1.09.2.3.3 Logarithmic rate law 187
1.09.2.4 Transport Properties of Metal Oxides 187
1.09.2.5 Wagner’s Theory of Metal Oxidation 188
1.09.3 Alloy Reactions 190
1.09.3.1 Thermodynamics of Alloy Oxidation 190
1.09.3.2 Criterion for the Sustained Exclusive Growth of a Protective Scale 191
1.09.3.3 Internal Oxidation 192
1.09.3.4 Transition from Internal Oxidation to External Scale Formation 193
1.09.4 Epilogue 193
References 193
1.09.1 Introduction
Symbols
ai Chemical activity of phase or component i
High temperature corrosion plays an important role
Di Diffusion coefficient of component i (m2 s1)
in the selection of materials in modern indus-
K Equilibrium constant for a given reaction
try. Numerous commercial processes such as electric
kl Linear rate constant (m s1) or (kg m2 s1)
power generating plants, aerospace, gas turbines,
kp Parabolic rate constant (m2 s1) or (kg2 m4 s1)
heat-treating, and mineral and metallurgical proces-
Ni Mole fraction of component i
sing operate at temperatures exceeding 500 C.1 Oxi-
Pi Partial pressure of gaseous species i
dation is often the most important high temperature
R Universal gas constant (8.314 J mol1 K1)
corrosion reaction in these commercial processes.
T Temperature (K)
Indeed, most high temperature alloys are designed to
t Time (s)
react with the oxidizing environment in such a way
Vm Molar volume of metal or alloy (m3 mol1)
that a protective oxide scale forms.2 The degradation
X Thickness of metal consumed due to scaling (m)
resistance of a high temperature alloy depends on
x Scale thickness (m)
sustaining the formation of this protective scale.
DG Gibbs free enthalpy in the standard state
The properties of the scale determine the extent to
(J mol1)
which protection can be provided. Ideally, the scale
DW Weight change (kg m2)
should exhibit a slow growth rate, good adherence to
gi Chemical activity coefficient of solute i
the alloy substrate, a high stability, and be continuous
n Stoichiometric factor for the oxide BOn
and free of defects such as microcracks or large
180
Thermodynamics and Theory of External and Internal Oxidation of Alloys 181
voids. In general, a chromia, alumina, or silica scale can 1.09.2.1 Thermodynamics of a

meet these requirements for high temperature oxida- Single-Oxidant Reaction
tion resistance, with chromia-forming alloys being the
Consider the high temperature reaction of a metal,
most extensively used in high temperature industrial
M, with an oxidant gas, in this case oxygen. The metal
applications.
initially absorbs oxygen and then chemical reaction
Almost all theoretical treatments of high
ensues to form an oxide. For the metals of relevance
temperature corrosion assume local equilibrium. In
to this review, the resulting oxide is solid. Thus, the
the majority of cases, this proves to be a reasonable
oxide first nucleates and then grows to form a scale on
assumption, although it is usually insufficient for pre-
the metal surface. Depending on its growth kinetics, the
dicting the nature and phase constitution of the reaction
scale may or may not protect the underlying metal.
product. This is because reaction kinetics dictate the
The formation of an oxide may be generally
pathway for scale development. The following discusses
described by the reaction
key fundamental aspects associated with the thermo-
dynamics and kinetics of high temperature corrosion. 2x 2
M þ O 2 ¼ Mx O y ½1
From a thermodynamic standpoint, it will be shown that y y
relative partial pressures of the gaseous components in Presented in this manner, the high temperature oxida-
the reacting atmosphere are important for predicting tion of metals may seem to be among the simplest of
the composition of the product scale. From a kinetic reactions; however, the reaction path and behavior often
standpoint, it will be shown that limiting equations can involve a number of phenomena and processes that
be established for predicting both scaling kinetics and depend on a variety of factors.3,4 In fact, at the same
critical concentrations in the alloy for transitioning to temperature for a given gas–metal reaction, one can
and sustaining protective scaling behavior. observe drastically different rates of reaction (linear,
parabolic, or any other rate) when the gaseous composi-
tion of the atmosphere is altered and the relative partial
pressures of the gaseous components are different.5
1.09.2 Pure Metal Reactions Under equilibrium conditions, the law of mass
action for reaction [1] gives6
In the context of this discussion, the metal and gas 2=y
react to form a solid surface scale. The prototypical gas aMx Oy
K1 ¼ ½2
will be oxygen; however, it should be realized that the aM2x=y aO2
fundamental treatments equally apply to other gases, where K1 is the temperature-dependent equilibrium
such as sulfur and nitrogen. Moreover, the oxidant may constant and ai is the chemical activity of species i. In
be simple, like O2, S2, and N2, or it may be more complex, most cases, the solids (metal and oxide) are assumed to
like H2O, H2S, and NH4. For the sake of clarity, the be in their pure standard state, so that their activities are
following discussion will consider oxide-scale formation. defined as unity. At relatively high temperatures and
An oxide scale may be a single layer or it may be moderate pressures, the oxidant gas can be treated as
comprised of two or more layers of varying compositions being ideal; that is, the activity of oxygen can be approxi-
that depend on the temperature and oxidizing condi- mated by its partial pressure in atmospheres. Thus, eqn
tions. There are two important factors in discussing the [2] simplifies to
oxidation of metals: thermodynamics and kinetics.
Metallic elements react with oxygen to form oxides if it K1 ¼ 1=PO2 ½3
is energetically feasible. Thermodynamics show whether where PO2 is the oxygen partial pressure.
or not a reaction can take place. When the oxidation Thermodynamically, reaction [1] for any metal
reaction is possible, kinetics shows how fast the reaction can take place spontaneously from left to right
will be. In practical applications, kinetics is of more when its overall Gibbs free energy change, DG, is
importance because it determines the extent of metal negative. For reaction [1], the Gibbs free energy
consumption and the overall reaction pathway (i.e., change under isobaric conditions is given as:
assemblage and structure of the reaction product). Oxi-
DG ¼ DG þ RT lnK1 ½4
dation theory of pure metals provides the foundation

for understanding the more complicated processes where DG is the standard Gibbs free energy
associated with alloy oxidation. of formation of the oxide at absolute temperature
T and R is the gas constant. If DG ¼ 0, the system is at usually called an Ellingham diagram,6 summarizes the
equilibrium, and if DG > 0, the reaction is thermody- temperature dependence of DG for various common
namically unfavorable. At equilibrium (DG ¼ 0), oxidation reactions at a standard state (pO2 ¼ 1 atm).
Such a diagram is shown in Figure 1 for oxides.
DG ¼ RT ln K1 ¼ RT ln PO2 ½5
The Ellingham diagram in Figure 1 shows the
Thus, knowing that at equilibrium both the forward relative thermodynamic stability of the indicated oxi-
and reverse reaction rates are equal, the dissociation des. The lower the line on the diagram, the more
pressure of the oxide can be defined from [5] as: negative the standard free energy of formation and,
hence, the more stable the oxide. For example, the
DG
PO2 ¼ exp
diss
½6 lines for Al2O3, SiO2, and Cr2O3 are lower than those
RT for FeO, NiO, and CoO in the Ellingham diagram, so
The metal M can only be oxidized to the oxide MxOy the former oxides are more stable and therefore meet
at the temperature T if the ambient partial pressure an important criterion for being protective scales. It is
of oxygen is larger than the dissociation pressure further seen in Figure 1 that the DG versus T lines for
defined by eqn [6]. most of the oxides are parallel and positively sloped.
For the usual conditions of constant temperature This is a consequence of the fact that the entropy of a
and pressure, the auxiliary function DG is usually gas is much larger than that of a solid. Thus, for the
described by the simple relation metal oxidation reactions represented by eqn [1],
DG ¼ DH T DS ½7 2 2x
DS ¼ SM x Oy
SM SO 2 SO 2 : ½8

y y
where DH is the enthalpy of reaction and DS is the
entropy change under standard-state conditions. Tabu- It follows from eqn [7] that qDG =qT ¼ DS SO 2 ,
lated values of DH and DS for the determination of which is greater than zero.
DG at any given temperature are readily available.7–9 The PO2 can be read directly from Figure 1 by
A small selection of useful values is given in Table 1. using the PO2 scale along the bottom and the right
Equation [7] shows that a plot of DG versus T side of the diagram. A straight line drawn from the
gives a straight line. This line would change in slope index point labeled ‘O’ (at DG ¼ 0, T ¼ 0 K) at the
when a new phase forms (i.e., at melting or boiling upper left of the diagram, through a specific temper-
temperature). A Gibbs energy–temperature diagram, ature point on an oxide line, intersects the PO2 scale
at the dissociation oxygen partial pressure (POdiss
2
) for
Table 1 Standard energies of reaction7,8 for pure solid
that oxide at that particular temperature. Accordingly,
metal (except liquid Al) and various common gasses the oxides lower on the diagram are more stable and
consequently have lower POdiss values. For instance,
DG = DH – TDS
2
Reaction from Figure 1, it is found that the dissociation pres-
(J mol1)
sure for NiO is 1010 atm at 1000 C, while that for
DH DS Cr2O3 is 1022 atm, and for SiO2 and Al2O3, it is
(J mol1) (J mol1 K) 1026 and 1034 atm, respectively. The significance of
this is that it is difficult thermodynamically to pre-
3 AlðlÞ þ 2 O2 ¼ 3 Al2 O2 565 900
2 1 1
128
Co þ 12 O2 ¼ CoO 233 886 70.7 clude the oxidation of Cr, Si, and Al.
3CoO þ 12 O2 ¼ Co2 O4 183 260 148.1 Similar free-energy diagrams, which can be inter-
3 Cr þ 2 O2 ¼ 3 Cr2 O2 373 420
2 1 1
86 preted in exactly the same way, have been constructed
Fe þ 12 O2 ¼ FeO 264 890 65.4
3FeO þ 12 O2 ¼ Fe2 O4 312 210 125.1
for sulfides, carbides, and nitrides.5 Moreover, Lou
2Fe2 O4 þ 12 O2 ¼ 3Fe2 O2 249 450 140.7 et al.10 presented an excellent review on the use of
Mn þ 12 O2 ¼ MnO 412 304 72.8 Ellingham diagrams for treating gas–solid reactions.
Ni þ 12 O2 ¼ NiO 234 345 84.3 Low oxygen partial pressures are in practice
2 Si þ 2 O2 ¼ 2 SiO2 451 040
1 1 1
86.8 achieved using oxygen-bearing gas mixtures. The
H2 þ 12 O2 ¼ H2 O 246 440 54.8
CO þ 12 O2 ¼ CO2 282 420 86.8
most common gas mixtures are H2O/H2 and CO2/
O2 þ 12 S2 ¼ SO2 362 420 72.4 CO. The partial pressures of oxygen are then estab-
NiO þ Cr2 O2 ¼ NiCr2 O4 1 376 880 332 lished from the equilibria:
NiO þ Al2 O2 ¼ NiAl2 O4 1 933 667 408
1=2
1 PH2 PO2
The calculated DG value would be for the mole numbers shown in H2 O ¼ H2 þ O2 ; K9 ¼ ½9
the particular reaction considered. 2 PH2 O
H2/H2O ratio 10–8 10–6 10–4 pO 2

CO/CO2 ratio 10–8 10–6 10–4 10–2
Temperature (F) 2
10
390 750 1110 1470 1830 2190 2550 2910 3270 3630 3990 4350
0 0 1
O3
Fe 2
2
=6 1
–100 +O M
CO 2 CO 2
e 3O
4
M M =2 =2 10–2
4F + O2 + O2
O O
2C 2C O 1
–200 u 2O iO = 2H 2
= 2C 2N H + O2
+ O2 = M 2 2
4Cu i+ O2 10–4
2N
–300 102
C + O2 = CO2 10–6
2
–400 10
ΔG = RT ln po2 (kJ mol–1 O2)
M
O 10–8
Zn B 2C
H –500 =2 O3 +O
+O 2 2 /3Cr 2 2 =2 104
n = CO
2Z + O2
C 4 /3Cr
M M B 10–10
–600
104
iO 2
=S
i+O
2
–700 S 10–12
106
–800 l 2O 3
2 3/A
= 10–14
+ O 2
106
4 3/ Al
–900 B
O
Mg
O Ca B 10–16
M
=2 =2 108
+O
2 a +O2
–1000 g 2C
2M
M Change of state Element Oxide
M 10–18
–1100 Melting point M M 108
Boiling point B B
1010
–1200 10–20
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400
Temperature (C)
CO/CO2 ratio 1014 1012 10–22
1010
OK H2/H2O ratio
po2 (atm)
10–200 10–100 10–70 10–60 10–50 10–42 10–38 10–34 10–30 10–28 10–26 10–24
Figure 1 Ellingham diagram showing the standard Gibbs energies of formation of selected oxides as a function of
temperature. Reproduced from Gaskall, D. R. Introduction to the Thermodynamics of Materials, 5th ed.; Taylor & Francis:
New York, 2008.
and these equilibrium constants if the equilibrium PCO/

1=2 PCO2 or PH2/PH2O ratios are known. In oxygen-lean
1 PCO PO2
CO2 ¼ CO þ O2 ; K10 ¼ ½10 gases containing both H2O and CO2, the PO2 is usu-
2 PCO2 ally determined by the H2–H2O reaction [9] since
where K9 and K10 are the temperature-dependent steam is more reactive than CO2. Moreover, for a
equilibrium constants for reactions [9] and [10], controlled laboratory experiment, it is preferred
respectively. Thus, the PO2 may be determined from practice to facilitate the equilibrium by using a
platinum-containing catalyst in the reaction zone of of which type of corrosion should predominate. In
the furnace. such cases, it may be necessary to conduct field-
In addition to the direct oxidation with oxygen, exposure tests to properly evaluate the corrosion
the overall metal oxidation reactions in these gas behavior of the candidate alloys. Even so, the assump-
mixtures are: tion of thermodynamic equilibrium provides a rea-
sonable starting point for assessment.
xM þ yH2 O ¼ Mx Oy þ yH2 ½11a
Quite often, two-dimensional phase stability dia-
and grams are used to assess the high temperature corrosion
of a metal M exposed to a dual-oxidant atmosphere.
xM þ yCO2 ¼ Mx Oy þ yCO ½11b
To illustrate this, the possible reactions of M exposed to
From Figure 1, the H2/H2O and CO/CO2 ratios oxidizing–sulfidizing atmosphere are determined by
can be obtained by using the same method as considering the following:
for determining POdiss
2
, using the index points labeled 1
‘H’ and ‘C’ at the left of the diagram instead of M þ O2 $ MO ½13
2
point ‘O.’
The growth of a scale is usually sufficiently slow 1
M þ S2 $ MS ½14
for local equilibrium to be closely approached. In 2
fact, it is found that the Gibbs phase rule6 can be
1 1
applied to rationalizing the phase assemblage of a MO þ S2 $ MS þ O2 ½15
growing scale. Specifically, the Gibbs phase rule 2 2
under isothermal and isobaric conditions is given as The equilibrium from reactions [13] and [14] defines
the critical PO 2 and PS2 values (i.e., dissociation pres-
PþF ¼C ½12 sures) for M/MO and M/MS equilibrium, respectively,
where P is the number of phases at a given location in while reaction [15] defines critical PS2/PO2 ratios for
the scale, F represents the degrees of freedom, and MS/MO equilibria. With regard to the latter, it can be
C number of components in the system. For the easily shown for reaction [15] that log(PS2) ¼ log(PO2) –
oxidation of a pure metal, C ¼ 2 (i.e., metal and oxy- 2logK15, where K15 is the equilibrium constant for this
gen), and for diffusion to occur there must be a reaction. Thus, a plot of log(PS2) versus log(PO2) should
gradient in the chemical activity across the scale, so have a line of slope equal to the unity that separates MS
that F ¼ 1. Thus, P ¼ 1, meaning that only one phase stability from MO stability. Figure 2 shows the general
can exist at a given plane (parallel to the metal/scale construction of an M–S–O phase-stability diagram. It is
interface) within the scale. In practical terms, this noted that a more complete diagram may include
means that a metal capable of forming multiple oxi- higher-order oxides and sulfides, as well as sulfates
des must form those oxides as distinct scale layers. (MSO4).11 Only the simpler diagram is discussed here
For example, at 1 atm O2 and temperatures above since the main important points can still be made.
560 C, pure iron oxidizes to form a triple-layered
scale in the sequence FejFeOj Fe3O4 jFe2O3 jO2, with
the oxide position being dictated by the necessity for
the oxygen content to progressively increase when
traversing from the iron to the oxygen at the scale MS
surface. A similar line of reasoning can be used to
show that a pure metal oxidized isothermally will not
2
log pS
form internal oxide precipitates, but instead must

form an external product. MO
M
1.09.2.2 Thermodynamics of Dual Oxidant
Reactions
The thermodynamic aspects of multioxidant corro-
log pO2
sion have been discussed by Giggins and Pettit.11
Often, however, the complexity of a given process Figure 2 Schematic representation of a simple
environment precludes an accurate determination phase-stability diagram for a metal and its oxide and sulfide.
The diagram in Figure 1 shows the stability range Figure 4 shows the phase-stability diagram for the
of a metal and its oxide and sulfide products as a Type 310 stainless steel at 875 C.12 Superimposed on
function of the two principal reactants: oxygen and this diagram are experimental data indicating the
sulfur. A given atmosphere is defined by either equi- type of scale formed on 310 under different PO2–PS2
librium or nonequilibrium PO2 and PS2 values and combinations. It is seen that sulfide-to-oxide transi-
would therefore be represented by a point in the tion at a given PS2 actually occurs at a higher PO2 than
stability diagram. The location of that point identifies that predicted from equilibrium calculations. The
the phase that is in stable equilibrium with that par- experimentally-determined boundary is dictated by
ticular atmosphere. However, as shown in Figure 3, kinetic factors and, accordingly, is referred to as the
other phases can form even if only oxide stability kinetic boundary. For the case of the 310 stainless
is predicted. This depends on the reacting gas, the steel shown in Figure 4, the actual PO2 values for the
PO2–PS2 combination, and whether the scale devel- transition from chromium–sulfide to chromium–
ops open pathways for gaseous penetration (e.g., oxide formation are about three orders of magnitude
microcracks). Internal sulfidation is particularly higher than the equilibrium values. The kinetic fac-
apt to occur in SO2-containing atmospheres due tors which influence the location of the kinetic
to the local equilibrium dictated by the reaction boundary include composition and surface finish of
O2 þ ‰ S2 ! SO2, which has an equilibrium constant the alloy, and gas composition. Although theoretical
denoted as K3. As indicated in Figure 3, the formation prediction of the location of a kinetic boundary is
of an internal sulfide beneath an external oxide scale not possible, LaBranche and Yurek13 showed that
can only be completely avoided if for H2–H2O–H2S gas mixtures there is a critical
qffiffiffiffiffiffiffi H2O/H2S ratio associated with the kinetic boundary.
PSO2 <K3 PO 2 PS2 ½16 The value of this ratio is dictated by the competitive
formation of the oxide and the faster-growing sulfide.
Although phase-stability diagrams are very useful for It was found by LaBranche and Yurek that oxide
interpreting reaction products and gaining insights formation on pure chromium at 900 C could only
into reaction pathways, they do not have any predic- occur when the area fraction of Cr2O3 was greater
tive capabilities from a practical standpoint. than 0.9 in the early stages of exposure, which
On the basis of equilibrium thermodynamics of corresponds to H2O/H2S > 10.
reaction [15], the transition from sulfidation to oxi-
dation of metal M should occur when
1=2
PO2
> K15 ½17 0
P S2
The thermodynamic boundary separating sulfide
FeS
and oxide stability in a phase stability diagram is Kinetic
–2
determined by replacing ‘>’ with ‘=’ in eqn [17]. boundary
log pS2 (Pa)
Cr3S4
min.
pSO = K3⋅pO
* ⋅ pS* –4
2 2 2
1 Cr2O3
SO2 or S2
MS CrS
1 MS
Sulfide-containing
log pS2
scale
MO 2 Alloy Pure oxide
SO2
M 2 MS
3 –20 –15 –10
3 log pO2 (Pa)
log pO2 MO Figure 4 Thermodynamic phase-stability diagram for
type 310 stainless steel at 875 C, showing the
M
experimentally-determined kinetic boundary. Reproduced
Figure 3 Application of the phase-stability diagram in from Stroosnijder, M. F.; Quadakkers, W. J. High Temp.
identifying possible modes of attack. Technol. 1986, 4, 141.
1.09.2.3 Kinetics of Scale Formation thickness, x, increases with time, t, and, since this
corresponds to the increasing diffusion distance,
The reaction of a metal with an oxidant to form a
the oxidation rate decreases. Thus, the instantaneous
product scale is concomitant with a weight gain due
oxidation rate is quite simply inversely proportional
to pick-up of the oxidant (e.g., grams of oxygen
to the oxide thickness, that is,
gained). Accordingly, the scaling rate may be quanti-
0
fied by the change in the scale thickness, x (cm), or by dx kp
weight gain, DW (mg cm2). In the case of oxidation, ¼ ½20
dt x
these two parameters are directly related by 0
where kp is a proportionality constant. Integration of
Vox eqn [20] gives
x¼ DW ½18
yMO x 2 ¼ kp t þ C ½21
3 1
where Vox is the molar volume of oxide in cm mol , where kp is taken to be the parabolic rate cons-
0
y is the stoichiometric amount of oxygen in oxide tant (kp ¼ 2 kp) with typical units of cm2 s1. Another
MxOy , and Mo is the atomic weight of oxygen. form of the parabolic rate equation is given by the
The formation of oxide scale is also related to the weight gain (g cm2):
consumption of metal, and the relationship between
DW and the thickness of the metal consumed (X) is DW 2 ¼ kp t þ C ½22
given by where the units of kp in this case are g2 cm4 s.
Vm The parabolic rate law is the standard for ana-
X¼ DW ½19
gMO lysis of high temperature oxidation kinetics, in which
diffusion through the relatively thick scale con-
where Vm is the molar volume of metal in cm3 mol1
trols reaction rates. The diffusion-controlled scale-
and g is the stoichiometric factor for the oxide scale
thickening process is thermally activated, meaning
product (i.e., g ¼ y/x for MxOy). For most metals, the
that the rate increases exponentially with temperature,
oxidation rates follow one or more of the three possi-
as given by the Arrhenius equation:
ble kinetic laws: linear, logarithmic, and parabolic.
Q
These kinetic laws are discussed in the following.
kp ¼ ko exp ½23
Formation of an oxide scale will separate the two RT
reactants, metal and gaseous oxygen. In order for the Here, ko is a constant that is a function of the oxide
reaction to proceed further, at least one of the reac- composition and the gas pressure, and Q is the activa-
tants must progress through the scale to form more tion energy for oxide-scale growth. Figure 51 shows
oxide at the oxide/gas, oxide/metal, or both inter- the temperature dependence of the kp values for the
faces. The mechanisms by which the reactants prog- oxides of Fe, Co, Ni, Cr, Al, and Si. The figure shows a
ress through the scale can therefore be an important range of kp values for Al2O3 and Cr2O3 growth because
part of the overall mechanism and kinetics by which these oxides do not show intrinsic behavior and are
high temperature oxidation reaction proceeds. Another instead very sensitive to impurities (i.e., doping).
aspect of the oxidation process, which can sometimes Deviations from parabolic kinetics are generally
be rate controlling, is the kinetics of the interfacial analyzed in terms of chemical and metallurgical effects
reaction steps. on the rates of the relevant diffusing process(es). The
parabolic rate law will not hold in the very early stages
1.09.2.3.1 Parabolic rate law of oxidation, before the scale has developed sufficient
At high temperature, initial scale growth is usually continuity and thickness. In the case of hafnium and
very rapid; however, the reaction rate will eventually zirconium, the time exponent in eqn [21] is found to be
decrease when scale thickness reaches 0.5 mm and 3 (cubic kinetics) rather than 2 (parabolic kinetics) at
the transport of reacting species through the scale high temperatures. This has been attributed to the
becomes rate controlling. When the rate-controlling simultaneous dissolution of oxygen into the substrate
step in the oxidation process is the diffusion of reac- metal during oxidation.14 It has also been shown for
tant(s) through the oxide layer and with the boundary NiO-15 and Al2O3-scale16 growth that subparabolic
conditions for diffusion being time independent, the kinetics can occur if both short-circuit diffusion
scaling kinetics will follow the parabolic rate law. through the scale predominates and the average grain
Parabolic kinetics results from the fact that the scale size of the scale increases with oxidation time.
T (C) microcracking and porosity may develop as the scale

1200 1100 1000 900 800 700 600 thickens, reducing the protectiveness of the oxide.
10–5 The parabolic rate law may then fail, and the kinetics
approaches linearity at some time after the start of
Parabolic scaling constant, k (g2 cm–4 s–1)
10–7 FeO/Fe3O4/Fe2O3 reaction and scale growth. That is, a constant oxida-
tion rate can develop after a period of parabolic
behavior. In this special case of parabolic scaling
10–9 superimposed on a relatively constant rate of scale
cracking and healing, the overall kinetics is said to
CoO
be ‘paralinear.’20
10–11
NiO
Cr2O3
1.09.2.3.3 Logarithmic rate law
At low temperatures (e.g., T < 300–400 C), oxidation
10–13 SiO2 on Al2O3 rates are often inversely proportional to time, that is,
MoSi2
dx k
¼ ½25
10–15 dt t
where k is a constant. Integration of [25] leads to the
10–17 logarithmic rate law
6 7 8 9 10 11 12 13
104/T (K–1) x ¼ ka logðkb t þ 1Þ ½26
Figure 5 Parabolic oxidation rate constant for various where ka and kb are constants. Logarithmic oxidation is
oxide scales as a function of temperature. Reproduced from usually obeyed for relatively thin scales at low
Gleeson, B. In Corrosion and Environmental Degradation, temperatures.
Vol. II: Volume 19 of the Materials Science and Technology
Series; Schütze, M., Ed.; Wiley-VCH: Weinheim,
Germany, 2000. 1.09.2.4 Transport Properties of Metal
Oxides
1.09.2.3.2 Linear rate law Metallic oxides are seldom stoichiometric, meaning
Under certain conditions, the oxidation of a metal that the metal-to-oxygen atom ratio is not exactly
proceeds at a constant rate according to a linear rate that given by the stoichiometric chemical formula,
law, that is, even though the compound is electrically neutral.
The same can be stated for sulfides and nitrides,
x ¼ kl t ½24
but the focus here will be on oxides as the prototypi-
where x is the scale thickness and kl is the linear rate cal systems. The ionic charge imbalance in a non-
constant. A linear rate law may result when a phase- stoichiometric oxide is compensated by electronic
boundary reaction controls the kinetics rather than a charges, that is, electron and electron holes. As a
transport process. An example is CO2 dissociation at consequence, nonstoichiometric oxides exhibit both
the scale surface controlling the oxidation kinetics of electronic and ionic conductivities. These conductiv-
steel in a CO2-rich atmosphere.17 ities are temperature dependent, which, from the
Linear kinetics is also possible if the oxide is electronic standpoint, classifies the nonstoichio-
volatile or molten, if the scale spalls or cracks, or if metric oxides as semiconductors.
a porous, nonprotective oxide forms on the metal.4 Electronic semiconductors are categorized as
Since the rate of oxidation never slows down, con- n-type (excess of electrons) or p-type (excess of elec-
sumption of the metal occurs in a relatively short tron holes). The n-type oxides may have either an
time in comparison to a metal scaling according to excess of cation interstitials or an excess of oxygen
parabolic kinetics. Examples of metals that scale in a vacancies (i.e., a deficiency in filled sites on the oxy-
nonprotective, linear manner due to continual scale gen sublattice, MO1x) as the principal ionic defects.
cracking are Nb and Ta.18 Some examples of n-type oxides are TiO2, Fe2O3,
In the early stages of a metal oxidation process, the NiFe2O4, ZnO, Al2O3, and SiO2.18 For n-type oxides,
scale may be sufficiently thin that linear oxidation the principal ionic defect concentration, C, is found
kinetics prevails. As the scale thickens, a transition to to be proportional to the negative power of the oxy-
parabolic kinetics will usually ensue.19 Conversely, gen partial pressure21:
1=n The addition of substitutionally dissolved foreign

C / PO2 ½27
cations of lower valence increases the concentra-
where n is a positive integer which depends upon the tion of electron holes and decreases the concentra-
charge of the ionic defect in the oxide and the oxide’s tion of cation vacancies.
stoichiometry. The addition of substitutionally dissolved foreign
The p-type oxides may have either an excess of cations of higher valence decreases the concentra-
cation vacancies (i.e., a deficiency in filled sites on the tion of electron holes and increases the con-
metal sublattice, M1xO) or an excess of oxygen centration of cation vacancies.
interstitials. Some common p-type oxides include
NiO, CoO, FeO, FeCr2O4, CoCr2O4, and NiAl2O4. Thus, for the p-type, metal-deficient NiO, doping
For p-type oxides, the principal ionic defect concen- with Cr3+, will cause an increase in the nickel vacancy
tration is related to the oxygen partial pressure by21: concentration, which in turn will cause an increase
1=n
in the effective diffusivity of nickel in the NiO. This
C / PO2 ½28 increase in diffusivity is manifested as an increase
During the oxidation process, ions transport by two in the kp for NiO-scale growth. Indeed, it is found
ways: the cation migrates outward to scale/gas inter- experimentally that NiO grows faster on dilute
face and the anion migrates inward to metal/scale Ni–Cr alloys than on pure Ni.22
interface (Figure 6). Thus, two reactions could
potentially happen at the two interfaces to result in 1.09.2.5 Wagner’s Theory of Metal Oxidation
scale growth, although it is typically found that one Carl Wagner’s theory of metal oxidation23,24 provides
reaction predominates. In other words, either metal a fundamental understanding of the essential features
or oxygen diffusion predominates in the thickening of of the high temperature growth of a continuous scale.
a given oxide scale. The model ideally assumes that the scale is dense,
For the defective structures treated so far, it has single-phase, continuous, and adheres to the metal over
been assumed that the oxides are pure; however, it is the entire metal surface. The basic assumption of the
thermodynamically impossible to produce perfectly theory is that lattice diffusion of the reacting atoms or
pure compounds or materials. It is therefore neces- ions through the dense scale is rate controlling (i.e.,
sary to consider the effects of impurities on the the microstructure of the oxide scale was not consid-
defective structure of oxides. It is noted, however, ered). Figure 7 gives a set-up of the model.14
that the presence of impurities can sometimes be As indicated in Figure 7, the scale growth involves
neglected if the intrinsic defect concentration in the fluxes of both ionic and electronic charged species.
host oxide is relatively large. As an example of the The driving forces for these fluxes are related to
so-called ‘doping effect,’21 the following is found for a the chemical potential gradient and electrostatic
p-type, metal-deficient oxide M1xO: field that develop in the growing scale. The relative
M MO O2 M MO O2
M2+ O2–
2e– 2h+
M = M2+ + 2e– M2+ + 2e– + ½O2 = MO M + O2– + 2h+ = MO ½O2 = O2– + 2h+
Figure 6 Ionic and electronic transport processes and interfacial reactions in the growth of a nonstoichiometric oxide
scale MO.
Metal M Oxide MO Gas O2
Cations
Anions
Electrons
Electron holes
aM = 1 PO2

ΔGMO
= exp
aM
ΔG 2RT
pO 2 = exp MO
RT ( pO2) 1/2
M = M2+ + 2e– or M2+ + 2e– + ½O2 = MO or

M + O2– = MO + 2e– ½O2 + 2e– = O2–
Figure 7 Transport processes according to Wagner’s theory. Adopted from Birks, N.; Meier, G. H. Introduction to High
Temperature Oxidation of Metals; Edward Arnold: London, 1983.
migration rates of cations, anions, electrons, and elec- Because the diffusion flux within the oxide is
tron holes must also be balanced such that no net proportional to the defect concentration, eqn [28]
charge build up occurs within the scale. can be used for a p-type oxide to arrive at the follow-
Wagner derived an expression for the parabolic ing relation for the oxidation rate:
rate constant in terms of the electronic and ionic 0
conductivities of the oxide or, alternatively, in terms k / ½ðPOo 2 Þ1=n ðPOi 2 Þ1=n ½31a
of the self-diffusion coefficients of the reacting ions in
where POo 2 and POi 2 are the oxygen pressures at the
which parameters can be measured relatively easily.
scale/gas interface and metal/scale interface respec-
Limiting cases of Wagner’s derivation are as follows14:
tively, and n is an integer related to the cation vacancy
mð0M
0 1 or oxygen interstitial charge. For an n-type oxide, use
k ¼ DM dmM ½29 of eqn [27] gives:
RT
m00M
k / ½ðPOi 2 Þ1=n ðPOo 2 Þ1=n
0
½32a
and
mð0X where in this case n is related to the oxygen vacancy
1
0
k ¼ DX dmX ½30 or cation interstitial charge. In most cases, the ambi-
RT ent oxygen pressure POo 2 is much greater than POi 2 ,
m00X
which is the dissociation pressure (POdiss
2
). Thus, eqns
where k 0 is the parabolic rate constant with units of [31a] and [32a] can be approximated to give
cm2s1, DM and DX are the self-diffusion coefficients for 0
metal, M, and nonmetal, X, through the scale, respec- k / ðPOo 2 Þ1=n for a p-type scale ½31b
tively, and mM and mX are the chemical potentials for the
metal and nonmetal. Equation [29] is valid when cation and
diffusion predominates and eqn [30] is valid when anion k / ðPOi 2 Þ1=n for a n-type scale
0
½32b
diffusion predominates. The good agreement between
parabolic rate constants calculated from diffusivities and These equations show that the growth rate of a
the oxidation rate constants measured experimentally p-type oxide is directly dependent on the oxygen
have provided validation for Wagner’s theory.25,26 partial pressure in the atmosphere. By contrast, the
growth rate of an n-type oxide is independent of the 1.09.3 Alloy Reactions

external oxygen partial pressure.
Oxides formed in practice are usually more com- The fundamentals of pure-metal oxidation provide a
plex than what was assumed by Wagner. First, a multi- basic understanding of alloy oxidation, but the latter
layer scale can form on a number of metals, such as is generally much more complex as a result of some,
Fe, Cu, and Co. For instance, and as discussed earlier, a or all, of the following14:
three-layered FeO/Fe3O4/Fe2O3 scale forms on iron
above 570 C at atmospheric pressure. Treatments (a) The various metal components in an alloy will
of the growth kinetics of multilayered scales have been have different affinities for oxygen, as indicated
presented by a number of authors.27–30 In general, these in the Ellingham diagram in Figure 1.
treatments relate overall oxidation scaling kinetics to (b) More oxides may be formed, including ternary
the intrinsic growth rates of the individual layers in the and higher oxides.
scale. Second, scales forming on metal surfaces are (c) A degree of solid solubility may exist between
usually polycrystalline in structure, and in many cases, the oxides.
these structures are fine-grained. The activation energy (d) The various metal ions will have different mobi-
for grain-boundary diffusion is lower than that for lattice lities in the oxide phases.
diffusion by up to a factor of three. As a consequence, (e) The various metals will have different diffusiv-
grain-boundary diffusion will tend to predominate at ities in the alloy.
lower temperatures and can cause the scaling kinetics to (f) The extra degree of thermodynamic freedom
be orders of magnitude higher than what would be provided by an additional metal component in
predicted based on lattice diffusion.26 the alloy may result in subsurface oxide precipi-
As indicated above, Wagner’s theory considered tation (i.e., internal oxidation).
the ideal case of scale formation in which the scale is When it is considered that scales can crack, contain
assumed to be a compact and perfectly adherent voids, spall, and even form multiple layers of irregu-
barrier, that is, free of voids, pores, and fissures. The lar thickness, the situation becomes even more com-
assumptions made by Wagner were indeed appropri- plex. The following provides a brief overview of the
ate; however, deviations from Wagner’s theory have fundamental aspects of alloy oxidation.
been shown.31 Many studies have shown that pores
form and develop preferentially along grain bound-
aries in the scale.32,33 The presence of voids in scales 1.09.3.1 Thermodynamics of Alloy Oxidation
clearly represents a deviation from ideal scale growth. As in the case of a pure metal, alloy oxidation is
In some extreme cases, it may be necessary to take driven by an overall decrease in free energy. How-
into account the void volume fraction in Wagner’s ever, the alloy components are not in their pure state,
model; but often the void formation may be consid- which corresponds thermodynamically to their chem-
ered to be a secondary process that does not signifi- ical activities being less than unity. For dilute solution
cantly affect the overall scaling kinetics. of metallic component B in A, the average solvent
As the scale grows, stresses develop due to differ- A atom exists essentially in the same chemical sur-
ences in the molar volume of metal and oxide(s). The rounding environment as in its pure state, with only a
resulting growth stresses are typically compres- small number of neighboring solute B atoms. Thus, in
sive.34,35 More significant stresses can develop under the limit of a very dilute solution, the A atoms follow
thermal cycling conditions due to a mismatch in the Raoult’s law,6 such that:
coefficient of thermal expansion (CTE) between
oxide scale and metal. Increasing stresses may even- lim aA ¼ NA ½33
NA !1
tually result in crack formation and even scale
detachment.35 Cracking of a protective oxide scale where aA and NA are the chemical activity and mole
can result in the parabolic oxidation kinetics being fraction of solvent A, respectively. Also, according to
interrupted by a sudden increase in rate when the gas Henry’s law, the chemical activity of solute B is given as:
can react directly with the metal surface. As oxide lim aB ¼ gB NB ½34
NB !0
begins to cover the metal surface again, parabolic
oxidation is resumed. The overall oxidation of the where gB is the activity coefficient of solute B for the
metal becomes approximately a paralinear process of A–B system. It can be easily shown that the dissociation
periodic cracking and healing of a protective oxide. PO2 for an AO scale in contact with an A–B alloy is
greater than what would be calculated for AO in con- remainder of this chapter in order to provide more
tact with pure A. Specifically, general kinetic-based expressions.) An example of
2 this could be the preoxidation of an alloy at a low
PO2 1
¼ ½35 PO2 followed by in-service exposure to a higher PO2
PO 2 aA
and/or different temperature. The sustained growth
where aA is the activity of A in the A–B alloy, and PO2 of a continuous BOu scale requires a sufficient supply
and PO 2 are the dissociation pressures for the A–B alloy of B from within the alloy to the alloy/scale interface.
and pure A, respectively. The resulting subsurface concentration gradient of
It is typically the activity of the solute that is solute B is represented schematically in Figure 8, in
most important when assessing the thermodynamics which NBi is the mole fraction of B in the alloy at the
of alloy oxidation; although, the results are often alloy/scale interface. The maximum possible rate of
of limited utility. This can be shown as follows. B supply in the alloy can be imposed by setting NBi
Consider the oxidation of an A–B alloy in which the equal to zero, thereby producing the steepest possible
solute B forms a more stable and protective oxide. diffusion gradient of B. Under steady-state conditions,
Assuming that the only stable oxides are AO and BO this maximum rate of B supply would have to equal
and that these oxides are insoluble in one another, the the rate of B consumption due to BOu scale growth.
important reaction to consider is Determination of the minimum B content in the alloy,
o
NBðminÞ , necessary for the sustained exclusive growth
AO þ B ! BO þ A ½36
of a BOu scale on an A–B alloy was originally consid-
where B and A correspond to B and A in the ered by Wagner,38 who further assumed that: (1) the
alloy solution. The law of mass action for this reaction diffusion coefficient of B in the alloy, DB, is indepen-
gives dent of concentration; (2) the BOu scale obeys para-
aA NA bolic thickening kinetics with a rate constant kp;
K36 ¼ ½37 (3) solvent metal A is insoluble in BOu; and (4) the
aB gB NB
recession of the alloy/scale interface may be neglected.
Thus, reaction [36] will only proceed to the right Wagner derived the criterion
if aaAB < K36 , which means that there is a critical aB, 1
or conversely NB, for this to occur. In most practical Vm pkp 2
NBðminÞ >
o
½38
cases of interest, the solute–metal oxide BO is signifi- nMO 2DB
cantly more stable than the base-metal oxide AO, where Vm is the equivalent molar volume of the alloy
causing K36 to be much larger than unity. This, in and MO is the atomic weight of the oxidant, which is
turn, results in a very small minimum NB value for oxygen in this discussion (M ¼ 16 g mol1). Many
BO to be thermodynamically stable in contact with researchers have used the criterion in eqn [38]
the A–B alloy. In the case of an Ni–Cr alloy, NCr must to predict the minimum content of B necessary for a
be greater than 109 at 1000 C for the displace- bare alloy to form an exclusive BOu scale layer. However,
ment reaction 3NiO þ 2Cr ! Cr2O3 þ 3Ni to be this is not a correct use of the criterion, since its deriva-
thermodynamically stable.36 Such a low Cr content is tion was based on supply rather than establishment. The
orders of magnitude below what is necessary kineti-
cally to sustain Ni–Cr/Cr2O3 stability, let alone to
kinetically establish a continuous Cr2O3 scale. For the A-B O2
crit BOν
latter, the critical Cr mole fraction, NCr , in Ni–Cr

alloy exposed to air at 1000 C is experimentally
found to be 0.2.37 N 0B
JB
1.09.3.2 Criterion for the Sustained

Exclusive Growth of a Protective Scale
If a continuous BOu scale on an A–B alloy is estab- N iB
lished under a particular set of conditions, it may be
necessary to determine if its growth can be sustained Figure 8 Schematic representation of the concentration
under a different set of conditions. (The stoichiomet- profile of B in a binary alloy A–B which is forming an
ric factor u for the oxide BOu will be used for the exclusive scale layer of BOn.
criterion given in [38] gives only the minimum possible diffusion-controlled and the depth of the internal
B content in the alloy necessary to supply B at a suffi- oxidation zone, x, in an alloy undergoing no external
cient rate to the alloy/scale interface for the sustained scale formation can be described by the following
growth of an established BOu scale layer. The actual kinetic expression, assuming no enrichment of B in
B content necessary for both the establishment and the internal oxidation zone44,45:
sustained growth of a BOu scale will very likely be higher s
o 2NO DO t 1=2
than NBðminÞ owing to transient and kinetic effects. x¼ ½41
nNBo
where NOs is the solubility of the oxidant in the alloy,
1.09.3.3 Internal Oxidation DO is the diffusivity of the oxidant in the alloy, NBo is
the initial solute concentration in the A–B alloy, and
Internal oxidation is used here in a generic sense to t is the time. The product NOs DO is the oxygen per-
represent a process in which a diffusing oxidant from meability in alloy. The solubility of oxygen in the
the surface reacts with a less-noble solute component alloy depends upon the oxygen partial pressure at the
in the alloy to form discrete particles.39 Internal oxi- alloy surface in accordance with Sievert’s law,6 that is,
dation is not desired because it changes the optimized
1=2
mechanical properties of an alloy and may result in NOs / PO2 ½42
internal stress and weakening of the grain boundary.
Thus, the depth of internal oxidation would be
For a dilute binary alloy A–B, the dissolved oxygen
expected to decrease with decreasing PO2 in the man-
atoms can react with less-noble B atoms in the alloy in 1=4
ner x / PO2 .
the manner,
The above discussion considers only internal pre-
BþuO ! BOu ½39 cipitation in the absence of external scale formation.
In the case of AOu scale formation, the value of NOs is
The necessary condition for BOu formation in the no longer fixed by the environment but is instead
alloy may be formulated in terms of the equilibrium fixed by the equilibrium between at the alloy/scale
solubility product Ksp , such that: interface (assuming no through-scale access by the
oxidant) according to the reaction,
½aB ½aO n > Ksp ½40
n
A þ O2 ¼ AOn ½43
The degree to which the solubility product must be 2
exceeded (i.e., the degree of supersaturation) before The effective thickness of the internal precipitation
precipitation occurs depends on a number of factors, zone in the situation of concurrent AOu-scale forma-
such as the stability, composition, and crystal structure tion decreases in a most pronounced manner from
of the precipitating BOu phase. For instance, if the that predicted in the absence of an external scale at
crystal structure and lattice dimensions of BOu are large kp and NBo values.46 A large kp value for AOu-
such that it can form a coherent interface with the scale growth corresponds to a large amount of metal
alloy matrix, then only a small degree of supersatura- recession, which means the total attack is still high,
tion is likely to be necessary for BOu precipitation. In even though the internal precipitation zone is rela-
general, the greater the stability of BOu , the lower will tively low.
be its Ksp value. It is for this reason that the internal The preceding theoretical considerations are
precipitates commonly observed in oxidized high based on models assuming predominant lattice diffu-
temperature alloys are of the stable oxides Al2O3, sion of both oxygen and the alloying element. As
SiO2, TiO2, and Cr2O3.40,41 Multiple internal oxida- such, the theoretical treatments are only expected
tion zones can also develop if more than one reaction to apply to relatively high temperatures or to very
product is stable.42,43 The sequence of the thermody- large-grained materials, with the extreme being sin-
namically possible phases progresses from metal-rich gle crystals. As the temperature is reduced, grain-
in the innermost zone of the alloy to oxidant-rich at boundary effects will become increasingly important.
the surface. This is due to relatively rapid diffusion along grain
The internal oxidation zone extends to the depth boundaries, enhanced concentration or segregation of
at which the activity of dissolved oxygen becomes too the oxidant and the alloying element at grain bound-
small for the formation of the oxide BOu . The kinet- aries, and a preferred tendency for nucleation at grain
ics of internal oxidation are generally found to be boundaries.4
1.09.3.4 Transition from Internal Oxidation experimental studies conducted over the past
to External Scale Formation 40 years have validated this criterion.1
The transition from internal to external oxide forma-
tion typically occurs with a relatively small increase 1.09.4 Epilogue
in the alloy content of the less-noble component B in
the A–B alloy, at which oxidation leads to the forma- Advances in the high temperature stability of materi-
tion of a surface scale of BOu, and the alloy no longer als are critically needed to realize the full perfor-
undergoes internal oxidation. Considering again the mance and potential of many current and future
alloy system A–B, there are two limiting possible commercial systems. These advances must lead to
situations: one is when only one of the components significantly enhanced capabilities that will allow
can oxidize and the other is when both of the com- high temperature components to operate robustly
ponents can oxidize. For the first case, if the oxygen for prolonged periods in harsh environments, such
partial pressure is low (i.e., below the stability of AOu as those involving aggressive gases, deposits, photon
formation), the component B in the alloy surface or radiation fluxes, stresses, high or low pressures,
nucleates as BOu in an A-rich matrix. If B can diffuse or some combination of these conditions. For exam-
fast enough to reach the surface and maintain a ple, coal gasification, biomass conversion, and gas-
sufficient supply for BOu formation, a complete sur- cooled nuclear reactor systems typically produce
face layer of BOu will be established. However, if this complex, multioxidant gaseous environments that
condition is not met, then BOu will precipitate as can be highly aggressive from the standpoint of
internal particles within the alloy. So, the internal or surface degrading structural components. The result-
external formation of BOu depends on the balance ing multioxidant process environments are often non-
between the outward flux of B and the inward flux of equilibrium and can involve both gaseous and deposit-
atomic oxygen into the alloy. induced attack. Fundamentally, the high temperature
Wagner47 proposed that the condition for the stability of a material in an aggressive environment
transition from internal to external BOu formation relates to reactions and transport at and across its
occurs when a critical volume fraction of BOu, f *, is external surfaces. Similar to what was discussed in
attained. Under this condition, the influx of oxygen is this chapter, these reactions are defined in terms of
so restricted that sideways growth of the internal BOu some combination of chemical potential, temperature
precipitates is kinetically favored to the extent that and pressure and can be highly complex. Different
BOu eventually forms as a continuous layer on the processes, many of them coupled, are involved from
alloy surface. Under conditions of no or a negligibly the onset of reaction, that is, the absorption and dis-
small rate of metal recession, the following criterion sociation of gaseous molecules at the surface, to the
for the transition from internal to external BOu for- steady-state growth of a protective surface scale that
mation was obtained: develops. Indeed, the ability to control the growth and
1=2 stability of this scale and to predict its behavior under
Vm NOS DO
N B f
o
p ½44 different types of extreme chemical environments for
Vox 2nDB extended periods of time will require a much greater
where NBo is the critical mole fraction of B in the basic understanding of the underlying reactions and
alloy, Vm is molar volume of the alloy, Vox is molar transport processes involved. The basic starting points
volume of the oxide, and DB is the diffusion coeffi- have been presented in this chapter, but still much
cient of solute B in the alloy. Rapp44 reported excel- more research and development are needed to improve
lent agreement between experimental and reasonably the reliability and durability of materials exposed to
predicted values of NIno as a function of oxygen aggressive conditions.
partial pressure in Ag–In alloys oxidized at 550 C.
The value of f * is usually taken to be 0.3 in accor-
dance with the results from this study by Rapp.
From the criterion given by eqn [44], it is seen that References
NBo can be decreased by decreasing the product
of NOs and DO (i.e., decreasing the permeability of 1. Gleeson, B. In Corrosion and Environmental Degradation,
Vol. II: Volume 19 of the Materials Science and Technology
oxygen) and/or increasing DB. Indeed, the direct Series; Schütze, M., Ed.; Wiley-VCH: Weinheim, Germany,
and indirect control of these variables in numerous 2000.
2. Lai, G. Y. High-Temperature Corrosion of Engineering 21. Kofstad, P. Nonstoichiometry, Diffusion and Electrical
Alloys; ASM International: Materials Park, OH, 1990. Conductivity in Binary Metal Oxides; Wiley-Interscience:
3. Kofstad, P. High Temperature Oxidation of Metals; John New York, 1972.
Wiley and Sons: New York, 1966. 22. Giggins, C. S.; Pettit, F. S. Trans. AIME 1969, 245, 2495.
4. Young, D. High Temperature Oxidation and Corrosion of 23. Wagner, C. Z. Phys. Chem. 1933, 21, 25.
Metals; Elsevier: Amsterdam, 2008. 24. Wagner, C. Atom Movements; ASM: Cleveland, 1951;
5. Alcock, J. C. B.; Easterbrook, E. Thermodynamics and p 153.
Kinetics of Gas-Metal Systems. In Corrosion, Volume 1: 25. Raynaud, G. M.; Clark, W. A. T.; Rapp, R. A. Metall. Trans. A
Metal/Environment Reactions, 3rd ed.; Shreir, L. L., 1984, 15A, 573.
Jarman, R. A., Burstein, G. T., Eds.; Butterworth- 26. Atkinson, A. Rev. Mod. Phys. 1985, 57, 437.
Heinemann: London, 1994. 27. Garnaud, G.; Rapp, R. A. Oxid. Met. 1977, 11, 193.
6. Gaskall, D. R. Introduction to the Thermodynamics of 28. Hsu, H. S. Oxid. Met. 1986, 26, 315.
Materials, 4th ed.; Taylor & Francis: New York, 2008. 29. Gesmundo, F.; Viani, F. Corros. Sci. 1978, 18, 217.
7. Kubaschewski, O.; Alcock, C. B. Metallurgucal 30. Wang, G.; Gleeson, B.; Douglass, D. L. Oxid. Met. 1989,
Thermochemistry, 5th ed.; Pergamon Press: Oxford, 1979. 31, 415.
8. Barin, I.; Platzki, G. Thermochemical Data of Pure 31. Yurek, G. J. Corrosion Mechanisms; Marcel Dekker Inc.:
Substances; VCH: Weinheim, 1995. New York, 1987; p 397.
9. JANAF Thermochemical Data, Army-Navy-Air Force 32. Dravinieks, A.; McDonald, H. J. Electrochem. Soc. 1948,
Thermochemical Panel, Dow Chemical Company, 94, 139.
Midland, MI, 1962–1963. 33. Mrowec, S. In Proceedings of JIM International
10. Lou, V. L. K.; Mitchell, T. E.; Heuer, A. H. J. Am. Ceram. Symposium on High Temperature Corrosion of
Soc. 1984, 68, 49. Metals and Alloys; Mt Fuji, Japan, 17–20 November
11. Giggins, C. S.; Pettit, F. S. Oxid. Met. 1980, 14, 363. 1982; p 115.
12. Stroosnijder, M. F.; Quadakkers, W. J. High Temp. 34. Stringer, J. Corros. Sci. 1970, 10, 1970, 513.
Technol. 1986, 4, 141. 35. Schütze, M. Protective Oxide Scales and Their
13. LaBranche, M. H.; Yurek, G. J. Oxid. Met. 1987, Breakdown; Institute of Corrosion and Wiley Series on
28, 73. Corrosion and Protection; John Wiley & Sons Ltd:
14. Birks, N.; Meier, G. H. Introduction to High Temperature England, 1991.
Oxidation of Metals; Edward Arnold: London, 1983. 36. Birks, N.; Rickert, H. J. Inst. Met. 1963, 91, 1963, 308.
15. Peraldi, R.; Monceau, D.; Pieraggi, B. Oxid. Met. 2002, 58, 37. Wood, G. C.; Wright, I. G.; Hodgkiess, T.; Whittle, D. P.
275. Werkst. Korros. 1970, 20, 900.
16. Naumenko, D.; Gleeson, B.; Wessel, E.; Singheiser, L.; 38. Wagner, C. J. Electrochem. Soc. 1952, 99, 369.
Quadakkers, W. J. Metall. Mater. Trans. A 2007, 38A, 39. Douglass, D. L. Oxid. Met. 1995, 44, 81.
2974. 40. Gleeson, B.; Harper, M. A. Oxid. Met. 1998, 49, 373.
17. Lee, V. H. J.; Gleeson, B.; Young, D. J. Oxid. Met. 2005, 41. Wei, F. I.; Stott, F. H. High Temp. Technol. 1989, 7, 59.
63, 15. 42. Schnaas, A.; Grabke, H. J. Oxid. Met. 1978, 12, 387.
18. Bradford, S. A. Fundamentals of Corrosion in Gases. In 43. Kane, R. H. Corrosion 1981, 37, 187.
Metals Handbook, 9th ed.; Corrosion; ASM International: 44. Rapp, R. A. Corrosion 1965, 21, 382.
Metals Park, OH, 1987; Vol. 13. 45. Wagner, C. Z. Elektrochem. 1959, 63, 772.
19. Pettit, F. S.; Wagner, J. B. Acta Met. 1964, 12, 35. 46. Gesmundo, F.; Viani, F. Oxid. Met. 1986, 25, 269.
20. Sheasby, J. S.; Smeltzer, W. W. Oxid. Met. 1981, 15, 215. 47. Wagner, C. Z. Elektrochem. 1959, 63, 772.
1.10 Oxidation of Metals and Alloys
P. Y. Hou
Lawrence Berkeley National Laboratory, Materials Sciences Division, 1 Cyclotron Rd., Berkeley, CA 94720, USA

1.10.2 Oxidation of Nickel, Cobalt, and Iron 197
1.10.2.1 Nickel 197
1.10.2.1.1 Transient stage oxidation 197
1.10.2.1.2 Oxide structures 197
1.10.2.1.3 Growth rate and transport processes through NiO 199
1.10.2.2 Cobalt 200
1.10.2.3 Iron 201
1.10.3 Oxidation of Refractory Metals and Their Alloys 203
1.10.3.1 Oxidation of Ta, Nb, Mo, and W Metals 203
1.10.3.1.1 Tantalum and niobium 203
1.10.3.1.2 Tungsten and molybdenum 205
1.10.3.2 Oxidation of Refractory Alloys 206
1.10.3.3 Coatings on Refractory Metals and Alloys for Oxidation Protection 207
1.10.4 Oxidation of Silica-Forming Alloys 208
1.10.4.1 Si-Containing Alloys 208
1.10.4.2 Silicides 209
1.10.5 Oxidation of Chromia-Forming Alloys 211
1.10.5.1 Transport Properties and Oxidation Rates 211
1.10.5.2 Cr2O3 Layer Development on Alloys 213
1.10.5.3 Scale Morphology and Breakaway Oxidation 214
1.10.6 Oxidation of Alumina-Forming Alloys 215
1.10.6.1 Scale Development 216
1.10.6.2 Oxidation Rates 218
1.10.6.3 Oxide Growth Mechanism 219
1.10.6.4 Scale Adhesion 223
1.10.7 Effect of Minor Alloying Elements and Impurities 224
1.10.7.1 Reactive Elements 224
1.10.7.1.1 Promoted selective oxidation 226
1.10.7.1.2 Decreased growth rate and changed growth direction 226
1.10.7.1.3 Improved scale adhesion 228
1.10.7.2 Nonmetallic Impurities 230
1.10.7.2.1 Sulfur 230
1.10.7.2.2 Other impurities 231
References 232
RE Reactive element
Abbreviations
SEM Secondary electron microscopy
AE Acoustic emission
SIMS Secondary ion mass spectrometry
AES Auger electron spectrometry
SNMS Secondary neutral mass spectrometry
EDS Energy dispersive spectroscopy
STEM Scanning transmission electron microscopy
EELS Electron energy loss spectroscopy
TEM Transmission electron microscopy
ppm Parts per million
195
196 Types of High Temperature Corrosion
UHV Ultra high vacuum and their alloys, and then by SiO2-, Cr2O3-, and
YAG Yttrium aluminum garnet Al2O3-forming alloys. In each section, results on the
nucleation, growth and protectiveness of the oxide
formed are presented with the following topics cov-
ered whenever data are available: initial stage scale
Symbols development, oxidation rates, transport mechanisms
b Oxide grain size and dominating diffusion paths, development of
D Diffusivity oxide phases including oxide phase transformation,
DG Gibb’s free energy and evolution of scale microstructure and its rela-
kp Parabolic rate constant tionship to the growth mechanism and scaling rate.
pO2 Partial pressure of oxygen A section (Section 1.10.7.2) on impurity segregation
R Gas constant (mainly S, but also C) at different oxide–alloy inter-
T Temperature faces, and its effect on oxide scale adhesion is included.
d Grain boundary width The data cover a range of Al2O3-forming alloys, with
or without Pt or reactive elements (REs), and some
Cr2O3-forming alloys. In Section 1.10.7.1, the RE
effect on Cr2O3- and Al2O3-forming alloys, and some
1.10.1 Introduction work with NiO on Ni, is presented. Mechanisms that
have been proposed for the RE effects on oxide nucle-
This chapter provides an overview of the corrosion ation, growth and adhesion are briefly reviewed.
behavior of different metals and alloys where oxygen Figure 1 compares the range of oxidation rates at
is the only oxidant. Isothermal oxidation studies con- different temperatures for the major oxides discussed
ducted in air and oxygen are considered, but any in this chapter. Above 1100 C, the oxide on Si has
effects due to the presence of nitrogen in air, or the slowest growth rate, but the establishment of a
humidity in the atmosphere are not included; these protective surface SiO2 layer is difficult; more details
topics are covered in Chapter 1.13, Nitridation of are presented in Section 1.10.4. The kinetics for
Alloys and Chapter 1.17, Oxidation in Steam and
Steam/Hydrogen Environments. Only pure metals
and model alloys are discussed to illustrate the oxida- Temperature (8C)
tion process; the behavior of more complicated com- 1100 700 500 300
1E–006
mercial alloys is addressed in Chapter 1.23, High
Temperature Corrosion of Chromia-forming Iron,
1E–007
Nickel and Cobalt-base Alloys and Chapter 1.24, Fe
High Temperature Corrosion of Alumina-forming
1E–008
Iron, Nickel and Cobalt-base Alloys. The principles
Oxidation rate kp (g2 cm–4 s–1)
that govern many aspects of high temperature oxida-

1E–009 Co
tion that are considered in this chapter, such as oxida-
tion kinetics, transport mechanisms through oxides, 1E–010
defect formation, stress generation and relaxation, Al
microstructural evolution, and scale failure mechan- 1E–011
isms, are thoroughly covered elsewhere in this volume. Ni
In this chapter, only experimental results on these phe- Si
1E–012
nomena will be summarized, without details of mecha- Cr
nistic discussions. The purpose here is to provide a Nb
1E–013
short view of the current understanding of the oxida-
tion behavior of a range of metals and alloys. More 1E–014
extensive discussions can also be found in textbooks on
the oxidation of metals and alloys.1–3 1E–015
The chapter is divided into sections on similar 0.0004 0.0008 0.0012 0.0016 0.002
types of oxides that are formed on metals and alloys. 1/T (K–1)
The oxidation of Ni, Co, and Fe is first summarized, Figure 1 Summary of parabolic oxidation rate constants
followed by that of refractory metals (Nb, Mo, Ta, W) for some of the metals discussed in this chapter.
refractory metals, such as Nb, are only parabolic at known of its oxidation mechanism and the scale
low temperatures. Linear kinetics and oxide volatiliza- microstructure that develops.6–12
tion are often observed at higher temperatures (Sec-
tion 1.10.3). The rate for Co oxidation shows two
1.10.2.1.1 Transient stage oxidation
activation energies that are associated with the forma-
According to the surface studies of Mitchell et al.,13,14
tion of Co3O4 above a growing CoO layer at lower
the initial stage oxidation of Ni proceeds in three
temperatures; more details are discussed in Section
stages. The first is the chemisorption of oxygen and
1.10.2.2. For iron, the lower temperature rate is domi-
the formation of a two-dimensional oxygen surface
nated by the growth of magnetite (Fe3O4), while, at
structure, which is different for different Ni orienta-
high temperatures, above 570 C, the rate is mainly
tions. The second involves nucleation and lateral
dominated by the growth of the very nonstoichiometric
growth of NiO. Nucleation occurs on sites that are
wustite (FeO); since deviation from stoichiometry
present at the onset of oxidation, but such sites were
decreases with increasing temperature, the activation
not identified; lateral growth takes place at the
energy is seen to diminish with temperature. Details of
periphery of nuclei by direct capture of oxygen from
iron oxidation are presented in Section 1.10.2.3.
the gas phase. Stage 3, which starts before the com-
pletion of stage 2, involves logarithmic thickening
of the oxide nuclei (up to their tested temperature
1.10.2 Oxidation of Nickel, Cobalt,
of 300 C). The nucleation and growth kinetics
and Iron
depend on temperature and the partial pressure of
oxygen, but not significantly on metal orientation.
The high temperature oxidation behaviors of Fe,
At higher oxidation temperatures, however, the
Ni, and Co are similar in that the oxide growth is
shape, size, and density of NiO nuclei have been
mainly controlled by the outward diffusion of cations
found to depend strongly on Ni orientation,15 in
because they form oxides that are metal-deficient at
agreement with later results16,17 showing that sub-
high temperatures and over a wide range of oxygen
strate orientation can strongly affect oxide grain size
partial pressures.4 From studies of nonstoichiometry,
and oxide-scale thickness within a wide range of
electrical conductivity, and self-diffusion coefficients,
oxidation temperatures, from 450 to 1200 C.
the predominant point defects in the major oxides
formed on these metals are determined to be metal
vacancies. Among these metals, iron oxidizes the fast- 1.10.2.1.2 Oxide structures
est, as iron oxide and, hence, has the highest concen- The structure of NiO can be either simplex (single
tration of defects and the highest diffusion rate. The layer), which consists of a layer of columnar grains, or
next lowest rate is for Co and then Ni (see Figure 1). duplex (double layer), which has a columnar outer
As Ni forms only one oxide, NiO, over a wide layer similar to that on the simplex scale, and an
range of temperatures and oxygen potentials,5 it is equiaxed, fine-grained, porous inner layer; examples
relatively the simplest system. Consequently, it has of these are shown in Figure 2. In both structures, the
often been chosen as a model to study the fundamen- columnar grains are textured,18,19 where the degree
tals of high temperature metal oxidation, where of texture becomes more apparent for thicker scales.
oxide growth is controlled by cation diffusion. In There have been experimental indications that the
this section, Ni oxidation is discussed first and in type of texture varies with Ni orientation,20,21 and
detail, because many concepts applicable to Ni are these differences may create different NiO grain
also relevant to the other metals, for example, initial boundary characters, thus affecting the oxidation
stage of oxidation, metal and oxygen transports, the rate.20 However, the mechanism by which a different
importance of oxide grain boundaries as transport texture develops and exactly how it affects transport
paths, the generation of oxidation stresses and the through the oxide is not yet clear. Even with an
development of duplex structures. Specific character- apparently single layer of columnar NiO, TEM stud-
istics for Co and Fe oxidation follow. ies22 have found that a layer of fine-grained oxide
with differently orientated grains can exist at the
oxide–metal interface, where the oxide grains are
1.10.2.1 Nickel
only submicron in size. The reason for their for-
Nickel forms only one oxide, NiO, over a wide range mation is not clear; it is also not known if such a
of temperatures and oxygen potentials,5 and much is layer exists under all oxidation conditions. The
consequence of having this layer of fine-grained It is not known if the fine-grained oxide layer at
oxide at the interface may be to relax the growth the oxide–metal interface is the prelude to the thicker
and even the thermal stresses, since these fine grains inner scale that is observed in a duplex structure.
may exhibit superplasticity. The thickness ratio of the inner and outer layers has
been noted to depend on the oxidation temperature,
the surface preparation23 and, most importantly, the
metal purity. For a given oxidation condition, less
pure Ni is more prone to develop duplex scales than
high purity Ni24, 25; furthermore, the duplex structure
is always found on dilute Ni-based alloys.10,26
Smoother surfaces and higher temperatures tend
to decrease the duplex layer formation. Figure 3
illustrates the range of temperatures and times at
(a) which different microstructures of NiO develop on
high purity Ni.17 At 900 C or above (up to 1200 C),
the scale is one single compact layer, but, at tempera-
tures below 850 C, it can be duplex or a porous
single layer, depending on the oxidation time. In both
cases, the first formed scale is a single layer, but less
dense oxide forms at higher temperatures; with time,
the scale can become duplex, where an inner, more
fine-grained layer develops. Greater amounts of
porosity are often found in the inner layer of the
(b) 10 μm duplex structure. The mechanisms by which they
develop are not clear, although several have been
Figure 2 Typical NiO microstructure (a) duplex, columnar proposed; a review of these proposals can be found
outer and fine grained, porous inner layers formed on normal
purity Ni, 99.6–99.7%. (b) Single columnar layer formed on
in the article by Kyung and Kim.27 Some of these
high-purity Ni, >99.95%. Reproduced from Hou P. Y.; mechanisms attribute oxide growth stress as the cause
Cannon, R. M. Mater. Sci. Forum 1997, 251–254, 325–332. for the onset of a porous, and eventually duplex, scale
1200 10
μm Simplex
compact
1100
5μ
1000 m
Temperature (°C)
900 30
μm
800
Duplex
700
1 μm
600
Simplex
500
porous
400
1E + 02 1E + 03 1E + 04 1E + 05 1E + 06 1E + 07
Oxidation time (s) (280 h) (2800 h)
Figure 3 Formation of different NiO microstructures in relation to oxidation temperature, time, and oxide scale
thickness. Oxidation was carried out on high purity Ni (>99.999%) in oxygen, after polishing to a 0.25 mm diamond surface finish
and annealing at 1350 C in Ar–5% H2. Reproduced from Peraldi, R.; Monceau, D.; Pieraggi, B. Oxid. Met. 2002, 58, 249–273.
formation28–30; others point to the importance of the rates and lower activation energies than those at the
inward transport of oxygen through the initially high temperatures. These differences may be due to
formed single layer.31,32 For whatever reason poros- differences in sample purity, orientation and surface
ities are present in the oxide, a quasiequilibrium treatment, all of which have been reported to affect
model of gaseous transport through these pores was NiO growth rates. Impurities from the Ni, or minor
developed to explain the formation of the inner alloying elements incorporated into the NiO, can
layer.33,34 In the case of Ni oxidation, it was suggested change its defect concentration and, hence, affect the
by Caplan et al.35 that the duplex structure was also Ni cation transport rates.42,43 Experimental measure-
related to carbon impurity in the Ni. It was found that ments,44 however, indicate that impurities often
carbon, segregated at Ni grain boundaries, could decrease the grain boundary diffusion rate of Ni.
induce grain boundary cavitation as a result of Thus, it would appear that the effects of impurities
metal creep under the oxide growth stress. These in increasing the oxidation rate of Ni most probably
cavities were then incorporated into the oxide as result from an altered oxide microstructure. Indeed,
the oxidation front progressed inward. In nondecar- the most significant way the oxidation rate can be
burized Ni, cavities that formed inside the oxide affected seems to be a change in the microstructure
contained C-containing gases, so CO–CO2 was sug- of the NiO scale, from simplex to duplex,12 where the
gested to act as a carrier gas for the inward transfer of inner layer of the duplex scale is often fine-grained
oxygen across these cavities. and porous.
Although the overall growth of the NiO scale is
1.10.2.1.3 Growth rate and transport largely controlled by the outward diffusion of Ni
processes through NiO through lattice or grain boundary diffusion via nickel
The growth of NiO at 700–1300 C generally follows vacancies,2 a noticeable degree of oxygen diffusion
parabolic kinetics, although, usually, these are only through the scale is often observed. As seen from the
exactly parabolic at temperatures above 1000 C. oxygen isotope studies conducted by Atkinson et al.7
Reported parabolic rate constants have recently been (Figure 4), the thicker the inner layer, the more
complied by Peraldi et al.36 over a wide range of inward penetration of oxygen, and the inner layer
temperatures, with data from different surface pre- thickness is directly related to Ni purity. Gaseous
parations, metal purities and orientations and partial transport of oxygen is believed to occur through a
pressures of oxygen. There has not been enough work porous oxide, whether in a porous simplex scale or in
done in the low-temperature range (<500 C), and a porous inner layer in a duplex scale and, hence,
existing data show too much discrepancy to warrant a contributes to, and increases, the oxidation rate. The
clear understanding of the behavior in this tempera- oxygen tracer studies show that most of the inner
ture regime. Oxidation rate values at very high tem- layer growth takes place at or near the scale–metal
peratures (>1000 C) have much less scatter than at interface via oxygen inward transport through the
intermediate temperatures (from 500 to 1000 C). scale.37 The path for oxygen transport, particularly
Furthermore, the activation energy for all the high through the outer layer, has been suggested, by
temperature data are similar, ranging between 180 Atkinson et al.,37 to be microcracks or fissures, in the
and 240 kJ mol1, which is similar to the activation scale caused by growth stresses. The conclusion was
energy for Ni lattice diffusion, at 154–254 kJ mol1. made simply because the rate of oxygen transport is
Tracer studies by Atkinson et al.37 have also shown too fast to be accounted for by solid state diffusion.
that NiO growth above 1000 C is controlled by Ni The topic of stresses in thermally grown NiO can
lattice diffusion, but the growth is dominated by be dated back to 1947 when Evans45 reported the
Ni grain boundary diffusion below this temperature.38 observations of buckling and curling of NiO scale
In the intermediate temperature range, a decrease after its removal from the underlying Ni substrate.
in the rate constant is often observed with oxidation This indicated the presence of a compressive residual
time,38–40 which can be attributed to a decreased stress in the NiO film, and a possible stress gradient
grain boundary area due to grain growth,41 attesting through it, with higher compression at the scale–
to the tracer studies38 that oxide growth is dominated metal interface.46 Later studies47 further demonstrated
by grain boundary, rather than lattice, diffusion. Results the existence of a compressive growth stress by mea-
from different works in this regime displayed greater suring the degree of Ni elongation during oxidation.
scatter; some data showed apparently slower oxidation Others48 have come to the same conclusion by
18O
1.0
Ni
c/cmax
0.5
18O
0
(a) NiO/Ni NiO/gas
interface interface
18O
1.0 Ni 1.0 18O
Ni
c/cmax
c/cmax
0.5 0.5
18O
18O
0 0
x x
(b) NiO/Ni NiO/gas (c) NiO/Ni NiO/gas
interface interface interface interface
Figure 4 Ni and 18O traces through NiO scales formed at 1000 C on (a) {100} Ni crystal, (b) 99.998% high-purity
polycrystalline Ni, and (c) 99% low-purity polycrystalline Ni. Each specimen was first oxidized in 18O2 then in equal
amount of time in 18O2, under 50 torr of total oxygen pressure. (After Atkinson, Taylor and Goode7).
studying the flexure of thin strips of Ni samples as a these fast gaseous transport paths, if they do exist,
result of oxide growth. More recent studies using unless microstructures can be studied in situ at the
in situ X-ray diffraction techniques, however, have oxidation temperatures.
found the growth stress to be slightly tensile49–51 at
900 C, but compressive at lower temperatures, that
1.10.2.2 Cobalt
is, 800 C52 and 600 C.51
An implication consistent with the stress studies Three cobalt oxides exist, namely Co3O4, Co2O3, and
seems to be that the first formed simplex scale is CoO, and have respectively the spinel, hexagonal,
under compression, but the growth stress in duplex and cubic structures. Co2O3 is unstable above
NiO scales is mildly tensile near the outer surface 300 C, so is not found in oxidation products of Co,
and strongly compressive near the scale–metal where tests are usually conducted between 400 and
interface, with the total net stress still being com- 1400 C.5 From the results reported in the review by
pressive.25 The tensile stress in the outer duplex Wood et al.,5 average activation energies are calculated
layer may assist microcrack formation, although to be 107 26 kJ mol1 between 500 and 800 C, and
direct observations of these gaseous transport paths 213 47 kJ mol1 between 800 and 1400 C. Clearly,
have not been made. Acoustic emission studies con- there is a sharp, but unexplained, increase in oxidation
ducted at 800 and 900 C53 have shown some activ- rate with temperature above 600–700 C. Kofstad2
ities during Ni oxidation, indicating possible has compiled some of these oxidation rates, between
cracking of the oxide. However, the experiment 800 and 1150 C, and noted yet another change in
was done on a thin film of Ni that was allowed to activation energy at about 950 C, which is associated
bend during oxidation, in order to determine the with the formation of Co3O4. The activation energy
level of growth stress. It is unclear how the bending for self-diffusion of Co in CoO in the range 800–
of the sample under the oxide growing condition 1300 C is 144 or 160 kJ mol1,4,54 which lies in the
affected the AE signals. Since oxidation is a dynamic range of oxidation activation energies. These results
process, where cracks can form and heal at any time suggest that CoO growth is controlled by cation
within the scale, it may be impossible to identify movement via vacant lattice sites.
Above 900–950 C, only CoO is formed, but

1300
below this temperature, Co3O4 can exist as a thin
layer at the scale–gas interface above the CoO. The 1200 Fe
FeO
proportion of Co3O4 in the scale seems to increase
Temperature (⬚C)
with decreasing temperature, from 800 to 600 C.55 1100
Fe3O4
The relative thickness of the two layers is a function
1000
not only of temperature, but also of the partial pres-
sure of oxygen. These dependencies, from 500 to 900
900 C, have been summarized by Hsu and Yurek.56
Both layers grow by cation outward diffusion, with 800
probably lattice and grain boundary diffusion dom- 700 Fe2O3
inating in the CoO and Co3O4 layers respectively,
due to the difference in grain size between the two 600
–20 –18 –16 –14 –12 –10 –8 –6
layers. The growth rate of CoO at 600 C has been Log PO2 (atm)
found to be about five times greater than that of
Co3O4,57 so the overall oxidation rate is usually con- Figure 5 Regions of stability for different iron oxides as a
sidered to be controlled by the growth of the CoO function of oxygen partial pressure and temperature.
Reproduced from Kofstad, P. High Temperature Corrosion;
layer. It should also be noted that, once a complete
Elsevier Applied Science: London, UK, 1988.
Co3O4 layer exits at the surface, the overall oxidation
rate becomes independent of the ambient oxygen
pressure, as the partial pressure of oxygen at the and above 570 C, the oxide scale is multilayered,
CoO/Co3O4 interface is given by the decomposition consisting of hematite (Fe2O3) at the scale–gas inter-
pressure of Co3O4 in equilibrium with CoO. face, magnetite (Fe3O4) in between and wustite (FeO)
Like NiO, the CoO layer can be single- or double- at the scale–metal interface. Below 570 C, wustite is
layered.55 It also appears that a double-layered struc- unstable, so the scale only consists of Fe2O3 and
ture is favored with alloying addition in dilute Co Fe3O4. While these observations are true for bulk
alloys,58 while pure Co tends to form a single layer.59 scales, FeO has been found to be stable in very thin
Pt markers placed on the sample surface prior to films at temperatures down to 400 C63 and within
oxidation reside at the inner–outer oxide layer inter- narrow cracks at 500 C.64 A cross-section of a scale
face,55,60,61 suggesting that, analogous to NiO, the formed at 525 C is given in Figure 6, where the scale
outer layer grows by the outward movement of cobalt consists only of Fe3O4 and Fe2O3, and the Fe3O4 layer
ions and the inner one by the inward diffusion of is seen to consist of two different layers.65 Similar to
oxygen gas or anions. More recent work using the NiO and CoO, duplex Fe3O4 or FeO layers are often
18
O isotope62 has confirmed these earlier studies; observed, and the amount of porosity in them
between 1000 and 1300 C. The CoO formed on increases with oxide thickness.
dilute Co–Cr alloys develop duplex structures, where Wustite is metal deficient, Fe1 yO. The nonstoi-
the inner layer thickness increased with increasing chiometry, y, is a function of the partial pressure of
Cr content. Growth of the scale is dominated by cation oxygen and temperature. Between 800 and 1250 C, it
outward transport, but oxygen inward diffusion varies from 0.05 at the iron–wustite phase boundary
becomes more apparent with a thicker inner layer. to about 0.1–0.15 (higher at higher pO2) at the
wustite–magnetite boundary.66–70 Unlike CoO and
NiO, the deviation from stoichiometry decreases
1.10.2.3 Iron
with increasing temperature at a constant partial
The rate of iron oxidation is governed by the stabi- pressure of oxygen, which means that the enthalpy
lities of its various oxide phases, FeO, Fe3O4, and of defect formation is negative. Magnetite has the
Fe2O3, which are in turn a function of the oxidation spinel structure. At pO2 values above the stoichio-
temperature and the ambient partial pressure of oxy- metric composition, the oxide is metal deficient
gen. The temperature and pO2 ranges at which each (Fe3 yO4), where y increases with increasing pO2
of these oxides are stable is shown in Figure 5 so, at and decreases with increasing temperature, as occurs
900 C and 1018 atm, for example, iron will not be for wustite. At lower pO2 values, the oxide has a cation
oxidized, because none of its oxides is thermodynam- excess, with the dominant defect being iron ions on
ically stable under such a condition. In air or oxygen interstitial sites.71 Because of these properties, Fe
Fe2O3
Oxide
Fe3O4
Fe3O4
Metal 2 μm
Figure 6 Focus ion beam image of ion milled cross-section of oxide formed on iron after 1 hr oxidation in oxygen at 525 C.
The oxide is seen to consists of three layers, with the top being hematite and the bottom two magnetite. Reproduced from
Jonsson, T.; Jardnas, A.; Svensson, J. -E.; Johansson, L. -G.; Halvarsson, M. Oxid Met. 2007, 67, 193–214.
diffusion in magnetite as a function of pO2 shows a Fe FeO Fe3O4 Fe2O3 O2

minimum near its stoichiometric composition for each
temperature.72 Among the three iron oxides, Fe2O3 has
the smallest degree of nonstoichiometry.
Both FeO and Fe3O4 grow predominantly by Fe2+ Fe2+ Fe3+
cation transport of Fe2þ in FeO and Fe2þ and Fe3þ
Fe3+
in Fe2O3. The dominant transport path in FeO O2−
seems to be Fe lattice diffusion,73 but oxide grain
boundaries have been shown to be more important
for Fe3O4 below 600 C, and at all temperatures for e− e− e−
Fe2O3.74–76 The measured activation energy for
Fe2O3 growth is about 169–222 kJ mol1,77 which is
much lower than that expected from measurements
of tracer diffusivities of iron or oxygen.78 Oxygen 1 2 3 4
tracer diffusion studies on the growth of Fe2O3 dur-
ing the oxidation of iron at 550 C have shown that, 1. Fe = Fe2+ + 2e−
while the primary diffusion species through the oxide 2. 4FeO = Fe2+ + 2e− + Fe3O4
is Fe ions, fast, inward oxygen diffusion occurs down 3. 3Fe3O4 = Fen+ + ne− + 4Fe2O3
cracks in the scale.79 A simplified schematic diagram 4. Fe2O3 = 2Fe3+ + 6e− + 3/2O2
for the diffusion-controlled growth of multilayered 1/2O2 +2e– = O2–
scales on iron above 570 C is illustrated in Figure 7, Figure 7 Schematic illustration of the oxide layers formed
and the reactions that occur at each interface are upon iron oxidation above 570 C, the dominant diffusing
noted. The range of oxidation rates, in an Arrhenius species through each layer and the reactions that take place
plot, can be found in Figure 1. Above 570 C, a at each interface are indicated.
distinct break associated with the formation of FeO
in the scale occurs. The activation energy above this Fe3O4,80 where a higher proportion of Fe3O4 exists
temperature is not constant, because of the large on the lower index planes close to (100), (110), and
change in stoichiometry of FeO with temperature. (111), than on the higher index planes. The initial
A strong orientation relationship exists between oxide composition and thickness are also found to
the iron surface and the composition and thickness of depend on surface preparation and the presence of
the initial oxide that forms. The first-formed oxide impurities81; for example, on cold-worked surfaces,
layer, at around 200 C, consists of Fe2O3 and the scale tends to contain less Fe3O4.82 Between 350
and 500 C, Fe3O4 nucleates first, and then grows high to completely preclude their use. Alloying has
laterally over the entire surface.83 Subsequent growth improved the high temperature strengths, as well as
of the Fe3O4 layer follows parabolic kinetics, where the oxidation resistances, allowing refractory alloys to
the rate is dominated by the diffusional transport of be used in a number of high temperature structural
Fe2þ and Fe3þ ions through oxide grain boundaries. applications, principally in areas of propulsion and
With time, and if the pO2 is sufficiently high, Fe2O3 energy conversion, such as rockets and reentry sys-
nucleates in the Fe3O4 surface.84 When the Fe2O3 tems.91 Among the five metals listed above, Nb, Ta,
overgrows to form a continuous layer, the overall Mo, and W have received greater attention in research
oxidation rate is substantially decreased, because the and have wider industrial applications; therefore, these
growth rate of Fe2O3 is slower than that of Fe3O4, and will be the ones discussed here.
growth of the Fe3O4 layer is decreased due to the
lower effective pO2 at the Fe3O4/Fe2O3 interface. 1.10.3.1 Oxidation of Ta, Nb, Mo, and
Therefore, the oxidation of iron often shows an W Metals
unusual dependence on oxygen pressure, with a
Although these are simple metals, their oxidation
higher overall oxidation rate at lower pO2 values,
behaviors are very different and more complicated
due to a slower rate of Fe2O3 nucleation, so a com-
than those of Fe, Ni, and Co described in the previous
plete surface layer of Fe2O3 is achieved less easily to
section. This is because in addition to the usual com-
help impede the faster growth of Fe3O4.
plications associated with oxidation, such as oxide
It is well known that most technical steels oxidize
nucleation, growth and diffusion processes in the
faster in the presence of water vapor than in dry air.
oxide and the metal, these metals can have relatively
The reasons are currently under intense investiga-
high oxygen solubility, and large numbers of oxides
tion. For unalloyed iron, a strong water vapor effect
that can form under different temperatures and oxygen
is not always observed. The difference in behavior
pressures and, in some cases, are volatile, especially
between dry and moist environments seems to be
oxides of Mo and W. The equilibrium vapor pressures
affected by the oxidation temperature,85,86 as well as
of some refractory metal oxides at pO2 ¼ 1 atm are
the experimental procedure and the purity of the
summarized92 in Figure 8. Most studies on the oxida-
samples.87 When a water vapor effect was observed,
tion of refractory metals, including reaction mechan-
it appears that the oxide scale is more plastic when it
isms, metal recession rates and their pO2 dependence,
was formed in the presence of moisture.85,87
were performed prior to 1970, and a thorough descrip-
tion can be found in Ref. 2. This section only briefly
summarizes the unique oxidation behavior of these
1.10.3 Oxidation of Refractory Metals
metals and discusses their alloys and coatings or sur-
and Their Alloys
face treatments that have been attempted to improve
oxidation resistance.
Refractory metals are by definition metals with high
melting points. If this were the only criterion, that is, 1.10.3.1.1 Tantalum and niobium
melting point greater than 1925 C,88–90 eleven Tantalum forms only one thermodynamically stable
metals would fall into this category. Other definitions solid oxide, Ta2O5. However, nucleation of this oxide
are more restrictive and include the requirements appears to be difficult, especially at lower tempera-
that the metal has a body-centered-cubic (bcc) crys- tures and oxygen pressures. As a result, oxidation of
tal structure and forms oxides whose melting points Ta during the initial stage involves extensive dissolu-
are lower than that of the metal.89 These are niobium tion of oxygen in the metal and the formation of
(Nb), tantalum (Ta), molybdenum (Mo), tungsten different metastable TaOx type oxides at different
(W), and rhenium (Re), in increasing melting points temperatures. Between 300 and 500 C, the oxides
that range from 2468 to 3180 C. Although the grow in the form of platelets on the metal surface
high melting points make these metals attractive and extend into the metal along preferred crystallo-
candidates for high temperature applications, their graphic orientations.93,94 As temperature increases,
usage has been limited due to several problems, Ta2O5 begins to nucleate on the surface platelets but,
such as insufficient yield strength at high tempera- below 800 C, it is not able to develop into a dense
tures, difficulties in fabrication and poor oxidation surface layer; reaction is believed to be controlled by
resistance. Even at 300 C, oxidation can be significant oxygen-chemisorption equilibrium and the nucleation
(Figure 1); above 1100 C, the rates are simply too and growth of Ta2O5, leading to a porous surface film
Temperature (⬚C)
1600 1400 1200 1000
1E–02
W4O12
1E–04
MoO3
1E–06 VO2
pMeO(g), atm
1E–08
SiO2
1E–10
1E–12
NbO2
1E–14
TaO2
1E–16
1E–18
0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85
1000/T (K)
Figure 8 Equilibrium vapor pressure of some refractory metal oxides and silicon dioxide at pO2 = 1 atm. Reproduced from
Ramberg, C. E.; Beatrice, P.; Kurokawa, K.; Worrell, W. L. In High Temperature Silicides and Refractory Alloys; Proceedings of
the Matererial Research Society Symposium, Boston, MA, Nov.–2 Dec. 1993; Briant, C.L., Petrovic, J. J., Bewlay, B. P.,
Vasudevan, A. K., Lipsitt, H. A., Eds.; Material Research Society: Pittsburgh, PA, 1994; pp 243–253.
and a linear oxidation rate.94 A consequence of oxygen oxygen dissolution during the initial stage and oxide
dissolution into the metal is metal embrittlement, film formation afterwards. The stress distribution
which can be significant and should not be ignored.95 across the scale to the metal region affected by oxy-
Above 800 C, between 1 atm to 0.01 torr O2, a gen dissolution cannot be determined from the bend-
complete Ta2O5 layer can develop after an initial ing test, but tensile stress can develop in the oxide
incubation time that is associated with the nucleation film at some stage if there is compression in the
and growth of Ta2O5 in similar ways as at lower substrate near the oxide–metal interface, and this
temperatures.96 Growth of the first formed Ta2O5 tensile stress can easily induce scale fracture. Under
layer exhibits parabolic kinetics, which should be compression, oxides are less easy to crack, but scale
dominated by oxygen inward transport through the failure, especially by buckling of the surface film, is
scale. This conclusion, although made only from the common under biaxial compression. The fact that
morphology of the scale, is consistent with the oxida- blisters were found on the oxides formed on Ta and
tion pressure dependence of the rate constants.96 Nb99 suggests this kind of failure.
Since conductivity experiments show that Ta2O5 Above 1050 C, the post-parabolic oxidation rate
exhibits p-type conductivity at pressures close to is no longer linear, but decreases with time,100 proba-
1 atm O2 and n-type conductivity at lower oxygen bly due to increased sintering of the oxide grains that
pressures,97 it was concluded that transport through densify the film, and enhanced plasticity to relieve
the compact Ta2O5 predominantly occurs via oxygen some of the growth stresses that cause cracking. At
vacancies at lower pO2 and by interstitials at higher temperatures greater than 1200 C, however, reaction
pO2. After the short parabolic period, the oxidation rates at 1 atm oxygen can be so fast that severe sample
rate becomes linear.96 This behavior was interpreted heating or even ignition can occur. Above 1500 C,
to involve a repetitive, continuous, and probably sta- evaporation of TaO and TaO2 becomes important.
tistically distributed cracking of the scale down to The oxidation behavior of niobium is in many ways
the metal, whereby exposing the metal–oxide inter- analogous to that of tantalum, but the behavior is even
face beneath the cracks to oxygen. Oxygen then che- more complex due to the fact that Nb forms several
misorbs on the freshly exposed metal surface and stable oxides: NbO, NbO2, which has the rutile struc-
reaction rates again become dominated by the nucle- ture, and several polymorphs of Nb2O5. During the
ation and regrowth of Ta2O5. Indeed, substantial initial stage, oxidation proceeds principally by dissolu-
amounts of compressive growth stresses have been tion of oxygen in the metal, and this is subsequently
found during oxidation up to 550 C by deflection of followed by the formation of NbO, NbO2 or a mixture
Ta, as well as Nb.98 The sources of the stresses are of both. At higher temperatures, 450–500 C, a
surface layer of NbO2 forms at low pO2 (104 to that desorbs above 900 C into predominantly vol-
105 torr)101,102 and Nb2O5 develops near atmospheric atile WO2, with very small amounts of WO and
pressures.103,104 Both types of oxides can be protective WO3.106 WO3x forms in the temperature range
for a period of time, showing parabolic growth rate 568–908 C, and its growth rate has been expressed
controlled by the inward diffusion of oxygen, but the as: D ¼ 6.83 102 exp (29 890/RT), with the acti-
surface layer eventually breaks down, giving rise to vation energy given in calories per mole.107 Growth of
linear kinetics, where the rate is controlled by oxygen the WO3x film was determined to be dominated by
adsorption and the nucleation and growth of Nb oxide oxygen diffusion via oxygen vacancies. At higher tem-
at the oxide–metal interface. Usually, oxidation kinetics peratures (1100–2900 C),108 the oxidation rate is
are parabolic to paralinear over the temperature range dominated by the formation of a variety of volatile
400–600 C, but linear from 700 to 900 C.105 Like Ta, products, namely WO2, W2O6, WO3, and W3O9,
oxidation can only be studied in low pO2 environments where the polymeric oxides, (WO2)2 and (WO3)2,
above 1200 C, because the rate is too rapid near atmo- are believed to form through the interaction between
spheric pressure.102 At and above 1500 C, oxidation is WO2 molecules. Due to the extensive volatilization,
affected by the presence of liquid Nb2O5, which melts the oxidation of tungsten is dominated by the adsorp-
at 1490 C; at even higher temperatures, evaporation of tion and dissociation of oxygen, the surface interaction
Nb2O5 becomes increasingly important. of oxygen atoms with the metal and the subsequent
desorption of tungsten oxide.109 The reaction rates are
1.10.3.1.2 Tungsten and molybdenum a strong function of not only temperature, but also
When W is heated in the temperature range 200– oxygen pressure, as illustrated in Figure 9, where the
700 C in oxygen, it grows a multilayered oxide film experimental data are those from Bartlett.110
⬚C
3410
3200
2800
2400
2000
1800
1600
1400
1
0.5
0.2
7.6 torr
Surface recession (mg cm–2 min–1)
0.1
0.05
~24.5
0.02 kcal mol–1 ~44 kcal mol–1
0.076 torr
0.01
0.005
0.002
0.001
2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5
104/T (K)
Figure 9 Tungsten surface recession rate as a function of temperature under high and low oxygen partial pressures.
Reproduced from Kofstad, P. High Temperature Corrosion; Elsevier Applied Science: London, UK, 1988.
Under low oxygen pressures, generally below temperatures.117 Dilute Nb–Hf and Nb–Zr alloys,
104 atm,111 the oxidation rate is governed by often considered for applications where strength and
surface processes or reactions; above 1300 C, the resistance to alkali metals are required, have also been
oxide evaporates as soon as it is formed, and the studied.119,120 Although the oxidation rate in air can be
metal remains free of surface oxide during oxida- decreased with the addition of Hf or Zr, the benefit is
tion.112 The metal recession rate is slower at higher not significant. Furthermore, internal oxide precipi-
pressures, because a surface boundary layer exists, tates of Hf and Zr develop, and they may further
which limits the rate of oxygen arrival at the metal embrittle the alloy. Careful heat treatments can control
surface. Furthermore, the surrounding gas acts as a the size, morphology, and distribution of the internal
blanket that reflects escaping molecules back to the particles, allowing them to produce a strengthening
surface; the fraction of this back-reflection process is a effect in the alloy,119 but this type of controlled mor-
function of the total gas pressure of the system. Con- phology is difficult to maintain in long-term applica-
sequently, metal recession rates are slower than tions. Also, protective surface oxides do not form
expected from surface reactions alone. Furthermore, on alloys such as Ta–8W–1Re–0.7Hf–0.25C and
the high pressure regime is affected by the gas flow Mo–36Re (all in wt%); with the Mo–Re alloy, weight
rate, because the thickness and effect of the boundary gains associated with the formation of MoO2 were
layer are decreased with higher flows. found at temperatures above 400 C for PO2 < 103 Pa
The oxidation of Mo is similar to that of W. At (105 torr) in vacuum, but weight losses took place
190 C in air, molybdenum is coated with an oxide with evaporation of MoO3 at 102 Pa.121 Oxygen dis-
film, which hinders further oxidation up to 350 C. solution and internal oxidation were also reported to
Over 350 C, oxidation recommences. At 500 C, the affect the mechanical property of these alloys.122 The
rate increases quickly as MoO3 sublimates inten- oxidation rate of Nb in air from 800 C to above
sively, exposing the metal. When oxidized in air, 1000 C can be decreased by alloying with Hf, Zr, W,
molybdenum oxides show valencies of 4 and 6.113 Mo, Ti, or Ta.123 However, the preferred fabricable
Oxidation behaviors at higher temperatures are dic- alloys still require further protection by coating.
tated again, as with W, by volatilization of oxides. The The development of oxidation resistant Nb alloys
principal products are the gaseous dioxide and triox- that can be used for aerospace applications had
ide, MoO2 and MoO3, respectively. Between 1200 received considerable interest,124 since Nb is less
and 2300 C,114 the trioxide was found to be the more dense but stronger than Ni-based superalloys at
important species at lower temperatures, and the high temperatures. Attempts have been made by
dioxide was the dominant product at higher tempera- alloying to promote the formation of an Al2O3
tures. At the lowest temperatures and highest oxygen scale as a protective layer. Svedberg125 conducted an
pressures, small amounts of polymeric oxides, (MoO2)2 extensive investigation of the oxidation resistance of
and (MoO3)2 were also detected. Nb-alloys in air at 1200 C and found the slowest rate
for NbAl3, which developed an alumina inner layer
adjacent to the metal–oxide interface and an NbAlO4
1.10.3.2 Oxidation of Refractory Alloys
outer layer at the oxide–gas interface. The oxidation
The focus of this section is not to review the effect of rate was approximately parabolic, but the rate con-
alloying additions on the oxidation of refractory stant was still 2 orders of magnitude higher than that
metals, but to discuss advances made that can improve of NiA1, which forms a protective alumina scale in
the oxidation behavior and, hence, the usefulness of O2 at 1200 C. Perkins and Meier126 have also studied
this class of material. factors affecting the selective oxidation of Al and the
Some of the investigators, who studied the oxidation formation of protective alumina scales on Nb–Al
behavior of refractory metals prior to 1970, also exam- alloys. It was shown that protective alumina scales
ined the oxidation mechanism of some refractory are not formed on NbAl3 in air at 1350 C due to the
alloys. For example, W addition to Ta has been studied formation of an Al-depleted Nb2Al layer at the scale–
to evaluate its effect on Ta oxidation; the amount alloy interface. Protective alumina scales can be
of W added varied from 10 wt%115–117 to 50 wt%.118 formed on Nb–Al alloys with 37.5–50 at.% Al in air
In most cases, no significant beneficial effects were at 1400–1600 C by adding Ti to increase the solubil-
found, except at 1200 C, where 50 wt% W was rep- ity and diffusivity of Al and adding Cr and/or V to
orted to decrease the Ta oxidation rate by nearly a decrease the solubility–diffusivity product of O in the
factor of 10, but the effect quickly diminished at higher alloy. An alloy with 25.4Nb–29.1Ti–2.8Cr–3.5V–39.2Al
(wt%) was developed and was able to form a protective a small mismatch in the thermal expansion coefficients
alumina scale in air at temperatures above 1000 C that between the base metal and the coating can lead to
was even stable up to 1400 C. Furthermore, this alloy cracks, especially when the coating is under residual
oxidized with kinetics comparable to those of NiAl.127 tension. The cracks can heal under some conditions,
However, the melting point of the alloy was relatively but often they allow oxygen penetration into the base
low, about 1600 C, which resulted in poor mechanical metal, which then oxidizes rapidly. Consequently, the
strengths at high temperatures. Attempts to raise the use of coatings to protect these metals is not always an
melting point were made by replacing Nb (melting adequate solution to the oxidation problem.
point 2500 C) with Ta (m.p. 3000 C), and led to Other types of coatings have continually been
an increase in the alloy melting point by 300 C. evaluated, but have showed little success. On tung-
However, the microstructures of the Ta–Al–Ti–V– sten, amorphous W–N–Ni coatings with different
Cr alloys are thermally unstable. Technical efforts N and Ni contents have been demonstrated to
have also been made to improve the strength and improve the oxidation resistance up to 800 C,135 by
oxidation resistance of the Nb–Al–Ti–V–Cr alloys forming two nickel-rich, external oxide layers of NiO
by adding dispersion particles.128 It was found that and NiWO4. The oxidation rate was controlled by the
Al2O3 and Y2O3 were thermally and chemically stable outward diffusion of either Ni2þ or W6þ ions through
with respect to Nb alloys, but SiC, Nb5Si3 and AlN the oxide layers. However, for the coating to remain
reacted extensively with the alloys at high effective, it has to stay amorphous. Crystallization
temperatures. induced a dimensional stress capable of destroying
Some efforts have also been carried out through the protective oxide layers. FeCrAl-, SiCrFe-, and
alloying of Cr to produce a refractory metal based MoSi2-based coatings have been deposited onto
alloy that is a good Cr2O3 former,129 but the results molybdenum with various techniques (plasma spray-
have not been successful, mainly due to the difficul- ing, slurry deposition and physical vapor deposition)
ties in controlling the alloy microstructure and also to different thicknesses with a range of varying ther-
from the formation of volatile oxides, especially those mal post-treatments.136 The best coating lifetime,
of W and Mo. however, was only a few thousand hours at 1200 C
In general, despite quite a bit of effort and some and a few hundred hours at 1450 C under isothermal
limited successes, successful development of high oxidation tests in air. Although a noticeable improve-
temperature refractory-based alloys with the ability ment from uncoated metal, durability has not been
to form a protective oxide scale over an extended proved for these coated systems for them to be used
period has not been achieved. Even though some of reliably in engineering applications.
these alloys have potentially useful properties, the Surface treatments have included laser alloying,
alloying elements usually lead to detrimental effects ion implantation and aluminizing, all of which aim at
in at least one crucial alloy property.130 These draw- modifying the surface composition of an alloy so that
backs have severely limited the applicability of it can develop a protective oxide scale, while not
refractory alloys. One approach to improve the oxi- changing the alloy bulk properties. Ion implantation
dation resistance has been to use coatings, and these of different species can have a beneficial influence on
will be briefly discussed in the next section. the thermal oxidation kinetics of niobium in pure
oxygen at temperatures below 500 C.137 Si and Al
were most beneficial. However, the treatment only
1.10.3.3 Coatings on Refractory Metals
delayed the appearance of the linear catastrophic
and Alloys for Oxidation Protection
kinetics, but showed no long-term effects. When Al
Coatings, whether applied as an external layer, or or Si powders were deposited onto niobium by laser
produced in the alloy surface region, have been used melting, forming aluminized or siliconized surface
to improve the oxidation resistance of refractory alloys. coatings, the effect was longer lasting, but the coat-
In some cases, they are very effective.131 These include ings still failed with time.138 Likewise, coatings pro-
MoSi2,132,133 niobium silicide, both in its pure form duced by pack cementation with Cr and Al also failed
and alloyed with Fe and Cr,130 (Mo,W)(Si,Ge)2,134 and under longer times and at higher temperatures; one
aluminides.127 The oxidation behavior of some of the of the reasons for the failure was interdiffusion, where
silicides is described in Section 1.10.4.2. Although the scale forming element is lost from the coating by
these coatings can, for the most part, provide adequate diffusing into the alloy, while alloying elements dif-
oxidation protection, they tend to be very brittle. Even fuse into the coating and change its oxidation
properties. Diffusion barriers have been tested to 1.10.4.1 Si-Containing Alloys

show some success. When a Re–Ni film that
Silicon additions to Fe, Co, and Ni-based alloys have
contained more than 70 at% Re was applied onto a
strong adverse effects on their mechanical properties;
Nb–5Mo–15W (wt%) alloy via electroplating, fol-
therefore, like Al, only small amounts can be added.
lowed by pack cementation with Cr and Al, the
Similar to aluminum, the presence of a third element
alloy was able to develop a protective Al2O3 scale,
in the alloy, such as Cr, can greatly enhance SiO2
with a Re–Cr rich s-phase layer in the inner layer of
formation, where a SiO2 healing layer is often found
the coating, which decreased the mutual diffusion
beneath either a Cr2O3 healing layer or an external
between the outer Al reservoir layer and the alloy
Cr2O3 scale. (An example of a partial SiO2 healing
substrate.139 However, the most severe condition
layer beneath an external Cr2O3 can be seen in
tested so far was isothermal oxidation at 1100 C for
Figure 10.) According to Stott et al.,147 healing layers
100 h in air. No cyclic oxidation tests have been
of SiO2 are harder to establish than Al2O3 because the
performed to evaluate the durability of the coating.
Gibbs free energy of formation (DG) for SiO2 and
Cr2O3 are closer than that between Al2O3 and Cr2O3
(125, 162, and 202 kcal mol1 O2 at 1000 C,
1.10.4 Oxidation of Silica-Forming respectively, for SiO2, Cr2O3, and Al2O3). The smal-
Alloys ler differences in G SiO2 and G Cr2 O3 give rise to a
lower driving force for SiO2 nucleation at, or near, the
Pure silica (SiO2) is one of the most impermeable scale–alloy interface, resulting in a lower density of
oxides at high temperatures. The low activation internal oxide precipitates, hence making the forma-
energy for the diffusion of oxygen in SiO2 makes tion of a healing layer more difficult.
it a more protective oxide scale than Al2O3 at tem- Experimentally, it has been suggested that SiO2
peratures above 1080 C (see Figure 1). However, forms as a complete layer beneath a Cr2O3 scale when
this rate can increase by a factor of 2–8 in the pres- a 20Cr–25Ni–Nb stainless steel containing 0.45–
ence of water vapor or certain impurities such as Na 0.75 wt% Si was oxidized at 750–900 C.148 However,
or Al.140,141 More than 15 different crystalline forms there was no strong evidence indicating that this layer
of SiO2 have been reported.142 Silica also exists in was not partial, but complete at the interface. Later
amorphous form (quartz glass). In both crystalline work on the same 20Cr–25Ni–Nb alloy oxidized at
and vitreous (fused) quartz, tracer studies have 825 C in CO2, has shown by TEM that the silica
shown that the diffusivity of oxygen is much faster layer at the interface was amorphous.149 The addition
than that of silicon.143 The diffusion rate of oxygen in of various amounts of Si to a similar steel, 20Cr–
fused quartz, is around 1013 cm2 s1 at 1200 C, and 25Ni–TiN, oxidized at 850 C in CO2-based gas con-
about 1.5 orders of magnitude lower at 1000 C.144,145 taining 2% CO showed a minimum parabolic rate at
Formation of volatile SiO, however, should limit the 0.9 wt% Si, and X-ray mapping of polished scale
application of SiO2-forming alloys in environments cross-sections showed the presence of a silica interlayer
of low oxygen activity.146 between the external Cr2O3 and the alloy.150 However,
Silica Chromia scale
Metal substrate
110nm
Figure 10 Cross-sectional transmission electron micrograph showing SiO2 healing layer beneath an external Cr2O3 scale
formed on commercial Ni–20 wt% Cr alloy containing 1.3 wt% Si, oxidized in air for 2 min at 950 C. Reproduced from Ahmad, B.;
Fox, P. Oxid. Met. 1999, 52, 113.
the decrease in alloy oxidation rate was no more than below 1000 C,149,154 but a crystalline form of either
2.5 times, and this rate was considered to be controlled cristobalite or tridimite developed at 1100 C,
by Cr diffusion through the SiO2 layer. For example, depending on alloy composition and oxidation
Cr2O3 internal oxides were often found beneath the time.156 In general, the amorphous phase is favored
silica interlayer, suggesting that the layer did not form at low temperatures and is the first formed phase,
continuously over the entire alloy surface. Similar while crystalline forms are favored at high tempera-
base metal cation transport through SiO2 layers was tures or after longer time periods. Transformation
also proposed for Fe–14 and –20 wt% Si alloys,151 from amorphous to the cristobalite phase involves a
where Fe seemed to diffuse through an apparently substantial volume change, which generates higher
complete SiO2 layer to form Fe2O3, resulting in a linear stresses within the silica layer that may contribute to
oxidation rate. the often observed preferential spalling.157
Whether the SiO2 layer, formed at the oxide–alloy In summary, even though SiO2 can provide effec-
interface, is continuous or intermittent is often con- tive protection against high temperature oxidation,
sidered to depend on the Si concentration in the Si added to Fe-, Co-, or Ni-based alloys is not able
alloy, as well as the oxidation time and temperature. to develop into protective SiO2 scales in the same
For example, Douglass and Armijo152 reported that way that Cr and Al can in forming Cr2O3 and Al2O3.
an almost continuous SiO2 layer formed after oxidiz- In most cases, the SiO2 may not be a complete layer
ing Ni–20Cr–1Si (wt%) at 1200 C for 2 weeks, while that acts as a diffusion barrier. Even if it were, base
Saito and Maruyama153 observed a continuous SiO2 metal cations and oxygen seem to diffuse rather rap-
layer on a similar Ni–20Cr–1Si alloy oxidized for idly through it, which may involve transport through
50 h at 1250 C. However, the method of detection short-circuit paths, such as cracks and pores.158 Fur-
can affect these results. While most works rely on thermore, the presence of a SiO2 layer, whether par-
microprobe or EDS/SEM analyses, where relatively tial or complete, often has a detrimental effect on
thick (close to a micron) layers are observed, much oxide scale adhesion, where preferential cracking and
thinner films can be detected by STEM. An amorphous spalling take place within this layer.
silica healing layer, a few nanometers thick, has been
found on a Ni–20Cr–1Si (wt%) alloy after only 5 min
1.10.4.2 Silicides
oxidation at 950 C in air.154 Upon oxidation, the initial
scale consisted almost exclusively of Cr2O3. Thin, par- Because of their high melting points and low den-
tial layers of silica were detected at the metal–oxide sities, silicides have long been considered for high
interface (Figure 10) after 2 min at 950 C, and they temperature applications, but their poor ductility is a
thickened locally with oxidation times. After >25 min, major drawback, which has hampered their develop-
these regions spread along the metal–oxide interface ment as high temperature structural components.159,160
until an almost continuous silica layer had formed. When a protective SiO2 scale can form, the silicides
Healing layers of silica have also been reported in should also possess adequate oxidation resistance. From
Ni–Si alloys.155 The addition of 1 wt% Si had little thermodynamic considerations, most Si-rich MxSiy
effect on the oxidation rate of nickel at 1000 C, but was silicides, where M ¼ Ti, V, Nb, Ta, Cr, Mo, W, Mn,
sufficient to establish a partial-healing layer of amor- Tc, Re, Fe, Ru, Os, Co, Rh, Ir, Ni, Pd, or Pt, are
phous SiO2. The layer was incorporated into the inner predicted to be capable of forming protective SiO2
part of the duplex NiO scale but did not react with the layers at temperatures greater than 900 C.161 However,
oxide. Increasing the silicon concentration to 4 or 7% like the Si-containing alloy discussed above, kinetic
facilitated the development of apparently continuous considerations ultimately determine whether or not a
amorphous SiO2 layers at the base of the scale, result- protective SiO2 scale will develop. Higher tempera-
ing in decreased rates of oxidation. However, these tures and longer oxidation times usually favor the for-
layers were unable to prevent the continued transport mation of a complete surface SiO2 layer; for example,
of Ni2þ ions into the NiO scale and oxygen into the SiO2 was found to cover the entire MoSi2 surface at
alloy. Furthermore, the SiO2 layers provided planes of temperatures greater than 1000 C, but at lower tem-
weakness that resulted in considerable damage under peratures, portions of the surface were covered by
thermal stresses during cooling, causing severe scale MoO3.162
spalling, where failure occurred within the SiO2 layer. The high temperature oxidation and corrosion of
The structure of the SiO2 that formed during high some silicides, namely MoSi2, Mo5Si3, TiSi2, Ti5Si3,
temperature oxidation seemed to be amorphous at or V5Si3, Cr3Si, and Fe and Ni silicides have recently
been summarized by Brady et al.163; emphasis was The rapid disintegration is usually preceded by
placed on Mo and Ti silicides, because their oxida- an incubation period,175 the duration of which is
tion behaviors have been most extensively studied. affected by microstructure, applied stress, tempera-
Readers interested in the details of silicide oxidation ture, and oxygen partial pressure. The susceptibility
should refer to this publication. Only brief summaries to pesting seems to depend on the material micro-
are given here. structure and stoichiometry. When the MoSi2 is
During the initial stages of silicide oxidation, oxides dense with small amounts of connected porosity,
of the metal and silica usually form simultaneously. and is free of cracks, pesting does not occur.176–179
For example, TaSi2 or NbSi2, upon oxidation, form An example of the pesting behavior and its sensitivity
mixed oxides of both silica and Ta2O5 or Nb2O5.164,165 to specimen porosity is seen in Figure 11, from
Diffusion across the surface scale at this point is Kurokawa et al.,179 where noticeable pesting of the
therefore not dictated by the transport through SiO2, porous sample took place shortly after 56 h at 500 C,
but through the mixed MeOx–SiO2 layer. Although leading to total disintegration of the sample after
the Nb and Ta silicides could eventually be protected 100 h, but weight gain of the dense sample was less
by a thick mixed-oxide layer, spallation problems than 2 103 g cm2 up to the end of the test, at
due to growth stresses become critical as scale thick- 111 h. The oxides that formed in various stages of
ness increases, and the oxide scales on TaSi2 and NbSi2 disintegrated MoSi2 are MoO3 and amorphous
can crack extensively.164,165 In some systems, the mixed SiO2.177,180,181 Pest disintegration appears to be
oxide layer can be rather protective. Several stud- caused by the formation of voluminous Mo oxides
ies166,167 have suggested that the oxidation kinetics of in extended defects. This introduces stresses that
TiSi2 are parabolic, due to the formation of a protective provide the driving force for fast crack propagation
scale that consists of a silica layer near the silicide–scale and fragmentation of the relatively brittle solid.
interface and islands of TiO2 at the scale–gas interface. It appears that susceptibility to pesting increases
Ti, which is soluble in SiO2 up to 10 at.%,168 appar- with Mo concentration for Mo:Si ratios >0.5 and
ently diffuses to the outer surface, forming islands of decreases when MoSi2 is Si-rich (68–70 at.%).181
TiO2 and leaving a protective silica–titania glass on the Mo–Si–B alloys, with melting points above
inside. In other systems, the metal oxide that coexists 2000 C, are being considered for high temperature
and continues to grow with the SiO2 may eventually applications, such as turbine blade materials.182
disrupt the protectiveness of the silica.169–171
Many of the metal oxides can be volatile under
some oxidation conditions, which complicates the
scale development process. Equilibrium vapor pres-
sures of some refractory metal oxides, together with
SiO2, are presented in Figure 8. Mo and W oxides, 400
for example, are highly volatile. As these oxides, 773 k
initially formed with silica on MoSi2 or WSi2, volatil-
Mass gain (⫻ 10–3 kg m–2)
ize, the scale layer becomes more silica-rich and oxy- 300
gen transport to the silicide–scale interface decreases.
This decrease in oxygen transport and the associated
decrease in the oxygen activity at the silicide–scale 200
interface continue as the metal oxides continue to
volatilize. Eventually, the oxygen activity at the inter- Porous
100 5 mm
face decreases to a value where the base metal oxide is
Dense
unstable, then the Mo or W silicide is protected by a
relatively pure SiO2 layer.172, 173 This process is favored
as the oxidation temperature is raised. At temperatures 0
0 100 200 300 400
>1200 C, volatilization is rapid, and the silica layer Oxidation time (ks)
can be purified within the first hour of oxidation.174
Pesting is often a problem with molybdenum sili- Figure 11 Oxidation kinetics of dense and porous MoSi2
at 500 C in air, and the appearance of the porous sample at
cides, where extensive oxidation disintegrates the four different times. Reproduced from Kurokawa, K.;
silicide to powder or fragments. For MoSi2, this Houzumi, H.; Saeki, I.; Takahashi, H. Mater. Sci. Eng. A
occurs in the temperature range 300–550 C.163 1999, 261, 292–299.
Several compositions of this alloy are able to develop steels have the base composition of Fe–Cr with
a protective surface borosilicate glass that offers good 13–18 wt% Cr to ensure the formation of a slow
oxidation resistance at high temperatures. The devel- growing scale of Cr2O3, or mixed Fe–Cr spinel and
opment of this glass layer is similar in principle to the Cr2O3, for corrosion protection. Common austenitic
oxidation of MoSi2. Above 800 C, the first formed steels contain higher concentrations of Cr (18–30
oxides are rich in MoO3, which volatilizes, while the wt%) and an austenite stabilizer, such as Ni or Mn.
silicon and boron oxides slowly form a complete layer The present chapter describes the development,
of borosilicate. Once this silicate layer is established, growth and adhesion of Cr2O3 scales observed for
the rate of alloy recession is decreased. However, Cr and model Cr2O3-forming alloys.
development of the protective silicate layer is very
sensitive to the oxidation temperature, alloy composi-
1.10.5.1 Transport Properties and
tion, particularly the B to Si ratio, and other factors.183
Oxidation Rates
Increasing the B/Si ratio decreases the viscosity of the
glass and allows it to cover the surface of the sample Cr2O3 is an electronic semiconductor.4 It changes
faster, but this also increases the oxygen diffusivity from a p-type to an n-type semiconductor from
through the glass and lowers the high temperature near-atmospheric oxygen pressures to oxygen activ-
oxidation performance. Furthermore, at low tempera- ities close to the Cr/Cr2O3 interface. There are evi-
tures, 650–750 C, the formation of a protective dences4 showing that Cr self-diffusion (mainly lattice
borosilicate layer is slow compared to the loss of Mo diffusion) depends on oxygen activity, and it can vary
by the formation and volatilization of MoO3, and this by 3–4 orders of magnitude under a given tem-
leads to pesting of the material.184,185 Current efforts perature. The dominant defect structures may be
in improving the oxidation resistance of these alloys Cr vacancies or interstitials respectively, at high and
involve the development of coatings.186 low oxygen activities.194 Earlier diffusion studies
by Hagel and Seybolt195 on sintered polycrystalline
Cr2O3 showed a three orders of magnitude higher Cr
diffusivity over that of oxygen. Compared with
1.10.5 Oxidation of Chromia-Forming Cr diffusion rates found in single crystal Cr2O3,196
Alloys which is nearly five orders of magnitude lower than
the rates reported by Hagel and Seybolt, the earlier
Chromium oxidizes at high temperatures to form one rates should be dominated by short circuit diffusion,
oxide, the corundum Cr2O3, which consists of hexago- mainly via oxide grain boundaries.74 Consequently,
nally close-packed oxygen ions with Cr cations occu- growth of Cr2O3 scales has often been considered to
pying two-thirds of the octahedral sites. The structure be dominated by cation outward transport, mainly
is the same as a-Al2O3 and a-Fe2O3. Cr2O3 reacts with through Cr2O3 grain boundaries.197 Location of inert
oxygen to form volatile CrO3 according to the markers subsequent to oxidation of essentially Cr2O3-
equation 0.5Cr2O3 þ 0.75O2 ¼ CrO3, at temperatures forming alloys seems to support this conclusion,198
above 850–900 C; the reaction rate increases with where markers that had been deposited on specimen
increasing partial pressure of oxygen in the atmo- surfaces prior to oxidation resulted near the scale–
sphere and with increasing gas flow rates187; it metal interface after oxidation, suggesting an outward
becomes rather pronounced at temperatures above growth mechanism.
1000 C,188 thus limiting the use of Cr2O3 forming More recent diffusion studies,199 utilizing oxygen
alloys to temperatures less than 900 C. In the pres- and chromium isotopes, 18O and 54Cr respectively,
ence of moisture, Cr2O3 reacts with water vapor to showed that the rates of O and Cr bulk diffusion are
form a number of volatile chromium hydroxides.189 similar, but Cr grain boundary diffusion is greater
The reaction rates depend on temperature, total than that of oxygen. The parabolic oxidation rate
pressure, partial pressure of water and the reaction constants calculated from these diffusion data agreed
gas flow rate. These volatilization reactions have been with experimentally determined values, but sug-
shown to cause severe problems in solid oxide fuel gested that Cr2O3 scale growth is in fact controlled
cell environments, where Cr2O3 forming alloys are by counter diffusions of oxygen and chromium along
being considered as a low-cost interconnect mate- oxide grain boundaries. Moreover, SIMS studies200
rial.190–193 The most common and extensively used on different areas of Cr2O3 showed different trans-
Cr2O3-forming alloys are stainless steels. Ferritic port properties on the same sample of Cr after it was
oxidized in natural oxygen (16O) and then, without be the relative adhesion of the Cr2O3 scale, resulting
cooling to room temperature, in oxygen enriched in in different degrees of scale buckling and convolution
18
O. Results showed that, in some places, the oxide during oxidation, or even local breakaway resulting in
grew primarily by chromium transport while, in small nodule formation (see Figure 13 for typical
others, it grew primarily by oxygen transport and, in microstructures). Buckling, driven by the compres-
still others, it grew by a mixture of both. It was sive growth stress in the oxide film,203 often occurs;
postulated that these apparent different transport an example is given in Figure 14. These buckles
processes were related to different degrees of high increase the amount of oxide per unit area of metal,
versus low angle oxide grain boundaries, presuming hence giving rise to an apparent increase in oxidation
that the two have different transport properties, rate. Continued scale growth beneath a buckle is
whether intrinsic or due to different degrees of impu- achieved by Cr vapor transport,204 which can be
rity segregation. faster than solid state diffusion through the scale.
Above 700 C in air or atmospheric oxygen, the Several other factors may also contribute to the scat-
kinetics of Cr2O3 growth are generally interpreted as ter. Formation of volatile oxide species, especially at
parabolic, although periodic increases in the reaction higher temperatures, can be a problem. Although
rates have been observed in some cases201; these will most studies corrected for the sample weight loss
be treated later under breakaway oxidation. The par- due to CrO3 formation, the possible effect of water
abolic oxidation rates on Cr metal prior to 1980 had vapor and the formation of other volatile species were
been compiled by Hindam and Whittle,202 with data not considered. Metal purity can also be a contributor
scattered between the two solid lines in Figure 12; all to rate variations. Dissolution of metallic impurities
of the experiments included here were conducted in in Cr2O3 can exert a doping effect, that is, changing
oxygen. Reported rates from more recent studies also its defect concentration and, hence, its oxidation
fall within this wide band of scattered data. It is rather kinetics. Nonmetallic impurities, such as S and C,
astonishing that such a large range of rates, over three have also been suggested to affect the transport rate
orders of magnitude, can exist for a simple metal. through Cr2O3.205,206 However, it is also possible that
The most likely explanation for these differences may these impurities caused a decrease in scale adhesion,
consequently, leading to increases in the rate by
forming more convoluted or buckled oxides. Differ-
10−9 ent surface treatments, such as electropolishing, etch-
ing or mechanical abrasion, have also been shown
to have profound effects on subsequent oxidation
rates,207 mainly due to the development of very dif-
10−10 ferent oxide grain sizes on these surfaces.208 Likewise,
sample pretreatment, such as annealing, cold work-
ing, rolling, etc., or the rate at which specimens are
Kg (g2 cm−4 s−1)
Cr alloys heated to the oxidation temperature can all have an

effect, since they can influence the development and
10−11
Cr2o3
Cr
10−12 m
et Cr
al (a)
10−13
1300 1200 1100 1000 900 800 (b) 30 μm

Temperature (⬚C)
Figure 13 Optical micrograph of polished cross-section
Figure 12 Range of parabolic rate constants from the of etched Cr oxidized for 42 h at 1200 C, giving examples of
oxidation of pure Cr (bonded between the solid lines) and localized oxide nodule, or ridge, formation on pure Cr.
binary Co, Ni and Fe–Cr alloys (the shaded area), complied Reproduced from Caplan, D.; Sproule, G. I. Oxid. Met. 1975,
by Hindam and Whittle.202 9, 459–472.
(a) 30 μm (b) 0.3 μm
(c) 3 μm (d) 2 μm
Figure 14 Examples of buckled Cr2O3 scale formed on commercial grade 99.8% Cr at 1000 C. (a) Scale surface and (c)
cross-section after oxidation for 30 min. (b) and (d) are metal surfaces after scale removal and oxidation for 5 and 10 min,
respectively.
grain size of the first formed Cr2O3. Although it 1E–07

appears that the grain size and microstructure of the
Cr2O3 scale dictates its growth rates, no clear corre-
lations can be identified between the rate and the
Rate constant (g2 cm–4 s–1)
various factors mentioned above.

The oxidation rates of Cr2O3 forming alloys of Ni,
Co, or Fe–Cr have also been complied by Hindam
and Whittle,202 and those data are also plotted, as a 1E–08
shaded band, in Figure 12. It is seen that, even with
different base metals, less scattering occurs during
alloy oxidation. This is probably related to better
Cr2O3 adhesion on alloys than on Cr. It is also possi-
ble that base metals incorporated into these Cr2O3
scales dominate other impurities that might influence
the oxidation rate. Both Ni and Co have small, but 1E–09
0 5 10 15 20 25 30 35
significant solubility in Cr2O3, while Fe2O3 and Chromium concentration (at.%)
Cr2O3 are completely isomorphous.
Figure 15 The change of oxidation rate, at 1000 C
760 torr oxygen, of Co–Cr as a function of alloy Cr
1.10.5.2 Cr2O3 Layer Development concentration. Data from Wright and Wood209 include
on Alloys parabolic rate constants from initial and steady states.
When a Cr-containing alloy is oxidized, its oxidation

rate usually changes with its Cr concentration, as illu- rate constant is rather unaffected by additional Cr
strated in Figure 15 for binary Co–Cr alloys.209 The until a critical Cr concentration is reached. This is
rate increases with the first few percent of Cr addition, the concentration necessary to form a complete
due to duplex scale formation and the increased oxi- Cr2O3 layer, whether as an external surface layer or
dation rate from oxygen inward transport, as described a healing layer beneath some base metal oxides.
for Ni and Co alloys in Section 1.10.2. Afterwards, the Once the Cr2O3 layer is formed, the oxidation rate
becomes dominated by its subsequent growth. Ni–Cr surface exhibits oxide grain ‘imprints,’ where the
and Fe–Cr alloys behave similarly.210 scale had maintained contact with the alloy. It is
The minimum Cr concentration necessary for the evident that the fraction of these debonded areas
establishment of a continuous Cr2O3 layer is usually can be significant, and the extent of scale detachment
around 13–25 wt% in oxygen at atmospheric pressure. is found to be strongly related to alloy purity.215
However, this selective oxidation process depends With longer oxidation times, Cr2O3 scales formed
heavily not only on partial pressures of oxygen, but on Cr can be heavily buckled (or convoluted) or show
also on oxidation temperature; sample surface treat- localized nodule formation (Figures 13, 14(a), and
ments and substrate grain size are also influencing 14(c)). Development of the nodules seen in Figure 13
factors. Consequently, alloys with compositions has been shown to be related to oxide grain size.
around the narrow transition range can develop non- When the Cr surface was etched prior to oxidation,
protective or protective scales with minor changes in large grained Cr2O3 developed on some metal orien-
operating conditions. For example, a Ni–14.6 wt% tations, while finer grained oxide formed on other
Cr211 can either form a protective Cr2O3 scale or a orientations and on alloy grain boundaries.208 The
Ni-rich oxide, depending on how it was heated to fine grained oxide grew considerably faster, and
the oxidation temperature. The former protective apparently inward and outward from the original
scale developed when heating was done in vacuum alloy surface, and is most likely due to faster transport
(105 torr), but the latter behavior was found when rates via oxide grain boundaries. Both kinds of oxide,
the sample was heated in air, attesting to the pO2 thin film or nodule, remained in good contact with
dependence of the selective oxidation process. Fine- the metal. However, such good contacts are not often
grained212 or cold-worked alloys213 can also assist the observed. Typically, scale buckling occurs, as seen in
establishment and maintenance of the Cr2O3 layer, Figure 14 in plan view and in cross-section. The
often at Cr concentrations lower than otherwise pre- buckles appeared to have formed preferentially
dicted. This is because alloy grain boundaries are faster along alloy grain boundaries, similar to those ob-
diffusion paths, so the healing Cr2O3 layer is preferen- served by Howes,216 who in a later paper217 investi-
tially established at the intersections of these grain gated buckling above alloy grain boundaries under a
boundaries with the surface. The effect has been scanning electron microscope. It was suggested that
clearly demonstrated on a Ni–Cr alloy.214 For Fe–Cr the preferential separation was due to faster Cr
alloys with comparable compositions to Ni–Cr, Cr2O3 boundary diffusion from the alloy, hence higher cat-
is more readily established initially because of the ion vacancy flux, but not all grain boundaries devel-
greater alloy interdiffusion coefficient in the bcc Fe– oped buckles, so the exact reason why scales are
Cr structure compared with that in the fcc Ni–Cr buckled on some locations but not on others is not
structure, and the lower oxygen solubility and diffu- clear. Depending on the sample surface treatment,
sivity in Fe–Cr. Consequently, less Fe than Ni oxides oxidation conditions and probably other factors,
are present above the Cr2O3 layer formed on Fe–Cr such as metal purity, the extent of buckling can be
and Ni–Cr respectively. Detailed discussions of the rather severe, resulting in very convoluted oxides.
selective oxidation process and factors affecting it Cracking of these extremely convoluted, or buckled
can be found elsewhere in this volume. oxides can occur at the oxidation temperature and
allow oxygen access to the bare metal surface under-
neath; repeated cracking and healing at temperature
1.10.5.3 Scale Morphology and
eventually leads to the formation of a multilayered
Breakaway Oxidation
oxide, as shown in Figure 16. In the case of
The first-formed, thin Cr2O3 scales are uniform and chromium-containing alloys, this process is unlikely,
adherent. However, under these apparently adherent except for high Cr contents,218,219 because selective
scales are often debonded areas or facetted interfacial oxidation of Cr to form Cr2O3 leads to an appreciable
voids (Figure 14(b)), which can be seen after the Cr depletion in the alloy beneath the scale–alloy
scale has been spalled away from the force of a interface. When the protective Cr2O3 is damaged,
scratch. The debonded areas had a smooth, often base metal oxides may develop, resulting in the for-
thermally etched appearance, where the scale and mation of large nodules that are usually double lay-
substrate were not in intimate contact at the reaction ered, with an outer layer of almost purely base metal
temperature, and their fraction increased with oxida- oxide, and an inner layer of mixed base metal and Cr
tion time (Figure 14(d)). Elsewhere, the metal oxides. Depending on the growth rate of the base
stresses in Cr2O3, although the level was not as high

as that found by Kitamura et al.223 Counter diffusion of
oxygen and chromium and internal oxide growth
resulting from these fluxes has been suggested47 to
be the cause of such stresses.
More severe Cr2O3 buckling and convolution have
Cr 20 μm been observed on Cr oxidized at low pO2 values.194
Figure 16 Optical micrograph showing layers of Cr2O3 The scales were heavily wrinkled or even totally
scales formed on Cr metal after oxidation at 1090 C for detached from the metal except at specimen corners.
21 h. Reproduced from Caplan, D.; Sproule, G. I. Oxid. Met. It was suggested that the difference in behavior in
1975, 9, 459–472.
lower pO2 is due to a higher oxide plasticity at low
oxygen pressures. The greater plasticity would indeed
metal oxide and the diffusion rate of Cr in the alloy, a allow the scale to deform without cracking, but there
healing layer of Cr2O3 may again form at the base of still appears to exist a higher driving force that causes
the nodule and thus stop its growth. Otherwise, the the greater deformation. It is possible that higher
fast growing nodule continues to grow vertically and growth stresses were generated at lower pO2. If the
laterally, and eventually destroys the protective stress generation mechanism was that of internal
Cr2O3 scale. growth, as proposed by Rhines and Wolf,47 it is possi-
Associated with the failure of the protective Cr2O3 ble that the relative diffusion rates of Cr and O may
layer to allow fast base metal oxide growth are accel- differ under different pO2 values.
erated oxidation kinetics that can be observed as a
fast increase in specimen weight gain, and called
‘breakaway oxidation.’ When the Cr2O3 forms again 1.10.6 Oxidation of Alumina-Forming
as a healing layer, the oxidation rate can revert to that Alloys
dominated by Cr2O3 scale growth. In cases where
breakaway and healing happen continuously at dif- At temperatures above 900 C, alloys or coatings that
ferent locations, the weight gain may still show an develop Al2O3 scales are most commonly employed
apparent parabolic behavior, but with a much faster for practical applications, due to the relative ease
rate than that found for Cr2O3 only. Breakaway oxi- with which alumina can develop into a complete
dation usually occurs more easily on Fe–Cr than Ni– layer and the slow growth rate and long-term chemi-
Cr or Co–Cr alloys,220 due to a poorer scale adhesion cal stability of Al2O3. There are mainly two types of
on Fe–Cr alloys, for reasons that are not clear. When commercial Al2O3-forming alloys: the aluminides,
breakaway does happen for Ni–Cr alloys,211 the situ- such as FeAl and NiAl, and the MCrAl type alloy
ation is worse, because chromium depletion at the that is often the base of superalloys, where ‘M’ repre-
alloy–oxide interface is greater and longer lasting sents the base metal and can be a mixture of Fe, Ni,
than for Fe–Cr alloys. and Co. For the aluminides, at least 19 at.% Al is
The driving force for buckling has to be the needed to achieve the selective oxidation of Al to
compressive stress that develops in the Cr2O3 develop a protective Al2O3 scale and suppress the
film.203 Although earlier works have concluded that oxidation and growth of iron oxides227,228; 22 at.%
oxidation stresses are negligible compared with ther- Al is needed for Ni–Al, where a thin layer of
mal stresses,221,222 direct measurements of stresses in Ni-containing oxide, NiO and/or NiAl2O4, is present
Cr2O3 at oxidation temperatures have been achieved at the scale outer surface and the Al2O3 forms at
more recently. Using in situ Raman spectroscopy, the scale–alloy interface as a healing layer. In accor-
Kitamura et al.223 determined that the growth stress dance with the theory on selective oxidation, these
in Cr2O3 formed at 900 C in air is highly compres- critical Al compositions can vary depending on oxi-
sive, with a maximum level of 700 100 MPa. dation temperature and environment, and also on
Although still a fraction of the thermal stress, alloying additions. The addition of about 18–20 at.%
measured to be about 3500 MPa, the magnitude is Cr in MCrAl type alloys decreases the amount of Al
higher than that observed on any Al2O3 forming necessary to establish the protective Al2O3 layer to no
alloys, recently performed with in situ synchrotron more than 10 at.%. Iron aluminides have limited
X-ray techniques.224,225 A study by Zhu et al.226 using industrial usage, due to their low high temperature
hot stage X-ray diffraction, also reported compressive strengths.163 Platinum is often added to nickel
aluminides229 to enhance the oxidation resistance, effect,240 where the alloy is suggested to initially
mainly by increasing the Al2O3 scale adhesion. develop Cr2O3, causing a decrease in the partial pres-
Small amounts of Ti, Zr, Hf, Y, or a combination of sure of oxygen at the scale–alloy interface to facilitate
these are added to MCrAl type alloys also to enhance the Al2O3 formation. Therefore, steady-state scales
Al2O3 adhesion; these are the so-called ‘REs,’ which that form on these alloys are always almost pure
have higher oxygen affinities than Al, and their effects Al2O3. However, the initial scale still has appreciable
on the oxidation behavior of Al2O3-forming alloys will amounts of the base metals, because all metals and
be presented in Section 1.10.7.1. alloys exposed in the ambient atmosphere contain a
thin layer of native oxide that is a few nanometers
thick. An example is given in Figure 17(a) for
1.10.6.1 Scale Development
FeCrAl where the surface oxide consists of an outer
Only Al2O3 is formed on b-NiAl230–232 and layer rich in Fe and an inner layer slightly enriched
b-FeAl233 at elevated temperatures (>800 C). On with Cr. When the alloy is heated to elevated tem-
g0 -Ni3Al, an external layer of Ni-rich oxide, either peratures in air or oxygen, this native oxide quickly
NiO and/or NiAl2O4, is always present, beneath thickens and aluminum starts to enrich at the scale–
which is the Al2O3 layer.234–236 Unlike Ni3Al, Fe3Al alloy interface (Figure 17(b)). Further oxidation
alloys do not seem to develop a noticeable Fe-rich results in a fully established aluminum oxide layer at
surface layer.237,238 When such alloys are heated the scale–alloy interface (Figure 17(c)). Subsequent
slowly to 1100 C in air,239 the first detected oxide, scale growth only involves thickening of this alumina
using Raman spectroscopy, was Fe2O3, whose inten- layer, where all the initially formed Fe and Cr oxides
sity increased from 500 to 700 C, then dropped to become incorporated into the Al2O3 layer. Integrating
zero with continued oxidation at higher tempera- the depth profile curves in Figure 17 over distance
tures. Above 900 C, the scale consisted of only shows that the initial stage oxide clearly contains high
Al2O3 and Fe2O3 was no longer detected. This differ- levels of base metals, in this case Fe and Cr, whose
ence between the Fe-based and the Ni-based alloys is concentrations decrease as the scale develops into a
believed to be associated with faster diffusion rates of complete layer of the thermodynamically most stable
Al relative to Fe in Fe3Al than of Al compared to Ni oxide, in this case Al2O3.241 This process of initial stage
in Ni3Al.238 base metal oxidation and the subsequent incorporation
On MCrAl type alloys, the addition of Cr promotes of their oxides into the final scale are expected for all
Al2O3 formation, by the well-known third element alloys.242 The amount that is incorporated depends
Concentration (at.%)
80 80
Fe
60 O 60 O Fe
40 40 Al
20 Cr 20 Cr
Al Fe
0 0
0 2 4 6 8 0 100 200 300 400
(a) Distance from surface (nm) (c) Distance from surface (nm)
80 80
O Fe Fe
60 60 O
40 40 Al
20 Cr Cr
20
Al Fe
0 0
0 20 40 60 80 100 0 100 200 300 400
(b) Distance from surface (nm) (d) Distance from surface (nm)
Figure 17 Auger depth profiles through the surface oxide on a FeCrAl alloy (a) the native oxide prior to oxidation, (b)–(d)
oxides formed after the alloy was introduced into a 1000 C furnace and oxidized in air for 1, 3, and 11 min, respectively.
Reproduced from Hou, P. Y. J. Am. Ceram. Soc. 2003, 86, 660–668.
strongly on the alloy composition, the oxidation tem- the growth stress. Surface smoothing of the platelets
perature and the partial pressure of oxygen in the occurs with time as a spheroidization phenomenon
environment, or sometimes even on the alloy surface driven by a decrease in surface energy.258
finish.243 The behavior of FeCrAl similar to that When a-Al2O3 is nucleated at the transition
reported here has been shown using other surface alumina–alloy interface, its orientation is random,
techniques244,245; NiCrAl246 also forms transient Ni and it forms an incoherent interface with the alloy.232
and Cr oxides. The effect of these base metals in On FeCrAl type alloys, these nuclei are finer
Al2O3 on its transport, and hence growth properties, grained250,261 than those on NiAl260,262 or NiPtAl,263
are addressed later in this section. where they tend to form large circular plates that
When Al2O3 is first developed, regardless of the extend faster laterally than vertically260,262,264;
alloy system, it is never the thermodynamically most an example of these circular plates found on Ni–
stable a form, especially upon heating, but consists of 55 at.% Al is shown in Figure 18(a). The number
alumina that have often been identified as the cubic density of these circular plates depends on the alloy
y230,231 or g232,235,247–250 phase, or a mixture of the grain orientation260 and increases on rougher sur-
two.230,252 The more stable hexagonal a-Al2O3 phase faces.263 The first formed a-Al2O3 is always under
later nucleates at the transition alumina–alloy inter- tension,224,225,260 which can be as high as 400–
face,249–251 while the initially formed transition alumina 600 MPa. Since there is 5% volume shrinkage asso-
transforms to the a form with time. The development ciated with the y-to-a transformation, this result
of a-Al2O3 is faster at higher temperatures252 and with indicates that most of this alpha phase is developed
the presence of Cr253 and/or Fe238,239 in the alloy, but by oxide phase transformation, rather than from oxi-
seems slower with higher concentrations of Al or the dation, that is, growth of a new oxide grain at the
presence of a RE, such as Hf, Y or Zr.254 interface. The high tensile stress, which exceeds the
The first-formed alumina are often referred to
as ‘transition’ or ‘metastable’ alumina. At present,
there are no clear indications as to why a certain
form of transition alumina is formed; either y or
g has been found without any systematic differences
in alloy type, oxidation temperature, detection meth-
ods or oxidation environment, for example, the pres-
ence of moisture. Even the d230 and k phases have
α-Al2O3
been reported in some cases. Since there are many patches
polymorphs of alumina, and many routes by which θ-Al2O3
scale
one phase can transform to the other,255 it is possible
that many different types of alumina can be pre- (a) Alloy
1 μm
sent during the initial stage of oxidation.256 TEM
studies230,232,235,236 often showed that the transition
alumina formed epitaxially, having a cube-on-cube
orientation relationship with the alloy, while the
degree of preferred orientation decreased with oxi-
dation time. Limited high-resolution TEM studies257
indicated that the interface between the transition
alumina and the alloy, in this case a single crystal
NiAl, was coherent, but became incoherent once
a-Al2O3 nucleated. The morphology of the transition
aluminas is often described as platelet or needle-like. (b) 1 μm
TEM analyses258,259 have identified the plane of the
platelets to be (110), whether they are y or g-Al2O3; Figure 18 Examples of alumina scale morphology
and the platelets are often heavily twinned along the resulting from the transformation of y to a alumina during
oxidation with SEM micrographs showing (a) cross-section
[111] direction.259 This structure is probably related
of scale formed on Ni–55 at.% Al, 1100 C, 40 min. (b)
to the high compressive growth stresses present in the Surface of oxide on the same alloy after 11 h at 1100 C.
initially formed transition alumina layer,260 whereas Reproduced from Hou, P. Y.; Paulikas, A. P.; Veal, B. W.
twinning and needle-like growth are means to relieve Mater. High Temp. 2006, 22, 373–381.
toughness limit of a-Al2O3,265 often causes cracking because transport through Al2O3 scales is less sensi-
of the first-formed a-Al2O3 grains. Ridges of oxides, tive to impurities.
which are associated with a-Al2O3 grain boundaries, As common base metals such as Fe, Ni, Co, and Cr
develop and fill the cracks after the transformation is can be incorporated into the Al2O3 scale during ini-
completed (Figure 18(b)). The alpha nuclei eventu- tial stage oxidation, and some of these elements have
ally impinge and develop into a complete a-Al2O3 been shown to have a profound effect on the high
layer above the alloy, and subsequent scale growth temperature creep properties of Al2O3,266–271 hence
becomes dominated by the transport of O and Al affecting O and/or Al transport; it is worthwhile to
through this alpha layer, while the initially formed examine whether these incorporated base metals
transition alumina transform to the a-Al2O3 phase affect oxidation rates. Before that, one needs to real-
with time. ize how much of these base metals are present in a
When Cr and/or Fe are present in the alloy, the growing a-Al2O3 scale and whether they segregate at
transformation to a-Al2O3 is accelerated. Since the grain boundaries. TEM studies have found a
Cr2O3 and Fe2O3 have the same corundum structure decrease from 4 to 0.3 at.% Fe and Cr in the scale as
as a-Al2O3, they have been suggested to serve as it increases from 0.4 to 5 mm for Al2O3 scales grown
templates that facilitate the nucleation and growth on FeCrAl alloy at 1000–1200 C,272 indicating that
of the a phase at a temperature below which this the scale becomes purer as it thickens with oxidation
transition happens in the bulk at 1000 C.239,253 time. In other words, all these impurities were
TEM studies have found small particles of a-Al2O3 incorporated during the initial stage, attesting to the
present in the first-formed transition alumina layer protectiveness of the Al2O3 scale. The Fe and Cr
on FeCrAl250 and Fe3Al238 alloys. Although it is not concentrations in the Al2O3 appeared uniform when
clear how their formation is related to the Cr2O3 relatively large areas, that is, 14 mm2, were examined,
and/or Fe2O3 that were present on the alloy surface but they varied greatly from point to point under a
at the onset of oxidation, their presence in the initial 17 nm beam size. This kind of distribution was also
Al2O3 layer must also help to facilitate the alumina found for Ni and Cr in Al2O3 formed on NiCrAl.293
transformation and the development of a complete These local high concentrations probably resulted
a-Al2O3 layer. from small grains of base metal oxides or spinels left
from the initial stage of oxidation. The lattice param-
eter of a-Al2O3 grains within the scale was enlarged
1.10.6.2 Oxidation Rates
with higher concentrations of Fe, Ni, and Cr,272 sug-
Oxidation of alumina scales usually follows parabolic gesting that at least some of these elements are dis-
kinetics, indicating diffusion-controlled growth. The solved in the Al2O3. Point analyses made at many
growth rate of the transition alumina is about 1 order grain boundaries and within the grains adjacent to the
of magnitude faster than that of the a-Al2O3.248 Due boundary found 38% of the boundaries with higher
to the phase transformation and the different growth concentrations of Fe, but none had higher concentra-
rates of the different aluminas, oxidation kinetics tions of Cr. Since Cr2O3 forms a complete solid
usually show two parabolic stages separated by a solution with Al2O3,273 Cr is expected to dissolve in
gradual transition. Al2O3 and not segregate on grain boundaries. Fe may
The parabolic rate constants of many binary and have segregated, but its segregation may be charge
ternary Al2O3-forming alloys, known prior to 1982, related, hence pO2 dependent. Recent work by
have been compiled by Hindam and Whittle.202 Harmer et al.274 using EELS shows that Fe2þ segre-
Their summary included alloys that contained a gates at Fe-doped Al2O3 grain boundaries, but not
RE, such as Y, Ce, or Th, and contributions from Fe3þ. The behavior is consistent with the fact that
transitional alumina phases were not always elimi- Fe3þ has an appreciable solubility in Al2O3, about
nated. It was also noted that, in some cases, strict 3–5 at.%,275 but the solubility is only a few parts per
parabolic behavior was not observed, so the rate con- million for Fe2þ.276
stants were only determined from the weight change In order to better correlate alloy composition to
near maximum exposure times. Despite these reser- oxidation rates, data cited in the Hindam and Whittle
vations, all the data fell in a scatter band of about 16, paper202 are reevaluated, and those representing
with scale thickness for a given time varying by a steady-state growth rates from Ni-, Fe-, or Pt-based
factor of 4. The scatter is much smaller than that alloys that are undoped with REs are compared with
observed for Cr2O3 scales (Figure 12), probably more recently published results.277–302 These rates
are plotted in Figure 19 as a function of the inverse Pt-based binary or ternary alloys without any RE
of the oxidation temperature. All the studies included addition can vary over two orders of magnitude, with
here are from laboratory tests performed on high an average activation energy of about 240 kJ mol1.
purity binary M–Al or ternary M–Cr–Al type alloys, Although base metal components are incorporated in
so M (Fe, Ni, or Pt) and Cr should be the most the growing alumina during the initial stage of oxida-
abundant impurities incorporated in the Al2O3 tion, and depending on their size and solubility in
scale. Most tests were carried out at one atmosphere Al2O3, may segregate to oxide grain boundaries, their
oxygen, but some in air. Different symbols in effect on oxidation rates seems small. In a way, this is
Figure 19 distinguish different types of alloys. The surprising, because Al2O3 is a very stoichiometic oxide,
data showed much scatter from study to study even whose transport property is dictated by impurity con-
for alloys with the same composition. Some general tents.305 No explanations are apparent for the large
trends, however, seem apparent. PtAl alloys, on aver- variation in reported oxidation rates. Perhaps part of
age, tend to have the slowest oxidation rates, but this this scatter is indeed due to the presence of different
conclusion is not supported by all the PtAl data. The types and/or amounts of impurities under different
presence of Cr does not seem to make any difference test conditions. To date, there is no good understand-
to oxidation rates. The Fe-based alloys seem to oxi- ing of which impurities these may be. Tolpygo et al.,306
dize with the highest rates, which would agree with for example, have shown that grit blasted coating
the higher creep rates observed for Fe-doped surfaces develop much thicker Al2O3 scales (by a
Al2O3.266 However, the higher oxidation rates may factor of 3) than other surfaces, due to the incor-
also be related to increases in surface area that poration of impurities, such as Li, Na, K, Ca, Mg,
resulted from wrinkling of the scale and the underly- and/or Si, in the surface
ing alloy.303,304 FeCrAl and Fe3Al type alloys are
especially prone to wrinkling. Such increase in sur-
1.10.6.3 Oxide Growth Mechanism
face area can be as high as 10%,304 but none of the
oxidation rates cited here had been corrected for it. The transition alumina has been shown to grow pre-
In summary, the growth rates of a-Al2O3 scales dominantly by aluminum outward transport.231,307
from 1000 to 1300 or 1400 C on Fe-, Ni-, Co-, and From several earlier inert Pt marker studies,301,308
a-Al2O3 was believed to grow predominantly by
O inward diffusion, since markers that were placed
1E–009 on the alloy surface prior to oxidation remained on
the scale surface afterwards, indicating inward growth
of the scale. The minor amount of aluminum outward
1E–010 transport is believed to contribute to lateral growth of
the scale, causing extensive scale convolution, or
wrinkling298; an example is given in Figure 20.
Based on the degree of such a convolution, whether
1E–011
kp (g2 cm–4 s)
the scale is detached from the substrate or wrinkled

with it, the extent of lateral growth on polycrystal-
line FeCrAl alloy, for example, can be as high as
1E–012 20–50%.303,304 This shows that the amount of Al
Fe3Al outward transport is not trivial. More recent two-
FeCrAl stage oxidation studies using 18O/16O309–315 indi-
FeAl
1E–013 Ni3Al cated clearly that a considerable amount of Al out-
NiCrAl ward transport indeed occurs during scale growth. In
NiAl
PtAl
this type of study, the oxidation at each stage is done
1E–014 in an atmosphere enriched with one of the oxygen
0.5 0.6 0.7 0.8 0.9 isotopes. The isotope distribution throughout the
1/T (3 10–3 K–1) scale is subsequently determined using spectroscopic
methods. Secondary ion mass spectroscopy (SIMS) is
Figure 19 Parabolic rate constants showing the Al2O3
scale growth rate on different Fe, Ni, and Pt-based alloys. the most commonly employed technique.311,313–315
Reproduced from Hou, P. Y. J. Am. Ceram. Soc. 2003, 86, Secondary neutral mass spectroscopy312 (SNMS)
660–668. and other techniques309,310 have also been used.
60
Oxide Alloy
Concentration (at·%)
O total
40
16O
20
18O
0
0 500 1000 1500 2000
Sputtering time (s)
Figure 21 An example of oxygen isotope profiles from
Quadakkers et al.312 after two-stage oxidation of an FeCrAl
(a) 10 μm alloy oxidized at 1000 C first in 16O2 for 2.5 h, then in 18O2
for 5 h.
formed oxide and at the scale–alloy interface. This is

the result of oxygen diffusion through the Al2O3 scale
forming new oxides at the scale–alloy interface. If the
outer and inner portions are integrated and their areas
(b) 5 μm compared, one obtains an outward to inward growth
ratio of 0.45. This method is of course not quantita-
tively precise, but can be used to compare the effect of
dopants on the transport properties through alumina
scales.
Although the isotope studies clearly show that
a-Al2O3 grows mainly by oxygen inward transport,
but with an appreciable amount of Al outward trans-
2 μm port, the transport path, that is, lattice, grain bound-
(c)
ary or other short circuit diffusion paths, cannot be
Figure 20 Wrinkling of alumina scale formed on FeCrAl at revealed. The grain size of transition alumina that grows
1000 C in oxygen: (a) oxide surface and (b) polished cross- initially on alloy surfaces upon oxidation is extremely
section after 3 h, (c) fractured cross-section after 120 h.
fine, in the range of a few nanometers.238,246 Even
with more mature scales, that is, a-Al2O3, that have
From the location and the distribution profiles of the grown to several microns thick, the oxide grain size is
second stage isotope, the dominant transport species still no more than 1 or 2 mm.281,282 Owing to these
and its path can be determined.37,310,316 A typical small grain sizes, and the slow oxygen lattice diffusiv-
profile from the results of Quakakkers et al.312 is ity found by tracer studies,317 grain boundaries have
presented in Figure 21 as an illustration. The alu- always been considered the dominant transport
mina scale was formed at 1000 C on a FeCrAl alloy paths through alumina scales. More recently, micro-
after oxidation first in 16O2 for 2.5 h to ensure a structural evidence has also shown the importance of
completely transformed and stable a-Al2O3 layer, grain boundary transport through a-Al2O3 scales.263
then without cooling, the specimen was continually While some oxide ridges, such as those shown in
oxidized in 18O2 for 5 h to allow new oxides to grow Figure 18, develop during the initial stage from
so its location could be determined later. A large peak alumina transformation, continued ridge growth can
of the second isotope, in this case 18O, is present at the still occur through an adherent a-Al2O3 layer. Fur-
scale–gas interface. Portions of it come from oxygen thermore, polished cross-sections, as seen in
isotope exchange but, due to the slow rate of this Figure 22, often show thicker scales above and below
process,37 its contribution should be small. The major- an Al2O3 grain boundary,263 giving clear indications
ity of 18O found at the outer surface is therefore due to that this is the preferred transport path through the
new oxide formation from Al that diffused through the scale, and growth of the oxide took place at the scale–
scale and reacted with oxygen from the atmosphere. alloy interface and on the oxide surface, as suggested by
A significant amount of 18O also exists within the first oxygen tracer studies.
Oxide
surface
Cross
section
2 μm
Figure 22 SEM image of polished cross-section of alumina formed on NiPtAl, showing oxide ridges above and below the
alumina grain boundary. Oxidation was carried out at 1200 C for 50 h in air. Reproduced from Tolpygo, V. K.; Clarke, D. R.
Mater. High Temp. 2000, 17, 59–70.
The growth of an Al2O3 oxide scale that follows 10 nm is chosen to convert grain boundary diffusivity,
parabolic kinetics is controlled by the diffusion of usually presented as dD with units of cm3 s1.
O and Al through the scale layer according to the Oxygen diffusion from tracer studies in single crys-
following equation tals undertaken by several investigators317–321 all agree
Z " ! !# within an order of magnitude and with similar activa-
dDAl dDO
kp ¼ 1:5 DLAl þ b þ DLO þ b d lnPO2 ½1 tion energies. Good agreement was also found with DLO
P O2 d d obtained from dislocation loop annealing studies by
where kp is the parabolic rate constant, d is the oxide Lagerloff et al.322 The apparent oxygen diffusivity
grain size and Ds are the respective diffusivities. determined on polycrystalline Al2O3 by Oishi and
Assuming the Ds are not functions of PO2 and that Kingery317 showed enhanced rates by 1–2 orders of
the contribution from lattice diffusion is insignificant, magnitude; the lower temperature branch may be
an effective diffusivity, Deff, which is a combination of related to impurities, but the exact reason is unknown.
the boundary diffusivities of Al and O, can be To deduce a boundary diffusivity from these data is
obtained as not straightforward, as the study only measured the
total uptake of 18O. The boundary diffusion of oxygen
Deff ¼ kp d =lnPO2 ½2
obtained by Prot and Monty323 has a drastically differ-
The PO2 is the oxygen potential difference across the ent activation energy from that of Oishi and King-
alumina scale. The parabolic rate constant, kp, in its ery.317 Tracer studies of Al diffusion have been limited
usual units of g2 cm4 s1, can be converted to the due to the low activity of 26Al. The only data are from
units of Ds in cm2 s1 using the density of alumina. Paladino and Kingery324 and a more recent work by
Assuming thermodynamic equilibrium at the Al2O3– Le Gall et al.325 Although Palodino and Kingery324
alloy interface, the PO2 there would be that of the used polycrystalline Al2O3, the grain size was
dissociation pressure of Al2O3 at the oxidation tem- 130–200 mm and the behavior was believed to be of
perature. From most reported microstructure of oxide bulk diffusion. In that case, this result should be com-
scales, the average oxide grain size can be taken as parable to those from Le Gall et al.,325 whose study was
roughly 0.5, 1, 1.5, and 2 mm respectively for oxida- performed on single crystal Al2O3, but was slightly
tion at 1000, 1100, 1200, and 1300 C. The range of Deff contaminated with Si.320 Yet the two sets of data differ
thus calculated using eqn [2] from the range of by nearly four orders of magnitude. The discrepancy
kps shown in Figure 19 is presented as a shaded area between these two tracer studies may lie in the differ-
in Figure 23. On this figure are also plotted different ence in the Si content of their samples, or simply due
cation and anion diffusivities experimentally deter- to experimental difficulties using the 26Al tracer.
mined in bulk alumina. A grain boundary width, d, of These results show that diffusion data on Al2O3 are
Temperature (⬚C)
1800 1600 1400 1200 1000
–7
O g.b.
–8 (Le Gall et al.)
–9
Al polycrystal
(Paladino and Kingery)
–10 From
oxidation
–11
O polycrystal
(Oishi and Kingery)
log D (cm2 s–1)
–12
1
–13
3 O single
5 crystal
2
–14 4
Reddy
–15
–16 (1) Oishi and Kingery317

6 (2) Reddy and Cooper318
Al single crystal (3) Oishi et al.319
(Prot and Monty)323 (4) Legall et al.320
–17 (5) Reed and Wuensch321
(6) Lagerloff et al.322
–18
4.4 4.8 5.2 5.6 6.0 6.4 6.8 7.2 7.6 8.0
104/T (K–1)
Figure 23 Comparison of diffusion rates calculated from oxidation kinetics with published tracer diffusivity data.
O
rather scattered. One can only conclude that Dgb (oxy- profiles, giving rise to a higher DL, thereby causing
gen boundary diffusion) is much faster than that of DLO , deconvolution to yield ratios of Db/DL that are too low
and the same may be true for Al. A conclusion on the by a few orders of magnitude.327
relative magnitude of the Al and O boundary diffusiv- Oxidation studies are usually performed at much
ity is more difficult. They may be more comparable lower temperatures than creep, sintering or tracer
than the difference between Al and O lattice diffu- diffusion studies, making direct comparisons less
sions. These general conclusions agree qualitatively straight forward, where diffusivities are extrapolated
with the current view on O and Al transports through assuming activation energies to be constant over a
Al2O3 scales, and also with diffusion rates compiled wide range of temperatures. The effective diffusivity
from creep and sintering studies.326 However, qualita- determined from the range of kps from a-Al2O3 scale
tively, diffusivities from creep tend to be at least an growth (shaded band in Figure 23) is seen to lie
order of magnitude higher than the rates determined slightly above the available dDbAl values from tracer
from tracer studies. This could be either due to con- studies. Compared to dDbO , it is more than two orders
tributions from other deformation mechanisms, such of magnitude higher than most of the tracer data,
as basal slip and cavitation, or from trivial amounts of except the calculated lower temperature branch of
grain boundary migration that perturb the tracer Oishi and Kingery.317 These comparisons suggest that
Al boundary transport in Al2O3 scale growth is not by others,344 suggesting that interfacial pore forma-
trivial, which agrees with the results of two-stage tion is a result of uneven stress distribution at the
oxidation studies using 18O2/16O2310 and with the scale–alloy interface.
conclusion obtained from oxide wrinkling analysis328 Although the exact mechanism of void formation
and oxide microstructures.263 is still not established, it is clear that on Ni or Fe
aluminides, voids almost 10 times larger than the
oxide grain size and as deep as the scale thickness
1.10.6.4 Scale Adhesion
can exist from the very beginning of oxidation,335
Although Al2O3 grows slowly to offer good oxidation even during the heating stage; an example is shown
protection, it spalls extensively during cooling on in Figure 24. On FeAl, it has been shown that most
alloys that do not contain any RE. Part of the reason voids nucleated during the initial stage of oxidation,
is the high thermal stress, as high as several giga- where the first-formed transition alumina grew pre-
Pascals,329 that exists between Al2O3 and these dominantly by cation outward transport. Moreover,
Al2O3-forming alloys, and partly due to a weak the density of these voids increased dramatically with
Al2O3–alloy interface. Unlike Cr2O3, Al2O3 almost the presence of surface impurities, such as C, Cl, P, S,
always fails adhesively, where the oxide delaminates or a combination of these.345 Impurities, whether
along the interface, exposing bare alloy surfaces after present at the alloy surface prior to oxidation or
scale spallation. Interfacial pores are often present inside the alloy that diffuses to the surface during
after oxidation, acting as interfacial defects, which oxidation, have been suggested to enhance pore for-
can further deteriorate scale adhesion. Sulfur impu- mation by lowering the metal surface energy.346
rity is known to weaken the interface and REs to Other than the often large pores found at oxide–
strengthen it. These effects are discussed in detail in metal interfaces, various types of porosity have also
the next section. Discussion of Al2O3 adhesion here been reported within the Al2O3 scale. One particularly
only involves alloys without any RE addition. common appearance is a row of fine pores at the
Intermetallic alumina formers, such as NiAl and boundary between the first-formed outer transition
FeAl, are particularly susceptible to forming large
cavities at the scale–alloy interface. Several experi- Epoxy
mental studies on NiAl232,235,330–332 and FeAl333–335 Al2O3
have investigated the pore formation behavior in these
systems. On NiAl,331 more voids were observed on the
Ni-rich alloys, where diffusion of Ni in the alloy is
much faster than that of Al. The authors attributed the
pore formation mechanism to the Kirkendall effect,
where the different Ni and Al diffusion rates can cause (Ni, Pt) Al
a net flux of vacancy to the scale–alloy interface, which
condenses out into voids. Indeed, the mechanism of (a)
pore formation has often been attributed to vacancy 10 μm
condensation. These can be vacancies that arrive at
the scale–alloy interface from the growing scale as
a result of cation outward transport,336,337 or from
any unequal diffusion of the alloying elements in
the substrate,338 such as that from the Kirkendall
effect. It has been suggested339,340 that vacancies
can be injected into the metal to be annihilated or
to form voids there. However, if the oxide–metal
interface acts as a perfect vacancy sink, such rejection
should not take place; instead, vacancies will be anni- (b)
hilated at the scale–metal interface.341,342 Harris,343 200 nm
on the other hand, disputed the vacancy condensation
Figure 24 Examples of voids formed at the alumina–
model entirely and suggested that tensile stress aluminide interfaces: (a) on a Ni–44Al–6Pt–3.5Co–1.5Cr
induced in the alloy by the compressive growth stress (at.%) alloy after 300 h at 1150 C,385 and (b) on Fe–40Al
in the oxide caused voiding. Similar views are shared after 3 min at 1000 C.335
alumina layer and the subsequently developed a-Al2O3 ‘REs,’357 such as Y, Ce, Hf, and Zr. They are often
inner layer. These pores most likely resulted from the added in the alloy and found in the protective oxide
alumina transformation that is associated with a vol- scales. Studies have shown that they segregate strongly
ume reduction. Other types of small pores, some at at alumina and chromia grain boundaries and alter the
Al2O3 grain boundaries, especially at triple point junc- scale growth process.358,359 The first part of this sec-
tions, and some even at oxide grain centers, are com- tion describes the RE effects and discusses some of
monly observed. A few explanations have been given the mechanisms that have been proposed to explain
for their occurrence,246,328,332 but there has not been these effects.
much discussion on this topic. As most of these pores Impurities that are incorporated into the growing
are small, about a fraction of the alumina grain size, and scale can come from many different sources. An appar-
randomly distributed inside the scale, they usually do ent one is the common surface contaminants that are
not contribute to scale failure. present before oxidation, such as Na and K,360 but
Buckling of Al2O3 scales at the growth tempera- their concentrations are usually low and can vary
ture, as often found with Cr2O3 scales, is rarely significantly from test to test. The most abundant
observed, except after extensive scale convolution,347 impurity sources are base metal components (dis-
sometimes referred to as rumpling348 or wrinkling.349 cussed in the previous section), such as Fe, Ni, or Co,
In the beginning of the convolution, the oxide re- which are incorporated as oxides during the initial
mains attached to the alloy at all locations, and the stage of oxidation.241,361 Nonmetallic impurities that
alloy and scale wrinkle together. An example of such are present in almost all commercial grade alloys may
morphology is seen in Figure 20(b). The compres- segregate to the scale–alloy interface during oxidation.
sive growth stress in the oxide is believed to be the In particular, the segregation of sulfur is believed to
driving force266,350,351 that causes an initially planar greatly deteriorate scale adherence.362,363 This phe-
oxide–alloy interface to convolute as a stress relaxa- nomenon is also addressed in this section, after the
tion mechanism. With further oxidation, the rum- discussion on RE effects.
pling amplitude increases and its wavelength
decreases. This phenomenon seems to occur only 1.10.7.1 Reactive Elements
for Fe-based Al2O3 formers without a RE addition,
such as FeAl and FeCrAl352; some believe the process The addition of oxygen active elements, such as Y, Hf,
is related to the alloy’s mechanical properties, but Ce, or Zr, is known to have profound beneficial
others have suggested that it may be related to the effects on the oxidation behavior of Cr2O3- and
rates of interface diffusion.351,352 It is important to Al2O3-forming alloys. This so-called ‘RE effect’ was
note that this type of wrinkling happens at the oxida- discovered 70 years ago364 and was first termed the
tion temperature, with only a small contribution dur- rare earth effect, because rare earth metals added as
ing the first few hundred degrees of cooling.353 The melt deoxidants to Nichrome (Ni–20 wt% Cr) heat-
phenomenon is very different from that observed on ing elements were found to increase cyclic lifetimes
coatings during thermal cycling oxidation.354,355 substantially by increasing the adhesion of Cr2O3
A consequence of the wrinkling morphology is the scale to the alloy. The amounts of the rare earth
establishment of very uneven stress distribution additions were 0.01–0.5 wt%. In a later patent,365 it
along the oxide–metal interface.353 High tensile was found that elements from groups IIA, IIIA, IVA,
stress thus present at the peaks can cause scale and VA of the periodic table could produce similar
separation from the alloy, and after prolonged oxida- beneficial effects. Indeed, subsequent work carried
tion, this indeed occurs. Therefore, even though a out by various other investigators confirmed that a
convoluted, rough interface like the ones discussed wide range of additions affect Cr2O3- and Al2O3-
here can impede crack propagation,356 the outcome is forming alloys similarly, as long as the additions
that it provides a tensile stress component at the form a more stable oxide than Cr2O3 or Al2O3.
interface that can act as a crack initiator. Early results have been extensively reviewed by
Whittle and Stringer,357 where they concluded that
the effects on Cr2O3-forming alloys are fourfold; they
1.10.7 Effect of Minor Alloying (1) enhance the selective oxidation of chromium,
Elements and Impurities (2) decrease the oxide growth rate, (3) change the
oxide growth mechanism from predominantly cation
The most beneficial minor alloying additions in com- outward to oxygen inward transport and (4) improve
bating high temperature oxidation are the so-called the oxide scale adhesion. For Al2O3-forming alloys,
however, the effect was only to improve the scale element to help promote Al2O3 formation; hence, any
adhesion. This last conclusion arises because the effect of RE, on the selective oxidation of Al, could
ratio of cation to anion transport is much lower in not be observed. Advances with improved analytical
Al2O3 than Cr2O3; also, Al2O3 growth rate is signifi- techniques and studies of NiAl and FeAl alloys have
cantly lower than that of Cr2O3, so any effect on since shown that actually effects (1) and (2) also apply
cation transport and oxidation rate would be small to Al2O3 formers.366 Each of these effects is described
and, therefore, difficult to measure. Furthermore, and discussed separately later in this section. Other
most Al2O3-forming alloys studied prior to that than these, RE additions also alter the oxide grain
time were MCrAl type alloys, where M was Fe, Ni structure from equiaxed to columnar367 (an example
or Co. These alloys already have Cr as an added third can be seen on Figure 25), and the average oxide
2 μm
(a) (b)
300
120
100
200
80
kcps
60
100 40
1 μm
20
0 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
(c) (d) (e)
2 μm
(a1) (b1)
1600 1600
kcps
1200 1200
800 800
1 μm
400 400
0
0
0 2 4 6 8 10 0 2 4 6 8 10
(c1) (d1) (e1)
Figure 25 Effect of Y addition (bottom set of figures) in FeCrAl on oxygen transport and oxide microstructure.315 (a) and (c)
are map and line traces of 16O; (b) and (d) are those for 18O. The specimens were oxidized at 1200 C first for 1.5 h in 16O2,
then 3.5 h in 18O2.
grain size is decreased by about a factor of 1.5–2. The system and for a given application is still an open
RE effects are most pronounced between 1000 and question and is currently under investigation.
1200 C. At 900 C or below, they become marginal; Another question worth pursuing is the benefits of
above 1300 C (tested for Al2O3 formers), the effects codoping,373 that is, adding more than one type of RE
break down. in an alloy. This may be the future area for optimiza-
Other than alloying additions of REs, a fine tion of the RE effect for commercial applications.
dispersion of a few volume percent of RE oxide
particles, usually incorporated by mechanical alloy-
ing, are just as effective. Furthermore, superficially 1.10.7.1.1 Promoted selective oxidation
applied RE or RE oxide,368 via ion-implantation or a The beneficial effect of REs on selective oxidation is
thin coating of RE oxides or nitrates that convert to to decrease the amount of Cr or Al in the alloy
the oxides, seems equally effective. However, these necessary to develop the protective Cr2O3 or
surface applications, although sometimes useful for Al2O3 scales. For alloys that already have sufficient
mechanistic studies,369,370 are not effective for long- amounts of Cr or Al, the effect is to decrease the
term applications that are often required for indus- time required for these scales to be established,
trial usage, due to insufficient supply of the RE. The hence decreasing the amount of base metal oxide
distribution368 and concentration278 of RE in alloys formation. RE oxide dispersions seem to be more
are extremely critical for their effectiveness on oxi- effective than RE additions in this respect, while
dation. While earlier studies usually use a few tenths surface applied RE or RE oxides do not show the
of a weight percent, more recent works suggest that a same effectiveness.374
more favorable concentration may be only a few It has been suggested that RE or its oxides in the
hundred parts per million.371 Whether this range of alloy increase Cr diffusion,375 or acts as preferential
concentration is ‘optimal’ for all applications and nucleation sites,376 to promote the establishment of
systems is, however, questionable. The most favorable the Cr2O3 layer. However, diffusion studies377 on
concentration may depend on the type of RE, the Y2O3- and ThO2-containing Ni–Cr and Co–Cr
alloy system, the oxidation conditions or a combina- alloys have shown that the presence of RE disper-
tion of them. Overdoping of RE can produce detri- sions alone does not enhance the diffusivity of chro-
mental rather than beneficial effects. Excess RE in mium in the alloy; rather, Cr diffusion increases with
the alloy can form a second phase, such as Y garnet, a decrease in alloy grain size. Indeed, both RE and
which oxidizes to YAG that promotes internal oxida- their oxide dispersions tend to stabilize a fine alloy
tion and acts as a crack initiator, hence decreases grain size; this has been shown for Cr2O3-378 and
scale adhesion. An uneven distribution of RE oxide Al2O3-forming alloys.379 Furthermore, subgrain
dispersions is also undesirable, for it causes nonuni- structures created by the oxide dispersions can also
form oxidation, where areas without any RE oxides enhance diffusion.377,380 It is known that a Cr2O3
can exhibit fast and nonprotective oxidation behavior. layer develops on Ni–Cr alloys more readily over
The relative effectiveness of the different REs is alloy grain boundaries than over the bulk of the
unknown, although there do appear to be differences. grains, and on cold-worked surfaces than on polished
It is also not yet clear whether one type of RE is more ones.381 Therefore, it appears that the major effect of
effective in a certain type of alloy system, although RE or RE oxides is to decrease the alloy grain size,
the same RE incorporated in different types of alloys thus allowing faster diffusion of the scale-forming
can show different effects.372 Hafnium, for example, element from the bulk to promote an earlier devel-
has been shown to be exceptionally effective in NiAl opment of a protective scale.
alloys, decreasing the oxidation rate by almost a
factor of 10.372 The early patent365 also noted that 1.10.7.1.2 Decreased growth rate and
the effectiveness of RE decreased on passing from changed growth direction
group II to group V, but increased with increasing Above 900 C, RE additions can decrease the growth
atomic weight within a particular group, which is rate of Cr2O3 scales by nearly an order of magni-
similar to recent suggestions made by Pint and tude.357 While the oxidation rates of undoped alloys
Alexander,283 pointing out that the relative effective- are approximately parabolic, those doped with RE
ness may be related to the atomic size of these ele- are often described as asymptotic or cubic. Similar
ments. In summary, the relative effectiveness of REs decreases can be achieved by surface-applied RE or
and their optimal concentrations in a given alloy RE oxides, or even with RE implanted into an
existing thin oxide film,370,382 suggesting that RE transport is decreased with the addition of REs; an
incorporation into the oxide is essential in decreasing example can be seen in Figure 25. An estimate from
the scale growth rate. Below 900 C, the effect is the relative percentage of the two oxygen isotope
minimal, with an approximately twofold decrease in distributions, averaged over seven data sets over the
rates.383 temperature range of 900–1100 C,241 showed that the
When rates on RE-doped and undoped Al2O3- outward transport of Al is decreased by about a factor
forming alloys are compared and plotted over a wide of four. These oxygen tracer studies, performed on
temperature range, the data are scattered over about different types of alloys with different REs, showed
two orders of magnitude and show no apparent effect similar results. However, microstructural observations
of RE on scale growth rate.202,241,357 Therefore, earlier on the extent of grain boundary ridge development on
reviews202,357 have concluded that RE has no effect on Al2O3 surfaces showed obvious differences between
Al2O3 scale growth rate. However, the scatter may be the type of alloy and RE.385,386 These studies were
related to differences in the process of adding RE to made during a second stage oxidation after polishing
the alloys, and the different experimental conditions the top layer of the first-formed oxide. In the first
and procedures. If data with or without RE additions study, it was found that different REs behaved simi-
obtained from the same research group on the same larly on FeCrAl-based alloys, but 0.05 at.% addition of
type of alloy are compared,241 the results show a Zr in Ni–50Al was more effective in decreasing Al
consistent decrease in oxidation rate with the addition outward transport than the same amount of Y addition.
of RE (Table 1). In this summary, no distinction was In a second, more quantitative study, Y was found to be
made on the different types of RE, even though some more effective on FeCrAl-based alloys than Zr. It was
RE may be more effective than others in some cases. concluded that the ratio of inward diffusion of oxygen
Some of the rates were averaged from several tests of to outward diffusion of aluminum depends not only
an undoped alloy, or from alloys doped with different on the oxidation temperature, but also on the RE
REs, giving rise to the relatively large error bars. Still, dopant, as well as its form in the alloy. There are
a clear trend can be seen from these results showing apparent discrepancies between the two reports,
that all RE additions in Ni- and Fe-based alloys but the results are interesting in that they point to
decrease the oxidation rate by about a factor of two. possible different degrees of the RE effect in different
On the PtAl alloy, the addition of Zr decreased the rate systems.
by a factor of four, but only one set of data exist for this Utilizing analytical TEM, Yurek and coworkers
alloy, so it would be premature to conclude that the investigated oxide scales in cross-sectional samples
RE effect on scale growth rate is greater on Pt-based and first demonstrated that RE segregates to
than on Ni- or Fe-based alloys. Cr2O3359 and Al2O3358 grain boundaries. A similar
Other than the decrease in rate, RE additions also segregation has since been observed for many RE-
affect the relative transport rates of oxygen and Cr or containing alloys, where different REs, such as Y, Zr,
Al through the scale. The effect was more prominent Hf, La, and Ti, added in different forms (metallic,
for Cr2O3-forming alloys, so it was often realized with oxide dispersion, ion-implantation, or surface coating)
earlier Pt marker studies.384 Not until oxygen isotopes and in different types of alloys, have been found to
were used, was the effect also verified for Al2O3-form- segregate at Al2O3 grain boundaries.237,281–283,358,387–392
ing alloys, where a noticeable amount of Al outward The amount that is segregated has been quantified to
Table 1 Effect of RE addition on oxidation rates
Alloy Oxidation Rate w/o RE RE type Rate with RE Refs

system condition
NiAl 1200 C (1.4 0.4) 1011 Ti, La, Zr, Hf, Y (8.0 4.1) 1012 277, 279, 280,
Ni3Al 7.1 1012 Y (4.3 3.1) 1012 283, 387
Fe3Al (1.3 0.2) 1011 Hf, Zr, Y (4.9 2.4) 1012
FeCrAl (1.8 0.6) 1011 Nb, Ti, Sc, Hf, Gd, (7.1 2.8) 1012
Ba, Nd, La, Zr, Y
PtAl (1.6 0.4) 1011 Zr 4.1 1012
FeCrAl 7.8 1012 Y 4.1 1012 287
FeCrAl 1150 C 8.5 1012 Ti 2 1013 290
NiCrAl 1100 C 1.1 1012 Y, Hf (4.5 2.1) 1013
be about 0.2 monolayer.392 When the segregation was the scale–alloy interface. Chromia scales without
followed as a function of oxidation temperature and any RE tend to detach from the alloy at the oxidation
time,393 it was determined that the RE not only segre- temperature, leading to widespread buckling, and,
gates at Al2O3 grain boundaries, but also diffuses hence, cracking and spalling during cooling. Alumina
through it to the scale surface, where it can precipitate scales, on the other hand, often develop many
out as fine oxide particles. discrete interfacial voids, which act as defects to assist
Based on the segregation result, and the change in in scale failure, or become wrinkled, as often seen in
the oxygen isotope distribution, Yurek and cowor- Fe-based alloys, which give rise to tensile stresses at
kers358 proposed that the presence of RE at the some locations at the interface that lead to scale
oxide interface decreases cation grain boundary dif- delamination. Apart from these microstructure
fusion, but probably does not affect oxygen diffusion; changes, there are also significant chemical changes
consequently, scale growth becomes dominated by at the scale–alloy interface, which can significantly
oxygen inward transport. This proposal has been affect the strength of the interface. This effect will be
explained in detail by Pint,393 stating that the oxygen discussed in more detail later in this section under
potential across the scale drives the RE to diffuse the ‘sulfur effect.’ Surface additions of REs or RE
from the alloy, through the scale to the outer surface, oxides have been shown to exert a similar beneficial
where the diffusion path is the oxide grain bound- effect on Cr2O3 scale adhesion,397 but do not seem to
aries, and this is where RE segregates. Although there do so on Al2O3-forming alloys.398
are ample evidences of RE segregation to growing Many earlier explanations of the possible role of
Cr2O3 and Al2O3 oxide grain boundaries, creep stud- REs in improving scale adhesion have been summar-
ies have also indicated that RE additions to sintered ized and described.357,366,399– 401 A few of them are no
Al2O3 significantly decreased the creep rate,270,271 longer being considered, except for special cases. For
implying a decrease in Al grain boundary diffusivity, example, in the case of fine-grained scales that
the exact mechanism by which RE at oxide grain develop on nanocoatings,402,403 that is, alloy coatings
boundaries affects cation, but not anion, diffusion is consisting of columnar grains with grain widths in the
not known. Analysis of the effect of impurities on tens of nanometer range, the small grains are envi-
grain boundary structure394,395 may shed some light saged to be highly plastic and, hence, are able to
on the boundary diffusional processes. Another ques- relieve much of the growth stress and decrease the
tion worth considering, which may be closely related driving force for spallation. However, recent in situ
to the transport mechanism of cations and anions in stress measurements using synchrotron radiation
oxide grain boundaries, is why Al and O do not react have shown that the magnitude of the growth stress
at the boundary, as both diffuse through it. Such a is far less important than the interface chemistry or
reaction should result in internal plating of oxide at microstructure in dictating scale failure.404 Oxide
the grain boundary, or lateral growth,47 which would pegging, where fine stringers of Al2O3 oxide grow
generate in-plane compressive stress. It is possible that internally from the external scale into the substrate,
once such stress is generated, it prevents further reac- have been suggested to impede crack propagation
tion, until the stress is relieved. Clarke,396 using a dislo- due to the large interfacial contact areas associated
cation climb model, has determined for Al2O3 scale with the peg microstructure, and to relieve stresses
that only 107 molecules per micron length of grain by the two-phase peg-containing region.405 Mor-
boundary can generate 0.5 GPa of growth stress. phology of the pegs is critical to their effectiveness,406
where a fine and more uniform distribution of small
1.10.7.1.3 Improved scale adhesion pegs is much more efficient in combating scale spall-
The most industrially relevant beneficial effect of the ation than large ones, which are even capable of
RE additions is the improved scale adhesion, espe- initiating scale failure. Although pegging may be
cially for Al2O3-forming alloys and coatings. It has important, there are several alloy systems, including
been demonstrated repeatedly that additions of RE or Cr2O3-forming alloys, which do not form any pegs,
RE oxides in the alloy greatly increase the spallation but still remain adherent under severe cyclic oxida-
resistance of Al2O3 and Cr2O3 scales during cooling, tion conditions.
thereby significantly increasing the lifetime of an To date, the most accepted mechanism by which
alloy or coating under cyclic oxidation conditions. RE affects scale adhesion is related to the chemical
Associated with the greater spallation resistance are change that occurs at the scale–alloy interface when
noticeable microstructural changes, particularly at RE is present. This can be a decrease in interface
S concentration and/or the presence of RE at the Although it is clear that S can segregate at oxide–
interface. The proposed mechanism related to sulfur metal interfaces and weaken the interfacial strength,
segregation362,407,408 is usually known as the ‘sulfur the role of the RE is still uncertain. The presence of
effect.’ The hypothesis is that the oxide–metal inter- RE in the alloy can indeed prevent S segregation to
faces are intrinsically strong, but the indigenous sul- the interface,423 but by what mechanism? It is con-
fur that is present in metals and alloys, possibly as low ceivable that RE can getter sulfur by forming a sul-
as around 5–30 ppm, segregates to the interface dur- fide, but RE oxide dispersions are equally effective in
ing oxidation, thus weakening the bonding and ren- improving scale adhesion. The RE oxides are more
dering the scale nonadherent. The role of the REs, as stable than their sulfides, so the gettering mechanism
a result of their strong sulfide-forming ability, is to does not seem to apply. Since REs are known to
react and tie up the sulfur, thereby preventing it from segregate to oxide–alloy interfaces,237,392 as they do
segregating to the interface. to oxide grain boundaries, it has been suggested393,424
There has been some debate on whether it is that RE at the interface prevents S from segregating
possible for a large and negatively charged sulfur there, by blocking available sites for S segregation,
ion to segregate to an intact oxide–metal inter- similar to surface site competition or poisoning. Pier-
face.346 Indeed, if the interface were always coher- aggi et al.424,425 further suggested that the segregated
ent, it might not have enough space to accommodate large RE ions can pin interfacial dislocations, which
a large segregand, especially one that is negatively are necessary in annihilating interfacial point defects
charged as are the oxygen ions in the oxide. How- generated by the scale growth process. Blocking these
ever, oxide–metal interfaces are not always coher- interfacial reaction steps can then retard the scaling
ent; coherency is often lost after prolonged kinetics and even alter the scaling mode, for example,
oxidation.409 At such interfaces, as with high angle from predominantly cation transport to that of an-
grain boundaries,410,411 segregation may be easier. ion transport. This ‘poisoned interface’ model
When pO2 and interfacial strains are taken into attempts to explain the RE effects on scale growth
consideration, first principle calculations show that kinetics, transport process and adhesion. However,
S segregation to Al2O3/Ni(Al) interfaces is in fact there has not been convincing experimental evidence
thermodynamically favored.412 Ample evidences to support it.
also exist from Auger analysis studies, showing that The presence of RE at the interface may exert a
sulfur can indeed segregate to growing oxide–metal bond strengthening effect. This has been demonstrated
interfaces,413,414 not only on interfacial void sur- experimentally423 on Fe3Al-based alloys that were
faces, but also on intact interfaces. The latter loca- undoped, doped with Zr and desulfurized. When
tion was further vindicated when sulfur was detected interfacial composition and strength were compared,
by analytical transmission electron microscopy (TEM) it was found that, although eliminating the sulfur in
on cross-sectional specimens of Cr2O3415 and the alloy greatly improved scale adhesion, the pres-
Al2O3416–419 scales, indicating that true interfacial seg- ence of RE at the interface further strengthened the
regation of S to an oxide–metal interface is possible, scale–alloy interface. The bond strengthening effect of
even though such detection is often difficult, due to RE at Al2O3–metal interfaces has also been established
beam-induced desorption.417 by earlier theoretical analyses,426 and, recently, by
The effect of S segregation on interfacial strength more rigorous first principle calculations.427
has been demonstrated experimentally on Al2O3 grown All the above discussions on the RE effect have
on NiAl420 and on diffusion bonded Al2O3 on Ni,421 been based mainly on Al2O3- and some on Cr2O3-
showing greatly decreased interfacial strength with forming alloys. A few studies have also been done on
submonolayer of S coverage. Furthermore, first princi- the oxidation of nickel and other pure metals.428 For
ple calculations also demonstrated a large bond- the work on nickel, surface application techniques
weakening effect of S at oxide–metal interfaces.412 were often used, such as sol–gel coatings of RE oxi-
When an alloy is desulfurized, for example by a high des429 or ion-implantation of REs.430 The observed
temperature H2-annealing process, its ability to retain effects are not as strong as those reported for Al2O3-
the scale under severe thermal cycling conditions is also and Cr2O3-forming alloys. Oxidation rates are usu-
greatly improved. This conclusion has been repeatedly ally decreased by 2–3 times and the oxide grain sizes
demonstrated for Al2O3-forming alloys,363 and also for are also decreased, by a factor of 2. Since the adhe-
Cr2O3-forming alloys.422 These results further attest to sion of NiO on Ni is usually good, it is not clear if
the detrimental effect of S on scale adhesion. these surface-applied REs affect NiO adhesion.
1.10.7.2 Nonmetallic Impurities was evaluated using conventional Auger electron

microscopy (AES) with a typical probe size of about
1.10.7.2.1 Sulfur
1 mm, after spalling the oxide scale in ultrahigh vac-
The segregation of indigenous sulfur impurity from
uum (UHV) by scratching an oxidized specimen sur-
an alloy to the scale–alloy interface during high tem-
face,420, 438,443,445,446 or bending a thin strip of oxidized
perature oxidation is often considered the major
sample.414 The chemical composition of the alloy sur-
cause that weakens the interface, a conclusion
face was then examined. With the scratch technique,
drawn mainly from works performed on Al2O3-form-
the underside of the oxide, or the oxide side of the
ing alloys.362,363,408,431 The originally proposed
interface, could also be examined, and that location
hypothesis362,408,431 states that the oxide–metal inter-
was always found to be free from segregands. Any
face is intrinsically strong, but is weakened by the
impurity, if present at the interface, resided on the
segregation of indigenous sulfur, which is often pres-
alloy side after scale spallation. One of these studies
ent in the alloy in parts per million levels, hence
also utilized a field emission AES,439 of probe size
making the scale nonadherent. Evidence of the detri-
30 nm, and verified that sulfur was indeed present
mental effect of sulfur on scale adhesion has been
at the interface, rather than at small interfacial void
provided by the performance of many different
surfaces. Furthermore, the amount of interface sulfur
desulfurized or low S alloys.292,432,433 When the sulfur
was found to be significantly higher than that calcu-
content in the alloy is decreased to less than 3 ppm or
lated from interface sweeping, that is, accumulation
so, often by a high temperature H2-annealing process,
of impurity atoms in the alloy onto the moving
scale spallation under thermal cycling can be signifi-
oxide–metal interface when it advances into the
cantly decreased.433
alloy as a result of an inwardly grown scale. Occasion-
The proposed ‘sulfur effect’ shows how interface
ally, sulfur could also be detected by using scanning
chemistry may affect scale adhesion. Sulfur is well
electron transmission microscopy (STEM) on thin
known to segregate to alloy grain boundaries and
cross-sectional samples,415–419 but such detection is
weaken them,434 and theoretical studies have shown
usually difficult, partly due to the small amount of
bond weakening effects of S on Al2O3–metal inter-
S segregated at the interface, and partly to
faces.412,426 Limited mechanical testing on the
S desorption under the electron beam.417 However,
strength of diffusion-bonded Al2O3–metal interfaces
results from these few TEM investigations have fur-
in the presence of a foreign element has also demon-
ther demonstrated that sulfur can segregate to intact
strated strong effects of interface chemistry.435,436
oxide–alloy interfaces that are pore-free.
However, most of these studies did not characterize
AES studies of the build-up of sulfur as a function
the effect that these segregands had on interface
of oxidation time on several Al2O3–alloy interfaces
microstructure, particularly on the defect type, size,
are summarized in Figure 26, where each datum
and distribution. This is especially important, as more
point was obtained from a different specimen oxi-
debonded areas seem to exist when impurities are
dized for a given time. Although sulfur was found to
present at these interfaces.412,435 As all commercial
be the major segregand at these scale–alloy interfaces,
metals and alloys contain parts per million levels of
the segregation behavior, in terms of rate and amount,
nonmetallic impurities that readily segregate to sur-
varied significantly for different alloys and differed
faces when the alloy is heated; it is often assumed that
from surface segregation. A few general conclusions
the same kind of segregation takes place at oxide–
can be made from these interfacial segregation
metal interfaces. Whether this is always the case is
results: (i) On MCrAl type alloys, the amount of
not obviously clear. Some researchers have ques-
interfacial sulfur is high due to strong co-segregation
tioned whether it is energetically favorable for a
of S with Cr413,438; cosegregation of C also occurs on
large and electronically negative sulfur to segregate
FeCrAl during cooling, but not on NiCrAl, due to the
to an oxide–metal interface.346,437
faster C diffusion rate in the bcc FeCrAl than in the
Studies of the chemical changes at Al2O3–alloy
fcc NiCrAl. (ii) On Fe or Ni aluminides, S is only
interfaces as a function of oxidation time have in
present at the Al2O3–alloy interface when a complete
recent years been carried out for MCrAl type of alloys,
layer of a-Al2O3 is developed.413,420 On FeAl, the
where M is Fe414,438 or Ni,413 iron aluminides,439–441
amount saturated at the Al2O3–alloy interface is
nickel aluminides,420,442 or NiPtAl443,444; a few studies
about 0.5 monolayer, while much higher concentra-
have also been conducted on Cr2O3/NiCr445 and
tions are found on surfaces of interfacial voids due to
Cr2O3/Cr.446 In most cases, the interfacial chemistry
30 10 ferritic steel, and promoted the formation of a pro-

tective Cr2O3 layer.448 In FeCrAl, samples containing
greater than 100 ppm sulfur seemed to develop Al2O3
25
Sulfur at Al2O3–alloy interfaces (at%)

Fe-20Cr-10Al
Sulfur at Al2O3–alloy interfaces (at%)
8 scales that were more adherent than those with 3 or

30 ppm S.449 It is not clear if this ‘improved adhesion’
Ni-15Cr-13Al
20 is related to the formation of a thin layer of sulfide at
6 the Al2O3–alloy interface; the mechanism is rather
unclear and the method of determining scale adhe-
15 sion not quantitative.
Fe-40Al Sulfur has also been suggested to affect the growth
4
10 mechanism of Al2O3 or Cr2O3 scales. Some studies
Ni-40Al have reported that the effect of excess sulfur, in the
alloy or from the oxidizing atmosphere, is to produce
2
5 a less protective surface oxide film, as seen from
the outgrowth of Cr2O3 nodules on a layer of Al2O3
Ni-50Al NiAl + Pt FeAl + Zr on a NiCrAl alloy,450 or an Al2O3 layer that was more
0
0.1 1 10 FeCrAlY 100 NiAl + Hf easily cracked at the oxidation temperature.432 These
Oxidation time (h) results can be explained by the less adherent scales
Figure 26 Summary of the amount of sulfur detected at that usually form in the presence of sulfur. However,
different alumina–alloy interfaces as a function of in other cases where scale damage or breakaway was
oxidation time. not apparent or not studied, excess sulfur seemed to
affect transport rates through the oxide, giving rise
to thicker scales with higher S concentrations in
cosegregation of Al and S.439 On NiAl, the saturation
the alloy.417 Consequently, sulfur in the oxide scale,
level is lower, only about 0.2 monolayer, which may
particularly at grain boundaries, was suggested to
be related to the lower level of sulfur impurity in
enhance cation outward transport,415,417,451 where
nickel aluminides compared with that in iron alumi-
S acts as an electron donor and increases the number
nides (a few parts per million vs. tens of parts per
of Al vacancies.451 However, the exact effect of S on
million). (iii) Segregation on NiAl depends on alloy
oxide growth rates is not fully substantiated, for there
stoichiometry, in that no S segregates at a-Al2O3/Ni–
are cases where sulfur was detected everywhere
50Al interfaces, but small amounts are present at a-
within the oxide layer, but it did not show any effect
Al2O3/Ni–40Al.420 Finally, Pt443 and REs (Hf, Zr, and
on scale growth rate.418
Y)423 prevent S segregation to the interface. Mechani-
cal testing on some of these alloy systems directly 1.10.7.2.2 Other impurities
related the amount of S at the interface to interfacial For all of the metals discussed in this chapter under
strengths,420,423 and found that the strength decreases most conditions, oxidation rates are controlled by
most sharply with the first few submonolayers of sulfur. diffusional transport across the oxide scale. There-
Another related detrimental effect of sulfur on scale fore, impurities that are incorporated into the grow-
adhesion is its ability to enhance interfacial void for- ing oxide must have an effect on the rates of cation
mation, probably by decreasing the energy of void and/or anion diffusion, on oxide creep and perhaps
nucleation by decreasing the metal surface energy also on the oxide microstructure. These impurities
when S is segregated to a void embryo.332,345 In fact, can be segregated at oxide grain boundaries, as in the
different types of common surface impurities, such as case of REs (Section 1.10.7.1), or dissolve in the
C, P, Cl, and Na, have also been shown to greatly oxide, to affect grain boundary transport, diffusional
increase the interfacial pore density on FeAl.345,447 creep, and/or change the oxide defect concentration,
Although it is now clear that sulfur, as an impurity thereby affecting the rates of cation or anion trans-
in the alloy, can segregate to the growing oxide–metal port through the oxide film. For NiO, it is clear that
interface to enhance interfacial void formation and impurities enhance duplex scale formation452 (see
decrease the interfacial bond strength, a few studies Section 1.10.2.1), in ways that are still not clearly
have shown a surprising beneficial effect of sulfur. For understood.
example, under steam oxidation environments, Cr Although the type and concentration of impurities
cosegregated with S to the alloy surface, a high Cr must be important in high temperature oxidation,
evaluations of their effects have not been made and chemistry affect their strength. The above exam-
systematically. Ample examples on the effect of ple is only one of many that can be raised. While
impurities, whether on sample surfaces from the engineering investigations are often system specific,
environment306,345 or from within the alloy,453,454 scientific research should provide knowledge that has
serve to underline their importance, as well as the broad relevance. Coupling of phenomena is an area
complexity of the oxidation process. that has received little attention. For example, it has
been proposed328 that the magnitude of the oxidation
stress should be related to the oxide microstructure
1.10.8 Concluding Remarks and vice versa. How structure and chemistry affect
stress generation and how the stress in turn affects
Oxidation is considered by some as a perfect topic structure due to relaxation are questions that remain
for teaching material science, because it covers all unclear. Many such couplings can be identified. Study-
aspects of the field, from basic thermodynamics to ing them undoubtedly raises the level of experimental
gas–solid interaction, defect chemistry, similar and difficulty, but with increasingly advanced instrumen-
dissimilar interfaces, bulk and short circuit diffusions, tation and theoretical tools, new results and insights
phase transformation, deformation, fracture mechan- should be attainable.
ics and more. Indeed, the oxidation of even pure
metals and simple binary alloys involves extremely
References
complicated processes, many of which are interre-
lated. The summary given here and the many refer-
1. Kubaschewski, O.; Hopkins, B. E. Oxidation of Metals
ences sited are intended to provide future researchers and Alloys, 2nd ed.; Academic Press: New York, 1962.
with a foundation of the extensive works that have 2. Kofstad, P. High Temperature Corrosion; Elsevier
been done, so that the same conclusions need not be Applied Science: London, UK, 1988.
3. Birks, N.; Meier, G. H.; Pettit, F. S. Introduction to the
repeated. In every section of this chapter, many ques- High-Temperature Oxidation of Metals, 2nd ed.;
tions obviously remain for future scientists to unravel. Cambridge University Press: Cambridge, UK, 2006.
It is important to bear in mind that quantitative analy- 4. Kofstad, P. Nonstoichiometry, Diffusion, and Electrical
Conductivity in Binary Metal Oxides; Wiley-Interscience:
sis is extremely important. Although some aspects of London, 1972; Chapter 11, pp 221–257.
the oxidation processes are difficult to quantify, one 5. Wood, G. C.; Wright, I. G.; Ferguson, J. M. Corros. Sci.
should strive to do it as much as possible, even if 1965, 5, 645–661.
6. Vasyutinskiy, B. M.; Kartmazov, G. N. Fiz. Metal
semiquantitatively. It is also important to always con- Metalloved 1963, 15, 132.
sider the dynamic nature of the oxidation process and 7. Atkinson, A.; Taylor, R. I.; Goode, P. D. Oxid. Met. 1979,
take notice of the knowledge gained from other dis- 13, 519–537.
8. Harris, A. W.; Atkinson, A. Oxid. Met. 1990, 34, 229–258.
ciplines, as oxidation covers a wide range of scientific 9. Sawhill, H. T.; Hobbs, L. W.; Tinker, M. T. Adv. Ceram.
principles. 1983, 6, 128–138.
What is most lacking in past research, in my view, 10. Moon, D. P. Oxid. Met. 1989, 31, 71–89.
11. Kim, C. K.; Fan, S. K.; Hobbs, L. W. In Microscopy of
is the understanding of the coupling of different Oxidation; Bennett, M. J., Lorimer, G. W., Eds.; Institute
phenomena and the identification of the generality of Metals: London, UK, 1991; pp 374–380.
of certain phenomenon, that is, how and/or whether 12. Peraldi, R.; Monceau, D.; Pieraggi, B. Mater. Sci. Forum
2001, 369–372, 189–196.
an experimentally observed trend for one system 13. Mitchell, D. F.; Sewell, P. B.; Cohen, M. Surf. Sci. 1976,
applies to others. An example of this is the sulfur 61, 355.
effect, which has been extensively studied and sub- 14. Mitchell, D. F.; Graham, M. J. Surf. Sci. 1982, 114,
546–562.
stantiated for Al2O3 formers, but it is not certain if the 15. Martius, U. M. Can. J. Physics 1956, 33, 466–472.
same applies for Cr2O3, NiO, Fe2O3, or other sys- 16. Balmain, J.; Savall, C.; Molins, R.; Severac, C.; Haut, C.;
tems. Related to this question is the nature of the Huntz, A. M. Mater. Sci. Forum 2001, 369–372, 125–132.
17. Peraldi, R.; Monceau, D.; Pieraggi, B. Oxid. Met. 2002,
oxide–metal interface on different types of alloys. 58, 249–273.
There seems to be a myth that interfaces formed on 18. Ueno, T. Trans. Jpn. Inst. Met. 1973, 14, 267.
alloys whose scales grow mainly by anion transport 19. Lebrun, J. L.; Huntz, A. M.; Beranger, G.; Aubry, A.;
Zhao, J. G. In Residual Stresses in Science and
are stronger than those that grow by cation transport. Technology; Ceck, G., Denis, D., Simon, A., Eds.;
Quantitative experimental work has to be conducted Elsevier Science: London, UK, 1988; p 135.
to provide answers. First principle calculations may 20. Czerwinski, F.; Zhilyaev, A.; Szpunar, J. A. Corr. Sci.
1999, 41, 1703–1713.
also be employed to help elucidate many questions 21. Desserrey, F.; Valot, C.; Montesin, T.; Larpin, J. P. Mater.
related to metal–oxide interfaces and how structure Sci. Forum 2002, 408–412, 1007–1012.
22. Hobbs, L. W.; Sawhill, H. T.; Tinker, M. T. Rad. Effects 60. Wagner, C.; Koch, E. Z. Phys. Chem. 1936, 1332, 439.
1983, 74, 291–297. 61. Mrowec, S.; Werber, T. Roczniki Chem. 1962, 36, 159.
23. Homma, H.; Khoi, N. N.; Smeltzer, W. W.; Embury, J. D. 62. Przybylski, K.; Szwagierczak, D. Oxid. Met. 1982, 17,
Oxid. Met. 1971, 3, 463–473. 267–295.
24. Kim, C. K.; Fan, S. K.; Hobbs, L. W. In Microscopy of 63. Gulbransen, E. A.; Ruka, R. Trans. AIME 1950, 188, 1500.
Oxidation; Bennett, M. J., Lorimer, G. W., Eds.; Institute 64. Pinder, L. W. C.E.G.B. Report, MID/SSD/80/0050/R,
of Metals: London, UK, 1991; pp 374–380. August 1980.
25. Hou, P. Y.; Cannon, R. M. Mater. Sci. Forum 1997, 251– 65. Jonsson, T.; Jardnas, A.; Svensson, J. -E.;
254, 325–332. Johansson, L. -G.; Halvarsson, M. Oxid. Met. 2007, 67,
26. Stawbridge, A.; Stott, F. H.; Wood, G. C. Corros. Sci. 193–214.
1993, 35, 855–862. 66. Vallet, P.; Raccah, P. Mem. Sci. Rev. Metall. 1965, 67, 1.
27. Kyung, H.; Kim, C. K. Mater. Sci. Eng. 2000, B76, 173–183. 67. Swarooop, B.; Wagner, J. B., Jr. Trans. AIME 1967, 239,
28. Mrowec, S. Corros. Sci. 1967, 7, 563. 215.
29. Evans, A. G.; Rajdev, D.; Douglass, D. L. Oxid. Met. 1972, 68. Bransky, I.; Hed, A. Z. J. Am. Ceram. Soc. 1968, 57, 231.
4, 151. 69. Fender, B. E.; Riley, F. D. J. Phys. Chem. Solids 1969, 30,
30. Kofstad, P. Oxid. Met. 1985, 24, 265. 793.
31. Atkinson, A.; Smart, D. W. J. Electrochem. Soc. 1988, 70. Giddings, R. A.; Gordon, R. S. J. Am. Ceram. Soc. 1973,
135, 2886. 56, 111.
32. Robertson, J.; Manning, M. I. Mater. Sci. Technol. 1988, 71. Dieckmann, R. In High Temperature Corrosion;
4, 1064. Rapp, R. A., Ed.; NACE: Houston, TX, 1983; p 70.
33. Gibbs, G. B.; Hales, R. Corros. Sci. 1977, 17, 487. 72. Dieckmann, R. In Oxidation of Metals and Associated
34. Yurek, G. J.; Schmalzried, H. Ber. Bunsenges. Phys. Mass Transport; Dayananda, M. A., Rothman, S. J.,
Chem. 1975, 79, 255. King, W. E., Eds.; Metallurgical Society: Warrendale, PA,
35. Caplan, D.; Hussey, R. J.; Sproule, G. I.; Graham, M. J. 1986; p 7.
Oxid. Met. 1980, 14, 279–299. 73. Garnaud, G.; Rapp, R. A. Oxid. Met. 1977, 11, 193.
36. Peraldi, R.; Monceau, D.; Pieraggi, B. Oxid. Met. 2002, 74. Atkinson, A. Rev. Mod. Phys. 1985, 57, 437–470.
58, 275–295. 75. Atkinson, A.; Taylor, R. I. High Temperature – High
37. Atkinson, A.; Taylor, R. I.; Goode, P. D. Oxid. Met. 1979, Pressure 1985, 14, 469.
13, 519–543. 76. Atkinson, A.; Taylor, R. I. J. Phys. Chem. Solids 1985, 46,
38. Atkinson, A.; Taylor, R. I.; Hughes, A. E. Philos. Mag. A 469.
1982, 45, 823–833. 77. Channing, D. A.; Graham, M. J. Corros. Sci. 1972, 12,
39. Gulbransen, E. A.; Andrew, K. F. J. Electrochem. Soc. 271.
1954, 101, 128. 78. Channing, D. A.; Dickerson, S. M.; Graham, M. J. Corros.
40. Herchl, R.; Khoi, N. N.; Homma, T.; Smeltzer, W. W. Oxid. Sci. 1973, 13, 933.
Met. 1972, 4, 35. 79. Francis, R.; Lees, D. G. Corros. Sci. 1976, 16, 847.
41. Rhines, F. N.; Connell, R. G., Jr. J. Electrochem. Soc. 80. Sewell, P. B.; Cohen, M. J. Electrochem. Soc. 1964, 111,
1977, 124, 1122. 501–508.
42. Horn, I. Z. Z. Metalkunde 1949, 40, 73. 81. Caplan, D.; Graham, M. J.; Cohen, M. Corros. Sci. 1970,
43. Wood, G. C.; Stott, F. H.; Forrest, J. E. Werk. Korr. 1977, 10, 1.
28, 395–404. 82. Svedung, I.; Hammar, B.; Vannerberg, N. G. Oxid. Met.
44. Atkinson, A.; Taylor, R. I. J. Phys. Chem. Solids 1986, 47, 1973, 6, 21.
315. 83. Howe, C. I.; McEnany, B.; Scott, V. D. Corros. Sci. 1985,
45. Evans, U. R. Symposium on Internal Stresses in Metals 25, 195.
and Alloys, London, 1947, pp 291–310. 84. Boggs, W. E.; Kachik, R. H.; Pellisier, G. E.
46. Ueno, T. Trans. Jpn. Inst. Met. 1974, 15, 167–172. J. Electrochem. Soc. 1965, 112, 539.
47. Rhines, F. N.; Wolf, J. S. Metall. Trans. 1970, 1, 85. Rahmel, A.; Robolski, J. Corros Sci. 1965, 5, 333.
1701–1710. 86. Fukumoto, M.; Hayashi, S.; Maeda, S.; Narita, T. J. Inst.
48. Ueno, T. Jpn. J. Appl. Phys. 1975, 14, 2081–2082. Met. 2001, 65, 115.
49. Liu, C.; Huntz, A. M.; Lebrun, J. L. Mater. Sci. Eng. A 87. Tuck, C. W.; Odgers, M.; Sachs, K. Corros. Sci. 1969,
1993, 160, 113–126. 9, 271.
50. Goedjen, J. G.; Stout, J. H.; Guo, Q.; Shores, D. A. Mater. 88. Wittnauer, J. J. Met 1900, 42, 7.
Sci. Eng. A 1994, 177, 115–124. 89. Buckman, R. W., Jr. In Alloying; Walter, J. L.,
51. Homma, T.; Pyun, Y. J. In Proceedings of JIMIS-3: High Jackson, M. R., Sims, C. T., Eds.; ASM International:
Temperature Corrosion Trans. Jpn. Inst. Met., Suppl. Metals Park, 1988; p 419.
1983, 161–166. 90. Pugh, J. J. Met. 1958, 212, 335.
52. Wolf, J. S.; Cavin, B. In Advances in X-ray Analysis; 91. Nieh, T. G.; Wadsworth, J. In High Temperature Silicides
Golfrich, J. V., et al. Ed.; Plenum, 1994; p 449. and Refractory Alloys; Proceedings of the Matererial
53. Przybilla, W.; Schutze, M. Oxid. Met. 2002, 58, 103–145. Research Society Symposium; Briant, C. L.,
54. Carter, R. E.; Richardson, F. D. J. Met 1954, 6, 1244. Petrovic, J. J., Bewlay, B. P., Vasudevan, A. K.,
55. Phalnikar, C. A.; Evans, E. B.; Baldwin, W. M. Lipsitt, H. A., Eds.; Mater. Res. Soc.: Pittsburgh, PA,
J. Electrochem. Soc. 1956, 103, 367–429. 1994; Vol. 322, p 315.
56. Hsu, H. S.; Yurek, G. J. Oxid. Met. 1982, 17, 55–76. 92. Ramberg, C. E.; Beatrice, P.; Kurokawa, K.;
57. Martin, M.; Hilbrandt, N.; Koops, U. In Proceedings of the Worrell, W. L. In High Temperature Silicides and
Symposium on High Temperature Corrosion and Refractory Alloys; Proceedings of the Matererial
Materials Chemistry; Electrochemical Society: Research Society Symposium, Boston, MA, Nov.–2 Dec.
Pennington, NJ, 1998; pp 106–117. 1993; Briant, C. L., Petrovic, J. J., Bewlay, B. P.,
58. Kofstad, P. K.; Hed, A. Z. J. Electrochem. Soc. 1969, 116, Vasudevan, A. K., Lipsitt, H. A., Eds.; Material Research
224–229. Society: Pittsburgh, PA, 1994; pp 243–253.
59. Pettit, F. S.; Wagner, J. B. Acta Metall. 1964, 12, 41. 93. Kofstad, P. J. Inst. Met. 1962, 90, 253–264.
94. Pawel, R. E.; Cathcart, J. V.; Campbell, J. J. Acta Metall. 133. Ux, M.; Wittenauer, J. P. J. Met. 1992, 44, 25.
1962, 10, 149. 134. Mueller, A.; Ge, W.; Rapp, R. A.; Courtright, E. L.
95. Pint, B. A.; DiStefano, J. R. Oxid. Met. 2005, 63, 33–55. J. Electrochem. Soc. 1992, 139, 1266–1275.
96. Kofstad, P. J. Electrochem. Soc. 1963, 110, 491. 135. Louro, C.; Cavaleiro, A. J. Mater. Proc. Tech. 1999,
97. Kofstad, P. J. Electrochem Soc. 1962, 109, 776. 92–93, 162–168.
98. Pawel, R. E.; Campbell, J. J. Acta Metall. 1966, 14, 1827. 136. Martinz, H. -P.; Rieger, M. Mater. Sci. Forum 1997,
99. Cathcart, J. V.; Campbell, J. J.; Smith, G. P. 251–254, 761–768.
J. Electrochem. Soc. 1958, 105, 442. 137. Pons, M.; Caillet, M.; Galerie, A. Mater. Chem. Phys.
100. Kofstand, P. J. Less Common Met. 1963, 5, 158. 1986, 15, 45–60.
101. Inouye, H. In Cohimbium Metallurgy; AIME Metallurgical 138. Pons, M.; Caillet, M.; Galerie, A. Mater. Chem. Phys.
Society Conferences, 10; Douglass, D. L., Kunz, F. W., 1987, 16, 423–432.
Eds.; Interscience Publishers: New York, 1961. 139. Fukumoto, M.; Matsumura, Y.; Hayashi, S.;
102. Kofstad, P.; Espevik, S. J. Electrochem. Soc. 1965, Sakamoto, K.; Kasama, A.; Tanaka, R.; Narita, T. Oxid.
112, 153. Met. 2003, 60, 355.
103. Earl, A.; Gulbransen, A.; Andrew, K. F. J. Electrochem. 140. Deal, B. E.; Grove, A. S. J. Appl. Phys. 1965,
Soc. 1958, 105, 4–9. 36, 3770.
104. Sheasby, J. S.; Wallwork, G. R.; Smeltzer, W. W. 141. Opila, E. J. J. Am. Ceram. Soc. 1994, 77, 730.
J. Electrochem. Soc. 1966, 113, 1255. 142. Duquesnoy, A.; Marion, F. C.R. Hebd. Seances Acad.
105. Strafford, K. N. Corros. Sci. 1979, 19, 49–62. Sci. 1963, 256, 2862.
106. King, D. A.; Madey, T. E.; Yates, J. T., Jr. J. Chem. Phys. 143. Kroger, F. A. In High Temperature Corrosion; Rapp, R. A.,
1971, 55, 3247–3253. Ed.; NACE: Houston, 1983; p 89.
107. Sikka, V. K.; Rosa, C. J. Corros. Sci. 1980, 20, 1201–1219. 144. Sucov, E. W. J. Am. Ceram. Soc. 1965, 46, 190.
108. Schissel, P. O.; Trulson, O. C. J. Chem. Phys. 1965, 43, 145. Haul, R.; Dumbgen, Z. Elektrockem 1962, 66, 636.
737–743. 146. Robinson, R. C.; Smialek, J. L. J. Am. Ceram. Soc. 1999,
109. Walsh, P. N.; Quets, J. M.; Graff, R. A. J. Chem. Phys. 82, 1817–1825.
1967, 46, 1144–1153. 147. Stott, F. H.; Wood, G. C.; Stringer, J. Oxid. Met. 1995, 44,
110. Bartlett, R. W. Trans. Metal Soc., AIME 1964, 230, 1097. 113–145.
111. Bartlett, R. W. Trans. Metall. Soc. AIME 1964, 230, 1097. 148. Evans, H. E.; Hilton, D. A.; Holm, R. A.; Webster, S. J.
112. Perkins, R. A.; Crooks, D. D. J. Met. 1961, 13, 490. Oxid. Met. 1980, 14, 235.
113. Zhetbaev, A. K.; Ibragimov, S. H. S. H.; Shokanov, A. K. 149. Bennett, M. J.; Desport, J. A.; Labun, P. A. Math. Phy.
J. Phys. 1980, 41, 387–388. Sci. 1987, 412, 223–230.
114. Berkowitz-Mattuck, J. B.; Cuchler, A.; Engelke, J. L.; 150. Evans, H. E.; Hilton, D. A.; Holm, R. A.; Webster, S. J.
Goldstein, N. J. Chem. Phys. 1963, 39, 2722–2730. Oxid. Met. 1983, 19, 1.
115. Voitovich, R. F. Fiz. Metal. Metalloved 1961, 12, 376. 151. Adachi, T.; Meier, G. H. Oxid. Met. 1987, 27, 347.
116. Kofstad, P. J. Inst. Met. 1962–63, 91, 411. 152. Douglass, D. L.; Armijo, J. S. Oxid. Met. 1970, 2, 207.
117. Dooley, R. B.; Stringer, J. J. Less Common Met. 1971, 153. Saito, Y.; Maruyama, T. Mater. Sci. Eng. 1987, 87, 275.
25, 15–26. 154. Ahmad, B.; Fox, P. Oxid. Met. 1999, 52, 113.
118. Klopp, W. D.; Maykutii, D. J.; Jaffeet, R. I. Trans. Am. 155. Stott, F. H.; Gabriel, G. J.; Wood, G. C. Oxid. Met. 1987,
Soc. Met. 1961, 53, 637. 28, 329–345.
119. Corn, D. L.; Douglass, D. L.; Smith, C. A. Oxid. Met. 1991, 156. Adachi, T.; Meier, G. H. Oxid. Met. 1987, 27, 347.
35(1–2), 139–173. 157. Saito, Y.; Maruyama, T.; Amano, T. Mater. Sci. Eng.
120. DiStefano, J. R.; Chitwood, L. D. J. Nucl. Mater. 2001, 1987, 87, 275–280.
295, 42–48. 158. Meier, G. Mater. Sci. Eng. 1989, 120, 1.
121. DiStefano, J. R.; Chitwood, L. D. AIP. AIP Conference 159. Sauthoff, G. Intermetallics; VCH: New York, 1995;
Proceedings, no. 552, 2001; pp 1076–1081. [Conference pp 99–117.
Paper; Journal Paper]. 160. Willaims, J. C. In Structural Intermetallics; Nathal, M. V.,
122. DiStefano, J. R.; Pint, B. A.; DeVan, J. H. Int. J. Refract. Darolia, R., Liu, C. T., Martin, P. E., Miracle, D. B.,
Met. Hard Mater. 2000, 18(4–5), 237–243. Wagner, W., Yamaguchi, M., Eds.; TMS: Warrendale, PA,
123. Babitzke, H. R.; Siemens, R. E.; Asai, G.; Kato, H. Bureau 1997; pp 3–8.
of Mines Report of Investigations 6558, US Department 161. Jiang, H.; Petersson, C. S.; Nicolet, M. A. Thin Solid Films
of the Interior, 1964. 1986, 140, 115.
124. Nieh, T. G.; Wadsworth, J. In High Temperature Silicides 162. Meier, G. H. Mater. Corros. 1996, 47, 595.
and Refractory Alloys; MRS Proceedings; Briant, C. L., 163. Brady, M. P.; Pint, B. A.; Tortorelli, P. F.; Wright, I. G.;
Petrovic, J. J., Bewlay, B. P., Vasudevan, A. K., Hanrahan, R. J., Jr. In Corrosion and Environmental
Lipsitt, H. A., Eds.; 1993; Vol. 322, p 315. Degradation; Materials Science and Technology; Vol II,
125. Svedberg, R. C.; Foroulis, Z. A.; Pettit, F. S. Schutze, M., Ed.; Wiley-VCH: Germany, 2000,
J. Electrochem. Soc. 1976, 331. pp 296–308.
126. Perkins, R. A.; Meier, G. H. J. Met. 1990, 42, 17–21. 164. Lublin, P.; Sama, L. Am. Ceram. Soc. 1967, 46, 1083.
127. Perkins, R. A.; Chiang, K. T.; Meier, G. H. Scripta Metall. 165. Lavendel, H. W.; Elliot, A. G. Trans AIME 1967, 239, 143.
1988, 22, 419–424. 166. Schwettmann, F. N.; Graff, R. A.; Kolodney, M.
128. Lee, J. S.; Stephens, J. J.; Nieh, T. G. In High J. Electrochem. Soc. 1971, 118, 1973–1977.
Temperature Niobium Alloys; Stephens, J. J., Ahmad, I., 167. Abba, A.; Galerie, A.; Cailet, M. Oxid. Met. 1982, 17,
Eds.; TMS: Warrendale, PA, 1991; p 143. 43–54.
129. Diliberto, S.; Rapin, C.; Steinmetz, P.; Vilasi, M.; 168. Evans, D. L. J. Am. Ceram. Soc. 1970, 53, 418.
Berthod, P. J. Mater. Sci. 2003, 38, 2063–2072. 169. Meschter, P. J. Metall. Trans. A. 1992, 23A, 1763.
130. Perkins, R. A.; Meier, G. H. J. Met. 1990, 42, 17. 170. Glushko, P. I.; Postogvard, G. I.; Pugachev, N. S.;
131. Wittenauer, J. J. Met. 1990, 42, 7. Dudnik, S. F.; Podtykan, V. P. Prot. Met. 1977, 13, 187.
132. Wadsworth, J.; Nieh, T. G.; Stephens, J. J. Int. Met. Rev. 171. Melsheimer, S.; Fietzek, M.; Kolarik, V.; Rahmel, A.;
1988, 33, 131. Schutze, M. Oxid. Met. 1997, 47, 139.
172. Berkowitz-Mattuck, J.; Dils, R. R. J. Electrochem. Soc. 206. Fox, P.; Lees, D. G.; Lorimer, G. W. Oxid. Met. 1991, 36,
1965, 112, 583. 491.
173. Mochizuki, T.; Kashiwagi, M. J. Electrochem. Soc. 1980, 207. Caplan, D.; Harvey, A.; Cohen, M. Corros. Sci. 1963, 3, 161.
127, 1128. 208. Caplan, D.; Sproule, G. I. Oxid. Met. 1975, 9, 459–472.
174. Ramberg, C. E.; Worrell, W. L. J. Am. Ceram. Soc. 2001, 209. Wright, I. G.; Wood, G. C. Oxid. Met. 1977, I1, 163.
84, 2607–2616. 210. Wood, G. C.; Hodgkiess, T.; Whittle, D. P. Corros. Sci.
175. Aitken, E. A. In Intermetallic Compounds; 1966, 6, 129–147.
Westrook, J. H., Ed.; Wiley: New York, 1967; p 491. 211. Wood, G. C.; Hodgkiess, T. J. Electrochem. Soc. 1966,
176. Berkowitz-Mattuck, J. B.; Blackburn, P. E.; Felten, E. J. 113, 319.
Trans. AIME 1965, 233, 1093. 212. Moulin, P.; Armanet, F.; Beranger, G.; Lacombe, P. Mere.
177. Berztiss, D. A.; Cerchiara, R. R.; Gulbransen, E. A.; Sci. Rev. Metall. 1977, 74, 143.
Pettit, F. S.; Meier, G. H. Mater. Sci. Eng. A 1992, 155, 165. 213. Warzee, M.; Maurice, M.; Sonnen, C.; Waty, J.;
178. Yanagihara, K.; Przybylski, K.; Maruyama, T. Oxid. Met. Berge, P. H. Rev. Met. 1964, 61, 593.
1997, 47, 277. 214. Stott, F. H.; Bartlett, P. K. N.; Wood, G. C. Mater. Sci.
179. Kurokawa, K.; Houzumi, H.; Saeki, I.; Takahashi, H. Eng. 1987, 8, 163.
Mater. Sci. Eng. A 1999, 261, 292–299. 215. Hou, P. Y. unpublished results.
180. Bartlett, R. W.; McCamont, I. W.; Gage, P. R. J. Am. 216. Howes, V. R. Corros. Sci. 1968, 8, 221.
Ceram. Soc. 1965, 48, 551. 217. Howes, V. R. Corros. Sci. 1968, 8, 729.
181. McKamey, C. G.; Tortorelli, P. F.; DeVan, J. H.; 218. Mortimer, D.; Post, M. L. Corros. Sci. 1968, 8, 499.
Carmichael, C. A. J. Mater. Res. 1992, 7, 2747. 219. Wood, G. C.; Whittle, D. P. J. Electrochem. Soc. 1968,
182. Dimiduk, D. M.; Perepezko, J. H. MRS Bull 2003, 28, 639. 115(126), 133.
183. Thom, A. J.; Summers, E.; Akinc, M. Intermetallics 2002, 220. Wood, G. C. Oxid. Met. 1970, 2, 11–57.
10, 555. 221. Daghigh, S.; Lebrun, J. L.; Huntz, A. M. Mater. Sci. Forum
184. Supatarawanich, V.; Johnson, D. R.; Liu, C. T. 1997, 251, 381–388.
Intermetallics 2004, 12, 721. 222. Hou, P. Y.; Stringer, J. Act. Metall. 1991, 39, 841–849.
185. Parthasarathy, T. A.; Mendiratta, M. G.; Dimiduk, D. M. 223. Kitamura, K.; Nishiyama, Y.; Otsuka, N.; Kudo, T. Mater.
Acta Metall. 2002, 50, 1857. Sci. Forum 2006, 522–523, 489–495.
186. Sakidja, R.; Rioult, F.; Werner, J.; Perepezko, J. H. 224. Tortorelli, P. F.; More, K. L.; Specht, E. D.; Pint, B. A.;
Scripta Mater. 2006, 55, 903. Zschack, P. Mater. High Temp. 2003, 20, 303–310.
187. Young, C. T.; Tenney, D. R.; Herring, H. W. Metall. Trans. 225. Hou, P. Y.; Paulikas, A. P.; Veal, B. W. Mater. Sci. Forum
1975, 6A, 2253. 2004, 461–464, 671–680.
188. Caplan, D.; Cohen, M. J. Electrochem. Soc. 1961, 108, 226. Zhu, D.; Stout, J. H.; Nelson, J. C.; Shores, D. A. Trans.
438. Tech. Pub. 1997, 251–254, 437–444.
189. Opila, E. J.; Jacobson, N. S. In Fundamental Aspects of 227. Tortorelli, P. F.; Natesan, K. Mater. Sci. Eng. 1998,
High Temperature Corrosion; Shores, D. A., Rapp, R. A., A258, 115.
Hou, P. Y., Eds.; The Electrochemical Society: 228. DeVan, J. H.; Tortorelli, P. F. Corros. Sci. 1993, 35, 1065.
Pennington, NJ, 1997; Vol. 96-26, pp 269–280. 229. Pint, B. A.; Wright, I. G.; Lee, W. Y.; Zhang, Y.;
190. Kofstad, P.; Bredesen, R. Solid State Ionics 1992, 52, 69. Pruessner, K.; Alexander, K. B. Mater. Sci. Eng. A 1998,
191. Huang, K.; Hou, P. Y.; Goodenough, J. B. Solid State 245, 201.
Ionics 2000, 129, 237. 230. Doychak, J.; Smialek, J. L.; Mitchell, T. E. Metall. Trans. A
192. Brylewski, T.; Nanko, M.; Maruyama, T.; Przybylski, K. 1989, 20, 499.
Solid State Ionics 2001, 143, 131. 231. Rybicki, G. C.; Smialek, J. L. Oxid. Met. 1989, 31, 275.
193. Quadakkers, W. J.; Piron-Abellan, J.; Shemet, V.; 232. Yang, J. C.; Nadarzinski, K.; Schumann, E.; Rühle, M.
Singheiser, L. Mater. High Temp. 2003, 20, 115–127. Scripta Metall. 1995, 33, 1043.
194. Kofstad, P.; Lillerud, K. P. J. Electrochem. Soc. 1980, 233. Smialek, J. L.; Doychak, J.; Gaydosh, D. J. Oxid. Met.
127, 2410–2419. 1990, 34, 259.
195. Hagel, W. C.; Seybolt, A. U. J. Electrochem. Soc. 1961, 234. Kuenzly, J. D.; Douglass, D. L. Oxid. Met. 1974, 8, 139.
108, 1146. 235. Doychak, J.; Rühle, M. Oxid. Met. 1989, 31, 431.
196. Atkinson, A.; Taylor, R. I. In Proceedings of the 3rd 236. Schumann, E.; Rühle, M. Acta Metall. Mater. 1994, 42,
International Conference on Nonstoichiometric 1481.
Compounds, State College: PA, 1984. 237. Alexander, K. B.; Prussner, K.; Hou, P. Y.; Tortorelli, P. F.
197. Hughes, A. E.; Atkinson, A.; Chadwick, A. T. In Defect In Microscopy of Oxidation; Newcomb, S. B., Little, J. A.,
Properties and Processing of High-Technology Eds.; The Institute of Metals, 1997; Vol. 3, pp 246–264.
Nonmetallic Materials; North-Holland: New York, NY, 238. Zhang, X. F.; Thaidigsmann, K.; Ager, J.; Hou, P. Y.
1984; pp 27–37. J. Mater. Res. 2006, 21, 1409–1419.
198. Giggins, C. S.; Pettit, F. S. Metall. Trans. 1971, 2, 1071. 239. Renusch, D.; Grimsditch, M.; Koshelev, I.; Veal, B. W.;
199. Tsai, S. C.; Huntz, A. M.; Dolin, C. Mater. Sci. Eng. A. Hou, P. Y. Oxid. Met. 1997, 48, 471.
1996, 212, 6–13. 240. Wagner, C. Corros. Sci. 1965, 5, 751.
200. Dong, W.; Bishop, H. E.; Johnson, D.; Lees, D. G.; 241. Hou, P. Y. J. Am. Ceram. Soc. 2003, 86, 660–668.
Lorimer, G. W. Oxid. Met. 2000, 54, 509. 242. Wood, G. C.; Hodgkiess, T. Nature 1966, 211,
201. Lillerud, K. P.; Kofstad, P. J. Electrochem. Soc. 1980, 1358–1361.
127, 2397–2410. 243. Sauer, J. P.; Rapp, R. A.; Hirth, J. P. Oxid. Met. 1982, 18,
202. Hindam, H.; Whittle, D. P. Oxid. Met. 1982, 18, 285–294.
245–262. 244. Jedlinski, J.; Borchardt, G.; Bernaski, A.; Scherrer, S.;
203. Evans, A. G.; Cannon, R. M. Mater. Sci. Forum 1989, 43, Ambos, R.; Rajchel, B. In Microscopy of Oxidation;
243–268. Newcomb, S. B., Bennett, M. J., Eds.; The Institute of
204. Kofstad, P.; Lillerud, K. P. Oxid. Met. 1982, 17, 177. Metals, 1993; Vol. 2, pp 445–454.
205. Ben Abderrazik, G.; Moulin, G.; Huntz, A. M. Oxid. Met. 245. Quaddakkers, W. J.; Elschner, A.; Speier, W.; Nickel, H.
1990, 33, 191–235. Appl. Surf. Sci. 1991, 52, 271–287.
246. Smialek, J. L.; Gibala, R. Metall. Trans. A 1983, 14A, Bennett, M. J., Eds.; The Institute of Materials, 1993;
2143–2161. Vol. II, pp 463–475.
247. Pint, B. A. In Fundamental Aspects of High Temperature 280. Pint, B. A.; Hobbs, L. W. Proc. Electrochem. Soc. 1992,
Corrosion; Shores, D. A., Rapp, R. A., Hou, P. Y., Eds.; 92–100.
Electrochemical Society, 1996; pp 74–85. 281. Pint, B. A.; Hobbs, L. W. Oxid. Met. 1994, 41, 203–233.
248. Brumm, M. W.; Grabke, H. J. Corros. Sci. 1992, 33, 282. Pint, B. A.; Garratt-Reed, A. J.; Hobbs, L. W. J. Phys IV.
1677–1690. 1993, C9-3, 247–256.
249. Yang, J. C.; Schumann, E.; Levin, I.; Ruhle, M. Acta 283. Pint, B. A.; Alexander, K. B. J. Electrochem. Soc. 1998,
Mater. 1998, 46, 2195–2201. 145, 1819–1829.
250. Andoh, A.; Taniguchi, S.; Shibata, T. Mater. Sci. Forum 284. Sheasby, J. S.; Jory, D. B. Oxid. Met. 1978, 12, 527–539.
2001, 369–372, 303–310. 285. Felten, E. J.; Pettit, F. S. Oxid. Met. 1976, 10, 189–223.
251. Yang, J. C.; Nadarzinski, K.; Schumann, E.; Rühle, M. 286. Giggins, C. S.; Pettit, F. S. J. Electrochem. Soc. 1971,
Scripta Metall. 1995, 33, 1043–1048. 118, 1782–1790.
252. Sigler, D. R. Oxid. Met. 1991, 36, 57–80. 287. Tien, J. K.; Pettit, F. S. Metall. Trans. 1972, 3, 1587–1599.
253. Hagel, W. C. Corrosion 1965, 21, 316. 288. Pettit, F. S. AIME Met. Soc. Trans. 1967, 239, 1296–1305.
254. Jedlinski, J. Oxid. Met. 1993, 39, 55–60. 289. Stasik, M. S.; Pettit, F. S.; Meier, G. H.; Ashary, A.;
255. Frémy, N.; Maurice, V.; Marcus, P. Surf. Interf. Anal. Smialek, J. L. Scripta Metall. Mater. 1994, 31, 1645–1650.
2002, 34, 519–523. 290. Sarioglu, C.; Stiger, M. J.; Blachere, J. R.;
256. Doychak, J.; Ruhle, M. Oxid. Met. 1989, 31, 431–452. Janakiraman, R.; Schumann, E.; Ashary, A.; Pettit, F. S.;
257. Yang, J. C.; Schumann, E.; Müllejans, H.; Rühle, M. Meier, G. H. Mater. Corros. 2000, 51, 358–372.
J. Phys. D: Appl. Phys. 1996, 29, 1716–1724. 291. Hindam, H. M.; Smeltzer, W. W. J. Electrochem. Soc.
258. Doychak, J. K. Doctoral thesis, Case Western Reserve 1980, 127, 1622–1630.
University: Cleveland, OH, 1986. 292. Smialek, J. L. Metall. Trans. A 1978, 9A, 309–320.
259. Yang, J. C.; Schumann, E.; Levin, I.; Rühle, M. Acta 293. Smialek, J. L.; Gibala, R. In High Temperature Corrosion;
Metall. 1998, 46, 2195–2201. Rapp, R. A., Ed.; NACE: Houston, TX, 1981; pp 274–283.
260. Hou, P. Y.; Paulikas, A. P.; Veal, B. W. Mater. High Temp. 294. Nanko, M.; Ozawa, M.; Maruyama, T. J. Electrochem.
2006, 22, 373–381. Soc. 2000, 147, 283–288.
261. El Kadiri, H.; Molins, R.; Bienvenu, Y.; Horstemeyer, M. F. 295. Xu, C. H.; Gao, W.; Gong, H. Intermetallics 2000, 8,
Oxid. Met. 2005, 64, 63–97. 769–779.
262. Lipkin, D. M.; Schaffer, H.; Adar, F.; Clarke, D. R. Appl. 296. Kuenzly, J. D.; Douglass, D. L. Oxid. Met. 1974, 6,
Phys. Lett. 1997, 70, 2550–2552. l39–l78.
263. Tolpygo, V. K.; Clarke, D. R. Mater. High Temp. 2000, 17, 297. Kumar, A.; Nasrallah, M.; Douglass, D. L. Oxid. Met.
59–70. 1974, 8, 227–263.
264. Steiner, C. J. -P.; Hasselman, D. P. H.; Spriggs, R. M. 298. Golightly, F. A.; Stott, F. H.; Wood, G. C. J. Electrochem.
J. Am. Ceram. Soc. 1971, 54, 412–413. Soc. 1979, 126, 1035–1042.
265. Schutze, M. In Proceedings of the Workshop on High 299. Kahn, A. S.; Lowell, C. E.; Barrett, C. A. J. Electrochem
Temperature Corrosion of Advanced Materials and Soc. 1980, 127, 670–679.
Protective Coatings; Saito, Y., Onay, B., Maruyama, T., 300. Nicolas-Chaubet, D.; Haut, C.; Picard, C.; Millot, F.;
Eds.; Elsevier Science, 1992; pp 39–49. Huntz, A. M. Mater. Sci. Eng. A. 1989, 120–121, 83–89.
266. Lessing, P. A.; Gordon, R. S. J. Mater. Sci. 1977, 12, 301. Amano, T.; Yajima, S.; Saito, Y. Trans. Jpn. Inst. Met.
2291–2302. 1979, 20, 431–441.
267. Ikuma, Y.; Gordon, R. S. J. Am. Ceram. Soc. 1983, 66, 302. Amano, T.; Yajima, S.; Saito, Y. Trans. Jpn. Inst. Met.
139–147. 1985, 26, 433–443.
268. Hollenberg, G. W.; Gordon, R. S. J. Am. Ceram. Soc. 303. Hou, P. Y.; Cannon, R. M.; Zhang, H.; Williamson, R. L.
1973, 56, 140–147. In Fundamental Aspects of High Temperature Corrosion;
269. Sugita, T.; Pask, J. A. J. Am. Ceram. Soc. 1970, 53, ECS Fall Meeting, San Antonio, TX, 6–11 October 1996;
609–613. Shores, D. A., Rapp, R. A., Hou, P. Y., Eds.;
270. Cho, J.; Harmer, M. P.; Chan, H. M.; Richman, J. M.; Electrochemical Society: Pennington, PA, 1997; ECS
Thompson, A. M. J. Am. Ceram. Soc. 1997, 80, 1013–1017. Proceedings, Vol. 96-26, pp 28–40.
271. Wang, C. M.; Cho, J.; Chan, H. M.; Harmer, P.; 304. Tolpygo, V. K.; Clarke, D. R. Acta Metall. 1998, 46,
Richman, J. M. J. Am. Ceram. Soc. 2001, 84, 1010–1016. 5135–5166.
272. Hou, P. Y.; Zhang, X. F.; Cannon, R. M. Scripta Metall. 305. Kroger, F. A. Ann. Rev. Mater. Sci. 1977, 7, 449–475.
2004, 51, 45–49. 306. Tolpygo, V. K.; Clarke, D. R.; Murphy, K. S. Metall. Trans.
273. Levin, E. M. Robbins, C. R.; McMurdie, H. F. Eds. Phase A 2001, 32, 1467–1478.
Diagram for Ceramists; The American Ceramic Society, 307. Jedlinski, J.; Borchardt, G.; Mrowec, S. Solid State Ionics
1964; p 309. 1992, 50, 67–74.
274. Drahus, M. D.; Chan, H. M.; Rickman, J. M.; 308. Hindam, H. M.; Smeltzer, W. W. Oxid. Met. 1980, 14,
Harmer, M. P. J. Am. Ceram. Soc. 2005, 88, 337–349.
3369–3373. 309. Young, E. W. A.; De Wit, J. H. W. Solid State Ionics 1985,
275. Muan, A. Am. J. Sci. 1958, 256, 413–422. 16, 39–46.
276. Novokhatskii, I. A.; Velov, B. F.; Gorokh, A. V.; 310. Reddy, K. P. R.; Smialek, J. L.; Cooper, A. R. Oxid. Met.
Savinskaya, A. A. Russ. J. Phys. Chem. 1965, 39, 1982, 17, 429–449.
1498–1499. 311. Jedlinski, J.; Borchardt, G. Oxid. Met. 1991, 36, 317–337.
277. Wright, I. G.; Pint, B. A.; Tortorelli, P. F. Oxid. Met. 2001, 312. Quadakkers, W. J.; Elschner, A.; Speier, W.; Nickel, H.
55, 333–337. Appl. Surf. Sci. 1991, 52, 271–287.
278. Pint, B. A.; More, K. L.; Tortorelli, P. F.; Porter, W. D.; 313. Prescott, R.; Mitchell, D. F.; Sproule, G. I.; Graham, M. J.
Wright, I. G. Mat. Sci. Forum 2001, 369–372, 411–418. Solid State Ionics 1992, 53–56, 229–237.
279. Pint, B. A.; Garratt-Reed, A. J.; Hobbs, L. W. 314. Pint, B. A.; Martin, J. R.; Hobbs, L. W. Oxid. Met. 1993,
In Microscopy of Oxidation; Newcomb, S. B., 39, 167–195.
315. Mennicke, C.; Schumann, E.; Ruhle, M.; Hussey, R. J.; 353. Wright, J. K.; Williamson, R. L.; Renusch, D.; Veal, B.;
Sproule, G. I.; Graham, M. J. Oxid. Met. 1998, 49, Grimsditch, M.; Hou, P. Y.; Cannon, R. M. Mater. Sci.
455–466. Eng. A 1999, 262, 246.
316. Basu, S. N.; Halloran, J. W. Oxid. Met. 1987, 27, 354. Balint, D. S.; Xu, T.; Hutchinson, J. W.; Evans, A. G. Acta
143–155. Metall. 2006, 54, 1815–1820.
317. Oishi, Y.; Kingery, W. D. J. Chem. Phys. 1960, 33, 355. Davis, A. W.; Evans, A. G. Metall. Trans. 2006, A37A,
480–486. 2085–2095.
318. Reddy, K. P. R.; Cooper, A. R. J. Am. Ceram. Soc. 1982, 356. Schutze, M.; Saito, Y.; Onay, B.; Maruyama, T.
65, 634–638. Proceedings of the Workshop on High Temperature
319. Oishi, Y.; Ando, V.; Suga, N.; Kingery, W. D. J. Am. Corrosion of Advanced Materials and Protective
Ceram. Soc. 1983, 66, C130–C131. Coatings; Saito, Y., Onay, B., Maruyama, T., Eds.;
320. Le Gall, M.; Huntz, A. M.; Lesage, B.; Monty, C.; Elsevier Science, 1992; pp 39–49.
Bernardini, J. J. Mater. Sci. 1995, 30, 201. 357. Whittle, D. P.; Stringer, J. Philos. Trans. R. Soc. London A
321. Reed, D. J.; Wuensch, B. J. J. Am. Ceram. Soc. 1980, 63, 1980, 295, 309.
88–92. 358. Przybylski, K.; Garrett-Reed, A. J.; Pint, B. A.; Katz, E. P.;
322. Lagerlof, K. P. D.; Peltka, B. J.; Mitchell, T. E.; Yurek, G. J. J. Electrochem. Soc. 1987, 134, 3207.
Heuer, A. H. Rad. Effects 1983, 74, 87–107. 359. Cotell, C. M.; Yurek, G. J.; Hussey, R. J.; Mitchell, D. F.;
323. Prot, D.; Monty, C. Philos. Mag. 1996, 73, 899–917. Graham, M. J. Oxid. Met. 1990, 34, 173–200.
324. Paladino, A. E.; Kingery, W. D. J. Chem. Phys. 1962, 37, 360. Tolpygo, V. K.; Grabke, H. J. Scripta Metall. 1998, 38,
957–962. 123–129.
325. Le Gall, M.; Lesage, B.; Bernardini, J. Philos. Mag. 1994, 361. Delaunay, D.; Huntz, A. M.; Lacombe, P. Corros. Sci.
70, 761–773. 1984, 24, 13–25.
326. Cannon, R. M.; Rhodes, W. H.; Heuer, A. H. J. Am. 362. Funkenbusch, A. W.; Smeggil, J. G.; Bornstein, N. S.
Ceram. Soc. 1980, 63, 46–53. Metall. Trans. 1985, 16A, 1164–1166.
327. Glaeser, A. M.; Evans, J. W. Acta Metall. 1986, 34, 1545. 363. Smialek, J. L. Metall. Trans. 1991, 22A, 739–752.
328. Cannon, R. M.; Hou, P. Y. In High Temperature Corrosion 364. Pfeil, L. B. U.K. Patent No. 459848, 1937.
and Materials Chemistry; Hou, P. Y., McNallan, M. J., 365. Pfeil, L. B. U. K. Patent No. 574088, 1945.
Oltra, R., Opila, E. J., Shores, D. A., Eds.; The 366. Pint, B. A. In Proceedings of John Stringer Symposium
Electrochemical Society, 1998; pp 594–607. on High Temperature Corrosion; ASM Materials Solution
329. Lipkin, D. M.; Clarke, D. R. Oxid. Met. 1996, 45, Conference, Indianapolis, IN, 5–8 November 2001;
267–280. Tortorelli, P., Wright, I. W., Hou, P. Y., Eds.; ASM
330. Hutchings, R.; Loretto, M. H.; Smallman, R. E. Met. Sci. International: OH, 2003; pp 52–62.
1981, 15, 7. 367. Pint, B. A. In Fundamental Aspects of High Temperature
331. Brumm, M. W.; Grabke, H. J. Corros. Sci. 1993, 34, 547. Corrosion; Shores, D. A., Rapp, R. A., Hou, P. Y., Eds.;
332. Pint, B. Oxid. Met. 1997, 48, 303. Electrochemical Society, 1997; Vol. 96-26, pp 74–85.
333. Rommerskirchen, I.; Eltester, B.; Grabke, H. J. Mater. 368. Stringer, J.; Hou, P. Y. In Proceedings of the Symposium
Corros. 1996, 47, 646. on Corrosion and Particle Erosions at High
334. Xu, C. H.; Gao, W.; Gong, H. High Temp. Mater. Proc. Temperatures; TMS Annual Meeting, Las Vegas, NV, 27
2000, 19, 371. February–3 March; Srinivasan, V., Vedula, K., Eds.; 1989;
335. Hou, P. Y.; Niu, Y.; Van Lienden, C. Oxid. Met. 2003, 59 pp 383–401.
(1/2), 41–61. 369. Jedlinski, J.; Godlewski, K.; Mrowec, S. Mater. Sci. Eng.
336. Dravnieks, A.; McDonald, H. J. J. Electrochem. Soc. A 1989, 120–121, 539–543.
1948, 94, 139. 370. Hou, P. Y.; Brown, I.; Stringer, J. Nucl. Inst. Meth. B 1991,
337. Dobson, P. S.; Smallman, R. E. Proc. R. Soc. Lond. A 59/60, 1345–1349.
1966, 293, 423. 371. Pint, B. A. J. Am. Ceram. Soc. 2003, 86, 686–695.
338. Kumar, A.; Nasrallah, M.; Douglass, D. L. Oxid. Met. 372. Pint, B. A. Oxid. Met. 1998, 49, 531–559.
1974, 8, 227. 373. Pint, B. A.; More, K. L.; Wright, I. G. Mater. High Temp.
339. Hales, R.; Smallman, R. E.; Dobson, P. S. Proc. R. Soc. 2003, 20, 375–386.
Lond. A 1968, 307, 71. 374. Hou, P. Y.; Stringer, J. Oxid. Met. 1988, 29, 45–73.
340. Douglass, D. L. Mater. Sci. Eng. 1969, 3, 255. 375. Davis, H. H.; Graham, H. C.; Kvernes, I. A. Oxid. Met.
341. Evans, H. E. Mater. Sci. Technol. 1988, 4, 1089. 1971, 3, 431.
342. Pieraggi, B.; Rapp, R. A.; Hirth, J. P. Oxid. Met. 1995, 376. Stringer, J.; Wilcox, B. A.; Jaffee, R. L. Oxid. Met. 1972,
44, 63. 5, 11.
343. Harris, J. Acta Metall. 1978, 26, 1033. 377. Seltzer, M. S.; Wilcox, B. A.; Stringer, J. Met. Trans. 1972,
344. Zimmermann, D.; Bobeth, M.; Ruhle, M.; Clarke, D. R. 3, 2391; Seltzer, V. J. Metal Trans. 1972, 3, 2357.
Zeit. Metall. 2004, 95, 84–90. 378. Nanni, P.; Stoddart, C. T. H.; Hondros, E. D. Mater.
345. Hou, P. Y.; Van Lienden, C. Mater. High Temp. 2003, Chem. 1976, 1, 297.
20, 357. 379. Pint, B. A.; Leibowitz, J.; DeVan, J. H. Oxid. Met. 1999,
346. Grabke, H. J.; Wiemer, D.; Viefhaus, H. Appl. Surf. Sci. 51, 183–199.
1991, 47, 243. 380. Pint, B. A.; Garratt-Reed, A. J.; Hobbs, L. W. Mater. High
347. Hou, P. Y.; Stringer, J. J. Phys 1993, 4(C9), 231–240. Temp. 1995, 13, 3–16.
348. Golightly, F. A.; Stott, F. H.; Wood, G. C. Oxid. Met. 1976, 381. Giggins, C. S.; Pettit, F. S. Trans. TMS–TIME 1969, 245,
10, 163. 2509.
349. Tolpygo, V. K.; Clarke, D. R. Acta Mater. 1998, 46, 5153. 382. Li, M. S.; Qian, Y. H.; Zhou, Y. C. Oxid. Met. 2004, 61, 529.
350. Wood, G. C.; Stott, F. H. Mater. Sci. Technol. 1987, 383. Huang, K.; Hou, P. Y.; Goodenough, J. B. Mater. Res.
3(7), 519. Bull. 2001, 36, 81–95.
351. Suo, Z. J. Mech. Phys. Solids 1995, 43, 829. 384. Giggins, C. S.; Pettit, F. S. Met. Trans. 1971, 2, 1071.
352. Yang, Z. G.; Hou, P. Y. Mater. Sci. Eng. A 2005, 391(1/2), 385. Tolpygo, V. K.; Clarke, D. R. Mater. High Temp. 2003, 20,
1–9. 261–271.
386. Nychka, J. A.; Clarke, D. R. Oxid. Met. 2005, 63, Hou, P. Y., Eds.; Electrochemical Society: Pennington,
325–352. NJ, 1997; pp 344–351.
387. Dickey, E. C.; Pint, B. A.; Alexander, K. B.; Wright, I. G. 418. Molins, R.; Rouzou, I.; Hou, P. Y. Mater. Sci. Eng. A 2007,
J. Mater. Res. 1999, 14, 4531–4540. 454–455, 80–88.
388. Ishii, K.; Kohno, M.; Ishikawa, S.; Satoh, S. Mater. Trans., 419. Prüßner, K.; Schumann, E.; Rühle, M. In Fundamental
JIM (Japan) 1997, 38, 787–792. Aspects of High Temperature Corrosion; Shores, D. A.,
389. Mennicke, C.; Schumann, E.; Le Coze, J.; Smialek, J. L.; Rapp, R. A., Hou, P. Y., Eds.; The Electrochemical
Meier, G. H.; Ruhle, M. In Microscopy of Oxidation; Society, 1996; p 344.
Newcomb, S. B., Little, J. A., Eds.; Institute of Materials: 420. Hou, P. Y.; Priimak, K. Oxid. Met. 2005, 63, 113–130.
UK, 1997; Vol. 3, pp 95–104. 421. Kiely, J. D.; Yeh, T.; Bonnell, D. A. Surf. Sci. 1997, 393,
390. Ramanarayanan, T. A.; Raghavan, M.; Petkovic-Luton, R. L126.
Oxid. Met. 1984, 22, 83–100. 422. Hou, P. Y.; Smialek, J. L. Scripta Metall. 1995, 33,
391. Versaci, R. A.; Clemens, D.; Quadakkers, W. J.; 1409–1416.
Hussey, R. Solid State Ionics 1993, 59, 235–242. 423. Hou, P. Y. Oxid. Met. 1999, 52, 337–351.
392. Schumann, E.; Yang, J. C.; Ruhle, M.; Graham, M. J. 424. Pieraggi, B.; Rapp, R. A. Mater. High Temp. 1994, 12,
Oxid. Met. 1996, 46, 37–49. 229–235.
393. Pint, B. A. Oxid. Met. 1996, 45, 1–37. 425. Pieraggi, B.; Rapp, R. A.; Hirth, J. P. Oxid. Met. 1995,
394. Duffy, D. M.; Tasker, P. W. Philos. Mag. A 1984, 50, 44(1–2), 63–79.
155–169. 426. Hong, Y.; Anderson, A. B.; Smialek, J. L. Surf. Sci. 1990,
395. Lartigue-Korinek, S.; Dupau, F. Acta Metall. Mater. 1994, 230, 175–183.
42, 293–302. 427. Yong, J.; Smith, J. R.; Evans, A. G. g-Ni/Al2O3 Adhesion:
396. Clarke, D. R. Acta Metall. 2003, 51, 1393–1407. Aluminum Activity and Impurity Effects, Presented at
397. Hou, P. Y.; Stringer, J. J. Electrochem. Soc. 1987, 134, the Workshop on Science and Technology of Thermal
1836–1849. Barrier Systems, University of California, Santa Barbara,
398. Hou, P. Y.; Shui, Z. R.; Chuang, G. Y.; Stringer, J. CA, 9–11 January 2007.
J. Electrochem. Soc. 1992, 139, 1119–1126. 428. Strawbridge, A.; Rapp, R. A. J. Electrochem. Soc. 1994,
399. Wood, G. C.; Stott, F. H. In High Temperature Corrosion; 141, 1905–1915.
Rapp, R. A., Ed.; NACE, 1981; p 227. 429. Czerwinski, F.; Szpunar, J. A.; Smeltzer, W. W.
400. Prescott, R.; Graham, M. J. Oxid. Met. 1992, 38(3/4), J. Electrochem. Soc. 1996, 143, 3000–3007.
233–254. 430. George, P. J.; Bennett, M. J.; Bishop, H. E.; Cotell, C. M.;
401. Strawbridge, A.; Hou, P. Y. Mater. High Temp. 1994, 12, Garratt-Reed, A. J. Surf. Coat. Technol. 1992, 51(1–3),
177–182. 45–51.
402. Yang, S. L.; Wang, F. H.; Niu, Y.; Wu, W. T. Mater. Sci. 431. Ikeda, Y.; Nii, K.; Yoshihara, K. Trans. Jpn. Inst. Met.
Forum 2001, 369–372, 361–368. Suppl. 1983, 24, 207.
403. Wu, Y. N.; Qin, M.; Feng, Z. C.; Liang, Y.; Sun, C.; 432. Smeggil, J. G. Mater. Sci. Eng. 1987, 87, 261–265.
Wang, F. H. Mater. Lett. 2003, 57(16–17), 2404–2408. 433. Smialek, J. L.; Jayne, D. T.; Schaeffer, J. C.;
404. Hou, P. Y.; Paulikas, A. P.; Veal, B. W. Mater. Sci. Forum Murphy, W. H. Thin Solid Films 1994, 235, 285–292.
2006, 522–523, 433–440. 434. Seah, P.; Hondros, E. D. Proc. R. Soc. Lond. A 1973, 335,
405. Hindam, H.; Whittle, D. P. Trans. Jpn. Inst. Met. Suppl. 191–212.
1983, 261. 435. Lipkin, D. M.; Clarke, D. R.; Evans, A. G. Acta Met. 1998,
406. Allam, I. A.; Whittle, D. P.; Stringer, J. Oxid. Met. 1978, 46, 4835–4850.
12, 35; 1979, 13, 381. 436. Bonnell, D. A.; Kiely, J. Phys. Stat. Sol. 1998, 166, 7–17.
407. Ideda, Y.; Nii, K.; Yoshihara, K. Trans. Jpn. Inst. Met. 437. Grabke, H. J.; Wagemann, B. Oxid. Met. 1995, 43,
Suppl. 1983, 24, 207. 97–114.
408. Lees, D. G. Oxid. Met. 1987, 27, 75. 438. Hou, P. Y. Mater. Corros. 2000, 51, 329.
409. Hou, P. Y.; Bennett, M. J.; Ruhle, M. In International 439. Hou, P. Y.; Moskito, J. Oxid. Met. 2003, 59(5/6), 559–574.
Workshop on High Temperature Oxidation; Gorisch, 440. Hou, P. Y. J. Mater. Sci. Lett. 2000, 19, 577–578.
Germany, 15–19 August 1994, Douglass, D. L., 441. Hou, P. Y. Mater. Sci. Forum 2001, 369–372, 23.
Kofstad, P., Rahmel, A., Wood, G. C., Eds.; Oxid. Met. 442. Rivoaland, L.; Maurice, V.; Josso, P.; Bacos, M. -P.;
1996; 45, pp 529–620. Marcus, P. Oxid. Met. 2003, 60, 137–157.
410. Balluffi, W. In Interfacial Segregation; Johnson, W. C., 443. Hou, P. Y.; McCarty, K. F. Scripta Metall. 2006, 54,
Blakely, J. M., Eds.; ASM: Metals Park, OH, 1997; 937–941.
pp 193–237. 444. Cadoret, Y.; Bacos, M. -P.; Josso, P.; Maurice, V.;
411. Seah, P.; Hondros, E. D. Proc. R. Soc. Lond. A. 1973, Marcus, P.; Zanna, S. Mater. Sci. Forum 2004,
335, 191–212. 461–464, 274.
412. Zhang, W.; Smith, J. R.; Wang, X. G.; Evans, A. G. Phys. 445. Hou, P. Y.; Stringer, J. Oxid. Met. 1992, 38, 323–345.
Rev. B. 2003, 67, 245414. 446. Hou, P. Y. In High Temperature Corrosion and Materials
413. Hou, P. Y. J. Corr. Sci. Eng. 2003, 6, Paper 75. Chemistry; Hou, P. Y., McNallan, M. J., Oltra, R.,
414. Al-Badary, H.; Tatlock, G.; Le Coze, J. In Microscopy of Opila, E. J., Shores, D. A., Eds.; The Electrochemical
Oxidation; Newcomb, J. B., Little, J. A., Eds.; The Institute Society, 1998; pp 198–210.
of Metals, 1997; Vol. 3, pp 105–114. 447. Hou, P. Y. In Proceedings of the John Stringer
415. Fox, P.; Lees, D. G.; Lorimer, G. W. Oxid. Met. 1991, 36, Symposium on High Temperature Corrosion; ASM
491–503. Materials Solution Conference, Indianapolis, IN, Nov.
416. Hou, P. Y.; Prüßner, K.; Fairbrother, D. H.; Roberts, J. G.; 5–8, 2001; Tortorelli, P., Wright, I. W., Hou, P. Y., Eds.;
Alexander, V. Scripta Metall. 1999, 40, 241–247. ASM International: OH, 2003; pp 164–173.
417. Prussner, K.; Schumann, E.; Ruhle, M. In Proceedings of 448. Murata, Y.; Nakai, M.; Nagai, K.; Morinaga, M.;
the Symposium Fundamental Aspects of High Sasaki, Y.; Hashizume, R. Mater. Sci. Forum 2006,
Temperature Corrosion; Shores, D. A., Rapp, R. A., 522–523, 147–154.
449. Amano, T.; Watanabe, T.; Michiyama, K. J. Jpn. Inst. 452. Molins, R.; Andrieux, M.; Huntz, A. M. Mater. High Temp.
Met. 1997, 61, 1077–1085. 2005, 22, 335–342.
450. Quadakkers, W. J.; Wasserfuhr, C.; Khanna, A. S.; 453. Ben Abderrazik, G.; Moulin, G.; Huntz, A. M. Oxid. Met.
Nickel, H. Mater. Sci. Technol. 1988, 4(12), 1990, 33, 191–235.
1119–1125. 454. Mayer, J.; Penkalla, H. J.; Dimyati, A.; Dani, M.;
451. Huang, T. T.; Richter, R.; Chang, Y. L.; Pfender, E. Metall. Untoro, P.; Naumenko, D.; Quadakkers, W. J. Mater.
Trans. A 1985, 16A, 2051–2059. High Temp. 2003, 20, 413–419.
1.11 Sulfidation and Mixed Gas Corrosion of Alloys
R. John
Shell Global Solutions (US) Inc., Westhollow Technology Center, P.O. Box 4327, Houston, TX 77210, USA
1.11.1 Sulfidation 240

