Anda di halaman 1dari 517

ISBN: 0-8247-0074-0

This book is printed on acid-free paper.


Headquarters
Marcel Dekker, Inc.
270 Madison Avenue, New York, NY 10016
tel: 212-696-9000; fax: 212-685-4540
Eastern Hemisphere Distribution
Marcel Dekker AG
Hutgasse 4, Postfach 812, CH-4001 Basel, Switzerland
tel: 41-61-261-8482; fax: 41-61-261-8896
World Wide Web
http://www.dekker.com
The publisher offers discounts on this book when ordered in bulk quantities. For more
information, write to Special Sales/Professional Marketing at the headquarters address
above.
Cover illustration: Scanning electron micrograph of an oxide/oxide composite exposed
to an environment of air and water at 1000C. Courtesy of C.A. Lewinsohn, Pacific Northwest National Laboratory.
Copyright 2001 by Marcel Dekker, Inc. All Rights Reserved.
Neither this book nor any part may be reproduced or transmitted in any form or by any
means, electronic or mechanical, including photocopying, microfilming, and recording,
or by any information storage and retrieval system, without permission in writing from
the publisher.
Current printing (last digit):
10 9 8 7 6 5 4 3 2 1
PRINTED IN THE UNITED STATES OF AMERICA

Preface

New materials, composites, and coatings are being developed at a rapid rate, and
there has been an increase in the substitution or replacement of one class of material
by another. More complex materials are being engineered and used in a new variety
of environments. Many materials are used in composite or coated forms to enhance
performance. Various combinations of metals, intermetallics, ceramics, and polymers are becoming more common. Composites with discontinuous, dispersed
phases within a matrix, and fiber and laminated reinforcements are being developed. Coatings also include combinations similar to those for bulk composite materials. Materials are being pushed to perform in a wider range of environments
than ever before. Aqueous and high-temperature environments, which may contain
varying amounts of corrosive species, are commonly encountered by advanced
materials. In other cases, the effect of environments such as water, solvents, wine,
and food thought to be relatively benign must be understood.
All these developments have made it difficult to locate information on the
effects of environment on the new materials, composites, and coatings. This comprehensive book describes such effects for a broad range of materials and environments,
filling the information gap and providing a comprehensive viewpoint for the scientist
or engineer interested in applying new materials to existing applications or old materials to new applications.
This book would not have been possible without the many hours given by each
contributor. Their effort and dedication are greatly appreciated. Also, the assistance of
B. H. Wardlow at PNNL in coordinating the manuscripts is greatly appreciated.
Russell H. Jones
iii

Contents

Preface
Contributors

iii
vii

I. Metallic Alloys
1. Ferrous Alloys (Ferritic and Martensitic)
Bruce Craig

2. Austenitic Stainless Steels


Russell H. Jones, Stephen M. Bruemmer, Mike J. Danielson,
and Bruce Craig

31

3. Nickel-Based Alloys for Resistance to Aqueous Corrosion


Paul Crook

55

4. Nickel-Based Alloys for Resistance to High-Temperature


Corrosion
Mark A. Harper and George Y. Lai
5. Corrosion of Copper and Its Alloys
Andrew James Brock

75

115
v

vi

Contents

6. Reactive and Refractory Alloys


Te-Lin Yau

151

7. Aluminum Alloys
N. J. Henry Holroyd

173

8. Magnesium Alloys
Mike J. Danielson

253

II. Intermetallic Alloys


9. Environmental Embrittlement of Nickel-Based and Iron-Based
Intermetallics
Norman S. Stoloff
III.

275

Ceramics

10. Nonoxide Ceramics


Nathan S. Jacobson and Elizabeth J. Opila

311

11. Oxide Ceramics


F. S. Pettit, G. H. Meier, and J. R. Blache`re

351

IV. Composites
12. Metal Matrix Composites
Russell H. Jones

375

13. Ceramic Matrix Composites


Russell H. Jones, C. H. Henager, Jr., Charles A. Lewinsohn,
and Charles F. Windisch, Jr.

391

14. Issues in Predicting Long-Term Environmental Degradation of


Fiber-Reinforced Plastics
Aaron Barkatt

419

V. Metallic Glasses
15. Amorphous and Nanocrystalline Alloys
Koji Hashimoto

459

Index

501

Contributors

Aaron Barkatt Department of Chemistry, The Catholic University of America,


Washington, D.C.
J. R. Blache`re Materials Science and Engineering Department, University of
Pittsburgh, Pittsburgh, Pennsylvania
Andrew James Brock
Haven, Connecticut

Metals Research Laboratories, Olin Corporation, New

Stephen M. Bruemmer Pacific Northwest National Laboratory, Richland,


Washington
Bruce Craig MetCorr, Denver, Colorado
Paul Crook
diana

Engineering and Technology, Haynes International, Kokomo, In-

Mike J. Danielson Pacific Northwest National Laboratory, Richland, Washington


Mark A. Harper Research and Development, Special Metals Corporation,
Huntington, West Virginia
vii

viii

Contributors

Koji Hashimoto Tohoku Institute of Technology, Sendai, Japan


C. H. Henager, Jr. Pacific Northwest National Laboratory, Richland, Washington
N. J. Henry Holroyd
erside, California

Research and Development, Luxfer Gas Cylinders, Riv-

Nathan S. Jacobson Materials Division, NASA Glenn Research Center, Cleveland, Ohio
Russell H. Jones Pacific Northwest National Laboratory, Richland, Washington
George Y. Lai Consultant, Carmel, Indiana
Charles A. Lewinsohn
Washington

Pacific Northwest National Laboratory, Richland,

G. H. Meier Materials Science and Engineering Department, University of


Pittsburgh, Pittsburgh, Pennsylvania
Elizabeth J. Opila Department of Chemical Engineering, Cleveland State University, Cleveland, Ohio
F. S. Pettit Materials Science and Engineering Department, University of Pittsburgh, Pittsburgh, Pennsylvania
Norman S. Stoloff Materials Science and Engineering Department, Rensselaer
Polytechnic Institute, Troy, New York
Charles F. Windisch, Jr.
Washington

Pacific Northwest National Laboratory, Richland,

Te-Lin Yau Te-Lin Yau Consultancy, Albany, Oregon

1
Ferrous Alloys
(Ferritic and Martensitic)
Bruce Craig
MetCorr, Denver, Colorado

I.

INTRODUCTION

This chapter addresses the corrosion behavior of ferrous alloys, specifically ferritic and martensitic irons and steels. The reason for this designation is to distinguish these alloys from the austenitic alloys that will be discussed in a later
chapter. However, the use of the terms ferritic or martensitic is not intended
to exclude pearlitic or bainitic microstructures, but is only intended as a convenience. Therefore, the discussion in this chapter addresses all low-alloy ferrous
materials and ferritic and martensitic stainless steels.
The largest group of ferrous alloys are steels which will be the emphasis
in this chapter, however, cast irons, several of which can be quite corrosion resistant, will also be mentioned. There are tens of thousands of different steels in
the world; however, they are usually referred to in groups as a function of their
chemical composition. Thus, carbon steels (also referred to as mild steels) contain
little or no alloy elements beyond the Mn, P, S, Si, and Al needed to produce a
good quality structural material. The low-alloy steels are the next group that can
be characterized by small additions of Cr, Mo, and Ni, usually in the range of
about 0.104.0% of each element, but generally less than 5% of the total alloying
elements. Higher additions of these elements form a group of steels referred to
as alloy steels. Generally, the alloying content is equal to or less than 10% (e.g.,
9 Cr1 Mo steel).
The distinction between low-alloy and alloy steels is not well defined nor
even well observed in practice. Often, all of these steels are lumped together
under the term low-alloy steel or alloy steel. As will be seen in this chapter,
1

Craig

this distinction is largely unnecessary from a corrosion standpoint because alloying of less than 10% for many environments is not sufficient to impart significant
corrosion resistance to steels.
In a similar vein, cast irons are used as structural or pressure-containing
alloys that have little natural corrosion resistance. Additions of Cr, Ni, and Si
are most often the primary means for improving corrosion resistance.
Unquestionably the most important alloying element in steels and irons
from a corrosion standpoint is Cr. Steels containing in excess of 11% Cr, will
display stainless (rust-resistant) qualities when exposed to the atmosphere; thus,
this group or, more properly, family of alloys is termed stainless steels. In
this chapter, the ferritic and martensitic stainless steels will also be discussed.
Although the number of alloys that are covered by the categories just presented are myriad, the general performance is relatively easy to address. The
corrosion performance of these ferrous alloys is of major importance not only
because they represent the largest tonnage of metals used by the world but because they represent the benchmark from which corrosion performance of other
alloys is compared.

II. CORROSION BEHAVIOR


A. General Corrosion
Carbon and low-alloy steels generally display active corrosion in the majority of
environments to which they are exposed. This means they will corrode unabated
at some corrosion rate determined by factors such as solution composition,
pH, fluid velocity, presence of oxidizers, temperature, and so forth. In many
of the environments to which ferrous alloys are exposed, there is little effect or
benefit of minor alloying element additions. Figure 1 illustrates the typical polarization behavior for steels in many environments. The anodic curve shows active corrosion with no tendency toward passivation. The environmental factors
mentioned earlier will determine the anodic and cathodic behaviors and ultimately the anodic current density (i.e., the corrosion rate). Figure 2 illustrates
the effect of increasing the conductivity of the solution (produced by increasing
chloride content) on the corrosion rate of carbon steel (1). Solution conductivity
plays a major role in the tendency for corrosion of alloys in a specific environment. For example, in many hydrocarbon environments, the conductivity and
polarizability of the solution are so low that corrosion cannot be established or
maintained. In these solutions, carbon steel is quite useful and cost-effective.
Likewise, in systems containing corrosive gases (i.e., CO2, H2S, etc.), if no water
is present, there is no electrolyte for corrosion, and carbon steels are adequate
for the service.

Ferrous Alloys

Fig. 1 Typical polarization behavior for mild steel under active corrosion.

In those environments in which corrosion of steel follows the behavior in


Fig. 1, other means of mitigating corrosion must be considered. These other
means are coatings, inhibitors, cathodic protection, or anodic protection. These
other methods are dealt with in more detail elsewhere (2).
In some very specific environments, steels may develop a protective corrosion product layer that essentially passivates the steel surface, reducing corrosion
to an acceptable level. Although the environments for which this phenomenon
occurs are few compared to those for active corrosion, they are notable. Examples
of such environments are steels exposed to concentrated sulfuric acid, hydrofluoric acid, and sodium hydroxide. In concentrated sulfuric acid, a soft protective
iron sulfate corrosion product is formed that inhibits further corrosion. However,

Craig

Fig. 2 Conductivity effects on mild steel in aqueous solutions, argon saturated, as a


function of NaCl content.

this film is not mechanically strong and is easily eroded. Thus, this film is not
suitable for exposure to high-velocity streams, yet it is beneficial from a sulfuric
acid storage standpoint because carbon steel containers can be used to handle
the acid under essentially static conditions.
Other environments produce this same behavior and Fig. 3 illustrates this

Ferrous Alloys

Fig. 3 Passive behavior of mild steel in environments such as Na2SO4.

development of passivity in the anodic curve that reflects a decrease in the anodic
corrosion current with the formation of a passive film (1).
Great care must be taken in applying this method of passivity, however,
because many factors in actual service can eliminate or degrade this protective
film, causing significant corrosion to occur. Velocity changes, temperature increases, the presence of impurities (i.e., chlorides), and concentration changes
can produce high-corrosion rates instead. A useful example of this change in

Craig

corrosion rate is carbon steel in 90% sulfuric acid at room temperature. Under
static conditions, the corrosion rate is about 0.5 mm/year. However, at a concentration of less than 50% H2SO4, the corrosion rate exceeds 5 mm/year. Thus, the
stability of the passive film is an important factor in the choice of any material
for a specific environment and that choice may be suitable only over a narrow
range of conditions. As will be discussed in later chapters, the stability of the
passive layer on nickel-based alloys, titanium alloys, and other materials is much
greater than for steels, thus the reason these alloys are more resistant to corrosive
environments. It is the great stability of the air-formed oxide on ferritic and martensitic stainless steels that produce their stainless quality when exposed to the
atmosphere. Yet, this passive film is not stable in all environments and care must
be taken in their application, as the oxide is particularly susceptible to attack by
halides.
B. Localized Corrosion
In addition to the uniform or general corrosion of ferrous alloys, there are numerous forms of localized corrosion that can cause failure of these alloys. Pitting
corrosion is a highly localized attack of the metal, creating pits of varying depth,
width, and number. Pitting may often lead to complete perforation of the metal
with little or no general corrosion of the surface. This can be a considerable
problem in steels and is one of the most common causes of failure for stainless
steels. At this time it is impossible to predict the remaining life of a pitted structure; thus, pitting remains one of the leading causes of failure for ferrous alloys.
Crevice corrosion is similar to pitting corrosion in its localized nature but
is associated with crevices. Stainless steels and some nickel-based alloys are particularly susceptible to this form of corrosion; however, steels are less susceptible
to this form of attack, except in aerated environments.
Intergranular corrosion is the preferential corrosion of grain boundaries in a
metal caused by prior thermal treatments and related to specific alloy chemistries,
especially in stainless steels and nickel alloys. Corrosion of this type is rare in
carbon and alloy steels but can be a problem in ferritic and martensitic stainless
steels.
Dealloying is the selective removal of one element (usually the least noble)
from an alloy by the corrosive environment. Also referred to as selective leaching
or dezincification, denickelification, and so forth, designating the element removed. Steels are not generally attacked by this mechanism, nor are ferritic or
martensitic stainless steels. However, some cast irons, especially gray iron, are
quite susceptible to dealloying. For gray cast iron, the graphite flakes are cathodic
to the surrounding ferritic matrix. Thus, the ferrite is selectively corroded away,
leaving a mechanically weak graphite structure.
Corrosion fatigue is the initiation and extension of cracks by the combined
action of an alternating stress and a corrosive environment. The introduction of

Ferrous Alloys

a corrosion environment often eliminates the fatigue limit of a ferrous alloy,


creating a finite life regardless of applied stress level. It is currently impossible
to predict the corrosion fatigue life of an alloy because of the difficulty in distinguishing the contributing effects of the corrosion portion and the mechanical
portion of corrosion fatigue.
Galvanic corrosion is the accelerated corrosion of the least noble metal
when coupled to one or more other metals. The more noble metals are protected
from corrosion by this action. This form of attack is one of the most common
causes of corrosion for all of the ferrous alloys, especially carbon and alloy steels.
More detail on this type of corrosion is provided later in this chapter.
Many forms of flow-assisted corrosion are often included under the term
erosioncorrosion such as cavitation, impingement, and corrosionerosion.
All of these types of attack are the result of accelerated corrosion due to flow of
solids, liquids, or gases, and the ferrous alloys are very susceptible to this form
of attack. Therefore, ferrous alloys are quite limited for applications where a
corrosive fluid, even one that is mildly corrosive, is combined with rapid flow.
Environmental cracking is the initiation and propagation of cracks by the
combined action of a corrosive environment and a tensile stress. Typically, under
anodic conditions, this form of attack is most often referred to as stress-corrosion
cracking (SCC). Generally, susceptibility to cracking increases with increasing
temperature, but not every alloy cracks in every environment. This form of corrosion causes significant damage to steels and stainless steels.
Another form of cracking is strictly related to hydrogen absorption into
ferrous alloys and the resultant cracking. In aqueous environments and in contrast
to SCC, this occurs under cathodic conditions. There are numerous forms of
damage associated with hydrogen, which are contained under the collective term
hydrogen damage (HD). For hydrogen embrittlement and hydrogen-stress
cracking, tensile stress and hydrogen atoms are necessary to cause failure. However, contrary to SCC, susceptibility is greatest near room temperature. Other
terms and forms are hydrogen-induced cracking (HIC), blistering, sulfide-stress
cracking (SSC), hydrogen stress-corrosion cracking, hydriding, and hydrogen attack. There are many other terms too numerous to mention. As with SCC, this
is a major problem in steels and martensitic stainless steels.
Although all of these corrosion mechanisms are of some concern for ferrous
alloys, the three most problematic and often observed forms are pitting, galvanic
corrosion, and environmental cracking (this term is frequently used to encompass
all forms of SCC and HD). Therefore, these three forms will be discussed in
greater detail as they relate to ferritic and martensitic steels.

1. Pitting Corrosion
Pitting corrosion is one of the most common and most insidious types of corrosion
attack on steels. Pitting may rapidly produce perforation of a metal or may take

Craig

many years to develop. Currently, there are no methods to accurately predict the
propagation rates of pits and, therefore, no valid means to estimate the remaining
life of a structure or component once pitting has initiated. There has been some
success in modeling pitting as a stochastic process, but, as yet, there is not an
accepted methodology. Because of this inability to predict pitting and remaining
service life, the primary focus in materials selection for a specific environment
is to choose a material that is either immune to a particular environment or at
least highly resistant to pitting in the first place. This can be a difficult task because often it is not the major component of the service environment that induces
pitting but rather the small concentration of some impurity that does. A good
example of this is shown in Table 1, where increasing the Cl content of H 2 SO4
requires a corresponding increase in Cr to the steel to resist pit initiation (3). As
discussed earlier, steels are generally resistant to concentrated H2SO4, but the
introduction of small amounts of Cl makes the solution particularly corrosive.
This same effect is observed for steels exposed to seawater. Seawater itself is
not very aggressive to ferrous alloys; however, it is the introduction of dissolved
oxygen that causes seawater to become corrosive, producing severe pitting attack.
Figure 4 illustrates the dramatic effect of oxygen, in only the parts-per-billion
(ppb) range, on the corrosion of steel (4).
The mechanism of pitting is well understood in a general sense. Pit initiation begins with the very localized breakdown of the passive film, leading to the
formation of a small pit bottom that acts as the anode and the remainder of the
passive surface as the cathode. Thus, there is a large driving force to continue
development and propagation of the pit. However, the nature of pitting is a selfsustained autocatalytic process that continues pit propagation. During the propagation process, the solution in the pit bottom becomes and remains very acidic,
further enhancing propagation. Moreover, the potential difference between the
steel surface and the pit bottom acts as a driving force for propagation. During

Table 1 Minimum Concentration of


Cl Necessary for Pit Initiation in 1N
H 2SO4 Solution
Alloy
Fe
Fe5.6 Cr
Fe11.6 Cr
18.6 Cr9.9 NiFe
20.0 CrFe
24.5 CrFe
29.4 CrFe

Cl (normality)
0.0003
0.017
0.069
0.1
0.1
1.0
1.0

Ferrous Alloys

Fig. 4 Effect of oxygen concentration on corrosion of mild steel in Pacific Ocean water.

this period, the original steel surface, which has not begun pitting, is effectively
protected from further corrosion by the resulting cathodic polarization.
The difficulty in predicting the remaining life of a structure during pitting
corrosion is due to continual pit initiation, propagation, repassivation, and repropagation. Not all pits in the same structure propagate at the same rate and
propagation is not linear but rather an exponential function.
It is generally recognized that pitting will initiate at microstructural discontinuities on the steel surface. These discontinuities can be grain boundaries,
second-phase particles, and so forth, but they are most often sulfide inclusions.
This latter feature is most commonly the origin for pits in stainless steels. Therefore, it is quite predictable that resulfurized steels, especially the resulfurized
stainless steels, will suffer pitting corrosion in an environment long before and
under less severe conditions than the lower sulfur version of the same steel. For
example, AISI 416 stainless steel, which contains 0.15% S minimum, compared

10

Craig

to its counterpart, AISI 410, which contains 0.030% S maximum, is highly susceptible to pitting corrosion and may pit in environments where AISI 410 does
not.
It is impossible to list all the environments in which steels pit because of
the numerous factors involved and the great variety of possible combinations of
chemicals. Moreover, great care must be taken when selecting an alloy for a
certain application in not simply scanning the large number of corrosion data
references and databases for alloys with low corrosion rates, as most of these
resources do not present pitting data, but rather provide only uniform corrosion
rates that can be very misleading. However, that said, it is often the case that
steels and, for that matter, many other alloys have a great tendency to pit in
environments that contain chlorides or, more generally, halides. Although chlorides are by far the most prevalent species in many environments bromides, iodides and fluorides can also induce pitting.
Therefore, the presence of halides in a process stream should be a signal
that pitting must be considered in the choice of alloys. Yet, the absence of these
species does not necessarily eliminate the possibility of pitting. An example of
pitting in the absence of halides is corrosion from CO2 gas dissolved in water.
This condition produces carbonic acid that can lead to pitting corrosion of carbon
and low-alloy steels. Of course, the situation becomes more complex as a function
of temperature and the introduction of chlorides. Figure 5 shows the envelope
of applicability of AISI Type 420 stainless steel (also referred to as 13 Cr) to a
combined environment of CO2 and chlorides as a function of temperature in the
absence of oxygen (5). Within this envelope, no pitting occurs and corrosion is
minimal but uniform. However, the introduction of small concentrations (ppb)
of oxygen creates a severe pitting attack of the 13 Cr even at ambient temperature,
thereby eliminating the use of this alloy.
Thus, prior experience or laboratory testing is often necessary to confirm
that a particular alloy will not be susceptible to pitting in a specific environment.

2. Environmental Cracking
As previously indicated, environmental cracking (EC) is a general term that encompasses all forms of cracking that are induced or accelerated by the service
environment. The two principal categories within this form of corrosion that are
pertinent to this discussion on ferrous alloys are SCC and HD. An in-depth review
of these types of cracking and their mechanisms can be found elsewhere (6). It
is simplest and consistent with much of the literature to discuss SCC in terms
of an active path corrosion coupled with a tensile stress (often referred to as
anodic cracking) and HD as all those forms of cracking that depend on hydrogen assistance (often referred to as cathodic cracking in aqueous environments,

Ferrous Alloys

11

Fig. 5 The corrosion resistance of 13 Cr (Type 420) stainless steel in CO2 /NaCl environments in the absence of O2.

hydrogen-stress cracking, HIC, blistering, and hydrogen embrittlement, to name


a few).
In many environments where steels are susceptible to SCC, a frequent precursor to crack initiation is pitting corrosion. In these environments when stresses,
either applied or residual, are relatively low, pitting ultimately produces failure.
However, as the stress level increases, SCC can become the controlling mode of
failure. Figure 6 illustrates this sequence of events (7). The important feature of
SCC is that cracking initiates and propagates at a subcritical level below the
scale of macroscopic flaws that would be considered critical from a linear elasticfracture mechanics (LEFM) standpoint. Therefore, LEFM by itself cannot be
used to predict the likelihood of EC.
One of the most significant factors affecting EC is the strength level of the
steel. High-strength steels are very susceptible in a variety of environments and
this susceptibility is a function of the yield strength. Figure 7 shows that many
steels fail in a simple marine environment at ambient temperature when the yield
strength exceeds about 180 ksi, regardless of alloy composition (8). However,
below 150 ksi yield strength, cracking does not occur in this environment. Figure

12

Craig

Fig. 6 Proposed sequence of crack initiation, coalescence, and growth for steels undergoing subcritical cracking in aqueous environments.

7 should not be construed to mean that ferrous alloys do not crack at all below
150 ksi, because quite low-strength steels are very susceptible to EC, just in
different environments. Moreover, the cracking of these high-strength steels in
seawater is thought to involve HD. Thus, it is convenient to further discuss the
EC of ferrous alloys in two groups; high-strength steels (150 ksi) and lowstrength steels (150 ksi).
a. Environmental Cracking of Low-Strength Steels Low-strength steels
(150 ksi yield strength) are quite susceptible to EC in certain specific environments. The yield strength of the steel in this strength range is not particularly
significant to the susceptibility to EC as it is for higher-strength steels. Rather,
other factors such as applied stress, steel composition, pH, solution composition,
potential, and temperature are much more critical. Increasing applied (or residual)
stress and increasing temperature enhances the SCC of low-strength steels as
does decreasing pH. Small concentrations of trace or impurity elements in the
alloy can have a profound effect on SCC of steels.
Some of the more common environments known to cause SCC of lowstrength steels are liquid NH3, CO2 /CO, carbonate/bicarbonate, hydroxide, nitrate

Ferrous Alloys

13

Fig. 7 Stress-corrosion behavior of steels exposed to marine atmosphere.

solutions, and amine solutions. Generally, as the concentration of the solution


increases, the susceptibility to SCC increases. Table 2 shows the effect of increasing nitrate concentration on the threshold stress of a plastically deformed low
carbonmanganese steel (9). The threshold stress (that stress below which cracking does not occur) decreases with increasing concentration of nitrate and is partially dependent on the specific cation associated with the nitrate anion.
Similar behavior has been observed for OH solutions and sufficient data
have been gathered to develop the useful engineering diagram shown in Fig. 8

Table 2 Threshold Stress Values (ksi) for Mild Steel in


Boiling Nitrate Solutions of Various Concentrations
Solution concentration
Nitrate

8N

4N

2.5N

1N

NH 4 NO3
Ca(NO3 ) 2
LiNO3
KNO3
NaNO3

2.2
5.6
5.6
6.7
9.0

3.4
7.8
9.0
4.5
9.5

7.8
13.4
21.3 (2N )
15.7
24.7

13.4
25.8
25.8
26.9
29.2

14

Craig

Fig. 8 Temperature and concentration limits for stress-corrosion cracking susceptibility


of carbon steels in caustic soda (NaOH).

(10). Area C can also be handled successfully with austenitic stainless steels.
This diagram illustrates the important effect of residual welding stresses on SCC
and the ability to extend the range of applicability of steels in OH simply by
reducing the residual stresses.
An area of great concern that has recently received increased attention is
the SCC of low-strength pipeline steels. The external SCC of pipeline steels has
occurred in two distinct environments. Early failures were in soil environments
that produced solutions of carbonate/bicarbonate with a pH of about 9.5 on the
outside of the pipe, causing intergranular SCC. Figure 9 shows the intergranular
SCC of a low-strength pipeline steel that failed in the high-pH environment. More
recently, transgranular SCC has been found to be the cause of several pipeline
failures. The pH in this latter case frequently falls in the range of 68. Many of

Ferrous Alloys

15

Fig. 9 Intergranular fracture of a low-strength pipeline steel from SCC. Magnification:


100.

the pipeline failures occurred in pipelines that are more than 20 years old that
have yield strengths around 52,000 psi. This long incubation time for crack initiation and propagation is typical for low-strength steels and in sharp contrast to
the often rapid initiation and fracture of high-strength steels. Yet, it would be
misleading to assume that SCC of low-strength steels is always a slow process.
Figure 10 shows that the crack growth rate in many low-strength steels is a strong
function of the solution composition and is directly related to the bare surface
current density (11). This current density, in combination with straining at the
crack tip, is the driving force for cracking and is frequently referred to as active
path or anodic cracking. It is generally believed, though not entirely agreed, that
SCC progresses by the rupture of the oxide film at the crack tip, thereby providing
a bare surface for the peak current to advance the crack tip a certain distance

16

Craig

Fig. 10 Crack velocities and peak current densities at the same potentials for a variety
of systems and alloys.

before the oxide re-forms and the crack arrests. These events may occur over
many cycles or just a few. However, it is important to recognize the difference
in this mode of cracking versus that due to cathodic cracking, where hydrogen
is the primary agent to assist cracking.
In general, martensitic stainless steels are used at higher strengths than
ferritic stainless steels because the former can be strengthened by heat treatment
and the latter cannot. Therefore, the martensitic stainless steels are discussed
under the high-strength section. Although ferritic stainless steels are generally
more resistant to SCC than austenitic stainless steels, especially in chloride solutions, they are not entirely immune. Small additions of Ni and plastic deformation
can each increase the tendency for SCC in chloride environments.
Hydrogen damage of low-strength steels typically occurs in steels that have
yield strengths less than 100 ksi. As with the SCC of low-strength steels, the
yield strength is not an important factor in HD. Moreover, residual and/or applied
stresses have little effect.

Ferrous Alloys

17

The primary cracking modes are stepwise cracking (SWC, also referred to
as HIC) and blistering (HB) or blister cracking. Both types of cracking are the
result of relatively high hydrogen input fugacities compared to the HD of highstrength steels and are often found together in the same steel. Both SWC and
HB are considered to occur by the classical hydrogen-pressure mechanism. According to this mechanism, hydrogen atoms enter the steel and combine at discontinuities (i.e., nonmetallic inclusions) to form molecular hydrogen, which is too
large of a molecule to diffuse back out of the steel. The molecules continue to
accumulate, increasing the local hydrogen pressure until a crack or blister forms.
Figure 11 shows an example of SWC in a low-strength steel exposed to H2S.
Hydrogen damage of steels occurs over the entire strength range of typical engineering applications. Figure 12 shows that regardless of the strength level of the
steel, some form of hydrogen cracking may occur and the only distinction is in
the morphology of cracking (12).
b. Environmental Cracking of High-Strength Steels The EC of highstrength steels (150 ksi) is highly dependent on strength, and in many environments, it is difficult to distinguish between the more classical HD and SCC mech-

Fig. 11 Stepwise cracking from hydrogen in a low-strength steel exposed to H2S. Magnification: 25.

18

Craig

Fig. 12 Critical hydrogen concentration in steel for cracking as a function of yield


strength and the morphology of cracking.

anisms. On the other hand, EC from caustic solutions are not obviously hydrogen
related, but the crack propagation rate is many times faster than for low-strength
steels in caustic.
Seawater and brackish waters do not typically produce EC failures of lowstrength steels, but they do produce the EC of high-strength steels, as demonstrated earlier in Fig. 7. Again, this is most likely a HD mechanism.
From an engineered materials sense, the actual mechanism is not as important as the fact that high-strength steels are so susceptible to EC, and the resulting
crack propagation rate so high that catastrophic failure in many otherwise benign
environments can easily occur. Because of this high risk of steel failure with
increasing strength beyond 150 ksi, it is common practice to select other alloys
and materials that have a greater overall corrosion resistance for high-strength
applications. These materials and their performance are dealt with in the remainder of this book.

Ferrous Alloys

19

As an example of the behavior of high-strength steels in some of these


environments, Fig. 13 shows the crack propagation rate of various alloy steels
(13). It is apparent that above 150 ksi yield strength, the crack growth rates are
so high that strength level becomes meaningless. Moreover, the speed at which
the crack propagates is too rapid for detection in actual service, often leading to
a catastrophic failure.
Martensitic stainless steels are generally resistant to chloride SCC when
heat treated to yield strengths less than about 100 ksi. However, above this yield
strength, they become increasingly susceptible to EC in seawater and H2S. Both
of these environments are known to produce hydrogen so that the failure of highstrength martensitic stainless steels in these cases is probably a HD mechanism.
When selecting an alloy or material for a specific application, it is common
practice to first ensure that EC will not be a potential problem in service. Once
this form of degradation is eliminated, the select material can be further evaluated
for resistance to other less catastrophic forms of attack.

Fig. 13 Comparison of stress-corrosion crack velocities in maraging and low-alloy


steels.

20

Craig

C. Galvanic Corrosion
Galvanic corrosion is one of the most common yet least recognized corrosion
problems for ferrous alloys. When two or more dissimilar metals are intimately
connected and placed in a solution, current will flow between them because of
the potential difference of the metals. The metal with the least resistance to corrosion (active metal) in the particular environment will become the anode, and the
more corrosion-resistant metal (noble metal) will become the cathode. Corrosion
of the anode will usually be more severe than if that metal was alone in the
same solution, whereas the cathode will achieve a degree of protection from the
environmentsometimes to the extent that corrosion is completely stopped on
the cathodic metal. These effects can be measured and have been done for couples
of metals in seawater at 25C. Table 3 provides a relative ranking of metals and
alloys regarding their resistance to corrosion in seawater (14). The greater the
distance between two metals in the table, the greater their potential difference and
the higher the probability that the active metal will suffer accelerated corrosion.
Note that some alloys and metals are listed twice in the table: once with
the word active following and once with the word passive. Some metals
and alloys become essentially immune to corrosion in certain environments beTable 3

Galvanic Series of Some Commercial Metals and Alloys in Seawater

Active or anodic
Magnesium
Magnesium alloys
Zinc
Galvanized steel
Aluminum 1100
Aluminum 2024
Mild Steel
Wrought Iron
Cast Iron
13% Chromium stainless steel
Type 410 (active)
18-8 Stainless steel
Type 304 (active)
Leadtin solders
Lead
Tin
Muntz metal
Manganese bronze
Naval brass
Nickel (active)
Nickel Alloy 600 (active)

Yellow brass
Admiralty brass
Red brass
Copper
Silicon bronze
7080 Cupro nickel
Gbronze
Silver solder
Nickel (passive)
Nickel alloy 600 (passive)
13% Chromium stainless steel
Type 410 (passive)
18-8 Stainless Steel
Type 304 (passive)
Silver

Graphite
Gold
Platinum
Noble or cathodic

Ferrous Alloys

21

cause of the formation of a surface film so thin that it is impossible to see with
the naked eye or even with an optical microscope. The stability of these films is
paramount to the enhanced corrosion resistance of these alloys. Moreover, corrosion films represent the controlling factor in almost all corrosion (15). The addition of chromium to iron can produce a passive alloy (18-8 stainless steel) of
considerable corrosion resistance compared to the original iron. However, in an
environment in which the passive film is not functional, the active surface becomes far less noble, as indicated in Table 3.
A more active metal in the series will corrode at the expense of a nobler
one. Thus, coupling zinc to steel will cause the zinc to corrode and will protect
the steel. This is the reason for galvanizing steel; when pinholes in the galvanizing
occur, the steel underneath, once exposed to the environment, will be protected
by the zinc. This is also the basis for cathodic protection. Sacrificial anodes are
made of metals or alloys that are more active than steel, which allows for the
consumption of the anode and the protection of the steel structure. However, if
steel is coupled to copper, the distance on the chart is large, so in this case, steel
will be the anode and have a greater tendency to corrode.
The galvanic series is useful for approximating the behavior of coupled
alloys; however, care must be used in its application. Several parameters affect
galvanic corrosion and, as such, may affect the actual behavior of a couple in
service. Two important factors in galvanic corrosion are the temperature and the
relative area of the metals. Increasing temperature in some cases may cause a
reversal in the anodecathode relationship. This reversal has been responsible
for failures of galvanized systems or systems protected with zinc sacrificial
anodes. These effects point to the need to measure the potential of a couple,
especially in cathodic protection, in the actual environment prior to its application. It must always be borne in mind that the ranking of alloys in Table 3 is
strictly true only for seawater and that extending it to other environments may
result in some changes in the position of metals and alloys in the series.
The other factor, the ratio of the area of the anode to the area of cathode,
is of considerable importance. If the anode area is smaller than the cathode area,
the corrosion rate may be increased many orders of magnitude as a function of
this ratio. However, if the anode area is greater than the cathode area, corrosion
of the anode will be less than for a 1: 1 anode/cathode ratio.

III. EFFECTS OF ALLOYING ELEMENTS


Small additions of alloying elements to ferrous alloys generally do not significantly improve their corrosion resistance. As stated earlier, at least 11% Cr is
needed to ensure that a steel becomes stainless and thus possesses a certain degree
of corrosion resistance. One important exception to this behavior is the class of

22

Craig

Fig. 14 Corrosion rate of steels in wet CO2 as a function of the chromium content of
the alloy.

steels referred to as weathering steels. Small additions (0.5%) of elements such


as Cr, Ni, Cu, and P greatly enhance these steels resistance to rusting in the
atmosphere. This is accomplished through the production of a tight tenacious
rust that forms on the steel after exposure to its environment. Hence, further
corrosion is stifled. However, for most other service environments, the small
variations in alloying or tramp elements are not sufficient to increase the corrosion
resistance of a steel. Figure 14 shows the benefit of increasing the Cr content on
steels exposed to water containing a high concentration of dissolved CO2 (16).
This behavior is typical for steels exposed to many acids as well as other environments. Thus, Cr is an important alloying element for a steels resistance to acid

Ferrous Alloys

23

attack. A similar benefit is observed for cast irons exposed to nitric and hydrochloric acid by alloying with either Si or Ni in excess of about 5%.

IV. EFFECTS OF HEAT TREATMENT


Carbon and low-alloy steels and the martensitic stainless steels can be and are
heat treated to enhance certain mechanical properties and, as a result, develop
different microstructures. Various combinations of ferrite, pearlite, bainite, and
martensite may be present in a particular steel, depending on its thermal history.
By and large, heat treatment and the subsequent phases formed are not a significant factor in the corrosion of steels. However, there are important exceptions.
As previously discussed, environmental cracking is highly dependent on steel
strength and, to some degree, its microstructure. Likewise, in certain corrosive
environments, a distinction can be observed in corrosion rate as a function of
heat treatment. One important example is shown in Fig. 15, for Type 420 stainless
steel in wet CO 2 (17). In actual practice, heat treatment/microstructure is not a
large concern, except for environmental cracking.

V.

EFFECTS OF SOLUTION VELOCITY

In general, the corrosion rate of ferrous alloys increases with increasing velocity.
This can be readily explained by the increased mass transport of ferrous ions
across the fluid boundary layer that is established under flowing conditions com-

Fig. 15 Effect of the tempering on the yield strength and on the corrosion resistance
of quenched-and-tempered 13% Cr stainless steel.

24

Craig

bined with the enhanced transport of corrodents to the metal surface across this
boundary layer. Of course, the picture is actually more complex when the Helmholtz double layer and the presence of a corrosion product film are included.
However, regardless of these issues, at some critical velocity these layers are
essentially stripped away and bare metal is continually exposed to the fluid
stream. At this point, one of two entirely divergent phenomenon may occur. The
erosioncorrosion rate becomes extremely high due to the loss of the ratedetermining mass transport across all of these layers or the erosioncorrosion
rate becomes much lower as a result of the inability of the corrodent to reach
the bare metal surface and have sufficient time to react. Both of these phenomena
are observed for ferrous alloys. Figure 16 illustrates the former case for carbon
steel in distilled water (18). At different pHs, the corrosion product formed can
be resistant to erosioncorrosion (pH 5 and 9), thus minimizing the erosion
corrosion rate, or the corrosion film is unstable (pH 4, pH 68), leading to
high erosioncorrosion rates. The introduction of solid particles such as sand
significantly lowers the erosioncorrosion threshold, making the selection of
alloys resistant to erosioncorrosion much more difficult. In the absence of solid
particles, it has been found that erosioncorrosion resistance is strongly a function of the nature of the oxide. Therefore, more corrosion-resistant alloys such
as stainless steels, nickel-based alloys, and titanium alloys have greater erosion
corrosion resistance than steels, even if the alloy is much softer, because of their
tighter more resilient oxides.

Fig. 16 Effect of pH of distilled water on erosion corrosion of steel at 50C and 12


m/s flow rate.

Ferrous Alloys

25

VI. GENERAL APPLICABILITY OF FERROUS ALLOYS


Every service environment is different and care should be taken in trying to generalize the performance of ferrous alloys, especially carbon and low alloys, in common environments. However, several important characteristics are worth emphasizing.
In almost every circumstance, the presence of dissolved oxygen in solution
causes an increased corrosion rate for steels and cast irons. However, for ferritic
and martensitic stainless steels, the presence of oxygen is not as critical and often
oxygen in only the ppb range is necessary to maintain the oxide film. Moreover,
under total anaerobic conditions at sufficiently high temperatures, oxygen is available from dissociation of the water molecule. The problem for steels and cast
iron is due to the fact that oxygen is a very effective cathodic depolarizer; that
is, it stimulates the cathodic reaction, and because the anodic and cathodic reactions are interdependent, it produces a net increase in the corrosion rate.
Likewise, steels are not resistant to corrosion in acidic pH environments
or even in many neutral pH environments, especially in the presence of dissolved
oxygen. It is for this reason that steels are most often painted (coated) or cathodically protected to ensure a satisfactory service life. Exposure to the atmosphere
is often sufficient to cause corrosion of ferrous alloys to the extent they become
either unserviceable or, more typically, aesthetically unpleasing. Moisture, temperature, periods of wet and dry, the presence of chlorides (coastal locations), and
industrial pollutants (oxides of sulfur and nitrogen) all contribute to atmospheric
corrosion of steels and cast iron.
Atmospheric corrosion requires a critical moisture content in the atmosphere and its rate generally increases when the humidity exceeds this critical
value (i.e., approximately 60%). Figure 17 illustrates this phenomenon for corrosion of many alloys, including steels, as a function of relative humidity (RH).
A form of atmospheric corrosion, wet corrosion, is considered to occur
when actual water layers or pools form on the surface of the metal, often from
dew, rain, or sea spray. This can be a very complex state because a thin layer
of water can act to dissolve a high concentration of gases from the atmosphere,
causing a concentrated solution at the metal surface, which produces a correspondingly high, short-term corrosion rate that produces a locally high metal ion
concentration, resulting in an oxide that stifles further corrosion or, if the corrosion product is soluble, continued localized attack.
Temperature can have many secondary effects on atmospheric corrosion
besides the primary effect of increasing reaction rate. Temperature influences the
relative humidity, dew point, and time of wetness; all important factors in themselves on atmospheric corrosion.
Contaminants, essentially airborne in nature, can profoundly affect atmospheric corrosion. Agents such as gases, like SO2, that can selectively absorb on

26

Fig. 17

Craig

Corrosion of ferrous alloys as a function of relative humidity.

metal surfaces which act as a catalyst to form SO3 and thus H 2 SO4 in moist
environments or particulates such as dust that can cause local cells to form by
aiding in the absorption of water and chlorides can accelerate corrosion of alloys
that normally would be resistant to atmospheric corrosion in a relatively clean
environment.
Climatic conditions have a variable effect on corrosion rate. For example,
in some regions, winter exposure may be more severe than summer if fuels are
used during cold spells that increase combustion products in the air such as NO x
and SO2. Conversely, if these fuels are not used in the region, then summer may
be worse due to the higher metal temperatures.
Likewise, periodic rainfall may be beneficial causing a rinsing action on
the surface compared to a climate where the surface is continually wet. Thus,
time of wetness and wet/dry cycles can also be quite important, especially if
frequent periods of wet and dry can limit the formation and development of a
protective oxide layer. Moreover, the existence of insoluble corrosion products
can act to entrain water during short dry cycles, keeping the metal surface sufficiently wet to continue corrosion.
Two other external environments that can cause significant deterioration
of ferrous alloys are soils and concrete. The majority of buried structures in the

Ferrous Alloys

27

world are made of cast irons and steels. Soils have a great variability in their
tendency to cause corrosion of ferrous alloys. Some of the more important factors
that contribute to the aggressiveness of soils are resistivity, pH, moisture, oxygen,
bacterial activity, and temperature of the ferrous alloy (i.e., hot pipelines) in contact with the soil. Decreasing resistivity and pH, increasing moisture content,
oxygen content, and bacterial activity all enhance corrosion of ferrous alloys in
soils. The elevated temperature of the steel surface not only can increase the
corrosion rate but also can lead to other forms of more serious attack such as
EC. As shown in Fig. 9, the external SCC of pipelines has become a particular
problem for those pipelines that operate above ambient temperature and the trend
in the future is to operate pipelines at even higher temperatures. Often, coatings
and cathodic protection can be used to limit these problems; however, they can
also exacerbate them if not properly maintained. One of the major reasons older
pipelines have become susceptible to SCC is that the coatings have degraded
over time and the cathodic protection systems cannot effectively limit corrosion.
Corrosion of steel reinforcing bar in concrete has gained attention due to
the widespread problem of the crumbling infrastructure (bridges, highways,
buildings, etc.) in many countries. Typically, the steel rebar corrodes as a result
of the pH of the cement paste in contact with the steel and the diffusion of chlorides and oxygen into and through the concrete. As corrosion products form on
the steel, they represent a larger volume than the original iron in place, thereby
spalling and cracking the concrete. It has been found that temperature and relative
humidity are important factors in rebar corrosion as well as chlorides and oxygen.
Thus, tropical climates that are hotter and more humid than temperate climates
would be expected to have a greater problem with rebar corrosion than cooler
climates. However, even northern climates have had problems for other reasons;
for example, deicing salts used on bridges and roadways to eliminate snow and
ice can cause severe rebar corrosion. Several of the methods currently used to
fight rebar corrosion are organic coated steel, galvanized steel, stainless-steel
rebar, and cathodic protection.
As can be appreciated from the foregoing comments, coatings on ferrous
alloys are an important means of extending the applicability and service life of
these materials. As this is a book on engineered materials, it is beneficial to at
least mention the general types of coatings used on ferrous alloys. Coatings can
be grouped into four general categories: organic, inorganic, conversion, and metallic. Table 4 lists some of the typical coatings under each of these categories.
Under the category for metallic coatings, the specific metal is not listed because
many metals can be applied; rather, the process is provided because it will determine which metal can be applied. The selection and use of a particular coating
is a function of many factors and care must be taken in selecting the right coating
for a ferrous alloy.
It must always be borne in mind that a coating is part of a system that

28
Table 4

Craig
General Categories of Coatings

Organic
Coal tars
Epoxy
Phenolics
Alkyds
Vinyls
Urethanes
Acrylics

Metallic
Galvanizing
Plating
Ion implantation
Cladding
Flame spray
Chemical vapor deposition
Physical vapor deposition

Conversion
Anodizing
Phosphating
Chromate
Molybdate

Inorganic
Silicates
Ceramics
Glass

Fig. 18

Coating degradation and corrosion of HY80 in artificial seawater.

Ferrous Alloys

29

includes the steel substrate; therefore, if the coating is damaged, corrosion often
occurs by different mechanisms than if only the steel were involved. For example,
galvanizing (Zn) acts as a sacrificial anode to the underlying steel if the coating
is damaged. Thus, the steel substrate is protected against corrosion. However, a
more noble coating such as Ni on steel can act as a large cathode, thereby accelerating corrosion at a damaged location (referred to as a holiday). Even organic
coatings can display accelerated corrosion at holidays in the coating depending on
the particular environment to which they are exposed (Fig. 18) (19,20). Generally,
coatings show a slow decrease over time in the coating (film) resistance, indicating the gradual permeation of water and other ionic species through the coating.
Thus, coating life in any environment is finite and organic coatings are not truly
barriers, as is so often mistakenly suggested. However, the application of the
correct coating can often double or triple the useful life of a ferrous alloy in
certain environments and is, therefore, an important factor in the selection of
ferrous alloys.
In conclusion, even though ferrous alloys (ferritic and martensitic) are the
most widely used engineered materials, they are also the least corrosion resistantreadily degrading in most environments. Because of this behavior, they
most often require additional means of controlling corrosion (i.e., coatings, cathodic protection, inhibitors) to provide a satisfactory service life.

REFERENCES
1. RL Martin. Application of Electrochemical Polarization to Corrosion Problems. St.
Louis, MO: Petrolite Corp., 1977.
2. ASM Metals Handbook, Vol. 13, Corrosion. ASM International, Materials Park,
OH, 1989.
3. ND Stolica. Pitting corrosion on FeCr and FeCrNi Alloys. Corrosion Sci 9:
455460, 1969.
4. D Wheeler. Treating and monitoring 450,000 b/d injection water. Petrol Eng Int
1975; November: Vol. 38, 6872.
5. BD Craig. Selection guidelines for corrosion resistant alloys in the oil and gas industry.
Technical Publication No. 10073, The Nickel Development Institute, Toronto, 1995.
6. BD Craig, RH Jones. Environmentally induced cracking. In: ASM Metals Handbook, Vol. 13, Corrosion, ASM International, Materials Park, OH, 1989, pp. 145
171.
7. FP Ford. Quantitative prediction of environmentally assisted cracking. Corrosion
51:375395, 1996.
8. EH Phelps. Stress corrosion behavior of high yield strength steels. Proc. Seventh
World Petroleum Congress. Amsterdam: Elsevier, 1967.
9. RN Parkins, R User. The effect of nitrate solutions in producing stress corrosion

30

10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.

Craig
cracking in mild steel. First International Congress on Metallic Corrosion, London,
1961, p. 289.
Corrosion Data SurveyMetal Selection, 6th ed. Houston, TX: NACE International, 1985, p. 176.
RN Parkins. Predictive approaches to stress corrosion cracking failure. Corrosion
Sci 20:147166, 1980.
E Sato, M Hashimoto, T Murata. Corrosion of Steels in a Wet H2S and CO2 Environment. Second Asian Pacific Corrosion Control Conf., Kuala Lumpur, 1981.
CS Carter. Fracture toughness and stress corrosion characteristics of a high strength
maraging steel. Met Trans 2:16211625, 1971.
Corrosion Basics. LS Van Delinder, ed. Houston, TX: NACE International, 1984,
p. 35.
BD Craig. Fundamental Aspects of Corrosion Films in Corrosion Science. New
York: Plenum Press, 1991.
NG Galindez Ruiz. The effect of crude oil on corrosion of alloys in H2S/CO2 environments. Masters thesis, Colorado School of Mines, 1993.
JL Crolet. Acid corrosion in wells (CO2, H2S): Metallurgical aspects. J Petrol Technol 35:15521558, August 1983.
MG Fontana. Corrosion Engineering, 3rd ed., New York: McGraw-Hill, 1986, p. 94.
BD Craig, DL Olson. Corrosion at a holiday in an organic coated-metal substrate
system. Corrosion 32:316321, 1976.
JR Scully. Evaluation or organic coating deterioration and substrate corrosion in
seawater using electrochemical impedence measurements. Corrosion/86, NACE,
Houston, TX, 1986.

2
Austenitic Stainless Steels
Russell H. Jones, Stephen M. Bruemmer, and Mike J. Danielson
Pacific Northwest National Laboratory, Richland, Washington

Bruce Craig
MetCorr, Denver, Colorado

I.

INTRODUCTION

Austenitic stainless steels derive their stainless properties from the presence
of a very effective passive film. This film forms spontaneously within aqueous
environments and the stability of this film in various environments determines
the corrosion resistance of stainless steels. Alloy composition is also very
significant in determining the stainless character of this family of alloys. Chromium concentration is the one element that directly affects the passive film stability, and Ni, Mn, Mo, C, and N also play a role. Passive behavior begins at about
10% Cr, with increasing film stability occurring with increasing Cr concentration.
Repassivation can occur in aqueous environments and it is the rate at which a
break in the passive film re-forms that also contributes to the stainless character of these materials. Under selected conditions, the passive film is not stable
and this can lead to general or localized corrosion phenomena. Austenitic stainless steels are often prepassivated in an acid bath, but the removal of surface
contaminants that would hinder passivation in aqueous solutions is the primary
purpose of this prepassivation treatment. Given a clean surface in an alloy with
31

32

Jones et al.

sufficient Cr, austenitic stainless steels will passivate upon immersion in an aqueous environment. Also, there will always be an air-formed oxide even without
this prepassivating treatment.
Austenitic stainless steels undergo all the common forms of corrosion, including (1) general, (2) galvanic, (3) pitting, (4) crevice, (5) intergranular, and
(6) stress corrosion. General corrosion occurs when stainless steel is immersed
in an environment in which the passive film is not stable, such as hot sulfuric
acid, boiling MgCl2, or another very aggressive environment. Pitting corrosion
occurs, as in other metals, because of a local discontinuity in the passive film
such as an inclusion or local chemistry change. Halogen ions are the most prevalent cause of pitting in stainless steels, with chloride being the most common
halogen to initiate pitting. Pit growth depends on a variety of factors, with the
localized corrosion environment in the pit being the most significant factor. Surface condition can also contribute to pitting, with the presence of deposits being
a significant factor. Prepassivating treatments are used to clean the surface of
deposits. Intergranular corrosion and intergranular stress-corrosion cracking are
related phenomena which occur when the grain-boundary microchemistry is altered by thermal treatment such as welding or heat treatment. The process that
alters the grain-boundary corrosion resistance is called sensitization and occurs
when chromium carbides precipitate at the grain boundaries.

II. CORROSION BEHAVIOR


A. Alloy Classification
Stainless steels are classified by the phases that are present. Table 1 shows the
classifications and compositions of the more commonly available alloys, but this
table is not exhaustive. The five classifications are austenitic, ferritic, duplex (containing both austenite and ferrite), martensitic, and precipitation hardening (PH).
Alloying affects the predominant phase that is present, and the phase has a profound effect on the mechanical and corrosion characteristics. Often, the mechanical properties are the driving force for choosing the alloy for the application, and
the corrosion properties must then be optimized within that alloy classification.
A brief description of each classification is given in the following subsections.

1. Austenitic Stainless Steels


Austenitic stainless steels have a face-centered-cubic (fcc) crystal structure. The
most commonly used stainless steels, the 300 series, belong to this group. The
austenite is a high-temperature phase that is stabilized by the addition of nickel,
manganese, or nitrogen, but they also contain significant amounts of chromium,
which gives them good overall corrosion properties. These alloys are fairly low

Austenitic Stainless Steels

33

in strength, but they exhibit a high fracture toughness and are used over a wide
range of temperatures. As a class, they have good resistance to hydrogen embrittlement. In recent years, a new class has emerged called superaustenitics. They
are very high in Mo (47%) and nickel, which results in the highest resistance
(within the austenitics) to any localized attack processes in chloride-containing
media.

2. Ferritic Stainless Steels


Ferritic stainless steels have the body-centered-cubic (bcc) crystal structure, using
chromium as the major alloying element. They can have higher strengths than
the austenitic steels at ambient temperatures, but they suffer from a lower fracture
toughness, particularly at lower temperatures. In an attempt to improve the fracture toughness and corrosion behavior, a new class of superferritics was developed. These materials are low in carbon and higher in Cr and Mo than the older
ferritics. Successful use of ferritics requires careful attention to detail in controlling the heat treatment.

3. Duplex Stainless Steels


Duplex stainless steels have lower amounts of nickel than the austenitic grades,
with the result that some of the austenite transforms to ferrite. Generally, the
alloying element and heat treatments are controlled to form equal amounts of
ferrite and austenite. The principal advantages over the fully austenitic grades
are higher strength, improved resistance to stress-corrosion cracking (SCC), and
a high immunity to sensitization. In order to improve the localized corrosion
behavior, a superduplex series of alloys has emerged. These are alloyed to contain
larger amounts of Cr, Mo, and N.

4. Martensitic Stainless Steels


Martensitic stainless steels contain significant amounts of chromium (1018%)
and carbon but are low in Ni. Although austenitic at high temperatures, they can
be transformed into the martensite structure by rapid cooling. These alloys are
very strong, but they suffer a loss of fracture toughness and have generally inferior localized corrosion resistance. Hydrogen embrittlement has been identified
as a cause of fracture. Recently, supermartensitics have been formulated that are
higher in Mo to improve localized attack behavior.

5. Precipitation-Hardening Stainless Steel


Precipitation-hardened stainless steels superficially resemble the 300 series austenitic steels in nickel and chromium composition, but, in addition, they contain
small amounts of copper, aluminum, or titanium that can be precipitated with

Table 1

Stainless Steel Compositions (wt%)

Austenitics
201
202
301
302
304
304L
304LN
316
316LN
316L
321
347
348
384
Superaustenitics
254 SMO
AL-6X
AL-6XN
20Cb-3

442
446

Mn

Si

Ni

Cr

S20100
S20200
S30100
S30200
S30400
S30403
S30453
S31600
S31653
S31603
S32100
S34700
S34800

0.15
0.15
0.15
0.15
0.08
0.03
0.03
0.08
0.03
0.03
0.08
0.08
0.08

5.57.5
7.510.0
2
2
2
2
2
2
2
2
2
2
2

1
1
1
1
1
1
1
1
1
1
1
1
1

16.018.0
17.019.0
16.018.0
17.019.0
18.020.0
18.020.0
18.020.0
16.018.0
16.018.0
16.018.0
17.019.0
17.019.0
17.019.0

3.55.5
4.06.0
6.08.0
8.010.0
8.010.5
8.012.0
8.012.0
10.014.0
10.014.0
10.014.0
9.012.0
9.013.0
9.013.0

0.06
0.06
0.045
0.045
0.045
0.045
0.045
0.045
0.045
0.045
0.045
0.045
0.045

0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03

S38400

0.08

15.017.0

17.019.0

0.045

0.03

S31254

0.02

0.8

19.5020.50 17.5018.50 0.03

N08366 0.035
N08367 0.03
N08020 0.07

2
2
2

1
1
1

20.022.0
20.022.0
19.021.0

23.525.5
0.03
23.5025.50 0.04
32.038.0
0.045

0.03
0.03
0.035

N08904 0.02

19.023.0

23.028.0

0.035

S40500
S40900
S42900
S43000
S43600
S43035

0.08
0.08
0.12
0.12
0.12
0.07

1
1
1
1
1
1

1
1
1
1
1
1

11.514.5
10.511.75
14.016.0
16.018.0
16.018.0
17.019.0

0.5

0.5

0.04
0.045
0.04
0.04
0.04
0.04

S44200
S44600

0.2
0.2

1
1.5

1
1

18.023.0
23.027.0

0.04
0.04

0.045

0.01

Other
0.25 N
0.25 N

0.100.16 N
2.03.0 Mo
2.03.0 Mo;
2.03.0 Mo
5 % C min Ti
10 % C min Nb
0.2 Co; 10 % C min Nb;
0.10 Ta

6.006.50 Mo; 0.501.00


Cu; 0.1800.220 N
6.07.0 Mo
6.007.00 Mo; 0.180.25 N
2.03.0 Mo; 3.04.0 Cu; 8
% C min to 1.00 max
Nb
4.05.0 Mo; 1.02.0 Cu

0.03
0.100.30 Al
0.045 6 % C min0.75 max Ti
0.03

0.03

0.03
0.03
0.15 Al; 12 % C min
1.10 Ti
0.03

0.03
0.25 N

Jones et al.

904L
Ferritics
405
409
429
430
436
439

UNS

34

Type

S44660

0.025

25.027.0

1.53.5

0.04

0.03

AL 29-4C

S44735

0.03

28.030.0

0.04

0.03

AL 29-4-2

S44800

0.01

0.3

0.2

28.030.0

2.02.5

0.025

0.02

Duplex
329
Uranus 50

S32900
S32404

0.2
0.04

1
2

0.75
1

23.028.0
20.522.5

2.505.00
5.58.5

0.04
0.03

0.03
0.01

S32550

0.04

1.5

24.027.0

4.506.50

0.04

0.03

S40300
S41000
S41400
S41600

0.15
0.15
0.15
0.15

1
1
1
1.25

0.5
1
1
1

11.513.0
11.513.5
11.513.5
12.014.0

1.252.50

0.04
0.04
0.04
0.06

S42000
S42200

0.15 min
0.200.25

1
1

1
0.75

12.014.0
11.513.5

0.51.0

0.04
0.04

440A
S44002
440B
S44003
440C
S44004
Lapelloy
S42300
Precipitation Hardening
138 Mo
S13800

0.600.75
0.750.95
0.951.20
0.270.32

1
1
1
0.951.35

1
1
1
0.5

16.018.0
16.018.0
16.018.0
11.012.0

0.5

0.05

0.2

0.1

12.2513.25 7.58.5

0.07
0.07
0.09
0.070.11

1
1
1
0.51.25

1
1
1
0.5

14.015.5
15.517.5
16.018.0
16.017.0

Ferralium
255
Martensitic
403
410
414
416
420
422

S15500
S17400
S17700
S35000

3.55.5
3.05.0
6.57.75
4.05.0

1.002.00 Mo
2.03.0 Mo; 1.02.0 Cu;
0.20 N
2.004.00 Mo; 1.502.50
Cu; 0.100.25 N

0.03

0.03

0.03

0.15
0.6 Mo(b)
min
0.03

0.03
0.751.25 Mo; 0.751.25 W;
0.150.3 V
0.03
0.75 Mo
0.03
0.75 Mo
0.03
0.75 Mo
0.025 2.53.0 Mo; 0.20.3 V

0.01

0.008

0.04
0.04
0.04
0.04

0.03
0.03
0.04
0.03

2.02.5 Mo; 0.901.35 Al;


0.01 N
2.54.5 Cu; 0.150.45 Nb
3.05.0 Cu; 0.150.45 Nb
0.751.5 Al
2.53.25 Mo; 0.070.13 N

35

15-5 PH
17-4 PH
17-7 PH
AM-350
(Type 633)

0.04
0.04
0.04
0.025

2.53.5 Mo; 0.2 4 (% C


% N) min to 0.8 max
(Ti Nb); 0.035 N
3.604.20 Mo; 0.201.00 Ti
Nb and 6 (% C %
N) min Ti Nb; 0.045 N
3.54.2 Mo; 0.15 Cu; 0.02
N; 0.025 max (% C
% N)

Austenitic Stainless Steels

Superferritics
Sea-Cure
(SC-1)

36

Jones et al.

the appropriate heat treatment. These alloys can develop extremely high strength
and can also share some of the good corrosion properties of the austenitics, but
hydrogen embrittlement has been identified as a cause of fracture. This grade of
alloy can be martensitic or austenitic.
These five classes of alloys can also be obtained as cast alloys. Microstructure is a key variable in controlling the strength, localized corrosion behavior,
and SCC behavior of all these classes. It needs to be pointed out that the existence
of cast microstructures increase the complexity of choosing the ideal material
and making certain the microstructure is under control.
B. Composition Effects on Corrosion
Stainless steels are vulnerable to all forms of corrosive attack. Their successful
use requires (1) knowledge of the alloying constituents that impact resistance,
(2) detailed information on the chemical composition and temperature of the test
environment, and (3) knowledge of the failure processes to which the particular
alloy or alloy class is subject. Each of the major alloying elements will be briefly
described to give the reader a general knowledge of how to fit the alloy composition to the environment.

1. Chromium
Chromium is the single most important element contributing to the stainless
behavior of these iron-based alloys, and in general, the higher the chromium level
(once it gets above 10%), the better the performance. Additions of chromium
greatly improve the behavior over that of iron in neutral and acidic pH ranges
but, curiously, has little effect in high-pH environments. The chromium appears
in the oxide film and acts to inhibit the transport of corrosion products across it,
leading to the formation of a passive film. In particular, the chromium acts to
decrease the general corrosion rate and improve crevice corrosion resistance.

2. Molybdenum
Molybdenum is perhaps the second most important alloying element from a corrosion standpoint. Small additions have a profound effect by enhancing the passive character of the passive film, particularly in chlorides and reduced sulfur
environments. In particular, the pitting and crevice corrosion behavior are improved relative to similar alloys without the molybdenum. Every class of stainless-steel alloy has members that are very high in Mo, and these are called super austenitics, martensitics, and so forth. The alloys with high Mo levels have
the disadvantage in that certain undesirable phases (sigma, chi, laves) can form
unless care is taken in heat treatment and welding.

Austenitic Stainless Steels

37

3. Nickel
Nickel is the most important austenite stabilizer, but it has a complex effect on
the corrosion behavior. It acts to decrease the general corrosion rate in reducing
acidic environments. In sufficiently high levels, it acts to improve the SCC behavior in chloride and caustic solutions, but at intermediate levels, it can decrease the
SCC resistance. The austenite structure results in a high resistance to hydrogen
embrittlement.

4. Manganese
At the low levels used in most stainless steels, manganese can be considered a
substitute for nickel. It controls the solubility of sulfur by precipitating the sulfur
as MnS inclusions. Modern steel-making practice is to keep the sulfur as low
as possible because localized corrosion events such as pitting initiate at MnS
inclusions.

5. Carbon
Carbon is used at low levels as a strengthener in stainless steels, but it also can
render the stainless steels vulnerable to localized attack or SCC from sensitization. In sensitization, the carbon reacts with chromium to precipitate a Cr 23C 6
carbide causing a very local decrease of the chromium concentration in the metal,
making this depleted region vulnerable to localized attack or SCC. In the austenitics, this has led to the L grades, which are low in carbon (less than approximately 0.03 at.%) and fairly immune to this problem.

6. Nitrogen
Nitrogen is used at very low levels and acts as a strengthener and austenite stabilizer. Nitrogen greatly improves the pitting and crevice corrosion performance
of the austenitics and superaustenitics, particularly in concert with Mo. It helps
prevent (slows down) the undesirable chi (and probably sigma, laves) phase from
forming during welding or heat-treating operations with the high-Mo alloys. Its
behavior is mixed in other classes of stainless steels.
C. Corrosion Behavior
Choosing the right material is a multistage process. Once the mechanical properties and certain other properties (e.g., wear resistance, weldability, machinability,
availability in the needed form, etc.) of the metal are defined for the application,
the next level requires a careful definition of the chemical (e.g., pH, chloride
level, sulfide level, temperature, velocity, oxygen, hydrogen gas) and exposure
environment (single or multiple phases present, occasional dryout, differential

38

Jones et al.

aeration, etc.). Once this multistage process is completed, the general class of
alloy can be defined. Because cost is always an important parameter for the
material choice, the tendency will be to use the lowest alloyed material because
nickel and chromium are expensive. As a general rule for iron-based stainless
steels, the higher the alloy is in nickel and chromium (and molybdenum), the
more corrosion resistant the material will be. The task of the materials scientist is to find the least expensive material that will perform adequately. Often,
the exact answer will not be found in the literature; consequently, testing of various representative alloys will be needed. Many test methods are described in
Ref. 1.
From their inception, stainless steels were developed to have low general
corrosion rates. This simplification is still fairly true as long as one avoids certain
environments such as reduced sulfur environments or highly acid environmentsonly highly alloyed stainless steels will perform in these environments.
The surprises with stainless steels are usually due to localized attack such as
pitting, crevice corrosion, and SCC. Because the chemical environment has infinite variability, the corrosion behavior for each class of stainless steel will be
examined using certain defined environments which will act as benchmarks or
models. Examples will be potable water (low chloride), seawater (high chloride),
and so forth. The book by Sedriks (2) is a particularly good reference.

1. Atmospheric Environments
Atmospheric environments are those exposed to the natural elements in the air
at ambient temperatures. City and marine environments present the most difficult
environments because they are contaminated with sulfur compounds and chlorides. The austenitics, particularly those containing molybdenum, give the best
outdoor performance from the standpoint of pitting and crevice corrosion, but it
should be noted that they will not be pit-free. Rather, the pit density and depth
will usually remain low enough that the cosmetic use of the material will not be
impacted. Localized corrosion rates for all classes of stainless steels are highest
where the chlorides and sulfides are highest. The ferritics have also been used
in this environment but not as successfully as the austenitics. Both the austenitics and ferritics have good immunity to SCC. The higher-strength alloys such
as the martensitics and precipitation-hardened alloys are prone to pitting and
SCC. This environment can get extremely aggressive if there is a source of heat
such that condensation and dryout takes place, resulting in very concentrated
solutions. One extreme of this environment might be found under insulation, and,
here, SCC and pitting can occur with all the standard stainless classes, even the
austenitics and ferritics. The high-Mo (super types) alloys provide the greatest
resistance to pitting and SCC when the atmospheric environment is the most
aggressive.

Austenitic Stainless Steels

39

2. Deionized (Pure) Water (Ambient Temperature)


Pure water with low concentrations of dissolved salts (although it may contain
dissolved carbon dioxide from the air) is the most ideal environment for stainless
steels. They should all perform well and be free of any localized corrosion problem and SCC.

3. Deionized (Pure) Water (High Temperature)


The nuclear power industry successfully uses austenitic stainless steels under
elevated temperature conditions. Initially, 304 stainless (high C) was used, which
when welded gave rise to sensitized grain boundaries which were very prone to
SCC. This was a very major problem which has been solved by using low-carbon
austenitics. Some ferritics and precipitation-hardened alloys have been successfully used in other components such as turbine blades. The key to successful use
is attention to heat treatment and maintaining the purity of the chemical environment. The other classes of stainless steel have not been used extensively in this
environment. There is an extensive literature on this subject and the reader is
directed to reviews by Hanninen (3) and Cragnolino (4).

4. Fresh (Potable) Water


Fresh water can contain a few hundred ppm chloride, and when used in a heat
exchanger, it can reach temperatures near 100C. The 300 series austenitics have
been used successfully in this environment, but as the temperature and chloride
level increase, the amount of alloying (particularly Mo) must increase to prevent
pitting and SCC. Sensitization is an issue and the low carbon grades must be
used. For years, there was a widely accepted belief that SCC did not occur below
60C, but long-term experience has now demonstrated that there is no threshold
temperature or chloride level for the 300 series austenitics (5). Consequently,
there will be a large variability in the observations of success with the 300 series.
The superaustenitics with high levels of molybdenum should perform very well
against pitting and SCC degradation. The standard ferritics are low in molybdenum and will be prone to pitting, but they can be resistant to SCC. The behavior
of the duplex stainless steels is complex and must be considered carefully. They
are more resistant to SCC relative to the 300 series of austenitics but may be
similarly prone to pitting. Both the martensitic and precipitation-hardening grades
are prone to pitting and SCC. The SCC process for these higher-strength steels
is considered to actually be hydrogen embrittlement.

5. Seawater
Seawater is the most challenging environment for stainless steels, and the problems are compounded if reduced sulfur species are present due to the action of

40

Jones et al.

bacteria or decaying vegetation. In general, the standard alloys within each grade
will be prone to pitting and SCC with much variability in the reported results. For
maximum safety and reliability, only the high-Mo alloys should be considered for
this environment. A review paper by Streicher (6) on 30- and 60-day ambienttemperature crevice tests clearly shows only the superaustenitics, superferritics,
high-Mo duplex, and nickel-based alloys are suitable. All localized corrosion
and SCC problems worsen at elevated temperatures. Here, the famous boiling
magnesium chloride test can give insight into a materials behavior. The 300
series alloys have completely unsuitable SCC behavior in this test (unless they
are cathodically protected). The nickel level has to be above 20% before the
austenitics show significant improvement in SCC behavior in the boiling magnesium chloride test. Molybdenum is also beneficial in raising the threshold stress
intensity for SCC. Austenitics (high Ni), ferritics, and duplex alloys all can show
resistance to SCC, but it is clear that they must be high in Mo. For the most
difficult applications, titanium or nickel-based alloys will need to be used.

6. Acidic Environments
Chromium is the most important element that imparts resistance to acidic environments; consequently, the highest resistance is associated with the highest Cr
levels within each grade of stainless steel. In general, the 300 series austenitic
stainless steels can be used in nitric acid over a wide concentration range (0
65%), even up to the boiling point. However, stainless steels are completely unsuitable for the HCl environment, and mixed acids containing HCl and other
halides are also very problematic. The alloy 20Cb-3, containing copper, was especially developed for use in sulfuric acid. Clearly, it is important to define the
acid compositions and temperatures and then to utilize the literature for alloy
recommendations. General corrosion and localized attack due to microstructural
problems are the major causes of failure. Much less is known about the SCC
behavior. Organic acids are less corrosive than the mineral acids.

7. Basic Environments
In general, all stainless steels are quite resistant to general corrosion in concentrated caustic solutions, even up to boiling temperatures. However, there is a
severe SCC problem, and the alloys can act in a very brittle manner. Alloys with
the highest nickel content are the most resistant to general attack and SCC. The
threshold for SCC is a function of the temperature, caustic concentration, and
nickel content. The austenitics 304 and 316 are resistant to SCC up to about 60C
in 60% concentrated caustic. At higher temperatures, nickel-based alloys must
be used. Microstructural effects are very important because sensitization makes
the materials more susceptible to SCC. Austenitic structures are more resistant
than ferritic structures.

Austenitic Stainless Steels

41

III. INTERGRANULAR STRESS-CORROSION CRACKING


Grain-boundary composition has been inferred to control intergranular (IG) fracture in a wide range of materials systems. Although many authors have attempted
to link grain-boundary composition and environmental cracking susceptibility,
few have made direct measurements. In most cases, bulk composition and/or heat
treatment is varied and it is assumed that interfacial segregation is systematically
changed. Indirect measurements are often made (e.g., IG corrosion tests) indicating the grain-boundary composition of an isolated element. Within selected wellunderstood cases, such approaches can give reproducible results. However, quantitative measurements of grain-boundary composition are essential to enable any
reasonable assessment of variables controlling cracking susceptibility. With the
commonplace use of high-resolution techniques such as analytical transmission
electron microscopy (ATEM) and scanning Auger microscopy (SAM), quantitative relationships have been established between interfacial composition and
cracking susceptibility for many metallic alloy systems (7). Perhaps the alloy
system that has been most closely examined has been austenitic stainless steel
due to its widespread use as a corrosion-resistant structural alloy in nuclear power
systems. The vast majority of failures have been in high-carbon, 300-series stainless steels thermally sensitized during fabrication. Extensive basic and applied
research activities were initiated about 25 years ago to develop a mechanistic
understanding of the IGSCC process and, more importantly, to identify remedial
actions and corrective measures to cracking problems in boiling-water reactor
(BWR) power plants. For the most part, those research activities were highly
successful. IGSCC of sensitized stainless steel is probably the best understood
and effectively modeled environmental cracking process (8). However, recent
observations of IG cracking in cold-worked or in irradiated stainless steels, have
been difficult to explain.
Austenitic stainless steels provide an example alloy system to demonstrate
the influence of grain-boundary composition on IGSCC. Emphasis is placed on
identifying equilibrium and nonequilibrium segregants that may promote susceptibility or improve resistance to cracking, respectively. In each case, current understanding of grain-boundary composition development in stainless steels is reviewed and assessed relative to IG fracture in corrosive environments.
A. Grain-Boundary Composition and IGSCC
The general conditions necessary to promote IGSCC are a susceptible material
microstructuremicrochemistry, a sufficiently corrosive environment, and the
presence of tensile stresses. Many of the important aspects controlling environmental crack advance are illustrated in Fig. 1. In nearly all cases of IG cracking,
grain-boundary composition plays a dominant role. Interfacial composition can

42

Fig. 1

Jones et al.

Schematic illustrating intergranular stress corrosion processes.

be significantly changed from the matrix by equilibrium (nonequilibrium) processes resulting in segregation (depletion) of alloying (impurity) elements and
precipitation of second phases. These compositional changes in the grain-boundary region can influence IG crack advance through effects on electrochemical
behavior (e.g., dissolution, repassivation, and hydrogen recombination) as well
as effects on interfacial mechanical behavior (e.g., deformation and cohesive
strength).
B. Precipitation and Grain-Boundary
Composition Changes
The dominant material variable controlling IGSCC susceptibility in austenitic
stainless steels results from the precipitation of Cr-rich M23C6 carbides at highenergy interfaces. This promotes the development of a Cr-depleted region adjacent to carbide precipitates. This depletion is controlled by the thermodynamics
of carbide formation and differences between the diffusivities of Cr and C.
ATEMEDS has enabled Cr-depletion profiles to be routinely measured demonstrating that interfacial Cr concentrations decrease (from 18% to 10%) as the
heat-treatment temperature is decreased due to changes in C and Cr activities. The
width of the depleted zone increases with time after IG carbides are nucleated.
The extent of grain-boundary Cr depletion has been directly linked to the
IG corrosion and SCC susceptibility of austenitic stainless steels (912). The
threshold concentration to promote IG degradation can be quite different for cor-

Austenitic Stainless Steels

43

Fig. 2 Grain-boundary Cr concentration width on IGSCC in BWR water environment.

rosion and SCC, as illustrated in Fig. 2. Classical IG corrosion is detected in a


standard sensitization test when the grain-boundary Cr concentration drops below
13.5 wt%. On the other hand, IGSCC in high-temperature aerated-water environments can be initiated during slow-strain-rate (SSR) tests when local Cr levels
drop below 17 wt%. Additional tests varying the width of the Cr-depletion
zone (and keeping boundary Cr concentration approximately constant) reveal that
only a very narrow (4 nm) width is necessary to promote cracking (Fig. 3).

Fig. 3 Grain-boundary Cr depletion on IGSCC in BWR water environment.

44

Jones et al.

IGSCC susceptibility is not sensitive to increases in depletion width beyond that


necessary to establish a continuous path for crack advance. On the other hand,
standard sensitization tests will show much more aggressive IG corrosion with
increasing depletion widths (11). The correlations presented in Fig. 2 point out
the critical importance of Cr depletion, and Cr minimums in particular, on IG
degradation of austenitic stainless steels. Specific relationships between IGSCC
and grain-boundary composition will always depend on many other critical factors, including mechanical loading characteristics and environmental conditions,
as well as secondary material variables. For example, the threshold grain-boundary Cr concentration has been shown to depend on the strain rate during SSR
tests (10,11)
C. Equilibrium Impurity Segregation
Impurity elements present at low levels in austenitic stainless steels can reach
high levels at grain boundaries due to equilibrium segregation. The most prevalent segregant is P, which can reach grain-boundary P contents 10 at% in commercial stainless steels after intermediate temperature heat treatments (500
750C). Thus, materials in the sensitized condition will most likely have considerable P segregation along with M23C6 carbides and Cr depletion defining the
local microchemistry. Segregation of other impurity elements to stainless-steel
grain boundaries has been observed, but this often requires high bulk contents or
special thermal treatments. Sulfur segregates rapidly to boundaries if preexisting
sulfides are dissolved by a high-temperature (1200C) exposure. Without such
treatment, grain-boundary S segregation is very slight even in doped alloys. Another element that has been shown to strongly segregate to austenitic stainlesssteel grain boundaries is N. Grain-boundary segregation of N in commercial 304
and 316 grades is likely because bulk N levels are typically greater than 0.02
wt% (higher in L grades).
Grain-boundary impurity segregation has been shown to promote hydrogen-induced cracking (HIC) in many iron- and nickel-based alloys (13). Phosphorus segregation induces HIC during low-temperature SSR tests at cathodic potentials, as illustrated in Fig. 4, but appears to have no effect on IGSCC in
high-temperature water, as indicated by the triangular points in Fig. 2. The influence of grain-boundary Cr content on IG cracking is not affected by P segregation
(10 at%). These results are consistent with the crack-growth-rate tests of Andresen and Briant (14), who found that grain-boundary P (and N) enrichment did
not promote IG cracking of 304L SS in high-temperature water, whereas S had
a small detrimental effect. However, there remains a need for additional crack
growth experiments to evaluate segregation effects on SCC in stainless steels
strengthened by cold work where cracking has been identified in laboratory tests
and in service without Cr depletion (15,16).

Austenitic Stainless Steels

45

Fig. 4 Grain-boundary P concentration on HIC in low-temperature acidic environment


at cathodic electrochemical potential.

IV. NONEQUILIBRIUM THERMAL SEGREGATION


Recent ATEM characterizations (1721) in annealed stainless steels have clearly
demonstrated that significant grain-boundary segregation occurs as a result of
high-temperature heat treatment and subsequent rapid cooling. Alloying elements
that are enriched are Cr and Mo compensated by Ni and Fe depletion, as illustrated in Fig. 5 for 316SS. Normal rapid air cooling from the solution anneal
temperature (1100C) can increase boundary Cr and Mo levels by 10 wt%
over that in the matrix. Isolated measurements on stabilized stainless steels indicate that Nb and Ti may also be enriched by a few percent in the solution-annealed
condition. This presegregation is commonly thought to result from a nonequilibrium, vacancy drag process with the degree of boundary enrichment dependent
on the annealing temperature and the cooling rate. Although data are limited,
maximum segregation appears to occur at higher annealing temperatures and at
intermediate cooling rates. The mechanism of this presegregation is not well understood, as indicated by Simonen and Bruemmer (22), who demonstrated that
Cr enrichments cannot be explained by simple solutevacancy interactions. Large
Cr-vacancy binding energies are required to achieve the observed segregation
during quenching, which are completely inconsistent with available data and with
nonequilibrium segregation during irradiation. The strongest nonequilibrium segregant in stainless alloys is probably B, which cannot be easily detected by ATEM
techniques. Quench-induced B segregation has been observed in stainless steels

46

Jones et al.

Fig. 5 Example of nonequilibrium grain-boundary segregation of Mo and Cr during air


cooling from annealing temperatures.

by atom-probe and radiography techniques and in conjunction with Cr and Mo


(20,21). It appears to be likely that B cosegregates with Cr and Mo to grain
boundaries during cooling and promotes the presegregation commonly detected
due to strong binding between B and vacancies and between B and transition
metals. Other elements such as C and N may also reach grain boundaries during
cooling and influence presegregation through interactions with Cr and Mo.
Interfacial enrichment of Cr (and Mo) should impact the local grain-boundary repassivation behavior and may be critical in the IGSCC resistance of stainless steels. Large heat-to-heat and processing variability can produce boundary Cr concentrations ranging from 18% to 30%. Although enhanced IGSCC
resistance might be expected by these large increases in Cr due to improved
passivation behavior, much depends on the mechanism of cracking. Austenite
stability at the boundary will certainly be altered as well as the oxidation characteristics. In addition, B cosegregation with Cr and Mo to interfaces may
influence the chemical and mechanical properties of the grain boundary. Research is needed to assess what role (if any) presegregation plays on the IGSCC
of cold-worked stainless steels. Interfacial enrichment of Cr (or other elements
that strongly oxidize in preference to the base metal) may be detrimental under
specific electrochemical conditions where internal oxidation can occur. Recent
results have indicated that this mechanism may control the IGSCC of NiCr
stainless alloys in high-temperature water at low electrochemical potentials
(23,24).

Austenitic Stainless Steels

V.

47

CRACK INITIATION PROCESSES

The factors controlling crack initiation are expected to be similar to those that
control stress-corrosion cracking, namely environment, stress, and material microstructure and microchemistry. However, surface features resulting from fabrication or heat treatment contribute to initiation of cracks but are not a factor in
crack growth. Examples of features contributing to stress-corrosion crack initiation in stainless steels have been given by Jones (25), Clarke and Gordon (26),
and Licina and Giannuzzi (27). Jones showed a surface feature resulting from
surface grinding in preparation for welding of type 304 SS that would allow
crevice/occluded cell conditions to occur. Clarke and Gordon found that interface
fracture between the matrix and titanium carbonitrides and dissolved silicates
and sulfides were potential crack initiation sites in type 304 SS tested in hightemperature water. Licina and Giannuzzi also showed several examples of surface
irregularities on ground surfaces of type 304 SS, which would act as crevices/
occluded cells and hence crack initiation sites. Therefore, surface grinding can
produce defects that may accelerate crack initiation, although crack initiation will
eventually occur at inclusions or grain boundaries if the environment/material/
stress conditions are within a cracking regime.
There have been a number of studies aimed at measuring crack initiation.
Fatigue crack initiation has been studied extensively, with several studies showing cracks emanating from pits (28). Under stress-corrosion conditions, Clarke
and Gordon (26) studied the sites at which cracks initiated in type 304 SS in
high-temperature water. They observed that fractures of titanium carbonitride
matrix interface and dissolved silicates and sulfides were sites for crack initiation
when these sites were on grain boundaries. Stewart et al. (29) have correlated
electrochemical transients with intergranular crack initiation of type 304 SS. They
performed slow-strain-rate tests in dilute thiosulfate and found numerous electrochemical current transients that the authors correlated with intergranular crack
advance prior to the advance of a single dominate crack. The authors estimated
that each current transient was associated with crack advance equal to about one
grain diameter and that current decay occurred because of crack arrest.
Isaacs (30) used an in situ scanning vibrating electrode technique to identify
locations of electrochemical potential variation on the surface of type 304 SS
stressed in tension to an unknown level in 10-ppm thiosulfate solution at room
temperature. The in situ probe was capable of detecting currents emanating from
growing stress-corrosion cracks, and it was shown that several cracks initiated
and repassivated prior to a dominate crack initiated and continued to propagate.
This result is very consistent with those of Stewart et al. and those obtained with
acoustic emission (25). Locci et al. (31) measured the elongation of samples
loaded uniaxially in a simple beam apparatus used for creep tests. They measured
the crack initiation behavior of ferritic stainless steels in chloride solutions using

48

Jones et al.

this apparatus and found the elongation versus time curve correlated with the
growth of corrosion trenches to sharp cracks. The initiation period correlated
with the transition from flat-bottomed corrosion trenches to 1020-m-long sharp
cracks emanating from the corrosion trenches in a time period of 1050 min for
a 25% Cr ferritic stainless steel loaded to 90% of its yield strength in boiling
42% LiCl at a potential of 460 mVSCE.

VI. FIELD EXPERIENCE


Whereas the relative resistance of stainless steels to pitting, crevice corrosion,
and SCC are well defined in the laboratory, their applications in the field are far
less predictable. This unpredictability is most often the result of poor knowledge
of the actual field conditions such as the stress state, the process constituents
(i.e., chlorides, oxidizers, pH, etc.), fluctuations in temperature, and process upset
conditions. As a result of these inadequacies of knowledge, there are frequent
failures of stainless steels in service.
Therefore, in order to minimize the potential for in-service failures, it is
useful to examine the field performance of the various stainless-steel families,
considering both their successes and failures. It is largely by understanding the
failures that a better appreciation of the limits of these alloys can be achieved.
A. Austenitic Stainless Steels
The austenitic stainless steels are the mainstay of most industries. Although Types
304 and 316 are the most commonly used austenitic stainless steels, there are a
myriad of standard and specialty austenitics that are used for various applications.
The application of these alloys are too numerous to mention here, however, several examples are given to illustrate their use.
The chemical process industry uses a large amount of various austenitic
stainless steels and, therefore, has accumulated a breadth of experience with these
alloys. Figure 6 shows the SCC experience over 10 years with Types 304 and
316 in various plant environments (32). Note that not just the process side can
cause SCC but also exposure to nonprocess streams such as steam and cooling
water can produce SCC of these alloys. In fact, one of the most significant problems with SCC of austenitic stainless steels for more than 40 years has been the
external SCC of austenitic stainless steels under thermal insulation as a result of
either chlorides leaching from the insulation or entrapment of chlorides due to
exposure to seawater or other chloride-containing waters at various times.
The nuclear power industry likewise has had problems with SCC of Type
304 piping and has expended considerable effort to quantify and model the problem. It has been found that among other factors, the solution conductivity has a
profound effect on the crack penetration rate (Fig. 7); that is, the lower the con-

Austenitic Stainless Steels

49

Fig. 6 Stress-corrosion experience for Type 304 and 316 in various plant environments.

ductivity of the solution, the lower the crack penetration rate (33). Factors such
as oxygen content, chloride content, and temperature significantly affect solution
conductivity and are also difficult to consistently maintain in a plant environment.
The need for greater pitting and crevice corrosion resistance of austenitic
stainless steels has led to the development of a new stainless-steel family referred

Fig. 7 Effect of solution conductivity on crack penetration rate in Type 304 stainlesssteel piping.

50

Jones et al.

Table 2 Critical Crevice Temperature (CCT) for Some 6


Month and Other Common Austenitic Stainless Steels
Temperature in
10% FeCl 3 H 2O (pH 1)
UNS No.

Grade

S31254
N08366
N08367
S30403
S31603
N08904

254 SMO
AL-6X
AL6XN
Type 304L
Type 316L
Alloy 904L

90.5
63.5
90.5
27.5
27.5
32

32.5
17.5
32.5
2.5
2.5
0

to as superaustenitics. These alloys generally have 6% Mo and often contain


higher N than the typical austenitics. Both of these elements are known to improve the pitting and crevice corrosion resistance of stainless steels. Table 2
shows the enhanced crevice corrosion resistance of these alloys in ferric chloride
solution compared to the standard austenitic stainless steels (34). This same improved performance has been observed in seawater service in many field applications.
The austenitic stainless steels have been extensively used for equipment in
urea plants. The variety of corrosive environments involved with urea synthesis
(ammonia, carbon dioxide, urea, and ammonium carbamate) and the associated
pressures and temperatures make this process a very complicated one for the
application of alloys. An example of such is the rapid failure of Type 316 and
cast CF8M (316 equivalent) to ammonium carbamate when small amounts of
ferrite are present in the alloy. From experience, it has been found that limiting
the ferrite content of CF8M to less than 2% and increasing the chromium content
can provide excellent resistance to corrosion from carbamate. However, contrary
to this trend and probably as a result of the increased chromium content, the
duplex stainless steels provide even better resistance to carbamate corrosion in
spite of their high ferrite content.
B. Ferritic Stainless Steels
The ferritic steels are the simplest of the stainless-steel family of alloys because
they are principally ironchromium alloys. They are widely used for applications
that require resistance to atmospheric corrosion and are highly resistant to chloride stress-corrosion cracking. They also provide good oxidation resistance at
moderate high temperature. The newer ferritic grades that contain relatively high

Austenitic Stainless Steels

51

Cr and Mo additions have found even wider use in the chemical process industry
in heat exchangers and other equipment. However, one of the major disadvantages of these alloys is their low strength and non-heat-treatability. Even though
the ferritic stainless steels are resistant to SCC from chlorides, they are quite
susceptible to pitting and crevice corrosion in environments containing chlorides,
so this family of alloys more often fails from localized corrosion rather than SCC.

C. Duplex Stainless Steels


The duplex stainless steels have gained wide use in many industries over the
last 20 years. They not only have generally better corrosion resistance than the
austenitics because of the higher Cr content but also have demonstrated superior
SCC. Moreover, they have almost twice the yield strength of the standard austenitics.
Although they often have greater resistance to SCC than the austenitics,
they are certainly not immune. Two recent failures illustrate this limitation. A
North Sea offshore separator vessel constructed from 22 Cr duplex stainless steel
failed by external SCC after it was determined that seawater was soaking through
the insulation onto the hot vessel such that high chlorides then became concentrated on the surface. The vessel operated at 175F (35).
A superduplex stainless-steel manifold was installed in the ocean on a subsea wellhead and exposed to the cathodic protection system which generates hydrogen. Failure occurred quickly as a result of hydrogen embrittlement and plastic
strain in critical areas of the manifold connector (36).
Duplex stainless steels have been successfully used in urea plants and in
the mining industry for portions of pressure acid-leaching plants. They have also
been successfully used for pipelines in the oil and gas industry to carry fluids
that contain CO2.

D. Martensitic Stainless Steels


The martensite stainless steels are widely used because of their combined corrosion resistance and the ability to heat treat them to relatively high strength. However, the high strength also makes these alloys prone to hydrogen stress cracking
and some forms of SCC. The absence of Ni in these alloys provides essential
immunity to chloride SCC from which the austenitic stainless steels often suffer.
Type 410 is one of the most commonly used martensitic stainless steels
and is used for a variety of applications, including fasteners. At high strength
(700 MPa yield), this alloy has suffered from HSC when exposed to seawater
and environments containing H2S. Type 420, a higher-carbon-content version of
Type 410, is used successfully in the oil and gas industry for tubing to resist

52

Jones et al.

corrosion from CO2; however, it is limited to well environments containing little


or no H2S.
Alloys CA6NM and F6NM have been used for many years for pumps compressors, valves, and so forth and were the precursor to the new family of supermartensitic stainless steels that contain 25% Ni and 13% Mo. These alloys
are considered to have superior pitting resistance in chloride-containing environments compared to the standard martensitic stainless steel.
Another martensitic stainless steel that is frequently used is the freemachining grade of Type 410, Type 416, which has reduced pitting and SCC
resistance compared to Type 410 due to the large concentration of nonmetallic
inclusions.
Type 440 stainless steel can achieve very high hardnesses (HRC50), thus
providing good wear resistance. This alloy is often used for surgical instruments
and cutlery, for which the requirement for corrosion resistance is minimal. However, it is highly susceptible to HSC in any environment in which hydrogen ions
are present.

