You are on page 1of 460

OXIDATION OF METALS

OXIDATION OF METALS
Karl Hauffe

Based on the German edition of


Oxydation von
Metallen und Metallegierungen

PLENUM PRESS
New York
1965
Translated by Karl Vorres from the 1956 German edition,
subsequently revised by the author. This edition produced
by special arrangement with the original publisher, Springer·
Verlag OHG, Berlin/Gottingen 1Heidelberg.

ISBN 978-1-4684-8922-4 ISBN 978-1-4684-8920-0 (eBook)


DOl 10.1007/978-1-4684-8920-0

Library of Congress Catalog Card Number 63·17648

©1965 Plenum Press


Softcover reprint of the hardcover 1st edition 1965
A Division of Consultants Bureau Enterprises, Inc.
227 West 17th Street· New York, N. Y.l0011
All rights reserved
No part of this publication may be rep1·oduced in any
form lcithout written pennission from the publisher
Preface to the English Edition

During the translation, the author had the opportunity to re-


view several chapters, taking into consideration the more recent
literature. As far as possible all new theoretical concepts and experi-
mental data published before 1963 have been quoted and discussed
under the theoretical viewpoint of this book.
A new chapter "Passivity and Inhibition During High-Tempera-
ture Oxidation" was introduced. Section 4.8 was enlarged by a dis-
cussion of the transition from internal to external oxidation.
The author very much appreciates the cooperation of the trans-
lator and of Plenum Press.
Gottingen, April 1.965

Karl Hauffe

v
Preface

The number of publications concerned with oxidation and cor-


rosion processes has become so copious that many engineers and
scientists find it practically impossible to obtain an overall view of
the growing body of knowledge and to bring order to the confusing
multiplicity of experimental data. As a result the need for a compre-
hensive survey of the present state of research in this field has be-
come more and more urgent.
It seemed particularly useful to start with a critical review of
structural defects in scaling layers formed in the process of metal
oxidation, since it is these defects in the protective layers that deter-
mine the mechanism and rate of oxidation, at least in those cases
where diffusion or transport processes are rate-determining. This
approach leads to a classification of oxidation processes from the
point of view of the theory of defects in the layers of reaction prod-
ucts and of the associated kinetics, in place of the more familiar
chemical classification as oxidation, sulfidation, halogenation, etc.
Since this classification is applicable only when the rate of oxidation
is not determined by a phase-boundary reaction, it has to be sub-
ordinated to a more general one, namely, that of oxidation processes
controlled on the one hand by transport phenomena and on the other
by phase boundary reactions.
The fundamentally recurring observation of the appearance of
various rate laws (e.g., parabolic, cubic, logarithmic, and reciprocal-
logarithmic) in various temperature regions can usually be explained
in terms of the generalized oxidation theory. Thus, in the high-tem-
perature region, where generally diffusion processes are rate-deter-
mining, we find thick, compact protective layers, which we call
scaling layers. Here, then, the metal oxidation is called a scaling
process (Wagner's scaling theory). However, in the middle- and
low-temperature regions, other phenomena, such as electric fields and
space charge effects, enter into the picture and give rise to a different
oxidation mechanism, which is much more difficult to describe than
vii
viii Preface

that at high temperatures. Here, in contrast to high-temperature


oxidation, we find thin protective layers, which we call tarnishing
layers. The oxidation process at lower temperatures may be called
a tarnishing process.
In the oxidation of alloys, in addition to the above-mentioned
processes we find the diffusion of the component metals and of the
attacking gas in the alloy and reaction product phases, which gives
rise to such supplementary phenomena as selective and internal oxi-
dation. Precisely these two are frequently the decisive mechanisms
in the oxidation of alloys used in practical applications. For this
reason, a separate chapter is devoted to the discussion of each of these
phenomena, and all the pertinent diffusion data are summarized in
tabular form. Next, an attempt is made to fit the mechanism for the
formation of passive layers and for the corrosion of metals with
passive layers into the general framework of oxidation processes.
Finally, a brief description of measuring methods is intended to
furnish the reader with an insight into at least the more easily ex-
ecuted experiments.
The author feels that a detailed treatment of the basic, generally
valid elementary processes of metal oxidation is of greater value than
a comprehensive reporting of the mass of published results. For this
reason many reports on the oxidation of alloys of industrial interest
have not been cited. However, it is hoped that this weakness is
counterbalanced by the fundamental nature of the discussion of the
possible reactions leading to the formation of protective layers on
metals. After all, this book is intended to stimulate professional col-
leagues employed in industry to further significant experimentation.
Since the author has been active in this field for a number of
years, it is inevitable that several of his own experimental results
and deliberations have found their way into this book.
Thanks to the cooperation of many colleagues in Germany and
abroad who so kindly sent reprints, preprints, and personal notes,
many otherwise inaccessible results and suggestions fitting into the
scope of this book could be taken into consideration.
Most of the work on this volume was completed while the author
was employed at the Central Institute for Industrial Research of the
Norwegian Research Council in Oslo. The author extends his heart-
felt thanks to the director of the Institute, Mr. Alf Sanengen, for his
generous support.
Frankfurt arn Main, May 1956

Karl Hauffe
Contents

1. Introduction to the Reactions Between Oxidizing Gases and Metals


and Alloys . 1
2. Lattice Defect Phenomena and Diffusion Processes in Ionic, Cova-
lent, and Metallic Crystals. 8
2.1. Lattice Defect Phenomena in Stoichiometric Ionic Crystals. 9
2.2. Defect Phenomena in Nonstoichiometric Ionic Crystals. 14
2.3. Lattice Defect Phenomena in Regions Near the Surface of
Nonstoichiometric Ionic Crystals. 25
2.4. Lattice Defect Phenomena and Diffusion Mechanisms in Metals 28
2.4.1. Vacancy Diffusion in Metals and Alloys . 32
2.4.2. Interstitial Diffusion in Metals . 34
2.4.3. Summary of Some Diffusion Data in Metals and Ionic
Crystals Which are Important for the Oxidation of Alloys 36
2.4.4. Summary of Self-Diffusion Coefficients in Metals and In-
organic Solid Materials. 37
2.4.5. Grain-Boundary and Surface Diffusion in Metal and
Ionic Crystals . 38
3. The Mechanism of Oxidation of Metals - Theory. 79
3.1. Diffusion and Transport Processes in Scaling and Tarnishing
Layers. 82
3.2. The Wagner Theory of Oxidation 87
3.3. Diffusion and Scaling Coefficients 91
3.4. Calculation of the Absolute Oxidation Rate Constants of Metals
for the Parabolic Rate Law . 94
3.5. The Effect of Electric Fields on Metal Oxidation . 97
3.5.1. Space-Charge Layers at Phase Boundaries. 97
3.5.2. Transport Processes in Space-Charge Layers 102
3.5.3. Formation of Thin Tarnishing Layers - The Parabolic
and Cubic Rate Law Caused by Participation of Field
Transports . 109