1.11.1.1 Types of Equipment and Processes 240
1.11.1.2 Corrosion Mechanism 240
1.11.1.3 Compilations of Sulfidation Corrosion Rate Predictions 243
1.11.1.4 Time Dependence to Predict Sulfidation 243
1.11.1.5 Laboratory Simulation 245
1.11.2 Corrosion by Mixed Gases 245
1.11.2.1 Thermochemical Calculations to Predict Oxide–Sulfide–Carbide–Nitrides on Alloys 247
1.11.2.2 Sulfidation/Oxidation by CO–CO2–COS and H2–H2O–H2S Gases 249
1.11.2.3 Thermochemistry and Corrosion Mechanism 250
1.11.2.4 Laboratory Simulation 250
1.11.2.5 Kinetics 255
1.11.2.6 Corrosion Influenced by Gas History 259
1.11.3 Oxidation/Nitridation by N2–O2 Gases 260
1.11.3.1 Types of Equipment and Processes Where Nitridation Occurs 260
1.11.3.3 Predicting Corrosion Product and Alloy Phase Formation in Nitriding/Oxidizing Conditions 262
1.11.4 Oxidation/Carburization by CH4–H2O Gases 265
1.11.4.1 Types of Equipment and Processes Where Carburization Occurs 265
1.11.4.3 Corrosion Products Prediction 266
1.11.5 Summary 267
References 270
1.11.1 Sulfidation and H2S gases) or PCO and PCOS (for CO and COS
gases), alloy composition and temperature. Sulfidation
1.11.1.1 Types of Equipment and Processes
corrosion forms sulfide corrosion products, damages
Corrosion of metals and alloys used in equipment the metal and alloy by losses in wall thickness (penetra-
processing high temperature, corrosive, gases contain- tion), and occurs upon exposure of metals to gases
ing sulfur, H2S, and COS is a concern in processes containing sulfur, COS, or H2S. Sulfidation by H2S is
used in gas processing, combustion gas process heaters, predominately discussed in this chapter. Sulfidation
petroleum refineries (hydrocracking, coking, vacuum reduces the useful thickness of the metal by the forma-
flashing, hydrotreating, and catalytic reforming) coal/ tion of surface sulfide scale and internal sulfide phases.
coke/oil gasification, petrochemical production, gasi- The sum of these two thicknesses is defined as the total
fication of black liquor in pulp/paper production, and metal penetration. The first step in determining the
fossil fuel-fired power generation. Corrosion is often potential for equipment to sulfide is to assume that sul-
the phenomenon, which defines the maximum allow- fidation is the dominant corrosion mechanism. The key
able temperature or maximum allowable gas species indicator of sulfidation is that most of the corrosion
concentrations for metals and alloys in equipment. products are sulfides. Sulfides will be found on the
exposed surface and within the alloy microstructure
near to the surface, depending upon the alloy and exp-
1.11.1.2 Corrosion Mechanism
osure conditions. X-ray analyses by diffraction and fluo-
The key variables, which influence the kinetics are rescence of a surface scale sample are common methods
exposure time, partial pressures PH2 and PH2 S (for H2 to determine the presences of sulfides. An alternate
240
Sulfidation and Mixed Gas Corrosion of Alloys 241
method is to evaluate the thermochemical interactions the effect of several alloys widely used in process
between the alloy and gas, and the compounds formed. equipment. Dashed lines represent extrapolations of
Surface metal loss (scaling) by sulfidation can be the available data, while solid lines represent interpo-
detected by methods, which measure the metal thick- lations of the available data. The line for carbon steel
ness, such as ultrasonic and mechanical methods. Sound stops for low concentrations of H2S because FeS is
metal loss by internal sulfidation can only be detected by not stable at low concentrations of H2S (as shown in
metallography, as shown schematically in Figure 1. Figure 4), and the carbon steel cannot corrode at
The time dependence of sulfidation has been con- combinations of sufficiently high temperature and suf-
troversial1–5 with reports of a parabolic time depen- ficiently low partial pressure of H2S. It is incorrect to
dence (metal loss proportional to time0.5), linear time indicate that steels with significant concentrations of
dependence (metal loss proportional to time), power Cr cannot corrode at conditions representing low
law dependence (metal loss proportional to timex), H2S concentrations and high temperatures indicated
and various combinations of these dependencies. If by this line. This is because alloys with Cr form CrS at
the corrosion product sulfide scale remains undis- lower H2S concentrations and higher temperatures
turbed on the alloy surface and sufficient time has
passed (in excess of 1000 h), currently available infor- 0.6
mation suggests that the time dependence is para-
bolic. Removal or cracking of the surface sulfide scale
tends to increase the rate of sulfidation, because the
presence of the scale is a partial barrier, which tends Penetration (mm)
0.4
to slow the sulfidation rate as time passes. The idea-
lized time parabolic dependence of sulfidation of
G10200 carbon steel at 350 C in 0.05 atm H2S in H2
after several thousand hours is shown in Figure 2.
0.2
Some studies suggest an initial linear time depen-
dence for several thousand hours. Sulfidation data
measured after several hundred hours (as is often
the case for available data) are unlikely to be useful
0.0
in estimating sulfidation corrosion rates for long-term 0 25 000 50 000 75 000 100 000
service. Sulfidation data are properly used when the Time (h)
time dependence is considered. Figure 2 Effect of time upon the sulfidation corrosion of
Increasing the concentration of H2S tends to carbon steel (G10200) after 1 year in 0.05 atm H2S and H2 at
increase the sulfidation rate of alloys. Figure 3 shows 623 K.
10.00
External scale
S50400
G10200
Penetration (mm)
1.00
Total
penetration S41000
Internal
penetration 0.10
N08810
S30400 S31000
Internal corrosion 0.01
products 1.E − 04 1.E − 03 1.E − 02 1.E − 01 1.E + 00
Corroded grain Uncorroded alloy P H2S (atm)
boundaries
Figure 3 Effect of H2S partial pressure upon the
Figure 1 Schematic view of total penetration sulfidation corrosion after 1 year in H2–H2S gases at a total
measurement for a typical corrosion product morphology. pressure of 1 atm at 723 K.
10 effective gas concentration of H2S could be estimated,

the approach discussed should be suitable for estimat-
ing the metal losses, as related to the maximum allow-
able temperatures/H2S concentrations and appropriate
1 materials of construction.
High-Ni alloys used either as base metals or as
welding filler metals are a special concern in sulfida-
tion conditions. Sulfidation of high-Ni alloys can be
H2S concentration (%)
0.1 especially rapid and yield corrosion rates greater than

Sulfidation of carbon/low 2.5 mm year1 if the temperature exceeds 530 C,
alloy (<5% Cr) steels which is the melting point of a potential corrosion
product which forms as a mixture of Ni and nickel
0.01 FeS sulfide. A reasonable approach for high-Ni alloys is
Fe that they should not be used in sulfidation conditions
when metal temperatures will exceed 530 C. High-Ni
No sulfidation of
carbon/low alloy alloys with high-Cr levels (such as N06625 or N08825)
0.001 (<5% Cr) steels can be very suitable, with low corrosion rates at lower
Sulfidation of temperatures.6–8 The phase behaviors of FeS and NiS
steels are described in Figures 5–8.
with >5% Cr Cr6S7 The first step in assessing the rate of sulfidation is
0.0001 Cr to evaluate the potential for the formation of sulfide
400 600 800 1000 1200 corrosion products. Confirmation of formation of
Temperature (⬚F) sulfides on existing equipment or a thermochemical
Figure 4 Conditions for possible sulfidation, based upon evaluation of the corrosive gas to produce sulfides
H2S concentrations in H2–H2S gases and temperatures may be done by the analyses of corrosion products,
above the FeS/Fe line for carbon/low alloy (<5% Cr) steels
use of well-known compilations,7–13 or ASSET pro-
and above the Cr6S7/Cr line for alloys with >5% Cr,
400–1200 F (204–648 C, 473–921 K). gram.10 Once sulfidation is expected, one can predict
sulfidation rates by using either the traditional sulfi-
dation curves for corrosion in H2S–H2 gases or the
than that needed to form FeS on steels. Figure 4 shows curves of the type shown in previous work,7–13
the limits of H2S concentration and temperature including the effects of temperature, gas composition,
corresponding to sulfide corrosion products of and alloy composition.
Fe and Cr, as shown beneath the lines. Alloys with Most corrosion data for alloys exposed to high
greater than 5% Cr can corrode in conditions where temperature gases have been reported in terms of
low alloy steels cannot corrode. The corrosion rates weight change/area for relatively short exposures
may be low (such as 0.025–0.25 mm year1) but still and inadequately defined exposure conditions. The
significant for conditions where steels are traditionally weight change/area information is not directly
thought to be immune to sulfidation. The ranges of relatable to the thickness of corroded metal (pen-
H2S concentration represented in Figures 3 and 4 etration), which is often needed to assess the
span the low H2S range of catalytic reformers to the strength of the equipment components. Corrosion
high H2S concentrations expected in modern hydro- is best reported in penetration units, which indicate
treaters in crude oil distillation equipment in petro- the sound metal loss, as discussed earlier.12,13 Cor-
leum refineries. These curves are in good agreement rosion in high temperature gases is affected by key
with the traditional data.1–5 parameters of the corrosive environments such as
Sulfidation caused by sulfur-containing compounds temperature, alloy composition, time, and gas com-
in liquid hydrocarbons in the absence of a gas phase position. Summaries of metal penetrations are used
(such as that found in petroleum distilling units) is not in this chapter, which goes beyond the traditional
discussed in this chapter. The sulfur concentrations in corrosion weight change data by reporting total
liquid hydrocarbons and the sulfur evolvable as gaseous metal penetration for an extensive number of alloys
H2S have not yet been clearly related. However, the over a wide range of conditions. Compositions of
effective H2 and H2S concentrations are likely domi- some alloys discussed in this chapter are shown
nant variables in the liquid hydrocarbons. If the in Table 1.
1800
1596
1600
1455
1400
Liquid(1) + Liquid(2)
1200
Liquid 1001 1023

1000
FCC + liquid 993
t/C
801
800
637 β Ni1-xS
NiS2
560 600
600 536
437
400
400 353
Ni7S6
302
Ni2S3
200 Ni3S4
Ni9S8
115
NiS
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
xs
Figure 5 Calculated condensed phase diagram with experimental data for Ni–S system.
1.11.1.3 Compilations of Sulfidation increase the sulfidation rate. Temperature increases of

Corrosion Rate Predictions 50 C will typically double the sulfidation rate, while
increasing the H2S concentration by a factor of 10 may
Sulfidation rate data for some commercial alloys are
be needed to double the sulfidation rate. Therefore,
summarized in Figures 9–12. Figure 9 shows sulfi-
likely changes of H2S concentrations in processes are
dation predictions of several simple metals (copper,
generally less significant, in terms of influencing corro-
carbon steel, and nickel). Figure 10 shows sulfidation
sion, than the likely temperature variations. Increasing
predications of a range of alloys often used in these
the Cr content of the alloy greatly slows the sulfidation,
types of conditions. Figure 11 shows sulfidation pre-
as seen in progression from 9% Cr (S50400), 12% Cr
dictions of some of the most resistant alloys available.
(S41000), 18% Cr (S30400), 20% Cr (N08810), and
Figure 12 shows sulfidation predictions of several
25% Cr (N08825 and N06625). The ranges of
alloys similar to the 18% Cr–8% Ni steels. These steels
H2S concentration represented in these figures span
experience different sulfidation rates, even though
the low H2S range of catalytic reformers to the high
the Cr and Ni concentrations are similar. The differ-
H2S concentrations expected in modern hydrotreaters.
ent concentrations of Mo, Ti, and Nb among these
These curves are in good agreement with the tradi-
alloys do significantly influence the sulfidation kinet-
tional data.8–13
ics which is not well known; it is widely assumed that
these steels corrode with kinetics similar to the com-
1.11.1.4 Time Dependence to Predict
mon 18% Cr–8% Ni steel UNS S30400.
Sulfidation
The effects of temperature and H2S concentration
upon sulfidation of steels often used in oil refining A large compilation of test data was evaluated with
services are shown in Figures 3 and 9–12, which respect to the time dependence of the corrosion. The
represent metal losses expected after 1 year of exposure. following rate equation can be used to calculate cor-
Increasing the temperature and H2S concentration rosion by sulfidation.8 The equation incorporates the
0.9
0.1
0.8
0.2
0.7
0.3
0.6
0.4
x Ni
xS
M1-xS
0.5
0.5
β
0.4
0.6
FCC+
M1-xS+
β
0.3
0.7
0.2
0.8
0.1
0.9
Fe 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Ni
xFe
Figure 6 Calculated isotherm of the Fe–Ni–S system at 998 K.
well-known time parabolic and linear behaviors, are calculated separately for each alloy. The form of
which have been observed. Parabolic behavior occurs the equation includes both linear and parabolic time
initially and then tends towards linear time depen- dependencies. The M factor empirically describes
dence as the surface scales crack or partially spall as how much time passes before the time dependence
they thicken and become less adherent. The transi- becomes linear and varies for each alloy.
tion towards linear dependence often results from Graphical illustrations of correlations for a couple
development of an effective steady-state thickness of example alloys (S30400 and N08810) are shown in
of surface corrosion product, as caused by cracking, Figure 13. These alloys are fairly typical and in that
spalling, or other failure mechanisms. Fe–Ni–Cr alloys are widely used in many different
types of process equipment, and they are also well
logððPenetration=Hours0:5 Þ þ ðM HoursÞÞ
correlated with the methodology described in this
¼ A þ B logðPH2 S Þ þ C logðPH2 Þ þ D=T ½1 article. In fact, the data are well correlated over
about three orders of magnitude of variation in the
In this equation, penetration represents sound metal
measured corrosion for wide ranges in the important
loss by corrosion, T is the absolute temperature, and
environmental variables of temperature, partial pres-
PH2 S and PH2 represent the partial pressures of the
sure of H2, and partial pressure of H2S.
respective species. A, B, C, D, and M are constants that
1400
1300
1200
1100
Liquid
1000
Liquid + M1-xS
900
t (⬚C)
800
M1-xS + β
700 β
600
500
M9S8 + β β + Ni3S2
400 M9S8 + β + Ni3S2

FCC + M9S8 + Ni3S2
300
0 0.2 0.4 0.6 0.8 1
XN3S2
Figure 7 Calculated section of the FeS–Ni3S2 with experimental data.
10
1.11.1.5 Laboratory Simulation
Laboratory simulation of sulfidation corrosion is
generally carried out with flowing H2–H2S gases or
1
CO–COS gases at constant temperature, starting
Sulfidation of nickel
H2S concentration (%)
with a polished metal surface. The combination of

gas flow rate and H2S (or COS) concentration must
0.1
Ni3S2 be high enough to prevent significant consumption
of H2S during the test exposures. Post-exposure
0.01 examination is carried out with weight change and
Ni microscopic methods to assess the amount of metal
consumed by scale formation and also the amount of
0.001 the metal containing internal corrosion products.
No sulfidation of nickel
0.0001 1.11.2 Corrosion by Mixed Gases

400 600 800 1 000 1 200
Temperature (⬚F)
Figure 8 Conditions for possible sulfidation of nickel
The corrosion mechanism names for mixed gas cor-
based upon combinations of H2S concentrations in rosion indicate the stable alloy corrosion products
H2–H2S gases and temperatures above the Ni/Ni3S2 line formed and result from a competition between the
for Ni-base alloys. 400–1200 F (204–628 C). different possibilities and are named after the corrosion
Table 1 Compositions of alloys discussed
UNS Fe Cr Ni Co Mo Al Si Ti W Mn Cu RE Nb
S30415 69.66 18.30 9.50 0.00 0.42 0.00 1.23 0.00 0.00 0.56 0.23 0.05 0.00
S30815 65.60 20.90 11.00 0.00 0.00 0.00 1.77 0.00 0.00 0.64 0.00 0.00 0.00
N06025 9.45 25.35 62.63 0.00 0.00 2.09 0.06 0.14 0.00 0.09 0.01 0.00 0.00
S41000 86.50 12.30 0.50 0.00 0.10 0.00 0.60 0.00 0.00 0.00 0.00 0.00 0.00
G10200 99.42 0.00 0.00 0.00 0.00 0.00 0.04 0.00 0.38 0.00 0.00 0.00
S30400 71.07 18.28 8.13 0.14 0.17 0.00 0.49 0.00 0.00 1.48 0.19 0.00 0.00
S31000 52.41 24.87 19.72 0.05 0.16 0.00 0.68 0.00 0.00 1.94 0.11 0.00 0.00
S31600 68.75 17.00 12.00 0.00 2.25 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
S34700 68.14 17.75 10.75 0.00 0.00 0.00 0.55 0.00 0.00 1.80 0.00 0.00 0.96
S44600 74.12 24.36 0.36 0.02 0.20 0.00 0.33 0.00 0.00 0.45 0.10 0.00 0.00
R30188 1.32 21.98 22.82 38.00 0.00 0.00 0.37 0.00 14.55 0.82 0.00 0.04 0.00
N07214 2.49 16.04 76.09 0.14 0.10 4.71 0.10 0.00 0.10 0.20 0.00 0.00 0.00
N06230 1.30 21.90 59.70 0.28 1.20 0.38 0.42 0.02 14.20 0.49 0.01 0.00 0.00
R30556 32.50 21.27 21.31 18.09 2.88 0.17 0.33 0.00 2.38 0.96 0.00 0.00 0.00
N06600 7.66 15.40 75.81 0.00 0.00 0.32 0.16 0.00 0.00 0.29 0.32 0.00 0.00
N06601 13.53 23.48 60.00 0.06 0.16 1.26 0.50 0.27 0.00 0.31 0.38 0.00 0.00
N06617 0.76 22.63 53.20 12.33 9.38 1.15 0.15 0.27 0.00 0.02 0.05 0.00 0.00
N06625 2.66 21.74 62.79 0.00 8.46 0.10 0.41 0.19 0.00 0.10 0.00 0.00 3.52
N08810 44.22 21.22 31.71 0.00 0.00 0.33 0.60 0.41 0.00 0.92 0.51 0.00 0.00
C11000 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 0.00 0.00
N08120 34.53 25.12 37.44 0.11 0.37 0.11 0.57 0.02 0.10 0.73 0.18 0.00 0.66
N12160 8.00 28.00 34.30 27.00 0.00 0.00 2.70 0.00 0.00 0.00 0.00 0.00 0.00
S67956 75.22 19.40 0.28 0.05 0.00 4.50 0.11 0.33 0.00 0.09 0.00 0.00 0.00
N02270 0.00 0.00 99.99 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
RE means rare earth elements.
1.E + 02 formation of corrosion products, which leads to metal

C11000 loss (penetration), and occurs upon exposure of metals
to gases, which contain the reactive species needed
1.E + 01 to form the corrosion products. An X-ray analysis by
diffraction of a surface scale sample is a common
Penetration (mm)
Nickel 201 method to determine the identity of corrosion pro-

ducts. An alternate method is to evaluate the gas
1.E + 00
composition. Surface metal loss (scaling) can be
detected by methods, which measure the metal thick-
ness, such as ultrasonic and mechanical methods.
1.E – 01 G10200
Sound metal loss by internal corrosion can only be
detected by metallography, as schematically shown
in Figure 1.
1.E – 2
200 300 400 500 600 700 The time dependence of high temperature corro-
Temperature (⬚C) sion is controversial3–5 and depends on the combination
Figure 9 Effect of temperature upon sulfidation of some
of alloy, temperature, time, and corrosion mechanism.
simple metals after 1 year in 0.5 atm H2 and 0.05 atm H2S. Reported time dependencies include: parabolic (metal
loss proportional to time0.5), linear (metal loss propor-
tional to time), power law (metal loss proportional to
products, which dominate. Examples are: sulfidation/ timex), and various combinations of these dependencies.
oxidation means that sulfides and oxides form, oxida- If the corrosion product scale remains undisturbed on
tion/carburization means that carbides and oxides the alloy surface and sufficient time has passed (in
form, oxidation/nitridation means that nitrides and excess of 1000 h), a reasonable approximation is that
oxides form, and so on. Corrosion leads to the the time dependence is parabolic. Removal or cracking
1.0E + 1 1.00
S50400
G10200
S32100
1.0E + 0
Penetration (mm)
S41000
Penetration (mm)
S31600
1.0E − 1
0.10 S30400
S30400
1.0E − 2
S31000 S34700
N08810 S30415
1.0E − 3
300 400 500 600 700 800 0.01
Temperature (⬚C) 300 400 500 600 700
Temperature (⬚C)
Figure 10 Effect of temperature upon sulfidation one year
corrosion of some alloys commonly used in sulfidizing
conditions after 1 year at 0.5 atm H2 and 0.05 atm H2S. Figure 12 Effect of temperature upon sulfidation
corrosion of ‘18-8’ steels exposed to 0.5 atm H2 and
0.05 atm H2S after 1 year.
1.00
S30815 Sulfidation/oxidation by CO–CO2–COS and H2–
H2O–H2S gases – key variables, which influence
S30400 the kinetics, are exposure time, PO2, PS2, alloy
S44600 composition, and temperature;
Penetration (mm)
Oxidation/nitridation by O2–N2 gases – key vari-

0.10
ables, which influence the kinetics, are exposure
S31000
time, partial pressures of O2 and N2, alloy compo-
sition, and temperature;
INCOLOY DS
Oxidation/carburization by CH4–H2O gases – key
variables, which influence the kinetics, are expo-
sure time, PO2, activity of carbon (aC), alloy com-
N12160 position, and temperature.
0.01
1.E − 04 1.E − 03 1.E − 02 1.E − 01 1.E + 00
PH2S (atm)
Figure 11 Effect of H2S partial pressure upon 1.11.2.1 Thermochemical Calculations to

sulfidation corrosion of some very resistant alloys at Predict Oxide–Sulfide–Carbide–Nitrides
550 oC after 1 year. on Alloys
Since many alloys may be used in a wide range of
of the surface scale will tend to increase the rate of gases, which can induce corrosion producing mix-
corrosion, because the presence of the scale is a partial tures of corrosion products, it is wise to examine the
barrier, which tends to slow corrosion as time passes, types and combinations of potential corrosion pro-
but corrosion still continues. Some studies suggest ini- ducts. One approach is to use thermochemical calcu-
tial linear time dependence for several thousand hours. lations to predict the stable corrosion products
Corrosion data measured after hundreds of hours (as defined by equilibrium reactions between the alloy,
is often the case for available data) are unlikely to be gas, and alloy temperature. The surface of equipment
useful in estimating sulfidation corrosion rates for exposed to large amounts of flowing gas can be
long-term service. Corrosion data are properly used approximated by a small amount of alloy exposed to
when the time dependence is considered. The corrosion a large amount of gas. For example, a ratio of 1 mol
mechanisms considered and some of the important alloy to 1000 mol of gas will not deplete the gas of
variables are listed below: important corrosives in the calculation. This
10.00 1.00
1.00
0.10
Calculated (mm)
Calculated (mm)
0.10
0.01
0.01
0.00 0.00
0.00 0.01 0.10 1.00 10.00 0.00 0.01 0.10 1.00
Measured (mm) Measured (mm)
Figure 13 Example of correlations of sulfidation data for S30400 (left) and N08810 (right) using the formalism described in
this chapter. Corrosion by mixed gases.
approach ignores differing reaction and diffusion Al–Co–Cr–Fe–Mn–Ni–Si–Ti in oxides,

rates of the species in the alloys and corrosion pro- Fe–Ni–Co–Cr in sulfides,
ducts. The calculation also assumes that the major W–Mo–Nb–Cu–Ni–Co–Fe–Mn–Cr–Ti–Si–Al in
elements in the alloys will determine the type of carbides and nitrides,
dominant corrosion products, which is typically W–La–Mo–Nb–Cu–Ni–Co–Fe–Mn–Cr–Ti–Si–
true and will ignore any subtleties of corrosion prod- Al–N–C in the alloy liquid and solid-solution
uct morphologies and kinetics. phases.
To this end, a capability to perform thermochem-
ical models for complex corrosion product phases was The effort included development of mathematical
developed to enable the required calculations. The models, based on the known solutions and described
elements considered in the thermochemical calcula- the thermodynamic properties as functions of tem-
tions are: perature and composition. The models were opti-
mized using all the available thermodynamic and
WLaMoNbCuNiCoFeMnCr phase diagram data from the literature in order
TiArClSSiAlFONCH to obtain a set of model parameters for the two-
component, three-component, and higher-order sub-
With all of the known corrosion product phases having
systems. The resulting models were then used to
been analyzed in terms of the thermochemical solu-
estimate the thermodynamic properties of the N-com-
tion behaviors of the known species in each phase,
ponent solutions from the database of parameters for
the assumption of pure and single component con-
lower-order subsystems. Examples of the species group-
densed phases, which is a common assumption, is no
ings are listed below:
longer required. In order to structure the calcula-
tions, the thermochemical literature was extensively Spinel – (Fe2+, Fe3+, Co2+, Co3+, Cr2+, Cr3+, Ni2+,
reviewed. Thermochemical data such as free energies Mg, Al) [Fe2+, Fe3+, Co2+, Co3+, Cr3+, Ni2+, Mg,
of formation, heat capacities, heats of mixing, heats of Al]2O4;
fusion, heats of evaporation, and second-order phase Corundum – Fe2O3–Cr2O3–Al2O3;
transition were also considered and compiled. Models Monoxide – FeO–CoO–NiO–CaO–MgO–ZnO–
were then constructed to describe the solution thermo- CuO.
dynamics of the complex sulfide, alloys, carbides,
oxides, and intermetallic phases, as described.22–24 For oxide phases in the Fe–Co–Ni–Cr–O system,
The elements considered to occur in the solution thermodynamic data are available mainly for the
phases, are: following subsystems:
Fe–O, Co–O, Ni–O, Cr–O Significant research has been carried out over
Fe–Cr–O, Fe–Ni–O, Fe–Co–O the last 20 years by investigators such as Perkins,14,15
Natesan,16–18 Bakker,19 and John20,21 to define how
The models were evaluated by comparison to the
the gas composition influences the kinetics of alloy
available thermodynamic and phase equilibrium
corrosion in sulfidizing/oxidizing conditions. It is
data. Comparison between the model results and
well accepted that the thermochemical interactions
known phase diagrams indicated that the model pre-
of the alloys and the exposure conditions influence
dictions could reproduce the measured data within
the stabilities (and formation) of the corrosion pro-
the experimental error limits from 25 C to above the
ducts and the rates of corrosion. The referenced
liquidus temperatures at all compositions and oxygen
studies and others show that the corrosion products,
partial pressures from metal saturation to equilibrium
which form may not necessarily be those expected, are
with oxygen. Examples of the high-quality fits of the
based upon simplified thermochemical expectations
thermochemical models to the measured data are
using Fe–S–O, Cr–S–O, and Ni–S–O systems for the
shown in Figures 5–7. The model for the sulfide
computations and assuming pure component corrosion
liquid phase contains Fe–Ni–Co–Cr–S and uses the
products. In fact, gas mixtures corroding real alloys
quasichemical model pair approximation22–24 to take
expected to form pure Cr2O3-type corrosion products
into account strong short-range ordering. For solid
actually require gas PO2 values of 2 or 3 orders of
nonstoichiometric sulfide solutions in the Fe–Ni–
magnitude greater than the calculated value thought
Co–Cr–S system, appropriate thermodynamic mod-
to form Cr2O3. This has been interpreted to be a
els were developed. Parameters of these models were
‘kinetic boundary,’ meaning that an excess PO2 value
optimized to reproduce all solid–solid and solid–
is needed to precipitate and grow the Cr2O3. This is
liquid phase equilibria.
different in concept from a thermochemical criterion
The main data that were considered in the models
for the relative stabilities of corrosion products, as
were:
defined by combinations of PO2 and PS2, at a given
Solid sulfide/liquid sulfide equilibrium, partial temperature. It was suggested that the high growth
pressure data of S in the liquid sulfide over the rates of sulfides cause high PO2s required to form the
whole composition range from pure metals to sul- Cr2O3-rich corrosion products. However, thermochem-
fur-rich liquid ical calculations possible with the models now available
Sulfur activity over the liquid sulfide and over can calculate the real corrosion product compositions.
solid (Fe, Ni)S solutions These thermochemical solution calculations now sug-
Sulfide/metal/oxide/gas equilibria gest that the ‘kinetic boundaries’ are really equilibrium
Heats of mixing, heat content, heat capacities, etc. phase boundaries, when the real phase compositions
Metal activity data and thermochemistry are considered.
Metal–liquid sulfide equilibrium data On the basis of the logic presented above, the
Liquid–oxide equilibrium data. strategy adopted by John20,21 has been used to corre-
late sulfidation/oxidation corrosion with the relative
This technology is embodied in the ASSET (Alloy
stabilities of the Cr2O3 and CrS phases as expressed
Selection System for Elevated Temperatures) tech-
in terms of PO2 and PS2, as shown below:
nology and is widely available.10,25,32,33
2CrS þ 3=2O2 ¼ Cr2 O3 þ S2 ½2
1.11.2.2 Sulfidation/Oxidation by
CO–CO2–COS and H2–H2O–H2S Gases P O3=2 K2 ACr2 O3
2
¼ ½3
Sulfidation corrosion in the presence of oxidizing PS2 ACrS2
gases is called sulfidation/oxidation, which occurs where K2 is the equilibrium constant for reaction [2].
when both sulfide and oxide corrosion products A measure of the relative stability of Cr2O3 and CrS
form. The key variables, which influence the kinetics, (or the activity factor AF), is defined below:
are exposure time, partial pressures of O2 and S2,
in gases like CO–CO2–COS or H2–H2O–H2S or AF ¼ logðACrS2 =ACr2 O3 Þ ½4
CO–CO2–COS–H2–H2O–H2S, alloy composition,
and temperature. The corrosion leads to metal losses where A indicates the activity of the subscripted
by both surface scaling and internal corrosion prod- species. High AF values correspond to high activities
uct formation (penetration). of CrS (ACrS) values and rapid corrosion, while low
AF values reflect low ACrS values and slow corrosion.

It has been experimentally well demonstrated that
sulfidation/oxidation kinetics correlate with AF and
that sulfidation/oxidation kinetics are parabolic with
time, a parabolic rate constant is written in terms of
the gas composition, as related by eqns [3] and [4]
and then used to correlate corrosion by sulfidation/
oxidation with exposure conditions:
logðKp Þ ¼ A þ BðlogðPS2 Þ 1:5logðPO2 ÞÞ þ C=T ½5
where Kp is a parabolic rate constant for the reactions
leading to total penetration, A, B, and C are experi-
mentally determined constants, and T is temperature.
This relationship has been demonstrated to correlate
alloy corrosion by sulfidation/oxidation for many
different alloys over wide ranges of temperatures
and gas compositions.8–10,20,21 An example of this
type of correlation between measured corrosion and
exposure conditions is shown in Figure 14 for a
common heat-resistant alloy N06600. Once these
types of correlations for an alloy are established,
alloy corrosion in sulfidation/oxidation conditions
can be made for many alloy/condition combinations.
The utility of these correlations is obvious. Accurate
predictions of sulfidation/oxidation for wide ranges
of conditions in terms of gas composition, tempera-
ture, and time can be made and expressed in very
practical engineering units related to change in wall
thickness.
1.11.2.3 Thermochemistry and Corrosion

Mechanism Figure 14 Illustration of sulfidation/oxidation correlation
for alloy N06600 in various combination of exposure
The utility to predict the formation of alloy corrosion conditions, in terms of temperature, time, PO2 and PS2.
products is illustrated in Figures 15–17. The results
compare very well with practical experience and sum-
marize the stable corrosion products for ranges of gas prevent the formation of the liquid sulfide. An iron-
compositions and temperatures. Nickel-base alloys based stainless steel, like S30400, does not form liquid
like N06600 and N06625 form liquid sulfide at tem- sulfide, but does form large amounts of solid sulfides.
peratures slightly above 1100 F (866 K), while an A Ni–Cr–Co alloy like N12160 with high concentra-
Fe–Cr–Ni stainless steel like S30400 and a heat-resis- tions of Cr, Si, and Co shows large concentrations of
tant Ni–Cr–Co alloy like N12160 form liquid sulfide oxides and no liquid sulfide formation, until higher
corrosion products only at much higher temperatures. H2S concentrations are present, which should be
Nickel-base alloys form liquid nickel sulfide corrosion excellent for resistance to corrosion in these condi-
products, if any portion of the corrosion product layer tions, as is the case for the Ni–Cr–Co alloy.
is in conditions, which stabilize the liquid. Through
the thickness of the corrosion product, one might
1.11.2.4 Laboratory Simulation
think of a gradient of H2S, as the corroding alloy at
the alloy/corrosion product interface consumes sulfur It is important to remember that the usual goal of
species. Increasing the amount of Cr will increase laboratory corrosion tests is to produce informa-
the amount of oxide in the corrosion product for a tion, which will help decisions related to industrial
nickel-base alloy (as found in N06025) but will not equipment. Industrial equipment often requires
S30400
N06600 100
100
Phase amount (mole %)

Phase amount (mole%)
80
80
60 60
40 40
20 20
0 0
1000 1100 1200 1300 1400 1000 1100 1200 1300 1400
Temperature (⬚F) Temperature (⬚F)
MeS (l) MeS (s) Me2O3 MeS (s) Me3O4
N06625 N12160
100
100

80 80
60 60
40 40
20 20
0 0
1000 1100 1200 1300 1400 1000 1100 1200 1300 1400
Temperature (⬚F) Temperature (⬚F)
MeS (l) MeS (s) Me2O3 MeS (l) MeS (s) Me2O3
Figure 15 Calculated corrosion products for a small amount of alloy exposed to a large amount of 5% H2–65% H2–30%
H2O at 1 atm, 1000–1400 F (811–1033 K).
lifetimes of years at high pressures, high temperatures the thickness of the subsurface metal layer that is
and requires considerable expense and time to fabricate. affected by the corrosion exposure. Such a subsurface
Laboratory tests aim to quickly and cheaply define the layer may have internal precipitates of corrosion
likely corrosion phenomena and kinetics prior to build- products, pores, voids, or cracks, which have been
ing the industrial equipment. For the case of sulfidiz- produced by reaction with the corrosive environ-
ing/oxidizing conditions, some important issues will ment. This is illustrated in the schematic view of
now be discussed, including the effects of temperature, a cross section of a corroded alloy shown previously
exposure time, and gas composition. in Figure 1. Some of the alloys studied are shown in
Laboratory corrosion tests often intend to acceler- Table 1.
ate corrosion relative to exposures in industrial equip- Many process gases contain H2–H2O–H2S gas
ment and to speed assessment of many alloys and their mixtures, which will determine the partial pressures
corrosion with coupons. This is cheaper, faster, and of O2 and S2 (PO2 and PS2) and which are in turn well
more reliable than gathering corrosion data by known to determine the sulfidation/oxidation corro-
inspecting full-scale equipment in operating processes. sion behaviors for Cr-containing alloys. As the fol-
Examples of process equipment where sulfidation/ lowing reactions show, chemical thermodynamics can
oxidation may occur are processes in petroleum refin- be used to calculate the PO2 and PS2 values for C-
ing such as hydrocrackers, hydrotreaters, coal/coke/ based gases (CO–CO2–COS):
oil gasifiers, crude-oil distilling columns, vacuum
CO þ 1=2O2 ¼ CO2 ½6
flashers, sulfur-removal plants, coking units, and
flexicokers. Corrosion will be discussed in terms of PCO2
the total metal penetration, which is defined as the Keqn½6 ¼ 1=2
½7
PO2 PCO
sum of the metal lost by surface scale formation plus
N06600
100 S30400
100

80
80
60 60
40 40
20 20
0 0
0.1 1 10 100 0.1 1 10 100
% Steam % Steam
MeS (l) MeS (s) MeS (s) Me3O4
Me2O3 Me3S4 Me3S4
N06625 N12160
100 100

80 80
60 60
40 40
20 20
0 0
0.1 1 10 100 0.1 1 10 100
% Steam % Steam
MeS (l) Me2O3 MeS (s) Me 2O 3

Me3S4 SiO 2
Figure 16 Calculated corrosion products for a small amount of alloy exposed to a large amount of gas of 5% H2S, 50% H2,
and X% H2O at 1 atm at 921 K.
2
PCO PH2 2 S
PO2 ¼ 2
2
2 ½8 PS 2 ¼ 2 ½14
Keqn½6 PCO Keqn½13 PH2 2
CO þ 1=2S2 ¼ COS ½9 H2 þ 1=2O2 ¼ H2 O ½15
PH2 O
PCOS Keqn½15 ¼ ½16
Keqn½9 ¼ 1=2
½10 1=2
PO2 PH2
PS2 PCO
PH2 2 O
2
PCOS PO 2 ¼ 2 ½17
PS2 ¼ ½11 Keqn½15 PH2 2
2 2
Keqn½9 PCO
where Keqn values are the equilibrium constants for
While the PO2 and PS2 values can be calculated for the reaction number indicated.
H-based gases (H2–H2O–H2S): The H-based gases are found in many industrial
processes. However, use of these gases as test gases in
H2 þ 1=2S2 ¼ H2 S ½12 small-scale laboratory testing involves some practical
difficulties, including producing, handling, and
PH2 S
Keqn½12 ¼ 1=2
½13 controlling small amounts of H2O vapor (steam).
PS2 PH2 Using H2O in the test gases requires attention to
S30400 at 977 K
S30400 at 921 K 100
100
Phase fraction (mole %)

80
80
60
60
40 40
20 20
0 0
0.01 0.1 1 10 0.01 0.1 1 10
H2S (mole %) H2S (mole %)
Me2O3 Me3O4 Me3O4 Me3S2 (s) Alloy
Alloy MeS (s) MeS (s) MeO
N06600 at 921 K N06600 at 977 K

100
100

80 80
60 60
40 40
20 20
0 0
0.01 0.1 1 10 0.01 0.1 1 10
MeS (l) Me2O3 Me3O4 MeS (l) Me2O3 Me3O4

Me3S2 (s) Alloy Me3S2 (s) Alloy
N12160 at 921 K N12160 at 977 K

80
100
60 80
60
40
40
20
20
0 0
0.01 0.1 1 10 0.01 0.1 1 10
MeS (l) Me2O3 Me3O4 Me3O4
MeS (l) Me2O3
Alloy MeS (s) SiO2 Me3S2 (s) Alloy MeS(s)
SiO2
Figure 17 Effect of H2S concentrations upon formation of liquid corrosion products for several alloys exposed to X% H2S,
60% H2, 30% H2O, and balance Ar at 1 atm.
several concerns: continuous production of small We can also use CO–CO2–COS gases to simulate
quantities of steam and heat tracing of the test appa- the PO2 and PS2 values of the H2–H2O–H2S gases. Use
ratus to prevent steam condensation during the tests. of CO–CO2–COS gases, rather than H2–H2O–H2S
Condensation of the steam would change the gas gases avoids the difficulties of managing steam in
composition and the corrosion conditions in the small amounts and losses due to condensation. The
tests. H2–H2O–H2S gases are widely used in corro- ambition is to make laboratory tests easier and more
sion testing. However, there is an alternative. reliable, particularly during long-term corrosion tests.
C-based gases can simulate the PO2 and PS2 values compositions shown. C-based and H-based gases can
and corrosion behaviors of H-based gases. An example interchangeably simulate equivalent thermochemistries
of a comparison of some values of PO2 and PS2 for a of the corrosive conditions, as defined in terms of PO2
C-based gas of 32% CO–1% COS–balance CO2 at and PS2. Figure 19 shows that the AF values for C-based
1 atm, which was chosen to simulate the H-based gas and H-based gases can also be closely matched. It is
of 10% H2–10% H2S–25% H2O–balance Ar at 1 atm is not important that the values of AF, PO2, and PS2 in
shown in Table 2. The PO2 and PS2 values closely match the laboratory test conditions exactly match the condi-
between the C- and the H-based gases. Figure 18 illus- tions of industrial interest, because correlations between
trates the excellent agreement for PO2 and PS2 values these variables can be defined and used to interpolate/
between these C- and H-based gases over the tempera- extrapolate/predict corrosion for ranges of conditions
tures of 1200–1800 F (648–982 C) for the example of interest, once a correlation is established.
Table 2 Summary of calculations comparing PS2, PO2, and AF values in C-based and H-based gases, where
AFðactivity factorÞ ¼ logððaCrS Þ2 Þ=ðaCr2 O3 Þ, corrosion correlates with, logðPenetration=Hours0:5 Þ ¼ A þ B½logðPS2 Þ 1:5logðPO2 Þþ
C=absolute temperature
Temperature ( F)
1200 1400 1600 1800
H2–H2S–H2O PS2 atm 8.3 E06 1.0 E04 7.5 E04 3.3 E03
gases PO2, atm 3.0 E22 3.2 E19 8.5 E17 7.8 E15
AF 7.54 6.51 5.69 5.04
MeS 79 79 79 79
Me2O3 18 18 18 18
CO–COS–CO2 PS2 atm 9.7 E05 9.7 E05 5.9 E04 1.7 E03
gases PO2, atm 4.9 E23 1.4 E19 8.0 E17 1.5 E14
AF 6.37 5.99 5.76 5.76
MeS 79 79 79 85
Me2O3 18 18 18 0
Me3O4 0 0 0 11
The table is calculated for 10% H2–10% H2S–25% H2O–balance Ar and 32% CO–1% COS–balance CO2 at 1 atm and corrosion product
quantities are calculated for UNS S34700.
1.E + 00
1.E – 05
PS2 or PO2 (atm)
1.E – 10
1.E – 15
1.E – 20
1.E – 25
1200 1400 1600 1800
Temperature (⬚F)
PS2 in H gases PO2 in H gases

PS2 in C gases PO2 in C gases
Figure 18 Summary of calculations comparing PS2 and PO2values in C- and H-based gases where C-based gas means
CO–CO2–COS and H-based gas means H2–H2O–H2S.
AF values in H and C gases 100

0 330 ⬚C 2
400 ⬚C
–2 50 500 ⬚C
600 ⬚C
AF value
1
–4 700 ⬚C
20 0.5
–6
600, 700 ⬚C
10 500 ⬚C
–8
Surface recession (mils)
Surface recession (mm)

1200 1400 1600 1800 0.2
Temperature (⬚F)
5
H gases C gases 0.1
Figure 19 Summary of calculations comparing AF in
C- and H-based gases where C-based gas means 2 400 ⬚C 0.05
CO–CO2–COS and H-based gas means H2–H2O–H2S.
1
0.02
0.5 330 ⬚C
2 0.01
Instantaneous log Kp (mg cm–2)2 h–1
UNS S41000 – gas A

1 0.2 0.005
UNS N08800 – gas B
0 0.1
100 500 1000 5000 10 000
Alloy 671 – gas B Time (h)
–1
Figure 21 Sulfidation/oxidation corrosion of alloy UNS
–2 N06625 in a gas of initial composition of 0.1% H2S–79.1%
Alloy 160 – gas B H2–20.8% CO at 1 atm.
–3
–4 it is independent of time). Exposure times greater than

0 100 200 300 400 500 600 500 h may be needed to measure real parabolic rate
Time (h)
constants useful to predict very long-time corrosion
Figure 20 Effect of time upon the instantaneous parabolic behavior. Indeed, after the initial time of changing
rate constants for sulfidation/oxidation of several alloys at kinetics, the long-term corrosion behavior can be
600 C. Gas A is 2% CO–2% COS–balance CO2 and Gas
B is 48% CO–48% CO2 and 4% COS.
shown to be parabolic for many thousands of hours
for many Fe–Ni–Cr heat resistance alloys, with an
example shown in Figure 21.
Once the time dependence is understood, the next
1.11.2.5 Kinetics
question becomes whether real corrosion behavior in
It is not sufficient to only confirm that AF, PO2, and PS2 equilibrium and nonequilibrium H-based and C-based
thermochemical values of gases can be correlated gases can be correlated in the same way. This question
between H-based and C-based gases in order to assure arises because many industrial processes handle gases,
that corrosion is adequately related between the gas which are not equilibrated and alloy corrosion occurs
types. The relation between measured corrosion by contact with nonequilibrium gases. Nonequilibrium
kinetics for the gas types and their compositions gases occur when the combinations of gas flow rates,
must also be examined. Many studies either suggest temperatures, and gas phase reaction rates do not
or assume that the time dependence of the extent of allow sufficient time to equilibrate the gases.
hot gas corrosion is parabolic. Figure 20 shows that Corrosion in equilibrated and nonequilibrated gas
the measured values of the instantaneous parabolic compositions has been compared. Previous corro-
rate constants for a number of alloys exposed to sulfi- sion tests25,26 showed that if the PO2 and the PS2 values
dizing/oxidizing conditions can vary greatly from the are matched between CO–CO2–COS gases and
initial exposure time. It may take up to 500 h before H2–H2O–H2S gases, then the corrosion behaviors
the parabolic rate constant becomes constant (meaning of Fe–Ni–Cr alloys are equivalent in terms of the
corrosion products formed, corrosion product is no distinction between the C-based and H-based
morphologies, and the measured corrosion kinetics. gases, in terms of correlating PO2 and the PS2 with
This is illustrated in Figure 22, where the gas com- corrosion behavior for alloy UNS N08800. This is
positions of various C-based and H-based gases at true for a variation of nearly six orders of magnitude
equilibrium and not at equilibrium were used to change in the gravimetric parabolic rate constants for
expose N08800. All the different types of gases alloy N08800, showing that C-based and H-based
induce similar corrosion behavior, once the AF (or gases can be used interchangeably to simulate corro-
PO2 and PS2 values) are considered. These gases cover sion conditions, as long as the PO2 and PS2 values are
variations of about 10 orders of magnitude in both assessed. This approach has been used for a large
PO2 and PS2. The open symbols are C-based gases, number of heat-resistant alloys based on various
while the closed symbols are H-based gases. The combinations of Fe–Ni–Cr–Co–Al.
solid points are equilibrium gases and open points After demonstrating that AF (or PO2 and PS2 values)
are nonequilibrium gases. Figure 22 shows that there can define the corrosion condition for the alloy; an
log (PS /PO3/2)

2 2
20 25 30 35 40
2
10
101
10–4
100
10–5
Kp (mg2 cm–4 h)
Kp (mm2 h–1)
10–1
10–6
Oxidation + Normal
Normal oxidation
sulfidation transition sulfi-
dation
10–2
10–7
(2 ⫻ 10–5 mg2/cm4-h)
10–3
–16 –12 –8 –4 0 4
log (a2CrS/aCr O )
2 3
Figure 22 Effect of gas composition expressed in terms of AF (comparing C-based and H-based gases) upon the
instantaneous parabolic rate constant (Kp) for UNS N08800 exposed to sulfidizing/oxidizing conditions at 600 C.
alternative approach is to prepare the phase stability lines. A potential reaction between CO and CO2,
diagram shown in Figure 23. The diagram includes which can lead to carbon deposition, is shown below.
iso-AF lines, which suggest lines of equal corrosion CO2 þ C ¼ 2CO ½18
kinetics, as defined by AF (or PO2 and PS2) values. The
next step is to then add isocorrosion contours, which 2
PCO
are parallel to the phase boundary between the CrS aC ¼ ½19
Keqn½18 PCO2
and Cr2O3 phase fields for the diagram for Cr–S–O–C
at 600 C, as shown in Figure 24. This accomplish- where aC is the activity of solid C described in reaction
ment is quite remarkable in using thermochemical [18] and calculated in eqn [19]. In order for the
information on corrosion product stability diagrams CO–CO2–COS gases to adequately simulate the cor-
to compile and correlate corrosion kinetics of real rosivity of the H2–H2O–H2S gases, conditions must
engineering alloys in complex gases. Indeed, a collec- also not allow solid carbon to form. This can be con-
tion of these types of diagrams could be constructed for firmed experimentally by analyses of the corrosion
numerous combinations of alloys and exposure condi- products and can also be assessed by calculating the
tions of gas compositions and temperatures. However, stabilities of the potential phases, while considering
the utility represented in ASSET,10,25,32,33,52,53 avoids oxides, carbides, sulfides, and alloy phases. Formation
the cumbersome preparation of many diagrams and of carbides in conditions which induce sulfidation/
allows accurate predictions as needed. oxidation is rare, since oxides and sulfides are usually
One concern with CO–CO2–COS gases in labora- much more stable than carbides. Safety and toxicity
tory tests is that compositions and temperatures must concerns with COS are similar to those with H2S,
be maintained to keep carbon activities (aCs) < 1, in and the methods to remove COS and H2S from exiting
order to avoid carbon deposition, which could present test gases by passage through aqueous NaOH solutions
practical problems in terms of unexpected changes in before exhausting them into the atmosphere. Per-
the gas composition and potential plugging of gas flow forming these types of corrosion tests requires careful
log (PH2S/PH2)2
–10 0
log (PCOS/PCO)2
0
Cr2(SO4)3(s)
20 20
–10
10
10
log (PH2O/PH2)2
log (PCO2/PCO)2
Cr2O3(s)
log PO2 (atm)
log (a2CrS/aCr2O3) values

–20
–15
0
–10 0
–5
–30 0
5
–10
–10
Cr3C2(s) CrS(s) Cr6S7(s)
–40
–20 –10 0
log PS2 (atm)
Figure 23 Calculated thermochemical phase stability diagram for C–S–O system at 600 C. Round points represent
H-based gases and square points represent C-based gases. Solid points represent equilibrium gases and open points are
nonequilibrium gases.
log(PH2S/PH2)2
–10 0
2
–10 log(PCOS/PCO) 0
0
Cr2(SO4)3(s)
20
20
–10
10
10
s
mm
log (PH2O/PH2)2
log (PCO2/PCO)2
log PO2 (atm)
Cr2O3(s) 7
0.0 9
n 0.0
–20 id atio 1
Ox 0.2
0
0.3
K p’s 6
0.6 0
–7 0
10 4
5 ⫻ –6 0.9
10 –6
10 n
5 ⫻ –5
atio
–30 10 –5 lfid
10 Su
5 ⫻ –4
10
–10
–10
Cr3C2(s) CrS(s) Cr6S7(s)
–40
–20 –10 0
log PS2 (atm)
Figure 24 Isocorrosion contours for alloy N08800 on the calculated Cr–S–O thermochemical phase stability diagram at
600 C. Values for Kp’s in square millimeter per hour and metal loss (in mm) after 1 year of exposure are shown.
control of the test atmosphere. For example, the air Let us examine the utility of these concepts. The
must be adequately removed from the test chamber presence of oxidizing gases such as H2O or CO2 slows
before the high temperature test gas contacts the alloy the sulfidation rate because they increase the stability
coupons. The test chamber must be purged with gases of slow-growing oxide corrosion products. The cor-
like N2 or inert gases until the O2 decreases to values rosion can be much slower than expected if only
below 0.001 atm. Then, H2 can be introduced to the H2S–H2 were present. This can be important because
test chamber at temperatures above 1000 F to scav- gases, sometimes thought to contain only H2S–H2,
enge the remaining low levels of O2 by reaction to also contain some H2O vapor, because of exposure of
form H2O before the corrosion test exposure starts. the gas to liquid water. A gas exposed to water at room
Another concern that might be raised is whether temperature (from processes such as a water wash)
C-based gases can adequately simulate H-based gases may contain up to 2–3% water in the gas, if saturated
in terms of any potential effects of water vapor upon with water at room temperature. Sulfidation rates pre-
the corrosion behavior. The reported observations to dicted using only the H2S–H2 concentrations would
date suggest that the effects of water vapor upon overestimate the corrosion, if H2O were really present.
oxidation are associated with oxidizing conditions This slower corrosion rate by sulfidation/oxidation
where volatile, oxidized, hydrated species form. The can be predicted using the approaches described in
conditions that produce sulfidation/oxidation have this article. This is illustrated in Figures 25 and 26 for
PO2 values of perhaps 1025 to 1010 atm, and the several alloys at 600 C in sulfidizing/oxidizing gases.
majority of corrosion products are sulfides. The PO2 The extent of corrosion occurring at the right-hand
values are well below those associated with condi- PO2 in Figure 26 and the low AF values in Figure 26
tions showing oxidation to be affected by water vapor. corresponds to simple oxidation, while progression to
Oxides Sulfides + oxides Sulfides
3 UNS N066000
2
UNS S41000
0
UNS N08800
(penetration
(penetration
after 1 year)
after 1year)
(mm)
(mils)
Log
1
Log
671 –1
Alloy 160
Alloy
0
–2
–1
–3
–2
–10 –8 –6 –4 –2 0
log(a2CrS/aCr2O3)
Figure 25 Influence of gas composition measured in terms of AF upon corrosion of high-chromium alloys corroding by
sulfidation/oxidation.
3E + 2 sulfidation in H2S–H2 gases. Several alloys are shown

S30400
1E + 2 in Figure 26, with both thick and thin portions of the
lines. The thick portions represent predictions based
1E + 1 upon interpolations for conditions for which there
Alloy 671
Penetration (mils)
are compiled corrosion data for that alloy. The thin

1E + 0 portions of the lines represent corrosion predictions
S44600 N08810
based upon extrapolations for conditions outside the
1E – 1 data represented by the data compilation. Figures 25
and 26 clearly show that data compilations for a num-
S67956
1E – 2
ber of commercially important alloys can represent
many orders of magnitude of variation in the exposure
conditions, as summarized in terms of the PO2, PS2,
1E – 3
1.00E – 20 1.00E – 10 1.00E – 10 temperature and exposure time. Figure 27 shows
P(O2) (atm) how variations in steam concentrations can affect the
Figure 26 Effect of O2 partial pressure upon sulfidation/
sulfidation/oxidation of S030400, which is typical for
oxidation of alloys after 1 year of time and exposure to S2 many alloys. Figure 28 shows an example of a corre-
partial pressure of 107 atm at 700 C. lation for sulfidation/oxidation for alloys S30400.
the left-hand side of the axis corresponds to lower PO2s

1.11.2.6 Corrosion Influenced by Gas
and the high values of AFs are expected in O2-
History
depleted conditions. When the PO2 becomes low
enough and oxides are no longer stable, sulfidation Another way to examine the approach to equilibrium
becomes the dominant corrosion mechanism. The for an alloy corroding by sulfidation/oxidation is to
minimum corrosion in this scenario is oxidation in examine how promptly changes in gas composition
O2-containing gases and the maximum corrosion is influence subsequent alloy corrosion behavior, since
100.0
0% H2O
0.1% H2O
10.0
Penetration (mils)
1% H2O
10% H2O
45% H2O
1.0
0.1
500 700 900 1100 1300 1500
Temperature (⬚F)
Figure 27 Effect of temperature and steam concentration upon metal penetration by sulfidation/oxidation of S30400 after
1 year in 5% H2S, 50% H2O, X% H2O and balance Ar.
the gas thermochemistry should determine the alloy mechanical properties in different kinds of service,
corrosion behavior. Data reported some time ago21 for such as vessels and pipelines in oil and gas equipment
sulfidation/oxidation of UNS N08800 exposed to and other components. Their high temperature corro-
gases, which induced sulfidation/oxidation, showed sion behaviors depend upon the composition of the
that corrosion kinetics changed promptly as the gas gases and temperatures. Surface scales and internal
composition and temperature were changed to corre- (subsurface) oxidation are both possible and seen in
spond to that measured when the alloy had always many different applications. Corrosion by both nitri-
been exposed to that environment. The previous expo- dation and oxidation are possible when alloys are
sure history of the corroding alloy had no effect upon exposed to atmospheres rich in nitrogen at tempera-
the corrosion behavior after a few hundred hours. This tures higher than 1000 F (537 C), nitrides such as
was demonstrated by comparing the parabolic rate con- CrN and Fe4N may form and affect the alloy mechan-
stants for the time periods corresponding to the same ical properties. Nitridation is similar to carburization
gaseous exposure, independently of the prior exposure. in that both are high temperature corrosion mechan-
The alloy corrosion kinetics change within only a hun- isms for metals and alloys which do not produce
dred hours to match those expected for the new gaseous surface scales and do not cause losses in wall thickness,
exposure. There have been several examples of sum- as experienced by other high temperature corrosion
ming the expected amounts of corrosion expected mechanisms, such as oxidation and sulfidation. The
because of different portions of complex exposure as if main consequence of nitridation (like carburization)
they were produced by individual exposures. is embrittlement of the alloys making them vulner-
able to fracture. There is also a slight increase in
volume (decrease in density) as the nitrogen dissolves
1.11.3 Oxidation/Nitridation by N2–O2 into the alloy (or carbon in the case of carburization).
Gases The interstitial solution of the nitrogen into the aus-
tenite phase distorts the lattice.41–44
1.11.3.1 Types of Equipment and
The high temperature oxidation of UNS 30400
Processes Where Nitridation Occurs
steel and other stainless steels in the presence of
Stainless steels and other heat-resistant alloys are used 60% H2–40% N2, NH3–H2O mixtures, and other
because of their excellent corrosion resistances and H2–N2 mixture, has already been studied by several
process occurs upon exposure of metals in gases

containing pure N2, N2 þ H2, NH3, air, and mixtures
of O2-depleted air and other gases at temperatures
above 300–430 C (600–800 F). This corrosion
depends upon the gas composition, alloy composi-
tion, time, and temperature.42,44
A recent corrosion evaluation study was per-
formed on stainless steel tubes used in as a compo-
nent part in a reactor.41 The material was UNS
S30400 and was exposed to a nitriding, carburizing,
and oxidizing environment, containing mainly NH3
and CO2, at temperatures between 390 and 450 C.
The atmosphere had a high nitridation potential.
Alloy nitrides were stable and formed. The authors
explained that the corrosion process includes ingress
(solution) of nitrogen into the steel and then forma-
tion of a nitrided layer, once the solubility products of
the nitrides are exceeded. This formation of CrN
particles at temperatures above 425 C in the near-
surface layer causes a volume expansion of about
11% and a corresponding increase in compressive
stresses in the surface that can lead to crack forma-
tion. Temperatures lower than 425 C lower the Cr
mobility, even if the N2 concentration in the atmo-
sphere is high.
Issartel et al.45,46 analyzed this occurrence of nitri-
dation and the effect of the temperature upon spal-
ling in terms of the corrosion products formed in
UNS S30400 stainless steel in air at 800 and 900 C
(1122 and 1652 F) by using X-ray diffraction and
SEM. Their results support the presence of austenite
at lower than expected temperatures in a nontreated
sample because the nitridation treatment on the steel
surface produced a solid solution of N in the austen-
ite. In this case, no nitrides form in the alloy surface.
Figure 28 Example of the analysis of the sulfidation/
Riffard et al.47 used in situ X-ray diffraction to follow
oxidation corrosion database for S30400 and the quality of the formation of nitrides at the testing temperature.
the corrosion correlation. At the beginning of the oxidation test, CrN formed
together with Fe2O3. Cr2O3 quickly appears and leads
to a protective oxide scale formation growing accord-
authors from the phase transformation point of ing to a parabolic rate law. During oxidation, in situ
view.45–47 It is also well known that nitrogen and X-ray diffraction also shows that Fe2O3 transforms to
carbon dissolving into alloys increase hardness yield FeCr2O4. These results confirm that nitridation
strength, tensile strength, and wear resistance48 and enhances internal oxidation at high temperature.
lower toughness. Nevertheless, there is almost no Internal nitridation is also observed on numerous
previous work dealing with the conditions, which commercial high temperature alloys.49–51 In the case
define the competition between nitridation and oxi- of nickel-base alloys, which are often used for high
dation at high temperatures for these materials. temperature applications, internal nitridation is aided
Nitridation primarily forms internal nitride corro- by repeated spalling and cracking of the protective
sion products and can be described in terms of oxide made of either Al2O3 or Cr2O3. Nitrogen pene-
sound metal loss (penetration) by measuring the trates into the near-surface area of the alloy and forms
depth of the formation of the internal nitrides. The stable nitrides, such as, TiN and AlN.51 These surface
precipitates embrittle the high temperature and low- 304 (UNS S30400) and 410 (UNS S41000), and a
temperature mechanical properties of the material. nickel-base alloy named 214 (UNS N07214). The
Additionally, dissolution of the high-volume fraction nominal chemical compositions of these alloys are
of the initial small precipitates of the cubic g0 -phase reported in Table 1. The stable corrosion products
occurs [Ni3(Al, Ti)], which provides the excellent for conditions of interest were predicted using
creep resistance of Ni-base superalloys at high tem- ASSET and are the likely surface corrosion products
perature. Formation of nitrides can reduce creep and for the alloys exposed to the O2–N2 gases. Corrosion
stress rupture properties of Ni-base superalloys at data reported in the literature have been compared to
high temperatures, lowering the maximum allowable these predictions, in terms of corrosion product for-
stress and/or temperature for a particular application. mation and the metal losses observed.
The transition from oxidation to nitridation in Figures 29 and 30 show the calculated phases for
O2–N2 gases of varying composition for several dif- UNS S30400 and UNS S41000 stainless steels, as
ferent types of alloys (Fe–Cr alloy UNS 410000, functions of the O2 partial pressure (in O2–N2
Fe–Ni–Cr alloy UNS 30400 and Ni–Cr–Al alloy gases) at 1 atm and 1600 F (871 C). The results
UNS N07214) is a function of temperature, O2 par- show that both alloys form the M2O3 phase at O2
tial pressure, and alloy composition in O2–N2 gases at partial pressures higher than 0.001 atm. The corro-
1600 F. The stable alloy phases and corrosion pro- sion products are about 75% Fe2O3 and 25% Cr2O3.
ducts were calculated by ASSET52–54 by thermo- Alloy UNS S30400 has around 8 wt% Ni and differs
chemical equilibrium calculations for varying gas from UNS S41000 in that a significant fraction of the
compositions. The predictions of the phase behavior spinel phase (Fe/Ni)3O4 forms over a broad range of
were compared to and agreed with experimental O2 partial pressures, as shown in Figure 29. The
observations made with carefully exposed coupons 0.5 wt% Ni composition in the alloy UNS S41000 is
of the alloys.53,54 insufficient to contribute to the significant formation
of the spinel phase (FeNi2O4), which occurs for only
a very narrow range of O2 partial pressure.
In the case of stainless steel UNS S30400 shown in
Mechanism
Figure 29, levels of O2 between 0.04 and 0.0634 atm
Oxidation/nitridation by O2–N2 gases occurs in induce formation of an MeO phase (where Me is
various industrial equipment involved in fossil fuel mostly Fe with small amounts of Cr and Ni). Lower
combustion, fertilizer manufacturing, or ammonia levels of O2 tend to produce some M2O3 phase (where
production. The key variables influencing the kinet- M is mostly Cr) with a maximum content of 15%; then,
ics are exposure time, partial pressures of O2 and N2, this phase decreases slowly. For this alloy, very low
alloy composition, and temperature. The effects of levels of O2, tend to form nitrides. Chromium nitrides
variation from a high PO2 to low PO2 can induce formation (CrN and Cr2N) starts at 0.0125 atm of O2
changes in the corrosion products and rates of corro- and continue to increase in concentration at lower PO2s
sion. The oxidation behavior can vary from predomi- until the Cr reacts to make nitride.
nantly surface oxidation (scaling) to oxidation by For the UNS S41000 stainless steel shown in
internal oxidation as the PO2 decreases depending Figure 30, a comparable result was observed. There
upon alloy composition. Progression to lower levels is more MeO phase (FeO) – maximum 85% at an O2
of PO2, will induce nitridation to dominate over oxi- partial pressure of 64 105 atm – and the presence
dation. These observations impact how laboratory of this phase increases at 0.075–0.0275 atm O2. At O2
tests can best simulate corrosion behavior by oxida- partial pressures below 3 105 to 15 105 atm O2,
tion or nitridation and also in terms of applications the formation of iron ferrite forms with the reduction
where low levels of PO2 may be different from the of austenite. This steel has 12.3 wt% of Cr. At very low
expected in O2–N2 gases. levels of O2, the formation of nitrides starts. As seen
from Figure 30, the UNS S41000 stainless steel forms
Cr nitrides (CrN and Cr2N) but less than the values
1.11.3.3 Predicting Corrosion Product and
found in UNS S30400. The lower Cr content (12.0 wt
Alloy Phase Formation in Nitriding/Oxidizing
%) of UNS S41000 naturally forms less nitride phase
Conditions
than UNS S30400 which has a higher Cr content
Three different alloys were investigated: two iron- (18.4 wt%). At even lower partial pressures of O2
base stainless steels, under the specification AISI (5 106 atm), the amount of Cr2N decreases to 3–4%.
100
80 (Fe, Cr)N
Cr2N
Phase quantity (mole %)
SiO2
60 FeO
Fe2O3
M3O4
40 Austenite (Fe, Ni)
Ferrite-Fe
(FeO)2 SiO2
20 M7C3
0
0.000001 0.00001 0.0001 0.001 0.01 0.1 1
PO (atm)
2
Figure 29 Effect of O2 partial pressure upon the stable phases for UNS S30400 stainless steel exposed to O2–N2 gases
at 1 atm and 871 C (1600 F).
100
80
(Cr, Fe)N
Cr2N
SiO2
60
FeO
(Fe, Cr)2O3
M3O4
40
Austenite (Ni, Fe)
MnSiO3
Mn2SiO4
20
0
0.000001 0.00001 0.0001 0.001 0.01 0.1 1
PO (atm)
2
Figure 30 Effect of O2 partial pressure upon the stable phases for UNS S41000 stainless steel exposed to O2–N2 gases
at 1 atm and 871 C (1600 F).
Increasing the amount of Ni, as found in the alloy Because alloy 214 (nickel-base UNS N07214)
UNS N07214 increases the amount of oxides, which has a high Ni content (76 wt%) and low Fe content
depend upon the presence of nickel, such as MeO (2.5 wt%), the results are very different. Figure 31
oxides, where Me can be Fe or Ni. Figure 31 illustrates shows the effect of the O2 partial pressure at 1 atm
this by showing stable NiO at O2 partial pressures above and 1600 F (871 C). At O2 partial pressures above
6 103 atm and spinel phases consisting of NiCr2O4 0.006 atm of O2, NiO is the dominant phase
and NiOAl2O3 at O2 partial pressures between (80 wt%), but this phase decreases dramatically at
3 103 and 6 103 atm, as shown in Figure 31. lower O2 partial pressures and disappears at 0.003 atm
100
80
CrN
Cr2N
60 AlN
NiO
Cr2O3
40 NiCr2O4
Austenite-Ni
(NiO) Al2O3 (s)
Al2O3 (s)
20
0
0.000001 0.00001 0.0001 0.001 0.01 0.1 1
PO2 (atm)
Figure 31 Effect of O2 partial pressure upon the stable phases for nickel-base alloy UNS N07214 exposed to O2–N2 gases
at 1 atm and 871 C (1600 F).
5E + 3 At 25 103 atm and above of O2, the formation of

Cr2O3 and Al2O3 occurs while NiCr2O4 and the
1E + 3
5E + 2 NiOAl2O3 (spinel structure) cannot form. At
103 atm of O2 and less, nitrides became stable.
1E + 2
Low levels of O2 and high levels of N2, such as
Penetration (mils)
5E + 1
410
found in 3 103 atm O2 and 0.997 atm N2 increase
1E + 1
5E + 0 the stability of the austenite phase (FCC), since nitro-
304 gen is well known as an austenite stabilizer. This
1E + 0
5E – 1 behavior was observed in the stainless steels consid-
214 ered in this study at 1600 F (871 C). The N2 partial
1E – 1
5E – 2 pressure of about 1 atm in the gas phase when O2
partial pressures are sufficiently low (104 to
3 103 atm), allows formation of nitrides – mostly
1E – 2
1.00E – 05 1.00E – 04 1.00E – 03 1.00E – 02 1.00E – 01 1.00E + 0 0
PO2 (atm) in the form of CrN and Cr2N with limited solubility
AISI 304 AISI 410 AISI 214 for Fe into the nitride phases. These compounds can
Figure 32 Predicted effect of PO2 upon oxidation be detected by bulk analysis of the corrosion product
corrosion of UNS S30400, UNS S41000, and UNS N07214 scale with X-ray diffraction.
at 871 C (1600 F) after 1 year of exposure time (obtained Figure 32 shows the predicted effect of O2 (in
by using ASSET predictions). The thickened portion of each O2–N2 gases) at 871 C (1600 F) after 1 year of expo-
line correspond to interpolations of actual data (corrosion
sure time obtained by using ASSET for UNS S30400,
rates) obtained in the range of pressure from 1 to 0.001 atm
for 304 and 410; and from 1 to 0.01 atm for 214. UNS S41000, and UNS N07214. Stainless steels UNS
S30400 and UNS S41000 have comparable penetra-
tions under different O2 partial pressures. Penetration
of O2. The reduction in the amount of NiO occurs of UNS S30400 is almost independent of O2 while
simultaneously with an increase in the amount of the penetration of UNS S41000 slightly decreases when
Ni-rich austenite phase. It is well known that the O2 concentration decreases. In the case of nickel-base
oxidation rates of nickel alloys such as UNS N07214 alloy UNS N07214, the penetration decreases with
depend strongly upon the O2 partial pressure in the increasing O2 concentration and internal oxidation is
gas. Low O2 partial pressures (about 0.01 atm, and less) the dominant oxidation mechanism. These predicted
favor subsurface oxidation and rapid total penetration. rates are based on the correlations of measured
S30400 N07214
100.0 100.0
Penetration (mm)
Penetration (mils)
10.0 10.0
1.0 1.0
0.1 0.1
0.001 0.010 0.100 1.000 0.001 0.010 0.100 1.000
(a) PO (atm) (b) PO (atm)
2 2
1400 F 1600 F 1800 F 1800 F 2000 F
Figure 33 Effect of O2 partial pressure upon metal penetration of (a) UNS S30400 and (b) UNS N07214 exposed to different
temperatures for 1000 h.
the amount of oxide present as Al2O3–Cr2O3–Fe2O3.

Figure 34 illustrates the different types of corrosion
expected over a range of N2–O2 gas composition.
Penetration
Internal oxidation competes with scaling as defined

Oxidation
by the PO2. At low O2 pressures, such as 104 atm of
O2, nitrides form and grow within the metal matrix.
Nitridation
Internal Scale 1.11.4 Oxidation/Carburization by
corrosion formation
CH4–H2O Gases
Log PO
2
1.11.4.1 Types of Equipment and
Figure 34 Schematic representation of the transition from Processes Where Carburization Occurs
oxidation to nitridation, for a fixed temperature, exposure
time, and alloy. Corrosion of metals and alloys used in equipment
processing high temperature, corrosive, gases contain-
ing carbon, hydrocarbons and gases like CO or CO2
oxidation data. The corrosion rates were predicted for have the potential to induce carburization. Examples
1.0 to 0.001 atm, as indicated in the bold lines in of some processes might be gas processing, combus-
Figure 32. tion gas process heaters, petroleum refineries (hydro-
As shown in Figure 33, the influence of O2 con- cracking, coking, vacuum flashing, hydrotreating, and
centration upon oxidation is complex over the tem- catalytic reforming) coal/coke/oil gasifying, petro-
perature range of 871–1177 C (1400–2000 F) for chemical production, gasification of black liquor in
alloys exposed to different O2 concentrations at a pulp/paper production, and fossil fuel-fired power
constant exposure time of 1000 h. Alloys S30400 and generation. Carburization may limit the maximum
S41000 show higher penetrations at higher PO2s and allowable temperature or maximum allowable gas spe-
lower PO2s, with a minimum at some intermediate cies concentrations for metals and alloys in equipment.
PO2. Microstructures show that penetration for these
alloys is mostly scale formation at high PO2s, while
internal oxidation dominates at low PO2s. The mini-
Mechanism
mum penetration at some intermediate PO2 is caused
by the transition between dominance by scale forma- Oxidation/carburization by CH4–H2O, CO, or CO2
tion and internal oxidation. gases is possible. The key variables influencing the
Al2O3-forming alloys such as N07214 exhibit kinetics are exposure time, PO2, activity of carbon
slower oxidation as the O2 concentration increased (aC), partial pressure of H2S, alloy composition, and
temperature. Alloys exposed to mixtures of C–CH4– 1.11.4.3 Corrosion Products Prediction

H2O can experience corrosion by oxidation and car-
ASSET was used to calculate stable corrosion pro-
burization. The transition between carburization and
ducts and alloy phases of several alloys exposed to
oxidation behaviors of commercial alloys will be
CH4–H2O gas mixtures at 1800 F (982 C). Three
discussed.
alloys are summarized: UNS S34700, UNS S44600,
In carburizing atmospheres, carbon is transferred
and UNS N06025. UNS S34700 is an Fe–Cr–Ni
from the environment into metal components. This
stainless steel stabilized with small additions of nio-
process is carburization, which is a high temperature
bium and tantalum. The alloy exhibits an excellent
corrosion phenomenon.57–66 This can change
resistance to intergranular corrosion and can be used
mechanical properties of metals and alloys mostly at
for prolonged service from 800 to 1500 F (427–
temperature >800 C when carbon diffuses into
816 C). UNS S44600 is an Fe–Cr ferritic heat-resist-
alloys and react with metals to form internal carbide
ing alloy with excellent resistance to oxidation and to
precipitates.
various forms of hot corrosion. The alloy is most
A result of carburization is an increase in hardness
commonly used for service between 1500 and
of the carburized materials. Low alloy steels are often
2200 F (815 and 1200 C). UNS N06025 is a high
exposed to carbonaceous environments to produce
carbon Ni–Cr–Fe alloy with additions of Al, Ti, Y,
hard surfaces to increase wear resistance, corrosion
and Zr. The alloy exhibits exceptional resistance to
resistance, and abrasion resistance. So, with the
oxidation at high temperatures and possesses very
proper heat treatment in a gas atmosphere, the con-
good high temperature carburization resistance.
sequence of carburization is beneficial.
This alloy can be used for pressure vessels with
Carburization is also a corrosion process that
service temperatures up to 1150 C.
affects high temperature alloys, especially in petro-
Corrosion products and alloy phases were calcu-
chemical industry processes. It is a common contri-
lated at 1800 F (982 C) under various CH4–H2O gas
bution to failure of ethylene pyrolysis tubes and can
mixtures by assuming that equilibrium conditions
degrade the corrosion resistance and mechanical
are satisfied, as shown in Figure 35 for alloy UNS
properties of metals and alloys. In some cases, coke
S34700, Figure 36 for alloy UNS S44600, and
deposition occurs at the inner walls of tubes during
Figures 37 and 38 for alloy UNS N06025. Each figure
hydrocarbon cracking. The decoking process is used
shows the stable corrosion products such as oxides
to remove coke deposits. Steam or steam–air mix-
and carbides, the alloy phases and their relative
tures are often used to remove coke. The removal
phase amounts.
processes oxidize the coke to CO and CO2 and by
The effect of gas composition on Me3O4 (spinel)
thermal shocking/spalling of the deposits. The
composition was investigated. The concentration of
decoking process also introduces oxidation as a cor-
metals (Fe, Cr, and Ni) in spinel were calculated, as
rosion concern.
shown in Figure 35 for UNS S34700, Figure 36 for
Carburization occurs when carbon enters the sur-
UNS S44600, and Figure 37 for UNS N06025. For
face layer of a metal or alloy. The dissolved carbon
these three alloys, the amount of spinel increases with
remains in solution or reacts with the metal to form
increasing H2O in gas mixture. The major metals in
carbides. Both mechanisms strengthen or harden the
the spinel are Fe and Cr. For UNS S34700 and UNS
surface of the metal. In most cases, carburization
S44600, the concentration of Fe in spinel increases
leads to precipitation of carbides like M23C6 and
with increasing H2O and the concentration of Cr in
M7C3 inside the alloy surface. The internal carbide
spinel (Me3O4) decreases with increasing H2O. For
formation causes a volume increase of the carburized
UNS S44600, which is a nickel-based alloy, the con-
zone and potential cracking which leads to deteriora-
centration of Ni in spinel (Me3O4) increases with
tion of mechanical properties of metals. The increased
increasing H2O.
carbon levels also lead to loss of ductility, which can
Carburization tests were performed on several
lead to various forms of fracture.
alloys at 1800 F (982 C) for 720 h under different
The carburization rate of an alloy is affected by
gas atmospheres. One group of samples was under
alloy composition, exposure gas composition, temper-
flowing pure CH4 gas, which decomposed to produce
ature, and time. This chapter discusses carburization
massive amounts of solid carbon. The other group was
in terms of the total sound metal loss (penetration),
under H2–1.2% CH4, which yields a carbon activity of
which is the sum of surface metal loss by scaling plus
one at the test temperature of 1800 F (982 C).
internal carburization.
80
Concentration of metals in Me3O4

70
60
50
( mole%) 40
30
20
10
Fe Cr Ni
0
50 60 70 80 90 100
Mole percent of H2O
Figure 35 Effect of gas composition upon Me3O4 composition in alloy UNS S34700 at 1800 F (982 C).
80
70
60
50
( mole%)
40
30
20
10
Fe Cr
0
50 60 70 80 90 100
Mole percent of H2O
Figure 36 Effect of gas composition upon Me3O4 composition in alloy UNS S44600 at 1800 F (982 C).
The carburization corrosion rates were measured and H2–1.2% CH4 gas mixture. Carbon activity would
shown in Figure 39 for samples under H2–1.2% CH4 have been one if the gas mixture had been at equilib-
gas mixtures and Figure 40 for samples under pure rium. Since available data for well-known alloys like
CH4. Also carburization rates for 1 year were calcu- UNS N08810, HP, and HK show increasing carbur-
lated based on the measured corrosion rates for 720 h ization at higher carbon activities, the carbon activity
and assuming parabolic kinetics for 1 year, as shown in must have been larger than one in the pure CH4 gas
Figures 39 and 40. The carburization results show atmosphere and the gas must not have been at equi-
that UNS S34700 has better carburization resistance librium at 1800 F (982 C).
than UNS S44600 and UNS N06025 in both testing
gas environments.
The carburization corrosion rates for alloys in 1.11.5 Summary
pure CH4 are much higher than those in H2–1.2%
CH4. Under pure CH4 gas atmosphere at 1800 F The various corrosion mechanisms are named for the
(982 C), tested alloys were heavily coated with corrosion products that form in high temperature gas
solid carbon on the surface and their carburization phase corrosion of metals and alloys, which occur in a
rates are much higher than the same samples in the variety of industrial processes. The thermochemistry
80
70

60
50
(mole%)
40
30
20
Fe Cr Ni
10
2O
0
50 60 70 80 90 100
Mole percent of H2O
Figure 37 Effect of gas composition upon Me3O4 composition in alloy UNS N06025 at 1800 F (982 C).
100
80
Phase amounts
60
Solid carbon is deposited
40
20
0
0 20 40 60 80 100
Mole percent of H2O
FCC alloy Cr3C2 Cr2O3

Me3O4 Al2O3
Figure 38 Equilibrium corrosion products of UNS N06025 under CH4–H2O gas mixture atmospheres at 1800 F (982 C).
of the interactions between the alloying elements and mixed gas corrosion have been discussed for several
the species in the gas phase determines which corro- classes of corrosion mechanisms. Any changes in gas
sion products form. In most instances, increasing the composition, which induce greater stability of slower
temperature increases the rate of corrosion. However, growing corrosion products, will tend to slow the rate
the effect of changes in the gas species concentrations of corrosion in mixed gases. For example, increasing
upon the rate of corrosion may have nonobvious the PO2 in a gas, which is causing sulfidation/oxida-
effects upon the rate of corrosion, particularly when tion, will tend to favor slower oxide growth at the
corrosion is producing a mixture of corrosion pro- expense of growth of faster growing sulfides, thereby,
ducts. Much of the effects of mixed gas corrosion can slowing the rate of corrosion. This chapter has
be thought of in terms of a competition between the reviewed the current knowledge of how thermo-
different corrosion products, which likely impact chemistry in terms of the PO2 and PS2 values of gases
corrosion behaviors in various ways. Effects of chang- can describe corrosion behaviors for sulfidation/oxi-
ing gas species concentrations upon corrosion for dation conditions and can even be used to select a test
300
Projection for one year
250
Corrosion rate (mpy)
After 720 hours
200
150
100
50
0
253MA 602CA MA956 446 353MA Ebrite- 347H
26
Figure 39 Carburization rates for alloys in H2–1.2% CH4 at 1800 F (982 C).
300
Projection for one year
250
After 720 h
200
150
100
50
0
253MA 602CA MA956 446 353MA Ebrite- 347H
26
Figure 40 Carburization rates for alloys in CH4 at 1800 F (982 C).
gas which is very different from the industrial gas in exposures, by summing the corrosion expected in the
question. Industrial corrosion in H2–H2O–H2S gases various sets of exposure conditions.
can be simulated in laboratory tests using CO–CO2– Corrosion can be well discussed and used in terms
COS, which are suggested because they are experi- of total metal penetration, which is the sum of metal
mentally more convenient. Test exposure times must thickness lost by scale formation plus the metal thick-
be long enough to achieve long-term corrosion ness affected by internal corrosion product forma-
behavior, as measured by the time needed for the tion. Alloys vary widely in their relative amounts of
parabolic rate constant to become independent of total metal penetration for various corrosion
time, which is at least 500 h at 600 C for a number mechanisms and exposure conditions, which is due
of alloys discussed. Once long-time (time indepen- to either internal penetration or scale formation.
dent) parabolic time dependence is achieved, then The corrosion behaviors of alloys vary from sur-
the extent of corrosion over thousands of hours can face to internal corrosion and were studied by asses-
be well correlated with a parabolic time dependence. sing the stable corrosion products and alloy phases as
Once the effects of gas composition, exposure time, functions of the ambient gas composition. There are
and temperature are well understood for a given alloy transitions in corrosion behavior as the gas exposure
and corrosion mechanism, it is also possible to predict changes, causing changes in the stable corrosion pro-
alloy corrosion for many hot-gas corrosion mechan- ducts. The metal loss information for alloys in several
isms for wide ranges of conditions using correlations, corrosive environments was reviewed and advice
even considering changing gaseous and temperature given on how to predict the probable corrosion
mechanism and amount of metal penetration as func- ‘‘Technical Information, Oxidation Resistance of Haynes
High Temperature Alloys’’.
tions of the important variables. A new capability was 7. Stoklosa, A.; Stringer, J. Oxid. Met. 1977, 11, 263.
developed and reviewed with the ability to predict 8. John, R. C.; Pelton, A. D.; Young, A. L.; Thompson, W. T.;
formation of the basic types of corrosion products Wright, I. G.; Besmann, T. M. Assessing corrosion in oil
refining and petrochemical processing, Presented at
and thickness losses by corrosion for wide ranges of Symposium on High Temperature Corrosion in Energy
conditions of gas composition, temperatures, and Related Systems, Angra dos Reis, Rio de Janeiro, 1–4
times for many alloys. September 2002.
9. John, R. C.; Fort, W. C., III; Tait, R. A. Mater. High Temp.
1993, 11, 124.
10. John, R. C.; Young, A. L.; Thompson, W. T. Corrosion/97;
Acknowledgments National Association of Corrosion Engineers: Houston, TX,
1997; Paper 142.
11. Lai, G. Y. In Corrosion/84; National Association of
Support by the following organizations are recognized Corrosion Engineers: Houston, TX, 1984; Paper 73. Data
and appreciated: US Department of Energy – Office of from brochures Haynes 188 and Pocket Guide to Haynes
Alloys from Cabot Corp, 1986.
Industrial Technologies via cooperative agreement 12. John, R. C. Second International Conference on Heat
DE-FC02-00CH11020, Shell Global Solutions (US) Resistant Materials, Gatlinburg, TN, 11–14 September
Inc.; Shell International E&P Inc, Arthur L. Young of 1995.
13. John, R. C. In Corrosion/96; National Association of
Humberside Solutions Ltd., Arthur D. Pelton of Centre Corrosion Engineers: Houston, TX, 1996; Paper 171.
for Research in Computational Thermochemistry in 14. Perkins, R. A. In High Temperature Corrosion; Rapp, R. A.,
Université de Montréal, William T. Thompson of Ed.; National Association of Corrosion Engineers:
Houston, TX, 1981; pp 345–353.
the Royal Military College of Canada, Oak Ridge 15. Perkins, R. A. Proceedings of the Third Annual Conference
National Laboratory, Foster Wheeler Development on Materials for Coal Conversion and Utilization.
Corporation, KEMA, Caterpillar, Special Metals Cor- U. S. Department of Energy, CONF-781018, October,
1978.
poration, Texaco, Haynes International, Usinor Indus- 16. Natesan, K.; Chopra, O. K. In Proceedings of the
teel, Kvaerner Pulping Oy, Materials Technology Conference on Properties of High-Temperature Alloys;
Institute of the Chemicals Process Industries (MTI), Foroulis, Z. A., Pettit, F. S., Eds.; The Electrochemical
Society: New Jersey, 1976; pp 493–510.
Dupont, Exxon/Mobil, Rolled Alloys, VDM Fluor 17. Natesan, K. In High Temperature Corrosion; Rapp, R. A.,
Daniel and Air Products. The experimental testing Ed.; National Association of Corrosion Engineers:
effort owes its deepest gratitude to Owen Kriegel and Houston, TX, 1981; pp 336–344.
18. Baxter, D. J.; Natesan, K. In High Temperature Corrosion
Paul F. Schmidt of Shell Global Solutions (US) Inc and in Energy Systems; Rothman, M. F., Ed.; TMS-AIME/MSD-
research staff at ORNL. Data contributions from ASM: Detroit, 1985, pp 237–252.
Haynes International, VDM and Special Metals Cor- 19. Bakker, W. T. Mixed oxidant corrosion in nonequilibrium
syngas at 540 C; Electric Power Research Institute: Palo
poration are acknowledged. Ian G. Wright of Oak Alto, CA, 1995.
Ridge National Laboratory is thanked for his many 20. John, R. C. In High Temperature Corrosion in Energy
contributuions. Technical contributions from Yali Li Systems; Rothman, M. F., Ed.; TMS-AIME/MSD-ASM:
Detroit, 1985, pp 501–514.
and Juan Gonzalez are acknowledged. 21. John, R. C. In High Temperature Materials/Corrosion –
Fundamental Aspects of High Temperature Corrosion II,
Shores, D. A., Yurek, G. J., Eds.; Electrochemical Society,
References 1987, pp 86–99.
22. Waldner, P.; Pelton, A. D. J. Phase Equilib. Diff. 2005, 26,
23–28.
1. Haycock, E. W. In High Temperature Metallic Corrosion of 23. Waldner, P.; Pelton, A. D. Thermodynamic modeling of the
Sulphur and Its Compounds; Foroulis, Z. A., Ed.; Ni–S system, in press.
Electrochemical Society: Princeton, NJ, 1970; p 110. 24. Waldner, P.; Pelton, A. D.; Thermodynamic modeling of
2. Haycock, E. W.; Sharp, W. H. Corrosion of ferrous alloys the Fe–Ni–S system, in press.
by H2S at high temperatures, Session from the 24th 25. John, R. C. In Life Prediction of Corrodible Structures;
Meeting of the American Petroleum Institute, New York NACE, ISBN 1-877914-60-6 1994; Vol. 1.
1959. 26. John, R. C. Meeting of Electrochemical Society, 4–9 May
3. Guzeit, J. In Process Industries Corrosion – The Theory 1986; paper no. 358.
and Practice; National Association of Corrosion 27. John, R. C. Fundamental Aspects of High Temperature
Engineers: Houston, TX, 1986. Corrosion – II, Simultaneous Sulphidation and Oxidation of
4. Backensto, E. B.; Drew, R. E.; Stapleford, C. C. In Alloy 800 in Mixed Gases; Electrochemical Society, 1987;
Corrosion; NACE, 1956; Vol. 12, p 22. p 86.
5. Sorell, G.; Hoyt, W. B. In Corrosion; NACE, 1956; Vol. 12, 28. Natesan, K. In High Temperature Corrosion; Rapp, R. A.,
p 213t. Ed.; NACE: Houston, TX, 1983; p 336.
6. Lai, G. Y. High-Temperature Corrosion of Engineering 29. Baxter, D. J.; Natesan, K. Oxid. Met. 1989, 31, 305.
Alloys; ASM International: Materials Park, OH, 1990; 30. Rhamel, A.; Schorr, M.; Valesco-Tellez, A.; Pelton, A. Oxid.
Literature from Haynes International Marketing, Met. 1987, 27, 199.
31. Perkins, R. A.; Morse, G.; Coons, W. C. Materials for Protection of Materials, Les Embiez, France, 20–24 May
syngas coolers, Report AP-2518, EPRI Research Report 2000.
1654-5, Palo Alto, CA, August 1982. 52. John, R. C.; Pelton, A. D.; Young, A. L.; Thompson, W. T.;
32. John, R. C.; Thompson, W. T.; Karakaya, I. In Corrosion/88 Wright, I. G. Predicting equipment lifetimes with high
1988; Paper 136. temperature corrosion data, Presented at the 6th
33. John, R. C.; Pelton, A. D.; Young, A. L.; Thompson, W. T.; International Symposium on High Temperature Corrosion
Wright, I. G. MS&T Conference, Cincinnati, OH, 15–19 and Protection of Materials, Les Embiez, France, 16–21
October 2006. May 2004; Proceedings of the Material Science Forum,
34. Kofstad, P. High temperature Corrosion; Elsevier Applied 2004; Vols. 461–464, 599–610.
Science, 1988. 53. Thompson, W. T.; Bale, C. W.; Pelton, A. D. Facility for the
35. Process Industries Corrosion; National Association of analysis of chemical thermodynamics (FACT), Ecole
Corrosion Engineers, 1975. Polytechnique, Royal Military College of Canada in
36. Norton, J. F.; Bakker, W. T. Eds. Materials for Coal Kingston, McGill University Montreal, Montreal, 1985.
Gasification Power Plant; Butterworth-Heinemann, 1993. 54. Christ, H. J.; Berchtold, L.; Sockel, H. G. Oxid. Met. 1986,
37. Bakker, W. T. Dapkunas, S.; Hill, V. Eds. Materials for Coal 26(1–2), 45–76.
Gasification; ASM International, 1988. 55. Pedraza, F.; Grosseau-Poussard, J. L.; Abrasonis, G.;
38. Embury, J. D. Ed. High Temperature Oxidation and Riviére, J. P.; Dinhut, J. F. J. Appl. Phys. 2003, 94(12),
Sulphidation Processes; Pergamon Press, 1990. 7509–7519.
39. Rapp, R. A. Ed. High Temperature Corrosion; National 56. Udyavar, M.; Young, D. J. Corros. Sci. 2000, 42, 961–883.
Association of Corrosion Engineers, 1983. 57. Grabke, H. J. Carburization: A High Temperature
40. Streiff, R. Stringer, J. Krutenat, R. C.; Caillet, M. Eds. High Corrosion Phenomenon; MTI, 1998.
Temperature Corrosion – Advanced Materials and 58. Kofstad, P. High Temperature Corrosion; Elsevier Applied
Coatings; Elsevier Science, 1989. Science, 1988.
41. Grabke, H. J. Mater. Corros. 2005, 56(6), 384–388. 59. Grabke, H. J.; Muller, E. M.; Speck, H. V.; Konczos, G.
42. Grabke, H. J.; Strauss, S.; Vogel, D. Mater. Corros. 2003, Steel Res. 1985, 56(5), 275–282.
54(11), 865–901. 60. Grabke, H. J.; Müller-Lorenz, E. M.; Schender, A. ISIJ Int.
43. Kothari, D. C.; et al. Mater. Sci. Eng. A 1989, 116, 89–95. 2001, 41, S1–S8.
44. Agudo, L.; Sanchez, C.; Nachez, L.; Feugeas, J. N.; 61. John, R. C.; Pelton, A. D.; Young, A. L.; Spencer, P.,
Staia, M. H. Revista de la Facultad de Ingenieria, Thompson, W. T. Alloy Carburization at Temperatures of
Universidad Central de Venezuela 2005, 20(1), 71–76. 1200–2100 F (650–1150 C); NACE – Corrosion, 2003,
45. Issartel, C.; et al. Corros. Sci. 2004, 46(9), 2191–2201. Paper 03474.
46. Issartel, C.; Buscail, H.; , Caudron, E.; et al. Appl. Surf. Sci. 62. Nava Paz, J. C.; Grabke, H. J. Oxid. Met. 1993, 39(5/6),
2004, 225, 14–20. 437–456.
47. Riffard, F.; Buscail, H.; Caudron, E.; Cueff, R.; Issartel, C.; 63. Ben-Haroe, I.; Rosen, A.; Hall, I. W. Mater. Sci. Technol.
Perrier, S. J. Phys. IV 2004, 118, 369–375. 1993, 9, 620–626.
48. Rawers, J. C. Proc. Inst. Mech. Eng. L J. Mater. Des. 64. Bullock, E.; Frampton, P. D.; Norton, J. F. Microstruct. Sci.
Applic. 2004, 218(L3), 239–246. 1981, 9, 215–224.
49. Han, S.; Young, D. J. Oxid. Met. 2001, 55(3–4), 223–242. 65. Klöwer, J. Mater. Corros. 1996, 47, 685–694.
50. Krupp, U.; Christ, H.-J. Oxid. Met. 1999, 52(3–4), 277–298. 66. Driver, D.; Barrand, P. J. Iron Steel Inst. 1964, 202,
51. John, R. C. In Proceedings of the 4th International 614–615.
Symposium on High Temperature Corrosion and
1.12 Carburization and Metal Dusting
D. J. Young
School of Materials Science and Engineering, University of New South Wales, UNSW Sydney, NSW 2033, Australia

1.12.2 Gaseous Carbon Activities 274
1.12.3 Carburization 274
1.12.4 Internal Carburization of Model Alloys 276
1.12.4.1 Reaction Morphologies and Thermodynamics 276
1.12.4.2 Carburization Kinetics 277
1.12.4.3 Carbide Microstructures and Distributions 280
1.12.5 Internal Carburization of Heat-Resisting Alloys 282
1.12.5.1 Effect of Carbon 283
1.12.5.2 Effect of Molybdenum 283
1.12.5.3 Effect of Silicon 283
1.12.5.4 Effect of Niobium and Reactive Elements 284
1.12.5.5 Effect of Aluminum 284
1.12.5.6 Alloying for Carburization Protection 284
1.12.6 Metal Dusting of Iron and Ferritic Alloys 285
1.12.6.1 Metal Dusting of Iron 285
1.12.6.2 Iron Dusting in the Absence of Cementite 288
1.12.6.3 Effects of Temperature and Gas Composition on Iron Dusting 288
1.12.6.4 Dusting of Low-Alloy Steels 290
1.12.6.5 Dusting of Ferritic Chromium Steels 291
1.12.6.6 Dusting of FeAl and FeCrAl Alloys 292
1.12.7 Dusting of Nickel and Austenitic Alloys 293
1.12.7.1 Metal Dusting of Nickel 293
1.12.7.2 Dusting of Nickel Alloys in the Absence of Oxide Scales 296
1.12.7.3 Effects of Temperature and Gas Composition on Nickel Dusting 297
1.12.7.4 Dusting of Austenitic Alloys 297
1.12.8 Protection by Oxide Scaling 298
1.12.8.1 Protection by Adsorbed Sulfur 300
1.12.8.2 Protection by Coatings 300
1.12.9 Conclusions 301
References 301
g i Activity coefficient of species i

Abbreviations
« Tortuosity factor, reflecting diffusional blocking
EDS Energy dispersive spectrometry (of X-rays)
by precipitate
FIB Focused ion beam
n Stoichiometric coefficient
ppm Parts per million (concentration)
ai Activity of species i
SAD Selected area diffraction
erf Error function
erfc Complementary error function
g Critical volume fraction necessary to form
external scale (of oxide or carbide)
Symbols kd Metal disintegration (dusting) rate constant
ðiÞ
a Ferrite; solubility kp Parabolic rate constant for internal carburisation
g Austenite rn Rate constant for reaction n
272
Carburization and Metal Dusting 273
t Time passing the gases through catalyst beds in externally

w/o Weight percent heated alloy tubes. Somewhat similar tubular reactors
x Position co-ordinate are used in the pyrolysis or ‘steam cracking’ process
A Surface area for making ethylene or propylene; for example,
Bi Mobility of species i
ðH2 OÞ þ C2 H6 ¼ C2 H4 þ H2 þ ðH2 OÞ ½5
Di Diffusion coefficient of species i
EA Activation energy for kinetic process In this case, the steam is added as a diluent in order
DGfo standard Gibbs free energy of formation to reduce the amount of solid carbon produced by
DHC Partial molar enthalpy of carbon dissolution gas-phase pyrolysis.
Kn Equilibrium constant for reaction n Feedstock materials, hydrocarbons, have been rising
Ksp Solubility product in price, and the need for improved process efficiencies
Ni Mol fraction of species i has led to higher operating temperatures. Simulta-
Ni(o) Original mol fraction of species i in solid prior to neously, tube wall thicknesses have been reduced to
diffusion improve heat transfer. Thus, process engineering
Ni(s) Mol fraction of species i at surface of alloy changes have led to higher tube metal temperatures
Q Activation energy for diffusion and reduced load-bearing sections. These increased
R General gas constant demands on material properties have been met by a
XS
DSC Partial molar excess entropy of carbon series of advances in alloy design. The tubes are cen-
dissolution trifugally cast, austenitic chromia formers. Their com-
T Temperature positions have evolved from the old HK 40 grade
Vi Molar volume of species i (Fe–25Cr–20Ni), through the HP grades (Fe–25Cr–
W Weight 35Ni) to high-nickel alloys containing 45% or even
X Scale thickness 60% Ni. The increased alloy levels have provided
Xi Width of internal precipitation zone significant improvements in creep properties, but the
materials are still subject to corrosion by carbon.
The process of coal gasification is of potential
1.12.1 Introduction importance in achieving high-efficiency electric power
generation. Steam is contacted with coal at high
Carbonaceous gases are generated by the combustion temperature, producing synthesis gas
of fossil fuels: coal, oil, and natural gas. The reactions
C þ H2 O ¼ CO þ H2 ½6
1
C þ O2 ¼ CO ½1 Although coal gasification has been employed for
2
1 a long time, its use on the scale and at the intensity
CO þ O2 ¼ CO2 ½2 required to support gas turbine-driven power gen-
2
CH4 þ 2O2 ¼ CO2 þ 2H2 O ½3 eration is relatively new. However, the carbon cor-
rosion mechanisms can be expected to be related
are often carried out using excess air, so that the to those observed in steam reforming and steam
exhaust gases contain free molecular oxygen plus cracking plants.
CO2 and H2O. Despite the use of excess air or oxy- Under combustion conditions, the oxygen potential
gen, locally reducing conditions can form and persist is usually high enough to guarantee oxide scale for-
in large-scale combustion units. Such conditions can mation. The questions of interest then concern inter-
lead to carbon attack, as will be discussed in the action of CO2 with the oxide. In contrast, synthesis gas
following sections. and other carbonaceous gases handled in the chemical
A number of other high temperature processes and petrochemical industries are characterized by high
are conducted under reducing conditions so as to aC values and low oxygen potentials. Typically, these
produce hydrogen or CO, rather than their oxidation gases are oxidizing to chromium and aluminum, but
products. Steam reforming of natural gas not to iron or nickel. The questions of interest, there-
CH4 þ H2 O ¼ CO þ 3H2 ½4 fore, concern the ability of Cr2O3 and Al2O3 scales to
exclude carbon from the alloy, and the consequences
at temperatures around 800–900 C is widely used to of oxide scale failure when carbonaceous gas species
produce hydrogen. This process is carried out by contact the metal surface.
We consider first the simple case where the oxy- As seen in Figure 1, temperature effects are very
gen potential is so low that no oxide forms, and different for these carbon-producing reactions. Thus,
corrosion by carbon alone results. If the carbon activ- methane, and hydrocarbons in general, can produce
ity, aC, is less than 1, but large enough to stabilize significant carbon activities only at high temperatures.
a metal carbide, the reaction is described as carbur- On the other hand, the synthesis gas and Boudouard
ization. Such reactions are rapid and destructive. reactions produce increasing carbon activities as the
However, if the gas is supersaturated with respect to temperature is lowered.
carbon (aC > 1), an even greater threat emerges. It is because these gases are difficult to equilibrate
Catalysis of carbon deposition by the metal can lead that they can become supersaturated with respect to
to its disruption and fragmentation in an extremely carbon; that is, aC > 1. An example is synthesis gas
rapid corrosion process known as metal dusting. produced in a steam reforming unit, which will nor-
Dusting mechanisms are different for ferritic and mally have aC < 0.5. As the gas is cooled from reac-
austenitic alloys, and they are discussed separately. tion temperature, however, the rate of reaction [7] is
Alloy design strategies for slowing carburization and so slow that the gas fails to adjust its composition
metal dusting are shown to be of limited value. How- by depositing carbon. The gas-phase carbon activity
ever, alloying to achieve protective oxide formation calculated from
can delay carbon access to the metal. This approach is K7 pH2 pCO
discussed together with other surface treatments in aC ½7 ¼ ½10
pH2 O
the final section of this chapter.
can then be much greater than 1. It is recognized that
the same gas also produces carbon via reaction [8],
and one can calculate
1.12.2 Gaseous Carbon Activities 2
K8 pCO
aC ½8 ¼ ½11
The common gas-phase processes producing carbon pCO2
are the synthesis gas reaction In general, aC ½7 6¼ aC ½8, because the gas is not at
equilibrium. In this situation, it is appropri-
CO þ H2 ¼ H2 O þ C ½7 ate to consider reactions [7]–[9] as separate,
the Boudouard reaction independent processes, as will be demonstrated in
Sections 1.12.6 and 1.12.7. Nonetheless, because
2CO ¼ CO2 þ C ½8 reaction [7] is usually faster than [8], it is common
to calculate aC from eqn. [1].
and hydrocarbon cracking, for example, Carburization experiments require aC 1. The
use of CH4/H2 gas mixtures to control carbon activ-
CH4 ¼ 2H2 þ C ½9
ity is inadvisable at temperatures below 1000 C,
The standard free energies of these reactions are because of the slow rate of reaction [9] and the
listed in Table 1. It is necessary to recognize that usually brief residence time of a laboratory reactor.
all three reactions are very slow as homogenous gas- It is preferable to use mixtures of H2 and C3H6, as the
phase processes. They will not reach equilibrium in a latter pyrolyzes readily.
typical laboratory reactor, unless catalyzed. Although
many materials of practical interest – iron, nickel,
cobalt, and their alloys – are catalytically active to 1.12.3 Carburization
these reactions, their oxide scales are inert.
Carbides are much less stable than oxides, as seen
from the examples in Table 2. Thermodynamic
Table 1 Gas phase equilibria relevant to carburizing
data for other carbides can be found in a review by

Reaction DGf ¼ A þ BTðJÞ Shatynski.1 Of the common alloy base metals, nickel
A B and cobalt do not form carbides under the conditions
of interest. Iron forms cementite, Fe3C, at tempera-
CO + H2 = H2O + C 134 515 142.37 tures below 763 C only if aC > 1. Exposure of these
2CO = CO2 + C 170 700 174.5
metals to reducing carbonaceous gases leads to dis-
CH4 = 2H2 + C +87 399 108.74
solution of carbon at the metal surface and its
H2 + CO → C + H2O
3
2
log10 (Kp)
1 2CO → C + CO2
CH4 → C + 2H2
–1
–2
300 400 500 600 700 800 900 1000
T (8C)
Figure 1 Equilibrium constants for gas-phase carbon producing reactions.
Table 2 Properties of metal carbides Table 3 Partial molar enthalpy and entropy for carbon

dissolution in metal

Carbide DGf ¼ A þ BTðJÞ VMCy* (cm )3
MP ( C)
Metal C (kJ mol1)
DG DSxs 1 1
c (J mol K ) Ref
A B
Ni 54 5 2
Cr23C6 411 200 38.7 7.91 1580 g-Fe 44.04 17.62 3
Cr7C3 174 509 25.5 8.26 1665
Cr3C2 84 353 11.53 8.98 1895
NbC 130 122 +1.67 13.47 3480
SiC 113 386 +75.7 13.70 2700
Al4C3 266 520 +96.2 15.24 1400 ðoÞ
NC NC x
Fe3C 29 037 28.0 8.31 1650
ðsÞ ðoÞ
¼ erfc pffiffiffiffiffiffiffiffi ½13
NC NC 2 DC t
*Volume per mole of metal
ðoÞ
Here, NC represents the original carbon level in the
metal prior to carburization. The rate at which the
diffusion inwards, but no carbide phase is formed if
carburization zone widens is given approximately by
aC < 1. If the metal surface is at equilibrium with the
gas phase, the surface concentration of dissolved car-
ðsÞ
Xi2 ¼ 4DC t ½14
bon NC (mole fraction), can be found from the rela-
tionship As DC values in austenite are high, the process is
fast. Using data for DC,4, 5 it is found that carbon
ðsÞ
NC ¼ KaC ½12 penetrates about 3 mm into each of g-Fe and Ni in
24 h at 1000 C.
Data for carbon dissolution in iron and nickel is The heat-resisting alloys used in contact with
summarized in Table 3, where the carbon solubility carbon-rich gases are usually chromia formers. As
in nickel is seen to be much lower than in g-Fe. If seen in Table 2, chromium also forms reasonably
inward carbon diffusion causes no phase change in stable carbides. It is commonly observed6–8 that
the solvent metal, and if furthermore DC is indepen- exposure of these alloys to gas compositions such
dent of composition, then the resulting carbon con- that no chromia scale can form leads to internal
centration profile is found by solving Fick’s second chromium carbide precipitation, rather than external
law to obtain scale formation. The conditions under which this
reaction morphology develops are now examined The conditions necessary for carbide precipitation
using model Fe–Cr, Ni–Cr, and Fe–Ni–Cr alloys, are now examined more closely. In the case of
for which the necessary data is available. Cr23C6, we can write
ðMÞ þ 23Cr þ 6C ¼ Cr23 C6 ðþMÞ; DGP ½16
1.12.4 Internal Carburization of the free energy change for which can be calculated
Model Alloys from data for carbide formation

1.12.4.1 Reaction Morphologies and 23Cr þ 6C ¼ Cr23 C6 ; DGf ½17
Thermodynamics
and alloy component dissolution
Chromium carbides are the expected reaction pro-
ducts and their formation within the alloy is the Cr
Cr ¼ Cr; DG ½18
outcome of competition between rival processes. C
C ¼ C; DG ½19
Scale formation is favored by rapid diffusion of chro-
mium from the alloy to its surface, whereas internal in the solvent metal, M. Thus,
precipitation is favored by rapid carbon ingress.
Wagner’s analysis9 of this situation allows calculation Cr 6DG
DGP ¼ DGf 23DG

C ½20
ðoÞ
of the minimum value of NCr at which scaling is and we evaluate the carbide solubility product
favored over internal precipitation
!1=2

Cr
NC ¼ Ksp ¼ exp½ðDGf 23DH
23 6
NCr
ðsÞ
ðoÞ p V A NC D C ½21
NCr ¼ gCrCv ½15 C Þ=RT
6DH
2v VCrCv DCr
i being the partial molar heat of dissolution.
with H
where g is the critical volume fraction necessary to A similar treatment for Cr7C3 leads to the result
form a continuous carbide layer, VA and VCrCv are the
molar volumes of alloy and carbide, DC and DCr are 7
NCr Cr
NC3 ¼ Ksp ¼ exp½ðDGf 7DH

the diffusion coefficients in the alloy of the indicated ½22

Cr Þ=RT
6DH
solutes, and v is the stoichiometric constant for the
carbide CrCv . Choosing Nimonic 75 (approximately
Standard values (Table 2) for DGf , carbon solu-
Ni–20Cr) as a basis for calculation, we can specify
bility data for g-Fe and Ni (Table 3), and activity
VA ¼ 6.58 cm3 mol1. Values for carbon permeability
ðsÞ coefficient data for Fe–Cr10 and Ni–Cr11 allow cal-
ðNC DC Þ and DCr listed in Table 4, together with
culation of carbide solubility product values shown in
VCrCv from Table 2 and the supposition g ¼ 0:3, lead
ðoÞ Tables 5 and 6.
to estimates of NCr ¼ 15, 20, and 37 required to form
Carbon solute levels in g-Fe and Ni in equilibrium
scales of Cr3C2, Cr7C3, and Cr23C6, respectively, at
with aC ¼ 1 are also shown in the tables, along with
1000 C. Of course, values of N > 1 lack physical
the corresponding minimum chromium concentra-
significance and result from inaccuracies in the data.
tions necessary to stabilize each carbide. Iron base
The conclusion is simply that Ni–Cr alloys are
alloys are predicted on this basis to be more suscep-
unlikely to form carbide scales exclusively, because
tible to internal carbide precipitation. This predic-
the inward carbon flux is so high and the molar
tion is tested by comparing the calculated minimum
volume of chromium carbide is small.
NCr values required for carbide precipitation with the
Table 4 Permeability data for carburization

ðsÞ
Table 5 Chromium carbide precipitation in Fe–Cr alloys
Alloy T ( C) NC DC DCr
Alloy 900 C 1000 C 1100 C
g-Fe 900 4.3 109 4.4 1013
ðsÞ
1000 1.4 108 3.7 1012 NC ðg-FeÞ 0.057 0.066 0.098
1100 5.5 108 2.3 1011 Ksp ðCr23 C6 Þ 1 1029 3.6 1027 2.6 1024
Ni 900 7.9 1010 8.0 1013 NCr ðminÞ 0.12 0.14 0.17
1000 3.6 109 7.2 1012 Ksp ðCr7 C3 Þ 3 1016 3.8 1015 3.4 1014
1100 1.4 108 4.7 1011 NCr ðminÞ 0.02 0.03 0.03
Table 6 Chromium carbide precipitation in Ni–Cr alloys before entering the single-phase M3C zone, in agree-

ment with experimental observation (Table 7). An
Alloy 900 C 1000 C 1100 C ðoÞ
alloy with NCr ¼ 0:18 is seen to develop a carbon
ðsÞ
NC 0.007 0.011 0.016 diffusion path which crosses successive two-phase
Ksp ðCr23 C6 Þ 9.9 1026 8.4 1024 3.7 1022 regions g þ M23C6 and g þ M7C3 before entering
NCr ðminÞ 0.29 0.32 0.34
the single-phase M7C3 field. Again, this corresponds
Ksp ðCr7 C3 Þ 9.8 1014 9.4 1013 6.5 1012
NCr ðminÞ 0.10 0.13 0.17 with the experimental observation (Table 7) of two
internal precipitation zones, with M23C6 forming in
the inner (lower aC) zone.
As iron solubility in the carbides increases with aC,
Table 7 Internal carbides found in Ni–Cr and Fe–Cr at
1000 C at ambient aC = 1 the Fe/Cr ratio in the precipitates is predicted to
decrease with increasing depth within the precipita-
ðoÞ
Alloy NCr Surface Internal carbides Ref. tion zone. Microanalysis in a transmission electron
carbides microscope of carbides precipitated within an Fe–
Ni–Cr 0.11 None None 12 Ni–Cr alloy16 revealed the partitioning of chromium
0.22 Cr3C2 Cr7C3 12 between precipitate and matrix in the carburized
0.33 Cr7C3 13 alloy. As seen in Figure 4, the results are in reason-
Fe–Cr 0.05 M3C M7C3a 14
able agreement with measured equilibrium values,
0.08 M3C M7C3 14
0.11 M3C M7C3 14 and again it is concluded that local equilibrium is
0.18 M7C3 M7C3 + M23C6 14 achieved throughout the precipitation zone. Two
0.26 M7C3 M7C3 + M23C6 14 more important inferences can be drawn from the
a
phase diagram of Figure 3. Carburization is pre-
M: chromium-rich (Cr + Fe).
dicted to transform the alloy matrix of a high-
chromium Fe–Cr alloy from ferrite to austenite as a
result of chromium depletion and carbon saturation.
experimental results for 1000 C summarized in As shown in Figure 2, this transformation is observed
ðoÞ
Table 7. The appearance of the carbide precipitation at the precipitation front. Secondly, if NCr is less than
zones in Fe–Cr alloys is illustrated in Figure 2. As 0.4, then (Fe,Cr)3C is predicted to form at or near
predicted, Fe–Cr alloys of high chromium content the alloy surface if the gas phase aC value is high
formed both carbides, whereas Ni–Cr formed only enough. This is important in metal dusting reactions
Cr7C3. Furthermore, the prediction that no carbide (Section 1.12.6), but can be ignored when studying
ðoÞ
should form in Ni–Cr with NCr < 0:13 is borne out. carburization reactions at aC 1. Furthermore, in
The success of this simple thermodynamic treat- austenitic alloys the nickel content destabilizes
ment indicates that local equilibrium is attained, and Fe3C, and the phase is not observed.
a steady-state diffusion description should be appli-
cable. However, whilst the assumption that the chro-
mium carbides are pure phases – the basis for eqns. 1.12.4.2 Carburization Kinetics
[21] and [22] – is reasonable for the Ni–Cr–C sys-
Internal carburization is a particular form of
tem, it is a poor approximation for Fe–Cr–C. As seen
internal oxidation, and its kinetics can therefore6–8,12
in the phase diagram of Figure 3, iron solubilities
be described using Wagner’s theory.9 Because
in the carbides are high, and cannot be neglected. ðsÞ ðoÞ
NC DC NCr DCr , the rate at which the carbide pre-
It is possible to calculate Ksp values for the mixed
cipitation zone deepens is given by
carbides (Cr,Fe)23C6 and (Cr,Fe)7C3, but a simpler
approach is to construct diffusion paths representing Xi2 ¼ 2kpðiÞ t ½23
the locus of phase constitutions along lines through
ðsÞ
the reaction zone. As DC DCr, these paths are con- ENC DC
kpðiÞ ¼ ðoÞ
½24
structed on the basis that only carbon diffuses and vNM
hence the NCr/NFe ratio remains unchanged within
ðoÞ
the reaction zone. where NM is the original alloy concentration of
ðoÞ
The diffusion path in Figure 3 for NCr ¼ 0:08 is metal M which forms carbide MCv. Thus, carburization
seen to cross the g þ M7C3 two-phase region, cor- rates are predicted to vary inversely with concentration
responding to internal precipitation of this carbide, of reactive solute metal. Data for a series of Fe–Cr
100 μm 15 μm
(a) (b)
Figure 2 Internal carburization of Fe–Cr at 1000 C (a) Fe–7.5Cr forms M7C3 precipitates and (b) Fe–17Cr forms innermost
zone of M23C6 precipitates.14
C 1
0.8
NCr /(NCr + NFe ) carbide
0.6
0.5 0.5 0.4

M7C3 + Gr
M3C + Gr M3C + M7C3 Cr3C2
0.2
␥ + Gr Cr7C3
Fe3C
+ ␥ Cr23C6
0
M 3C + ␥ M 7C
3
0 0.1 0.2 0.3 0.4
M23C6 + ␥ NCr /(NCr + NFe ) matrix
M23C6 + ␣
Fe Cr Figure 4 Partitioning of Cr between precipitates and
0 σ+γ 0.5 1
matrix in carburized alloy16 (filled symbols) and in
Atom fraction
equilibrium studies15 (open symbols). Circles for M7C3 and
Figure 3 Isothermal section at 1000 C of the Fe–Cr–C squares for M23C6.
phase diagram.15 Dotted lines show carburization diffusion
paths for DC DCr.
concentration of chromium in the matrix being close
to zero. However, as will be seen in Section 1.12.4.3,
alloys14,17 is shown in Figure 5. The kinetics are seen to the effect on carburization rates is small.
ðoÞ
be parabolic, and plots of kpðiÞ against 1=NCr are linear A further prediction of eqn. [24] is that carburiza-
ðoÞ
except at high NCr values. The slopes of these lines tion rates are determined by the permeability of the
were used together with v ¼ 0:71 (for (Cr0.6Fe0.4)7C3 metal matrix, regardless of the identity of the pre-
formed by low-chromium alloys) and the assumption cipitating carbide, providing that changes in the stoi-
E ¼ 1 to calculate carbon permeabilities. Comparison chiometric coefficient v are taken into account.
ðsÞ
in Table 8 with values found from NC 3 and DC 4 Permeability values calculated for nickel by Allen
measurements shows good agreement, demonstrating and Douglass12 from their carburization measure-
the utility of eqn. [24] in describing carburization ments of Ni–V, Ni–Cr, and Ni–Nb alloys are seen
rates. This is at first sight surprising, as eqn. [24] is in Table 9 to be in approximate agreement with each
based on the assumption that Ksp 1, and the other and with values found from independently
ðsÞ
45 Table 8 Carbon permeabilities NC DC (cm2 s1) in
Fe–Cr
40 Fe-5%Cr
900 C 1000 C 1100 C
35
Carburization 6.6 109 2.5 108 6.2 108
Depth2 (cm2) ⫻ 10−4
30 kinetics
ðsÞ
From NC and DC 4.3 109 1.4 108 5.5 108
25 Fe-7.5%Cr
20
ðsÞ
15 Table 9 Carbon permeabilities ðNC DC Þ (1010 cm2 s1)
Fe-10%Cr deduced from carburization kinetics of Ni–base alloys12
10
Alloy T ( C)
5 700 800 900 1000
0 Ni–12V 0.40 2.2 11 44
0 50 100 150 200 250 300
Ni–20Cr 0.21 0.8 8 55
(a) Time (min)
Ni–3Nb 0.30 6 14
ðsÞ
From NC and DC 0.19 1.6 7 36
6
1100⬚C
stability of the carbides, and the simple description

107kp (cm2 s−1)
4 of eqn. [26] cannot be expected to apply. In the case

of Ni–Cr alloys, however, nickel dissolves to only
a small extent in the carbides, and Cr7C3 is the only
1000⬚C stable internal carbide over a wide range of tempera-
2 ð0Þ
ture for NCr
0:2. The value of EA ¼ 190 kJ mol1
measured by Allen and Douglass12 for Ni–20Cr
900⬚C agrees with the prediction of eqn. [26] based on
DH C ¼ 54 kJ mol1 (2) and Q ¼ 138 kJ mol1 (5).
0 5 10 15 20 Chromia-forming alloys are usually based on
(b) 8()
1/NCr Fe–Ni, and the applicability of eqn. [24] to Fe–Ni–
Cr model alloys is now tested. Carburization of 25Cr
Figure 5 Carburization of Fe–Cr alloys (a) representative
kinetics at 1000 C and (b) effect of alloy chromium content alloys at 1000 C and aC ¼ 1 produces two reaction
on carburization rate.14,17 zones. Diffraction studies identified the near-surface
zone precipitates as M7C3, and those in the inner zone
as M23C6. Carburization kinetics of a series of Fe–
Ni–Cr alloys have been found18 to be parabolic at
ðsÞ
measured values of NC 2 and DC .5 It is concluded that 1000 C, and the rate constants are seen in Figure 6
internal carburization of both Fe and Ni base alloys is to vary considerably with alloy nickel content. If nearly
controlled by interstitial diffusion of carbon through all the chromium is precipitated as carbide, then the
the depleted metal matrix lattice. The tempera- reaction is sustained by carbon dissolution in and
ture effect on the rate is described by the empirical diffusion through the remaining Fe–Ni matrix. Ignor-
equation: ing the dissolution of some of the iron into carbide, we
approximate the matrix as having the same NNi/NFe
kpðiÞ ¼ koðiÞ expðEA =RT Þ ½25 ratio as the parent alloy. On this basis, one can use
ðsÞ
values of NC measured by Wada et al.2 for Fe–Ni
Logarithmic differentiation of eqn. [24] yields alloys and for DC measured by Bose and Grabke19 to
C þ Q predict carburization rates from eqn. [24]. Measured
EA ¼ DH ½26
and predicted rates were found18 to agree very well.
where Q is the activation energy of diffusion. In the It is therefore concluded that the Wagner theory pro-
case of Fe–Cr alloys, the extent of iron dissolution in vides a satisfactory basis for describing the car-
the carbides varies with temperature as does the burization of alloy compositions close to those of
7.5
6.0
109kp(i) (cm2 s−1)
4.5
10 mm
3.0 Figure 7 Cuboidal and lath-shaped M23C6 precipitates at

the reaction front in Fe–20Ni–25Cr carburized at 1000 C.
1.5
alloys.18,20 A distinctive common feature in these
cases is the formation of a grain boundary at the
0 carbide precipitation front. In the case of ferritic
0 0.1 0.2 0.3 0.4 0.5 0.6
NNi alloys (Figure 2(b)), the boundary corresponds to
the phase transformation
Figure 6 Variation of carburization rate at 1000 C with Ni
content of Fe–Ni–25Cr alloys.18
a-ðFe; CrÞ þ xC ! g-ðFe;CÞ þ Cr23 C6 ½28
commercial heat-resisting alloys. Before going on
In high-nickel alloys, lamellar carbides develop in
to consider practical alloys, it is worth considering
colonies in the alloy interior. It seems likely that they
the microstructures and distributions of carbide
nucleated at alloy grain boundaries and then grew
precipitates.
into the adjacent grain in a discontinuous precipita-
tion process which is now described.
1.12.4.3 Carbide Microstructures and A clear view of the boundary at the carbide pre-
Distributions cipitation front is shown in Figure 8 for an Fe–20Ni–
25Cr alloy first subjected to a brief period of internal
Particles of M7C3 precipitated in austenite are glob-
nitridation to form a boundary just beneath the sur-
ular, and develop no rational orientation relationship
face. Subsequent carburization led to the pictured
with the matrix. In contrast, M23C6 possesses a cubic
microstructure: the dark grain on the right is
structure and develops a strong cube-in-cube orien-
unreacted austenite, the light grain on the left is
tation relationship with the fcc g-matrix20
chromium-depleted matrix, and the precipitates are
M23C6. The selected area diffraction pattern shows
½001g ½001M23 C6 : ð100Þg ð100ÞM23 C6 ½27
the same cube-in-cube orientation relationship [27]
Usually, the M23C6 precipitates are small, cuboidal, between precipitate and matrix. The crystallographi-
or needle-shaped particles (e.g., Figure 7). The small cally oriented sides of the precipitates were always
precipitate size reflects the fact that they grew for the close-packed (111) planes. No rational orienta-
only a short time: a result of the continued nucleation tion relationships were found between unreacted
of new particles as the reaction front advanced into austenite and either reacted austenite or carbide.
the alloys. This in turn was due to rapid carbon Microanalysis results in Figure 8 show a step func-
diffusion, which quickly produced sufficient super- tion change in NCr at the austenite/depleted austen-
saturation to favor homogeneously distributed nucle- ite grain boundary, but no sign of lateral diffusion
ation. In other circumstances, the same carbide can within the matrix. The mechanism is that of discon-
form elongated lamellar or lath shaped precipitates tinuous precipitation20,25
oriented parallel to the reaction direction (Figure 7). g þ C ¼ gD þ Cr23 C6 ½29
As the value of kpðiÞ is 30–50% higher when the
aligned microstructure is adopted, the reasons for where gD denotes depleted austenite. The change in
its development are of interest. crystallographic orientation from parent g to prod-
Lamellar or cellular M23C6 microstructures have uct gD is obvious in Figure 8. This reorientation
been reported to develop in ferritic alloys,14,16,21,22 resulted from the free energy reduction achieved
in high-nickel austenitics,18 and in a variety of heat- when the austenite formed coherent interfaces with
resisting alloys carburized at low temperatures.23,24 It the precipitates which grew approximately unidirec-
is also observed in previously nitrided Fe–Ni–Cr tionally, parallel to the carbon diffusion direction.
1200
Reacted matrix Parent alloy
1000 Fe
Relative intensity
800
600
400 Ni
200
Cr
0
0.2 0.3 0.4 0.5
Distance (um)
1 um
Carbide Product Carbide
Relative intensity
Fe
Cr
Ni
0 0.2 0.4 0.6 0.8 1.0 1.2

Distance (um)
Figure 8 M23C6 precipitation front in Fe–20Ni–25Cr after brief prenitridation and subsequent carburization at 1000 C,
aC ¼ 1.20,25
The incoherency of the g/gD interface is evident in precipitation zone increases, aC and NC decrease caus-
its curvature, and this provided rapid chromium dif- ing NCr to increase according to the solubility product
fusion towards the advancing carbide precipitate tips, equilibria [21] and [22]. As a result, the amount of
sustaining their growth and producing the discontin- chromium precipitated is less. The qualitative effect
uous change in NCr seen at the interface. on carburization rate can be seen from eqn. [24]:
ðoÞ
It is concluded that lamellar carbide precipitates because the effective value of NCr is lowered, the
develop when a grain boundary is present. The penetration rate is faster. This effect has been ana-
boundary provides more rapid chromium diffusion lyzed26,27 for the general case of low-stability precipi-
to the precipitates, favoring their continued growth tates. Deviation from the Wagner assumption of
rather than nucleation of new ones. As seen in vanishingly small Ksp values is expressed via a solubil-
Figure 5(b), carburization rates for Fe–Cr alloys ity parameter
with high NCr values are higher than predicted by
ðsÞ ðoÞ
eqn. [24]. These are the alloys which form lamellar a ¼ 1 NCr =NCr ½30
precipitates, and the acceleration is attributed to
ðsÞ
boundary diffusion of carbon along the multiple car- where NCr is the matrix equilibrium chromium con-
ðsÞ
bide–austenite interfaces. As NC DC in austenite is in centration at the surface of the reacted alloy. If a ¼ 1,
any case large, the increase in rate is relatively small: Ksp ¼ 0 and the Wagner model applies; if a ¼ 0,
30% faster in Fe–Cr and 30–50% faster in Fe–Ni–Cr no precipitation occurs and an error function solu-
alloys at 1000 C. tion describes the carbon profile. Considering Fe–Cr
Carbide precipitate distributions are nonuniform, alloys, we find from eqn. [22] and the data of Table 5
ðsÞ
because Ksp is not small. Thus, as depth within the that NCr ¼ 0:028. Taking a representative value of
ðoÞ
NCr ¼ 0:25, then a value of a ¼ 0.9 is arrived at. high-nickel alloys and M7C3 on low-nickel materials.
For this value, Ohriner and Morall26 calculate that The difference arises from the changed Fe/Cr
kpðiÞ is increased by a factor of 2 above that predicted ratios.35 Thus, reaction morphologies are consistent
from eqn. [24]; that is, the penetration depth is in- with local equilibrium. Moreover, as the internal car-
creased by 40%. However, uncertainties in measured burization kinetics are parabolic, it is clear that the
ðsÞ
values of DC, NC , and kpðiÞ total at least this amount. process is diffusion controlled.
A similar conclusion is reached for Cr23C6 precipita- Despite the complexity of these alloys, their rela-
tion. It is, therefore, concluded that despite the ap- tive performance under nonoxidizing conditions can
proximate nature of its assumptions, the simple result be understood in terms of Wagner’s equation [24].
eqn. [24] provides good order of magnitude predic- The procedure is to model the heat-resisting alloys
tion for model alloy carburization. A more precise as Fe–Ni–Cr ternaries, and approximate the carbur-
prediction can be made using the numerical treat- ization zones as chromium carbide precipitates in
ment of the Ohriner–Morall theory developed by an Fe–Ni matrix. Carburization rates are then pre-
Christ.28 dicted from eqn. [24] to change with carbon perme-
ðsÞ
ability, NC DC . This permeability shows a minimum
at Ni=Fe
4 : 1,2,10 which was shown by Rahmel
et al.47 to be reflected in relative penetration depths
1.12.5 Internal Carburization of
of a range of alloys. Clearly, the Fe/Ni ratio has a
Heat-Resisting Alloys
significant effect on carburization rate. However, this
approach neglects the effect of NM, and a more
Many laboratory studies of the carburization of heat-
detailed calculation is required.
resisting alloys have been reported. 6–8,28–45 These are
The quantity NM represents mainly chromium. In
in general agreement with the findings from industrial
calculating its value, the alloy chromium content
exposures46 that carburization rates vary with Fe/Ni
must, therefore, be corrected for the amount already
ratio and decrease with increased levels of Cr, Nb, Si,
removed from the matrix into interdendritic carbide
and sometimes W and Ti.
during casting. This is done on the assumption that
The usual appearance of a cast heat-resistant alloy
all of the alloy carbon was precipitated as pure
after carburization is shown in Figure 9. Murakami’s
Cr23C6. Added contributions to NM are calculated
etch reveals finely precipitated, cuboidal M23C6 par-
on the basis of NbC and other MC carbide formation,
ticles in the inner zone and coarser, spherical M7C3
as well as the molybdenum carbides (Mo2Fe)C and
particles near the surface. The original cast alloy
(CrMoFe)C. As seen earlier, the application of eqn.
structure is seen to the right: austenite dendrites
[24] is nonetheless an approximation because M7C3
with primary interdendritic M23C6. Near the carbur-
and M23C6 carbides contain substantial levels of
ization front, preferential precipitation is seen along
iron.15 Consequently, the value of NM calculated as
dendrite boundaries. Some fragmentary external
described above is an underestimate. On the other
carbide scale is also seen. This is usually Cr3C2 on
hand, however, an overestimate of NM results from
the error in the mass balance underlying eqn. [24].
This latter error arises because the solubility pro-
ducts of Cr7C3 and Cr23C6 are large, and significant
levels of chromium remain unreacted in the depleted
matrix.
Carburization leads to approximately equimolar
amounts of M7C3 and M23C6; hence, a value of
v ¼ 0:345 is used in eqn. [24]. No value for e is
available. Predicted kp values based on E ¼ 1 are
compared in Table 10 with measured40 quantities
for a selection of alloys. It is seen that close order of
magnitude agreement is achieved for the 30, 35, and
50 µm
45 Ni grades, but not for the 60Ni grades. The
latter contain aluminum, and will be discussed
Figure 9 Cast alloy Fe–45Ni–19Cr after carburization for below. We consider first the effects of other alloy
24 h at 900 C. components.
Table 10 Carburization rate constants (kp) (107 cm2 s1) for heat-resisting alloys
Cast Alloy 1100 C 1000 C 900 C
Measured Calc Measured Calc Measured Calc
Fe–30Cr–30Ni (Nb, Ti, Zr) 1.45 2.05 0.13 0.33 0.10 0.11
Fe–25Cr–35Ni (Nb) 0.64 2.15 0.28 0.37 0.18 0.10
Fe–26Cr–36Ni (Nb, low C) 0.44 1.32 0.17 0.24 0.06 0.06
Fe–35Cr–45Ni 0.44 0.50 0.10 0.08 0.04 0.03
Fe–19Cr–45Ni (Mo,Hf) 0.41–0.43 0.99 0.13–0.22 0.18 0.05–0.08 0.03
Fe–30Cr–45Ni (Hf) 0.63 0.62 0.10–0.15 0.15 0.04–0.05 0.02
Fe–25Cr–60Ni–2.4Al 0.14 0.87 0.04 0.17 0.03 0.03
Fe–25Cr–60Ni–3.6Al 0.01 0.95 0.01 0.19 0.02 0.03
Fe–25Cr–60Ni–4.8Al 0.02 0.90 ND 0.18 ND 0.03
602CA (2.3 Al, wrought) 0.14 0.82 0.04 0.17 0.03 0.03
ND, not determined.
1.12.5.1 Effect of Carbon of carbon to a level of 2.5 w/o chromium, forming

Cr7C3. The value of NM in the molybdenum-bearing
Cast alloys usually contain high levels of carbon, which
alloy is on this basis calculated to be 13.6% higher
segregates as M23C6 during alloy solidification, thereby
than for the comparison alloy. Taking into account
affecting NM as described above. The success of this
the effects of alloy compositional changes in DC, NC,
description is tested by comparing carburization rates
and NM gives predicted relative reductions in the kp
for the alloy Fe–25Cr–35Ni–Nb with high and low
values when Mo is added of 40% at 900 C, 23% at
carbon. At 1000 C, the ratio kp ðHigh CÞ=kp ðLow CÞ
1050 C, and 10% at 1150 C. The measured reduc-
predicted from eqn. [24] to be 1.5 compares well with
tions were 44% at 900 C, 24% at 1050 C, and 16%
the measured value of 1.7. Agreement at the other
at 1150 C. This suggests that eqn. [24] enables the
temperatures is also good. It is concluded that the
effect of molybdenum to be modeled, subject to the
method used to calculate the effect of original alloy
reliability of the NC and DC data. Furthermore, it also
carbon is successful. On this basis, it would follow that
predicts correctly the effect of temperature on the
cast alloys have lower carburization resistance than
efficacy of this element in reducing carburization.
their wrought (low carbon) equivalents. In fact, the
opposite effect is found,8,38 as a result of rapid grain
boundary diffusion of carbon in wrought alloys. 1.12.5.3 Effect of Silicon
It has long been known29 that silicon slows carbur-
ization, even under gas conditions where no silicon-
1.12.5.2 Effect of Molybdenum
rich oxide can form. The stability of SiC is a great
Molybdenum can be added for solution strengthen- deal less than that of Cr23C6 and Cr7C3 and no SiC
ing of an alloy, and is also a carbide former. Two will form in these chromium-rich alloys. Increasing
alloys containing 24Cr, 32Ni, 0.8Nb, and 0.44C, the alloy silicon content therefore has no effect on
with and without the addition of 3 w/o Mo were NM. Under the reducing conditions of these experi-
found38 to carburize at aC ¼ 1 at significantly differ- ments no SiO2 is formed, and the beneficial effects to
ent rates. The carbides Mo3C and Cr7C3 are of com- be expected of silicon on carburization rates under
parable stability, and can therefore co-exist if the reducing conditions, therefore, result from modifica-
metals are at similar activity levels. In fact, the car- tion of the carbon solubility and/or diffusivity. These
bides (Mo2Fe)C, (CrFeMo)C, and Cr7C3 were all modifications are due to thermodynamic interaction
identified by X-ray diffraction analysis of the carbur- between the dissolved silicon and carbon.
ðsÞ
ized alloy C. Equation [24] is now used to test the Silicon is known to reduce both NC and DC. Roy
48
possibility that precipitation of molybdenum car- et al. have examined the effect of silicon on carbon
bides slows the rate. diffusion in Fe–Si–C. A comparison of carburization
A level of 3 w/o molybdenum, forming a carbide rates of two cast heat-resisting steels which differed
of stoichiometry Mo1.5C (an average of the two only in their silicon levels showed38 that increasing
observed carbides), is equivalent in its consumption the silicon level decreased the rate by more than
would be predicted from Roy’s diffusion data. The dendrite boundaries (Figure 9), reactive metal carbide
other major contributory factor is the depression of precipitation at these interfaces could exercise a dis-
carbon solubility by silicon. The effect has been proportionate effect on the overall rate.
measured in liquid iron alloys, where the resulting
change in carbon solubility is significant, but no data
directly applicable to heat-resisting alloys is available. 1.12.5.5 Effect of Aluminum
The 60Ni alloys in Table 10 were predicted from
1.12.5.4 Effect of Niobium and Reactive eqn. [24] to carburize at rates similar to the 45Ni
Elements alloys. At 900 C, the wrought alloy 602 CA and
two cast versions reacted at close to the predicted
Niobium is often added to cast heat resistant alloys for
rates. At higher temperatures, the rates were much
strengthening purposes. It is also found in some
slower than predicted. The explanation is clear from
wrought alloys, where it improves weldability. Reac-
the micrographs of Figure 11, where a protective,
tive elements such as Ce and Hf are added to modify
external scale is seen to form on high-NAl alloys. The
carbide shapes and to improve oxide scale spallation
scale was a-Al2O3, which is thermodynamically sta-
resistance. All are strong carbide formers and have
ble at the water vapor impurity levels which are
strong effects on carburization resistance. The benefits
unavoidable in reaction gases. However, external
of niobium have been reported several times.8,45,49 The
scale formation is possible only when a sufficient
variation of kpðiÞ with niobium concentrations is shown
flux of alloy solute aluminum is available, and inter-
in Figure 10 for several heat-resisting alloys. The
nal oxidation can be avoided (see eqn. [15]). This flux
effect of niobium can be distinguished from variations
ðsÞ increases with both alloy NAl and temperature,
in NC DC =DM also plotted in this figure. Even if all
through its effect on DAl, qualitatively accounting
the alloy niobium was available in solution to precipi-
for the observed pattern of behavior. When this
tate NbC, the effect of adding 1–2 w/o Nb on NM is
scale formed, it functioned as a barrier limiting car-
very small, much less than the substantial reductions
bon access to the underlying metal. Thus, at 900 C,
in kpðiÞ seen at higher niobium levels.
no alumina formed and the 60Ni alloys all carburized
Similar effects have been noted for additions of
at the expected rates. At 1000 and 1100 C, alumina
Ce39 and Hf.40 These elements are present at low
scales grew on all alloys, and carburization was
concentrations, typically around 0.1–1.0 w/o, and
slowed. The scale was discontinuous on the low-
their effect on the value of NM is negligible. Nonethe-
aluminum alloys, and carburization was not com-
less, their addition is found to reduce carburization
pletely suppressed. A minimum aluminum content
rates substantially. It is possible that carbides of Nb,
of 4 w/o was required to achieve complete protection.
Ce, and Hf precipitate preferentially at sites where
interference with carbon diffusion is maximal. As car-
bon penetration is more rapid at primary carbide/
1.12.5.6 Alloying for Carburization
Protection
4
5 The kinetics of alloy carburization are very well
36X described by diffusion theory, and a rational approach
3 4 to alloy design is therefore available. Unfortunately,
10–3DcNc /Nm (μm2 h−1)
36XS however, most methods of suppressing internal attack

10–3 Kp(μm2 h−1)
3 on chromia-forming alloys are either impractical or

2 35CW
only modestly successful. It is not possible to adjust
2 chromium levels to achieve exclusive external car-
2325Nb
1 bide growth. Modifications of alloy carbon perme-
1 ability through adjusting the Fe/Ni ratio or alloying
with other metals yield only small improvements in
0 0
0 0.5 1 1.5 2
carburization rates. Silicon decreases carbon solubil-
w/o Nb ity and diffusivity quite strongly, but metallurgical
Figure 10 Carburization rate constants for commercial
limits on alloy silicon concentrations mean that only
25Cr–35Ni alloys at 1000 C as a function of niobium modest improvements in carburization resistance can
ðsÞ
content (○) compared with NC DC =NM (●). Data from Ref. 8. be obtained. The only really successful alloy additive
40 μm 50 μm
(a) (b)
Figure 11 Effect of Al2O3 formation on carburization of 60Ni alloys (a) low-Al at 900 C and (b) high Al at 1100 C.
is aluminum, and it functions by forming an oxide the carbide. The carbon activity at the cementite
scale. The general question of protection against car- surface is then supposed to be unity (rather than
burization by oxide scale formation is considered in the supersaturated value of the gas phase); the
Section 1.12.8. cementite becomes unstable and decomposes via
the reaction
Fe3 C ¼ 3Fe þ C ½31
1.12.6 Metal Dusting of Iron and
Ferritic Alloys producing finely divided iron and carbon. The iron
particles produced in this way are catalytically active,
Metal dusting is a catastrophic form of corrosion in and lead to accelerated carbon deposition. The kinet-
which metals exposed to carbon-supersaturated gas ics of carbon deposition were observed61 in the short
disintegrate, forming metal-rich particles (the ‘dust’) term to follow the quadratic rate law
dispersed in a voluminous carbon deposit. Early DWC =A ¼ kC t 2 ½32
reports of industrial failures50–53 were followed by
the laboratory research of Hochman54–56 concerning where DWC =A is the carbon weight per unit area,
the dusting of iron, nickel, cobalt, and chromia- before becoming approximately linear. The form of
forming ferritic and austenitic alloys. Subsequently, eqn. [32] was explained by Grabke et al.61 as being due
work by Grabke57–61 quantified and extended Hoch- to the generation of catalytically active particles by
man’s observations. The description of the process, as metal consumption in the dusting process:
provided by Hochman and Grabke, for pure iron is DWM =A ¼ kd t ½33
shown schematically in Figure 12.
where DWM =A is the metal wastage expressed as a
mass loss per unit area. If the catalytic particles are
1.12.6.1 Metal Dusting of Iron of uniform surface area and activity, then the rate of
carbon deposition is proportional to the mass of
When iron is exposed to carbon-rich gas at oxygen metal consumed
potentials too low to form iron oxide, the metal
catalyzes reactions such as [7]–[9], but the resulting dðDWC =AÞ
¼ k kd t ½34
carbon is dissolved in the metal. Hochman and dt
Grabke suggested that this leads to carbon supersat- Integration of this expression yields eqn. [32]. In the
uration of the iron, and the subsequent precipitation long term, individual particles are encapsulated by
of metastable Fe3C phase which they observed. The carbon and deactivated as catalysts. The rate at which
appearance of the cementite is shown in Figure 13. this happens approximately balances the rate of new
According to the proposed mechanism, once the iron particle generation, as coking rates become approxi-
surface is covered with cementite, carbon deposits on mately constant.
(1) (2) (3) (4)

Fe Gas Fe Fe3C Gas Fe Fe3C C Gas Fe Fe3C C + Fe
ac > 1
JC JFe
Fe3C 3Fe + C
Figure 12 Hochman–Grabke model for dusting of pure iron.
the carbon tube axis.71 This allows Fe3C planes in the

{010} and {100} families to be parallel to the tube
Coke
axis. The d-spacing of the (020) plane in 0.337 nm and
that of the (300) plane is 0.169 nm. These correspond
closely to the (0002) and (0004) d-spacings of graph-
Fe3C Fe matrix ite (0.337 and 0.168 nm, respectively). Accordingly, it
is suggested that alignment of these planes with the
graphite basal planes, which form the multiple walls
of the nanotube, leads to formation of low-energy
graphite–carbide interfaces and a preferred growth
20 μm orientation for the carbon tubes.
A mass transport model for filamentary coke deposit
Figure 13 Cementite produced by reaction of pure iron
is shown in Figure 14(c). The exposed Fe3C facets are
with CO/H2/H2O (aC ¼ 2.9, pO2 ¼ 2 1023 atm).62 in contact with the gas, and catalyze carbon production.
This carbon diffuses through the particle to the Fe3C-
graphite interface where growth of the carbon tube
Transmission electron microscopy (TEM)63–66 continues. The model is analogous to that originally
has revealed that the carbon at the cementite surface proposed by Baker et al.72 for catalysis of carbon fila-
is mainly graphite at temperatures above 500 C. It ments grown by metallic particles. There remains, how-
was suggested that the iron resulting from cementite ever, the question of how the Fe3C particles are formed.
decomposition dissolved in the graphite, diffused If iron does, in fact, dissolve in graphite, it could
outwards and precipitated as metal particles which diffuse outward if a carbon activity gradient was in
catalyzed further carbon deposition. The evidence effect, as illustrated in Figure 15(a). The iron flux is
for this was the measurement by EDS of a small given by
concentration of iron in the graphite. However, it is
@mC
difficult to understand the driving force which would JFe ¼ BFe NFe ½35
cause iron to diffuse from a low-activity source, the @x
cementite, towards a high activity destination, metal- and results from the thermodynamic Fe–C interaction.
lic iron particles. Here, BFe is the mobility of iron in graphite. When the
Further examination of the particles by both solute iron reaches a position at which aC is high
X-ray and electron diffraction23,67–70 has established enough to stabilize cementite, that phase precipitates.
that they are Fe3C. As seen in Figure 14, much of the To test this model, one needs values for iron solubility
coke deposit is filamentary. These filaments usually and its diffusion coefficient in graphite, and these are
carry faceted Fe3C particles at their tips. The parti- lacking. It is clear that an extremely high value for DFe
cles are oriented with their [001] direction parallel to would be required to explain why the decomposition
C (gr) Fe3C Fe (C)
aC
JFe
1 µm
aC = 1
(a)
(a) C (gr)
+ Fe3 Fe3C Fe (C)
(10)
(001)
(101) aC
J
JC
C
25 nm
(b)
JC
Graphite Fe3C (b)
JC
Figure 15 Mass transport models for metal dusting when
cementite is formed: (a) cementite decomposition and iron
diffusion through graphite and (b) cementite disintegration
aC = 2.9 coupled with inward graphite growth.
ac = 1
(c) its activity by a coefficient gC, with aC ¼ gC C.
Figure 14 Coke filaments with Fe3C particles at their tips Approximating gC as a constant and setting aC ¼ 1
(a) SEM view, (b) TEM image, and (c) mass transfer model. at the iron–graphite interface, we obtain

Fe C=Fe
reaction [31] leads to 75 at % iron, but produces none Rate ¼ DC aC 3 1 =gC ½37
of that phase at the decomposition site.
Fe C=Fe;
Studies of cementite decomposition in CH4–H2 Using standard data for DC 74 and aC 3 75
it is
gas mixtures73 have shown that the reaction products found that the temperature dependence of cementite
are iron and graphite. In that case, the reaction is decomposition predicted by eqn. [37] is in very good
controlled by the diffusion of carbon through the agreement with experimental observation (Figure 16).
product ferrite, driven by the carbon activity gradient Cementite decomposition by that mechanism is
between the high value at the Fe3C/Fe phase bound- clearly not occurring in the iron dusting reaction
ary and its value of unity at the Fe/C(gr) boundary: depicted in Figures 13 and 14.
The alternative mechanism of Fe3C particle pro-
Rate ¼ DC C Fe3 C=Fe C C=Fe ½36 duction is mechanical disintegration resulting from
volume expansion.67,76 Because Fe3C is a carbon dif-
where C Fe3 C=C and C C=Fe are carbon concentrations fuser,77,78 the cementite layer grows inward and is
at the ferrite–cementite and ferrite–graphite inter- consequently under compressive stress. Precipitation
faces, DC is the carbon diffusion coefficient in ferrite, of graphite could occur at internal defects in a nucle-
and the variation in the diffusion path length is ation and growth process.67 Such a process would be
ignored. The concentration of carbon is related to similar to the way carbon forms and grows at the rear
5 2 respect to graphite by alloying with silicon to raise

Rate of decomposition (s–1 3 104)
the solute carbon activity. Unfortunately, silicon also

4 oxidizes in the gases under discussion, as will be dis-
1.5
cussed in the following sections. Germanium, on the
(aC–1) 3 Dx107
3 other hand, forms a much less stable oxide, and by
1 virtue of its chemical similarity to silicon, might be
2 expected to suppress Fe3C formation. This is indeed
0.5
the case, as shown in Figure 18, where graphite is seen
1 to be growing directly into a ferritic Fe–Ge alloy, in
the absence of any cementite. The nanoparticulate
0 0 material near the disintegrating interface is also a-Fe
450 550 650 750 850
Decomposition temperature (8C) (Ge), as are the particles found on coke filaments.
Alloying with germanium suppressed Fe3C for-
Figure 16 Rate of Fe3C decomposition measured (▲) in
H2/CH4 and calculated (▪) from DC and aC in ferrite.73
mation, but did not prevent metal dusting. Instead,
dusting occurred more rapidly by the growth of
graphite directly into the alloy. Metal particles were
of Fe3C particles (Figure 14). Growth of these preformed by disintegration of the bulk metal, as the
cipitates could then disrupt the cementite surface. graphite grew inwards. Again it is suggested that
Examination of the C(gr)/Fe3C interface in this was a consequence of the volume expansion
Figure 17 shows that graphite is growing into the accompanying nucleation and growth of graphite
cementite layer, and that nanoparticles of Fe3C are within the metal. This process was more rapid than
detached from the bulk carbide. For these to exist, the corresponding one involving Fe3C. It is note-
aC must be high enough to stabilize the phase. worthy that the Fe–Ge/graphite interface morphol-
It is therefore concluded that the graphite layer ogy is similar to that developed between nickel and
does not function as an effective barrier to the gas, graphite (where no carbide forms) during dusting
and that aC at the coke–cementite interface is proba- (see Section 1.12.7). However, the dusting rate was
bly close to the value in the ambient gas. much faster for the ferritic material. It is concluded
This conclusion is supported by the observation79 that suppression of Fe3C formation does not prevent
that the surface cementite layer continues to thicken, dusting when this alternative mechanism is available.
and that the carbon content of the iron specimen
increases as dusting proceeds. It is clear that carbon is
1.12.6.3 Effects of Temperature and Gas
diffusing through the cementite scale and into the
Composition on Iron Dusting
iron, and that the description of Figure 15(b) applies.
However, these experiments were of limited duration As noted by Grabke,82 iron dusting and coking kinet-
(up to 40 h) and at a single temperature. Once the ics are very complex, and more detailed studies are
iron sample reaches a steady state of carbon supersatu- needed to arrive at a comprehensive, self-consistent
ration, the mechanism may well change80 when an picture. As seen in Figure 19, somewhat different
inward flux of carbon is no longer possible. Moreover, temperature dependencies have been reported
it is likely that the mechanism changes with gas compo- under different gas conditions. Grabke et al.61 con-
sition and temperature. Zhang et al.81 have reported sidered the temperature dependence at T < 540 C
that at T ¼ 700 C, and low pCO values, the surface to reflect rate control by cementite decomposition,
Fe3C scale decomposes to form a surface layer of ferrite. which is independent of gas composition. Ramaraya-
At still higher temperatures, no cementite layer is seen, nan et al.65,84 identified two temperature regimes:
and graphite deposits directly into the metal. T < 450 C where the coke was amorphous and the
rate was controlled by physical disintegration of
Fe3C, and 450 < T < 570 C where the chemical
1.12.6.2 Iron Dusting in the Absence of
decomposition of Fe3C was thought to become
Cementite
important. Its rate increased as the graphitization of
Given the important role played by cementite in the coke increased with temperature, which was thought
dusting of iron, it is reasonable to enquire whether to provide a diffusion pathway for dissolved iron. The
dusting might be prevented if Fe3C formation was decline in dusting rates reported by Ramarayanan
suppressed. Cementite can be destabilized with et al. at T > 570 C was attributed by them to a
Coke X Graphite
X
Fe3C Fe
Fe3C
1 μm
(a) (b)
Figure 17 Graphite–cementite interface developed during dusting of iron at 680 C: (a) FIB milled section and (b) TEM bright
field with SAD identifying nanoparticles as Fe3C.62
(a) (b)
Figure 18 Fe–10Ge alloy after 10 h reaction in H2/H2O/CO (aC ¼ 2.9, pO2 ¼ 2 1023 atm): (a) FIB cross-section and (b) TEM
cross-section with SAD identifying a-Fe.62
decrease in aC with increasing temperature. Muller- around 550 C, both aC and pCO can vary consider-
Lorenz and Grabke83 observed an increased dusting ably in a nominal CO–H2 gas mixture.
rate in the range 540–620 C, reporting it to vary with The dependence of both coking and dusting rates
the product pCO pH2 , and concluded that carbon trans- on the composition of CO/H2/H2O gases deter-
fer was rate-controlling. mined by Muller-Lorenz and Grabke83 is shown in
Part of the reason for this confusion is the way in Figure 20. Similar results were found for iron dusting
which aC varies with temperature (Figure 1) and gas at 550 C by Chun et al.65 In both cases it was con-
composition (eqns. [10] and [11]) in CO/H2/H2O gas cluded that the rate-determining step in the dusting
mixtures. Thus, the driving force for carbon precipi- process was reaction [7] leading to
tation (aC 1) is related to gas composition variables
Rate ¼ r7 pCO pH2 ½38
which themselves appear in kinetic expressions. Dis-
tinguishing the two effects can be difficult, and is However, this analysis neglects the effect of aC.
impossible if the gas compositions are not carefully Experiments85 in which aC was constant revealed
controlled. Thus, the use of CO/H2 gases without that keeping the product pCOpH2 constant, but vary-
H2O to buffer the composition means that aC is ing the individual partial pressures changed both
uncontrolled, and will vary with the extent of carbon coking and dusting rates. Obviously, the simple
deposition. As carbon deposition rates are rapid description of eqn. [38] fails.
⬚C
600 500 400 2.5 50
1.0 Metal dusting kinetics
at 650⬚C
2.25Cr Mo
1Cr Mo
Coke formation rate r2 mg cm2 h1

2.0 40
Metal wastage rate r1 mgM cm2 h1

2¼Cr Mo
Coke formation
0.1 1
1.5 30
–2 –1
mg cm h
Fe
mm/y
2¼Cr Mo
1.0 20
Fe
0.01 0.1
Metal wastage
0.5 10
cm–2 h–1 0.1 0.3 0.5 0.7 Pco (bar)

1.1 1.2 1.3 1.4 1.5
–1 Figure 20 Dependence of coking and dusting rates on
103 T
pCO in H2/CO/H2O mixtures.83
1500 Regime I
4500
II III
1¼Cr½Mo and 5Cr steel (mpy)
50CO:50H2 4000
Metal dusting rate of Fe (mpy)
1250 t = 48 h More information is required on the variation of

3500
Metal dusting rate of
dusting kinetics with gas composition. Available data

1000
1¼Cr½Mo
3000 for dusting in CO/H2/H2O mixtures at 500 C86
steel 2500 indicates the rate increases with aC. Data at 550 C84
750
2000 for metal consumption rates of iron exposed to CO/H2
Fe
500 1500 mixtures reveal a maximum at pCO ¼ 0:5 atm ¼ pH2 . If
the unavoidable water vapor impurity level was the
1000
250 5Cr steel same in all gases used, then aC ¼ K7 pCO pH2 =pH2 O also
500
had its maximum at this composition.
0 0
300 400 500 600 700 800 900 1000
Temperature (⬚C) 1.12.6.4 Dusting of Low-Alloy Steels
Figure 19 Temperature effects on metal dusting for iron
and low-alloy steels: LHS in CO/H2/H2O83; RHS in CO–H2.84
Dusting of 214Cr–1Mo and 1Cr–12Mo steels is seen in
Figures 19 and 20 to be similar to pure iron in rate
and dependence on temperature and gas composi-
Considering the coking process first, it is seen that
tion. Reaction morphologies are also similar,82 and it
the Boudouard reaction [8] is likely to be important
may be concluded that mechanisms are the same.
at high pCO values. Furthermore, when pH2 is high, it
The reasons for the slightly faster dusting rates
is likely that methanation (the reverse process in [9])
observed on 214Cr–1Mo steel have not been estab-
will occur under catalyzed conditions. Ignoring the
lished. However, it is to be noted that in CO/H2/
reverse reactions in [7] and [8] along with the for-
H2O gases, the pO2 values are high enough to oxidize
ward process [9], we can write
the chromium. Although no Cr2O3 scales can form on
dðDWC =AÞ such a dilute alloy, conversion of the steel surface to
¼ r7 pCO pH2 þ r8 pCO
2
r9 pH
2
½39 Fe3C may lead to encapsulation of chromium-rich
dt 2
oxide particles. These might act as nuclei for graphite

where ri are the rate constants. As seen in Figure 21, precipitates, thereby accelerating the cementite
this expression is successful in describing coking disintegration.
rates with r7 ¼ 35:5;r8 ¼ 4:5, and r9 ¼ 5:6 mg cm2 Addition of silicon to iron has two effects: a partial
atm2 h1. It is concluded that coke deposition can destabilization of Fe3C with respect to C(gr) and the
be described in terms of gas–solid kinetics. promotion of SiO2 formation at the oxygen potentials
9 scale is formed and retained. If the alloy chromium

level is too low to form a Cr2O3 scale, the steel will
8
dust at essentially the same rate as a 214Cr–1Mo
7 steel.88 If the steel forms a continuous, adherent
6
chromia scale, resistance to dusting under isothermal
conditions is very good, because the scale prevents
r mg cm2 h1
5 carbon entry. The factors determining the success or

4 otherwise of a steel in resisting dusting are those
governing its ability to quickly form a continuous
3
Cr2O3 scale by diffusing chromium to the surface.
2 The effect of temperature on DCr is clear from
1
the studies of Grabke et al.60 Steels containing 17 and
26 Cr showed complete resistance to dusting at
0
0 1 2 3 4 5 6 7 8 9
650 and 600 C, but underwent a slight extent of
k1pCOpH2 + k2pCO22 + k3 pH22 cm–2 h–1
attack at 550 C. Thus, the susceptibility to dusting
increased as the temperature and DCr decreased. The
Figure 21 Variation of coking rates on iron according to effective value of DCr can be increased at the low
eqn. [38].85
temperatures involved here by creating a deformed
and fine-grained alloy surface. This is done by sur-
face grinding, shot peening, etc., and has been
Internal Cementite shown88 to lead to better dusting resistance. In the
silica absence of such treatment (or after its effects have
Graphite been annealed out), the ferritic nature of the alloy is
itself important, because of the higher DCr value
compared to austenitic materials. A comparison of
the dusting performance of model ferritic and aus-
tenitic 25Cr alloys in Figure 23 illustrates this point.
These alloys had been electropolished to remove any
cold-worked surface material, so that alloy chromium
transport was via lattice diffusion.
Alloys which successfully develop continuous,
Beam pA Mag Det 04/13/05 Tilt 2 μm protective chromia scales are nonetheless subject to
30.0 kV 25.0 25.0 kX CDM-E 12:31:05 45.4⬚
long-term dusting attack. Under isothermal exposure
Figure 22 FIB image of Fe–xSi after dusting in CO/H2/ conditions, growth stress accumulation in the scale
H2O (aC ¼ 2.9, pO2 ¼ 1023 atm) at 680 C.87
leads ultimately to mechanical failure. A series of
such events can exhaust the capacity of an alloy to
of CO/H2/H2O gases. At low-silicon alloy levels, the
regrow its protective scale, and metal dusting
SiO2 forms as a dendritic internal precipitate rather
ensues.88,89 Discontinuous exposures combine the
than an external scale (Figure 22). Thus, the SiO2
effects of accumulated growth stress and occasional
provides little or no protection against carbon access
thermal shock. These have also been shown to pro-
to the metal. The Fe3C layer formed by Fe–Si alloys
duce alloy depletion and eventual dusting.60,82,88,90
is thinner than on iron, coking is faster, and metal
Short-term thermal cycling experiments induce the
wastage is also more rapid.87 The internal SiO2 pre-
same failure mode, and are useful as accelerated cor-
cipitates are incorporated into the Fe3C scale as it
rosion tests.91 In all cases, the chromia scale
grows into the alloy. These might act as graphite
breaks down locally, allowing carbon to enter the
nucleation sites, thereby accelerating cementite dis-
chromium-depleted metal. Rapid inward diffusion of
integration and dusting.
carbon leads to internal precipitation of chromium
carbides, thereby preventing any subsequent rehealing
1.12.6.5 Dusting of Ferritic Chromium
of the scale. The depleted iron matrix forms a surface
Steels
cementite layer23,92 which disintegrates producing
The behavior of these alloys when exposed to CO/ numerous cementite particles which catalyze further
H2/H2O gas mixtures depends on whether a chromia coke deposition.23 This localized attack produces a
1000
Fe–25Cr–25Ni
Fe–25Cr
Coke deposits mg cm—2

100 Fe–25Cr–10Ni
Fe–25Cr–5Ni
10
Fe–25Cr–2.5Ni
1
0.1
0.01
1 10 100 1000
(a) Exposure cycles (h)
7
Mass change of substract mg cm—2
Ni
2 Fe60Cr
Fe25Cr2.5Ni
−3
Fe25Cr
−8 Fe25Cr5Ni
−13
Fe25Cr25Ni
Fe25Cr10Ni
−18
0 30 60 90 120 150 180 210 240
(b) Exposure cycles (h)
Figure 23 Coke deposition and metal wastage kinetics for electropolished 25Cr alloys at 680 C in CO/H2/H2O (aC = 2.9,
pO2 ¼ 1023 atm).23
pitted surface (Figure 24). As the reaction proceeds, who reported extensive attack at localized pits.
more pits form and they widen and coalesce until the After addition of 2.2% Cr to the alloy, dusting was
attack becomes general. confined to the unpolished specimen edges. With
If the alloy chromium level is high enough, dust- 4.8% Cr and 0.15% Zr, pitting was completely sup-
ing of ferritics can be prevented. An Fe–60Cr alloy pressed, and only a thin coke layer formed. Dusting
survived one thousand 1-h cycles at 680 C, forming was associated with formation of a surface layer of
only Cr2O3,23 which was impermeable to carbon. Fe3C. Further work on Fe–15Al and Fe–26Al by
Furthermore, the chromia was catalytically inactive, Schneider and Zhang94,95 showed that dusting was
and no coke deposited. also associated with internal precipitation of the k-
carbide, Fe3AlCx. Attack on Fe–15Al was general, but
was reduced to localized pitting by alloying with
1.12.6.6 Dusting of FeAl and FeCrAl Alloys
2.9% of Nb or Ta, and almost stopped by adding
Iron aluminides and FeCrAl alloys are able to 2% of either V or Ti. Increases in temperature or
develop alumina scales, and their ability to resist alloy aluminum content led to reduced dusting rates.
dusting is, therefore, of interest. The high-diffusion As pointed out by the authors, the observations are
coefficients characteristic of the ferritic FeCrAl consistent with protection against dusting due to
materials (typically Fe–20Cr–5Al) means that they Al2O3 scale formation. However, at the relatively
are able to reheal scales quickly, thereby preserving low temperature of 650 C, the binary intermetallic
the surface barriers to carbon entry. does not reliably form a continuous scale. Alloy addi-
Dusting of the intermetallic Fe3Al at 650 C in tions of Cr, Nb, Ta, V, and Ti all improve alumina
CO–H2–H2O was investigated by Strauss et al.,93 scale formation. Nonetheless, once the scale is
catalysis of reactions [7]–[9] producing carbon.

Hochman56 and Schneider et al.95 reported the rate
of carbon uptake to be much slower than the
corresponding process on iron. The kinetics are
Spinel
approximately linear94,95 after an induction period
of length varying with temperature and gas composi-
tion. Metal consumption kinetics have not been
(Fe,Cr)3C layer measured directly. It is usually assumed that the
carbon deposit contains an approximately constant
nickel concentration (1–2 w/o), and on this basis
linear dusting kinetics would be deduced. Chun
et al.96 measured metal surface recession after 7 h
reaction in unbuffered CO/H2 gas mixture of pro-
portions 25:1, and found average rates to be of order
Figure 24 Onset of dusting for Fe–25Cr reacted at 680 C 1 mm year1 at temperatures above 600 C.
in CO/H2/H2O (aC = 2.9, pO2 = 1023 atm) showing local Reaction morphologies vary with ambient condi-
internal carburization, surface cementite layer and its
tions, and the available information is incomplete.
disintegration.
The 25:1 CO/H2 gas produced external attack,
whereas at a CO to H2 ratio of 1:1, extensive internal
graphitization also resulted. The appearance of
damaged, rapid carbon entry leads to internal carbur-
coarse-grained nickel after exposure to a CO/H2/
ization of the alloy and prevents subsequent alumina
H2O gas is shown in Figure 25. Cold working the
rehealing. The mechanism is, thus, very similar to
metal surface prior to reaction induced recrystalliza-
that of attack on ferritic Fe–Cr alloys.
tion of the near-surface region, and graphite forma-
The FeCrAl materials provide much better dust-
tion at the multiple grain and sub-grain boundaries. It
ing resistance. Baker and Smith90 reported that an
is clear that carbon dissolves in the nickel and diffuses
oxide dispersion-strengthened alloy, MA956, demon-
inwards, supersaturating the metal until graphite pre-
strated very good dusting resistance at 621 C up to
cipitates nucleate and grow at favorable sites. Nava
9000 h in a CO–H2–H2O gas which was oxidizing
Paz and Grabke58 reported earlier that CO/H2/
to aluminum. Good performance has also been
H2O mixtures with low pCO led to internal graphiti-
reported80,91 for FeCrAl materials at 650 C in simi-
zation, whereas high pCO mixtures induced surface
lar atmospheres. Internal precipitation of the k-
deposition.
carbide was observed after several thousand hours,80
Coke accumulated on the external nickel surface
and filamentary coke growth was catalyzed by Fe3C
in three forms: a more-or-less uniform layer on to the
particles91 when the scale was damaged by repeated
surface, clusters of approximately spherical particles,
thermal shocks.
and filaments.97 The spheres contain nickel particles
and the filaments carry particles at their tips. The
presumed correlation between coke mass and metal
1.12.7 Dusting of Nickel and consumption is seen to be reasonable. Nickel is cata-
Austenitic Alloys lytically active to all of reactions [7]–[9], and it was
proposed long ago72 that reactions such as
Metal dusting of nickel and austenitic alloys differs
from the reaction of ferritic materials in that cementite CO þ H2 ¼ H2 OþC ½40
is not formed and the corresponding nickel carbide is where C represents dissolved carbon, occur on the
unstable. An examination of the dusting behavior of bare facets of the nickel particles. The carbon then
pure nickel provides a good basis for understanding diffuses rapidly through the particle and precipitates
the reaction of austenitic, heat-resisting alloys. at the rear faces, causing elongation of the carbon
filament. If the exposed nickel faces cannot dissolve
carbon quickly enough, the particle is encapsulated
1.12.7.1 Metal Dusting of Nickel
with graphite, forming a roughly spherical particle.
Exposure of nickel to carbon-rich gases at oxygen The accumulation of the outer, loose-coke deposit is
potentials where the metal does not oxidize leads to thereby explained. Of more interest, however, is the
Polished
500 μm
Etched
Figure 25 Graphitization of nickel exposed to CO/H2/H2O (aC = 19) for 100 h at 650 C.
development of coke at the metal surface, and that a single-phase graphite–nickel solution might
the way in which parent metal is ‘dusted’ to form the form under other reaction conditions and the nickel
catalytic nanoparticles. diffusion model could apply. In the example of
Zeng and Natesan98 used Raman spectroscopy, Figure 26, it appears that mechanical disintegration
which is sensitive to the degree of carbon crystallin- of the metal is a consequence of the inward growth of
ity, to show that the surface carbon layer was more graphite and the accompanying volume expansion.
graphitic than the outer coke. Grabke et al.,99,100 Such a process was in fact deduced from the original
Pippel et al.,64 and Chun et al.96 all used TEM to electron microscopy studies,64,96,99,100 and the pro-
examine the nickel–carbon interface. These authors posed diffusion of nickel through graphite is of sec-
agreed that the carbon was graphite, that the graphite ondary importance.
basal planes were oriented approximately normal to The mechanism of graphite nucleation and
the nickel surface when dusting occurred, and that growth is of fundamental importance to the dusting
nickel was dissolved (1–2 w/o) in the graphite. The process. It is proposed96,99,102 that the free edges of
mechanism deduced from these observations was one graphite basal planes act as attachment sites for car-
of outward diffusion of solute nickel through the bon atoms, permitting their extension into the metal
graphite, followed by precipitation of nickel particles (Figure 27, left). The supply of carbon necessary for
in the outer regions of the graphite layer. The funda- this process can only be maintained if direct gas
mental difficulty with this mechanism is essentially access to the metal continues throughout the reac-
the same as was identified earlier in the iron dusting tion. Even when the surface is covered with graphite
case. No driving force is apparent for mass transfer and coke, no effective barrier to the gas is formed.
from bulk nickel to particulate metal, which would Most of the coke is obviously porous, and even the
presumably be at a higher energy level as a result of more dense graphite layer is extensively fissured.
its surface area. The factors controlling graphite formation on
A TEM image of the graphite layer and nickel nickel have been investigated intensively because car-
concentrations analyzed within it using EDS101 are bon fouling (coking) of industrial nickel catalysts is an
shown in Figure 26(b). Very little concentration important practical problem. Direct surface observa-
gradient is apparent, suggesting either that no diffu- tion103 using low energy election diffraction showed
sion occurs or that DNi in graphite is extremely high. that a preferred epitaxial relationship developed
Examination of the microstructure in Figure 26 between the graphite basal plane (0001) and Ni(111)
shows that in fact nickel nanoparticles are distributed faces. A computer simulation of this arrangement is
throughout the graphite layers. Thus, the surface shown in Figure 27 (right). Electron diffraction104
layer is a two-phase, two-component material in confirmed that (111), (113), and (220) nickel faces
which isothermal diffusion could not occur if local were found at filament–metal interfaces. The same
equilibrium was in effect. It is nonetheless possible preferred epitaxies are observed in metal dusting.
300
250 NiK
200
Intensity
150
100
50
0
0 0.5 1 1.5 2 2.5
(a) Distance (µm )
(b) 100 nm 100 nm
Figure 26 (a) TEM bright field view and EDS line scan through uniform graphite layer on nickel and (b) bright and dark field
images using (111) nickel reflection reveal particulate metal in graphite (CO/H2/H2O, aC ¼ 19, T ¼ 680 C).101
Ni
C
Figure 27 Left: Schematic view of graphite growth into nickel. Right: Computer simulation of the epitaxial relationship
between the graphite basal plane and a Ni(111) surface.
Examination by TEM of reacted nickel single surface graphite grew into the nickel, causing its
crystal and polycrystalline surfaces64,96,100 revealed disintegration. This pattern of behavior is clearly
that graphite basal planes developed parallel to nickel consistent with the reaction model of Figure 27,
(111) and (110) surfaces, but at right angles to a (100) which requires the graphite basal planes to be ori-
surface. In the short term, no dusting occurred ented at an angle to the surface. Of course, the model
at the (111) or (110) surfaces, but on the (100) is somewhat over simplified. Consider, for example,
a (111) surface which is in fact intersected by planes morphologies were the same as for pure nickel. The
such as (111), providing favorable inward growth dependence of dusting rate on alloy composition can
directions for graphite. These will be accessible at be understood in part from carbon permeability data.
surface jogs, kinks, etc., and dusting does, in fact, The higher permeability of iron-rich alloys would
ultimately commence on these surfaces.64,96,100 provide a greater flux to the graphite nucleation
Figure 28 shows graphite growing into the metal, and growth sites, supporting more rapid graphitiza-
along nickel (111) and (113) planes. Penetration of tion and metal dusting. On this basis, however, pure
graphite basal planes between adjacent planes of the nickel would be predicted to dust more rapidly than
ðsÞ
nickel lattice destroys its structure. It has been sug- alloys with 80% Ni, which have the lowest NC DC
gested98,101 that the graphite nucleates within the values. In fact, the metal dusted more slowly than the
metal interior. This is self evidently the case for alloy. More information is required for high Ni/Fe
internal graphitization (Figure 25). Such a process ratios, which are typical of Inconel alloys.
is analogous to the dissolution–precipitation mecha- Alloying copper with nickel has been found106,107
nism producing carbon filaments from nickel nano- to decrease coking and dusting rates sharply
particles. Some insight into the process can be gained (Figure 29). The coke deposit on alloys containing
from a consideration of alloying effects. at least 10 w/o copper consisted solely of filaments.
Thus, metal wastage via the process leading to graphite
particle clusters was suppressed. Copper is known to be
1.12.7.2 Dusting of Nickel Alloys in the
immune to dusting attack, but its effect on nickel alloy
Absence of Oxide Scales
dusting was much greater than one of simple dilution.
The dusting of austenitic Ni–Fe alloys at 650 C in Similar results have been reported108–110 for the effect
CO/H2/H2O gases such that no oxidation occurred of copper on catalytic coking by nickel. This can be
was studied by Grabke et al.,105 who found that both understood108,110,111 if the catalytically active sites con-
coking and metal wastage rates increased monotoni- sist of y near-neighbor atoms. Then the carbon deposi-
cally with iron concentration. The changes in coking tion rate on an alloy, r, is described by
rate reflected a combination of changing catalytic
activity and dust particle size. Regardless of alloy r ¼ rNi ð1 NCu Þy ½41
iron content within the range 0–70%, the reaction where rNi is the rate on pure nickel. The effect of
copper can be described by this equation with y ¼ 18,
as shown by the calculated lines in Figure 29.
A catalytic site of 18 near-neighbor atoms is physi-
111 cally unrealistic if surface reactions of simple mole-
cules (e.g., eqn. [40]) are involved. However, if graphite
113
nucleation is the process being catalyzed, then a stable
220
10
9 20 h
8 Lines: calculation, points: experiment
50 h
− 111 N: 20
113 7
111 100 h
6
150 h
5
4
− 111 3
111 2
1
111 0
–1
0 0.1 0.2 0.3 0.4 0.5
Cu content (at%)
Figure 29 Carbon uptake on Ni–Cu alloys at 680 C in

Figure 28 Nickel facets at graphite–metal reaction front CO/H2/H2O (aC = 19). Continuous lines calculated from
determined by SAD.101 eqn. [41].106
nucleus presumably requires at least one hexagonal 0.2

carbon ring. As seen in Figure 27 (right), it would
require 7 near-neighbor nickel atoms on a free surface
r, experimental (mg cm–2h–1)

or 14 atoms on adjacent (111) planes for internal 0.15
nucleation. It is therefore concluded that the copper

effect is consistent with internal graphite nucleation.
0.1
Copper alloying also affects carbon solubility in
the metal. The solubility is reported112 to be reduced
from a maximum of 0.18% in nickel to 0.01% in
0.05
Ni–90Cu. However, Mclellan and Chrashka113
showed that carbon solubility was unaffected by the
presence of up to 40% copper. More information is 0
required for carbon permeability, but on the basis of 0 0.05 0.1 0.15 0.2
r, calculated (mg cm–2h–1)
existing data, graphite nucleation appears to be more
strongly affected. Figure 30 Carbon uptake kinetics on nickel according to
eqn. [39].97
1.12.7.3 Effects of Temperature and Gas

Composition on Nickel Dusting
success of this description with r7 ¼ 0.73, r8 ¼ 0.06,
Average metal recession rates in a 50–50 mixture of and r9 ¼ 0.27 mg cm2 atm2 h1. This indicates
CO–H2 were found96 to increase with temperature to that coke formation is controlled by the CO þ H2
a maximum at about 800 C, and to remain constant at reaction [7] at moderate pCO levels and by the Bou-
higher temperatures. The carbon activity in those douard reaction [8] at high pCO levels. The methana-
experiments was uncontrolled, and interpretation of tion process is important when pH2 is significant. It
the high temperature results is therefore difficult. The needs to be recognized that different dependencies are
low-temperature results were correlated with an likely at different temperatures, and that the relation-
observed increase in carbon graphitization with ship between coking and dusting rates is likely also to
increasing temperature. Chun et al.96 suggested that be temperature dependent. More work is required to
dusting was controlled by outward diffusion of nickel obtain a full understanding of the effects of environ-
dissolved in graphite, and therefore accelerated with mental variables on nickel dusting.
increased graphitization of the coke. However, it could
also be argued that only graphite, and not amorphous
1.12.7.4 Dusting of Austenitic Alloys
carbon, was capable of growing into the metal, because
graphite can develop a crystallographic orientation Grabke et al.102,105 found that the dusting of binary
relationship with the metal. Increasing graphitization Fe–Ni alloys varied in reaction morphology and rate
would, therefore, increase the extent of nickel disinte- with nickel content. Essentially, low-nickel content
gration. Carbon uptake rates in a gas mixture of H2– alloys behaved like pure iron, forming a surface layer
24CO–2H2O were found by Schneider et al.99 to have a of cementite, whereas high-nickel alloys graphitized
maximum at about 625 C, and decreased sharply at directly without forming carbide. The nickel level
higher temperatures, reflecting the lower carbon activ- necessary to suppress cementite formation at 650 C
ities reached at higher temperatures in a gas of fixed was reported as 30%105 and also as 5–10%.102
composition. The dusting of austenitic chromia-forming alloys is
Direct measurements of the dependence of metal prevented for so long as the oxide scale acts as a barrier
consumption (i.e., dusting) rates on gas composition are to carbon ingress.56 The onset of dusting has been
lacking, but data is available for coking rates. At a fixed characterized by Grabke et al.58,60,105 and the general
temperature of 650 C, carbon uptake rates vary with features of the process are now clear. Selective oxida-
gas composition in a complex way. Experiments in tion of chromium produces a chromium-depleted sub-
which pCO and aC (as calculated from eqn. [10]) were surface alloy region. Local scale damage allows gas
varied independently97 showed that carbon uptake access to the metal. If sufficient chromium remains,
rates were not directly related to aC. Using instead an the Cr2O3 scale reheals; if not, other reactions follow. If
elementary kinetic description for reactions [7]–[9], pO2 is too low for nickel or iron oxides to form, carbon
one arrives at eqn. [39]. Figure 30 demonstrates the enters the alloy, precipitating chromium carbides.
At these low-temperatures, DCr is small, and the car- alloy 800, just as copper decreases the dusting of nickel
bides are very fine. Removal of chromium from the itself. This effect is limited by the alloy solubility for
matrix renders future oxide healing of the surface copper, which increases with nickel concentration.
impossible, and gas access to the chromium-depleted Szakalos et al.24,80,122 have pointed out that the fine
surface continues. The metal is essentially an Fe–Ni internal carbides formed after scale failure can be oxi-
alloy, and at high-nickel levels it undergoes graphitiza- dized in situ, leading to disruption of the metal and
tion and disintegration in the same way as pure nickel. contributing to the dusting process. This is, in fact, the
Thermal cycling dusting studies23 on model ‘green rot’ corrosion process,123 in which the large vol-
Fe–xNi–25Cr alloys revealed considerable variation ume expansion accompanying carbide oxidation frac-
in metal wastage rate with nickel content (Figure 23). tures the metal. Under dusting conditions, this could
A 2.5Ni alloy was ferritic, and formed a surface layer occur simultaneously with graphitization of the
of M3C, which disintegrated into cementite dust. chromium-depleted surface metal. The two possible
Alloys with 5 and 10Ni had duplex a + g microstruc- reactions for internal carbides near the surface are the
tures, in which the austenite was clearly carburized oxidation process and simple dissolution, providing a
more rapidly than the ferrite. Dusting produced chromium flux toward the surface. The competition
nanoparticles of M3C from the 5Ni alloys and both between the two processes will depend on oxygen and
M3C and austenite from the 10Ni alloy. A 25Ni alloy carbon permeabilities, and DCr within the subsurface
was fully austenitic, and disintegrated to yield aus- alloy region. Rates of carbon and oxygen dissolution into
tenite dust. This shift from carbide to austenite par- the region are also likely to depend on gas compositions.
ticles with increasing nickel levels is the same as that
of binary Fe–Ni alloys,102,105 and reflects the mecha-
nism of attack on chromium depleted surfaces. 1.12.8 Protection by Oxide Scaling
The variation of dusting rate with Fe/Ni ratio
shown in Figure 23 reflects mainly the difference As noted earlier, industrial gas streams which cause
in DCr accompanying the change from ferritic to carburization are almost always oxidizing to chro-
austenitic structures. Because the alloy surfaces mium, and therefore also to silicon and aluminum.
were electropolished and any cold-worked surface Heat-resisting alloys used at temperatures up to
regions removed, chromium was available to the sur- about 1000 C are usually chromia formers, and the
face only via lattice diffusion. Thus, rehealing was protective nature of their scales is what preserves the
more effective, and frequency of dusting initiation alloys from carburization. Using radioactive 14C, Wolf
less, in the alloy sequence a > (a þ g) > g. At still and Grabke124 showed that the solubility of carbon in
higher nickel levels, improved performance resulted Cr2O3 and Al2O3 at 1000 C is below the detectabil-
from the lowering of alloy carbon permeability. ity limit of 0.01 ppm. Nonetheless, chromia scales
A growing body of results on the dusting resis- grown on alloy surfaces can be permeable to carbon,
tance of austenitic alloys is becoming available. It is presumably by transport through defects or along
generally agreed60,90,105,114 that higher nickel levels internal surfaces. Grabke et al.125 showed that radio-
are beneficial and that a minimum chromium level tracer carbon in a CO/CO2/H2/H2O gas mixture
of about 25% is required.114,115 At these levels, slowly permeated scales on preoxidized Fe–Cr alloys.
scale breakdown allows formation of two internal Simultaneous internal carburization and external
carbide zones, usually spheroidal M7C3 near the sur- Cr2O3 growth has been observed126 on Fe–28Cr
face and lamellar or Widmanstatten M23C6 at greater exposed to CO/CO2 at 900 C. However, a Ni–
depths.23,116 Alloy additions of silicon and aluminum 28Cr alloy reacted in the same way formed no inter-
improve the ability of the scale system to exclude nal carbides, indicating a more gas-tight scale.
carbon by forming oxide subscales.59,90,117–119 Addi- Cast heat-resisting steels form scales consisting of
tions of carbide forming metals (Mo, W, Nb) form mixed carbides and oxides, the proportion of oxide
stable carbides. Their ability to getter carbon allows increasing with ambient pO2 . At high-oxygen activ-
unreacted chromium to reheal the surface scale, delay- ities, the scale is mainly Cr2O3 with an outermost
ing the onset of dusting.119 However, subsequent oxi- layer of manganese-rich spinel and, depending on
dation of these refractory metal carbides leads to alloy silicon levels, a more or less continuous SiO2
volume expansion, and disruption of the protective layer at the alloy–scale interface. For as long as they
scale.120 It has been shown121 that additions of copper maintain their mechanical integrity, these scales
improve the dusting resistance of 310 stainless steel and completely block carbon access to the underlying
alloys. The appearance of scales grown at 1000 C, 9

3Cr2 O3 þ 4C ¼ 2Cr3 C2 þ O2 ½42
low pO2 values and aC ¼ 1 and the corresponding 2
diffusion paths mapped on the thermochemical dia-
DG ¼ 3192100 797:3T ½43
gram are shown in Figure 31. The scale grown at
pO2 ¼ 1022 atm is a mixture of oxide (dark) and In the presence of a coke deposit, aC ¼ 1, and the
carbide (light) with a sublayer of SiO2 (black). The value of pO2 corresponding to the equilibrium [42] is
protectiveness of these scales depends on alloy silicon calculated to be as shown in Figure 33. In a steam
content, as shown in Figure 32, where a level of cracking reactor, the H2/H2O ratio is approximately
about 1.8 w/o is seen to reduce the carburization 1. The equilibrium pO2 values calculated for
rate dramatically at 1050 C. Kane29 reported a
1
value of 2 w/o at 1093 C. At pO2 ¼ 1024 atm, H 2 þ O2 ¼ H 2 O
Cr2O3 is unstable, but SiO2 still forms (Figure 31). 2
Exposure to these conditions127 led to a scale of car- are seen to be much higher than the Cr3C2/Cr2O3
bide over a thin silica layer at the alloy surface. This values, and the oxide is stable. However, as pointed
scale was not protective, and alloys carburized rapidly, out by Grabke et al.,6 the oxygen activity beneath a
even at silicon levels up to 2.4 w/o. It is concluded that carbon deposit can be a great deal less. If the carbon
conditions producing both SiO2 and Cr2O3 are neces- deposit is gas tight, then the gas species will be CO
sary to provide a carbon-resistant scale. and CO2, with the ratio pCO 2
=pCO2 set through the
The location of the stability boundary between Boudouard equilibrium [42] with unit carbon activ-
chromium carbide and oxide shown in Figure 31 is ity. If the total pressure pCO þ pCO2 ¼ 1 atm, the
temperature dependent: corresponding pO2 is found from the thermodynamics
(a) 20 μm (b) 20 μm
Cr3C2 T = 12738K
R2 R1
0
SiC Cr7C3
Cr23C6
–4
log10aC
Si SiO2 Cr2O3
–8
Cr
–12
–30 –25 –20 –15

(c) log10 po2
Figure 31 Scales grown on 25Cr–35Ni heat-resisting steels at 1000 C and aC ¼ 1; (a), pO2 ¼ 1022 atm, (b) pO2 ¼ 1024
atm127 and (c) diffusion paths.
8 T (°C)
1150 1100 1050 1000
−10
6
kP/(μm2h–1 ⫻ 103)
−12
4 H2/H2O
−14
log10po2
2
−16
Cr3C2/Cr2O3
0
0 1 2 3 −18 C/CO/CO2
Alloy silicon content w/o
Figure 32 Dependence of carburization rate on alloy −20
silicon content at 1050 C, aC ¼ 1, pO2 ¼ 3 1020 atm.128 7 7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 7.9 8
I(104 K )
of [2] to be low at high temperatures (Figure 33). Figure 33 Thermodynamics of oxide to carbide
If T 1100 C, carbon will reduce the oxide and conversion compared with CO/CO2 mixture at pT ¼ 1 atm in
degrade the protective nature of the scale. equilibrium with graphite, and with oxygen potential
Both Cr2O3 and Al2O3 scales provide effective controlled by H2/H2O equilibrium in steam cracking
furnaces.
barriers to carbon entry and metal dusting, but will
eventually fail by cracking or spallation. If sufficient
chromium or aluminum remains at the alloy surface, preoxidation conditions so that the oxides formed at
then scale rehealing occurs. If not, carbon dissolves in that stage are stable during subsequent service.
the depleted alloy and diffuses inward, to precipitate
carbides.129 At high carbon activities, metal dusting 1.12.8.1 Protection by Adsorbed Sulfur
follows. The competition between outward metal
diffusion to form a scale and inward carbon diffusion The introduction of gaseous sulfur species such as
should in principle be described by an equation such H2S to industrial process steams is widely practiced
as eqn. [15]. Unfortunately, no rigorous test of this in order to minimize carburization and metal dusting.
description is available. Qualitative support is Sulfur adsorbs on the metal surface, impeding carbon
provided by the finding91,130 that several heat- access.34,132,133 Under these conditions, rehealing of
resisting alloys can be ranked in metal dusting resis- damaged oxide scales is favored over carbon penetra-
tance during temperature cycling according to their tion. The effect increases with pS2 , but the sulfur
ðoÞ ðsÞ pressure must be kept below the value at which CrS
NCr DCr =NC DC values.
A practical problem arises in the use of oxide can form.
scales for protection against carbon. Preoxidation Adsorbed sulfur also provides protection against
procedures used to develop a chromia scale prior to metal dusting, delaying the onset of the process and
service can also develop an outer scale layer of spinel, allowing more time for oxide rehealing. Data assem-
MCr2O4. If subsequent service conditions provide an bled by Schneider et al.134 for the effect on iron is
oxygen potential below the spinel stability level, it is shown in Figure 34. The H2S/H2 ratios required to
reduced leaving particles of metal, yield protection increase with temperature because
sulfur adsorption is strongly exothermic. In the case
1
MCr2 O4 ¼ Cr2 O3 þ M þ O2 ½44 of iron, sulfur adsorbs on cementite and prevents the
2 nucleation of graphite.
These particles act as catalytic sites, accelerating the
onset of coking. This in turn can lead to scale disin-
1.12.8.2 Protection by Coatings
tegration and the commencement of dusting. Expos-
ing austenitic chromia formers to oxidizing and As is clear, long-term protection against metal dust-
carburizing conditions has been shown131 to lead to ing and carburization can only be achieved by form-
rapid scale failure, accelerated carburization, and in ing a stable oxide scale which is capable of rapid
some cases the commencement of dusting. In order to rehealing. Coatings with high concentrations of
avoid this effect it is necessary to adjust the scale forming elements can be used to provide this
1000 900 800 700 600 500 ⬚C the complete gas composition, including minority
pH2S species, as well as temperature. Considerably more
pH2 work is needed to define temperature and gas com-
10 –3 position effects on dusting rates.
Sulfur adsorbed FeS
→ No metal dusting Ferritic materials at moderate temperatures form
10 –4 Fe3C. This phase is either disintegrated by precipita-
Fe
tion within it of graphite or, in other gases, decom-
10 –5
posed to yield metallic iron. Unfortunately, the
boundaries between the two regimes are still not
10 –6
defined. Austenitic materials form no iron carbide,
10 –7
and are disintegrated by precipitation and growth of
No sulfur adsorbed graphite within the metal. A similar mechanism oper-
→ Metal dusting
10 –8 ates for ferritic materials when cementite formation is
prevented by high temperatures or alloying.
10 –9 Protection against carburization and dusting
8 9 10 11 12 104/T requires the provision of a surface barrier between
Figure 34 Effect of sulfur on metal dusting. The hatched metal and gas, either an adsorbed sulfur layer or an
region represents the transition to an iron surface saturated oxide scale. When attack by carbon does occur, it is
with sulfur.134 catastrophically rapid. For this reason, protective
measures must be employed, and their continued
protection. Chemical vapor deposition135–140 and effectiveness monitored.
flame spraying141 have been used to produce carbur-
ization resistant coatings, and their utility under
metal dusting conditions has also been tested.142 Alu- References
minum diffusion coatings were found to be protective
for a series of ferritic steels (2.25–28Cr) and the 1. Shatynski, S. R. Oxid. Met. 1978, 13, 105.
2. Wada, T.; Wada, H.; Elliott, J. F.; Chipman, J. Met. Trans.
austenitic alloy 800. However, long-term exposure 1971, 2A, 2199.
led to pore development under the scale and crack- 3. Wada, T.; Wada, H.; Elliott, J. F.; Chipman, J. Met. Trans.
ing. Silicon diffusion coatings did not develop pro- 1972, 3, 2865.
4. Smith, R. P. Acta Met. 1953, 1, 578.
tective scales. A flame sprayed g-TiAl coating was 5. Grabke, H. J.; Peterson, E. M. Scripta Met. 1978, 12, 1111.
successful on a ferritic material, but failed on alloy 6. Grabke, H. J.; Gravenhorst, U.; Steinkusch, W. Werkst. u.
800 as a result of thermal mismatch. This work Korros. 1976, 27, 291.
7. Schnaas, A.; Grabke, H. J. Oxid. Met. 1978, 12, 387.
showed that alumina scales provided better protec- 8. Smith, G. M.; Young, D. J.; Trimm, D. L. Oxid. Met. 1982,
tion against carbon than did chromia. 18, 229.
9. Wagner, C. Z. Elektrochem. 1959, 63, 772.
10. Kubaschewski, O.; Alcock, C. B. Metallurgical
Thermochemistry, 5th ed.; Pergamon: Oxford, 1983.
1.12.9 Conclusions 11. Mazandarany, F. N.; Pehlke, R. D. Met. Trans. 1973, 4A,
2070.
12. Allen, A. T.; Douglass, D. L. Oxid. Met. 1999, 51, 199.
Carburization reactions at aC 1 are described by the 13. Ramanarayanan, T. A.; Slolovitz, D. J. J. Electrochem.
classical theory of internal oxidation. Local equilibrium Soc. 1985, 132, 2268.
is achieved within the reacting alloy, solid-state diffu- 14. Ahmed, O.; Young, D. J. In High Temperature Corrosion
and Materials Chemistry; McNallan, M. J., Opila, E. J.,
sion of dissolved carbon controls the rate, and parabolic Maruyama, T., Narita, T., Eds.; The Electrochemical
kinetics result. Wagner’s diffusion theory provides good Society, Inc.: Pennington, NJ, 2000; Vol. II; p 77.
quantitative predictive capability, despite the approxi- 15. Benz, R.; Elliott, J. F.; Chipman, J. Met. Trans. 1974, 5,
2235.
mate nature of some of its assumptions. 16. Ford, S. Ph.D. Thesis, University of New South Wales,
Metal dusting reactions at aC > 1 proceed accord- 2005.
ing to complex mechanisms which are still not fully 17. Young, D. J.; Ahmed, O. Mater. Sci. Forum 2001, 369–
372, 93.
defined. Local equilibrium is not achieved within the 18. Udyavar, M.; Young, D. J. Corros. Sci. 2000, 42, 861.
gas or at the gas–solid interface. It is, therefore, nec- 19. Bose, S. K.; Grabke, H. J. Z. Metallk. 1978, 69, 8.
essary to consider both the thermodynamic state of 20. Ford, S.; Young, D. J.; McGrouther, D.; Munroe, P. R.
Mater. High Temp. 2005, 22, 351.
the gas and the kinetics of the several parallel gas– 21. Bunin, K. P. Izv. Chern. Metall. 1973, 2, 123.
solid reactions possible. This requires specification of 22. Mozchan, V. I. Izv. Chern. Metall. 1979, 8, 92.
23. Toh, C. H.; Munroe, P. R.; Young, D. J. Oxid. Met. 2002, 60. Grabke, H. J.; Krajak, R.; Muller-Lorenz, E. M. Werkst.
58, 1. Korros. 1993, 44, 89.
24. Szakalos, P.; Lundberg, M.; Petterson, R. Corros. Sci. 61. Grabke, H. J.; Bracho-Troconis, C. B.; Muller-Lorenz,
2006, 48, 1679. E. M. Werkst. Korros. 1994, 45, 215.
25. Ford, S.; Munroe, P. R.; Young, D. J. In John Stringer 62. Motin, M. A. A.; Munroe, P. R.; Brady, M. P.; Young, D. J.
Symposium; Tortorelli, P. F., Wright, I. G., Hou, P. Y., Scripta Mater. 2007, 56, 281.
Eds.; ASM International: Materials Park OH, 2003; p 77. 63. Pippel, E.; Woltersdorf, J.; Grabke, H. J.; Strauss, S.
26. Ohriner, E. K.; Morall, J. E. Scripta Met. 1979, 13, 7. Steel Research 1995, 66, 217.
27. Kirkaldy, J. S. Canad. Met. Q. 1969, 8, 35. 64. Pippel, E.; Woltersdorf, J.; Schneider, R. Mater. Corros.
28. Christ, H. J. Mater. Corros. 1998, 49, 258. 1998, 49, 309.
29. Kane, R. H. Corrosion 1981, 37, 187. 65. Chun, C. M.; Mumford, J. D.; Ramanarayanan, T. A. J.
30. Chu, W. F.; Rahmel, A. Oxid. Met. 1981, 15, 331. Electrochem. Soc 2002, 149, B348.
31. Roach, D. B.; Wright, I. G. The Study of the Carburisation 66. Schmid, B.; Walmsley, J. C.; Grong, O.; Odegard, R. Met.
Resistance of Heat Resistant Casting Alloys; 1st Interim Mat. Trans. A. 2003, 34, 345.
Report on Project No 60, Batelle, Columbus Labs, OH, 67. Zeng, Z.; Natesan, K.; Maroni, V. A. Oxid. Met. 2002, 58,
1974. 147.
32. Ledgeff, K.; Rahmel, A.; Schorr, H. Werkt. Korros. 1980, 68. Zhang, J.; Schneider, A.; Inden, G. Corros. Sci. 2003, 45,
31, 121. 1329.
33. Ramanarayanan, T. A.; Petkovic-Luton, R. Corrosion 69. Zhang, J.; Schneider, A.; Inden, G. Corros. Sci. 2003, 45,
1981, 37, 712. 281.
34. Barnes, J.; Corish, J.; Norton, J. F. Oxid. Met. 1986, 26, 70. Zhang, J.; Schneider, A.; Inden, G. Mater. Corros. 2003,
333. 54, 770.
35. Kinniard, S. P.; Young, D. J.; L Trimm, D. Oxid. Met. 1986, 71. Toh, C. H.; Munroe, P. R.; Young, D. J. Mater. High
26, 417. Temp. 2003, 20, 527.
36. Mitchell, D. R. G.; Young, D. J. Foundry Trade J. 1992, 72. Baker, R. T. K.; Barber, M. A.; Harris, P. S.; Yeates, F. S.;
166, 253. Waite, R. J. J. Catal. 1972, 26, 51.
37. Mitchell, D. R. G.; Young, D. J. J. Mat. Sci. Lett. 1993, 12, 73. Longbottom, R.J.; Ostrovski, O.; Zhang, J.; Young, D.J.
1076. Met. Mat. Trans. B. 2007, 38, 175.
38. Mitchell, D. R. G.; Young, D. J.; Kleeman, W. Mater. 74. Smith, R. P. Transactions, AIME 1962, 224, 105.
Corros. 1998, 49, 231. 75. Turkdogan, E. T. Physical Chemistry of High Temperature
39. Becker, P.; Ouamara, F.; Young, D. J. In High Technology; Academic Press: New York, 1980.
Temperature Corrosion Materials Chemistry IV; Opila, E., 76. Koszman In Proc. Symp. on High Temp. Gas-Metal
Hou, P., Moruyama, T., Pieraggi, B., Schifler, D., Reactions in Mixed Environments; Jannson, S. A.,
Wachina, E., Eds.; The Electrochemical Society, Inc.: Foroulis, Z. A., Eds.; Met. Soc. AIME: New York, 1973;
Pennington, NJ, 2003; p 178. p 155.
40. Becker, P.; Young, D. J. Oxid. Met. 2007, 67, 267. 77. Ozturk, B.; Fearing, V. L.; Ruth, J. A.; Simkovich, G. Met.
41. Klower, J.; Heubner, U. Mater. Corros. 1998, 49, 237. Trans. A. 1982, 13, 1871.
42. Young, D. J. Mater. Corros. 1999, 50, 675. 78. Schneider, A.; Inden, G. Calphad 2007, 31, 141.
43. Forseth, S.; Kofstad, P. Mater. Corros. 1998, 49, 266. 79. Young, D. J.; Motin, M. A. A.; Zhang, J. Defects and
44. Grabke, H. J. Carburisation – A High Temperature Diffusion, accepted for publication.
Corrosion Phenomenon; MTI Publication No. 52, 80. Szakalos, P. Mater. Corros. 2003, 54, 752.
Materials Technology Institute of the Chemical 81. Zhang, J.; Schneider, A.; Inden, G. Mater. Corros. 2003,
Processing Industries: St. Louis, MI, 1997. 54, 763.
45. Grabke, H. J.; Wolf, I. Mater. Sci. Eng. 1987, 87, 23. 82. Grabke, H. J. Mater. Corros. 2003, 54, 736.
46. Lewis, H. B. Corros. J. 1968, 3, 166. 83. Muller-Lorenz, E. M.; Grabke, H. J. Mater. Corros. 1999,
47. Rahmel, A.; Grabke, H. J.; Steinkusch, W. Mater. Corros. 50, 614.
1998, 49, 221. 84. Chun, C. M.; Ramanarayanan, T. A. Oxid. Met. 2004, 62,
48. Roy, S. K.; Grabke, H. J.; Wepner, W. Arch. 71.
Eisenhuttenwes. 1980, 51, 91. 85. Yin, H. ME Thesis, University of New South Wales 2006.
49. Hemptenmacher, J.; Grabke, H. J. Werkst. Korros. 1983, 86. Schneider; Vielhaus, H.; Inden, G.; Grabke, H. J.;
34, 333. Muller-Lorenz, E. M. Mater. Corros. 1998, 49, 336.
50. Camp, E.; Phillips, C.; Cross, L. Corrosion 1954, 10, 149. 87. Motin, M. A. A. PhD Thesis, University of New
51. Hubbell, W. G. The Iron Age 1946, 157, 56. Southwales 2009.
52. Burns, L. Corrosion 1950, 6, 169. 88. Grabke, H. J.; Muller-Lorenz, E. M.; Eltester, B.;
53. Lefrancois, P. A.; Hoyt, W. B. Corrosion 1963, 19, 360t. Lucas, M.; Monceau, D. Steel Research 1997, 68, 179.
54. Hochman, R. F. Proc. 3rd Int. Cong. Met. Corrosion, 89. Frabiezewski, A. S.; Watkins, W. R.; Hoffman, J. J.;
University of Moscow Press, 1969. Dean, S. W. Corrosion 2000; NACE: Houston, TX, 2000;
55. Hochman, R. F.; Klett, M. G. Proc. 5th Int. Cong. Met. paper no. 532.
Corrosion, NACE: Houston, TX, 1974. 90. Baker, B. A.; Smith, G. D. Corrosion 2000; NACE:
56. Hochman, R. F. In Proc. Symp. Properties of High Houston, TX, 2000; paper no. 257.
Temperature Alloys with Emphasis on Environmental 91. Toh, C. H.; Munroe, P. R.; Young, D. J.; Foger, K. Mater.
Effects; Foroulis, Z. A., Pettit, F. S., Eds.; Electrochemical High Temp. 2003, 20, 129.
Society: Pennington, NJ, 1977; p 571. 92. Grabke, H. J. Mater. Corros. 1998, 49, 303.
57. Grabke, H. J.; Hemptenmacher, J.; Munker, A. Werkst. 93. Strauss, S.; Krajak, R.; Palm, M.; Grabke, H. J. Mater.
Korros. 1984, 35, 543. Corros. 1996, 47, 701.
58. Nava Paz, J. C.; Grabke, H. J. Oxid. Met. 1993, 39, 437. 94. Schneider; Zhang, J. Mater. Corros. 2003, 54, 778.
59. Grabke, H. J.; Krajak, R.; Nava Paz, J. C. Corros. Sci. 95. Schneider; Zhang, J. J. Corros. Sci. Eng. 2003, 6, paper
1993, 35, 1141. H043.
96. Chun, C. M.; Mumford, J. D.; Ramanarayanan, T. A. 119. Strauss, S.; Grabke, H. J. Mater. Corros. 1998, 49, 321.
J. Electrochem Soc. 2000, 147, 3680. 120. Litz, J.; Rahmel, A.; Schorr, M. Oxid. Met. 1988, 29, 95.
97. Zhang, J.; Young, D. J. Corros. Sci. 2007, 49, 1496. 121. Zhang, J.; Young, D. J. Corros. Sci. 2007, 49, 1450.
98. Zeng, Z.; Natesan, K. Chem. Mater. 2003, 15, 872. 122. Szakalos, P.; Pettersson, R.; Hertzman, S. Corros. Sci.
99. Schneider, R.; Pippel, E.; Woltersdorf, J.; Strauss, S.; 2002, 44, 2253.
Grabke, H. J. Steel Research 1997, 68, 326. 123. Betteridge, W. The Nimonic Alloys; E. Arnold: London,
100. Wei, Q.; Pippel, E.; Woltersdorf, J.; Strauss, S.; 1959.
Grabke, H. J. Mater. Corros. 2000, 51, 652. 124. Wolf; Grabke, H. J. Solid State Commun. 1985, 54, 5.
101. Zhang, J.; Munroe, P. R.; Young, D. J. Acta Mater. 2008, 125. Grabke, H. J.; Ohla, K.; Peters, J.; Wolf, I. Werkst. Korros.
56, 68. 1983, 34, 495.
102. Pippel, E.; Woltersdorf, J.; Grabke, H. J. Mater. Corros. 126. Zheng, X. G.; Young, D. J. Oxid. Met. 1994, 42, 163.
2003, 54, 747. 127. Tomas, P.; Young, D. J.; L Trimm, D. In Proc. Int.
103. Shelton, J. C.; Patil, H. R.; Blakely, J. M. Surf. Sci. 1974, Congress Met. Corros. National Research Council of
43, 493. Canada: Ottawa, 1984; Vol. 1; p 58.
104. Yang, R. T.; Chen, J. P. J. Catal. 1972, 26, 51. 128. Mitchell, D. R. G.; Young, D. J.; Kleeman, W. Corrosion
105. Grabke, H. J.; Krajak, R.; Muller-Lorenz, E. M.; 92; NACE: Houston, TX, 1992; paper 302.
Strauss, S. Mater. Corros. 1996, 47, 495. 129. Dillinger, L.; Buchheit, R. D.; Van Echo, J. A.; Roach,
106. Zhang, J.; Cole, D. M. I.; Young, D. J. Mater. Corros. D. B.; Hall, A. M. Microstructures of Heat-Resistant-
2005, 56, 756. Alloys; Steel Founders’ Society of America: Rocky River,
107. Nishiyama, Y.; Otsuka, N. Mater. Sci. Forum 2006, OH, 1970.
522–523, 581. 130. Young, D. J. Mater. Sci. Forum 2006, 522–523, 15.
108. Bernardo, C. A.; Alstrup, I.; Rostrup-Nielsen, J. R. J. 131. Harper, M. A.; Ducasse, M.; Young, D. J. Corrosion 1995,
Catal. 1985, 96, 517. 51, 191.
109. Alstrup; Tavares, M. T.; Bernardo, C. A.; Sorensen, O.; 132. Grabke, H. J.; Peterson, E. M.; Srinivasan, S. R. Surface
Rostrup-Nielsen, J. R. Mater. Corros. 1998, 49, 367. Sci. 1977, 67, 501.
110. Tavares, M. T.; Alstrup, I.; Bernardo, C. A. Mater. Corros. 133. Grabke, H. J.; Moller, R.; Schnaas, A. Werkst. Korros.
1999, 50, 681. 1979, 30, 794.
111. Dalmon, J. A.; Martin, G. A. J. Catal. 1998, 49, 367. 134. Schneider; Viefhaus, H.; Inden, G.; Grabke, H. J.;
112. DKI German Copper Institute Booklet: Copper nickel Muller-Lorenz, E. M. Mater. Corros. 1998, 49, 336.
alloys: Properties, Processing, Application. http://www. 135. Southwell, G.; MacAlpine, S.; Young, D. J. J. Mater. Sci.
copper.org/applications/cuni/txt_DKI.html. Eng. 1987, 88, 81.
113. Mclellan, R. B.; Chraska, P. Mater. Sci. Eng. 1970, 6, 136. Ganser, B.; Wynns, K. A.; Kurlekar, A. Mater. Corros.
176. 1999, 50, 700.
114. Grabke, H. J.; Muller-Lorenz, E. M.; Klower, J.; 137. Rapp, R. A. Corrosion 89; NACE: Houston, TX, 1989;
Agarwal, D. C. Mater. Performance 1998, 37, 58. paper 532.
115. Maier, M.; Norton, J. F.; Puschek, P. Mater. High Temp. 138. Streiff, R. In Elevated Temperature Coatings: Science
2000, 17, 347. and Technology II; Dahotre, N. B., Hampikian, J. M.,
116. Perkins, R. A.; Coons, W. C.; Radd, F. J. In Properties of Eds.; TMS, 1996; p 407.
High Temperature Alloys; Foroulis, Z. A., Pettit, F. S., 139. Wachtell, R. L. In Science and Technology of Surface
Eds.; Electrochemical Society, Inc.: Princeton, NJ, 1976; Coating; Chapman, B. N., Anderson, J. C., Eds.;
p 733. Academic Press: London, 1974; p 105.
117. Maier, M.; Norton, J. F. Corrosion 99; NACE: Houston, 140. Kung, S. C.; Rapp, R. A. Oxid. Met. 1989, 32, 89.
TX, 1999; paper 75. 141. Schutze, M. German Patent: DE 197 43 421.5 (1998).
118. Zeng, Z.; Natesan, K.; Grimoditch, M. Corrosion 2004, 142. Rozado, C.; Schutze, M. Mater. Corros. 2003, 54, 831.
60, 632.
1.13 Nitridation of Alloys
U. Krupp
Faculty of Engineering and Computer Sciences, University of Applied Sciences Osnabrück, Albrechtstraße 30, 49076 Osnabrück,
Germany

1.13.2 Thermodynamics and Kinetics of Nitridation Processes 305
1.13.2.1 Nitrogen-Containing Gas Atmospheres 305
1.13.2.2 Diffusion-Controlled Internal Nitridation 306
1.13.2.3 Influence of the Chemical Composition – Thermodynamics of Internal Nitridation 308
1.13.2.4 The Transition from Internal to External Nitridation 309
1.13.3 Internal Nitridation of Heat-Resistant Alloys 310
1.13.3.1 Internal Nitridation of Fe- and Ni-Based Superalloys 310
1.13.3.2 Mechanical Effects on Internal Nitridation Kinetics 311
1.13.3.3 Protection Against Internal Nitridation Attack 312
1.13.4 Modelling and Computer Simulation of Internal Nitridation 313
References 315
p(O2) Oxygen partial pressure (Pa)

Abbreviations R Universal gas constant (J mol1 K1)
CALPHAD Calculation of phase diagrams
t Exposure time (s)
CSL Coincident site lattice
T Temperature (K, C)
x, y Coordinates (m)
Vi Molar volume of a species i (m3 mol1)
DG0 Standard free energy of formation (J mol1)
Symbols « Labyrinth factor
ci Concentration of a species i
g Parameter
cjB Concentration of the nitride-forming alloying
n Stoichiometric coefficient
element B at the reaction front x
j Depth of the internal reaction front (precipitation
c0B Initial concentration of the nitride-forming
depth) (m)
alloying element B in the alloy
c0N Nitrogen concentration at the surface (can be
set equal to the max. nitrogen solubility in the
alloy)
1.13.1 Introduction
cjN Nitrogen concentration at the reaction front x
Di Diffusivity of a species i (m2 s1)
Nitridation of alloys refers mostly to a kind of surface-
g* Critical volume fraction for the transition from
modification process applied to high-strength steels,
internal to external oxidation/nitridation
to improve wear and fatigue resistance (case harden-
jbulk Diffusive flux through the bulk (m2 s1)
ing).1 Here, the nitridation process is carried out in an
jif Diffusive flux along interfaces (precipitates, grain
ammonia-containing atmosphere, exhibiting high
or phase boundaries) (m2 s1)
nitrogen activity, leading to the formation of Fe
jprec Diffusive flux through the precipitates (m2 s1)
nitrides, as well as V, Cr, and Al nitrides, in the case
of alloy steels. However, nitrogen-based gas atmo-
LSP Solubility product
spheres are often considered inert and used for weld-
P Permeability (m2 s1)
ing and heat-treatment purposes. Nitridation as a kind
p(H2) Hydrogen partial pressure (Pa)
of high temperature corrosion becomes relevant for
p(NH3) Ammonia partial pressure (Pa)
technical components operating at temperatures well
p(N2) Nitrogen partial pressure (Pa)
above 800 C, for example, gas-turbine blades or tubes
304
Nitridation of Alloys 305
in power generation and chemical industry. At these activities up to the range of aN ¼ 1010. The resulting
temperatures, dissociation of N2 into atomic nitrogen atomic-nitrogen concentration at the metal surface
at metal surfaces is sufficient, so that alloying elements can be calculated using the following equations2:
with a high affinity to nitrogen, such as Ti, Al, and Cr, qffiffiffiffiffiffiffiffiffiffiffiffi
1
can form nitrides. Provided that diffusive penetration N2 , Ndissolved ; cðNdissolved Þ ¼ K pðN2 Þ ½1
2
of nitrogen is fast, the nitrides form internally; it is a 3
NH3 , H2 þ Ndissolved ;
form of corrosive attack that is very similar to internal 2
oxidation and has been termed internal nitridation.2 pðNH3 Þ
Figure 1 shows an example of internal nitridation cðNdissolved Þ ¼ K ½2
pðH2 Þ3=2
leading to failure of a natural-gas-fired burner tube
made of the Ni-base alloy 601. It should be mentioned that at moderate and low tem-
In most technical cases, internal nitridation has peratures atomic nitrogen (eqn [2]) recombines to form
been observed in heat-resistant alloys together with molecular N2. Hence, nitridation requires either a high
superficial oxide-scale formation and internal oxida- flux of nitriding gas or high temperatures.
tion. Hence, quantitative prediction of the degree of The relevant nitride-forming alloying elements Ti,
nitridation attack requires a thorough analysis of the Al, and Cr are at the same time prone to oxidation.
relevant diffusive transport mechanisms and the ther- The formation of the respective nitrides or oxides
modynamic stabilities of the species participating in depends on the standard free energy of formation
the high temperature corrosion processes. The rela- (cf. Section 1.13.2.3) and the relationship between
tionship between the gas atmosphere and the forma- the nitrogen and oxygen partial pressures. With the
tion of metal nitrides is discussed in Section 1.13.2, help of thermodynamic data for the stability of the
while Section 1.13.3 focuses on some implications different oxide and nitride compounds, this relation-
during internal corrosion reactions occurring during ship can be represented by means of stability diagrams.
the operation of components made of technical heat- The competitive situation between oxidation and
resisting alloys. Section 1.13.4 provides an insight nitridation is prevalent during high temperature exp-
into applying computer modeling to complex inter- osure to gases containing both oxygen and nitrogen.
nal-corrosion processes. This is the case for air and for nearly all combustion
atmospheres. In the case of cracks in thermally-grown
oxide (TGO) scale and/or metal substrate, the nitro-
1.13.2 Thermodynamics and Kinetics gen activity can reach much higher values than the
of Nitridation Processes oxygen activity, which is reduced due to oxidation.
1.13.2.1 Nitrogen-Containing Gas This can lead to situations where locally the p(N2)/p
Atmospheres (O2) ratio reaches values at which Al, Ti, and Cr form
nitrides instead of oxides.3,4 An example of such a
The risk of nitridation of alloys is prevalent in situation is shown in Figure 2.
atmospheres with high nitrogen activities aN, that is,
nitrogen-based atmospheres (N2) or, in particular,
ammonia-bearing atmospheres (NH3) with nitrogen
Cr2O3
Al2O3
Al2O3
AIN TiN
20 μm 5 μm
Figure 1 Internal-oxidation and -nitridation attack of a Figure 2 Crack in the superalloy CMSX-6 caused
failed natural-gas burner tube of alloy 601 operated at by fatigue at 950 C showing internal oxidation and
T ¼ 1100 C. nitridation.
Concentration c Internal nitridation

depth x
CB0
CB
Alloying element B
C NS DN DB
Nitrogen
CN
Alloy AB
B νN
Superficial
oxide scale Distance from surface x
(with cracks) Reaction front
Figure 3 Concentration profiles during internal precipitation of BnN according to Wagner’s classical theory of internal
oxidation.
1.13.2.2 Diffusion-Controlled Internal solubility product LSP ¼ cB cN of the respective nitride

Nitridation compound BnN (cf. Figure 3 and Section 1.13.2.3)
x pffiffiffiffiffiffiffiffi
Provided that the degree of N2 dissociation at the metal cNs
cN
cN ðx; t Þ ¼ cN
s
erf x=2 DN t ½5
surface is fast enough, that is, p(N2) can be assumed to be erf g x
equal to the nitrogen concentration in the gas phase, the cB0 cB pffiffiffiffiffiffiffiffi
cB ðx; t Þ ¼ cB0 qffiffiffiffiffi erfc x=2 DB t ½6
nitridation process is determined by inward diffusion of erfc g DDNB
nitrogen into the alloy and outward diffusion of the
nitride-forming alloying element, cf. Figure 3. Under
and taking into account that the diffusive flux of nitro-
these conditions, the nitrogen concentration at the sur-
gen and the nitride-forming alloying element at the
face is constant (cN ðx ¼ 0Þ ¼ cN s
), while the concentra-
reaction front must be equivalent (according to the
tion of the nitride-forming alloying element far away
stoichiometry BnN).
from the internal-nitridation zone is always equal to the " #

nominal bulk concentration (cB ðx > xÞ ¼ cB0 ). In the @cN @cB
most simple case of just one nitride compound pre- lim DO ¼ nDB ½7
"!0 @x x ¼ x" @x x ¼ x þ "
cipitating in a specimen or component where two- or
three-dimensional geometrical features are negligi- For nitride compounds of high stability, like the
x
ble, the nitridation process can be described by Fick’s common ones TiN and AlN, the concentrations cN
x
second law of diffusion with cN, cB, DN, DB being and cB can be neglected, that is, LSP ¼ 0 (type I
the concentrations and diffusivities of nitrogen N in behavior). Furthermore, interstitial nitrogen diffusion
the alloy and the nitride-forming alloying element B, (cf. Table 1) should be by orders of magnitude faster
respectively: than substitutional diffusion of the metallic alloying
elements, that is, DB =DN cNs
=cB0 1. According to
@cN=B @ 2 cN=B Wagner’s theory of internal oxidation, the internal-
¼ DN=B ½3
@t @x 2 nitridation depth can be calculated as
s
In general, internal nitridation can be treated in an cN DN
x2 ðt Þ ¼ 2 t ½8
analogous way to internal oxidation, where the classical ncB0
|fflffl{zfflffl}
theory of Carl Wagner5 can be applied. Assuming para- kN
bolic progress of the internal-nitridation depth x, that is,
pffiffiffiffiffiffiffiffi with kN being the nitridation constant. Even though
xðt Þ ¼ 2g DN t ½4 eqn [8] provides a good estimate for many internal
carburization, oxidation, and nitridation phenomena,
the diffusion equation [3] can be coupled with the con- technical corrosion processes involve mostly simulta-
x
centrations cN and cBx at the reaction front x based on the neous formation of various compounds of different
thermodynamic stability,6,7 for example, in NiCrAlTi The assumption that internal nitridation is gov-
alloys, the progress of the internal TiN precipitation erned by bulk diffusion of nitrogen according to
front is governed by the consumption of nitrogen due to Fick’s second law is an oversimplification. As demon-
CrN/Cr2N and AlN formation and the surface nitro- strated by Stott and Wood,13 for internal oxidation,
s
gen concentration cN , which might be decreasing the diffusive flux j of nitrogen through the precipita-
because of the possible growth of an outer Al2O3 tion zone should be threefold (cf. Figure 4): bulk
and/or Cr2O3 scale. This is schematically represented diffusion through the solvent jbulk, diffusion through
in Figure 4. the precipitates jprec (should be negligible), and dif-
The problem becomes more complex when fusion along the precipitate/bulk interfaces or grain
nitrides of moderate or low thermodynamic stability boundaries jif . Since the flux contributions have to be
like CrN, Cr2N, or Fe nitrides have to be accounted multiplied by the respective diffusion area, the block-
for. For such situations, where the solubility product ing effect of the precipitates (jprec ffi 0) plays
LSP is considerably high (eqn [9]), the theory of an important role for the transition from internal
Laflamme and Morral8 (type II behavior) or numerical nitridation to external nitride-scale formation (cf.
approaches9,10 (Section 1.13.4) should be considered. Section 1.13.2.4). A phenomenological approach to
account for the flux-blocking effect of the internal
0 < LSP =cN cB < 1
s 0
½9
precipitates was suggested by Schnaas and Grabke14
In most studies, nitrogen diffusion coefficients were using a labyrinth factor e in eqn [8]:
not determined directly, but by evaluating nitridation "cN
s
DN
kinetics,11 for example, experimental determination of x2 ðt Þ ¼ 2 t ½10
ncB0
the permeability P ¼ cN DN by measuring the precip-
itation depth x12 (cf. eqn [8]). Diffusion data for nitro- Douglass11 reports diffusion coefficients for nitrogen
gen in Ni and g-Fe alloys are given in Table 1. in nickel being two orders of magnitude higher than
for oxygen and attributes this observation to trapping
Table 1 Diffusion coefficients (m2 s1) for nitrogen in Ni effects, where a higher binding energy between oxy-
alloys and Fe–20Ni gen atoms and the traps could exist.
N in Ni alloys N in Fe–20Ni It has been observed in several studies15,16 that
13 12
internal nitridation (similar to internal oxidation17–19)

700 C 9.510 to 2.310 1.21012
does not only imply diffusion of the reacting species,
800 C 3.21012 to 8.51012 3.91012
900 C 1.41011 to 4.01011 1.51011 but also mechanical stresses arising from the higher
1000 C 2.21011 (24) – specific volumes of the nitride compounds as com-
pared to the substrate. This leads to a flux of the
Source: Douglass, D. L. J. Met. 1991, 11, 74–79. nonreacting alloying elements towards the surface,
N, O
Bulk diffusion j bulk
Cr, Al
Interface diffusion j if
(precipitates, grain boundaries)
Precipitate diffusion j prec
(negligible)
Gas phase Al2O3/ Al2O3 AlN AlN + Cr2N AlN Alloy
(air, 1100 ⬚C) Cr2O3
CO
CN C Cr
C Al
Figure 4 Schematic representation of the concentration profiles during oxidation and nitridation of Ni–Cr–Al alloys6 and
relevant diffusion fluxes.13
50
Fe4N Mo2N
0
CrN Si3N4
Cr2N
Standard free energy of formation ΔG (kJ mol–1 N2)

–50 NbN
–100 BN
AlN
–150
TaN
–200
–250
20 mm –300
–350
Figure 5 Formation of solvent surface protrusions after
TiN
internal nitridation of Ni–10Cr–2Ti for 150 h at 1100 C in –400
low-oxygen N2–45% He–5% H2 atmosphere. ZrN
–450
–500
where they form protrusions (Figure 5). Hence, –550

internal nitridation kinetics must be influenced by –600
additional factors like Nabarro–Herring creep,18 dis- –650
location pipe diffusion,18 and/or grain-boundary
sliding.19 (a) 0 500 1000 1500 2000
Temperature (⬚C)
CrN
Cr2N 100
Cr Cr2O3 Environment
1.13.2.3 Influence of the Chemical TiN
10–10
Composition – Thermodynamics of AlN
Possible reaction path
Internal Nitridation Al
10–20 for internal nitridation
Al2O3
p(N2) (bar)
Although diffusive fluxes are relevant for the lateral TiO2
10–30 Ti
progress of the reaction front, the kind of nitride that
is formed depends on thermodynamic-stability con-
ditions: local thermodynamic equilibrium can be 10–40 10–30 10–20 10–10 100
(b) p(O2) (bar)
assumed in cases of sufficiently high reaction rates.
Figure 6(a)20 represents the condition for thermody- Figure 6 (a) Standard free energy diagram of selected
namic equilibrium in terms of the standard free nitridation reactions and (b) stability diagram for nitridation
energy of formation DG0, which is correlated with and oxidation in N2/O2 gas mixtures (1000 C).
the equilibrium nitrogen partial pressure p(N2):
DG0
lnpðN2 Þequ ¼ ½11 for example, in the software packages FactSage™,
RT
ChemApp™, and ThermoCalc™, is applied.
In addition to the technically most relevant nitrides In general, from the thermodynamic point of view,
Fe4N, e-Fe2–3N, TiN, AlN, CrN, and Cr2N, in Ni- the susceptibility to nitridation depends on (i) the
based alloys with high Cr contents, the ternary p concentration of the nitride-forming elements, and,
phase can also occur.21–23 The p phase of a stoichi- in particular, on (ii) the nitrogen concentration being
ometry of Cr10Ni7N3 is formed by a peritectic reac- dissolved in the alloy. Although the nitrogen solubil-
tion of Cr2N with Ni.22 An estimation of the nitrides/ ity in steels at elevated temperatures is rather high (in
oxides being formed for given nitrogen and oxygen the order of 101 at.%), it drops substantially when
partial pressures p(O2) and p(N2), respectively, is increasing the Ni or Co content in the alloys. This
possible with the help of thermodynamic stability was shown by Barnes and Lai2 experimentally and
diagrams. An example for the system Ni–Cr–Al– by Krupp et al.12,24 by means of CALPHAD calcula-
Ti–N–O is shown in Figure 6(b). To account for tions, using the thermodynamic software ChemApp™.
complex higher-order interaction terms determining It should be mentioned that the solubility data
the internal-corrosion behavior of complex techni- reported for nitrogen in Ni alloys show large scatter,
cal alloys, the CALPHAD method, implemented, ranging from 0.0015 at.% (Wriedt and Gonzalez25 for
Nitrogen absorption (mg cm–2)

15
TiN
10
0
0 20 40 60 80
(a) Ni + Co content (wt%)
(a)
100 μm
0.010
Dissolved N concentration (at.%)
0.008 1100 ⬚C
15
0.006
5
0.004 TiN
1000 ⬚C
0.002
900 ⬚C
0
0 10 20 25 30 35 (b) 100 μm
(b) Cr content (wt%)
Figure 7 Effect of the chemical composition of (a) the
nitridation resistance of Ni-, Fe-, and Co-base alloys2 and
(b) the nitrogen solubility in Ni–xCr alloys (ChemApp™
calculation according to Gorr et al.24). TiN
pure Ni) and 0.004 at.% (Chang26 derived from mea-

surements (LECO nitrogen analyzer) for Ni–10%Cr)
to 0.16 at.% (Allen and Douglass for Ni–V27 alloys).
(c) 100 μm
Figure 7 shows data for the nitrogen concentration
absorbed or dissolved in Ni-containing alloys.
Figure 8 Internal-nitridation attack (TiN) of Ni–xCr–yTi
The low nitrogen solubility in Ni-based alloys, model alloys after 100 h exposure at 1100 C to low-oxygen
which can be assumed to be equal to the maximum N2–45%He–5%H2 atmosphere: (a) 5 wt% Cr–2 wt% Ti,
s (b) 20 wt% Cr–2 wt% Ti, (c) 20 wt% Cr–6 wt% Ti.12
N concentration in the alloy surface layer (cN ), is
directly related to the susceptibility to internal nitri-
dation, where, according to eqn [8], the precipita-
s
tion depth depends on the permeability cN DN, the
exposure time t, and the concentration of the nitride-
forming alloying element cB0 . These relationships are approximately equal to the original mole fraction cB0
represented by results obtained for the nitridation of the nitride-forming alloying element B. However,
of various Ni–xCr–yTi model alloys,12 as shown in if the diffusivity of B is not negligible as compared
Figure 8. to the nitrogen diffusivity, that is, DN cN s
DB cB0 ,
then the mole fraction of nitrides BnN in the
internal-nitridation zone becomes substantially
higher than cB0 . According to Wagner,5 coalescence
1.13.2.4 The Transition from Internal to
of the nitride precipitates into a compact nitride
External Nitridation
scale can be expected once a critical volume fraction g*
Assuming eqn [8] to be valid, the mole fraction of of the nitrides (typical values: g ¼ 0:2; . . . ; 0:3)
nitrides BnN in the internal-nitridation zone is is exceeded, corresponding to a solute concentration
1.13.3 Internal Nitridation of

Heat-Resistant Alloys
1.13.3.1 Internal Nitridation of Fe- and
Ni-Based Superalloys
Because of the very low nitrogen solubility in Ni
TiN
compared to Fe, Ni-based alloys, for example, alloy
600 and 601,2,28 exhibit a substantially higher resis-
tance to internal nitridation than Fe-based alloys, such
as alloy 800/800H or AISI 314.2,29 In the case of NH3-
5 μm
containing atmospheres, the high temperature corro-
(a) sion behavior of both, Ni- and Fe-based alloys can be
catastrophic, generally leading to external and/or
internal nitridation attack.30 Very high nitrogen activ-
ities in combination with some amount of oxygen
TiN
lead to the formation of nonprotective porous
oxide–nitride scales. According to Strauß et al.,31 in
low- and high-alloy Cr steels, sponge-like scales of
Fe3O4 or FeCr2O4 are formed in 70% NH3–30%
H2O atmosphere being obviously permeable for NH3
with the consequence of strong internal nitridation in
the form of needle-like Cr nitrides. Because of the high
specific volume of the Cr nitrides, compressive stresses
1 μm are induced in the internal-nitridation zone, leading to
(b)
intercrystalline cracks and a disintegration of the poly-
Figure 9 Internal-nitridation attack (TiN) of Ni–yTi crystalline microstructure. Ni alloys may form very
model alloys after 800 h exposure at 900 C to low-oxygen
instable Ni3N that releases gaseous nitrogen when
N2–45% He–5% H2 atmosphere: (a) 2 wt% Ti and
(b) 6 wt% Ti. decomposing, leading to pore formation. In the case
of Cr-containing Ni-based alloys, the combination of
pores and internal Cr nitrides causes fast surface-layer
pg DN V s degradation in ammonia atmosphere.31
cB0 < c ½12 Since Cr nitrides are of only moderate thermody-
2nDB VBn N N
namic stability, their formation is limited to atmo-
for which internal nitridation instead of superficial spheres with very high nitrogen activities or alloys
nitride-scale formation occurs, with V and VBn N being containing Cr concentrations above
20 wt%, hence
the molar volumes of the alloy and the nitride com- exposure of most technical NiCr alloys to air does
pound, respectively. Hence, the critical solute con- not cause any nitridation. However, alumina-forming
centration cB0 increases with (i) increasing nitrogen superalloys, often alloyed with Ti for g0 strengthen-
s
solubility cN , (ii) increasing nitrogen diffusivity DN, ing, often show massive internal nitridation under
and (iii) decreasing solute diffusivity DB. This is the service conditions, for example, gas-turbine blades
reason why for Ni–6Ti with very low N solubi- operating in air or combustion atmosphere.32 Below
lity, superficial TiN-scale formation was observed a superficial Cr2O3 scale, the oxygen partial pressure
(Figure 9, Ni–2Ti shows little internal-nitridation drops to a value sufficient only for the formation of
attack under the same conditions), while Ni–20Cr–6Ti Al2O3 and TiO2. On the other hand, Cr2O3 is
shows deep internal-nitridation attack (cf. Figure 8). observed to be permeable to atomic nitrogen;33
Equation [12] has to be considered when design- hence the ratio p(N2)/p(O2) increases below the chro-
ing alloys to form superficial nitride scales for mia scale. For conditions where no dense Al2O3 sub-
improved wear resistance at elevated temperatures. scale can be formed because of overall or local
However, in such cases, the high specific volume depletion of the Al concentration below a critical
of the nitrides compared with that of the substrate value, a situation that is often promoted by service
alloy, needs to be taken into account (cf. Section under thermal-cycling conditions, further penetration
1.13.3.1). of nitrogen is possible, allowing stable AlN and TiN
Hence, the formation of AlN nuclei requires much

NiO + TiO2 higher supersaturation and particle growth must be
Cr2O3 partly controlled by Al diffusion in a much more
Al2O3 pronounced way than is the case for TiN precipitates.
Ni-based alloys with Cr concentrations above
TiN
20wt% are prone to the formation of a third group
γ of nitrides: Cr2N and the ternary p phase (Cr10Ni7N3,
cf. Section 1.13.2.3). The p-phase nitrides reported
by several authors21–23 are of coarse morphology
and were discussed as a possible strengthening
γ+γ⬘ phase in the bulk34 or as surface layer of heat-
10 μm resistant alloys.
Figure 10 Internal nitridation in the single-crystalline

Ni-base superalloy SRR99 after 100 h exposure at 1.13.3.2 Mechanical Effects on Internal
1000 C to air. Nitridation Kinetics
Generally, the mechanisms governing internal-
corrosion cannot be analyzed without taking mechanical
precipitates to be formed according to the local com- stresses into account. In particular, large, needle-like
bination of p(N2) and p(O2) (cf., thermodynamic stabil- precipitates act as crack initiation sites and cause an
ity diagrams, Figure 6(b)). The process of internal embrittlement of the internally-corroded surface layer.
nitridation of alumina-forming Ni- or Fe-based super- According to a study by Klöwer et al.,29 the small sub-
alloys is shown for the single-crystalline superalloy micrometer TiN precipitates are much less harmful than
SRR99, as an example, in Figure 10. the blocky AlN precipitates. However, in cases where
For superalloys with Cr concentrations not above the initial Ti concentration is rather high (cTi 2 wt%),
20 wt% and nitrogen activities below aN ¼ 1, occur- the alloy grain boundaries can be completely covered by
rence of only the most stable nitride TiN, the some- TiN, leading to intercrystalline cracking as shown for a
what less stable nitride AlN, and, depending on the nitrided model alloy in Figure 11(a). Hence, nitridation
gas and alloy composition, NbN and VN,27 is leads to a dangerous decrease in the ductility of technical
possible. high temperature alloys.
In cast Ni-based superalloys, for example, the As a second effect, the nitridation reaction is accom-
single-crystalline alloys used for the first stages gas- panied by a strong increase in the specific volume
turbine blades, the susceptibility to internal nitrida- v of the nitrides, for example, vCrN ¼ 0.17 cm3 g1,
tion varies locally due to dendritic segregations. The vAlN ¼ 0.19 cm3 g1, vTiN ¼ 0.31 cm3 g1, as com-
high nitrogen diffusivity (cf. Section 1.13.2.2) leads pared to the specific volume of Ni-base superalloys,
to a fast progress of the internal-precipitation front, vNi-base ffi 0.13 cm3 g1. This volume increase has been
which is accompanied by a dissolution of the shown by various authors to cause damage in the form
strengthening g0 phase (Ni3(AL,Ti)) due to the nitri- of cracks or grain disintegration14,31 and/or to drive
dation (and oxidation-) caused consumption of the outward solvent diffusion leading to substrate–metal
g0 -forming alloying elements Al and Ti (see Figure 10). protrusions.15–19 Figure 11(b) shows the tip of a
A particular deep dissolution can be observed in wedge-shaped specimen of the Ni-based superalloy
alloys where the g0 phase is stabilized by a high Ti CMSX-4 after thermal cycling at 1100 C (1 h dwell
content (cTi >2 wt%). times interrupted by 15 min cooling periods).35 The
Even though it can be assumed that internal nitri- combination of geometry-induced growth stresses and
dation of superalloys is controlled by nitrogen diffu- differences in the coefficient of thermal expansion
sion, the parabolic rate law was shown in an explicit (CTEoxideCTEalloy) led to repeated spalling of the
way only for model systems or in pure nitriding alumina scale, near-surface depletion of the Al reser-
gas atmospheres.6,12 It can be assumed that the mor- voir, and, eventually, strong internal-oxidation and
phology of the internal reaction products itself is nitridation attacks in combination with cracks and fast
of significance for the precipitation kinetics. The oxidation of the solvent (‘breakaway’ oxidation).
AlN precipitates are by orders of magnitude larger As mentioned earlier, stress-driven and stress-
than the TiN precipitates (see Figures 10 and 14). enhanced diffusion contributes vitally to the kinetics
TiN
AIN 10 μm
TiN
200 μm
(a)
(a) 10 μm
Internal oxidation/
nitridation
Al2O3
NiO
TiN
TiN
Non-protective
Cracks 20 μm
‘breakaway’ oxidation
(b)
30 MPa 10 μm
Figure 11 Cracks in an internally nitrided surface layer in (b)
alloy 80A (100 h, 1000 C, low-oxygen N2–45% He–5% H2
atmosphere) due to tensile testing and (b) ‘breakaway Figure 12 Internal nitridation within (a) a mechanically
oxidation’ and internal oxidation/nitridation of a wedge- unloaded specimen and (b) a creep-loaded specimen
shaped specimen of the single-crystalline Ni-base (s ¼ 30 MPa) of the Ni-base superalloy alloy 80A exposed at
superalloy CMSX-4 after thermal cycling exposure at T ¼ 900 C for t ¼ 48 h to low-oxygen N2–45% He–5% H2
1100 C – 1 h dwell times in air. atmosphere.
of internal nitridation. Figure 5 shows NiCr protru-

sions on top of an internally nitrided Ni–10Cr–2Ti very low oxygen partial pressures in the range of
alloy. Chang et al.15 could prove the direct relationship p(O2)<1023 bar, a very thin alumina scale of sub-
between the depth of internal TiN precipitation and micrometer thickness is formed, which is not able
the degree of coverage by surface protrusions. The to block nitrogen permeation.4 Even though preox-
most probable mechanism for the respective outward idation of an alumina-forming alloy or any other
transport of solvent has been assumed to be dislocation- alloy that has been aluminized could prevent inter-
pipe diffusion15,18 rather than Nabarro–Herring creep nal oxidation, there is still the risk of local spalling
or grain-boundary sliding. Furthermore, it was shown and/or cracking of the scale. If under service con-
by Krupp et al.36 and Dong et al.37 that the internal- ditions no healing layer can be formed, such a
nitridation rate was substantially increased when super- situation would cause a transition to internal oxi-
imposing a mechanical creep load (Figure 12); again, dation and nitridation.
this can be attributed to dislocation-pipe diffusion, As a second possibility to protect a material against
contributing to internal precipitation when the disloca- internal nitridation, nitride-forming alloying elements,
tion density is increasing due to dislocation creep. such as Ti and Al, should be avoided. However, this is
difficult because of the high significance of Al for
alumina-scale formation and g0 stabilization. The lat-
1.13.3.3 Protection Against Internal
ter effect, g0 destabilization because of Al and Ti
Nitridation Attack
reduction, can be partly accommodated by adding
In the literature on high temperature corrosion of Ta and Nb, though the latter also is a nitride former.27
technical superalloys, there is agreement that a dense Similar to many alloying elements in high-alloyed
and adherent Al2O3 scale provides the most effective Ni-based superalloys, the risk of brittle-phase forma-
protection against internal nitridation. Nitrogen per- tion, for example, by high Nb concentrations, needs to
meation is obviously very slow. However, it should be be taken into account when considering new alloy
mentioned that in a nitrogen-based atmosphere with compositions. Attempts to replace Al as the most
effective TGO former at very high temperatures, for ChemApp

example, by Si, has never been successful; hence, even nitride precipitation
for coated alloys, a minimum Al concentration for Gibbs free energy G = min
c(x,y ,t +Δt )
healing-layer formation must be maintained.
In the case of polycrystalline alloys, the suscepti-
bility to internal-oxidation and internal-nitridation
attack might be reduced by grain-boundary-
engineering-type processing. It was shown for oxida-
tion of alloy 718 in air that by increasing the fraction t +Δt Δyr
of low-energy CSL boundaries (coincident site lat- y
tice) the oxidation rate was reduced.38 There is a Δyl
chance that a combination of near-surface grain t
refinement and high temperature exposure could
Δx l x Δx r
(i) reduce the ratio between the inward oxygen flux
and the outward aluminum flux, that is, promoting Figure 13 Schematic representation of the
superficial Al2O3 formation even at low initial Al two-dimensional finite-difference mesh (Crank–Nicolson
concentrations, and (ii) reduce the inward nitrogen scheme) for simulation of internal-corrosion processes in
flux by an increasing number of CSL boundaries combination with ChemApp™.
exhibiting low grain-boundary diffusivity.
according to the Crank–Nicolson scheme (Figure 13,
l ¼ left-hand side, r ¼ right-hand side)40.
1.13.4 Modelling and Computer cðx; y; t þ Dt Þ cðx; y; t Þ
Simulation of Internal Nitridation ¼
Dt

Dx ðx; yÞ cðx Dx; y; t Þ 2cðx; y; t Þ þ cðx þ Dx; y; t Þ
According to the preceding sections, internal nitrida- 2 Dxl ðx; yÞ Dxr ðx; yÞ

tion is governed by solid-state diffusion of the react- cðx Dx; y; t þ Dt Þ 2cðx; y; t þ Dt Þ þ cðx þ Dx; y; t þ Dt Þ
þ
ing metallic and nonmetallic species (i) through a Dxl ðx; yÞ Dxr ðx; yÞ

superficial oxide scale, (ii) through an internal- Dy ðx; yÞ cðx; y Dy; t Þ 2cðx; y; t Þ þ cðx; y þ Dy; t Þ
þ
precipitation zone, and (iii) through the substrate 2 Dyl ðx; yÞ Dyr ðx; yÞ

alloy. In general, this problem can be treated by cðx; y Dy; t þ Dt Þ 2cðx; y; t þ Dt Þ þ cðx; y þ Dy; t þ Dt Þ
þ
the diffusion differential equation for all the species Dyl ðx; yÞ Dyr ðx; yÞ
participating in the nitridation process, using either ½14
the gradients in the chemical potential or, in a sim-
plified way, the concentration gradients.9,10,39 In the Equation [14] applied simultaneously to all the species
latter case, the equations need to be coupled with a involved in the internal-nitridation process allows
mass balance and a Gibbs-energy criterion, account- the stepwise solution of two-dimensional diffusion
ing for the local solubility product, that is, the con- problems in a two-dimensional area with location-
centrations of the diffusing species in thermodynamic dependent diffusion coefficients accounting for the
equilibrium. different diffusion paths, that is, bulk diffusion, grain-
Possible ways to solve the diffusion equation boundary-diffusion, etc. Coupling of the equations
can be established by the integrable thermodynamics
@c
¼ rðDrcÞ ½13 interface ChemApp™, which represents the ther-
@t modynamic software FactSage™. The interface recal-
numerically are provided by the finite-difference and culates the set of concentrations at a certain time step
the finite-element method. Although the latter one is of according to thermodynamic equilibrium until the
a large flexibility with respect to diffusion in complex diffusion calculation continues.
geometries, the finite-difference approach is easier to The numerical approach can be applied to solve
handle and to link with computational thermodynam- complex internal-nitridation phenomena, which cannot
ics. Here, the derivatives in eqn [13] (simplified for be treated by Wagner’s classical theory of internal oxi-
two-dimensional diffusion problems) are replaced by dation, for example, (i) internal precipitation of com-
difference quotients using time steps of a length Dt and pounds of moderate thermodynamic stability such as
location steps of flexible widths Dx and Dy, respectively, Cr nitrides (cf. Figure 14(a)), (ii) simultaneous internal
CCr,CrN CN
(at. %) (at. %)
25 0.0006
20 0.0005
Cr
0.0004
15
0.0003
10 CrN 0.0002
5 N
0.0001
CrN
10 μm 0 0.0000
0 10 20 30 40 50
(a) Distance from surface (μm)
CTi,Al,TiN,AIN CN
(at. %) (at. %)
6 0.003
AlN
AIN 5 Al 0.0025
4 N 0.002
3 Ti 0.0015
TiN
TiN
2 0.001
1 0.0005
20 μm 0 0
0 100 200 300 400 500
(b) Distance from surface (μm)
Figure 14 Cross-sectional view and simulated concentration profiles of (a) internal precipitation of CrN in Ni–20Cr (800 C,
200 h) and (b) simultaneous internal precipitation of TiN and AlN in Ni–20Cr–2Al–2Ti (1100 C, 150 h) in both cases exposure
to low-oxygen N2–45% He–5% H2 atmosphere.
and/or intergranular precipitation of more than one nitrogen affinity, such as Al, Ti, Cr, Nb, and V, imply
nitride and/or oxide compound (cf. Figure 14(b)), superimposition of oxidizing and nitriding environ-
or (iii) internal corrosion as a consequence of local dam- ments. Hence, whether a superficial oxide scale can
age of a superficial oxide scale by cracking or spalling. protect the substrate against internal corrosion attack
depends strongly on the mechanical integrity of the
scale and the ability to form a healing layer by means
1.13.5 Concluding Remarks of sufficient supply of the oxide-forming alloying
element to the metal surface. Here, the critical factors
Although for many technical applications nitridation are (i) the local ratio between the oxygen and the
treatments are used to produce wear-resistant surface nitrogen activity, and (ii) the solute concentration
layers on steels, in high temperature corrosion, the being required for transition from internal oxidation
term nitridation refers to a deep embrittlement attack to external scale formation. These factors may change
caused by nitride compounds. Nitridation occurs not when mechanical stresses are superimposed, which
only in ammonia atmospheres with a high nitrogen arise due to thermal cycling or due to volume
activity but also in pure nitrogen or even in air changes caused by internal precipitation. The mutual
or combustion atmospheres at temperatures above interactions between the diffusivity of the reacting
800 C, where N2 dissociation is fast enough to species through a superficial oxide scale and the
allow atomic nitrogen to penetrate into the metal metallic bulk, the thermodynamics and nucleation
surface by solid-state diffusion. The nitridation rate kinetics of nitrides, and the parameters caused by
depends mainly on the diffusivity of nitrogen in the the service conditions, that is, temperature changes,
alloy and the thermodynamic stability of the nitrides composition of the gas phase, etc., are complex and
being formed internally; the precipitation depth as a only touched upon in this chapter. A thorough anal-
function of exposure time follows in many cases a ysis is not possible by means of an analytical
parabolic rate law as was proposed by Wagner5 for approach; it requires a numerical treatment taking
internal oxidation. Most technical applications of the various aspects of solid-state reactions into
heat-resistant alloys with alloying elements of high account.
Acknowledgments 17. Douglass, D. L. Oxidat. Met 1995, 44, 81–111.