REFERENCES
1. ASTM Volume 3.02, Metals Test Methods and Analytical Procedures. West Conshohocken, PA: American Society for Testing Materials, 1998.
2. AJ Sedriks. Corrosion of Stainless Steels, 2nd ed. New York: WileyInterscience,
1996.
3. G Cragnolino. Cracking of austenitic alloys. Int Metals Rev 3:85135, 1979.
4. G Cragnolino, DD Macdonald. Intergranular SCC of austenitic stainless steel at temperatures below 100CA review. Corrosion 38:406424, 1982.
5. DR McIntyre. Experience SurveySCC of Austenitic Stainless Steels in Water. St.
Louis, MO: Materials Technology Institute, 1987.
6. MA Streicher. Analysis of crevice corrosion data from two seawater exposure tests
on stainless steels. Mater Perform 22:3750, 1983.
7. GS Was, SM Bruemmer eds. Grain Boundary Chemistry and Intergranular Fracture,
Materials Science Forum, 1989, p. 46.
8. PL Andresen, FP Ford, Mater Sci Eng A103:167, 1988.
9. SM Bruemmer. Corrosion 98, 1998, Paper 138.
10. SM Bruemmer, BW Arey, LA Charlot, Corrosion. 48(1):42, 1992.
11. SM Bruemmer, BW Arey, LA Charlot. In: RE Gold, EP Simonen, eds. 6th Int.
Symp. Environmental Degradation of Materials in Nuclear Power SystemsWater
Reactors. The Minerals, Metals & Materials Society, 1993, p. 277.
12. SM Bruemmer, LA Charlot, EP Simonen. In: EP Simonen, RE Gold, DE Cubicciotti.
5th Int. Symp. Environmental Degradation of Materials in Nuclear Power Systems
Water Reactors. American Nuclear Society, 1992, p. 821.
13. RH Jones and SM Bruemmer. In: RP Gangloff, MB Ives. Proc. Environment-In-

Austenitic Stainless Steels

14.

15.
16.

17.
18.

19.
20.
21.
22.
23.
24.
25.

26.
27.
28.
29.
30.

31.
32.
33.

34.
35.
36.

53

duced Cracking of Metals, NACE-10, National Association of Corrosion Engineers,


1990, p. 287.
PL Andresen, CL Briant. In: GJ Theus, JR Weeks. Proc. 3rd Int. Sym. Environmental
Degradation of Materials in Nuclear Power SystemsWater Reactors. The Metallurgical Society, 1988, p. 371.
S Tahtinen, H Hanninen, and M Trolle, ibid 5, pp. 265.
TM Angeliu, et al. In: SM Bruemmer, AR McIlree. 8th Int. Symp. on Environmental
Degradation of Materials in Nuclear Power SystemsWater Reactors. American
Nuclear Society, 1997, p. 649.
P Doig and PEJ Flewitt. Metall Trans A 18A:399, 1987.
J Walmsley, P Spellward, S Fisher, A Jenssen. In: SM Bruemmer, AR McIlree, RE
Gold. 7th Int. Symp. on Environmental Degradation of Materials in Nuclear Power
SystemsWater Reactors. National Association of Corrosion Engineers, 1996,
p. 985.
AW James, CM Shepherd. Mater Sci Technol 5:33, 1989.
L Karlsson, et al. Acta Metall 36(1):1, 1988.
S Dumbill, RM Boothby, TM Williams. Mater Sci Technol 1991, 7:385.
EP Simonen, SM Bruemmer. ibid 19, pp. 751.
PM Scott, M Le Calvar. CorrosionDeformation Interactions, EUROCORR 96. The
Institute of Materials, 1997, p. 384.
SM Bruemmer, LE Thomas, J Daret, PM Scott. Corrosion, 1998.
RH Jones. In Proceedings: Workshop on Initiation of Stress Corrosion Cracking
Under LWR Conditions. EPRI Report NP 5828. Electric Power Research Institute,
Palo Alto, CA, 1988, p. 4-1.
WL Clarke, GM Gordon. ibid, p. 9-1.
GL Licina, AJ Giannuzzi, ibid, p. 13-1.
L Hagn. Mater Sci Eng A 103(1):193, 1988.
J Stewart, PM Scott, DE Williams. CORROSION/88, Houston, TX, 1988, Paper
285.
HS Isaacs. Initiation of stress corrosion cracking of sensitized Type 304 stainless
steel. EPRI report RP1167-8. Electric Power Research Institute, Palo Alto, CA.
1985.
IE Locci, HK Kwon, RF Heheman, AR Troiano. Corrosion 43(8):465, 1987.
M Nakahara. Preventing stress corrosion cracking of austenitic stainless steels in
chemical plants. Nickel Development Institute, 1992.
F Ford, PL Andresen. Unresolved modeling issues and their effects on quantitative
predictions of environmental cracking. Corrosion/89. National Association of Corrosion Engineers, 1989.
RM Davison, JD Redmond. Practical guide to using 6 mo austenitic stainless steel,
Mater Perform 27:39, December 1988.
I Oystetun, KA Johannson, OB Anderson. Offshore Technology Conf., Houston,
TX, 1993, Paper 7207.
TS Taylor, T Pendlington, R Bird. Foinaven superduplex materials cracking investigation. Offshore Technology Conf., Houston, TX, 1999. Paper 10965.

3
Nickel-Based Alloys for Resistance
to Aqueous Corrosion
Paul Crook
Haynes International, Kokomo, Indiana

I.

ADVANTAGES OF NICKEL AS A CORROSION


ALLOY BASE

In the world of metallic materials for (aqueous) corrosion resistance, nickel and
its alloys fill the wide performance gap between the stainless steels and the exotic
materials, such as tantalum. Within this same performance band reside the titanium alloys; however, these have more specific uses.
The advantages of many nickel-based alloys relative to the stainless steels
include the following:
1. Much higher resistance to stress corrosion cracking
2. Higher uniform corrosion resistance, especially in reducing acids, such
as hydrochloric, hydrofluoric, and low to moderate concentrations of
sulfuric
3. Higher resistance to localized attack (pitting and crevice corrosion),
particularly in the presence of chlorides
These advantages stem from three attributes of nickel. First, it is more noble
than iron; second, it exhibits a ductile, face-centered-cubic (fcc) structure at all
temperatures in its solid form; third, it has a high tolerance for useful solutes
(alloying additions), such as chromium and molybdenum.
The nickel-based corrosion alloys may be grouped in several ways. For the
purpose of this chapter, however, they are characterized in terms of the major elemental constituents. Representative, wrought compositions from each group (or
55

56

Crook

family) are given in Table 1. From this table, it is evident that the chief alloying
elements used in the corrosion-resistant nickel alloys are chromium, copper, molybdenum, tungsten, iron, and silicon. Of these, copper, molybdenum, and tungsten
are used to enhance the nobility of nickel; chromium and silicon are used to enhance
passivation; iron is used either to lower the cost of the alloys or to shift positions on
alloy phase diagrams, so that deleterious microstructural precipitates are avoided.
The issue of precipitation within the nickel-based corrosion alloys is extremely complex. However, some discussion of the subject is warranted here. It is
generally accepted that, from a corrosion standpoint, the ideal microstructure is one
that consists of a single phase. A primary concern, therefore, during the design
of the nickel-based alloys has been to avoid phases other than fcc. This has been
accomplished not by restricting alloying additions within the soluble range at room
temperature but by ensuring that a (metastable) single-phase microstructure is possible at room temperature through a process of annealing and quenching. In other
words, the important factors during design have been the high-temperature solubilities and the kinetics of the intermediate-temperature precipitation reactions.
As a result of this approach, it is possible for secondary precipitates to form
during subsequent excursions to intermediate temperatures (e.g., in heat-affected
zones during welding). These precipitates typically form at the grain boundaries
of wrought alloys, as they represent ideal nucleation sites. The extent to which
these precipitates affect the corrosion resistance and the mechanical properties
is related to their atomic structure, shape, composition, and how they influence
the composition of the surrounding fcc solid solution.
To minimize these precipitation reactions, minor elements such as carbon
and silicon are generally held at low levels in the nickel-based wrought alloys.
Silicon is a problem, however, in most nickel-based cast alloys for corrosion
service, because it is necessary for fluidity. An added problem with castings is
elemental segregation (inhomogeneity).
This chapter concerns mostly the performance of the wrought, nickel-based
alloys. These are typically electric-arc melted, then refined by argon-oxygen decarburization. In many cases, they are remelted (e.g., by the electroslag process) for
further refinement and to optimize the structure of the ingot for further processing.
Wrought products such as bars and plates are normally made by hot forging
and/or hot rolling. Sheets are generally made from hot-rolled coils and are typically cold finished prior to final annealing, to achieve tight tolerances and to
control grain sizes.

II. CORROSION-RESISTANT, NICKEL-ALLOY SYSTEMS


A. Ni Alloys
For many applications, the use of commercially pure nickel is warranted. The
most common wrought grade is Nickel 200, the composition of which is given

Nominal Chemical Compositions of Representative Corrosion-Resistant Nickel-Based Alloys (wt%)


Ni

Ni alloys
Nickel 200
NiCr alloys
Alloy 625
NiCu alloys
Alloy 400
Alloy K-500
NiMo alloys
Alloy B-2
B-3 alloy
NiCrMo alloys
Alloy C-4
C-22 alloy
Alloy C-276
C-2000 alloy
Alloy 59
Alloy 686
NiCrSi alloys
D-205 alloy
NiFeCr alloys
Alloy 825
Alloy G-3
G-30 alloy

Cr

99.0 min.

Cu

Mo

0.25 max.
9.0

Fe

Mn

max.
0.40 max. 0.35 max. 0.35 max. 0.15

5.0 max.

max.
2.50 max. 0.50 max. 2.00 max. 0.3
max. 2.7
2.00 max. 0.50 max. 1.50 max. 0.25

61.0

21.5

66.5
66.5

31.5
29.5

69
65

28

2 max.
1 max.
1.5
1.5
0.2 max. 28.5 3 max.

65
56
57
58.5
Bal.
Bal.

16
22
16
23
23
21

1.6

65

20

2.0

2.5

42
44

21.5
22

2.2
2.5

3
7

43

30

1.7

5.5

Si

16
13 3
16 4

16
15.75
61
3.7

3 max.
3
5
3 max.
1.5 max.
5 max.

0.6

0.08 max. 1 max.


0.08 max. 0.5 max.
0.08 max. 1 max.
0.08 max. 0.5 max.
0.1 max. 0.5 max.
0.08 max. 0.75 max.

max.
0.03

30
1.5 max. 19.5

0.5 max.
1 max.

1 max.
1 max.

2.5

1 max.

1.5 max.

0.05max. 0.2 max.


0.015

max.
0.03 max.

0.9

Zr

0.7 max.
0.01 max.

0.01 max.

0.01 max.
0.01 max. 0.5 max.

0.01 max. 0.25


0.14
0.01 max.

0.1 max. 1 max. 0.01 max.

0.1 max. 3 max.


0.01 max. 0.5 max. 0.2 max.

15

0.50 max. 0.50 max. 0.10max. 0.40 max. 0.40 max. 3.6

Cb Ta

Ti

Al

0.2 max. 0.1max.

0.35 max.
0.35 max.

0.5 max.
0.8

Ni-Based Alloys and Aqueous Corrosion

Table 1

57

58

Crook

in Table 1. Several other grades, with tighter limits on the residual elements, are
also available for specific uses (e.g., where enhanced physical or mechanical
properties are needed).
A review of the physical and mechanical properties of annealed Nickel 200
in Table 2 reveals that this group of materials is characterized by moderate density, a high melting range, low tensile strength, and moderately high-tensile elongation. Fabrication (forming and welding) of commercially pure nickel presents
no difficulties, and Nickel 200 is covered by the ASME Boiler Code.
A primary use of commercially pure nickel is hardware for handling caustic
soda (sodium hydroxide), as it resists practically all concentrations and temperatures of this compound. For example, Nickel 200 is favored for the construction
of evaporators, shell and tube heat exchangers, pumps, crystallizers, valves, and
fittings used in the concentration and handling of caustic soda (1). Commercially
pure nickel is also suitable for caustic potash (potassium hydroxide) service.
With regard to the performance of commercially pure nickel in sulfuric acid,
the principal applications of Nickel 200 are at room temperature, in unaerated solutions, or where an organic inhibitor is present (2). In hydrochloric acid, Nickel 200
is useful only at room temperature in air-free solutions, up to a concentration of 10
wt%; corrosion rates at higher temperatures are generally unacceptable (3).
B. NiCr Alloys
This group of materials may also be considered a subset within the NiCrMo
family. However, they have significantly lower molybdenum contents, so are not
quite as resistant to nonoxidizing media and localized attack (pitting, crevice
corrosion, and underdeposit corrosion) as the standard NiCrMo alloys.
Alloy 625 was developed for use as both an aqueous corrosion-resistant
material and as an alloy for use in gaseous, high-temperature environments. One
of its attributes is moderately high strength, as compared with the NiCrMo
alloys, allowing the use of thinner structures. The strength level can be controlled,
to some extent, by varying the annealing temperature. The strength of Alloy 625
is a result of the presence of niobium (columbium) in the solid solution. Alloy
625 is also slightly age-hardenable, by virtue of the sluggish precipitation of the
intermetallic Ni3Cb, although the alloy is not normally used in this condition.
Alloy 625 is very popular in the marine and off-shore industries because
of its high resistance to localized attack in seawater, as compared with many
stainless steels. Undersea applications include exhaust ducts for Navy utility
boats and sheathing for communication cables. Alloy 625 is covered by the
ASME Boiler Code.
C. NiCu Alloys
Copper is very soluble in nickel (in fact, they are mutually soluble in all proportions) and enhances its nobility. The basic NiCu alloys, such as Alloy 400, are

Physical and Typical Room-Temperature Mechanical Properties of Representative Corrosion-Resistant Nickel-Based Alloys
Density

Ni alloys
Nickel 200
NiCr alloys
Alloy 625
NiCu alloys
Alloy 400
Alloy K-500a
NiMo alloys
Alloy B-2
B-3 Alloy
NiCrMo alloys
Alloy C-4
C-22 alloy
Alloy C-276
C-2000 alloy
Alloy 59
Alloy 686
NiCrSi alloys
D-205 alloy
NiFeCr alloys
Alloy 825
Alloy G-3
G-30 alloy

Ultimate tensile strength

0.2% Offset
yield strength

Tensile
elongation
(%)

g/cm3

lb/in.3

8.89

0.321

14351446

26152635

379552

5580

103207

1530

4055

8.44

0.305

12881349

23502460

8271034

120150

414655

6095

3060

8.83
8.47

0.319
0.306

12991349
13161349

23702460
24002460

483621
8961138

7090
130165

172345
586827

2550
85120

3560
2035

9.22
9.22

0.333
0.333

13701418

25002585

896917
862883

130133
125128

400414
400421

5860
5861

5561
5358

8.64
8.69
8.89
8.50
8.60
8.73

0.312
0.314
0.321
0.307
0.311
0.315

13571399
13231371

24752550
24152500

13101360
13381380

23902480
24402516

765807
765800
786793
752779
690 min.
722848

111117
111116
114115
109113
100 min.
105123

338421
359407
359365
345393
340 min.
359421

4961
5259
5253
5057
49 min.
5261

5263
5770
5961
6268
40 min.
5671

7.99

0.288

11711299

21402370

786 (sheet)

114 (sheet)

338 (sheet)

49 (sheet)

57 (sheet)

8.14
8.30
8.22

0.294
0.300
0.297

13711399

25002550

586724
689 (plate)
676689

85105
100 (plate)
98100

241448
310 (plate)
310352

3565
45 (plate)
4551

3050
58 (plate)
5565

K-500 tensile data in the annealed age-hardened condition.

MPa

ksi

MPa

ksi

59

Melting range

Ni-Based Alloys and Aqueous Corrosion

Table 2

60

Crook

therefore ductile, single-phase materials with excellent resistance to non-oxidizing media, such as dilute hydrochloric acid (at low temperatures), hydrofluoric
acid, and dilute sulfuric acid.
For higher strength levels, age-hardenable NiCu alloys, such as Alloy K500, are available. These contain aluminum and titanium to encourage the controlled precipitation of submicroscopic particles of the intermetallic Ni3(Ti, Al).
As may be deduced from Table 2, Alloy K-500 in the aged condition has about
twice the strength of Alloy 400.
With regard to applications of the NiCu alloys, these are wide and varied,
particularly within the chemical process industries and marine engineering. The
chemical process industry uses include pressure vessels (Alloy 400 is covered
by the ASME Boiler Code), heat exchangers, valves, and pumps. Marine engineering uses include propellers and pumps because the NiCu alloys possess
high resistance to degradation in both seawater and brackish water under highvelocity conditions (4). Typical applications of the age-hardenable Alloy K-500
include pump shafts, blades, and scrapers.
D. NiMo Alloys
The first NiMo alloy (known as Alloy B) was developed in the 1920s for use
in cast form. As such, it was limited in performance by elemental segregation
and by precipitates induced by significant carbon and silicon contents. Alloy B
contained 28 wt% molybdenum and 5 wt% iron, a combination which fortuitously
placed the alloy in a fairly safe phase field (i.e., the precipitation of Ni4Mo was
avoided).
B-2 and B-3 alloys (Table 1) are the modern wrought equivalents. In Alloy
B-2, the carbon, silicon, and iron levels were reduced to enhance corrosion resistance, although the reduction in iron caused the alloy to fall in the phase
field, where is the fcc phase and corresponds to the ordered intermetallic
compound Ni4Mo. This phase is extremely deleterious because it forms rapidly
in the temperature range 550800C (especially in cold-worked microstructures)
and reduces both ductility and resistance to stress-corrosion cracking.
By the deliberate addition of minor elements within specific narrow ranges,
the precipitation of Ni4Mo has been avoided in B-3 alloy while maintaining the
very high corrosion resistance (5). Instead, the phase, Ni3Mo, is the stable precipitate. This ordered intermetallic compound takes considerably longer to form,
as it requires more diffusion of molybdenum. By virtue of this slower precipitation reaction the B-3 alloy is much more forgiving of slow cooling, following
hot forging, hot rolling, and annealing, and it is much more tolerant of elevated
temperature excursions during welding.
A study of the physical and mechanical properties of the NiMo alloys
(Table 2) reveals that they possess moderately high tensile strengths, high ductili-

Ni-Based Alloys and Aqueous Corrosion

61

ties, and high melting ranges. With regard to welding and fabrication, care and
knowledge of the precipitation propensities are necessary to achieve good results.
Both B-2 and B-3 alloys are covered by the ASME Boiler Code.
The applications of the NiMo alloys are very specific and related to pure
solutions of nonoxidizing acids (organic and inorganic). B-2 and B-3 alloys, for
example, are resistant to pure sulfuric and hydrochloric acids at nearly all concentrations and temperatures, up to the boiling points. One of the most important
applications, in recent years, has been hardware for making and handling acetic
acid.
E.

NiCrMo Alloys

The NiCrMo alloys are the multipurpose materials of the chemical process
and allied industries. They combine the benefits of molybdenum in nickel with
the advantages of passivity due to chromium. Whereas the NiMo alloys are
unsuitable in the presence of oxidizing contaminants, such as ferric and cupric
ions, and dissolved oxygen, the NiCrMo alloys can cope with such conditions.
They are also extremely resistant to localized attack (pitting, crevice corrosion,
and underdeposit corrosion) in the presence of chlorides.
From Table 1, it is evident that the chromium contents of the modern
wrought NiCrMo alloys range from 16 to 23 wt%, and the molybdenum levels
range from 13 to 16 wt%, in some cases augmented by tungsten. Other deliberate
additions can include iron, to allow the use of less expensive raw materials or
to alter positions within the NiCrMo alloy phase field, copper, to enhance
resistance to nonoxidizing acids, aluminum (for the control of oxygen), manganese (for the control of sulfur), and elements capable of tying up residual carbon,
such as titanium and vanadium.
Although all the NiCrMo alloys are versatile, they each have specific
attributes which are taken into account during alloy selection. The high-chromium alloys (C-22, C-2000, and 59), for example, possess much higher resistance
to oxidizing media than the low-chromium alloys. Also, a high combined molybdenum plus tungsten content (C-276 and 686) is beneficial in nonoxidizing acids,
as is the presence of copper (C-2000). Alloy C-4 has the highest thermal stability
(resistance to sensitization).
From Table 2, it is evident that the NiCrMo alloys possess high ductility,
which is helpful in fabricating components. They are also of moderate strength
in the annealed condition, thus limiting the thicknesses required for pressure
containment. All but the most recently developed NiCrMo alloys are covered
by the ASME Boiler Code, and applications are in process for the newest materials.
With regard to the general performance of the NiCrMo alloys, they are
resistant to a wide range of chemicals, even in the presence of chlorides. They

62

Crook

are generally suitable for use in sulfuric acid up to moderate temperatures [e.g.,
up to 100C in the case of C-2000 alloy (except in the concentration range 75
85 wt%)] (6). The NiCrMo alloys are also suitable for use in hydrochloric
acid, although the concentration and temperature limitations are more severe.
However, several of these alloys provide high resistance to dilute (less than 5
wt%) hydrochloric acid, in some cases up to the boiling points.
The NiCrMo alloys are among those materials able to handle hydrofluoric acid. Also, they withstand organic acids and alkaline media, such as sodium
hydroxide.
With regard to industrial applications of the NiCrMo alloys, these are
many and varied; however, they include reaction vessels, piping, heat exchangers,
tanks, valves, nozzles, and pumps. In the power industry, the NiCrMo alloys
have become the premier materials for lining flue gas desulfurization ducts. Also,
in the oil and gas industry, C-276 is used for pipework in some of the harshest
downhole conditions.
F.

NiCrSi Alloys

D-205 alloy, the composition of which is given in Table 1, is the sole representative of this alloy group. It was developed as an alternate to the high-silicon stainless steels, which also possess outstanding resistance to superoxidizing media,
such as concentrated commercial sulfuric acid.
The advantages of nickel over iron as a basis for a high-silicon alloy include
lower work-hardening rates (which are important in the pressing of heat-exchanger plates) and freedom from sigma formation during elevated temperature
excursions (instead, a less deleterious intermetallic, Ni3Si, which can actually be
used to age-harden the material, forms in D-205 alloy).
In the annealed condition, D-205 alloy is characterized by moderate
strength and high ductility. Unlike the other wrought nickel-based alloys described in this chapter, however, it is not suitable for use in the as-welded condition because of a continuous, brittle, silicon-rich eutectic phase which forms in
the weld metal. The tensile elongation of D-205 weld metal (applied using the
gas tungsten arc process) is only about 3%. Annealing at 1040C (1900F) is
necessary to provide sufficient ductility to these welds for most industrial uses.
Table 2 indicates that D-205 alloy also exhibits a relatively low density, compared
with the other nickel-based alloys, and a low solidus temperature. This low solidus limits the annealing temperature range.
So far, the predominant use of D-205 alloy has been in the form of plate
heat exchangers for the cooling of hot, concentrated sulfuric acid, where it has
replaced cast iron (cascade coolers) and anodically protected stainless steel (shell
and tube heat exchangers).

Ni-Based Alloys and Aqueous Corrosion

G.

63

NiFeCr Alloys

Although grouped together because of their major alloying elements, Alloy 825,
Alloy G-3, and G-30 alloy (Table 1) have distinctive characteristics and are suitable for different market segments.
Alloys 825 and G-3 were designed as multipurpose materials to fill the gap
between the high-nickel stainless steels (which are more prone to stress-corrosion
cracking) and the NiCr and NiCrMo alloys. Both 825 and G-3 offer good
resistance to sulfuric acid, by virtue of their combined molybdenum and copper
contents. With its higher molybdenum level, Alloy G-3 is significantly more resistant to pitting and crevice corrosion in the presence of chlorides.
G-30 alloy was designed specifically for service in commercial phosphoric
acid, which is the primary chemical in the agrichemical industries. For this type of
service, a high chromium content was found to be desirable, in addition to copper
and molybdenum. As a result of its high chromium content, G-30 alloy is also
suitable for strong, oxidizing media, such as nitric-acid-based pickling solutions.
A review of their respective physical and mechanical properties (Table 2)
reveals that the NiFeCr alloys are characterized by moderate strengths, in the
annealed condition, and high ductilities. All three alloys chosen as representatives
of this category are covered by the ASME Boiler Code.
Applications of Alloy 825 include equipment for containing and handling
sulfuric acid sludges in petroleum refineries (i.e., tanks, heat exchangers, piping,
valves, and pumps) and for petrochemical processes which employ phosphoric
acid as a catalyst. The main use of G-30 alloy in the agrichemical industries has
been for shell and tube heat exchangers.
III. NICKEL ALLOYS IN SULFURIC ACID
Sulfuric acid is one of the most important industrial chemicals and pervades not
only the chemical process industries but also the mining/metal extraction, metal
finishing, and agrichemical industries.
The performance of the nickel-based alloys in sulfuric acid is very much
dependent on acid concentration, temperature, and the presence of other species
in the solution.
In Ref. 7, three concentration regimes are defined as follows:
1. Low-concentration acid
2. Intermediate-concentration acid
3. High-concentration acid
By compiling and comparing data for many nickel-based alloys and stainless
steels, it was deduced that the most beneficial elements in low-concentration acid

64

Crook

are nickel itself, molybdenum (or tungsten), and chromium. At low concentrations, sulfuric acid is nonoxidizing (reducing), and the cathodic reaction is hydrogen evolution.
The intermediate-concentration range is defined as approximately 2060
wt%. Here too, the cathodic reaction is believed to be hydrogen evolution; hence,
the solutions are nonoxidizing. In this concentration range, the NiMo alloys are
outstanding, so it is presumed that molybdenum is one of the most beneficial
alloying elements. Reference 7 also infers that copper and silicon are beneficial
alloying elements within this concentration range.
The high-concentration acid is complex, both in electrochemical terms (the
redox potential exhibits a steep increase once the concentration exceeds about 70
wt%) and in terms of the influence of the elements on performance. Certainly, the
NiMo alloys continue to resist corrosion at these high concentrations, and silicon
can become very beneficial above about 90 wt%. However, the corrosion performance of the NiCu alloys deteriorates markedly as the redox potential increases.
A review of the data in Ref. 7 leads to the following conclusions:
1. Of the nickel alloys, those based on the NiMo system are the most
resistant to pure sulfuric acid, at concentrations up to about 95 wt%;
however, these should not be used in aerated systems or where other
oxidizing species are present.
2. The NiCrMo and NiFeCr systems constitute the next best choice
at concentrations up to about 95 wt%; these alloys will also resist sulfuric acid in the presence of oxidizing species.
3. In the low- and intermediate-concentration ranges, the NiCu alloys
are also worthy of consideration.
4. At very high concentrations of sulfuric acid, the NiCrSi alloy may
be the most suitable alloy, especially in the presence of impurities.
Iso-corrosion diagrams for B-3 and C-2000 alloys in sulfuric acid are presented in Figs. 1a and 1b, respectively. These indicate the very safe (0 to 5 mpy),
moderately safe (520 mpy), and unsafe (over 20 mpy) concentration/temperature regimes for these alloys in pure sulfuric acid.
For a perspective, it should be stated that the NiCrMo and NiFeCr
alloys are within the same performance band as Alloy 20 (stainless steel) at concentrations up to about 60 wt%; however, they are considerably better than Alloy
20 in the approximate range of 6080 wt%. Relative to 316L stainless steel, the
NiCrMo and NiFeCr possess much higher resistance to sulfuric acid at
concentrations up to about 85 wt%.
A comparison of the NiCrSi alloy (D-205) and a high-silicon stainless
steel (Fe17 Cr20 Ni5 Si) is shown in Fig. 2. This indicates their relative
resistance to commercial-grade sulfuric acid at 130C, in the concentration range

Ni-Based Alloys and Aqueous Corrosion

65

(a)

(b)
Fig. 1 (a) Iso-corrosion diagram for B-3 alloy in sulfuric acid; (b) iso-corrosion diagram
for C-2000 alloy in sulfuric acid.

9699 wt%. A plot of corrosion rate versus temperature for D-205 alloy in 99
wt% acid is shown in Fig. 3; this indicates an upper temperature limit of 150C.

IV. NICKEL ALLOYS IN HYDROCHLORIC ACID


Hydrochloric acid is another extremely important industrial chemical. It is both
a feedstock and by-product in the organic chemical industry. It is very corrosive,

66

Fig. 2

Crook

Comparative corrosion rates in commercial-grade sulfuric acid at 130C.

and only a few alloy systems are suitable for use in the acid at temperatures
above ambient. As a result of its volatility, high concentrations are rarely encountered, except in pressurized systems. In a flask/condenser system, for example,
the highest concentration that is stable at the boiling point is 20 wt%.
Many nickel-based alloys can be used in hydrochloric acid at temperatures

Fig. 3 Corrosion rate versus temperature for D-205 alloy in 99% commercial-grade sulfuric acid.

Ni-Based Alloys and Aqueous Corrosion

67

(a)

(b)
Fig. 4 (a) Iso-corrosion diagram for B-3 alloy in hydrochloric acid; (b) iso-corrosion
diagram for C-2000 alloy in hydrochloric acid.

at and above ambient. Again, their performance is generally a function of temperature, concentration, and the impurities present. The most useful nickel-based
alloys for hydrochloric acid service are those of the NiMo and NiCrMo systems. As shown in the iso-corrosion diagram for B-3 alloy (Fig. 4a), the NiMo
alloys exhibit low corrosion rates in pure (reagent-grade) hydrochloric acid, in
the concentration range 120 wt%, even up to the boiling points. As in sulfuric
acid, however, the NiMo alloys should not be used in aerated systems or in the
presence of other oxidizing species, such as ferric ions. The same restrictions

68

Crook

apply to the Ni and NiCu alloys, which are useful in low concentrations of
hydrochloric acid at low temperatures.
The NiCrMo alloys not only exhibit low corrosion rates in hydrochloric
acid over reasonably wide concentration and temperature ranges (as shown in
the iso-corrosion diagram for C-2000 alloy, in Fig. 4b), but they also are tolerant
of oxidizing species. However, there are considerable differences in performance
between individual compositions within this alloy system, so attention to existing
corrosion data or field testing are warranted.

V.

NICKEL ALLOYS IN HYDROFLUORIC ACID

Although not as widely used as hydrochloric acid, hydrofluoric acid is becoming


increasingly important as a feedstock and by-product of the organic chemical
industry. With it come serious safety concerns, because it such a hazardous compound, and difficulties of containment, because hydrofluoric acid attacks glass.
The corrosion rates for several nickel-based alloys in hydrofluoric acid at
79C (175F) in the concentration range 548 wt% are given in Table 3. Also
presented in Table 3, for comparison, are corrosion data for several stainless
steels (i.e., the most widely used austenitic grade, one from the high-molybdenum
family, one from the alloy 20 group, and one ferriticaustenitic/duplex stainless

Table 3 Corrosion Data for Nickel-Based Alloys and Stainless Steels in


Hydrofluoric Acid
Corrosion rates (mpy) at 79C (175F)

Alloy 625
Alloy 400
Alloy B-2
B-3 alloy
C-22 alloy
Alloy C-276
C-2000 alloy
Alloy 825
G-30 alloy
316L stainless steel
904L stainless steel
20Cb-3 alloy
Alloy 255

5% HF

10% HF

20% HF

48% HF

44.7
16.6
11.8
12.4
25.0
15.9
10.8
31.8
34.0
3,877.0
457.0
30.0
802.0

349.0
16.6
13.8
16.3
31.1
19.0
21.5
44.1
89.3
11,043.0
780.0
29.0
1,528.0

934.0
22.4
17.8
20.7
51.6
34.7
20.2
40.8
83.6
17,760.0
2,449.0
37.3
4,682.0

1,576.0
40.4
25.5
30.6
27.0
36.8
19.4
127.0
269.0
21,336.0

59.7
7,247.0

Ni-Based Alloys and Aqueous Corrosion

69

steel). From these data, it is evident that the NiCu alloys, the NiMo alloys,
and certain NiCrMo alloys can be used in hydrofluoric acid at this temperature,
but with moderate rates of attack. The NiCr and NiFeCr alloys appear to be
unsuitable at this temperature and within this concentration range, as do the austenitic and ferriticaustenitic stainless steels. At much lower temperatures, the
NiCr and NiFeCr alloys may be used in dilute hydrofluoric acid, as may the
copper-containing stainless steel, 20Cb-3 alloy.
The performance of the NiCu and NiMo alloys in hydrofluoric acid is
strongly influenced by the presence of oxygen, as it is in sulfuric and hydrochloric
acids. Considerable increases in the corrosion rates of Alloy 400 in 38% and
48% hydrofluoric acid (boiling) have been recorded, even in the presence of 500
ppm oxygen (9).
The Ni alloys, such as Nickel 200, are not as resistant as Alloy 400 to
hydrofluoric acid. They are also more strongly affected by oxygen in this acid.
Their use is therefore limited to oxygen-free solutions, at temperatures below
about 79C (175F) (9).

VI. NICKEL ALLOYS IN PHOSPHORIC ACID


Two distinctly different types of phosphoric acid are encountered in industry
(10). The pure (reagent-grade) acid is made from elemental phosphorus, derived
from phosphate rock. This is oxidized, then reacted with water.
The acid which pervades the agrichemical industries, on the other hand, is
made by reacting phosphate rock with sulfuric acid and contains several impurities, such as hydrofluoric acid, sulfuric acid, silica, and chlorides. The levels of
these impurities vary depending on the source of the rock, and different batches
of this so-called commercial grade of phosphoric acid can vary considerably in
their corrosivity.
Generally, the commercial grade of phosphoric acid is more corrosive than
the reagent grade, and in the commercial grade, a high chromium content has
been shown to be extremely beneficial. Test data in 60 wt% commercial phosphoric acid at 116C are presented for several nickel-based alloys and stainless
steels in Fig. 5. These tests were performed under laboratory conditions in a
solution provided by an agrichemical company.
From these data, it is evident that a high chromium content is beneficial,
whether the alloy is nickel based or a stainless steel. Of the materials tested, G30 alloy (30 wt% chromium) exhibited the lowest corrosion rate, whereas Alloy
28 (27 wt% chromium) was the best of the stainless steels.
Several nickel-based alloy systems are suitable for use in pure (reagentgrade) phosphoric acid. These include the NiCr, NiCu, NiMo, NiCrMo,
and NiFeCr alloys.

70

Crook

Fig. 5 Corrosion rates for nickel-based alloys and stainless steels in 60% commercial
phosphoric acid at 116C.

VII. NICKEL ALLOYS IN NITRIC ACID


Nitric is a strong, oxidizing acid. Thus, chromium is an extremely beneficial
alloying element in nitric acid solutions, as it readily provides passivation. In
general terms, the stainless steels are more resistant to nitric acid than the chromium-bearing nickel-based alloys. However, there are occasions when nickelbased alloys are preferred:
1. In heat exchangers, where the stainless steels might not possess sufficient pitting resistance on the cooling water side
2. In multiple-purpose chemical systems, where a batch process involving
nitric acid might be followed by another involving a different acid,
such as hydrochloric
3. In acid mixtures, where the second acid induces the degradation of
stainless steels
From Fig. 6, a chart comprising the corrosion rates for several nickel-based
alloys and stainless steels in boiling 65% nitric acid, it is evident that care must
be taken in choosing a suitable alloy. For example, Alloy C-276, with a chromium
content of only 16 wt%, corrodes rapidly in nitric acid. Depending on the requirements for pitting resistance and the other corrosive species encountered, a highchromium NiCr, NiFeCr, or NiCrMo might be appropriate.

Ni-Based Alloys and Aqueous Corrosion

71

Fig. 6 Corrosion rates for nickel-based alloys and stainless steels in 65% boiling nitric
acid.

VIII. NICKEL ALLOYS IN CHLORIDES


Chlorides, even in small amounts, are among the most damaging of chemicals,
because they can induce pitting and crevice corrosion in materials which normally
exhibit passive behavior. Once initiated, these forms of attack progress at unpredictable rates, often causing equipment to perforate, with a subsequent spillage.
Of course, the most abundant chloride-containing solution is seawater, which
is commonly used as a coolant in heat exchangers. At high concentrations and
temperatures, chlorides can cause stress-corrosion cracking.
Chlorides are particularly damaging to the austenitic stainless steels, which
are not as resistant to localized attack (i.e., pitting and crevice corrosion) and
stress-corrosion cracking as most chromium-bearing nickel-based alloys.
Ferric and cupric chlorides are of particular concern, because ferric and
cupric ions can markedly alter the nature of the electrochemical process, leading
to much higher potentials, hence higher corrosion rates, in the absence of passivation. Thus, alloys such as those in the NiCu and NiMo systems should not
be used when these compounds are present.
With regard to the suitability of the various nickel-alloy systems for service
in chlorides, those in the NiCrMo family are the most appropriate, because
they not only possess very high resistance to pitting, crevice corrosion, and stresscorrosion cracking, but also they are passive at high potentials, such as those
induced by ferric and cupric ions. Alloys in the NiFeCr and NiCr families

72

Crook

Fig. 7 Critical pitting temperatures for nickel-based alloys and stainless steels in an
oxidizing NaCl/HCl solution.

exhibit moderate resistance to localized attack and are therefore suitable under
mild conditions.
To provide some perspective, pitting and crevice corrosion data for several
nickel-based alloys and stainless steels are presented in Figs. 7 and 8. The values
represent the lowest temperatures at which pitting (Fig. 7) and crevice corrosion
(Fig. 8) occur in a solution of 4 wt% NaCl 0.1 wt% Fe2 (SO4)3 0.01 M HCl.
From these data, it is evident that the temperature required to induce crevice
corrosion is generally lower than that needed for pitting to occur. It is also evident
that the NiCrMo alloys are by far the best nickel-based alloys, with regard to
resistance to localized attack in oxidizing chloride solutions.

IX. NICKEL ALLOYS IN HYDROXIDES


Sodium hydroxide (caustic soda) is the most widely used alkaline material (1). As
with acids, the concentration, temperature, and impurities are the most important
factors with regard to alloy performance. At low temperatures, iron and steels
can be used for handling sodium hydroxide; however, at elevated temperatures,
these are subject to caustic embrittlement, and other materials must be considered.
Preeminent among the alloy choices for elevated-temperature caustic soda is
commercially pure nickel (Nickel 200), which exhibits corrosion rates of less
than 1 mpy in boiling solutions up to a concentration of about 50 wt%.

Ni-Based Alloys and Aqueous Corrosion

73

Fig. 8 Critical crevice corrosion temperatures for nickel-based alloys and stainless steels
in an oxidizing NaCl/HCl solution.

With regard to the performance of other nickel alloy systems in sodium


hydroxide, it has been reported that the NiCu alloys are practically as resistant
as commercially pure nickel (1). Other nickel-based alloys can be used in sodium
hydroxide over fairly wide concentration and temperature ranges, although there
is some evidence that the NiCr and NiCrMo alloys are susceptible to caustic
stress-corrosion cracking at very high temperatures and concentrations.
In general, the performance of the nickel-based alloys in potassium hydroxide (caustic potash) mirrors that in caustic soda. However, in ammonium hydroxide, which is also important in industry, commercially pure nickel and the Ni
Cu alloys are not recommended, whereas most other nickel-based alloys resist
all concentrations up to the boiling points (1,10).

REFERENCES
1. Resistance of Nickel and its Alloys to Corrosion by Caustic Alkalies. Corrosion
Engineering Bulletin 2, The International Nickel Company.
2. Resistance of Nickel and High Nickel Alloys to Corrosion by Sulfuric Acid. Corrosion Engineering Bulletin 1, The International Nickel Company.
3. Resistance of Nickel and High Nickel Alloys to Corrosion by Hydrochloric Acid,
Hydrogen Chloride and Chlorine. Corrosion Engineering Bulletin 3, The International Nickel Company.
4. MONEL NickelCopper Alloys, 3rd ed. The International Nickel Company, 1978.

74

Crook

5. DL Klarstrom. A new NiMo alloy with improved thermal stability. Proceedings


of the 12th International Corrosion Congress, 1994.
6. P Crook, ML Caruso. The corrosion resistance of NiMo and NiCrMo alloys
in sulfuric and hydrochloric acids. Proceedings of the 13th International Corrosion
Congress, 1996.
7. N Sridhar. Behavior of high-performance alloys in sulfuric acid. Mater Perform
1988; 27(3):40.
8. P Crook, B Ornberg. A new nickelchromiumsilicon alloy for plate heat exchangers. Proceedings of the 10th European Corrosion Congress, EFC, 1993.
9. Corrosion Resistance of Nickel-Containing Alloys in Hydrofluoric Acid, Hydrogen
Fluoride and Fluorine. Corrosion Engineering Bulletin 5, The International Nickel
Company.
10. N Sridhar. Behavior of nickel-base alloys in corrosive environments. Metals Handbook, 9th ed., ASM International, 1987, Vol. 13, p. 643.

4
Nickel-Based Alloys for Resistance
to High-Temperature Corrosion
Mark A. Harper
Special Metals Corporation, Huntington, West Virginia

George Y. Lai
Consultant, Carmel, Indiana

I.

INTRODUCTION

Similar to (aqueous) corrosion resistance alloys, the advantages of the Ni-based


alloys become evident when the combination of environment plus temperature
become too severe for the stainless steels. For materials at high temperatures
(e.g., T 1000F), corrosion problems can be very complex, with relatively
small amounts of impurities (e.g., Cl in an O2 environment) causing significant
changes in the corrosion behavior of a particular alloy. Depending on the environment, which will dictate the general characteristics of the corrosion reaction, eight
different corrosion modes can typically be identified in an industrial process.
They are oxidation, sulfidation, carburization (including metal dusting), nitridation, halogen corrosion, ash- and salt-deposit corrosion, molten-salt corrosion,
and molten-metal corrosion.
Although, the criteria for the formation of an oxide scale is simply governed
by the thermodynamics of the system (i.e., enough oxygen must be present such
that an oxide scale is thermodynamically stable), the continued growth of the
scale and its ability to reform when damage, spalling, and so forth occurs is
usually governed by kinetic factors. Most industrial environments contain enough
oxygen such that the oxidation reaction participates in the corrosion process and
a protective oxide scale is usually relied upon by most high-temperature alloys
for protection against the various modes of high-temperature attack. Also, it is
logical that the formation and growth of the oxide scale and its ability to protect
75

76

Harper and Lai

the underlying alloy are more probable in an oxidizing environment (i.e., high
oxygen activity, excess oxygen) than in a reducing environment (i.e., low oxygen
activity, no excess oxygen). With the formation and growth of oxide scales occurring more slowly under reducing conditions, these types of environments are
usually much more aggressive/corrosive than oxidizing environments.
Regarding the corrosion reaction(s) occurring in an industrial process, it is
important to note that the oxygen activity present in the system is usually an
important influence on the rate of attack. For example, sulfidation attack is controlled by both the sulfur and oxygen activities, with an increase in the oxygen
activity causing a decrease in the sulfidation attack, and vice versa. Carburization
and nitridation behave in a similar manner. However, halogen corrosion, moltensalt corrosion, and molten-metal corrosion behave differently (i.e., oxidizing environments tend to be more corrosive than reducing environments).
Thus, an alloy selection process must take into consideration the nature of
the corrosive environment and the mode of corrosion attack. Once these two
items have been determined, an adequate corrosion database is required in order
for a materials engineer to make an informed decision on the appropriate alloy
selection.
The nominal compositions of several Ni-based wrought alloys used for
high-temperature service are shown in Table 1. A more complete list of alloys
can be found in Ref. 1. As shown in this table, the levels of carbon, and sometimes
nitrogen, in these alloys are relatively high. This is primarily due to the use of
the various metal carbides to provide high-temperature strength. Other elements
such as Mo, Nb, and W are used for solid-solution strengthening. Although not
discussed in this chapter, the high-temperature strength, creep, and thermal fatigue usually play as important a role as the high-temperature corrosion resistance
of the alloy. The remainder of this chapter will discuss the various modes of
high-temperature corrosion and a review of the available data that exist for Nibased alloys in the various types of environment. Using this information, a materials engineer can make a more informed decision on the appropriate alloy selection
for a given application.

II. OXIDATION
Oxidation is the most common and important high-temperature corrosion reaction, and many alloys rely on the protective oxide scale to prevent attack of the
alloy from sulfidation, carburization, and so forth. When the environment contains corrosive impurities (e.g., sulfur, chlorine, etc.), the principal corrosion
mode may no longer be oxidation, even though it may still be part of the overall
corrosion reaction. For this reason, this section considers oxidation primarily associated with air and clean combustion atmospheres generated by using clean

Nominal Composition of Alloys (wt%)

Alloy name
214 alloy

UNS alloy
No.

Cr

Ni
b

Co

Fe
3

0.05

16

75

600 alloy
S alloy
601 alloy
625 alloy
230 alloy
617 alloy
X alloy
263 alloy
HR-160 alloy
HR-120 alloy
RA330 alloy
800HT alloy
556 alloy

N06600

N06601
N06625
N06230
N06617
N06002

N12160
N08120
N08330
N08811
R30556

0.10 a
0.02 a
0.05
0.10 a
0.10
0.07
0.10
0.06
0.05
0.05
0.05
0.08
0.10

15.5
16
23
21
22
22
22
20
28
25
19
21
22

72 c
67 b
Bal
62 b
57 b
Bal
47 b
52 b
37 b
37
35
32.5
20

2a

1a
5a
12.5
1.5
20
30
3a

18

7
3a
14.1
5a
3.0a
1.5
18
0.7 a
3.5 a
33 b
Bal
Bal
31 b

RA85H alloy

S30615

0.20

18.5

14.5

Bal

Mo

Others

4.5 Al, 0.5 Mn , 0.2 Sia , 0.1 Zra , 0.01 Ba , 0.01


Y

2.25 NbTa, 1 Mn a , 0.75 Si a , 0.5 Cu a


15
0.5 Mn, 0.4 Si, 0.25 Al, 0.015 Ba , 0.002 La
1a

1.4 Al, 0.5 Mn, 0.2 Si


9

3.7 NbTa, 0.5 Mna , 0.5 Sia , 0.4 Ala , 0.4 Tia
2
14
0.5 Mn, 0.4 Si, 0.3 Al, 0.02 La, 0.015 B a
9

1 Al
9
0.6
1 Mn a , 1 Si a , 0.008 B a
6

2.4 Ti a , 0.6 Al a , 0.6 Mn a , 0.4 Si a , 0.2 Cu a


a
a
1.0
1.0
2.75 Si, 1 Nb a , 0.5 Mn
2.5 a
2.5 a 0.7 Mn, 0.7 Nb, 0.6 Si, 0.2 N, 0.1 Al, 0.004 B
1.25 Si

0.8 Mn, 0.5 Si, 0.4 Cu, 0.4 Al, 0.4 Ti

3
2.5
1 Mn, 0.6 Ta, 0.4 Si, 0.2 N, 0.2 Al, 0.02 Zr,
0.02 La
3.4 Si, 1 Al

Ni-Based Alloys and High-Temperature Corrosion

Table 1

Note: 214, 230, HR-160, HR-120, and 556 are trademarks of Haynes International. RA330 and RA85H are trademarks of Rolled Alloys, Inc. 800HT is a
trademark of the INCO Family of companies.
a
Maximum.
b
As balance.

77

78

Harper and Lai

fuels (e.g., natural gas or number one and/or number two fuel oil). These fuels
generally contain low levels of contaminants such as sulfur, chlorine, alkali metals, and vanadium.
Many oxidation problems result from using an alloy in a temperature region
that exceeds its protective capability; that is, for a given alloy in service at excessively high temperatures, significant scaling occurs. In response to this problem,
a large database on the oxidation of commercial alloys, ranging from carbon
steels to superalloys, is available in the literature. Although not discussed in this
section, the comparative oxidation resistance of carbon and low-alloy steels, 9
Cr and 12 Cr steels, 17 Cr and 27 Cr stainless steels and austenitic stainless
steels at temperatures between 480C, and 930C is summarized in The Making,
Shaping and Treating of Steel (2). Also, Eiselstein and Skinner (3) summarized
the comparative oxidation resistance of austenitic stainless steels, FeNiCr
alloys, and Ni-based alloys at 980C. Nickel-based alloys performed significantly
better than stainless steels.
Most commercial high-temperature alloys rely on a chromia (Cr 2O3 ) scale
for protection at elevated temperatures and these alloys are often called chromia
formers. However, at temperatures above 1000C, Cr2O3 exhibits significant
volatilization to CrO3 and, thus, diminishes the protective capability of a chromiaforming alloy to the oxidation attack. Several chromia-forming Ni-based alloys
have been found to exhibit good oxidation resistance at 982C (1800F) and
higher. For example, Incos Inconel 601 and 625 alloys and Haynes Internationals Hastelloy S and X alloys, as well as the Haynes 230 alloy, suffered less
than 25 m after 1008 h in air at 982C. For the same conditions at 1093C
(2000F), many alloys (e.g., 600, 230, S, and 617) exhibited less than 50 m of
total attack. Even at 1149C (2100F), several alloys (e.g., 600, 230, S, and 617)
showed less than 100 m of oxidation attack. However, at 1204C (2200F), the
oxidation rate increases significantly for the chromia formers, and only a few
alloys exhibited less than 250 m of attack for 1008 h exposure in air. These
results, along with the results of other Ni-based alloys and a few austenitic stainless steels, are shown in Table 2 (4). At this temperature, alloys that form and
use an alumina (Al2O3) scale for protection are preferred. For example, the 214
alloy exhibited less than 18 m of oxidation attack when exposed to air at 1204C
for approximately 1000 h.
It is important to evaluate an alloys long-term oxidation behavior and potential for breakaway oxidation. Harper et al. (5) conducted long-term oxidation
studies on several Ni-based alloys at elevated temperatures for periods of 1 and
2 years. Figure 1 shows the weight change as a function of exposure time for
the three alloys (HR-120, 800HT, and RA85H) that were exposed to still air at
982C for 720 days. This figure demonstrates the importance of long-term testing
and provides a good example of breakaway oxidation. As shown in this graph,

982C

Alloy

Metal
loss
(m)

Average metal
affected a
(m)

Metal
loss
(m)

Average metal
affected a
(m)

Metal
loss
(m)

214
S
230
617
601
600
X
625
556
800H
RA 330
310
304
316

1.8
4.6
6.4
7.9
13.5
8.1
8.6
8.1
9.9
23.9
10.2
8.9
140.7
314.2

5.1
12.4
18.0
33.3
32.0
22.9
23.9
18.3
26.7
45.5
108.5
28.7
205.7
362.0

2.0
11.2
11.4
16.3
30.7
27.9
37.8
83.1
24.6
136.9
20.8
24.6
Consumed
Consumed

2.0
32.8
32.3
46.5
67.1
41.4
69.1
121.9
65.3
187.7
170.2
57.4
689.4 b
1737.4b

3.8
25.7
58.2
27.4
59.9
43.9
114.3
405.4
236.5
191.0
40.6
75.4
Consumed
Consumed

a
b

1204C

1149C

1093C

Average metal
affected a
(m)
4.1
42.2
87.4
85.1
133.9
72.6
148.1
462.3
295.7
255.0
221.7
112.8
549.9 b
2667 b

Metal
loss
(m)

Average metal
affected a
(m)

5.6
Consumed
115.1
269.5
112.3
129.8
Consumed
Consumed
Consumed
286.3
95.8
201.9
Consumed
Consumed

16.5
805.2 a
201.4
316.7
191.5
213.9
899.2 b
1209 b
3810 b
344.2
209.6
260.4
1725.9 b
3566.2 b

Ni-Based Alloys and High-Temperature Corrosion

Table 2 Metal Loss and Average Metal Affected for Various Alloys Exposed to Flowing Air for 1008 h and Cycled to Room
Temperature Once a Week

Average metal affected Metal loss Average internal penetration.


Extrapolated to 1008 h.

79

80

Harper and Lai

Fig. 1 Weight change versus time for long-term oxidation in still-air testing of HR-120,
RA85H, and 800HT alloys at 982C for 720 days. (From Ref. 5.)

for exposures up to 180 days (4320 h), all three alloys appear to be equivalent
from a weight change perspective. However, between 180 and 210 days of exposure, the 800HT sample exhibited an accelerated rate of weight loss, characteristic
of breakaway oxidation. The RA85H sample exhibited similar behavior after 360
d of exposure. This breakaway oxidation is a result of the transition from a protective Cr 2O3 scale to a less protective Fe,Ni,Cr-spinel oxide scale (6). Table 3 shows
the results of the testing conducted at 982C for a total exposure period of 720
days. The samples were exposed to still air and were subjected to a thermal cycle
to room temperature once every 30 days. Figure 2 provides an easy comparison
of the alloys exposed in this test and shows the relative amounts of attack that
occur via metal loss (i.e., actual thickness loss of the sample) and internal attack
caused by internal oxidation and/or void formation. Table 1 also shows the results
of similar testing for 360 days of exposure at 1093, 1149, and 1204C, and Figs.
35 show the corresponding graphs of the metal loss plus internal attack. Similar
to the results described above, very few chromia-forming alloys can survive longterm exposure at temperatures approaching 1200C. In contrast, Table 1 and Fig.
5 show the excellent oxidation resistance of an alumina-forming alloy at these
high temperatures.

982C720 days

Alloy

Metal
loss
(mm)

Average metal
affected
(mm)

214
230
617
601
556
HR-160
HR-120
RA85H
800HT

0.00
0.00
0.01
0.02
0.06
0.04
0.16
0.53

0.15
0.24
0.57
0.39
0.42
0.27
1.36
2.03

1093C360 days
Metal
loss
(mm)

Average metal
affected
(mm)

0.05

0.14
0.36
0.09
0.83
0.45
1.13

Metal
loss
(mm)

0.27

1.15
0.54
0.74
0.97
2.04
1.30

Average metal
affected
(mm)

0.28
0.54
0.32
Consumed
0.19
1.11
0.51
1.66

Metal
loss
(mm)

0.86
0.94
1.85
6.29
1.49
1.35
2.41
1.79

Average metal
affected
(mm)
0.02

0.01

1204C360 days

1149C360 days

1.51
2.23
0.70

0.42
Consumed
0.78
Consumed

2.64
6.36
2.95

2.62
6.37
6.39
6.35

Ni-Based Alloys and High-Temperature Corrosion

Table 3 Metal Loss and Average Metal Affected for Various Alloys Exposed to Still Air and Cycled to Room Temperature Once
Every 30 Days

81

82

Harper and Lai

Fig. 2 Average metal affected for various alloys exposed to still air at 982C for 720
days. Samples cooled to room temperature once every 30 days.

Fig. 3 Average metal affected for various alloys exposed to still air at 1093C for 360
days. Samples cooled to room temperature once every 30 days.

Ni-Based Alloys and High-Temperature Corrosion

83

Fig. 4 Average metal affected for various alloys exposed to still air at 1149C for 360
days. Samples cooled to room temperature once every 30 days.

Fig. 5 Average metal affected for various alloys exposed to still air at 1204C for 360
days. Samples cooled to room temperature once every 30 days.

84

Harper and Lai

Fig. 6 Approximate rate constants as a function of temperature for various oxides. (From
Ref. 7.)

The results of the 214 alloys shown in Table 2 and 3 are not surprising
because alumina formers are known to be more oxidation resistant than chromia
formers. Figure 6 shows a plot of the parabolic rate constants for various oxides
(7). As shown in this figure, the growth of Al2O3 is at least two orders of magnitude slower than the growth of Cr2O3. However, it should be noted that some
studies have shown that these alloys are prone to the formation of internal, randomly distributed voids. In particular, the mechanically alloyed, oxide-dispersion-strengthened (ODS) alloys have been shown to be more susceptible to this
type of void formation than conventional wrought alloys (810).
In addition to the use of alumina-forming alloys at temperatures approaching 1200C, these alloys are finding use in applications/processes where
contamination of the product is a critical issue. An example of this type of application is the furnace equipment used to process electronic components (e.g., semiconductors, capacitors, etc.), glass, and chinaware. In these processes, the major
source of contamination is the oxide spalled off of the furnace components, such
as wire mesh belts, baskets, and fixtures. These components are typically made
from chromia-forming alloys. However, due to the slower growing and more

Ni-Based Alloys and High-Temperature Corrosion

85

adherent oxide scale formed on an alumina-forming alloy, furnace equipment


fabricated from these alloys results in a much cleaner environment. In addition,
these alloys can be put into service with a preformed Al 2O3 scale on the surface
and thus eliminate any contamination caused by the transient oxidation of the
alloy. For example, preoxidized alloy 214 baskets and wire mesh belts are used
for processing of semiconductor components. In terms of product cleanliness,
the 214 components perform significantly better than the NiCr alloy previously
used (11).

III. SULFIDATION
One of the most common high-temperature corrosion modes responsible for plant
component failures is sulfidation, with two conditions typically responsible for
this type of attack. The first involves attack by a gaseous environment, which
can be either reducing with H 2 S or oxidizing with SO2. The second involves salt
deposits in oxidizing atmospheres with low concentrations of SO2 (less than 1.0%
SO2 ). The first condition will be discussed in this section, and the second type
of attack will be discussed in the section on ash- and salt-deposit corrosion.
One method of approaching the sulfidation problems associated with the
attack by gaseous environments is to segregate the environments into three different types: (1) H 2 H 2S mixtures or sulfur vapor with oxygen activities below the
thermodynamic stability region of Cr 2O3 , thus the sulfides are the stable phases,
(2) reducing, mixed-gas environments containing H2, H2O, CO, CO2 , H2S, and
so forth, and oxygen activities high enough to form Cr 2O3 , and (3) SO2-bearing
environments.
A. Sulfur Vapor and H2 H2S Mixtures
A review of the sulfidation of metals and alloys in sulfur-vapor and H 2 H 2 S
environments has been conducted by Mrowec and Przybylski (12), Mrowec (13),
and Young (14). Most studies have been conducted in sulfur-vapor environments
with sulfur pressures greater than 103 atm and in H 2 H2 S environments with
sulfur partial pressures less than 102 atm. These sulfur potentials and the very
low oxygen activities in the system resulted in the formation of sulfides. Among
the FeCr, NiCr, and CoCr alloy systems studied, a significant difference
in performance was not noted; however, increasing the chromium content within
a particular alloy system generally improved its sulfidation resistance.
One application where the H 2 H 2S mixture is observed is in the gas stream
of hydrotreating units for petroleum refining. Severe corrosion attack of the processing equipment has been reported (1517). The sulfidation behavior of various
alloy systems in H 2 H 2S mixtures has been described by iso-corrosion rate curves

86

Harper and Lai

as a function of H 2 S concentration and temperature, with data on chromium


steels, FeCr alloys, and austenitic stainless steels reported by Backensto and
Sjoberg (18). Sorrell (16) summarized an extensive set of corrosion data for the
H 2 H 2S mixtures typically found in catalytic re-forming units. These data were
generated by laboratory, pilot-plant, and field tests and include inspections of
commercial operating equipment. Austenitic stainless steels were found to be
most resistant followed by straight Cr stainless steels (1216 wt% Cr). Lowchromium steels (09 wt% Cr) were reported to perform poorly.
B. Reducing, Mixed-Gas Environments
Reducing, mixed-gas environments typically contain H 2 , H 2O, CO, CO2, H 2S,
and other gaseous components (e.g., N 2 ) and usually have oxygen and sulfur
activities sufficient to form oxides and sulfides on most high-temperature alloys.
Thus, the corrosion reaction in these environments usually involves oxidation
and sulfidation. In most cases, an alloy will remain relatively protected during
an oxidation period where a protective Cr 2O3 scale prevents attack of the alloy
by sulfur in the environment. However, at some point, failure of the Cr 2O3 scale,
coupled with the alloys inability to quickly reform this scale, results in breakaway
corrosion, which is followed by rapid sulfidation attack.
Regarding sulfidation problems in coal gasification and similar environments, a large engineering database on commercial alloys was generated during
the 1970s and 1980s under several Metal Properties Councils (MPCs) programs.
These programs evaluated over 80 commercial alloys and coatings, with results
documented in MPC annual reports (19) and a summary report published by
Howes (20). Overall, high-nickel alloys (e.g., Alloy 600) were very susceptible
to sulfidation attack. The NiNi3S 2 eutectic melts at 635C, and molten sulfide
slags can easily destroy the chromium oxide scale and cause catastrophic sulfidation attack. The most important alloying element for improving the sulfidation
resistance of iron-, nickel-, and cobalt-based alloys was identified to be chromium. Also, Nagarajan et al. (21) and Norton et al. (22) have studied the effect
of silicon on the sulfidation resistance of various FeCr alloys in simulated coal
gasification atmospheres. Nagarajan et al. found that an Fe18 Cr2 Si alloy
performed significantly better than an Fe18 Cr0.5 Si alloy in a 24% H 2 39
H 2O18 CO12 CO2 5 CH4 1 H 2S atmosphere at 980C. In a reducingsulfidizing atmosphere containing 0.8 vol% H 2S at 450C, Norton found a sharp decrease in the corrosion kinetics of an Fe12 Cr alloy as the Si content of the
alloy was increased from 0.5 to 4.0 wt%.
One alloy that has shown excellent sulfidation resistance, both in laboratory
and industrial testing, is the HR-160 alloy. This alloy contains high chromium
and silicon, along with cobalt (29 Co28 Cr2.75 Si), and has been found to
perform significantly better than stainless steels and FeNiCr alloys (e.g., Alloy

Ni-Based Alloys and High-Temperature Corrosion

87

Table 4 Metal Loss and Average Metal Affected for


Alloys Tested in a H 2 25 Vol% CH 4 14.8 N 2 4 CO0.6
CO2 0.6 H 2S Atmosphere at 899C for 500 h
Alloy

Metal loss
(mm)

Average metal
affected (mm)

HR-160
HR-120
556
RA85H
253MA
800H
310

0.14
0.59
0.48
0.14
0.44
0.47
Consumed

0.71
1.29
1.30
3.18
3.18
3.18
3.18

800H) (23,24). One example of this behavior is shown in Table 4. These data
were generated from a 500-h exposure to a H 2 25 vol% CH 4 14.8 N 2 4 CO
0.6 CO2 0.6 H 2S atmosphere at 899C (1650F) (25). In iron-based alloys, aluminum has been shown to be beneficial (26) and Santorelli et al. (27) have reported
the sulfidation behavior of two advanced iron-based alumina formers: MA 956
and U.K. Atomic Energys Fecralloy alloy. However, nickel-based alumina formers showed poor sulfidation resistance (28).
C. SO2-Bearing Environments
Most sulfidation studies of SO2-bearing environments have been conducted using
pure SO2 or SO2 O2 mixtures containing high concentrations of SO2. Primarily,
pure metals (e.g., Fe, Ni, and Cr) and binary alloys (e.g., NiCr alloys) have
been studied. An extensive study of NiCr alloys with various amounts of Cr
was conducted by Vasantasree and Hocking (29), with increased amounts of chromium resulting in decreased rates of sulfidation attack. Reviews of the corrosion
behavior of metals and alloys have been published by Kofstad (30). However,
very few data have been published for commercial alloys.
Sulfur furnaces used for the manufacturing of sulfuric acid are the most
typical application where environments containing high levels of SO2 are experienced. In this process, sulfur is used as a feedstock for combustion with excess
air at approximately 1150C. The product gas typically contains about 1015%
SO2, (plus 510% O2, balance N 2 ), which is then converted to SO3 for sulfuric
acid production. One study that examined this type of environment looked at the
behavior of a Type 304 stainless steel and the 556 alloy in an oxidizing environment with and without SO2 present (31). Table 5 summarizes the results of this
work and suggests that oxidation resistance may be at least one criterion to be
used in selecting an alloy for use in an SO2-bearing environment.

88

Harper and Lai

Table 5 Metal Loss and Maximum Depth of Attack for Type 304 Stainless Steel and
the 556 Alloy Exposed at 982C for 550 h in an Oxidizing Environment With and
Without SO2
Type 304 SS

556 alloy

Test gas

Metal
loss
(mm)

Maximum
depth of attack
(mm)

Metal
loss
(mm)

Maximum
depth of attack
(mm)

Ar5 O2 5 CO2
Ar5 O2 5 CO2 10 SO2

0.31
0.61

0.44
0.61

0.005
0.06

0.056
0.10

The other type of environment where SO2 is usually present, albeit in much
lower quantities than discussed earlier, is in the combustion of sulfur-bearing
fuels such as oil and coal. The combustion flue gas stream produced in a coalor oil-fired boiler typically contains less than 1% SO2. Again, little work has been
published for corrosion occurring in these relatively low level SO2-containing
environments. One study by Viswanathan and Spengler (32) found that a Ni15
Cr alloy suffered more attack in a 0.2% SO2 bal N 2 atmosphere than pure SO2
at 870C, with the addition of oxygen significantly reducing the corrosion rate.
Studies on commercial alloys in environments containing low levels of SO2, particularly those with no excess oxygen, are needed, as this type of atmosphere is
relevant to the localized reducing zones that are frequently developed in some
industrial boilers.

IV. CARBURIZATION AND METAL DUSTING


Carburization attack typically occurs when alloys are exposed to an environment
containing CO, CH 4 , or other hydrocarbon gases at elevated temperatures. This
type of attack results in the formation of internal carbides, causing an embrittlement of the alloy and an overall degradation of its original mechanical properties.
Carburization of commercial alloys occurs in many industrial environments; however, most reported problems are experienced in (1) the pyrolysis
furnace tubes used in ethylene production and (2) heat-treating equipment such
as furnace retorts, baskets, fans, and other components used for the case hardening of steels by gas carburizing. Calciners and furnace components used in the
production of activated carbon and carbon fibers are also susceptible to carburization.
One of the most effective alloying elements for improving carburization
resistance to FeNiCr alloys is silicon (3336), with increased chromium con-

Ni-Based Alloys and High-Temperature Corrosion

89

tents generally being beneficial also. Steel and Engel (37) studied FeNiCr
alloys with chromium contents between 15 and 35 wt%, and found that chromium
has a definite beneficial effect on alloys containing 225 wt% nickel. Increased
chromium levels were less effective for alloys containing 2645 wt% nickel, and
for alloys containing 4670 wt% nickel, the chromium additions were slightly
detrimental. Nickel is known to reduce the diffusivity of carbon in Fe15 Cr
Ni alloys (38) and increased nickel contents in FeNiCr alloys exhibit improved
carburization resistance (37). Figure 7 shows this effect of nickel on the carburization resistance of FeNiCr alloys. Also, Grabke et al. (39) showed that FeNi
Cr alloys that had a nickel to iron ratio of 4: 1 exhibited maximum carburization
resistance, this being in general agreement with the minimum value for the product of the carbon solubility and diffusivity (40). One other beneficial alloying
element is aluminum, because alumina formers (e.g., the 214 and MA 956 alloys)
have been found to be more carburization resistant than chromia formers (41,42).
An example of this behavior is shown in Table 6, where the mass of carbon
absorbed by the 214 alloy is compared to other chromia-forming alloys when
exposed to a carburizing gas at 1093C for 24 h (43). As shown in this table,
the 214 alloy absorbed approximately one-third the amount of carbon as the best

Fig. 7 Effect of nickel on the carburization resistance of FeNiCr alloys. (From Ref.
37.)

90

Harper and Lai

Table 6 Carbon Absorption


for Various Alloys Exposed to
an Ar5 H 2 5 CO5 CH 4 Gas
at 1093C for 24 h
Alloy

Carbon absorption
(mg/cm2 )

214
600
625
230
X
S
304
617
316
800H
330

3.4
9.9
9.9
10.3
10.6
10.6
10.6
11.5
12.0
12.6
12.7

performing chromia-forming alloy. The excellent carburization resistance of this


alloy has been attributed to the Al2O3 scale formed on the surface of the alloy,
which has been confirmed by Auger analysis (41). Similar results have been
reported (42) for the MA956 alloy (see Table 7).
At temperatures below 815C, carburization is usually not a problem for
industrial equipment because of relatively slow kinetics. However, a problem
know as metal dusting can occur in strongly carburizing atmospheres (i.e.,
activity of carbon 1) at temperatures generally between 480C and 815C.
This type of attack exhibits a catastrophic deterioration of metallic materials, with

Table 7 Weight Gain for


Various Alloys Exposed to a
H 2 2 CH 4 Gas at 1000C for
100 h
Alloy
MA-956
601
800
310

Weight gain
(mg/cm2 )
0.3
10.0
19.0
36.0

Ni-Based Alloys and High-Temperature Corrosion

91

Fe-, Ni-, or Co-based alloys decomposing into a dust of metal particles, carbon, and sometimes oxides and carbides. Depending on the alloy, maximum attack is observed at approximately 650C, with alloys typically showing signs of
pitting.
Metal dusting has been reported for the following: a 1 Cr0.5 Mo steel in
the waste-heat boiler of a partial oxidation syngas production unit at a chemical
plant (44); stainless steels in the waste-heat boiler of a synthesis gas reactor (45)
and a plant producing gasoline from coal (46); Alloy 800 in the preheater of a
gasifier in a coal gasification plant (47) and in a bypass line in a hydrogen reformer (48); Alloy 600 in the re-former of a natural gassynthetic fuel conversion
plant (49); and for various Fe- and Ni-based alloys in carburizing furnaces (50).
The environments are typically enriched in H 2 and CO.
The fundamentals of metal dusting related to Fe-based alloys have been
explained by Grabke et al. (5153), and a fracture mechanism for metal disintegration during metal dusting has been proposed by Katsman et al. (54). Regarding
Ni-based alloys, Grabke et al. (55) have recently published a study on the metaldusting behavior of several commercial alloys and concluded the following: (1)
any metallic material is susceptible to metal dusting if carbon is possible at an
activity of carbon greater than unity, (2) the mechanism that applies for iron and
low-alloy steels does not apply for nickel and Ni-based alloys, and (3) a protective
oxide scale or surface poison is required for protection. Also, Klower et al. (56)
have found that Ni-based alloys with chromium contents of at least 25 wt%
showed no significant evidence of metal dusting for exposures up to 10,000 h,
thus supporting work that has shown that a protective chromia scale retards a
metal-dusting attack.
Compared to the available data on the carburization resistance of commercial alloys, very little data has been published on their metal-dusting behavior.
Clearly, more studies are needed not only to develop a better understanding of
the mechanism of a metal-dusting attack on Ni-based alloys but also to obtain
engineering data regarding the relative resistance of various alloys in various
metal-dusting-prone environments.

V.

NITRIDATION

All alloys are susceptible, to some degree, to nitridation attack in ammonia-bearing atmospheres at elevated temperatures and this type of environment is common
in the chemical processing industries when ammonia, nitric acid, melamine, and
nylon 6-6 (57,58) are produced. Also, ammonia is widely used in the heat-treating
industry as a nitriding gas for the case hardening of steels.
Similar to carburization, nitridation usually results in the formation of internal nitrides, thus causing an alloy component to become embrittled. When the

92

Harper and Lai

Table 8 Nitrogen Absorption and Depth of Nitride Penetration of


Various Alloys Exposed to Ammoniaa at 649C for 168 h

Alloy

Alloy
base

Nitrogen
absorption
(mg/cm2 )

C-276
230
HR-160
600
625
RA333
601
S
617
214
X
825
800H
556
316
310
304

Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Iron
Iron
Iron
Iron
Iron

0.7
0.7
0.8
0.8
0.9
1.0
1.1
1.3
1.3
1.5
1.7
2.5
4.3
4.9
6.9
7.4
9.8

Depth of
nitride
penetration
(mm)
0.02
0.03
0.01
0.03
0.01
0.03
0.03
0.03
0.03
0.04
0.04
0.06
0.10
0.09
0.19
0.15
0.21

100% NH 3 in the inlet gas and approximately 30% NH 3 in the exhaust gas.

exposure temperatures are relatively low (e.g., 650C), a surface nitride layer
typically forms (59), with the kinetics of the nitridation process depending on
the usual system parameters (i.e., temperature, ammonia concentration in the gas
phase, and alloy composition). In general, austenitic stainless steels have been
successfully used for the processing equipment in ammonia-bearing environments (58,6062). However, when the environment is too severe for the stainless steels, nickel-based alloys are known to be more nitridation resistant than
iron-based alloys (61). Barnes and Lai (59) studied the nitridation resistance of
a wide range of iron-, nickel-, and cobalt-based alloys, with Table 810 showing the amount of nitrogen absorption and depth of internal nitride penetration
exhibited by the nickel-based alloys when exposed to ammonia for 168 h at
649C, 982C, and 1093C, respectively. Based on this work, Barnes and Lai
constructed a plot showing the amount of nitrogen absorption as a function of

Ni-Based Alloys and High-Temperature Corrosion

93

Table 9 Nitrogen Absorption and Depth of Nitride Penetration of Various


Alloys Exposed to Ammoniaa at 982C for 168 h

Alloy

Alloy
base

Nitrogen
absorption
(mg/cm2 )

214
600
S
601
230
617
HR-160
625
X
RA333
800H
825
316
556
304
310

Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Iron
Nickel
Iron
Iron
Iron
Iron

0.3
0.9
0.9
1.2
1.4
1.5
1.7
2.5
3.2
3.7
4.0
4.3
6.0
6.7
7.3
7.7

Depth of
nitride
penetration
(mm)
0.04
0.12
0.18
0.17
0.12
0.38
0.18
0.17
0.19
0.42
0.28
0.58
0.52
0.37
0.58
0.38

100% NH 3 in the inlet gas and less than 5% NH 3 in the exhaust gas.

nickel plus cobalt in the alloys tested at 982C (see Fig. 8). Clearly, the beneficial
effect of nickel and cobalt on the nitridation resistance of an alloy is evident from
this graph.
Nitrogen-containing atmospheres can also be nitriding. For example, when
exposed to pure nitrogen at 1093C for 900 h, Alloy 600 and Alloy 800 showed
1.85 mm and 3.81 mm of attack, respectively (63). Swaminathan and Lukezich
(64) observed internal nitridation of nickel-based alloys that were exposed to
high-velocity combustion gases generated in a gas turbine. Also, Lai (65) observed internal nitridation in four nickel-based alloys (230, 617, 263, and X) that
were tested in a laboratory dynamic oxidation burner rig at 982C for 1000 h.
Figure 9 shows an optical micrograph of the cross section of each alloy, with
the nitrides highlighted. Certainly, more studies are needed, given that data for
the nitridation attack of commercial alloys in this type of environment are rather
limited and the fact that nitrogen-containing atmospheres are being used more
and more in the production of sintered powder-metallurgy products and other
heat-treating operations.

94

Harper and Lai

Table 10 Nitrogen Absorption and Depth of Nitride Penetration of Various


Alloys Exposed to Ammoniaa at 1093C for 168 h

Alloy

Alloy
base

Nitrogen
absorption
(mg/cm2 )

600
214
S
230
617
HR-160
601
625
316
304
X
556
825
RA333
800H
310

Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Nickel
Iron
Iron
Nickel
Iron
Nickel
Nickel
Iron
Iron

0.2
0.2
1.0
1.5
1.9
2.5
2.6
3.3
3.3
3.5
3.8
4.2
5.2
5.2
5.5
9.5

Depth of
nitride
penetration
(mm)
0.00
0.02
0.34
0.39
0.56
0.46
0.58
0.56
0.91
0.58
0.58
0.51
0.58
0.71
0.76
0.79

100% NH 3 in the inlet gas and less than 5% NH 3 in the exhaust gas.

Fig. 8 Effect of Ni Co content on nitridation resistance of Fe-, Ni-, and Co-based


alloys. Alloys were exposed to ammonia gas at 982C for 168 h (100% NH 3 in inlet gas
and 5% NH 3 in exhaust gas). (From Ref. 59.)

Ni-Based Alloys and High-Temperature Corrosion

95

Fig. 9 Optical micrographs showing internal oxidation and internal nitridation of four
alloys tested in dynamic oxidation burner rig at 980C for 1000 h with a 30-min thermal
cycling. (a) Alloy 230, blocky nitrides (arrows), fine carbide precipitates were due to aging
during testing; (b) Alloy 617, blocky nitrides (arrows), large needle nitrides and tiny needle
nitrides; (c) Alloy 263, tiny needle nitrides; (d) Alloy X, blocky nitrides. (From Ref. 65.)

VI. HALOGEN CORROSION


In contrast to the behavior of alloys in an environment containing oxygen at high
temperatures, where an oxide scale forms and protects the material, the exposure
of metallic materials to halogen gases results in the formation of volatile corrosion
products consisting of metal halides. In a gaseous environment with both oxygen
and a halogen, corrosion of the alloy will involve a combination of an oxide scale
and the volatile halides, with the volatile halides causing a significant increase
in the spalling and overall degradation of the oxide scale.
With respect to industrial processes, the chlorination process has been used
to produce titanium, zirconium, tantalum, niobium, and tungsten (6668), as well
as to extract nickel from iron laterites (69) and detinning (70). The manufacture
of TiO2 , SiO2 , and ethylene dichloride (EDC) can also involve chlorine. Also,
chlorine-containing environments are typically generated during calcining operations used to produce lanthanum, cerium, and neodymium. Thus, for these operations, the reactor vessels, calciners, and other process equipment require alloys
that are resistant to high-temperature chlorination attack. It is also important to
note that the use of fuels and/or feedstocks that are contaminated with impurities
such as chlorine, sodium, potassium, and zinc can result in the reaction of the

96

Harper and Lai

Table 11 Melting Point, Boiling Point, and Temperature at Which Certain


Metal Chlorides Reach 104 atm Vapor Pressure

Chloride

Melting
point (C)

Boiling
point (C)

104 atm
temperature
(C)

FeCl 2
FeCl 3
NiCl 2
CoCl 2
CrCl 2
CrCl 3
CrO2Cl 2
MoCl 5
WCl 5
WCl 6
TiCl 2
TiCl 3
TiCl 4
AlCl 3

676
303
1030
740
820
1150
95
194
240
280
1025
730
23
193

1026
319
987
1025
1300
945
117
268

337

750
137

536
167
607
587
741
611

58
72
11
921
454
38
76

chlorine with the other impurities to form chloride salts. The corrosion that occurs
under these will be discussed in Section VII.
Returning to the subject of volatile halides, the melting and boiling points
of some relevant metal chlorides, as well as the temperatures at which the vapor
pressures of these chlorides reach 104 atm, are shown in Table 11 (71,72). These
data illustrate the relatively high volatility and low melting points of the metal
chlorides.
Nickel and nickel-based alloys are widely used in chlorine-bearing environments. Kane (73) studied the behavior of several commercial iron- and nickelbased alloys in an Ar30% Cl2 atmosphere at temperatures between 400C and
704C and found that higher nickel contents resulted in lower chloridation attack.
Table 12 shows the amount of weight loss experienced by various alloys when
exposed to the above-mentioned atmosphere and temperatures.
The above-discussed data were the result of testing in a chlorine-bearing
environment with no measurable oxygen present. However, in many industrial
environments, Cl 2 and O2 are usually present, and under these conditions, the
formation of both a condensed oxide and the volatile metal halides may occur.
The corrosion behavior of various commercial alloys, especially iron and nickel
based, in O2 Cl 2 environments has been studied by several investigators (74
78). The most notable results of this work has been the observation that alloy

Ni-Based Alloys and High-Temperature Corrosion

97

Table 12 Descaled Weight Loss of Several Alloys After Exposure to an Ar30 Cl 2


Atmosphere at 400, 500, 600, and 704C for 500 h
Descaled weight loss (mg/cm2 )
Alloy

400C

500C

600C

704C

Ni-201
600
601
625
617
800
310
304
347

0.2
0.02
0.3
0.7
0.6
6
28
108
215

0.3
5
3
7
7
13
370
1100
Consumed

47101
127180
85200

200270

97
160
215
180
190
890
820
1000
Consumed

Fig. 10 Comparison of the corrosion behavior for the 214, S, and 800H alloys when
exposed to an Ar20 O2 0.25 Cl 2 atmosphere for 400 h at temperatures between 700C
and 1000C. (From Refs. 76 and 77.)

98

Table 13 Metal Loss and Total Depth of Attack for Several Alloys After Exposure to an Ar20 O2 0.25 Cl 2 Atmosphere at 700,
800, 850, 900, and 1000C for 400 h
700C

800C

850C

1000C

900C

Alloy

Metal
loss
(mm)

Total
depth
(mm)

Metal
loss
(mm)

Total
depth
(mm)

Metal
loss
(mm)

Total
depth
(mm)

Metal
loss
(mm)

Total
depth
(mm)

Metal
loss
(mm)

Total
depth
(mm)

214
600
800H
310
S
C-276

0.010

0.025

0.079
0.033

0.010

0.033

0.081
0.046

0.018
0.020
0.023
0.036
0.145
0.066

0.061
0.086
0.046
0.053
0.150
0.071

0.018
0.038
0.031
0.031
0.224
0.163

0.066
0.132
0.097
0.061
0.257
0.175

0.023
0.127
0.043
0.086
0.315
0.300

0.150
0.252
0.191
0.152
0.353
0.320

0.013
0.330
0.203
0.191
0.419
0.419

0.051
0.386
0.424
0.246
0.472
0.450

Harper and Lai

Ni-Based Alloys and High-Temperature Corrosion

99

additions of molybdenum and tungsten, particularly at high concentrations, are


detrimental to chlorination resistance. One investigator (75) has attributed this
behavior to the formation of molybdenum and tungsten oxychlorides, both of
which have very high vapor pressures. The addition of aluminum has been shown
to be beneficial to chlorination resistance, especially at temperatures high enough
to ensure that the formation and growth of Al 2O3 prevails on alumina-forming
alloys. An example of this behavior is shown in Fig. 10, where the total depth
of attack as a function of temperature is plotted for three different alloys exposed
to an Ar20% O2 0.25% Cl 2 atmosphere between 700C and 1000C (77,78).
The two chromia-forming alloys (800H and Hastelloy S) experienced an increase
in attack with increasing temperature; however, the 214 alumina-forming alloy
experienced maximum attack at 900C, with a sudden decrease in corrosion when
the alloy was exposed at 1000C. This has been attributed to the formation and
growth of a protective Al 2O3 scale at the higher temperature, whereas at the lower
temperatures, the lower growth kinetics of Al 2O3 prevented a completely protective alumina scale from being established on the surface of the alloy. A summary
of the results for all the alloys tested in this work is shown in Table 13.
The corrosion of several commercial alloys exposed to gaseous HCl has
been studied by several investigators (74,79,80). In general, nickel and nickelbased alloys were shown to be more resistant to chlorination attack than ironbased alloys. In contrast to the corrosion studies involving O2 Cl 2 atmospheres,
molybdenum was found to improve an alloys resistance to chloridation in reducing environments containing HCl. Hossain et al. (80) found that NiCrMo
alloys (e.g., 625 and Hastelloy C-4) performed the best among various nickelbased alloys, including nickel.

VII. ASH- AND SALT-DEPOSIT CORROSION


In many industrial environments, deposits gather on component surfaces, with a
subsequent accelerated corrosion attack observed. In most cases, the deposit contains some type of salt, which can cause a damaging chemical reaction between
the salt and protective oxide scale. Also, the breakdown of the oxide scale and
attack of the alloy can be quite severe when the salt deposit is liquid. The actual
formation of the salts usually occurs in the vapor phase, when sulfur and/or chlorine react with other impurities (e.g., K, Na, V, Zn, Pb, etc.) in the fuel or feedstock during combustion. These salt vapors then deposit on cooler components
and, in many cases, contain ash from incombustible mineral matter in the fuels.
In gas turbines, a type of high-temperature attack known as hot corrosion
is both well known and well understood. Sulfur from the fuel reacts with sodium
chloride from ingested air during combustion to form sodium sulfate. The sodium
sulfate subsequently deposits on hot-section turbine components (e.g., nozzle

100

Harper and Lai

guide vanes and rotor blades), causing a fluxing/breakdown of the protective


oxide scale that is present on the alloy. The topic of hot corrosion has been
extensively reported (8185) and the mechanism of salt fluxing has been discussed in detail by Goebel et al. (86) and Rapp (87,88).
Burner-rig testing is typically used to simulate gas turbine conditions and
thus has been used to generate hot corrosion data for several commercial nickeland cobalt-based alloys (8992). The most important alloying element has been
found to be chromium, with alloys containing less than 15% generally showing
poor performance in a hot-corrosion-type environment. The effects of other alloying elements was found to be less well defined.
In coal-fired boilers, salt deposits may contain sulfur, sodium, potassium,
and chlorine, along with the fly ash from the incombustible mineral matter in
the coal. The corrosion of superheater and reheater tubes has been very well
characterized, with the corrosion rate of austenitic stainless steels exhibiting a
bell-shaped curve with respect to temperature (93,94); that is, the rate of attack
increases with temperature to a maximum and then decreases with increasing
temperature. The formation of molten alkali metal iron trisulfate [(Na,K) 3Fe
(SO4 ) 3 ] has been related to the accelerated and then decelerated corrosion shown
by the bell-shaped curve. The testing of coextruded tubes using a high-chromium
alloy outer layer [310, Inconel 671, and CR35A [45 Ni35CrFe)] has shown
that these alloys are much more resistant than the austenitic stainless steels such
as 304H, 316H, 312H, and 347H (9597). Results from laboratory testing where
various alloys were covered with a synthetic ash (37.5 mol% Na 2SO4 , 37.5
K 2SO4 , 25 Fe 2O3 ) and exposed to a synthetic flue gas (80% N2 15 CO2 4 O2
1 SO2 ) for 50 h at different temperatures are shown in Fig. 11 (97).
Another severe deposit-type corrosion is found in oil-fired boilers that burn
low-grade fuels (e.g., Bunker C) containing high levels of vanadium, sulfur, and
sodium. Typically referred to oilash corrosion, vanadium pentoxide and sodium
sulfate are the principal phases responsible for this type of corrosion, with reactions between these two constituents resulting in the formation of low-meltingpoint vanadates (98). For uncooled components in the boiler (e.g., hangers and
tube supports), alloys containing high levels of chromium, such as 50 Ni50 Cr,
generally perform much better than stainless steels (99101). In the superheater
sections, coextruded tubes containing an outer layer of high-chromium alloys
such as 446, 50 Ni50 Cr, and CR35A also perform better than austenitic stainless steels (102).
Fireside corrosion in waste incinerators also involves ash- and salt-deposit
corrosion. Corrosive impurities of Cl, S, Na, K, Zn, Pb, and P are usually present
in municipal and/or industrial waste. As a result of these impurities in the feedstock, many complex salts, particularly chlorides and sulfates, are formed, thus
causing sulfidation and/or chloridation attack. Sulfidation attack of several
nickel-based alloys, including 825, 600, 601, 800H, X, and 690, has been ob-

Ni-Based Alloys and High-Temperature Corrosion

101

Fig. 11 Laboratory test results for various alloys coated with a synthetic ash (37.5 mol%
Na 2SO4 , 37.5 K 2SO4 , 25 Fe 2O3 ) and exposed to a synthetic flue gas (80% N 2 15 CO2
4 O2 1 SO2 ) for 50 h at different temperatures. (From Ref. 97.)

served in municipal waste incinerators (103,104). Also, Whitlow et al. (105)


found that chloride-accelerated corrosion attack was responsible for a severe attack of alloys 188, 316 stainless steel, and 825 in a municipal waste incinerator.
The HR-160 alloy was found to perform significantly better than many other
commercial alloys (106) at temperatures above 650C (see Table 14); however,
for superheater, reheater, and furnace-wall tube applications where the metal temperatures are less than 650C, recent work has shown that NiCrMo alloys
perform the best (107). The corrosion reaction at these lower temperatures is still
not completely understood. Krause et al. (108) and Krause (109) looked at the
corrosion exhibited by various alloys exposed to superheater and furnace-wall
conditions. The results showed that molten-salt-deposit corrosion may be a likely
mechanism.
Certain calcining operations experience ash- and salt-deposit corrosion.

102

Harper and Lai

Table 14 Field Test Results from a Municipal Waste


Incinerator Where Uncooled Specimens Were Exposed at
700760C for 2 months
Alloy
HR-160
556
625
214
825
304

Metal loss
(mm)

Maximum metal
affected (mm)

0.00
0.24
0.42
0.50
0.55
0.72

0.05
0.28
0.44
0.55
0.62
0.75

Fig. 12 Results of a field rack test in a NaClKClBaCl 2 salt bath at 840C for 1 month.
(From Ref. 111.)

Chemical feedstocks are commonly contaminated with S, Cl, K, and Na impurities, and salts are formed during the high-temperature calcining process. Recuperators used in industrial furnaces can also suffer sulfidation and/or chloride attack.
For example, severe corrosion attack can be experienced by stainless-steel recuperator tubes used in aluminum melting operations. The flue gas stream that exits
from an aluminum remelting furnace typically contains Cl, S, K, Na, and other
impurities as a result of the flux used in the aluminum melting process. These
impurities lead to sulfide and/or chloride salt deposits on the recuperator tubes

Ni-Based Alloys and High-Temperature Corrosion

103

Table 15 Results of
Laboratory Tests in a NaCl Salt
Batha at 840C for 100 hours

Alloy
188
556
601
214
304
316
X
310
800H
625
RA330
617
230
S
600

Total depth
of attack b
(mm)
0.050
0.060
0.060
0.070
0.080
0.080
0.090
0.100
0.100
0.110
0.110
0.120
0.140
0.160
0.190

Fresh salt bath used for each test run;


air used for cover gas.
b
Mainly intergranular attack; no metal
wastage.

and rapid attack of the alloy. In one case, a Type 310 recuperator suffered approximately 3.9 mm of attack after 15 months of service at about 650C.

VIII. MOLTEN-SALT CORROSION


Separate from salt-deposit corrosion, molten-salt corrosion is related to the corrosion of containment materials that are in contact with a molten salt. Typical examples are molten-salt pots and heat exchangers containing molten salt. The heattreating industry use molten salts extensively for the heat treatment of metals
and alloys, with the furnace equipment and other components in contact with the
molten salt typically suffering corrosion problems. Other applications of molten
salts include heat-transfer and energy-storage media used in solar energy and
nuclear systems, high-temperature batteries, fuel cells, and metallurgical extraction processes.