ix
x Contents

3.5.4. Formation of Very Thin Tarnishing Layers - The Log-


arithmic and Reciproc-Logarithmic Rate Laws. 125
3.5.5. The Nonparabolic Tarnishing Law as a Result of Block-
ing and Cavity Formation at the Metal/Oxide Phase
Boundary 137
4. Scaling Processes in Metals and Alloys with Formation of Thick
Protective Layers. 144
4.1. Scaling Systems with Ion-Conducting Protective Layers. 148
4.2. Scaling Systems with Electron-Conducting Protective Layers 158
4.2.1. Scaling Systems with p-Type Conducting Protective
Layers. 159
4.2.1.1. The Rate of Oxidation of Copper and Copper
Alloys . 159
4.2.1.2. The Rate of Oxidation of Nickel and Nickel
Alloys . . . . . 171
4.2.1.3. The Rate of Oxidation of Other Metals and Alloys
with p-Type Conducting Protective Layers. 193
4.2.2. Scaling Systems with n-Type Conducting Protective
Layers. 201
4.2.2.1. Rate of Oxidation of Zinc and Zinc Alloys. 202
4.2.2.2. The Rate of Oxidation of Titanium and Titan-
ium Alloys 209
4.2.2.3. The Oxidation Rate of Zirconium. 228
4.2.2.4. The Oxidation Rate of Niobium . 236
4.2.2.5. Other Scaling Systems with Probable n-Conduct-
ing Protective Layers. 239
4.3. Catastrophic Oxidation . 250
4.4. Scaling Systems with Rate-Determining Phase-Boundary Re-
actions . 253
4.5. Scaling Systems with Protective Layers Containing Several
Phases. 267
4.5.1. Joint Parabolic and Linear Growth of Two Adjacent
Oxide Layers . 288
4.5.2. Passivity and Inhibition During High-Temperature Oxi-
dation. 293
4.6. Scaling of Iron Alloys 296
4.7. The Influence of Metal Diffusion in the Alloy Phase on the
Scaling Rate. 314
Contents xi

4.7.1. Formation of a Smooth Alloy/Scale Phase Boundary. 315


4.7.2. Formation of a Rugged Alloy/Scale Phase Boundary. 329
4.7.3. The Mechanism of the Formation of an Oxide Layer Con-
sisting of Two Oxides . 331
4.8. The Mechanism of Internal Oxidation of Alloys. 335
5. The Mechanism of the Attack of Sulfur and Sulfur Compounds on
Metals and Alloys. 365
6. The Oxidation Mechanism of Metal-Carbon Alloys and Carbides. 392
7. The Mechanism of Oxide Layer Formation in Aqueous Electrolytes 402
7.1. The Phenomenon of Passivity. 404
7.2. The Mechanism of Passive Layer Formation on Metals and
Alloys . 409
7.2.1. The Theory of Passive Layer Formation on Metals 410
7.2.2. Formation and Structure of the Passive Layer. 417
7.3. The Solution Current of Metals with Passive Layers. 424
8. A Few Approved Methods of Measurement of Coating Growth. 431
8.1. Use of the Microbalance in Oxidation Apparatus. 431
8.2. Gas Volumetric and Manometric Methods for Measurement of
the Oxidation Rate. 435
8.3. Further Methods for the Measurement of the Thickness of
Tarnishing Layers . 438
Author Index . 441
Subject Index . 450
1. Introduction to the Reactions Between
Oxidizing Gases and Metals and Alloys

The tremendous chemical and technological progress made by industry


in the past few years has imposed increasingly greater demands on the
mechanical and chemical properties of metallic materials. In what follows,
we will be concerned with chemical aspects of the problem, and primarily
with only one-of course very important-part of this field: the oxidation
and scaling resistance of the metallic materials. The terms metal oxidation,
tarnishing, and scaling will be used whenever oxidizing gases such as oxygen,
sulfur, the halogens, or water vapor attack a metal or an alloy-either at
low or high temperatures-i.e., when a chemical reaction takes place.
Oxidative attack upon metals can take place under the most varied con-
ditions-from the "mild" oxidizing conditions which exist in air at room
temperature to the "severe" conditions present during reactions of hot
furnace gases on metallic structural elements. Especially stringent require-
ments are placed upon the scaling resistance of metallic materials used for
chemical apparatus in high-temperature reactions, which frequently proceed
under high pressures, as well as on the materials used for the construction
of gas turbines (turbine blades), hot air motors, and flaming-gas jet-pro-
pulsion systems, and those employed in the construction of high-pressure
steam boilers, where, for reasons of economy, low-alloy ferritic steels are
used. It is the last example which poses the greatest demands on the scaling
resistance.
Industrial materials without sufficient scaling resistance frequently
fail even in a short time as a result of rapid oxidation, which owing to poor
adhesion of the oxide layers, is also often accompanied by spaUing of the
scale. As a result, the permissible limits of wear are often exceeded, and
expensive apparatus and machinery parts have to be replaced prematurely.
It is obvious, then, that the metal industry makes every effort, at a cost of
millions, to meet the demands, not simply for heat-resistant alloys but for
materials which also possess the mechanical properties required at high
working temperatures.
Forty years ago, when the metal industry started to become concerned
1
2 1. Introduction

with such development problems, neither complete thermodynamic data on


the oxidation reactions at various temperatures nor the necessary kinetic
data, or even useful working hypotheses, were available to serve as a basis
for a program to develop scaling-resistant alloys. In view of the scarcity of
data at that time, it is not surprising that the investigators initially pro-
ceeded with the solution of these problems in a purely empirical manner,
but it was not until they gradually determined the most important thermo-
dynamic data on the oxides, sulfides, halogens, etc. and discovered the rate
law of the oxidation reaction and its dependence on conditions of temperature,
gas composition of the attacking atmosphere, and pretreatment of the
materials that they succeeded in developing scaling-resistant alloys.
Available thermodynamic data, such as the heat of formation of the
oxidation product combined with the Arrhenius equation (logarithm of the
rate proportional to the activation enthalpy), served as the basis for a working
hypothesis, in terms of a classical model, for the investigators in the metal
industry. However, since the kinetics remained rather vague owing to
ignorance of the true reaction mechanism, not just for alloys but even for
pure metals, clearly the then current working hypotheses could not possibly

I
explain the true state of affairs. This becomes obvious when we examine,
for example, the following chemical reaction equations:

Ni+ iO(~)...-""NiO
Ni + i S(~)...-""NiS (1.1)
Ag + t Br(~)...-"" AgBr
The formation of a metal oxide, sulfide, or halide as a result of the action
of oxygen, sulfur vapor, or halogen gas on a metal at elevated temperatures
appears to be one of the simplest reactions, but this is not really so. The
reaction product frequently appears as a compact phase with the reacting
substances spatially separated from each other, and further reaction is then
possible only if at least one of the reactants diffuses through the scaling
layer to the other reaction partner. In such cases, the course of the reaction
is no longer determined by the overall chemical process described by equation
(1.1), but by diffusion processes and phase-boundary reactions, for which
the mechanism, as we will show in detail, can be quite complex. Generally,
one must always consider the several partial processes listed below. Of these
processes, one is the slowest and therefore the rate-determining step:

1. Phase-boundary reactions (chemisorption of the nonmetal molecules


with simultaneous electron exchange and splitting of the molecules
at the gas/oxide interface and transfer of the metal from the metallic
phase, in the form of ions and electrons, to the scale at the metal/oxide
1. Reactions Between Oxidizin~ Gases and Metals and Alloys 3

interface with further reaction of the individual reactants and


formation of the reaction products), nucleation, and crystal growth.
2. Diffusion or transport of cations, anions, and electrons through the
scale complicated by a special migration mechanism because of the
appearance of chemical and electrical potential gradients in the
scaling or tarnishing layer.
3. Predominant transport processes in space-charge boundary layers
in the case of thin tarnishing layers, especially at low temperatures.

Two further factors are also significant for the formation, composition,
and structure of the scaling layer:

4. The thermodynamic stability of the oxide that is formed.


5. The crystal structure of the scaling layer and of the metal or the
alloy, which determines the adhesion between the scaling layer and
the underlying metal.