18. Yi, C.; Guan, S. W.; Smeltzer, W. W.; Petric, A. Acta Met.
Mater. 1994, 42, 981–990.
The author acknowledges the experimental and 19. Stott, F. H.; Shida, Y.; Whittle, D. B.; Wood, G. C.;
theoretical work of Dr. S.Y. Chang, Dr. R. Orosz, Bastow, B. D. Oxidat. Met. 1982, 18, 127.
20. Bürgel, R. Handbuch Hochtemperatur-Werkstofftechnik;
Dr. V. Braz da Trindade Filho, and Dr. U. Buschmann, Vieweg: Braunschweig Wiesbaden, 1998.
as well as the strong support by Prof. Dr. H.-J. Christ, 21. Krupp, U.; Chang, S. Y.; Christ, H.-J. Z. Metallkunde,
head of Institut für Werkstofftechnik of the University 2000, 12, 2006–2012.
22. Kodentsov, A. A.; Van Dal, M. J. H.; Cserhati, C.;
of Siegen, Germany, where a great part of the research Daroczi, L. Acta Mater. 1999, 47, 3169–3180.
work discussed in this chapter was carried out. The 23. Ono, N.; Kajihara, M.; Kikuchi, M. Met. Trans. A 1992, 23,
research was financially supported by Deutsche For- 1389–1393.
24. Gorr, S.; Chang, S.Y.; Krupp, U.; Christ, H.-J. Unpublished
schungsgemeinschaft and the European Commission, data.
which is gratefully acknowledged. 25. Wriedt, H. A.; Gonzalez, O. D. Trans. AIME 1961, 221, 532.
26. Chang, S. Y., PhD thesis, University of Siegen, 2001.
27. Allen, A. T.; Douglass, D. L. Oxidat. Met. 1999, 52,
505–536.
References 28. Tjokro, K.; Young, D. J. Oxidat. Met. 1995, 44, 453–474.
29. Klöwer, J.; Brill, U.; Rockel, M. Mater. Corros. 1997, 48,
511–517.
1. Gupton, P. S. The Heat Treatment Source Book; ASTM: 30. Rubly, R. P.; Douglass, D. L. Oxidat. Met. 1991, 35,
Metals Park, 2001. 259–278.
2. Lai, G. L. High-Temperature Corrosion of Engineering 31. Strauß, S.; Grabke, H. J.; Vogel, D. Mater. Corros. 2003,
Alloys; ASM International, Materials Park, 1990. 54, 895–902.
3. Bennett, M. J.; Nichols, J. R.; Borchardt, G.; Strehl, G. 32. Barnes, J.; Lai, G. Y. In Proceedings Corrosion and
Mater. High. Temp. 2002, 19, 117–125. Particle Erosion at High Temperatures; Srinivasan, V.,
4. Krupp, U.; Christ, H.-J. Met. Mat. Trans. 2000, 31, 47–56. Vedula, K., Eds.; TMS: Warrendale, 1989; 617.
5. Wagner, C. Z. Elektrochem. 1959, 63, 772–782. 33. Krupp, U.; Chang, S.-Y.; Christ, H.-J. In Proceedings
6. Han, S.; Young, D. J. Oxidat. Met. 2001, 55, 223–242. EFC-Workshop Life Time Modelling of High-Temperature
7. Young, D.J. In Proceedings Corrosion Science in the 21st Corrosion Processes; Schütze, M., Quadakkers, W. J.,
Century, 7–11 July 2003, Manchester, UK; J. Corros. Sci. Nichols, J. R., Eds.; EFC Publications No. 34, 2001;
Eng [online]. pp 148–164.
8. Laflamme, G. R.; Morral, J. E. Acta Met. 1978, 26, 1791. 34. Brill, U. Habilitation thesis, RWTH Aachen Technical
9. Nesbitt, J. Oxidat. Met. 1995, 44, 309–338. University Aachen, Germany, 1997.
10. Engström, A.; Höglund, L.; Ågren, J. Met. Mat. Trans. 35. Orosz, R.; Krupp, U.; Christ, H.-J. Mater. Corros. 2006, 57,
1994, 25, 1127–1134. 154–158.
11. Douglass, D. L. J. Met. 1991, 11, 74–79. 36. Krupp, U.; Orosz, R.; Christ, H.-J.; Buschmann, U.;
12. Krupp, U.; Christ, H.-J. Oxidat. Met. 1999, 52, 277–298. Wiechert, W. Mater. Sci. Forum 2004, 461–464, 37–44.
13. Stott, F. H.; Wood, G. C. Mater. Sci. Technol. 1988, 4, 37. Dong, J. X.; Sawada, K.; Yokokawa, K.; Abe, F. Scripta
1072–1078. Mater. 2001, 44, 2641–2646.
14. Schnaas, A.; Grabke, H. J. Werkst. Korrosion 1978, 29, 38. Krupp, U.; Wagenhuber, Ph. E. G.; Kane, W. M.;
635–644. McMahon, C. J., Jr. Mater. Sci. Technol. 2005, 21,
15. Chang, S. Y.; Krupp, U.; Christ, H.-J. Mater. Sci. Eng. 1247–1254.
2001, 301, 196–206. 39. Krupp, U.; Christ, H. J. Oxidat. Met. 1999, 52, 299–319.
16. Kodenstov, A. A.; van Dal, M. J. H.; Cserhati, C.; 40. Krupp, U.; Trindade, V. B.; Christ, H.-J.; Buschmann, U.;
Daróczi, L.; van Loo, F. J. J. Acta Mater. 1999, 11, 3169. Wiechert, W. Mater. Corros. 2006, 57, 263–268.
1.14 Corrosion in Molten Salts
M. Spiegel
Salzgitter-Mannesmann-Forschung GmbH, Ehinger Strasse 200, 47259 Duisburg, Germany

1.14.2 Principles and Mechanisms of Corrosion in Sulfate Melts 317
1.14.2.1 Thermodynamic Considerations 317
1.14.2.2 Solubility of Gases in Sulfate Melts 317
1.14.2.3 Oxidizing and Reducing Reactions of Sulfate Melt Species 317
1.14.3 Principles and Mechanisms of Corrosion in Carbonate Melts 319
1.14.3.1 Solubility of Gases 319
1.14.3.2 Oxidizing and Reducing Reactions of Carbonate Melt Species 319
1.14.4 Solubility of Oxides in Sulfate and Carbonate Melts 320
1.14.5 Corrosion Mechanisms in Sulfate Melts: The Rapp and Goto model 322
1.14.6 Methods of Investigation of Corrosion in Molten Salts 323
1.14.6.1 Electrochemical Techniques 323
1.14.6.2 Other Methods 325
1.14.7 Examples of Hot-Salt Corrosion in Industrial Practice 325
1.14.7.1 Gas Turbines: Corrosion Beneath Na2SO4 Deposits 325
1.14.7.2 Waste Incineration: Corrosion in Chloride Melts 326
1.14.7.3 Corrosion in Molten Carbonate Fuel Cells 328
References 330
for severe corrosion-induced failures. For many years,

Abbreviations
salt deposits on gas turbines (especially Na2SO4) were
CE Counter electrode
identified as the major corrosive species that limit the
RE Reference electrode
lifetime of the turbine blades. Molten salts are also
WE Working electrode
present as electrolytes, for example, ((Li,K)2CO3) in
molten carbonate fuel cells. As the economic impact
of corrosion due to molten salts (i.e., hot corrosion) is
Symbols significant, great research efforts have been con-
ai Activity of species i ducted on the mechanisms and prevention of molten
E Voltage (V, mV) salt-induced corrosion.
I Current (A cm2) High temperature corrosion processes beneath mol-
K Equilibrium constant ten salts are generally complicated as many different
Mn+ Cation of charge n+ steps are involved. The metal is separated from the gas
n Charge phase by the melt and transport of oxidizing species
pi Partial pressure of species i from the gas phase to the metal–melt interface that
Z Complex resistivity (O) occurs. Hence, the solubility of gases in the salt is an
important factor in hot corrosion. Subsequent to diffu-
sional transport of the dissolved gas species through the
molten salt, interfacial reactions take place at the metal–
1.14.1 Introduction melt interface and corrosion products are formed, par-
tially soluble in the salt, and transported towards lower
Molten salts are present in many technical processes. concentrations by diffusion in the salt film.
As sulfate- and chloride-containing deposits on boiler The nature of the molten salts is generally ionic
tubes from conventionally-fired plants such as waste-, and the corrosion process involves electrochemical
biomass-, and coal-fired boilers, they are responsible reactions, that is, oxidation of the metal and reduction
316
Corrosion in Molten Salts 317
of melt components and dissolved gases. This requires under consideration need to be soluble in the molten
investigation by electrochemical techniques. The salt, where they can act as oxidizing agents for the
complexity of the corrosion mechanism and the intermetal and, therefore, detrimentally influence the cor-
dependence of many physical and chemical steps rosive nature of the salt.
require a systematic description of the processes Experimental investigations of the solubility of O2
involved. Due to the different chemical and electro- and SO2 in molten Na2SO4 were carried out by
chemical properties of a particular melt, it is impossi- Andresen et al.3 For O2 and SO2, a low solubility was
ble to give a general approach, as this is the case for observed, following Henry’s law. Flood and Forland2
oxidation processes. published some work on the solubility of SO3 in the
As a consequence, the following chapters look melt. At low concentration of SO3, the solubility fol-
much more deeply into corrosion mechanisms in lows Henry’s law but, at higher concentrations, devi-
sulfate and carbonate melts, as examples of very ation from Henry’s law was observed according to a
different mechanisms involved. reactive solubility of SO3 with the sulfate ion, eqn [5]:
4 þ SO3 ¼ S2 O7
SO2 ½5
2
1.14.2 Principles and Mechanisms of In summary, the solubility of O2 and SO2 in

Corrosion in Sulfate Melts Na2SO4 is negligibly low, whereas the solubility of
1.14.2.1 Thermodynamic Considerations SO3 is significantly high.
Sulfate melts contain SO2

4 oxyanions. According to
the models of Flux1 and Flood and Forland and 1.14.2.3 Oxidizing and Reducing Reactions
Motzfeld,2 the SO2 of Sulfate Melt Species
4 anion dissociates, giving SO3
and O2 ions, eqn [1]: From an electrochemical point of view, the corrosion
SO2 ¼ SO3 þ O 2
½1 process is only possible if the anodic oxidation of the
4
metal as it occurs in corrosion is forced by a comple-
þ
As cations (e.g., Na ) are also present in the melt, mentary cathodic reduction of either the melt com-
equilibrium [1] can be written as eqn [2]: ponents or the dissolved gases. In order to study the
Na2 SO4 ¼ Na2 O þ SO3 ½2 electrochemical behavior of a sulfate melt, polariza-
tion experiments in a pure Na2SO4 melt have been
2
The basicity of the melt is defined by the p(O ) performed with noble metal electrodes when the
value, eqn [3]: following anodic and cathodic reactions were identi-
pO2 ¼ log aO2 ½3 fied, eqns [6]–[11]:
Reduction (cathodic) reactions (increasing over-
According to eqn [3], a small pO2 value is given by a potential):
high O2 activity and the melt is defined to be basic.

4 þ 2e ¼ SO3 þ O
SO2 ½6
2 2
Correspondingly, a high pO2 value is given by a small
O2activity and the melt is defined to be acidic.

4 þ 6e ¼ S þ 4O
SO2 ½7
2
The overall basicity of the melt can be changed by
the gas phase composition. From eqn [2], it becomes

4 þ 8e ¼ S
SO2 þ 4O2 ½8
2
clear that:
aðO2 Þ ¼ KpðSO3 Þ1 ½4 Oxidation (anodic) reactions:
Hence, by variation of p(SO3) in the gas phase, the 1
SO2 4 ¼ SO3 þ O2 þ 2e ½9
value of a(O2) is increased. 2
Reduction of SO3 and O2, produced by equilib-
rium eqn [1], is also possible:
1.14.2.2 Solubility of Gases in Sulfate Melts
SO3 þ 2e ¼ SO2 þ O2 ½10
From eqn [4], it becomes immediately clear that the
1
melt chemistry is influenced by variations in the gas O2 þ 2e ¼ O2 ½11
phase in the atmosphere, especially by variations in 2
p(O2) and p(SO3) values. In order to reach chemical The decomposition of the sulfate as described
equilibrium through the whole melt film, the gases previously only occurs in highly basic melts under
p(SO ) + p(SO ) × 103 (atm)

2 3
2.0 4.0 6.0 8.0 10.0
Limiting current density (mA cm–2)

6
p(SO2), p(SO3)varied
p(O2) ≅ 1 atm
2
0
0 5.0 10 15 20 25
p(SO3) × 104 (atm)
Figure 1 Limiting diffusion current at Pt electrodes beneath a thin Na2SO4 melt as a function of p(SO2,SO3) at constant
p(O2) at 896 C.
p(SO3) + p(O ) •104 (atm)

showing that the limiting diffusion current increases
2
0 1 2 3 4 with increasing p(SO3, SO2) values at constant p(O2)
(Figure 1) and also increases with increasing p(SO3)
Limiting current density (mA cm–2)
values at constant p(O2, SO2) (Figure 2).

15 As SO3 is dissolved as S2 O2
7 in the melt, the major
oxidizing species in a thin Na2SO4 melt film in the
presence of SO3 is S2 O2
7 ions. The reduction reac-
10 tions were investigated by chronopotentiometric
measurements in a Na2SO4 melt at 900 C in an
O2 þ 1% (SO2 þ SO3) atmosphere5 and the results
5 show two transition times (Figure 3), attributed to
p(SO ), p(O ) varied
3 2
p(SO ) ≅ 1 atm
the following reduction steps, eqns [12] and [13]:
2

7 þ e ¼ SO4 þ SO3
S2 O2 ½12
0 2
0 1 2 3 4
p(SO 4
) •10 (atm)
3
SO
3 þ e ¼ SO3 ¼ SO2 þ O
2 2
½13
Figure 2 Limiting diffusion current at Pt electrodes
beneath a thin Na2SO4 melt as a function of p(SO3) at The reduction of O2 is also possible in an O 2
constant p(SO2, O2) at 896 C.
and SO 3 containing atmosphere, but to a much smal-
ler extent, due to the low solubility of O2 in the
reducing conditions or at high cathodic or anodic sulfate melt.
overpotentials. In principle, this is possible in deep For very basic melts without SO2/SO3 being pres-
melts or if the melt is trapped in pores within the ent in the gas atmosphere, the reduction of oxygen
oxide scale. ions occurs, as shown by chronopotentiometric mea-
The situation is completely different if gas species surements in a Na2SO4 melt, with the addition of
are dissolved in the melt. This has been shown by 5 mol% Na2O2 in air at 900 C. Two major reactions
measurements of the limited diffusion current in a were identified, eqns [14] and [15]:
thin Na2SO4 film at 896 C.4 The experiments were
carried out in an O2 – (SO3 þ SO2) gas mixture, O
2 þ 3e ¼ 2O
2
½14
0.4 on impedance measurements in a (Li,K)2CO3 melt at

615–800 C, the following reduction steps for oxygen
0.2 species take place by dissolution in the carbonate
τ1a melt,6 eqns [18] and [19]:
0
1. Peroxide mechanism:
3 þ O2 ¼ 2CO2 þ 2O2
2CO2 ½18
τ2a 2
−0.2
−0.4
Peroxide ions are mainly present in K-rich melts.
τc1
E (V)
2. Superoxide mechanism:
−0.6

3 þ 3O2 ¼ 2CO2 þ 4O2
2CO2 ½19
τc2
−0.8
The superoxide mechanism does not take place in
pure Li2CO3 melts, that is, superoxide-ions are not
−1.0 present in predominately Li-rich melts.
Current
reversed According to experiments by Cleas et al.7 in a
−1.2 molten Na2CO3–K2CO3 melt at 800 C, the solubil-
ity of CO2 in carbonate melts is rather high. The
0 1 2 3 4 5 6 7 results of cyclovoltametric experiments indicate a
Time (s)
very weak anodic reduction peak for CO2 to CO,
Figure 3 Chronopotentiometric curve for Na2SO4 in O2 + but a significant solubility of CO2 in the carbonate
1% (SO2 + SO3) at 900 C. melt by chemical reaction of CO2 with the carbonate
ion, according to eqn [20]:
2O CO2 þ CO2

3 ¼ C 2 O5
2
½20
2 þ 4e ¼ 2O þ O2 ½15
2
2
In general, corrosion mechanisms in sulfate melts

mainly involve gaseous species dissolved in the mol-
ten salt. The same is true for molten carbonates, 1.14.3.2 Oxidizing and Reducing Reactions
which is considered in the next section. of Carbonate Melt Species
Anodic polarization of a carbonate melt with inert
Au-electrodes is limited by the oxidation of the
1.14.3 Principles and Mechanisms of
carbonate ion according to the reaction given in
Corrosion in Carbonate Melts
eqn [16]. The cathodic process is generally the depo-
sition of alkali-metal according to:
Molten carbonates are also oxyanion melts, contain-
ing O2 ions, stemming from the dissociation of the Mnþ þ ne ¼ M ½21
carbonate ions, eqn [16]:
There is also evidence that, when in contact with
3 ¼ CO2 þ O
CO2 ½16
2 O2- and CO2-containing gases, some contributions to
the cathodic current are due to the following reduc-
In accordance with the definition of the acid/base tion reactions, eqns [22]–[24]:
characteristic of sulfate melts, the basicity of carbon-
1
ate melts is defined by eqn [17]: CO2 þ O2 þ 2e ¼ CO2
3 ½22
2
pO2 ¼ log aO2 ½17
1
O2 þ 2e ¼ O2 ½23
2
1.14.3.1 Solubility of Gases
CO2 þ 2e ¼ CO þ O2 ½24
The dissolution of oxygen in carbonate melts takes
place by a chemical reaction between the O2 mole- Special interest has to be paid to the electrochem-
cules and the basic component of the melt (i.e., ical behavior of oxygen species, stemming from
Na2O); hence, it does not obey Henry’s Law. Based the reactive dissolution of oxygen in the carbonate.8
The following reduction steps take place after reac- Na2SO4 at 1200 K. The melt basicity was fixed by a
tive dissolution by the peroxide and superoxide variation in the gas atmosphere, that is, p(SO2) and
mechanisms in a (Li,K)2CO3 melt at 615–800 C for p(O2), and by the addition of Na2O2 to the molten
oxygen species, eqns [25]–[30]: salt. Figure 5 shows solubility plots for various oxides
in molten Na2SO4 at 1200 K as a function of melt
1. Peroxide mechanism:
basicity, that is, log a(Na2O). The characteristic
3 þ O2 ¼ 2CO2 þ 2O2
2CO2 ½25
2
shape of the plots represents the amphoteric dissolu-
tion behavior of oxides in molten sulfates. Basic solu-
peroxide reduction steps: bility occurs at high a(Na2O) (left hand side of each

2 þ 2e ¼ 2O ½26
O2 2 curve), acidic solubility occurs at low a(Na2O) (right
hand side of each curve). A characteristic minimum is
O2 þ CO2 ¼ CO2
3 ½27
2. Superoxide mechanism: –log p(SO
3)
−15 −10 −5 0 5

3 þ 3O2 ¼ 2CO2 þ 4O2
2CO2 ½28
0
Fe2(SO4)2
superoxide reduction steps:

O
2 þ e ¼ O2
2
½29
−5
2 þ 2e ¼ 2O
O2 ½30
2 α-Fe2O3
NaFeO2
–log p(O2)
Hence, in corrosion studies in oxygen-containing

environments, the reduction of oxygen species is the FeSO4
relevant reduction reaction. Fe3O4
−10
1.14.4 Solubility of Oxides in Sulfate Na2SO4 FeS2

and Carbonate Melts Na2FeO2
Na2S
Metal surfaces are protected by a dense and adherent −15
oxide layer, formed upon high temperature oxidation. 0 5 10 15 20
Corrosive attack of metals and alloys in molten salts is –log aNa2o
due to the solubility of oxide scales by basic and Figure 4 Na–Fe–S–O phase diagram for prediction of
acidic dissolution whereas the solubility of oxides is corrosion of iron in a Na2SO4 melt at 1200 K.
a function of melt basicity and depends on the chem-
ical composition of the passive layer.
)
By a combination of thermodynamic stability dia- 4

metal Ions
Cr2O3
4
grams of the salt phase to be investigated and the oxide

mole Na SO
Fe2O3
2
phase under consideration, phase stability diagrams

can be constructed to predict the behavior of any 3
( mole
oxide in any molten salt. Figure 4 shows the Na–Fe– NiO

S–O phase diagram for prediction of corrosion of
log conc. ppm
iron in a Na2SO4 melt at 1200 K. This diagram was 2
constructed from superposition of the Na–S–O- and Co3O4 Al2O3
Fe–S–O-stability plots. The diagram shows clearly the SiO2

1 (900 °C)
basic corrosion products: NaFeO2 at high a(Na2O) and
the acidic corrosion products FeSO4 and Fe2(SO4)3 at 5 6 7 8 9 10 11 12 13 14 15 16 17
–log aNa O
low a(Na2O). 2
In order to investigate the solubilities of oxides in Figure 5 Solubility plots of various oxides in molten
molten salts as a function of melt basicity, experi- Na2SO4 at 1200 K as a function of melt basicity, for
ments were carried out by Rapp et al.8 in molten example, a(Na2O).
reached, where the dissolution mechanism changes Hwang and Rapp13 have shown that the solubility
from basic to acidic. of different oxides is not independent and a syner-
Basic dissolution of oxide scales occurs when gistic effect is observed, showing accelerated dis-
oxide ions are present in the molten salt from the solution of both oxides with respect to the single
dissociation of the sulfate ion according to the reac- oxide phase. At 1200 K, Fe2O3 and Cr2O3 powder
tion given in eqn [31]: were dissolved in a Na2SO4 melt with a certain
1 basicity, given by an atmosphere of oxygen containing
4 ¼ SO2 þ O2 þ O
SO2 ½31
2
1 vol.% SO2. In Figure 5, this condition is placed
2
between the solubility minima of Cr2O3 and Fe2O3
During oxidation of the bare metal surface by in the melt.
SO2
4 in a basic melt, metal oxide and additional Hence, Cr2O3 is dissolved by basic fluxing,
oxide ions are formed according to eqn [32]: eqn [35]:
M þ SO2
4 ¼ MO þ O
2
þ SO2 ½32 3
Cr2 O3 þ O2 þ 4Na2 SO4 ¼
2
Hence, the O2 ion concentration on top of the
metal oxide is higher than in the entire melt and 2Na2 CrO4 þ 2Na2 S2 O7 ½35
dissolution of oxide takes place according to eqn [33]:
The presence of Na2S2O7 leads to acidic fluxing of
MO þ O2 ¼ MO2
2 ½33 the Fe2O3, according to eqn [36]:
In the case of basic dissolution, the metal oxide is
Fe2 O3 þ 3Na2 S2 O7 ¼ Fe2 ðSO4 Þ3 þ 3Na2 SO4 ½36
dissolved as a complex oxide ion.
Such dissolution reactions were investigated by The accelerated dissolution is due to the short
Rapp and coworkers12 for several oxides, such as diffusion distance for SO3 in the melt.
Al2O3,9 Fe2O3,10 Cr2O3,11 NiO, and Co3O4, in a As in the case of sulfates, the solubility of oxides in
Na2SO4 melt at 1200 K. carbonate melts also depends on the partial pressure
Acidic dissolution occurs by SO3, which is dis- of CO2 in the gas phase, that is, on the basicity of the
solved in the molten sulfate as S2 O2
7 ions. In princi- melt. As NiO is the cathode material in molten car-
ple, the following reaction takes place, eqn [34]: bonate fuel cells, its solubility is of major interest and
has been investigated in detail. Figure 6 shows the
MO þ SO3 ¼ M2þ þ SO2
4 ½34
solubility plot for NiO in a (Li0.62K0.38)2CO3 mixture
The oxides are dissolved as metal ions in the as a function of p(CO2) in the gas phase at tempera-
sulfate melt and metal sulfates are formed. tures of 923–1023 K.14
–3
Li/K = 62/38
–4
log x (mole fraction)
–5
–6
1023 K
973 K
923 K
–7
–5 –4 –3 –2 –1 0 1
log p(CO2)
Figure 6 Solubility plot of NiO in a (Li0.62K0.38)2CO3 mixture as a function of p(CO2) in the gas phase at temperatures of
923–1023 K.
Two different solubility regimes are observed, the melt film towards the melt/gas or the gas/oxide
showing a negative slope (low p(CO2)) and a positive interface. In addition to the concentration gradient in
slope (high p(CO2)). These two regimes correspond the melt film, a basicity gradient also exists, from
to the basic (low p(CO2) dissolution and acidic disso- regions of high p(O2) at the melt/oxide interface to
lution (high p(CO2) of NiO in the carbonate melt, low p(O2) at the melt/gas interface. By reaching
according to eqns [37] and [38]: regions of low p(O2), the solute is no longer stable
and precipitation of solid oxide occurs by the reverse
basic dissolution:
of the dissolution reactions, eqn [43]:
NiO þ CO2
3 þ NiO2 þ CO2
2
½37
2 ¼ MO þ O
MO2 ½43
2
acidic dissolution: This precipitation creates a permanent sink for

NiO þ CO2 ¼ Ni 2þ
þ CO2 ½38 the solute, promoting the continuous flux of the
3
dissolved oxide in the melt film. On the other hand,
Further investigations of the basic dissolution of the consumption of oxidant gas O2 by the formation
NiO in other carbonates have shown that the solubil- of O2 ions and subsequent dissolution reaction at
ity also depends on the nature of the carbonate. The the melt–oxide interface creates the gradient for
solubility of NiO at 910 C in a CO2–O2–Ar atmo- oxidant diffusion. Figure 7 shows a schematic plot
sphere is highest for Rb2CO3 and lowest for Na2CO3, of the gradients and transport processes within the
especially in the acidic regime.15 melt film.
Evidence is also given to the fact that the basic Rapp and Goto16 analyzed the basic fluxing con-
solubility of NiO in K2CO3 melts at constant p(CO2) ditions and developed a criterion for the continuous
depends slightly on the p(O2) value in the gas phase. fluxing of oxide scales, eqn [44]:
This dependence is due to the following basic disso-
dðsol:Þ
lution reaction: 0 ½44
dx x¼0
1
K2 O þ 2NiO þ O2 ¼ 2KNiO2 ½39 As long as the negative solubility gradient (Rapp–
2 Goto criterion) is valid, basic fluxing proceeds and
the rate of the corrosion process is enhanced.
An important proof of the Rapp–Goto criterion
was given later by Otsuka and Rapp,17 who performed
1.14.5 Corrosion Mechanisms an electrochemical experiment (potentiometric mea-
in Sulfate Melts: The Rapp and surements) on preoxidized nickel at 1173 K in an
Goto model O2–0.1% SO2 gas atmosphere and covered by a thin
film of Na2SO4. The authors continuously recorded
As discussed earlier, the oxidation of the metal beneath the basicity and oxygen activity of the melt as a func-
a highly basic molten sulfate film generates O2-ions tion of time. Figure 8 shows the results, plotted on the
at the oxide–melt interface by the reduction of O2 or Ni–Na–S–O phase diagram. At the beginning of the
by the reduction of the sulfate-ion, eqns [40]–[42]. corrosion reaction, the melt was found to be basic.
M ¼ M2þ þ 2e ½40
Metal MO Salt melt Gas phase
O2 þ 2e ¼ O2 ½41
[MO]diss.
and:
M þ SO2 M2+
4 ¼ MO þ O þ SO2 ½42
2
[O2−]
Accordingly, the region close to the oxide surface
is much more basic than the entire salt film and basic
dissolution of the oxide occurs. In the ideal case, the
dissolved species form a separate layer, comparable to
a Nernst concentration interface. From this interface, Figure 7 Schematic plot of the solubility gradient and
transport of the dissolved oxide species and also oxi- transport processes within the melt film. The oxide
dant gas occur within their concentration gradients in precipitates at the melt–gas interphase.
0
11
NiSO4
NaNiO2 Basic 10 Acidic
−4 dissolution
9 dissolution
8
4 7 Start of
NiO 6 corrosion
3
log p(O2)
−8 2
Na2NiO2
10 min
30 min 5 14 min
1
20 min
−12
18 min
Ni
−16 16 min
Ni3S2
NiS NiS2
−20
−4 −0 4 8 12 16 20
–log a(Na2O)
Figure 8 Basicity and oxygen activity measurements during corrosion of preoxidized Ni beneath a Na2SO4 film at 1173 K in
O2–SO2, plotted in the Ni–Na–S–O phase diagram.
During the corrosion reactions, the basicity of the melt

increases and Ni-sulfide becomes stable. As the com- (d)
position of the gas atmosphere was not changed, the
(e)
results show quite clearly that the local chemistry of the
melt is influenced by the local corrosion reactions.
Acidic fluxing requires a higher p(SO3) value in
the gas phase and/or some alloying elements such as (c)
Co or Ni, from which molten eutectic mixtures are
able to form with the sulfate deposit (alloy-induced
acidic fluxing). The main feature of acidic fluxing is
the solubility of certain metal sulfates, formed by
reaction of the alloying elements and SO3 from the
gas phase, within the salt deposit. Generally, no sul-
fide formation occurs by acidic fluxing.
(b) (a)
1.14.6 Methods of Investigation of
Corrosion in Molten Salts
Figure 9 Typical set-up of the inner crucible of a high
1.14.6.1 Electrochemical Techniques temperature electrochemical cell with counter-electrode (a),
the working-electrode (b), the reference electrode (c), the gas
Being electrochemical in nature, fundamental studies of inlet tube (d), and the thermocouple (e).
hot corrosion mechanisms require electrochemical
techniques. The simplest electrochemical measurement and breakthrough potentials. For such studies, the melt
is polarization of the sample and the measurement of is placed in an alumina crucible, situated in a vertical,
the induced current. As in aqueous corrosion, polariza- closed-end furnace, as shown in Figure 9.
tion studies provide information mainly about the In order to avoid electronic distortions from the
potential dependent behavior of a piece of metal in furnace heating wires, a cylindrical vessel made of
the melt, that is, active dissolution, the passive range stainless steel is usually placed between the alumina
crucible and the furnace. The vessel is electrically manner in the potential range of approximately 0.25
connected to the furnace in order to avoid electrical to 0.25 V; this is followed by an increasing current.
charges on the surface. The alumina crucible is Investigation of the sample within the passive region
closed by a heat-resistant pyrex-glass flange with shows the formation of a porous Fe2O3 scale on the
spaces for the electrodes, the thermocouple and the metal surface.19
gas-inlet tube. The electrode wires and the thermo- Extended investigations are possible by using
couple are shielded by small-diameter alumina tubes. cyclic voltammetry. Figure 11 shows a cyclic voltam-
These alumina tubes are fixed with cayon fittings in the mogram of pure iron in a eutectic 62 wt% Li2CO3–
glass flange. 38 wt% K2CO3 melt at 650 C in synthetic air,
For the polarization studies, a three electrode set- containing 25 vol.% CO2. The details regarding ref-
up is necessary, consisting of a working electrode erence electrode and set-up are described elsewhere.20
(WE), a counter electrode (CE), and a reference Upon anodic polarization, the voltammogram shows
electrode (RE). The WE is usually the metal to be three different anodic peaks (A, B, C) and two cathodic
investigated or, for studies of the polarization behav- peaks (D, E).
ior of the melt itself, Au or Pt are appropriate; the According to detailed investigation by SEM and
latter are also suitable for the counter electrode. XRD on samples, held at different anodic potentials,
The main difference in aqueous corrosion studies is the following reactions were identified:
the design of the reference electrodes. The most useful A: Fe þ CO2
3 ¼ FeO þ CO2 þ 2e ½45
types are closed-end tubes of either internally-plati-
nized Y2O3-stabilized ZrO2, mullite or glass mem- B: Fe þ 4FeO þ 4CO2 þ
3 þ 5Li ¼
branes, depending on the composition of the melt and
the temperature. Glass membranes, especially Pyrex®, Li5 Fe5 O8 þ 4CO2 þ 3e ½46
Supremax® and Duran 50®, are Naþ and Kþ ion- C: Li5 Fe5 O8 þ 2CO2
3 ¼
conducting membranes and are also useful for studies LiFe5 O8 þ 2Li2 CO3 þ 4e ½47
in molten salts, especially at temperatures up to 700 C.
Impedance spectroscopic measurements are occa-
Similar to the mullite reference electrode, a silver wire
sionally used in molten-salt corrosion studies. In
is dipped into the molten chloride or sulfate to be
general, most of the impedance spectra emphasize
investigated; usually this melt additionally contains
diffusion-controlled kinetics for the active corrosion
AgCl2 or Ag2SO4 in concentrations of 1–10 mol%.18
of metals in molten salts. This behavior is expected as
For example, Figure 10 shows a polarization curve
the activation energy for charge-transfer reactions is
for a 2.25Cr–1Mo–steel in a eutectic (Ca–K–Na) sul-
easily reached at higher temperatures.
fate mixture at 850 C in an N2–5 vol.% O2 gas mix-
ture, using a mullite–Ag/AgSO4 reference electrode.
0.75
After an active region, the material behaves in a passive
C
B
0.5
0.50
A
0.4
0.3 0.25
Pt
i (A m−2)
2.25Cr–1Mo
i (A cm–2)
0.2
0.00
0.1
0.0 −0.25 E D
–0.1
–1.0 –0.5 0.0 0.5 1.0 1.5
−0.50
E (V) vs. (Ag/Ag+) −1500 −1000 −500 0
E (mV)
Figure 10 Polarization curve of a 2.25Cr–1Mo–steel in
a eutectic (Ca–K–Na)–sulfate mixture at 850 C in an Figure 11 Cyclic voltammogram of pure iron in a eutectic
N2–5 vol% O2 gas mixture; the curve for Pt is shown for 62 wt% Li2CO3–38 wt% K2CO3 melt at 650 C in synthetic
comparison. air, containing 25 vol% CO2.
The paper by Zheng et al.21 gives some theoretical full salt immersion (crucible test, thick salt film)
impedance models for hot corrosion of metals in surface coverage of the sample (thin salt film)
molten salts. In general, the authors propose large slurry or from an aqueous solution (thin salt film)
semicircles in the Nyquist plot for nonactive metals,
Determination of the extent of corrosion can be
resulting from a rate-limiting charge-transfer reac-
obtained by measurements of the scale thickness or of
tion. For active metals, the total impedance may
the remaining metal thickness or by the determina-
result from the scaling of the metals when the
tion of the total weight loss after removal of the
Nyquist plot should show two capacitance loops,
corrosion products. The last method is particularly
and with the rate-limiting step being the charge-
favored if a high amount of oxide is dispersed in the
transfer reaction. For porous scales on the surface, a
solidified salt.
line should be shown in the low frequency range; this
According to the standard guidelines for expo-
is related to a diffusion-controlled process.
sures of samples coated with thin salt films, as devel-
One example is the hot corrosion of a preoxidized
oped in the TESTCORR project,23 the extent of
nickel specimen under a thin Na2SO4 melt film in a
corrosion should be determined on a logarithmic
0.1 wt% SO2–O2 gas mixture at 1200 K.22 By varia-
time scale, that is, after 1, 3, 10, 30, 100, 300, and
tions in the oxide scale thickness and the purity of the
1000 h of exposure and redeposition of salt should be
material, different regimes of corrosion were investi-
made after every 20 h.
gated, that is, the passive state, the pseudopassive state
In order to study short-term kinetics, thermo-
and the active state. The passive state for 99.9975% of
gravimetric experiments are useful; these allow the
pure nickel, preoxidized in pure O2 for 2 h at 1200 K, is
identification of possible incubation times or the influ-
controlled by diffusion of S2 O2 7 ions in the salt melt.
ence of gas phase composition. However, it has to be
The corresponding Nyquist plots of impedance data
shown that the overall mass change is not influenced
show linear behavior in the low frequency range with a
by reactions of the salt itself, for example, significant
slope of 45 , as shown in Figure 12.
evaporation or reaction of the salt with the gas phase.
1.14.6.2 Other Methods

Multisample exposure tests are the most useful tests 1.14.7 Examples of Hot-Salt
for estimating long-term corrosion data behavior of a Corrosion in Industrial Practice
certain material by fitting the corrosion data to gen- 1.14.7.1 Gas Turbines: Corrosion Beneath
eral kinetic laws. The furnace set-up, used for the Na2SO4 Deposits
tests, is connected to a gas-flow device. In addition,
salt needs to be deposited on the sample surface. This The need for the early work on hot corrosion beneath
can be done by different techniques: Na2SO4 melts was based on gas-turbine blade failures
as a consequence of Na2SO4 deposition. It has been
evaporation of the salt with subsequent condensa- established that deposition of Na2SO4, formed by salt
tion on the cooler sample surface (Dean test) from the air and sulfur from the fuel, is responsible
for these kinds of attack. Especially the work by
16 Goebel and Pettit24 as well as by Bornstein and
DeCrescente25,26 recognized that the formation of
sulfides on the metallic engine parts results from
Imag Z (Ω)
the presence of a thin salt film on the metal surface.

8 As a consequence, extensive research was conducted
on alloys coated with Na2SO4 and how they reacted
in different environments. In the following discus-
sion, some general principles on the reaction
0 mechanisms will be discussed. The corrosion kinetics
0 8 16 24 32
Real Z (Ω) of metals are significantly enhanced by several orders
of magnitude compared to the oxidation without any
Figure 12 Nyquist plot of impedance data of the passive
state of 99.9975% pure nickel in a Na2SO4 melt in an
salt, as shown in Figure 13 for nickel reaction
0.1 wt% SO2–O2 gas mixture at 1200 K, showing linear beneath a Na2SO4 deposit at 900 C in an O2–SO2
behavior in the low frequency range, with a slope of 45 . containing atmosphere.27
10−1
900 ⬚C, Ni SO3 pressure at which
50 O2 + 4% SO2, 1 atm liq. sulfate observed
300 p*SO3 (NiSO4)

40 Luthra
Thickness of NiO (mm)

Weigt gain (mg cm–2)
10−2 + shores
Na2SO4 + NiSO4
liquid solution
30
P(SO3) (atm)
With Na2SO4 200
20 10−3
100
10
Without Na2SO4 Na2SO4 + NiSO4
solid solution
10−4
0 0
0 5 10
Time (h) 700 800 900
Figure 13 Comparison of the reaction kinetics of pure Temperature (⬚C)
nickel with and without deposit of Na2SO4 in 1 atm Figure 14 Comparison of the theoretical values of the
O2–4% SO2 at 900 C. Reproduced from Edeleanu, C.; critical SO3 pressure needed to form a Na2SO4 + NiSO4 melt
Littlewood, R. Electrochim. Acta 1960, 3, 195. with the lowest pressures occurring where a molten sulfate
was observed and ‘hot corrosion’ was initiated.28,29,34,35
Two main corrosion mechanisms have been iden-
tified: Type I and Type II hot corrosion. Type I formed. In only O2-containing gases, Co2þ ions are
usually occurs at temperatures above the melting dissolved from the oxide and transported to the melt–
point of Na2SO4 (T > 884 C); here, the und die gas interface at which they are oxidized by O2 to
oxide scale is dissolved in the salt melt due to basic Co3þ ions and Co3O4 is formed. Some Co3þ ions
fluxing (see Section 1.14.4). Type II hot corrosion also diffuse back to the melt–oxide interface where
occurs at temperatures below the melting point of they are reduced to Co2þ ions again. Alloying ele-
Na2SO4. One example for this is the corrosion of ments Mo, W, and, also, V (present as a fuel impurity)
nickel beneath a solid Na2SO4 salt deposit in SO3- form the acidic oxides: MoO3, WO3, and V2O5. The
containing gas. In the early stages of corrosion, a effect of NaVO3 on the acidic fluxing of CeO2, HfO2,
Na2SO4–NiSO4 solid solution is formed: and Y2O3 was studied in detail by Zhang and Rapp.31
NiO þ SO3 þ Na2 SO4 ¼ Na2 SO4 NiSO4 ðssÞ ½48 The authors show that fluxing was accelerated by the
presence of acidic solutes in the melt. This behavior
In the preceding reaction, the amount of NiSO4 was shown to result from complexing reactions
increases, the Na2SO4–NiSO4 mixture starts to melt between the solute VO3 ions and oxide ions,
and accelerated corrosion is observed. The melting provided by the acidic dissolution of ceria in the
point of the mixture strongly depends on the p(SO3) respective oxide.
value in the gas phase, that is, it decreases with the
increasing value of p(SO3) and the melting point of the
Na2SO4–NiSO4 mixture decreases,28,29 Figure 14.
1.14.7.2 Waste Incineration: Corrosion in
In aircraft engines, mostly Type II hot corrosion is
Chloride Melts
mainly observed in nickel- and cobalt-based materi-
als. Extensive studies were preformed on Co–Cr, Co– In waste-fired boilers, solid and molten salts are
Al, Ni–Cr, Ni–Cr–Al, and Co–Cr–Al–Y alloys.35,36 formed by condensation of aerosols within the flue
A detailed corrosion mechanism for Co-based alloys gas on cooled material surfaces that function as heat-
in O
2 and SO3 containing gases was presented by exchanger tubes. The presence of salts accelerates the
30
Luthra. In this model, the SO3 is dissolved in the corrosion process, especially if they are associated
melt and transported via SO2 4 ions from the melt– with the formation of molten phases on the material
gas to the melt–oxide interface at which CoSO4 is surface or in the oxide scales.
Figure 15 shows a characteristic failure situation, ions and initially free of oxygen, which is different
in which the salt-covered part of the tube is signifi- from sulfates. In some cases, molecular species exist,
cantly damaged. for example AlCl3 or HgCl2.
The molten phases are mainly chlorides, such as Theoretical thermodynamic considerations by
KCl, NaCl, ZnCl2, and PbCl2.32 Due to the signifi- Edeleanu and Littlewood34 have shown that the cor-
cant amount of heavy metals, these salt mixtures form rosion tendency of a metal in chloride melts depends
molten eutectics at temperatures as low as 250 C, as on the stability of the specific metal chloride. Fur-
shown in Table 1. thermore, the solubility of the metal chloride or
A typical salt melt-induced attack of a superheater metal oxide depends on the total pressure and,
tube (T 550 C) from a waste-fired boiler is shown more importantly, on contaminants in the molten
in Figure 16. The morphology of the scale is typical salt, especially oxygen. In pure chloride melts, a pas-
for molten salt attack, that is, a thick and dense sivation of the metal is impossible, because most of
chromium-rich oxide scale in contact with the the metal chlorides are soluble. Electrochemical
metal and iron-oxide and nickel-oxide precipitates experiments on iron and nickel in a KCl–NaCl melt
in the former molten salt.33 at 700 C in argon showed a very negative corrosion
The behavior of metals and their oxides in molten potential with respect to platinum and no passivation
chlorides has not been studied as extensively as in of the metal was observed. As corrosion products,
sulfate systems. Chloride melts are mostly simple Fe3O4, Fe2O3, and NiO were observed on top of the
melt at the melt–gas interface.36 Obviously, oxidation
of the soluble metal chlorides occurs in contact with
the oxygen-containing gas.
The cathodic reaction during metal oxidation is
the reduction of contaminants in the melt, especially
dissolved oxygen. Ishitsuka und Nose35 studied the
solubility of some metal oxides in pure NaCl–KCl
melts at 727 C and in NaCl–KCl–Na2SO4–K2SO4
melts at 550 C. The basicity of the melt was set
at log a 2O ¼ 2 for basic, log a O ¼ 7 for
2
intermediate and log a O 17 for acidic condi-

2
tions. The results show a typical amphoteric solubil-

ity, as already known from sulfate melts. For some
oxides, no solubility minima were observed, that is,
for example, NiO/Cr2O3 is thought to be dissolved as
Figure 15 Characteristic failure situation of a superheater chromate, corresponding to a basic solubility.
tube from a waste-fired boiler. The ash-covered part of
the tube is significantly damaged.
Table 1 Eutectic mixtures and melting points of salts

found in deposits from waste fired plants
Composition (wt%) Melting point ( C)
48ZnCl2–52KCl 250
82ZnCl2–18KCl 262
84ZnCl2–16KCl 262
73ZnCl2–27PbCl2 300 1000 μm
31NaCl–69PbCl2 410
21KCl–79PbCl2 411
17NaCl–83PbCl2 415
39ZnCl2–50KCl–11PbCl2 275 Figure 16 Typical melt-induced attack of a
35ZnCl2–48NaCl–17PbCl2 350 superheater tube (T 550 C) from a waste-fired boiler.
16NaCl–40KCl–44PbCl2 400 The morphology of the scale is typical for molten salt
K2SO4–Na2SO4–ZnSO4 384 attack that is, a thick and dense chromium-rich oxide scale
KCl–ZnCl2–K2SO4–ZnSO4 292 in contact with the metal and iron-oxide and nickel-oxide
precipitates in the molten salt.
In summary, it seems to be evident that corrosion chloride diffuses outward towards the oxygen-con-
in chloride melts is based on the solubility of metals taining gas and is oxidized to oxide again.
and oxides as metal chlorides in the molten salt. It has been found that iron-based low alloyed
In deposits from waste incineration, mainly KCl– steels show much more corrosion than nickel-based
ZnCl2 mixtures are present in the ashes and fast materials. The proposed model is supported by ther-
corrosion occurs at relatively low temperatures. The modynamic calculations of the solubility of iron,
main corrosion mechanism is dissolution of metal in chromium, and nickel beneath a ZnCl2–KCl melt.
the molten salt as soluble metal chlorides. Figure 17 Figures 18–20 show ternary phase diagrams for the
shows a schematic plot of the mechanisms involved, FeCl2–ZnCl2–KCl, CrCl2–ZnCl2–KCl, and NiCl2–
as described by Ruh.36 ZnCl2–KCl systems at 320 C.
At the metal–salt interface, the metal is dissolved The solubility of FeCl2 is quite high whereas the
as metal chloride by forming a ternary eutectic with solubility of CrCl2 is low and almost no solubility
the salt deposit. The concentration of dissolved metal occurs for NiCl2. This reflects the stability of the
chloride is high at the inner interface and low at different alloys based on the corrosion mechanism
the salt–gas interface. As a consequence, the metal as described above.
Ar–O2 1.14.7.3 Corrosion in Molten Carbonate

Gas flow p(O2) p(Cl2)
Fuel Cells
3
2FeCl2 + — O2 = Fe2O3 + 2Cl2
2 Molten carbonate fuel cells are high temperature fuel
cells, working at 650 C with a molten 62 wt%
Li2CO3–38 wt% K2CO3 eutectic mixture as the elec-
Melt Fe + Cl2 FeCl2 (dissolution) trolyte, CO2 and O2 as oxidant gases and H2 as the
Metal Fe fuel. The relevant reactions are:
Cathode:
Figure 17 Reaction scheme for the mechanisms involved
1
in chloride-induced corrosion beneath a molten ZnCl2–KCl CO2 þ O2 þ 2e ¼ CO2
3 ½49
eutectic from deposits in waste-incineration plants. 2
FeCl2
#: Liquid + K2ZnCl4 (s)

*: Liquid + KFeCl3 (s)
l3 (s .8
0.2
§: Liquid + K2FeCl4 (s)

0
)+
$: Liquid + KFeCl3 (s) + K2FeCl4 (s)

l2 (s KFeC
0.6
0.4
+
)
uid
Liq
FeC
0.4
0.6
$ Liquid + FeCl2(s)
∗
§
0.2
0.8
KCl (s) +
K2ZnCl4 (s)
#
+ K2FeCl4 (s) Liquid
KCl 0.8 0.6 0.4 0.2 Zncl2

Mole fraction
Figure 18 Solubility of FeCl2 in a ZnCl2–KCl eutectic mixture at 320 C.
CrCl3
0.8
0.2
0.6
0.4
l4 (s)
)
l4 (s
+ K2 ZnC
2 ZnC
0.4
0.6
(s) +
CrCl (s)
3
3
rCl
+C
0.2
Liquid + CrCl3 (s)
0.8
Liquid +
l (s)
KC
Liquid
KCl 0.8 0.6 0.4 0.2 ZnCl2

Mole fraction
Figure 19 Solubility of CrCl2 in a ZnCl2–KCl eutectic mixture at 320 C.
NiCl2
#: Liquid + K2ZnCl4 (s)

*: Liquid
0.8
0.2
4 )
2 ZnCl (s
0.6
0.4
NiCl (s l3 (s) +
l4 (s)
2 )+K
KNiC
+ K2 ZnC
0.4
0.6
Liquid + NiCl2 (s)

NiCl (s)
2
0.2
0.8
KCl (s) +
Liquid +
KNiCl3 (s) +
K2ZnCl4 (s)
KCl 0.8 # 0.6 ∗ 0.4 0.2 ZnCl2

Mole fraction
Figure 20 Solubility of NiCl2 in a ZnCl2–KCl eutectic mixture at 320 C.
Anode: The main corrosion problems occur at the cathodic

side, where NiO is used as the cathode material and
H2 þ CO2
3 ¼ H2 O þ CO þ 2e ½50
stainless steel as the current collector. Both parts
In practice, the stack technology is applied, by are in contact with the gas phase (O2, CO2) and
stapling some of the modules to a bigger device. the molten carbonate, leading to hot corrosion
attack. NiO forms as the cathode materials starts to 13. Hwang, Y.-S.; Rapp, R. A. J. Electrochem. Soc. 1990, 137,
1276.
dissolve in the melt and reprecipitates in the elec- 14. Ota, K.; Mitsushima, S.; Kato, K.; Kamiya, N. In
trolyte, leading to a short circuit and breakdown of Proceedings of the second Symposium in MCFC-
the cell voltage. Furthermore, corrosion of the Technology; Electrochemical Society, 1990; Vol. 90–16,
318.
cathodic-current collector by the molten carbonate 15. Orfield, M. L.; Shores, D. A. J. Electrochem. Soc. 1988,
occurs and the cell voltage decreases due to the 135, 1662.
growth of poorly-conducting oxide phases, that is, 16. Rapp, R. A.; Goto, K. In Proceedings of the Fused Salt
Symposium; Braunstein, J., Ed.; The Electrochemical
LiCrO2. Laboratory research is still going on in Society: Princeton, NJ, 1979; Vol. 2.
order to find metallic materials with good corrosion 17. Otsuka, N.; Rapp, R. A. J. Electrochem. Soc. 1990,
resistance, that form corrosion products with rea- 137, 46.
18. Lee, K. N.; Shores, D. A. J. Electrochem. Soc. 1990,
sonable electric conductivity, such as spinels, doped 137, 859.
with Co and/or Mn. 19. Spiegel, M. Proceedings EUROCORR’00, London, 2000.
20. Biedenkopf, P.; Spiegel, M.; Grabke, H. J. Mater. Corros.
1997, 48, 731.
References 21. Zheng, C. L.; Eang, W.; Wu, W. T. Corros. Sci. 2001, 43,
787.
22. Wu, Y. M.; Rapp, R. A. J. Electrochem. Soc. 1991, 138,
1. Lux, H. Z. Elektrochem. Angew. Phys. Chem. 1939, 45, 2683.
303. 23. Tomkings, A. B., et al. TESTCORR, Code of Practice, ERA
2. Flood, H.; Forland, T.; Motzfeld, K. Acta Chem. Scand. Report 2000-0546, 2001, Unpublished data.
1952, 6, 257. 24. Goebel, J. A.; Pettit, F. S. Metall. Trans. 1970, 4, 1943.
3. Andresen, R. E. J. Electrochem. Soc. 1979, 126, 328. 25. Bornstein, N. S.; DeCrescente, M. A. Trans. Met. Soc.
4. Shores, D. A.; Fang, W. C. J. Electrochem. Soc. 1981, 128, AIME 1969, 24, 1947.
346. 26. Bornstein, N. S.; DeCrescente, M. A. Metall. Trans. 1971,
5. Fang, W. C.; Rapp, R. A. J. Electrochem. Soc. 1983, 130, 2, 2875.
2335. 27. Lillerud, K. P.; Kofstad, P. Oxid. Met. 1984, 21, 233.
6. Nishina, T.; Uchida, I.; Selman, J. R. J. Electrochem. Soc. 28. Luthra, K. L.; Shores, D. A. J. Electrochem. Soc. 1980,
1994, 141, 1191. 127, 2202.
7. Claes, P.; Thirjon, B.; Glibert, J. Electrochim. Acta 1996, 29. Luthra, K. L. Metall. Trans. A 1982, 13, 1647, 1843, 1853.
41, 141. 30. Luthra, K. L. J. Electrochem. Soc. 1985, 132, 1293.
8. Rapp, R. A. Corrosion 1986, 42, 568. 31. Zhang, Y. S.; Rapp, R. A. Corrosion 1987, 43, 384.
9. Jose, P. D.; Gupta, D. K.; Rapp, R. A. J. Electrochem. Soc. 32. Spiegel, M. Mater. Corros. 1999, 50, 373.
1985, 132, 735. 33. Spiegel, M. Mater. Corros. 2000, 51, 303.
10. Zhang, Y. S.; Rapp, R. A. J. Electrochem. Soc. 1985, 132, 34. Edeleanu, C.; Littlewood, R. Electrochim. Acta 1960,
734. 3, 195.
11. Zhang, Y. S. J. Electrochem. Soc. 1986, 133, 655. 35. Ishitsuka, T.; Nose, K. Mater. Corros. 2000,
12. Gupta, D. K.; Rapp, R. A. J. Electrochem. Soc. 1980, 127, 51, 177.
2194. 36. Ruh, A.; Spiegel, M. Corros. Sci. 2006, 48, 679.
1.15 High Temperature Tribocorrosion
I. A. Inman, P. S. Datta, and H. L. Du
Ellison Building, Northumbria University, Newcastle upon Tyne, NE1 8ST, UK
C. Kübel
Group Leader Electron Microscopy, Forschungszentrum Karlsruhe, Institute for Nanotechnology, Postfach 3640, 76021 Karlsruhe,
Germany
P. D. Wood
Ricardo Plc, Shoreham Technical Centre, Shoreham-by-Sea, West Sussex, BN43 5FG, UK

1.15.2 Wear Theory 336
1.15.2.1 Early Wear Theory 336
1.15.2.1.1 Archard and Hirst – Distinction between mild and severe wear 336
1.15.2.1.2 Classification by mechanism 338
1.15.2.2 ‘Two and Three Body’ Wear 340
1.15.2.2.1 Overview 340
1.15.2.2.2 Surface films and preoxidation – Third body or not? 341
1.15.2.2.3 Behavior of particles at the interface 342
1.15.2.2.4 The effect of forces of attraction between third bodies 343
1.15.2.3 Mild Wear and Mechanisms of Compact Oxide Formation 344
1.15.2.3.1 Introduction to compacted oxides or ‘glazes’ 344
1.15.2.3.2 Quinn’s oxidational wear model 345
1.15.2.3.3 Modification of Quinn’s oxidational wear model for discontinuous contact 346
1.15.2.4 Effect of Load and Sliding Speed 347
1.15.2.4.1 Early work 347
1.15.2.4.2 Wear of cobalt-based alloys 349
1.15.2.4.3 The effect of load and sliding speed – Stellite 6 351
1.15.2.5 Effect of a Second Phase on Wear 353
1.15.3 High Temperature Wear Behavior of Advanced Materials 354
1.15.3.1 Oxide Dispersion Strengthened (ODS) Alloys 354
1.15.3.1.1 Introduction 354
1.15.3.1.2 Observations on high temperature wear behavior of some ODS alloys 354
1.15.3.1.3 Effects of load 358
1.15.3.2 Intermetallics 359
1.15.3.2.1 Introduction 359
1.15.3.2.2 Wear of TiAl – Metallic counterfaces 360
1.15.3.2.3 Wear of TiAl – Ceramic counterfaces 362
1.15.3.3 Nimonic Alloys 364
1.15.3.3.1 Incoloy 800HT counterface 364
1.15.3.3.2 Stellite 6 counterface 366
1.15.3.3.3 Effect of Nimonic material processing route on wear 371
1.15.3.3.4 Nimonic 80A sliding wear – Comparisons between various wear rig configurations 371
1.15.3.3.5 Si3N4 counterface 371
1.15.3.4 Effects of Environmental Variables 371
1.15.3.4.1 Oxygen levels and partial pressure 371
1.15.3.4.2 Effect of water vapor and relative humidity 373
1.15.3.4.3 Other atmospheres 373
1.15.3.5 Effects of Pretreatment of Sliding Surfaces 374
1.15.3.5.1 Preoxidation 374
1.15.3.5.2 Presliding 375
331
1.15.3.5.3 Ion implantation 375

1.15.4 ‘Glaze’ Formation – Microscale and Nanoscale Investigations 376
1.15.4.1 Introduction 376
1.15.4.2 Microscale Studies of ‘Glaze’ Formation 376
1.15.4.3 Third Body Interaction in Relation to Compact Oxide Formation 377
1.15.4.4 Nanoscale Investigations of ‘Glaze’ Formation 379
1.15.4.4.1 Wear data 381
1.15.4.4.2 Studies of wear-affected surfaces produced during sliding of Nimonic 80A
against Stellite 6 at 20 C 383
1.15.4.4.3 Nano-scale microscopy of ‘glazed’ layers formed during high temperature
sliding wear at 750 C 387
1.15.4.4.4 Other systems 389
1.15.5 Wear Maps: A Useful Design Aid for Selecting Wear-Resistant Materials and
Surfaces 389
1.15.5.1 Introduction 389
1.15.5.2 Work by Lim 392
1.15.5.3 Dissimilar Interfaces 393
1.15.5.3.1 Oxide chemistry 396
1.15.6 Summary 396
References 396
Glossary Adhesion is also more effective in a vacuum,

Abrasion The removal of material from a solid where there is no surrounding atmosphere to
surface by the mechanical action of another affect it. Adhesion is normally used to
solid. The second solid may be in the form of describe attraction between dissimilar
a second body (the opposing sliding surface) molecules and cohesion between like
or third body (wear debris). molecules.
Abrasive wear The removal of surface material Asperity A protruding, raised area of (sliding)
from an object by the action of another agent surface.
or medium. This may be the surface of Block-on-cylinder A test configuration used to
another object or by hard particles trapped study unidirectional sliding wear, with the
between the two interacting surfaces – ‘block’ (either stationary or in reciprocating
referred to as ‘two body’ and ‘three body’ motion) as the sample material and the
abrasion respectively. The hard particles or ‘cylinder’ (normally rotating) as the
surface must be 1.3 times harder than the counterface.
soft material from which material is removed. Coefficient of friction Dimensionless value
Adhesion/cohesion The action of intermolecular describing the resistance experienced by
forces at surfaces or interfaces in close two surfaces in movement against each
contact, holding those surfaces together. other. The higher the value, the greater the
Such forces may include chemical bonding ability to resist movement. The ‘frictional
(chemical adhesion), intersolubility (diffusive force’ is that which needs to be applied to
adhesion), Van der Waals forces (dispersive overcome this resistance to movement.
adhesion) and electrostatic forces Compacted oxide layer Often wear-protective
(electrostatic adhesion). Surface interlocking layers of oxide are formed as a
may also occur (mechanical adhesion) via ‘tribocorrosion’ product, when two metals (or
material filling surface voids or pores. a metal and ceramic) are slid against each
Adhesion is of greater influence during other at high temperature in an oxygen-
contact of clean metallic surfaces (and thus containing atmosphere.
during severe wear), as there are no Compacted particles Debris forming a
contaminants to prevent this contact. compacted oxide layer.
High Temperature Tribocorrosion 333
Counterface (countersurface) The material or HIPped (hot isostatically pressed) A method of

surface sliding against the sample. alloy or material production by powder
Critical oxide thickness (j) According to Quinn’s metallurgy methods, using a combination of
oxidational wear model,1 this is the temperature and (isostatic) pressure to
maximum thickness of compacted oxide produce the final item.
layer or ‘glaze’ that can develop and remain Intermetallic Materials composed of defined
mechanically stable. Beyond this, the oxide proportions of two or more metallic elements
becomes unstable, generating a wear and possibly one or more nonmetallic
particle. elements. Such materials typically have a
Critical transition temperature The minimum highly ordered crystal lattice structure
temperature according to Jiang2 at which composed of the constituent parts, but not
sintering can occur within a compacted necessarily the same as any of the
oxide layer, allowing a solid wear-protective constituent parts.
layer (or ‘glaze’) to form on top of it before it is Ion-implantation The implantation of ions of one
broken down. material into the crystal lattice of another
Cross-slip Transfer of (screw) dislocation glide solid, thereby bringing about a desired
from one slip plane to another during change in the physical properties of that solid
deformation. (i.e., doping).
Cutting The physical deformation due to a harder Material transfer Transfer of material at the wear
asperity or third body being pushed through interface from one wear surface to the other.
a softer material where adhesion is high. The Mechanical alloying The mixing of two or more
movement of the harder material over the solid phase materials to produce material of
softer material results in the creation of a different composition and possibly phase to
deeper groove upon the sample surface than either of the original materials. During sliding
seen with ‘wedge-forming’ or ploughing, with wear (i.e., severe wear), this may happen
long strips of debris forming at the point of where the two parent sliding surfaces are of
contact. different composition, due to material
Dissimilar interface system A configuration where transfer between and mixing at the sliding
the sample and counterface are of different interface. This mechanically alloyed material
materials. may then potentially readhere to either
Ejection The removal of wear debris parent sliding surface.
from the sliding wear interfaces such that it Mild wear Wear in which direct contact between
can no long play an ‘active’ part in the wear the sliding surfaces are separated by
process. corrosion debris (generally oxide), often less
Fretting Damage due to repeated relative motion than 1 mm diameter. This debris may take the
of sliding surfaces against each other, for form of compacted oxide or ‘glaze’ layers.
example, due to vibration or oscillation. Where the two sliding surfaces are metallic,
Glaze Compacted oxide layer, usually referring to a the debris being generally nonmetallic
more wear protective, sintered form prevents adhesion and (metallic) transfer.
generated (normally) at elevated Coefficient of friction values are thus usually
temperatures. much lower. Wear values are generally, but
High angle annular dark field in STEM Images not always, lower than severe wear,
formed by gathering scattered electrons with however, such debris can act abrasively and
an annular dark field detector. An ‘annular instead enhance wear.
dark field’ (ADF) image generated only by ‘Orowan’ stress The stress required
very high angle, incoherently scattered to bow a dislocation between particles or
electrons, is highly sensitive to variations in impurities, acting as obstacles within a
atomic number of the elements in the crystal lattice.
sample (Z-contrast images) and is hence Oxide dispersion strengthened (ODS) alloy An
known as ‘high angle annular dark field’ or alloy in which a fine dispersion of oxides have
HAADF. been introduced, to strengthen the material
by interfering with the operation of slip created from secondary electrons that are
systems during material deformation. ejected from it.
Phase transition Thermally or mechanically Scanning transmission electron microscopy
induced change in crystallographic structure, (STEM) A variation of TEM in which the electrons
for example, from ‘face-centered cubic’ to pass through the specimen and the electron
‘body-centered cubic’ in iron or mild steel, or optics focus the beam into a narrow spot
‘face-centered cubic’ to ‘hexagonal close- over the sample in a raster. The rastering of
packed’ in cobalt-rich alloys the beam across the sample is used to
Pin-on-disk A test configuration used to study perform various analyses such as annular
sliding wear, with the ‘pin’ (or in some cases dark field (ADF) imaging, energy dispersive
‘button’ or ‘coupon’) as the sample material X-ray (EDX) analysis and electron loss
and the ‘disk’ as the counterface. The ‘pin’ spectroscopy (EELS) simultaneously,
may be held stationary against a rotating ‘disk’ allowing direct correlation of image and
to study unidirectional sliding wear, or moved quantitative data.
backwards and forward against a stationary Scanning tunneling microscopy (STM) An
‘disk’ to study reciprocating or fretting wear. imaging technique based on the concept of
Ploughing A general term, covering the three quantum tunneling. In STM a conducting
related terms of ‘ploughing,’ ‘wedge- surface is brought very near to the surface
forming’ and ‘cutting,’ being defined as the (metallic/semiconductor) when a bias
physical deformation due to a harder between the two allows electrons to tunnel
material ploughing through a softer material through the vacuum between them.
or rather the work required to carry this out. At low voltage the tunneling current is a
Ploughing is also referred to more function of local density of states (LODS) at
specifically as the physical deformation due the Fermi level of the sample. The changes in
to a harder material being pushed through a current accompanying the tip
softer material where adhesion forces are movement over the surface are translated
weak. The grooves created are shallower into an image.
than with ‘cutting’ or ‘wedge-forming,’ with Scanning tunneling spectroscopy (STS)
lower penetration of the harder asperity or A technique employed within an STM, to
third body into the softer material. Formation probe the local density of electronic states,
of wear debris particles cannot be clearly and band gap of surfaces and materials on
seen at the point of contact. surfaces at the atomic scale. The technique
Positron annihilation (PA) A major technique in allows observation of changes in constant
Materials Science, originally applied in current topographies with tip-sample bias,
condensed-matter physics, now widely used local measurements of IV curves and
in metals and alloys to provide information on tunneling conductance dI/dV. This allows
defect structures. investigations of small areas (5 Å diameter,
Retention The continued residency of wear debris the area in which the tunneling current flows).
at the wear interface, such that it may Scuffing Roughening of surfaces by plastic flow,
continue to play an active part in the wear whether or not there is material loss or
process. transfer.
‘Run-in’ wear A normally brief period of severe Selected area diffraction pattern (SAD) This
wear due to initial metal-to-metal contact technique is generally performed inside a
almost always observed before mild wear is TEM, in which electrons pass through a
established. thinned sample (see ‘TEM’). As the wave
Sample The material or surface undergoing wear length (a fraction of a nanometer) and the
testing. spacing between atoms are comparable, the
Scanning electron microscopy (SEM) A type of atoms act as diffraction gratings to the
microscope in which the surface of a sample electrons. Some of the electrons are
is scanned with a high energy beam of scattered to particular angles determined by
electrons. An image of the sample surface is the material crystal structures (allowing
identification and analysis) and form a spot Tribology The study of the science and
pattern image on the TEM screen. technology of interacting surfaces in relative
Severe wear The wear encountered typically motion, and including the related subjects of
where two metallic surfaces are in direct lubrication, friction and wear. Derived from
sliding contact, without an intervening the Greek verb tribo (‘tribo’), meaning ‘to
compacted oxide or ‘glaze’ layer, or other rub.’
protective coating. Such wear is ‘Two body’ wear The interaction of two wear
characterized by high coefficient of friction surfaces during sliding, in the absence of a
values, high levels adhesion, plastic third agent (i.e., wear debris) between the
deformation and to varying degrees (mainly wear surfaces.
metallic), material transfer between the Unidirectional sliding Where the sample and
surfaces. Also typical of severe wear is the counterface move in only one direction
generation of large, flat, angular, generally relative to each other.
metallic wear debris with sizes of up to 0.1 Van der Waals’ forces Attractive forces between
mm (or greater) due to ‘delamination.’ unbonded atoms and molecules (significant
Slip direction The direction on the slip plane with for very small particles of less than 5 mm in
the highest linear density. diameter).
Slip plane The crystallographic close packed Vickers hardness number (VHN) A parameter
plane in which dislocation movement can defining the hardness of an engineering
occur. material. The Vickers hardness number is
Slip system A combination of slip planes and that obtained by application of a predefined
directions defined by crystallographic load through a diamond pyramidal (or
arrangement (i.e., body-centered cubic, Vickers) indenter. The size of the indent
face-centered cubic, hexagonal close- created is used to obtain a measure of
packed). materials’ hardness.
‘Third body’/‘three body’ wear The interaction of Wear behavior, wear regime, wear mechanism
two wear surfaces during sliding, in The type of wear observed (i.e., mild or
combination with a third agent (i.e., debris). severe wear).
This third agent may be termed ‘active’ (it is Wear debris The material removed from sliding
retained between the wear surfaces and surfaces during wear. This may be metallic
modifies the wear process) or ‘passive’ (it is (dominant during severe wear) or oxide
ejected and has no effect). (dominant during mild wear).
Torque transducer A device for measuring Wear map, wear mechanism map A plot that
torque in a rotating system. In tribology, shows the effect on wear mechanism of two
changes in torque due to sliding surface or more sliding variables (i.e., load,
contact can be used to determine the temperature, sliding speed), allowing
coefficient of friction. prediction of wear mechanism where the
Transmission electron microscopy (TEM) values of these variables are known.
A widely used technique in Materials Science Wear rate The rate of material removal during
where an electron beam is transmitted sliding.
through a thinned sample placed in vacuum Wear volume The volume of material removed
with a typical energy of 100–400 keV. The during sliding.
electron beam is then focused by a series of Wedge-forming The physical deformation due to a
magnetic field lenses into a typical spot of harder asperity or third body being pushed
diameter 1–10 nm. A TEM image is created through a softer material where adhesion is
from those electrons that pass through the moderate. Material is pushed up ahead of
sample. asperities on the counterface, resulting in a
Tribocorrosion, tribo-oxidation The enhanced grooved wear scar with transverse cracks.
generation of corrosion product (normally The depth of groove created is lower than for
oxide) during sliding wear. The conjoint the ‘cutting’ model, but higher than for
action of tribology and corrosion/oxidation. ‘ploughing.’
forming the contacting surfaces, and changes in adhe-

Symbols sion between these surfaces caused by the joint action
d Distance for which sliding contact is maintained of temperature and tribological parameters. Efforts to
f Mass fraction of oxide prevent wear have included the use of oxidation resis-
p Mean pressure (load/area) tant and thermally stable materials, coatings, and mate-
r Distance affecting mutual action of Van der Waals’ rials with preoxidized surfaces.3–10 However, the
forces choices of coatings and materials are severely restricted
t i) Time of sliding by high temperature environmental conditions.3–5
ii) Time required to grow oxide layer of critical An alternative method of generating wear-resistant
thickness surfaces on coated and uncoated materials is to take
v Speed advantage of certain events that accompany the high
A i) True area of contact temperature wear process, such as oxidation, debris
ii) Hamaker constant generation, and elemental transfer between contacting
Ap Arrhenius constant surfaces.3–5,11 These events under certain conditions of
D Depth of wear temperature, pressure, and speed can lead to ‘glaze’
E Activation energy for bonding formation on contacting surfaces, enhancing resistance
F i) Force of attraction to further wear.2–5,11–30 Although ‘glaze’ formation and
ii) Contact frequency the general issues relating to wear at elevated tempera-
H i) Material indentation hardness tures have been extensively studied,2–5,11–30 it is still
ii) Indentation hardness difficult to predict the precise conditions that promote
Ka Probability of a wear particle being generated ‘glazed’ surface formation. Many aspects of high
Kp Oxidation rate (parabolic rate constant) temperature wear, including debris generation, debris
K1 ‘K factor’ (K1=Ka/H) oxidation, material transfer between surfaces, and
K1c Fracture toughness enhanced elemental diffusion, still need clarification.
L Sliding distance At present, there is incomplete understanding of the
P Applied load formation, nature, microstructure, and defect structure
Qp Activation energy for oxidation of high temperature wear-resistant surface ‘glaze.’
R i) Radius of particle of sphere This chapter is not intended to catalogue all pre-
ii) Gas constant vious work. Instead, the objective is to concentrate on
T, To Temperature in Kelvin (absolute temperature) generic principles, modeling, and scientific theories,
V Sliding speed which where possible will facilitate understanding of
W Work volume the high temperature wear process. There is a heavy
g Surface energy reliance on the information generated in the authors’
go Surface energy at standard temperature and own laboratory, where extensive use has been made
pressure of current microscopical techniques.
r Oxide average density The chapter is structured as follows: Section
t Length of time of each encounter (contact 1.15.2 deals with some general principles of wear,
between surfaces) Sections 1.15.3 and 1.15.4 consider published
j Critical oxide layer thickness research while Section 1.15.5 discusses some results
Dm Growth of oxide per unit area of nanoscale investigations of ‘glaze’ formation;
1/Ka Number of encounters (contacts between Section 1.15.6 concludes with some final remarks.
surfaces) required to generate an oxide layer
of critical thickness
1.15.2 Wear Theory
1.15.2.1 Early Wear Theory
1.15.1 Introduction 1.15.2.1.1 Archard and Hirst – Distinction
between mild and severe wear
High temperature wear is a serious problem in many In 1956, Archard and Hirst31,32 categorized wear into
situations, including power generation, transport, two groups: mild wear and severe wear.
materials processing, and turbine engines.3–7 Such ‘Mild wear’ occurs when the debris produced
wear is accentuated by faster surface oxidation kinetics, (generally oxide) prevents direct metal-to-metal con-
loss of mechanical hardness and strength of materials tact. Although Quinn33 does not specifically mention
the oxidation reaction in his review of oxidational Incoloy MA956 against Incoloy 800HT, in the work
wear when discussing the definition of mild wear, of Wood3 and Rose4 (these authors refer to these
the majority of studies into sliding wear have concen- alloys as MA956 and Incoloy 800); the 270 C case
trated on the oxidation reaction. Debris produced is of shown in Figure 1 is one such example. Debris parti-
very small size (<1 mm) and complete coverage is not cles tend to be large, flat and angular, up to 0.1 mm or
necessarily achieved, with oxide in many cases only greater. Contact resistance on surfaces that have
forming on load-bearing areas such as asperities. Elec- undergone severe wear tends to be very low, due to
trical contact resistance is high because of the pres- the exposure of the metallic surface.
ence of the oxide on the wear surfaces. Temperature affects the nature of wear as it influ-
The absence of such oxides allows contact between ences the kinetics of oxidation. Other factors such as
the metallic interfaces, with adhesion, plastic defor- relative humidity34 and partial pressure of oxygen
mation, and, to varying degrees, material transfer also affect the nature of wear as explained by Lan-
between the surfaces. This is typical of the ‘severe caster26 and Stott et al.14,35 Figure 2 shows a smaller
wear’ situation, examples of which have already been range of severe wear in an oxygen atmosphere at
observed in the lower temperature sliding wear of 300 C, compared to that in air.
270 °C 100 µm 750 °C 100 µm
Figure 1 Wear surfaces produced during the sliding of Incoloy MA956 against Incoloy 800HT at 270 and 750 C4 (load ¼ 7 N,
sliding speed ¼ 0.654 m s1, sliding distance ¼ 9418 m). At 270 C, a highly worn, heavily deformed surface is produced.
There is direct metal-to-metal contact, allowing high levels of adhesion accompanied by plastic deformation, material transfer
and the production of large flat angular debris. At 750 C, a compacted oxide layer has been created from the debris, giving
physical protection via enhanced hardness and also separating the metallic surfaces, preventing contact and adhesion.
1
20 °C in air –
10–4 to 10 m s–1 300 °C in air –
10–3 to 1 m s–1
Wear rate (mm–3 m–1)
10–1
10–2
300 °C in O2 –
5 ⫻ 10–2 to 0.4 m s–1
400 °C in air -
10–2 to 0.5 m s–1
10–3
10–4 10–3 10–2 10–1 1 10

Sliding speed (m s–1)
Figure 2 Variation in wear rate with sliding speed at 20, 300, and 400 C in air and also 300 C in pure oxygen for a/b brass
sliding against steel26 (approximate sliding speed ranges for severe wear at each temperature given in italics).
The model that Archard and Hirst31 proposed coefficient of friction and also in the variability of the
assumes a true area of contact, occurring between a coefficient of friction versus time trace. For example,
small number of asperities on the contacting surfaces. during early severe wear at 750 C, mean values typi-
The true area of contact can be calculated by eqn [3]: cally between 0.7 and 1 could be obtained for Nimo-
nic 80A versus Incoloy 800HT. On reaching the mild
A ¼ P=H ½1
wear stage, values of between 0.4 and 0.5 were typi-
where P is the applied load and H is the indentation cally obtained.
hardness of the material. If W is the worn volume and A direct link between hardness and wear rate is
L is the sliding distance producing the wear, then not always observed. Archard and Hirst proposed the
W/L is dependent on, and is therefore proportional theory of mild and severe wear discussed in this
to, the area of the friction junctions or true area of section, to resolve these difficulties.31
contact.
1.15.2.1.2 Classification by mechanism
W =L / A or W =L ¼ Ka L ½2
Many approaches to classifying wear have been
This gives: attempted; Quinn’s 1983 review33 lists a number of
these and tries to classify them under Archard and
W =L ¼ Ka P=H or W ¼ Ka PL=H ½3
Hirst’s mild and severe wear31 headings (Table 1). For
the dimensionless parameter Ka being the constant of example, Burwell and Strang38 propose seven different
proportionality and also the probability of a wear classifications of mild and severe wear, which Quinn
particle being generated. It is also referred to as the argues are actually special cases of Archard and Hirst’s
‘wear coefficient.’ An alternative form (K1 ¼ Ka/H ) is: mild and severe wear (with many of their classifica-
tions, including elements of both).31 Ludema39 talks in
W ¼ K1 PL ½4
terms of ‘scuffing’ (‘roughening of surfaces by plastic
31
K1 being referred to as the ‘K factor.’ Taking eqn [3] flow, whether or not there is material loss or transfer’)
and rearranging allows Ka to be expressed in terms of and ‘run-in,’ Quinn seeing ‘scuffing’ as a form of
wear depth, sliding velocity, and pressure. adhesive wear and ‘run-in’ as corrosive or mild wear
Dividing by the apparent area of contact gives: (due to the generation of oxides on the wear surface).
Finally, Tabor40 does not distinguish between adhesive
d =L ¼ pðKa =H Þ ½5
and corrosive wear, preferring to classify both as adhe-
where d is the depth of wear (volume divided by area) sive wear, with Burwell and Strang’s remaining
and p is the mean pressure (load over area). If v is mechanisms being referred to as ‘nonadhesive wear’
sliding speed and t is the time of sliding, L = vt, then (surface fatigue, abrasion, and fretting) or a mixture of
both (cavitation and erosion).
d =vt ¼ pðKa =H Þ or t ¼ dH =Ka pv ½6
The classification of wear mechanisms clearly
remains a matter of debate, although Quinn33 pro-
Ka ¼ dH =pvt ½7
poses that each of the forms of wear described should
The above implies that the level of wear is propor- be considered in terms of either mild or severe wear
tional to the sliding distance and applied load; Arch- in any given situation. However, it is the view of the
ard and Hirst showed this to be true over a limited current authors that it is not possible to talk simply in
range31 (Figure 3). terms of mild and severe wear. For example, when
The Archard model31 is effective assuming that material is lost by abrasion, the loss of material by the
there are no changes in the wear surface as a result of ploughing action of asperities on a wear surface does
the sliding process. However, changes do occur in not necessarily require adhesion or corrosion to
many cases, leading to changes in wear rate resulting remove material. Also, loss of material by delamina-
from little or no variation in experimental or operation wear is a fatigue-related process, caused by
tional parameters,27,28,36,37 and thus, changes in Ka repeated loading and unloading of surface layers as
value may be observed. Previous experimental work3–5 asperities of the opposite surface pass over it, assisting
has demonstrated that such changes can occur, with the propagation of subsurface cracks, eventually lead-
changes in friction being coincident with a switch ing to material loss. Although adhesion or reaction
from early severe wear to mild oxidational wear, with a corrosive environment may accelerate the
without any alteration of experimental parameters. process, again neither is necessary for fatigue and
This was usually denoted by decreases in the mean crack propagation to occur.
5
Linear or near linear relationship in all
B
cases – no significant deviations
Volume removed (10–3 cm3)

4
A
3 C
2
D
E
1
F
0
1 2
(a) Sliding distance (3 103 m)
A – mild steel with a 50 g load, B – ferritic stainless steel at 250 g, C – 70/30

brass at 80 g, D – stellite at 2500 g, E – hardened tool steel at 300 g, load 1 kg
3
Volume removed (10–3 cm3)
B
A
Wear rate (cm3 cm–1)
10–7
Linear relationship
up to ~2 3 103 g, then 2
10–8 deviation to higher Linear relationship
wear rate up to ~2 3 103 m,
followed by
10–9 1
reduction in wear
rate (or volume
removed)
10–10
0
10 102 103 104 0 1 2 3 4
(b) Load (g) Sliding distance (⫻103 m)
A. Wear rate against load for a ferritic stainless steel pin worn
against high speed tool steel ring, load 1 kg, speed 1.8 m s–1
B. Wear rate against sliding distance for a 70/30 brass pin worn
against hardened tool steel ring, load 1 kg, speed 1.8 m s–1
Figure 3 Experimental data from Archard and Hirst’s Work on like-on-like sliding at 1.8 m s1. Reproduced from Archard, J. F.;
Hirst, W. Proc. Royal Soc. London A 1956, 236, 397–410.
Table 1 Quinn’s comparison of the various classifications of wear
Burwell and Strang [25] Tabor Ludema Quinn [16]

Adhesive wear Adhesive wear Scuffing Severe wear
Corrosive wear 9 9 ‘‘Run-in’’ Mild wear
8
> >>
= >
> Surface fatigue>
>
< >
>
(pitting) >
>
>
=
> Abrasion >
> Nonadhesive wear Mechanisms of scuffing Mechanisms of mild and
>
: ;
Fretting >
> and ‘‘run-in’’ severe wear
>
>
Cavitation >
> Mixtures of adhesive and Not covered in
>
>
Erosion ; non-adhesive wear Ludema’s review
Source: Reproduced from Quinn, T. F. J. Tribol. Int. 1983, 16, 257–270.

The different behaviors of oxide debris are not At intermediate temperatures up to 630 C for
accounted for either. Inman et al.5,21–24 observed that Nimonic 80A versus Stellite 6 and 510 C for Incoloy
during sliding wear, an oxide could either be protective MA956 versus Stellite 6, debris was ejected as larger
or abrasive or even in limited quantities affect the severe metallic particles that did not remain at the wear
wear process. It was thus suggested that mild and severe interface and thus did not separate the two wear
wear could be subcategorized into ‘standard severe wear’ surfaces. This is an example of ‘passive’ debris.
(no oxide present, adhesion and delamination dominat- At higher temperatures, larger amounts of fine
ing), ‘abrasion-assisted severe wear’ (oxide abrasively oxide debris were generated; in the case of Incoloy
assists the adhesion/delamination dominated wear pro- MA956 versus Stellite 6, this again separated the inter-
cess), ‘protective mild wear’ (oxide prevents metallic faces and also formed protective compacted oxide or
contact and forms a protective layer as loose debris or ‘glaze’ layers. Nimonic 80A slid against Stellite 6 also
‘glaze’), and ‘abrasive mild wear’ (oxide debris prevents formed protective layers at 0.314 m s1; however, at
metallic contact, but remains loose and promotes wear 0.654 and 0.905 m s1, wear levels increased because
by abrasion).24 Another suggested classification system of oxide debris acting abrasively against the Nimonic
is that of Rabinowicz,41 who identifies abrasive and 80A surface. These latter observations are once again
fatigue wear, as well as adhesive and corrosive wear, examples of ‘active’ debris.
as distinct categories of wear in their own right. Each In addition to promoting mild wear, the negative
of these four categories of wear is now discussed. effect of abrasion has also been noted; as with hard
and soft surfaces in the two body wear models already
discussed, for a third body to have an abrasive effect
1.15.2.2 ‘Two and Three Body’ Wear
at the wear interface possibly leading to increased
1.15.2.2.1 Overview wear, it is normally expected that the hardness of the
Much of the discussion so far has been concerned third body will be at least 1.3 times greater than that
with what happens when two surfaces move relative of either of the contacting materials.33
to one another, generating wear debris. The consid- Active participation of third bodies has been noted
eration of wear without the interaction of debris is by other researchers, notably Iwabuchi et al.,43–45 who
referred to as ‘two body’ wear. However, the genera- studied the effects of the introduction of iron oxide
tion of debris particles introduces a ‘third body’ into (Fe2O3) particles to the wear interface, noting that,
the sliding process, which can then go on to have a where the particles were supplied under fretting test
significant effect on the wear process. conditions, the severe running-in wear volume for a
This debris may be retained within the interface standard carbon steel (S45C) was decreased 10-fold.45
area, where it may become an ‘active’ participant42 in Increased surface roughness also proved to be a posi-
the wear process or may be ejected immediately after tive factor in that the particles were more effectively
its formation, in which case, it is referred to as ‘passive’ retained. Introduction of a large enough quantity of
debris. Active debris tends to be fine and may be a Fe2O3 particles eliminated the severe wear running
mixture of metallic and oxide particles. On the other in stage.
hand, ‘passive’ debris particles are in general much Iwabuchi also studied the effects of particle size43
larger and, because of their immediate removal from and found that with a surface roughness of maximum
the wear interface on formation, may retain much of asperity height 20 mm, a particle size of 0.3 mm (the
their original form and structure. In metallic wear, smallest studied) was found to give the lowest wear
‘passive debris’ is more likely to be metallic. rate during mild wear. However, particle sizes of 1 mm
The work of Rose4 and Inman et al.5,21–23 clearly were observed to decrease the wear rate in the severe
illustrates this difference. For sliding of Nimonic 80A wear stage the most. Ideally, the same size particle
and Incoloy MA956 versus Stellite 6, at lower tem- would result in both; for practical use to be made of
peratures (390 C and lower at 0.654 m s1, 270 C such observations, a compromise needs to be made. It
and lower at 0.905 m s1), debris particles tended to is also to be noted that Iwabuchi expressed uncer-
be of a fine nature; these were largely retained at the tainty as to whether the 0.3 mm particle size is the
interface, commuted, and converted to a layer of most effective only for the 20 mm maximum asperity
loose, discrete (usually) oxide particles. The reten- height or for other asperity heights as well.
tion of this debris as a ‘third body’ then acted to keep The current authors suggest that the optimum size
the interfaces separated and wear values low. This is will vary and also change as the wear test proceeds,
an example of ‘active’ debris. for the simple reason that, as the larger asperities are
removed, until such time as an oxide layer forms, the temperatures for Fe–18% Cr–9% Ni stainless steel.
particle size that can be contained in the recesses will The effect of temperature was quite marked, with
also decrease. Conversely, during severe wear, rough- removal of the debris at room temperature leading to
ening of the wear surface may allow for a larger increased wear. At temperatures of 400 and 500 C,
particle size to be retained. A limiting factor is that formation of compacted oxide layers did still occur;
there will probably be an upper limit of particle size, this was considered to be due to the ready sintering
partially due to comminutation and ejection of larger and adherence of the loose oxide particles at these
debris and also a limit to the size of particle that can temperatures. The attractive effects of Van der Waals’
convert from metal to oxide. forces (acting to retain smaller particles between the
Iwabuchi also points out that the tendency to form wear surfaces) were also considered, as detailed in
a protective compacted layer of oxide is affected by Sections 1.15.2.2.3 and 1.15.2.2.4. Colombie et al.47
other parameters; varying test parameters such as load obtained similar results with increased wear on blow-
and sliding speed showed that, for certain conditions, ing nitrogen through the wear interface at frequent
the oxide particles acted as an abrasive and increased intervals; the more frequent the intervals, the greater
the wear rate. Figure 4 is a wear map of load versus the observed wear. Magnetic fields can also have an
amplitude, where different combinations of load and effect with ferromagnetic materials; Hiratsukam et al.48
amplitude produce different outcomes as regards the showed that the field direction could influence
influence of the introduced oxide particles.42 For very whether debris was retained (thus decreasing wear)
low loads and amplitudes, the introduction of the or ejected (thus increasing wear).
oxide particles has a positive effect, lowering the Rigney49 also questioned the validity of the tradi-
level of wear. For medium loads and amplitudes, tional models of wear if they do not take into account
the oxides have a negative effect and wear levels are third bodies resulting from transfer and mixing; in
higher because of the abrasive action of the oxides. addition to separating or screening ‘first bodies’ (sam-
Moreover, increasing load for moderate amplitude or ple and counterface), third bodies can have other
amplitude for moderately high load again results in a effects as they are able to flow and transmit load,
positive influence and wear levels are low. accommodate velocity gradients, and are also cre-
Leheup and Pendlebury46 took a different ated, destroyed, and regenerated during sliding.
approach, by the use of an interfacial air flow in a Resulting behavior can depend on dimensions, com-
like-on-like ‘cup-on-flat surface’ sliding test at various positions (may be the same as either of the ‘first body’
materials or a mixture of both), properties (the mate-
rials may undergo stresses close to their mechanical
limits), and hardness. Chemical composition is a key
10
O O X X
to the observed properties of the third body.
Despite the evidence, it is still the case that many
models of wear do not take into account the action of
8
O X O O the third body, with debris assumed to be ejected on
formation. Much work has been done using experi-
mental rigs in which debris is retained at the wear
Load (N)
6
X X X O interface, including fretting wear and much of the
pin-on-disk work.1,2,6,10,14,17–19,33,35,43,46,50–71 Even in
4
situations where ejection is favored, such as the
O X X X block-on-cylinder approach,3,4 debris has remained
2 at the interface and has played a significant part in the
promotion of mild wear.
O O X X
0
0 50 100 150 200 1.15.2.2.2 Surface films and preoxidation –
Amplitude (µm) Third body or not?
Also absent from the traditional models, is consider-
Figure 4 Effectiveness of wear reduction on S45C plain
ation of the effect of any surface films, adsorbed gases
carbon steel due to the introduction of Fe2O3 particles.42 (○)
Positive effect, oxide separates wear surfaces and wear is or other volatile species that may be present in most
low, () negative effect, oxide acts as an abrasive and situations. Clearly, the wear process will be affected
increases wear. by the nature of the surfaces.72,73
This is also true for preoxidized films,3,48,49,52 steel), there was nothing to suggest the generation of
where early metal-to-metal contact has been oxide debris and thus all interactions can be assumed
decreased or eliminated to varying degrees, followed to be metal–metal in their case. Hence, a direct
by earlier formation of compacted oxides in a number comparison between their work and that of Halliday
of cases. Although not a ‘third body’ in their own right and Hirst may not be wholly appropriate. However,
(they are directly attached to one or both ‘first body’ Rose4 and Inman et al.5,21,22,24 report high levels of
surfaces at the beginning of sliding), the sliding action wear for Nimonic 80A versus Stellite 6 at elevated
will create extra loose material from the surface oxide temperature, where only oxide debris was generated
layer, which will proceed to act as a third body where during all but the earliest part of the sliding process.
retention is preferred over ejection. Here, it is the oxide debris that acts as a third body
and aids material removal.5,21,22,24
1.15.2.2.3 Behavior of particles at the Rice et al.75 showed that the effects of debris para-
interface meters on coefficient of friction are more important
The behavior of third body material at the wear than asperity parameters. On modeling and compar-
interface goes beyond just acting as an agent for sepa- ing the influence of variations in density and size of
ration of the sliding surfaces or removing material particles at the interface with that of asperities on
from them. Rigney’s observations49 on load transmis- coefficient of friction, the sensitivity of coefficient of
sion, velocity gradients and debris particle creation, friction to changes in particle density and size was
destruction, and regeneration are one example of this. much the greater.
In addition, the particles may undergo varying When particle parameters alone were considered, it
degrees of motion at the interface because of the was observed that increases in density lead to a much
movement of the sliding surfaces. Halliday and faster increase in the coefficient of friction to a high
Hirst53 noted accumulation and ‘rolling’ of debris steady state value. Increased particle sizes also lead to
particles during fretting corrosion tests on mild increases in the coefficient of friction values, or con-
steel specimens, which they commented as being versely, smaller particle sizes are preferred for a lower
responsible for the decrease in friction during testing. friction regime. Of most interest in the work of Rice
They also noted that some sliding debris must be et al.,75 was the observation that, if particle density was
present, as the observed coefficient of friction would increased while particle size was decreased (which could
have been in the region of 0.002 for rolling alone, be regarded as analogous to particle break-down early
rather than the values near or below 0.05 observed in the wear process), then there was a sharp increase in
after the run-in phase was completed. Halliday and friction, followed by a steady decrease with time.
Hirst also established that the presence of the oxide It is to be noted that Suh and Sin,74 Rice et al.75 did
particles prevented wear due to welding (adhesion) of not consider the movement of debris, as the debris
the surfaces together. was ‘entrapped.’ However, their approach75 does
Conversely, Suh and Sin74 noted an increase in allow for estimation of friction where particle size
friction and wear by ploughing due to the presence means that movement is not a primary consideration;
of debris particles; this was confirmed by the removal Hesmat76 proposes a lower size limit of 20 mm for
of the debris, after which the coefficient of friction fell debris to be a contributor to abrasive wear. At higher
from a high ‘steady-state’ value and gradually rose values, the results of Suh and Sin74 and Rice et al.75
again as further debris was generated. It is to be com- are, therefore, more applicable.
mented here, that Suh and Sin used a unidirectional Debris movement for smaller particles is clearly a
crossed rotating-cylinder-on-stationary-cylinder con- key consideration, as the fall in friction due to, for
figuration, while Halliday and Hirst conducted tests instance, rolling invalidates any models that fail to
with an ‘oscillatory-cylinder-on-two-block-fretting- consider it. Jiang et al.19 proposed the following series
wear’ configuration. Thus, it would have been likely of mechanisms for wear debris movement under
that debris retention in Halliday and Hirst’s work sliding conditions, each mechanism accounting for
would have been much greater, allowing for the a different mode of perceived particle movement
observed generation of Fe2O3 oxide particles. (Figure 5).
Also, Suh and Sin’s work74 largely describes a
metal–metal interaction, with most testing being car- 1. Rotating: A particle is entrapped in front of another
ried out in a purified argon atmosphere. Even where fixed or immovable particle or object. It cannot
an inert atmosphere is not used (as with AISI 1020 move from its current position when impacted by
Load
Sliding direction
1) 2) 3) 4) Adhesion
Rolling Rotating Rolling

Skidding
A A A A
Interlocking
and fracture
Figure 5 Mechanisms of possible movement of particles during sliding of particulate materials.19 1 – Rotating, 2 – relative
skidding, 3 – rolling, 4 – adhesion or sintering affected rolling mechanism.
an asperity from the opposite surface and is only However, it is to be pointed out that the debris
able to rotate around its own center. particles produced using the block-on-cylinder con-
2. Relative skidding: A fixed or locked grain in one half figuration,3,4 in the 5–20 mm size range specified by
of the sliding pair slides or ‘skids’ against a fixed Hesmat for flow, were noted to be highly irregular in
grain attached to the opposite surface. shape, being ‘flat and angular’ as a result of delamina-
3. Rolling: The particles are able to freely move or roll tion wear. Also, they were metallic, meaning that
along with the opposite sliding surface (and are adhesion between metallic surfaces would also be a
not entrapped by it as with the ‘rotation’ major influence; Hesmat’s work was with oxide parti-
mechanism). cles. The combination of these influences would thus
4. Adhesion or sintering affected rolling mechanism: Adhe- inhibit any flow process and eliminate the quasihy-
sion developing between adjacent particles hin- drodynamic flow region; high levels of metallic wear
ders the rolling of particles, affecting (increasing) were observed3,4 during the ‘severe wear’ stages, with
friction compared with the ‘free’ rolling associated high levels of adhesion due to transfer and
mechanism. back-transfer of material between metallic interfaces.
Where oxidational wear did occur, particles were
The mechanisms of movement can change and any
5 mm in size or less, when, as Hesmat points out, Van
particle may be subject to a different mechanism at
der Waals’ and other attractive forces and compaction
different time points. It is also possible that a particle
come into play. This means that quasihydrodynamic
will become ‘interlocked’ or ‘entrapped’54 and be
flow cannot play a significant part in the sliding wear
unable to move, and thus be liable to fracture, due to
process as observed between the selected test alloys in
a build-up of stress upon it. Further appraisal of these
the current configuration, favoring ejection of mate-
mechanisms54 does not specifically mention the
rial from the wear interface. Hesmat’s lower limit of
fourth mechanism, adhesion and sintering general
20 mm for abrasive wear could, therefore, be revised
factors affecting the resulting particle layer as a whole.
downward to the 5 mm upper limit for forces of attrac-
It is also possible that for a given range of particle
tion and compaction.
sizes, powder ‘flow’ may occur. Hesmat76 observes
that particles of an intermediate size, between 5 and
20 mm, may undergo ‘quasihydrodynamic flow’ under 1.15.2.2.4 The effect of forces of attraction
sliding conditions. Larger particles will behave as between third bodies
in situ elastic bodies and contribute to abrasive wear, As already noted,76 attractive forces such as Van
while smaller particles will compact and behave as a der Waals’ forces become a factor when considering
nearly solid body, moving along with the ‘first body’ very small particles (for oxides, 5 mm or less). Elec-
interface and protecting it. At these smaller sizes, trostatic forces between particles, and between par-
Van der Waals’ forces and other attractive forces ticles and tribosurfaces, also exist, but these are only
will become important factors and inhibit the flow about 1% of Van der Waals’ forces for nonconductive
process. solids.77
If F is the force of attraction and assuming the traditional fine particles of around 5 mm did not
particle is spherical, this can be directly related to show a significant response (in terms of shrinkage)
particle size by: until above 500 C. This does not mean that sintering
AR did not occur with the larger particles; it is more that
F¼ ½8 a greater response was obtained for the smaller par-
6r 2
ticles because of their greater relative surface area,
where A is the Hamaker constant; R, the radius of the and hence, the contact area. With particles of 300–
particle or sphere; r is the equilibrium separation. 400 nm, a ready sintering response may be detectable
The Hamaker constant50 is dependent on the sur- in a wear situation.
face energy of the sphere: The effect of shrinkage of the compacts, as occurs
A ¼ 9pr 2 g ½9 in all powder compacts on sintering, is worth noting.
As sintering and thus shrinkage undoubtedly occur in
where r is the distance affecting mutual action of accumulations of wear debris at not-necessarily-
Van der Waals’ forces (which can be estimated at greatly-elevated temperatures, this will no doubt
0.3–0.5 mm), and g is the surface energy. have an influence on the formation of compacted
Van der Waals’ forces are extremely short ranged oxide layers during the wear process. To date, no
in nature, produced by dipole induction between attempt has been made to address this issue or its
neighboring bodies. This accounts for the attraction importance on the sliding wear process.
observed between the smaller wear particles
(<5 mm)76 and the wear surfaces.78
1.15.2.3 Mild Wear and Mechanisms of
Surface energy is temperature-dependent, thus
Compact Oxide Formation
making the Hamaker ‘constant’ and the force of
attachment F temperature-dependent. An Arrhenius 1.15.2.3.1 Introduction to compacted oxides
relationship exists between surface energy and tem- or ‘glazes’
perature: As stated in Section 1.15.2.1, Archard and Hirst31,32
defined mild wear as that occurring where the surface
E
g ¼ g0 exp ½10 is extremely smooth and consisting partially or
RT wholly of the product of reaction between the mate-
where R is the gas constant; E, the activation energy rial under sliding and the surrounding atmosphere or
for ‘bonding’; and T is the absolute temperature. fluid. An oxidation reaction is required for the crea-
The amount of work done in relation to wear is tion of the corrosion product, although, as already
fairly minimal, compared with the formation of com- stated, the corrosion product itself need not neces-
pacted oxides and ‘glazes,’ as discussed in Section sarily be an oxide.
1.15.2.3, with the work of Hesmat76 being one of the The term ‘glaze’ is misleading, as it implies a glassy
more comprehensive studies of the potential behav- amorphous material, which was thought to be the case
ior of debris and particles at the wear interface in until Lin et al.6 proved it to be crystalline by means of
relation to their size. It is suggested, here, that the electron diffraction; the term ‘compacted oxide layer’
role of Van der Waals’ forces in the accumulation and is a more accurate description. ‘Glaze’ and ‘compacted
grouping together of particles is perhaps more oxide’ tend to be used fairly interchangeably through-
important, especially at more elevated temperatures. out literature, and despite the crystalline nature of
For sintering and fusion of finer oxide particles to these surfaces, the term ‘glaze’ has remained in com-
form glazes, an attractive force must be present to mon use. More recently, Datta et al.15,80 have demon-
hold the particles in position long enough to allow strated that the compacted oxide layers formed in
significant sintering reactions to occur. some systems and under certain conditions have a
The importance of sintering in the formation of nanocrystalline structure and are highly disordered
debris layers in itself cannot be ignored. Zhou et al.79 as discussed further in Section 1.15.4.3.
showed that for iron, nickel, and iron–25% nickel Compacted oxide layers tend to form under con-
powders, significant sintering can occur on raising ditions of moderately high temperature, low loading,
the temperature above room temperature (denoted and usually low sliding speed. For a compacted oxide
by the significant shrinkage of specially prepared layer to form and a resultant decrease in wear to
compacts), where the particle size is defined as ‘ultra- occur, many researchers specify that a minimum
fine,’ in this case, 30–40 nm. In comparison, more temperature that is dependent on conditions and
alloy composition must be exceeded. With increasing tangentially on them, and suffer failure and break
temperature, the formation of these compacted oxi- away. Combining Quinn’s and Archard’s models31
des becomes more rapid, with a consequent decrease allows the critical thickness to be used to relate
in the ‘severe wear’ running in period, the debris from wear rate to material oxidation rate, showing an
which can be a major source of material needed for increase in wear rate with temperature.
the formation of the compacted oxide layers. Com- From Archard’s wear equation [5], the wear vol-
pacted oxide formation is not restricted to low sliding ume W is related to the applied load P and the total
speed, as the effect of frictional heating due to high sliding distance L, giving
sliding speed can raise the temperature above the
Ka PL P
critical temperature required for the formation of W ¼ or W ¼ Ka AL A ¼
the compacted oxide layers. H H
Many researchers have extensively studied the de- Ka can be regarded as the probability of a wear
crease in wear by the formation of compacted oxides at particle being generated in any given encounter
high temperatures. In addition to Archard and Hirst,31,32 (assuming unit sliding distance in the following deri-
others such as Lancaster,26 Bhansali,81 Razavizadeh and vation, making W the wear rate). Thus, 1/Ka is the
Eyre,82 Stott et al.,2,6,10,14,17–19,35,50,55–61 Quinn1,33,62–64 number of encounters required to produce a wear
and Garcia,83 have contributed to the knowledge of particle.
compacted oxide formation. Recent work3–5,20–25 has In oxidational wear, a wear particle cannot be
additionally concentrated on the sliding of dissimilar produced until the critical thickness for mechanical
interfaces and their effects on the mode of wear. The stability x is reached; thus 1/Ka is the number of
key issues on oxide formation are discussed as follows. encounters required to generate an oxide layer of
Wear that is dependent on the tendency to oxi- this thickness. If t is the time required to grow the
dation is highly dependent on temperature. The layer and t is the length of time of each encounter,
tendency to oxidation with respect to temperature then
can be described by the following Arrhenius type
t
equation: t¼ ½12
Ka
Qp
Kp ¼ Ap exp ½11 If V is the sliding speed and d is the distance along
RT0
which the sliding contact is maintained, then
where Kp is the oxidation rate, Ap is the Arrhenius
t
constant, Q p is the activation energy (for oxidation), t¼ ½13
R is the gas constant, and T0 is the absolute tempera- V
ture in Kelvin. and eqn [14] can be modified to
Stott in conjunction with Lin and Wood,6,55–59 and
t
later Glascott and Wood10,50,54,60,61 proposed various t¼ ½14
VKa
modes and mechanisms of ‘glaze’ formation based on
the generation and behavior of oxide debris at the In a given time period t, the growth of oxide per unit
sliding interface. Jiang et al. building on studies of area will be Dm. Assuming parabolic oxidation
third body interactions19 (later revised by Stott54)
and adhesive processes between particles,17 proposed Dm2 ¼ Kp t ½15
a descriptive model for debris behavior at the sliding where Kp is the parabolic rate constant. If f is the
interface,2 based on their generation, commutation, mass fraction of oxide, that is, oxygen, and r is the
sintering to form ‘glaze’ layers (above a critical tem- oxide average density, then
perature), layer break-up, and debris removal. These
mechanisms and modes of formation are discussed in Dm ¼ f xr ½16
more detail in Section 1.15.4. and
1.15.2.3.2 Quinn’s oxidational wear model f 2 x2 r2 ¼ Kp t ½17

62
Quinn’s model of wear assumes a dependence upon Substituting eqn [2] into eqn [16] gives
the wear rate of oxide film formation on the wear
dKp
interface. At a critical thickness x, the oxide layers Ka ¼ ½18
are no longer able to withstand the forces acting Vf 2 x2 r2
The rate constant can normally be calculated using The slower heat flow from the wear interface and
an Arrhenius relationship, as described by eqn [13]: thus higher interface temperature, caused by lower
thermal conductivity of the oxides;
Qp The presence of multilayer scales (formed partic-
Kp ¼ Ap exp
RT0 ularly at high temperature, e.g., with iron), pre-
venting the calculation of the temperature based
where Ap is the Arrhenius constant, Q p is the activa-
on the oxide type generated33;
tion energy (in this case for oxidation), R is the gas
The expected presence of increased levels of sur-
constant, and T0 is the absolute temperature of
face defects at the contacting surfaces, increasing
reaction. Thus
the oxidation rate constant.

dAp exp Q p =RT0 The continued removal of oxides, leading to expo-
Ka ¼ ½19 sure of fresh surfaces and a decrease in oxide
Vf 2 x2 r2
thickness (hence diffusion distance), thereby
Substituting into Archard’s equation31 gives a final enhancing the tribooxidation rate; and
expression for wear rate: The presence of defects such as cracks and grain
boundaries, allowing ingress of oxygen (in ion and
dAAp exp Q p =RT0 molecular form), leading to an enhanced oxidation
W ¼ ½20
Vf 2 x2 r2 rate.
This expression relates directly to the oxidational A schematic presentation of Quinn’s models84 is
properties of the materials, plus the critical environ- shown in Figure 6.
mental variables affecting the wear process, such as
temperature at the interface at the time of contact
1.15.2.3.3 Modification of Quinn’s oxidational
and sliding speed relative to the opposing interface,
wear model for discontinuous contact
as well as the critical thickness of the oxide. Knowing
The validity of Quinn’s model1 is conditional upon
the critical thickness of the oxide62 and the static
the maintenance of contact between the two surfaces,
oxidation properties of the wearing materials, wear
which is clearly often not the case. Quinn’s model was
rate prediction can be easily achieved.
further modified by Garcia et al.83 to take into
However, there are many complications that can
account discontinuous contacts between the contact-
prevent the accurate prediction of the wear rate:
ing surfaces.
The difficulty in estimating the wear interface For this modified model, it is assumed that 1/Ka
temperature, particularly in the presence of oxides; contact events are required for x, the critical oxide
Metal
C Oxide
Cracks or grain
boundaries B
B⬘ Oxygen
C⬘ Oxide
Metal
Figure 6 Oxygen transport between oxide plateaux and cracks in the oxides.84 Quinn’s oxidation model supposes that
any oxygen species would have to diffuse right from the surface of the oxide (B or B0 ) to the metal. The presence of cracks
and grain boundaries act as points of ingress for oxygen ions and in the case of cracks, molecular oxygen. This means that the
distance for diffusion is significantly less where cracks in the outer layers of oxide are prevalent. Diffusion need only take
place across the underlying oxide from the crack tips (C and C0 ) to the metal and thus the rate of oxidation is greater.
thickness, to be achieved; this is the same as for approach provides a good approximation in cir-
Quinn’s model. However, in this case, the time cumstances where Quinn’s model is less effective
required to reach this critical thickness depends on in predicting wear rates (i.e., where contact is not
the contact frequency F, which is the inverse of the maintained).
elapsed time between two contacts at a given point
between the contacting surfaces; while this can
1.15.2.4 Effect of Load and Sliding Speed
clearly be related to the sliding speed, the frequency
of contact can also be changed by varying the length 1.15.2.4.1 Early work
of the wear track without any need to vary the slid- A significant contribution to the effects of load and
ing speed. It is clear that each asperity will not sliding speed came from Welsh’s work27,28 on low-
make contact each time the disk rotates; however, carbon steels using a pin-on-rotating-ring (cylinder)
Garcia83 comments that the probability of a con- configuration with applied loads of up to 2 kgf and
tact (and hence a wear particle being generated) sliding speeds of up to 2.66 m s1. The wear process
is included in the statistical meaning of the wear was characterized by two transitions: T1, which is
coefficient, Ka. mild-to-severe at low load and sliding speed, and
Therefore T2, which is severe-to-mild at high load and sliding
speed. Increasing the sliding speed decreased the
1=Ka 1
t¼ ¼ ½21 critical load at which these transitions occurred
F FKa (Figure 7), the lower transition being eliminated in
The constant of proportionality Ka in Archard and some cases, leaving only the severe wear to mild wear
Hirst’s model31 is thus defined by transition. In extreme cases, these transitions could be
decreased enough to be eliminated from the experi-
Kp
Ka ¼ 2 x2 r2
½22 mental data – mild wear could be observed over the
Ff whole range. The variation in the upper transition
and substituting the rate constant as defined by eqn from the intermediate severe wear back to higher
[11] gives speed mild wear was observed to be the more sensi-
tive to the sliding speed.
Low-speed, low-load mild wear was attributed to
Ap exp Q p =RT0
Ka ¼ ½23 the presence of loose oxide debris at the sliding
Ff 2 x2 r2
interface and intermediate severe wear to direct
metal-to-metal wear. The mild wear encountered at
Substituting into Archard’s equation [5]
high speed and high load was attributed to hardening,
accompanied by the development of an adherent
Ka PL P
W ¼ or W ¼ Ka AL A ¼ oxide film, as a result of frictional heating. The hard-
H H
ening came about as a result of the low-carbon steels
gives a final expression for wear rate (again as the undergoing phase changes, due to high localized tem-
sliding distance L is assumed to be 1, the wear volume peratures around points of contact being sufficient
W can be taken to represent the wear rate), the to produce a transformation to austenite, followed
frequency of contact events being the determining by rapid cooling by conduction of heat into the bulk
parameter for the input of energy for oxide growth, metal producing a structure at the surface not too
rather than the sliding speed: dissimilar to martensite.
A critical hardness had to be exceeded by these
AAp exp Q p =RT0 phase changes for mild wear to be reestablished
W ¼ ½24
Ff 2 x2 r2 under high-speed, high-load conditions; the transi-
tion back from severe wear to mild wear is in fact a
The validity of Garcia’s model83 is dependent upon a two part transition, with T2 referring to the point
wear particle being generated on any given contact. where sufficient phase hardening occurs to suppress
For this to work, it is necessary for a wear particle to severe wear without the intervention of an oxide film
be generated each time a contact is made; this in itself (the development of which further acts to protect the
is highly improbable. However, the lower level of data wear surface) and a T3 transition approximately
scatter achieved by Garcia’s discontinuous contact matching the point where permanent phase change
model indicates that this alternative frequency-based hardening occurs.
T2 T2
10–6
T1 T1
Wear rate (cm3 cm)

10–8
Increasing Increasing
10–10 speed speed
(a) (b)
10 102 103 104 10 102 103 104

(a) Load (g)
Figure 7 Effect of sliding speed on wear rate/load – 0.52% carbon steel.27 (a) 1.00–2.66 m s1 (● – 1.00 m s1; – 1.33 m s1;
○ – 2.00 m s1; + – 2.66 m s1), (b) 0.17–1.00 m s1 ( – 0.017 m s1; ○ – 0.067 m s1; 5 – 0.33 m s1; + – 0.66 m s1;
● – 1.00 m s1). T1 – Transition from low speed, low load mild wear to severe wear. T2 – Transition from severe wear to high
speed, high load mild wear. (Arrows show movement of transitions with increasing sliding speed).
Subramanian85 conducted a series of sliding tests delamination or plastically deformed material at higher
of an Al–12.3 wt% Si alloy in pin form against various speeds. It is not stated whether the particles produced
rotating ‘ring’ counterfaces, including mild steel in at any particular speed are metallic or oxide.
the rolled condition, quenched and tempered die The critical speed was observed to be dependent
steel and copper with varying levels of aluminum, on counterface material and a higher transition was
during which the sliding speed was increased at vari- noted for harder, more thermally conductive alloys.
ous times. The wear rate of the Al–12.3 wt% Si alloy Decreased mutual solubility also led to a higher
pin (Figure 8) decreased with increased sliding speed transition speed.
up to a critical value of usually 1 m s1, regardless of Welsh27,28 discusses the existence of lower limits of
counterface material or applied pressure used. Fur- load and speed, marking the transition from mild to
ther increases in speed above this critical value led to severe wear and also an upper limit, marking the
progressive increases in wear. transition back to mild wear. So,86 on the other hand,
Subramanian explains the decrease in wear with discusses only a single limit or critical value for both
increasing sliding speed is due to increasing strain load and speed for the transition from mild wear at low
rates and due to increased hardness and flow strength speeds and loads, to severe wear at high speeds and
of the wear surface. The true area of contact is thus loads. Comparison with the work of Welsh would make
and with a lower level of contacts between the wear- So’s transition equivalent to the lower transition, with
ing surfaces, a lower wear rate results. In competition no mention of an upper transition despite the use of
with this is the effect of increased temperature higher sliding speeds. So quotes values of 400 C and
because of frictional heating (which was observed to 5 MPa contact pressure as being limiting conditions
occur), softening the material at the wear interface. for mild wear for many steels. In one test of note, a
This results in an increase in the true area of contact high carbon steel sample underwent a mild-to-severe
and thus an increase in the wear rate. Subramanian wear transition at a contact pressure of 4.43 MPa, on
does not go into detail with his reasoning; however, raising the sliding speed from 3 to 4 m s1. A Stellite
the softening of the material must allow for deforma- sample remained in the mild wear state at a contact
tion and ‘spreading’ of asperities and also increased pressure of 8.85 MPa (just under twice as much
contact at other points. pressure) under similar conditions. As So used a
Changes in wear mechanism were observed at Sub- pin-on-disk configuration, compared to the pin-on-
ramanian’s ‘1 m s1’ transition, with equiaxed particles rotating-cylinder configuration of Welsh, this may
produced below this critical or ‘transition’ speed, com- account for the differences in results; the pin-on-disk
paction of these particles and delamination of the configuration may not have generated sufficiently
compacted particles around the transition speed, and severe conditions for the upper transition to occur.
10–10 10–10
Strain rate Pressure 0.1 MPa Pressure 0.1 MPa

effects dominant Strain rate
effects dominant
10–11 10–11 Thermal
Wear rate (m3 m–1)

Thermal
effects
effects
dominant
dominant
PSZ
Cu
Mild steel
Cu–4.6% Al
10–12 Cu–7.5% Al 10–12
Die steel
10–13 10–13
0.1 1 10 0.1 1 10
(a) Sliding speed (m s–1) (b) Sliding speed (m s–1)
10–10 10–10
Strain rate Pressure 0.5 MPa

Strain rate Pressure 0.5 MPa effects dominant
effects dominant Mild steel
Thermal
10–11 effects 10–11 PSZ Thermal
dominant effects
Cu
dominant
Cu–4.6% Al
Cu–7.5% Al
10–12 10–12
Die steel
No transition
10–13 10–13
0.1 1 10 0.1 1 10
(c) Sliding speed (m s–1) (d) Sliding speed (m s–1)
Figure 8 Effect of sliding speed on wear rate of Al–12.3 wt% Si versus various counterface materials.85 (a) Cu, Cu–4.6 wt%
Al, and Cu–7.5 wt% Al counterfaces, applied pressure 0.1 MPa, (b) mild steel, die steel, and partially stabilized zircona
counterfaces, applied pressure 0.1 MPa, (c) Cu, Cu–4.6 wt% Al, and Cu–7.5 wt% Al counterfaces, applied pressure 0.5 MPa,
(d) mild steel, die steel, and partially stabilized zircona counterfaces, applied pressure 0.5 MPa.
Most load and sliding speed work done to date has Incoloy MA956 was worn against Stellite 6 or Incoloy
concentrated on what happens at room temperature, 800HT, at 25 N, the ‘glaze’ layer formed on the Inco-
with little work at elevated temperature. One exam- loy MA956 was beginning to show signs of breaking
ple is the work of Rose,4 where a series of experi- away (as described later).
ments were conducted at 750 C with loads between
7 and 25 N, during which an apparent transition from 1.15.2.4.2 Wear of cobalt-based alloys
mild-to-severe wear was noted at 15 N, when Nimo- Valuable information can be gleaned from work on
nic 80A was worn against Incoloy 800HT. Also, when cobalt-based alloys. The most comprehensive work
carried out on the sliding wear of cobalt was that by The formation of compacted oxide layers dur-
Buckley,87 who compared the sliding wear of cobalt ing like-on-like fretting wear tests of cobalt-based
with that of copper in vacuum. Lower friction and Stellite 31, between room temperature and 800 C
adhesion levels noted for cobalt were attributed to its (293–1074 K), was observed by Stott et al.90 The for-
hexagonal close-packed structure, compared to cop- mation of these oxides was by a similar route to that
per’s face-centered cubic structure. observed for iron-based and nickel-based alloys, with
The observed increase in friction and cohesion for alloying components present in the oxides to roughly
cobalt tested at temperatures 300 C, with com- the same proportions as the original alloy.
plete welding at 450 C, was considered to result At temperatures between room temperature and
from a phase change from hexagonal close-packed 300 C, Stellite 31 undergoes a much lower level of
to face-centered cubic because of increased tempera- wear than do various nickel- and iron-based alloys,
ture (to above 417 C; the phase change temperature which Stott also attributes to the smaller number of
for cobalt), caused by frictional heating. slip planes in the hexagonal close-packed structure of
The differences between the sliding behavior of cobalt. He specifies an initial period of low wear for
metals in hexagonal close-packed phase and face-cen- up to an hour, followed by the production of a bright,
tered cubic phase are due to the greater number of active rough metallic wear scars showing characteristics of
slip systems available in face-centered cubic structures. abrasion and evidence of material transfer. This he
There are 12 primary slip systems within a typical face- attributes to a probable change in phase from hexag-
centered cubic metal (4 slip planes each with 3 slip onal close-packed to face-centered cubic and thus a
directions), which are all crystallographically similar, loss in wear resistance. Later, the bright worn surface
compared to only three primary slip systems in cobalt, is lost with increasing amounts of oxide being pro-
these being based on the basal plane with the highest duced, although the load-bearing areas remain
atomic density (i.e., 1 slip plane with 3 slip directions). metallic.
Cross slip is also more difficult, as with hexagonal close- Both the ‘phase changes’ and oxide production
packed structures such as cobalt, screw dislocations are have been attributed to temperature increases at the
required to move out of the primary basal glide plane wear interface. However, the phase transformation
onto planes that, unlike face-centered cubic structures, temperature for cobalt is 417 C; Stellite 31 contains
are crystallographically different. Hexagonal close- 26% chromium, which has the effect of significantly
packed materials are thus less deformable. raising the hexagonal close-packed to face-centered
The effect of these hexagonal close-packed struc- cubic transition temperature; 20% Cr is estimated to
tures on wear was further elaborated on by Pers- raise the transition temperature to 847 C (expressing
son88,89 on studying the low friction tribological uncertainty as to the effects of the other alloying
properties of Stellite 21 and Stellite 6. Persson com- components). This suggests a far greater influence
ments that although a metastable face-centered cubic due to frictional heating and localized flash tempera-
structure may be retained in both Stellite alloys down tures due to asperity interactions; for a phase change
to room temperature, transformation to a hexagonal to readily occur, the temperature at the immediate
close-packed structure can be induced by application interface would have to reach 500 C above ambient.
of sufficiently high shear stresses (i.e., during sliding). In a fretting wear situation, it is difficult to see how
A thin, easily sheared layer can develop at the sliding this could occur.
surface due to the shear-induced alignment of the However, the level of alteration of temperature for
hexagonal close-packed basal plane parallel to the any phase transitions will also depend on the effects
direction of sliding.89 This alignment significantly of other alloying components in cobalt-based alloys.
reduces friction and improves galling resistance, As already stated, chromium will raise the transition
with shear and adhesive transfer restricted to this temperature quite dramatically. Other works68,87–89
layer. This sliding regime persists even as this layer suggest that tungsten and molybdenum also raise this
is removed, as it is easily regenerated. Also, the transition, while nickel and iron (also, magnesium
removal of material in such thin sections may at and carbon88,89) have the effect of stabilizing the
least in part explain the ready generation of fine higher temperature face-centered cubic structure
Stellite 6-sourced Co-based oxide debris observed (due to increases in stacking fault energy88,89) and
elsewhere.3–5,20–25 These sections may be more easily suppressing this transition. It is possible that the
commutable to a small size and oxidized, providing a presence of nickel to 10.5% and iron to 2% may be
ready supply of material for ‘glaze’ formation. sufficient to retard the effect of the chromium and
7.5% tungsten to a much lower level. Thus, a much to 156.8 N and sliding speeds of up to 4 m s1 used.
smaller increment in temperature due to frictional After sliding for up to 10 000 m, the Stellite 6 layer
heating and flash temperatures may be needed to was observed to be mostly covered by an oxide layer
effect any phase transition, confirming the conclusion reported to consist of W3O, CrO, and Co2O3. Where
that the damage observed can be attributed to phase this oxide layer spalled, a new oxide film was observed
changes, and hence, a decrease in resistance to to replace it readily. This applied for all combinations
deformation. of loads (19.6, 39.2, 78.4, and 156.8 N) and sliding
Stott et al.90 observed that compacted oxide for- speeds (1, 2, 3, and 4 m s1), with the exception of
mation was not accompanied by any decreases in the most severe tested combination, 156.8 N and
friction observed during similar experimentation 4 m s1, when severe wear was observed for the Stellite
with nickel-based alloys at elevated temperatures. 6 as the pin material. The experimental data obtained
With friction levels already low due to the hexagonal from these tests are presented in Figure 9.
close-packed structure, it is possible that no signifi- The steels underwent increased wear compared to
cant difference between the friction levels before and the Stellite 6, despite being of much greater hardness,
after elimination of metallic contact by higher tem- and So et al. concluded that the oxide layer formed on
perature ‘glaze’ formation may be discerned. the Stellite 6 must be tougher than that formed on the
The face-centered cubic to hexagonal close-packed steels. When used as a disk material, only a thin layer of
transformation observed in Co-based alloys is con- oxide material was formed on the AISI 4340, compared
sidered a martensitic transformation.91 Attempts to to the thicker layer formed on the Stellite 6; the wear
utilize Co-free hard wearing alternatives making rate of the steel was seven times that of the Stellite 6
similar use of the austenite–martensite transition in laser-clad pin. As the pin material, severe wear was
Fe-based alloys have been limited by typical upper observed for the AISI 4140 steel, the rate of wear
useful temperatures of 180 C88,89 (not discussed in being 10 times higher than that of the Stellite 6 laser-
the current work). clad disk. For all but the highest load, the wear rate of
the Stellite 6 pin actually decreased when the sliding
1.15.2.4.3 The effect of load and sliding speed was increased from 1 to 2 m s1, with only a slight
speed – Stellite 6 increase at intermediate loads on raising the sliding
Following on from previously discussed work speed to 4 m s1 (Figure 9). At the lowest load used,
(Section 1.15.2.4.1) with a Stellite material,86 So the decrease in wear continued up to 4 m s1.
et al. went on to test Stellite 6 clad mild steel against The increasing wear rate for specimens under a
AISI 4140 and 4340 steels in the martensitic phase, load of 156.8 N was ascribed to softening of material
first as the pin material (4.75 mm in diameter) and due to the higher flash temperatures encountered,
then the disk material.30 Measured hardness values especially at higher sliding speed. The higher flash
were 580 Vickers hardness number (VHN) for the temperatures also led to changes in the oxide phases
Stellite 6 layer and 750 VHN for the steels, increasing that were reported to form on the respective wear
to 970 VHN after an unspecified heat treatment surfaces. A shift was noted from Fe2O3 to FeO on the
followed by water quenching, with test loads of up steels and from W3O, through Co2O3, CrO, Cr2O3,
50 50
19.6 N 1 m s–1
39.2 N 2 m s–1
40 78.4 N 40
3 m s–1
W (10–13 m3 m–1)
156.8 N
W (10–13 m3 m–1)
4 m s–1
30 30
20 20
10 10
0 0
0 1 2 3 4 5 10 60 110 160
(a) Speed (m s–1) (b) Load (N)
Figure 9 Variation in wear rate (W) with sliding speed (a) and load (b) for the rubbing of laser-clad Stellite 6 pins with AISI
4340 steel disks. Reproduced from So, H. Wear 1996, 192, 78–84.
to Cr5O12 on the laser-clad Stellite 6 layer with higher the cobalt content, the better was the resistance
increasing temperature; this was accompanied by a to metal–metal wear at temperatures up to 750 C.
decrease in friction. So does not offer an explanation Where no cobalt was present within the alloy,
for the change in oxide with temperature, though as wear rates were observed to be highest. Above this
for the oxidation of iron, it appears that this can be temperature (at 1000 C), all alloys exhibited low
attributed to changes in oxidation state of the chro- wear with a protective oxide layer forming across
mium in Stellite 6, with preferential oxidation of the wear surface. Increases in wear were observed
tungsten and cobalt respectively at lower tempera- for all combinations with increased contact pressure,
tures. It is curious to note here that, in the work of though at high load, increases became less severe for
Wood3 and Rose,4 no such shift was observed for cobalt–chromium and cobalt–iron–chromium alloys.
Stellite 6, with Cr2O3, Co3O4, or a combined oxide Of particular note is the response to increasing the
of the two being consistently observed from XRD sliding speed by an order of 10 from the 7.06 104
results. No evidence of tungsten phases was found, m s1 used for all their other tests, to 7.88 103 m
although this could be attributed to the sensitivity of s1, carried out at 500 C and 20.69 MPa. For the high
the measurement and characterization equipment. cobalt–chromium alloys, including Stellite 1 and
Also of note was the fact that the oxide layers Stellite 6, there was a slight decrease in the observ-
formed in So’s work were more reminiscent of those ed wear rate. Where cobalt levels were low or non-
created at higher ambient temperature in the work of existent, the converse was true and increases in wear
Wood and Rose (>510 C), indicating extremely high were observed.
temperatures at the points of contact. So’s measure- Crook and Li69 attributed the superior wear
ments indicate a rapid rise in temperature with resistance of the cobalt–chromium alloys, first, to
increasing load and speed (speed having less of an the superior galling resistance, and second, to the
effect than load). So comments that, under the most tendency of alloys when in the face-centered cubic
severe conditions (156.8 N and 4 m s1), the mean form to undergo phase changes and become hex-
surface temperature at the point of contact reaches agonal close-packed, which as discussed earlier is
over 700 C and, because of this, wear becomes severe less prone to deformation, due to a smaller number
due to softening. This may be more to do with the of available slip planes. Conversely, they point out
load and speed conditions than to temperature (even that high nickel alloys have a poor galling resis-
accounting for phase transitions), as Wood and Rose tance, yet specifically quote the work of Stott
tested Incoloy MA956 and Nimonic 80A against et al.55,56,58 as examples of nickel–chromium alloys,
Stellite 6 at an ambient temperature of 750 C, with in particular, exhibiting low levels of wear and
oxide layers being obtained in both cases on the developing ‘glaze’ during the wear process at high
Stellite 6 counterface. temperature.
This implies that So has underestimated the In both experimental programs, a low amplitude
temperature at the sliding interface in this case. The ‘button-on-disk’ system suitable for fretting wear
softening may again be attributable to phase changes studies was used. However, Stott et al. concentrated
from hexagonal close-packed to face-centered cubic. solely on one material (Nimonic 80A), whereas
In the case of Stellite 6, there are far fewer alloying Crook and Li’s comparative work on a range of alloys
additions to offset the effects of chromium (present at showed that, although wear was still low for nickel–
27%) and tungsten (5%) on this transition. The tran- chromium alloys, the wear resistances were inferior
sition between the two phases for 27% chromium is to those of cobalt-containing alloys. In both cases, it is
880–900 C,5,92 almost 200 C higher than So’s not possible to say that in an extreme high wear
700 C estimate. environment (e.g., high speed, high load) similar
Crook and Li69 carried out a comparative ‘like-on- observations of low wear would be made. In the
like’ sliding study of Stellite 6 and a number of case of Stellite 6, So’s work30 does indicate continued
other hard-facing alloys of various cobalt contents, low wear rates during moderately high speed, high
including Stellite 1 (with higher levels of carbon, load unidirectional sliding wear (up to 156.8 N and
chromium, and tungsten than Stellite 6), Stellite 4 m s1, with frictional temperatures of up to 700 C
2006 (a 33% cobalt–iron–chromium alloy), Haynes being generated); however, if the works of Wood3 and
No. 716 (a nickel–iron–chromium alloy with 11% Rose4 are considered, high rates of wear are observed
cobalt), and Haynes No. 6 (a nickel–chromium alloy with Nimonic 80A at elevated temperature (750 C,
with no cobalt). They observed that in general the 0.654 m s1, 7 N, 9418 m sliding distance) when
undergoing unidirectional wear against a Stellite 6 systems where the properties of the metallic matrix can
counterface. Even in a like-on-like situation, wear become less robust.
properties of Nimonic 80A are inferior to those of During the wear process, where second phases are
Stellite 6.4,93 harder than the matrix material, it is not sufficient to
assume that their presence will have no effect on the
wear process and that they will simply be ‘worn away’
The presence of carbides in Stellite 6
with the matrix as sliding proceeds; this only occurs if
In both the work of Stott90 and So,86 no mention was
the second phase is of similar or lesser hardness.
made of the effect of carbides that would have formed
Vardavoulias94 studied a number of steels into
with both Stellite 31 and especially Stellite 6, carbon
which hard ceramic phases of various sizes were
being present to 0.5% and 1.1%, respectively. In the
introduced; these included titanium carbide (modi-
cast form, carbon combines with chromium to form a
fied to a much finer carbon nitride phase by nitrogen
chromium carbide phase at the grain boundaries; in
annealing), copper phosphide, and alumina. It is
Stellite alloys, these are of the form M7C3 and
assumed following that the substrate metal is oxi-
M23C6.30 In the wrought and hot isostatically pressed
dized and is the main source of compacted oxide or
forms, these carbides instead form discrete particles
oxide debris and, thus, the oxide–metal interface is
dispersed evenly in the microstructure (the effect of
effectively moving into the metal.
Stellite processing or carbide position on sliding wear
If the sizes of the second phase particles are less
has not, to the knowledge of the authors, been
than the critical oxide thickness (x from Quinn’s
investigated).
oxidational wear theory62), then these particles pass
The presence of these hard, difficult-to-deform
into the oxide layer as the metal is oxidized. They
carbides may have had a number of effects on both
may not protrude above the surface of the oxide layer
sets of experimental work. First, they may have fur-
and thus cannot directly protect the matrix or
ther inhibited deformation of the mainly cobalt
impinge on the counterface material. The particles
matrix during sliding wear, over and above the effect
are lost as the oxide layer breaks up at the critical
expected from the hexagonal close-packed structure,
thickness to form debris. The only contribution may
blocking the operation of the fewer slip planes pres-
be to enhance the load-carrying capacity of the
ent. Second, the removal of material from the Stellite
metallic matrix in supporting the oxide film.
alloys may have released some of these carbides into
Where the second phase particle size is only
the sliding interface, increasing the levels of wear
slightly greater than the critical oxide thickness
observed due to increased abrasion effects. The
(between x and 3x), there is a transition in the wear
enhanced wear of the AISI 4140 and 4340 steels,
mechanism – this is referred to as the ‘first stage’ –
when worn against Stellite 6 clad mild steel in So’s
while a small quantity of the second phase may be
work30 may have in addition been partially attribut-
removed with the oxide, as it breaks up, most will
able to this.
remain embedded in the substrate or matrix. These
There is also the possibility that the carbides (up
particles protrude above the nominal surface of the
to 30 mm in size) within the Stellite 6 may affect the
interface and the counterface slides over them. This
formation of ‘glaze’ (only a few micrometers thick) on
continues until the oxide layer can reform and during
opposing wear surfaces4 as the Stellite 6 is worn down
this stage of wear, the matrix cannot influence the
and the carbides are exposed; Inman,5 however, sug-
wear process; this is the ‘second stage.’
gested that the exposed carbides are not hard enough
The mechanical properties of the second phase
to do this. The possible effects of carbides are discussed
particles influence what happens next during the
in more detail in Sections 1.15.2.5 and 1.15.3.3.2.
‘second stage.’ If the particles are able to resist the
sliding action, then the matrix will be protected for a
prolonged period and the ‘second stage’ is extended.
1.15.2.5 Effect of a Second Phase on Wear
Enhanced wear of the counterface material by abra-
The role of second phases in the wear process is often sion may occur in this stage. If they are unable to
neglected in wear studies. In many studies, experimen- resist the sliding process and fail under the load from
tation has concentrated only on single-phase alloys. the counterface, then contact between matrix and
However, second phases are used in many alloys for counterface is quickly restored and the ‘first stage’
various reasons, including enhancement of strength of wear will be repeated. A further possibility is
and creep resistance, especially in high temperature detachment of second phase particles as the oxide
breaks up, with these particles acting as third body Ni-base-Inconel alloys MA 6000, MA754,
abrasives; this occurs where cohesion between matrix MA758, and MA760, and Fe-base Incoloy MA956,
and second phase is poor. PM2000, and PM2000SD are well-known examples
If the mean particle size is much greater than the of ODS superalloys. Such ODS alloys possess good
critical oxide thickness, the particles show increased high temperature strength and resistance to HT cor-
efficiency in providing oxidational wear protection to rosion and oxidation. They derive their strength from
the material subject to wear. After break up of the the insoluble, deformation resistant and thermally
oxide layer (end of the ‘first stage’), the harder second stable dispersoids such as Y2O3 (as used in Incoloy
phase particles remain embedded in the matrix. MA956)96 introduced during fabrication by mechan-
Again, the main interaction is between the particles ical alloying. This process offers great opportunities
and the counterface, and this process controls the in the selection of dispersoids–matrix combination.
wear mechanism; the matrix plays no direct part. Strengthening by insoluble, deformation resistant
The majority of the particles are surrounded by the and inert dispersoids is far more effective than other
matrix; thus, break up is more difficult and detach- methods such as solid solution and precipitation
ment is almost impossible. While this means that hardening, as thermodynamics imposes limitation
the matrix is well protected against wear, the counter- on their continued effectiveness. At certain critical
face may undergo high rates of wear and thus become temperatures, solubility limits may be exceeded,
the main source of debris. The inference here is that leading to the onset of precipitation and eventually
the first stage cannot resume until these larger parti- Oswald ripening of the precipitates.
cles wear to near the level of the rest of the sample During high temperature deformation, most ODS
surface; as other particles will continue to be exposed alloys exhibit a threshold stress s0 below which creep
elsewhere on the surface, first stage wear with pro- becomes negligible; the threshold stress s0 is less than
tective oxide layer formation cannot readily happen Orwan stress sor . Several models, such as dislocation
and severe wear will continue. climb and dislocation detachment (local climb), have
On the basis of the apparent failure of carbides been advanced to explain the existence of the thresh-
in Stellite 6 to disrupt ‘glaze’ formation for cer- old stress. None of these models have been univer-
tain wear combinations, Inman5 suggested that the sally accepted.
second phase particles must also be harder than the To increase the temperature capability of
generated ‘glaze’ on the opposing sliding surface ODS alloys, the powder produced by mechanical
(Section 1.15.3.3.2). If this is not so, the second alloying is subjected to hot extrusion and the fine
phase particles may wear in preference to the ‘glaze’ grained extruded structure is then subjected to a
layer and not promote its break-up. This may prove high temperature secondary recrystallization anneal.
beneficial, with a harder second phase conferring The overall enhancement of HT capability may be
wear resistance but allowing protective ‘glazes’ to associated with:
form. Tribaloy alloys, consisting of a hard Laves
large grain size produced by secondary
intermetallic second phase in a Ni-based or more
recrystallization;
notably a Co-based matrix,95 are possible examples
high GAR (grain aspect ratio) produced by recrys-
(not discussed in the current work).
tallization under a high temperature gradient and
minimizing the detrimental effect of transverse
grain boundary effect;
1.15.3 High Temperature Wear formation of serrated grain boundaries allowing
Behavior of Advanced Materials
grain interlocking; and
1.15.3.1 Oxide Dispersion Strengthened minimum grain boundary hardening.
(ODS) Alloys
1.15.3.1.1 Introduction 1.15.3.1.2 Observations on high temperature
This section discusses the HT wear behavior of fer- wear behavior of some ODS alloys
ritic ODS alloys, Incoloy MA956, and its variants. There is no evidence to suggest any improved wear
The section starts with a brief introduction to ODS resistance conferred by alloy pretreatment for the
alloys to provide a context and facilitate interpreta- three near-identical ODS alloys, Incoloy MA956,
tion of the experimental observations on their HT PM2000, and PM2000SD3 (Figure 10). Testing
wear characteristics. against various counterfaces (Stellite 6 and Si3N4)
–3.2 St6 wheel at 750 ⬚C

PM2000SD
–14.4 St6 wheel at RT
–280.3 In800 wheel at 750 ⬚C
–0.7 In800 wheel at RT
–1.4 St6 wheel at 750 ⬚C

PM2000 –22.8 St6 wheel at RT
–594.7 In800 at 750 ⬚C
–1.4 Si3N4 wheel at 750 ⬚C

–95.3 Si3N4 wheel at RT
MA956 –7.4 St6 wheel at 750 ⬚C
–10.2 St6 wheel at RT
–1020.7 In800 wheel at 750 ⬚C
–1200 –1000 –800 –600 –400 –200 0

Weight loss (mg cm–2)
Figure 10 Weight change of the ODS alloys worn against various counterfaces at room temperature and 750 C3 (sliding
speed 0.654 m s1, load 7 N).
and at various temperatures (room temperature to Table 2 Description and properties of the ODS alloys
750 C) does not indicate a definitive trend in rela- investigated
tion to grain size and hardness. Alloy Description Hv Av. grain
500g size (mm)
Ma956 ODS alloy 303 3000

Incoloy 800HT counterface
strengthened
Only on sliding against Incoloy 800HT there is an through a yttria
apparently strong preference towards lower grain size dispersion in
and increased hardness (Table 2) favoring higher Fe–Cr–Al matrix
wear.3 The severity of wear was of the order PM2000 ODS alloy 311 500
strengthened
PM2000SD, PM2000, and Incoloy MA956.
through a yttria
Poor wear resistance and high friction coefficient dispersion in
characterize the wear behavior of the ODS alloys Fe–Cr–Al matrix
worn against Incoloy 800HT at 750 C (reciprocating PM2000SD ODS alloy 363 80
rig, 0.314, 0.654, and 0.905 m s1 sliding speed, strengthened
through a yttria
7 N load) as indicated by higher weight losses com-
dispersion in
pared to that at room temperature (Figure 10 – Fe–Cr–Al matrix.
0.654 m s1).3 The apparent improved wear resistance Rolled at higher
observed at room temperature is attributable to trans- temperature than
fer3–5 and work hardening of a layer of Incoloy PM2000 imparting
greater ductility
800HT or back-transferred Incoloy MA956,5 protect-
ing the ODS alloy surface from sustained wear Reproduced from Wood, P. D. Ph.D. Thesis, Northumbria
(Figure 11(a) – 0.654 m s1). University, UK, 1997.
Severe wear with high metallic

transfer from Incoloy 800HT
counterface to Incoloy MA956
sample
Torn, metallic surface typical of

adhesive wear, no traces of
oxide
(a) ⫻100 0000 20 kV 500 ␮m
Severe wear with reduced

transfer from Incoloy 800 HT
counterface to Incoloy MA956
sample
Torn, metallic surface typical of

adhesive wear, no clear traces
of oxide
(b) ⫻300 0000 20 kV 100 ␮m
Smooth ‘glaze’ layer formation

with little debris following period
of early severe wear with
damage to underlying sample
(c) ⫻300 0000 20 kV 100 ␮m
Figure 11 SEM micrographs of Incoloy MA956 wear surfaces after sliding at 0.654 m s1 against an Incoloy 800HT
counterface at room temperature, 510 and 750 C (sliding distance 9418 m).4 (a) Room temperature (shown), also typical of
270 C, (b) 510 C (also typical of 390, 450, 570, and 630 C), (c) 750 C (shown), also typical of 690 C.
The loss of strength and hardness suffered by the metallic transfer layer seen at room temperature5
ODS alloys at higher temperatures10–12 (Figure 12) and wear thus increases with temperature. Some
probably undermines the ability of the alloy to sup- transfer of Incoloy 800HT material still occurs at
port the process of ‘glaze’ development. Severe wear high temperature; however, continued sliding
continues (Figure 11(b) – 0.654 m s1) and the low removes this layer.3,5 Such transfer tends to be
levels of oxide produced act only to inhibit metallic greater at higher sliding speeds, and hence, the
adhesion, preventing the formation of the protective weight loss is lower at 0.905 m s1 than at
12
Incoloy MA956
10 Nimonic 80A
Stellite 6
Knoop hardness (GPa)

Incoloy 800HT
8
6 1) 2) 3)
0
0 100 200 300 400 500 600
Temperature (⬚C)
Mean Knoop hardness (GPa)
Temperature (°C) Stellite 6 Nimonic 80A MA956 Incoloy 800
30 9.39 3.33 1.99 3.38
150 6.28 2.75 1.71 2.16
270 4.37 2.58 1.45 1.68
390 3.61 2.50 1.24 1.52
450 3.55 2.37 1.16 1.34
510 3.08 2.01 0.97 0.55
Figure 12 Mean Knoop hardness (hot hardness, 50 g load, 12 s dwell time) from room temperature to 510 C,4 with wear
regimes with respect to temperature identified for the Nimonic 80A/Stellite 6 and Incoloy MA956/Stellite 6 systems.4,5
1. Low-temperature oxidational wear, moderate falls (large for Stellite 6) in hardness. 2. Low-temperature oxidational wear at
0.314 m s1, metallic severe wear only at 0.905 m s1, little change in hardness. 3. ‘Glaze’ formation for Nimonic 80A vs.
Stellite 6 at 0.314 m s1 with high oxidational wear replacing metallic severe wear at 0.905 m s1. Also, early ‘glaze’ formation
for Incoloy MA956 versus Stellite 6, with extended early severe wear at 0.905 m s1.
0.314 m s15 Only at 690 and 750 C (Figure 11(c) – (at least for Incoloy MA956 versus Stellite 6) have
0.654 m s1 example shown) is an oxide3 or ‘glaze’4,5 been observed to depend on sliding speed.5,23 A slow
layer able to develop on the Incoloy MA956 surface sliding speed (0.314 m s1)5 favors greater Stellite 6
(sourced from the Incoloy MA956) after an initial wear and thus higher levels of Co-based oxide
period of severe wear.5 debris. Faster sliding speeds (0.654 m s1 4and espe-
cially, 0.905 m s1)5 encourage greater Incoloy
Stellite 6 counterface MA956 wear and thus higher levels of Fe and Cr
The improved wear resistance of ODS alloys when oxide contribution (Figure 15). At 750 C, the
worn against Stellite 6 coincides with rapid oxide presence of Co in the debris promotes more rapid
development; such oxide is sourced from both sam- ‘glaze’ formation23 and prevents continued early
ple and counterface. At low temperature, this debris wear.
remains loose and does not readily form ‘glaze’ Only at intermediate temperatures (390 and
(Figures 13(a) and 14(a)). At high temperature, it 450 C)4,5 and at 0.654 and 0.905 m s1 does the
sinters rapidly to form more extensive ‘glaze’ layers oxide fail to separate wear surfaces and allow severe
(Figures 13(c) and 14(c)).3–5 The relative contribu- wear to occur (Figure 14(b)). Below these tempera-
tions of sample- and counterface-sourced debris tures, the oxide separates the surfaces even in the
Loose oxide separating wear surfaces,

plus some compacted oxide layers –
isolated patches of ‘glaze’ at 450 ⬚C only
(not shown)
(a) ⫻300 0000 20 kV 100 ␮m
Areas of smooth, protective ‘glaze’ layer

formation
Some accompanying loose debris.
(b) ⫻300 0000 20 kV 100 ␮m
Smooth, protective ‘glaze’ layer

formation with very little loose debris
(greater debris incorporation)
(c) ⫻300 0000 20 kV 100 ␮m
Figure 13 SEM micrographs of Incoloy MA956 wear surfaces after sliding at 0.314 m s1 against a Stellite 6 counterface
at room temperature, 510, and 750 C.23 (a) Room temperature (shown), also typical of 270, 390, and 450 C, (b) 510 C,
and (c) 750 C (shown), also typical of 570, 630, and 690 C.
form of the aforementioned loose debris. At progres- 1.15.3.1.3 Effects of load

sively higher temperatures, severe wear is increas- Rose4 also examined the effect of load (7–25 N,
ingly restricted to only the initial sliding period as 0.654 m s1 only) on the sliding behavior of Incoloy
continued sliding promotes debris sintering and MA956 when slid against Incoloy 800HT and Stellite
‘glaze’ formation. Above 630 C, ‘glaze’ development 6 counterfaces at 750 C (Figure 16). ‘Glaze’ layers
is so rapid that severe wear is all but eliminated. No were observed for all loads up to 20 N for both counter-
severe wear is observed at 0.314 m s1, with the loose faces and no change in oxide behavior was observed
debris progressively developing into ‘glaze’ as sliding with increasing load (although greater material working
temperature is increased5 (Figure 13 shows the was reported as being necessary to provide material for
debris (a) loose at room temperature, (b) partially ‘glaze’ when an Incoloy 800HT counterface was used).
forming a ‘glaze’ layer at 510 C, and (c) forming a No significant increases in weight loss were observed
comprehensive ‘glaze’ layer at 750 C). with increasing load up to 20 N (Figure 17). These
Loose oxide debris spread across

wear surface (more smeared
than at 0.314 m s–1), plus some
compacted oxide formation and
odd fragment of metallic debris
(not shown)
(a) ⫻300 0000 20 kV 100 ␮m
450 ⬚C 510 ⬚C
450 ⬚C-torn, metallic surface

typical of adhesive wear, no
traces of any oxide debris
Limited
glaze
Torn 510 ⬚C and 570 ⬚C-very limited
metallic ‘glaze’ formation, more
surface comprehensive at 570 ⬚C
(b) ⫻600 50 ␮m ⫻300 100 ␮m
Smooth ‘glaze’ layer formation

with little debris (greater debris
incorporation)
(c) ⫻300 0000 20 kV 100 ␮m
Figure 14 SEM micrographs of Incoloy MA956 wear surfaces after sliding at 0.654 m s1 against a Stellite 6 counterface
at room temperature, 450, 510, and 750 C4 (also representative of 0.905 m s1).23 (a) Room temperature (shown), also
typical of 270 and 390 C, (b) 450 C (left) and 510 C (right – also typical of 570 C), (c) 750 C (shown), also typical of 630
and 690 C.
observations were irrespective of whether the ‘glaze’ 1.15.3.2 Intermetallics

was primarily Stellite 6-counterface sourced or Incoloy
1.15.3.2.1 Introduction
MA956-sample sourced (when the counterface was
Strong, predominantly metallic bonding between
Incoloy 800HT). Higher losses were reported at
unlike atoms leads to the formation of intermetallics
25 N (most noticeably with an Incoloy 800HT counter-
and intermetallic compound phases. From such
face – Figure 17) and, although ‘glaze’ layers still
bonding comes crystal structure, ordering, high
formed, the Incoloy MA956 sample substrate was
strength at low and high temperature, low ductility,
unable to provide sufficient support for the ‘glaze’
and low K1c, particularly at low temperature.
layer to remain protective.
Damaged Incoloy MA956 surface

due to transfer of surface material
to the Si3N4 counterface
Loose oxide debris generated

later due to Incoloy MA956
sample sliding against Incoloy
MA956 transfer layer on Si3N4
counterface
(a) ⫻100 0000 20 kV 500 ␮m
Thin ‘glaze’ layer source from

both Incoloy MA956 sample and
Si3N4 counterface
(b) ⫻150 0000 20 kV 200 ␮m
Figure 15 SEM micrographs of Incoloy MA956 wear surfaces after sliding at 0.654 m s1 against a Si3N4 counterface at (a)
room temperature and (b) 750 C (load 7 N, sliding distance 9418 m). Reproduced from Wood, P. D. Ph.D. Thesis,
Northumbria University, UK, 1997.
The low ductility, low K1c, and high strength of 1.15.3.2.2 Wear of TiAl – Metallic counterfaces
intermetallics stem from such critical factors as com- In discussing intermetallic wear resistance, attention
plex crystal structures, the large Burgers Vector, high is focused on TiAl and TiAl-based intermetallics
lattice stress, the inadequate slip systems and the because of increasing interest in using these materials
inability to cross slip. The complex interplay between in many industrial applications, including automo-
these parameters makes the prediction of intermetal- tive, aerospace, and power generation.
lic wear behavior difficult. TiAl when worn against Incoloy 800HT at room
The expected improvement in wear resistance temperature (‘reciprocating-block-on-rotating-cylinder’
from high strength, ordering and an adequate slip rig, load 7 N, sliding speed 0.654 m s1) shows a similar
system may be off-set by low K1c and low fracture wear pattern to the ODS and Nimonic alloys, and
strain. The advantage in providing high wear resis- involves a similar mechanism3; the transfer of Incoloy
tance due to ordering may eventually be lost by the 800 to the TiAl surface, followed by work hardening of
destruction of this ordering due to elemental diffu- the transferred layer, and formation of a wear-resistant
sion from the counterface into the intermetallic lat- oxide layer (NiCr2O4/Fe2O3). The development of this
tice. At elevated temperature, increased K1c and hardened, wear-resistant layer causes some improve-
fracture strain are likely to improve the wear resis- ment in wear resistance without any observable ‘glaze’
tance; however, some of the effects of these para- formation (Figure 18(a)).
meters may be masked by the formation of wear In contrast, ‘glaze’ formation has been observed in
resistance surface ‘glaze’ layers. the same system at 750 C (7 N, 0.654 m s1), the ‘glaze’
Incoloy 800HT counterface Stellite 6 counterface
Smooth
compacted
oxide / ‘glaze’ Smooth
layers with a little compacted
loose debris oxide / ‘glaze’
(underlying layers
sample damaged
due to early
severe wear)
(a) ⫻300 0000 20 kV 100 ␮m (b) ⫻300 0000 20 kV 100 ␮m
Smooth
High Temperature Tribocorrosion

compacted
oxide / ‘glaze’ Smooth
layers and a little compacted
loose debris, with oxide / ‘glaze’
some cracking layers, with some
due to high load cracking due to
(underlying high load
sample damaged
due to early
severe wear)
(c) ⫻300 0000 20 kV 100 ␮m (d) ⫻300 0000 20 kV 100 ␮m
Figure 16 SEM micrographs of Incoloy MA956 worn against an Incoloy 800HT counterface at applied loads of (a) 7 N and (b) 25 N, and against a Stellite 6 counterface at
applied loads of (c) 7 N and (d) 25 N (750 C, sliding speed 0.654 m s1, sliding distance 9418 m). Reproduced from Rose, S. R. Ph.D. Thesis, Northumbria University, UK, 2000.
361
0.35
0.3
Incoloy MA956 vs. Stellite 6
0.25
Incoloy MA956 vs. Incoloy 800HT
Weight loss (g)
0.2 Nimonic 80A vs. Stellite 6
Nimonic 80A vs. Incoloy 800HT

0.15
0.1
0.05
0
0 5 10 15 20 25 30
Load (N)
Figure 17 Weight losses for Incoloy MA956 and Nimonic 80A worn against Incoloy 800HT and Stellite 6 counterface at
applied loads of between 7 and 25 N (750 C, sliding speed 0.654 m s1, sliding distance 9418 m). Reproduced from Rose,
S. R. Ph.D. Thesis, Northumbria University, UK, 2000.
(Figure 18(b)) containing mainly oxidized Fe/Ni/Cr While the debris generated (sourced from both sample
material from Incoloy 800 with little Ti and Al.3 and counterface) forms loose oxide platforms (or ‘pla-
These observations are in conflict with other work teaux’ – Figure 20(a)), these are formed and removed
where the ‘glaze’ contained the elements of TiAl. It is as sliding continues and do not form a wear protective
believed that the use of a different test system (‘pin-on- layer. The high TiAl wear has been attributed to the
disk’ rather than ‘block-on-cylinder’) in the latter work elements of the counterface material interfering with
accounts for this difference. The role of TiAl is con- the ordered structure of the intermetallic, as small
sidered to provide a deformation and wear-resistant atomic radii Si and N enter the lattice and increase
substrate for the ‘glaze’ to develop, to reside, and to its susceptibility to wear. A mixture of abrasive and
be sustained. adhesive wear with some evidence of stick-slip, possi-
When worn against Stellite 6 (7 N, 0.654 m s1)3, bly due to high mutual chemical compatibility
initial transfer of Stellite 6-based material on the TiAl between the Si3N4 and the TiAl, has been reported98
surface results initially in a like-on-like sliding regime, (Si3N4 has been previously observed to have high
with the Stellite 6 counterface sliding against this mutual chemical compatibility with both Ti and
transferred layer. At room temperature (Figure 18(c)), Al99), leading to moderate wear of the TiAl and high
instability of this layer leads to a continual process of wear of the Si3N4. In contrast, TiAl undergoes lower
transfer and removal from both surfaces and a loose wear when slid against silicon nitride at 750 C (at
oxide containing Ti and Al from the TiAl, and Co and 0.31497 and 0.654 m s1),3 due to the rapid formation
Cr from the Stellite 6 is generated. Only at 750 C (at of a thin wear-resistant oxide layer (Figures 19(b) and
0.314 m s1,97 as well as 0.654 m s1)3 does the trans- 20(b)). The formation of this layer, sourced from both
ferred Stellite 6 material form a primarily Co–Cr-oxide the TiAl and Si3N4, prevents substantial wear. It is
‘glaze’ layer on the TiAl surface (Figures 18(d) and suggested that any high mutual compatibility
19(a)), this layer being thinner than that created when enhances the very early stages of wear, providing the
sliding against Incoloy 800HT. necessary material for the wear-resistant layer.
TiAl undergoes higher wear when worn against
1.15.3.2.3 Wear of TiAl – Ceramic counterfaces Al2O3 at room temperature97 with enhanced material
TiAl suffers moderate wear when slid against a silicon removal by abrasion, the debris from which does not
nitride interface at room temperature (0.654 m s1). form a protective layer. The material generated from
Mixed oxide
Work hardened debris layer of
Incoloy 800HT Stellite 6-sourced
transfer layer on Co/Cr and TiAl-
TiAl surface sourced Ti and
Al.
(a) 1.0 k 0000 20 kV 50 ␮m (c) ⫻2.0 k 0000 20 kV 25 ␮m
High Temperature Tribocorrosion

Oxidized layer
consisting of
Incoloy 800HT- Primarily Co–Cr
sourced Fe, Ni oxide layer on
and Cr, with a TiAl surface
little TiAl-sourced
Ti and Al
(b) ⫻2.0 k 0000 20 kV 25 ␮m (d) ⫻2.0 k 0000 20 kV 25 ␮m
Figure 18 SEM micrographs of TiAl worn against an Incoloy 800HT counterface at (a) room temperature and (b) 750 C, and against a Stellite 6 counterface at (c) room
temperature and (d) 750 C, sliding speed 0.654 m s1 (load 7 N, sliding distance 9418 m). Reproduced from Wood, P. D. Ph.D. Thesis, Northumbria University, UK, 1997.
363
‘Glaze’ layer up to 3 µm thickness

with limited fragmentation and
break-away, most evident near
top and top right
Also some loose debris evident
(lighter areas to left and right)
(a)
Mag HFW WD Spot HV 300.0 µm
500 ⫻ 0.54 mm10.0 mm 5.0 20.0 kV AMRI
‘Glaze’ layer up to 3 µm thickness

overlaid by a few fragements of
broken-away ‘glaze’ of between 1
and 15 µm
Also some loose debris and very
limited break-away (streaks in
bottom right hand corner)
(b)
500 ⫻ 0.54 mm10.0 mm 5.0 20.0 kV AMRI
Smooth ‘glaze’ layer up to 3 µm

overlaid by fragments of
broken-awy ‘glaze’ of between
1 and 20 µm in size
Little evidence of loose debris
(c)
500 ⫻ 0.54 mm10.0 mm 5.0 20.0 kV AMRI
Figure 19 SEM micrographs of g-TiAl wear scar surfaces after wear against Stellite 6, Si3N4 and Al2O3 counterfaces at
750 C, sliding speed 0.314 m s1 (load 7 N, sliding distance 4522 m).97 (a) Stellite 6 counterface, (b) Si3N4 counterface, and
(c) Al2O3 counterface.
an initial period of high TiAl-wear against Al2O3 at 1.15.3.3 Nimonic Alloys

750 C,97 however, readily forms a ‘glaze’ layer
1.15.3.3.1 Incoloy 800HT counterface
(Figure 19(c)) which provides some protection for
Wood3 reported a correlation between Nimonic
the TiAl surface and prevents excessive wear. How-
80A cast, Nimonic 80A HIPped, and Nimonic 90
ever, despite the ‘glaze’ layer, the abrasive nature of
when tested against different counterfaces at various
the oxide generated (a mixture of Ti and Al oxides)
temperatures (0.654 m s1, 7 N load and reciprocating-
still promotes continued material removal. In con-
block/sample-on-rotating cylinder/counterface con-
trast, the Al2O3 undergoes little wear at both room
figuration). The results of room temperature tests
temperature99 and 750 C,97 with it having no poten-
seemed to be greatly influenced by the degree of
tial chemical compatibility with TiAl; hence, no
material transfer to the Nimonic surface, this transfer
adhesive mechanism can occur.
Nonprotective loose oxide layer

overlaying damaged TiAl surface
(caused by Ti and Al mutual
chemical compatibility with
Si3N4? [15])
(a) ⫻200 0000 20 kV 250 ␮m
TiAl- and Si3N4-sourced thin and

wear protective oxide layer
(b) ⫻200 0000 20 kV 250 ␮m
Figure 20 SEM micrographs of TiAl worn against a Si3N4 counterface at (a) room temperature and (b) 750 C, sliding speed
0.654 m s1 (load 7 N, sliding distance 9418 m). Reproduced from Wood, P. D. Ph.D. Thesis, Northumbria University, UK,
1997.
itself being influenced by counterface hardness; the Rose (0.654 m s1)4 and Inman (0.314 and
softer the counterface the higher the amount of 0.905 m s1)5 examined temperatures between room
transfer. temperature and 750 C (Figures 21 and 22), with
Metallic transfer readily occurred from the rela- Rose4 noting that Incoloy 800HT-sourced transfer
tively soft Incoloy 800HT counterface onto the sur- layers continued to be observed up to 570 C.
faces of each of the Nimonic materials at room Inman5 reported the development of transfer layers
temperature. The transferred material formed a up to 750 C, observing an increase in transfer with
wear-resistant layer which protected the Nimonic sliding speed. At 0.314 m s1, transfer was less, due
material surfaces and led to very small weight to increased amounts of oxide that interfered with
changes. Inman5 later reported the development metallic adhesion, but did not form ‘glaze.’ At
of metallic transfer layers at 0.314 and 0.905 m s1 0.905 m s1, transfer was greater than at either 0.314
(Figures 21 and 22 – greater transfer was observed at or 0.654 m s1; any surface oxidation was more readily
0.905 m s1) when Nimonic 80A was slid against removed, allowing greater adhesion of transferred
Incoloy 800HT; these layers were work hardened. material.
At 750 C, the Nimonic alloys showed a large ‘Glaze’ layers were observed only between 630
range of behaviors when tested against the same and 750 C,4,5 with its formation being favored by
three counterfaces.3 When slid against Incoloy increased test temperature4,5 and lower sliding
800HT, mainly Incoloy 800HT-sourced layers that speed (due to decreased material removal).5 These
protected the Nimonic alloys against wear, together ‘glaze’ layers were reported to form after transfer
with low weight changes, were observed. Rose4 and from the Incoloy 800HT counterface to the Nimonic
Inman5 additionally reported ‘glaze’ overlying any 80A surface5; effectively, the ‘glaze’ was generated
transferred material when Nimonic 80A was worn by ‘like-on-like’ sliding of the transfer layer against
against Incoloy 800HT at this temperature. the Incoloy 800HT counterface.
Metallic transfer layer with moderate

levels of metallic debris generation at
room temperature (RT) and 270 ⬚C
(a) ⫻300 0000 20 kV 100 mm
Limited patchy metallic transfer at

510 ⬚C, surface mostly exposed sample
material
(b) ⫻300 0000 20 kV 100 mm
‘Glaze’ at 750 ⬚C, overlying slightly more

developed metallic transfer layer than
seen at 510 ⬚C
(c) ⫻300 0000 20 kV 100 mm
Figure 21 SEM micrographs of Nimonic 80A wear surfaces after sliding at 0.314 m s1 (load 7 N, sliding distance
4522 m) against an Incoloy 800HT counterface at room temperature, 270, 570, and 750 C.5 (a) Room temperature
(shown), also typical of 270 C, (b) 510 C (shown), also typical of 390, 450, 570, and 630 C, and (c) 750 C (shown), also
typical of 690 C.
Effects of load changed from protective to abrasive due to the increased

Rose4 also examined the effect of load (7–25 N, load. This was reflected by a rapid increase in weight
0.654 m s1 only) on sliding behavior for Nimonic 80A loss above 15 N (Figure 17).
when slid against Incoloy 800HT at 750 C. Limited
compacted oxide layers formed at 7 N (Figure 23(a)) 1.15.3.3.2 Stellite 6 counterface
and 10 N load, protecting the Nimonic 80A surface. At Wood (0.654 m s1, 7 N load)3 reported only limited
loads between 15 and 25 N (Figure 23(b)), an abrasive, material transfer from the harder Stellite 6 counterface
oxidational wear regime dominated, with no debris to the Nimonic material surfaces at room temperature.
build-up or layer formation; the behavior of the oxide Thin mixed oxide layers were formed on the Nimonic
Metallic transfer layer with moderate

levels of metallic debris generation at
room temperature (RT) and 270 ⬚C,
greater transfer of metal onto sample
surface than at 0.314 m s–1 (Fig. A1)
(a) ⫻300 0000 20 kV 100 mm
Comprehensive metallic transfer

continuing at 510 ⬚C (greater than at
0.314 m s–1 – Fig. A1) – metallic debris
generation from both sample and
counterface
(b) ⫻300 0000 20 kV 100 mm
‘Glaze’ at 750 ⬚C, overlying more limited

metallic transfer layer
(c) ⫻300 0000 20 kV 100 mm
Figure 22 SEM micrographs of Nimonic 80A wear surfaces after sliding at 0.905 m s1 (load 7 N, sliding distance
4522 m) against an Incoloy 800HT counterface at room temperature, 270, 570, and 750 C.5 (a) Room temperature and
270 C (shown), also typical of 390 C, (b) 570 C (shown), also typical of 450, 510, and 630 C, and (c) 750 C (shown), also
typical of 690 C.
surfaces; these contained material from both the Nimo- separated the wear surfaces, preventing metallic con-
nic and Stellite 6 wear surfaces and provided only tact; a low temperature mild wear regime resulted.
limited protection. Testing of Nimonic 80A against a Behavior at intermediate and high temperature
Stellite 6 counterface at 0.654 m s1(4) and at 0.314 and depended significantly on sliding speed. At
0.905 m s1(5,21,22) (Figures 24(a) and 25(a)) resulted 0.314 m s1(5,21,22) the loose debris mild wear regime
in preferential wear of the Stellite 6 counterface, (Figure 24(a)) continued up to 450 C. Increased
generating loose Co–Cr-based oxides. This debris sintering was observed at 390 C, isolated ‘glaze’ at
368
Types of High Temperature Corrosion
Limited compacted
oxide layers Grooved, worn
overlying Nimonic surface due to
80 A surface abrasive action
(a) ⫻300 0000 20 kV 100 ␮m (c) ⫻300 0000 20 kV 100 ␮m
Grooved, worn
surface due to Damage (grooves)
abrasive action more pronounced at
higher load
(b) ⫻300 0000 20 kV 100 ␮m (d) ⫻300 0000 20 kV 100 ␮m
Figure 23 SEM micrographs of Nimonic 80A worn against an Incoloy 800HT counterface at applied loads of (a) 7 N and (b) 25 N, also against a Stellite 6 counterface at
applied loads of (c) 7 N and (d) 25 N (750 C, sliding speed 0.654 m s1, sliding distance 9418 m). Reproduced from Rose, S. R. Ph.D. Thesis, Northumbria University, UK, 2000.
Loose oxide separating wear surfaces,

plus some compacted oxide layers –
isolated patches of ‘glaze’ at 450 ⬚C only
(not shown)
(a) ⫻300 0000 20 kV 100 µm
Areas of smooth, protective ‘glaze’ layer

formation
Some accompanying loose debris.
(b) ⫻300 0000 20 kV 100 µm
Smooth, protective ‘glaze’ layer

formation with very little loose debris
(greater debris incorporation)
(c) ⫻300 0000 20 kV 100 µm
Figure 24 SEM micrographs of Nimonic 80A wear surfaces after sliding at 0.314 m s1 (load 7 N, sliding distance 4522 m)
against a Stellite 6 counterface at room temperature, 510 and 750 C.22 (a) Room temperature (shown), also typical of 270,
390, and 450 C and (b) 510 C, (c) 750 C (shown), also typical of 570, 630, and 690 C.
450 C, with increasingly comprehensive ‘glaze’ for- metallic debris sourced from the Nimonic 80A was
mation above 510 C (Figure 24(b)) and most notably generated by delamination wear. At 630 C, however,
between 630 and 750 C (Figure 24(c)).5,21,22,24 oxide was observed and this assisted severe wear by
The loose debris mild wear regime was observed abrasion.5,21,22,24 A similar pattern was observed at
up to 390 C at 0.654 m s1(4) and 270 C at 0.905 m s1(5,21,22) though enhanced frictional heating
0.905 m s1(5,21,22); at higher temperatures, there was led to oxide generation at 570 C as well as 630 C.
a transition to intermediate severe wear. Such severe On increasing temperature to 690 and 750 C, fur-
wear was observed at 0.654 m s1 up to 570 C,4 ther increases in oxide production were sufficient to
without any evidence of oxide (Figure 25(b)); only separate completely the Nimonic 80A sample and
R.T. 270 ⬚C
Loose oxide separating wear surfaces
(a) ⫻300 0000 20 kV 100 µm
390 (0.905 m s−1 only), 450, 510, and

570 ⬚C–torn, metallic surface typical of
adhesive wear, no traces of any oxide
debris
630 ⬚C–adhesive wear persists,

however, some oxide present and
assisting wear by abrasion (not shown)
(b) ⫻300 0000 20 kV 100 µm
Parallel grooves due to abrasion (holding

some loose oxide debris)
(c) ⫻300 0000 20 kV 100 µm
Figure 25 SEM micrographs of Nimonic 80A wear surfaces after sliding at 0.654 m s1 (load 7 N, sliding distance 4522 m)
against a Stellite 6 counterface at room temperature, 270, 570, and 750 C (also typical of 0.905 m s1).22 (a) Room
temperature and 270 C (shown), also typical of 390 C (0.654 m s1 only), (b) 570 C (shown), also typical of 390 (0.905 m s1
only) 450, 510, and 630 C, and (c) 750 C (shown), also typical of 690 C.
Stellite 6 counterface; a mild wear regime was thus adherent ‘glaze’ that was continuously removed. In
established. However, high wear rates continued; contrast, Rose4 reported no ‘glaze’ formation on Nimo-
although technically a mild wear regime, the high nic 80A and the generation of only loose oxides that
wear rates indicates that the oxide was not protective. acted abrasively (NiO and Cr2O3), thereby enhancing
Two slightly different forms of oxidational behavior removal from the Nimonic 80A surface. These latter
were observed, depending on researcher. At observations were confirmed by Inman5,22,24 at both
0.654 m s1, Wood3 reported the formation of a poorly 0.654 m s1 (Figure 25(c)) and 0.905 m s1, who
observed greater oxide generation at 0.905 m s1 but 1.15.3.3.4 Nimonic 80A sliding wear –
still without ‘glaze’ layer formation. Comparisons between various wear rig
Rose4 stated that layers were unable to form due to configurations
insufficient debris adhesion to the Nimonic 80A sam- The apparently poor sintering and ‘glaze’ forming
ple surfaces and lack of debris cohesion, caused by the characteristics of NiO and Cr2O3 generated from
ploughing of sample surfaces by hard carbide particles Nimonic 80A (at 0.654 and 0.905 m s1)4,5,21,22,24 at
in the Stellite 6 counterface. However, Inman5 argued first seems to contradict the extensive studies carried
that the carbides were not hard enough to have this out into ‘glaze’ formation with Nimonic 80A-based
ploughing effect and inhibit the development of oxide systems by Jiang et al.2,16–19 However, the last group
layers; it was possible that the carbides were softer used lower sliding speeds with a like-on-like recipro-
than the ‘glaze’ layers and wore in preference to them. cating ‘pin-on-disk’ configuration (mean sliding speed
This inability of the Nimonic 80A-sourced NiO and 83 mm s1); such systems have a higher degree of
Cr2O3 oxides to form a ‘glaze’ was instead attributed debris retention, providing greater opportunity for
to their relatively poor sinterability. If this failure to the oxide debris to sinter together to form ‘glaze’ layers.
form ‘glaze’ layers had been due to carbide ploughing, The ‘block-on-cylinder’ configuration used by Rose4
then these layers would not have formed in other and Inman et al.5,21,22,24 is a unidirectional sliding wear
systems involving Stellite 6. For example, Stellite 6- system that promotes debris mobility and ejection over
sourced Co–Cr layers formed in the Nimonic 80A/ retention, especially at higher sliding speeds. The
Stellite 6 (counterface) system at 0.314 m s1.5,22,24 decreased residency and greater mobility of the debris
Also, Incoloy MA956-sourced Fe–Cr-based layers do not allow sufficient contact time between debris
formed at 0.905 m s1 in the Incoloy MA956/Stellite particles for sintering and welding processes to occur,
6 (counterface) system.5,23 as is necessary for ‘glaze’ formation.
Inman et al.22 collated the available data to create a
temperature versus sliding speed wear map for the
1.15.3.3.5 Si3N4 counterface
Nimonic 80A versus Stellite 6 (counterface) system
Wood (0.654 m s1, 7 N load)3 reported limited mate-
(later further developed by Inman and Datta24), cov-
rial transfer from the harder Si3N4 counterface to the
ering a variety of different modes of wear behavior
Nimonic material surfaces at room temperature.
that are dependant on load–sliding speed combina-
Thin mixed oxide layers were formed on the Nimo-
tion. This is discussed further in Section 1.15.5.3.
nic surfaces; these contained material from both wear
Effects of load surfaces that provided only limited protection.
Rose also studied the effect of load on sliding behavior Wood3 observed a very different wear resistance
for Nimonic 80A worn against Stellite 6 at 750 C.4 At when Nimonic 80A (cast) and Nimonic 90 were worn
all loads (7–25 N), the wear mechanism remained pre- against Si3N4 at 750 C. Nimonic 80A (cast) showed
dominantly oxidational, with elements of abrasive wear, very poor resistance, with no ‘glaze’ formation in
regardless of applied load (Figures 23(c) and 23(d)); contrast to Nimonic 90 where ‘glaze’ formation and
no compacted debris layers were formed across the very low weight losses were observed. Wood attrib-
range of loads (7–25 N) at 750 C. There was no evi- uted the difference to the presence of Co in the
dence that changing load had any significant effect on Nimonic 90, which somehow improved ‘glaze’ adhe-
wear regime in this case; however, the grooves on the sion and/or strength on the Nimonic 90 surface.
damaged surface became more pronounced at high load.
This was reflected by increased weight loss at high
load (Figure 17). 1.15.3.4 Effects of Environmental Variables
1.15.3.3.3 Effect of Nimonic material 1.15.3.4.1 Oxygen levels and partial pressure
processing route on wear Even in environments with low oxygen partial pres-
Very similar wear properties were observed for sure19,100 (effectively removing much of the oxygen
Nimonic 80A (cast) and Nimonic 80A (HIPped)3 from the system), stable oxide layers are still able to
and it was not possible to conclude that processing form. Under vacuum, increasing pressure from high
route affected wear resistance. Wear behavior was vacuum conditions to 102 Pa was enough to result in
also similar for Nimonic 90, the only exception a decrease in friction in sliding of an iron–chromium
being the enhanced wear resistance of Nimonic 90 alloy. Buckley101 noted during the like-on-like sliding
over Nimonic 80A (cast) at 750 C (Figure 26). of clean iron that a pressure of 400 Pa (or 3 Torr) was
sufficient to prevent seizure. Lancaster26 noted that, Barnes et al.100,102,103 investigated the effects of
at 300 C, the range of sliding speeds over which partial pressure on iron–chromium alloys, ranging
severe wear was observed when a 60/40 brass was from pure iron to iron–40% chromium (Figure 27).
slid against tool steel was greatly in an oxygen atmo- Initial work in a normal atmosphere indicated
sphere compared to the level of wear observed in air high friction and seizure at 450 C for iron and
(Figure 2). between 500 and 600 C (rising slightly with
Nimonic 80 A sample Nimonic 90 sample
Loose oxide Loose oxide

(sourced from (sourced from
both Nimonic both Nimonic 90
80A and Si N ) and Si N )
3 4
providing only 3 4
providing only
limited limited protection
protection to to Nimonic 90
Nimonic 80A surface
(cast) surface
(a) ⫻1.0K 0000 20 kV 50 µm (c) x100 0000 20 kV 500 µm
‘Glaze’ Damaged Areas of ‘glaze’

Nimonic 80A formation
(cast) surface protecting
with no Nimonic 90 from
protective enhanced wear
‘glaze’ layer
(b) ⫻5.0K 0000 20 kV 10 µm (d) x2.0 k 0000 20 kV 25 µm
Figure 26 SEM micrographs of Nimonic 80A (cast) at (a) room temperature and (b) 750 C, also Nimonic 90 at (c)
room temperature and (d) 750 C, worn against a Si3N4 counterface, sliding speed 0.654 m s1 (load 7 N, sliding distance
9418 m). Reproduced from Wood, P. D. Ph.D. Thesis, Northumbria University, UK, 1997.
2.0 10−5
Coefficient of friction (m)
Wear rate (mm3 mm−1)
1.5
1.0 10−6
0.5
0.0 10−7
0 10−6 10−1 10 105 0 10−6 10−1 10 105
(a) Oxygen pressure (Pa) (b) Oxygen pressure (Pa)
Figure 27 Variation of coefficient of friction (a) and wear rate (b) of Fe–4.9% Cr with oxygen partial pressure during like-on-
like sliding at 20 C. Reproduced from Barnes, D. J.; Wilson, J. E.; Stott, F. H.; Wood, G. C. Wear 1977, 45, 634–640.
chromium content) for various iron–chromium rate on raising the relative humidity from 0 to 10%,
alloys; the only exception to this was Fe–40% Cr followed by a rapid decrease and minimum values in
where there was no seizure up to the maximum test the region of 50–70%, then by a small increase
temperature of 850 C. towards atmospheric saturation. Increasing wear was
Adhesive wear and seizure (here defined as the demonstrated for titanium at up to 30% relative
coefficient of friction rising above a nominal value humidity, followed by an erratic decline in wear
of 3.5) were dominant at oxygen partial pressures values up to saturation levels. Nickel showed a
of 106 and 105 Pa, despite there apparently being sharp decrease in wear levels between 0 and 10%
sufficient oxygen present to prevent this. On raising relative humidity, followed by increasing wear with
the partial pressure to 104 Pa, however, signifi- relative humidity up to saturation.
cant amounts of oxide were observed and areas of Oh et al.17 suggested a transition from severe to
compacted debris had developed. These ‘islands’ mild wear with increasing relative humidity for car-
were specified as the reason for the switch from bon steels. At low relative humidity, severe wear was
severe to mild wear, with even more rapid develop- encountered, with total losses amounting to
ment of these oxides on raising the partial pres- between 0.130 and 0.190 g. This remained the case
sure to 101 Pa. The compacted debris was either up to 50% relative humidity. The level of wear
completely oxidized or oxide-covered metallic dropped rapidly after this ‘transition point,’ with
debris; its formation was accompanied by decreases mild wear being observed at relative humidity levels
in friction (Figure 27). However, despite the pres- of 70% and losses totaling no more than 0.002 g. The
ence of this oxide debris, the wear rate remained amounts of carbon in the steel were observed to
high until the oxygen partial pressure reached 1 Pa affect this transition, which occurred at higher
or above. values with increasing carbon content. Friction was
Changes in partial pressure were also made during also observed to fall rapidly, from between 0.62 and
sliding tests,100 with oxygen in some cases being 0.68 at 35% relative humidity to between 0.44 and
removed from the wear system (the pressure was 0.48 at 70% relative humidity.
from 101 to 106 Pa). When this occurred, the oxide
debris and the compacted oxide layers remained at the 1.15.3.4.3 Other atmospheres
wear interface, showing continued stability and wear The wear process in atmospheres other than air or
resistance even without a continued supply of oxygen. oxygen will depend on whether the atmosphere is
oxidizing or reducing. In most practical situations,
1.15.3.4.2 Effect of water vapor and corrosion product will not form in a reducing (non-
relative humidity oxidizing) atmosphere and, thus, the formation of
The presence of water vapor in the atmosphere can wear protective layers is not possible. Only the pres-
have a positive or negative effect on oxide develop- ence of adsorbed gases or other volatiles will act to
ment, depending on relative humidity levels and separate the wear surfaces, adhesion, and therefore,
materials. For mild steel a decrease in wear was levels of wear and friction.108
observed with increasing relative humidity under In other oxidizing atmospheres, only carbon dioxide
fretting conditions104,105 and similarly with carbon has been examined to any significant extent. Sullivan
steel under sliding wear conditions.17 It has been and Granville51 showed that a compacted oxide layer
suggested that adsorbed moisture might have a dual was formed when a Fe–9% Cr steel was tested in a pin-
effect,106 in that on the debris surface, it might act as a on-disk rig in carbon dioxide, between 200 and 550 C.
lubricant, promoting speedier debris dispersal and, Smith109 observed the formation of compacted oxides
thus, less abrasive wear. From this, it was proposed on wear testing of 316 stainless steel in carbon dioxide
that the hydrated form of the iron oxide that develops at temperatures between 20 and 600 C. The carbon
in the presence of the moisture might be a less abra- dioxide acted as the oxidizing agent in each case:
sive medium. Such debris does, however, have the
nCO2 þ M ! nCO þ MOn ½I
potential to enhance interface contact and bring
about adhesive wear. with the wear mechanisms being very similar to those
Experimental work by Bill107 demonstrated that in air.
this could be the case, with the relationship between Research in other environments is extremely lim-
relative humidity and wear rates becoming quite ited. Bill107 obtained ‘prodigious amounts of black
complex. Iron showed a significant increase in wear debris’ on testing titanium in supposedly pure dry
nitrogen. It may be that this is removed oxide that was Iwabuchi et al.65 studied the effects of preoxidation
present prior to sliding (or formed during sliding by of a number of samples of S45C carbon steel;
reaction with the trace oxygen invariably present in for each sliding test, a moving disk specimen was
bottled gases) or even a mixture of metallic titanium rotated against a fixed ring specimen, with both
and oxide. The possibility of titanium nitride is preoxidized samples and nonoxidized samples under-
extremely unlikely, if not impossible, due to the tests going unidirectional sliding for 1000 m at room
being carried out at room temperature. However, no temperature. The preoxidation treatment was carried
attempt was made to analyze the debris or explain the out at 300 C for 5 min, 1 h, or 3 h. Times of 5 min
result. and 1 h did result in progressive decreases in wear
(Figure 28); however, increasing the time to 3 h
produced no further improvement. The observed
1.15.3.5 Effects of Pretreatment of Sliding
decreases in wear occurred because of break-down
Surfaces
of the oxide layer to form debris, the presence
1.15.3.5.1 Preoxidation of which prevented metal contact and adhesion.
Stott and Mitchell52 carried out preoxidation (oxida- Iwabuchi et al. conducted similar experiments
tion prior to commencement of sliding) on Jethete with 304 stainless steel specimens,65 with specimen
M152 (a high chromium steel) and 321 stainless configurations identical to those for S45C plain
steel. Elimination of metal-to-metal contact was carbon steel. Preoxidation was carried out at 300 C
observed immediately on commencement of sliding for times of up to 10 h. Although there was some
in the case of Jethete M152 and also the immediate scatter in the data after 300 m of sliding, there
establishment of compacted oxide in the case of was no evidence of any effect of preoxidation on
the 321 stainless steel. They concluded that the pre- overall wear, regardless of the time of preoxidation
oxidation provided an extra supply of oxide debris and it was concluded that preoxidation had no effect
that led to the more rapid establishment of ‘glaze’ under the prescribed test conditions. This was due
surfaces. to ‘selective oxidation’ of chromium at the surface;
(a) Standard test, no preoxidation (d) Preoxidation at 300 ⬚C for 3 h

(b) Preoxidation at 300 ⬚C for 5 min (e) Presliding for 300 m at 300 ⬚C
(c) Preoxidation at 300 ⬚C for 1 h (Loose accumulated oxide layer formed)
4
Wear volume (mm3)
(a)
3
(b)
(d)
2
(c)
1 (e)
0
0 200 400 600 800 1000
Sliding distance (m)
Figure 28 Effects of preoxidation and presliding on wear of S45C at 20 C.34 (a) Standard test, no preoxidation, (b)
preoxidation at 300 C for 5 min, (c) preoxidation at 300 C for 1 h (loose accumulated oxide layer formed), (d) preoxidation at
300 C for 3 h, and (e) presliding for 300 m at 300 C.
this decreased the level of oxidation of the stainless Table 3 Wear rates of case-hardened steels before and
steel, with the result that there was insufficient after implantation of oxygen ions, 400 m sliding distance
oxide to decrease the severe wear rate by debris Wear rate (1016 m3 m1)
generation from this layer.
Thus, although preoxidation can decrease or elim- Steel ball Tungsten carbide
ball
inate early metal–metal contact, this cannot be guar-
anteed. Under certain circumstances, possibly due to 30% RH 80% RH 30% RH 80% RH
the effects of variations in alloy composition on the AISI 52100
nature of the surface oxidation, the production of Untreated 1000 100 17.5 17.5
suitable surface oxide required for the promotion Implanted 1 3.4 1.8 0.3
of early ‘glaze’ formation may not occur. AISI 440B
Untreated 19 25 34 85
Implanted 1.4 2.5 2.8 1.8
1.15.3.5.2 Presliding AISI M2
Iwabuchi et al.65 additionally looked at samples of Untreated 11.5 2.9 11.2 20
Implanted 3.3 1.9 1.6 3.3
S45C carbon steel that had undergone presliding for
300 m at 300 C; this was sufficient to create an accu- Ion implantation conditions 5 1017 cm2, 50 keV; test conditions
mulated loose oxide layer. Subsequent tests indicated in air at 30% or 80% relative humidity RH, 28 103 m s1.
the complete elimination of the severe wear regime Reproduced from Stott, F. H.; Wood, G. C.Tribol. Int. 1978, 11,
211–218.
(Figure 28), with only mild wear being observed,
regardless of test temperature. In the case of preox-
severe wear period. Certain aspects of heat treat-
idation, the oxide layer served at best only to reduce
ment of the alloys and the sliding conditions were
(dependent on combination) the severe wear ‘run-in’
observed to affect this. For example, in the case of
period (Section 1.15.3.5.1).
the chromium and carbon steels, the improvement
A decrease in severe wear was also observed when
in wear resulting from oxygen ion implantation was
304 stainless steel underwent presliding for 100 m at
noticeably less for AISI 52100 and AISI 440B in the
room temperature, due again to the presence of an
annealed form compared to the martensitic form;
accumulated loose oxide layer. In both cases, the
this can be seen from the data in Tables 3 and 4.
availability of preexisting oxide debris acted to pre-
Relative humidity has a marked effect, as in the case
vent contact between the metallic interfaces.
of AISI 52100 steel, where the oxygen ion implanted
Presliding was additionally carried out for 304
material actually undergoes a higher level of wear
stainless steel over a distance of 300 m at room tem-
than the untreated material. These observations
perature, 200, and 400 C (at which temperature the
were attributed to the higher plasticity of the
oxide layer formed was a ‘glaze’), with the severe
annealed samples.
wear stage being eliminated in each case during
A change in the form of the debris was also
subsequent sliding. The presence of accumulated
observed, from a smooth oxide layer for the marten-
oxide from presliding did not, however, lead to a
sitic samples to loose debris for the annealed sam-
decrease in the rate of wear during mild wear.
ples. The one exception was for AISI 440B steel,
where the decrease was greater in the annealed
1.15.3.5.3 Ion implantation state and sliding was accompanied by a change in
Langguth et al.66 carried out oxygen ion implantation the state of the oxide debris from the loose form to
on a series of chromium and carbon steels (AISI the oxide layer form.
52100, AISI 440B, AISI M2) and 321 stainless steel, In comparison to standard preoxidation treatment
followed by a series of sliding experiments using a (Table 4), wear tended to be less, with the exception
pin-on-disk rig. A sliding speed of 28 103 m s1 of annealed AISI 52100, where the preoxidized sam-
was used over a sliding distance of 400 m, with rela- ples produced superior results, regardless of the
tive humidity at 30% or 80%. levels of relative humidity.
In general, recorded levels of wear for the ion- Langguth et al. note also that implantation has
implanted samples were much lower than for their been tried using different ions, including nitrogen,
untreated equivalents (Table 3). This was due to the carbon, and boron, with varying degrees of success,
oxygen in the surface layers assisting the formation though not to the same extent as oxygen, as these
of oxidized debris and, thus, decreasing the initial alternatives do not promote wear track oxidation.
Table 4 Wear rates of case-hardened steels before and emphasis on generic principles and wear systems
after implantation of oxygen ions, 100 m sliding distance have been selected that facilitate the mechanistic
Wear rate (1016 m3 m1) understanding of the ‘glaze’ formation process.
Steel ball Tungsten carbide
1.15.4.2 Microscale Studies of ‘Glaze’
ball
Formation
30% RH 80% RH 30% RH 80% RH
From a study of mainly nickel- or nickel–iron-based
AISI 52100 alloys containing significant quantities of chromium
Untreated 611 7.7 38 1.5 in like-on-like sliding, and in some cases, cobalt, it
Implanted 68 11.2 38 2.2
Oxidized 1.1 0 0.8 0.8
was established by Stott et al.6 that the oxides formed
AISI 321 have elemental ratios that differ little from the origi-
Untreated 1400 72 4000 32 nal base alloys. It was thus concluded that the
Implanted 215 72 2000 16 observed low wear and friction arise from the physi-
Oxidized 140 72 0 32 cal properties and condition of the glaze, rather than
AISI 440B
Untreated 98 9.1 42 42
their chemical compositions.
Implanted 33 0.2 84 42 Further research6,50,55–59 allowed the identifica-
tion of the following modes of compacted layer
Ion implantation conditions 5 1017 cm2, 50–100 keV in case of formation.
AISI 52100; test conditions in air at 30% or 80% relative humidity
RH, 28 103 m s1. Reproduced from Stott, F. H.; Wood, G. C. The first mode is characterized by the formation of
Tribol. Int. 1978, 11, 211–218.
transient oxides, followed by the oxide thickening
by continued oxidation by oxygen diffusion to the
substrate–oxide interface and through physical
defects.
1.15.4 ‘Glaze’ Formation – Microscale The second mode of formation is characterized by
and Nanoscale Investigations two stages. Stage one involves the formation of an
insufficiently thick layer due to unfavorable tem-
1.15.4.1 Introduction
perature and low alloy strength, possibly involving
It is clearly recognized that ‘glaze’ formation on the an extended preglaze or severe run-in period. Stage
contacting surfaces affords significant protection two involves the formation of a sufficiently thick
against wear damage and degradation. The generation oxide layer through continued break-up and con-
of HT wear-resistant surfaces in situ overcomes the solidation of ‘glaze’ debris produced in stage one.
serious limitations on materials and coatings imposed In the third mode, the ‘glaze’ does not remain
by HT wear conditions. ‘Glaze’ formation is very stable during sliding and areas of compacted
useful as it takes advantage of important events that oxide continually break down and reform.
accompany the processes of HT wear, such as oxida-
Stott et al.10,50,55,61,112 later produced a further set of
tion, debris generation, and elemental transfer
three modified mechanisms, based on their studies of
between the contacting surfaces.3–6,14,110 These events,
the elevated temperature (200–600 C) fretting wear
under certain conditions of temperature, pressure, and
of iron-based alloys. These mechanisms50 were seen
speed3–6,14,20–22,24,25,111 lead to the formation of sur-
as limiting cases for oxide debris generation, after
faces with self-functionalized HT wear resistance.
which the build-up of oxide to form compacted
Although the phenomenon of ‘glaze’ formation and
layers continued:
related general issues of HT wear have been exten-
sively studied, it is still not possible to predict the Oxidation–scrape–reoxidation: This involves a two-
precise conditions which promote ‘glazed’ surfaces. stage process. In the first step, oxide generation
Here, it has been recognized that a step forward takes place in the areas of contact between the two
would be to gain detailed knowledge of ‘glaze’ layer sliding surfaces, with general oxidation over the
microstructure and nanoscale structure evolution so apparent sliding area of contact and, also, at asperity
that the precise mechanisms of ‘glaze’ formation can contacts where temperatures exceed the general tem-
be established. This section summarizes and updates perature in the region of the sliding area of contact. In
the studies carried out on ‘glaze’ formation during the second stage, this oxide is removed by subsequent
wear of various systems. It has been written with traversals of the sliding interfaces, exposing fresh
metal for further oxidation. The debris formed may interface, as proposed by Jiang et al.19 – these
then be either completely removed from the inter- included (1) rotation, (2) skidding, (3) rolling, and
face, act as a third body abrasive, thereby contributing (4) adhesion/sintering affected rolling (Figure 5).
to the wear process or be compacted to form a wear- Jiang et al. observed that the friction levels for mech-
protective oxide layer. anism (4) are highly dependent on the adhesion force
Total oxidation : Under certain conditions, particu- between particles in the sliding system. While this
larly high ambient temperatures, oxide generated adhesion force is weak, friction levels for mechanism
during sliding or even present prior to the com- (4) are lower than those for mechanism (2); however,
mencement of sliding, is not completely removed increasing this adhesion force above a critical level
by subsequent traversals of the sliding interfaces, results in a situation where the reverse is the case.
allowing the oxide to thicken with time. Provided Skidding then becomes the dominant mechanism,
this layer is coherent and adherent to the metal with no relative movement between neighboring
substrate and can withstand the stresses of sliding, particles – an increase in adhesion force locks them
a plastically deformed wear-protective oxide layer in place. A stable compact layer can result and, at
can develop. higher temperatures, a wear-resistant ‘glaze type’
Metal debris : Debris particles generated during the layer is possible (with the particles locked together,
early stages of wear are broken up by the sliding there is sufficient time for sintering). Stott’s later mod-
action, with any fresh areas of exposed metal being ification of this approach111 (the inclusion of adhesion
subject to further oxidation. There may be a high and sintering effects in rotation, skidding and rolling
level of oxidation of the debris surfaces, due to the mechanisms) better recognizes the fact that adhesion
relatively large exposed surface area of metal. and sintering has a more general effect on the various
particles making up the particle layer, regardless of
Enhanced oxidation is promoted by heat of deforma- whether they are entrapped or in relative motion.
tion and increased energy of the particles due to Adhesive forces and sintering tend to take effect
increased defect density and surface energy (the at more elevated temperatures, as demonstrated by
exposed surface area of debris material will increase experimental work carried out by Jiang et al.17 on the
as particle size decreases). There is also an input due sliding wear of Nimonic 80A at 20, 150, and 250 C.
to the heat of oxidation, promoting oxidation of the At 20 C, a thick layer of compacted, fine wear
metallic particles, the finer particles undergoing debris was formed, with some evidence of solidifica-
complete and spontaneous oxidation. The resulting tion and sintering in some areas; however, these
oxide can later develop into a wear-protective layer. layers were found to be predominantly particulate
The formation of compacted layers has also been in nature. There is a transition from metal–metal
observed by Wood et al.,3,12 Rose,4 Inman et al.,5,21–25 wear to contact between these primarily oxide par-
Datta et al.,20 and Du et al.110 ticle layers, at which point increases in contact
The third mechanism proposed in Stott’s original resistance and decreases in levels of wear are coin-
work is similar to that of Lin and Wood,6 both cident. At 250 C, sintering becomes a significant
depending on the generation of larger debris from factor and there is a tendency to form smooth
the wear substrate and the comminutation of this ‘glaze’ layers on top of these compacted oxide layers.
debris to fine oxide particles as the wear process The 150 C case was intermediate, with some devel-
continues to develop. opment of smooth load-bearing areas between the
The one major difficulty with these mechanisms is particulate layers. Removal of this more loosely
that they were developed from work on low speed compacted material by ultrasonic cleaning in ace-
reciprocating sliding wear, where frictional heating is tone left behind the more compacted debris, load-
not such an important factor.61 At sliding speeds of bearing areas.
greater that 1 m s1,104 frictional heating increasingly These observations clearly indicate that tempera-
becomes an issue. ture is a major driving force for adhesion between
particles and formation of load-bearing compacted
debris layers. This was demonstrated further17 by
1.15.4.3 Third Body Interaction in Relation
a heat treatment for 90 min at 600 C, for samples
to Compact Oxide Formation
preslid at 20 C. The compacted layers formed
In Section 1.15.2.2.3, a brief discussion was made of during the sliding phase of the test became solidly
four mechanisms of particle behavior at the sliding sintered together as a result of the subsequent
heating of the samples. The effect of a very small ðKp ¼ Ap expðQ p =RT0 ÞÞ. Shrinkage and thermal
particle size would be to increase the available sur- stress-induced cracks were, however, observed on
face energy, due to the resultant increase in relative layers sintered at 600 C.
surface area. This would act to drive the adhesion From experimental observations, Jiang et al.
and sintering processes and allow for observable sin- proposed a descriptive model of the sliding wear
tering at temperatures where sintering of the larger process.2,112 Figure 29 shows this diagrammatically,
particles used in powder technology applications with possible modifications to it for the reincorpora-
would not be noticeable. As adhesion itself is tion of debris for broken-down compacted oxide and
temperature-dependent, increases in temperature ‘glaze’ layers:
due to ambient or frictional heating would accelerate
1. Generation of wear particles due to the relative
the adhesion and, therefore, the sintering process.
movement of the metal surfaces;
An Arrhenius relationship1 influences the process
Relative movement of material surfaces
Generation of
wear debris particles
Retained between Removed from

rubbing surfaces rubbing surfaces
Partially removed
Remain within the rubbing

surfaces and undergo:
‘Recycling’ or
reincorporation
Comminution and oxidation
Agglomerated and compacted,

Protective
giving compact layers
Compacted oxide
layers in absence of
No debris retention?
Is T > Tc?
Yes
Sintering: Break-down
chemical plus thermal of the layers
‘Glaze’ layers More protective
Break-down Wear
Figure 29 ‘Modified’ version of Jiang’s diagrammatic representation of sliding wear processes at various temperatures.2,5
The original diagrammatic representation2 does not contain any ‘feedback’ loops to account for recycling of broken down
‘glaze’ layers and debris. A decision box is inserted to describe what happens with the actual contact temperature T in
relation to Jiang’s critical temperature Tc. Also, what happens in the case of partial retention, plus the formation of compact
layers when there is no significant debris retention, are not covered. Suggested modifications are indicated in bold.
Reproduced from Inman, I. A. Ph.D. Thesis, Northumbria University, UK, 2003; published by ‘Dissertation.com’, 2006.
2. Removal of some particles from the wear tracks to Jiang estimated also that, to cause a transition from
form loose wear particles; severe to mild wear, only 20% of the surface needed
3. Retention of other particles within the wear track; to be covered by ‘glaze’-type layers, compared to a
4. Comminution of the retained particles by value of greater than 30% for nonglaze compacted
repeated plastic deformation and fracture, with layers, demonstrating that ‘glaze’ layers offer more
particles freely moving between the rubbing sur- physical protection.
faces and undergoing partial or even complete Jiang’s model is based on experimental work done
oxidation, due to continued exposure of fresh on fretting wear systems and Rose4 questioned
metallic surfaces during comminutation; the applicability of Jiang’s model in a low debris reten-
5. Continued fragmentation and agglomeration at tion system. However, Jiang does account for debris
various sites on the wear surfaces, due to adhesion removal from the system leading to wear as would
forces between solid surfaces originating from sur- occur in, for example, high-speed unidirectional
face energy and the formation of relatively stable sliding3–5,20–25,110 or debris removed by introduced
compact layers. interfacial airflow.46 Also, in Rose’s own ‘reciprocat-
ing-block-on-rotating-cylinder’ experimental work (a
This has two effects, that is: higher speed unidirectional sliding configuration),
1. First, material loss is reduced by a material recy- examples of debris retention with ‘glaze’ formation and
cling effect of the wear debris particles. Material compacted layers do occur (i.e., Incoloy MA956 vs.
breaking away from the compacted debris may Stellite 6), despite the more adverse sliding conditions.
rejoin it. However, as Rose points out, there are examples where
2. Second, due to heavy deformation and oxidation compacted oxide layers do form and there is no signifi-
of the wear debris particles, the layers formed are cant third body debris retention,3–5,23,58 thus suggesting
hard and wear-protective. that Jiang’s model requires further modification to
account for this; Figure 29 includes some suggestions
Two competitive processes then occur during by Inman5 to cover these points.
subsequent sliding, that is:
(a) The compacted layers are continually broken 1.15.4.4 Nanoscale Investigations of
down, the debris generated promoting wear ‘Glaze’ Formation
(though, again, reincorporation may occur).
This section reports recent findings on the evolution
(b) Continuing sintering and cold welding between
particles within the layers, leading to further of structures and substructures of ‘glaze’ layers at the
consolidation. nanoscale level,20,25,113 from extensive research car-
ried out in the authors’ laboratory. Data relating to
For the latter case to predominate, the temperature transmission electron microscopy (TEM) and scan-
must be high enough (in excess of a critical tempera- ning tunneling microscopy (STM) studies are pre-
ture) to encourage the sintering processes required to sented in Figures 30–44. Scanning tunneling
ensure the formation of a solid wear-protective layer on spectroscopy (STS) and current imaging tunneling
top of the compacted particle layers before the layers spectroscopy (CITS) data are not shown here.
are broken down. The effects of this can be seen in the The work presented here is generic in nature, but
experimental work of Jiang et al. – at 20 C, this critical refers to a particular system in order to facilitate
temperature was not reached and the debris, although understanding of the underlying mechanisms for
undergoing compaction, did not sinter to form a ‘glaze’ nanostructure evolution in the surface ‘glaze’ layers
layer. At 250 C, ‘glaze’ was clearly visible, while at and their improved HT wear resistance. In this
150 C, closer to and possibly just above the critical case, the ‘Nimonic 80A (sample) versus Stellite 6
temperature, more limited sintering meant; although (counterface)’ system is considered, using a counter-
some ‘glaze’ areas were formed, there were also sub- face surface sliding speed of 0.314 m s1, with a
stantial areas of loose debris still present. 7 N applied load and total sliding distance of
Also of note is the formation of compacted debris 4522 m.5,23–25 The sample also underwent a recipro-
layers close to the center of the wear scars, the cating action against the counterface for a distance of
location of highest debris retention,18 where coverage 12 mm, three times a minute, thereby limiting debris
by high resistance compacted layers was estimated to retention to between 20 and 30% (most of the debris
be 20–50%. was allowed to escape).
0.05
B C
Weight change (g)
-0.05
-0.1
0.314 m s–1
0.654 m s–1
0.905 m s–1
-0.15
-0.2
0 100 200 300 400 500 600 700 800
Temperature (°C)
Figure 30 Weight change versus temperature for Nimonic 80A slid against Stellite 6 at 0.314, 0.654, and 0.905 m s1
(load 7 N, sliding distance 4522 m).22 At 0.314 m s1: (A) Loose oxide debris spread across wear surface, plus some compacted
oxide formation, room temperature to 390 C; smooth ‘glaze’ layers with only a little loose debris (450 C to 750 C). At 0.654
and 0.905 m s1: (B). Loose oxide debris spread across wear surface, plus some compacted oxide formation (room temperature
and 270 C at both 0.654 m s1 and 0.905 m s1, also 390 C at 0.654 m s1). (C) Torn, metallic surface typical of adhesive wear,
no traces of any oxide debris (390 C to 510 C) or only very limited oxide debris formation (at 0.905 m s1 – 570 C; at
0.654 m s1 and 0.905 m s1 – 630 C). (D) Loose oxide but with no compact oxide layers at 0.654 m s1 and only isolated
build-ups of oxide at 0.905 m s1; fine parallel grooves in direction of sliding on Nimonic 80A surface (690–750 C).
1.2
Increased ‘run-in’ friction
1.0
Coefficient of friction (µ)
0.8
0.6
0.4
More settled ‘steady state’ phase
0.2
0
0 1000 2000 3000 4000 5000
Sliding distance (m)
Room temperature 750 °C
Figure 31 Coefficient of friction versus time for Nimonic 80A versus Stellite 6 at 0.314 m s1, room temperature and
750 C.5,21,25
(a) (b)
1 µm 1 µm
(c) (d)
200 nm
2 3
(e) (f)
200 nm 200 nm
Figure 32 TEM overview of a cross-section of the wear affected surface (Nimonic 80A vs. Stellite 6, 0.314 m s1, 20 C,
load 7 N, sliding distance 4522 m): (a) bright-field overview image of the Nimonic 80A and the ‘glaze’ layer; (b) dark-field
image of the nanocrystalline ‘glaze’ layer (obtained with a 10 mm objective aperture as indicated in the SAD pattern by );
(c) selected area diffraction (SAD) pattern of the cross-section; (d) bright-field image of the interface; (e) dark-field image of the
nano-crystalline ‘glaze’ layer close to the interface (as indicated in the SAD pattern by ); (f) dark-field image of a larger crystal
in the Nimonic 80A close to the interface (as indicated in the SAD pattern by ). Reproduced from Du, H. L.; Datta, P. K.; Inman,
I.; Geurts, R.; Kübel, C. Mater. Sci. Eng. A 2003, 357, 412–422.
1.15.4.4.1 Wear data 450 and 750 C, with maxima in the mean values at
Figure 30 displays the wear data for the Nimonic 510 and 630 C of 0.001(1) and 0.001(4) g respectively.
80A/Stellite 6 system as a function of temperature at It should be noted that the wear data at 570 C show
0.314 m s1. Weight changes after 4522 m of sliding a departure from the general trend and repeated
were extremely low for all temperatures, with the experiments indicate the 570 C situation needs fur-
largest mean change being 0.002(4) g at 270 C. Slight ther attention. The following discussion focuses on
gains were observed for all test temperatures between the situation at room temperature and 750 C.
1 µm
200 nm 200 nm
Figure 33 HAADF-STEM (200 mm camera length) overview image and close-up of the interface and the nanocrystalline
‘glaze’ layer (Nimonic 80A vs. Stellite 6, 0.314 m s1, 20 C, load 7 N, sliding distance 4522 m). Reproduced from Du, H. L.;
Datta, P. K.; Inman, I.; Geurts, R.; Kübel, C. Mater. Sci. Eng. A 2003, 357, 412–422.
CCD pattern
3
1µm
CCD pattern CCD pattern
Figure 34 HAADF-STEM overview image with examples from a CCD line trace (Nimonic 80A vs. Stellite 6, 0.314 m s1,
20 C, load 7 N, sliding distance 4522 m). Reproduced from Du, H. L.; Datta, P. K.; Inman, I.; Geurts, R.; Kübel, C. Mater. Sci.
Eng. A 2003, 357, 412–422.
The coefficient of friction values measured dur- produced by fragmentation (Figures 32(b) and 32(e)),
ing the tests showed an initial period of rapid of typical diameter of 5–20 nm. Although the adhesion
change, before, in many cases, settling down into a between the ‘glaze’ layer and Nimonic 80A substrate
‘steady state’ situation with less variation. The data in was generally good, cracks were visible in some areas.
Figure 31 show no significant changes with time Figure 33 shows a HAADF-STEM overview
after an initial peak at the very beginning of the image of the FIB prepared cross-section described
test. This indicates the rapid onset of debris genera- in the previous paragraph.113 This is consistent with
tion (up to 450 C) or ‘glaze’ layer formation (450 C the structural variations observed by TEM. Two dif-
and above, almost immediate at 750 C), with the lack ferent areas are revealed for the substrate: a uniform
of frictional variation during the ‘steady state’ stage ‘bulk’ area and a large grained structure about one
being due to the continued presence of debris or micron below the interface with the ‘glaze’ layer. The
‘glaze’ on the worn surfaces (Figure 24). Coefficient latter exhibits a fine-grained structure (5–20 nm)
of friction levels (Figure 31) are significantly lower for with irregular shaped grain boundaries. The particles
‘glaze’ covered surfaces (i.e., 750 C) than for loose are larger close to the interface (up to about 50 nm).
debris covered surfaces (i.e., room temperature). Furthermore, the strong Z-dependence of the
HAADF-STEM image reveals several metal parti-
cles in the (oxidized) ‘glaze’ layer as well as low-
1.15.4.4.2 Studies of wear-affected surfaces
density material at the interface.
produced during sliding of Nimonic 80A
Figure 34 shows examples from a STEM line trace
against Stellite 6 at 20 C
of nano-diffraction patterns, taken using a 30 nm
probe.113 As previously, only small randomly oriented
Scanning electron microscopy (SEM)
crystals are observed in the ‘glaze’ layer and at the
The top layer of the wear surface after sliding
interface. The larger grains in the alloy exhibit various
at 20 C (Figure 24) consisted of a loosely bound
crystallographic orientations; some of them well
noncompacted oxide that was easily detachable by
defined and similar to those expected for Nimonic 80A.
brushing or light polishing. Underneath this layer,
Figure 35 shows EDX analyses of the ‘glaze’
there was a more compacted layer that was not
layers, together with their reference areas.113 The
clearly apparent using optical and scanning electron
quantification is based on theoretical k-factors and
microscopy.113
uses a thickness correction for an estimated 150 nm
Analysis of the surfaces reveals the presence of
sample thickness (EELS thickness measurements
oxygen and elements from Nimonic 80A (typically
indicate a thickness of 2–3 nm mean free path).
Ni/Cr) and Stellite 6 (Co/Cr). The results indicate
Quantitative analysis of the Nimonic 80A layer
elemental transfer from the counterface to the speci-
(Area 1) gives the characteristic composition of the
men and mixing of the transferred and host element
bulk alloy (Table 5), apart from a slightly higher
oxides. In this section, attention is focussed on the
silicon concentration and a small amount of cobalt.
surface layers produced at 20 C.113
Analysis of the ‘glaze’ layer shows that it is es-
sentially oxidized Stellite 6, with a significant
Transmission electron microscopy (TEM) contribution of between 15–20% Ni from the
Figure 32(a) shows a cross-sectional TEM image of Nimonic 80A.
the wear-affected Nimonic 80A surface, revealing the The interface layer consists of a mixture of Nimo-
interface between the Nimonic 80A substrate and the nic 80A and Stellite 6, with a higher than average
compact ‘glaze’ layer (the loosely bound uncom- titanium concentration.113 HAADF-STEM EELS/
pacted oxide debris layer was not visible in this EDX line traces across the interface (Figure 36)
area).113 The associated selected area diffraction show the transition between the ‘glaze’ layer and the
(SAD) pattern (Figure 32(c)) is dominated by Nimonic 80A substrate, based on the cobalt and
Debye rings and some well-defined Bragg reflections. nickel concentrations; the composition of the ‘glaze’
Dark-field images (Figures 32(e) and 32(f)) reveal that layer is locally inhomogeneous (Figures 36 and 37).
the latter correspond to irregular shaped Nimonic 80A There are some variations in the chromium, cobalt
crystals of several hundred nanometres diameter, in a and oxygen concentrations as well as a few distinct
layer up to 1 mm from the ‘glaze’ layer/substrate inter- particles with a high nickel concentration. Neverthe-
face. The ‘glaze’ layer shows a very different morphol- less, the overall concentrations are relatively uniform
ogy, consisting of small and misoriented crystals throughout the ‘glaze’ layer, for example, no nickel
2000
O EDX area 3
Co
1500
Counts
Ni Cr
1000
3
Ni
2 500
1 Si
C Cr Cu Ni
Al Ti Fe
2 µm 0
0 2 4 6 8 10 12
Energy (keV)
1500
Ni EDX area 1 EDX area 2
4000 Ni
1000
3000
Counts
Counts
Ni
Cr
2000
Ni Cr
500
1000 O
Cu Ni Co Ni
C Cr Co C AlSi Ti Cr Cu
Al Si Ti Fe Cu Fe Cu
0 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Energy (keV) Energy (keV)
Figure 35 HAADF-STEM (200 mm camera length) overview image and local EDX analysis (Nimonic 80A vs. Stellite 6,
0.314 m s1, 20 C, load 7 N, sliding distance 4522 m). Reproduced from Du, H. L.; Datta, P. K.; Inman, I.; Geurts, R.; Kübel, C.
Mater. Sci. Eng. A 2003, 357, 412–422.
concentration gradient was observed. The line traces traces reveal a low oxygen concentration for the bright
further reveal the preferential formation of titanium areas, indicating that the image mostly reflects the
and aluminum oxides in a 100 nm thick layer at the oxidation state. Furthermore, a low oxygen concentra-
interface. Slight chromium depletion is observed tion coincides with a low chromium concentration;
in the substrate close to the interface, with a local EDX analysis shows a low chromium and oxygen
corresponding chromium enrichment in the first concentration for the particles that appear bright in the
300 nm of the ‘glaze’ layer. image. This implies that some of the nickel and cobalt
Similar areas of enrichment/depletion have particles are not completely oxidized.
been found at other locations along the interface; an EDX line traces in an area up to about one micron
example is given at a higher spatial resolution in below the interface reveal the preferential segrega-
Figure 37.113 Furthermore, this line trace also con- tion of light elements, especially aluminum and in
firms aluminum enrichment at the interface; more- some cases titanium (Figure 39).113 This segregation
over, this Al2O3 layer is between the TiO2 layer and is also visible in the HAADF-STEM images as thin
the Nimonic 80A substrate. Aluminum (in Al2O3) lines (<10 nm diameter) oriented parallel to the
enrichment is further responsible for the dark areas interface (Figures 33–39). It is noted that the peaks
in the HAADF-STEM image, leading from the inter- of Ti and Al are overlapped away from the interface,
face into the substrate (Figure 26). which indicates the presence of sub-microscopic
The atomic numbers of chromium (24), cobalt g’-phase, Ni3Al or Ni3(TiAl) particles, as expected
(27) and nickel (28) are similar and compositional following ageing.113
variations of these main elements do not explain the These results clearly indicate the formation of a
strong contrast observed in the HAADF-STEM wear-resistant nanostructured surface during sliding
image (Figure 38).113 However, the EDX/EELS line wear of Nimonic 80A against Stellite 6 at 20 C using
1 µm
35 000 STEM HAADF detector 1000 1 Ni

Co
30 000 800
Count
600
Count
25 000
400
20 000
200
15 000
0
0 1000 2000 0 0.5 1.0 1.5 2.0 2.5
Position (nm) Position (µm)
1 O 500 1 Cr
Ti Ti
400 O
10 000
Count
Count
300
5000 200
100
0
0
0 0.5 1.0 1.5 2.0 2.5 0 0.5 1.0 1.5 2.0 2.5
Position (µm) Position (µm)
Ti O Ti Cr Cr Cr EDX
2000 30
Ni
1500
Count
Count
20 O
1000 Ti
Ni
Si
500 10 Ni
Cr Co
Ti Fe
0
0
300 400 500 600 700 0 2 4 6 8 10 12
Energy (eV) Energy (keV)
Figure 36 HAADF-STEM EELS/EDX line trace revealing the compositional variations of the wear affected surface (Nimonic
80A vs. Stellite 6, 0.314 m s1, 20 C, load 7 N, sliding distance 4522 m); line trace performed with 1 nm probe size and spectra
collected every 10 nm. Reproduced from Du, H. L.; Datta, P. K.; Inman, I.; Geurts, R.; Kübel, C. Mater. Sci. Eng. A 2003,
357, 412–422.
a speed of 0.314 m s1 under a load of 7 N.113 The The nickel concentration (from the Nimonic 80A)
analyses reveal the complex structure of this surface, is uniform throughout the whole layer, except for
which consists of multiple layers: the presence of several larger nonoxidized particles
that were randomly dispersed in the glaze.
1. A loose, uncompacted, highly oxidized layer.
3. Larger particles with a diameter of up to 50 nm
2. A nano-crystalline (5–20 nm grain diameter) com-
were observed in a layer of about 300 nm at the
pacted layer mostly consisting of oxides of elements
base of the ‘glaze’ layer. The compositions indicate
originating both from Stellite 6 and Nimonic 80A.
35 000 STEM HAADF detector

1
Counts
1 30 000
1 25 000
0 200 400 600 800 1000

Position (nm)
1
30 EDX HAADF detector line 1
Counts
20
Cr
Ti
O
10 Ti Ni
Co Ti Cr
Al Si Co Ni W
500 nm 0
0 2 4 6 8 10
Energy (keV)
1 500 1
Ti Ni
100 Al 400 Co
Counts
O Counts 300 Cr
50 200
100
0 0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Position (µm) Position (µm)
Figure 37 HAADF-STEM (200 mm camera length) overview and EDX line trace revealing the compositional variations of the
wear interface structure (Nimonic 80A vs. Stellite 6, 0.314 m s1, 20 C, load 7 N, sliding distance 4522 m); the line trace was
performed with a 1 nm probe size and spectra collected every 2 nm. Reproduced from Du, H. L.; Datta, P. K.; Inman, I.;
Geurts, R.; Kübel, C. Mater. Sci. Eng. A 2003, 357, 412–422.
the presence of elements from both Stellite 6 and contacting surface/particles subjected to high mechani-
Nimonic 80A, but chromium enriched. cal stress due to rotation of the counterface. The ‘glaze’
4. The chromium-rich phase gradually gave way to a layer, consisting mainly of the oxides of the elements
titanium (oxide)-rich phase of about 100 nm thick- originating from the contacting surfaces, acted as a
ness, followed by an aluminum (oxide)-rich phase barrier that separated the substrate from the environ-
of about 50 nm thickness, at the interface with the ment. Although the mechanical effect would continue to
Nimonic 80A substrate. influence the ‘glaze’ layer and subsurface deformation,
5. Adjacent to the interface with the ‘glaze’ layer, the the subsequent oxidation process is controlled by oxy-
Nimonic 80A substrate exhibited a large-grained gen species and substrate element diffusion, and the
structure (of diameter of several hundred nano- thermodynamic nature of the relevant oxides. The
metres). Slight chromium depletion was observed. observed sequential existence of the oxides Cr2O3/
The precipitation lines (diameter < 10 nm) TiO2/Al2O3/substrate at the interface can be under-
indicated that aluminum-, and in some cases tita- stood in terms of thermodynamics. Figure 40(a) indi-
nium-, enriched phases formed near the large- cates the oxidation tendencies of relevant elements.113
grained structures. The larger dark areas perpen- The dissociation partial pressures for Al2O3, TiO2,
dicular to the interface were also due to aluminum Cr2O3, NiO, and Co3O4 follow the same trend in the
enrichment. range of temperatures 25–1000 C. Al2O3, TiO2
6. A uniform ‘bulk’ Nimonic 80A composition was and Cr2O3 are more likely to develop with increasing
observed at depths greater than one micron below temperatures. Although the temperature generated was
the interface. not measured, the relatively fast kinetics of oxidation, as
evidenced by the rapid formation of the ‘glaze’ layer,
The first key process in the development of nano- indicates the generation of relatively high temperatures,
structured ‘glaze’ layers involves oxidation of the sufficient for such oxidation to occur.
31 000 50
STEM HAADF EDX HAADF detector
detector line 1
1 40
30 000
30
Counts
Counts
Ni
1 29 000 20 Co
Ni
1
10
28 000 O Al Cr
Si Ti Co Ni
0
200 nm 0 100 200 300 400 0 2 4 6 8 10 12
Position (nm) Energy (keV)
80 1 1
Ti Ni
O 500 Co
60 Al Cr
400
Counts
Counts
300
40
200
20
100
0 0
0 100 200 300 400 0 100 200 300 400
Position (nm) Position (nm)
Figure 38 HAADF-STEM (50 mm camera length) overview and EDX line trace revealing the aluminum (oxide)
segregation at the top of the Nimonic 80A phase (Nimonic 80A vs. Stellite 6, 0.314 m s1, 20 C, load 7 N, sliding distance
4522 m); the line trace was performed with a 1 nm probe size and spectra collected every 2 nm. Reproduced from Du, H. L.;
Datta, P. K.; Inman, I.; Geurts, R.; Kübel, C. Mater. Sci. Eng. A 2003, 357, 412–422.
Additionally, it is apparent that the high Cr dominant presence of Co, Ni, Cr and O on the
activity in both counterface and sample materials ‘glaze’ surface.20,25 Quantification of the results on
leads to the formation of Cr2O3 preferentially in the average gives 34.2% Co, 36.2% Cr, 16.7% Ni, 3.8%
early stages of the process. When oxygen diffuses Si, and 1.3% Fe (all at%). However, analysis by
inwards, Ti in the g0 phase is favorably oxidized. TEM-EDS showed some location-to-location varia-
The formation of TiO2 gives rise to a decrease in tion. The dominant phases identified by XRD
Ti activity and, correspondingly, the Al activity in the included CoCr2O4 and Ni2.9Cr0.7Fe0.36.
g0 phase increases and Al2O3 becomes a favorable Figure 41(a) is a cross-sectional composite TEM
product, as described for oxidation of TiAl in micrograph of the surface formed during wear.20,25 It
air.115,116 NiO and Co3O4 are much less stable than reveals the surface layer (glazed surface), the
TiO2 and Al2O3, which is the reason why some par- deformed substrate and the ‘glazed’ layer/substrate
ticles of Ni and Co have not been oxidized in the interface. The wear-affected region (total thickness
wear process. These oxides (TiO2 and Al2O3) formed 3 mm) consists of three layers; the top-most layer
beneath the ‘glaze’ layer, play an important role in (the ‘glaze’ layer) has a uniform grain structure of size
providing support additional to that provided by 5–15 nm, some of the grains displaying contrast,
the substrate, to sustain the ‘glaze’ layer and prevent while the dislocation density in this area is low. The
it from collapsing.4 interfacial layer consists of grains of 10–20 nm and
has a higher dislocation density. The layer just
1.15.4.4.3 Nano-scale microscopy of ‘glazed’ beneath the interfacial layer shows subsurface defor-
layers formed during high temperature mation and elongated grains. The SAD pattern from
sliding wear at 750 C the ‘glaze’ layer (Figure 42) consists of spots
The spectrum from EDX analysis of the surface arranged in concentric circles, indicating the pres-
generated under sliding wear at 0.314 m s1 and ence of small grains with high angle boundaries,
750 C for 4522 m, Figure 40(b), reveals the multiple boundaries and large misorientations
STEM HAADF detector

35 000
1
Counts
34 000
33 000
32 000
31 000
0 50 100 150 200
Position (nm)
50 1
1 Al
40
100 nm
Counts
30
20
EDX HAADF detector line 1 10
25
0 50 100 150 200
Position (nm)
20 Ni
Counts
60 1
15 Cr Ti
50
Counts
10 40
Co Ti
O 30
5 Al
Si Cr 20
0
0 1.0 2.0 3.0 4.0 5.0 6.0 0 50 100 150 200
Energy (keV) Position (nm)
Figure 39 HAADF-STEM (50 mm camera length) overview and EDX line trace revealing the precipitation of light
elements (aluminum and titanium) in the Nimonic 80A close to the wear interface (Nimonic 80A vs. Stellite 6, 0.314 m s1,
20 C, load 7 N, sliding distance 4522 m). Reproduced from Du, H. L.; Datta, P. K.; Inman, I.; Geurts, R.; Kübel, C. Mater. Sci.
Eng. A 2003, 357, 412–422.
(formation of misorientated lattice-fragmentation). both contacting surfaces is responsible for the gener-
The poorly-defined irregular boundaries indicate ation of the ultrafine structured surfaces. Moreover,
nonequilibrium high-energy configuration. The the detailed TEM studies presented here has enabled
indexed SAD pattern also revealed the presence of understanding of the formation mechanisms of wear-
oxides of Ni, Cr, and Co (indexing not shown here). resistant nano-structured surfaces.
Sub-surface deformation is illustrated in It is clear that the initial processes responsible for
Figure 41(b).20,25 Dislocations, present as networks generating the ‘glazed’ layer at both 20 and 750 C are
inside the deformed (elongated) grains, have been ‘deformation of the surface,’ ‘intermixing of the
observed in the deformed substrate. Shearing defor- debris generated from the wear and the counterface
mation took place in the substrate as a response to the surfaces,’ ‘oxidation,’ ‘further mixing,’ and ‘repeated
sliding process. welding and fracture.’ These processes are aided by
The present results clearly indicate the formation of high temperature oxidation and diffusion. The posi-
a nano-structured ‘glaze’ oxide layer during high tem- tron annihilation studies confirmed the presence of
perature sliding wear under the specified conditions. vacancy clusters consisting of five vacancies.110
The creation of nano-structures is confirmed by the The next stage in the process involves deforma-
STM topography, indicating grains of between 5 and tion of oxides and generation of dislocations, leading
10 nm (Figure 43).5,20,21,25 These results again demon- to the formation of sub-grains. These sub-grains are
strate that such a nano-structured surface is extremely then further refined, with increasing misorientation
effective in conferring high resistance to wear. resulting in nano-structured grains with high angle
It has been indicated by various authors5,20,21,110 boundaries (a process called ‘fragmentation’), a non-
that, in many systems, surfaces with ultra-fine struc- equilibrium state indicated by poorly defined and
ture are generated during high temperature sliding irregular grain boundaries. High internal stress is
wear. Mechanical mixing involving repeated welding, created inside the grains; the dislocation density and
fracture and re-welding of the debris generated from arrangement depend on the grain size, with smaller
−20
−40
−60
−80
log pO2
−100
−120
log ( pO2)(Cr2O3)
−140 log ( pO2)(AI2O3)
log ( pO2)(TiO2)
−160 log ( pO2)(NiO)
log ( pO2)(Co3O4)
−180
−200
5 7.5 10 12.5 15 17.5 20 22.5 25 27.5 30 32.5 35
(a) 1/T (K) ⫻ 104
0
Counts
Ni
Co
Si Cr
C Al Co
Ti Ni
0.80 1.60 2.40 3.20 4.00 4.80 5.60 6.40 7.20 8.0
(b) Energy (eV)
Figure 40 Plots showing (a) dissociation partial pressures of Al2O3, TiO2, Cr2O3, NiO, and Co3O4 versus reciprocals of
temperatures20 and (b) EDX surface spectrum for Nimonic 80A vs. Stellite 6 (0.314 m s1, 750 C, load 7 N, sliding distance
4522 m).20, 25
grains containing fewer dislocations. The process Fe–Cr-dominated debris at 0.905 m s1; the sliding
leads to the formation of high energy grain bound- speed and composition having no noticeable affect on
aries with a high defect density.117–123 ‘glaze’ structure with this latter system.
1.15.4.4.4 Other systems

Such nanostructured ‘glaze’ layers are not unique to
Nimonic 80A when slid against Stellite 6 (0.314 m s1, 1.15.5 Wear Maps: A Useful Design
7 N load, 4522 m sliding distance). For example, using Aid for Selecting Wear-Resistant
AFM, Incoloy MA956 has also been shown to form Materials and Surfaces
nanostructured ‘glaze’ layers when slid against Stellite 623
1.15.5.1 Introduction
and Incoloy 800HT at 750 C (Figure 44).114 In the case
of the Incoloy MA956/Stellite 6 system, this nanostruc- Lancaster, Welsh,27,28 So,29,30 Rose,4 Inman,5,21–24
26
tured ‘glaze’ formed has been found to originate from and others42,85 have shown that various combinations
both Co–Cr-dominated oxide debris at 0.314 m s1 and of load, temperature and sliding speed can
Surface layer
– grain size 5–15 nm
‘Glaze’ layer
– total thickness
3.5 µm
Intermediate layer
(some metallic grains)
– grain size 10–20 nm
Interface region
Deformed
1 µm Nimonic 80A
metallic
substrate
(a) (b)
Figure 41 Nanoscale characterization of ‘glaze’ layers formed (Nimonic 80A vs. Stellite 6, 0.314 m s1, 750 C, load
7 N, sliding distance 4522 m) showing TEM bright field images of (a) a wear-induced polycrystalline ‘glaze’ layer plus
substrate deformation and (b) the interface of the ‘glaze’ layer and deformed substrate.5,20,21,25
50 nm 50 nm
Figure 42 TEM morphological and structural details of the ‘glaze’ layer formed (Nimonic 80A vs. Stellite 6, 0.314 m s1,
750 C, load 7 N, sliding distance 4522 m).20,25
1200
1000
800
600
nm
400
200
0
20
15
10
0
12 nm
10
8
6
4 1.2 mm ⫻ 1.2 mm
2
0
−2
0 200 400 600 800 1000 1200
nm
Figure 43 STM surface line profile results on the ‘glaze’ layer formed on Nimonic 80A (Stellite 6 counterface, 0.314 m s1,
750 C, load 7 N, sliding distance 4522 m).5,20,21,25
Figure 44 AFM image of ‘glaze’ layer surfaces produced on Incoloy MA956 (slid against Incoloy 800HT) at 750 C and
sliding speeds of (a) 0.314 m s1 and (b) 0.905 m s1. Reproduced from Inman, I. A. Nano-scale studies of ‘glaze’ layers
formed on Incoloy MA956 when slid against Incoloy 800HT at 750 C, Northumbria University, 2003, Unpublished work.
significantly affect wear behavior, regardless of systems based on load/pressure and sliding speed.
whether or not a protective ‘glaze’ can form. Other wear-related parameters have also been used
Several authors have also constructed wear maps in and Kato and Hokkirigawa127 developed an abrasive
an attempt to present wear data in a more easily wear map using ‘degree of penetration (of asperities)’
understood format, allowing prediction of likely wear and ‘shear strength at the contact interface’ as key
mode under specified sliding conditions. Lim124,125 parameters (Figure 46(b)). Adachi et al.134 used ‘sever-
(Figure 45), Childs126 (Figure 46(a)), and more ity of contact’ and ‘thermal severity of contact’ for
recently, Riahi and Alpas,128 Chen and Alpas,129 Yang ceramic wear.
et al.,130 Grimanelis and Eyre,131,132 and Elleuch The following section discusses recent develop-
et al.133 have constructed wear maps for various sliding ments in high temperature wear maps made by the
present authors, following a brief review of earlier
Table 5 Quantification of the EDX spectra in Figure 35 work by Lim.
Element Area 1 (Nimonic Area 3 (Glaze
80 A substrate) layer) 1.15.5.2 Work by Lim
wt.% at.% wt.% at.% Lim124 attempted to collate the available data for mild
steels into a single comprehensive wear map (Figure 45)
Al 0.7 1.4 0.4 0.9
using dimensionless parameters of normalized velocity
Si 0.7 1.3 1.1 2.3
Ti 1.8 2.1 0.4 0.5 versus normalized load. This showed that the selection
Cr 18.7 20.2 25.0 27.4 of sliding conditions and configuration can greatly affect
Fe 0.7 0.7 1.4 1.4 the wear behavior and transitions observed, with load,
Co 1.0 0.9 49.8 48.2 sliding speed, and temperature potentially having a large
Ni 76.6 73.4 19.0 18.4
influence on the boundaries between the different
W – – 2.9 0.9
modes of wear.
Sliding velocity v (m s–1)

10−4 10−2 1 102
10
Load-dependent Distance-dependent Steel
Welsh (1965) Sata (1960) Wear-transition map
Ward (1970) Welsh (1957)
Norse and Sasada (1980)
Clayton (1979) Okoshi and Sakai (1941)
Quinn et al (1973) Present work
Dunckley and Quinn Seizure
(1978)
Farrell and Eyre (1970)
Velocity-dependent Severe
10−1
Normalized pressure (F )
~
Kehl and Siebel (1936)

Nakajima and Mizutani (1969)
Wear
Okoshi and Sakai (1941) ~ (a)
Welsh (1965) W
~
Archard (1958/59) W ~
F (b)
~ ~
Severe wear F W
~
v
(c) Mild wear
Transition
between ~
10−3 mild and
W
~ (b)
severe W
Distance
wear ~
~ v ~ (c)
W W
~ Distance
v
~ ~ (b)
W W
Mild wear
~ ~
F v
10−5
10−2 1 102 104
~
Normalized velocity (v )
Figure 45 Wear transition map for steels showing regions of mild and severe wear – sliding conditions corresponding to
different types of wear transitions observed are also indicated. Reproduced from Lim, S. C. Tribol. Int. 1998, 31(1–3), 87–97.
Abrasive-wear diagram
Kato and Hokkirigawa (1985)
Wear-regime map Ploughing Wedge Cutting

soft steels Brass
Childs (1980) S45C
Contact pressure/yield stress
1.0 SUS304
1.0
Severe metallic wear 0.8
Degree of penetration (Dp)

0.5
0.6
B C D E Cutting
0.4
A
0 Wedge
10−2 10−1 1 10 102 0.2
Sliding speed (m s−1) Ploughing
0 0.2 0.4 0.6 0.8 1.0

Shear strength at the contact interface (f )
(a) (b)
Figure 46 Examples of ‘alternative parameter’ wear maps: (a) Childs soft steels wear-regime map126; and (b) Kato and
Hokkirigawa’s abrasive-wear diagram for brass, S45C medium carbon steel and SUS304 stainless steel. Reproduced from
Kato, K.; Hokkirigawa, K. In Proceedings of the Eurotrib’85; Elsevier: Amsterdam, 1985; Vol. 4, Section 5.3, pp 1–5.
For example, it can be seen that, if a relatively high For both systems,23 mild wear with low weight loss
fixed load is used, with increasing sliding speed, a dominates at 0.314 m s1 regardless of temperature
transition from severe-to-mild wear is observed.27 (Figures 47 and 48). Up to 450 C, a low temperature
More complex forms result at lower loads when mild wear regime occurs, with the wear surfaces sepa-
sliding speeds are at much higher values.27,32 Simi- rated by a layer of Co–Cr oxide particles (i.e., loose
larly, if speed is fixed at an arbitrarily low value, debris – Figures 13(a) and 24(a) – potentially under-
increasing load may see a switch from mild to severe lain by a more compacted layer – Section 1.15.4.3.2),
wear124 while a higher speed can result in a more primarily sourced from the Stellite 6. Increased
complex curve.27 To summarize, when two surfaces agglomeration and sintering of the debris has been
are worn against each other, the outcome can be a observed at 390 C, with isolated patches of ‘glaze’
variety of apparently contradictory results. Thus, at 450 C (although most of the oxide remained as
mapping is necessary to understand the relative loose debris). Mild wear persists between 510 C
behaviors in different tests and the potential outcome (Figures 13(b) and 24(b)) and 750 C (Figures 13(c)
under a given set of sliding conditions. and 24(c)), with the still Stellite 6-sourced oxide
sintering to form comprehensive ‘glaze’ layers with
nanostructured surfaces5,20,21,23,25 (discussed in Section
1.15.5.3 Dissimilar Interfaces
1.15.4.4); with extremely low weight change with some
As discussed previously, Inman et al.5,22,23 studied the very slight weight gains due to oxide development
effects of temperature and sliding speed on several (Figures 47 and 48). At all temperatures, there was
superalloys using dissimilar interfaces. This section virtually no initial severe wear period, with sufficient
refers specifically to Nimonic 80A22 and Incoloy MA Co–Cr oxide debris forming extremely rapidly.
95623 worn against Stellite 6. For both systems, the data The Nimonic 80A versus Stellite 6 system22 is char-
have been used to create simple ‘temperature versus acterized by three distinct wear regimes at 0.654 m s1
sliding speed’ wear maps22,23 (Figures 47 and 48). The (Figure 47); up to 390 C, a low temperature ‘low weight
weight change data are presented in Figures 30 and 49. loss’ mild wear regime (Figure 30) was observed, with a
No oxide—greater Greater oxide generation at 0.905 m s −1

wear at 0.905 m s–1 – higher wear at 0.654 m s–1
Severe wear High (mild) wear
Increased Increased oxide
0.905 oxide debris at 0.905 m s–1,
content at isolated oxide build-up
0.905 m s−1 Loose, abrasive
Nimonic 80A
Slding speed (m s–1)
Increasing
NiO and Cr2O3 debris
-sourced metallic from nimonic 80A
oxide
debris visible only
debris Oxide generation
0.654 with but no build-up
Low temperature mild temp- at 0.654 m s–1
wear – mainly Co/Cr erature
(high losses from nimonic 80A)
loose oxide debris
generated from Stellite 6
debris with isolated

Co/Cr loose oxide
patches of ‘glaze’
High Co/Cr ‘glaze’ sourced
0.314
from Stellite 6–mild wear
0 270 390 450 510 570 630 690 750

Temperature (⬚C)
Weight change data: <0 g (weight gain) 0.001 g

0.01 g 0.05 g >0.1 g
Figure 47 Wear map for Nimonic 80A versus Stellite 6 (load 7 N, sliding distance 4522 m), with weight loss (contour) data
superimposed. Reproduced from Inman, I. A.; Rose, S. R.; Datta, P. K. Wear 2006, 260, 919–932.
Severe wear with ‘glaze’ More limited severe wear

only after extended sliding [1] before ‘glaze’ formation
4) Mainly Incoloy
5) Severe
MA956-sourced
0.905 wear 7) Mostly
Fe–Cr loose oxide
only mild wear,
debris
6) Initial severe wear Fe–Cr ‘glaze’
Sliding speed (m s–1)
Incoloy (reducing with initial

MA956- temperature), followed severe
sourced by high Fe-Cr wear
0.654 Low temperature metallic ‘glaze’ formation minimal
mild wear debris
‘Glaze’ forming
2) Co–Cr loose
temperatures
with isolated
oxide debris
1) Mainly Stellite 6-sourced

patches of
Co–Cr loose oxide debris 3) High Co–Cr ‘glaze’

0.314 sourced from Stellite 6
‘glaze’
0 270 390 450 510 570 630 690 750

Temperature (⬚C)
Weight change data: <0 g (weight gain) 0.001 g

0.01 g 0.05 g >0.1 g
Figure 48 Wear map for Incoloy MA956 slid against a Stellite 6 counterface (load 7 N), with weight loss (contour) data
superimposed. Reproduced from Inman, I. A.; Rose, S. R.; Datta, P. K. Tribol. Int. 2006, 39, 1361–1375.
layer of Stellite 6-sourced Co–Cr oxide particles separ- lower the severe-wear-to-abrasive-mild-wear transi-
ating the wear surfaces (Figure 41(a)). A metallic severe tion from 690 C.
wear (especially of the Nimonic 80A) regime dominated The behavior for the Incoloy MA956 versus Stellite
between 450 and 630 C (Figure 25(b)) with no visible 6 system follows the same general pattern at 0.654 and
oxide between 450 and 570 C. This severe wear 0.905 m s1 as that of Nimonic 80A versus Stellite
regime was also observed at 630 C, now accompanied 623 (Figure 48). However, in this case, the high
by a small amount of NiO and Cr2O3 generated by temperature mild wear regime confers protection,
thermal and frictional heating; however, this oxide was with the mixed Fe–Cr and Co–Cr oxides at
insufficient to impede metal-to-metal contact and may 0.654 m s1 or the largely Fe–Cr oxides at 0.905 m s1
instead have assisted wear by abrasion. readily sintering to form ‘glaze’ layers. Such ‘glaze’ is
A mild oxidational wear regime was evident at 690 observed from 510 C upwards (450 C after extended
and 750 C, although Nimonic 80A weight losses sliding at 0.905 m s1), after a period of early serve wear
(Figure 30) remained high due to abrasion. This (Figure 14(b)) that decreases in length with increasing
abrasion was caused by large amounts of Nimonic temperature. At and above 630 C, rapid ‘glaze’ forma-
80A-sourced NiO and Cr2O3, which (unlike Co–Cr tion (Figure 14(c)) almost completely eliminates early
oxides generated at 0.314 m s1) showed little ten- severe wear (Figure 49).
dency to sinter and form ‘glaze.’ A later study by Inman and Datta24 looked to im-
The sliding behavior at 0.905 m s1 exhibited the prove Nimonic 80A versus Stellite 6 wear map resolu-
same three sliding regimes observed at 0.654 m s1.22 tion between 630 and 750 C by adding extra test sliding
However, greater frictional heating resulted in a down- speeds. Rudimentary wear maps were also developed
ward shift in the low-temperature-mild-wear-to- separately for Nimonic 80A and Incoloy MA956 as
severe-wear transition from 390–450 to 270–390 C. sample materials versus an Incoloy 800HTcounterface.5
Increased amounts of loose NiO and Cr2O3 debris Such sliding studies indicate the potential for
occurred at 570 C upwards, but not sufficient to complex behavior during sliding of dissimilar
0.05
A B C
0.00
Weight change (g)
-0.05
-0.10
0.314 m s−1, 4522 m