104

Harper and Lai

Given that molten salts are usually good fluxing agents which remove oxide
scale from a metal surface, the corrosion reaction proceeds by oxidation of the
alloy followed by dissolution of the oxide in the molten salt. For this reason, the
presence of oxygen and water vapor can accelerate the rate of molten-salt corrosion. This type of corrosion can also take place via mass transfer due to a thermal
gradient in the melt. This mode of attack involves the dissolution of an alloying
element at hot spots and deposition of that element at cooler spots, which can
subsequently result in fouling and plugging in a circulating system.
Chloride salts are commonly used in the heat-treating industry for annealing
and normalizing of steels at temperatures between 760C and 980C. Neutral salt
baths, as they are commonly called, typically consist of one or more of the following chlorides: barium, sodium, and potassium. Compositions of five of the more
common neutral salt baths are as follows (110):

50
50
20
25
21

NaCl50 KCl
KCl50 Na2CO3
NaCl25 KCl55 BaCl 2
NaCl75 BaCl 2
NaCl31 BaCl 2 48 CaCl 2

Jackson and LaChance (110) conducted an extensive study on molten-salt corrosion of cast FeNiCr alloys in a 20 NaCl25 KCl55 BaCl 2 salt bath. In this
study and as typically found in molten chloride salts, they found that the alloys
suffered intergranular attack significantly more than metal loss. Also, the results
showed that resistance to the molten salt increased with decreasing chromium
content and increased nickel content and that the intergranular attack generally
followed grain-boundary carbides. Thus, lowering the carbon content of a given
alloy could significantly improve its molten-salt corrosion resistance.
Lai et al. (111) conducted field testing on various iron-, nickel-, and cobaltbased wrought alloys in a NaClKClBaCl 2 salt bath at 840C for 1 month, with
the results shown in Fig. 12. In contrast to the above-discussed results for cast
alloys, two of the high-nickel alloys (Alloy 600 and Alloy 601) suffered more
corrosion attack than the stainless steels 304 and 310. Results from testing in a
NaCl salt bath at 840C are shown in Table 15 (111,112), and similar to the field
test results, CoNiCrW and FeNiCoCr alloys performed the best. At
lower temperatures, corrosion from molten salts typically decreases. Susskind et
al. (113) studied various alloys in a molten NaClKClMgCl 2 salt bath at temperature between 450C and 500C and found many alloys resistant to molten chlorides (see Table 16).
Nitrate or nitratenitrite salt baths are also used for heat-treating purposes,
with typical salt bath temperatures between 160C and 590C. Applications also
exist for use as a medium for heat transfer or energy storage. Slusser et al. (114)
evaluated the molten-salt corrosion of various alloys in a NaNO3 KNO3 salt bath

Ni-Based Alloys and High-Temperature Corrosion

105

Table 16 Corrosion Rates of Alloys in


Molten Eutectic NaClKClMgCl 2 Salt at
450500C for 1000 h
Alloy
1020
2.25 Cr1 Mo
304
310
316
347
410
430
446
600
N
Molybdenum
Tantalum

Maximum penetration
(mm/year)
0
0.08
0.01
0
0.01
0.12
0.03
0.05
0.01
0.05
0.05
0
0.07

at 675C for 336 h. In general, nickel-based alloys performed better than ironbased alloys. However, pure nickel exhibited a rapid rate of corrosion attack.
The corrosion rates of the various alloys plotted as a function of the nickel content
are shown in Fig. 13 (114). Longer-term testing (1920 h) showed corrosion rates
similar to the 336-h tests, except for Alloy 800, as shown by a comparison of
Fig. 13 and Table 17 (114). Also shown in Table 17 is the result of testing at
700C, where corrosion rates became much higher.
Exposure of metals to molten sodium hydroxide (NaOH) results in the formation of metal oxide, sodium oxide, and hydrogen (115). Nickel shows the best
resistance to molten NaOH (116119), particularly low-carbon nickel such as
the Ni 201 alloy (120). The corrosion rates of several nickel-based alloys exposed
to molten NaOH at temperatures between 400C and 680C were reported by
Gregory et al. (119) and are shown in Table 18. Based on these results, molybdenum and silicon were detrimental to the molten NaOH salt corrosion resistance.
Also, molybdenum and iron were found to be selectively removed from nickelbased alloys with less than 90% nickel, leading to the formation of internal voids
(121).
The molten-salt nuclear reactor uses a LiFBeF 2 base salt containing various amounts of UF 4 , ThF 4 , and ZrF 4 , as a fuel salt, and the reactor coolant is a
mixture of NaBF 4 NaF (122). Thus, the corrosion of alloys in molten fluoride
salts has been extensively studied for nuclear applications. The most corrosion-

106

Harper and Lai

Fig. 13 Corrosion rates of various alloys as a function of nickel content in molten


NaNO3 KNO3 salt at 675C. (From Ref. 114.)

Table 17 Corrosion Rates of Selected Alloys at 675


and 700C in a SodiumPotassium NitrateNitrite Salt
Corrosion rate (mm/year)
Alloy
214
600
N
601
800

675C
1920 h

700C
720 h

0.41
0.25
0.23
0.48
1.85

0.53
0.99
1.22
1.25
6.60

Ni-Based Alloys and High-Temperature Corrosion

107

Table 18 Corrosion Rates of Selected Alloys Obtained from Static Tests in Molten
Sodium Hydroxide
Corrosion rate (mm/year)
Alloy

400C

500C

580C

680C

Ni-201
C
D
400
600
301SS

0.023

0.018
0.046
0.028
0.043

0.033
2.540
0.056
0.130
0.060
0.080

0.06
(a)
0.25
0.45
0.13
0.26

0.96

(a)

1.69
1.03

(a) Severe corrosion.

resistant alloy in this environment has proven to be the nickelbased alloy N


(123). Kroger (122) reported a corrosion rate for this alloy of less than 0.0025
mm per year at 704C in the LiFBeF 2 base salt and approximately 0.015 mm
per year at 607C in the NaBF 4 NaF coolant salt. A variety of commercial alloys
were tested on a molten LiF19.5 CaF 2 salt at 797C for 500 h by Misra and
Whittenberger (124). This salt was being considered for a heat-storage medium
in an advanced solar space power system. The tests were conducted in alumina

Table 19 Results of Corrosion Tests in LiF19.5 CaF 2


at 797C for 500 h
Depth of attack (m)
Alloy
Mild Steel
304
310
316
RA330
N
S
X
600
718
188

General a

Grain boundary b

15
90

90
45

155
185
130
165
270
15

140
30
120
105

Note: Tests were conducted in alumina crucibles under argon.


a
Intragranular voids near surface.
b
Intergranular voids.

108

Harper and Lai

crucibles with an argon cover gas and the results are shown in Table 19. Chromium appeared to be detrimental in the nickel-based alloys, but no such effect
was seen in the iron-based alloys.
Molten carbonates are generally less corrosive than molten chlorides or
hydroxides. Coyle et al. (125) evaluated a eutectic sodiumpotassiummagnesium chloride salt (33 NaCl21.5 KCl45.5 MgCl 2 ), a sodium hydroxide salt,
and a eutectic sodiumpotassium carbonate salt (58 Na 2CO3 42 K 2CO3 ) for a
possible heat-transfer and energy-storage medium capable of operating at 900C
for a solar power generation system. Both the NaClKClMgCl 2 and NaOH were
too corrosive for many commercial alloys; however, the Na 2CO3 K 2CO3 showed
promise because of its less aggressive nature. Results of testing in this molten

Table 20 Corrosion Results


of Selected Alloys Tested
in Molten Eutectic Sodium
Potassium Carbonate at 900C
for 504 h

Alloy

Total depth
of attack a
(mm)

X
214
188
556
600 b
600 b
N
304
316
230
Nickel
800 b
800 b
S

0.12
0.19
0.22
0.26
0.34
0.44
0.51
0.54
0.63
0.77
0.30
0.25
0.8
1.43

Note: N2 0.1 CO2 (110 O2 ) used


for cover gas.
a
All alloys showed metal loss, except
for nickel, which suffered 0.2 mm
metal loss and 0.11 mm intergranular attack.
b
Two samples from different suppliers.

Ni-Based Alloys and High-Temperature Corrosion

109

carbonate salt at 900C for 504 h are shown in Table 20. The NiCrMo Alloy
S was severely corroded; the NiCrFeMo Alloy X performed the best. However, no systematic trend between alloying elements and performance was noted.

IX. MOLTEN-METAL CORROSION


Similar to molten-salt corrosion, molten-metal corrosion relates to the corrosion
of a containment material in contact with a molten metal and/or alloy. Liquid
metals are sometimes used as a heat-transfer medium because of their excellent
heat-transfer properties and various metals have been investigated for use as a
coolant in nuclear reactors. Other applications of liquid metals exist in the heattreating industry (e.g., molten lead) and power generation (126).
The molten metal corrosion behavior of a containment material is usually
related to its solubility in the molten metal. Thus, a containment material with
a higher solubility in the molten metal generally exhibits a higher corrosion rate.
Molten aluminum is very aggressive and iron-, nickel-, and cobalt-based alloys
are rapidly attacked by this liquid metal. Molten zinc is less of a problem; however, nickel and nickel-based alloys react readily with molten zinc and are not
recommended for use in this environment. Nickel and nickel-based alloys also
have a relatively high solubility in molten lead, molten lithium, and molten sodium and, thus, are typically not recommended in these applications.

REFERENCES
1. GY Lai. High-Temperature Corrosion of Engineering Alloys. Metals Park, OH:
ASM International, 1990.
2. HE McGarrow, ed. The Making, Shaping and Treating of Steel. Pittsburgh, PA:
United States Steel Corporation, 1971, p. 1136.
3. HE Eiselstein, NE Skinner. ASTM STP No. 165. Philadelphia, PA: ASTM, 1954,
p. 162.
4. MF Rothman. Internal Technical Report No. 12034, Haynes International, 1985.
5. MA Harper, JE Barnes, GY Lai. CORROSION/97. Houston, TX: National Association of Corrosion Engineers, 1997, Paper 132.
6. B Gleeson, MA Harper. Oxid Met, Vol. 49, Nos. 3/4, 1998, p. 373.
7. JL Smialek, GM Meier. High-temperature oxidation. In CT Sims, NS Stoloff, WC
Hagel, eds. Superalloys II. New York: John Wiley & Sons, 1987, pp. 293323.
8. RJ Hendricks, KD Sheffler. Materials for advanced turbine enginesProject 3 design, fabrication and evaluation of an oxide dispersion strengthened sheet alloy
combustor liner. NASA CR-17491, February 1984.
9. JL Gonzalez-Carrasco, V Guttmann, H Fattori. Met Trans A 26A: 915, 1995.
10. GY Lai. Haynes International, unpublished research, 1991.

110

Harper and Lai

11. J Bailey. Volatile Cr contamination reduction in APCVD systems by alloy oxidation engineering. J Electrochem Soc, in press.
12. S Mrowec, K Przybylski. High Temp Mater Proc 6(1,2): 1, 1984.
13. S Mrowec. Oxid Met 44(1/2):177, 1995.
14. DJ Young. Rev High Temp Mater 4(4):299, 1980.
15. G Sorrell, WB Hoyt. Collection and Correlation of High Temperature Hydrogen
Sulfide Corrosion Data. Houston, TX: National Association of Corrosion Engineers, 1956.
16. G Sorrell. Compilation and Correlation of High Temperature Catalytic Reformer
Corrosion Data. Houston, TX: National Association of Corrosion Engineers, 1957.
17. EB Backensto. Corrosion in Catalytic Reforming and Associated Processes. 22nd
Midyear Meeting of APIs Division of Refining, Philadelphia, 1957.
18. EB Backensto, JW Sjoberg. Iso-Corrosion Rate Curves for High temperature Hydrogen-Hydrogen Sulfide. Houston, TX: National Association of Corrosion Engineers, 1958.
19. AO Schaefer. A program to discover materials suitable for service under hostile
conditions obtained in equipment for the gasification of coal and other solid fuels,
Metal Properties Council annual reports, 1976 through 1983.
20. MAH Howes. High temperature corrosion in coal gasification systems. Gas Research Institute Report GRI-8710152, 1987.
21. V Nagarajan, RG Miner, AV Levy. JECS 129(4):782, 1982.
22. JF Norton, M Maier, WT Bakker. Corrosion of candidate heat exchanger alloys
in complex simulated coal gasification atmospheres at 450C. CORROSION/97.
Houston, TX: National Association of Corrosion Engineers, 1997, Paper 144.
23. JF Norton, FG Hodge, GY Lai. High Temperature Materials for Power Engineering.
Netherlands: Kluwer Academic, 1990, p. 167.
24. GY Lai, JF Norton, FG Hodge. The corrosion behavior of a new sulfidation-resistant alloy in a sulfidizing/oxidizing/carburizing atmosphere. Proceedings of the
First International Conference on Heat-Resistant Materials, September 2226,
1991. Metals Park, OH: ASM International, 1991, p. 211.
25. MA Harper, JE Barnes. Haynes International, unpublished research, 1997.
26. RW Bradshaw, RE Scholts, DR Adolphson, Sandia National Laboratories Report
SAND 77-8277, 1977.
27. RL Santorelli, JF Norton, F Bregani. Werkst Korros 41:669, 1990.
28. GY Lai. In: MF Rothman, ed. High Temperature Corrosion in Energy Systems.
Warrendale PA: TMS, 1985, p. 227.
29. V Vasantasree, MG Hocking. Corros Sci. 16:261, 1976.
30. P Kofstad. High Temperature Corrosion. New York: Elsevier Applied Science,
1988.
31. JJ Barnes, GY Lai. CORROSION/90. Houston, TX: National Association of Corrosion Engineers, 1990, Paper 276.
32. R Viswanathan, CJ Spengler. Corrosion 26(1):29, 1970.
33. W Steinkusch. Werkst Korros 30:837, 1979.
34. LH Wolfe. Mater Perform 38, April 1978.
35. RH Kane. CORROSION/83. Houston, TX: National Association of Corrosion Engineers, 1983, Paper 266.

Ni-Based Alloys and High-Temperature Corrosion

111

36. U Van den Bruck, CM Schillmoller. CORROSION/85. Houston, TX: National Association of Corrosion Engineers, 1985, Paper 23.
37. C Steel, W Engel. AFS Int Cast Metals J 28, September 1981.
38. O Demel, E Keil, P Kostecki. SGAW Report 2538. Studiengesellschaft fur Atoenergie, Osterreichische.
39. HJ Grabke, U Gravenhorst, W Steinkusch. Werkst Korros 27:291, 1976.
40. SK Bose, HJ Grabke. Z Metallk 69:8, 1978.
41. GY Lai. In: MF Rothman, ed. High Temperature Corrosion in Energy Systems.
Warrendale, PA: TMS, 1985, p. 551.
42. RH Kane, GM McColvin, TJ Kelly, JM Davison. CORROSION/84. Houston, TX:
National Association of Corrosion Engineers, 1984, Paper 12.
43. GY Lai, CR Patriarca. Metals Handbook, 9th ed. Metals Park, OH: ASM International, 1987, p. 1311, Vol. 13.
44. GM Tanner. Eng Fail Anal 1(4):289, 1994.
45. F Eberle, RD Wylie. Corrosion 15(12):622t, 1959.
46. WB Hoyt, RH Caughey. Corrosion 15(12):627t, 1959.
47. RA Perkins, WC Coons, FJ Radd. Properties of High Temperature Alloys. Pennington, NJ: Electrochemical Society, 1976.
48. RL Codwell, Corrosion in the Petrochemical Industry. Metals Park, OH: ASM International, 1994, p. 231.
49. ML Holland, HJ de Bruyn. Int J Pres Ves Piping 66:125, 1996.
50. GY Lai. J Metals 37(7):14, 1985.
51. HJ Grabke, R Krajak, JC Nava Paz. Corros Sci 35:1141, 1993.
52. HJ Grabke, CB Bracho-Troconis, EM Muller-Lorenz. Werkst Korros 45:215,
1994.
53. E Pippel, HJ Grabke, S Straub, J Woltersdorf. Steel Res 66:217, 1995.
54. A Katsman, L Klinger, LA Levin, T Werber. In: RY Lin, YA Chang, RG Reddy,
and CT Liu, eds. Design Fundamentals of High Temperature Composites, Intermetallics, and Metal-Ceramic Systems. Warrendale, PA: TMS, 1995, p. 413.
55. HJ Grabke, R Krajak, EM Muller-Lorenz, S Straub. Mater Corros 47:495, 1996.
56. J Klower, HJ Grabke, EM Muller-Lorenz, DC Agarwal. CORROSION/97. Houston, TX: National Association of Corrosion Engineers, 1997, Paper 139.
57. GL Swales, In: I Kirman, JB Marriott, M Merz, PR Sahm, and DP Whittle, eds.
Behavior of High Temperature Alloys in Aggressive Environments, Proceedings
of the Petten International Conference. London: Metals Society, 1980, pp. 45
82.
58. K Rorbo. Environmental Degradation of High Temperature Materials. London: Institution of Metallurgists, 1980, Series 3, No. 13, Vol. 2.
59. JJ Barnes, GY Lai. High temperature nitridation of Fe-, Ni-, and Co-base alloys.
In Proceedings of the TMS/AIME Symposium on Corrosion and Particle Erosion
at High Temperatures. Warrendale, PA: TMS, 1989, p. 617.
60. KM Verma, H Ghosh, JS Rai. Br Corros J 13(4):173, 1978.
61. JJ Moran, JR Mihalism, EN Skinner. Corrosion 17(4):191t, 1961.
62. DW McDowell Jr. Mater Protect 1(7):18, 1962.
63. GD Smith, PJ Bucklin. CORROSION/86. Houston, TX: National Association of
Corrosion Engineers, 1986, Paper 375.

112

Harper and Lai

64. VP Swaminathan, SJ Lukezich. Degradation of Transition Duct Alloys in gas Turbines. In: Proceedings of ASM 1993 Materials Congress Materials Week 93. Materials Park, OH: ASM International, 1993, pp. 99111.
65. GY Lai. Nitridation attack in a simulated gas turbine combustion environment.
In: D Coutsouradis et al. eds. Materials for Advanced Power Engineering, Part II.
Netherlands, Kluwer Academic, 1994, pp. 12631272.
66. WJ Kroll. Method of Manufacturing Ti and Alloys Therof, U.S. Patent 2205854,
January 1940.
67. WA Henderson. J Metals 16:155, 1964.
68. SM Shelton. In: B Lustman and F Kerze Jr, eds. The Metallurgy of Zirconium.
New York: McGraw-Hill, 1955, p. 59.
69. I Iwasaki, Y Takahasi, H Kahata. Trans SME AIME 243:308, 1966.
70. CL Mantell. Tin New York: Reinhold, 1949.
71. PL Daniel, RA Rapp. In: MG Fontana, and RW Staehle, eds. Advances in Corrosion
Science and Technology. New York: Plenum Press, 1970, Vol. 5, p. 55.
72. O Kubaschewski, E Evans. Metallurgical Thermochemistry, New York: Pergamon
Press, 1958.
73. RH Kane. In: BJ Moritz, WI Pollock, eds. Process Industries Corrosion. Houston,
TX: National Association of Corrosion Engineers, 1986, p. 45.
74. S Baranow, GY Lai, MF Rothman, JM Oh, MJ McNallan, MH Rhee.
CORROSION/84, Paper 16, Houston, TX: National Association of Corrosion Engineers, 1984.
75. JM Oh, MJ McNallan, GY Lai, MF Rothman. Met Trans A 17A:1087, 1986.
76. MH Rhee, MJ McNallan, MF Rothman. In: MF Rothman, ed. High Temperature
Corrosion in Energy Systems. Warrendale, PA: TMS, 1985, p. 483.
77. MJ McNallan, MH Rhee, S Thongtem, T Hensler. CORROSION/85. Houston, TX:
National Association of Corrosion Engineers, 1985, Paper 11.
78. S Thongtem, MJ McNallan, GY Lai. CORROSION/86. Houston, TX: National
Association of Corrosion Engineers, 1986, Paper 372.
79. MH Brown, WB DeLong, JR Auld. Ind Eng Chem 39(7):839, 1949.
80. MK Hossain, JE Rhoades-Brown, SRJ Saunders, K Ball. Proc. U.K. Corrosion/
83, p. 61.
81. J Stringer. Hot corrosion in gas turbines. Battelle Report MCIC-72-08, 1972.
82. J Stringer, RI Jaffee, TF Kearns, eds. High Temperature Corrosion of Aerospace
Alloys. London: Harford House, 1973.
83. JW Fairbanks, I Machlin, eds. Proceedings of the 1974 Gas Turbine Materials in the
Marine Environment Conference, Columbus, OH: Battelle Columbus Laboratories,
1974.
84. Hot Corrosion Problems Associated with Gas Turbines, Philadelphia, PA: ASTM,
1967.
85. AB Hart, AJB Cutler, eds. Deposition and Corrosion in Gas Turbines, London:
Applied Science Publishers, 1973.
86. JA Goebel, FS Pettitt, GW Goward. Met Trans 4:261, 1973.
87. RA Rapp. Corrosion 42(10):568, 1986.
88. RA Rapp, YS Zhang. J Metals 46(12): 4755, 1994.
89. PA Bergman, AM Beltran, CT Sims. Development of hot corrosion-resistant alloys

Ni-Based Alloys and High-Temperature Corrosion

90.
91.
92.
93.
94.

95.
96.

97.
98.
99.
100.
101.
102.
103.
104.
105.
106.
107.

108.

109.
110.
111.

113

for marine gas turbine service. Summary Report to Marine Engineering Laboratory,
Navy Ship R&D Center, October 1, 1967.
MJ Zetlmeisl, DF Laurence, KJ McCarthy. Mater Perform 41, June 1984.
PA Bergman, CT Sims, AM Beltran. Hot Corrosion Problems Associated with Gas
Turbines. Philadelphia, PA: ASTM, 1967, p. 38.
GY Lai, JE Barnes. A burner rig investigation of the hot corrosion behavior of
several wrought superalloys and intermetallics. ASME Paper 91-GT-21, 1991.
C Cain Jr, W Nelson. J Eng Power Trans. ASME 468, October 1961.
RW Borio, AL Plumley, WR Sylvester. In: RW Bryers, ed. Ash Deposits and Corrosion Due to Impurities in Combustion Gases. New York: Hemisphere Publishing/
McGraw-Hill, 1978, p. 163.
T Flatley, CW Morris, UK Corrosion 83, Birmingham, UK: Institution of Corrosion Science & Technology, 1982, p. 71.
AH Rudd, JM Tanzosh. In: A Armor et al., eds. Proceedings of the 1st International
Conference on Improved Coal-Fired Power Plants, Palo Alto, CA: Electrical Power
Research Institute, 1986, pp. 2145.
J Stringer. Metals Handbook, 9th ed. Corrosion. Materials Park, OH: ASM International, 1987, p. 998, Vol. 13.
WT Reid. In: JM Beer, ed. External Corrosion and DepositsBoilers and Gas
Turbines. New York: American Elsevier, 1971.
DW McDowell Jr, JR Mihalisin. ASME Paper 60-WA-260, 1960.
BF Spafford. UK Corrosion 83. Birmingham, UK: Institution of Corrosion Science & Technology, 1982, p. 67.
GL Swales, DM Ward. CORROSION/79. Houston, TX: National Association of
Corrosion Engineers, 1979, Paper 126.
N Bolt. Proceedings of the 10th International Congress on Metallic Corrosion. New
Delhi: Oxford and IBH Publishing, 1988, p. 3593.
SK Srivastava, GY Lai, DE Fluck. CORROSION/91. Houston, TX: National Association of Corrosion Engineers, 1991, Paper 398.
JA Harris, WG Lipscomb, GD Smith. CORROSION/87. Houston, TX: National
Association of Corrosion Engineers, 1987, Paper 402.
GA Whitlow, PJ Gallagher, SY Lee. CORROSION/89. Houston, TX: National
Association of Corrosion Engineers, 1989, Paper 204.
GY Lai. CORROSION/91. Houston, TX: National Association of Corrosion Engineers, 1991, Paper 249.
JL Blough, GJ Stanko, MT Krawchuk. In Situ Materials Testing in a Waste-toEnergy Power Plant, 2nd International Conference on Corrosion in Advanced
Power Plants, Tampa, FL, 1997.
HH Krause, PW Cover, WE Berry. Proceedings of the 1st Conference on Advanced
Materials for Alternative Fuel Capable for Use in Directly Fired Heat Engines.
Castine, ME: Maine Maritime Academy, 1979.
HH Krause. CORROSION/91. Houston, TX: National Association of Corrosion
Engineers, 1991, Paper 242.
JH Jackson, MH LaChance. Trans ASM 46:157, 1954.
GY Lai, MF Rothman, DE Fluck. CORROSION/85. Houston, TX: National Association of Corrosion Engineers, 1985, Paper 14.

114

Harper and Lai

112. GY Lai. Haynes International, unpublished results, 1986.


113. H Susskind, FB Hill, L Green, S Kalish, L Kukacka, WE McNulty, E Wirsing.
Chem Eng Prog 56(3):57, 1960.
114. JW Slusser, JB Titcomb, MT Heffelfinger, BR Dunbobbin. J Metals 24, July 1985.
115. GP Smith. Corrosion of materials in fused hydroxides. USAEC Report ORNL2048, 1956.
116. CM Craighead, LA Smith, RI Jaffee. Screening tests on metals and alloys in contact
with sodium hydroxide at 1000 and 1500F. USAEC Report BMI-705, 1951.
117. EM Simmons, NE Miller, JH Stang, C Weaver. Corrosion and components studies
on systems containing fused NaOH. USAEC Report BMI-1118, 1956.
118. RA Lad, SL Simon. A study of corrosion and mass transfer of nickel by molten
sodium hydroxide. Corrosion 10(12):435, 1954.
119. JN Gregory, N Hodge, JVG Iredale. The static corrosion of nickel and other materials in molten caustic soda, Report AERE-C/M-272, 1956.
120. RR Miller. Thermal properties of sodium hydroxide and lithium metal. Quarterly
Progress Report May 1Aug. 1, 1952, NRL-3230-201/52, 1952.
121. GP Smith, EE Hoffman. Corrosion 13:627t, 1957.
122. JW Kroger. Corrosion 29(3):115, 1973.
123. JW Kroger. Corrosion 30(4):125, 1974.
124. AK Misra, JD Whittenberger. Fluoride Salts and Container Materials for Thermal
Energy Storage Applications in the Temperature Range 970 to 1400K. NASA Tech.
Memo. 89913. Cleveland, OH: NASA Lewis Research Center, 1987.
125. RT Coyle, TM Thomas, GY Lai. In: MF Rothman, ed. High Temperature Corrosion
in Energy Systems. Warrendale, PA: The Metallurgical Society of AIME, 1985,
p. 627.
126. DL Katz. In: RN Lyon, ed. Liquid-Metals Handbook. NAVEXOS, P-733, Rev.
Washington, DC: US Government Printing Office, 1952, p. 1.

5
Corrosion of Copper and Its Alloys
Andrew James Brock
Metals Research Laboratories, Olin Corporation, New Haven,
Connecticut

I.

INTRODUCTION

Copper and its alloys are some of the earliest metals known to man. The use of
copper was recorded in northern Iraq as early as 8500 b.c. and in Egypt in 7000
b.c. Copper articles found in the Sinai Peninsula have been dated as being made
in 4000 b.c. The ease with which copper oxide ores can be reduced to copper,
together with the ability of the metal to alloy advantageously with other elements,
resulted in copper playing a critical role in the development of civilization.
The present wide use of copper results from a combination of good corrosion resistance in a multitude of environments, together with other desirable properties. These include its high electric and thermal conductivities, its mechanical
properties, and its ease of being formed. Not the least of its uses has been for
decorative purposes where the attractive color of the pure element and of alloys
such as the brasses, together with their ease of accepting a polish, have considerable aesthetic appeal. Copper and its alloys find other decorative uses such as
for the roofs or domes on churches and other institutional buildings. In this case,
slow corrosion of the metal at less than 7.5 mils in every 100 years is accompanied by the formation of characteristic green coatings or verdigris. Perhaps the
classic example of this is the verdigris on the Statue of Liberty. In other natural
environments, such as freshwater and saline waters and soil, copper and copper
alloys have high corrosion resistance. They are extensively used for underground
water lines, for plumbing tubes, and for condenser tubes in the power utility
industry.
Copper is a noble metal in that its electrochemical potential is above that
of hydrogen. Accordingly, it does not discharge hydrogen from nonoxidizing
115

116

Brock

acids. Thus, in oxygen-free sulfuric acid, copper is essentially immune. In the


presence of air, dissolved oxygen promotes dissolution of copper by providing
an alternate cathodic reaction, namely
O2 4H 4e 2H2O
This supports the anodic reaction of
Cu Cu2 2 e
In neutral solutions, such as freshwater or seawater, oxygen also participates in the corrosion reaction. However, the anodic reaction results in the formation of cuprous oxide:
2Cu H2O Cu2O 2H 2e
This oxide film generally forms according to an approximately parabolic
rate law and reaches an environment-specific final thickness which is essentially
constant. The oxide film provides resistance to the migration of ions and electrons, hence providing a degree of resistance to further corrosion. Cuprous oxide
is a p-type semiconductor. The electronic resistance of such oxides can be decreased by doping with higher-valence elements. This is achieved by alloying
copper with elements such as aluminum, nickel, tin, zinc, and iron. The oxide
films which form on these alloys are then doped with these elements. These doped
oxides, together with the presence of corrosion products of the alloying elements,
can significantly further increase the corrosion resistance of the metal.
A broad range of copper alloys are in use, the alloying elements providing
for a required combination of properties. Some of the more common wrought
and cast copper alloys and their UNS numbers are listed in Table 1.

II. TYPES OF ATTACK


Depending on the environment to which the metal is exposed, copper and its
alloys can suffer general corrosion or the various types of localized corrosion
which are suffered by other metals. It is appropriate to characterize these types of
corrosion prior to describing the behavior of the metal in specific environments.
A. General Corrosion
General corrosion is the term which describes the uniform attack of a metal. It
leads to surface roughening on a microscale. Such attack is free of localized
corrosion phenomena, which are described below. High rates of localized attack
of copper and its alloys result from contact with oxidizing acids. This property is
utilized in the cleaning and etching of the metal by immersion in acidhydrogen
peroxide mixtures, nitricsulfuric acid mixtures, and oxidizing acid salts. Expo-

Corrosion of Copper and Its Alloys

117

Table 1 Types and Alloying Constituents in Selected Wrought and Cast Copper
Alloys
Alloy
Wrought
Coppers
High coppers
Brasses
Leaded brasses
Tin brasses
Phosphor bronzes
Aluminum bronzes
Silicon bronzes
Coppernickels
Cast
Coppers
High coppers
Brasses, leaded brasses, and
Manganese bronzes
Manganese and leaded
Silicon brasses and bronzes
Tin bronzes and leaded tin bronzes
Nickeltin bronzes
Aluminum bronzes
Coppernickels
Nickelsilvers

UNS No.

Alloying elements

C10100C15760
C16200C10600
C20500C28580
C31200C38590
C40400C49080
C50100C52400
C60600C64400
C64700C66100
C70000C79900

99.3% Cu
96% Cu
Zn
ZnPb
ZnSnPb
SnP
AlNiFeSiSn
SiSn
NiZn

C80100C81100
C87300C82800
C83300C85800

99% Cu
94% Cu
ZnSnPb

C86100C86800
C87300C87900
C90200C94500
C94700C94900
C90520C95800
C96200C96800
C97300C97800

ZnMnFePb
ZnSnSi
SnZnPb
NiSnZnPb
AlFeNi
NiFe
NiZnPbSn

sure in environments such as certain soils can result in similar etched surfaces
but only after several years.
B. Pitting
In certain environments, the corrosion of copper alloys can result in pitting. In
some cases, pitting occurs over the entire exposed surface of the alloy; in others,
pits are formed at discrete locations. In either case, the formation of pits is undesirable because it can lead to structural weakening of the alloy, or perforation in
the case of tubes or vessels.
Pitting is not the predominant type of attack seen with copper alloys. Environments which typically promote pitting are specific and these will be discussed
in later sections. Typically, these environments include sulfide-polluted seawater,
water under conditions of stagnation, and certain potable waters with very specific
chemistries. The pits often do not continue to grow with time, but attain a certain
depth beyond which further increases in depth do not occur. Alloys most resistant

118

Brock

to pitting include the high-copper alloys and tin bronzes. Those least resistant to
pitting are the low-zinc brasses and aluminum bronzes, whereas the copper
nickels and tin bronzes have intermediate resistance.
C. Crevice Corrosion
This is a form of attack which can occur where there is a crevice on the exposed
alloy surface. Such a crevice can be formed between two copper alloy surfaces
or between the alloy and a nonmetallic surface. The latter includes deposits on
the metal surface. In the case of copper-alloy condenser tubes, such deposits
could be fragments of shells or vegetation or localized deposits of silt. Cleaning
of the condenser tubes is one method of eliminating this form of attack.
Classically, two types of crevice corrosion are known. In one type, the area
inside the crevice is preferentially attacked because it can become depleted in
oxygen with respect to the environment outside. The crevice then serves to isolate
the anode reaction of alloy dissolution inside the crevice from the cathodic reduction of oxygen outside the crevice. Attack is then essentially due to differential
aeration of the metal inside the crevice and that outside it. In the second type of
crevice corrosion, initial corrosion of the alloy within the crevice results in an
increase in the concentration of metal ions in the solution. On the alloy surface
immediately outside the crevice, the metal ions can readily diffuse away or be
swept away by the flow of the environment. The overall results is that the areas
outside of the crevice suffer dissolution and the metal inside the crevice is ennobled because of the higher metal ion concentration. This form of attack is due
to the formation of a metal ion concentration cell. It is the type of crevice corrosion which is more typical of copper alloys. As with pitting, crevice corrosion
is a statistical phenomenon. In waters, it is promoted by high temperatures and
by the flow of the environment outside the crevice. The depth of attack is usually
less than that seen with pitting.
D. Dealloying
This is a form of localized attack which occurs in alloys where the alloying
constituents differ in activity. It has been observed in copper alloys containing
Ni, Al, Mn, or Zn. The attack is due to the selective dissolution of the more
active component, leaving behind a structurally weak deposit of the less active
component. Among copper alloys, dealloying of brasses is most common. It occurs in alloys with more than 15% zinc, the attack being termed dezincification.
In waters, this form of attack is favored by high temperature, water stagnation,
the presence of crevices where aeration is restricted, a high ratio of chloride ion
to bicarbonate ion, and a relatively high pH (1,2).
In single-phase brass alloys, the dealloying occurs over the entire alloy

Corrosion of Copper and Its Alloys

119

surface and is known as layer-type dezincification (3). In two-phase alloys, the


phase is attacked preferentially at a much higher rate than that of the phase
(4). The corroded metal then contains discrete regions where penetration has
occurred, leaving plugs of porous copper. For this reason, such an attack is termed
plug dezincification. Figure 1 shows cross sections through dezincified regions
of leaded copperzinc alloys, C36000, which had been immersed in an acidified
copper chloride solution for 7 days. It illustrates in one case a uniform dezincified
layer. In the other case, dezincification in a cast alloy has penetrated along the
phase which formed at grain boundaries. This represents an extreme case of
plug-type dezincification.
The dealloying of copper alloys has been the subject of extensive investigation. Early work suggested that the mechanism was one of dissolution of both
active and less active components, with the less active component being redeposited on the metal (5,6). Later work suggests that the mechanism was one of selective dissolution of the reactive element (7,8). This mechanism requires diffusion
of the more active component through the alloy. Studies of the dealloying of
coppernickel, coppermanganese, and copperzinc alloys showed that the dealloying kinetics were highest for CuMn alloys and lowest for CuNi alloys, but
they were not consistent with diffusion data (8,9). These results led to the conclusion that dealloying rates are primarily controlled by the difference in the potential of the alloy in solution and the reversible potential of the solute element.
Layer-type dezincification can be completely inhibited by the addition of
low levels of As, P, or Sb (10,11). Bismuth, a similar element present in bismuth
brasses, does not play a role in inhibiting dezincification. The role of As, Sb,
and P led to the development of Inhibited Admiralty alloys C44300, C44400,
and C44500, containing 28% Zn and 1% Sn, with low levels of As, Sb, and P,
respectively. However, these elements do not prevent dealloying of the phase
in high-zinc brasses. Tin decreases the rate of attack on the phase and is found
at the 1% level in both naval brass (C46400) and manganese bronze (C67500)
for this reason.
E.

Stress-Corrosion Cracking

Stress-corrosion cracking is attack which can occur due to a combination of stress


and corrosion where either acting alone would not lead to the development of
cracks. With certain susceptible alloys, stress corrosion can lead to very rapid
failure. The crack path can be intergranular or transgranular and is perpendicular
to the direction of stress. The alloys most susceptible are typically those also
most susceptible to dealloying, such as the high-brass and manganese-containing
alloys. The attack can result in brittle fracture of the metal. In brasses, this type
of failure has also been called season cracking because of the cracking of brass
cartridge cases during hot, rainy seasons in former British colonies.

Fig. 1 Cross sections of leaded brass alloys showing dezincification after exposure to
acidified copper chloride solution for 7 days: (a) shows fairly uniform layer-type attack
and (b) shows attack along phase in a cast alloy. Magnification 100.

Corrosion of Copper and Its Alloys

121

The atmospheres which promote this form of attack are especially those
containing ammonia together with water or water vapor. The presence of oxygen
and carbon dioxide accelerate the rate of failure (12). Other corrodents which
have been claimed to promote stress corrosion in brasses are amines (13), citrate
and tartrate solutions (14), nitrites, carbonates, phosphates, and alkalis (15), sulfur
dioxide (16), and seawater (17).
Testing for susceptibility to stress-corrosion cracking is typically conducted
in Mattssons solution at pH 7.3. This is a solution of ammonium sulfate, copper
sulfate, and water with a copper ion content of 0.05 g ions/L and an ammonium
ion content of 1.0 g ions/L (18). In one form of the test, strip samples, 30 mils
thick and measuring 6 0.5 in. are bent around a 3/4-in. mandrel to provide a
permanent 90 set. The sample is then bent into a U and fixed in a jig so that
the legs are 3/4 in. apart. The stressed sample is then immersed in the Mattssons
solution. Withdrawal of the sample is practiced at intervals. The degree of
springback of the samples is compared against that of a control sample not immersed in the solution. Any decrease in springback over that of the control is
then due to stress corrosion. When the springback falls to below 80%, the sample
is said to have failed. A more rapid test is to expose similar U-bend samples
over 28% ammonia in a closed container.
Table 2 shows the response of various alloys to such testing (19). The
results indicate that many alloys are immune to cracking in the Mattssons solution. Those containing Zn are susceptible, the susceptibility increasing with increasing Zn content. Alloy C66900, containing Mn and Zn, is particularly susceptible. In moist ammonia, other alloys also show susceptibility to stress corrosion.
Of the alloys tested, only alloys C11000, C19400, and C65400 and the copper
nickel alloys are immune. Figure 2 illustrates the typical intergranular paths of
stress-corrosion cracks which formed in U-bends of alloy C26000 after only 4
h exposure over 28% ammonia. The many branches to the main cracks are characteristic of this corrosion phenomenon.
Another test used as an indicator of whether a component will suffer stress
corrosion consists of immersion in a mercurous nitrate solution (20). This test
is strictly only a means of indicating the presence of residual stress in an alloy,
the failure occurring by mercury embrittlement of grain boundaries. As such, it
also reveals the presence of residual stresses in alloys, such as the coppernickels,
which are not susceptible to stress-corrosion cracking. Therefore, this test should
be confined to testing parts fabricated from brass.
F.

Corrosion Fatigue

Corrosion fatigue, like stress-corrosion cracking, is a form of attack caused by


the combination of mechanical and corrosive actions. Metals subjected to cyclic
stress in a corrosive environment may be able to withstand a much reduced level

122

Brock

Table 2 Stress-Corrosion Data for Selected Copper Alloys in Moist Ammonia and
Mattssons Solution, pH 7.3
Time to failure (h)
Alloy
C11000
C19400
C22000
C23000
C24000
C26000
C26000
C35300
C35300
C42200
C42200
C42500
C42500
C44300
C44300
C51000
C63800
C65400
C66900
C68800
C70600
C71500
C72200
C75200
C75200
C76200
C77000
a

Condition
Hard
Hard
Extra hard
Cold rolled
Spring
Hard
Extra hard
Extra hard
Cold rolled
Cold rolled
Extra hard
Cold rolled
Spring
Cold rolled
Cold rolled
Extra hard
Extra hard
Cold rolled
Cold rolled
Cold rolled
Cold rolled
Cold rolled
Cold rolled
Extra hard
Cold rolled
Cold rolled
Extra hard

50%

50%
40%
40%
10%
40%

25%
25%
40%
50%
50%
10%
50%
50%

Composition (wt%)
99.9% Cu
2.4% Fe, 0.13 Zn, 0.04 P
10% Zn
15% Zn
20% Zn
30% Zn
30% Zn
35% Zn, 2% Pb
35% Zn, 2% Pb
12% Zn, 1% Sn, 0.2% P
12% Zn, 1% Sn, 0.2% P
9.3% Zn, 2% Sn, 0.2% P
9.3% Zn, 2% Sn, 0.2% P
28% Zn, 1% Sn, 0.04 As
28% Zn, 1% Sn, 0.04 As
5% Sn, 0.2% P
2.8% Al, 1.8% Si, 0.4% Co
3% Si, 1.6% Sn
15% Zn, 12% Mn
23% Zn, 3.4% Al, 0.4% Co
10% Ni, 1.4% Fe
30% Ni, 0.5% Fe
17% Ni, 0.75% Fe, 0.5% Cr
17% Zn, 18% Ni
17% Zn, 18% Ni
30% Zn, 12% Ni
27% Zn, 18% Ni

Mattssons
a

NF
NF
NF
34
15
4
4.7
2
2
NF
NF
NF
NF
42
13
NF
NF
NF
0.6
500
NF
NF
NF
880
530
29
60

Ammonia
NF
NF
11
2.4
1.2
1.2
3
17
10
37
29
1.2
16
393
28
NF
37
1
NF
NF
NF

422
16

NF no failures.

of stress compared to the stress level for the same number of cycles in air (21).
Failure under such conditions is termed corrosion fatigue. The process is characterized by cracks in the metal which are perpendicular to the tensile stress. Their
rate of propagation is usually faster than that of stress-corrosion cracks and they
are generally much straighter and with less crack branching. In contrast to stresscorrosion cracks, there is generally only one crack associated with failure by
corrosion fatigue. Crack initiation often occurs at the base of corrosion pits. Figure 3 shows a typical fatigue crack initiated on a pit in the outside of an Admiralty
alloy condenser tube. Cyclic stresses resulted from vibration of the tube within

Corrosion of Copper and Its Alloys

123

Fig. 2 Cross section showing typical branched intergranular stress-corrosion cracks in


alloy C26000 after immersion in a moist ammonia environment for 4 h. Magnification
320.

the condenser, the corrosive environment being ammoniated condensate. The micrographs reveal that the crack is associated with a small pit on the alloy surface
and that the crack path is transgranular. Copper alloys most resistant to corrosion
fatigue are those with a high fatigue limit and a high resistance to corrosion in
the particular environment. Typical of such alloys are the berylliumcoppers,
phosphor and aluminum bronzes, and the coppernickels.
G.

Intergranular Corrosion

As its name implies, this is a form of corrosion in which attack penetrates along
grain boundaries, often to a depth of several grains. The more rapid attack of the
grain boundaries is generally the result of a difference in composition between
the metal in the grain boundary and that in the bulk. Such differences in composition can result from segregation of impurities at the grain boundaries. As intergranular corrosion proceeds in an alloy, the rate of metal loss can increase with

124

Brock

Fig. 3 Cross section through the wall of an alloy C44300 tube showing (a) fatigue crack
initiating from small pit at 50 and (b) the transgranular nature of crack at 500.

Corrosion of Copper and Its Alloys

125

Fig. 4 Cross section of an alloy C44300 condenser tube revealing integranular corrosion
on the water side at 1000.

time. This is due to grains being isolated and removed from the alloy bulk. Copper
alloys most susceptible to this form of attack include Admiralty metal, aluminum
brasses, and silicon bronzes. Figure 4 shows an intergranular attack which occurred on the water side of an Admiralty alloy condenser tube which had been
in service for close to 30 years with lake water as coolant. Although the depth
of attack shown in the micrograph is only some three grains deep, intergranular
attack with removal of alloy grains by the flowing water had resulted in a loss
in wall thickness of the tube of up to 20 mils.
H. ErosionCorrosion
When copper alloys are exposed to an environment under conditions of flow,
such as in water tubes or condenser tubes, the flow rate can be significant in
determining the corrosion rate. This is because at high fluid velocities, the shear
stresses exerted by the fluid on protective films can lead to their removal. Syrett
and Lapel (22) defined a breakaway velocity beyond which the effects of erosion
cause the oxide to be partly removed. The rate of corrosion in the oxide-free
regions increases and is further accelerated by the galvanic action between the
oxide-free and oxide-covered areas. The nature of the attack often takes the form

126

Brock

of horseshoe-type pits with undercutting in the direction of flow. At higher flow


rates, more of the alloy surface becomes free of oxide and the rate of erosion
corrosion decreases because of a decrease in the galvanic effect.
Fluid flow in heat-exchanger tubes and condenser tubes is usually fully
developed turbulent flow except at the entrance of the tubes. In these regions,
local turbulence can promote erosioncorrosion. Additionally, high local fluid
velocities can arise around partial blockages in a tube, promoting erosioncorrosion.
Erosioncorrosion can also be associated with a phenomenon known as
cavitation. This is essentially the wearing away of protective films and metal by
repeated impact of blows resulting from the collapse and formation of vacancies
or bubbles within a fluid. Such can result from the application of ultrasonics to
a fluid and this is utilized in the ultrasonic cleaning of metal. Here, the formation
and collapse of vapor bubbles within the fluid results in a scouring of the metal
surface with the removal of soils even from recessed parts of the metal surface.
The collapse of the vapor bubbles at the metal surface can result in instantaneous
stresses of up to 200 ksi. Instances where erosion is promoted by cavitation in
corrosive environments include ships propellers and in pumps and piping systems subject to severe vibration. Because cavitation is essentially a mechanical
form of degradation, the harder forms of copper alloys, such as the aluminum
bronzes, are most resistant to it.
I.

Galvanic Corrosion

When two dissimilar metals are immersed in a solution and there is electrical
contact between the two, accelerated corrosion of the more electronegative metal
can occur while corrosion on the other is reduced. Such accelerated attack is
known as galvanic corrosion. Copper and its alloys are usually noble with respect
to other metals with which they come into contact. Thus, copper coupled to aluminum or steel will increase the corrosion of the aluminum and the steel while
being cathodically protected itself. In these instances, copper is the cathode in
the coppersolutionaluminum and coppersolutionsteel cells. The degree of
attack on the more active component is greatest at the point of contact with the
copper. Additionally, the attack is greater, the greater the ratio of the area of the
copper member to the more active member. An example of poor design with
respect to galvanic corrosion would be the use of steel rivets in copper plates or
steel nails in copper roof flashing. Very rapid deterioration of the steel would be
anticipated in such cases.
The electrochemical potential of a metal in solution is a function of the
solution concentration and type. Such potentials are not usually known. Table 3
(23) can be used as a general guideline for designing so that there is no problem
with galvanic corrosion. It lists the electrochemical potentials, with respect to a

Corrosion of Copper and Its Alloys

127

Table 3 Electrochemical Potential of Metals in


Seawater with respect to Saturated Calomel Electrode
Alloy

Electrochemical
potential (V)

Magnesium
Zinc
Aluminum alloys
Mild steelcast iron
Low-alloy steel
Aluminum bronze
Naval brassyellow brass
Tin
Copper
Leadtin solder
Admiralty brassaluminum brass
Manganese bronze
Silicon bronze
Nickel silver
9010 copper nickel
Stainless steel 430
Lead
7030 coppernickel
Nickel 20
Silver
Stainless Steel, 302, 304, 321, 347
Stainless steel, 316, 317
Titanium
Platinum
Graphite

1.6 to 1.63
0.99 to 1.2
0.97 to 0.99
0.6 to 0.71
0.58 to 0.63
0.31 to 0.42
0.3 to 0.4
0.3 to 0.33
0.3 to 0.36
0.28 to 0.37
0.28 to 0.35
0.28 to 0.33
0.25 to 0.28
0.24 to 0.29
0.22 to 0.27
0.2 to 0.27
0.19 to 0.24
0.18 to 0.23
0.1 to 0.2
0.1 to 0.13
0.05 to 0.1
0 to 0.1
0.05 to 0.04
0.25 to 0.18
0.3 to 0.2

Source: Ref. 23.

saturated calomel electrode, of various metals in seawater. The relative position


of the metals will be similar in most neutral solutions and dilute acids. Copper and
its alloys occupy the cathodic end of the table. Their corrosion can be galvanically
promoted by contact with graphite or titanium. Galvanic corrosion due to contact
between two different copper alloys is usually minimal because of the small differences in their electrochemical potentials. The above is illustrated by the results
in Table 4, which shows the effect of coupling to other metals on the corrosion
rates of some copper alloys in seawater (24).
The rate of galvanic corrosion is greatest when the difference in the electrochemical potentials of the two metals in contact is high. Therefore, the coupling

128

Brock

Table 4 Corrosion Rates of Alloys


in Coupled and Uncoupled Condition
After 2 Years of Immersion in
Flowing Seawater
Alloy

Uncoupled
C70600
C71500
C61400
Carbon steel
Titanium
Coupled
C70600
C61400
C70600
Carbon steel
C70600
Titanium
C71500
C61400
C71500
Carbon steel
C71500
Titanium

Corrosion
rate
(mils/year)
1.2
0.8
1.7
1.3
0.08
1
1.7
0.12
31
8.2
0.08
0.7
2.5
0.12
28
4.2
0.08

Source: Ref. 25.

of such metals in a corrosive environment should not be practiced unless the


coupling is via an insulating member.

III. OXIDATION OF COPPER ALLOYS


In air or oxygen at high temperatures, copper and high-copper alloys oxidize to
form films of cuprous oxide, Cu2O, or cupric oxide CuO, or, in many instances,
an outer layer of cupric oxide with an inner layer of cuprous oxide. The rate of
oxidation and hence the thickness of the oxide films formed in a given time
increases with increasing oxidation temperature. At temperatures below 200C,
the rate of oxidation of copper has been described in terms of an inverse logarithmic law (25) and a logarithmic law (26). The expressions are

Corrosion of Copper and Its Alloys

inverse logarithmic law

logarithmic law

1
t
k1 log
1
m
k2
m k 3 log

t
1
k4

129

where m is the weight of oxide formed, t is the time, and the other terms are
constants. At these temperatures, the thickness of the oxide film is usually of the
order of the wavelength of light. Because of this, oxides formed are often translucent and colored because of the property of light interference. The thickness of
the oxide films can be estimated from the color of the oxide as illustrated in
Table 5 (27).
At higher temperatures, the rate of oxidation is usually described in terms
of a parabolic rate equation (28) and sometimes in terms of a cubic rate equation
(29). The expressions are
m 2 k 5 t k 6 parabolic rate equation
m 3 k 7t k 8 cubic rate equation
where the terms have the same meaning as given previously. At temperatures in
the range 3501000C, thick oxide films are formed which spall off from the
metal when the samples are cooled.

Table 5 Color of Copper Oxide


Films as Function of Thickness
Oxide color
Dark brown
Red brown
Dark purple
Dark violet
Dark blue
Pale blue-green
Pale silvery green
Yellowish green
Yellowish green
Old gold
Orange
Red brown
Source: Ref. 27.

Thickness
)
(A
380
420
450
480
500
830
880
970
980
1110
1200
1260

130

Brock

Thermodynamic considerations would suggest that oxidation in air at 1 atm


and temperatures below 300C would favor the formation of cupric oxide, with
cuprous oxide being favored as an increasing fraction of the overall film as the
temperature increases above 300C. Results reported in the literature are generally in accord with this. Thus, Hickman (30), using electron diffraction studies,
demonstrated that between 20C and 300C, cupric oxide films were formed. At
350C, films consisting of both cuprous oxide and cupric oxide were identified,
and above 450C, the films consisted entirely of cuprous oxide.
Alloying copper with elements such as Fe, Mn, or Ni has only a minor
effect on their oxidation behavior (31). Additions of Al, Be, Mg, Si, or Zn can
significantly decrease the oxidation rates. This is because oxides of the alloying
constituents form beneath the copper oxide layer. At high-alloying contents, such
as 15% Zn (32) or 5% Al (33), oxidation at temperatures in excess of some 500C
results in the formation of a complete layer of zinc oxide or alumina adjacent to
the surface, which significantly decreases the oxidation rate of the metal.
Copper alloys are not used as materials for oxidation resistance at high
temperatures. However, the above types of oxides can result when the alloys
are heat treated in air at high temperatures. Removal of copper oxides is generally accomplished by immersion in 1020% sulfuric or hydrochloric acids.
Refractory oxides, such as silica or alumina, must be removed by treating with
1020% sulfuric acid made oxidizing by the addition of 3% hydrogen peroxide. In this case, the oxides are removed by dissolution of the alloy layers under
them.
The formation of copper oxides can be reduced by annealing in nitrogen
atmospheres and can be completely prevented by annealing in nitrogen containing
14% hydrogen. Such reducing atmospheres are useful because they do not form
explosive mixtures with air. Reducing atmospheres with much higher hydrogen
contents can be used to anneal brasses without oxidation of Zn. However, these
atmospheres will not prevent the oxidation of reactive elements such as Si, Al,
or Mg. These either oxidize to form internal oxides, often preferentially at grain
boundaries, or they form uniform layers of the oxide of the alloying elements.
Transition from internal to external oxidation is favored by increasing the temperature and increasing the alloying content of the reactive phase. Figure 5 demonstrates this for coppersilicon alloys with increasing silicon content oxidized at
600C in an atmosphere of nitrogen4% hydrogen (19).

IV. ATMOSPHERIC CORROSION


Copper and its alloys have a high resistance to corrosion in the atmosphere. Although copper oxidizes on exposure to the atmosphere, corrosion is prevented
by the formation of an adherent layer of corrosion products. In some instances,

Corrosion of Copper and Its Alloys

131

Fig. 5 Cross sections of copper silicon alloys oxidized for 16 h at 600C in nitrogen
4% hydrogen showing transition from internal oxidation to external film formation as
silicon content increases from (a) 0.2% through (b) 0.5% to (c) 2.5% at bottom. Magnification 800.

this coating takes the form of a green patina. On copper on roofs or statues, this
patina provides for an aesthetically pleasing appearance.
Vernon and Whitby (33) characterized the composition of the patina formed
on copper exposed in England for times ranging up to 300 years. Under most
conditions of exposure, they showed that the major constituent was brochantite,
a basic copper sulfate. A basic copper chloride, atacamite, was found in the corrosion products formed on copper near the sea. Lesser amounts of malachite, a
basic copper carbonate, were also detected.
It is evident from work described in the literature that the rate of corrosion
of copper in the atmosphere depends on the specific environment, be it urban,
marine, or rural, Other factors which play a role are the temperature and humidity,
the degree of pollution, and the relative time spent between wet and dry conditions. From measurements of the increase in resistance of copper and copper-

132

Brock

alloy wires exposed in an urban atmosphere, Hudson (34) showed that over a
time of close to 5 years, copper corroded at an average rate of 0.195 mils/year
Slightly lower corrosion rates were observed with silicon and tin bronzes but
significantly higher corrosion rates for 8020 and 7030 cupronickel alloys, 70
30 brass, and an aluminum bronze.
Thompson et al. (35) used weight-loss measurements to determine the rate
of corrosion of sheet and wire samples of 11 different copper alloys in various
atmospheres over a time of up to 2 years. Over this time, they determined average
corrosion rates of 3447, 2737, and 1217 in. per year in industrial, marine,
and rural atmospheres, respectively. Tracy (36) also summarized the corrosion
rates of copper and several copper alloys exposed for 20 years at eight different
sites. For most of the alloys, the corrosion rates were within the ranges 230,
1090, and 50120 in. per year. Exceptions to the behavior were two brasses,
for which the rates of attack were 180450 mil/year.
Holm and Mattsson (37) evaluated sheet and rod samples of a wide range
of copper alloys after exposure for times up to 16 years in rural, marine, and
urban atmospheres in Sweden. They observed that dark brown coatings consisting essentially of copper oxides together with some copper salts and alloying
constituents formed in the first year of exposure. In urban and marine environments, green coatings appeared on copper after 67 years and even earlier on
phosphor bronzes. However, green coatings were not formed on high-zinc alloys.
Such coatings were not formed on any of the alloys exposed at rural sites
even after 16 years exposure. Instead, black or brown films were formed. They
showed that the average rate of loss of metal by general corrosion after 7 years
was much lower than that after only 2 years of exposure but similar to that after
exposure for 16 years. The values were 1220 in./year in rural atmospheres,
2035 in./year in marine atmospheres, and 3552 in./year in urban atmospheres.
Penetration of the brasses by dezincification occurred at rates significantly
higher than those described. It was greatest for two-phase alloys and after 16
years had penetrated to 3.55.6 mils. Their results showed that two-phase alloys
containing additions of aluminum, tin, or arsenic improved the dezincification
resistance. Single-phase brasses containing arsenic additions were shown to have
good resistance to dezincification.
Costas (38) evaluated a range of copper alloys after exposure for up to 20
years in rural, marine and industrial locations. He observed that at industrial sites,
green coatings predominated on samples free of zinc or nickel. Blue or green
hues were seen on a few alloys exposed at both marine and rural sites but not
to the same degree as at the industrial sites. The average corrosion rates varied
from 9 to 90 in./year and were greatest at the industrial site.
Figure 6 shows weight losstime plots for alloy C26000 exposed at three
different sites and illustrates the dependency of the corrosion rate on the site

Corrosion of Copper and Its Alloys

133

Fig. 6 Weight losstime plots for alloy C26000 exposed to the atmosphere at three
different sites.

conditions (19). The New Haven site is representative of an industrial atmosphere, Daytona Beach represents a marine environment with negligible industrial pollution, and East Alton represents a severe industrial atmosphere. The
corrosion rate of the brass is much less at the Daytona Beach site, reflecting the
absence of industrial pollutants. Figure 7 shows similar plots for three alloys
exposed at Daytona Beach. The corrosion rate of the high-copper alloy, C11000,
is significantly greater than those of brass, C26000, and of the coppernickel
alloy, C70600.
The plots also demonstrate that, with time, the corrosion rates generally

Fig. 7 Weight losstime plots for selected copper alloys exposed in New Haven.

134

Brock

Table 6
Sites

Corrosion Rates of Copper Alloys After 2 Years of Exposure at Various

Alloy
C11000
C22000
C23000
C26000
C42500
C51000
C65400
C68800
C70600
C76200

Corrosion rates (mils/year)

Alloying
elements
(wt%)

Daytona
Beach

New
Haven

East
Alton

99% Cu
10% Zn
15% Zn
30% Zn
9.3% Zn, 1% Sn, 0.2% P
5% Sn, 0.2% P
3% Si, 1.6% Sn
2.8% Al, 1.8% Si, 0.4% Co
10% Ni, 1.4% Fe
3% Zn, 12% Ni

0.1
0.07
0.03
0.03
0.08
0.15
0.28
0.02
0.06
0.02

0.04
0.03
0.03
0.05
0.04
0.05
0.04
0.03
0.03
0.05

0.04
0.06
0.04
0.04
0.06
0.04
0.04
0.03
0.06
0.04

Source: Ref. 19.

decrease because the corrosion product layers provide some resistance to further
attack. The rate of corrosion at any time is given by the slope of a particular plot
at that time. Table 6 shows the corrosion rates so determined for several alloys
after 2 years of exposure at the three sites.
Stress-corrosion cracking of certain copper alloys can also occur during
exposure to the atmosphere. Table 7 shows stress-corrosion data for copper-alloy
samples exposed as U-bends at New Haven, Daytona Beach, and Brooklyn, NY.
In contrast to the data shown in Table 3 for laboratory testing in moist ammonia
and the Mattssons solution, brasses with 15% or less Zn did not fail within the
10-year exposure time. However, failures were observed for alloy C66900 with
only 15% Zn but with 12% Mn. The results also demonstrate a sensitivity to the
site, with the marine site being least aggressive in promoting stress-corrosion
cracking.
Under certain conditions, much more rapid atmospheric corrosion of copper
alloys can occur. Typically, such conditions include the presence of sulfides such
as hydrogen sulfide, which leads to the formation of black, much less protective
films of copper sulfide.

V.

WATER STAINING

Very rapid staining of copper alloys can occur under specific conditions. Typically, this occurs on freshly cleaned strip or parts which have not been fully

Corrosion of Copper and Its Alloys

135

Table 7 Time to Failure by Stress-Corrosion Cracking of Selected Copper Alloys


Exposed as U-Bends at Various Atmospheric Sites
Alloy
110
194
220
230
260
353
353
422
425
425
443
443
510
638
654
669
688
706
715
722
752
752
762
770

Condition
Hard
Hard
Extra hard
Cold rolled
Hard
Extra hard
Cold rolled
Extra hard
Cold rolled
Extra hard
Cold rolled
Cold rolled
Extra hard
Extra hard
Cold rolled
Cold rolled
Cold rolled
Cold rolled
Cold rolled
Cold rolled
Extra hard
Cold rolled
Extra hard
Extra hard

50%

50%
40%
10%
40%

25%
20%
40%
50%
50%
10%
50%

Alloying elements
(wt%)

New
Haven

Daytona
Beach

East
Alton

99.5% Cu
2.4% Fe, 0.13% Zn, 0.04% P
10% Zn
15% Zn
30% Zn
35% Zn, 2% Pb
35% Zn, 2% Pb
12% Zn, 1% Sn, 0.2% P
9.3% Zn, 2% Sn 2% P
9.3% Zn, 2% Sn, 0.2% P
28% Zn, 1% Sn, 0.04% As
28% Zn, 1% Sn, 0.04% As
5% Sn, 0.2% P
2.8% Al, 1.8% Si, 0.4% Co
3% Si, 1.6% Sn
15% Zn, 12% Mn
23% Zn, 3.4% Al, 0.4% Co
10% Ni, 1.4% Fe
30% Zn, 0.5% Fe
17% Ni, 0.75% Fe, 0.5% Cr
17% Zn, 18% Ni
17% Zn, 18% Ni
30% Zn, 12% Ni
27% Zn, 18% Ni

NF a
NF
NF
NF
133
856
79
NF
NF
NF
NF
69
NF
NF
NF
11
2080
NF
NF
NF
NF
3100
667
396

NF
NF
NF
NF
140
NF
1256
NF
NF
NF
NF
61
NF
NF
NF
7.5
NF
NF
NF
NF
NF
NF
1105
NF

NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
9.5
NF
NF
NF
NF
NF
NF
NF
146

Source: Ref. 19.


a
NF no failures.

dried. The staining occurs under water drops, which are the last to evaporate.
The situation contrasts with tarnishing in bulk water, which only occurs slowly
because the tarnishing process requires diffusion of oxygen to the metal surface.
Under a water drop, the diffusion distance of the oxygen from the air to the metal
surface is small. The rates of tarnishing are then much higher, and brown stains
of copper oxides can result in times of only a few minutes.
Under conditions of high humidity and temperatures, such rapid tarnishing
can also occur even if the metal is originally dry. On cleaned dry parts held in
a barrel, moisture can condense out when the temperature falls, as, for example,
during the night. This water is trapped at points where components are in contact
and held there by capillary action. Rapid tarnishing, because of the ease of diffu-

136

Brock

sion of oxygen, through the thin water layers results. Similar rapid tarnishing can
occur between two alloy strips when one is placed on top of the other. Moisture
condensing from the atmosphere is trapped in the space between the two strips
and drawn into the center of the strip by capillary action. During severe conditions
of humidity and temperature, tarnishing can be such as to result in the formation
overnight of violet films on the surfaces of the strips which are in contact, with
no tarnishing on surfaces not in contact.
The rapid tarnishing described occurs on all copper alloys and the rate of
tarnishing is worse if there are contaminants on the alloy surface, such as those
resulting from ineffective rinsing following an acid clean. The tarnishing can be
reduced by maintaining a low ambient humidity and temperature in the area
where parts are dried and stored. Significant decrease in the tarnishing rates can
also be obtained by final rinsing in water containing 0.10.5 wt% benzotriazole
(BTA). Such treatment results in the formation of a protective film of a BTA
thick. This serves to inhibit attack of the
copper complex only some 2040 A
alloy surface.

VI. SOIL CORROSION


Copper generally has excellent resistance to soil-side corrosion. The degree of
attack varies with the nature of the soil. Thus, Gilbert (39) examined the corrosion
of copper buried in soils at seven sites in England for times of up to 10 years.
He showed that the most corrosive soils were a wet acid peat with a pH of 4.3
and a moist acid clay with a pH of 4.6. The corrosion rates in these averaged up
to 260 in./year with a pitting rate of up to 1.8 mils/year. In the less corrosive
soils, the general corrosion rate ranged from 2 to 10 in./year and there was no
pitting. In a further test, a 5-year exposure of phosphorus-deoxidized copper in
cinders resulted in high corrosion rates of up to 260 in./year, with pitting at a
rate of 12.6 mils/year. These high corrosion rates were shown to be due to the
presence of sulfides and sulfate-reducing bacteria.
Logan and Romanoff (40) evaluated samples exposed for up to 14 years
in 14 different soils in the United States. The greatest attack was observed in
soils where the backfill contained cinders or had high organic or inorganic acid
content. Losses in wall thickness ranged from 0.2 to 2 mils and pit depths were
up to 51 mils.
From the above and work conducted by Denison (41), Meyers and Cohen
(42) described conditions which could render soils corrosive to copper. These
included elevated sulfate and chloride contents, poor drainage, inorganic and organic acids, cinder fills, sulfate-reducing bacteria, and ammoniacal compounds.
Other factors affecting the corrosion behavior are the aeration characteristic of
the soil, differential aeration, and stray currents.

Corrosion of Copper and Its Alloys

137

Fig. 8 Weight losstime plots for selected copper-alloy tubes resulting from New Haven
tap water flowing at 8 ft/s at ambient temperature.

Cohen and Brock (43) recently described the soil-side corrosion of copper
water service lines removed from various streets in Billings, Montana after service in the range 1070 years. Their findings attested to the excellent resistance
of copper in the alluvial soil characteristic of this location. In some instances,
tubes in use for 70 years had essentially no evidence of corrosion on the soil
side. The degree of corrosion varied from site to site, but in the worst case, pits
only 11 mils deep were found, and these were on a tube which had been in service
for 35 years.

VII. WATER CORROSION


A. Potable Water
Copper and its alloys generally have an excellent resistance to corrosion in freshwater and are widely used in water distribution systems and as plumbing tubes.
In the United States, some 375,000 miles of plumbing tube is installed each year.
Mostly, this is in the form of phosphorus deoxidized, alloy C12200, tubes. In
Ref. 43, these alloy tubes were shown to have suffered little if any water-side
attack even after 70 years of service.
The corrosion rates of copper alloys in water are also a function of alloy
type. Figure 8 shows weight losstime plots for three alloy tubes exposed to New
Haven tap water flowing at 7 ft/s under ambient conditions (19). The rate of
corrosion decreases with time due to the formation of protective copper oxide
films. Table 8 summarizes the corrosion rates in New Haven water after 1 year.
The lowest corrosion rate was observed with alloy C19400 containing 2.4% Fe

138

Brock

Table 8 Corrosion Rates of Copper Alloys After 1


Year in Flowing New Haven Tap Water at 7 ft/s
Alloying
elements
(wt%)

Corrosion
rate
(mils/year)

0.02% P
2.4% Fe, 0.13% Zn, 0.04% P
15% Zn
20.5% Zn, 2% Al
10% Ni, 1.4% Fe

0.32
0.1
0.4
0.2
0.27

Alloy
C12200
C19400
C23000
C68700
C70600

Source: Ref. 19.

and the highest for alloy C230 with 15% Zn. The corrosion rates are low and
all less than 0.5 mils/year.
Despite the outstanding corrosion resistance, there are isolated instances
where corrosion can lead to problems. Pitting of copper tubes, although infrequently found, can occur in certain aggressive waters and lead to early failure.
In water at temperatures above 140F, pitting is particularly unusual. Waters
which promote such pitting are soft, have a pH less than 7.4, and have a ratio
of bicarbonate to sulfate ion content of less than 1 (44). Failures have occurred
in Canada, Sweden, Europe, and the United Kingdom. The author has seen rapid
failure of copper tubes in hot, soft well water in Connecticut (U.S.A.). Pits which
form in hot water are deep and narrow and contain cuprous oxide. They are
generally capped by black or greenish-black mounds of copper oxide and basic
copper sulfate. The pits are often surrounded by a deposit high in alumina. In
some hot waters, the presence of manganese can promote pitting and form somewhat larger pits which are surrounded by black deposits of manganese dioxide.
Cold-water pitting has been widely described in the literature, the studies
mainly relating to failures in the United States, the United Kingdom, and Belgium. It requires an aggressive water. Such pitting typically arises in hard waters
from deep wells. The pits are usually hemispherical in nature and form under
green mounds of malachite, a basic copper carbonate. When such nodules are
removed, a brown film of cuprous oxide is revealed with a central hole (Fig. 9a).
Close examination of this reveals that it has an underlying crystalline nature (Fig.
9b). In cross section (Fig. 9c), this oxide is seen to form a membrane which
covers the underlying pit. The base of the pits are filled with copper chloride.
Lucey (45) proposed a mechanism by which such pits are formed. The copper
oxide membrane is critical to his theory. He proposed that the copper oxide membrane initially promotes the formation of pockets of copper chloride adjacent to

Corrosion of Copper and Its Alloys

139

the metal surface. In continuing pit growth, the membrane then separates the
anodic process within the pit from the cathodic process of oxygen dissolution on
the water side of the membrane. His generally accepted mechanism clearly requires the presence of chloride and bicarbonate ions in the water, as well as
dissolved oxygen.
Cohen and Lyman (46) analyzed 65 waters in the United States where
pitting had occurred, and they showed that they typically contain over 5 ppm
carbon dioxide, have a pH value in the range 7.07.8, contain 1012 ppm oxygen,
and have a sulfate content generally three to four times that of the chloride ion
content. These findings are in accord with the work of OBrecht et al. (47), who
showed that aggressive waters contained high concentrations of dissolved carbon
dioxide and oxygen, a pH in the range 7.07.8, and a sulfate-to-chloride ratio
of from 34 to 1. Following analysis of many aggressive waters in the United
Kingdom, Lucey (48) produced a nomogram, the use of which permits termination of the pitting propensity of the water as determined from its sulfate, sodium,
chloride, and bicarbonate ion content, the concentration of dissolved oxygen, and
the pH.
The presence of carbon films can promote pitting in aggressive waters
(49,50). Such films can be present on a half-hard tube as a result of the decomposition of drawing oil during annealing operations. The films can promote pitting
by isolating pockets of copper chloride adjacent to the alloy surface and enhance
pitting by providing a large cathode-to-anode ratio.
In certain potable waters, corrosion of plumbing fixtures, which are usually
fabricated from leaded brass alloys, can occur. The factors which promote dezincification are high temperatures, water stagnation, the presence of crevices where
aeration is restricted, a high ratio of chloride ion to carbonate ion, and a relatively
high pH (2). In terms of plumbing fixtures, such an attack causes problems not
only because of metal degradation but also because of the formation of voluminous corrosion products which can cause blockages of valves and freezing of
valve stems.
B. Freshwater Cooling Systems
The excellent corrosion resistance of copper alloys in water results in their widescale utilization in power utility condensers and heat exchangers where the source
of water is from rivers or lakes. Alloys typically used for condenser applications
include C14200, C19400, C44300, C68700, C70600, and C71500. In systems
in which the cooling water is circulated through cooling towers, evaporation,
blowdown, and makeup of the water results in an increase in concentration of
the salts initially present in the water. The waters can then become more corrosive. Coppernickel alloys find more application in such systems.

140

Brock

Corrosion of Copper and Its Alloys

141

Fig. 9 Micrographs showing characteristics of pits formed on copper alloys in potable


water: (a) perforated membrane of cuprous oxide exposed after removal of corrosion product nodule at 25; (b) details of a membrane revealing its crystalline nature at 1000;
(c) cross section through small nodule revealing pit, corrosion products in pit, oxide membrane, and overlying nodule at 200.

C. Saltwater
Copper alloys have a high resistance to corrosion in seawater. Although copper
itself has good resistance, the coppernickel alloys are among the most resistant.
This is illustrated in the weight-loss results shown in Table 9 obtained for panels
of selected copper alloys exposed for various times on racks below the low-tide
level at Daytona Beach (19). These results are consistent with those obtained
from much longer immersion in tidal seawater (51). In these, the corrosion rates
of alloys C70600 and C71500 after immersion for 14 years were shown to be
only 1.1 and 0.8 m/year.
The good corrosion resistance of copper alloys in saltwater or seawater has
resulted in their wide-scale use in ships and in tidal power station condensers.
Alloys used in such applications include the inhibited Admiralty alloys C44300,
C44400, and C44500, and alloys C61300, C68700, C70600, C71500, and
C71640. Instantaneous corrosion rates and weight losses of selected copper alloys
after 1 year obtained from laboratory studies in 3.5% sodium chloride solution

142

Brock

Table 9 Weight Loss of Copper Alloys After Immersion Below Low-Tide Level for
Various Times at Daytona Beach

60 Days

156 Days

365 Days

99.5% Cu
2.4% Fe, 0.13% Zn, 0.04% P
15% Zn
30% Zn
10% Ni, 1.4% Fe
27% Zn, 18% Ni

10.6
9.6
8.9
17.2
3.6
9.7

14.6
11
102
15.6
3.4
14.2

19.7
16
16.1
21.9
5.1
20.9

Alloy
C11000
C19400
C23000
C26000
C70600
C77000

Weight loss
(mg/cm 2)

Alloying
elements
(wt%)

Source: Ref. 19.

at 40C and flowing at 5 ft/s are shown in Table 10 (19). The various weight
losses reflect the more rapid rate of corrosion of the alloys which occurs during
the early stages of the tests.
The corrosion resistance of copper alloys in nonpolluted seawater results
from the formation of protective films over the surface of the alloy. These have
been shown to be cuprous oxide with outer layers enriched in iron and nickel
(52). Analysis of the protective films formed on a coppernickel condenser tube
removed from tidal power stations alloys confirms these findings (19). Highresolution scanning electron microscopy of cross sections of such films always
reveals an inner layer of cuprous oxide with layers high in nickel and then iron

Table 10 Corrosion Rates and Weight Losses of Copper Alloys After 1 Year
with 3.5% Sodium Chloride at 40C Flowing at 7 ft/s

Alloy

Alloying
elements
(wt%)

Corrosion
rate
(mils/year)

Weight
loss
(mg/cm 2)

C12200
C19400
C44300
C68700
C70600
C71500

0.02% P
2.4% Fe, 0.13% Zn, 0.04% P
28% Zn, 1% Sn, 0.04% As
20.5% Zn, 2% Al
10% Ni, 1.4% Fe
30% Ni, 0.5% Fe

0.6
0.1
0.45
0.1
0.1
0.1

38.6
16.0
10.3
4.0
4.7
3.8

Source: Ref. 19.

Corrosion of Copper and Its Alloys

143

oxides above this. Above these, are a layer of paratacamite, Cu2(OH)Cl, and,
invariably, a layer of material deposited from the water, which usually has a high
silica content often with significant concentrations of oxides of iron.
Examination of Alloy C687 condenser tubes from a coastal power station revealed that some 20% of the surface was covered with an orange layer
containing iron, copper, oxygen, and sodium (53). The remaining surface was
covered with a double layer, with the outer layer being a porous material
consisting mostly of iron oxide and an inner later consisting of hydrotacite,
Mg 6 Al 2 (OH) 16 CO3 4H2O, and lesser amounts of paratacamite. Further work on
tubes removed from many power stations revealed that there was always an outer
layer of -FeOOH with a thin underlying layer containing Mg, Al, Zn, and Cu
(54).
Selection of alloys for seawater condenser tubes is based on factors other
than their intrinsic corrosion resistance. An important consideration is the resistance of the alloy to erosioncorrosion resistance of the material. This determines
not only the maximum coolant velocity but also the resistance to increased turbulence round partial blockages in the tubes. Acceptable maximum velocities obtained from both laboratory and service performance are listed in Table 11 (55).
The high resistance of coppernickel alloys to erosioncorrosion by seawater
was demonstrated in the early pioneering work of Stewart and LaQue (56). They
showed that at velocities of 30 ft/s, the erosioncorrosion resistance of Cu10%
Ni alloys increased with increasing iron content and leveled off at the 1.4% iron

Table 11 Accepted Maximum Tubular Design


Velocities for Some Copper Alloys
Alloying
elements
(wt%)

Maximum
velocity
(ft/s)

0.02% P
28% Zn, 1% Sn, 0.04% As
5% Al
7% Al, 0.3% Sn
20.5% Zn, 2% Al
1.5% Si
10% Ni, 1.4% Fe
30% Ni, 0.5% Fe
17% Ni, 0.75% Fe, 0.5% Cr

23
46
9
9
8
3
1012
15
30

Alloy
C12200
C44300
C60800
C61300
C68700
C65100
C70600
C71500
C72200

Source: Ref. 25.

144

Brock

content. Some of this resistance was lost when the alloy was annealed. This precipitates the iron out in an ironnickelcopper phase. These findings were instrumental in setting the iron content of Alloy C70600 close to 1.4% with magnetic
permeability measurements being conducted to ensure that the iron is essentially
retained in solid solution. Similar increases in erosioncorrosion resistance of
copper30% nickel were observed when the iron content was increased from
0.05% to 0.46%. Increasing the iron content above these levels can result in a
increased tendency for pitting corrosion (57). The presence of 2% Mn and 2%
Fe in an alloy with 15% Ni, C71640, and the Cr addition in Alloy C72200 also
leads to high resistance to erosioncorrosion and in preventing attack in severe
conditions where there is entrained sand (57).

VIII. FOULING RESISTANCE


During seawater service, fouling of materials can occur. This phenomenon consists of the attachment of marine organisms on the surface of the materials. These
organisms include algae, sponges, barnacles, oysters, and mussels. Copper alloys
have a high resistance to such fouling, with Alloy C70600 being the best and
superior in this respect to Alloy C71500 and Alloy C68700 (58). Studies of the
fouling resistance of copper alloys have attributed the affect to poisoning of the
organisms by the slow release of copper ions from the alloy surface (59) and to
the toxicity of cuprous oxide to marine organisms as well as the sloughing off
of outer corrosion product layers of basic copper oxides (60).
Marine fouling can result in the buildup of substantial layers of marine
organisms on boat hulls, resulting in damage and added resistance to movement
through the water. The resistance of coppernickel alloys to biofouling has resulted in their successful use for hulls of yachts and shrimp trawlers (61). In
tidal power stations, coppernickel alloys have proved useful as seawater in the
condenser screens because of the resistance to both corrosion and biofouling.

IX. EFFECT OF POLLUTION


The high corrosion resistance of copper alloys in waters is adversely affected
when pollutants are presentin particular, sulfides. The sulfides are introduced
into the water by sulfate-reducing bacteria which under anaerobic conditions reduce sulfate ions to sulfides. Alternatively, sulfides can be introduced into waters
by the decomposition of plant or animal matter. The decrease in corrosion resistance is associated with the incorporation of sulfides into the corrosion product

Corrosion of Copper and Its Alloys

145

films. These render the films far less protective to subsequent corrosion. Pitting
of copper alloys is typical in sulfide-polluted conditions.
The sensitivity of the corrosion resistance to sulfide ions is remarkable, and
sulfides at the level of only 0.01 ppm can promote attack on Alloy C706 (62).
Other work has demonstrated that the adverse affect of sulfides increase with
increasing coolant velocity (63).
Polluted waters can lead to early failures of condensers. Elimination of the
source of the pollutant is an obvious step in preventing this type of attack. Decaying plant and animal life can be prevented from entering condensers by the
use of screens and filters. Aeration of the water is also beneficial in removing
sulfides. The intentional formation of hydrated iron oxides is also beneficial in
decreasing the rate of attack by polluted waters. Iron can be introduced into waters
either by an iron anode or by intermittently introducing ferrous ions into the
water by the addition of ferrous sulfate (64). The protective films are then deposited over the surface of the existing corrosion product films.

X.

STEAM

In the absence of carbon dioxide, ammonia, and oxygen, copper alloys are resistant to attack by steam. This property is important in the wide use of copper
alloys in condenser applications. At high steam velocities, such as may be seen
where steam enters a condenser, erosion of copper alloys can occur. Additionally,
if the steam is wet, then water droplets impinging on the metal can promote severe
attack in the form of a high frequency of narrow, deep pits. This is illustrated in
Fig. 10 for an Admiralty alloys tube which had been exposed to such an impingement attack (19). The attack can be averted by incorporation of appropriate baffles
to prevent the water droplets and steam from directly impinging on the tubes.

XI. AMMONIACAL SOLUTIONS


Concentrated ammoniacal solutions are corrosive to copper alloys and, as described previously, can promote stress-corrosion cracking. In service conditions
copper-alloy condenser tubes often are exposed to condensate which contain low
levels of ammonia. The ammonia originates from the decomposition of oxygen
scavengers such as hydrazine or morpholine. The concentrations of ammonia are
highest in the air-removal section of the condensers. Although the concentrations
of ammonia are only a few ppm, its presence together with oxygen can promote
attack. Coppernickel alloys are most resistant to attack by ammoniated conden-

146

Brock

Fig. 10 Micrograph (at 100) of pits on condensate side of an alloy C706 condenser
tube resulting from impingement of water droplets in wet steam.

sate. This is supported by the data for various copper alloys under both laboratory
and field test conditions, as shown in Table 12 (65).

XII. CORROSION IN OTHER ENVIRONMENTS


The preceding sections describe the behavior of copper alloys in environments
in which they are most used. The metals are also used in a wide range of chemical
equipment, including pipelines, fractionating columns, heat exchangers, and condensers. Space does not permit detailed descriptions of these. Such information
is available in a number of publications (66,67). The subsequent subsections give
a general description of the behavior of copper and its alloys in chemical environments.
A. Acid Solutions
All copper alloys are attacked by oxidizing acids such as nitric acid and strong
sulfuric acid. In deaerated nonoxidizing acids, copper is essentially immune to
corrosion. The presence of dissolved air in the acid will, however, result in attack

Corrosion of Copper and Its Alloys

147

Table 12 Comparison of Field and Laboratory Data for Condensate-Side Corrosion


of Copper Alloys
Corrosion rate (mils/year)
Alloy

Plant A

Plant B

Plant C

Laboratory

C71500
C72200
C70600
C44300

0.0083
0.016
0.019
0.05

0.004
0.016
0.014
0.031

0.015
0.015
0.018
0.024

0.002
0.008
0.043
0.09

Note: Corrosion rates are after 2 years exposure with laboratory data extrapolated from 1000 h in 2
ppm ammonia at pH 9.4.
Source: Ref. 19.

of the metal. The rate of attack in acids is therefore proportional to the acid
concentration, the temperature, the degree of aeration, and the flow rate. Under
fairly mild conditions, copper alloys are successfully used for handling hydrofluoric, sulfuric, phosphoric, acetic, and other organic acids (6871). Alloys most
resistant are the tinbronzes, the aluminum bronzes, the silicon bronzes, and the
cupronickels.
B. Alkali Solutions
Copper alloys have good resistance to alkali solutions other than ammonium
hydroxide. Coppernickel alloys are most resistant and corrosion rates of less
than 0.2 mils/year are typical of Alloy C71500 in 1N to 2N sodium hydroxide
solutions. For the phosphorbronze alloys, the corrosion rates under the same
conditions are some 10 mils/year.
C. Neutral Solutions
Copper and its alloys are suitable for handling most neutral solutions of nonoxidizing salts. They are used for handling solutions of nitrates, sulfates, and chlorides. Solutions of oxidizing salts, such as those containing chromate, ferric, or
stannic ions, can promote rapid attack. Similarly, salts of metals more noble than
copper will promote attack while plating out on the metal surface.
D. Organic Compounds
Copper alloys are resistant to a wide range of organic compounds such as amines,
ester, ethers, ketones, alcohols, aldehydes, naphtha, and gasoline. In amines, the

148

Brock

corrosion resistance is significantly decreased if there is contamination by water.


These conditions can also promote stress-corrosion cracking of brass. Alloys
C44300 and C71500 are used in equipment for refining gasoline. Copper is extensively used for kettles in the brewing of beer and for evaporators and heating
coils in the manufacture of cane and beet sugar.

REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.

MED Turner. Proc Soc Water Treat Exam 10:162, 1961.


MED Turner. Proc Soc Water Treat Exam 14:81, 1965.
D Bengough, RM Jones, R Pirret. J Inst Metals 23:65158, 1920.
GD Bengough, R May. J Inst Metals 32:81269, 1924.
L Kenworthy, GO Driscoll. Corros Technol 2:247, 1955.
HW Pickering. J Electrochem Soc 117:815, 1970.
HW Pickering, PJ Byrne. J Electrochem Soc 116:14921496, 1969.
MJ Pryor, RG Blundy. Corros Sci 12:6575, 1972.
DS Keir, MJ Pryor. J Electrochem Soc 127:21382144, 1980.
GD Bengough, R May. J Inst Metals 32:81, 1924.
R May. Trans Inst Mar Eng 49:171, 1937.
E Edmunds, EA Anderson, RK Waring. Symposium of Stress Corrosion Cracking
of Metals. Philadelphia: ASTM, 1944.
H Rosenthal, AL Jamiseon. Trans AIME 156:212, 1944.
HE Johnson, J Leja. Corrosion 22:178189, 1966.
AV Boblyer. Intercrystalline Corrosion of Metals Under Stress. New York: Consultants Bureau, 1962, p. 298.
RG Johnson. Sheet Metal Ind 145:11971199, 1211, 1940.
FL Laque. Marine Corrosion Causes and Prevention. New York: John Wiley & Sons,
1975, p. 6.
Annual Book of ASTM Standards Part 10, Metals-Physical, Mechanical, Corrosion
Testing, G37, 1982, pp. 9991002.
Technical Reports, Metal Research Laboratories, Olin Corporation, New Haven, CT.
Annual Book of ASTM Standards, Section 2, Vol. 02.01 Copper and Copper Alloys,
B154-92, 1994, pp. 254256.
DJ McAdam Jr. Proc ASTM 27, 1927.
BC Syrett. ErosionCorrosion of Copper-Nickel Alloys. Stanford Research Institute,
1974.
FL Laque. Marine Corrosion Causes and Prevention. New York: John Wiley & Sons,
1975, p. 179.
Metals Handbook Ninth Addition, Volume 13, Corrosion. Metals Park, OH: ASM
International, 1987, p 624.
FW Young, JV Cathcart, AT Gwathmey. Acta Metall 4:145153, 1956.
B Lustman, F Mehl. Trans AIME 143:246, 1941.
FH Constable. Proc Royal Soc. A117:376, 1928.
NB Pilling, RE Bedworth. J Inst Metals 29:529591, 1929.

Corrosion of Copper and Its Alloys

149

29. RF Tylecote. J Inst Metals 78:259350, 19501951.


30. JW Hickman. Proceedings, Pittsburgh International Conference on Surface Reactions. Corrosion Publication Co., 1948, p. 156.
31. RL Levin, C Wagner. J Electrochem Soc. 108:954959, 1961.
32. H Nishimura. Chem Abstr 33:1252, 1939.
33. WHJ Vernon, L Whitby. J Inst Metals 44:389408, 1930.
34. JC Hudson. Metals Ind (London) 44:415, 1934.
35. DH Thompson, AW Tracy, JR Freeman JR. Symposium on Atmospheric Corrosion
of Non-Ferrous Metals. Philadelphia; American Society for Testing and Materials,
1955, p. 67.
36. AW Tracy. Corrosion 7:373, 1951.
37. R Holm, E Mattsson. In: SW Dean Jr, EC Rhea, eds. Atmospheric Corrosion Tests
of Copper and Copper Alloys in Sweden16 Year Results. Philadelphia: American
Society for Testing and Materials, 1982, pp. 85105.
38. LP Costas. In: SW Dean Jr, EC Rhea eds. Atmospheric Corrosion tests of copper
and copper alloys in Sweden16 Year Results. Philadelphia: American Society for
Testing and Materials, 1982, pp. 106115.
39. PT Gilbert, J Inst Metals 73:139174, 1947.
40. KH Logan, M Romanoff. J Res Natl Bur Stand 33:172, 1944.
41. LA Denison. Trans Electrochem Soc 81:111, 1942.
42. JR Meyers, A Cohen. JAWWA 76:68, 1984.
43. A Cohen, AJ Brock. Mater Perform 34:5157, 1995.
44. E Mattsson. Br Corros J. 15:613, 1980.
45. VF Lucey. Br Corros J. 2:175185, 1967.
46. A Cohen, WS Lyman. Service experience with copper plumbing tube. Corrosion
71. Houston, TX: National Association of Corrosion Engineers, 1971, Paper 84.
47. MF OBrecht, WE Sartor, JM Keyes. Nodular pitting of copper tube by cold potable
water supplies. Sixth International Congress on Metallic Corrosion, Sydney, 1975.
48. VF Lucey. Pitting corrosion of copper in supply waters: The effect of water composition. B.N.F.M.R.A. Research Report No A1838, December 1972.
49. HS Campbell. J Inst Metals 77:345356, 1950.
50. FJ Cornwell, G Wildsmith, PT Gilbert. Pitting Corrosion in Copper Tubes in Cold
Water Service. Philadelphia: American Society for Testing and Materials, 1976, pp.
155179.
51. KD Efird, DB Anderson. Mater Perform 14:11, 1975.
52. FP Ijselling, JM Krougman. Proc. 6th European Congress on Metallic Corrosion,
London, 1977, p. 181.
53. JE Castle, DC Eppler, DB Peblow. Corros Sci 16:145157, 1976.
54. JE Castle, DC Eppler. Corrosion 35:451455, 1986.
55. Metals Handbook, 9th ed. Metals Park, OH: ASM International, 1987, p. 624.
56. WC Stewart, FL LaQue. Corrosion 8:259, 1952.
57. C Pearson. Br. Corros J 7:6168, 1972.
58. JJ Obrzut. Iron Age 3641, 1977.
59. FL Laque, WF Clapp. Trans Electrochem Soc 87, 1945.
60. KD Effird. Mater Perform 15:4, 1976.
61. JL Manzolillo, EW Thiele, AH Tuthill. CA-706 coppernickelalloy hulls: The cop-

150

62.
63.
64.

65.
66.
67.
68.
69.
70.
71.

Brock
per mariners experience and economics. Society of Naval Architects and Marine
Engineers, 1976.
JP Gudas, HP Hack. Corrosion 78. Houston, TX: National Association of Corrosion
Engineers, 1978, Paper 23.
TS Lee, HP Hack, DG Tipton. Proceedings of the Fifth International; Congress on
Marine Corrosion and Fouling, Barcelona, 1980.
S Sato, K Nagata, S Yamauchi. Evaluation of various protective measures against
corrosion of copper alloy condenser tubes by sea water. Corrosion 81. Houston,
TX: National Association of Corrosion Engineers, Toronto, 1981, Paper 195.
GP Sheldon, NW Polan. Field testing of power utility condenser tube alloys. J Mater
Energy Syst 6(4), 1985.
Metals Handbook, Volume 13, Corrosion, 9th ed. Metals Park, OH: ASM International, 1987, pp. 627633.
H Leidheiser. The Corrosion of Copper, Tin, and Their Alloys. New York: John
Wiley & Sons, 1971, pp. 107119, 135143.
HH Uhlig. Corrosion Handbook, New York: John Wiley & Sons, 1948.
RJ McKay, R Worthington. Corrosion Resistance of Metals and Alloys. New York:
Reinhold, 1936.
FN Speller. Corrosion, Causes and Prevention. New York: McGraw-Hill, 1951.
JA Lee. Materials of Construction for Chemical Process Industries. New York:
McGraw-Hill, 1950.

6
Reactive and Refractory Alloys
Te-Lin Yau
Te-Lin Yau Consultancy, Albany, Oregon

I.