On bases of qualitative considerations it is not surprising that rate


laws observed in the oxidation of metals and alloys may take various forms.
At high temperatures and high gas pressures we often find a parabolic rate
law of the form

(1.2)

where g is the thickness of the scale, k' the parabolic scaling constant, and t
the oxidation time. On the other hand, at intermediate temperatures, a
cubic rate law is often approached, which has the form

(1.3)

Furthermore, at lower temperatures we find a normal and a reciprocallogarith-


mic rate law of oxidation, which may be put in the following integrated forms:

~ = ~o In(t + to) - const (1.4)


and
1/~=A - Blnt (1.5)

Finally, there is a linear rate law, which in its integrated form reads

~ =kt (1.6)

Here, the thickness of the layer is directly proportional to the experimental


4 1. Introduction

time. The linear rate law is always observed when the diffusion and transport
processes are sufficiently rapid.
The problem now is to match the observed rate law to a reasonable
mechanism and to discover which of the steps is rate-determining. We shall
not be in a position to avoid laborious empirical attempts to find practical
methods of slowing down scaling rates in metals and alloys until we isolate
the step that determines the overall course of the oxidation. As noted before,
the reactants (e.g., oxygen and the metal) are spatially separated from one
another by the compact reaction product formed on the metal, so unless the
reactants-or at least one of them-can diffuse through the scaling layer,
no further reaction is possible.
In general, the thick oxide layers which form on the metal during
oxidation are adherent and pore-free if the criterion formulated by Pilling
and Bedworth,l is obeyed, namely, that the ratio of the molar volume of
the reaction product to that of the metal is greater than one. However,
a more exact approach indicates that this volume quotient is probably
important, but not decisive. 2 Schottky was recently able to show that the
essential requirement for the formation of a compact, adherent scale is
a capability for plastic flow of both the metal and the oxide. While the
contraction mechanism in the metal phase by a vacancy precipitation at
dislocations does not involve any theoretical difficulties, the appearance of
a contraction in the oxide phase is understandable only if oxygen ions as well
as metal ions diffuse. Then, according to Schottky, the oxygen ion transport
can be several orders of magnitude smaller than the metal ion transport
required for the growth of the layer. 3
The question may now be raised as to the form in which the reactants
enter the lattice of the scale and the manner in which these species diffuse
through the lattice. Obviously, penetration of the reactants into the scale
starts with a chemisorption process. However, as we will show later, the
mechanism of this initial reaction can be quite complicated. This is under-
standable if one considers that the chemisorption of a gas molecule proceeds
by electron transfer to the metal or the oxide layer and simultaneous
dissociation of the molecule into atoms. The subsequent infiltration of this
chemisorbed atom into the scale is possible only if some ofthe lattice positions
in the scale are vacant, or if the nature of the lattice structure is such as
to permit occupation of interstitial lattice positions or displacement of cations
from their regular lattice positions to the chemisorbent. Thus, a nonideal

1 Pilling, N. B., and R. E. Bedworth: .f. Inst. ""Ietals 29, 529 (1923).
2 Jaenicke, W.: (in Passivierungs- u. Anlaujvorgange an ltIetalioberflachen, edited by
H. Fischer, K. Hauffe, and W. 'Wiederholt, Springer, Berlin/Gi.ittingen/Heidelberg,
1956) was able to show that additional conditions are required.
3 Schottky, W.: Z. Elektrochem. 63, 784 (1959).
1. Reactions Between Oxidizing Gases and Metals and Alloys 5

ordering of the crystal lattice is a prerequisite for penetration into the lattice
and diffusion through the scaling layer. For a comprehensive understanding
of the oxidation processes of metals, it is necessary to clarify the nature and
the extent of lattice defects in real ionic and covalent crystals, as well as the
possible mechanisms for diffusion. With such oxidation experiments with
metals and alloys, simultaneous investigations should be conducted to
determine the relationships between the nature of the lattice defect in the
scaling layer that is obtained (e.g., on a simple oxide) and the conditions of
temperature, gas composition, and foreign-ion additives. Furthermore, if one
now makes use of the working hypothesis formulated by Wagner! that the
species which migrate through the lattice ofthe reaction product, i.e., through
the scale, are not metal and nonmetal atoms but rather metal ions and
electrons or anions and electrons, respectively, then the first problem appears
to be limited to the study of the mechanism by which these particles migrate.
This migration mechanism, however, must be closely related to the defects
in the lattice of the scale and the difference between the chemical potentials
at the scale/gas and scale/metal phase boundaries. It is well known from
the chemical physics of semiconductors2 that the presence of an electric
field increases the mobility via lattice defects of ions and electrons. Thus
measurements of electrical conductivity, the Hall effect, and thermoelectric
power give indirect information on the kind and extent of the lattice defects
present and allow conclusions to be drawn about the kind and extent of
migration processes of ions and electrons in scaling layers.
On basis of the previous considerations, it is apparent that there is
no direct connection between the oxidation rate of a metal and the magnitude
of the negative free energy of formation L1F. Thus, for example, the oxidation
rate of aluminum with the scaling constant k" (in g2Jcm4 -hr) at 600°C equal
to about 3 x 10-11 is several orders of magnitude less than that of copper
with the oxidation constant k" equal to 1.1 x 10-6 under the same experi-
mental conditions, although the free energy of formation at 600°C for
aluminum oxide (L1FAI,O. = - 220 kcalJmole of oxygen) is. considerably
larger than that for CU20 (L1F cu•o = - 55 kcalJmole of oxygen).3 As we
shall later show in detail, the reason for the variant behavior of the oxidation
rates is the difference in the natures of the defects in the two oxide lattices.
While Cu 20 in the presence of oxygen shows a considerable concentration
of copper ion vacancies and holes (equal to the concentration of divalent
copper ions), which constitutes a good prerequisite for a rapid diffusion of
the copper ions and electrons, the Al 20 3 lattice exhibits only a small lattice
1 Wagner, C.: Z. physik. Chem. (B) 21, 25 (1933); Angew. Chem. 49. 737 (1936).
2 See for example K. Hauffe: Ergeb. exakt. Naturw. 25. 193ff (1951).
3 See for example C. W. Dannatt and H. J. T. Ellingham: Discussions Faraday Soc.
4. 126 (1948).
6 1. Introduction

defect concentration at high temperatures and low oxygen pressures, so that


considerably fewer paths for diffusion are available. Moreover, the energy
barriers, or potential barriers, which must also be considered in a quantitative
approach to diffusion, are evidently especially high in Al 20a.
The approach in this introductory work must be to first delineate the
oxidation mechanism of pure metals, because not until this problem is solved
can one predict the influence of alloying additives on the mechanism and
the rate of the oxidation process. For this reason, in the following chapters
we must be concerned first with the lattice defect phenomena in ionic and
covalent crystals, where we make use of the results of the Wagner-Schottky
theory! of lattice defects, which were later extended by Wagner,2 Verwey,a
and Hauffe 4 to heterotype semiconducting solid solutions. Since for various
problems, metal diffusion in the alloy phase assumes a significance apart
from participation in the diffusion process in the scaling layer, we will also
have to consider the lattice defect phenomena and diffusion processes in
metal and alloy systems. A general presentation of the diffusion and trans-
port processes in ionic and covalent crystals is given which includes the
scaling and tarnishing laws (1.2) to (1.6). Although Wagner's theory of
metal oxidation is included in the general presentation, we will treat it later
in greater detail and illustrate its application. The Wagner theory enjoys
special treatment not merely for historical reasons-Wagner 5 was the first
to give a complete theory of the parabolic oxidation law equation (1.2),
experimentally observed by Tammann6-but also because many high-
temperature oxidation processes in metals and alloys can be described by
this formulation, which Wagner 7 himself later extended to include oxidation
systems in which metal diffusion in the alloy phase plays an additional role.
It should be emphasized now, at the outset, that the Wagner theory is
only applicable when diffusion processes alone are rate.determining, which
is generally the case at high temperatures. Furthermore, it assumes that cation
and anion diffusion are equally likely, which in the more recent literature
is not always considered and understood. 8 The Wagner theory is not applicable
in oxidation processes at low temperatures; neither is it valid for special