-0.15
0.654 m s−1, 4522 m
0.905 m s−1, 4522 m
-0.20
0 100 200 300 400 500 600 700 800
Temperature (⬚C)
Figure 49 Weight change versus temperature for Incoloy MA956 slid against Stellite 6 at 0.314, 0.654, and 0.905 m s1
(load 7 N, sliding distance 4522 m) – effect on transition temperatures on increasing sliding speed from 0.654 to 0.905 m s1
are indicated.23 (a) Decrease in low temperature mild to severe wear transition from 390 C at 0.654 m s1 to 270 C at
0.905 m s1, (b) Decrease in temperature of first appearance of ‘glaze’ from 510 C at 0.654 m s1 to 450 C after extended
sliding at 0.905 m s1, and (c) More rapid ‘glaze’ formation at 0.905 m s1.
materials. In such cases, the necessity for mapping 4. Rose, S. R. Ph.D. Thesis, Northumbria University, UK,
2000.
wear behavior to assist prediction of mild or severe 5. Inman, I. A. Ph.D. Thesis, Northumbria University, UK,
wear is important, if potentially catastrophic material 2003; published by ‘Dissertation.com’, 2006.
failure is to be avoided. 6. Stott, F. H.; Lin, D. S.; Wood, G. C. Corros. Sci. 1973, 13,
449–469.
7. Johnson, M.; Moorhouse, P.; Nicholls, J. R. DTI Industry
1.15.5.3.1 Oxide chemistry Valve Project 1990; pp 61–68.
These studies also indicate that the readiness for a 8. Aoh, J.-N.; Chen, J.-C. Wear 2001, 250–251, 611.
wear-generated oxide to form ‘glaze’ layers depends 9. Singh, J.; Alpas, A. T. Metall. Mater. Trans. A 1996, 27,
3135–3148.
not just upon the sliding conditions, but also on the 10. Stott, F. H.; Glascott, J.; Wood, G. C. Wear 1984, 97,
chemical composition of the oxide. For example, Co– 93–106.
Cr-based oxides readily form wear protective ‘glaze’ 11. Gee, M. G.; Jennett, N. M. Wear 1995, 193, 133–145.
12. Wood, P. D.; Datta, P. K.; Burnell-Gray, J. S.; Wood, N.
layers at high temperature,3,20–25 while Fe–Cr-based Mater. Sci. Forum 1997, 251–254, 467–474.
oxides3,23 can also form fairly robust wear protective 13. Wisbey, A.; Ward-Close, C. M. Mater. Sci. Technol. 1997,
layers under adverse sliding conditions. NiO and 13, 349–355.
14. Jiang, J.; Stott, F. H.; Stack, M. M. Wear 1997, 203–204,
Cr2O3 oxides produced from Nimonic 80A do not 615–625.
too readily sinter together to develop into ‘glaze’ and 15. Li, X. Y.; Tandon, K. N. Wear 2000, 245, 148–161.
can actually enhance wear by abrasion3,6,7; however, 16. Jiang, J.; Stott, F. H.; Stack, M. M. Wear 2004, 256,
973–985.
NiO (produced from Nickel 200)3 has been shown to 17. Jiang, J.; Stott, F. H.; Stack, M. M. Tribol. Int. 1998, 31–5,
easily form a ‘glaze’ in the absence of Cr2O3. 245–256.
18. Jiang, J.; Stott, F. H.; Stack, M. M. Tribol. Int. 1997, 30–7,
517–526.
1.15.6 Summary 19. Jiang, J.; Stott, F. H.; Stack, M. M. Wear 1994, 176,
185–194.
20. Datta, S.; Inman, I.; Du, H. L.; Luo, Q. Microscopy of
In writing this chapter strong emphasis has been placed ‘glazed’ layers formed during high temperature wear,
on scientific principles underpinning the phenomenon Invited Talk at the Institute of Materials, Tribology
Meeting, London, November 2001.
of high temperature wear. At the outset a review of 21. Inman, I. A.; Datta, S.; Du, H. L.; Burnell-Gray, J. S.;
some of the well-known and relevant wear theories and Pierzgalski, S.; Luo, Q. Tribol. Int. 2005, 38, 812–823.
models, supported by experimental findings on con- 22. Inman, I. A.; Rose, S. R.; Datta, P. K. Wear 2006, 260,
919–932.
ventional and advanced materials, has been presented. 23. Inman, I. A.; Rose, S. R.; Datta, P. K. Tribol. Int. 2006, 39,
This background information has provided a frame- 1361–1375.
work to discuss new areas of high temperature wear. In 24. Inman, I. A.; Datta, P. S. Wear 2008, 265, 1592–1605.
25. Inman, I. A.; Datta, S.; Du, H. L.; Burnell-Gray, J. S.;
this context the high temperature wear behavior of Pierzgalski, S.; Luo, Q. Wear 2003, 254, 461–467.
those materials which have provided new information 26. Lancaster, J. K. Proc. Royal Soc. London A 1962, 273,
has been considered. Particular attention has been 466–483.
27. Welsh, N. C. Philos. Trans. A 1965, 257, 31–50.
focused on high temperature wear behavior of oxide 28. Welsh, N. C. Philos. Trans. A 1965, 257, 51–70.
dispersion strengthened and Nimonic alloys, and inter- 29. So, H. Tribol. Int. 1996, 25(5), 415–423.
metallic materials involving like-on-like and unlike- 30. So, H. Wear 1996, 192, 78–84.
31. Archard, J. F.; Hirst, W. Proc. Royal Soc. London A 1956,
on-unlike combinations. The most significant part of 236, 397–410.
this chapter includes the exposition of the phenomena 32. Archard, J. F.; Hirst, W. Proc. Royal Soc. London A 1957,
of glaze formation at fundamental levels. Here the 238, 515–528.
33. Quinn, T. F. J. Tribol. Int. 1983, 16, 257–270.
novel aspect is the inclusion of nanoscale elaboration 34. Oh, H.-K.; Yeon, K.-H.; Kim, H. Y. J. Mater. Process.
of the processes of glaze formation underpinned by Technol. 1999, 95, 10–16.
fundamental information gathered from TEM and 35. Stott, F. H.; Wood, G. C. Tribol. Int. 1978, 11, 211–218.
36. Bowden, F. P.; Tabor, D. Friction, An Introduction to
STM/STS investigations. In this area significant reli- Tribology; Heinemann: London, 1973.
ance has been placed on the authors’ own research. 37. Eyre, T. S.; Maynard, D. Wear 1971, 18, 301–310.
38. Burwell, J. T.; Strang, C. D. J. Appl. Phys. 1952, 23, 18–28.
39. Ludema, K. Scuffing, run-in and the function of surface
References films, particularly oxides; Review Paper for
Interdisciplinary Collaboration in Tribology Project,
NASA-Lewis, 1981.
1. Quinn, T. F. J. Wear 1992, 153, 179–200. 40. Tabor, D. Proceedings of the International Conference on
2. Jiang, J.; Stott, F. H.; Stack, M. M. Wear 1995, 181–183, Wear of Materials, St. Louis, Missouri in April 1977;
20–31. ASME, 1978; pp 1–10.
3. Wood, P. D. Ph.D. Thesis, Northumbria University, UK, 41. Rabinowicz, E. C. Friction and Wear of Materials; Wiley:
1997. New York, 1965.
42. Blau, P. J. Wear 1981, 72, 55–66. 78. Shaw, D. J. Colloid and Surface Chemistry, 4th ed.;
43. Iwabuchi, A. Wear 1990, 151, 301–311. Butterworth Heinemann, 1992.
44. Iwabuchi, A.; Kubosawa, H.; Hori, K. Wear 1988, 128, 79. Zhou, Y. H.; Harmelin, M.; Bigot, J. Scr. Metall. 1989, 23,
123–137. 1391–1396.
45. Iwabuchi, A.; Kubosawa, H.; Hori, K. Wear 1990, 139, 80. Datta, P. K.; Du, H. L.; Kuzmann, E.; Inman, I. A. Near
319–333. surface structural changes of ‘glaze’ layers formed during
46. Leheup, E. R.; Pendlebury, R. E. Wear 1991, 142, high temperature sliding wear, to be published in Wear.
351–372. 81. Bhansali, K. J. Wear 1980, 160, 95–110.
47. Colombie, C.; Berthier, Y.; Floquet, A.; Vincent, L. Trans. 82. Razavizadeh, K.; Eyre, T. S. Wear 1982, 79, 325–333.
ASME 1984, 106F, 194–201. 83. Garcia, I.; Ramil, A.; Celis, J. P. Wear 2003, 254,
48. Hiratsukam, K.; Sasada, T.; Norose, S. Wear 1986, 110, 429–440.
251–261. 84. Mølgaard, J.; Srivastava, V. K. Tribol. Int. 1983, 16,
49. Rigney, D. A. Wear 2000, 245, 1–9. 305–314.
50. Stott, F. H.; Glascott, J.; Wood, G. C. Proc. Royal Soc. 85. Subramanium, C. Scr. Metall. 1991, 25, 1369–1374.
London A 1985, 402, 167–186. 86. So, H. Tribol. Int. 1996, 25(5), 415–423.
51. Sullivan, J. L.; Granville, N. W. Tribol. Int. 1984, 17, 87. Buckley, D. H. Cobalt 1968, 38, 20–28.
63–71. 88. Persson, D. H. E. Licentiate of Philosophy Dissertation,
52. Stott, F. H.; Mitchell, D. R. G. In Surface Engineering; Uppsala University, 2003.
Datta, P. K., Gray, J. S., Eds.; Vol. 1: Fundamentals of 89. Persson, D. H. E. Ph.D. Thesis, Uppsala University, 2003.
Coatings; Royal Society of Chemistry, 1993; 90. Stott, F. H.; Stevenson, C. W.; Wood, G. C. Met. Technol.
pp 141–150. 1977, 4, 66–74.
53. Halliday, J. S.; Hirst, W. Proc. Royal Soc. London A 1956, 91. Antony, K. C. J. Met. 1983, 35(2), 52–60.
236, 411–425. 92. Brandes, E. A.; Brook, G. B. Smithells Metals Reference
54. Stott, F. H. Tribol. Int. 2002, 35, 489–495. Book, Butterworth Heinemann, 1992.
55. Lin, D. S.; Stott, F. H.; Wood, G. C.; Wright, K. W.; 93. Inman, I. A. High temperature ‘like-on-like’ sliding of
Allen, J. H. Wear 1973, 24, 261–278. Nimonic 80A under conditions of limited debris retention,
56. Lin, D. S.; Stott, F. H.; Wood, G. C. Trans. ASLE 1973, 17 Northumbria University, 2003, Unpublished work.
(4), 251–262. 94. Vardavoulias, M. Wear 1994, 173, 105–114.
57. Stott, F. H.; Lin, D. S.; Wood, G. C. Proceedings of 95. Wood, P. D. Unpublished work.
the 5th European Congress on Corrosion, 1973; pp 96. Incoloy Alloy MA956; Special Metals Corporation, Publ.
452–455. No. SMC-008.
58. Stott, F. H.; Lin, D. S.; Wood, G. C. Nat. Phys. Sci. 1973, 97. Inman, I. A. The effects of pre-oxidation on the high
242, 75–77. temperature wear of g-TiAl, Internal report, Northumbria
59. Stott, F. H.; Lin, D. S.; Wood, G. C.; Stevenson, C. W. University/University of Birmingham, 2005.
Wear 1976, 36, 147–174. 98. Li, C. X.; Xia, J.; Dong, H. Wear 2006, 261, 693–701.
60. Stott, F. H.; Glascott, J.; Wood, G. C. Wear 1984, 97, 99. Takadoum, J.; Houmid-Bennani, H.; Mairey, D. J. Eur.
155–178. Ceram. Soc. 1998, 18, 553–556.
61. Stott, F. H. Tribol. Int. 1998, 31, 61–71. 100. Barnes, D. J.; Wilson, J. E.; Stott, F. H.; Wood, G. C. Wear
62. Quinn, T. F. J. Wear 1971, 18, 413–419. 1977, 45, 634–640.
63. Quinn, T. F. J. Wear 1996, 199, 169–180. 101. Buckley, D. H. Influence of chemisorbed films on
64. Quinn, T. F. J. Wear 1994, 175, 199–208. adhesion and friction of clean iron, NASA Center for
65. Iwabuchi, A.; Hori, K.; Kudo, H. Proceedings of the Aerospace Information, NASA-TN-D-4775, 1968.
International Conference on Wear of Materials, New York 102. Barnes, D. J.; Stott, F. H.; Wood, G. C. Wear 1977, 45,
1987; pp 211–220. 199–209.
66. Langgath, K.; Kluge, A.; Ryssel, H. Wear 1992, 155, 103. Barnes, D. J.; Wilson, J. E.; Stott, F. H.; Wood, G. C. Wear
343–351. 1977, 45, 97–111.
67. Mølgaard, J. Wear 1976, 40, 277–291. 104. Feng, J.-M.; Uhlig, H. H. J. Appl. Mech. 1954, 21(4),
68. Kuzucu, V.; Ceylan, M.; Çelik, H.; Aksoy, I. J. Mater. 395–400.
Process. Technol. 1997, 69, 257–263. 105. Wright, K. H. R. Proc. Inst. Mech. Eng. 1952–1953, 1B
69. Crook, P.; Li, C. C. Wear of Materials; ASME, 1983; pp (11), 556–574.
272–279, Pub. No. 110254. 106. Waterhouse, R. B. Fretting Corrosion; Pergamon Press:
70. Bartsch, M.; Wasilkowska, A.; Czyrska- New York, 1972.
Filemonowicz, A.; Messerschmidt, U. Mater. Sci. Eng. A 107. Bill, R. C. Wear of Materials; ASME: New York, 1979; pp
1999, 272, 152–162. 356–370.
71. Fujita, A.; Shinohara, M.; Kamada, M.; Yokota, H. Isij Int. 108. Hutchings, I. M. Tribology: Friction and Wear of
1998, 38, 291–299. Engineering Materials; Edward Arnold, 1992.
72. Soda, N.; Sasada, T. ASME Trans. J. Lubr. Technol. 109. Smith, A. F. Tribol. Int. 1986, 19, 65–71.
1978, 100(4), 492–499. 110. Du, H. L.; Datta, P. K.; Inman, I. A.; Kuzmann, E.;
73. Wu, W.-Z.; Xing, J.-D.; Su, J.-Y. Wear 1997, 210, Suvegh, K.; Marek, T.; Vértes, A. Tribol. Lett. 2005, 18(3),
299–303. 393–402.
74. Suh, N. P.; Sin, H. C. Wear 1981, 69, 91–114. 111. Chattopadhyay, B.; Wood, G. C. Oxidat. Met. 1970, 2,
75. Rice, S. L.; Moslehy, F. A.; Zhang, J. In Proceedings of 373–399.
the 18th Leeds-Lyon Symposium on Tribology, 3–6 112. Stott, F. H. Tribol. Int. 2002, 35, 489–495.
September 1991, Lyon, France; Dowson, D., Ed.; 113. Du, H. L.; Datta, P. K.; Inman, I.; Geurts, R.; Kübel, C.
Elsevier, 1991; pp 463–467. Mater. Sci. Eng. A 2003, 357, 412–422.
76. Hesmat, H. Tribol. Trans. 1991, 34(3), 433–439. 114. Inman, I. A. Nano-scale studies of ‘glaze’ layers
77. Ludema, K. C. In Proceedings of the 18th Leeds-Lyon formed on Incoloy MA956 when slid against Incoloy
Symposium on Tribology, 3–6 September 1991, Lyon, 800HT at 750 C, Northumbria University, 2003,
France; Dowson, D., Ed.; Elsevier, 1991; pp 155–160. Unpublished work.
115. Du, H. L.; Aljarany, A.; Datta, P. K.; Burnell-Gray, J. S. Deformation; Lowe, T. C., Valiev, R. Z., Eds.; Kluwer
Corros. Sci. 2005, 47, 1706–1723. Academic, 2000; pp 231–240.
116. Aljarany, A. A. Ph.D. Thesis University of Northumbria, 124. Lim, S. C. Tribol. Int. 1998, 31(1–3), 87–97.
UK, 2002. 125. Lim, S. C. Tribol. Int. 2002, 35(11), 717–723.
117. Gleiter, H. Prog. Mater. Sci. 1989, 33, 223–315. 126. Childs, T. H. C. Tribol. Int. 1980, 13, 285–293.
118. Valiev, R. Z.; Islamgalier, R. K.; Alexandrov, I. V. Prog. 127. Kato, K.; Hokkirigawa, K. In Proceedings
Mater. Sci. 2000, 45, 103–189. of the Eurotrib’85; Elsevier: Amsterdam, 1985; Vol. 4,
119. Lowe, T. C.; Valiev, R. Z. JOM 2000, 52, 27–28. Section 5.3, pp 1–5.
120. Ghosh, A. K.; Huang, W. In Investigations and 128. Riahi, A. R.; Alpas, A. T. Wear 2003, 255, 401–409.
Applications of Severe Plastic Deformation; 129. Chen, H.; Alpas, A. T. Wear 2000, 246, 106–116.
Lowe, T. C., Valiev, R. Z., Eds.; Kluwer Academic, 130. Yang, S. H.; Kong, H.; Yoon, E.-S.; Kim, D. E. Wear 2003,
2000; pp 29–36. 255, 883–892.
121. Mishra, R. S.; McFadden, S. X.; Mukherjee, A. K. Mater. 131. Grimanelis, D.; Eyre, T. S. Wear 2007, 262, 93–103.
Sci. Forum 1999, 304–306, 31–38. 132. Grimanelis, D.; Eyre, T. S. Surf. Coat. Technol. 2006,
122. Mishra, R. S.; Mukherjee, A. K. In Superplasticity and 201–6, 3260–3268.
Superplastic Forming; Ghosh, A. K., Bieler, T. R., Eds.; 133. Elleuch, K.; Elleuch, R.; Mnif, R.; Fridrici, V.; Kapsa, P.
TMS Warrendale, 1998; pp 109–116. Tribol. Int. 2006, 39–4, 290–296.
123. Mishra, R. S.; McFadden, S. X.; Mukherjee, A. K. In 134. Adachi, K.; Kato, K.; Chen, N. Wear 1997, 203–204,
Investigations and Applications of Severe Plastic 291–301.
1.16 Types of Environments
B. A. Baker
Special Metals Corporation, 3200 Riverside Drive, Huntington, WV 25705, USA

1.16.2 Oxidizing Environments 400
1.16.3 Nitridation 400
1.16.4 Environments Promoting Carbon Ingress 401
1.16.5 Environments Containing Halides 402
1.16.6 Sulfur-Containing Environments 403
1.16.6.1 Coal-Fired Boiler Corrosion 404
1.16.6.2 Oil-Fired Boiler Corrosion 404
1.16.7 Molten Metals 405
1.16.8 Molten Salts 405
References 405
molten salts are analogous to aqueous corrosion pro-

Abbreviations cesses in that an electrolyte is present; galvanic effects
AHF Anhydrous hydrofluoric acid
may thus be operable. Lower-melting metals and alloys
LME Liquid metal embrittlement
are another source of possible corrosion.
MSW Municipal solid waste
Laboratory studies have contributed significantly
over the years to deepen the knowledge base of high
temperature corrosion processes. Prediction of cor-
1.16.1 Introduction rosion rates for many high temperature corrosion
phenomena may be performed based upon the model-
High temperature corrosion in gaseous environments ing of the kinetics of governing reactions; predictive
is commonly interpreted as that taking place above software tools have been developed which provide this
the temperature at which corrosive liquids may con- function for a wide range of high temperature alloys.2
dense, causing dew point corrosion. While most high Extraneous factors, which may be difficult to sim-
temperature corrosion processes typically take place ulate in the laboratory, often account for the difficulty
at quite high temperatures, above about 450 C, such in mimicking commercial processes. Such factors
phenomena can take place at even lower tempera- include, but are not limited to, pressure, gas velocity,
tures. For example, formation of pyrosulfates of impurities, and mechanical and thermal stress.
potassium and sodium can contribute significantly Pressure: The effect of pressure upon the mobility
to corrosion rate in coal-fired boiler furnace walls of reacting ions through a surface layer such as an
within the temperature range of 250–350 C.1 oxide is not pronounced. However, if the dissociation
While interaction with gaseous environments pressure of a corrosion product is within the given
comprises much of the breadth of knowledge of range of conditions, then pressure may exert a pro-
high temperature corrosion, interdiffusion can occur found effect. The effect of temperature is far more
between high temperature materials and other solids, prominent, typically.
for example, between a silicon carbide furnace hearth Gas velocity: Very high velocity gas streams may
and a furnace belt carrying product through the promote erosion; this, in conjunction with corrosive
furnace, or between a graphite sample holder and species which may be present, can result in accelerated
metallic furnace components. Contact with molten wastage rates via an erosion–corrosion mechanism.
materials may occur, such as interaction between alkali Conversely, exposure of materials high in molybdenum
salts of sulfur or halides and metal alloys, which can or vanadium to stagnant high temperature oxidizing
result in the formation of lower-melting eutectics. Pro- atmospheres can lead to what is termed ‘catastrophic
cesses governing corrosion resulting from contact with oxidation’ – molybdenum trioxide, MoO3, melts
399
400 High Temperature Corrosion Environments
at 795 C and vanadium pentoxide, V2O5, melts at pCO2 and pH2/pH2O ratios. In such ‘reducing’ com-
690 C. If adequate gas flow is not provided, these bustion atmospheres, oxidation, while occurring con-
oxides do not volatilize and serve to promote addi- currently, may not often be the predominant mode of
tional oxidation.3,4 corrosion. For example, if the sulfur level in the
Impurities: Simulation of industrial corrosion combusted fuel is sufficiently high, sulfidation could
environments is frequently very difficult because of become the principal mode of corrosion.
the improbability of pinpointing the concentration of Oxidation in pure steam is a significant consider-
various low-level impurities. This often stems from a ation in the specification of materials for steam
lack of knowledge about the specific nature of the superheater tubes and steam header pipes in power
industrial environment. generating plants. Recent advances in power plant
Stress: Stress-accelerated grain boundary oxi- design have targeted very high steam temperatures
dation has been found to contribute to cracking up to 760 C and increased pressure levels in boost-
phenomena in high temperature materials. For exam- ing the efficiency of future plants. Oxidation rates of
ple, Soontrapa5 documented a case occurring in an age- austenitic stainless steels and even nickel-based alloys
hardenable Fe–Ni material. Interaction between this are concerns under these conditions.12,13
mechanism and stress-corrosion cracking in simulated Some industrial processes utilize pure oxygen at
primary water at 360 C, mimicking conditions found elevated temperatures; for example, conversion of
in a pressurized water reactor, has been reported by titanium tetrachloride to titanium oxide. Pure oxygen
Panter6; cracking in austenitic stainless steels at 610 C rather than air can be utilized for combustion of coal
has also been attributed to this mechanism.7 and other fuels (oxy-fuel combustion) in the interest
Thermal cycling can produce stresses sufficient to of reducing NOx levels in emissions, and producing a
cause cracking and buckling of corrosion layers, pos- purified CO2 stream for enhanced utilization or dis-
sibly leading to eventual spalling of the corrosion posal.14 Reforming processes, which involve combin-
layer completely from the surface. The tendency for ing a hydrocarbon feedstock (typically methane) with
this to occur can depend upon the differential expan- steam to produce syngas and ultimately hydrogen,
sion coefficients between the oxide and substrate as can also utilize oxygen in place of a fraction of sup-
well as the strength of the oxide itself versus that of plied steam.
the substrate. Subsurface oxides can serve to provide
mechanical stability during thermal cycling as well.
1.16.3 Nitridation
1.16.2 Oxidizing Environments Nitridation can give concurrent results with oxida-
tion in air, or in combustion environments employing
Oxidation is certainly the most important mode of air as the oxidizing gas. This is predominant in mate-
high temperature corrosion. While certain processes rials having additional elements such as aluminum,
do involve simple oxidation by air, such as the exter- titanium, and niobium – nitrides formed are typically
nal surfaces of electrically heated components, it is internal – for which nitride formation is particularly
more common that oxidizing species are accompa- thermodynamically favorable. Such nitridation can
nied by other reacting components from the atmo- occur under static conditions, but is more prevalent
sphere. The combustion of natural gas is one very upon exposure to dynamic loading conditions. Inter-
common example, where typically a percentage of nal nitrides are often found to be associated with the
excess air is employed, producing a gas which con- tips of creep cracks formed in air or combustion
tains primarily O2, N2, H2O, CO2, and CO. While the environments.15 In nickel-based alloys, the diffusivity
majority of the oxidizing potential is provided by O2 of nitrogen is determined to be nearly two orders
in such an atmosphere, other species may have very of magnitude greater than that of oxygen.16 This
meaningful effects upon observed oxidation behavior. explains the observation of internal nitrides extend-
For example, water vapor is known to have a signifi- ing to greater depths than internal oxides after expo-
cant effect upon ferritic and austenitic stainless steel sure to such environments.
oxidation behavior, in the formation and volatiliza- Nitrogen or cracked ammonia is often used as a
tion of CrO2(OH)2.8–11 In cases where combustion is protective atmosphere during the heat treatment of
carried out with a substoichiometric air-to-fuel ratio, stainless steels. Sintering of powder metal compo-
the oxygen pressure becomes a function of the pCO/ nents is performed in a nitrogen-based atmosphere.
Purposeful nitridation of the surface of metallic com- Fe–Ni–Cr materials, often having chromium levels at
ponents is frequently carried out to impart strength or above 25% and silicon additions, are commonly
and wear resistance. Furnace components exposed used as radiant tubes in ethylene-cracking furnaces.
repeatedly to the same nitriding atmosphere must Under normal operating conditions, chromium oxide
possess good resistance. Typically, in nitrogen and protects the tubes against rapid carburization in the
ammonia-based environments, alloys high in nickel carburizing–oxidizing environment at the ID of the
content (having low nitrogen solubility and diffusiv- tubes. At very high temperatures, at and above
ity) and low in elements which are strong nitride 1100 C where chromium oxide is not stable, silicon
formers (aluminum, titanium, niobium, chromium) oxide can afford protection, as shown by Nishiyama.18
are favored for sustained service. Uncracked ammo- Coke formation also occurs in the ethylene cracking
nia is a particularly harsh nitriding environment. environment. Laboratory testing to truly simulate
Interaction of the uncracked ammonia with the such an environment would thus need to incorporate
metal surface results in direct diffusion of nascent an atmosphere for which the carbon activity exceeds
nitrogen. Ammonia, which has been dissociated into unity. Production of carbon fiber products also typi-
H2 and N2, possesses less nitriding potential com- cally involves exposure to solid carbon deposits at
pared to pure nitrogen. elevated temperatures.
A mixture of nitrogen and water vapor is utilized Steam reformers, converting natural gas or other
in processes employing the Kalina Cycle. Corrosion hydrocarbon feed stocks to a gas rich in hydrogen and
effects in such atmospheres have been studied by carbon monoxide, can promote carburization if tem-
Grabke.17 Interestingly, the performance of austenitic peratures exceed the normal 700–850 C, or if carbon
stainless steels was found to supersede the perfor- activity becomes excessive in the normally oxidizing
mance of nickel-based materials in this specific gas atmosphere.19
environment. Gas carburizing of steels takes place in an endo-
thermic combustion environment with added hydro-
carbons. Such atmospheres typically have a carbon
1.16.4 Environments Promoting activity below unity. Carburization of furnace com-
Carbon Ingress ponents such as supporting grates, fans, and radiant
furnace tubes is often observed; heat treatment bas-
Carburization, a high temperature form of corrosion kets and belts are also affected.
promoted by carbon ingress and subsequent internal Another form of carburization, termed ‘green rot’
carbide precipitation, may proceed via one of the involves alternating exposure of a nickel-based mate-
following reactions resulting in carbon formation: rial having an intermediate chromium level (first
encountered with 80% Ni–20% Cr heating element
2CO ¼ CO2 þ C ½1
materials) to a nonoxidizing (reducing) carburizing
atmosphere followed by an environment containing
CO þ H2 ¼ H2 O þ C ½2
sufficient oxygen to oxidize chromium. Internal car-
bides formed during the reducing cycle are subse-
CH4 ¼ 2H2 þ C ½3
quently preferentially oxidized upon exposure to the
High temperature carburization in hydrocarbon- oxidizing environment. This phenomenon has been
based environments is encountered commonly in combated by the use of nickel-based materials having
heat treatment processes. Here again, materials hav- higher Cr levels, or niobium addition.
ing high nickel content (low carbon solubility and Carbon dioxide cooled nuclear reactors, which
diffusivity) and low levels of strong carbide formers operate at temperatures lower than the helium-
(chromium, titanium, niobium, tungsten, molybde- cooled variant, can experience carburization of com-
num) are commonly favored. High levels of alumi- ponents, despite the fact that the CO2 environment is
num or silicon can enhance the performance in such oxidizing and has low carbon activity. This occurs at
environments as well, where even very low levels of the oxide scale–metal interface if CO2 penetrates the
oxygen may allow the formation of a protective scale. scale.19 Numerous reactors employing CO2 cooling
Environments that are concurrently oxidizing are in operation in the United Kingdom.
and carburizing allow for a wider range of material Carburization can occur in coal combustion and
usage, if the oxygen pressure is high enough to pro- coal gasification processes, despite the fact that the
mote the formation of chromium oxide. For example, prevailing form of attack is typically sulfidation, often
via contact with incompletely combusted carbona- Municipal solid waste (MSW) combusted to pro-
ceous material at elevated temperatures. duce electrical power typically contains sufficient
Metal dusting is a form of carburization which chloride-containing plastic materials (primarily PVC)
occurs when carbon activity exceeds unity at in- to promote a severe form of corrosion on the fireside
termediate temperatures, 400–800 C. The most wall of waterwall and superheater tubing. Metal tem-
frequently encountered occurrence of this phenome- peratures range typically from 400 to 650 C. Deposi-
non is seen in processes which form syngas via the tion of alkali chlorides (often containing heavy metals
combination of steam and hydrocarbon feedstock such as lead and zinc), which may be molten or partially
(reforming) at high temperature (750–850 C). The molten, onto metal surfaces results in the formation of
resulting syngas mixture contains H2, CO, CO2, and metal chlorides. In addition, the flue gas contains HCl
H2O as well as residual CH4 and other hydrocarbons. vapor. The HCl vapor can react with oxygen in the
Industrial processes involving syngas production oxidizing combustion environment to form chlorine
include production of hydrogen, fertilizers, and syngas, according to the following equation:
thetic fuels as well as gasification of various feed- 1
stocks including coal and biomass. Typically, higher 2HCl þ O2 ¼ Cl2 þ H2 O ½4
2
ratios of pCO/pCO2 and pH2/pH2O produce higher
calculated carbon activity and consequently higher pro- Volatilized metal chlorides, formed via interaction with
pensity for metal dusting. Carbon activity, by virtue of the generated chlorine, may react with oxygen in the
reactions [1] and [2], rises as temperature is lowered; flue gas to form additional chlorine, creating a self-
at some point as temperature drops, sluggish kinetics sustaining mechanism:
result in decreased metal dusting activity. Nickel- FeCl2 þ 2 CrCl2 þ 2O2 ¼ FeCr2 O4 þ 3Cl2 ½5
based alloys having high levels of oxide scale forming
elements, chromium and aluminum, and low iron Vapor pressures for various metal chlorides as a func-
content tend to exhibit the most favorable perfor- tion of temperature are shown in Figure 1.22
mance in these environments. Environments with Typical MSW combustion gas also contains SO2;
very high CO levels encountered in iron ore reduc- residual levels of HCl in the gas commonly range from
tion promote metal dusting as well. Metal dusting 400 to 600 ppm, whereas SO2 levels are typically in the
corrosion is also encountered occasionally in the range of 100–200 ppm. It is generally agreed that the
heat treating industry. Petroleum refineries experi- mechanism of corrosion in MSW boilers is related
ence metal dusting in hydrodealkylation processes primarily to the presence of chlorine-containing spe-
and catalyst regeneration systems. Nuclear plants cies. To combat this form of corrosion, metallic coat-
employing CO2 for cooling can experience metal ings are typically applied via weld overlay or thermal
dusting as well. Metal dusting can occur in molten spray onto the fireside tube surface. These materials are
carbonate and solid oxide fuel cells. Exposure in typically Ni–Cr–Mo alloy systems. At temperatures
hydrocarbon-based environments can also result in above 800 C in oxidizing–chloridizing atmospheres,
metal dusting attack.20 the high vapor pressure of molybdenum oxychlorides
Incorporation of even ppm levels of H2S into makes molybdenum-containing materials a less desir-
environments, known to promote metal dusting, has able choice.23 At such high temperatures, Ni–Cr–Al
been found to suppress the phenomenon.20,21 materials typically exhibit better performance.
Combustion of biomass in boilers can have a
mechanism of corrosion in waterwalls and superhea-
1.16.5 Environments Containing ters similar to that seen in MSW boilers. Wood and
Halides straw products, for example, contain significant levels
of potassium and chlorine, producing KCl in the gas
Chlorine and hydrogen chloride are found in numer- phase. Condensation on cooler metal surfaces from
ous industrial processes. Concentrations, which may the gas phase in turn can create more complex alkali
have an impact upon high temperature corrosion salts with alloying elements from the metal substrate.
behavior, range from ppm levels to pure gas streams. As in MSW boilers, metallic coatings are typically
Other gaseous components can have an effect on applied by weld overlay, thermal spray, etc. Materials
material performance; environments which contain utilized to combat the intermediate-temperature
oxygen and/or water vapor can accelerate corrosion form of chloride salt attack are also typically Ni–Cr–Mo
behavior for some materials. alloy systems.
100
FeCl2
10−1
NiCl2
FeCl3
NaCl
10−2
Vapor pressure (atm)
ZnCl2 CrCl2
10−3
KCl
PbCl2
10−4
CrCl3
10−5
10−6
0 200 400 600 800 1000 1200 1400
Temperature (⬚C)
Figure 1 Vapor pressure versus temperature for metal chlorides.
Production of ethylene dichloride, an intermedi- Bromine and iodine have similar effects, though
ate of vinyl chloride monomer, involves creation of less severe than those of chlorine or fluorine.
chlorine-bearing high temperature environments.
Production of titanium and silicon dioxides
involves oxidation of a chloride precursor at high 1.16.6 Sulfur-Containing
temperature. Environments
Formation of metal chlorides from ores is a com-
mon step in metals production, capitalizing upon the Sulfidation occurs in a variety of forms and envi-
very high vapor pressure of many metal chlorides. ronments, and can result from a number of differ-
Chemical vapor deposition processes may incor- ent sulfur-containing species. Exposure to H2–H2S
porate the utilization of high temperature HCl vapor. gas mixtures in refining environments is a classical
Certain coals can contain low levels of chlorine, form of sulfidation. Exposure of nickel-containing
which can exacerbate coal ash corrosion effects in materials in such an environment above the Ni–NiS
boiler tubing and superheater tubing.24 Chloride layers eutectic melting temperature (645 C) can have
have been discovered at the scale-to-metal interface in catastrophic results. However, at lower temperatures,
exposed tubes. Even a very modest quantity of chlorine nickel-based materials can provide very good perfor-
at high temperatures can evoke severe corrosion effects mance. Attack rate is dependent simply on the partial
because of the previously described self-perpetuating pressure of sulfur, determined by the pH2/pH2S ratio.
mechanism. These conditions are encountered in petroleum refin-
Fluorine can promote very severe forms of corro- ing in hydrotreating, hydrocracking, and hydrodesul-
sion as well. Anhydrous hydrofluoric acid (AHF) is furising processes.
used in the alkylation catalysis for gasoline manufac- Environments containing significant partial press-
ture and for the manufacture of chlorofluorocarbon ures of both sulfur and oxygen are termed sulfidizing-
refrigerants and propellants. Uranium oxide is fluori- oxidizing; this is typical of most sulfur-containing
nated at high temperatures with hydrofluoric acid environments. Corrosion behavior is dictated by both
to produce HF6. Centrifuging then typically follows the oxygen and sulfur partial pressures. Again, in general,
for the separation of HF6 containing U235 from high nickel alloys are not desired for such environ-
that incorporating U238. Production of modern re- ments; however, with sufficient levels of scale-forming
frigerants involves elevated-temperature exposure to elements such as chromium or aluminum, protection
chlorine- and fluorine-containing environments. can be afforded. An example of such an environment
would be that produced by the combustion of natural Steam superheaters and reheaters in coal-fired
gas containing significant levels of H2S. Commercial boilers are subject to severe coal-ash corrosion pro-
food preparation operations involving meat products blems as well, because of deposition of combustion
can produce sulfidation–oxidation in oven components ash containing sulfate salts. Temperatures are signifi-
by virtue of the sulfur present in the disulfide bonds of cantly higher than that at the waterwall surface. Such
proteins. Oil-fired boilers commonly burn low-grade conditions often necessitate not only high-strength
fuel containing significant sulfur levels. austenitic materials for tubing, but also consideration
of protective claddings. Nickel–chromium materials
have been found to perform well; tubes clad with a
1.16.6.1 Coal-Fired Boiler Corrosion
50Ni–50Cr material are often utilized.
The sulfur content of coal used for energy generation Recent interest in developing the technology re-
can be as high as 5% or more, producing combustion quired for the construction and operation of coal-fired
gases containing H2S, SO2, and SO3 and resulting in plants having steam temperatures of up to 760 C has
the deposition of flue ash containing alkali sulfate necessitated consideration of a different class of mate-
salts onto metal surfaces cooler than the gas (furnace rials. Elevating the operating temperature of the water-
walls, superheaters and reheaters). These salts may wall and superheaters will have an effect upon coal-ash
react with SO2 or SO3 to form pyrosulfates or more corrosion dynamics. One recent study by McDonald
complex alkali–iron trisulfates. Low-melting combi- involved insertion of a test loop into the low-NOx
nations of the various species may be formed. For boiler burning Ohio coal containing 3% sulfur, in
example, potassium sulfate (K2SO4) and potassium the interest of simulating conditions in a superheater
pyrosulfate (K2S2O7) form a molten salt mixture at operating at temperatures higher than conventional; at
407 C when the SO3 concentration is 150 ppm. The 1050–1210 F.25 Results indicated that the conditions
mode of sulfidation promoted by deposits containing were harsh for many materials and that, tubes weld-
such salt mixtures is commonly termed ‘coal-ash overlaid or clad with Ni–Cr materials having very high
corrosion.’ Cr levels exhibited the most favorable performance.
While the sulfidizing–oxidizing environments
noted so far are typically strongly oxidizing in nature,
1.16.6.2 Oil-Fired Boiler Corrosion
certain environments exhibit a higher ratio of sulfur
pressure to oxygen pressure. These types of environ- Fuel oils consisting of residuals from petroleum refin-
ments are generated by combustion under significan- ing cause oil-ash corrosion in boilers. Compounds
tly substoichiometric conditions. Typically, oxides formed from residual vanadium, sulfur, and sodium
afford protection to the point where breakaway cor- account for this mode of corrosion. Vanadium forms
rosion initiates, followed by significant sulfidation V2O5; similarly, sodium forms sodium oxide (Na2O) or
attack. Coal gasification processes typify this type of sodium sulfate (Na2SO4). These compounds can com-
environment. Low-NOx burners in coal combustion bine to form fused salts that can have very low melting
operate under conditions which produce a reducing temperatures.26 The resulting molten vanadate com-
environment in the vicinity of the burners, with pounds cause fluxing of otherwise protective oxide
metal temperatures possibly as high as 600 C. This scales, promoting very aggressive corrosion attack
can result in the presence of H2S rather than SO2 in and metal wastage. The concentration of SO3 in the
the combustion gas. High wastage rates of waterwall flue gas exerts significant influence upon the stability
tubes in such boilers can result. Weld overlays are of the fused vanadate salts, as discussed by Seiersten
applied to combat corrosion; Ni–Cr–Mo welding and Kofstad.27 As in coal-fired boilers, Ni–Cr overlay
products are commonly utilized. Thermal spray coat- coatings are commonly applied over the ferritic or
ings and chromium/chromium–silicon coatings are austenitic steels used in steam superheaters and rehea-
also often utilized. ters in the interest of optimizing corrosion resistance.
Blowing deposits from superheaters and reheaters The mode of corrosion resulting from exposure to
using steam soot blowers is commonly employed, deposited sulfate is termed ‘Type I’ hot corrosion in
which promotes a form of erosion–corrosion. Water gas turbines, whereby sulfur-containing salts become
lances or water canons may be used to remove slag molten, resulting in dissolution of protective oxides,
deposits from waterwalls via thermal shock. This can and resultant ingress of sulfur and formation of inter-
contribute to circumferential cracking by a mecha- nal sulfides. Type I hot corrosion occurs typically in
nism involving corrosion-assisted thermal fatigue. the temperature range of 850–950 C, and primarily
in aerospace applications. Coatings are normally used in nuclear and solar energy systems as a medium for
for protection of metallic components against this heat transfer and storage because of high thermal
mode of corrosion; for example nickel- and cobalt- conductivity and heat capacity. A mixture of sodium
based MCrAlY systems have been used with success. and potassium nitrates has been used for energy stor-
Cobalt-based coatings offer particularly good resistance age in solar collectors.
to Type I hot corrosion. Type II hot corrosion, also Extraction of aluminum is performed by dissol-
resulting from sulfate salt deposition, typically occurs at ving alumina in cryolite. Extraction of various reac-
temperatures below 800 C. This form of hot corrosion tive metals, such as titanium, magnesium, sodium and
results from localized formation of low-melting eutec- others is accomplished using molten salts, often
tics, resulting in localized pitting with the absence of chlorides.
gross sulfide formation, and is more commonly encoun- Molten salt batteries use molten salts as an elec-
tered in land-based and marine gas turbine applications. trolyte. They are used in services where very high
Typically, cobalt-based coatings are not as resistant to power density is required. These characteristics make
Type II hot corrosion as nickel-based coatings. such batteries a promising technology for powering
electric vehicles. Operating temperatures as high as
700 C, however, pose significant challenges.
1.16.7 Molten Metals
Molten carbonate fuel cells are currently being
developed for natural gas and coal-based power
Liquid metals are often considered for use as heat
plant applications, as well as other industrial and
transfer media, because of their attractive physical
military applications. These fuel cells do not require
properties – high thermal conductivity and heat capac-
an external reformer to produce hydrogen. Instead,
ity. Also, their operating pressures are very low com-
fuels are converted to hydrogen within the fuel cell
pared to water. Lead, lead–bismuth, and sodium have
itself.
all been considered as coolants for nuclear reactors.
Molten salts are used frequently in the heat treat-
Sodium metal is used to fill hollow high-performance
ing industry for case hardening, annealing and stress
engine valve stems, in the interest of transmitting heat
relieving, quenching and tempering, as well as clean-
along the valve stem and to the cooling jacket. The
ing. As corrosion rate in molten salts is typically a
properties of molten potassium make it suitable as the
form of oxidation, attack is commonly exacerbated at
working fluid for a two-phase Rankine cycle, enabling
the salt-to-air interface.
operation at temperatures higher than that for typical
In the chemical industry, molten salts are used for
steam-based systems.28
cracking and catalysis, curing and vulcanizing plas-
Solar Stirling engines have employed the use of
tics, synthesis of organics and gases, and in pyrolysis
sodium as the working fluid as well. The range of heat
of scrap and hazardous materials.
capacities and boiling points for the various metals
used for heat transfer applications serve to dictate the
range for the use. For example, lead is useful up to
500 C, sodium 700 C, and lithium 1300 C.29 References
Molten aluminum is an extremely harsh environ-
ment for all types of metals and alloys. 1. Reid, W. T. External Corrosion and Deposits; American
Molten zinc is commonly used in the hot-dip gal- Elsevier: New York, 1971.
2. John, R. C., et al. Mater. Sci. Forum 2004, 461–464,
vanizing process for coating steel for corrosion resis- 599–610.
tance enhancement. Galvanizing tanks and handling 3. Meijering, J. K.; Rathenau, G. W. Nature 1950, 165, 240.
accessories require resistance. 4. Brennor, S. S. J. Electrochem. Soc. 1955, 102(1), 16.
5. Soontrapa, C. Master’s Thesis, Massachusetts Institute of
When under stress, metals may become suscepti- Technology, 2005.
ble to liquid metal embrittlement (LME). Instances of 6. Panter, J.; Viguer, B.; Cloué, J. M.; Foucault, M.;
LME are often associated with welding, brazing, or Combrade, P.; Andrieu, E. J. Nucl. Mater. 2006, 348(1–2),
213–221.
soldering. 7. Le Calvar, M.; Scott, P. M.; Magnin, T.; Rieux, P. Corrosion
1998, 54(2), 101–105.
8. Asteman, H.; Svensson, J. E.; Johansson, L. G.; Norell, M.
1.16.8 Molten Salts Oxid. Met. 2000, 54(1–2), 11–26.
9. Young, D.; Pint, B. Oxid. Met. 2006, 66(3–4), 137–153.
10. Peraldi, R.; Pint, B. A. Oxid. Met. 2004, 61(5–6), 463–483.
Molten-salt technology is critical for several indus- 11. Rakowski, J. M. J. Eng. Gas Turbines Power 2004, 126(4),
tries.30 As with molten metals, molten salts are utilized 768–873.
12. Program on Technology Innovation. Oxide Growth and 23. Oh, J. M.; McNallan, M. J.; Lai, G. Y. Metall. Trans. A 1986,
Exfoliation on Alloys Exposed to Steam EPRI, Palo Alto, 17A, 1087–1094.
CA, 2007; 1013666. 24. Effect of Coal Chlorine on Waterwall Wastage in Coal-
13. Holcomb, G. R. Oxid. Met. 2008, 69(3–4), 163–180. Fired Boilers with Staged Low NOx Combustion Systems,
14. Natesan, K.; Rink, D. L. In Proceedings of the 21st Annual EPRI, Palo Alto, CA; American Corporation, St. Louis, MO;
Conference on Fossil Energy Materials, Oak Ridge Dairyland Power Cooperative, La Crosse, WI, 2002,
National Laboratory/U.S. Department of Energy, April 30 1004082.
to May 2, 2007. 25. McDonald, D. K. Coal Ash Corrosion Resistant Materials
15. Welker, M.; Rahmel, A.; Schütze, M. Metall. Trans. A 1989, Testing Program Evaluation of the First Section Removed
20A, 1541–1551. in November 2001, The Babcock and Wilcox Company,
16. Douglass, D. L. JOM; 1991, 74. submitted to DOE (DE-FC26-99FT40525) and OCDO
17. Grabke, H. J.; Strauss, S.; Vogel, D. Mater. Corros. 2003, (CDO/D-98-2).
54(11), 895. 26. Reid, W. T. External Corrosion Deposits in Boilers and Gas
18. Nishiyama, Y.; Otsuka, N.; Nishizawa, T. Corrosion 2003, Turbines; Elsevier: Amsterdam, 1971.
59(8), 689–700. 27. Siersten, M., Kofstad, P. High Temp. Technol. 1987, 5(3),
19. Grabke, H. J. Carburisation – A High Temperature 115.
Corrosion Phenomenon; Materials Technology Institute of 28. Radovanovic, M. Fluidized Bed Combustion; Taylor &
the Chemical Process Industries: St. Louis, MO, 1998. Francis: London, 1986; 267.
20. Grabke, H. J. In European Corrosion Congress, 29. Baboian, R. Corrosion Tests and Standards: Application
Eurocorr’99; Aachen, September 1999. and Interpretation, 2nd ed.; ASTM International: West
21. Grabke, H. J.; Müller-Lorenz, A.; Schneider, A. ISIJ Int. Conshohocken, PA, 2005.
2001, 41(Supplement), S1–S8. 30. Lovering, D. G. Ed. Molten Salt Technology; Plenum
22. Brandes, E. A.; Brook, G. B. Eds. Smithells Metals Press: New York, 1982; 1.
Reference Book; Butterworth-Heinemann: Oxford, 1992.
1.17 Oxidation in Steam and Steam/Hydrogen Environments
W. J. Quadakkers and J. Żurek
Forschungszentrum Jülich, Institute of Energy Research (IEF-2), 52428 Jülich, Germany

1.17.2 Thermodynamic Considerations 409
1.17.2.1 Thermodynamics of the Gas Atmosphere and Stability of Solid Oxides 409
1.17.2.2 Volatile Reaction Products in H2O-Containing Gases 409
1.17.2.3 Considerations for Material Testing in Steam Environments 411
1.17.3 Growth Mechanisms of Oxide Scales 413
1.17.3.1 Kinetics of Surface Reactions 413
1.17.3.2 Effect of Water Vapor and Hydrogen on Lattice Diffusion in Oxide Scales 415
1.17.3.3 Molecular Diffusion in Oxide Scales in Wet Gases 415
1.17.4 Metallic High Temperature Components in Steam and Steam/Hydrogen
Environments 416
1.17.4.1 General Remarks 416
1.17.4.2 Oxidation of Chromium and High-Chromium Alloys 418
1.17.4.2.1 Oxidation of chromium 418
1.17.4.2.2 Oxidation of chromia-forming Fe- and Ni-based alloys 418
1.17.4.2.3 Mechanisms of chromia scale growth 419
1.17.4.2.4 Commercial chromia-forming Fe- and Ni-based alloys 422
1.17.4.2.5 Effect of common minor alloying additions 423
1.17.4.3 Effect of Environment on Scale Formation in Case of ‘Borderline Alloys’ 426
1.17.4.3.1 Nonprotective oxidation in water vapor 426
1.17.4.3.2 Internal oxidation of Cr as trigger for breakaway oxidation 427
1.17.4.3.3 Conditions for external chromia formation in dry and wet gases 427
1.17.4.3.4 Breakaway oxidation mechanisms of FeCr alloys in water vapor 430
1.17.4.3.5 Borderline NiCr alloys in water vapor 430
1.17.4.4 Oxidation of Ferritic and Austenitic Steels in Steam 431
1.17.4.4.1 General remarks 431
1.17.4.4.2 Steam oxidation mechanisms of 9–12% Cr steels 433
1.17.4.4.3 Void and gap formation 435
1.17.4.4.4 Long-term behavior 436
1.17.4.4.5 Temperature dependence of oxidation during long-term exposure 440
1.17.4.4.6 Austenitic steels 445
1.17.4.4.7 Effect of pressure on steam oxidation 449
1.17.4.5 Alumina-Forming Alloys and Coatings 449
1.17.4.5.1 a-Alumina scales 449
1.17.4.5.2 Metastable alumina 451
1.17.4.5.3 Borderline alumina forming alloys 452
References 454
REs Reactive elements

Abbreviations
SEM Scanning electron microscopy
BC Bond coat
SNMS Sputtered neutrals mass spectrometry
EDX Energy dispersive X-ray spectroscopy
SOFC Solid oxide fuel cell
IEP Isoelectric point
TBC Thermal barrier coating
ODS Oxide dispersion strengthened
PSEs Protective scale forming elements
407
XPS X-RAY photoelectron spectroscopy environment as the main oxidizing gas species or in
XRD X-Ray diffraction addition to larger amounts of ‘free oxygen.’ From the
viewpoint of equilibrium oxygen partial pressure and
oxygen/water vapor ratio, the service environments
may be divided into four groups:
Symbols
A. Wet air or gases originating, for example, from
h Positive hole
combustion processes with a high oxygen (air) to
kp(Kp) Parabolic rate constant (mg2 cm4 s1,
fuel ratio. In such environments the oxygen con-
mm2 h1)
tent may be as high as 10–20%.2
n Oxidation rate exponent
B. Combustion or exhaust gases originating from
p Vapor pressure (bar, Pa)
combustion processes with low oxygen (air) to
pi Partial pressure of component (bar, Pa)
fuel ratio. In such gases, which, for example, pre-
t Time (h, s)
vail in car exhaust gas or in the combustion gas of
x Material loss (cm, mm)
lignite/coal fired power plants, the oxygen con-
x Oxide scale thickness (cm, mm)
tent may be as low as 1% or even smaller.3
A Surface area (m2, cm2)
~FeCr Alloy interdiffusion coefficient (m2 s1) C. Environments based on steam such as, for exam-
D
ple, prevailing in chemical processes or in steam
DO Diffusivity of oxygen (m2 s1)
power plants.4
Hi Interstitial proton
D. Service environments that contain hydrogen in addi-
JO2 Oxygen flux (mol cm2 s1)
tion to water vapor which may prevail in chemical
K15 Equilibrium constant for reaction (15)
plants, gasification plants and (solid oxide) fuel cells.5
K Oxidation rate (mm2 h1)
K0 Oxidation parameter In the first two types of environments (A and B), the
M Molecular mass (g mol1) equilibrium oxygen partial pressure is hardly affected
ð1Þ ð2Þ
NCr ; NCr Critical Cr mole fraction for external scale by the presence of water vapor, whereas in environ-
formation ments C and D it is mainly governed by the absolute
ðsÞ
NO Oxygen solubility (mol/mol) water vapor partial pressure or the H2O/H2 ratio.
Q Activation energy (J mol1) The equilibrium oxygen partial pressures in these
R Universal gas constant (8.314 J mol1 K1) environments are orders of magnitude smaller than
T Temperature (K, C) in gases of type A and B. Consequently, oxide phases
000
VCr Cr vacancy and vapor pressures of volatile species that form by
Vm Molar volume of alloy (cm3 mol1) reaction of metallic components with gases of group
VO Oxygen vacancy A and B may substantially differ from those formed in
DGo Standard free energy (J mol1) gases of groups C and D. When comparing the oxida-
Dm Weight change (mg cm2) tion behavior of metallic materials in dry oxygen with
that in water vapor-containing gases, it is therefore
important to distinguish between the various types of
water vapor-containing atmospheres.
1.17.1 Introduction The following sections deal with the oxidation
properties of metallic construction materials in gases
In practical applications, metallic construction mate- of type C and D. The main emphasis will be put on
rials for high temperature components are subjected describing the oxidation mechanisms in these environ-
to a large variety of service environments. Stability ments compared to the behavior in dry oxygen, oxygen–
and growth rate of the oxide scales that form on the inert gas mixtures or air. The examples shown will
component surfaces during exposure are commonly mainly deal with alloy systems that are commonly
described on the basis of equilibrium oxygen partial used as construction materials in practical applica-
pressures of the various environments. However in tions where typical application temperature ranges
many of these applications, water vapor has been for the respective materials are being considered.
found to substantially alter the technologically rele-
vant properties, mainly growth rate and adherence, of Nomenclature
the surface oxides.1 Thereby it has to be considered In the following sections, the behavior of various
whether water vapor can be present in the service materials in (Ar=N2 )H2O and (Ar/N2–)H2–H2O is
Oxidation in Steam and Steam/Hydrogen Environments 409
compared with that in air, oxygen or various mixtures similar, but in the calculations the change in water
of oxygen and inert gases. The equilibrium oxygen vapor content due to formation of H2, O2, and other
ðgÞ
partial pressures of the two mentioned H2O-contain- gas species such as HO2 and OH(g) has to be taken
ing gas types at, for example, 800 C are typically into account to calculate the oxygen partial pressure
in the range 108 to 1018 bar. This is obviously orders of the fully equilibrated gas or gas mixture. The
of magnitude lower than the pO2 in the mentioned O2- equilibrium oxygen partial pressure in steam and in
based gases. Therefore, in the following discussion the gases with various H2/H2O ratios in the temperature
mentioned H2-containing gases are frequently termed range 500–1100 C is presented in Figure 1.
‘low-pO2 gases,’ and the O2-based environments ‘high- Figure 2 compares the equilibrium oxygen partial
pO2 gases.’ pressure of an Ar–50 vol.% H2O mixture and an
Ar–H2O/H2 mixture (H2O/H2 ratio ¼ 1/1) with
that of the dissociation pressures of selected oxides.
1.17.2 Thermodynamic It is apparent that oxides of Cr, Al, Si, and Ti are stable
Considerations in both atmospheres in the temperature range
500–1100 C. Fe will easily oxidize in Ar–H2O
1.17.2.1 Thermodynamics of the Gas but only at lower temperatures in the Ar–H2–H2O
Atmosphere and Stability of Solid Oxides mixture. In the latter gas, Ni and Co will not oxidize.
In mixtures of hydrogen and water vapor, the oxygen Equation [2] reveals that the equilibrium oxygen
partial pressure is substantially lower than in air partial pressure of an H2O/H2 mixture is governed
where pO2 is approximately 0.2 bar. The equilibrium by the H2O/H2 ratio and thus independent of the
oxygen partial pressure can be derived from: total gas pressure. This is not the case in pure steam
or in mixtures of water vapor with an inert gas.8 In these
2H2 O , 2H2 þ O2 ½1 gases, the equilibrium oxygen partial pressure increases
and with increasing total pressure (Figure 3).

DG pH2 2
pO2 ¼ exp = ½2
RT pH2 O 1.17.2.2 Volatile Reaction Products in
H2O-Containing Gases
in which DG is the standard free energy of
formation.6 The oxidation behavior of metallic materials may be
In pure steam or Ar–H2O mixtures, the equilib- substantially affected by the formation of volatile
ðgÞ
rium hydrogen and oxygen partial pressures are oxides. Well known examples are CrO3 , which
1100 1000 900 800 700 600 500 Temperature ( ⬚C)

0
Air
−5
−10 H2O(g)
−15
log (pO2 bar–1)
−20
H2O(g)/H2(g)
−25
10/1
1/1
−30 1/10
1/100
1/1000
−35
1/10 000
1/100 000
−40
7 8 9 10 11 12 13 14
1/T(104/K)
Figure 1 Equilibrium oxygen partial pressure as function of temperature in various gas atmospheres (calculations
carried out with the Program Fact Sage7).
1100 1000 900 800 700 600 500 T(⬚C)

−5
Ar−50%H2O
−10
Fe3O4/Fe2O3
−15
Ni/NiO
FeO/Fe3O4
log (pO2 bar–1)
−20
Fe/FeO Co/CoO
−25
Ar−H2O/H2
−30
Cr/Cr2O3
−35
Si/SiO2
−40 Ti/TiO
Al/Al2O3
−45
7 8 9 10 11 12 13 14
1/T (104/K)
Figure 2 Comparison of the equilibrium oxygen partial pressure of an Ar–50 vol.% H2O mixture and an Ar–H2O/H2
mixture (H2O/H2 ratio ¼ 1:1) with that of the dissociation pressures of selected oxides (calculations carried out with the
Program Fact Sage7).
−4
−5
300 atm
−6
100 atm
10 atm
−7
log (pO2 atm–1)
1 atm
−8 0.1 atm
−9
−10
−11
−12
400 500 600 700 800 900
T(⬚C)
Figure 3 Calculated equilibrium oxygen partial pressure in steam at various total pressures. Calculations carried out with
the Program Fact Sage [7].
forms in equilibrium with Cr2O3, and SiO(g) in equi- water vapor, a number of additional volatile species
librium with Si or SiO2. The vapor pressure of SiO(g) may form. For the three most important protective
ð gÞ
increases with decreasing pO2 , whereas CrO3 surface oxides on metallic materials, alumina, chro-
becomes more significant at high pO2 . In case of mia, and silica, the following reactions in equilibrium
competition between oxide scale growth and volatili- with the respective oxide phases seem to be the most
zation, a simple Tedmon model,9 in which the overall important ones:10
rate of metal wastage is a function of the parabolic ðgÞ ðgÞ
rate constant for oxide growth and the linear rate 1=2Cr2 O3 þH2 OðgÞ þ3=4O2 ¼ CrO2 ðOHÞ2 ½3
constant for volatilization, is frequently used to ðgÞ
describe the observed kinetics. In the presence of SiO2 þ2H2 OðgÞ ¼ SiðOHÞ4 ½4
ðgÞ from eqn. [3] that the loss of chromium by volatilization

SiO2 þH2 OðgÞ ¼ SiOðOHÞ2 ½5 ðgÞ
of CrO2 ðOHÞ2 should have a significant impact only
on the oxidation process in gases in which partial pres-
ðgÞ
1=2Al2 O3 þ3=2H2 OðgÞ ¼ AlðOHÞ3 ½6 sures of both oxygen and water vapor are high,13 and it
Various data sets11 exist for vapor pressure data of should not be significant for the overall oxidation pro-
these volatile oxyhydroxides, and in the case of cess in the low-pO2 gases considered in the present
ðgÞ
CrO2(OH)2 this literature data substantially differ paper. In such gases, in fact CrðOHÞ3 is the most
(Figure 4). More recent data show, for example, volatile chromium compound over the entire consid-
for the partial pressure of this oxyhydroxide at ered temperature range but the vapor pressure at
850 C in Air + 2% H2O, a variation between 108 700 C, for example, is only about 1013 atm. Galerie
and 109 bar.12 On the basis of own experimental data et al.,14 Michalik et al.,15 and Ehlers et al.16 confirmed that
and an extensive evaluation of the various data sources, at low-pO2 CrO2(OH)2 does not possess sufficient vola-
Opila11 showed that the rates of evaporation increase tility to be considered, for instance, as a possible cause
ðgÞ ðgÞ of breakaway oxidation of ferritic steels in water vapor.
in the following order: AlðOHÞ3 <SiðOHÞ4 <
ðgÞ
CrO2 ðOHÞ2 , as shown in Figure 5. It is apparent Young and Pint17 showed that the volatile Cr oxyhydr-
oxide may become significant in high-pressure steam.
The data of Opila11 shows that the reaction pro-
Temperature ( ⬚C) ducts of water vapor with chromia and silica have
900 700 500 400 300 significant volatility at temperatures above about
800 C, whereas for alumina this is the case only at
−4
Ebbinghaus
Opila et al.
temperatures in excess of 1400 C. It is important to
Gindorf et al. note that the vapor pressures of the most important
−5 IVTANTHERMO volatile Al- and Si-containing species are indepen-
dent of pO2 (eqns [4] and [6]) and they can therefore
Log K
−6
be equally important for oxidation processes in H2O
−7 (–H2)-based gas mixtures and, for example, in wet air.
Also, volatilization of other alloying elements has
−8 been considered to play a role in the oxidation process
1/2Cr2O3(s) + 3/4 O2(g)
+ H2O(g) = CrO2(OH)2(g) in wet environments. For reaction of the ferritic 9% Cr
−9
steel P92 in oxygen/water vapor gas mixtures at 650 C,
8 10 12 14 16 18 formation of volatile species was considered by
10 000/T (K)
Milewska et al.18 as function of pO2 but with a constant
Figure 4 Temperature dependence of the equilibrium water vapor content of 10%. At a low pO2 of 1019 atm
ðsÞ ðgÞ ðgÞ ð gÞ
constant K for the reaction 1=2Cr2 O3 þH2 OðgÞ þ3=4O2 ¼ the most volatile species are FeðOHÞ2 , MoO2 ðOHÞ2 ,
ðgÞ
CrO2 ðOHÞ2 . Compiled by Opila11 using data from various ðgÞ
and WO2 ðOHÞ2 . With increasing pO2 up to 101
sources. ðgÞ ðgÞ
atm, MoO2 ðOHÞ2 and WO2 ðOHÞ2 remain the
ðgÞ
most volatile compounds but CrO2 ðOHÞ2 and
ð gÞ
Mo3 O2 also show significant volatility. The hydrox-
−5
CrO2(OH)2 ðgÞ
ide FeðOHÞ2 as possible active volatile species in the
Log (partial pressure/atm)
Si(OH)4 oxidation of 9% Cr steels in presence of water vapor,

−6
especially at high total pressure, was also proposed by
−7 Thiele et al.,19 Ehlers et al.,16 and Davis and Dinsdale.8
AI(OH)3
−8 This was based on an evaluation by Surman and Cas-
tle,20 but the authors indicated that better data was
−9
required to confirm its significance for oxidation
−10 processes.
p(H2O) = 0.5 atm, p(O2) = 0.5 atm
700 800 900 1000 1100 1200 1300 1400 1500

1.17.2.3 Considerations for Material
Temperature (⬚C)
Testing in Steam Environments
Figure 5 Partial pressures of primary volatile species
calculated for each oxide at 1 atm total pressure: 0.5 atm When testing materials in water vapor-containing
O2, 0.5 atm H2O.10 gases, impurities in the test gas may have an
important effect on the equilibrium oxygen partial tested.21,22 The latter effect was claimed to be responsi-
pressure that establishes in the gas. If, for example, an ble for differences in oxide scale composition and thick-
inert gas such as Ar is bubbled through water at ness in the gas flow direction when testing ferritic steel
controlled temperature to establish a certain partial specimens in slow-flowing Ar–H2O (Figure 7).22 The
pressure of H2O(g), the equilibrium oxygen partial effect did not occur if the test gas contained substantial
pressure of the gas mixture at temperatures between amounts of hydrogen in addition to water vapor.23
500 and 900 C is in the range 109 to 106 bars The possible formation of volatile species due to
(Figure 6). Therefore, traces of oxygen as low as, reaction of oxides with water vapor has important
for example, 1 ppm in the Ar carrier gas may, in this implications for materials testing. Owing to its excel-
temperature range, substantially affect the equilib- lent thermal shock resistance, quartz is frequently used
rium values. This will not be the case in (Ar–)H2– as construction material for reaction tubes and/or
H2O mixtures containing a few percent of H2. specimen holders in oxidation test facilities. In envir-
Hydrogen will at high temperature react with the onments containing substantial amounts of water vapor,
O2 impurity such that the initial content of oxygen the formation of volatile Si-containing gas species such
ð gÞ
in the carrier gas will become insignificant for the as, for example, SiðOHÞ4 can, however, substantially
overall equilibrium oxygen partial pressure. How- affect the oxidation process, as illustrated in Figure 8.
ever, in pure steam or in Ar–H2O mixtures, not In such test gases, reaction tubes and specimen holders
only traces of oxygen but also hydrogen in the initial should therefore, for instance, be made of alumina, in
gas mixture can have a substantial effect on the pO2 of spite of problems that might occur due to the poorer
the equilibrated gas. For example, the presence of thermal shock resistance of this material.
ðgÞ
1 ppm H2 in an Ar–4% H2O gas mixture would, in a A further consequence of the SiðOHÞ4 formation
large temperature range, be sufficient to destabilize is that materials or coatings that rely on formation of
haematite, while traces of oxygen would have the oppo- silica-based surface scales may not be suitable for
site effect (Figure 6). It should be mentioned that traces material protection in steam environments. This is
of hydrogen may, for example, be present in the inert especially the case in high-pressure steam because
ðgÞ
carrier gas, but may also originate from the reaction of eqn [4] shows that the vapor pressure of SiðOHÞ4 is
2
the water vapor with the metallic material to be proportional to ½ pH2 O .
1500 1300 1100 1000 900 800 700 600 500 T(⬚C)
0
Dissosiation pressure of Fe2O3
−5 Ar−50%H2O + 1 ppm O2
Ar−50%H2O
log (pO2 bar–1)
−10
Ar−4%H2O
Ar-4%H2O+1 ppm H2
−15 Ar-4%H2O+10 ppm H2

Ar-50%H2O + 1 ppm H2
Ar-4%H2O+100 ppm H2
Ar-50%H2O + 10 ppm H2
Ar-50%H 2O + 100 ppm H2

−20
−25
5 6 7 8 9 10 11 12 13 14 15
1/T (104/K)
Figure 6 Calculated equilibrium oxygen partial pressures in various Ar–H2O(–H2) gas atmospheres compared
with the dissociation pressure of Fe2O3, showing effect of trace amounts of O2 and H2. Reproduced from Żurek, J.;
Michalik, M.; Schmitz, F.; Kern, T. U.; Singheiser, L.; Quadakkers, W. J. Oxid. Met. 2005, 63(5/6), 401–422,
calculated using FactSage.
Ni-coating
Fe2O3 Ni-coating
Gap Fe3O4 Fe3O4
(Fe,Cr)3O4 + Fe3O4
(Fe,Cr)3O4 + Fe3O4
Internal oxidation
Internal oxidation
Steel
(a) Steel (b)
20 μm 20 μm
Fe3O4
Ni-coating
a
Gas flow
b direction
Internal oxidation
Steel (c)
20 μm
Figure 7 Cross sections in different locations of rectangular specimen (20 9 2 mm) oxidized for 5 h at 15 l h1 gas flow in
Ar–10% H2O at 650 C. Macropicture shows the flow patterns and analysis areas. The specimen was placed in a reaction
tube with a diameter of 10 mm and the linear gas flow rate was 5.31 cm s1.22
Ni-coating Mn/Si-oxide Mn/Si-oxide
(Mn,Cr)3O4
Cr2O3
Alloy 2 μm 1 μm
Figure 8 Formation of mixed Mn/Si oxide on high-Cr ferritic steel during exposure in Ar–H2–H2O mixture with 18.5% H2O(g)
at 800 C using a quartz reaction chamber.
1.17.3 Growth Mechanisms of by an approximate parabolic time dependence. How-

Oxide Scales ever, especially in the early stages of oxidation, the
kinetics of the surface reactions may be the rate-limiting
1.17.3.1 Kinetics of Surface Reactions step in the oxidation process.
Upon long-term oxidation, the growth of oxide scales is Michalik et al.24 studied the early stages of oxidation
frequently governed by solid-state transport processes of Cr at 1000 C in an Ar–1% O2–2% H2O mixture
within the scale. This results in an oxide thickening, using H182 O as tracer (Figure 9). The authors found that
which can, according to the classical oxidation theory the chromia scale formation for up to exposure times of
developed by Wagner,6 in many cases be described 3 h virtually contained only 18O, thus showing that in
100
80
Ototal
60 18O
Cr
40
20
16O
0
0 5000 10 000 15 000 20 000
Sputtering time (s)
Figure 9 SNMS depth profile after oxidation of chromium for 3 h in Ar–1%16O2–2% H18 o 24
2 O at 1000 C.
this stage of the oxidation process oxygen transfer via N2–H2–H2O atmosphere with the same equilibrium
ð gÞ
H2O(g) is much faster than that via O2 . Also, Jacob oxygen partial pressure. The authors explained
et al. reported for chromium oxidized in 18 O2 =H16
25
2 O the results in terms of competitive adsorption of the
mixtures that oxygen derived from water vapor was various gas species on oxide surfaces and grain
the major component of the oxide even though the boundaries. Thus, as pH2 O= pO2 increases, preferential
ðgÞ ðgÞ
pH2 O =pO2 ratio was only 0.05 in that case. adsorption of H2O becomes favored. Evidently, nitro-
In several studies it was found that whisker gen can enter and penetrate dry Cr2O3 grain bound-
formation was promoted by the presence of water aries, but cannot do so when large numbers of water
vapor.26,27 Whiskers are believed to be formed at molecules are present. This confirms the ability of
dislocations in the oxide28 and have hollow cores H2O(g) to interact with the interior of chromia scales
such that rapid surface diffusion within the whisker modifying their transport properties.30
can take place and their growth is then limited by the Akermark and Hultquist31 showed that with an
availability of oxidant at the whisker tip.28 It was Fe–20% Cr–10% Al alloy oxidized at 920 C, water
proposed26 that the faster growth of whiskers in dissociation rates decreased with decreasing oxida-
moist atmospheres is related to the generally faster tion rates, and they interpreted this as the result of
dissociation of water molecules than for most other fewer defects in the more mature alumina scale. In
oxidant molecules. general, it appears that dissociation of a gas molecule
Anghel et al.29 studied the adsorption of N2, H2, such as H2O(g) is favored by defects in the surface,
CO, and H2O on oxidized Cr- and Zr-based materials and the stoichiometric oxides such as alumina may
in the temperature range 400–600 C. When adding therefore be less active than, for example, FeO.
water vapor to CO gas, the dissociation rate of CO on A number of authors have proposed that the rela-
Cr decreased, which can be interpreted as being the tive acidity/basicity of oxides controls the reaction
result of a blocking effect by water. On the basis of with water vapor,32 where water vapor will react
the results of these authors, the following ranking more strongly with more acidic oxides (Figure 10).
for the tendency of adsorption at high temperatures The commonly used method of PZC (point of zero
was made: N2< H2< CO < H2O. This might explain charge) or IEP (isoelectric point) is believed to yield
the results of Michalik et al.,24 who found pure Cr to a good relative assessment for oxide ranking, while
form subscale nitrides during exposure at 1000 C in use of chemical shift measurements by X-ray photo-
N2–O2 but not in N2–H2–H2O. Zheng and Young30 electron spectroscopy (XPS) or heats of adsorption
found that for pure Cr at 900 C subscale nitride by calorimetry or gas chromatography yield better
formation occured in N2–CO–CO2 but not in absolute values.33
4
Ta2O5
log (ki/mg cm–2 h–1)
2
Acidic Basic
VO2
TiO2
0 ZrO2
NbO2
−2
−4 Cr2O3 CoO
NiO
−6
−12 000 −10 000 −8000 −6000 −4000 −2000 0
Cation hydration enthalpy (kJ mol–1)
Figure 10 Correlation between the oxidation rate of metals by water vapor at 800 C and the hydration enthalpy of the
corresponding cation.32
1.17.3.2 Effect of Water Vapor and which, in general, is exothermic so that protons are
Hydrogen on Lattice Diffusion in dominant at low temperatures and oxygen vacancies
Oxide Scales at high temperatures. It is also important to consider
the parallel reaction of dissolution from the gas phase:
The possible effects of hydrogen/water vapor incor-
poration in the transport properties of oxides were H2 OðgasÞ ¼ H2 OðdissÞ ½10
considered by several authors. Hydrogen dissolves in
Thus, different types of hydrogen-containing species
the oxide as protons and is bound to oxide ions to
may be present simultaneously, that is, molecular
form the substitutional hydroxide ðOHÞo .34 These
water and OH groups, as is frequently found in
defects may equivalently and more simply be
glasses.35 In close-packed structures such as alumina
described as interstitial protons Hi . Generally it is
and chromia, dissolution of molecular water would be
found that the concentration of Hi increases with
less favored. Transport of ‘water’ in oxide scales is
increasing pH2 O and decreasing temperature. The
claimed to occur by a proton ‘hopping’ process in
dissolution of protons in an oxide can be written as:
which protons localized at oxide ions move by transfer
1 ðgÞ from one oxygen to another. Since oxygen is available
H ¼ Hi þ e0 ½7
2 2 everywhere, the net effect of proton diffusion is diffu-
or, by considering gas equilibria, sion of water, and since proton diffusion rates are high
compared with those of metal or oxygen, diffusion of
1 1 ðgÞ
H2 OðgÞ ¼ Hi þ e0 þ O2 ½8 ‘water’ is easy. Although the general principles of the
2 4 effects of protons can be outlined, the uncertainties
Thus, at sufficiently low pH2 O or pH2 , protons are regarding transport in the oxides scales of interest are
minority defects. The concentrations of native such that it is not possible to predict with any cer-
defects are independent of pH2 O or pH2 (at constant tainty the effects of water vapor on lattice transport.
pO2 ), while the concentrations of protons increase
with pH2 O ½ or pH2 ½. At higher pH2 O or pO2 , protons
may become the dominant defect, compensated for 1.17.3.3 Molecular Diffusion in Oxide
by defect electrons, metal vacancies, oxygen intersti- Scales in Wet Gases
tials or acceptor dopants. Many investigations focus Already in earlier studies it was claimed that, depend-
on the equilibrium between water vapor, oxygen ing on the scale type, the oxidation kinetics in H2O-
vacancies and protons as being the most likely: containing gases at least partially depends on the

ingress of water molecules through the oxides scales.
H2 OðgÞ þ OxO þ V
O ¼ 2ðOHÞO ½9 For this to happen, ‘microchannels’ must be present,
or, as in the case of silica, an open network structure Void transport by the Rahmel–Tobolski mechanism
ðgÞ
allows the transport of large molecular species. or generation of volatile FeðOHÞ2 obviously requires
It is important to consider gaseous or volatile spe- water vapor to exist within the oxide scale. In addition to
cies that might be created during oxidation reactions proton diffusion, the presence of cracks and fissures in
involving water vapor and how these are transported the oxide is a possible mechanism to allow water ingress.
within the scale. A well-accepted process for vapor or Ehlers et al.16 have found that scales grown in water
gaseous transport is the dissociative mechanism first vapor mixtures on the 9% Cr steel P91 are more per-
proposed by Rahmel and Tobolski.36 This mechanism meable than those grown under dry conditions, which
is applicable only for scales grown by cation transport was demonstrated by 16 O2 =H18 2 O studies (Figure 12).
and involves interaction with void formed by vacancy A competitive adsorption between water and oxygen
condensation.6 If a void contains both hydrogen and occurred at inner surfaces, with water being preferen-
water vapor, oxidation of the surface nearest the metal tially adsorbed if the pH2 O =pO2 ratio is 1.
will occur by reaction with water to form new oxide
and the gaseous reaction product hydrogen (Figure 11).
Adapting this suggestion to Cr2O3 and using the 1.17.4 Metallic High Temperature
Kröger–Vink notation, one can write37 Components in Steam and Steam/
Hydrogen Environments
ðgÞ 000
3H2 OðgÞ ¼ 3H2 þ 3OxO þ 2VCr þ 6h ½11
1.17.4.1 General Remarks
This process occurs in the forward direction, on the side Nearly all high temperature metallic construction
of the cavity nearest the metal, and in the reverse materials rely for their oxidation protection on chro-
direction at the outer surface of the cavity. It provides mia or alumina scales that form on the component
for rapid inward diffusion of oxygen while the void surface during high temperature service. Only in a
gradually moves outwards from the metal/oxide inter- few cases, silica, mostly found as sublayer beneath
face to the oxide/gas interface. chromia base scales, is of importance for metallic
As mentioned in Section 1.17.2.2, Surman and materials protection. Pure silica scales are of far
Castle20 and Ehlers et al.16 have proposed that volatile more importance for environmental protection of a
ðgÞ
FeðOHÞ2 might participate in the scale-formation number of ceramic materials, for example, of the type
process in wet gases and might be an important factor Si nitride or carbide.
for pore generation in an oxide formed on a 9% Cr steel When comparing the behavior of metallic high
in wet gas. Using available thermodynamic data, how- temperature components in steam and steam/hydro-
ever, Ehlers et al.16 calculated that the vapor pressure of gen gases with that in oxygen or air, the literature
ðgÞ
FeðOHÞ2 is too low to sustain the reaction, but pointed data indicate that the service environment may affect
out that these data need verification. The process may the growth of the protective scale (chromia or alu-
be of more importance at high steam pressures.17 mina) as well as that of the less protective oxides of
the base elements (Fe, Ni, Co). Therefore, the effect
of the environment on the overall material behavior
H2O(g) H2O(g) of alloys with high concentrations of protective scale-
forming elements (PSEs) chromium and aluminum
may differ from that of lower alloyed materials that
Oxide scale form oxide scales of the base elements or alloys that
can be considered as ‘borderline’ alloys between pro-
H2O(g) tective and nonprotective behavior.
void H2 Figure 13 shows the qualitative dependence of the
oxidation behavior of Fe-, Ni-, and Co-based alloys as a
function of the amount of PSEs, as frequently measured
when exposing the alloys to oxygen or air.6 With small
Steel
additions of PSEs (group I alloys), the total oxygen
uptake frequently tends to be slightly higher than in
Figure 11 Schematic illustration showing oxygen transfer case of pure base metal oxidation because of doping
across in-scale voids in water vapor-containing gas. (From effects and/or occurrence of internal oxidation.6,38
Ref. 16 based on the Rahmel–Tobolski mechanism36). With a further increase in concentration of PSEs, the
100
Fe2O3 Fe3O4 Fe3O4 + (Fe,Cr)3O4 Alloy
80
60 Ototal
Fe
40
18O
20 16O
Cr
0
0 2 4 6 8 10 12 14 16 18 20
Depth (μm)
Figure 12 MCs+ secondary ion mass spectrometry profile of P91 steel after 30 h oxidation at 650 C in N2–1%16O2–4%
H18
2 O showing differences in distribution of
18
O and 16O in the Fe-based oxide scale. Arrow indicates location of gap in outer
magnetite layer. Reproduced from Ehlers, J.; Young, D. J.; Smaardijk, E. J.; Tyagi, A. K.; Penkalla, H. J.; Singheiser, L.;
Quadakkers, W. J. Corros. Sci. 2006, 48, 3428–3454.
surface treatment. Alloys with a concentration of the

I II III PSE near the steep drop in oxidation rate (group II
alloys) can thus be considered as ‘borderline’ alloys
between protective and nonprotective scale formation.
The commercially available alloy systems for
log (oxidation rate)
high temperature applications are mostly materials

of group II and III in Figure 13. Therefore, in the
following sections the effect of service environment
on the behavior of these types of alloy systems will be
described. First, chromia scale formation on pure Cr
as well as high-Cr Fe- and Ni-based alloys (group III
materials) will be discussed. Subsequently, the scale
formation on corresponding ‘borderline alloys’ with
intermediate Cr contents (group II) will be treated.
Owing to its technological importance, special
Concentration of protective scale forming elements emphasis will be put on steam oxidation behavior of
Figure 13 Schematic illustration showing qualitatively the ferritic/martensitic and austenitic steels in the temper-
effect of Cr and/or Al content on the oxidation rates of Ni-, ature range 550–650 C. The final sections will deal
Fe- and Co-based alloys during exposure in air or oxygen. with alumina-forming alloys and coatings, including
(The oxidation rates of group I and group III alloys are not the behavior of ‘borderline’ alumina formers.
necessary independent of the concentration of PSEs, as
explained, for example, by Birks et al.6).
Nomenclature
oxidation rate decreases as a result of the formation of In the following discussion the chromia and alumina
larger amounts of more protective oxides in the surface scales will be frequently termed as ‘protective scales’
scale. Above a critical concentration of the PSEs, the whereas the oxides of the base elements Fe, Ni, and,
alloy starts forming a slowly growing scale of chromia for example, Ti will be termed ‘nonprotective oxides.’
or alumina. The exact concentration at which this This expression is strictly speaking not correct
fundamental change in oxidation mechanism occurs because oxides with relatively high growth rates still
depends on numerous parameters such as temperature, offer protection of the metal or alloy. The expressions
time, gas composition, minor alloying additions, and will, however, be used for convenience.
1.17.4.2 Oxidation of Chromium and hydroxyl species. At high oxygen activities, molecular
High-Chromium Alloys oxygen adsorbs preferentially on the entire surface,
overwhelming any local variation due to catalysis of
1.17.4.2.1 Oxidation of chromium
H2O dissociation. This explained the absence of whis-
In earlier studies, Quadakkers et al.39 found the oxide
kers in high-oxygen-content atmospheres.
scale on pure chromium in an Ar/H2/H2O mixture
Figure 14 shows results of Michalik et al.15
to grow faster than that formed in air. During cyclic
comparing weight-change data after isothermal
oxidation, however, the overall oxidation resistance of
exposures of Cr in various low- and high-pO2 gases.
the material was far better in Ar/H2/H2O because of
Also shown are the weight changes after cooling to
the better adherence of the scales than that of the
room temperature. The data confirm that samples oxi-
scales formed in air. The oxide scales formed in the
dized in low-pO2 gases (Ar–H2O and Ar–H2–H2O)
low-pO2 gas showed a smaller tendency to buckle, but
generally showed, during cooling, far less extensive
a larger tendency to whisker formation, than found
scale spalling than samples oxidized in high-pO2 gas.
during air exposure.
The most pronounced spalling was found for the
The lower growth rate and/or poorer adherence
specimen oxidized in Ar–20% O2. Figure 15 shows
might be related to the subscale nitride formation
transmission electron microscopy (TEM) micrographs
which is known to occur during air exposure of Cr.
from cross sections of specimens oxidized in Ar–20%
However, similar observations in respect to differences
O2 and Ar–2% H2O at 950 C for 0.5 h. The oxide
in adherence and growth rate were made by Hänsel
scale formed in the high-pO2 gas exhibits pores and
et al.40 when comparing the behavior in Ar–H2–
cracks, mainly at the alloy–oxide interface. The oxide
H2O with that in Ar–O2. Hultquist et al.41 found the
formed during oxidation in the low-pO2 gas appears to
chromia growth rates at 900 C to be approximately
be very dense, the alloy–oxide interface is intact and
doubled in 2% water vapor (20 mbar) compared with
the oxide adheres well to the metal (Figure 15(b)).
the rate in dry oxygen at the same total pressure.
Michalik et al.15,24 carried out two-stage oxidation
A similar acceleration was observed when the Cr
tests with Cr at 1000 C where the gas was, without
contained dissolved hydrogen prior to exposure.42
intermediate cooling, changed from Ar–O2 to Ar–H2–
The adhesion of the scale was improved when it was
H2O. An oxide layer with a morphology typical for
formed in water vapor-containing atmospheres, and
scales formed in Ar–H2–H2O was found at the
this was proposed to be related to a greater fraction
scale/metal interface, beneath the buckled layer grown
of scale growth at the metal–oxide interface. The fact
in Ar–O2.
that adherence of chromia-based scales formed in
The oxidation of Cr and chromia-forming alloys
Ar/H2/H2O is better than those formed in air or
in dry, high-pO2 gases is known to be influenced by
oxygen was also observed by Malkow et al.43 for a
reactive evaporation at the scale/gas interface which
number of high-Cr ferritic steels. ðgÞ
releases CrO3 as the primary species,6 whereas in
At higher levels of water vapor (Ar–15% H2O),
the low-pO2 gas the partial pressure of volatile Cr
Henry et al.44 observed that the growth rate of chromia
species should be extremely low.11,12 The differences
on pure chromium at 900 C was higher than in an
in oxide growth rates observed in the various envir-
Ar–15% oxygen gas mixture. The authors proposed
onments might thus be related to differences in the
that the increased rate of growth was due to enhanced
formation of volatile species in Ar–O2 and Ar(–H2) –
inward scale growth as a result of the smaller ionic
H2O (see Section 1.17.2.2). However, Michalik
radius of OH (95 pm) compared with that of
et al.15 illustrated that, under the used experimental
O2 (140 pm). It was also found that scale adhesion
conditions, this effect could not explain the observed
was significantly improved when water vapor was pre-
differences in scale growth rates.
sent compared with the case for dry oxygen. Hänsel
et al.45 showed that for Cr oxidized at 1000 C the 1.17.4.2.2 Oxidation of chromia-forming
oxidation rates in Ar–H2–H2O mixtures were equal Fe- and Ni-based alloys
or greater than those in Ar–O2 or Ar–O2/H2O and Essuman et al.46 studied the isothermal oxidation
the formation of blades was observed at the oxide behavior of FeCr alloys at 900 C in Ar–O2 and
surface. They proposed a model in which oxide growth Ar–H2–H2O. Energy dispersive X-ray (EDX) analysis
in Ar–H2–H2O was sustained by diffusion of chromium showed that the surface oxide scales formed in both test
vacancies and adsorption of H2O on oxide exposed to gases on Fe–20% Cr consisted essentially of chromia,
low oxygen-activity gas led to the formation of and no indication for the presence of Fe in the scale
Ar–4%H2–2%H2O
Ar–7%H2O Ar–4%H2–7%H2O
3
Ar–20%O2
Weight change (mg cm–2)
2 Ar–1%O2
After isothermal oxidation

−1
After cooling
−2
Figure 14 Weight changes after 72 h isothermal oxidation and after subsequent cooling to room temperature of pure Cr in
different high- and low-pO2 gases at 1000 C. Reproduced from Michalik, M.; Hänsel, M.; Żurek, J.; Singheiser, L.;
Quadakkers, W. J. Mater. High Temp. 2005, 22(3/4), 213–221.
Cr2O3 Cr2O3
Cr
Cr
2 μm 2 μm
(a) (b)
Figure 15 TEM cross sections of oxide scales formed on pure chromium after 0.5 h oxidation at 950 C in (a) Ar–20% O2
and (b) Ar–2% H2O. Reproduced from Michalik, M.; Hänsel, M.; Quadakkers, W. J. Report Forschungszentrum Jülich, 67
(2007), ISBN 978-3-89336-486-2.
could be found. As described above for pure Cr, the 1.17.4.2.3 Mechanisms of chromia
chromia scale that formed in the Ar–H2–H2O gas exhib- scale growth
ited a higher growth rate than that that formed in Ar–O2. The differences in chromia growth in the various
Figure 16 shows gravimetric data during isother- gases appear to be similar for all three discussed mate-
mal oxidation of a binary Ni–25Cr alloy in low-pO2 rials, that is, Cr as well as high-Cr Ni- and Fe-based
gases Ar–(4% H2–)7% H2O and in Ar–20% O2. alloys. The scales exhibit higher growth rates but are
The growth rate in the latter gas was smaller than more compact, show better adherence and frequently
that in the low-pO2 gases and the scale was found to exhibit a flatter metal/scale interface when formed in
have developed voids and internal pores at and near the low-pO2 gas.
oxide–alloy interface. In contrast, the oxide grown on The better adherence of the oxide in the low-pO2
this alloy in Ar–4% H2–7% H2O was compact and gases might be related to the H2O/H2 bridges
closely adherent to the substrate.47 The latter phe- in voids described by Rahmel and Tobolski
nomenon had already been found in the earlier studies (see Section 1.17.3.3). In Ar–O2 the outward-grow-
of Wood et al.48 for scales grown on Fe–28% Cr in ing scale will result in vacancy condensation, prefer-
Ar–10% H2O compared to scales grown in oxygen. entially at the alloy/oxide interface, resulting in poor
3.5
Ni–25Cr
3.0
Weight chnage (mg cm–2)

2.5 Ar–7%H2O
2.0 Ar–4%H2–7%H2O
1.5
1.0 Ar–20%O2
0.5
0.0
0 10 20 30 40 50 60 70 80
Time (h)
Figure 16 Weight change versus time during isothermal oxidation of Ni–25Cr model alloy at 1050 C in Ar–20% O2,
Ar–7% H2O and Ar–4% H2–7% H2O. Reproduced from Essuman, E.; Meier, G. H.; Żurek, J.; Hänsel, M.; Norby, T.;
Singheiser, L.; Quadakkers, W. J. Corros. Sci. 2008, 50(6), 1753–1760.
adherence. The Rahmel–Tobolski mechanism pro- small tendency for the grain size to increase towards
vides an inward oxygen transport resulting in void the scale/alloy interface.
healing and a net transport of the void away from the Figure 18 shows the distribution of oxygen iso-
metal scale interface, which increases scale adher- topes in the scale formed on alloy Ni–25Cr during
ence. However, the mechanism cannot explain why, the two-stage oxidation in Ar–H2–H16 18
2 O/H2 O at
18
in fact, less voidage is seen in the scales formed in the 1050 C. The small O peak at the scale/alloy inter-
low-pO2 gases (Figures 15 and 17). face is a clear proof of inward transport of oxygen-
The growth of chromia scales is generally believed bearing species along rapid diffusion paths, likely grain
to be controlled by boundary diffusion rather than boundaries.51 The continuously decreasing 18O con-
lattice diffusion.37 Nonetheless, if the mobile species centration in the outer part of the scale is, at least
are ionic and at local equilibrium with the surround- partly, related to extensive isotope exchange between
ing oxide, then Wagner’s description6,49 of the depen- 18
O at the grain boundaries and 16O within the grains
dence on pO2 of scale growth rate can be applied, due to the extremely fine size of the chromia grains.51
providing that a suitable effective diffusion coeffi- A similar difference in growth mechanism in a low-
cient is used. If the predominant defects are fully and high-pO2 environment was also found by Bamba
charged vacancies, then it is predicted for the para- et al.52 for the alloy Fe–15Cr–0.5 Si at 850 C.
bolic rate constant kp.49 In earlier studies it was proposed that the better
3 adherence of chromia scales on pure chromium
kp constðpO2 Þ16 ½12
formed in water vapor-containing atmosphere may
However, if the principal defects are interstitials and be related to enhanced scale plasticity.53,54 It would
the ambient pO2 value is significantly above the Cr/ seem that in terms of defect concentration the main
Cr2O3 equilibrium value, then kp is independent of effect of water vapor is to increase the concentration
pO2 . Both predictions are not in agreement with the of the majority species so that one might not expect
experimental observations. to see an effect on creep rates.27 The improved plas-
From the TEM cross sections of the oxidized Cr ticity in the low pO2 gas could be related to the
(Figure 15) and Ni–25Cr specimens (Figure 17), it is smaller chromia grain size compared to that formed
apparent that the oxide scales formed in the high-pO2 on chromia in dry atmospheres. However, on the
gas show large grains with a columnar morphology basis of the newer results, a more likely explanation
which is typical, as frequently described by several is that the observed differences in adherence as well
authors,50 for scales that mainly grow by outward as growth rate of the scales found in the high- and
cation diffusion. The oxides formed in the two low- low-pO2 gases are just a result of the difference in
pO2 gases are, however, extremely fine grained with a scale growth mechanisms. By the substantial inward
Ni-coating
Cr2O3
Cr2O3
(a) Alloy
500 nm (b) 500 nm
Alloy
Figure 17 TEM cross sections of oxide scales formed on Ni–25Cr model alloy during 4 h oxidation in (a) Ar–20% O2 and
(b) Ar–4% H2–2% H2O at 1050 C.
100
16O
10
18O
0.1
Oxide Alloy
0.01
0 2000 4000 6000 8000 10000 12000 14000 16000 18000
Sputtering time (s)
Figure 18 Semi-logarithmic plot of the 18O and 16O profiles in the oxide scale on a Ni–25Cr model alloy after two-stage
oxidation at 1050 C: first stage 0.5 h in Ar–4% H2–2% H16 18
2 O; second stage 2 h in Ar–4% H2–2% H2 O. Reproduced from
Żurek, J.; Young, D. J.; Essuman, E.; Hänsel, M.; Penkalla, H. J.; Niewolak, L.; Quadakkers, W. J. Mater. Sci. Eng. A 2008,
477, 259–270.
oxygen transport in the low-pO2 gas, growth of voids their movement and thus grain growth. This would
as a result of vacancy condensation at the scale–metal explain the extremely fine oxide grains formed in
interface is suppressed, resulting in better scale Ar(–H2)–H2O.51 However, it could also be argued
adherence. The extremely high grain boundary den- that the stronger contribution of inward scale growth
sity favors inward oxygen transport, resulting in a might not be the result of the finer grain size: rather,
higher oxide growth rate in the low-pO2 gas. the finer grain size could result from the modified
The remaining questions concern the mecha- scale growth process induced by H2O(g). The avail-
nism whereby H2O(g) alters the grain size in the low- able results do not allow a distinction to be drawn
pO2 gas and the way in which inward oxygen transport between the two possibilities. Nonetheless, it is clear
occurs.37 One possibility is that the presence of that the effect of water vapor during oxidation in Ar–
H2O adsorbed at the oxide grain boundaries hinders (H2–)H2O environments is twofold: water molecules
incorporated into the scale provide accelerated mass (SOFCs), as in these devices metallic interconnects
transport while simultaneously promoting the forma- are exposed on one side to the oxidant gas (mostly
tion of a fine-grained, inwardly growing, and there- air) and on the other side to moist hydrogen-based
fore better adhering, oxide scale. environments. Figure 19 shows the oxide scales on a
high-Cr ferritic steel Crofer APU (nominal composi-
1.17.4.2.4 Commercial chromia-forming tion in mass%, 22Cr–0.5Mn–Ti–La) after exposure
Fe- and Ni-based alloys for1000 h at 800 C in air and Ar–H2–H2O. In both
Nearly all commercially available chromia-forming cases a two-layered scale is formed consisting of
alloys contain intentionally added alloying elements Cr2O3 and an outer Cr/Mn-spinel. During air expo-
and impurities which may substantially affect the oxi- sure the spinel is present in the form of well-devel-
dation mechanisms and thus also differences in oxida- oped crystals, whereas in Ar–H2–H2O it possesses a
tion behavior in the mentioned low and high-pO2 gases. blade-like morphology (Figure 20). The scale
The alloying elements and impurities may, for exam- formed in the low-pO2 gas tends to form voids at/
ple, lead to doping of the chromia layer and/or result near the interface between the chromia and the spi-
in formation of other mixed oxides, for example, of the nel layer (Figure 19). Traces of Ti are frequently
spinel type. Only a few systematic studies that deal present in the scales: that is, dissolved in the Cr2O3
with the effect of such impurities and minor alloying near the scale–alloy interface and in form of TiO2
additions on the differences in the oxidation behavior particles near the scale–gas interface.59,60 Presence of
in low- and high-pO2 gases are available. minor concentrations of Al and Si (a few tenths of a
England and Virkar55,56 studied the oxidation of a percent) results in internal oxide precipitates of those
series of thin-foil nickel-based alloys (Inconel 625, elements in the subsurface layer. In some cases the
Inconel 718, Hastelloy X, and Haynes 230) in air and oxides contain traces of reactive elements (REs), for
wet hydrogen ( pO2 ¼ 1021 to 1016 atm, pH2 O ¼ 0.07 example, La or Y, which are added to the ferritic
atm) at 700–1100 C. At 800 C the oxidation rate in steels in metallic form or in the form of an oxide
wet hydrogen was increased by factors ranging dispersion to lower the oxidation rates and/or
between 8 and 30, but at 1100 C the effect of wet improve the scale adherence.53,54
hydrogen was to decrease the oxidation rate com- Contrary to the cases described in the previous
pared to that in air by factors ranging between 1.5 section, the Cr-rich surface scales on the ferritic steel
and 13. Hussain et al.57,58 reported that Hastelloy C-4 show a slightly slower growth rate in low-than in
formed a protective scale in 100% steam at all tem- high-pO2 gas (Figure 19). This is in agreement with
peratures in the range 600–1200 C, whereas the findings of other authors,60,61 who observed that the
scale spalled when formed in air. growth rates of Cr-rich oxide scales on high-Cr fer-
Extensive studies concerning the behavior of ritic steels at 800–900 C in the H2/H2O-based gases
chromia forming-alloys, especially high-Cr ferritic are similar to, or smaller than, in air. The reason
steels, in low- and high-pO2 gases have been carried why these ferritic steels behave, in respect to gas
out in the development of solid oxide fuel cells composition dependence of the oxidation rate,
Ni−coating
Ni−coating
(Mn,Cr)3O4
Cr2O3
(Mn,Cr)3O4 Cr2O3
Steel
Steel
Air 1μm Ar–4%H2–2%H2O 1μm
Figure 19 SEM cross sections of high-Cr ferritic steel Crofer 22 APU after 1000 h oxidation at 800 C in air and Ar–H2–H2O.59,60
(Mn, Cr)3O4
(Mn, Cr)3O4
Air 2 μm Ar–4%H2–2%H2O 2 μm
Figure 20 Surface scale morphologies of high-Cr ferritic steel Crofer 22 APU after 1000 h exposure at 800 C in air and
Ar–H2–H2O. Adapted from Piron-Abellan, J.; Quadakkers, W. J. Report Forschungszentrum Jülich, Jül-4170, 2005; ISSN
0944-2952.
differently from that of Cr and model alloys of the NiCr, and FeCr. This is in agreement with recent
type NiCr and FeCr (see Sections 1.17.4.2.1 and results from Essuman et al.,46 who found for an Y2O3
1.17.4.2.2) cannot be unequivocally derived from dispersion strengthened Ni–Cr-based alloy a similar
the mentioned SOFC-related studies. Also, the atmosphere dependence of the oxidation rate as
differences in behavior mentioned above for the observed for Ni–25Cr.
commercial Ni-based alloys are not immediately The decreased growth rate and improved adher-
apparent from the respective papers. A likely expla- ence of chromia scales grown in air and other oxygen-
nation is that a change in scale growth mechanism rich environments due to the Y addition is commonly
imparted by one or more of the mentioned minor explained by a reduction of outward cation diffusion
alloying additions (Mn, Ti, Si, Al, rare earths) results as a result of Y segregation to the oxide grain bound-
in a change of the gas composition dependence of the aries.54 The predominant inward grain boundary
oxidation rate. Therefore, possible effects of some oxygen transport in combination with the high grain
common minor alloying additions on the scale for- boundary density of the scales formed in Ar–4% H2–
mation of chromia-forming alloys in low- and high- 7% H2O (Section 1.17.4.2.3) has the result that Y-
pO2 gases will be discussed in the following section. addition to the Ni–Cr alloy does not substantially
alter the scale growth rate, unlike the effect observed
1.17.4.2.5 Effect of common minor alloying during oxidation in O2-rich environments.
additions Mn addition apparently enhances the oxidation
Figure 21 shows weight change data of a number of rate and decreases the scale adherence during Ar–O2
ternary Ni–25Cr-based model alloys in low- and exposure, but it decreases the growth rate in Ar–H2–
high-pO2 gas after isothermal exposure at 1050 C. H2O (Figure 21).51 Similar low oxidation rates of the
Also presented are the weight changes after subsequent Mn-containing alloys were found during exposure in
cooling to room temperature. The data shows that Ar–CO–CO2,62 thereby indicating that the underlying
Y reduces, as frequently demonstrated,53,54 the scale mechanism is related to the low-pO2 rather than the
growth rate and spallation from the NiCr alloy during presence of water vapor and/or hydrogen. The ternary
oxidation in Ar–O2. However, the Yaddition has hardly NiCrMn alloy formed a double-layered oxide scale
any effect on the growth rate in the low-pO2 gas. The consisting of Cr2O3 and a thin layer of Cr/Mn spinel
measured weight changes in this gas are even slightly in the outer part of the scale. In the case of Ar–20% O2
higher than those of the corresponding Y-free alloy, exposure, the spinel layer was slightly thinner than
which was found to be related to an additional oxygen that formed in Ar–4% H2–7% H2O; additionally,
uptake due to internal Yoxidation.51 Like Ni–25Cr, the NiO was locally observed. The oxide scale formed on
Ni–25CrY alloy showed a substantially higher oxide alloy NiCrMnY had a morphology similar to that
growth rate in Ar–H2O–H2 than in Ar–O2. Apparently, on alloy NiCrMn; however, no clear indication of
the reactive element addition does not fundamentally Ni-containing oxide was found during Ar–O2 exposure,
alter the atmosphere dependence observed for Cr, and the scale adherence was dramatically improved.
These results strongly indicate that the atmosphere On the basis of the phase diagrams (Figure 22) of
dependence of the oxidation rate of the NiCrMn alloy Naoumidis et al.,63 the spinel phase formed in the
and also that described in the previous section for the low-pO2 gas is stoichiometric MnCr2O4, whereas in
high-Cr ferritic steels are related to the formation of high-pO2 gas it is Mn1+xCr2–xO4. This is confirmed by
the Cr/Mn spinel on top of the chromia scale. Man- the secondary neutral mass spectrometry (SNMS)
ganese has a higher solubility in Cr2O3 at high oxygen results in Figure 23 showing a nearly constant
potentials than at low pO2 values.63 One might specu- Cr/Mn ratio in the outer part of the oxide scale
late that in the low-pO2 gas Mn hinders inward oxygen after exposure of a Mn-containing, high-Cr ferritic
transport by segregating at oxide grain boundaries. As steel in Ar–H2–H2O but a changing ratio as function
far as known to the authors, this effect has not been of depth, and thus as function of pO2 , when formed in
studied extensively and no unequivocal evidence air. Retaining the double-layered oxide structure
could be derived from previous work. Grain boundary during longer time exposure in Ar–H2–H2O would
segregation of Mn in the scale formed in the low-pO2 require a balanced flux of Cr and Mn ions through
gas is not unlikely, considering the large size of the the oxide scale. If, because of the low Mn concentra-
Mn+2 ion compared to that of Cr+3. tion in the alloy the flux of Mn is reduced after longer
3.0 3.0
After 72 h exposure After 72 h exposure
2.5 After cooling to room temperature 2.5 After cooling to room temperature
2.0 2.0
NiCrY
1.5 1.5 NiCr
NiCrMn
NiCr NiCrMnY NiCrMnY
1.0 1.0 NiCrMn
NiCrY
0.5 0.5
0.0 0.0
(a) (b)
Figure 21 Weight changes of Ni–25Cr-based model alloys after isothermal exposure at 1050 C and after subsequent
cooling to room temperature in (a) Ar–20% O2 and (b) Ar–4% H2–7% H2O. Mn and/or Y addition were 0.50 and 0.15%
respectively. Reproduced from Żurek, J.; Young, D. J.; Essuman, E.; Hänsel, M.; Penkalla, H. J.; Niewolak, L.; Quadakkers,
W. J. Mater. Sci. Eng. A 2008, 477, 259–270.
TiO2 Ti4O7
Cr2Tin–2O2n–1
o
io
ati
Ti
rat
Tit
CrTi2O5
cr
to
om
mi
ic
me
Cr2Ti3O9
tom
ato
eta
tal
la
MnTiO3
tal
MnTiO3
la
ato
eta
me
tom
mi
m
to
cr
ic
to
Mn2TiO4
Cr
rat
ati
Cr
o
io
Cr2O3 MnCr2O4 Mn3O4 Cr2O3 MnCr2O4 MnO

(a) Mn to metal atomic ratio (b) Mn to metal atomic ratio
Figure 22 Phase equilibria in the quasi-ternary system: (a) Cr2O3–Mn3O4–TiO2, after annealing at 1000 C in air;
(b) Cr2O3–MnO–Ti4O7 after annealing at 1000 C in a gas with an equilibrium oxygen partial pressure of 10 21 bar.
Reproduced from Naoumidis, A.; Schulze, H. A.; Jungen, W.; Lersch, P. J. Eur. Ceram. Soc. 1991, 7, 55–63.
60 OTotal
Fe
Concentration (at. %) 18
O
40
Cr
16O
20
Mn
0
(a) Sputtering time (s)
OTotal
60
Concentration (at. %)
16O Fe
18O
40
Cr
20
Mn
0
(b) Sputtering time (s)
Figure 23 SNMS profiles of high-Cr ferritic steel Crofer 22 APU after two-stage oxidation at 900 C for (a) 2 h in Ar–20%
16 18
O2, 4 h in Ar–20% 18O2; (b) 2 h in Ar–4% H2–2% H16
2 O, 4 h in Ar–4% H2–2% H2 O. Reproduced from Ertl, S. T. Diss. RWTH
Aachen, 2006.
exposure times,61 a Mn deficit will occur in the spinel using thin components,61 will obviously alter the
layer. In the pO2 gradient existing in the scale, this will described effect of Mn additions on the gas composi-
result in a vacancy flux from the spinel surface tion dependence of the oxide growth kinetics.
towards the spinel–chromia interface. Here these Figure 24 shows metallographic cross sections of
vacancies tend to condense and result in void forma- a high-Cr ferritic steel with a very similar composition
tion, as illustrated in Figure 19. In air, the Mn deficit as that shown in Figure 19 but with minor additions of
in the spinel may be compensated by Cr cations from Si and Al, after exposure in air and H2/H2O at 800 C.
the inner chromia layer. This would result in an The volume increase accompanied by the internal
enhanced outward flux of Cr, which could explain oxide formation leads to outward metal flow61,64 re-
why during air exposure the presence of Mn in the sulting in formation of metallic nodules in the scale
alloy increases not only the overall thickness of the and damage of the oxide scale.61 In spite of this differ-
oxide but also that of the inner chromia scale.51 ence in scale morphology compared to that shown in
It should be mentioned that the rapid incorpora- Figure 19, both steels exhibit a double-layered surface
tion of Mn into the oxide layer may easily lead to oxide of Cr2O3 and Cr/Mn spinel that exhibits a
complete exhaustion of the Mn reservoir in the test slightly higher growth rate in high – than in low – pO2
specimen. This effect, which will be especially pro- gas. Apparently, the atmosphere dependence is mainly
nounced at high oxidation temperatures and/or when governed by presence of the Cr/Mn spinel and is not
Ni-coating
Ni-coating (Mn,Cr)3O4
(Mn,Cr)3O4
Cr2O3
Cr2O3
Al2O3
Al2O3
Steel
Air 2 μm Ar–4%H2–2%H2O 2 μm
Figure 24 Metallographic cross sections of high-Cr ferritic steel Crofer 22 A after 1000 h exposure in Air and in Ar–4%
H2–2% H2O at 800 C. The steel had a very similar composition as that shown in Figure 19; however, it contained 0.1%
Al and 0.1% Si.
fundamentally affected by presence of the minor alloy- dry conditions.6,16,22,49,66–68 These high rates are
ing additions (0.1%) of Al and Si. This may be different related to a change in oxide composition from a chro-
if higher amounts (e.g., a few tenths of a percent) of mia base to an Fe-oxide base scale. They are some-
Si/Al are present. In that case, a more extensive inter- times preceded by an initial period of protection after
nal oxidation and formation of metallic protrusions which breakaway oxidation occurs as a consequence
result in a more severe damage of the external scale of failure of the protective chromium-rich oxide layer.
with the result that molecular transport substantially Referring to Figure 13, the critical Cr content
contributes to the scale growth process.61 It is obvious required for obtaining protective chromia scale for-
that then a substantially different atmosphere depen- mation in water vapor is thus higher than that
dence may occur. This can be derived from results required during O2 or air exposure.
of Hammer et al.,65 who studied the behavior of high- Ehlers et al.4 suggested that during exposure of 9%
Cr ferritic steels containing Al and/or Si up to 1 wt% Cr steels, in Ar–50% H2O at 600–650 C, volatile
ð gÞ
in low- and high-pO2 gas between 700 and 900 C. For FeðOHÞ2 may form in the inner layers of the scale
such alloys the gas composition dependence described which is then converted to hematite or magnetite in the
earlier for alloys forming ‘ideal’ gas-tight two-layered outer layers of the scale thereby contributing to the
spinel/chromia layers is not clearly observed. deterioration of the protective Cr-rich oxide. However,
Further studies on the effect of minor alloying based on the available thermodynamic data, the vapor
additions on the oxidation behavior in low- and pressure of the hydroxide may be too low to pro-
high-pO2 gases were carried out by Galerie et al.14 vide sufficient mass transport by the volatile species.16
The authors investigated the behavior of 18% Cr Galerie et al.14 showed that the first stage of breakaway
ferritic steels containing minor additions of Nb or of Fe–15% Cr at 800–1000 C in Ar–15% H2O corre-
Ti. Both elements affected the growth rate of the sponds to a separation of the chromia scale from the
protective scale but did not fundamentally alter alloy and consequently scale equilibration, allowing the
the relative differences between Ar–O2 and Ar– formation of hematite at the metal–oxide interface.
H2O exposure. Then rapid growth of hematite at this interface disrupts
the chromia scale and breakaway is initiated. The
authors propose that the more acidic nature of hematite
1.17.4.3 Effect of Environment on Scale
compared to chromia encourages more rapid dissocia-
Formation in Case of ‘Borderline Alloys’
tion of water and therefore continued hematite growth is
1.17.4.3.1 Nonprotective oxidation in favored, whereas in the absence of water vapor chromia
water vapor would continue to form and protection is maintained.
For iron–chromium alloys with intermediate Cr con- Fujii and Meussner68 oxidized a range of FeCr alloys
tent in the range 8–20%, it has frequently been found with up to 25% Cr at temperatures between 700 and
that oxidation rates in water vapor may be several 1100 C in Ar–10% water vapor and observed a period
orders of magnitude higher than those found under of protective oxidation before occurrence of high
oxidation rates for the alloys with a Cr content greater despite the presence of transient oxides. In the gas
than 20%. They proposed that the high rates of oxida- with the lowest oxygen potential (Ar–H2–H2O), the
tion were related to hydrogen/water dissociation in Cr is oxidized internally and no external Fe oxide
pores (see Section 1.17.3.3) within the wüstite layer. scale is formed because of the low oxygen activity
Mikkelsen and Linderoth67 explained the breakaway in the gas (Figure 2). This in contrary to Ar–H2O in
oxidation of Fe–22% Cr at 800 C in a H2/Ar gas which the alloy matrix, which is Cr-depleted due to
mixture (7:93) with addition of 12% water vapor to be the internal oxidation, oxidized because of the higher
the consequence of the stability of fast-growing wüstite, pO2 , resulting in non-protective scale formation.
whereas with 1% water vapor wüstite is not stable. A similar effect of water vapor on internal oxidation
Hayashi et al.69 observed for diluted Fe–Al alloys a was found by Ani et al.71 These authors studied the
very similar gas composition dependence as observed behavior of FeCr alloys at 800 C in dry and wet
for Fe–Cr alloys by other authors. gas (pH2 O = 3 103 bar) at a fixed low pO2 of
Summarizing, it can be said that the breakaway 1.1 109 bar. The transition from internal to exter-
type oxidation of ‘borderline’ FeCr-based alloys with nal oxidation was shifted from 8% Cr in the dry gas to
intermediate contents of protective scale-forming 12% Cr in the wet gas. In low Cr alloys, the depth of
elements exhibits a number of common features internal oxidation in the wet gas was substantially
which strongly indicates that the responsible process larger than in the dry gas.
is independent of the type of oxide formed. The These results thus indicate that the inward flux of
mechanisms occurring at high temperatures (700– oxygen in the FeCr alloys is greater in the water
1000 C) seem to be not fundamentally different vapor-containing gases than in the dry gases. It is
from those at lower temperature (500–650 C). Mod- well known that hydrogen from water vapor readily
ified transport processes in, for example, wüstite permeates ferritic steels at high temperatures. Rahmel
due to presence of water vapor have clearly been and Tobolski36 observed hydrogen uptake in the sam-
verified, but are unlikely to be responsible for the ples oxidized in wet gases. Also a ‘dual atmosphere’
initiation of breakaway because it also occurs effect has been observed for ferritic steel intercon-
at temperatures at which wüstite is unstable. nects in SOFCs where the oxidation on the cathode
A mechanism that explains the breakaway oxidation (air) side is influenced by the presence of a hydrogen/
at higher as well as lower temperature will be dis- water vapor mixture on the anode (fuel) side.72 Fur-
cussed in the following sections. thermore, the intrinsic diffusivity of hydrogen in bcc
iron at 900 C is approximately 4 103 cm2 s1,73
which is 3 orders of magnitude larger than that
1.17.4.3.2 Internal oxidation of Cr as trigger of oxygen, which is 2 106 cm2 s1.74 The values
for breakaway oxidation in fcc iron are 7 105 cm2 s1 for hydrogen75 and
Figure 25 shows SEM cross sections of an Fe–10Cr 1.8 107 cm2 s1 for oxygen.74
model alloy after isothermal oxidation for 72 h at The reasons for the enhanced internal Cr oxida-
900 C in Ar–20% O2, Ar–4% H2–2% H2O and tion by hydrogen incorporation into the alloy will
Ar–7% H2O. In Ar–O2 the scale contains substantial now be discussed.
amounts of Fe and consists of Fe2O3 near the scale–
gas interface, an intermediate layer of spinel, and a
1.17.4.3.3 Conditions for external chromia
thin chromia layer in contact with the alloy. Appar-
formation in dry and wet gases
ently, significant transient oxidation occurred before
The development and maintenance of a continuous
the chromia became continuous, and thus Fe–10Cr
external chromia scale require that two criteria are
can be considered as a ‘borderline alloy’ when
satisfied. First, the Cr concentration in the alloy must
exposed to Ar–O2.70 In Ar–H2–H2O, only internal
exceed that necessary for outward diffusion to pre-
oxides, mainly of Fe/Cr spinel, have formed.62 Very
vent internal oxidation.76 This criterion for the case
rapid oxidation occurred in Ar–H2O because the
of Fe–Cr alloys is:
oxide scale consisted mainly of Fe-rich oxides. Simi-
lar observations were made at 1050 C.46 1=2
ð1Þ pg ðSÞ DO V m
These results clearly show that internal oxidation NCr > N ½13
3 O D~FeCr VCrO1:5
of Cr is promoted in the gases containing water vapor.
ð SÞ
In the gas with the highest oxygen potential (Ar–20% where NO is the oxygen solubility in the alloy, DO is
O2), a continuous external chromia layer develops the diffusivity of oxygen in the alloy, D~Fe-Cr is the alloy
interdiffusion coefficient, and Vm and VCrO1.5 are the is, a gas in which Fe oxide can hardly form, it is
ð1Þ
molar volumes of the alloy and oxide, respectively. It apparently between 10 and 20 mass% (NCr between
was assumed here that the internal oxide was chromia. 0.11 and 0.21). In the wet gas Ar–H2O, in which
The factor g* is generally approximated as about 0.3.77 Fe can easily oxidize, it is larger than 20 mass%
ð1Þ
Equation [13] was derived under conditions where (NCr > 0.21 ).
only Cr could oxidize. Under conditions where both Essuman et al.80 have shown that for FeCr alloys
ð1Þ
Fe and Cr can oxidize, the transient oxidation of Fe the parameter regarding internal oxidation NCr (eqn
must also be considered. Gesmundo and Viani78 [13]) moves to higher values as the temperature is
showed that the critical solute concentration increases increased (see Figure 27). Figure 27 shows that this
as the rate constant for the growth of the transient, is indeed the case for exposure in Ar–O2 as well as in
nonprotective oxide increases. Ar–H2O. Thus, for Fe–20Cr, non-protective oxida-
The second criterion is that, once a continuous tion occurs in both gases at 1050 C, only in Ar–
external scale is formed, diffusion in the alloy must be H2O at 900 C and in none of the gases at 800 C.
rapid enough to supply the solute at least at the rate it Consideration of eqn [13] indicates that promot-
is being consumed by scale growth.79 This criterion ing internal oxidation by dissolution of hydrogen
may be expressed as: could result from a reduction of D~Fe-Cr ; however,
1=2 based on data of Park et al.81, this seems to be unlikely.
ð2Þ Vm pkp
NCr ¼ ½14 The second possibility is that water vapor is increasing
32n D~FeCr ð SÞ
the effective value of NO . The rapid diffusion of
where kp is the parabolic rate constant for growth of hydrogen into the specimen may shift the H–O–
the external scale, measured in terms of scale thick- H2O equilibrium such that a metastably high concen-
ness. Based on data Essuman et al.46, it was derived tration of oxygen exists at the specimen surface and
ð1Þ
that, for example, at 900 C NCr for FeCr alloys results in an increased inward flux of oxygen.46 Con-
is
0.03. Comparison of the results in Figures 25–27 sidering the equilibration of water vapor at the speci-
shows that for the FeCr alloys the critical Cr content men surface
for external scale formation in Ar–20% O2 at 900 C
ð1Þ
is near 10 mass% (NCr 0.11 ). In Ar–H2–H2O, that H2 OðgÞ ¼ 2H þ O ½15
Ni-coating Ni-coating
Fe2O3
Fe/Cr spinel
Internal Fe/Cr spinel
Cr2O3
Alloy Alloy
(a) Ar–20%O2 10 µm (b) Ar–4%H2–2%H2O 10 µm
Fe/Cr oxides
FeO + Fe3O4
Fe/Cr oxides
Internal oxides
Alloy Internal oxides Alloy

2 µm
(c) Ar–7%H2O 100 µm (d) Ar–7%H2O
Figure 25 Cross-section SEM micrographs of Fe–10Cr model alloy after isothermal oxidation for 72 h at 900 C in various
atmospheres: (a) Ar–20% O2; (b) Ar–4% H2–2% H2O: (c) Ar–7% H2O; (d) Ar–7% H2O high magnification of the internal
oxidation zone in Fig. (c). Note the differences in magnification.46,62
60
(a) 900 ⬚C
Weight change after 72 h (mg cm–2)

50
Fe–10Cr
40
30
20
Fe–20Cr
10
0
Ar–20%O2 Ar–7%H2O Ar–4%H2–2%H2O
(b) Ni-coating (c)
Cr2O3
FeO + Fe3O4
(Fe, Cr)3O4
Fe/Cr oxides
Ar–7%H2O 20 µm Ar–4%H2–2%H2O 10 µm
Figure 26 (a) Weight change of Fe–10Cr and Fe–20Cr model alloys after 72 h isothermal exposure in different test
environments at 900 C. The dotted/dashed lines do not show the real dependence of the gas composition but are just
inserted for easier visibility of the data points. SEM cross sections of the Fe–20Cr model alloy showing oxide scales formed
during 72 h isothermal oxidation in (b) Ar–7% H2O, and (c) Ar–4% H2–2% H2O at 900 C.46,62
Weight change after 72 h exposure (mg cm–2)
40
Fe–20Cr
30
Ar–7%H2O
20
10
Ar–20%O2
0
800 900 1050
Temperature ( ⬚C)
Figure 27 Weight changes after 72 h isothermal oxidation of a binary Fe–20Cr model alloy in Ar–O2 and Ar–H2O at different
temperatures. Reproduced from Essuman, E.; Meier, G. H.; Żurek, J.; Hänsel, M.; Singheiser, L.; Quadakkers, W. J.
Mater. Sci. Forum 2008, 595–598, 699–706.
the concentration of oxygen at the metal surface may If the hydrogen concentration is reduced by rapid
be expressed as: inward diffusion, the gas phase will establish an
pH O
NO ¼ K15 22 ½16 increased concentration of dissolved oxygen, resulting
NH in a concomitantly higher inward oxygen flux.
The final possibility is that the dissolved hydrogen oxidized internally. Ehlers et al.4,16 showed that an
increases the diffusivity of oxygen. It is known from internal oxidation zone was present after 250 h oxi-
studies on austenitic iron that lattice distortion pro- dation of a 9% Cr steel in Ar–50% H2O at 650 C.
duced by interstitials can affect diffusivities. For exam- After subsequent exposure of the specimen in air, this
ple, the intrinsic diffusivity of carbon increases with internal oxidation zone vanished and a thin, protec-
increasing carbon content, which is usually interpreted tive, Cr-rich layer was formed at the scale–alloy
in terms of carbon interstitials expanding the Fe lattice interface and the scale growth rate in this second
and thus increasing their own mobility.82 It is conceiv- oxidation stage was extremely small. This change in
able that the rapidly diffusing hydrogen interstitials composition and morphology of the inner interface
expand the iron lattice sufficiently so that the diffusiv- upon changing from water vapor to a high-pO2 atmo-
ity of the larger oxygen interstitials is increased. sphere or vice versa can be explained by changes in
transport processes in the scale by one or more of the
mechanisms described in Section 1.17.3. However,
1.17.4.3.4 Breakaway oxidation mechanisms
based on the discussions in the previous sections, the
of FeCr alloys in water vapor
main change in oxidation mechanism that occurs
As stated in Section 1.17.4.3.1, enhanced, breakaway
when changing from water vapor to a high-pO2 gas
type oxidation of ‘borderline’ Fe(Ni)Cr-based alloys in
is the decreased tendency of Cr to become internally
water vapor-containing environments has been
oxidized. The formation of the rapid, breakaway-
described in a large number of publications as well as
type, Fe-rich oxide scales when exposing ‘borderline
in textbooks.1,6 If enhanced internal oxidation of Cr
alloys’ in water vapor is just a result of Cr depletion in
due to dissolution of hydrogen in the alloy is the main
the alloy matrix because Cr oxidizes internally rather
reason for the occurrence of breakaway oxidation in
than in the form of a protective external scale. This
steam or Ar–H2O, then this process will be promoted
mechanism would be promoted by a direct reaction
if molecules of water vapor or hydrogen have direct
of H2O(g) molecules with the metal surface, thus
access to the steel surface. The significance of molec-
explaining the observations described above that
ular transport of water vapor was identified by several
transport of H2O(g) through a preexisting scale is
authors as an important factor for the occurrence of
required to initiate breakaway oxidation. In H2O(g)/
breakaway in O2/H2O mixtures.16 Ikeda et al.83 pro-
O(g)
2 mixtures, competitive adsorption of oxygen and
posed molecular transport as a major pre-requisite for
water vapor molecules at external and internal
the breakdown of initially protective scales on 5–30%
surfaces of the oxide scale was found to govern
Cr steels in O2–10 vol.% H2O at 750 and 900 C.
the growth of the Fe-rich oxide scales at high H2O/
Similar conclusions were derived by Ehlers et al.16
O2 ratios. It would therefore be expected that occur-
from thermogravimetry studies of a 9% Cr steel,
rence of breakaway oxidation is determined by the
where the gas was in situ changed from dry Ar–1%
H2O(g)/O(g) ratio in the gas, as was indeed ex-
O2 to an Ar–1% O2–7% H2O mixture at 650 C. 2
perimentally observed for a 9% Cr steel at 600 C.16
Galerie et al.14,32 proposed that the change from a
Also, the oxidation behavior of Fe-Al alloys with Al
protective oxide scale formed in dry gas into break-
contents of 5 wt% was found to depend on the
away-type oxidation of an Fe–15% Cr alloy at 800–
H2O(g)/O(g)2 ratio.
85
1000 C during subsequent exposure in wet gas was
initiated by molecular transport of water vapor via
microcracks. 1.17.4.3.5 Borderline NiCr alloys in
In many of the studies in which Fe-rich, break- water vapor
away-type oxide scales were observed during expo- Essuman et al.47 studied the oxidation behavior of Ni–
sure of Fe–Cr alloys in water vapor, a zone with 25Cr and Ni–20Cr as well as the ‘borderline alloy’
internal precipitates of Cr-rich oxide was found (see Ni–10Cr in various dry and wet Ar-based gas mix-
Figure 25 and, e.g., Ueda et al.84). In classical oxida- tures. Decreasing the oxygen content in dry gas from
tion theory, the presence of this zone can be 20% to 1% was found to promote the selective oxi-
described as a logical consequence of the formation dation of Cr in the ‘borderline alloy’ Ni–10Cr. It
of the nonprotective Fe-based oxide scale; that is, the could be shown that this is related to a decrease of
oxygen partial pressure below that scale at the scale– the NiO growth rate with decreasing oxygen partial
alloy interface is sufficiently high for Cr to become pressure (Figure 28), which, according to Gesmundo

Ar–20%O2
Ar–1%O2
2
1
Ar–7%H2O
0
0 1 2 3 4 5
Time (h)
Figure 28 Weight change versus time during isothermal oxidation of pure Ni at 1050 C in Ar–20% O2, Ar–1% O2 and
Ar–7% H2O.
et al.,78 promotes selective oxidation of Cr. Exposure because the equilibrium oxygen partial pressure in a
in Ar–H2O would then be expected to result in a gas with a H2O/H2 ratio of 1 is very near to the
further improvement in selective oxidation due to the dissociation pressure of Fe oxide (Figure 2).
very slow growth rate of NiO in this low-pO2
gas.32 However, the Cr-rich surface scale formed in
1.17.4.4 Oxidation of Ferritic and Austenitic
Ar–H2O contained Ni-rich oxide. Also, internal oxi-
Steels in Steam
des of Cr could be found. These results clearly indi-
cated that, like in FeCr alloys, hydrogen promotes 1.17.4.4.1 General remarks
internal oxidation of Cr also in NiCr alloys. A technologically important example in which
In spite of this similarity in mechanism, the gas ‘borderline’ FeCr- and FeNiCr-based alloys are sub-
composition dependence of borderline NiCr alloys jected to steam at relatively high temperatures is
substantially differs from that of corresponding FeCr encountered in many power generation systems. In
alloys, the most obvious difference being that the conventional power plants, steam-carrying compo-
high oxidation rates observed for Fe–10Cr and Fe– nents are commonly designed using low-Cr (e.g.,
20Cr in Ar–H2O46 are not found for Ni-based alloys 2¼% Cr) steels. The behavior of this type of materi-
with the same Cr content (Figure 29). The reason is als at steam temperatures up to 530 C has fre-
that the growth rate of NiO strongly decreases with quently been described86,87 and will not extensively
decreasing pO2 , with the result that in Ar–H2O the be treated here. Increasing the steam temperature to
NiO growth rate is extremely small,32 that is, in fact 600 C and even higher for obtaining higher energy
similar to that of chromia. Even if the Cr content was conversion efficiencies requires construction materi-
so low that a protective chromia scale could not form, als that possess a higher creep strength than that of
the overall oxidation rate would still be quite low the low-Cr steels. This led to introduction of mar-
because of the very slow growth rate of NiO in tensitic steels with Cr contents of 9–10% for tubing
Ar–H2O (Figure 28), and, obviously, in Ar–H2– in heat exchanging components (e.g., T91, T92) as
H2O the NiO will not form at all. well as for various components in steam turbines.88
Fe oxide exhibits quite high growth rates in Ar– Depending on the actual application, steels with
H2O mixtures,14,16,36 and for an Fe–10Cr alloy in higher Cr contents (e.g., 12% Cr in X20CrMoV12 1
Ar–H2O the enhanced internal oxidation of Cr will, or HCM 12) are considered (Table 1); however, these
thus, in contrary to Ni–10Cr, result in formation of generally possess a lower creep strength than the 9%
rapidly growing Fe-rich oxide scales. The behavior of Cr steels.89 Based on the results in the previous
borderline FeCr alloys in Ar–H2O–H2 mixtures section, martensitic 9–12% Cr steels may, from the
strongly depends on the exact gas composition viewpoint of oxidation resistance, be considered as
100
Fe–10Cr
10
Ni–10Cr
1
(a) Ar–20%O2 Ar–1%O2 Ar–7%H2O Ar–4%H2–2%H2O
100
Fe–20Cr
10
Ni–20Cr
1
(b) Ar–20%O2 Ar–1%O2 Ar–7%H2O Ar–4%H2–2%H2O
Figure 29 Weight change of Ni–10Cr and Fe–10Cr (a), as well as Ni–20Cr and Fe–20Cr (b) after 72 h isothermal oxidation
at 1050 C in different test environments. Dotted/dashed lines do not present a real pO2 dependence of the oxidation rate;
they are just inserted for clearer visibility of the data points.
Table 1 Typical examples of ferritic, martensitic, and austenitic steels as well as Ni-based alloys presently considered as
construction materials for steam-carrying components in fossil fuel-fired power plants
Steel designation Nominal composition Steel designation Nominal composition
Ferritic and martensitic steels Austenitic steels

30CrMoNiV5–11 1Cr–1Mo 1.4910 17Cr–13Ni–Mo–N
HCM2S (T23) 2¼Cr–1.5W–V Super 304H 18Cr–10Ni
NF616 (P92) 9Cr–2W–Mo–V–Nb–N TP347HFG 18Cr–10Ni–Nb
P91 9Cr–1W–Mo–V–Nb–N NF709 20Cr–25Ni
E911 9Cr–1Mo–1W–V–Nb–N Save 25 23Cr–18Ni–3Cu–1.5W–Nb-N
X7CrCoNiMo10–6 10Cr–6Co–1Mo HR3C 25Cr–20Ni–Nb–N
VM 12 11Cr–0.5Si–Mn AC66 28Cr–32Ni–Ce–Nb
HCM 12 12Cr–1Mo–1W–V–Nb Ni-based alloys
HCM12A (P-122) 12Cr–1.5W–Mo–V–Nb–Cu-N Alloy 617 22Cr–12Co–9Mo
X20CrMoV12–1 12Cr–1Mo–V Alloy 740 24Cr–20Co–0.5Mo
AISI 430 17Cr–0.5Mn–0.5Si Alloy 263 22Cr–20Co–0.6Al
Nominal concentrations of alloying additions given in mass%.

Ar–50%H2O
Fe3O4
50
Ar–50%H2O
40
Mass change (mg cm–2)
Fe3O4 + (Fe,Cr)3O4
30
20 200 μm
10 Air
Air
0
0 2000 4000 6000 8000 10 000 Cr-rich oxide
Time (h)
20 μm
Figure 30 Weight change versus time for oxidation of 9% Cr steel P92 during exposure at 650 C in air and Ar–50% H2O.
Metallographic cross sections show differences in scale morphology after 10 000 h exposure. Adapted from Ehlers, J.;
Quadakkers, W. J. Report Forschungszentrum Jülich, Jül-3883, 2001; ISSN 0944-2952.
‘borderline’ alloys. This is confirmed by the results probably caused by vacancy condensation resulting
in Figure 30, which shows the weight change data of from the rapid outward growth of the outer magne-
the 9% Cr steel P92 during exposure at 650 C in air tite layer (Figure 31). Further growth of the scale
and in Ar–50% H2O. depends to a large extent on transport processes within
this gap, likely by the so-called H2/H2O bridges (see
1.17.4.4.2 Steam oxidation mechanisms of Section 1.17.3.3). This explains why there is a relative
9–12% Cr steels movement of the ‘transient gap’ towards the magnetite
During exposure in (simulated) steam in the tempera- layer in the outer part of the scale during prolonged
ture range 550–650 C, all ferritic 9– 12% Cr steels tend exposure. As the overall scale thickness increases, the
to exhibit an incubation period in which the oxidation Fe activity at the scale–gas interface gradually
rates are very low (Figure 31). This incubation period decreases because of the increasingly difficult trans-
can range from minutes to several hundreds of hours port of Fe cations to the oxide surface as a result of the
and it increases with increasing amount of protective presence of the in-scale gap (Figures 31 and 32). This
scale-forming elements (e.g., Cr, Si) and the extent of decrease in Fe activity in the outer scale eventually
cold work. In this oxidation stage, oxide layers of the results in formation of hematite, that is, the oxide in
type (Fe,Cr)2O3 and/or Cr-rich (Fe,Cr)3O4 covered by a equilibrium with the gas atmosphere (Figure 32), on
hematite layer are formed.4,19,90 top of the scale. The formation of hematite starts at the
After longer exposure times, breakdown of this grain boundaries of the magnetite layer (Figure 32(b)).
protective scale occurs, which is accompanied by Additionally, molecular gas transport occurs through
formation of rapidly growing magnetite and a zone the outer scale.83
with internal Cr2O3 precipitates (Figure 31).46 At For a 10% Cr steel at 650 C, time t1 in Figure 31
temperatures above approximately 580 C, wüstite is
10–15 min and t5
5 h91. Exact times, however,
may be present near the scale–steel interface.16,84,87 differ with the water vapor content, exact steel com-
The inner and outer layers are separated by a gap, position and surface deformation, for example,
t1 t2
H2
H2O H2 H2O
Protective spinel
Protective Fe3O4
Original alloy surface spinel
Fe3O4 + (Fe,Cr)3O4
Internal oxide
Alloy Alloy
t3 t4 t5
H2O H2
H2O H2
Fe2O3
H2O H2
Fe3O4
Original alloy surface Fe3O4 Fe3O4 Original alloy surface

Fe3O4 + (Fe,Cr)3O4 Fe3O4 + (Fe,Cr)3O4
Fe3O4 + (Fe,Cr)3O4 Internal oxide Internal oxide
Internal oxide
Alloy Alloy Alloy
Figure 31 Proposed mechanism for the oxide scale formation on ferritic 10% Cr steel in Ar-H2O mixtures. Times t1, t2, t3, t4, t5
represent subsequent time steps during the oxidation process. Reproduced from Quadakkers, W. J.; Ennis, P. J.; Żurek, J.;
Michalik, M. Mater. High Temp. 2004, 22(1/2), 37–47.
introduced by grinding or cold work. The depth of exposure at 650 C.16 If, however, H2O(g) is present in
the internal oxidation zone tends to decrease with the void, its partial pressure will be determined by
increasing temperature, as will be more extensively the equilibrium:
discussed later. ðgÞ 1 ðgÞ
The change from protective to non-protective H2 OðgÞ ¼ H2 þ O2 ½18
2
oxidation (t1! t2 in Figure 31) can be explained by
the enhanced oxygen transport in voids by Calculating the equilibrium gas composition at a total
H2/H2O as proposed by Fujii and Meussner.92 Dur- pressure of 1 atm, assuming the gas Ar–50% H2O to
ing exposure in oxygen or air, and during the incuba- be in equilibrium with the dissociation pressure of
tion period in wet gas (t1, Figure 31), voids may be the oxide (1034 atm), reveals for pH2O a value of 106
formed at the interface between steel and the protec- atm. Putting this value in the Langmuir equation, it is
tive Cr-rich oxide as result of vacancy condensation. found that an oxygen flux of 5 106 g cm2 s1 can
In the case of Cr2O3, pO2 at the scale–alloy interface be sustained. This is substantially larger than the
at 650 C is
1034 atm and the flux of oxygen ( JO2 ) actually observed oxidation rate in the wet gas.
across a void can be calculated using the Hertz– Thus, oxygen transport via the H2/H2O bridges36 in
Langmuir equation:93 voids within the scale can in principle explain the
p breakdown of the protective Cr-rich oxide and can
J O2 ¼ ½17 account for the high oxidation rates in Ar–H2O.
ð2pMRT Þ0:5
However, based on the considerations in the pre-
where p is the vapor pressure and M the mass of the vious sections, the change from protective to nonpro-
gas molecules. The flux is found to be extremely low tective oxidation will likely be triggered by enhanced
(in the range 8 1034 g cm2 s1), thus giving Cr internal oxidation of Cr due to hydrogen dissolution
sufficient time to diffuse from the bulk alloy towards in the alloy. Water vapor and hydrogen can be trans-
the inner surface of the void, which can, therefore, ported through the oxide layer, for example, via
heal by again forming a Cr-rich scale. Such a healing microcracks formed as a result of growth stresses.66
process could be illustrated by the TEM cross sec- Nakagawa et al.94 considered the case where different
tions of an oxide scale on a 9% Cr steel after air oxidation processes were occurring on two sides of
2 1 Ni-coating
Fe2O3
Gap
Fe3O4
Steel
(a) (b) 2 μm
20 μm
1 Measured spectrum
Standard spectrum for Fe2O3
2 Measured spectrum
Standard spectrum for Fe3O4
200 400 600 800 1000 1200 1400

(c) Raman shift (cm−1)
Figure 32 Metallographic cross section (a) and SEM picture (b) of oxide surface of 10Cr–1Mo–1W (low Si) steel after 5 h
oxidation in Ar–4% H2O at 650 C. Point 1 and 2 in (a) relate to Raman analyses in (c). Reproduced from Żurek, J.;
Michalik, M.; Schmitz, F.; Kern, T. U.; Singheiser, L.; Quadakkers, W. J. Oxid. Met. 2005, 63(5/6), 401–422.
ferritic steels when used as a boiler tube material for literature is vacancy condensation as a result of the fast
power stations. It was proposed that hydrogen gener- outward migration of metal ions through the scale.6,49
ated on one side of the tube wall diffuses to the side During long-term exposure, void formation seems
with the higher pO2 and deteriorates the protective to occur at different locations in the scale. First voids
properties of the air-formed chromia base scale. In (‘transient gap’) start to form already after very short
many cases cited above, hydrogen diffusion into the oxidation times as shown in Figure 31. If the iso-
metal was observed, a clear sign of direct reaction of thermal exposure is continued, the ‘transient’ gap
the water vapor in the oxidation process. may heal by growth of the inner oxide scale as water
vapor molecules gain access through the outer layer,
1.17.4.4.3 Void and gap formation and the gap may eventually completely close which
The oxide scales after oxidation of 9–10% Cr steels in results in a more or less compact scale. Transport of
water vapor contain nearly always a substantial amount Fe cations to the surface then becomes again possible
of pores and/or microvoids (Figure 33), finally result- and thus the Fe activity at the oxide surface increases.
ing in the formation of gaps that may promote spall- At a sufficiently large flux of Fe cations, the outer
ation of the scales during thermal cycling. The main hematite transforms into magnetite, and after longer
reason for void formation in the oxide scale given in the times (at 650 C, for example, a few hundred hours)
Fe2O3 Buckling
Ni-coating
Fe3O4
Fe3O4
Fe3O4+(Fe,Cr)3O4 Fe3O4+(Fe,Cr)3O4
Internal oxide
Steel
Internal oxide Steel
(a) (b)
20 μm 50 μm
Figure 33 (a) Isothermal oxidation of a 10Cr–Mo–W–Si steel at 625 C in Ar–7% H2O showing buckling of oxide separation at
hematite–magnetite interface, (b) oxidation of 10Cr–Mo–W–Si steel at 600 C in Ar–50% H2O (intermediate cooling every 250 h)
showing formation of compact oxide layer. No evidence of spalling was found up to the total exposure time of 1000 h.
Reproduced from Quadakkers, W. J.; Ennis, P. J.; Żurek, J.; Michalik, M. Mater. High Temp. 2004, 22(1/2), 37–47.
the scale appearance is then similar to that shown in The time required for complete healing of the
Figure 33(b). ‘transient’ gap tends to increase with decreasing tem-
If a thermal cycle is introduced in the early stages perature. After 550 C exposure, remnants of the gap
of the oxidation process, that is, before healing of the within the magnetite can sometimes be found up
‘transient gap,’ scale spallation occurs. Further expo- to quite long exposure times of more than 1000 h
sure then results in growth of the freely exposed (Figure 34(a)), whereas after the same exposure
magnetite layer eventually leading to a scale mor- time at higher temperatures only a dense magnetite
phology similar to that shown in Figure 33(b). The layer, sometimes with isolated voids in the outer
question of whether spallation or healing of the outer scale, is found (Figure 34(b)).
layer above the gap occurs thus depends on the time
at which, for a given temperature, the thermal cycle is 1.17.4.4.4 Long-term behavior
introduced. This could be shown for a 10Cr–Mo–W– The long-term behavior of the various types of com-
Si steel after exposures at 600 and 625 C in Ar– mercially available martensitic Cr steels in steam in
H2O.91 If the specimen was cooled after 70 h exposure, the temperature range 550–650 C can be rationa-
spallation of the top layer occurred because of the lized on the basis of a graph similar to that shown in
presence of the in-scale gap. If the first temperature Figure 13. The borderline between the various oxi-
cycle was introduced after 250 h, sufficient time was dation regimes is substantially affected by alloying
available for healing of the gap and thus excellent additions, as seen by the data in Figure 35, with the
oxide adherence was found even after extended expo- result that a large variation in oxidation rates for
sure for several thousand hours. group II alloys with the same Cr contents may be
As long as the scale contains an outer hema- observed. As will be shown later, the exact location of
tite layer, local buckling by a separation at the the borderline between the various types of oxide
magnetite–hematite interface is sometimes observed scales depends on temperature. Taking these restric-
(Figure 33(a)). This might be related to a mechanism tions into account, it can be said that ferritic steels
which is similar to that frequently described for oxide with a Cr content of less than approximately 8%
films on, for example, pure chromium,39 that is, lat- (group I) exhibit at 600–650 C high oxidation rates
eral growth of the scale due to in-scale oxide forma- and the main oxide phases present in the surface
tion at scale grain boundaries. The cation and anion scales are magnetite and (Fe,Cr) spinel. Steels of
diffusion in the Fe2O3 lattice are known to be much group III exhibit very low oxidation rates and the
smaller than in magnetite and wüstite. Therefore, scale oxide scales mainly consist of Cr2O3 and (Cr,Fe)2O3.
growth is likely to occur to a large extent via and at The steels of group II are ‘intermediate cases’ because
the oxide grain boundaries, and thus in-scale growth the oxidation rates and scale morphologies may sub-
can occur by reaction of inwardly and outwardly stantially differ as a function of time, temperature,
moving species. minor alloying additions, and surface treatment.
There is general consensus that steam-side oxida- in the temperature range 600–650 C, the outer mag-
tion of ferritic and martensitic boiler tubes occurs at a netite layer is mostly compact and no hematite can be
rate given by: found (Figure 36). The magnetite frequently contains
cracks perpendicular to the oxide surface, which are
x ¼ K 0t n ½19
caused by tensile cracking during cooling due to the
where x is the oxide scale thickness, t is the exposure high thermal expansion coefficient of the magnetite.95
time, n is the rate exponent, and K 0 is the proportion- Upon prolonged exposure, the scales start to
ality coefficient, which is a function of temperature, develop more and more microvoidage. This first
alloy composition, oxygen partial pressure, stress occurs in the inner scale (Figure 36). Here the larg-
level, and radius of curvature of the tubing. For tub- est extent of void formation seems to develop at the
ing of ferritic/martensitic steels from groups I and II, interface between the internal oxidation zone, if pres-
n has been found to vary from 1/2 to 1/3, and is ent, and the inner oxide layer (Figure 36). The
believed to be a function of time and temperature.88 eventual pore morphology and distribution in the
During exposure of the ferritic steels of group I or inner scale depend on the way in which the vacancies
II (Figure 35) for longer times of, for example, 1000 h within the scale and/or at the scale–alloy interface
Fe2O3
Gap Fe3O4
Fe3O4
Fe3O4 + (Fe, Cr)3O4

Fe3O4 + (Fe,Cr)3O4
Internal oxide
(a) (b) Internal oxide

20 μm 50 μm
Figure 34 Effect of temperature on morphology and void/gap formation of scale on 10Cr–1Mo–1W (low Si) steel during
exposure in Ar–50% H2O for 1000 h: (left) 550 C; (right) 600 C. Note the differences in magnification. Reproduced from
Quadakkers, W. J.; Ennis, P. J.; Żurek, J.; Michalik, M. Mater. High Temp. 2004, 22(1/2), 37–47.
12
Group I Group II Group III

Weight change after 250 h (mg cm–2)
10
0
0 5 10 15 20
Cr content (wt.%)
Figure 35 Weight change of various ferritic and martensitic steels after 250 h oxidation in Ar–50% H2O at 600 C, derived
from data in Ehlers and Quadakkers.4 Designations ‘Group I–III’ are explained in the text.
(a) high-purity elements after 1000 h exposure in Ar–

50% H2O. Although the thicknesses of the scales on
the two materials do not substantially differ, the void
and crack morphologies exhibit tremendous differ-
ences. The most striking difference is the large
gap occurring in the case of the model alloy near
the alloy–steel interface, but not in case of the com-
mercial steel. This is likely related to differences in
the mechanisms of vacancy condensation, due to the
absence of carbide precipitates in the model alloy.
This results in a concentration of the voidage at the
scale–steel interface, whereas in case of the com-
mercial steel the voids at first tend to be evenly
50 μm
distributed in the whole inner layer.
(b) (c)
During long exposure of the commercial steels of
group II, voidage also starts to occur in the outer
magnetite layer (Figure 36). These voids can be
unevenly distributed or they can coalesce to a gap
at the interface between the inner and outer layers
(Figure 39(a)). Indications have been found that the
latter is caused by limitations of the cation transport
in the inner scale. Here three possible reasons may be
20 μm 20 μm mentioned97:
1. The continuing formation of voids in the inner
Figure 36 Oxide scale formed on 10Cr–1Mo–1W steel
during exposure for 1000 h in Ar–50% H2O at 650 C: scale decreases the transport area available for
(a) overview; (b) higher magnification of scale–steel cation diffusion.
interface; (c) steel microstructure. Reproduced from Ehlers, 2. The two-phase mixture of FeCr spinel and mag-
R. J.; Ennis, P. J.; Singheiser, L.; Quadakkers, W. J.; Link, T. netite in the inner scale changes, near the interface
In EFC Workshop 2001; Schütze, M.; Quadakkers, W. J.,
between outer and inner scale, to a single-phase
Nicholls, J. R., Eds.; The European Federation of Corrosion
Publications, 2001; Vol. 34, Paper 12, pp 178–193. layer. Magnetite, which acts as ‘rapid diffusion
path’ for Fe cations, vanishes with the result that
the transport of Fe cations from the inner to the
coalesce to voids. Figure 36 shows a clear difference
outer scale decreases.
in pore morphology in the outer pure Fe3O4 layer
3. The formation of a protective, Cr-rich spinel layer
and in the inner two-phase layer consisting of Fe3O4 +
at the oxide–alloy interface.
(Fe,Cr,Mn)3O4 and in the internal oxidation zone
adjacent to the alloy. Numerous tiny voids are present The limitation of cation transport has the result that
in the inner two-phase layer (Figure 36) which seem the growth rate of the outer magnetite in the existing
to be correlated with the (Fe,Cr,Mn)3O4 bands oxygen activity gradient can no longer be sustained.
embedded in the Fe3O4 matrix.4 Apparently, the inter- Consequently, inwardly moving vacancies in the
faces between these bands and the magnetite act as magnetite tend to condense at the interface between
nucleation sites for the voids. For a given alloy, the the inner and the outer layer resulting in void and gap
morphology and distribution of the ‘bands’ depend on formation (Figure 39(a)).
temperature, as will be illustrated later. The formation of a Cr-rich layer at the oxide–
This type of void formation, as well as the alloy interface (No. 3 above) preferentially occurs at
morphologies of the inner spinel layers and the higher temperatures (at and above 600 C) because
zone of internal oxidation, is affected by the underly- then diffusion of Cr in the alloy and/or the rate of
ing steel microstructure (Figure 37), including car- spinel formation are sufficiently fast. It is also favored
bides preferentially formed at steel grain boundaries by increasing the amounts of protective scale-forming
and/or martensitic laths.96 Figure 38 compares the elements (Cr, Si), and, therefore, the tendency for
scale cross sections of a 10Cr–Mo–W steel with that extensive void formation in the outer magnetite layer
of a Fe–9% Cr model alloy manufactured from and/or at the interface between inner and outer scale
Oxide scale Oxide scale
Steel
Steel
10 μm (a) 10 μm (b)
Figure 37 Cross-sections of oxide scales on specimens of 9% Cr steel P92 after oxidation for 1250 h at 650 C in Ar–50%
H2O. Prior to exposure the specimens were heat-treated so that they had a martensitic (a) or ferritic (b) microstructure.
Reproduced from Ehlers, R. J.; Ennis, P. J.; Singheiser, L.; Quadakkers, W. J.; Link, T. In EFC Workshop 2001; Schütze, M.;
Quadakkers, W. J., Nicholls, J. R., Eds.; The European Federation of Corrosion Publications, 2001; Vol. 34, Paper 12,
pp 178–193.
Voids
Fe3O4
Gap
Fe3O4 + (Fe, Cr)3O4
Fe–9%Cr model alloy

Steel
(a) (b)
50 μm 50 μm
Figure 38 Metallographic cross sections of oxide scales on (a) commercial 10Cr–Mo–W steel and (b) Fe–9% Cr model
alloy (manufactured from high-purity materials) after 1000 h isothermal exposure in Ar–50% H2O at 625 C, showing
differences in void/gap formation in the scale. Reproduced from Quadakkers, W. J.; Ennis, P. J.; Żurek, J.; Michalik, M. Mater.
High Temp. 2004, 22(1/2), 37–47.
Initiation
of spallation
Initiation
of spallation
Steel Steel
(a) (b)
100 μm 100 μm
Figure 39 Different types of scale spallation during oxidation on 10Cr–1Mo–1W low Si steel after long-term exposure in
Ar–50% H2O at (a) 625 C and (b) 650 C.
will in principle be larger for steels of group II in 700 ⬚C 620 ⬚C 600 ⬚C 570 ⬚C
4.0
Figure 13 than for lower Cr steels (group I).
STBA24
As explained in Section 1.17.4.4.2, extensive
(a) 10 MPa
void/gap formation will eventually result in the 2.0 HCM2S
formation of hematite on top of the outer magnetite
ln (Kp) (mg2cm–4h–1)
layer. Macroscopic observation of hematite is thus 0.0
frequently a strong indication for the formation of a NF616
heavily voided/cracked magnetite layer, which is
–2.0 HCM12A
expected to be prone to spallation. Nishimura Fe
et al.98 described a similar mechanism for the hema- S/H-A
tite formation and claimed that the transformation –4.0 S/H-B
of magnetite into hematite plays a role in oxide
spalling. An interesting feature is that the outer –6.0
surface of the hematite is in some cases flat; in
other cases it exhibits a whisker type morphology, R/H-B
–8.0
possibly related to a mechanism described in 1.00 1.05 1.10 1.15 1.20 1.25
Section 1.17.3.1 for whisker formation on chromia 1/T (1/K•1000)
scales. The reasons for these differences in morphol-
Figure 40 Arrhenius-type plots of oxidation rate constant
ogy of the hematite are, as far as known to the Kp measured at 10 MPa. Reproduced from Yi, Y. S.;
authors, not fully clarified. Watanabe, Y.; Kondo, T.; Kimura, H.; Sato, M. J. Pressure
On the basis of the described mechanisms of Vessel Technol. 2001, 123, 391–397.
void/gap formation, it can generally be said that
the crack initiation/propagation modes occurring
can result in three different types of scale spalla- A similar difference in temperature dependence of
tion: that is, spalling occurring in the outer magne- various steel types was found by Ehlers et al.4 after
tite layer (Figure 39(a)); near the interface between exposure in Ar–50% H2O at 1 bar pressure. Also, the
the oxide and the alloy or the internal oxidation zone data from Husemann,101 derived from service
(Figure 39(b)); or at the interface between inner and exposed tubes, show that at low temperatures
outer scale (Figure 39(a)). The latter mode prefer- (
550 C) the oxidation rates of the 9% Cr steels
entially occurs in case of group II alloys with higher are not substantially lower than those of the low alloy
Cr (and Si) contents and is similar to that frequently steels (Figure 41). These data also show that large
found for austenitic steels (see Section 1.17.4.4.6). differences in oxidation rates may occur between
different batches of one steel, here shown for X20
1.17.4.4.5 Temperature dependence of ‘low Cr’ and ‘high Cr.’ This was also found by other
oxidation during long-term exposure authors during laboratory tests of 10.5% Cr and
Watanabe et al.99 investigated the oxidation behavior 11.3% Cr batches of this steel type (Figure 41(c)).
of various steels in steam in an autoclave in the range The results in Refs. 4 and 101 indicate that at 550 C
560–700 C. Low alloy steels (group I) showed the a clear difference exists between low alloy steels
same activation energy (Q ) for the oxidation rate and 12% Cr steel. This contradicts the results of
constant, that is, 350–400 kJ mol1, over the entire Yi et al.100
temperature range (Figure 40). On the other hand, The differences in temperature dependence of the
the 9% Cr and 12% Cr steels (group II) showed a oxidation rates for various steel types (Figure 40) are
bilinear behavior. Below 600 C they had the same Q mainly related to the enhanced incorporation of
but exhibited much lower values in the range 600– chromium and other protective scale-forming ele-
700 C; the higher the Cr content, the lower the Q ments into the scale upon temperature increase.
values. HCM12A had a Q value of 40 kJ mol1. The tendency to form surface scales which are richer
The data show that the oxidation behavior of in Cr upon increasing the temperature from, say,
steels with Cr concentrations between 2 and 8% in 550–650 C is in some steels even more pronounced
superheated steam at temperatures below 600 C is if the temperature is even further increased. This
virtually independent of the Cr content.100 The dif- then has as result that the oxidation rate of, for
ference between 9% Cr steels and 12% Cr steels is example, P91 at 700 C is, at least during short-term
also small at 600 C but becomes marked at 650 C. oxidation, smaller than at 600 C.102
–4
–4.5
Unalloyed steel
–5 2,25 Cr 1Mo
9Cr
–5.5
–6
lg K
18/8 Austenitics
–6.5
–7 TP 347 H fine
TP 347 H fine grain
–7.5
12 Cr
–8
Service test, Sumitomo, Manning + Metcalfe
–8.5
480 500 520 540 560 580 600 620 640 660 680 700 720 740 760
(a) Temperature ( ⬚C)
–5
–5.5
TP347 H
–6
1.4910
–6.5 35 266 h
T91 Esshete
lg K
–7
X20 (10%Cr)
9 Cr TP347 H fine grain
–7.5
–8
–8.5 35 266 h
12 Cr
AC 66
–9
480 500 520 540 560 580 600 620 640 660 680 700 720 740 760
(b) Temperature ( ⬚C)
30
Laboratory tests
Weight change (mg cm–2) after 250 h
in Ar–50%H2O
25
20
15 P92
1 Cr 1 Mo
10Cr steels
10
X20 “low Cr”
5
X20 “high Cr”

0
540 560 580 600 620 640 660 680
(c) Temperature ( ⬚C)
Figure 41 (a, b) Oxidation rates for various steels derived from literature data and service-exposed tubes. The rate
constant K here refers to the material loss x in mm defined as x2 ¼ Kt whereby t is the time in hours.101 (c) Weight
changes for various steels after laboratory tests of 250 h in Ar–50% H2O.4 Designation ‘X20’ relates to X20CrMoV12–1
(see Table 1).
Depending on the exact steel composition, a 625 C, the weight changes are clearly lower than
decreasing oxidation rate with increasing tem- those observed after the 550 C exposure, whereas
perature can already occur at lower temperatures. at 600 C the value is only slightly smaller than that
Figure 42 shows the mass change data after 1000 h at 550 C. It is important to note that the high value at
exposure of different steels at different temperatures 600 C was already reached after 250 h; in other
in Ar–50% H2O. Samples of 1Cr–Mo–V and 10Cr– words, the oxidation rate between 250 and 1000 h is
Mo–W steels clearly exhibit increasing oxidation as small as that observed at 625 C.
rates with increasing temperature (Figure 42). How- Figure 43 shows the oxide scales formed on the
ever, the 12% Cr steel exhibits lower oxidation rates Co-containing steel after 1000 h of exposure at 550
at 625 C than at 550 C, whereas the Co-containing and 600 C. In agreement with the mass change data
11% Cr steel shows a quite complex temperature in Figure 42, the thicknesses of the scales formed at
dependence (Figure 42). After 1000 h exposure at 550 and 600 C are quite similar. However, the scale
25
Weight change (mg cm–2) after 1000 h
1Cr–1Mo–V
20
10Cr–1Mo–1W low Si
15
10
10Cr–6Co–1Mo after 250 h

5 10Cr–6Co–1Mo
12Cr–1Mo–V
0
540 550 560 570 580 590 600 610 620 630 640 650 660
Temperature ( ⬚C)
Figure 42 Weight changes of four different ferritic steels after exposure for 1000 h in Ar–50% H2O showing different
types of temperature dependence of the oxidation rates. Also inserted is the weight change of the 10Cr–6Co–1Mo steel
after 250 h, showing that between 250 and 1000 h hardly any weight gain occurred. Reproduced from Żurek, J.; Wessel, E.;
Niewolak, L.; Schmitz, F.; Kern, T. U.; Singheiser, L.; Quadakkers, W. J. Corros. Sci. 2004, 46(9), 2301–2317.
5 mm Steel 5 mm
550 ⬚C Internal oxidation 600 ⬚C
Cr-distribution Cr-distribution
Figure 43 Backscattered electron images and Cr distribution in scales formed on 10Cr–Co–Mo steel after 1000 h exposure
in Ar–50% H2O at 550 and 600 C showing the temperature dependence of Cr distribution in the inner scale. Reproduced
from Żurek, J.; Wessel, E.; Niewolak, L.; Schmitz, F.; Kern, T. U.; Singheiser, L.; Quadakkers, W. J. Corros. Sci. 2004, 46(9),
2301–2317.
morphologies show substantial differences. At 550 C, is visible in a narrow band near the scale–alloy inter-
the inner part of the scale exhibits a wide zone of face. Beneath this, a Cr depletion in the alloy as a
internal oxide precipitates which are embedded in a result of the selective oxidation of Cr is clearly visi-
Co-rich alloy matrix. After exposure at 600 C, the ble. The enhanced diffusion and the associated
internal oxidation zone is absent. increase in the selective oxidation of Cr become
At 550 C, chromium diffusion in the alloy and more pronounced if the temperature is increased to
in the oxide is so slow that Cr can be considered as 625 C. The Cr-rich oxide then becomes stabilized
being virtually immobile. It becomes preferentially already in the very early stages of oxidation, resulting
incorporated in the inner part of the scale as a result in an extremely slow oxidation rate during the whole
of oxidation of the Cr-rich carbides at the steel grain 1000 h of exposure (Figure 42).
boundaries and/or martensite laths prevailing in Qualitatively, these features in respect to the
the alloy.97 The Cr-rich internal oxides become sub- adverse temperature dependence of the oxidation
sequently embedded in the scale by the inwardly behavior can also be observed in the case of the
moving oxidation front.96 The Cr-rich bands in 12% Cr steel (Figure 42), in agreement with earlier
the inner scale thus exhibit a sort of ‘fingerprint’ findings of Haarmann et al.103 for this type of steels.
of the typical microstructure in this type of steel At 550 C, the Cr-enriched regions almost exclu-
(Figures 36(b) and 36(c)). sively consist of oxides initially formed on grain
On the basis of the initial rapid mass change boundaries and martensite laths, thus expressing the
(Figure 42), the scale formation mechanism at original microstructure in the steel. After exposure at
600 C is more similar to that at 550 C than that at 625 C, the alloy exhibits Cr-rich stringers in the
625 C. However, after longer times the oxidation inner oxide layer, arranged parallel to the initial
rate resembles that found at the higher temperatures, steel surface, indicating more rapid Cr diffusion
which means that somewhere in the scale a protective from the bulk alloy in the direction of the scale.104
Cr-rich oxide becomes stabilized after about 250 h of This preferential oxidation of Cr results in a Cr
exposure. The element mapping in Figure 43 shows depletion beneath the Cr-rich spinel layer in the
that the overall Cr enrichment in the inner part of the alloy. After continued Cr depletion, the growth of
scale is not very pronounced, but a strong enrichment the Cr-rich spinel can eventually no longer be
Steel
2 mm 3 mm
550 ⬚C Steel 625 ⬚C
Cr-distribution Cr-distribution
Figure 44 Backscattered electron images and Cr distributions in inner scale formed on 12Cr–Mo–V steel after 1000 h
exposure at 550 and 625 C showing temperature dependence of Cr incorporation and distribution in inner scale.
Reproduced from Z˙urek, J.; Wessel, E.; Niewolak, L.; Schmitz, F.; Kern, T. U.; Singheiser, L.; Quadakkers, W. J. Corros. Sci.
2004, 46(9), 2301–2317.
16
650 ⬚C
Mass change after 250 h (mg cm–2)

14
12
10
600 ⬚C
8
6
550 ⬚C
4
0
9 10 11 12 13 14 15 16
Cr (%)
Figure 45 Weight changes of binary Fe–Cr model alloys after 250 h exposure in Ar–50% H2O in the temperature range
550–650 C. Adapted from Nieto-Hierro, L.; Quadakkers, W. J. Unpublished results, Forschungszentrum Jülich, FRG, 2006.
16
Weigh change after 250 h (mg cm–2)
14
12
10
Fe–9Cr
8
6 Fe–9Cr–4Co
4
Fe–8Cr–0.2Si
2 Fe–9Cr–0.2Si–0.4Mn
0
550 600 650
Temperature ( ⬚C)
Figure 46 Weight changes of Fe–Cr model alloys with various additions of minor elements, after 250 h exposure in Ar–50%
H2O in the temperature range 550–650 C. Derived from Nieto-Hierro and Quadakkers.105
sustained, and as a result Fe-rich oxides are subse- model alloys in Figure 45 illustrate that this is also
quently formed. Repetition of this process leads to the case for steam oxidation in the temperature range
the microstructure in the inner part of the scale 550–600 C.105 If a different temperature dependence
shown in Figure 44. is sometimes observed for the commercial ferritic
The adverse temperature dependence (Figure 42) steels, then apparently factors other than just temper-
can thus be expected in steels that exhibit a higher ature dependence of Cr and oxygen diffusion in the
tendency to form protective Cr-rich scales (group II, alloy affect protective scale formation. On the basis of
Figure 13), that is, it increases with increasing studies with a number of ternary and quaternary
amounts of elements (Cr, Si, etc.) that participate in model alloys, parameters that affect this process are
the formation of protective scales. alloying elements such as Si and Mn which partici-
The adverse temperature dependence, shown for pate in the oxidation process, C content in the alloy
two steels in Figure 42, in fact seems to contradict which apparently affects the Cr-distribution and
the results and considerations in Section 1.17.4.3.3 activity, as well as enhanced dissolution of carbides
which showed that for binary FeCr alloys in the with increasing temperature thereby ‘releasing’ Cr
temperature range 800–1050 C the tendency to for diffusion. The results in Figure 46 confirm the
form protective scales decreases with increasing tem- significance of minor additions of Si and Mn for
perature. The results obtained for high-purity Fe–Cr decreasing oxidation rates upon temperature increase.
O Cr
10 mm
Fe Co
10 mm
Figure 47 Backscattered electron images (left) and X-ray mappings of oxide scale formed on model alloy Fe–9Cr–4Co after
1000 h of exposure at 650 C in Ar–50% H2O. Pictures on left side show an overview of oxide scale and higher magnification of
scale–steel interface, respectively. Adapted from Nieto-Hierro, L.; Quadakkers, W. J. Unpublished results,
Forschungszentrum Jülich, FRG, 2006.
For Si-containing steels, the occurrence of the adverse Temperature (⬚C)

temperature dependence was also found by other 750 700 650 600 550
researchers,106,107 and a detailed description concern- 1
AA-1
ing the effect of Si for disappearance of the internal
oxidation zone and the resulting formation of barrier
0
layers was given by Ueda et al.84 Single additions of Super304H
NF709
other common alloying elements such as W, Co,
log Kp(μm2/h)
and Nb did not fundamentally alter the T-depen- –1

SAVE25
dence of the oxidation rates.4,105 However, Co con-
tents of more than 4% appeared to substantially
HR3C
affect the oxidation behavior of the Fe–Cr model as –2
illustrated in Figure 46.
An Fe–9% Cr–4% Co model alloy forms at 650 C Super304H
–3
an inner and outer scale (Figure 47), as shown for shot blasting
other steels before. However, here, the selective oxi-
dation of Cr, and subsequently Fe, leads in the inner –4
part of the scale to an apparent enrichment of Co in 0.0009 0.001 0.0011 0.0012 0.0013
the remaining metallic regions due to the high disso- 1/T (K–1)
ciation pressure of Co oxide (Figure 47). X-ray dif- Figure 48 Scale growth rate Kp of various austenitic steels
fraction (XRD) revealed even the presence of pure versus reciprocal temperature during steam oxidation. Here
cobalt near the interface between inner and outer Kp is defined as Kp = x2/t with x the scale thickness and t the
scale.105 This results in a microstructure in which time. Reproduced from Muramatsu, K. In Development of
Ultrasupercritical Plant in Japan, Advanced Heat Resistant
the transport of Fe cations is strongly hampered. In
Steel for Power Generation; Viswanathan, R., Nutting, J. W.,
this way, sufficiently large additions of Co (e.g., 4%) Eds.; IOM Communications: London, UK, 1999; Book 708,
substantially decrease the Cr content required to form pp 543–559.
protective Cr-rich oxide scales (Figure 46); however,
no adverse temperature dependence is found.
highest temperature components to be designed
1.17.4.4.6 Austenitic steels using austenitic materials, sometimes from the view-
Attempts to further increase the steam temperature point of mechanical properties and in other cases
in power generation systems frequently requires the for obtaining improved steam oxidation resistance.
The various austenitic steels that are potentially suit- activation energy over the entire temperature range
able as construction materials for heat-exchanging between 550 and 750 C, whereas the activation
components widely differ in composition, for exam- energy Q decreases as Cr concentration increases
ple, in respect to Cr and Ni content (Table 1), and (Figure 48).
consequently steam oxidation resistance may sub- Oxide scales formed on austenitic steels are some-
stantially differ between various materials. times more prone to spallation than those on ferritic
Depending on the composition, the commercially or martensitic steels. Figure 49 shows the gross and
available austenitics may, in respect to steam oxida- net mass changes of an austenitic 25Ni–15Cr–Ti
tion resistance, be described as group II or group III steel as a function of time during exposure in Ar–
alloys in Figure 13. Steels with Cr contents in the 50% H2O in the temperature range 550–650 C. As
range 17–20% may be considered as group II alloys described for some of the high-Cr group II ferritics
and thus show oxide scales that exhibit many features (Section 1.17.4.4.5), different from the materials
similar to those described for the high-Cr ferritics of studied by Muramatsu108 this austenitic steel shows
group II. Muramatsu108 found that the oxidation rates an ‘adverse temperature dependence’: the growth rate
of the austenitic steels are characterized by a single of the initially formed surface scale at 650 C is
40
Net weight
30 changes
20 Martensitic steels
10
Ni-base alloy Austenitic steels

–10
–20
–30
0 2000 4000 6000 8000 10 000 12 000
(a) Time (h)
12
600 ⬚C gross
10
8
6 650 ⬚C gross
4 650 ⬚C net
550 ⬚C gross
2
0
550 ⬚C net
–2
–4
600 ⬚C net
–6
–8
0 2000 4000 6000 8000 10 000 12 000
(b) Time (h)
Figure 49 Mass changes as function of time during oxidation in the temperature range 550–650 C in Ar–50% H2O:
(a) Typical examples showing differences in behavior of martensitic 9–10% Cr steels, austenitics and Ni-base alloy; (b)
Temperature dependence for onset of scale spallation for austenitic steel 25Ni–15Cr–Ti. ‘Net’ and ‘gross’ weight changes
correspond to specimen mass changes without and with spalled oxide, respectively.
smaller than that at 600 C. The data illustrates that obtain protective chromia base scale formation will
no clear correlation exists between the oxidation rate generally be higher for austenitic than for ferritic
and the time for the onset of scale spallation. For steels. In coarse-grained austenitics, the change from
example, at 600 C, the first spallation occurs after formation of Fe/Cr spinel to pure Cr2O3 was reported
3000 h where the net mass change is
5 mg cm2, to require a Cr content of 25% such as prevailing
while at 650 C spallation occurs after 5000 h, that is, in HR3C.88 Owing to the slow diffusion, Cr trans-
when the mass change equals
3.5 mg cm2. At port along grain boundaries in the alloy plays a more
550 C, spallation starts to occur after 2000 h expo- important role in austenitics for the overall scale
sure, where the mass change exceeds
1 mg cm2. formation than in case of ferritic steels. This has
The onset of spallation can thus not easily be corre- the result that the steam oxidation behavior can
lated with a critical thickness of the oxide scale. be strongly affected by the steel grain size or by
The metallographic cross sections presented in ‘imperfections’ introduced by surface deformation
Figure 50 illustrate the typical scale morphologies (Figure 51). An interesting feature sometimes found
on austenitic steels. Spallation generally occurs at the in austenitic steels is that the depletion of Cr
interface between the inner and the outer scale, that is, beneath a Cr-rich surface scale may differ from
near the original metal surface. After the total outer grain to grain (Figure 52).
scale has spalled, the scale regrowth is determined by Beneficial effects due to grain refining have been
transport processes in the thick inner scale and thus pursued as a proven method for improving oxidation
occurs at a rate similar to that prevailing just before the resistance of austenitic steels. A typical example is
occurrence of spalling. Owing to the high Cr content, TP347HFG, which has been applied widely in the
Fe diffusion in the inner scale is slow, and thus the grain-refined condition, and favorable service expe-
regrown oxide exhibits features as described in the rience up to 50 000 h is available. Currently, the
previous sections, which means that it mainly consists technique is being extended to other stainless steels
of hematite on top of a gap. The spalling rate is thus such as SUPER304H. Sato et al.110 have described
governed by an initial smaller weight loss, a slow scale substantial beneficial effects on oxidation resistance
re-growth and consequently a smaller weight loss dur- of austenitic stainless steels due to shot-blasting. This
ing the following spalling steps. This is, for example, technique has been applied to conventional austenitic
clearly illustrated by the 600 C curve in Figure 49. steels in Japan for more than 20 years. The effect of
There are, however, a number of major differ- cold work pretreatment on the corrosion rate111 is
ences in oxide scale formation between 17% and explained by the improved diffusivity of Cr in the
20% Cr austenitic steels and most of the commer- defect structure, preferential oxidation, and forma-
cially available 9–12% Cr martensitic steels. The tion of Cr-rich solid solution (FeFe2xCrxO4) spinel-
effect of carbide formation on martensite laths for type oxide scales.112
the inner scale morphology does obviously not occur Another important feature is the role of Ni in the
in the austenitic steels. Also, chromium diffusion in overall oxidation process. The thermodynamic
austenite is slower by approximately a factor 10 than driving force for oxidation of Ni in the steam envi-
in ferrite.104,109 Therefore, the critical Cr content to ronment is much smaller than that for Cr or Fe.
Mounting Mounting
Fe2O3 Fe2O3 Ni-coating
Ni-coating Fe3O4
Fe3O4 + (Fe, Cr)3O4
Fe3O4 + (Fe, Cr)3O4

Steel
Steel
550 ⬚C 600 ⬚C
20 mm 50 mm
Figure 50 Metallographic cross sections of 25Ni–15Cr–Ti after oxidation in Ar–50% H2O at 550 C after 3000 h and 600 C

after 10 000 h.
Therefore, Ni is generally not present in the outer because in that area of the scale the oxygen partial
magnetite–hematite scale. As described for Co in Co- pressure is apparently lower than the dissociation
containing martensitic steels in Section 1.17.4.4.5, pressure of NiO or Ni/Fe/Cr spinel. The chromium
Ni is frequently found in the inner scale where it and oxygen distribution clearly show that the steel
sometimes prevails in metallic form (Figure 53) initially formed a Cr-rich surface layer, remnants of
Oxide scale
Steel Steel
Sand blasted Polished

100 mm 50 mm
Sand blasted Polished

50 mm 20 mm
Figure 51 Metallographic cross sections showing effect of surface treatment on oxide scale formation of austenitic steel
TP347HFG during exposure in Ar–50% H2O for 1000 h at 625 C. Adapted from Piron-Abellan, J.; Quadakkers, W. J.
Unpublished results, Forschungszentrum Jülich, FRG, 2007.
10 mm Fe
O Cr
Figure 52 Backscattered electron image and X-ray mappings of oxide scale on steel 310N (HR3C) after exposure at 600 C
for 10 000 h in Ar–50% H2O. Adapted from Piron-Abellan, J.; Quadakkers, W. J. Unpublished results, Forschungszentrum
Jülich, FRG, 2007.
O Fe
30 mm
Cr Ni Nb
Figure 53 Backscattered electron image and X-ray mappings of oxide scale on steel Super304H after exposure at 625 C
for 10 000 h in Ar–50% H2O. Adapted from Piron-Abellan, J.; Quadakkers, W. J. Unpublished results, Forschungszentru
Jülich, FRG, 2007.
which can be seen between the inner and outer layer. revealed either no effect of pressure or a change in
In the inner scale, Cr-rich oxide is present as contin- the pressure effect with temperature depending on
uous bands at the alloy grain boundaries and as inter- interpretation.
nal oxides within the grains. Areas with Ni-rich alloy An apparent influence of steam pressure on the
matrix prevail in the inner scale especially near the scale morphology is illustrated in Figure 54 for a
initial steel surface. ferritic 12Cr–1Mo–V steel oxidized for 250 h at
650 C. When exposed to Ar–50% H2O at ambient
1.17.4.4.7 Effect of pressure on steam pressure, hematite was formed at the outer surface,
oxidation whereas only magnetite was present after the high-pres-
Data regarding the influence of pressure on steam sure oxidation. However, the high-pressure exposure
oxidation behavior show a considerable amount of was carried out in an autoclave in stagnant gas, and the
variation.113 Otoguro et al.114 measured the maximum observed effect is probably due to a build-up of hydro-
scale thickness on four austenitic alloys (SUS347 gen in the sealed system. This leads to a decrease in
HTB, 17/14CuMo, 20Cr/25Ni and 22Cr/35Ni) oxygen partial pressure, thus inhibiting hematite for-
after exposure to steam for 500 h at 650 and 700 C mation, as explained in Section 1.17.2.3. The build-up
and pressures up to 35 MPa. On increasing the pres- of H2 may also have affected the high-pressure results
sure at 650 C, two of the alloys (SUS347 HTB, 17/ of Yi et al.,100 which were carried out in stagnant gas.
14CuMo) investigated showed a decrease in scale
growth rate of
20%, one of the alloys (22Cr/
1.17.4.5 Alumina-Forming Alloys and
35Ni) showed no influence of pressure on the scale
Coatings
thickness and one alloy (20Cr/25Ni) showed an
increase in scale thickness by a factor of 4. The 1.17.4.5.1 a-Alumina scales
same trend in results was observed at 700 C. Mon- There are conflicting reports about the behavior of
tgomery and Karlsson115 have summarized the data alumina-forming alloys and coatings in water vapor-
on the influence of pressure on scale growth kinetics containing gases. In an extensive review of the effects
during exposure of martensitic steels to steam envir- of water vapor on the oxidation of metals,1 it was
onments. The high-pressure data for HCM12 indi- concluded that water vapor had little effect on the
cated a decrease in scale growth rate with increasing oxidation behavior of alumina-forming alloys, although
pressure, whereas the high-pressure data for P92 some evidence was cited that hydrogen defects were
(a) (b)
Fe2O3
Fe3O4
Fe3O4
2 μm 2 μm
Figure 54 Surface scale morphology of 12Cr–1Mo–V steel after 250 h exposure at 650 C: (a) flowing Ar–50% H2O at 1 bar
total pressure; (b) static steam at 300 bar total pressure. Adapted from Ehlers, J.; Quadakkers, W. J. Report
Forschungszentrum Jülich, Jül-3883, 2001.
2 μm
Ar–20%O2 2 μm Ar–4%H2–7%H2O
Figure 55 In-lens SEM images of oxide scales formed on an FeCrAlY model alloy after 100 h isothermal oxidation in
low- and high-pO2 gas at 1250 C. Reproduced from Naumenko, D.; Quadakkers, W. J. Unpublished results,
present in Mg-doped alumina and that scale adhesion poor scale adherence. Kirchheim122–124 actually
was reduced by water vapor. It should be noted that the observed hydrogen to segregate at metal–oxide inter-
mentioned studies mainly relate to wet air or wet faces during internal oxidation of Pd-based alloys. In
oxygen rather than low-pO2 gases. recent studies, Smialek125 attributed the poor alumina
Kolb-Telieps et al.116 showed the a-alumina scale scale adherence in wet gases to hydrogen embrittle-
on an FeCrAlY alloy in the temperature range 1200– ment of the scale–alloy interface.
1300 C to grow more slowly in Ar–H2–H2O mixtures Toscano et al.126 studied the thermal cycling
than in Ar–O2 or air (Figure 55). Kochubey et al.117 behavior of a thermal barrier coated (TBC) coated
found a pO2 dependence of the sub-parabolic oxidation Ni-based alloy where the bond coat (BC) was a
rates assuming oxygen grain boundary diffusion to be NiCoCrAlY alloy containing 10 mass% Al. The
the dominating scale growth process whereby the a-alumina scale that formed during cyclic oxidation
oxide grains size increased in scale growth direction.118 at the BC surface beneath the TBC appeared to grow
Also in a low-pO2 exhaust gas116 and in H2/H2O-based faster in air than in Ar–H2–H2O. As a result, the life
shield gases,119 the growth rate of a-alumina on an time of the TBC at 1100 C, which was governed by
Y-doped FeCrAl alloy was found to be smaller than spallation of the a-alumina scale at the BC surface,
that in air. In the low-pO2 exhaust gas, Naumenko was substantially longer in Ar–H2–H2O than in air.
et al.120 found, however, the adherence of the scales to This gas composition dependence was found only
be worse than that of the scales formed in air or Ar–O2. in case of optimized reactive element (RE) doped
Poorer scale adherence of alumina scales due to pres- alloys or coatings which formed a scale that does not
ence of water vapor therefore seems to be a feature allow molecular gas transport. FeCrAl alloys that
which is independent of the pO2 of the test gas. It contained Zr (300 ppm) as a second RE in addition
was also observed by Janakiraman et al.121 during oxi- to Y developed in Ar–O2 as well as in Ar–H2–H2O
dation of Ni-based superalloys and MCrAl coatings in alumina scales, which exhibited substantial porosity.
wet air. Dieckmann35 proposed hydrogen segregation This was related to internal oxide particles of zirconia
at the oxide–metal interface to be a possible reason for that became incorporated into the inwardly growing
a-alumina scale during long-term exposure.127 Poros- morphology may be the reason for the fact that Rama-
ity appeared to develop at the interfaces between the narayanan et al.128 found in the oxidation of an FeCrAl-
alumina scale and the zirconia particles. Therefore the based oxide dispersion strengthened (ODS) alloy at
scale did not develop the ideal columnar, large-grained 1000–1200 C no clear effect of oxygen partial pressure
microstructure shown in Figure 55, and probably on scale growth kinetics.
molecular transport participated in the scale growth
process. This had the result that the pO2 dependence of
the oxidation rate was far less obvious than in case of 1.17.4.5.2 Metastable alumina
the Zr-free alloy (Figure 56). At higher temperature Naumenko et al.129 also studied the effect of gas
(1300 C) in low-pO2 gas, variations in oxide thickness composition on the oxidation of Y-doped FeCrAl
over the specimen were found and locally the scale was alloys at lower temperatures. The authors found
even thicker than after Ar–O2 oxidation (Figure 57). that the dependence of the oxidation behavior on
Indications were found that the Ar–H2–H2O exposure gas composition in the temperature range 800–
modified the internal oxidation of the Zr in a manner 950 C was completely different from that at tem-
similar to that described for FeCr alloys in Section peratures well above 1000 C. At the lower tempera-
1.17.4.3.2, and in this way the Ar–H2–H2O indirectly tures, the Ar–H2–H2O mixture tended to promote
changed the alumina growth rate. The strong depen- the formation of rapidly growing, metastable alumi-
dence of the a-alumina growth rate on scale num oxide phases (e.g., g, Y) compared to Ar–O2
2.5
1200 ⬚C
Ar–20%O2
2.0
Y + Zr
1.5 Ar–4%H2 –7%H2O
Ar–20%O2
1.0 Y
Ar–4%H2 –7%H2O
0.5
0.0
0 10 20 30 40 50 60 70 80 90 100
Exposure time (h)
3.0
1300 ⬚C Y + Zr
Ar–20%O2
2.5
Ar–4%H2 –7%H2O
2.0
Ar–20%O2 Y
1.5
Ar–4%H2 – 7%H2O
1.0
0.5
0.0
0 10 20 30 40 50 60 70 80 90 100
Exposure time (h)
Figure 56 Weight changes during isothermal oxidation of FeCrAlY alloy (Y content 500 ppm) without or with 300 ppm Zr in
high and low-pO2 gas at 1200 and 1300 C. Reproduced from Naumenko, D.; Quadakkers, W. J. Unpublished results,
Al2O3 Al2O3
ZrO2
Embedded
metal ZrO2
Ar–20%O2 10 μm Ar–4%H2 – 7 % H2O 10 μm
Figure 57 SEM images of oxide scales on FeCrAlY + Zr alloy (compare Figure 56) after 100 h oxidation at 1300 C in various
atmospheres.117 Reproduced from Kochubey, V. Ph.D. Thesis, Bochum University, 2005.
TiO2
Ni-coating
Al2O3/ TiO2 Al2O3

Al2O3 Z-phase
Alloy α2–Ti3Al + Z-phase

Alloy
Ar–20%O2 10 μm Ar–4%H2 – 2 % H2O 10 μm
Figure 58 Oxide scales on Ti–46.5Al–4(Cr, Nb, Ta, B) alloy after 96 h oxidation at 800 C in Ar–20% O2 and Ar–4% H2–2%
H2O. Reproduced from Niewolak, L. Unpublished results, Forschungszentrum Jülich, FRG, 2003.
exposure. This effect was also found by Götlind steel containing 1.2% Al protective alumina scale for-
et al.130 when comparing the behavior of a commer- mation after 1000 h exposure at 800 C in air. However,
cial FeCrAl alloy in O2 with that in O2–H2O, thereby in simulated SOFC anode gas (Ar–H2–H2O–(CO)), the
indicating that presence of water vapor promotes alloy tended to form a Cr-rich surface oxide and internal
metastable alumina growth, independent of the oxy- oxidation of Al. The results for these borderline alloys
gen partial pressure in the gas. In a recent study, might be related to enhanced internal oxidation of the
Lodziana et al.131 demonstrated that strongly bonded protective scale-forming element Al due to hydrogen
hydroxide groups form on Y-Al2O3 in the presence incorporation in the alloy, in a similar manner as
of water vapor. It was suggested that g-Al2O3, when described for FeCr alloys in Section 1.17.4.3.2.
formed on FeCrAl alloys at 900 C, can generate a Another alloy system for which retarded protec-
strongly bonded surface hydroxide similar to tive alumina scale formation due to the presence of
Y-Al2O3. This is supported by the observation that water vapor was described is the intermetallic phase
g-Al2O3 powders can absorb significant amounts of g-TiAl. On the basis of numerous studies, g-TiAl
water vapor (
3% by weight) at 900 C.132 can be considered as ‘borderline alloy’ because pro-
tective alumina scale formation in the temperature
1.17.4.5.3 Borderline alumina forming alloys 700–900 C has been shown to depend, for example,
In earlier studies, Boggs133 showed that the oxidation on the exact Al content, microstructure, surface treat-
rate of Fe–5% Al was significantly increased at 800 C ment, alloying additions of, for example, Cr or Ag,
when about 3% water vapor was present either alone and presence of microdopants such as halogens,
or with oxygen. g-Alumina formed initially in dry etc.134–136 Also, for this alloy system most of the
oxygen which later transformed to a-alumina, with studies in respect to the effect of water vapor relate
about 1% Fe-rich nodules. Hayashi et al.69 found for to wet oxygen or wet air and only a few studies are
an Fe–5% Al alloy at 800 C a substantially lower available in low-pO2 gases.
growth rate in N2–20 vol.% O2 than in N2–12.2 vol. % Niewolak137 found protective alumina scale for-
H2O. Malkow et al.43 observed for an 18% Cr ferritic mation on a Ti–46.5Al–4(Cr, Nb, Ta, B) alloy after
0.40 void would only be 1014 atm. This would, using the
650 ⬚C
Langmuir equation (Section 1.17.4.4.2), result in a
0.35 maximum rate of oxygen transfer of 5 1013 g
21 mbar cm2 s1, which is far too low to sustain the oxidation
0.30
rates observed by Niewolak137 in the wet gas.
⌬m/A (mg cm–2)
0.25 Another possible explanation for the nonprotec-

13 mbar tive scale formation in Ar–H2–H2O would be that
0.20 6 mbar water vapor increases the growth rate of the nonpro-
tective oxide, here TiO2, thereby, according to Ges-
0.15
2 mbar mundo et al. (Section 1.17.4.3.3), increasing the
0.10 critical Al content required for protective scale for-
mation. Such an explanation may be derived from the
0.05 results of Wouters et al.,138 who reviewed the oxida-
tion behavior of Ti in H2O containing gas. For exam-
0.00 ple, after 25 h oxidation of Ti at 850 C, the mass gain
0 10 20 30 40 50 60
was higher by about a factor of 3 in 20 mbar H2O
Time (min)
compared with the rate in 20 mbar O2. In the tem-
2.80
850 ⬚C
perature range 650–850 C, the rate of oxidation was
2.45
directly proportional to the water vapor content, and
approximately parabolic kinetics were observed
2.10 17 mbar (Figure 59). Hydrogen was found in the metal after
exposure and there was no change in the conductivity
⌬m/A (mg cm–2)
11.5 mbar
1.75 of the oxide, which remained n-type. Wouters et al.138
proposed that the increased rate of oxidation results
1.40
from incorporation of OH in the titania lattice and
1.05 the more rapid diffusion associated with the smaller
5 mbar
ionic radius than the oxide ion.
0.70 Niewolak137 investigated the oxidation kinetics of
1.32 mbar g-TiAl alloys in a number of dry and wet, high- and
0.35 low-pO2 gases. He found, that the change from a
0.00
protective alumina base scale to a poorly protective
0 10 20 30 40 50 60 TiO2-rich oxide layer (Figure 58) also occurred
Time (min) when decreasing the oxygen content in dry Ar–O2
Figure 59 Weight changes during isothermal oxidation of
gases from 20% to less than 5%. The results thus
Ti in Ar–H2O showing influence of temperature and water indicate that the gas composition dependence shown
vapor pressure. Reproduced from Wouters, Y.; Galerie, A.; in Figure 59 is at least to certain extent related to the
Petit, J. P. Solid State Ionics 1997, 104, 89–96. differences in oxygen partial pressures of the test
gases. Several authors139–141 have shown that the
96 h of isothermal exposure at 800 C in Ar–20% O2 dominating point defects in TiO2 are oxygen vacan-
(Figure 58). However, when oxidized in Ar–4% H2– cies at high oxygen partial pressure, whereas titanium
2% H2O, rapidly growing, mixed Al2O3/TiO2 base interstitials dominate at low pO2. At 800 C the oxy-
scales were formed (Figure 58). As above, this effect gen partial pressure at which this change from one
might have been be caused by an enhanced internal dominating defect type to another occurs is
103
oxidation of Al due to hydrogen incorporation in the bar.142 It has also been shown143–145 that the solubility
alloy. Considering the morphology of the mixed of Al in TiO2 increases with decreasing oxygen par-
Al2O3/TiO2 scales, also H2O/H2 bridges (Section tial pressure, a factor frequently illustrated to be
1.17.3.3) in voids could principally explain this effect, of great importance for oxide scale formation on
although in this alloy system this mechanism seems g-TiAl-based alloys. Decreasing the oxygen partial
unlikely. The oxygen partial pressure in a void at the pressure of the gas thus not only changes the growth
scale–alloy interface will be extremely low, for rate of the ‘nonprotective’ oxide TiO2 but also the
instance, at 800 C in the range 1043 atm. Therefore, competition between formation of the alumina scale
the equilibrium water vapor partial pressure in that and its dissolution in TiO2.
References 29. Anghel, C.; Hörnlund, E.; Hultquist, G.; Limbäck, M. Appl.
Surf. Sci. 2004, 233, 392–401.
30. Zheng, X. G.; Young, D. J. Oxid. Met. 1994, 42(3/4),
1. Douglass, D. L.; Kofstad, P.; Rahmel, A.; Wood, G. C. 163–190.
Oxid. Met. 1996, 45(5/6), 529–620. 31. Åkermark, T.; Hultquist, G. Oxid. Met. 1997, 47(1),
2. Asteman, H.; Svensson, J. E.; Johansson, L. G.; 117–137.
Norell, M. Oxid. Met. 1999, 52(1), 95–111. 32. Galerie, A.; Wouters, Y.; Caillet, M. Mater. Sci. Forum
3. Thiele, M.; Quadakkers, W. J.; Schubert, F.; Nickel, H. 2001, 369–372, 231–238.
Report Forschungszentrum Jülich, Jül-3712,1999; ISSN 33. Sun, C.; Berg, J. C. Adv. Colloid Interf. Sci. 2003,
0944-2952. 105(1–3), 151–175.
4. Ehlers, J.; Quadakkers, W. J. Report Forschungszentrum 34. Norby, T. J. Phys. IV 1993, 3(C9), 99–106.
Jülich, Jül-3883, 2001; ISSN 0944-2952. 35. Dieckmann, R. Mater. High Temp. 2005, 22(172),
5. Quadakkers, W. J.; Piron-Abellan, J.; Shemet, V.; 93–103.
Singheiser, L. Mater. High Temp. 2003, 20(2), 115–127. 36. Rahmel, A.; Tobolski, J. Corros. Sci. 1965, 5, 333–346.
6. Birks, N.; Meier, G. H.; Pettit, F. S. Introduction to the 37. Young, D. J. Mater. Sci. Forum 2008, 595–598,
High Temperature Oxidation of Metals; Cambridge 1189–1197.
University Press: Cambridge, UK, 2006. 38. Wood, G. C. Oxid. Met. 1970, 2(1), 11–57.
7. Bale, C. W.; Pelton, A. D.; Thompson,, W. J.; Eriksson, G.; 39. Quadakkers, W. J.; Norton, J. F.; Canetoli, S.;
Hack, K.; Chartrand, P.; Degterov, S.; Melancon, J.; Schuster, K.; Gil, A. In 3rd International Conference on
Petersen, S. FactSage 5.3, Thermfact, 1976–2004. Microscopy of Oxidation; Newcomb, S. B., Little, J. A.,
8. Davies, H.; Dinsdale, A. Mater. High Temp. 2005, 22, Eds.; The Institute of Materials: Cambridge, 1996;
15–25. pp 609–619.
9. Tedmon, C. S. J. Electrochem. Soc. 1966, 113(8), 40. Hänsel, M.; Quadakkers, W. J.; Singheiser, L.; Nickel, H.
766–768. Report Forschungszentrum Jülich, Jül-3583,1998;
10. Opila, E. J. Mater. Sci. Forum 2004, 461–464, 765–774. ISSN 0944-2952.
11. Opila, E. J.; Jacobson, N. S.; Myers, D. L.; Copland, E. H. 41. Hultquist, G.; Tventen, B.; Hörnlund, E. Oxid. Met. 2000,
JOM, 2006, 22–28. 54(1/2), 1–10.
12. Stanislowski, M. Report Forschungszentrum Jülich, 54 42. Tveten, B.; Hultquist, G.; Norby, T. Oxid. Met. 1999,
(2006), ISBN 3-89336-432-2. 51(3), 221–233.
13. Segerdahl, K.; Svensson, J. E.; Halvarsson, M.; Panas, I.; 43. Malkow, T.; Quadakkers, W. J.; Singheiser, L.; Nickel, H.
Johansson, L. G. Mater. High Temp. 2005, 22, 69–78. Report Forschungszentrum Jülich, Jül-3589,1998;
14. Galerie, A.; Henry, S.; Wouters, Y.; Mermoux, M.; ISSN 0944-2952.
Petit, J. P.; Antoni, L. Mater. High Temp. 2005, 21(4), 44. Henry, S.; Mougin, J.; Wouters, Y.; Petit, J. P.; Galerie, A.
105–112. Mater. High Temp. 2000, 17(2), 231–235.
15. Michalik, M.; Hänsel, M.; Żurek, J.; Singheiser, L.; 45. Hänsel, M.; Quadakkers, W. J.; Young, D. J. Oxid. Met.
Quadakkers, W. J. Mater. High Temp. 2005, 22(3/4), 2003, 59(3/4), 285–301.
213–221. 46. Essuman, E.; Meier, G.; Żurek, J.; Hänsel, M.;
16. Ehlers, J.; Young, D. J.; Smaardijk, E. J.; Tyagi, A. K.; Quadakkers, W. J. Oxid. Met. 2008, 69(3), 143–162.
Penkalla, H. J.; Singheiser, L.; Quadakkers, W. J. Corros. 47. Essuman, E.; Meier, G. H.; Żurek, J.; Hänsel, M.;
Sci. 2006, 48, 3428–3454. Norby, T.; Singheiser, L.; Quadakkers, W. J. Corros. Sci.
17. Young, D. J.; Pint, B. A. Oxid. Met. 2006, 66(3/4), 2008, 50(6), 1753–1760.
137–153. 48. Wood, G. C.; Wright, I. G.; Hodgkiess, T.; Whittle, D. P.
18. Milewska, A.; Hierro, M. P.; Trilleros, J. A.; Bolivar, F. J.; Werkstoffe Korrosion, 1970, 21(11), 900–910.
Pérez, F. J. Mater. Sci. Forum 2004, 461–464, 321–326. 49. Kofstad, P. High Temperature Corrosion; Elsevier
19. Thiele, M.; Teichmann, H.; Schwarz, W.; Quadakkers, W. J. Applied Science: London, 1988.
VGB Kraftwerkstechnik 1997, 77(Heft 2), 135–140. 50. Przybylski, K.; Yurek, G. J. J. Electrochem. Soc. 1988,
20. Surman, P. L.; Castle, J. E. Corros. Sci. 1969, 9, 771–777. 135(2), 517–523.
21. Tomlinson, L.; Cory, N. J. Corros. Sci. 1989, 29(8), 51. Żurek, J.; Young, D. J.; Essuman, E.; Hänsel, M.;
939–965. Penkalla, H. J.; Niewolak, L.; Quadakkers, W. J. Mater.
22. Żurek, J.; Michalik, M.; Schmitz, F.; Kern, T. U.; Sci. Eng. A 2008, 477, 259–270.
Singheiser, L.; Quadakkers, W. J. Oxid. Met. 2005, 52. Bamba, G.; Wouters, Y.; Galerie, A.; Borchardt, G.;
63(5/6), 401–422. Shimada, S.; Heintz, O.; Chevalier, S. Scr. Mater. 2007,
23. Żurek, J.; Michalik, M.; Singheiser, L.; Quadakkers, W. J. 57(8), 671–674.
Mater. Sci. Forum 2006, 522–523, 155–162. 53. Whittle, D. P.; Stringer, J. Philos. Trans. R. Soc. A 1980,
24. Michalik, M.; Hänsel, M.; Quadakkers, W. J. Report 295, 309–329.
Forschungszentrum Jülich, 67(2007), ISBN 978-3- 54. Pint, B. A. In John Stringer Symposium on High
89336-486-2. Temperature Corrosion; Tortorelli, P. F., Wright, I. G.,
25. Jacob, Y. P.; Haanappel, V. A. C.; Stroosnijder, M. F.; Eds.; ASM International: USA, 2003; pp 1–10.
Buscail, H.; Fielitz, P.; Borchardt, G. Corros. Sci. 2002, 55. England, D. M.; Virkar, A. V. J. Electrochem. Soc. 1999,
44, 2027–2039. 146(9), 3196–3202.
26. Raynaud, G. M.; Rapp, R. A. Oxid. Met. 1984, 21(1/2), 56. England, D. M.; Virkar, A. V. J. Electrochem. Soc. 2001,
89–102. 148(4), A330–A338.
27. Saunders, S. R. J.; Monteiro, M.; Rizzo, F. Prog. Mater. 57. Hussain, N.; Qureshi, A. H.; Shahid, K. A.; Chughtai, N.
Sci. 2008, 53(5), 775–837. A.; Khalid, F. A. Oxid. Met. 2004, 61(5), 355–364.
28. Tallman, R. L.; Gulbransen, E. A. J. Electrochem. Soc. 58. Hussain, N.; Shahid, K. A.; Khan, I. H.; Rahman, S. Oxid.
1967, 114(12), 1227–1230. Met. 1994, 41(3), 251–269.
59. Piron-Abellan, J.; Quadakkers, W. J. Report 93. Langmuir, I. Phys. Rev. 1913, 5(2), 329–342.
Forschungszentrum Jülich, Jül-4170, 2005; 94. Nakagawa, K.; Matsunaga, Y.; Yanagisawa, T. Mater.
ISSN 0944-2952. High Temp. 2001, 18(1), 51–56.
60. Huczkowski, P.; Quadakkers, W. J. Report 95. Armitt, I.; Holmes, D. P. The spalling of steam grown
Forschungszentrum Jülich, 65, 2005; oxide from superheater and reheater tube steels,
ISBN 978-3-89336-484-8. Technical Planning Study 76-655, Final Report, EPRI,
61. Huczkowski, P.; Ertl, S.; Pirón-Abellán, J.; USA,1978.
Christiansen, N.; Höfler, T.; Shemet, V.; Singheiser, L.; 96. Ehlers, R. J.; Ennis, P. J.; Singheiser, L.; Quadakkers, W. J.;
Quadakkers, W. J. Mater. High Temp. 2006, 22, 253–262. Link, T. In EFC Workshop 2001; Schütze, M.,
62. Żurek, J.; Quadakkers, W. J. Unpublished results, Quadakkers, W. J., Nicholls, J. R., Eds.; The European
Forschungszentrum Jülich, FRG, 2007. Federation of Corrosion Publications, 2001; Vol 34, Paper
63. Naoumidis, A.; Schulze, H. A.; Jungen, W.; Lersch, P. 12, pp 178–193.
J. Eur. Ceram. Soc. 1991, 7, 55–63. 97. Żurek, J.; Wessel, E.; Niewolak, L.; Schmitz, F.; Kern, T. U.;
64. Krupp, U.; Christ, H. J. Oxid. Met. 1999, 52(3), 277–298. Singheiser, L.; Quadakkers, W. J. Corros. Sci. 2004, 46(9),
65. Hammer, J.; Laney, S.; Jackson, R.; Coyne, K.; Pettit, F.; 2301–2317.
Meier, G. Oxid. Met. 2007, 67(1), 1–38. 98. Nishimura, N.; Komai, N.; Hirayama, Y.; Masuyama, F.
66. Schütze, M.; Renusch, D.; Schorr, M. Mater. High Temp. Mater. High Temp. 2005, 22(1/2), 3–10.
2005, 22(1/2), 113–120. 99. Watanabe, Y.; Yi, Y. S.; Kondo, T.; Inui, K.; Kishinami, T.;
67. Mikkelsen, L.; Linderoth, S. Mater. Sci. Eng. A 2003, Kimura, H.; Sato, M. In Proceedings of the 9th
361(1–2), 198–212. International Conference on Pressure Vessel
68. Fujii, C. T.; Meussner, R. A. J. Electrochem. Soc. 1964, Technology, ICPVT-9, Sydney, Australia; Price, J., Ed.;
111(11), 1215–1221. 2000; pp 545–552.
69. Hayashi, S.; Narita, T. Oxid. Met. 2001, 56(3), 251–270. 100. Yi, Y. S.; Watanabe, Y.; Kondo, T.; Kimura, H.; Sato, M. J.
70. Essuman, E.; Meier, G. H.; Żurek, J.; Hänsel, M.; Pressure Vessel Technol. 2001, 123, 391–397.
Singheiser, L.; Quadakkers, W. J. Scr. Mater. 2007, 57(9), 101. Husemann, R. U. VGB Kraftwerkstechnik 1999, 10,
845–848. 146–149.
71. Ani, M. H. B.; Ueda, M.; Kawamura, K.; Maruyama, T. 102. Ehlers, J.; Smaardijk, E. J.; Penkalla, H. J.; Tyagi, A. K.;
In Proceedings of the European Corrosion Congress, Singheiser, L.; Quadakkers, W. J. In Proceedings of
Paper 183, Eurocorr 2007, 2007. International Corrosion Congress, Cape Town, South
72. Yang, Z.; Xia, G.; Singh, P.; Stevenson, J. W. Solid State Africa, 1999; Vol. 3, Paper 336.
Ionics 2005, 176, 1495–1503. 103. Haarmann, K.; Vaillant, J. C.; Bendick, W.; Arbab, A. The
73. Eichenauer, W.; Künzig, H.; Pebler, A. Z. Metallk. 1958, T91/P91 Book; Vallourec & Mannesmann Tubes, 1999.
49, 220. 104. Piehl, C.; Tökei, Z.; Grabke, H. J. Defect Diffusion Forum
74. Swisher, J. H.; Turkdogan, E. T. Trans. Metall. Soc. AIME 2001, 194–199, 1689–1694.
1967, 239, 426–431. 105. Nieto-Hierro, L.; Quadakkers, W. J. Unpublished results,
75. Fromm, E.; Gebhardt, E. Gase und Kohlenstoff in Forschungszentrum Jülich, FRG, 2006.
Metallen; Springer: Berlin, 1976. 106. Khanna, A. S.; Rodriguez, P.; Gnanamoorthy, J. B. Oxid.
76. Wagner, C. J. Electrochem. Soc. 1952, 99(10), 369–380. Met. 1986, 26(3), 171–200.
77. Rapp, R. A. Acta Metall. 1961, 9(8), 730–741. 107. Ishitsuka, T.; Inoue, Y.; Ogawa, H. Oxid. Met. 2004, 61(1),
78. Gesmundo, F.; Viani, F. Oxid. Met. 1986, 25(5), 269–282. 125–142.
79. Wagner, C. Z. Elektrochem. 1959, 63, 772. 108. Muramatsu, K. In Development of Ultrasupercritical Plant
80. Essuman, E.; Meier, G. H.; Żurek, J.; Hänsel, M.; in Japan, Advanced Heat Resistant Steel for Power
Singheiser, L.; Quadakkers, W. J. Mater. Sci. Forum Generation; Viswanathan, R., Nutting, J. W., Eds.; IOM
2008, 595–598, 699–706. Communications: London, UK, 1999; Book 708,
81. Park, E.; Hüning, B.; Grabke, H. J.; Spiegel, M. Defect pp 543–559.
Diffusion Forum 2005, 237–240, 928–933. 109. Peraldi, R.; Pint, B. A. Oxid. Met. 2004, 61(5), 463–483.
82. Ågren, J. Scr. Metall. 1986, 20(11), 1507–1510. 110. Sato, T.; Fukuda, Y.; Mitsuhara, K.; Sakai, K. In Advances
83. Ikeda, Y.; Nii, K. Trans. Nat. Res. Inst. Met. 1984, 26(1), in Materials Technology for Fossil Power Plants;
52–62. Viswanathan, R., Gandy, D., Coleman, K., Eds.; ASM
84. Ueda, M.; Nanko, M.; Kawamura, K.; Maruyama, T. International, 2005.
Mater. High Temp. 2003, 20(2), 109. 111. Piron-Abellan, J.; Quadakkers, W. J. Unpublished
85. Hayashi, S.; Narita, T. Oxid. Met. 2002, 58(3), 319–330. results, Forschungszenrum Jülich, FRG, 2007.
86. Griess, J. C.; De Van, J. H. Mater. Perform. 1980, 19(6), 112. Leistikow, S.; Wolf, I.; Grabke, H. J. Werkstoffe Korrosion
46–52. 1987, 38, 556–562.
87. Cory, N. J.; Herrington, T. M. Oxid. Met. 1987, 28(5), 113. Osgerby, S.; Quadakkers, W. J. Mater. High Temp. 2005,
237–258. 22(1/2), 27–33.
88. Viswanathan, R.; Sarver, J.; Tanzosh, J. M. J. Mater. Eng. 114. Otoguro, Y.; Sakakibara, M.; Saito, T.; Ito, H.; Inoue, Y.
Perform. 2006, 15(3), 255–274. Trans. Iron Steel Inst. Jpn. 1988, 28, 761–768.
89. Ennis, P. J.; Quadakkers, W. J. VGB PowerTech 2001, 115. Montgomery, M.; Karlsson, A. VGB Kraftwerkstechnik
8, 87–90. 1995, 75(3), 258–264.
90. Mayer, P.; Manolescu, A. V. In High Temperature 116. Kolb-Telieps, A.; Strehl, G.; Naumenko, D.;
Corrosion; Rapp, R. A., Ed.; National Association of Quadakkers, W. J.; Newton, R. High Temperature
Corrosion Engineers, 1983; pp 368–379. Corrosion of FeCrAlY/Aluchrom YHf in environments
91. Quadakkers, W. J.; Ennis, P. J.; Żurek, J.; Michalik, M. relevant to exhaust gas systems, Materials Aspects in
Mater. High Temp. 2004, 22(1/2), 37–47. Automotive Catalytic Converters; Wiley-VCH: Weinheim,
92. Fujii, C. T.; Meussner, R. A. J. Electrochem. Soc. 1963, Germany, 2001.
110(12), 1195–1204. 117. Kochubey, V. PhD thesis, Bochum University, 2005.
118. Naumenko, D.; Gleeson, B.; Wessel, E.; Singheiser, L.; 133. Boggs, W. E. J. Electrochem. Soc. 1971, 118(6),
Quadakkers, W. J. Metall. Mater. Trans. 2007, 38A, 906–913.
2974–2983. 134. Niewolak, L.; Shemet, V.; Thomas, C.; Lersch, P.;
119. Bennett, M. J.; Newton, R.; Nicholls, J. R. Mater. High Singheiser, L.; Quadakkers, W. J. Intermetallics 2004, 12
Temp. 2003, 20(3), 347–356. (12), 1387–1396.
120. Naumenko, D.; Quadakkers, W. J. Unpublished results, 135. Donchev, A.; Richter, E.; Schütze, M.; Yankov, R.
Forschungszentrum Jülich, FRG,2007. Intermetallics 2006, 14(10–11), 1168–1174.
121. Janakiraman, R.; Meier, G. H.; Pettit, F. S. Metall. Mater. 136. Rakowski, J. M.; Meier, G. H.; Pettit, F. S.;
Trans. A 1999, 30A, 2905–2913. Dettenwanger, F.; Schumann, E.; Rühle, M. Scr. Mater.
122. Kirchheim, R. Acta Metall. 1981, 29(5), 835–843. 1996, 35(12), 1417–1422.
123. Kirchheim, R. Acta Metall. 1981, 29(5), 845–853. 137. Niewolak, L. Unpublished results, Forschungszentrum
124. Kirchheim, R.; Hirth, J. P. Scr. Metall. 1982, 16(4), Jülich, FRG,2003.
475–478. 138. Wouters, Y.; Galerie, A.; Petit, J. P. Solid State Ionics
125. Smialek, J. L. JOM 2006, 29–35. 1997, 104, 89–96.
126. Toscano, J.; Naumenko, D.; Gil, A.; Singheiser, L.; 139. Bak, T.; Nowotny, J.; Rekas, M.; Sorrell, C. C. J. Phys.
Quadakkers, W. J. Mater. Corros. 2008, 59(6), 501–507. Chem. Solids 2003, 64(7), 1043–1056.
127. Wessel, E.; Kochubey, V.; Naumenko, D.; Niewolak, L.; 140. Nowotny, J.; Bak, T.; Nowotny, M. K.; Sheppard, L. R.
Singheiser, L.; Quadakkers, W. J. Scr. Mater. 2004, 51 Int. J. Hydrogen Energy 2007, 32(14), 2630–2643.
(10), 987–992. 141. Kofstad, P. Nonstoichiometry, Diffusion and Electrical
128. Ramanarayanan, T. A.; Ayer, R.; Petkovic-Luton, R.; Conductivity in Binary Metal Oxides; Wiley Interscience:
Leta, D. P. Oxid. Met. 1988, 29(5), 445–472. New York, 1972.
129. Naumenko, D.; Quadakkers, W. J.; Galerie, A.; 142. Nowotny, J.; Radecka, M.; Rekas, M.; Sugihara, S.;
Wouters, Y.; Jourdain, S. Mater. High Temp. 2003, 20(3), Vance, E. R.; Weppner, W. Ceram. Int. 1998, 24(8),
287–293. 571–577.
130. Götlind, H.; Liu, F.; Svensson, J. E.; Halvarsson, M.; 143. Rahmel, A.; Schütze, M.; Quadakkers, W. J. Mater.
Johansson, L. G. Oxid. Met. 2007, 67(5), 251–266. Corros. 1995, 46(5), 271–285.
131. Lodziana, Z.; Topsoe, N. Y.; Norskov, J. K. Nat. Mater. 144. Rahmel, A.; Spencer, P. J. Oxid. Met. 1991, 35(1),
2004, 3, 289–293. 53–68.
132. Götlind, H. Thesis for the degree of doctor of 145. Becker, S.; Rahmel, A.; Schorr, M.; Schütze, M. Oxid.
engineering. Chalmers University of Technology: Met. 1992, 38(5), 425–464.
Göteborg, Sweden, 2614, 2007, ISBN 978-91-7291-933-4.
1.18 Fireside Corrosion
N. Otsuka
Corporate R&D Laboratories, Sumitomo Metal Industries, Ltd., 1-8 Fusocho, Amagasaki, Japan

1.18.1.1 Fuel Chemistry 459
1.18.1.2 Flue Gas Composition 460
1.18.1.3 Combustion Conditions 461
1.18.1.4 Deposit Chemistry 461
1.18.2 Gas-Phase Corrosion 464
1.18.2.1 Oxidizing Conditions 464
1.18.2.2 Reducing Conditions 465
1.18.3 Molten Salt Corrosion in Power Generating Systems 468
1.18.3.1 Vanadium Attack 470
1.18.3.2 Sulfate-Induced Corrosion 472
1.18.3.3 Fireside Corrosion Induced by Chlorine in Fuel 477
References 481
Glossary hemicelluloses. It is fired in boilers to obtain

Acidic dissolution (of metal oxides) A dissolution heat and to recover sodium as sodium
reaction to form acidic solutes such as sulfide and sulfates from it.
Fe2(SO4)3, NiSO4, Cr2(SO4)3. See basic Borio index An index obtainable from coal
dissolution. chemistry. This index is used to
Basic dissolution (of metal oxides) A dissolution estimate the relative severity of high-S coal
reaction to form basic solutes such as regarding the fireside corrosion of metal
NaFeO2, NaNiO2, Na2CrO4. See acidic components.
dissolution. Catastrophic oxidation See hot corrosion.
Basicity (of fused sodium sulfate) Defined as –log Fireside corrosion A type of corrosion
aNa2 O . The basicity of sodium sulfate is encountered for high temperature metal
affected by the SO3 concentration of the components exposed to hot combustion
ambient gas atmosphere. gases.
Biomass Plant matter grown for use as a biofuel. Fluxing of oxide scale A reaction that takes
Biomass is grown from several plants, place when protective oxide scales form on
including swichgrass, hemp, corn, willow, high temperature steels and alloys come in
and sugarcane. Chips of wood are also contact with molten salts. Protective metal
included. oxides dissolve in molten salts and lose
Bitumen A mixture of organic liquids that are highly their protection against corrosive
viscous, black, sticky, entirely soluble in environments.
carbon disulfide, and composed primarily of High temperature hot corrosion (HTHC) See
highly condensed polycyclic aromatic Type I hot corrosion.
hydrocarbons. Hot corrosion An accelerated (or catastrophic)
Bituminous coal A relatively hard coal containing a oxidation of metals and alloys whose surface
tar-like substance called bitumen. It is of is coated with fused salt and/or corrosive
higher quality than lignite coal but of poorer slag films in high temperature gaseous
quality than anthracite coal. environments.
Black liquor A solvent of the cooking process for Kinetic boundary (of Cr2O3 protection) A
Kraft pulping. Black liquor is rich in sodium boundary shown in the PO2–PS2 phase
and organic portions such as lignin and stability diagram of chromium which
457
separates the regime of the uniform Salt-fluxing model A corrosion mechanism based
formation of protective Cr2O3 scale from that on the destructive dissolution of originally
of the formation of fast-growing Cr/Fe sulfide protective metal oxide scale into thin fused
scale on high temperature stainless steels salt film, followed by reprecipitation of these
and alloys. metal oxides as thick, porous, nonprotective
Lignite coal Often referred to as brown coal, reaction products.
which is the lowest rank of coal and used Sulfidation A type of corrosion caused by reaction
almost exclusively as fuel for steam-electric of metal and alloys with sulfur-bearing
power generation. It is brownish-black and species such as gaseous hydrogen sulfide or
has a high inherent moisture content, liquid alkali sulfates. The major reaction
sometimes as high as 66%, and very high products are metal sulfides.
ash content compared to bituminous coal. It Superheater A heat exchanger made of steel tubes
is also a heterogeneous mixture of which are used in steam generating boilers in
compounds for which no single structural order to raise the temperature and pressure
formula will suffice. of steam (to superheat) to be used in steam
Low-temperature hot corrosion (LTHC) See turbines. Their metal temperature is among
Type II hot corrosion. the highest in boiler tubes, and they are
Naphtha A group of various liquid prone to fireside corrosion.
hydrocarbon intermediate-oil-refining Syngas (or synthetic gas) A gas mixture
products used primarily as feedstock for containing CO and H2. Syngas is used as a
producing a high-octane gasoline fuel gas for combustion or as a raw material to
component via the catalytic reforming process. obtain ammonia, methanol, and hydrogen.
Naphtha is also used in the petrochemical Type I (or high temperature) hot corrosion
industry for producing olefins in steam Occurs at a relatively high temperature
crackers and in the chemical industry for (900–1000 C) at which pure Na2SO4 (melting
solvent applications. point 884 C) is liquid.
Oil ash corrosion See vanadium attack. Type II (or low-temperature) hot corrosion
Orimulsion A new type of fossil fuel coming from Occurs at 600–750 C, well below the melting
vast reserves of bitumen in the Orinoco Belt point of pure sodium sulfate (884 C). This
in Venezuela. This fuel is dispersed by water corrosion results from the reaction of
in a condition of emulsion. protective oxide scale with liquid phase of
Oxidizing gas A gas atmosphere containing sodium and potassium complex metal
oxidizing gas species such as O2, CO2, SOx, sulfates.
and H2O. Generally, oxidation of steels and Vanadium attack A type of corrosion caused by the
alloys takes place upon exposure to this reaction of molten vanadate compounds with
environment. protective metal oxide scales. The major
Reheater A heat exchanger made of steel tubes degradation reaction is the dissolution of the
which are used in steam generating boilers to protective metal oxide scales into molten
reheat steam coming from a steam turbine in oxides, and fast transport of the dissolved
order to be used again in a steam turbine. oxygen in molten oxide layer from flue gas to
Their metal temperature is among the metal surface supports the corrosion reaction.
highest in boiler tubes, and they are prone to
fireside corrosion.
Reducing gas A gas atmosphere containing
reducing gas species such as H2, CO, and Abbreviations
hydrocarbon (CH4, etc). For some BLRB Black liquor recovery boiler
combustion gases where sulfur-bearing gas HTHC High temperature hot corrosion
species such as H2S are included, sulfidation LTHC Low temperature hot corrosion
(see sulfidation below) may take place for MSW Municipal solid waste
steels and alloys upon exposure to this gas ppm parts per million
atmosphere. PVC Polyvinyl chloride plastic
Fireside Corrosion 459
type of fuel, combustion conditions, and design fac-

Symbols tors of respective system. In this chapter, fireside
kp Parabolic rate constant (cm2 s1) corrosion is explained from the point of
log aNa2 O Basicity of melt fuel chemistry, and the effect of combustion condi-
PO2 Oxygen partial pressure (atm) tions and design factors on corrosion is emphasized.
Ps2 Sulfur partial pressure (atm) Corrosion resistance of steels and alloys is intro-
duced as well.
1.18.1.1 Fuel Chemistry

1.18.1 Introduction
Examples of elemental analyses of major fuels used
Fireside corrosion is often encountered in metal for generating electricity are shown in Table 1.
components of energy-converting systems which Fuels consist of combustible elements such as car-
are aimed at generating steam and electricity upon bon and hydrogen, with slight addition of impurities
combusting fossil fuel such as coal and oil, biomass, such as nitrogen, sulfur, and chlorine. In this table,
municipal and industrial solid waste. Turbine blades municipal solid waste (MSW) burnt in waste incin-
and vanes, boiler tubes and plate-fin type heat- erators and black liquor fired in Kraft pulping mill
exchangers, hangers and tube supports, etc., which are shown for comparison. From the point of fireside
are exposed to hot combustion gases, are the major corrosion, vanadium, sulfur, and chlorine in fuel are
concerns. Corrosion is normally influenced by the the important impurity elements.1 Alkali metals
Table 1 Elemental analyses of major fuel (wt.%, example)
Element Heavy oil Coal Natural gasa Municipal solid wasteb Black liquor c
Ultimate
C 85 76.2 76.5 22 28.4
H 12 5.13 23.5 8.7 5.2
O 0.35 6.15 – 60.4 44.3
N 0.2 1.63 – 0.3 –
S 2.4 1.73 – 0.02 2.7
Cl – 0.275 – 0.4 3.0
Ash 0.05 8.89 – 8.18 –
Metals
Fe – 1.1 – 0.2 –
V 0.01 – – – –
Ca – 0.1 – 0.9 –
Mg – 0.04 – 0.1 –
Na 50 ppm 0.03 – 0.06 13.0
K 50 ppm 0.03 – 0.04 3.4
Zn – – – 70 ppm –
Pb – – – 10 ppm –
Heat valued 10 400 7509 10 800 2440 3700
a
Liquified natural gas (LNG): CH4–6% C2H6–4% C3H8–2% C4H10.
b
51 wt.% water involved.
c
For closed cycle, S/Na2 = 0.3, 20% water involved.
d
kcal kg1, high-heat value.
Source
Heavy oil: Thermal and Nuclear Power Engineering Society, The Thermal and Nuclear Power, 1988, 39, 1453–1485.
Coal: Blough, J. L.; Stanko, G. J. In Fireside Corrosion testing of Candidate Superheater Tube Alloys, Coatings, and Claddings-Phase II,
Corrosion/97, NACE International: Houston, TX, 1997; Paper no. 140.
MSW: Otsuka, N. In Thermodynamic Equilibrium Calculations of Deposits on Superheater Tubes in Waste Incinerators, Corrosion/2000;
NACE International: Houston, TX, 2000; Paper 00229.
Black Liquor: Pejryd, L.; Hupa, M. In Bed and Furnace Gas Composition in Recovery Boilers-Advanced Equilibrium Calculations,
Proceedings of TAPPI Pulping Conference, 1984; pp 579–590.
Reproduced from Otsuka, N. Corr. Sci. 2002, 44, 265–283.
sodium and potassium, along with alkali earth 1.18.1.2 Flue Gas Composition
metals calcium and magnesium, relate to corrosion
Examples of flue gas composition in boilers firing
as well. In the case of oils used in utility and in-
various types of fuels are presented in Table 2.
dustrial boilers, various types of oils such as crude
These values may vary depending on the location
oil, heavy oil, light oil, and naphtha, or mixtures
of boilers, combustion conditions such as air–fuel
of these are combusted. Similarly, with respect to
ratio as described in the following section, and the
coals, lignite , bituminous , anthracite coals, etc. with
degree of combustion reactions proceeding toward
varying sulfur and chlorine contents, are fired.
the final thermodynamic equilibrium conditions.
Therefore, a large variety does exist in the fuel
Note that the flue gas of heavy oil contains SO2 of
chemistry. For most of the oils fired in boilers, vana-
the order of 0.1%; that of coal has similar SO2
dium is included typically from 0 to as high as
concentration with slight HCl; those of natural gas
300 ppm1 (parts per million by weight). Sulfur con-
and black liquor are virtually free of SO2 and HCl;
tent ranges from 0.1 to 3.2 wt%2, and sodium and
and that of municipal solid waste contains HCl of the
potassium, from 0 to 100 ppm.1 Sulfur concentration
order of 0.1% with traces of SO2.
generally correlates with vanadium content; high-
sulfur oil contains high concentration of vanadium.2
Chlorine content of oil is in most cases low, up to
100 ppm.3 For fuel coal, sulfur and chlorine are the 10 Black liquor
significant impurities relevant to corrosion. Similar to
Cl wt% in fuel
oil, their composition varies depending on their ori- 1 MSW
gin and how they are blended. Sulfur is generally
present from less than 0.5% to 3% (in extreme Coal
cases up to 7%),1 while chlorine, from less than 0.1
0.01% to as much as 0.8%.1 Fuel coal contains Natural gas
sodium from 0 to 700 ppm while potassium from 100 ppm Oil
0 to 2500 ppm.4 Vanadium content of coal is generally
negligible. In contrast, natural gas does not virtually 100 ppm 0.1 1 10
contain impure elements of vanadium, sulfur, and S wt% in fuel
chlorine. This is considered one major reason why Figure 1 Chlorine and sulfur contents of major fuels.
fireside corrosion in gas-fired boilers and turbines is
slight. For municipal solid waste (MSW), almost half
of their weight is water, and they contain chlorine in Table 2 Flue gas composition for complete combustion
(vol.%, example)
the order of 1000 ppm. The sources of chlorine are
inorganic, primarily as NaCl and organic such as Heavy Coal Natural Municipal Black
polyvinyl chloride plastic (PVC). Note that sulfur oil gas solid liquor
concentration of waste is minor. Heavy metals such waste
as lead and zinc in waste are noticed as well. Black O2 3 4.2 3.3 9.0 1.2
liquor is a solvent of cooking process for Kraft pulp- CO2 14.5 13.6 9.1 10.9 14.2
ing and is rich in sodium with organic portion such as H2O 10.5 5.5 16.7 19.5 16.8
lignin and hemicelluloses. This fuel, burnt in boilers SO2 0.1 0.12 – 19 ppm –
HCl – 0.02 – 0.11 4 ppm
to recover sodium and heat, contains large amounts
of sulfur and chlorine. Ranges of chlorine and sulfur Data taken from
contents of various fuels are illustrated in Figure 1. Heavy Oil: Thermal and Nuclear Power Engineering Society,
From the point of chlorine and sulfur concentrations, Therm. Nucl. Power 1988, 39, 1453–1485.
Coal: Calculated for the coal in Table 1 for air–fuel ratio of 13 in
coal and black liquor are categorized as high-chlorine weight at 1300 C.
high-sulfur fuel, oil as high-sulfur low-chlorine fuel, Natural gas: Calculated for the natural gas in Table 1 for air–fuel
municipal solid waste as high-chlorine low-sulfur fuel, ratio of 20 in weight.
MSW: Kawahara, Y.; Hagiwara, H.; Nakamura, M.; Shibuya, E;
and natural gas as low-chlorine low-sulfur fuel. Accord- Yukawa, K. In An Evaluation of Corrosion Resistant Alloys by Field
ing to the literature,5 biomass fuel can be categorized as Corrosion Test in Japanese Refuse Incieration Plants,
low-sulfur medium-chlorine fuel, since their sulfur CORROSION/95; NACE International: Houston, TX, 1995;
Paper no. 564.
content ranges from 0.02% to 0.1%, whereas chlorine, Black Liquor: Calculated for the black liquor in Table 1 for air–fuel
from 0.01% to just below 1%. ratio of 3.77 in weight at 1000 C.
1.18.1.3 Combustion Conditions efficiency, fuel is fired in slightly air-excess conditions.