INTRODUCTION

This chapter covers three metals each of Group IVB [titanium (Ti), zirconium
(Zr), and hafnium (Hf )] and Group VB [vanadium (V), niobium (Nb), and tantalum (Ta)] in the periodic table. Because of their similarities, these metals are
referred to as the following:
1. Reactive metals, as they are highly active in the electromotive force
(EMF) series. Practically, the reactivity of these metals allows them
to spontaneously form protective oxide films in air and makes them
corrosion-resistant materials.
2. Refractory metals, as their melting points are above the range of iron,
cobalt, and nickel.
Although these metals are similar in many ways, they also show very significant differences. The attractive strength-to-density ratio makes titanium an
important structural material in the aerospace industry. The contrast between zirconium and hafnium in thermal neutron absorption cross section makes them
complementary in nuclear applications. Vanadium is a vital alloying element to
make steel and titanium strong and tough. Niobium has hot and cold applications
because of its strength at elevated temperatures and superconducting characteristics at low temperatures. Tantalum is widely used in electrolytic capacitors resulting from the high dielectric constant of its surface oxide film.
In addition to these standout applications, materials and corrosion professionals often specify Ti, Zr, Nb, Ta, and their alloys to handle highly corrosive environments. Moreover, current trends in the chemical process industries favor the increas151

152

Yau

ing usage of these metals. These metals are highly corrosion resistant over wide
ranges of media. When properly used, these metals can realize high returns-oninvestments due to low maintenance and replacement costs, improved process efficiency, added values of high-quality products, and compliance with safety and environmental protection requirements. However, there are some striking differences
among these metals, too. It is essential to fully understand the capabilities of these
metals. Otherwise, any misuses may result in costly mistakes. Consequently, this
chapter discusses environmental effects on reactive and refractory metals.

II. GENERAL CHARACTERISTICS


Physical and some typical mechanical properties of unalloyed reactive and refractory metals are given in Table 1. These metals have high to very high melting
points and their densities range from low to high. Their coefficients of thermal
expansion are typically low. Their thermal conductivities are from better than
that of type 304 stainless steel to matching that of steel.
Their mechanical properties vary over a wide range depending on metallurgical states and impurities levels. The presence of interstitial impurities, such as
oxygen and nitrogen, are particularly pronounced in affecting mechanical properties. In fact, oxygen can be used to modulate mechanical properties.
It is important to properly specify any metal to match requirements in a
specific application. There are different grades of unalloyed reactive and refractory metals by varying impurity levels. Also, there are alloys developed for improved mechanical and/or corrosion properties. Readers should review the
ASTM specifications shown in Table 2 before specifying them.

III. ENVIRONMENTAL EFFECTS


Reactive and refractory metals are often regarded as materials that possess exceptional corrosion resistance in a wide range of environments. One can expect that
these metals are similar in corrosion resistance in many environments, such as
seawater. However, it is not always clear which metal is the right choice for a
specific application. One should consider these metals as friendly rivals that display contrasts. Very often, the selection is made based on cost and availability.
This approach may work simply because these metals are overlapping in corrosion properties. However, this approach may result in costly mistakes, too. It is
important to fully understand their corrosion properties. Candidates can be selected by matching the capabilities of these metals to the characteristics of the
environments. Then, the optimal choice can be made by considering other factors,
such as cost, requirements of product quality, strength, and fabricability.

Physical and Mechanical Properties at Room Temperature

Property
Atomic number
Melting point (C)
Density (g/cm3)
Coefficient of thermal expansion (106 /C)
Thermal conductivity (W/mC)
Modulus of elasticity (GPa)
Tensile strength (MPa)
Yield strength (MPa)
Stress-relieving temperature (C)

Ti

Zr

Hf

Nb

Ta

22
1677
4.51
8.9
17
52
240655
170720
480

40
1857
6.51
5.89
22
99
165440
170310
540

72
2227
13.3
5.9
22.3
135
440
225
650

23
1917
6.11
9.4
31
120
200500
100200
900

41
2468
8.57
7.1
523
188
170500
75200
800

73
2996
16.69
6.6
544
185
170285
100170
900

Reactive and Refractory Alloys

Table 1

153

154

Yau

Table 2

ASTM Specifications

Products

Ti

Zr

Hf

Nb

Ta

Plate
Pipe
Tube
Wire
Ingots
Castings
Fittings
Forgings

B-265
B-337
B-338
B-348

B-367
B-363
B-381

B-551
B-658
B-523
B-550
B-495
B-752
B-653
B-493

B-776

B-737

B-393

B-394
B-392
B-391

B-708

B-521
B-365
B-364

Table 3 gives a general comparison on the corrosion resistance of reactive


and refractory metals. Hafnium and vanadium are omitted because they are not
being used in the chemical process industries. The major purpose of this comparison is to show that there are major differences among them. It should be noted that
the magnitude of the differences depends on metallurgical and chemical factors.
A. Titanium
Titanium is most popular among all reactive and refractory metals. In fact, Ti is
the fourth most abundant structural metal in the world. The once exotic image

Table 3

General Comparison of the Resistance of Reactive Metals

Condition

Ti

Zr

Nb

Ta

Hydrochloric acid
Sulfuric acid
Nitric acid
Oxidizing acids w/o Cl
Oxidizing acids with Cl
Acids with F
Caustics
Hydrogen peroxide
Dry chlorine
Wet chlorine
Ignition in oxygen
Abrasives

Excellent
Very good
Excellent
Excellent
Cautious
Poor
Excellent
Excellent
Very good
Cautious
Very good
Very good

Good
Good
Excellent
Excellent
Excellent
Fair
Poor
Fair
Excellent
Excellent
Excellent
Good

Very good
Excellent
Excellent
Excellent
Excellent
Poor
Poor
Fair
Excellent
Excellent
Excellent
Good

Poor
Poor
Very good
Very good
Excellent
Poor
Poor
Poor
Poor
Excellent
Good
Good

Ranking (excellent being the best and poor being the worst): Excellent, very good, good, fair, and
poor. Cautious means that other factors such as surface condition is important.

Reactive and Refractory Alloys

155

of Ti is long gone. Titaniums popularity can be attributed to its several attractive


properties, including competitive cost, good corrosion resistance, high availability, superior structural efficiency, and great fabricability.
In addition to the well-known aerospace applications, Ti is widely used
in industrial, marine, and commercial applications that include pulp and paper,
desalination plants, dental materials, jewelry, architectural, marine, and sporting
equipment, medical implants, automotive, electrochemical anodes, food, brewing, pharmaceutical, flue gas desulfurization, steam turbines, petrochemical refineries, nuclear waste storage, metal extraction equipment, and cookware.
By itself, Ti is not suitable for most of these applications because of inadequate corrosion resistance, strength, and formability. However, unlike other reactive and refractory metals, Ti can be readily alloyed to modify its properties.
Basically, there are three types of titanium alloy:
1. Alpha alloys are non-heat-treatable and are very weldable. They possess low to medium strength, good notch toughness, adequate ductility,
and excellent cryogenic mechanical properties. The more highly alloyed alpha and near-alpha alloys offer improved high-temperature
creep strength and oxidation resistance.
2. Alpha-beta alloys are heat treatable and most are weldable. They have
medium to high strength. Their hot forming properties are good. They
have lower high-temperature creep strength than most alpha alloys.
3. Beta and near-beta alloys are readily heat treatable and generally weldable. They are capable of achieving high strengths. They have good
creep strength to intermediate temperatures. Solution-treated alloys
have excellent formability.
In some cases, it takes less than 0.5% of alloying elements to significantly
improve Tis corrosion resistance. In other cases, it takes large amounts of alloying elements to substantially increase Tis strength. Also, technologies such as
powder metallurgy and superplasticity have been developed to fabricate Ti into
complicated components. It would be too extensive for this chapter to cover all
the details. Only some information relating to environmental effects will be discussed here.

1. Water and Steam


Titanium and its alloys are highly resistant to water, natural waters, and steam
to temperatures up to at least 316C. They may acquire a tarnished appearance
on their surfaces in hot water or steam due to the formation of a protective oxide
film. This is normal and causes no alarm.
Natural waters often contain contaminants, such as iron and manganese

156

Yau

oxides, sulfides, carbonates, and chlorides, that do not affect Tis corrosion resistance. Also, chlorination treatments used to control sliming and biofouling have
no adverse effects on Ti.

2. Seawater and Other Salt Solutions


Seawater is a complicated corrosive. It is difficult for common metals and alloys
to handle. Frequently encountered corrosion problems include general corrosion,
pitting, stress-corrosion cracking (SCC), microbiologically induced corrosion
(MIC), and erosion.
Titanium resists corrosion by seawater, regardless of chemistry variations
and pollution effects, to temperatures up to 260C. The compatibility between
Ti and seawater makes Ti a vital material in marine applications. Titanium equipment has provided reliable service for decades in the chemical, oil refining, and
desalination industries.
Titanium does not encounter the common problems like other metals and
alloys in seawater. It is practically immune to pitting, MIC, and stress-corrosion
cracking. However, some highly alloyed Ti alloys are susceptible to stresscorrosion cracking. Titanium can withstand seawater impingement and flow
velocities in excess of 30 m/s.
Similarly, titanium resists attack by chloride solutions and other brines over
the full concentration range with pH between 3 and 11. Oxidizing chloride solutions, such as ferric and cupric chlorides, chlorites, hypochlorites, chlorates,
perchlorates, and chlorine dioxide, extend titaniums resistance to lower pH
levels.
Nevertheless, titanium is not perfect in hot seawater and other chloride
solutions. One of the major concerns is its susceptibility to crevice corrosion
within tight physical crevices. It is affected by several, often interacting, factors,
including temperature, solution chemistry/pH, nature of the crevice, alloy composition, metal surface condition, and metal potential. One can apply one or more
measures for preventing crevice corrosion on Ti equipment and components.
These measures include selecting the right alloy (e.g., palladium-containing
alloys), noble metal surface treatment, pickling for smeared surface iron particles,
and avoiding incompatible gaskets/sealants.
Another major concern is galvanic corrosion leading to hydrogen embrittlement. Normally, titanium is the cathode when it is in contact with common
metals, such as steel and aluminum. Titanium does not corrode, but the coupling
metal becomes the anode and experiences accelerated corrosion. Consequently, the excessive hydrogen generated resulting from the contact may induce
hydrogen embrittlement in titanium. To avoid galvanic corrosion, it is important
to use two metals that are close in the galvanic series. Other preventative

Reactive and Refractory Alloys

157

measures include insulating joints, coatings, and cathodically protecting the


anode.

3. Inorganic Acids
Titanium is regarded as a resistant metal in oxidizing acids, such as nitric and
chromic acids, over a wide range of concentrations and temperatures. It is fair
in mildly reducing acids, such as sulfurous acid, but is rather poor in strongly
reducing acids, such as sulfuric, hydrochloric, hydrobromic, and phosphoric
acids. Still, titanium is not as corrosion resistant as zirconium and tantalum to
nitric acid. Titanium resists nitric acid over a wide range of concentrations at
temperatures below boiling temperatures. At boiling and above, titaniums corrosion resistance is very sensitive to nitric acid purity. Generally, titanium is not
corrosion resistant in pure nitric acid. Similarly, titanium may be attacked in the
vapor of nitric acid where condensates may form. The higher the metallic ion
content of the acid, the better titanium will perform. In particular, the corrosion
of titanium to produce small amounts of titanium ions will result in the inhibitive
effect on titaniums corrosion in nitric acid. Therefore, titaniums corrosion may
decrease rapidly in nitric acid under a closed system. Unlike stainless steels,
titanium is uniquely suitable for handling recycled nitric acid. Consequently, titanium has many commercial applications.
When titanium experiences corrosion problems in nitric acid, the problems
cannot be solved by switching to normally more resistant titanium alloys, such
as Pd-containing alloys. These types of alloys are useful in improving Tis resistance in reducing acids. Because nitric acid is not reducing but oxidizing, the
switch will not offer any improvement. Moreover, titanium should not be considered for handling red fuming nitric acid because of the danger of pyrophoric
reactions.
Titanium has limited usefulness in strongly reducing acids (e.g., up to about
7% hydrochloric or sulfuric acid at room temperature). The resistance decreases
rapidly with increasing temperature. It improves when the acids contain small
amounts of oxidizing impurities, such as ferric ions or chlorine. For example,
the addition of 2 g/L ferric chloride will reduce the corrosion rate of Ti in 3%
HCl at boiling from 14 mm/year to less than 0.01 mm/year. Fortunately, it is
common to have this type of impurities in industrial acids. Titanium does have
numerous industrial applications involving reducing acids, such as in the mining
industry.
Titanium can have various corrosion problems in handling reducing acids,
such as crevice corrosion and vapor-phase attack or when the concentration gets
too strong. The addition of up to 0.25% Pd to Ti will significantly improve Tis
resistance in reducing acids. Traditionally, the Pd content is controlled in the

158

Yau

0.15% range. Still, this addition greatly increases the cost. Recently, new grades
of Ti alloys have been added by just having 0.05% Pd for improved corrosion
resistance with a minimum increase in cost.

4. Organic Solutions
Organic acids normally are mildly reducing. Titanium shows good corrosion resistance in most organic acids and other chemicals. It would be vulnerable to
corrosion in strong organic acids, such as formic, oxalic, sulfamic, and trichloroacetic acids. Generally, the presence of oxygen due to aeration and water presence
improves Tis resistance in organic media.
On the other hand, certain anhydrous organic media may attack titanium.
Without oxygen, it would be difficult for Ti to maintain its passivity. For example,
dry methyl alcohol can cause SCC in titanium, probably due to the breakdown
of passive film by halides. Once the passive film is broken down, there will not
be any repair if there is not any oxygen or water. Indeed, the combination of the
absence of water and the presence of halogens or halides is the major reason that
titanium experiences corrosion problems in organic solutions. The addition of
1.5% water can suppress titaniums susceptibility to SCC in methyl alcohol.
Another major problem for Ti in organic media is its susceptible to hydrogen embrittlement. Under reducing conditions, Ti will slowly absorb hydrogen
even when the corrosion rate is very low.

5. Alkalis
Titanium resists most alkalis except hot, strongly alkaline solutions. The major
problem is the excessive hydrogen uptake and eventual embrittlement of titanium
at temperatures above 80C when the pH is at or above 12. The presence of a
strong oxidizer, such as chlorine, makes Ti highly suitable for processing alkalis.
In fact, Ti is a useful structural material in the dual production of caustic soda
and chlorine by an electrochemical process.

6. Gases
Titanium has excellent resistance to air and oxygen at temperatures up to 370C.
Above this temperature but below 450C, titanium may form colored surface
films that thicken slowly with time. Above 650C, titanium will become brittle
due to poor oxidation resistance. Scales form rapidly at 930C. Because the oxidation is an exothermal reaction, titanium may ignite in pressurized oxygen under
a confined condition.
Nitrogen reacts much more slowly with titanium than oxygen. It reacts with
nitrogen to form a gold-colored film starting at 540C. Above 800C, excessive

Reactive and Refractory Alloys

159

diffusion of the nitride will occur and may cause metal embrittlement. Properly
formed nitride layer can enhance Tis resistance to abrasives.
The surface oxide film can protect Ti from absorbing hydrogen at temperatures below 80C. Absorption of several hundred parts per million (ppm) of hydrogen results in embrittlement and the possibility of cracking under conditions
of stress. The presence of moisture in hydrogen (e.g., as low as 2%) can effectively reduce hydrogen uptake.
Titanium is resistant to corrosion by sulfur dioxide and water-saturated
sulfur dioxide. Water-saturated sulfur dioxide may form the highly corrosive sulfurous acid that does not affect titanium. This capability allows Ti to be used in
various sulfur dioxide scrubber systems.
Titanium resists attack by wet and dry hydrogen sulfide. It is well known
that hydrogen can induce hydrogen embrittlement on many metals and alloys.
This possibility exits in Ti, too. Unlike most metals and alloys, Ti does not become brittle in wet hydrogen sulfide. However, in galvanic couples with certain
metals, such as steel, the presence of hydrogen sulfide in an aqueous solution
will promote hydriding in titanium.
Titanium is among the most resistant metal in wet chlorine and other chlorine chemicals because of their strongly oxidizing natures. This has been the
primary factor for using Ti in industrial service. Titanium equipment has been
relatively free of corrosion problems for decades. However, titanium is incompatible with dry chlorine that can cause a rapid attack of Ti and may even cause
ignition. As little as 1% water is sufficient for repassivation after mechanical
damage to Ti in chlorine gas under static conditions at room temperature.

7. Liquid Metals
Titanium has good resistance to many liquid metals at moderate temperatures.
It has been used in processing molten aluminum. Rapidly flowing molten aluminum, however, will erode titanium. Also, some liquid metals, such as mercury
and cadmium, can cause SCC in titanium.
B. Zirconium
Zirconium has the reputation of being one of the most corrosion-resistant metals.
It has a very strong affinity for oxygen. It is one of very few metals that even can
react with oxygen in water under highly reducing conditions to form an adherent,
protective oxide film on its surface. This film is self-healing and protects the base
metal from chemical and mechanical attack at temperatures to 350C.
Many engineering metals, such as iron, nickel, chromium, and titanium,
produce metal ions of a variable valency. Uniquely, zirconium is predominantly
quadrivalent in its oxides and many other compounds. It forms very few unstable

160

Yau

compounds in which its valency is other than 4. The chemistry of zirconium is


characterized by the difficulty of achieving an oxidation state less than 4. This
character, along with high oxygen affinity, allows zirconium to form protective
oxide films even in highly reducing media, such as hydrochloric acid and dilute
sulfuric acid. Under these conditions, common metals and alloys may form subordinate oxides or other compounds of low or no protective capability.
Moreover, metal ions of a constant valency imply their stability. This is
an important requirement in many applications. For example, zirconium can
maintain the stability of certain chemicals, such as hydrogen peroxide. Ions of a
variable valency are the common decomposition catalysts for hydrogen peroxide.
Another advantage is that zirconium ions are colorless. This is important when
the color stability of products is a major concern. Most transition metals produce
ions of different colors depending on their valency state.
Protective oxide films are difficult to form on zirconiums surface in a few
media, such as hydrofluoric acid, concentrated sulfuric acid, and oxidizing chloride solutions. Consequently, zirconium is not suitable or needs protection measures for handling these media.

1. Water and Steam


Zirconium has excellent corrosion and oxidation resistance in water and steam
at temperatures exceeding 300C. Zirconium has a great capability to take oxygen
from water for the formation of a protective oxide film. This capability is not
weakened even when zirconium is in a highly reducing medium. Most passive
metals form protective oxide films in aqueous solutions only when the solutions
are somewhat oxidizing. Consequently, zirconium is uniquely suitable for nuclear
applications because water-cooled reactors operate with an oxygen- or hydrogencharged coolant at temperatures from 280C to 300C.
However, corrosion and oxidation of unalloyed zirconium in high-temperature water and steam were found to be irregular. This behavior is related to variations in the impurity content in the metal. Nitrogen and carbon impurities are
particularly harmful. The oxidation rate of unalloyed zirconium increases markedly when nitrogen and carbon concentrations exceed 40 and 300 ppm, respectively. The irregular behavior of unalloyed zirconium stimulated alloy development programs. Zircaloys (Zr1.5% Sn-based alloys), Zr2.5 Nb and Zr1 Nb
are the most important ones developed for nuclear applications because they are
more reliable and predicable for use in hot water and steam in addition to being
stronger.
As compared to unalloyed zirconium, zircaloy-2 has an improved character
in oxide formation at elevated temperatures. A tightly adherent oxide film forms
on this alloy at a rate that is at first quasicubic but undergoes a transition to linear

Reactive and Refractory Alloys

161

behavior after an initial period. Unlike the white, porous oxide films on unalloyed
zirconium, the oxide film on zircaloy-2 remains dark and adherent throughout
transition and in the posttransition region.
Zircaloy-4 differs in composition from zircaloy-2 in having a slightly
higher iron content but no nickel. Both variations are intended for reducing hydrogen pickup with little effect on corrosion resistance in reactor operation. For
example, in water at 360C, hydrogen pickup for zircaloy-4 is about 25% of
theoretical, or less than half that for zircaloy-2. In addition, hydrogen pickup for
zircaloy-4 is less sensitive to hydrogen overpressure than that for zircaloy-2. For
both alloys, hydrogen pickup is greatly reduced when dissolved oxygen is present
in the medium.
Zr2.5 Nb is considered to be somewhat less resistant to corrosion than
the zircaloys, with exception. Nevertheless, Zr2.5 Nb is suitable for many applications, such as pressure tubes in the primary loops of some reactors. Furthermore, the corrosion resistance of Zr2.5 Nb can be substantially improved by
heat treatments. Also, the Zr2.5% Nb alloy is superior to zircaloys in steam at
temperatures above 400C.

2. Saltwater
Zirconium has excellent corrosion resistance to seawater, brackish water, and
polluted water. Zirconiums advantages include its insensitivity to variation in
factors like chloride concentration, pH, temperature, velocity, crevice, and sulfurcontaining organism. Some of the results are summarized as follows.
Zirconium specimens with or without a crevice attachment were placed in
the Pacific Ocean at Newport, Oregon, for up to 129 days. All welded and nonwelded specimens exhibited nil corrosion rates. Marine biofouling was observed;
however, no attack was found beneath the marine organisms or within the crevices.
Laboratory tests were performed on Zr 702 (unalloyed Zr) and Zr 704 (nonnuclear grade of zircaloys) in boiling seawater for 275 days and in 200C seawater for 29 days. Both alloys were resistant to general corrosion, pitting, and crevice corrosion. Tests of U-bend specimens, with or without steel coupling, of Zr
702, nickel-containing Zr 704, and nickel-free Zr 704 were conducted in boiling
seawater for 365 days. No cracking was observed during the test period. Overstressing of the tested U-bends indicated that all specimens remained ductile except for the welded nickel-containing Zr 704 with steel coupling. Steel-coupled
nickel-containing Zr 704 showed much higher hydrogen and oxygen absorption
and formed hydrides. Chemical analyses and metallographic examinations on
other U-bends did not show evidence of hydrogen absorption and hydride formation.

162

Yau

3. Halogen Acids
Zirconium resists attack by all halogen acids, except hydrofluoric acid (HF). HF
vigorously attacks it at all concentrations. Acidic fluoride solutions are highly
corrosive to zirconium, too. Zirconiums corrosion resistance is not as poor in
fluoride salt solutions as in HF until fluorides become HF in pH 3 solutions.
This fact is taken advantage of in preparing zirconium surfaces using mixtures
of hydrofluoric and nitric acids for various fabrication steps and for improved
corrosion resistance in certain nuclear and chemical applications.
In recent years, it appears that the chance to have fluorides in the process
media has increased somewhat. One of the possibilities is the increased usage of
recycled chemicals. For example, recycled sulfuric acid may contain more than
100 ppm fluorides. When zirconium equipment faces fluoride-containing acids,
inhibitors that form strong fluoride complexes should be added for protecting
zirconium equipment. Effective inhibitors include zirconium nitrate, zirconium
sulfate, and phosphorus pentoxide.
The other halogen acids [i.e., hydrochloric (HCl), hydrobromic (HBr), and
hydriodic (HI) acids] do not attack zirconium. Yet, one of the most impressive
corrosion properties for zirconium is its excellent resistance in HCl at all concentrations and temperatures even above boiling. Because of its strong reducing
power, it is very difficult for most metals to form protective oxide films in HCl.
The presence of even a small amount of HCl in a medium may cause common
metals and alloys to suffer general corrosion, pitting, and SCC.
Zirconium is suitable for handling HCl at all concentrations. Moreover,
zirconium is not as susceptible to hydrogen embrittlement in HCl as tantalum is.
For example, tantalum became brittle when tested in 11M HCl and 11M HCl
7% GaCl3 for 1000 h at 70C. Under the same conditions, zirconium remained
unattacked and retained 100% of its ductility.
Zirconium is susceptible to localized corrosion, such as pitting, intergranular corrosion, and SCC when it is anodically polarized. Zirconium is susceptible
to pitting in 20% HCl, but to intergranular corrosion in 20% HCl. The same
types of corrosion problem may be developed in HCl when highly oxidizing ions,
such as ferric and cupric ions, are present. The presence of ferric ions may polarize the zirconium surface to a potential exceeding the breakdown potential. Thus,
a local breakdown of the passive surface at preferred sites occurs, and the condition favors the occurrence of localized corrosion. To eliminate preferred sites by
pickling zirconium in a mixture of hydrofluoric and nitric acids can suppress the
breakdown process of passive films. Alternatively, maintaining zirconium at a
potential in its passive region, which is arbitrarily set at 50100 mV below the
corrosion potential, can counteract the detrimental effects resulting from the presence of ferric ions.

Reactive and Refractory Alloys

163

4. Nitric Acid
Nitric acid (HNO3), because of its oxidizing power, is not considered to be a
difficult acid for passive metals to handle. Nevertheless, HNO3 becomes highly
corrosive when its temperature is high, its concentration is too high or not high
enough, or its purity is poor. The oxidizing power favors the formation of oxide
films, but it may also cause the passive films to break down.
Zirconium is considerably more suitable than most passive metals for handling HNO3, particularly when the acid is hot, impure, and/or variable in concentration. In certain conditions, zirconium is even more resistant than the noble
metals to the acid. Zirconiums temperature limit is somewhat higher than that
of noble metals. Traces of chloride may lead to rapid attack of noble metals, but
not of zirconium.
The excellent corrosion resistance of zirconium in HNO3 has been recognized for more than 30 years. Zirconium resists nitric acid up to the boiling point
and at 98% HNO3, and up to 250C and at 70% HNO3. Moreover, the corrosion
rates are nil in boiling 3070% HNO3 with up to 1% FeCl3, 1% NaCl, 1% seawater, 1% iron, or 1.45% type 304 stainless steel at 205C. These results indicate
that the presence of heavy metal ions and chlorides has little effect on the corrosion resistance of zirconium.
Zirconium is normally susceptible to pitting in oxidizing chloride solutions.
However, the NO3 ion is an effective inhibitor for the pitting of zirconium. The
minimum [NO3]/[Cl] molar ratio required to inhibit pitting of zirconium was
determined to be 15. Still, the presence of an appreciable amount of HCl should
be avoided because zirconium is not resistant to aqua regia.
The slow strain-rate technique can reveal zirconiums SCC susceptibility
in HNO3. The primary concern for using zirconium in HNO3 service would be
SCC in concentrated acid. Zirconium seems to be resistant to SCC in concentrated
acids when it is stressed below the yield point.
Other concerns include the accumulation of chlorine gas in the vapor phase
and the presence of fluoride ions. Chlorine gas can be generated by the oxidation
of chlorides in HNO3. Areas that can trap gases should be avoided when Cl is
present in HNO3; or, zirconium equipment should be pickled for much improved
resistance to pitting in wet-chlorine-containing vapors. As indicated previously,
the corrosion of zirconium in fluoride-containing acids can be controlled by adding an inhibitor, such as zirconium compounds, to convert fluoride ions into noncorrosive complex ions.

5. Sulfuric Acid
Sulfuric acid (H2SO4) is the most important acid for use in the manufacture of
many chemicals. For example, it is used as a dehydrating agent, an oxidizing

164

Yau

agent, an absorbent, a catalyst, a reagent in chemical syntheses, and so on. The


consumption of sulfuric acid indicates a nations industrial activity. These highly
versatile capabilities can be attributed to the complicated nature of this acid.
Dilute solutions are reducing, which make passive metals vulnerable to corrosion.
In fact, hot, dilute solutions can be used to pickle steel and stainless steel. Solutions become increasingly oxidizing at or above 65%. The usefulness of common
metals depends strongly on acid concentration, temperature, and the presence of
other chemicals.
Zirconium resists attack by H2SO4 at all concentrations up to 70% and at
temperatures to boiling and above. In 7080% H2SO4, the corrosion resistance
of zirconium depends strongly on temperature. In higher concentrations, the corrosion rate of zirconium increases rapidly with concentration.
In the range in which zirconium shows corrosion resistance in H2SO4, a
passive film is formed on zirconium that is predominantly cubic zirconium oxide
(ZrO2) with only traces of the monoclinic phase. Zirconium corrodes in highly
concentrated H2SO4 (e.g., 80%) because loose films are formed that prove to
be zirconium disulfate tetrahydrate [Zr(SO4)2 4H2O]. Also, at the higher acid
concentrations, films that flake off are formed and are probably partly zirconium
hydrides.
Zirconium cannot tolerate much of strong oxidizing agents in 40%
H2SO4. However, in 40% H2SO4, zirconium can tolerate large amounts of strong
oxidizing agents. Consequently, zirconium equipment is often used in steel pickling.
The presence of chlorides in H2SO4 has little effect on the corrosion resistance of zirconium unless oxidizing agents are also present. Therefore, in the
presence of oxidizing agents, chloride ions should be controlled within a limit
to avoid detrimental attack.
Zirconium weld metal may corrode preferentially when H2SO4 concentration is approximately 55% and higher. Heat treatment at 775 15C for 1 h per
25.4 mm of thickness can be applied to restore the corrosion resistance of the weld
metal to that of the parent metal. However, this high-temperature heat treatment is
not suitable for equipment made of zirconium/steel clad materials because of the
large difference in thermal expansion coefficients between these two alloys. Heat
treatment at a much lower temperature, such as 425530C, should be applied
when there is a concern for SCC. Zirconium is susceptible to SCC in a narrow
range of H2SO4, (i.e., 6469%).
For zirconium equipment, it is very important to maintain acid concentration within the applicable limit. When the limit is exceeded, zirconium may corrode rapidly. In 65% H2SO4, the vapor phase is almost entirely water. However,
the concentration change is negligible when a system is under a pressurized condition. Acid concentration may change significantly when, for example, the sys-

Reactive and Refractory Alloys

165

tem is imperfectly sealed. In a leaking system, the acid concentration can exceed
the concentration limit. Acid concentration can easily increase when the system
is under vacuum operation because water vapor is continuously taken away.
When the corrosion resistance limits of zirconium in H2SO4 are exceeded,
a pyrophoric surface layer may be formed on zirconium under certain conditions.
The pyrophoric surface layer on zirconium formed in 77.5% H2SO4 200 ppm
Fe3 at 80C consists of -hydride, ZrO2, Zr(SO4)2, and fine metallic particles.
The combination of hydride and fine metallic particles is suggested to be responsible for the pyrophoricity. Treating in hot air or steam can eliminate this tendency.

6. Phosphoric Acid
Phosphoric acid (H3PO4) is less corrosive than other mineral acids. Many stainless
alloys demonstrate useful resistance in the acid at low temperatures. As often
occurs, corrosion rates of common alloys in the acid increase with increasing
temperature, concentration, and impurities. Areas like the liquid-level line or the
condensing zones are particularly vulnerable to attack.
Zirconium resists attack in H3PO4 at concentrations up to 55% and temperatures exceeding the boiling point. Above 55% H3PO4, the corrosion rate could
increase substantially with increasing temperature. The most useful area for zirconium would be dilute acid at elevated temperatures. Zirconium outperforms common stainless alloys in this area.
If H3PO4 contains more than a trace of fluoride ions, attack on zirconium
may occur. Because fluoride compounds are often present in the ores for making
H3PO4, the use of zirconium has always been questioned. However, because P2O5
is an effective fluoride inhibitor and is usually present in large amounts in H3PO4
processes, tests should be conducted to determine zirconiums suitability in the
actual stream. Furthermore, zirconium compounds can be used to complex fluorides.

7. Other Acids
Zirconium has excellent corrosion resistance in up to 30% chromic acid at temperatures to 100C. It is not suitable for handling chrome-plating solutions that
contain fluoride catalysts.
Zirconium is also resistant to some mixed-acid systems. It can be in acid
mixtures of sulfuricnitric, sulfurichydrochloric, and phosphoricnitric. The
sulfuric acid concentration must be below 70%. Zirconium is aggressively attacked in 1:3 volume mixtures of nitric and hydrochloric acids (aqua regia). In
the 1:1 volume mixture, zirconium is attacked but much slower than in the 1:3
mixture. In mixtures greater than the 3:1 ratio, zirconium is resistant.

166

Yau

8. Alkalis
Zirconium resists attack in most alkalis, which include sodium hydroxide, potassium hydroxide, calcium hydroxide, and ammonium hydroxide. This makes zirconium distinctly different from other highly corrosion-resistant materials, such
as titanium, tantalum, graphite, glass, and polytetrafluoroethylene (PTFE).
Zirconium U-bend specimens have been tested in boiling 50% NaOH. During the test period, the concentration changed from 50% to about 85% and the
temperature increased from 150C to 300C. The PTFE washers and tubes used
to make the U-bends dissolved. However, the zirconium U-bends remained ductile and did not show any cracks after 20 days. It should be noted that stainless
steel is susceptible to SCC in alkaline solutions including NaOH solutions.
Zirconium coupons were tested in a white-liquor, paper-pulping solution,
which contained NaOH and sodium sulfide, at 120C, 175C, and 225C. All
coupons showed nil corrosion rates. In the same solution, graphite and glass both
corroded badly at 100C. Zirconium also exhibits excellent resistance to SCC in
simulated white liquors.

9. Salt Solutions
Zirconium is resistant to most salt solutions, which include halide, nitrate, carbonate, and sulfate salts. Corrosion rates are usually very low at temperatures at least
up to the boiling point. Solutions of strong oxidizing chloride salts, such as FeCl3
and CuCl2, are examples of the few exceptions. In strong oxidizing chloride solutions, zirconiums performance is very dependent on surface condition. Zirconium becomes quite resistant to pitting when it has a good surface finish, like
the pickled surface.
Although zirconium has good corrosion resistance in sodium fluoride and
potassium at low temperatures, resistance decreases rapidly with increasing temperature or decreasing pH. Consequently, zirconium is not ideal for handling
most fluoride-containing solutions unless fluoride ions are complexed.
Zirconium is considerably more resistant to chloride SCC than stainless
steels are. No failure was observed in U-bend tests conducted in boiling 42%
magnesium chloride (MgCl2). Another attractive property of zirconium is its high
crevice corrosion resistance. Zirconium is not subject to crevice corrosion even
in acidic chloride at elevated temperatures. No attack was observed on zirconium
in a salt spray environment.
Unlike many common metals, zirconium has very little affinity for sulfur.
Zirconiumsulfur compounds form only at temperatures above 500C. Furthermore, there is no instance of zirconiumsulfur bonds forming in aqueous systems.
Hence, hydrogen sulfide (H2S), which is highly corrosive to common metals and
alloys, is not expected to participate in the corrosion reactions of zirconium in
sulfide-containing solutions. Zirconium coupons and U-bends were tested in nu-

Reactive and Refractory Alloys

167

merous NaClH2S solutions at temperatures to 232C. No general corrosion, pitting, crevice corrosion, or SCC was observed.

10. Organic Solutions


Zirconium has excellent corrosion resistance in most organic solutions, except
certain chlorinated compounds. It has been extensively tested in organic-cooled
reactors where the coolant consisted of mixtures of high-boiling aromatic hydrocarbons (e.g., terphenyls). These coolants are noncorrosive to zirconium. However, early experiments in the organic coolants indicated that hydriding was a
major concern. It was found that chlorine impurity in the organic coolants was
the major cause of gross hydriding. Elimination of the chlorine and maintenance
of a good surface oxide film by ensuring the presence of adequate water (50
ppm) alleviates the hydriding problem.
Indeed, the combination of the absence of water and the presence of halogens or halides is the major reason why zirconium experiences corrosion problems in organic solutions. For example, the addition of some water can suppress
zirconiums susceptibility to SCC in alcohol solutions with halide impurities. On
the other hand, zirconium shows excellent corrosion resistance in certain chlorinated carbon compounds (e.g., carbon tetrachloride and dichlorobenzene) at temperatures up to 200C.
From a corrosion point of view, organic halides can be classified into three
groups: water soluble, water insoluble, and water incompatible.
Water-soluble halides, such as aniline hydrochloride, chloroacetic acid, and
tetrachloroethane, are not corrosive to zirconium. They may become more corrosive when the water content is low and/or zirconium is highly stressed. More
active halides, such as dichloroacetic and trichloroacetic acids, are more corrosive
to zirconium. It is suspected that these halides may attack zirconium or intermetallic compounds at grain boundaries to form organometallic compounds. It
should be noted that certain organic compounds, such as alkyl and aryl halides,
are the common ones that react with most metals, including noble metals, to
form organometallic compounds. These reactions can be suppressed when there
is some water present in the media. Consequently, water addition and/or stress
relieving would be effective in preventing the corrosion of zirconium in watersoluble halides. However, water addition may increase the corrosivity of many
organic solutions toward common metals and alloys, but it seems to be always
beneficial to zirconium.
Water-insoluble halides, such as trichloroethylene and dichlorobenzene, are
not corrosive to zirconium, probably because of their stability. They will not
dissolve in water and they will not exclude water, either. They and water can be
physically mixed together.
Water-incompatible halides, such as acetyl chloride, may be highly corro-

168

Yau

sive to zirconium. They are not stable. They react violently with water. There is
no chance for water to be present in this type of halides, which are the most
undesirable organics for zirconium, and possibly other metals, to handle.

11. Gases
Zirconium forms a visible oxide film in air at about 200C. The oxidation rate
becomes high enough to produce a loose, white scale on zirconium at temperatures above 540C. At temperatures above 700C, zirconium can absorb oxygen
and become embrittled after prolonged exposure.
Zirconium reacts more slowly with nitrogen than with oxygen because it
has a higher affinity for oxygen than for nitrogen and it is normally protected by
a layer of oxide film. Once nitrogen penetrates through the oxide layer, it diffuses
into the metal faster than oxygen because of its smaller size. Clean zirconium
starts the nitride reaction in ultrapure nitrogen at about 900C. Temperatures of
1300C are needed to fully nitride the metal. The nitriding rate can be enhanced
by the presence of oxygen in the nitrogen or on the metal surface.
The oxide film on zirconium provides an effective barrier to hydrogen absorption up to 760C, provided that small amounts of oxygen are also present
in hydrogen for healing damaged spots in the oxide film. In an all-hydrogen
atmosphere, hydrogen absorption will begin at 310C.
Zirconium will ultimately become embrittled by forming zirconium hydrides when the limit of hydrogen solubility is exceeded. Hydrogen can be effectively removed from zirconium by prolonged vacuum annealing at temperatures
above 760C.
The corrosion resistance of zirconium and its alloys in steam is of special
interest to nuclear power applications. They can be exposed for prolonged period
without pronounced attack at temperatures up to 425C. In the 360C steam, up
to 350 ppm chloride and iodide ions, 100 ppm fluoride ions, and 10,000 ppm
sulfate ions are acceptable for zirconium in general applications but not in nuclear
power applications.
Zirconium is stable in NH3 up to about 1000C, in most gases (CO, CO2,
and SO2) up to about 300400C and in dry halogens up to about 200C. At
elevated temperatures, zirconium forms volatile halides. Depending on the surface condition, zirconium may or may not be resistant in wet chlorine. Zirconium
is susceptible to pitting in wet chlorine unless it has a properly cleaned surface.

12. Molten Salts and Metals


Zirconium resists attack in some molten salts. It is very resistant to corrosion by
molten sodium hydroxide to temperatures above 1000C. It is also fairly resistant
to potassium hydroxide. The oxidation properties of zirconium in nitrate salts are
similar to those in air.

Reactive and Refractory Alloys

169

Zirconium resists some types of molten metals, but the corrosion resistance
is affected by trace impurities, such as oxygen, hydrogen, or nitrogen. Zirconium
has a corrosion rate of less than 25 m/year in liquid lead to 600C, lithium to
800C, mercury to 100C, and sodium to 600C. The molten metals known to
attack zirconium are zinc, bismuth, and magnesium.

C. Hafnium
Hafnium and zirconium are chemically very similar. They occur naturally in ores.
However, hafnium has a high thermal neutron absorption cross section in contrast to the very low one of zirconium. Because of this dramatic difference, they
have to be separated from each other for nuclear applications but not corrosionresistant applications.
Similar to zirconium, hafnium has excellent corrosion resistance to many
media, including hydrochloric, nitric, and nonconcentrated sulfuric acids and alkalis. In fact, hafnium is superior to zirconium in corrosion resistance in water,
steam, molten alkali metals, and air.
Hafnium is soluble is hydrofluoric and concentrated sulfuric acids and aqua
regia. It begins to react slowly with air or oxygen at about 400C and nitrogen
at about 900C and rapidly with hydrogen at about 700C.
An interesting capability for hafnium is the combination of high neutron
absorption and excellent corrosion resistance to nitric acid. This capability makes
hafnium a uniquely reliable material for use in spent nuclear fuel reprocessing
plants. Hafnium is being used not just for its corrosion resistance but also for
preventing criticality.

D. Vanadium
Compared to other reactive and refractory metals, vanadium is inferior in aqueous
corrosion resistance. There is no known application for vanadium in the chemical
process industries. There is no established ASTM specification for vanadium,
either.
Vanadium resists attack by oxygen, nitrogen, and hydrogen at ambient temperatures. It oxidizes in air at different temperatures to form various oxides (i.e.,
trioxide, tetroxide, and pentoxide). It reacts with chlorine at temperatures greater
than 180C.
Vanadium is resistant to seawater, reducing acids, such as hydrochloric and
dilute sulfuric acids, and to alkaline solutions. It is poor in oxidizing acids, such
as nitric and concentrated sulfuric acids, and hydrofluoric acid.
Vanadium has one important corrosion property: its resistance in liquid
metals including lithium and sodium. This, combined with its neutron economy

170

Yau

and high-temperature strength, makes vanadium a leading candidate for first wall
and blanket structures of liquidmetal cooled-fusion energy systems.
E.

Niobium

The corrosion properties of niobium are often compared to those of tantalum.


Like tantalum, niobium forms very stable oxide films under anodic conditions.
The films are high in dielectric constant and breakdown potential. These properties, coupled with its excellent electrical conductivity, have led niobium to be
used as a substrate for platinum-group metals in impressed-current cathodic protection anodes. Consequently, it would be great if one can cut costs and weights
by replacing tantalum with niobium in anticorrosion applications. However, in
most cases, niobium has its own unique capabilities and is not interchangeable
with tantalum in severe-corrosion-resistant applications.
Niobium resists most inorganic and organic acids, except hydrofluoric acid,
at all concentrations and temperatures below 100C. It is especially resistant under oxidizing conditions, such as strong sulfuric acid containing ferric or cupric
ions. It has excellent corrosion resistance in most salt solutions, except those that
hydrolyze to form alkalis.
At room temperature, niobium is resistant to sulfuric acid at all concentrations up to 95%. The corrosion resistance decreases rapidly with increasing temperature. Compared to tantalum, niobium is much less useful in sulfuric acid
applications.
Niobium is excellent in oxidizing acids like nitric acid. However, niobium
is susceptible to hydrogen embrittlement in reducing acids, such as hydrochloric
acid. Hydrogen embrittlement can be prevented by converting the reducing condition into an oxidizing condition. For example, niobium becomes corrosion resistance in mixtures of nitric and hydrochloric acids.
Like other reactive and refractory metals, niobium is not corrosion resistant
in hydrofluoric acid. However, niobium has an adequate resistance in certain
fluoride-containing contaminated acids. For example, niobium is suitable for handling chromium plating solutions when small amounts of fluorides are added as
a catalyst. Similarly, niobium can be used in the reprocessing of spent chromium
plating solutions. The presence of a fluorides makes tantalum unsuitable.
At ambient temperatures, niobium is resistant in alkaline solutions. However, at higher temperatures, niobiums corrosion rates may stay low in alkaline
solutions, but it may become brittle even in dilute solutions of sodium hydroxide
or potassium hydroxide.
Niobium oxidizes easily in air. The oxidation becomes noticeable as the
temperature approaches 200C. It becomes rapid when the temperature exceeds
500C. In pure oxygen, the attack is catastrophic at 390C. Oxygen diffuses freely
through the metal and this causes embrittlement. Still, within the temperature
limit, niobium is much more resistant to ignition than titanium in pure oxygen.

Reactive and Refractory Alloys

171

To reduce the cost, a TiNb alloy can be used in ignition-sensitive areas. For
example, the Ti45% Nb alloy has found applications that employ dilute sulfuric
acid and pressurized oxygen, as in the gold mining industry.
Below 100C, niobium is inert in most common gases, including bromine,
chlorine, hydrogen, nitrogen, and sulfur dioxide (wet or dry). The resistance to
wet and dry chlorine is particularly interesting in corrosion-resistant applications.
Niobium reacts with nitrogen above 350C, with steam above 300C, with chlorine above 200C, and with carbon dioxide, carbon monoxide, and hydrogen
above 250C.
In the absence of nonmetallic impurities, such as gases, niobium has excellent corrosion resistance in liquid metals. The temperature limits for niobium in
liquid metals are 510C for bismuth, 400C for gallium, 850C for lead, 1000C
for lithium, sodium, potassium, and sodiumpotassium alloys, 600C for mercury, 850C for thoriummagnesium eutectic, 1400C for uranium, and 450C
for zinc.
The excellent resistance of niobium in sodium vapor leads to the use of
the Nb1 Zr alloy as the end caps in high-pressure sodium vapor lamps used for
highway lighting. Furthermore, liquid metals are excellent heat-transfer media;
they are ideal for use in compact thermal systems, such as fast breeder reactors,
reactors for space vehicles, and fusion reactors.
F.

Tantalum

Tantalum has the reputation to be one of the most versatile corrosion-resistant


materials. Like glass, tantalum is inert in most inorganic and organic solutions
at temperatures to at least 150C. The exceptions to this include fluorides, oleum,
oxalic acid, and strong alkalis. In fact, tantalum parts are often used to repair
glass-lined equipment because of their close corrosion properties and thermal
expansion coefficients. Of course, unlike glass, tantalum can withstand thermal
shocks and has a thermal conductivity comparable to carbon steel.
As shown in Table 1, tantalum is quite low in strength at room temperature.
The strength decreases quickly with increasing temperature. For the strength requirement at elevated temperatures, a tantalum alloy, such as Ta2.5% W or Ta
10% W, can be used. These alloys are as corrosion resistant as tantalum in all
environments.
One of the most important corrosion properties for tantalum is its resistance
to the complicated sulfuric acid. Sulfuric acid changes from the highly reducing
character of dilute solutions to the strong oxidizing character of concentrated
solutions. Even many high-performance alloys only have very restrictive usefulness in sulfuric acid. It is unusual that tantalum resists sulfuric acid at all
concentrations up to 98% and temperatures to at least 200C. Yet, the Ta2.5%
W alloy has been shown to be even more corrosion resistant than tantalum in
concentrated sulfuric acid. Due to the presence of sulfur trioxide, tantalum cor-

172

Yau

rodes in fuming sulfuric acid even at room temperature. Tantalum equipment


would be reliable for use in processes such as the recovery of strong sulfuric
acid, the nitration, and thermal cracking of organics.
Tantalum is practically inert to nitric acid, including the fuming grade, and
temperatures up to 300C. Its excellent corrosion resistance, high melting point,
and high thermal conductivity make tantalum one of the most suitable materials
for handling fuming nitric acid. Fuming nitric acid is not just highly corrosive
but also strongly oxidizing for inducing ignition on metals.
Tantalum resists hydrochloric acid at all concentrations up to 190C, although above 25%, the corrosion rate starts to rise noticeably. In addition, the
entry of hydrogen in concentrated acid may cause embrittlement in tantalum.
However, the presence of small amounts of oxidizing impurities will greatly reduce hydrogen absorption to avoid hydrogen embrittlement. Also, hydrogen embrittlement can be avoided by embedding platinum particles onto tantalums surface in area ratios of 1: 1000 platinum to tantalum. The embedding method can
be plating or welding. The low hydrogen overvoltage and cathodic character of
platinum makes it very easy for atomic hydrogen to form molecular hydrogen.
Thus, absorption of hydrogen into the tantalum is prevented.
Tantalum is useful in handling phosphoric acid at concentrations up to 85%
and temperatures up to 200C, provided that the fluoride impurity is very low.
Commercial grades of phosphoric acid may contain more than 100 ppm fluorides.
When tantalum equipment is involved, the fluoride content should not exceed 5
ppm.
As discussed previously, some metals, including titanium and zirconium,
are vulnerable in organic halides in the absence of water and oxygen. Tantalum
seems to be much less vulnerable than most metals under the same conditions.
Without the interference of water and oxygen, organic halides may attack metals
by forming organometallic compounds. Maybe, because of its large atomic size,
it would be more difficult for organic halides to incorporate tantalum into forming
a new compound.
Tantalum is fairly resistant to dilute alkaline solutions. It is attacked by
concentrated alkaline solutions, even at room temperature. One may not consider
tantalum for caustic services. However, caution should be exercised when strong
alkaline solutions are used to clean tantalum equipment.
Tantalum is excellent for handling wet or dry halogens at temperatures up
to at least 250C. It, however, reacts with large numbers of gases, such as oxygen,
nitrogen, hydrogen, carbon, carbon monoxide, carbon dioxide and methane, at
temperatures above 300C. This ability makes tantalum ideal as a getter in vacuum systems and gas purification systems.
Generally, tantalum possesses good resistance to most liquid metals. Liquid
metals that attack tantalum include aluminum, magnesium, cadmium, and zinc.

7
Aluminum Alloys
N. J. Henry Holroyd
Luxfer Gas Cylinders, Riverside, California

I.

INTRODUCTION

The aim of this chapter is twofold: first, to briefly describe the various modes
of corrosion (time-dependent environment-induced degradation) suffered by aluminum and its alloys; second, to discuss and review the influence played by
surface films during these corrosion processes. This aspect has been chosen because although its importance is consistently highlighted in the published texts
on aluminum corrosion (14), little detailed discussion ensues. A good example
is the recently published book Corrosion of Aluminum Alloys (4) which reviews
many aspects of the corrosion of aluminum and its alloys. The detailed discussion
on the role of surface films during corrosion is limited to the statement, Aluminum owes its excellent corrosion resistance and its usage as one of the primary
metals of commerce to the barrier oxide film that is bonded strongly to its surface and to a page of text giving basic information on aluminum films. Hopefully, this chapter will contribute to addressing this position.

II. BACKGROUND
Aluminum and its alloys are currently used in a wide range of applications, including aerospace, automotive, building, electrical, marine, packaging, and transportation. A comprehensive summary of the specific alloys (wrought and cast alloys)
used for different applications has recently been published (4). Reasons why a particular alloy is selected are application dependent. For instance, for structural appli173

174

Holroyd

cations, a potential user requires an alloy to provide an attractive strength-to-weight


ratio, the ability to be joined using welding, adhesive bonding, and so forth and
an acceptable corrosion resistance, all to be available at a competitive cost. For a
nonstructural application such as a lithographic sheet, the users requirements are
more focused on the uniformity of the alloys surface properties and their subsequent response to chemical and electrochemical processes.
There are eight main groups of aluminum alloys:

Alloy series
1000
2000
3000
4000
5000
6000
7000
8000

Main alloy groups


99.099.99% aluminum
AlCu
AlMn
AlSi
AlMg
AlMgSi
AlZnMg(Cu)
Others, some including AlLi

In most commercial applications the corrosion of aluminum and its alloys is not
an issue during service life because the thin surface oxide film formed during
heat-treatment and\or during product fabrication reduces corrosion to negligible
rates. In benign service environments this remains the case even if these preexisting surface films are mechanically disrupted because another protective film
rapidly forms locally preventing the initiation of corrosion. Situations do arise,
however, where this is not the case, and a surface that was protective, locally
loses this capability and a corrosion process initiates.

III. TYPICAL MODES OF CORROSION EXPERIENCED


BY ALUMINIUM AND ITS ALLOYS
Over the years, aluminum alloys have been reported susceptible to many forms
of corrosion, some of which are rare and specific to aluminum alloys (for instance,
snowflake corrosion or fingerprint corrosion) (3). The more common forms
of corrosion potentially suffered by aluminum and its alloys are both alloy-system
and alloy-temper sensitive.
A. General Corrosion
General corrosion occurs when a preexisting surface film is no longer protective. This usually occurs in either strongly acidic or alkaline conditions, with

Aluminum Alloys

175

the initial surface film dissolving and the new film that develops being
nonprotective. The extent of the damage induced depends on the specific chemical composition of the electrolyte, its temperature, and its volume. At one extreme, the alloy can completely dissolve; at the other, the corrosion product consumes all the free water, resulting in a gel layer forming that can stifle further
attack and eventually may become a glasslike solid layer.
In highly acidic solutions (pH 2), general corrosion usually occurs via
a crystallographic mode, whereas in highly alkaline solutions (pH 11), attack
is usually associated with the formation of shallow cusps. General corrosion
is not usually experienced during a commercial products service life unless the
environmental conditions encountered stray well outside those anticipated for its
application.
B. Localized Corrosion
Localized corrosion is the most common form of corrosion experienced by aluminum alloys during service applications. It occurs when the surface oxide film,
for some reason, locally can no longer prevent the initiation of corrosion. The
most common forms of localized corrosion are pitting, intergranular corrosion,
crevice corrosion, galvanic corrosion, exfoliation, stress corrosion, and corrosion
fatigue. Less common forms, although in some cases commercially important
forms, include filiform corrosion, erosioncorrosion, poultice corrosion, biological corrosion, snowflake corrosion, and fingerprint corrosion.
C. Pitting Corrosion
Pitting corrosion is the most common form of localized corrosion suffered by
aluminum and its alloys, and although numerous pitting studies have been undertaken since the 1930s, the universal agreement on how pits initiate is still awaited.
An example of the pitting of an aluminum alloy is shown in Fig. 1 and a polished
metallographic cross section through a typical pit is shown in Fig. 2.
D. Intergranular Corrosion
Intergranular corrosion (IGC), second to pitting corrosion, is probably the next
most common form of corrosion suffered by aluminum alloys. A typical example
of intergranular corrosion is given in Fig. 3. It is potentially more damaging than
pitting because for a given loss of metal, the percentage reduction in a structures
load-bearing capability is significantly higher for IGC and the observed tendency
for self-stifling is less common than is found for pitting corrosion.
Numerous studies have been published over the years, and compared to
pitting corrosion, there is a greater consensus on the mechanisms involved; albeit,
details may be alloy system and temper dependent. In general, underaged tempers

176

Holroyd

Fig. 1 Pitting morphology before and after cleaning an aluminum alloy after long-term
exposure to a saline environment (reduction 3).

Fig. 2 A polished metallurgical cross section through a typical pit shown in Fig. 1 (magnification 200).

Aluminum Alloys

177

Fig. 3 Typical intergranular corrosion in an aluminum alloy (magnification 200).

are more susceptible to intergranular corrosion that either peak or overaged tempers.
E.

Crevice Corrosion

An opportunity for crevice corrosion occurs when the local geometry provides
a region where a high surface area of the metal can be exposed to a relatively
small volume of an aqueous solution that has a restricted access to the bulk environment.
It has been argued that crevice corrosion is a variant of pitting corrosion,
where the initiation phase of pitting is eliminated by the crevice itself, providing
the restricted geometry (i.e., a very large pit) needed for the development of the
local environmental conditions for corrosion initiation. Although this is mechanistically incorrect, it is not an unreasonable visualization of why crevices should
be avoided in good designs and how tight crevices could lead to the initiation
of localized corrosion.
F.

Galvanic Corrosion

Galvanic corrosion of aluminum alloys usually only occurs as a result of poor


design or incorrect material selection (Fig. 4). The observed form of attack depends on the actual environmental conditions, the aluminum alloy system, its
temper, and the magnitude of the galvanic stimulation. Galvanic corrosion generally manifests itself as one of the other forms of localized corrosion, usually

178

Holroyd

Fig. 4 Several localized corrosion modes occurring simultaneously: pitting, galvanic


corrosion (mixed metal), and erosioncorrosion (sometimes referred to as impingement).

either pitting or crevice corrosion; however, severe galvanic stimulation can lead
to general corrosion.
G.

Exfoliation Corrosion

Exfoliation corrosion is sometimes referred to as lamella or layer corrosion


because, in cross section, it has the visual appearance of flaky pastry (Fig. 5).
It usually only occurs in highly worked wrought alloys and the phenomena may be

Aluminum Alloys

179

Fig. 5 Section of an agricultural vehicle showing severe exfoliation corrosion: magnification (a) 3/4; (b) 15.

180

Holroyd

regarded as corrosion occurring simultaneously along multiple parallel paths associated with a given microstructural region adjacent to the alloys external surface.
Exfoliation corrosion has been observed in both heat-treatable and nonheat-treatable alloys. For heat-treatable alloys, the corrosion paths follow grain
boundaries and/or regions adjacent to them, whereas for non-heat-treatable alloys, the regions followed involve those rich in transitional metals and the regions
immediately adjacent.
H. Stress-Corrosion Cracking
Stress-corrosion cracking (SCC) is a complex phenomena requiring the concurrent action of both local strain and corrosion. Generally, it is only of practical
concern for the higher-strength aluminum alloys, and then only when heat-treatable alloys [e.g., AlCuMg, AlZnMg(Cu), or AlLi] are in underaged or
peak-aged tempers. Non-heat-treatable AlMg alloys with magnesium concentrations above around 3% (wt) may be susceptible to SCC if they become sensitized
by a grain-boundary precipitation of an active phase during service life. (This
process tends to be accelerated in highly worked alloy microstructures.)
Medium-strength heat-treatable alloys such as AlMgSi (6xxx series)
alloys are rarely susceptible to SCC other than when highly alloyed and subjected
to unusual noncommercial laboratory heat treatments (3,5). Several extensive and
complete reviews of the SCC of aluminum alloys have been published (59) and
the reader is directed to these works, as this aspect of aluminum corrosion will
not be covered in this chapter.
I.

Corrosion Fatigue

Corrosion fatigue (CF) is a complex phenomenon requiring the concurrent action


of both local cyclic strain and corrosion. CF can occur in a wide spectrum of
conditions, involving crack-propagation mechanisms ranging from where local
corrosion is contributing to a process that is essentially pure fatigue through to
a situation in which the cyclic loading is enhancing the crack growth rate in a
stress-corrosion process. As with SCC, the reader is directed to an extensive and
complete review of the subject (10).

IV. ENVIRONMENTAL EFFECTS ON ALUMINUM ALLOYS


Environmental effects on aluminum and its alloys are dictated initially by how
the surface oxide responds to its new environment. Before discussing these
interactions, it is appropriate to review the types of surface oxide film that
typically are present on the aluminum surfaces prior to their environmental exposure.

Aluminum Alloys

181

A. Surface Oxide Films on Aluminum


Aluminum is an active metal and its stable behavior depends on its exposed
surfaces being covered by a protective thin oxide film. The nature of this film,
its stability over time, and the consequences following its local disruption in a
given set of environmental conditions usually dictates the corrosion performance
of aluminum-based alloys in those conditions. Despite this knowledge, it is interesting to note how little attention has been paid to this detail over the years by
the numerous research workers studying the corrosion and electrochemistry of
aluminum alloys. The first reaction is to remove the irreproducible as-received
surface and replace it with something else.
Good examples of where the corrosion performance of a given alloy in a
particular environment is strongly influenced by the nature of the pre-existing
surface oxide prior to the alloy being exposed to an environment include the
following:
1. Godards (11) statement: Experiments I have performed indicate that
the thickened films developed in pure water at room temperature increase the resistance of the surface to corrosion; if such films can be
developed before some corrosion conditions are encountered no corrosion will occur, whereas a freshly exposed surface with only an air
formed film will be subject to corrosion.
2. The blue corrosion phenomenon (12), experienced in the early
1980s, where users of household aluminum foil (99% Al) observed
small blue patches developing on foil surfaces. This corrosion issue
occurred because the foil alloy contained unexpected trace levels of
lithium (10 ppm) that, during thermal treatment, resulted in the oxide
film having localized areas of lithium surface enrichment of up to
4000: 1 (13,14) which upon exposure to humid air (as often found in
domestic kitchens) suffered corrosion and generated a local blue
color, whereas regions without lithium suffer no attack.
3. Ranking of the pitting corrosion susceptibilities of aluminum alloy systems using well-known standard methods can be influenced by the surface pretreatment employed prior to testing (15).
B. Oxide Films Formed on Aluminum Alloys in
Dry Environments
The surface films forming on aluminum exposed to the natural atmosphere, dry
air, or oxygen at room temperature initiate instantaneously. These films consist
thick) covered by a slightly thicker, less
of a compact inner layer (1015 A
dense outer layer of an amorphous noncrystalline aluminum oxide that reaches
after a few hours (16).
a limiting thickness, typically less that 50 A
These duplex oxide films are produced on all aluminum alloys for tem-

182

Holroyd

peratures up to around 400C, as long as the alloy does not contain alloying
elements, such as magnesium or lithium, that react selectively to form their own
oxides. The principle difference for these higher-temperature films is that the
-thick compact
thickness of the outer -Al2O3 layer, forming over the 1015-A
inner layer, increases with temperature. Film thickness measurements quoted in
the literature vary significantly, reflecting that, as with corrosion, the surface preparation prior to oxidation often has a significant influence on the subsequent oxidation behavior (17). Notwithstanding this, it is reasonable to assume oxides will
thick [normally 100 A
(18)] for temperatures up to at 400C
be less than 200 A

and increase to around 400 A at 600C (19). Alloying elements such as copper
and manganese have been reported to enter and dope -Al2O3 at these higher
temperatures (20).
At temperatures above about 425450C, the amorphous outer layer of
the duplex film initially forming on aluminum alloys undergoes a discontinuous
structure change. This allows the rapid migration of oxygen to the oxidemetal
surface and the nucleation and growth of crystalline -Al2O3 as a new phase below
the amorphous layer (21,22).
The presence of alloying elements that react selectively, forming their own
oxides, slightly complicates the oxidation of aluminum alloys. For commercial
aluminum alloys, the principal elements doing this are magnesium and lithium.
For aluminum alloys containing more than 0.5 wt% magnesium, the oxides generated at temperatures up to around 300C are independent of the magnesium addition. However, for higher temperatures, the outermost surface of the
oxides formed is essentially a dense layer of MgO crystals (23). Electron microscopy (24) has shown that MgO crystals can also form at the metaloxide interface, thereby disrupting the inner compact alumina layer.
Zinc additions to aluminum alloys appear to have minor effects on an
alloys oxidation characteristics (25,26) with no evidence existing for zinc incorporation into the films formed on high-purity AlZnMg alloys (27) and only
limited evidence for commercial AlZnMgCu alloys (26,28). Experimental
evidence for these alloys shows that the alloy grain boundaries provide a shortcircuit diffusion path for magnesium to participate in the oxidation process (27).
The presence of lithium in aluminum alloys even at low concentrations has
a dominant effect on the oxidation behavior of aluminum alloys, overriding those
due to magnesium and other alloy additions to commercial alloys (25).
C. Oxide Films Formed on Aluminum Alloys in
Wet Environments
As stated earlier, the natural oxide formed on aluminum in dry air at room
-thick barrier layer next to the
temperature is duplex with a compact 1015-A
metal covered with a thicker less dense layer. The thickness of the outer layer,
although initially around 50 angstroms, is highly sensitive to the prevailing envi-

Aluminum Alloys

183

ronmental conditions (11,29). Although researchers have studied the filming characteristics of aluminum alloys in water for decades, accurate prediction of the
expected film thickness of aluminum alloy is not yet possible. This is due to the
multiplicity of variables involved which include: the nature of the pre-existing
surface oxide, details of the water chemistry composition, temperature, pH, oxygen content, etc. and the alloy composition.
An excellent review of the films forming in the aluminum-water system
has been provided by Alwitt (29), who states: It is an interesting fact that those
phases commonly present as surface films on aluminumpseudo-boehmite, bayerite, boehmite and -Al2O3 are all metastable phases of the Al2O3 water system and the film composition is often the most kinetically accessible product,
which often is only an intermediate product on the path to thermodynamic equilibrium.
Film forming on aluminum exposed to water at various temperatures can
be categorized by four temperature regimes: a) room temperature up to around
40C; b) above 40C to 100C; c) 100C to 150C and d) above 150C.

1. Films Formed at Temperatures at and Below 40C


These films differ from those formed at temperatures up to 100C (2931), with
the films formed at 40C being basically either exclusively bayerite or a very
thin of layer pseudo-boehmite covered by a thicker layer of bayerite. An explanation of this behavior at temperatures below around 45C could be that although
pseudo-boehmite can form on aluminum surfaces immersed in water at these
temperatures, its growth rate is so slow and bayerite precipitation occurs because
it is kinetically more favorable. Alternatively, the pseudo-boehmite could start
to form, but it then dissolves at the filmsolution interface and reprecipitates as
the thermodynamically more stable product bayerite (29,30), thereby generating
a duplex film with a very thin partially protective pseudo-boehmite layer covered
by a thick bayerite layer. The net effect is that the films formed at temperatures
below 45C are thicker, more porous, less protective, and more brittle, with a
tendency to locally spall away from the metal substrate (29,30) (e.g., see Fig. 12
in Ref. 29).
A clear indication that the filming characteristics change at temperatures
around 45C is provided by the temperature dependence of the film formation
induction times (Fig. 6). The implied activation energy is 18.7 kcal/mol for temperatures in the range 50100C, whereas it is significantly higher for lower
temperatures.

2. Films Formed at Temperatures Above 40 C and


Below 100 C
These films have a duplex structure, consisting of an inner layer of pseudoboehmite [a poorly crystallized oxyhydroxide, similar to boehmite (AlOOH) but

184

Holroyd

Fig. 6 Temperature dependence of the induction time for film growth on pure aluminium
exposed to distilled water in the temperature range 30100C showing that filming process
change occurs at temperatures around 45C. from Ref. 18; from Ref. 29; from
Ref. 89; and from Ref. 33.

containing excess water (18)] and an outer layer of bayerite crystals [a form of
aluminum hydroxide (32)].
Film growth for this temperature range proceeds in three stages: (1) an
induction period, during which no significant weight change is observed, (2) a
period during which pseudo-boehmite forms, and (3) bayerite crystallization onto
the pseudo-boehmite layer. Stages 2 and 3 can overlap, with bayerite crystal
nucleation occurring on the boehmite while the boehmite is still growing (29).
It has been suggested that the cessation of film growth is dictated by the completion of the bayerite layer and is independent on the formation of the underlying
pseudo-boehmite layer, which is considered to reach a limiting thickness rapidly (29).
The precise mechanism is not fully established. Scamans and Rehal (32)
have made a reasonable proposal based on published work, suggesting that the
pre-existing amorphous film thickens during the induction period [as proposed

Aluminum Alloys

185

by Hart (33)] with either proton, hydroxyl ions, and/or water penetrating the
film to establish a soluble aluminum species. A dissolutionprecipitation process
follows, leading to the rapid formation of a pseudo-boehmite layer on the aluminum surface that, once complete, can only thicken slowly via a process limited
by the diffusion of water into and through the film, as opposed to the outward
diffusion of Al3 ions. These authors suggest that the mechanistic details of the
bayerite formation process awaits clarification. However, it probably involves a
further dissolutionprecipitation process, in this case, occurring at the pseudoboehmitesolution interface with a typical bayerite layer thickness growing to
23 m at 70C.

3. Films Formed at Temperatures in the Range 100150 C


These films on aluminum in water and steam have been studied in great detail
and reviewed by Altenpohl (34). According to Altenpohl (35,36), typical boehmite films developed on superpurity aluminum immersed in boiling distilled water
involve two types of boehmite. The first forms within 10 min in boiling water
and will redissolve fairly readily, whereas the second, which is highly insoluble,
forms subsequently as a thinner layer beneath the thicker, more soluble form of
boehmite. A schematic of these boehmite films is given in Fig. 7 and typical film

Fig. 7 Schematic representations as given by Altenpohl (34) for cross sections through
boehmite films formed on superpure aluminum immersed in boiling distilled water for 1
-thick barrier layer at
h and 5 h. [Note: Subsequent studies indicate that a retained 10-A
the aluminumfilm interface is unlikely to be present (29).]

186

Holroyd

Fig. 8 Typical film thicknesstime data according to Altenpohl (34) for film growth on
AA1145 immersed in boiling distilled water.

thicknesstime data for their film growth on AA1145 exposed to distilled water
is shown in Fig. 8.
It is worth noting that Altenpohls filming model (34) envisions that the
thin, compact barrier oxide layer on the aluminum surface prior to any reaction
with water remains in situ throughout the boehmite filming process. Alwitt (29),
in his review of the aluminumwater system states that in his opinion, there
appears to little experimental evidence for such a film.
The benefits offered by boehmite films can be impressive. For instance,
data from Leidheisser and Kellermans study of straining aluminum alloy wires in
aqueous environments (37) shows that immersing commercially pure aluminum
(AA1100) in boiling water for 15 min prior to rapidly straining in a 0.1M NaCl
solution significantly improves the surface films ductility compared to that given
by an air-formed film or an anodized film exposed to boiling water (Fig. 9).
Another good example is the improved pitting-resistant boehmite films can
provide in potentially aggressive tap waters, as is shown in Table 1. Based on

Aluminum Alloys

187

Fig. 9 Effect of surface films on the electrochemical potential response after commercially pure aluminum (AA1100) wires are rapidly plastically strained 1.5% in a deaerated
0.1M NaCl solution, pH 5.5. (Data from Ref. 37.)

the data in Table 1 and other data, a significant interest was generated during the
1960s to attempt to utilize boehmite films commercially to render aluminum
cans suitable for packaging beer, milk, and other products (34).
Attractive as the boehmite films are, their commercial exploitation to
date has been restricted by issues associated with the conditions needed to successfully generate these films in either boiling water or in high-temperature steam.
For instance,
1. Protective boehmite films have only commercially been reproducibly formed on high-purity aluminum.
2. Trace additions of contaminants in boiling water or steam can either
stifle film growth or promote corrosion and/or the growth of a nonprotective film.
Re-examination of the boehmite film thickness data provided by Altenpohl
(34) for six superpure aluminum alloy compositions subjected to different surface
pretreatments prior to their immersion in boiling high-purity distilled water for

188

Holroyd

Table 1 Number of Pits Generated on Filmed Commercially Pure Aluminum


(AA1145) After Exposure to Fairly Hard Tap Water for 9.5 Months at Room
Temperature
No. of
corrosion pits
per dm2 in tap
water

Type of distilled water used for


boehmite film formation

50C

70C

A
B
C
A 0.1N NH4OH

5
30
12
0

2
3
0
0

B 0.1N NH4OH
C 0.1N NH4OH
Air-formed film

0
0
18

0
0
3

Specific resistance of
distilled water (cm)
8 105
7 105
5 105
2.18 104 to 1.45 104
(start)
(finish)

Note: Boehmite films produced by immersion in boiling distilled waters for 4 hours, with and without
a 0.1N ammonium hydroxide addition.
Source: Data from Ref. 34; see reference for details on distilled-water compositions.

various times reveals that the film thickness generated is effectively controlled
by an alloys iron, copper, and silicon contents. This conclusion is based on the
following observation of high-purity aluminum (i.e., concentrations of iron, copper, and silicon will always be low):
1. Thicker films only occur if an alloys copper and iron contents are
below 0.0015 and 0.002 wt%, respectively (Fig. 10).
2. When copper and iron levels are sufficiently low, increasing an alloys
silicon weight percent favors thicker films (Fig. 10).
The latter observation is interesting in view of the well-known detrimental effects
trace additions of silicon to the alloy have on film formation at temperatures
above 150C or dissolved silicate ions additions to boiling distilled water have
on boehmite film formation (Fig. 11). Based on the latter, it is reasonable to
suggest that it is the free silicon in the alloy that is detrimental to boehmite
film formation, and when at appropriate low concentrations, iron and silicon can
form relatively benign precipitates, thereby allowing copper levels to control
boehmite film thickness (see Fig. 10).
Some published data on the influence of anions on the growth behavior of
boehmite films on aluminum immersed in boiling water are available (34,38,39).
Data presented by McCune et al. (38) for 99.99% aluminum (AA1100) in boiling

Aluminum Alloys

189

Fig. 10 Effect of silicon content on the boehmite film thickness developed on a range
of superpurity aluminum compositions immersed in boiling distilled water for either 2 or
8 days. (Data from Ref. 34.)

190

Holroyd

Fig. 11 Boehmite film thickness developed on 99.99% aluminum after 4 h immersion


in boiling distilled water containing various silicate ions additions up to 15 ppm. (Data
from Ref. 34.)

distilled and deionized water containing a range of inhibitor anions as single


additions suggests that in all cases the film growth was less than the pure water
case after 1 h (Fig. 12). Comparison of these data with that provided by Altenpohl
(34) indicates that although there is good agreement between the experimental
data, the addition of anions does not necessarily lead to film thickness reductions.

4. Films Formed at Temperatures Above 150 C


At atmospheric pressure, boehmite is the only phase consistently reported (40
43) to form on aluminum surfaces exposed to water or steam in the temperature
range 150374C, the critical temperature of water. Hart and Maurin (41) have
observed that Dispore can form as a thin outer layer over the boehmite after
extended periods at temperatures above 300C. For temperatures above 374C,
-alumina and corundum are formed (41).
In the mid-1950s, there was considerable interest in using aluminum to
clad the fuel elements used in water-cooled nuclear power reactors, with the

Aluminum Alloys

191

Fig. 12 Apparent pseudo-boehmite film thickness formed on 99.99% (AA1100) aluminum immersed in boiling distilled water containing various single-anion additions as sodium salts (1 g/L nitrate, 2 g/L sulfate, and 1 g/L silicate) for various up to 1 h. (Data
from Ref. 38.)

aluminum being exposed to water at temperatures up to around 350C. Aluminums attractions for this application were its low cost, ease of fabrication, and
low-adsorption cross section for neutrons. An intense research and development
(R&D) effort ensued in North America to overcome two issues:
1. To improve aluminums environmental performance in water at elevated temperatures and to avoid the situation where, above a critical

192

Holroyd

temperature, surface films lose their protective nature and allow intergranular corrosion and rapid loss of metal
2. To increase the strength of aluminum at temperatures around 300C
Aluminum alloy development was the R&D approach adopted to achieve the above.
This decision was based on the belief that appropriate alloying additions would
exist that could simultaneously increase an alloys strength while also increasing
the critical temperature above which aluminum corrodes rapidly, which was known
to increase as the alloys purity decreased (44). [Typical critical temperatures are
105C for 99.99% aluminum, 150C for 99.85% aluminum, 180C for 99.5% aluminum and a higher temperature for 2S aluminum (99.099.3% aluminum), which
normally performs well in water up to temperature around 200C (34).]
Work in the United States at the Argonne National Laboratories in 1955
established that a 1 wt% nickel addition to AA1100 prevented intergranular attack
at 350C (45), changing the mode to uniform corrosion (40). This led to the
development of the experimental alloy X-8001, which was superseded in 1957
by two further alloys based on AlNiFeTi (A203X and A198X), that both
offered a superior corrosion performance in water at 360C, which was attributed
to these alloys having a finer dispersion of second-phase precipitates.
Corrosion performances for the initial commercially cast versions of the
above AlNiFeTi alloys in high-temperature water were extremely variable,
and after detailed studies, it became evident that the corrosion performance was
extremely sensitive to the alloys purity levelin particular, the silicon weight
percent. A further alloy development program was undertaken by the General
Electric Company to generate an alloy that provided yet a better corrosion performance in water at 360C while also having a sufficient tolerance to the alloys
purity level that it could be readily cast by conventional casting practices. An
alloy based on Al1.8 Fe1.2 Ni resulted (46) that, as well as being sufficiently
tolerant to the alloys purity levels, was also tolerant of long-term thermal treatment at 550C and provided a superior corrosion performance.
In parallel with the United States effort, Krenz (42), working in Canada,
reported two AlNiFeSi alloys (Al2 Ni and Al0.5 Ni, both containing 0.5%
Fe and 0.2% Si) that performed reasonably well when exposed to water at 300C.
Initial attack of the alloys was rapid, but further attack slowed down to a linear
rate that was marginally lower for the higher-nickel-containing alloy.
MacLennan (47) exposed the above two AlNiFeSi alloys along with
2S aluminum to water at 300C for short periods of time (minutes) and then took
replicas of the inner and outer surfaces of the oxide films and evaluated them
using electron microscopy. Based on this study and further work involving longer
exposure times in water at temperatures in the range 150340C (48), MacLennan attributed the improved corrosion performance for the nickel-containing
alloys to their oxide films providing a local stress relief effect. This stress
relief effect was deemed to minimize any tendency for the films to crack, as is

Aluminum Alloys

193

thought to occur to the oxides formed on aluminum exposed to high-temperature


water (49). Improved film ductility was suggested to be provided by the
second-phase particles and the adjacent aluminum-forming local cells where
the aluminum rapidly dissolves until the particle is isolated from the metal.
Other workers in Canada simultaneously evaluated the oxide films forming
on these AlNiFeSi alloys after exposure to pure water at high temperatures
(200300C). Grenblatt and McMillan (50), using electrochemical polarization
techniques, established that the oxide films differed from those formed on
AA1100, and Greenblatt (51), using optical metallography to study the oxide
thickness and the corrosion pit depths, showed that the corrosion mechanism for
AA1100 differed from that for AlNiFeSi alloys.
Early operational experience for aluminum-alloy-clad fuel elements used in
pressurized and boiling-water nuclear reactors quickly revealed that the corrosion
performance was significantly inferior in the pressurized-water reactor case with
the aluminum alloys oxide film dissolving in the high-temperature water at
appreciable rates (52). The reason for this is that in the pressurized-water case,
unlike for the boiling-water case, the oxide surface temperature is not uniform
and the solution layer in contact with the oxide is not always nearly saturated
with dissolved oxide.

5. Electrochemical PotentialpH Diagrams


Electrochemical potential versus solution pH diagrams for the aluminumwater
system have been presented for temperatures in the range 25300C (53,54).
Construction of these diagrams are based on calculations using thermodynamic
data which inevitably involves making assumptions that are known to introduce
oversimplifications. For instance, for the calculations, it is normally assumed that
the chosen stable oxide film is in equilibrium with soluble aqueous species such
as Al3 and Al(OH)4 ions, although it is well established that at intermediate
pHs, the hydrolysis products of the Al3 ion dominate, and at higher temperatures, polynuclear complexes readily form. Despite these oversimplifications,
qualitatively useful information may be gleaned from these diagrams. For example, (a) the pH range for oxide stability is highly temperature dependent [410
at 25C, 37 at 100C, and 1.53.5 at 300C (54)] and (b) the solution pH for
minimum corrosion as a function of temperature are correctly predicted [6.5
at 50C (40), 5.7 at 92C (55), and 3.0 for 170300C (40)]. A typical potentialpH diagram for aluminum in water at 25C is given in Fig. 13.
V.

OXIDE FILM DEVELOPMENT AND CORROSION


INITIATION IN AQUEOUS ENVIRONMENTS

Aluminum surfaces are normally covered by a protective thin oxide film. The
specific nature of this film depends on the conditions under which it was formed

194

Holroyd

Fig. 13 Electrochemical potentialsolution pH diagram (often referred to as a Pourbaix


diagram) showing thermodynamically predicted regions for corrosion, passivation, and
immunity of aluminum in water at 25C, assuming that the oxide formed is bayerite.

and subsequently stored. Time-dependent processes are initiated when these surfaces are exposed to aqueous environments, the results of which are dependent
on several factors, including the aluminum alloy composition, temper and microstructure, the nature of the pre-existing surface oxide film itself, and the prevailing local electrochemical and environmental conditions (e.g., electrode potential, solution chemistry, temperature, pH, oxygen concentration, etc.).
When exposed to sufficiently aggressive aqueous environments, the protection provided by the pre-existing oxide film on aluminum alloys is negated
and general corrosion of the underlying metal is initiated. For the vast majority
of structural or engineering applications of aluminum alloys, these situations are
of no interest and these conditions need to be avoided at all costs. Exceptions
to this usually involve situations in which the alloy is intentionally being used
as (a) a battery anode [e.g., an aluminumair battery (56)] or (b) a sacrificial
anode (57) or aluminum-based spray coating (5860) to protect another material

Aluminum Alloys

195

such as a steel tension leg in an offshore oil platform structure (61). In these
cases, the general corrosion behavior itself has to be controlled to maximize
coulombic efficiencies (e.g., by alloy composition and microstructure manipulation). Another exception is the situation in which an aluminum alloy is subjected
to a short period of alkaline etching aimed at removing the pre-existing surface
film and the preferential removal of second-phase particles that otherwise would
act as a cathodic reaction sites during any subsequent corrosion processes (15).
Although it is true that aluminum alloys are only resistant to general corrosion when the solutions pHs are neither too high nor too low, it is now generally
accepted that an alloys corrosion performance is controlled by the nature of the
aluminum salt involved rather than the solutions pH produced by the hydrolysis
of the aluminum cation (62). Published information on the inhibition of aluminum
alloy corrosion in highly acidic (6365) and highly alkaline (6668) environments is available.
To simplify the following discussion we will initially concentrate on the
behavior of nonalloyed aluminum (i.e., high-purity and commercial purity grades
of aluminum alloys). Following this, we will examine the effects of single-alloy
additions (i.e., binary alloys) and, finally, consider the commercial alloy systems.
A. Unalloyed Aluminum Alloys
Although some controversy still exists in the literature on the details of corrosion
initiation, there is now a strong opinion that the initial process occurring when
an oxide-filmed aluminum surface is first introduced into an aqueous environment, although being sensitive to electrochemical potential, is predominantly a
chemical process. In a comprehensive review covering over 70 years of published literature Foley (69) came to the conclusion that the initiation of localized
corrosion of aluminum could be described by four steps involving the following:
1. Adsorption of a reactive species (usually an anion) on the oxidecovered aluminum alloy surface
2. A chemical reaction of the adsorbed species with the aluminum ion in
the aluminum oxide lattice or precipitated aluminum hydroxide
3. The effective thinning of the oxide film by dissolution or
anions penetrating the film
4. Direct attack of the exposed metal by the anion, possibly assisted by
anodic potentials (sometimes, Steps 3 and 4 can occur simultaneously)
Step 1: Adsorption A range of analytical tools, including autoradiography, secondary ion mass spectroscopy (SIMS), and X-ray photoelectron
microscopy (XPS), have provided substantial experimental evidence for
the competitive adsorption of anions on aluminum surfaces [e.g., chloride (7075), sulfate (38,70,72,75), perchlorate (72), nitrate (38,72),

196

Holroyd

chromate (72,76), and molybdate (38,77,78)] and the subsequent influence upon whether or not aluminum will subsequently undergo dissolution. This is particularly the case for the adsorption of chloride ions as
a preliminary step to aluminum alloys suffering pitting corrosion (70
75).
Observed adsorption behavior is dependent on the anion species involved.
For example, experimental evidence provided by Tomcsanyi et al. (70)
using a radio-tracer technique on 99.99% pure aluminum at its free-corrosion potential in various aqueous solutions has shown the following:
1. Adsorbed labeled sulfate ions can be displaced by further additions of
nonlabeled sulfate ions, whereas they are not displaced by the further
additions of chloride ions (Fig. 14).
2. Chromate ions displace adsorbed chloride ions and electrochemical potentials move to more positive values (Fig. 15), whereas further additions of nonlabeled chloride ions do not displace adsorbed labeled chloride ions.

Fig. 14 Accumulation of labeled sulfate ions on pure aluminum surfaces: (A) exchange
when additional nonlabeled sulfate ions are added and (B) effect of adding 5 103M
sodium chloride. (Data from Ref. 70.)

Aluminum Alloys

197

Fig. 15 Effect of chromate ion additions to the surface accumulation of chloride ions
on pure aluminum in a 1M sodium sulfate solution: (A) a 0.1M Na2CrO4 addition to a
solution containing 4 102M sodium chloride and (B) a 5 102M Na2CrO4 addition
to a solution containing 2 102M sodium chloride. (Data from Ref. 70.)

3. The concentration of adsorbed chloride is both time and concentration


dependent (Fig. 16).
4. The excess surface concentration of chloride ions is insensitive for all
electrode potentials up to around 750 mV on the saturated calomel
electrode scale (SCE) other than for cathodic potentials where measurable hydrogen evolution occurs and adsorbed concentrations significantly decrease with decreasing potential.
5. Surface concentrations of adsorbed chloride ions are not fully reversible when the electrode potential is moved from a given anodic value
to a sufficiently cathodic one and then back to the original anodic potential (Fig. 17).
These results are important. The first two give a good indication of how
chromate and sulfate ions initially inhibit the detrimental effects of chloride ions [note: Augustynski (72) suggests that nitrate ions retard the

198

Holroyd

Fig. 16 Time dependency of the surface accumulation of chloride ions on pure aluminum exposed to various bulk solution chloride concentrations in a 1M sodium sulfate
supporting electrolyte. (Data from Ref. 70.)

adsorption of chloride ions], whereas the next two are consistent with the
adsorbed chloride ions chemically rather than electrochemically
reacting with the oxide film. Justification of the latter suggestion is
that the adsorption behavior for a given surface condition is insensitive
to electrochemical potential and that although the concentration of the
adsorbed chloride increases with increasing bulk chloride solution, the
absolute concentrations always start decreasing after 1 h. Rationalization of the final observation that the adsorption behavior was not reversible was that the adsorption is sensitive to the surface roughness of
the aluminum substrate, which will be modified by the electrode potential
being held at a cathodic potential.
Tomcsanyi et al.s (70) suggestion that chloride adsorption on oxide-covered aluminum is insensitive to electrode potential would seem inconsistent with Augustynskis (72) results showing that adsorbed chloride concentrations increased from around 3 at% at the free-corrosion potential to
around 1213 at% at potentials approaching the critical pitting potential.

Aluminum Alloys

199

Fig. 17 Evidence that the concentration of accumulated chloride ions on aluminum surfaces depend on both time and the current surface condition. Effects of cathodic polarization were not reversible. (Data from Ref. 70.)

Rationalization of these differences is probably associated with the timedependent changes occurring after the chloride ions are initially adsorbed;
this will be discussed below in Step 2 of the proposed process, which
covers the chemical reaction of the adsorbed species with the substrate.
Further evidence supporting the proposal that chloride ions adsorb on aluminum surfaces and then react with the oxide is provided by the simple but elegant experimental work published by Nguyen and Foley (73),
who measured the uptake of chloride ions and the production of Al3
ions as a function of chloride concentration when 10 g of either aluminum powder or aluminum oxide were mixed with 100 mL of an aqueous
sodium chloride solution for 16 h. The results generated showed that the
chloride ions adsorption is consistently higher on aluminum oxide than
on freshly generated aluminum powder, and for both substrates, the adsorption increased with increasing bulk solution chloride concentrations
up to a maximum at around 0.8M sodium chloride. The results for alumina are summarized in Fig. 18, along with data from similar experiments on aluminum powder conducted by Drazic et al. (74), who used

200

Holroyd

Fig. 18 Adsorption of chloride ions onto alumina from a range of sodium chloride solutions. from Ref. 73; from Ref. 74.

more dilute sodium chloride solutions. (Drazic et al.s aluminum powder


behaved as if it was alumina, so it was assumed it was coated with a
relatively thick layer of aluminum oxide.)
Having established that anion adsorption on oxide-covered aluminum immersed in aqueous environments is a competitive process, with certain
ions promoting passivity and others film degradation, the logical question
to ask is whether anion adsorption is a uniform or localized phenomenon.
An argument for anion adsorption on alumina being relatively uniform
can be made on the basis that the measured isoelectric point for alumina
(79) and aluminum-oxide-coated aluminum (80) is in the range 8.99.1,
irrespective of the chloride concentration (79). Hence, it is reasonable to
expect that negatively charged species such as anions will be uniformly
attracted to aluminum-oxide-coated aluminum surfaces. In line with this,
Nguyen and Foleys (73) results show that chloride ion adsorption on
aluminum oxide exposed an aqueous sodium chloride solution of given

Aluminum Alloys

201

concentration increased linearly on a one-for-one basis with the increasing surface area of alumina.
An argument for anion adsorption being locally enhanced can be made
using the fact that metal surfaces are rarely homogenous and enhanced
localized adsorption and surface activity will tend to occur preferentially at imperfections, weak spots, and flaws in the surface oxide. In the
early 1970s, Richardson and Wood (81,82) proposed that oxide flaws
in aluminum oxide act as active centers for the initiation of localized
corrosion. They supported their proposal with experimental evidence
from transmission electron microscopy (TEM) studies for the existence
of large flaws in aluminum oxide layers electrochemically grown on
aluminum substrates. Further studies by Wood and later by Thompson
and their co-workers have led to numerous publications generating evidence demonstrating that the initiation of the pitting of aluminum could
be via an oxide flaw mechanism. Irrefutable evidence is awaited;
however, the recent evidence from ultramicrotomed sections of pure aluminum exposed to an aqueous environment (83) is very convincing.
Step 2: Chemical Reaction of the Adsorbed Species with the Surface
Oxide Evidence for the solubility of oxide films on aluminum substrates immersed in aqueous environments was provided by Lorking and
Mayne (84,85) in the early 1960s. These workers generated nonporous
anodized layers on aluminum and then exposed them to a wide range of
anions in aqueous environments which included chloride, sulfate, benzoate, acetate, citrate, phosphate, and chromate anions. In all cases, dissolved aluminum species were detected within 24 h and the concentrations detected were directly related to the initial corrosion rates for all
the environments evaluated other than chloride. On the basis of these
findings, the authors concluded that the initial anodic reaction in the chloride solutions was the formation of soluble aluminum chloride, whereas
in all the other aqueous solutions, the anions promoted the repair of the
anhydrous oxide.
Perhaps the clearest evidence that a chemical reaction occurs between aluminum oxide and an aqueous environment is provided by the experiments conducted by Nguyen and Foley (73). These experiments involved
the mixing of 10 g of aluminum oxide or 10 g of fresh aluminum powder
with 100 mL of various aqueous saline solutions for 16 h and the subsequent measurement of the chloride ions adsorbed and the aluminum species generated. The data obtained indicated that while the chloride ion
concentration adsorbed on the alumina increases with increasing initial
chloride concentration (Fig. 18), the concentration of the dissolved aluminum species found in the aqueous solution decreases, with the highest
concentration being recorded when the alumina is exposed to doubly-

202

Holroyd

Fig. 19 Aluminum ion concentration (Al3) resulting from immersing 10 g of alumina or


fresh aluminum powder in various aqueous sodium chloride solutions. (Data from Ref. 73.)

distilled water (Fig. 19). Results for the aluminum powder were different
in that the dissolved aluminum concentrations were considerable lower
and independent of the solution chloride concentration, and although the
adsorbed chloride ions increased with increasing solution chloride concentration, the absolute concentrations were always considerable lower
than on alumina (73).
Based on the above results, it is reasonable to conclude that aluminum oxide
in the presence of water undergoes hydration and that when chloride ions
are available, the positive surface charge will lead to adsorption and the
generation of aluminum hydroxychloro species such as Al(OH)2Cl and
Al(OH)Cl2 or soluble complexes of the form Al(OH)2Cl2 (86).
Even more compelling evidence supporting the latter proposal was provided by Nguyen and Foley (73) when they mixed 10 g of alumina or
aluminum powder with 100 mL in a 1N sodium chloride solution acidified with 0.01N aluminum chloride. In the experiments involving alu-

Aluminum Alloys

203

mina, the solutions pH became alkaline along with the dissolved aluminum concentration falling dramatically from its initial value of 83 mg/
L to a final value of 3.29 mg/L and the formation of a heavy gelatinous
precipitate, presumably via a chemical reaction of the type
Al3 2OH Cl Al(OH)2Cl
For the aluminum powder case, the dissolved aluminum concentration
increased from 82.9 to 140.0 mg/L and the solution pH remained acidic.
Presumably, in this case, the dissolution of the thin layer of oxide
on the aluminum powder occurs without the solutions pH increasing
significantly and the surface layers of the aluminum powder can then
dissolve in the acidic solution.
The generation of a heavy gelatinous precipitate in the case when Nguyen
and Foley (73) mixed alumina with an acidified saline solution provides
clear evidence of the chemical reactivity of aluminum oxide in acidic
aqueous environments. Another good indicator of the chemical reactivity of oxide films on aluminum surfaces exposed to aqueous saline
environments is given by Tomcsanyi et al.s (70) observation that the
concentration of excess chloride adsorbed on aluminum surfaces decrease with time (Fig. 16). This phenomenon was deemed by Tomcsanyi
et al. (70) to be a multistep heterogeneous transformation of the original
air-formed oxide into a mixture of oxo-hydroxo- and chloro- complexes,
with the details of the process being influenced by the initial surface
oxide and the local solution chemistry at the oxidesolution interface.
It could be wrongly concluded that chemical reactivity between the initial
surface oxide and an aqueous environment necessitates the presence
of aggressive species. Clear evidence that aggressive species are not essential is provided by evidence published by Scamans and Rehal (32).
In their electron microscopy studies, small surface blisters were observed
to form in the oxide when a wide range of aluminum-oxide-covered aluminum alloys (pure Al, AlMg, AlMgSi and AlZnMgCu) were
exposed to water-vapor-saturated air at 70C. Further evidence is provided by the experimental observation for 99.95% purity aluminum artificial crevices exposed to high-purity water, where the local solutions
pH conditions within the crevices become acidic, following a similar
time dependence as observed for most aluminum alloys exposed to aqueous saline solutions (87).
On the basis of the above discussion, it is reasonable to conclude that the
initial surface oxides chemically react with aqueous environments to
form modified surface films that, in the absence of aggressive species,
will exist in a state of equilibrium with the aqueous environment.
Although not yet proven, it would seem reasonable that inhibitor anions

204

Holroyd

such as chromate initially function by competitive adsorption on the initial oxide surface followed by the adsorbed species somehow modifying
the subsequent film-formation process and the film properties. Augustynskis (72) results from XPS experiments supports this view and
he concluded that after the chromate anions adsorb on aluminum oxide
surfaces, chemical reactions lead to hydrated chromium III and aluminum III oxides as well as chromium IV species being present in the
protective film.
Step 3: Effective Thinning of the Oxide Film by Dissolution or
Anion Penetrating the Film The traditional view of the oxide film
on aluminum participating in a corrosion reaction in an aqueous environment has been that it acts essentially as an anhydrous inert barrier
(88,89). Despite all the evidence to the contrary, it is still common for
the oxide to be considered as chemically inert in aqueous solutions
with pHs in the range 49. This is not an accurate view of the oxide
films.
A more realistic view of surface films participating in corrosion reactions
should be of hydrated aluminum oxide layers, probably in a colloidal
state (69) or as a semipermeable membrane (90). Unfortunately, these
ideas of films existing in a colloidal state or as semipermeable membranes have not received the attention they deserve. An exception to this
is the work conducted by Liepina and co-workers during the 1950s and
1960s (9195), in which the corrosion process was envisaged in terms
of colloidalchemical effects occurring on the metal surface. For aluminum in an aqueous potassium chloride solution, these authors suggest the
potential reaction sequence AlAlCl3 polyoxychloride intermediates
amorphous gelsboehmitebayeritehydrargillite, with the proviso that
if the aluminum oxide is in a colloidal state, the chloride ion would
peptize it and render it dispersible. [Interestingly, in 1967, Hoar (96)
suggested that the strong negative charge caused by adsorbed anions
could peptize oxide films. However, to the authors knowledge, this idea
was not pursued.]
As will be discussed further in Step 4 of the proposed process, the modified
oxide that is generated on aluminum surfaces exposed to aqueous
environments must have a significantly higher ductility than those reported for anhydrous alumina films either stripped from or while adhered
to an aluminum substrate (97).
In the very early investigations of the pitting of aluminum in chloride solutions, it was believed that the chloride ion had the ability to penetrate
the surface oxide (88,89), with the anion directly diffusing through the
aluminum oxide lattice. This mechanism has now been discounted and

Aluminum Alloys

205

the concept of penetration has been replaced with ones involving the
following:
1.
2.