1 Wagner, C., and W. Schottky: Z. phY8ik. Ohern. (B) 11, 163 (1930). C. Wagner: Z.
phY8ik. Ohern. (Bodenstein-Festband) 177, (1931); (B) 22,181 (1933).
2 Wagner, C.: J. Ohern. PhY8. 18, 62 (1950).
3 Verwey, E. J. W., P. W. Haayman, and F. C. Romeyn: Ohern. Weekblad 44,705 (1948).
4 Hauffe, K.: Ann. PhY8ik. (6) 8,201 (1950); "Fehlordnungserscheinungen und Leit-
ungsvorgange in ionen- und elektronenleitenden fest en Stoffen," in Ergeb. exakt. Naturw.
25, 193 (1951).
5 Wagner, C.: Z. phY8ik. Ohern. (B) 21, 25 (1933); 32, 447 (1936).
6 Tammann, G.: Z. anorg. u. allgem. Ohern. 111, 78 (1920).
7 Wagner, C.: J. Electrochem. Soc. 99, 369 (1952).
8 Lindner, B. R.: J. Ohern. PhY8. 23, 410 (1955).
1. Reactions Between Oxidizinl1 Gases and Metals and Alloys 7

cases at higher temperatures. His theory is not generally applicable in cases


where the chemical potential gradients of the lattice defects or of the corre-
sponding ions and electrons are not the sole rate-determining factors, i.e.,
where electric boundary layer fields caused by space charges-appearing pre-
dominantly in thin layers-also become responsible for mass transport
through the tarnishing layer. In these cases the Cabrera-Mott theory of
metal oxidation has been found to be valid,l and has been presented by
Engell, Hauffe, and Ilschner2,3 in a somewhat extended and modified form.
At lower temperatures, thin to very thin layer formation frequently takes
place. After attaining a critical thickness, the thin layer practically stops
growing. In order to distinguish between the two oxidation regions in the
following discussion, we will designate oxidation processes at high tempera-
tures, where thick oxide layers generally appear, as scaling processes, and
oxidation processes at lower temperatures, where thin oxide layers are
formed, as tarnishing processes.
The reader should be cautioned at the outset against too optimistic an
evaluation of the theoretical discussions in regard to their application for
the solution of technological problems. Some gratifying partial successes
notwithstanding, the current course of research has not yet led us to a
complete understanding of the oxidation mechanisms of all the scaling-
resistant alloys used in industry. In spite of considerable effort, we are still
at the beginning stages of development. Many questions about the formation
of scaling and tarnishing layers still remain unanswered; nevertheless,
the partial results which we have obtained through the use of the Wagner-
Schottky lattice defect theory and through the theory of migration processes
-caused by chemical potential gradients and electrical fields in such scales-
are not to be overlooked. These works form the basis for the explanation
of oxidation mechanisms and the development of scaling-resistant alloys.
In the following sections we will describe experimental results with
oxidation processes in metals and alloys at greater length, as far as possible
in terms of the mechanisms sketched above, in the hope of stimulating
further significant experiments.
1 Cabrera, N., and N. F. Mott: Repts. Progr. in Phys. 12, 163 (1949).
2 Hauffe, K., and B. Ilschner: Z. Elektrochem. 58, 478 (1954).
3 Hauffe, K.: Reaktionen in und an festen StojJen, Springer, p. 547jJ, BerlinfGiittillgenf
Heidelberg, 1955.
2. Lattice Defect Phenomena and Diffusion
Processes in Ionic, Covalent, and
Metallic Crystals

Chemists have known for a long time that many inorganic chemical
compounds, e.g., the oxides (CuzO, FeO, NiO, etc.) and sulfides (CuzS, AgzS,
NiS, etc.) and intermetallic phases in alloy systems do not have a stoichio-
metric composition, but rather exhibit a more-or-less large excess or deficit
of one or another of the constituent components of the crystal. In such
cases, the fact that the crystal lattice is not ideally filled is easily under-
standable. However, there are also compounds with stoichiometric composi-
tion, e.g., the alkali and silver halides, which can exhibit considerable disorder
in their lattice site occupation, as we shall see later. It can be said, in general,
that these lattice building blocks (atoms and ions) leave their lattice sites
with increasing frequency as the temperature increases and go either to an
interstitial lattice site or, if this is not possible for spatial energetic reasons,
"break out" to the surface and leave behind an unoccupied lattice site,
which serves again as an empty site, which can be occupied, in turn, by
particles lying deeper in the interior. Thus, the unoccupied lattice site
(vacancy) moves toward the interior, while the ions migrate to the surface.
A condition of equilibrium is obtained when the particle and vacancy
currents, which are in opposite directions, are of equal magnitude. At
constant pressure or volume and constant crystal composition, the concen-
tration of the lattice defects is determined solely by the temperature. In
succeeding sections, we group the above events following Frenkel,! Jost,Z
and Schottky and Wagner 3 ,4 under the term lattice defect phenomena.

1 Frenkel, J.: Z. Physik. 35, 652 (1926).


2 Jost, W.: J. Chem. Phys. 1,466 (1933); Trans. Faraday Soc. 34,860 (1938).
3 Wagner, C., and W. Schottky: Z. physik. Chem. (B) 11, 163 (1930).
4 'Wagner, C.: Z. physik. Chem. (Bodenstein-Festband) 177, (1931); (B) 22, 181 (1933).

8
2.1. In Stoichiometric Ionic Crystals 9

2.1. Lattice Defect Phenomena in Stoichiometric Ionic (rystals


Following the historical development, we now consider the mechanism
for the formation of the ionic lattice defects in stoichiometric ionic crystals.
An ionic defect in this case may appear in such a manner as to affect pre-
dominantly only the cation or anion sublattice, or it may affect both the
anion and cation sublattices to the same extent. According to Schottky,!
we have to distinguish among four types of lattice defects in stoichiometric
crystals:
Limiting Type I: Cations in interstitial lattice sites and vacancies in
the cation sublattice (Frenkel type-Fig. 1).
Limiting Type II: Anions in interstitial lattice sites and vacancies in
the anion sublattice (anti-Frenkel type).
Limiting Type III: Cations and anions in interstitial lattice sites (anti-
Schottky type).
Limiting Type IV: Vacancies in cation and anion sublattice sites
(Schottky type-Fig. 2).