Therefore, except at locations where equilibrium is
Flue gas composition is significantly affected by the
not established, flue gas generally oxidizes the metal
combustion conditions. Concentration of gaseous spe-
components. Partial combustion is encountered for
cies upon firing 1.7% S coal (chemical composition is
gasification process of coal and residual oil. For these
given in Table 1) was calculated at 1300 C, assuming
cases, flue gas, called synthetic gas (syngas) in some
thermal equilibrium, and presented in Figure 2 as a
cases, is generally reducing, and contains gaseous
function of air–fuel ratio. In this chapter, air–fuel ratio
H2S, resulting from sulfur in fuel. For some industrial
is defined by the ratio of their weights; air–fuel ratio of
boilers, slightly air-deficient combustion is favored at
10 represents combustion of 100 g fuel with 1000 g air.
their furnaces, in order to minimize formation of gas-
Where combustion takes place in complete or in air-
eous nitrogen oxides. Gaseous H2S exists in the flue
excess conditions such as for the air–fuel ratio of more
gases locally. Oxygen and sulfur partial pressures of
than 10, oxidizing gas species such as O2, SO2, and NO
these combustion gases are calculated as a function of
prevail. Complete combustion is termed here when
air–fuel ratio at 1300 C and presented in Table 3. As
carbon and hydrogen in fuel is totally oxidized to
expected, low PO2-high PS2 gas atmospheres are indi-
carbon dioxide and water vapor, and no additional
cated upon combustion at low air–fuel ratio, whereas
oxygen is present in the flue gases. This corresponds
high PO2–low PS2 gas atmospheres are characterized for
to air–fuel ratio of 10 in Figure 2. In contrast, when
combustion at high air–fuel ratio. Therefore, from the
combustion reactions proceed in air-deficient condi-
equilibrium point of view, corrosion environments for
tions where air–fuel ratio is less than 10, reducing gas
air-excess combustion are categorized as ‘oxidizing,’
species such as CO, H2, and H2S replace the oxidizing
while that for partial combustion as ‘sulfidizing’ when
gas species of O2, SO2, and NO. This is termed the
fuel contains considerable amounts of sulfur. The effect
partial combustion. Concentration of CO2 and H2O is
of temperature on flue gas composition upon combust-
not significantly affected by the air–fuel ratio. Carbon
ing 1.7% S coal is basically minor for conditions of air-
dioxide and water vapor can coexist with oxidizing gas
excess combustion (Table 4 for air–fuel ratio of 12),
species of O2 and SO2 and reducing gas species of
but slight changes are noticed for conditions of partial
CO and H2. Normally, from the point of combustion
combustion (Table 5 for air–fuel ratio of 6).
100% 1.18.1.4 Deposit Chemistry

N2
Equilibrium concentration of gaseous species (vol.%)
CO CO2
10% Fireside corrosion of metal components is signifi-
H2O cantly affected, in many cases, by the chemical and
H2 O2 physical properties of deposits piled up on metal
1%
surface from flue gas. Usually, chemical and physical
H2S SO2 properties of these deposits vary drastically,
0.1%
NO
Table 3 Equilibrium oxygen partial pressure PO2 and
100 ppm HCl sulfur partial pressure PS2 of flue gases upon combustion
S2 COS of 1.7% S coal as a function of air–fuel ratio at 1300 C
(calculated)
10 ppm
SO3 Air–fuel ratio PO2 PS2
1 ppm
NH3 5 3 1012 8 105
6 2 1011 1 104
0. 1 ppm 7 8 1011 2 104
8 3 1010 2 104
9 2 109 3 105
6 8 10 12 14 10 5 108 4 108
Air-fuel ratio (by weight) 11 0.01 4 1019
12 0.03 8 1020
Figure 2 Equilibrium concentration of gaseous species in 13 0.04 3 1020
flue gas upon combustion of 1.7% S coal at 1300 C.
Calculated based on the coal chemistry presented in Table 1. Coal composition is taken from Table 1. Air-fuel ratio is defined by
Air–fuel ratio is given by weight; for example, air–fuel ratio of weight; for example, air-fuel ratio of 10 represents combustion of
10 represents combustion of 100 g coal with 1000 g air. 100 g coal with 1000 g air.
Table 4 Equilibrium flue gas composition upon combusting 1.7% S coal in an air-excess condition (vol.%, air–fuel ratio
of 12, calculated)
Temperature N2 O2 CO2 H2O SO3 SO2 NO OH CO H2 PS2a

( C) (ppm) (ppm) (ppm) (ppm) (ppm)
200 76.4 2.8 14.7 5.9 1252 – – – – – 3 1086

400 76.4 2.8 14.7 5.9 1239 0.001 – – – – 6 1056
600 76.4 2.8 14.7 5.9 787 0.04 3 – – – 6 1040
800 76.4 2.8 14.7 5.9 152 0.11 27 – – – 4 1031
1000 76.3 2.9 14.7 5.9 31 0.12 134 4 – – 2 1025
1200 76.3 2.9 14.7 5.9 9 0.12 425 30 3 – 2 1021
1400 76.3 2.9 14.7 5.9 4 0.12 1022 144 43 5 2 1018
a
Partial pressure. Coal composition is taken from Table 1. Air-fuel ratio is defined by weight; for example, air-fuel ratio of 12 represents
combustion of 100 g coal with 1200 g air.
depending on the type of fuel and location of the react primarily with sulfur to form their sulfates,
component in the system. A layered structure is usu- and the remaining react with chlorine to generate
ally observed for these deposits,6,7 with different their chlorides. Hence, upon firing municipal solid
chemistry and porosity in each position. Deposit waste, SOx concentration in flue gas is generally very
chemistry of the innermost layer, which contacts low, since sulfur in waste is totally captured upon
directly the metal to its oxide scale, is considered reaction to form sulfate salts. The important point is
important to examine the corrosion. Examples of that the chemical affinity of sulfur to react with
the deposit chemistry obtained for superheater sodium and potassium is much greater than that of
tubes of boilers firing various fuels are presented in chlorine; for fuel abundant in sulfur, there is no
Table 6. Tube deposit of heavy oil is rich in vanadium chance for chlorine to react with sodium/potassium
and sulfur, and the major constituents are Na2SO4 and to form their chlorides, and for this fuel chloride salts
V2O5 (explained later). Tube deposit of coal comprises are generally not incorporated in the deposits. For
ash constituents of silicon and aluminum oxides, and black liquor, abundant sodium enables reaction with
sulfur compounds such as (Na,K)2SO47 are also indi- sulfur and chlorine in the fuel to form its sulfide and
cated. For those firing municipal solid waste, chlor- chloride. In this case, sulfur and chlorine in the fuel
ides of sodium and potassium with slight inclusion of are completely consumed, and the remaining sodium
lead and zinc salts are characterized. Deposit of black can react with carbon to form its carbonate. There-
liquor consists of mixtures of sodium and potassium fore, in tube deposits of black liquor recovery boilers
sulfates, carbonates, and chlorides. Corrosive com- sulfates, chlorides, and carbonates are present. In this
pounds in these deposits can be predicted to a certain way, one may predict the deposit chemistry, to the
extent, from the fuel chemistry, assuming that the first-order approximation, upon knowing the fuel
corrosive compounds are vanadium oxides, sodium chemistry of the respective system. Critical tempera-
and potassium salts such as sulfates, chlorides, and ture at which the corrosive salts in the deposits start
carbonates, and compounds of heavy metals such as to melt varies with the salt systems. For V2O5–
lead and zinc.8 This is schematically illustrated in Na2SO4 salt mixtures relating to oil deposits, this is
Figure 3. This sketch is constructed with the assump- reported to be 600 C9; for Na2SO4–K2SO4–Fe2O3
tion that combustion terminates at thermodynamic salt mixtures representing coal deposits, it is recog-
equilibrium, and deposition of corrosive salts such as nized as 552 C10; for (Na,K)2SO4–(Na,K)Cl (–(Na,
sulfates and chlorides takes place by vapor condensa- K)2CO3) salt mixtures relevant to black liquor depos-
tion from flue gas on metal surface. For example, in its, it is estimated to be around 500 C11; for sulfates
the case of high-Cl high-S coal, sodium and potas- and chlorides of sodium, potassium, and heavy metals
sium in the fuel do not react with chlorine, but they such as lead and zinc representing municipal solid
preferentially react with sulfur to generate sulfate waste deposits, it can be as low as 300 C.12 Involve-
salts.8 Chlorine in coal reacts to form gaseous HCl. ment of chloride salts of sodium and potassium in the
Therefore, sodium/potassium chlorides are normally deposits generally lowers their melting points, and its
not found in their tube deposits. Similarly, for munic- effect on temperature is drastic for sulfates and chlor-
ipal solid waste, sodium, potassium, lead, and zinc ides of heavy metals such as lead and zinc.
Table 5 Equilibrium flue gas composition upon combusting 1.7% S coal in a partial combustion condition (vol.%, air-fuel ratio of 6, calculated)
Temperature ( C) N2 CO2 CO CH4 H2O H2 H2S SO2 COS NH3 PO2a PS2a
200 73.7 15.3 5 ppm 0.46 10.2 0.08 0.24 – 2 ppm 11 ppm 4 1045 3 1014
400 72.9 16.1 0.35 0.68 7.7 2 0.24 – 26 ppm 34 ppm 3 1032 2 1011
600 67.5 12 10.2 0.23 3 6.8 0.21 – 0.02 23 ppm 3 1025 2 109
800 64.9 7.9 17 12 ppm 3 6.9 0.19 – 0.02 6 ppm 8 1020 2 107
1000 64.9 7 18 – 3.9 6 0.19 – 0.02 2 ppm 1 1015 5 106
1200 64.9 6.4 18.6 – 4.6 5.4 0.17 9 ppm 0.03 – 1 1012 6 105
1400 64.9 5.9 19.1 – 5 4.9 0.13 0.01 0.02 – 2 1010 2 104
a
Partial pressure. Coal composition was taken from Table 1. Air–fuel ratio is defined by weight; for example, air–fuel ratio of 6 represents combustion of 100 g coal with 600 g air.
Fireside Corrosion
463
Table 6 Composition of Superheater Deposits (wt%, example)
Heavy oil Coal Municipal Black liquor

solid waste
Outer layer Intermediate layer Inner layer
Si as SiO2 – 23.5 23.3 7.6 15.3 –

Al as Al2O3 – 14 11.5 1.7 9.5 –
Fe as Fe2O3 11.2 36 11 70.5 1.8 –
Ti as TiO2 – 0.9 < 0.1 < 0.1 1.4 –
Ca as CaO 1.17 1.3 < 0.1 < 0.1 17.4 –
Mg as MgO 0.88 1.3 1.1 < 0.1 2.9 –
Na as Na2O 17.8 0.3 1.7 0.15 8.4 44.3
K as K2O – 2.9 13.5 1.3 12.3 5.2
V as V2O5 49.7 – – – – –
Ni as NiO 2.24 < 0.1 < 0.1 0.3 – –
Cr as Cr2O3 – < 0.1 < 0.1 7.0 1.2 –
Pb as PbO – – – – 0.1 –
Zn as ZnO – – – – 1.0 –
C as CO2 1.6 – – – 0.4 14.3
S as SO3 21.6 7.3 27.5 10.0 15 34.6
Cl – 0.02 < 0.01 < 0.01 9.2 1.5
pH 3.8 3 2.2 4.3 8.2 –
Water soluble % – 9 45.4 9.0 – –
Data taken from

Heavy oil: superheater deposit for 1.6–1.8%S, 130–150 ppm V2O5, from Fukahori, K.; Uera, H.; Harada, Y. The Thermal and Nuclear
Power, 1977, 28, 639–649, Coal: Blazewicz, A.J.; Gold, M. ASME publication, 79-WA/Fu-6, 1980, Municipal solid waste: Otsuka, N.
CORROSION/97, paper no.157, NACE: Houston Texas, 1997, Black liquor: Backman, R.; Skrifvars, B.-J.; Hupa, M.; Siiskonen, P.;
Mäntyniemi, J. J.Pulp and Paper Sci., 1996, 22, J119–J126.
1.18.2 Gas-Phase Corrosion behavior is shown in Figure 4. For steels of more

than 9%Cr, stainless steels, and nickel-base alloys,
In this section, fireside corrosion caused not by slag uniform formation of Cr2O3 scale is established at
deposits but by gaseous species in flue gas is lower temperatures, and their weight gains are slight.
explained. Gas-phase corrosion can be separated Above a critical temperature, which depends on
into two categories, that is, high temperature oxida- respective alloy system, breakaway oxidation results
tion in oxidizing gas atmospheres and sulfidation in for these steels and alloys and Cr2O3 scale loses its
reducing gas atmospheres. protection to high temperature oxidizing gases.14
Temperature limits shown in Table 7 for stainless
steels and nickel-base alloys are determined from the
1.18.2.1 Oxidizing Conditions
onset of breakaway oxidation. Parabolic rate constants
In most energy-converting systems, combustion is of major metal oxide scales are summarized in Figure 5.
conducted under slightly excess-air conditions and Among the metal oxide scales which can uniformly
their flue gases normally contain oxidizing gas species form on metal surfaces, the growth rates of Cr2O3,
such as O2, CO2, H2O, and SOx . From engineering Al2O3, and SiO2 are relatively low, compared to those
point of view, scaling of metal components exposed to of Fe3O4, CoO, and NiO. Therefore, these metal
these oxidizing gas atmospheres can generally be oxide scales can be used as ‘protective’ oxide scales
predicted from the available data reported in air, for steels and alloys as far as breakaway oxidation
since the effects of CO2 and SOx on scaling is in does not take place. The effect of water vapor on
most cases minor.13 Approximate temperature limits oxidation of engineering steels and alloys forming
of steels and alloys considered to be applicable for Cr2O3 scale must be pointed out. As shown in Figure 6,
these oxidizing gas conditions are summarized and water vapor in flue gas atmosphere can significantly
shown in Table 7. For carbon steel and low-alloyed enhance the breakaway oxidation, especially of stainless
steels of up to 5% Cr, exposing these steels to flue steels.14 For some stainless steels, oxidation experiments
gases of higher temperature results in heavy scaling, conducted in laboratory air during hot, wet season (e.g.,
and metal loss due to scaling increases exponentially PH2O of 0.1 atm at 40 C with 100% humidity) can
upon raising the oxidation temperature. This be different from those performed in cold, dry air
Elements Salts, gas species
Elements Salts, gas species Na, K Na2SO4, K2SO4
V V2O5
Na,K Na2SO4, K2SO4
S
SOx(g)
S
SOx(g)
Cl HCl(g)
(a) Heavy oil (b) Coal
S Na2SO4, (K2SO4)
Na Cl NaCl, (KCl)
Na2SO4, K2SO4
(K)
PbSO4, ZnSO4

S Na2CO3
Na, K
NaCl, KCl (K2CO3)
Pb, Zn
Cl PbCl2 ZnCl2
HCl(g)
(c) Municipal solid waste (d) Black liquor

Figure 3 A simplified relationship between the elements in fuel and corrosive salts in tube deposits.
(e.g., PH2O of 0.006 atm at 0 C with 100% humidity). cases, reducing gas species of H2 and CO coexist
Partial pressure of water vapor as a function of dew with CO2 and H2O, in flue gases and H2S is
point of flue gas can be found in the literature.15 Since incorporated when fuel contains sulfur.17 For high
flue gases resulting from combustion of natural gas temperature components exposed to these gas mix-
contain greater amount of water vapor, its effect on tures, sulfidation of steels and alloys occurs. For
oxidation of steels and alloys should be examined care- reducing conditions, gas atmospheres are normally
fully. Oxidation rates of boiler tube materials exposed high in PS2 and low in PO2, and simultaneous forma-
to 100% steam, often used to estimate the metal temp- tion of metal sulfides and oxides takes place. Since
eratures for ferritic steels, are summarized in the the driving force for the formation of metal oxide
literature.16 scale is small, uniform formation of protective metal
oxide scales is relatively difficult to be established. In
these cases, formation of metal sulfides overwhelms
1.18.2.2 Reducing Conditions
that of metal oxides, since the growth rate of metal
For gasifiers of coal, residual-oil, etc., combustion is sulfide scales is generally much higher than those
regulated under air-deficient conditions. In these of metal oxide scales.18,19 Corrosion rates are
Table 7 Approximate temperature limits ( C) in air
Steel Composition (wt%) Fontana & Green Morris Kane
Intermittent service Continuous service
Plain carbon 0.1C 480 – 510

T12 1Cr–0.5Mo – – 565
T22 2.25Cr–1Mo – – 580
T5 5Cr–0.5Mo 620 – 620
T9 9Cr–1Mo 670 – 650
AISI type 410 11Cr 760 815 705 700
AISI type 430 17Cr 840 870 815 840
AISI type 442 21Cr 950 1035 980 –
AISI type 446 25Cr 1030 1175 1095 –
AISI type 304, 321, 347 18Cr–8Ni–(Ti,Nb) 900 870 925 900
AISI type 316 18Cr–10Ni–2Mo 900 870 925 –
AISI type 309 23Cr–12Ni 1090 980 1095 1040
AISI type 310 25Cr–20Ni 1150 1035 1150 1090
Hastelloy X Ni-base superalloy 1200 – –
Hastelloy C Ni-base superalloy 1150 – –
Cr 900 – –
Ni 780 – –
Cu 450 – –
Brass 70Cu–30Zn 700 – –
Temperature below which oxidation rate in negligible. Negligible is defined as less than about 0.31 mg cm2 h1 (for weight gain, Fontana &
Greene). For Morris and Kane, criteria is not presented.
Source: Fontana, M. G.; Greene, N. D. Corrosion Engineering; McGraw-Hill, 1967; p 369.
Morris, L. A. Met. Eng. Q. Am. Soc. Met. 1968, 8, 30–47.
Kane, R. H. In Process Industries Corrosion; Moniz, B. J., Pollock, W. I., Eds.; NACE: Houston, TX, 1986; pp 45–65.
Temperature (⬚C)
600 700 800 900
3500
Carbon steel, C-0.50Mo,
3000 2.25Cr-0.5 Mo, 2.25Cr-1 Mo,
1.75Cr-0.75 Mo-0.75Si
Mo
2500
Gain in weight (g m−2)
5Cr-0.5 Mo, 5Cr-0.5 Mo-Ti

0.5
5Cr-0.5 Mo-Cb
Cr-
12
2000
r-Al
12C
1500
Si
.5
1
o
1000
o-
1M
3Cr-0.5 Mo-1.5Si Cr
.5
-
17
9Cr
0
r–
5C
500
18.8Cb 18.8-Mo
18-8
{ 18.8-Ti
27Cr
0
900 1000 1100 1200 1300 1400 1500 1600 1700 1800
Temperature (⬚F)
Figure 4 Oxidation behavior of engineering steels and alloys exposed to air for 1000 h. Reproduced from Kane, R. H.
In Process Industries Corrosion – The Theory and Practice, Moniz, B. J., Pollock, W. I., Eds.; NACE: Houston, TX, 1986;
pp 45–65.
1400 ⬚C 1200 ⬚C 1000 ⬚C 800 ⬚C 700 ⬚C 600 ⬚C 500 ⬚C
FeO in O2
–6 Davies, Simnad, Birchenall (1951)
FeO in air
Paidassi
(1958)
CoO in air
Mrowec and
–8 Przybylski (1977)
NiO in O2
Fueki and
Wagner
(1965) Fe3O4 in air
log kp (cm2 s–1)
Paidassi (1958)
Davies et al. (1951)
– 10
Fe3O4 in O2
Al2O3 Fe2CrO4 in CO2

Compiled by Smith (1981)
– 12 Hindam and Whittle (1983)
SiO2
Cr2O3
Compiled by Bastow,
Compiled by Whittle and Wood (1978)
– 14 Kircher and Courtright (1992)
6 8 10 12
1/ T ⫻ 104 (K–1)
Figure 5 Parabolic rate constants (in scale thickness) of major oxide scales. Data were taken from Davies, M. H.;
Simnad, M. T.; Birchenall, C. E. J. Met. 1951, 889–896; Paidaiis, J. Revu Metall. 1957, LIV, 569–585; Mrowec, S.;
Przybylski, K. Oxid. Met. 1977, 11, 365–381; Fueki, K.; Wagner, J. B. J. Electrochem. Soc. 1965, 112, 384–388; Smith, A. F.
Werkst. Korros. 1979, 30, 100–104; Hindam, H.; Whittle, D. P. Oxid. Met. 1982, 18, 245–284; Bastow, B. D.; Whittle, D. P.;
Wood, G. C. Oxid. Met. 1978, 12, 413–438; Kircher, T. A.; Courtright, E. L. Mater. Sci. Eng. A 1992, 155, 67–74.
influenced not by slow-growing metal oxide scales H2S–H2 gas mixture of naphtha desulfurizers. These
but by fast-growing metal sulfide scales. Therefore, data represent exclusive formation and growth of
sulfidation can cause serious corrosion damage for sulfide scales where formation of protective oxide
metal components. In order to estimate the sulfidation scales is minimized. For certain flue gas system con-
rates of industrial steels and alloys, data taken from taining not only H2S and H2 but also some oxidizing
simple binary H2S/H2 gas mixtures are often gas species of CO2 and H2O, formation of the protec-
referred,20 since these data are considered to indicate tive Cr2O3 scale may result for stainless steels and
the maximum corrosion rates of steels and alloys in nickel-base alloys, if thermodynamic condition favors.
respective PS2 gas atmospheres. Iso-corrosion rate Once formation of uniform Cr2O3 scale is achieved,
curves of engineering steels are presented in Figure 7 sulfidation by H2S is hindered, and steels and alloys
as a function of temperature and H2S concentration in are protected by the uniform Cr2O3 scale. In coal
gasifiers, PS2–PO2 boundary of uniform formation of

Cr2O3 scale on steels and alloys deviates from that
predicted from the thermodynamics. This boundary
is referred as the kinetic boundary of Cr2O3 scale,
shown in Figure 8 for example. Higher oxygen poten-
tial (or lower sulfur potential) is needed to secure the
uniform formation of Cr2O3 scale, since greater
driving force is needed for the growth of Cr2O3 to
compete with that of the fast-growing metal sulfides.
Oxidized in100% steam 1.18.3 Molten Salt Corrosion in

Power Generating Systems
In this section, fireside corrosion caused predomi-

nantly by molten salts is outlined. High temperature
metal components subject to fireside corrosion in
energy-converting systems are exposed to hot com-
bustion gases, and their metal temperatures are nor-
mally lower than the ambient flue gases, especially
20 μm for heat exchanger materials placed in hot combustion
gases to obtain heat from them. Depending on fuel
Oxidized in laboratory air chemistry, sodium/potassium sulfates and their chlor-
Figure 6 Cross section of TP321H steel (18% Cr–8%
ides, vanadium oxides, and chlorides and sulfates of
Ni–0.4% Ti) reacted with 100% steam (top) and laboratory heavy metals such as lead and zinc, which are present
air (bottom) at 650 C for 1000 h under atmospheric as vapor phases in flue gases, are considered as corro-
condition. sive salts. They tend to vapor-condense on ‘cold’ metal
surface from surrounding hot gas atmospheres. These
vapor-condensed salts concentrate at the innermost
metal-side of the deposit layer, since these sites are
10 the coldest location in the deposits and hence the
18-8 driving force for vapor-condensation of these salts is
H2S in H2S/H2 gas mixture (Mole %)
Corrosion considered the greatest. For conditions where these

rate of
1 more than salts become molten, reaction of these salts with pro-
0.1 mm/y tective metal oxide scales may initiate and fluxing
12Cr
reaction would take place, resulting in degradation of
these protective metal oxide scales.6,21 Metal oxide
0.1 9Cr-1Mo
scales, initially formed as protective oxide layer on
Carbon metal surfaces by gas-phase reaction, react with fused
steel salts and dissolve into the salts. These dissolved metal
ions precipitate out as oxide particles at the gas-side of
0.01
5Cr-0.5Mo the salt layer. Hence, once fluxing of metal oxide scale
initiates, protective metal oxide scale is converted into
No corrosion
porous, nonprotective oxide particles. At the reaction
0.001 front, oxidation of metal constituents takes place in
200 300 400 500 much higher rates than the case where the formation
Metal temperature (°C)
of protective oxide scales is established. At these sites,
Figure 7 Isocorrosion rate curves of engineering steels for sulfidation as well as chlorination may simultaneously
high temperature H2S/H2 corrosion in naptha desulfurizers. occur depending on the chemistry of molten salts.
Corrosion rate of 0.1 mm year1. Adapted from Gutzeit, J.
In Process Industries Corrosion – The Theory and Practice;
Since the final product of this corrosion is generally
Moniz, B. J., Pollock, W. I., Eds.; NACE: Houston TX, 1986; metal oxides, the overall reaction of molten salt corro-
pp 367–372. sion can be interpreted as ‘oxidation’ of metals. This is
–6 Thermodynamic
boundary
Cr3S4 + FeS + Ni3S2
Kinetic
boundary
–8
Cr3S4 + FeS
log Ps2 (atm)
–10
Cr3S4 Scale
Cr2O3 Cr/Fe sulfide
Sulfide/oxide
Cr-rich oxide
–12
Incoloy 800
CrS 1023 K (750 ⬚C)
Alloy
–14
–30 –25 –20 –15
log Po2 (atm)
Figure 8 Kinetic boundary of Cr2O3 protection found for Alloy 800 as a function of oxygen and sulfur partial pressures
in sulfidizing gas atmospheres at 750 C. Reproduced from Natesan, K. In High Temperature Corrosion, 2–6 March 1981,
San Diego, CA; Rapp, R. A., Ed.; NACE: Houston, TX, 1981; pp 336–344.
one reason why the molten salt corrosion is termed as

‘accelerated oxidation’ or ‘catastrophic oxidation.’ 6 In
Corrosion thickness loss
Molten salts in
molten salt corrosion, severe corrosion does not take tube deposits
place when the surrounding gas atmosphere does not
contain any oxygen, even though sufficient amounts of
molten salts are present on metal surfaces. Generally,
Salts in tube deposits
oxygen is the predominant oxidant for fireside corro- Incubation do not melt
sion, while carbon dioxide and water vapor are not that time
influencing. The role of molten salts is interpreted to
enable fluxing of metal oxide scales,21 and the ability
to facilitate fast transportation of oxygen in molten salt
Corrosion time
layer to reaction sites is another important property of
molten salts to maintain the high reaction rates of this Figure 9 Schematic corrosion behavior caused by
fused salts.
corrosion.9 A schematic behavior of molten salt corro-
sion is presented in Figure 9. Frequently, incubation
time is experienced at the beginning of the corrosion,6 great importance to know the melting point of the
where the corrosion rates remain relatively low. In this deposits and to compare it with the metal temperature
period, formation of protective oxide scales and con- of the component. Finally, the importance of flue gas
densation of corrosive salts are considered to take temperature on corrosion should be pointed out. It is
place. For conditions where salts become fused, that generally accepted that the corrosion rate of this type
is, metal temperatures exceed that of the melting depends not only on metal temperature, but also on
points of the salt mixture, accelerated oxidation takes flue gas temperature to which the metal components
place, and fireside corrosion initiates. When the metal are exposed. Exposing metal components to flue gas of
temperatures remain lower than the melting point of higher temperature generally result in higher corro-
the salt mixture, accelerated oxidation do not take sion rates.22,23 This is interpreted from greater con-
place, and corrosion remains slight. Therefore, it is of centration of corrosive salts in the deposits, since the
flue gas of higher temperature normally contains 1285⬚

higher concentration of vapor species of corrosive
salts, and the driving force for vapor-condensation of 1200
these salts onto metal surface is apparently greater
than those exposed to flue gas of low temperatures, Liquid
which is attributed to the temperature difference 2Na2O
–V2O5
between the flue gas and the base metal.8 1000
Temperature (⬚C)
Na2O-V2O4 Na2O-V2O5
–5V2O5
1.18.3.1 Vanadium Attack
Vanadium attack, referred as oil ash corrosion as well, 800 4Na2O-V2O4
has been recognized as severe fireside corrosion for –11V2O5
700⬚
boiler tubes, turbine components, furnace tubes, and 675
653⬚
reformer tubes in boilers, combustors, and reformers 650⬚ 626⬚ 3Na2O
(5%)
which fire oil containing vanadium.3,9 Oil, rich in 600 580⬚
–V2O5
530⬚ 540⬚
Na4V2O7
vanadium and sodium of approximately more than
Na8V24O63
(59%)
NaV6O15
Na3VO4
(40%)
1–10 ppm, is recognized to have potential corrosivity
NaVO3
to metals.24 Vanadium reacts with oxygen in combus- 426⬚
400
tion air to form V2O3, V2O4, and finally volatile V2O5
in flue gases.2 Gaseous V2O5 vapor condenses on 0 20 40 60 80 100
metal surface as ash deposits. Depending on the depo- Mol% Na2O
V2O5
sition rates of V2O5 and of other vapor-condensed
Figure 10 Melting point of Na2O–V2O5 system.
species such as Na2SO4, etc., the deposit chemistry Reproduced from Roth, R. S.; Dennis, J. R.; McMurdie,
varies, but vanadium (as oxide) can constitute up to H. F. Phase Diagrams for Ceramists; The American Ceramic
80% of the deposits although vanadium concentra- Society, 1987; Vol. VI, p 98.
tion in fuel is typically less than 300 ppm.3 In many
cases, vanadium oxides react with other constituents
such as sodium to form low melting point compounds. for V2O5–20 mol% Na2SO4 salt mixture.9 This
This is shown in Figure 10. Note that for Na2O–V2O5 was explained from the ability for vanadium pentox-
system, melting points can be as low as 530–540 C, ide–alkali salt mixtures to absorb oxygen. Indeed,
which corresponds to the metal temperature of super- vanadium pentoxide–alkali salt mixtures absorb
heater and reheater tubes in practical boilers.3 In oxygen upon heating from solid to liquid phase
addition to vanadium pentoxide, sodium vanadates through the reversible change from vanadylvanadate
of Na2O–V2O4–5V2O5, 4Na2O–V2O4–11V2O5, to alkali acid vanadates, according to the following
Na2O–V2O5, and 3Na2O–V2O5 can be involved equation9
in superheater and reheater deposits. Melting points
mNa2 O ðn pÞV2 O5 pV2 O4 þ 1=2pO2 ¼ mNa2 O nV2 O5
of superheater and reheater deposits taken from
350–500 MW boilers firing oil of 30–40 ppm V, Capability to accommodate great amount of oxygen
8–15 ppm Na, and 1–1.5% S are plotted as a function in the melt and ability for fast diffusion of oxygen
of (Na þ S)/V atomic ratio of their deposits and shown through its layer to metal surface seem to be the
in Figure 11.25 Clearly, vanadium-rich deposits start to characteristics of this liquid phase.
melt at 480–500 C, whereas those lean in vanadium The effect of metal temperature on corrosion is
remain solid until heated to above 800 C. Sulfate-rich significant. In order to minimize the corrosion, limiting
deposits have normally high melting points. metal temperature to below the critical temperature,
Vanadium attack is interpreted to result from which relates to the melting points of the ash deposits,
fluxing of protective metal oxide scales and fast trans- is considered an effective measure to alleviate the
fer of dissolved oxygen in molten oxide layer.9 corrosion.3 The effect of flue gas temperature on
A laboratory corrosion test revealed that accelerated corrosion is reported to be important as well. When
corrosion of TP304 stainless steel occurred when placing heat-exchanger tubes and plates to flue gases
reacted with pure molten V2O5, but the corrosion of higher temperature, vapor-deposition of corrosive
rate even increased when sodium sulfate was added vanadium and sodium compounds enhances, and
to vanadium pentoxide; maximum corrosion resulted concentration of these compounds on metal surface
S and V (as V2O5) weight % in heavy oil
S : 2.4 ∼ 2.5% S : 1.6 ∼ 1.8%, V: 130 ∼ 150 ppm
V : 200 ∼250 ppm S : 2.7 ∼ 2.8%

S : 0.2 ∼ 0.3%
V : 45 ∼ 65 ppm
V : 1 ∼ 3 ppm
900
800
Temperature (⬚C)
700
600
500 Fuel
S : 1.0 ∼ 1.5%
Na : 8 ∼ 15 ppm
400 V : 30 ∼ 40 ppm
0.1 1 10 100
(Na + S)/V atomic ratio of tube deposits
Figure 11 Melting point of tube deposits taken from superheater and reheater tubes of 350–500 MW boilers firing heavy oil.
Steam temperatures were 540–571 C. Reproduced from Harada, Y.; Nakamori, S. Boushoku Gijutsu 1980, 29, 615–621.
increases. This facilitates formation of molten salts at ⬚C F

these sites, and the corrosion aggravates accordingly. 2600
1400
For boiler tubes, regions of gas and metal tempera-
tures where corrosion may occur are empirically
2400
recognized in practice. This is presented in Figure 12. 1300
Corrosive zone
This criterion depends on several factors such as fuel
temperature
chemistry, boiler design, and combustion conditions, 1200 2200

Gas
but in order to increase metal temperature, the lower-

ing of the flue gas temperature is generally needed. 1100 2000
Lowering excess air used for combustion of oil should
have a certain effect to reduce the corrosion rate, 1000
since removing ‘free’ oxygen in the flue gas obviously 1800 Stable
would help in minimizing supply of major oxidizing zone
gas species of the corrosion reaction. However, using 900
1600
low excess combustion air is reported not always 950 1050 1150 1250
successful when applied to practical power plants, Metal temperature, F
because of the complexity of combustion and flue 550 600 650 ⬚C
gas mixing in real boilers.3 The effect of inhibitors
Figure 12 Regions where boiler tubes are subject to
and fuel additives on corrosion has been widely fireside corrosion as a function of tube metal temperature
recognized, and addition of inhibitors to fuel oil is and flue gas temperature. Reproduced from Hansen, W. H.;
successfully applied to some boilers in practice.25 Kessler, G. W. Trans. ASME 1965, 210–214.
These inhibitors are designed to react with vanadium in Venezuela has become recognized as an useful fuel
to form compounds of high melting points. Out of a to generate electricity.28 This fuel, arising from the
number of inhibitors, MgO and its derivatives are technology for dispersing natural bitumen in water,
used in practice because of its effectiveness and rela- is called orimulsion. Orimulsion contains water of
tively low cost.3,25 For example, addition of water- around 30 wt%. An example of the elemental analysis
soluble Mg(OH)2 to fuel oil, at an atomic Mg/V is 60.1% carbon, 10.1% hydrogen, 26.4% oxygen,
ratio of 2, is reported to have been successful 0.35% nitrogen, 2.85% sulfur, and 0.2% ash (by
in reducing the corrosion rate of TP321H (18% weight).28 It should be noted that this fuel contains
Cr–10% Ni–0.4% Ti) superheater tubes in a prac- vanadium of 310 ppm and sodium of 30 ppm. Vana-
tical boiler to about one-third, the steam tempera- dium attack on metal components is predicted. Field
ture of which was 571 C.25 tests were conducted in real boilers firing orimulsion,
According to Figure 13, high-chromium steels and fireside corrosion of boiler materials turned out
and alloys seem to have better resistance to vanadium to be not that severe. This was attributed to high
attack. Nickel-based alloys have generally higher cor- concentration of magnesium in the fuel, that is,
rosion rates than the iron-based alloys, suggesting that 370 ppm, which acted as inhibitor to raise the melting
nickel is detrimental compared to iron. For boiler tube points of the ash deposits, and hence alleviating cor-
application, high-Cr ferritic steels of 9% Cr and 12% rosion for metal components. Information on the
Cr are reported to have better resistance than the concentration not only of corrosive impurities but
austenitic 18-8 stainless steels.26,27 This is interpreted also of other impurities such as alkali earth elements
to be attributed, again, to nickel in 18-8 steels, since is needed to evaluate the corrosivity of fuel.
nickel is a harmful alloying element to combat
the corrosion, the reason for which still seems to
1.18.3.2 Sulfate-Induced Corrosion
remain unclear.
Recently, a new type of fossil fuel coming from Fireside corrosion has often been encountered in
vast reserves of bitumen existing in the Orinoco Belt blades and vanes of land-based and aircraft gas
(mm year–1)
25.4 1000
Relationship between
Na and V content in fuel
oil to hot ash Fe based alloys
corrosion rates Fe and Ni based alloys
(5)
2.54 100 (4) 50 Cr/50 Ni alloy

(2) (3)
0.254 10 (1)
Fe based alloys
Ni based alloys
50 Cr/50 Ni alloy
0.0254 1.0
0.1 1.0 10 100 1000
Sodium and vanadium in fuel oil (ppm)
Figure 13 Relationship between Na + V content in fuel oil to corrosion rates. Reproduced from Demo, J. J. Mater. Perform.
1980, 9–15. Data of iron-based alloys are from boiler and furnace/reformer tubes, those of nickel-base alloys predominantly
from turbine blades, and those of 50Cr–50Ni alloy from tube supports and hangers. Solid marks with numbers represent
industrial and semiworks experience points of iron-based reformer alloys.
turbines because of thin film of pure liquid sodium Recommended

sulfate, which is vapor-condensed from flue gas onto maximum tube
temperature
metal surface.6,21 The source of sodium comes from
either the fuel or the air intake. Sodium, commonly
present in sea salt as sodium chloride, reacts with 0.5 1400 ⬚C
sulfur in fuel to form sodium sulfate, according to
the well-known reaction6:
2NaCl þ SO3 þ 1=2O2 ¼ Na2 SO4 þ Cl2 0.4 640 700
Metal loss (mm year–1)

‘Bell shaped’ form of
corrosion rate/temperature 1200 ⬚C
At turbine conditions, sodium chloride is considered relation from probe trials
thermodynamically unstable. Metal temperatures of 0.3

the components are generally above the melting
point of sodium sulfate, that is, 884 C. This is often
1000 ⬚C
termed as Type I hot corrosion or high temperature 0.2
hot corrosion (HTHC).21 For utility boilers firing
high-sulfur coal, fireside corrosion of superheater
and reheater tubes can take place. This is categorized 0.1
800 ⬚C
as Type II hot corrosion, or low-temperature hot Gas
corrosion (LTHC), since this corrosion occurs in temperature
the temperature range of 600–750 C, well below 0
the melting point of pure sodium sulfate.21,29 In this 580 600 620 640
section, fireside corrosion of Type II hot corrosion is Temperature of tube surface (⬚C)
introduced firstly. Figure 14 An example of the corrosion rates of
It is well recognized that for boilers firing pulver- superheaters and reheaters as a function of tube surface
temperature and flue gas temperature. The inset shows the
ized coal of high sulfur concentration, external cor-
‘bell-shaped’ corrosion behavior of this type of corrosion.
rosion of superheater and reheater tubes takes Reproduced from Cutler, A. J. B.; Flatley, T.; Hay, K. A.
place.7,10,29,30,31 The corrosion rate of boiler tubes Metallurgist and Materials Technologist, 1981,
starts to significantly increase at around 565 C February, 69–81.
(1050 F) when the metal temperature exceeds the
melting point of their tube deposits, exhibits a maxi-
mum at around 650–700 C, and decreases drastically trisulfate, Na3Fe(SO4)3 becomes fused at 624 C,
upon further increase in the metal temperature.7,10 whereas potassium iron trisulfate, K3Fe(SO4)3, at
The drastic decrease in the corrosion rate at above 618 C.10 The melting point of these complex sulfates
700 C is interpreted to be from the solidification of is affected by the sodium–potassium ratio, and the
liquid phases, the temperature of which depends lowest melting-point 552 C is obtained for com-
upon the partial pressure of SO3 in flue gas and the pounds between 1:1and 2:1 molar mixture of potas-
chemistry of the complex sulfates.32 Hence, the cor- sium to sodium iron trisulfates.10 Hence, for most
rosion behavior is characterized as ‘bell-shaped’ laboratory tests simulating fireside corrosion of
form.7,10 Corrosion rate of tube steels is influenced Type II attack, synthetic ash of Na1.5K1.5Fe(SO4)3
not only by the metal temperature, but also by the (1.5 mol Na2SO4–1.5 mol K2SO4–1 mol Fe2O3 mix-
flue gas temperature, suggesting a great effect of ture) is used as predominant corrosives to simulate
vapor-condensation of corrosive salts from flue gas this environment.33 It should be noted that these
on metal surface.23 This is shown in Figure 14. Tube sodium and potassium complex iron trisulfates are
deposits contain corrosive salts of sodium and potas- stable only in gas atmospheres containing SO2. Tem-
sium sulfates and iron oxide (hematite, Fe2O3), in perature range for the liquid-phase stability of these
addition to ash constituents such as complex oxides complex sulfates seems to be affected by SO2 (and
of aluminum, silicon, etc. Sodium and potassium therefore SO3) concentration in the flue gases. For
sulfates can form complex sulfates of low melting- example, temperature regime where severe corrosion
points upon reaction with hematite, that is, sodium occurs for TP321H steel (18% Cr–10% Ni–0.4%
and potassium iron trisulfates of (Na,K)3Fe(SO4)3. Ti) becomes broader when exposed to gas atmo-
Under a simulated flue gas atmosphere of 0.25% spheres of higher SO2 concentration. This is shown
SO2–3.6% O2–15.0% CO2–balanceN2, sodium iron in Figure 15.33 For flue gas of low SO2 concentration,
and contents at least four times with small amounts of

850.0
water, and make it up to 100 ml volume. This solution
should be carefully measured.’4 As expected, corrosivity
Corrosion temperature (⬚C)
800.0 4 4
150 mg/cm2 of high-S coal is related to the concentration of acid-
125
750.0 2 3 2 117 151 soluble sodium and potassium compounds, hematite
144 content, and calcium and magnesium contents. To
700.0 121
3 6 13 107 116 combat the corrosion, high-Cr steels and alloys perform
133 100 mg/cm2
well, which will be explained later in detail.
650.0 5 8 50 56 53 51
The mechanism of Type II hot corrosion seems to
48
600.0
resemble that of Type I hot corrosion. Therefore,
4 41
10 mg/cm2 corrosion reaction offered for Type I hot corrosion
550.0 4 3 is explained. For Type I hot corrosion observed at
turbine blades and vanes, corrosion products consist
0.05 0.1 0.2 0.5 1 2 5
generally of relatively thin outer oxide layer, a thick
SO2 concentration (vol.%) mixed oxide/metal layer, and a thin metal zone con-
Figure 15 The effect of SO2 vol.% in gas atmosphere on taining sulfides below the mixed oxide/metal layer.6
high temperature corrosion of TP321H steel specimen Sulfides are generally rich in chromium, but in some
reacted with 1.5 mol Na2SO4–1.5 mol K2SO4–1 mol Fe2O3 in cases nickel-rich or cobalt-rich sulfides are found
x% SO2–5% O2–15% CO2–N2. Numbers in the figure for materials which experienced rapid degradation,
designate corrosion weight loss. Reproduced from Fukuda,
S. T.; Hamada, H.; Sakaguchi, Y. In Proceedings of the 32th
presumably, due to liquid formation of Ni–Ni3S2 or
Meeting of the Japanese Society of Corrosion Engineering, Co–Co4S3 eutectics.6 The eutectic temperatures
1985, August, Sapporo, B-302, pp 239–246. are 645 and 877 C, for Ni–Ni3S2 and Co–Co4S3
eutectics respectively.34 Early interpretation of this
corrosion was derived to explain the formation of
sulfates of potassium and sodium do not seem to react sulfides in the base metal and called the sulfidation
with iron oxides to form stable complex sodium and model.34 According to this model, corrosion can be
potassium iron trisulfates, and these ash deposits divided into two stages. At a triggering stage, which is
remain solid at tube metal temperatures. This result ‘sporadic and unpredictable,’ reduction of sodium
clearly demonstrates the importance of SO2 (and sulfate by an unspecified reducing agent generates
therefore SO3) concentration in flue gases, since one elemental sulfur, which reacts with base metal compo-
can predict the occurrence/absence of fireside cor- nents to form their sulfides. Hence, for this model,
rosion of superheater and reheater tubes of coal-fired oxidizing agent is considered sodium sulfate. These
boilers, when the SO2 concentration of flue gas metal sulfides react with base metal elements to form
is specified. Therefore, for firing low-S coal where low melting point eutectics such as Ni–Ni3S2. At an
the SO2 concentration of flue gas is less than 0.1%, enhanced oxidation stage, metal is penetrated by the
sulfate-induced corrosion is expected to be slight. metal–metal sulfide liquid eutectics, and oxidation of
There is an index representing the corrosivity of these liquid eutectics by oxygen was proposed to pro-
high-S coal of 1.4–8.2% S (proximate, by weight) ceed more rapidly than the sulfur-free alloy. Forma-
from the coal chemistry. This index is often referred tion and enhanced oxidation of metal–metal sulfide
to as the Borio index, determined from the concentra- eutectics was the early mechanism. Since chromium
tion of acid-soluble alkali and alkali earth elements (in sulfides are normally present in the corrosion products
ppm), and Fe2O3 (in wt%) of coal.4 Procedure to of the degraded turbine blades and vanes, and metal–
obtain this index is shown in Figure 16. In order to metal sulfide eutectics such as Ni–Ni3S2, etc. are not
obtain acid-soluble sodium and potassium concen- always indicated,6 more generalized interpretation
trations in coal, precise measurement seems to be seems to be necessary. Chromium does not form any
necessary. Test procedure for this is described in the metal–metal sulfide eutectic. Recent understanding of
literature as follows,4 ‘set 2 g portion of coal sample this type of corrosion is based on destructive dissolu-
(70% through 200 mesh) to a 250-m L Erlenmeyer tion of originally protective metal oxide scale into thin
flask and add 40 m L of 5% HCl solution. Fix air fused salt film, followed by reprecipitation of these
condenser and digest the sample for 16 h at a gentle metal oxides as thick, porous, nonprotective reaction
boil. After digestion, filter through #40 Whatman filter products.21 This is called the salt-fluxing model.
paper of 15 cm into 100-ml volumetric flask. Wash filter The overall reaction is again accelerated/catastrophic
0 0.5 1.0 1.5 2.0 2.53.5 5.0 12.0
900
)
al basis 800
pm co
Na 2O (p
oluble 750
Acid s 700
sis)
650
l ba
600 coa
550 (%
O 3
500 Fe 2
400
0-300
0 5 8 9 10 11 12 13 14 15 16 17 18 19 20
Corrosion
Acid soluble K2O (expressed as Na2O eq., ppm x 10–2 coal basis) index
22
0.0
0.4
20
0% corrosion reduction
0.8
n
tio
1.2
18 s)
uc
basi 1.6
ed
l
ion 16 coa 2.0
nr
u ct eq.% 2.4
sio
d n O 2.8
re tio 14 Ca
uc as
rro
3.2
n n d
io ed ctio sse 3.6
nr
co
s u re
rro s io red 12
exp 4.0
O(
%
co rro on
co osi g
25
% orr 10 O+M
50 % c Ca
75 0%
10 8
4
Example (Coal F)
Acid sol. K2O (expressed as Na2O eq.) 1410 ppm
2 Acid sol. Na2O 540 ppm
CaO + MgO (expressed as CaO eq.) 0.50%
0 Fe2O3 1.9%
0 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00 3.25 3.50 3.75 4.00 4.25 4.50 Quantity of CaO eq. required to reduce
CaO equivalent (%, coal basis) Corrosion rate by 100% (theor) 1.60%
(all values are measured on coal basis)
Fireside Corrosion
Relationship of coal
constituents to high temperature corrosion* *[1] Probe temperatures maintained at average of 1100 ⬚F
[2] Probe metal - 321 55
Figure VII-13. Relationship of cool constituents to high temperature corrosion VII-29

(a) (b) (c)
Figure 16 Relation of coal constituents to high temperature corrosion of boiler tubes. Reproduced from Borio, R. W.; Wilson, E. B. The control of high temperature fire-side
corrosion in utility coal-fired boilers, OCR R6D Report No. 41, April 1969, Office of Coal Research, Department of the Interior, Washington, DC.
475
oxidation of metals, and deterioration of protective In order to establish self-sustaining hot corrosion of
metal oxide scale through dissolution/fluxing reac- steels and alloys, it seems necessary for fused sodium
tions is an important step for initiating and maintain- sulfate to locate its melt chemistry at a specific con-
ing the corrosion. Oxygen in flue gas is necessary to dition where simultaneous dissolution and reprecipi-
maintain the corrosion; without oxygen, reaction can- tation of metal oxide scales are sustained. The
not proceed even when abundant fused salts are pres- basicity at which self-sustaining hot corrosion can
ent on metal surfaces. The fluxing reactions of metal occur may depend upon the alloy system; for pure
oxide scales are significantly affected by the ‘basicity’ nickel, this corresponds to the basicity at which basic
of fused salts, which is defined as log aNa2O for pure dissolution of NiO is secured.35 One explanation for
sodium sulfate.21 Basicity of sulfate melts is affected by it may be provided by the solubility gradient model,
the activity of SO3 in flue gas, according to the follow- which can be found in the literature.36 For fused salt
ing reaction: film of sodium sulfate, maintaining its melt basicity in
basic conditions is considered rather difficult since
Na2 SO4 ¼ Na2 O ðin the meltÞ þ SO3 ðgÞ
the melt chemistry enabling NiO basic dissolution is
This is similar to the thermal stability of the complex quite far from the melt conditions equilibrated with
sodium/potassium iron trisulfates where the stability acidic flue gas atmosphere containing SO3. In acidic
of its liquid phase is influenced by SO3 concentra- melts, pure nickel, if protected by dense metal oxide
tion.33 It should be pointed out that basicity of fused scales, did not produce any rapid attack.35 Current
sulfate salts is influenced by the partial pressure not of understanding for this is to incorporate the effect of
SO2, but of SO3. Generally, SO3 concentration of flue sulfidation reaction of the base metal by fused sodium
gas is approximately two orders of magnitude less than sulfate, which generates O2 ion and shifts its basicity
SO2 (see Table 4, for example). to more basic conditions.21 For pure nickel, the sulfi-
Solubility of major metal oxides in fused sodium dation reaction is given by the following equation:
sulfate at 927 C is presented in Figure 17.21 Clearly,
4 þ 4Ni ¼ 3NiO þ NiS þ O
SO2 2
solubility of metal oxides depends on the basicity of
sulfate melt, and two types of dissolution reactions Direct contact of fused salts with the base metal,
are noticed, except for SiO2. The one is categorized presumably at cracks, flaws, and grain boundaries of
as the acidic dissolution, which takes place in an its protective oxide scale results in altering the basic-
acidic melt lean in O2 ion.21 For NiO, for example, ity to more basic, where self-sustaining corrosion is
the reaction is written as NiO ¼ Ni2+ þ O2. The facilitated. Since formation of protective oxide scales
other one is termed as the basic dissolution, which is considered to precede vapor-condensation of mol-
occurs in a basic melt rich in O2 ion. The reaction ten salts, ‘incubation time’ is usually required for
is designated as 2NiO þ O2 þ 1/2O2 ¼ 2NiO 2. fused salts to permeate the protective oxide scale37
and to arrive at the point where direct contact of
fused salt and base metal becomes possible. For acidic
Log conc. ppm (mole metal ions/mole Na2SO4)
melts, oxide solubility is the lowest for Cr2O3 and

4
Al2O3, as shown in Figure 17. Therefore, these metal
Cr2O3 oxide scales, along with SiO2, is expected to function
Fe2O3 as protective oxide scales in acidic melts. This may be
3 one reason why the steels and alloys forming uniform
Cr2O3, Al2O3, and SiO2 scales have good resistance to
NiO hot corrosion. On the contrary, molybdenum, tung-
2 Co3O4 sten, and vanadium, commonly added to strengthen
Al2O3 their alloys, are regarded as harmful alloying ele-
ments for hot corrosion.38 Oxides of these alloying
SiO2 (900 ⬚C) elements can be categorized as ‘acidic’ oxides, which
1
are thermodynamically stable in acidic melts. Ther-
5 6 7 8 9 10 11 12 13 14 15 16 17 modynamic stability of major metal oxides relevant to
Basic –Log aNa O Acidic alloy constituents of high temperature materials
2
Figure 17 Oxide solubility of major oxides in fused

is calculated at 927 C, PO2 of 1 atm and shown in
Na2SO4 at 927 C and 1 atm O2. Reproduced from Figure 18. In this figure, unit activities of condensed
Rapp, R. A. Corrosion 1986, 42, 568–577. phases in mutual equilibrium are assumed. Oxides of
1200 K PO2 = 1 atm Minimum

solubility 40
Corrosion weight loss (mg cm–2)

Base elements 17–14 CuMo
NaFeO2 Fe2O3 Fe2(SO4)3 321H
NaNiO2 NiO NiSO4 30 AN31
Esshete 12R72V
NaCoO2 Co3O4 CoSO4 –1250
TP347H
Na2CrO4 Cr2O3 CrSO4 SUS347H
Protective 20 15–15N 617
elements Na2SiO3
Na2
Si2O5 SiO2 HR11N
625
TP347HFG
NaAlO2 Al2O3 Al2(SO4)3 HK4M
10 800H SZ 35Cr–54Ni-Nb
NaVO3 V2O5(l) VOSO4
Addition 807 IN–671
elements Na2MoO4 MoO3(I) 310S Chromized
Na2Ti3O7 TP310HCbN
Na2WO4 WO3
Na2Ti2O5 10 20 30 40 50
Na2TiO3 TiO2 Cr (%)
CuO CuO·CuSO4 Figure 19 Laboratory corrosion test results of high
MnO MnSO4 temperature steels and alloys. The test was conducted at
650 C for 5 h in 1% SO2–5% O2–15% CO2–N2 Test
Fused salts MgO MgSO4
specimens were reacted with 1.5 mol Na2SO4–1.5 mol
constituents CaO CaSO4 K2SO4–1 mol Fe2O3 salt mixture by salt-coating method.
Na2O2 Na2SO4
corrosion resistance, whereas with niobium the oppo-
K2O K2SO4
site occurs. The beneficial effect of niobium on cor-
rosion seems to remain unclarified, although recent
0 10 20 modern 18-8 stainless steel tubes contain small
Basic –log aNa2O Acidic
amount of niobium.
Figure 18 Basicity of major oxides at 1200 K, PO2 ¼ 1 atm This explanation is valid for fireside corrosion
relevant to high temperature corrosion. encountered in components exposed to environments
firing fossil fuel containing sulfur. In general cases,
molybdenum, tungsten, and vanadium are stable in SO2 concentration in these flue gases lies in the range
acidic melts, while oxides of sodium and potassium of less than several thousand parts per million, and
are stable in basic melts. In ‘neutral’ melts, basic disso- SO3 concentration in the range of less than 100 ppm.
lution may take place for acidic oxides; described for However, for extreme cases such as boilers firing solid
molybdenum oxide as follows: sulfur, SO3 concentration of flue gas may surpass
hundreds of parts per million, and fused-salt corro-
MoO3 ðsÞ þ O2 ¼ MoO2
4
sion attributed to liquid sodium/potassium pyrosul-
This reaction shifts the basicity of molten salts to more fates of Na2S2O7 and K2S2O7 can take place.39 Since
acidic. At the reaction front of metal/salt interface of the melting point of Na2SO4–Na2S2O7 and K2SO4–
nickel-base alloys containing molybdenum and the K2S2O7 systems is reported to be low, down to around
like, oxidation and subsequent dissolution of these 400 C (Figure 20), fireside corrosion resulting from
alloying elements may take place simultaneously and the formation of liquid phase of sodium and potassium
a very acidic condition can be established locally, pyrosulfates can occur for boiler tubes. Again, it must
which forces acidic fluxing of otherwise protective be pointed out that these liquid phases are only stable
Cr2O3 and Al2O3 scales. This may be one explanation for conditions where SO3 (not SO2) concentration in
for the detrimental effect of molybdenum and tung- flue gas exceeds 200 ppm for potassium pyrosulfate
sten on corrosion. and 2000 ppm for sodium pyrosulfate, which is very
For Type II hot corrosion, the beneficial effect rare for boilers firing conventional fuel.
of chromium on corrosion is prominent as well, as
shown in Figure 19 for example. High-Cr alloys such
1.18.3.3 Fireside Corrosion Induced by
as alloy 671 are used to combat fireside corrosion in
Chlorine in Fuel
these environments. It should be pointed out that for
18-8 stainless steel family additional elements of Fireside corrosion relating to chlorine in fuel is encoun-
molybdenum and titanium to steels degrade their tered for boiler components in waste incinerators and
500
Solid phase
Sodium system
Liquid phase
490 Solid phase Potassium system
Liquid phase
Potassium
480
470
460
Solid
Temperature (⬚C)
450 K2SO4
Sodium
440
430 Solid
Liquid
K2SO4 Na2SO4
420 Liquid
–K2S2O7 Na2SO4.Na2S2O7
410
Solid K2S2O7
400
Solid Na2S2O7
390
20 200 2000
p.p.m.
102 103 104

Concentration of SO3 (ppm)
Figure 20 Temperature range of liquid K2S2O7–K2SO4 and Na2S2O7–Na2SO4 in the presence of various partial pressures
of SO3. Reproduced from Coats, A. W.; Dear, D. J.; Penfold, D. J. Inst. Fuel 1968, 129–132.
in black liquor recovery boilers (BLRB). It is gener- in waste incinerators is presented in Figure 21.41
ally accepted that chloride salts, deposited on metal Since hydrogen chloride is always present in flue
surface via vapor-condensation, play an important gases in the order of hundreds and thousands of
role in corrosion.40 In order to deposit chloride salts parts per million, the early interpretation of corro-
from flue gas on metal surface, chloride vapor species sion was the gas-phase attack of metals by hydrogen
must be present in flue gas atmospheres. chloride.42 Since one of the reaction products can be
Fireside corrosion of waste incinerators has been highly volatile iron trichloride (Fe2Cl6, boiling point
intensively studied for steel tubes of their heat rec- 315 C, see Figure 22), corrosion products are
overy boilers.40–42 Except for aqueous corrosion expected not to function as corrosion barrier for
experienced at below the dew-point temperatures, boiler tubes, and high corrosion rates can be expected
corrosion can start at above 320 C, which is indeed even at low temperatures. However, laboratory tests
surprisingly low compared to other types of fireside examining gas-phase corrosion of tube steels by
corrosion encountered in real systems. A schematic hydrogen chloride turned out that its corrosion rate
temperature dependence of boiler tubes on corrosion is significantly low at 400 C, which cannot explain
Corrosion caused by
molten salts of sulfates
Dew point
corrosion
Corrosion caused by
molten salts of chlorides
Corrosion rate
involving heavy metals
320 480 Corrosion by gas
0 100 200 300 400 500 600 700 800

↑
Dew point
Tube wall temperature (°C)
Figure 21 Corrosion behavior of carbon steel boiler tube in waste incinerator as a function of tube wall temperature
(schematic). Adapted from Von Fa¨ßler, K.; Leib, H.; Spa¨hn, H. Mitteilungen der VGB 1968, 48, 126–139.
760 may corrode steels through the formation of sodium

117 ⬚C 315 ⬚C CrCl3 and potassium pyrosulfates ((Na,K)2S2O7) of low
CrO2Cl2 melting points, which react with tube steels (and
Vapor pressure (mmHg)
NiCl2 their protective oxide scales) to form liquid alkali

FeCl3
iron trisulfates ((Na,K)3Fe(SO4)3) at below 500 C.41
100
Gas-phase corrosion by chlorine, and corrosion by
liquid phases are the early mechanisms.40,41 Accord-
ing to this mechanism, corrosion of carbon steel
boiler tubes by chlorine and sulfur occurs simulta-
FeCl2 CrCl2 neously. Corrosion products of these reactions are the
corresponding iron chlorides (FeCl2, Fe2Cl6), iron
oxides (Fe3O4, Fe2O3), and iron sulfide (FeS) for
10
200 400 600 800 1000 steel boiler tubes.40
Temperature (⬚C) According to the Deacon reaction, increasing cor-
Figure 22 Vapor pressure of Fe, Cr, and Ni chlorides. rosion temperature should result in lowering the
Data taken from Kubaschewski, O.; Alcock, C.B. partial pressure of Cl2, and hence would lead to a
Metallurgial Thermochemistry, 5th ed.; Pergamon Press, decrease in the corrosion rate of tube materials. How-
1979.
ever, in real furnaces, severe fireside corrosion of
boiler tubes normally occurs when metal tempera-
the corrosion occurring in real boilers.43 Therefore, ture of tube steels exceeds certain critical tempera-
gas-phase corrosion by chlorine was proposed instead ture, which depends on several factors such as tube
as the root cause for the corrosion since the corrosion location, operational conditions, boiler design, waste
rate of tube steel by chlorine gas is appreciably higher chemistry, etc.44 This is apparently inconsistent with
than hydrogen chloride.41 Gaseous chlorine is never the model based on the Deacon reaction. In addition,
found in furnace flue gases. Therefore, local forma- in order to stabilize molten phases of pyrosulfate
tion of chlorine gas in tube deposits through the salts, high concentration of gaseous SO3 (of more
following Deacon reaction: than several hundreds parts per million) is needed
in the ambient flue gases (see Figure 20), which is
4HCl þ O2 ¼ 2Cl2 þ 2H2 O
apparently not the case in real waste incinerators.
was suggested.41 Tube deposits were explained to The concentration of SO3 in flue gas of waste incin-
function as catalyst of the Deacon reaction. Sulfur erators is normally below ppm level. Extensive field
tests conducted in the late 1990s revealed that fire- lower the melting points of the deposits, but may also
side corrosion of boiler tubes becomes significant for augment the volume fraction of molten phases at a
the area where their metal temperatures exceed the certain temperature, resulting in an increased corro-
melting point of tube deposits.12,46 Most slag deposits sion rate for boiler tubes.
contain not only sodium and potassium chlorides, but Generally, among iron, chromium, and nickel, iron
also chlorides and sulfates of heavy metals such as and chromium have poor resistant to high temperature
lead, zinc, tin, which can lower the melting point chlorine environments. This is interpreted from the
of tube deposits, down to below 300 C.12 Therefore, ability to form highly volatile chlorides at low tempera-
the corrosion can be interpreted as molten salt corro- tures, for example, iron trichloride for iron and chro-
sion as well. In many cases, molten phase is indicated mium oxychloride for chromium. This is shown in
in tube deposits where fireside corrosion is promi- Figure 22. Corrosion resistance of steels and alloys in
nent. Similar to fused sodium sulfate, acid–base equi- waste incinerators seems to follow this trend, and high-
librium for chloride melts with hydrogen chloride Ni alloys have normally good resistance to the environ-
and water vapor in flue gas is proposed,46 and solu- ment. However, according to the field tests, chromium
bility measurements of major metal oxides in eutectic seems to be effective as well, which is different from the
NaCl–KCl and NaCl–KCl–Na2SO4–K2SO4 melts as argument of chloride volatility. Nickel-base alloys with
a function of the melt ‘basicity’ were performed high chromium seem to perform well. It is interesting
experimentally.47 Since tube deposits in waste incin- to note that the effect of molybdenum on corrosion
erators generally contain considerable amounts of is beneficial, which is opposite to those experienced
solid ash constituents such as CaSO4, Al2O3, and for sulfate-induced corrosion. Corrosion resistance of
SiO2, which do not melt at tube metal temperatures, steels and alloys seems to increase with increasing their
tube deposits are expected to be in partial fusion Cr þ Mo þ Ni concentration, shown in Figure 24 for
when corrosion reaction takes place. Laboratory test example. High-Cr, high-Ni, high-Mo alloys such as
revealed that the corrosion rate of tube materials alloy 625 (22% Cr–61% Ni–9% Mo) are recognized
drastically increases when the tube deposits start to to be resistant in this environment.
melt, and the enhanced corrosion rate is influenced
by the volume faction of fused salts in tube deposits.48 550 ⬚C/6000 h Gas flow
T22 347H
For more molten phases in the deposit, higher corro- 7
HR3C
Maximum corrosion thickness loss (mm 6000 h–1)
sion rate is expected. This is schematically illustrated 1st row

in Figure 23. Therefore, enrichment of heavy metals 6
825 JHN24
such as lead and zinc in tube deposits will not only HR11N
625M*
625
5 2nd row
HC-22
C-276 * * Mark:
4 Weld
overlay
3 Plant A
Corrosion rate
1st 2nd
2 Plant B
1st 2nd
1 Plant C
1st 2nd
–
0
0 20 40 60 80 100
[Cr+Ni+Mo] concentration (wt%)
Boiler tube deposits Figure 24 Tube wall thickness loss of corrosion probes
as a function of alloy Cr þ Mo þ Ni concentration after
Volume fraction of fused salts in tube deposits
exposure at 550 C for 6000 h in waste incinerators.
at corrosion temperature
Reproduced from Kawahara, Y.; Orita, N.; Nakamura, M.;
Figure 23 A schematic illustration explaining the effect of Ayukawa, S.; Hosoda, In Corrosion 97; NACE: Houston, TX,
volume fraction of fused salts on corrosion. 1999; paper no. 91.
For black liquor recovery boilers, chloride salts are 23. Cutler, A. J. B.; Flatley, T.; Hay, K. A. Metallurgist and
Materials Technologist, 1981, February, pp 69–81.
incorporated in superheater and reheater tube depos- 24. Demo, J. J. Mater. Perform. 1980, 19, 9–15.
its, similar to those in waste incinerators.49 However, 25. Harada, Y.; Nakamori, M. Boushoku Gijutsu 1980, 29,
their melting points are generally above 500 C, 615–621.
26. Alexander, P. A.; Marsden, R. A.; Nelson-Allen, J. M.;
considerably higher than the waste incinerators. Stewart, W. A. J. Inst. Fuel 1964, 37, 59–69.
This is attributed to the absence of heavy metals 27. Holland, N. H. J. Inst. Fuel 1970, 43, 97–103.
such as zinc and lead in the fuel. For black recovery 28. Marruffo, F.; Chirinos, M. L.; Sarmiento, W.; Hernandez-
Carstents, E. In 17th Congress of the World Energy
boilers, stress corrosion cracking of their floor tubes Council, Division II, 13–18 September; World Energy
seems to be the major concern.50 Council: Houston, TX, 1988; Vol. 73, pp 73–87.
29. Plumley, A. L.; Jonakin, J.; Vuia, R. E. A review study of
fire-side corrosion in utility and industrial boilers,
Presented at Corrosion Seminar at MacMaster University
References and Engineering Institute of Canada, Hamilton, Ontario,
19–20 May 1966.
1. Stringer, J. Mater. Perform. 1993, 53–59. 30. Blazewicz, A. J.; Gold, M. An ASME Publication 79-WA/
2. Fukahori, K.; Uera, H.; Harada, Y. Therm. Nucl. Power Fu-6, The American Society of Mechanical Engineers,
1977, 28, 33–43. 1979.
3. Paul, L. D.; Seeley, R. R. In Corrosion/90; NACE: Houston, 31. Blough, J. L.; Stanko, G. J.; Bakker, W. T. In
TX, 1990; Paper no. 267. Ultra-Supercritical Power Plants, CORROSION/2000,
4. Borio, R. W.; Wilson, E. B. The control of high-temperature NACE: Houston, TX, 2000, Paper no. 250.
fire-side corrosion in utility coal-fired boilers, OCR R6D 32. Luthra, K. L.; Shores, D. A. J. Electrochem. Soc. 1980,
Report No.41, April 1969, Office of Coal Research, 127(10), 2202–2210.
Department of the Interior: Washington, DC. 33. Fukuda, Y.; Sato, T.; Hamada, H.; Sakaguchi, Y.
5. Simms, N. J.; Kilgallon, P. J.; Oakey, J. E. In In Proceedings of the 32th Meeting of the Japanese
Materials for Advanced Power Engineering 2006, Society of Corrosion Engineering, August 1985, Sapporo,
Proceedings of the 8th Liège Conference Part III, B-302, pp 239–246.
Forschungszentrum Jülich GmbH Zentralbibliothek, 34. Simons, E. L.; Browning, G. Y.; Liebhafsky, H. A. Corrosion
Verlag, 2006; pp 1503–1512. 1955, 11, 505t–514t.
6. Stringer, J. Ann. Rev. Mater. Sci. 1977, 7, 477–509. 35. Otsuka, N.; Rapp, R. A. J. Electrochem. Soc. 1990, 137,
7. Nelson, W.; Cain, C., Jr. Trans. ASME, Series A 1960, 82, 46–52.
194–204. 36. Rapp, R. A.; Goto, K. S. In Proceedings of the 2nd
8. Otsuka, N. Corros. Sci. 2002, 44, 265–283. International Symposium on Molten Salts; Braunstein, J.,
9. Cunningham, G. W.; Brasunas, A. Corrosion 1956, 12, Ed.; The Electrochemical Society: Pennington, NJ, 1979;
389t–405t. pp 159–177.
10. Cain, C.; Nelson, W. Trans. ASME, Series A 1961, 83, 37. Pettit, F. S.; Meier, G. H. In Oxidation and Corrosion of
468–474. Superalloys; Superalloys 85, The Metallurgical Society of
11. Levin, E. M.; Robbins, C. R.; McMurdie, H. F. Phase AIME: Warrendale, PA, 1985; pp 651–687.
Diagrams for Ceramists; The American Ceramic Society, 38. Goebel, J. A.; Pettit, F. S.; Goward, G. W. Met. Trans.
1964; Vol. I, p 495. 1973, 4, 261–278.
12. Daniel, P. L.; Paul, L. D.; Barna, J. In Corrosion/1987; 39. Coats, A. W.; Dear, D. J. A.; Penfold, D. J. Instit. Fuel 1968,
NACE: Houston, TX, 1987; Paper no. 400. 41, 129–132.
13. Edström, J. O. JISI 1957, 185, 450–466. 40. Krause, H. H. J. Mater. Energy Syst. 1986, 7(4), 322–332.
14. Kofstad, P. High Temperature Corrosion; Elsevier Applied 41. Von Fäßler, K.; Leib, H.; Spähn, H. Mitteilungen der VGB
Science, 1988. 1968, 48(2), 126–139.
15. Weast, R. C.; Astle, M. J.; Beyer, W. H. CRC Handbook of 42. Von Huch, R. Bresnnst.-Wärme-Kraft 1966, 18, 76–79.
Chemistry and Physics, 68th ed., CRC Press: Boca Raton, 43. Brown, M. H.; DeLong, W. B.; Auld, J. R. Ind. Eng. Chem.
FL, 1987; pp D189–D190. 1947, 39, 839–844.
16. Armitt, J.; Holmes, R.; Manning, M. I.; Meadowcroft, D. B.; 44. Wright, I. G.; Krause, H. H.; Dooley, R. B. In Corrosion/95;
Metcalfe, E. The spalling of steam-grown oxide from NACE: Houston, TX, 1995; Paper no. 562.
superheater and reheater tube steels, EPRI FP-686, TPS 45. Krause, H. H.; Wright, I. G. In Corrosion/95; NACE:
76-655, Final Report, February 1978. Houston, TX, 1995; Paper no. 561.
17. Bakker, W. T. Mixed oxidant corrosion in nonequilibrium 46. Otsuka, N.; Kudo, T. In High Temperature Corrosion of
syngas at 540 C EPRI TR-104228, March 1995. Advanced Materials and Protective Coatings; Saito, Y.,
18. Mrowec, S.; Werber, T. In Corrosion of Metals; Önay, B., Maruyama, T., Eds.; North-Holland: Tokyo,
Department of Commerce, National Technical Information Japan, 1992; pp 205–211.
Service, 1978; Chapter 13, pp 444–474. 47. Ishitsuka, T.; Nose, K. Mater. Corros. 2000, 51, 177–181.
19. Young, D. J. Rev. High-Temp. Mater. 1980, 4(4), 299–346. 48. Otsuka, N.; Tsukaue, Y.; Nakagawa, K.; Kawahara, Y.;
20. Gutzeit, J. In Process Industries Corrosion – The Theory Yukawa, K. In Corrosion/97; NACE: Houston, TX, 1997;
and Practice; Moniz, B. J., Pollock, W. I., Eds.; NACE: paper no.157.
Houston, TX, 1986; pp 367–372. 49. Tran, H. N.; Barham, D.; Hupa, M. Mater. Perform. 1988,
21. Rapp, R. A. Corrosion 1986, 42, 568–577. 27(7), 40–45.
22. Hansen, W. A.; Kessler, G. W. Trans. ASME 1965, 87, 50. Barna, J. L.; Mattie, R. J.; Rogan, J. B.; Allison, S. F. Pulp
210–214. Paper 1989, 63(6), 90–98.
1.19 High Temperature Corrosion Issues for Metallic
Materials in Solid Oxide Fuel Cells
L. Singheiser, P. Huczkowski, T. Markus, and W. J. Quadakkers

1.19.2 Metallic Materials for High Temperature Applications 484
1.19.3 Chromium-Base Metallic Interconnectors 485
1.19.4 Volatile Cr Species 485
1.19.5 Behavior of Cr-Based Alloys in H2/H2O-Based Gases 488
1.19.6 Mixed Gas Corrosion of Cr-Based Alloys 489
1.19.7 Oxidation Rates of Chromium-Based Alloys 490
1.19.8 Electronic Conductivity of Chromia-Based Scales 492
1.19.9 High-Chromium Ferritic Steels 492
1.19.10 Behavior of High-Cr Ferritic Steels in Anode Gas 494
1.19.11 Ferritic Steels in Anode Gases Containing Carbonaceous Gas Species 497
1.19.12 Ferritic Steels Designed for SOFC Application 499
1.19.13 Effect of Component Thickness on Oxidation Behavior 502
1.19.14 Oxidation Behavior of Ferritic Steels Under Dual Atmosphere Conditions 507
1.19.15 Interaction of Chromia-Forming Interconnects with Cathode Side Materials 507
1.19.16 Protection Methods to Minimize Cr Vaporization 509
1.19.17 Interaction of Chromia-Forming Interconnects with Anode Side Materials 510
1.19.18 Interaction of Metallic Interconnects with Glass Sealants 512
References 514
Kp0 Apparent oxidation rate constant (g2 cm4 s1)

Symbols L Characteristic length (cm)
a Activity
M Molecular weight of volatile species (g mol1)
CB Critical chromium concentration in the
n Oxidation rate exponent
component (wt%)
p Vapor pressure bar (Pa)
DV F Volume change associated with the formation
P Partial pressure of volatile species (atm)
of the relevant point defect (cm3)
r Grain size (mm)
Cv Vacancy concentration ((z) cm3)
R Gas constant ((cm3 atm) K1 mol1)
Cvo Equilibrium concentration of vacancies in the
Re Reynold’s number
unstressed state ((z) cm3)
Sc Schmidt number
C0 Initial chromium concentration in the
t Time (h, s)
component (wt%)
tB Time to breakaway (h)
d Component thickness (mm)
T Temperature (K, C)
D Diffusivity (m2 s1)
x Oxide scale thickness (cm, mm)
D Interdiffusion coefficient of volatile species in
d Grain boundary width (nm)
laminar boundary layer (cm2 s1)
r Alloy density (mg cm3)
Do Oxygen grain boundary diffusion
s Electrical conductivity (S cm1)
coefficient (m2 s1)
sh Hydrostatic component of the stress tensor (Pa)
J Mass flux (mg cm2 h1)
sxx Biaxial compressive stress (Pa)
k Oxidation rate constant (mg cm2 hn)
Dm Oxygen uptake per unit area
k0 Parabolic oxidation rate constant (cm2 s1)
(Weight change) (mg cm2)
K Equilibrium constant of the reaction *
Dm Critical weight change at which the onset of
KP Parabolic oxidation rate constant (mg2 cm4 s1)
scale spallation occurs (mg cm2)
482
High Temperature Corrosion Issues for Metallic Materials in Solid Oxide Fuel Cells 483
DH Enthalpy (kJ mol1) vibration, and therefore, the noise pollution commonly
DH* Activation enthalpy (kJ mol1) associated with power generation is eliminated.
DS* Activation entropy (kJ mol1 K1) In an SOFC system, the single cell is constructed
DV* Activation volume (m3 mol1) with an electrolyte arranged between two porous elec-
Dmo Oxygen – potential gradient across the scale trodes (Figure 1). When the oxygen in the flowing
(J mol1) air at the cathode side contacts the cathode–electrolyte
interface, it catalytically acquires electrons from the
cathode and splits into oxygen ions. The oxygen ions
diffuse into the electrolyte material and migrate to the
1.19.1 Introduction anode side of the cell where they react with the fuel at
the anode–electrolyte interface thereby giving off
A solid oxide fuel cell (SOFC) is an electrochemical water (and/or carbon dioxide if a CO-containing fuel
device that converts the chemical energy in fuels into is used), heat, and electrons. The electrons are trans-
electrical energy by utilizing the natural tendency ported through the anode to the external circuit and
of oxygen and hydrogen to react.1 Fuel cells are able back to the cathode, providing a source of useful elec-
to convert hydrogen-based fuels with higher effi- trical energy in an external circuit.1
ciency compared to engines and thermal power plants. Two possible design configurations for SOFCs
Compared with other fuel cell systems, the main have been considered, they are: the tubular and the
advantage of the SOFC is its ability to use not only planar design, the latter drawing more attention
hydrogen but also the currently available fossil fuels because it allows higher power densities than the
(such as methane, butane, or even gasoline and diesel), tubular configuration. In the planar design, the com-
thus reducing operating costs and flexibility. SOFCs ponents are assembled in flat stacks, with air and fuel
possess no moving parts, so the cells are free of flowing through channels, in most cases built into the
Interconnect
Anode
Electrolyte
Fuel Cathode
Interconnect
Air Anode
Fuel
(a)
Interconnect
Ni-wire mesh
Anode
Electrolyte
Cathode
Contact layer
Interconnect
(b)
Figure 1 Schematic configuration of a planar SOFC stack (a) and schematic cross-section showing means of contacting
between interconnect and electrodes (b).
interconnect (Figure 1). In an SOFC stack, the inter- gas-proof separation between fuel gas and oxidant.
connect thus provides the separation of the gas atmo- Besides, it distributes the gases in co-, cross, and/or
spheres, acts as current collector, and provides the counter flow and acts as the constructional connection
electrical connection between the various single to the external inlets and outlets. Therefore, the use of
cells.1 In most planar designs, two additional compo- metals has a number of significant advantages over
nents are present to improve the electrical contact La-chromite-based ceramics,6–9 because they are eas-
between the interconnect and the anode and cathode, ier, and therefore cheaper to fabricate than ceramics,
respectively (Figure 1): they are less brittle, easier to machine, and they can be
joined with a number of standard welding and brazing
a Ni-wire mesh at the anode side,
techniques. Additionally, they possess higher electri-
a ceramic contact layer with high electronic con-
cal and thermal conductivities than most ceramics.
ductivity at the cathode side.
The most important properties required for the inter-
connector material are high electronic conductivity, 1.19.2 Metallic Materials for
thermal stability in the cathode and anode side gas at High Temperature Applications
the high-service temperatures (650–1000 C), and a
thermal expansion coefficient (Table 1) similar to When using metallic interconnect materials, it has
that of the ceramic, electro-active components. Most to be considered that the reaction with the anode
designs use yttria-stabilized zirconia (YSZ) as electro- and cathode side service environments at the high
lyte, (La,Sr)MnO3 as cathode, and Ni/ZrO2-cermet operating temperatures may not only lead to dimen-
as anode.2,3 Ceramic, perowskite type materials on the sional changes and loss in load-bearing cross sections
basis of La-chromite have been shown to possess the of the components but also to the formation of oxide
property combination required for SOFC inter- scales on the component surface10,11 which mostly
connects.3–5 However, in planar cell designs, the possess low electrical conductivities. The use of
interconnector also acts as the mechanical support noble metals could avoid these problems, however,
for the thin electroactive ceramic parts and it is the because of the limited availability and high cost, this
solution has been abandoned for large scale applica-
tion, and conventional high temperature alloys12,13
Table 1 Thermal expansion coefficients (between RT have received by far the most attention as possible
and 1000 C) of ceramic materials used in SOFCs com- candidate metallic interconnector materials.
pared with that of a Cr-based alloy and a high-Cr ferritic From the viewpoint of oxidation resistance, alloys
steel
of the type NiCrAl, CoCrAl, and especially FeCrAl
Material TEC (106 K1) would be the materials to be chosen as construction
materials for interconnects.10 These types of high
8YSZ- 10.5
Ni-YSZ-Cermet 11–12
temperature alloys are able to form extremely slowly
(La,Sr)MnO3 11–14 growing alumina scales on the material surface upon
LaCrO3 9.5 high temperature exposure. Alumina-forming inter-
(La,Sr)CrO3 11 metallics on the basis of NiAl have been considered
LaCoO3 20–23 as interconnector materials14 because they not only
(La,Sr)CoO3 18–23
(La,Ca)(Cr,Co)O3 9.5–23
possess superior oxidation resistance but also the
Cr-based ODS alloy 9–10 required low thermal expansion coefficient. However,
Ferritic steel Crofer 22 APU 12 like all alumina-forming metallic materials, they have
a major drawback; that is, the surface oxide scales pos-
Source: Minh, N. Q. J. Am. Ceram. Soc. 1993, 176(3), 563–588. sess an extremely low electronic conductivity15–17
Steele, B. C. H. Solid State Ionics 2000, 134, 3–20.
Singhal, S. C. In Solid Oxide Fuel Cell (SOFC-V); Stimming, U., and the use of such materials would thus require
Singhal, S. C., Tagawa, H., Lehnert, W., Eds.; measures to overcome this problem by developing
The Electrochemical Society Proceedings Series, Pennington, NJ, special stack designs to assure suitable, long-term
1997; p 88, PV 97-40.
Stolten, D.; Späh, R.; Schamm, R.; In Solid Oxide Fuel Cell stable electrical connections of the interconnector
(SOFC-V); Stimming, U., Singhal, S. C., Tagawa, H., Lehnert, W., with anode and cathode.
Eds.; The Electrochemical Society Proceedings Series, The second type of surface oxide scale which is
Pennington, NJ, 1997; p 88, PV 97-40.
Quadakkers, W. J.; Piron-Abellan, J.; Shemet, V.; Singheiser, L. potentially suitable to provide oxidation/corrosion
Mater. High Temp. 2003, 20(2), 115–127. protection to high temperature alloys is silica.
However, the formation and long-term stability of a been attributed to oxide formation within the scale
protective silica scale requires quite a large amount of resulting from the parallel diffusion of chromium and
silicon additions in nickel-, cobalt-, or iron-based oxygen along chromia grain boundaries.11,25
alloys. This leads, in nearly all cases, to substantial An improvement in scale adherence and reduction
embrittlement, thus making the alloys unsuitable as in scale growth rate can be obtained by the addition
construction materials. Besides, silica has a very low of the so-called reactive elements (REs) such as Y, La,
electronic conductivity, leading to similar problems Ce, and Zr, either in metallic form or in the form of
in SOFC application as mentioned above for alumina an oxide dispersion.28–31 This RE-imparted change of
surface scales. the most important oxide scale properties is fre-
Most commercially available high temperature quently claimed to be related to a suppression of
alloys rely on the formation of chromia-based surface the outward transport of chromium ions.27–31
scales for their oxidation resistance during the high A major drawback of metallic chromium is its high
temperature service. Chromia provides less oxida- ductile to brittle transition temperature (DBTT)
tion/corrosion protection than alumina or silica,11 which is known to be significantly affected by inter-
however, its electronic conductivity in the envisaged stitially dissolved gaseous impurities such as oxygen
SOFC operation temperature range of 600–1000 C and nitrogen.23 Extensive research work at Plansee
is orders of magnitude larger than that of alumina or led to the development of oxide dispersion strength-
silica. Due to the required similarity of the thermal ened (ODS) chromium which showed substantially
expansion coefficient with that of the commonly used improved mechanical properties compared to con-
electroactive ceramic cell materials,18,19 the chromia- ventional chromium alloys.6,7 The ODS materials are
forming alloys, most frequently studied for SOFC manufactured by elemental mixing of the starting
interconnector application, are based on the binary powders (abbreviated as MIX alloy) or by mechanical
alloy system Fe–Cr. The thermal expansion coeffi- alloying (MA), that is, by high-energy ball milling of
cient in Fe–Cr alloys decreases with increasing chro- chromium and oxide powders in a nonoxidizing envi-
mium content20 and consequently, chromium-based ronment.23,32 The latter process can produce a very
materials have in most cases been considered for finely distributed oxide dispersion (typical size
application in zirconia electrolyte-based SOFC con- 15 nm) embedded in the chromium matrix, compared
cepts.6,21–23 If a slightly higher thermal expansion to the coarser dispersions in MIX materials, as shown
coefficient is tolerable (typically 12 106 K1), in Figure 2.6,7 The alloys manufactured from the
for example, the case in Ni-cermet anode-based cell compacted powders possess superior mechanical
concepts or in cells using ceria as electrolyte, high- properties, especially with respect to creep resistance
chromium ferritic steels are commonly considered (Figure 3). The improvement in ductility depends on
as construction material for the interconnect.20 a number of factors, for example, dispersion type and
distribution. A substantial ductility increase of Cr was
obtained by using MgO dispersions,33 however, no
1.19.3 Chromium-Base Metallic further data on other mechanical properties have yet
Interconnectors been reported. The commercial dispersion strength-
ened Cr alloys especially developed for SOFC appli-
The chromia scales forming on pure chromium and cation mostly contain a few percent of iron to adjust
chromium-based alloys at high temperatures exhibit the thermal expansion coefficient as close as possible
slow growth rates. However, when formed during to that of the yttria-stabilized zirconia electrolyte.6,7
exposure in air or oxygen, the scales tend to buckle As mentioned above, the RE-oxide dispersion not
and possess poor adherence to the metallic substrate. only improves the mechanical properties but also
The buckling is related to substantial compressive decreases the oxide growth rate and improves the
growth stresses in combination with vacancy conden- adherence of the oxide to the metallic substrate.28–32
sation at the scale–metal interface. Tracer studies
using 18O isotopes revealed that the latter is related
to outward chromium diffusion being the dominating 1.19.4 Volatile Cr Species
transport mechanism in chromia scales whereas the
contribution of oxygen transport is substantially smal- During the operation of planar SOFCs with Cr-based
ler.24–27 The buckling that occurs at oxidation tem- interconnects, rapid degradation of the cell perfor-
perature as a result of lateral growth24 has frequently mance has frequently been found and was claimed to
form volatile (oxy-) hydroxide species, however, the

published data on the actual thermodynamic stability
of these vapor species still show major differences. The
rate of material loss by volatilization is highly depen-
dent on the partial pressures of the volatile species as
well as the gas flow rate. It is known that under laminar
flow conditions, which are representative of many
power generation and propulsion environments, the
rate of material loss, J (in mg cm2 h1) by evapora-
tion, is given by the following expression37:
DPM
(a)
100 nm J ¼ 0:664 Re1=2 Sc1=3 ½1
LRT
where the terms are: J is the mass flux (mg cm2 h1);
Re, the Reynold’s number; Sc, the Schmidt number; D,
the interdiffusion coefficient of volatile species in lam-
inar boundary layer (cm2 s1); P, the partial pressure
of volatile species (atm); M, the molecular weight of
volatile species (g mol1); L, the characteristic length
(cm); R, the gas constant (cm3 atm) K1 mol1; T is the
absolute temperature (K).
At high temperatures in dry air, the Cr2O3 surface
scale present on Cr or Cr-based alloys evaporates
according to the following reactions, forming volatile
2 µm CrO, CrO2, and CrO3:
(b)
1
Cr2 O3 ðsÞ þ O2 ðgÞ $ 2CrO2 ðgÞ ½2
Figure 2 TEM pictures showing distribution of Y2O3 2
dispersions in Cr-based ODS alloys, manufactured by (a)
mechanical alloying (MA) and (b) elemental mixing (MIX) of 1
Cr2 O3 ðsÞ $ 2CrOðgÞ þ O2 ðgÞ ½3
the metal and oxide powders. Reproduced from 2
Quadakkers et al. Mater. High Temp. 2003, 20(2), 115–127.
3
Cr2 O3 ðsÞ þ O2 ðgÞ $ 2CrO3 ðgÞ ½4
2
Ebbinghaus et al.38 summarized and evaluated the
be caused by the release of gaseous Cr species, desig- available literature data and found CrO3(g) to be
nated in the following sections as ‘Cr vaporization,’34 the most abundant vapor species formed upon vapor-
from the metallic materials. The volatile Cr species ization in dry air.
are reduced at the triple-phase boundaries between In the development of SOFCs that use chromia-
cathode, electrolyte, and air thereby forming solid forming alloys or conductive chromia-containing
Cr2O3 and other Cr-rich phases which inhibit the ceramics as interconnect, the loss of chromium in
electrochemical processes of the cell. This effect is the presence of air and water vapor has frequently
often called ‘poisoning’ of the cathode and/or the been identified as a major cause for cell degradation.
cathode–electrolyte interface. A detailed description Therefore, a detailed knowledge on the vaporization
of the degradation mechanisms was given by Hilpert behavior and equilibria of chromia in the cathode
et al.35 and Jiang et al.36 side environments, especially in the presence of
The gas on the cathode side of the SOFC is in water vapor, is of great importance.
most cases normal air which contains the usual minor It has been experimentally proven that in wet air
amounts of water vapor. Several authors have shown the volatile CrO2(OH)2 is formed by the reaction35,39:
that even small amounts of water vapor may sub-
1 3
stantially increase the ‘Cr vaporization’ in the high Cr2 O3 ðsÞ þ O2 ðgÞ þ H2 OðgÞ $ CrO2 ðOHÞ2 ðgÞ
temperature environment. Chromia are known to 2 4
react with water vapor at elevated temperatures to ½5
90
80 Fracture strain as function of temperature
70
60 Cr–0.4La2O3
Strain (%)
50
40
30
20
10
Cr–5Fe–1Y2O3
0
0 200 400 600 800 1000 1200 1400
(a) Temperature (°C)
2.4
Creep rupture data of various alloys at 900 °C
2.2
2 Cr-ODS alloy
log (stress)/(N mm−2)
1.8
Inconel 617
1.6
1.4
1.2
Alloy 800
1 Manaurite 36x
0.8
0.6
1 2 3 4 5
(b) log (time (h))
Figure 3 Typical mechanical properties of Cr-based ODS alloys. Reproduced from Quadakkers, W. J.; Greiner, H.;
Köck, W. In Proceedings of the First European Solid Oxide Fuel Cell Forum, Bossel, U., Ed.; European SOFC Forum,
Dr. Ulf Bossel, Morgenacher Str. 2F, CH-5452: Oberrohrdorf, Switzerland, 1994; Vol. 2, p 525.
At the cathode–electrolyte interface, the oxyhydrox- which in some cases resulted in large uncertainties of
ide reacts according to: the derived thermodynamic data. Assuming the validity
of eqn [5], the equilibrium constant Kp of the reaction
ðgÞ ðsÞ
2CrO2 ðOHÞ2 þ 6e ! Cr2 O3 þ 2H2 OðgÞ þ 3O2 can be calculated from experiments using the vapor
½6 transpiration method. Stanislowski et al.40 have carried
out equilibrium vaporization experiments with Cr2O3(s)
Ebbinghaus38 derived thermodynamic data for a large in humid air under SOFC-relevant conditions. The data
number of volatile chromium oxyhydroxides in 1993 derived are confirmed by the work of Opila et al.41 The
using the limited available experimental data at that results are shown in Figure 4. For comparison, the
time, in combination with the molecular constant calculated equilibrium constants according to the ther-
method. Empirical relationships were employed, where modynamic database IVTANTHERMO,42 as well as
needed, to estimate unknown molecular properties, the data from Ebbinghaus38 and the experimental results
T (⬚C)
1200 1000 800 600 400
10−3
10−4
10−5
10−6
Kp
10−7
Stanislowski et al.
Gindorf et al.
10−8 Opila et al.
IVTANTHERMO
10−9 Ebbinghaus
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6
T−1(103 K−1)
Figure 4 Equilibrium constant for reaction [5], comparing the data of various authors. Reproduced from Stanislowski, M.;
Wessel, E.; Hilpert, K.; Markus, T.; Singheiser L. JES 2007, 154(4), A295–A306.
from Gindorf et al.43 are shown. From those results, the 1.19.5 Behavior of Cr-Based Alloys
enthalpy and entropy of formation for the volatile spe- in H2/H2O-Based Gases
cies CrO2(OH)2(g) can be calculated. The recom-
mended value resulting from the aforementioned Several reviews on oxidation of chromium in oxy-
comparison is DH0f,298 ¼767.7 11 kJ mol1. gen or air have been published and substantial infor-
In technical systems, the actually occurring chro- mation can be found in text books.10,11,48,49 Far less
mium evaporation also depends on a number of other information is available on behavior of Cr and Cr alloys
factors such as the morphology of the outer oxide in SOFC-relevant anode side H2/H2O-based gases.
layer as well as its growth mechanism and its adher- Hänsel et al.50 found the oxidation rate of Cr-based
ence to the metallic material. Therefore, the actual ODS alloys at 950–1050 C in an Ar/H2/H2O-mixture
chromium release from Cr-based alloys under oper- (equilibrium oxygen partial pressure 1015 bar at
ating conditions may quantitatively differ from that 1000 C) to be higher than in high-pO2 environments
measured for pure chromia samples. (air, Ar/O2), whereby the difference could only
Although the significance of reaction [6] occurring partly be explained by formation of volatile oxides
at the triple-phase boundaries has been identified in a and hydroxides in the high-pO2 gases (Figure 5). In
number of studies, several authors have claimed that the Ar/H2/H2O-environment, where the formation
the driving force for cell degradation is related to of volatile Cr species is negligible, the alloys tended
spinel formation in the cathode. This occurs by the to form whisker type oxide morphologies, the extent
reaction of volatile Cr species with elements such as of whisker formation being decreased by an ‘optimum’
Mn, Co, and Fe present in the perowskite type cath- addition of a reactive element (oxide dispersion).
ode and/or contact layer. The extent of degradation Chromia scales formed on Cr-based ODS alloys in
thus depends on the cathode composition.36 Several Ar/H2O/H2 are far less voided and exhibit substan-
protection methods have been proposed to minimize tially better adherence to the metallic substrate than
evaporation of volatile Cr-species, such as coating of those formed in air or oxygen.24,50 This effect of
the interconnector with La-chromite44 or -manganite,45 atmosphere composition on scale adhesion is even
metallic layers,46 oxide layers,46 or aluminum surface more dramatic in the case of non-RE-doped alloys
enrichment to promote alumina surface scale forma- and elemental chromium. Water vapor can eliminate
tion on interconnector areas, where electrical conduc- or at least decrease oxide porosity by providing rapid
tivity is not a major issue.46,47 This subject will be gas phase transport of oxygen within the void space. As
extensively treated in Section 1.19.16. proposed earlier,51 the oxygen transport is facilitated
8
Cr Cr–5Fe Cr–5Fe–1Y203 (MA)
6
After isothermal
oxidation
Weight change (mg cm−2)
After
2 cooling
−2
−4
Ar–O2 Ar–H2–H2O Ar–O2 Ar–H2–H2O Ar–O2 Ar–H2–H2O

−6
Figure 5 Weight changes of Cr, Cr–5Fe, and the ODS alloy Cr–5Fe–Y2O3 after isothermal oxidation in Ar–20% O2 and
Ar–4% H2–2% H2O at 1000 C and after subsequent cooling to room temperature. The results illustrate, especially for the
RE-free materials, the importance of the test atmosphere for oxide adherence. Reproduced from Hänsel, M.; Quadakkers, W. J.;
Singheiser, L.; Nickel, H. Report Forschungszentrum Jülich, Jülich FRG, Jül-3583, 1998; ISSN 0944-2952.
by the reaction between H2O(g) and oxide. Water non-RE-doped chromium and chromium alloys,
vapor reacts with the oxide on the side of the cavity reaction with nitrogen is of major concern because
nearest to the metal, whereas the formed hydrogen it leads to environmentally induced embrittlement
reacts with the outer surface of the void. These redox during high temperature exposure.23 This is caused
reactions are coupled with gas phase mass transfer, by nitrogen dissolution in the metal or the alloy as
and the inward oxygen transport through the scale well as by nitride formation beneath the chromia
via this mechanism leads to new oxide formation at scale (Figure 6). The sensitivity for nitrogen uptake
the scale–metal interface, reducing the probability of depends on the gas tightness of the chromia layer
void formation and subsequent scale detachment.52 because nitrogen transport between gas atmosphere
Enhanced inward oxygen transport, which was proven and metal or alloy is known to occur via gas molecules
by two-stage oxidation tests using Ar–H2–H18 2 O,
53
rather than via solid state diffusion through the scale.
accounts satisfactorily for the observed decrease in Generally, it can be said that the thick chromia scales
the scale porosity and improved scale adhesion. The formed on non-RE-doped Cr and Cr alloys are more
change in the dominating growth mechanism of chro- permeable for nitrogen than the thinner, that is, more
mia in the high- and low-pO2 environments is accom- protective, scales on RE- or RE-oxide-doped metals or
panied by a substantial change in oxide grain size, alloys. This observation in fact indicates that also in a
the grains being much smaller when formed in the purely oxidizing environment, the protective proper-
low pO2 environment. More detailed information on ties of the chromium oxide layer are to a large extent
the differences in chromia scale formation in low- and determined by its ability to prevent molecular trans-
high-pO2 gases can be found in the chapter ‘Oxidation port of oxidizing species rather than by the exact
in Steam and Steam/Hydrogen Environments.’ concentration of point defects in the chromia lattice.
The second reactive gaseous species which is of
importance during air oxidation of Cr-based materials
1.19.6 Mixed Gas Corrosion of is water vapor. As extensively described in the previ-
Cr-Based Alloys ous section, the first well-known effect of water vapor
is that it promotes the formation of volatile chromium
During service in the SOFC cathode side environ- oxyhydroxides.35,38,39,54 A second, less apparent effect
ment, the reaction of oxygen with the surface of the was observed by Hänsel et al.50 The authors found
Cr-based interconnect may be affected by the pres- the scale growth rate in the temperature range
ence of nitrogen and/or water vapor. Especially, for 950–1050 C to be enhanced if water vapor was
Spallation
area
Cr-oxide
Cr-oxide
Cr-nitride
Cr-ODS
10 μm 2 μm
Cr
Figure 6 Metallographic cross-sections (SEM images) of Cr and Cr ODS alloy after 1000 h oxidation at 800 C in air.
Oxide scale
Oxide scale
MIX-alloy
MA-alloy
10 μm
Figure 7 Oxide scales on Cr-based ODS alloys, manufactured by different processes (cf. Figure 2) after 1000 h exposure
in a H2O–H2 mixture at 950 C. Reproduced from Hänsel, M.; Quadakkers, W. J.; Singheiser, L.; Nickel, H. Report
Forschungszentrum Jülich, Jülich FRG, Jül-3583, 1998; ISSN 0944-2952.
present in air or Ar–O2, and this effect was related to carbon attack, which occurs by molecular transport of
the enhanced oxidation of cracks which formed in the the carbon-containing gas species through the oxide
chromia layers as a result of growth stresses even scale, decreases with the decreasing growth rate of the
during the isothermal oxidation. The enhanced oxi- oxide layer.55 This again is a strong indication that in
dation may therefore be related to a dissociation the case of chromia scales, protectiveness means to a
mechanism occurring in voids and cracks,51 similar large extent resistance against molecular transport
to that described in the previous section. from the gaseous atmosphere to the metal or alloy.
Depending on the cell operating conditions and In the case of Cr-ODS alloys, the protective scale
the actually used fuel, the mixed gas corrosion of properties depend on the dispersion type, and even
Cr-based alloys also has to be considered in the more on the dispersion distribution50,55,56 and thus on
anode side gas. If carbon-containing gas species such the alloy manufacturing process (Figures 7 and 8).
as CO and/or CH4 are present, all Cr-based alloys
tend to form a subscale layer containing Cr-carbide,
that is, mainly Cr23C6, sometimes in combination with 1.19.7 Oxidation Rates of
Cr7C3.21,55 In the case of a CH4-based test gas, the Chromium-Based Alloys
carbides exist as a near-continuous layer,21 whereas in
the case of a CO-based gas, the subscale layer mainly If chromia growth proceeds by diffusion of point
consists of an oxide–carbide mixture.55 The extent of defects in the oxygen and/or chromium sublattice,
Oxide scale
Ni-coating
Oxide scale
Cr-carbide
MA-alloy MIX-alloy
10 μm
Figure 8 Oxide scales on Cr-based ODS alloys, manufactured by different processes (cf. Figure 2) after 1000 h exposure
in a CO-containing H2O–H2 mixture at 950 C. Reproduced from Hänsel, M.; Quadakkers, W. J.; Singheiser, L.; Nickel, H.
Report Forschungszentrum Jülich, Jülich FRG, Jül-3583, 1998; ISSN 0944-2952.
scaling rates could be calculated by Wagner theory,11 growth proceeds to a large extent via rapid diffusion
if the self-diffusion coefficients of chromium and paths, such as oxide grain boundaries,26,27,58 the den-
oxygen were known. Lattice self-diffusion coeffi- sity of which is in most cases not time-independent.
cients of both chromium and oxygen in chromia are For ODS alloys, the parabolic rate constant K 0 (in
reported to be extremely small, whereby the litera- cm2 s1) for such cases can be written as62:
ture data show considerable scatter. At 1100 C, the Do d mo
self-diffusion coefficients are in the range 1017 to K0 ½9
r RT
1018 cm2 s1.57–59 These low values indicate that it is
in which Do is the oxygen grain boundary diffusion
extremely difficult to achieve true equilibrium, and
coefficient; d, the grain boundary width; r, the grain
this might be an important reason for the discrepan-
size; and Dmo, the oxygen-potential gradient across
cies in the literature diffusion data.
the scale. In many cases, r increases in scale growth
The oxidation rate of Cr-based ODS alloys is
direction, and/or with oxidation time.62,63 This fre-
substantially slower than that of pure Cr or Cr alloys
quently leads to a K 0 -value which decreases with
without addition of a RE (oxide).48–50 It is difficult to
increasing oxidation time (Figure 9).
give a quantitative figure of this reduction in oxida-
Another factor which has to be considered in
tion rate by the RE addition because Cr as well as Cr-
evaluating oxide growth rates is that the scales are
based alloys frequently do not exhibit the classical
not completely gas-tight due to the formation of
parabolic time dependence for oxide scale growth.
micro voids and micro cracks and consequently,
Classical Wagner’s oxidation theory11 predicts the
molecular gas transport contributes to the growth
scale growth to obey a parabolic time dependence,
process. Furthermore, the above-mentioned forma-
if oxidation is controlled by diffusion of metal and/or
tion of volatile oxides and/or hydroxides can affect
oxygen ions through the oxide lattice:
chromia growth, especially, at high temperatures in
x2 ¼ K 0t ½7 oxygen-rich environments (see previous sections).
In the case of ODS alloys, it has to be considered
in which x is the scale thickness, t the time, and k the that the desired low oxide growth rate of the scales
parabolic oxidation rate constant. Expressed as oxy- requires the RE to become incorporated in the sur-
gen uptake per unit area (Dm), eqn [7] is mostly face oxide layer.31,64 This occurs by embedding the
written in the following form: dispersions in the inwardly growing oxide scale and
ðmÞ2 ¼ Kp t ½8 subsequent diffusion of the RE to the oxide grain
boundaries.60 Consequently, the beneficial effect does
For Cr, Cr alloys, and especially Cr-based ODS alloys, not immediately become apparent but appears only
eqn [8] is not often followed50,51,60 because the after a certain oxidation time, with the consequence
assumptions made to describe scale growth by eqn that a rapid oxide growth rate occurs which then
[8] are mostly not completely fulfilled.61–63 Chromia subsequently decreases to very low values.50,60,62,63
1.E−09
1000 °C
1.E−10
Cr
Kp (g2 cm−4 s−1)
1.E−11
Cr5Fe
1.E−12 Cr5FeY2O3-mix
1.E−13
Cr5FeY2O3-MA
1.E−14
0 10 20 30 40 50 60 70 80 90
Time (h)
Figure 9 Typical time dependence of the apparent oxidation rate Kp0 during isothermal oxidation of Cr and Cr alloys in
Ar/H2/H2O at 1000 C. Reproduced from Hänsel, M.; Quadakkers, W. J.; Singheiser, L.; Nickel, H. Report Forschungszentrum
Jülich, Jülich FRG, Jül-3583, 1998; ISSN 0944-2952.
As a reduction in scale growth rate in a RE (oxide) presence of impurities or dopants.68–73 The electrical
containing Cr alloy requires the RE to become conductivity of bulk chromia samples at 800–1000 C
incorporated into the scale,64,65 the RE-imparted (Figure 10) has been reported to be in the range
reduction in chromia growth rate is strongly affected 1 10–2 to 5 10–2 S cm1,68–71,74 although lower
by size and distribution of the oxide dispersion.50 Small values have also been found.75–77 Below 1000 C Mg,
dispersoid sizes and small distances between the dis- doped chromia is a p-type conductor, whereby the
persoids are required to obtain an optimum growth rate electronic conductivity only marginally depends on
reduction. The oxidation rates of the ODS alloys thus oxygen partial pressure.68 In H2–H2O mixtures, the
depend on the manufacturing process.50 In fact, tracer conductivity is affected by dissolved protons,68,70
studies using 18O2 revealed that even by the optimum whereas in TiO2-doped chromia, no effect of hydrogen
dispersion distribution, which can be achieved in com- on conductivity is detected.69 In the temperature range
mercial alloys, the resulting RE distribution in the 400–1000 C, the Ti-doped chromia is an n-type con-
growing scale does not completely block outward chro- ductor at low oxygen partial pressures and a p-type
mium transport. A virtually complete suppression of conductor at near atmospheric oxygen partial pressure.
outward growth can be obtained by a strong surface Nagai et al.75,78,79 found the electronic conductivity to
enrichment of the RE, for example, by applying a RE- increase by Y2O3 or La2O3 doping and decrease with
oxide coating or by RE-ion implantation.65,66 increasing oxygen partial pressure, although this effect
is not very pronounced, especially at high temperatures
(Figure 11). The highest increase of the conductivity
1.19.8 Electronic Conductivity of was observed for doping by NiO, whereby in this case
Chromia-Based Scales no oxygen partial pressure dependence was found. Also
doping by CeO2 or Fe2O3 increased the conductivity,
Chromia is an electronic conductor and at high tem- however, the effect was much smaller than that
peratures (>1000 C) the electrical conductivity is observed for NiO-doping.
found to be independent of the oxygen partial pres-
sure.67–69 At lower temperatures, the concentration
of electronic defects due to the intrinsic electronic 1.19.9 High-Chromium Ferritic Steels
equilibrium becomes so small that the chromia
changes into an extrinsic electronic conductor, the High-chromium ferritic steels are to be preferred
electronic conductivity being dominated by the over Cr-based (ODS) alloys if lower SOFC service
3.5
2.5
log (σT) (S K cm−1) 2

Holt and Kofstad
1.5
Su
1 Hagel and Seybolt
0.5
Crawford and Vest
0 Nagai et al.
−0.5
5 7 9 11 13 15
1/T (104 K−1)
Figure 10 Electrical conductivity of Cr2O3 as function of reciprocal temperature in air (derived from Holt,68–71
Hagel and Seybolt,74 Nagai and Fujikawa,75 Su,76 and Crawford and Vest77).
3.5
Kofstad and Holt
(2 mol.%MgO)
2.5
Nagai et al. (1 mass% NiO)
log (σT) (S K cm–1)
Kofstad and Holt

(nondoped)
1.5
Nagai et al. (1 mass% La O )

2 3
0.5
Nagai et al. (1 mass% Y2O3)
Nagai et al. (nondoped)

–0.5
6 8 10 12 14 16 18
1/T (104 K−1)
Figure 11 Effect of dopants on electrical conductivity of Cr2O3 as a function of reciprocal temperature (derived from
Nagai and Fujikawa,75 Nagai and Ohbayashi,78 Nagai et al.79).
temperatures (<900 C) prevail, especially in the case resistance generally increases with increasing Cr con-
of anode-supported cell designs, in which the CTE is tent.80 On low-Cr steels (<5% Cr), the scales consist
mainly governed by the anode rather than the elec- of nearly pure Fe-oxide accompanied by internal
trolyte. In commercially available ferritic steels, for oxide precipitates of Cr2O3 and/or FeCr2O4-spinels.
high temperature application, the Cr content varies With increasing Cr content, the scales become richer
between 7 and 28 mass% (Table 2). During air in spinel and chromia, which is accompanied by a
exposure, for example, in the temperature range decrease of the scale growth rate. Formation of
700–1000 C, the oxide scale composition strongly a highly protective, single-phase chromia layer
varies with Cr concentration and the oxidation requires a chromium content of 17–20%, the
Table 2 Main alloying elements in selected commercially available, high-Cr ferritic steels
Material Commercial name DIN-Nr. Composition of Cr, Al, Si, Mn, and others (mass%)
X10CrAl7 Ferrotherm 4713 1.4713 6–8 0.5–1 0.5–1

X10Cr13 Nirosta 4006 1.4006 11.9 <0.02 0.49 0.30
1.4509 15–16 <1.25 <0.5 <1 Ti/Nb
X10CrAl18 Ferrotherm 4742 1.4742 17.3 1.04 0.93 0.31
X10CrAl24 Ferrotherm 4762 1.4762 23.5 1.82 1.01 0.38
24.4 1.38 1.14 0.54
Fe-25Cr-Mn RA 446 24.2 0.43 0.67
Fe-26Cr-1Mo E Brite 25.8 0.24 0.02
X18CrN28 Sandvik 4C54 1.4749 26.5 <0.01 0.47 0.70
Fe-29Cr-4Mo-Ti Al29–4C 27.3 0.26 0.28 Ti
exact concentration depending on surface treatment, poorly protective Fe- and Cr-rich heterogeneous
minor alloying additions, environment, and impuri- oxide mixtures accompanied by internal oxidation
ties (e.g., C, S, P). The critical Cr content for obtaining and nitridation of aluminum on coarse ground sur-
protective chromia scale formation on ferritic steels faces but tended to form a protective alumina scale
increases with increasing temperature. on polished surfaces.80,82,83
Minor alloying additions of Mn, Ti, Si, and Al may This finding illustrates that only slight changes
substantially alter the oxidation behavior of commercial in alloy composition, service temperature, or
ferritic steels (Figure 12). During high temperature design-related parameters can fundamentally change
exposure, the two first mentioned elements tend to the oxidation properties of high-chromium ferritic
become incorporated in the scale,81 although titanium steels (Figure 12). Additionally, a change from one
can also be present in the form of internal titania pre- oxide scale type into another may occur during long-
cipitates in the subsurface region. Titanium can become term service due to the depletion of minor alloying
dissolved in chromia at low oxygen partial pressures, elements such as Al, Si, Mn, and/or Ti in the alloy
that is, in the inner part of the scale, whereas at higher matrix because of their low concentrations.
oxygen partial pressures, that is in the outer part of the External silica scale formation is rarely observed
scale, it tends to become reprecipitated. It is therefore during oxidation of commercial ferritic steels. Some-
frequently found in the form of titania crystals at the times silica, mostly in the amorphous state, locally
scale–gas interface.81 The incorporation of Ti into the forms relatively dense subscales at the interface between
oxide scale tends to increase the chromia growth rate, alloy and chromia-based external layer (Figure 13).
especially at higher temperatures. Mn possesses only a This effect is frequently accompanied by an increased
small solubility in chromia and it is frequently found in tendency of the oxide scales to spallation during thermal
the form of a Cr–Mn-spinel layer in the outer part of cycling, probably due to the extremely small thermal
the oxide scale. expansion coefficient (CTE) of the silica compared to
Oxides of Si and Al mostly prevail as internal that of the alloy and/or due to the transformation, for
oxides rather than being incorporated into the exter- example, into crystoballite.
nal scale. This is related to the higher thermody- As already mentioned before, the formation of
namic stability of those oxides compared to that of alumina or silica in the form of a closed (sub-) layer
chromia. If the Al and/or Si contents in the steel are is generally not desired in SOFC application, because
increased to 1%, the internal silica and/or alumina it can dramatically increase the contact resistance
precipitation can change into protective, external between the interconnector and the electrodes or
scale formation, which is accompanied by a strong contact materials.
decrease in oxidation rate.80 The exact Si and/or
Al levels, at which this effect occurs, depend on
temperature, Cr content, concentrations of minor 1.19.10 Behavior of High-Cr Ferritic
alloying additions, and component surface treatment. Steels in Anode Gas
The latter effect was, for example, clearly demon-
strated during air oxidation at 800 C of an 18Cr, A change of the test atmosphere from air to an anode
1Al steel (DIN designation 1.4742) which formed gas simulating H2O–H2 mixture for high-Cr ferritic
Spallation Ni-coating
Ni-coating
Internal oxidation zone Oxide scale
Oxide scale Breakaway

Steel
50 μm
Steel 50 μm
1.4509 1.4016
Spallation
Oxide scale
Internal oxidation zone Oxide scale
Steel Steel
50 μm 50 μm
ZMG232 Alloy 446
Ni-coating Oxide scale

Ni-coating Oxide scale
Internal oxidation zone

Steel Steel
50 μm 50 μm
Crofer 22 A Crofer 22 APU

Figure 12 Different types of oxide scale formation on high-Cr ferritic steels after 1000 h oxidation at 900 C in air.
steels has similar effects as described for the Cr-based whereas the spinel in simulated anode gas grows via
alloys. The morphology of the chromia and spinel- grain boundary diffusion86 resulting in plate and
rich surface layers is slightly modified84 and the whisker type oxide morphologies and microvoids at
adherence of the scale is improved.80 The oxide the chromia–spinel interface (Figures 14 and 15).
growth rates of ferritic steels with Cr contents above Further details can be found in the section on ‘Oxida-
about 20% in H2O–H2 mixtures are generally similar tion in Steam and Steam/Hydrogen Environments.’
or slightly smaller than those in air. This observation Due to the substantially lower equilibrium oxygen
seems at first sight to be surprising because chromia partial of the anode gas compared to that of the cathode
scales on Cr-based alloys showed an opposite gas side environment, the thermodynamic driving force for
composition dependence of the oxidation rates, as oxide formation in anode gas is in fact substantially
described in Section 1.19.5. The behavior of the smaller than that in the cathode gas. In ferritic steels
high-chromium ferritic steels in anode and cathode with Cr contents which are lower than 20%, the
side gases was shown85 to be governed by the differ- anode side environment can, for instance during the
ences in growth mechanisms of a Cr–Mn-spinel layer transient stages of oxidation, suppress formation of
which forms on top of the inner chromia scale. In oxides with low thermodynamic stability.80 This may
anode gas, the spinel tends to be stoichiometric for such steels result in a type and morphology of
MnCr2O4, whereas in air it is Mn1þxCr2 xO4. The external and internal oxide formation which completely
latter grows by the outward lattice diffusion of cations differs from that occurring during air or oxygen
(a) 2 μm
(b) 2 μm
(c) 2 μm
Figure 13 Oxide scale formation on ferritic steels with 22% Cr after 1000 h exposure at 800 C in air showing effect of
Si/Al content: (a) 0.01% Si/Al, (b) 0.1% Si/Al, and (c) 0.4% Si/0.2% Al. Reproduced from Quadakkers et al. Mater. High Temp.
2003, 20(2), 115–127.
A more dramatic effect of atmosphere depen-

dence of oxidation rates can be found for ferritic
Ni coating
steels with intermediate Cr contents of 10–20%.
In atmospheres in which water vapor is the main
(MnCr)3O4 oxidizing species, the tendency to form a protective
external chromia-based scale is substantially sup-
pressed compared to that during air or oxygen expo-
sure.88,89 Presence of high amount of water vapor thus
Cr2O3
shifts the critical Cr content for obtaining protective
chromia scale formation towards higher concentra-
tions. This has as a result, that a steel with an interme-
diate Cr content, which may form a protective
chromia-based scale during air or oxygen exposure,
may form a poorly protective, Fe-base oxide scale
Alloy during exposure in a high water vapor-containing
2 μm environment. The tendency to nonprotective scale
formation is suppressed by the small size of alloy
Figure 14 TEM bright field image of oxide scale (spinel þ grain and surface deformation of the steel. Details
chromia) on high-Cr ferritic steel after exposure in air at
on the mechanisms of nonprotective scale formation
900 C for 1000 h. Reproduced from Essuman et al. Corros.
Sci. 2008, 50, 1753–1760. in wet gases can be found in the chapter on ‘Oxidation
in Steam and Steam/Hydrogen Environments.’
exposure.80 This, however, does not mean that the The formation of poorly protective, Fe-rich oxides
oxidation rates are generally smaller in anode than in will, of course, occur only if the equilibrium oxygen
cathode gas. An example of higher growth rates in partial pressure in the gas is higher than the dissocia-
anode gas was already illustrated for the case of pure tion pressures of the Fe oxides. In an H2–H2O gas
chromia scales formed on chromium-based alloys mixture at typical SOFC temperatures in the range
(Section 1.19.5) and was also found to occur in the 600–900 C, this is the case if the H2/H2O ratio is
case of Mn-free, high-purity Fe–Cr and Ni–Cr model lesser than 1. Thus, if pure hydrogen is used as fuel
alloys with Cr contents well above 20%. in an SOFC stack with an intermediate Cr content,
Cr Mn
Whiskers
Voids
Oxide scale Steel

0.5 μm
Figure 15 TEM bright field image and EELS maps for Mn and Cr of the oxide scale on high-Cr ferritic steel during 15 h
exposure at 800 C in Ar–4%H2–2% H2O. Reproduced from Essuman et al. Corros. Sci. 2008, 50, 1753–1760.
ferritic interconnect, the poorly protective oxides Table 3 Typical gas composition at SOFC gas inlet and
will not form in areas near the gas entrance of the gas outlet (assuming 80% fuel utilization) when using
reformed gasoline as fuel
stack but only in gas flow direction nearer to the gas
outlet, the critical areas being determined by the Gas species Gas composition (vol.%)
actually applied fuel utilization.
Stack inlet Stack outlet
H2 20 4
CO 22 4.4
1.19.11 Ferritic Steels in Anode N2 52 52
Gases Containing Carbonaceous CO2 3 20.6
H2O 2.5 18.5
Gas Species
CH4 0.5 0.2
Applying internal or external reforming, SOFCs

can handle a variety of fuels containing carbon- When using carbonaceous fuels, the interconnect
containing gas species. The simulated anode gas will may be subjected to environments in which the oxy-
then contain gas species such as CO, CO2, and/or gen partial pressure and the carbon activity substan-
CH4 in addition to H2 and H2O. tially change as a function of time. For illustration, we
The carbon transfer which may occur from the use the application of an SOFC stack in mobile
service environment to the alloy can, depending on application using reformed gasoline as fuel (Table 3).
the prevailing gas composition, the temperature, and During the heating process of the stack to the opera-
the properties of the interconnect material, result in a tion temperature, the interconnect is subjected to a
number of material degradation mechanisms, that gas with a high-carbon activity and thus a substantial
is, carburization, metal dusting, and C deposition. danger for occurrence of metal dusting exists, especially
The latter two effects occur only if the C activity of in the temperature range 400–650 C (Figure 16).
the environment is larger than unity. Thereby When reaching the SOFC operation temperature of
C deposition may hamper the cell performance, for 800 C, the carbon activity is lower than unity and
example, by blocking narrow gas passages, but it does metal dusting attack is no longer possible. Due to the
not necessarily affect the properties of the intercon- electrochemical reactions occurring at the anode side of
nect adversely. Metal dusting, however, is a process the cell, the concentrations of CO, CH4, and H2 de-
which commonly results in a very rapid damage of crease in gas flow direction, whereas the concentrations
metallic materials. The mechanisms of the metal dust- of CO2 and H2O increase. As shown in the Cr–O–C
ing process are still a subject of scientific debate and a stability diagram in Figure 17, chromia is the stable
detailed description of it falls outside the scope of this phase in equilibrium with the gas atmospheres at
chapter. The reader is referred to a number of refer- 800 C. However, it is well established55,90 that under
ences90,91 in which the various steps in the degrada- such conditions, formation of internal Cr carbide, for
tion process are described. example, Cr23C6, may occur even if the equilibrium
aC
0 aC
−5
log10(activity)
−10 Stack inlet Stack outlet
−15
pO2
−20
pO2
−25
−30
550 600 650 700 750 800 850 900
Temperature (°C)
Figure 16 C activities and oxygen partial pressures of gases in Table 3 as a function of temperature.
Cr3C2 Gas inlet
–1
–2 Cr7C3
log10aC(s)
–3
Gas outlet
Cr23C6
–4
Cr Cr2O3
–5
–6
–40 –38 –36 –34 –32 –30 –28 –26 –24 –22 –20 –18 –16 –14 –12 –10 –8 –6 –4 –2 0
log10(PO2(atm))
Figure 17 Cr–O–C stability diagram at 800 C and the location of the equilibrium oxygen partial pressures and carbon
activities of the atmospheres in Table 3.
carbon activity in the gas is smaller than the activity The formation of these Cr-rich carbides (Cr23C6 and/
required to form the carbide. This is related to a decrease or Cr7C3) leads to material embrittlement and addi-
in oxygen partial pressure and a parallel increase in tionally depletes Cr from the ferritic matrix in the
carbon activity upon molecular transport of the gaseous surface-near region of the alloy. The latter may result
carbonaceous species through the surface oxide layer.90 in a reduced Cr availability for transport to the sur-
Internal carburization in the case of ferritic steels face required for protective scale formation and thus
does not generally result in formation of a continuous in a deterioration of the steel oxidation resistance.
subscale carbide layer, but in formation of finely During stack operation, the steel interconnect
distributed carbide precipitates in the alloy matrix. may thus be subjected to service environments
which, from the viewpoint of thermodynamic driving optimum combination of low scale growth rate and
forces, cycle between metal dusting and oxidizing/ excellent scale adherence. La appeared to be the
carburizing conditions. It is generally accepted that preferred RE addition, mainly because it does not,
metal dusting as well as internal carburization can be contrary to other commonly used REs, form interme-
prevented by the presence of a protective surface tallic compounds with iron. This allows La to become
oxide layer which prohibits molecular transport of evenly distributed in the alloy rather than being pre-
carbonaceous species to the metal surface. The resis- cipitated in the form of intermetallics on the alloy
tance against metal dusting and carburization thus grain boundaries, for example, observed if Y or Ce
increases with increasing protective properties of additions98 are being used. The actual scale growth
the surface oxide layer. Ferritic steels with Cr con- rate strongly depends on commonly present minor
tents well above 20% showed higher metal dusting alloying additions/impurities such as Mn and Ti
and carburization resistance under SOFC conditions (Figures 18 and 19). Although in RE-doped steels,
than lower Cr steels and, for example, 18–20% Cr very protective behavior can be obtained if the Cr
austenitics.92 content is just 16–18%, a substantially higher level
was chosen to obtain a larger ‘Cr reservoir’ to coun-
teract the danger for the occurrence of breakaway
1.19.12 Ferritic Steels Designed for oxidation due to the presence of high water vapor
SOFC Application contents in the anode gas,99 Cr loss due to interaction
with contacting materials100 or rapid exhaustion of
The optimum steel composition for an SOFC inter- the Cr reservoir in very thin interconnects. Small
connector will depend on the used cell design (e.g., amounts of Mn and Ti were added to obtain external
electrolyte or anode supported concept) and the ser- spinel formation which is known to decrease the
vice conditions (temperature, thermal cycling, fuel gas formation of volatile Cr species.43,97,101 Minor Ti
composition, and required service life).93 If a YS–ZrO2 additions were added to obtain fine internal oxide
electrolyte-supported design is considered, a low ther- precipitates of titania (Figure 20) which results in
mal expansion coefficient of 10 106 K1 is a cru- strengthening of the surface-near alloy region thus
cial requirement for the interconnector. Using binary reducing the tendency for occurrence of surface
Fe–Cr alloys with Cr contents commonly present in wrinkling caused by relaxation of oxidation-induced
ferritic steels (see Table 2), such a low value cannot be stresses during thermal cycling. By defining low Al
achieved.20 Some authors, therefore, alloyed steels on and Si contents (Table 4), formation of oxide scales
the basis of Fe–20% Cr with large amounts of refrac- with a very low electrical resistance was obtained for
tory elements such as Mo or W94,95 to decrease the alloys of the type FeCrMn(La/Ti) (Figures 21 and 22).
CTE. The oxidation resistance could be improved by The low Al/Si content also prevents extensive inter-
adding small amounts of REs. However, it has to be nal oxide formation which was found to affect not
demonstrated whether the scaling rates are sufficiently only the mechanical properties of the surface-near
small for the materials to be used in SOFCs with the regions but also the growth rate of the external chro-
required high-application temperature of 1000 C in mia scale. The latter is related to a volume increase
electrolyte-supported designs. imparted by the internal oxide formation resulting
Substantially lower service temperatures between in outward metal flow and inclusions of metallic
600 and 800 C seem to be more appropriate if fer- nodules in the scale.86,102
ritic steels are to be used as interconnectors. In this Further indications for development of steels
temperature range, for example, ceria should be used which are especially designed for SOFC application
as electrolyte3 or an anode-supported cell design are described elsewhere,103–105 however, no detailed
allowing an extremely thin YS–ZrO2 electrolyte information in respect to steel composition prevail.
should be applied.82,83 The choice of such concepts Recently, developments on the basis of 22% Cr-
would require interconnectors with CTEs similar containing ferritic steels with systematic variations
to that of the Ni/ZrO2-anode, that is, in the range of minor alloying additions were described.106
11–12 106 K1. Such CTE values can be achieved Honnegger et al.107 developed high-Cr ferritic mate-
by ferritic steels with Cr contents of 20%. rials for SOFC application that were produced
Quadakkers et al.96,97 studied the suitability of by powder metallurgical techniques, thus allowing
model steels with Cr contents of 16–25% and eval- incorporation of oxide dispersions to achieve dis-
uated the effect of various RE additions to achieve an persion strengthening. The mechanical alloying
1E−10
Isothermal oxidation 800 °C, Ar/O2
1E−11
Fe–Cr–Ti–Y
Kp⬘ (g2 cm−4 s−1)
1E−12
Fe–Cr–Mn–Y
1E−13
Fe–Cr–Y
1E−14
Fe–18Cr–1Al (1.4742)
1E−15
0 5 10 15 20 25 30 35 40 45 50
Time (h)
Figure 18 Apparent oxidation rate Kp0 as function of time for various types of FeCr alloys during isothermal oxidation at
800 C in Ar/oxygen. Reproduced from Piron-Abellan, J.; Quadakkers, W. J. Development of ferritic steels for application
as interconnect materials for intermediate temperature solid oxide fuel cells (SOFCs), Report Forschungszentrum Jülich,
Jül-4170, March 2005; ISSN 0944-2952.
1.E−10
Cr (in H2/H2O)
Instantaneous Kp⬘ at 50 h (g2 cm−4 s−1)
Cr5Fe (in H2/H2O)

1.E−11
Cr5FeY2O3 - mix FeCrYTi
(in H2/H2O)
1.E−12 FeCrMn(La/Ti)
Cr5FeY2O3 - mix
1.E−13
Cr5FeY2O3 - MA Fe−24Cr−Mn
(in H2/H2O)
1.E−14
Cr5FeY2O3 - MA FeCrY
1.E−15
7 7.5 8 8.5 9 9.5 10 10.5
1/T (104 K−1)
Figure 19 Instantaneous apparent oxidation rates at 50 h of FeCr-RE and FeCrMn(La/Ti) alloys during isothermal oxidation
at various temperatures compared with data of various Cr alloys. If not indicated, data relate to exposures in Ar/20% O2
(compiled from Refs. 24, 50, 85).
process allows to produce materials with substantially can also be obtained by applying a RE-containing
higher concentrations of REs (here Y) than can be coating. In this way, the oxidation rate of a com-
obtained by conventional alloy manufacturing meth- mercial, ferritic 18% Cr steels can substantially be
ods (Table 4). Therefore, the ferritic ODS steels decreased.
generally exhibit lower oxidation rates than RE con- A major disadvantage of high-Cr ferritic steels
taining wrought alloys produced by conventional is their tendency to form s-phase and a0 precipitates
methods. A high RE content in the surface regions at lower temperatures. As both processes lead to
(Mn,Cr)3O4 TiO2 (Mn,Cr)3O4 TiO2 Mn-rich

regions
Cr2O3 Mn-rich Cr2O3

intrusion SiO2
Steel TiO2 Steel Al2O3

2 μm 2 μm

Figure 20 Cross sections of oxide scales on type FeCrMn(La/Ti) steels after 6000 h exposure in air at 800 C showing effect
of minor Al/Si additions. Left: Crofer 22APU, right: Crofer 22A. (Mark the differences in magnification). Reproduced from
Huczkowski, P.; Christiansen, N.; Shemet, V.; Niewolak, L.; Piron-Abellan, J.; Singheiser, L.; Quadakkers, W. J. Fuel Cells
2006, 2, 93–99.
Table 4 Compositions of selected ferritic steels specially developed for SOFC application
Steel Composition (mass%)
Fe Cr Mn Ti La Si Al Ni Zr Others
Crofer 22A Bal. 22 0.4 0.05 0.09 <0.01 <0.01 <0.002

Crofer 22APU Bal. 22 0.4 0.06 0.07 0.11 0.12 0.16
Crofer 22H Bal 22 0.4 0.06 0.09 0.3 0.01 0.1 Nb, W
ZMG232 Bal. 22 0.5 0.36 0.19 0.31 0.13
Sanergy HT Bal. 22 <0.5 <0.3 Nb, Mo, Ce
IT11 Bal. 26 0.04 0.2 0.03 Mo; Y2O3
100 000
Commercial 18% Cr ferritic steel (1.4742)
10 000
Contact resistance (mΩ cm2)
Preoxidation
1000 time
Commercial 24% Cr ferritic steel (alloy446)
100
10
Crofer22 variants
1
0 100 200 300 400 500 600
Time (h)
Figure 21 Contact resistance of various steels during isothermal exposure at 800 C in air. Reproduced from Piron-Abellan,
J.; Quadakkers, W. J. Development of ferritic steels for application as interconnect materials for intermediate temperature
solid oxide fuel cells (SOFCs), Report Forschungszentrum Jülich, Jül-4170, March 2005; ISSN 0944-2952.
2.5
2
log (sT ) (S K cm–1)
1.5 Crofer22APU
Holt et al. for pure chromia
Sakai et al. for
1 Cr2MnO4 spinel
Crofer22A Cr (99.96 wt %)
0.5
0 Nagai et al. for

pure chromia
Cr-ODS
–0.5
9 9.5 10 10.5 11 11.5 12
4
1/T (10 K–1)
Figure 22 Temperature dependence of ‘apparent’ conductivity for various steels during exposure at 800 C in air.
Reproduced from Huczkowski, P.; Christiansen, N.; Shemet, V.; Niewolak, L.; Piron-Abellan, J.; Singheiser, L.; Quadakkers,
W. J. Fuel Cells 2006, 2, 93–99.
embrittlement of the steels, the exact alloy composi- long-term formation of protective chromia-based
tion has to be carefully adapted to the actual envi- oxide scales is not only governed by the alloy chro-
saged application temperature. For instance, steels mium content but also by the chromium reservoir in
with a Cr content of more than 25% are prone to the prevailing interconnect. The growth of the chro-
s-phase formation even at temperatures as high as mia-based surface scale consumes chromium from
800 C, whereby the s-phase formation is known to the metallic material with the result that after long-
be promoted, for example, by alloying additions of term service, the initial chromium concentration in
Mo and Si. Hammer et al.102 illustrated that the s- the component (C0) may decrease to such a low level
phase nucleation is not only affected by the alloy (CB) that the growth of the protective scale can no
purity but also by the actual prevailing service envi- longer be sustained. The time until depletion of the
ronment. Steels with Cr contents substantially lower chromium content down to this critical concentration
than 20% behave more ‘forgiving’ in respect to will be shorter in thin than in thick components.
embritteling precipitate formation. However, they Huczkowski et al.109 used a simplified expression
are prone to breakaway oxidation in high water derived from studies on breakaway oxidation of
vapor-containing anode side gases.108 They also may alumina-forming FeCrAl alloys to predict the time
easily run into breakaway oxidation in cathode side (tB) until the occurrence of breakdown of the pro-
gases in the case of thin components, as will be illu- tective chromia scale for selected high-chromium
strated in the next section. ferritic steels as a function of temperature and com-
ponent thickness (Figure 23)

1.19.13 Effect of Component 3 rd 1=n
Thickness on Oxidation Behavior tB ¼ 2:3 10 ðC0 CB Þ ½10
k
Especially in mobile applications of SOFCs, a reduc- Here k and n are the oxidation parameters; d, the
tion in weight of the fuel cell stack is a major design component thickness; and r, the alloy density. In
requirement. Weight reduction can be achieved by deriving this expression, it was assumed that the
reducing the interconnect thickness from the com- chromium depletion profiles beneath the oxide scales
mon few millimeters to a few tenths of a millimeter. are virtually ideally flat. Experimental data showed
When using such thin components of high-chromium that in the temperature range 800–900 C, this
ferritic steels, it has to be considered that the assumption was a close description of the actual
1.E+07
1.E+06
Time to breakaway (h) 800 ⬚C

1.E+05
1.E+04
Constant Kp
1.E+03
1.E+02 900 °C
1.E+01
0.01 0.1 1 10
Specimen thickness (mm)
90 μm
Figure 23 Upper figure shows calculated life time for plate material of Crofer 22A in air at 800 and 900 C (ignoring scale
spalling and thickness dependence of Kp). Lower figure shows example for loss of protective scale properties for 0.2 mm
specimen after 1000 h at 900 C. Reproduced from Huczkowski, P.; Quadakkers, W. J. Effect of geometry and composition of
Cr steels on oxide scale properties relevant for interconnector applications in solid oxide fuel cells (SOFCs), Report
Forschungszentrum Jülich, Energy Technology; 2007; Vol. 65, ISBN 978-3-89336-484-8.
occurring profiles because the ratio between oxida- alloys. Figure 24 shows mass change data for the
tion rate k (in cm2 s1) and diffusion coefficient (D) of high-Cr ferritic steel Crofer 22 A109,112 during cyclic
chromium in the ferritic steel is quite small.109 oxidation at 900 C for specimens of different thick-
For cases in which oxide scale spallation occurred, a ness. The thickening rate of the two-phase oxide
modified expression based on observations for FeCrAl layer consisting of Cr2O3 and MnCr2O4 appears to
alloys110,111 may be derived: increase with decreasing specimen thickness. This
is confirmed by the metallographic cross sections
tB ¼ 2:3 103 ðC0 CB Þrd ðkÞ1=n ðm
Þð1=nÞ1 ½11 in Figure 25 which shows that the difference in
Here, Dm* is the critical weight change at which the overall scale thickness on high-chromium ferritic
onset of scale spallation occurs. steels is mainly related to the difference in thickness
A major finding of Huczkowski et al.112 was that, in of the inner chromia layer, whereas the thickness of
these life time calculations, it had to be taken into the outer spinel layer differs only marginally.109,112
account that not only the time until occurrence A second effect of specimen thickness is that the
of breakaway oxidation depended on component extent of internal oxidation may become limited by
thickness but also the intrinsic oxidation rates of the the reservoir of the minor elements Ti and Al in the
3.5
0.3 mm
3.0 0.1 mm

0.5 mm
2.5
2.0 2 mm
1.5
1.0
0.5 900 °C, air
0.0
0 200 400 600 800 1000 1200 1400
Time (h)
Figure 24 Weight change during cyclic air oxidation of ferritic steel Crofer 22 A at 900 C showing effect of specimen
thickness on oxidation rate. Reproduced from Huczkowski, P.; Quadakkers, W. J. Effect of geometry and composition of
Cr steels on oxide scale properties relevant for interconnector applications in solid oxide fuel cells (SOFCs), Report
Forschungszentrum Jülich, Energy Technology; 2007; Vol. 65, ISBN 978-3-89336-484-8.
Cr2O3 MnCr2O4
MnCr2O4
Cr2O3
Alloy 10 μm Alloy 10 μm
(a) (b)
Figure 25 Metallographic cross sections (SEM images) of Crofer 22 A after 1000 h cyclic oxidation at 900 C in air:
(a) 0.3 mm thick specimen and (b) 2 mm thick specimen. Reproduced from Huczkowski, P.; Quadakkers, W. J. Effect of
geometry and composition of Cr steels on oxide scale properties relevant for interconnector applications in solid oxide fuel
cells (SOFCs), Report Forschungszentrum Jülich, Energy Technology; 2007; Vol. 65, ISBN 978-3-89336-484-8.
component (Figure 25). At 800 C, the thickness thickness. Strong indications were found that the dif-
dependence of the oxidation rate is less pronounced ference in growth rate of the chromia-based scales on
than at 900 C, but is still clearly seen and of techni- thick and thin specimens is the result of compressive
cal importance (Figure 26). growth stresses in the scales which can relax by plastic
First attempts to explain the specimen thickness deformation of the substrate for thin specimens but are
dependence of the oxidation rates related to depletion maintained at significant levels for thick specimens.
of major (Cr) and minor alloying elements (Al, Si, La, For an oxide under a biaxial compressive stress
Ti, and/or Mn) as well as scale microcracking occur- sxx, the diffusivity may be written as114,115:
ring during thermal cycling.109,112 However, more
2

S H
3 V sxx
recent experimental results for the ferritic steels113 in D ¼ A exp exp exp ½12
combination with data for Ni-base alloys114 indicate R RT RT
that all these factors quantitatively affect the oxidation in which DH * and DS * are the activation enthalpy and
rates, however, they do not seem to be the main reason entropy, respectively, DV * is the activation volume, and
for the oxidation rate dependence on specimen A is a pre-exponential constant. Additionally, there is
1E−10
1E−11
Kp (g2 cm–4 s–1) 900 ⬚C
1E−12
800 ⬚C
1E−13
1E−14
0.10 0.1 1 10
Figure 26 Measured Kp values as a function of specimen thickness during cyclic oxidation of high-Cr ferritic steel
Crofer 22 A in air. Reproduced from Huczkowski, P.; Quadakkers, W. J. Effect of geometry and composition of Cr steels on
oxide scale properties relevant for interconnector applications in solid oxide fuel cells (SOFCs), Report Forschungszentrum
Jülich, Energy Technology; 2007; Vol. 65, ISBN 978-3-89336-484-8.
the possibility of a stress gradient over the oxide scale decreasing component thickness. Studies on steel
whereby both cations and anions are transported compositions such as Crofer 22H which were mod-
from regions where the stress is compressive to ified by adding the solid solution strengthening ele-
regions where the stress is tensile resulting in a net ment tungsten in combination with the Laves phase
transport of oxide into the tensile regions. The forming element niobium, indeed showed that, espe-
vacancy concentration can be written as cially the oxidation rates of thin components of a
few tenths of a millimeter at 800 C were markedly
sh V F
CV ¼ CV0 exp ½13 reduced compared to corresponding specimens made
RT of high-purity, high-chromium ferritics (Figure 27(a)).
where C0v is the equilibrium concentration of vacan- A further beneficial property of this type of creep-
cies in the unstressed state and DVF is the volume resistant steels of the type Crofer22H (Table 4),119,120
change associated with the formation of the relevant is that the Laves phase was found to dissolve substantial
point defect. Of course, in the case of oxidation, the rate amounts of silicon,121 an element which is commonly
is controlled by the faster diffusing ion. An estimate of added in minor amounts as deoxidant of the alloy melt.
the reduction caused by the effect of an assumed com- As explained above, these minor additions (commonly
pressive stress116–118 revealed for chromia that the a few tenths of a percent) of silicon in conventional
effect of stress on the defect concentrations may be alloys result, during high temperature exposure, in
more significant than the direct effect on diffusivity.114 undesired internal oxidation and/or formation of a
This explanation for the effect of specimen thick- silica sublayer beneath the chromia scale. The incor-
ness on oxidation rate implies that occurrence of the poration of silicon in the Laves phase leads to a
effect depends on the extent of growth stress devel- decreased activity of this element in the steel matrix
opment in the scale. This explains the observation with the result that no internal oxidation of silicon
that the specimen thickness dependence differs from occurs (Figure 27(b)) if the niobium/silicon ratio is
environment to environment and is, for instance, far carefully adjusted.119 In parallel, the incorporation of
less pronounced in anode than in cathode gas.86 The silicon stabilizes the Laves phase with the result that
explanation also implies that an increase in creep the adverse effect of niobium on the chromia growth
strength of the ferritic steel would decrease the ten- rate is suppressed. This finding allowed the definition
dency for the oxidation rates to increase with of ferritic steel compositions that contained a minor
7
Crofer 22 APU Crofer 22 H
6 Air Air
0.5 mm Ar–4%H2–20%H2O Ar–4%H2–20%H2O
1.0 mm
5
Oxide thickness (μm)
2.0 mm
3
0.3 mm 0.5 mm 2.0 mm
1.0 mm
2
(a) 0
Oxide scale
Oxide scale
SiO2
Crofer 22APU + Si addition

10 μm 10 μm
Oxide scale
Oxide scale
Nb-rich oxide
+ Nb addition + Nb/Si addition

10 µm 10 μm
(b)
Figure 27 (a) Oxide thicknesses on Crofer 22 APU and Crofer 22 H after 1000 h discontinuous oxidation at 800 C (100 h
cycles) in different gases. (b) Effect of Si, Nb, and combined Si/Nb addition on oxide scale formation (1000 h, 900 C, air) on
Crofer22 type materials. Alloy designated as ‘Nb/Si addition’ corresponds to Crofer 22H in Table 4. Reproduced from
Froitzheim, J. Ferritic steel interconnectors and their interactions with Ni base anodes in solid oxide fuel cells (SOFC),
RWTH Aachen, Report Forschungszentrum Jülich, Energy and Environment, Vol. 16, 2008; ISBN 978-3-89336-540-1.
amount of silicon which are desired from the viewpoint oxidation properties could be obtained for components
of alloy manufacturing by conventional methods, but of a few millimeter thickness which were similar to
did not exhibit the disadvantageous effect of silicon on those of the high-purity steels thus allowing to
the oxide scale formation mechanisms. In this way, decrease the alloy manufacturing costs compared to
those of the high-purity, single-phase ferritic steels. 122 is known to be very fast131 but may be substantially
For components with a thickness of a few tenths of a reduced by presence of protective, perfectly gas-
millimeter, this creep resistant steel showed lower oxi- tight surface oxide layers of chromia.131 The modi-
dation rates than the high-purity steel. fied air side oxidation behavior of ferritic steels
under dual atmosphere conditions is similar to that
found when such steels are exposed to wet air88,132
1.19.14 Oxidation Behavior of or wet oxygen.133 The extent of enhanced attack
Ferritic Steels Under Dual Atmosphere depends on steel composition, interconnect thickness
Conditions as well as exact composition of the anode gas envi-
ronment and tends to be reduced by a preoxidation
Most laboratory oxidation testing of candidate mate- treatment.
rials for interconnect application is based on short-
or long-term exposures in cathode gas and various
types of anode gases. However, in reality, the metallic 1.19.15 Interaction of
interconnect is simultaneously exposed to air at the Chromia-Forming Interconnects with
cathode side and fuel at the anode side and thus Cathode Side Materials
experiences a chemical potential gradient from the
fuel side to the air side. A number of investigations In evaluating the growth rates and the electronic
have shown that under such conditions, the oxidation properties of the surface scales of metallic intercon-
behavior of metallic interconnects on the cathode nectors, it has to be borne in mind that parts of the
side may differ from that observed during normal interconnects are at the cathode side in direct contact
air oxidation.123,124 Especially, ferritic stainless steels with the cathode or with ceramic contact materials.
with relatively low Cr contents of less than 20%, The cathode and contact materials are in most
under dual atmosphere conditions, show a tendency cases La-based perowskites. Due to interaction
to form oxide scales with less protective properties between these compounds and the interconnect,
than formed during normal air oxidation. The 17% two important processes can occur: a change of the
Cr ferritic steel AISI 430 exhibited hematite nodules oxide composition on the interconnect surface and/or
on the air side of an air–hydrogen sample during a change in the alloy composition in the surface-near
isothermal exposure at 800 C, whereas no hematite region of the interconnect, mainly due to Cr trans-
formation was seen on the air–air sample. The oxida- port into the perowskite.106,134 These processes can
tion behavior on the hydrogen side of the air–hydrogen change the contact resistance as well as the oxidation
sample was similar to that on the hydrogen–hydrogen resistance of the interconnect.134,135 With perowskite
sample. The detrimental effects of the dual exposures type contact materials commonly used in SOFCs,
are far less pronounced in the case of steels with higher formation of spinel layers (e.g., CoCr2O4, MnCr2O4)
Cr contents, such as Crofer 22APU or E-brite, is frequently observed at the interface between the
although a modification of the outer spinel layer was surface oxide film and the perowskite.134–138
observed125,126 and some Fe-oxide formation was found The changes in interface composition can affect
at higher exposure temperatures. Dual atmosphere the contact resistance in a positive as well as negative
effects were observed for austenitic stainless steels,127 way. The relatively pure chromia layers formed
Ni–Cr-base alloys,128 alumina-forming alloys125 as well on Cr-based (ODS) alloys are frequently reported
as pure metals such as nickel or silver.125,129 The obser- to exhibit high contact resistances which can even
vations are in qualitative agreement with observations be at 1000 C in the range O cm2.139 Larring and
of Nakagawa et al.130 when studying the oxidation Norby140 observed, for contact layers of Sr-doped
behavior of martensitic steels under conditions in La-manganite and La-cobaltite on a Cr-ODS alloy
which the inner surface of a tube was exposed to steam at 900 C, an extrapolated contact resistance of
and the other side to air. 24 m O cm2 after 10 000 h. It was claimed that the
The poorer oxidation resistance of ferritic steels formation of CoMn-spinel at the interface reduced
with relatively low Cr contents under dual atmo- the transport of Cr from the interconnector into the
sphere conditions is commonly attributed to the perowskite. This chromium transport has been shown
transport of hydrogen through the metal substrate to at least partly occur via volatile species,141 as
from the fuel side to the air side. Hydrogen perme- already discussed in Section 1.19.4. Due to the signif-
ation through metallic materials at high temperatures icance of this effect for cell performance, some authors
tried to select the composition of the perowskite-based Several authors found that Sr or Ca dopants added
contact layer in such a way that it not only acted as to increase the electrical conductivity of conact layers
contacting material but also as ‘gettering agent’ for the on the basis of La-manganite, -chromite, and -cobalt-
evaporating chromium species. Other authors pro- ite are easily transported from the cathode or contact
posed that a dense and gas-tight Cr-free spinel layer layer in the direction of the Cr-based or ferritic steel
can be formed on top of a protective chromia scale by interconnect surface.141,143 This not only alters the
reaction of manganese, which outwardly diffuses initial electrical conductivity of the perowskite con-
through the chromia scale on the ferritic steel, with tact layer but also that of the chromia surface scale
contact layers containing large amounts of spinel- (Figure 29) due to formation of compounds such as
forming elements such as Co or Ni (Figure 28). The SrCrO4 and/or Sr3Cr2O8.
dense, Cr-free spinel layer, formed by interdiffusion, In the case of ferritic steels, the chemical compat-
acts as barrier against vapor-phase transport of volatile ibility between alloy and perowskite mainly depends
chromium oxides and hydroxides.101,140,142 on the relative thermodynamic stability of the mixed
La
Intensity (a.u.) 1K
33.9 μm Co
Mn
Cr
0
20 μm 0 10 20 30
(a) (b) Distance (μm)
Figure 28 Interface between steel of the type FeCrMn(La/Ti) and LaCoO3 contact layer after 1000 h exposure at 800 C in
air. (a) SEM cross section and (b) element profiles measured along trace indicated in (a). Reproduced from Quadakkers, W. J.;
Piron-Abellan, J.; Shemet, V.; Singheiser, L. Mater. High Temp. 2003, 20(2), 115–127.
100
80
Alloy (La,Sr)CoO3 coating

60
Cr
40
Co
Sr
20 La
Fe
0
−40 0 40 80 120
Distance (μm)
Figure 29 Element profiles measured near interface between Cr-ODS alloy and (La,Sr)CoO3 contact layer after 500 h
exposure at 950 C in air. Reproduced from Quadakkers, W. J.; Piron-Abellan, J.; Shemet, V.; Singheiser, L. Mater. High
Temp. 2003, 20(2), 115–127.
oxides (e.g., spinels) which can form due to reaction 1.19.16 Protection Methods to
of Cr from the alloy with the B-type elements (Co, Minimize Cr Vaporization
Mn, Cr, Fe, Cu, etc.) in the perowskite contact
layer.80,134–136 From this viewpoint, it seems obvious Different strategies are used to minimize the cell deg-
that manganites and especially chromites provide the radation as a result of the Cr release from metallic
best compatibility with the ferritic steels, whereas interconnector materials in SOFCs. They include a
cobaltites and ferrites tend to result in extensive decrease in the operation temperature of the SOFC,
reaction with the ferritic steel, thereby decreasing the application of less Cr-sensitive cathode materials,
its oxidation resistance.80,144–148 Also, the presence the design of alternative multilayer interconnects, the
of Cu in the perowskite substantially enhances the development of interconnect alloys with a lower Cr
oxidation rates of interconnector steels.149 release, and the use of barrier coatings against
The effect of these interface reactions on the Cr release.
contact resistance cannot simply be correlated with It has been observed that the Cr vaporization can
the thickness of the reaction layer. The resistance be reduced by the formation of outer spinel layers
depends on the type of perowskite used and also on formed on the inner protective chromia scale on
perowskite and oxide layer morphology. The results high temperature alloys97 and therefore, ferritic
are also affected by the way in which the perowskite is interconnect steels such as Crofer 22APU, with
applied, that is, to a bare or to a preoxidized metal. defined additions of Mn and Ti in combination with
Indications were found that the dense inner chromia reactive elements (REs) such as Y, Ce, Zr, or La, have
layer frequently present at the interface with the alloy been developed (Table 4). A number of studies on Cr
possesses a higher specific resistivity than the outer vaporization of a number of these steels are avail-
mixed oxide layers (e.g., of the spinel type) formed able.40,43 The potential of the Cr retention capability
either due to oxidation of steel alloying elements or of alloys with different types of oxide layers on the
due to reaction of the chromia layer with perowskite outer surface, such as for example (Cr,Mn)3O4 layers
constituents. Several surface modifications were there- or other oxide scales, is described elsewhere.156
fore proposed to decrease the scale growth rate or to The Cr vaporization rates of ferritic stainless steels,
modify the composition of the inner chromia layer. To specifically developed for application as interconnec-
achieve this goal, a number of methods such as RE tor materials in SOFCs such as Crofer 22 APU, ZMG
implantation or application of RE-nitrate solutions 232, and IT-10 and IT-11, are very similar. These
have been used, sometimes in combination with Ni alloys form a well-adhering outer (Cr,Mn)3O4-spinel
deposition to increase chromia conductivity.66,150–153 layer that reduces Cr vaporization rates in humid air
From the viewpoint of long-term stack perfor- at 800 C by 61–75% compared to alloys which form
mance, it has to be considered that the electronic pure chromia scales such as Cr-based ODS alloys or
conductivity of actually formed reaction products E-Brite. Substantially lower oxidation rates were
alone does not govern the relevant contact resistance found for Fe-, Ni-, or Co-base alloys with relatively
of the interface between interconnector and contact low Cr contents. This type of alloys formed thicker
layer. The reaction between the perowskite and the oxides of the base material (e.g., Co-oxide on Co-
chromia-based surface scales may also lead to mor- based alloy) on top of the inner chromia layer, how-
phological changes of the perowskite layer near the ever, with the disadvantage of high oxidation rates.
interface with the surface oxide. These morphologi- For electrically insulating SOFC components,
cal changes sometimes, for example, result in sub- alumina-forming alloys are sometimes considered.
stantial void formation which increases the overall At high temperatures, these alloys at their surfaces
contact resistance. Therefore, a number of authors154 form slow-growing alumina scales which possess far
considered the use of Co–Mn or Co–Mn–Fe-spinels better protective properties and resistance against Cr
as possible contact layer materials. These spinels evaporation than chromia-forming materials. How-
were found to exhibit less extensive reactions with ever, due to the low electronic conductivity of alu-
the steel surfaces than most perowskites, thus ensur- mina, these materials do not seem to be suitable as
ing a slower increase in contact resistance during construction materials for interconnectors.
long-term cell operation. Yang et al.155 used a combi- Numerous authors aimed to reduce the Cr vaporiza-
nation of Co-based spinels and RE-oxides, the latter tion from ferritic steels by development of coating sys-
being added to improve the adherence of the oxide tems which possess sufficient electronic conductivity
scale to the steel surface. at the SOFC operating temperature.43,137,141,148,157,158
10−9
Cr vaporizaton rate (kg m−2 s−1)

10−10 Crofer (KMT) uncoated
Crofer + MnO + LCC10
Crofer + MCF + LCC10

10−11
Crofer + MCF + LCC13 (WPS)
10−12
Crofer + MCF (APS)
10−13
0 200 400 600 800 1000
Time (h)
Figure 30 Cr evaporation rate of Crofer 22 APU with and without various types of coatings during exposure in wet air at
800 C.158 LCC: La-base perowskite; MCF: Mn/Co/Fe-spinel.
The coatings mostly are perowskites layers on the basis with the chromia-based surface scales present on
of (La,Sr)MnO3, (La,Sr)CoO3, and (La,Sr)CrO3 or com- the ferritic steels. With such coatings, Cr retentions
binations of these (Figure 30). There were also attempts of more than 99% up to the maximum test time of
to use reactive coatings of La2O3 and SrO, that is, 1200 h at 800 C in humid air could be achieved. The
compounds which form protective (La,Sr)CrO3 layers coatings seem to be applicable for a wide range of
by reaction with Cr in the steel.48 The most common stainless steels and fulfill the requirements for mobile
coating techniques used to apply such protection layers and stationary long-term applications. These types of
are vacuum plasma spraying (VPS), air plasma spraying coatings were found to possess electronic conductiv-
(APS), and wet powder spraying.43,48,140,141 ities which are 1–2 orders of magnitude larger than
Sputtered ceramic Cr barrier coatings of La0.8Sr0.2 that of Cr2O3 and even tend to reduce the oxidation
CrO3, La 0.99 (Cr0.77Mg0.05Al0.18)O3, La0.80Sr0.20MnO3, of the ferritic substrate.
and La0.65Sr0.30MnO3 on substrates of Crofer 22 APU Figure 31 shows a comparison of the chromium
showed only a minor reduction of the Cr release. This release during 1000 h exposure in wet air at 800 C
is mainly related to the formation of pores due to for different types of ferritic steels with and without
crystallization of the amorphous sputter coatings dur- various types of coatings. It is obvious that suitable
ing high temperatures exposure. The LSM coatings coating compositions can effectively reduce the chro-
showed a high tendency to spallation after thermal mium release by approximately two orders of magni-
cycling, which might be prevented by a proper adjust- tude provided that they can be applied in the form
ment of the manufacturing parameters such as the of a dense layer with suitable adhesion to the metallic
thickness of the coating. In general, the practical appli- substrate.
cation of perowskite type Cr retention layers seems
to be questionable due to high costs and difficulties
with the fabrication of dense coatings. 1.19.17 Interaction of
Promising results have been obtained with coat- Chromia-Forming Interconnects
ings on the basis of spinel type oxides, for example, with Anode Side Materials
on the basis of Co and/or Ni. The coatings were
either directly applied in the form of respective oxi- At the anode side, the interconnect is in direct con-
des or as metallic coatings of Co or Ni with a thick- tact with the Ni/ZrO2 cermet anode or, in most cases,
ness of about 10 mm. The metallic layers are with a Ni wire mesh which provides the electrical
transformed into oxides during high temperature contact between the interconnect and the anode.
exposure and form the spinel layers by reactions If, for instance, the Ni net is spot welded to the
10−3
10−4
Cr release (g cm−2)
10−5
10−6
10−7
--
+L 0
B 11+ 3
2 A +M F ( A 3
+M +L 10
LC 0
r2 NB
ro 17 T 1
IT Mn 4+L ferH
+ LC F
F1 n C PU
+M F ( )
)
(A )
ro MC er+ )
PU C PS
PS
C PS
1
f e IT 1 M C C 1
+M O 4 C 1
S
F+ 1
F IT 1
F+ MC
C
C
11 O4 C
C CC
P
+M M
2A
A
C
(A
C
7+ O ro
F
F
fe
f
T N IT
r+ o r
fe C
3
3
11 3
C
H
r2 1
M
r
fe
r+
IT
fe
ro
7
C
F1
ro
C
C
ro
C
Figure 31 Total Cr release during 1000 h exposure in air with standard humidity of 60% rel. at 25 C. LCC: Perowskite type
coating; MCF: Mn/Co/Fe-spinel coating. Reproduced from Trebbels, R. Ph.D. Thesis, RWTH Aachen, 2009.
interconnector, excellent contact resistance values

are initially observed. In the case of a Cr-based inter-
connector, however, interdiffusion can result in
Ni-mesh
excessive Kirkendall-void formation at the Cr–Ni
interface, eventually resulting in deterioration of the
contact. Ni plating of the interconnector has been
applied to reduce this problem,135 however, extensive
outward diffusion of Cr into the Ni with subsequent
chromia formation on the Ni layer has been found to
Cr-rich
occur after long-term exposure. oxide
Interdiffusion between a ferritic steel interconnect Austenite
and the spot-welded nickel wire mesh will lead to Ni Crofer 22 APU
transport into the steel resulting in local austenite 200 μm
formation (Figure 32), thereby changing the steel Figure 32 Ni mesh in direct contact with the interconnect
oxidation resistance as well as the thermal expansion plate of an SOFC stack after 3000 h at 800 C. Reproduced
coefficient. As a result of interdiffusion these prop- from Brandner, M.; Bram, M.; Froitzheim, J.; Buchkremer,
erty changes will be of special concern if very thin H.P.; Stover, D. Solid State Ionics 2008, 179(27–32),
1501–1504.
interconnects are being used.
Vice versa, transport of Fe, Cr, and other steel
constituents into the wire mesh renders the latter as (Figure 33). A number of externally applied diffusion
an alloy which will tend to form surface oxide layers barrier layers have been proposed and quite
in the anode side environment resulting in a promising results were obtained with sputtered
decreased catalytic activity.159 ceria layers.162,163 The intrinsic electronic conductiv-
Preoxidation layers have been examined as possi- ity of ceria in the low-pO2 anode gas is substantially
ble barriers to minimize interdiffusion processes smaller than that of Ni, but it is sufficiently high to
between the ferritic steel and the Ni wire mesh or obtain low contact resistance values because layers of
the Ni/ZrO2-anode. However, the preformed oxide only a few micrometer thickness appeared to be
scales tended to become dissociated at the Ni–steel sufficient to virtually completely suppress the inter-
interfaces,160,161 so after longer exposure time, the diffusion phenomena.87 During long-term operation,
unwanted interdiffusion processes appeared again however, the outward growing Cr–Mn-spinel oxide
1.19.18 Interaction of Metallic

Interconnects with Glass Sealants
Ni-coating
Mounting
Oxide scale Most glass sealants used in solid oxide fuel cells are
based on the system, barium-calcium-aluminosilicate
(BCAS).164–167 The glasses frequently contain further
Preoxidation layer minor amounts of other oxides, for instance, to
Austenite increase the wetting properties. The extent of the
reaction of high-chromium ferritic steels with the
glass ceramics depends on the exposure conditions
Crofer 22 APU and the distance of the interface of sealing glass and
100 μm
ferritic steel to the ambient air.168–170 Near the triple-
phase boundary, steel/glass/air volatile chromium
Figure 33 Cross section of sample with a Ni coating on species react with BaO, thereby forming BaCrO4
(100 h) preoxidized Crofer 22 APU after 1000 h exposure in (Figure 35). In the interior seal regions, hardly any
Ar–4H2–2H2O at 800 C showing austenite formation,
dissociation of the preformed oxide layer, and chromia
access of oxygen from the air is possible and thus only
formation on top of the Ni coating. Reproduced from a very minor reaction with the steel surface occurs
Froitzheim, J.; Niewolak, L.; Brandner, M.; Singheiser, L.; and the chromium tends to slightly dissolve into the
Quadakkers, W. J. J. Fuel Cell Sci. Technol. in press 6(4). glass. The extent of reaction depends on the steel
composition. Crofer 22 APU was found to exhibit
better chemical compatibility and bonding with the
BCAS-based glass-ceramics than most other ferritic
Mounting steels due to the growth of its specific oxide scale
Ni-coating on the alloy during high temperature exposure.170
CeO2 A further suppression of interface reactions and
Cu
BaCrO4 formation was obtained when an alumina-
forming alloy was used.169 Thus, one approach to
improve the sealing effectiveness is to engineer the
Oxide scale alloy surface, for example, by local aluminizing171 in
places where electrical conductivity is not required.
More severe reactions may occur if the glass con-
tains minor amounts of wetting agents in the form of
thermodynamically less stable oxides such as PbO or
Crofer 22 APU
ZnO. Depending on the ferritic steel composition, the
10 μm use of such glasses may result in rapid short circuiting
of SOFC stacks by excessive iron oxide formation on
Figure 34 Cross section of multilayer ceria/Cu coating
between preoxidized Crofer 22 APU and Ni layer after
the ferritic steels near the triple-phase boundary air/
1000 h of exposure in Ar–4H2–2H2O at 800 C. Reproduced glass/metal.172–175 It was however found that the
from Froitzheim, J.; Niewolak, L.; Brandner, M.; Singheiser, trigger of this detrimental effect occurred near the
L.; Quadakkers, W. J. J. Fuel Cell Sci. Technol. in press 6(4). triple-phase boundary at the anode side of the stack.
Here, the less stable oxides were reduced to, for
example, metallic Pb, which possesses a substantial
on the steel deteriorated the chemical contact vapor pressure (104 bar at 800 C) at the prevail-
between oxide scale and ceria layer. An improvement ing high service temperature of 800–900 C. As con-
of the electrical contact could be obtained by apply- firmed by laboratory tests,159 the presence of Pb
ing a duplex layer of ceria and copper on a preox- vapor resulted in severe grain boundary attack in
idized ferritic steel (Figure 34). The high oxygen the high-chromium ferritic steels accompanied by
solubility in Cu prevented the dissociation of the excessive internal oxidation of chromium (Figure 36).
oxide layer and the geometrical deterioration of the Due to the low pO2 in the anode side gas, no external
contact was reduced because the geometrical irregu- iron oxide formation occurred. However, the internal
larities were compensated by plastic deformation of oxidation resulted in a large volume increase and thus
the Cu layer.87 in a local bulging of the steel. In the actual SOFC stack,
Air side
Glass sealing
BaCrO4
SiO2
Crofer 22 H Crofer 22 H
20 μm 10 μm
FZJ-IEF 2008 EHT = 20.00 kV Detector = QBSD WD = 24 mm FZJ-IEF 2008 EHT = 20.00 kV Detector = QBSD WD = 25 mm
Figure 35 SEM/BSE pictures of glass/metal microstructure on cathode side (left) and interseal region (right) of SOFC stack
after 1000 h operation at 800 C. Reproduced from Shemet, V.; Quadakkers, W. J. Unpublished results.
Glass
Glass
Internal oxidation
Ferritic steel 10 μm Ferritic steel 10 μm

(a) (b)
Figure 36 Interaction of ferritic steel with PbO-containing BCAS glass at 800 C in (a) air and (b) Ar–H2–H2O. Reproduced
from Shemet, V.; Quadakkers, W. J. Unpublished results.
this initiated a cracking at/near the steel–glass interface scale towards the steel surface resulting in the above
which allowed the Pb vapor-triggered internal oxida- described enhanced oxidation attack. High-purity
tion to propagate in the direction of the triple-phase steels (for instance Crofer 22APU) with very low Al
boundary glass/steel/air. When reaching this point, the and Si contents in the range 0.01% were less prone to
chromium-depleted steel matrix started to form rapidly Pb vapor attack than steels (such as, e.g., Crofer 22A)
growing iron-base oxide which eventually ‘bridged’ the with more common Al/Si contents of 0.1%.
small distance between two interconnector plates. The A further effect imparted by an indirect interac-
high electronic conductivity of the Fe-base oxide in this tion of the steel with the ceramic sealant is related to
way resulted in a short circuiting of the stack. the high silicon content of the glass. In dry and wet
The rate at which this Pb vapor-induced reaction gases, silicon may form a number of volatile oxides
occurred, appeared to depend on the amount of the and (oxy-)hydroxides
minor alloying additions Si and Al in the ferritic steel.
Concentrations of a few tenths of a percent results, as 2SiO2 ¼ 2SiOðgÞ þ O2 ½14
explained in Section 1.19.9, in formation of internal
ðgÞ
silicon- and aluminum oxides. The minor volume SiO2 þ 2H2 OðgÞ ¼ SiðOHÞ4 ½15
increase imparted by this effect leads to outward
metal flow, noticeable by formation of metallic ðgÞ
SiO2 þ H2 OðgÞ ¼ SiOðOHÞ2 ½16
nodules in the surface oxide scale86,102 and is accom-
panied by formation of micro cracks. The latter can act It is apparent that reaction [14] is only significant in
as channels for transport of Pb vapor through the oxide low-pO2 environments. Reactions [15] and [16]
are independent of pO2 and the vapor pressure of Yokokawa, H., Singhal, S. C., Eds.; The Electrochemical
Society Proceedings Series: Pennington, NJ, 2001;
the volatile species is proportional to p2(H2O) and p 130, PV 2001-16.
p(H2O), respectively.176 Due to these volatile species, 20. Malkow, T.; Crone, U. v. d.; Laptev, A. M.; Koppitz, T.;
silicon-containing oxides are frequently found at the Breuer, U.; Quadakkers, W. J. In Solid Oxide Fuel Cells
(SOFC V); Stimming, U., Singhal, S. C., Tagawa, H.,
external surface at the anode side of ferritic steel inter- Lehnert, W., Eds.; The Electrochemical Society
connects after long-term stack operation, mainly in the Proceedings Series: Pennington, NJ, 1997; p 1245,
form of Mn-silicate.177 For further details on the for- PV 97-40.
21. Thierfelder, W.; Greiner, H.; Köck, W. In Solid Oxide Fuel
mation of volatile silicon-containing species, see the Cells (SOFC V); Stimming, U., Singhal, S. C., Tagawa, H.,
chapter on ‘Oxidation in Steam and Steam/Hydrogen Lehnert, W., Eds.; The Electrochemical Society
Environments.’ Proceedings Series: Pennington, NJ, 1997; p 1306,
PV 97-40.
22. Schmidt, H.; Brückner, B.; Fischer, K. In Solid Oxide Fuel
Cells (SOFC IV); Dokiya, M., Yamamoto, O., Tagawa, H.,
References Singhal, S. C., Eds.; The Electrochemical Society
Proceedings Series: Pennington, NJ, 1995; p 869,
PV 95-1.
1. Singhal, S. C. MRS Bull. 2000, 25(3), 16–21. 23. Quadakkers, W. J.; Greiner, H.; Köck, W. In Proceedings
2. Minh, N. Q. J. Am. Ceram. Soc. 1993, 176(3), 563–588. of the First European Solid Oxide Fuel Cell Forum
3. Steele, B. C. H. Solid State Ionics 2000, 134, 3–20. Bossel, U., Ed.; European SOFC Forum, Dr. Ulf Bossel,
4. Singhal, S. C. In Solid Oxide Fuel Cell (SOFC-V); Morgenacher Str. 2F, CH-5452: Oberrohrdorf,
Stimming, U., Singhal, S. C., Tagawa, H., Lehnert, W., Switzerland, 1994; Vol. 2, p 525.
Eds.; The Electrochemical Society Proceedings Series: 24. Quadakkers, W. J.; Norton, J. F.; Penkalla, H. J.;
Pennington, NJ, 1997; p 88, PV 97-40. Breuer, U.; Gil, A.; Rieck, T.; Hänsel, M. In Proceedings of
5. Stolten, D.; Späh, R.; Schamm, R. In Solid Oxide Fuel the 3rd International Conference on Microscopy of
Cell (SOFC-V); Stimming, U., Singhal, S. C., Tagawa, H., Oxidation, Cambridge, 16–18 September 1996;
Lehnert, W., Eds.; The Electrochemical Society Newcomb, S. B., Little, J. A., Eds.; The Institute of
Proceedings Series: Pennington, NJ, 1997; p 88, PV 97-40. Materials, 1996; p 221.
6. Quadakkers, W. J.; Piron-Abellan, J.; Shemet, V.; 25. Graham, M.; Hussey, R. Oxid. Met. 1995, 44, 339–374.
Singheiser, L. Mater. High Temp. 2003, 20(2), 115–127. 26. Cotell, C.; Yurek, G.; Hussey, R.; Mitchel, D.; Graham, M.
7. Köck, W.; Martinz, H.; Greiner, H.; Janousek, M. In Solid J. Electrochem. Soc. 1987, 134, 1871.
Oxide Fuel Cells (SOFC IV); Dokiya, M., Yamamoto, O., 27. Quadakkers, W. J.; Holzbrecher, H.; Briefs, K. G.;
Tagawa, H., Singhal, S. C., Eds.; The Electrochemical Beske, H. In Proceedings of the Europ. Coll. on the Role
Society Proceedings Series: Pennington, NJ, 1995; of Active Elements in the Oxidation Behaviour of High
p 841, PV 95-1. Temperature Alloys, 12–13 December 88; Petten, N. L.,
8. Yang, Z. Int. Mater. Rev. 2008, 53(1), 39–54. Lang, E., Ed.; European Communities, 1988; p 155.
9. Stöver, D.; Diekmann, U.; Flesch, U.; Kabs, H.; 28. Hou, P.; Stringer, J. Oxid. Met. 1992, 38, 323.
Quadakkers, W. J.; Tietz, F.; Vinke, I. In Solid Oxide Fuel 29. Whittle, D.; Stringer, J. Philos. Trans. Roy. Soc. London A
Cells (SOFC VI); Singhal, S. C., Dokiya, M., Eds.; The 1980, 295, 309.
Electrochemical Society Proceedings Series: 30. Golightly, F.; Stott, H.; Wood, G. Oxid. Met. 1976, 10(3),
Pennington, NJ, 1999; p 812, PV 99-19. 163.
10. Lai, G. High Temperature Corrosion of Engineering 31. Pint, B. Oxid. Met. 1996, 45, 1.
Alloys; ASM International: Materials Park, OH, 1990. 32. Quadakkers, W. J.; Greiner, H.; Köck, W.;
11. Kofstad, P. High Temperature Corrosion; Elsevier Buchkremer, H. P.; Hilpert, K.; Stöver, D. In Proceedings
Applied Sciences/Chapman and Hall: London, 1988. of the 2nd European Solid Oxide Fuel Cell Forum;
12. England, D.; Virkar, A. V. J. Electrochem. Soc. Thorstensen, B., Ed.; European SOFC Forum, Dr. Ulf
1999, 146(9), 3196. Bossel, Morgenacher Str. 2F, CH-5452: Oberrohrdorf,
13. Wersing, W.; Ivers-Tiffee, E.; Schiessl, M.; Greiner, H. Switzerland, 1996; Vol. 2, p 297.
In Proceedings of the International Symposium 33. Brady, M. P.; Wright, I. G.; Anderson, I. M.; Sikka, V. K.;
on Solid Oxide Fuel Cells; Nagoya, Japan, 13–14 Ohriner, E. K.; Walls, C.; Westmoreland, G.;
November 1989; Yamamoto, O., Dokiya, M., Tagawa, H., Weaver, M. L. In Ductilization of Cr via Oxide Dispersions;
Eds.; Scince House: Tokyo, Japan, 1989; p 33. Proceedings of the 15th International Plansee
14. Yamamoto, O. Electrochim. Acta 2000, 45, 2423. Symposium; Plansee, AG, Reutte, Kneringer, G.,
15. Kofstad, P. In Proceedings of the 2nd European Solid Rodhammer, P., Wildner, H., Eds.; 2001;
Oxide Fuel Cell Forum; Thorstensen, B., Ed.; European pp 108-1–108-13.
SOFC Forum: Switzerland, 1996; Vol. 2, p 479. 34. Badwal, S. P. S.; Deller, R.; Foger, K.; Ramprakash, Y.;
16. Kofstad, P. In Proceedings of the 17th Risø International Zhang, J. P. Solid State Ionics 1997, 99, 297.
Symposium on Materials Science: High Temperature 35. Hilpert, K.; Das, D.; Miller, M.; Peck, D. H.; Weiß, R.
Electrochemistry, Ceramics and Metals; Poulsen, F., J. Electrochem. Soc. 1996, 143, 3642.
Bonanos, N., Linderoth, S., Mogenson, M., Zachau- 36. Jiang, S. P.; Zhang, J. P.; Apateanu, L.; Froger, K.
Christiansen, B., Eds.; Riso National Laboratory: J. Electrochem. Soc. 2001, 148, 447.
Roskilde, Denmark, 1996; p 55. 37. Odell, S. P.; Ding, G. L.; Tewari, S. N. Metall. Mater.
17. Kofstad, P.; Bredesen, R. Solid State Ionics 1992, 52, 69. Trans. A 1999, 30A, 2159–2165.
18. Matsuzaki, Y.; Yasuda, I. Solid State Ionics 2000, 132, 38. Ebbinhaus, B. B. Comb. Flame 1993, 93, 119.
271. 39. Das, D.; Miller, M.; Nickel, H.; Hilpert, K. In Proceedings
19. Yasuda, I.; Baba, Y.; Ogiwara, T.; Yakabe, H.; of the First European Solid Oxide Fuel Cell Forum;
Matsuzaki, Y. In Solid Oxide Fuel Cells (SOFC VII); Bossel, U., Ed.; European SOFC Forum, Dr. Ulf Bossel,
Morgenacher Str. 2F, CH-5452: Oberrohrdorf, 65. Hou, P.; Brown, I.; Stringer, J. Nucl. Instrum. Meth. B
Switzerland, 1994; Vol. 2, p 703. 1991, 59/60, 1345.
40. Stanislowski, M.; Wessel, E.; Hilpert, K.; Markus, T.; 66. Quadakkers, W. J.; Breuer, U.; Tyagi, A.; Gil, A.;
Singheiser, L. JES 2007, 154(4), A295–A306. Stroosnijder, M. F.; Becker, S.; Hänsel, M. Report,
41. Opila, E. J.; Myers, D. L.; Jacobson, N. S.; et al. J. Phys. Forschungszentrum Jülich, Jülich, FRG, Jül-1372, p. 97,
Chem. A 2007, 111, 1971–1980. ISSN 0944–2952.
42. Yungman, V. S.; Medvedev, V. A.; Veits, I. V. and 67. Park, J. H.; Natesan, K. Oxid. Met. 1990, 33(1/2), 31.
Bergman, G. A.; Eds. IVTANTHERMO – 68. Holt, A.; Kofstadt, P. Solid State Ionics 1994, 69, 137.
A Thermodynamic Database and Software System for the 69. Holt, A.; Kofstadt, P. Solid State Ionics 1999, 117, 21.
Personal Computer; CRC Press/Begell House: Boca 70. Holt, A.; Kofstadt, P. Solid State Ionics 1997, 100, 201.
Raton, FL, 1993. 71. Holt, A.; Kofstad, P. Solid State Ionics 1994, 69, 127.
43. Gindorf, C.; Singheiser, L.; Hilpert, K. Steel Res. 2001, 72. Liu, H.; Stack, M.; Lyon, S. Solid State Ionics 1998, 109,
72, 528. 247.
44. Gindorf, C.; Hilpert, K.; Nabielek, H.; Singheiser, L.; 73. Huczkowski, P.; Christiansen, N.; Shemet, V.;
Ruckdäschel, R.; Schiller, G. In Proceedings of Niewolak, L.; Piron-Abellan, J.; Singheiser, L.;
the Fourth European Solid Oxide Fuel Cell Forum; Quadakkers, W. J. Fuel Cells 2006, 2, 93–99.
McEvoy, J., Ed.; European SOFC Forum, Dr. Ulf Bossel, 74. Hagel, W. C.; Seybolt, A. U. J. Electrochem. Soc. 1961,
Morgenacher Str. 2F, CH-5452: Oberrohrdorf, 108(12), 1146–1152.
Switzerland, 2000; Vol. 2, p 845. 75. Nagai, H.; Fujikawa, T. Trans. JIM 1983, 24(8), 581.
45. Batawi, E.; Plas, A.; Straub, W.; Honegger, K.; 76. Su, M. Y. Diss. Abstr. Int. 1988, 48, 3049.
Diethelm, R. In Solid Oxide Fuel Cells (SOFC VI); 77. Crawford, J. A.; Vest, R. W. J. Appl. Phys. 1964, 35(8),
Singhal, S. C., Dokiya, M., Eds.; The Electrochemical 2413–2418.
Society Proceedings Series: Pennington, NJ, 1999; 78. Nagai, H.; Ohbayashi, K. J. Am. Ceram. Soc. 1989, 72,
p 767, PV 99-19. 400.
46. Quadakkers, W. J. German Patent DE 195 47 699 C2; 79. Nagai, H.; Ishikawa, S.; Amano, N. Trans. JIM 1985,
1995. 26(10), 753.
47. Quadakkers, W. J.; Baumanns, F.; Nickel, H. German 80. Malkow, Th.; Quadakkers, W. J.; Singheiser, L.;
Patent DE 44 10 711; 1995. Nickel, H. Report Forschungszentrum, Jülich, Jülich,
48. Linderoth, S. Mater. Processes 1998, 17, 217. FRG, Jül-3589,1998; ISSN 0944-2952.
49. Linderoth, S.; Hendriksen, P.; Mogensen, M.; 81. Ennis, P. J.; Quadakkers, W. J. In Proceedings of the
Langvad, N. J. Mater. Sci. 1996, 31(9), 5077. Conference High Temperature Alloys – Their Exploitable
50. Hänsel, M.; Quadakkers, W. J.; Singheiser, L.; Nickel, H. Potential, JRC Petten, NL, 15–17 October 1985;
Report Forschungszentrum Jülich, Jülich FRG, Jül-3583, Marriot, J. B., Merz, M., Nihoul, J., Ward, J. O., Eds.;
1998; ISSN 0944-2952. Elsevier Applied Science: London, 1987; p 465.
51. Rahmel, A.; Tobolski, J. Corros. Sci. 1965, 5, 333. 82. Buchkremer, H. P.; Diekmann, U.; Stöver, D. In
52. Young, D. J. Mater. Sci. Forum 2008, 595–598, Proceedings of the 2nd European Solid Oxide Fuel Cell
1189–1197. Forum, Thorstensen, B., Ed.; European SOFC Forum,
53. Zurek, J.; Young, D. J.; Essuman, E.; Hänsel, M.; Dr. Ulf Bossel, Morgenacher Str. 2F, CH-5452:
Penkalla, H. J.; Niewolak, L.; Quadakkers, W. J. Mater. Oberrohrdorf, Switzerland, 1996; Vol. 2, p 221.
Sci. Eng. A 2008, 477, 259–270. 83. De Haart, L.; Vinke, I. C.; Janke, A.; Ringel, H.; Tietz, F. In
54. Miyake, Y.; Yasuo, T.; Taniguichi, Y.; Kadowaki, M.; Solid Oxide Fuel Cells (SOFC VII); Yokokawa, H.,
Kawamura, H.; Saitoh, T. In Proceedings of the 4th Singhal, S. C., Eds.; The Electrochemical Society
International Symposium on Solid Oxide Fuel Cells Proceedings Series: Pennington, NJ, 2001; p 111, PV
(SOFC IV); Dokiya, M., Yamamoto, O., Tagawa, H., 2001-16.
Singhal, S. C., Eds.; The Electrochemical Society 84. Horita, T.; Xiong, Y.; Yamaji, K.; Sakai, N.; Yokokawa, H.
Proceedings Series: Pennington, NJ, 1995; p 100, In Proceedings of the 5th European Solid Oxide Fuel Cell
PV 95-1. Forum, Huijsmans, J., Ed.; European SOFC Forum, Dr.
55. Quadakkers, W. J.; Hänsel, M.; Rieck, T. Mater. Corros. Ulf Bossel, Morgenacher Str 2F, CH-5452: Oberrohrdorf,
1998, 49, 252. Switzerland, 2002; Vol. 1, p 401.
56. Crone, U. v. d.; Hänsel, M.; Quadakkers, W. J.; 85. Piron-Abellan, J.; Quadakkers, W. J. Development of
Vaßen, R. Fresenius J. Anal. Chem. 1997, 358, 230. ferritic steels for application as interconnect materials for
57. Atkinson, A.; Taylor, R. In Transport in Non- intermediate temperature solid oxide fuel cells (SOFCs),
Stoichiometric Compounds; Simkovich, G., Stubican, G., Report Forschungszentrum Jülich, Jül-4170, March
Eds.; NATO ASI Series, Series B: Physics Col. Vol. 129, 2005; ISSN 0944-2952.
Plenum Press: New York, 1984; p 285. 86. Huczkowski, P.; Ertl, S.; Piron-Abellan, J.; Christiansen,
58. Sabioni, A.; Huntz, A.; Millot, F.; Monty, C. Philos. Mag. A N.; Hoefler, T.; Shemet, V.; Singheiser, L.; Quadakkers,
1992, 66, 351. W.J. Mater. High Temp. 2005, 22(3/4), 253–262.
59. Sockel, H.; Saal, B.; Heilmayer, M. Surf. Interf. Anal. 87. Froitzheim, J. Ferritic steel interconnectors and their
1988, 12, 531. interactions with Ni base anodes in solid oxide fuel cells
60. Gil, A.; Penkalla, H. J.; Hänsel, M.; Norton, J.; Köck, W.; (SOFC), RWTH Aachen, Report Forschungszentrum
Quadakkers, W. J. IX Conference on Electron Jülich, Energy and Environment, Vol. 16, 2008; ISBN
Microscopy of Solids, 6–9 May 1996, Krakow-Zakopane, 978-3-89336-540-1.
Poland 1996, Fotobit: Krakow, Poland; p 441. 88. Fujii, C. T.; Meussner, R. A. Corros. Iron & Steel 1964,
61. Quadakkers, W. J. J. Phys. IV 1993, 3(part I), 177. 111, 1215–1221.
62. Quadakkers, W. J. Werkst. Korros. 1990, 41, 659. 89. Essuman, E.; Meier, G. H.; Zurek, J.; Hänsel, M.;
63. Bongartz, K.; Quadakkers, W. J.; Pfeifer, J. P.; Becker, Quadakkers, W. J. Oxid. Met. 2008, 69, 143–162.
J. S. Surf. Sci. 1993, 292, 196. 90. Grabke, H. J. Mater. Corros. 2003, 54(10), 736–746.
64. Quadakkers, W. J. Mater. Sci. Forum 2001, 91. Young, D. J. Mater. High Temp. 2007, 24(3), 225–232.
369–372, 77. 92. Huczkowski, P.; Quadakkers, W. J. Unpublished results.
93. Dulieu, D.; Cotton, J.; Greiner, H.; Honegger, K.; 550 bis 650 C, Report Forschungszentrum Jülich, Jülich,
Scholten, A.; Christie, M.; Seguelong, T. In 3rd European FRG, Jül-3883, ISSN 0944–2952, June 2001.
SOFC Forum; Stevens, P., Ed.; European Fuel Cell 109. Huczkowski, P.; Christiansen, N.; Shemet, V.; Piron
Forum: Oberrohrdorf, Switzerland, CH-5452, Abellan, J.; Singheiser, L.; Quadakkers, W. J. Mater.
1998; p 447. Corros. 2004, 55(11), 825–830.
94. Ueda, M.; Taimatsu, H. In Proceedings of the Fourth 110. Quadakkers, W. J.; Bongartz, K. Werkst. Korros. 1994,
European Solid Oxide Fuel Cell Forum, McEvoy, J., Ed.; 45, 232–241.
European SOFC Forum, Dr. Ulf Bossel, Morgenacher Str. 111. Gurrappa, I.; Weinbruch, S.; Naumenko, D.;
2F, CH-5452: Oberrohrdorf, Switzerland, 2000; Vol. 2, Quadakkers, W. J. Mater. Corros. 2000, 51, 224–235.
p 837. 112. Huczkowski, P.; Christiansen, N.; Shemet, V.; Piron
95. Ueda, M.; Kadowaki, M.; Taimatsu, H. Mat. Trans. JIM Abellan, J.; Singheiser, L.; Quadakkers, W. J. J. Fuel Cell
2000, 41(2), 317–322. Sci. Technol. 2004, 1, 30.
96. Quadakkers, W. J.; Malkow, T.; Piron-Abellan, J.; 113. Huczkowski, P.; Quadakkers, W. J. Effect of geometry
Flesch, U.; Shemet, V.; Singheiser, L. In Proceedings of and composition of Cr steels on oxide scale
the Fourth European Solid Oxide Fuel Cell Forum, properties relevant for interconnector applications in
McEvoy, J., Ed.; European SOFC Forum, Dr. Ulf Bossel, solid oxide fuel cells (SOFCs), Report
Morgenacher Str. 2F, CH-5452: Oberrohrdorf, Forschungszentrum Jülich, Energy Technology; 2007;
Switzerland, 2000; Vol. 2, p 827. Vol. 65, ISBN 978-3-89336-484-8.
97. Piron-Abellan, J.; Shemet, V.; Tietz, F.; Singheiser, L.; 114. Zurek, J.; Meier, G. H.; Essuman, E.; Hänsel, M.;
Quadakkers, W. J. In Solid Oxide Fuel Cells (SOFC VII); Singheiser, L.; Quadakkers, W. J. J. Alloys Comp. 467
Yokokawa, H., Singhal, S. C., Eds.; The Electrochemical (2009) 450–458.
Society Proceedings Series: Pennington, NJ, 2001; 115. Mehrer, H. Defect Diffus. Forum 1996, 57, 129–130.
p 811, PV 2001-16. 116. Huntz, A.-M.; Daghigh, S.; Piant, A.; Lebrun, J. L. Mater.
98. Bredesen, R.; Kofstad, P. In Proceedings of the 17th Risø Sci. Eng. A 1998, 248, 44.
International Symposium on Materials Science: High 117. Galerie, A.; Toscan, F.; Dupeux, M.; Mougin, J.;
Temperature Electrochemistry, Ceramics and Metals; Lucazeau, G.; Valot, C.; Huntz, A.-M.; Antoni, L. Mater.
Poulsen, F., Bonanos, N., Linderoth, S., Mogenson, M., Res. 2004, 7, 81.
Zachau-Christiansen, B., Eds.; Risø National Laboratory: 118. Gray, S.; Berriche-Bouhanek, K.; Evans, H. E. Mater. Sci.
Ros Rilde, Denmark. 1996; p 187. Forum 2004, 755, 461–464.
99. Quadakkers, W. J.; Thiele, M.; Ennis, P. J.; 119. Froitzheim, J.; Meier, G. H.; Niewolak, L.; Ennis, P. J.;
Teichmann, H.; Schwarz, W. In Proceedings of the Hattendorf, H.; Singheiser, L.; Quadakkers, W. J.
EUROCORR 97, Trondheim, Norway, 22–25 September J. Power Sources 2008, 178, 163–173.
1997, European Federation of Corrosion, 1997; Vol. II, 120. Quadakkers, W. J.; Niewolak, L.; Ennis, P. J. Patent
p 35. number(s): WO2007093148-A1; DE102006007598-A1.
100. Huang, K.; Hou, P.; Goodenough, J. Solid State Ionics Application date, 18 February 2006 (DE), PCT/DE2007/
2000, 129, 237. 000166.
101. Pirón-Abellán, J.; Tietz, F.; Shemet, V.; Gil, A.; 121. Niewolak, L.; Huczkowski, P.; Froitzheim, J.; Ennis, P. J.;
Ladwein, T.; Singheiser, L.; Quadakkers, W. J. In Singheiser, L.; Quadakkers, W. J. In Proceedings of the
Proceedings of the 5th European Solid Oxide Fuel Cell 7th European SOFC Forum, 3–7 July 2006, Lucerne,
Forum, Huijsmans, J., Ed.; European SOFC Forum, CH, CD-ROM, Paper B084.
Dr. Ulf Bossel, Morgenacher Str 2F, CH-5452: 122. Niewolak, L.; Quadakkers, W. J. Unpublished results.
Oberrohrdorf, Switzerland, 2002; Vol. 1, p 248. 123. Yang, Z.; Walker, M. S.; Singh, P.; Stevenson, J. W.
102. Hammer, J. E.; Laney, S. J.; Jackson, R. W.; Coyne, K.; Electrochem. Solid State Lett. 2003, 6, B35–B37.
Pettit, F. S.; Meier, G. H. Oxid. Met. 2007, 67(1/2), 1–38. 124. Nakagawa, K.; Matsunaga, Y.; Yanagisawa, T. Mater.
103. Ghosh, D.; Tang, E.; Perry, M.; Prediger, D.; Pastula, M.; High Temp. 2003, 20, 67–73.
Boersma, R. In Solid Oxide Fuel Cells (SOFC VII); 125. Yang, Z.; Xia, G.-G.; Walker, M. P.; Wang, C.-M.;
Yokokawa, H., Singhal, S. C., Eds.; The Electrochemical Stevenson, J. W.; Singh, P. Int. J. Hydrogen Energy 2007,
Society Proceedings Series: Pennington, NJ, 2001; 32, 3770.
p 101, PV 2001-16. 126. Yang, Z.; Walker, M. S.; Singh, P.; Stevenson, J. W.;
104. Kung, S.; Cal, T.; Moris, T.; Barringer, E.; Elangovan, S.; Norby, T. J. Electrochem. Soc. 2004, 151, B669–B678.
Hartvigsen, J. In Proceedings of the 2000 Fuel Cell 127. Ziomek-Moroz, M.; Cramer, S. D.; Holcomb, G. R.;
Seminar, Portland OR, 30 October–2 November 2000; Covino, B. S., Jr.; Bullard, S. J.; Singh, P. In Corrosion
2000; p 585. 2005; NACE International: Houston, TX, 2005;
105. Krumpelt, M.; Ralph, J.; Cruse, T.; Bae, J.-M. In Paper no. 10.
Proceedings of the 5th European Solid Oxide Fuel Cell 128. Yang, Z.; Xia, G.-G.; Singh, P.; Stevenson, J. W. J. Power
Forum, Huijsmans, J., Ed.; European SOFC Forum, Sources 2006, 160, 1104–1110.
Dr. Ulf Bossel, Morgenacher Str 2F, CH-5452: 129. Jackson, R. W.; Pettit, F. S.; Meier, G. H. Journal of
Oberrohrdorf, Switzerland, 2002; Vol. 1, p 215. Power Sources 2008, 185(2), 1030–1039.
106. Uehara, T.; Ohno, T.; Toji, A. In Proceedings of the 5th 130. Nakagawa, K.; Matsunaga, Y.; Yanagisawa, T. Mater.
European Solid Oxide Fuel Cell Forum, Huijsmans, J., High Temp. 2001, 18, 51–56.
Ed.; European SOFC Forum, Dr. Ulf Bossel, 131. Hecker, R.; Stover, D.; Jonas, H.; Buchkremer, H. P.
Morgenacher Str 2F, CH-5452: Oberrohrdorf, J. Nucl. Mater. 1990, 171(1), 84–93.
Switzerland, 2002; Vol. 1, p 281. 132. Wood, G. C.; Wright, I. G.; Hodgkiess, T.; Whittle, D. P.
107. Honegger, K.; Plas, A.; Diethelm, R.; Glatz, W. In Solid Werkst. Korros. 1970, 21, 900–910.
Oxide Fuel Cells (SOFC VII); Yokokawa, H., Singhal, 133. Ehlers, J.; Young, D. J.; Smaardijk, E. J.; Tyagi, A. K.;
S. C., Eds.; The Electrochemical Society Proceedings Penkalla, H. J.; Singheiser, L.; Quadakkers, W. J. Corros.
Series: Pennington, NJ, 2001; p 803, PV 2001-16. Sci. 2006, 48, 3428–3454.
108. Ehlers, J.; Quadakkers, W. J. Oxidation von 9–12% 134. Tietz, F.; Simwonis, D.; Batfalsky, P.; Diekmann, U.;
Cr-Stählen in wasserdampfhaltigen Atmosphären bei Stöver, D. In Degradation Phenomena during Operation
of Solid Oxide Fuel Cells Proceedings of the 12th IEA 152. Brylewski, T.; Nanko, M.; Maruyama, T.; Przybylski, K.
Workshop on SOFCs: Materials and Mechanisms; Solid State Ionics 2001, 143, 131.
Nisancioglu, K., Ed.; International Energy Agency, 153. Quadakkers, W. J. German Patent DE 19 54 66 14, 1995;
1999; p 3. Quadakkers, W. J.; Hänsel, M. German Patent DE 19 60
135. Teller, O.; Meulenberg, W.; Tietz, F.; Wessel, E.; 98 13, 1996.
Quadakkers, W. J. In Solid Oxide Fuel Cells (SOFC VII); 154. Montero, X.; Jordan, N.; Piron-Abellan, J.; Tietz, F.;
Yokokawa, H., Singhal, S. C., Eds.; The Electrochemical Stover, D.; Cassir, M.; Villarreal, I. J. Electrochem. Soc.
Society Proceedings Series: Pennington, NJ, 2001; 2009, 156(1), B188–B196.
p 895, PV 2001-16. 155. Yang, Z. G.; Xia, G. G.; Nie, Z. M.; et al. Electrochem.
136. Batfalsky, P.; Buchkremer, H.-P.; Froning, D.; Solid State Lett. 2008, 11(8), B140–B143.
Meschke, F.; Nabielek, H.; Steinbrech, R. W.; Tietz, F. In 156. Stanislowski, M.; Froitzheim, J.; Niewolak, L.;
Proceedings of the 3rd International Fuel Cell Conference; Quadakkers, W. J.; Hilpert, K.; Markus, T.; Singheiser, L.
New Energy and Industrial Technology Development J. Power Sources 2007, 164, 578–589.
Organization (NEDO) and Fuel Cell Development 157. Batawi, E.; Honegger, K.; Diethelm, R. In Proceedings of
Information Center (FCDIC), Japan, 1999; p 349. the 6th IEA Workshop on Advanced SOFC, published by
137. Urbanek, J.; Miller, M.; Schmidt, H.; Hilpert, K. In ENEA, Energy Department, 22–29 February 1994; Rome,
Proceedings of the 2nd European Solid Oxide Fuel Cell Italy, p 175.
Forum, Thorstensen, B., Ed.; European SOFC Forum, Dr. 158. Trebbels, R. Ph.D. Thesis, RWTH Aachen, 2009.
Ulf Bossel, Morgenacher Str. 2F, CH-5452: 159. Shemet, V.; Quadakkers, W. J. Unpublished results.
Oberrohrdorf, Switzerland, 1996; Vol. 2, p 503. 160. Ertl, S. T. Ph.D. Thesis, RWTH Aachen, 2006.
138. Quadakkers, W. J.; Mallener, W.; Grübmeier, H.; 161. Froitzheim, J.; Niewolak, L.; Brandner, M.; Singheiser, L.;
Wallura, E. In Proceedings of the 5th IEA Workshop on Quadakkers, W. J. J. Fuel Cell Sci. Technol. in press 6(4).
SOFC; Materials, Process Engineering and 162. Brandner, M.; Bram, M.; Froitzheim, J.;
Electrochemistry, Jülich Germany, 2–4 March 1993, Buchkremer, H. P.; Stover, D. Solid State Ionics 2008,
Forschungszentrum Jülich, Jülich, FRG 1993; p 87; 179(27–32), 1501–1504.
ISBN 3-89336-127-8. 163. Brandner, M.; Niewolak, L.; Froitzheim, J.; Quadakkers,
139. Badwal, S.; Deller, R.; Foger, K.; Ramprakash, Y.; W. J.; Bram, M. Patent number(s): DE102006056251-A1;
Zhang, J. Solid State Ionics 1997, 99, 297. WO2008064938-A1.
140. Larring, Y.; Norby, T. J. Electrochem. Soc. 2000, 147, 164. Menzler, N. H.; Sebold, D.; Zahid, M.; Gross, S. M.;
3251. Koppitz, T. J. Power Sources 2005, 152(1), 156–167.
141. Quadakkers, W. J.; Greiner, H.; Hänsel, M.; Pattaniak, A.; 165. Weil, K. S. JOM 2006, 58, 37–44.
Khanna, A. S.; Mallener, W. Solid State Ionics 1996, 91, 166. Eichler, K.; Solow, G.; Otschik, P.; Schafferath, W. J. Eur.
55. Ceram. Soc. 1999, 19, 101–104.
142. Tietz, F.; Mohsine, Z.; Quadakkers, W. J.; Shemet, V. 167. Sohn, S.-B.; Choi, S.-Y.; Kim, G.-H.; Song, H.-S.;
U.S. Patent, US 7,407,717, B2, PCT/DE2004/000024, Kim, G.-D. J. Amer. Ceram. Soc. 2004, 87, 254–260.
Application Data 18 February 2003. 168. Yang, Z.; Meinhardt, K. D.; Stevenson, J. W. J.
143. Kadowaki, T.; Shiomitsu, T.; Matsuda, E.; Nakagawa, H.; Electrochem. Soc. 2003, 150, A1095–A1101.
Tsuneizumi, H.; Maruyanama, T. Solid State Ionics 169. Yang, Z.; Stevenson, J. W.; Meinhardt, K. D. Solid State
1993, 67, 65. Ionics 2003, 160, 213–225.
144. Maruyama, T.; Ioue, T.; Nagata, K. In Solid Oxide Fuel 170. Yang, Z.; Xia, G.-G.; Meihardt, K. D.; Weil, K. S.;
Cells (SOFC IV); Dokiya, M., Yamamoto, O., Tagawa, H., Stevenson, J. W. J. Mater. Eng. Perform. 2004, 13,
Singhal, S. C., Eds.; The Electrochemical Society 327–334.
Proceedings Series: Pennington, NJ, 1995; p 889, PV 95-1. 171. Yang, Z.; Coyle, C. A.; Baskaran, S.; Chick, L. A. US
145. Oishi, N.; Namikawa, T.; Yamazaki, Y. Surf. Coatings Patent, 6,843,406, 2005.
Technol. 2000, 132, 58. 172. Haanappel, V. A. C.; Shemet, V.; Vinke, I. C.;
146. Yoo, Y.; Dauga, M. In Solid Oxide Fuel Cells (SOFC VII); Quadakkers, W. J. J. Power Sources 2005, 141,
Yokokawa, H., Singhal, S. C., Eds.; The Electrochemical 102–107.
Society Proceedings Series: Pennington, NJ, 2001; 173. Haanappel, V. A. C.; Shemet, V.; Vinke, I. C.;
p 837, PV 2001-16. Gross, S. M.; Koppitz, Th.; Menzler, N. H.; Zahid, M.;
147. Godfrey, B.; Föger, K.; Gillespie, R.; Bolden, R.; Quadakkers, W. J. J. Mater. Sci. 2005, 40, 1583–1592.
Badwal, S. J. Power Sources 2000, 86, 68. 174. Haanappel, V. A. C.; Shemet, V.; Gross, S. M.;
148. Hou, P.; Huang, K.; Bakker, W. In Solid Oxide Fuel Cells Koppitz, Th.; Menzler, N. H.; Zahid, M.;
(SOFC VI); Singhal, S. C., Dokiya, M., Eds.; The Quadakkers, W. J. J. Power Sources 2005, 150,
Electrochemical Society Proceedings Series: 86–100.
Pennington, NJ, 1999; p 737, PV 99-19. 175. Menzler, N. H.; Batfalsky, P.; Blum, L.; Bram, M.;
149. Tietz, F.; Sebold, D. Mater. Sci. Eng. B 2008, 150(2), Gross, S. M.; Haanappel, V. A. C.; Malzbender, J.;
135–140. Shemet, V.; Steinbrech, R. W.; Vinke, I. Fuel Cells 2007,
150. Ager, F. J.; Respaldiza, M.; Botella, J.; Soares, J.; Da 7(5), 356–363.
Silva, M.; Benitez, J.; Adriozola, J. Acta Mater. 1996, 44, 176. Opila, E. J.; Jacobson, N. S.; Myers, D. L.; Copland, E. H.
675. JOM 2006, 58, 22–28.
151. Ruiz, M.; Heredia, A.; Botella, J.; Adriozola, J. J. Mater. 177. Niewolak, L.; Shemet, V. Forschungszentrum Jülich,
Sci. 1995, 30(20), 5146. 2008, Unpublished results.
1.20 Gas Turbine Oxidation and Corrosion
J. R. Nicholls and N. J. Simms
Cranfield University, Cranfield, Bedfordshire MK43 0AL, UK

1.20.2 Gas Turbine Operating Conditions 520
1.20.3 Gas Turbine Hot Gas Path Components and Materials 522
1.20.4 Oxidation and Thermal Cycling 524
1.20.4.1 Oxide Formation and Spallation 524
1.20.4.1.1 Oxidation processes 524
1.20.4.1.2 Modeling oxide growth and spallation 526
1.20.4.2 Thermal Barrier Coating Systems 527
1.20.4.2.1 Damage mechanisms 527
1.20.4.2.2 Modeling TBC spallation 529
1.20.5 Hot Corrosion 529
1.20.5.1 Hot Corrosion Mechanisms 530
1.20.5.2 Deposit Formation and Fuel Specifications 532
1.20.5.3 Damage Locations 535
1.20.5.4 Rates of Hot Corrosion Attack and Modeling Hot Corrosion Processes 535
1.20.6 Current/Future Issues 538
1.20.7 Summary 538
References 539
Abbreviations Symbols
APS Air plasma spray y Reciprocal value of the C to H ratio in mol
CVD Chemical vapor deposition A/F Air to fuel mass flow ratio
EBPVD Electron beam physical vapor MO Metal oxide
deposition S/F Steam (water) to fuel mass flow ratio
EFC European Federation of Corrosion XA Contaminant concentration in inlet air (ppm) by
EPSRC Engineering and Physical Sciences weight
Research Council (United Kingdom) XF Contaminant concentration in fuel (ppm) by weight
EU European Union XS Contaminant concentration in injected steam or
FOD Foreign object damage water (ppm) by weight
HP High pressure XT Total contamination in fuel equivalent (ppm) by
HRSG Heat recovery steam generator weight
HVOF High velocity oxy-fired l Amount of excess air: actual mass flow divided by
IGCC Integrated gasification combined cycle the air mass flow for stoichiometric
ISO International Organization for combustion
Standardisation
LPPS Low pressure plasma spray
NGV Nozzle guide vane
1.20.1 Introduction
NMAB National Materials Advisory
Board (USA)
Gas turbines have been developed rapidly over the
NRC National Research Council (USA)
last 60 years, with applications currently ranging
PLPS Photoluminescence piezospectroscopy
from aeroengines to marine propulsion to industrial
TBC Thermal barrier coating
mechanical power to utility scale electrical power
TGO Thermally grown oxide
generation.1,2 Whatever be the intended application
518
Combustor
Air inlet Compressor Exhaust

Power
turbine
Figure 1 Schematic diagram of simple cycle gas turbine showing fundamental components.
of the gas turbine, the basic features and principles of Thus, within these constraints, there are several
operation are common. Figure 1 illustrates the fun- routes to improve the efficiency of gas turbines:
damental parts of a gas turbine. increasing the power turbine inlet temperature; this
Details of the operation of gas turbines (which use
is the traditional route with limitations being set by
the Brayton cycle) are available from numerous
the materials of construction for the hot gas path,
sources1,2 but can be summarized as follows for the
but now also being set by cooling technologies and
purposes of this chapter:
thermal barrier coating (TBC) performance;
air enters the gas turbine through the compressor; lower the exit temperature;
the air is compressed (and heated adiabatically) as using the heat in the exit gases in another way; for
it passes through a series of vane/rotor stages of example, by passing the gases through a heat
the compressor; recovery steam generator (HRSG) to generate
this air is passed into combustion chamber(s) steam which is then passed through steam turbine
where fuel is introduced and burnt raising the to generate power (i.e., using a conventional Ran-
temperature of the gas stream; kine cycle); the overall system is known as a com-
this hot, high pressure gas stream is then passed bined cycle, and overall efficiencies for such
through the power turbine where its temperature systems have been reported to exceed 60%.
and pressure are reduced as it passes through a series In addition, there are several other variations of
of vane/rotor stages which drive both the compressor gas turbine cycles that continue to be investigated:
and mechanical/electrical power conversion unit.
Indirect firing; where the compressed air is heated
The Carnot efficiency of the gas turbine (as with by an external heat source via heat exchangers
any heat engine) is related to the difference between before passing through the power turbine;
the maximum and minimum absolute temperatures Oxy-firing; where the fuel is burnt with oxygen to
of the working fluid (in which case, the inlet and generate a combusted gas stream of mostly CO2 and
outlet temperatures of the power turbine are divided steam, from which the steam can be condensed to
by the inlet temperature). For a current gas turbine produce a CO2 rich gas stream for recycling into the
with an inlet temperature of 1400 C (1673 K) and an compressor or for storage (after further cleanup).
exit temperature of 600 C (873 K), this would give a In terms of the sizes of gas turbines:
Carnot efficiency of 48%. Real cycle efficiencies
the largest utility power turbines are rated at
are always lower than Carnot efficiencies, and in this
case, for a simple cycle gas turbine, the efficiency 340 MWe in simple cycle mode and 530 MWe
would be closer to 38–40%.3 If the inlet tempera- in combined cycle mode4;
jet engines are rated in terms of their thrust, with
ture is to be raised to 2100 C or 2373 K (just
below stoichiometric combustion for maximum the largest generating up to 100 000 lbs.1
flame temperature1 – a near impossible achievement) From the foregoing, it is evident that materials in
and the exit temperature remains at 600 C (873 K), gas turbines need to be capable of operating at high
then the maximum thermal efficiency will be 63%, bulk temperatures with high static and fluctuating
without dropping the exit temperature to extract stresses, while by also withstanding the surrounding
energy from this lower grade heat source. environments. The environments generated within
the hot gas paths of gas turbines can be both physi- Excess air is always used in the process to ensure
cally and chemically aggressive, with particles pro- that the fuel is completely consumed, to reduce the
ducing erosion or deposition while gaseous and temperature of the combusted gas stream, and to
vapor-phase species may produce different forms of increase the mass flow through the power turbine.
deposition, as well as oxidation and hot corrosion. The This produces a high temperature combusted gas
balance between these various degradation mecha- stream that is oxidizing. For some industrial gas tur-
nisms is component and turbine dependent (as are bines, water/steam is also added to increase the mass
the temperatures and stresses involved). flow through the power turbine and/or reduce NOX
Thus, from a generic materials point of view, emissions5–9 (in some cases, steam-cooled compo-
issues of concern for gas turbine hot gas path components also exhaust into the main combusted gas flow).
nents include: However, unwanted products of combustion pro-
cesses can be produced:
creep,
fatigue (thermal, low cycle, and high cycle), CO from incomplete combustion of the fuel;
thermal cycling, NOX from thermal deposition of air or fuel-
oxidation, derived nitrogen species;
hot corrosion, SOX from oxidation of fuel borne impurities.
erosion/foreign object damage (FOD),
Various approaches have been developed to con-
synergistic oxidation/corrosion/mechanical prop-
trol these unwanted combustion products, such as
erty interactions.
improved combustor design, staged combustion,
This chapter focuses on the oxidation and hot water injection, etc. Not all are appropriate (or feasi-
corrosion performance of materials in a range of ble) for every gas turbine application, so their use
different gas turbines and on how thermal cycl- varies. However, increasingly stringent environmental
ing, erosion, and mechanical loading affect this regulations are being developed and applied in various
performance. countries; especially to utility scale gas turbines.
Unfortunately, the combusted fuel gas stream also
includes species derived from minor and trace ele-
1.20.2 Gas Turbine Operating ments in the fuels and oxidant streams: these elements
Conditions commonly include sulfur, alkali metals, chlorine,
magnesium, calcium, lead, vanadium, etc. However,
Gas turbines have conventionally been fired using when gasification processes are considered, even
hydrocarbon fuels (such as diesels, jet fuels, and more elements can be involved, including cadmium
natural gases) burnt using air as the oxidant. More and arsenic.10 These various elements can combine to
recent fuels have included syngases produced by coal cause hot corrosion damage to the hot gas path of the
or biomass gasification processes. In addition, use of gas turbine. The reactions involved, their effects, and
oxygen as the oxidant is now being actively investi- methods to limit them (including fuel specifications)
gated for utility gas turbines as one potential route to are described in detail in Section 1.20.5.
enable CO2 capture technologies to be introduced. During the last 60 years, there has been a continu-
In all cases, the basic reaction of the fuels with the ous drive to increase the efficiencies of all types of gas
oxidant can be summarized as: turbine, but they still fall short of the Carnot theoreti-
cal optimum, which would be achieved at stoichiomet-
Fuel þ air ! Combusted gases
ric combustion combined with no mechanical losses.
or As a result, the firing temperatures and pressures
of gas turbines have increased steadily (Figure 2).
C þ yH þ lð1 þ y=4ÞðO2 þ 3:719N2 þ 0:043ArÞ
Current industrial gas turbines for power generation
! CO2 þ ðy=2ÞH2 O þ lð1 þ y=4Þð3:719N2 applications have firing temperatures of up to
1430 C and pressures of up to 34 bar, while for
þ 0:043ArÞ þ ðl 1Þð1 þ y=4ÞO2
jet engine applications, firing temperatures are up to
where y is the reciprocal value of the C to H ratio in 1550 C and pressures up to 45 bar.1
mol, l is the amount of excess air actual mass flow Figure 3 illustrates the temperature and pres-
divided by the air mass flow for stoichiometric sure variations expected through a generic gas tur-
combustion. bine. There are wide variations in the actual
1300 24
SGT800
1200 SGT400 21
SGT700
Firing temperature (⬚C)
1100 18
Pressure ratio
SGT100
1000 15
SGT200
900 12
TB5000
800 9
700 6
1975 1980 1985 1990 1995 2000 2005
Year
Figure 2 Increase in turbine entry temperatures and pressure ratios over time for Siemens industrial gas turbines
(4–50 MW). Reproduced from Hannis, J. M. In Power Generation in an Era of Climate Change; Proceedings of the 7th
International Charles Parsons Turbine Conference; Strang, A. et al. Eds.; IOM Communications, 2007; pp 37–49.
20 1400
Maximum gas pressure
18 Minimum gas pressure 1300
Tempest-gas pressure in/out
16 Cyclone-gas pressure in/out 1200
Metal temperature (°C)

TB5000-gas pressure in/out
14 1100
Gas pressure (bar)
Typhoon-gas pressure in/out

12 Other GTs-gas pressure in/out 1000
Maximum metal temperature
10 Minimum metal temperature 900
8 800
6 700
4 600
2 500
0 400
400 500 600 700 800 900 1000 1100 1200 1300 1400
Gas temperature (°C)
Figure 3 Variation in temperature and pressure of the gas stream passing through typical industrial gas turbines.
Reproduced from Simms, N. J.; Nicholls, J. R.; Oakey, J. E. In Lifetime Modelling of High Temperature Corrosion Processes;
Schütze, M. et al. Eds.; Maney, 2001; pp 379–397, EFC No. 34.
temperatures and pressures that are produced by (with flight times of up to 12 h), with peak operating
different gas turbines. This results from a number of temperatures for short periods during takeoff and
factors such as: landing. In contrast, industrial gas turbines designed
for utility scale electrical power generation should be
power requirement of the turbine,
ramped slowly to their optimum operating tempera-
design approach,
ture and then ideally remain there for hundreds if not
duty and cycle required,
thousands of hours. Other differences include weight
materials/lifting restrictions.
limitations and the various contaminants that can
The gas turbines used for the different applications enter the gas turbines during their operation. Thus
outlined in the introduction, while obviously related the gas turbine usage defines the maximum gas tem-
by their basic principles of operation, have different perature and pressure, duty cycle, operating time at
detailed requirements. For example (Figure 4), civil temperature, number of shut-down cycles, and mate-
aeroengines go through a routine cycle on every flight rial operating environment.
1300
1200
1100
Temperature (°C) 1000
900
800
700
600
500
400
0 20 40 60 80 100
(a) Time (min)
1300
1200
1100
1000
Temperature (°C)
900
800
700
600
500
400
0 10 20 30 40 50 60 70 80 90 100
(b) Time (h)
Figure 4 Examples of gas temperature cycles for an aeroengine flight and an industrial gas turbine.12 (a) Example of short
duration aeroengine cycle (Note: 4000–7500 cycles between overhauls, depending on cycle duration). (b) Example of
industrial gas turbine cycle (Note: desired component lives of >24 000 h).
1.20.3 Gas Turbine Hot Gas Path improved mechanical properties and thermal stabili-
Components and Materials ties. As these compositions have been progressively
optimized, new materials have been developed with
As efficiencies of all types of gas turbines increase improved high temperature mechanical properties
with firing temperature and pressure, there has (Figure 5). Until the 1970s, such improvements in
always been a drive for the hot gas paths to use high temperature mechanical properties could be
materials close to their mechanical and environmen- coupled to improved oxidation and hot corrosion
tal operating limits.1,2 However, viable component resistance,2,14 with alloys being designed to have suf-
lives for the hot gas path of the gas turbines have ficient chromium and aluminum contents to form
only been achieved as a result of significant materials self-protective oxide scales. However, more recent
developments coupled with improved manufacturing alloy developments, targeted at improved mechanical
methods and engine designs (specifically efficient properties, could only be achieved to the detriment
cooling). Originally steel parts were used, but as the of oxidation and hot corrosion resistance of these
severity of operating conditions was increased, it alloys. Thus there is now a requirement for many
was necessary to invent and develop whole series hot gas path components to be coated to enable
of nickel-based and cobalt-based superalloys with them to resist the surrounding environments and
Uncooled Cooled Coated

2100 Demonstrator
blades blades blades
technology
Turbine entry temperature (K)

1900
Trent
Production
1700 technology
1500
Conway RB211
1300 Spey SX cast
W1 Avon
Dart DS cast
1100
Cast alloys
Wrought alloys
900
1940 1960 1980 2000 2020
Year of introduction
Figure 5 Improvements in materials strength with time and new alloy development. Reproduced from Rickerby, D. S.;
Hicks, M. A. In Power Generation in an Era of Climate Change, Proceedings of the 7th International Charles Parsons Turbine
Conference; Strang, A. et al. Eds.; IOM Communications, 2007; pp 229–240.
to permit economically viable component lives. Higher TBCs means that first stage blades/vanes can now be
temperature alloy mechanical performance, as well as operated at gas temperatures in excess of the melting
associated engine capability, has also been enhanced points of the alloys used for these components.3,4
through improvements in casting technologies. These Critical components within the hot gas path of gas
have permitted a change in the microstructure of turbines include:
nickel-based superalloys from equiaxed to direction-
ally solidified and single-crystal blade materials combustor can/tiles/transitions ducts,
(eliminating the grain boundaries that partly dictate vanes and blades (both in multiple stages),
high temperature creep damage). Originally these seals,
casting technologies were developed for small aero- discs.
engine blades, but now larger industrial gas turbine Examples of alloys used for these components are
blades/vanes can also be processed.3 given in Table 1.
The development of a range of increasingly sophis- A wide range of coatings have been developed and
ticated component cooling technologies (including air used for protection along the gas turbines hot gas path.
impingement, film cooling, etc. Figure 6) has enabled The different routes for coating application that have
gas stream temperatures to be decoupled from metal developed are favored by different gas turbine manu-
temperatures along the gas turbine hot gas path. Most facturers. Thus, the situation of using a ‘standard’
of the cooling systems rely on air diverted from coating for protection against hot corrosion damage
selected locations in the air compressor, but there are has not developed. Instead, there are several coatings
some land-based gas turbines that use steam cooling. from the different classes of coating materials available
The use of such cooling technologies has enabled gas that are currently being used (Table 2):
temperatures to be progressively increased, much fur-
ther than would be expected from alloy developments. Overlay coatings:
The introduction of cooled components has typically Ni–Co–15–23% Cr–8–12% Al–0.1–
allowed the adoption and development of TBCs as 0.9% Y,
part of the turbine thermal management system. applied by high velocity oxy-fuel (HVOF)
These low conductivity ceramic coatings protect the or low pressure plasma spray (LPPS) coating
underlying metallic materials from the surrounding processes,
high temperature gas streams and allow a balance some newer coatings also include Re, Hf, Ta, or
between component life, component operating tem- Si in some combination.
perature, and cooling air requirements to be addressed.
The principal of TBCs is illustrated in Figure 7. The Diffusion coatings:
use of advanced air cooling technologies coupled with Aluminizing, chromizing, or chrome-aluminizing;
Front chamber double end Air passes

feed to leading edge holes through holes
in impingement
plate cooling
the aerofoil
Baffle plate
Air exits leading

edge holes to Air exits through
cool NGV trailing edge
pedestal bank
Film cooling
HP NGV cooling
flows, showing
Front chamber Single end feed impingement and
double end feed baffle plates
Figure 6 Example of air cooling systems in advanced gas turbine aerofoil.1 Courtesy of Rolls Royce.
Thermal Platinum-aluminizing;
Combustion barrier Bond
gases coating coat Substrate Cooling air Electroplate of platinum followed by CVD
aluminizing process.
Slurry coatings;
Tgas
Silicon-aluminizing.
Coating performance varies with base alloy
Temperature
compositions, which have a notable effect in

terms of hot corrosion resistance, with such
coatings on lower Cr-based alloys performing
significantly worse than on higher Cr substrates.
Tcoolant
Distance
1.20.4 Oxidation and Thermal Cycling
Figure 7 Principal of operation of thermal barrier coatings
(TBCs). 1.20.4.1 Oxide Formation and Spallation
1.20.4.1.1 Oxidation processes
Chemical vapor deposition (CVD) processes Gas turbine materials will oxidize in the combustion
applied by in-pack, above pack or gaseous gases produced in all gas turbine systems. The earlier
routes; superalloys were relatively rich in chromium and
Gaseous routes enable coating of internal so formed slow growing protective Cr2O3 scales at
cooling passages in advanced aero-foil designs; their temperatures of use. However, as component
Table 1 Examples of materials used in gas turbine hot gas path components (in wt%)
Ni Co Fe Cr Al Ti Mo W Others
CMSX-4 Bal 9.0 6.5 5 1 0.6 0.1 Hf

IN738LC Bal 8.5 16 3.4 3.4 1.7 2.6 1.7 Ta, 0.9 Nb
GTD111 Bal 9.5 14 3 4.9 1.5 3.8 2.8 Ta
IN939 Bal 19 22.5 1.9 3.7 2.0 1.4 Ta, 1.0 Nb
IN6203 Bal 19 22 2.3 3.5 2.0 1.1 Ta, 0.8 Nb, 0.8 Hf
GTD222 Bal 14 22.5 1.2 2.3 2.0 1.0 Ta, 0.8 Nb
IN792 Bal 9.0 12.4 3.1 4.5 1.9 3.8 3.9 Ta
FSX-414 10 Bal 1 29 7.5
MarM 247 Bal 8.3 5.5 1 0.7 10 1.5 Hf
Hastelloy X Bal
Haynes 230 Bal <3 22 0.3 2 14
RR1000 Bal 18.5 15 3 3.6 5 2 Ta, 0.5 Hf
IN718 Bal 18.5 19 0.5 0.9 3 5.1 Nb
Astroloy Bal 15 15 4.4 3.5 5.2
Waspalloy Bal 13.5 <2 19.5 1.4 3 4.3
U500 Bal 19 4 19 3 3 4
U720Li Bal 15 16 2.5 5 3 1.2
Haynes 214 Bal 3 16 4.5 0.1 Zr, 0.01 Y
PM2000 Bal 20 5.5 0.5 0.5 Y as Y2O3
Aluchrom YHf 0.2 Bal 20 5.8 0.5 Y, 0.05 Hf, 0.05 Zr
Table 2 Examples of gas turbine coating compositions The development of superalloys for high metal tem-
(wt%) peratures resulted in materials with progressively less
Overlays Ni Co Cr Al Y Other
chromium and more aluminum. This resulted in
alloys that formed slower growing Al2O3 scales at
Amdry 963 Bal – 25 6 0.4 high temperatures that were not volatile in the sur-
Amdry 995 32 Bal 21 8 0.5 rounding gas streams.2
Amdry 997 Bal 23 20 8.5 0.6 4 Ta
LCO22 32 Bal 21 8 0.5
At temperatures higher than 950–1000 C,
CT102 32 Bal 21 8 0.6 a-Al2O3 scales are the preferred scale to be formed.
CT103 Bal 22 17 12 0.6 Considerable research targeted at a range of base
CT104 Bal 23 22 11 0.4 4 Hf, 0.4 Si alloys and different applications has been carried
GT29 – Bal. 29.0 6.0 0.3 out on the growth of these scales over the year.2,16
SV20 Bal – 25 5.5 0.7 1 Ta, 2.7 Si
SV30 Bal. – 25.0 5.0 0.7 0.1 Ta,
However, at lower metal temperatures, more com-
0.35 Ti plex oxides may form on the surfaces of superalloys:
SC2231 30 Bal 28 8 0.5 0.5 Si various transition aluminas (e.g., y- or g-Al2O3) as
SC2453 Bal 10 28 12 0.6 3 Re well as mixed (Cr, Al)2O3 oxides, spinel oxides of Cr
and/or Al with base metal oxides, and oxides of other
Diffusion Description
minor alloying elements that are present in these
Aluminized Applied in-pack, above pack or by gas complex superalloy materials may form.
phase CVD In general, the combination of metal temperature
Chromized Applied in-pack, above pack or by gas
and a ‘clean’ oxidizing environment (i.e., a combus-
phase CVD
RT22 Variations on platinum electro-plate followed tion gas stream with trace metal contaminants at too
CN91 by aluminizing low a level to cause any hot corrosion) will result in
MDC150 the formation of slow-growing stable oxide scales for
SermaLoy Slurry application the appropriate blade and vane materials, plus any
1515
coating system, in the gas turbine hot gas path (but it
should be noted that such materials will vary between
temperatures rose, so did the rate of Cr2O3 formation, gas turbine manufacturers, and with the intended use
and other oxidation mechanisms, such as formation of of gas turbine).
volatile oxide and oxyhydroxide species in the fast There will, therefore, be a difference between
flowing combustion gases, also became signifcant.2,15 gas turbines: the higher peak temperatures of
aeroengines will encourage a-Al2O3 scale formation, At high temperatures, coatings will degrade by
compared to industrial machines operating close to interdiffusion with the substrate as well as surface
oxide transition temperature. Another important dif- oxidation; this can be particularly important for alu-
ference in the exposure conditions will be number minum which tends to diffuse inwards as well as
and severity of thermal cycles. Aeroengine tempera- being consumed due to oxidation. The development
tures are cycled much more frequently and rapidly of diffusion barriers to reduce or prevent such
than industrial gas turbines. As a result, the slow- inwards diffusion is an active research topic.
growing scales may spall from the component sur-
faces and a new oxide will form during the next 1.20.4.1.2 Modeling oxide growth and
period of operation: whether this is the same oxide spallation
or not will depend on the composition of the under- Oxide growth and spallation models have been devel-
lying alloy/coating. oped for alumina type oxide scales by several research
The coating of superalloys with most overlay and groups around the world, most notably: NASA,18
diffusion coatings (Table 2) results in a surface rela- Jülich–Cranfield19,20 and Toulouse.21 These models
tively rich in aluminum, which assists in the forma- are essentially all empirical, but are based on the math-
tion, retention, and reformation of a slow-growing ematical understanding of oxidation processes:
Al2O3-based (or at least alumina rich) oxide scale.
The NASA cyclic oxidation and spallation (COSP)
Overlay coatings traditionally include ‘reactive ele-
model is based on rate constants determined from
ments’ such as Y or Hf that are well known to enhance
mass change data and uses these combined with
the adhesion of alumina scales (see Chapter 1.23,
idealized parameters that describe the spallation
High Temperature Corrosion of Chromia-forming
process to predict the damage to a metal in terms
Iron, Nickel and Cobalt-base Alloys). Scales can
of mass change or metal loss for different time
develop through transient oxides (i.e., various
periods, cycle durations, etc.
Al2O3) and other oxides (e.g., NiO, spinels) and mul-
The Toulouse p–kp model uses a spalling proabil-
tiple oxide layer developments (e.g., undergrowths of
ity (p) and parabolic rate constant (kp) to predict
Cr2O3 and a-Al2O3) before reaching a slow growing
the mass change of materials as a function of time
stable structure (e.g., Figure 8). This development
and temperature.
depends on the exposure temperature and coating–
The Jülich–Cranfield model uses the concept of an
base alloy composition.
aluminum reservoir to calculate the metal loss due
Diffusion coatings have also progressively devel-
to alumina growth and spallation. Spallation pro-
oped to assist in the growth and retention of alumina
cesses are assumed to commence after some criti-
scales, in this case, using metals such as Pt and Pd.
cal mass gain (or oxide thickness) which is
These coatings will also develop scales through the
influenced by thermal cycling.
formation of transient oxides (i.e., various Al2O3) and
other oxides (e.g., NiO) and multiple oxide layer Continuous oxide growth and coating interdif-
developments (i.e., undergrowth of a-Al2O3) before fusion models are being developed by several
reaching a slow growing stable structure. As for over- research groups, most notably NASA,22 DTU,23 and
lay coatings, this development depends on the expo- Toulouse.24,25 These use various finite difference
sure temperature and coating–base alloy composition. or finite element methods, in some cases combined
α-Al2O3 Cr2O3 NiAl2O4 Nio
2 μm
(a) Back-scattered image (b) Gray scale image
17
Figure 8 Scale grown on Ni–27Cr–8Al at 1353 K. (a) Back-scattered image and (b) grey scale image.
with thermodynamic phase predictions, to calculate as well as to provide a transition zone between the
elemental profiles throughout the oxide–coating– metal and the ceramic layer. Thus bond coat materi-
substrate system as a function of time and tempera- als are often classically established environmental
ture, again using the ideal a-Al2O3 scale formation. protection coatings, for example, a platinum alumi-
nide diffusion coating system or an MCrAlY overlay
coating. However, bond coatings for APS and EB
1.20.4.2 Thermal Barrier Coating Systems
PVD TBCs require different initial characteristics
1.20.4.2.1 Damage mechanisms to enable the coatings to be successfully applied.
Two types of TBCs are now frequently found in gas APS TBCs generally need to be applied to deliber-
turbines. These are generally classified by their appli- ately roughened bond coat surfaces to provide
cation routes: air plasma sprayed (APS) and electron mechanical keying, whereas EBPVD TBCs are usu-
beam physical vapor deposition (EB PVD) (Figure 9). ally applied to smoother, polished, or media finished
However, currently their compositions are generally bond coat surfaces.
the same: 7 wt% Y2O3 – balance ZrO2. New thermal The spallation of TBCs after exposure at high tem-
barrier ceramics are also being researched,27 but a perature is now well recognized as the life-limiting
review of these is beyond the scope of this chapter. factor dictating TBC performance and has been fre-
Both types of TBCs are applied to gas turbine quently reported.28 However, the causes of this spall-
components after they have had a bond coating ation remain controversial (e.g., Figure 10). Features
applied. The purpose of this bond coating is to reported include28:
encourage and maintain the formation of a stable a-
spallation is usually observed on samples/compo-
Al2O3 scale at the interface between metal and ceramic,
nents after the growth of 5–10 mm of oxide scale;
cracks grow roughly parallel to the bond coat/
thermally grown oxide/TBC interfaces, before
large areas of TBC spall;
cracks may grow:
in TBC near TGO interface (between bond-
coat peaks, or not),
within the TGO,
along bond coat/TGO interface, depending on
the system;
bondcoat/TGO/interfaces may roughen with
exposure (for EBPVD TBCs on Pt–Al bond coats);
mixed oxides (Cr,Al)2O3 or spinels (e.g., (Ni,Co)
100 μm
(a) Cr2O4) may form above or below the stable alu-
mina scales in areas close to crack growth.
Both academic and industrial research is actively
continuing in this area, but has moved from charac-
terization towards methods of finding damaged TBCs
nondestructively and also developing predictive
models for TBC spallation (see below). But it is
worth noting that there are many significant differ-
ences between the preparation and processes used for
applying bond coatings, initiating the growth of ther-
mally grown oxide and applying the TBC layer that
are used by different coating companies, all of which
100 μm may interact and therefore influence susceptibility to
(b) TBC spallation.
Particular interest has been taken in finding ways
Figure 9 Two main types of TBCs.26 (a) Air plasma
sprayed thermal barrier coating (APS TBC) and (b) electron
of detecting the cracks in/close to TGO using non-
beam physical vapor deposition thermal barrier coating destructive inspection methods, before the cracks
(EB PVD TBC). have a chance to coalesce and cause TBC spallation
TBC
Transient oxides
and detached areas TBC
Oxide protrusions TGO

and inclusions TGO
NiCoCrAIY bond coat NiCoCrAIY bond coat
(a) (b)
Defects
TBC TBC
Defects
TGO TGO
Crack
Crack
NiCoCrAIY bond coat NiCoCrAIY bond coat
(c) (d)
(A) Possible causes of TBC spallation [70]
Thermochemical
α-Al2O3 TBC TBC

α-Al2O3 Impacting
particle
Spinel
(aAI<10–17) Delamination
BC BC
IAI
(a) AI depletion (c) Undulating interface Tg

Compacted Erosion
area
TBC TBC
TBC
Spinel α-Al2O3 Hot

BC
α-Al2O3 area
IN Crack of BC
BC inclusion (e) Foreign object damage
(b) Outward Ni transport (d) Local oxide thickening
(B) Thermomechanical
Figure 10 Schematic diagrams of alternative views of the causes of TBC spallation. (A) Possible causes of TBC spallation29
and (B) possible causes of TBC spallation. Reproduced from Simms, N. J.; Kilgallon, P. J.; Roach, C.; Oakey, J. E.
In Microscopy of Oxidation, Proceedings of 5th International Conference, Newcomb, S. and Tatlock, G. Eds.; Science
Reviews, 2003; pp 273–280.
to occur. Promising nondestructive techniques cur- photoluminescence piezospectroscopy (PLPS) –

rently include30: this uses lasers to excite the Cr3+ contaminant ions
that are always present in Al2O3 oxide scales in these
images produced by differential heat conduction material systems and detects the shift in the charac-
(using lasers or other heat sources) – essentially teristic peaks produced: the greater the shift the more
relying on cracked regions of the coating system stressed the TGO. This technique can be used to
not allowing heat to pass through as readily as fully examine a coated TBC component and to detect
adherent coated regions; variations in stresses in the TGO (cracked areas
Top coat valley
TGO centre
Top coat peak
Figure 11 Finite element approach to TBC spallation modeling as a function of TGO growth. Reproduced from Hermosilla, U.
PhD Thesis, Nottingham University 2008; UK EPSRC Supergen Plant Life Extension Consortium.
showing up as low stressed TGO). This technique 6.6 6.8 7 7.2 7.4 7.6 7.8 8 8.2
0
works better on EB-PVD TBCs than APS TBCs due –0.5 Activation energy from
NiCoCrAIY bond coats kp α-Al2O3 growth on
to the columnar microstructure of the former that –1 Pt.Al Alloys = 85 kcal/mole
allows light signals to pass through more easily. –1.5 Same slopes Activation energy
–2 = 85.1 kcal/mole
log 1st
–2.5
1.20.4.2.2 Modeling TBC spallation –3
Platinum aluminide Pt overlayer on NiCoCrAIY
Several industrial and academic research groups are –3.5 bond coats
14791
working on the development of models to predict –4
spallation of TBCs.31–37 There are essentially two –4.5
–5
distinct approaches:
104/T
Finite element modeling of a selected area of an
Figure 12 Observations of TBC spallation that form a
idealized TBC system to predict development of basis for analytical approaches to life modeling.
stresses around TGO as it grows (e.g., Figure 11); Reproduced from Yanar, N. M.; Kim, G.; Hamano, S.;
these models are being developed by several EU Petit, F. S.; Meier, G. H. In Microscopy of Oxidation,
and US research groups.32–34,36,37 Proceedings of 5th International Conference, Newcomb, S.
Phenomenological approaches based on observa- and Tatlock, G., Eds.; Science Reviews, 2003; pp 249–260.
tion of the performance of many TBC-coated

more rapidly and can be life limiting for all compo-
samples under a range of conditions to identify
nents. As a result, hot corrosion of gas turbine hot gas
the most important variables and their effects as a
path materials has been the subject of active research
function of exposure time and temperature (e.g.,
and industrial experience for more than 50 years.2
Figure 12); these types of models are being dev-
The potential problems that may be encountered in
eloped particularly by Dechema and Cranfield in
gas turbines fired on conventional natural gas, diesel,
Europe31,35 and Pittsburg in USA.29
and jet fuels have been well characterized,2,15,38–45 and
fuel standards derived (Table 3). In recent years, the
1.20.5 Hot Corrosion ranges of fuels that need to be considered have wid-
ened in response to changes in availability and price. In
Although oxidation of gas turbine hot gas path addition, investigations have also been carried out into
materials is rarely life-limiting, unless localized over- the potential effects on gas turbines from using fuel
heating has occurred (e.g., following TBC spallation), gases derived from solid fuels (e.g., gasification of coal
the hot corrosion of these materials can occur much and/or biomass).47
Table 3 Fuel specifications46
Fuel type True distillates Ash-forming fuels
Kerosene No.2 Distillate Blended residuals and crude Heavy residuals
Sulfur (%) 0.01/0.1 0.1/0.8 0.2/3 0.5/4

Ash (as delivered) (ppm) 1/5 2/50 25/200 100/1000
Ash (inhibited) (ppm) – – 25/250 100/7000
Trace metal contaminants (untreated) (ppm)
Sodium + potassium 0/0.5 0/1 1/100 1/350
Vanadium 0/0.1 0/0.1 0.1/80 5/400
Lead 0/0.5 0/1 0/1 0/25
Calcium 0/1 0/2 0/10 0/50
100 μm
(a) 600 μm (b)
Type I hot corrosion Type II hot corrosion

Figure 13 Typical appearance of type I and type II hot corrosion during their propagation stages.50 (a) Type I hot corrosion
and (b) type II hot corrosion.
These studies have successfully identified the two widely recognized forms of hot corrosion, distin-
causes of hot corrosion and some methods to limit guished by their propagation modes in gas turbines
the damage caused. A considerable amount of data on (Figure 13)2,15,38–45:
hot corrosion damage for a wide range of materials,
Type I hot corrosion has traditionally been
under different exposure conditions, has been gen-
observed in a temperature band between the melt-
erated during the course of research projects and
ing point of the surface deposit and vapor deposi-
industrial usage.2,42 This dataset has permitted,
tion dew point for the deposit (above this dew
experimentally determined, safe operating conditions
point deposit is unstable and will evaporate). An
to be produced. However, the development of quan-
example of this could be the formation of sodium
titative predictive models for hot corrosion damage
sulfate deposits. This form of corrosion damage is
to materials of current interest in gas turbines is an
characterized by the appearance of internal sulfide
on-going activity,48,49 with some embryonic models
phases in a zone of selective alloy depletion beneath
produced for these complex corrosion phenomena.
a fairly smooth scale–metal interface. The typical
temperature ranges quoted for this type of attack
1.20.5.1 Hot Corrosion Mechanisms
are typically 750–900 C. This form of attack has
Hot corrosion of materials in a power turbine occurs been called high temperature hot corrosion.
when a liquid deposit forms on the surface of a com- Type II hot corrosion occurs at a lower range of
ponent. After a period of incubation as the deposit temperatures (typically quoted as 600–750 C).
forms on the surface and reacts with the initially This form of attack relies on the formation of a
protective oxide, hot corrosion reactions move into complex mixed metal sulfate deposit: an example
their more rapid propagation stages. There are of this could be the formation of a mixed nickel
sulfate–sodium sulfate deposit. Such deposits are 4
(mole metal ions/mole Na2SO4)

not very stable and require the presence of signifi- Cr2O3
Log concentration ppm

cant levels of SO3 to stabilize them (hence the term
Fe2O3
‘gas phase induced acidic fluxing’). The corrosion 3
occurs over the stability range of the liquid deposit NiO
from its melting point to its dissociation tempera-
Al2O3
ture. The attack is characterized by the formation of 2
pitted or uneven surfaces with layers of corrosion Co3O4
products/deposits on the surface and no subsurface
depletion zones or internal sulfide formations 1 SiO2(900 ⬚C)
(although a layer of sulfides may be observed at 5 7 9 11 13 15 17
the metal–scale interface). This form of attack has –Log aNa2O
been called low temperature hot corrosion.
Figure 14 Oxide solubility in molten Na2SO4 at 927 C.
Reproduced from Sims, C. T.; Stoloff, N. S.; Hagel, W. C.
The mechanisms of hot corrosion are generally
Eds. Superalloys; Wiley: New York, USA, 1987; Vol. II.
believed to be a result of the interaction of the surface
oxide scales at first and then of the underlying alloys
with molten surface deposits.2,15,38,39,42 These are KCl-NaCl-K2SO4-Na2SO4
commonly described as fluxing mechanisms. Both Na2SO4 K2SO4
acidic and basic fluxing reactions are possible depend- 884⬚ ~932⬚ 1069⬚
ing on the deposit composition and the composition of 95
0
oxide/underlying coating/alloy. P108⬚
P696⬚ 85
In basic fluxing, oxide ions are produced in a P686⬚ P736⬚ 0
deposit of Na2SO4 by removal of oxygen and sulfur [II]
75
0
[I] [III]
from the deposit by reaction with underlying coating/ Na 2S
O4
65 K
628⬚ 0
alloy. Then, the oxide scales (e.g., MO) that form can ⬚ P567⬚ 2S
O
P576P ⬚
react with the oxide ions via reactions such as2,15,38: 536⬚ 534 P550⬚
4
55 P
N
0
aC
MO þ O2 ¼ MO2
l
2 0 KCl 690⬚
65 60
In acidic fluxing, the oxide scale dissolves to 0
donate oxide ions to the deposit melt2,15,38:

0
70
MO ¼ M2þ þ O2
75
0
Acidic conditions can be developed in a molten

deposit by at least two different processes, leading 800⬚ ~658⬚ 774⬚
to the terminology gas phase acidic fluxing and alloy- Na2Cl2 K2Cl2
Mol %
induced acidic fluxing.
As a result of these different possible hot corrosion Figure 15 Na2SO4–K2SO4–NaCl–KCl phase diagram.
Reproduced from Phase Diagrams for Ceramists; American
mechanisms, the solubility of commonly formed oxi-
Ceramic Society, 1964.
des in molten deposits by acidic and basic reactions has
been the subject of several studies (e.g., Figure 14).
In practice, the temperature ranges of type I and combustion gas stream (e.g., calcium, magnesium,
type II hot corrosion are affected by the composition lead, or chlorine) or elements from the components
of deposits formed. For example, the presence of being corroded (e.g., molybdenum or tungsten).
potassium in a deposit can significantly lower the Some of these elements may also participate in the
melting point of the deposit and lower the tempera- hot corrosion reactions between the deposit and the
ture range for type I when compared to sodium component materials.
sulfate alone (Figure 15). Similarly, NaCl may also Along with the distinctive propagation morpholo-
lower deposit melting points51 and also assist the gies of types I and II hot corrosion, it is also possible
breakdown of any protective oxide scales. for an intermediate form of attack to develop which
Other elements can also affect the melting ranges has some of the features of both. This has been
of these deposits: these may be contaminants from described as mixed mode attack or a transition-type
attack. It is characterized by the formation of layered combustor, these contaminants can join those exit-
corrosion products/deposit on the component surface ing in the compressor and react as they pass through
and intermittent areas of internal sulfides/depletion the combustor and power turbine.
zones and an uneven scale/metal interface. This form Salts entering the compressor can be diverted from
of attack has been observed at temperatures between the main air flow into smaller flows that are used to
traditional type I and type II corrosion regimes. cool components in the power turbine; these salts
An alternative mode of hot corrosion attack is seen do not get the chance to react with the fuel derived
in turbines in which the combustion gas stream con- contaminants as they pass through the combustor,
tains vanadium (usually a contaminant that arrives so different compositions of deposits can form
via a fuel oil). This form of hot corrosion also from this contaminant routing.
depends on the formation of liquid deposits on the
For gas turbines that operate under steady condi-
surfaces of components, but these low-melting-point
tions (i.e., some larger land-based gas turbines), it is
deposits do not require the presence of high levels of
possible to use knowledge of these conditions to
SO3. In the presence of sodium-containing deposits,
calculate the total contaminants entering the system
mixed deposits, for example, sodium vanadyl vanadate
(whether from fuel or oxidant or coolant) giving rise
may form with melting points as low as 535 C.51 These
to an equivalent fuel contaminant level49,54,55:
sodium vanadate compounds can flux the protective
oxide from component surfaces and then continue XT ¼ ðA=F ÞXA þ ðS=F ÞXS þ XF
with a fluxing reaction with the metallic substrates
where XT is the total contamination in fuel equiva-
below. It has been found that this form of hot corrosion
lent (ppm) by weight; A/F is the air to fuel mass flow
can be suppressed by deliberately adding dopants (such
ratio; S/F is the steam (water) to fuel mass flow ratio;
as magnesium or chromium compounds) to the fuel;
XA is the contaminant concentration in inlet air
these are designed to react with the vanadium species as
(ppm) by weight; XS is the contaminant concentra-
they pass through the combustion process and deposit
tion in injected steam or water (ppm) by weight; XF is
onto component surfaces as solid deposits with high
the contaminant concentration in fuel (ppm) by
melting points that remain stable and unreactive.
weight.
This approach provides a method of assessing
1.20.5.2 Deposit Formation and Fuel
potential corrosiveness of gas streams within the gas
Specifications
turbine and is one route to setting specifications for
For hot corrosion to occur, a liquid or partially mol- safe operation of such turbines.
ten deposit must be created on the surface of compo- It is also possible to consider how the various
nents. Such deposits can be formed by a number of contaminant species that enter the power turbine
different mechanisms,52,53 but all require the com- can deposit out onto components. There are several
busted gas streams passing through the gas turbine well-known deposition processes (and associated
hot gas path to contain both trace metal species (such models52,53,56–62) that can occur within a gas turbine
as alkali metals) and reactive gas species (such as hot gas path (illustrated schematically in Figure 16
SO2/SO3). The source of the contaminants may be for an aerofoil):
either the fuel or the oxidant (usually air) or compo-
Direct inertial impact ion; larger particles
nent coolant (usually air, but sometimes steam in
(>5 mm) from the gas stream impact on compo-
large land based gas turbines) or a mixture of all
nents instead of following the gas flows around
three. For example:
them; depending on their properties and the sur-
Sea salt contains a mixture of metal chlorides and face condition of the components, these particles
lower levels of sulfates: these contaminants may may stick to the component surface, bounce off (or
enter a turbine through its air inlet, pass through partially stick).
the compressor and the combustor before enter- Eddy diffusion; smaller particles may become
ing the power turbine. The salts may build up as entrained in the turbulent eddies that can form in
deposits on the compressor vanes/blades before certain locations in the gas paths (e.g., towards the
shedding and passing further into the turbine as trailing edge of aerofoil suction surfaces, or around
particles, or may pass through as vapor or aerosols. blade platforms, disc edges, seals, etc.); these par-
Gas turbine fuels usually contain sulfur compounds ticles can then impact on component surfaces, and
together with a number of trace metal species; in the again depending on their properties and the surface
Diffusion
High velocity
thin boundary Suction
layer side
Film cooling Air cooling

provides barrier increases DT
to deposition and thermophoresis Alkali deposition when
surface temperature
below dewpoint
Inertial
Low-velocity deposition
thicker boundary layer
Pressure
side
Figure 16 Potential deposition around a gas turbine aerofoil surface. Reproduced from Grimethorpe Topping Cycle
Project, UK DTI Clean Coal Programme, Summary Report 1995.
condition of the components, these particles may temperature gradient) are well known but are
stick to the component surface, bounce off (or par- believed to be less probable in the environments
tially stick). It is quite likely that the composition of present within a gas turbine.
the particles will vary with size, so this may cause It is clear from the brief descriptions of the differ-
the behavior of larger and smaller particles to differ, ent types of deposition processes above that the
as well as the differences in impact conditions. deposition mechanisms that will operate will depend
Vapor deposition; this transfers trace elements on a large number of potential variables that include
from the bulk gas stream to the surface of compo- ones resulting from (a) the composition of the com-
nents; however, there are several variations of this busted gas flow and (b) gas turbine component opera-
type of process which include: tions/geometry.
condensation of a vapor of a particular species Combusted gas stream factors (local to area of
from a hot gas stream onto a cooler metal sur- deposit formation):
face (e.g., gaseous sodium sulfate condensing
trace metal contaminant levels,
out as solid or liquid sodium sulfate);
condensation of a vapor of a particular species SO2/SO3 levels,
particle sizes,
from a hot gas stream into a solid or liquid of
particle composition.
another species on a cooler surface (e.g., gaseous
sodium hydroxide or sodium chloride condens- Gas turbine component operations and geometri-
ing out as solid or liquid sodium sulfate); reac- cal factors (local to area of deposit formation):
tion with SO2/SO3 could happen during the
condensation process or on the component sur- bulk gas temperatures,
face after condensation; component temperatures,
these can be complex processes that depend on difference between bulk gas and metal temperatures,
the activities and relative stabilities of all the gas pressure,
contaminants involved in the gas stream and at component shape and size, affecting:
the surface of the components, as well as the gas flows around component,
temperature differences between the bulk gas development of eddy flows,
stream and component surface, system pressure, development and thickness surface boundary
boundary layer thickness, and other flow related layers.
parameters.
The source of the contaminant can alter the form
Other deposition processes such as Brownian diffu- in which it enters the gas turbine, passes through it
sion or thermophoresis (i.e., diffusion along a and deposits onto components in the power turbine.
For example, sodium chloride can enter the system Pressurized combustion systems (such as pressur-
from the fuel or combustion air, pass through the ized fluidized bed combustion) in which solid fuels
combustor as a vapor or as particles (or both), and are burnt, and then the hot pressurized gas stream
then deposit via direct inertial impaction, eddy produced cleaned and passed through the power
impaction, and/or vapor condensation on different turbine part of the gas turbine (in some cases, the
areas of an aerofoil. compression part could be used to produce the
For a given gas turbine design, only factors related pressurized air stream needed for the combustion
to the combusted gas stream composition may be open process);
to an operator to control (and so minimize deposition Use of natural gases with increasingly high
and hot corrosion). Gas turbine manufacturers, some H2S levels (beyond the range of previously accept-
gas turbine operators, and standards bodies have all able gas compositions);
considered the potential damage that can be caused to Natural gases or syngases are processed to reduce
gas turbines by fuel contaminants. As a result, a num- their carbon contents and enhance their hydrogen
ber of fuel specifications have been produced that contents;
include limitations on elements related to hot corro- Bio-oil fuels;
sion (the fuel standards also include limitations on Gas turbines fired using oxygen and any of the
many other fuel properties for other purposes). Exam- other potential fuels to produce a combusted gas
ples of such fuel specifications are given in Table 3. stream of mostly CO2 and steam, from which the
All these specifications have the aim of reducing the steam can be condensed to produce a CO2 rich gas
damage caused by contaminants to levels acceptable to stream for recycling into the compressor or for
gas turbine manufacturers or users. storage (after further cleanup).
Along with controlling the contaminants entering
the gas turbine hot gas path via the fuel, for many gas All these systems can be expected to have similar
turbines (land and marine based systems), it is also contaminants (e.g., S, Cl, alkali metals, Ca, Pb, V) to
possible to limit the contaminants entering the system those traditionally found in the various oil or natural
via the combustion air (using air filters to remove gas systems fired on air. However, both the absolute
particles and aerosols). For aeroengines, the contami- and relative levels of these contaminants will be
nants entering the system via the air intakes depend on different in each of the different systems, as a result
the types of flight path, its height, and location in the of different fuels being used in them. In addition,
world: for example, for civil jet engines, island hopping other trace metals (e.g., Cd, Sb, Zn), which may give
with short flights at low altitudes over the sea versus rise to vapor phase species when different fuels
long haul flights with long periods at high altitudes.63 are gasified, may affect the deposition and hot cor-
In addition, it should also be recognized that con- rosion processes in the gas turbines that use the fuel
taminants can be accidentally put onto gas turbine gases produced.49,50,60,64 Coal, biomass, and waste
components via cleaning fluids, greases, and other products have significantly different compositions
lubricants used during component manufacturing, to each other, and these variations persist to differ-
inspection and/or assembly. However, every effort ing degrees for each element/fuel utilization process
should obviously be made to avoid this deposit for- by the time the fuel gas streams reach the gas
mation route wherever possible. turbine.
Fuel specifications have been derived mainly from Thus, new gas turbine fuel specifications will
experience with traditional gas turbine fuels such as be needed for new types of fuels that are being con-
diesels, jet fuels, heavy fuel oils and natural gas. How- sidered. These are currently under active consi-
ever, in recent years, economic and environmental/ deration and in some cases research.14,49,50,60,62
regulatory pressures have encouraged the develop- However, in considering the use of gas turbines for
ment of power systems that use a wider range of new applications, it is necessary to consider both the
fuels, as well as new types of power systems3,14,60,64–66: composition of the combusted gas stream expected in
the power turbine and detailed design of gas turbine as
Integrated gasification combined cycles (IGCC) both should be optimized for the new gas turbine
in which solid fuels such as coal, biomass, and applications, to minimize deposition and subsequent
waste products are gasified, and then the fuel gases hot corrosion (as well as allowing for new bulk gas
(or syngases) produced cleaned up and passed into a compositions in some of the novel power generating
gas turbine for combustion and power generation; cycles).
1.20.5.3 Damage Locations corrosion in different areas of components. It should

be noted that as operating temperatures are slowly
The potential locations for hot corrosion along the
being raised, potential areas for damage are not
gas turbine hot gas path are controlled by deposition
restricted to aerofoil surfaces traditionally consid-
processes (outlined in Section 1.20.4) and the heat
ered. In addition, platforms, shrouds, roots, and
transfer processes between bulk gas stream and var-
shanks may also develop hot corrosion damage as
ious components that enclose it. Heat transfer pro-
may internal cooling air passages. In future advanced
cesses (via radiation, convection, and conduction
gas turbines, one may expect that even outer parts of
mechanisms) cause unique distributions of metal
discs and seals may also suffer from oxidation or
temperatures to be set up on each type of compo-
corrosion damage as the gas temperatures and com-
nent; for example, the vanes and blades at each stage
ponent operating temperatures are raised further.
of any turbine will have different temperature dis-
tributions. The air-cooled vanes/blades in the
higher pressure stages may have wider temperature 1.20.5.4 Rates of Hot Corrosion Attack and
distributions than the ‘uncooled’ vanes/blades in Modeling Hot Corrosion Processes
lower pressure stages. However, the temperature
The rates of hot corrosion attack on a component
profiles on cooled vanes/blades depend on the cool-
depend on the alloy and coating used as well as on
ing methods used (e.g., convection cooling, film
exposure conditions (including local environment).
cooling), details of the blade designs, and whether
As a result of the potential for hot corrosion to be
TBCs are used. Thus, the surface temperature dis-
a life-limiting failure mechanism for some hot gas
tributions on all components along a gas turbine hot
path components within gas turbines, there have been
gas path are unique (and are often commercially
many studies of the mechanisms of hot corro-
sensitive). Figure 17 gives examples of predicted
sion damage in gas turbines over the years.2,15,38–45
temperature distributions on a cooled and an
Failed or damaged components from operating gas
uncooled turbine aerofoil surface.
turbines have confirmed that the various potential
It is quite possible for cooled blades or vanes in the
hot corrosion mechanisms are all possible in practice
higher pressure stages to have the combination of
and have enabled these potential problems to be well
surface temperature and deposition conditions to
characterized. Thus, operating experience and
cause oxidation as well as type I and type II hot
research studies have identified the main factors
that will influence the rates of hot corrosion; in no
particular order, these are2,15,38–45:
metal temperature,
gas temperature,
gas pressure,
rate of deposit formation (or deposition flux),
deposit composition,
gas composition,
component geometry,
alloy composition,
coating composition,
exposure time (hot corrosion damage goes through
an incubation stage before reaching the character-
istic propagation stage).
Some of these factors relate more to deposition
processes than hot corrosion processes, but are nec-
essary to set up conditions needed for hot corrosion
to proceed.
(a) Cooled aero-foil (b) Uncooled aero-foil Unfortunately, each type of component in every
Figure 17 Examples of temperature distributions on (a)
gas turbine operating with different fuels provides a
cooled and (b) uncooled gas turbine aerofoil surfaces. different combination of exposure conditions. Thus
Courtesy of Siemens Industrial Turbo-Machinery Ltd. carrying out systematic studies on all the different
potential variables is not a viable option. However, Such studies have confirmed that base alloys con-
information is available from: taining higher chromium (e.g., IN939, IN6203DS,
IN738LC, and FSX-414) tend to perform well under
manufacturers and operators of fleets of gas tur-
hot corrosion conditions. Such materials are still used
bines, that use fuels from different sources and
for components in many turbines. However, gas turbines
turbines sited or operated in different locations;
that use high component operating temperatures have
rainbow trials of components in operating gas
moved to using single crystal superalloys for blades and
turbines;
vanes in the higher pressure stages. These lower chro-
burner rigs:
mium materials (e.g., CMSX-4) do not have good hot
high velocity,
corrosion resistance, so they require coating to pro-
low velocity,
vide them with the required surface protection.14,49,50
laboratory tests:
Figure 19 gives a qualitative indication of the
deposit recoat methods,
relative performance of different coating types (dat-
dean rig,
ing from the late 1980s).69,70 The introduction of
buried in ash methods.
single crystal materials with poor hot corrosion per-
Moving down the above list, control over some of formance and the differences between type I and
the exposure variables generally increases and their type II hot corrosion mechanisms, that are now
cost decreases, but their relevance to gas turbine known, have complicated this picture.
service also decreases unless care is taken to match Quantitative data on hot corrosion of gas turbine
exposure conditions to particular in-service condi- substrate materials and coatings under well-
tions. In addition it is necessary to ensure that mate- characterized conditions are not frequently reported
rials performance data is produced in a form that is due to the difficulties of carrying out such
useful for comparison with component performance experimental programs and the commercial sensitiv-
determined from use in real environments (i.e., metal ities of the results if they are targeted at particular
loss data rather than weight change data). operating conditions or materials systems.14
Figure 18 illustrates hot corrosion damage found However, over the last 15 years, a series of experi-
for two example gas turbine materials in burner rig mental programs have been carried out at Cranfield
studies carried out to compare the performance of University with the aim of generating systematic data
base alloys and coatings under well-characterized sets of the performance of generic gas turbine mate-
operating conditions. rials (e.g., IN738LC and CMSX-4) under ranges of
1200
1200
1000
Corrosion (μm)
Corrosion (μm)
1000
800
800
600
600 1200 1200
1000 400 1000
)
400 (h
) (h
800 e 800 e
m
Ti m Ti
200 600 200 600
400 400
700 750 200 200
800 850 700 750
800 850
900 Tempera 900 Tempera
tu re (⬚C) ture (⬚C
(a) IN792 )
(b) IN738
Figure 18 Examples of hot corrosion performance of two example materials in burner rig tests as a function of metal
temperature and exposure time.67 (a) IN792 and (b) IN738.
exposure conditions that reflect hot corrosion condi- standards. The datasets generated illustrate some of
tions in industrial and aeroengines that have been the features of hot corrosion damage in gas turbine
accelerated to differing degrees. These datasets are environments (Figures 20 and 21).
in the form of metal losses determined at a number of Figure 20 illustrates three different regimes for
locations to give a cumulative probability distribution hot corrosion damage11,47,48,50:
for the damage observed under specific exposure
conditions. The methods being used for this work At low deposition fluxes; accelerated oxidation pro-
have been included in EU guidelines for hot corro- cesses are observed and the hot corrosion is still in
sion testing that now form parts of draft ISO an incubation phase after this exposure time (500 h);
Ni
Co NiCrAIY
C
rA
IY
Oxidation resistance
CoNiCrAIY
Pl
at
inu
m
alu
m
Aluminides
ini
CoCrAIY
de
s
High chromium
Corrosion resistance - chromium content

Figure 19 Qualitative indication of the relative performance of different coating types.27,68–70
600
100 vppm SOx, 100 vppm HCl
‘Maximum’ corrosion damage (μm)
500 3000 vppm SOx, 300 vppm HCl
‘Buried in ash’
400
300 Flux limited

corrosion
‘Protective’
200 scale formation
(incubation)
100
0
0.1 1 10 100 1000
Alkali sulphate deposition flux (μg cm–2 h–1)
Figure 20 Dependency of type II hot corrosion damage as a function of deposition flux for IN738LC. Reproduced from
Simms, N. J.; Nicholls, J. R.; Oakey, J. E. In Lifetime Modelling of High Temperature Corrosion Processes; Schütze, M. et al.
Eds.; Maney, 2001; pp 379–397, EFC No. 34.
100
μg/cm2/h
Flux 1.5 ug/cm2/h
90 μg/cm2/h
Flux 5.0 ug/cm2/h
CMSX-4 μg/cm2/h
Flux 15.0 ug/cm2/h
(1.5 mg cm–2 h–1) 80 μg/cm2/h
Flux 1.5 ug/cm2/h
μg/cm2/h
Flux 5.0 ug/cm2/h
70 μg/cm2/h
Flux 15.0 ug/cm2/h
Probability [ % ]
60
50
40
30
20
10
0
0 200 400 600 800 1000
Incubation time (h)
Figure 21 Dependence of type II hot corrosion incubation times on deposition flux for CMSX-4 and IN738LC.
Reproduced from Nicholls, J. R.; Simms, N. J.; Encinas-Oropesa, A. In Materials Issues Governing the Performance of
Advanced 21st Century Energy Systems; Levi, T. et al. Eds.; Science Reviews, 2008; pp 35–48.
At intermediate deposition fluxes; corrosion rates process. In terms of hot corrosion, there are several
are proportional to deposition fluxes; issues that are believed to be of increasing interest in
At high deposition fluxes; corrosion rates are not the near future3,14,60,64–66:
sensitive to deposition flux as it is no longer the Corrosion of internal blade cooling passages (from
rate limiting process in hot corrosion reactions
contaminants in the cooling air);
(this is equivalent to previous and older ‘buried Deposit-induced corrosion damage below blade
in ash’ test methods).
platforms/on disc rims (due to increasing compo-
Figure 21 illustrates the much shorter incubation nent temperatures);
times found for the single crystal material CMSX-4 Much higher SOx levels than traditionally consid-
when compared to that of the older alloy IN738LC.48 ered (from dirtier/cheaper fuels) modifying type
In addition, the effect of increasing deposition flux on the I and type II hot corrosion reactions;
reduction of incubation times for the IN738LC is shown. Gasifier fuel gas derived deposits with calcium,
In such laboratory tests, the exposure conditions magnesia, alumina, silica, and iron compounds;
have been adjusted to accelerate the hot corrosion Effects of novel environments (e.g., high steam
damage relative to that expected in operating gas levels from H2 firing, or high CO2/steam mixtures
turbines. However, by determining the sensitivity of from oxy-fired systems) combined with traditional
the damage to changes in exposure conditions, it is gas turbine degradation issues;
possible to determine the expected degree of accel- Interaction between mechanical properties (e.g.,
eration linked to specific operating conditions and fatigue) and hot corrosion;
then compare the data generated to that observed in Development of single crystal alloys with composi-
practice under service conditions. tions that produce higher hot corrosion resistance.
The overall data set being generated by this work
is being used to produce empirical-mechanistic based
models of hot corrosion damage to assist with the
1.20.7 Summary
assessment and prediction of component lives within
the hot gas paths of gas turbines.
Gas turbines are currently used in a wide variety of
applications: from jet engines to large-scale power
1.20.6 Current/Future Issues generation. Each of these applications presents its
own particular set of challenges for the use of specific
The development of gas turbine technologies for gas turbine designs. However, many of these systems
many different potential applications is a continuing experience damage related to oxidation and hot
corrosion in the hot gas paths of gas turbines. These 14. Simms, N. J.; Bale, D. W.; Baxter, D.; Oakey, J. E. In
Materials for Advanced Power Engineering 2002,
various forms of damage are related to a combination Lecomte-Beckers, J. et al. Eds.; Jülich, 2002; p 73, 675–690.
of operating conditions of the particular gas turbine 15. Birks, N.; Meier, G. H.; Petit, F. S. Introduction to the High
and its components together with contaminants Temperature Oxidation of Metals; Cambridge University
Press: Cambridge, 2006.
entering hot gas paths via air or fuel streams. The 16. Schütze, M. et al. Eds. Lifetime Modelling of High
various forms of oxidation and hot corrosion that are Temperature Corrosion Processes; Maney, 2001, EFC
expected under such exposure conditions are well No. 34.
17. Nijdam, T. J.; van der Pers, N. M.; Sloof, W. G. In Novel
known and are characterized with qualitative descrip- Approaches to Improving High Temperature Corrosion
tion of the potential mechanisms developed mostly Resistance; Schütze, M. and Quadakkers, W. J., Eds.;
during the 1970s. However, quantitative data on the Woodhead, 2008; pp 582–598, EFC No. 47.
18. Smialek, J. L.; Nesbitt, J. A.; Barrett, C. A.; Lowell, C. E. In
various damaging reactions are less readily available, Cyclic Oxidation of High Temperature Materials;
and models to enable the prediction of such damage Schütze, M. et al. Eds.; IOM Communicvations, 1999,
are subjects of current research activity. The use of Chapter 9, EFC No. 27.
19. Nicholls, J. R.; Newton, R.; Bennett, M. J.; Evans, H. E.;
new materials designed for strength rather than cor- Al-Badairy, H.; Tatlock, G. J.; Naumenko, D.;
rosion resistance is proving to be problematic for Quadakkers, W. J.; Strehl, G.; Borchardt, G. In Lifetime
some gas turbine applications. The on-going wish to Modelling of High Temperature Corrosion Processes;
Schütze, M. et al. Eds.; Maney, 2001; pp 83–106, EFC No. 34.
increase gas turbine efficiencies and at the same time 20. Nicholls, J. R. In Life Assessment of Hot Section Gas
widening the range of potential fuels and including Turbine Components; Townsend, R. et al. Eds.; IOM
new types of power generation system, to meet envi- Communications, 2000; pp 135–154.
21. Poquillon, D.; Monceau, D. Oxid. Met. 2003, 59, 409.
ronmental and regulatory wishes, is presenting a new 22. Nesbitt, J. A. In Lifetime Modelling of High Temperature
series of challenges in terms of the oxidation and hot Corrosion Processes; Schütze, M. et al. Eds.; Maney,
corrosion performance of components in various gas 2001; pp 359–378, EFC No. 34.
23. Dahl, K. V.; Hald, J. In Materials for Advanced Power
turbine hot gas paths. Engineering 2006; Lecomte-Beckers, J. et al. Eds.; Jülich,
2006; pp 675–690.
24. Bertrand, N.; Desgranges, C.; Nastar, M.; Girardin, G;
References Poquillon, D.; Monceau, D. Mater. Sci. Forum 2008,
595–598, 463–472.
25. Bacos, M. P.; Josso, P.; Vialas, N.; Poquillon, D.;
1. Rolls Royce: The jet engine. Rolls Royce, 2005. Pieraggi, B.; Monceau, D.; Nicholls, J. R.; Simms, N.;
2. Sims, C. T.; Stoloff, N. S.; Hagel, W. C. Eds. Superalloys; Encinas-Oropesa, A.; Ericsson, T.; et al. Appl. Therm. Eng.
Wiley: New York, USA, 1987; Vol. II. 2004, 24, 1745–1753.
3. Hannis, J. M. In Power Generation in an Era of Climate 26. Simms, N. J.; Kilgallon, P. J.; Roach, C.; Oakey, J. E. In
Change; Proceedings of the 7th International Charles Microscopy of Oxidation, Proceedings of 5th International
Parsons Turbine Conference; Strang, A. et al. Eds.; IOM Conference, Newcomb, S. and Tatlock, G. Eds.; Science
Communications, 2007; pp 37–49. Reviews, 2003; pp 273–280.
4. Data sheet for SGT5-8000H; Siemens AG, 2007. 27. Nicholls, J. R. Mater. Res. Soc. Bull. 2003, 28(9),
5. Pavri, R.; Moore, G. D. General Electric Reference 659–670.
Document, GER-4211, 2001. 28. Newcomb, S. and Tatlock, G. Eds.; Microscopy of
6. De Backer, W.; Bopp, C. In Proceedings of ASME Oxidation; Proceedings of 5th International Conference
TURBOEXPO 2000, Munich, Germany 2000; Paper Science Reviews, 2003.
2000-GT-0307. 29. Yanar, N. M.; Kim, G.; Hamano, S.; Petit, F. S.; Meier, G. H.
7. Ingistov, S. Proceedings of ASME TURBOEXPO 2000, In Microscopy of Oxidation, Proceedings of 5th
Munich, Germany, 2000; Paper 2000-GT-305. International Conference, Newcomb, S. and Tatlock, G.,
8. Horlock, J. H. Proceedings of ASME TURBOEXPO 2001, Eds.; Science Reviews, 2003; pp 249–260.
New Orleans, Louisiana, 2001; Paper 2001-GT-0343. 30. Osgerby, S.; Rinaldi, C.; De Maria, L. In Materials for
9. Bhargava, R.; Meher-Homji, C. B. In Proceedings of ASME Advanced Power Engineering 2006; Lecomte-Beckers, J.
TURBOEXPO 2001, Amsterdam, The Netherlands 2002; et al. Eds.; Jülich, 2006; pp 217–232.
Paper 2002–30560. 31. Renusch, D.; Schütze, M. Mater. Sci. Forum 2008,
10. Bradshaw, A.; Simms, N. J.; Nicholls, J. R. Fuel 2008, 87, 595–598, 151–158.
3529–3536. 32. Busso, E. P.; Lin, J.; Sakurai, S.; Nakayama, M. Acta
11. Simms, N. J.; Nicholls, J. R.; Oakey, J. E. In Lifetime Mater. 2001, 49, 1515.
Modelling of High Temperature Corrosion Processes; 33. Busso, E. P.; Evans, H. E.; Wright, L.; McCartney, L. N.;
Schütze, M. et al. Eds.; Maney, 2001; pp 379–397, EFC Nunn, J. E.; Osgerby, S. Mater. Corros. 2008, 59(7),
No. 34. 556–565.
12. ADSEALS – Investigation in advanced high temperature 34. Hermosilla, U. PhD Thesis, Nottingham University 2008;
seals, European Union FP5 project, G4RD-CT-2000- UK EPSRC Supergen Plant Life Extension Consortium.
00185, 2004 35. Nicholls, J. R. DARP: Advanced Aero-engine Materials
13. Rickerby, D. S.; Hicks, M. A. In Power Generation in an Era (ADAM) – Next Generation Single Crystal and Lifing
of Climate Change, Proceedings of the 7th International Methodologies; GR/S26149/01, Final Report, EPSRC, 2007.
Charles Parsons Turbine Conference; Strang, A. et al. 36. Padture, N. P.; Gell, M.; Jordan, E. H. Science 2002, 296,
Eds.; IOM Communications, 2007; pp 229–240. 280–284.
37. He, M. Y.; Hutchinson, J. W.; Evans, A. G. Mater. Sci. Eng. 57. Fackrell, J. E.; Tabberer, R. J.; Young, J. B. In The Impact
A 2003, 345, 172–178. of Ash Deposition in Coal Fired Plants; Williamson, J. et al.
38. Kofstad, P. High Temperature Corrosion; Springer, 1988. Eds.; Taylor and Francis, 1994; pp 123–134.
39. Giggins, C. S.; Pettit, F. S. Hot Corrosion of Metals and 58. Hepworth, J. K.; Fackrell, J. E.; Pinder, L. W.; Wilson, J. D.
Alloys – A unified Theory, PWA Report No. FR-11545, 1979. In Life Assessment of Hot Section Gas Turbine
40. Corrosion, 9th ed.; ASM International: Ohio, USA, 1987; Components, Townsend, R. et al. Eds.; IOM
Vol. 13, ISBN 0-87170-007-7. Communications, 2000; pp 281–292.
41. Stringer, J.; Wright, I. G. Oxidation Metals 1995; 44(1/2), 59. Simms, N. J.; Oakey, J. E.; Stephenson, D. J.; Smith, P. J.;
265–308. Nicholls, J. R. In Proceedings 8th International Conference
42. Special Issue on Hot-salt corrosion standards, test on Erosion by Liquid and Solid Impact; Wear 1995;
procedures and performance, High Temp. Technol. 186–187(Part 1), 247–255.
1989, 7(4). 60. DeCorso, M.; Newby, R.; Anson, D.; Wenglarz, R.;
43. Rapp, R. A. Corros. Sci. 2002, 44, 209–221. Wright, I. Coal/biomass fuels and the gas turbine:
44. Rapp, R. A.; Goto, K. S. In Symposium on Fused Salts; Utilization of solid fuels and their derivatives; ASME,
Braunstein, J., Selman, J. R., Eds.; The Electrochemical Paper 96-AT-76, 1996.
Society, 1979; p 159. 61. Fackrell, J. E.; Tabberer, R. J.; Young, J. B.; Wu,
45. Shores, D. A. In High Temperature Corrosion; Rapp, R. A., Z. Gas turbines fired on coal-derived gases –
Ed.; NACE, 1983; p 493. Modelling of particulate and vapour deposition, UK
46. Foster, A. D.; von Doering, H. E.; Hilt, M. B. Fuels Flexibility DTI Report No. COAL R280, DTI/Pub URN 05/661,
in Heavy Duty Gas Turbines; GE: Schenectady, NY, 1983; March 2005.
GER3428a. 62. Grimethorpe Topping Cycle Project, UK DTI Clean Coal
47. Simms, N. J.; Nicholls, J. R.; Oakey, J. E. Mater. Sci. Programme, Summary Report 1995.
Forum 2001, 369–372, 833–840. 63. Leggett, A. J.; Simms, N. J.; Rickerby, D. S. In Power
48. Nicholls, J. R.; Simms, N. J.; Encinas-Oropesa, A. In Generation in an Era of Climate Change, Proceedings of
Materials Issues Governing the Performance of Advanced the 7th International Charles Parsons Turbine Conference;
21st Century Energy Systems; Levi, T. et al. Eds.; Science Strang, A. et al. Eds.; IOM Communications, 2007; pp
Reviews, 2008; pp 35–48. 241–251.
49. Bordenet, B.; Bossman, H. P. Mater. Sci. Forum 2004, 64. Wenglarz, R.; Oakey, J. E. In Materials for Advanced
461–464, 907–916. Power Engineering 2006, Lecomte-Beckers, J. et al. Eds.;
50. Simms, N. J.; Encinas-Oropesa, A.; Nicholls, J. R. Mater. Jülich, 2006; pp 233–244.
Sci. Forum 2004, 461–464, 941–948. 65. Ruth, L. A. In Life Cycle Issues in Advanced Energy
51. Phase Diagrams for Ceramists; American Ceramic Systems, Norton, J. F. et al. Eds.; Science Reviews, 2003;
Society, 1964. pp 7–14.
52. Hamed, A.; Tabakoff, W.; Wenglarz, R. J. Propul. Power 66. Paisley, M. A.; Welch, M. J. In ASME Turbo Expo 2003;
2006, 22(2), 350–360. ASME, 2003; GT2003-38294.
53. Young, J. B.; Wu, Z. Particulate and vapour deposition in 67. High Temperature Corrosion of Aerospace Alloys,
gas turbines fired on coal-derived gases, Final report DTI AGARD-AG-200, 1975.
Cleaner Coal R&D Programme. Project 142, 2004. 68. Coatings for High Temperature Structural Materials,
54. Foster, A. D.; et al. Fuels flexibility in heavy-duty gas NMAB report of the NRC; National Academic Press:
turbines, General Electric Reference Document, Washington DC, 1996.
GER-3428a; GE: Schenectady, NY, 1983. 69. Mom, J. A. NLR Report MP 81003U, Amsterdam, 1981.
55. Wolf, J.; Werning, H. J. Proceedings of VGB Conference 70. Novak, R. C. Coating development and use: Case studies.
on ‘Corrosion and Corrosion Protection in Power Plant Presentation to the Committee on Coating for High-
Technology 1995 (VGB, 1995), paper 27. Temperature Structural Materials; National Materials
56. Fackrell, J. E.; Tabberer, R. J.; Young, J. B.; Fantom, I. R. Adisory Board; National Research Council: Irvine;
ASME International Gas Turbine & Aeroengine Congress, California, April 18–19, 1944. Cited in references [27]
The Hague, 1994; Paper 94-GT-177. and [68].
1.21 Design of High Temperature Alloys
P. F. Tortorelli and M. P. Brady
Oak Ridge National Laboratory, Oak Ridge, TN 37831-6115, USA

1.21.2 Protective Oxidation 542
1.21.3 Selective Oxidation and Establishment of a Protective Scale on an Alloy 543
1.21.4 Minimizing Oxide Growth and Improving Scale Adhesion by Alloying 546
1.21.5 Maximizing Oxidation Lifetime 547
1.21.6 Alloy Selection for Specific Environments 549
1.21.7 Conclusions 555
References 555
protective surface layer. Such surface layer formation