Formation of soluble compounds or transitory species at critical sites


(69,98,99) and\or
The transport of species through a semipermeable membrane,
formed as a result of the initial oxide chemically reacting with an
aqueous environment (90)

Step 4: Direct Attack of the Exposed Metal It is most improbable that


the direct attack of film-free aluminum surfaces ever occurs during
the corrosion of aluminum in aqueous environments: an exception perhaps being during stress corrosion (or corrosion fatigue) when the operative crack propagation mechanism, such as hydrogen embrittlement,
involves periodic brittle crack jumps linking the main crack with subsurface internal cracks with surfaces that have not previously been exposed to the aqueous environment. Quantitative electrochemical data describing the behavior of fresh aluminum surfaces when exposed to
aqueous environments is now available (100) based on data from a recently developed test method known as the guillotine method (101). Another exception is an air-stable phosphate surface directly bonded to pure
aluminum without an intermediate oxide layer that has been generated
by Rotole and Sherwood (102) under precisely controlled conditions
generated during in situ experiments in ultrahigh-vacuum surface analytical equipment.
Increasing experimental evidence now suggests that corrosion initiation and
some corrosion processes can occur beneath the surface films that have
formed as a result of the original surface oxide chemically reacting
with its local environment (90,103). A good example of the above is the
experimental evidence provided by Bargeron and Givens (104107) and
Bargeron and Benson (108), who exposed 99.999% aluminum with preanodized oxide surfaces to aqueous chloride solutions. In their initial
studies, Bargeron and Givens (104) observed that small circular blisters
(diameter 100 m) developed between the aluminum surface and the
oxide when the systems electrochemical potential was given a sufficiently anodic short-duration pulse. Further work by Bargeron and Givens (105) for high-purity aluminum in 0.5M KCl indicated the following:
1. Blisters only formed if the steady-state electrochemical potential is
above the pitting potential and the minimum potential for blisters dependence on the chloride concentration closely resembles that reported
by others for the pitting potential (Fig. 20).

206

Holroyd

Fig. 20 Blister growth potential given by Bargeron and Givens (105) compared with
the pitting potentials for pure aluminum in various aqueous chloride solutions quoted by
various authors. from Ref. 120; from Ref. 144.

2. Blister growth proceeds along the metaloxide interface with the oxide
exhibiting plastic behavior.
3. Blisters generally are circular, but the shape can be locally distorted
by grain boundaries and other microstructural features.
4. Corrosion initiates locally in a peripheral zone toward the blisters perimeters, prior to the blister bursting due to excessive internal gas pressure.
5. An oxide skin of some description remains in place over the ruptured blister which is capable of maintaining an occluded cell and
a localized solution chemistry.
6. Gas compositions generated from the blister/pits are specific to the
anions in the aqueous environment; for instance, Bargeron and Benson
(108) detected hydrogen exclusively for chloride solutions and de-

Aluminum Alloys

207

tected a range of gases including hydrogen, nitrogen, and nitrous oxide


for nitrate solutions.
Further work by Bargeron and Givens (106) matched the hydrogen-bubblegeneration activity to the current fluctuations recorded during potentiostatic experiments on aluminum specimens with a single active pit immersed in an aqueous 1M KCl solution. They claimed that irrespective
of bubble rate or size, the current noise was directly related to bubble
generation.
To date, attempts to generate and observe blister formation on pure aluminum surfaces with air-formed oxide films have failed unless the films
. Despite this, Bargeron and
are artificially thickened to around 100 A
Givens (105) and others (109) firmly believe that the blistering process
is the precursor to the pitting of aluminum alloys. Sound evidence in
favor of this argument include the following:
1.

2.

Current transients similar to those found when blisters are observed


are also detected during pit initiation when blisters are not resolvable.
Natishan and McCafferty (109) have observed blister formation in
-thick oxide layers on ion-implanted pure aluminum surfaces
30-A
containing 4 or 12 at% Cr, Si, Zr, Nb, Mo, or Cr Mo. (In these
cases, it was suggested that ion implantation somehow modifies the
properties of the oxide forming after the reaction with the aqueous environment, thereby allowing larger blisters which are detectable to form.)

1. Influence of Electrochemical Potential on


Corrosion Initiation
The conventionally accepted view, as proposed by Pourbaix and his co-workers
(53,110), is that two characteristic potentials are associated with the pitting corrosion of aluminum; namely the critical (or breakdown) potential Ep (a potential
below which pits cannot activate) and the protection (or repassivation) potential
ER; a more active potential below which pre-existing pits will no longer remain
active or grow); see Fig. 21. (Throughout this chapter, these two potentials will
be referred to as Ep, the pitting potential, and ER, the repassivation potential.)
The accepted view was challenged in the early 1970s when results from
studies conducted at the University of Trondheim in Norway (111113) suggested that the potentials Ep and ER were identical and, therefore, there was only
one characteristic potential associated with the pitting corrosion of aluminum in
saline solutions. In these detailed studies, Broli and Holtan (111) measured Ep

208

Holroyd

Fig. 21 Schematic of a typical anodic polarization curve for aluminum and its alloys
in a saline environment.

and ER values for 1S aluminum (99.099.3% purity aluminum) in de-aerated 3%


NaCl using three electrochemical methods: (a) potentiodynamic (1100 mV/
min), (b) a quasi-stationary method, and (c) a potentiostatic method typically
lasting 46 days [later adding galvano-kinetic polarization as a fourth method
(113)].
During further detailed studies at the University of Trondheim by Nisancioglu and Holtan (114,115), it was realized that some of the results published by
Broli and Holtan were flawed due to crevice conditions being established
during some of the electrochemical experiments. Once the experimental problems
were resolved, two definitive papers emerged in 1978 by Nisancioglu and Holtan
(114,115) and conclusions from these studies for 1S aluminum (99.099.3% purity aluminum) in de-aerated 3% NaCl at 30C may be summarized as follows:
The critical pitting potential, Ep (114):
1. It is a fairly reproducible quantity for aluminum in saline solutions if
measured using static electrochemical methods. However, high-

Aluminum Alloys

2.

3.

209

potential scan rates during potentiodynamic polarization methods


can lead to appreciable errors, and scan rates less than a few millivolts per minute are recommended.
It is a strong function of the aqueous solutions chloride concentration and is independent of surface treatment, film thickness, and solution stirring rates up to 2040 rpm.
When measured by a potentiostatic method and is typically 0.76
V (SCE), with

Pits
Pits
Pits
Pits

almost never initiating at E 0.76 V (SCE)


seldom initiating at E 0.76 V (SCE)
often initiating at E 0.75 V (SCE)
always initiating at E 0.75 V (SCE)

4.

It is not an ideal criterion to determine an aluminum alloys pitting


susceptibility.
The protection potential, ER (115):
1. It is typically about 100 mV more active than Ep.
2. It is the lowest potential for pit growth. Pits grown at potentials
above ER but below Ep involve growth from pre-existing occluded
microscopic sites via crystallographic pitting or tunneling, whereas
pitting at potentials above Ep is generally macroscopic in nature.
3. It may be explained in terms of active chloride participation in competition with the hydroxide in the metal dissolution reaction (116),
whereas the basic factors determining the critical pitting potential,
Ep, remain unclear, although its measurement is reasonably reproducible under a given set of experimental conditions.

2. Ep Dependence on Dynamic Versus Static Polarization:


Induction and Statistical Effects
Differences between Ep measurements from dynamic and static polarization methods are mainly associated with the induction time needed for pits
to initiate. Several authors have studied this phenomenon using various electrochemical techniques, and reviews are provided elsewhere (117,118).
Foroulis and Thubrikar in the mid-1970s (98,119,120) evaluated the influence of various factors on Ep values for 99.99% pure aluminum in de-aerated
aqueous potassium chloride solutions with concentrations in the range 0.01M
3M and temperatures in the range 560C. In these studies, the aluminum surfaces were initially mechanically polished to produce a mirror finish and then
anodized using one of two methods to produce oxide films of known thickness.
Critical pitting potentials were determined using two methods, one potentiostatic and the other a quasi-potentiodynamic method. In a typical potentiostatic

210

Holroyd

experiment, the electrode potential would be increased anodically by a potential


step of 1020 mV and then held constant with the current continuously monitored
for up to 24 h if a significant current increase was not recorded. After this, the
aluminum electrode surface was examined for pit initiation using a low-power
optical microscope.
The critical pitting potential, Ep, from these experiments was defined as the
lowest potential that the pits were optically observed, which turned out also to
be the lowest potential where the current significantly increased after the potential
increment. A typical result from a static polarization experiment is given in
Fig. 22, which shows that for the particular conditions used, Ep is 0.81 V (SCE)
and that the current transient at Ep has an induction time, , of about 8 min
during which the current remains negligible prior to it increasing significantly at
longer times (98,119,120).
Induction-time effects lead to Ep values from potentiodynamic polariza-

Fig. 22 Typical potentiostatic polarization curves for pre-anodized 99.99% pure aluminum in 3M KCl. [Note the 8-min induction time for pitting at E 0.81 V (SCE).] (Data
from Ref. 120.)

Aluminum Alloys

211

tion studies being more noble than values obtained from potentiostatic methods.
The reason for this is that by the time the induction time has elapsed, the electrochemical potential in a dynamic experiment has moved on from the potentiostatic Ep to a more positive value. Typical examples of this effect are provided
by the Ep data published by Foroulis and Thubrikar (119,120) from potentiostatic
and potentiodynamic polarization experiments on pure aluminum (99.99%) in
aqueous potassium chloride solutions at various chloride concentrations for a
given temperature and at various temperatures for a given solution chloride concentration. These data are reproduced in Figs. 23 and 24, along with comparable
data from other studies.
Induction times are a common feature of pit initiation under potentiostatic test conditions. Published data indicate that it is a statistical variable
(121,122) with induction times, :
Decreasing with increasing chloride concentration (98,122), potential
(112,122), or temperature (98,122),

Fig. 23 Pitting potentials for pure aluminum as a function of chloride concentration


given by potentiostatic and potentiodynamic polarization methods. , from Ref. 120;
from Ref. 123; from Ref. 127; from Ref. 126; from Ref. 125.

212

Holroyd

Fig. 24 Effect of temperature on the pitting potential for pure aluminum in a 3M KCl
solution given by potentiostatic and potentiodynamic polarization methods. , from
Ref. 120; from Ref. 123.

Being independent of solution pH (98) or chloride concentrations below


0.5M (98)
(98) or being to Increasing with oxide film thickness above 1300 A
tally independent (123)
Shibata and Sudos study on pure aluminum (99.99%) in de-aerated 3.5% NaCl
(121), in addition to giving further data on potential sweep rate effects on Ep
measurements from anodic potentiodynamic polarization experiments, also provides excellent examples of the statistical nature of Ep measurements. Measured
Ep values in the form of cumulative probability plots for various potential sweep
rates and mean Ep values as a function of potential sweep rate for various temperatures (298323 K) are shown in Figs. 25 and 26. From these data, it is apparent
that the absolute value and the statistical spread of the Ep measurements both
increase with increasing potential scan rate, and the effects become less dominant
as the solution temperature increases. Similar statistical data have recently been
published by Sato and Newman (124) for pure aluminum (99.999%) in a 0.5M
NaCl solution at room temperature.

Aluminum Alloys

213

Fig. 25 Experimental pitting potentials determined obtained using various anodic potential sweep rates for pure aluminum in de-aerated 3.5% NaCl and plotted on normal probability paper. (Data from Ref. 121.)

3. Influence of Chloride Concentration, Solution pH, and


Temperature and the Nature and Thickness of Oxides
on Ep and ER
a. Chloride Concentration Although, it is true that static polarization methods provide more accurate Ep and ER measurements than dynamic
methods, for convenience most researches have employed dynamic polarization methods. No significant problems arise as long as dynamic data are used
to evaluate trends rather than making absolute judgments. With this in mind, we
can use published polarization data to characterize the effect of a range of variables on the characteristic pitting potentials for pure aluminum in aqueous environments.
The dependence of Ep on chloride concentration (shown in Fig. 23) is well
known (120,123,125127) and several authors have published relationships, for
instance (120):
Ep 0.80 0.120 log[Cl] V (SCE)

214

Holroyd

Fig. 26 Median values of the pitting potentials for pure aluminum in a de-aerated 3.5%
NaCl solution at various temperatures, plotted as a function of the square root of the anodic
polarization potential sweep rate. (Data from Ref. 121.)

b. Solution pH Until recently, it was generally accepted that Ep and ER


values are effectively insensitive to the solutions pH for pHs in the range 4
9. More recently, it has been recognized that the local solutions pH changes
can occur at aluminum surfaces during the electrochemical experiments used to
measure Ep and ER values and when these changes are compensated for the resultant Ep and ER values can be influenced (128). Lampeas and Koutsoukos (128)
showed that the cathodic polarization characteristics of 99.99% purity aluminum
in aqueous saline solutions are relatively insensitive to solution pH changes,
whereas those for anodic polarization are sensitive in the pH range 67, where
films are more protective. Interpretation of these data with respect to the initiation
of pitting needs further evaluation, particularly if initiation occurs beneath hydrated surface films, as has been suggested by several researchers (90,103109).
c. Temperature Potentiodynamic anodic polarization experiments on
99% purity aluminum (AA1100) and a range of other aluminum alloys were
conducted in the de-aerated synthetic seawater at various temperatures in the
range 25150C during the early 1970s in a program to evaluate the suitability
of the alloys for use in desalination applications (129,130). Findings from these
studies were as follows:
1. Free-corrosion potentials increase over the first 40 h or so exposure to

Aluminum Alloys

215

the aqueous environments with the stabilizing values increasing with


increasing temperature
2. Repassivation potentials, ER, are insensitive to soak time and decrease
linearly with increasing temperature, from 760 mV at 25C to 800
mV (SCE) at 150C
3. Critical pitting potentials, Ep, increase with soak time and the stabilization times and potential shifts increase significantly at higher solution
temperatures.
The observed temperature dependencies for free-corrosion potentials (FCP),
Ep and ER, are understandable if the latter potential is deemed independent of
surface filming characteristics, as opposed to the other two parameters being considered highly dependent on surface filming characteristics. This is justifiable
because (a) the FCP is associated with establishing and maintaining a surface
film, (b) Ep is associated with the local breakdown of surface films, and (c) ER
is defined as the potential below which activity no longer occurs when the local
film has previously already suffered breakdown.
B. Aluminum Alloys: Oxide Development and Corrosion
Initiation in Aqueous Environments
Effects alloying additions have on the filming characteristics of aluminum alloys
in aqueous environments and any subsequent corrosion behavior are dependent
on the additions themselves, the actual microstructural form(s) adopted in a given
alloy-temper condition, and the prevailing local environmental conditions for film
formation. Considerable research effort has been directed toward answering the
question, How do alloying additions modify the inherent corrosion resistance of
aluminum alloys? Suggested mechanisms include the following:
The intermetallic providing local anodic and/or cathodic sites
Generation of oxide weak spots and oxide flaws associated with second phases
Alloying additions entering and modifying the properties of passive
layers
Modified local solution chemistries developing within corrosion sites
influencing growth kinetics
The extremes a given alloying addition may introduce into an alloy microstructure are that the addition may be effectively insoluble in aluminum and therefore be present only as a second phase or an addition may be highly soluble and
be present in a single-phase solid solution. The most common situation for most
major alloying additions in commercial aluminum alloys (e.g., Cu, Mg, Mn, and
Li) is that the additions have a limited solid solubility in aluminum, and at low

216

Holroyd

concentrations, they are present in the form of a solid solution, whereas at higher
concentrations, they are also present as a second phase. For heat-treatable aluminum alloys, the situation is further complicated by the fact that heat treatment
may modify the microstructural form adopted in the alloy, and for non-heattreatable alloys, the same may be promoted by cold or warm working processes.
In view of the above situation, it is not too surprising that the precipitation
of electrochemically distinguishable second phases (e.g., CuAl2 in AlCu alloys
and Mg3Al2 in AlMg alloys) at preferential microstructural sites means that
alloys can suffer additional forms of localized corrosion to the pitting suffered
by unalloyed aluminum. These various forms of localized corrosion will be covered in the following discussion below, which will initially focus on binary aluminum alloys, move on to ternary alloys, and, finally, deal with the more complicated commercial aluminum alloys that contain multiple deliberate alloying
additions along with others present as impurities.
C. Binary Alloys
Many studies have been published on the corrosion of binary aluminum alloys
in aqueous environments. Initial studies concentrated on binary systems involving
the elements used as major alloying additions (i.e., AlCu, AlMg, AlZn, Al
Mn, and AlLi). In more recent times, in response to the goal of generating a
stainless aluminum alloy, corrosion studies have involved binary aluminum
alloys bases containing elements known to enhance the passivity of stainless
steels (e.g., Cr, Mo, Ta, Zr, Nb, and W). Because these elements all exhibit extremely low solid solubilities in aluminum (typically less than 1 at%), these researchers have had to employ novel techniques such as rapid solidification processes (e.g., splutter deposition or melt spinning) or ion implantation to generate
the alloys.
Detailed studies have also been conducted on various binary alloys containing alloying elements that activate aluminum (e.g., Hg, In, Sn, Zn) (57,131
133). This aspect of aluminum alloy corrosion will not be discussed in detail
here, as our focus is on aluminum alloys used in structural and engineering applications rather than those employed as sacrificial anodes or battery anodes.
The earliest corrosion studies on aluminum binary alloys not surprisingly
were conducted on the AlCu system, which, after being discovered in Germany
in 1906, resulted in the alloy duralumin and the first commercial exploitation of
an age-hardenable aluminum alloy in 1908 (134). It has been appreciated since
the early 1940s that the copper in a Al4% Cu (wt%) binary alloy may be retained
in solid solution if the alloy is solution heat treated and sufficiently rapidly
quenched. In this metallurgical state, the dissolved copper is thermodynamically
unstable and will preferentially precipitate as CuAl2 at grain boundaries if the
alloy is exposed to temperatures above around 120C or will have already precipi-

Aluminum Alloys

217

tated to some extent if the quench rate employed after solution heat treatment
was insufficient. This miscrostructural condition promotes a susceptibility to intergranular corrosion and researchers quickly realized that the localized corrosion
was related to the grain-boundary precipitation of CuAl2 (135,136). Dix and coworkers developed an electrochemical based model for the intergranular corrosion promoted in these situations. In this model (137), they suggest that the intergranular corrosion occurs along a narrow zone adjacent to the grain boundary
that is relatively depleted in copper and anodic to both the main body of the
grain and the CuAl2 precipitates on the grain boundary. (A schematic of the local
microstructure is given in Fig. 27.)
The weakness of the Dix et al. model for intergranular corrosion as it was
initially presented is that it relies solely upon whether a given phase is anodic
or cathodic with respect to another phase when exposed to a saline solution doped
with hydrogen peroxide. No consideration is given to the magnitude of the current
that would flow in a local cell or whether the poly-phase systems free-corrosion
potential would be greater or smaller than that needed to promote local corrosion.
Ideas presented by Galvele and his co-workers (138140) show that the Dix et
al. model for the intergranular corrosion of AlCu alloys can be justified if it
includes a consideration of the breakdown potentials of the phases present in the
system (141).

Fig. 27 Schematic representation of an aged AlCu (wt%) alloy microstructure.

218

Holroyd

An excellent overview of the corrosion behavior of AlCu binary alloys


in aqueous saline environments is provided by Muller and Galveles studies
(142). They established that a clear relationship existed between alloy microstructure, electrode potential, and corrosion behavior in de-aerated 1M sodium chloride at 25C as a function of alloy composition and thermal aging. The effect of
copper content on the pitting potentials of AlCu solid solutions is shown in Fig.
28 and the effect that aging time at 240C has on the pitting potential, alloy
hardness, and corrosion behavior for an Al3.3% Cu (wt%) alloy as a function
of electrode potential and the calculated copper content of the solute depleted
zone associated with the grain boundary is shown in Fig. 29. From these data,

Fig. 28 Pitting potentials for various aluminum binary alloys in a de-aerated 1M NaCl
solution at 25C. from Ref. 140; from Ref. 140; from Ref. 142.

Aluminum Alloys

219

Fig. 29 Corrosion mode, pitting potential, and alloy hardness for an A13.3% Cu (wt%)
as a function of aging time at 240C. (Data from Ref. 142.)

it should be evident that the observed corrosion mode is directly related to an


alloys microstructure and the prevailing electrochemical potential. For instance:
1. For an as-solution treated and rapidly quenched alloy microstructure,
the alloy is passive at potentials below the pitting potential and pitting
only occurs at higher potentials.
2. For aging times where the pitting potential for the grain-boundary regions is below that for the grain interiors, a potential regime will exist
where only intergranular corrosion occurs (Fig. 29).
3. After long aging times, the potential domain for intergranular corrosion
disappears (Fig. 29).

220

Holroyd

Based on the above, it would seem reasonable that suitably heat-treated AlCu
binary alloys should display two pitting potentials during anodic polarization
studies: one for the grain-boundary region and a second at a higher electrode
potential for the grain interiors. Such results have been reported by Urushino and
Sugimoto (143) for an Al4 Cu alloy aged at 170C and polarized in a de-aerated
1M NaCl solution with its pH adjusted to 10. These results are reproduced in
Fig. 30 and the good fit with the Muller and Galvele data is obvious.
Muller and Galveles corrosion studies also included work on the AlMg
and AlZn binary-alloy solid solutions (140). Significant differences were readily
apparent between the various alloy systems. Alloying additions up to around 5%
(wt) led to pitting potentials in an aqueous 1M NaCl solution displaying positive
shifts for AlCu alloys, small effects for AlMg alloys, and shifts in the negative
direction for AlZn alloys (Fig. 28). A further significant difference noted was
the electrochemical potentials for the intermetallic phases precipitating from supersaturated solid solutions and those for the denuded zones adjacent to these
precipitates. In the case of AlCu alloys, the intergranular precipitation of Al2Cu-

Fig. 30 Effect of aging time at 170C on the grain boundary and matrix pitting potentials
for an A14 Cu (wt%) in a deaerated 1M NaCl solution with its pH adjusted to 10. (Data
from Ref. 143.)

Aluminum Alloys

221

generated adjacent regions that were relatively anodic, whereas for the Mg3Al2
precipitation in AlMg alloys, the precipitate itself was relatively anodic; see
data given in Table 2.
Pitting attack morphologies for the AlMg alloys were crystallographic,
closely resembling that previously reported by Galvele and de Micheli (144) for
pure aluminum, and there was no evidence of any magnesium buildup within
the pits. The pitting attack for the AlZn alloys was dependent on an alloys zinc
concentration, and microprobe studies indicated that zinc accumulation occurred
within pits. For low zinc levels, the pitting was crystallographic; however, as the
zinc level increased, the pits became more irregular and more tunnellike, probably
with some subsurface propagation (140).
Based on results from their experiments involving scratching oxide surfaces
during anodic polarization experiments, Muller and Galvele (140) concluded that
the oxides forming on various aluminum alloys had different properties. The
presence of oxide films on pure aluminum, AlCu, and AlZn did not interfere
with or accelerate the initiation of pitting, whereas films forming on the AlMg
alloys acted as a barrier to the pitting process, with pits for these alloys, unlike
the others, nucleating preferentially on the scratch lines.
Muller and Galvele (140) noted during their polarization studies on the
higher magnesium binary alloys (Al3 Mg and Al5 Mg) that the measured
currents became unsteady at potentials just below the pitting potentials. The ob-

Table 2 Published Pitting Potentials, Ep, for Various Aluminum Binary Alloys and
Aluminum-Based Precipitates in De-aerated 1M NaCl at 25C and 53 g/L NaCl 3
g/L H2O2

Alloy

Pitting potential of
de-aerated 1M NaCl
[V (SCE)] (Ref.)

Al2 Cu
Al4 Cu

0.774
0.777
0.654
0.594

Al1 Zn
Al3 Zn
Al3 Mg
Al5 Mg
Al2Cu
Al3Mg2

0.864 (140)
0.994 (140)
0.814 (140)
0.824 (140)
0.654 (144)
0.994a

Pure aluminum

Extrapolated from data given in Ref. 145.

(144)
(145)
(142)
(142)

Pitting potential
53 g/L NaCl 3 g/L H2O2
[V (SCE)] (Ref.)
0.764 (144)
0.644 (141)
0.598 (135)
0.604 (141)
0.874 (141)
0.994 (141)
0.784 (141)
0.794 (141)
0.644 (6, 144)
0.978 (135)

222

Holroyd

served transient peak currents were two to three times those needed to maintain
the passive film and occurred at increasing frequencies as the potential approached the alloys pitting potential. These authors did not comment on whether
these effects also occurred during their experiments on AlZn binary alloys; however, as will be discussed later, these effects are now known to occur for Al
Zn binary alloys in aqueous saline solutions (124,145).
Perhaps the most significant outcome from the above studies by Galvele
and his co-workers is that the results are consistent with his previously proposal
pitting model, where the pitting potential is the minimum potential for sufficient
localized acidification to be maintained at the metalsolution interface within a
pit (138). According to Galvele (138), the pitting potential is given by
Ep E p* Einh

(1)

where Ep is the pitting potential, E p* is the pitting potential of the alloy in a pitlike
environment, is the overpotential necessary to draw a net anodic current
through the pit, is the potential gradient through the pit, and Einh is the contribution due to inhibitors present in the solution.
Substitution of appropriate values for 99.99% purity aluminum, Al3 Cu,
Al3 Mg, and Al3 Zn into Eq. (1) yields the results given in Table 3. The
predicted pitting potentials are in excellent agreement with those determined experimentally. [Values for E p* are based the alloys corrosion potential measured
in a saturated AlCl3 solution after 2 h exposure (140). Values for were taken
as the polarization above the corrosion potential necessary to maintain an alloys
anodic current density at 1 mA/cm2 in a saturated AlCl3 solution; see Table 3.
for a 1M NaCl solution is approximately 0.050 V (138) and because there are
no inhibitors Einh 0.]
Sato and Newman (145) have recently published elegant experimental studies on the role zinc additions up to 5% (wt) play during the activation of pitting
corrosion for high-purity binary AlZn alloys in a de-aerated aqueous 0.5M NaCl
solution. Their results are consistent with Zn effects on the pitting potential being
the following:

Table 3 Comparison of Pitting Potentials Calculated Using Eq. (1) with


Experimental Values Obtained in De-aerated 1M NaCl at 25C
Alloy
99.99% aluminum
Al3 Cu
Al3 Mg
Al3 Zn

E *p

ETheory

Ep

1.00
0.76
0.10
1.06

0.17
0.04
0.15
0.01

0.05
0.05
0.05
0.05

0.78
0.67
0.80
1.0

0.77
0.64
0.81 to 0.77
0.99

Source: Data from Refs. 140 and 142.

Aluminum Alloys

223

1. Predictable from Zns effects on dissolution kinetics in pitlike environments


2. A pure Galvele effect adhering to his local acidification model of
the pitting potential
3. Independent of zinc needing to modifying properties of the surface
films
Furthermore, their results clearly showed the following:
1. The same number of potential pit sites exist irrespective of a binary
alloys zinc level (presumably they depend on the nature of the oxide
and sites generated with time by passive dissolution).
2. Zinc additions have little or no effect on pit nucleation or its frequency,
but effects are solely associated with pit propagation by the enhancement of local dissolution kinetics within the active pits.
3. Zinc has no direct effect on the local film breakdown and repair events
occurring at potentials either just below or just above the measured
pitting potential.
4. Zincs role is to act as an activator during the propagation stage of
pitting making pits grow faster and, in the case of metastable pitting,
allowing pits to exist for longer times and in some circumstances grow
sufficiently to aid their transition to stability.
Justification of the above statements for the AlZn binary alloys containing up
to 0.13% Zn (wt) was that the pitting potentials and repassivation potentials for
the binary alloys were statistically indistinguishable from those for 99.999% purity aluminum (Fig. 31), as was the frequency of the metastable pitting events
as a function of exposure time in the saline solution. (The 0.13% Zn addition,
however, did extend the lifetime of the pits and so zinc was activating aluminum
dissolution.) Justification for the higher-zinc-containing binary alloys was not
possible by the direct comparison of pit nucleation frequency with those for pure
aluminum at a given potential because no events occur on pure aluminum at the
lower potentials associated with the pitting potentials of the higher-zinc-containing binary alloys. However, pit nucleation behavior was compared with pure
aluminum by measuring the standard deviation of the passive current, which indicated passive film breakdown associated with pit nucleation in the presence of
the chloride ion (145). In these experiments, the authors showed that the current
noise for pure aluminum exposed to an aqueous sodium borate solution containing 0.5M NaCl decreased to that of the borate solution without the chloride
addition when the potential was decreased below the pitting potential of a Al
2 Zn alloy. This, the authors claim, successfully confirms that pitting of AlZn
alloys is the result of nucleation events that occur with about the same potential
dependence and possibly the same frequency as on pure aluminum.

224

Holroyd

Fig. 31 Normal probability plot of the pitting potential for superpure (99.999%) aluminum and A10.13 Zn (wt%) in a de-aerated aqueous 0.5M NaCl solution. (Data from
Ref. 124.)

Aluminum Alloys

225

Corrosion studies on AlLi-based alloys increased when the aerospace industry became interested in the alloy systems potential promise of a 10% increases in stiffness combined with a 10% density reduction over the conventional
high-strength aluminum alloys (146). Published work on AlLi binary alloys
indicates that the corrosion resistance in saline solutions is no worse than that
observed for pure aluminum. Corrosion occurs as crystallographic pitting and is
insensitive to thermal aging, irrespective of whether the AlLi binary alloy is
single phase, with the lithium totally in solid solution, or the lithium also present
in the alloy as the normal age-hardening precipitate, , Al3Li (147,148). The
only exception is when AlLi binary alloys are subjected to extended overaging
heat treatments and enhanced corrosion rates are observed with pits initiating
and spreading from grain boundaries (147). The latter has been attributed to the
extended overaging promoting grain-boundary precipitation of , AlLi, a highly
anodic phase (149,150).
The reason why, unlike AlCu and AlMg binary alloys, the corrosion
behavior of AlLi binary alloys are insensitive to the alloy addition being either
in solid solution or precipitated as a hardening phase is twofold. First, the Al3Li
phase promoting age hardening is a coherent phase and the electrochemical potential differences between it, the adjacent solute-depleted zones, and the grain interiors are believed to be minimal and, second, lithium dissolution can lead to
beneficial modifications to the surface films generated in aqueous environments.
Over the last 20 years, attempts have been made to develop stainless
aluminum alloys that will offer high resistances to localized corrosion in chloride
environments. The approach adopted has been to use nonequilibrium methods
that will simultaneously do the following:
1. Eliminate the second phases, usually present in conventionally cast
alloys, that are known to be detrimental to corrosion resistance
2. Produce supersaturated alloys containing elements thought to enhance
passivity
Nonequilibrium methods employed to introduce a wide range of elements in aluminum include ion implantation (109,151153) and rapid solidification methods
such as melt-spinning (154) and sputter deposition (155161).
One of the earliest studies involved using ion implantation to introduce
Mo ions or Ar ions into 99.99% pure aluminum and a high-strength AlZn
MgCu alloy, 7075-T6 (151). Implantation of Mo ions into the pure aluminum
promoted a measurable improvement in the pitting corrosion resistance in an
aqueous saline environment with the pitting potential, Ep, increasing by around
200 mV, whereas for 7075-T6, the effect was less pronounced with the Ep only
increasing by around 100 mV. The authors suggested that these beneficial effects
were due to either the incorporation of Mo into the passive film or the reprecipitation of some Mo-containing species on the passive film. Effects due to Ar ion

226

Holroyd

implantation into either pure aluminum or the 7075-T6 were limited to small
increases in the measured Ep values which were attributed to a slight thickening
of the air-formed oxide films induced by the ion implantation.
Corrosion data in de-aerated 0.5N NaCl has been provided by Yoshioka et
al. (154) for a wide matrix of rapidly solidified aluminum binary alloys containing
2, 4, and 6 (at%) Mg, Ti, Mn, Cr, Fe, Ni, Cu, Zn, Zr, and Si produced by a meltspinning process. Pitting potentials for all the alloys were significantly ennobled
(and for pure aluminum) except for the alloys containing Mg, Fe, and Zn. Corrosion rates at the free-corrosion potential showed similar trends that reflected the
observed decreases in both the anodic and cathodic current densities. These beneficial effects were attributed to both the supersaturation of the aluminum solidsolution phase with solute atoms and to the elimination or decrease in the number
of intermetallic phases available to initiate pitting.
Extensive work studying the evolution of passive film chemistry on a range
of sputter-deposited AlX binary alloys, where X included Cr, Ta, Mo, and W,
was conducted at the Martin Marietta Research Laboratories by Moshier and
coworkers during the mid-1980s and the early 1990s (155159). Beneficial effects on pitting resistance promoted by the various additions should be obvious
from Fig. 32, showing a compilation of anodic polarization curves from these
studies for pure aluminum and a range of the supersaturated aluminum binary
alloys obtained in 0.1M NaCl. X-ray photospectroscopy (XPS) studies on the
passive films formed on AlMo, AlCr, and AlTa alloys after exposure to a
0.1M NaCl solution revealed that the films contained 510% of the oxidizing
solute as MoO4, CrOOH, and Ta2O5, respectively, with the concentrations increasing fivefold as the overpotential increased from the open-circuit potential
to the pitting potential (156,157). On the basis of these results, the authors proposed that the enhanced pitting corrosion resistance resulted from the above species somehow impeding chloride ingress through the passive film. A different
explanation was invoked for AlW alloys because although W was very effective
in improving the pitting resistance (Fig. 32), the XPS studies failed to detected
any W in the passive films. Here, it was proposed that an undefined synergistic
effect between W and Al2O3 somehow led to a more stable oxide layer at the
metaloxide interface (158).
Natisham et al. (152) offer an alternative hypothesis to account for the
beneficial effects on the pitting corrosion resistance found for a range of aluminum binary alloys produced by an ion-implantation method. Here, the authors
assume that the rate-determining step (RDS) for pitting is the adsorption of chloride ions on to an alloys filmed surface and, hence, it is critical that the oxide
film covering the alloy must be positively charged so that chloride ions will be
adsorbed. As evidence supporting this proposal, they presented a correlation between the pitting potentials for Ta, Ti, Cr, Cu, Fe, Al, Zn, and Mg and the pHZCH

Aluminum Alloys

227

Fig. 32 Anodic polarization curves for various sputter-deposited aluminum binary


alloys in an aerated aqueous 0.1M NaCl solution. (Data from Ref. 158.)

(pH of zero charge) for the metals hydrated oxide in an aqueous 1N NaCl solution
(152). More recent studies have questioned the validity of Natisham et al.s mechanistic interpretation and the relevance of the correlation between the pitting potential and pHZCH. Perhaps the most convincing of these arguments is provided
by the work of Vijh (162), who suggests that the correlation between Ep and pHZCH
exists because a linear relationship also exists between pHZCH and the metalmetal
bond energy b(MM) (163). Vijh suggests that the critical relationship is the one
between Ep and b(MM) (Fig. 33) and not the one cited by Natisham et al. The
acceptance of b(MM) as the dominant parameter rather than pHZCH is relatively
straightforward if the RDS for pitting is attributed to a process such as the devel-

228

Holroyd

Fig. 33 A plot of pitting potentials of various aluminum alloys against the metalmetal
bond energy b(MM) values of the alloying element in the binary alloy. (Data from Ref.
162.)

opment of appropriate local solution conditions within a pit rather than one involving an aggressive species surface adsorption on to or its transport through
a passive film.
Szklarska-Smialowska (164) has proposed yet another mechanistic interpretation. The RDS for stable pit growth in her model is deemed to be the establishment of an appropriate solution chemistry within a pre-existing defect in the
film that can support active dissolution of the metal at the metaloxide interface.
This is a variant of the Galvele model for pitting (138), with the main difference
being an extra caveat in that the envisioned role for the alloying addition is to
establish the level of acidification needed to dissolve the alloying additions oxide

Aluminum Alloys

229

in the pit environment. For instance, in this model, an alloying addition can either
cause Ep to increase by providing an oxide that is less soluble than alumina in
an acidified pit environment (e.g., Cr and W oxides) or to decrease Ep by providing an oxide that is more soluble. Justification for the passive films themselves
not being involved in the RDS is based on the following premise:
1. The observation that the passive current densities for pure aluminum
and AlX alloys in a given saline environment are effectively identical
(see data in Fig. 32) and, hence, so are the physical and/or electrical
properties of the passive films.
2. The passive films forming on pure aluminum in aqueous saline environments are not effective obstacles to the penetration of chloride ions
and water to the metal surface (165)
3. Support for item 2 is provided by the immediate generation of electrochemical current noise when aluminum is immersed in chloride solutions as reported by Uruchurtu and Dawson (166)
Inturi and Szklarska-Smialowska (165) used the model to explain the beneficial
effects on the pitting characteristics of sputter-deposited Al4 Cr, Al9 Cr, Al
10 Ta, and Al17 Ta (at%) alloys exposed to a de-aerated 0.1M NaCl solution.
The observed open-circuit potentials (OCP) and the passive current densities for
these alloys were similar to pure aluminums, other than the OCP for the Al
17 Ta alloy, which was more positive. This was attributed to the Al17 Ta alloys
air-formed oxide containing Ta5 species as well as Al3 species, whereas Auger
electron spectroscopy only detected Al3 cationic species in the air-formed oxide
films on the other alloys.
Vijh (162) has suggested that the correlative trend between the aluminum
binary-alloy pitting potentials and the solubility of the alloying element oxide
quoted by Szklarska-Smialowska (164) is heavily weighted by the presence of
W and Ta among the nine elements studies. He presents an alternative interpretation of the data presented by Inturi and Szklarska-Smialowska (165) based on a
Galvele-type model with active dissolution of bare metal initiating in Richardson
and Wood (81,82) type of microfissures and defects in the oxide films. In support
of his interpretation, Vijh demonstrates that the pitting potentials of the binary
aluminum alloys quoted by Inturi and Szklarska-Smialowska (165) show the expected relationship with the solid-state cohesion metal-to-metal bond energies
b(MM) of the alloying element (163); see Fig. 33.
Some of the controversies outlined above may be reconciled using data
and ideas from a study on the pitting resistance of sputter-deposited aluminum
alloys presented by Frankel et al. (161). These authors used thin-film sputter thick) and measured pitting potentials, Ep, and
deposited alloys (10002000 A
the repassivation potentials, ER, in an aerated aqueous 0.1M NaCl solution. The

230

Holroyd

alloys evaluated included AlNb, AlMo, AlCr, and AlW with solute concentrations up to around 35 (at%).
A major advantage for the Frankel et al. study was that the thin films used
in the corrosion studies led to two-dimensional pits that accurately simulate the
behavior of small three-dimensional pits (167). This overcomes the usual problems encountered with pit depth and ohmic potential drop and/or diffusional path
issues and facilitates the generation of highly reproducible repassivation potentials, ER (167). Important findings from this study include the following:
1. For a given alloy system, the anodic pit current densities just prior to
repassivation do not increase with the pitting resistance or an alloys
solute content.
2. Although the pitting potentials of many of the sputter-deposited alloys
were significantly higher than that for pure aluminum, the Ep for a
given alloy was only slightly higher than its repassivation potential.
Based on these findings, the authors concluded that although alloy solute enrichment in the alloys influenced the local environmental conditions needed for pit
growth, any effects on the passive films themselves were relatively minor, even
when the beneficial effects due to film modification from long-term aging
(laboratory air exposure for 4 years), displayed by some of the alloying systems,
was taken into account.
In essence, the work by Frankel et al. provides a link between the nonequilibrium aluminum binary alloys and the earlier pioneering work of Galvele and
his co-workers on conventional cast-aluminum binary alloys.
D. Ternary and Commercial Alloys
As discussed earlier, the corrosion behavior of AlLi binary alloys in saline solutions is independent of alloy heat treatment other than when alloys are grossly
overaged and ,AlLi, a highly anodic phase, forms at grain boundaries
(149,150). Copper additions to AlLi alloys significantly modifies this situation
with thermal aging at temperatures around 170C promoting copper-containing
precipitation at subgrain boundaries, T1(Al2CuLi) followed by T2(Al6CuLi3) at
high-angle grain boundaries when tempers approach peak strengths (168). The
intergranular corrosion susceptibility for both AlLiCu alloys and the commercial alloy 2090 have been explained by the selective dissolution of a copperdepleted zone (147), as previously proposed for AlCu alloys (137,142). The
situation for AlLiMgCu alloys (e.g., 8090) is slightly different, with the presence of magnesium leading to different precipitates being developed during
aging. The influence of alloy temper on the mode of localized attack and pitting
potentials for these alloys, including 2091 (169), however, closely resembles that

Aluminum Alloys

231

Fig. 34 Effect of aging time at 170C on the grain boundary and matrix pitting potentials
for 2024 in a de-aerated 1M NaCl solution adjusted to 10. (Data from Ref. 143.)

observed for the AlCu and the AlMgCu alloy systems, (compare Figs. 30,
34, and 35 for Al3.3 Cu, 2024, and 8090, respectively).
Similar trends have been reported for AlZnMgCu alloys by Maitra and
English (170). They report that the temper dependence of the pitting potentials
for the AlZnMgCu alloy 7075 in an aqueous saline solution is remarkably
similar to that reported by Muller and Galvele (142) for AlCu alloys. Their
anodic polarization studies on 7075 in a de-aerated 3.5% NaCl solution show the
following:
1. A single pitting potential was detected for the solution heat-treated and
rapidly quenched condition.
2. Two pitting potentials existed for the peak aged temper, the less noble
one being associated with the local breakdown in the grain-boundary
region and second corresponding to pitting of the matrix.
3. Sufficiently overaged alloy microstructures displayed a single pitting
potential that corresponded to pitting of the copper-depleted solidsolution matrix.

232

Holroyd

Fig. 35 Efect of aging time at 170C on the grain boundary and matrix pitting potentials
for 8090 in a de-aerated 3.5 (wt%) NaCl solution adjusted to 6.5. (Data from Ref. 148.)

Although the microstructural details of the precipitates formed differ for the various age-hardening aluminum alloy systems, the influence of alloy temper on the
modes of localized corrosion suffered are self-consistent for the AlCu(Mg)-,
AlZnMgCu-, and AlLi-based systems (i.e., 2xxx, 7xxx, and 8xxx series
alloys).

VI. CONCLUDING REMARKS: ROLE OF SURFACE FILMS


DURING THE CORROSION OF ALUMINUM ALLOYS
Based on available published information, it is reasonable to conclude that the
oxide layers existing on aluminum or its alloys will chemically react when
exposured to aqueous environments and this will be a precursor to any subsequent
corrosion process that may subsequently initiate.
This hypothesis is compatible with previous suggestions for the initiation
of localized corrosion because it is a global surface process occurring prior
to any localized processes such as blister development (104109, 152) and/

Aluminum Alloys

233

or the development of critical local solution chemistries at potential initiation


sites for pitting (138,164,171), grain boundary (137,142), or crevice (172175)
corrosion. These latter processes are usually associated with some heterogeneous
alloy microstructural feature at the metalfilm interface alloy surface beneath the
surface film or possibly associated with a flaw/defect in the oxide (81,82).
The chemical reaction experienced by the oxide film is highly dependent on the environmental conditions to which it is exposed. The nature of the
oxide and its thickness may influence these interactions with these effects
usually being kinetic, although there are instances where the type of pre-existing
oxide can determine whether corrosion initiates (11,12,15).
When aluminum or aluminum alloys are exposed to aqueous environments
(or water vapor), the oxide surface films are modified via hydration and/or
dissolution processes. The extent and details of these processes are dictated
by the ionic species present and the solution pH, temperature, pressure, and flow
rate of the environment.
For most instances when aluminum and its alloys are exposed to aqueous
solutions with pHs in the range 49, a porous hydroxide (often pseudo-boehmite
and/or bayerite) forms by nucleation and growth from soluble aluminum species
generated from the pre-existing oxide (29,39,73). The overall reaction is controlled by what happens to the dissolved aluminum species immediately after
immersion. Typical situations include the following:
1. Oxide hydration with growth and densification of a hydroxide layer
retarding reaction rates and eventually promoting self-stifling. A typical example of this is the filming of pure aluminum when exposed to
distilled water, as described earlier in this chapter.
2. Corrosion inhibition by insoluble complexes forming on the oxide surface and preventing further oxide hydration [as suggested by Vermilyea and Vedder (39)]. A good example of this is the filming behavior reported for pure aluminum in dilute aqueous phosphate solutions
(103M102M), where the hydration of the oxide is thought to be completely suppressed with no evidence of oxide dissolution other than
dissolved aluminum species interacting with phosphate species to form
zero-charge complexes that are stable on the oxide and hinder the penetration of water molecules through the oxide (176).
3. Oxide hydration initiates as in situation 1 but restricted geometry
(e.g., a tight crevice) promotes local concentration buildup of dissolved
aluminum species causing local solution pHs to become increasingly
acidic. Depending on the anions present, the possibility that film dissolution and modification occurs in some circumstances can lead
to the onset of localized corrosion. An excellent example of this is
given in the detailed study of the crevice corrosion of pure aluminum

234

Holroyd

in 1M NaCl conducted by Baumgartner and Kaecshe (175). These researchers showed the following:
(a) Crevice corrosion can be triggered by unstable micropitting at potentials as low as around 1.040 V (SCE), some 0.3 V below the
pitting potential Ep and at least 0.1 V (SCE) below the repassivation potential ER for the 1M NaCl bulk environment.
(b) Unstable micropitting occurred during the incubation period for
crevice corrosion, which also occurs when aluminum surfaces in
noncrevice situations are exposed to aluminum chloride or acidified NaCl solutions.
(c) When a critical solution chemistry is established inside a crevice,
the mode of attack switches to a more general type of attack, operating at a low current density of around 10 A/cm2.
Based on these findings, Baumgartner and Kaecshe (175) suggest that
the buildup of a critical acidic electrolyte is a sufficient requirement for
the initiation of crevice corrosion. Other researchers have disagreed,
suggesting that the requirement is either a critical concentration of aluminum ions that is the chloride ion concentration dependent (174) or
the presence of basic aluminum chloride (171).
4. Aggressive ionic species in the aqueous environment modify the oxide hydration process and gel-like anion-selective surface membranes form (90), allowing water entry and an inward migration of
specific anions. In this scenario, dissolved aluminum species (mainly
Al3) and aggressive anions (e.g., Cl) build up at the metalfilm interface beneath the gel-like films and lead to the initiation of localized
corrosion at active microstructural sites (Fig. 36). (It may be argued
that situation 3 is a special case of situation 4) for which the local
bulk environment inside a restricted geometry (e.g., a crevice or
crack) changes and then process 4 occurs.)
Recent time-lapse video studies (103) have shown that continuous gel-like
corrosion films form on aluminum alloy surfaces exposed to saline environments.
Gas bubble generation occurs at sites beneath these films and, in some cases,
leads to local eruptions and the exposure of sites where localized corrosion has
become established. Although it is clear that the mechanism of localized corrosion initiation needs further detailed study, it is reasonable to conclude that the
initiation of localized corrosion for aluminum and its alloys in aqueous environments normally occurs by either situation 3 or 4, irrespective of whether it occurs
under total immersion, thin-film, or wetdry environmental conditions.
Surface films may have several roles during the propagation phase of localized corrosion:

Aluminum Alloys

235

Fig. 36 Schematic representation of a anion-selective gel-like surface film that has


formed on an aluminum alloy surface and has facilitated local acidification and the possibility of localized corrosion initiation beneath the film (based on the idea presented by
Sato (90)) (not to scale).

1. Preventing extensive lateral spread and the initiation of general corrosion.


2. Maintaining the occluded nature of the corrosion site by providing a
total, partial, or intermittent seal-restricting migration, diffusion, or significant hydrodynamic flow into or out of the corrosion site.
3. A surface film coating the inside of a local corrosion site can effectively
limit localized corrosion rates.
The films fulfilling the above functions will tend to be different. An example of
this is shown schematically in Fig. 37, in which the surface films formed on the
surfaces of an aluminum alloy suffering localized corrosion in an aqueous saline
environments are envisaged to change as a function of the depth of a localized
corrosion site. Some experimental justification for this proposal exists (177,
178):

236

Holroyd

Fig. 37 Typical surface films and the solution is pHs developed as a function of depth
within restricted geometries for aluminum and its alloys exposed to saline solutions.

1. Le and Foley (177) measured the solution pH within stress corrosion


cracks for 7075-T651 in a saline solution and reported that the solutions pHs decreased as a function of crack length from 4.6 to 5.0 in
the precracked region, through a region with a pH of 4.24.6 with
evidence of corrosion product to the advancing crack-tip region with
a pH of 3.03.2.
2. Holroyd et al. (178) reported that the solutions pHs within artificial
crevices for aluminum alloys exposed to saline solutions vary with
both time and crevice depth (Fig. 37) and that the measured pHs
are consistent those known for the aluminum oxychloride products
(Table 4).

Table 4 Saturation Concentrations and Solution


pHs (at 90% Saturation) for Aqueous Solution of
Aluminium Oxychloride Salts (179) and Aluminum
Chloride (180)
Aluminum
oxychloride salt
Al(OH)Cl2
Al(OH)2Cl
Al2(OH)5Cl
AlCl3

Saturation
concentration
(M)

Solution pH
(90% sat.)

3.02
3.12
2.34
3.11

2.85
2.94
3.67
0.15

Aluminum Alloys

237

Further support for these ideas is provided by Wong and Alkire (179), who have
characterized solution chemistries developing within naturally occurring 100m-deep pits in 99.999% purity aluminum in 1M NaCl. Their results indicate
the following:
1. The solutions pHs in pits are between 3 and 4 and the acidity can
be explained by the hydrolysis of Al3 ions, as has been suggested by
others for solutions within pits (181), crevices (87,182,183), and cracks
(87,183188).
2. Eighty-four percent of the measured dissolve aluminum in the pits is
present as monomeric species and the remainder is present as polymeric species with no evidence of solid gels.
3. The hydrolyzed aluminum species react within the pit environment to
form basic salts that have nuclear magnetic resonance spectra very similar to those given by synthetic solutions of Al(OH)Cl2 or Al(OH)2Cl.
4. Current densities from single pits decreased roughly with the square
root of time and are consistent with the corrosion rate being controlled
by the rate of dissolution of an aluminum oxychloride salt film formed
by the precipitation of dissolution products.
Several authors (181,185,186,189,190) in the past have suggested that the
solutions pHs developed within restricted geometries (e.g., pits, crevices, and
cracks) for aluminum alloys exposed to aqueous saline environments are controlled by acid hydrolysis involving Al(OH)3. For instance, Pryor (181) suggested
that in practical situations, the solutions pH in aluminum pits would not fall
below 3.54.0 and he calculated a minimum possible pH of 2.8 using thermodynamic data for the chemical equilibrium process:
Al3 3OH Al(OH)3

(2)

A major flaw in the above argument is that the species Al(OH)3 is unstable in
solutions with pHs below 5 (87,177,191,192) and, therefore, it is unreasonable
to assume that pHs are being controlled by Eq. (2). Recent theoretical work by
Galvele (192) indicates that no solid-reaction products are expected to form in
aluminum crevice solutions with pHs of 3 and below, irrespective of bulk solution pHs, and only 10% of the products would be solid in a crevice solution
with a pH of 4.
Several researchers have suggested that the solutions pHs within restricted geometries are well explained by the hydrolysis of Al3 ions to form
AlOH2 and H ions via the equilibrium process (62,87,177,178):
Al3 H2O AlOH2 H

(3)

The more recent work of Wong and Alkire (179) is consistent with Eq. (3) but
indicates the difficulty in differentiating between Al(OH)2Cl and Al(OH)Cl2 and

238

Holroyd

determining which is dominant within restricted geometries. One possibility is


that neither is dominant and their balance within an occluded cell reflects the
local solutions pH. An example of this is shown schematically in Fig. 37 and
a qualitative description of the chemical equilibrium processes controlling the
local solution and film-formation processes within a restricted geometry for an
aluminum alloy exposed to an aqueous solution is given by
Al3 H2O

AlOH2 H

Cl

Cl

AlCl2

Al(OH)Cl
Al(OH)2Cl2

In view of the experimental evidence now available, the previous suggestions


that the presence of AlCl3 is needed to activate passive aluminum (171) or saturated AlCl3 must form at the bottom of pits for them to remain active
(126,171,193) are clearly inappropriate. Perhaps the most direct evidence for this
is the excellent agreement between the measured solution chemistries within restricted geometries [e.g., dissolved aluminum concentrations, the solutions pHs
and the NMR (nuclear magnetic resonance spectroscopy) data] and those measured for species other than AlCl3.
If it is accepted that saturated AlCl3 does not form at the base of restricted
geometries, we need to modify the data we input into the Galvele pit model earlier
in this chapter [see Eq. (1) and Table 3] so that the pitting potential, E *p , reflects
the actual environmental conditions developed within a pitlike geometry rather
than that developed when an aluminum alloy is exposed to a saturated aqueous
solution of AlCl3 for 2 h (140). Typical published data for electrochemical and
environmental conditions developed within aluminum crevices and cracks are
given in Table 5.
If we assume that the potentials developing at the base of active crevices
or propagating stress corrosion cracks are realistic values for E p* and substitute
into Eq. (1),
Ep E p* Einh

(1)

along with published pitting potentials and taking Einh 0, we can calculate
values for where is the overpotential needed to draw a net anodic current
through a restricted geometry and is the potential gradient within the restricted
geometry. Experimental work is needed to validate the ( )calc values given
Table 6.
Vermilyea and Vedder (39) studied the filming behavior of pure aluminum

Aluminum Alloys

Table 5

Environmental Conditions Developed Within Restricted Geometries for Various Aluminum Alloys

Restricted geometry

Alloy

Solution
pH

Internal
potential

External
potential

Al3 conc.
(M)

Cl conc.
(M)

Bulk NaCl
conc. (M)

Ref.

Crevices
1 mm diameter
10 mm deep
80 mm deep, 90 m wide
15 mm diameter 30 mm
Cracks
6 mm long
15 mm long
20 mm long
Pit
100 m deep

1199
2017
3003
6063
7475-T6
7475-T6
LC4-T6
7075-T6

4
4
3.5
3.5
4

7475-T6
LC4-T6

3.4
33.2
2.7
3.2

S.P. Al

34

0.830
0.825
0.844
0.916
0.940
0.870
0.920

0.925

0.780
0.773
0.890
0.780
0.790
0.830
0.850
0.832
0.830

0.003
0.25
0.0250.37
0.1

0.75

23
1.82

1.0
0.5
0.5
0.5
0.6
0.51
0.6
1.0
1.0
0.51
0.6

187
182
182
182
194
87
184
187
177
87
184
179

239

240

Holroyd

Table 6 Calculated Values for ( ) from Eq. (1) Using Published Data
Alloy
Pure Al
(in 1M NaCl)
3003
(in 0.5M NaCl)
7475-T6
(in 0.6M NaCl)

Ep (Ref.)

E*p (Ref.)

( )calc.

0.740 (222)

0.830 (187)
crevice
0.844 (182)
crevice
0.870 (194)
crevice

0.090

0.713 (222)
0.760 (170)

0.131
0.110

in boiling water containing 103M of various anions. Takahashi et al. (176) have
classified these results into four groups as a function of their ability to inhibit
hydration:
1. No inhibition: borate, MnO4, ClO4, NO3, NO2, Cl, and CO32
2. Moderate inhibition: IO3, SO42, SeO42, GeO42, CrO42, citrate, oxalate, and MNO42
3. Strong inhibition: Teo42, SiO42, WO42, AsO42, and IO42
4. Extremely strong inhibition: PO43
Generalization of these findings needs care because the behavior of some anions
are sensitive to their concentration and/or pH and temperature. Phosphate ions
is a good example. At low concentrations (103M102M), oxide hydration is
completely suppressed (as discussed earlier in this chapter); at intermediate concentrations (101M), oxide films initially suffer both dissolution and hydration
with a dissolutiondeposition process generating a surface film; at higher
phosphate concentrations, the dissolution of the oxide is accelerated by the formation of anion complexes that themselves hydrolyze to form Al-hydroxo-phosphate
complexes that then undergo polymerization to deposit basic phosphate or
hydrated oxides containing phosphate layers (176).
The presence of several anions in an aqueous solution may promote unexpected results with regard to the corrosion performance of aluminum alloys in
aqueous environments. Three examples follow.
Becerra and Darby (195) found that the corrosion rates for AA1100-H14,
AA5052-H32, and AA6063-T4 sheet materials all increased with additions of
sodium bicarbonate (0225 ppm) or cupric sulfate (02 ppm) added as single
additions to a 3.5 wt% NaCl solution. The increased corrosion rates were greater
than simply additive when the additions were added simultaneously. The synergistic effect was attributed to copper plating out on the aluminum alloy surface
and acting as a cathode [as suggested by Davies (196) and Fraker and Ruff (197)]
and the bicarbonates buffering capability limiting the local solutions pH

Aluminum Alloys

241

changes at the cathodes and reducing the tendency for local protective film formation.
Overaged AlZnMgCu alloys usually provide a good resistance to intergranular corrosion in aqueous saline solutions (170). Maitra and English (198)
have shown that small additions of nitrate and sulfate ions (particularly nitrate
ions) can cause 7075-T7351 to suffer intergranular corrosion during laboratory
testing. They also found that the relative susceptibilities promoted by the different
solution chemistries were reflected by the difference between the grain-boundary
and matrix pitting potentials. These results suggested the possibility that overaged
high-strength AlZnMgCu alloys, although being highly resistant in a saline
environments containing just chloride ions, may be prone to either intergranular
corrosion or stress-corrosion cracking (SCC) in certain industrial environments
(198). To date, there is no evidence of service performance issues with localized
corrosion occurring in overaged high-strength AlZnMgCu alloys. However,
it is interesting to note that the stress-corrosion performance of 7075-T7551 and
7075-T7351 during a extensive ongoing test program on various aluminum alloys
in rural, industrial, and seacoast environments (199) has identified that these alloy
tempers are significantly more susceptible to SCC in an industrial environment.
Yet another example of synergetic effects between anions influencing the
localized corrosion behavior of aluminum alloys is that reported for the AlLi
MgCu alloy (8090) and the AlLiCu alloy (2091) where small additions of
sodium sulfate to aqueous sodium chloride solutions have had marked effects
on the alloys SCC susceptibility (200,201). For 8090-T651, Craig et al. (200)
demonstrated that whereas 8090-T6 was resistant to SCC when under totalimmersion conditions in an aqueous 0.6M sodium chloride solution, it became
susceptible with small additions of sulfate ions but highly resistant with higher
levels of sulfate additions (Fig. 38). Similar findings have been reported by Marsac et al. (201) for 2091, albeit this alloy simultaneously suffers a susceptibility
to intergranular corrosion that seems to be independent of the solutions sulfate
concentration. A likely explanation of the sulfate effect at low concentrations is
that the surface filming process is modified by the introduction of a thermodynamically more attractive reaction path allowing the formation of aluminum oxysulfate species which will form in competition with the aluminum oxychloride
species, as has been described by Foley and Nguyan (202). For the high concentrations of sulfate ions, it could be that basic aluminum sulfate films form.
Based on their findings, Craig et al. (200) predicted that 8090-T651 would
suffer SCC when under total-immersion conditions in artificial or natural seawater. This was observed, and the susceptibility being lower than predicted on the
basis of the solutions sulfate-to-chloride ratio (Fig. 38) was attributed to filming
effects promoted by the additional soluble additives in seawater (e.g., magnesium
salts).

242

Holroyd

Fig. 38 Time to failure for peak aged 8090 stress-corrosion samples tested under totalimmersion conditions in a range of aqueous environments containing various concentrations of sodium sulfate and sodium chloride at 20C. (Data from Ref. 200.)

A. Hydrostatic Pressure Effects


To the authors knowledge, the available published literature on the effects of
hydrostatic pressure on the films formed during the localized corrosion of aluminum alloys is limited. Although most laboratory (203205) and field testing
(206208) studies indicate that the pitting susceptibilities of aluminum and its
alloys exposed to seawater generally increase with increasing hydrostatic pressure, data have been published showing that corrosion rates may be insensitive
or actually decrease with increasing pressure (209,210). In the late 1960s, the
increased pitting susceptibilities were attributed to hydrostatic pressure increasing
chloride ion activity (211) and its penetration into passive layers. It is now realized that the surface films are important and, in some instances, can control corrosion performance (205,210). A good example of the importance of surface films is
provided by Beccaria and co-workers (205,210), results showing that increasing
hydrostatic pressure increased the corrosion rate for pure aluminum exposed to
102 M Na2SO4 (205) but decreased corrosion rates for 6061-T6 exposed to seawater (210).

Aluminum Alloys

243

In the case of the pure aluminum exposed to aqueous Na2SO4 solutions,


Beccaria and Poggi (205) found that sulfate ion additions in the range 102M
101M reduced corrosion rates for all the hydrostatic pressures tested in the
range of 1300 atm. Corrosion rates in solutions containing 102M sulfate increased with increasing hydrostatic pressure, whereas corrosion rates in solutions
containing 101M sulfate were insensitive to hydrostatic pressure. The observed
behavior at the lower sulfate concentration was deemed independent of sulfate
ions and associated with hydrostatic pressure promoting the incorporation of aluminum hydroxide into surface films that would be more soluble at higher pressures and therefore less protective. For the higher sulfate concentration, the
authors suggested that aluminum sulfates formed in competition with the hydroxides and the net effect was that corrosion rates were insensitive to increasing
pressure.
Experimental results reported by Beccaria et al. (210) showing that the
corrosion rates for AA6061-T6 exposed to seawater decreased with increasing
hydrostatic pressure would seem to be contrary to their previous work and that
reported by others. Their explanation, based on surface film analysis using Xray photospectroscopy and infrared spectroscopy, was that hydrostatic pressure
promotes a more protective and less hydrated surface film containing magnesium
oxide and pseudo-boehmite, as opposed to a less protective hydrated bayerite
[Al(OH)3] film that forms at atmospheric pressure. These findings are extremely
interesting and deserve further evaluation because they suggest that hydrostatic
pressure can induce significant changes to the surface films forming on aluminum
alloys in aqueous solutions.
B. Influence of Flowing Conditions
Erosion corrosion of aluminum alloys in aqueous environments due to flowing
conditions usually become evident when flow rates exceed 3 m/s and the loss
of metal increases sharply with increasing flow rates in excess of about 9 m/s
(212,213). Localized corrosion is sensitive to solution flow rates below 3 m/s,
with evidence of low flow rates up to around 1.5 m/s being beneficial (214,215)
or at least not detrimental, and higher flow rates up to 3 m/s having minor or a
gradually increasing effect on corrosion rates (212,213). These effects may differ
depending on whether or not localized corrosion has initiated and is well established. In the case of corrosion initiation, low flow rates may lead to surface films
that prevent or delay initiation. For localized corrosion that is well established
and in a propagation mode, the influence of external solution flow is dependent
on whether solution chemistries are influenced within the local corrosion sites.
As with the hydrostatic pressure effects, the available literature on the effect
of flowing aqueous conditions on the localized corrosion of aluminum and its
alloys is limited (215,216). In theory, mathematical modeling is feasible for the

244

Holroyd

flowing conditions, despite the combination of localized corrosion, general corrosion, and bulk environmental flow, creating a complex set of differential equations that are difficult to solve (217).
Most published studies on the effect of flowing conditions on the corrosion
of aluminum alloys have involved marine environments (212216,218220) and
are associated with applications such as marine vessels, structural and nonstructural uses on offshore oil platforms, ocean thermal energy conversion (OTEC) heat
exchangers, and desalination plants. A notable exception being Witts (221) studies on the effects of flowing conditions have on aluminum alloy (AA6063 type)
domestic radiator systems.
Godard and Booth (218) reported that the pitting corrosion rates for seawater flowing through 1S-H18 aluminum (0.02 Cu, 0.21 Fe, 0.11 Si, 0.016 Ti) pipes
initially decreased with increasing flow rates and then increased with further increases in flow rate. Unfortunately, only relative flow rates are available. Beneficial effects of low flow rates on localized corrosion rates have also been reported
for various alloys used in OTEC heat exchangers (214,215) and in domestic hotwater radiator systems (221).
Larson-Basse and co-workers (214216,219) have evaluated the corrosion
performance of a range of aluminum alloys [AA3003 and 5052 and Alcald (7072)
3003 and 3004] in flowing seawater in OTEC heat exchangers using flow rates
in the range 1.42.4 m/s. The observed localized corrosion behavior in the tropical surface water differs significantly from that promoted in the cold, deep ocean
water (214,216). In the warm surface water, a surface film forms after an initial
short period (510 days) of relatively rapid corrosion which is comprised of both
scale minerals precipitated from the seawater and aluminum corrosion products
that are alloy dependent (214). This film is protective the corrosion rates fall to
around 3 m/year for all the alloys tested. Unfortunately, the film thickness increases in a parabolic manner and is sufficiently detrimental to the systems heattransfer characteristics that it has to be regularly removed. This introduces challenging problems with both material selection and choice of method to remove
scales (216).
Exposure of aluminum alloys to the cold, deep ocean water under flowing
conditions as used in OTEC heat exchangers introduces little or no tendency for
biofouling; however, the alloys usually suffer localized pitting, with the incubation times being sensitive to small variations in solution chemistry (214) and
pitting tendencies increasing significantly with decreasing flow rates (214).
Witt (221) has reported that residual water left in a hot-water radiator system for a few hours after a successfully operating system has been emptied can
locally disrupt the protective film that had previously provided protection, and
intergranular corrosion can result during the subsequent operation of the radiator
system. Once again, this provides clear evidence that the corrosion performance
of an aluminum alloy in a given aqueous environment is not always predictable

Aluminum Alloys

245

and can be dictated by the surface film already present or one forming upon
immersion in an aqueous environment.

REFERENCES
1. RB Mears. Aluminum and Aluminum Alloys. In: HH Uhlig, ed. Corrosion Handbook. London: J Wiley, 1948, p. 48.
2. WW Binger. Aluminum and its alloys. In: FL LaQue, HR Copson, eds. Corrosion
Resistance of Metals and Alloys, 2nd ed. New York: Reinhold Publishing Corp.,
1963, p. 181.
3. HP Godard. Aluminum. In: HP Godard, WB Jepson, MR Bothwell, RL Kane, eds.
The Corrosion of Light Metals. New York: J Wiley, 1967, p. 3.
4. JR Davis, ed. Corrosion of Aluminum Alloys. Materials Park, OH: ASM International, 1999.
5. DO Sprowls, RH Brown. Stress corrosion mechanisms for aluminium alloys. In:
RW Stechle, AJ Forty, D van Rooyen, eds. Fundamental Aspects of Stress Corrosion Cracking. Houston, TX: National Association of Corrosion Engineers, 1966,
p. 466.
6. RH Brown, DO Sprowls, MB Shumaker. The resistance of wrought high strength
aluminium alloys to stress corrosion cracking. In: Stress Corrosion of MetalsA
State of the Art. Philadelphia: American Society for Testing and Materials, STP
518, 1972, p. 87.
7. MO Speidel. Metall Trans 6A:631, 1975.
8. NJH Holroyd. Environment induced cracking of high strength aluminium alloys.
In: RP Gangloff, MB Ives, eds. Environment-Induced Cracking of Metals. Houston,
TX: National Association of Corrosion Engineers, 1989, p. 311.
9. TD Burleigh. Corrosion 47:89, 1991.
10. RP Gangloff. Corrosion fatigue crack propagation in metals. In: RP Gangloff, MB
Ives, eds. Environment-Induced Cracking of Metals. Houston, TX: National Association of Corrosion Engineers, 1989, p. 55.
11. HP Godard. Aluminum. In: HP Godard, WB Jepson, MR Bothwell, RL Kane, eds.
New York: Wiley, 1967, p. 11.
12. H Dolling, FE Faller, W Gruhl, E Cossack, G Scharf, T Singe. Erzmetallurgica,
32:161, 1979.
13. M Textor, R Grauer. Corros Sci 23:41, 1983.
14. A Csanady, T Turmezey, I Ime-Baan, A Griger, D Marton, L Fodar, L Vitolis.
Corros Sci 24:237, 1984.
15. O Lunder, K Nisancioglu. Corrosion 44:414, 1968.
16. K Wafers, C Misra. Alcoa Technical Paper No. 19 (Revised), Alcoa Laboratories,
Pittsburgh, PA, 1987.
17. PE Blackburn, EA Gilbransen. J Electrochem Soc. 107:944, 1960.
18. W Vedder, DA Vermilyea. Trans Faraday Soc. 65:561, 1969.
19. CN Coehran, WC Sleppy. J Electrochem Soc. 108:322, 1961.
20. AJ Brock, MJ Pryor. Corros Sci 13:199, 1973.

246

Holroyd

21.
22.
23.
24.
25.

AF Beck, MA Heine, EJ Caule, MJ Pryor. Corros Sci 7:1, 1967.


PE Doherty, RS Davies. J Appl Phys 34:619, 1963.
K Wafers. Aluminium 57:722, 1981.
GM Scamans, EP Butler. Metall Trans 6A:2055, 1975.
DJ Field. Oxidation of aluminum alloys. In: AK Vasudevan, RD Doherty, eds.
Aluminum AlloysContemporary Research and Applications. New York: Academic Press, 1989, p. 523.
A Csanady, D Marton, I Geleji-Neubauer, S Hofmann, JM Sanz. Corros Sci 22:
689, 1982.
RK Wiswanadham, TS Sun, JAS Green. Corrosion 36:275, 1980.
JAS Green, RK Wiswanadham, TS Sun, WG Montague. Grain Boundary Segregation and Stress Corrosion Cracking of Aluminum Alloys, Proceedings of Corrosion
77 Conference, San Francisco, 1977, Paper 17.
RS Allwitt. The aluminumwater system. In: JW Diggle, ed. Oxide and Oxide
Films. New York: Marcel Dekker, 1976, p. 169.
RS Allwitt. J Electrochem Soc 121:1322, 1974.
RS Allwitt, LC Archibold. Corros Sci 13:687, 1973.
GM Scamans, AS Rehal. J Mater Sci 14:2459, 1979.
RK Hart. Trans Faraday Soc 53:1020, 1957.
DG Altenpohl. Corrosion 18:143t, 1962.
DG Altenpohl. Aluminium 31:10, 1955.
DG Altenpohl. Aluminium 31:62, 1955.
H Leidheisser, E Kellerman. Corrosion 26:99, 1970.
RC McCune, RL Shilts, SM Ferguson. Corros Sci 22:1049, 1982.
DA Vermilyea, W Vedder. Trans Faraday Soc 66:2644, 1970.
JE Draley, WE Ruther. Corrosion 12:441t, 1956.
RK Hart, JK Maurin. Corrosion 21:222, 1965.
FH Krenz. Corrosion 13:575t, 1957.
VH Trouter. Corrosion 15:9t, 1959.
DG Altenpohl. Z.F. Metallkunde 48:306, 1957.
JE Draley, WE Ruther. Corrosion resistant aluminum alloys above 200C, Argonne
National Labs. Report ANL-5430, July 1955.
RL Dillon, HC Bowen. Corrosion 18:406t, 1962.
DF MacLennon. Corrosion 17:181, 1961.
DF MacLennon. Corrosion 17:239, 1961.
WE Tragget. J Electrochem Soc 106:903, 1959.
JH Greenblatt, AF McMillan. Corrosion 19:146, 1963.
JH Greenblatt. Corrosion 19:295, 1963.
DR Dickinson. Corrosion 21:19, 1965.
E Deltombe, M Pourbaix. Corrosion 14:496t, 1956.
DD MacDonald, P Butler. Corros Sci 13:259, 1973.
RL Dillon, VH Troutner. Atomic Energy Commission Report HW-51849,
1957.
NJ Fitzpatrick, GM Scamans. New Scientist 111(1517):34, 1986.
MC Reboul, PH Gimenez, JJ Rameau. Corrosion 40:366, 1984.
WH Thomason. Mater Perform 24:20, 1985.

26.
27.
28.

29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.

Aluminum Alloys

247

59. VE Carter, HS Campbell. Br Corros J 4:15, 1969.


60. DJ Scott. Trans Inst Metal Finishing 49:111, 1971.
61. MM Salama, RJ Myers, WH Thomason, MH Joosten. US Patent 4,619,577, Oct
28, 1986.
62. K Sotoudeh, TH Nguyen, RT Foley, BF Brown. Corrosion 37:358, 1981.
63. F Tirbonod, C Fiaud. Corros Sci 18:139, 1978.
64. SM Hassan, YA Elawady, AI Ahmed, AO Baghlaf. Corros Sci 19:951, 1979.
65. SM Hassan, MNH Moussa, FIM Taha, AS Fouda. Corros Sci 21:439, 1983.
66. JD Talati, RM Modi. Corros Sci 19:35, 1979.
67. H Scholl, MM Davila Jimenez. Corros Sci 33:1967, 1992.
68. E McCafferty, P Trzaskoma-Paulette. Corrosion 53:179, 1997.
69. RT Foley. Corrosion 42:277, 1986.
70. L Tomcsanyi, K Varga, I Bartik, G Horanyi, E Maleczki. Electrochim Acta 34:
855, 1989.
71. GC Wood, JA Richardson, MF Abd Rabbo, LB Mapa, WH Sutton. In: RP Frankenthal, J Kruger, eds. Passivity of Metals. Electrochemical Society, Pennington, NJ,
1973, p. 989.
72. J Augustynski. In: RP Frankenthal, J Kruger, eds. Passivity of Metals. Electrochemical Society, Pennington, NJ, 1973, p. 898.
73. TH Nguyen, RT Foley. J Electrochem Soc 127:2563, 1980.
74. DM Drazic, SK Zecevic, RT Atanasoski, AP Despic. Electrochim Acta 28:751,
1983.
75. WC Moshier, GD Davis, JS Ahearn. Corros Sci 27:785, 1987.
76. HB Kono, S Kobayashi, H Takahashi, M Nagayama. Corros Sci 22:913, 1982.
77. WC Moshier, GD Davis. Corrosion 46:43, 1990.
78. AKh Bairamow, S Zakiour, C Leygraf. Corros Sci 25:69, 1985.
79. GA Parks. Chem Rev 65:177, 1965.
80. L Campanella F Croce, P Mazzoni. Oberflaeche Surface 19:224, 1978.
81. JA Richardson, GC Wood. Corros Sci 10:313, 1970.
82. JA Richardson, GC Wood. J Electrochem Soc 120:193, 1973.
83. GM Brown, K Shimizu, K Kobayashi, GE Thompson, GC Wood. Corros Sci 34:
2009, 1993.
84. KF Lorking, JEO Mayne. J Appl Chem (London) 11:170, 1961.
85. KF Lorking, JEO Mayne. Br Corros J 1:181, 1966.
86. K Videm. The electrochemistry of uniform corrosion and pitting corrosion of aluminum. Kjeller Report KR-149, Institute for Atommenergi, Norway, 1974.
87. NJH Holroyd, GM Scamans, R Hermann. Environmental conditions within crevices
and stress corrosion cracks in aluminium alloys. In: A Turnbull, ed. Corrosion
Chemistry within Pits, Crevices and Cracks. London: HMSO, 1987, p. 495.
88. SC Britton, UR Evans. J Chem Soc (London) 1773, 1930.
89. WJ Bernard, JJ Randall. J Electrochem Soc. 108:822, 1961.
90. N Sato. Corros Sci 27:421, 1987.
91. L Liepina, Z Osis. Latvijas PSR Zinatnu Akad Vestis No. 6 (Whole No. 35): 35,
1950; No. 8 (Whole No. 61):107, 1952.
92. L Liepina, A Vaivade, Z Osis, A Striprais. Latvijas PSR Zinatnu Akad Vestis No.
3 (Whole No. 80):107, 1954.

248

Holroyd

93. L Liepina, A Vaivade. Latvijas PSR Zinatnu Akad Vestis Kim Ser No. 3:297, 1963;
No. 6:683, 1963; 441, 1964; No. 4:527, 1964.
94. L Liepina, V Kadek. Corros Sci 6:177, 1966.
95. L Liepina. Latvijas PSR Zinatnu Akad Vestis No. 3:55, 1967.
96. TP Hoar. Corros Sci 7:341, 1967.
97. JC Grossman. J Electrochem Soc 116:1232, 1969.
98. ZA Foroulis, MJ Thubrikar. J Electrochem Soc 122:1296, 1975.
99. RT Foley. J Electrochem Soc 122:1493, 1975.
100. GT Burstein, RJ Cinderey. Corros Sci 33:475, 1992.
101. GT Burstein, RJ Cinderey. Corros Sci 3:1195, 1991.
102. PMA Sherwood, JA Rotole. U.S. Patent 6,066,403, May 23, 2000.
103. CJ Newton, NJH Holroyd. Time lapse-video techniques in corrosion testing aluminum alloys. In: VS Agarwala, GM Ugiansky, eds. New Methods for Corrosion
Testing of Aluminum Alloys. Philadelphia: ASTM, STP 1134, 1992, p. 153.
104. CB Bargeron, RB Givens. J Electrochem Soc 124:1845, 1977.
105. CB Bargeron, RB Givens. Corrosion 36:618, 1980.
106. CB Bargeron, RB Givens. J Electrochem Soc 124:1230, 1977.
107. CB Bargeron, RB Givens. J Electrochem Soc 129:340, 1982.
108. CB Bargeron, RC Benson. J Electrochem Soc 127:2528, 1980.
109. PM Natishan, E McCafferty. J Electrochem Soc. 136:53, 1989.
110. M Pourbaix. Corrosion 26:431, 1970.
111. A Broli, H Holtan. Corros Sci 13:237, 1973.
112. A Broli, H Holtan. Werkstaffe Korros 26:629, 1975.
113. A Broli, H Holtan. Corros Sci 17:59, 1978.
114. K Nisancioglu, H Holtan. Corros Sci 18:835, 1978.
115. K Nisancioglu, H Holtan. Corros Sci 18:1011, 1978.
116. DA Vermilyea. J Electrochem Soc 118:529, 1971.
117. N Nilsen, E Bardal. Corros Sci 17:635, 1977.
118. DM Aylor, PJ Moran. J Electrochem Soc 133:868, 1986.
119. ZA Foroulis, MJ Thubrikar. Electrochim Acta 21:225, 1976.
120. ZA Foroulis, MJ Thubrikar. Werkstaffe Korros 26:350, 1975.
121. T Shibata, M Sudo. Denki Kagaki 58:227, 1990.
122. TH Nguyen, RT Foley. J Electrochem Soc 126:1855, 1979.
123. H Bohni, HH Uhlig. J Electrochem Soc. 116:906, 1969.
124. F Sato, RC Newman. Corrosion 54:955, 1998.
125. BN Stirrup, NA Hampson, IS Midgley. J Appl Chem 5:229, 1975.
126. H Kaesche. Z Phys Chem 34:87, 1962.
127. A Bond, G Rolling, H Domian. J Electrochem Soc. 113:773, 1966.
128. N Lampas, PG Koutoukos. Corros Sci 36:1011, 1994.
129. RA Bonewitz. Corrosion 29:215, 1973.
130. RA Bonewitz. Corrosion 30:53, 1974.
131. JT Reding, JJ Newport. Mater Protect 5:15, 1966.
132. CDS Tuck, JA Hunter, GM Scamans. J Electrochem Soc 134:2070, 1987.
133. Q Song, RC Newman, RA Cottis, K Sieradzki. Corros Sci 31:621, 1990.
134. JT Staley. History of wrought aluminum alloy development. In: AK Vasudevan,
RD Doherty, eds. Aluminum AlloysContemporary Research and Applications.
New York: Academic Press, 1989, p. 3.

Aluminum Alloys

249

135. E Dix. Trans AIME 137:11, 1940.


136. RP Mears, RH Brown. Ind Eng Chem 33:1003, 1941.
137. RP Mears, RH Brown, E Dix. A generalized theory of the stress corrosion cracking
of alloys. In: Symposium on Stress Corrosion Cracking of Metals, New York:
ASTM/AIME, 1945, p. 323.
138. JR Galvele. J Electrochem Soc 123:464, 1976.
139. SB de Wexler, JR Galvele. J Electrochem Soc 121:1271, 1974.
140. IL Muller, JR Galvele. Corros Sci 17:995, 1977.
141. SM de Micheli, Rev Coatings Corros 2:73, 1977.
142. IL Muller, JR Galvele. Corros Sci 17:179, 1977.
143. K Urushino, K Sugimoto. Corros Sci 19:225, 1979.
144. JR Galvele, SM de Micheli. Corros Sci 10:795, 1970.
145. F Sato, RC Newman. Corrosion 55:3, 1999.
146. WE Quist, GH Narayanan. Aluminumlithium alloys. In: AK Vasudevan, RD Doherty, eds. Aluminum AlloysContemporary Research and Applications. New
York: Academic Press, 1989, p. 219.
147. C Kumai, J Kusinski, G Thomas, TM Devine. Corrosion 45:294, 1989.
148. S Ohsaki, T Sato, T Takahashi. Aluminium 66:565, 1990.
149. P Niskanen, TH Sanders, M Marek, M Rinker. AluminumLithium 2. New York:
Metallurgical Society/AIME, 1983, p. 347.
150. P Niskanen, TH Sanders, M Rinker, M Marek. Corros Sci 22:283, 1982.
151. AH Al-Satter, V Ashworth, AK Bairamov, DJ Chivers, WA Grant, RPM Procter.
Corros Sci 20:127, 1980.
152. PM Natisham, E McCafferty, GK Hubler. J Electrochem Soc 135:321, 1988.
153. PM Natisham, E McCafferty, GK Hubler. J Electrochem Soc 133:1061, 1986.
154. H Yoshioka, S Yoshida, A Kawashima, K Asami and K Hashimoto. Corros Sci
26:795, 1986.
155. WC Moshier, GD Davis, JS Ahearn, HF Hough. J Electrochem Soc 134:2677,
1987.
156. WC Moshier, GD Davis, GO Cote. J Electrochem Soc 136:356, 1989.
157. GD Davis, WC Moshier, TL Fitz, BJ Rees. J Electrochem Soc 137:422, 1990.
158. GD Davis, TL Fitz, BJ Rees, BA Shaw, WC Moshier. Annual Report MM1-TR9028c, Martin Marietta Laboratories, 1991.
159. BA Shaw, TL Fitz, GD Davis, WC Moshier. J Electrochem Soc 137:1317, 1990.
160. RB Inturi, Z Szklarska-Smialowska. Corrosion 34:1201, 1993.
161. GS Frankel, RC Newman, CV Jahnes, MA Russak. J Electrochem Soc 140:2192,
1993.
162. AK Vijh. Corros Sci 36:1615, 1994.
163. AK Vijh. Mater Chem Phys 20:371, 1988.
164. Z Szklarska-Smialowska. Corros Sci 33:1193, 1992.
165. RB Inturi, Z Szklarska-Smialowska. Corrosion 34:705, 1993.
166. JC Uruchurtu, JL Dawson. 43:19, 1987.
167. GS Frankel. Corros Sci 30:1203, 1990.
168. RG Buchheit, JP Moran, GE Stoner. Corrosion 46:610, 1990.
169. R Ambat, ES Dwarakadasa. Corros Sci 33:681, 1992.
170. S Maitra, GC English. Metall Trans 12A:535, 1981.
171. T Hagyard, JR Santhiapillai. J Appl Chem 9:323, 1959.

250

Holroyd

172.
173.
174.
175.

IL Rosenfeld, IK Marshakov. Corrosion 20:115t, 1964.


DW Siitari, RC Alkire. J Electrochem Soc 129:481, 1982.
K Herbert, RC Alkire. J Electrochem Soc 130:1001, 1983.
M Baumgartner, H Kaesche. The nature of crevice corrosion of aluminum in chloride solutions. In: HS Isaacs, U Bertocci, J Kruger, S Smialowski, eds. Advances
in Localized Corrosion. Houston, TX: National Association of Corrosion Engineers, 1990; p. 407; also Werkstaffe Korros 39:129, 1988.
H Takahashi, M Mukai, M Nagayama. Breakdown of oxide films on aluminum
in neutral solutions containing organic and inorganic anions. Proceedings of 9th
International Congress on Metallic Corrosion, Toronto, 1984, p. 155.
AH Le, RT Foley. Corrosion 40:195, 1984.
NJH Holroyd, MR Jarrett, CJ Newton. Submitted to Corros Sci.
KP Wong, RC Alkire. J Electrochem Soc 137:3010, 1990.
T Beck. Electrochem Acta 29:485, 1984.
MJ Pryor. The influence of defect structure of aluminum oxide films on the pitting
of aluminum in chloride solutions. In: RW Staehle, BF Brown, J Kruger, A Agrawal, eds. UR Evans Conference on Localized Corrosion, Houston, TX: National
Association of Corrosion Engineers, 1974, p. 3.
K Kitamura, E Sato. J Jpn Inst Light Metals 29:563, 1979.
BF Brown, CT Fujii, EP Dahlerg. J Electrochem Soc 116:218, 1969.
ZX Haung, YD He. Electrochemical behaviour of within stress corrosion cracks
of aluminum alloy LC-4. Proceedings of 9th International Congress on Metallic
Corrosion, Toronto, 1984, p. 495.
AJ Sedriks, JAS Green, DL Novak. Corrosion 27:198, 1971.
JA Davis. Use of microelectrodes for study of stress corrosion of aluminum alloys.
In: RW Staehle, BF Brown, J Kruger, A Agrawal, eds. UR Evans Conference on
Localized Corrosion. Houston, TX: National Association of Corrosion Engineers,
1974, p. 168.
TH Nguyen, BF Brown, RT Foley. Corrosion 38:319, 1982.
NJH Holroyd, GM Scamans, R Hermann. Environmental interaction with the crack
tip region during environment sensitive fracture of aluminium alloys. In: RP Gangloff, ed. Embrittlement by the Localized Crack Environment. New York: TMS
AIME, 1984, p. 327.
C Edelenenanu, UR Evans. Trans Faraday Soc 47:1121, 1951.
AJ Sedriks, JAS Green, DL Novak. Corrosion processes and solution chemistry
within stress corrosion cracks in aluminum alloys. In: RW Staehle, BF Brown, J
Kruger, A Agrawal, eds. UR Evans Conference on Localized Corrosion. Houston,
TX: National Association of Corrosion Engineers, 1974, p. 569.
CF Bates, RE Mesmer. The Hydrolysis of Cations. New York: J Wiley, 1976.
JR Galvele. Corros Sci 21:551, 1981.
SM de Micheli. Corros Sci 18:605, 1978.
A Alavi, RA Cottis. Corros Sci 27:443, 1987.
A Becerra, R Darby. Corrosion 30:153, 1974.
DE Davies. J Appl Chem 9:651, 1959.
AC Fraker, AW Ruff. Corrosion 27:151, 1971.
S Maitra, GC English. Metall Trans 13A:161, 1982.

176.

177.
178.
179.
180.
181.

182.
183.
184.

185.
186.

187.
188.

189.
190.

191.
192.
193.
194.
195.
196.
197.
198.

Aluminum Alloys
199.
200.
201.
202.
203.
204.
205.
206.
207.
208.
209.
210.
211.
212.
213.
214.
215.

216.
217.
218.

219.
220.
221.

251

BF Lifka. Aluminum 63:12, 1987.