Br- Br- Ag+ p;.;r-


Ag+ Br- Ag+ Ag+
Ag"-
Br- Ag+ Br- Ag+ Br- O/Br- Ag+
Ag+ Br- Agt Br- Agt Br- Ag+ Br-
Fig. 1. Lattice defect model according to
Frenkel (Limiting Type I) presented for the Br- Ag~Br- Agt Br- Ag+ Br- Agt
example of silver bromide. For this case, Agt Br- 0 Br- Agt BrAg.'Agt,.. 5r-
only silver ions can move via interstitial
lattice positions or vacancies. Br- Agt Br- Agt Br- 'Agt Br- Ag'

Nat CI- Na~- Na+ CI- Na+ CI- Na+


Fig. 2. Lattice defect model according CI- Nat CI- CI- Nat CI- Nat CI-
to Schottky (Limiting Type IV) for Nat CI- Na+ CI- Nat 0 Nat CI- Nat
example of NaCI with equivalent
number of Nat and Cl- vacancies CI- Nat CI- Na+ CI-/ Na+ e- Nat CI-
illustrated. For this case, a migration
of cations and anions takes place only
Nat 0 Nat CI- Nat CI- o CI- Nat
-:-........
via vacancies. Clj"Nat CI- Nat CI- Nat CI
.
Na~ c:-
In principle, all types of lattice defects can occur in a stoichiometric
crystal, but a single type is usually preferred, or predominant. Investiga-
tions have also shown that lattice defects corresponding to Limiting Types
I and IV are energetically preferred. Thus, Frenkel defects are predominant
in the silver halides, as shown by numerous investigations, especially those
of Koch and Wagner,2 Stasiw,3 and Teltow. 4 As is seen in the lattice defect
1 Schottky, W.: Z. physik. Chem. (B) 29, 335 (1935).
2 Koch, E., and C. Wagner: Z. physik. Chem. (B) 38, 295 (1937).
3 Stasiw, 0., and J. Teltow: Ann. Physik. (6) 1, 261 (1947).
4 Teltow, J.: Ann. Physik. (6) 5, 63,71 (1949); Z. physik. Chem. 195, 213 (1950).
10 2. Lattice Defect Phenomena and Diffusion Processes

scheme for AgBr reproduced in Fig. 1, the bromide ion lattice is completely
occupied while some of the Ag+ ions in the cation sublattice are displaced
into interstitial lattice positions. From considerations of electroneutrality,
the concentration of silver ions in interstitial lattice positions XAgo· must
be equal to the number of silver ion vacancies XAgO', where x is the con-
centration stated as a mole fraction; an interstitial lattice position is indi-
cated by a circle and a vacancy by a square; the primes and dots next to
these circles and squares denote the number of negative or positive excess
charges, respectively.
Thus we obtain
(2.1)
with the 0 denoting the pure AgBr phase without added impurity salts.
The Frenkel lattice defects may also be called intrinsic lattice defects of the
cations, and the overall process may be compared with the intrinsic dissocia-
tion of water. For the ion lattice defect product we write
(2.2)

When an electrical field is applied to the AgBr crystal, the only migrating
entities are those Ag+ ions which move in the direction of the cathode via
either interstitial lattice positions or vacancies. On this basis, it is under-
standable that Tubandtl observed a pure silver ion conduction in accordance
with Faraday's law (per current equivalent = 1 Faraday = 1 gram-atom
separated on the cathode) on basis of transference measurements at higher
temperatures. In contrast to these, alkali halides have lattice defects of the
Schottky type. For instance, in NaCI, from the electroneutrality require-
ment, there must be equal concentrations of sodium ion and chloride ion
vacancies:
(2.3)
The mass action law under ideal conditions (similar to the ideal conditions
in very dilute electrolytes) is expressed as

XXao" XC/o' = K (2.4)

It is quite reasonable, on basis of lattice defect relationships (Fig. 2),


to find that when an electrical field is applied to an alkali halide crystal, there
is a contribution to the current transport from the halogen as well as from
the alkali ions, since, from equation (2.3), even in crystals with Schottky
lattice defects, the anions and cations have an equal number of migration
possibilities available. The fact that the cations provide a larger contribution

1 Tubandt, C.: Handbuch Exp. PhysikXII 1, Leipzig, 1932, pp. 394ff.


2.1. In Stoichiometric Ionic Crystals 11

to the current than would be expected can be attributed to the somewhat


lower height of the energy barrier which the cations have to overcome in
order to jump into an adjacent vacancy. Frequently, under given conditions,
the energy barrier for anions is higher than that for cations, so that in spite
of the statistically equal jump probability, only a small fraction of the anions
arrive at the anode.
Anti-Frenkel lattice defects were previously observed only in the
alkaline earth fluorides, especially in CaFz by Ure1 and probably in SrFz by
Croatto. z With these, all of the ionic lattice defect possibilities have been des-
cribed, but they must all still be treated in nonstoichiometric ionic and covalent
crystals, which will be discussed later, and must also be brought into reason-
able relationships with the free-electron and hole possibilities to be considered.
Jost 3 and Mott et al. 4 have derived quantitative relationships between
defect concentration and lattice defect energy. If we designate no as the
number of ions in interstitial lattice positions, no as the number of vacancies,
N as the total number of ions, and No as the total number of interstitial
lattice positions (all quantities refer, e.g., to a crystal of 1 cm3 ), there then
exists a relationship between these quantities and the energy expended,
Eo or E oo , to create an ion vacancy or to dislodge an ion and bring it to an
interstitial lattice position with the simultaneous production of a vacancy:
Frenkel Lattice Defects
no = viN No exp( - E oo/2kT) (2.5)
Schottky Lattice Defects
no = N exp( - Eo/kT) (2.6)
These relationships are valid only if the volume change accompanying
the formation of defects and the influence of the defect position in the ionic
lattice on the ion oscillation frequency are both neglected. Frequently,
however, these factors cannot be ignored.
In a discussion of the migration of ions via lattice defects-which may
be caused by an electric field or by a chemical potential gradient set up
in the scaling layer during the oxidation of a metal-we must consider both
the lattice defect energy given in equations (2.5) and (2.6) and the average
energy U for a change of place, which is required to surmount the potential
barriers in the migration between interstitial lattice positions and vacancies.
1 Ure, R. W., Jr.: J. Ohem. PhY8. 26, 1363 (1957).
2 Croatto, U., and A. Mayer: Gazz. Ohim. ltal. 73, 199 (1943); U. Croatto and M.
Bruno: Gazz. Ohim. ltal. 78, 95 (1948).
3 Jost, W.: Trans. Faraday Soc. 34, 860 (1938).
4 Mott, N. F., and M. J. Littleton: Trans. Faraday Soc. 34,485 (1938); N. F. Mott and
R. W. Gurney: Electronic Proce88e8 in Ionic Ory8ta18, Oxford, New Jersey, 1948.
12 2. Lattice Defect and Phenomena and Diffusion Processes

Mott and Gurney have given the following simple expressions for the elec-
trical conductivity:

" = "g exp{ - (-~Eoo + U)jkT} (2.7)


o
" = "0 exp{ - (ED + U)jkT} (2.8)
where all constant terms contained in ,,0 are calculable according to the
special model concept presented by Mott and Gurney.!
Now, in what ways can the lattice defects be changed? One possibility-
through change of the temperature-has just been discussed. A further
possibility is immediately apparent if we remember the continually pre-
vailing electroneutrality principle, which states that the total number of
excess positive charges must be equal to the total of excess negative charges per
unit volume of the lattice. If we now dissolve divalent metal halides, e.g., CdBr2,
PbBr2, CdCI 2 , in silver halides, then according to the symbolic lattice defect
equation
CdBr2 = Cd.·(Ag) + AgO' + 2AgBr (2.9)