Abbreviations
involves a complex interplay of reactive and diffusive
AA Accelerated attack
processes, mediated by both material and environ-
AAS Accelerated attack with spallation
mental factors. There are well known principles to
POS Protective oxide scale
overcome this design problem, particularly for the
simpler alloy systems that exhibit behaviors that are
a close match to the assumptions made to produce
Symbols analytical expressions of reaction rates and species
d Specimen thickness fluxes.1 However, designing of alloys for maximal
DAB The alloy interdiffusion coefficient in binary corrosion resistance at high temperatures is often a
alloy A–B where B is the metal that forms the qualitative endeavor, due to the large set of possi-
protective oxide BxOy ble alloy and environmental permutations that make
DO Diffusion coefficient of oxygen in alloy A–B quantification and prediction difficult. Therefore,
k Rate constant where (kt)n is the rate equation for material specifications for corrosion resistance are
B consumption by oxidation often based on empirical databases and experiences
M Mass conversion factor for metal and oxide that lack the quantification needed for directly link-
NB Concentration of B needed to form ing materials selection to the component design life-
protective BxOy time. Nevertheless, there are a number of cases where
NB,c Critical concentration below which a alloy composition can be manipulated, based on an
protective BxOy cannot be maintained underlying understanding of corrosion principles and
NB Concentration of B in the bulk alloy A–B their verification, to derive good or excellent corro-
NO,s Solubility of oxygen in alloy A–B sion behavior in appropriate high temperature envir-
t Time of exposure onments. In this chapter, certain approaches based on
tb Time to breakaway some of these basic principles and empirical findings
x Depth of internal oxidation will be reviewed with selected alloy systems and
r Alloy density environments used as examples. Many more specific
examples of alloy design or optimization for high
temperature corrosion resistance can be found in
other chapters of this book.
1.21.1 Introduction The essence of the alloy design effort for high
temperature corrosion resistance is to achieve a con-
The design of alloys for corrosion resistance at high tinuous and protective surface layer or scale. Once a
temperatures involves specification and manipulation layer of this type is established, the access of the alloy
of alloy composition and microstructures that assure to the environment is effectively isolated and, there-
preferential formation and long-term stability of a after, is controlled by the transport properties and
541
542 High Temperature Materials
stability of the surface scale as long as its integrity is of protective scales for the reasons discussed above. In
maintained. This type of scale is most often an oxide this hypothetical A–B alloy notation scheme, element
because virtually all high temperature, technically A is typically a base metal such as Fe, Ni, or Co, while
relevant environments contain reactive oxygen. Fur- B forms the desired protective oxide:
ther, at equivalent gas activities and, frequently at
xB þ ðy=2ÞO2 ¼ Bx Oy ; ½1
very small concentrations of O2 in many mixed-gas
environments, oxides are more thermodynamically with possible modification of the oxide by other major
stable than their equivalent carbides, nitrides, sul- and minor alloying elements. Principal examples of
fides, etc. It is important to note, though, that the B would be Al, Cr, and Si for Al2O3, Cr2O3, and SiO2,
approaches to alloy design for protective scale for- respectively. The desired design goal is to formulate an
mation described here, while typically discussed alloy composition and microstructure that will produce
in terms of oxides, are also applicable to situations a continuous layer of BxOy at the expense of less protec-
where other reactive species are present, that is, the tive base metal A oxides. To assure this, BxOy must be
formation of surface or internal carbides, nitrides, thermodynamically stable in the subject environment,
sulfides, etc. by ‘oxidation,’ a chemical process by and, typically, have significantly greater stability than
which the valence state of the metal species is the oxide(s) of A. This thermodynamic tendency can be
increased. Such product formation can be of interest readily assessed qualitatively in terms of the (negative)
as a synthesis route to functional surface structures2 free energies of formation (DG s) as a function of tem-
but, at elevated temperatures in most technically- perature. Such an Ellingham diagram is schematically
relevant environments, these nonoxide phases are shown in Figure 1 for representative metal oxides.
less stable than their equivalent oxides and hence, From an empirical alloy design experience, preferential
can subsequently be converted to oxides. Even if formation of BxOy is found to be easier in alloy systems
thermodynamic stability is achieved, most nonoxides, for which the difference between DGs of the A and
and many oxides, cannot meet the other key criterion B oxides is large. Hence, it is much easier to assure
of protective scale formation: the kinetic standard of Al2O3 formation on Ni–Al than on Ti–Al. However,
a slowly thickening layer that, by itself, is an effective kinetic factors rather than thermodynamic factors are
permeation barrier between the reactive environ- often more important in determining whether a contin-
mental species and the alloy on which it is growing. uous protective scale can be established.
For temperatures in excess of 600 C, the protective In the present context, ‘protective’ has several
oxides relevant for alloys in service applications are attributes. First, BxOy must form preferentially, and
Cr2O3, Al2O3, and, to a lesser extent, SiO2 (as evi-
denced by the fact that most ‘heat-resistant’ alloys
rely on these surface oxides for corrosion resistance).
Increasing thermodynamic stability in oxygen
At lower temperatures, iron, nickel, cobalt, and zirco-

NiO
nium oxides can be considered protective, but in
this chapter, design examples focus on the former MoO2 FeO
three types. Under certain conditions, protective oxi-
des can also serve as efficient barriers to environ-
mental species other than oxygen (such as water Cr2O3
vapor, carbon, or sulfur), thus alloys designed to
form these surface layers may have attractive appli- a 2O 5
O,T SiO 2
cations in high temperature mixed-gas environments Nb
(see Section 1.21.6 on alloy selection for specific O3 es
TiO Al 2 xid
environments). )-o
e,…
,La, C
(Zr, Y
RE
O,
Be
1.21.2 Protective Oxidation
For this study, binary alloys, generally denoted as Increasing temperature

A–B, will be considered with other major alloying Figure 1 Schematic representation of free energies of
elements being treated as minor additions to the formation of oxides of interest. Increasing stability means
binary system, and with oxides as the exclusive form larger negative energies.
Design of High Temperature Alloys 543
reduce the permeation between the reactive species 1.21.3 Selective Oxidation and
in the environment and in the alloying elements – Establishment of a Protective Scale
that is, both the solubility and diffusivity of anionic on an Alloy
species (from the environment) and cationic species
(from the alloy substrate) in the oxide scale must be Managing the inherent instability of a metallic material
low. Therefore, the product must be continuous across in a reactive environment so as to limit the amount of
the alloy surface and exhibit long-term stability in the metal wastage through control of thermodynamic and
environment. Note that this does not necessarily mean kinetic factors is the goal of alloy design for high
that no other products can grow or exist. Rather, the temperature applications. This is first attempted by
protective layer must form in a way that it rapidly the application of well founded, although somewhat
controls the transport across the entire surface of idealized, fundamental principles of selective oxidation
the alloy – see, for example, Figure 2, which shows of alloys that effect preferential formation of a contin-
the development of a continuous Al2O3 scale on a uous protective scale (BxOy) on alloy A–B. (Some
Nb–33Ti–40Al (wt%) alloy under the fast-growing guidelines for choosing BxOy for a given set of envi-
Nb-rich transient (nonprotective) oxides. The protec- ronmental factors are discussed below.) For a single-
tive scale should also be dense (no interconnected phase alloy, A–B, the formalism is based on pioneering
porosity or large pores), mechanically sound (no work of Wagner,6 further studied and elaborated by a
through cracks), and adherent to the alloy substrate host of others.7,8 For multiphase alloys, the develop-
on which it grows. This last attribute, strong adhesion ment of relevant principles is more recent9–11 and
to the alloy, is often problematic even when other takes advantage of analytical advances originating
chemical and physical characteristics are met; these from more extensive work on single-phase alloy oxi-
oxides tend to have coefficients of thermal expansion dation. The treatment in this section follows the out-
(CTEs) that are quite different from those of the alloy line of a recent description of principles for protective
and, thus, large stresses (typically compressive) are scale formation on alloys by Brady et al.12 Specific
encountered upon cooling from high temperature application of these approaches to intermetallics can
such that spallation of the scale eventually occurs.3 be found in Brady and Tortorelli.13
This is a disruptive event; the protective scale regrows As an illustration and reference to the discus-
during the next heating cycle (hopefully) and acceler- sion that follows, it is appropriate to consider the
ates the metal wastage (as well as the attendant subsur- schematic descriptions of the two limiting cases,
face depletion of B) due to reinitiation of the product namely, fully nonprotective behavior (Figure 3, left
growth process.4 Stresses can also develop in the scale side) and the best possible outcome for the alloy design
due to the growth process itself 5 but, typically, these process being addressed here (Figure 3, right side)
are not of the same order of magnitude as the thermal under conditions where O2 is the reactive species,
stresses on cooling, at least at a steady state. A and B oxides are mutually insoluble, and the oxygen
partial pressure is sufficiently high to oxidize either
both A and B or just B. In the former (nonprotective)
Transient oxides limiting case, internal oxidation of B occurs, with the
possible formation of rapidly growing A oxides while,
in the latter limiting case, a slowly growing, exclusive,
continuous, adherent BxOy layer develops.
In the former case, the depth of internal oxidation,
x, at time t can be described functionally as14
2
Continuous Al2O3 x 2 / NO;s =NB DAB DO t ½2
Internal BxOy External BxOy O2

O2
Alloy
5 µm x
A-B A-B
Figure 2 Scanning electron micrograph of cross-section B
B
of Nb–33Ti–40Al (at.%) exposed at 1400 C in air for 15 min.
Reproduced from Brady, M. P.; Gleeson, B.; Wright, I. G. Figure 3 Schematic illustration of internal oxidation and
JOM 2000, 52, 16–21. external scale formation by selective oxidation.
where DO is the diffusion coefficient of oxygen in than NB such that an exclusive protective scale does
A–B, NO,s is the solubility of oxygen in A–B, NB is the not form. Therefore, it is important from an alloy
concentration of B in the bulk alloy, and DAB is the design viewpoint to be aware of how alloying (for
alloy interdiffusion coefficient. From a practical alloy whatever purpose) affects this solubility term to
design viewpoint, the extent of internal oxidation can avoid unintended deleterious effects on the ability of
be decreased most directly by increasing the concen- an alloy to develop a protective surface product.
tration of B. Internal oxidation can also be decreased It is well known that protective scale development
by decreasing the permeability of oxygen in A–B can be promoted by the ‘third-element effect,’ defined
(NO,sDO) via ternary and higher-order alloying addi- generically as an alloying addition that decreases the
tions. Microstructure can impact the extent of inter- amount of B needed to form a protective layer of BxOy
nal oxidation if the morphology of these internal (N B*). This can be an important design principle if
oxides enhances the transport of oxygen along their it is necessary to minimize the concentration of B to
interfaces with the alloy matrix (increased DO).15 achieve other desirable properties of the alloy (for
It is this type of mechanism that can explain the example, decreasing the aluminum concentration to
presence of stable (essentially inert) Al2O3 reinforce- avoid brittleness). For example, such an effect has
ments in a Ni3Al matrix composite structure that been observed for (Ni,Co,Fe)–Al–Cr alloys, where
compromises the otherwise good oxidation resistance the addition of Cr lowers the NA* l value for protective
of the aluminide.16 Al2O3 formation, and is, in fact, the basis for the design
From the Wagner perspective, as NB is increased, of commercial heat-resistant alloys.19
the volume fraction of internal BxOy precipitates There are several mechanistic variants to the
increases to the point where an external layer of third-element effect. The ‘classic’ version is generally
BxOy forms (Figure 3, right side). Under this assump- associated with a secondary gettering mechanism,1,20
tion, the critical concentration needed for external whereby X is added to A–B such that its oxide is
formation of BxOy , NB*, is a function of metal and of intermediate stability compared to that of the
oxygen diffusivities in the alloy, and the solubility of oxides of A and B, and decreases the effective par-
oxygen6 such that tial pressure at the metal–oxide scale interface such
that continuous BxOy layer formation can occur.
NB / NO;s ðDO =DAB Þ1=2 ½3 A second type of third-element effect can be consid-
ered to be the influence of an alloying addition on
Therefore, external BxOy formation is favored (that is, early-stage oxidation, such that a transient oxide can
the critical concentration of B is decreased) by compo- preferentially nucleate the growth of BxOy.21,22 Addi-
sitional and microstructural conditions that decrease tionally, just as microstructure can enhance metal
the solubility and inherent diffusivity of the oxidizing diffusivity in the alloy lattice (see above discussion),
species in the alloy and/or increase the mobility of a third element can be used to influence diffusional
the metallic elements (cf. Figure 3). One way to pref- effects so that protective oxide layer formation is pro-
erentially increase the effective value of DAB and, thus, moted (DAB is increased and/or NO,s and DO are
decrease the value of NB*, is to provide a greater decreased).23
density of short-circuit diffusion paths – for example, The activity(ies) of the reactant(s) in the subject
by decreasing the grain size17 or increasing the dislo- environment must be a part of alloy design consid-
cation density at the alloy surface.18 Although such erations for protective scale development.12 For
structural pathways may not be stable over long per- example, an alloy with NB > N B* at a given reactant
iods of time at elevated temperature, their existence (oxygen) partial pressure may not be able to form a
during the period when the BxOy layer is being estab- protective BxOy layer at higher pressures where the
lished is critical and effective. oxygen solubility in A–B increases in accordance
In theory, ternary and higher order alloying addi- with Sievert’s law and possibly raises the value of
tions that decrease the solubility of the environmental N B* above that of NB (eqn [3]) or where the higher
species in the alloy should, using eqns [2] and [3], pressures result in altering the nature and/or stability
decrease the value of NB* and promote protective of the competing A oxides. Similar considerations
scale formation. Obviously, the opposite is also true: must be given to service temperature, which not
compositional modifications that increase reactant sol- only affects the reactant solubility in the alloy, but
ubility in the alloy can increase NB* to a value greater also other factors that influence the value of N B*.
Despite the importance of multiphase alloys to secondary-gettering mechanism described earlier.

real-world applications, relatively little work in Rather, the addition of Cr introduced a second
high temperature corrosion has been dedicated to (Laves) phase, Ti(Cr,Al)2, which promoted the forma-
comprehensive studies of such materials.9–11,24–28 tion of a protective Al2O3 layer by equilibrating
Nevertheless, some important and intriguing impli- with, and stabilizing, the g-TiAl phase that acted
cations of alloy design have begun to emerge from as an effective Al reservoir and did not transform to
developments in the understanding of how multi- the less oxidation-resistant a2-Ti3Al phase due to
phase alloys oxidize. In this regard, two limiting Al depletion.31,32 Such an example highlights the
cases of high temperature oxidation of a multiphase variety of phase stability relationships in multiphase
alloy9 are shown in Figure 4. In (a), the component alloys that can be exploited as design routes to pro-
phases react independent of each other, and this can tective scale formation. Similar approaches based on
result in corrosion behavior that is worse than that thermodynamic and kinetic pathway factors are cur-
displayed by either phase singly, and is certainly rently being exploited, for example, in the design of
inferior to the other limiting case where the phases multiphase coatings used as bond coats in thermal
act together to form a continuous, exclusive protec- barrier systems33 or as alloy templates for synthesiz-
tive layer, as depicted in (b) of Figure 4. In the latter ing composite structures by controlling gas–metal
case, a solute-rich phase acts as an efficient source or reactions.2,13
‘reservoir’ of B, resulting in a subscale zone that is This section has briefly described the underlying
depleted of this component. The important factors principles for the promotion of a protective scale
include size, shape, distribution, volume fraction, and via alloying or microstructural manipulation. These
composition(s) of the minority phase(s). Normally, a serve as a qualitative guide in assuring that the selec-
finer dispersion of such phases is more efficient in tive oxidation process produces the desired surface
promoting protective scale formation.24,29 A striking layer, as opposed to internal oxidation or a mixed
example of this is shown in Figure 5, where the external scale that is not truly protective. For an A–B
as-cast version of an alloy of Fe–15Cr–0.5C (wt%) alloy, this is achieved by considering conditions that
composition did not form a protective Cr2O3, but a minimize the solubility and diffusion of environmental
version of the same alloy with a refined dispersion of reactants in the alloy and maximize the flux of B to the
carbides did.29 reacting surface – by increasing its initial concentra-
A second approach using alloys to form a reservoir tion, its availability, or its mobility. Once a protective
phase is best exemplified by the addition of Cr to scale is formed, there are other considerations for
g-TiAl, facilitating the formation of a protective maximizing the corrosion resistance and component
layer of Al2O3. Without a minimum of 8–10 at.% lifetime; these are discussed below in terms of alloying
Cr in place of Ti, 60–70 at.% Al was needed for a for minimizing the scale growth and spallation as well
protective alumina scale to form.30 Because Cr is sig- as for materials selection for specific environments.
nificantly more noble in oxygen than Ti or Al, this
alloying benefit was not the traditional third-element
Fe2O3 Cr2O3
A-oxide Carbide-depleted zone
B-oxide
FeO + FeCr2O4
B-oxide 2nd phase-free zone
Alloy
A-rich Alloy
metal matric
100 µm 20 µm
(a) (b)
B-rich (a) (b)
precipitate
Figure 5 Cross-sectional micrographs of Fe–15Cr–0.5C
Figure 4 Schematic depiction of (a) independent and (wt%) steel exposed at 850 C for 72 h in 1 atm O2 (a) as-
(b) cooperative oxidation based on concepts presented in cast and (b) hot-forged. Reproduced from Durham, R. N.;
Gesmundo, F.; Gleeson, B. Oxid. Met. 1995, 44, 211–238. Gleeson, B.; Young, D. J. Oxid. Met. 1998, 50, 139–165.
1.21.4 Minimizing Oxide Growth and the way it is added to or incorporated into the alloy,
Improving Scale Adhesion by Alloying and its concentration (see Figure 6), as well as how it
is dispersed in the alloy matrix.35–38 These additions
It is now a known fact that dilute additions of reactive, can add significant costs to alloy production and,
or rare-earth elements (REs) can substantially decrease because of their reactivity, REs can be difficult to
the oxide scale growth rate and improve the scale control to specified levels during processing, especially
adhesion of alloys that form protective layers of chro- for coatings,40,41 where laboratory-scale benefits have
mia or alumina.34–38 Oxides of REs are highly stable not yet been scaled up for commercialization. It is,
(see Figure 1) and are readily taken up into the therefore, particularly important to weigh all factors
protective oxide layer, where they can help decrease affecting the RE concentration and incorporation into
the steady-state growth rate.35 These active elements the alloy, some of which (such as the use of multiple
can also improve the scale adhesion, in many cases REs and the effects of other minor alloy constituents –
by offsetting the detrimental effects of indigenous see below) have only been widely discussed relatively
impurities, such as sulfur.35,39 Therefore, the most recently.
oxidation-resistant commercial alloys contain minor The choice of a particular RE can be guided by
(<0.1 wt%) concentrations of REs. As an example, empirical findings or rules. For example, as single
Figure 6 illustrates the beneficial effects of certain additions, Hf and Zr are preferable for aluminides,
RE additions on cyclic oxidation lifetime for two while Yis superior in Ni–Cr–Al and Fe–Cr–Al alloys.37
Al2O3-forming alloy systems. For chromia-forming alloys, La seems to be preferred
While all these active elements qualitatively respond based on its large ion size and its higher solubility,
in a similar manner upon exposure to an oxidizing particularly in ferritic alloys.42 In this regard, the low
environment, there are differences in the effective- solubility of Yin many alumina-forming alloys limits its
ness of specific REs. Different effects on oxidation effectiveness.37,43 When a RE exceeds its solubility
behavior are observed depending on the RE chosen, limit, it can be more readily internally oxidized given
Dopants in FeCrAl
None
ODS 0.2 Y (MA956HT)
ODS 0.22 Y (PM2000)
0.08 Y
0.06 Zr (APM)
0.05 Hf
Dopants in Fe–28Al–2Cr
None
0.04 Y
0.10 Y
0.62 Y
ODS 0.21 Y as oxide
0.03 Zr
0.10 Zr
0.05 Hf
0.10 Hf
0.03 mischmetal
0 2000 4000 6000 8000 10 000

Normalized lifetime in 1-h cycles at 1200 ⬚C
Figure 6 Effect of various RE additions to a base Fe–20Cr–10Al or Fe–28Al–2Cr (at.%) alloy in terms of normalized
oxidation lifetime for 1-h thermal cycles at 1200 C. Reproduced from Pint, B. A. In Proceedings of John Stringer Symposium;
Tortorelli, P. F., Hou, P. Y., Wright, I. G., Eds.; ASM International: Materials Park, OH, 2003; pp 9–19.
the active nature of these elements (Figure 1). This Copper

can compromise the oxidation resistance and/or
mechanical integrity (if extreme). The effective use
of a RE requires optimization of its concentration.38
NiAl + Hf Hf/C = 1.63
Figure 6 clearly indicates that the oxidation resistance (a)
10 µm
is quite sensitive to RE concentrations. Copper
The use of two reactive elements (so-called
‘co-doping’) is often more effective than the use of a
single RE,38,43–45 at least for alumina-forming alloys. NiAl + Hf Hf/C = 1.35
(b)
Several commercial alumina-forming alloys and
coatings contain multiple REs.43,44 It is thought
(at least for Fe–Cr–Al alloys) that co-doping allows
Y to be fully effective by keeping it below its solubil-
NiAl + Hf Hf/C = 1.09
ity limit and then supplementing it with a more (c)
soluble RE, such as Hf or Zr. Thus, in general, it 10 µm
Copper
appears that co-doping allows the total RE concen-
tration to be decreased while maintaining good scale
adhesion, and low oxide growth rates. In many cases,
(d) NiAl + Hf Hf/C = 0.94
this then decreases the amount of internal oxidation
relative to the single dopant case.43 Figure 7 Polished cross-sections of the Al2O3 scale
It is clear that attention to microalloying additions formed on coatings of NiAlþHf with differing Hf/C ratios,
and/or impurity control is important in maximiz- after 10 100-h thermal cycles at 1200 C. Reproduced from
ing the oxidation resistance of alloys using REs. Pint, B. A. In Proceedings of John Stringer Symposium;
Sulfur (and possibly other) impurities can have a Tortorelli, P. F., Hou, P. Y., Wright, I. G., Eds.; ASM
International: Materials Park, OH, 2003; pp 9–19.
detrimental effect on scale adhesion.4,46–48 Further,
such indigenous impurities, as well as purposeful
additions for microstructural tailoring and properties There are potential detrimental effects of other
enhancement (such as carbon, nitrogen, boron, etc.), alloying elements on oxidation resistance if they are
can mitigate the positive effects of REs on oxidation sufficiently active. For example, single additions of
resistance.38,45,49,50 For example, in Hf-doped NiAl, 1 at.% Re, Ti, Ta, or Cr (commonly present in coat-
the Hf/C ratio has been found to be important in ings and Ni-base superalloys) were found to degrade
optimizing scale adhesion.49,50 Figure 7 shows that the oxidation resistance of Hf-doped NiAl.50 The Re
NiAl suffered greater scale spallation with decreasing and Cr additions formed precipitates in the alloy and
Hf/C ratios. It has been suggested that, if the Y:S negatively affected the oxide–metal adhesion, while
and/or Hf:C ratios are <1, the alloy will have less the Ti and Ta were oxidized and incorporated into the
than optimal performance.38 Therefore, the active scale as potential sites for scale breakdown.51,52 On
nature of REs makes alloy design with these elements the contrary, certain precious metal additions to NiAl
a delicate optimization process. can improve the oxidation resistance by improving
It should be emphasized that, while effective when the scale adhesion,53,54 presumably by mitigating the
concentrations are optimized for given levels of impu- detrimental effect of sulfur, and promoting alumina
rities or microalloying additions, REs generally do not formation on low-Al (gþg0 ) substrates.33
promote protective scale formation, that is, they do
not make marginal Al2O3 or Cr2O3-forming composi-
tions into robust ones.37 Other approaches, as outlined 1.21.5 Maximizing Oxidation Lifetime
in the previous sections, must be used. In this regard,
however, any indirect effects of RE additions need to be There has been an increasing focus on oxidation
evaluated and separated from the direct influence lifetime, particularly in terms of approaches used
of RE’s on scale growth and adherence. For example, for predicting it accurately for a given material and
a sufficient concentration of RE-rich precipitates in the exposure condition.55 The primary reason for this
alloy matrix can result in a finer grain size and, this trend is that lifetime values can lead to quantification
in itself, can promote protective scale development of oxidation behavior in a way that can be better used
(as discussed in a preceding section).37 by researchers, engineers, systems designers, plant
operators, etc. As the most corrosion-resistant mate- or elsewhere in this book), that slows the rate of
rials have very long lifetimes, verified life prediction scale growth or spallation (decreases kn) or enhances
models provide a way to predict end of life without the promotion of a protective layer (decreases NB,c).
necessarily having to conduct extensive, extended In addition, tb can be increased by increasing the
exposures. Also, lifetime models can provide guid- concentration of B, assuming no alteration of other
ance to the alloy design process by quantifying the factors (e.g., solubilities or DAB) that would destabilize
effect of specific approaches taken to improve high the stability of the protective BxOy layer. However,
temperature corrosion resistance. Some trends based increasing B can affect physical or mechanical prop-
on these are discussed in this section. erties that, in turn, can influence the lifetime through
Usually, lifetime is defined as the time-to-break- effects on k. For example, for Fe–Al systems, simply
away (tb) – the time at which the protective scale fails increasing the concentration of Al can actually
and cannot be reformed, thereby resulting in the decrease the lifetime. This is because the ordered
growth of faster-growing, nonprotective products. structures at higher Al levels result in significantly
The cross-sectional image in Figure 8 captures the higher values of alloy CTE than those of other iron-
process at tb, that is, the transition from protective to based alumina-forming alloys (see Figure 9).60
nonprotective behavior, and illustrates the radical Therefore, greater compressive stresses are experi-
difference in the nature of the respective aluminum enced by the scales on cooling, leading to higher
and iron oxides. After 1101 h, the protective Al2O3 spallation rates and, thus, higher values of k. Hence,
layer on the right-hand side of the image in Figure 8 iron aluminides do not necessarily show substantially
reached the point where it could no longer be main- greater lifetimes than Fe–Cr–Al alloys despite much
tained by the supply of remaining Al from the alloy.56 greater values of NB.56
Therefore, iron oxides formed and rapidly consumed The critical concentration for maintaining the
the remaining metal. Following Quadakkers et al.,57,58 ability to reform a protective layer, NB,c, is dependent
the time-to-breakaway for alloy A–B that forms a on the failure mode of the BxOy layer (intrinsic
protective BxOy layer can be expressed as: chemical failure or mechanically-induced chemical
failure).61 Intrinsic failure is when the actual level of
tb ¼ ½ NB NB;c =100 d =2 r M 1 k1 1=n ½4 B falls below the equilibrium value for formation of
where NB is the initial B concentration, NB,c is the the protective oxide by continued selective oxidation;
critical concentration below which a protective BxOy here, NB,c can be quite low and only achieved under
layer cannot be maintained (different from NB*), d is very limited circumstances when the alloy is very
the thickness of the specimen, r is the alloy density,
M is the mass conversion factor for metal and oxide,
and k is the rate constant, where (kt)n is the rate for 30
consumption of B by oxidation. This k represents the 28
Mean thermal expansion (ppm/⬚C)
total consumption of B by oxide growth as well as 26 40Al

spallation. It can also be dependent (inversely) on d 24
for thin cross-sections.59 22
From an alloy design viewpoint, oxidation lifetime 20 24Al
can be increased by any approach (discussed above 18 28Al
16
22Al 20Al
14
13Al
12 15Al
18Al
Al2O3 10
formation Fe-based 8 a-Al2O3
oxides 6
Figure 8 Cross-sectional image of an Fe–28Al–5Cr–0.2Zr 0 200 400 600 800 1000 1200
(at.%) alloy exposed in air for 1101 h at 1300 C. Other data
Temperature (⬚C)
from these exposures reported in Bennett, M. J.; DeVan,
J. H.; Tortorelli, P. F. The Oxidation Behavior of Iron Figure 9 Mean thermal expansion coefficient versus
Aluminides in Air at 1300 C. In Microscopy of Oxidation 3; temperature for a-Al2O3 and various Fe–Al alloys.
Newcomb, S. B., Little, J. A., Eds.; The Institute of Materials: Reproduced from Pint, B. A.; Porter, W. D.; Wright,
London, 1997; pp 233–245. I. G. Mater. Sci. Forum 2008, 595–598, 1083–1092.
thin (to relieve stresses readily by creep) and tem- internal precipitates and limited means to predict and
peratures very high.61–63 The mechanically induced control their formation.67,68 However, in other cases
chemical failure mode is the more typical one and of multiple reactive species, reactions with oxygen
occurs when the concentration of B becomes lower still predominate and can be used constructively to
than the value needed for kinetically controlled scale promote the formation of a POS that is an effective
rehealing, and the accumulated strain energy is suffi- barrier to other reactive species in the environment.
cient for scale cracking and/or decohesion.3 It may, This is usually enabled, at least in part, by the greater
therefore, be possible to use alloying additions to stability of oxides relative to other products so that
control this process and lower NB,c (raise tb). For oxide formation is possible even at very low oxygen
alumina forming alloys, Cr can increase the lifetime partial pressures. A good example in this regard is the
by allowing Cr2O3 scales to extend protection after preferential establishment of a protective Al2O3 layer
Al2O3 scales can no longer form (intrinsic failure even under the so-called ‘reducing’ sulfidizing environ-
mode)61 or to promote reestablishment of the latter ments, where the partial pressure of oxygen ( pO2)
by a type of third-element effect (mechanically is of the order of 1020 atm while pS2 is 12–14
induced chemical failure mode).63 In this regard, orders of magnitude greater.69 Under these conditions,
alloying additions or heat treatments that affect the approaches to promote the stability, exclusiveness,
mechanical strength of the substrate can also influence integrity, and maintenance of POSs can still have a
oxidation lifetime by modifying the effective strain role in alloy design and selection for mixed-oxidant
energy available for protective scale breakdown37,64 environments. Some examples of alloy selection for
as reflected by k (and n) in eqn [4]. This is another specific environments are given below. Further details
example of unintended, or secondary, consequences of of the relevant corrosion processes are described in
alloy manipulation via effects on chemical, physical, or other chapters of this book.
mechanical factors that affect either the ability to form
a protective scale or to reform it after a failed event. Air/Nitridation. Because of the typically greater
There are additional concerns about the lifetime of stability of the oxides, the alloys that form protective
corrosion-resistant coatings designed for use at high Al2O3, SiO2, and/or Cr2O3 scales often perform well
temperatures. Because of the dependence of tb on the in air with little interference from nitrogen. However,
total amount of B available for protective oxide scale subscale nitride formation is possible if there is a way
(POS) formation (manifested through d in eqn [4]), for the nitrogen to penetrate the oxide scale or dis-
coatings, particularly thin ones, have limited reser- rupt its ability to form a continuous surface layer, or if
voirs of B. Additionally, the depletion of B by oxida- the scale is mechanically breached and cannot heal.
tion (k) is accompanied by loss of B by back-diffusion (This is true, of course, for many oxygen-second
from the coating into the substrate on which it is oxidant systems.) For example, Fe-based Al2O3-form-
deposited.65 Alloy manipulation can be used to con- ing alloys usually perform similarly in oxygen and air
trol this process to a certain extent.33 Barrier layers at but, recently, Pint et al. reported (nonclassical) inter-
the coating–substrate interface may also have to be nal oxidation and nitridation of Fe–Al alloys upon air
considered to prevent undesirable (from an oxidation exposure.70 This accelerated attack was eliminated by
perspective) alloying of the coating by elements diffus- additions of REs or Ti to the Fe–Al alloy, by promot-
ing from the substrate.66 As discussed above, it is often ing a less defective scale. For TiAl intermetallics,
difficult to synthesize coatings in a controlled manner TiN formation during the initial stages of oxidation
with appropriate levels and distributions of REs; so in air has been shown to disrupt the continuity of the
attempts to minimize k in order to maximize lifetime Al2O3 layer, resulting in less protective (alumina þ
are rather more limited in these cases. titania)-based mixed-oxide scales.71 The addition of
Ag or Cr to TiAl helps promote continuous Al2O3
layer formation and ameliorates this problem in
1.21.6 Alloy Selection for Specific part by modifying the nature of the compositional
Environments changes at the alloy–scale interface during the initial
stages of oxidation which can otherwise trigger local
Most high temperature environments involve more TiN formation, and prevent the development of con-
than one potential reactant. In many mixed reactants tinuous alumina layers.32,72
involving O2, N2, H2O, S2, CO, etc., the reaction of The participation of nitrogen as a reactant for air
alloys can be quite complex, with multiple types of exposures can also be triggered by the presence of
ambient water vapor in the environment. As an exam- As already noted, internal nitridation can occur
ple, while Cr2O3 formed on a Cr–Nb alloy in both dry in air environments if a continuous, protective oxide
and humid air, subsurface nitridation only occurred in layer cannot form or is breached. It can also be
the latter case (Figure 10).73 Similar behavior was also observed in nitrogen-rich environments that are
observed for a gamma-based Ti–Al–Cr alloy, with a reducing with respect to even the most stable oxides.
borderline Cr content to promote alumina scale for- Under such conditions, alloys with elements that
mation; protective Al2O3 formed in dry air at 1000 C, minimize nitrogen solubility and diffusivity are less
but subscale nitridation was observed after exposure to susceptible to attack (eqn [2]). In this regard, nickel
humid air.32 For Cr–Nb at 950 C, little difference and cobalt have strong beneficial effects, with the
between dry and humid air was observed, but signifi- added advantage that neither Ni nor Co tend to
cant subscale nitridation was observed when the oxy- form thermodynamically stable nitrides under most
gen and sulfur impurities in the alloy were decreased, technologically relevant nitrogen-containing envir-
with possible synergistic effects between S and water onments.74,75 This impact of low nitrogen permeabil-
vapor.73 Experience suggests that such complicated ity is illustrated in Figure 11, which shows nitrogen
effects tend to occur in alloy composition ranges update as a function of the sum of Ni and Co con-
which fall near the kinetic boundary for continuous, centrations in Fe–Ni–Co alloys and in Figure 12,
POS formation. where reaction rates of internal nitridation are shown
Cr2O3
41Cr–0.2Nb–58.8O–0N
10Cr–25Nb–64O–1N
33Cr–40Nb–3O–24N*
60Cr–28Nb–0O–12N
68Cr–32Nb
(a) 40 µm
Cr2O3
41Cr–0.3Nb–58.6O–0N
17Cr–17Nb–66O–0N (light)
38Cr–2Nb–60O–0N (dark)
14Cr–30Nb–21O–35N*
**43Cr–22Nb–0O–35N (matrix)
**53Cr–24Nb–0O–23N (matrix)
61Cr–29Nb–0O–10N
N decreasing
(b) 40 µm
Figure 10 Cross-sectional micrographs of Cr–31Nb (at.%) oxidized for 24 h at 1100 C. (a) filtered dry air (b) humid air.
Reproduced from Brady, M. P.; Tortorelli, P. F.; Walker, L. R. Oxid. Met. 2002, 58, 297–318.
10 alloys to combine good oxidation resistance with

attractive high temperature mechanical properties at
relatively low alloy cost. Resistance to attack derives
8 from the development and maintenance of a continuous
Nitrogen absorption (mg cm2)
Cr2O3 scale. Consequently, alloy design approaches

that generally promote chromia formation are also a
6 benefit in carburizing environments. Alloying that leads
to the development of an Al2O3 or SiO2 layer or sub-
layer can proffer greater protection under carburization
4 conditions that are too severe for chromia-forming
alloys.77,78 Recently, it was shown that increasing the
nickel content of Fe–Cr–Ni alloys from 30–35 wt% to
2
45% or greater, significantly decreased the carburiza-
tion rate due to decreased carbon solubility and diffu-
0 sivity in the alloy (eqn [2]).78 Furthermore, the
0 20 40 60 80 addition of a few % Al to 60% Ni alloy increased
Ni + Co (wt%)
the carburization resistance substantially due to alu-
Figure 11 Effect of Ni þ Co content in Fe–Ni–Co alloys on mina scale formation at higher temperatures.78
nitridation at 650 C for 168 h (NH3). Data of Barnes and Metal dusting is a catastrophic form of metal degra-
Lai as shown in Lai, G. Y. High Temperature Corrosion of
dation in carbonaceous environments that occurs over
Engineering Alloys; ASM International: Materials Park,
OH, 1990. a lower temperature range (typically 400–800 C),
but accompanied by higher carbon activities than
the carburization phenomenon described above.79,80 It
10
253MA can result in the disintegration of alloy into powder,
1100 ⬚C
153MA
1200 ⬚C and occurs for nickel- as well as iron-based alloys

kp (mg2 cm–4 h–1)
309S (although it is less severe in the former case). As indi-

cated above, alloying or microstructural approaches
5 310S
that promote the exclusive formation of a robust pro-
tective chromia scale improve the resistance to metal
353MA dusting, including the use of cold work or surface
601 deformation to increase outward Cr diffusivity in the
0
0.000 0.200 0.400 0.600 0.800 1.000 alloy.81 Alloys with 26 wt% Cr and a few % of Si or
XNi Al should show improved resistance to metal dusting.77
Figure 12 Parabolic rate constants for nitridation in NH3
It is expected that silica or alumina scales should resist
as a function of Ni concentration in various alloys. Reproduced the uptake of carbon more than chromia scales and,
from Tjokro, K.; Young, D. J. Oxid. Met. 1995, 44, 453–474. therefore, be suitable for metal dusting resistance.
Indeed, any alloy improvements that extend the time
to decrease strongly with nickel concentration of the for localized oxide scale breakdown, and the initia-
alloy. Under these special circumstances, alloying addition of metal dusting at such locations, will also confer
tions of Cr, Al, or Ti will lead to internal precipitation a benefit from any processes that can be used to reoxi-
of their respective nitrides and, therefore, will compro- dize a component during operational shutdowns.82
mise the nitridation resistance of the nickel or cobalt. In Alloying Ni with Cu has been found to improve
particular, the introduction of Cr in Ni–Al or Ni–Ti the resistance to metal dusting.83 Rates of reaction of
base alloys strongly increased the amount of internal Ni–Cu alloys decreased significantly with increasing
attack due to increased nitrogen permeability.76 Cu up to 10 wt% with virtually no metal dusting
at 20% Cu. Because Cu appears to be noncatalytic
Carburization and Metal Dusting. Like many cases towards carbon deposition and has virtually no solubil-
of nitridation, carburization is manifested as an inter- ity for C, its positive effect on dusting of Ni–Cu alloys
nal precipitate formation (of carbides), ahead of the probably relates to a dilution effect.83
alloy–scale interface. Chromia-forming materials
are often used for service where carburization can Sulfidation. Pure sulfidation, that is gas–metal reac-
occur due to the many options available among such tion in an environment completely devoid of oxygen,
poses a challenge for the design of traditional heat- 6

resistant high temperature alloys that normally depend Fe–27%Cr–19%Ni
on the development of Cr2O3, Al2O3, or SiO2 scales84 5
Mass change (mg cm2)

This is because the growth rates of many metal sulfides
4 Fe–18%Cr–12%Al
are quite high (due to their highly defective structures)
and they cannot function as protective scales. On the
3
contrary, refractory metal sulfides based on Mo or
Nb are slow growing,84 but little success in alloying 2
Fe–Al with Mo or Nb has been reported.77 The most
promising system is Fe–Al–Mo, due to the formation 1
of a protective layer of Al0.55Mo2S4.85 Fe3Al (Fe–28%Al)
As mentioned earlier in this section, there are 0
applications in which the environment is highly sul- 0 50 100 150 200 250 300
Time (h)
fidizing, but there is a presence of oxygen ( pO2 of the
order of 1018–1024 atm or perhaps even lower). Figure 13 Mass change of Fe-based alloys in H2–H2S–H2O
Under these circumstances, there are greater oppor- at 800 C ( pO2 1022 and pS2 106 atm). Concentrations are
in at.%. Reproduced from DeVan, J. H. In Oxidation of
tunities for design of more practical alloy systems. Intermetallics; Grobstein, T., Doychak, J., Eds.; TMS:
This is because Al2O3 and SiO2 are thermodynami- Warrendale, PA, 1989; pp 107–115.
cally/kinetically stable at these oxygen partial pres-
sures, while Cr2O3 may or may not be.86 At 700 C
with pS2 108 atm and pO2 10201024 atm, chro- alloy systems, alloying approaches that promote exclu-
mia scales can form on Fe–Cr(–Ni) alloys, at least at sive formation of protective Al2O3 layers in this system
the highest oxygen pressures, with scales formed on result in good sulfidation resistance.91 This is specu-
Fe–Cr (M2O3 type with low Fe levels, M ¼ metallic lated to result, in part, from the unusual mechanism of
element) being significantly more protective that the third-element effect in the Ti–Al–Cr system,
those on Fe–Cr–Ni (M3O4, with higher Fe con- mechanistically based on sub-scale alloying depletion
tents).86 Consequently, minor alloying elements that effects during reaction rather than the more classical
promote exclusive formation of M2O3-type scales gettering type behavior for Fe–Cr–Al.25 There have
(Si, Al) also improved sulfidation resistance.86 been other reports of good sulfidation resistance for
Work over a number of years has consistently coatings based on Ti–Al,92 and a low reaction rate for
shown that alumina-forming materials have signifi- Ti–6Al–4V.93
cantly better resistance to corrosion under highly sul- Alloys that form SiO2 scales also show some
fidizing conditions than chromia-forming alloys,69,87–89 promise for the highly sulfidizing-oxidizing environ-
particularly for iron-based systems. Figure 13 comments described above. Good resistance of MoSi2 to
pares mass change data at 800 C for model Fe–Cr–Ni Ar–H2–H2S has been reported,92 presumably due
(stainless steel) and FeCrAl alloys with that of an to the slow reaction rate of Mo in sulfur84 and the
Fe–28Al (at.%) alloy in a H2S–H2–H2O gas mixture, lack of a stable sulfide of Si. More recently, work
where pO2 was 1022 and pS2 106 atm. The excep- with multiphase Mo–Si–B alloys in the same mixed
tionally low rate of reaction in this very aggressive sulfidizing-oxidizing gas as used for the data in
environment is due to the formation of a protective Figure 13 showed low reaction rates at 800 C, with
Al2O3 scale for the iron aluminide while sulfides the formation of MoSi2 and SiO2 scales.94 Despite the
formed on the other alloys. Additions of Cr to the susceptibility of nickel to reaction with sulfur or SO2,
base Fe–28Al composition increased the rate of reac- silica scales seemed to proffer good sulfidation resis-
tion at a concentration level of 4% and above.69 The tance to iron and nickel silicides.95
presence of REs in alumina-forming alloys was As pS2 is decreased and pO2 increased, SO2 becomes
reported to have a beneficial effect on sulfidation resis- the dominant reactive species. Such environments
tance by improving the scale integrity.90 In contrast, no are typically found in combustion of sulfur-bearing
influence of REs on sulfidation behavior of chromia- fuels. Under these conditions, chromia is often stable,
forming systems, where scale breakdown was not by although its range of stability is usually deter-
mechanical disruption of the oxide scale, was found.90 mined by kinetic factors.96 However, the overall
Recent work with TiAlCr alloys in H2S–H2–H2O corrosion resistance still involves approaches for
environments indicated that, in contrast to Fe–Cr–Al manipulating alloy composition and microstructure
to promote the formation and maintenance of a pro- Table 1 Approximate upper use temperature ( C) for
tective Cr2O3(-rich) scale, as discussed above. Many oxides based on partial pressure of all volatile species
equal to 1
107MPa
studies of sulfidation and POS breakdown under
such oxidizing–sulfidizing conditions have been con- Air: 0.1 MPa 0.1 MPa 1 MPa total
ducted.74,77,96 Alloys that form protective Al2O3 and total 2.1
total 102 0.1 MPa O2
SiO2 scales can, in theory, show good resistance to 102 MPa O2 MPa O2 102 0.1 MPa H2 O
103 MPa H2O MPa H2O
corrosion by SO2. On the other hand, nickel aluminides
were not particularly resistant to corrosion in the pres- Cr2O3 1122 1042 499
ence of SO2,97 although iron aluminides should be.89 If SiO2 1575 1370 967
conditions are such that sulfate salts deposit on alloys, Al2O3 * 1864 1345
chromia scales are most resistant compared to Al2O3
*limited by melting point of oxide rather than volatility.
and SiO2 scales under the so-called type II hot corro- Source: Adapted from Opila, E. J.; Jacobson, N. S.; Myers, D. L.;
sion conditions (lower temperature mode), while alu- Copland, E. H. JOM 2006, 58, 22–28.
mina scales can be preferable for resistance to type-I
attack.98 ferritic and austenitic steels, which are extensively used
in the 500–800 C range (see other chapters in this
Water vapor. The effects of water vapor (H2O) on book).102–110 However, proper alloying can improve
high temperature oxidation have received increas- oxidation resistance of these chromia-forming alloys
ingly greater attention. In large part, this is because in the presence of water vapor. This is shown schema-
most power-generation applications (gas turbines, tically for Cr and Ni in Figure 14, which indicates the
combustion and steam-generation systems, solid compositional regions where behavior associated with
oxide fuel cells, etc.) involve environments with rela- development of a POS consisting of Cr-rich (Cr,Fe)2O3
tively high levels of H2O. As such, there is a critical was observed at 650 and 800 C.106 As expected nor-
need for materials that are relatively nonreactive with mally (cf. eqn [2]), increasing Cr has a beneficial
water vapor under the aggressive conditions imposed effect on promoting POS formation in water vapor-
by the increased operating temperatures and pres- containing environments. At levels above 16 wt% Cr,
sures needed to achieve high power-generation increasing the Ni level in the alloy also provides some
efficiencies and decreased emissions. As described improvement in oxidation resistance in the presence of
elsewhere in this book, the presence of water vapor water vapor.106 While increasing the Cr levels above
in the environment typically leads to higher oxida- those shown in Figure 14 may improve the oxidation
tion rates, particularly for chromia- or silica-forming resistance even more, the onset of brittle sigma phase
alloys where H2O can dramatically enhance oxide formation in the alloy and its detrimental effect on the
volatilization and/or impact oxide structure/stability mechanical properties must be considered.109 As dis-
and solid-state transport properties.99,100 In particular, cussed in the previous sections, a finer alloy grain size
water vapor results in the formation of volatile oxy- can help promote POS formation on an alloy.17 This has
hydroxides, which can significantly compromise the been observed in water vapor-containing environments
formation of POS by constant thinning of the oxide for chromia-forming alloys.111
surface layer due to volatilization.99 Because chromia Minor individual alloying additions of Mn (0.1–
has the highest volatilization susceptibility of the 3.5 wt%), Si (0.2%), Ti, and B to form TiB2 ( 1%),
major high temperature protective oxides,100 Mn is and Nb and C to form NbC ( 1%) in model Fe–
added to chromia-forming steels to decrease the rate 16 wt% Cr base alloys have been reported to have a
of volatilization substantially by promoting the devel- beneficial effect on oxidation resistance in air þ10%
opment of (Cr,Mn)3O4 surface layers.101 In contrast to H2O.111 The effects of Ti and Nb may result, in part,
Cr2O3 and SiO2 scales, Al2O3 scales are far more stable by decreasing the alloy grain size by formation of
in H2O-containing environments (see Table 1)100 and, TiB2 and NbC, which retard the grain growth during
as discussed below, are generally less susceptible to alloy processing. No beneficial effect of RE-type
accelerated oxidation in the presence of water vapor. additions (0.3% La, 0.1% Y, 0.03% Zr) on oxidation
Accelerated oxide scale growth rates, decreased resistance in water vapor was noted.111 As mentioned
time-to-breakaway (associated with formation of iron previously, Mn additions to Fe–Cr alloys in sufficient
oxide nodules), and increased spallation susceptibility concentrations to promote a Cr–Mn spinel layer can
under thermal-cycling conditions have been widely be beneficial by decreasing oxide volatilization.101
reported for water vapor effects on chromia-forming Additions of Mn and Si to Fe–20Cr–20Ni (wt%)
POS AA AAS
Ni Ni
650 ⬚C 800 ⬚C
0 10 15 20 25 30 0 10 15 20 25 30
10 F 10 F
12 F 12 F
14 F 14 F
Cr Cr
16 F F+A A A A A 16 F F+A A A A A
Figure 14 Maps of three behaviors (POS – protective oxide scale, AA – accelerated attack, AAS – accelerated attack
with spallation) as a function of Ni and Cr concentrations (wt%) in Fe-based alloys (F – ferritic, A-austenitic). Based on
observations for 100 1-h thermal cycles in air þ 10 vol. %H2O at indicated temperatures. Reproduced from Peraldi, R.;
Pint, B. A. Oxid. Met. 2004, 61, 463–483.
40 0.8
Fe9Cr Fe9Cr 304L/Al
Specimen mass change (mg cm–2)
20 Specimen mass change (mg cm–2) 0.6

Fe–15Al + Hf 304L/Al
0 0.4
Fe–20Al + Hf Fe9Cr/Al Fe–15Al + Hf
−20 0.2
Fe9Cr/Al
−40 0 Fe–20Al + Hf
−60 −0.2
304L 304L
−80 −0.4
0 100 200 300 400 500 0 100 200 300 400 500
(a) Number of 1-h cycles at 800 ⬚C (b) Number of 1-h cycles at 800 ⬚C
Figure 15 Mass changes for 1-h thermal cycles at 800 C in air þ 10 vol. % H2O. Selected data from (a) are shown in (b) over
a smaller ordinate range. ‘Fe9Cr/Al’ and ‘304/Al’ denote data for aluminized Fe–9Cr–1Mo steel and type 304L stainless steel,
respectively. Concentrations are in wt%. Reproduced from Zhang, Y.; Pint, B. A.; Haynes, J. A.; Tortorelli, P. F. Oxid. Met.
2004 62, 103–120.
have been reported to have a positive effect on oxi- Kvernes et al. found that additions of Al to a
dation in airþ10% H2O, with the best performance Fe–13Cr alloy eliminated the detrimental influence
observed when both elements were present.111 Henry of water vapor in the oxidizing environment (studied
et al. have also indicated a positive effect of Si on up to 2.3% H2O) if present in sufficient concentra-
oxidation resistance in water vapor.105 However, recent tion to form a continuous protective layer of Al2O3
work of Hammer et al. indicates an opposite effect for (1% Al was needed at 680 C and 4 at.% 980 C).102
Si because of the presence of SiO2 at the oxide–metal While there are some reported detrimental effects of
interface,109 a reason suggested by Pint et al. for its H2O on Al2O3-forming alloys,114 these are relatively
beneficial effect.111 These observations illustrate the minor compared to those reported for chromia- and
need to balance the effect of Si as an alloying addition silica-forming alloys, and protective alumina scales
to chromia-forming alloys (for water vapor-containing provide substantially improved resistance for alloys
environments as well for oxidizing conditions, in gen- in water vapor-containing environments.115,116 For
eral). The formation of a silica layer beneath the example, Figure 15 shows the mass changes of an
chromia scales can decrease the overall scale growth Fe–Al alloy and ferritic, and austenitic steels (Fe–9Cr
rate relative to that of chromia, but only if the Si and type 304L stainless steel) with and without an
concentration is insufficient to form a brittle continu- aluminized coating in air þ 10% H2O under cyclic
ous subscale that promotes scale spallation.112,113 oxidation exposures at 800 C. Both the Fe–Al alloy
and the aluminized specimens (which developed of alloys used at high temperatures, Cr2O3, Al2O3,
Fe–Al-based surface compositions) showed excellent and SiO2, with emphasis on the first two. At its
oxidation resistance despite the presence of H2O, essence, alloy design for high temperature corrosion
whereas the uncoated chromia-forming steels oxi- resistance relates to the tailoring of composition and
dized rapidly. microstructure to promote the preferential formation
Given the attractiveness of alumina-forming alloys of the desired POS by selective oxidation, and its
for water vapor (and other) applications, the alloy subsequent stability and integrity, while balancing
design problem, therefore, becomes one of assuring the effects on other properties. The attractiveness of
the development and maintenance of protective alumina as a protective scale for higher temperatures
Al2O3 scales and devising routes to achieving adequate as well as in a variety of aggressive multireactant
strength and ductility, particularly under creep con- environments (e.g., water vapor, sulfur) is evident,
ditions, at higher temperatures and H2O pressures but widespread applications of appropriate composi-
envisioned for advanced uses of such materials. In this tions await further alloy and/or coating development
system, increasing the Al content (that is, increasing that must optimize high temperature corrosion resis-
NB) is limited because aluminum stabilizes the body- tance, desired mechanical performance, and/or phys-
centered cubic form of Fe, which inherently exhibits ical properties.
poor high temperature creep strength. (Austenitic Ni- The goal of the alloy design or materials selection
base alumina-forming alloys are available, but their processes should, in most cases, be to maximize the
high cost limits their use to a small subset of higher overall oxidation lifetime by accounting for all factors
performance applications.) To this end, there has been (thermodynamic, kinetic, mechanical, physical, met-
some recent progress in the design of creep-resistant, allurgical, etc.) that influence it. Looking into the
alumina-forming austenitic stainless steels based on future, it is hoped that understanding and control of
Fe–(20–25)Ni–(12–15)Cr–(2–4)Al (wt%) with fine these factors can be developed to the point that
dispersions of MC (M = Nb, V, Ti) and intermetallic improved lifetime prediction will enable quantification
B2–NiAl and Fe2Nb Laves precipitates.117–119 At of the effectiveness of various alloying schemes for
600–800 C, Nb was found to promote Al2O3 layer given application environments and component
development, while both Ti and V had the reverse designs. To this end, it is anticipated that the ability
effect for the range of compositions with good creep to describe adequately and manipulate reaction prod-
resistance.118 In order to maintain a single-phase aus- uct development and kinetic pathways for complex,
tenitic matrix microstructure, these compositions con- multiphase alloys and material systems in multireac-
tain lean levels of Al and Cr (via a third-element effect) tant, high temperature environments will provide the
for protective Al2O3 scale formation. Therefore, under foundation for approaches to be successful and effi-
certain conditions, enhanced internal oxidation and a cient in alloy design, and selection for a variety of
transition to nonprotective Fe-based oxide nodule for- future applications.
mation were observed in water vapor-containing envir- This research was sponsored by the U.S. Depart-
onments compared to exposures in dry air.119 ment of Energy under Contract No. DE-AC05–
Enhanced internal oxidation of chromia-forming alloys 00OR22725. By accepting the article for publication,
in the presence of water vapor has also been the publisher acknowledges that the United States
reported.110 This effect has been speculated to be the Government retains a nonexclusive, paid-up, irrevo-
result of hydrogen ingress into the alloy from decom- cable, worldwide license to publish or reproduce the
position of the water vapor, possibly shifting the bal- published form of this work, or allow others to do so,
ance of internal/external oxidation by increasing the for United States Government purposes.
oxygen solubility (cf. eqns [2] and [3]).110
References
1.21.7 Conclusions
1. Stott, F. H.; Wood, G. C.; Stringer, J. Oxid. Met. 1995, 44,
Given the vast number of possible alloy and environ- 113–145.
mental combinations, this chapter has focused on 2. Brady, M. P.; Tortorelli, P. F.; More, K. L.; Payzant, E. A.;
general principles and illustrative (but far from Armstrong, B. L.; Lin, H. T.; Lance, M. J.; Huang, F.;
Weaver, M. L. Mater. Corros. 2005, 56, 748–755.
exhaustive) examples related to the formation of the 3. Schütze, M. Protective Oxide Scales and their
most protective scales needed for corrosion resistance Breakdown; John Wiley and Sons: Chichester, 1997.
4. Smialek, J. JOM 2000, 52, 22–25. 42. Quadakkers, W. J.; Malkow, T.; Phon-Abellan, J.;
5. Stott, F. H.; Atkinson, A. Mater. High Temp. 1994, 12, Flesch, U.; Shemet, V.; Singheiser, L.; McEvoy, A. J., Ed.;
195–207. In Proceedings of the 4th European Solid Oxide Fuel cell
6. Wagner, C. Z. Elektrochem. 1959, 63, 772–790. Forum; Elsevier: Amsterdam, 2000; Vol. 2,
7. Whittle, D. P. In High Temperature Corrosion; Rapp, R. A., pp 827–836.
Ed.; National Association of Corrosion Engineers: 43. Pint, B. A.; More, K. L.; Wright, I. G. Mater. High Temp.
Houston, TX, 1981; pp 171–183. 2003, 20, 375–386.
8. Stott, F. H. Rep. Prog. Phys. 1987, 50, 861–913. 44. Gupta, D. K.; Duvall, D. S. In Superalloys; Gell, M., et al.
9. Gesmundo, F.; Gleeson, B. Oxid. Met. 1995, 44, Eds.; TMS: Warrendale, PA, 1984; pp 711–720.
211–238. 45. Kochubey, V.; Al-Badairy, H.; Tatlock, G.; Le-Coze, J.;
10. Gesmundo, F.; Castello, P.; Viani, F. Oxid. Met. 1996, 46, Naumenko, D.; Quadakkers, W. J. Mater. Corros. 2005,
383–398. 56, 848–853.
11. Gesmundo, F.; Niu, Y. Oxid. Met. 1998, 50, 1–26. 46. Funkenbush, A. W.; Smeggil, J. G.; Bornstein, N. S. Met.
12. Brady, M. P.; Gleeson, B.; Wright, I. G. JOM 2000, 52, Trans. 1985, 16A, 1164–1166.
16–21. 47. Sigler, D. R. Metals 1989, 32, 337–355.
13. Brady, M. P.; Tortorelli, P. F. Intermetallics 2004, 12, 48. Smialek, J. L.; Jayne, D. T.; Schaeffer, J. C.; Murphy,
779–789. W. H. Thin Solid Films 1994, 253, 285–292.
14. Rapp, R. A. Corrision 1965, 21, 382–401. 49. Wright, I. G.; Pint, B. A. In Proceedings of SF2M;
15. Stott, F. H. Mater. Charact. 1992, 28, 311–325. Journees d’automne, Soc. Francaise de Metallurgie et de
16. Tortorelli, P. F.; DeVan, J. H.; McKamey, C. G.; Materiaux: Paris, 2000; pp 86–87.
Howell, M. Ceram. Trans. 1991, 19, 961–968. 50. Pint, B. A.; More, K. L.; Wright, I. G. Oxid. Mat. 2003, 59,
17. Yurek, G. J.; Eisen, D.; Garratt-Reed, A. Metall. Trans. A 257–283.
1982, 13, 473–485. 51. Pint, B. A.; Wright, I. G.; Lee, W. Y.; Zhang, Y.; Prüßner, K.;
18. Caplan, D. Corros. Sci. 1966, 6, 509–515. Alexander, K. B. Mater. Sci. Eng. 1998, A245, 201–211.
19. Wright, I. G. Oxidation of Iron-, Nickel-, and Cobalt-Base 52. Wright, I. G.; Pint, B. A.; Lee, W. Y.; Alexander, K. B.;
Alloys, Metals and Ceramics Information Center Report Prüßner, K. In High Temperature Surface Engineering;
MCIC-72–07 (Columbus, OH, Battelle Columbus Nicholls, J., Rickergy, D., Eds.; Institute of Materials:
Laboratories, 1972). London, 2000; pp 95–113.
20. Wagner, C. Corros. Sci 1965, 5, 751–764. 53. Haynes, J. A.; More, K. L.; Pint, B. A.; Wright, I. G.;
21. Renusch, D.; Veal, B.; Natesan, K.; Grimsditch, M. Oxid. Cooley, K.; Zhang, Y. Mater. Sci. Forum 2001, 369–372,
Met. 1996, 46, 365–381. 679–686.
22. Niu, Y.; Wang, S.; Gao, F.; Zhang, Z. G.; Gesmundo, F. 54. Cadoret, Y.; Bacos, M. P.; Josso, P.; Maurice, V.;
Corros. Sci. 2008, 50, 345–356. Marcus, P.; Zanna, S. Mater. Sci. Forum 2004, 461–464,
23. Guan, S. W.; Smelzer, W. W. Oxid. Met. 1994, 42, 247–254.
375–391. 55. Schütze, M. Quadakkers, W. J.; Nicholls, J. R. Eds.
24. Wang, Ge; Gleeson, B.; Douglass, D. L. Oxid. Met. 1991, Lifetime Modeling of High Temperature Corrosion
35, 333–348. Processes; Maney Publishing: London, 2001.
25. Gesmundo, F.; Viani, F.; Niu, Y. Oxid. Met. 1994, 42, 56. Bennett, M. J.; DeVan, J. H.; Tortorelli, P. F. The
409–429. Oxidation Behavior of Iron Aluminides in Air at 1300 C
26. Gesmundo, F.; Niu, Y.; Douglass, D. L. Oxid. Met. 1994, In Microscopy of Oxidation 3; Newcomb, S. B., Little, J. A.,
42, 465–484. Eds.; The Institute of Materials: London, 1997; pp 233–245.
27. Gesmundo, F.; Viani, F.; Niu, Y. Oxid. Met. 1996, 45, 51–76. 57. Quadakkers, W. J.; Bongartz, K. Mater. Corros. 1994, 45,
28. Carter, P.; Gleeson, B.; Young, D. Acta Mater. 1996, 44, 232–241.
4033–4038. 58. Guruppa, I.; Weinbruch, S.; Naumenko, D.;
29. Durham, R. N.; Gleeson, B.; Young, D. J. Oxid. Met. Quadakkers, W. J. Mater. Corros. 2000, 51, 101–108.
1998, 50, 139–165. 59. Huczkowski, P.; Shemet, V.; Piron-Abellan, J.;
30. Perkins, R. A.; Meier, G. H. In Proceedings of the Singheiser, L.; Quadakkers, W. J.; Christiansen, N.
Industry-University Advanced Materials Conference II; Mater. Corros. 2004, 55, 825–830.
Smith, F. W., Ed.; Advanced Materials Institute: Denver, 60. Pint, B. A.; Porter, W. D.; Wright, I. G. Mater. Sci. Forum
1989; pp 92–99. 2008, 595–598, 1083–1092.
31. Brady, M. P.; Smialek, J. L.; Terepka, F. Scripta Met. 61. Evans, H. E.; Donaldson, A. T.; Gilmour, T. C. Oxid. Met.
1995, 32, 1659–1664. 1999, 52, 379–402.
32. Brady, M. P.; Smialek, J. L.; Smith, J.; Humphrey, D. L. 62. Newton, R.; et al. In Modeling of High Temperature
Acta Mater. 1997, 45, 2357–2369. Corrosion Processes; Schütze, M., Quadakkers, W. J.,
33. Gleeson, B.; Mu, N.; Hayashi, S. J. Mater. Sci. 2009, 44, Nicholls, J. R., Eds.; Maney Publishing: London, 2001;
1704–1710. pp 15–36.
34. Whittle, D. P.; Stringer, J. Philos. Trans. Royal Soc. 63. Nicholls, J. R.; Newton, R.; Bennett, M. J.; Evans, H. E.;
London 1980, 295(A), 309–329. Al-Badairy, H.; Tatlock, G. J.; Naumenko, D.;
35. Pint, B. A. Oxid. Met. 1996, 45, 1–37. Quadakkers, W. J.; Strehl, G.; Borchadt, G. In Modeling
36. Quadakkers, J.; Singheiser, L. Mater. Sci. Forum 2001, of High Temperature Corrosion Processes; Schütze, M.,
369–372, 77–92. Quadakkers, W. J., Nicholls, J. R., Eds.; Maney
37. Pint, B. A. In Proceedings of John Stringer Symposium; Publishing: London, 2001; pp 83–106.
Tortorelli, P. F., Hou, P. Y., Wright, I. G., Eds.; ASM 64. Evans, H. E. Inter. Mater. Rev. 1995, 40, 1–40.
International: Materials Park, OH, 2003; pp 9–19. 65. Nesbitt, J. A. Oxid. Met. 1995, 44, 309–338.
38. Pint, B. A. J. Am. Ceram. Soc. 2003, 86, 686–695. 66. Haynes, J. A.; Zhang, Y.; Cooley, K. M.; Walker, L.;
39. Sigler, D. R. Oxid. Met. 1993, 40, 555–583. Reeves, K. S.; Pint, B. A. Surf. Coat. Technol. 2004,
40. Warnes, B. M. Surf. Coat. Technol. 2001, 146–147, 7–12. 188–189, 153–157.
41. Naveos, S.; Oberlaender, G.; Cadoret, Y.; Josso, P.; 67. Giggins, C. S.; Pettit, F. S. Oxid. Met. 1980, 14, 363–413.
Bacos, M. P. Mater. Sci. Forum 2004, 461–464, 375–382. 68. Young, D. J.; Watson, S. Oxid. Met. 1995, 44, 239–264.
69. DeVan, J. H. In Oxidation of Intermetallics; Grobstein, T., Environments, Oak Ridge National Laboratory report,
Doychak, J., Eds.; TMS: Warrendale, PA, 1989; pp 107–115. ORNL/Sub-83–43346/02, 1989.
70. Pint, B. A.; Dwyer, M. J.; Deacon, R. M. Oxid. Met. 2008, 96. Baxter, D. J.; Natesan, K. Oxid. Met. 1989, 31, 305–313.
211–231. 97. Natesan, K. Oxid. Met. 1988, 30, 53–83.
71. Dettenwanger, F.; Schumann, E.; Rühle, M.; Rakowski, 98. Nicholls, J. R. JOM 2000, 52, 28–35.
J. M.; Meier, G. H. Oxid. Met. 1998, 50, 269–307. 99. Opila, E. J. J. Am. Ceram. Soc. 1994, 77, 730–736.
72. Shemet, V.; Tyagi, A. K.; Becker, J. S.; Lersch, P.; 100. Opila, E. J.; Jacobson, N. S.; Myers, D. L.; Copland, E. H.
Singheiser, L.; Quadakkers, W. J. Oxid. Met. 2000, 54, JOM 2006, 58, 22–28.
211–235. 101. Stanislowski, M.; Wessel, E.; Hilpert, K.; Markus, T.;
73. Brady, M. P.; Tortorelli, P. F.; Walker, L. R. Oxid. Met. Singheiser, L. J. Electrochem. Soc. 2007, 154,
2002, 58, 297–318. A295–A306.
74. Lai, G. Y. High Temperature Corrosion of Engineering 102. Kvernes, I.; Oleveira, M.; Kofstad, P. Corros. Sci. 1977,
Alloys; ASM International: Materials Park, OH, 1990. 17, 237–252.
75. Tjokro, K.; Young, D. J. Oxid. Met. 1995, 44, 453–474. 103. Asteman, H.; Svensson, J.-E.; Norell, M.; Johansson, L.-G.
76. Krupp, U.; Christ, H.-J. Oxid. Met. 1999, 52, 277–298. Oxid. Met. 2000, 54, 11–26.
77. Gleeson, B. In Materials Science and Technology: 104. Rakowski, J. M.; Pint, B. A. NACE Paper 00–517,
Corrosion and Environmental Degradation; Shütze, M., Proceedings of Corrosion 2000; NACE International:
Ed.; Wiley-VCH: Weinheim, 2000; Vol. 2, Chapter 5. Houston TX, 2000.
78. Becker, P.; Young, D. J. Oxid. Met. 2007, 67, 267–277. 105. Henry, S.; Galerie, A.; Antoni, L. Mater. Sci. Forum 2001,
79. Hochman, R. F. In Properties of High Temperature Alloys; 369–372, 353–360.
Foroulis, Z. A., Pettit, F. S., Eds.; The Electrochemical 106. Peraldi, R.; Pint, B. A. Oxid. Met. 2004, 61, 463–483.
Society: Pennington, NJ, 1976; pp 715–732. 107. Pint, B. A. Paper GT2005–68495, Proceedings of
80. Nava Paz, J. C.; Grabke, H. J. Oxid. Met. 1993, 39, ASME Turbo Expo 2005; ASME International: New York,
437–456. 2005.
81. Grabke, H. J.; Müller-Lorenz, E. M.; Strauss, S.; 108. Ehlers, J.; Young, D. J.; Smaardijk, E. J.; Tyagi, A. K.;
Pippel, E.; Woltersdorf, J. Oxid. Met. 1998, 50, 241–254. Penkalla, H. J.; Singheiser, L.; Quadakkers, W. J. Corros.
82. Zeng, Z.; Natesan, K. Oxid. Met. 2006, 66, 1–20. Sci. 2006, 3428–3454.
83. Zhang, J.; Cole, D. M. I.; Young, D. J. Mater. Corros. 109. Hammer, J. E.; Laney, S. J.; Jackson, R. W.; Coyne, K.;
2005, 56, 756–764. Pettit, F. S.; Meier, G. H. Oxid. Met. 2007, 67, 1–38.
84. Mrowec, S. Oxid. Met. 1995, 44, 177–209. 110. Essuman, E.; Meier, G. H.; Zurek, J.; Hänsel, M.;
85. Gleeson, B.; Douglass, D. L.; Gesmundo, F. Oxid. Met. Quadakkers, W. J. Oxid. Met. 2008, 69, 143–162.
1990, 34, 123–150. 111. Pint, B. A.; Peraldi, R.; Maziasz, P. J. Mater. Sci. Forum
86. Wright, I. G.; Srinivasan, V.; Vedula, K. M. Mater. High 2004, 461–464, 815–822.
Temp. 1993, 11, 159–166. 112. Bennett, M. J.; Desport, J. A.; Labun, P. A. Oxid. Met.
87. Green, S. W.; Stott, F. H. Oxid. Met. 1991, 36, 239–252. 1984, 22, 291–306.
88. Stott, F. H.; Chuah, K. T.; Bradley, L. B. Mater. Corros. 113. Hoelzer, D. T.; Pint, B. A.; Wright, I. G. J. Nucl. Mater.
1996, 47, 695–700. 2000, 283–287, 1306–1310.
89. Tortorelli, P. F.; Natesan, K. Mater. Sci. Eng. A 1998, 258, 114. Maris-Sida, M. C.; Meier, G. H.; Pettit, F. S. Metall. Mater.
115–125. Trans. A 2003, 34A, 2609–2619.
90. Stott, F. H.; Chong, F. M. F.; Stirling, C. A. Mater. Sci. 115. Zhang, Y.; Pint, B. A.; Haynes, J. A.; Tortorelli, P. F. Oxid.
Forum 1989, 43, 327–362. Met. 2004, 62, 103–120.
91. Brady, M. P.; Tortorelli, P. F.; More, K. L.; Walker, L. R. 116. Pint, B. A.; Shingledecker, J. P.; Brady, M. P.; Maziasz, P. J.
Oak Ridge National Laboratory, Oxid. Met. 2009 paper GT2007–27916, Proceedings of the ASME Turbo
(in press). Expo 2007l; ASME International: New York, 2007.
92. Schütze, M.; Nöth, M. NACE Paper 517, Proceedings of 117. Yamamoto, Y.; Brady, M. P.; Lu, Z. P.; Maziasz, P. J.;
Corrosion/98; NACE International: Houston, TX, 1998. Liu, C. T.; Pint, B. A.; More, K. L.; Meyer, H. M.;
93. Du, H. L.; Datta, P. K.; Lewis, D. B.; Burnell-Gray, J. S. Payzant, E. A. Science 2007, 316, 433–436.
Oxid. Met. 1996, 45, 507–527. 118. Brady, M. P.; Yamamoto, Y.; Santella, M. L.; Pint, B. A.
94. Tortorelli, P. F.; Schneibel, J. H.; More, K. L.; Pint, B. A. Scripta Mater. 2007, 57, 1117–1120.
Mater. Sci. Forum 2004, 461–464, 1063–1070. 119. Brady, M. P.; Yamamoto, Y.; Santella, M. L.; Maziasz, P. J.;
95. Kim, G. M.; Meier, G. H. Breakdown Mechanisms of Pint, B. A.; Liu, C. T.; Lu, Z. P.; Bei, H. JOM 2008, 60,
Al2O3, Cr2O3, and SiO2 Scales in H2/H2O/H2S 12–18.
1.22 High Temperature Corrosion of Low Alloy Steels
L. W. Pinder
E.ON Engineering Limited Technology Centre, Ratcliffe on Soar, Nottingham, NG11 OEE
K. Dawson and G. J. Tatlock
Department of Engineering, University of Liverpool, Liverpool, L69 3GH
This article is a revision of the Third Edition article 7.2 by L. W. Pinder, volume 1, pp 7:16–7:52, ß 2010 Elsevier B.V.

1.22.2 Factors Governing Oxidation Behavior 559
1.22.3 The Oxidation of Iron 560
1.22.4 General Alloying Effects on Oxidation 562
1.22.5 Effects of Specific Alloying Elements on the Oxidation of Iron 564
1.22.5.1 Carbon 564
1.22.5.2 Aluminum 565
1.22.5.3 Silicon 566
1.22.5.4 Manganese 567
1.22.5.5 Sulfur 567
1.22.5.6 Phosphorus 567
1.22.5.7 Nickel 567
1.22.5.8 Chromium 567
1.22.5.9 Molybdenum 568
1.22.5.10 Boron 568
1.22.5.11 Others 569
1.22.5.12 Effects of Heat Flux and Cold Work 569
1.22.6 Stress Effects 570
1.22.6.1 Growth Stresses 570
1.22.6.2 System-Applied Stresses 571
1.22.6.3 Thermal Stresses 571
1.22.7 Commercial Low-Alloy Steels in Air or Oxygen 572
1.22.8 Industrial Environments 572
1.22.8.1 Steam 572
1.22.8.2 Combustion Gases 575
1.22.8.3 Chemical Environments 576
1.22.8.4 CO/CO2 578
References 579
1.22.1 Introduction which steels may be exposed. Novel manufacturing

techniques such as mechanically alloyed steels pro-
Low alloy steels are generally considered to comprise duced via powder metallurgy techniques are also
plain carbon steels and steels with a total alloying offering new groups of steels designed for utilization
content of up to 12 wt%. This group of steels is in high temperature applications.
exploited for many applications, including those in The drive for increased efficiency in power gen-
the heavy engineering industries. While these mate- eration requires higher operating temperatures. Low
rials are often selected based on mechanical proper- alloy ferritic steels with attractive mechanical prop-
ties, extensive use in high temperature applications erties, including low coefficients of thermal expan-
often involving harsh environments including power sion, have been optimized for high temperature
generation and the petrochemical industries demands performance. Since they possess adequate creep
a degree of oxidation resistance. Emerging industries strength for their designed capability, the service
creating energy from waste and biofuels have intro- life of these Fe–Cr alloys is now often limited by
duced more varied and complex environments to their steam-side oxidation resistance.
558
High Temperature Corrosion of Low Alloy Steels 559
Alloying additions of Si, Al, and Cr which form protection to the substrate.1 In other cases, a solid
stable, slow growing, protective scales on steels are oxide scale forms that may, or may not, be continu-
extensively used; however, these additions can be finan- ous. The extent to which a solid surface scale protects
cially inhibitive and/or detrimental to mechanical the metal depends upon the extent of surface cover-
properties. In terms of low alloy steels in general, the age. This behavior can largely be categorized accord-
restricted quantities (12 wt%) of alloying additions may ing to whether the volume of oxide produced is less
not enable the most protective form of oxide to grow. or greater than the volume of metal consumed during
The extent to which low alloy steels react to high the oxidation reaction. This principle was originally
temperature corrosive environments is the subject of advanced by Pilling and Bedworth,2 with the oxide:
this chapter. In view of the commercial importance of metal volume ratio being known as the Pilling
these steels, the published literature on this topic is Bedworth Ratio (PBR), where
extensive and being continually enlarged. The reader Volume of oxide produced
is encouraged to refer to many excellent papers and PBR ¼
Volume of metal consumed
current issues of journals, referenced at the end of the
chapter, for more detailed and contemporary infor- If the PBR is less than unity, the oxide will be non-
mation on the topic. protective and oxidation will follow a linear rate law,
governed by surface reaction kinetics. However, if the
PBR is greater than unity, then a protective oxide
1.22.2 Factors Governing scale may form and oxidation will follow a reaction
Oxidation Behavior rate law governed by the speed of transport of metal or
environmental species through the scale. Then the
In the absence of stress, the survival of a given com- degree of conversion of metal to oxide will be depen-
ponent is largely determined by the extent to which dent upon the time for which the reaction is allowed to
the material of construction reacts with the envi- proceed. For a diffusion-controlled process, integration
ronment in which it resides. This can largely be of Fick’s First Law of Diffusion with respect to time
categorized by the extent to which the following yields the classic Tammann relationship,3 commonly
two criteria are satisfied: referred to as the Parabolic Rate Law:
1. Is the material thermodynamically stable in the x 2 ¼ Kt ½1
environment?
2. If not, will the reaction rate between the environ- where x is a measure parameter, K is the rate constant,
ment and the material be slow enough to give an and t is time. The progress of oxidation (x) is generally
acceptable life? measured by weight gain, weight loss, scale thickness,
or retreat of the metal surface. However, other para-
In the vast majority of cases at elevated temperatures, meters, such as loss of oxidant in the environment, may
the answer to the first question is ‘no.’ With the be used. The time interval may be measured in sec-
exception of gold and platinum, which are generally onds, hours, thousands of hours and years. As a conse-
too expensive for large-scale industrial use, are in short quence, many different units of the measured oxidation
supply, and do not have the required mechanical proper- parameter and the time period have been reported in
ties, most materials will react to some degree with their the literature, leading to many differing units for the
environment. Most highly alloyed steels and superalloys quoted rate constant.
satisfy the second criterion for a wide range of high In some circumstances, the reaction rates may not
temperature environments. However, these materials be exactly parabolic, and even initially parabolic rates
may be far too expensive for large-scale industrial appli- may be influenced by changes within the oxide scale
cations. The oxidation behavior of most metals and with time. As an oxide scale grows, the build-up of
alloys in a high temperature environment is, therefore, inherent growth stresses, externally applied strains,
governed by the degree of protection afforded by any and chemical changes to either the oxide scale or
oxide scale that forms.1 This depends upon the oxide the metal may all compromise the initial protection
melting point, its mechanical integrity, and the rate of offered by the scale, leading to scale breakdown
diffusion of elemental species present within both the and ultimately to a partial or complete loss of protec-
environment and the alloy, through the oxide scale. tion; paralinear or linear kinetics may ensue. In other
In certain cases (e.g., Cr above 1000 C), the CrO3 circumstances, as will be seen later in this chapter,
oxide that forms is volatile, and clearly affords no very small additions of contaminants to the
environment may radically modify the oxidation

response, by either favoring nonprotective scales
Key
from the outset or causing rapid breakaway corrosion Mild steel
10−8
following an initially protective period. Hence, there 1/
2% Cr 1/2% Mo 1/4% V
are many hidden dangers in extrapolating short-term 1 1/4% Cr 1/2% Mo 1/4% V
oxidation kinetic data to long periods, and great care 2 1/4% Cr 1% Mo
10−9
Parabolic rate constant (cm2 s−1)

5% Cr 1% Mo
is required when utilizing published oxidation data to
assess potential corrosion rates and component lives in
industrial environments.
The temperature dependence of a diffusion- 10−10
controlled reaction is typically described by the

Arrhenius relationship:
10−11
ðQ Þ
K ¼ a exp ½2
RT
where a is a constant, Q is the activation energy, R is
10−12
the gas constant, and T is the temperature (K). Hence
the rate constant (K ) will be influenced by the tem- 850 800 750 700 650 600 550 500
perature of exposure because of the increased ion
T ( ˚C)
mobility at higher temperatures, via the activation
9 10 11 12 13
energy term. Again, great care is required when 104/T (K)
extrapolating data beyond the temperature range of
the initial experimental data. Sharp changes in the Figure 1 Arrhenius plot for the oxidation of mild steel
and low-alloy steels in air showing a sharp break in the slope
Arrhenius slope may occur with the appearance of and curvature due to the appearance of FeO in the scale
a new, thermodynamically favored, species within the above 570 C.
scale. In addition, an Arrhenius plot of the tempera-
ture dependence of the reaction rate may reveal cur-
vature because of the progressive change from one
rate determining process to another. Both the features
are observed in the Arrhenius plot for low-alloy steels 1600
(Figure 1). Fe3O4
1400
1.22.3 The Oxidation of Iron α − Fe
+ FeO
For iron in most oxidizing environments, the PBR is 1200

approximately 2.2 and the scale formed is protective.
FeO
The oxidation reaction forms a compact, adherent 1000
⬚C
scale, the inner and outer surfaces of which are in FeO + Fe3O4
thermodynamic equilibrium with the metal substrate

and the environment respectively, and ion mobility 800
α − Fe
through the scale is diffusion controlled. + FeO
The rate of oxidation of iron is then governed by 600
the stabilities of the various phases, which are in turn
a function of the temperature and oxygen partial
400 α − Fe + Fe3O4
pressure of the environment. Examination of the
Fe–O2 phase diagram (Figure 2) reveals that the prin- Fe3O4 + Fe2O3
cipal solid oxide phases below 570 C will be Fe3O4 200

10 20 30 40 50 60 70 80 90
(magnetite) and Fe2O3 (haematite). Above 570 C, FeO FeO FeO Fe2O3 Fe2O3
(wüstite) appears as a third phase within the scale. Weight %
These phases are present within the scale as individual Figure 2 Fe–O phase diagram showing the principal solid
layers, with the layer sequence dictated by the equilib- phases stable on iron and wide stoichiometry limits of FeO.
rium oxygen partial pressure ( pO2) for phase stability Adapted from White, J. CEGB Private Communication.
prevailing at the given temperature. Hence, the oxide that demonstrate a high degree of nonstoichiometry;
phase stable at the lowest pO2 (FeO at >570 C) is with vacancy–hole complexes possibly being pre-
found closest to the metal substrate, whereas the sent at the higher deviations.11 Single defects occur
phase stable at the highest pO2 (Fe2O3) is found closest only when these oxides are closest to their stoichiomet-
to the oxidizing environment. If, however, the pO2 of ric composition. These considerations lead to a com-
the environment is low enough, only FeO will be plex dependence of the cation diffusion coefficient on
formed. At intermediate values of the pO2, Fe3O4 will the oxide stoichiometry,12 and hence on temperature.
also form, and for most industrial environments, the Mass transport measurements have shown that
pO2 is sufficient for the formation of an outer layer of cation transport predominates in FeO (Fe2þ) and
Fe2O3. While these observations are true for bulk Fe3O4 (Fe2þ, Fe3þ), whereas anion transport predo-
scales, FeO has also been found to be stable in very minates in Fe2O3 (O2). This leads to the well-
thin films at temperatures down to 400 C4 and within accepted growth scheme for multilayered scale
narrow cracks at 500 C.5 growth on iron as shown in Figure 3, the governing
In the initial stages of oxidation, nucleation of equations for individual layer growth being
oxide occurs at favored crystallographic sites, fol-
lowed by preferential lateral spread from these nuclei Fe3 O4 þ Fe ¼ 4FeO above 570 ∘ C
to form a continuous thin film over the iron surface. 4Fe2 O3 þ Fe ¼ 3Fe3 O4
At low temperatures (<200 C), thin films, tens of 2Fe3 O4 þ O ¼ 3Fe2 O3
nanometers thick, form exceedingly rapidly. Ion
mobility is driven by space charges between gas– It was originally considered that the growth of these
oxide and oxide–metal interfaces. As the scale thick- layers was largely controlled by lattice diffusion, fol-
ens, the initial rapid growth rate is superseded by an lowing the theories expounded by Wagner.13,14 How-
extremely slow, logarithmic or inverse logarithmic ever, experimental work has shown that, for Fe3O4
growth law. below 600 C and Fe2O3 at all temperatures, theoreti-
For iron oxidizing at temperatures between 350 and cal calculations of oxide growth, using lattice diffu-
500 C, Fe3O4 nucleates first and grows laterally over sion coefficients are unable to predict the growth rates
the surface.6 Once complete coverage is achieved, the observed in practice.15,16 Tracer studies during the
Fe3O4 thickens parabolically. Ultimately, nucleation growth of Fe3O4 on Fe at 500 C have shown that
and lateral overgrowth by Fe2O3 occur, slowing the Fe3O4 growth is dominated by outward diffusion of
growth rate of the Fe3O4 appreciably, because of the Fe ions along short-circuit paths in the oxide.17 These
lower effective pO2 at the Fe3O4 surface, that is, the pO2 short-circuit paths are largely considered to be the
falls from that of the environment to that governed by oxide grain boundaries.15 For Fe3O4 above 600 C, the
the equilibrium between Fe3O4 and Fe2O3 at the pre- observed parabolic growth rate constants are within
vailing temperature. Both FeO and Fe3O4 nucleate first
and grow out of the surface of pure Fe and Fe–3%Cr II III IV
alloy at 700–800 C in low pressure O2 (103 Pa).7,8
While Fe2O3 is thermodynamically stable, it can form
only when the oxygen available for Fe2O3 growth has
increased sufficiently. Initially, the available oxygen is
continually depleted by the formation of Fe3O4. The
length of this induction period prior to Fe2O3 growth is Fe Fe3O4
Fe O Fe2O3 O2
a direct function of the prevailing pO2, being 3 h at Fe2+
Fe2+ O2−
40 Pa, but only 1 min at 101 kPa.7 Fe3+
Above 570 C, a distinct break occurs in the
e− e− e−
Arrhenius plot for iron, corresponding to the appear-
ance of FeO in the scale. The Arrhenius plot is then
nonlinear at higher temperatures. This curvature @ II Fe3O4 + Fe 4 Fe O (wüstite growth)
is due to the wide stoichiometry limits of FeO: @ III 4 Fe2O3 + Fe 3 Fe3O4 (magnetite growth)
limits that diverge progressively with increasing tem- Also @ III 2Fe3O4 + O 3 Fe2O3 (haematite growth)
perature. Diffraction studies have shown that com- Figure 3 Simplified scheme for the diffusion-controlled
plex clusters of vacancies exist in Fe(1 x)O.9,10 growth of multilayered scales on pure iron and mild
Such defect clustering is more prevalent in oxides steel above 570 C.
an order of magnitude of those calculated from tracer chemical degradation by the environment, and should
diffusion of Fe in Fe3O4.18 A similar correspondence be easily fabricated. These requirements are often
between theory and practice has been found for the conflicting, and a compromise is required for alloy
growth of Fe3O4 by the solid state reactions from FeO design. Simple iron alloys form oxides that are not
and Fe2O3 between 600 and 1200 C.18 The growth normally protective enough at temperatures above
rate of FeO is within 10% of the theoretical rate about 550 C.33 Therefore, the addition of other ele-
expected from Fe lattice diffusion, calculated accord- ments is needed to form a more protective scale.
ing to the Wagner theory.19 These elements usually comprise chromium, alumi-
The measured activation energy for Fe2O3 growth num, and/or silicon. For these elements to confer
is of the order of 169–222 kJ mol1.20 This activation adequate protection, the scale that they develop
energy is much lower than that expected from mea- should be stoichiometric (to minimize ionic transport
surements of the tracer diffusion of Fe or O21 and the rates), free of gross defects, stress free at temperature,
measured rate constant for Fe2O3 growth is approxi- should resist spalling, and should not be volatile by
mately ten times greater than that calculated according further reaction with the environment.34
to Wagner.13 O18 tracer diffusion studies of the growth It has already been shown that bulk lattice dif-
of Fe2O3 during the oxidation of Fe at 823 K have fusion is not generally considered to be the rate-
shown that, although the primary diffusion species controlling process for the oxidation of iron in most
through the crystal lattice is Fe ions, fast, inward O2 real situations. Hence, the classical Wagner treatment,
diffusion occurs down cracks in the scale.22 whereby the valency of the alloying element increases
The oxidation rate of pure Fe in O2 has been shown or decreases the number of lattice defects, provides
to be affected by the specimen shape and the original little clue as to the overall effects of alloying elements
surface profile.23 For iron oxidizing in O2, 50:50 on the oxidation of iron alloys. Rather, studies have
O2 þ H2O and O2 þ CO2 rough surfaces have been shown that the effects of alloying elements on the
found to oxidize more slowly than smooth surfaces, oxidation of iron alloys are largely brought about by
since surface irregularities hinder the oxide flow.23,24 their mobility (or lack of mobility) in the iron oxide
The oxide is unable to deform sufficiently to maintain lattice, and the effect this has on scale morphology,
intimate contact with the metal surface such that phase structure, and oxide plasticity. The addition of
porous scales are formed.25 even small quantities of alloying elements can pro-
For the oxidation of pure Fe at 500 C, the early foundly influence the scale morphology produced
reaction rate is more rapid on cold-rolled than annealed and hence the subsequent oxidation behavior.
surfaces. On cold-worked surfaces, the scale tends to be The Fe3O4 unit cell comprises an inverse spinel
rough and poorly adherent and contains less Fe3O4.26 structure, with Fe2þ and Fe3þ cations occupying octa-
Cold-work enhances the oxidation rate by mopping up hedral and tetrahedral interstitial sites within a close-
vacancies so that pore-free scales are formed which packed oxygen lattice. The occupation of an interstitial
maintain contact with the substrate.27–31 One of the site is accompanied by a specific site energy and partic-
authors32 has found an effect of stored cold-work dur- ular cations show a preference for occupation of that site
ing the oxidation of fracture surfaces at 600 C, which in which they sit most comfortably, that is, the lowest-
disappears with increasing oxidation because of the energy site (Figure 4). The difference between the
effect of surface annealing and consumption of potential energy of a cation in a preferred site to that
the cold-worked layer by oxidation. At long oxidation in a nonpreferred site is termed the crystal field prefer-
times (>200 h), the oxidation rate of a fracture surface ence energy35 (CFPE) and is principally determined
becomes indistinguishable from that of an annealed by the number of electrons on the d-shell. The direc-
emery-ground surface.32 At temperatures above tion of travel of a cation through the magnetite lattice
approximately 625 C, there is no effect of cold-work.23 is via alternate octahedral–tetrahedral–octahedral site
transfers.36 These interstitial sites are interconnected by
saddle points within the lattice, and in order to transfer
1.22.4 General Alloying Effects from one interstitial site to another, the diffusing ion
on Oxidation must acquire enough energy to surmount this energy
barrier and pass through the saddle point. If the direc-
The general requirements for higher temperature tion of movement of the cation is from a favored to a
alloys is that they should be cheap, should possess nonfavored site, additional energy is required to over-
adequate mechanical strength, should be resistant to come the CFPE. If, however, the direction of cation
Activation energy Oct Tet = xp + xa At temperatures below approximately 600 C,

Activation energy Tet Oct = xa numerous studies have shown that the thickness of
the inward growing layer approximates to that of the
Tetrahedral
site outwardly growing magnetite þ hematite scale, with
the location of the spinel–magnetite interface approx-
imately located at the original metal surface.34,37,44–46
Potential energy
xa
Hence, the new spinel formed approximately bal-
ances the volume of metal consumed, implying
Octahedral
equal amounts of oxygen and cation transport in
site duplex film growth34 and also implying that the
xp = CFPE rate of growth of the two scales are linked to one
another. Moreover, it is now considered that the over-
all growth rate is mainly determined by the rate of
cation diffusion outward through the outer layer,
Reaction co-ordinate since the cations must first diffuse out to provide
Figure 4 Potential energy/lattice position diagram for space for inner layer growth. Hence, the extent of
occupation of interstitial sites in Fe3O4 lattice alloying inner layer growth is dictated by the recession of the
element cations. oxide–metal interface.
These observations have been rationalized by
movement is in the opposite direction, there is a net Robertson and Manning,43 who suggest that, for single-
reduction in the total lattice energy equivalent to the layer growth, the metal ions are transported as metal
CFPE. It is apparent that, if the cation shows a high vacancies to the scale–metal interface. This inter-
CFPE between interstitial sites, then cation transfer is face consists of incomplete planes of metal atoms, or
much easier in one direction than the other and diffu- ledges, on which there are incomplete rows or jogs,
sion of the cation through the lattice will be severely analogous to interfacial edge dislocations. A metal
hindered. The CFPE for Cr, Mo, and V in octahedral vacancy in the oxide, on arrival at the interface, aligns
sites in magnetite is high, whereas that for Mn is virtu- with a jog atom on the metal surface. Oxidation then
ally identical to that of Fe, both of which are very low. occurs when the jog atom moves into the oxide and
Hence Cr, Mo, and V are virtually immobile in the annihilates the vacancy. Repetition of this process
magnetite lattice, while Mn is found throughout the causes continual stripping of the metal ledge and
scale at atom ratios to the iron similar to those found in allows oxide to fall into the space created (analogous
the original metal.37,38 to the climb of an edge dislocation). The surrounding
As a result of this lack of mobility of certain alloying oxide lattice then relaxes, maintaining adhesion. Since
elements, a new phase layer appears in the scale. This Cr is much less mobile in Fe3O4 than Fe, a Cr ion
new layer grows inward from the original metal surface, tends to remain at the site where it first enters the
is in intimate contact with the substrate alloy,39 and has oxide and inhibits dislocation movement.47 This pre-
an alloying element composition approximately 1.5 vents the scale falling into the space created by the
times that of the original metal.37 Moreau40 identified outward diffusion of Fe ions and microvoids (a few
this inward growing layer on Fe–Cr steels as FeCr2O4 atom sites) are formed at the scale–metal interface.
globules within a wüstite matrix. Further, Rahmel39 This initiates duplex growth by creating the space
determined that the inner layer of four-layered scales required for inner layer growth, without compromis-
grown on iron alloys containing Cr, Mo, V, and Si ing adhesion.
consisted of FeO containing (Fe, X )3O4, where X is For many low-alloy steels, therefore, the scale phase
Cr, Mo, V, or Si. The oxygen transport through the sequence is as shown in Figure 5 and the governing
inner layer is thought to take place via pores within equations for individual layer growth are similar to
the inner layer, since solid-state diffusion of oxygen those for pure iron, with the addition of
through the magnetite lattice is five orders of magni-
tude too low,41 and grain boundary diffusion is also too Fe3 O4 ¼ Fe2þ þ 2Fe3þ þ 4O2
low, to account for the observed growth rates.42 The
oxygen pathways are now thought to comprise grain at the spinel–magnetite interface and
boundary triple points and transient microvoids, con-
tinuously created and rearranged by creep.43 3ðFeX Þ þ 4O2 ¼ ðFeX Þ3 O4
The oxidation resistance of low-alloy steels By adding a third element (C ), it is possible to

0 I II III IV increase the likelihood of forming compound BO, at
a lower concentration than would be required in the
pure binary alloy A–B, provided that the third element
C has an affinity for O intermediate between that of
A and B.49 If the mobility of B in the ternary alloy is
high enough and element C has sufficient thermody-
Fe (FeCrMo V)3O4 FeO Fe3O4 Fe2O3 O2
namic activity in the alloy to form CO, then the
potential of O may be lowered to such an extent that
BO forms more readily, that is, element C acts as a
O2-
getter for element B. Such behavior is characteristic
of, for instance, Fe–Al–Si and Fe–Cr–Si alloys.
At 0 3 (Fe, Cr, Mo, V) + 40 (Fe, Cr, Mo, V)3 O4 1.22.5 Effects of Specific Alloying
FeO Fe + O
I.
II III and IV as mild steel (Figure 3)
Elements on the Oxidation of Iron
Figure 5 Simplified scheme for the diffusion-controlled Although in reality, binary low alloy steels are rarely
growth of multilayered scales containing spinel layers on
low-alloy steels.
used in industry, the effects of single element additions
to iron will be discussed. However, the effects of fur-
ther additions whether intentional or as impurities to
at the metal–spinel interface. The Fe2þ and Fe3þ ions form ternary, quaternary or multi-element alloys can
diffuse through the overlying magnetite (or wüstite) be large and may alter the behavior of the previous
to promote higher oxide reduction as before. The system to which they were added to a great extent.
O2 ions percolate through pores and microfissures
in the underlaying spinel to promote growth of the 1.22.5.1 Carbon
spinel layer.
The addition of carbon to iron principally affects the
With some alloying elements, for example, Si and
subsequent oxidation response via the oxidation of the
Al, the concentration of element required in the
carbon in the steel to form CO and CO2. Runk and
lattice to form their own oxide is so low that prefer-
Kim50,51 and latterly Nosek and Werber52 have all
ential formation of pure Al2O3 or SiO2, which are
reported similar observations on the oxidation of
both highly protective, may occur. For the reaction
Fe–C alloys at temperatures up to 400 C. It appears
A O þ Bðin alloyÞ ¼ Aðin alloyÞ þ B O that magnetite nucleates first over the ferrite, the
growth reaction obeying two-stage logarithmic kinetics
the limiting mole fraction (NA) of A required for with electron transfer through the scale being the rate-
protection is given by controlling process. Due to the formation of a kinetic
NA 1 barrier of CO and CO2 at the carbide–oxide interface,
¼ exp ðDGAO DGBO Þ ½3 only a thin film (15 nm) of randomly oriented crystal-
1 NA RT
lites of g-Fe2O3 of approximately 7 nm mean grain
where DGAO and DGBO are the free energies of for- diameter form over the cementite. Continued oxida-
mation of the oxides AO and BO, respectively. tion results in grain growth in the g-Fe2O3 due to
Provided the mole fraction of A does not fall below strain-induced grain boundary migration and a transi-
NA, the oxide AO will be formed exclusively. The tion from g- to a-Fe2O3. This latter transformation is
important criterion is the ratio of the oxidation para- accompanied by a 2.3% volume decrease, which causes
bolic rate constant to that of the diffusion coefficient cracking and allows the carbon oxides to escape. Ulti-
of A.48 For Al in Fe, the parabolic rate constant is very mately, lateral magnetite overgrowth of the cementite
low, while the diffusion coefficient is relatively high, occurs, and once Fe3O4 has spread over the whole
whereas the diffusion coefficient of Cr is much lower. surface, protective kinetics prevail, since the magnetite
Hence, the bulk alloy composition of Al in iron is impermeable to the carbon oxides.
required for the exclusive formation of Al2O3 at any Bohnenkamp and Engell53 and Caplan et al.54,55
given temperature is lower than the Cr concentration have also reported rapid carbon loss from the steel
required for the exclusive formation of Cr2O3. during the initial stage of oxidation at higher
temperatures (850 C) followed by a much lower, or of aluminium to steels produces a dramatic change in
zero, loss of carbon later. Caplan et al. measured the their mechanical properties with a reduction in
CO2 evolution by infrared gas analysis and reported toughness and an increase in ductile-to-brittle tran-
that the percentage of carbon loss from 0.1% C, 0.4% sition temperature (DBTT), which is detrimental in
C, 0.8% C, and 1.2% C steels was overall very small many circumstances. There is a need, therefore, to
and may be redistributed in the metal. There was no realize the protective benefit of aluminum at as low
carbon loss detected at 700 C. The overall oxidation a concentration as possible within the steel. Al com-
rates were all found to be parabolic at 850 C and less positions of 8 wt% results in severe embrittlement
than the oxidation rate of pure Fe. In the Fe–C alloys, at room temperature, and the oxidation of iron–
the individual phases of FeO, Fe3O4, and Fe2O3 were aluminum alloys has been reviewed in detail by
found to be less regular than those of pure Fe and Prescott and Graham.57 At aluminum concentrations
were often highly porous. below 2.4 wt%, bulky stratified scales, comprising
Malik56 reported that, at temperatures between 600 Fe2O3 and Fe3O4, with an inner layer of Al2O3 or
and 850 C, in 101 kPa oxygen, the oxidation rate of FeAl2O4 are formed at 800 C.58 At 2.5% aluminum,
Fe–5% M–C steels (where M was Si, Ti, V, Nb, Ta, Cr, large areas of Al2O3 were always observed with iron
W, or Ni) fell as the carbide stability increased. The oxide nodules. Formation of these iron oxide nodules
oxidation of all of the alloys obeyed parabolic kinetics, is suppressed only when the aluminum content
although some breakaway occurred following an incu- exceeds approximately 7 wt%.58–60
bation period. This breakaway was attributed to scale Ahmed and Smeltzer,61 Pons et al.,62 and Smith
disruption, as a result of CO2 evolution, with the et al.63 have all found that, at 1173 K, iron alloys
carbon loss being most rapid during the first 5 min. containing around 5–6% aluminum initially form a
Although the amount of carbon loss increased with rapidly growing duplex scale comprising an outer
the carbon content of the alloy, as did the oxidation a-Fe2O3 layer overlaying an inner (FeAl)3O4 layer.
rate constant, the total carbon loss was very much lower However, with prolonged oxidation, an Al2O3 layer
than that available. Those alloys forming a pure carbide eventually forms at the oxide–metal interface and
phase were found to have a lower oxidation rate than precipitates within the alloy. The reaction rate grad-
those alloys comprising a solid solution phase or ually decreases as the Al2O3 at, or near, the surface
cementite. All of the binary Fe–5% M alloys displayed coalesces eventually to form a continuous film that
a similar reaction rate, which was approximately one virtually stops the outward diffusion of Fe ions. Elec-
order of magnitude lower than that of pure Fe, because tron backscattered Mössbauer spectroscopy studies
of the formation of mixed oxides or spinels in the scale. have shown that the outer Fe2O3 contains approx-
The Fe–5% M–C alloys always showed two-layered imately 10% Al3þ and the inner layer comprises
scales, with an inner mixed oxide or spinel overlaid Al2O3 with some Fe3þ.(64) The benefits afforded by
with Fe2O3. The scales formed on the high-carbon alumina scale formation are well documented: slow
alloys were generally more compact and adherent (fol- growth as a result of low defect concentrations and
lowing initial scale disruption by C loss) because of stability at high temperature. A disadvantage asso-
carbide dispersion improving scale integrity. Malik56 ciated with alumina scales on iron is the relatively
argued that the carbide-forming elements retard the poor scale adherence which can result in spallation of
diffusion of carbon in austenite, reducing the overall the protective layer.65
scaling rate. In non-carbide-forming alloys, such as Alumina generally develops as either a-Al2O3
Ni and Si, the oxidation rate was greater because of or g-Al2O3 during the oxidation of Fe-Al alloys.
a higher carbon mobility in the steel. a-Al2O3 forms, and is stable at higher temperatures
At 850 C and 1.2% C, the oxidation rate was found and is more desirable for conferring oxidation resis-
to be in the order Fe–Cr–C > Fe–C > Fe–Ni–C > Fe– tance.57 Initial transient stages of alloy oxidation show
Ti–C > Fe–Ta–C > Fe–Nb–C > Fe–V–C > Fe–W–C, scale compositions controlled by kinetics rather than
which is nearly, but not exactly, the sequence of carbide those predicted by thermodynamics. The growth of
stabilities. convoluted transient scales containing all of the
alloying elements is quickly replaced by a slow grow-
ing protective alumina scale dependent on composi-
1.22.5.2 Aluminum
tion and temperature. The critical Al content for the
On oxidation, aluminum forms the highly refractory, exclusive formation of Al2O3 has been found to be
and hence protective, Al2O3. However, the addition raised by the presence of Ti and B.66
The beneficial effects of adding both aluminum only approximately 0.05%.77 However, at higher oxy-
and chromium to steels has been demonstrated by gen potentials, the higher growth rate of transient
Tomaszewicz and Wallwork66 during oxidation stud- iron oxides suppresses the growth of this continuous
ies of Fe–Al–Cr alloys at 800 C in pure oxygen at SiO2 scale. Even at 5% Si in the steel, no continuous
26.6 kPa. They showed that Al acts as a primary getter SiO2 layer formed during oxidation in air, whereas
for oxygen, nucleating Al2O3, with Cr acting as a such a film was seen to form during oxidation in Ar at
secondary getter, nucleating Cr2O3. If there had a pO2 of 104.
been sufficient Cr and Al in the alloy, no iron oxides Logani and Smeltzer78–80 have observed that,
would have been formed. The total Al þ Cr content for Fe–1.5% Si at 1000 C in CO/CO2, the initial
to suppress nodule formation was found to be in the slow reaction rate was followed by regions of linear
range of 7–8%, with 7% Al required at 0% Cr, but behavior because of the amorphous SiO2 film being
only 3% Al required at 5% Cr. consumed by the growth of wüstite–fayelite nodules
The diffusion coefficient for S in Al2O367 at during the early stages. These wüstite–fayelite nod-
950 C is approximately 100 times lower than that ules were nucleated at alloy grain boundaries and
in Cr2O3.68 Hence, for high temperature applications then grew laterally to inundate the SiO2 films over
in S environments, aluminum confers a much greater alloy grains.
degree of protection than that afforded by chromium. During the oxidation of high Si content steels in
high-pressure CO2, the oxidation reaction can sud-
denly switch to a highly protective mode that pro-
1.22.5.3 Silicon
ceeds extremely slowly after an initial incubation
The addition of silicon to iron has been reported period.70 This follows the formation of a healing
by many authors to confer significant corrosion pro- layer, comprising a line of amorphous silicon-rich
tection. Rahmel and Tobolski69 found that binary alloys oxide along the oxide–metal interface with Cr
containing up to 4% Si displayed a limiting corrosion enrichment to 30–40 at% just above the healing
rate due to the formation of an iron-silicate layer when layer and Cr depletion in adjacent metal sites. Slow-
exposed to pure oxygen, O2 þ H2O or CO2 in the ing of the oxidation reaction by the SiO2 permits
temperature range 750–1050 C. However, Robertson selective oxidation of Cr. Cr and Si are then syner-
and Manning70 indicate that Fe–Si oxides are generally gistic, with less Si being required in Cr-containing
immiscible. Several authors71–74 have reported that the steels than in straight Fe–Si steels, because of second-
corrosion protection apparent in Fe-Si alloys arises ary gettering. For Fe–Si, then, 2.5–3% Si is required
from the formation of a SiO2 healing layer, beneath for healing layer formation, irrespective of the tem-
the magnetite, which acts as a barrier to outward trans- perature. For Fe–Cr–Si, however, the critical Si
port of metal ions. At the low oxygen potentials at the content decreases with increasing Cr content and
base of the scale, and for very thin films, charged effects temperature.
in the amorphous SiO2 network and electronic carriers Silicon has a major effect on the rate of oxidation
control the growth of the SiO2.42,75 of iron. Yang et al.81 found that small additions (less
Adachi and Meier76 studied the oxidation of Fe–Si than 2 wt% Si) result in a marked decrease in the rate
alloys under isothermal and cyclic oxidation condi- of oxidation up to 1000 C in dry moving air, however,
tions, in air, between 900 and 1000 C. They found at 1200 C the formation of a liquid phase result-
that oxidation rates in air decreased with the silicon ing from the melt of FeO and SiO2 substantially
content, such that at 10 wt% Si the oxidation rate was increased the oxidation rate. Adachi and Meier76
lower than that conferred by Cr, because of the for- showed that the addition of 5 wt% Si is observed to
mation of a continuous film of SiO2. The oxidation decrease the oxidation rate of iron by more than
kinetics of these steels was found to be linear because two orders of magnitude and additions of 10 wt%
of the diffusion of Fe through a film of SiO2 of con- Si or more results in oxidation rates slower than Fe-26
stant thickness. Outward diffusing Fe subsequently wt% Cr. Lashin et al.82 studied Fe – 6 at.%Si steels at
dilutes the SiO2 to form Fe2SiO4 and produced the 500 C – 540 C in ambient air of various pressures. The
Fe2O3 as an outer layer. At greater than 10% Si, SiO2, tests were designed to show the oxidation behavior of
overlaid with Fe2O3, comprised the total scale. steels intended for use in electro-magnetic applications.
At the low oxygen potentials found in CO/CO2 Ishitsuka et al.83 report a reduction in the steam
environments, the critical concentration of Si for the side oxidation rate of 9Cr- 0.5Mo 1.8W (P92) with
selective formation of a continuous film of SiO2 is increasing Si content. Improved oxidation resistance
was apparent at all temperatures investigated (500 C, 1.22.5.6 Phosphorus

550 C, 600 C, 650 C and 700 C). The largest
Like sulfur, phosphorus appears to have only a small
reduction in corrosion rate was observed at 700 C,
effect on the overall scaling of iron alloys in air. For
oxidation rates at all other temperatures showed sim-
example, minor editions (<0.1 wt%) have been shown
ilar rate reductions. At 700 C the remarkable reduc-
to increase the oxidation resistance of iron at 500 C,
tion in oxidation rate was attributed to the formation
but above this amount the oxide layers begin to break
of a protective amorphous SiO2 film grown at the
and become nonprotective.94 However, phosphorus
scale metal interface. At temperatures of 600 C, or
may play a role in suppressing breakaway oxidation
less, only dissolved silicon in the Fe–Cr spinel lattice
in carbon steels in CO/CO2 environments. Donati
contributes to the effect of silicon on oxidation rates.
and Garaud95 found that the tendency for breakaway
was lower over ferrite, where P segregates. To confirm
1.22.5.4 Manganese
this, the authors doped pure Fe with P and found that
Since Mn is both soluble in iron oxides and mobile to the breakaway rate slowed down at 350 ppm phospho-
the same extent as Fe, the addition of Mn to steels has rus and was totally suppressed at 900 ppm phospho-
little effect on the overall scaling rate in air or oxygen. rus. A similar benefit has been reported by Dewanckel
Jackson and Wallwork84 have shown that between 20% et al.88
and 40% manganese must be added to steel before the
iron oxides are replaced by manganese oxides. How-
ever, Mn suppresses breakaway oxidation in CO/CO2 1.22.5.7 Nickel
possibly by reducing the coalescence of pores in the The addition of Ni to Fe dramatically reduces the
oxide scale. It also appears to be important in the oxidation rate due to the virtual insolubility of NiO
healing of microcracks in protective chromia/spinel in FeO.96 The oxidation of iron nickel alloys in
oxide scales formed on highly alloyed steels.85 general was reviewed by Foley in 1962.97 During
The addition of up to 15% Mn to pure Fe, under the oxidation of Fe–Ni alloys, and in the absence of
sulfidizing conditions at 1073 K, leads to a small nickel oxide formation, Ni concentrates at the FeO-
increase in the scaling rate.86 At 2% Mn, MnS forms substrate interface as a result of Fe depletion. Oxida-
as stringers in the subscale, but these do not form a tion rates decrease due to the reduced Fe activity at
coherent layer even at a concentration of 15% Mn. the base of the FeO scale and thus the stability of the
The increase in scaling rate is possibly due to FeO layer is diminished. Menzies and Lubkiewicz98
increased short-circuit diffusion, since metal diffusion found that the oxidation of an Fe-12%Ni alloy in O2
in MnS is much slower than in FeS.87 obeyed parabolic kinetics at all temperatures between
700 C and 1000 C. At 700 C, the Ni suppressed the
1.22.5.5 Sulfur formation of FeO, with the scale comprising only
The presence of small quantities of S in steels has little Fe3O4 and Fe2O3. Progressive enrichment of Ni
effect on the initial scaling rates in air, but may be occurred in the alloy as the substrate was consumed
detrimental to long-term scale adhesion. Sulfur has, by oxidation. When the concentration of Ni reached
however, been shown to be detrimental to breakaway 50–60%, the Ni entered the spinel phase, leading to
oxidation in CO/CO2 environments.88 However, it the formation of Ni(x) Fe(3-x)O4 with x approximately
has been shown to reduce the total uptake of carbon equal to 0.24 near the alloy surface and less than 0.01
in the steel under CO/CO289 and reduce the scale near the Fe2O3. At 900 C – 1000 C, the Ni entered
thickening rate. In this context, free-cutting steels the spinel in the early stages of oxidation, with x
were found to oxidize at a significantly lower rate, as values of 0.4 at 900 C. This led to a reduction in
did steels subjected to pretreatment in H2S. the parabolic rate constant. At 1000 C FeO is the
Modern steel making techniques are capable of stable oxide phase formed on Fe–Ni alloys, even up
routinely producing alloys of sulfur contents to 80% Ni, since FeO is more stable than NiO.99
< 50 ppmw;90 however, recent research by a number
of groups91–93 indicates that removing sulfur to
1.22.5.8 Chromium
extremely low levels may have a detrimental effect
on the oxidation resistance of 9–12Cr steels in steam Of all of the alloying elements added to steels, Cr has
and it may, therefore, be necessary to intentionally been the most used for improving the corrosion prop-
add sulfur to these alloys. erties. In terms of high temperature oxidation, steels
containing approximately 10% Cr are capable of form- Many studies have shown that surface pretreat-
ing a continuous, highly protective film of Cr2O3.33 ment of Fe–Cr alloys has a strong effect on the scale
Rhys-Jones et al. 100 reported that the isothermal oxi- morphology and subsequent oxidation rate.27,104–106
dation kinetics of Fe10Cr at 1000 C were parabolic. For instance, Caplan27 indicated that several Fe–Cr
However, significant reductions in the oxidation rate alloys showed improvements in corrosion resistance
are also realized at lower Cr concentrations because of because of cold work, with greater than 16% Cr
the formation of FeCr spinels101 and the suppression of required to show the optimum benefit. Khanna and
FeO formation to temperatures in excess of 570 C.102 Gnanamoorthy104 examined the effect of cold work
At 700 C, FeO exhibits a very narrow stability on 2.25% Cr–1% Mo steels at temperatures between
range on the Fe–Cr–O phase diagram (Figure 6) for 400 and 950 C over 4 h in 1 atm O2. They found that
Cr up to 6%.103 The wüstite stability range is almost up to 90% reduction by cold rolling had a negligible
negligible on Fe–0.5% Cr because of the high reactiv- effect on the oxidation rate up to 700 C. However,
ity of Cr toward O2 at 1000 C. However, while Cr2O3 above 700 C there was a general reduction in the
is normally protective on steels in air or O2 up to kinetics because of enhanced Cr diffusion in the alloy,
approximately 900 C, volatile CrO3 may form at leading to the formation of a Cr-rich spinel. In addition,
higher temperatures.33 Dewanckel et al.88 found that they found that breakaway corrosion occurred at
although the addition of Cr to steels is normally con- 900 C on annealed samples, but this did not occur on
sidered beneficial, low quantities (<500 ppm Cr) were samples that had been cold-worked. For Fe–10% Cr at
detrimental to the breakaway performance of low-alloy 600 C in air, Hossain found that the oxidation resis-
and carbon steels. tance increased with the degree of cold work.105
1.22.5.9 Molybdenum
-60 M2O3
In the temperature range 500–1000 C, additions of
M3O4 between 0.5% and 5.6% of Mo decreases the oxida-
-80 tion rate of iron by a factor of almost 10, with the
RT In pO2 (K cal)
M3O4
Fe O + maximum benefit being obtained at a concentration
M2O3
of approximately 2%.107 Like many alloying ele-
At 700 ⬚C
-100 M1 + spinel ments, Mo promotes the formation of a duplex spi-

nel–magnetite scale, with the Mo retained within the
-120
M1 + M2O3 inner spinel layer.105
M1
-140
1.22.5.10 Boron
Fe 20 40 60 80 Cr Tsipas and Rus108 and Suwattananont and Petrova109
found that thermally diffusing boron into the surface
of steel gave a boride layer with enhanced oxidation
M2O3
-80 resistance at temperatures up to 900 C by acting as a
M3O4 barrier to inward diffusing oxygen. Rowley et al.110
M3O4 also found that adding small amounts of boron to
-100 +
RT In pO2 (K cal)
M2O3 Fe–Cr alloys strongly affects the materials oxidation

M1 + spinel characteristics in super-heater steam. Enhanced
At 400 ⬚C
-120
manganese and/or chromium content of the oxide
films in comparison to those seen in undoped 9Cr
M1 + M2O3 steels provides greater protection and hence a reduc-
-140 tion in oxidation rates. Rowley proposes two likely
models for boron doped oxide growth:
M1 M1 + M2 M2
-160 1. The possibility that Schottky electron emission
Fe 20 40 60 80 Cr from the metal substrate into the conduction
Figure 6 Fe–Cr–O phase diagram for Fe–Cr alloys at band of the oxide layer and/or subsequent elec-
400 C and 700 C. tron diffusion may control oxide growth kinetics.
These processes are governed by the space charge oxidation behavior of boiler steels, for example.
in the oxide film which is due to the electric field Most industrial components in service will be sub-
produced by an ionic defect gradient through the jected to a thermal gradient across them, and the
layer. Existing theories indicate that this mecha- effect of this has been investigated recently by Fry
nism is applicable for oxide thicknesses ranging et al.114,115 They showed that the oxide grain size and
from 5–100 nm in the temperature region 300– morphology could change in subtle ways when sub-
600 C, both of these conditions are satisfied in jected to a heat flux, and that this could have impor-
the case of boron doped oxide films. tant implications under some conditions. Trindade
2. It is possible that a boron containing phase of low et al.116 have also shown that surface finish and cold
ionic conductivity is formed at grain boundaries. work can influence the substrate grain size and hence
This may lead to a progressive reduction in ionic the high temperature oxidation of pure Fe and
mobilities via a mechanism similar to that postu- 2.25Cr–1Mo steel. Decreasing grain size increases
lated for the reactive elements effect111 (cited by oxidation rates by facilitating oxygen diffusion down
Rowley et al.). Although boron does show a ten- grain boundaries. Internal oxidation including the
dency to migrate to grain boundaries, it is experi- formation of FeCr2O4 and Cr2O3 along grain bound-
mentally observed that the limiting oxide thickness aries was observed. Oxidation kinetics decreased
is reached extremely rapidly indicating an almost strongly with an increase in grain size as a result of
instantaneous formation of a blocking phase. Fur- the reduction of grain boundary density leading to
thermore, the ultrafine grain size exhibited by boron less internal oxidation per unit volume. Fluctuation
doped oxides would provide a much larger number in grain size did not appear to affect the rate or
of short-circuit diffusion pathways than in control structure of external oxide scales.
oxides. This suggests that ionic blocking must be The effect of cold working on high temperature
extremely effective for this mechanism to be tenable. oxidation was also investigated. In agreement with
Ostwald,117 Trindade116 suggests the increased dislo-
cation density associated with cold working allows
1.22.5.11 Others
greater diffusion rates of chromium in bcc steel; this
Small additions of Ce have been shown to have increased mobility enables Cr enrichment of the spi-
a favorable influence on the oxide growth of several nel layer in comparison to non-cold worked speci-
Fe–Cr alloys by improving scale adherence and mens. The increased Cr content within the spinel
acting as nucleation sites for Cr2O3.112 In a scale affords improved oxidation resistance. Recrys-
H2H2OCOCO2 atmosphere at 700 C cerium tallization/recovery of specimens subjected to
levels as low as 0.024% reduce the carbon uptake of greater than 42% cold work resulted in low disloca-
steels in carbonaceous atmospheres by several orders tion densities and polygonization; the ensuing fine
of magnitude. However, Sroda et al.113 found addi- grained low dislocation density microstructure led
tions of Ce to low alloy steels to have a negative effect to an overall increase in oxidation rates.
on corrosion resistance at 500 C, especially, in HCl Samples of pure iron were heat treated to establish
containing atmospheres. The techniques used in the microstructures of two different grain sizes and then
study did not enable the author to detect Ce in the oxidised in laboratory air at 550 C for 72 h. Gold
oxide scale and thus it was not possible to determine markers were used to indicate the position of the
the role of cerium in the corrosion process (although original metal surface and to define the mass trans-
thermodynamic calculations show that the formation port direction. Studies reveal that the oxide scale in
of cerium oxides is possible). Trace concentrations of pure iron also grows by outward Fe migration and
As and Sn have been found to improve the breakaway inward oxygen diffusion and that, similarly to 2.25Cr–
properties of scales formed on mild and low-alloy 1Mo steels, oxidation kinetics of pure iron also
steels in CO/CO2, whereas Cu has been found to decrease with an increase in substrate grain size.
be detrimental.88 However, Caplan and Cohen30 studied the effects
of cold work on the oxidation rates of pure iron, where
in contrast to the findings of cold worked 2.25Cr–1Mo
1.22.5.12 Effects of Heat Flux and
steels, oxidation rates of cold worked pure Fe were
Cold Work
observed to increase as a result of cold work. It was
Another factor which has been largely ignored in proposed that as iron diffused outwards, vacancies are
earlier work is any influence of heat flux on the annihilated at dislocations or lattice steps on the cold
worked Fe surface. As a result, the oxide layer remains protection. For pure metals which oxidize in a para-
in good contact with the Fe substrate in the absence of bolic manner, repetitive scale spalling may produce a
pore formation and rapid oxide growth ensues. much increased total metal loss over time, since the
reaction kinetics continually revert to the initial rapid
period of the oxidation curve and paralinear kinetics
1.22.6 Stress Effects ensue.
The majority of heat-resisting alloys contain at
While alloying elements may initially determine the least one element that is selectively oxidized (e.g.,
protective nature of an oxide scale, the response of Cr, Si, Al). The resulting scale is highly protective,
this oxide scale to stress is often crucial in determin- but depletion of the secondary element may occur if
ing the long-term oxidation performance of an alloy the diffusion rate of this element in the alloy is low.
in an industrial environment.118 The stresses in an Then repetitious loss of the scale can be profoundly
oxide scale may arise both internally (because of detrimental to long-term performance, since depletion
growth stresses) and externally (because of applied of the alloying element near the surface may occur to
stresses).119 In general, oxide scales do not possess such an extent that its concentration falls below the
sufficient slip systems for plasticity and thus rely critical level for the exclusive reformation of the
upon diffusion-controlled creep for plastic deforma- highly protective scale.126,127
tion and stress relief.120 At temperatures below approx-
imately 600 C, only elastic deformation is possible and
1.22.6.1 Growth Stresses
stress relief by cracking is likely to occur. If the scale–
metal interface is strong, and the bulk oxide is weak, As indicated earlier, protective oxide scales typically
through-scale cracking results when the combined have a PBR greater than unity and are, therefore, less
stresses exceed a critical value.121 If the interface has dense than the metal from which they have formed.
a lower strength than the oxide bulk strength, decohe- As a result, the formation of protective oxides invari-
sion occurs first. A minimum energy is required for ably results in a local volume increase or a stress-free
cracking or decohesion.122,123 Evans and Lobb121 have oxidation strain.128,129 If lateral growth occurs, then
calculated that cracking occurs when the strain energy compressive stresses can build up, and these are
per unit volume (w*) of oxide contained in layer thick- intensified at convex and reduced at concave inter-
ness (t) equals the work required for internal cracking faces by the radial displacement of the scale due to
(Gc) or decohesion (Gd). For internal cracking, outward cation diffusion (Figure 7).48,129,130
The reduced oxidation near sample corners is
8PGc 4Gc
w ¼ ½4 related to these stress effects, either by retarded
fL fL diffusion or by modified interfacial reactions.128
Manning131 described these stresses in terms of the
where L is the side length of unit volume; f, the fraction
conformational strain and distinguished between
of stored energy in the oxide; Gc, the energy for unit
anion and cation diffusion, and concave and convex
area of cracked oxide surface; and P is the geometric
surfaces. He defined a radial vector M, describing the
parameter (0.5). For decohesion,
direction and extent of displacement of the oxide
Gd layer in order to remain in contact with the retreating
tw ¼ ½5
f metal surface, where
where f is the energy/unit area of fresh metal at the Vox

M¼ ð1 aÞ1ð1 uÞ ½6
oxide–metal interface. Vme
Through-scale cracking may not necessarily be
detrimental to an alloy if rapid scale healing can and Vox/Vme is the PBR, a is the amount of oxide
occur,124 and scale delamination can reduce the total formed at the surface, and u is the volume gain due to
scaling rate if the scale remains adjacent to the alloy vacancy injection into the metal. If u is zero, the sign
surface.119,125 In this instance, scale separation con- of M depends upon the predominance of anion
founds the transport of metal ions across the interface (a ¼ 0) or cation (a ¼ 1) diffusion. The enforced radial
into the oxide. However, if through-scale cracking displacement (dy) of the oxide layer results in a tan-
occurs in conjunction with delamination, scale gential or radial reaction stress in both the oxide and
spalling may follow,121 with an attendant loss of the metal. The strain rate (de/dt) of the oxide layer
Old oxide
New oxide
Metal
Anion diffusion Cation diffusion
Convex
M
y= M h
R R M
R
afh y h (1-V )
a=0 a=0
d = f h (1– a) f h
M=1 M = −1
V
–
M = f (1 – a) - (1 – V )
(f = 2 V = 0)
R R
M
M
Concave
Figure 7 Oxide growth stresses on curved surfaces. Reproduced from Hsueh, C. H.; Evans, A. G. J. Appl. Phys.
1983, 54, 6672.
then depends upon the radius of curvature (R) and the 2.5 104.(136) Components in service may be stressed
rate of metal loss (dh/dt) as beyond these failure strains, leading to scale cracking.
Ward et al.137 have shown that, under cyclical load-
de MðdhÞ
¼ ½7 ing, the oxidation rate of steels is similar to that under
dt Rðdt Þ
unstressed isothermal conditions, provided the fatigue
This leads to high strains on regions of high curvature stress is below the stress required to exceed the scale
and zero strain on a flat surface. If new growth occurs failure strain. If, however, the failure strain is exceeded,
stress free, then the oldest region of oxide will be the the oxidation rate is accelerated because of repetitive
most highly stressed, this being located at the oxide scale failure, and linear kinetics are observed.
surface for anion-controlled growth and at the oxide– Low-cycle fatigue loading of 9.5% Cr steel at
metal interface for cation-controlled growth. Hsueh 650 C in air has been shown to enhance uniform
and Evans128 carried out a similar analysis and calcu- scale formation as well as promote nodular scale for-
lated the distribution of radial and tangential stress mation at cracks.124 However, no cracks were found in
within the oxide scale as a function of scale thickness the scale after exposure, indicating that any cracks
and specimen geometry. that form must heal very quickly. This healing of
Whisker growth is a process often reported for the cracks was attributed to the overgrowth of the
relief of compressive growth stresses during oxi- chromium-rich oxide by an iron-rich oxide.
dation.132–134 This whisker formation and scale buckl-
ing of thin hematite layers have been linked to the 1.22.6.3 Thermal Stresses
stresses arising from oxide formation at the magne-
Under thermal cycling conditions, the principal source
tite–hematite interface as a result of the countercur-
of stress within the oxide scale is the temperature
rent of cation and anion diffusion in both oxides.130
change.121 Christl et al.125 have noted that, when cool-
Acoustic emission studies of 2.25% Cr–1% Mo steel
ing 2.25% Cr–1% Mo steel from 600 C in air, com-
at 900 C have shown that, as the oxide thickness
pressive stresses build up in the hematite, while tensile
increases, growth stress builds up in the scale and is
stresses build up in the magnetite and spinel layers.
relieved by scale cracking.135
This arises because the thermal expansion coefficients
of the individual oxide layers increase in the order
1.22.6.2 System-Applied Stresses
a metal < a spinel < a magnetite < a hematite.119
The tensile failure strain of oxides grown on EN2 steel Multilaminated scales have been reported follow-
between 600 and 900 C lies in the range 1 104 to ing thermal cycling,138 and Rolls and Nematollahi139
have studied the influence of thermal cycling on the 25% Fe2O3 at all temperatures below 600 C
oxidation of 1% Cr–0.5% Mo low-carbon steel. The (Figure 8). Above 600 C, a coarse, columnar-grained
oxidation kinetics were found to be mostly parabolic, layer of FeO occupying approximately 90% of the
with thin scales (10 mm) more prone to spalling than scale thickness was observed. For the Cr-containing
thick scales (20 mm). The authors reported that the steels, a distinct spinel phase was observed at tem-
higher the temperature drop on cycling, the greater peratures in excess of around 615 C, with all of the
the degree of scale disruption. They derived a quali- Cr, Mo, and V incorporated within this layer, beneath
tative relationship to describe this behavior: the Fe3O4 and Fe2O3. FeO was not observed in the
scale until temperatures exceeded 650 C, the FeO
f ðNOÞ
Y ¼ ½8 appearing as a fourth layer between the spinel and
x Fe3O4 layers. Parabolic oxidation kinetics were
where Y is the degree of scale detachment, N is observed for all steels at all temperatures. Below
the number of cycles, x is the thickness, and O is the 600 C, there was little difference in the oxidation
cooling rate. This implies that thicker scales cooled kinetics between the chromium steels until the Cr
slowly show less detachment than thinner scales cooled level reached around 5%, this difference mainly aris-
rapidly. The oxide scale comprised spinel, magnetite, ing because of an increasing activation energy for
and wüstite, with only the spinel layer remaining adher- Fe3O4 growth with increasing Cr content. There
ent during thermal cycling. From this, the authors was no significant difference in the activation energy
concluded that the bulk scale cohesive strength was for Fe2O3 growth, irrespective of the Cr content of
less than the inner spinel adhesive strength. Scale the steel (Figure 9).
detachment was observed at voids and microcracks
produced at the outer–spinel interface during isother-
1.22.8 Industrial Environments
mal oxidation. Hence, the number of scale layers that
became detached during thermal cycling was gov- 1.22.8.1 Steam
erned by the number of parallel rows of voids in the
Many low alloy steels are utilized as pressure vessels
scale and not by the number of thermal cycles.
in high temperature applications in a number of
1.22.7 Commercial Low-Alloy Steels

Thickness of phase as a percentage of total scale thickness
in Air or Oxygen 100
90
Simms and Little140 have examined the early stages FeO
of scale growth on 2.25% Cr–1% Mo steel at 600 C 80
in dry flowing oxygen. Between 1 and 22 h, they
found that a thick oxide layer spreads laterally over 70
a thin oxide layer. After 50 h, no thin areas of oxide
layer were left. Whiskers gradually developed on the 60
outer surface and these were well defined after 100 h.
50
Fracture sections of the oxide revealed that the thin
scales were duplex, whereas the thicker scale was 40
triplex, with a middle layer of Fe3O4 which spread
laterally with time. The authors concluded that the 30
first phases to develop were fine equiaxed a-Fe2O3
overlying a doped spinel. Later, nucleation and lateral 20
spread of coarse columnar grains of Fe3O4 occurred
10 Fe3O4
between the two first-formed layers. The Fe2O3
Fe2O3
appears to reach a limiting thickness at which time 0
Fe3O4 nucleates. 500 550 600 650 700 750 800 850
One of the authors141 has carried out exposures of Temperature (°C)
a range of low-alloy steels (up to 5% Cr) in labora- Figure 8 Relative thickness of wüstite, magnetite and
tory air between 500 and 850 C. Plain carbon steel hematite on mild steel as a function of formation
displayed a duplex scale, comprising 75% Fe3O4 and temperature.
Key
offered a possible insight into the increased steam side
Mild steel oxidation of low alloy steels.144–146
10-9
½ % Cr ½ % Mo ¼% V In recent years the group of modified 9%Cr steels
1¼ % Cr ½% Mo ¼% V has been developed for utilization in high tempera-
2¼ % Cr 1% Mo
ture environments including power generation.
5% Cr 1% Mo
10-10 Although the mechanical properties of these steels
have been optimized, it is now appreciated that steam
Parabolic rate constant (cm2 s-1)
side oxidation of these alloys may be the factor that

10-11
limits their maximum service temperature. Oxide
growth rates have been shown to increase exponen-
tially with temperature142 up to 700 C, where the
isothermal oxidation rates are observed to be para-
10-12 bolic. However, investigations (COST-522, Europe’s
Thermie and the U.S. Vision 21 programs) aimed at
improving thermodynamic efficiency by increasing
10-13 service temperatures suggested that linear kinetics
prevailed at temperatures above 700 C for steels
850 800 750 700 650 625 600 550 500
with Cr contents in the range 1–15 wt%.142,147
T (⬚C) 9% Cr ferritic martensitic steels show acceptable
9 10 11 12 13 oxidation resistance in dry air at high temperatures
104/T (K)
due to the formation of a protective scale; between
Figure 9 Arrhenius plot for the hematite thickening rate on 400 C and 600 C the scale comprises of FeO, Fe2O3,
mild steel and low-chromium alloy steels in air. and Cr2O3. However, operation atmospheres com-
posed of water vapour or steam þ oxygen are much
more corrosive than dry air. Subsequent scales
diverse and often aggressive environments. Metal that grow are less protective than Fe2O3, or Cr2O3.
wastage by oxidation and corrosion reduces the load Sanchez et al.148 predicts the formation of Fe3O4,
bearing cross sectional area of a vessel wall and can Fe2O3 and (FeCr)3O4 scales in 100% H2O atmosphere
have grave consequences in terms of creep rupture at 650 C on 9%Cr steels, in contrast, at 800 C the
life, especially of thin walled components. Different scale is mainly composed of Fe3O4 and (FeCr)3O4.
environment parameters including temperature, gas/ Itagaki et al.149 found that an addition of 3%Pd to
fluid composition, pressure, flow rates and alloy com- steel NF616 (similar to modified 9Cr–1Mo) gave signif-
position will define the oxidation mechanisms and icant improvements in oxidation resistance in steam at
rate of material wastage. It is therefore imperative 650 C as the growth of a Cr rich (Fe2Cr)2O3 rhombo-
that a material gains protection from the formation hedral structured oxide was observed in the absence of
of a stable slow growing oxide scale. Wright and a magnetite layer. Czyrska-Filemonowicz et al.150 state
Pint142 predict that in steam environments the pro- that oxidation (weight gain) of P91 steel can increase
tective behavior associated with the formation of a 30 fold in atmospheres containing water vapour com-
complete Cr2O3 layer is not necessarily observed in pared to dry air.
FeCr alloys with Cr contents below 20%. Tuurna et al.151 also studied the effects of water
Jonsson et al.143 compared the oxidation of iron at vapour on the high temperature oxidation of low
600 C in dry O2 and wet O2 (40% H2O) where a alloy steels. Tests on 2.25Cr–1Mo steels at constant
three layer scale is formed consisting of a haematite temperature (550 C) with variable moisture content
outer layer, a magnetite middle layer and a wüstite (5%–10%–15%) revealed an increasing linear rela-
inner layer forming the metal–scale interface. All tionship between moisture content and mass gain after
three layers grow with time, but with different growth 360 hours. Further tests at fixed moisture (15%) and
rates, the overall rate being roughly parabolic in both variable temperature result in a steep increase in oxi-
environments. While the detrimental effects of water dation occurring at higher temperatures as the oxide
vapour on oxidation rates are well reported, there is no scale becomes nonprotective; the use of 2.25Cr–1Mo
consensus regarding the causes of this effect, however, steels is thus restricted to temperatures below 580 C.
recent investigations into the effect of hydrogen-accel- Nakai et al.152 report the effects of preoxidation of
erated oxidation on scale growth and morphology has Fe, 10%Cr, 0.08%C ternary alloy on subsequent steam
oxidation.143 Cr2O3 scales containing a small amount of these scales has a propensity to occur along a plane
Fe2O3 are formed during preoxidation offering between one of the sets of duplex layers or at the
short term improvements in oxidation resistance. scale–metal interface and is likely to be a mechanism
After long-term oxidation in steam nodular like oxides of stress relief. There is evidence that there are con-
are formed on the preoxidised specimen showing a ditions where the scales formed on ferritic steels in
clear resemblance both chemically and structurally to steam can remain perfectly adherent for long times;
the oxides formed on non-preoxidised specimens. for instance, from tests in a side loop on an operating
It is understood that hydrogen dissolves inward boiler, Griess and Maxwell154 reported that alloy T22
when steam oxidation takes place on the surface of formed adherent scales throughout exposures for 28kh
steels.145,146,152 Thus it appears that hydrogen plays in 105 bar steam at 482 C; at 538 C exfoliation
an important role in the accelerated growth of initiated after approximately 12kh. Alloys 9Cr–1Mo
the oxide scale when the steels are exposed to a (0.46Si), 9Cr–2Mo, and Sandvik HT9 (11.4Cr) fol-
wet gas atmosphere. Nakai et al.152 proposes a model, lowed essentially the same kinetic behavior (and
where hydrogen dissolves as a proton (H+ ion) in rates) as T22, but exhibited no signs of exfoliation at
(Fe, Cr)3O4 modifying the point defect structure 482 and 538 C after exposures of 28kh, 19kh, and
inducing significant changes in the ionic diffusivity 28kh, respectively.
in the oxide. For example, the dissolved hydrogen The laminated scales of low alloy steels are rarely
will decrease the diffusion rate of Cr ions in observed on T/P91 steel, instead a largely duplex
(Fe, Cr)3O4. The ensuing inhibited ionic Cr supply scale is observed with an incomplete outer layer of
to the oxide layer results in the attendant decrease of haematite. The duplex layer consists of an outer mag-
the Cr concentration in (Fe, Cr)3O4. Subsequently an netite layer underlayed with a spinel layer rich in Cr,
apparent increase in the diffusion rate of Fe ions in Si and Mo in comparison to the alloy composition.
the oxide is observed. As the oxidation rate of Fe–Cr 12%Cr alloy steels display duplex oxide scale
steels is determined mainly by the Fe diffusivity in similar to those initially formed on the lower alloy
the (Fe, Cr)3O4 layer,153 hydrogen dissolution ferritic steels. The outer scale is predominantly mag-
induces the accelerated growth of the oxide scale on netite and the inner layer is a Cr rich spinel. The
the steels. inner layer shows morphology similar to that often
Low alloy steels containing a maximum Cr con- observed in P91 scales of incomplete Cr rich bands.
tent of 9 wt% but excluding T/P91 display similar As quoted in the EPRI report on oxidation of
scale morphologies and tendency/mechanisms to alloys in steam90 ‘‘There are five possible scenarios
spallation showing a laminated oxide scale. The ini- by which the scale morphologies observed on ferritic
tial formation of a magnetite outer layer and an inner steels in steam might develop. Thus, it is suggested
spinel layer is followed by the formation of similar that the development of the (initial) double-layered
underlying duplex layers (Figure 10), possibly as a scales on ferritic steels in steam (under isothermal
result of delamination of the original scale at the conditions before the onset of any transition to multi-
oxide–substrate interface.90 Repetition of this process layered scales) involves the following steps; note that,
develops a multilayered scale consisting of repeated in order to reflect the uncertainty in the dominant
duplex magnetite and spinel layers, successive duplex transport modes in these scales, five possible pathways
layers becoming noticeably thinner. Exfoliation of for oxide growth are suggested (Figure 11)’’:
Fe2O3 Steam
1
Original duplex
2
Exfoliation
3
Second duplex
4
Ferritic steel
(X100)
(a) (b)
Figure 10 (a) A cross section from a 2.25Cr–1Mo steel tube oxidised in steam. (b) A schematic representation defining the
duplex layered system.155
H2O(g) Û OH(ads) + O(ads) Û H2 + O(in oxide)

3Fe2+ + 4O2· + 2vac+ Û Fe3O4 H2O(g)?
Fe3O4 O 2· Fe2+,Fe3+
Fe3O4 Û 2Fe3+ + Fe2+ + 4O2·
Fe3O4/Fe2xCrxO4 2
1 2 3 4 5 O
Ferritic steel Fe2+ 2Cr3+ + 3O2· Û Cr2O3

Cr3+/Fe2+ Cr3+
2Cr3+ + 3O2· Û Cr2O3 H2O(g) Û 2H+ + O2·
(a) 2Cr3+ + 3O2· Û Cr2O3
Cr content of Fe3O4 increases;

outward flux of Fe decreases with time
Fe2+,Fe3+
O2· Voids form at Inner-outer scale interface
Fe3O4
Incorporation of Cr2O3 precipitates
FeFe2xCrxO4 into inner layer
Formation of internal precipitates of
Fe2+ Cr2O3
(b) Cr3+ Cr3+
Outward flux of Fe < inward flux of O2

2Fe3O4 + O2·+ 2vac+ = 3Fe2O3 O2· O2·
O2·
Fe2O3
Fe2O3
Fe O O2· Fe2+,Fe3+ Fe O 3 4
Fe2+,Fe3+
3 4
FeFe2xCrxO4 FeFe2xCrxO4
Fe2+
Cr3+ Cr3+
Fe2+ Cr3+
(c) (d) Formation of intermittent layer of Cr2O3
Figure 11 Schematic representation of the oxidation mechanisms observed in low alloy steels containing chromium.158
Path 1: outward diffusion of iron ions, which unlikely, unless the scale has significant, inter-
react at the oxide-steam interface with oxygen connected porosity.160
from the dissociation of steam to form fresh Path 5: which involves the dissociation of magne-
magnetite.156,157 tite at the interface with the inner and outer
Path 2: dissociation of steam at the oxide-steam layers,160,161 with the iron ions released diffusing
interface, and diffusion of oxygen ions inwards via to the oxide-gas interface to react and form new
defects in the oxygen lattice, or ‘‘hydrogen oxide, and the released oxygen diffusing to the
defects’’159; these react with iron ions diffusing alloy-oxide interface to react with Cr in the alloy
from the alloy to form new oxide at the interface to form internal particles of Cr2O3 (or Fe–Cr
between the inner and outer oxide layers. spinel).162
Path 3: dissociation of steam at the oxide-steam
This whole area of steam oxidation has been
interface, and diffusion of oxygen ions inwards to
reviewed recently by Wright and Dooley.162
the alloy-oxide interface, where they react with
chromium in the alloy to form discrete internal
particles of chromia.
1.22.8.2 Combustion Gases
Path 4: suggests that molecular steam is transported
through the scale, dissociating in the scale (in this A considerable amount of work has been carried out
case at the alloy-oxide interface) to form iron or investigating the corrosion of steels in the gases pro-
chromium oxides. This path is considered to be duced during the combustion of fossil fuel due to
extensive use of low-alloy steels as heat exchanger ashes containing chlorides and sulphates. Wright and
tubes in power generation. Combustion gases contain Krause168,169 predict corrosion by either combustion
many species, such as CO, CO2, SO2, SO3, H2S, and gases and/or molten deposits; erosion corrosion is
HCl, arising from elements within the fuel. The dif- also problematic due to flow rates and the presence
ferent combinations of operating temperature and of silica and alumina–silica particulates.
chemical stoichiometry of combustion reactions lead Fluidized-bed systems produce higher combustion
to many possible complex corrosion reactions. intensities at lower temperatures than combustion of
In coal-fired power stations, severe corrosion of the pulverized fuel in conventional fossil-fuel-fired boi-
steam-generating tubes in the furnace walls has been lers. The mineral matter for corrosion does not form
attributed to reducing conditions with high pCO, low fused salts and is not expected to release corrosive
pO2, high pHCl, and high local heat fluxes and flame species. Fluidized bed combustors can, therefore,
impingement.163,164 In areas of high corrosion, large burn lower grade, cheaper fuel in smaller plant with
concentrations of CO are almost always present, and better pollution control than traditional boilers.170
thick Fe3O4 scales, with islands and bands of FeS are Minchener171 reports that the bubble phase of
observed.163,164 Under the most severe reducing con- atmospheric fluidized bed combustion has a pO2 in
ditions, thick columnar scales of almost pure FeS have the range 2 101 to 2 102. Combustion in the
been observed.164 The high corrosion rate areas are dense phase is substoichiometric, with the pO2 as
also subject to a significant flux of sulfur-bearing car- low as 1013, and SO2 and SO3 present in the range
bonaceous material in the flame envelope163 which 500–5000 ppm. Low Cr–Mo steels show heavy scaling
may locally exacerbate reducing conditions and raise in these conditions, whereas 9–12% Cr steels show
the pH2S.164 Outside the high corrosion rate areas, the good resistance to sulfidation up to 650 C. Roberts
furnace atmospheres are found to be relatively O2-rich et al.,172 however, report that for pressurized fluidized-
and the corrosion scales comprise protective Fe3O4.163 bed combustion, ferritic steels at or below 9% Cr show
The emergence in recent times of biomass fuels heavy general corrosion above 540–560 C.
intended to reduce the reliance on fossil fuels for
steam power generation has introduced new chal-
1.22.8.3 Chemical Environments
lenges to the corrosion engineer. The use of biomass
fuels based on a wide range of agricultural waste The oil industry frequently uses stainless steels or
products, such as straw, forestry waste, sawdust, and exotic bonded alloys for the processing of crude oil in
purpose grown crops including coppiced willow and the temperature range 200–600 C. These materials
miscanthus have been investigated.165 Concerns over are very expensive and there is a strong economic
potential high rates of metal wastage have been raised incentive for finding cheaper alloys that are resistant
as the fireside environment is known to be particularly to H2S and some gaseous organic sulfides arising
aggressive since it often contains HCl and different from the S content of the crude oil.173
alkali salts which condense on heat exchanger surfaces Metal sulfides show the same type of predominant
promoting corrosion. Presently, it is necessary to oper- defects as metal oxides, that is, cations in Fe(1 x)S,
ate biomass and energy from waste plants at lower Cr(2 þ y)S3. The defect concentration in most sulfides
temperatures than traditional coal fired systems so is much higher than those in the corresponding
that tolerable corrosion rates prevail.166 However, pro- oxides, but the defect mobilities are only slightly
posed methods of modifying combustion gas chemis- higher. Thus, the higher diffusivities and growth rates
try where corrosive KCl is converted to the far less are determined by the higher defect concentrations.
corrosive K2SO4 by the addition of sulphur may enable Cr and Al only slightly reduce the corrosion rate,
higher service temperatures to be attained.167 and much higher Al is needed than that required for
Similarily, refuse incineration plants are now oxidation protection. Very protective scales are formed
based on the Waste-to-Energy concept, and it is only at a S pressure lower than that for the formation
necessary to optimize thermal transfer between com- of the base-metal sulfide.174
bustion gases and water vapour. A significant problem Mrowec et al.175 examined the resistance to high
with these installations concerns the fireside corro- temperature corrosion of Fe alloys with Cr contents
sion of the tubes. Combustion of waste generates between 0.35 and 74 at% Cr in 101 kPa S vapor. They
highly corrosive flue gas often containing HCl, SO2, found that the corrosion was parabolic, irrespective
NaCl, KCl, alkali and heavy metal chlorides and of the temperature or alloy composition, and noted
that sulfidation takes place at a rate five orders The authors propose three stages for scale develop-
of magnitude greater than oxidation at equivalent ment. First, a thin compact layer forms because of the
temperatures. At less than 2% Cr, the alloys formed penetration of S into the alloy with preferential for-
Fe(1 x)S growing by outward diffusion of Fe ions, with mation of sulfide from those metals with the highest
traces of FeCr2S4 near the metal core. affinity for S. Fe(1 x)S also forms because of the
Narita and Nishida176 examined the sulfidation of abundance of Fe in the alloy.
low Cr–Fe alloys at 700–900 C in 101 kPa of pure S. The outer layer then dissociates to release sulfur,
They found that the addition of small quantities of which dissolves in the grain boundaries of the alloy to
Cr significantly decreased the corrosion rate because form Cr and Al sulfides. The Fe released by this disso-
of the formation of FeCr2S4 in the inner reaches of ciation sulfidizes again at the interface between the
the scale. The scale comprised an outer FeS layer, two layers. The volume increase associated with the
with an inner layer of FeS, FeCr2S4, and Cr3S4 in conversion of metal to sulfide generates a mechani-
varying amounts depending upon the Cr content. cal stress that causes the outer layer to break up and
The corners of specimens corroded more rapidly permits permeation of S. This initiates a second stage,
than flat faces because of breakaway conditions. where growth of the scale is linear and comprises
At low Cr contents, the rate was increased, but above a porous outer layer, with FeS, Cr2S3, and Al2S3
4–6% Cr the parabolic rate constant decreased. evenly distributed, possibly as FeCr2S4, FeAl2S4, and
Above 7.4% Cr, an intermediate layer containing FeCr(x)Al(2 x)S4. S then diffuses through the pores
FeCr2S4 and varying amounts of Cr3S4, the propor- to the scale–metal interface. The third stage com-
tion of the latter increasing with increasing Cr con- prises the formation of an outer compact layer of
tent, formed between the inner and outer layers. Fe(1x)S and continued thickening of the inner layer.
In view of this potentially rapid degradation of Addition of Cr to Fe–Al alloys aids the formation of
Cr-containing steels by high temperature sulfidation Cr sulfides and Al2S3, which together markedly reduce
in petrochemical and coal gasification reactors, Al is the sulfidation rate.173 Between 2% and 5% Cr then,
much used in the Fe alloys for these applications.177–180 more than 3% Al is required to obtain protection. At
Al in Fe reduces the sulfidation rate in S2 vapor by up 9% Cr, however, only 1% Al is needed to give protec-
to two orders of magnitude181 because of the high tion, since the Cr is sufficiently active to lower the
thermodynamic stability of aluminum sulfide relative S potential seen by Al (secondary gettering). Thus, all
to iron sulfide, the low rate of sulfidation of Al com- Fe–Cr–Al alloys initially form FeS, Cr sulfides, and
pared with pure Fe and the large PB ratio of Al2S3 (3.7). Al2S3. The Cr and Al are then exposed to a much
The addition of 5% Al in Fe in 101 kPa S2 vapor lower pS2 at the scale–metal interface and Cr sulfides
between 500 and 700 C resulted in the rate of reaction and Al2S3 grow preferentially if their activity is
decreasing by a factor of 10. Paralinear kinetics were high enough. Ultimately, a protective layer of Al2S3 or
observed, with the inner layer of a duplex FeS scale Al2S3 þ Cr sulfides develops at the scale–metal inter-
containing a finely dispersed Al2S3 phase which acts as face and the reaction rate decreases substantially.
a diffusion barrier to Fe2þ migration. Increasing the Karlson et al.183 found, from on-site experience of
temperature to 800 C resulted in a rapid take off of the cement-producing plant, that corrosion of Fe surfaces
corrosion rate, with catastrophic corrosion rates above may occur in gases containing O2, SO2, and alkali
800 C due to the large volume of Al2S3 causing an chlorides such as NaCl and KCl between 300 and
increase in scale porosity. 500 C. They reported that the corrosion rates may
Condit et al.182 examined the sulfidation of several be extraordinarily high (5–10 mm month1) implying
Fe–Cr–Al alloys under a variety of sulfidizing condi- liquid-phase corrosion. Laboratory simulation of the
tions. They noted that, in the early stages of sulfida- plant conditions demonstrated the need for both SO2
tion, a thin compact inner layer forms that is high in and the alkali chloride in the environment. The prin-
Cr and Al. Subsequently, a thicker microcrystalline cipal corrosion reaction was found to be
outer layer forms with a uniform Fe, Cr, and Al com- 2Fe2 O3 þ 12½K; NaClðs; lÞ þ 12SO2 ðgÞ þ 9O2 ðgÞ
position. Formation of the outer compact layer was
favored by increasing pS2 and decreasing tempera- ! 4½K; Na3 FeðSO4 Þ3 ðs; lÞ þ 6Cl2 ðgÞ
ture, with the layer forming much more rapidly in A thermodynamic evaluation of this equation indi-
H2S than in pure S2. The sublayer disappeared more cated that the reaction could proceed with SO2 levels
or less rapidly depending upon the alloy composition. as low as 100 ppm.
1.22.8.4 CO/CO2 Fe3O4.188–190 Growth of this layer is controlled by

outward grain boundary diffusion of Fe ions.190–192
Failures of mild steel components in Magnox reactors
The inward counter-current of vacancies is ini-
in the United Kingdom and Italy after approximately
tially annihilated at the metal surface, but eventually
5 years of operation alerted the world to the potential
vacancy condensation at the scale–metal interface
for breakaway oxidation of low-alloy steels in CO/
gives decohesion, the scale develops microporosity,
CO2 environments.95 The CO2, 1% CO, 300 vppm
an inner layer grows within the space created by the
CH4, 250 vppm H2O, 100 vppm H2 environment
departing metal ions and a duplex scale forms.193–195
used in CAGRs was selected on the need to minimize
Then for Cr-containing steels, the protective scale
oxidation of the graphite reactor core and deposition
comprises an outer, coarse columnar-grained magne-
of C from the coolant gas.184 Corrosion rate tests of
tite layer and an inner, slightly porous, Cr-rich spinel
15 000–20 000 h, in the limited range of conditions
of fine (0.1 mm) equiaxed grains.70,194,196 As with low-
anticipated by the designers, showed that the maxi-
alloy steels in other environments, the Cr is not
mum reduction in corrosion rate of ferritic steels in
mobile in the scale, but is oxidized in situ.195,196 Stud-
CO2 at 600 C is realized at around the 9% Cr
ies have shown that the M3O4 spinel nucleates at
level.185–187 Therefore, 9Cr1Mo steel was chosen for
asperities on the surface and duplex growth is also
the evaporator and primary superheater sections of
known to be favored in the vicinity of inclusions and
the CAGR.185 However, in the late 1960s, Taylor
specimen corners.197 Once initiated, the inner layer
(reported in Holmes et al.) identified evidence for a
grows by inward diffusion of O, probably as CO2,
significant change in the corrosion mechanism for
down microfissures and micropores,17,136,196 in both
9% Cr steels at around 550 C. This change could
the lateral and vertical directions, until a complete
lead to rate inversion with increasing temperature
layer is obtained.197 The growth of this layer subse-
in steels containing 0.7–0.8% Si, or breakaway in
quently follows the parabolic rate law.
0.4–0.5% Si steels.
During this protective stage, decreasing the water
Because of their importance to the nuclear power
and CO content of the gas appears to decrease
generation industry, these observations initiated a
the rate constant of Cr-containing steels,184 but has
vast amount of research into the oxidation of low-
little effect on the rate constant of carbon steels.193,198
alloy steels in CO/CO2 environments. It is now clear
The rate constant of all steels has been found to
that low-alloy steels exhibit three types of behavior
decrease with decreasing temperature and increasing
when exposed to CO/CO2, that is, protective, transi-
Si content.184,185,195,199 Ferguson et al.198 have also
tional, and linear-breakaway (Figure 12), with the
reported that increasing the S content of carbon
time to breakaway and the breakaway rate being of
steels reduces the parabolic rate constant.
crucial importance in determining component life.
An increase in the Si content of the steel has been
For mild and low-alloy steels in CO2, the first scale
reported to give a significant increase in the duration
to form is a compact coarse columnar layer of
of the protective regime for carbon steel.188,200,201
A similar benefit has also been reported for Cr-
containing steels.193,198,202 Increases in the duration
of the protective regime are also realized with reduc-
Breakaway tions in the CO and H2O contents of the gas,198 in the
temperatures,185,193,198 and for carbon steels, in the
surface roughness,193 or for 9Cr steels, with increased
Weight gain
surface cold work. Robertson and Manning43 found that

w = at breakaway may also be delayed by some S-containing
gases.
Following the initial protective period, under cer-
Protective tain conditions of temperature, alloy and gas composi-
Transitional tion,185 the oxidation goes through a transitional stage
w = atb
into breakaway. Several authors have reported that
breakaway oxidation is initiated once the scale reaches
Time a critical thickness184,185 or weight gain70 and only
Figure 12 Schematic diagram of stages of low-alloy steel occurs below an initially protective duplex layer.43,195
oxidation in CO/CO2. Both single-layer protective and duplex-layer
protective scales have a PBR of 2.1, whereas breakaway For Fe in CO/CO2 at atmospheric pressure,
scales show a PBR of approximately 2.7, containing Surman210 found that if
around 23% porosity,195 and graphitic carbon grains pCO
(up to 6% by weight) between the oxide grains.193,198 103 < < 0:3
pCO þ pCO2
During this transitional period, and once full break-
away is established, growth occurs throughout the oxide then the oxidation is parabolic and very little C
and close to the scale–metal interface.203 Growth below deposition occurs. He concluded that magnetite is
the oxidant surface requires oxidant transport through not sufficiently catalytic to promote the Boudouard
the scale, but solid-state lattice diffusion is much too reaction unless CO > 10% and moisture is present
slow to account for the rates observed.17 Therefore, a and surmised that H2O promotes the formation of a
porous scale is required. Mechanisms for pore growth Boudouard catalyst. If the CO is greater than 0.4 in
have been postulated by several authors204–207 who have this expression, oxidation and carbon deposition
suggested that the transition from single-layer to duplex occur simultaneously at a linear rate.211
growth is due to the initiation of pores within the scale. German and Littlejohn193 have observed that
Then breakaway was thought to be caused by increased increasing the Si content of carbon steel reduces the
porosity giving unlimited access of oxidant to metal linear rate constant during breakaway and Banks and
surface. However Atkinson47 has shown single layer Lorimer208 have shown that Cr and Si reduce the
scales to be slightly porous. Robertson and Manning43 creep rate of Fe3O4, thus reducing the postbreakaway
have proposed that oxidant access is always available rate of Cr-containing steels. However, Si has been
and the type of scale that develops is dependent upon reported as having no effect on the postbreakaway
conditions at the scale–metal interface. Breakaway rate on 9Cr steels.184,199 Small changes to the CO and
occurs if space is created by continuous scale deforma- H2O content of the environment and the temperature
tion or creep.208 have no significant effect on the breakaway rate con-
The inward penetration of CO2 via cracks and stant, but a large reduction in the CO can cause
micropores gives reversion to protective kinetics,211 as does reducing
the CO2 pressure to atmospheric.212 Increasing the
CO switches breakaway on again.211
3Fe þ 4CO2 ¼ Fe3 O4 þ 4CO
2CO ¼ O2 þ 2C via the Boudouard reaction
References
136,196
at the scale–metal interface.
The carbon initially diffuses into the steel, but 1. Holmes, D. R.; Stringer, J. In Corrosion of Steels in CO2;
Holmes, D. R., Hill, R. B., Wyatt, L. M., Eds.; British
ultimately, the steel may become saturated with Nuclear Energy Society: London, 1974; p 165.
C.208 There is no detectable solubility of C in FeO, 2. Pilling, N. B.; Bedworth, R. E. J. Inst. Met. 1923, 29, 529.
Fe3O4, MnO, or Cr2O3,209 and C can only permeate 3. Tammann, G. Z. Anorg. Allgem. Chem. 1920, 111, 78.
4. Gulbransen, E. A.; Ruka, R. Trans. AIME 1950, 188, 1500.
through pores or faulty scales.112 Thus, the C then 5. Pinder, L. W. C.E.G.B Unclassified Report, MID/SSD/80/
deposits at the scale–metal interface and prevents 0050/R August 1980.
the formation of a coherent protective layer once it 6. Howe, C. I.; McEnany, B.; Scott, V. D. Corr. Sci. 1985,
25(3), 195.
reaches a critical activity.208 Carbides in the steel 7. Goursat, A. G.; Smeltzer, W. W. Oxid. Met. 1973, 6(2), 101.
have been found to be the preferred sites for break- 8. Kuroda, K.; Labun, P. A.; Welsch, G.; Mitchell, T. E.
away. These are either pre-existing carbides or car- Oxid. Met. 1983, 19(3/4), 117.
9. Koch, F.; Cohen, J. B. Acta Crystallogr. 1969, B 25, 275.
bides precipitated by C injection during oxidation.190 10. Cheetham, A. K.; Fender, B. E. G.; Taylor, R. I. J. Phys. C.
Precarburized or graphite-painted steel breaks 1971, V, 2160.
away rapidly, as do thin foils, because of the smaller 11. Catlow, C. R. A.; Mackrodt, W. C.; Norgett, M. J.;
Stoneham, A. M. Philos. Mag. A. 1979, 40(2), 161.
C sink available.190 Pritchard et al.89 reported that the 12. Chen, W. K.; Peterson, N. L. J. Phys. Chem. Solids 1975,
proportion of C in the scale increased with scale 36, 1097.
thickness and water content of the gas and was higher 13. Wagner, C. Z. Phys. Chem. 1933, B621, 25.
14. Wagner, C. Atom Movements, ASM: Cleavland, 1951,
in breakaway oxide. For 9% Cr steel, breakaway p 153.
oxidation is associated with heavy carburization of 15. Atkinson, A. Rev. Mod. Phys. 1985, 57, 437.
the metal and C deposition within the oxide, with 16. Atkinson, A.; Taylor, R. I. J. Phys. Chem. Solids 1985, 46,
469.
preferential breakaway occurring at corners and 17. Atkinson, A.; Taylor, R. I. High Temperature – High
edges.196 Pressure 1982, 14, 571.
18. Dieckmann, R.; Kohne, M. Phys Chem. 1983, 87, 495. 65. Fox, P.; Tatlock, G. J. Mater. Sci. Technol. 1988, 4, 439.
19. Garnaud, G.; Rapp, R. A. Oxid. Met. 1977, 11, 193. 66. Tomaszewicz, P.; Wallwork, G. R. Oxid. Met. 1983,
20. Channing, D. A.; Graham, M. J. Corr. Sci. 1972, 12, 271. 20(3–4), 75–109.
21. Channing, D. A.; Dickerson, S. M.; Graham, M. J. Corr. 67. Wagner, J. B. In Defects and Transport in Oxides;
Sci. 1973, 13, 933. Smeltzer, M. S., Jaffe, R. I., Eds.; Plenum Press:
22. Francis, R.; Lees, D. G. Corr. Sci. 1976, 16, 847. New York, 1974; p 283.
23. Rahmel, A. Werkst. Korros. 1965, 16(10), 837. 68. Seybolt, A. U. Trans. Met. Soc. AIME 1968, 242, 752.
24. Rahmel, A. Korrosion 1966, 18, 41. 69. Rahmel, A.; Tobolski, J. Werkst. Korros. 1965, 16(8), 662.
25. Eubanks, K. G.; Moore, D. G.; Pennington, W. A. 70. Robertson, J.; Manning, M. I. Mater. Sci. Technol. 1989,
J. Electochem. Soc. 1962, 109, 382. 5(8), 741–753.
26. Svedung, I.; Hammar, B.; Vannerberg, N. G. OXid. Met. 71. Darken, L. S. Trans. AIME 1942, 150, 157–171.
1973, 6(1), 21. 72. Tuck, C. W. Corros. Sci. 1965, 5(9), 631–634, IN3–IN7,
27. Caplan, D. Corr.Sci. 1966, 6, 509. 635–643.
28. von Fraunhofer, J. A.; Pickup, G. A. Corr. Sci. 1970, 10, 253. 73. Svedung, I.; Vannerberg, N. G. Corros. Sci. 1974, 14(6),
29. Janssen, S.; Lehtinen, B. Metallurgie 1967, 7, 61. 391–399.
30. Caplan, D.; Cohen, M. Corr. Sci. 1966, 6, 321. 74. Wood, G. C.; Richardson, J. A.; Hobby, M. G.; Boustead, J.
31. Price, W. R. Corr. Sci. 1967, 7, 473. Corros. Sci. 1969, 9(9), 659–668, IN7–IN12, 669–671.
32. Pinder, L. W. C.E.G.B Unclassified Report, MID/SSD/80/ 75. Rochet, F.; Rigo, S.; Froment, M.; D’Anterroches, C.;
0057/R August 1980. Maillot, C.; Roulet, H.; Dufour, G. Philos. Mag. B. 1986,
33. Stott, F. H. Mat. Sci. Tech. 1989, 5(8), 743. 55(2), 309.
34. Atkinson, A. Mat. Sci. Tech. 1988, 4(12), 1046. 76. Adachi, T.; Meier, G. H. Oxid. Met. 1987, 27(5–6),
35. Dunitz, J. D.; Orgel, L. E. J. Phys. Chem. Solids 1957, 3, 347–366.
318. 77. Atkinson, A. Corros. Sci. 1982, 22(2), 87–102.
36. Azaroff, L. V. J. Phys. 1969, 32, Part 9, 1658. 78. Logani, R. C.; Smeltzer, W. W. Oxid. Met. 1971, 3(3),
37. Cox, M. G. C.; MacEnany, B.; Scott, V. D. Phil. Mag. 279–290.
1972, 26, 839. 79. Logani, R.; Smeltzer, W. W. Oxid. Met. 1969, 1(1), 3–21.
38. Hodge, J. D. Electrochem. Soc. 1978, 125(2), 55c. 80. Logani, R. C.; Smeltzer, W. W. Oxid. Met. 1971, 3(1),
39. Rahmel, A. Z. Electrochem. 1962, 66(4), 363. 15–32.
40. Moreau, J. Compte. Rendu. 1953, 236, 85. 81. Yang, C.-H.; Lin, S.-N.; Chen, C.-H.; Tsai, W.-T. Oxid.
41. Surman, P. L.; Castle, J. E. Corr. Sci. 1969, 9, 771. Met. 2009, 72(3), 145–157.
42. Atkinson, A. In Oxidation of Metals and Associated Mass 82. Lashin, A.-R.; Schneeweiss, O.; Houbaert, Y. Corros. Sci.
Transport; Dayananda, M. A., et al. Ed.; The Metallurgical 2008, 50(9), 2580–2587.
Society of the AIME: Warrendale, PA, 1987; p 29. 83. Ishitsuka, T.; Inoue, Y.; Ogawa, H. Oxid. Met. 2004,
43. Robertson, J.; Manning, M. I. Mat. Sci. Tech. 1988, 4, 61(1–2), 125–142.
1064. 84. Jackson, P. R. S.; Wallwork, G. R. Oxid. Met. 1983,
44. Bruckman, A.; Emmerich, R.; Mrowec, S. Oxid. Met. 20(1–2), 1–17.
1972, 5(2), 137. 85. Huczkowski, P.; Ertl, S.; Piron-Abellan, J.; Christiansen, N.;
45. Fromhold, A., Jr.; Sato, N. Oxid. Met. 1981, 16(3/4), 203. Höfler, T.; Shemet, V.; Singheiser, L.; Quadakkers, W. J.
46. Harrison, P. L. Oxid. Met. 1984, 22(1/2), 35. Mater. High Temp. 2005, 22(3–4), 253–262.
47. Atkinson, A. Philos. Mag. B 1987, 55, 637. 86. McAdam, G.; Young, D. J. Oxid. Met. 1987, 28(3–4),
48. Whittle, D. P. Acta Met. 1967, 15, 1421. 165–181.
49. Wagner, C. Corr. Sci. 1965, 5, 751. 87. Nishida, K.; Narita, T.; Tani, T.; Sasaki, G. Oxid. Met.
50. Runk, R. B.; Kim, H. J. Oxid. Met. 1970, 2(3), 285. 1980, 14(1), 65–83.
51. Kim, H. J.; Runk, R. B. Oxid. Met. 1970, 2(3), 307. 88. Dewanckel, B.; Leclercq, D.; Dixmier, J.; Holmes, D. R.
52. Nosek, E.; Werber, T. Oxid. Met. 1986, 25(3/4), 121. Corrosion of unalloyed or low-alloyed steels in
53. Bohnenkamp, V. J.; Engell, H. J. Arch. Eisenhuettenw. pressurized carbon dioxide, 1974; p 3.
1962, 33, 359. 89. Pritchard, A. M.; Antill, J. E.; Cottell, K. R. J.; Peakall, K. A.;
54. Caplan, D.; Sproule, G. I.; Hussey, R. J.; Graham, M. J. Truswell, A. E. In Corrosion of Steels in CO2; Holmes, D. R.,
Oxid. Met. 1979, 13, 255. Hill, R. B., Wyatt, L. M., Eds.; British Nuclear Energy
55. Caplan, D.; Sproule, G. I.; Hussey, R. J.; Graham, M. J. Society: London, 1974; p 73.
Oxid. Met. 1978, 12, 67. 90. Program on Technology Innovation: Oxide Growth
56. Malik, A. U. Oxid. Met. 1985, 25(5/6), 223. and Exfoliation on Alloys Exposed to Steam; EPRI,
57. Prescott, R.; Graham, M. J. Oxid. Met. 1992, 38(1–2), Palo Alto, CA: 2007. 1013666.
73–87. 91. Morinaga, M.; Murata, Y.; Hashizume, R.; Sawaragi, Y.
58. Tomaszewicz, P.; Wallwork, G. R. Oxid. Met. 1983, ISIJ Int. 2001, 41(3), 314–316.
19(5–6), 165–185. 92. Kutsumi, H.; Itagaki, T.; Abe, F. Tetsu-To-Hagane/J. Iron
59. Tomaszewicz, P.; Wallwork, G. R. International Corrosion Steel Inst. Jpn. 2002, 88(9), 520–525.
Conference Series; 1983. 93. Greeff, A. P.; Louw, C. W.; Terblans, J. J.; Swart, H. C.
60. Boggs, W. E. J. Electrochem. Soc.: Solid State Sci. 1971, Corros. Sci. 2000, 42(6), 991–1004.
118, 906. 94. Vannerberg, N. G.; Svedung, I. Corros. Sci. 1971, 11(12),
61. Ahmed, H. A.; Smeltzer, W. W. J. Electrochem. Soc. 915–927.
1986, 133(1), 212–216. 95. Donati; Garaud. In Corrosion of Steels in CO2
62. Pons, M.; Caillet, M.; Galerie, A. Corros. Sci. 1982, 22(3), Holmes, D. R., Hill, R. B., Wyatt, L. M., Eds.; British
239–249. Nuclear Energy Society: London, 1974; p 28.
63. Smith, P. J.; Beauprie, R. M.; Smeltzer, W. W.; 96. Menzies, L. A.; Tomlinson, W. J. JISI 1958, 204, 1239.
Stevanovic, D. V.; Thompson, D. A. Oxid. Met. 1987, 97. Foley, R. T. J. Electrochem. Soc. 1962, 109(4), 278–284.
28(5–6), 259–276. 98. Menzies, I. A.; Lubkiewicz, J. Oxid. Met. 1971, 3(1),
64. Ahmed, H. A.; Underbill, R. P.; Smeltzer, W. W.; Brett, M. E.; 41–58.
Graham, M. J. Oxid. Met. 1987, 28(5–6), 347–351. 99. Dalvi, A. D.; Coates, D. E. Oxid. Met. 1972, 5(2), 113–135.
100. Rhys-Jones, T. N.; Grabke, H. J.; Kudielka, H. Corros. 136. Hancock, P.; Hurst, R. C. In Corrosion of Steels in CO2;
Sci. 1987, 27(1), 49–63, 65–73. Holmes, D. R., Hill, R. B., Wyatt, L. M., Eds.; British
101. Wood, G. C. Oxid. Met. 1970, 2(1), 11–57. Nuclear Energy Society: London, 1974; p 320.
102. Yearian, H. J.; Randell, E. C.; Longo, T. A. Corrosion 137. Ward, G.; Hockenhull, B. S.; Hancock, P. Metall. Trans.
1956, 12, 515t–525t. 1974, 5(6), 1451–1455.
103. Douglass, D. L.; Gesmundo, F.; de Asmundis, C. Oxid. 138. Forrest, J. E.; Bell, P. S. Relationship between the
Met. 1986, 25(3–4), 235–268. deformation of substrate and scale in the mechanism of
104. Khanna, A. S.; Gnanamoorthy, J. B. Oxid. Met. 1985, multilayer oxide formation on low alloy ferritic steels;
23(1–2), 17–33. Commission of the European Communities, (Report)
105. Hossain, M. K. Corros. Sci. 1979, 19(7), 1031–1045. EUR, 1981.
106. Caplan, D.; Sproule, G. I. Oxid. Met. 1975, 9(5), 459–472. 139. Rolls, R.; Nematollahi, M. Oxid. Met. 1983, 20(1–2), 19–35.
107. Rahmel, A.; Jarger, W.; Becker, K. Arch. Eisenhuttenwes 140. Simms, N. J.; Little, J. A. Mater. Sci. Technol. 1988, 4(12),
1959, 30, 351–359. 1133–1139.
108. Tsipas, D. N.; Noguera, H.; Rus, J. Mater. Chem. Phys. 141. Pinder, L. W. C.E.G.B. Unclassified Report SSD/MID/
1987, 18(3), 295–303. R58/77, November 1977.
109. Suwattananont, N.; Petrova, R. Oxid. Met. 2008, 70(5), 142. Wright, I. G.; Pint, B. A. In NACE 2002; 2002.
307–315. Paper no. 02377.
110. Rowley, P. N.; Brydson, R.; Little, J.; Saunders, S. R. J.; 143. Jonsson, T.; Pujilaksono, B.; Fuchs, A.; Svensson, J. E.;
Sauer, H.; Engel, W. Oxid. Met. 1991, 35(5), 375–395. Johansson, L. G.; Halvarsson, M. In 7th International
111. Hussey, R. J.; Mitchell, D. F.; Graham, M. J. Werkst. Symposium on High Temperature Corrosion and
Korros. 1987, 38(10), 575–583. Protection of Materials; Trans Tech Publications Ltd.:
112. Wolf, I.; Grabke, H. J.; Schmidt, P. Oxid. Met. 1988, Les Embiez, France, 2008.
29(3–4), 289–306. 144. Yang, Z.; Xia, G.-G.; Walker, M. S.; Wang, C.-M.;
113. Sroda, S.; Baxter, D.; Arponen, M. Mater. Corros. 2005, Stevenson, J. W.; Singh, P. Int. J. Hydrogen Energy 2007,
56(11), 791–795. 32(16), 3770–3777.
114. Fry, A.; Banks, J.; Osgerby, S. The influence of heat flux 145. Nakagawa, K.; Matsunaga, Y.; Yanagisawa, T. Mater.
on the oxidation of boiler steels. Parson Conference; High Temp. 2001, 18, 51–56.
2007. 146. Nakagawa, K.; Matsunaga, Y.; Yanagisawa, T. Mater.
115. Fry, A. T.; Roberts, S.; Chapman, L. Measuring heat High Temp. 2003, 20, 67–73.
transfer through steam grown oxide scales. NPL 147. Fujii, C. T.; Meussner, R. A. J. Electrochem. Soc. 1964,
Measurement Note - DEPC(MN)043, 2007; p 9. 111(11), 1215–1221.
116. Trindade, V. B.; Borin, R.; Hanjari, B. Z.; Yang, S.; 148. Sánchez, L.; Hierro, M.; Pérez, F. Oxid. Met. 2009, 71(3),
Krupp, U.; Christ, H. J. Mater. Res. 2005, 8(4), 365–369. 173–186.
117. Ostwald, C.; Grabke, H. J. Corros. Sci. 2004, 46(5), 149. Itagaki, T.; Kutsumi, H.; Haruyama, H.; Igarashi, M.;
1113–1127. Abe, F. Corrosion 2005, 61(4), 307–316.
118. Baxter, D. J.; Natesan, K. Rev. High Temp. Mater. 1983, 150. Czyrska-Filemonowicz, A.; Zielińska-Lipiec, A.; Ennis, P. J.
5(3–4), 149–250. J. Achiev. Mater. Manufact. Eng. 2006, 19(2), 43–48.
119. Schutze, M. Mater. Sci. Technol. 1988, 4, 407–414. 151. Tuurna, S.; Yli-Olli, S. Heikinheimo. L. Mater. Sci. Forum.;
120. Stringer, J. Corros. Sci. 1970, 10(7), 513–543. 2008.
121. Evans, H. E.; Lobb, R. C. Corros. Sci. 1984, 24(3), 152. Nakai, M.; Nagai, K.; Murata, Y.; Morinaga, M. Corros.
209–222. Sci. 2006, 48(11), 3869–3885.
122. Kubaschewski, O.; Hopkins, B. E. Oxidation of 153. Fujikawa, H.; Otsuka, N. Nippon Kagaku Kaishi/Chemical
Metals and Alloys Butterworths: London, 1967; Vol. 50, Society of Japan – Chem. Indust. Chemis. J. 1998, 1,
pp 83–249. 51–52.
123. Manning, M. I.; Metcalfe, E. In EROCOR ’77, Eur Congr 154. Griess, J. C.; Maxwell, W. A. The long-term oxidation of
on Met Corros, 92nd Event of the Eur Fed of Corros. selected alloys in superheated steam at 482 and 538 C.;
1977; pp 121–127. 1981.
124. Barbehön, J.; Rahmel, A.; Schütze, M. Oxid. Met. 1988, 155. Dooley, R. B.; Paterson, S. J.; Wright, I. G. Presented at a
30(1–2), 85–94. EPRI-NPL Workshop on Scale Growth and Exfoliation in
125. Christi, W.; Rahmel, A.; Schütze, M. Oxid. Met. 1989, Steam Plant.; National Physical Lab, 2003.
31(1–2), 1–34. 156. Gala, A.; Grabke, H. J. Archiv. Fur. Das.
126. Whittle, D. P. Oxid. Met. 1972, 4(3), 171–179. Eisenhuttenwesen 1972, 43(6), 463.
127. Deadmore, D. L.; Lowell, C. E. Oxid. Met. 1977, 11(2), 157. Grabke, H. J.; Viefhaus, H. Berichte Der
91–106. Bunsen-Gesellschaft – Phys. Chem. Chem. Phys. 1980,
128. Hsueh, C. H.; Evans, A. G. J. Appl. Phys. 1983, 54(11), 84(2), 152–159.
6672–6686. 158. Wright, I. G.; Wood, G. C. Dilute Fe–Cr alloys; 1969.
129. Douglass, D. L. Oxidation of Metals and Alloys; 159. Norby, T. In 3rd International Symposium on High
American Society of Metals: Metals Park, OH, 1971; Temperature Corrosion and Protection of Materials;
p 1137. Editions Physique: Les Embiez, France, 1992.
130. Mitchell, T. E.; Voss, D. A.; Butler, E. P. J. Mater. Sci. 160. Mayer, P.; Manolescu, A. V. In International Corrosion
1982, 17(6), 1825–1833. Conference Series; 1983.
131. Manning, M. I. Corros. Sci. 1981, 21(4), 301–316. 161. Stott, F. H.; Wright, I. G.; Hodgkiess, T.; Wood, G. C.
132. Evans, A. G.; Crumley, G. B.; Demaray, R. E. Oxid. Met. Oxid. Met. 1977, 11(3), 141–150.
1983, 20(5–6), 193–216. 162. Wright, I. G.; Dooley, R. B. A review of the oxidation
133. Norin, A. Oxid. Met. 1975, 9(3), 259–274. behaviour of structural alloys in steam; in press.
134. Appleby, W. K.; Tylecote, R. F. Corros. Sci. 1970, 10(5), 163. Clarke, F.; Morris, C. W. In Corrosion Resistant Materials
325–341. for Coal Combustion Systems; Meadowcroft, D. B.,
135. Jha, B. B.; Raj, B.; Khanna, A. S. Oxid. Met. 1986, Manning, M. I., Eds.; Applied Science Publishers:
26(3–4), 263–273. London, 1983; p 47.
164. Lees, D. G.; Whitehead, M. E. In Corrosion Resistant 190. Gibbs, G. B.; Pendlebury, R. E.; Wooton, M. R. In
Materials for Coal Combustion Systems; Corrosion of Steels in CO2; Holmes, D. R., Hill, R. B.,
Meadowcroft, D. B., Manning, M. I., Eds.; Applied Wyatt, L. M., Eds.; British Nuclear Energy Society:
Science Publishers: London, 1983; p 63. London, 1974; p 59.
165. Coleman, K. E.; Simms, N. J.; Kilgallon, P. J.; Oakey, J. E. 191. Atkinson, A.; Taylor, R. I. High Temp. High Pressure
7th International Symposium on High Temperature 1982, 14, 571.
Corrosion and Protection of Materials; Trans Tech 192. Cox, M. G. C.; McEnaney, B.; Scott, V. D. High Temp.
Publications Ltd.: Les Embiez, France, 2008. High Pressure 1982, 14, 247.
166. Johansson, L. G.; Svensson, J. E.; Skog, E.; Pettersson, J.; 193. German, P. A.; Littlejohn, A. C. In Corrosion of Steels in
Pettersson, C.; Folkeson, N.; Asteman, H.; Jonsson, T.; CO2; Holmes, D. R., Hill, R. B., Wyatt, L. M., Eds.; British
Halvarsson, M. J. Iron Steel Res. Int. 2007, 14(5 Suppl. 1), Nuclear Energy Society: London, 1974; p 1.
35–39. 194. Hussey, R. J.; Sproule, G. I.; Caplan, D.; Graham, M. J.
167. Henderson, P.; Szakálos, P.; Pettersson, R.; Andersson, C.; Oxid. Met. 1977, 11(2), 65–79.
Högberg, J. Mater. Corros. 2006, 57(2), 128–134. 195. Gibbs, G. B.; Pendlebury, R. E.; Wooton, M. R. In
168. Krause, H. H. Mater. Perform. 1994, 33(3), 63–69. Corrosion of Steels in CO2; Holmes, D. R., Hill, R. B.,
169. Krause, H. H.; Wright, L. G. Mater. Perform. 1996, 35(1), Wyatt, L. M., Eds.; British Nuclear Energy Society:
46–53. London, 1974; p 59.
170. Perkins, R. A. In Corrosion Resistant Materials for Coal 196. Harrison, P. L.; Dooley, R. B.; Lister, S. K.;
Combustion Systems; Meadowcroft, D. B., Manning, M. I., Meadowcroft, D. B.; Nolan, P. J.; Pendlebury, R. E.;
Eds.; Applied Science Publishers: London, 1983; p 219. Surman, P. L.; Wooton, M. R. In Corrosion of Steels in
171. Minchener, A. J. In Corrosion Resistant Materials for CO2; Holmes, D. R., Hill, R. B., Wyatt, L. M., Eds.; British
Coal Combustion Systems; Meadowcroft, D. B., Nuclear Energy Society: London, 1974; p 220.
Manning, M. I., Eds.; Applied Science Publishers: 197. Cox, M. G. C.; McEnaney, B.; Scott, V. C. In Corrosion of
London, 1983; p 299. Steels in CO2; Holmes, D. R., Hill, R. B., Wyatt, L. M., Eds.;
172. Roberts, A. G.; Raven, P.; Lane, G.; Stringer, J. In British Nuclear Energy Society: London, 1974; p 247.
Corrosion Resistant Materials for Coal Combustion 198. Ferguson, J. M.; Garrett, F.; Lloyd, B. In Corrosion of
Systems; Meadowcroft, D. B., Manning, M. I., Eds.; Applied Steels in CO2; Holmes, D. R., Hill, R. B., Wyatt, L. M.,
Science Publishers: London, 1983; p 323. Eds.; British Nuclear Energy Society: London, 1974; p 15.
173. Zelankof, P. D.; Simkovich, G. Oxid. Met. 1974, 8(5), 199. Grandison, N. O.; Facer, R. I. In Corrosion of Steels in
343–360. CO2; Holmes, D. R., Hill, R. B., Wyatt, L. M., Eds.; British
174. Mrowec, S.; Przybylski, K. Oxid. Met. 1985, 23(3–4), Nuclear Energy Society: London, 1974; p 208.
107–139. 200. Ferguson, J. M.; Garrett, J. C. P.; Lloyd, B.
175. Mrowec, S.; Walec, T.; Werber, T. Oxid. Met. 1969, 1(1), In Corrosion of Steels in CO2; Holmes, D. R., Hill, R. B.,
93–120. Wyatt, L. M., Eds.; British Nuclear Energy Society:
176. Narita, T.; Nishida, K. Oxid. Met. 1973, 6(3), 181–196. London, 1974; p 15.
177. Setterlund, R. B.; Prescott, G. R. Corrosion 1961, 17(6), 201. Rowlands, P. C.; Garrett, J. C. P.; Popple, L. A.;
277–282. Whittaker, A.; Hoaksey, A. Nucl. Energy 1986, 25(5), 267.
178. Backensto, E.; Prior, B.; Sjoberg, J. E. Corros. Sci. 1962, 202. Rowlands, P. C.; Garrett, J. C. P.; Popple, L. A.;
18, 253–258. Whittaker, A.; Hoaksey, A. Nucl. Energy 1986, 25(5),
179. Malinowski, E. Metal 1962, 94(4). 267–275.
180. Burns, F. J. Corrosion 1969, 25, 119. 203. Gleave, C.; Calvert, J. M.; Lees, D. G.; Rowlands, P. C.
181. Strafford, K. N.; Manifold, R. Oxid. Met. 1969, 1(2), Proc. R. Soc. London Ser. A. 1982, 379, 409.
221–240. 204. Gibbs, G. B. Oxid. Met. 1973, 7(3), 173–184.
182. Condit, R. H.; Hobbins, R. R.; Birchenall, C. E. Oxid. Met. 205. Gibbs, G. B.; Hales, R. Corros. Sci. 1977, 17(6), 487–497,
1974, 8(6), 409–455. 499–507.
183. Karlsson, A.; Mãller, P. J.; Johansen, V. Corros. Sci. 206. Kofstad, P. Oxid. Met. 1985, 24(5–6), 265–276.
1990, 30(2–3), 153–158. 207. Mrowec, S. Corros. Sci. 1967, 7(9), 563–578.
184. Rowlands, P. C.; Garrett, J. C. P.; Garrett, F.; Hicks, F. G.; 208. Banks, P.; Lorimer, G. W. In Materials to Supply the
Lister, B. K.; Lloyd, B.; Twelves, J. A. In Corrosion of Steels Energy Demand; Hawbolt, E. B., Mitchell, A., Eds.;
in CO2; Holmes, D. R., Hill, R. B., Wyatt, L. M., Eds.; British American Institute of mechanical Engineers: New York,
Nuclear Energy Society: London, 1974; p 193. 1982; p 231.
185. Holmes, D. R.; Mortimer, D.; Newell, J. Proc. BNES 1974, 209. Wolf, I.; Grabke, H. J. Solid State Commun. 1985, 54(1),
p 151. 5–10.
186. Newell, J. E. Nucl. Energy Int. 1972, 17, 637. 210. Surman, P. L. Corros. Sci. 1973, 13(11), 825–830,
187. Taylor, J. W.; Trotsenberg, P. V. In Corrosion of Steels in IN1–IN5, 831–832.
CO2; Holmes, D. R., Hill, R. B., Wyatt, L. M., Eds.; British 211. Surman, P. L.; Brown, A. M. In Corrosion of Steels in CO2;
Nuclear Energy Society: London, 1974; p 180. Holmes, D. R., Hill, R. B., Wyatt, L. M., Eds.; British
188. German, P. A.; Littlejohn, A. C. In Corrosion of Steels in Nuclear Energy Society: London, 1974; p 85.
CO2; Holmes, D. R., Hill, R. B., Wyatt,, L. M., Eds.; British 212. Goodison, D.; Harris, R. J.; Goldenbaum, P. In British
Nuclear Energy Society: London, 1974; p 1. Joint Corrosion Group Symposium on Metal–Gas
189. Hussey, R. J.; Sproule, G. I.; Caplan, D.; Graham, M. J. Reactions in Atmospheres Containing CO2; London,
Oxid. Met. 1977, 11, 65. March 1967.
1.23 High Temperature Corrosion of Chromia-forming Iron,
Nickel and Cobalt-base Alloys
A. Galerie
SIMaP, Grenoble INP/CNRS/UJF, Domaine Universitaire, BP 75, 39402 Saint Martin d’Héres, France

1.23.2 Underlying Principles of High Temperature Corrosion Protection by Chromium 584
1.23.2.1 Various Types of Chromium-Containing Alloys 584
1.23.2.2 Growth of Cr-Containing Oxides 587
1.23.2.2.1 Chromia growth 588
1.23.2.2.2 Growth of spinel phases 588
1.23.2.3 Influence of Alloy Minor Elements 589
1.23.3 High Temperature Oxidation Behavior of Chromium Alloys 589
1.23.3.1 Oxidation of Martensitic and Ferritic Stainless Steels 589
1.23.3.2 Oxidation of Austenitic Stainless Steels (AISI 3XX Grades) 591
1.23.3.3 High Temperature Oxidation of Ni- and Co-Containing Alloys 592
1.23.3.3.1 Ni–Cr alloys 592
1.23.3.3.2 Ni–Fe–Cr alloys 593
1.23.3.3.3 Co–Cr alloys 593
1.23.3.3.4 Various comparisons between alloy grades 594
References 605
and less often on cobalt, they can offer mechanical

Abbreviations properties which can be adjusted to the situa-
bcc body-centered cubic
tions they will encounter, with additional resis-
fcc face-centered cubic
tance to chemical attack by the various gaseous
AISI American Iron and Steel Institute
atmospheres.
CTE coefficient of thermal expansion
Over the years, there have been arguments about
DIN Deutsches Institut für Normung
who invented stainless steels, the first chromium-con-
EN European norms
taining alloys. It seems that Pierre Berthier in France
MICF mechanically-induced chemical failure
was the first to report 1820 that addition of sufficient
REE reactive element effect
chromium to iron suppressed attack by water and
SOFC solid oxide fuel cell
acids. However, he could not, at that time, produce
UNS Unified Numbering System
mechanically resistant steels. Harry Brearley of Shef-
field was the first to produce a 13% Cr–0.24%
C martensitic stainless steel in 1913 and to observe
Symbols its excellent resistance to corrosion. It is amazing to
a bcc ferritic solid solution note that this discovery was made as Brearley looked
g fcc austenitic solid solution s for better erosion (and not corrosion) resistance of gun
s brittle intermetallic sigma phase barrels. More interesting is that his work was made in a
technical laboratory cofinanced by two main Sheffield
steel companies of the day. It seems however that the
Krupp German company in Essen (first austenitic
1.23.1 Introduction stainless steel patent: 1912) as well as Swedish steel-
makers produced stainless steels at this time. In fact, all
In industrial applications where high temperatures industrial countries could produce such materials after
are required or produced, chromium-containing World War I. In 1924, grade 304 (18Cr–8Ni) was
alloys are widely used. Based on iron and/or nickel, produced at Sheffield by W. H. Hatfield.
583
Concerning nickel based alloys, they were devel- in air; it is generally considered that this maximum
oped a decade later, for example, by Haynes Interna- service temperature must be reduced by at least
tional, USA. The term Superalloy appeared just after 25 C in cycling conditions, sometimes by larger
World War II for designing new alloys for aircraft values in more aggressive atmospheres (metal dusting
engine applications. or sulfidizing environments).
Cobalt based alloys were also first produced at the At present, simple scientific principles can explain
beginning of the twentieth century, for example, by the results of Hatfield; they are recalled in the fol-
Stellite in 1907 in Canada. lowing section.
The first published systematic comparative study
of the behavior of numerous chromium-containing
alloys by Hatfield appeared in the Journal of the Iron 1.23.2 Underlying Principles of
and Steel Institute in 1927.1 High Temperature Corrosion
The same author published the first quantified Protection by Chromium
values in 1928, using a ‘scaling index’ (SI) defined as
the temperature giving a certain weight gain (in g m2) High temperature alloys must be protected in service
at the end of a seven 6-h cycle oxidation run.2 The by an external slow-growing oxide scale. In the case
higher the SI, the better the behavior of the alloys. His of chromium-containing alloys, this protective scale
results for SI10 (10 g m2), SI20 (20 g m2), and SI100 is based on chromia Cr2O3.
(100 g m2) are presented in Table 1.
Nowadays, many results can be found either in
1.23.2.1 Various Types of Chromium-
the scientific literature or in technical notes provi-
Containing Alloys
ded by the steelmakers, giving the maximum service
temperature acceptable for all commercial alloys. As may be observed on Fe–Cr, Ni–Cr, and Co–Cr
Figure 1 presents such data for isothermal oxidation binary system phase diagrams (Figures 2–4),
Table 1 Scaling indexes ( C) for a series of chromium-containing alloys
Cr Ni Si Mn Others SI10 ( C) SI20 ( C) SI100( C)
0.1 0.1 0.25 0.82 <400 600 620

3.1 0.3 0.26 0.55 Mo <500 600 640
3.1 0.9 0.22 0.61 Mo, V <500 600 615
8.5 0.3 3.37 0.50 950 960 1010
9.0 0.4 0.19 0.48 Mo <500 640 695
12.9 0.3 0.21 0.37 800 820 835
12.8 0.5 0.31 0.31 790 800 825
16.6 2.5 0.29 0.41 750 790 850
12.3 0.3 0.19 0.46 Mo 800 810 820
10.9 0.6 0.42 0.86 Mo, V 800 805 820
10.6 0.8 0.43 1.06 Mo, V, Nb 800 810 815
11.4 2.6 0.48 0.73 Mo, V, Nb 750 760 770
10.7 0.6 0.48 0.89 Mo, V, Nb, Co 800 810 825
15.9 4.0 0.55 0.26 Nb, Cu, 780 807 818
13.9 5.5 0.32 0.78 Mo, Nb, Cu 750 770 830
16.0 0.3 0.42 0.37 850 860 875
20.5 0.2 0.39 0.73 900 1000 1075
28.9 1.8 1.49 1.34 N 900 1120 1160
13.1 0.1 0.55 0.85 Al 1060 1100 1125
18.7 10.0 0.51 1.04 860 880 915
18.3 9.0 0.76 0.80 Ti 820 860 890
18.3 10.0 0.40 0.76 Nb 820 860 890
16.4 10.5 0.19 1.69 Mo 700 820 845
22.0 14.4 0.47 1.17 980 1060 1070
25.2 20.2 0.48 1.49 1030 1090 1180
20.3 33.5 0.60 1.12 Ti, Al 970 1010 1140
Oxidizing atmosphere: combustion gas with 50% excess air.

High Temperature Corrosion of Chromia-forming Iron, Nickel and Cobalt-base Alloys 585
HA
1.4710
409 1.4713
430Ti A286 242
321 1.4729
HF 12SR 1.4740
304 316
429 436 441 444
800H 810 Multimet R41
309 309S 810 625
1.4849 1.4865
310S 153MA Manaurite 900
800HT 811 617 HD HN HP HQ HW Manaurite XM

230 446 310 314 330 253MA 600 HC HL HX Nimon75 XTM
353MA GX 40NiCrNb45-35
331 601 602CA 50W GX40NiCrCoW35-25-15 2.4778
35K
MO-RE-2
600 700 800 900 1000 1100 1200 1300
Temperature (⬚C)
Figure 1 Maximum isothermal service temperature for various chromia-forming alloys, estimated from literature and
producer data.
Atomic percent chromium

0 10 20 30 40 50 60 70 80 90 100
1900
1863 ⬚C
L
1700
1538 8C 1513 8C
1500 19.8
1394 8C
1300
Temperature ( ⬚C)
1100 (g Fe)
11.2 13.4
912 8C
900 846 8C
(aFe, d Fe) 830 8C
6.5 45
770 8C TC
700 s (Cr)
500
47.2
300
0 10 20 30 40 50 60 70 80 90 100
Fe Weight percent chromium Cr
Figure 2 Fe–Cr phase diagram.

0 10 20 30 40 50 60 70 80 90 100
1900 1863 ⬚C
1700
L
1500 1455 ⬚C
1345 ⬚C
Temperature (⬚C)
1300 (Cr)
1100
(Ni)
900
700
590 ⬚C
g⬘
500
0 10 20 30 40 50 60 70 80 90 100
Ni Weight percent chromium Cr
Figure 3 Ni–Cr phase diagram.

0 10 20 30 40 50 60 70 80 90 100
2000
1863 ⬚C
1800
L
1600
1495 ⬚C
Temperature (⬚C)
~41
1400
37 43.9
1395 ⬚C 1283 ⬚C (Cr)
40
~1260 ⬚C ~56
1200
(aCo)
1121 ⬚C
1000 33.6 967 ⬚C

50.5 s
800
(eCo)
600
0 10 20 30 40 50 60 70 80 90 100
Co Weight percent chromium Cr
Figure 4 Co–Cr phase diagram.
chromium is a highly soluble element in iron, (700–1100 K), with very bad mechanical properties
nickel, and cobalt.3 However, the intermetallic (high brittleness) and its formation must be avoided.
sigma phase may appear during long time anneal Ternary or quaternary Fe–Ni–Co–Cr solid solutions
of iron–chromium alloys at medium temperatures are also stable.
The particular situation of iron, where bcc–fcc allo- Chromium steels containing 12% Cr and more
tropic transformation occurs at 912 C, has to be con- are usually qualified as stainless steels. Commercial
sidered. Increasing chromium content of bcc Fe–Cr grades are 12% and 17% Cr, with a new composition
alloys to 5 wt% first reduces this value to 850 C recently appeared at 14%. Higher chromium is also
before increasing it to 1050 C for 15 wt%. For available, up to 30%. These alloys are ferritic or
higher Cr contents, the steels retain the bcc structure austenitic depending on the fraction of g-stabilizing
up to the melting temperature. Fe–Cr ferritic stainless element (Ni or Mn). Both ferritic and austenitic
steels with a chromium content falling in the g þ a stainless steels are protected in high temperature
loop, if not ‘stabilized,’ undergo phase transformation service by a chromia scale. However, that does not
on heating, leading to a biphasic structure. As described mean they can be used indifferently in any industrial
in the following, the appearance of the fcc g phase situation. Austenitic stainless steels are preferably
significantly modifies two main properties of the used when mechanical resistance at high tempera-
alloy: diffusion coefficient of chromium and coefficient tures is required, despite their higher price. Con-
of thermal expansion (CTE). versely, ferritic grades are used in situations where
Nickel is used as g-promoter in chromium permanent or cyclic mechanical loading in service
alloys, giving better mechanical properties than the remains low. However, a careful design of parts made
ferritic structure. Considering not only the high price of of ferritic grades may authorize their use in various
nickel, but also the instability of its market, efforts have industrial situations.
been made to replace it by other gamma-stabilizing Due to their difference in crystallographic struc-
elements. Manganese-containing austenitic stainless ture, ferritics, and austenitics do not allow iden-
steels, such as 204Cu grade, have been produced and tical chromium diffusion rates; the more open bcc
tested in high temperature situations.4 ferritic structure allow quicker chromium diffusion than
Carbon and nitrogen minor elements in Fe–Cr the compact fcc austenitic structure. As a result, alloy
alloys are g-stabilizers. Their concentration is there- chromium depletion under chromia scales will be
fore not strictly limited in austenitic alloys and can be more marked, but to shorter depths, in austenitics.
adjusted according to the destination of the materials. A second property which makes ferritics and
Conversely, their amounts must be controlled in austenitics different is their CTE, which lies in the
ferritic steels. range (12–13) 106 K1 for the bcc grades and
Metallic elements such as Ti, Nb, Mo, intrinsi- (19–20) 106 K1 for the fcc grades. Due to the
cally ferrite promoters, have the additional property lower CTE of chromia ((8–12) 106 K1 according
to strongly fix carbon and nitrogen as carbides, to various authors), chromia scales grown on austen-
nitrides, and carbonitrides. Due to this fact, they itic stainless steels are subjected to more intense
increase the stability domain of ferrite and are largely compressive stresses on cooling than the same scales
used in ‘stabilized’ ferritic stainless steels. on ferritics.
Regarding these considerations, high temperature A particular phenomenon occurring during tem-
chromium-rich alloys are generally described as mar- perature increase (decrease) of the nonstabilized
tensitic steels (iron based, with ferritic structure), ferritic grades is the volume increase (decrease) asso-
stainless steels (iron based, with ferritic or austenitic ciated with the ferrite–austenite transformation
structure) and refractory alloys (iron, nickel, or cobalt which can be deleterious for the integrity of the
based, with austenitic structure). oxide scale, producing early breakaway corrosion in
Martensitic low chromium steel grades are first thermal cycling conditions.
hardened by rapid cooling, and then tempered at a Refractory alloys contain high chromium levels
temperature in excess of the proposed service tem- and strengthening elements to increase their mecha-
perature. They are largely used for pressurized water nical properties at high temperatures. Based on the
or steam tubes, vessels and heat exchangers. Initially, austenitic structure, they are mainly Fe or Ni-rich,
based on the composition Fe–2.25Cr–1Mo (T22 and less often Co-rich.
steel), these grades had to follow the regular increase
in temperature and pressure of processes to fit the
1.23.2.2 Growth of Cr-Containing Oxides
required decrease in CO2 emissions; they now
include 9% Cr (T9 steel) and several similar grades To discuss the behavior of the various chromia-
with minor additions that increase their creep resis- forming alloys used in high temperature industrial
tance in service (steels T91, T92, T911). applications, several physical, chemical, and
mechanical mechanisms are of importance. Although Maintaining protection against oxidation necessi-
they have been mostly described in other chapters, tates that chromia scales remain compact and adher-
they are briefly recalled here for an efficient reading ent to the alloy substrates. Understanding the
of this chapter. mechanisms of scale spallation was the subject of
much work in the past decades.8,9 Many results of
1.23.2.2.1 Chromia growth stress measurements appear in the literature, showing
Regarding M–Cr–O systems (M ¼ Fe, Ni, Co) sub- mean values of 1–3 GPa compressive stresses in chro-
jected to oxidizing atmospheres, chromia Cr2O3 is the mia scales, counterbalanced by tensile MPa values in
most stable oxide. It is therefore expected to inhibit the alloy substrates. However, it was experienced that
M oxidation by creating a chromia barrier to oxygen, oxide stress is not the key parameter in scale spall-
lowering the oxygen chemical potential in contact with ation, but that scale adhesion energy, defined as the
the alloy to a value too low for oxidation of the major metal–oxide interfacial fracture energy, was of greater
element M. In fact, the minimum chromium content for importance.10
effectively growing a chromia scale is 14% for iron,
20% for nickel, and 25% for cobalt.5 Chromia is 1.23.2.2.2 Growth of spinel phases
regarded as an amphoteric semiconductor, with pre- In any situation where a chromia scale cannot grow or
dominant chromium interstitials at low oxygen poten- be maintained on any Fe-containing alloy surface,
tials (i.e., near the metal–oxide interface) and magnetite enriched in chromium becomes the major
predominant chromium vacancies at higher oxygen oxidation product. This compound exhibits less pro-
potentials (i.e., near the oxide–gas interface).6 In all tective properties than chromia, allowing outward
situations where chromium alloys are oxidized in air, transport of iron and inward transport of oxygen.
chromia growth takes place by chromium transport That results in a typical duplex morphology where
through the scale. Inversion of transport may occur by the external subscale is chromium free (hematite or
blocking cation diffusion using yttrium or rare earth magnetite) and the internal one has the limit stoichi-
trace additions to the alloys, resulting in slower growth ometry of FeCr2O4 (chromite spinel phase).
by oxygen transport (reactive element effect, REE). Iron chromite affords lower protection against
Water vapor or steam atmospheres also produce chro- oxidation compared to chromia, with larger mobility
mia scales where inward transport of H-assisted oxygen of oxide ions leading to more internal oxide growth.
is predominant.7 Nickel–chromium spinel (nickel chromite
Due to the high compactness of chromia, and NiCr2O4) formation may occur concurrently with chro-
also due to the necessity for less oxidizable Fe mia on nickel based low chromium alloys, leading
and Ni elements to enter it in the atomic state to some loss of protection. Increasing chromium content
(atomic radius 125 pm compared to ionic (þII) in the alloy tends to suppress this formation.
radius 77 pm), this oxide is a good diffusion barrier Manganese–chromium spinel (manganese chro-
for these major elements of stainless steels and refrac- mite, Mn3xCrxO4, with x generally between 1.5
tory alloys. Conversely, Mn, Ti and, to a lesser extent, and 2) is known to constitute the external part of
Nb, which are more oxidizable than chromium, rap- the scale grown on high temperature alloys contain-
idly diffuse through chromia scales as small cations ing fractions of manganese, as low as 0.5%. This oxide
(rMn2+ ¼ 80 pm, rTi4þ ¼ 68 pm). overlays the internal chromia part of the scale and is
The behavior of iron also has to be considered, as the result of manganese diffusion from the alloy
Fe3+ and Cr3+ ions are completely exchangeable in through this chromia part. The formation of this
the rhombohedral chromia structure. All range of oxide is not considered as deleterious as it remains
solid solutions between Fe2O3 (hematite) and Cr2O3 generally compact and adherent and it was even
(chromia) are known, with the general formula claimed that it is better resistant than chromia to
Cr2xFexO3. In high temperature processes, it is of chromium carbide transformation in high carbon
importance to limit the values of x to the minimum activity atmospheres.11 It also decreases chromia
by rapidly heating the steels to 800 C or using a activity and consequently inhibits chromium (VI)
reducing atmosphere for hematite during heating. vaporization.12,13
This is of particular importance when iron-free sur- Competition between chromia and iron–nickel
faces are needed, for example, in the petrochemical chromite is sometimes a key factor for long lasting
industry where iron catalyzes graphite deposition use of high temperature alloys. Catastrophic break
(catalytic coking). away oxidation occurs generally by local than by
generalized transformation of chromia to a spinel temperature corrosion. All furnace elements working
phase.14,15 Factors that promote spinel appearance are in air or combustion atmospheres are subject to
oxide scale growth. It should be noted that oxidation,
high temperatures,
generally considered as a degradation mechanism,
low chromium and/or subsurface dechromization,
may be used as a beneficial phenomenon, as in stain-
thermal cycling, and
less steel processing where iron–chromium thick
water vapor atmospheres.
scales formed during reheating before hot rolling
Promoting chromia growth may be obtained by consume surface defects and form a ductile lubricat-
increasing the chromium content of alloy or by any ing coating preventing sticking between the metal
chemical or mechanical operation that enhances the and the roller.
number of fast diffusion paths for chromium from
the alloy to the metal–oxide interface: laser glazing
induced microcrystallization or work hardening gen- 1.23.3.1 Oxidation of Martensitic and
erating high concentration of dislocations. Ferritic Stainless Steels
Martensitic grades are possibly chromia-formers in
1.23.2.3 Influence of Alloy Minor Elements
oxygen or air in low aggressive conditions, but
Properties of today’s industrial chromia-forming exhibit rapid chromia destabilization with increases
alloys are tailored to their exact application by adjust- in temperature and time. This destabilization, lead-
ing the amount of minor metallic (Mn, Ti, Nb, Mo, Y, ing to the so-called breakaway corrosion, is also
La, Ce, Zr) or nonmetallic (C, S, Si, B) elements. accelerated by thermal cycling. Post-breakaway
The effect of manganese was described above. scales exhibit classical duplex morphology, with
Titanium, in the form of carbide, nitride, carboni- internal Cr-containing magnetite subscale and
tride, and eventually excess solid solution Ti, slightly external hematite. In water vapor or steam, chromia
increases the parabolic rate constant of oxidation, but destabilization is easier and these grades can no
concurrently promotes scale adhesion,16 due to TiO2 longer be considered as chromia-formers. Duplex
precipitation at the alloy–scale interface. External iron-rich scales are observed, as in air, generally
diffusion of titanium through chromia may also with external magnetite instead of hematite due to
occur, forming titania and/or Cr2TiO5 surface low values of the oxygen chemical potential in the
nodules. Niobium in the form of carbide, nitride, steam environment. For such a steam situation, an
carbonitride, but also of Fe2Nb intergranular inter- empirical law was proposed by Gabrel et al.18 for the
metallic precipitates, may slowly oxidize to NbO at parabolic increase of scale thickness z
the alloy–scale interface and also slowly diffuse to the
scale surface, forming NbO2 (in solid solution with 2 16200
z ¼ ð8:099Þ 2:35:10 exp
2 10
TiO2 when this latter element is present). Molybde- 8:32T
num is not oxidized at the alloy–chromia interface,
ð1 SiÞ187 ð1 CrÞ36:7 t
remaining as an intermetallic phase. Silicon and
boron, more oxidizable than chromium, enrich as
High chromium (18%) ferritic stainless steels
oxides at the metal–scale interface. Sulfur also segre-
(AISI 4XX grades) oxidize with near-parabolic slow
gates at this interface in the elementary form,
growth of chromia-rich scales. The Fe/Cr ratio of the
strongly reducing scale adhesion.17 Reactive elements
rhombohedral oxide increases with increasing tem-
(Y, La, Ce, Zr) decrease oxidation rates and promote
perature and oxidation time. For these grades, passive
scale adhesion. Mechanisms for that have been dis-
chromia scales can form for long periods, giving excel-
cussed at length. Interface sulfur removing and scale
lent oxidation resistance. As a consequence, increasing
growth inversion are probably of major importance in
chromium concentration confers better oxidation
these beneficial phenomena.
resistance in isothermal conditions, as observed in
Figure 5. It is noted that small aluminum additions
1.23.3 High Temperature Oxidation to ferritic stainless steels generate internal precipita-
Behavior of Chromium Alloys tion of alumina between the steel substrate and the
chromia scale, leading to strong improvements in the
Oxidation, in its original meaning of oxygen fixation, oxidation behavior, as shown for the 12SR (12Cr–
is the most frequently encountered mode of high 1Al) and 18SR (18Cr–2Al) in Figure 5. Such alloys
are chromia-formers in air, but may become exclu- In thermal cycling situations, good protection is
sively alumina-formers in high temperature situa- obtained up to MICF (mechanically induced
tions of decreased oxygen chemical potential where chemical failure), as proposed by Evans,19 where
chromia becomes unstable. Ferritic FeCrAls, exclu- breakaway oxidation occurs by forming nonprotec-
sively alumina-formers need a minimum composition tive iron-rich oxides. In such thermal cycling situa-
of 20Cr–4Al. tions, the same beneficial influence of chromium
was noted, with also clear improvement by the fer-
rite stabilizers Ti and Nb (Table 2).20 The 441
8 grade is commonly proposed for exhaust parts in
409
430
the automotive industry.
304 High temperatures or water vapor in the gas may
6 442
441 promote chromium (VI) oxide or oxohydroxide
446 volatilization, strongly impeded by manganese addi-
18SR
4 12SR tions which form an external MnCr2O4 spinel phase,
decreasing chromium (III) oxide activity and there-
fore the gaseous chromium (VI) compound’s partial
2 pressure as seen in Table 3.
The growing interest in solid oxide fuel cells
(SOFCs) working at intermediate temperatures
0
800 900 1000 1100 1200 (<800 C) has led to increasing work on ferritic stain-
Temperature (⬚C) less steels which are possibly used as interconnectors,
Figure 5 Weight gains of various stainless steel grades
separating the cathode side (air) from the anode side
after 100 h isothermal oxidation in air (data from producers). (wet hydrogen) of the cell. Good technical solutions
need decreased chromium (VI) poisoning of the cath-
Table 2 Time to breakaway for cyclic oxidation of fer-
ode and manganese additions help by lowering the
ritic stainless steels (in and out 20 min cycles in laboratory partial pressure of undesired gas species. Another
air at 950 C) important parameter for this application is the low
contact resistance between scale and metal and ele-
Steel grade 409 439 441 436
ments forming internal oxide enrichment at the
Composition 12 Cr 18 Cr 18 Cr–Ti, 18 Cr–1 metal–scale interface (mainly Si and Al) must be
Nb Mo avoided. Following these observations, special grades
Time to 150 900 >1200 >1200
to be used in SOFCs have been produced and are
breakaway (h)
now available in the market (Table 4).
Table 3 Calculated partial pressures of gaseous chromium (VI) oxide or oxyhydroxide in equilibrium with solid Cr2O3 or
MnCr2O4
Atmosphere Solid compound Predominant gas 900 (K) 1200 (K) 1500 (K)
N2 + 20% O2 Cr2O3 CrO3 1.5 108 1.4 106 3.3 105

N2 + 20% O2 MnCr2O4 CrO3 5.0 1010 1.1 107 4.3 106
N2 + 20% O2 + 15% H2O Cr2O3 CrO2(OH)2 1.6 107 9.4 107 3.7 106
N2 + 20% O2 + 15% H2O MnCr2O4 CrO2(OH)2 5.2 109 7.3 108 3.5 107
Data for CrO3 from Barin,21 for CrO2(OH)2 from Opila et al.22 for MnCr2O4 from Jung.23 It is reminded that volatilization is assumed to be
negligible for gas pressure <1 nbar and catastrophic for gas pressure >0.1 mbar.
Table 4 Composition of ferritic grades specially adjusted to be used as SOFC interconnectors
Name Cr C Si Mn Cu Al Ti Ni Others
CROFER22APU ThyssenKrupp 22 <0.03 <0.5 0.3–0.8 0.5 0.5 0.03–0.2 La 0.04–0.2

ZMG232 22 0.02 0.4 0.5 0.05 0.26 0.04 La, 0.22 Zr
Hitachi metals
Fe–Cr–W 20 0.3 0.5 0.3 Zr
Sumitomo metals 7W
1.23.3.2 Oxidation of Austenitic Stainless service temperature needs increased chromium up to

Steels (AISI 3XX Grades) 20–25%, with larger silicon additions up to 2–2.5%
producing silica enrichment at the alloy–chromia
For temperatures lower than 800 C, 18Cr–8Ni stain-
interface, decreasing chromium incorporation into the
less steels 304 and 321 are commonly used. Increasing
scale. With such compositions, grades 309, 310, and 314
are known as ‘refractory stainless steels.’ All these
100 12/0 grades are more mechanically resistant than ferritics
18/0
27/0 and can be further strengthened by increasing carbon
Metal loss rate (mm yr–1)
18/10
18/36 and/or nitrogen contents.
25/12
10 25/20 Care must be taken when using high chromium
20/80
15/80
austenitic grades at intermediate temperatures where
s-FeCr may precipitate (310, 314, 253 MA grades, for
example), leading to brittleness. Possible formation of
1
this phase can be roughly predicted from the Fe–Ni–Cr
phase diagram, with additional influence of minor
elements by using empirical relations.24
0.1
In Figure 6, it is shown that the austenitic grades are
900 1000 1100 1200 more scaling resistant in isothermal conditions than the
Temperature (⬚C)
ferritic ones for equivalent chromium content.
Figure 6 Isothermal air behavior of various austenitic In cyclic conditions, heavy spallation is observed
grades compared to three ferritics with equivalent clearly due to strong differential contraction on cool-
chromium contents; legend: Cr/Ni in wt%. Redrawn from
Brasunas, A. de S.; Gow, J. T.; Harder, O. E. Proc. ASTM
ing. Here also, increasing nickel content leads to
1946, 46, 870. better behavior (Figure 7).
Alloy 625
0 Alloy 600
80Ni–20Cr
60Ni–15Cr
22Cr–32Ni
25Cr–20Ni
–20
Weight change (%)
20Cr–25Ni
–40
18Cr–8Ni (+Nb)
–60 19Cr–14Ni
18Cr–8Ni
0 200 400 600 800 1000

Time (h)
Figure 7 Weight loss in cyclic conditions (15 min heating at 980 C, 5 min cooling) of austenitic grades with different
chromium and nickel contents. Adapted from Lai, G. A. High Temperature Corrosion of Engineering Alloys; ASM International:
Metals Park, OH, 1990.
500 An influence of steel structure on spallation

Cyclic oxidation 850 ⬚C
450 appears also in the recent results from a study by
F18TNb (AISI 441) Ugine & ALZ (today ArcelorMittal Stainless Europe)
400
18-9ED (AISI 304) for exhaust systems (Figure 8),25 confirming older
350 R20-12 (AISI 308Si) results (Table 5 and Figure 9) showing that ferritic
m/s (g m–2)
300 stainless steels are better resistant in short cycle

250 situations than austenitic ones. This is due to a
200
lower thermal expansion coefficient which prevents
high stress states from appearing during cooling. Fer-
150
ritic materials also relax by high temperature creep,
100 further decreasing oxide stress. However, the good
50 behavior of ferritic stainless steels in cyclic conditions
is lost as soon as they are used at temperatures enter-
0
0 250 500 ing the (a þ g) domain.
(a) t (h) Internal alumina precipitation for the 18SR grade
is also beneficial as seen in Table 5.
100
Cyclic oxidation 850 ⬚C
50
0
1.23.3.3 High Temperature Oxidation of
–50
Ni- and Co-Containing Alloys
–100
1.23.3.3.1 Ni–Cr alloys
m/s (g m–2)
–150
–200 Binary nickel–chromium alloys have been recog-
–250 nized very early as high temperature oxidation resis-
–300 tant. Table 6 summarizes results published in 1957
–350 showing the ability of Cr to drastically decrease
–400 the parabolic rate constant of nickel oxidation.28 It
F18TNb (AISI 441)
–450 18-9ED (AISI 304) can be noted that higher chromium is necessary
R20-12 (AISI 308Si)
–500
when the temperature increases. Commercial alloys
0 250 500 mainly contain 20–30% Cr to stabilize a protective
(b) t (h) chromia scale and always contain several minor
Figure 8 Comparative air cyclic oxidation of ferritic and additions to improve their mechanical and/or
austenitic grades (20 min cycles). (a) gross mass gain; chemical properties: typical ‘Nichrome’ (UNS
(b) mass change after descaling. N06003, EN/DIN 2.4869) contains 1% Si and
Table 5 Weight changes of austenitic and (stabilized) ferritic grades of equivalent composition in air thermal cycling
conditions
Alloy Average Cycling between 870 and 930 ( C) Cycling between 980 and 1040 ( C)
composition
288 480 750 958 130 368 561 753 1029

cycles cycles cycles cycles cycles cycles cycles cycles cycles
309 20 Cr, 12 Ni, 0.3 4.6 23.7 32.6 24.2 77.5 178 242 358
2 Si
310 25 Cr, 20 Ni, n.t.a n.t. n.t. n.t. 1.5 11.3 29.3 62.8 107
2 Si
442 20 Cr, 1 Mo, 0.7 1.2 1.5 1.5 n.t. n.t. n.t. n.t. n.t.
1 Si
446 25 Cr, 0.5 Si 0.3 0.4 0.3 0.1 0.4 0.5 0.2 7.0 19.4
18SR 18 Cr, 1 Si, 0.3 0.4 0.5 0.6 0.7 1.1 1.5 2.2 3.0
0.4 Ti, 2 Al
a
n.t.: not tested.
Source: Reproduced from Lasday, S. B. Ind. Heating 1979, 12. Reproduced from Lai, G. A. High Temperature Corrosion of Engineering
Alloys; ASM International: Metals Park, OH, 1990.
200 20
409 304 SS
405
446 Alloy 800 H

310 Alloy 188
100 18
CTE (K–1 ⫻ 106)

430
R20-12
304 Alloy 600
Alloy X
0 16 Alloy 230
–100 14
12
Figure 9 Cyclic oxidation weight change of several ferritic 200 400 600 800 1000 1200
and austenitic stainless steels after 400 cycles in air at ΔT from room temperature (⬚C)
1000 C (30 min in furnace, 30 min out). Reproduced from
Kado, S., Yamazaki, T., Yoshida, K., Yabe, K., Kobayashi, H., Figure 10 Reduction of the coefficient of thermal
Trans. Iron Steel Inst. Jpn. 1978, 18, 387. expansion (CTE) by minor element additions for
several chromia-forming alloys. Data from Haynes
International.
Table 6 Parabolic rate constant of oxidation in air or in

pure oxygen (atmospheric pressure) of pure nickel and vari-
ous nickel–chromium alloys 1.23.3.3.2 Ni–Fe–Cr alloys
Chromium Atmosphere Temperature Parabolic rate Increasing nickel content from stainless steels stabi-
content ( C) constant 100 lizes the gamma phase and prevents sigma phase
(wt%) (g2 cm4 s1) formation. High mechanical resistance, particularly
0.0 Air 900 0.28
to creep in service, is realized by adding interstitial
1.97 Air 900 4.9 nonmetallic elements (C, N, B) and/or substitutional
4.12 Air 900 5.8 refractory elements (Mo, W, Ta, Nb) in solid solution
5.89 Air 900 8.2 in the g phase. Carbide or intermetallic precipitation
8.0 Air 900 0.0 also participates in alloy strengthening.
0.0 Oxygen 1000 3.48
0.3 Oxygen 1000 15.0
Other effects of minor element additions are to
1.0 Oxygen 1000 25.8 slow down grain growth in service, to inhibit sigma
3.0 Oxygen 1000 28.3 phase formation and to lower the thermal expansion
10.0 Oxygen 1000 5.55 coefficient. Figure 10 shows that Ni-base alloys (600,
0.0 Oxygen 1096 5.48 X, and 230) have naturally lower CTE than Fe-based
0.32 Oxygen 1096 23.6
0.92 Oxygen 1096 29.7
(304, 800H) and also that the strongly-charged 230
2.0 Oxygen 1096 39.6 alloy (14% W, 2% Mo) has lower CTE than Alloy
3.45 Oxygen 1096 46.8 X (9% Mo) and alloy 600 (no W, Mo). Decreasing
5.67 Oxygen 1096 58.5 CTE enhances resistance to scale spallation in ther-
7.64 Oxygen 1096 67.8 mal cycling service.
8.71 Oxygen 1096 30.8
11.1 Oxygen 1096 3.79
Scaling resistance of nickel-based alloys in oxidiz-
14.9 Oxygen 1096 0.35 ing atmospheres is far better compared to stainless
20.0 Oxygen 1096 0.07 steels (Figure 11) and these grades are of major impor-
tance in all very high temperature processes using
Source: Zima, G. E., Trans. Am. Soc. Met. 1957, 49, 924.
furnaces or combustion chambers for energy transfor-
mation and chemical or metallurgical reactions.
0.01–0.04 reactive element, whereas Nimonic 75

contains 0.5% Ti. 1.23.3.3.3 Co–Cr alloys
Nickel–chromium alloys are mainly used in the Industrial cobalt–chromium alloys are strengthened
form of wires for electrical heating and for making by solid solution and carbide formation. They present
heat resistant parts in situations where mechanical very good resistance to high temperature creep, even
considerations are low. for temperatures above 1000 C. Due to their high
2000
>2667 >3566
* 982 ⬚C
1093 ⬚C
1500
Affected metal (μm)
1149 ⬚C
1204 ⬚C
*
1000
**
500
0
316 446 800 H 625 Hast. X 601 230
10 microns
Figure 11 Metal loss plus average internal penetration
measured on various iron- and nickel-base alloys after Figure 13 Cross section of alloy HK40 oxidized at
1008 h under flowing air (213.4 cm min1). Metal loss from 1100 C in air for 50 h, showing large internal penetrations.
weight gain, penetration from cross-section observations.
Samples cycled to room temperature once a week.
Asterisk: actual value larger. Data from Haynes
International. 1.23.3.3.4 Various comparisons between
alloy grades
One point which may be emphasized concerns the
high concentrations of minor oxidizable elements,
600 particularly for Ni(Co)–Cr grades, which lead to
internal inter- and intragranular oxidation to depths
500 Haynes 230: Ni-22Cr-14W-2Mo
often much larger than the chromia scale thickness
Haynes 188: Co-22Ni-22Cr-14W
Metal affected (μm)
(Figure 13). This may be deleterious for the mechan-

400
ical resistance of these alloys in situations where the
300
total weight gain seems to be not catastrophic. For
engineering comparison between different grades,
200 figures concerning metal loss or scale thickness
must necessarily be discussed together with the
100 depth of the affected metal (Table 7).
As seen earlier, it is difficult to compare the resis-
0
1000 1050 1100 1150 1200 tance to high temperature oxidation of various indus-
Temperature (⬚C) trial grades using only one parameter. Trying to
overcome this difficulty, an important work done at
Figure 12 Comparative oxidation damage of one
NASA was presented by Barrett where not only
nickel- and one cobalt-based alloy of approximate identical
composition showing better performance of the nickel kinetics were taken into account but also the visual
alloy at the highest temperatures. aspect of the samples.27 Many industrial alloys were
submitted to 10 cycles of 1000 h each at 982 C in air.
Kinetics were evaluated by determining linear (kl)
and parabolic (kp) rate constants from weight change
chromium contents, they also resist oxidation well,
measurements according to
but it was reported that alloy 188 is sensitive to water
vapor in the oxidizing atmosphere which reduces Dm=A ¼ kl t þ kp t 1=2
its life time. They are of interesting use when nickel
The attack parameter, called KB3, was defined by
is prohibited, for example in sulfidizing situations
where Ni-containing low melting temperature eutec- KB3 ¼ kp þ 100jkl j
tics can form.
for positive values of kp, or by
The drawback is that their market price is high.
When no particular situation imposes the choice of KB3 ¼ 250jkl j
cobalt, it seems preferable to use nickel alloys as
shown in Figure 12. for negligible or negative values of kp.
Table 7 Comparative metal loss and depth of affected metal for Fe–Ni–Co grades after six 1-week cycles (data from
Haynes International)
Alloy T ( C)
980 1095 1150 1205
Metal Affected Metal Affected Metal Affected Metal Affected

loss metal loss metal loss metal loss metal
(mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm)
Iron base alloys
253MA 0.033 0.074 0.018 0.210 0.210 0.41 0.465 0.73
310 0.009 0.029 0.023 0.033 0.075 0.113 0.2 0.26
316 0.315 0.365 >1.75 >1.74 >2.7 >2.67 >3.57 >3.57
330 0.010 0.109 0.02 0.170 0.041 0.221 0.096 –
446 0.035 0.060 0.335 0.370 >0.55 >0.551 >0.59 >0.592
556 0.010 0.027 0.027 0.065 0.237 0.296 >3.8 >3.8
800H 0.024 0.045 0.137 0.188 0.191 0.225 0.29 0.35
Multimet 0.100 0.033 0.226 0.295 >1.2 >1.2 >3.7 >1.7
RA 85H 0.013 0.210 0.074 0.660 0.092 >1.47 0.097 >1.47
Nickel base alloys
214 0.002 0.005 0.002 0.003 0.004 0.008 0.005 0.018
230 0.006 0.018 0.011 0.032 0.058 0.087 0.11 0.201
263 0.018 0.145 0.089 0.361 0.18 0.409 >0.91 >0.91
600 0.008 0.023 0.027 0.041 0.044 0.073 0.13 0.213
601 0.013 0.033 0.031 0.066 0.061 0.135 0.11 0.191a
617 0.080 0.033 0.015 0.046 0.028 0.086 0.27 0.32
625 0.008 0.018 0.083 0.122 0.405 0.462 >1.2 >1.21
HR-120 0.008 0.094 0.030 0.190
HR-160 0.015 0.147 0.042 0.257 0.068 0.325 0.102 0.605
R-41 0.018 0.122 0.086 0.295 0.210 0.442 >0.73 >0.73
S 0.005 0.012 0.011 0.033 0.026 0.042 >0.81 >0.81
Waspaloy 0.015 0.079 0.036 0.137 0.079 0.231 >0.4 >0.4
X 0.009 0.024 0.038 0.069 0.114 0.148 >0.9 >0.89
Cobalt base alloys
25 0.010 0.018 0.234 0.259 0.430 0.490 >0.96 >0.96
188 0.005 0.015 0.010 0.033 0.184 0.204 >0.55 >0.55
a
Porosity formation in the metallic substrate.
In addition to KB3, the post-test appearance of 1 < KB4 < 5: poor

samples was ranked from 1 to 5: KB4 > 5: catastrophic.
rank 1: excellent, The results for Fe-, Ni-, and Co-chromium con-
rank 2: good, taining alloys appear in Figures 14–16 respectively.
rank 3: fair, None of the iron-base alloys were ‘excellent,’
rank 4: poor, with only four in the ‘good’ range. A minimum of
rank 5: catastrophic 20% Cr seemed necessary for this ‘good’ ranking.
Nickel-base alloys behave better, with four grades in
and the global rating parameter (KB4) was defined by
the ‘excellent’ family and nine in the ‘good’ one. It
KB4 ¼ KB3 ½1 þ 0:1ðrank 1Þ should be observed that Hasteloy-S, which behaved
the best, contains the reactive element (La) that
Using this global parameter, the behavior of alloys
improves scale adhesion, but Chromel-A, having a
in long term thermal cycling situations was ranked
quite similar KB4, does not. In fact, this test exhibit-
according to the following rule:
ing very long isothermal periods and only a few
KB4 > 0.2: excellent thermal cycles was probably not discriminating in
0.2 < KB4 < 0.5: good terms of scale spallation. Cobalt-base alloys pre-
0.5 < KB4 < 1: fair sented the best behavior, with five alloys out of
RA-26-1 H-150
Increasing Cr content
310 SS X-40
RA 310
309 SS & RA 309 P-3

Multimet
S-57
IN-800
304 SS MM-509
334 SS
RA-330 HA-188
T-439
347 SS WI-52
316 SS
L-605
321 SS
430 SS
410 SS 0.01 0.1 1 10
409 SS
KB4
0.01 0.1 1 10 100 Figure 16 KB4 parameter for several cobalt-base

KB4 chromia-forming alloys.
Figure 14 KB4 parameter for various iron-base

chromia-forming alloys.
10 000
9Cr 1Mo
Metal penetration (mm)

IN-671
1.000
Ni-40Cr
Trophet 30
IN-804
RA-333
IN-601 0.100
Hasteloy G
Hasteloy X 617 304

IN-617 410
Chromel A 0.010 310
Chromel AA
DH-241
DH-242 800 H
IN-706 0.001
IN-738 500 700 900 1100
IN-600
HC-276 Temperature (⬚C)
Hasteloy S
Chromel C Figure 17 Comparative metal penetration (metal loss +
IN-750
internal oxidation) of various alloys after 1 year oxidation in
René 80
Chromel P air as proposed by the ASSET system. Reproduced from
Hasteloy N John, R. C.; Pelton, A. D.; Thomson, W. T.; Young, A. L.;
Wright, I. G. High Temperature Corrosion Materials
0.01 0.1 1 10
Chemistry III. Electrochemical Society Meeting, March
KB4
25–30, 2001, Washington, DC, USA.
Figure 15 KB4 parameter for various nickel-base
chromia-forming alloys.
Another attempt to present comparative sets of
eight tested ranking ‘good’ to ‘excellent.’ They results was initiated by SHELL under the name
all have high chromium content (25–28%). ‘ASSET’ (Alloy Selection System for Elevated Tem-
MAR-M-509 alloy, with 23.5% Cr has a ‘fair’ behav- peratures), in collaboration with several industrial and
ior probably due to 7% W and both alloys with academic organizations in five countries.28 This infor-
lower chromium and higher W (WI-52 and L-605, mation system proposes a compilation of high tem-
21Cr–11W and 20Cr–15W respectively) behaved perature corrosion data, from literature and from
‘catastrophic’ and ‘poor.’ specific testing, arranged so that engineering deci-
It has to be reminded that the proposed test was sions can be made on lifetime, maximum tempera-
performed at 982 C, a temperature where chromia ture and influence of gas composition. In Figure 17
volatilization was not detrimental. The observation is presented the oxidation behavior of six common
of the complete set of results also showed that alu- alloys exposed to air for 1 year. It is remarkable that,
mina-forming alloys behaved much better than except the two ferritics (9Cr, 1Mo, and 410), all
chromia-forming alloys, with 10 grades showing alloys behave similarly at 800 C. Major differences
KB4 < 0.1. are found at the lower and higher test temperatures,
with the worst alloys at 500 C behaving the best at 3–5 for the best alloys to 15–40 for the worst ones.
1100 C. However, the general ranking of alloys established by
Attention has to be paid to transposition of oxida- furnace testing is not strongly modified.
tion data from laboratory tests to industrial practice.
Alloys exposed in service to combustion atmospheres
are better tested in burner rigs, a procedure really 1.23.4 Concluding Remarks
aggressive due to high rates of gas flow and possible
particle impact. Figure 18 shows comparative data The number of technical chromia-forming alloys pro-
from burner and furnace testing for various Fe-, Ni-, posed on the market is very large (see Appendices
and Co-base alloys. It is clear that burner oxidation is 1–6) and the choice of the right material is generally
more severe than still air by a factor extending from very difficult. The first step to a right choice is the
careful observation of the service conditions which
will be supported by the alloy, with clear definition of
Air, 6 ⫻ 1week cycles the possible accidental situations. To select the best
metal loss
grade, laboratory tests have to be performed at a tem-
500 Air, 6 ⫻ 1week cycles
affected metal depth perature and in an atmosphere as close as possible to the
Burner rig, 1000 hr real ones, taking into account possible thermal cycling.
400 metal loss
Burner rig, 1000 hr Comparison with specimens exposed to the real indus-
affected metal depth trial situations may validate the laboratory tests.
Thickness (μm)
300 Attention has also to be paid to the desired shape,

as all alloys are not available in all shapes, and some-
200 times, the shape influences the corrosion resistance.
The cost of the material is also an important
100 parameter, and the technical engineers must always
decide whether they want to use an expensive alloy
which will function for a long time or a cheaper one
0
Alloy grade which will have to be replaced after limited service.
A last criterion to have in mind is the availability
Figure 18 Metal loss and average metal penetration for
various alloys exposed to still air oxidation (980 C, six
of the chosen grade, which must be rapidly delivered
1-week cycles, conditions of Table 7) and burner test rig at and possibly by several suppliers to avoid a monopoly
980 C for 1000 h. Data from Haynes International. situation.
Appendix 1 Denomination and composition of refractory ferritic stainless steels
Alloy EN UNS Fe Cr Mo Si C Ti Others
409 1.4512 S40900 Bal 10.5–12.5

430 1.4016 S43000 Bal 16–18 1 <1 <0.08 6(C+N)<Ti <0.65
430Ti 1.4510 S43036 Bal 16–18 1 <1 <0.05 4(C+N)
+ 0.15–0.8
436 1.4526 S43600 Bal 16–18 0.8–1.4 <1 <0.08 7(C+N)
+ 0.1–1 Nb
439 1.4510 S43035 Bal 17.5 0.35 0.02 0.4
441 1.4509 S44100 Bal 17.5–18.5 <1 <0.03 0.1–0.6 3 C+0.3–1 Nb
442 S44200 Bal 18–23 1 1 0.2
444 1.4521 S44400 Bal 17–20 1.8–2.5 <1 <0.025 4(C+N) 0.45 (Ti+Nb)
+ 0.15–0.8
446 1.4763 S44600 Bal 23–26 1.5 <1 <0.1
CROFER 1.4760 S44535 Bal 20–24 <0.5 <0.03 0.03–0.2 0.04–0.2 La
22APU
ZMG232 Bal 22 0.4 0.02 0.04 La, 0.02 Zr
12SR 12 0.6 0.02 0.3 1.2 Al
0.6 Nb
18SR 1.4736 Bal 18 1 0.02 0.4 2 Al
R2012 S482800 Bal 20 2
598
High Temperature Materials
Appendix 2 Denomination and composition of refractory austenitic stainless steels
Alloy EN UNS Fe Ni Cr Mo Si C Others
302B/309 1.4828 S30900 Bal 8–10 19–21 1.5–2.5 <0.2

308 S30800 Bal 10–12 19–21 1 0.08
347 1.4550 S34700 Bal 9–13 17–19 <0.08 10 C<Nb+Ta<1
310/314 1.4841 S31400/S31000 Bal 19–22 24–26 1.5–2.5 <0.2
25–20 1.4843 19–21 23–25 1.5–2.5 <0.1 <0.1 RE
0.3 Al
31 1.4562 N08031 Bal 30–32 26–28 6–7 <0.3 <0.015 1–1.4 Cu
0.15–0.25 N
33 1.4591 R20033 Bal 30–33 31–35 0.5–2 <0.5 <0.015 0.3–1.2 Cu
0.35–0.6 N
153MA 1.4818 S30415 Bal 9–11 18–20 1–2 0.04–0.08 0.03–0.08 Ce
0.12–0.2 N
204Cu S20430 Bal 1.5–3 16–18 <1 <1 <0.15 0.05–0.25 N
2–4 Cu; 6.5–9 Mn
253MA 1.4835 S30815 Bal 10–12 20–22 1.4–2.5 0.05–0.12 0.03–0.08 Ce
0.12–0.2 N
304 1.4301 S30400 Bal 8–11 18–20 0.75 0.08
309S 1.4833 S30908 Bal 12–14 22–24 <1 <0.15
310S 1.4845 S31008 Bal 19–22 24–26 <1.5 <0.1
316 1.4401 S31600 Bal 10–14 16–18 2–3 0.75 <0.08
321 1.4878 S32100 Bal 9–12 17–19 <1 <0.1 0.5(C–0.8)Ti
353MA 1.4854 S35315 Bal 34–36 24–26 1.2–2 0.04–0.08 0.053–0.08 Ce
0.12–0.2 N; 1.5 Mn
RA85H S30615 Bal 14.5 18.5 3.5 0.2 1 Al
High Temperature Corrosion of Chromia-forming Iron, Nickel and Cobalt-base Alloys
Appendix 3 Denomination and composition of wrought refractory Fe–Ni–Cr alloys (balance: Fe)
Alloy EN UNS Ni Co Cr Al Mo Si C Ti Others
1.4872 6–8 24–26 <1 0.2–0.3 0.2–0.4 N; 8–10 Mn

1.4877 31–33 26–28 <0.3 0.04–0.08 0.05–0.1 Ce; 0.6–1 Nb
1.4887 33–37 20–23 1–2 <0.15 1–1.5 Nb; <0.11 N
30–20 1.4860 30–31 19.5–21.5 <0.3 1.8–3 <0.1 <0.1 RE
330 1.4886 N08330 33–37 17–20 1–2 <0.15 1.5 Mn; <0.11 N
556 1.4971 1.4833 R30556 20 18 22 0.2 3 0.4 0.1 2.5 W; 0.02 La; 0.6 Ta
800 1.4958 N08800 30–35 19–23 0.15–0.6 <1 <0.1 0.15–0.6 0.3–1.2 Ti+Al; 0.8 Mn
800H 1.4876 N08810 30–35 19–23 0.15–0.6 <1 0.05–0.1 0.15–0.6 0.3–1.2 Ti+Al; 0.9 Mn
800HT 1.4876 N08811 30–35 19–23 0.25–0.6 <1 0.06–0.1 0.25–0.6 0.85–1.2 Ti+Al; 0.9 Mn
803 S35045 32–37 25–29 0.15–0.6 <1 0.06–0.1 0.15–0.6 0.9 Mn
864 S35135 30–38 20–25 0.3 4–4.8 0.8 <0.08 0.6 0.4 Mn
903 N19903 38 15 0.7 1.4 3 Nb
907 N19907 38.4 13 0.03 0.15 1.5 4.7 Nb
A-286 1.4980 S66286 24–27 13.5–16 <0.35 1–1.5 <1 <0.08 1.9–2.35 <0.01 B; 1.3 Mn; 0.1–0.5 V
DS 1.4864 N08332 34.5–41 17–19 1.9–2.6 0.1 0.2 0.8–1.5 Mn; 0.5 Cu
Multimet R30155 20 20 21 3 1 0.08 2.5 W; 1Nb+Ta
Nitronic 50 S20910 12.5 22 2.25 5 Mn; 0.3 N; 0.2 Nb; 0.2 V
Nitronic 60 S21800 8.5 17 4 8 Mn; 0.15 N
599
600
Appendix 4 Denominations and composition of wrought refractory Ni–(Fe)–Cr alloys (balance: Ni)
Alloy EN UNS Fe Co Cr Al Mo Si C Ti Others
Waspaloy 2.4654 N07001 <2 12–15 18–21 1.2–1.6 <0.75 0.02–0.1 2.75–3.25 Zr, B
X 2.4665 N06002 17–20 0.5–2.5 20.5–23.5 <0.25 <1 0.05–0.15 0.2–1 W
S N06635 2 15.5 0.3 0.4 0.02 0.2 Mn; 0.05 La; 0.01 B; 0.8 W
Chromel A 0.5 20 1.4 0.03 0.2 Mn
Chromel AA 10 20 2 0.03 0.2 Mn
Chromel C 24 15 1.4 0.06 0.2 Mn
Chromel P 0.2 10 0.4 0.01
G N06007 20 22 0.4 0.03 1 Nb; 1 Ta; 1.3 Mn; 2 Cu; 0.5 W
N N10003 <5 7 1 0.06 0.3 Cu
N4 7.5 9.8 4.2 0.06 3.5 0.5 Nb; 5 Ta; 0.15 Hf; 6 W
N6 12.5 4.2 5.75 5.4 RE; 7.2 Ta; 0.15 Hf; 6 W
Astroloy 17 15 4 0.06 3.5 0.03 B
X-750 2.4669 N07750 5–9 14–17 0.4–1 <0.5 <0.08 2.25–2.75 1(Nb+Ta); 0.5 Mn
HR-160 2.4880 N12160 2 30 2828 2.8 0.05 0.5 0.5 Mn
HR-120 2.4854 N08120 35 25 0.1 0.6 0.05 0.1 0.7 Nb; 0.7 Mn; 0.004 B
R-95 8 14 3.5 0.15 2.5 0.01 B; 0.05 Zr; 3.5 Nb; 3.5 W
R-41 2.4973 N07041 5 10 18 1.5 0.5 0.09 3.1 0.005 B
G-30 N06030 15 5 29.5 0.4 0.03 2 Cu; 2.5 W
G-3 2.4619 N06985 18–21 <5 21–23.5 <1 <0.015 <0.5(Nb+Ta); 2 Cu; <1.5 W
70–30 2.4658 N06003 <1 29–31 <0.2 1–1.5 <0.07 0.01–0.04 RE
45-TM 2.4889 N06045 21–25 26–29 2.5–3 0.05–0.12 0.03 Zr; Ce
59 2.4605 N06059 <1.5 22–24 0.1–0.4 <0.1 <0.01
60–16 2.4867 N06004 Bal 14–17 <0.3 1–1.75 <0.1 <0.04 RE
80–20 2.4869 N06003 <1 19–21 <0.2 1–1.5 <0.08 0.01–0.04 RE
Nimonic 75 2.4951 N06075 <5 18–21 <1 0.08–0.15 0.2–0.6 0.3 Mn
Nimonic 80A 2.4952 N07080 <3 <2 18–21 1–1.8 <1 <0.1 1.8–2.7 0.3 Mn; 0.06 Zr; 0.003 B
Nimonic 90 2.4632 N07090 <1.5 15–21 18–21 1.2 <1 <0.13 2–3 0.3 Mn; 0.06 Zr; 0.003 B
Nimonic 105 2.4634 <1 18–22 14–15.7 4.5–9 4.5–5.5 <1 <0.12 0.9–1.5 0.3 Mn; 0.1 Zr; 0.005 B
Nimonic 115 2.4636 <1 13–15.5 14–16 4.5–5.5 3–5 <1 0.12–0.2 3.5–4.5 0.016 B; 0.04 Zr

214 2.4646 N07214 3.5 16 4.5 0.04 0.05 Y; 0.03 Zr
230 2.4633 N06230 <3 22 0.3 2 0.4 0.1 14 W; 0.02 La; 0.5 Mn; 0.004 B
242 8 25 0.03 14 W
Nimonic 263 2.4650 N07263 <0.7 19–21 19–21 <0.6 5.6–6.1 <0.4 0.04–0.08 1.9–2.4 0.4 Mn; 0.02 Zr; 0.001 B
333 2.4608 N06333 18 3 25 3 1.25 0.05 3W
Udimet 400 14 17.5 1.5 4 0.06 2.5 0.5 Nb; 0.06 Zr; 0.008 B
Monel 400 2.4360 N04400 <2.5 0.1 0.02 <0.5 <0.3 0.4 28–34 Cu; 0.8 Mn
Udimet 500 18.5 18 2.9 4 0.08 2.9 0.05 Zr; 0.006 B
Udimet 520 12 19 2 6 0.05 3 1 W; 0.005 B
600 2.4816 N06600 6–10 14–17 <0.3 <0.5 <0.08 0.2–0.6 <1 Mn
601 2.4851 N06601 Bal 21–25 1–17 <0.5 <0.1 0.4 0.5 Mn
602CA 2.4633 N06025 8–11 24–26 1.8–2.4 <0.5 0.15–0.25 0.1–0.2 0.1 Zr; 0.001–0.1 Y
617 2.4663 N06617 <2 10–13 20–23 0.5–1.5 8–10 <0.7 0.05–0.1 0.2–0.6
625 2.4856 N06625 <5 <1 20–23 <0.4 8–10 <0.5 <0.1 <0.4 3.5–4.15(Nb+Ta)
625LCF N06626 <5 <1 20–23 <0.4 8–10 <0.15 <0.03 <0.4 3.5–4.15(Nb+Ta); <0.02 N
Udimet 630 18 18 0.5 3 0.03 1 6.5 Nb; 3 W
671 46 0.3 0.05 0.3
686 2.4606 N06686 <5 19–23 15–17 <0.08 <0.01 0.02–0.25 3–4.4 W
690 2.4642 N06690 7–11 27–31 <0.5 <0.05
693 N06693 2.5–6 27–31 2.5–4 <0.5 <0.15 <1 0.5–2.5 Nb
Udimet 700 17 15 4.3 5 0.06 3.5 0.08 Zr; 0.03 B
702 N09702 0.5 15.4 3.1 0.04 0.4
706 N09706 39–44 14.5–17.5 <0.4 <0.35 <0.06 1.5–2 2.5–3.3(Nb+Ta)
Udimet 710 15 18 2.5 3 0.07 5 1.5 W; 0.02 B
718 2.4668 N07718 20–24 <1 17–21 0.2–0.8 2.8–3.3 <0.35 <0.08 0.65–1.15 4.75–5.5(Nb+Ta)
Udimet 720 15 16 2.5 3 0.03 5 1.3 W; 0.033 B; 0.03 Zr
751 N07751 5–9 14–17 0.9–1.5 <0.5 0.04 2–2.6 0.5 Mn; 0.7–1.2(Nb+Ta)
825 2.4858 N08825 <22 19.5–23.5 <0.2 2.5–3.5 <0.5 <0.05 0.6–1.2 1.5–3 Cu; 0.4 Mn
601
602
Appendix 5 Denomination and composition of wrought refractory Co–Cr alloys (balance: Co)
Alloy EN UNS Fe Ni Co Cr Al Mo Si C Ti Others
25/L605 2.4964 R30605 <0.3 9–11 Bal 19–21 <0.3 0.05–0.15 14–16 W; 1–2 Mn
Belgian S-57 Bal 25 3 5 Ta; 0.5 Y
Belgian 29.5 12 Bal 25 0.05
P-3
6B 3 3 Bal 30 <1.5 2 1.2 4.5 W; 2 Mn
150 20 1 Bal 28 0.7 0.08 0.6 Mn; 0.02 P; 0.002 S
188 2.4683 R30188 <0.3 20–24 Bal 20–24 0.2–0.4 <0.15 13–16 W; <0.12 La; 0.004 B
MAR-M-918 20 Bal 20 0.05 7.5 Ta; 0.1 Zr
1233 R31231 3 Bal 26 5 2W
S-816 R30816 4 Bal 20 4 4 Nb; 4 W
MP-35N R30035 Bal 20 10
Ultimet R31233 3 9 Bal 25 5 0.3 0.06 2 W; 0.6 Mn
Appendix 6 Denomination and composition of cast refractory Fe-, Ni-, and Co-base alloys
EN UNS Alloy Fe Ni Co Cr Al Mo Si C Others