JG Craig, RC Newman, MR Jarrett, NJH Holroyd. J Phys C3:481, 825, 1987.
S Marsac, G Mankowski, F Dabosi. Br Corros J 27:50, 1992.
RT Foley, TH Nguyen. J Electrochem Soc 129:464, 1982.
AM Beccaria, G Poggi. Br Corros J 20:163, 1985.
AM Beccaria, G Poggi. Corros Preven Control 34:51, 1987.
AM Beccaria, G Poggi. Corrosion 43:153, 1987.
FM Reihart. Corrosion of metals and alloys in deep ocean. Report 834, US Naval
Engineering Laboratory Port Hueneme, Alexandria, VA, 1976.
IB Ulanowskii, VA Egorova. Zashch Met 14:176, 1978.
IB Ulanowskii. Zashch Met 16:156, 1980.
J Perkinks, JR Cummings, KJ Graham. J Electrochem Soc 129:137, 1982.
AM Beccaria, G Poggi, D Gingaudo, P Castello. Br Corros J 29:65, 1994.
RA Horne. Marine Chemistry. New York: J Wiley, 1969, p. 73.
GA Gehring Jr. Corrosion of aluminum alloys in high velocity sea water. Fifth
International Congress on Marine Corrosion and Fouling, Madrid, 1980.
GA Gehring Jr, MH Peterson. Corrosion 37:232, 1981.
J Larson-Basse. J Metals 37:24, 1985.
S Jain, J Larson-Basse. Effect of flow velocity on the corrosion of some aluminum
alloys in deep ocean seawater. Proceeding of the 42nd NACE Conference, 1986,
Paper 224.
J Larson-Basse. Mater Perform 23:16, 1984.
DC Silverman. Corrosion 41:679, 1985.
HP Godard, FF Booth. Corrosion behaviour of aluminum alloys in seawater. In:
Congres International de al Corrosion Marine et des Salissures. Editions du Centre
du Rechercher et lEtudes Oceaniques, Paris, 1964, p. 37.
J Larson-Basse, Y Park. Mater Perform 28:51, 1989.
WH Ailor. Flowing sea water corrosion potentials of aluminum alloys. Proceedings
of 26th NACE Conference, Houston, TX, 1970, p. 200.
CA Witt, Aluminium 56:398, 1980.

8
Magnesium Alloys
Mike J. Danielson
Pacific Northwest National Laboratory, Richland, Washington

I.

INTRODUCTION

Aluminum and zinc are the two major alloying elements added to magnesium
to increase its strength, ductility, and resistance to general corrosion attack. Magnesium alloys are of engineering interest because of their low density and high
strength-to-weight ratio. Magnesium alloys are about two-thirds the density of
aluminum alloys and about 25% lighter than aluminum alloys with similar stiffness. Other advantages are good machinability and castability into complex configurations. Magnesium alloys are used in applications where strength and
lightness are at a premium, such as aircraft, space applications, automotive applications, and certain consumer uses (i.e., luggage, computer frames, audio equipment, lawnmowers). In 1990, most alloy use was in die castings involving one
composition, AZ91D (1). Under dry or low-humidity applications, these alloys
have excellent materials properties, but when immersed or at high relative humidity, they can suffer severe localized corrosion attack, stress-corrosion cracking,
and corrosion fatigue.
There have been a number of reviews (26) on the materials performance
of magnesium alloys. This review will focus on the effect of the environment
on the materials performance, including general and localized corrosion attack,
stress-corrosion cracking, and corrosion fatigue under a variety of relevant engineering environments. This review will (a) summarize the early work and (b)
cover the more recent work since 1993, the most recent review. The intent will
be (a) to describe the underlaying principles that control the behavior of these
alloys, (b) describe the failure modes and their character, (c) describe methods
253

254

Danielson

that mitigate the failure processes, (d) direct the engineer in searching and evaluating the literature, and (e) assist the engineer in carrying out new testing in
relevant environments. The literature is quite complete on the subjects of generalized and localized attack but very sparse on stress-corrosion cracking and corrosion fatigue. On these latter two topics, benchmark papers are pointed out that
will be used to reveal the general character of magnesium alloys. After reading
this chapter, the engineer will have a good grounding in using these alloys in
practical applications and avoiding corrosion-related failures.

II. NOMENCLATURE
Successful use of these alloys requires an understanding of how the alloys are
identified. The American Society for Testing and Materials (ASTM) created a
standard system for naming magnesium alloys and tempers which has been in
use since 1952 (7). This system is clearly described by Hough et al. (8), and an
abbreviated version of their original table is shown in Table 1. The system is
most easily learned by following an example while referring to Table 1, using
AZ91E-T6 as the magnesium alloy. The letters AZ refer to the two alloying
elments in highest concentration, aluminum and zinc (highest concentration
shown first). The next two numbers, 91, refer to the approximate amount of each
alloying element, namely 9 wt% Al and 1 wt% Zn, rounded up to the largest
whole number. The third part of the description, E, is an additional designator
of the alloy and may have importance in specifying additional compositional
details. The fourth part, T6, refers to the temper, which in this case is a solution
anneal followed by some artificial aging. Table 2 shows the composition and
yield strength of some commonly used cast magnesium alloys.

III. CORROSION BEHAVIOR


A. Basic Metallurgy of Magnesium Alloys Affecting the
Materials Performance
Aluminum and zinc are the two elements most often used to alloy with magnesium to improve its strength and corrosion properties, and their phase diagrams
are shown in Figs. 1 and 2, respectively (9). Generally, the aluminum content
does not exceed about 10 wt% or the zinc content about 6 wt% in commercial
alloys. This is because the solubility of aluminum in magnesium is 12.8 wt% at
437C, and exceeding this concentration will result in the formation of the phase, Mg17Al12. However, the solid solubility of Al significantly decreases with
temperature (1.5 wt% at ambient), and some Mg17Al12 will usually be present
depending on the quenching rate and tempering. The solubility of zinc in magne-

Magnesium Alloys

255

Table 1 ASTM Identification System for Magnesium Alloys and Temper


First part
Indicates the two
principal alloying elements using two coded
letters:
A
B
C
D
E
F
G
H
K
L
M
N
P
Q
R
S
T
W
Y
Z

aluminum
bismuth
copper
cadmium
rare earth
iron
magnesium
thorium
zirconium
lithium
manganese
nickel
lead
silver
chromium
silicon
tin
yttrium
antimony
zinc

Second part

Third part

Two numbers indicating the


weight percent
(wt%) the
amounts of the
two principal elements

A letter designation that distinguishes between


two alloys with
the same composition of the two
principal elements

Fourth part
Temper designation:
F as fabricated
O annealed
H10, H11
slightly strain
hardened
H23, H24, H26
strain hardened and partially annealed
T4 solution
heat treated
T5 artificially
aged, only
T6 solution
heat treated
and artificially
aged
T8 solution
heat treated,
cold worked,
and artificially
aged

Source: Refs. 7 and 8.

sium is 6.2 wt% at 340C, and exceeding this concentration will result in the
formation of a variety of complex intermetallics, such as MgZn2. Because the
solubility of zinc also decreases with temperature (1.7 wt% at 150C), zincrich intermetallics may also be present. Mn and, more recently, Zr are added to
the alloy to act as grain refiners to improve the strength. Due to their fairly low
solubility, they will also be present as intermetallics in the matrix. Fe, Ni, and
Cu are very insoluble impurities, which above a certain threshold concentration
are well known to have extremely detrimental effects on the localized corrosion
behavior. They are present as intermetallics which act as cathodic depolarizers
in the creation of a localized microgalvanic cell. Magnesium alloys make use of
a number of other minor metal additions of low solubility such as the rare earths,

256

Table 2

Nominal Compositions and Yield Strengths of Selected Magnesium Cast Alloys

Alloy

Mn

Zn

10.1
6.0
7.6
8.7
8.7
9.0

0.1
0.15
0.13
0.13
0.13
0.10

3.0
0.7
0.7
0.7
2.0
2.7
3.3
3.3

1.0

0.250.75

6.0
4.2
5.8
5.7
4.6
6.0

Other

Th

0.7
0.6
0.7
0.7
0.7
0.7
0.7
0.7
0.7
2.7
0.7
0.7

Zr,
Zr,
Zr
Zr
Zr
Zr,
Zr,
Zr,
Zr,
Cu
Zr,
Zr,

1.8
0.7 Zr
0.7 Zr

Note: In most cases, Cu (0.08%), Ni (0.01%), and Fe (0.005%) are kept at very low values.
Source: Refs. 7 and 8.

1.5 Ag, 2.1 Di


3.3 RE

2.5
2.5
4.0
5.2

Ag, 2.1 Di
Ag, 1.0 Di
Y, 3.4 RE
Y, 3.00 RE

1.2 RE
2.6 RE

Yield strength (MPa)


150
130
83
145
145
150
195
110
105
90
55
195
205
165
172
125
140
190
170
165
195

Danielson

AM100A-T61
AZ63A-T6
AZ81A-T4
AZ91C
AZ91E-T6
AZ92A-T6
EQ21A-T6
EZ33A-T5
HK31A-T6
HZ32A-T5
K1A
QE22A-T6
QH21A-T6
WE43A-T6
WE54A-T6
ZC63A-T6
ZE41A-T5
ZE63A-T6
ZH62A-T5
ZK51A-T5
ZK61A-T6

Al

Magnesium Alloys

257

Fig. 1 Phase diagram of the AlMg system.

which improve certain mechanical properties but also may form precipitates (contribute to the strengthening by precipitation hardening) within the matrix.
An article by Beldjoudi et al. (10) examined the microstructure of Mg9
Al, AZ91, and Mg3 Al alloys. In the T4 (solution annealed) condition, the alloys
were all single phase (dendritic substructure) with the Al uniformly distributed
in the Mg matrix, there being no evidence of segregation. Precipitates of the type
AlMnFe (globular) and Mg2Si (polygonal) were distributed uniformly throughout
the matrix. Aging the Mg9 Al alloy to the T6 condition resulted in the decrease
of solid-solution Al from 9% to 3% and the formation of a lamellar -Mg17Al12
at the grain boundaries. The T6 condition did not affect the size and distribution
of the AlMnFe-containing precipitates, but the Mg2Si were now present primarily
along grain boundaries. Zn did not affect the microstructure. The observations
of Lunder (11,12) on AZ91 in the F, T4, and T6 tempers reinforce those of
Beldjoudi, with the additional observation that the Zn is always in solid solution.
The as-cast condition (F) contained a great deal of -Mg17Al12 at the grain boundaries as well as being distributed throughout the matrix, and the solid-solution
Al was distributed nonuniformly. The intermetallics, AlMnFe and Mg2Si, were
present throughout the matrix and did not appear to be altered by heat treatment.

258

Fig. 2

Danielson

Phase diagram of the MgZn system.

In summary, the presence of a variety of intermetallics in the matrix, the


grain boundaries, and at other inhomogeneities has a profound affect on the corrosion performance of this entire class of alloys. Heat treatment can change the
distribution and microchemistry of the intermetallics and it should be expected
that the corrosion degradation processes will be a strongly dependent on the heat
treatment. The detailed interaction between the intermetallics and the degradation
mechanisms has not been adequately studied to date.
B. Basic Electrochemistry of Magnesium Alloys Affecting
the Materials Performance
Magnesium is one of the most anodic (reactive) metals in aqueous solution, being
next to the very reactive metals Li and Na in the electromotive series. A Pourbaix
diagram (13) is shown in Fig. 3 that demonstrates that the electrochemical potential for the reaction
Mg 2 2e Mg 0

Magnesium Alloys

259

Fig. 3 Pourbaix diagram for the MgH 2 O system.

is more than 2 V below the hydrogen evolution a line, indicating that there
is a large thermodynamic driving force for the spontaneous reaction (corrosion)
of the Mg with water with the simultaneous evolution of hydrogen gas. Pourbaix
diagrams only show the thermodynamic propensity for reaction; the actual kinetic
rate must be determined by a direct physical measurement. Figure 4 shows the
measured open-circuit corrosion potential of various Almagnesium alloys in a
common testing environment, 3.5 wt% NaCl. The significant feature of this plot
is that the potential (written as a reduction potential) becomes more negative as
the magnesium content increases, indicating that the driving force for the corrosion reaction increases with Mg content. Because many of the alloying agents

260

Danielson

Fig. 4 Effect of the magnesium composition in AlMg alloys on the open-circuit corrosion potential exposed to 3.5% NaCl.

have limited solubility in the magnesium matrix, intermetallics and impurities


will be commonly present in the matrix as precipitates. For example, -Mg17 Al 12
(56 wt% Mg) will be present near grain boundaries or other precipitates, and
because its corrosion potential is less negative than the surrounding Mg matrix
(see Fig. 4), it will be cathodically protected by the adjacent matrix that is richer
in magnesium. One of the unusual features of magnesium alloys is that these
precipitates are cathodic relative to the matrix; consequently, they act as sites for
hydrogen evolution. This results in the creation of microgalvanic cells that are
in intimate contact with each other. It appears that the corrosion rate of magnesium alloys is controlled by the rate of the hydrogen evolution reaction, and often
these intermetallic cathodic sites have greater hydrogen evolution kinetic rates
than the magnesium matrix itself. This explanation was used early in the development of Mg alloys to explain why the Fe, Cu, and Ni intermetallics, which support
unusually large hydrogen evolution rates, are worse than any others for driving
the galvanic process (14). Recently, Lunder et al. (15) synthesized a number of
the normally present intermetallics [e.g., Al 6 Mn, Al 6 Mn(Fe), Al 3 Fe, -Mg 17Al 12,
Mg 2Si, and Al 4MM, where MM Misch metal] and (a) measured their opencircuit corrosion potential in pH 10.5, 5% NaCl solution saturated with Mg(OH) 2
and (b) determined the hydrogen evolution rate when polarized to 1.6 V (saturated calomel reference electrode scale), the open-circuit potential of magnesium.
The hydrogen evolution rate was greatest for the Fe-containing intermetallics and
least for Al 6Mn and Mg 2 Si. The -Mg17Al12 gave intermediate values between
these two extremes. Clearly, the composition of the intermetallics can have a
very important role in controlling the localized corrosion performance, and the
alloy purity is the starting point for controlling the intermetallic composition and
improving the corrosion behavior of Mg alloys. Because the precipitates are less

Magnesium Alloys

261

reactive than the matrix, an unusual type of localized attack occurs: the precipitates are protected and attack starts adjacent to them in the matrix. Careful microstructural evaluation is needed to reveal this type of morphology.
Magnesium corrodes in neutral and basic water to form a magnesium hydroxide film that can be very protective under certain specific conditions:
Mg 0 2H 2O Mg(OH) 2 H 2 (g)
The pH and certain anions, such as chloride, can damage the protective film,
leading to large increases in the corrosion rate. The composition of the film is a
function of the alloy composition and the impurities in the environment (16,17).
In summary, magnesium alloys are vulnerable to corrosive attack because
of the extreme reactivity of the magnesium within the alloy and the intimate
mixture of intermetallics that can act as localized cathodes, creating microgalvanic cells. The composition of the intermetallics can strongly affect the corrosion
performance of the magnesium alloy.
C. Localized and General Corrosive Attack
(Aqueous Solutions)
Magnesium alloys are prone to attack in aqueous solutions, particularly saltwater.
The presence of certain metallic impurity phases that makes the alloys vulnerable
has been recognized for over 50 years. Hannawalt et al. (14) carried out a pioneering study in 1942 that defined the tolerance limits for a number of the common
impurities. Their effects on the corrosion rate in saltwater are shown in Fig. 5.
Basically, Na, Si, Pb, Sn, Mn, and Al have negligible effects up to 5 wt% and
Fe, Ni, Co, and Cu have extremely detrimental effects at very low concentrations.
These tolerance limits are approximately 0.017, 0.1, and 0.004 wt% for Fe, Cu,
and Ni, respectively, and these elements must be maintained at very low concentrations in order for magnesium alloys to have adequate corrosion resistance.
Tolerance limits (3) are shown for some modern cast alloys in Table 3. Manganese is commonly added to commercial magnesium alloys at the nominal 0.2%
level to act as a grain refiner and to increase the tolerance for Fe. Recent work
indicates that the corrosion resistance is better correlated to the Fe/Mn ratio than
to the Fe content (18). Manganese is believed to act by (a) forming AlMnFe
intermetallic particles that precipitate out and fall to the bottom of the crucible,
thereby reducing the iron content, and (b) the fact that the AlMnFe intermetallics
remaining within the matrix are less efficient as cathodes as compared to metallic
Fe, reducing the driving force for the galvanic cell. Table 4 (19) is included in
order to benchmark the corrosion rate of a common magnesium alloy, AZ91, as
a function of purity (Fe content) and heat treatment under ASTM B-117 salt
spray conditions. Clearly, the corrosion rate of the commercial purity alloy (C
suffix) is unacceptable in saltwater conditions in any heat-treatment condition,

262

Fig. 5

Danielson

Corrosion rate of magnesium alloys as a function of various constituents.

being nominally 15 mm/year. The high-purity alloy (E suffix) has a dramatically


lower corrosion rate, and this reinforces the principle that improved corrosion
resistance must come from improved purity, there having never been found any
alloying elements that greatly inhibit the corrosion of magnesium. The results in
Table 4 clearly show that heat treatment affects the localized corrosion behavior
of the alloys. For instance, aging (T6 condition) results in the precipitation of

Table 3 Tolerance Limits for Cast Alloys (ppm)


Element
Fe (with 0.2% Mn)
Ni
Cu

AZ91

AM60

AS41

AE42

64
50
400

42
30
100

20
40
200

40
200
1000

Magnesium Alloys

Table 4

Alloy
AZ91C
AZ91C
AZ91E
AZ91E
AZ91E
AZ91E

Corrosion Behavior of AZ91 in Saltwater


Grain
size (m)
187
66
146
78
160
73

Temper and corrosion rate (mm/year)

Mn%

Ratio
Fe/Mn

T4

T6

T5

0.18
0.16
0.23
0.26
0.33
0.35

0.087
0.099
0.008
0.008
0.004
0.004

18
17
0.64
2.2
0.35
0.72

15
18
4
1.7
3
0.82

15
15
0.15
0.12
0.22
0.1

0.12
0.12
0.12
0.1

263

264

Danielson

the -Mg 17Al 12 phase at the grain boundaries which are cathodic to the matrix,
and yet this heat treatment results in the most beneficial condition for decreasing
localized attack. The solution anneal case (T4) shows an anomolously high corrosion rate relative to the as-cast (F) and aged conditions (T6) (12,19) that is not
satisfactorily explained (the solution annealed alloy should be more homogeneous
that the cast or aged material). Beldjoudi et al. (10) reproduced these effects
of heat treatment and clearly showed by electrochemical measurements that Mg 17Al 12 is less reactive than the matrix. Lunder et al. (11) examined the attack
morphology on AZ91 in saltwater and found the pitting attack initiated adjacent
to the intermetallic particles consisting of type AlMnFe, -Mg 17Al 12 , or Mg 2Si.
Attack followed the dendritic arms in the as-cast material, whereas filiform corrosion was observed for the T4 condition. Other alloying agents (Zr, Y, and rare
earths Nd, Gd, Tb, Dy) are used to increase the strength of magnesium alloys,
and they appear to have a small effect on the corrosion resistance as reported by
King et al. (20) and Kamado et al. (21) in salt spray tests. King et al. (20) report
corrosion rates for WE54 and WE45 of about 0.7 mg/cm 2 /day (1.4 mm/year),
which are still substantial. Lunder et al. (12) determined that rare earths and
silicon additions have a beneficial effect on the corrosion resistance, whereas
zinc has a negligible effect.
Magnesium alloys are attacked by acidic solutions but are quite resistant
to alkaline environments and many hydrocarbon environments. Two exceptions
to the acidic generalization are HF and chromic acid solutions, in which the alloy
can be quite resistant (2,3). The alloys are very resistant to deionized water but
will pit in potable water, but the threshold halide concentration for pitting is not
defined in the literature. Except for fluoride, any of the halides, sulfates, and
nitrates are damaging. Solutions containing dissolved Cu, Ni, or Fe ions are especially damaging because the ions will be reduced onto the alloy surface and act
analogously to what happens when they are present within the alloy as impurities.
Froats et al. (22) have a useful chart of organic and inorganic environments with
the corrosion response of the magnesium alloys. Many of the results are prefaced
with the excellent advice that the material should also be tested in the actual
engineering environment to ensure a satisfactory result.
Hallopeau et al. (23) examined the behavior of certain traditional inhibitor
anions (SiO 32 , PO 43 , CrO 42 , MoO 42 ) on a Mg9 Al alloy in a 0.5M sodium
sulfate solution. The corrosion behavior was determined by using potentiodynamic and polarization resistance methods, which only evaluate the short-term
performance. As expected, the Mg was selectively dissolved from the alloy, leading to a surface film enriched in Al. Some inhibition was observed in alkaline
solutions, with SiO 32 and CrO 42 having an inhibiting effect in neutral solutions.
No effect was observed for MoO 42 and PO 43.
The effect of increasing the temperature is to increase the severity of attack.
As an example, the corrosion rate, which is negligible in deionized water at ambi-

Magnesium Alloys

265

ent temperature, increases to about 0.5 mm/year for the AZ alloys at 100C (16).
Clearly, magnesium alloys do not have adequate corrosion resistance for applications above ambient temperature. The literature does not give any guidance as
to the existence of a threshold temperature, above which the material will corrode
excessively.
In summary, magnesium alloys have their poorest corrosion behavior in
immersed, aqueous environments. Generally, the corrosion rates are reported in
terms of weight loss rather than penetration depth, and this can be misleading
because this material is prone to pitting rather than uniform attack. As a caveat,
the corrosion behavior of magnesium alloys under any immersion condition
should be regarded with suspicion until proven safe.
D. Localized and General Corrosive Attack (Gas Phase
and Partial Immersion)
Under most conditions, a partial immersion (part gas phase and part liquid phase)
is at least as damaging as a full immersion, so that all the observations applicable
to full immersion are relevant to a partial immersion. There is the additional
complication that the current lines are focused at the liquid side of the gasliquid
phase, resulting in a knifeline attack that may increase the penetration rate. The
presence or absence of oxygen does not appear to affect the process.
Magnesium alloys have insignificant corrosion rates in clean air if the relative humidity is below about 65% (2); if the surface remains clean of salt deposits,
the alloys can remain corrosion-free up to relative humidities of 90%. When
the surface becomes contaminated with dirt (which contains salts), the corrosion
rates increase and the damage is revealed as shallow pits.
Benchmark results for a dirty industrial environment are shown in Table
5. These results can be quite variable, with a rural environment having lowest
corrosion rates and a marine environment might be somewhat greater. These
results indicate that air exposures are relatively benign for magnesium alloys
compared to total immersion.
Dry gases such as chlorine, iodine, bromine, and fluorine result in little
corrosion of magnesium alloys under ambient temperature conditions. However,
the addition of moisture can greatly increase the corrosion rate.
Magnesium alloys have linear reaction kinetics in oxygen gas at elevated
temperatures and these rates can be very high. For instance, as a benchmark rate,
Leontis and Rhines (24) have measured a penetration rate of about 1.4 mm/year
for a 9% Almagnesium alloy at 400C. The rate would be lower in air and at
lower temperatures, but it is not clear what would constitute safe operating conditions; this would have to be determined on a case-by-case situation. Magnesium alloys can be used successfully under inert or low-reactivity environments.
There is a successful application (350C) in the gas-cooled nuclear industry for

266

Danielson

Table 5 Penetration Rates for Various


Magnesium Alloys Exposed for 2 Years
in an Industrial Atmosphere
Alloy
A8
AZ91
Z5Z
ZRE1
TZ6
A8, high purity

Penetration rate
(mm/year)
0.049
0.047
0.069
0.074
0.084
0.058

Source: Ref. 2.

zirconium- or beryllium-containing magnesium alloys where they are used to


encapsulate uranium fuel. Clearly, this has to be under inert gas conditions so
that the penetration rate by the oxidation process is essentially zero (2). Magnesium alloys are reported to be very corrosion resistant in elevated-temperature
H 2 S or SO 2 environments (25). Oxidation rates are low in BF 3 and SF 6 , and these
gases are used as cover gases in foundry operations (22). Usually, magnesium
alloys are not used at these elevated temperatures because of a significant loss
of creep strength.
In summary, magnesium alloys have excellent corrosion behavior in lowhumidity environments and at elevated temperatures under oxygen-free atmospheres. Partial immersion and high-humidity environments must be examined
on a case-by-case basis.
E.

Stress-Corrosion Cracking and Corrosion Fatigue


(Aqueous Solutions)

A serious limitation to the structural use of magnesium alloys is their propensity


for SCC (stress-corrosion cracking) in a wide variety of environments, alloy compositions, and metallurgical conditions. Most of the cracks are transgranular and
initiate at a surface pit or defect. The SCC review by Miller et al. (26) remains
the best overview in this field to the year 1991, and it will be quoted extensively
in this section. The Al-containing alloys are considered the most susceptible to
SCC, and this susceptibility increases with Al content. Zinc also increases the
susceptibility, and the most commonly used class of alloys, AZ, is considered to
be particularly susceptible. Those alloys containing no aluminum or zinc are the
most resistant, with M1 (Mnmagnesium alloy) being one of the most resistant.
SCC is observed in almost any aqueous environment, including deionized water.

Magnesium Alloys

267

Halide ions are particularly aggressive toward SCC, but almost any dissolved
salt increases the rate over that observed in deionized water. The environments
that do not induce SCC are those in which the magnesium alloys are perfectly
passive, such as dilute alkalies, concentrated HF, and chromic acid. Nitrates and
carbonates also have an inhibiting effect. The threshold concentrations at which
dissolved ions can cause SCC are not well defined in the literature, but dilute
solutions are capable of accelerating SCC. A popular laboratory test environment
for ranking alloys is one containing NaCl K 2Cr 2O 4 because it results in rapid
SCC. Unfortunately, the results obtained in this environment have not correlated
well with field failure experience. The literature is filled with conflicting observations about the effects of wrought versus cast alloys, the effects of cold work,
and the effects of heat treatment on susceptibility to SCC. Curiously, the cracking
process is slowed or even stopped by cathodic polarization while it is increased
with anodic polarization (relative to the open-circuit electrochemical potential);
yet, the cracking process is hypothesized to be driven by hydrogen embrittlement.
Fracture-mechanics-based SCC crack growth rate data are uncommon in
the literature, but some benchmark fracture mechanics data for this process are
shown in Fig. 6 in a work by Speidel et al. (27) with a somewhat SCC-resistant
alloy, ZK-60A [Al-free alloy containing zirconium, yield strength 296 MPa
(43 ksi)]. The cracking rate in the two salt solutions was greater than 3 10 4
cm/s, a very high rate, with the K 1scc 7.7 MPa m 1/2 (7 ksi in.1/2 ). The effects
of the dissolved salt are to increase the cracking rate and lower the K 1scc relative
to the distilled-water environment. The authors make the important comment
that in their experience, all high-strength magnesium alloys behave similarly to
environment-enhanced subcritical crack growth and that this behavior does not
vary significantly with alloy composition and heat treatment. Both SCC and corrosion fatigue cracks propagated in a mixed transgranular and intergranular mode
in this testing.
The same environments that cause SCC also reduce the corrosion fatigue
(CF) performance. Magnesium alloys have their best CF performance in vacuum,
followed by air, deionized water, and salt solutions. The CF strengths can be as
low as 10% of those in air. As with SCC, the aluminum- and zinc-containing
alloys are especially susceptible, even in mildly corrosive environments (28). A
significant amount of SN data exist in the literature, but most of it is old. The
majority of the modern fracture-mechanics-based fatigue crack growth data
comes from Russia. The effects of wrought versus cast alloys, the effects of cold
work, and the effects of heat treatment are largely unexplored with CF. Some
fracture-mechanics-based (da/dNK) benchmark data for CF are shown in Fig.
7 (27). This work clearly demonstrates that salt solutions significantly increase
the CF rate and that distilled water is more aggressive than dry argon (particularly
in stage III crack growth). In a more recent study, Stephens et al. (29) examined
the CF response of a modern, high-purity sand-cast alloy, AZ91E-T6, in a 3.5

268

Danielson

Fig. 6 The effect of stress intensity and environment on the velocity of SCC cracks for
a high-strength magnesium alloy.

wt% NaCl solution and air (R 0.05 and 0.5, respectively). Their CF testing
was limited to the stage II regime, but the results were very similar to the results
of Spiedel et al., implying that their earlier comment about the similarity of all
magnesium alloys with regard to SCC behavior might also be extended to their
CF behavior. The threshold K for CF is 2 MPa m 1/2 which is even lower than
the K1scc for SCC. These are very low values.
Increasing the temperature generally acts to increase the propensity for SCC
and CF, but the literature data are extremely sparse on this variable (6).
In summary, the SCC and CF behavior of magnesium alloys is seriously

Magnesium Alloys

269

Fig. 7 The effect of cyclic stress intensity range and environment on the velocity of CF
cracks for a high-strength magnesium alloy.

degraded by immersion in aqueous solutions and probably imposes a greater


limitation to the structural use of these alloys than their general and localized
attack properties. Careful mechanical design (staying below the K 1scc or threshold
K) will be needed to successfully use magnesium alloys for structural components.
F.

Stress-Corrosion Cracking and Corrosion Fatigue


(Gas Phase and Partial Immersion)

Magnesium alloys have their best CF behavior in a vacuum environment. The


CF strength is lower in dry air and continues to decrease as the relative humidity

270

Danielson

increases. Bothwell (16) reports that an AZ31 alloy lost CF strength once the
relative humidity exceeded 50%, and at 93% relative humidity, it had a fatigue
strength of about 75% that of dry air.
G.

New Processing Methods

Two rapid-cooling methods have been utilized to improve the corrosion resistance of magnesium alloys: rapid solidification (RS) and laser annealing (LA).
Both methods are based on rapid cooling of the melt such that metastable, supersaturated solid solutions are obtained. The basic idea is to minimize the amount of
the various intermetallic compounds that could act as initiation sites for localized
corrosion processes. The RS methods result in the formation of fine powders or
foils that must be later extruded into structural forms. LA takes place after the
part is nearly machined to the final dimensions, whereupon a high-energy laser
beam is focused on the surface and melted to a depth of a few micrometers, and
then the part is machined to the final dimensions. The work of Chang and coworkers (30,31) demonstrated the improved general and localized corrosion resistance of the RS alloys as compared to conventional alloys in 3% NaCl. These
studies were carried out on RS-formed MgAlZn alloys with Mn, Si, and rare
earth additions. The materials were extruded into a bar before corrosion testing.
The measured corrosion rate for the MgAlZnY alloy was about 400 m/
year (10 mpy), one of the lowest rates ever measured for a magnesium alloy in
a salt solution. Makar et al. (32) studied the SCC of RS MgAl alloys in NaCl
K 2 CrO4 solution using the slow-strain-rate method. The materials showed the
classic transgranular SCC morphology, but no crack growth rate data were reported, so that a comparison could be made with conventional materials. The RS
alloy showed an improved general corrosion behavior and very rapid repassivation kinetics when the electrode was scratched.
The LA technique has been applied to the commercial alloy AZ91C by
first depositing a thin layer of Al, Cr, Zn, and so forth on the surface, followed
by laser annealing to melt the surface in order to incorporate the coating into the
surface alloy layer (33,34). Both Zn and Al deposits resulted in an improved
general corrosion rate behavior, but this was determined by potentiodynamic
scans which measures only short-term behavior. Once the thin modified layer
is perforated by corrosive attack, the material will probably revert to its bulk
behavior.
Another new surface modification technique is ion implantation (5). With
this technique, the part must be placed in a vacuum chamber and ions (to be
implanted) are accelerated and slammed into the surface to a penetration depth
of 50500 nm. The method is only mentioned here for information because the
penetration depth is extremely shallow and the results have had little effect on
the corrosion rate.

Magnesium Alloys

271

Magnetron sputter deposition is another surface-modification technique for


two or more alloying components. Miller et al. (26) have fabricated nonequilibrium YMg alloys in the 922 at% Y range in which the Y remained in solid
solution. The metal films were very thin (2 m thick). Potentiodynamic scans
with a potentiostat in 0.1M NaCl indicated improved corrosion resistance over
that of conventional magnesium alloys.
In summary, both the RS and LA alloys have improved the general corrosion behavior, but there is a lack of SCC and CF rate data that would demonstrate
any superiority over conventional materials. Coatings can increase the time for
failure from localized attack and potentially for SCC and CF processes by increasing the initiation time of the degradation process, but once the coating is breached,
the propagation phase will likely proceed at the same rate as in the conventionally
bulk material.
H. Coatings
Coatings are useful in improving the cosmetic appearance of finished products
and extending the service life under certain marginal corrosive environments.
There are a great many coating schemes in practice which fall into two broad
categories although they are complementary: chemical or electrochemical methods which produce a corrosion-resistant film and coatings which require the application of an impervious top coat. The reader is directed to Refs. 2 and 22 for a
comprehensive coverage of this topic. Briefly, the coating process involves cleaning the magnesium-alloy surface, first by degreasing followed by chemical or
electrochemical etching (anodizing). Etching often has the beneficial effect of
removing the surface-exposed intermetallic particles or heavy metal contamination (i.e., embedded by grit blasting) which can act as cathodes and drive the
galvanic attack. Next, some type of conversion coating is applied which may
contain a corrosion inhibitor (i.e., chromium), which prepares the surface for
adhesion of an organic top coat (i.e., paint). The top coat serves as a barrier
between the metal and the outside environment. The creation of the corrosionresistant film follows a similar processing scheme (although certain details such
as the composition of the anodizing solution may differ), except that there is no
top coat applied. Once the coating is breached by wear or abrasion, the bare metal
will be subject to the corrosion processes at the uncoated rate. Coatings would
be very useful in certain marginal environments such as alternate wet/dry exposure where general and localized attack are problems but would be dangerous in
environments where the bare alloy has a serious corrosion problem, unless extra
effort is directed toward frequent inspection and repair of the coating. Coatings
for alloys used in structural applications may give inadequate protection from
SCC and CF degradation processes because they only extend the initiation time
and could act to impede the visual inspection that would be necessary.

272

I.

Danielson

Macroscopic Galvanic Attack

The presence of intermetallic compounds, particularly iron compounds, results


in an extremely important form of microgalvanic attack that has been examined
extensively in this review. A macroscopic form of this galvanic attack is similarly
present when a Mg alloy is fastened (coupled) to a metal of different composition.
Although increasing the purity of the alloys has had a very beneficial effect on
the microgalvanic attack processes, it is of no avail for the macrogalvanic process.
This is a very serious, practical problem which requires attention to detail for
the immersed or occasionally immersed condition. Froats et al. (22) have written
about this subject in a comprehensive manner, and Olsen (35) has focused on
galvanic corrosion of automobile driveline components to which the reader is
directed. Basically, prevention of macrogalvanic attack requires attention to the
following: (1) minimizing the entrapment of water, (2) selecting metals in direct
contact with the Mg alloys that have high hydrogen overvoltages, (3) inserting
electrical insulators between the metallic fasteners, if possible, and (4) using protective coatings. Galvanic attack is least with pure aluminum (aluminum alloys
are somewhat worse) in electrical contact with a Mg alloy, and carbon and stainless steels are the worst. A compromise solution might be to use an aluminumalloy washer between the steel fastener and the Mgalloy component. Applying
a coating over the cathode (i.e., steel fastener) helps to reduce the surface area;
hence, the electrochemical driving force for the galvanic process. By attention
to detail, it is possible to make acceptable compromises that result in adequate
performance of Mg alloys. The effect of dissimilar materials is unknown on SCC
and CF processes.

IV. CONCLUSIONS
Magnesium alloys have very desirable structural properties from the standpoint of weight and strength, but these properties are offset by their unusual propensity for corrosive attack by general, localized, stress-corrosion cracking, and
corrosion fatigue under aqueous environments. The first line of defense is using
high-purity alloys containing low amounts of Fe, Cu, and Ni. There appears to
be no inhibitor for these degradation processes in most engineering environments. Coatings can improve the cosmetic appearance and extend the useful operation life in marginal environments. Accelerated testing in the laboratory
has a history of not predicting field experience, and the technical literature
usually reports test results in these accelerated environments. Clearly, laboratory
testing must realistically replicate the application environment, particularly with
respect to SCC and CF processes, if these alloys are to be economically and
safely utilized.

Magnesium Alloys

273

REFERENCES
1. MO Pekguleryuz, MM Avedesian. Magnesium alloyingSome metallurgical aspects. In: BL Mordike, F Hehmann, eds. Magnesium Alloys and Their Applications.
Oberursel: Informationgesellschaft-Verlag, 1992, pp. 213220.
2. KG Adamson, DS Tawil. Magnesium and magnesium alloys. In: LL Shreir, RA
Jarman, GT Burstein, eds. Corrosion, 3rd ed. Boston: Butterworth/Heinemann,
1994, pp. 4:984:115.
3. JE Hillis. Magnesium. In: R Baboian, ed. Corrosion Tests and Standards: Application and Interpretation. Philadelphia: ASTM, 1995, pp. 438446.
4. A Luo, MO Pekguleryuz. Cast magnesium alloys for elevated temperature applications. J Mater Sci 29:52595271, 1994.
5. GL Makar, J Kruger. Corrosion of magnesium. Int Mater Rev 38:138153, 1993.
6. WK Miller. Stress-corrosion cracking of magnesium alloys. In: RH Jones, ed. StressCorrosion Cracking. Materials Park, OH: ASM, 1992, pp. 251263.
7. Standard Practice for Codification of Certain Nonferrous Metals and Alloys, Cast
and Wrought. B 275-95a. In: ASTM, Volume 2.02, Aluminum and Magnesium
Alloys. West Conshohochon, PA: ASTM, 1996.
8. S Housh, B Mikucki, A Stevenson. Selection and application of magnesium and
magnesium alloys. In: Metals Handbook, Properties and Selection: Nonferrous
Alloys and Special-Purpose Materials, 10th ed. Materials Park, OH: ASM, 1990.
9. TB Massalski (Editor in Chief). Binary Alloy Phase Diagrams. Materials Park, OH:
ASM, 1986.
10. T Beldjoudi, C Fiaud, L Robbiola. Influence of homogenization and artificial aging
heat treatments on the corrosion behavior of MgAl alloys. Corrosion 49:738745,
1993.
11. O Lunder, JE Lein, SM Hesjevik, TK Aune, K Nisancioglu. Corrosion morphologies
on magnesium alloy AZ91. Werkstoffe Korros 45:331340, 1994.
12. O Lunder, M Videm, K Nisancioglu. Corrosion resistant magnesium alloys. Soc
Automotive Eng, 1995, Paper 950428.
13. M Pourbaix. Atlas of Electrochemical Equilibria in Aqueous Solutions. Houston,
TX: National Association of Corrosion Engineers, 1974.
14. JD Hannawalt, CE Nelson, JA Peloubet. Corrosion studies of magnesium and its
alloys. Trans AIME 147:273299, 1942.
15. O Lunder, K Nisancioglu, RS Hansen. Corrosion of die cast magnesiumaluminum
alloys. Soc Automotive Eng, 1993, Paper 930755.
16. MR Bothwell. In: H Godard, WB Jepson, MR Bothwell, RL Kane, eds, The Corrosion of Light Metals. 1967, pp. 259311.
17. GR Bengough, L Whitby. Trans Inst Chem Eng 11:176189, 19318.
18. RS Hanson. Review of corrosion behavior of Mg-alloys. In: BL Mordike, F Hehmann, ed. Magnesium Alloys and Their Applications. Oberursel: Informationgesellschaft-Verlag, 1992, pp. 327334.
19. KN Reichek, KJ Clark, JE Hillis. Controlling the Salt Water Corrosion Performance
of Magnesium AZ91 Alloy. Soc Automotive Eng, 1985, Paper 850417.
20. J King, S Thistlethwaite. New corrosion resistant wrought magnesium alloys. In:

274

21.

22.

23.

24.
25.
26.
27.

28.
29.

30.
31.
32.
33.
34.

35.

Danielson
BL Mordike, F Hehmann, ed. Magnesium Alloys and Their Applications. Oberursel:
Informationgesellschaft-Verlag, 1992, pp. 327334.
S Kamado, Y Kojima, Y Negishi, S Iwasawa. Corrosion resistance of Mgheavy
rare earth metal alloys. In: Light Metals Processing and Applications, Proceedings
of the International Symposium on Light Metals Processings and Applications, 1993,
pp. 849858.
A Froats, TK Aune, D Hawke, W Unsworthm, J Hillis. Corrosion of magnesium
and magnesium alloys. In: Metals Handbook, Volume 13, Corrosion, 9th ed. Metals
Park, OH: ASM 1987, pp. 740754.
X Hallopeau, T Beldjoudi, L Robbiola, C Fiaud. Electrochemical behavior of Mg
9 Al alloy in aqueous electrolyte solutions containing XO yn inhibiting anions. 8th
European Symposium on Corrosion Inhibitors, 1995, Vol. 2, 913927.
TE Leontis, FN Rhines. Rates of high temperature oxidation of magnesium and
magnesium alloys. Trans AIME 166:265294, 1946.
O Kubaschewski, BE Hoptins. Oxidation of Metal and Alloys. London: Butterworths, 1962, pp. 210277.
PL Miller, BA Shaw, RG Wendt, WC Moshier. Improving corrosion resistance of
magnesium by nonequilibrium alloying with yttrium. Corrosion 49:947950, 1993.
SMO Speidel, MJ Blackburn, TR Beck, JA Feeney. Corrosion fatigue and stress
corrosion crack growth in high strength aluminum alloys, magnesium alloys, and
titanium alloys exposed to aqueous solutions. In: AJ McEvily, RW Staehle, ed. Corrosion Fatigue: Chemistry, Mechanics and Microstructure. Houston, TX: National
Association of Corrosion Engineers, 1972, pp. 324345.
VV Ogarevic, RI Stephens. Fatigue of magnesium alloys. Annu Rev Mater Sci 20:
141177, 1990.
RI Stephens, CD Schrader, KB Lease. Corrosion fatigue of AZ91E-T6 cast magnesium alloy in a 3.5% NaCl aqueous environment. J Eng Mater Tech 117:293298,
1995.
CF Chang, SK Das, D Raybould, A Brown. Met Powder Rep 41:302308, 1986.
CF Chang, SK Das, D Raybould. In: Rapidly Solidified Materials. Metals Park, OH:
ASM, 1986, pp. 129135.
GL Makar, J Kruger, K Sieradzki. SCC of rapidly solidified MgAl alloys. Corros
Sci 34:13111342, 1993.
S Akavipat, EB Hale, CE Haberman, PL Hagans. Mater Sci Eng 69:311316, 1984.
P Hagnas, C Langhoff, D Moll, D Perettie, R Yates. In: Laser Assisted Deposition,
Etching, and Doping, Proceedings, Volume 459. Society of Photo-Optical Instrumentation Engineers, 1984, pp. 103107.
AL Olsen. Designing galvanic corrosion out of magnesium driveline components.
Soc Automotive Eng, 1987, Paper 870364.

9
Environmental Embrittlement of
Nickel-Based and Iron-Based
Intermetallics
Norman S. Stoloff
Rensselaer Polytechnic Institute, Troy, New York

I.

INTRODUCTION

Many investigations have shown that low ductility and brittle fracture in intermetallics are caused not only by intrinsic factors (such as lack of sufficient deformation modes, poor cleavage strength, weak grain boundaries, etc.) but also by extrinsic factors. Environmental degradation, an extrinsic factor, is found to be a
major cause of brittle fracture in many ordered intermetallics, particularly those
with high crystal symmetries (i.e., cubic L12 and B2 and hexagonal DO19), as
outlined in several recent reviews (13).
There are two types of environmental embrittlement observed in intermetallics. One is hydrogen-induced embrittlement occurring at ambient temperatures.
Many intermetallic alloys show a substantial decrease in room-temperature tensile ductility due to moisture-induced hydrogen embrittlement in air, or as a result
of direct exposure to hydrogen by precharging or testing in hydrogen gas. The
other is oxygen-induced embrittlement in air at elevated temperatures. In most
cases, the embrittlement is due to a dynamic effect involving generation and
penetration of an embrittling agent (i.e., hydrogen or oxygen) during testing.
In this chapter, the two types of environment-sensitive degradation are treated
separately. The concluding section discusses metallurgical means of alleviating
environmental degradation.

275

276

Stoloff

II. EMBRITTLEMENT AT AMBIENT TEMPERATURES


Hydrogen, introduced by exposure to moist environments, by cathodic charging,
or by testing in hydrogen gas, has been shown to embrittle a large number of
single-phase intermetallics. Table 1 lists those nickel- and iron-containing intermetallic alloys that have been tested for susceptibility to environmental embrittlement (2). Most notable is the susceptibility to moisture or hydrogen of all undoped
L12 intermetallics that have been examined to date, except for Ni3Fe, which is
embrittled only when hydrogen is charged into the alloy. Therefore, it is highly
likely that all L12 intermetallics containing substantial amounts of the transition
metals Fe, Ni, or Co will prove to be susceptible to contact with hydrogen or
hydrogen-containing environments. Except for the iron aluminides, data for other
intermetallics remain sparse.
A. L12 Alloys
In terms of environmental embrittlement, ordered L12 intermetallics can be
grouped into two categories: (a) alloys containing no reactive elements and (b)
alloys containing reactive elements (e.g., Al or Si). In the first group, the alloys
are severely embrittled only when charged with hydrogen, such as by cathodically
charging. For the second category, the alloys themselves are capable of generat-

Table 1

Alloys Embrittled by Moisture or Hydrogen at Ambient Temperature

Alloy
Ni3(Al,Ti)
(single crystal)
Ni3AlB
Ni3AlBe
Ni3(Al,Mn)
Ni3Si
Ni3(Si,Ti)
Ni3(Si,Ti)B
(Co,Fe)3V
Ni3Fe
FeAl
Fe3Al

Crystal
structure

Environmental
embrittlementa

Ref.

L12

L12
L12
L12
L12
L12
L12
L12
L12
B2
DO3

5
6
7
8
6
9
10
11
12

: Environmental embrittlement observed in moist environments; : not observed in moist environments, but embrittled by hydrogen.
Source: Ref. 2.
a

Embrittlement of Ni- and Fe-Based Intermetallics

277

ing hydrogen from hydrogen-containing environments at ambient temperatures.


The most striking case is severe embrittlement of iron aluminides in moist air at
room temperature.

1. Ni3Fe
Ni3Fe is a model intermetallic for the study of hydrogen embrittlement in alloys
displaying a critical ordering temperature, Tc. This has permitted a direct comparison to be made between the ordered and disordered conditions of Ni3Fe, as shown
in Fig. 1 (13). In each case, hydrogen was cathodically charged for 1 h prior to
tensile test at room temperature. Ni3Fe is slightly embrittled by precharging in
both the ordered and disordered conditions; simultaneous charging and testing
causes greater loss of ductility than precharging, with an especially severe effect
for the ordered condition, perhaps due to the transport of hydrogen by moving
dislocations. Fractographic studies showed a change in fracture mode from transgranular microvoid coalescence to intergranular fracture in hydrogen accompanying the reduction in elongation. However, the intergranular embrittlement zone
was only about one-third as deep in precharged samples of ordered Ni3Fe as in
disordered samples. Chia and Chung (14) report that Ni3Fe is not embrittled by
moisture in air or by water, in marked contrast to the effects of hydrogen.

2. Ni3 Al
a. Effects of Composition Although the embrittling effects of hydrogen
on fracture behavior of Ni3Al alloys have been well documented (6,15), only
recently has evidence been provided for moisture-induced cracking in these
alloys.
Liu (16) has reported that the tensile ductility of Ni-24 at% Al is increased
from 2.6% to 7.2% by testing in oxygen rather than in air. A similar result was
noted with Ni23.5 at% Al (2.5% versus 8.2% elongation in air and oxygen,
respectively). Higher ductilities are observed in vacuum than in air for boronfree Ni23.4 at% Al, but the difference is strain rate sensitive; see Fig. 2 (17).
Also, the vacuum level plays a role, with the ductility of polycrystalline Ni23.4
at% Al increasing from 7.9% to 23.4% as the vacuum level changes from 101
to 3.6 108 Pa (17).
Additional striking effects of the environment have been reported for a Ni
22.65 Al0.26 Zr alloy, recrystallized from a 110-oriented single crystal and
tested at room temperature in two orientations; see Table 2 (18). This boron-free
alloy displayed 8.7% elongation in water, 13.2% in air, and 50.6% in oxygen in
the 0 orientation. Fracture was predominantly intergranular in all environments.
Although the results could have been influenced by texture (unknown in these
alloys) or by a preponderance of fracture-resistant grain boundaries in this material, the authors suggested that the principal role of boron in ductilizing Ni3Al

278

Fig. 1

Stoloff

Embrittlement of Ni3Fe by cathodically charged hydrogen. (From Ref. 13.)

Embrittlement of Ni- and Fe-Based Intermetallics

279

Fig. 2 Effect of strain rate on room-temperature tensile ductility of polycrystalline Bfree Ni3Al (23.4% Al) in air and vacuum. (From Ref. 17.)

may be to suppress environmental embrittlement. Grain-boundary strengthening,


permitting transgranular failure, would be a secondary effect. [Another factor is
the zirconium content, as it has been shown that zirconium improves the ductility
of Ni3Al in air and, especially, in oxygen (19).]
Additional evidence that boron suppresses environmental embrittlement is
summarized in Table 3, which shows that undoped Ni23.4% Al is severely
embrittled in air, whereas a Ni24 Al0.2 B alloy displays high ductility in air,
water, and oxygen (17). These results are consistent with an earlier report by
Takasugi et al. (5), who showed that the ductility of boron-doped Ni3Al was
nearly the same in vacuum and in air. The minimum amount of boron required
to eliminate environmental embrittlement is between 50 and 100 wppm.
Wan et al. (20) have reported that the tensile ductility of Ni23 at% Al
120 wppm B is only 0.6% in 0.1 MPa hydrogen gas and rises to 10.1% in air.
However, when tested in 1.3 103 Pa vacuum or in 0.1 MPa oxygen, ductilities
of 20% or more are achieved. Although preoxidation can suppress embrittlement
by cathodically charged hydrogen (15), embrittlement by hydrogen gas was little
affected by preoxidizing for 24 h at 900C (20). Previous work by Takasugi and
Izumi (8), carried out on Ni3Al with 500 wppm B, had shown no effect of the
test environment on ductility. Wan et al. (20) concluded that higher boron contents may suppress grain-boundary decohesion in the presence of hydrogen released from water vapor. Clearly, however, boron does not prevent hydrogen
embrittlement under cathodic charging conditions (15,21).

280

Table 2 Effects of Environment and Specimen Orientation on Tensile Properties of Ni22.65% Al0.26% Zr, 25C,
5.3 103 s1
Specimen
orientation
0

45

Test
environment

Elongation to
fracture (%)

Yield strength
(MPa)

Ultimate tensile
strength (MPa)

10 5a
(MPa)

15 5a
(MPa)

Water
Air
Oxygen
Water
Air
Oxygen

8.7
13.2
50.6
6.3
10.7
47.8

322
324
326
331
341
327

528
661
1451
473
603
1438

133
127

127
143

250

271

a
Difference in flow stress at 10% and 5% (or 15% and 5%) strain.
Source: Ref. 18.

Stoloff

Effect of Test Environment on the Room-Temperature Tensile Properties of B-Free and B-Doped Ni3Al

Alloy chemistry
(at.%)
Ni23.4 Al

Ni24 Al0.2 B

Test
environment

Elongation to
fracture (%)

Yield strength
(MPa)

Ultimate strength
(MPa)

Air
Oxygen
UHV
Water
Air
Oxygen

3.1
15.8
23.4
36.8
39.3
42.8

308
336
a
288
280
289

392
681

1199
1241
1315

Embrittlement of Ni- and Fe-Based Intermetallics

Table 3

a
Not measured.
Source: Ref. 17.

281

282

Stoloff

This observation suggests that hydrogen, generated either by the moisture


aluminum reaction or by hydrogen charging, diffuses mainly along grain boundaries and causes intergranular fracture. Note that a similar reduction in ductility
and change in fracture mode has been observed in the other L12 intermetallic
alloys listed in Table 1.
The role of boron in suppressing environmental embrittlement of Ni3Al
alloys has been studied extensively, but the precise mechanism of enhancement
of environmental resistance remains elusive. Cohron et al. (22,23) have shown
that at three different boron levels, 50, 150, and 500 wppm, the ductility of Ni3Al
decreases with increasing pressure of hydrogen gas; see Fig. 3. The 500 wppm
boron alloy had previously been considered to be immune to the environment.
A shift in fracture mode from transgranular to intergranular is observed as the
ductility drops, suggesting that the level of intergranular fracture correlates well
with ductility in this intermetallic. The same study showed that a boron-free alloy
was more ductile than alloys containing 50 or 100 wppm boron at hydrogen

Fig. 3 Elongation to fracture of three B-doped Ni3Al (24% Al) alloys as a function of
hydrogen pressure. Open and closed symbols refer to the ion gauge on and off, respectively. (From Refs. 22 and 23.)

Embrittlement of Ni- and Fe-Based Intermetallics

283

pressures exceeding about 1 Pa. Because atomic hydrogen is the cause of embrittlement, this surprising result suggests that boron promotes the dissociation of
molecular to atomic hydrogen. This study also showed that when all sources of
hydrogen were minimized, the ductility of polycrystalline Ni3Al exceeded 40%,
with a predominantly transgranular fracture path.
Zirconium, as cited previously, is another important dopant that can reduce
environmental embrittlement in Ni3Al. George et al. (19,24) and Lin et al. (25)
suggested that zirconium enhances grain-boundary cohesion. Chiba et al. (26), on
the other hand, proposed that zirconium reduces the extent of sulfur segregation to
grain boundaries, thereby inhibiting sulfur-induced loss of cohesion. Itoh et al.
(27) recently attempted to resolve these conflicting views by studying the ductility
of sulfur-doped Ni3Al containing different levels of zirconium. Although they
confirmed that zirconium alleviated environmental embrittlement and found that
ZrS particles formed when sulfur levels were high, they could not definitively
identify the mechanism of the improved behavior. Zirconium has a similar beneficial effect on ductility and fatigue crack growth rates in Fe3Al alloys [see Sec.
II.C.2, (28)], but it is difficult to see how environmental effects in the two alloy
systems are related because moisture induces intergranular fracture in Ni3Al but
does not change the characteristic transgranular crack path in Fe3Al. Clearly,
much more work has to be done to identify the cause of the benefits bestowed
by small additions of zirconium.
Apart from boron and zirconium, several other solutes can increase the
ductility and change the crack path of polycrystalline Ni3Al in air; see Table 4
(29). Transition metals such as chromium, iron, and manganese improve ductility

Table 4 Effect of Alloying Additions on Room-Temperature Ductility and Fracture


Behavior of Ni3Al Alloys Prepared by Conventional Melting and Casting
Alloy
element
B
B,Fe
Mn
Fe
Pd
Pt
Co
Cu
Zr
Source: Ref. 29.

Alloy composition
(at.%)

Tensile ductile
(%)

Fracture
mode

Ni3Al (24% Al)


Ni24 Al0.5 B
Ni20 Al10 Fe0.2 B
Ni16 Al9 Mn
Ni10 Al15 Fe
Ni23 Al2 Pd
Ni23 Al2 Pt
Ni23 Al2 Co
Ni23 Al2 Cu
Ni22.65 Al0.26 Zr

13
3554
50
16
8
11
5
4
6
13

Intergranular
Transgranular
Transgranular
Transgranular
Mixed
Intergranular
Intergranular
Intergranular
Intergranular
Intergranular

284

Stoloff

in air, but the effects of these elements may be due to the introduction of the
disordered -phase in the microstructure. Recently, palladium has been reported
to increase ductility in air of single-phase Ni3Al (30). Palladium in the range 1
3 wt% also increases ductility in the presence of hydrogen, although considerable
scatter in the data was noted (31). Further, Pd-free Ni3Al containing 5 wppm
hydrogen exhibited cleavagelike facets, whereas an alloy containing 0.5% Pd
exhibited microvoids on the fracture surface. A similar beneficial effect of Pd
has been noted in steels, for which the fracture toughness and threshold stress
in constant-load tests were significantly improved by Pd additions (32).
The partial replacement of aluminum by titanium in Ni3Al single crystals
does not change the pattern of embrittlement by moisture, water vapor, and hydrogen; see Fig. 4 (33). A strain rates approaching 103, embrittlement disappeared
in all environments. This work also demonstrated that when grain boundaries are
absent, embrittlement still can occur, along low-index cleavage planes.
b. Effects of Predeformation Moisture-induced embrittlement of Ni3Al
(34) and Ni3Si,Ti (35) can be reduced by predeformation (e.g., by shot peening
or by tensile prestrain), as shown in Fig. 5. The effect varies with temperature
of the predeformation; elongation increases with decreasing predeformation temperature. Corresponding to increased tensile ductility, the fracture mode changes
from intergranular to transgranular. The effect of prior deformation was attributed

Fig. 4 Effects of environment and strain rate on the ductility of 100-oriented Ni3Al,
Ti single crystals. (From Ref. 33.)

Embrittlement of Ni- and Fe-Based Intermetallics

(a)

285

(b)

Fig. 5 Effect of prestrain on environmental embrittlement in Ni3(Si,Ti) polycrystals at


room temperature: (a) simply deformed and (b) prestrained up to 11% in liquid nitrogen
and then strained to fracture. (From Ref. 35.)

to trapping of hydrogen atoms at vacancies or dislocations or an increase of soluble hydrogen in the lattice.
c. Stress-Corrosion Cracking Ricker et al. (36) have shown that stresscorrosion cracking of Ni3Al alloyws such as IC-50 (Ni22 at% Al, 0.01 Zr, 0.08
B) and IC-218 (Ni17 at% Al, 8.18 Cr, 0.127 Zr, 0.098 B) in acidic solutions
(HNO3, H2SO4) arises from the liberation of hydrogen. Slow strain-rate tests in
low-pH solutions show a large reduction in ductility and a change in fracture
mode from transgranular to intergranular compared to tests in neutral or alkaline
solutions. Preexposure to sulfuric acid also reduces ductility in a subsequent slowstrain-rate test conducted in air, again due to release of hydrogen at the specimen
surface (36,37).

3. Ni3Si
Intergranular fracture and environmental embrittlement have been studied in
alloys based on Ni3Si (7,38). Ni3Si showed no appreciable plastic deformation
when tested in moist air but displayed an elongation of 7.5% when tested in dry
oxygen (Table 5), demonstrating that Ni3Si is prone to environmental embrittlement (39). Because the elimination of the environmental effect by testing in dry
oxygen does not lead to extensive ductility (e.g., 30% or more) and complete
suppression of intergranular fracture in Ni3Si, moisture-induced hydrogen embrittlement does not appear to be the sole source of grain-boundary brittleness in
the silicide. Boron additions segregate strongly to grain boundaries in Ni3Si and
suppress environmental embrittlement, as is clearly shown in Table 5 (39). Note

286

Stoloff

Table 5 Effect of Test Environment on Room-Temperature Tensile Properties of


Ni3Si (22.5% Si) With and Without Boron Additions

Test environment

Tensile ductility
(%)

Yield strength
[MPa (ksi)]

Ultimate
tensile strength
[MPa (ksi)]

0 ppm Ba
Air
Vacuum
Oxygen

0
4.7
7.5

677 (98.3)
685 (99.4)

627 (91.0)
853 (124)
1040 (151)

774 (112)
813 (118)

1151 (167)
1268 (184)

655 (95.0)
659 (95.6)
703 (102)

819 (119)
823 (120)
837 (136)

610 (88.6)
590 (85.6)

875 (127)
854 (124)

50 ppm Bc
Air
Vacuum

8.1
8.9
100 ppm Ba

Air
Vacuum
Oxygen

5.0
5.0
5.9
150 ppm Bc

Air
Oxygen

7.0
6.6

Fractured prior to macroscopic yielding.


Specimens were annealed 3 days/950C 1 day/600C.
c
Specimens were annealed 1 day/950C 1 day/600C.
Source: Ref. 39.
b

that as little as 50 ppm boron provides essentially the same ductility in air and
in vacuum, whereas the undoped alloys display no ductility in air and 4.7% in
vacuum. Other elements that reduce the effects of moisture in air or of distilled
water on the ductility of Ni3(Si,Ti) alloys are chromium, manganese, and iron,
in that order of effectiveness (40). Accompanying improved ductility is a reduction in the extent of intergranular fracture. Reduced strain rates are detrimental
to embrittled alloys, as is common with other intermetallics. However, the mechanism of improvement in ductility with transition metal solutes is not clear, with
several hypotheses being offered. Furthermore, the addition of these transition
elements was less effective in reducing embrittlement induced by hydrogen gas.
This implies that decomposition of H2 into H at the surface is not markedly affected by these solutes.
B. NiAl
Little is known about possible environmental embrittlement of NiAl. Lahrman
et al. (41) reported that the tensile properties of NiAl single crystals are about

Embrittlement of Ni- and Fe-Based Intermetallics

287

the same in air, in vacuum, and in oxygen at room temperature. However, Bergmann and Vehoff (42) have reported that gaseous hydrogen, charged into single
crystals at 1000C, promotes stable crack growth in 110-oriented single crystals
at temperatures above 475K. Extensive scatter in the data precluded any conclusions as to the possible influence of hydrogen on the toughness of polycrystals.
C. Iron Aluminides
The two iron aluminides, FeAl and Fe3Al, form back-centered-cubic (bcc)-ordered crystal structures of B2 and DO3 type, respectively. Despite the difference
in crystal structure, these aluminides show somewhat similar patterns of embrittlement in hydrogen-containing atmospheres at room temperature. Both aluminides exhibit only a few percent ductility (15%) when tested at ambient temperature in air. This led many researchers to conclude that these alloys are inherently
brittle. However, it has been demonstrated conclusively that when water vapor
and hydrogen are eliminated from the external environment, both alloys exhibit
considerable ductility.

1. FeAl
Tensile elongations up to 19% have been achieved in FeAl alloys tested in dry
environments such as oxygen (11). An increase in ductility from 2% in water to
18% in oxygen was accompanied by a change in fracture mode from transgranular
cleavage in air to mainly grain-boundary separation in dry oxygen. These observations suggest that cleavage planes are more susceptible to embrittlement than
are grain boundaries. The maximum degree of moisture-induced embrittlement
occurs on either side of room temperature; see Fig. 6 (1). At higher temperatures,
in situ protective oxide films can form readily on specimen surfaces, whereas at
lower temperatures, the aluminummoisture reaction is slowed and the equilibrium moisture content in air also is lowered.
a. Effects of Aluminum Content The ductility of FeAl decreases with increasing aluminum content in the range 3848 at%, as shown in Fig. 7 (43).
Environmental embrittlement of FeAl, defined as the difference in ductility between tests in air and in oxygen, is considerably reduced as the aluminum content
increases. For Fe40% Al, ductility is about 4% in air, in 2 107 torr vacuum,
or after hydrogen charging (44). Gaydosh and Nathal (45) reported that the ductility of Fe40% Al is sensitive to microstructure (annealed and furnace-cooled
material displayed 9% elongation in vacuum compared to 5% for an as-extruded
sample). However, boron increased ductility of as-extruded Fe40% Al to 9%
in vacuum. Fe50% Al, on the other hand, was brittle in vacuum for both the
as-extruded and furnace-cooled conditions. For Fe43 at% Al, the ductility is
nil in air as well as in dry oxygen; all specimens fail intergranularly. This differ-

288

Stoloff

Fig. 6 Effect of test temperature on tensile elongation of FeAl (36.5% Al) in air. (From
Ref. 1.)

Fig. 7 Effect of aluminum concentration and test environment on the room temperature
tensile ductility of B-doped and B-free FeAl. (From Ref. 43.)

Embrittlement of Ni- and Fe-Based Intermetallics

289

ence in behavior with aluminum content suggests that grain boundaries in FeAl
alloys with Al 38% are intrinsically weak. Therefore, environmental embrittlement and intrinsic effects must be distinguished in order to establish strategies for
reducing brittleness. It has been demonstrated that the intrinsic grain-boundary
brittleness in FeAl and other intermetallics can be alleviated by microalloying
with boron (45,46).
Figure 7 shows that environmental embrittlement does not occur in borondoped Fe48% Al, but ductility is low in both inert and aggressive environments
(42). As aluminum content decreases, embrittlement is increasingly severe in
boron-doped alloys, as measured by the difference in ductility between oxygen
and air. It is clear from Fig. 7 that the intrinsic ductility of boron-doped FeAl
drops sharply with increasing aluminum content, paralleling observations on boron-free material cited earlier.
b. Effects of Annealing Temperature It has been widely reported that furnace cooling from elevated temperatures or the imposition of low-temperature
anneals markedly increases the ductility of FeAl alloys (4751). This effect has
been linked to the elimination of excess vacancies, which cause hardening, as a
result of these treatments. The influence of annealing temperature on environmental embrittlement of Fe36% Al has been studied by Lynch and Heldt (49,50).
Ductility in air is much lower than in dry oxygen at all annealing temperatures,
but there is a detrimental effect of raising annealing temperature in either environment. It was concluded that excess vacancies resulting from cooling affect the
intrinsic ductility of iron aluminides and not their susceptibility to moisture.
c. Effects of Grain Boundaries There have been no systematic studies of
the influence of grain size on the susceptibility of FeAl alloys to environmental
embrittlement. However, smaller grain sizes tend to favor increased ductilities
for similar heat treatments, as noted by both Klein and Baker (51) and Gaydosh
et al. (52).
Lynch et al. (53) have shown that the tensile elongation of single crystals
of Fe35% Al is extremely sensitive to the environment; see Fig. 8. Note that
the ductility in air is about 4%, independent of orientation, whereas ductilities
in oxygen are 18.5% for crystal No. 1, oriented with the tensile axis 40 from
the (100) pole and 22.7% for crystal No. 2, oriented with the tensile axis 40
from the (100) pole. These differences in ductility are similar to those observed
for polycrystals, suggesting that grain boundaries are not necessary for moistureinduced embrittlement of FeAl alloys. (The same is true for Fe3Al alloys, in which
transgranular cleavage is usually observed in air as well as in oxygen or vacuum.)
The similarities between water-vapor-induced embrittlement and hydrogen
embrittlement are now well established for both FeAl and Fe3Al alloys. For example, lower strain rates exacerbate embrittlement in air, exactly as would be expected for a hydrogen-related phenomenon (54,55). Recently, Kasul and Heldt

290

Fig. 8

Stoloff

Elongation of Fe35 Al single crystals in oxygen and air. (From Ref. 53.)

(56) have provided additional evidence to link both types of embrittlement. Subcritical crack velocities were measured in Fe35 at% Al under constant-load
conditions; these were compared with strain-rate effects on ductility in air, as
well as with results of cathodic charging experiments. It was shown that the strain
rate has essentially no effect on ductility in 103 Pa vacuum, whereas ductility
was sharply reduced with decreasing strain rate in air (56). Similar results in air
have been reported by Shan and Lin (55), but they are at variance with a report
by Nagpal and Baker (54) of a sharp drop in ductility at a particular strain rate.
At a strain rate of about 0.3 s1 the ductilities in air and vacuum were equal; this
strain rate corresponded reasonably well with the established critical strain rate
above which enhanced hydrogen transport on moving dislocations is no longer
likely (56). The role of moving dislocations in transporting hydrogen to potential
crack initiation sites has been discussed by several research groups (13,56), but
there is, as yet, no consensus that this mechanism is applicable to environmental
embrittlement phenomena.
d. Cathodic Charging Experiments Cathodic charging experiments on
Fe35 at% Al have shown a clear correlation among hydrogen content, charging
time, and ductility (56). For example, after 24 h of charging in 1N H2 /SO4, the
hydrogen content is about 1.3 ppm and the ductility in subsequent tests in vacuum
has dropped from 8% to 1.3% (56). Baking treatments of sufficient duration at
high temperature (800C) had previously been shown by Kasul and Heldt (57)

Embrittlement of Ni- and Fe-Based Intermetallics

291

to completely restore the ductility of cathodically charged Fe3Al. For Fe35% Al,
baking at 400C for more than 1 h was sufficient to completely restore ductility. A
valuable result of this study was an estimate of the diffusivity of hydrogen in
Fe35% Al at room temperature: about 4 1012 cm2 /s. Because this diffusivity
is too low to allow significant hydrogen penetration ahead of a crack moving
under constant stress in stage 2, it was suggested that hydrogen transport by
moving dislocations may occur.
e. Fracture Toughness Relatively few studies of fracture toughness in the
iron aluminides have been reported. Klein et al. (58) have shown that the fracture
toughness of FeAl increases from 23 to 30 MPa m1/2 for blunt notched specimens
as the strain rate increased. A similar trend was noted for precracked samples.
Ko et al. (59) used the standard multispecimen JIc procedure on side-grooved
specimens to estimate fracture toughness of Fe35 at% Al as a function of test
environment. It was shown that toughness increased from air to vacuum to oxygen environments; see Fig. 9. Also shown are toughness data for an Fe3Al alloy,
Fe28 at% Al, tested in both the B2 and DO3 conditions. For both conditions,
toughness in the vacuum and oxygen environments was lower than for the FeAl

Fig. 9 Fracture toughness of iron aluminides in different environments. (From Ref. 59.)

292

Stoloff

alloy, whereas tests in air resulted in the lowest toughness in FeAl. These results
were very similar to tensile data reported for the two alloys by the same group.
Under both tensile and fracture toughness loading conditions transgranular cleavage was the dominant fracture mode seen in air. Some intergranular fracture
regions were seen in B2-ordered Fe28 Al and Fe35 A1 specimens tested in
oxygen. This study indicated that Fe35 Al is more susceptible to environmental
embrittlement than is Fe28 Al and that grain boundaries are intrinsically weaker
in the FeAl alloy. However, complicating the interpretation of the data is the fact
that the B2 condition of Fe3Al alloys is more ductile and displays higher toughness than the DO3 condition. This suggests that it is the higher Al content in the
Fe35 Al alloy, rather than the crystal structure, that is responsible for the greater
effect of the environment.
f. Fatigue Crack Growth Recent work in our laboratory has shown that
hydrogen severely embrittles FeAl under cyclic loading conditions; see Fig. 10
(60). It appears that moisture is embrittling only at high cyclic stress intensities
for this alloy.

2. Fe3 Al
Fe3Al alloys form in the range of about 1832 at% Al. When near-stoichiometric
alloys are quenched from above the critical ordering temperature, Tc, they display
a partially ordered B2 structure. Slow cooling through Tc or extended annealing

Fig. 10 Effects of environment on fatigue crack growth of Fe35 at% Al at 25C. (From
Ref. 60.)

Embrittlement of Ni- and Fe-Based Intermetallics

293

just below Tc results in a highly ordered DO3 structure at room temperature. There
is considerable evidence that environmental embrittlement, resulting from moisture in air, is severe in binary Fe3Al alloys. For example, tensile elongation in
air is 4% but 19% in vacuum, whereas less than 1% strain is noted after precharging with hydrogen, as shown in Fig. 11 (44). Ductility of Fe28% Al alloys in
air is significantly increased by the addition of chromium (61). Although the
mechanism of improvement in ductility is unknown, it may result from modification of the Al2O3 coating that naturally forms on Fe3Al in such a way that hydrogen liberation from water vapor is reduced. It is unlikely that Cr2O3 replaces
Al2O3 as the protective oxide at these chromium levels. Unfortunately, chromium
additions do not suppress embrittlement by gaseous hydrogen or by hydrogen
introduced by electrolytic charging (44).
a. Effects of Strain Rate Reducing strain rate sharply reduces the ductility
of Fe24 at% Al tested in air at room temperature, as was discussed previously
for Ni3Al (Fig. 2) and FeAl alloys (60). Such behavior further supports the concept of environmentally induced embrittlement, because lower strain rates provide more time for release of hydrogen from the water vapor in air, and for
penetration into the alloy. The influence of strain rate on ductility of FeAlCr
alloys in air has been confirmed by Shan and Lin (55). Further, Shea et al. (44)
have shown the same trend of increasing embrittlement with lower strain rate in
hydrogen-charged Fe23.5 at% Al.

Fig. 11 Effect of environment on room-temperature ductility of Fe25 at% Al. (From


Ref. 44.)

294

Stoloff

b. Effects of Composition and Microstructure A fundamental question


with respect to environmental embrittlement of Fe3Al alloys relates to the role
of composition and, to lesser extent, the type and degree of long-range order.
Early work on FeAl alloys showed that alloys with less than 8.5 wt% Al (16.3
at% Al) exhibited substantial ductility at room temperature (62,63). Recently,
Vyos et al. (64) have shown that binary Fe16.3 at% Al, which is a singlephase disordered alloy, is not susceptible to environmental embrittlement at 25C.
Microstructural changes seemed to have no effect on ductility, which was near
25% for the 16.3 at% Al alloy in the fully recrystallized condition. In contrast,
an alloy containing 22 at% Al (displaying a two-phase ordered disordered
microstructure) showed tensile elongation of about 5% in air, with slightly higher
elongations in vacuum. Unfortunately, the 16.3 at% Al alloy displays very high
fatigue crack growth rates in air, suggesting that environmental embrittlement
may be exacerbated under cyclic loading (3).
With respect to chromium content, beneficial effects of chromium on environmental embrittlement in tension of Fe28% Al alloys have been documented
(61). The highest ductilities are produced in a partially recrystallized, quenched
condition which produces partial B2-type order.
c. Cyclic Loading Unlike the case for tensile ductility, under cyclic loading a marked susceptibility of a Fe28 at% Al4.8 Cr0.47 Nb0.2 C alloy (FA129) to environmental embrittlement has been noted, as shown in Figs. 12a and
12b (65). The effect is small in the B2 condition, but it is significant in the slow-

Fig. 12 Effects of environment on fatigue crack growth of Fe3Al alloy FA-129 at room
temperature. (From Ref. 65.)

Embrittlement of Ni- and Fe-Based Intermetallics

295

cooled DO3 condition. Embrittlement does not seem to be influenced by the microstructure, as both fully recrystallized and partly recrystallized samples showed
similar crack growth rates. The differing behavior between monotonic and cyclic
loading may be due to the repeated rupturing of oxide films that occurs under
cyclic loading. Therefore, any beneficial effect that chromium confers on the
protectiveness of Al2O3 under monotonic conditions is lost under cyclic loading.
Other solutes also have an effect. For example, the addition of 0.5% Zr to
Fe28 Al5 Cr results in a decrease in the crack growth rate as compared with
other alloys; see Fig. 13 (28). Carbon additions have been found to increase the
critical stress intensity with little effect on the crack growth rate. There exists a
limit to the beneficial effects of alloying with Zr, as a 1% Zr alloy has a higher
crack growth rate than either of the 0.5% Zr alloys.
The fatigue crack growth behavior of the iron aluminides demonstrates
typical corrosion fatigue characteristics. A more inert environment results in a
lower crack growth rate. In the alloys studied, the environment was found to
have an effect on the threshold and critical stress intensities, except in the Zrcontaining alloys. In those alloys, the threshold stress intensity was found to be

Fig. 13 Fatigue crack growth of iron aluminides in air at 25C. Note low growth rates
in alloys containing Zr. (From Ref. 28.)

296

Stoloff

insensitive to environment. (These values are for comparison purposes only, as


ASTM E-399 conditions for a valid fracture toughness test were not met; such
values would probably be considerably lower than the Kc values recorded.)
d. Stress-Corrosion Cracking Alloy FA-129, which contains 28Al, 5 Cr,
and 0.5% Nb, also has been shown to be susceptible to intergranular stress-corrosion cracking (SCC) in neutral-pH chloride-containing solutions (66). However,
little effect of cathodic potential on SCC of Fe3Al was noted, in marked contrast
to the high susceptibility of a TiAlNbVMo alloy under similar charging
conditions. These results conflict with the earlier findings of Shea et al. (44) and
Kasul and Heldt (67), who showed that binary Fe24.6 at% Al is very susceptible
to hydrogen charging (see Fig. 14) as well as work that showed that the maximum
stress intensity at fracture of cyclically loaded FA-129 is very sensitive to the
atmosphere (65). These differences may be due in part to pH effects, as Ricker
et al. (36) have reported that hydrogen-induced SCC of Ni3Al occurs only in lowpH solutions. In the case of Fe24.6 at% Al ductility is reduced under cathodic
conditions in both acidic and basic solutions. Buchanan et al. (68) also have
reported on hydrogen-induced stress corrosion cracking of iron aluminides.
D. Other Intermetallics
There have been few reports of embrittlement of ordered iron- or nickel-based
intermetallics by hydrogen or moisture, apart from those systems described in
Secs. II.A and II.B. The influence of hydrogen charging on the tensile ductility
of Ni2Cr was found to be a maximum for disordered and nearly ordered Ni2Cr
(69). Recent fatigue-crack-growth studies have shown that nearly ordered Ni2Cr
displays the lowest crack growth rate and highest threshold in air and in 3.5%
NaCl, whereas disordered Ni2Cr had the highest growth rates and lowest threshold
(70). Better fatigue-crack-growth resistance in partially ordered ane nearly ordered Ni2Cr was attributed to the occurrence of planar slip and the resultant cyclic-slip reversibility. However, cracks grow much more rapidly and thresholds
are significantly lower in 3.5% NaCl than in air for all conditions of order (S
0.5 or 0.9).
Limited embrittlement studies have been carried out on the B2 alloy FeCo
2% V (15). This alloy can be disordered by quenching from above the critical
temperature, Tc, of 720C. Hydrogen embrittlement occurs in the fully ordered
(S 1), partially disordered (S 0.4), and disordered (S 0) conditions. Brittle
transgranular cleavage is observed in ordered material in both air and in hydrogen. Dimpled fracture is observed in partially ordered and disordered samples
tested in air, whereas cleavage is again observed in hydrogen. These results support the conclusion that transgranular cleavage can be induced by hydrogen in
B2 polycrystals (FeAl, Fe3Al, and FeCo2% V) as well as in L12 single crystals

Embrittlement of Ni- and Fe-Based Intermetallics

297

(Ni3Al,Ti; Co3Ti). The influence of hydrogen on cleavage energy of these cubic


alloys needs to be established by both theoretical and experimental methods.
Theoretical calculations have been provided thus far only for the influence of
hydrogen on cleavage energy of FeAl (71), as described in the next subsection.
E.

Embrittlement Mechanisms

As mentioned in the previous sections, many ordered intermetallics display environmental embrittlement in hydrogen (charged or uncharged) environments or
in moist air at ambient temperatures. It has been shown that some intermetallics
containing reactive elements, such as FeAl and Fe3Al, exhibit more severe embrittlement in moist air than in dry hydrogen. The proposed chemical reaction for
the moisture-induced embrittlement of aluminides is (1)
2 M 3 H2O Al2O3 6 H

(1)

An alternative reaction between H2O and Ni3Al,Ti, as revealed by x-ray photoelectron microscopy (XPS) is (72)
Al x H2O Al(OH)x x H

(2)

It is the high-fugacity atomic hydrogen that rapidly penetrates into crack


tips and causes severe embrittlement. For the B2 structure, the {200} planes offer
the maximum amount of aluminum to react with water vapor. Fracture in FeAl
is by transgranular cleavage in moisture or hydrogen, mixed mode in vacuum,
and intergranular in dry oxygen. The underlying mechanism of moisture-induced
embrittlement in FeAl, Fe3Al, Ni3Al, Co3V, and other intermetallics is undoubtedly embrittlement by hydrogen, with the principal difference being the manner
in which atomic hydrogen is generated and absorbed at crack tips.
The reaction kinetics of D2O molecules with single crystals of Ni3(Al,Ti)
have been shown to strongly depend on the crystallographic orientation of the
surface (73). Chemically adsorbed D2O reacts with {100} planes and generates
deuterium when heated to at least 200 K. On the other hand, no deuterium was
detected on {111} surfaces. Therefore, the decomposition of moisture depends
on the atomic arrangement and chemical composition of crystallographic planes.
Atomic hydrogen also has been detected from the reaction of moisture with an
iron aluminide by laser desorption mass spectrometry (74). Cohron et al. (23)
have shown that when low-pressure hydrogen is disociated in contact with an
ionization gauge, the embrittlement of Ni3Al is severe. However, when the gauge
is turned off, low-pressure hydrogen is not very embrittling. These results demonstrate that the dissociation of molecular hydrogen is a necessary precursor to
severe embrittlement. The yield strength of the intermetallics is found to be insensitive to the test environment, as is typical with conventional alloys. Strain-rate
effects are very significant, due to the time dependence of hydrogen diffusion.

298

Stoloff

The highest ductility is generally obtained in dry-oxygen environments because


oxygen suppresses the reaction of Eq. (1) and allows more rapid formation of
oxides than in the presence of moisture.
Hydrogen embrittlement is a very complex phenomenon even in conventional metals and alloys. The underlying mechanisms suggested for hydrogen
embrittlement in ordered intermetallics resemble those for structural alloys and
can be grouped into four categories:
1.
2.
3.
4.

Reduction of atomic bonding across cleavage planes


Reduction of cohesive strength across grain boundaries
Reduction of dislocation mobility and crack tip plasticity
Formation of brittle hydrides

Environmental and hydrogen embrittlement in the bcc-ordered iron aluminides and in FeCoV occurs along cleavage planes rather than grain boundaries,
suggesting that cleavage strength is reduced by absorbed or adsorbed hydrogen.
Experimental results are supported by the previously cited first-principles quantum-mechanical calculations, which indicate that absorbed hydrogen significantly
reduces the cleavage strength and energy of FeAl (by as much as 2070%, depending on the hydrogen concentration) (71). Superdislocations have been suggested to be the carriers for enhanced diffusion of hydrogen at crack tips (21). For
face-centered-cubic (fcc)-ordered intermetallics, hydrogen embrittlement usually
causes intergranular fracture, suggesting that reduction of cohesive strength along
the boundaries is responsible. However, single crystals of various L12 alloys also
are embrittled (8) so that grain boundaries clearly are not necessary to observe
reduced ductility. Bond et al. (75) have shown by in situ transmission electron
spectroscopy (TEM) that hydrogen enhances dislocation mobility and crack
growth in Ni3Al, but these authors also suggested that decohesion is the failure
mechanism. More work is needed to determine whether enhanced dislocation
mobility in the presence of hydrogen occurs in other intermetallic systems.

III. EMBRITTLEMENT AT ELEVATED TEMPERATURES


Environmental degradation in iron- and nickel-based ordered intermetallics occurs also at elevated temperatures; see Table 6 (2). Hydrogen is the major embrittling agent and oxygen is beneficial at room temperature; oxygen is the major
embrittling agent at elevated temperatures (typically above 300C). At present,
only a few intermetallic systems have been studied for environmental degradation
at elevated temperatures and data are available principally for Ni3Al and Ni3Si
alloys.

Embrittlement of Ni- and Fe-Based Intermetallics


Table 6

299

Elevated-Temperature Embrittlement of Ordered Alloys

Alloy

Crystal
structure

Environmental
embrittlementa

Ref.

Ni3Al Hf B
Ni3Al Cr
(Ni,Co)3Al
Ni3Si
Ni3Si Cr
Ni3(Si,Ti) B
(Fe22Co78)3V
Ni3(Si,Ti) B
FeAl
Fe3Al

L12
L12
L12
L12
L12
L12
L12
L12
B2
DO3

76
76
93
80
80
38
81
38
95
95

: Environmental embrittlement is observed when tested in oxidizing environments; : not observed.


b
No difference in elevated-temperature tensile ductility between air and vacuum, but environmental
embrittlement is possibly masked because of a poor vacuum. Observed but reduced by alloying
with Cr.
Source: Ref. 2.
a

A. L12 Alloys

1. Ni3Al
Tensile properties of Ni3Al are sensitive to test temperature and environment.
Figure 14 compares the tensile elongation of a Ni3Al alloy (Ni21.5 Al0.5 Hf
0.1 B) (IC-145) tested in air and vacuum (103 Pa) as a function of test temperature (76). The alloy tested in air showed appreciably lower ductility than that
tested in vacuum at temperatures above 300C, and the severest embrittlement
occurred near 750C, despite the fact that Ni3Al alloys exhibited good oxidation
resistance in air. The loss in ductility generally is accompanied by a change in
fracture mode from ductile transgranular to brittle intergranular. Similar embrittlement has been observed in other Ni3Al alloys, such as B-doped Ni3Al containing iron or hafnium (77). In these cases, oxygen has been identified as the
embrittling agent. The similarity in the shape of the two curves in Fig. 14 further
indicates that embrittlement cannot be completely suppressed by a conventional
vacuum of 103 Pa. The role of air pressure in embrittlement of Ni3Al alloys has
been shown clearly for IC-136 (Ni23 at% Al0.5% Hf0.07% B) tested at
760C. There is a rapid increase in elongation as air pressure is reduced from 1
to 103 torr, and continued to increase in ductility to a pressure of 107 torr (78).

300

Stoloff

Fig. 14 Tensile ductility of IC-145 (Ni21.3 at% A10.5 Hf0.1 B) in air and in vacuum. (From Ref. 76.)

Fortunately, alloying with 68 wt% Cr considerably diminishes the degree of


embrittlement (76).
Test environments also influence the fatigue life of boron-doped Ni3Al (24
at% Al) at elevated temperatures (79). The alloy showed a sharp drop in fatigue
life at temperatures above 500C when tested in a conventional vacuum (103
Pa); see Fig. 15 (80). The decrease in the fatigue life was accompanied by a
change in fracture mode from transgranular to intergranular.

2. Ni3Si Alloys
As in the case of Ni3Al, Ni3Si alloys exhibit severe environmental embrittlement
in oxidizing environments at elevated temperatures. For Ni3Si and Ni3(Si,Ti)
alloys doped with and without boron, tensile ductility decreases sharply at temperatures above 300C in moist air. The ductility of low-Si alloys exhibits a
minimum at 600C, as seen in Fig. 16 (79); above that temperature, the ductility
increases sharply. Ductility in vacuum at 600C increases by a factor of 20. As
in the case of Ni3Al, Cr additions in the range 26 at% effectively reduce embrittlemen (81). For high-(SiTi) alloys (e.g., 21%) ductility decreases continuously
with increasing temperature and approaches zero at temperatures above 600C.
No improvement in ductility results in a conventional vacuum at these temperatures.

Embrittlement of Ni- and Fe-Based Intermetallics

301

Fig. 15 Influence of temperature on fatigue life, Nf , of Ni3AlB, tested in vacuum.


(From Ref. 80.)

B. Other Intermetallics
(CoFe)3V alloys, which are susceptible to embrittlement by water or moist air (9),
also display a sharp loss in ductility at temperatures above 500C. The minimum
ductility in both air and vacuum (5 104 Pa) is noted near Tc (910C) (81).
Ductility is lower in air than in vacuum between 500C and 910C. Embrittlement
below Tc was attributed to oxygen-induced penetration of grain boundaries. These
alloys are not embrittled at temperatures above Tc, probably due simply to the
lack of long-range order.
Iron aluminide alloys based on Fe3Al or FeAl do not appear to be susceptible to elevated-temperature embrittlement in oxidizing environments, in spite of
their extreme ambient-temperature embrittlement in moist air. The absence of
elevated-temperature embrittlement in moist air. The absence of elevated-temperature embrittlement in these aluminides is possibly related to the lack of a substantial yield anomaly, together with rapid formation of protective oxide films

302

Stoloff

Fig. 16 Yield strength and elongation of Ni18.9 at% Si, tested in air as a function of
temperature. (From Ref. 79.)

due to rapid diffusion. Further experimental studies are required to clarify these
points.
C. Embrittling Mechanisms
Embrittlement (other than via the pest reaction) has been suggested to be caused
by a dynamic effect simultaneously involving high localized-stress concentration,
elevated temperature, and gaseous oxygen (2). Such a dynamic effect involves
repeated weakening and cracking of grain boundaries as a result of oxygen absorption and penetration at crack tips. Based on a detailed study of crack growth
in Ni3Al alloys tested in oxidizing environments, a fracture mechanism of stressassisted grain-boundary oxygen penetration has been suggested to explain the
elevated-temperature embrittlement (82). This model consists of four sequential
steps: (1) occurrence of gaseous oxygen to the crack tips where a high localized
stress field is involved, (3) oxygen penetration on its atomistic form to the stress
field ahead of tips, and (4) inward development of surface cracks preferentially
along the grain boundaries, leaving some secondary cracks (83). Steps (2) and (4)
proceed continuously and repeatedly during deformation, leading to premature

Embrittlement of Ni- and Fe-Based Intermetallics

303

fracture and severe loss in ductility at elevated temperatures in oxidizing environments.


The fatigue crack growth rate of Ni3Al and the Ni3Al-based alloy IC-221
increases with increasing temperature, between 25C and 600C, even when testing is carried out in moderate vacuum (83). This behavior indicates that environmental embrittlement due to oxygen is occurring, especially because the flow
stress increases with temperature over the same range. These observations were
consistent with those of Hippsley and Devan (82) for static crack growth. By
contrast, when Fe3Al is tested at elevated temperatures in air the fatigue crack
growth rate decreases, perhaps because moisture-induced embrittlement is maximized near room temperature, as in the case of structural steels.
Alternative mechanisms for high-temperature embrittlement also have been
suggested by observations of low elevated-temperature ductility in Ni3Al-based
single crystals tested in air (8486). The increase in flow stress to a maximum
in the temperature regime of minimum ductility may play a significant role in
elevated-temperature embrittlement. Also, Yizhang et al. (87) suggest that in cast
polycrystalline Ni3Al alloys, chromium may enhance ductility through its effect
on - eutectic size and distribution and the replacement of brittle grain
boundaries by tough boundaries. Clearly, the role of chromium in improving ductility in air requires further study.

IV. ALLEVIATION OF EMBRITTLEMENT


Environmental degradation has been identified as a main cause of the low ductility and brittle fracture in many ordered intermetallics. This problem has to be
solved satisfactorily in order to use intermetallic alloys as engineering materials.
Despite their different embrittling agents, ambient-temperature and elevated-temperature embrittlement can be treated together because both involve surface reactions and are sensitive to localized-stress concentrations. Results generated to
date indicate that embrittlement can be alleviated or reduced by (a) control of
surface conditions, (b) control of grain size and shape, (c) alloy additions, (d)
processing techniques, and (e) prestrain.
A. Surface Conditions
Control of surface conditions is a straightforward way to alleviate environmental
degradation involving surface reactions. In several cases, preoxidation and formation of protective oxide scales were proven to be beneficial in reducing environmental embrittlement at ambient and elevated temperatures. Preoxidation at
1000C effectively reduced ambient-temperature embrittlement in B-doped Ni3Al
charged with hydrogen (15). Formation of protective oxide scales increases the

304

Stoloff

tensile ductility of FeAl and Fe3Al alloys in air at ambient temperatures and the
ductility of B-doped Ni3Al alloys at elevated temperatures (76). Unfortunately,
the oxide films crack after stretching a few percent and their protective effect
disappears. Surface coatings also should be useful in protecting underlying alloys
from hydrogen or oxygen penetration along grain boundaries or through the lattice; however, this effect has not yet been well demonstrated.
B. Grain Size and Shape
Columnar-grained structures have proven to be effective in reducing environmental embrittlement in Ni3Al alloys of varying aluminum contents, tested in moist
air at room and elevated temperatures (8890). For example, formation of a columnar-grained structure in boron-doped Ni3Al produced by directional levitation-zone remelting increases the ductility in air from 0.2% to 33% at temperatures in the range 600700C (88). The loss of ductility in air is accompanied
by a change in fracture mode from microvoid coalescence to intergranular in
equiaxed material. However, the columnar structure displayed mainly transgranular failure in both air and vacuum. The beneficial effect of the columnar-grained
structure with grain boundaries oriented parallel to the stress axis is attributed
to minimizing the normal stress across grain boundaries, thereby suppressing
nucleation and propagation of cracks along boundaries even when those boundaries are weakened by oxygen penetration. Hirano (89) has reported high ductility
in stoichiometric Ni3Al tested both parallel and perpendicular to the growth direction. More recent work shows that elongation increases with decreasing solidification rate; see Fig. 17 (90). The technique also is effective in Al-rich Ni3Al,
leading the authors to suggest that unidirectional solidification may be effective
in improving ductility of two-phase Ni3AlNiAl alloys. A detailed analysis of
grain-boundary chemistry and misorientation is required to fully understand these
results. However, it appears that the mostly low-angle and 3-type boundaries
resulting from unidirectional solidification are less affected by moisture than are
general grain boundaries in conventionally processed material (91). Studies of
unidirectional solidification should be extended to other intermetallic systems.
The role of grain size seems to be the same for intergranular and transgranular fracture in that refining grain size tends to reduce susceptibility to embrittlement both at room and elevated temperatures. Takeyama and Liu (92) have shown
that heat treatment of Ni3Al alloys in oxidizing atmospheres can cause embrittlement in subsequent tests at both ambient and elevated temperatures. The degree
of embrittlement is essentially zero for fine-grained (20 m) alloys, but it is very
pronounced at 200 m, as shown in Fig. 18 (92). Apparently, a thin protective
Al-rich oxide film is formed on fine-grained material, whereas a less protective,
predominantly Ni-rich film forms on coarse-grained samples. The aluminum-rich
oxide on fine-grained material may form as a result of short-circuit diffusion of
aluminum atoms from the interior to the surface.