Owing to the requirement for electroneutrality, the introduction of


divalent cadmium causes the appearance of additional excess negative
charge carriers-in this case AgO' species. Every divalent cadmium ion
introduced into the lattice is the carrier of a single excess positive charge.
We designate this by using a black circle as the symbol for the substitution
and a dot for the charge. The ion in parentheses is the one that was replaced.
Ag' Br- Ag' Br- Ag'
B,'- e Br- Ag' Br-
Ag' Br- Cd2+ Br- Ao~

Br- Ag' Br- e Be Fig. 3. Section of the lattice defect model of a heterotype solid
Cd 2+ Br- Ag+ Br- Ag+ solution of AgBr-CdBr2' according to Wagner. For this case
XAgo' = XCdBr 2 == XCd • ·(Ag)·

The introduction of the impurity ion is shown schematically in Fig. 3 for


a crystal plane. The concentration of ions in interstitial lattice positions
XAgo' decreases in accordance with the mass action law equation (2.2)
because of an increase in the concentration of vacancies XAgo" and electro-
neutrality condition reads
(2.10)

From this and from consideration of equations (2.1) and (2.2), it follows that
1A good review of ionic conductivity is: A. B. Lidiard, "Ionic Conductivity," Handbuch
der Physik, Volume XX, Springer.Verlag, Berlin, 1957, p. 246. (This work discusses the
importance of the correlation factor for self· diffusion. )
2.2. In Stoichiometric Ionic Crystals 13

the ratio of the vacancy concentration of a solid solution of AgBr-CdBr2


(XAgO') to pure AgBr (x~gO') is

XAgO' XCdBr.
--=--+
o 2 0
[( XCdBr.)2
-
2 0
- +1 ]! (2.11)
X AgO' X AgO' X AgO'
With larger additions of CdBr2, i.e., when xCdBr, = XCde ·(Ag) ~ x1 go" we
have

XAgO' XCdBr.
--~--- (2.12)
o
XAgO ' 0
XAgO '

Therefore, to a first approximation we can immediately set the CdBr2


concentration equal to the vacancy concentration XAgo' in the solid solution,
which, as defined by Laves,! is a heterotype 80lid 8olution. We will have
considerable occasion to make use of this fact in subsequent discussions
of oxidation processes in metal alloys. The increase in electrical conductivity
in the solid solution associated with the increase in XAgO' (corresponding
to increased PbBr2 content) is illustrated in Fig. 4, after Koch and Wagner. 2
The heterotype solid solutions with Schottky lattice defects may be

~r------+------1-~---1

~~------~~----~----~

tZOk---=----+---t-------I
>0:

Fig. 4. The electrical conductivity of AgBr-


PbBr2 solid solutions as a function of PbBr2
content, according to Koch and Wagner.

1 Laves, F.: Chemie 57,30 (1944).


2 The conductivity minimum appearing at the beginning indicates a greater mobility
of silver ions via interstitial lattice positions-see E. Koch and C. Wagner: Z. physik.
Chern. (B) 38, 295 (1937).
14 2. Lattice Defect Phenomena and Diffusion Processes

treated in a completely analogous way in alkali halide solid solutions. 1 How-


ever, since the alkali halide crystals are not involved in the following oxida-
tion processes, a treatment of their lattice defect mechanism will not be
included.

2.2. Defect Phenomena in Nonstoiehiometrie Ionic Crystals


The ionic crystals considered in the preceding section, with their
heteropolar character and stoichiometric composition, show predominantly
ionic defects with a practically negligible concentration of electronic defects
(ionic conductor). On the other hand, in ionic crystals of nonstoichiometric
composition, the number of electronic defects is of the same order of magni-
tude as the number of ionic defects, resulting in completely different phys-
icochemical behavior of the crystals. Since the mobility of electronic defects
is orders of magnitude larger than that of ionic defects, if, for example, one
applies an electric field the conduction is predominantly electronic, and the
ionic contribution to the current is often smaller than one part in a thousand.
These defect phenomena are particularly marked in oxides, sulfides, selenides,
etc., which, because of their electrical properties, are called semiconductors.
In general, these semiconducting compounds can be divided into three
groups: i-type or intrinsic semiconductors, n-type or electron-excess semi-
conductors, and p-type or electron-deficit semiconductors. In the present
state of the art, only n-type and p-type semiconductors are of interest in
problems of metal oxidation, because only for these cases does one know
the ionic as well as the electronic defect characteristics, and knowledge of
both is requisite for treatment of the diffusion and transport phenomena
in semiconducting scaling and tarnishing layers. Scaling layers with i-type
semiconducting crystals, e.g., CuO, at present admit of no quantitative
treatment of their diffusion phenomena, since our knowledge of the nature
and number of ionic defects in these compounds is still very inadequate. We
also find the same situation wit.h regard to spinels and mixed spinel phases,
which are important in the formation of tarnishing layers, which provide
protection against oxidation. A metal excess or nonmetal deficiency is
exhibited in n-type oxides, and such nonstoichiometry may appear, for
example, in a metal oxide if it gives off a certain amount of oxygen at
elevated temperatures.
We thus obtain either an excess of metal ions, through their introduction
into interstitial lattice positions MeO" (in the case of divalent metal ions), or
a deficiency of oxygen, through formation of oxide ion vacancies 00". In
1See, for example, K. Hauffe: Reaktionen in und an feBten Stoffen, Springer, Berlin!
G6ttingen!Heidelberg, 1955, pp. 73ff.
2.2. In Nonstoichiometric Ionic Crystals 15

any case, from a consideration of electroneutrality, we know that there is


an equivalent number of excess electrons e (so-called conduction electrons).
At lower temperatures these are associated to a fairly high degree with the
ionic lattice defects carrying the positive excess charge. Zinc oxide is a typical
example of an electron excess or n-type oxide and, according to the above
explanations, it can dissociate in four ways:
(2.13a)
In (2.13a) a conduction electron is bound to a zinc ion in an interstitial
lattice position Zn 0·· according to Zn 0·· + e ~ Zn O· as a limiting case or
(2.13b)
and
(2.14a)
In this reaction a conduction electron e is captured by an 0 2 - vacancy 00··
according to 00·· + e ~OO· (a color or F center appears, which may be
recognized, in sodium halides, for example, by a blue coloration, as Pohl
and co-workers were able to showl ) or
(2.14b)
Null denotes a stoichiometric crystal without lattice defects. Even
though the interstitial lattice defects in equation (2.13) predominate in the
case of ZnO (see also Fig. 5), (supported by exchange experiments with

Zn.'· 0- Zn.'·
0- e Zn.'· 0-
Zn.+
0= Zn.'·
Fig. 5. Idealized representation of the lattice defects 0- Zn.'. 0= e
in ZnO. (The ratio of uni· and divalent zinc ions in
interstitial lattice positions depends on the temperature.) Zn.'· e 0=

radioactive Zn65 by Banks,2 Miller,3 and Lindner,4) we have expressed,


in equation (2.14b), the other possibility, i.e., a lattice defect with free
electrons. This possibility appears to exist, for example, in Ti0 2, as investi-
gations by Ehrlich5 and by Buessem and Butler 6 have shown. On
1 Pohl, R. W.: PhY8ik. Z. 36, 732 (1935); 39, 36 (1938).
2 Banks, F. R.: PhY8. Rev. 59, 376 (1941).
3 Miller, P. H., Jr.: PhY8. Rev. 60, 890 (1941).
4 Lindner, R.: Acta Chem. Scand. 6, 457 (1952).
5 Ehrlich, P.: Z. Elektrochem. 45, 362 (1939).
6 Buessem, W. R., and S. R. Butler: Kinetic8 of High-Temperature Proce88es, John
Wiley, New York, 1959, p. 13.
16 2, Lattice Defect Phenomena and Diffusion Proces~es