Iron-base
1.4729 Bal 12–14 1–2.5 0.3–0.5
1.4740 Bal 16–19 1–2.5 0.3–0.5
1.4745 Bal 23–26 1–2.5 0.3–0.5
1.4777 Bal 27–30 1–2.5 1.2–1.5 0.5–1 Mn
1.4743 Bal 17–19 1–2.5 1.4–1.8
HA Bal 8–10 0.9–1.2 <1 <0.2
1.4776 J92605 HC Bal <4 27–30 1–2.5 0.3–0.5
1.4823 J93005 HD Bal 3–6 25–28 1–2.5 0.3–0.5
J93403 HE Bal 8–11 26–30 0.5–2 0.2–0.5
1.4825 J92603 HF Bal 8–10 17–19 0.5–2 0.15–0.35
1.4826 Bal 9–11 21–23 1–2.5 0.3–0.5
1.4832 Bal 13–15 19–21 0.5–2.5 0.15–0.35
1.4855 Bal 23–25 23–25 1–2.5 0.3–0.5 0.8–1.8 Nb
1.4837 J93503 HH Bal 11–14 24–27 1–2 0.3–0.5
J93503 HH1 Bal 12 26 1–2 0.4 0.5 Mn
J94003 HI Bal 14–18 26–30 0.5–2 0.2–0.5
1.4848 J94224 HK Bal 19–22 24–27 1–2.5 0.4–0.5
J94204 HK40 Bal 20 26 1.4 0.4 0.6 Mn
J94604 HL Bal 18–22 28–32 0.5–2 0.2–0.6
J94213 HN Bal 23–27 19–23 1–2 0.2–0.5
1.4852 J95705 HP Bal 33–36 24–27 1–2.5 0.3–0.5 0.8–1.8 Nb
J94605 HT Bal 33–37 13–17 1.7 0.35–0.75 0.9 Mn
1.4865 Bal 36–39 18–21 1–2.5 0.3–0.5
N08004 HU Bal 37–41 17–21 1–2.5 0.35–0.75
1.4849 Bal 36–39 18–21 1–2.5 0.3–0.5 1.2–1.8 Nb
1.4859 Bal 31–33 19–21 0.5–1.5 0.05–0.15 0.5–1.5 Nb
1.4857 Bal 33–36 24–27 1–2.5 0.3–0.5 1.2–1.8 Nb
GX40NiCrCo 20–20–20 Bal 18–22 18.5–22 19–22 <1 0.35–0.65 2–3 W; 0.75–1.25 Nb
Nickel-base
13–23- Bal 14–16 24–26 1–2 0.45–0.55 4–6 W
11–21 Bal 32.5–37.5 1.5–2 0.35–0.45
N08001 HW 12–28 Bal 12–18 1–2.5 0.35–0.65 1–1.5 Mn; 1.5–2 Nb
N08006 HX 10–20 Bal 15–19 1–2.5 0.35–0.75
NA22H 24.4 Bal 25 0.55
GX40CrNiSiNb35–17 38–50 Bal 17–20 1–2.5 0.3–0.5 0.8–1.8 Nb
GX40CrNiSi 35–17 40–45 Bal 16–18 1–2.5 0.3–0.5 0.8–1.8 Nb
1.4865 GX40CrNiSi 38–19 37–45 Bal 18–21 1–2.5 0.3–0.5
603
Continued
604
Appendix 6 Continued
EN UNS Alloy Fe Ni Co Cr Al Mo Si C Others
GX40NiCrCoW35–25–15 22–33 Bal 14–16 24–26 1–2 0.45–0.55 4–6 W

GX40NiCrNb 45–35 10–20 Bal 32.5–37.5 1.5–2 0.35–0.45 1–1.5 Mn 1.5–2 Nb
B-1900 Bal 10 8 6 6 0.1 4.3 Ta; 1 Ti; 0.1 Nb; 0.08 Zr;
0.15 B
1N-100 Bal 15 10 5.5 3 0.18 5.5 Ti; 1 V; 0.05 Zr; 0.015 B
IN-713 LC Bal 12 6 4.5 0.05 2 Nb; 0.6 Ti; 0.1 Zr; 0.01 B
IN-738 Bal 8.5 16 3.4 1.7 0.17 3.4 Ti; 1.7 Ta; 0.9 Nb; 0.1 Zr;
0.01 B
MAR-M-200 Bal 10 9 5 0.15 2 Ti; 1 Nb; 0.05 Zr; 0.015 B
Rene´ 120 Bal 10 9 4.3 2 4 Ta; 4 Ti; 0.07 Zr; 0.015 B
Rene´ 80 Bal 9.5 14 3 4 0.17 5 Ti; 0.03 Zr; 0.015 B
TAZ-8A Bal 6 6 4 0.13 8 Ta; 2.5 Nb; 1 Zr; 0.004 B
TRW-VIA Bal 7.5 6 5.4 2 0.13 9 Ta; 1 Ti; 0.5 Nb; 0.4 Hf;
0.13 Zr; 0.02 B
WAZ-20 Bal 6.5 0.15 0.15 Zr
N07765 IN-657 Bal 50 0.03 1.5 Mn
Cobalt-base
X-40 10.5 Bal 26 0.5
WI-52 Bal 21 0.45 2 Nb
MAR-M-509 10 Bal 24 0.6 3.5 Ta; 0.5 Zr; 0.2 Ti
References 16. Toscan, F.; Antoni, L.; Wouters, Y.; Dupeux, M.; Galerie, A.
Mater. Sci. Forum 2004, 461–464, 705–712.
17. Belogolowsky, I.; Hou, P. Y.; Jacobson, C. P.; Visco, S. J.
1. Hatfield, W. H. J. Iron Steel Inst. 1927, 115, 517. J. Power Sources 2008, 182, 259–264.
2. Hatfield, W. H. J. Inst. Fuel 1938, 11, 245. 18. Gabrel, J.; Coussement, C.; Verest, L.; Blum, R.;
3. Massalski, T. B. Binary Alloys Phase Diagrams; ASM Chen, Q.; Testani, C. Mater. Sci. Forum 2001, 369–372,
International: Materials Park, OH, USA, 1990. 931–938.
4. Antoni, L. Mater. Sci. Forum 2001, 369–373, 345–352. 19. Evans, H. E.; Donaldson, A. T.; Gilmour, T. C. Oxid. Met.
5. Gleeson, B. In Corrosion and Environmental Degradation; 1999, 52(5/6), 379–402.
Schütze, M., Ed.; Wiley-VCH: Weinheim, Germany, 2000; 20. Antoni, L.; Baroux, B. La Revue de Métallurgie-CIT/
Vol. 2, pp 173–228. Science et Génie des Matériaux 2002, 177–188.
6. Kofstad, P. High Temperature Corrosion; Elsevier, 1988. 21. Barin, I. Thermochemical Data of Pure Substances; VCH:
7. Bamba, G.; Wouters, Y.; Galerie, A.; Borchardt, G.; Weinheim, Germany, 1989.
Shimada, S.; Heintz, O.; Chevalier, S. Scripta Mater. 2007, 22. Opila, E. J.; Myers, D. L.; Jacobson, N. S.; Nielsen, I. M. B.;
57(8), 671–674. Olminsky, J. K.; Allendorf, M. D. J. Phys. Chem. 2007, 111,
8. Evans, H. E. Int. Mater. Rev. 1995, 40(1), 1–40. 1971–1980.
9. Schütze, M. In Series on Corrosion and Protection; 23. Jung, I. H. Solid State Ionics 2006, 177, 765–777.
Holmes, D. R., Ed.; John Wiley: Chichester, UK, 1997; Vol. 1. 24. Sourmail, T. Mater. Sci. Technol. 2001, 17, 1–12.
10. Galerie, A.; Dupeux, M.; Wouters, Y.; Toscan, F. Mater. 25. Santacreu, P. O. ArcelorMittal Stainless Europe;
Sci. Forum 2006, 522–523, 441–450. personal communication, 2007.
11. European Patent 1,325,166 B1, 2006. 26. Lai, G. A. High Temperature Corrosion of Engineering
12. Holcomb, G. R.; Alman, D. E. Scripta Mater. 2006, 54(10), Alloys; ASM International: Metals Park, OH, 1990.
1821–1825. 27. Barrett, C. A. 10 000 hour cyclic oxidation behavior at
13. Stanilowski, M.; Wessel, E.; Hilpert, K.; Markus, T.; 982 C (1800 F) of 68 Co-, Fe- and Ni-base alloys; NASA
Singheiser, L. J. Electrochem. Soc. 2007, 154(4), Technical Memorandum 107394, 1997.
A295–A306. 28. John, R. C.; Pelton, A. D.; Thomson, W. T.; Young, A. L.;
14. Essuman, E.; Meier, G. H.; Zurek, J.; Haensel, M.; Wright, I. G. High Temperature Corrosion Materials
Norby, T.; Singheiser, L.; Quadakkers, W. J. Corros. Sci. Chemistry III. Electrochemical Society Meeting, March
2008, 50(6), 1753–1760. 25–30, 2001, Washington, DC, USA.
15. Halvarsson, M.; Tang, J. E.; Asteman, H.; Svensson, J. E.;
Johansson, L. G. Corros. Sci. 2006, 48, 2014–2035.
1.24 High Temperature Corrosion of Alumina-forming Iron,
Nickel and Cobalt-base Alloys
B. A. Pint
Materials Science and Technology Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831-6156, USA

1.24.1.1 Functionality of Alumina-Forming Alloys 607
1.24.1.2 Other Applications for Thermally-Grown Alumina Layers 611
1.24.2 Selective Oxidation 612
1.24.2.1 M–Al Alloys 612
1.24.2.2 MCrAl and M–X–Al Alloys 613
1.24.2.3 Alloy Grain Size Effect 615
1.24.2.4 Precious or Platinum Group Metal (PGM) Effect 616
1.24.3 Transient Oxidation 617
1.24.3.1 Base Metal Oxide Formation 617
1.24.3.2 Cubic Alumina Phases 618
1.24.4 Steady-State Oxidation 621
1.24.4.1 Scale Growth Rate 621
1.24.4.2 Mechanism 624
1.24.4.3 Morphology 626
1.24.4.4 Adhesion 627
1.24.4.4.1 Reactive elements 628
1.24.4.4.2 Interstitials (S, C, etc.) 630
1.24.4.4.3 Precious metals 631
1.24.4.4.4 Role of CTE 631
1.24.4.4.5 Role of substrate strength 631
1.24.4.4.6 Effect of cycle frequency 632
1.24.4.5 Internal Oxidation 633
1.24.4.6 Comparison of Fe-, Ni-, and Co-Based Alumina Forming Alloys 634
1.24.5 Breakaway Oxidation 634
1.24.5.1 Failure Mechanisms 636
1.24.5.2 Unusual Failure Mechanisms 636
1.24.5.3 Strategies for Extending Lifetime 637
1.24.6 Effect of Environment 637
1.24.6.1 Water Vapor 637
1.24.6.2 Sulfur and Hot Corrosion 638
1.24.6.3 Carbon 639
1.24.6.4 Nitrogen 639
1.24.7 Summary 640
References 640
EPMA Electron probe microanalysis using

Abbreviations wavelength dispersive spectroscopy
COSIM Coating oxidation and substrate
NASA National Air and Space Administration
interdiffusion model
ORNL Oak Ridge National Laboratory
COSP Cyclic oxidation simulation program, NASA
PM Powder metallurgy
computer model
RE Reactive element
CTE Coefficient of thermal expansion
SIMS Secondary ion mass spectroscopy
606
High Temperature Corrosion of Alumina-forming Iron, Nickel and Cobalt-base Alloys 607
alloys. However, in Table 1, commercial alloy com-

Symbols positions are given in weight or mass percentage,
Cb Critical Al content of metal at breakaway which is the common convention, while Table 2
oxidation, nominal or average provides compositions in atomic percentage. The oxi-
Cbi Critical Al content at breakaway oxidation at the dation behavior of alumina-forming alloys will be dis-
metal-oxide interface cussed in a general sense in Sections 1.24.2–1.24.5 for
Cmin Minimum initial Al content needed for alumina air or oxygen environments. Section 1.24.6 covers the
formation behavior of these materials in different oxidants, for
Co Initial Al content example, H2O, S, C, or N. This discussion assumes
d Specimen thickness some basic knowledge of the high temperature oxida-
DAl Self-diffusion coefficient of aluminum tion concepts that were discussed in Chapter 1.10,
DO Self-diffusion coefficient of oxygen Oxidation of Metals and Alloys. The terminology is
Eox Elastic modulus of oxide defined in the Glossary. It is almost impossible to
M Metal: Fe, Ni or Co discuss the oxidation behavior of alumina-forming
0
pO 2
Partial pressure of oxygen at the metal-oxide alloys without discussing the reactive element (RE)
interface, Al/Al2O3 equilibrium effect.1–11 Low levels (0.1 at.%) of RE dopants have
00
pO 2
Partial pressure of oxygen at the oxide-gas a dramatic effect on oxidation performance and most
interface commercial alloys contain one or more additions. The
tb Exposure time to breakaway oxidation examples given will rely heavily on the oxidation
tz Exposure time to zero mass gain database of the Corrosion Science and Technology
W* Strain energy Group in the Materials Science and Technology Divi-
Wad Work of adhesion sion at Oak Ridge National Laboratory (ORNL) in
[X] Concentration of element X, atom percent Oak Ridge, TN.12
a Coefficient of thermal expansion
gint Surface energy of the metal-oxide interface
gmet Surface energy of the metal substrate 1.24.1.1 Functionality of Alumina-Forming
gox Surface energy of the oxide scale Alloys
r Alloy density
n Poisson’s ratio In a general sense, applications for alumina-forming
j Scale thickness alloys are only for the highest temperatures
j f Critical scale thickness at scale spallation (>900 C) or most aggressive conditions. At lower
Da Change in coefficient of thermal expansion temperatures, chromia-forming alloys, such as stain-
DT Change in temperature less steels, usually have sufficient high temperature
oxidation resistance for most applications. In fact,
Ni–Cr alloys are often discussed for applications
as high as 1200 C.13 However, metal consumption
1.24.1 Introduction rates for Fe- and Ni-based chromia forming alloys
become significant at temperatures above 1000 C.14
This chapter will discuss the oxidation behavior of These differences are not apparent at short intervals,
alumina-forming alloys, particularly MCrAl and but become more important after long intervals.
M(Al) where M is Fe, Ni and to a lesser extent, Co. Figure 1 shows the total mass gain and the spalled
(Information on intermetallic aluminides, such as mass of alloy coupons after 15 kh at 1000 C in lab-
NiAl, TiAl, or Fe3Al is included in Chapter 1.25, oratory air. (Total mass gain (i.e., specimen mass
High Temperature Corrosion of Intermetallic gain þ spalled oxide) will be frequently used in this
Alloys) For the most part, these alloys owe their chapter as it is directly proportional to metal wastage
high temperature oxidation resistance to the for- and gives more information than just the specimen
mation of a dense, protective, external alumina mass change which is typically lower than the total
scale; although some portion of the discussion will mass. Total mass is obtained by conducting the expo-
concern alloys that benefit from Al additions but do sure in a preannealed alumina crucible.15 The spalled
not exclusively form alumina scales. Compositions mass can be larger than the total mass because the
generally will be discussed in atomic percentage, spalled mass includes the metal cations while the
with Al contents typically 22% in the binary total mass reflects only O uptake.) One of the most
Table 1 Alloy chemical compositions (mass or wt%) determined by inductively coupled plasma analysis and combustion
analysis (interstitials)
Material Fe Ni Al Cr Si Ti C S Other
Commercial chromia-forming alloys

800H 43.2 33.8 0.7 19.7 0.3 0.5 0.08 <3 0.2 Mo, 1 Mn
330RA 42.6 35.9 0.1 19.0 0.1a 0.1 0.06 <3 0.1 Mo, 1.5 Mn
45TM 23.7 47.8 0.1 27.4 0.2a 0.1 0.08 <3 0.04 Ce, 0.02 La
X 17.9 46.8 0.1 22.1 0.3 0.01 0.07 <3 9.5 Mo, 1.8 Co, 0.6 W, 0.7 Mn
617 0.6 53.5 1.1 22.4 0.2 0.3 0.06 <3 12.5 Co, 9.2 Mo
556 31.3 22.2 0.1 21.6 0.3 0.03 0.11 3 18.5 Co, 3 Mo, 1 W, 0.3 Ta, 0.03 La
353MA 36.6 35.7 0.02 25.1 0.2 0.01 0.06 <3 0.05 Ce, 0.02La, 0.01 Nd
625 3.3 63.8 0.2 23.1 0.2 0.2 0.02 <10 8.9 Mo, 0.2 Nb
230 1.5 63.4 0.3 22.6 0.4 0.01 0.10 9 12.3 W, 1.4 Mo, 0.5 Mn, 0.01 La
602CA 9.5 62.3 2.4 24.9 0.08 0.15 0.19 <10 0.1 Zr, 0.1 Y
ODS Ni–25Cr (lab.) 0.5 74.2 0.01 24.3 0.06 < 0.04 220 0.08 Y, 0.6 O
Commercial alumina-forming alloys
214 3.5 75.9 4.3 16.0 0.08 0.01 0.05 3 0.02 Zr, 0.004 Y
693 4.8 62.1 3.1 28.6 0.05 0.4 0.03 <3 0.6 Nb, 0.2 Mn, 0.05 Zr
702 0.04 81.9 2.9 14.7 0.01 0.3 < <1 0.04 Mg, 0.02 Zr
René N5 0.07 63.5 6.3 6.8 0.07 0.01 0.05 4 7 Co, 6 Ta, 5 W, 3 Re, 0.2 Hf, 0.005 Y
Nimonic 105 0.8 53.0 4.9 15.5 0.13 1.36 0.14 20 19 Co, 4.7 Mo, 0.12 Zr
MA956 74.2 0.08 4.5 20.0 0.07 0.4 0.01 25 0.4 Y (oxide)
MA956HT 71.2 0.05 5.6 21.9 0.06 0.4 0.04 44 0.4 Y (oxide)
PM2000 73.2 0.02 5.5 20.1 0.02 0.4 0.01 31 0.4 Y (oxide)
ODM751 75.8 0.02 5.0 16.1 0.04 0.56 0.01 57 1.5 Mo, 0.4 Y
APM 73.3 5.5 20.4 0.2 0.03 0.03 10 0.1 Zr
APMT 69.2 0.2 5.0 21.1 0.6 0.02 0.03 <10 2.8 Mo, 0.2 Hf, 0.25 Y, 0.1 Zr
Ohmalloy 30 82.8 0.6 2.6 12.6 0.3 0.3 0.02 <3 0.4 Mn, 0.012 Mg
Ohmalloy 40 81.7 0.5 3.6 12.7 0.2 0.3 0.02 <3 0.4 Mn, 0.012 Mg
SUH21 78.6 0.1 3.0 17.7 0.3 0.1 0.01 <3 0.1 Mn, 0.01 Ce
Aluchrom YHf 72.9 0.1 6.1 20.3 0.2 < 0.02 <10 0.2 Mn, 0.06 Y, 0.05 Zr, 0.03 Hf
a
Higher value in product literature.
<
Below the detectability limit of <0.01% or <0.001% for interstitials.
oxidation-resistant, commercial, chromia-forming Above 1000 C, the benefit of alumina-forming

alloys is the Ni-based, alloy 230, Table 1. Even after alloys becomes much more evident. Instead of just
15 kh at 1000 C (although there were only scale comparing the mass change of various alloys, con-
spallation-inducing, thermal cycles every 500 h, see sider the example of designing a heat exchanger
Section 1.24.4.4.6), the mass gain for alumina-forming where the wall thickness will determine the lifetime
alloys (e.g., 214, MA956, PM2000, see Table 1) is not of the component. Figure 2 shows the time to reach
significantly lower than that for alloy 230. However, 20–40 mg cm2 total mass gains as a function of spec-
the amount of spalled oxide for the alumina-forming imen thickness for alloy 230 (Ni–Cr, Table 1) at
alloys is typically much lower (Figure 1). This has 1100 C in laboratory air (exposed with 100 h per
significant implications for extrapolations of oxidation- cycle). These times are compared with the times to
limited lifetimes for real applications where 40 kh is breakaway oxidation (tb, see Section 1.24.5) for oxide
often considered a short interval between equipment dispersed FeCrAl alloy MA956. As will be discussed
overhauls and fossil or nuclear energy plant life in Section 1.24.5, Ni-based alloys do not have such a
can be 300 kh.16 Also, alloy 230 was one of the best definitive, binary-like, transition time and instead show
commercial alloys tested in the study. A number of an increase in the rates of mass gain and spallation.
chromia-forming alloys, particularly those without Thus, for a 2 mm thick alloy 230 specimen, almost 7 kh
a RE addition (e.g., 330RA and 800H, while 230 con- of exposure at 1100 C was needed to reach 20 mg
tains La, Table 1) showed significantly more metal cm2 of total mass gain but in less than 1000 h
wastage in the study and a few had to be stopped early more exposure, the mass gain increased to 40 mg cm2
because of the degree of attack. (Figure 2). These times were not significantly different
Table 2 Alloy chemical compositions (at.%) determined by inductively coupled plasma analysis and combustion analysis
(interstitials)
Material Fe Ni Al Cr Si Ti C S Other
Commercial alumina-forming commercial alloys

214 3.4 70.5 8.7 16.7 0.16 0.01 0.23 5 0.012 Zr, 0.003 Y
693 4.7 57.7 6.2 30.0 0.10 0.47 0.12 <4 0.4 Nb, 0.2 Mn, 0.03 Zr
702 0.04 77.7 6.1 15.7 0.02 0.38 < <1 0.09 Mg, 0.012 Zr
René N5 0.1 64.9 13.9 7.8 0.15 0.01 0.25 7 7 Co, 2 Ta, 2 W, 1 Re, 0.05 Hf, 0.003 Y
Nimonic105 0.8 49.6 10.0 16.4 0.25 1.56 0.64 34 17.8 Co, 2.7 Mo, 0.07 Zr
MA956 69.4 0.07 8.8 20.1 0.13 0.40 0.06 41 0.24 Y
MA956HT 65.7 0.04 10.7 21.7 0.11 0.43 0.17 71 0.23 Y
PM2000 67.8 0.02 10.6 20.0 0.04 0.44 0.03 50 0.23 Y
APM 68.1 10.7 20.4 0.43 0.03 0.13 16 0.06 Zr
APMT 65.2 0.1 9.7 21.3 1.11 0.02 0.13 <10 1.6 Mo, 0.15 Y, 0.07 Hf, 0.06 Zr
Ohmalloy 30 79.6 0.5 5.2 13.0 0.50 0.38 0.077 <4 0.03 Mg, 0.4 Mn
Ohmalloy 40 77.8 0.5 7.1 12.9 0.42 0.38 0.095 <4 0.03 Mg, 0.4 Mn
SUH21 75.0 0.1 5.9 18.1 0.55 0.12 0.04 <4 0.1 Mn, <0.01 Ce
Aluchrom YHf 67.3 0.1 11.7 20.1 0.35 < 0.09 <10 0.2 Mn, 0.036 Y, 0.028 Zr, 0.008 Hf
Model alumina-forming alloys
NiAl þ Hf < 51.6 48.3 < < < 0.034 <4 0.056 Hf
Ni3Al þ Hf < 75.1 24.8 < < < 0.041 <4 0.057 Hf
Ni–22Al–5Pt þ Hf 0.02 72.8 22.1 < 0.02 < 0.024 <10 5.0 Pt, 0.016 Hf
Ni–22Al– 0.02 68.1 21.7 0.01 < < 0.012 <4 9.7 Pt, 0.39 Hf
10Pt þ Hf
NiCrAl þ Hf 0.01 61.8 19.2 18.9 < < 0.043 8 0.058 Hf
NiCrAl þ Y 0.01 60.9 19.0 20.0 < < 0.043 8 0.052 Y
NiCrAl þ Y,Hf 0.01 61.4 18.9 19.5 0.02 < 0.042 5 0.046 Hf, 0.006 Y
NiCoCrAlY 0.02 40.4 23.5 17.4 < < 0.042 8 0.147 Y
FeNiCrAl 56.5 18.7 5.1 15.0 0.29 0.39 2 Mn, 1.4 Mo, 0.6 Nb
FeCrAl 70.3 < 9.6 20.1 < < < 85
FeCrAl–0.04Y 69.1 < 10.2 20.6 < < 0.01 <10 0.041 Y
FeCrAl–0.08Y 69.8 < 9.9 20.2 < < < 16 0.082 Y
FeCrAl þ Hf 70.1 < 10.0 19.8 < < 0.044 41 0.053 Hf
FeCrAl þ Mm 70.3 < 10.0 19.6 < < 0.043 20 0.19 Ce, 0.007 Nd, 0.004 La, 0.006 Hf
FeCrAl þ 0.04Hf,Y 70.1 0.01 9.8 20.1 < < 0.037 16 0.035 Hf, 0.003 Y
FeCrAl þ 0.02Hf,Y 69.6 0.01 10.1 20.3 < < 0.010 16 0.018 Hf, 0.002 Y
FeCrAl þ Al2O3 69.2 0.04 9.6 19.6 0.17 0.01 0.17 44 0.84 O
FeCrAl þ Ho2O3 67.9 0.04 9.4 19.4 0.11 < 0.086 45 1.7 O, 0.9 N, 0.15 Ho
Fe–13Al þ Hf 86.5 0.01 13.4 0.01 < < 0.004 37 0.044 Hf
Fe–17Al þ Hf 83.1 < 16.7 < 0.05 < 0.005 67 0.057 Hf
Fe–17Al þ Y,Hf 82.8 < 17.1 < 0.02 < 0.009 8 0.023 Hf, 0.010 Y
Fe–17Al þ Y,Ti 82.6 < 16.9 < < 0.44 0.012 < 0.069 Y
Fe–17Al–10Cr þ Y 72.5 < 17.4 10.0 0.05 < 0.014 <4 0.062 Y
Fe–18Al þ Y 81.9 < 18.0 0.02 0.02 < 0.013 <15 0.057 Y
Fe–20Al þ Hf 79.7 < 20.1 < 0.02 < 0.003 39 0.056 Hf
Fe–20Al þ Y2O3 76.9 0.06 19.8 0.11 0.09 0.01 0.08 41 0.13 Y, 2.5 O, 0.4 N
Fe–40Al þ Hf 60.5 < 39.5 < < < 0.007 21 0.047 Hf
<
Below the detectability limit of <0.01% or <0.001% for interstitials.
for alloy 230 specimens that were <1 mm thick. In with a minimum lifetime service of 25 kh it would be
contrast, a drastic increase in the time to breakaway extremely difficult to predict any wall thickness of
oxidation was observed for MA956 coupons in this alloy 230 that would be applicable but an extrapolation
exposure. The example of MA956 is given because a of the MA956 data would suggest a minimum wall
number of other alloys (typically with higher Al con- thickness of 3 mm. Oxidation-limited lifetime pre-
tents) have such high values for tb that it is difficult to dictions are further discussed in Section 1.24.5.
put them on the same plot with alloy 230.14 It is easy to The upper temperature limit for alumina-forming
see that if one were designing a 1100 C heat exchanger alloys is, of course, application dependent.
70
^ 129.4
1000 °C Total mass gain
5 000 h
60 15 000 h
10 000 h
lab.air Spall
50
Mass gain (mg cm–2)

40
30
20
10
0
556
617
353
230
693
214
800H
PM2000
602CA
MA956
330RA
45
APM
APMT
X
Figure 1 Bar graph of the total mass gain (specimen mass gain þ spalled oxide) and spalled oxide mass per unit area
for various commercial alloy specimens (compositions in Table 1) at 1000 C in laboratory air after 30, 500 h cycles (15 kh total
exposure). Data from Pint12 and Pint and Keiser.14
9000 temperatures of 1250 C for NiAl þ Zr18 which

Time to FeOx contains a substantially higher initial Al content
MA956
(FeCrAl)
(50 at.%) than most alumina-forming alloys. Even
40 mg cm–2
Exposure time at 1100 ⬚ C (h)
8000 n = 0.9 with a longer cycle time between cooling to room

30 mg cm–2 temperature, the maximum temperature used to
25 mg cm–2 achieve respectable lifetimes will not be significantly
higher. This is because scale spallation must also be
7000
considered. Specifically, the strain energy per unit
volume (W*) is defined19 as
Time to 20 mg cm–2
6000 Alloy 230 (NiCr) W ¼ Eox ðDT DaÞ2 ð1 nÞ ½1
n = 5.9
(where Eox is the elastic modulus of the oxide, DT is
the temperature drop, Da is the coefficient of thermal
5000 expansion (CTE) difference between metal and
0.5 1.0 1.5 2.0 oxide, and n is Poisson’s ratio) in a thick alumina
Specimen thickness (mm) scale (x > 10 mm) will exceed the work of adhesion
Figure 2 Time to accelerated attack as a function of (Wad), that is, the work needed to form unit area of
specimen thickness for NiCr alloy 230 and alloy MA956 (see oxide and metal interfaces from unit area of interface
Table 1 for compositions) for exposure in 100 h cycles at of the metal–oxide interface20:
1100 C in laboratory air. For MA956, the time is for the
onset of Fe-rich oxide formation and increases significantly Wad ¼ gox þ gmet gint xW ½2
with thickness. For NiCr, the times are to total mass gains of
20–40 mg cm2 that do not change significantly with where the g’s are the surface energies of the oxide,
thickness. Values of n are given for eqn [5]. Data from Pint12 metal and interface. This leads to spallation and the
and Pint and Keiser.14 Al consumed to regrow the spalled oxide signifi-
cantly increases the Al consumption rate, after spall-
However, the lifetime predictions made by NASA ation begins. For example, the time needed to form
using modeling methods like Cyclic oxidation simula- a 10 mm thick alumina scale is 200 h at 1200 C
tion program (COSP)17 for b-NiAl and NiAl þ Zr, but 10 kh at 1050 C. Thus, the prime temperature
usually with criteria of 250 mm metal consumption in range of interest for commercial applications of
<10 kh and 1 h thermal cycles, showed maximum alumina-forming alloys is 1050–1250 C.
Because of the high temperatures involved, appli- Metals) are discontinuing production because of the
cations for alumina-forming alloys are often limited limited market. Less expensive dispersion strength-
by their creep strength. For ferritic alumina-formers ened alloys such as Kanthal alloys APM® and
such as FeCrAl, some of the primary applications are APMT® also have been developed and are more
as heating elements (e.g., Kanthal alloys) and in foil widely available. Because of their higher strength,
(30–70 mm) form as catalyst supports in automotive dispersion strengthened alloys have better dimen-
applications21 as well as catalytic combustion to sional stability at high temperature and APMT is
lower emissions from turbine engines.22 Some of the available in tube form. Because of its strength and
less discussed applications are: (1) dissipative resistor oxidation resistance, APMT was selected as a heat
banks, which are used to eliminate excess electricity shield material in the heat exchanger of a biomass
for regenerative braking of diesel electric locomo- gasifier.14
tives,23 (2) reinforcement for molded refractory pro- Nickel-based alumina-forming alloys generally
ducts, and (3) thermal protection shields. have better high temperature strength than wrought
Oxide dispersion strengthened (ODS) FeCrAl ferritic alloys and can be considered for a broader
alloys have been the subject of considerable research range of applications. Alloys such as 214 have similar
and will be discussed throughout this chapter. The oxidation resistance as FeCrAl alloys and are available
use of yttria dispersions increases the creep strength in foil form, as in Figure 3. The second-generation
of these materials but the drawback is that the powder of single crystal (SX) and dispersion strengthened
metallurgy (PM) processing24 increases their cost Ni-based super alloys contain 11–13 Al and have
significantly. These alloys have been investigated for fairly good oxidation resistance, even without a coat-
a variety of high temperature applications including ing.16,27,28 To further improve their oxidation resis-
heat exchangers running at 1100 C.25 Figure 3 tance, Ni-based superalloys are often coated with a
gives the creep rupture life for 125 mm thick higher Al content NiAl or NiCoCrAl composition
PM2000 foil compared to 100 mm thick Ni-based (dealt elsewhere in Chapter 1.27, High Temperature
alloys. The higher creep strength of this material Coatings: Protection and Breakdown). When suffi-
allows a wider range of applications to be considered. cient Al is added to an alloy for alumina scale forma-
However, manufacturers (e.g., Plansee and Special tion, other properties such as strength, phase stability,
and weldability may be adversely affected Neverthe-
less, cast NiCrAlY alloys have been developed for
steam reforming applications where the formation of
10 000
a chemically inert alumina scale has an additional
Alloy 214 100 MPa
benefit of reducing surface coke formation.29
Cobalt-based alumina-forming alloys are also pri-
1000
marily used as coatings, especially for hot corrosion (see
Rupture life (h)
Section 1.24.6.2) or high temperature wear resistance

applications. The addition of Co in MCrAlY coatings
100 PM2000
Alloy 625 is known to change the phase composition and
mechanical behavior.30,31 Minimal research has been
reported on Co-based alumina-forming alloys in the
10
last 20 years and their application seems somewhat
specialized and limited at present.
1
700 800 900 1000 1100
1.24.1.2 Other Applications for
Temperature (⬚C)
Thermally-Grown Alumina Layers
Figure 3 Creep rupture life at 100 MPa for 100 mm thick
foil of Ni-base alloys 214 and 625 and for 125 mm thick alloy One of the more interesting aspects of alumina-forming
PM2000 foil (alloy compositions in Table 1). Adapted from alloys is the usage of the alumina layer in other applica-
Pint, B. A.; Swindeman, R. W.; More, K. L.; Tortorelli, tions besides for oxidation resistance. After forming a
P. F. Materials selection for high temperature (750–1000 C) dense-adherent alumina layer at high temperature, this
metallic recuperators for improved efficiency microturbines,
ASME Paper #2001-GT-445, Presented at the International
layer could be used at lower temperatures where fur-
Gas Turbine and Aeroengine Congress and Exhibition, ther alumina formation would be limited. Some of the
New Orleans, LA, June 4–7, 2001. properties of interest are electrical (insulating),
chemical inertness (for chemical processing applica- needed to reform Al after an alumina scale had
tions) and wear resistance. Besides microelectronics, already been established, Cb.41,42 The basic concept
one application where alumina is attractive for its elec- of selective oxidation was discussed in Chapter 1.09,
trical resistance is, as a coating to decrease the Thermodynamics and Theory of External and
magneto-hydrodynamic force in magnetic confine- Internal Oxidation of Alloys. For alumina-forming
ment fusion reactors. Alumina could be used as a coat- alloys, the general interest is the behavior of M–Al and
ing or inside a flow channel insert.32 Another property M–X–Al, where X is normally Cr but can also be
of interest is hydrogen or tritium permeability.33 Alu- another element, such as Mn, with nobility between
mina has a hydrogen permeability several orders of M and Al, the so-called third element.43
magnitude lower than most metals (Figure 4). Because
of the recent interest in the use of hydrogen as a clean-
1.24.2.1 M–Al Alloys
burning fuel and in the use of more traditional weapons
and nuclear energy applications (e.g., hydride fuels), For Ni–Al alloys, an early work by Pettit44 consid-
permeation barriers are widely desired but frequently ered the thermodynamics of the system as well as
ineffective.35 Also nuclear-related, alumina can be a experimental evidence which indicated that kinetic
corrosion barrier in liquid Pb (or Pb eutectics like issues dominate Cmin. The thermodynamic calculations
Pb–Li or Pb–Bi) systems36 but not liquid Li, which indicated <1 ppm Al at the alloy surface was needed
dissolves alumina37 because of the thermodynamic for NiO or NiAl2O4 (spinel) to be stable. However,
stability of LiO2.38,39 Even with a decreased Li activ- experimental work, summarized in Figure 5 showed
ity in Pb–Li eutectic, the Li can transform a pre- that much higher Al contents were needed for transi-
formed a-Al2O3 scale on alloy PM2000 to LiAlO2 tion from Al internal oxidation with an external NiO
after exposure to Pb–Li at 600–800 C.40 scale (Region I) to an external alumina formation
(Region III) in 0.1 atm O2. (An over-lying spinel
layer is formed at up to 40 at.% in the b-NiAl
1.24.2 Selective Oxidation phase field.45) The effect of temperature was to
slightly lower Cmin at higher temperatures. Even in a
The first issue to consider as far as alumina-forming CO/CO2 mixture (2:1) where NiO was not stable, the
alloys are concerned is the need for selective oxida-
tion of Al in the alloy to form alumina and the Transition regions
minimum alloy Al content, Cmin, needed to form 1300
alumina in the alloy. This minimum value is not
necessarily equal to the minimum concentration 1200 Region III
Temperature (⬚C)
Region
700 ⬚ C 500 ⬚ C 300 ⬚ C 1100
II
1E–09
1E–10
Fe 1000
Permeability (mol m–1s1 Pa–½)
1E–11
Region I Ni–22Al + Hf
1E–12
Ni 900
1E–13 0 10.3 19.5 27.7 35.2 42.0
1E–14 Mo Aluminum content (at.%)
1E–15
Figure 5 Relationship between oxidation temperature
1E–16 Al and Al content on the reaction products formed on Ni–Al
1E–17 alloys oxidized in 0.1 atm O2. The various regions are:
I – primarily NiO þ Al internal oxidation, II – initial spinel
1E–18
Au (NiAl2O4) þ Al2O3 formation, but eventual NiO, III – primarily
1E–19 Al2O3 Al2O3. The bold line indicates the transition between internal
1E–20 and external oxidation in CO–CO2 mixtures where Al2O3
1.0 1.2 1.4 1.6 1.8 2.0 was the only stable oxide. Adapted from Pettit44 with the Al
content changed to atomic%. The dashed line is an
1000/temperature (K–1)
observation for Ni–22Al þ Hf where the initial NiAl2O4 þ
Figure 4 Hydrogen permeability as a function of Al2O3 scale was stable for 1000 h at 1000–1200 C. Pint, B. A.
reciprocal temperature for various pure metals and alumina. Unpublished oxidation data, ORNL, Oak Ridge, TN,
Data from Fowler et al.33 and Steward.34 1994–2008.
solid bold line in Figure 5 shows a fairly high Al interface after oxidation at 1100 C but underneath
content needed for transition from internal to exter- is a fully-developed columnar alumina scale.
nal alumina formation. Thus, kinetic factors such as Subsequent sections will discuss some of the other
the scale growth rate and diffusivities in the metal and reasons for the scatter in Figure 6. Recent calcula-
oxide dominate the selective oxidation process. tions by Zhang et al.53 using relevant Fe–Al rate con-
Values of Cmin for Fe–Al alloys have been stants predicted much lower Cmin values (0.5–4.5%
reviewed46 and Figure 6 shows a summary of Cmin depending on the assumptions) which does not agree
values found during oxidation in air or O2.47–52 As with the experimental results summarized in
discussed by Pint et al.,51 a distinct difference between Figure 6.
complete alumina scale formation and Fe-rich oxide One issue that can affect Cmin is time, that is, the
formation is clear at 1000 C but can be more length of exposure, and whether the alloy can main-
difficult to differentiate at lower temperatures, espe- tain the alumina scale during extended exposures.
cially at 600–700 C where an exclusive alumina scale Region II in Figure 5 indicates where an initial
is difficult to form on binary Fe–Al alloys and there is external alumina scale was not sustained and a faster
no significant difference in mass gain between high growing NiO formed. The likely explanation was Al
and low Al contents. One reason for the large range of depletion in the metal which did not allow the sus-
values in Figure 6 may be because of the interpreta- tained formation of alumina.44 This point might be
tion. If the criterion is that no Fe-rich oxide formed more relevant to the discussion of lifetime, i.e., the
using a visual assessment of scale color, then the amount of time an alloy can sustain the formation of
contents of 18–20%Al are typically needed. How- an alumina scale (discussed in Sections 1.24.4.4 and
ever, the same requirement for Ni–Al alloys in air 1.24.5). Nevertheless, the usefulness of Cmin plots
would require Cmin > 35%. A more appropriate cri- may be questionable if the specified level of Al can-
terion is that a continuous alumina scale forms not sustain alumina formation. For example, Hf-
underneath the transient, base metal oxide. This is doped Fe–11Al at 1200 C (Figure 6) only survived
the criteria used in Figure 5. Figure 7 gives an for 300 h before breakaway oxidation.52 Thus, these
example for Y2O3-dispersed Fe–20Al where large plots provide a defined minimum Al level and not a
faceted Fe-rich grains were observed at the gas target Al content for alloy development.
1.24.2.2 MCrAl and M–X–Al Alloys

Cast Fe−Al Cast Fe−Al
20 DeVan and Tortorelli, 1993 Sykes and Bampfylde, 1934 Most commercial alumina-forming alloys are based
on MCrAl rather than M–Al compositions. The main
Critical aluminum content (at.%)
18 Cast Fe−Al
Saegusa and Lee, 1966
Faceted Fe-rich oxide grains
16
PM Fe−Al + Al2O3
Pint, et al., 1999 (corrected)
14
PM Fe-Al + Y2O3
Pint, et al., 1999
(corrected)
12 Cast Fe−Al + Hf
Pint and wright, 2004
10
700 800 900 1000 1100 1200 1300

Temperature (⬚C) 1 μm
Columnar α-Al2O3 scale Substrate
Figure 6 Summary of critical Al contents for Fe–Al alloys
from various studies. The values for powder metallurgy (PM) Figure 7 SEM secondary electron image of scale
alloys51 were decreased by 2% because this Al was tied fracture cross section after 100 h at 1100 C in air on
up as oxides in these materials. Generally, these values Fe–20Al–Y2O3. An outer Fe-rich oxide formed over the
were determined for exposures of 100 h. Data from Sykes inner a-Al2O3 layer because of the relatively low Al content
and Bampfylde,47 Saegusa and Lee,48 Devan and in this alloy. Adapted from Pint, B. A.; Leibowitz, J.;
Tortorelli,50 Pint et al.,51 and Pint and Wright.52 DeVan, J. H. Oxid. Met. 1999, 51, 181–197.
reason is that the addition of Cr (typically 10–25%, and mass gain but did not decrease the long-term
see Table 2) allows alloys with much lower Al con- growth rate. With the presence of Cr in these alloys,
tents to form a continuous alumina-scale than the Al level needed for alumina formation, 6%, is
observed in M–Al compositions. Since Al tends to significantly lower than the Al levels discussed in
have detrimental effects on mechanical and fabrication binary Ni–Al alloys (Figure 5).
(i.e., rolling, joining) properties, minimizing the Al Composition effects can be summarized in so-
content is often a goal in alloy development. called oxide maps.57–59 Figure 9 shows a Ni–Cr–Al
The effect of Cr is described as a third element oxide map with some of the data points from Figure 8.
effect.43,53–55 There is still a disagreement as to The three oxide regions, very similar in composition to
whether the third element (1) acts as an oxygen ‘get- the work of Giggins and Pettit57 (and not at all like
ter,’ (2) changes the Al activity, (3) increases the Al Wallwork and Hed58) are: I – Ni-rich scale with Al
diffusivity, (4) increases the volume of internal oxides internal oxidation, II – Cr-rich scale with Al internal
in the alloy, (5) simply decreases the concentration of oxidation (e.g., alloys 602 and 617), and III – continu-
base metal in the alloy, and/or (6) acts by some other ous Al2O3 scale. Obviously, the most oxidation-
mechanism. Most agree that the third element effect is resistant alumina-forming alloys will be in region III.
more complex than the early explanation of Wagner.54 A similar map was constructed for CoCrAl.57 The
To illustrate the performance of Ni–Cr–Al alloys main difference between CoCrAl and NiCrAl was
(Figure 8) shows examples of Ni–(17–30)Cr alloys that much higher Cr and Al contents were needed
with various Al additions oxidized in 100-h cycles at to achieve alumina scale formation, on the order of
1200 C. Lower Al additions, 2.5–4.8% such as in 20% Cr and 20% Al. The need for higher Cr and Al
alloys 617 and 602CA resulted in high mass gains contents in Co-based alloys was also observed in a
similar to alloy 230 without Al. Alloy 602CA does separate study.60Alloys based on TiCrAl are covered
not form an external alumina layer at 1200 C; only in a separate chapter.
internal Al oxidation is observed.56 However, the Similarly to NiCrAl, in Fe-based alloys, as little as
6% Al addition in alloys 702 and 693 resulted in a 13Cr is added to Ohmalloy® alloys combined with
significantly lower mass gain after the initial transient 5.2%Al to form alumina, Table 2.23 In both Fe-
formation of Ni- and Cr-rich oxides. Further in- and Ni-based alloys, a reasonable starting point for
creases in Al content, for example, alloy 214, de- a standard alumina-forming MCrAl composition
creased the transient (i.e., Ni-rich) oxide formation
Al
20 30
Alloy 230 Ni−27Cr−4W + La
Alloy 617 Ni−25Cr−12Mo

6Mo−2.5Al γ9
Total mass gain (mg cm–2)
15
20
γ + γ9
)
.%
(at
10 Alloy 602 Ni−26Cr−9Fe-4.8Al + Y, Zr

Al
Alloy 693 Ni−30Cr−6.2Al + Ti, Zr 10 III

214
5
702 693
Alloy 702 Ni−16Cr-6.1Al + Ti, Zr
Alloy 214 Ni−17Cr−8.7Al + Y, Zr 602
I II 617
Ni−48Al + Hf
0 Ni Cr
0 200 400 600 800 1000 10 20 30
Time (h) in 100 h cycles at 1200 ⬚ C Cr (at.%)
Figure 8 Total mass gain (specimen þ spalled oxide) Figure 9 Schematic of the oxide map for Ni–Cr–Al and the
during 100 h cycles at 1200 C in laboratory air for various location of the commercial alloys from Figure 8. Region I is
Ni–Cr and Ni–Cr–Al alloys (see Table 1). Generally, the for alloys which form Ni-rich oxide and Al internal oxide,
mass gain decreases with Al content due to the formation Region II alloys form a Cr-rich oxide and Al internal oxide,
of a protective alumina scale with NiAl þ Hf having one of and Region III alloys are able to form a continuous alumina
the slowest growth rates of any alumina-former. Data scale. Map based on Giggins and Pettit,57 alloy data
from Pint.12 from Pint.12
is 20% Cr and 10% Al (5 wt%). Obviously, 6 at.% Al in Fe–15Cr–19Ni, increasing the Nb content
lower Cr and Al contents can form alumina, as decreased the mass gain after 500 h at 800 C. At low
demonstrated by both commercial and laboratory levels of Nb, mass losses were observed due to the
alloys (Tables 1 and 2). Higher Cr contents formation of a thick Fe- and Cr-rich oxide. The effect
(>25%) lead to a decrease in workability and ductil- of Y on selective oxidation also has been investigated.
ity and in FeCrAlY to a decrease in the incipient Castello and coworkers70 reported improved selective
melting temperature of any Y-rich precipitate.61 oxidation in Ni–12.9Al–11.9Cr–7.5C–0.11Y compared
While Cr is the most popular third element addi- to the Ni–Al þ Cr3C2 alloy without Y. They attributed
tion, other elements are used or have been investi- the effect to Y reducing the growth rate of the
gated. To replace Cr, Mn additions have been initially-formed Cr-rich scale. Figure 6 shows that
investigated in Fe-based alloys by several authors.62,63 Hf-additions to Fe–Al alloys decreased the critical Al
These austenitic alloys have good wear properties but content needed to form alumina.52 Studies of Y2O3
their oxidation resistance, particularly above 900 C, additions in PM Fe–Al alloys did not find a benefit of
is compromised of the formation of Mn-rich oxides. Y on selective oxidation compared to PM alloys with
The addition of Si has been investigated in Fe–Al an Al2O3 dispersion.51 Rather, the fine-grain size of the
alloys64 and has been shown to exhibit a third ele- PM Fe–Al alloys was thought to influence selective
ment effect in Ni–Si–Al alloys.65 (Large Si additions oxidation.
are mainly an academic exercise as Si has detrimental
effects on ductility and fabricability.) In Fe–Al–Si
1.24.2.3 Alloy Grain Size Effect
alloys, higher Si contents led to a decrease in oxida-
tion rate. Also, Ni additions to Fe–Al alloys have been Alloy grain size effects on selective oxidation are not
attempted.66 However, the addition of Ni did not appear widely appreciated in high temperature oxidation
to improve the oxidation behavior at 600–800 C but are extremely important and often ignored as
compared to binary Fe–Al. specimen alloy grain size is not always specified in
Quaternary additions on selective oxidation have literature papers. Early work on chromia-forming
been explored. In the case of alumina-forming auste- stainless steels showed that a fine-grained substrate
nitics (AFA, see example in Table 2), where low levels was more oxidation resistant.71–73 More grain bound-
by Al are needed to retain a fully austenitic matrix aries increased the flux of Cr to the surface along
(and retain creep-resistance67,68), additions of Nb have these fast diffusion paths. This concept was used in
been found to influence whether the alloy forms an alumina-forming alloys by Maloney and Yurek74,75
alumina scale at 800 C (Figure 10).69 At constant 5 or where rapid solidification was used to achieve a fine
grain size in Fe–2.8Cr–4.7Al alloys and TiB2 disper-
sions were used to stabilize the fine grains. As dis-
Base: Fe–15Cr–19Ni–2Mn–0.3Si–0.4C cussed in the previous section, these low levels of Cr
0.12 800 ⬚ C, 500 h exposure, laboratory air and Al are well below commercial FeCrAl alloys,
0.2Nb, 0.1Ti
Table 2. Figure 11 shows the low mass gain at
0.10 600 C and 700 C for the fine-grained material
0.6 Nb compared to the conventionally-cast, coarse-grained
0.08 Fe–2.8Cr–4.7Al specimens. While the TiB2 additions
0.9Nb, 0.1Ti
1.5Nb, 0.1Ti may have contributed to the effect, Figure 12 shows
0.06
a thicker oxide formed over a large grain in the TiB2-
0.04 1.8Nb, 0.2V 0.4Nb, 0.1Ti dispersed FeCrAl specimen. A thin Al-rich oxide
No Mn, Si 12Cr (arrow) formed over the small grains in the substrate.
0.02 Goedjen and Shores76 investigated the effect of
0.1Nb (−0.53) grain size on NiCrAlY alloys. They found a decreased
0.00 0.1Nb, 0.1Ti, 0.1V (−7.7) transient mass gain (i.e., less Ni-rich oxide formation)
5 6 7 8 with finer grains but no change in the steady-state
Al content (at.%) alumina scale growth rate. More recently, coating pro-
cesses that produce fine-grained NiCrAl coatings have
Figure 10 Specimen mass gain for Fe–15 at.% Cr–19Ni
alloys with various Al and carbide forming elements (Nb, Ti,
been found to have better selective oxidation.77,78
V) after 500 h (5, 100-h cycles) in laboratory air at 800 C. Collectively, these results suggest that for low Al
Data from Brady et al.69 content alloys, reducing the grain size will improve
Fe–2.5Cr–2.5Al
3
Conventionally processed
Fe–2.5Cr–2.5Al–1Ti–0.6B
Spray formed at 9 in
Spray formed at 25 in
Powder processed
0
300 400 500 600 700 800 900 1000 1100
Temperature (⬚C)
Figure 11 Mass change for conventional processed and fine-grained Fe–2.5 wt% Cr–2.5Al (2.8 at.% Cr–4.7Al) alloys
made by several spray and powder processes. By using TiB2 to keep a fine grain size, an Al-rich oxide was formed at
600–700 C, thereby decreasing the mass gain of these alloys compared to conventional, coarse-grained material with the
same Cr and Al content. Maloney, M. J. Ph.D. Thesis, Massachusetts Institute of Technology, Cambridge, MA, 1989,
Maloney, M. J.; Yurek, G. J. In Composites/Corrosion-Coatings of Advanced Materials, Proceedings IMAM-4; Kimura, S.,
Kobayashi, A., Umekawa, S., Nii, K., Saito, Y., Yoshimura, M. Eds.; Materials Research Society: Pittsburgh, PA, 1989; Vol. 1,
pp 383–388.
Cu plate 1.24.2.4 Precious or Platinum Group

FeOx Metal (PGM) Effect
One of the more intriguing results of this decade in
Large grain
alumina-forming alloys is the work on g–g0 Ni–Pt–Al
compositions.79–81 The addition of a PGM such as Pt
PM Fe–2.8Cr–4.7Al–1.5Ti–2.3B
Fine grains (or Ir, Pd, Rh, etc.)81–83 to a base Ni–(16–22)Al alloy
improves selective oxidation (i.e., decreases the for-
mation of Ni-rich oxide) such that the alloy becomes
a primary alumina former, similar to b-NiAl. How-
10 μm ever, TEM (or careful metallography) has found that
a thin Ni-rich oxide layer does form on these alloys
Figure 12 SEM secondary electron image of a polished even with 30% Pt.82,83 Improved selective oxidation
cross-section of Fe–2.8 at.% Cr–4.7Al–1.5Ti–2.3B after 50 h had been previously observed in NiCrAl þ Pt alloys
at 700 C. A thick Fe-rich oxide formed over a large grain in
by Felten,84 Pt-containing superalloys by Tatlock and
the substrate, elsewhere a thin Al-rich oxide formed over the
finer-grained regions (arrow). Adapted from Maloney, M. J. Hurd85 and with Pt on the surface of MCrAl alloys.86
Ph.D. Thesis, Massachusetts Institute of Technology, To illustrate the PGM effect on transient oxida-
Cambridge, MA, 1989. tion, Figure 13 shows the mass gain after a 500 h
isothermal exposure at 1000 C as a function of
the possibility of forming an alumina scale. Further- PGM addition in cast Ni–22Al Hf alloys.82,83 The
more, this effect will probably be most critical at lower mass gain for Ni–48Al þ Hf serves as a baseline for
temperatures 600–1000 C where selective alumina alumina formation with no Ni-rich oxide formation.
formation is more difficult because of slow diffusion The alloys without a PGM addition show a higher
rates and less critical at higher temperature >1100 C mass gain due to the formation of a Ni-rich scale with
where diffusion in the alloy is much faster. some contribution due to Hf internal oxidation.
Base: Ni–22Al
2 Air at 1000 ⬚C, 500 h
~4000 ppma Hf
Total mass gain (mg/cm−2)
1.6 ~500 ppma Hf

(a)
10 μm
No Hf
1.2
Pd addition
~800 ppma Hf
0.8
(b)
~200 ppma Hf
0.4 ~200 Hf
Ir addition NiAl + 0.05Hf

~600 Hf (Al2O3 baseline)
~700 ppma Hf
0
0 5 10 15 20 25 30
Pt or PGM addition (at.%) (c) 10 μm
Figure 13 Total mass gain (specimen þ spalled oxide) Figure 14 Light microscopy of polished cross-sections
after a 500 h isothermal exposure at 1000 C in laboratory air after 2000, 1 h cycles at 1100 C in O2 (a) Ni–22Al–10Pt,
for various precious metal (PM) and Hf additions to Ni–22Al (b) Ni–50Al–5Pt, and (c) Ni–22Al–10Pt þ 0.4Hf. Adapted
alloys. Data from Pint et al.82,83 from Pint, B. A.; More, K. L.; Wright, I. G. Oxid. Met. in
preparation.
With increasing PGM content, the mass gain

decreased. An Ir addition showed a similar benefit out of an Al-rich coating), and (3) their high strength
as Pt at the 5% level but 10–20% Pd, which is compared to b-NiAl and MCrAlY coatings82 is
significantly less expensive than Pt, was not as effec- thought to increase their resistance to thermal
tive due to formation of additional phases.83 As an mechanical fatigue. When the g–g0 coating contains
example of the alumina scale formed on these alloys, 5–10% Cr, either by interdiffusion from the substrate
Figure 14 shows scale cross-sections after 2000, 1-h or by addition as a part of a secondary coating pro-
cycles at 1100 C. Without a Hf addition, the alumina cess, the hot corrosion resistance is improved.91 One
scale on Ni–22Al–10Pt is similar in thickness to that concern about these coatings is the depletion of Pt
formed on b-Ni–50Al–5Pt (Figures 14(a) and 14(b)). due to interdiffusion with the super alloy, which is
The addition of 0.4 Hf to Ni–22Al–10Pt decreased the significant at 1100 C.92
oxide thickness but some internal oxidation of Hf was
evident (Figure 14(c)).
The g–g0 Ni–Pt–Al compositions are primarily 1.24.3 Transient Oxidation
considered for coatings on NiCrAl-based superalloys
1.24.3.1 Base Metal Oxide Formation
but some effort has been dedicated to alloy develop-
ment.87 By suppressing the formation of Ni-rich oxi- As already discussed in the previous section on selec-
des with the associated weak Al2O3–NiAl2O4 tive oxidation, during the initial or transient stage of
interface,88,89 these compositions can be used as oxidation, alumina-forming alloys will tend to form
bond coatings in thermal barrier coatings (TBCs). some base-metal oxide (e.g., NiO, FeOx, or M–Al–O
Such g–g0 coatings are attractive because (1) they spinel, etc.), see Figure 7. This is illustrated in
can be formed simply by interdiffusing a plated Figure 15 for Fe–Al alloys oxidized at 900 C. The
Pt layer into the surface of the super alloy (i.e., a Pt decreasing mass gain with increasing Al content is
diffusion coating rather than a Pt-modified b-phase due to a decrease in the initial transient Fe-rich oxide
aluminide), (2) Al interdiffusion is less of an issue formation.93 Because of the subsequent slow alumina
because the phase composition of the coatings is the scale growth, the mass gain after 5 kh is dominated by
same as the super alloy substrate79 and Pt appears to the mass gain after the first 500 h exposure (triangles
decrease the Al activity in the coating,90 thereby in Figure 15). In addition to the alloy composition,
creating a chemical potential gradient for Al to dif- the transient oxide thickness and phase composition
fuse from the substrate into the coating (rather than are affected by many factors associated with the
15
mass gains reflect the formation of a Ni- and Cr-rich
Lab. Air, 900 ⬚C transient oxide which is nonuniform (e.g., Figure 16(b)).
(Even at 1100 C, a thin (0.5–1 mm) external layer of
Open symbol-undoped, 5 kh
Filled symbol-doped, 5 kh Ni(Al,Cr)2O4 formed on this alloy with only 8.7 at.%
Al (see Table 2). But this exposure was sufficient to

10 develop a continuous, underlying a-Al2O3 layer.94)
With the preoxidation (<0.2 mg cm2 mass gain), no
Undoped further mass gain was measured during 10 kh expo-
+0 Cr
sures in air at 700 or 800 C.12
The differences in oxidation behavior among Co-,
5 Ni-, and Fe-based alumina-forming alloys were stud-
5 000 h
500 h ied at 1000 and 1200 C by Stott et al.60 The critical
Doped
+0 Cr factors identified were DAl and O permeability in the
+20 Cr alloy. With similar Cr and Al contents, Co-based alloys
formed thicker transient oxides that were attributed to
0
the fast growth rate of CoO and the relatively slow Al
10 15 20 25 30
Al content (at.%)
diffusivity in Co-based alloys. (Stated differently,
higher Cr and Al contents were needed in Co-based
Figure 15 Total mass gain (specimen þ spalled oxide) alloys to quickly form alumina.) The growth rate of
after 1 or 10, 500-h cycles at 900 C in air for Fe(Al) alloys as
a function of Al content. The doped alloys contained Ti, Hf,
FeO is faster, which is important at breakaway (see
Zr, or Y and the Fe–20Cr–10Al specimen contained no Section 1.24.5), but faster Al diffusivity in ferritic
dopant. For doped Fe–Al alloys (dashed line fit), the majority FeCrAl alloys resulted in rapid alumina formation.
of the mass gain occurred during the first 500 h (triangle) A less comprehensive study of Fe, Ni, and Co–Al
when Fe-rich oxide formed. The higher mass gains for the alloys by Hagel found similar results.95
undoped Fe–Al alloys (open squares) were due to an internal
oxidation/nitridation with nearly linear kinetics. This internal
The work of Kear et al.96 was one of the earliest
oxidation was not observed for FeCrAl. Data from Pint12 microstructure studies of the transient stage oxida-
and Pint et al.93 tion of NiCrAl. By reducing the oxygen partial pres-
sure, NiO and Ni-rich spinel-type oxide formation
thermodynamics and kinetics of oxide formation. was suppressed at 1000 C resulting in a mixture of
Examples of thermodynamic factors are temperature Cr2O3 and Al2O3. A subsequent characterization was
and the partial pressure of the oxygen (e.g., H2– reported by Smialek and Gibala.97 On FeCrAl, there
H2O or CO–CO2 gas mixtures). Some of the kinetic is typically very little transient base-metal oxidation.
factors are heating rate, alloy grain size (affecting Quadakkers and coworkers98 used the Fe and Cr in
the flux of Al in the alloy) and surface preparation the outer scale to mark the original metal interface
which can affect the number of fast-diffusion paths and study the alumina growth mechanism. Figure 17
in the near surface region (e.g., grinding versus acid gives an example of the type of transient scale on
etching). ODS FeCrAl after 2 h at 1000 C in O2.40 The Fe and
Oxidation temperature is a critical factor for tran- Cr maps (Figures 17(b) and 17(c)) show that Cr and
sient oxide because of the slow-formation of alu- Fe were concentrated in a thin (100 nm) outer layer
mina below 800 C. (A slow heating rate to a higher after these conditions.
temperature can have a similar effect.) Alumina may
be the thermodynamically stable phase but kinetics
1.24.3.2 Cubic Alumina Phases
often dominate, allowing rapid base-metal oxidation.
A simple strategy to enhance alumina formation is to One of the most critical transient oxidation issues for
preoxidize at a higher temperature where an alumina alumina-forming alloys is the formation of metastable
scale forms more readily. Figure 16 gives an example cubic Al2O3 phases (e.g., g, d, y phases). Their forma-
of this strategy for alloy 214 foil (50 mm thick) pre- tion can be considered as a transient phenomenon
oxidized for 2 h at 1100 C. Because of the difficulty in because, given time and temperature, these metastable
forming alumina at low temperatures, the parabolic phases generally transform to the stable a phase. How-
rate constant after 5000 h was higher at 700 C ever, because these cubic phases are more defective
(8
1016 g2 cm4 s1) than at 800 C (2
1016 g2 than a, they allow faster diffusion and grow at a faster

1016 g2 cm4 s1) (Figure 16(a)). The observed rate.95,99,100
0.3 W
800 ⬚C
Specimen mass gain (mg cm–2)
kp = 2⫻10–16 g2 cm–4s–1
0.2 Alloy 214 ~50 μm foil
700 ⬚C
kp = 8x10−16 g2 cm–4s–1 Alumina

0.1
650 ⬚C
kp = 3x10−16g2 cm–4s–1
Preoxidized
700 ⬚C and
800 ⬚C
0.0
0 1000 2000 3000 4000 5000

(a) Time (h) in 500 h cycles ODS FeCrAl
(a) 100 nm
Cu plate
(b) Alloy 214

10 μm
(c) Pre-oxidized alloy 214

Fe (b) Cr (c)
Figure 16 (a) Specimen mass gains for alloy 214 (NiCrAl,
see Table 2) foil (50 mm) with and without preoxidation
exposed for 500 h cycles in laboratory air at 650 C, 700 C,
and 800 C. Polished cross-sections for the specimens
shown in (a) after exposure in laboratory air at 800 C,
(b) as-received foil specimen oxidized for 5000 h, and
(c) preoxidized (2 h at 1100 C in dry O2) foil after 10 000 h.
Data from Pint.12
Y (d) Ti (e)
The formation of cubic alumina phases on Figure 17 Cross-section of the a-Al2O3 scale formed on
alumina-forming alloys was discussed in the early PM2000 (ODS FeCrAl, see Table 1)) after 2 h at 1000 C in
literature.95 However, the possibility of forming dry O2 (a), STEM high angle annular dark field image and
these metastable phases, rather than a, was not dis- EDS X-ray maps from the box in (a): (b) Fe, (c) Cr, (d) Y, and
(e) Ti. Adapted from Pint, B. A.; More, K. L. J. Nucl. Mater.
cussed for many years and probably led to the con- 2008, 376, 108–113.
troversy regarding the growth mechanism of alumina
scales in the 1980s. The observation of growth by
outward Al transport was assumed to be for an morphologies on commercial FeCrAl foil are shown
a-Al2O3 scale but, based on the oxidation time and in Figure 18. The outward growing features are
temperature, was likely y-Al2O3.101 This issue will be minor at 700 C but are very evident at 800 C
further discussed in Section 1.24.4.2. (Figure 18(b)). After long-term exposure at 900 C,
The hallmark of cubic alumina phases is the gas- there is little evidence of outward transport and
side blade or whisker morphology often observed. the scale has completely transformed to a-Al2O3
Much of the cubic alumina scale characterization (Figure 18(c)). However, aluminized foils (alumi-
work has been done on b-NiAl,102–105 because of the nized to increase the Al reservoir) showed more
cubic epitaxy with the b phase and because Fe and Cr distinct remnants of outward growth (Figure 18(d)).
in the substrate tend to help nucleate a phase.100,106 This is consistent with work on NiAl where surface
However, at lower temperatures (<950 C), cubic enrichment of Al by ion implantation led to more
alumina scales form even on FeCrAl. Typical surface cubic alumina formation.107 Cubic alumina phases
1.2
kp = 1 ⫻ 10–16 g2 cm–4 s–1
kp = 6 ⫻ 10–14 g2 cm–4 s–1
1.0 800 °C

First 2500 h
0.8
900 °C
0.6
1
–4 s–
(a) 1μm (b) – 15 g2 cm
0.4 –2 ⫻ 1 0
kp = 0.9
0.2 kp = 1 ⫻ 10–15 g2 cm–4 s–1 700 °C
0.0
20 30 40 50 60 70 80 90 100 110
Square root time (h0.5)
Figure 19 Specimen mass gains for aluminized FeCrAl þ
(c) 2 μm (d)
Mm foil (50 mm) exposed for 500 h cycles in laboratory
air at 700, 800, and 900 C. The higher rate for cubic
Figure 18 SEM secondary electron plan view images of
alumina phases is seen at 700 C and initially at 800 C.
the alumina scale morphology formed on 50 mm
Data from Pint12 and Dryepondt et al.110
commercial FeCrAl foils after exposure in laboratory air: (a)
FeCrAl þ Y,Hf, 700 C 10 kh, (b) FeCrAl þ Y,Hf, 800 C,
10 kh, (c) FeCrAl þ Mm (mischmetal), 900 C, 50 kh, and (d)
aluminized FeCrAl þ Mm, 900 C, 50 kh. Adapted from Pint,
B. A. Unpublished oxidation data, ORNL, Oak Ridge, TN, gains are relatively low. However, with the foil there
1994–2008. is a limited Al reservoir. Figure 20 shows the starting
and final Al contents in these foils.110 The starting Al
such as y also have been observed on bulk FeCrAl profile in the aluminized FeCrAl is not flat because of
alloys at 900 C.104,108,109 For Ni-based alloys, cubic the aluminization step. As expected, the Al content
aluminas are less common because they tend to form drops with exposure time at 900 C with very little Al
Ni-rich oxides first and the nucleation of alumina remaining after 50 kh. However, because of the faster
beneath this layer apparently favors a-Al2O3 forma- growth of the cubic alumina scale, the foil was almost
tion. The alloy 214 (NiCrAl) foils in Figure 16 do completely depleted of Al after only 10 kh at 800 C
not show the high parabolic rate constants observed (Figure 20) is similar to other observations for
on FeCrAl foils (Figure 19). FeCrAl foils in this temperature range.111
The study of the alumina phase may seem an Because of the use of these foils in automotive
academic exercise except that there is a strong effect catalytic converters and other applications operating
on performance of commercial MCrAl foils in the in this temperature range, there is considerable con-
700–950 C temperature range. Figure 19 shows the cern about the increasing Al depletion due to tran-
kinetics for commercial aluminized FeCrAl foil oxi- sient oxide formation. Various treatments have been
dation in air at 700–900 C. At 900 C, the initial developed to speed the transformation to an a-Al2O3
jump in mass gain is due to the initial formation of scale.112 Small ion dopants such as Ti are known to
the cubic phase. Based on the mass gain, more Al was speed the transformation from y to a-Al2O3.113,114
consumed during the first 500 h exposure than during Therefore, a TiO2 slurry was an effective surface
the next 10 kh of exposure, after the cubic phase had modification.112 Studies have shown a faster transfor-
transformed to a slower growing a-Al2O3 scale. mation to a-Al2O3 in the presence of H2O,112,115
Because of the faster growth of cubic-aluminas, the although an earlier study found the opposite
parabolic rate constant at 700 C (1
1015 g2 cm4 effect.116 Thus, a higher temperature pretreatment
s1) is comparable to the rate at 900 C (Figure 19). in H2–H2O is another strategy to quickly form an
At 800 C, where the blade structure was most evi- a-Al2O3 scale.
dent (Figure 18(b)), the mass gain after 10 kh is Once a continuous a-Al2O3 scale has formed, the
higher than at 900 C, although the rate decreased transient stage of oxidation is completed and the
significantly during the last 2.5 kh. All of these mass steady-state period of oxidation begins.
implantation.6,118 The amount of addition can vary

40
from the ppm level to 0.5% and one of the significant
Aluminized FeCrAl
(tradename durafoil) issues is optimizing the RE benefit to achieve the
35
longest alloy lifetime.
Normalized Al content (at.%)
30
1.24.4.1 Scale Growth Rate
25
The previous section on transient cubic alumina for-
Starting mation began the discussion about growth rates.
20
(circles) Starting with the derivation of Wagner,119 the
10 kh, 900 ⬚C steady-state rate limiting step should be solid state
15 (diamonds)
diffusion through the alumina scale. Therefore, as
the diffusion distance increases with time due to
10
the formation of new oxide, the reaction kinetics are
expected to be parabolic. Typically, parabolic rate
5 50 kh, 900 ⬚C constants have units in a form such as g2 cm4 s1
10 kh, 800 ⬚C
for changes in mass data. The rate also could be
0 expressed in terms of oxide thickness change or
0 5 10 15 20 25 30 35 40 45 50 55
metal loss but mass change is the easiest parameter
Distance (μm)
to quantify. For alumina in the temperature range
Figure 20 Electron probe microanalysis using of most interest (1050–1250 C), the diffusion is
wavelength dispersive spectroscopy Al profiles of the
along short-circuiting paths, likely the scale grain
aluminized FeCrAl þ RE foil after various exposures in
laboratory air. The higher depletion after 10 kh at 800 C boundaries. As the oxide grain size increases with
compared to 10 kh at 900 C is consistent with the mass oxidation time, it has been argued that the kinetics
gain data in Figure 19 and is attributed to the formation of a are actually cubic.120,121 The steady-state compres-
faster-growing metastable cubic alumina scale. Data from sive growth stresses in an alumina scale122–124 also
Pint12 and Dryepondt et al.110
could cause a change in the kinetics as Evans and
coworkers argued for ZrO2.125 However, as will be
shown for lifetime predictions of bulk alloys in the
1.24.4 Steady-State Oxidation next section, the onset of scale spallation results in Al
consumption kinetics which is nearly linear with
At the completion of the transient stage of oxidation, time. Nevertheless, parabolic rate constants are a
a complete alumina scale would have formed and standard and conventional way to quantify and com-
subsequent oxidation should be controlled by the pare scale growth rates.
properties of this layer. This section will consider As a diffusion-controlled process, rate constants
the performance of this layer and the factors that are expected to follow an Arrhenius-type relationship
affect it. This will be a general discussion with exam- with temperature. Figure 21 shows an Arrhenius plot
ples given for various materials. The issues rele- of rate constants, mostly for RE-doped FeCrAl. As
vant to steady-state alumina formation have been mentioned in the previous section, the rates are
reviewed several times.9,46,117 Nominally, if the scale higher for cubic alumina scales such as y-Al2O3.
remains dense and adherent, steady-state oxidation Since cubic alumina is metastable, these scales are
should continue until the substrate is depleted in Al. only present at lower temperatures, <1000 C. At
This depletion will occur more rapidly if scale spall- higher temperatures, only a-Al2O3 scales are formed.
ation occurs, requiring reformation of the oxide. Table 3 gives some values for FeCrAl with and
Thus, adhesion of the oxide is also an important without RE dopants and the associated activation
issue. Most commercial alumina-forming alloys con- energies over the range 1000–1300 C. The activation
tain a RE addition, for example, Y, Hf, Zr, La, Ce, etc. energies for undoped a-Al2O3 are lower than that
which strongly affects the steady state oxidation for RE-doped a-Al2O3. The activation energy for
behavior. The role of RE additions will be discussed y-Al2O3 is also similar to undoped a-Al2O3, 235 kJ
throughout. These dopants can be added as alloy mol1 for FeCrAl foils which is comparable to the
additions, oxide dispersions or less effectively, as 200 kJ mol1 found for the cubic alumina scales
surface treatments or coatings such as slurries or ion formed on b-NiAl.99
Table 3 Parabolic rate constants (g2 cm4s) and activation energy (kJ mol1) for several FeCrAl alloys. Data from Pint.129
Temperature Undoped FeCrAl 0.08Y 0.038Hf þ Y 0.020 Hf þ Y

13 13 14
1000 C
3.1
10 (1.1–2.0)
10 4.8
10 2.5
1014
1100 C 3.2
1012 (4.2–4.4)
1013 1.4–3.7
1013 0.98–1.5
1013
1200 C 1.4
1011 7.5
1012 5.6
1012 3.5
1012
1300 C (2.4–3.3)
1011 (3.6–4.4)
1011 2.2–2.3
1011 1.1
1011
Activation energy (kJ mol1) 244 352 367 366
As can be seen in Table 3, RE additions can 1300 8C 1100 8C 900 8C 700 8C

decrease the scale growth rate. However, this RE 1E–10
effect was not widely accepted until the past decade,
particularly when it was discovered by Doychak126
that Hf-doping in aluminides decreased the parabolic 1E–11

rate constant by a factor of 10.127–129 One of the
complicating factors is the additional mass gain in θ (NiAl)
RE-doped alloys due to the internal oxidation of 1E–12 (200 kJ mol–1)
the RE addition. Figure 22 illustrates this issue com-
paring the mass gains for FeCrAl and FeCrAl–0.08Y
1E–13
at 1100 C. Little difference was observed after a (235 kJ mol–1)
200 h exposure. However, for Y2O3-dispersed FeCrAl θ ( FeCrAl)
(PM2000) a slower rate was measured. For PM2000, 1E–14
there was no mass gain associated with the oxidation
of Y because the Y was added to the alloy as an oxide. α (doped)
By minimizing the RE addition to limit internal RE 1E–15 (330 kJ mol–1l)
oxidation, the rate constant was further decreased
(Figure 22). Figure 23 illustrates the effect on the
oxide thickness after 1000 h at 1200 C by minimizing 1E–16
the RE addition.94,129 In these backscattered SEM 0.0006 0.0008 0.0010
images, the bright phases in the a-Al2O3 scale are 1/temperature (K–1)
RE-rich oxides which are defects in the scale allow-
Figure 21 Summary of parabolic growth rates reported in
ing fast O transport.130 Minimizing the RE addition,
the literature for RE-doped FeCrAl and calculated activation
minimizes the number of these particles as well as energies. Rates can be split into those for a-Al2O3 and
RE-rich oxide penetrations that penetrate deeper y-Al2O3. For comparison, the y-Al2O3 rate data for NiAl is
into the substrate, typically referred to as pegs. included.99 The highlighted rates (diamonds) represent the
Note that the lowest RE addition resulted in the lowest reported rates at 1100 C and 1200 C for Hf-doped
NiAl. Other data points from Pint12 and Pint.129
thinnest alumina scale on FeCrAl (Figure 23).
Several RE optimization studies131–133 have been
conducted to determine the minimum alumina than smaller ions (Figure 25).11 A similar relation-
growth rate possible, as in Figure 24. With very low ship has not been established for alumina-forming
RE additions, little benefit was observed. With very alloys,135 where ion size does not predict the strong
high additions, the internal oxidation and entrapped beneficial effect of Hf. However, larger ions with high
RE-oxides increase the rate. However, the optimal oxygen affinity, for example, Y, Zr, Hf, La, Ce, are all
RE levels found in laboratory scale heats may not be beneficial in reducing the scale growth rate. Smaller
viable in commercial alloys with standard impurity dopant ions such as Ti, Nb and Ta do not strongly
levels of interstitials such as S, C, and N.10 For ODS affect the scale growth rate.135
alloys, the level of Y2O3 may be selected to optimize As mentioned earlier, the parabolic rate constant
creep strength rather than a lower level that may is related to the rate of solid state ionic diffusion
minimize the oxidation rate.109,134 The type of dop- through the scale. However, there is no simple rela-
ant can affect the oxidation rate. A clear effect of tionship and a number of authors have had difficulty
dopant ion size was found for chromia-forming alloys relating the published a-Al2O3 diffusion data to alu-
where larger ions decreased the rate more effectively mina scale growth rates.136 Using the derivation of
1.2
FeCrAl + 0.08Y 3.7 3 10–12 g2 cm–4s–1 0.8
1.0 undoped CoCrAl + Hf
FeCrAl 3.2 3 10–12 g2 cm–4s–1
Mass gain (mg cm–2)

(circles) 0.6
0.8
CoCrAl + Y
0.6
0.4
PM2000 NiAl + Hf
0.4 1.8⫻10–12 g2 cm–4s–1 0.2
100 h at 1100 °C
0.2 FeCrAl + 0.02Hf, Y 0
1.3⫻10–12 g2 cm–4s–1 0 0.2 0.4 0.6 0.8 1
0.0 Percentage Y or Hf (mass%)
0 1 2 3 4 5 6 7 8 9 10 Figure 24 Mass gain for CoCrAl and NiAl after 100 h at
Square root time (h0.5) 1100 C with various reactive element additions. CoCrAl
Figure 22 Specimen mass gain for various FeCrAl alloys data are from Figure 3 in Whittle and Stringer,1 NiAl data are
(Table 2) during isothermal oxidation at 1100 C in dry from Pint.12
flowing O2. The addition of Y did not decrease the reaction
rate compared to undoped FeCrAl due to the additional
mass gain associated with the internal oxidation of Y. When
2.0
the addition is made as Y2O3 or lower alloy additions, lower Nil Al Ta
oxidation rates were measured. Data from Pint.12 1.8 Sc Ca
Mn
1.6
Epoxy Nb
1.4 Ti
Mg
1.2 Cr
Alumina
1.0 Hf
0.8 Zr
0.6 Base alloy: Ni-25 wt% Cr
RE-rich oxides most additions: 0.1at.% Y
0.4 Tb
La
(a) FeCrAl + 0.08%Y 5 μm 0.2 Gd SmNd
Ce
Epoxy 0.0
0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1
Dopant (oxide dispersion) cation ion size (Å)
Figure 25 Total mass gain for Ni–25 wt% Cr alloys with
various oxide dispersions added at a nominal level of 0.1
cation% after 100 h at 1000 C. The mass gain decreased
(b) FeCrAl + 0.020%Hf + Y for larger cation additions. The addition of an Al2O3
dispersion had little effect on the scale growth rate. Data
Figure 23 SEM backscattered electron images of the from Pint.11
polished cross-section of the scale formed after 10, 100 h
cycles at 1200 C on FeCrAl with additions of (a) 0.08%
Y and (b) 0.020% Hf/Y. Adapted from Pint, B. A.; More,
K. L.; Wright, I. G. Mater. High Temp. 2003, 20, 375–386. where kp is parabolic oxidation rate constant, P 0O2 is
the decomposition pressure of Al2O3, P 00O2 is the
Wagner119 and the interpretation of Yurek,137 the ambient oxygen pressure, and DAl and DO are the
parabolic oxidation rate constant for the formation respective self diffusion coefficients for aluminum
of Al2O3 can be related to the solid state diffusion and oxygen. This equation can be greatly simpli-
constant by the following equation for Al2O3: fied137 given the observation that the alumina growth
ð 0 rate is independent of oxygen content in the environ-
1 P O2
kp ¼ ½3=2DAl þ DO d ln PO2 ½3 ment, P 00O2.108 Extrapolating published diffusion
2 PO00
2 rates for bulk a-Al2O3138,139 to 1200 C, Table 4
Table 4 Parabolic rate constants at 1200 C calculated 100% a-Al2O3.150 In the undoped scale (Figure 26(a)),
from diffusion data from the literature138,139 and compared the 16O is enriched in both the inner and outer scale
to measured values for RE-doped and undoped alumina-
indicating both Al and O transport. With the addition of
forming alloys7,127,203
Zr, only a minor 16O peak was observed at the specimen
Predominant diffusion pathway kg (g2 cm4 s1) at surface. A stronger 16O peak was observed underneath
and species 1200 C the outer 18O peak indicating predominantly inward
Calculated O boundary transport.
Grain boundary diffusion of O 1.4
1012 The explanation for the RE effect on transport has
Grain boundary diffusion of Al 1.5
1011 long been thought to be the segregation of RE ions to
Lattice diffusion of O 9.3
1014 scale grain boundaries in both alumina and chromia
Measured
RE-doped FeCrAl (3.0–4.4)
1012
scales.8,154–157 Segregation of dopant ions is com-
Hf-doped NiAl (0.76–1.0)
1012 monly observed in bulk alumina because of the
Undoped FeCrAl, or NiAl (0.9–1.4)
1011 low solubility of many cations in the closed-packed
a lattice.158 Examples of Y and Ti grain boundary
segregation are shown in Figure 17. The mechanism
shows that parabolic rate constants can be calculated by which RE ions inhibit cation transport could be
which are of the same order of magnitude as those blocking or site competition or a space charge
measured.140 A recent literature review concluded effect.159 However, the RE valence has not been
that diffusion in a-Al2O3 is not fundamentally well shown to be an important variable.135 A dynamic
understood, making it difficult to correlate diffusion blocking mechanism was suggested, where large,
data with oxidation data.141 slow-diffusing RE ions on the boundaries inhibit the
diffusion of Al.8 This mechanism is based on the
observed outward transport of RE dopants dur-
1.24.4.2 Mechanism
ing oxidation. Figure 27(a) shows a schematic of
While there was some confusion about the growth this phenomenon which is even observed on oxide-
mechanism of Al2O3 scales in the 1980s, the issue dispersed alloys like PM2000. After extended times at
seems to be largely settled today. There were several temperature, RE-rich oxide particles nucleate at the
complicating issues. First, Pt does not wet the surface metal–oxide interface and grow during expo-
of a-Al2O3 (based on work with Pt photolithographic sure.8,135,160 Figures 27(b) and 27(c) show an example
markers142). Thus, traditional inert Pt markers used to at 1100 C where the faceted Y3Al5O12 particles at the
determine the scale growth mechanism floated on the gas interface increase in volume with time and do not
alumina scale surface whether the scale grew inward just coarsen. This mechanism explains why smaller
or outward. In all cases, the markers showed inward dopants, Ti, Nb, and Ta, segregate to a-Al2O3 grain
growth.61,143 For Cr2O3 scales, inert markers showed a boundaries but do not have the same effect on the
predominant Cr outward growth without RE dopants scale growth rate as larger RE dopants.135 These smal-
and predominantly O inward growth with RE dopants ler ions rapidly diffuse outward and thus do not effec-
and this was confirmed, without controversy, using tively inhibit Al boundary transport.
18
O tracer experiments.144,145 Second, the transient, Recently, there has been renewed interest in
cubic Al2O3 phase formation confused some of the studying the effect of RE dopants on diffusion and
early 18O studies.101,146 The issue was later clarified, sintering in bulk ceramics161,162 because of the obser-
showing that y-Al2O3 scales predominantly grow by vation that RE dopants like Y and La strongly
an Al outward mechanism,104 which is intuitive given improve the tensile creep strength of polycrystalline
their morphology147 (Figure 18(b)). Third, unlike alumina.163 The boundary electronic structure model
Cr2O3, there is no distinct difference between un- proposed as an alternative to a simple dopant ion
doped and RE-doped a-Al2O3, both show some sized model to explain the behavior of various
growth by inward O diffusion. With RE-doping, the dopants164 may be relevant to the effect of RE
outward Al transport is diminished and the scale grows dopants on scales. Similar to results for bulk ceramics,
predominantly by inward transport. This has been an increase in the creep resistance of Hf-doped alu-
shown by several 18O tracer studies.134,148–153 An mina scales has recently been demonstrated.165
example of the tracer profiles is shown in Figure 26. As mentioned previously, the O and/or Al trans-
The 18O tracer was used first followed by 16O. The first port is likely along the scale grain boundaries because
stage was 20 min at 1200 C to ensure that the scale was bulk diffusion rates in a-Al2O3 are too slow to
0.9
Undoped
0.8
%O-16
0.7 %O-18
Oxygen isotope fraction 0.6
0.5
0.4
0.3 20 min in 18O

40 min in 16O
0.2
0.1
0 20 40 60 80 100
(a) Sputter time (min)
0.9
With 0.11 % Z r
0.8
0.7
Oxygen isotope fraction
0.6
%O-16
0.5
%O-18
0.4
0.3
0.2
0.1
0 20 40 60 80 100
(b) Sputter time (min)
Figure 26 Secondary ion mass spectroscopy (SIMS) sputter depth profiles of 100% a-Al2O3 scales formed isothermally at
1200 C after 20 min in 18O and 40 min in 16O on (a) undoped NiAl and (b) NiAl þ 0.11%Zr. Data from Pint et al.150
account for the observed growth rates. Figure 17 scale growth rate167–169 and electrical properties
shows a typical cross-section of an alumina scale measurements indicate that x is 2.170
where very few dislocations or other defects were A final point about scale growth is that there
observed that might accommodate ionic transport. is now considerable interest in studying scale
Atkinson and coworkers166 calculated that a high growth stresses, particularly with the advanced char-
dislocation density was needed for dislocation trans- acterization techniques now available for in situ mea-
port through the grains. It could be argued that the surements.124,165,171–174 These in situ results are
dislocations were present at a temperature but were considerably better than measuring the stress after
annihilated during cooling.141 However, the cooling cooling to room temperature by photostimulated
rate for the thin scale is rapid which would not luminescence and attempting to separate the growth
support this hypothesis. Experimental work has also stress from the thermal stress.175,176 This recent in situ
indicated that, rather than neutral species, transport work has confirmed the observation that these growth
in alumina scales is via charged ions, that is, Ox or stresses can be tensile,171,172 rather than the com-
Alþy, as applied voltages have been shown to alter the mon assumption of compressive growth stresses.177
Gas interface
RE-rich oxide particles
nucleate and grow
1 atm
Equiaxed
grains
RE ions diffuse out O2

along scale grain
boundaries
Columnar
grains
Alloy 10–25 atm

RE-rich oxide dispersion
(a)
YAG
YAG
YAG
YAG
(b) 2 000h 1 μm 10 000 h (C)
Figure 27 (a) Schematic representation of the outward diffusion of RE ions from oxide particles in the substrate to
incorporation in the scale.8 (b) and (c) show plan view SEM secondary electron images of the scale formed on PM2000 (ODS
FeCrAl) after exposures in air at 1100 C for (b) 2000 h and (c) 10 000 h.12 The size and volume fraction of Y-rich oxide particles
such as Y3Al5O12 (YAG) has increased with exposure time. Titanium, also present in the alloy, is highly enriched at the gas
interface.
The stresses can also decay with time as they are (e.g., level of polishing for laboratory specimens)
accommodated by substrate or scale creep.165 Never- often affects the morphology.181 At higher magnifica-
theless, these stresses are generally <10% of the ther- tion, fine ridges can be seen that also suggest outward
mal stresses generated on cooling due to the CTE transport (Figure 28(a)). (These ridges are not the
mismatch between metal and scale (see Section same as the ridges on b-NiAl formed by the y–a
1.24.4.4.4). Thus, they are often ignored when calcu- phase transformation.102) Undulations of the undoped
lating the effect of stress on adhesion, eqn [1].19 oxide scales also occur on FeCrAl and Fe(Al), some-
times accompanied by equivalent deformation of
the underlying metal but also due to the formation
1.24.4.3 Morphology
of interfacial voids.7,182–187 The undulations are not
Figures 28 and 29 give some examples of typical found when undoped FeCrAl substrates have low
scale morphologies on alumina-forming alloys that S contents185 or an Al2O3 oxide dispersion is added
do not form a base metal transient oxide and are (Figure 29(a)).187,188 These differences suggest that
commonly observed in the literature.7,135,178–181 The the interfacial strength or mechanical properties of
surface morphology of transient cubic aluminas was the oxide and scale may control this phenomenon.188
shown in Figure 18. Without a RE addition, undoped In the cross-section, the scale typically has an
a-Al2O3 often shows a wrinkled morphology with equiaxed grain structure with or without wrinkles
oxide whiskers at the gas interface7,117,143,178–181 (Figure 29(a)).
which hint at the outward transport of Al. As alumina With a RE addition, the surface ridges and whis-
scales are usually thin, the original surface roughness kers are not present, indicative of the RE’s role in
less angular.135,189 The most distinctive RE effect

on morphology is on the grain structure where the
change in growth mechanism results in a columnar
grain structure with fine equiaxed grains near the gas
interface.178 In many alloys, the layer of equiaxed
grains does increase in thickness with time because
of the small amount of outward Al transport.190 How-
ever, the scale thickening is predominantly due to an
increase in the length of the columnar grains. Unlike
(a) 1 μm (b) 2 μm some studies showing no texture, an x-ray pole figure
on the scale formed on PM2000 at 1200 C showed a
Figure 28 SEM secondary electron images of the a-Al2O3 preferred growth direction.191 The grain structure of
scale formed on undoped FeCrAl at 1200 C in dry
RE-doped a-Al2O3 is remarkably similar indepen-
flowing O2. (a) Plan-view image of undoped FeCrAl after
2 h and (b) cross-section of the scale formed on dent of the RE type and substrate, that is, M–Al or
Al2O3-dispersed FeCrAl after 100 h. In (a), the ridges can M–Cr–Al, likely because the scales on all these sub-
be referred to as intrinsic ridges to differentiate them strates grow by a similar mechanism. Some less com-
from the extrinsic ridges formed on NiAl due to the mon RE dopants are shown in Figure 29 to illustrate
alumina phase transformation. Adapted from Pint,
that a wide range of dopants produce similar effects.
B. A. Unpublished oxidation data, ORNL, Oak Ridge,
TN, 1994–2008. Varying the type and amount of dopant can alter the
size and structure of the columnar grains.192
1.24.4.4 Adhesion
The most critical performance issue for alumina-
forming alloys in long-term high temperature appli-
cations is scale adhesion. Adherent scales grow with
parabolic (or even cubic) kinetics which results in
very slow Al consumption rates from the substrate.
(a) 1 μm However, when the scale spalls, the oxide must
Gas regrow or heal the spalled region, resulting in a
significant increase in the rate of Al consumption
Scale
due to the initial high rate of oxidation. As described
Metal 1 μm in Section 1.24.1, the major cause of scale spallation
(b)
is the thermal or cooling stress due to the CTE
mismatch (Da) between the substrate and oxide, see
eqn [1]. An additional issue is the formation of voids
or other defects at the exposure temperature that lead
to decohesion. Without defects, the scale will eventu-
ally reach a critical thickness, xf, where the strain
(c) 1 μm energy (W*) is sufficient to cause spallation, eqn [2].
Defects will decrease the interfacial adhesion result-
Figure 29 SEM secondary electron images of fracture
ing in failure at a lower scale thickness.
cross-sections of scales formed on oxide dispersed FeCrAl
after 2 h at 1200 C in dry flowing O2. (a) Al2O3, (b) Ho2O3, For undoped alumina-formers, void formation at
and (c) Er2O3. Adapted from Pint, B. A. Unpublished the metal–alumina interface is typically one of the
oxidation data, ORNL, Oak Ridge, TN, 1994–2008; Pint, critical issues that determines adhesion. Provenzano
B. A.; Alexander, K. B. J. Electrochem. Soc. 1998, 145, and coworkers193 found in CoCrAl compositions that
1819–1829. Arrows in (a) indicate internal voids.186
the interfacial void area increased with temperature
from 700 to 1000 C. Typically, these voids are
suppressing outward Al transport.148–153 As shown in thought to be formed by a Kirkendall-type mecha-
Figure 27, instead of ridges, fine RE-rich oxide par- nism either due to diffusion in the alloy or the oxide
ticles form on the surface. Additions such as Ti will or both. Because of the outward transport of Al in the
tend to make the surface grains more rounded and undoped scale, a counter current vacancy flux moves
towards the metal-scale interface. Also, because of the breakaway for various FeCrAl models and commercial
faster undoped scale growth rate, there is increased alloys, all with 10%Al. In order to accelerate the test,
diffusion in the substrate with Al diffusing to the 1 h cycles were used at 1200 C. The time has been
interface and other elements moving in the opposite normalized to a 1.5 mm thick specimen. Thus, all the
direction. Contributing to the nucleation or growth specimens begin with a similar Al reservoir and the
of interfacial voids may be surface-active interstitial rate of scale growth and adhesion are thus key para-
impurities such as S,186,194 which will be discussed meters in determining lifetime (see Section 1.24.5).
further in Section 1.24.4.4.2. An undoped wrought FeCrAl specimen, which spalls
One problem with discussing scale adhesion is readily during thermal cycling, had a lifetime of only
that it is difficult to quantify.195,196 A well-formed 300, 1-h cycles. From Figure 31, it is obvious that
scale is more tenacious than most epoxies used in RE additions play a critical role in increasing scale
quantifiable pull tests with an interface fracture adhesion and thus increasing life.
toughness of >90 MPa.197 Specimen mass gain curves
can be used with a decrease in mass suggesting 1.24.4.4.1 Reactive elements
spallation. However, consider some NiCrAl alloy The main beneficial effect of RE additions (e.g., Y, Zr,
examples in Figure 30. The specimens that spall Hf, La, Ce) on alumina-forming alloys is the
and rapidly lose mass are obvious but a higher speci- improvement in scale adhesion. Figure 31 shows
men mass gain does not necessarily indicate a more that the addition of 0.08 Y increased the lifetime of
adherent oxide. A higher mass gain could indicate a FeCrAl by a factor of 30 compared to undoped
faster growing oxide coupled with minor spallation. FeCrAl. Most studies have observed a reduction in
The lower mass gain for NiAl þ Hf is due to a thin- the interfacial void formation. A number of hypoth-
ner, adherent oxide. Furthermore, some oxides eses have been proposed to explain the RE effect on
remain attached but are friable and therefore not void formation. Tien and Pettit143 proposed that the
protective,198 resulting in a false indication of adhe- RE dopants in the alloy formed a vacancy complex
sion. Many early RE optimization studies relied on a that prevented vacancies from coalescing at the
relatively short term cyclic oxidation data to con- metal scale interface. Using a similar logic, Kumar
clude that a certain composition was superior.131,133 et al.199 suggested that RE internal oxides in NiCrAl
A more absolute way of quantifying scale adhesion acted as a vacancy sink. However, there is no
is cyclically testing specimens through the steady-state experimental evidence of these hypotheses. The
oxidation period until they reach breakaway oxidation formation of RE internal oxides or ‘pegs’ was thought
(see Section 1.24.5). Figure 31 shows the time to to create a mechanical interlock that improved
scale adhesion.61,131,132,200 While a small number of
1.0
Y + Hf: (60 Y, 460 Hf ppma)
Y/S = 12 Hf/C = 1.1 Dopants in FeCrAl
0.8 None
(3 specimens) N5 superalloy ODS None
0.6 7Cr–13Al + Hf,Y ODS 0.2 Y (MA956)
ODS 0.2 Y (MA956HT)
ODS 0.22 Y (PM2000)
0.2Y (APMT)
0.4 0.06 Zr (APM)
0.08 Y
NiAl + Hf 0.035 Y
0.2 0.05 Hf
Base alloy: 0.1 Hf
0.04 Mischmetal
0.0 Ni–20Cr–19at.%(10 wt%)Al 0.14 Hf, Y
0.038 Hf, Y
0.020 Hf, Y
–0.2 0.018 Hf, (Y)
Hf-doped Y-doped 0.060 Zr, (Y)
(600 ppma Hf) (500 ppma Y)
–0.4 0 2000 4000 6000 8000 10 000 12 000
0 100 200 300 400 500 600 700 800 900 1000 Normalized lifetime in 1 h cycles at 1200 °C
Number of 1 h cycles at 1100 °C
Figure 31 Normalized time to breakaway oxidation
Figure 30 Specimen mass change of various Ni–20Cr–19Al (as an indicator of scale adhesion) for various FeCrAl
alloys during 1 h cycles at 1100 C in dry flowing O2. The alloys (see Table 2) at 1200 C in 1 h cycles. The lifetime is
codoping of Y and Hf produced better scale adhesion normalized to a 1.5 mm thick specimen. Model wrought
compared to Hf or Y additions alone. Data from Pint12 and alloys doped with Y and Hf showed the longest lifetime.
Pint et al.94 Data from Pint.12,129
pegs is likely to be beneficial in preventing inter- 3000 Sr

facial (metal–scale) crack propagation, pegs are not 2750
Hf
Normalized time to breakaway (h)

necessary as ODS alloys achieve good scale adhesion 2500
Exposure: 1 h cycles at 1200 °C La
without forming pegs. For example, the lifetimes for 2250 Normalized to 1.5 mm thickness
several ODS alloys are shown in Figure 31. 2000 Yin MA956
Another hypothesis was that the finer grained RE- 1750
Lu Nd
doped alumina scale resulted in a more plastic oxide Y Ba
1500 Ho
that accommodated the strain better.61,201 Ironically, 1250
Er
Gd
Nb Zr Dy
the recent work on RE-doped bulk alumina161,163,164 1000 Mg Tb
and scales165 has shown that RE-doping actually 750 Ta

Ti
Tm
Sm
increased the creep strength by several orders of 500 Life for cast, undoped Yb
FeCrAl Mn Ca
magnitude. The change in mechanical response 250
Si V Sc Ce Eu
of the oxide scale with Y doping can explain the 0
Al In Te Bi
higher residual stresses measured in Y-doped alu- 0.2 0.4 0.6 0.8 1.0 1.2 1.4
mina scales.176 Weaker, undoped scales can wrinkle Dopant (oxide dispersion) cation ion size (Å)
to accommodate the growth stress at temperature. Figure 32 Normalized time to breakaway oxidation
One of the issues that confounded potential (as an indicator of scale adhesion) at 1200 C in 1 h cycles
hypotheses to explain the role of RE additions was for various oxide (VN for V, BaTiO3 for Ba) dispersions
(0.025–0.2 cation% additions) in Fe–20Cr–10Al. Lifetimes
the reported benefit of any oxide dispersion, includ- were normalized to a 1.5 mm specimen thickness. Low
ing Al2O3 on scale adhesion.143,202 This result could lifetimes for diamonds (Sc, Ce, Eu, and Yb) were attributed
not be repeated in more recent work on Fe3Al, NiAl, to overdoping with 0.2% additions. In general, all the
or FeCrAl.135,188,203 No benefit on scale adhesion, lifetimes are somewhat lower than in commercial ODS
growth rate, or scale morphology was observed with alloys because of the laboratory processing. Data from
Pint.11,12 and Pint and Alexander.135
only an Al2O3 dispersion. In Figure 31, an Al2O3
dispersion reduced the lifetime from 300 h for
wrought undoped FeCrAl to 70 h. normalized lifetime for various dopants added as
A current interpretation8 of the role of RE addi- 0.2 cation% oxide dispersion additions to FeCrAl.
tions on scale adhesion involves several issues: (1) the (In general, the lifetimes for these laboratory-made
inhibition of outward Al transport decreases the alloys were not as good as that of the higher-quality
counter vacancy flux to the interface, (2) the segrega- commercial alloys (Figure 31).) There is a weak
tion of RE ions to the interface decreases the interfa- correlation between lifetime and larger ion size with
cial energy (gint), thereby increasing the critical scale many dopants not fitting this explanation. Several
thickness7,8,204 (see eqn [2]), and (3) the RE addition large ions, such as In, Te and Bi, have low oxygen
prevents the interfacial segregation of S which has a affinity and never become incorporated into the oxide.
detrimental effect on void growth. The observed Dopants such as Sc, Yb, Ce, Hf, La, and Eu produced
outward flux of RE ions from the alloy into the negative effects with 0.2% additions associated with
scale results in RE ions segregated at the metal– over-doping.135 The data points for Hf and La, two of
scale interface.7,8 the highest lifetimes, were additions of 0.05%.
Mechanistically, the role of RE additions may For alloy additions, Figure 31 suggests that the
be resolved, but, practically the issue of optimizing best RE dopant is Y (compared to Hf, Zr, or misch-
that benefit has not. The type of dopant, the amount, metal, a combination of Ce, La, and Nd). However,
and the method of addition has still not been there was a considerable difference in lifetime be-
completely resolved. However, a few issues have tween 0.08 and 0.035 Y additions. Figure 33 shows
been clarified. Surface doping with RE oxide coatings lifetime as a function of specimen thickness for several
or ion implantation does not produce optimal bene- Y additions. It is interesting to note that the lower
fits.6,118,205 Alloy additions and oxide dispersions are Y contents, which would result in less internal oxida-
clearly superior. Effective RE dopants have two qua- tion, produced lower lifetimes than the specimens
lities: large ion size and high oxygen affinity.135 To with 0.08% Y. A wider range of Y additions needs to
avoid the issue of dopant solubility, which affects be studied to resolve this issue and identify an opti-
internal oxidation and other issues, a screening test mum for lifetime. However, given the weak correlation
was performed on various dopants added as oxide with specimen thickness, these results suggest that the
dispersions. Figure 32 shows the 1200 C, 1 h cycle thicker specimens may become depleted in Y due to
14 000 1.24.4.4.2 Interstitials (S, C, etc.)

FeCr–9.8Al–350Hf–30Y
16 ppma S Identifying the detrimental role of S, and possibly
12 000 1200 °C
FeCr–10.1Al-180Hf–20Y 1 h cycles
other interstitials, has caused a major shift in the
Time to breakaway (h)
10 000 16 ppma S thought process on scale adhesion of alumina-

FeCr–9.9Al–820Y
forming alloys in the past 25 years.28,185,204,208–216
8000 The early work by Smeggil and Smialek on NiCrAl
16 ppma S
6000 FeCr–10.2Al–410Y and later on FeCrAl has resulted in almost every
<16 ppma S commercial high temperature alloy being pro-
4000 duced with <5 ppma S, Table 2. While an early
2000
hypothesis217 suggesting that S could explain all RE
FeCr–9.8Al–350Y < 4 ppma S effects was disproved,8,204 the current mechanistic
0 understanding is that S segregates to the metal–
0.5 1 1.5 2
scale interface thereby increasing the interface
energy, which in turn increases the formation of inter-
Figure 33 Time to breakaway oxidation (i.e., Fe-rich facial voids and weakens the interface.186,194,218,219
oxide formation) at 1200 C using 1 h cycles for various
Sulfur segregates to all free surfaces220 and alloy
RE-doped FeCrAl alloys (dopant contents in ppma) as a
function of specimen thickness. For both Y-doped and grain boundaries; thus, it is not surprising that a high
co-doped (Y and Hf) alloys, there is a decrease in lifetime S content in nano-grained FeCrAl did not have
with very low RE additions. Data from Pint.12 a detrimental effect on alumina scale adhesion.221
The large number of alloy grain boundaries provided
spallation and not get as full a benefit as with higher alternative locations for S segregation besides the
Y additions. A similar result has been observed with metal–scale interface.
RE doping in NiCrAl coatings.206 An alternative explanation, that other interstitial
Figure 30 shows better spallation resistance for a elements besides S, such as Na and Ca, has been
NiCrAl with both Y and Hf compared to Y or Hf proposed.222 However, a follow-up study by Smialek223
alone. Figures 31 and 33 show the highest lifetimes which examined 41 elements in a Ni-based superalloy
for a combination of Y and Hf. This so-called co- found that hydrogen annealing of superalloys, which
doping results in low scale growth rates, minimal improved alumina scale adhesion, primarily lowered
internal oxidation, and more optimal performance. the S and C contents. It has been suggested that La
Commercially, some alloys contain two or more additions to superalloys was beneficial in countering a
dopants such as NiCrAl alloy 214 (Y and Zr), SX detrimental effect of P.224 However, a study of Ni–7Cr–
superalloys 1487 and N5þ (Hf and Y), and Aluchrom 13Al þ Y alloys with intentional P additions (70 ppma)
YHf (Y, Zr, Hf, and Ti) (Tables 1 and 2). A Pratt & showed no effect on scale adhesion during 1 h cyclic
Whitney NiCoCrAl coating (PWA286) also contains testing at 1100 C.12
Y and Hf.207 Co-doped alloys rely on a higher solu- A common RE–S explanation is that RE dopants
bility dopant (Hf or Zr) along with lower levels of the ‘getter’ S away from the metal–scale interface by
low solubility Y dopant. Again, dopant depletion can forming RE–S compounds.210 However, this mecha-
be an issue because the lowest level of Y and Hf nism has not been proven for ODS alloys (which
dopant did not produce the longest lifetime in a contain high S levels, Table 2) where dispersoids
model FeCrAl alloy (Figures 31 and 33). Co-doping did not contain S and S was not found segregated to
levels in model alloys were selected to match the the dispersoid–alloy interface.8 The observation of a
alloy interstitial levels to achieve a [Y]/[S] >1 and Y–S compound225,226 is not proof of a gettering
a [Hf]/[C] >1.129 This interstitial-dopant ratio strat- mechanism, only that Y forms a stable sulfide.
egy was originally proposed by Sigler.208 In commer- A segregation-based RE mechanism suggests that
cial alloys, the interstitial levels are higher than in RE ions segregate to the metal–scale interface and
the laboratory, which may require slightly differ- lowers the energy such that S segregation is no longer
ent doping strategies.10 Trying to target narrow com- thermodynamically favorable.8 Whether this involves
position windows is probably not realistic for site competition for segregation or another mecha-
commercial production, particularly considering nism is not known. Nevertheless, there is a clear
that a co-doped alloy with a Y/S <1 will have very relationship where the Y/S ratio needs to be greater
poor performance lifetime, for example, FeCrAl with than unity for good scale adhesion in FeCrAl and
0.018Hf(þY) in Figure 31. NiCrAl.28,129
1.24.4.4.3 Precious metals 20
Mean thermal expansion (ppm ⬚C–1)

Fe–28Al–2Cr + Hf
Alumina scale adhesion has been improved by Pt
and other Pt group metals (PGMs).79,84–86,227–229 18
NiCoCrAlY
René N5
The majority of observations have been for Ni-
based alloys and coatings, especially b-NiAl. Pt addi- MA956
16
tions also have been used to improve the performance NiAl + Hf
of CoCrAlY coatings.227 While the effect of Pt on Ni–22Al–5Pt
+ Hf
14
adhesion is similar to RE additions as both are likely Fe–18Al + Y
to lower the metal–scale interface energy, the mech-
12
anism is quite different as 2 at.% Pt or Ir is needed
as an alloy addition to improve adhesion. Also, Pt Al2O3
10
does not alter the scale growth rate and does not
become incorporated into the alumina scale or seg- 400 600 800 1000 1200
regate to scale grain boundaries, even on Pt–Al Temperature (⬚C)
alloys.230 Figure 34 Mean coefficient of thermal expansion as a
There is some disagreement about the effect of Pt function of temperature for several Ni- and Fe-based alloys
additions in FeCrAl-based alloys. One study sug- (see Table 2) showing the difference between these metals
and alumina. Data from Haynes et al.,30 Pint et al.,82 Wright
gested that PGM additions benefited scale adhesion et al.,232 and Pint et al.233
on FeCrAl.229 However, that result could not be
reproduced with 0.1 or 3 at.% Pt additions.12 Unlike
in Ni-based compositions, the formation of a Pt-rich a-Al2O3. Alloys that have a phase transformation
phase in FeCrAl þ 3Pt, likely due to limited solubil- such as NiCoCrAlY or b-FeAl have much higher
ity of Pt in ferrite, may explain the lack of benefit CTE than the materials that remain a single phase
compared to Ni- and Co-based alloys and coatings. across the 25–1200 C temperature range.
For Ni-based alloys, it is difficult to draw an exact
correlation between CTE mismatch and scale adhe-
1.24.4.4.4 Role of CTE sion, again because it is difficult to quantify scale
Thermal expansion mismatch between the metal and adhesion. One criterion used previously for lifetime
scale is a critical factor in understanding the oxida- modeling was the time to zero specimen mass gain, tz,
tion resistance of various alumina-forming alloys. in a cyclic oxidation test.234 For example, in Figure 31
The effect is most evident in the Fe–Al system the tz for Hf-doped NiCrAl would be 85 h and
because of the very high CTE of the aluminide 425 h for Y-doped NiCrAl. In Figure 35, the
phases, particularly b-FeAl.52,128,231–233 The differ- CTE mismatch is plotted versus tz at 1150 C for
ence in CTE can explain the scale spallation pro- several Ni-based alloys.30 The general trend in the
blems of RE-doped aluminides compared to ferritic data is that a smaller Da results in a longer tz. The
FeCrAl and Fe(Al) alloys. The effect in RE-doped various NiCrAl and NiCoCrAlY compositions have a
Fe–Al alloys can be quantified to a certain extent by higher CTE and spall more readily. One complicat-
using an approximation of the critical scale thickness, ing issue for this type of plot is that RE additions have
xf, from eqns [1] and [2]: a strong influence on scale adhesion (and thus tz) but
xf /ðDT DaÞ2 ½4 have no effect on CTE.30 Therefore, the horizontal
arrows show the increase associated with RE-doping
Above this thickness, the strain energy in the scale or RE optimization (in the case of NiCrAlYHf). The
exceeds the interface strength and spallation occurs. dashed line represents an assumed optimum for RE
Assuming that the Al content does not affect the doping. (Note that the 2 000 h value for Hf-doped
interface or oxide properties, if xf = 10 mm for NiAl was an approximation, the mass gain did not
FeCrAl, then it is only 1 mm for b-phase Fe–40Al reach zero after 2000 1 h cycles at 1150 C, although
and 2 mm for Fe3Al.233 it was decreasing towards zero.12)
Figure 34 gives some typical mean CTE values
for alumina-forming alloys as a function of tempera- 1.24.4.4.5 Role of substrate strength
ture. The values for NiAl, FeCrAl and Ni–22Al–5Pt One of the alloys that does not fit the trend in
are all remarkably similar at 1000–1200 C and all Figure 35 is the single-crystal (SX) superalloy, N5,
these materials have the lowest CTE mismatch with which contains both Y and Hf. There are two possible
11 9000
Ni–20Cr–19Al + Y 1200°C
Ni–20Cr–19Al + Y/Hf Fe–18Al + Y
8000 1h cycles
10
Ni–7Cr–13Al 7000 Wrought
Limit with optimal APM

FeCrAlY
RE addition (?)
ΔCTE at 1150 °C
9 6000
Ni–7Cr–13Al + Y APMT
NiCo–17Cr–14Al + Y 5000
ODS
8
4000 PM2000
Ni–10Cr–19Al + Y ODS
7 3000 MA956HT
N5 superalloy (Y + Hf)
Ni 3Al+Y 2O3 2000 ODS
NiPtAl + Hf
6 NiAl Ni3Al + Hf MA956
1000
Improvement due to RE NiAl + Hf
5 0
100 1000 10 000 0.5 1 1.5 2 2.5
Number of 1 h cycles to zero mass gain at 1150 °C (a) Specimen thickness (mm)
Figure 35 Difference in CTE between the alloy and 6000

alumina at 1150 C for various Ni-based alloys (see Table 2) ODS FeCrAl
plotted versus the number of 1 h cycles to zero mass gain at MA956
5000
1150 C. The number of cycles reflects the relative scale 1100 °C
adhesion of these materials. Intermetallic alloys have much 100 h cycles
lower CTE values than the multiphase NiCrAl alloys that 4000 1200 °C
undergo phase transformations between RT and 1150 C. 100 h cycles
Adapted from Haynes, J. A.; Pint, B. A.; Porter, W. D.;
3000 1200 °C
Wright, I. G. Mater. High Temp. 2004, 21, 87–94.
1 h cycles
2000
reasons for this deviation, one is that the spallation of 1250 °C
the transient oxide on this alloy increases the speci- 100 h cycles
men mass loss but does not actually reflect the adhe- 1000
1300 °C
sion of the alumina scale. The other possible reason is 100 h cycles
that this very strong SX substrate does not accommo- 0
date the stresses in the scale (e.g., by creep) like a 0.5 1 1.5 2 2.5
weaker substrate, such as NiAl, resulting in earlier (b) Specimen thickness (mm)
scale spallation.235,236 Figure 36 Time to breakaway oxidation (i.e., Fe-rich
A substrate creep strength hypothesis has been oxide formation) as a function of initial specimen thickness
used to explain the decreased alloy lifetimes of (a) at 1200 C using 1 h cycles for various FeCrAl and
ODS FeCrAl alloys compared to weaker wrought Fe–Al alloys (b) for alloy MA956 under various cyclic
conditions (see Table 2 for compositions). Data from Pint12
FeCrAlY or Fe(Al)Y (Figures 31 and 36(a)).233 One and Pint et al.233
exception to this negative effect of alloy strength is
the commercial FeCrAl alloys APM and APMT.233
With the addition of Mo and lower Cr content, APMT characterized because it can be detrimental to the
has a higher creep strength than APM237 but essen- performance of FeCrAl catalyst supports.238
tially no decrease in lifetime (Figure 36(a)). However, In general, it is difficult to quantify the mechanical
APMT also contains Y and Zr so that RE optimization effects of typical alumina-forming alloys because the
compared to APM may be responsible for maintaining tensile properties and creep (primary, as well as
the cyclic oxidation lifetime. steady-state) rates are not commonly measured at
Substrate strength also affects the performance of the temperatures relevant for high temperature
foil specimens. In this case, the thin metal section oxidation.187
deforms due to stresses in the oxide and little scale
spallation occurs in most cases. The compressive 1.24.4.4.6 Effect of cycle frequency
stress generated in the growing oxide can cause the A final note about adhesion is that increasing the cycle
foil to lengthen due to the tensile stress in the adja- frequency, that is, reducing the time per cycle usually
cent metal. This increase in length has been decreases scale adhesion by increasing the number of
thermal cycles per time period.15,198,234,239 Figure 36(b) 1.24.4.5 Internal Oxidation
gives examples of the time to breakaway oxidation
The formation of oxide (or, more generally, any oxi-
for ODS alloy MA956 under various cyclic conditions.
dation reaction product) within the alloy, beneath the
Of course increasing the temperature decreases the
external scale, is generally regarded as detrimental
lifetime using 100 h cycles at 1200–1300 C. However,
because it can degrade the mechanical properties of
increasing the cycle frequency from 100 to 1 h per
the substrate and interfere with the long-term chem-
cycle at 1200 C, produced a reduction in lifetime
ical or mechanical stability of the protective external
similar to increasing the test temperature by 50 C.
scale. For example, an internal oxide can help to
Figure 37 gives an example for FeCrAlY where dur-
nucleate a crack in the substrate. As also described
ing 100 h cycles to a total time of 5000 h cycles, virtu-
previously (Section 1.24.4.1), internal oxidation gen-
ally no scale spallation occurred. However, in 1 or 2 h
erally increases the overall oxidation rate and can
cycles, spallation began after a few hundred hours.
serve as fast transport paths for oxidants. The issue
One-hour cycles are used in lifetime testing to accel-
of internal oxidation of Al could be related to insuffi-
erate the experiment by inducing scale spallation
cient Al content to form an alumina scale, which
more quickly.
essentially is a problem of low Al content alloys
Exceptions to the inverse relationship between
and/or alloys with poor selective oxidation.
cycle frequency and scale adhesion were observed
Internal oxidation behavior has been studied on
for some Ni-based superalloys.15 This behavior was
dilute M–Al alloys to understand the transition from
attributed to the spallation of transient scale during
internal to external oxidation. As oxidation is a diffu-
longer cycles but could also be due to the time-
sion controlled process, parabolic kinetics are gener-
dependent growth of defects such as interfacial
ally observed.240,241 The internal versus external
voids. In short cycles, the defect may never reach
oxidation condition given by Wagner240 suggested
the critical size necessary to cause spallation or only
that the oxidant solubility and diffusivity in the
cause small areas to spall. Longer cycles, with long
alloy were critical. Wolf and Evans242 showed that
holds at temperature, may allow defects to grow to a
the internal oxidation rate of Ni–Al alloys was nearly
sufficient size or accumulate such that scale spall-
independent of the pO2 in the environment. Whittle
ation occurs over large areas.
et al.243 found that the internal formation of Al2O3 in
dilute Ni–Al alloys accelerated O diffusion within
the alloy, likely at the Al2O3 precipitate–metal inter-
face. This could explain the faster oxidation kinetics
6
100 h cycles
of dilute Ni–Al alloys compared to unalloyed Ni.244
5 In alloys with sufficient Al to form an external
4 10 h cycles alumina scale, internal oxidation of Al has been

3 observed in Fe(Al) þ 0–5Cr alloys at 800–1000 C in
2 Fe–20Cr–10Al-0.08Y air.93 Internal oxides as well as nitrides were observed
2 h cycles on undoped alloys with defective alumina scales. It
1
was concluded that the defective scale allowed
0
N permeation through the scale into the substrate.
–1 The less-stable AlN precipitates were displaced by
–2 more-stable oxides and nitride formation moved fur-
–3 1h cycles
ther (deeper or laterally) into the substrate with time.
–4 The high mass gains (with linear kinetics) for the
undoped ferritic Fe(Al) alloys are shown in Figure 15.
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Figure 15 also shows that internal oxidation did not
Total exposure time (h) at 1200 °C
occur in air for undoped Fe–20Cr–10Al.
Figure 37 Specimen mass change for FeCrAl–0.08Y as a Internal oxidation of other alloy elements may
function of total exposure time for various cycle frequencies occur, typically only those with a higher affinity for
of 1–100 h per cycle. In 100 h cycles, very little spallation
was observed but reducing the cycle time increases the
O than Al such that oxidation will occur below the
number of thermal cycles and increases the amount of Al/Al2O3 equilibrium pO2 at the metal–scale inter-
spallation, particularly for 1 and 2 h cycles. Data from Pint12 face. All RE additions fall into this category.135 The
and Pint et al.15 issue of RE internal oxidation has already been
lower Cr and Al additions needed to form alumina

and better alumina adhesion. With two-phase alloys,
the Al-poor phase will tend to form more base-metal
oxide. Furthermore, alloy phase transformations typ-
ically increase the CTE. The low CTE of FeCrAl
results in less thermal stress which drives scale spall-
(a) 10 μm ation. Faster Al diffusion in the alloy minimizes Al
concentration gradients in the alloy and decreases the
critical Al content needed to form alumina. The rapid
growth of FeO (106 times faster than Al2O3) results
in more catastrophic failure at the end of life com-
pared to NiCrAl (NiO forms 103 times faster than
Al2O3), however, the other factors tend to outweigh
20 μm
(b) this problem.60
Figure 38 Light microscopy of polished cross-sections of
Several studies agree that the oxidation behavior
FeCrAl þ 0.038(Y,Hf) after oxidation in laboratory air for of CoCrAl alloys are decidedly inferior to NiCrAl or
(a) 5000 h at 1000 C and (b) 100 h at 1300 C.12 More FeCrAl.60,245 For the formation of an alumina scale,
internal oxidation occurred at low temperature than Co-based alloys require higher Cr and Al contents
at high temperature where a much thicker external because of the rapid growth of CoO (105 times
oxide formed.
faster than Al2O3) and the relatively low diffusivity
in the alloy compared to Ni (102 slower) and
discussed in terms of minimizing the internal oxida- Fe-based (103 slower) alloys.60 In cyclic oxidation
tion by reducing the RE addition (e.g., Figure 23). testing, CoCrAl alloys can be improved by RE addi-
The effect of oxidation temperature on internal oxi- tions.132,245,246 However, their cyclic oxidation resis-
dation is complicated by the effects of solubility and tance is inferior to NiCrAl. The main interest for
diffusivity.240 The solubilities and diffusivities of both Co-based alumina-forming materials appears to be
O and the RE increase with temperature but at dif- as coatings for Co-based alloys.
ferent rates. An example of the effect of temperature
on internal oxidation is shown for FeCrAlYHf in
Figure 38. Over the range of 1000–1300 C, it is 1.24.5 Breakaway Oxidation
difficult to make direct comparisons. However, rela-
tive to the external scale thickness, the internal attack The final stage of oxidation is breakaway which is
at 1000 C was deeper than at 1300 C. The solubility characterized by rapid oxidation due to the depletion
of the particular RE addition in the alloy also could of Al from the substrate and the formation of faster-
play a factor in the temperature effect. growing reaction products. Figure 39 shows a failed
For g–g0 NiPtAl alloys, the addition of Pt specimen of FeCrAl alloy APMT cycled to failure at
decreased the internal oxidation of Hf.80 The effect 1200 C using 1 h cycles. One edge had been
was particularly strong with 10–15%Pt and Hf levels completely consumed by Fe-rich oxide formation.
up to 0.4%. This Pt effect was attributed to Pt sup- Typical of many dispersion strengthened alloys,247
pressing the Hf activity;80 however, it could also be an the center contains Kirkendall voids due to the
effect of Pt on the O solubility. depletion of Al. In stronger ODS FeCrAl, the voids
are often perfectly round. In weaker wrought
1.24.4.6 Comparison of Fe-, Ni-, and FeCrAlY, more deformation of the substrate occurs
Co-Based Alumina Forming Alloys and center voids are not very evident.
The time to breakaway, tb, is thus the useful life-
Only a few authors have explicitly examined the time of a component with thickness, d. It was used in
differences in oxidation behavior due to the base several figures in the previous section (Figures 31–33
metal.60,95 In general, FeCrAl alloys are the best and 36). A common model for lifetime of Fe-based
alumina-formers because they are single-phase, alloys is based on the Al reservoir in the component
have a lower CTE, faster diffusion in the alloy (due and the rate of consumption of that reservoir41,42:
to their bcc structure) and lower O solubility. This
combination of diffusivity and solubility results in drðC0 Cb Þ ¼ Aktbn ½5
Cu plate
APMT
FeOx
A l 2O3
400 μm
Figure 39 Light microscopy of polished cross-sections of alloy APMT after cycling to failure in 1 h cycles at 1200 C.12
One edge has been consumed and only Fe-rich oxide remains. In the center of the specimen, large Kirkendall voids remain
from the depletion of Al.
where r is the alloy density, C0 is the starting Al 1.4

Center -1300 °C (triangles)
content in the alloy, Cb is the Al content where the
material can no longer form a protective alumina 1.2
scale, A is a constant, k is a rate constant, and n = 0.5 Normalized Al content (at.%)
1.0 Center -1250 °C
for parabolic kinetics and 1 for linear kinetics. For
(squares)
FeCrAl-type alloys with 20% Cr, Cb is close to zero.
0.8
With lower Cr contents, Cb increases to 10% Al in
Fe–Al binary alloys.248,249 This type of equation 0.6
allows the lifetime to be predicted as a function of Center -1200 °C
(diamonds)
component thickness or for a thickness to be selected 0.4 End -1250 °C
to achieve a desired service lifetime. (circles)
One of the key assumptions of the reservoir model 0.2
is that there are no Al gradients in the substrate; this
greatly simplifies behavior. It is a reasonable assump- 0
0 100 200 300 400 500 600 700
tion for foil (e.g., Figure 20) and thin (<1 mm) com-
ponents.248 However, for thicker components, Al Distance (μm)
gradients do occur and cause significant deviations Figure 40 EPMA Al line scans from two different MA956
in lifetime behavior.249 For example, Figure 40 shows (ODS FeCrAl) specimens oxidized to breakaway at
the gradients in several MA956 specimens oxidized 1200—1300 C and metallographically sectioned. The
center scans are across the entire specimen thickness near
to failure at 1200, 1250, and 1300 C. The center of
the center of the specimen while the end scan is from an end
the specimen contained >1%Al while the end of the of the specimen with distance from the edge towards the
specimen, near where the breakaway formation of Fe- center of the specimen. Data from Pint et al.249
rich oxide occurred, had 0.1% Al. Earlier studies
suggested that Cb was a function of thickness.250 on this alloy after cyclic oxidation. Specimen geome-
While Cb effectively increases with thickness because try (e.g., corners on thin specimens) can also lead to
of Al gradients in thick substrates, mechanistically, in early failure. Therefore the specimen surface area to
all cases, the value of Cb at the metal–scale interface, volume also can affect lifetime and round versus flat
Cbi, at breakaway is close to zero.249 specimens may behave differently.252
Gradients also drive early failures in some materi- As shown for alloy 214 (Figure 41), the behavior
als (Figure 41). For Fe–17Al þ Hf and NiCrAlYZr of Ni-based alumina-forming alloys12 may not be
(alloy 214), increasing the specimen thickness explained well by the Quadakkers reservoir model.
resulted in virtually no increase in tb. It was shown The slower diffusion of Al in austenitic alloys and the
that Al gradients formed in Fe–17Al þ Hf resulted formation of multiple phases (g, g0 , b) in NiCrAl
in early failures for the thicker specimens.52,249 alloys may result in complications that are better
A similar failure mechanism may occur on alloy modeled by a diffusion based model as suggested by
214,251 but less characterization work was performed Nesbitt and Heckel253 or the spallation-based COSP
3500 Fe-based alloys that do not spall as orderly cannot

1200°C be modeled in this manner (compare Ni-base spall
Aluchrom YHf 1h cycles
3000 behavior in Figure 30 to Fe-based spallation in
Figure 37). In general, measuring the residual Al con-
Wrought ODS
2500 FeCrAlY PM2000 tent and determining the rate of Al loss in the alloy is
ODS
MA956 the most accurate method of predicting lifetime.255
2000
1500 1.24.5.1 Failure Mechanisms

SUH21
Fe–17Al + Hf In practice, two types of oxidation-induced failure
1000 Ohm.30
can be distinguished.256 A chemical failure occurs
500 NiCrAl (Alloy 214) when the substrate is sufficiently depleted in Al that
Ohm.40 it can no longer form alumina. The most extreme
0 example of such a failure is for FeCrAl(þRE) foil
0.5 1 1.5 2
where no scale spallation occurs because the foil can
deform and accommodate strain energy from the scale.
Figure 41 Time to breakaway in 1 h cycles at 1200 C When the foil is depleted in Al, a chromia scale has
versus specimen thickness for various alumina-forming been observed to form beneath the alumina scale.257
alloys (see Tables 1 and 2). For alloy 214 and Hf-doped
As Cr2O3 is less stable than Al2O3, normally this would
Fe–17Al, there was little increase in lifetime with specimen
thickness due to the formation of alumina gradients. Data not be thermodynamically possible. It occurs when Al
from Pint12 and Pint et al.233 is removed from the metal and the interface oxygen
potential increases allowing Cr2O3 to form.
model.17 These models and Nesbitt’s COSIM (Coat- Prior to a chemical failure, a mechanical failure
ing Oxidation and Substrate Interdiffusion Model)254 can cause a premature failure by cracking the scale
are well suited for studying the behavior of NiCrAl and requiring reformation of alumina at a bare inter-
coatings where the Al content decreases both due to face.256 The mechanical failure could be due to the
oxidation and interdiffusion with the coating substrate. typical thermal stress (which can induce scale spall-
As shown in eqn [5], the consumption side of the ation), an applied stress to the component in service
equality is very simplistic. Parabolic kinetics would or a one-time upset or impact which removes the
only be observed for alloys where no spallation surface oxide. In this case, the Cb at failure is higher,
occurred. As shown in Figure 41 and in previous reducing tb.
figures, the nearly linear relationship between speci-
men thickness and tb suggests that linear kinetics, that
1.24.5.2 Unusual Failure Mechanisms
is, n = 1, are more relevant. However, this approxima-
tion would overestimate the Al loss at early times, Some unusual circumstances can cause early failure
when the scale is adherent and Al consumption fol- of alumina scales. Additional oxidants (discussed in
lows parabolic kinetics. A more complicated con- the next section) can cause early failure as will envir-
sumption equation probably involves several time onments that can otherwise degrade the alumina
stages with different k and n values for each stage.251 scale. As mentioned previously, the formation of Al
If a critical scale thickness could be identified to concentration gradients in the alloy will result in ear-
determine the change from the protective scale lier failures. This is generally not a problem for Fe-
phase to the spallation phase (with higher Al con- based alloys with relatively rapid Al diffusion (e.g., flat
sumption and higher n), such a model would be more Al profiles in Figures 20 and 40). However, it
mechanistic and accurate. appeared to be a problem for Fe–17Al þ Hf, Ni-
Other models are available to predict lifetime. based alloy 214 (see Figure 41) and is likely a prob-
The NASA-developed COSP model uses the scale lem for Co-based alloys. There is little data in the
growth and spallation rates to predict lifetime.17 The literature for Ni- and Co-based alloys on residual
spallation mass loss rate is attributed to an area frac- Al profiles or lifetime as a function of specimen
tion of spallation and regrowth. This model has been thickness.
applied to an extensive oxidation database of typi- Another unusual failure mechanism for Fe-based
cally Ni-based and ODS FeCrAl alloys that exhibit alloys has been attributed to the internal oxidation of
near-linear spallation rates. However, wrought Cr carbides on alloy grain boundaries resulting in
accelerated Fe-rich oxide formation at alloy grain optimization. However, in commercial alloys, the
boundaries, sometimes called ‘broccoli’ failures possibility of achieving narrow composition targets
because of the eruption of the voluminous Fe-rich may be an unrealistic goal. In terms of increasing C0,
oxide.258 This type of failure is found in alloys with alloy mechanical properties and the phase equilibria
higher C contents but no strong carbide former, for limit the amount of Al that can be added without
example, Y-doped FeCrAl and Fe–Al. The problem significant negative consequences to other properties.
can be eliminated by the addition of a strong carbide Figure 41 shows the relatively low lifetimes for some
former such as Hf or Ti, which is a common addi- commercial alloys, Ohmalloy 30 and 40 and SUH21,
tion in commercial alloys, Table 2. For example, all with lower Al contents (Tables 1 and 2). However,
Figure 42 shows the mass gain of ferritic Fe–17Al these alloys are easier to fabricate, process and join
alloys tested at 1250 C to accelerate their failure. than alloys with higher Al content such as ODS alloys
With only Y or Y and Cr additions, tb was relatively and Aluchrom YHf. Also, FeCrAl alloys become very
short compared to the specimen with a Hf addition. brittle above 12 at.% Al.61 Decreasing Cb is virtually
The early failures showed broccoli failures. However, impossible for FeCrAl alloys because Cb is already
when a small amount of Ti or Hf was added, the close to zero. For ferritic Fe(Al) alloys with higher Al
lifetime increased dramatically. contents (up to 19%), attempts to lower Cb from
10% for binary alloys have been unsuccessful.12
Additions of ‘third elements’ Mn and Cr were
1.24.5.3 Strategies for Extending Lifetime
explored with little success.12,233 Manganese appears
One goal of alloy development is to design alloys with to become quickly incorporated into the scale near
longer lifetimes. Despite shortcomings, eqn [5] does the end of life and Cr additions appear to increase the
focus attention on some of the key lifetime variables: CTE resulting in increased spallation. A general
the Al reservoir (C0 Cb) and the Al consumption strategy of lowering the CTE, possibly with refrac-
rate by oxidation (k and n). It suggests that lifetime tory metal additions, may be possible but has not
could be increased by decreasing the rate of con- been successful for Fe-based alloys to date.233 A low
sumption, or by increasing the Al reservoir either CTE Ni-based bond coating has been developed.259
by increasing C0 or decreasing Cb. As discussed pre-
viously, the scale growth rate can be decreased by RE
1.24.6 Effect of Environment
+Y,Ti
The previous sections discussed the oxidation behav-
25 +Y
ior of alumina-forming alloys in O2 or air with low
FeCrAl –0.08Y
levels of water vapor in the ambient laboratory envi-

20 (dashed)
10Cr + Y
ronment. In practice, high temperature applications

+Hf
tend to involve more complex environments. How-
15 ever, because of the inherent chemical inertness
+Y,Hf (i.e., thermodynamic stability) of a-Al2O3, alumina-
10 forming alloys tend to have an advantage over conven-
tional chromia-forming alloys under these conditions.
5
Base: Fe–17Al
1.24.6.1 Water Vapor
0
0 4000 8000 12 000 16 000 In actual applications, particularly those involving
Time (h) in 100 h cycles at 1250 °C combustion, water vapor is commonly found in the
environment. While Kofstad called the effect of water
Figure 42 Total mass gain (specimen þ spalled oxide) for
ferritic Fe–17Al alloys (see Table 2 examples) during 100 h
vapor on oxidation a ‘forgotten’ problem in 1991,260 it
cycles at 1250 C in laboratory air. Specimens of alloys with has become a significant topic for research in the past
only Y additions showed relatively short times to breakaway decade because of increases in temperature applica-
oxidation due to ‘broccoli’ formation.258 Specimens with tion and the desire to use new materials. Also, com-
strong carbide-forming elements: Hf, Y þ Hf or Y þ Ti, bustion of hydrogen or synthetic natural gas from
prevent this form of attack associated with accelerated
oxidation of grain boundary carbides. The co-doped alloy
coal would increase the water content in the combus-
specimens had a much longer time to breakaway oxidation. tion product from 10 to 15% for natural gas or jet fuel
Data from Pint.12 to 60% for hydrogen, making water vapor more of a
concern.261 As shown by Opila,262 chromia and silica 1

NiCrAl + Y,Hf 10%
scales readily form oxy-hydroxides in the presence of 0.9

H2O and O2 resulting in accelerated oxidation attack
0.8 0%
for chromia-forming alloys and silica-based cera-
50%
mics.263,264 For typical stainless steels at 650–800 C, 0.7
90%
the Cr consumption becomes linear due to the 0.6 90%
evaporation.265 0.5
50%
10%
As a-Al2O3 is more stable than Cr2O3 or SiO2, the 0%
0.4
reaction with water vapor is minimal until higher
temperatures >1300 C.262 The effect of Al additions 0.3 Ni–48Al + Hf
on resistance to water vapor is quite dramatic. 0.2
Kvernes et al.266 showed that small Al additions to
0.1
Fe–Cr alloys were very effective in improving the
oxidation resistance. Likewise, Pint et al.,267 showed 0
0 200 400 600 800 1000 1200
that additions of Al to austenitic Fe–15Cr–15Ni were
Number of 1 h cycles at 1100 °C
very effective in reducing the detrimental role of
water vapor. However, that study also showed that Figure 43 Specimen mass gain for cast Ni–20Cr–10Al þ
the addition of Al caused the formation of ferritic Y,Hf and Ni–48Al–0.05Hf specimens (see Table 2) exposed
phases in the steel that decreased the 750 C creep to 1 h cycles at 1100 C in dry O2 or air with various vol% of
H2O from 10 to 90%. The addition of water vapor caused a
life by 2 orders of magnitude, making the resulting
minor increase in the mass gain for Hf-doped NiAl and a
alloy unacceptable for many applications. Recently, slight increase in spallation (mass losses) after 800 cycles
Brady and coworkers67–69 showed that careful com- 12
for NiCrAlYHf. Data from Pint and Pint et al.261
position control and minor element additions could
produce an Fe-based alumina-forming austenitic H2O content. (2) A small increase in the oxidation
steel with very good creep strength and resistance rate. The mass gain of NiAl þ Hf increases with
to water vapor environments. This alloy is a potential H2O content suggesting a slight increase in the rate
solution to the problem of degradation of heat constant (Figure 43). This was also noted for Fe-
exchangers or recuperators for small turbines.268 based alloys.276 (3) A decrease in selective oxidation.
Compact, high-efficiency primary surface recupera- On low Al alloys, more Ni-rich spinel oxide was
tors for these turbines require <100 mm thick foil formed with the addition of water vapor. This is a
with good creep and oxidation resistance. particular concern for uncoated superalloys and g–g0
The excellent resistance of alumina scales to water NiPtAl coatings with low Al content.82,261 An expla-
vapor has led to extensive work on alumina-forming nation for this decrease based on H injection into the
coatings on Fe-based alloys for combustion and steam substrate may apply to both alumina- and chromia-
environments. Thin coatings can improve oxidation forming alloys.277 Nevertheless, for adherent
resistance to steam for Fe–9Cr alloys for >40 kh at alumina-forming alloys with adequate Al, these
650 C.269 Ferritic Fe(Al,Cr) weld overlaid coatings effects are relatively minor as shown by the data in
have shown promising results for boiler applications Figure 43 where even 90% water vapor caused a
at <600 C.270 Coatings are further addressed in a minor effect at 1100 C.
separate chapter.
Several studies have examined the effect of water
1.24.6.2 Sulfur and Hot Corrosion
vapor on Ni-based alloys.261,271–274 The general con-
clusions focused on three effects of water vapor on A wide range of environments contain sulfur, sulfur-
alumina-scales are: (1) Water vapor increases the containing compounds (i.e., hot corrosion) or, more
spallation rate of alumina scales, particularly for mar- commonly, mixtures of S and O such as in coal-
ginally adherent scales. Smialek275 suggested that this fired combustion (SOx) or gasification (H2S, low
could be due to water vapor increasing the crack pO2), which was extensively investigated in the
growth rate at the metal–scale interface. Figure 43 1980s. For purely sulfidizing and hot corrosion (salt)
shows the specimen mass gain for NiAl þ Hf and environments, alumina-forming alloys are only mar-
NiCrAlYHf specimens in dry O2 and air with ginally effective since Al does not really form a selec-
10–90 vol.% H2O. For NiCrAlYHf, the drop in mass tive, protective sulfide layer and can be fluxed away
gain suggests scale spallation which increases with by sodium–sulfate salts. As a result, the sulfidation
reaction of the base metal is a key issue. Ni-based alumina scales and S was found segregated to a-
alumina-forming alloys will be catastrophically Al2O3 grain boundaries using STEM after exposure
affected by the formation of a liquid Ni–S reaction to a H2–H2O–H2S environment.293 Cracks in the
product above 700 C.278,279 Cobalt-based alloys are scale are another possible route for S penetration.
thought to be more resistant to sulfidation. Therefore, Hot corrosion, generally associated with Na2SO4
CoCrAl alloys and coatings have been developed for salts, can be decreased by the formation of an adher-
their sulfidation and hot corrosion resistance.280,281 ent alumina scale. For example, on g–g0 NiPtAl
For Fe-based alumina-formers, some improvement is compositions, preoxidation was very effective in
associated with the addition of Al compared to FeCr reducing hot corrosion attack.91 Therefore, RE addi-
alloys but the sulfidation rates are still relatively tions and Pt may be somewhat beneficial in hot
high.282 For example, a 150 mm thick scale was formed corrosion.294 However, because of the fundamental
in 7 h at 800 C with a pS2 = 0.1 Pa. fluxing reaction controlling hot corrosion,295 Cr is
For oxidizing–sulfidizing environments, the per- much more effective than Al or Pt in reducing hot
formance of Fe-based alumina-forming alloys in corrosion attack.296
petroleum environments has been investigated for
more than 40 years.283 Since then a great deal of
1.24.6.3 Carbon
work has been conducted on Fe(Al) and FeCrAl
alloys in low oxygen, high sulfur pressure environ- For environments with high C activity where internal
ments.50,284–288 Under these conditions, binary (low carburization and metal dusting occur,297 alumina-
Cr) Fe–Al alloys are generally superior to FeCrAl forming alloys appear to be oxidation resistant. Com-
alloys because of the possible sulfidation reaction pared to chromia-forming stainless steels, where the
with Cr. However, 19%Al is needed in Fe(Al) for Cr2O3 scale appears to allow the ingress of C, alu-
this sulfidation resistance.286 A similar level of Al was mina scales appear to prevent C penetration.298,299
needed at the surface of Fe–Al coatings for a protec- The addition of Al or alumina-forming coatings
tive scale to form.289 However, these experiments appears to solve many C-related oxidation problems.
were typically performed at 800 C. At lower tem- Again, the issue of alloy strength and joining is a
peratures, 400–600 C, where reaction rates are lower critical one for alloy selections in many relevant
and it is more difficult to form alumina, the presence applications. Alumina-forming alloys may not have
of Cr may be beneficial and work better than Fe–Al the necessary strength or can be difficult to fabricate
materials. This has been observed for Fe3Al and or join for commercial applications where high car-
FeCrAl porous metal filters used in sulfidizing bon activities are a concern. Recently, cast NiCrAlY
environments.12 alloy tubes have been developed for steam reforming
Several studies suggest that additions of Y and applications.29 The authors reported no major drop
Zr are beneficial to corrosion resistance in sulfur- in creep strength or weldability. In addition, the
containing environments at 600–800 C.284,285 In some relatively inert alumina scale had an additional
mixed O–S environments it is possible to form an benefit of reducing surface coke formation.
alumina scale that could be protective. Another com-
mon strategy has been to preoxidize the alloy in an
1.24.6.4 Nitrogen
oxidizing environment prior to exposure to S-bearing
gases, particularly for FeCrAl.290–292 If metastable, Nitrogen containing environments appear to be
cubic alumina phases form during preoxidation, the one condition that can cause problems for alumina-
volume reduction and associated scale cracking during forming alloys. It appears that N can permeate through
the transformation to a-Al2O3 will likely be detrimen- alumina scales and internally nitride alumina-forming
tal to performance.292 Thus, the preoxidation step must alloys.93,299 Most likely, the N permeation is not diffu-
be controlled to assure an a-Al2O3 preoxidation scale. sion through dense alumina but penetration in the
While preoxidation has been successful in laboratory cracks in the scale. However, diffusion along a-Al2O3
tests, in practice it has been less effective. In some grain boundaries, like S, cannot be ruled out. For
applications, the problem is probably due to the dam- Fe-based alumina-forming alloys, internal nitridation
age of the preoxidation scale during subsequent has been observed in atmospheres with predominantly
handling and joining. N and low pO2.299–301 Both Al and Ti (when present) are
Even when the alumina scale is adherent and internally nitride. Internal nitridation of Al is observed
dense, it has been shown that S will penetrate at higher pO2 levels when the Al content is decreased
during extended exposure. As mentioned previously, References

internal nitridation has been observed for undoped
Fe(Al) alloys in air but not for undoped FeCrAl.93 1. Whittle, D. P.; Stringer, J. Philos. Trans. R. Soc. London,
For Ni-based alloys, Han and Young302 found that Ser. A 1980, 295, 309–329.
2. Lang, E. Ed. The Role of Active Elements in the Oxidation
internal oxidation and nitridation of Ni–Al–Cr Behavior of High Temperature Metals and Alloys; Elsevier
alloys was a result of heavy spallation which Applied Science: London, 1989.
depleted the alloy in Al and caused a phase trans- 3. King, W. E. Ed. Material Science Forum 43, The Reactive
Element Effect on High Temperature Oxidation – After
formation. The internal nitridation of Ni-based Fifty Years; Trans Tech Publications: Switzerland, 1989.
superalloys is commonly observed303 and was briefly 4. Moon, D. P. Mater. Sci. Technol. 1989, 5, 754–764.
reviewed by Krupp and Christ.304 Their study of 5. Strawbridge, A.; Hou, P. Y. Mater. High Temp. 1994, 12,
177–1781.
model NiCrAlTi alloys concluded that higher Cr 6. Pint, B. A. Mater. Res. Bull. 1994, 19(10), 26–30.
contents increased the N solubility and diffusivity in 7. Pint, B. A.; Garratt-Reed, A. J.; Hobbs, L. W. Mater. High
NiCrAlTi alloys. Temp. 1995, 13, 3–16.
8. Pint, B. A. Oxid. Met. 1996, 45, 1–37.
Bennett and coworkers305 also found internal 9. Stott, F. H. Mater. Sci. Forum 1997, 251–254, 19–32.
nitridation of FeCrAl by ‘crevice corrosion.’ In this 10. Quadakkers, J.; Singheiser, L. Mater. Sci. Forum 2001,
case, it was proposed that in crevices in components, 369–372, 77–92.
11. Pint, B. A. In Progress in Understanding the Reactive
O could be locally depleted in the gas due to reaction Element Effect Since the Whittle and Stringer Literature
with the surrounding metal and stagnant flow. The Review, Proceedings of John Stringer Symposium on
resulting high N and low O environment could cause High Temperature Corrosion; Tortorelli, P. F., Wright, I. G.,
Hou, P. Y., Eds.; ASM International: Materials Park, OH,
internal attack. 2003; pp 9–19.
12. Pint, B. A. Unpublished oxidation data, ORNL, Oak
Ridge, TN, 1994–2008.
1.24.7 Summary 13. Agarwal, D. C.; Brill, U. Performance of Alloy 602CA
(UNS N06025) in high temperature environments up to
Alumina-forming alloys are the most oxidation resis- 1200 C (2200 F), NACE Paper 00-521, Houston, TX,
Presented at NACE Corrosion 2000, Orlando, FL,
tant alloys due to their low oxidation rates and ther- March 2000.
modynamically stable scale. However, because of the 14. Pint, B. A.; Keiser, J. R. Alloy selection for high
negative effects of Al on alloy mechanical properties, temperature heat exchangers, NACE Paper 06-469,
Houston, TX, Presented at NACE Corrosion 2006,
fabrication, and joining, alumina-forming alloys are San Diego, CA, March 2006.
not widely used except when needed in very high 15. Pint, B. A.; Tortorelli, P. F.; Wright, I. G. Oxid. Met. 2002,
temperature applications or in particular environ- 58, 73–101.
16. Pint, B. A.; DiStefano, J. R.; Wright, I. G. Mater. Sci. Eng.
ments such as those with water vapor (or steam) and A 2006, 415, 255–263.
high C activities. As application temperatures 17. Lowell, C. E.; Barrett, C. A.; Palmer, R. W.; Auping, J. V.;
increase in search of higher generation and produc- Probst, H. B. Oxid. Met. 1991, 36, 81–112.
18. Nesbitt, J. A.; Vinarcik, E. J.; Barrett, C. A.; Doychak, J.
tion efficiencies, the use of alumina-forming alloys or Mater. Sci. Eng. A 1992, 153, 561–566.
coatings will inevitably increase. 19. Evans, H. E.; Lobb, R. C. Corros. Sci. 1984, 24,
209–222.
20. Przybylski, K.; Yurek, G. J. Mater. Sci. Forum 1989, 43,
Acknowledgments 1–74.
21. Heck, R. M.; Farrauto, R. J. Appl. Catal. A 2001, 221,
443–457.
The author would like to thank his colleagues at ORNL, 22. Kajita, S.; Dalla Betta, R. Catal. Today 2003, 83, 279–288.
acknowledge the discussions with I. G. Wright, appre- 23. Stainless Steel OhmaloyW 30 and OhmaloyW 40
Resistance Alloys; ATI Allegheny Ludlum: Brackenridge,
ciate the comments on the manuscript from M. P. Brady PA, 2007.
and P. F. Tortorelli, and gratefully acknowledge the 24. Benjamin, J. S. Metall. Trans. 1970, 1, 2943–2951.
assistance provided by the dedicated technical staff, 25. Nicholls, J. R.; Newton, R.; Simms, N. J.; Norton, J. F.
Mater. High Temp. 2003, 20, 93–108.
L. D. Chitwood, G. W. Garner, J. L. Moser, H. F. Long- 26. Pint, B. A.; Swindeman, R. W.; More, K. L.; Tortorelli,
mire and L. R. Walker for the experimental work. The P. F. Materials selection for high temperature (750–
manuscript preparation was sponsored by the U.S. 1000 C) metallic recuperators for improved efficiency
microturbines, ASME Paper #2001-GT-445, Presented at
Department of Energy, Fossil Energy Advanced Re- the International Gas Turbine and Aeroengine Congress
search Materials Program and work at the SHaRE and Exhibition, New Orleans, LA, June 4–7, 2001.
User Facility by the Division of Scientific User Facilities, 27. Wright, I. G.; Pint, B. A.; Lee, W. Y.; Alexander, K. B.;
Prüßner, K. In High Temperature Surface Engineering;
under contract DE-AC05-00OR22725 with UT- Nicholls, J., Rickerby, D., Eds.; Institute of Materials:
Battelle, LLC. London, UK, 2000; pp 95–113.
28. Smialek, J. L.; Pint, B. A. Mater. Sci. Forum 2001, 369– 60. Stott, F. H.; Wood, G. C.; Hobby, M. G. Oxid. Met. 1971,
372, 459–466. 3, 103–113.
29. Kirchheiner, R. R.; Becker, P.; Young, D. J.; Durham, R. 61. Wukusick, C. S.; Collins, J. F. Mater. Res. Stand. 1964, 4,
Improved oxidation and coking resistance of a 637–646.
new alumina forming Alloy 60 HT for the petrochemical 62. Sauer, J. P.; Rapp, R. A.; Hirth, J. P. Oxid. Met. 1982, 18,
industry, NACE Paper 05-428, Houston, TX, Presented at 285–294.
NACE Corrosion 2005, Houston, TX, April 2005. 63. De Freitas Cunha Lins, V.; Freitas, M. A.; De Paula
30. Haynes, J. A.; Pint, B. A.; Porter, W. D.; Wright, I. G. e Silva, E. M. Corros. Sci. 2004, 46, 1895–1907.
Mater. High Temp. 2004, 21, 87–94. 64. Boggs, W. E. Oxid. Met. 1976, 10, 277–289.
31. Toscano, J.; Gil, A.; Huttel, T.; Wessel, E.; Naumenko, D.; 65. Wu, Y.; Niu, Y. Corros. Sci. 2007, 49, 1656–1672.
Singheiser, L.; Quadakkers, W. J. Surf. Coat. Technol. 66. Sakiyama, M.; Tomaszewicz, P.; Wallwork, G. R. Oxid.
2007, 202, 603–607. Met. 1979, 13, 311–330.
32. Malang, S.; Deckers, H.; Fischer, U.; John, H.; 67. Yamamoto, Y.; Brady, M. P.; Lu, Z. P.; Maziasz, P. J.;
Meyder, R.; Norajitra, P.; Reimann, J. J.; Reiser, H.; Liu, C. T.; Pint, B. A.; More, K. L.; Meyer, H. M.;
Rust, K. Fusion Eng. Des. 1991, 14, 373–399. Payzant, E. A. Science 2007, 316, 433–436.
33. Fowler, J. D.; Causey, R. A.; Chandra, D.; 68. Brady, M. P.; Yamamoto, Y.; Santella, M. L.; Maziasz, P.
Elleman, T. S.; Verghese, K. J. Vac. Sci. Technol. 1976, J.; Pint, B. A.; Liu, C. T.; Lu, Z. P.; Bei, H. JOM 2008, 60(7),
13, 401–402. 12–18.
34. Steward, S. A. Review of hydrogen isotope permeability 69. Brady, M. P.; Yamamoto, Y.; Santella, M. L.; Pint, B. A.
through materials, LLNL Report #UCRL-53441, Lawrence Scr. Mater. 2007, 57, 1117–1120.
Livermore National Laboratory, Livermore, CA, 1983. 70. Castello, P.; Stott, F. H.; Gesmundo, F. Corros. Sci. 1999,
35. Konys, J.; Aiello, A.; Benamati, G.; Giancarli, L. Fusion 41, 901–918.
Sci. Technol. 2005, 47, 844–850. 71. Merz, M. D. Metall. Trans. A 1979, 10, 71–77.
36. Müller, G.; Heinzel, A.; Konys, J.; Schumacher, G.; 72. Baer, D. R.; Merz, M. D. Metall. Trans. A 1980, 11,
Weisenburger, A.; Zimmermann, F.; Engelko, V.; 1973–1980.
Rusanov, A.; Markov, V. J. Nucl. Mater. 2002, 301, 73. Yurek, G. J.; Eisen, D.; Garratt-Reed, A. J. Metall. Trans.
40–46. A 1982, 13, 473–485.
37. Lauf, R. J.; DeVan, J. H. J. Electrochem. Soc. 1992, 139, 74. Maloney, M. J. Ph.D. Thesis, Massachusetts Institute of
2087–2091. Technology, Cambridge, MA, 1989.
38. Hubberstey, P.; Sample, T.; Terlain, A. Fusion Technol. 75. Maloney, M. J.; Yurek, G. J. In Composites/Corrosion-
1995, 28, 1194–1199. Coatings of Advanced Materials, Proceedings IMAM-4;
39. Hubberstey, P. J. Nucl. Mater. 1997, 247, 208–214. Kimura, S., Kobayashi, A., Umekawa, S., Nii, K.,
40. Pint, B. A.; More, K. L. J. Nucl. Mater. 2008, 376, Saito, Y., Yoshimura, M., Eds.; Materials Research
108–113. Society: Pittsburgh, PA, 1989; Vol. 1, pp 383–388.
41. Quadakkers, W. J.; Bennett, M. J. Mater. Sci. Technol. 76. Goedjen, J. G.; Shores, D. A. Oxid. Met. 1992, 37,
1994, 10, 126–131. 125–142.
42. Quadakkers, W. J.; Bongartz, K. Mater. Corros. 1994, 45, 77. Wang, F. Oxid. Met. 1997, 48, 215–224.
232–241. 78. Chen, G.; Lou, H. Oxid. Met. 2000, 54, 155–162.
43. Stott, F. H.; Wood, G. C.; Stringer, J. Oxid. Met. 1995, 44, 79. Gleeson, B.; Wang, W.; Hayashi, S.; Sordelet, D. Mater.
113–145. Sci. Forum 2004, 461–464, 213–222.
44. Pettit, F. S. Trans. Metall. Soc. AIME 1967, 239, 80. Izumi, T.; Gleeson, B. Mater. Sci. Forum 2005, 522–523,
1296–1305. 221–228.
45. Pint, B. A. Surf. Coat. Technol. 2004, 188–189, 71–78. 81. Kartono, R.; Young, D. J. Mater. Corros. 2008, 59,
46. Prescott, R.; Graham, M. J. Oxid. Met. 1992, 38, 455–462.
233–254. 82. Pint, B. A.; Haynes, J. A.; More, K. L.; Schneibel, J. H.;
47. Sykes, C.; Bampfylde, J. W. J. Iron Steel Inst. Adv. Copy Zhang, Y.; Wright, I. G. In Superalloys 2008;
1934, 12, 22. Reed, R. C., et al. Ed.; TMS: Warrendale, PA, 2008;
48. Saegusa, F.; Lee, L. Corrosion 1966, 22, 168–177. pp 641–650.
49. Boggs, W. E. J. Electrochem. Soc. 1971, 83. Pint, B.A.; More, K.L.; Wright, I.G. Oxid. Met. in
118, 906–913. preparation.
50. DeVan, J. H.; Tortorelli, P. F. Corros. Sci. 1993, 35, 84. Felten, E. J. Oxid. Met. 1976, 10, 23–28.
1065–1071. 85. Tatlock, G. J.; Hurd, T. J. Oxid. Met. 1984, 22, 201–226.
51. Pint, B. A.; Leibowitz, J.; DeVan, J. H. Oxid. Met. 1999, 86. Fountain, J. G.; Golightly, F. A.; Stott, F. H.; Wood, G. C.
51, 181–197. Oxid. Met. 1976, 10, 341–345.
52. Pint, B. A.; Wright, I. G. Mater. Sci. Forum 2004, 461–464, 87. Ballard, D. L.; Martin, P. L.; Gleeson, B.; Menon, S.
799–806. Structure and oxidation behavior of PGM modified
53. Zhang, Z. G.; Gesmundo, F.; Hou, P. Y.; Niu, Y. Corros. nickel-base alloys, Presentation at 18th AeroMat
Sci. 2006, 48, 741–765. Conference and Expo, Baltimore, MD, ASM, June 2007.
54. Wagner, C. Corros. Sci. 1965, 5, 751–764. 88. Doychak, J.; Rühle, M. Oxid. Met. 1989, 31,
55. Zhang, X. J.; Wang, S. Y.; Gesmundo, F.; Niu, Y. Oxid. 431–452.
Met. 2006, 65, 151–165. 89. Pint, B. A.; Garratt-Reed, A. J.; Hobbs, L. W. Oxid. Met.
56. Pint, B. A.; Paul, L. D. Oxidation behavior of welded and 2001, 56, 119–145.
base metal UNS N06025, NACE Paper 07-470, Houston, 90. Copland, E. J. Phase Equilib. Diff. 2007, 28, 38–48.
TX, Presented at NACE Corrosion 2007, Nashville, TN, 91. Deodeshmukh, V.; Mu, N.; Li, B.; Gleeson, B. Surf. Coat.
March 2007. Technol. 2006, 201, 3836–3840.
57. Giggins, C. S.; Pettit, F. S. J. Electrochem. Soc. 1971, 92. Zhang, Y.; Haynes, J. A.; Pint, B. A.; Wright, I. G. Surf.
118, 1782–1790. Coat. Technol. 2005, 200, 1259–1263.
58. Wallwork, G. R.; Hed, A. Z. Oxid. Met. 1971, 3, 171–184. 93. Pint, B. A.; Dwyer, M. J.; Deacon, R. M. Oxid. Met. 2008,
59. Wallwork, G. R. Rep. Prog. Phys. 1976, 39, 401–485. 69, 211–231.
94. Pint, B. A.; More, K. L.; Wright, I. G. Mater. High Temp. 125. Evans, H. E.; Norfolk, D. J.; Swan, T. J. Electrochem.
2003, 20, 375–386. Soc. 1978, 125, 1180–1185.
95. Hagel, W. C. Corrosion 1965, 21, 316–326. 126. Doychak, J. In Intermetallic Compounds, Principles;
96. Kear, B. H.; Pettit, F. S.; Fornwalt, D. E.; Lemaire, L. P. Westbrook, J. H., Fleischer, R. L., Eds.; John Wiley &
Oxid. Met. 1971, 3, 557–569. Sons: New York, NY, 1994; Vol. 1, pp 977–1016.
97. Smialek, J. L.; Gibala, R. Metall. Trans. A 1983, 14, 127. Pint, B. A.; Wright, I. G.; Lee, W. Y.; Zhang, Y.;
2143–2161. Prüßner, K.; Alexander, K. B. Mater. Sci. Eng. A 1998,
98. Quadakkers, W. J.; Elschner, A.; Speier, W.; Nickel, H. 245, 201–211.
Appl. Surf. Sci. 1991, 52, 271–287. 128. Pint, B. A.; More, K. L.; Tortorelli, P. F.; Porter, W. D.;
99. Rybicki, G. C.; Smialek, J. L. Oxid. Met. 1989, 31, Wright, I. G. Mater. Sci. Forum 2001, 369–372, 411–418.
275–304. 129. Pint, B. A. J. Am. Ceram. Soc. 2003, 86, 686–695.
100. Brumm, M. W.; Grabke, H. J. Corros. Sci. 1992, 33, 130. Hindam, H.; Whittle, D. P. J. Electrochem. Soc. 1982,
1677–1690. 129, 1147–1149.
101. Young, E. W. A.; de Wit, J. H. W. Oxid. Met. 1986, 26, 131. Kvernes, I. Oxid. Met. 1973, 6, 45–64.
351–361. 132. Allam, I. M.; Whittle, D. P.; Stringer, J. Oxid. Met. 1978,
102. Doychak, J.; Smialek, J. L.; Mitchell, T. E. Metall. Trans. A 12, 35–66.
1989, 20, 499–518. 133. Pivin, J. C.; Delaunay, D.; Roques-Carmes, C.; Huntz, A.
103. Prüßner, K.; Bruley, J.; Salzberger, U.; Zweygart, H.; M.; Lacombe, P. Corros. Sci. 1980, 20, 351–373.
Schumann, E.; Rühle, M. In Microscopy of Oxidation 2; 134. Clemens, D.; Bongartz, K.; Speier, W.; Hussey, R. J.;
Newcomb, S. B., Bennett, M. J., Eds.; Institute of Metals: Quadakkers, W. J. Fresenius J. Anal. Chem. 1993, 346,
London, 1993; pp 435–444. 318–322.
104. Pint, B. A.; Martin, J. R.; Hobbs, L. W. Solid State Ionics 135. Pint, B. A.; Alexander, K. B. J. Electrochem. Soc. 1998,
1995, 78, 99–107. 145, 1819–1829.
105. Yang, J. C.; Schumann, E.; Levin, I.; Rühle, M. Acta 136. Balmain, J.; Loudjani, M. K.; Huntz, A. M. Mater. Sci. Eng.
Mater. 1998, 46, 2195–2201. A 1997, 224, 87–100.
106. Bye, G. C.; Simpkin, G. T. J. Am. Ceram. Soc. 1974, 57, 137. Yurek, G. J. In Corrosion Mechanisms; Mansfeld, F., Ed.;
367–371. Marcel Dekker: New York, 1987; pp 397–446.
107. Pint, B. A.; Jain, A.; Hobbs, L. W. In High Temperature 138. Oishi, Y.; Kingery, W. D. J. Chem. Phys. 1960, 33, 480–486.
Ordered Intermetallics V, Symposium Proceedings; 139. Paladino, A. E.; Kingery, W. D. J. Am. Ceram. Soc. 1962,
Baker, I., Darolia, R., Whittenberger, J. D., Yoo, M. H., 37, 957–962.
Eds.; Materials Research Society: Pittsburgh, PA, 1993; 140. Pint, B. A.; Deacon, R. M. J. Appl. Phys. 2005, 97(11),
Vol. 288, pp 1013–1018. 116111.
108. Ramanarayanan, T. A.; Raghavan, M.; Petkovic- 141. Heuer, A. H. J. Eur. Ceram. Soc. 2008, 28, 1495–1507.
Luton, R. In Proceedings of JIMIS-3, High Temperature 142. Kim, C. K. Oxid. Met. 1996, 45, 133–152.
Corrosion Transactions Supplement Japan Institute of 143. Tien, J. K.; Pettit, F. S. Metall. Trans. A 1972, 3,
Metals: Tokyo, Japan, 1983; Vol. 24, pp 199–206. 1587–1599.
109. Quadakkers, W. J.; Schmidt, K.; Grubmeier, H.; 144. Cotell, C. M.; Yurek, G. J.; Hussey, R. J.; Mitchell, D. F.;
Wallura, E. Mater. High Temp. 1992, 10, 23–32. Graham, M. J. J. Electrochem. Soc. 1987, 134,
110. Dryepondt, S.; Pint, B. A.; Lara-Curzio, E. Acta Mater. in 1871–1872.
preparation. 145. Cotell, C. M.; Yurek, G. J.; Hussey, R. J.; Mitchell, D. F.;
111. Bennett, M. J.; Newton, R.; Nicholls, J. R.; Graham, M. J. Oxid. Met. 1990, 34, 173–200.
Al-Badairy, H.; Tatlock, G. J. Mater. Sci. Forum 2004, 146. Jedlinski, J.; Borchardt, G.; Mrowec, S. Solid State Ionics
461–464, 463–472. 1992, 50, 67–74.
112. N’Dah, E.; Galerie, A.; Wouters, Y.; Goossens, D.; 147. Felten, E. J.; Pettit, F. S. Oxid. Met. 1976, 10, 189–223.
Naumenko, D.; Kochubey, V.; Quadakkers, W. J. Mater. 148. Reddy, K. P. R.; Smialek, J. L.; Cooper, A. R. Oxid. Met.
Corros. 2005, 56, 843–847. 1982, 17, 429–449.
113. Burtin, P.; Brunelle, J. P.; Pijolat, M.; Soustelle, M. Appl. 149. Quadakkers, W. J.; Holzbrecher, H.; Briefs, K. G.;
Catal. 1987, 34, 225–238. Beske, H. Oxid. Met. 1989, 32, 67–88.
114. Pint, B. A.; Treska, M.; Hobbs, L. W. Oxid. Met. 1997, 47, 150. Pint, B. A.; Martin, J. R.; Hobbs, L. W. Oxid. Met. 1993,
1–20. 39, 167–195.
115. Pijolat, M.; Dauzat, M.; Soustelle, M. Thermochim. Acta 151. Versaci, R. A.; Clemens, D.; Quadakkers, W. J.;
1987, 122, 71–77. Hussey, R. Solid State Ionics 1993, 59, 235–242.
116. Moseley, P. T.; Hyde, K. R.; Bellamy, B. A.; Tappin, G. 152. Prescott, R.; Mitchell, D. F.; Graham, M. J.; Doychak, J.
Corros. Sci. 1984, 24, 547–565. Corros. Sci. 1995, 37, 1341–1364.
117. Hindam, H. M.; Whittle, D. P. Oxid. Met. 1982, 18, 153. Mennicke, C.; Schumann, E.; Rühle, M.; Hussey, R. J.;
245–284. Sproule, G. I.; Graham, M. J. Oxid. Met. 1998, 49,
118. Hou, P. Y.; Shui, Z. R.; Chuang, G. Y.; Stringer, J. 455–466.
J. Electrochem. Soc. 1992, 139, 1119–1126. 154. Ecer, G. M.; Singh, R. B.; Meier, G. H. Oxid. Met. 1982,
119. Wagner, C. Atom Movements; American Society of 18, 55–81.
Metals: Cleveland, OH, 1951; pp 153–173. 155. Kofstad, P. Oxid. Met. 1985, 24, 265–276.
120. Quadakkers, W. J. Mater. Corros. 1990, 41, 659–668. 156. Ramanarayanan, T. A.; Raghavan, M.; Petkovic-
121. Quadakkers, W. J.; Naumenko, D.; Wessel, E.; Luton, R. J. Electrochem. Soc. 1984, 131, 923–931.
Kochubey, V.; Singheiser, L. Oxid. Met. 2004, 61, 17–37. 157. Przybylski, K.; Garratt-Reed, A. J.; Pint, B. A.; Katz, E. P.;
122. Saunders, S. R. J.; Evans, H. E.; Li, M.; Gohil, D. D.; Yurek, G. J. J. Electrochem. Soc. 1987, 134, 3207–3208.
Osgerby, S. Oxid. Met. 1997, 48, 189–200. 158. Li, C. W.; Kingery, W. D. In Structure and Properties of
123. Tolpygo, V. K.; Dryden, J. R.; Clarke, D. R. Acta Mater. MgO and Al2O3 Ceramics, Advances in Ceramics;
1998, 46, 927–937. American Ceramic Society: Columbus, OH, 1984;
124. Hou, P. Y.; Paulikas, A. P.; Veal, B. W. Mater. Sci. Forum Vol. 10, pp 368–378.
2004, 461–464, 671–680. 159. Kingery, W. D. Solid State Ionics 1984, 12, 299–307.
160. Pint, B. A.; Garratt-Reed, A. J.; Hobbs, L. W. J. Am. 194. Grabke, H. J.; Weimer, D.; Viefhaus, H. Appl. Surf. Sci.
Ceram. Soc. 1998, 81, 305–314. 1991, 47, 243–250.
161. Fang, J.; Thompson, A. M.; Harmer, M. P.; Chan, H. M. 195. Jedlinski, J.; Bennett, M. J.; Evans, H. E. Mater. High
J. Am. Ceram. Soc. 1997, 80, 2005–2012. Temp. 1994, 12, 169–175.
162. Bedu-Amissah, K.; Rickman, J. M.; Chan, H. M.; 196. Hou, P. Y.; Saunders, S. R. J. Mater. High Temp. 2005,
Harmer, M. P. J. Am. Ceram. Soc. 2007, 90, 1551–1555. 22, 121–130.
163. Cho, J.; Harmer, M. P.; Chan, H. M.; Rickman, J. M.; 197. Hou, P. Y.; Priimak, K. Oxid. Met. 2005, 63, 113–130.
Thompson, A. M. J. Am. Ceram. Soc. 1997, 80, 198. Stott, F. H.; Golightly, F. A.; Wood, G. C. Corros. Sci.
1013–1017. 1979, 19, 889–906.
164. Yoshida, H.; Ikuhara, Y.; Sakuma, T. Acta Mater. 2002, 199. Kumar, A.; Nasrallah, M.; Douglass, D. L. Oxid. Met.
50, 2955–2966. 1974, 8, 227–263.
165. Veal, B. W.; Paulikas, A. P.; Gleeson, B.; Hou, P. Y. Surf. 200. Giggins, C. S.; Kear, B. H.; Pettit, F. S.; Tien, J. K. Metall.
Coat. Technol. 2007, 202, 608–612. Trans. 1974, 5, 1685–1688.
166. Atkinson, A.; Taylor, R. I.; Hughes, A. E. Philos. Mag. A 201. Antill, J. E.; Peakall, K. A. J. Iron Steel Inst. Adv. Copy
1982, 45, 979–998. 1967, 205, 1136–1142.
167. Sheasby, J. S.; Jory, D. B. Oxid. Met. 1977, 12, 527–539. 202. Kingsley, L. M.; Stringer, J. Oxid. Met. 1989, 32, 371–378.
168. Huntz, A. M.; Moulin, G.; Lesage, B. Contribution of 203. Pint, B. A. Oxid. Met. 1998, 49, 531–560.
Combined Techniques for Studying Oxidation 204. Hou, P. Y. Oxid. Met. 1999, 52, 337–351.
Mechanisms, Proceedings of the 9th International 205. Hou, P. Y.; Stringer, J. J. Phys. IV 1993, 3(C9), 231–240.
Congress on Metallic Corrosion; National Research 206. Quadakkers, W. J.; Huczkowski, P.; Naumenko, D.;
Council of Canada: Ottawa, Canada, 1984; Vol. 2, Zurek, J.; Meier, G. H.; Niewolak, L.; Singheiser, L. Mater.
pp 400–405. Sci. Forum 2008, 595–598, 1111–1118.
169. Nicolas-Chaubet, D.; Huntz, A. M.; Millot, F. Coll. Phys. 207. Gupta, D. K.; Duvall, D. S. In Superalloys; Gell, M., et al.
1990, 51, 1015–1020. Eds.; TMS: Warrendale, PA, 1984; pp 711–720.
170. Ben Abderrazik, G.; Millot, F.; Moulin, G.; Huntz, A. M. 208. Sigler, D. R. Oxid. Met. 1989, 32, 337–355.
J. Am. Ceram. Soc. 1985, 68, 307–314. 209. Ikeda, Y.; Nii, K.; Yoshihara, K. Trans. Jpn. Inst. Met.
171. Tortorelli, P. F.; More, K. L.; Specht, E. D.; Pint, B. A.; 1983, 24, 207–214.
Zscack, P. Mater. High Temp. 2003, 20, 303–310. 210. Funkenbush, A. W.; Smeggil, J. G.; Bornstein, N. S.
172. Veal, B. W.; Paulikas, A. P.; Hou, P. Y. Nat. Mater. 2006, Metall. Trans. A 1985, 16, 1164–1166.
5, 349–351. 211. Smeggil, J. G.; Funkenbusch, A. W.; Bornstein, N. S.
173. Hou, P. Y.; Paulikas, A. P.; Veal, B. W. Mater. Sci. Forum Metall. Trans. A 1986, 17, 923–932.
2006, 522–523, 433–440. 212. Smialek, J. L. Metall. Trans. A 1987, 18, 164–167.
174. Reddy, A.; Hovis, D. B.; Heuer, A. H.; Paulikas, A. P.; 213. Quadakkers, W. J.; Wasserfuhr, C.; Khanna, A. S.;
Veal, B. W. Oxid. Met. 2007, 67, 153–177. Nickel, H. Mater. Sci. Technol. 1988, 4, 1119–1125.
175. Lipkin, D. M.; Clarke, D. R. Oxid. Met. 1995, 45, 267–280. 214. Smialek, J. L.; Jayne, D. T.; Schaeffer, J. C.; Murphy, W. H.
176. Christensen, R. J.; Tolpygo, V. K.; Clarke, D. R. Acta Thin Solid Films 1994, 253, 285–292.
Mater. 1997, 45, 1761–1766. 215. Meier, G. H.; Pettit, F. S.; Smialek, J. L. Mater. Corros.
177. Rhines, F. N.; Wolf, J. S. Metall. Trans. 1970, 1, 1995, 46, 232–240.
1701–1710. 216. Sarioglu, C.; Stinner, C.; Blachere, J. R.; Birks, N.;
178. Golightly, F. A.; Stott, F. H.; Wood, G. C. Oxid. Met. 1976, Pettit, F. S.; Meier, G. H. In Superalloys; Kissinger, R. D.,
10, 163–187. Deye, D. J., Anton, D. L., Cetel, A. D., Nathal, M. V.,
179. Golightly, F. A.; Stott, F. H.; Wood, G. C. J. Electrochem. Pollack, T. M., Woodford, D. A., Eds.; TMS: Warrendale,
Soc. 1979, 126, 1035–1042. PA, 1996; pp 71–80.
180. Tolpygo, V. K.; Grabke, H. J. Oxid. Met. 1994, 41, 217. Lees, D. G. Oxid. Met. 1987, 27, 75–81.
343–364. 218. Hou, P. Y.; Stringer, J. Oxid. Met. 1992, 38, 323–345.
181. Tolpygo, V. K. Oxid. Met. 1999, 51, 449–477. 219. Hou, P. Y. Annu. Rev. Mater. Res. 2008, 38, 275–298.
182. Stott, F. H.; Wood, G. C.; Golightly, F. A. Corros. Sci. 220. Briant, C. L.; Luthra, K. L. Metall. Trans. A 1988, 19,
1979, 19, 869–887. 2099–2108.
183. Fox, P.; Tatlock, G. J. Mater. Sci. Technol. 1988, 4, 221. Yang, S.; Wang, F. Oxid. Met. 2006, 65, 195–205.
439–445. 222. Tolpygo, V. K.; Grabke, H. J. Scr. Mater. 1997, 38, 123–129.
184. Tolpygo, V. K.; Clarke, D. R. Acta Mater. 1998, 46, 223. Smialek, J. L. Oxid. Met. 2001, 55, 75–86.
5153–5166. 224. Harris, K.; Wahl, J. B. In Superalloys; Green, K. A.,
185. Mennicke, C.; Schumann, E.; Ulrich, C.; Rühle, M. Mater. Pollock, T. M., Harada, H., Howson, T. E., Reed, R. C.,
Sci. Forum 1997, 251–254, 389–396. Schirra, J., Walston, S., Eds.; TMS: Warrendale, PA,
186. Pint, B. A. Oxid. Met. 1997, 48, 303–328. 2004; pp 45–52.
187. Dryepondt, S.; Pint, B. A. Mater. Sci. Eng. A 2008, 497, 225. Schumann, E.; Yang, J. C.; Graham, M. J. Scr. Mater.
224–230. 1996, 34, 1365–1370.
188. Pint, B. A. Mater. Sci. Forum 1997, 251–254, 397–404. 226. Mendis, B. G.; Livi, K. J. T.; Hemker, K. J. Scr. Mater.
189. Pint, B. A.; Garratt-Reed, A. J.; Hobbs, L. W. J. Phys. IV 2006, 55, 589–592.
1993, 3(C9), 247–255. 227. Lowrie, D.; Boone, D. H. Thin Solid Films 1977, 45,
190. Nychka, J. A.; Clarke, D. R. Oxid. Met. 2005, 63, 491–498.
325–352. 228. Allam, I. M.; Akuezue, H. C.; Whittle, D. P. Oxid. Met.
191. Lance, M. J.; Pint, B. A. Unpublished data; ORNL, Oak 1980, 14, 517–530.
Ridge, TN, 1999. 229. Amano, T.; Takezawa, Y.; Shiino, A.; Shishido, T. J. Alloys
192. Naumenko, D.; Kochubey, V.; Le Coze, J.; Wessel, E.; Compd. 2008, 452, 16–22.
Singheiser, L.; Quadakkers, W. J. Mater. Sci. Forum 230. Dickey, E. C.; Pint, B. A.; Alexander, K. B.; Wright, I. G.
2004, 461–464, 489–495. J. Mater. Res. 1999, 14, 4531–4540.
193. Provenzano, V.; Sadananda, K.; Louat, N. P.; Reed, J. R. 231. Smialek, J. L.; Doychak, J.; Gaydosh, D. J. Oxid. Met.
Surf. Coat. Technol. 1988, 36, 61–74. 1990, 34, 259–275.
232. Wright, I. G.; Pint, B. A.; Tortorelli, P. F. Oxid. Met. 2001, 266. Kvernes, I.; Olivera, M.; Kofstad, P. Corros. Sci. 1977, 17,
55, 333–357. 237–252.
233. Pint, B. A.; Porter, W. D.; Wright, I. G. Mater. Sci. Forum 267. Pint, B. A.; Peraldi, R.; Maziasz, P. J. Mater. Sci. Forum
2008, 595–598, 1083–1092. 2004, 461–464, 815–822.
234. Smialek, J. L.; A Nesbitt, J.; Barrett, C. A.; Lowell, C. E. In 268. Pint, B. A.; More, K. L.; Trejo, R.; Lara-Curzio, E. J. Eng.
Cyclic Oxidation of High Temperature Materials; Gas Turbines Power 2008, 130, 012101.
Schütze, M., Quadakkers, W. J., Eds.; Institute of 269. Agüero, A.; Muelas, R.; Gutiérrez, M.; Van Vulpen, R.;
Materials: London, UK, 1999; pp 148–168. Osgerby, S.; Banks, J. P. Surf. Coat. Technol. 2007, 201,
235. Evans, H. E. Int. Mater. Rev. 1995, 40, 1–40. 6253–6260.
236. Nicholls, J. R.; Evans, H. E.; Saunders, S. R. J. Mater. 270. Regina, J. R.; DuPont, J. N.; Marder, A. R. Oxid. Met.
High Temp. 1997, 14, 5–13. 2004, 61, 69–90.
237. Jönsson, B.; Berglund, R.; Magnusson, J.; Henning, P.; 271. Leyens, C.; Fritscher, K.; Gehrling, R.; Peters, M.;
Hättestrand, M. Mater. Sci. Forum 2004, 461–464, Kaysser, W. A. Surf. Coat. Technol. 1996, 82, 133–144.
455–462. 272. Janakiraman, R.; Meier, G. H.; Pettit, F. S. Metall. Mater.
238. Herbelin, J. M.; Mantel, M. Mater. Sci. Forum 1997, Trans. A 1999, 30, 2905–2913.
251–254, 349–356. 273. Onal, K.; Maris-Sida, M. C.; Meier, G. H.; Pettit, F. S.
239. Lowell, C. E.; Deadmore, D. L. Oxid. Met. 1980, 14, Mater. High Temp. 2003, 20, 327–337.
325–336. 274. Maris-Sida, M. C.; Meier, G. H.; Pettit, F. S. Metall. Mater.
240. Wagner, C. J. Electrochem. Soc. 1952, 99, 369–380. Trans. A 2003, 34, 2609–2619.
241. Douglass, D. L. Oxid. Met. 1995, 44, 81–111. 275. Smialek, J. L. JOM 2006, 58(1), 29–35.
242. Wolf, J. S.; Evans, E. B. Corrosion 1962, 18, 129–136. 276. Al-Badairy, H.; Tatlock, G. J. Mater. High Temp. 2000,
243. Whittle, D. P.; Shida, Y.; Wood, G. C.; Stott, F. H.; 17, 133–137.
Bastow, B. D. Philos. Mag. A 1982, 46, 931–949. 277. Essuman, E.; Meier, G. H.; Zurek, J.; Hänsel, M.;
244. Hindam, H. M.; Smeltzer, W. W. J. Electrochem. Soc. Quadakkers, W. J. Oxid. Met. 2008, 69, 143–162.
1980, 127, 1622–1635. 278. Alcock, C. B.; Hocking, M. G.; Zador, S. Corros. Sci.
245. Barrett, C. A.; Lowell, C. E. Oxid. Met. 1978, 1969, 9, 111–122.
12, 293–311. 279. Vineberg, E. J.; Douglass, D. L. Oxid. Met. 1986, 25,
246. Allam, I. M.; Whittle, D. P.; Stringer, J. Oxid. Met. 1979, 1–28.
13, 381–401. 280. Wright, I. G. In Proceedings of the 1974 Gas Turbine
247. Bennett, M. J.; Romary, H.; Price, J. B. In Heat Resistant Materials in the Marine Environment Conference;
Materials; Natesan, K., Tillack, D. J., Eds.; ASM: Materials Fairbanks, J. W., Machlan, I., Eds.; Battelle: Columbus,
Park, OH, 1991; pp 95–103. OH, 1975; pp 357–377.
248. Gurrappa, I.; Weinbruch, S.; Naumenko, D.; 281. Nakamori, M.; Kayano, I.; Tsukuda, Y.; Takahashi, K.;
Quadakkers, W. J. Mater. Corros. 2000, 51, 224–235. Torigoe, T. Mater. Sci. Forum 1997, 251–254, 633–640.
249. Pint, B. A.; Walker, L. R.; Wright, I. G. Mater. High Temp. 282. Narita, T.; Przybylski, K.; Smeltzer, W. W. Oxid. Met.
2004, 21, 175–185. 1984, 22, 181–200.
250. Wilber, J. P.; Bennett, M. J.; Nicholls, J. R. Mater. High 283. Setterlund, R. B.; Prescott, G. R. Corrosion 1961, 17,
Temp. 2000, 17, 125–132. 103–108.
251. Wright, I. G.; Pint, B. A.; Hall, L. M.; Tortorelli, P. F. In 284. Huang, T. T.; Lin, Y. C.; Shores, D. A.; Pfender, E.
Lifetime Modelling of High Temperature Corrosion J. Electrochem. Soc. 1984, 131, 2191–2196.
Processes; Schütze, M., Quadakkers, W. J., Nicholls, J. R., 285. Saxena, D.; Prakash, S.; Mehta, M. L.; Saraswat, I. P.
Eds.; Maney: London, 2001; pp 299–318. Oxid. Met. 1987, 28, 127–153.
252. Wright, I. G.; Peraldi, R.; Pint, B. A. Mater. Sci. Forum 286. DeVan, J. H.; Tortorelli, P. F. Mater. High Temp. 1993,
2004, 461–464, 579–590. 11, 30–35.
253. Nesbitt, J. A.; Heckel, R. W. Oxid. Met. 1988, 29, 75–102. 287. Nyassi, A.; Larpin, J. P.; Bendriss, A. Oxid. Met. 1995,
254. Nesbitt, J. A. NASA Glenn Research Center: Cleveland, 43, 543–560.
OH, 2000NASA/TM, 2000-209271. 288. Banovic, S. W.; Du Pont, J. N.; Marder, A. R. Metall.
255. Pint, B. A.; Walker, L. R.; Wright, I. G. Mater. High Temp. Mater. Trans. A 2000, 31, 1805–1817.
2009, in press. 289. Pint, B. A.; Zhang, Y.; Tortorelli, P. F.; Haynes, J. A.;
256. Evans, H. E.; Donaldson, A. T.; Gilmour, T. C. Oxid. Met. Wright, I. G. Mater. High Temp. 2001, 18, 185–192.
1999, 52, 379–402. 290. Mari, P. A.; Chaix, J. M.; Larpin, J. P. Oxid. Met. 1982, 17,
257. Hiramatsu, N.; Stott, F. H. Oxid. Met. 1999, 51, 315–328.
479–494. 291. Stott, F. H.; Chong, F. M. F.; Stirling, C. A. In Proceedings
258. Kochubey, V.; Naumenko, D.; Wessel, E.; Le Coze, J.; of the 9th International Congress on Metallic Corrosion
Singheiser, L.; Al-Baidary, H.; Tatlock, G. J.; National Research Council of Canada: Ottawa, Canada,
Quadakkers, W. J. Mater. Lett. 2006, 60, 1564–1568. 1984; Vol. 2, pp 1–16.
259. Fritscher, K.; Leyens, C.; Peters, M. Mater. Sci. Eng. A 292. Taniguchi, S.; Shibata, T. Oxid. Met. 1989, 32, 391–404.
1995, 190, 253–258. 293. Kim, Y.-K.; Przybylski, K.; Yurek, G. J. In Fundamental
260. Kofstad, P. In Microscopy of Oxidation I; Bennett, M. J., Aspects of High Temperature Corrosion; Shores, D.,
Lorimer, G. W., Eds.; Institute of Metals: London, 1991; Yurek, G. J., Eds.; Electrochemical Society: Pennington,
pp 2–9. NJ, 1986; Vol. II, pp 259–281.
261. Pint, B. A.; Haynes, J. A.; Zhang, Y.; More, K. L.; Wright, I. G. 294. Schaeffer, J.; Kim, G. M.; Meier, G. H.; Pettit, F. S.
Surf. Coat. Technol. 2006, 201, 3852–3856. In Lang, E., Ed.; The Role of Active Elements in the
262. Opila, E. J. Mater. Sci. Forum 2004, 461–464, 765–774. Oxidation Behavior of High Temperature Metals and
263. Asteman, H.; Svensson, J.-E.; Johansson, L.-G.; Alloys; Elsevier Applied Science: London, 1989;
Norell, M. Oxid. Met. 1999, 52, 95–111. pp 231–267.
264. Opila, E. J.; Hann, R. E., Jr. J. Am. Ceram. Soc. 1997, 80, 295. Rapp, R. A. Mater. Sci. Eng. 1987, 87, 319–327.
197–205. 296. Leyens, C.; Wright, I. G.; Pint, B. A. Mater. Sci. Forum
265. Young, D. J.; Pint, B. A. Oxid. Met. 2006, 66, 137–153. 2001, 369–372, 571–578.
297. Grabke, H. J. Mater. Corros. 1998, 49, 303–308. 302. Han, S.; Young, D. J. Oxid. Met. 2001, 55, 223–242.
298. Loudjani, M. K.; Pivin, J. C.; Huntz, A. M.; Davidson, J. H. 303. Stott, F. H.; Berg, S.; Sang, M.; Karim, N. In High
Corros. Sci. 1988, 28, 1075–1088. Temperature Materials for Power Engineering; Bachelet,
299. Jönsson, B.; Svedberg, C. Mater. Sci. Forum 1997, 251– et al. Eds.; Kluwer Academic Publishers: Dordrecht,
254, 551–558. Netherlands, 1990; Vol. 1, pp 213–226.
300. Turker, M.; Hughes, T. A. In Microscopy of Oxidation; 304. Krupp, U.; Christ, H.-J. Metall. Mater. Trans. A 2000, 31,
Newcomb, S. B., Bennett, M. J., Eds.; Institute of Metals: 47–56.
London, 1993; Vol. 2, pp 310–320. 305. Bennett, M. J.; Nicholls, J. R.; Borchardt, G.; Strehl, G.
301. Bennett, M. J.; Newton, R.; Nicholls, J. R. Mater. High Mater. High Temp. 2002, 19, 117–125.
Temp. 2003, 20, 347–356.
1.25 High Temperature Corrosion of Intermetallic Alloys
J. W. Fergus
Materials Research and Education Center, 275 Wilmore Laboratories, Auburn University, Auburn, AL 36849, USA

1.25.1.1 General Uses of Intermetallic Alloys 647
1.25.1.2 Common Intermetallic Alloys 647
1.25.1.2.1 Nickel aluminides 647
1.25.1.2.2 Iron aluminides 648
1.25.1.2.3 Titanium aluminides 649
1.25.2 Oxidation 649
1.25.2.1 Nickel and Iron Aluminides 650
1.25.2.1.1 Scale properties/structure 650
1.25.2.1.2 Effect of atmospheres/impurities 654
1.25.2.1.3 Effect of alloying additions 655
1.25.2.2 Titanium Aluminides 656
1.25.2.2.1 Scale properties/structure 656
1.25.2.2.2 Effect of atmosphere and pretreatment 657
1.25.2.2.3 Effect of alloying additions 658
1.25.2.3 Other Aluminides 659
1.25.3 Hot Corrosion 660
1.25.3.1 Sulfur-Containing Environments 660
1.25.3.1.1 Gaseous 660
1.25.3.1.2 Molten salts 661
1.25.3.2 Chlorine-Containing Environments 661
1.25.3.2.1 Gaseous 661
1.25.3.2.2 Molten salts 662
1.25.3.3 Others 662
1.25.4 Coatings 662
1.25.4.1 Aluminide Coatings with Same Base as Substrate 663
1.25.4.1.1 Simple aluminide coatings 663
1.25.4.1.2 Alloyed aluminide coatings 663
1.25.4.2 Aluminide Coatings with Different Base as Substrate 665
1.25.5 Summary and Conclusions 666
References 666
Glossary Kirkendall effect Vacancy formation in a diffusion

B2 Crystal structure formed by NiAl and FeAl (see couple because of differences in the
Figure 1(a)). interdiffusion coefficients of the two atoms.
Corundum Crystal structure formed by a-Al2O3. Laves phase Intermetallic crystal structure.
D03 Crystal structure formed by Fe3Al (see L10 Crystal structure formed by TiAl (see
Figure 1(d)). Figure 1(b)).
D019 Crystal structure formed by Ti3Al (see L12 Crystal structure formed by Ni3Al (see
Figure 1(e)). Figure 1(c)).
Hot corrosion Corrosion in a high temperature Pack cementation Chemical vapor deposition
environment where molten salts are present. process performed in a packed bed.
646
High Temperature Corrosion of Intermetallic Alloys 647
because of the difficulty of slip in their ordered struc-

Rutile Crystal structure formed by TiO2.
tures, which leads to low fracture toughness and cre-
Spinel Crystal structure formed by NiAl2O4 and
ates challenges in the fabrication of intermetallic alloys
other mixed oxides.
and their implementation in engineering structures.
X-phase Mixed titanium–aluminum oxide phase.
The synthesis of intermetallic alloys is complicated
Z-phase Mixed titanium–aluminum oxide phase.
because of the larger differences in melting tempera-
ture among the constituent elements, which limits the
available fabrication processes. In addition, the proper-
ties of intermetallic alloys depend strongly on small
Abbreviations changes in composition, so compositional control dur-
CVD Chemical vapor deposition ing synthesis and processing is particularly important.
NCP New cubic phase (formed in Ti–Al–O system) Despite these significant engineering challenges,
TBC Thermal barrier coating the attractive high temperature properties of inter-
TCP Topographically close packed (phases) metallic alloys have led to their use in a variety of
TGO Thermally grown oxide high temperature applications.1–3 One of the most
promising applications of intermetallic alloys is for
components in aeroturbine engines, where increases
in operating temperature and decreases in weight
1.25.1 Introduction
improve engine efficiency and performance. Although
low density is particularly important in aerospace
Metal alloys, such as those discussed in the chapter on
applications, intermetallic alloys have also been used
High Temperature Corrosion of Alumina-forming
in other applications, such as valves and turbocharger
Iron, Nickel and Cobalt-base Alloys can be strength-
rotors in high-performance internal combustion
ened by the formation of intermetallic compounds, but
engines for automotive applications.
alloys consisting entirely of intermetallic compounds
are stronger. These intermetallic compounds typically
have higher melting temperatures compared with their
1.25.1.2 Common Intermetallic Alloys
constituent metals, and are thus attractive materials for
high temperature applications. The most widely used Many of the promising intermetallic aluminide
intermetallic compounds for high temperature appli- alloys have been developed from the nickel- and
cations are aluminides, because of their low densities iron-based high temperature alloys described in the
and their ability, or at least potential, to form a protec- chapter on High Temperature Corrosion of
tive alumina scale. Silicides are also used for some high Alumina-forming Iron, Nickel and Cobalt-base
temperature applications, but their oxidation behavior Alloys. Intermetallic alloys have also been devel-
is similar to that of nonoxide ceramics, such as silicon oped in the titanium–aluminum system. In all cases,
carbide and silicon nitride, so they are discussed in the the intermetallic compounds form ordered crystal
chapter on High Temperature Corrosion of Cera- structures, some examples of which are shown in
mics and Refractory Materials. Thus, the focus of Figure 1. Plastic deformation by dislocation movement
this chapter is on the high temperature corrosion behav- in such ordered structures is more difficult than in
ior of aluminide intermetallic alloys. disordered metallic structures, so these compounds
have high strength and low ductility. Some of the char-
1.25.1.1 General Uses of Intermetallic acteristics of these widely used systems are described in
Alloys the following sections.
The low density of intermetallic aluminides is primar-
ily due to their high aluminum content, which also 1.25.1.2.1 Nickel aluminides
promotes the formation of an aluminum scale for good Alloys based on both NiAl and Ni3Al have been
high temperature corrosion resistance. Intermetallic developed. As shown in Figure 2, NiAl, which
compounds also have good high temperature strength forms the cubic B2 structure (Figure 1(a)), has a
and creep resistance, but the mechanical properties at much wider range of stoichiometry than Ni3Al,
low temperatures (including ambient room tempera- which forms the cubic L12 structure (Figure 1(c)).
ture) are inferior to those of metal alloys. In particular, Ni3Al-based alloys exhibit an anomalous tempera-
intermetallic compounds tend to have low ductility ture dependence of flow stress, in which the strength
increases with increasing temperature up to about

700 C and then decreases at higher temperatures.
Polycrystalline Ni3Al is difficult to fabricate because
a of its low deformability. Although their ductility can
be improved with boron, Ni3Al-based alloys have
relatively large densities and do not offer significant
advantages over nickel-based superalloys. NiAl-based
a
a alloys, however, have a higher aluminum content,
(a)
which decreases density and also improves oxidation
resistance. NiAl-based alloys have a higher melting
temperature and better creep resistance compared to
nickel-based superalloys. The strength can be
improved with the addition of niobium or tantalum,
c
which leads to the formation of Laves phases, and the
ductility can be improved with alloying additions
such as chromium, molybdenum, rhenium, and iron.
a However, the creep resistance of NiAl above 1000 C
(b) a is not good.
1.25.1.2.2 Iron aluminides

The iron–aluminum system is similar to the nickel–
a aluminum system, in that both Fe3Al and FeAl com-
pounds are formed. Like NiAl, FeAl forms the cubic
B2 structure (Figure 1(a)), but Fe3Al forms the D03
structure (Figure 1(d)). The tendency towards order-
a
a ing in the iron–aluminum system is not as strong as
(c) that in the nickel–aluminum system, as indicated by
the large disordered ferrite region shown in the phase
diagram in Figure 3. As in the nickel–aluminum
system, the strengths of Fe3Al-based alloys increase
with increasing temperature to about 500 C and
a
then decrease at higher temperatures. However,
FeAl offers more advantages over superalloys than
NiAl. In particular, the fracture toughness and spe-
a
cific properties of FeAl-based alloys are better than
(d) a those of NiAl-based alloys, but ductility is low com-
pared to metal alloys. The ductility can be improved
with alloying additions such as zirconium. Another
y-120°
difference between the iron–aluminum and nickel–
aluminum systems is that carbon is often used in the
iron–aluminum system to improve strength through
c
carbide strengthening.
structure type AuCu, space Group P4/mmm. (c) L12 (Ni3Al),

Pearson symbol cP4, structure type Cu3Al, space group
a Pm3m. (d) D03 (Fe3Al), Pearson symbol cF16, structure type
a BiF3, space group Fm3m. (e) D019 (Ti3Al) Pearson symbol
(e)
hP8, structure type Ni3Sn, space group P63/mmc.
Figure 1 Crystal structures of selected intermetallic Reproduced from Villars P., Calvert L.D. (Eds.), Pearson’s
compounds in the Ni–Al, Fe–Al, and Ti–Al systems. (a) B2 Handbook of Crystallographic Data for Intermetallic Phases,
(NiAl, FeAl), Pearson symbol cP2, structure type CsCl, 2nd ed., ASM International, Materials Park, OH, 1991, with
space group Pm3m. (b) L10 (TiAl), Pearson symbol tP4, permission from ASM International.
Weight percent nickel

0 10 20 30 40 50 60 70 80 90 100
1800
1638 ⬚C
1600 L
1455 ⬚C
1400
AlNi
1395 1385
(Ni)
Temperature (⬚C)
1200
1133 ⬚C
1000
854 ⬚C
660.452
800
~700 ⬚C
639.9 ⬚C
600
Al3Ni
Al3Ni2
(Al) Al3Ni5 AlNi3
400
0 10 20 30 40 50 60 70 80 90 100
Al Atomic percent nickel Ni
Figure 2 NiAl phase diagram. Reproduced from Binary Alloy Phase Diagrams, 2nd ed., ASM International, Materials
Park OH, 1996, with permission from ASM International.
1.25.1.2.3 Titanium aluminides high reactivity of titanium creates challenges in attain-

Titanium aluminide alloys have the lowest densities ing corrosion resistance, as will be discussed later, as
among the three common aluminide systems and well as in reliable alloy production.
thus offer the best specific properties. The phase
diagram in Figure 4 shows that Ti3Al and TiAl have
similar ranges of stoichiometry. TiAl forms the tetrag- 1.25.2 Oxidation
onal L10 structure (Figure 1(b)), while Ti3Al forms
the hexagonal D019 structure (Figure 1(e)). Both A major difference between titanium aluminides and
Ti3Al- and TiAl-based alloys have been developed, the other two common aluminide systems (nickel and
but the best ductility is achieved with two-phase iron) is that titanium is much more active than either
(Ti3Al + TiAl) alloys. Production of the desirable iron or nickel. This is illustrated in the Ellingham
two-phase microstructure requires careful control of diagram shown in Figure 5. Titanium can change
both the chemical composition and the heat treatment valence from Ti2+ to Ti4+, so there are several tita-
process. The aluminum content in Ti3Al is typically nium oxides, a few of which are included in Figure 5.
not sufficient for good oxidation resistance, so the The two important titanium oxides are TiO, which is
most promising alloys consist entirely or mostly of in equilibrium with titanium metal, and TiO2, which
the g-phase (TiAl in the L10 structure). The mechan- is stable in air. The free energy of TiO (per mole of
ical properties can be improved by adding substitu- oxygen) is about the same as that of Al2O3, indicating
tional elements such as chromium, as well as there is a strong competition between the formation of
interstitial elements such as carbon and boron. The titanium and aluminum oxides on TiAl. In comparison,
Weight percent aluminum

0 10 20 30 40 50 60 70 80 90 100
1600
1538 ⬚C
1400 L
1394 ⬚C
1310 ⬚C
(α Fe) 1232
1200 1169 ⬚C ~1160 ⬚C
ε
Temperature (⬚C)
(γ Fe)
1102 ⬚C
1000
912 ⬚C
FeAl
800
Fe2Al5
FeAl3
FeAl2
770 ⬚C
655 ⬚C
660.452 ⬚C
600
Fe3Al (Al)
400
0 10 20 30 40 50 60 70 80 90 100
Fe Atomic percent aluminum Al
Figure 3 FeAl phase diagram. Reproduced from Binary Alloy Phase Diagrams, 2nd ed., ASM International, Materials
Park OH, 1996, with permission from ASM International.
nickel and iron oxides are much less stable than Al2O3, in the scale formed during oxidation. However, nickel
so the amount of aluminum in the alloy required for and iron oxides can also exist in the scale, either
the formation of an Al2O3 scale is much less in nickel as pure oxides (e.g., NiO, Fe2O3) or mixed oxides
and iron aluminides than in titanium aluminides. (e.g., NiAl2O4 spinel). In addition, with long-time
Because of this significant difference in oxide stabili- exposure, the growth of an Al2O3 scale depletes the
ties, nickel and iron aluminides have, in many ways, alloy of aluminum, so the amount of aluminum
similar oxidation behavior and will be discussed required to maintain a protective scale is higher
together in the next section. This will be followed by than the minimum for thermodynamic stability or
a section on titanium aluminides, and then a section on initial alumina formation. Pettit10 characterized the
some less commonly used aluminide alloys. oxidation behavior of nickel–aluminum alloys and
found that Ni3Al alloys form a scale with alumina
and NiO or the NiAl2O4 spinel, whereas NiAl alloys
1.25.2.1 Nickel and Iron Aluminides
form a protective alumina scale. Once the critical
There are several general reviews of the oxidation of aluminum content for a protective alumina scale is
intermetallic alloys4–6 as well as reviews that focus on reached, additional aluminum has little effect on the
either the nickel–aluminum7,8 or iron–aluminum9 oxidation rate for short-term exposures. However,
system. with longer exposures, the aluminum content of the
alloy beneath the scale can decrease, which leads to
1.25.2.1.1 Scale properties/structure an increase in the oxidation rate. Changes in the
As a result of the large differences in oxide stability aluminum content can also affect the microstructure.
(Figure 5), Ni3Al, NiAl, Fe3Al, and FeAl are in For example, the decrease in the aluminum content
equilibrium with Al2O3, so Al2O3 is always present in FeAl can lead to a disordered structure (i.e., FeAl
Weight percent aluminum

0 10 20 30 40 50 60 70 80 90 100
1800
1700
1670 ⬚C
1600 L
1500
1400
(b Ti)
1300 ~1285
d
Temperature (⬚C)
TiAl
1200
~1125
1100
Superlattice structures
1000 Ti3Al
900
882 ⬚C TiAl3
800 (a Ti)
700 665 ⬚C
660.452 ⬚C
600
TiAl2
aTiAl3 (Al)
500
0 10 20 30 40 50 60 70 80 90 100
Ti Atomic percent aluminum Al
Figure 4 TiAl phase diagram. Reproduced from Binary Alloy Phase Diagrams, 2nd ed., ASM International, Materials Park
OH, 1996, with permission from ASM International.
to a-Fe in Figure 3) below the oxide scale. Such segregation of particular elements to the alloy–scale
changes can affect the oxidation behavior, as diffusion interface can lead to weakening of the interface.
rates in the alloy will determine the magnitude of any Porosity is often formed beneath the scale formed on
changes in alloy composition at the alloy–scale inter- nickel and iron aluminide alloys. The formation of poros-
face, which determines the oxide phases formed in ity is generally attributed to the coalescence of vacancies
the scale. formed during the growth of alumina. Figure 6 shows
In addition to alloy composition, microstructure the transport processes that occur during the growth of
also affects the oxidation behavior. For example, a an alumina scale on an aluminide alloy. Bulk alumina
finer alloy grain size can lead to reduced oxidation grows by the outward diffusion of aluminum through
rates, because the diffusion of aluminum to replenish the inward diffusion of aluminum vacancies, which are
the aluminum used to form the alumina scale can created at the oxide surface from oxygen in the gas
reach the scale faster by grain boundary diffusion. phase. The vacancies are filled by aluminum from the
A fine-grained microstructure has also been shown existing oxide, so that the new oxide is formed on the
to improve the adherence of an alumina scale on outer surface. The aluminum vacancies migrate to
NiAl.11 Scale adherence is particularly important dur- the alloy and are filled with aluminum from the alloy,
ing thermal cycling, as thermal stresses are generated which creates vacancies in the alloy. Once the concen-
in the scale due to differences between the coefficients tration of these vacancies exceeds the maximum con-
of thermal expansion of the scale and the alloy. How- centration in the alloy, pores will form.
ever, stresses can be generated in the scale even in The growth process at grain boundaries is gener-
isothermal oxidation due to growth or transforma- ally attributed to the inward migration of oxygen
tion stresses in the oxides. In addition, porosity or rather than to the outward migration of aluminum.
−100
1/3 Fe
2O3
NiO
−150
Free energy of formation (kJ mol–1) ¼Fe3O4
FeO
−200
−250
−300
1/
7Ti4O7
−350 1/ Ti O
½TiO2 5 3 5
1/
3Ti2O3
−400
−450 TiO
1/ Al O
3 2 3
−500
600 700 800 900 1000 1100 1200 1300
Temperature (°C)
Figure 5 Ellingham diagram for nickel, iron, titanium, and aluminum oxides. Data from Kubaschewski, O., Alcock, C.
B. Metallurgical Thermochemistry, 5th ed., Pergamon Press, 1979; Chase M.W., Jr. (Ed.), NIST-JANAF Thermochemical
Tables, 4th ed., J. Phys. Chem. Ref. Data, Monograph No. 9, 1998.
Grain boundary Bulk

O2
••
Al + 3V O } + 2 O2 = 2V⬘⬘⬘+ 3VxO + 3p•
3
{2V⬘⬘⬘ Al
1
O2 = O⬘⬘i + 2p•
2 New oxide
JO O⬘⬘i p• Al2O3 V⬘⬘⬘ p• JAl

Al
New oxide
Alalloy + V⬘⬘⬘+ • x
Al 3p = Al Al + Valloy
JAl
JAl NiAl JNi
JNi
2Alalloy + 3O⬘⬘i + 6p• + {2V⬘⬘⬘+

Al
3VO••} = 2AlxAl + 3VxO + 2Valloy
Figure 6 Transport during growth of alumina scale on nickel aluminide alloys. The predominant inward diffusing oxygen
species has not been determined, so oxygen interstitials are used as an example.
Although the specific oxygen defect by which oxygen the Kirkendall effect, as, for nickel-rich compositions,
is transported is not well established, for illustrative the diffusion constant for aluminum in NiAl is smal-
purposes, oxygen interstitials are shown in Figure 6. ler than that for nickel in NiAl.12 Thus, the diffusion
Regardless of the specific predominant oxygen defect, of nickel away from the scale into the alloy is faster
however, inward migration of oxygen will lead to the than the diffusion of aluminum from the bulk toward
forming of new oxide at the alloy–scale interface. the scale, which leads to a net movement of mass
Inward-growing scales generally contain less porosity away from the scale and the formation of vacancies
at the alloy–scale interface as compared to outward- in the metal beneath the scale. The concentration of
growing scales. However, in the Ni–Al system, poros- vacancies continues to increase with increasing oxi-
ity can still form within the inward-growing scale. dation and eventually leads to the formation of pores.
The formation of porosity is generally attributed to The Kirkendall effect can occur regardless of where
the new oxide forms. However, the Kirkendall effect humidity. The transient alumina phases eventually
does not occur for aluminum-rich compositions of transform to the more stable a-Al2O3 phase, which
NiAl, because, in this case, the diffusion of aluminum leads to a reduction in the oxidation rate, but volu-
is not slower than that of nickel. This mechanism is metric changes associated with the phase transforma-
supported by the observation that the number of tion can generate stresses that can affect the
voids formed decreases with decreasing grain size, morphology of the oxide scale.
as the larger amount of grain boundary diffusion in The stresses generated in the a-Al2O3 are tensile,
the finer grained material would reduce the amount but relax with time.24 Relaxation can occur by defor-
of vacancies formed.13 mation in either the alloy or the oxide, and thus
The most important effect of the formation of depends on the microstructure. Fine grains in either
porosity at the alloy–scale interface is that it weakens phase will enhance this deformation (i.e., creep of the
scale adhesion. The voids can act as stress concen- oxide or grain-boundary sliding in the alloy), and
trators and detachment of the oxide has been thus accelerate relaxation of the stresses. The stresses
observed at the pores.14 Pore formation does not, can also lead to convolution of the alloy–scale inter-
however, appear to significantly affect the oxidation face, which is commonly referred to as rumpling.25
rate. Although the loss of contact between the alloy Although the interface of the rumpled scale is irregu-
and the scale at pores or voids might be expected to lar, contact is maintained between the alloy and the
affect the scale growth, the AlOx vapor pressure is scale. The rumpling is associated with swelling of the
large enough for vapor-phase transport of aluminum scale, which is attributed to the different diffusion rates
across the void to occur. The number and size of of nickel and aluminum in NiAl (i.e., the same Kirken-
voids increase with oxidation time and the growth dall effect that leads to void formation beneath the
of voids in directions parallel to the surface is faster scale).26 This has led to the suggestion that rumpling
than that perpendicular to the surface.15 Void forma- is initiated by swelling of individual grains in the alloy.
tion has been attributed to the growth of the oxide The a-Al2O3 forming from the transient oxides
around perturbations at the alloy–scale interface.16 has been reported to nucleate at both the scale–gas
Corundum (a-Al2O3) is the most stable form of interface27 and at the alloy–scale interface.28 After
alumina and the scale formed on nickel- and iron nucleation, the transformation progresses in direc-
aluminide alloys is predominantly a-Al2O3. How- tions both perpendicular to the surface (inward or
ever, the first alumina phase to form is generally outward) and parallel to the alloy surface (lateral) and
one of the metastable alumina phases – usually y or can lead to the formation of ridges as illustrated in
g, but occasionally d. These transient phases form on Figure 7.6 The lateral growth of a-Al2O3 from the
nickel and iron aluminide alloys, but are especially nucleation site results in low-angle grain boundaries,
prevalent on NiAl. The crystallographic orientation which are oriented radially from the nucleation site.
of the y- or g-Al2O3 phases is typically aligned with The transformation of the transient alumina phase
that of the alloy, and the oxides can grow epitaxially, creates tensile stresses in the a-Al2O3 that can lead to
as observed by transmission electron microscopy17 the formation of radial cracks with associated accel-
and low energy electron diffraction.18 Epitaxial erated growth, which can result in ‘spokes’ along the
growth is maintained for a longer time on NiAl radial cracks (not shown in Figure 7). The outer
crystals with the (001) and (012) orientations as com- ridges form because the transient phases in this
pared to those with (011) and (111) orientations.19 region are the last to transform to a-Al2O3 and the
The transient phases grow by outward diffusion of higher growth rate of the transient oxides leads to a
aluminum, but at a higher growth rate (with lower thicker scale in the region. The ridges do eventually
activation energy) as compared with a-Al2O3.20,21 transform to a-Al2O3 and high-angle grain bound-
The transport of aluminum in y-Al2O3 is sufficiently aries are formed in the ridges at the boundaries
high for the overall growth rate of the scale to be between a-Al2O3 grown from adjacent nuclei. The
controlled in some cases by diffusion of aluminum high angle boundaries allow for more short-circuit
from the bulk of the alloy to the alloy–scale interface, migration of oxygen (as shown in Figure 6), so the
rather than the diffusion of aluminum through the growth rate of the ridges can continue to be higher
scale to the surface.22 As expected for meta-stable than that inside the ridge, where only low-angle
phases, the transient alumina phases are more preva- boundaries are present. The transformation stresses
lent at lower temperatures.23 They are also more are relaxed with time and the scale eventually con-
prevalent at low oxygen partial pressures and high sists of columnar a-Al2O3 grains.
High-angle
grain boundaries α-Al2O3 nucleation site
in center Low-angle
grain boundaries
Radial and inward growth of α-Al2O3 from transient Al2O3

phases creates tensile stress and leads to radial cracking
Ridges: transient (θ/Y) Al2O3

growth by outward cation diffusion
Figure 7 Ridge formation in alumina scale formed on nickel aluminide alloys.6
In iron aluminides, the nucleation of a-Al2O3 can which, as discussed earlier, leads to higher growth
be enhanced by Fe2O3, which can be present in the rates. Nickel dissolves more readily in the transient
oxide scale and act as a template or nucleation site for oxides, so the outward diffusion of nickel is more
the a-Al2O3. This transformation of the transient prevalent in low oxygen partial pressures.29 The dis-
oxides to a-Al2O3 enhances rumpling or wrinkling solution of nickel has been postulated to increase the
of the scale. The degree of wrinkling increases with ionic transport number, but the effect on the para-
increasing temperature and aluminum content and bolic rate constant is not significant.30 At very low
also depends on the crystallographic structure oxygen partial pressures, the formation of nickel
(ordered or disordered) and orientation. Another dif- oxide can be completely eliminated. Although this
ference between iron- and nickel aluminide alloys is would seem to be beneficial, since alumina, rather
that the Fe2O3 formed on iron aluminides is often than NiO, provides the good oxidation resistance, the
mixed with the alumina, whereas the NiO or spinel absence of NiO can lead to higher stresses in the
formed on nickel aluminides is usually located on the oxide scale.31
outer scale surface. The presence of water vapor has less effect on the
oxidation of alumina-forming alloys compared with
1.25.2.1.2 Effect of atmospheres/impurities its effect on chromia-forming alloys. Although water
In systems in which the active component provides vapor often does not significantly affect the oxidation
the oxide for good corrosion resistance, such as nickel rate of aluminides, it can lead to increased spallation
and iron aluminides, a reduced oxygen partial pres- and oxidation rates for NiAl.32,33 The increased spall-
sure is typically beneficial to the corrosion resistance, ation has been attributed to the hydration of the
because it promotes the formation of the protective alumina scale. Hydrate phases are not stable at oxi-
scale. If the oxygen pressure is sufficiently low, the dation temperatures, but they can form during cool-
oxide of the base metal could become unstable. How- ing and generate stresses in the scale, which promote
ever, even if the oxide of the base metal is stable, scale spallation.32 Although water vapor generally has
the dissolution of oxygen into the alloy is reduced little effect on the oxidation rate of iron aluminides,
at lower oxygen partial pressures, which reduces increased rates have been observed in some experi-
the formation of internal oxidation and promotes the ments.34 Such variable results suggest that water
formation of an external protective scale. Because the vapor may influence transient phase transformations,
low oxygen partial pressure can promote a protective which can be affected by minor changes in scale
scale, preoxidation in low oxygen partial pressures is microstructure or morphology.
often used to improve the oxidation resistance. These The most important impurity for the oxidation of
trends apply to the nickel– and iron–aluminum sys- alumina-forming oxides is sulfur, which can increase
tems as the amount of alumina in the scale is typically porosity at the alloy–scale interface and enhance
larger for oxidation in lower oxygen partial pressures, crack propagation.35,36 Sulfur diffuses to the scale and
but there are exceptions. For example, the lower initially adsorbs to the surface of pores. Once a contin-
oxygen partial pressure stabilizes the transient oxides, uous alumina scale is formed, the sulfur segregation
expands to the alloy–scale interface. As increased observed at temperatures lower than those at which
porosity and sulfur segregation weaken the scale the general reactive element effects are observed, and
adherence, sulfur is detrimental to oxidation resis- is attributed to the formation of hafnium oxides that
tance. The diffusion rate of sulfur in NiAl decreases protrude into the scale and provide anchors, which
with increasing aluminum content; therefore the improve adhesion.44 Hafnium also improves the plas-
amount of segregation, and thus the associated detri- ticity of the scale, which increases the magnitude of
mental effects, are greater for lower aluminum con- stresses that can be accommodated without spall-
tents. In iron aluminides, sulfur initially adsorbs to the ation. The addition of hafnium can significantly
pore surface and subsequently cosegregates with alu- extend the lifetime of aluminide alloys, particularly
minum. The presence of sulfur increases the scale when exposed to cyclic oxidation.45
convolution, which, as discussed earlier, results from Zirconium is also a commonly added alloy, and,
growth and transformation stresses. like yttrium, can lead to the formation of an addi-
tional intermetallic compound, NiZr5, which can
1.25.2.1.3 Effect of alloying additions lead to the formation of ZrO2 in the scale; therefore,
The oxidation resistance of aluminide alloys can be the amount of zirconium added must be limited.
improved with alloying additions. Reactive elements Small amounts of zirconium can improve adherence
have been shown to be beneficial to the oxidation of and reduce oxide scale growth rate by increasing
aluminide alloys by reducing the amount of porosity grain size. However, zirconium can also be detrimen-
that forms at the alloy–scale interface.37,38 One of the tal by impeding the formation of a-Al2O3 and pro-
mechanisms proposed for this beneficial effect of reac- moting outward scale growth, which increases growth
tive elements is that the reactive elements segregate to rates. Zirconium is a strong carbide former, so if
the grain boundaries in the oxide scale and form oxides zirconium is added with carbon, which is commonly
that block the outward diffusion of cations.36,38,39 added to iron aluminides, the ratio of zirconium to
Decreasing the outward diffusion of cations leads to carbon must be controlled, so that the appropriate
an increase in the proportion of growth by inward amount of zirconium is available to beneficially affect
diffusion of oxygen, which, as discussed earlier, the oxidation behavior.
reduces the amount of porosity formed at the alloy– Other reactive elements, such as titanium and
scale interface and thus improves scale adherence. cerium, have also been shown to decrease the oxi-
This mechanism appears to be more prevalent in dation rate, either by decreasing the outward diffu-
nickel aluminides compared to iron aluminides. sion of aluminum or by increasing the grain size of
The reactive element most commonly added to the oxide. However, not all reactive elements are
aluminide alloys is yttrium, which can be added as a beneficial to oxidation behavior; e.g., lanthanum
metallic alloy or as an oxide in oxide dispersion leads to both a higher oxidation rate and lower scale
strengthened (ODS) alloys. In nickel aluminide adherence.
alloys, the addition of yttrium can lead to the forma- Transition metals are often used as alloying additions
tion of a NiYx phase, which promotes spinel forma- for aluminide alloys. Chromium is commonly added to
tion and thus is detrimental to oxidation behavior. intermetallic alloys to improve ductility and minimize
The formation of NiYx is more likely when yttrium environmentally assisted crack embrittlement. Chro-
is added as a metallic alloy than when added as an mium is beneficial to the alumina-forming nickel-
oxide phase.40 As too large an addition of yttrium can and iron-based alloys, because it reduces the amount
increase stresses in the oxide,41 yttrium is most effec- of aluminum required to avoid internal oxidation and
tive when added in small amounts as an oxide phase. At form a protective scale. Chromium has a similar effect
low temperatures, yttrium additions promote a-Al2O3 in Ni3Al as it reduces the amount of aluminum
formation and thus decrease the oxidation rate. How- required to maintain an effectively rehealing scale
ever, at high temperatures, where a-Al2O3 forms easily for long oxidation exposures. Chromium affects the
with or without yttrium additions, yttrium is detrimen- transformation of y-Al2O3 to a-Al2O3, although the
tal by decreasing the grain size of the oxide scale, which reported effects are mixed as chromium has been
increases the amount of short-circuit grain-boundary reported to both inhibit42 and promote21 a-Al2O3
diffusion and thus increases the oxidation rate.42 formation. Chromium dissolves in y-Al2O3, which
Another common reactive element used with alu- would explain the inhibitive effect, as the chromium
minide alloys is hafnium, which improves the adhe- diffusion may be required for a-Al2O3 formation.46
sion of the oxide scale.43 The improvement is At the same time, chromium appears to accelerate
nucleation of a-Al2O3. The effects of chromium on interface, and improves the long-term oxidation
the oxidation rate are likewise mixed as it has been resistance of the alloy.
reported to both increase and decrease the oxidation As mentioned earlier, carbon is used to strengthen
rate. Chromium can segregate to the alloy–scale iron aluminide alloys. Carbon is generally detrimen-
interface and lead to additional phase formation, tal to oxidation resistance. The oxide scale formed on
which decreases adherence and is therefore detrimen- the carbide phase (Fe3AlCx) that strengthens the
tal to oxidation resistance. alloy is thicker than, but has the same morphology
The effects of niobium and tantalum are more as, the scale formed on the metal phase. As mentioned
consistent than those of chromium. Like chromium, earlier, when used with strong carbide formers (e.g.,
niobium and tantalum are added to improve the zirconium), the effects of carbide formation must be
mechanical properties. In both cases, a mixed oxide, considered in optimizing the alloy composition.45
NiTa2O6 or NiNb2O6, forms and is detrimental to Boron, another interstitial additional alloy, is also
oxidation resistance. However, if added in small generally detrimental to oxidation behavior, but can
amounts, niobium and tantalum improve the mechan- be beneficial when added with others, such as zirco-
ical properties with an acceptably minimal effect on nium or hafnium.
oxidation resistance. Although increasing the nio-
bium content is detrimental to the oxidation resis-
1.25.2.2 Titanium Aluminides
tance, very large niobium contents (i.e., NiAl + NbAl3
alloys) can lead to the formation of Laves phases, The high reactivity of titanium, relative to iron and
which promote the formation of a protective alumina nickel, creates additional challenges for the formation
scale during oxidation. and maintenance of an alumina scale, because of the
Refractory metals, such as molybdenum, tungsten, competition between aluminum and titanium for oxide
and vanadium, are generally detrimental because of formation. However, as mentioned earlier, titanium
the formation of liquid or gaseous oxides. For exam- aluminides have excellent specific properties and so
ple, although molybdenum does not significantly have been developed for aerospace application. There
affect short-term oxidation behavior, for longer expo- are several reviews of their oxidation behavior.48–51
sures, the formation of molybdenum oxide vapor
phases is detrimental. 1.25.2.2.1 Scale properties/structure
Noble metals, such as Pt, Re, and Rh, can be The minimum aluminum content required for alu-
beneficial to the adherence of the oxide scale.6 For mina formation on titanium aluminide alloys is much
example, the beneficial effect of platinum to the higher than that for alumina formation on nickel or
oxidation behavior of NiAl is generally attributed to iron aluminide alloys. TiAl is in equilibrium with
its interaction with sulfur.47 Platinum, like sulfur, TiO and Al2O3, but any lower compositions are not
segregates to the pore surfaces and alloy–scale inter- in equilibrium with Al2O3; so, even before any alu-
faces. Any platinum present at these locations dis- minum depletion occurs, TiAl is at the borderline
places sulfur, and thus decreases the detrimental aluminum composition for Al2O3 formation. Thus,
effects associated with sulfur segregation. Platinum scales formed on titanium aluminides generally con-
can also be detrimental by stabilizing y-Al2O3, which tain both aluminum and titanium oxides.
inhibits the formation of a-Al2O3, thus increasing The typical scale morphology of the scale formed
the oxidation rate.23 However, because of the cata- on TiAl is shown schematically in Figure 8.
strophic effects of scale spallation, the beneficial
effects of improved adherence can often outweigh Gas
the detrimental effects of increased growth rates.
Scale spallation is particularly critical when the alu- TiO2 outer layer
minide is used as a bond for thermal barrier coatings,
which will be discussed in a later section. Large
amounts of platinum additions can lead to the forma- Al2O3 + TiO2
tion of a platinum aluminide phase (PtAl2), which
N-rich inner layer
minimizes depletion of aluminum from the alloy–
scale interface. Maintenance of a sufficient aluminum Alloy
level at the alloy–scale interface prevents the forma- Figure 8 Schematic of scale formed during oxidation of
tion of Ni3Al, which can lead to local stresses at the titanium aluminide alloys.
Although TiO is the titanium oxide in equilibrium inferior to those of nickel and iron aluminides, so
with the alloy, TiO2 is stable in air. While there may titanium aluminide alloys are usually considered for
be some titanium oxides with valence lower than +4 lower temperature applications.
at the alloy–scale interface, the major portion of the
titanium oxide is rutile TiO2. Rutile is an oxygen- 1.25.2.2.2 Effect of atmosphere
deficient oxide, for which the predominant ionic and pretreatment
point defects are oxygen vacancies at high oxygen The presence of a nitrogen-containing layer suggests
partial pressures and titanium interstitials at low oxy- that nitrogen may influence oxidation behavior.
gen partial pressures. Platinum marker experiments Nitrogen is often considered essentially inert for
on the oxidation of pure titanium have shown that the many oxidation processes. However, titanium, in par-
new oxide forms in the middle of the oxide, indicat- ticular, and aluminum to a lesser extent, are nitride
ing that the oxygen vacancies are predominant in the formers, so nitrogen does affect oxidation behavior.
outer scale and titanium interstitials are predominant The oxidation behavior of titanium aluminide alloys
in the inner scale. The oxidation of titanium alumi- in air is different from that in argon–oxygen mix-
nide is further complicated, as the scale also contains tures.49,50 Although the results are somewhat varied,
aluminum oxides. As shown in Figure 8, the alumi- the oxygen rates in air are typically higher than those
num oxide is often dispersed in the rutile in the inner in oxygen without nitrogen for the same oxygen
portion of the scale. The outer portion of the scale is partial pressure. The detrimental effect of nitrogen
usually TiO2. Sometimes, as shown in Figure 8, there is attributed to the nitrogen-containing layer promot-
is a layer in the middle of the scale where the amount ing the formation of TiO2. Although the nitride layer
of alumina is higher than that in the inner portion of contains aluminum and titanium, the amount of tita-
the scale. These layers could be in part due to the nium is typically higher than that of aluminum.
oxygen partial pressure dependence of the ionic Nitrogen has also been shown to be beneficial to the
point defects in rutile as described earlier. Although oxidation behavior of some alloys – particularly alloys
the defect chemistry has not been well established, with lower aluminum content. The beneficial effect of
platinum marker experiments have shown that, as in nitrogen is attributed to nitrogen doping the rutile and
the case of pure titanium, the new oxide forms in the decreasing the concentration of the mobile defect,
middle of the scale. There is typically an inner layer, which decreases transport through the scale. Nitrogen
which contains nitrogen if the oxidation is performed may also promote the formation of alumina in the
in air. This inner layer may also contain titanium in scale by reducing the titanium activity in the alloy.
a lower valence and/or mixed oxides. The mixed Because of the delicate balance between alumina
titanium–aluminum oxide phases (X-phase, Z-phase, and titania formation, small changes can determine
NCP) are not thermodynamically stable at the oxi- which phase nucleates and grows; subtle differences
dation temperature, but can form at the interface and can therefore affect oxidation behavior. For example,
affect the oxidation behavior. surface finish has been shown to significantly affect
The growth rate of rutile is much higher than that the oxidation rate in oxygen, but not in air.52 The
of alumina, so an alumina scale is desired for oxida- surface finish most likely affects nucleation, which
tion resistance. A continuous alumina layer would implies that the superior performance in oxygen is
provide the lowest oxidation rate and has been due to the formation of a phase that is inhibited by
achieved on some titanium aluminide alloys. How- either the presence of nitrogen or the surface rough-
ever, a mixed scale, as shown in Figure 8, is more ness. Such phenomena are difficult to characterize
typical and the growth rate of such a scale is between because they likely depend on transient phases and
the rates for alumina and titania growth. Although microstructures, which are not present when the
increasing the amount of alumina is beneficial for sample characterization is performed after the oxida-
oxidation resistance, the morphology is also important. tion experiment is completed.
For example, the more continuous the aluminum-rich The alloy microstructure also affects oxidation
layer in the middle of the scale (Figure 8), the slower behavior. For the same alloy composition, the oxida-
is the growth rate. Thus, control of the oxidation rate tion rate with a lamellar microstructure is lower than
of titanium aluminum alloys depends critically on the that for a duplex microstructure.53 One of the chal-
scale morphology. Because of the mixed scale that lenges in the application of titanium aluminide alloys
grows at a rate higher than pure alumina, the oxidation is control of the microstructure, as the microstructure
resistances of titanium aluminide alloys are generally also affects the mechanical properties. Production of
parts with the same microstructure throughout the Another common alloying addition for titanium alu-
sample, as well as between samples, is a significant minide alloys is chromium, which is added primarily to
engineering challenge. Microstructural inhomogene- improve the mechanical properties of the alloy. Small
ities can lead to locally high corrosion rates or local amounts of chromium lead to an increase in the oxida-
mechanical weaknesses. tion rate. This increase is typically attributed to the
Water vapor in the atmosphere can increase the chromium doping the rutile scale and increasing the
oxidation rate. This increase is generally attributed to concentration of oxygen vacancies and/or titanium
increasing water vapor transport in the scale. interstitials. Larger amounts of chromium, however,
Although water vapor does not significantly affect are beneficial to oxidation resistance. The beneficial
the growth of alumina, it does affect transport in effect has been attributed to chromium increasing
rutile, which is generally the continuous phase in the aluminum activity in the alloy. While increases in
the oxidation scale and thus dominates the overall the aluminum activity have been measured, the mag-
scale growth rate. nitude of the increase is not sufficient to explain the
Preoxidation and presulfidation have been used to observed improvements. Another explanation is related
improve the oxidation resistance of titanium alumi- to the formation of Laves phases, which form in alloys
nide alloys. There are often conditions in which a with high chromium content. The Laves phases have
protective scale would not form, but could be main- low oxygen solubility and lower titanium activity, both
tained if it had already formed. Under such conditions, of which promote a more protective alumina scale.
a pretreatment in carefully controlled conditions can Niobium and chromium are sometimes added
be used to form a protective scale, which subsequently together and, fortunately, the oxidation behavior
grows at a slow rate. appears to be dominated by niobium rather than by
chromium. The oxidation rates of alloys with a
1.25.2.2.3 Effect of alloying additions combination of niobium and other alloying additions
The aluminum content in titanium aluminide alloys is such as manganese, zirconia, and hafnium are also
more critical than in nickel and iron aluminide alloys. similar to those of alloys with niobium as the only
The oxidation resistance of Ti3Al is generally poor, and alloying addition.
in binary titanium–aluminum alloys, at least 48–50 at% Like chromium, other transition metals, such as iron,
aluminum is generally required for good oxidation nickel, manganese, and vanadium, are generally detri-
resistance. The amount of aluminum required, how- mental to the oxidation of titanium aluminide alloys,
ever, can be reduced with alloying additions. while refractory metals, such as tungsten and molybde-
The alloying addition most widely used for improv- num, are beneficial. The beneficial effects of tungsten,
ing the oxidation resistance of titanium aluminide like those of niobium, are attributed to decreasing the
alloys is niobium. During oxidation, the alloy beneath oxygen solubility in the alloy (to inhibit internal oxida-
the scale is enriched with niobium, but niobium is also tion) and doping of the rutile (to reduce oxygen vacan-
incorporated in the scale. Several explanations for the cies and/or titanium interstitial concentrations).
beneficial effects of niobium additions have been pro- Silicon additions can be beneficial to the oxidation
posed. One explanation is that niobium increases the resistance of titanium aluminide alloys. If silicon is
activity of aluminum in the alloy, but this has not been present in the alloy, silica will usually form in the
supported by subsequent thermodynamic measure- scale. However, silica is often dispersed so that is it is
ments of the aluminum activity.49 Another explanation not effective for reducing the oxidation rate. In some
is that niobium increases the diffusion of aluminum cases, silicon additions lead to a continuous silica
and/or decreases the diffusion of oxygen in the alloy, layer, which provides a barrier to oxygen diffusion
which reduces the amount of internal oxidation of and thus improves the oxidation resistance. However,
aluminum and promotes a more continuous alumina this layer leads to stresses during thermal cycling and
scale. The beneficial effect of niobium has also been is thus detrimental to the cyclic oxidation resistance.
attributed to niobium doping the rutile scale and Noble metals have mixed effects on the oxidation
decreasing the concentration of oxygen vacancies of titanium aluminide alloys. Platinum and gold are
and/or titanium interstitials. Niobium has also been detrimental to oxidation resistance, but silver is ben-
reported to improve the mechanical integrity of the eficial.54 Silver promotes the formation of y-Al2O3,
scale, which improves the resistance to cyclic oxidation. which, as discussed earlier, grows at a faster rate than
Large amounts of niobium, however, can be detrimen- a-Al2O3 and thus would seem to be detrimental.
tal if a Nb2O5 phase forms in the oxide scale. However, the a-Al2O3 formed from transient alumina
is continuous and the oxidation rate of the alloy is the Gas

same as that for the growth of an alumina scale. The 4AlCl(g) + 3O2(g) = 2Al2O3(s) + 2Cl2(g)
beneficial effect of silver decreases with increasing
temperature, presumably because it results from the
transient formation of y-Al2O3, which is metastable
and thus becomes less prevalent with increasing tem-
perature. The addition of small amounts (1–3%) of Scale
chromium to silver-doped alloys is detrimental to Cl2(g) AlCl(g)
oxidation resistance, but with larger amounts (5–7%)
the oxidation rate is reduced to that of an alumina-
forming alloy.55 The excellent oxidation resistance of
2Al(alloy) + Cl2(g) = 2AlCl(g)
these alloys, like the high-chromium-content alloys
Alloy
described earlier, is attributed to the presence of
Laves phases in the alloy microstructure. Figure 9 Vapor transport mechanism for beneficial effect
Reactive elements are used as alloying additions in of halogen additions on the oxidation of titanium aluminide
alloys.58,59
titanium aluminide alloys, but not to the extent, or to
the same beneficial effect, as in nickel and iron alu-
minide. Small amounts of reactive elements, such as range: it must be high enough to transport sufficient
zirconium, hafnium, and yttrium, are beneficial to aluminum to maintain the protective scale, but if it is
oxidation resistance. Hafnium has been shown to too high, the partial pressure of a titanium chloride
improve scale adherence. Zirconium, while beneficial species may exceed that of the aluminum-containing
for short oxidation exposures, eventually leads to the species. When the pressure of the titanium species is
formation of zirconia, the transformation of which too high the alumina forms in a whisker-like morphol-
creates stresses in the oxide scale. ogy, which does not provide adequate protection.
One of the most successful recent approaches for To impart its beneficial effect on oxidation resis-
improving the oxidation resistance of TiAl-based tance, the halogen element is needed only at the
alloys is the addition of halogen elements, such as surface, so ion implantation is often used to add halo-
fluorine, chlorine, or bromine. The addition of the gen elements to titanium aluminide alloys. The same
appropriate amount of a halogen element can lead to is true for other alloying elements added for improved
a continuous alumina scale and kinetics following the corrosion resistance, especially those which may have
growth of an alumina scale. Various explanations detrimental properties on the mechanical properties of
have been proposed for the beneficial effect of halo- the alloy. Thus, ion implantation is used to alloy tita-
gen additions. One explanation is that chlorine nium aluminide alloys with other metallic elements, as
doping of the scale decreases the oxygen vacancy well as nonmetallic elements, such as phosphorous.60
and/or titanium interstitial concentration (i.e., the This technique minimizes the amount of alloying ele-
same as that of niobium and tungsten doping).56 ments used by localizing the addition to the location at
Other explanations are that the formation of TiO2 which the element is needed. This technique would not
is inhibited by the presence of the halogen element57 be effective for an element that is consumed during
or by a reduction in titanium activity due to volatili- oxidation, as there is a limited source, which could
zation of titanium. The explanation that best illus- become depleted after long exposures. However, in the
trates the long-term effect is shown in Figure 9.58,59 case of halogen additions, the halogen element appears
In this model, aluminum is transported from the alloy to be recycled, and ion implantation has been shown to
to the scale–gas interface in a gas phase that diffuses improve oxidation resistance for long periods of time.
through pores or microcracks in the scale. Once the
aluminum-containing vapor reaches the scale surface,
1.25.2.3 Other Aluminides
it reacts with oxygen to form alumina and evolves
chlorine, which can diffuse back to the alloy and Most of the aluminide alloys used are from the nickel–,
form more of the aluminum-containing species. The iron–, or titanium–aluminum system. However, there
chlorine that evolves is thus recycled and provides a are a few other systems that have been developed,
mechanism for the transport of aluminum to heal including aluminides of noble and active metals.
pores and/or cracks in the scale. For this mechanism As described earlier, the beneficial effect of plati-
to work, chlorine partial pressure must be in a critical num additions to NiAl is due, in part, to the formation
of a platinum aluminide phase at the alloy–scale inter- corrosion of intermetallic alloys. The corrosion can
face. Since platinum does not form a stable solid oxide, be enhanced if the additional species is in the gas
the only oxide that can form on a platinum aluminide phase, but corrosion is further accelerated if a
alloy is alumina. The microstructure of the alumina condensed molten salt phase is formed.
scale is similar to that formed on NiAl, but the adhesion
is better on platinum aluminide alloys.61 The scale
1.25.3.1 Sulfur-Containing Environments
growth occurs predominately through short-circuit
grain boundary diffusion, so the grain size of the The sulfidation behavior of intermetallic alloys has
oxide scale affects the oxidation rate. The adhesion recently been reviewed.67 The effects of sulfur on the
can be further improved with zirconia additions, corrosion rate depend on the oxygen partial pressure
which segregate to the alloy–scale interface and lead and whether a condensed phase is formed.
to a convoluted, but strong, interface between the alloy
and the scale.62 A platinum aluminide coating is clearly 1.25.3.1.1 Gaseous
expensive and thus is used in applications where a small The addition of SO2 to an oxidizing environment
amount of material is required, such as in a bond coat, does not significantly affect corrosion behavior.
which will be discussed in a later section. While the scale formed in SO2 contains oxides and
Other noble metal aluminides include those of sulfides, scales formed in oxygen with SO2 contain
ruthenium and iridium. Ruthenium aluminide has mostly oxides with little or no sulfides.68 One notable
better ductility than most aluminide alloys and has difference between oxidation with and without sulfur
good oxidation resistance. The scale morphology is that silver, which as discussed earlier, is beneficial
suggests that the scale grows by outward diffusion of to the oxidation behavior of TiAl, is detrimental to the
aluminum, although voids are not formed beneath corrosion of TiAl in sulfur-containing atmospheres.
the scale.63 Two-phase Ru(Al) + RuAl are particu- The corrosion behavior changes as the oxygen par-
larly attractive because of their good ductility. How- tial pressure is reduced. In H2–H2S–H2O mixtures,
ever, the oxidation resistance decreases as the amount the oxygen partial pressure is typically sufficiently
of the metal phase increases, because the high oxygen high for alumina to form and alumina is present in
diffusivity of the metallic phase leads to internal the outer portion of the scale. However, the inner
oxidation.64 The oxidation resistance can be improved portion of the scale consists of sulfides, which can be
with platinum additions, but weight loss still occurs mixed, but usually contain little or no aluminum (e.g.,
during cyclic oxidation.65 Iridium aluminide is used TiS for TiAl69 or FeS for FeAl70). As expected from the
because iridium has very low oxygen permeability low aluminum content in the sulfides, increasing the
and thus is an excellent oxygen barrier. The oxidation aluminum content in the alloy increases the amount of
rates of iridium–aluminum alloys are representative alumina in the scale. For the corrosion of TiAl, TiO2 is
of pure alumina scale growth.66 also formed in the scale. In addition, the rapid transport
An aluminide alloy system that has been investi- of titanium in the inner TiS layer leads to depletion of
gated because of its high melting point and low titanium in the alloy, so phases with higher aluminum
density is niobium aluminide.6 However, niobium content, such as TiAl2 and TiAl3, form beneath the
aluminides, even with large aluminum concentra- scale. The addition of niobium improves the sulfidation
tions (e.g., NbAl3), form mixed oxide scales rather resistance of both FeAl and TiAl. Niobium oxides and
than alumina scales. In particular, the scales contain sulfides form in the scale and the scale adherence is
AlNbO4 and Nb2O5 and are thus not protective. improved with niobium additions.71
Zirconium aluminide has been investigated but has With further decrease in the oxygen partial pres-
poor oxidation resistance because of the formation of sure in H2–H2S environments, the oxides become
ZrO2. ZrO2 not only has high oxygen permeability, less stable and the scale consists almost entirely of
but also undergoes phase transitions, which generate sulfides. For FeAl alloys, FeS and mixed (Fe,Al)S
stresses in the oxide scale. phases are formed in sulfidizing environments. The
predominant phase formed during sulfidation of
NiAl is Ni3S2, which can also contain Al2S3 or
1.25.3 Hot Corrosion NiAl2S4 inclusions. For low H2S levels the sulfidation
rate is reasonably low, but as the H2S content
The presence of other reactive species, such as sulfur increases a Ni–Ni3S2 eutectic forms and leads to
and chlorine, in the environment can accelerate the large (4 orders of magnitude) increases in corrosion
rates. The addition of chromium leads to the forma- (Na,K)2SO4

tion of (Cr,Al)3S4 and further degrades the sulfidation
resistance.72 Similarly, zirconium additions increase TiO2 outer layer
the amount of internal corrosion and increase the
sulfidation rate. In general, nickel aluminides have
poor sulfidation resistance. Al2O3 + TiO2
TiAl forms a layered sulfide structure in low
oxygen-partial-pressure sulfidizing atmospheres.73
The scale typically contains Ti3S4, TiS, and Al2S3. TiS
The sulfide scale typically contains more titanium TiAl3
than aluminum, so the alloy beneath the scale is
TiAl
depleted in titanium, and aluminum-rich phases,
such as TiAl2 and TiAl3, form. Alloying addition Figure 10 Schematic of scale formed during corrosion of
Ti–50Al in (Na,K)2SO4 at 900 C.76
affect the specific sulfide phases formed, but in most
cases do not significantly affect the overall sulfidation
rate. Silver and copper additions, however, increase
the amount of sulfidation. shown schematically in Figure 10.76 As with sulfida-
tion in a gaseous environment, the scale contains
1.25.3.1.2 Molten salts more titanium than aluminum, so the alloy beneath
The corrosion behavior changes when sulfur is pres- the scale is depleted in titanium and a layer of TiAl3
ent in a condensed salt on the alloy surface. An oxide is formed. Also, as with other sulfidation and oxida-
scale that is relatively stable in a gaseous environment tion processes, increasing the amount of aluminum in
may dissolve in the salt and lead to increased corro- the alloy increases the amount of alumina in the scale
sion rates. and improves corrosion resistance. Niobium is bene-
The reaction between FeAl and Na2SO4 produces ficial to the oxidation resistance as it improves scale
a complex mixture of oxygen- and sulfur-containing adherence.76 Chromium aids the hot corrosion resis-
compounds (e.g., NaAlO2, NaAl(SO4)2, Fe2O3, Al2O3). tance, and, as will be discussed later, is often added to
The corrosion of the alloy occurs by dissolution of protect aluminide coatings. Vanadium additions,
the oxide in the molten salt and internal sulfidation however, increase the amount of scale spallation and
of the alloy. Corrosion resistance can improve with are thus detrimental.
increasing temperature as the more rapid growth of
the alumina results in a more protective scale.74 Simi-
1.25.3.2 Chlorine-Containing Environments
larly, the addition of dispersed Al2O3 to the alloy
improves the corrosion resistance by stabilizing the Chlorine enhances the corrosion rate of most materi-
Al2O3 in the scale. Carbon additions have been shown als and intermetallic alloys are no exception. Chlo-
to improve the corrosion resistance of FeAl by form- rine can be present in the gas phase or in a molten
ing carbides that inhibit sulfur diffusion into the alloy. salt – either a chloride or mixed salt.
Boron additions, however, have mixed effects as they
have been shown to be both beneficial and detrimen- 1.25.3.2.1 Gaseous
tal to hot corrosion resistance. The primary effect of chlorine in the gas-phase cor-
Aluminization of nickel-based superalloys has been rosion of aluminide alloys is the formation of volatile
used to improve hot corrosion resistance. The oxide chloride species.77 In particular, the vapor pressure
scale formed on nickel aluminide alloys during hot of AlCl3 is relatively high and can lead to loss of
corrosion in molten sulfates contains an outer layer of aluminum and/or void formation in the scale, both
NiO or NiAl2O4. The latter has better hot corrosion of which are detrimental, as loss of aluminum
resistance because of the lower solubility of the spinel, decreases the amount of alumina in the scale, and
as compared to NiO, in the molten salt.75 The resis- void formation degrades scale adherence. The addi-
tance to molten sulfate corrosion can be improved with tion of zirconium has been shown to reduce the loss of
the addition of platinum, palladium, or chromium. aluminum and thus improve corrosion resistance. In
The oxide formed during the corrosion of tita- some cases, the liquid chlorides can form from reac-
nium aluminide alloys in molten sulfates contains tion with the chlorine in the gas and can enhance the
TiO2 and Al2O3 with an inner layer of sulfides as corrosion rate.
1.25.3.2.2 Molten salts Na2SO4 + NaCl