Embrittlement of Ni- and Fe-Based Intermetallics

305

Fig. 17 Effect of growth rates (mm/h) on ductility of Ni26 at% Al. (From Ref. 90.)

Fig. 18 Effect of grain size on ductility of bare and preoxidized Ni23 at% Al0.5 Hf
0.5 B at 600C and 760C in vacuum. (From Ref. 92.)

306

Stoloff

C. Alloy Additions
Several striking examples of alleviation of embrittlement by alloying may be
cited. The ductilizing effect of boron on low-temperature fracture of Ni3Al and
Ni3Si in air is clearly associated with the inhibition of moisture-induced embrittlement (22,23). Similarly, chromium enhances high-temperature ductility of Ni3Al,
presumably by altering the kinetics of oxide film formation (76). Chromium also
has a beneficial effect on the low-temperature ductility of Fe3Al alloys in air (61).
However, chromium does not prevent environmental embrittlement under cyclic
loading conditions (44,60). Finally, zirconium additions have been shown to improve both ductility and fatigue crack growth behavior in air of Fe28 Al5 Cr
alloys, although the mechanism for such improvements remains obscure (28).
D. Processing Techniques
Boron-doped Ni3Al alloys prepared by rapid solidification are more prone to environmental embrittlement than alloys produced by conventional casting and thermomechanical processing (93). The latter exhibit much higher ductility at 750C
in vacuum than does the rapidly solidified material, although at 600C, there is
little effect of microstructure. The same workers reported that spray-formed Ni
10 at% Co24% Al with boron exhibited zero ductility in vacuum at 760C,
but ductility could be substantially improved by subsequent thermomechanical
processing. A number of factors could be influencing the results obtained with
differing processing techniques, including grain size and shape, grain-boundary
energy, oxygen and other impurities, mobile dislocation density, thermal history,
and texture. Of these, the only factor definitely linked with low ductility in many
alloys is impurity content.
E.

Prestrain

There is now considerable evidence that prestrain can reduce the severity of environmental embrittlement in Ni3Al and Ni3Si alloys (35,40,94). The magnitude of
the effect depends on the degree of prestrain as well as the temperature at which
it is performed. It remains to be seen whether this effect also can be found in
other intermetallics, especially the iron aluminides. There are several possible
mechanisms for the prestrain effect, such as dislocations acting as traps for hydrogen or the creation of compressive residual stresses in the case of prestrain by
shot peening. It has been argued that the beneficial effects of prestrain must mean
that dislocation-assisted transport of hydrogen does not occur. However, it must
be pointed out that the dislocation transport mechanism depends on the simultaneous application of strain and exposure to hydrogen, a condition clearly unmet by
the prestrain experiments.

Embrittlement of Ni- and Fe-Based Intermetallics

V.

307

SUMMARY

Many nickel- and iron-based intermetallics are embrittled by moisture and other
hydrogen-containing environments at low temperatures and by oxygen at elevated
temperatures. Crystal structure, composition, metallurgical variables, and test
conditions all play a role in determining the degree of embrittlement. Dopants
such as boron and zirconium have proven to be effective in inhibiting embrittlement under some conditions, but the effects are complex. The emphasis in the
past has been on empirical observations of behavior; relatively little theoretical
modeling has been conducted. Nevertheless, several methods of alleviating embrittlement have been noted, including compositional variations, control of grain
structure, prestrain, and the use of coatings. Such efforts must be pursued vigorously in order for these intermetallics to reach their potential as structural materials.

ACKNOWLEDGMENTS
The author is grateful to the Department of Energy, Fossil Energy AR&TD Materials Program for financial support under Subcontract No. 19X-SF521C with Martin Marietta Energy Systems.

REFERENCES
1. CT Liu. In: O Izumi, ed. Intermetallic CompoundsStructure and Mechanical Properties. Sendai: Japan Institute of Metals, 1991, pp. 703711.
2. M Takeyama, CT Liu. Mater Sci Eng A153:538547, 1992.
3. NS Stoloff. In NS Stoloff, VK Sikka, eds. Intermetallic Compounds, Physical Metallurgy and Processing. New York, Chapman & Hall, 1996, pp. 479516.
4. T Takasugi. Acta Metall 39:21572167, 1991.
5. T Takasugi, N Masahashi, O Izumi. Scripta Metall 20:13171321, 1986.
6. N Masahashi, T Takasugi, O Izumi. Metall Trans 19A:353358, 1988.
7. CT Liu, WC Oliver. Scripta Metall 25:19221937, 1991.
8. T Takasugi, O Izumi. Acta Metall 34:607618, 1986.
9. C Nishimura, CT Liu. Scripta Metall 25:791794, 1991.
10. CT Liu, M Takeyama. Scripta Metall 24:15831586, 1990.
11. CT Liu, EH Lee, CG McKamey. Scripta Metall 23:875880, 1989.
12. CT Liu, CG McKamey, EH Lee. Scripta Metall 24:385390, 1990.
13. GM Camus, N Stoloff, DJ Duquette. Acta Metall 37:14971501, 1989.
14. WJ Chia, YW Chung. Intermetallics 4:283288, 1996.
15. AK Kuruvilla, NS Stoloff. Scripta Metall 19:8387, 1985.
16. CT Liu. Scripta Metall 27:2528, 1992.

308

Stoloff

17. EP George, DP Pope. In: JA Horton, I Baker, S Hanada, RD Noebe, DS Schwartz,


eds. High Temperature Ordered Intermetallic Alloys VI. Pittsburgh, PA: Materials
Research Society, 1995, Vol. 364(2), pp. 11311135.
18. EP George, CT Liu, DP Pope. Scripta Metall 27:365370, 1992.
19. EP George, CT Liu, DP Pope. Scripta Metall 28:857862, 1993.
20. XJ Wan, JH Zhu, KL Jing. Scripta Metall 26:473477, 1992.
21. AK Kuruvilla, NS Stoloff. Metall Trans 16A:815820, 1985.
22. JW Cohron, EP George, L Heatherly, CT Liu, RH Zee. In: WO Soboyejo, TS Srivatsan, HL Fraser, eds. Deformation and Fracture of Ordered Intermetallic Materials
III, Warrendale, PA: TMS, 1997, pp. 265273.
23. JW Cohron, EP George, L Heatherly, CT Liu, RH Zee. Intermetallics 4:497502,
1996.
24. EP George, CT Liu, H Lin, DP Pope. Mater Sci Eng A 192(3):277288, 1995.
25. D Lin, Y Gu, J Guo. In: JA Horton, I Baker, S Hanada, RD Noebe, DS Schwartz,
eds. High Temperature Ordered Intermetallic Alloys VI, Pittsburgh, PA: Materials
Research Society 1995, Vol. 364(2), pp 885890.
26. A Chiba, S Hanada, S Watanabe. Mater Trans JIM 35(4):286290, 1994.
27. G Itoh, R Chikaizumi, M Kanno. In: WO Soboyejo, TS Srivatsan, HL Fraser, eds.
Deformation and Fracture of Ordered Intermetallic Materials III. Warrendale, PA:
TMS, 1997, pp. 275283.
28. NS Stoloff, DA Alven, CG McKamey. In: SC Deevi, VK Sikka, PJ Maziasz, RW
Cahn, eds. International Symposium on Nickel and Iron Aluminides: Processing
Properties and Applications, Materials Park, OH: ASM, 1997, pp. 6572.
29. CT Liu, DP Pope. In: JH Westbrook, RL Fleischer, eds. Intermetallic Compounds,
Principles and Practices, Vol. 2. New York: J Wiley, 1994, pp. 751.
30. A Chiba, S Hanada, S Watanabe. Acta Metall 39:17991805, 1991.
31. L Yang, RB McLellan. J Mater Res 11:862864, 1996.
32. JB Lumsden, BE Wilde, PJ Stocker. Scripta Metall 17:971976, 1983.
33. T Takasugi. In: NS Stoloff, AF Giamei, DJ Duquette, eds. Critical Issues in the
Development of High Temperature Structural Materials. Warrendale, PA: TMS,
1993, pp. 399413.
34. T Takasugi, S Hanada. In CC Koch, CT Liu, NS Stoloff, A Wanner, eds. High
Temperature Ordered Intermetallic Alloys VI, Pittsburgh, PA: Materials Research
Society 1997, Vol. 460, 587592.
35. T Takasugi. Intermetallics 4:S181S187, 1996.
36. RE Ricker, U Bertocci, JL Fink, MR Stoudt. In: RH Jones, RE Ricker, eds. Environmental Effects on Advanced Materials. Warrendale, PA: TMS, 1991, pp 213
225.
37. R Ricker, DE Hall, JL Fink. Scripta Metall 24:291296, 1990.
38. T Takasugi, H Suenaga, O Izumi. J Mater Sci 26:11791186, 1991.
39. CT Liu, EP George, WC Oliver. Intermetallics 3:7781, 1995.
40. CL Ma, T Takasugi, S Hanada. In: JA Horton, I Baker, S Hanada, RD Noebe, DS
Schwartz, eds. High Temperature Ordered Intermetallic Alloys VI. Pittsburgh, PA:
Materials Research Society 1995, Vol. 364(2), pp. 11591164.
41. DF Lahrman, RD Field, R Darolia. Scripta Metall 28:709714, 1993.
42. G Bergmann, H Vehoff. Mater Sci Eng A192/193:309315, 1995.

Embrittlement of Ni- and Fe-Based Intermetallics

309

43. EP George, CT Liu. In: JA Horton, I Baker, S Hanada, RD Noebe, DS Schwartz,


eds. Pittsburgh, PA: High Temperature Ordered Intermetallic Alloys VI, Pittsburgh,
PA: Materials Research Society 1995, Vol. 364(2), pp. 11311145.
44. M Shea, A Castagna, NS Stoloff. In LA Johnson, DP Pope, JO Stiegler, eds. High
Temperature Ordered Intermetallic Alloys IV, Pittsburgh, PA: Materials Research
Society 1991, Vol. 213, pp. 609616.
45. DJ Gaydosh, MV Nathal. Scripta Metall 24:12811284, 1990.
46. CT Liu, EP George. Scripta Metall 24:12851290, 1990.
47. I Baker, P Nagpal. In: R Darolia, JJ Lewandowski, CT Liu, PL Martin, DB Miracle,
MV Nathal, eds. Structural Intermetallics. Warrendale, PA: TMS, 1993, pp. 463
473.
48. P Nagpal, I Baker. Metall Trans 21A:22812282, 1990.
49. RJ Lynch, LA Heldt. Scripta Metall 30:895898, 1994.
50. RJ Lynch, LA Heldt, In JH Schneibel and MA Crimp, eds. Proceedings of the Symposium on Processing, Properties and Applications of Iron Aluminides. Warrendale,
PA: TMS, 1994, pp. 287300.
51. O Klein, I Baker. Scripta Metall 30:627632, 1994.
52. DJ Gaydosh, SL Draper, RD Noebe, MV Nathal. Mater Sci Eng A150:720, 1992.
53. RJ Lynch, KA Gee, LA Heldt. Scripta Metall 30:945950, 1994.
54. P Nagpal, I Baker. Scripta Metall 25:25772580, 1991.
55. A Shan, D Lin. Scripta Metall 27:9599, 1992.
56. DB Kasul, LA Heldt. Metall Trans 25A:12851290, 1994.
57. DB Kasul, LA Heldt. In: RH Jones, RE Ricker, eds. Environmental Effects on Advanced Materials. Warrendale, PA: TMS, 1991, pp. 6775.
58. O Klein, P Nagpal, I Baker. In: JA Horton, I Baker, S Hanada, RD Noebe, DS
Schwartz, eds. High Temperature Ordered Intermetallic Alloys VI, Pittsburgh, PA:
Materials Research Society, 1995, Vol. 364(2), pp. 11311145.
59. SH Ko, R Gnanamooorthy, S Hanada. J Mater Sci Eng A222:133139, 1997.
60. NS Stoloff. In: AW Thompson, NR Moody, eds. Hydrogen Effects in Materials.
Warrendale, PA: TMS, 1996, pp. 523537.
61. CG McKamey, CT Liu. Scripta Metall 24:219222, 1990.
62. W Justusson, VF Zackay, ER Morgan. Trans ASM 49:902921, 1957.
63. MJ Marcinkowski, ME Taylor, FX Kayser. J Mater Sci 10:406414, 1975.
64. S Vyos, S Viswanathan, VK Sikka. Scripta Metall 27:185190, 1992.
65. A Castagna, NS Stoloff. Scripta Metall. 27:673678, 1992.
66. B Bavarian, S Harutoouni, M Zamanzadeh. J Mater Sci Eng A153:613618,
1992.
67. DB Kasul, LA Heldt. In: RH Jones and RE Ricker, eds. Environmental Effects on
Advanced Materials. Warrendale, PA: TMS, 1991, pp. 6775.
68. RA Buchanan, JG Kim, RE Ricker, LA Heldt. In: NS Stoloff and VK Sikka, eds.
Physical Metallurgy and Processing of Intermetallic Alloys. New York: Chapman
and Hall, 1996, pp. 517558.
69. BJ Berkowitz, C Miller. Metall Trans 11:18771881, 1980.
70. PS Pao, SJ Gill, CR Feng, DJ Michel. Mater Sci Eng A153:532537, 1992.
71. CL Fu. In: O Izumi, ed. Intermetallic CompoundsStructure and Mechanical Properties. Sendai: Japan Institute of Metals, 1991, pp. 387396.

310
72.
73.
74.
75.
76.
77.

78.
79.

80.

81.
82.
83.
84.
85.

86.
87.

88.
89.
90.
91.
92.
93.
94.
95.

Stoloff
WJ Chia, YW Chung. J Vac Sci Technol A13:1687, 1995.
WJ Chia, YW Chung. Intermetallics 3:505509, 1995.
YF Zhu, CT Liu, CH Chen. Scripta Metall 35:14351439, 1996.
GM Bond, IM Robertson, HK Birnbaum. Acta Metall 37:14071413, 1989.
CT Liu, VK Sikka. J Metals 38(5):1921, 1986.
CT Liu. In: NS Stoloff, CC Koch, CT Liu, O Izumi, eds. High Temperature Ordered
Intermetallic Alloys II. Pittsburgh, PA: Materials Research Society, 1987, Vol. 81,
pp. 355367.
DJ Alexander. Technology Transfer Conference and Workshop on Nickel and Aluminides, Oak Ridge, TN, 1991.
WC Oliver. In: CT Liu, A Taub, NS Stoloff, CC Koch, eds. High Temperature
Ordered Intermetallics III. Pittsburgh, PA: Materials Research Society, 1989, Vol.
133, pp. 397402.
NS Stoloff, GE Fuchs, AK Kuruvilla, SJ Choe. In: NS Stoloff, CC Koch, CT Liu,
O Izumi, eds. High Temperature Ordered Intermetallic Alloys II. Pittsburgh, PA:
Materials Research Society, 1987, Vol. 81, pp. 247261.
S Miura, CT Liu. Scripta Metall 26:17531758, 1992.
CA Hippsley, JH Devan. Acta Metall 37:14851496, 1989.
W Matuszyk, G Camus, DJ Duquette, NS Stoloff. Metall Trans 21A:29672976,
1990.
K Aoki, O Izumi. J Mater Sci 14:18001806, 1979.
Z Jishan, Z Zhiya, T Yajun, Z Jinghua, Y Yang, L Yengao, D Dongsheng, H Zhiangqi. In: CT Liu, A Taub, NS Stoloff, CC Koch, eds. High Temperature Ordered
Intermetallic Alloys III. Pittsburgh, PA: Materials Research Society, 1989, Vol. 133,
pp. 549554.
M Nazmy. ABB Power Generation Ltd., Baden, Switzerland, unpublished work.
Z Yizhang, Z Tianxiang, T Yingjie, Z Bingda, Z Yaoxiao, H Zhuangqi. In: CT Liu,
A Taub, NS Stoloff, CC Koch, eds. High Temperature Alloys III. Pittsburgh, PA:
Materials Research Society, 1989, Vol. 133, pp. 555559.
CT Liu, WC Oliver. J Mater Res 4:294299, 1989.
T Hirano. Scripta Metall 25:17471752, 1991.
T Hirano, T Mawari. In: NS Stoloff, RH Jones, eds. Processing and Design Issues
in High Temperature Materials. Warrendale, PA: TMS, 1996, pp. 425432.
EP George, M Imai, T Hirano. Intermetallics 5:425432, 1997.
M Takeyama, CT Liu. Acta Metall 37:26812688, 1989.
A Taub, KM Chang, CT Liu. Scripta Metall 20:16131618, 1986.
CL Ma, T Takasugi, S Hanada. Scripta Metall 34:16331639, 1996.
I Baker, DJ Gaydosh. In: NS Stoloff, CC Koch, CT Liu, D Izumi, eds. High Temperature Ordered Intermetallic Alloys II. Pittsburgh, PA: Materials Research Society,
1987, Vol. 81, pp. 315320.

10
Nonoxide Ceramics
Nathan S. Jacobson
NASA Glenn Research Center, Cleveland, Ohio

Elizabeth J. Opila
Cleveland State University, Cleveland, Ohio

I.

INTRODUCTION

Nonoxide ceramics such as silicon carbide (SiC) and silicon nitride (Si3N4) are
promising materials for a wide range of high-temperature applications. These
include such diverse applications as components for heat engines, high-temperature electronics, and reentry shields for space vehicles. Table 1 lists a number
of selected applications. Most of the emphasis here will be on SiC and Si3N4.
Where appropriate, other nonoxide materials such as aluminum nitride (AlN) and
boron nitride (BN) will be discussed. Proposed materials include both monolithic
ceramics and composites. Composites are treated in more detail elsewhere in this
volume; however, many of the oxidation/corrosion reactions discussed here can
be extended to composites. In application, these materials will be exposed to a
wide variety of environments. Table 1 also lists reactive components of these
environments.
It is well known that SiC and Si3N4 retain their strength to high temperatures. Thus, these materials have been proposed for a variety of hot-gas-path
components in combustion applications. These include heat-exchanger tubes,
combustor liners, and porous filters for coal combustion products. All combustion
gases contain CO2, CO, H2, H2O, O2, and N2. The exact gas composition is dependent on the fuel-to-air ratio or equivalence ratio. The equivalence ratio (EQ) is
311

Selected Applications Which Are Proposed for Nonoxide Ceramics

Nonoxide ceramic

Application
Turbine engine components
Combuster liners
Blades and vanes
Piston engine components
Pistons
Valves
Industrial furnaces
Heat exchangers
Coal combustion
Particulate filters
Chemical process vessels, coal gasifiers, waste incinerators
Reentry shields

312

Table 1

Approximate use
temperatures
(C)

SiC, Si3N4
Composites

9001400

SiC, Si3N4
Composites

9001400

SiC
Composites
SiC
Composites
SiC, Si3N4

9001400
7001000
9001400

Electronic substrates

AlN

Fiber Coatings for composites

BN

Crucibles, insulators
Liquid metal containers
Processing, heat transfer
Pump bearings, cooling lines for
nuclear factors

BN
Various carbides and nitrides

9001400
6001400

SiC, Si3N4

300600

10001500
Use: 600
Processing: 1200
Use: 200
Processing: 1000
9001400

Ref.

Combustion gases; deposits: Na,


Mg, Ca sulfate, sodium vanadates
Combustion gases

Combustion gases, various deposits


Combustion gases, slag deposits

Various gases including air, H2S,


HCl
Reduced-pressure N2,O2, CO2,
N, O
Air
Air
Reduced-pressure combustion
gases
Vacuum, inert gases
Vacuum or inert gas and liquid
metals
High-pressure fluids, 10100 atm

4
5
6, 7
8
9
10
11
12
13

Jacobson and Opila

High-temperature semiconductors

SiC
Composites
SiC

Environment

Nonoxide Ceramics

313

a fuel-to-air ratio, with total hydrocarbon content normalized to the amount of


O2 and defined by EQ 1 for complete combustion to CO2 and H2O.] Figure 1
is a plot of equilibrium gas composition versus equivalence ratio. Note that as
a general rule, all combustion atmospheres are about 10% water vapor and 10%
CO2. The amounts of CO, H2, and O2 are highly dependent on equivalence ratio.
Other proposed applications for SiC include high-temperature semiconductors and for AlN include electronic substrates. In these situations, high-temperature oxidation behavior is a prime issue. Reentry shields have long been a useful

Fig. 1 Equilibrium gas composition versus equivalence ratio. (From Ref. 24.)

314

Jacobson and Opila

application of ceramic materialshere, the environment is a complex mixture


of atoms and molecules. Finally, a growing area is the application of ceramics
as pump components and ball bearings. The environment depends on the fluid
of interest, but it is generally high-temperature high-pressure water.
In this chapter we will discuss the interaction of nonoxide ceramics with
some representative environments. We begin with pure oxidation, which is important for nearly all the applications, and then proceed to the more systemspecific environments such as molten salts and high-pressure water.

II. HIGH-TEMPERATURE OXIDATION


Figure 2 summarizes the oxidation behavior for a range of nonoxide ceramics
(14). These data are presented in terms of recession in 100 h. Clearly, longterm operation requires very low recession. The key oxidation reactions can be
summarized as follows:
1. Silica formers: SiC, Si3N4, MoSi2
MSi 3/2O2(g) SiO2 MO

(1)

The oxidation of SiC and Si3N4 is described in detail in Sec. II.A.


2. Alumina formers: primarily alloys, but also AlN
2Al 3/2O2(g) Al2O3

(2)

The underline indicates that aluminum is at less than unity activity, as


would be found in an alloy.
2AlN 3/2O2(g) Al2O3 N2(g)

(3)

The oxidation of AlN is discussed more fully in Sec. II.B.


3. Borides
2MB 5/2O2(g) B2O3(l) 2MO

(4)

4. Carbides
MC O2(g) MO CO(g)

(5)

5. Nitrides
MN O2(g) MO NO(g)

(6)

or
2MN O2(g) 2MO N2(g)

(7)

Nonoxide Ceramics

315

Fig. 2 Recession due to oxidation for selected materials, with protective oxide scale
indicated. (From Ref. 14.)

Most of these materials oxidize according to a parabolic rate law, but some oxidize according to linear kinetics. Hence, Fig. 2 gives oxidation kinetics in terms
of recession to account for both kinetic laws. The reader is referred to Ref. 14 and
references contained therein for more details on the kinetics of these reactions.
The important point from Fig. 2 is that only materials which form protective
silica or alumina films are useful for long times in oxidizing environments. Most
of the borides, carbides, and nitrides do not form protective metal oxide scales.
Hence, the focus of this chapter will be on silica-forming ceramics, SiC and Si3N4,
and the alumina-forming ceramic, AlN. We shall also include a brief discussion
of BN, which is an important material in many areas of technology.

A. Oxidation of SiC and Si3N4

1. Oxygen Transport in Silica


The low oxidation rates of silica-forming ceramics are due to remarkably low
oxygen transport rates in silica. Thus, we begin with a brief discussion of oxygen

316

Jacobson and Opila

transport in silica. Amorphous silica, found for short-time oxidation or in very


clean systems, consists of a random network of SiO tetrahedra as shown schematically in Fig. 3a (15). Cristobalite, the crystalline form of silica most often
observed after long-term oxidation or oxidation in less clean environments, is
composed of the same SiO tetrahedra, but arranged in an orderly structure as
shown in Fig. 3b.
Oxygen transport through these structures occurs by two different mechanisms. First, molecular oxygen can move through the interstices between the
tetrahedra by a permeation mechanism. Alternatively, ionic oxygen can move
from network site to network site by a bond-breaking exchange mechanism. The
permeation mechanism requires less activation energy because bond breaking is
not required. Permeation through cristobalite is expected to be slower than
through amorphous silica because the regular structure is restricted to six-member
oxygen rings, whereas the irregular structure found in amorphous silica allows
for seven- and eight-member rings (15).
Permeation rates of molecular oxygen in amorphous silica have been determined (16); however, the corresponding rates in cristobalite have not been measured. The measurements are difficult because bulk cristobalite is not available
and must be nucleated and characterized at temperature. In addition, the presence
of short-circuit transport paths along cristobalite grain boundaries and cracks
formed from the to cristobalite phase transformation upon cooling make the
measurement of intrinsic oxygen transport rates in cristobalite impossible.

2. Oxidation of Silica-Forming Ceramics: Experimental


Techniques
Oxidation kinetics for these materials are determined by measuring either weight
changes using a sensitive thermogravimetric balance as shown in Fig. 4 or oxide
thickness changes using optical techniques. An example of weight change versus
time for SiC oxidation in dry oxygen is shown in Fig. 5. Because the oxidation
rate is controlled by diffusion of oxygen through the silica scale, the weight
change decreases with time according to a parabolic law and is described by the
parabolic rate constant kp. Accurate determinations of oxidation kinetics are more
difficult for silica-forming materials than for other metal-oxide-forming materials
because silica grows at such a slow rate and because silica growth kinetics are
affected by even small amounts of impurities in the sample or oxidation environment (17). A comparison of the parabolic oxidation rates for silicon, SiC, and
Si3N4 as a function of temperature is shown in Fig. 6. Similarities and differences
in the oxidation kinetics of these materials are discussed in Secs. II.A.3II.A.6.
It is instructive to begin with the relatively simple case of silicon oxidation. We will then focus on very pure SiC and Si3N4 in order to understand the

Nonoxide Ceramics

317

Fig. 3 Structure of SiO2: (a) amorphous; (b) crystalline; (c) with sodium cations.
(Adapted from Ref. 15.)

318

Fig. 4

Jacobson and Opila

Schematic of electrobalance and vertical tube furnace used for oxidation studies.

fundamental oxidation behavior of these materials. These pure materials are also
important in microelectronic applications.

3. Oxidation of Silicon
Silicon oxidation has been studied in great detail due to its application in microelectronic devices. The model of Deal and Grove (18) has been used to successfully describe silicon oxidation behavior at all but the shortest times. This model
considers two possible rate-limiting steps for the oxidation of silicon. First, the
reaction of oxygen with silicon according to a linear rate law controls the oxidation rate at short times or for thin scales. Second, transport of oxygen through

Nonoxide Ceramics

319

Fig. 5 Oxidation of SiC at 1300C in 1 atm oxygen, showing parabolic kinetics.


(Adapted from Ref. 20.)

320

Jacobson and Opila

Fig. 6 Arrhenius plot of log kp versus 1/T obtained in dry oxygen for Si, chemicallyvapor-deposited (CVD) SiC, CVD Si3N4, and Si3N4 with additives. (Data from Refs. 18,
20, and 35.)

the growing silica scale controls the oxidation rate according to a parabolic rate
law for long times and for thick scales. These two rate laws are combined into
a single expression that is valid for all times:

t
xo
1 2
1/2 A
A /4B

1/2

(8)

where xo is the oxide thickness, t is time, is the offset time which corrects for
the presence of an initial oxide layer, B is the parabolic rate constant, and B/A
is the linear rate constant. At short times, this expression simplifies to the linear
rate law xo (t )B/A; at long times, the expression simplifies to the parabolic
law x 2o Bt. For long-term applications of silica-forming materials, the parabolic
law usually describes the oxidation kinetics adequately.
Deal and Grove have derived the parabolic rate constant in terms of the
properties of silica; that is,
B

2DeffC*
N

(9)

Nonoxide Ceramics

321

and
C* kP n

(10)

where Deff is the effective diffusivity of the oxidant in silica, C * is the equilibrium
concentration of the oxidant in the oxide, N is the number of oxidant molecules
incorporated into a unit volume of the oxide layer, P is the oxidant pressure, n
is a power-law exponent, and k is a constant. The parabolic rate constant was
found to vary linearly with the oxidant pressure for both oxygen and water vapor
(i.e., n 1), indicating that the oxidant did not dissociate and molecular permeation through the silica is the rate-limiting step for oxidation in the parabolic
regime.

4. Oxidation of SiC
The oxidation of SiC is similar to the oxidation of silicon because a silica scale
forms in both cases. In this case, however, the gases generated, as shown in the
following oxidation reactions, will cause some differences:
SiC 3/2O2(g) SiO2 CO(g)

(11)

SiC 2O2(g) SiO2 CO2(g)

(12)

The linear reaction rate is different because the oxidation of C to CO(g) or CO2(g)
occurs in addition to the oxidation of Si. The parabolic oxidation rate could be
different if the outward transport of CO or CO2, rather than the inward transport
of oxygen, limited the oxidation rate. It has been shown that the oxidation rates
of SiC are about a factor of 2 slower than silicon (1921) due to the extra consumption of oxygen in the reaction with C, as predicted (22). In addition, the
activation energy for oxidation (20) is nearly identical to that of silicon (18) (see
Fig. 6) and the permeation of molecular oxygen through silica (16). Finally, the
oxidation rate of SiC is found to depend on the oxygen partial pressure (23).
Therefore, it is generally agreed that oxygen transport inward is the rate-limiting
step for parabolic oxidation of SiC. This issue has been discussed more fully in
Refs. 24 and 25.
The dependence of the parabolic rate constant on the oxygen partial pressure, given by the power-law exponent, n, in Eq. (10), gives information about
the type of oxygen transport occurring in the silica. For n 1, molecular permeation of oxygen occurs. For n 1, some dissociation of oxygen into a charged
species occurs and network diffusion of oxygen by a bond-breaking process is
likely. Zheng et al. (23) have determined the power-law exponent for the oxidation of SiC to vary between 0.6 and 0.3 at temperatures from 1200C to 1500C.
This implies that some combination of permeation and network diffusion limits
the oxygen transport through the silica scale grown on SiC, with network diffusion increasing with temperature. This is confirmed by 18O tracer diffusion stud-

322

Jacobson and Opila

ies, which show isotope exchange with network oxygen becomes increasingly
important as the temperature is increased (26). However, the similar activation
energies for molecular oxygen permeation of silica (16) and for oxidation of SiC
up to temperatures of 1500C (20) indicates that oxygen transport is dominated
by the permeation mechanism.
The effect of crystallization of silica on SiC oxidation is a complex topic.
Some investigators have used crystallization to explain observed rate changes
and/or activation energy changes. However, current evidence indicates that many
of these deviations in rate law and activation energies can be explained by impurities in either the environment or the SiC (17,20,27,28). Ogbuji (29) has shown
that silica scales fully crystallized during argon anneals at 1300C do result in
slower oxidation rates by about a factor of 30, but that fully crystalline scales
are not found in actual practice because amorphous silica is continually formed
during the course of SiC oxidation.

5. Oxidation of Si3N4
Like SiC, the oxidation of Si3N4 results in the formation of a silica layer and the
generation of gaseous products according to the following simplified reaction:
Si3N4 3O2(g) 3SiO2 2N2(g)

(13)

An additional complication, however, is the formation of a suboxide layer of


amorphous silicon oxynitride of variable stoichiometry (30,31). The oxidation
reaction can be written as (32)
Si3N4(1x) O6x 3xO2(g) Si3N4(1xdx) O6(xdx) 2xN2(g)

(14)

for x varying between 0 and 1. Measurements of the oxidation kinetics of Si3N4


have shown they are parabolic but significantly slower than oxidation of both
silicon (33) and SiC (20) to temperatures of 1500C. Pure Si3N4 is the slowest
oxidizing material known today. The higher activation energy for Si3N4 reflects
the additional energy required in the breaking of bonds for the nitrogenoxygen
substitution reaction. The oxidation rates were found to be dependent on the
oxygen partial pressure, but independent of the nitrogen partial pressure (33). In
this case, the oxidation reaction is limited by oxygen transport and reaction in
the oxynitride layer (32) rather than by oxygen transport in silica.

6. Oxidation of Additive-Containing Materials


The above discussions have considered only very pure materials. SiC and Si3N4
materials used for structural applications often contain additives to aid sintering.
These additives affect long-term oxidation behavior. First, additives can diffuse
into the silica scale during oxidation (3438). Impurities present in silica increase
oxygen transport rates by modification of the silica network structure and thereby

Nonoxide Ceramics

323

increase oxidation rates of the silica-forming material (39). Second, additives


such as Y2O3, La2O3, MgO, and CeO2 present in significant amounts (410%)
will diffuse to the silica surface and form discrete particles of silicates often of
the M2Si2O7 phase where M Y, La, Ce. In these cases, the oxidation kinetics
are still parabolic to times as long as 1000 h (D.S. Fox, personal communication,
1997), but are limited by the outward diffusion of metal cations to the matrix
oxide interface (34,35). In addition, for silicon nitride ceramics, the presence of
impurities (40) and additives (41) prevent the formation of the silicon oxynitride
inner layer; thus, oxidation rates of Si3N4 in practical applications are never as
low as those found for pure chemically-vapor-deposited (CVD) Si3N4. As additive
levels and impurities increase further, the oxidation kinetics often do not follow
simple parabolic rate laws. These more complex oxidation kinetics have been
modeled in various ways (42,43). Finally, for systems containing large amounts
of additives, 10%, at intermediate temperatures of 10001200C, reactions to
form silicates occur in the grain boundaries with a volume expansion large
enough to cause disintegration of the ceramic (44).

7. Thermal Cycling Effects


Long-term applications of structural ceramics, such as heat engines, require thermal cycling. Thermal cycling is a concern for several reasons. First, upon cooling,
cristobalite undergoes the to phase transformation with an accompanying
3% volume contraction (45). This volume change causes cracks to form in cristobalite upon cooling. Second, the thermal mismatch between SiC or Si3N4 and
cristobalite, shown in Fig. 7, results in tensile stresses in the oxide layer upon
cooling. Cracks in the oxide may then form. However, cyclic oxidation tests at
1300C in 5-h cycles for 1000 h have shown few deleterious effects on the oxidation kinetics when measured by weight change (46), as shown in Fig. 8. One
possible explanation is that upon reheating, stress is relieved and the cracks heal.
Note that the tensile stresses formed in these ceramicoxide systems contrast
with those in superalloyoxide systems where compressive stresses form in the
scale on cooling and oxide spallation is typically observed.
B. Oxidation of Other Nonoxide Ceramics
BN is a useful crucible material in a vacuum or inert atmosphere. However, in
oxygen, it readily forms a liquid B2O3 scale which is not protective (47). Furthermore, this B2O3 scale readily reacts with water vapor and forms stable volatile
HBO(g) species, such as HBO2(g), H3BO3(g), and H3B3O6(g). These species
have strongly negative free energies of formation and form even with parts-permillion (ppm) levels of water vapor. A similar situation exists with TiB2, which
is an attractive structural material due to its high strength (48).

324

Jacobson and Opila

Fig. 7 Thermal expansion of SiC, Si3N4, amorphous SiO2, and crystalline SiO2 as a
function of temperature. (From Ref. 24.)

AlN is currently a candidate for electronic substrates due to its high thermal
conductivity (9). As an alumina former, it is expected to exhibit slow oxidation
kinetics. However, it appears to oxidize quite rapidly above about 1000C, with
rates quite dependent on grain size, porosity, the presence of second phases, and
impurity content (49). In many instances, linear kinetics have been reported, in
contrast to alumina formation on metal alloys (50). In addition, it is well established that water vapor in the oxidizing stream leads to extensive attack (51).
The reasons for this extensive alumina formation in dry and wet oxygen are not

Nonoxide Ceramics

325

Fig. 8 Cyclic oxidation kinetics of SiC and Si3N4 obtained in air at 1300C. (Adapted
from Ref. 46.)

clear. Clearly, nitrogen must escape through the alumina scale, which may lead
to micropore formation. It has been suggested that the injection of N3 leads to
excess vacancy formation in the alumina and more rapid diffusion rates (52). In
addition, the coefficient of thermal expansion mismatch between AlN and Al2O3
may lead to scale cracking (53).

III. COMPLEX ENVIRONMENTS


In most applications, ceramic materials are subjected to more aggressive environments than high-temperature oxygen alone. These are outlined in Table 1. As in
the case of pure oxidation, most of the available data on interactions of nonoxide
ceramics in complex gas mixtures are for SiC and Si3N4.
A. Water Vapor
There is general agreement that water vapor enhances the oxidation rate of silicon, SiC, and Si3N4. There is a large amount of disagreement as to the magnitude
of this effect for SiC and Si3N4. This disagreement arises in part from the complex
effects water vapor has on the growing silica scale. These effects include enhanced impurity transport to the silica scale in water vapor containing atmo-

326

Jacobson and Opila

spheres, enhanced solubility of water in the silica scale, alterations in the silica
scale viscosity and structure and, finally, formation of volatile silicon hydroxide
and oxyhydroxide species. Each one of these topics will be discussed separately
below. Again, references to the oxidation behavior of silicon in water vapor are
made for comparison.
Oxidation of silicon by water vapor occurs by the following reaction:
Si 2H2O(g) SiO2 2H2(g)

(15)

The enhanced growth rates of SiO2 by this reaction, relative to dry oxygen, has
commercial application in the more rapid fabrication of microelectronic devices.
Oxidation of SiC by water vapor occurs by the reactions (54)
SiC 3H2O(g) SiO2 CO(g) 3H2(g)

for T 1400K

(16)

SiC 2H2O(g) SiO2 CH4(g)

for T 1400K

(17)

and for Si3N4 by the reaction


Si3N4 6H2O(g) 3SiO2 2N2(g) 6H2(g)

(18)

Note that in each case, the solid product SiO2 is formed with the generation of
additional gaseous products.

1. Impurity Transport
Impurities normally found in water-vapor-containing environment form M
OH(g) species, where M is the impurity element such as Na, K, Fe, and so forth.
Because MOH(g) species are so thermodynamically stable, the quantity of impurities transported to a silica scale forming on SiC or Si3N4 in a water-vapor
containing environment is increased. This increased contamination of the silica
scale results in faster transport rates of oxidant through the scale, and thus increased oxidation rates of SiC or Si3N4. This effect has been identified for both
Na (55) and K contamination (56) during the oxidation of SiC.

2. Enhanced Solubility of Water in the Silica Scale


Deal and Grove (18) have shown that the parabolic oxidation rate of silicon in
water vapor is increased by about an order of magnitude over the rate found in
dry oxygen. This is explained by examination of Eq. (9). Although the diffusivity
of water vapor in silica is almost two orders of magnitude slower than molecular
oxygen, the solubility of water is nearly three orders of magnitude larger than
oxygen. The net result is that parabolic oxidation rates of silicon in water vapor
are more than one order of magnitude larger than those observed in dry oxygen.
This explanation is directly applicable to the discussion for SiC and Si3N4 because
silica is the oxidation product for these materials. This enhancement in parabolic

Nonoxide Ceramics

327

oxidation rate constants has been observed for SiC (55,57) and in some cases
for Si3N4 (58).
Deal and Grove (18) have also demonstrated that the parabolic oxidation
rate has a power-law dependence on the water-vapor partial pressure, with the
power-law exponent, n [Eq. (10)], equal to 1, indicating that molecular water
diffusion is the rate-controlling step in the oxidation of silicon. Power-law exponents for the oxidation of SiC of 0.67 (59) and 0.67 to 0.85 (54) have been
obtained. This indicates dissociation of water into a singly-charged species is
probable. The discrepancy between the findings for silicon and SiC is unexpected
because the transport properties of silica should be independent of the substrate
material. A complex relationship between parabolic rate constant and water-vapor
partial pressure has been observed for Si3N4 (58), but this may be explained by
silica volatility, as described in Sec. III.A.4.

3. Alterations in the Structure of Silica


It has been shown that the viscosity of amorphous silica decreases as the hydroxyl
content increases (60). Hydroxyl groups are effective in breaking SiEOESi
bonds in amorphous silica. It is suggested that the resulting silica allows the more
rapid permeation of molecular oxygen (57). The reduced viscosity of amorphous
silica in conjunction with increased amounts of gaseous products results in the
formation of bubbles in the scale formed on SiC (54), as shown in Fig. 9. These

Fig. 9 Bubbles formed in SiO2 due to oxidation of SiC in 90% water vapor/10% O2 at
1200C. (From Ref. 54.)

328

Jacobson and Opila

bubbles, in turn, create shorter transport paths for the oxidant to the SiC/SiO2
interface by decreasing the effective oxide thickness, thereby increasing the oxidation rate. In contrast, bubbles are not observed in the scales formed on SiC in
dry oxygen or on silicon in wet oxygen.

4. Silica Scale Volatility


Water vapor reacts with the silica scale formed on SiC and Si3N4 to form volatile
hydroxide and oxyhydroxide species by the following reactions:
SiO2 2H2O(g) Si(OH)4(g)
SiO2 H2O(g) SiO(OH)2(g)

(19)
(20)

These volatile silicon hydroxides and oxyhydroxides have been identified experimentally using the transpiration technique (61) as well as mass spectrometry (62).
The reaction of water vapor with SiC or Si3N4 involves the oxidation reaction
[Eqs. (16)(18)] and the simultaneous linear volatilization reaction [Eqs. (19)
and (20)] resulting in overall paralinear kinetics (63). Paralinear kinetics, as measured by weight change, are shown in Fig. 10. At long times, the weight change
and recession can be approximated by the linear volatilization rate alone. At this
time, a steady state is achievedthe silica scale is consumed at the same rate it
is formed, leaving a constant oxide thickness. The volatilization rate of silica is
controlled by transport of the volatile species through a gaseous boundary layer

Fig. 10

Schematic showing components of paralinear kinetics. (From Ref. 63.)

Nonoxide Ceramics

329

(63). This boundary-layer-controlled volatility rate can be expressed in terms of


the application conditions such as gas velocity, total pressure, and water-vapor
partial pressure. For Si(OH)4(g) formation, the volatility rate has the following
dependences:
kl

v 1/2P(H2O)2
1/2
P total

(21)

where kl is the linear volatility rate, v is the gas velocity, and P is pressure. Thus,
in high-pressure and high-velocity applications, such as a gas turbine engine,
the recession of a SiC or Si3N4 component by the volatility mechanism can be
significant.
In summary, water vapor has deleterious effects on the durability of silica
formers through a number of different mechanisms: increased transport of impurities to the oxide surface, increased oxide formation rates due to impurity effects,
enhanced solubility of water vapor, short-circuit paths for oxidant transport, increased permeability of the oxidant, and, finally, consumption of the silica scale
and component recession due to silica volatility.
B. Carbon Dioxide
Although CO2 is a major component of combustion gases (Fig. 1), there are only
a few limited oxidation studies of SiC in CO2 (6467). It has been found that
the oxidation rates of SiC in CO2 are less than those in oxygen. Because the
oxide growth rates are so low, it is difficult to determine whether the kinetics
are linear, parabolic, or more complex. The oxidation weight gain for SiC in CO2
is shown compared to that observed in oxygen and a 50% water vaporoxygen
mixture in Fig. 11 (67). Thus, in a complex combustion environment, the effects
of CO2 as an oxidant are negligible.
C. Effects of Low P(O2), Reducing Gases, H2S, Cl2

1. SiO(g) Formation
A unique, but very important issue with silica formers is the highly stable volatile
suboxide, SiO(g). Consider the free energy of formation at 1500 K:
/2Si 1/2O2(g) 1/2SiO2(s), G 322 kJ/mol

(22)

G 227 kJ/mol

(23)

Si /2O2(g) SiO(g),
1

Note that these equations are normalized to 1 mol of oxygen atoms. The free
energies of formations are close, indicating that SiO(g) can readily form.
There are two conditions which lead to SiO(g) formation (68): active oxidation and oxidation in mixed oxidizing/reducing gases. In the active oxidation

330

Jacobson and Opila

Fig. 11 Comparison of SiC oxidation in 50% H2O/O2, O2, and CO2 at 1200C. (From
Ref. 67.)

case, the partial pressure of oxygen is too low to form a stable SiO2 film, but
sufficient to form SiO(g). This can occur in certain heat-treating environments
(69). Let us begin with a bare SiC surface. As the partial pressure of oxygen is
increased, SiO(g) will form in increasing quantities. Then, SiO(g) formation will
stop and a stable SiO2 film will form. This is the active-to-passive transition.
Wagner has derived this for pure silicon (70). His results can be easily extended
to SiC and Si3N4 (71). The active-to-passive transition occurs when sufficient
SiO(g), via Eq. (23), is generated to satisfy the SiC/SiO2 condition for equilibrium. There is some controversy about the exact equilibrium condition, but
reasonable agreement with measurements is obtained from the following equilibria:
2SiC SiO2 3Si 2CO(g)

(24)

The transition P(O2) for active-to-passive oxidation is calculated based on equilibrium conditions and diffusion through the gas boundary layer to the sample.
Calculated and measured active-to-passive transitions (7274) are shown in
Fig. 12.
Now, consider the case of passive-to-active transition. Beginning with a
stable SiO2 film on SiC; as the partial pressure of oxygen is lowered, the film

Nonoxide Ceramics

331

Fig. 12 Calculated active-to-passive and passive-to-active transitions and measured


active-to-passive transitions. (Data from Refs. 7274.)

will become unstable and SiO2 will react to form SiO(g). The transition pressure
is calculated from the decomposition of the protective SiO2 film (70):
SiO2(s) SiO(g) 1/2O2(g)

(25)

The calculated passive-to-active line is shown in Fig. 12note that it is several


orders of magnitude lower than the active-to-passive transition.
A second route to SiO(g) formation occurs in a mixture of oxidizing and
reducing gases (68). These can be present in a fuel-rich combustion situation

332

Jacobson and Opila

with appreciable amounts of CO2, CO, H2O, and H2. It has also been observed
in laboratory experiments with CO/CO2 (73) and H2 /H2O (72). Thus, a sequence
of reactions might be
SiC 3H2O(g) SiO2 CO(g) 3H2(g)

(26)

SiO2 H2(g) SiO(g) H2O(g)

(27)

This leads to paralinear kineticssimultaneous oxide growth and linear reduction kinetics.

2. Effects of H2S and Cl2


Hydrogen sulfide is found in coal gasifiers and various other petrochemical environments (5). At low oxygen potentials, when SiO2 does not form, both SiO(g)
and SiS(g) may form:
Si3N4 3H2O(g) 3SiO(g) 3H2(g) 2N2(g)

(28)

Si3N4 3H2S(g) 3SiS(g) 3H2(g) 2N2(g)

(29)

These reactions can lead to material consumption (5); however, when the oxygen
potential is high enough to form SiO2, the corrosion rate drops substantially.
Chlorine, which may be found in certain chemical process environments,
also leads to volatile products (75). In the case of SiC, reaction occurs with the
silicon, leading to formation of various silicon chlorides with residual carbon.
As the oxygen potential is increased, attack becomes less severe (75).
Current interest focuses on chlorine as HCl in waste incineration applications (76). Here, the situation is quite complex, involving not only HCl, but also
a range of deposits.
D. Oxidation in the Presence of Impurities and Deposits

1. Low Levels of Na and K


Small amounts of impurities have several possible effects on the structure and,
thus, the transport properties of silica. First, impurities can nucleate the formation
of cristobalite at temperatures and times where amorphous silica would be expected. Second, impurities can modify the silica network, as shown in Fig. 3c.
Breaking up the network tends to increase the oxygen transport rate through silica. Small amounts of alkali metals act as network modifiers and can increase
oxidation rates an order of magnitude (56,77,78).

2. Na2SO4 Deposits
As impurity levels increase, actual deposits may form. The chemistry of these
depends on application, as shown in Table 1. Corrosive deposits have long been

Nonoxide Ceramics

333

known to be an important issue for metals and alloys in various high-temperature


applications (79). Here, we shall focus on sodium sulfate deposits on silica. Many
of the general principles described for this system apply to the other cases.
It is difficult to simulate the effect of an actual deposit. Simple laboratory
experiments involve a one-time deposition of a salt on a test coupon, followed
by a heat treatment. This type of experiment has the advantage of precise control
of experimental variables such as temperature and gas composition. However,
the actual situation involves a continuous deposition process. This can be done
in a laboratory furnace. However, this can be accomplished more effectively by
seeding a flame in a burner rig, as illustrated in Fig. 13. Figure 14 is a comparison
of SiC treated in a burner without salt and with a flame seeded with NaCl. Note
the extensive corrosion in the latter case.
The formation of Na2SO4 occurs when ingested NaCl reacts with sulfur
impurities in the fuel (80,81):
2NaCl(v) SO3(g) H2O(g) Na2SO4(l) 2HCl(g)

(30)

The source of NaCl depends on the application: It may be from a marine environment in the case of a heat engine, or from the process chemicals in the case of
an industrial furnace. Under the appropriate conditions, Na2SO4 forms as a condensed phase, depositing on parts. Figure 15 shows the calculated dew points

Fig. 13 Schematic of jet fuel burner used for corrosion studies.

334

Jacobson and Opila

Fig. 14 Optical micrographs of sintered SiC coupons with carbon and boron additives,
oxidized in a burner rig at 1000C (a) 46 h with no sodium and (b) 13.5 h with a sodiumchloride-seeded flame. (From Ref. 24.)

for Na2SO4 deposition (81). The rates of deposition are also critical and have
been treated in detail (82).
The most useful interpretation of this process is with the acidbase theory
of molten salts. Na2SO4 decomposes to Na2O, which is the key reactant:
Na2SO4 Na2O SO3(g)

(31)

Note that the overpressure of SO3 sets the chemical activity of Na2O. A high
activity of Na2O is a basic molten salt; a low activity of Na2O is an acidic molten
salt. Because SiO2 is an acidic oxide, it is readily attacked by a basic molten salt:
2SiO2(s) Na2O Na2O2(SiO2)(l)

(32)

Nonoxide Ceramics

335

Fig. 15 Calculated dew points for Na2SO4 deposition. (From Ref. 24.)

This describes the product observed in Fig. 14: A solid, protective SiO2 layer
has been replaced by a liquid nonprotective sodium silicate layer. The liquid layer
allows rapid diffusion of oxygen inward and carbon monoxide outward, leading
to accelerated oxidation. The accelerated oxidation provides additional SiO2 for
reaction (32) and can lead to rapid consumption of the SiC.
The conditions for reaction (32) to proceed can be calculated from basic
thermodynamics. We can assume unit activities for the SiO2 and Na2O2(SiO2)(l)
and, hence, the threshold Na2O activity [a(Na2O)] for SiO2 dissolution is given by
a(Na2O) exp


G
RT

(33)

Here, G is the free-energy change for reaction (32). At 900C, the minimum
a(Na2O) is 1011, meaning any activity greater than this will lead to silicate formation. Because the activity of Na2O is set by P(SO3) in reaction (31), this means
that a P(SO3) of 5 105 bar or less will lead to silicate formation.
These predictions hold in the actual burner, where the P(SO3) is set by the
sulfur content of the fuel. Figure 16 shows a series of quartz coupons treated in
a burner with a low- (0.05%) and high- (0.5%) sulfur fuel. The low-sulfur fuel
has a low P(SO3) (basic salt) and, thus, allows silicate formation, whereas, the
high-sulfur fuel has a higher P(SO3) (acidic salt), which suppresses silicate
formation.

336

Jacobson and Opila

Fig. 16 Quartz coupons treated in burner rig for 1 h with a 2 ppm sodium (as NaCl)
seeded flame: (a) No. 2 diesel fuel (0.5% sulfur), 5 h; (b) Jet A fuel (0.05% sulfur). (From
Ref. 81.)

Actual ceramic systems are more complex than the quartz coupons used
in the preceding experiment. Ceramic systems are silica films on SiC and Si3N4
and may contain additives such as refractory oxides and carbon. We have seen
that carbon, in silicon carbide and as an additive, leads to extensive salt corrosion.
Using an electrochemical sensor, it can be shown that carbon tends to drive so-

Nonoxide Ceramics

337

dium sulfate more basic. The exact chemical mechanism for this is not clear, but
it may involve a Na2S intermediate (81).

3. Mixed Sulfates, Vanadates, Slags


This general concept of acidbase reactions in molten-salt corrosion extends to
other deposits as well. Ingested sea salt leads to MgSO4 and CaSO4 deposits as
well as Na2SO4 deposits (83). This mixed sulfate has been described by moltensalt solution models (84) and leads to Mg and Ca silicates. Impure fuels lead to
vanadate deposits. V2O5 is an acidic molten salt and, thus, does not react with
SiO2. However, it is reported to accelerate Si3N4 oxidation, possibly due to the
solubility of SiO2 in V2O5, which may be further enhanced by the presence of a
Y2O3 in Si3N4 (85).
A more complex case is that of a molten slag, which may contain up to
10 different oxides. Here, it is more difficult to define basicity; however, a useful
index has been the weight percent ratio of basic to acid oxides (86). The basic
and acidic oxides encountered are
Basic oxides: Na2O, K2O, MgO, CaO, Fe2O3
Acidic oxides: SiO2, Al2O3, TiO2

Fig. 17 CVD SiC oxidized for 1 h at 1800C. (Courtesy of D. Fox, NASA Glenn.)

338

Jacobson and Opila

Fig. 18 (a) Total pressure [P(SiO) P(CO)] generated from SiC/SiO2 interactions.
(b) Total pressure [P(SiO) P(N2)] generated from Si2N2O/SiO2 interactions. (From
Ref. 24.)

Nonoxide Ceramics

339

As in the Na2SO4 case, the basic slag tends to cause more rapid material degradation than the acidic slag. Under some conditions, metal silicides may form with
a basic coal slag (4).
E.

High-Temperature Effects

Silica melts at 1723C. Transport rates are very high in a liquid scale and rapid
oxidation occurs, as shown in Fig. 17. Some of the bubbles in this sample are
from escape of CO(g), but some are also from the interaction of SiC and SiO2
at high temperatures:
SiC(s) 2SiO2(s) 3SiO(g) CO(g)

(34)

Figure 18a is a plot of total pressures generated by this reaction (87). Note that
for carbon saturated SiC, the pressures are much higher, leading to a lower upper
use temperature. An analogous situation exists for Si3N4, but based on the Si2N2O
interaction with SiO2:
Si2N2O(s) SiO2(s) 3SiO(g) N2(g)

(35)

The total pressure is shown in Fig. 18b.

IV. REFRACTORY OXIDE COATINGS


One possible solution to the preceding issues of volatility and molten-salt corrosion is the use of refractory oxide coatings on SiC and Si3N4. Refractory oxides
are generally more chemically inert and this approach offers the possibility of
combining the benefits of both materials. Table 2 lists some coefficients of thermal expansion (CTEs) for SiC, Si3N4, and refractory oxides. There is a good

Table 2 Coefficients of Thermal Expansion of SiC,


Si3N4, and Several Refractory Oxides
Material
SiC
Si3N4
Mullite (3Al2O32SiO2)
Alumina (Al2O3)
Partially stabilized zirconia
(0.08 Y2O3ZrO2)
Source: Adapted from Ref. 89.

CTE (106 K1)


5.2
3.2
5.4
9
10

340

Jacobson and Opila

Fig. 19

Micrograph of polished cross section of mullite on SiC. (From Ref. 88.)

match between SiC and mullite. Mullite coatings on SiC were first developed at
Solar Turbines (San Diego, CA) and further developed at NASA Glenn (88,89).
The critical processing issues are surface roughening of the SiC for adherence
and application of a fully crystalline mullite coating. A mullite coating on a SiC
coupon is shown in Fig. 19. There are clear advantages with mullite coatings
in Na2O-induced corrosion due to formation of high-melting sodium-aluminosilicates, as opposed to lower-melting sodium silicates. However, the chemical
activity of silica in mullite is only about 0.4, so reactions which volatilize SiO2
can still occur readily.
It may be possible to apply other refractory oxides to SiC. As Table 2
shows, the problem of CTEs must be considered. One approach is graded coatings
from SiC to a refractory oxide such as alumina or zirconia.

V.

HYDROTHERMAL CORROSION OF SiC AND Si3N4

Hydrothermal corrosion involves the attack by water at high temperatures and


high pressures. Typical conditions are shown in Table 1. These conditions are
often near the critical point of water. Table 1 also shows some potential applications in which these environments may be encountered. Another promising application of SiC and Si3N4 is a vessel for oxidation of waste materials to form safe
products using supercritical water (90).

Nonoxide Ceramics

341

The hydrothermal corrosion of SiC may form silica by the following reactions (91,92):
SiC 2H2O(g) SiO2 CH4(g)

(36)

SiC 4H2O(g) SiO2 CO2(g) 4H2(g)

(37)

Given nearly equal amounts of SiC and water vapor at high temperatures, free
carbon can form (92,93):
SiC 2H2O(g) SiO2 C 2H2(g)

(38)

This provides a method for synthesis of carbon films. However, in most applications, there is excess water and SiO2 is formed as a protective film. The hydrothermal corrosion of Si3N4 is similar and occurs by
Si3N4 6H2O(g) 3SiO2 4NH3(g)

(39)

At temperatures above 430C, NH3(g) decomposes to N2(g) and H2(g).


In both cases, the silica is eventually dissolved by the high-pressure
water as
SiO2 OH HSiO3

(40)

In the case of Si3N4, this leads to attack of the grain-boundary phases and dislodging of the grains (94). Methodical studies have been done on the effect of additives in Si3N4 on its hydrothermal corrosion behavior (13,94). The morphology
of attack has been correlated with the type of additive and its resistance to hydrothermal attack.
Ceramics which form oxides other than silica behave differently under hydrothermal conditions. AlN forms AlOOH (95). Other ceramics such as TiC and
ZrC, which do not form protective oxides at high temperatures, also do not form
protective oxides under hydrothermal conditions (91).

VI. LIQUID METALS


Ceramics have long been used as containers for processing of liquid metals. More
recent applications involve use of liquid metals as a heat-transfer medium. A
wide range of nonoxide ceramics have been considered for these applications
(12,96,97). Dissolution of the ceramic and capillary action are two important
issues in liquid metal corrosion (12). Dissolution can be diffusion controlled or
interface controlled. Local variations in dissolution due to phase and structural
differences can lead to surface roughening. Capillary action leads to internal at-

342

Jacobson and Opila

tack of porous ceramics. More importantly it can lead to substantial grain-boundary attack, depending on the relation of the solidliquid interfacial energy and
the grain-boundary energy (12).

VII. EFFECT OF OXIDATION/CORROSION ON


MECHANICAL PROPERTIES
It is evident from current materials research that we cannot view one phenomenon
such as oxidation or corrosion alone, but rather must view its effect on the entire
system. Nearly all the applications listed in Table 1 utilize the ceramic in some
type of load-bearing situation. A critical question that arises is how oxidation
and corrosion alter the ability of the ceramic to bear a load (98112). Again, the
emphasis here is on SiC and Si3N4.
Two recent reviews summarize the critical issues involved in corrosion
mechanical property interactions (98,99). Many of the studies in this area deal
with the change in mechanical properties after corrosion. However, these processes often act together. The oxidation rate for various types of Si3N4 is increased
by the application of either compressive or tensile stresses (100,101).
It is well established that short-term oxidation tends to blunt cracks and
may actually increase strength (103). Long-term isothermal oxidation (5000 h)
for SiC containing B, C, and Si leads to strength increases, whereas the same
treatment for SiC containing Al2O3 and WC leads to strength decreases (104).
Little strength degradation of a variety of commercial SiC and Si3N4 materials
was observed for 3500-h burner rig tests with 12-min cycles (105). Recent research (106) shows that the effects of water vapor on strength are complex; however, in some cases, the rapid oxidation rates lead to flaw healing. In the case of
stress rupture for Si3N4, water vapor has little effect (107).
Molten-salt corrosion (108110) leads to extensive pitting and strength reduction. Figure 20 is a fracture origin due to a corrosion pit from molten-salt
attack (108). This indicates the need for tougher, greater flaw-resistant ceramics.
In the case of Si3N4, it has been shown that corrosion reactions often affect the
grain-boundary phase. Various mechanical properties are often related to the
stability of this phase. Compositional changes in this grain-boundary phase
may lower the threshold stress intensity for crack growth and also increase the
creep rate.
Corrosion reactions at a growing crack tip are quite important as evidenced
by moisture-assisted crack growth of ceramics (111). Henager and Jones (112)
have shown that the presence of Na2SO4 increases the slow crack growth velocity
by a factor of 2 over air at 1573K.
Similar effects occur in the case of hydrothermal corrosion. Pitting and
dislodging of grains in various types of Si3N4 leads to substantial strength reduc-

Nonoxide Ceramics

343

Fig. 20 Fracture origin for sintered -SiC reacted with Na2SO4 for 48 h at 1000C. The
area noted in (b) is enlarged in (c). (c) The preferential grain-boundary attack is observed
at arrows. (From Ref. 108.)

344

Jacobson and Opila

tions. Analogous to the high-temperature situation, changes in the grain-boundary


phase of Si3N4 lead to changes in the mechanical behavior. Weakening and dissolution of this phase leads to substantial strength reductions. Similarly, liquid metal
attack of grain boundaries will lead to strength reductions (96).

VIII. CONCLUSIONS
Nonoxide ceramics are a promising class of materials for a wide range of applications at high temperatures. Before application, their interactions with the hightemperature environment must be well understood. The silica-forming ceramics
(SiC and Si3N4) exhibit the best oxidation behavior and the focus of this chapter
is on them. However, some information is included on AlN and BN, which show
promise for several specific applications.
High-temperature oxidation is critical to most applications. The oxidation
of SiC is rate limited by the diffusion of oxygen inward through the growing
silica scale. The oxidation of Si3N4 is more complex: Here, the oxynitride scale
plays a role. Most engineering ceramics contain additives, which tend to increase
oxidation rates.
Actual applications involve complex gas mixtures. The effects of water
vapor, carbon dioxide, low oxidant pressures, and mixed oxidizing/reducing
gases are discussed. Water vapor enhances oxidation, whereas carbon dioxide is
a less effective oxidant. Low oxidant pressures and mixed oxidizing/reducing
gases lead to SiO(g) formation.
Corrosive deposits are also encountered in some applications. These include sodium sulfates, vanadate, and slags. Deposit-induced corrosion can be
described with the acidbase theory of oxides. Because silica is an acidic oxide,
it is readily attacked by a basic molten salt.
These problems may be minimized with the application of a refractory
oxide coating. Mullite shows a good thermal expansion match to SiC and is a
promising starting point. It may be possible to develop other more refractory
oxide coatings.
Some proposed applications involve hydrothermal conditions: water at high
temperatures and pressures. Hydrothermal corrosion is analogous to high-temperature oxidation in many ways. Ceramics which form protective high-temperature
oxides also form protective hydrothermal oxides. The form of corrosive attack
is also similar.
Finally, the interaction of corrosive attack and degradation of mechanical
properties is discussed briefly. It is essential to understand this before these materials can be applied. Microstructural changes such as pitting and grain-boundary
attack lead to a concurrent loss of mechanical integrity. The development of
tougher ceramics and coatings are necessary to minimize this problem.

Nonoxide Ceramics

345

REFERENCES
1. HB Probst. Structural ceramics in heat enginesThe NASA viewpoint. Proceedings of International Symposium on Ceramic Components for Heat Engines, Japan,
1983, pp. 4558.
2. AF McLean. Materials approach to engine component design. Proceedings of the
Second International Symposium on Ceramic Materials and Components for Engines, Bad Honnef, Germany, 1986, pp. 10231034.
3. JI Federer, TN Tiegs, DM Kotnick, D Petrak. Analysis of candidate silicon carbide
recuperator material exposed to industrial furnace environments. Report No.
ORNL/TM-9677, Oak Ridge National Laboratory, 1985.
4. MK Ferber, J Ogle, VJ Tennery, T Henson. Characterization of corrosion mechanisms occurring in a sintered SiC exposed to basic coal slags. J Am Ceram Soc
68:191197, 1985.
5. FA Costa Oliviera, RAH Edwards, RJ Fordham, JHW De Wit. Factors limiting the
application of silicon nitride ceramics in sulphur-containing environments of low
oxygen potential at high temperatures. In: K. G. Nickel, ed. Corrosion of Advanced
Ceramics Measurement and Modelling. Dordrecht: Kluwer, 1994, pp. 177188.
6. A Lacombe, C Bonnet. Ceramic matrix composites, key materials for future space
plane technologies. Proceedings of the AIAA Second International Aerospace
Planes Conference, Orlando, FL, 1990.
7. NS Jacobson, RA Rapp. Thermochemical degradation mechanisms for the reinforced carbon/carbon panels on the space shuttle. NASA Technical Memorandum
106793, 1995.
8. VE Chelnokov, AL Syrkin. High temperature electronics using SiC: Actual situation and unsolved problems. Mater Sci Eng B 46:248253, 1997.
9. PA Janeway. Aluminum nitride: Making the grade in demanding electronic applications. Ceram Ind 137:2832, 1991.
10. CH Henager Jr, RH Jones. Effects of an aggressive environment on the subcritical
crack growth of a continuous-fiber ceramic matrix composite. Ceram Eng Sci Proc
13:411419, 1992.
11. A Lepp, KA Schwetz, K Hunold. Hexagonal boron nitride: Fabrication, properties,
and applications. J Eur Ceram Soc 5:39, 1989.
12. R Sangiorgi. Corrosion of ceramics by liquid metals. In: K. G. Nickel, ed. Corrosion
of Advanced Ceramics Measurement and Modelling. Dordrecht: Kluwer, 1994, pp.
261284.
13. K Oda, T Yoshio, Y Miyamoto, M Koizumi. Hydrothermal corrosion of pure, hot
isostatically pressed silicon nitride. J Am Ceram Soc 76:13651368, 1993.
14. NJ Shaw, JA DiCarlo, NS Jacobson, SR Levine, JA Nesbitt, HB Probst, WA Sanders, CA Stearns. Materials for engine applications above 3000Fan overview.
NASA Technical Memorandum 100169, 1987.
15. MA Lamkin, FI Riley, RJ Fordham. Oxygen mobility in silicon dioxide and silicate
glasses: A review. J Eur Ceram Soc 10:347367, 1992.
16. FJ Norton. Permeation of gaseous oxygen through vitreous silica. Nature 171:701,
1961.

346

Jacobson and Opila

17. EJ Opila. Influence of alumina reaction tube impurities on the oxidation of chemically-vapor-deposited silicon carbide. J Am Ceram Soc 78:11071110, 1995.
18. BE Deal, AS Grove. General relationship for the thermal oxidation of silicon. J
Appl Phys 36:37703778, 1965.
19. JA Costello, RE Tressler. Oxidation kinetics of silicon carbide crystals and ceramics: I. In dry oxygen. J Am Ceram Soc 69:674681, 1986.
20. LUJT Ogbuji, EJ Opila. A comparison of the oxidation kinetics of SiC and Si3N4.
J Electrochem Soc 142:925930, 1995.
21. CE Ramberg, G Cruciani, KE Spear, RE Tressler, CF Ramberg Jr. Passive-oxidation kinetics of high-purity silicon carbide from 800 to 1100C. J Am Ceram Soc
79:28972911, 1996.
22. K Motzfeldt. On the rates of oxidation of silicon and silicon carbide in oxygen,
and correlation with permeability of silica glass. Acta Chem Scand 18:15961606,
1964.
23. Z Zheng, RE Tressler, KE Spear. Oxidation of single-crystal silicon carbide, Part
II. Kinetic model. J Electrochem Soc 137:28122816, 1990.
24. NS Jacobson. Corrosion of silicon-based ceramics in combustion environments. J
Am Ceram Soc 76:328, 1993.
25. KL Luthra. Some new perspectives on oxidation of silicon carbide and silicon nitride. J Am Ceram Soc 74:10951103, 1991.
26. Z Zheng, RE Tressler, KE Spear. Oxidation of single-crystal silicon carbide. Part
I. Experimental studies. J Electrochem Soc 137:854858, 1990.
27. WC Tripp, JW Hinze, MG Mendiratta, RH Huff, AF Hampton, JE Stroud, ET
Rodine. Internal structure and physical properties of ceramics at high temperatures.
Aerospace Research Laboratories Report No. 75-0130, 1975, pp. 116125.
28. CE Ramberg. PhD dissertation, University of Pennsylvania, Philadelphia, 1997.
29. LUJT Ogbuji. Effect of oxide devitrification on oxidation kinetics of SiC. J Am
Ceram Soc 80:15441550, 1997.
30. LUJT Ogbuji, DT Jayne. Mechanism of incipient oxidation of bulk chemical vapor
deposited Si3N4. J Electrochem Soc 140:759766, 1993.
31. LUJT Ogbuji, SR Bryan. The SiO2 Si3N4 interface. Part I: Nature of the interphase.
J Am Ceram Soc 78:12721278, 1995.
32. BW Sheldon. Silicon nitride oxidation based on oxynitride interlayers with graded
stoichiometry. J Am Ceram Soc 79:29932996, 1996.
33. H Du, RE Tressler, KE Spear, CG Pantano. Oxidation studies of crystalline CVD
silicon nitride. J Electrochem Soc 136:15271536, 1989.
34. D Cubicciotti, KH Lau. Kinetics of oxidation of hot-pressed silicon nitride containing magnesia. J Am Ceram Soc 61:512517, 1978.
35. D Cubicciotti, KH Lau. Kinetics of oxidation of yttria hot-pressed silicon nitride.
J Electrochem Soc 126:17231728, 1979.
36. JA Costello, RE Tressler, IST Tong. Boron redistribution in sintered -SiC during
thermal oxidation. J Am Ceram Soc 64:332335, 1981.
37. DM Mieskowski, WA Sanders. Oxidation of silicon nitride sintered with rare-earth
oxide additions. J Am Ceram Soc 68:C160C163, 1985.
38. P Andrews, FL Riley. The microstructure and composition of oxide films formed

Nonoxide Ceramics

39.
40.
41.

42.
43.
44.
45.
46.
47.
48.

49.
50.
51.
52.
53.
54.
55.
56.
57.

58.

347

during high-temperature oxidation of a sintered Si3N4. J Eur Ceram Soc 5:245


256, 1989.
SC Singhal, FF Lange. Effect of alumina content on the oxidation of hot-pressed
silicon carbide. J Am Ceram Soc 58:433435, 1975.
DS Fox. Oxidation behavior of chemically vapor-deposited silicon carbide and silicon nitride from 12001600C. J Am Ceram Soc 81:945950, 1998.
M Backhaus-Ricault, YG Gogotsi. Identification of oxidation mechanisms in silicon nitride ceramics by transmission electron microscopy studies of oxide scales.
J Mater Res 10:23062321, 1995.
P Andrews, FL Riley. Silicon nitride oxidation/reoxidation. J Eur Ceram Soc 7:
125132, 1991.
J Persson, P-O Kall, M Nygren. Parabolic non-parabolic oxidation kinetics of Si3N4.
J Eur Ceram Soc 12:177184, 1993.
JK Patel, DP Thompson. The low-temperature oxidation problem in yttria-densified
silicon nitride ceramics. Br Ceram Trans J 87:7073, 1988.
RB Sosman. The Properties of Silica. New York: The Chemical Catalog Company,
1927, p. 364.
EJ Opila, DS Fox, CA Barrett. Cyclic oxidation of monolithic SiC and Si3N4 materials. Ceram Eng Sci Proc 14(78):367374, 1993.
NS Jacobson, S Farmer, A Moore, H Sayir. High temperature oxidation of boron
nitride. Part IMonolithic boron nitride. J Am Ceram Soc 82:393398, 1999.
K Vedula, W Williams, AH Heuer, TE Mitchell, F Lisy, A Abada, DC Wang. Ultra
high-temperature ceramic-ceramic composites. Proceedings of Ultra-High Temperature Composite Materials Meeting, Dayton, OH, 1988.
A Bellosi, E Landi, A Tampieri. Oxidation behavior of aluminum nitride. J Mater
Res 8:565572, 1993.
D Robinson, R Dieckmann. Oxidation of aluminum nitride substrates. J Mater Sci
29:19491957, 1994.
T Sato, K Haryu, T Endo, M Shimada. High temperature oxidation of hot-pressed
aluminum nitride by water vapor. J Mater Sci 22:22702280, 1987.
DA Robinson, G Yin, R Dieckmann. Oxide film formation on aluminum nitride
substrates covered with thin aluminum layers. J Mater Sci 29:23892394, 1994.
H Kim, AJ Moorhead. Oxidation behavior and flexural strength of aluminum nitride
exposed to air at elevated temperatures. J Am Ceram Soc 77:10371041, 1994.
EJ Opila. Variation of the oxidation rate of SiC with water vapor pressure. J Am
Ceram Soc 82:625636, 1999.
EJ Opila. Oxidation kinetics of chemically vapor-deposited silicon carbide in wet
oxygen. J Am Ceram Soc 77:730736, 1994.
V Pareek, DA Shores. Oxidation of silicon carbide in environments containing potassium salt vapors. J Am Ceram Soc 74:556563, 1991.
RE Tressler, JA Costello, Z Zheng. Oxidation of silicon carbide ceramics. In: A.
J. Hayes, ed. Industrial Heat Exchangers. Warrendale, PA: American Society of
Metals, 1985, pp. 307314.
DJ Choi, DB Fischbach, WD Scott. Oxidation of chemically-vapor-deposited silicon nitride and single crystal silicon. J Am Ceram Soc 72:11181123, 1989.

348

Jacobson and Opila

59. JE Antill, JB Warburton. Oxidation of silicon and silicon carbide in gaseous atmospheres at 10001300C. In: PG Atkinson, DJ Dawton, O Lutz, R Monti, J Surugue, eds. Reactions between Solids and Gases, AGARD CP-52. Paris: Advisory
Group for Aerospace Research and Development, 1970, pp. 10-110-14.
60. G Hetherington, KH Jack, JC Kennedy. The viscosity of vitreous silica. Phys Chem
Glasses 5:130136, 1964.
61. A Hashimoto. The effect of H2O gas on volatilities of planet-forming major elements: I. Experimental determination of thermodynamic properties of Ca-, Al-, and
Si-hydroxide gas molecules and its application to the solar nebula. Geochim Cosmochim Acta 56:511532, 1992.
62. EJ Opila, DS Fox, NS Jacobson. Mass spectrometric identification of SiOH(g)
species from the reaction of silica with water vapor at atmospheric pressure. J Am
Ceram Soc 80:10091012 (1997).
63. EJ Opila, RE Hann Jr. Paralinear oxidation of CVD SiC in water vapor. J Am
Ceram Soc 80:197205, 1997.
64. AC Lea. The oxidation of silicon-carbide refractory materials. J Soc Glass Technol
33:2750, 1949.
65. E Fitzer, R Ebi. Kinetic studies on the oxidation of silicon carbide. In: RC Marshall,
JW Faust Jr., CE Ryan, eds. Silicon Carbide 1973. Columbia: University of South
Carolina Press, 1974, pp. 320328.
66. W Bremen, A Naoumidis, H Nickel. Oxidation behavior of pyrolytically deposited
-SiC in an atmosphere composed of COCO2 gas mixtures. J Nucl Mater 71:56
64, 1977.
67. EJ Opila, QN Nguyen. Oxidation of silicon carbide in carbon dioxide. J Am Ceram
Soc 81:19491952, 1998.
68. EJ Opila, NS Jacobson. SiO(g) formation from SiC in mixed oxidizingreducing
gases. Oxidation of Metals 44:527544, 1995.
69. DP Butt, RE Tressler, KE Spear. Durability of SiC materials in gaseous N2 H2
CO heat treat environments I. Ind Heat 58:4448, 1991.
70. C Wagner. Passivity during the oxidation of silicon at elevated temperatures. J Appl
Phys 29:12951298, 1958.
71. SC Singhal. Thermodynamic analysis of the high-temperature stability of silicon
nitride and silicon carbide. Ceram Int 2:123130, 1976.
72. HE Kim, DW Readey. Active oxidation of SiC in low dew-point hydrogen above
1400C. Proceedings of Silicon Carbide 87, Columbus, OH, 1989, pp. 301
312.
73. T Narushima, T Goto, Y Yokoyama, Y Iguchi, T Hirai. High-temperature active
oxidation of chemically vapor-deposited silicon carbide in COCO2 atmosphere.
J Am Ceram Soc 76:25212524, 1993.
74. T Narushima, T Goto, Y Iguchi, T Hirai. High-temperature active oxidation of
chemically vapor-deposited silicon carbide in an ArO2 atmosphere. J Am Ceram
Soc 74:25832586, 1991.
75. JE Marra, ER Kreidler, NS Jacobson, DS Fox. Reactions of silicon-based ceramics
in mixed oxidation chlorination environments. J Am Ceram Soc 71:10671073,
1988.
76. P Steinmetz, C Rapin. Corrosion of Metallic Materials in Waste Incinerators. In:

Nonoxide Ceramics

77.
78.

79.
80.

81.
82.

83.
84.

85.

86.
87.
88.
89.
90.

91.

92.

93.
94.

349

R Streiff, J Stringer, RC Krutenat, M Caillet, RA Rapp, eds. High Temperature


Corrosion and Protection of Materials, Zurich: Trans Tech, 1997, pp 505517.
Z Zheng, RE Tressler, KE Spear. The effect of sodium contamination on the oxidation of single-crystal silicon carbide. Corros Sci 33:545566, 1992.
Z Zheng, RE Tressler, KE Spear. A comparison of the oxidation of sodium-implanted CVD Si3N4 with the oxidation of sodium-implanted SiC-crystals. Corros
Sci 33:569580, 1992.
FS Pettit, CS Giggins. Hot corrosion. In: CT Sims, NS Stoloff, WC Hagel, eds.
Superalloys II. New York: Wiley, 1990, pp. 327354.
NS Jacobson, JL Smialek, DS Fox. Molten salt corrosion of SiC and Si3N4. In: NP
Cheremisinoff, ed. Handbook of Ceramics and Composites, Vol. 1, Synthesis and
Properties. New York: Marcel Dekker, 1990, pp. 99136.
NS Jacobson. Sodium sulfate: Deposition and dissolution of silica. Oxidation of
Metals 31:91103, 1989.
CA Stearns, FJ Kohl, DE Rosner. Combustion system processes leading to corrosive deposits. Proceedings of High Temperature Corrosion, NACE, San Diego,
1981, pp. 441450.
DS Fox, MD Cuy, TE Strangman. Sea salt hot corrosion and strength of a yttriacontaining silicon nitride. J Am Ceram Soc 80:27982804, 1997.
Y Dessureault, J Sangster, AD Pelton. Coupled phase diagram/thermodynamic
analysis of the nine common-ion binary system involving the carbonates and sulfates of lithium, sodium, and potassium. J Electrochem Soc 137:29412950, 1990.
MA Lamkin, FL Riley, RJ Fordham. Hot corrosion of silicon nitride in the presence
of sodium and vanadium compounds. Proceedings of High Temperature Corrosion
of Technical Ceramics, London, 1990, pp. 181191.
PF Becher. Strength degradation in SiC and Si3N4 ceramics by exposure to coal
slags at high temperatures. J Mater Sci 19:28052814, 1984.
NS Jacobson, KN Lee, DS Fox. Reactions of silicon carbide and silicon (IV) oxide
at elevated temperatures. J Am Ceram Soc 75:16031611, 1992.
KN Lee, RA Miller, NS Jacobson. New generation of plasma-sprayed mullite coatings on silicon carbide. J Am Ceram Soc 78:705710, 1995.
KN Lee, NS Jacobson, RA Miller. Refractory oxide coatings on SiC ceramics.
MRS Bull 19:3538, 1994.
N Boukis, N Claussen, K Ebert, R Janssen. Corrosion screening tests of high-performance ceramics in supercritical water containing oxygen and hydrochloric acid. J
Eur Ceram Soc 17:7176, 1997.
M Yoshimura, JI Kase, M Hayakawa, S Somiya. Oxidation mechanism of nitride
and carbide powders by high-temperature, high-pressure water. Ceram Trans 10:
337354, 1990.
NS Jacobson, YG Gogotsi, M Yoshimura. Thermodynamics and experimental
study of carbon formation on carbides under hydrothermal conditions. J Mater
Chem 5:595601, 1995.
YG Gogotsi, M Yoshimura. Formation of carbon films on carbides under hydrothermal conditions. Nature 367:628630, 1994.
T Yoshio, K Oda. Aqueous corrosion and pit formation of Si3N4 under hydrothermal
conditions. Ceram Trans 10:367386, 1990.

350

Jacobson and Opila

95. T Yoshio, K Oda. Durability of AlN ceramics in aqueous environment. Proceedings


of the Annual Meeting of the Ceramic Society of Japan, 1995, pp. 223228.
96. JW Cree, MF Amateau. Degradation of silicon carbide by molten lithium. J Am
Ceram Soc 70:C318C319, 1987.
97. U Schwabe, LR Wolff, FJJ van Loo, G Ziegler. Corrosion of technical ceramics
by molten aluminum. J Eur Ceram Soc 9:407415, 1992.
98. RE Tressler. Environmental effects on the long-term reliability of SiC and Si3N4
ceramics. Ceram Trans 10:99123, 1990.
99. KG Nickel, R Danzer, G Schneider, G Petzow. Corrosion and oxidation of advanced ceramics. Powder Metall Int 21:2934, 1989.
100. GA Gogotsi, YG Gogotsi, VP Zavada, VV Traskovsky. Stress corrosion of silicon
nitride based ceramics. Ceram Int 15:305310, 1989.
101. YG Gogotsi, GG Grathwohl. Stress-enhanced oxidation of silicon nitride ceramics.
J Am Ceram Soc 76:30933104, 1993.
102. TE Easler, RC Bradt, RE Tressler. Strength distributions of SiC ceramics after
oxidation and oxidation under load. J Am Ceram Soc 64:731734, 1981.
103. K Jakus, JE Ritter Jr, WP Rogers. Strength of hot-pressed silicon nitride after high
temperature exposure. J Am Ceram Soc 67:472475, 1984.
104. PF Becher. Strength retention in SiC ceramics after long-term oxidation. J Am
Ceram Soc 66:C120C121, 1983.
105. LJ Lindberg. Durability testing of ceramic materials for turbine engine applications.
Proceedings of the Twenty Fourth Automotive Development Contractors Coordination Meeting, Detroit, 1986, pp. 7986.
106. HS Rho, NL Hecht, GA Graves. The behavior of two Si3N4 and two SiC ceramics
subject to oxidation environments. Ceram Eng Sci Proc 18:633643, 1997.
107. M Kaji, T Ono, M Higashi, A Kokaji. Stress rupture behavior a silicon nitride under
combustion gas environment. Proceedings Yokohama International Gas Turbine
Congress, 1995, pp. 5966.
108. JL Smialek, NS Jacobson. Mechanism of strength degradation for hot corrosion of
-SiC. J Am Ceram Soc 69:741752, 1986.
109. JJ Swab, GL Leatherman. Strength of structural ceramics after exposure to sodium
sulfate. J Eur Ceram Soc 5:333340, 1989.
110. WJ Tomlinson, DM Caslin. Strength of SiC and SiC7.4 vol% TiB2 composite
after corrosion with Na2SO4 and NaCl deposits. Ceram Int 17:6166, 1991.
111. SM Wiederhorn. Moisture-assisted crack growth in ceramics. Int J Fract Mech 4:
171177, 1968.
112. CH Henager Jr, RH Jones. Molten salt corrosion of hot-pressed Si3N4 /SiC-reinforced composites and the effects of molten salt exposure on slow crack growth
of hot-pressed Si3N4. Ceram Trans 10:197210, 1990.

11
Oxide Ceramics
F. S. Pettit, G. H. Meier, and J. R. Blache`re
University of Pittsburgh, Pittsburgh, Pennsylvania

I.

INTRODUCTION

In discussing environmental effects on oxide ceramics, it is necessary to select


a manageable number of environments and oxide systems to illustrate most of
the salient points while also attempting to provide some systematic logic to the
degradation effects in general. An oxide ceramic is a ceramic that contains a
substantial amount of oxygen as a component. We can have binary oxides, ternary
oxides, as well as more complex oxides. Examples of nonoxide ceramics are
Si3N4 and SiC. In this chapter, we will consider environmental effects on binary
oxides, namely Al2O3, Cr2O3, and SiO2. Environments affect oxide ceramics when
the thermodynamic conditions are such that reactions occur between the environment and the ceramic. The extent of these reactions is determined by kinetic
processes, as well as by the driving force for reaction to occur. Therefore, the
particular way that the environmental conditions are imposed upon the ceramic
play an important role. For example, the forms of the environmental effects are
significantly different for oxides exposed to gas environments compared to liquid
environments. In this chapter, we will consider environmental effects induced by
gas environmentsin particular, O2, N2, and some gas mixtures such as O2
SO2, O2 CO2, and H2O O2. The liquid environments will be considered by
using molten salts and liquid metals. As discussed by McCauley (1), oxide ceramics can also be affected by solid environments. High-temperature acidbase reactions between solids such as SiO2 and MgO are major problems which can lead
to catastrophes by the reaction between refractories placed in contact. The low
melting salts formed in the reaction can lead to the melt down of structures.
351

352

Pettit et al.

Not placing acid and basic refractories in contact is a major requirement in the
design of furnaces such as arc furnaces for steel-melting and glass-melting furnaces. Although reactions between solid oxide ceramics are important, such reactions will not be considered specifically in this chapter. However, the processes
discussed in the liquid environment section are relevant to solid reactions. Finally,
it is important to emphasize that environmental effects on ceramics can extend
from room temperature to extremely high temperatures. In this chapter, a very
wide range of temperatures will be used to illustrate the various environmental
effects.

II. GASEOUS ENVIRONMENTS


A. Oxygen
Most oxide ceramics are rather insensitive to oxygen, especially at low temperatures. As the oxygen pressure is varied, oxide ceramics will adjust their stoichiometry or transform to higher or lower oxide phases at rates commensurate with
the temperature and the oxygen pressure differential. In the case of changes in
stoichiometry, even though the changes may be small, the effects on properties
may be substantial. This is shown in Fig. 1A, where -Al2O3 can be either an
n- or p-type electronic conductor, or an ionic conductor, depending on the oxygen
pressure and temperature (2). A KrogerVink diagram showing defect concentrations that could account for the observed conductivities at 1400C is also included
in Fig. 1B. Comparison of Figs. 1A and 1B show that at 1400C for oxygen
pressures of 1 atm, the principal point defects are cation vacancies and electron
holes, and Al2O3 is a p-type electronic conductor. As the oxygen pressure is decreased, the principal defects become aluminum ion interstitials and cation vacancies and the Al2O3 is an ionic conductor. At even lower oxygen pressures, Al2O3
becomes an n-type electronic conductor.
For some oxide ceramics, as the oxygen pressure is changed, another oxide
phase can be formed. For example, Cu2O could be converted to CuO. To determine if such phase changes may occur, standard free energies of formation can
be used to calculate the equilibrium pressures for the phases of concern. For
example, in the case of Cu2O and CuO, the equilibrium oxygen pressure can be
determined as follows:
Cu2O 1/2O2 2CuO
P O2 exp

(1)

2G CuO G Cu2O
2RT

P O2 104 atm at 1000 K

(2)

Oxide Ceramics

353

Fig. 1 Diagram showing the type of conduction predominating in -Al2O3 as a function


of temperature and oxygen pressure (A), and a KrogerVink diagram to account for these
conductivity regimes (B).

354

Pettit et al.

The effect of the oxygen pressure change depends on the function the oxide
ceramic may be serving. In the case of Cr2O3, which is often used to develop
oxidation resistance on nickel-, cobalt-, and iron-based alloys, the following reaction must be considered:
Cr2O3 3/2O2 2CrO3(g)

(3)

which gives
PCrO3 P O3/42 exp

G Cr2O3 2G CrO3
2RT

(4)

Equation (4) shows that the pressure P CrO3 increases with oxygen pressure. Therefore, the formation of CrO3 causes Cr2O3 to be a less effective protective oxide
barrier and this condition increased with oxygen pressure and temperature increases (G Cr2O3 2G CrO3 is negative). The vapor pressures of oxide phases are
always a factor that must be considered. The vapor pressure of Al2O3 is relatively
low and the pressures of other oxides involving aluminum and oxygen are also
low, and so, Al2O3 is an oxide ceramic that can be used at temperatures up to
13001400C with little effects of vaporization. On the other hand, MoO3 has
high vapor pressures and cannot be used at temperatures above 400500C. Gulbransen and Jansen (3) have proposed that vaporization of oxides begins to cause
problems at pressures greater than 109 atm. An interesting situation arises in the
case of SiO2, as has been discussed by Wagner (4). As indicated by the equations
SiO2 SiO(g) 1/2O2
P SiO

exp [(G SiO2 G SiO)/RT]


P O1/22

(5)
(6)

the pressure of SiO(g) increases as the oxygen pressure decreases. Therefore, at


low oxygen pressure, SiO formation can affect the oxidation of silicon, as well
as some nonoxide ceramics such as Si3N4.
When silica glass is exposed to elevated temperatures, it can devitrify or
change from the glass state to the crystalline state (5). Devitrification results in
significant changes in the properties of the silica. The cause of devitrification is
not completely understood. High temperatures are certainly a factor, but oxygen
in the gas, or water vapor, have been proposed to be also necessary.
B. Pure Gas Environments Other Than Oxygen
When oxide ceramics are exposed to gases other than oxygen, depending on the
conditions, the oxide ceramics can be changed to phases determined by the other

Oxide Ceramics

355

gas. For example, when the gas is pure nitrogen and the oxide ceramic is Al2O3,
the following reaction can be used to determine what will occur:
Al2O3 N2 2AlN 3/2O2
PN2
P O3/22

exp

(7)

2G AlN G Al2O3
RT

(8)

At 1500 K, PN2 /PO3/22 3 1030, and, therefore, if the P N2 /P O3/22 ratio in the gas is
greater than this value, the Al2O3 will be converted to AlN. Of course, the rate
at which this will occur will be determined by kinetics, which, in turn, is determined by factors that usually require some experimentation to understand. For
example, the AlN will form as a layer covering the Al2O3 and the reaction will
involve nitrogen diffusing inward and oxygen diffusing outward. If the oxygen
cannot be removed, pressures could develop that could rupture the AlN layer and
cause the rate of the reaction to be changed.
In principal, reactions, such as those described for nitrogen with Al2O3,
could occur for sulfur and carbon, as well as numerous other gases, but it is
important to emphasize that almost all metals have much greater affinities for
oxygen than for these other reactants, and, therefore, the oxide ceramics are comparatively stable. At 1500 K, the ratio PN2 /P O3/22 for equilibrium between Al2O3
and AlN is 3 1030. At a total pressure of 1 atm, if the nitrogen gas contains
even a small amount of oxygen, the oxide phase will remain stable.
C. Mixed Gases Containing Oxygen
When oxide ceramics are exposed to gases containing oxygen and another reactant, as discussed previously, the oxygen usually is not displaced unless the
oxygen pressure is extremely low. Even though the oxygen is not displaced from
the ceramic, the second reactant can affect the oxide ceramic by reactions of the
following type, where sulfur is the second gaseous reactant:
Al2O3 9/2O2 3/2S2 Al2O3 3SO3 Al2(SO4)3

(9)

The conditions for which such reactions can occur can be described by constructing thermodynamic diagrams such as the one presented in Fig. 2 for the
aluminumoxygensulfur system. Again, the rate of conversion of Al2O3 to
Al2(SO4)3 (Fig. 2) will be determined by kinetic processes. Reactions such as that
described by Eq. (9) can occur for carbon and nitrogen, as well as other gases.
Usually, most gas environments encountered in industrial practice are not appropriate for reactions of this type to take place. However, the stabilities of these
phases do vary. An interesting example is CaO or MgO, which is often present
in Al2O3 as impurities. The sulfates of these two oxides are more stable than

356

Pettit et al.