basis of the dissociation mechanism equation (2.13a), it follows from the


mass action law that
XZn o' xe = K PO,1/2 (2.15)
Since, according to equation (2.13a), the concentration of free electrons
Xe is equal to that of univalent zinc ions in interstitial lattice positions,
XZno' = X e , it then follows for the oxygen pressure dependence of the
defects that
Xe = xZ no '= constpo.1/ 4 (2.16a)
A corresponding expression may be obtained from the existence of lattice
defect equilibrium, as expressed in equation (2.13b):
Xe = 2xzno" = constpo.1/ 6 (2.16b)
These relationships may be easily checked by electrical conductivity measure-
ments (x as a function of Po,), since Xe '" x, The experiment in the tempera-
ture region between 500 and 700°0 gave "'" p(j;/4.5 to p(j;/5 (Fig. 6).1

log xt-3
(roO:) a7t---+----',,""",,-
loa Xi-'!-
7sso/
a51---+----j---
Fig. 6. Dependence of the elec·
trical conductivity" of sintered
ZnO samples on the oxygen
aJ~~r.~O-----1.~.;----~~~8-1.o--~2~2----~2£~m-m-o~ pressure, according to Baum·
9pq, bach and Wagner.

In contrast to crystals in which ionic conduction predominates, in


those now being considered the external gas atmosphere influences ionic
as well as electronic defects. This is significant in the kinetic treatment
of lattice defects, which comes later. The question now arises as to what
factors, besides temperature and nonmetal partial pressure, could influence
the lattice defect concentration. From the lattice defect equations, we can
consider the effect of introducing Li 20 and Ab03 as examples of metal ions
of lower and higher valence, respectively. The addition of the lower-valent
oxide to ZnO
(2.17a)
or
(2.17b)

1 Baumbach, R. R., and C. Wagner: Z. physik. Ohern. (B) 22,199 (1933).


2.2. In Nonstoichiometric Ionic Crystals 17

decreases the concentration of free electrons and increases that of the zinc
ions in interstitial lattice positions. The addition of the higher-valent oxide,

ZnO' + Al 0
2 S = 2Al.·(Zn) + e + 3ZnO (2.ISa)
or
(2.ISb)
increases the concentration of free electrons and decreases that of zinc
ions in interstitial lattice positions. The experimental proof of the validity
of these assumptions was obtained by Wagnerl and Hauffe 2 (Fig. 7). The
above example should be adequate to illustrate the influence of cations of
different valences on the lattice defect concentration, which determines
the oxidation mechanism and oxidation rate in an n-type scaling layer.

-1
V -.........
,/ ~J

If \
117

rr ~J
"

\ /

Fig. 7. Dependence of the electrical


\ \ I }~O
conductivity" of heterotype ZnO mixed
phases on the concentration of the
foreign oxide at 395°0 and 1 atm air, Q 1 e J f of
according to Hauffe and Vierk. Foreign Oxide, 'Mol %-

Introduction of ions of different valences influences the lattice defects


in other electron excess or n-type compounds; e.g., CdO,3,4 Ti02 ,5 Th02 ,6
1 Wagner, O.:J. Chem.Phys.18, 62 (1950).
2 Hauffe, K., and A. L. Vierk: Z. physik. Chern. 196, 160 (1950).
3 Baumbach, H. H., and O. Wagner: Z. physik. Chern. (B) 22, 199 (1933).
4 Bauer, G.: Ann. Physik (5) 30,433 (1937); O. A. Hogarth and J. P. Andrews: Phil.
Mag. (7) 11, 272 (1949); R. Giang: Diss. Darmstadt 1955.
5 Earle, M. D.: Phys. Rev. 61, 56 (1942). See also T. Hurlen: J. Inat. Metals 89. 128
(1960); K. Hauffe, H. Grunewald, and R. Tranckler·Greese: Z. Elektrochem. 56, 937
(1952); W. A. Weyl, and T. F0rland: Ind. Eng. Chern. 12,257 (1950); G. H. Johnson:
J. Am. Ceram. Soc. 36,97 (1953). See also F. A. Grant: Rev. Mod. Phys. 31,646 (1959).
6 Foex, M.: Compt. rend. 215. 534 (1942).
18 2. Lattice Defect Phenomena and DUfusion Processes

AlzOa,l Ta Z05,1 PbCr04,z SnOz,a NiS,4 AgzS,5 SnSe,6 FezOa,7 and Mo0a8 are
influenced in the same way. This influence is significant in determining the
oxidation rate.
In contrast to the compounds listed above, there are oxides, sulfides,
and halides which always have a nonmetal excess or a metal deficiency. Such
a nonstoichiometry may appear if oxygen is introduced into interstitial
lattice positions of a metal oxide, such as NiO, which has completely occupied
cation and anion sublattices, but this is not very likely because of spatial
energetic reasons. The other type of lattice defect with which we will have to
deal generally arises in the formation of metal ion vacancies, where, according
to the model, the lattice defect is not due to an oxygen excess but rather to a
nickel ion deficiency, NiD", with an equivalent number of electron holes EB.
Upon application of an electric field, we also observe in this type of crystal
a pure electron conduction which is due to the fact that the mobility of the
holes is several orders of magnitude greater than that of the metal ion
vacancies. The conduction mechanism-in this case hole or p-conduction-
is characterized by positive signs for the thermoelectric power and the Hall
effect.
The penetration of oxygen into an oxide like NiO always proceeds
through chemisorption, e.g.,
-(0")
!O(:)~ Ocbem + EB(R) (2.19a)
where the symbols (0-) and (R), which we will use more precisely later,
denote the surface and the crystal region near the surface (space-charge
layer), respectively. At sufficiently high temperatures a surface reaction
with the nickel ions of the lattice takes place immediately with simultaneous
production of an equivalent number of nickel ion vacancies:
O;;-A~~l~NiO + NiD" + EB (2.19b)
However, since at such high temperatures a vaporization of NiO can be
neglected,9 and the chemisorption of oxygen is "overrun" by the surface
reaction, the lattice defect equation for high temperature is written by
combining equation (2.19a) and (2.19b) in the following way (Fig. 8):
!O(~) ~ NiO + NiD" + 2EB (2.20)
1 Hartmann, W.: Z. Physik 102, 709 (1936).
2 Lashof, T. W.: J. Chern. Phy.~. 11, 196 (1943).
3 Hauffe, K., and A. L. Vierk: Z. physik. Chern. 196, 160 (1950).
4 Hauffe, K., and H. G. Flint: Z. physik. Chern. 200, 199 (1952).
5 Klaiber, F.: Ann. Physik (5) 3, 229 (1929); C. Wagner: J. Chern. Phys. 21, 1819 (1953).
6 Davidenko, V. A.: Z. Physik. SSSR 4,170 (1941).
7 Verwey, E. J. W., P. W. Haayman, and F. C. Romeyn: Chern. lVeekblad 44,705 (1948).
8 Stahelin, P., and G. Busch: Helv. physica Acta 23,530 (1950).
9 Grimley, R. T., R. P. Burns, and M. G. Inghram: J. Chern. Phys. 35, 557 (1961).
2.2. In Nonstoichiometric Ionic Crystals 19

From this follows the law of mass action,


2 1/2
XNiD" X E!! = Kpo. (2.21)
from which the oxygen pressure dependence of the lattice defect concentra-
tion-with the condition x(j) = 2xNio,,-may be found:
X(j) = 2XNiO" = constpK6 (2.21a)

NiZ+ 0'- Ni'+ 0'- Ni'+ 0 2-


0'- Ni3+ 0'- e 0'- Ni"
Ni2+ OZ- Ni'+ 0'- NiH 0'-
Fig. 8. Idealized picture of the ion lattice defects 0'- e 0'- Ni'+ 0'- Ni'+
and electron holes in pure NiO. The trivalent Ni 2+ 0'- Ni3+ 0'- Niz+ Oz-
nickel ions appearing in the lattice represent the
locations of the electron holes. Oz- Niz+ 0'- Ni'+ 0'- Nil>

Recently, Mitoff1 found that this relation is in good agreement with the
experimental results x ,...., p~.6, while Baumbach and Wagner2 found x,...., 4 •5 • P6.
Thus, since x,...., x(j) it follows that x(j) is equal to a constant times either
1/6
po. 1/4.5 (
or Po, see F'Ig. 9) .