Molten chlorides can also be formed from salts
deposited on the alloy surface. For example, the
reaction of titanium aluminides with sodium chloride Al2O3 + TiO2
results in the formation of TiO2, Al2O3, and NaTiO2.
The increase in corrosion rate due to the presence of
sodium chloride is generally attributed to vapor Z-phase
phase transport via chloride gas species.78 As dis- Ag
cussed earlier, controlled amounts of chlorine, or TiAl(Ag)
other halogens, can improve the oxidation of titanium
Figure 11 Schematic of scale formed during corrosion of
aluminide alloys, but for this beneficial effect to
Ti–48Al–(1–3)Ag in Na2SO4 + NaCl at 800 C.80
occur, the chlorine content must be limited so that
transport of aluminum, but not titanium, occurs. The
chlorine levels present during hot corrosion in NaCl aluminum and the scale is generally not protective.
are high enough for the transport of both titanium Silicon additions increase the amount of alumina, and
and aluminum from the alloy to the scale surface. thus improve the oxidation resistance. Oxide (Y2O3)
Under these conditions, the alumina forms as whis- dispersions can also improve the corrosion resistance
kers, rather than as a protective layer, so the corrosion by improving scale adherence.81
rate is increased.
Hot corrosion in lithium-containing chloride salts,
1.25.3.3 Others
such as LiCl–Li2O or KCl–Li2O, results in the forma-
tion of multiple mixed oxides. For TiAl, in addition to Molten carbonates can also lead to enhanced corro-
TiO2 and Al2O3, LiAlO2 forms in the outer portion of sion of aluminide alloys. Corrosion in lithium-
the scale, while Li2TiO3 forms in the inner portion containing carbonates is similar to that in other
of the scale. LiAlO2 also forms during corrosion of lithium-containing salts in that LiAlO2 is present in
NiAl and FeAl in lithium-containing salts, but NiAl the corrosion scale. Because of their high aluminum
is more resistant than FeAl to chloride attack.79 content, NiAl and FeAl, are used as replacements or
The corrosion of titanium aluminide alloys in coatings on stainless steels and superalloys in molten
mixed sulfide–chloride molten salts is similar to carbonate environments.
that in chloride salts, although in some cases the As mentioned earlier, vanadium is detrimental to
corrosion in a molten chloride is faster than that in the corrosion of TiAl in sulfate salts. Vanadium is also
a mixed salt. The chloride vapor transport mecha- detrimental when present in a corrosive salt. Vanadium-
nism described above is active in the corrosion of containing salts are more corrosive than sodium
TiAl in a mixed chloride–sulfide salt. Chromium chloride to iron aluminides. The scale formed in
and silver additions to TiAl improve the hot corro- V2O5-containing salts consists of several mixed oxi-
sion resistance by improving the adherence of the des and can be quite complex. For example, the
mixed oxide phase (Z-phase) formed at the alloy– weight loss of FeAl in a V2O5–Na2SO4 mixture
scale interface.76,80 The morphology of the scale increases with increasing temperature up to about
formed on titanium aluminide containing 1–3% sil- 700 C and then decreases at higher temperatures
ver is shown in Figure 11.80 Niobium, which, as due to the formation of a more protective scale. The
discussed earlier, is generally beneficial to the corro- addition of Al2O3 to the alloy provides some
sion of TiAl, can be detrimental when added in large improvement in scale adherence.82
amounts because of sulfide phase formation.
The corrosion of nickel aluminide alloys in mixed
sulfide–chloride molten salts is similar to that in 1.25.4 Coatings
sulfide melts, presumably due to the absence of sig-
nificant partial pressures of nickel chloride gas spe- Aluminide alloys are attractive as oxidation resistant
cies to provide vapor phase transport analogous to coatings, because ductility, while desirable, is less
that which occurs in titanium aluminide alloys. The critical in a coating than it is in a structural load-
oxide scale formed on nickel aluminide alloys con- bearing member. Aluminide coatings can be used as
sists of a layer of NiO and Al2O3 above an inner layer coatings for other aluminide alloys with inferior oxi-
of sulfides. The alloy beneath the scale is depleted in dation resistance, but superior mechanical properties.
In addition, aluminide coatings can be applied to and complicated alloy systems will be discussed. While
metallic alloys that have high ductility, but contain alloying additions from the substrate alloy can compli-
little or no aluminum and thus may have poor oxida- cate the coating process, the coating process can also
tion resistance. Coatings can be applied by a variety be used to introduce desired alloying additions for
of techniques, some of which will be discussed in improved coating performance.
the chapter on High Temperature Coatings: Pro-
tection and Breakdown. However, diffusion coatings 1.25.4.1.1 Simple aluminide coatings
are widely used for aluminide intermetallic alloys The aluminum for growth of the aluminide coating
and will be discussed in this chapter. Aluminum is can be introduced through several different methods.
typically supplied in a gaseous or liquid phase, but One method is to sputter an aluminum, or aluminum-
solid-state diffusion is required for the growth of the containing, layer on the surface of the alloy. The
coating. For aluminide coatings of the same base alloy sputtering process is typically followed by a high
(e.g., NiAl on nickel-based alloys), only aluminum temperature anneal for diffusion of the aluminum
need be added to form an aluminide phase, although into the alloy. Another method is to place the sub-
other elements are often added for improved proper- strate in molten aluminum, which provides good
ties. However, for aluminide coatings with a different contact with the alloy as well as protection from
base than the substrate (e.g., NiAl on steel), both oxidation. The most common method for producing
elements of the aluminide phase (e.g., nickel and aluminide coatings, however, is pack cementation.
aluminum for NiAl on steel), must be provided, The pack in the pack cementation process consists
which complicates the coating process. of 10–30% of an aluminum source (aluminum metal
or aluminum-containing alloy) and a few percent of
an activator (typically a halide species), which are
1.25.4.1 Aluminide Coatings with Same
dispersed in an inert filler (typically alumina). Upon
Base as Substrate
heating, aluminum halide gaseous species are formed
Increasing the aluminum concentration at the surface and transport aluminum to the substrate material.
is especially useful in the titanium–aluminum sys- Pure aluminum can be used as the aluminum source
tem, since, as discussed earlier, TiAl is a borderline for process temperatures below the melting point
alumina-former. The formation of TiAl3 coatings on of aluminum, while aluminum-containing alloys
TiAl-based alloys can improve oxidation resistance.83 must be used for higher temperature deposition.
A similar approach can be used in the nickel or iron Pack cementation processes are referred to as high
systems, but an increase in the aluminum content of activity or low activity, based on the aluminum activ-
NiAl- or FeAl-based alloys is less critical, because, ity in the aluminum source. Coatings formed in a
with the lower stability of the nickel- and iron-oxide high-activity process are inward-growing as alumi-
phases, the alumina phase is more stable and the num diffuses into the alloy. Low-activity processes
additional aluminum content only provides signifi- (e.g., Al/Ni < 1), however, result in the outward
cant benefit for long exposure times when aluminum growth as both elements (e.g., Ni and Al) are depos-
depletion occurs. However, aluminide coatings can ited. The low-activity process is commonly referred
significantly improve the oxidation resistance of to as chemical vapor deposition (CVD), as both, or
metallic alloys with little or no aluminum. For exam- all, components of the coating are deposited. One of
ple, aluminide coatings have been used for more than the advantages of a low-activity process is that impu-
30 years to improve the oxidation resistance of super- rity elements, such as sulfur and phosphorus, are
alloys.84 Such alloys contain several alloying additions, removed during the process, so that high-quality
the concentrations of which can change during the coatings are produced. The low-activity CVD pro-
diffusion process, and can lead to undesired micro- cess is more complicated, but provides flexibility to
structural changes, such as the formation of topo- deposit a variety of coatings.87
graphically close packed (TCP) phases.85 Similarly,
the corrosion resistance of steel can be improved by 1.25.4.1.2 Alloyed aluminide coatings
aluminization to form an iron aluminide coating. In As with bulk materials, the corrosion resistance of
this case, the presence of carbon, which can lead to aluminide coatings can be improved with alloying
undesired carbide precipitation, places restrictions on additions, so the coating processes are often modified
the coating process.86 In the following sections, the to incorporate other elements in the coating. One
application of aluminide coating processes to simple example is the introduction of chromium to improve
the resistance to corrosion in chlorine-containing envir- formation of silica, which provides excellent oxidation
onments. Chromium, as well as reactive element addi- resistance because of its low oxygen permeability and is
tions, can be deposited simultaneously with aluminum particularly resistant to acidic melts. Like chromium,
in a pack cementation process by using an alloy con- silicon can be added in the pack cementation process,
taining the desired elements as the source.88 This and, also like chromium, a two-step process can be
process is complicated when steel is used as the sub- used to produce a coating with the desired composition
strate, because carbon in the steel reacts with chro- distribution.91 Although silicon can be introduced
mium to form chromium carbides. For low carbon through the vapor phase, the eutectic in the alumi-
contents, the carbide particles form separate precipi- num–silicon system provides an opportunity for the
tates, but for higher carbon contents, even as low as introduction of silicon and aluminum through liquid-
0.2%, a chromium carbide layer forms and disrupts the phase processes. This includes simple dipping pro-
coating process. This problem can be overcome by cesses as well as surface melting processes, such as
using a two-step process, which is illustrated in Figure laser surface melting. Siliconizing is most widely used
12.89 In the first step, the pack is heated to a relatively for titanium aluminide alloys. Although such processes
low temperature where there is a significant partial can result in an outer titanium silicide layer, the scale
pressure of the aluminum chloride vapor species, but formed during oxidation is similar to that formed on
the vapor species of the chromium chloride species is uncoated titanium aluminide alloys. The outer scales
low, so that aluminum, but not chromium, is deposited. consist of titanium and aluminum oxides, but a silica
Aluminum stabilizes the ferrite structure, so dissolu- layer is formed in the inner portion of the scale and
tion of aluminum causes a phase change in the iron improves the oxidation resistance.92
from austenite to ferrite. Because of the low carbon Aluminide coatings can also be used as the bond
solubility of ferrite, carbon is rejected into the austen- coating between superalloys and thermal barrier
ite layer below the ferrite. In the second step, the pack coatings (TBC). In this application, in addition to the
is heated to a higher temperature where the partial inherent oxidation behavior, the interface between the
pressure of the chromium vapor species increases, so thermally grown oxide (TGO) on the bond coating
chromium is now deposited and diffuses in the ferrite. and the TBC must be strong, so that the TBC
Because of the low concentration of carbon in the remains attached to the alloy during use (particularly
ferrite, chromium carbides do not form and a high- during thermal cycling). Thus, the adherence of the
quality coating is produced. A similar process has been scale is of greater importance than the growth rate,
used with a nickel-based superalloy to produce a coat- because a thicker TGO is less of a problem than the
ing with an outer high-chromium layer, which pro- loss of the TBC. Because of the importance of adher-
vides good hot corrosion resistance.90 Chromium, in ence, one of the common additions to NiAl bond
larger amounts, can also improve the oxidation resis- coating is platinum.93,94 Platinum-containing coat-
tance of titanium aluminide alloys and is often ings maintain good scale adherence even with signif-
incorporated in titanium aluminide coatings. icant amounts of sulfur, which, as discussed earlier,
Another element that is commonly added to alumi- generally degrades scale adherence. Platinum is typi-
nide coatings is silicon. Silicon additions result in the cally added as a coating prior to aluminization. There
are several commercial pack cementation processes for
producing platinum-containing coatings, including
Step 1 (low temperature) Gas (high AlCl, low CrCl2)
high-activity inward-growing coating with the sub-
Ferrite (low C – stabilized by Al) C Al strate in the pack (e.g., Chromalloy RT22), high-activ-
C-enriched austenite (C rejected from ferrite) ity inward-growing coating with the substrate above
the pack (e.g., Howmet SS82A), and low-activity out-
Steel
ward-growing coating (e.g., Thermatech MDC150L).95
Step 2 (high temperature) Gas (reduce AlCl, increase CrCl2)
The addition of platinum eliminates void formation at
Al Cr the coating–TGO interface, but voids are observed at
Ferrite (low C – stabilized by Al)
C the alloy–coating interface, as shown schematically in
Figure 13.96 Void formation beneath the scale can lead
C-enriched austenite (C rejected from ferrite)
to spallation of the scale during thermal cycling, which
Steel is especially critical if the coating is used as a bond coat,
Figure 12 Two-step process for codeposition of since such spallation will lead to loss of the TBC. Void
aluminum and chromium on steel.89 formation at the coating–scale interface has been
attributed to both the Kirkendall effect and stresses more complicated than simply increasing the alumi-
associated with the transformation of b-NiAl to num content of the surface. For example, as discussed
g0 -Ni3Al.97 The void formation at the alloy–coating earlier, alumina is more stable on NiAl than on TiAl,
interface is less critical than that at the coating–scale so NiAl can provide improved long-term corrosion
interface; as both NiAl and Ni3Al are more ductile than protection, but TiAl is less dense and has better
alumina, crack propagation is less likely. specific strength. Titanium aluminide coated with
As mentioned earlier, large amounts of platinum NiAl combines the advantages of the two materials
additions can lead to the formation of PtAl2, which in and can be produced by first electroplating the TiAl
addition to improving the supply of aluminum with nickel and then aluminizing the nickel layer by
needed to maintain an a-Al2O3 layer, also limits the pack cementation.99 The same process can be used
transport of transition metal impurities outward from for applying a NiAl coating to steel and allows for a
the aluminide coating into the scale. This is particu- reduction in the processing temperature to avoid
larly important in coatings for superalloys, which grain growth or carbide formation in the steel sub-
contain elements that can degrade the corrosion pro- strate.100 Similarly, steel, and other non-titanium
tection provided by the coating. alloys, can be coated with titanium aluminide by
Other noble metals used in NiAl bond coatings first depositing a layer of titanium and then aluminiz-
include rhenium and palladium. Rhenium acts ing the titanium to form the corrosion resistant tita-
like platinum in that it promotes the formation of nium aluminide coating.
an a-Al2O3 layer and forms a compound that acts as a An analogous process can be used to produce a
diffusion barrier. Palladium helps to stabilize the ruthenium aluminide coating on a nickel-base super-
NiAl phase and maintain sufficient aluminum in the alloy.101 In this case, the coating microstructure
scale. Noble metals can also be used with reactive depends on whether the ruthenium layer is alumi-
elements. For example, hafnium and platinum have nized using a high-activity pack cementation process
been used together, where, in addition to improving or a low-activity CVD process. As shown schemati-
the oxidation resistance, the formation of a Hf–Pt cally in Figure 14, in a high-activity process, the
layer provides a diffusion barrier to prevent loss of coating grows by inward diffusion of aluminum,
aluminum and incorporation of transition metals in which results in an outer layer of RuAl and an inner
the coating and scale. layer of NiAl that are separated by a zone containing
Titanium aluminide alloys can also be used with Ru(Al,Ta) precipitates. In a low-activity process, the
TBCs. The TBC has relatively good adherence to the coating grows by outward diffusion of aluminum,
alloy, which is attributed to the better match in the which results in an outer layer of NiAl. The coating
coefficients of thermal expansion as compared to with the outer NiAl layer has superior oxidation
nickel-based superalloys. However, a bond coat, such resistance. However, the presence of the inner RuAl
as TiAlCr or TiAl3, is needed for long lifetimes.98 layer is beneficial as it acts as a diffusion barrier,
which can prevent undesired elements from the sub-
strate from entering the outer coating and oxidation
1.25.4.2 Aluminide Coatings with Different
scale.
Base as Substrate
Deposition of an aluminide phase of an element not Low activity CVD
in the substrate can provide useful properties, but is aluminization
NiAl
Ru
Al2O3scale (TGO) Al-rich layer JNi RuAl
Superalloy substrate Superalloy substrate
Location High activity pack

β-NiAl coating of voids β-(Ni,Pt)Al coating aluminization
Ru JAl RuAl Ru(Al,Ta)
Al-rich layer ppt.
Y⬘(interdiffusion zone) Y⬘(interdiffusion zone) NiAl
Superalloy substrate Interdiffusion
zone
Superalloy substrate Superalloy substrate Superalloy substrate
Figure 13 Effect of platinum additions on pore formation Figure 14 Ruthenium-modified coatings by low-activity
in b-nickel aluminide coating after oxidation at 1150 C.96 CVD aluminization and high-activity pack aluminization.101
1.25.5 Summary and Conclusions 15. Hou, P. Y.; Niu, Y.; Van Lienden, C. Oxidation Metals
2003, 59, 41–61.
16. Pint, B. A. Oxidation Metals 1997, 47, 303–328.
Intermetallic aluminide alloys are attractive for high 17. Yang, J. C.; Nadarzinski, K.; Schumann, E.; Rühle, M.
temperature applications. The corrosion resistance of Scripta Metall. Mater. 1995, 33, 1043–1048.
18. McCarty, K. F. Surf. Sci. 2001, 474, L165–L172.
aluminide alloys is achieved by the formation and 19. Doychak, J.; Smialek, J. L.; Mitchell, T. E. Metall. Trans. A
maintenance of a protective alumina, or alumina- 1989, 20A, 499–518.
containing, scale. Of the three common aluminide 20. González-Carrasco, J. L.; Pérez, P.; Adeva, P.; Chao, J.
Intermetallics 1999, 7, 69–78.
systems, nickel, iron, and titanium, the formation of 21. Grabke, H. J.; Brumm, M.; Steinhorst, M. Mater. Sci.
an alumina scale is most challenging on titanium Technol. 1992, 8, 339–344.
aluminide alloys because of the thermodynamic sta- 22. Rivoaland, L.; Maurice, V.; Josso, P.; Bacos, M. P.;
Marcus, P. Oxidation Metals 2003, 60, 159–178.
bility of titanium oxide. However, titanium aluminide 23. Cadoret, Y.; Monceau, D.; Bacos, M. P.; Josso, P.;
alloys also offer the greatest benefit in terms of Maurice, V.; Marcus, P. Oxidation Metals 2005, 64,
improvements in specific properties. In addition to 185–205.
24. Heuer, A. H.; Reddy, A.; Hovis, D. B.; Veal, B.;
the initial thermodynamic stability of the oxides, Paulikas, A.; Vlad, A.; Rühle, M. Scripta Mater. 2006, 54,
long-term corrosion resistance requires maintenance 1907–1912.
of the protective scale in the presence of stresses 25. Zhang, Z. G.; Hou, P. Y. Mater. Sci. Eng. A 2005, 391,
1–9.
(thermal and transformation) and changes in alloy 26. Tolpygo, V. K.; Clarke, D. R. Acta Mater. 2004, 52,
composition with time. In applications where the 5129–5141.
intermetallic alloys have inadequate mechanical 27. Doychak, J.; Rühle, M. Oxidation Metals 1989, 31,
431–452.
properties for use as structural components, they 28. Yang, J. C.; Schumann, E.; Levin, I.; Rühle, M. Acta
can be applied as coatings to provide corrosion resis- Mater. 1998, 46, 2195–2201.
tance for structural alloys. Improvements in corro- 29. Bobeth, M.; Pompe, W.; Schumann, E.; Rühle, M. Acta
Metall. Mater. 1992, 40, 2669–2676.
sion resistance can be achieved with alloying 30. Balmain, J.; Huntz, A. M. Oxidation Metals 1996, 45,
additions, but the effects of these elements on other 183–196.
factors, such as mechanical properties, cost, and man- 31. Yi, H. C.; Smeltzer, W. W.; Petric, A. Oxidation Metals
1996, 45, 281–299.
ufacturability, must also be considered. 32. Kartono, R.; Young, D. J. Electrochem. Soc. Proc. 2004,
16(High Temperatures Corrosion and Materials
Chemistry V), 43–55.
33. Pint, B. A.; Haynes, J. A.; Zhang, Y.; More, K. L.; Wright, I. G.
References Surf. Coatings Tech. 2006, 201, 3882–3886.
34. Pint, B. A.; Regina, J. R.; Prussner, K.; Chitwood, L. D.;
Alexander, K. B.; Tortorelli, P. F. Intermetallics 2001, 9,
1. Lasalmonie, A. Intermetallics 2006, 14, 1126–1129. 735–739.
2. Dimiduk, D. M. Mater. Sci. Eng. 1999, A263, 281–288. 35. Zhang, Z. F.; Thaidigsmann, K.; Ager, J.; Hou, P. Y.
3. Stoloff, N. S.; Sikka, V. A. Physical Metallurgy and J. Mater. Res. 2006, 21, 1409–1419.
Processing of Intermetallic Compounds; Chapman & 36. Hou, P. Y. J. Am. Ceram. Soc. 2003, 86, 660–668.
Hall: New York, NY, 1996. 37. Kuenzly, J. D.; Douglass, D. L. Oxidation Metals 1974, 8,
4. Brady, M. P.; Gleeson, B.; Wright, I. G. JOM 2000, 52(1), 139–178.
16–21. 38. Pint, B. A. Oxidation Metals 1996, 45, 1–37.
5. Meier, G. H.; Pettit, F. S. Mater. Sci. Eng. A 1992, 153, 39. Schumann, E.; Yang, J. C.; Rühle, M.; Graham, M. J.
548–560. Oxidation Metals 1996, 46, 37–49.
6. Doychak, J. In Intermetallics Compounds Principles and 40. Pint, B. A.; Hobbs, L. W. Oxidation Metals 2004, 61,
Practice. Vol. 1. Principles; Westbook, J. H., 273–292.
Fleischer, R. L., Eds.; John Wiley, 1995; pp 977–1016. 41. Berlanga, M.; González-Carrasco, J. L.;
7. Cathcart, J. V. Mater. Res. Soc. Proc. 1985, 39 Montealegre, M. A.; Muñoz-Morris, M. A. Intermetallics
(High-Temperature Ordered Intermetallic Alloys), 2004, 13, 205–212.
445–459. 42. Montealegre, M. A.; Strehl, G.; González-Carrasco, J. L.;
8. Grabke, H. J. Intermetallics 1999, 7, 1153–1158. Borchardt, G. Intermetallics 2005, 13, 896–906.
9. Tottorelli, P. F.; Natesan, K. Mater. Sci. Eng. A 1998, 258, 43. Taniguchi, S.; Shibata, T. In Oxidation of
115–125. High-Temperature Intermetallics; Grobstein, T.,
10. Pettit, F. S. Trans. Metall. Soc. AIME 1967, 239, Doychak, J., Eds.; The Minerals, Metals & Materials
2196–1305. Society: Warrendale, PA, 1989; pp 17–29.
11. Godlewska, E. Electrochem. Soc. Proc. 2003, 16 44. Pint, B. A.; Wright, I. G. Materials Science Forum 2004,
(High Temperature Corrosion and Materials 461–464(High Temperature Corrosion and Protection of
Chemistry IV), 51–61. Materials 6, Part 2), 799–806.
12. Katsman, A.; Grabke, H. J.; Levin, L. Oxidation of Metals 45. Pint, B. A.; Schneibel, J. H. Scripta Mater. 2005, 52,
1996, 46, 313–331. 1199–1204.
13. Liu, Z.; Gao, W. Intermetallics 2000, 8, 1385–1391. 46. Pérez, P.; Adeva, P. Oxidation Metals 2006, 65, 15–38.
14. Choi, S. C.; Cho, H. J.; Kim, Y. J.; Lee, D. B. Oxidation 47. Hou, P. Y.; McCarty, K. F. Scripta Mater. 2006, 54,
Metals 1996, 46, 51–72. 937–941.
48. Fergus, J. W. Mater. Sci. Eng. A 2002, 338, 108–125. 75. Lee, W. H.; Lin, R. Y. Mater. Chem. Phys. 2003, 77,
49. Brady, M. P.; Pint, B. A.; Tortorelli, P. F.; Wright, I. G.; 86–96.
Hanrahan, R. J., Jr. In Corrosion and Environmental 76. Tang, Z.; Wang, F.; Wu, W. Oxidation Metals 1999, 51,
Degradation, Vol. II; Schütze, M., Ed.; Wiley-VCH Verlag 235–250.
GmbH, Weinheim, Germany, 2000; pp 229–326. 77. Kim, I.; Cho, W. D. Mater. Sci. Eng. A 1999, 264,
50. Rahmel, A.; Quadakkers, W. J.; Schütze, M. Mater. 269–278.
Corros. 1995, 46, 271–285. 78. Nicholls, J. R.; Leggett, J.; Andrews, P. In Oxidation
51. Leyens, C. In Titanium and Titanium Alloys; Leyens, C., of Intermetallics; Grabke, H. J., Schütze, M., Eds.;
Peters, M., Eds.; Wiley-VCH, 2003; pp 187, 230. Wiley-VCH Verlag GmbH: Weinheim, Germany, 1998;
52. Haanappel, A. C.; Hofman, R.; Sunderkötter, J. D.; pp 329–344.
Glatz, W.; Clemens, H.; Stroosnijder, M. F. Oxidation 79. Li, Y. S.; Spiegel, M. Corrosion Sci. 2004, 46, 2009–2023.
Metals 1997, 48, 263–287. 80. Zhang, K.; Gao, W.; Liang, J. Intermetallics 2004, 12,
53. Dettenwanger, F.; Schumann, E.; Rühle, M.; Rakowski, J.; 539–544.
Meier, G. H. Oxidation Metals 1998, 50, 269–307. 81. Liang, J.; Gao, W.; Li, Z.; He, Y. Mater. Lett. 2004, 58,
54. Shemet, V.; Tyagi, A. K.; Becker, J. S.; Lersch, P.; 3280–3284.
Singheiser, L.; Quadakkers, W. J. Oxidation Metals 2000, 82. Amaya, M.; Espinosa-Medina, M. A.; Porcayo-
54, 211–235. Calderon, J.; Martinez, L.; González-Rodriguez, J. G.
55. Tang, Z.; Shemet, V.; Niewolak, L.; Singheiser, L.; Mater. Sci. Eng. A 2003, 349, 12–19.
Quadakkers, W. J. Intermetallics 2003, 11, 1–8. 83. Gauthier, V.; Dettenwanger, F.; Schütze, M.;
56. Kumagai, M.; Shibue, K.; Kim, M. S.; Yonemitsum, M. Quadakkers, W. J. Oxidation Metals 2003, 59, 233–255.
Intermetallics 1996, 4, 557–566. 84. Lindblad, N. R. Oxidation Metals 1969, 1, 143–170.
57. Kim, Y. J.; Kim, I. B.; Kim, M. S.; You, B. D.; Han, W.; 85. Rae, C. M. F.; Hook, M. S.; Reed, R. C. Mater. Sci. Eng. A
Hwang, W. S. Metals Mater. 2000, 6, 455–459. 2005, 396, 231–239.
58. Donchev, A.; Richter, E.; Schütze, M.; Yankov, R. 86. Xiang, Z. D.; Datta, P. K. Metall. Mater. Trans. A 2006,
Intermetallics 2006, 14, 1136–1142. 37A, 3347–3358.
59. Donchev, A.; Gleeson, B.; Schütze, M. Intermetallics 87. Warnes, B. M. Surf. Coatings Tech. 2003, 163–164,
2003, 11, 387–398. 106–111.
60. Yoshihara, M.; Taniguchi, S.; Zhu, Y. C. Metall. Mater. 88. Bianco, R.; Rapp, R. A. J. Electrochem. Soc. 1993, 140,
Trans. A 2003, 34A, 2253–2261. 1181–1190.
61. Felten, E. J.; Pettit, F. S. Oxidation of Metals 1976, 10, 89. Zheng, M.; Rapp, R. A. Oxidation Metals 1998, 49, 19–31.
189–223. 90. Bai, C.-Y.; Luo, Y.-J.; Koo, C.-H. Surf. Coatings Technol.
62. Dickey, E. C.; Pint, B. A.; Alexander, K. B.; Wright, I. G. 2004, 183, 74–88.
J. Mater. Res. 1999, 14, 4531–4540. 91. Brossard, J. M.; Hierro, M. P.; Trilleros, J. A.; Carpintero,
63. Soldera, F.; Ilić, N.; Brännström, S.; Barrientos, I.; M. C.; Sánchez, L.; Bolı́var, F. J.; Pérez, F. J. Surf.
Gobran, H.; Mücklich, F. Oxidation Metals 2003, 59, Coating Technol. 2006, 201, 5743–5750.
529–542. 92. Xiong, H. P.; Mao, W.; Xie, Y.-H.; Ma, W.-L.; Chen, Y.-F.;
64. Ilić, N.; Soldera, F.; Mücklich, F. Intermetallics 2005, 13, Li, X.-H.; Li, J.-P.; Cheng, Y.-Y. Acta Mater. 2004, 52,
444–453. 2605–2620.
65. Cao, F.; Nandy, T. K.; Stobbe, D.; Pollock, T. M. 93. Pint, B. A. Surf. Coatings Technol. 2004, 188–189, 71–78.
Intermetallics 2004, 15, 34–43. 94. Pint, B. A.; Speakman, S. A.; Rawn, C. J.; Zhang, Y. JOM
66. Kuppusami, P.; Murakami, H. Surf. Coatings Technol. 2006, 58(1), 47–52.
2004, 186, 377–388. 95. Angenete, J.; Stiller, K.; Langer, V. Oxidation Metals
67. Natesan, K.; Datta, P. K. In Intermetallic Compounds, 2003, 60, 47–82.
Principles and Practice, Vol. 3, Progress; Westbrook, J. 96. Zhang, Y.; Haynes, J. A.; Lee, W. Y.; Wright, I. G.; Pint, B.
H., Fleischer, R. L., Eds.; Wiley, 2002; pp 707–719. A.; Cooley, K. M.; Liaw, P. K. Metall. Mater. Trans. A
68. Natesan, K. Oxidation Metals 1988, 30, 53–83. 2001, 32A, 1727–1741.
69. Kai, W.; Chang, M. T.; Bai, C. Y. Oxidation Metals 2001, 97. Angenete, J.; Stiller, K.; Bakchinova, E. Surf. Coatings
56, 191–214. Technol. 2004, 176, 272–283.
70. Kai, W.; Huang, R. T. Oxidation Metals 1997, 47, 59–86. 98. Leyens, C.; Braun, R.; Fröhlich, M.; Hovsepian, P. Eh.
71. Kai, W.; Lee, S. H.; Chiang, D. L.; Chu, J. P. Mater. Sci. JOM 2006, 58(1), 17–21.
Eng. A 1998, 258, 146–152. 99. Izumi, T.; Nishimoto, T.; Narita, T. Intermetallics 2005, 13,
72. Godlewska, E. Mater. Corros. 1997, 48, 687–699. 727–732.
73. Izumi, T.; Yoshioka, T.; Hayashi, S.; Narita, T. 100. Xiang, Z. D.; Datta, P. K. J. Mater. Sci. 2005, 40,
Intermetallics 2000, 8, 891–901. 1959–1966.
74. González-Rodrı́guez, J. G.; Luna-Ramı́rez, A.; 101. Tryon, B.; Feng, Q.; Wellman, R. G.; Murphy, K. S.;
Salazar, M.; Porcayo-Calderon, J.; Rosas, G.; Martı́nez- Yang, J.; Levi, C. G.; Nicholls, J. R.; Pollock, T. M. Metall.
Villafañe, A. Mater. Sci. Eng. A 2005, 399, 344–350. Mater. Trans. A 2006, 37A, 3347–3358.
1.26 High Temperature Corrosion of Ceramics and
Refractory Materials
K. G. Nickel
Applied Mineralogy, Institute for Geosciences, University Tübingen, Wilhelmstr. 56, D-72074 Tübingen, Germany
P. Quirmbach and J. Pötschke
Deutsches Institut für Feuerfest und Keramik GmbH, An der Elisabethenkirche 27, D-53113 Bonn, Germany

1.26.2 Differences between Metals and Ceramics 670
1.26.3 Corrosion Modes and Kinetics 671
1.26.4 Corrosion Measurement 673
1.26.5 Oxides 674
1.26.5.1 Alumina 674
1.26.5.2 Zirconia 674
1.26.5.3 Cordierite 674
1.26.6 Carbides 675
1.26.6.1 Boron Carbide 675
1.26.6.2 Silicon Carbide 675
1.26.7 Nitrides 678
1.26.7.1 Silicon Nitride 678
1.26.7.2 Boron Nitride 679
1.26.7.3 Silico-Carbonitrides 680
1.26.8 Titanium Compounds 680
1.26.9 Ultrahigh Temperature Ceramics 680
1.26.10 Refractories 681
1.26.10.1 Chemical Dissolution 681
1.26.10.2 Erosion 681
1.26.10.3 Mechanical Wear 681
1.26.11 Chemical Dissolution 681
1.26.11.1 Diffusion/Blast Furnace 681
1.26.11.2 Marangoni Convection/SEN 682
1.26.11.3 Oxidation/Ladle 683
1.26.12 Erosion 684
1.26.12.1 Blast Wear/Pipes 684
1.26.12.2 Flowing Melts/Converter 684
1.26.13 Mechanical Wear 685
1.26.13.1 Thermal Shock Resistance/Ladle 685
1.26.13.2 Carbon Bursting/Blast Furnace 685
1.26.13.3 Destruction due to Impact/Converter 686
1.26.14 Infiltration by Steel and Slag 686
1.26.14.1 Basic Equations 686
1.26.15 Examples 687
1.26.15.1 Slag/MgO 687
1.26.15.2 Steel/MgO 687
1.26.15.3 Steel/MgO–C 688
References 689
668
High Temperature Corrosion of Ceramics and Refractory Materials 669
Glossary Abbreviations
Converter The containment for the industrial SEN Submerged entry nozzle
conversion of ore to metal and its purification SSiC Pressureless sintered silicon carbide
by slag formation.
Ellipsometry Measurement of the change of light
polarization upon reflection or transmission.
In corrosion, it is usually done in reflection Symbols
and used to infer the thickness of a very thin A Surface area (m2)
layer (down to the angstrom-level) of reaction a Thermal diffusivity (m2 s1)
product. C0 Concentration of carbon in liquid iron (wt %)
Flue gas Exhaust gases from a fireplace, oven, CS Saturation concentration of carbon in liquid
furnace, or other production process. iron (wt %)
Hot corrosion Historically evolved name for the Di Diffusion coefficient of component i (m2 s1)
attack of a material by or involving salt melts. Dp Pore-diffusion coefficient (m2 s1)
A typical representative is a Na2SO4 melt. E Modulus of elasticity (GPa)
Hydrothermal Conditions which involve a water GF Crack energy (J m2)
fluid at elevated temperatures. Commonly DG0 Free enthalpy in the standard state (J mol1)
used for conditions above 100 C at g Acceleration of gravity (m s2)
pressures above the atmospheric pressure. h Height (m)
Ladle Container for transport and pouring of (metal) hl Diameter of the convection loop (m)
melts. hg Characteristic length of the corrosion groove (m)
Rate constant Constant which relates the K Equilibrium constant
quantitative progress of a process, example, klinear Linear rate constant (m s1 or kg m2 s1)
a growth of a layer on an infinite plate, to time. klog Logarithmic rate constant
Slag By-product of ore smelting or refining. It kp Parabolic rate constant (m2 s1 or kg2 m4 s1)
usually consists mainly of oxides, with or l Length (m)
without other constituents like sulfides, Pi Partial pressure of species i
phosphates, and metals. R Universal gas constant (8314 J mol1 K1)
Steady-state conditions Conditions in which a T Temperature (K)
system has a constant behavior with time. In t Time (s)
corrosion, this often implies that external u Stoichiometry factor
conditions such as the atmospheric partial VI Molar volume of phase I (m3 mol1)
pressure of the attacking medium are kept v Flow velocity (m s1)
constant and a dynamic equilibrium is Dw Mass change (kg m2)
established, where two or more reversible x Mass change per unit area (kg m2) or scale
processes occur at one rate. thickness (m)
Thermogravimetry Continuous instrumental a Thermal expansion coefficient (K1)
determination of mass changes of a sample b Mass transfer coefficients (m s1)
as a function of time, with or without regular g Labyrinth factor, tortuosity
changes in temperature, pressure, or gas d Boundary layer of diffusion (m)
composition during the observed time. dPr Prandtl’s boundary layer thickness (m)
Wetting angle Contact angle at which the liquid– dN Nernst’s diffusion-boundary-layer
vapor interface of a fluid and the solid–vapor thickness (m)
interface of a contacting solid meet. pc Open porosity of the decarburized zone (%)
Describes the tendency of the fluid to form a ec Critical elongation
film (perfectly at 0 C) or a ball resting on the r Density (kg m3)
surface (perfectly above 150 C). Dynamic viscosity (kg m s1)
m Poisson ratio
y Kinematic viscosity (m2 s1) A refractory material is firstly a ceramic that has
s Interfacial tension (N m1) not reached the softening point, at which a special
sc Critical stress (MPa) pyramid (Seger-cone SK17) would bend when heated
u Wetting angle ( ) to 1500 C. However, the material has to fulfill other
t Correction time (s) criteria at high temperatures to be called refractory,
namely, it has to keep its shape, survive thermal fluc-
tuations, and withstand attack by slags or melts for a
sufficiently long time.5,6 Refractories are in most cases
1.26.1 Introduction porous composites, making the behavior complex.
There is an overlap in materials chemistry
Ceramics are nonmetallic–inorganic materials, which between refractories and technical ceramics. For
are usually made by forming from a mass or a powder example, alumina and silicon carbide are manufac-
and firing at high temperature, upon which the mate- tured both as refractories and as dense, advanced
rial becomes consolidated and obtains its properties. engineering ceramics. For this reason, refractories
This implies that ceramics have thermal and chemi- are dealt with in the form of typical examples from
cal stability at the firing temperature in the firing the steel- and iron-making industries. Other applica-
environment. ‘High temperature’ is a very blurred tions have similar principal problems and, from a
term, but most people would accept temperatures chemical point of view, are contained within the
above 800 C as high temperature. An exception is ceramics section.
the behavior in hydrothermal conditions. Hot, pres-
surized water is very reactive and many compounds
have a high level of solubility in it. For this reason, 1.26.2 Differences between Metals
we will describe the hydrothermal behavior of cera- and Ceramics
mics at lower temperatures.
A classical division of ceramics is silicate, oxide, and In the corrosion of metals, we usually find oxidation
nonoxide ceramics. Textbooks and handbooks on their of the metal to form some phase, which is then lost to
properties and manufacture have been published1–4 the environment by spallation, dissolution, or some
and contain chapters on their corrosion resistance. other mechanisms, resulting in mass loss or the for-
Metals and intermetallics do survive at high tem- mation of a surface layer, which can have protective
peratures in many environments only because they characteristics. Basic cases like the oxidation of iron
form phases, commonly oxides, which protect them are well known7 and may be described by simple laws
from further attack. In this respect, many chapters such as a parabolic weight gain in certain regimes.
in this book deal with ceramics. In order not to The complexities of scale evolution often necessi-
repeat too many issues, readers are referred to appro- tate a more complex modeling of the transport in
priate chapters, in particular the ones on Oxidation oxide scales, and finite difference calculations are
of Metals and Alloys and High Temperature Coat- used to predict the process progression.8
ings: Protection and Breakdown. In this chapter, we The importance of the oxide development in
will deal with bulk materials. silicon-based semiconductors has led to the devel-
There are many applications of ceramics at high opment of sophisticated detailed models for high
temperatures: one deals with materials that contain temperature oxidation, whether related to initial
and protect from melts in the metal or glass industry, periods9 or general laws.10
another is the function to transport or contain heat in Ceramics are often oxides, carbides, or nitrides. In
predominantly gaseous environments. For some tasks, general, nonmetals have a higher electronegativity
structural loads under adverse conditions have to be than metals. The bonding is mostly a mixed ionic–
met, whereas for others, several forms of attack may covalent one, but formally the electrons have to be
occur, e.g., waste incineration. A rather new applica- assigned to the nonmetal. High temperature oxida-
tion is the use of SiC as an electronic material, tion of these carbides and nitrides thus oxidizes the
commonly in the form of single crystal wafers. carbon or the nitrogen and leaves the metal or semi-
Although the use is at low or moderate temperatures, conductor in its oxidation state. Also, in contrast to
a high temperature treatment to produce insulating metals, where often the metal is transported as an ion
silica is a common step in the production of these to the surface to become oxidized, in ceramics we
new functional materials. usually have the transport of the corroding agent to
the scale–substrate interface where reaction occurs. boundary phase removal concurrent with a more
The oxidation process is therefore better compared or less inert matrix material gives us the picture
with semiconductors rather than with ordinary metals. of increasing pathway lengths for the attacking
Oxide ceramics are stable in environments having agent with time. This is very likely to slow down
sufficient oxygen; their corrosion is based on reac- the corrosion progress. In such cases, eqn. [1] is
tions that form other phases, becoming dissolved or correct for neither the bulk material nor the indi-
evaporated. vidual phase.
Porous materials have another prominent prob-
lem: establishing contact with the medium so that
the development of transport pathways becomes
1.26.3 Corrosion Modes and Kinetics complex. Wetting behavior is of utmost importance,
as it governs the penetration and hence the true
To understand the principles, we first assume the
contact surface. Wetting depends on surface energies
material to be single phased, isotropic, dense, and to
and these are likely to change with time due to
have perfect surfaces. Just like metals, ceramics can
gradient developments, saturation effects, etc. Such
show both active and passive corrosion behaviors.
problems are dealt with in detail in Section 1.26.14
Unless retarding phases are formed, steady-state pro-
on refractory corrosion in the form of typical exam-
cesses in active corrosion have simple and constant
ples and equations.
mass loss rates for a given set of physical boundary Although the chemical removal of phases is a corro-
conditions. It does not matter whether the chemical
sive process, it may well be influenced by erosive or
reaction rate, the dissolution kinetics, or the diffusion
mechanical stresses accompanying the attack. This is
from or into the corrosive medium is rate determin-
also described in the refractory section of this chapter
ing. The constancy of boundary conditions such as
(Section 1.26.10). At this point, we only want to hint
constant gradients in the attacking agent necessitates
that the formation of a solid reaction product during
linear kinetics following eqn. [1], in which x denotes
corrosion may not always retard the corrosion progress.
either the thickness of the layer lost or the mass of
Such a special form of active corrosion, which is par-
material loss per unit area, t the time, and klinear is the ticular to nonoxide ceramics, has been described for
linear rate constant:
silicon carbide at very high temperatures,11 where a
xðm or kgm2 Þ ¼ klinear ðms1 or kg m2 s1 Þ t ðsÞ ½1 high pressure of CO can create bubbles, which are
easily removed by spallation and erosion.
The time to reach those steady-state conditions is not The formation of solid phases during corrosion
covered by the model, so initial periods with differing can often result in scales or layers. It depends on
rates are possible. In the initial period, transport- their physical nature whether they hamper the cor-
related problems are less likely to dominate, so for- rosion process or not. Oxide scales on nonoxide
mation or removal of surface adlayers may lead to ceramics often do not have such a high mismatch in
incubation periods. Alternatively, the buildup of a thermal expansion coefficients as oxidized metals.
gradient may be fast in the beginning and turn into Hence, stresses in the scale can be lower and spall-
a steady-state rate later. ation is a less common phenomenon. Nonetheless,
The ease of this formulation hides the problems the oxides can undergo phase transitions with volume
associated with it. Everything rests on the constancy changes large enough to cause stresses and cracking
of external conditions, and therefore the measured during thermal cycling. Multiphase scales likewise
rates depend on a large number of boundary condi- can develop stresses that induce cracks. Bubbles are
tions. Laminar or turbulent flows of the attacking another way of potentially short-circuiting the scale.
medium, convection or still conditions, and flow In all such cases, eqn. [1] is also valid for passive
velocity of the attacking gas or fluid are only a few corrosion, because the scale that develops is not pro-
obvious parameters. tective. The equation has then positive constants for
For polyphasic materials, the resistance against mass gain. Also in any real process where a scale
attack is usually different for the different phases. becomes protective, there is an initial period, in
Here, eqn. [1] is only valid for each phase individu- which it is not thick enough to be rate determining.
ally and only as long as the processes do not influence This has been established for the oxidation of silicon
each other. The example of a preferential grain by Deal and Grove,10 where the linear character of
boundary phase attack is easy to understand. Here, the initial period has been shown.
In both metals and ceramics, corrosion in which the to be calculated to relate rate constants to each other
transport of the reacting agent through a growing layer by eqn. [7]:
is governed by the diffusion of this agent or the reac-
r2Product ðkg m3 Þ2
tion product through the scale, we have long-time kp ðm s1 Þ ¼ kp ðkg2 m2 s1 Þ ½7
behavior approximated by a simple parabolic law u2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi A third basic law is the logarithmic or asymptotic law
xðm or kgm2 Þ ¼ kp ðm2 s1 or kg2 m4 s1 Þ tðsÞ ½2 x ¼ c þ klog logðt Þ ½8
In eqns. [2]–[5], x now denotes either the thickness of This may be used for corrosion processes, which are
the layer formed or the mass of material gained per effectively stopped by the formation of a material
unit area, t is the time, and kp is the parabolic rate barrier during corrosion. It basically describes the
constant. Following the reasoning above, both parts of decrease in the effective exposed surface area. The
the process can be joined with the linear–parabolic barrier may be a crystalline phase with a much lower
model10 to give diffusion coefficient.
x 2 þ Ax ¼ Bðt þ tÞ ½3 Kinetic breaks are observed in cases where a phys-
ical effect interferes with the corrosion process. An
in which several physical parameter are contained in example is the repeated cracking of the scale. If a
the factors A and B and a correction time t is intro- critical thickness is responsible for the critical stress,
duced, which accounts for an initial scale thickness. By which initiates the cracking process, pseudolinear
setting reasonable boundary conditions, it is found that behavior with a succession of periods of parabolic
B equals kp for long times and thus eqn. [3] is approxi- growth and intervening periods of renewal of short-
mated by eqn. [2], B/A equals klinear and thus eqn. [3] is circuiting paths follows.
approximated by eqn. [1] for short times. Note that Internal oxidation is observed in multiphase cera-
‘short times’ is an expression relating to corrosion mics. A grain boundary phase is often the weakest
progress and can last, for example, for days at 700 C material in terms of corrosion resistance and provides
in the oxidation of silicon. a transport path for the exchange of matter between
It was claimed that the oxidation data for silicon the inside and the outside of the structure. The pro-
can equally well be fitted by a simple power law12 duction of selectively corroded phases in the interior
x ¼ at b ½4 induces internal stresses because of volume increases.
After a critical stress level is reached, mechanical
This method has not been used extensively in the failure of parts of the component near the surface
literature for oxidation of semiconductors or cera- may follow. If this cannot be healed during the process,
mics. The lack of a good physical explanation for the a large new surface area is opened and a ‘break-away’
parameters of eqn. [4] is probably the main reason. situation is likely, i.e., a change to fast linear corrosion.
For easy fitting, a variant of eqn. [3] may be Another typical break situation is observed in
preferred: porous materials. The closure of pore openings by
x2 x the formed scale diminishes the exposed surface
þ t¼t ½5 markedly and almost instantaneously. A much slower
kp kl
progress follows after closure.
A difference to be noted between ceramics and Simple basic laws of passive processes are only
metals is that the corrosion progress x is not directly true for constant diffusion coefficients. In all cases,
transferred from thickness to weight change units via where extrinsic or intrinsic impurities, additives, etc.,
densities r, because in nonoxides there is a loss of the are present and enter the scale forming material, an
accompanying light element. As described else- influence on the diffusion coefficient must be envi-
where,13 a stoichiometry factor relating to the reac- saged. An increase or decrease in the coefficient
tion equation causes the mass change curve to become steeper or
ðn MÞCondensed product flatter relative to parabolic behavior. Simulations
u¼ ½6 show that for strong effects of this kind linear [1] or
ðn MÞCondensed product ðn MÞMaterial
asymptotic behavior [7] is approximated.
where n is the number of moles in the chemical Therefore, we have good reasons to envisage com-
reaction equation and M the molecular weight, has plex kinetic behavior during corrosion to be possible.
Complex strictly physical models are a matter of Scale growth monitoring measurements, e.g., by
model calculations still and hard to generalize. ellipsometry, which are common for oxidation of
Attempts to model complex situations by semi- or silicon, have rarely been used in ceramics. Only in
purely empirical curve fitting procedures, e.g., by recent years, with SiC wafers for electronic applica-
using a multiple linear approach in the form tions becoming available, have such data been pro-
duced in greater numbers.
pffiffi
x ¼ k0 þ klin t þ kp00 t þ klog logðt Þ ½9 The most common experiments in ceramics cor-
rosion are still exposures in special environments,
followed by weighing and inspecting with micro-
have been proposed14, 15 and shown to be quite suc- scopic and other available analytical techniques.
cessful. Attempts to introduce an even higher degree Melt corrosion is often studied by ordinary drop,
of sophistication made the equations less robust crucible, and finger tests; these should only be used to
to the errors inherited from the uncertainties of rank materials, because they give qualitative data
measurement.16 inside a narrow frame set by the standardized experi-
A special case of complex kinetics arises from mental setup.
situations in which active and passive processes Burner rigs are used to test for salt melt corrosion
occur simultaneously. This has been discussed exten- (‘hot corrosion’).21 Salts are injected into the flame of
sively because it is of great importance to applica- a burner and deposited at constant rates onto sam-
tions such as gas turbines.17 Corrosion in which a ples, which are positioned and agitated in front of the
scale is formed and evaporated at the same time has apparatus exhaust.
a typical pattern in plots of mass change against time, Post-inspection of samples from furnace tests is
because the initial gain will turn into a linear loss most common for the investigation of interactions
after longer times. Opila18 has shown that the equa- with gases. As a typical example for the contamina-
tion for metal oxidation with evaporation reported by tion problems associated with the method, we can
Tedmon19 in the form look at protective tubes made from alumina. Most
2 gases, even expensive high-purity types, do contain
u kp 2kl Dw1 2kl Dw1
t ¼ 12 ln 1 water at the ppm-level. Unless the alumina tube was
2kl u1 kp u1 kp fired and flushed at high temperatures for times in
½10
Dw2 the order of 1000 h, it may contaminate the sample
with sodium, and as long as extreme measures have
u2 kl
not been taken to dry the atmosphere (e.g., by P2O5),
where u1 and u2 are stoichiometric factors, which there may be substantial contamination by alumina in
account for the mass balance of the appropriate reac- the form of transfer as hydroxide.22 In deliberately
tions, Dw1 and Dw2 are weight gain and weight loss moistened atmospheres, the effect is quite high.23
by scale growth and volatilization, respectively, is suc- The effect on oxidation rates for silica formers has
cessful in describing this steady-state situation, leading been shown to be extensive even for low contamination
to the establishment of a steady-state scale thickness levels,24 because the contaminants are good glass for-
and progressing corrosion of the material. The engi- mers, and may change the properties of growing scales.
neering basis for the loss function was also explored Apart from contamination, local atmospheres may play
and simulated with parameters for the flow of fluids.20 a role, which may explain the observed influences of
furnace size on experimental oxidation rates.25
1.26.4 Corrosion Measurement High temperature experiments have always an ex-
tended level of contamination possibilities. An increased
In situ experiments of ceramic corrosion can only be level of uncertainty in corrosion data comes also from
done for simple cases. Simple oxidation of a nonoxide in sample inhomogeneities. Preferential corrosion of sec-
a defined atmosphere can be followed continuously by ondary or grain boundary phases is often calculated
thermogravimetry (TGA), which allows proper kinetic directly relative to the total sample surface and yields
modeling. Resolution, accuracy, precision, and long- a corresponding bias in parameter evaluation. Adding
time stability of the apparatus have to be controlled. the increased uncertainty levels of temperature mea-
Other problems, e.g., contamination or the creation of surement and physical side effects such as the relative
local atmospheres, are identical to those discussed below. large areas in the vicinity of an edge in the usually small
samples, it is plain that one has to allow for some these are incorporated as grain boundary phases
systematic and statistical error in the experimental data. with accordingly selective preferential damage. In
particular, silica and alkalis are typical impurities
for alumina in engineering ceramics and even small
1.26.5 Oxides
amounts have been shown to be responsible for the
corrosion in hydrothermal conditions.32 Similarly
A number of oxides with high melting points could
corrosion by salt melts is related to impurities and
qualify as high temperature materials. Out of these
grain boundary phases.33
only a small number is used as bulk engineering mate-
rial, and therefore only alumina and zirconia are dealt
with here in detail. The other oxides are either cov- 1.26.5.2 Zirconia
ered by the chapter on ‘‘Oxidation of Metals and
Zirconia (ZrO2) is a high temperature phase with a
Alloys’’ or are covered by the chapter on Oxidation
melting point above 2500 C. Extremely high values
of Metals and Alloys or are contained within the
even for the lowest eutectic temperatures in the binary
refractory sections of this chapter.
systems with alumina and silica make it a prime mate-
rial for refractory systems by itself or with those oxi-
1.26.5.1 Alumina
des. For many conditions, it will behave in an almost
The only form of alumina that is stable at high temper- inert manner in contact with silica or alumina.
ature is corundum or a-Al2O3. It is well known for its The section on thermal barrier coatings in this
excellent chemical resistance, even at high tempera- book discusses the corrosion aspects in those high
tures. For this reason, it is a common tube material in temperature applications.
high temperature furnace systems with oxidizing atmo- As a pure monolithic material, it is more problem-
spheres. The applications at high temperature are often atic, because martensitic phase transformations with
limited by other properties rather than corrosion, volume changes occur and tend to embrittle the
namely creep and poor mechanical properties. material during thermal cycling. The addition of
Nonetheless, there are some restrictions for the use Mg, Ca, Y, and Ce stabilizes the higher temperature
of alumina based on its chemical behavior. At tem- modifications (for a review on the basic science of
peratures of 1700 C in Ar/H2O mixtures, evaporation these transformations see Hannink et al.34), and with
of alumina was reported and excessive grain growth appropriately high concentrations the cubic modifi-
may have enhanced the problems.26 Reducing envir- cation is stable down to room temperature.
onments can also foster carbothermic reduction of In a stabilized modification, the use of zirconia
alumina, e.g., in the presence of SiC,27 but in general as a refractory material is possible. Such refractories
a good resistance to flue gases of the aluminum- are susceptible to corrosion by slags, because slags
remelting industry has been reported.28 The attack with low basicity leach MgO, and highly basic slags
by alkali metal vapors is impurity dependent. can dissolve the material completely. Applications
Corrosion by aggressive gases can be estimated via in remelting furnaces showed a good resistance to
calculation of the partial pressures of volatile species flue gases.
over alumina.29 This has been shown for halogens30 The stabilization can have deleterious effects on
and is certainly true for other such agents. The Ce-doped materials because the variation of external
vaporization in water containing systems is mainly oxygen fugacity can induce corrosion due to a change
in the form of Al(OH)3. Therefore, in high-velocity in the oxidation state. Corrosion by water is a serious
gas streams containing water, there is a limitation for problem because even at low temperatures it can enter
long-time applications. the structure of zirconia and destabilize the modifica-
Solid-state reactions for silica are not restricted to tions. Appropriate phase transformations have been
phase equilibria with mullite, because the system is observed under hydrothermal conditions.35
complicated and characterized by metastable immis-
cibility gaps at temperatures significantly below the
1.26.5.3 Cordierite
eutectic value close to 1600 C. Alumina is a common
component in, and therefore can be dissolved by, Cordierite (Al3Mg2(Si5AlO18)) is a rock-forming
silicate slags. For the details of the process dependen- mineral, and the pure Mg-end member of the solid
cies, the reader is referred to example references.31 solution with Fe melts incongruently at 1460 C to
The corrosion resistance of alumina ceramics is yield mullite and liquid. It is a widely applied
strongly influenced by impurities because some of ceramic for soot filters or heat exchanger, because it
has a low thermal expansion coefficient and therefore advanced ceramics, and in recent times, increasingly
is very tolerant to thermal shocks and cycling. from the need to understand the behavior to obtain
High temperature corrosion of cordierite is known insulating parts within a semiconductor.
from treatments with molten salts (hot corrosion). Hydrothermal corrosion is significant only at tem-
peratures exceeding 500 C, even though initial reac-
tions may occur at 300 C. The hydrothermal
oxidation yields silica along with CH4, COx C.
1.26.6 Carbides
The attack is commonly an active corrosion, because
1.26.6.1 Boron Carbide silica is dissolved in H2O. The dissolution rate of silica
is the key to the kinetics of the process and can be
Hydrothermal corrosion of B4C proceeds at low
modified by increasing the pH of the water. The onset
pressures and C:H-ratios according to reactions of
of hydrothermal oxidation in silicon nitride, likewise
the type
forming silica, is at much lower temperatures. The
B4 C þ 8H2 O , 2B2 O3 þ CO2 þ 8H2 ½11 difference indicates protection by the formation of
carbon layers at temperatures below about 500 C,
in which CO2 is replaced by methane or carbon
which has been observed for SiC-based fibers45 and
monoxide under appropriate carbide/water ratios.
in halogen containing environments.46
In analogy with water vapor systems, reactions start
Silicides are thermodynamically often more stable
at temperatures as low as 250 C because no carbon
than SiC and a number of low-temperature eutectics
layer is produced.
exist with SiC. Therefore, metals can corrode SiC
Experiments on the attack by oxygen on pow-
and a compilation41 gives indications for reactions
ders,36 coatings, and sintered materials confirm that
with Al (750 C), Au (950 C), Bi (600 C), Cr
at low and moderate temperatures the oxidation
(1000 C), Cu (950 C), Co (1150 C), Fe (1000 C),
yields boric acid
Li (815 C), Mg (800 C), Ti (575 C), Mo (1200 C),
B4 C þ 4O2 , 2B2 O3 þ CO2 ½12 Nb (1300 C), Ni (500 C), Pd (600 C), Ta (1200 C),
and W (1500 C). Resistance to the attack was noted
and the kinetics are governed by oxygen diffusion
for Ag (962 C), Ca (1180 C), Cd (500 C), Na
through it. In the high temperature regime, starting
(350 C), Pb (815 C), and Sn (600 C).
at about 800 C, boric acid evaporates and the process
Most of those data were obtained in vacuum,
becomes paralinear with an overall loss at longer
where the wetting behavior, which is essential for
times. Evaporation is stronger in wet atmospheres,
the form and strength of attack, is different from
because the reaction
oxidizing conditions. It should also be noted that
B2 O3 þ 3H2 O , 2H3 BO3 ½13 SiC containing free Si reacts more readily than SiC.
Temperatures of possible reactions may be evaluated
dominates at temperatures below about 1000 C
with the aid of phase diagrams of the appropriate Si–
with high partial pressures. Boron carbide was inves-
metal system. In appropriate wear conditions, a reac-
tigated in steam because it is an important material
tion of SiC with a metal is therefore feasible.
in the nuclear industry. The dependence of the
For attack by salt melts (hot corrosion), the reviews
kinetics on partial pressures of water and flow speed
by Jacobson and coworkers47–49 are still fundamental.
was confirmed and modeled.37 It should be noted that
As explained in the appropriate chapter in this book,
many commercial bulk boron carbides contain impu-
the attacking salts are best represented by Na2SO4,
rities or additives, which strongly influence the
which forms in combustion environments by
process.38
2NaCl þ SO2 þ 0:5O2 þ H2 O ,
½14
1.26.6.2 Silicon Carbide Na2 SO4 þ 2HCl
Corrosion data collections for SiC-based materials A first condition for a window of effective hot corro-
are found in a number of handbooks and text- sion is given by the melting and dew points of Na2SO4.
books,39–42 are contained in specific conference pro- The melting point is 884 C, the dew point varies with
ceedings14,43,44 and a vast number of publications. pressure, S content of the fuel, and NaCl concentra-
The interest in the high temperature behavior of this tion in the atmosphere. In gases with NaCl > 1 ppm
material comes historically from its use as a refrac- and combustion pressures of 100 bar, the dew point
tory material, later from the structural engineering ranges between 1100 and 1200 C (from eqn.[14]).
The corrosive effect is not a reaction with the commonly causes pitting with a change in exposed
sulfate. It is due to the activity of Na2O from the surface and nonuniform attack. Pitting may not in all
decomposition cases be a secondary phase problem, because it can be
initiated by the immiscibility of two liquids.51 It seems
Na2 SO4 , Na2 O þ SO3 ½15
to be fairly difficult to model the damage by hot
and reaction with the protective SiO2 scale corrosion in a general way. One approach is to relate
the pit diameter to strength data, i.e., to treat them as
xSiO2 þ Na2 O , Na2 O ðSiO2 Þx ½16
critical flaw sizes in the Griffith equation.
Therefore, other compounds releasing alkali oxide on Hot corrosion protection has been investigated
decomposition, e.g., Na2CO3, also cause hot corrosion. using mullite coatings.17 Some temporary protection
Reaction [15] shows that a low P(SO3) value drives may even come from simple silica coatings.52
reaction [16] to the right side, typical for the external The corrosion of SiC by hot gases has been the
P(SO3) from the combustion of low-S fuels (<0.5%, subject of many studies. Active corrosion is a form
a condition met by many modern fuels) at tempera- of attack by gases, which do not form condensed
tures of about 1000 C. species with Si at the temperatures of interest. Thus,
However, corrosion starts at the melting point of halogens induce active corrosion, and calculated pres-
Na2SO4. Free carbon is often present in commercial sures of species such as SiF4 or SiCl4 over SiC will be
SSiC (=pressureless sintered silicon carbide) and as high as the concentration of F and Cl in the gas.
promotes the dissociation of Na2SO4 by making the These pressures will be significantly decreased
silicate melt more basic. Accordingly, very strong hot and the damage is less severe in the presence of
corrosion of SiC with free carbon is known.50 oxygen, as evidenced by a study on Cl2 corrosion.53
Even without free carbon hot corrosion of SiC Corrosion of pure SiC by H2 starts at tempera-
should occur as an internally governed process anal- tures above 1300 C, while grain boundaries and
ogous to studies51 on Si3N4, because SiC may act as secondary phases in sintered SiC are attacked at
the reducing agent to promote sulfate dissociation temperatures as low as 1000 C.
directly or by establishing an oxygen gradient in the The formation of silica is therefore essential for
silica scale consumed by reaction [16]. the stability of SiC. The boundary for active-to-
Hot corrosion tests are often only qualitative. Even passive oxidation is described in Figure 1. Wagner-
burner-rig methods have problems for quantifying the type models predict the transition for diffusion-related
attack. In terms of degradation of the mechanical situations and are relatively successful. From experi-
properties, it should be noted that hot corrosion mental evidence at temperatures of 1760 C,54
1.00.105 Wagner-model Passive
1.00.104
Oxygen partial pressure [Pa]
Transition region
1.00.103
Balat (PLS) Nickel (CC)
1.00.102
Gulbransen (SC)
1.00.101
1.00
Active
1.00.10–1 Song (SC)
1.00.10–2
1325 1375 1425 1475 1525 1575 1625 1675
Temperature (°C)
Figure 1 Active-to-passive transition of SiC thermal oxidation calculated (CC) from Wagner-type models and volatility
diagrams and experimental values on single crystals (SC) and SSiC (PLS), see Presser and Nickel.56
a secondary active-to-passive transition was formu- compatible with network oxygen diffusion in amor-
lated, at which the formed silica spalled because of a phous silica rather than molecular diffusion. It has
CO interface pressure exceeding ambient pressures. been argued that this is related to extrinsic impurities
Oxygen pressures in Figure 1 should not be con- from the experimental setups.59
fused with data from COx experiments, because CO In a plot compiling rate constants from a large
is a reactive and an oxidizing gas species for SiC41 number of data for fairly pure SiC (CVD, single
despite a low oxygen partial pressure. Therefore, crystals), including differing crystallographic faces,
active oxidation in CO is only observed at very high the situation is not so straight-forward,56 but a fairly
temperatures.55 high spread of parabolic rates is evident (Figure 2).
The silica formation in the passive regime at high Commercial qualities of sintered SiC often have
temperatures is in detail a complex process, which low impurity or additive contents and have rates as
includes several stages of amorphous silica formation low as pure materials, indicating extremely good
and its crystallization. A review will be given else- oxidation resistance up to very high temperatures.
where.56 At temperatures below about 1200 C a lin- This is certainly not really true for T above 1600 C,
ear–parabolic model in the sense of Deal and Grove10 because at very high temperatures the long-time
is necessary to explain the oxidation data, whereas at kinetics are not adequately described by the para-
higher temperatures a simple parabolic law is usually bolic law, the problem of active II oxidation is impor-
sufficient. tant and in addition bubble-and-pit formation, which
The process is still not well understood, because is usually attributed to additives containing boron
the rate constants are clearly dependent on the and/or aluminum, is likely to endanger applications
exposed crystallographic face of SiC,57 and at the (Figure 3). It should also be noted that bubble-and-pit
same time it is believed that it depends on the diffu- formation can occur after an ‘incubation period’ of
sion of oxygen through silica, because the kinetics are several 100 h at T > 1600 C, while oxidation studies
parabolic and activation energies fit this process.58 run usually only for times <100 h. Therefore long-
However, at temperatures above about 1300 C, time studies may be necessary for a true qualification
there are data sets indicating a high activation energy, of SiC at extremely high temperatures.
Temperature (°C)
1800 1600 1400 1200 1000 800
−22 10–9
Branch 1
Ea = 285 k J · mol –1 C-face 10–10
−24
n = 19 / R 2 = 0.98 Ea = 104 kJ · mol –1
n = 21/ R2 = 0.86
−26 10–11
Branch 2
Ea = 154 kJ · mol –1 10–12
kp [kg2 · m–4 · s–1]
−28
n = 25 / R 2 = 0.96
ln (kp)
10–13
−30
CVD SiC:
Ramberg et al. 2001 10–14
−32 Bartlett
Pampuch et al. Siliconized SiC:
Filipuzzi et al. Ramberg et al.
Sibieude et al. 2001 10–15
−34 Ogbuji & Opila Single crystal SiC:
Ramberg et al. - C-face Harris & call - C-face
Tressler et al. - C-face Si-face
Ramberg et al. - Si-face 10–16
−36 Fox Tressler et al. - Si-face
Ea = 270 kJ · mol –1
Narushima et al. - 1. phase Zheng et al. - C-face
Narushima et al. - 2. phase Zheng et al. - Si-face n = 16 / R 2 = 0.98
10–17
−38
4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5
10000/T [K–1]
Figure 2 Comparative plot of parabolic rate constants. The plot comes from Presser and Nickel56 where the data sources
are referenced.
1 mm
5 mm 10 mm
Figure 3 Bubbles at the surface of SSiC oxidized at 1500 C for 24 h under humid conditions and pits beneath
(1500 C, 50 h) revealed by HF-etching of oxidized SSiC from Ref. 100.
Furthermore, good resistance to oxidation by O2 Hydrothermal corrosion of silicon nitride starts at

or air does not guarantee equally good behavior in about 130 C because of the high solubility of silica in
complex industrial environments. Other components water under high pressure. The reaction
of the atmosphere or extrinsic impurities can change
Si3 N4 þ 6H2 O , 3SiO2 þ 4NH3 ½17
the behavior markedly and make a prediction diffi-
cult in cases of strong variations in temperature and does not passivate because the dissolution of NH3
gas composition, even though resistance of SiC-based makes the solution caustic, which strongly enhances
materials to other aggressive environments, e.g., such silica solubility. Therefore, reaction kinetics are lin-
as SO2 or H2S, has been reported.14 ear and active corrosion prevails.
For such environments, case studies give an indica- The hydrothermal corrosion of the grain bound-
tion but are not unambiguous.14 SSiC was reported to ary phase of Si3N4 ceramics can be very strong and
resist flue gases of aluminum-remelting furnaces up to pit formation was observed at 300 C and 8.6 MPa.62
1000 h without much change in strength in one study From experiments on hot isostatically pressed silicon
but had problems in others. The strength of SSiC nitride (HIPSN), which were free of additives other
was unchanged in nitrogen-based gases of the system than silica, we know that preferential hydrothermal
N2–H2–CO, but decreased up to 50% in endothermic leaching of silica as the grain boundary phase can
gases at 1200–1300 C in long-time experiments. cause grain dislodgment.63
Water enhances the oxidation of SiC only slightly, A compilation of data for the reactions of Si3N4
but the transport of extrinsic impurities because of with metals and alloys has been given in Ref. 64. The
water can strongly enhance oxidation and the limita- interaction is important for bonding to metals; often
tions in applications with high gas flow rates because silicides or other ternary compounds are formed. The
of vaporization of the silica scale inducing paralinear interaction of the related SiAlONs (silicon nitrides,
behavior are known.60 which incorporate alumina and yttria in their lattice)
and alloys was found to be similar. In inert gas atmo-
sphere or under vacuum, Si3N4 reacts readily with
1.26.7 Nitrides most transition elements below 1000 C and with
many other metals at various temperatures, while
1.26.7.1 Silicon Nitride
protection by silica in oxidizing conditions is effec-
Silicon nitrides are often discussed together with tive for some conditions.64
silicon carbides, because both materials are silica Hot corrosion of Si3N4 attacks the silica scale and
formers. For this reason, data collections and text- the behavior is therefore similar to that of SiC, as
book sections are identical to those already cited for confirmed by studies at NASA labs.48,65 However,
carbides.13,14,39,40,42–44 For an additional, detailed the initial process may be governed by intrinsic fac-
study see Ref. 61. A typical difference from the sili- tors, because surface tension differences between sul-
con carbides is the much higher additive content of fate and silicate melts result in a fractal pattern of
almost all commercial sintered silicon nitrides. Only hillocks of silicate melt beneath the sulfate. This was
liquid-phase sintered SiC types (‘LP-SiC’) have com- reported to enhance local corrosion and induce pit
parably high amounts. formation. The process slows down with time and
therefore Si3N4-based ceramics can serve well under oxidation resistance for pure Si3N4 compared with
hot corrosion conditions. Even though there is a that for SiC at temperatures of about 1500 C is evi-
degradation in mechanical properties similar to that dent and contrasts with the behavior of sintered Si3N4.
of SiC,21,66 the remaining strength after hot corrosion This is plain from Figure 4, which shows calculated
of the nitride is higher than that of SiC. critical times tc for 100 mm penetration on pure and
Active corrosion of Si3N4 ceramics by hot gases is sintered nitride and carbides, assuming parabolic
often stronger than comparable attack on SiC, growth and using the data compiled earlier.13
because the grain boundary phase removal causes The presence of grain boundary phases is the
grain dislodgment and inhomogeneous internal cor- main reason for the pronounced decrease in oxida-
rosion. In particular, halogens create high partial tion resistance. In detail, a number of processes,
pressures of Si-bearing species and accordingly including the leaching of grain boundary constituents
strong active corrosion, but small amounts of oxygen to the scale, crystallization, internal oxidation, bubble
or water can significantly decrease the effect of Cl2 or formation, and scale cracking, occur and vary with
sulfidizing environments by silica formation. The amount and exact composition of the additives. The
active-to-passive transition occurs at similar condi- discussion of such processes is referenced in the text-
tions as the one for SiC (Figure 1). books cited at the beginning of this chapter.
From thermodynamic considerations67 the passive Again, it must be noted that the data in Figure 4
oxidation of pure Si3N4 should involve a two-step should serve as an indication only, because the values
oxidation were calculated assuming a truly simple parabolic
behavior, which is often not really fulfilled.
2Si3 N4 þ 112O2 , 3Si2 N2 O þ N2 ½18
Si2 N2 O þ 112O2 , 2SiO2 þ N2 ½19
forming a duplex layer in an oxygen potential gradi- 1.26.7.2 Boron Nitride
ent. A definitive proof for a sublayer of Si2N2O was In hydrothermal conditions, BN is attacked at tem-
only provided for pure nitrides,68 but a superior peratures as low as 150 C, following linear kinetics.69
Hexagonal BN is often applied in the production of
the metal and for thermocouples in Fe alloy melts,
6
High-purity SN which is ample evidence for a good corrosion resis-
HP/SSN tance against metal melts and slags at much higher
5
SSiC temperatures, namely alkali halides, lithium borate,
Powder & SC and cryolite. This is due to the poor wetting behavior
4
CVD of such melts on BN.
Only molten alkali carbonates and hydroxides
3
decompose boron nitride by the reactions of the type
BN þ K2 CO3 , KBO2 þ KOCN ½20
Log (tc[a])
2
The product of the oxidation of boron nitride at
1 atmospheric and low pressures is boric acid
0 4BN þ 3O2 , 2B2 O3 þ 2N2 ½21

and therefore the material has a similarity to boron
−1 carbide. The melting point of boric acid at 410–
450 C keeps the formed oxide liquid at high tem-
−2 peratures. Accordingly, at temperatures exceeding
700 C, there is paralinear oxidation with the loss of
−3 boria by evaporation.70 Water enhances the evapora-
800 1000 1200 1400 1600 1800 2000 tion process, because the partial pressure of hydrogen-
Temperature (°C) bearing species is high71 and the oxidation can be
Figure 4 Logarithm of calculated critical times in years for described as
100 mm penetration for silicon nitrides (full symbols) and
silicon carbides (open symbols) vs. temperature from BN þ 3H2 O , H3 BO3 þ NH3 ½22
compiled parabolic rate constants.13
1.26.7.3 Silico-Carbonitrides TiB2 is significant above 700 C and shows parabolic

kinetics at those temperatures.82 In powder experi-
The precursor-derived amorphous silico-carbonitrides
ments, an onset of oxidation was found at 400 C and
have high thermal stabilities, a boron-containing type
TiBO3 was an oxidation product at low tempera-
reportedly up to 2000 C.72
tures.83 At 950 C, the evaporation of B2O3 becomes
Crystallization may induce detrimental reactions
significant and paralinear oxidation occurs. How-
as was shown for pure carbontirides at 1400–
ever, unlike for boron carbide and nitride, the kinet-
1500 C.73 Their nominal stoichiometry has C in
ics become parabolic again at higher temperatures
addition to Si3N4 and SiC and therefore the reaction
(>1200 C). This is explained by the fact that, when
Si3 N4 þ 3C , 3SiC þ 2N2 ½23 B2O3 is completely evaporated, the process is con-
trolled by diffusion through the titanium dioxide
takes place at about the thermodynamically pre-
film.
dicted temperature (1405 C at 1 bar N2). Oxidized
Impurities and metal binders were found to
samples without such a reaction74 were observed to
improve the oxidation resistance of TiB2, because
survive because of an internal pressure buildup of
the rutile scale sinters and becomes dense.84
several bars.
The use of TiB2 as a component of composites and
As suggested earlier,75 carbonitrides are suscepti-
their oxidation and corrosion behaviors are the sub-
ble to hot corrosion because the excess carbon will
ject of a large number of papers. The behavior can be
tend to make a melt basic. In terms of oxidation
fairly complex because other phases such as alumi-
resistance, the data76–78 suggest that they have very
num borates become stable and influence the
low oxidation rates at temperatures up to 1600 C.
behavior.
The problem of bubble formation seems to be avoid-
able by Al incorporation in the ceramic. However, the
penetration rates are probably not much better than
1.26.9 Ultrahigh Temperature
those for pure silicon nitrides, and the problem of
Ceramics
paralinear behavior in fast-flowing gases is valid for
these materials as well.
With the realization of the limits of silica-forming
ceramics in fast-flowing wet atmospheres, the search
for so-called ‘ultrahigh temperature ceramics’ was
1.26.8 Titanium Compounds started. The classes looked at were either the
polymer-derived ceramics85,86 or those forming very
TiC and TiN show a very wide range of stoichiome- refractory oxides, particularly ZrO2 or HfO2. How-
try and form solid solutions in the Ti–C–N system. ever, the precursor-derived ceramics are silica for-
Thermodynamic equilibria in water–carbide sys- mers, and therefore at least at temperatures above
tems79 indicated that carbon, corresponding oxides, 1650 C the problems of silica are inherent. The
CH4, CO2, CO, and H2 are the main products of zirconia and hafnia formers have the problem of fast
hydrothermal corrosion of metal carbides in the tem- ionic oxygen transport and thus materials such as
perature range 300–1000 C. No protective oxide ZrB2 or HfB2 oxidize extremely fast87 and are only
layers were formed upon hydrothermal corrosion of candidates for very short-lived applications.
carbides of transition metals, but Raman spectra from Together with the fact of decreasing mechanical
hydrothermally treated TiC powders demonstrated properties for all metallic materials at very high tem-
the presence of graphite. peratures, the only current vision is the development
Oxidation of TiN powders, films, and sintered of composites, which are protected from oxidation
bodies80,81 showed that oxidation occurs at very low by some sort of coating. Therefore, environmental
temperatures and becomes significant at about 500 C, barrier systems are a subject of current investigation.
with the formation of TiO2. Scales from titania are The old dream of a carbon-based system with its high
usually not protective at temperatures above 800 C, temperature properties protected by coatings is still
which limits the use of monolithic titanium com- alive and investigated repeatedly with new approaches
pounds at high temperature. and systems. It must still be regarded as one of the
TiB2 forms titania and boria and is known pri- most promising ways to maintain a strong system at
marily for coatings and composites. Oxidation of very high temperatures!
1.26.10 Refractories 1.26.11 Chemical Dissolution

1.26.11.1 Diffusion/Blast Furnace
Formed and unformed refractory products have a
wide process engineering spectrum of applicability In a shaft-shaped, double conic built blast furnace,
and therefore satisfy various demands and reciprocal iron ore is constantly reduced to pig iron by carbon
actions. The following text will present the corrosion and carbon monoxide, respectively. The liquid iron
of refractory materials with typical examples of the drips into the hearth of the blast furnace, gathers on
iron and steel industries. The effecting part- the ground, and is tapped periodically and in a batch
mechanisms can be compared with other industrial wise manner into transport ladles. The fluid slag in
sectors and can be divided as follows. the hearth floating on top of the pig iron fixes the
impurities contained in the ore and is extracted
1.26.10.1 Chemical Dissolution continuously.
The hearth, which is water cooled via its outer
The driving force is the existing thermochemical wall, is lined with carbon and graphite containing
nonequilibrium between the two involved phases bricks. The furnace lifetime is typically 10 years or
(liquid steel/slag/atmosphere – refractory lining) of more with a wall thickness of up to 2 m. In addition to
a system. The thermodynamic stability of a phase high heat conductivity, another advantage of this
(refractory material) is described by the free enthalpy lining is the nonwettability of the graphite by slag.
of formation DG and is heavily affected by the vari- No infiltration takes place.
ables of temperature, pressure, and concentration. Furthermore, the iron melt approaching from
Besides the dissolution itself, reduction and oxidation above is saturated with carbon, which strongly
processes as well as vaporization and evaporation decreases attack of the lining. The saturation with
play essential roles. carbon is also due to the presence of a bed of coke
in the hearth, which is highly porous and filled with
1.26.10.2 Erosion slag and pig iron. In addition elements such as Si and
Mn accompanying the pig iron decrease the satura-
The driving force is the flow of melts by gas flushing,
tion concentration of carbon.88
vacuum treatment, tapping, and filling the metallur-
Any iron melt not saturated in carbon causes a
gical reactor respectively, and continuous casting and
linear dissolution rate for a unidirectional water-
gases charged with solids. During the first process, a
cooled carbon plate described by the equation:
mostly predamaged structure is washed out. During
the latter, the damage is affected by solid particles Dx r CS C0
¼ b Fe ðcm s1 Þ ½24
impacting onto the refractory lining because of their Dt rC 100
inertia, and which break off the material because of
surface disruption. where b is the mass transfer coefficient (cm s1); rFe,
the density of liquid iron (6.67 g cm3), rC, the den-
sity of carbon/graphite (2.25 g cm3); CS, the satura-
1.26.10.3 Mechanical Wear tion concentration of carbon in liquid iron (1500 C:
Thermal shocks and steep temperature gradients (e.g., 5.16 wt %); C0 is the concentration of carbon in liquid
during heating up, change of temperature when fill- iron.
ing a ladle, etc.) cause thermal stresses and sometimes The mass transfer coefficient b is given by
structure changes. Furthermore, the handling in steel
DC 4DC v 1=2 DC 1=6
plants (e.g., filling a converter with scrap, the cleaning b¼ ¼ ðcm s1 Þ ½25
process with skull breakout and removal of slag, the d phl u
filling and transport of a ladle, etc.) cause wear by where DC is the diffusion coefficient of carbon in
mechanical contacts. liquid iron (104 cm2 s1); d, the boundary layer of
These three mechanisms are superimposed and diffusion (cm); v, the flow velocity of the melt
accelerated by infiltration, i.e., weakening the struc- (cm s1); hl, the diameter of the convection loop
ture of the refractory materials by the ‘inner’ dissolu- (assumption: 1 cm); u is the kinematic viscosity of
tion of the matrix. In practice, chemical dissolution, liquid iron (103 cm2 s1).
erosion, mechanical wear, and infiltration appear A laminar flow is predominant (free convection,
together with varying degrees of importance. Re 2100) because of low temperature differences at
the phase interface (DT < 1 K). The speed of the carbon-containing refractory material, there is a
convection loop moving at the phase interface can slightly increased carbon concentration in the region
be estimated by using the Reynolds number Re : close to the three-phase interface, which makes the
Re u interfacial tension sII (Figure 5) higher than sI.
v ðcm s1 Þ ½26 A part of the carbon dissolved in the slag reacts
hl
with the steel, thereby forming carbon monoxide
Using the above values, the flow velocity is (CO) and this causes a locally decreased oxygen
u 21 cm s1. Inserting this upper limit in eqn. [25], content at point II. The corresponding increase in
a mass transfer coefficient b of 0.035 cm s1 is the interfacial tension decreases the capillary activity
obtained. For a temperature of 1500 C, the carbon of the slag and the steel tends to convect from
concentration in the liquid iron will be 5.15 wt %.89 point I to point II in order to decrease the interfacial
Then, eqn. [24] yields the linear dissolution rate: tension there.
Dx 6:67 5:16 5:15 In a lot of cases, no infiltration occurs, because the
¼ 0:035 corrosion progress is faster than the speed of infiltra-
Dt 2:25 100 ½27
tion. To estimate the extent of corrosive wear, one
¼ 105 cm s1 or 0:04 cm h1 may use the approximation of eqn. [24]. The mass
At such a rate, the lining in the hearth of the blast transfer coefficient b is estimated well by eqn. [25].
furnace would be fully worn out within a few hours of The flow velocity of the steel and slag along the phase
operation. However, because of the above-described interface is then90:
mechanisms, the presence of a coke bed and the
hg ds ds
decrease in the carbon saturation by the additional vM ¼ grad C þ grad T ðcms1 Þ ½28
4S dC dT
elements present in the pig iron, a long service life of
up to 2 years is achievable. where hg is the characteristic length of the corrosion
groove (assumption: 1 cm) and S is the dynamic
viscosity of steel (1600 C: 0.04 g cm1 s1).
1.26.11.2 Marangoni Convection/SEN In this equation, the change in the interfacial
Preferential wear along the three-phase steel/slag/ energy of the steel/slag interface in relation to the
refractory material interface, such as at a submerged oxygen content in the liquid steel is estimated as
entry nozzle (SEN) used in a mould or in a ladle, can ds 200
be explained by the temperature and particularly by ¼ 1 ½mN m1 wt%1 ½29
dC C½O
the concentration-dependent interface phenomenon
of Marangoni convection (Figure 5). The refractory the approximated concentration gradient of the oxy-
material reacts with the steel and the slag. If it con- gen along the phase interface is
tains carbon, it dissolves in the steel which is con-
strained by the oxide in the refractory material. In
I
C½O C½O
II
grad C ¼ ¼ 0:001 wt% cm1 ½30
turn, the dissolution of the oxide in the refractory hg
material into the slag is constrained by the graphite.
the change in the interfacial energy in relation to the
Nonetheless, 0.1 wt % of carbon dissolves into the
temperature is91
slag. Because of this reaction, a local concentration
gradient develops at the steel/slag interface. For ds
¼ sT ¼ 0:2 mN m1 K1 ½31
dT
and the approximated temperature gradient because
Slag of the heat removal to the refractory material is
T I T II
gradT ¼ ¼ 10 K cm1 ½32
sI h
sII ⬎ sI With an assumed oxygen content CI[O] = 0.004 wt %
sII (40 ppm) in eqn. [19], the insertion of the values from
Steel eqns. [30]–[32] into eqn. [28] yields a flow velocity
Refactory vM of 325 cm s1.
Ca. 1 mm
Assuming that the back-flow velocity along the
Figure 5 Advanced wear by Marangoni convection. refractory material/slag interface is about the same
as vM, the thickness of Prandtl’s flow-boundary-layer 1.26.11.3 Oxidation/Ladle

dPr can be easily calculated:
The crude steel coming from the melting aggregates

Sl h 1=2 (converter or electric arc furnace) is partially already
dBL ¼ ðcmÞ ½33
rSl uM post-treated during tapping into the ladle and then
again within the ladle. This treatment, which is
where Sl is the dynamic viscosity of the slag referred to as secondary metallurgy, includes the
(2 g cm1 s1, 0.2 Pa s1) and rSl is the density of the process steps92 of degassing, decarburization, desul-
slag (2.7 g cm3). phurization, dephosphorization as well as removal of
The thickness of the Nernst’s diffusion-boundary- trace elements, fine adjustment of alloying contents,
layer dN is then deoxidization, inclusion transformation, improve-

Sl 1=3 ment in the purity levels as well as the solidifying
dN ¼ dPr ðcmÞ ½34 structure, adjustment and homogenization of temper-
rSl DSl
ature and chemical composition.
where DSl is the diffusion coefficient of the corrosion To minimize the thermal stresses and the cor-
determining component (1650 C: 1
107 cm2 s1) responding wear because of the actual temperature
With these material values, we get gradient when filling with crude steel (T up to
dPr = 1.51
102 cm and dNi = 1.67
104 cm. 1800 C), the cold ladle lined with MgO–C bricks is
Accordingly, the mass transfer coefficient b is calcu- preheated for 2 hours in the steel mill. A newly lined
lated by ladle does not have a slag layer protecting the bricks
DSl against oxidation of the binding carbon (graphite,
b¼ ðcm s1 Þ ½35 5–15 wt %). As an instantaneous result of this pre-
dN
heating procedure, the graphite oxidizes. The carbon
The result is b = 6
104 cm s. Assuming a slag bonding and its advantageous properties such as the
whose starting content of Al2O3 is 13 wt %, and good thermal shock resistance and the nonwettability
whose saturation concentration at 1650 C is in the by slag are lost. At the same time, an enormous pore
range of 60 wt % A12O3, the wear of a corundum- volume (about 35 vol %) is opened. Steel and slag can
enriched, graphite-containing refractory material easily infiltrate into the microstructure and therewith
(rR = 3.3 g cm3) due to Marangoni convection lead to ‘inner’ dissolution.
approximately amounts to The temporal progress of the reaction
2:7
vcorr ¼ 360
6
104 ð60 12Þ 2C þ O2 ! 2CO ½37
3:3 ½36
¼ 8:5 mm h1 in a MgO–C brick can be described by the following
differential equation:
which is in accordance with laboratory experiments.
The question is how to decrease this extremely dxi n0 ni
severe corrosion. The process cannot be influenced ¼ 2VC DP ½38
dt x0 xi
easily. In the case of the continuous casting process,
the concentration gradient and the temperature gra- The sharp transition from the bright, decarburized
dient are nearly fixed. The temperature and concen- outer zone to the black core suggests a topochemical
pffiffi
tration dependency of the interfacial energy are reaction which is controlled by diffusion ( t -law).
natural values, and the same applies for the dynamic The determining factor is the pore-diffusion coeffi-
viscosity of the steel. Therefore, the only way left is to cient DP93
use a refractory material that is as chemically inert as
possible. DP ¼ DO2 gpc ðcm2 h1 Þ ½39
This is, for instance, achieved by the implementa-
tion of a sleeve made of ZrO2 at the three-phase with
interface of a SEN. According to eqn. [24], the slag DO2 ¼ 2:7
102 T 1:79 ðcm2 h1 Þ ½40
and the mould powder, respectively, can be enriched
with Al2O3 up to a certain amount, in order to main- and
tain the difference in the saturation concentration as pffiffiffi
small as possible. g ¼ 1:84
pc 0:3 ½41
where DO2 is the diffusion coefficient of oxygen; T, because of erosion. The reason for this is the segre-
the temperature (1373 K); g, the labyrinth factor; gation of a solid loaded gas flow because of the mass
and pc is the open porosity of the decarburized inertia of its solid particles. These particles impact
zone (0.35). mostly in areas of flow redirection (elbow) onto
The insertion of these values results in the refractory lining, whereby a cyclic mechanical
DP = 3081 cm2 h1. The solution of the differential stress occurs at permanently changing surface areas.
equation [38] for an infinite plate is given by14 For brittle materials, such as refractories, these alter-
nating local impacts and corresponding multiaxial
ð1 fÞ2 ¼ t ½42
tensile and compression loads finally cause material
with removal by surface disruption. It should be noted that
xi the resulting wear is always a system characteristic
f¼ ðxi ¼ x0 at t ¼ 0Þ ½43 and is influenced by numerous factors.
x0
At room temperature, the testing of refractory
where xi is the position of the reaction front in the plate materials with respect to their wear by perpendicular
(cm) and x0 is the total thickness of the plate (cm). impacting solid particles is performed according to
t ASTM C 704. For this purpose, 1000 g of SiC particles
t¼ ½44
t0 of predefined granularity are accelerated by a pressur-
where t is the time (h) and t0 is the time for total ized air flow within 450 15 s and shot over a distance
decarburization of a plate with x0 (h) and of 203 mm perpendicular onto a surface of a test sam-
ple. The resulting wear volume (in cm3) is determined
x02 by weighing the geometrically well-defined sample.
t0 ¼ ðhÞ ½45
2zVC DP ðn0 ni Þ However, the results thus achieved can hardly be
where VC is the mole volume of graphite transferred to the temperatures applied in operation.
(5 cm3 mol1), z, the geometry factor (infinite plate: That is why a special experimental setup was developed,
z = 2); n0, the oxygen content at the plate surface which is as far as possible in accordance with ASTM C
(20 vol %; 2
106 mol cm3); and ni is the oxygen 704 and whereby the wear by impacting particles can be
content at the reaction front (ni n0, ni ! 0). determined at temperatures of up to 1400 C. The main
A MgO–C brick with x0 = 10 cm would be totally item of this experimental setup is a ring burner by which
decarburized within t0 = 1014 h. The time t required the pressurized air flow, the blasting abrasive (SiC par-
to decarburize 1 cm (xi = 9 cm) of the brick results as ticles), and the test sample are heated to the relevant test
follows: temperature. It has been shown that the wear volume of
heterogenic refractory materials decreases with an
9 2 increase in temperature, which correlates quite well
t¼ 1 ¼ 0:01
10 ½46 with the change in thermomechanical properties.
While brittle–elastic deformation behavior occurs
with t ¼ t t0 ¼ 0:01
1014 h ¼ 10:14 h at room temperature, viscoplastic deformation behavior
Numerous experiments demonstrated that the process is dominant at high temperatures, where molten phases
is in fact controlled by diffusion and is extensively from the binder phase are present. The local stress
blocked by specific additives, such as >10 wt % ZrB2. peaks, which are generated by the impacting SiC parti-
However, the addition of 3 wt % metallic aluminum as cles, are decreased by plastic deformation and by stress
an antioxidant has hardly any effect, since the porosity relaxation of the binder phase, respectively. In this con-
of the decarburized refractory material is high, and text, particularly wear-resistant grains (SiC, corundum)
during the coke reaction at T 800 C, the metallic incorporated inside the microstructure show no wear-
aluminum is consumed by the formation of Al3C4, and reducing action.94 Moreover, these grains break upon
above 1200 C by spinel formation. impact of a SiC particle because of their poor ductility.
1.26.12.2 Flowing Melts/Converter

1.26.12 Erosion
During filling of a torpedo ladle or a converter
1.26.12.1 Blast Wear/Pipes
with pig iron, or of a ladle with crude steel, the
Pneumatic conveying systems and pipes in thermo- melt often ‘falls’ at about 2 m. Because of the high
technical aggregates are subject to severe wear flow speed, the lining in the collision area of the melt
stream, which is possibly predamaged by chemical Both equations were derived from nonlinear frac-
dissolution, thermal shock stress, or infiltration, is ture mechanics, which is necessary when dealing with
washed out. However, a quantitative prognosis is not refractory materials, since they have a very heteroge-
possible. The same applies for the circulation current neous structure (grains, binder phase, pores).
generated during gas stirring treatment within a con- Together, both crack resistance parameters describe
verter or in a ladle. In connection with the high flow the so-called structure flexibility. From eqns. [47] and
velocities of the melt, cavitation erosion might play a [48], it is plain that an increase in the required crack
significant role, but this has not been examined fur- energy and the crack propagation work GF with both
ther in the field of steel production. unstable as well as ‘quasistationary’ crack propagation
is advantageous in order to reach a high value for
the respective resistance parameter and, therefore, to
1.26.13 Mechanical Wear achieve a high thermal shock resistance.
This can be obtained in structures with coarse
1.26.13.1 Thermal Shock Resistance/Ladle
grains in a fine-grained matrix, where entanglement
In practice, mechanical wear as an instantaneous of the grains increases the crack length. The modulus
result of a poor thermal shock resistance is commonly of elasticity, however, has the opposite effect, which is
found. During the filling of a ladle with crude steel, dependent on the mechanism of the crack propaga-
the lining is heated up rapidly from 1000 to 1800 C tion alone. With unstable crack propagation, the
within a few minutes. High tensile stresses develop in crack itself and the incipient crack arising must be
the surface, which frequently results in a single incip- as big as possible to reach the area of the ‘quasista-
ient crack, which can grow catastrophically. Off- tionary’ crack propagation without reaching the cata-
bursts occur in the surface region of the lining, strophic breakpoint. In the area of the stationary crack
which is also referred to as ‘spalling.’ This mechanism propagation, the modulus of elasticity should be as
is promoted by a modified structure of surface areas small as possible, which calls for an optimization of
containing slag infiltrations. the structure concerning both crack propagation
An assessment of various refractory materials mechanisms. A crack then gradually propagates with
concerning their thermal shock resistance can be per- every higher and more critical temperature difference.
formed by the quick quenching of a 950 C test sample In spite of great experimental endeavor, no defi-
by blowing pressurized air (DIN EN V 993-11) onto its nite connection between the resistance parameters
surface, or by dipping the test sample into water (DIN and the standard tests for judging the thermal shock
51068-1). Then tensile stresses occur, starting from the resistance has been found.96 At the moment, values
surface, which result in a large amount of incipient must be inserted into the aforementioned equations,
cracking. Usually, a catastrophic break does not occur. which have been measured isothermally.
No isothermal boundary conditions can be found Critical heating rates may be calculated according
in these cases and therefore a theoretical treatment to the example of a hot plate made out of sintered
becomes difficult. Modeling of ‘quasistationary’ crack corundum (99% Al2O3) at 1000 C:
propagation is possible, but during rapid heating up ec ð1 2mÞ 16a
of a ladle lining, unstable crack propagation occurs. T_ ¼ ðK s1 Þ ½49
3a x2
Therefore, characterization of the thermal shock
resistance of refractory materials uses the resistance where ec is the critical elongation (0.001); m, the
parameter R during unstable crack propagation and Poisson ratio (0.21); a, the thermal expansion coeffi-
Rst during stationary crack propagation95: cient (1
105 K1); a, the thermal diffusivity
(1
106 m2 s1); and x is the thickness of the plate
GF E
R0000 ¼ L ðmmÞ ½47 (0.02 m).
s2c Inserting the values results in T = 0.77 K s1. If the
rffiffiffiffiffiffiffiffi sample cools down/warms up faster, a breakage
GF (crack) occurs.
Rst ¼ ðm1=2 KÞ ½48
a2 E
where GF is the crack energy ( J m2); E, the modulus
1.26.13.2 Carbon Bursting/Blast Furnace
of elasticity (GPa); sc, the critical stress (MPa); a, the
thermal expansion coefficient (K1); and L is the In industrial processes (chemical plants, blast furnace
characteristic length (mm). shaft) in the temperature range from 400 to 800 C,
carbon bursting of refractory material in atmospheres process, the dissolved carbon is oxidized to a certain
containing CO is known. This phenomenon is due amount. Iron and those elements accompanying the
to the temperature dependence of the Boudouard pig iron such as P, Si, and Mn are oxidized and form a
equilibrium: slag. Temperatures of above 2000 C are reached by
strong exothermal reactions. It is, therefore, prefera-
2CO $ CO2 þ C ½50
ble that a large part of the crude steel charge is
When crystalline graphite precipitates in refractory replaced by scrap steel, whose melting offers a cool-
materials, it can create an enormous crystallization ing effect for the melt.
pressure, capable of bursting the structure. This pre- The lining of an emptied converter is 1200 C at
cipitation is said to be catalytically promoted by the hot bottom and is therefore under both thermal
metallic iron, hematite (Fe2O3), and magnetite and mechanical stresses upon refilling with scrap
(Fe3O4). In examination of the resistance to CO steel.
according to ASTM C 288, refractory materials are To estimate the procedures during charging, the
exposed to a pure CO atmosphere after a thermal transmitted mechanical energy of a falling scrap
pretreatment at 500 C for 200 h. piece mgh onto an impact area A is equated with the
A small amount of hydrogen (3%) accelerates the required crack energy AnGF for a number of n cracks.
carbon precipitation and thus we assume a catalytic The prevailing impact area and the number of pro-
effect of hydrogen. After the test, a visual evaluation of duced cracks are unknown, since the position of the
the refractory material leading to a classification scrap piece at the moment of impact is incidental and
concerning its resistance to CO is performed. The the shock wave spreads into the floor. It is assumed
resulting crystallization pressure (p 1) can be calcu- that a scrap piece falling from a height h = 4 m with a
lated using the following equation: mass m = 100 kg influences a surface area A = 0.1 m2.
The measured crack energy of the MgO–C material
DG 0 RT ln KC
p1¼ ½51 at 1200 C is GF = 500 J m2. The number of
R
DVC R emerging cracks is
where △G0 is the free enthalpy in the standard mgh 100
9:81
4
state ( J/mol); R, the universal gas constant (8.314 J n¼ ¼ 80 ½54
AGF 0:1
500
mol1 K1); R*, the gas constant (84.8 atm cm3
mol1 K1); T, the temperature (K); KC, the equilibrium and accordingly MgO–C material may break out of
constant; and DVC is the mole volume of graphite the surface. In the following refining process, tem-
(5 cm3 mol1). peratures of above 1800 C develop, so there is also a
At 500 C (773 K), magnetite is reduced to metal- good chance that cracks will heal again. To improve
lic iron, if the volumetric contents of CO and CO2 are the structural flexibility of such stressed components,
equal; the formation of graphite results for the earlier metallic aluminum powder and steel needles are
stated reaction [50] with added to the MgO–C material.97
PCO2
K ¼ 2 ¼2 with aC ¼ 1 ½52
PCO
1.26.14 Infiltration by Steel and Slag
DG 0 ¼ 169452 172:6T ; 1.26.14.1 Basic Equations
½53
1
DG ð773KÞ ¼ 35522 Jmol
0
The contact of a pure oxide and a graphite-
Inserting all values in eqn. [51] results in a crystalli- containing refractory material with liquid steel or
zation pressure of 63 375.3 atm, corresponding to slag in a thermodynamic equilibrium is described
62 150 bars, thereby destroying the structure of the by wetting theory. The vectors of the surface tensions
refractory material. of the three involved phases acting at the point of
contact of all three phases are in a mechanical equi-
librium state:
1.26.13.3 Destruction due to Impact/
Converter sR sS cos y sR=S ¼ 0 ½55
In the converter, crude steel is produced by blowing where sR is the surface tension of the refractory
oxygen onto and/or into the molten pig iron. In this material (mN m1); sS, the surface tension of the
steel or slag (mN m1); sR/S, the interfacial tension 1.26.15 Examples
between refractory material/steel or refractory mate-
rial/slag (mN m1); and y is the wetting angle ( ). 1.26.15.1 Slag/MgO
The interactions of the three phases involved are A MgO brick in contact with a slag with a minimal
characterized by the wetting angle of contact: surface energy at 1600 C (sS = 0.51 N m1, S = 1 Pa s)
sR sR=S will show a penetration depth of l cm within a pore
cosy ¼ ½56 d = 50 mm according to eqn. [59] after an interaction
sS
period of 10 min. The dependence on the wetting
and angle disregarding a slag-column above the pore entry
is outlined in Table 1. Note that this estimation is
y > 90; if sR sR=S < 0 ðnon-wettabilityÞ obtained with simplifications, namely that there is nei-
y ¼ 90; if sR sR=S ¼ 0 ther chemical interaction between the slag and the
y < 90; if sR sR=S > 0 ðwettabilityÞ refractory material nor a temperature gradient within
the inner structure.
The wetting angle of contact y and the surface
tension ss are the decisive values for the infiltration 1.26.15.2 Steel/MgO
of a capillary or pore by steel or slag, which can be Liquid steel (1600 C: sS ¼ 1.5 N m1, S ¼ 6.8
103
determined experimentally. The equilibrium of Pa s, rS ¼ 7.700 kg m3) and a MgO brick show
forces in a partly filled pore is then: completely different infiltration because of the wetting
angle of y ¼ 120 . According to eqn. [57], steel can
4sS cos y
rS gh ¼ ½57 only penetrate into open pores, if their diameter is
d sufficiently large. For this, Table 2 gives the infiltration
where rS is the density steel/slag (kg m3); g , the
acceleration of gravity (m s2); h, the height of Table 1 Penetration depth of a slag in relation to the
wetting angle of contact (d = 50 mm and t = 600 s)
steel-/slag-column above the pore entry (m); d is
the pore diameter (m). ( ) l (cm)
If the wetting angle of contact y is smaller than
0 1.96
90 , capillary ascension is present. For instance, a 10 1.94
liquid slag penetrates the pore or capillary system 20 1.90
of an oxidic refractory material without the effect of 30 1.82
any external force. In the case of liquid steel and 40 1.71
oxide refractory material, capillary depression is 50 1.57
60 1.38
present, since the wetting angle of contact is y > 90 . 70 1.14
Thus, an external (ferrostatic) pressure is necessary 80 0.81
to allow the liquid steel to penetrate into the pore 89 0.26
or capillary system. The same is true for a slag that 90 0
does not wet the graphite binder phase of the oxide
refractory material. Table 2 Infiltratable pore diameter d and penetration
The infiltration speed of the slag or steel into depth l in relation to the steel pool depth h
the structure results from the Hagen–Poiseuille
h (m) d (m) l (m). d = 200 m
equation98:
0.25 172 0.73
dl d 2 4sS cos y
¼ þ rgh ½58 0.50 86 2.10
dt 32l d 0.75 57 2.80
1.00 43 3.48
Integration leads to the penetration depth l, in rela- 1.25 34 4.00
tion to the time t: 1.50 29 4.46
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1.75 25 4.87
d sS cos y d 2 rS gh 2.00 22 5.25
l¼ þ t ½59 2.25 19 5.61
4S 16S
2.50 17 5.94
where l is the penetration depth (m); S, the dynamic 2.75 16 6.26
3.00 14 6.56
viscosity of steel/slag (Pa s); and t is the time (s).
pore diameter d in relation to the depth of the steel order to form secondary MgO that precipitates at the
pool, h (ferrostatic pressure). As we can see, the greater interface (heterogeneous nucleation) and, thus,
the depth, the higher the risk of infiltration of smaller restrains further infiltration into the structure.
pores. 4. The carbon solubility in liquid iron is also
Considering a depth, h, the penetration depth of affected by higher aluminum levels and the reaction
the steel into a pore having a diameter of d = 200 mm thereof with oxygen. If the MgO–C material contains,
after 10 min is calculated according to eqn. [59]. The for instance, metallic aluminum as an antioxidant,
penetration depth increases with a growing pool this metallic aluminum melts on contact with the
depth h and amounts to a about 2 m (see Table 2). steel and dissolves into the steel, infiltrating the
open pores. During this process, an oxidation reac-
tion according to the following equation occurs:
1.26.15.3 Steel/MgO–C
2Al þ 3O ! Al2 O3
The graphitic binder phase of MgO–C bricks is resis- ½64
tant to slag because graphite is not wetted and only with DG 0 ¼ 460 kJ mol1
minimally dissolved. In contact with liquid steel, the
situation is different. For example, carbon is soluble The equilibrium constant K of this reaction at
in iron at a temperature of 1600 C, up to 5.41 wt %. 1600 C is
Therefore MgO–C should not be useful in practice
Al2 O3
¼ eDG =RT ¼ 3
1013
0
and the question arises: ‘Why can MgO–C be used?’ K ¼ ½65
1. From the Vatcher–Hamilton equilibrium 1
between dissolved oxygen in iron and carbon88: In general, the steel penetrating the pores contains
0.1 wt % of dissolved aluminum. The content of
C þ O ! fCOg ½60 dissolved oxygen then in the equilibrium state is,
At 1600 C the partial pressure PCO is according to eqn. [65], 3 ppm. As a consequence,
the oxygen potential is decreased so that the bumping
C O ¼ 0:0024PCO ½61 effect of the CO gas and the formation of a MgO
If a steel melt with a dissolved oxygen content of protective layer are suspended. The infiltration and
30 ppm tries to enter into a pore and therefore dissolution of the graphitic binder material are pro-
comes into contact with the graphitic binder phase, moted. Accordingly, the content of carbon dissolved
the carbon activity aC at this specific point is, thus, 1. in the infiltrate, from eqn. [61], is
Inserted into eqn. [45] 0:0024
0:05
C¼ ¼ 0:4 wt% ½66
1
0:003 0:0003
PCO ¼ ¼ 1:25 bar ½62
0:0024 which is much less than 5.4%.
At the contact point, therefore, CO gas is produced, In summary, infiltration influences the corrosion
which separates the reaction partners from each profoundly. It may have beneficial effects, if the reac-
other and retards the dissolution and the infiltration tion with the refractory material leads to a surface
kinetically. compaction (formation of an MgO protective layer)
2. If steel would enter into the pore despite the CO or to stiffening of the infiltrate (slag), but it also can
cushion, the solubility of the carbon would be have detrimental effects, if chipping (spalling) occurs.
strongly decreased. The partial pressure PCO inside
the structure is only 0.05 bar. From eqn. [61], we have
C = 0.04 wt % for the dissolved carbon, which can be 1.26.16 Concluding Remarks
supported by experimental data.99
3. Furthermore, the formation of a MgO protec- As can be seen from this chapter about ceramics and
tive layer may take place because of the carbothermal refractories, the subject is fairly complex, because
reaction inside the structure: often the interplay of physical boundary conditions
and chemical interaction is governing corrosion pro-
MgO þ C ! fMgg þ fCOg ½63 cesses. Furthermore, the microstructure of the mate-
rials, particularly when it comes to refractories, is
Gaseous magnesium dissolves in the steel and likewise adding complexity, because several phases
reacts with the oxygen also dissolved in the steel in interact simultaneously and the ubiquitous porosity is
also causing problems in investigating the processes. 20. Opila, E. J. J. Am. Ceram. Soc. 2003, 86(8), 1238–1248.
21. Baxter, D. J. In Corrosion of Advanced Ceramics –
Thus, despite its industrial importance there are a lot Measurement and Modelling; Nickel, K. G., Ed.; Kluwer
of things to be learned and certainly there are good Academic Publishers: Dordrecht, NL, 1994; Vol. NATO
possibilities to improve the corrosion resistance. ASI E 267; pp 249–260.
22. Opila, E. J. J. Am. Ceram. Soc. 1994, 77(3), 730–736.
We are sure that the progress, which was made in 23. Maier, N.; Nickel, K. G. Key Eng. Mater. 2006, 326–328,
the investigations of the technical ceramics and is 1780–1783.
described in the first parts of this chapter, will help 24. Opila, E. J. Am. Ceram. Soc. 1995, 78(4), 1107–1110.
25. Maeda, M.; Nakamura, K.; Tsuge, A. J. Mater. Sci. Lett.
both to extend the use of those materials in high 1989, 8, 195–197.
temperature processes and to find economic viable 26. Tai, W.-P.; Watanabe, T.; Jacobson, N. J. Am. Ceram.
solutions for the refractory world. Doing so could Soc. 1999, 82(1), 245–248.
27. Mah, T.; Keller, K. A.; Sambasivan, S.; Kerans, R. J.
become a key issue in helping us to work more J. Am. Ceram. Soc. 1997, 80(4), 874–878.
efficiently and less harmfully for our environment 28. Federer, J. I. In Corrosion and Corrosive Degradation of
in many processes. The efforts to protect our climate Ceramics; Tressler, R. E., McNallan, M., Eds.; Martinus
Nijhoff: Boston, MA, 1990; pp 425–443.
could have a significant input from the work on 29. Jacobson, N.; Myers, D.; Opila, E.; Copland, E. J. Phys.
corrosion resistant ceramics and refractories. Chem. Solids 2005, 66(2–4), 471–478.
30. Readey, D. W. J. Eur. Ceram. Soc. 1998, 18, 2383–2387.
31. Sarpoolaky, H.; Zhang, S.; Lee, W. E. J. Eur. Ceram. Soc.
References 2003, 23, 293–300.
32. Oda, K.; Yoshio, T. J. Am. Ceram. Soc. 1997, 80(12),
3233–3236.
1. Kingery, W. D.; Bowen, H. K.; Uhlmann, D. R. 33. Hirata, T.; Ota, S.; Morimoto, T. J. Eur. Ceram. Soc. 2003,
Introduction to Ceramics, 2nd ed.; John Wiley: New York, 23, 91–97.
1976; 1038. 34. Hannink, R. H. J.; Kelly, P. M.; Muddle, B. C. J. Am.
2. Barsoum, M. W. Fundamentals of Ceramics; Drexel Ceram. Soc. 2000, 83(3), 461–487.
University: Philadelphia, 2002; 600. 35. Herrmann, M.; Schilm, J.; Michael, G.; Seipel, B.;
3. Richerson, D. W. Modern Ceramic Engineering, 3rd ed.; Nickel, K. G. J. Eur. Ceram. Soc. 2005, 25(10),
CRC Press, 2006; 707. 1805–1812.
4. Carter, C. B.; Norton, M. G. Ceramic Materials Science 36. Li, Y. Q. Mater. Sci. Eng. A 2007, 444, 184–191.
and Engineering; Springer: Berlin, 2007; 740. 37. Steinbruck, M.; Veshchunov, M. S.; Boldyrev, A. V.;
5. Carniglia, S. C.; Barna, G. L. Handbook of Industrial Shestak, V. E. Nucl. Eng. Des. 2007, 237(2), 161–181.
Refractories Technology – Principles, Types, Properties 38. Gogotsi, Y. G.; Yaroshenko, V. P.; Porz, F. J. Mater. Sci.
and Applications William Andrew Publishing: Noyes, Lett. 1992, 11, 308–310.
USA, 1992. 39. Gogotsi, Y. G.; Lavrenko, V. A. Corrosion of High-
6. Schacht, C. Refractories Handbook; CRC, 2004; 520. Performance Ceramics; Springer Verlag: Berlin, F.R.G.,
7. Goursat, A. G.; Smeltzer, W. W. Oxidation of iron Rev. 1992; 181.
High-Temp. Mater. 1973, 1(4), 351–395. 40. Morell, R. Handbook of Properties of Technical &
8. Nesbitt, J. A. Oxid. Metals 1995, 44(1/2), 309–338. Engineering Ceramics HMSO Books: London, 1987.
9. Irene, E. A. CRC Crit. Rev. Solid State Mater. Sci. 1988, 41. Haase, V. In Gmelins Handbuch der Anorganischen
14(2), 175–223. Chemie; Schröder, F., Ed.; Springer: Berlin, 1986;
10. Deal, B. E.; Grove, A. S. J. Appl. Phys. 1965, 36(12), Vol. B3; pp 325–395.
3770–3778. 42. Nickel, K. G.; Quirmbach, P. In Technische Keramische
11. Nickel, K. G. J. Eur. Ceram. Soc. 1992, 9, 3–8. Werkstoffe; Kriegesmann, J., Ed.; Deutscher
12. Reisman, A.; Nicollian, E. H.; Williams, C. K.; Merz, C. J. Wirtschaftsdienst: Köln 1991; pp 1–76.
J. Electron. Mater. 1987, 16(1), 45–55. 43. Tressler, R. E.; McNallan, M. Corrosion and Corrosive
13. Nickel, K. G.; Gogotsi, Y. G. In Handbook of Ceramic Degradation of Ceramics American Ceramic Society:
Hard Materials; Riedel, R., Ed.; VCH-Wiley: Weinheim, Westerville, OH, 1990. Vol. 10; 493.
2000; Vol. 1; pp 140–182. 44. Blachere, J. R.; Pettit, F. S. High Temperature Corrosion
14. Nickel, K. G. Corrosion of Advanced Ceramics – of Ceramics. Noyes Data Corporation: Park Ridge, NJ,
Measurement and Modelling; Kluwer Academic 1989; 188.
Publisher: Dordrecht, NL, 1994; Vol. 267; 467. 45. Gogotsi, Y. G.; Yoshimura, M. J. Mater. Sci. Lett. 1994,
15. Nickel, K. G. J. Eur. Ceram. Soc. 2005, 25(10), 13, 395–399.
1699–1704. 46. Gogotsi, Y. G.; In-Deok, J.; McNallan, M. J. J. Mater.
16. Ogbuji, L. U. J. T. J. Electrochem. Soc. 1998, 145(8), Chem. 1997, 7(9), 1841–1848.
2876–2882. 47. Jacobson, N. S.; Smialek, J. L.; Fox, D. S. In Handbook of
17. Jacobson, N. S.; Fox, D. S.; Smialek, J. L.; Opila, E. J.; Ceramics and Composites; Cheremisinoff, N. P., Ed.;
Tottorelli, P. F.; More, K. L.; Nickel, K. G.; Hirata, T.; Marcel Dekker: New York, 1990; pp 99–136.
Yoshida, M.; Yuri, I. In Ceramic Gas Turbine Component 48. Jacobson, N. S. J. Am. Ceram. Soc. 1993, 76(1), 3–28.
Development and Characterisation; Van Roode, M., 49. Jacobson, N. S.; Smialek, J. L.; Fox, D. S. In Corrosion
Ferber, M. K., Richerson, D. W., Eds.; ASME Press: of Advanced Ceramics – Measurement and Modelling;
New York, 2003; pp 607–640. Nickel, K. G., Ed.; Kluwer Academic Publishers:
18. Opila, E. J.; Hann, R. E., Jr. J. Am. Ceram. Soc. 1997, Dordrecht, 1994; Vol. NATO ASI Series E 267;
80(1), 197–205. pp 205–222.
19. Tedmon, C. S. J. J. Electrochem. Soc. 1966, 142(3), 50. Jacobson, N. S. J. Am. Ceram. Soc. 1985, 68(8),
925–930. 432–439.
51. Berthold, C.; Nickel, K. G. J. Eur. Ceram. Soc. 1998, 5, 77. Müller, A.; Gerstel, P.; Butchereit, E.; Nickel, K. G.;
2365–2372. Aldinger, F. J. Eur. Ceram. Soc. 2004, 24, 3409–3417.
52. Nickel, K. G.; Berthold, C. In High Temperature Corrosion 78. Wang, Y.; Fei, W.; An, L. J. Am. Ceram. Soc. 2006, 89(3),
and Materials Chemistry; McNallan, M. J., Opila, E. J., 1079–1082.
Maruyama, T., Narita, T., Eds.; Electrochemical Society: 79. Jacobson, N. S.; Gogotsi, Y. G.; Yoshimura, M. J. Mater.
Pennington, NJ, 2000; Vol. PV 99–38; pp 429–438. Chem. 1995, 5(4), 595–601.
53. Ip, S. Y.; McNallan, M. J.; Park, D. S. J. Am. Ceram. Soc. 80. Gogotsi, Y. G.; Dransfield, G.; Porz, F. In Microscopy of
1992, 75(7), 1942–1948. Oxidation; Newcomb, S. B., Bennett, M. J., Eds.;
54. Schneider, G. A.; Nickel, K. G.; Petzow, G. In Physics and The Institute of Materials: Cambridge, 1993; Vol. 2;
Chemistry of Carbides, Nitrides and Borides, NATO ASI pp 509.
Series E: 185; Freer, R., Ed.; Kluwer Academic: 81. Shimada, S.; Takada, Y.; Tsujino, J. J. Eur. Ceram. Soc.
Dordrecht, NL, 1990; pp 387–401. 2005, 25, 1765–1770.
55. Narushima, T.; Goto, T.; Yokoyama, Y.; Iguchi, Y. J. Am. 82. Basu, B.; Raju, G. B.; Suri, A. K. Int. Mater. Rev. 2006, 51
Ceram. Soc. 1993, 76(10), 2521–2524. (6), 352–374.
56. Presser, V.; Nickel, K. G. CRC Crit. Rev. Solid State 83. Kulpa, A.; Troczynski, T. J. Am. Ceram. Soc. 1996, 79(2),
Mater. Sci. 2008, 33(1), 1–99. 518–520.
57. Costello, J. A.; Tressler, R. E. J. Am. Ceram. Soc. 1986, 84. Voitovich, V. B.; Lavrenko, V. A.; Adejev, V. M. Oxid.
69(9), 674–681. Metals 1994, 42(1), 145–161.
58. Ogbuji, L. U. J. T.; Opila, E. J. J. Electrochem. Soc. 1995, 85. Raj, R.; Riedel, R.; Sorarù, G. D. J. Am. Ceram. Soc.
142(3), 925–930. 2001, 84(10), 2158–2159.
59. Ramberg, C. E.; Worrell, W. L. J. Am. Ceram. Soc. 2001, 86. Wang, Z. C.; Gerstel, P.; Kaiser, G.; Bill, J.; Aldinger, F. J.
84, 2607–2716. Am. Ceram. Soc. 2005, 88(10), 2709–2712.
60. Opila, E. J. J. Am. Ceram. Soc. 1999, 82(3), 625–636. 87. Nguyen, Q. N.; Opila, E. J.; Robinson, R. C.
61. Sangster, R. C.; Kämpf, P.; Nohl, U. Silicon Suppl. B 5d2; J. Electrochem. Soc. 2004, 151(10), B558–B562.
Springer-Verlag: Berlin, 1995; Vol. 15; 304. 88. Czichos, H. HÜTTE - Die Grundlagen der
62. Yoshio, T.; Oda, K. In Corrosion and Corrosive Ingenieurwissenschaften, 29th ed.; Springer-Verlag:
Degradation of Ceramics; Tressler, R. E., McNallan, M., Berlin, 1991; 1417.
Eds.; American Ceramic Society, 1990; pp 367, 386. 89. Markert, J. Gravimetrische und elektrochemische
63. Oda, K.; Yoshio, T.; Miyamoto, Y.; Koizumi, M. J. Am. Untersuchung der Reaktion zwischen Kieselsäure und
Ceram. Soc. 1993, 76(5), 1365–1368. Korund in desoxidierten Eisenschmelzen. RWTH
64. Sangster, R. C.; Bär, G. In Gmelin Handbook of Inorganic Aachen, Aachen, 1977.
and Organometallic Chemistry; Schröder, F., Ed.; 8th ed.; 90. Levich, V. G. Physiochemical Hydrodynamics; Prentice
Springer-Verlag: Berlin, 1995; Vol. Suppl. B 5d1; p 255. Hall: Englewood Cliffs, NJ, 1962; pp 373–390.
65. Fox, D. S.; Smialek, J. L. J. Am. Ceram. Soc. 1990, 73(2), 91. Turkdogan, E. T. Physiochemical Properties of Molten
303–311. Slags and Glasses; Metals Society: London, 1983; pp
66. Fox, D. S.; Jacobson, N. S.; Smialek, J. L. In Corrosion 163–174.
and Corrosive Degradation of Ceramics; Tressler, R. E., 92. Eisenhüttenleute, V. D. Stahlfibel; Verlag Stahleisen
McNallan, M., Eds.; Martinus Nijhoff: Boston, MA, 1990; mbH: Düsseldorf, 1989.
pp 227–249. 93. von Bogdandy, L. Die Reduktion der Eisenerze; Verlag
67. Singhal, S. C. Ceramurgia Int. 1976, 2(3), 123–129. Stahleisen mbH: Düsseldorf, 1967.
68. Ogbuji, L.; Smialek, J. J. Electrochem. Soc. 1991, 138 94. Brüggmann, C.; Krause, O.; Pötschke, J.; Simmat, R. In
(10), 51–53. High Temperature Abrasion Resistance of Cement-
69. Oda, K.; Yoshio, T. J. Ceram. Soc. Jpn. 1993, 101(8), Based Refractories 48th International Colloquium on
855–859. Refractories, Aachen; 2005, pp 35–39.
70. Jacobson, N.; Farmer, S.; Moore, A.; Sayir, H. J. Am. 95. Harmuth, H.; Tschegg, E. K. Veitsch-Radex Rundschau
Ceram. Soc. 1999, 82(2), 393–398. 1994, 1–2, 465–480.
71. Jacobson, N. S.; Eckel, A. J.; Misra, A. K.; Humphrey, D. 96. Alapin, B.; Ollig, M.; Pötschke, J. In Thermomechanical
L. J. Am. Ceram. Soc. 1990, 73(8), 2330–2332. Properties of Selected Refractory Materials up to 1500 C
72. Riedel, R.; Kienzle, A.; Dreßler, W.; Ruwisch, L.; Bill, J.; 46th International Colloquium on Refractories, Aachen,
Aldinger, F. Nature 1996, 382, 796–798. D, 2003, pp 107–112.
73. Friess, M.; Bill, J.; Aldinger, F.; Szabo, D. V.; Riedel, R. 97. Jansen, H. Stahl und Eisen 2005, 125(11), 43–49.
Key Eng. Mater. 1994, 89–9195–100.. 98. Lee, W. E.; Zhang, S. Int. Mater. Rev. 1999, 44(3),
74. Riedel, R.; Kleebe, H.-J.; Schönfelder, H.; Aldinger, F. 77–104.
Nature 1995, 374(6), 526–528. 99. Ollig, M. Reaktionen kohlenstoffhaltiger
75. Nickel, K. G. In Precursor-Derived Ceramics; Bill, J., MgO-Werkstoffe mit flüssigem Eisen. RWTH Aachen,
Wakai, F., Aldinger, F., Eds.; VCH-Wiley: Weinheim, 1999. Aachen, 2000.
76. Butchereit, E.; Nickel, K. G.; Müller, A. J. Am. Ceram. 100. Schumacher, C.; Nickel, K. G. Ceram. Eng. Sci. Proc.;
Soc. 2001, 84(10), 2184–2188. 2000.
1.27 High Temperature Coatings: Protection and
Breakdown
H. E. Evans
School of Metallurgy and Materials, The University of Birmingham, Birmingham B15 2TT, UK
1.27.1 The Need for Coatings 692

1.27.2 Underlying Requirements of Coatings 693
1.27.2.1 The Protective Oxide Layer 693
1.27.2.2 Coating Composition and Solute Supply 694
1.27.3 Examples of Main Coating Types 696
1.27.3.1 Overlay Coatings 696
1.27.3.1.1 MCrAlY coatings 696
1.27.3.1.2 Examples of other metallic coatings 697
1.27.3.1.3 Oxide overlay coatings 698
1.27.3.2 SMART MCrAlY Overlay Coatings 699
1.27.3.3 Diffusion Coatings 700
1.27.3.3.1 Basic types 700
1.27.3.3.2 Aluminide coatings 701
1.27.3.3.3 Platinized coatings 702
1.27.3.3.4 Silicide diffusion coatings 702
1.27.3.4 Thermal Barrier Coatings 704
1.27.4 Case Study I: Oxidation and Lifing of MCrAlY Overlays 705
1.27.4.1 Protective Oxidation 705
1.27.4.2 Spallation and Cracking of the Protective Alumina Layer 706
1.27.4.3 Aluminum Depletion 709
1.27.4.3.1 Uniform depletion 709
1.27.4.3.2 Diffusion cells and chemical failure 710
1.27.5 Case Study II: Oxidation-Induced Failure of TBC Systems 712
1.27.5.1 Failure Times under Oxidizing Conditions 712
1.27.5.2 Failure Characteristics 713
1.27.5.3 Strain Energy Considerations 715
1.27.5.4 Formation of Subcritical Cracks 716
1.27.5.4.1 Transformation strains 716
1.27.5.4.2 Mechanical instabilities 717
1.27.5.4.3 Roughness of the BC surface 718
1.27.5.4.4 Chemical failure 719
1.27.5.5 Surface Roughness: A Common Factor? 720
References 722
Em Young’s modulus of the alloy or metallic coating

Symbols
(Pa or GPa)
a Amplitude of bond coat surface roughness (m)
Eox Young’s modulus of the oxide (Pa or GPa)
b Wavelength of bond coat surface roughness (m)
fB Bulk fractional concentration of solute
C Concentration (at.%)
fl Fractional concentration of solute at oxide–
CAl Concentration of aluminum (m3)
coating interface
Cb Bulk concentration (at.%)
f(x) Fractional concentration at position x
D Interdiffusion coefficient (m2 s1)
h thickess of oxide layer (m)
DAl Interdiffusion coefficient for aluminum (m2 s1)
ho Preexisting oxide thickness (m)
691
JAl Flux of aluminum (m2 s1) oxidation/hot-corrosion protection. As a result, the

kn Oxidation rate constant (mn s1) intrinsic protection capability of many superalloys is
kp Parabolic oxidation rate constant (m2 s1) adequate only for short-term exposures and coatings
kp0 Mass gain parabolic rate constant (g2 m4 s1 or are required to permit extended operation under service
mg2 cm4 s1) conditions. This point will be demonstrated in the later
Mo Atomic weight of oxygen (kg) text for the particular case of Ni-based superalloys
Mox Molecular weight of the oxide (kg) under oxidizing conditions.
n Dimensionless number High-strength nickel-based alloys are used exten-
R Radius of decohesion zone (m) sively for components in gas turbines; an excellent
t Exposure time (s) account of their development and properties is
T Temperature (K) provided by Reed.1 Single crystal alloys, for example,
Tox Oxidation temperature (K) CMSX4, are used for high-pressure blades in
DT Amplitude of temperature change (K) aeroengines and in small (10 MW) land-based tur-
DTb Critical temperature change for oxide buckling bines, but, in larger machines, polycrystalline alloys
(K) (e.g., IN738LC) are used. Some compositions of
DTc Critical temperature change for oxide nickel-based alloys are given in Table 1. These are
spallation (K) given in ‘weight percentage’ as is customary, but, in
v Volume of oxide formed per aluminum atom (m3) assessing the ability to form and maintain a protective
w Half the specimen or cell thickness (m) alumina layer at high temperatures, it is the atomic
W* Strain energy (J m3) percentage of available aluminum which is a key
WAl Atomic weight of aluminum (kg) factor. For alloys having high nickel, chromium, and
WAI2O3 Molecular weight of alumina (kg) cobalt contents, the atomic concentration of alumi-
x Spatial coordinate and distance from the oxide– num is roughly twice the weight percentage value. In
metal interface (m) more complex alloys containing appreciable amounts
xh Section loss (m) of heavy elements such as tungsten, rhenium, or
am Linear thermal expansion coefficient of alloy or ruthenium, this factor is nearer 2.5. Thus, even
coating (K1) though in most of the alloys (shown in Table 1) the
aox Linear thermal expansion coefficient of oxide aluminum concentration is 5.6 wt%, its atomic con-
(K1) centration increases from 11 at.% in simpler alloys
« Strain such as RR2000 to 13 at.% in PWA1497. Much of
f Pilling–Bedworth ratio this aluminum is associated with the strengthening g0
gF Effective fracture energy (J m2) phase in these alloys.
n Poisson’s ratio The importance of both aluminum and chromium
rAI2O3 Density of alumina (kg m3) concentrations to the ability of a Ni-based alloy to
rm Density of the alloy (kg m3) form a protective alumina layer is summarized in
rox Density of the oxide (kg m3) Figure 1. This diagram is similar to that given by
sox In-plane oxide stress (Pa or MPa) Evans and Taylor,2 but has been modified to reflect
the low-Cr compositions of the more recent alloys
shown in Table 1 and to include g/g0 coatings. For
convenience in comparing between alloys, concen-
1.27.1 The Need for Coatings trations in weight percentage are used. The experi-
mentally based lines shown in Figure 1 derive from
High temperature superalloys, usually based on iron-, the work of Giggins and Pettit3 on NiCrAl alloys at
cobalt-, or nickel-rich compositions,1 are optimized for temperatures within the range 1000–1200 C. They
mechanical properties, such as creep and fatigue resis- show that in simple Ni–Al binary alloys, at least 6 wt
tance, since they are used for highly-stressed compo- % aluminum is required to form an alumina layer at
nents. In all cases, however, these alloys will be exposed these temperatures. This critical value decreases with
to oxidizing or hot-corrosion environments during ser- increasing alloy chromium content to 3wt% alumi-
vice and will undergo some form of degradation pro- num at chromium contents 10 wt%. At low alumi-
cess resulting from this interaction. The composition num and chromium contents, that is, below
optimization required to achieve good mechanical prop- the downward sloping line to the left-hand side of
erties, however, often conflicts with the requirements for Figure 1, the principal surface oxide is expected to
High Temperature Coatings: Protection and Breakdown 693
Table 1 Compositions (wt%) of some representative nickel-based superalloys
Alloy Ni Co W Cr Al Ta Re Hf Ru Ti Mo Nb
RR2000 Balance 15.0 – 10.0 5.5 2.8 – – – 4.0 3.0 –

IN738LC Balance 8.5 2.6 16.0 3.4 1.7 – – – 3.4 1.7 0.9
CM186 Balance 9.3 8.5 6.0 5.7 3.4 3.0 1.4 – 0.5 0.2 –
CMSX4 Balance 9.8 6.4 6.4 5.6 6.5 3.0 0.1 – 1.0 0.6 0.4
CMSX10 Balance 3.0 5.0 2.0 5.7 8.0 6.0 0.03 – 0.2 2.9 –
PWA1497 Balance 16.5 6.0 2.0 5.6 8.3 6.0 0.15 3.0 – 2.0 –
TMS162 Balance 5.8 5.8 2.9 5.8 5.6 4.9 0.1 6.0 – 3.9 –
30 is exacerbated with recent generations of alloys

in which chromium levels have been decreased to
β-forming
diffusion coatings
2–2.5 wt% (Table 1), making it less favorable for a
25
protective layer of alumina to form (see Figure 1). In
addition, elements such as rhenium and ruthenium
20
have been introduced, which have volatile oxides
Al2O3
that can be released at high temperatures, at least in
wt% Al
15 the absence of a protective surface oxide layer.

γ/γ9-forming Another important consideration when assessing the
diffusion coatings Overlay
10 coatings oxidation resistance of an alloy is its ability to reform
Typical a protective alumina layer should that which formed
superalloys
5 initially spall or crack. This is an arduous requirement
for borderline alloys, since the formation and growth of
NiO + sub-surface Cr2O3 + subsurface the original alumina layer is a process of selective oxida-
0
0 5 10 15 20 25 30 tion and necessarily depletes the alloy of aluminum.
wt% Cr Thus, even though an alumina layer may have formed
100 wt% Ni initially on the superalloy, residual aluminum concen-
Figure 1 A schematic map derived from Evans and trations are unlikely to be able to maintain protective
Taylor,2 based on experimental data3 showing the type of conditions when repeated spallation events occur.
oxide formed in the NiCrAl system over the temperature It is for this combination of reasons that Ni-based
range 1000–1200 C. Superimposed are the composition superalloys operating at elevated temperatures, say
ranges of typical superalloys and coatings.
1000 C, are coated with Al-rich layers. The com-
position ranges of two important classes of such coat-
ings, overlays and diffusion coatings, are shown in
be NiO (in NiCrAl alloys), with internal oxidation of Figure 1. These will be discussed shortly but, clearly,
aluminum and chromium. To the right-hand side of both types are designed to lie within the alumina-
Figure 1, corresponding to alloys of high chromium forming region in the diagram. It is the combination
and lowaluminum content, chromia will be the expected of high-strength superalloy and oxidation-resistant
surface oxide and aluminum will oxidize internally. coating that permits component operation for
Also shown schematically, as the cross-hatched extended periods at high temperatures.
ellipse in Figure 1, is the approximate range of
aluminum and chromium contents often used in
Ni-based superalloys. It is immediately apparent 1.27.2 Underlying Requirements of
that they tend to lie on or near the bounding lines Coatings
for protective alumina formation. In some cases, alu-
1.27.2.1 The Protective Oxide Layer
mina formation can be expected but not in others.
As a group, these Ni-based superalloys can be consid- High temperature coatings rely, for protection, on the
ered to be borderline alumina formers and nonprotec- formation of a dense and adherent surface oxide layer
tive oxidation has been reported in this temperature that is chemically stable, thickens slowly with time at
range for IN738LC4,5 and CMSX10.6 The situation temperature, and has good adherence to the coating.
These are, of course, the same requirements as are It must be borne in mind, though, that the formation
needed in the design of oxidation-resistant alloys in of the protective layer is a process of selective oxidation
general, as discussed elsewhere in this book, and, as in in which the solute atom is preferentially removed
that case, the choice is limited to chromia (Cr2O3), from the coating, oxidized, and incorporated into the
silica (SiO2), and alumina (a-Al2O3). oxide layer. This removal necessarily depletes the coat-
Chromia layers are particularly useful in conferring ing of the solute and will generate a concentration
protection against Type-II hot corrosion7 and interme- profile as shown schematically in Figure 2 for the
diate-temperature oxidation, say 600–800 C, but can case of aluminum removal from, nominally, a NiCrAlY
be converted to CrO3 at higher temperatures. The rate overlay coating. This is assumed to have had an initial
of formation of this gaseous oxide can be significant in aluminum concentration of 10 wt% and can be com-
both dry and moist air at temperatures 950 C and pared with the uncoated alloy having an initial concen-
chromia-forming coatings have limited usage at high tration of 5.6 wt%.
temperatures in such environments. Silica layers are The second requirement of a protective coating is
capable of providing protection to very high tempera- that the residual solute concentration after initial for-
tures, for example, via ceramic layers on carbon-based mation of the surface layer is still sufficient to reform a
materials, to 1800 C.8 They also suffer, however, protective layer in the event of cracking or spalling of
from concerns with chemical stability in moist envir- this first-formed layer. As shown in Figure 2, the
onments. Alumina layers do not suffer from such con- coating has sufficient aluminum, and hence even after
cerns and are the favored means of protection under alumina formation, there is still sufficient solute at the
Type-I hot-corrosion and high temperature oxidation oxide–metal interface to reform this layer. By contrast,
conditions. The majority of coatings are now designed because the superalloy had barely sufficient aluminum
to form a continuous alumina layer under these condi- to form the alumina layer initially, the interfacial con-
tions and to have sufficient aluminum content to pro- centration falls below the critical value after formation
vide a healing capability in the event of cracking or of the protective layer. Subsequent spallation of this
spallation of this layer. The majority of this chapter will layer would then mean that rapid oxidation of the alloy
place emphasis on the behavior of such alumina-form- would take place until the local concentration had
ing coatings designed for use under high temperature reached the critical value for reforming the alumina
oxidizing conditions, but interesting exceptions with layer. The quantity of metal consumed is shown by the
commercial relevance are also included. hatched area in Figure 2 as 7 mm but it will be much
larger if spallation occurs later in life when more pro-
nounced depletion has occurred. Typically, the thick-
1.27.2.2 Coating Composition and
ness of this at-risk zone will increase parabolically with
Solute Supply
exposure time up to the first spallation event.
The initial priority of oxidation-resistant coatings is to The shapes of the depletion profiles shown in
form a protective layer that can be achieved simply by Figure 2 are determined by the need to maintain a
incorporating sufficient amount of the relevant ele- mass balance at the oxide–metal interface, such that
ment into the coating composition. However, this the flux of aluminum leaving the metal is equal to
approach can impair physical and mechanical proper- that required to produce the observed thickening rate
ties, for example, by decreasing melting temperatures of the oxide.9 Considering the ideal case that the
or coating fracture toughness and ductility, and a bal- oxide thickness, h, increases parabolically with time,
ance must always be sought. The composition ranges of t, and forms at the start of the exposure period, the
three types of alumina-forming coatings are shown in rate constant, kp, is then defined as
Figure 1, from which it is clear that all lie in the
kp ¼ h2 =t ½1
appropriate composition range for protective oxide
formation. Detailed discussion of the coating types is This means that the flux of aluminum atoms, JAl,
provided later, but it is clear that aluminum concentra- arriving at the oxide–metal interface is
tions of 10 wt% are typical for both the MCrAlY
@CAl 1 dx 1
overlays and g/g0 diffusion coatings. These compare JAl ¼ DAl ¼ ¼ kp0:5 t 0:5 ½2
@x x¼0 V dt 2
with the minimum requirements for alumina formation
(solid lines) of 3 and 6 wt% aluminum, respectively. where it is assumed, reasonably for protective condi-
The b-forming class of diffusion coating has substan- tions, that the recession of the oxide–metal interface
tially larger amounts of aluminum at 25 wt%. is much smaller than the diffusion distances within
12
Alumina layer NiCrAlY coating

10
Al concentration (wt%)
8
Al depletion profiles
Bare alloy
6
4
Critical healing concentration of 3 wt% Al
Depth of alloy consumed by oxide if 2

the alumina layer cracks or spalls.
0
0 20 40 60 80 100
Distance from Oxide–metal interface (μm)
Figure 2 Selective oxidation of aluminum from a NiCrAlY coating and superalloy produces depletion profiles in that
element. They result in interfacial aluminum concentrations insufficient to reform the alumina layer on the superalloy without
the prior consumption of some of the underlying metal. The critical interfacial concentration for reforming an alumina layer is
taken as 3 wt%, that is, of similar magnitude to those shown as solid lines in Figure 1.
the coating. In eqn. [2], DAl is the interdiffusion function. The interfacial concentration, fI, is constant
coefficient for aluminum within the coating, CAl is provided that parabolic oxidation and semiinfinite con-
the aluminum concentration (number/unit volume), ditions apply, but its value is directly related to the rate
x is the spatial coordinate measured from the oxide– constant kp and can be determined from the need to
metal interface, and V is the volume of oxide formed conserve the aluminum flux at the oxide–metal inter-
per aluminum atom. The depletion profile within the face. For simplicity, taking f as a weight fraction, the
metal can then be obtained from solution of Fick’s interfacial concentration during the growth of alumina
second law is given as11
@CAl @ 2 CAl " #
¼ DAl ½3 pkp 1=2 rA12 O3 WA1
@t @x 2 f1 ¼ fB ½5
DA1 rm WA12 O3
using eqn. [2] as a boundary condition, that is, the
flux of aluminum at the oxide–metal interface varies as
Here, r is density, W is molecular or atomic weight, the
t1/2. At relatively short exposure times, the depletion
suffix ‘m’ is the coating, and Al2O3 the oxide.
distance will be small compared with the thickness of
The concentration in the coating at its interface with
the coating so that the latter can then be considered as
the oxide will be maintained provided that there is
semiinfinite. Under these conditions, the depletion
sufficient aluminum reservoir and thickness of the coat-
profile is obtained as10
! ing to approximate to semiinfinite conditions. Protec-
f ðxÞ fI x tion should, thus, be provided by these generic coatings
¼ erf ½4 until end-of-life effects, initiated often by aluminum-
fB fI 2ðDAl t Þ0:5
reservoir depletion, intervene. At this stage, the interfa-
where the fs are consistent concentration parameters, cial concentration will begin to decrease and a level will
for example, weight (mass) fractions, f (x) is the value at eventually be reached where the protective alumina
distance x, fI is the value in the metal at the oxide– layer cannot be maintained. These issues will be dis-
metal interface, fB is the bulk value, and erf is the error cussed in detail in the following sections.
1.27.3 Examples of Main Coating relate to the coatings shown in the table. Details of
Types these techniques and the historical development of
protective coatings can be found elsewhere.13–18
1.27.3.1 Overlay Coatings Even though the details of the coating techniques
1.27.3.1.1 MCrAlY coatings will be covered elsewhere in this book, it is important
Overlay coatings consist of a discrete coating layer to note that different application methods will pro-
deposited onto the surface of the component, with little duce different coating structures as can be appre-
interdiffusion and chemical reaction with the alloy ciated from Figure 3. In both of the plasma sprayed
substrate. These overlay coatings are often metallic coatings, (a) and (b), the sequential spray runs are
alloys, 150–300 mm thick, based on MCrAlY composi- evident in the lateral dark stringers of oxide formed
tions where M is Ni, Co, sometimes Fe, or a combina- around individual alloy particles during the deposi-
tion of these. Early development of these overlay tion process. These stringers can be continuous in the
coatings centered on cobalt-based materials containing APS coating in which extensive oxidation in the air
20–40 wt% Cr, 12–20 wt% Al, and 0.5 wt% Y.12 On the environment can result. Some oxide formation is
basis of hot corrosion and mechanical property tests, an inevitable also under LPPS, as can be appreciated
optimal composition of Co25Cr14Al0.5Y (wt%) was from Figure 3(b), but in this case, continuous strin-
selected. This initial composition is similar to those of gers are not usually formed (e.g., see Figure 4). These
the more recent Co-based coatings, ATD 2B and ATD differences affect the longevity of the coatings, even
5B, given in Table 2. Later trends, however, have been for a given nominal composition, as will be discussed
to replace some or all of the cobalt content with nickel later with reference to aluminum reservoir issues. The
in order to improve toughness and to limit interdiffu- EB-PVD coating, by contrast, forms as vertical col-
sion effects when protecting high temperature umns with negligible oxidation. These vertical features
Ni-based superalloys. There has also been a tendency can be seen on closer inspection in Figure 3(c), and
to decrease aluminum content, again to improve coating their presence provides sites of potential mechanical
toughness. A few examples of the composition of current weakness, which can be removed by postprocessing,
MCrAlY overlay coatings are given in Table 2. for example, peening, to improve lateral bonding.
It is sometimes the case, as can be inferred from The MCrAlY coatings (shown in Table 2) have a
Table 2, that overlay coatings of the same or similar nominal b/g structure, where b can be NiAl or CoAl
compositions have different designations, for exam- and g is face-centered cubic Ni- or Co-rich solid solu-
ple, LCO22 and CN122. This reflects the different tion. In practice, however, the actual microstructure will
possible deposition methods in commercial use. depend on cooling rate and temperature as well as on the
These include air plasma spraying (APS), argon- detailed composition. Ni-rich phases, for example, can
shrouded plasma spraying (ASPS), low-pressure undergo the following four-phase reaction at 980 C19:
plasma spraying (LPPS), high-velocity oxy-fuel b þ g ! g0 þ a ½6
(HVOF) spraying, electrodeposition, and physical
0
vapor deposition (PVD) processes such as sputtering where g is the Ni3Al phase and a is the a-Cr solid
and electron beam (EB) PVD. Some of these methods solution. It can be appreciated that a given coating
Table 2 Compositions (wt%) of some MCrAlY overlay coatings
Coating or feedstock Al Cr Ni Co Fe Y Others
ATD 2B 12 23 – Balance – 0.3

ATD 5B 11 18 – Balance – 0.5
LCO 22 8 22 32 Balance – 0.5 –
CN 122 8 22 32 Balance – 0.5 –
LN 34 11 20 Balance – – 0.5 0.5Mo
ATD 8 11 24 – – Balance 0.6 –
Amdry 963 6 25 Balance – – 0.4 –
Amdry 995 8 21 32 Balance – 0.5 –
Amdry 997 8.5 20 Balance 23 – 0.6 4Ta
LN and LCO coatings deposited by argon-shrouded plasma spraying, ATD deposited by EB-PVD, CN deposited by low-pressure plasma
spraying, Amdry are feedstock powders.
CMSX4
substrate
20 μm
MCrAlY
APS
coating
(a)
Ni plate
Figure 4 The duplex microstructure of a LPPS CoNiCrAlY

overlay coating. Courtesy of Dr. M.P. Taylor.
MCrAlY
LPPS
coating effect, however, possibly because the strains arise at
temperatures where the coating is ductile and stress
Interdiffusion zone relaxation rates are high.21,22 Nevertheless, the con-
cern has been one of the historic factors influencing
CM186 alloy
the incorporation of cobalt into many coating com-
positions in order to decrease the stability of the g0
(b)
phase. Even so, g0 phases have been reported20 in Ni-
MCrAlY based MCrAlY coatings containing 20 wt% Co.
EB-PVD
coating
Minor alloying additions, in addition to yttrium,
have also been made to MCrAlY coatings either in
attempts to improve alumina formation or adherence
IN738LC or to improve mechanical properties. Some examples
alloy include silicon23 and hafnium24 and also tantalum
(Table 2), titanium,23 molybdenum (Table 2), nio-
(c)
bium, rhenium, and zirconium. These are often used
Figure 3 Examples of the structure of MCrAlY coatings in combinations, but detailed compositions of current
deposited by (a) APS,2 (b) LPPS,2 and (c) EB-PVD 13
coatings tend to be proprietary. MCrAlY coatings have
techniques.
also been modified by means of surface enrichment
with silicon25 and aluminum (e.g., GE coatings 29+
composition can have a b/g microstructure at higher and 33+) and, in lower-density air plasma sprayed
temperatures, but at lower temperatures – often structures, by inward percolation of liquid NiAl3.26
corresponding to service conditions – can be g0 /a. In
reality, particularly if fairly rapid temperature 1.27.3.1.2 Examples of other metallic coatings
changes take place, the microstructure may be non- Much effort has been made to deposit metallic coat-
equilibrium and contain a mixture of the various ings other than MCrAlYs in attempts to improve
phases.20 An example of a typical duplex microstruc- substrate properties15; some examples designed to
ture is provided in Figure 4. Further complications confer oxidation resistance are given here.
may arise near the interface with the substrate alloy if Ti–Al alloys have considerable commercial poten-
g0 -stabilizing elements such as tantalum, titanium, tial because of their high specific creep strength, that
and niobium have diffused into the coating; the g0 is, the combination of good creep strength and low
phase can then exist to temperatures above 980 C at density. Their disadvantage lies in their poor oxida-
this location. tion resistance resulting from the competitive growth
Reaction [6] is also accompanied by linear con- of alumina and rutile (TiO2). Both of these oxides are
traction strains of 0.7% and there has been concern very stable, but the latter has a much higher growth
that this could prejudice the integrity of the coating. rate than alumina and the combination makes it
There is no particular evidence for any deleterious difficult to form and sustain protective conditions.
The situation can be improved by the application of the formation of silica coatings. An early evaluation of
surface coatings of increased aluminum to titanium silica coatings, produced from silane onto a 9Cr ferritic
ratios having greater propensity to form alumina. steel, has been made by Crouch and Dooley.33 Later
Interesting examples are the use of a sputtered Ti– developments have used plasma-assisted vapor deposi-
51Al–13Cr (at.%) overlay27 to improve oxidation tion (PAVD) resulting from the decomposition of tetra-
performance of a near-a titanium alloy at 750 C in ethoxysilane. Very dense, uncracked, and adherent
air. A similar principle has been adopted28 through amorphous silica coatings up to 10 mm thick have been
the use of either graded or discrete sequential layers produced by this method on, for example, Incoloy
of Ti3Al, TiAl, and TiAl3, again on a near-a alloy. 800H,34 IN 738LC35 and 20Cr25Ni, Nb-stabilized aus-
In an alternative approach,29 it is recognized that tenitic steel.36 The improvements in oxidation resistance
at typical operating temperatures for Ti–Al (say obtained from these coatings have been substantial, for
700–750 C), it is not necessary to use an alumina example, decrease by a factor of 5 in mass gain for IN
layer for oxidation protection, since a chromia layer 738LC after 5000 h exposure in air at 900 C.35
will provide the necessary improvements. As a conse- A striking demonstration of the improvement
quence, it is possible to use Al-free chromia-forming obtained on the 20Cr25Ni austenitic steel at 1300 C
coatings of high fracture toughness and ductility to is given in Figure 5.37 In this figure, the top micro-
confer oxidation protection.29 graph shows how the entire cross-section of the
Overlay coatings may also allow the potential of uncoated specimen has been oxidized after 1h expo-
other reactive alloys with attractive mechanical prop- sure at this temperature in a CO2-based gas. In con-
erties to be realized. Important examples are those trast, the silica-coated specimen (lower micrograph)
alloys based on niobium, which have good creep is largely still intact. This example is particularly
properties at high temperature but poor oxidation interesting, since the 20Cr25Ni, Nb-stabilized
resistance. Alumina-forming FeCrAlY and silica- alloy is used as thin-walled (0.38 mm) tubes to contain
forming SiCrAlY overlays, deposited by APS, can UO2 fuel in the UK’s Advanced Gas-Cooled Reactors
produce oxidation protection to molybdenum for (AGRs). The test time and temperature used relate to
significant periods, 300 and 2000 h, respectively, at postulated fault scenarios with restricted coolant flow.
1200 C.30 An advanced coating architecture for the These substantial improvements in oxidation
protection of a Nb–5Mo–15W alloy has been devel- resistance arise because the silica coating acts as a
oped by Narita and coworkers,31 in which the outer barrier both to the inward diffusion of oxygen and to
Al-rich reservoir within the coating is protected from
interdiffusion by underlying Re–Cr-rich layers.
1.27.3.1.3 Oxide overlay coatings

Directly depositing a surface coating of a protective
oxide layer ought, in principle, to be an attractive
option, since it bypasses the need for the metallic
element to be present in significant quantities in the
underlying alloy. This is particularly the case for alu-
minum and silicon, whose presence in large amounts
can impair the mechanical properties of the alloy. (a)
A low-cost method for the deposition of an oxide
coating is through the use of sol–gels.15,32 In this,
the coating is formed by evaporation of a colloidal
suspension of oxide or hydroxide particles and subse-
quent consolidation of the deposit to form a dense layer.
A common feature, though, is the formation of cracks 225 μm
(b)
during the curing process, which necessitates the use of
multilayers to confer reasonable oxidation protection. Figure 5 Optical micrographs of cross-sections of
Even so, crack-like defects will exist within the layer specimens of a 20Cr25Ni, Nb-stabilized alloy after oxidation
for 1 h at 1300 C in a CO2/1% CO gas.37 (a) Complete
thickness, which will lead to decreased coating ductility. oxidation of the uncoated alloy and (b) relatively intact
A better approach is to produce the oxide layer from the silica-coated specimen. The magnification is the same in
vapor phase and this has had considerable success with each case.
the outward diffusion of the alloy constituents. and the need to protect against Type-I hot corrosion
A disadvantage, though, is that the coating offers no and high temperature oxidation. In the former case,
self-healing capability in the event that through- high chromium levels and chromia protective layers
thickness cracks or localized spallation occurs. In are optimum whereas, in the latter case, high alumi-
other words, there is no reservoir of silicon available num concentrations and protective alumina layers are
to reform the silica layer. Cracks in the layer will more effective.7,39 These coatings respond reasonably
allow oxygen ingress to the underlying alloy and the but not as well as coatings designed for specific local
continuing formation of nonprotective oxides.32 The conditions of temperature and environment within,
situation is aggravated in the case of amorphous silica say, a large land-based gas turbine. Furthermore,
because of its low linear thermal expansion coeffi- the range of protection afforded by these coatings
cient of 0.5 106 K1. Because the coating layer is is limited and hampers moves to a more flexible
much thinner than the metal thickness, it is reason- operating regime in which various alternative fuels
able to assume that the planar dimensions of could be employed, for example, ranging from natural
the coating will follow those of the underlying alloy gas through kerosene, diesel oils to gaseous fuels made
during a temperature change, DT ¼ (T Tox), from coal, biomass, or waste.
(where T is the new temperature and Tox is the SMART MCrAlY overlay coatings have been
original oxidation temperature). The in-plane strain, designed40,41 to address this problem of flexibility
e, within the coating is then given as and also the need to increase fuel efficiency by
higher-temperature operation. Under these aggres-
e ¼ DT ðametal asilica Þ ½7
sive conditions, high temperature oxidation and both
where a is the respective thermal expansion coeffi- Type-I and Type-II hot corrosion may be experi-
cient. A detailed discussion of such mechanical issues enced across a single component. The intention is
has been given elsewhere,38 but it is instructive here for SMART coatings to respond to the local environ-
to calculate the thermal expansion strain in the silica ment and to generate the optimum protective system
coating of Figure 5(b). Assuming that the silica layer to it, that is, either a chromia or an alumina layer. The
was deposited at 800 C, DT ¼ 500 C for the tem- coatings are chemically graded to provide surface
perature rise to the test temperature of 1300 C. layers enriched in aluminum and chromium, respec-
Taking also ametal ¼ 17.5 106 K1 then gives a tively, overlying a MCrAlY overlay coating. The
tensile in-plane strain within the silica coating of structure is shown schematically in Figure 6. The
0.85%. Interestingly, this is much higher than the
fracture strain values at 500 C of 0.27–0.45% found
Alumina layer
by Crouch and Dooley33 for sol–gel-deposited silica.
It can be surmised from Figure 5(b) that the silica
layer formed by PAVD remains uncracked, however, MCrAlY
High
even after the 500 C temperature rise, since there is b-NiAl
temp.
continued oxidation protection. It implies that the a-Cr Substrate
PAVD structure either contains fewer crack-like alloy
MCrAlY
defects and, hence, has higher intrinsic ductility
than the sol–gel equivalent or that creep processes
at the oxidation temperature relax stresses at crack Substrate
alloy Chromia layer
tips and, again, increase ductility.
Eventual degradation of the amorphous silica MCrAlY
layer is associated both with its transformation to
the crystalline state (a-cristobalite) and with reaction Low
Substrate
with cations from the alloy.34 temp.
alloy
1.27.3.2 SMART MCrAlY Overlay Coatings Figure 6 Schematic diagram of the chemically-graded
SMART coating structure showing the outer b-NiAl layer
The compositions of current MCrAlY coatings
and the intermediate Cr-rich zone overlying the base
(Table 2) reflect a compromise between the need to MCrAlY coating. The right-hand figures show the expected
provide protection against low-temperature Type-II response under different oxidation and hot-corrosion
hot corrosion and intermediate-temperature oxidation regimes.
outer layer consists of b-NiAl produced by aluminiz- montage of SEM micrographs of cross-sections of
ing but before this treatment, the base MCrAlY coat- corroded specimens shown in Figure 7.40,41 The test
ing had been pretreated to form an intermediate layer duration in all cases was 500 h in an air–400 vpm
rich in chromium. Manufacturing details of this pro- SO2/SO3 environment and were recoated every 20h in
cess are currently proprietary,41 but it is the presence an 80% Na2SO4/20% K2SO4 eutectic mix. Figure 7
of this intermediate Cr-rich layer that distinguishes shows a comparison of a SMART coating variant
SMART coatings from over-aluminized variants, for (SmC 155H) with a Pt-aluminized diffusion coating
example, GT33þ, currently available. (RT22) and an over-aluminized CoNiCrAlY overlay
Under high temperature oxidation or Type-I hot coating in the as-produced condition and after testing
corrosion conditions, the outer layer of the SMART at 700 and 800 C. It can be seen that both commer-
coating reacts to form an alumina layer and provide cial coatings have been severely corroded at both
optimum protection. The Cr-rich intermediate layer temperatures whereas the SMART variant is rela-
acts as a diffusion barrier limiting interdiffusion and, tively unaffected at 800 C and shows penetration
particularly, the loss of aluminum from the outer only through the outer aluminized layer at the
layer into the base coating. Under Type-II hot corro- lower temperature.
sion conditions, say at 600–800 C, the outer region of
the coating forms an alumina layer sluggishly and is
1.27.3.3 Diffusion Coatings
intended to be sacrificed to allow the corrosion front
to encounter the intermediate Cr-rich layer at which 1.27.3.3.1 Basic types
a chromia protective layer will form. The rate of Diffusion coatings are formed by direct interaction of
pitting attack, associated with Type-II corrosion, the substrate alloy with elements, typically chro-
will then be substantially decreased at this stage. mium, aluminum, and silicon, which are capable of
Extensive testing of these coatings40,41 under hot forming protective oxide layers. Typically, coatings
corrosion conditions has demonstrated that these are produced through either immersion in a
concepts do work, as can be appreciated from the cementation-type pack or by vapor transfer of the
As manufactured 500 h at 700 ⬚C 500 h at 800 ⬚C
RT22
platinum
aluminide
100 μm 100 μm 100 μm
Over aluminized
CoNiCrAlY
(amdry 995)
200 μm 200 μm 200 μm
SmC 155H
smartcoat
treated
amdry 995
200 μm 200 μm 200 μm
Figure 7 Comparison of the corrosion morphology of a SMART coating variant with a RT22 (platinum-aluminide) coating
and an over-aluminized MCrAlY coating.41 See text for further details.
active species to the component remote from the unless formed by vapor deposition above pack) and
pack at temperatures where concurrent diffusion exists over a wide range of stoichiometry47 as shown
in the alloy develops the coating layer. Details can schematically in Figure 1. These coatings provide
be found elsewhere.13–18 Unlike overlays, diffusion good resistance to Type-I hot corrosion and to
coatings produced by these methods have limited high temperature oxidation through the formation
flexibility to vary composition, although it is possi- and maintenance of a protective alumina layer.
ble to codeposit more than one reactive species, Their disadvantage is that they are relatively thin
for example, Al/Cr,42 Al/Si,43 and Cr/Si44 and to and, although having higher aluminum contents
use predeposited layers, for example, platinum, to than the thicker MCrAlY overlay coatings, have a
vary coating structure.45 More flexibility with regard smaller total reservoir of aluminum. Loss of alumi-
to coating composition can be achieved through num through the oxidation process as well as by
the use of a slurry of reactive species, which can be interdiffusion with the substrate alloy limits their
painted or sprayed onto a work piece and sub- effectiveness after extended periods at high tempera-
sequently diffusion annealed. By repetition of this tures, say >1050 C.
process, multilayered coatings can be produced.46 A particularly successful modification to alumi-
Important examples include Cr–Si- and Al–Si- nide coatings has been the incorporation of platinum
rich compositions, for example, Sermaloy J and into the coating by deposition of a thin layer, 10 mm
Sermetel 1515. thick, prior to the aluminizing stage. The b layer
Chromizing can be used with both iron- and which is formed, again 50 mm thick, has the compo-
nickel-based alloys to provide protection at moderate sition (Ni, Pt)Al with platinum contents of 8 at.%.
temperatures (700 C) against Type-II hot corrosion This phase can also exist over a wide range of alumi-
and intermediate-temperature oxidation. The con- num and platinum contents.47 The microstructure of
centration of chromium in the alloy surface can be such a coating is shown in the as-received condition
raised to 40–80 wt% and this is sufficient to form a in Figure 8(a) as a bondcoat (BC) within a thermal
protective chromia layer. Chrome–aluminizing is also
used to improve protection, particularly at slightly
higher temperatures. It has become the case, however,
as temperatures increase in the pursuit of improved 25 μm EB-PVD YSZ topcoat
fuel efficiency, that more attention has been paid
to the production of aluminide layers and the Alumina TGO
subsequent formation of alumina protective layers. β-(Ni,Pt)Al bondcoat
1.27.3.3.2 Aluminide coatings IDZ
Important applications of aluminide coatings arise in

the protection of Co- and, particularly, Ni-based CMSX-4 superalloy
superalloys, since both form aluminide phases
(a)
which fix high concentrations of aluminum as a sur- Grit and
face layer, typically 30–70 mm thick. The coatings can possibly
voids
be produced from packs of either high or low alumi-
num activity. The former is processed at lower tem- β IDZ
peratures and the latter at higher temperatures in the
γ′
typical range 800–1100 C. High-activity aluminizing Pt-aluminide
results in the formation of a Ni2Al3 layer produced 25 μm coating
predominantly by inward diffusion of aluminum.
Constituents of the alloy can become incorporated
within this layer, which has relatively poor toughness. (b)
Improvements result in subsequent heat treatment to Figure 8 Cross-section of Pt-aluminide diffusion coatings.
convert the higher aluminide to NiAl. Low-activity (a) In the as-received condition as a bondcoat (BC) in a TBC
processing results directly in the formation of a NiAl system showing the fully b-(Ni, Pt)Al layer and an extensive
interdiffusion zone (IDZ) and (b) after 100h oxidation at
layer, which forms principally by the outward diffu- 1100 C showing partial transformation of the original b
sion of nickel. This layer is then relatively free of phase to g0 resulting from aluminum depletion. Courtesy of
alloy constituents (but can contain pack residues Miss F. Schenach.
barrier coating (TBC) system, which will be flux of aluminum from the alloy into this zone. Con-
described later. It can be seen that there is also an centrations of aluminum in the Ni–Al–Pt coating
extensive interdiffusion zone beneath the b layer. need to be 10 wt%, that is, approximately twice
This contains a dispersion of phases rich in alloy the base alloy concentration, to stabilize the g/g0
constituents such as tantalum and titanium. phase structure at 1100 C.47 The likely range of
The addition of platinum to the b-phase appears aluminum contents within the coating, shown sche-
to improve the oxide spallation resistance of the coat- matically in Figure 1, is similar to that of MCrAlY
ings, possibly by decreasing the segregation of sulfur coatings and will be sufficient both to form an initial
to the oxide–metal interface.48,49 It also decreases the alumina layer and to reheal this in the event of its
chemical activity of aluminum47 and then should be failure (Figure 2). The apparent ‘uphill’ diffusion of
expected to decrease the extent of aluminum loss by aluminum out of the alloy into the coating, that is, its
interdiffusion with the substrate alloy. The experi- diffusion against its concentration gradient, can be
mental evidence is ambiguous, however, because explained again by the strong bonding between plati-
platinum additions also seem to increase the princi- num and aluminum. It must mean that the chemical
pal diffusion coefficient of aluminum.47 Continued activity of aluminum within the coating is lower than
exposures at high temperature will inevitably result that within the alloy so that the enrichment within
in aluminum depletion as a consequence of both the coating is simply a consequence of aluminum
alumina formation and interdiffusion and this will diffusing down its activity gradient. As recognized,47
cause phase transformation within the coating along the coating has a replenishing capability by being
the sequence able to transfer aluminum into itself from the alloy,
as the continued growth of the alumina layer removes
b ! g0 ! g
aluminum from its outer surface.
An example of a Pt-aluminized coating with par-
1.27.3.3.4 Silicide diffusion coatings
tial transformation to the g0 phase after oxidation at
Silicide coatings have the potential of offering signif-
1000 C for 100 h is shown in Figure 8(b). It should
icant improvements in oxidation and hot corrosion
be noted, though, that both phases have adequate
resistance through the formation of a protective layer
aluminum contents to maintain a protective alumina
of silica. Siliconizing of iron and simple carbon steels
layer.
was achieved by Fitzer in 195453 using pack cemen-
tation. These coatings were found to improve resis-
1.27.3.3.3 Platinized coatings tance to both Type-I and Type-II hot corrosion.
An important variation on the conventional Pt- They consist principally of an outer layer of ferrite
aluminized coatings described above is to omit the alu- containing up to 10 wt% Si and are susceptible to
minizing stage in coating manufacture but to retain cracking during mechanical and thermal loading.54
the step of platinum deposition and subsequent Silicon-rich coatings on Ni-based alloys have been
inward diffusion into the Ni-based superalloy. This achieved using chemical vapor deposition (CVD)55
approach was patented in 1999 by Rolls-Royce and but there are concerns with the Ni–Si system over
Chromalloy UK50 and has subsequently been devel- the presence of relatively low melting temperature
oped to the stage of commercial deployment as BCs in eutectics. NiSi, in particular, forms an eutectic with
TBCs – see later text for a description of these. Unfor- Ni2Si which melts at 943 C. More attractive
tunately, few of the results of the underlying research approaches for iron- and nickel-based alloys are
work supporting this development have appeared in likely to involve the incorporation of silicon into
the open literature.51,52 chromized or aluminized layers.36,56
The key requirement of the coating is that it Direct reaction of silicon with the alloy substrate to
should be capable of forming and maintaining a pro- form a high melting-point surface layer is of particular
tective alumina layer. It is not intuitively obvious, interest with molybdenum-,57 niobium-,58 and tita-
however, that this should be so because no aluminum nium-based57,59–62 alloys. The last, in particular, are
is added and, as is recalled from Section 1.27.1 and of current relevance in aeroengines because of their
1.27.2, the Ni-based superalloy has borderline prop- good density-compensated creep properties but are
erties in this respect. Nevertheless, a 50-mm-thick g/ handicapped by poor long-term oxidation resistance.
g0 surface zone of enhanced aluminum concentration A wide range of titanium silicides can be produced
is formed indicating that there must be an outward in sequential layers on the alloy surface by using a
cementation pack, as shown by Cockeram and The subsequent reaction [9] produces the silicide:
Rapp.57,61 These coatings can be 100 mm thick and
5Tiþ3SiO ! Ti5 Si3 þ1‰O2 ½9
substantially improve oxidation resistance, particu-
larly when doped with germanium when oxidation The overall reaction is
rates in the temperature range 700–1000 C are simi-
5Tiþ3SiO2 ! Ti5 Si3 þ3O2 ½10
lar to those for the growth of silica itself (Figure 9).
Protection is maintained even under thermal cycling This reaction will only proceed in the direction shown
but the layers are brittle and cracking should be at low oxygen partial pressures, calculated by Gray
expected. This is accentuated by the thickness of et al. 62 at 7.7 1022. This seems to have been achieved
the coating which increases the probability of crack- in a sealed silica capsule, evacuated from an argon
like initiating defects being present. atmosphere, without the use of extraneous oxygen
Thinner coatings (1–10 mm) can be produced by getters, such as titanium chips, but by using the
siliconizing, in which the test piece is buried in silicon a2 phase of the alloy as an intrinsic getter. As with the
powder, at temperatures between 800 and 1100 C.59,60 above, a substantial improvement in oxidation resis-
Higher temperatures in this range favor the formation tance, a factor 5 in weight gain in this case, was achieved
of the Ti5Si3 phase rather than TiSi2 as the outermost over the uncoated Ti–Al control over the test duration
layer produced in the cementation pack.57,61 These of 500 h at 900 C. The parabolic rate constants of the
thin Ti5Si3 layers were uncracked and the layers more protective coatings are plotted in Figure 9 and
again conferred significant improvements in oxidation compared with the behavior of unalloyed titanium63
resistance of the underlying Ti-rich alloy, as can be and pure silicon.64 It can be seen that only the Ge-
appreciated from Figure 9. A rather novel approach62 doped coatings actually achieve the level of protection
to producing a thin, 1–2 mm, layer of Ti5Si3 is to use expected of a pure silica layer.
reaction (B) below to transport silicon to a Ti–Al test Broadly similar behavior to the Ti5Si3 coatings
piece via gaseous SiO. has been found by Taniguchi et al.65 for the bulk
phase. In this case, rutile (TiO2) had formed above
3SiO2 ! 3SiO þ 1‰O2 ½8 the silica layer and clearly contributed to the weight
10–2
Unalloyed Ti (Kofstad)
10–3
Gray et al.
Parabolic rate constant (mg2 cm–4 s–1)
Li et al.
10–4
Cockeram and Rapp, B-modified
Cockeram and Rapp, Ge-doped
10–5
10–6
10–7
10–8
Pure Si
(Deal and Grove)
10–9
7.5 8.0 8.5 9.0 9.5 10.0 10.5
104/T (K–1)
Figure 9 Parabolic rate constants for some silicide-coated titanium-rich alloys compared with the behavior of unalloyed
titanium and silicon.
gain of the specimen. An implication is that the The most recent development is the use of cast, sin-
beneficial influence of Ge-doping (Figure 9) arises gle-crystal alloys, such as CMSX-4 (see Table 1), for
by limiting the transport of titanium through the blades which remain cooled and also are now coated
silica layer. with TBC systems. The use of these can improve
engine efficiency and can also extend service life-
times.66,67 This same general developmental trend is
1.27.3.4 Thermal Barrier Coatings
mirrored in land-based turbines although TETs in
When discussing the need for high temperature coat- these cases tend to be lower and the moves toward
ings in Section 1.27.1, emphasis on placed solely on TBC usage lag behind those of aeroengines. A cross-
achieving adequate oxidation and corrosion resis- section of a typical TBC system is shown schemati-
tance. However, under modern turbine conditions cally in Figure 11.
and in other applications, such as diesel engines, it A TBC system consists of the outer thermal bar-
can also be necessary to protect the substrate alloys rier topcoat which is attached to the alloy substrate
from combustion temperatures by means of TBCs. by means of a mechanically-compliant, oxidation-
This developmental trend for aeroengines and asso- resistant BC. The topcoat is generally zirconia,
ciated Ni-based superalloys is shown in simplified containing 7–8 wt% yttria. This addition provides
form in Figure 10. partial stabilization of the tetragonal (and cubic) zir-
It can be seen how in the first 20 years or so, blade conia phases and avoids the volume changes asso-
alloys had adequate properties in the wrought and ciated with the formation of the monoclinic phase.
unprotected form to satisfy turbine requirements. In This yttria-stabilized zirconia (YSZ) layer (sometimes
the early 1960s, a significant change in blade design termed yttria partially stabilized zirconia (YPSZ)) has a
permitted the use of some intake air to provide cool- low thermal conductivity, 1 Wm1 K1, which is
ing and this increased the turbine entry temperature approximately an order of magnitude lower than that
(TET) significantly above the inherent capabilities of for alumina. It means that for cooled turbine blades, a
the substrate alloys. During this period from 1940 to temperature drop of perhaps 150 C can be obtained
approximately 1990, successive generations of super- across a 250–300-mm-thick topcoat during service
alloys were developed with improved mechanical which provides considerable benefit to the load-bear-
properties and, hence, higher temperature capabil- ing capability of the Ni-based superalloy. It needs to be
ities. This has been shown simply as a linear trend in emphasized, however, that the YSZ topcoat is perme-
Figure 10 but more details can be found elsewhere.1 able to oxygen and will not provide oxidation protec-
The main change during this period was to move from tion. This is achieved by means of the BC which is
wrought to cast, directionally-solidified structures. intended to be alumina-forming and can be based on
Uncooled Cooled Coated

2100 blades blades blades
1900
Turbine
TET capability k
1700 capability
Superalloy
1500 capability
1300
Cast Cast
1100 single-crystal
polycrystalline
Wrought alloys
alloys
900 alloys
1940 1960 1980 2000

Date
Figure 10 A comparison of the historical increase in turbine entry temperature (TET) for aeroengines and the associated
temperature capability of the alloys used for turbine blades.
aeroengines, platinized diffusion coatings. A typical

TBC topcoat
cross-section of a system consisting of an EB-PVD
topcoat and a Pt-aluminide diffusion BC is shown in
Figure 8(a). The columnar nature of the EB-PVD
TGO topcoat is evident and can be contrasted with the splat
Bondcoat structure of that produced by APS. A much more
extensive interdiffusion zone than in the MCrAlY
Substrate alloy system is also formed during coating manufacture.
To date, the commercial exploitation of TBC
systems has been dominated by applications using
Ni-based alloys but there is also foreseeable potential
for their use on other high temperature structural
materials such as the Ti–Al series. Significant prog-
Figure 11 Schematic cross-section of a thermal barrier ress has already been made in the production of
coating system. aluminide diffusion coatings on Ti-based alloys69–71
and recently these have been used as a BC under an
APS YSZ topcoat in a TBC system.72
APS YSZ topcoat
MCrAlY bondcoat
1.27.4 Case Study I: Oxidation and
Lifing of MCrAlY Overlays
IN738 superalloy
1.27.4.1 Protective Oxidation
In the initial stages of exposure of a complex alloy
Figure 12 Cross-section of a typical land-based TBC
coating system consisting of an air plasma sprayed (APS) such as the (Ni, Co)CrAlYs, all constituents which
YSZ topcoat with a MCrAlY BC68 on an IN738 Ni-based can oxidize will do so and nonprotective oxides such
superalloy substrate. Note the relatively rough BC surface as NiO together with spinels such as NiCr2O4 are
and the limited interdiffusion zone compared with the TBC likely to form. Eventually, however, a continuous
system shown in Figure 8(a).
a-alumina layer will develop underneath these oxi-
des and oxidation rates will rapidly decrease. The
formation of metastable aluminas, even at the rela-
one of the MCrAlYoverlays discussed previously or on tively low temperature of 950 C, was not found in a
diffusion coatings of the Pt-aluminide or platinized study73 using in situ high temperature XRD on a
variants. The oxidation of these BCs forms a layer of CoNiCrAlY coating. This transitional behavior may
thermally grown oxide (TGO) over time between not be an issue with these systems at relevant operat-
the BC and TBC topcoat as shown schematically in ing temperatures although with FeCrAlY alloys at the
Figure 11. As will be discussed in some depth in slightly lower temperatures of 900 and 925 C, the
Section 1.27.5, the formation of the TGO plays a transition from y- to a-alumina took tens of hours.74
significant role in the failure by spallation of the TBC. The growth kinetics of the a-alumina layer on
TBC systems used in practice on Ni-based super- MCrAlY-type alloys have been variously reported
alloys for turbine applications tend to be of two dis- to be parabolic (eqn. [1])75,76 or subparabolic77 as in
tinct types. For large, land-based units, a system the general eqn. [11] for increase in oxide thickness,
consisting of a MCrAlY BC, deposited by any of the h, with exposure time, t, at a constant temperature.
processes discussed previously, with a YSZ topcoat
h ¼ ho þ kn t 1=n ½11
produced by APS is usually used. An example of this
system is shown in Figure 12. In smaller machines, Here, ho is the preexisting oxide thickness at t ¼ 0, n is a
such as aeroengines or, increasingly, small land-based time-invariant numerical constant, and kn is
turbines, YSZ layers deposited by electron-beam the corresponding rate constant. Ideal parabolic con-
physical vapor deposition (EB-PVD) are favored. ditions exist when ho ¼ 0 and n ¼ 2 (eqn. [1]) but
These TBC systems can use MCrAlY BCs also but subparabolic kinetics obtain when n > 2. The reasons
more typically have Pt-aluminide or, in Rolls-Royce for the differences in kinetics are poorly understood
although there are considerable implications for the 10–9

longevity of the coatings. Subparabolic kinetics can CoCrAl Ni3Al
CoCrAlY Ni–25Al
arise, in principle, as a result of an increase with time FeCrAl Ni–32Al
of compressive growth stress within the oxide layer.38,78 FeCrAlSc NiCrAl
They will also arise79 if oxide grain size increases FeCrAlY NiCrAl
10–10 FeCrAl(Y) NiCrAl
during exposure and diffusion of reactive species
occurs principally by grain boundary diffusion. It is FeCrAl(Ce) NiCrAlY
FeCrAlY NiCrAl (Th, Y)
also feasible, particularly with a stoichiometric com- NiAl Pt-6Al
pound such as alumina, that the gradual incorporation
10–11
kp (g2 cm—4 s)
of coating or alloy (minor) constituents affects any
doping of the oxide layer so as to decrease diffusion
rates progressively with time. Apparent subparabolic
kinetics will also occur in those systems and exposure
conditions in which metastable alumina forms and 10–12
transforms over time to the a phase.80 The situation is
unclear and, for present purposes, it will be assumed
that the a-alumina layer grows parabolically with time
with ho ¼ 0 and n ¼ 2 in eqn. [11]. This assumption 10–13
leads to conservative estimates of coating lifetimes.
The temperature dependence of the parabolic
rate constant for the growth of a protective a-alumina
layer on various coating-type alloys is shown in 10–14
1300 1200 1100 1000 900
Figure 13,75,76 but expressed in terms of mass gain, T (⬚C)
Dm, per unit surface area. A reasonable mid-range
expression for the rate constant, kp0, from Figure 13 is Figure 13 Parabolic rate constants for the formation of
alumina on coating-type alloys.76
275 000
kp0 ¼ 2:68 104 exp ; mg2 cm4 s1 ½12
8:314T
where T is absolute temperature and the mass gain has h 20
xh 15:6 mm ½15
been expressed in the more common units of mg cm2. f 1:28
This equation can be expressed in terms of the para- where f is the Pilling–Bedworth ratio for alumina
bolic rate constant kp, which uses oxide thickness rather (1.28). This level of loss in an MCrAlY coating some
than mass gain, by means of eqn. [13]. 200 mm thick after this exposure period would be

Mox 2 0 excellent behavior but, unfortunately, other degrada-
kp ¼ k ½13 tion processes also occur which limit lifetimes. Princi-
3Mo rox p
pal among these is the loss of aluminum from the
Here, rox is the density of a-alumina (taken as 4000 mg coating and the consequences of this. Depletion will
cm3), Mox is the molecular mass of alumina and Mo is occur as a result of selective oxidation, as described in
that of oxygen; the factor three arises because there are Section 1.27.2.2, and will increase with time at tem-
three oxygen atoms in Al2O3. This conversion equates perature. In addition, it will be exacerbated by spall-
1 mg cm2 mass gain to an 5.3 mm thick layer of ation or cracking (and rehealing) of the protective
a-alumina. Using eqn. [13] to substitute into eqn. [12] oxide layer, geometrical factors, internal oxidation,
then gives: and interdiffusion with the substrate alloy. All of these

275 000 factors will be considered in this section.
kp ¼ 7:56 107 exp m2 s1 ½14
8:314T
1.27.4.2 Spallation and Cracking of the
The protection offered by the alumina layer can be
Protective Alumina Layer
appreciated by using eqn. [14] in eqn. [1] to show that
the oxide thickness after 30000h at 1000 C is still only Spallation and cracking of the protective oxide layer
20 mm. This corresponds to a section loss, xh, of the results from the generation of strains and associated
coating of 38 stresses arising from the oxide growth process,
from temperature changes or imposed upon the com- relative to the metal substrate, the equation can be
ponent. The issues are particularly complex since simplified to (compare eqn. [7]):
they depend on the nature of the oxidation process
Eox T ðam aox Þ
(the balance between growth by cation or anion sox ¼ ½17
fluxes), on the curvature of the substrate and on the ð1 uÞ
elastic, plastic and, particularly, creep properties of It is also reasonable to assume that the underlying
the respective layers of the coating system. A back- superalloy is much thicker than the coating and, of
ground review is available38 and phenomenological course, the oxide so that the alloy’s dimensional
approaches to failure which combine strains from changes are imposed upon the oxide. This means
multiple sources have also been developed.81–83 that in eqn. [17], the expansion coefficient, am, should
These predictive approaches have been formulated be that for the superalloy rather than for the coating.
specifically for TBC systems but contain general Using representative values of, u ¼ 0.3, Eox ¼ 387
principles. GPa, aox ¼ 7.9 106 K1, and am ¼ 15.0 106
This chapter emphasizes solely the influence of K1 in eqn. [17] predicts a residual in-plane oxide
thermal strains in producing oxide spallation from a stress of 3.9 GPa after a temperature drop of
MCrAlY substrate in order to establish mechanisms. 1000 C. Values of stress of this magnitude have
Strains deriving from oxide growth will be signifi- been measured in related alloys and coatings.68,85,86
cant only if creep relaxation rates are low or, on Since the values of expansion coefficients and
curved surfaces, if the oxide thickness is of similar Young’s modulus are assumed to be independent
magnitude to the local radius of curvature.38 For of temperature, the elastic oxide stress varies pro-
now they are neglected but, if significant, can sim- portionately with temperature change and will be
ply be introduced as an additive term.82 The impli- tensile for temperature increases above the oxida-
cation is that the alumina–MCrAlY interface is tion temperature.
sensibly flat, that is, there are no locations where Spallation under in-plane compression can
the local radius of curvature is of similar magnitude occur by two nominally distinct routes depending
to the oxide thickness. During cooling there are on the relative strengths of oxide and oxide–coating
then negligible out-of-plane strains, provided the interface,38,39 as shown schematically in Figure 14.
oxide remains adherent, but a substantial biaxial The critical stress for buckling of the oxide layer
compressive in-plane strain can develop within the (Figure 14(a)) can be obtained from standard
oxide layer. If elastic conditions are assumed, this text87 and converted38 to a critical temperature
strain corresponds to an in-plane oxide stress, sox, drop, DTb, of
given by84: 2
1:22 h
Eox T ðam aox Þ Tb ¼ ½18
sox ¼ ½16 ð1 u2 Þðam aox Þ R
ð1 uÞ 1 þ EEmoxwh
where the symbols have the same meaning as previ-
Here, DT (¼T Tox) is the temperature drop, ously and R is the radius of the circular zone of prior
taken as a negative quantity, Eox and Em are the decohesion at the oxide–coating interface. It can be
Young’s moduli of, respectively, oxide and metal; a appreciated that for a given value of R, the tempera-
is the corresponding linear thermal expansion coef- ture drop required to effect buckling increases with
ficient; u is the Poisson’s ratio, assumed to be the oxide thickness. For typical values of the various
same for oxide and metal; h is the oxide thickness, parameters, the zone diameter needs to be approxi-
and w is half the specimen thickness. For thin oxides mately at least 20 the thickness of the oxide layer
Oxide
layer Compression
Metal
(a) substrate (b)
Figure 14 Spallation of the oxide layer under in-plane compression from the metallic MCrAlY substrate can occur by
either (a) a buckling or (b) a wedging process.
before the buckle configuration can develop. Thus, 10

for a typical oxide thickness of 5 mm, a circular inter- ΔTc = 397 ⬚C
facial zone of some 100 mm would be required.

8 102 ⬚C h–1
This is a problematic issue since such large areas of 104 ⬚C h–1
decohesion are not seen at the alumina–MCrAlY
Crack length (μm)

interface in these simple overlay coatings. The 6
intrinsically good adherence of the alumina layer is
related to the addition of yttrium into these coatings
which appears to mitigate the potentially harmful 4
effects of sulfur segregation in lowering interface
adhesion.88,89
2
In the presence of good adhesion between oxide
and coating, a thick protective layer can develop but
this, as found on related systems, still contains poros- 0
ity and crack-like defects.90,91 At some critical value
0 100 200 300 400 500 600
of the compressive thermal stress generated during
Temperature drop (⬚C)
the cooling transient, a shear crack can be initiated
from one or more of these defects, as shown schema- Figure 15 Finite-element predictions of the kinetics of
interfacial wedge crack growth for an alumina layer on
tically in Figure 14(b). Continued cooling and con- Haynes 214 cooled from 1100 C at the rates shown. The
traction of the oxide layer will produce further shear dashed line reflects ideal behavior for spallation to occur at
on the crack-like surface and the development of out- a critical value of oxide strain energy.
of-plane tensile stresses at the oxide–coating inter-
face. A consequence is that a wedge crack will grow
along this interface as cooling continues and may oxide–metal interface at a temperature drop DT is
cause spallation of the overlying oxide layer. Exten- given as95:
sive finite-element (FE) modeling of this wedging hW ¼ hEox ð1 uÞðT Þ2 ðam aox Þ2 ½19
process has now been undertaken for chromia and
alumina layers on various metallic substrates.92–94 Equating this to the fracture energy, gF, of unit area of
A consistent result from this work is that the growth interface gives the critical temperature drop for spall-
rate of the interfacial wedge crack is significantly ation as:
retarded during cooling as a result of the relaxation " #1=2
gF
of crack-tip stresses by creep of the adjoining metal. Tc ¼ ½20
Some results93 for a 5-mm-thick alumina layer on hEox ðam aox Þ2 ð1 uÞ
Haynes 214 are given in Figure 15. They show that
the wedge crack nucleates quickly but its growth rate This shows that the critical temperature drop
is then stopped until some lower temperature at decreases with oxide thickness, that is, spallation
which rapid propagation occurs over a narrow tem- becomes easier for thicker oxides which is a common
perature range. This period of quiescence corre- observation. It contrasts with the buckling mode (eqn.
sponds to the period of significant stress relaxation [18]) which becomes more difficult with increasing
by alloy creep and its extent decreases with increased oxide thickness.
cooling rate, as shown. In applying eqn. [20] for prediction it should be
The general behavior shown in Figure 15 can be noted that the fracture energy is an effective value
approximated by the dashed horizontal and vertical since most of the strain energy is lost by creep relax-
lines superimposed, for clarity, only on the FE results ation during the cooling transient.92–94 Its value can
for a cooling rate of 104 C h1. They can be inter- be obtained from eqn. [20] using the FE predictions
preted as showing that the oxide remains adherent to of DTc and is always larger than the intrinsic fracture
the alloy or coating until a critical temperature drop, energy of the interface when relaxation processes are
DTc, at which stage the stored strain energy, W*, per present. Thus, for the example shown in Figure 15,
unit volume in the oxide layer is sufficient to produce gF is calculated to be 33 J m2 for the higher cool-
decohesion of the oxide–metal interface. For a thin ing rate and 62 J m2 for the lower. Using a similar
planar biaxially strained oxide layer of thickness h, approach for a free-standing NiCrAlY coating,93 the
the strain energy in the oxide layer per unit area of values of the effective fracture energy shown in
Table 3 Estimated effective fracture energies, gF, for shown in that section, how the aluminum concentra-
the spallation of a 5-mm-thick alumina layer from a free- tion can be decreased beneath the oxide but that the
standing NiCrAlY coating92
aluminum reservoir within the bulk of the coating
Cooling rate Computed DTc ( C) g F (J m2) could maintain concentrations sufficiently to ensure
( C h1; C s1) reformation of the alumina layer in the event of its
spallation or cracking. Nevertheless, the depletion of
102 745 40
103 620 28 aluminum by selective oxidation will destabilize Al-
104 550 22 rich phases in the vicinity of the surface oxide. In
105 382 10 particular, dissolution of the b-NiAl phase will occur
106 314 7 in a zone near the alumina layer100,101 and, as
expected for parabolic oxidation kinetics, the depth
of this b-free zone will thicken parabolically with
Table 3 are obtained as a function of cooling rate. time. In addition, loss of aluminum by interdiffusion
This decrease of gF with increasing rate reflects the with the substrate can also result in the formation of a
decreased contribution from creep relaxation and b-free zone at this inner interface. With continued
shows that no unique value of the fracture energy exposure, the entire coating can become free of the
can be assumed. It is possible, however, to interpolate b-phase and this has been used as a failure condition
between these results for any given conditions within for the coating.100 This is a useful pragmatic approach
the range shown. The values of gF given in the table but is conservative since sufficient aluminum will
are typically 10–100 greater than the likely intrinsic usually still be present to provide a rehealing
interfacial fracture energy but are of similar magni- capability.
tude to those measured by blister testing96 and spall- Even so, substantial exposure periods can be
ation93,94,97,98 on other systems. obtained while still retaining some b-phase. Anton
This discussion has highlighted the benefits of et al.,101 for example, find a total b-depletion zone,
stress relaxation within the coating to the spallation from inner and outer coating surfaces, of 160 mm
resistance of the oxide during a simple cooling tran- in a 250-mm-thick LPPS CoNiCrAlY coating after
sient. It should not necessarily be generalized from 10000h at 1000 C. This particular coating contained
this, however, that weak coatings should be preferred only 8 wt% Al but exhibited substantial interdiffu-
over strong. An important consideration is that creep sion with the IN738LC alloy substrate and formation
relaxation during thermal cycling can generate a of Kirkendall voids within the interdiffusion zone
mechanical hysteresis loop94,99 which can place the (IDZ). The extent of b depletion was decreased, as
oxide layer into in-plane tension on the return to expected, by increasing the aluminum content to
temperature. Consequent tensile cracking will, of 12 wt% in a NiCoCrAlY coating but rhenium addi-
course, impair the protective function of the oxide tions also decreased the extent of interdiffusion. In
layer. Strong coatings are also desirable to maintain this case the total b-depletion zone was only 75 mm
mechanical stability in rapidly rotating components. after 10 000 h at 1000 C.101
The balance of creep properties between high and Although the extent of aluminum depletion
low strength will depend on the application, for resulting from selective oxidation can be predicted
example, the magnitude of external stresses, the fre- based on the concepts outlined in Section 1.27.2.2, it
quency of thermal cycling, the rate of change of is much more difficult to model the interdiffusion
temperature, and the presence of hold periods within process. Analytical and numerical attempts have been
the creep regime. made for MCrAlY systems83,102 but it is difficult to
deal with diffusion and concurrent precipitation
without knowing the appropriate interaction terms
1.27.4.3 Aluminum Depletion
for the diffusion coefficients or, equivalently, the
1.27.4.3.1 Uniform depletion chemical activities of the diffusing species. Taylor,
The second main form of coating degradation is et al.,102 for example, report how the diffusion of
through the loss of aluminum and the consequent aluminum from a LPPS CoNiCrAlY coating into a
inability to maintain and reform the protective alu- CM186 substrate alloy, in the early stages of exposure
mina layer. A general discussion of the depletion at 1200 C, is limited by the formation of Al-rich
occurring as a result of the selective oxidation of phases, associated with tantalum, within the IDZ.
aluminum has been given in Section 1.27.2.2. It is With continued exposure and further aluminum
25 depletion within the coating, these phases dissociate

and act as a reservoir to resupply aluminum back into
Isolated particle the coating. These counter-flows of diffusive flux
20 reflect the changing aluminum activities within the
Al concentration (at.%)
coating and IDZ. They emphasize that, in such com-

15 Initial concentration plex situations, modeling using single-valued diffu-
No diffusion sion coefficients and chemical potential gradients
cells approximated by gradients in concentration can
10 lead to erroneous results.
Interesting future developments to limit inter-
D = 2.40 ⫻ 10–14 m2 s–1 diffusion are likely to involve the control of aluminum
5 kp = 1.16⫻ 10–16 m2 s–1 chemical activity within the coating103 or the use of
diffusion barriers such as a-Cr or Re-based alloys.104
0
0 5 10 15 20 25 30 35 40 45 50 55 1.27.4.3.2 Diffusion cells and chemical failure
(a) Distance (μm) In practice, exceptions to the ideal condition of
uniform depletion will exist. These include the diffu-
25
sional isolation of volumes of the coating as a result of
internal oxidation or geometric factors. As high-
D = 2.40 ⫻ 10–14 m2 s–1 lighted in Figure 3, MCrAlY coatings deposited by
20 kp = 7.87 ⫻ 10–17 m2 s–1
plasma spraying in air (APS) are of relatively low
density as a result of oxidation of the sprayed parti-

15 Initial concentration cles during coating formation. Bonding between the
splats within the coating is impaired and internal
No diffusion cells
oxidation at the splat surfaces can continue during
10 subsequent exposure. ASPS and even HVOF coatings
probably also suffer from this but to a lesser extent.
The alumina layers formed internally in APS
5
MCrAlY coatings have been shown105 to act as diffu-
sion barriers which isolate the splat volume from the
0 reservoir of aluminum in the remainder of the coat-
0 5 10 15 20 25 ing. As a consequence, aluminum depletion within
(b) Distance (μm)
the relatively small splat volume occurs at a much
25 greater rate than would be expected for the coating as
a whole. This is shown in Figure 16 for an APS
D = 2.40 ⫻ 10–14 m2 s–1 CoNiCrAlY coating oxidized in air at 1100 C.105
20 kp = 6.30 ⫻ 10–17 m2 s–1 Here the measurements of aluminum concentration
were obtained across separate diffusion cells which
were located at the surface of the coating and

15 Initial concentration bounded around their surface by an alumina layer.
No diffusion cells
The lines in the figures are predictions from eqn. [21]
which considers the cell to be approximated as a long
10
2-dimensional slab, on both flat surfaces on which
alumina grows parabolically.
5
The points are measured using SEM/EDS and the solid line
0 is the best-fit solution to eqn. [21] using the values shown for
0 5 10 15 20 25 30 35 40 45 50 55
D and kp. See text for more details. The dashed line is the
(c) Distance (μm)
expected depletion profile, using the same values of D and
Figure 16 Enhanced aluminum depletion measured in kp, if diffusion cells were absent. Oxidation was performed in
individual diffusion cells in an APS CoNiCrAlY coating. air for (a) 4 h, (b) 4 h, and (c) 100 h.105

50 pkp 1=2 500 20
C ¼ Cb :
f D 200
" ! !# 10
X1
Exposure time at 1100 ⬚C (h)

2nwþx 2ðnþ1Þw x 100 Chemical failure
þ erfc ½21
Alumina thickness (μm)

erfc 1 1
n0 2ðDt Þ2 2ðDt Þ2 50 5.0
Here, C is the aluminum concentration in at.%, Cb is 3.0

20
the initial value within the cell, D is the aluminum 10
2.0
interdiffusion coefficient, 2w is the cell width, x is
5
the distance from the oxide–metal interface, and 1.0
t is the exposure time. The predictions were made No chemical failure
2
using a single-valued interdiffusion coefficient 2.4 0.5
1
1014 m2 s1 (N.B. the sites were remote from any 0.3
0.5
interdiffusion effects from the substrate) consistent 7 10 20 30 40 50 70 100 200
with measurements on NiCrAl106 and CoCrAl107 Cell width, 2w (μm)
systems. The corresponding kp values to give the
Figure 17 A chemical failure (CF) map for a MCrAlY
observed fit vary within a factor 2 but lie within the coating (M¼Co,Ni). The left-hand ordinate refers to
data spread of Figure 13 (shown there in terms of exposure times at 1100 C but the oxide thickness of the
mass gain). The expected depletion profile without right-hand ordinate is independent of temperature.
diffusion cells is shown by the dashed line. The
agreement with the data is good and confirms that
oxidation is continuing at similar rates at both the initiating CF is relevant to the MICF condition
external and internal surfaces in these coatings. It can assuming a critical aluminum concentration of 5 at.
be seen that depletion is much more pronounced in %. The exposure times shown as the left-hand verti-
the small cell of Figure 16(b) than in the larger one cal axis refer to 1100 C but the oxide thicknesses
of Figure 16(a) even after 4 h at 1100 C. After 100 h (right-hand axis) apply at other temperatures also.
exposure, aluminum has been essentially consumed Clearly, small cells are susceptible to early MICF
in the 55 mm cell, Figure 16(c). leading to the consumption of the entire cell volume
The decrease in aluminum concentration can with the formation of porous and friable nonprotec-
make these cell volumes susceptible to chemical fail- tive oxides. For dense coatings, however, for example
ure (CF) at which stage fast-growing nonalumina those produced by LPPS, internal oxidation is negli-
oxides will form. A framework theory of CF can be gible and small diffusion cells do not form.110
found elsewhere.108 The cell will be susceptible to MCrAlY coatings produced by spraying tend to
mechanically induced chemical failure (MICF) when have convoluted surfaces consisting of protuberances
the aluminum concentration anywhere within it with dimensions relating to those of the original
(usually, at the center plane) falls below that needed sprayed particles. They may also be postprocessed
to reform the alumina layer should the original crack to change this topography but usually have a signifi-
or spall. Intrinsic chemical failure (InCF), unlike cant residual surface roughness. The significance is
MICF, is a thermodynamic boundary at which the that a surface protuberance will have a surface-to-
aluminum concentration at the oxide–metal interface volume ratio larger than that of a flat coating. Alumi-
falls to that below which alumina can be reduced by num depletion will then occur at an enhanced rate
another of the coating constituents, chromia for and early CF of the protuberance can be expected.
example. This critical concentration is very close to Calculation of the rate of depletion is difficult to
zero for the reduction of alumina by chromium but undertake analytically because the protuberance is
there is experimental evidence for the formation of not isolated from the bulk of the coating, as with the
Cr-rich oxides beneath an alumina layer in NiCrAlY diffusion cells discussed earlier. Some resupply of
coatings.109 aluminum will occur via the neck of the protuber-
A CF map for MCrAlY coatings can be con- ance. Useful progress in prediction can be made,
structed using eqn. [21] as shown in Figure 17. however, using numerical finite-difference methods
The calculations were performed using D ¼ and an example of some results, obtained using the
2.4 1014 m2 s1, as above, and a mid-range value 2-dimensional ODIN (Oxidation and INterdiffusion)
of kp ¼ 9.0 1017 m2 s1. The threshold line for code, is shown in Figure 18. These results apply to a
0 10 20 30 40 50 longevity. In order to achieve this and to develop the

0
full potential of TBC systems safely and reliably, it is
Protuberance necessary to understand and predict the degradation
1 at.%
–10 and failure processes within the system leading to
2
4 spallation of the insulating topcoat. As a consequence,
–20 6 a tremendous amount of research activity has been
at.% Al
undertaken worldwide, particularly in the last decade,
8 to try and achieve these goals. At the time of writing,
–30
9
success has been elusive although large amounts of
experimental data have been produced and a plethora
–40 of theoretical models have been suggested. It is beyond
Coating the scope of this present work to provide a detailed and
–50 comprehensive review of this large body of work which
is often contradictory. Rather, a personal view will be
Figure 18 Finite-difference predictions of the Al depletion
presented which will attempt to identify key experi-
produced in a surface protuberance of a MCrAlY coating
after oxidation at 1100 C for 1 h. The left-hand and top mental observations and materials’ properties and favor
axes give distances in microns. The right-hand axis gives theoretical models which are consistent with these.
the aluminum concentration corresponding to the In some applications, for example, on rotating
iso-concentration lines drawn on the figure. Courtesy of blades in gas turbines, TBC systems will experience
Dr. W.M. Pragnell.
mechanically-imposed strains in addition to those
developed by the oxidation process. As discussed in
100 mm thick MCrAlY coating oxidized at 1000 C for relation to MCrAlY overlay coatings (Section
1 h but with, for purposes of illustration, a relatively 1.27.4.2), it may be feasible to sum these various
low initial concentration of 10 at.% Al. It can be strains and define a critical total value at which TBC
appreciated that selective oxidation to form the alu- failure will occur.81–83 The approach is phenomeno-
mina layer causes a significant decrease in aluminum logical in that no detailed mechanisms of failure are
content at the surface of the bulk coating with concen- considered. It is, therefore, difficult to generalize com-
tration increasing with distance from this interface. plex situations where strains may be tensile or com-
More details can be found elsewhere.111 This level of pressive and may act in different directions at different
depletion is similar to the schematic line for an locations within the TBC system. Nevertheless, the
uncoated alloy, of similar aluminum content, shown approach does attempt to address the difficult subject
in Figure 2. In this particular example, there remains of the interaction between applied external stresses
sufficient aluminum at depths of 10 mm to reform and those produced by oxidation. For present pur-
the protective alumina layer. This contrasts with the poses, however, as earlier with MCrAlY overlays, the
situation within the protuberance, however, where intention is to examine only the role of oxidation and
concentrations have dropped everywhere to <1 at.% thermal loading in TBC failure.
after 1 h exposure. Early MICF of this is then likely It was pointed out in Section 1.27.3.4 that oxygen
and, indeed, was observed.111 This failure time is can penetrate the YSZ TBC layer and oxidize the
somewhat less but of similar magnitude to that underlying BC. The TGO, designed to be alumina,
shown for diffusion cells in Figure 17. forms as an interlayer between the topcoat and BC as
shown schematically in Figure 11. The development
of this layer has been implicated in TBC spallation
1.27.5 Case Study II: Oxidation- for many years and a simple failure criterion, used in
Induced Failure of TBC Systems practice, is to assume that spallation will occur at a
critical TGO thickness. As shown in Figure 19, this
1.27.5.1 Failure Times under Oxidizing
can be a useful approach.
Conditions
The time-to-failure data shown in Figure 19
An overview of TBC systems and principal types derive from previous compilations,112,113 but have
has been given in Section 1.27.3.4. They are an been simplified to distinguish only between the
essential part of present-day strategies to operate common types of TBC system in commercial use:
high temperature machinery, particularly aeroengines Pt-aluminide/EB-PVD YSZ, MCrAlY/EB-PVD
and gas turbines, with increased fuel efficiency and YSZ, and MCrAlY/APS YSZ. Results on experimental
systems, which can show improved behavior to the eqn. [14], assuming an initially bare BC surface. It
above, have been omitted from this graph intentionally should be noted, though, that those systems with an
although they are prominent in current research liter- EB-PVD topcoat will usually have a preformed alu-
ature. These include the use of platinum over- and mina layer, 0.5 mm thick on the BC surface. The
under-layers to MCrAlY BCs,114 preoxidation of agreement with the solid-line prediction is, neverthe-
MCrAlY BCs,115,116 polishing of the BC surface,117 less, quite reasonable and supports the use of this
and platinized BCs51,52 see also Section 1.27.3.3.3). simple criterion as a pragmatic lifing method. It also
This body of work is of interest and may point the way supports the view,118,119 that time and temperature are
to future improvements in TBC systems. The results the principal parameters affecting lifetimes. The scat-
have been omitted from Figure 19 simply because ter can then be thought of as reflecting the underlying
they add other variables which will confuse the trends variability in the oxidation process together with the
present in the failure-life compilation for conventional influence of uncontrolled variables, many of which are
systems. likely to arise from the manufacturing process. These
The data shown in Figure 19 include results from will include BC surface roughness,112,113,117 defects
tests ranging from nominally isothermal exposure to within the topcoat which may initiate interfacial
those using hourly thermal cycles, for which the cracks,120 segregation of elements such as titanium
accumulated time at peak temperature is used in from the substrate alloy,52 and variable sulfur contents,
the figure. A striking feature of the results is the particularly in Y-free BCs and alloys.91,121
very large scatter in lifetime – a factor 10 – at any
given temperature. Within this scatter, there appears
1.27.5.2 Failure Characteristics
to be no systematic difference in behavior between
the coating types or with the frequency of cycling.112 The failure times plotted in Figure 19 refer to the
This is a valuable working premise with which to find loss of typically 20% or more of the area covered by
a path through the diverse competing models of the YSZ topcoat. Often this occurs by the spallation
failure. The solid line in the figure is the expected of the topcoat in sheets measuring some mm across as
behavior if topcoat spallation occurred at a critical shown in Figure 20.117 The morphology of these
TGO thickness of 4 mm. It is calculated from failed regions in terms of the ratio of the diameter
Pt-aluminide BC/EBPVD TBC

MCrAlY BC/EBPVD TBC
3000
MCrAlY BC/APS TBC
Time to failure (h)
2000
4-μm oxide growth

1000 during exposure
0
950 1000 1050 1100 1150 1200 1250
Temperature (°C)
Figure 19 The temperature dependence of the time to spallation of the ceramic topcoat in various TBC systems in
general commercial use.
possible to construct a map showing the extent of

these low-stress regions as a function of exposure
time. A particularly satisfying example is shown in
Figure 21129 for a platinized BC system exposed to
repeated thermal cycles from 1150 C.
The darker regions in this figure are those of
relatively low in-plane compressive stress compared
with the lighter regions. The low-stress regions can
be interpreted as damaged, involving cracks at either
the TGO–BC interface or within the TGO itself. For
this particular example, it is clear that the extent of
the damaged regions increases markedly only after
190 h leading to spallation of some of the TBC top-
coat after 225 h.
The attractions of using this nondestructive tech-
nique to evaluate the residual life of TBC systems
in situ within turbines are considerable but there are
Figure 20 Decohesion of large sheets of the YSZ TBC on cautionary factors. The first to note is that although an
a coated 25 mm diameter specimen.117 EB-PVD YSZ topcoat is transparent to the laser light
there is lateral spread of the beam so that return signals
are obtained from an area of 20–60 mm diame-
of the area of decohesion to the thickness of the ter.128,129 It follows that early-stage cracks, perhaps
coating (20) is consistent with buckling occurring 5 mm long, will not readily be detected. This may be
during cooling (Section 1.27.4.2). This is supported why there appears to be an extensive damage-free
by examination of cross-sections of partially failed period in the example shown in Figure 21 whereas, in
coatings122 and by extensive modeling.123 Edge crack- reality, small cracks may already have nucleated. For
ing can also, in principle, produce spallation of large APS topcoats, the applicability of the technique is less
areas of the coating123 but this appears to be a much clear since the lateral splat boundaries now form inter-
more variable event than buckling. It has not occurred, nal interfaces across the path of the impinging laser
for example, in the specimen shown in Figure 20 but beam and significant attenuation of the return signal is
there are other cases of laboratory samples where likely.
coating failure at edges has occurred117 but also The other important consideration is the location
accompanied by failure remote from these sites.124 of the spallation interface. In practice, cracks can form
As discussed in Section 1.27.4.2, catastrophic within the topcoat near the TGO, at the TGO–top-
buckling will occur during cooling only when a delam- coat interface, within the TGO, and at the TGO–BC
ination zone of critical radius has formed under the interface. The luminescence technique relates damage
coating. Preexisting zones of this size are not present to a decrease in stress within the TGO and this will
in the as-manufactured coating and must form during occur only if cracking occurs within the TGO itself or
the high temperature exposure. The relatively slow along its interface with the BC. Fracture damage
growth of these subcritical defects determines the life occurring at the other locations will not be detected
of the TBC system and, understandably, there has been by the luminescence technique. For the example of the
much interest in techniques which can assess the extent platinized BC shown in Figure 21, fracture damage
of this damage nondestructively. Studies of the use of did indeed develop both at the TGO–BC interface
impedance spectroscopy have been made125 but a more and within the TGO. The luminescence technique
developed technique to date is that of laser-activated would detect this, subject to the limits on spatial reso-
luminescence of trace Cr3+ ions within alumina. Appli- lution. There are many examples in the literature of
cation of this to TBC systems has been pioneered by crack formation in TBC systems and a perceived
Clarke and coworkers.126–128 A particularly powerful tendency for the spallation path to be at the TGO–
aspect of the technique is its ability to determine BC interface in Pt-aluminide/EB-PVD YSZ systems.
regions in which the alumina TGO within TBC sys- This is indeed frequently the case but it does not seem
tems has relatively low stress after cooling to room to be an intrinsic feature of this system, however, since
temperature. By laser scanning the specimen, it is fracture along the TGO–topcoat or within the topcoat
25 h 98 h
900
720
Distance (AT)
540
360
180
132 h 190 h
900
720
Distance (AT)
540
360
180
206 h 225 h
900
720
Distance (AT)
540
360
180
0
0 180 360 540 720 900 0 180 360 540 720 900
Distance (AT) Distance (AT)
Figure 21 Map showing the development of low-stress regions (dark areas) with increasing number of cycles, containing
1 h hold periods at 1150 C, for a TBC system consisting of an EB-PVD YSZ topcoat and platinized BC. Note that a single
specimen was examined sequentially but not the same area.129
has also been observed.117 This cracking is often asso- crack, marked A, is hundreds of microns long and lies
ciated with BC protuberances, that is, with increased at the base of the topcoat. It intersects sites, marked B,
roughness of the BC. An example of the early stages of of porous nonprotective oxides resulting from
cracking within the topcoat of a (rough) Pt-alumi- CF. There is also a crack, marked C, along the
nide/EB-PVD YSZ system is shown in Figure 22. TGO–BC interface which appears to be linked to
By contrast, the perceived tendency with an inclined shear crack traversing the thickness of
MCrAlY/APS YSZ systems is for the fracture path the TGO. This is a similar morphology to that
to lie mainly within the topcoat. Again, this is not an expected for wedge cracking (Section 1.27.4.2).
intrinsic property of the system but is associated with
the generally rough BC surface. An example of mul-
1.27.5.3 Strain Energy Considerations
tiple fracture sites, that is, within the topcoat, within
the TGO and at the TGO–BC interface in such a The driving force for spallation within the TBC
system is shown in Figure 23. Here, the principal system is the release of the stored strain energy
vary through the coating thickness and is likely to

reach its highest value at its base. Nevertheless, for
purposes of illustration, consider a topcoat of thick-
ness 200 mm with an average Young’s modulus of
50 GPa, which is representative of literature values.
Using linear thermal expansion coefficients of
12 106 K1 for the topcoat and 15 106 K1
for the superalloy and 0.1 for the Poisson’s ratio of
the topcoat within eqn. [19] yields a value for the
stored strain energy within the topcoat of 81 J m2 on
cooling to 100 C from 1100 C. It is the release of
this energy that will develop cracks at the TGO–
topcoat interface or within the topcoat near the
TGO. It should be noted that this strain energy may
change with time as a result of sintering of the top-
coat and an increase in its Young’s modulus130 or it
30.00 μm
may decrease as a result of thinning due to chemical
⫻750
or physical erosion.131
Figure 22 SEM image of a section through a TBC system The specific strain energy within the TGO will
consisting of a Pt-aluminide bond coat with an EBPVD YSZ increase as the oxide thickens and, in principle,
top coat that has been held at 1200 C for 3 h. Small cracks should be added to that within the topcoat to produce
have developed, arrowed, at the flanks of BC
cracking either within the oxide layer or at the inter-
protuberances. Courtesy of Dr. M.P. Taylor.
face between the TGO and the coating. The strain
energy within the TGO can be calculated from
B eqn. [19]. For an oxide thickness of 4 mm, that is,
A Top coat
corresponding to the failure line drawn in Figure 19,
and using similar parametric values as previously,
TGO gives a specific stored energy within the TGO of
C
Bond coat 53 J m2. The total energy available to produce frac-
100.00 μm ture within the TGO or along the oxide–coating
Substrate
interface is then 134 J m2 for the example used.
Figure 23 Section through a TBC system consisting of an These values of specific strain energy should cer-
electroplated CoNiCrAlY BC and an APS topcoat after tainly be sufficient to produce cracking even with the
oxidation in air at 1100 C for 600 h. Different cracking sites dissipation of energy by creep anticipated for typical
are indicated and described in the text. See text for (MCrAlY) BCs – see, for example, Section 1.27.4.2
discussion. Courtesy of Dr. M.P. Taylor.
and Table 3. Interestingly, the stored energy within
the topcoat is available to produce cracks within the
TBC layer or at its interface with the TGO from the
within the respective layers. It is an analogous situa- onset of exposure. The reason that cracking is
tion to the spallation of the alumina layer from a deferred and that TBC systems survive beyond the
MCrAlY coating discussed in Section 1.27.4.2. As in first thermal cycle is that there must also develop a
that case, eqn. [19] can be used as a first approach but mechanism for producing cracking and releasing this
it needs to be noted that this is a ‘thin-layer’ approxi- energy. This is a product of the high temperature
mation which makes no allowance for any strain exposure and various suggested possibilities will be
energy stored within the coating and alloy substrates. described briefly in the next section.
As pointed out earlier, it also assumes that negligible
stress exists at the exposure temperature and that the
1.27.5.4 Formation of Subcritical Cracks
strain energy derives only from thermal loading.
The stored energy within the topcoat will depend 1.27.5.4.1 Transformation strains
linearly on the in-plane Young’s modulus (eqn. [19]) The removal of aluminum from the BC both by
but this parameter is not well defined, particularly for selective oxidation and interdiffusion will result in
EB-PVD topcoats. Here the in-plane modulus will phase transformation and associated volume changes.
The most obvious of these, evident in Figure 8(b) for out-of-plane tensile stresses may be produced across
example, is the transformation from b to g0 phases. the TGO–BC interface at the transformation tem-
Tolpygo and Clarke132 have estimated an associated perature of 450–500 C.139 The magnitude of these
volume contraction during this process of 38% and is unclear since the strains may also be accommo-
have suggested that this transformation strain could dated by void formation within the original b grain
induce BC rumpling and decohesion. A difficulty and some thermal or athermal relaxation may still
with the calculation is that, in the vicinity of the be possible at this temperature. There is a strong
TGO, the transformation occurs as a direct result of circumstantial argument, however, that the transfor-
the removal and oxidation of aluminum. The volume mation is not a significantly damaging event. If it
change at this location should then also include the were so, both Pt-aluminide and MCrAlY coatings
expansion arising from formation of alumina. When would tend to fail before b-depletion had occurred.
this is done,133 the transformation (and concurrent In practice, much of the life of MCrAlY coatings, for
oxidation), is found to produce a volume expansion, example, is spent with a b-depleted zone adjacent to
rather than contraction, which is of similar magni- the TGO layer. Platinized g/g0 coatings, of course,
tude to that produced by direct oxidation of alumi- never form the b phase.
num. Little effect of the b to g0 transformation on
TGO adherence and mechanical stability of the BC 1.27.5.4.2 Mechanical instabilities
surface should then be expected and this seems to be The alumina TGO layer is under compression dur-
borne out by observations.134,135 It can also be seen ing cooling and will also be so at temperature if
from Figure 8(b) that there is no marked indentation growth stresses are compressive. The strain energy
of the coating–oxide interface above the large g0 within the layer can be released and contribute to
phase. It is worth noting, though, that at the inner decohesion if the oxide layer can relax and extend to
surface of the BC where interdiffusion is occurring its stress-free dimensions. The linear relaxation strain
with the alloy substrate, the volume change asso- per cooling transient will be no more than 0.5%,
ciated with alumina formation should not be equivalent to an extension of 0.1 mm over a 20 mm
included in the calculation of total volume. The distance. Large dimensional changes of the TGO
transformation may result in contraction strains in layer over these small lengths would then require
this case which could contribute to the cavity devel- multiple thermal cycles.
opment sometimes found at these locations. It is not One suggested mechanism is that of ratchet-
easy to calculate these transformation strains, how- ing.135,141 The initiating event, in this case, is the
ever, since other, often poorly characterized, phases formation of small cracks within the TBC above
will form within the interdiffusion zone and the vol- preexisting undulations of the BC, essentially as
ume changes associated with these should also be shown arrowed in Figure 22. These cracks remove
taken into account. mechanical constraint from the underlying TGO
Another transformation of interest is that of mar- which, as discussed earlier, may be under a compres-
tensite formation of the b phase during cooling. This sive growth stress at the oxidation temperature but
has been recognized for many years for b-NiAl136,137 will certainly experience large in-plane compressive
and has also recently been confirmed in b-(Ni, Pt)Al stresses during cooling. Within the model, it is envi-
coatings.134,138,139 Typical of martensitic transforma- saged that, with increasing number of cycles, these
tions, a critical cooling rate is required but this seems stresses can drive the oxide layer as an indent into the
remarkably low in this case where martensite has BC and, in doing so, enlarge the crack at the base of
been found after air cooling 138 and even after furnace the topcoat. The model seems feasible, in principle,
cooling.139 The former, at least, is consistent with and experimental support has been provided135
earlier observations on b-NiAl.140 These recent although the incidence of failure by this route does
results suggest that martensitic transformation of not seem widespread. It is often observed, for exam-
the b-(Ni, Pt)Al phase should be a common occur- ple, that the main fracture site in TBC systems with
rence during cooling. such diffusion BCs is at the TGO–BC inter-
The formation of martensite will result in a volume face51,52,128 rather than within the TBC topcoat.
change estimated at 2%, that is, a linear contrac- A key aspect is that there needs to be either a certain
tion of 0.7%.138,139 Unlike the b ! g0 transformation roughness to the BC surface so that cracks such as
discussed earlier, the expansion due to alumina forma- those in Figure 22 can form or there needs to be a
tion will not mitigate this transformation strain and defect at the base of the topcoat which offers no
constraint to the downward migration of the TGO. 1.27.5.4.3 Roughness of the BC surface
The issue of surface roughness will be addressed in It has been known for some time that the roughness of
later sections. the BC surface will have a significant influence on the
There will probably exist a window of BC strengths development of stresses in the TBC system during
within which the ratcheting process could occur. It is cooling.148,149 Various attemspts have been made to
not possible to quantify this because modeling the evaluate these using FE methods113,130,133,150–153 which
progressive indentation is difficult and has tended to incorporate materials’ behavior of varying degrees of
assume135,141 that the BC deforms by athermal plastic- relevance. Most recently, it has also become clear that
ity characterized by yield behavior. In fact, creep pro- out-of-plane tensile stresses can also develop at the
cesses will dominate,142,143 even during cooling.133 Low oxidation temperature as a consequence of the volume
creep strength is necessary to move the BC away from increase on oxide formation.112,113
the indenting oxide but this deformation occurs at con- The development of these numerical models has
stant volume. There must, thus, also be an upwelling of to some extent mirrored the expansion in computing
the displaced matter against the TBC topcoat adjacent power. Thus, initial approaches tended to assume
to the indent. In this case, the creep strength of the BC entirely elastic behavior of the respective layers but,
must be high enough to displace the topcoat upward and importantly, creep behavior has been subsequently
allow continued penetration of the indenting oxide layer incorporated and recognition made that the TGO
into the BC. These two contradictory requirements of thickens during exposure. The most recent mod-
the BC creep strength may be satisfied only within a els113,130,133 incorporate creep of both BC and
limited temperature and composition range. TGO, allow for growth of the latter together with
Another interesting mechanical instability is rum- the associated volume change at temperature during
pling. This has been recognized for some years the computation. In addition, they permit densifica-
and has been shown to occur in aluminized and Pt- tion of the YSZ topcoat, again during the simulated
aluminized coatings under thermal cycling condi- exposure time, and consider a variation in BC surface
tions132,144,145 but in the absence of a topcoat. It is roughness. Some predicted results for two different
manifested by a roughening of the specimen surface BC-roughness geometries are shown in Figure 24.133
and is clearly associated with thermal cycling because As described in the caption, the stress contours of
isothermally-tested specimens do not show the phe- Figure 24 apply to the situation at the test tempera-
nomenon.144 The basic underlying reasons for the ture of 1100 C. For the case examined, the total
surface instability are well understood146,147 and are thickness of the alumina TGO had increased to
broadly the same as discussed above for the ratchet- 3.3 mm during the simulated 100 h isothermal expo-
ing process. In both cases, deformation of the coating sure at this temperature. The parameters b and a
permits the compressively stressed TGO to increase are the amplitude and half-wavelength, respectively,
its length and, thereby, relax its strain energy. As of the surface undulations shown in Figure 24. For
pointed out earlier, this increase in length per tem- the rougher surface, Figure 24(a), out-of-plane ten-
perature transient is small, 0.5%, and repeated sile stresses of 150 MPa, have developed within the
thermal cycles are needed to produce readily observ- topcoat along the flank of the BC protuberance.
able deformation. For the less-rough surface, however (Figure 24(b)),
The rumpling process has been implicated147 in these stresses are absent, showing the strong sensitiv-
failure of the TBC system because the instability ity of isothermal stress development to BC surface
would develop significant out-of-plane stresses in the roughness. In both cases, tensile stresses also develop
vicinity of the TGO with associated crack initiation. in the valleys of the undulations.
Modeling of the process again assumes plastic yielding The out-of-plane tensile stresses on the flanks of
rather than creep behavior and, although the presence the BC protuberance arise from continuity strains112
of the topcoat is considered, no mechanical constraint which are required if the topcoat is to remain
is attributed to it. This constraint will be present in uncracked. Their origins can be understood by
real cases and, for the reasons given above when dis- recognizing that the outward displacement due to
cussing ratcheting, is likely to limit the applicability of the formation of the TGO is normal to the BC
the process in TBC systems. Again, exceptions will interface and, so, changes direction around the pro-
arise when defects exist in the topcoat such that, tuberance shown in Figure 24. The out-of-plane
locally, mechanical constraint is absent and upward component, the x22 direction, of this displacement
displacement of the TGO and BC becomes feasible. also varies around the protuberance and reaches a
s22 (MPa)
153
120
YSZ 90
60
30
A 0.0
−134
TGO
(3.3 μm)
BC
X2
X1
(a) b/a − 0.52 (b) b/a − 0.25

Figure 24 FE predictions 133
of the s22 stress contours at 1100 C after 100 h oxidation at the same temperature for an
EB-PVD YSZ TBC system containing a Pt-aluminide BC. Creep of the BC and TGO, sintering of the topcoat and outward
strains due to TGO growth are considered in the stress analysis. The BC creep properties used were those of Taylor et al.,143
The arrowed locations mark the general sites of high stress at the flank (a) and valley (b) regions.
minimum on its flank. To maintain continuity of the These regions of high stress correspond to loca-
topcoat, tensile stresses must then develop to strain tions where cracks form in TBC systems. Examples of
the topcoat in these locations to account for the out- those found at flank and valley locations are arrowed
of-plane deficit in the displacements resulting from in Figure 22 and may well have formed at tempera-
oxide growth. ture. Crack initiation, presumably during cooling, has
The effect, on the s22 stress distribution, of cool- also been reported113 at the TGO–BC interface at
ing the specimen to 25 C after the isothermal the peak of the BC protuberances. These nucleation
exposure for 100 h at 1100 C is shown in Figure 25, processes will occur, however, only with relatively
again for the two BC surface roughnesses.133 The rough BCs. Cracking events should be less easy and
out-of-plane flank stresses in the rougher specimen, TBC lifetimes would be expected to be longer with
Figure 25(a), are largely unaffected by this transient flatter BC surfaces. This has been clearly demon-
and no new stress is developed in these locations strated recently by Yanar et al.117 who found an
in the less-rough specimen, Figure 25(b). There is approximate factor 10 increase in the lifetime of a
a large increase, however, in the tensile stress devel- NiCoCrAlY/EB-PVD YSZ system by hand polishing
oped in the topcoat in the valley regions and across the BC prior to deposition of the topcoat.
the TBC–TGO interface. For the rougher surface,
this is 700 MPa but is appreciably smaller (200 1.27.5.4.4 Chemical failure
MPa) for the smoother surface. Significant tensile A discussion of CF processes in relation to aluminum
stresses are also developed at the TGO–BC inter- depletion and the endurance of overlay coatings has
face during cooling as a result of differential con- been given earlier in Section 1.27.4.3.2. Identical con-
traction strains between oxide and coating (see cepts apply to BCs in TBC systems but, in this case,
Ref. [37] for a general review of such issues). Broadly, spallation of the YSZ topcoat may be induced prior to
the same behavior has also been found with MCrAlY complete CF of the BC.105 This can arise because
BCs where, again, appropriate creep and oxide CF will result in the fast growth of nonalumina, break-
growth kinetics were incorporated into the FE away-type oxides which will produce upward displa-
modeling.113 cements of the topcoat. This is an oxide-jacking
σ22 (MPa)
717
500
YSZ 375
250
105
0.0
−407
TGO
(3.3 μm)
BC
X2
X1
(a) b/a = 0.52 (b) b/a = 0.25

Figure 25 FE contour plots of the s22 stress for the same specimen as shown in Figure 24 but after cooling to 25 C
after exposure at 1100 C for 100 h.133 The arrows indicate the general areas in which out-of-plane stresses had developed
prior to cooling.
process which will result in the development of out-of-

Ceramic
plane tensile stresses within the topcoat when the CF topcoat
sites are discrete (Figure 26). Possible localized sites
for enhanced aluminum depletion and the early initi- Tensile stress
at periphery of
ation of CF are diffusion cells (splats) within sprayed oxide mound
MCrAlY BCs,105 protuberances (roughness) at the BC
surface,109,111,153 and locally thin cross-sections.
The jacking associated with the formation of the
mound of fast-growing (Ni, Cr)-rich breakaway oxides,
shown schematically in Figure 26, will tend to nucle- Mound of
Alumina
ate and propagate delamination cracks within the top- breakaway
coat near the TGO. This process will occur at the oxide
oxidation temperature and will not need thermal cycles Bondcoat
to produce this damage. It is likely that this type of
process was responsible for the long crack (marked A) Figure 26 A schematic diagram showing how the growth
in the cross-section shown in Figure 23. This particu- of fast-growing ‘breakaway’ oxides resulting from CF will
lar specimen had been exposed isothermally until the generate out-of-plane tensile stresses within the topcoat.
final cooling transient and several discrete regions of
breakaway oxides (marked B), associated, in this case, importance of others (phase transformation strains,
with BC protuberances, had formed. rumpling) to TBC failure is less clear and awaits
more robust experimental support. Of those mechan-
isms which have been shown to nucleate and propa-
1.27.5.5 Surface Roughness: A Common
gate subcritical cracks, some are either aided by or,
Factor?
indeed, require BC surface roughness (e.g., CF, out-
A number of the degradation mechanisms described of-plane tensile stress development at temperature or
in Section 1.27.5.4 have been observed experi- during cooling, ratcheting) and others (wedging) are
mentally and, consequently, have validity. The likely to operate only on relatively flat interfaces. The
experiments of Yanar et al.,117 in particular, show very shown schematically in Figure 27 which relates
well that decreasing the surface roughness of a TBC lifetimes to the BC surface roughness.
MCrAlY BC can increase the lifetime of a TBC Relatively flat surfaces (low values of b/a) provide
system by an order of magnitude. The improved the longest lifetimes since opportunities for crack
behavior was similar in their experiments to the behav- nucleation caused by continuity strains and CF will
ior of a Pt-aluminide BC which had similarly flat BC be limited. Studies of failure processes in such ideal
surfaces. As can be appreciated from Figure 22, how- structures are limited although in localized areas of
ever, flat Pt-aluminide BCs with EB-PVD topcoats otherwise rough BCs, wedge-type cracks have been
need not always be the case and will depend on the found (Figure 23 but see also Section 1.27.4.2).
extent of grit blasting performed by the manufacturer. The red line, representing continuity cracking in
The factor 10 improvement in lifetimes found by Figure 27, shows a decrease in lifetimes as BC
Yanar et al. is of the same order of magnitude found roughness increases. It refers to the development of
in the compilation of published lifetime results from subcritical cracks at the flanks of BC protuberances at
commercial systems shown in Figure 19. It is certainly temperature, possible ratcheting nucleation, and the
tempting to suggest that much of this scatter in life- nucleation of cracks at the TGO/BC interface during
times can be attributed to an uncontrolled variation in cooling. The stresses produced increase with oxide
the surface roughness of the BCs. There will be other thickness and, so, time at temperature is an important
sources of scatter, inevitably included within these factor. They will also increase with the roughness
various tests, but the first-order importance of BC factor b/a (Section 1.27.5.4.3)113 and it is expected
surface roughness to the scatter at any given tempera- that TBC lifetimes will decrease as this increases. For
ture can be taken as a working hypothesis.113 large values of b/a, aluminum depletion within BC
The dependence of the various delamination protuberances will be enhanced (e.g., Figure 18) and
mechanisms on surface roughness are only now CF with the formation of breakaway-type oxides will
being assessed quantitatively using realistic numeri- take place. These will increase the stresses produced
cal models which, as described in earlier sections, within the TBC layer at temperature and failure will
allow for creep processes and the volume expansion occur earlier than in the absence of breakaway. This
due to oxide formation. Work in this area is incom- trend is shown as the downward-sloping blue line in
plete but some rather qualitative trend lines are Figure 27. Again, time at temperature will also be an
Wedge
failure
Chemical failure
Wedge cracking along BC interface
Continuity
cracking
Time to failure
Cracking in topcoat due

to continuity strains
and ratcheting
Continuity cracking in
topcoat enhanced by
chemical failure of BC
Increasing roughness, b/a
Figure 27 A schematic map showing how variously proposed mechanisms for the nucleation and growth of subcritical
cracks may depend on the surface roughness of the BC. The colored lines show possible dependences of continuity cracking
and CF on roughness.
important parameter since the depletion of alumi- 19. Taylor, A.; Floyd, R. W. J. Inst. Met. 1952–53, 81,
451–464.
num will be directly related to this. 20. Muñoz-Arroyo, R.; Clemens, D.; Tietz, F.; Anton, R.;
As a final note of caution, it should be emphasized Quadakkers, W. J.; Singheiser, L. Mater. Sci. Forum
that the discussion in this section relates to typical 2001, 369–372, 165–172.
21. Brindley, W. J.; Whittenberger, J. D. Mater. Sci. Eng. A
laboratory experience in which the test specimens 1993, 163, 33–41.
have relatively simple shapes. The map of Figure 27 22. Taylor, M. P.; Evans, H. E.; Ponton, C. B.; Nicholls, J. R.
then provides a mechanistic framework for these con- Surf. Coat. Technol. 2000, 124, 13–18.
23. Fritscher, K.; Leyens, C.; Peters, M. Mater. Sci. Eng. A
ditions but the concepts are also expected to be valid 1995, 190, 253–258.
for service conditions although some modification 24. Gupta, D. K.; Duval, D. S. US Patent 1986, 4, 585, 481.
might well be required. There may be thin regions, 25. Goebel, J. A.; Giggins, C. S.; Krasij, M.; Stringer, J. EPRI
Report 2639SR; Palo Alto, CA, 1981.
for example, within the complex geometry of actual 26. Chan, W. Y.; Evans, H. E.; Ponton, C. B.; Nicholls, J. R.;
components in which the local total aluminum reser- Simms, N. J. Mater. High Temp. 2000, 17, 173–178.
voir is insufficient to maintain a protective TGO. Bulk 27. Leyens, C.; van Liere, J.-W.; Peters, M.; Kaysser, W. A.
Surf. Coat. Technol. 1998, 108–109, 30–35.
depletion and breakaway oxidation could then occur 28. Leyens, C.; Peters, M.; Kaysser, W. A. Mater. Sci. Forum
prematurely and failure by this route might then be 1997, 251–254, 769–776.
only weakly dependent on the surface roughness of the 29. Gray, S.; Ponton, C. B.; Jacobs, M. H.; Evans, H. E.
United States Patent 2002, 6, 387, 541 B1, May 14.
BC. In another example, local regions of components 30. Martinz, H.-P.; Rieger, M. Mater. Sci. Forum 1997,
may be designed with a high radius of convex curvature 251–254, 761–768.
and could develop significant out-of-plane tensile 31. Fukomoto, M.; Matsumara, Y.; Hayashi, S.;
Sakamoto, K.; Kasama, A.; Tanaka, R.; Narita, T. Oxid.
strains during cooling. These would contribute to the Met. 2003, 60, 335–346.
continuity strains developed solely from the surface 32. Bennett, M. J. In Coatings for High Temperature
roughness of the BC. As discussed earlier, similar issues Applications; Lang, E., Ed.; Applied Science, 1983;
pp 169, 192.
would apply for the case of imposed service strains. 33. Crouch, A. G.; Dooley, R. B. Corros. Sci. 1976, 16,
341–347.
34. Coley, K. S.; Tuson, A. T.; Saunders, S. R. J.; Bennett, M. J.;
References Knights, C. F. Mater. Sci. Eng. A 1989, 121, 461–466.
35. Bennett, M. J.; Tuson, A. T.; Knights, C. F.; Ayres, C. F.
Mater. Sci. Technol. 1989, 5, 841–852.
1. Reed, R. C. The Superalloys: Fundamentals and 36. Pennisi, F. J.; Gupta, D. K. Thin Solid Films 1981, 84,
Applications; Cambridge University Press: Cambridge, 49–58.
2006. 37. Lobb, R. C.; Bennett, M. J. Oxid. Met. 1991, 35, 35–52.
2. Evans, H. E.; Taylor, M. P. Proc. IMechE, Part G: 38. Evans, H. E. Int. Mater. Rev. 1995, 40, 1–40.
J. Aerospace Eng. 2006, 220, 1–10. 39. Nicholls, J. R.; Lawson, K. J.; Al Yasiri, L. H.; Hancock, P.
3. Giggins, G. S.; Pettit, F. S. J. Electrochem. Soc. 1971, Corros. Sci. 1993, 35, 1209–1223.
118, 1782–1790. 40. Nicholls, J. R.; Simms, N. J.; Neseyif, S.; Ponton, C.;
4. Bennett, M. J.; Tuson, A. T.; Knights, C. F.; Ayres, C. F. Evans, H. E.; Taylor, M. In High Temperature Corrosion
Mater. Sci. Technol. 1989, 5, 841–852. and Materials Chemistry; The Electrochemical Society
5. Chan, W. Y.; Evans, H. E.; Ponton, C. B.; Nicholls, J. R.; Proceedings; McNallan, M., Opila, E., Maruyama, T.,
Simms, N. J. Mater. High Temp. 2000, 17, 173–178. Narita, T., Eds.; 2000; Vol, 99–38, pp 270–281.
6. Akhtar, A.; Hook, M. S.; Reed, R. C. Metall. Mater. Trans. 41. Nicholls, J. R.; Simms, N. J.; Chan, W. Y.; Evans, H. E.
A 2005, 36A, 3001–3017. Surf. Coat. Technol. 2002, 149, 236–244.
7. Luthra, K. L.; Shores, D. A. J. Electrochem. Soc. 1980, 42. Choquet, P. A.; Naylor, E. R.; Rapp, R. A. Mater. Sci. Eng.
127, 2202–2210. A 1989, 121, 413–418.
8. Strife, J. R.; Sheehan, J. E. Ceram. Bull. 1988, 67, 369. 43. Pérez, F. J.; Trilleros, J. A.; Hierro, M. P.; Milewska, A.;
9. Wagner, C. J. Electrochem. Soc. 1952, 99, 369–380. Carpintero, M. C.; Bolivar, F. J. Mater. Sci. Forum 2004,
10. Crank, J. The Mathematics of Diffusion; Oxford 461–464, 313–320.
University Press: Oxford, 1975. 44. Harper, M. A.; Rapp, R. A. In Surface Modification
11. Evans, H. E.; Hilton, D. A.; Holm, R. A.; Webster, S. J. Technologies; Sudarshan, T. S., Bhat, D. G., Jeandin, M.,
Oxid. Met. 1980, 14, 235–247. Eds.; AIME, 1991; Vol. IV, 415–428.
12. Talboom, F. T.; Elam, R. C.; Wilson, L. W. NASA Report 45. Lehnert, G.; Meinhardt, H. Surface Treatment 1972
CR7813, NASA: Houston, 1970. 1, 72.
13. Saunders, S. R. J.; Nicholls, J. R. Mater. Sci. Technol. 46. Meelu, M. C.; McMordie, B. G.; Loretto, M. H.; Jones, A.
1989, 5, 780–798. In The Processing, Properties and Applications of
14. Mévrel, R. Mater. Sci. Eng. A 1989, 120, 13–24. Metallic and Ceramic Materials; Loretto, M. H.,
15. Hocking, G.; Vasantasree, V.; Sidkey, P. S. Metallic and Beevers, C. J., Eds.; The University of Birmingham,
Ceramic Coatings; Longman: Harrow, 1989. Birmingham, UK, 1992; pp 1247–1252.
16. Stern, K. H. Ed., Metallurgical and Ceramic Protective 47. Gleeson, B.; Wang, W.; Hayashi, S.; Sordelet, D. Mater.
Coatings; Chapman and Hall: London, 1996. Sci. Forum 2004, 461–464, 213–222.
17. Goward, G. W. Surf. Coat. Technol. 1998, 108–109, 48. Cadoret, Y.; Bacos, M.-P.; Josso, P.; Maurice, V.;
73–79. Marcus, P.; Zanna, S. Mater. Sci. Forum 2004, 461–464,
18. Coddet, C. Mater. Sci. Forum 2004, 461–464, 193–212. 247–254.
49. Zhang, Y.; Haynes, J. A.; Lee, W. Y.; Wright, I. G.; Pint, B. A.; 80. Pragnell, W. M.; Evans, H. E.; Naumenko, D.;
Cooley, K. M.; Liaw, P. K. Metall. Mater. Trans. 2001, 32A, Quadakkers, W. J. Mater. High Temp. 2005, 22, 561–566.
1727. 81. Miller, R. A. Trans. ASME 1987, 109, 448–451.
50. Rickerby, D. S.; Bell, S. R.; Wing, R. G. United States 82. Renusch, D.; Echsler, H.; Schütze, M. In Lifetime
Patent 1999, 5, 981 091, November 9. Modelling of High Temperature Corrosion Processes;
51. Bouhanek, K.; Adesanya, O. A.; Stott, F. H.; Skeldon, P.; Schütze, M., Quadakkers, W. J., Nicholls, J. R., Eds.;
Lees, D. G.; Wood, G. C. Mater. Sci. Forum 2001, Maney, Publishing: London, 2001; pp 324–336.
369–372, 615–622. 83. Renusch, D.; Echsler, H.; Schütze, M. Mater. Sci. Forum
52. Adesanya, O. A.; Bouhanek, K.; Stott, F. H.; Skeldon, P.; 2004, 461–464, 729–736.
Lees, D. G.; Wood, G. C. Mater. Sci. Forum 2001, 84. Tien, J. K.; Davidson, J. M. In Stress Effects and the
369–372, 639–646. Oxidation of Metals; Cathcart, J. V., Ed.; AIME,
53. Fitzer, E. Archiv. Eisenhüttenwesen 1954, 25, 45–5. New York, 197; pp. 200–219.
54. Grünling, H. W.; Bauer, R. Thin Solid Films 1982, 95, 85. Lipkin, D. M.; Clarke, D. R.; Hollatz, M.; Bobeth, M.;
3–20. Pompe, W. Corros. Sci. 1997, 39, 231–242.
55. Itzhak, D.; Tuller, F. R.; Schieber, M. Thin Solid Films 86. Echsler, H.; Martinez, E. A.; Singheiser, L.;
1980, 73, 379–384. Quadakkers, W. J. Mater. Sci. Eng. A 2004, 384, 1–11.
56. Young, S. G.; Deadmore, D. L. Thin Solid Films 1980, 73, 87. Timoshenko, S. Theory of Elastic Stability; McGraw-Hill:
373–378. New York, 1936; 367.
57. Cockeram, B. V.; Rapp, R. A. Oxid. Met. 1995, 45, 88. Hou, P. Y. Mater. Sci. Forum 2001, 369–372, 23–28.
375–425. 89. Smialek, J. L.; Pint, B. A. Mater. Sci. Forum 2001, 369–372,
58. Gaillard-Allemand, B.; Vilasi, M.; Belmonte, T.; 459–466.
Steinmentz, J. Mater. Sci. Forum 2001, 369–372, 90. Banks, J. P.; Gohil, D. D.; Evans, H. E.; Hall, D. J.;
727–734. Saunders, S. R. J. In Materials for Advanced Power
59. Li, X. Y.; Taniguchi, S.; Matsunaga, Y.; Nayagawa, K.; Engineering; Coutsouradis, D., et al., Eds.;
Fujita, K. Intermetallics 2003, 11, 143–150. Kluwer Academic: The Netherlands, 1994; Part II,
60. Voytĕch, D.; Kubatik, T.; Jurek, K.; Maixner, J. Oxid. Met. pp 1543–1552.
2005, 63, 305–323. 91. Pint, B. A.; Wright, I. G.; Lee, W. Y.; Zhang, Y.;
61. Cockeram, B. V.; Rapp, R. A. Mater. Sci. Forum 1997, Prüßner; Alexander, K. B. Mater. Sci. Eng. 1998, 245,
251–254, 723–736. 201–211.
62. Gray, S.; Jacobs, M. H.; Ponton, C. B.; Voice, W.; 92. Evans, H. E.; Mitchell, G. P.; Lobb, R. C.; Owen, D. R. J.
Evans, H. E. Mater. Sci. Eng. A 2004, 384, 77–82. Proc. R. Soc. A 1993, 440, 1–22.
63. Kofstad, P. High Temperature Corrosion; Elsevier, 93. Evans, H. E.; Strawbridge, A.; Carolan, R. A.; Ponton, C.
1988. B. Mater. Sci. Eng. A 1997, 225, 1–8.
64. Deal, B. E.; Grove, A. S. J. Appl. Phys. 1965, 36, 94. Evans, H. E.; Osgerby, S.; Saunders, S. R. J. In Proceedings
3770–3778. of John Stringer Symposium on High Temperature
65. Taniguchi, S.; Minamida, T.; Shibata, T. Mater. Sci. Corrosion; Tortorelli, P. F., Wright, I. G., Hou, P. Y., Eds.;
Forum 1997, 251–254, 227–234. ASM International: Ohio, 2003; pp 122–130.
66. Wortman, D. J.; Nagaraj, B. A.; Dunderstadt, E. C. Mater. 95. Evans, H. E.; Lobb, R. C. Corros. Sci. 1984, 24, 209–222.
Sci. Eng. A 1989, 121, 433–440. 96. Galerie, A.; Dupeux, M.; Wouters, Y.; Toscan, F. Mater.
67. Rhys-Jones, T. N.; Toriz, F. C. High Temp. Technol. Sci. Forum 2006, 522–523, 441–450.
1989, 7, 73–81. 97. Bouhanek, K.; Oquab, D.; Pieraggi, B. Mater. Sci. Forum
68. Saunders, S. R. J.; Banks, J. P.; Chunnilall, C. NPL 1997, 251–254, 33–40.
Report MATC(A)113, An evaluation of thermal barrier 98. Baleix, S.; Bernhart, G.; Lours, P. Mater. Sci. Forum
coated specimens exposed to long-term ageing tests, 2001, 369–372, 539–546.
Teddington, July 2002. 99. Osgerby, S.; Berriche-Bouhanek, K.; Evans, H. E. Mater.
69. Munro, T. C.; Gleeson, B. Mater. Sci. Forum 1997, Sci. Eng. A 2005, 412, 182–190.
251–254, 753–760. 100. Chan, K. S.; Cheruvu, N. S.; Leverant, G. R. Mater. Sci.
70. Mabuchi, H.; Asai, T.; Nakayama, Y. Scripta Metall. 1989, Forum 2001, 369–372, 623–630.
23, 685–689. 101. Anton, R.; Birkner, J.; Czech, N.; Stamm, W. Mater. Sci.
71. Gauthier, V.; Dettenwanger, F.; Schütze, M.; Forum 2001, 369–372, 719–726.
Shemet, V.; Quadakkers, W. J. Oxid. Met. 2003, 59, 102. Taylor, M. P.; Pragnell, W. M.; Evans, H. E. Mater. Sci.
233–254. Forum 2004, 461–464, 239–246.
72. Gauthier, V.; Dettenwanger, F.; Schütze, M. 103. Sato, A.; Harada, H.; Kawagishi, K. Mater. Sci. Forum
Intermetallics 2002, 10, 667–674. 2006, 522–523, 361–368.
73. Czech, N.; Kolarik, V.; Quadakkers, W. J.; Stamm, W. 104. Matsumaru, H.; Hayashi, S.; Narita, T. Mater. Sci. Forum
Surf. Eng. 1997, 13, 384–388. 2006, 522–523, 285–292.
74. Naumenko, D.; Quadakkers, W. J.; Galerie, A.; 105. Evans, H. E.; Taylor, M. P. Oxid. Met. 2001, 55, 17–34.
Wouters, Y.; Jourdain, S. Mater. High Temp. 2003, 20, 106. Nesbitt, J. A.; Heckel, R. W. Metall. Trans. 1987, 18A, 2075.
287–293. 107. Roper, G. W.; Whittle, D. P. Met. Sci. 1981 15, 148.
75. Hindam, H.; Whittle, D. P. Oxid. Met. 1982, 18, 245–284. 108. Evans, H. E.; Donaldson, A. T.; Gilmour, T. C. Oxid. Met.
76. Natesan, K.; Liu, Y. Y. Mater. Sci. Eng. A 1989, 121, 1999, 52, 379–402.
571–580. 109. Niranatlumpong, P.; Ponton, C. B.; Evans, H. E. Oxid.
77. Clemens, D.; Vosberg, V. R.; Penkalla, H. J.; Breuer, U.; Met. 2000, 53, 241–258.
Quadakkers, W. J.; Nickel, H. Fresenius J. Anal. Chem. 110. Taylor, M. P.; Evans, H. E. Mater. High Temp. 2003, 20,
1997, 358, 122–126. 461–466.
78. Evans, H. E.; Norfolk, D. J.; Swan, T. J. Electrochem. 111. Taylor, M.P.; Pragnell, W.M.; Evans, H.E. Mater. Corros.
Soc. 1978, 115, 1180–1185. 2008, 59, 508–513.
79. Sabol, G. P.; Dalgaard, S. B. J. Electrochem. Soc. 1975, 112. Evans, H. E.; Taylor, M. P. J. Corros. Sci. Eng. 2003, 6,
122, 316–317. Paper H011.
113. Busso, E. P.; Wright, L.; Evans, H. E.; McCartney, L. N.; 134. Reid, M.; Pomeroy, M. J.; Robinson, J. S. Mater. Sci.
Saunders, S. R. J.; Osgerby, S.; Nunn, J. Acta Mater. Forum 2004, 461–464, 343–350.
2007, 55, 1491–1503. 135. Mumm, D. R.; Evans, A. G.; Spitsberg, I. T. Acta Mater.
114. Yanar, N. M.; Meier, G. H.; Pettit, F. S. Scripta Mater. in press.
2002, 46, 325–330. 136. Rosen, S.; Goebel, J. A. Trans. AIME 1968, 242, 722.
115. Lih, W.; Chang, E.; Wu, B. C.; Chao, C. H. Oxid. Met. 137. Smialek, J. L. Metall. Trans. 1971, 2, 913.
1991, 36, 221–238. 138. Zhang, Y.; Haynes, J. A.; Pint, B. A.; Wright, I. G.; Lee, W. Y.
116. Nijdam, T. J.; Jeurgens, L. P. H.; Chen, J. H.; Sloof, W. G. Surf. Coat. Technol. 2003, 163–164, 19–24.
Oxid. Met. 2005, 64, 355–377. 139. Chen, M. W.; Ott, R. T.; Hufnagel, T. C.; Wright, P. K.;
117. Yanar, N. M.; Pettit, F. S.; Meier, G. H. Metall. Mater. Hemker, K. J. Surf. Coat. Technol. 2003, 163–164,
Trans. 2006, 37A, 1563–1580. 25–30.
118. Tolpygo, V. K.; Clarke, D. R.; Murphy, K. S. Surf. Coat. 140. Smialek, J. L.; Hehemann, R. F. Metall. Trans. 1973, 4,
Tech. 2001, 146–147, 124–131. 1571.
119. Kim, G. M.; Yanar, N. M.; Hewitt, E. N.; Pettit, F. S.; 141. Karlsson, A. M.; Hutchinson, J. W.; Evans, A. G. Mater.
Meier, G. H. Scripta Mater. 2002, 46, 489–495. Sci. Eng. A 2003, 351, 244–257.
120. Evans, A. G.; Mumm, D. R.; Hutchinson, J. W.; Meier, G. H.; 142. Pan, D.; Chen, M. W.; Wright, P. K.; Hemker, K. J. Acta
Pettit, F. S. Prog. Mater. Sci. 2001, 46, 505–553. Mater. 2003, 51, 2205–2217.
121. Rouzou, I.; Molins, R.; Rémy, L.; Jomard, F. Mater. Sci. 143. Taylor, M. P.; Evans, H. E.; Busso, E. P.; Qian, Z. Q. Acta
Forum 2004, 461–464, 101–108. Mater. 2006, 54, 3241–3252.
122. Padture, N. P.; Gell, M.; Jordan, E. H. Science 2002, 296, 144. Deb, P.; Boone, D. H.; Manley, T. F. J. Vac. Sci. Technol.
280–284. A 1987, 5, 3366–3372.
123. Choi, S. R.; Hutchinson, J. W.; Evans, A. G. Mech. Mater. 145. Tolpygo, V. K.; Clarke, D. R. Acta Mater. 2004, 52,
1999, 31, 431–447. 5129–5141.
124. Nunn, J.; Saunders, S. R. J.; Banks, J. P. In Microscopy 146. Suo, Z. J. Mech. Phys. Solids 1995, 43, 829–846.
of Oxidation Tatlock, G. J., Evans, H. E., Eds.; Science 147. He, M. Y.; Evans, A. G.; Hutchinson, J. W. Acta Mater.
Reviews, 2006; Vol 6, pp 219–226. 2000, 48, 2593–2601.
125. Md Shawkat Ali; Shenhua Song; Ping Xiao J. Eur. Ceram. 148. Miller, R. A.; Lowell, C. E. Thin Solid Films 1982, 95,
Soc. 2002, 22, 101–107. 265–273.
126. Christensen, R.; Lipkin, D. M.; Clarke, D. R.; Murphy, K. 149. Bartlett, A. H.; Maschio, R. D. J. Am. Ceram. Soc. 1995,
Appl. Phys. Lett. 1996, 69, 3754–3756. 78, 1018–1024.
127. Peng, X.; Clarke, D. R. J. Am. Ceram. Soc. 2000, 83, 150. Chang, G. C.; Phucharoen, W.; Miller, R. A. Surf. Coat.
1165–1170. Technol. 1987, 32, 307–325.
128. Tolpygo, V. K.; Clarke, D. R.; Murphy, K. S. Surf. Coat. 151. Freborg, A. M.; Ferguson, B. L.; Brindley, W. J.;
Technol. 2001, 146, 124–131. Petrus, G. J. Mater. Sci. Eng.A 1998, 245, 182–190.
129. Selçuk, A.; Atkinson, A. A. Acta Mater. 2003, 51, 535–549. 152. Kerkhoff, G.; Vaßen; Stover, D. In Cyclic Oxidation of
130. Busso, E. P.; Qian, Z. Q. Acta Mater. 2006, 54, 325–338. High Temperature Materials; Schütze, M.,
131. Wellman, R. G.; Nicholls, J. R. Surf. Coat. Technol., 2004, Quadakkers, W. J., Eds.; IOM Communications: London,
177, 80–88. 1999; pp 373–382.
132. Tolpygo, V. K.; Clarke, D. R. Acta Mater. 2000, 48, 153. Hsueh, C.-H.; Haynes, J. A.; Lance, M. J.; Becher, P. F.;
3283–3293. Ferber, M. K.; Fuller, E. R., Jr.; Langer, S. A.;
133. Busso, E.P.; Qian, Z.Q.; Taylor, M.P.; Evans, H.E. Acta Carter, W. C.; Cannon, W. R. J. Am. Ceram. Soc. 1999,
Mater., in press. 82, 1073–1075.

Alumina Preparia

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Alumina Preparia

Uploaded by

Copyright:

Available Formats

SHREIR’S CORROSION

This page intentionally left blank

Amsterdam • Boston • Heidelberg • London • New York • Oxford

# 2010 Elsevier Ltd. All rights reserved

First Edition 1963

British Library Cataloguing in Publication Data

Library of Congress Catalogue Number: 2009940135

For information on all Elsevier publications

Printed and bound in Spain

Contributors vii CHAPTER 1.08

CHAPTER 1.18 CHAPTER 1.23

CHAPTER 1.22 CHAPTER 1.27

Michael Graham graduated from the University of Liverpool with BSc

thin film formation on metals and semiconductors, high temperature

Robert Lindsay received a BSc in Chemistry from the University

Stuart Lyon is Professor of Corrosion Control Technology and Head of

Tony Richardson graduated from UMIST in the 1960s with BSc

Howard Stott graduated from the University of Cambridge with a BA

PART II CHAPTER 1.15

CHAPTER 1.07 CHAPTER 1.16

CHAPTER 1.08 CHAPTER 1.17

CHAPTER 1.23 CHAPTER 2.05

CHAPTER 2.12 CHAPTER 2.25

CHAPTER 2.13 CHAPTER 2.26

CHAPTER 2.16 PART IV

Atmospheric Corrosion 1051 EXPERIMENTAL TECHNIQUES FOR

CHAPTER 2.17 CHAPTER 2.30

CHAPTER 2.18 CHAPTER 2.31

PART V CHAPTER 3.06

PART I CHAPTER 3.13

CHAPTER 3.04 CHAPTER 3.17

CHAPTER 3.05 CHAPTER 3.19

CHAPTER 3.20 CHAPTER 3.32

CHAPTER 3.21 CHAPTER 3.33

CHAPTER 3.23 PART V

CHAPTER 3.24 CHAPTER 3.36

CHAPTER 3.27 PART I

CHAPTER 4.05 PART IV

PART VI CHAPTER 4.38

CHAPTER 4.36 CHAPTER 4.43

G E Thompson, OBE FREng,

G E Thompson, OBE FREng,

R The gas constant (8.314 J K1 mol1)

chemical composition gives DG ¼ ðcmC þ d mD Þ ðamA þ bmB Þ < 0 ½51

where DG is the change in free enthalpy. It follows that S ui mi

emf of a cell or the standard equilibrium potential

Cu2þ þ 2Hg ! Hg2þ

which shows how the equilibrium potential E is so that

½81 pressures used for the reaction, and the equation

1.01.4.1 Sign Convention for Equilibrium or

ß 2010 Elsevier B.V. All rights reserved.

IUPAC International union of pure and applied

OHP Outer Helmholtz plane Rp Polarization resistance (O)

Sc NiOHads !Ni2þ þ OH þ e ½10

Compact Diffuse layer

absence of change in the reaction free energy, in

ΔFint = Fmetal – Fsolution

Pt electrode Ag electrode Au3+/Au 1.43

region. It is important to appreciate that the position of E E 2

1.02.3.5.2 Purely pH-dependent equilibrium

and O2/H2O, characterized at 25  C by the following 2.8

are the hydrogen evolution (in acidic media) and the

a Note that both cathodic reactions yield an alkaliniza-

and O2/H2O, characterized at 25 C by the following 2.8

For KI KISCC, an initiated crack will continue to