Fig. 2 Stability diagram for Al2O3 as a function of O2 and SO3 pressures at 700C where
the boundaries for the CaO/CaSO4 and MgO/MgSO4 are also included.

Al2(SO4)3, and these sulfates have been observed (5) to form on Al2O3 without
any Al2(SO4)3 formation (Fig. 3).
In some gas mixtures, the second reactant may not enter directly into a
chemical reaction with the oxide ceramic, but it still exerts a profound effect on
the mechanical properties of the ceramic. An excellent example is the effect of
water vapor on crack propagation in soda-lime silica glass as has been described
in articles by Weiderhorn (6,7).
The strength of glass is known to deteriorate in atmospheres containing
moisture. Under constant tensile load below the fracture strength at room temperature, the glass may eventually fail in a process called static fatigue. This delayed
fracture is associated with preferential attack by water of stressed SiEO bonds
at the crack tip. This is usually considered a stress corrosion associated with the
reaction between the glass and water, as indicated schematically in Fig. 4, in
which an oxygen bridge of the glass structure is broken. This was proposed by
Orowan and many others.
Wiederhorn found that the velocity of crack propagation in soda-lime silica
glass depended on the humidity according to three regimes as shown in Fig. 5.
Over a range of stresses, the velocity of crack propagation depended on the relative humidity as expected from Fig. 4. However, at higher stresses, the velocity
of the crack propagation was limited apparently by the rate of transport of the
water to the crack tip. This transport was usually assumed to be by surface diffu-

Oxide Ceramics

357

Fig. 3 Scanning micrograph of alumina exposed to 7 103 atm of SO3 in O2 at 700C


where Mg(SO4) and Ca(SO4) were detected but not Al2(SO4)3.

Fig. 4 Schematic representation of proposed reaction between water and a strained


SiEOESi bond at a crack tip. Step 1 involves adsorption of water to the SiEO bond
and step 2 shows the reaction involving simultaneous proton and electron transfer. In step
3, surface hydroxyl groups are formed.

358

Pettit et al.

Fig. 5 Dependence of crack velocity in silica on the applied force. The percent relative
humidity for each set of runs is given on the right-hand side. Roman numerals identify
the different regions of crack propagation. In region I, the crack growth rate is proposed
to be controlled by adsorption or reaction of water vapor at the crack tip. In region II,
the rate is supposedly controlled by diffusion of water vapor in the crack. Region III
involves a crack propagation mode not involving water vapor.

sion along the crack or by vapor transport. At higher stresses yet, the crack velocity does not depend on humidity.
Tomazoa (8) recently discussed that either crack initiation or crack growth
can dominate static fatigue. He emphasized the evidence for water penetration
into the glass. The water would diffuse into the glass as molecular water and
react with the glass structure as shown in Fig. 4.
The interaction of water with the strength of glass is well established and
is expected to affect the strength of many ceramics which contain a glassy phase.
Michalski and Freiman (9) proposed that such processes can take place in other
oxides and species containing polar molecules. It has been found that such an

Oxide Ceramics

359

effect can occur in Al2O3. Furthermore, similar processes have been proposed as
the cause of the excessive spalling of alumina scales from metallic alloys exposed
to gases containing water vapor (10).

III. LIQUID ENVIRONMENTS


In discussing the effects of liquid environments on oxide ceramics, a variety of
types of liquid must be considered, and the thickness of the liquid is also an
important factor affecting the types of interaction that can occur. For example,
in the case of thin liquid molten-salt deposits, the gas environment above the
liquid deposit can enter into the reaction scheme. In this chapter, the effects of
thin liquid molten-salt deposits will be considered first and then additional effects
arising from thin deposits of other types of liquids will be discussed. Finally, the
changes in these types of interactions as the liquid becomes very thick will be
examined.
A. Thin Liquid Deposits of Molten Salts
Hot corrosion (1113) is a process whereby liquid deposits, usually molten deposits such as sulfates and vanadates, cause corrosion of metals and ceramics.
In the case of the hot corrosion of oxide ceramics, three types of process must
be considered (14):
Reactions between the ceramic and the deposit
Diffusion of various constituents through the molten deposit
Reactions between the gas and the molten deposit
These three processes are coupled to each other, with the slowest determining
the overall reaction rate.
To comment on specific processes that may be occurring, it is necessary
to consider some specific oxides, liquid deposits, and gas compositions; but before doing that, some generalizations are worthwhile.

1. Theoretical Considerations
The most important reactions between oxide ceramics and molten deposits can
be viewed as involving oxide ions, where considering the metal M, oxide ions
can be donated to the melt,
MO2 M4 202

(10)

or can be taken from the melt


MO2 O2 MO32

(11)

360

Pettit et al.

Depending on the characteristics of the melt, the oxide ions may not exist
as major species but can be defined by reactions of the type
SO3 O2 SO42

(12)

2SO3 O2 S2O72

(13)

V2O5 O2 2VO3

(14)

In view of such reactions, at times it is useful to view the oxide ions as a basic
component in the melt where in the previous equations the acidic components
would be SO3 or V2O5.
Stability diagrams can be used to identify the reactions that may occur
between an oxide ceramic and a molten deposit. For example, in the case of a
Na2SO4 deposit, the diagram in Fig. 6 defines the stability range of Na2SO4 at a
fixed temperature in terms of oxygen and SO3 pressures. The SO3 pressure is the
acidic component in the melt and is related to the basic component, Na2O, or
O2, via the equation
Na2O SO3 Na2SO4

(15)

Fig. 6 Stability diagram for the NaOS system at 1000C; the dashed lines are sulfur
isobars.

Oxide Ceramics

361

where for pure Na2SO4


K 15

1
a Na2OP SO3

(16)

The basic component can also be considered to be oxide ions because


SO42 SO3 O2

(17)

The stability diagram for the oxide ceramic can now be superimposed
upon the stability diagram for Na2SO4. In Fig. 7, the stability diagram for the
AlOS system has been superimposed on Fig. 6. Inspection of Fig. 7 shows
that depending on the composition of the Na2SO4, aluminum may be present as
NaAlO2 (AlO2 ions), Al2O3, Al2(SO4)3 (Al3 ions), or Al2S3. It is important to
note that these are the phases of aluminum which exist at unit activity. Hence,
in the region of Fig. 7 where Al2O3 is stable, an important equilibrium is
Al2O3 2Al3 3O2

(18)

Fig. 7 Stability diagram showing the phases of aluminum that are stable in Na2SO4 at
1000C; the dashed lines indicate sulfur isobars and the arrows indicate compositional
changes of Na2SO4.

362

Pettit et al.

where K 18 a 2Al3a 3O2 and, therefore, as the activity of oxide ions decreases, the
activity of Al3 ions increases until a value of unity is reached at the Al2O3 Al3
boundary. Another reaction of importance is
Al2O3 O2 2AlO2

(19)

and the activity of AlO2 ions increases as the oxide ion activity increases until
a value of unity is reached at the AlO2 Al2O3 boundary.
Rapp (13) has experimentally determined the solubilities of a number of
oxides in Na2SO4, and some of the results from these studies are presented in
Fig. 8. The slopes of these solubility curves can be used to verify the reactions
by which the oxide dissolves. For example, in the case of Al2O3, basic reactions
can be written as
Al2O3 Na2SO4 2NaAlO2 SO3

(20)

Al2O3 O2 2AlO2

(21)

or
which both yield slopes of 1/2. The acidic dissolution reactions can be written
as
Al2O3 3Na2SO4 Al2(SO4)3 3Na2O

(22)

Fig. 8 Solubilities of some oxides in Na2SO4 at 1200 K as determined by Rapp (13).


Dissolution reactions consistent with the solubility curves for Al2O3 are included on this
diagram. The dashed lines give calculated solubilities for Al2O3 obtained by using Temkins rule for ionic melts and the indicated reactions for Al2O3.

Oxide Ceramics

363

or
Al2O3 2Al3 3O2

(23)

which both have slopes of 3/2.


When considering more complex melts, such as solutions of Na2SO4
NaVO3, it is necessary to relate the two acidic components, SO3 and V2O5. This
can be done by using equations such as
Na2SO4 V2O5 2NaVO3 SO3

(24)

where
K 24

(1 X Na2SO4)2PSO3
X Na2SO4a V2O3

(25)

and K 24 is the equilibrium constant, XNa2SO4 is the mole fraction of Na2SO4 in the
Na2SO4 NaVO3 solution, which is assumed to be ideal, PSO3 is the pressure of
SO3 and aV2O5 is the activity of V2O5 in equilibrium with the melt. In order to use
Eqs. (24) and (25), it is necessary to determine the major components in the melts
of interest. For example, the vanadium may exist as V2O5, NaVO3, Na4V2O7, or
Na3VO4. Equilibrium conditions for reactions such as
V2O5 4Na3VO4 3Na4V2O7

(26)

V2O5 Na4V2O7 4NaVO3

(27)

2NaVO3 Na2O V2O5

(28)

can be used to determine the proportions of the important species. Rapp (13)
has developed diagrams for solutions of Na2SO4 NaVO3 and typical results are
presented in Fig. 9 for a temperature of 900C. These results show that for SO3
pressures between 102 and 107 at 900C, Na2SO4 NaVO3 melts contain predominantly NaVO3. The amount of NaVO3 in the melt versus Na3VO4, Na3V2O7,
or V2O5 also depends on temperature. When Na2SO4 and NaVO3 are the major
species in Na2SO4 NaVO3 melts, Eq. (25) can be used to relate the activities of
the two acidic species, SO3 and V2O5.
The solubilities of oxides in Na2SO4 NaVO3 melts are influenced by both
the basic and acidic components. In Fig. 10, the solubilities of some oxides in
Na2SO4 and in a Na2SO4 30 mol% NaVO3 melt are compared using results obtained by Hwang and Rapp (15). Because the basic component is the same in
both melts, the solubility curves are identical for both melts in regions where
dissolution involves the basic component. The differences arise when the dissolution reactions involve acidic species with higher solubilities occurring for V2O5
than SO3.

364

Pettit et al.

Fig. 9 Mole fractions of V2O5, NaVO3, Na2V2O7, and NaVO3 in a Na2SO4 20 mol%
NaVO3 melt at 900C as a function of SO3 pressure.

Another point to be considered in molten deposits on alloys and ceramics


involves the ionic species that are responsible for transport. In the case of Na2SO4
deposits, Na and SO42 ions predominate, but it has been shown (11) that the
transport of SO3 occurs via S2O72 ions. In Na2SO4 NaVO3 melts, the preceding
ions are also important, but in view of the results presented in Fig. 9, VO3,
V2O73, and VO43 ions are important, with VO 3 ions having larger concentrations
than V2O74 or VO43 in the melts exposed to SO3 pressures in the range between
102 and 107.
Two final points worth noting in the solubility curves of oxides in molten
deposits are dissolutionreprecipitation processes (16) and synergistic dissolution
(17). Dissolutionreprecipitation occurs when the gradient of the oxide solubility
is negative at the oxidemelt interface. This means that the oxide solubility is

Oxide Ceramics

365

Fig. 10 Solubilities of Al2O3 in pure Na2SO4 and in Na2SO4 30 mol% NaVO3 at 900C
as determined by Hwang and Rapp (15).

greatest at the oxidemelt interface. The oxide can, therefore, dissolve but precipitate out in the melt where the solubility is lower. Synergistic dissolution occurs
when two oxides react with melts in ways that the dissolution of one accelerates
the dissolution of the other. For example, if the solubility curves for Al2O3 and
Fe2O3 in Fig. 8 are considered at a value of log aNa2O 14, the dissolution of
Al2O3 occurs via
Al2O3 O2 2AlO2

(29)

As this reaction proceeds, oxide ions are consumed, and the melt becomes more
acidic. The Fe2O3 dissolution involves the reaction
/ Fe2O3 2/3Fe3 O2

13

(30)

where oxide ions are donated. The sum of these two reactions is
/3Fe2O3 Al2O3 2AlO2 2/3Fe3

(31)

The combined dissolution process of both oxides is such that the melt composition does not change with respect to oxide ion concentration and, hence, both
reactions proceed more rapidly when occurring concomitantly.

2. Alumina Corrosion
The corrosion of Al2O3 when exposed to deposits of molten salts is dependent
on purity (18,19). High-purity (99.9% pure) -Al2O3 single crystals lost weight

366

Pettit et al.

as a function of time when exposed to deposits of Na2SO4 and Na2SO4 NaVO3


in gas mixtures of oxygen and SO3 at temperatures of 700C and 900C. In the
sulfate melts, the weight-loss rates decreased as the SO3 pressure was increased,
which shows that the dissolution process is probably
Al2O3 SO42 2AlO2 SO3

(32)

The alumina dissolves by basic dissolution. The SO3 pressure is greater at the
Al2O3 interface than the gas interface and SO3 is lost from the deposit. As dissolution of the alumina continues, the increased concentration of AlO2 ions causes
the SO3 at the Al2O3 interface to approach that of the gas interface, and, eventually, the dissolution process will stop unless fresh Na2SO4 is applied.
The weight-loss rates of Al2O3 increased as the activity of V2O5 was increased in Na2SO4 NaVO3 deposits. The dissolution reaction is proposed to be
Al2O3 3VO3 2Al3 3VO43

(33)

with the following reaction occurring at the meltgas interface:


VO43 SO3 VO3 SO42

(34)

This observed dissolution of Al2O3 in the Na2SO4 NaVO3 melt corresponds to


melt compositions on the acidic dissolution side of the solubility curve in Fig.
10, with the melt at the gas interface being more acidic than the melt at the Al2O3
interface.
Impurities can play a significant role in the hot corrosion of Al2O3. A variety
of conditions can occur depending on the type of impurity, the concentration,
the melt composition, the gas composition, and the temperature. In the case of
polycrystalline -Al2O3 of 99.8% purity which contained SiO2, MgO, CaO and
Na2O as impurities, a hot corrosion attack caused a porous zone to be developed
(Fig. 11). The thicknesses of these zones conformed to the parabolic rate law
and rate constants were dependent on the activity of V2O5 in the molten deposits
(Fig. 12) with little effect of temperature. These polycrystalline specimens were
attacked more rapidly than the high-purity single crystals due to the impurities
present in the former specimens. It is also important to note that for activities of
V2O5 above 105, transport through the melt via VO3 ions determines the rates.
However, at lower values, the rates are controlled by transport via S2O72 ions.
A good example of impurity effects has been observed (18) during the hot
corrosion of Al2O3 induced by Na2SO4 in air at 700C and 1000C. Under these
conditions, no attack of the high-purity single crystals was observed. Highpurity polycrystalline Al2O3 (0.1% MgO, 0.1% SiO2), however, developed features showing that some attack had occurred. At 700C, some preferential dissolution of the grains was evident and some silicon-rich needles had developed on
the surface of the Al2O3 (Fig. 13a). At 1000C, a consistent pattern of sodium
aluminum silicate along grain boundaries and sodium magnesium aluminum sili-

Oxide Ceramics

367

Fig. 11 Cross section of polycrystalline alumina specimen exposed to hot corrosion


conditions at 700C that caused a porous zone to be developed.

cate at triple points was evident (Fig. 13b). These results show that the Na2O
activity in these melts is high enough to promote reaction with silicates present
at grain boundaries of the Al2O3. As Na2O is removed from the deposit, the SO3
pressure is increased to levels at which Al2O3 can dissolve by acidic dissolution.
Synergistic dissolution has occurred.

3. Silica Corrosion
The solubility for vitreous silica in acidic melts is extremely low (20), but silica
does react with basic melts to form Na2SiO3 (21) and there are significant solubilities (22). Furthermore, the silica devitrifies under certain conditions and sodium
in melts accelerates this devitrification (23). In view of these conditions, silica
is very resistant to acidic melts but rather susceptible to hot corrosion by basic
melts. Typical reaction products developed on silica with a Na2SO4 deposit in
air for different times at 1000C are presented in Figs. 14 and 15. After 1 h, the
Na2SO4 wetted the silica and a layer of cristobalite spherulites formed under the

368

Pettit et al.

Fig. 12 Parabolic rate constants for growth of porous zones on polycrystalline alumina
as a function of activity of V2O5 in the deposit. Horizontal lines indicate the rate constants
for sulfate deposits.

salt. After 24 h, nothing remained of the spherulitic surface morphology because


tridymite formed at the silicasalt interface. Cristobalite separated the tridymite
from the vitreous silica. After 10 h, a thin sodium silicate layer over tridymite
and cristobalite layers was evident (Fig. 15). The crystalline layers spalled extensively on cooling from the test temperature as anticipated from their phase transitions.
The major reactions between Na2SO4 deposits and vitreous silica consist
of formation and dissolution of silicate via
SiO2 Na2O Na2SiO3

(35)

and diffusion of sodium into vitreous silica. In some deposits, the SiO32 ions can
convert to silica out in the melt when the acidity is sufficiently high. Sodium
does diffuse into fused silica and may be incorporated in the glass as a network
modifier (24) via the reaction
Na2O SiEOESi 2SiEO 2Na

(36)

The penetration of Na2O is expected to increase rapidly with the activity


of Na2O in the melt because the driving force is increased. Initially, the sodium
must enter the glass via reaction (36), which may be rate controlling, but the
situation is modified by crystallization of the silica glass. Diffusion of sodium
through cristobalite is very difficult, because it does not contain the channels of
vitreous silica.

Oxide Ceramics

369

Fig. 13 High-purity polycrystalline alumina exposed in basic conditions at (a) 700C


for 24 h and (b) at 1000C for 405 h. (a) Silica-rich needles and (b) silicate reaction
products form at grain boundaries on washed substrates.

370

Pettit et al.

Fig. 14 Fused silica exposed to basic conditions at 1000C for 1 h developed globular
arrays of spherulites.

B. Molten Metals
The attack of oxide ceramics by molten metals has been discussed by McCauley
(1). This type of attack is encountered in the manufacture of steel and nonferrous
metals such as aluminum and copper. A primary cause of the corrosion process
is reaction of the molten metals with oxygen in the oxide ceramic. Consequently,
the thermodynamic stability of the oxide ceramic compared to that of the metal
oxide is an important factor. In the case of alumina exposed to some metal M,
the important reaction is
Al2O3 M (molten metal) 3MO 2Al (molten metal)

(37)

with the equilibrium conditions


a 2Al
aM

eq

exp [(3G MO G Al2O3)]


RT

(38)

where it is assumed that there is no solubility of MO in Al2O3 and that no ternary


oxides (e.g., MAl2O4) are formed. If the standard free energy of MO per gramatom of oxygen is large compared to that for alumina, the right-hand side of Eq.

Oxide Ceramics

371

Fig. 15 Cross section of fused silica exposed to basic conditions at 1000C for 100 h
with thermal cycling; three layers cover the smooth glass: sodium silicates, tridymite, and
cristobalite, which is in contact with the glass.

(38) can be a large positive number and so conditions will be favorable for Eq.
(37) to proceed to the right. Even when MO is less stable than alumina, in principle some aluminum must be in the molten metal, and so some reaction can occur.
The extent of oxide ceramicmetal reactions is determined by the distribution
of the reaction product MO. When MO forms as a continuous layer over the
alumina, the reaction rate will be controlled by transport through MO and can
be expected to be small. However, if the MO is discontinuous or as particles in
the molten metal, then reaction rates can be rapid.
C. Corrosion in Thick Melts
Corrosion of oxide ceramics in deep melts involves many of the same reactions
that have been discussed previously, but the gas phase usually does not play a
significant role. In cases where a component in the gas phase is involved in the
corrosion process, this component can be substantially depleted from the liquid
phase, and, consequently, the corrosive characteristics of the liquid can change
substantially. Such conditions usually are established during the corrosion of metals and alloys in thick melts where, for example, oxygen is depleted from molten
salts. To illustrate the effects of corrosion of oxide ceramics, only examples where
the gas phase does not play a significant role will be considered. Many of the

372

Pettit et al.

Fig. 16 Schematic to show the products developed during exposure of sapphire in CaO
MgOAl2O3 SiO2 slags.

factors that affect the corrosion or dissolution of oxides in liquids have been
described by Cooper and Kingery (25) in studies concerned with the dissolution
of sapphire in CaOAl2O3 SiO2 slags, where it was shown that rates were controlled by mass transport in the liquid. The dissolution of sapphire in this slag
was controlled by mass transport in the liquid even when transport rates were
increased by rotating the immersed specimens. These investigators also analyzed
the effects of molecular diffusion, natural convection, and forced convection with
self-consistent results. Sandage and Yurek (26,27) studied the corrosion of sapphire and (Al, Cr)2O3 in CaOMgOAl2O3 SiO2 slags and showed that a solid
spinel reaction product (MgAl2O4) could be formed upon the surface of the sapphire depending on the experimental conditions. At times, the spinel did not form
a continuous layer, but it did affect the dissolution of the sapphire. It was shown
that in cases where the spinel formed as a continuous layer upon the sapphire,
a steady-state condition was achieved for which a constant spinel thickness was
established. The proposed mechanism for this process is shown schematically in
Fig. 16. The net effect of this process is to transfer Al2O3 from the sapphire
to the melt, but because the dissolution process involves transport through an
intermediate phase, it is called indirect dissolution.

IV. CONCLUDING REMARKS


The effects of various environmental conditions on oxide ceramics are not remarkably different from nonoxide ceramic or metallic alloys. The differences are
in the magnitude of the effects, due to the stability of oxide ceramics compared
to nonoxide ceramics and especially metallic alloys in the environments that are
usually encountered in practice.

Oxide Ceramics

373

REFERENCES
1. RA McCauly. Corrosion of Ceramics. New York: Marcel Dekker, 1995, pp. 194
197.
2. K Kitazawa, RL Coble. Electrical conduction in single-crystal and polycrystalline
Al2O3 at high temperatures. J Am Ceram Soc 57:245250, 1974.
3. EA Gulbransen, SA Jansen. Thermochemistry of gasmatal reactions. In: DL Douglass, ed. Oxidation of Metals. Materials Park, OH: American Society for Metals,
1971, pp. 6386.
4. C Wagner. Passivity during the oxidation of silicon at elevated temperatures. J Appl
Phys 29:1295, 1958.
5. HR Kim. Gaseous hot corrosion of oxide ceramics. MS dissertation, University of
Pittsburgh, Pittsburgh, PA, 1983.
6. SM Wiederhorn. Moisture assisted crack growth in ceramics. Fracture Mech 4:171
177, 1968.
7. SM Wiederhorn. Influence of water vapor on crack propagation in soda-lime glass.
J Am Ceram Soc 50:407444, 1967.
8. M Tomozawa. Fracture of glasses. Annu Rev Mater Sci 12:4374, 1996.
9. TA Michalski, SW Freiman. A molecular mechanism for stress corrosion in vitreous
silica. J Am Ceram Soc 66:284288, 1983.
10. RJ Janakiraman, GH Meier, FS Pettit. The effect of water vapor on the oxidation
of alloys that develop alumina scales for protection. Met and Mat Trans 30A:2905
2913, 1999.
11. KL Luthra. Low temperature hot corrosion of cobalt-base alloys: Part I. Morphology
of the reaction product. Part II. Reaction mechanism. Met Trans A 13A:18431864,
1982.
12. FS Pettit, CS Giggins. Hot corrosion. In: C Sims, N Stoloff, W Hagel, eds. Superalloys II. New York: John Wiley & Sons, 1987, pp. 327358.
13. RA Rapp. Hot corrosion of materials. In: O Johannesen, A Andersen, eds. Selected
Topics in High Temperature Chemistry. New York: Elsevier, 1989, pp. 291329.
14. FS Pettit, GH Meier, JR Blachere. Hot corrosion of oxide ceramics. In: KG Nickel,
ed. Corrosion of Advanced Ceramics. Dordrecht, The Netherlands: Kluwer, 1994,
pp. 235248.
15. YS Hwang, RA Rapp. Thermochemistry and solubilities of oxides in sodium sulfatevanadate solutions. Corrosion 45:933937, 1989.
16. RA Rapp, KS Goto. The hot corrosion of metals by molten salts. In: J Braunstein,
JR Selman, eds. Proceedings of the Electrochemical Society Symposium on Molten
Salts. The Electrochemical Society, Pennington, NJ, 1981, pp. 159177.
17. YS Hwang, RA Rapp. Synergistic dissolution of oxides in molten salts. In: T
Grobstein, J Doychak, eds. Oxidation of High Temperature Intermetallics. Warrendale, PA: TMS, pp. 257270.
18. MG Lawson, FS Pettit, JR Blachere. Hot corrosion of alumina. J Mater Res 8:1964
1971, 1993.
19. BM Warnes. The influences of vanadium on the sulfate induced hot corrosion of
thermal barrier coating materials. PhD dissertation, University of Pittsburgh, Pittsburgh, PA, 1990.

374

Pettit et al.

20. DZ Shi, RA Rapp. J Electrochem Soc 133:849, 1986.


21. NS Jacobson. Sodium sulfate: Deposition and dissolution of silica. Oxide Metals
31:91103, 1989.
22. GM Kim. The effects of contaminants and solubilities of SiO2 in fused Na2SO4 at
1200 K, MS dissertation, University of Pittsburgh, Pittsburgh, PA, 1982.
23. MG Lawson, HR Kim, FS Pettit, JR Blachere. Hot corrosion of silica. J Am Ceram
Soc 73:989995, 1990.
24. WD Kingery, HK Bowen, DR Uhlman. Introduction to Ceramics, 2nd ed. New York:
John Wiley & Son, 1976, p. 103.
25. AR Cooper Jr, WD Kingery. Dissolution in ceramic systems: I. Molecular diffusion,
natural convection, and forced convection studies of sapphire dissolution in calcium
aluminum silicate. J Am Ceram Soc 47:3743, 1964.
26. KH Sandage, GJ Yurek. Indirect dissolution of sapphire into silicate melts. J Am
Ceram Soc 71:476489, 1988.
27. KH Sandage, GJ Yurek. Indirect dissolution of (Al, Cr)2O3 in CaOMgOAl2O3
SiO2 (CMAS) melts. J Am Ceram Soc 74:19411954, 1991.

12
Metal Matrix Composites
Russell H. Jones
Pacific Northwest National Laboratory, Richland, Washington

I.

INTRODUCTION

Composite reinforcements are added to metals to increase their strength, elastic


modulus, wear performance, thermal conductivity or to alter their thermal expansion properties. Applications for metal matrix composites (MMCs) include aerospace, automotive, military hardware, and so forth. The high specific properties
are an obvious advantage to aerospace, but they are also very attractive for helping the automotive industry achieve its goals to build lighter-weight automobiles.
Antenna, aircraft support structures, vertical tail fins, inertial guidance, and precision optical systems are some of the aerospace applications for MMCs as well
as the cargo bay stiffeners on the space shuttle.
Both continuous and discontinuous reinforcements are added to MMCs.
Continuous reinforcements are usually graphite or boron fibers, whereas discontinuous reinforcements are usually SiC or Al2O3 particles or, in a few cases,
whiskers. Graphite and boron fibers can induce galvanic corrosion, whereas SiC
and Al2O3 particles are nonconducting and therefore will not induce galvanic
corrosion. Clusters of particles can induce pitting corrosion, whereas both fibers
and particles can induce localized corrosion. There is a considerable amount of
corrosion data on aluminum matrix MMCs reinforced with graphite and boron
fibers and SiC and Al2O3 particles with a lesser amount of data on magnesium
and titanium MMCs. Hihara and Latanision (1) have summarized the corrosion
behavior of metal matrix composites. However, there is relatively little data on
the stress corrosion and corrosion fatigue performance of MMCs.

375

376

Jones

II. CORROSION BEHAVIOR


A. Discontinuous Reinforcement

1. Aluminum Matrix Composites


The addition of a nonconducting reinforcing phases such as SiC or Al2O3 particles
do not induce galvanic currents but can induce pitting corrosion. Trzaskoma et al.
(2) and Trzaskoma (3) measured the pitting behavior of aluminum alloys 2024,
6061, and 5456 reinforced with 20 vol% SiC whiskers with 0.51.0-m diameters
and lengths up to 50 m. Pitting potentials were determined electrochemically
in 0.1N and 0.6N NaCl with and without dissolved oxygen. The pit initiation is
unaffected for the 6061 and 5456 Al alloys by the presence of the SiC whiskers,
and the 2024 Al alloy was affected by their presence. SiC whiskers caused a
100-mV shift in the pitting potential of the 2024 Al alloy in the de-aerated 0.1N
NaCl solution. The pit morphology of the 6061 Al alloy was changed from large
irregularly shaped pits in the unreinforced material to shallower round pits in the
SiC reinforced material. Therefore, although SiC whiskers did not affect the pitting potential in the 6061 Al alloy, it did alter the pitting morphology. The relatively small 100-mV shift in the pitting potential of the 2024 Al alloy is not too
significant because this alloy has a higher potential potential than the 6061 and
5456 Al alloys. Shimizu et al. (4) also found that pit initiation was unaffected
by the presence of 10 vol% SiC whiskers, but that once a pit was initiated, the
possibility for accelerated corrosion at the crevices between the reinforcement
and the matrix existed. The electrochemical polarization curves for both 10 and
20 vol% SiC whisker reinforced 6061 Al in 3.5% NaCl at 25C showed no shift
in the open-circuit potential but an increased anodic current density. In contrast,
SiC whisker reinforcement had no affect on the anodic polarization of 7075 alloy
tested in the same solutions; however, in both alloys, the cathodic current density
was greater for the reinforced material relative to the unreinforced. The authors
suggested that this increased cathodic current could be the result of an interfacial
layer between the matrix and SiC whiskers.
Paciej and Agarwala (5) also examined the corrosion behavior of a SiCwhisker-reinforced (20 vol%) aluminum alloy. Their matrix alloy was AA 7090,
which is an age-hardenable alloy with high Zn (6%) with Mg and Cu and is
therefore a different matrix composition than the 2024, 5456, and 6061 Al alloys
studied by Trzaskoma and co-workers (2,3). Paciej and Agarwala (5) also studied
the effect of heat treatment on the corrosion behavior of their composite material
that was a powder metallurgy alloy MA-87 of the same composition as the AA
7091 for comparison. A primary conclusion of their study was that the skin of
the processed plate exhibited different corrosion behavior than the core in 3.5%
NaCl. This difference was suggested as being due to Fe enrichment and increased
porosity in the skin and Cu enrichment in the core. The steady-state corrosion

Metal Matrix Composites

377

potential of the core region was about 100 mV anodic as compared to the MA87 material. Paciej and Agarwala (6) found that AA 7091 reinforced with particulate SiC (20 vol%) did not exhibit the same behavior as the whisker-reinforced
material. Clusters of particles were observed to serve as regions for localized
corrosion, but there was no difference between the skin and the core. The pitting
potential in 3.5% NaCl ranged from 100 mV to no difference for the composite
material relative to the MA-87, depending on the heat treatment. The small 100mV shift is similar to that observed by Trzaskoma and co-workers (2,3) for SiCwhisker-reinforced AA 2024.
The effects of variable volume fraction (1540%) of SiC particles on the
corrosion behavior of AA 6061 was studied by Sun et al. (7). Studies were conducted in aerated and de-aerated solutions of NaCl. These authors found no shift
in the corrosion potential with respect to the volume fraction of SiC, but they
did see increased corrosion rates and less stable films with increasing volume
fraction of SiC. A significant density of small pits was observed for NaCl concentrations ranging from 0.035% to 3.5% for potentials beyond the pitting potential,
and these pits led to exfoliation-type cracking.
Nunes and Ramanathan (8) compared the corrosion behavior of SiC (5 and
10 vol%) and Al2O3 (5 and 20 vol%) particle-reinforced aluminum matrix material. The matrix alloys were Al7.5%, Si1% Mg, and AA 2014 and the composite was produced by a melt-stirring process rather than a powder process as that
reported by Trzaskoma and co-workers (2,3), Paciej and Agarwala (5,6), and Sun
et al. (7). The matrix composition of the Nunes and Ramanathan (8) material is
also different from the previous studies, with the exception of the AA 2024 reported by Trzaskoma and co-workers. Both types of reinforcement caused increased corrosion during immersion testing in 3.5% NaCl. The increased corrosion rate was associated with pits or microcrevices near the particlematrix
interface and from particle dropout. Pits in the SiC-reinforced material were
deeper than in the Al2O3-reinforced composite. The pitting potential of the composite materials was shifted to lower potentials compared to the alloy, with this
shift ranging from 95 mV in the aerated solution to 119 mV in the de-aerated
solution. These shifts were similar for both particles, although there is some uncertainty in the data, and were similar in magnitude to that observed by
Trzaskoma and co-workers (2,3) and Paciej and Agarwala (5,6).
Further information on the effect of SiC particles on the interfacial corrosion of an Al 2024 alloy in a NaCl solution was presented by Yao and Zhu (9).
They noted that the interfacial preferential dissolution (IPD) zone size was similar
to the plastic accommodation size around the particles. They concluded from this
that IPD was caused by the low integrity of the particle surface oxide film and
not by chemical, metallurgical, or galvanic coupling effects.
The corrosion characteristics of a cast AlSi alloy containing 3 wt% graphite was evaluated by Saxena et al. (10). They measured the weight change follow-

378

Jones

ing immersion in 3.5% NaCl and a salt spray and noted that the composite showed
a factor of 16 for the immersion test and 6 for the salt spray test at 25C. Potentiodynamic polarization tests in 3.5% NaCl showed considerably higher anodic and
cathodic current densities for the composite relative to the base material. The
authors concluded that the galvanic coupling of the graphite to the aluminum
could explain the increased corrosion rates of the composite relative to the matrix
material. They also observed localized corrosion at the interface between the
graphite particles and the aluminum matrix as observed for composites with SiC
and Al2O3 reinforcement.
In summary, the primary affect of ceramic-phase reinforcement of aluminum alloys on corrosion is that of providing sites for pitting and crevice corrosion.
This affect is manifested in decreased pitting potentials and increased corrosion
rates from localized corrosion at particlematrix interfaces and microcavities.
Graphite particles have a much larger affect on the corrosion rate of aluminum
composites than SiC or Al2O3 particle reinforcement because of the galvanic coupling and electrical conductivity of graphite.

2. Magnesium Matrix Composites


Magnesium matrix composites offer attractive properties for aerospace and automotive applications because of their low density and excellent specific properties.
Increases in the elastic modulus is one rationale for adding ceramic-phase reinforcements to magnesium. Magnesium has been reinforced with SiC (11,12) particles and Al2O3 (12) fibers, and because of the stability of SiC in magnesium,
composites can be produced by conventional foundry processes. Luo (11) found
good wetting between Mg and SiC, with no evidence of reaction at the Mg
SiC interface following liquid mixing and casting of the composite for the pure
magnesium matrix composite. However, a reaction product of Mg2Si was present
at the MgSiC interface in the AZ91 alloy matrix composite. The composite
showed a 56% increase in the yield strength, but a decrease in toughness relative
to the AZ91 alloy without reinforcement.
Nunez-Lopez et al. (12) studied the corrosion behavior of a MgZnCu
(ZC71) alloy reinforced with 12 vol% SiC particles. The ZC71 alloy is not as
corrosion resistant as the AZ91 series, so this composite would require a coating,
as would the unreinforced matrix material. The composite was made by blending
the SiC particles with the molten metal and then cast. Salt spray and electrochemical measurements were made in a 3.5% NaCl solution at 25C. The salt spray
corrosion rates for the composite material in the as-extruded and T6 condition
were very similar to that of the unreinforced material in the same heat-treatment
condition. However, localized corrosion is dominant in both reinforced and unreinforced material, and the local penetration rate for the reinforced material is
about three times faster than the unreinforced material. This alloy develops a

Metal Matrix Composites

379

passive film at cathodic potentials in both the reinforced and unreinforced condition and there is no evidence for galvanic corrosion between the matrix and SiC
particles. The electrochemical polarization response of the reinforced and unreinforced material were nearly identical.
Reinforcement of AZ91, the most corrosion-resistant magnesium alloy,
with Al2O3 (13) fibers produced a shift in the electrochemical polarization response for high-pH (10.5) solutions. There was a shift of about 150 mV of the
open-circuit potential in the anodic direction for the reinforced material relative
to the unreinforced material with and without NaCl addition (3.5%). The shift
in the open-circuit potential with Cl concentration was similar for both reinforced and unreinforced material, whereas the corrosion rates were similar up to
about 1% NaCl. The corrosion rate of the reinforced material increased much
more rapidly than the unreinforced material for Cl ion concentrations greater
than 1%. The corrosion rate for the composite was three times greater than the
matrix material at 3.5% NaCl.

B. Continuous Reinforcement

1. Aluminum Matrix Composites


Continuous fiber reinforcement offers a number of advantages over discontinuous
reinforcement of metals. Most of these advantages are in the mechanical properties such as specific strength and stiffness. Specific strengths of 500 MPa/mg/
m3 and specific stiffness of 80 GPa/mg/m3 are possible with continuous-fiberreinforced aluminum (14) where only graphite/epoxy has higher values. Graphite,
boron, and alumina fibers have been used to reinforce aluminum, but the transverse properties and corrosion performance of continuous-fiber composites are
not as good as those of discontinuous reinforced composites. Some other continuous reinforced composites include unidirectionally solidified eutectics such as
AlAl4Ca (15) and metalpolymer laminates (16).
Continuous-fiber-reinforced composites have been characterized by Trzaskoma (17) as Type I, where the reinforcing fibers are exposed on four of six sides
of a cube, and discontinuous reinforcement as Type II where the reinforcement is
exposed on all six sides of a cube. Crevice corrosion, galvanic corrosion, and
pitting are possible in both Type I and II configurations; however, continuous
fibers offer the potential for deep crevices in the direction of the fibers and graphite fibers produce significant galvanic effects.
Both Pohlman (18) and Sedriks et al. (19) concluded that localized corrosion of BAl composites was not the result of galvanic effects. Pohlman (18)
evaluated BAl couples and found no galvanic current, whereas boron fibers
extracted from a composite that had a BAl intermetallic on the surface did show
galvanic effects. Therefore, a galvanic effect is expected for hot-pressed BAl

380

Jones

composites where this intermetallic formed at the fibermatrix interface. Pohlman (17) associated the localized corrosion observed at the fibermatrix interface
with the presence of the BAl intermetallic phase. Sedriks et al. (19) measured
the corrosion behavior of B-fiber-reinforced 2024 aluminum alloy. They reported
the loss in tensile strength of 2024 aluminum reinforced with 15% and 40% B
following heat treatment for various times at 190C and exposed with and without
stress to a solution of 53 g of NaCl per liter of water. The maximum strength
loss occurred for aging treatments of 12 h, with the matrix losing 45%, the
202415% B composite losing 50%, and the 202440% B composite losing 60%.
Specimens exposed to this same environment but with a stress of 90% of the
yield strength again showed an effect of the volume fraction of B reinforcement
where specimens with the matrix material did not fail in 4000 h, 202415% B
specimens failed in 1000 h, and 202440% B specimens failed in 40 h. Evans
and Braddick (20) noted a preferential attack at the fibermatrix interface for Bfiber-reinforced aluminum composites. They concluded that this localized corrosion was likely the result of an Al boride that formed at the interface during hot
pressing. Weight-loss results show a rapid increase with time over the first 80
h, reaching a limiting weight loss of about 5 103 g/cm2. Similar results were
obtained for a C-fiber-reinforced aluminum composite.
One of the earliest reports of the corrosion behavior of CAl composites
was that by Dull et al. (21) for 6061 aluminum reinforced with carbon fibers.
These materials were made by laying up alternating layers of fibers and foils and
hot pressing to achieve the desired density. Corrosion tests were conducted in
3.5% NaCl at temperatures between 25C and 75C. The corrosion rate of the
composite was 15 times greater than that of the alloy when tested at 25C and
increased substantially for temperatures exceeding 50C. A graphite fiber6061
Al composite was also evaluated for its corrosion performance by Aylor et al.
(22), although the material was produced by infiltrating graphite fiber tows with
6061 aluminum alloy and hot pressing these tows. The graphite6061 Al tows
were hot pressed between 0.3-mm-thick alloy foils so that the outer skin had no
exposed fibers, although the edges did have exposed graphite as in the Type I
configuration discussed by Trzaskoma (16). Aylor et al. (22) also found that the
composite exhibited considerably faster corrosion than the matrix material in
splash and spray tests, as did Dull et al. (21). Pitting of the 6061 Al outer foil,
which was the same on the composite and unreinforced 6061 Al, preceded the
rapid corrosion of the underlying composite. Galvanic effects of the graphite
Al couple were suggested as the primary cause of the accelerated corrosion.
Shimiziu et al. (4) noted that the pitting potential for graphite-fiber-reinforced
6061 Al was the same as the unreinforced alloy but that pit initiation was suggested as being easier because of the galvanic couple between graphite and the
6061 Al. They also noted a 100-mV shift of the open-circuit potential in the

Metal Matrix Composites

381

anodic direct and a much higher cathodic current density than the unreinforced
material or material reinforced with SiC whiskers or Al2O3 fibers.
In studies to evaluate the corrosion mechanism in fiber-reinforced aluminum composites, Hihara and Latanision (23,24) found residual chloride in a
graphite6061 Al composite and measured galvanic corrosion for SiC, C, and
TiB2 fibers coupled to 6061 Al. The chloride was traced to the aluminum infiltration process and was suggested as a factor in accelerated corrosion of these composites. The galvanic current density for the graphite fiberaluminum couple was
30 times that of SiC or TiB2 fiberaluminum couples. The galvanic current density was controlled by the rate of O2 reduction on the graphite and, therefore, the
galvanic current density was greater in aerated solution than nonaerated solutions.
In an effort to develop a corrosion-resistant graphitealuminum composite,
Wendt et al. (25) evaluated the corrosion performance of graphite-reinforced Al
Mo alloys. Alloys with 1123% Mo were produced by sputter deposition and
the corrosion behavior measured in 0.1M NaCl at a pH of 8 and a temperature
of 25C. Galvanic current densities for the Al18% Mo and Al23% Mo alloys
coupled to graphite were three orders of magnitude less than for pure Al coupled
to graphite.
A directionally solidified eutectic of Al reinforced with lamina of Al4Ca
intermetallic (15) can be considered in the same category as continuous-fiberreinforced metal matrix composites because of the potential for continuous crevice corrosion in the direction of the Al4Ca intermetallic. Corrosion testing was
conducted in 3.5% NaCl at pH of 7 and a temperature of 25C. The electrochemical response of the eutectic was essentially equal to that of the extruded alloy
while being better than an AlCu alloy.

2. Magnesium Matrix Composites


Magnesiumgraphite composites with very high specific strengths and stiffnesses
and low coefficients of thermal expansion are attractive materials for use in space
applications. Although these materials could perform very well in the dry vacuum
of space, they must exhibit sufficient corrosion resistance to survive manufacture,
storage, and transport on Earth in the presence of moist salt air. Magnesium is
one of the most reactive metals, so the galvanic effects between graphite and
magnesium is of considerable importance. Trzaskoma (26) evaluated the corrosion performance of graphite-fiber-reinforced AZ91C which had AZ31B outer
foils. Open-circuit potential-time and galvanic short-circuit current measurements
were conducted in a borated boric acid solution containing 1000 ppm of NaCl
at a pH of 8.4. Severe localized corrosion occurred at the edges where there was
exposed graphite fibers after 5 days of immersion in this solution. This corrosion
was evident (a) along the face plate, (b) in the underlying metal, and (c) as damage

382

Jones

to the fibers. The open-circuit potential shifted with time in a very similar manner
for the reinforced and unreinforced AZ91C material. A galvanic corrosion rate
for the exposed edges of this composite was determined to be 1.8 mg/cm2 /day,
which is very fast for structural materials.
A potentially more stable Mg matrix composite reinforced with SiC fibers
was evaluated by Hihara and Kondepudi (27). The composite was a unidirectionally aligned SiC monofilaments in a matrix of ZE41 aged at 329C for 2 h to a
T5 temper. The volume fraction of SiC monofilaments was about 50%. Corrosion
studies were conducted in near-neutral 0.5M NaNO3 solutions at 30C, de-aerated
with high-purity N2. The magnesium alloy ZE41 had the following composition:
4.2% Zn, 0.7% Zr, 1.2% rare earth. Hihara and Kondepudi (27) found that the
presence of the SiC monofilament caused the MMC to corrode faster in aerated
than de-aerated solutions, unlike the matrix alloy in which the corrosion behavior
was insensitive to oxygen concentration. This difference in corrosion behavior
was evidenced by a 100-mV shift in the open-circuit potential of the composite
material in aerated solutions relative to de-aerated solutions. The effect of O2
was explained by the reduction of O2 on the SiC. The corrosion rate of both the
matrix material and the composite were very high at small voltages anodic to
the open-circuit potential. A corrosion current density of about 3.2 103 A/
cm2 was reported for the composite.

3. Titanium Matrix Composite


Titanium reinforced with monofilament SiC have potential application in drive
shafts, turbine engine disks, compressor disks, and hollow fan blades and were
being considered for the skin of the National Aerospace Plane prior to the cancellation of this program. The most common matrix materials are Ti6Al4V and
Ti15V3Cr3Sn3Al (Ti-15-3) and the most common fiber is the Textron Specialty Materials SiC. Hihara and Tamirisa (28) examined the corrosion resistance
of Ti-15-3 reinforced with monofilament SiC fibers. The composite had nine
plies and was unidirectionally reinforced with SCS-6 fibers to volume fraction
of 40%. The composite was given a heat treatment of 790C for 30 min and aged
at 510C for 8 h.
The Ti-15-3 matrix exhibited passive behavior in a solution of 3.15 wt%
NaCl, whereas the composite material did not. The current density was at least
a factor of 10 greater for the composite. Hihara and Tamirisa (28) calculated the
polarization behavior of the composite from corrosion measurements conducted
separately on the matrix and fibers and calculated using mixed potential theory.
The experimental and calculated polarization curves were identical, which led
the authors to conclude that the composite fabrication process had not altered
the electrochemical behavior of the constituents. At open-circuit potential, the
composite material exhibited excellent corrosion resistance.

Metal Matrix Composites

383

III. STRESS CORROSION CRACKING AND CORROSION


FATIGUE BEHAVIOR
Metal matrix composites have the potential to exhibit stress-corrosion properties
superior to those of the base alloy because the reinforcement provided by particles, whiskers, or fibers reduces the crack tip strain rate. Of course, this potential
can only be realized if the corrosion performance of the composite is not degraded
relative to the matrix material. As discussed earlier, this is most likely the case
for chemically stable, nonconducting, ceramic reinforcements such as SiC or
Al2O3. The localized corrosion that has been noted for many of the MMCs could
impact the stress-corrosion or corrosion fatigue performance of MMCs. The following sections summarize the crack growth observations for MMCs with discontinuous and continuous reinforcements. Much of the data are for cyclically loaded
specimens because stress-corrosion measurements are more difficult because of
the high K Iscc /K Ic ratio.
A. Discontinuous Reinforcement
Some of the first measurements of the effect of particle and whisker reinforcement
on corrosion fatigue were made by Yau (29) and Hasson et al. (30). Yau measured
the fatigue crack growth behavior of the 6061 AlSiC composite in air, water,
and water with 3.5% NaCl. The behavior in water with and without NaCl was
identical but exhibited slightly greater crack velocities than the tests in air. This
suggests that water and water plus NaCl likely caused some increased corrosion
rate and degradation of the composite relative to the matrix material. However,
the reinforced material exhibited a slower crack velocity in water plus 3.5% NaCl
than the unreinforced material. Therefore, if the localized corrosion effect of the
reinforcement can be eliminated, these results suggest the possibility for a beneficial effect of reinforcement on stress corrosion and corrosion fatigue.
Hasson et al. (30) evaluated unreinforced 6061-T6 and 6061-T6 reinforced
with 20% SiC whiskers and particles. The l/d ratio for the whiskers ranged from
2 to 5 following thermalmechanical processing. Their results show the cycles
to failure as a function of alternating stress in laboratory air and moist salt air.
The whisker-reinforced material had 100 times greater cycles to failure than the
unreinforced material when tested in moist salt air at an alternating stress of 96
MPa. This is a significant improvement imparted by the presence of the SiC
whiskers. Hasson et al. (30) did not report corrosion results for their test conditions, but it is likely that moist salt air caused less localized corrosion and degradation than aqueous corrosion tests where localized corrosion has been noted for
MMCs. The improvement in corrosion fatigue imparted by the SiC whiskers was
reduced at high alternating stresses. For instance, at an alternating stress of 144
MPa, the whisker reinforcement imparted less than a factor of 10 increase in

384

Jones

cycles to failure. Also, in contrast to the observations of Yau (29), the cycles to
failure of the reinforced material tested in moist salt air exceeded the cycles to
failure of the unreinforced material tested in laboratory air.
Corrosion fatigue results of 6061 Al reinforced with 10% Al2O3 and tested
in air and 3.5% NaCl have recently been reported by Bertolini et al. (31). They
observed that there was little difference in the da/dn versus K curve between air
and 3.5% NaCl for material extruded and forged 17%, but there was an increase in
da/dn for material extruded 8.3%. However, the increase in da/dn was only about
a factor of 5. Pitting at particlematrix interfaces was noted following potentiodynamic corrosion tests, as noted by others. The authors concluded that the pitting
corrosion contributed to the corrosion fatigue crack growth results; however, they
did not evaluate an unreinforced 6061 Al alloy to determine if the particleinduced localized pitting was a significant factor in the da/dn versus K behavior
of the composite.

B. Continuous Reinforcement

1. Aluminum Matrix Composites


Experimental results on the stress corrosion or corrosion fatigue of continuousor fiber-reinforced material is equally sparse as it is for discontinuous particle or
whisker-reinforced material. Berkeley et al. (32) measured the remaining strength
of 6061 Al reinforced with Nextel 440 fibers after exposure to NaCl at pH levels
of 1.5 and 2.0. Nextel 440 has a composition of 70% Al2O3, 28% SiO2, and 2%
B2O3. The composite contained 45% Nextel fibers, and the specimens were exposed to the ASTM G44 solution stressed to 80% of their yield strength. The
strength of the composite following exposure was used as a measure of the environmental degradation of the composite. Although this method may measure the
effects of stress-corrosion cracking by cracks that grow into the material, it may
only be measuring an embrittling effect. Preferential corrosion at the fibermatrix
interface following the length of the fiber was observed following exposure to
this low-pH solution and was connected to the strength reduction observed. The
residual strength of the composite material following 5 days of exposure to the
ASTM G44 solution was about 80% regardless of whether a stress was applied
during the exposure. For the same exposure conditions, the 6061 Al alloy had a
residual strength of 95%. The authors concluded that localized corrosion along
the length of the fiber at the fibermatrix interface was the primary cause of the
larger strength reduction for the composite compared to the matrix alloy.
Sedriks et al. (33) evaluated diffusion-bonded 2024 aluminum alloyboron
filament composites and reported localized corrosion of the transverse ends with
and without an applied stress. Stress-corrosion tests of samples loaded in the
transverse direction in a NaCl solution to 90% of the yield strength failed in 1000

Metal Matrix Composites

385

h for 15% B fibers and 40 h for 40% B fibers. The failures were intergranular
in the matrix and interfacial between the matrix and fibers. No failures were
observed for times up to 4000 h in the 2024 alloy tested in the longitudinal
direction. Stress-corrosion tests of samples loaded in the longitudinal direction
showed that the composite outperformed the 2024 Al in NaCl solution. No failures occurred in the composite for times up to 1000 h, whereas the matrix failed
in less than 10 h. The inferior stress-corrosion properties of the composite in the
transverse direction resulted from the exposed transverse cross sections.
A unique composite comprised of metal matrix and polymer composite
laminates has been evaluated by Wanhill (16). Corrosion fatigue tests were conducted on 2024 Al-T3 with carbonepoxy laminates and 7475 Al-T761 with
carbonepoxy laminates in air and air with a water spray. The specimens were
a symmetrical laminate of Al face sheets over a unidirectional carbon fiberepoxy
composite core. These materials were being evaluated for aerospace applications
so they were loaded with a flight simulation loading program with a mean load
of about 20% of the ultimate load capacity of the composite. The possibility of
a galvanic potential between the carbon fibers in the epoxy composite and the
aluminum alloys existed, but there was no evidence for accelerated corrosion
from galvanic corrosion. The results are reported in number of flights to produce
a crack extension of 40 mm. In the air plus water spray tests, the 2024-T3 carbon
epoxy composite was superior to the 2024-T3 material and the other composites.
In an environment of air plus water spray, the 2024-T3 carbonepoxy composite
required 8300 flights for 40 mm of crack extension, whereas the matrix required
only 4800 flights and the 7475-T761 carbonepoxy composite only 1800 flights.

2. Magnesium Matrix Composite


Stress-corrosion tests of a magnesium alloy reinforced with alumina fibers reported by Evans (34) is possibly the strongest case for the potential benefit of
composite reinforcement on stress corrosion and corrosion fatigue. A magnesium
alloy, ZE41A, with a composition of 4.5 Zn, 0.7 Zr, and 1.0% rare earths was
reinforced with 55 vol% alumina fibers. Tests were conducted with fibers oriented
perpendicular (90) and parallel (0) to the principal stress in a solution of NaCl
and K2CrO4 and reported as time to failure versus applied stress and crack velocities (determined from the change in the bending moment with time) versus time.
Both unnotched and chevron-notched specimens were loaded in a cantilever configuration. For the unnotched specimens, the failure time was measured as a function of the applied stress and a clear benefit of the alumina fibers was noted for
the 0 composite. Composites with the 90 fiber orientation had properties similar
to that of the matrix. A threshold stress of 600 MPa was found for the 0 composite and about 200 MPa for both the matrix and 90 composites. For the chevronnotched specimens, a crack growth velocity of about 105 mm/s was estimated for

386

Jones

the 0 composite. However, the K th for the 90 composite was only 6.5 MPam1/2
while it was 16.2 and 11.2 MPam1/2 for the 0 composite and the matrix, respectively.

3. Titanium Matrix Composites


Mahulikar and Marcus (35) evaluated the fatigue crack growth behavior of B4C
B and BORSIC reinforced Ti6Al4V in humid air. Fatigue tests were conducted
at an R of 0.1 to measure crack closure effects and compared to results at an R
of 0.5 and with variable water-vapor pressures. The primary observation was that
a transition from interface separation to fiber splitting occurred at a water-vapor
pressure of 100 MPa. The fiber splitting was explained as resulting from crack
closure effects rather than a direct embrittlement of the fiber by the environment.
Increased crack growth velocities accompanied the occurrence of fiber splitting.
Ti6Al4V/carbonepoxy laminate composites were also evaluated by
Wanhill (16) along with the 2024 Al and 7475 Al laminate composites described
earlier. The Ti6Al4V/carbonepoxy composite required only 1400 flights for
a crack extension of 40 mm when tested in an air plus water spray environment
and about 3600 cycles in air. This is in contrast to 8300 flights for the 2024 Al/
carbonepoxy and 4800 flights for the 7475/carbonepoxy composites. The poor
performance of the Ti6Al4V was related to cracking of the carbonepoxy
layer.

4. Iron Matrix Composite


The stress-corrosion performance of a laminate composite of maraging steel and
Armco iron was determined by Floreen et al. (36). The laminate was produced
by hot-rolled bonding plates of the 18% Ni maraging steel with a sheet of Armco
iron between each plate. Tests were performed with a notch-bend specimen with
the notch oriented in a notch-arrester orientation (cracks running through the
thickness of the plates) and crack-divider orientation (cracks running parallel to
the layer interfaces). Composites with two, four, and eight layers of maraging
steel were tested. Tests were performed in an aerated 3.5% NaCl solution. In all
cases, the composite failure time was over 240 h for the crack-arrester orientation
but ranged between 63 and 80 h for the crack-divider orientation for two, four,
and eight layers of maraging steel. Stress-corrosion cracks in the crack-arrester
specimens propagated through the maraging steel but were diverted and branched
in the Armco iron. The Armco iron layer was an effective crack-arrest layer.

5. Model of Stress-Corrosion Cracking in Composite Materials


A model of the stress-corrosion crack growth rate of metal matrix composites
has been presented by Jones (37). This model considers the length-to-diameter

Metal Matrix Composites

387

(l/d) ratio and volume fraction of reinforcing phase and the matrix stress exponent. The model is based on the assumption that several stress-corrosion mechanisms are controlled by crack tip strain rate and that the presence of a reinforcing
phase reduces the crack tip strain rate. Processes such as the passive film rupture
rate, dislocationsurface interaction rate, surface reactivity rate, and crack opening effects on transport of species into and out of the crack and crack wall corrosion rates are all factors that depend on the crack tip strain rate.
The resulting predictions of this model are presented in Figs. 1 and 2, along
with data by Yau (29), Hasson et al. (30), and Jones (37). The model predictions
given in Fig. 1 show that the value of R, the ratio between the stress corrosion
velocity of the composite to that of the unreinforced matrix, is a strong function
of the l/d ratio and moderately dependent on n, the matrix creep stress exponent.
Increasing values of the matrix stress exponent causes a decrease in the value of
R. The predictions given in Fig. 2 show that R has only a small dependence on
the volume fraction of the reinforcing phase over the range 540%. This model
assumes that the reinforcement does not accelerate the corrosion rate and induce
other damage. As discussed earlier, discontinuous reinforcement with ceramic
phases often cause localized corrosion around clusters of particles or the particle
matrix interface. This localized corrosion very likely influences the quality of the

Fig. 1 Calculated and experimental values of R versus l/d and n for 7090 Al/SiC and
6061 Al/SiC.

388

Fig. 2

Jones

Calculated values of R versus l/d and Vf.

comparison between the model and experiment. The results of Jones (37) and
Yau (29) are consistent with the model and matrix stress exponent of 36. Hasson
et al. (30) results are plotted at an l/d of 100, which is based on the whisker
dimensions prior to processing the composite, but they reported that the l/d was
reduced to about 5 after processing. Therefore, their results should be shifted
substantially to the left as the arrow indicates.

IV. CORROSION PROTECTION


Many of the desirable properties of metal matrix composites are compromised
by corrosion, especially for continuous reinforcement with graphite. Because of
this, several schemes have been evaluated to provide corrosion protection of these
materials. Aylor et al. (38) evaluated the effects of electroless nickel coatings
and anodization on the corrosion performance of 6061 Al with continuous graphite fibers. Tests were performed by immersion in seawater, splash, spray, and
atmospheric exposure. Corrosion on the uncoated composite began with pitting
and progressed to galvanic corrosion between the graphite fibers and the aluminum matrix. Electroless nickel coating resulted in severe galvanic corrosion of
the 6061 Algraphite substrate because of flaws in the coating and the high Ni/
Al ratio, and therefore cathode-to-anode ratio, at these flaws. Anodization increased the corrosion resistance of the composite with some thinning of the anod-

Metal Matrix Composites

389

ized layer that did not impact the corrosion resistance for times up to 180 days.
Anodization was judged to be a useful method for protecting metal matrix composites. Mansfeld et al. (39) concluded that chromate conversion coatings and
immersion in CeCl3 both produced stable, protective coatings on a SiCAl composite. Cathodic protection is another route for protecting these composites, but
Hihara and Latanision (40) have demonstrated that cathodic overprotection can
lead to accelerated corrosion for aluminum matrix composites. This occurs because of the production of OH and the instability of aluminum in high-pH environments. The effect of cathodic overprotection was the most severe in the 6061
Algraphite composite and less so in the 6061 Al/SiC composite and 6061 Al.
Modifying the matrix corrosion behavior through alloy compositional changes
is a route to improving the corrosion performance of Algraphite composites
considered by Wendt et al. (25). Aluminum and AlMg alloys with 11%, 18%,
and 23% molybdenum were prepared by sputter deposition and the galvanic corrosion response of these alloys with graphite determined from galvanic diagrams
constructed from the polarization data. The authors concluded that an Al18%
Mo/graphite composite would exhibit good galvanic corrosion behavior.
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.

14.
15.
16.
17.

LH Hihara, RM Latanision. Int Mater Rev 39(6):245, 1994.


PP Trzaskoma, E McCafferty, CR Crowe. J Electrochem Soc 130:804, 1983.
PP Trzaskoma. Corrosion 46:402, 1990.
Y Shimizu, T Nishimura, I Matsushima. Mater Sci Eng A 198:113, 1995.
RC Paciej, VS Agarwala. Corrosion 42:718, 1986.
RC Paciej, VS Agarwala. Corrosion/87. Houston, TX: National Association of Corrosion Engineers, 1987, Paper 221.
H Sun, EY Koo, HG Wheat. Corrosion 47:741, 1991.
PCR Nunes, LV Ramanathan. Corrosion 51:610, 1995.
H-Y Yao, RZ Zhu. Corrosion 54:499, 1998.
M Saxena, OP Modi, AH Yegneswaran, PK Rohatgi. Corros Sci 27:249, 1987.
A Luo. Metall Mater Trans A 26A:2445, 1995.
CA Nunez-Lopez, P Skeldon, GE Thompson, P Lyon, H Karimzaden, TE Wilks.
Corros Sci 37:689, 1995.
WM Chan, FT Cheng, LK Leung, RJ Horylev, and TM Yue. Corrosion behavior
of magnesium alloy AZ91 and its MMC in NaCl solution. Corrosion/97. Houston,
TX: National Association of Corrosion Engineers, 1997, Paper 441.
HE Deve, C McCullough. J Metals 47:33, 1995.
LP Bicelli, C Romagnani, D Sinigaglia. Corros Sci 19:553, 1979.
RJH Wanhill. Fatigue Eng Mater Struct 2:319, 1979.
PP Trzaskoma. Localized corrosion of metal matrix composites. In: RH Jones, RE
Ricker, eds. Proceedings of Environmental Effects on Advanced Materials. Warrendale, PA: TMS, 1991, p. 249.

390
18.
19.
20.
21.

22.
23.
24.
25.
26.
27.
28.
29.
30.

31.

32.
33.
34.
35.
36.
37.
38.
39.
40.

Jones
SL Pohlman, Corrosion 34:156, 1978.
AJ Sedriks, JAS Green, DL Novak. Met Trans A 2:871, 1971.
JM Evans, DM Braddick. Corros Sci 11:611, 1971.
DL Dull, WC Harrigan, MF Amateau. In: FH Meyer, ed. Proceedings of the 1974
Triservice Corrosion of Military Equipment Conference, Vol. 1 (Dayton, OH, Air
Force Materials Laboratory Report, AFML-TR-75-42, 1975), p. 399.
DM Aylor, RJ Ferrara, RM Kain. Mater Perform 23:32, 1984.
LH Hihara, RM Latanision. Mater Sci Eng A126:231, 1990.
LH Hihara, RM Latanision. Corrosion 48:546, 1992.
RG Wendt, WC Moshier, B Shaw, P Miller, DL Olson. Corrosion 50:819, 1994.
PP Trzaskoma. Corrosion 42:609, 1986.
LH Hihara, PK Kondepudi. Corros Sci 34:1761, 1993.
LH Hihara, C Tamirisa. Mater Sci Eng A198:119, 1995.
S Yau. Crack propagation in corrosion-fatigue of metal matrix composites. PhD
thesis, North Carolina University, Raleigh, 1983.
DF Hasson, CR Crowe, JS Ahearn, DC Cooke. In: Proceedings of Failure Mechanisms in High Performance Materials, Gaithersburg, Maryland, May 13, 1984.
New York: Cambridge University Press, 1985, p. 147.
L Bertolini, MF Brunella, S Candiani. Effect of chloride environments on fatigue
behavior of AA6061Al2O3 particle composite. In: Corrosion/98. Houston, TX: National Association of Corrosion Engineers, 1998, Paper 738.
DW Berkeley, HEM Sallam, H Nayeb-Hashemi. Corros Sci 40:141, 1998.
AJ Sedriks, JAS Green, DL Novak. Metal Trans. 2:781, 1971.
JT Evans. Acta Metall 34:2075, 1986.
D Mahulikar, HL Marcus. Metal Trans A 15A:209, 1984.
S Floreen, HW Hayden, N Kenyon. Corrosion 27:519, 1971.
RH Jones. In: RH Jones, RE Ricker, eds. Proceedings of Environmental Effects on
Advanced Materials. Warrendale, PA: TMS, 1991, p. 283.
DM Aylor, RJ Ferrara, RM Kain. Mater Perform 32, July 1984.
F Mansfeld, S Lin, K Kim, H Shih. Corros Sci 27:997, 1987.
LH Hihara, RM Latanision. Scripta Metall 22:413, 1988.

13
Ceramic Matrix Composites
Russell H. Jones, C. H. Henager, Jr., Charles A. Lewinsohn, and
Charles F. Windisch, Jr.
Pacific Northwest National Laboratory, Richland, Washington

I.

INTRODUCTION

Ceramic matrix composites (CMCs) are being developed to take advantage of


the high-temperature properties of ceramics while overcoming the low fracture
toughness of monolithic ceramics. Toughening mechanisms, such as matrix
cracking, crack deflection, interface debonding, crack-wake bridging, and fiber
pullout, are being incorporated in CMCs to reduce the tendency for catastrophic
failure found in monolithic ceramics. Ceramics reinforced with particulate,
whiskers, and continuous fibers exhibit varying aspects of these toughening
mechanisms; however, reinforcement with continuous fibers offers the greatest
improvements in toughness. Composites with carbide, oxide, glass, and carbon
matrices are being utilized in the development of CMCs. In the case of carbide,
oxide, and glass matrix CMCs, the matrix exhibits excellent high-temperature
corrosion resistance so that a goal of the composite development is to not detract
from this preexisting property. This is not the case for carbon matrix composites,
which frequently need coatings to provide adequate corrosion protection. The
purpose of this chapter is to review the database and understanding of corrosion
behavior of CMCs with the intent that this information will be useful in the
development of materials with improved performance and reliability.
Composite materials are chemically and microstructurally heterogeneous,
consisting of matrix, matrixreinforcement interface, and reinforcement constituents. The corrosion behavior of each of these constituents will likely not be
equal whether evaluated individually or within the composite. The composite
corrosion resistance may be more complex than each constituent because of inter391

392

Jones et al.

actions between corrosion reaction products and the composite constituents. An


example of this interaction would be a reaction product from the corrosion of
the matrix that protects a less corrosion-resistant reinforcement or reinforcement
matrix interface. However, there is far more data on the corrosion behavior of
the constituents as monolithic materials than there are on the composites made
from these constituents. It is important to note that the constituents of a composite
may differ slightly or substantially in chemical composition, crystal structure,
and microstructure from their monolithic equivalent. Examples of these differences are the SiC fibers and matrix produced from polycarbosilane or its variants
that may contain considerable oxygen or may be nanocrystals embedded in an
amorphous matrix. There is considerable value in evaluating the corrosion resistance of the monolithic equivalent of the composite constituents because these
generally represent the baseline corrosion behavior. An effort will be made to
identify where significant differences are likely to occur between the composite
constituents and the monolithic equivalent.
High-temperature composites are being developed to operate in a variety
of environments containing alkali elements, mildly oxidizing mixed gases such
as HeO2, highly oxidizing environments, and H2. Not all ceramic composites
are being considered for each of these environments, so the corrosion data are
not available for each ceramic composite in every environment.

II. SILICON CARBIDE MATRIX COMPOSITES


A. Corrosion Reactions for Silicon Carbide
Silicon carbide will chemically react with O2, H2, and H2O according to the following reactions:
SiC(s) O2(g) SiO(g) CO(g): low pO2

(1a)

SiC(s) / O2(g) SiO2(s) CO(g): high pO2

(1b)

32

SiC(s) 2H2(g) Si CH4(g)

(2)

SiC(s) 2H2O(g) SiO(g) CO(g) 2H2(g)

(3)

Silicon carbide is thermodynamically unstable in O2, H2, and H2O environments


under certain conditions. However, the kinetics of these reactions can be affected
by the formation of a protective layer of SiO2 at high pO2 such that SiC is stable
in many corrosive environments. The stability of the passive layer then becomes
critical to the stability of SiC. The O2 pressure for the SiC activepassive transition was determined by Gulbransen and Jansson (1) as shown in Fig. 1. At temperatures of 8001000C, this transition occurs at O2 pressures of 108 atm. Therefore, the pertinent reactions then become those between SiO2 and specific gaseous

Ceramic Matrix Composites

393

Fig. 1 Transition pressures for SiC activepassive oxidation versus temperature, according to Gulbransen and Jansson (1).

or molten-salt environments, except in molten Li with a low O2 activity, where


SiO2 is unstable. Some relevant reactions with SiO2 are below:
SiO2(s) H2(g) SiO(g) H2O(g)

(4)

xSiO2(s) Na2SO4(P) Na2OX x(SiO2)(P) SO3(g)

(5)

where M is an alkali element such as Na and Li. A phase diagram for the Na2O
SiO2 system is shown in Fig. 2. Alkali elements such as Na and Li cause a breakdown of the passive SiO2 film by the formation of low-melting alkali silicates
such as those which occur at 800C in the Na2 SiO2 system and 1024C in the
Li2OSiO2 system. The eutectic temperature in the Li2OSiO2 system is about
250C higher than that in the Na2OSiO2 system and, therefore, Li is expected
to have less effect on the passive film on SiC at temperatures below 1000C than
Na.
A summary of the behavior of SiC in gasmolten-salt environments as
presented by McKee and Chatterji (2) is shown in Fig. 3. Passivation occurs at
high pO2, and active oxidation (formation of gaseous SiO) occurs at low pO2. A
basic salt or salt melt with a low pO2 at the saltSiC interface will cause active
corrosion, as depicted by reaction scheme (4) and (5). McKee and Chatterji suggest that SiC will not react with H2; however, more recent analysis by Herbell
et al. (3) indicates that the equilibrium partial pressure of CH4 [Eq. (2)] is 104

394

Fig. 2

Jones et al.

Phase diagram for the system Na2OSiO2.

atm at temperatures of 8501400C for a H2 pressure of 1 atm. While McKee


and Chatterji measured the sample weight change in the test environment as a
function of time in 1 atm of H2 at 900C, they observed no reaction between SiC
and H2 (scheme 1, Fig. 3). However, their gas may have had sufficient O2 or H2O
to cause passivation. Herbell et al. (3) calculated the SiO(g) partial pressure for
Eq. (4) with 1 atm of H2 containing 1 ppm of H2O to be 107 atm. Therefore, it
would appear that a very small amount of H2O mixed with O2 to promote SiO2
formation would be sufficient to cause a significant reduction in the reaction rate
of SiC.

Ceramic Matrix Composites

395

Fig. 3 Possible modes of behavior of SiC in gasmolten-salt environments.

Jacobson (4) evaluated the kinetics and mechanisms of the corrosion of


sintered -SiC in molten salts at 1000C. In the reaction of Na2SO4 /O2 with SiC,
the reaction occurred primarily in the first few hours with the formation of a
protective SiO2 layer. This observation was demonstrated with results showing
the total weight of the corrosion products/unit area reaching 6 mg/cm2 (SiO2
Na2Ox(SiO2) after a few hours and remaining constant up to 20 h. Jacobson
and Smialek (5) also noted that SiC is subject to pitting corrosion in molten
salts. Pits occurred at structural discontinuities and bubbles formed during the
formation of SiO2. Pitting corrosion is detrimental because it demonstrated that
the passive SiO2 layer has been degraded and because the pits act as flaws resulting in reduced fracture strength.
Corrosion measurements of SiC immersed in a crucible of molten Na2SO4
were conducted by Tressler et al. (6). Weight change results for both SiC and
Si3N4 in 100% Na2SO4 at 1000C are given in Fig. 4, where it is evident that
considerable corrosion occurs for both materials, with SiC showing a higher corrosion rate than Si3N4. Both materials exhibited a weight loss because the reaction

396

Jones et al.

Fig. 4 Weight change versus exposure time to Na2SO4 at 1000C for several ceramic
composites. (From Ref. 6.)

product was removed prior to measuring the weight change. An in situ measurement would probably exhibit a weight gain up to the time at which reaction
product spallation occurs. Corrosion data obtained on SiCSi3N4 and SiCSiC
composite materials by Henager and Jones (7) is also presented in Fig. 4. The
corrosion rate of the composite material was found to exceed that of the monolithic material for a NicalonSiC-reinforced, hot-pressed Si3N4 and a Nicalon
SiC reinforced CVI SiC. A SiC-whisker-reinforced, hot-pressed Si3N4 exhibited
a similar corrosion rate to the unreinforced hot-pressed Si3N4. The activation energy for the corrosion of the fiber-reinforced material was 880 kJ/mol, whereas
the whisker-reinforced matrix material had an activation energy of 1280 kJ/mol.
The difference in corrosion rate and activation energy for the fiber-reinforced
material was associated with preferential corrosion of the Nicalon fiber. The fiberreinforced material had a carbon interfacial layer between the fiber and matrix,
and the whisker-reinforced material had a thin amorphous glass layer. However,
the difference in corrosion behavior is not thought to be associated with the interfacial layer but with the structure of the Nicalon fiber following hot pressing.
As produced, Nicalon has an amorphous/microstryalline structure while the hotpressed structure was crystallized with evidence for a Mg silicate in the grain
boundaries. The source of the Mg was the MgO sintering aid used for the hotpressing process. Corrosion of the Nicalon fiber along the glass-enriched grain

Ceramic Matrix Composites

397

boundaries probably accounted for the high corrosion rate of the Nicalon-fiberreinforced composite material.
The corrosion rate of a SiCSiC composite produced by the CVI process
was also greater than that of monolithic SiC (Fig. 4). Microstructures of CVI
SiCSiC composites made by the ORNL process and the Du Pont process both
exhibit porosity, which exists in materials made by this process. Composites containing fibers coated with C, BN, CB4CBN, and no coating were evaluated.
The corrosion rate was highest for the material made with fibers with no coating
and the BN coating, whereas composites made with C and CB4CBN coatings
exhibited smaller corrosion rates. However, the corrosion rate of these materials
at 900C was considerably greater than that of monolithic material at 1000C.
Matrix corrosion contributed the majority of the weight change, and while some
fiber corrosion was noted, although this was a relatively small factor compared
to the matrix corrosion. The open porosity and penetration of the molten salt into
the composite are considered the primary factors causing the high corrosion rates
for the SiCSiC composite.
Fox et al. found similar results in burner rig corrosion studies of CMCs to
those obtained in molten salt (8). The burner rig tests utilize a hot-combustion
gas and would be expected to differ from a molten Li or He gas environments;
however, molten deposits are one mechanism for corrosion in hot-combustion
gas studies. Fox et al. evaluated Si3N4 reinforced with 30 vol% SiC whiskers and
fibers. The method of procesing for the whisker-reinforced materials was not
reported; however, it was probably hot pressed, whereas the fiber-reinforced material was reaction bonded and had a residual porosity of 30%. Tests were conducted for 40 h at 1000C in an environment containing 2 ppm Na added as
NaCl. Monolithic Si3N4 exhibited a 10% decrease in the fracture stress and the
whisker-reinforced material a 35% decrease. The fiber-reinforced material exhibited an excessive corrosion rate, but no fracture studies were conducted. The high
residual porosity in the reaction-bonded material and the free C from the carboncoated fibers were postulated as the cause for the high corrosion rate in the fiberreinforced material. In the molten-salt corrosion studies (8), the more porous CVI
composite material also exhibited a higher corrosion rate than did the hot-pressed
Si3N4 /fiber- or whisker-reinforced-material.
Effects of corrosion on the mechanical integrity of SiCSiC structural components is an important factor in the durability of components constructed of
SiC/SiC. Decreases in the fracture strength would be possible by corrosion penetration along grain boundaries. Smialek and Jacobson (9) also found that moltensalt-induced pits were responsible for up to 50% of the observed strength reduction of SiC after 48 h at 1000C in Na2SO4 SO3. A relationship of fracture stress
versus (pit depth)2 illustrated the flaw-induced fracture relationship. Henshall et
al. (10) demonstrated that combustion gases containing alkali salts also contribute
to accelerated subcritical crack growth of monolithic SiC (Fig. 5). The crack

398

Jones et al.

Fig. 5 Crack velocity versus stress intensity for monolithic SiC exposed to air and hotgas environments. (From Ref. 10.)

velocity in the hot-combustion gas was several orders of magnitude faster at a


50 K lower temperature than in air. The primary mechanism for this accelerated
crack velocity is penetration of alkali ions into the grain-boundary glass phase
and the reduction in the viscosity of this phase. The reduced viscosity causes
increased crack tip creep rates and creep damage.
B. Stability of SiC in Mildly Oxidizing Environments
Spear et al. (11) measured the oxidation rate of hexagonal -SiC platelets with
both (0001) C and (0001) Si faces. The only solid corrosion product observed
was SiO2 [Eq. (1b)], with CO(g) formed at the SiCSiO2 interface. The O2 pressure was 103 to 1 atm at 12001500C, which is above the activepassive transition; thus, the presence of SiO(g) would not be expected. An activation energy
of 120 kJ/mol was found for SiC, which is very comparable to the value of 112

Ceramic Matrix Composites

399

kJ/mol found for Si. Silicon nitride has a much higher activation energy of 486
kJ/mol because of the formation of a silicon oxynitride phase between the Si3N4
and the SiO2 phases. A growth rate of 1.1 nm/min was determined for SiC at
1300C in 1 atm of O2. Oxygen diffusion was postulated as the rate limiting
process up to 1350C, above which ionic oxygen diffusion was rate controlling.
The protective properties of the passive SiO2 film are very sensitive to the
impurities of the SiC from which the film is formed. Comparison between the
oxidation behavior of several batches of SiC to that of very high-purity SiC produced by chemical vapor deposition was made by Fergus and Worrell (12). They
observed an activation energy of 142 kJ/mol for CVD SiC and 217 kJ/mol for
sintered -SiC. This compares with the value of 120 kJ/mol reported by Spear
et al. for single-crystal SiC platelets. Both the CVD and SiC platelets are of much
higher purity than the sintered -SiC to which sintering aids were added. Vaughn
and Maahs (13) in a review of the activepassive transition for SiC showed that
this transition was quite variable depending on the source of the SiC and other
experimental conditions such as the gas flow velocity. The Gulbransen data (Fig.
1) was very close to that determined theoretically, but most other materials exhibited much higher transition temperatures, presumably due in part to variable impurity concentrations in the materials.
Results of oxidation studies of ceramic composites have not been reported,
although Luthra (14) has presented some theoretical concepts on the oxidation of
ceramic composites. Luthra assessed the gas-phase diffusion of oxygen through
microcracks, solid-state diffusion of oxygen through protective oxide layers, formation of gaseous reaction products, and the combined reaction of the matrix
reinforcement and oxygen. In particular, the diffusion of oxygen through microcracks to react with the reinforcing fiber or the interface between the fiber and
matrix is a definite possibility for SiCSiC composites produced by the CVI
process because they have at least 10% residual porosity. Thermal cycling is also
likely to induce matrix microcracking because of the thermal expansion mismatch
between the reinforcing fiber and the matrix. For SiC fibers in a CVI SiC matrix,
this mismatch will be small, but it is not zero. Luthras analysis is primarily for
oxide matrix composites, where the oxygen will not react with the matrix,
whereas for SiC where the oxygen could react with the matrix, the transport
would be considerably slower.
The structural behavior of CMCs in an inert gas environment will be dependent on (a) whether the material is undergoing active or passive oxidation, (b)
the stability of the fibermatrix interfacial layer in O2, and (c) the stability of the
fiber in O2. Kim and Moorhead (15) have recently evaluated the flexural strength
of -SiC at room temperature following a 10-h exposure to ArO2 at 1400C.
Flexural strengths following exposure to ArO2 mixtures with O2 partial pressures of 7 107 to 2 104 MPa showed a strength reduction with increasing
pO2 up to about 2 105 MPa and a complete restoration of the as-polished

400

Jones et al.

strength at pO2 2 105 MPa (Fig. 6). The minimum strength (60% of unexposed material) at 2 105 MPa was the result of active corrosion creating
strength-reducing flaws along grain boundaries and at pits. Easler et al. (16) found
that the room-temperature flexure strength both increased and decreased upon
exposure to air at 1370C with and without an applied load. The strength increases
were thought to result from oxidative blunting, whereas the decreases which occurred under load were due to subcritical crack growth producing larger flaws.
Minford et al. (17) also found that stress enhanced the uptake of O2 into SiC.
Oxygen penetration occurred along cation-enriched grain boundaries. A stress
intensity threshold of about 1 MPa m1/2, below which crack blunting occurred
and above which crack extension occurred was found.
Oxidation can alter the dynamic crack growth behavior of SiC as well as
reduce the fracture strength due to corrosion-induced flaws. A reduction in the
K th of -SiC after being loaded to different stress intensities for 4 h at 1200C
and 1400C was reported by Minford et al. (18). They reported a reduction in
the K th from 2.25 in the unoxidized to 1.75 MPa m1/2 in the oxidized conditions
at 1200C and from 1.75 to 1.25 MPa m1/2 at 1400C for samples tested in
air or preoxidized. They concluded that the O2 caused a change from diffusioncontrolled crack growth to viscous cavity growth and linkage with a corresponding reduction in K th. The K th was defined as the transition between crack blunting

Fig. 6 Room-temperature flexural strength of sintered -SiC after exposure for 10 h at


1400C in Ar atmospheres with various pO2.

Ceramic Matrix Composites

401

and growth as determined from the fracture strength following the 4-h static loading at various K values. McHenry and Tressler (19) found that the K th and subcritical crack growth rate was independent of the pO2 for pressures ranging from 108
to 104 atm and temperatures of 9001100C. The crack growth rate exhibited
an Arrhenius temperature dependence with an activation energy of 20 kcal/mol,
which is consistent with viscous flow of a grain-boundary glass phase. The lack
of a dependence on pO2 does not necessarily indicate that O2 did not induce crack
growth but that the effect of O2 was saturated at pressures above 108 atm.
C. Oxidizing Environments
A key feature of the oxidation behavior of SiCSiC in O2 at pressures of 2
104 Pa (atmospheric pressure of O2) observed by Windisch et al. (20) is that only
a weight loss was observed. No SiO2 formation occurred in any of the materials
with graphite interphases, although some boron-containing glass phase was observed for the material with a BN interphase. The kinetics of mass loss is shown
as a function of pressure and temperature in Figs. 7 and 8. Complete burnout of
the graphite interphase occurred in less than 104 s in the small test samples at a

Fig. 7 Mass loss versus exposure time for SiCSiC with a C interphase exposed to
various O2 partial pressures at 1100C.

402

Jones et al.

Fig. 8 Mass loss versus time for SiCSiC with a C interphase exposed at 2.5 103 Pa
O2 at various temperatures.

pressure of 2.4 104 Pa and a temperature of 1373 K. It is also clear that the
reaction rate increases with increasing temperature. An activation energy of about
50 kJ/mol was reported by Windisch et al. (20), which could be explained as
diffusion controlled through a boundary layer or as being reaction rate controlled.
An interphase recession rate was determined from the weight-loss measurements
and by direct physical measurement. Both methods gave very similar recession
rate equations with the physically measured equation as follows:
log(RR) 0.9 log(pO2) 9.9

(6)

It should also be noted that only weight loss was observed over the temperature
range 10731373 K, which borders on the temperature range (8731073 K) suggested by Evans et al. (21) for the pest phenomena.
Oxidation of SiCSiC composites with carbon interphases can also result
in the formation of SiO2 and a weight gain following an initial weight loss (22,23)
or a reduced weight loss with increasing temperature (24). Tortorelli et al. (23)
observed an initial weight loss followed by a weight gain in a Nicalon-fiberreinforced SiCSiC composite with a 0.3-m-thick graphite layer exposed to dry
air (pO2 of 2 104 Pa) at 1223 K. Following the initial weight loss from oxidation

Ceramic Matrix Composites

403

of the carbon, they found that SiO2 formation occurred within the interfacial
region previously occupied by the carbon. For the Nicalon-reinforced composite
material, complete carbon depletion occurred within 15 min at 1223 K, followed
by weight gain from SiO2 formation. Unal et al. (24) observed their largest weight
loss (5%) at 1223 K for an exposure of 50 h in dry O2 and a decreasing weight loss
with increasing temperature up to 1673 K (2%). Their material was a Nicalonreinforced SiCSiC with a 0.5-m-thick fibermatrix carbon interphase. Kleykamp et al. (25) observed the following reactions of air with SiC-fiber-reinforced
SiC composites: (a) oxidation of free carbon at temperatures of 800965 K and
(b) a fast exothermic reaction and weight gain beginning at 1073 K and containing
up to 1773 K and up to times of 1 h followed by (c) the diffusion-controlled
oxidation of bulk SiC. Sebire-Lhermitte et al. (26) identified the presence and
location of SiO2 formation in SiCSiC composites using transmission electron
microscopy. They noted the presence of 15-nm-thick SiO2 layers at both the fibercarbon and matrixcarbon interfaces following an exposure of 1 h at 1123
K in air.
That Windisch et al. (20) only observed a weight loss while others (22
24) observed a weight gain following the weight loss could be the result of lower
O2 pressures, exposure time, and perhaps, carbon-layer thickness. The lower O2
pressure would reduce the SiO2 formation rate and therefore the chance for a
measurable weight gain during the 5-h exposures. Tortorelli et al. (22,23) used
exposures of up to 150 h, whereas Unal et al. (24) used 50-h exposures. Measurable SiO2 formation at pO2 2 103 Pa would need a much greater exposure
time than that used by Windisch et al. (20) and even greater than the time used
by Tortorelli and More (22). The existence of subcritical crack growth, as described in the next section, that coincides with only weight loss or interphase
removal without the embrittling effect of SiO2 or other solid-product formation
is the primary difference between an interphase removal mechanism (IRM) and
oxidation embrittlement mechanism (OEM) of crack growth. The oxidation results of Windisch et al. (20) demonstrate that the results of Henager et al. (27,28)
at temperatures ranging from 1073 to 1473 K and pO2 2 103 Pa occurred
by IRM only. The OEM is dependent on SiO2 formation, which depends on O2
pressure, temperature, and time. The interfacial layer thickness may also impact
this regime if a solid product can seal off the interface from further reaction. A
clear demonstration of this possibility has not been presented but remains as an
open issue needing further evaluation.
The effect of oxygen on the subcritical crack growth velocity of SiCSiC
is clearly demonstrated by the data given in Fig. 9. Oxygen has little effect on
the midpoint displacement (i.e., crack velocity) for about 2 104 s, but a marked
increase in the crack velocity is noted for longer times. These tests were performed in the O2 pressure, temperature, and time regime where only weight loss
was observed during oxidation studies (20). Therefore, the embrittling effect of

404

Jones et al.

Fig. 9 Midpoint displacement of a single-edge notch-beam specimen of SiCSiC with


a C interphase in gettered argon and 3.1 102Pa of O2 Ar at 1200C.

a solid reaction product should not be a factor, but only the effect of fiber creep
and interfacial removal should have contributed to the crack growth rate. However, even if SiO2 or other glassy phases were present, they would have low
viscosity at very high temperatures and would not likely affect the crack growth
behavior or cause brittle fracture.
The dependence of the crack velocity on O2 partial pressure up to 10% that
of air is given in Fig. 10 for tests at 1373 K. There is a sharp increase in the
crack velocity at low pressure and a slower increase from pressures of about 0.25
102 Pa up to 2.5 103 Pa. Material with a BN interface exhibited about a
factor-of-10 slower crack velocity. Some glass-phase formation was noted in this
material as a result of these exposures, but there was no evidence that the crack
growth behavior was affected by the presence of this glass phase.
The IRM has been observed to cause crack growth in SiCSiC at temperatures of 10731473 K at pressures ranging from 2 102 to 2 103 Pa for stressed
samples (29,30). Weight-loss measurements suggest that the IRM operates over
these same temperatures and pressures and at 1373 K and a pressure of 2 104
Pa (20). An example of the crack velocity versus crack length for tests on specimens with Hi-Nicalon-reinforced material tested in gettered Ar and Ar 2
102 Pa of O2 is given in Fig. 11. The acceleration in the crack velocity induced
by the presence of O2 is clearly shown by these data, whereas the effect of temperature on the Ar O2 test is only apparent after a crack extension of 3.5 mm.

Ceramic Matrix Composites

405

Fig. 10 Minimum crack velocity versus O2 partial pressure for SiCSiC with a C or
BN interphase exposed at 1100C.

Fig. 11 Crack velocity versus crack length for SiCSiC with a C interphase exposed
to 202 Pa of O2 Ar (dashed curves) or gettered Ar (solid curves) at 1175C (1448 K)
and 1200C (1473 K).

406

Jones et al.

Dynamic (sample stressed during exposure) OEM has been observed (31
33) to occur in SiCSiC at temperatures up to 1073 K, and 1223 K in air (O2
pressure of 2 104), whereas static (sample unstressed during exposure) OEM
was also observed in room-temperature tests following elevated-temperature exposures at 1223 K in air (23) and up to 1673 K in dry O2 at a pressure of 105
Pa (24). Heredia et al. (31) reported an upper pest temperature of 1073 K for
SiCSiC tested at elevated temperature in air, whereas Lin and Becher (32) and
Raghuraman et al. (33) observed OEM in air at 1223 K. Tortorelli et al. (23) and
Unal et al. (24) found the OEM to operate in room-temperature tests following
elevated-temperature exposure tests conducted in air without the application of
stress. In summary, the upper temperature limit for the dynamic operation of
OEM is between 1073 and 1223 K, and the formation of a glass phase at temperatures greater than 1223 K can cause OEM to occur when specimens are tested
at lower temperatures. Because the OEM results from the formation of a brittle
glass phase, this mechanism must depend on the growth rate and viscosity of the
glass phase. The growth rate will increase with increasing temperature and pO2,
but the viscosity decreases with increasing temperature. Therefore, there must
be a temperature at which the effectiveness of OEM is maximum.
The results of Heredia et al. (31), Lin and Becher (32), and Raghuraman
et al. (33) appear to define the upper temperature and O2 pressure limits for OEM
in SiCSiC at 10731223 K and O2 pressures of 2 104 Pa and above. However,
the IRM appears to operate over temperatures of at least 10731473 K at O2
pressures of 2 103 Pa and below. It may also operate at temperatures below
1073 K, within the OEM range, at low pressures, but this has not been observed
because the crack velocities are too low for experimental measurements. Bouchetou et al. (34) and Fr