~/
-til

/-1'V
-ac ,..;t/

v+ ~
.....+ / I
-tl~- /~ L+"" I
::;>l /
.....i'"
.w'e
-tl.
...../
w u ~ u u u u ~ U U M
logp..:;_
Fig. 9. Oxygen pressure dependence of the electrical
conductivity of NiO, according to Baumbach and Wagner.

The exponent of the oxygen pressure for individual compounds depends


upon the lattice defect equation that is applicable. The principal difference
between the nand p defect types is in the sign of the exponent, i.e., n-type
compounds always show a decrease in the lattice defect concentration with
increasing nonmetal partial pressure, while for p-type compounds an
1 Mitoff, S. P.: J. Chem. Phys. 35, 882 (1961).
2 Baumbach, H. H., and C. Wagner: Z. physik. Chem. (B) 22, 199 (1933).
20 2. Lattice Defect Phenomena and Diffusion Processes

increase in the lattice defect concentration is observed. As we shall see, the


penetration of ions of different valence causes a noticeable inverse process.
Thus, for example, the introduction of lithium ions into the NiO lattice
(in contrast to ZnO) causes an increase in the concentration of holes-in
this case the Ni3+ sites~and a corresponding decrease in the vacancy
concentration (Fig. 10):
(2.22a)
and
Li2 0 + KiD" = 2Li.'(Ki} + NiO (2.22b)

Ni,z . . 0'- LL+ 0'- Nie~ 0.2-


c'- N1.3i- 0'- NLl+ 0'- N~Z+

NiZ+- 0'- N"


'0 0'- Li.+ 0:-
0'- 0'- 0'- Fig. 10. Idealized representation of the ion
Lt' 0 Nt" lattice defects and electron holes in the hetero·
Ni." 0'- Ni'+ 0'- NLl+ 0'- type solid solution, NiO-Li20.

The introduction of higher-valent cations, e.g., Cr3+, in a manner governed


by the substitution equations
(2.23a)
and
Cr20g = 2Cr.· (Ki) + Ki D + 3NiO
II (2.23b)
lowers the concentration of holes and increases the concentration of the
nickelion vacancies (Fig. ll). This behavior has been confirmed experimentally.

NiH 0'- NiZ+ 0'- Nit+ 0'-


0'- 0 0'- Nil> 0'- Ni l +
NiZi- 0'- Cr'+ 0'- Ni 2+ 0'-

0'- NiZ+ 0'- NiZ+ 0'- Cr"


Ni" 0'- 0 0'- Ni ZT o~-
Fig. 11. Idealized representation of the ion lattice defects
and electron holes in the heterotype solid solution NiO-
0'- NiH 0'- er'+ 0'- Ni~+ Cr203, according to Wagner.

The electrical conductivity of the NiO is greatly increased (Fig. 12)1 by the
introduction of Li 20 according to equation (2.22a) and considerably decreased
by introduction of Cr203 (Fig. 13)2 according to equation (2.23a). Recently,
Schlosser3 conducted an extensive investigation into' the lattice defect
1 Verwey, E. J. W., P. W. Haayman, and F. C. Romeyn: Chem. Weekblad. 44, 705 (1948).
2 Hauffo, K.: Ann. Physik (6) 8, 201 (1950).
3 Schlosser, E. G.: Z. Elektrochem. 65, 453 (1961).
2.2. In Nonstoichiometric Ionic Crystals 21

8aw~--~---+--~~--+---~

t .aw~--~---+~~~--T---~
:>(I>l'
Fig. 12. Increase in conductivity of the ¥aw~--~~-+--~r---+---~
NiO-Li20 mixed oxide with increasing Li20
content, calculated from the experimental
data of Verwey, Haayman, and Romeyn.
(Measurements were carried out at approxi- as'TO-a 1-10-;: t.S'/(J-a c'f(J-" 2.s"0-;:
mately 20°C). j.:cu,o
conduction properties of both pure NiO and NiO with Li 20 and Ga20a
impurities, between 100 and 600°C in air, and from thermoelectric power
measurements was able to calculate both the distance of the acceptor level
from the valence band edge and the mobility of the holes. 1
Similarly, a change in the concentration of ionic and electronic defects
is to be expected from the introduction of ions of different valences in
other p-conducting oxides and sulfides. The following compounds may be
identified with high probability as p-type conductors with metal ion vacancies:
BhOa,2 Cr20a,3 Cu 20,4 CoO,5.6 FeO,5.7 Pr20a,s M00 2,9 ThO,9 CuI,lO and SnS. ll

(J

-!

\
-j' \
Fig. 13. Conductivity of NiO-Cr20a mixed oxide I
with increasing Cr20a content at 400°C in air
at 1 atm, according to Hauffe.
1 See also S. van Houten: J. Phys. Chern. SolidB 17, 7 (1960).
2 Mansfield, R.: Proc. Phys. Soc. (B) 62, 476 (1949).
3 Hauffe, K., and J. Block: Z. physik. Chem. 198, 232 (1951); Fischer, W. A., and G.
Lorenz: Arch. Eisenhuttenw. 28, 497 (1957); Z. phys. Chem. [NF] 18,308 (1958).
4 Gundermann, J., K. Hauffe, and C. Wagner: Z. physik. Chem. (B) 37, 148 (1937); C.
Wagner and H. Hammen: Z. physik. Chem. (B) 40, 197 (1938); N. N. Greenwood
and J. S. Anderson: Nature (London) 164, 346 (1949).
5 Wagner, C., and E. Koch: Z. physik. Chem. (B) 32, 439 (1936).
6 Carter, R. E., and F. D. Richardson: J. Metals 6, 1244 (1954).
7 Hauffe, K., and H. Pfeiffer: Z. Metallk. 44, 27 (1953).
8 Martin, R. L.: Nature (London) 165, 202 (1950).
9 Hochberg, B. M., and M. J. Sominski: Physik. Z. Sowjetunion 13, 198 (1938).
10 Nagel, K., and C. Wagner: Z. physik. Chem. (B) 25,71 (1934); R.J. Maurer:J. Chern.
Phys. 13, 321 (1945).
11 Anderson, J. S., and M. C. Morton: Proc. Roy. Soc. (A) 184, 82 (1945).
22 2. Lattice Defect Phenomena and Diffusion Processes

In all these crystals the vacancy concentration was increased by the intro-
duction of higher-valent metal ions and decreased by the introduction of
lower-valent metal ions in the parent lattice.
The mechanism of the influence of lattice defects on ionic crystals with
intrinsic semiconducting and amphoteric defect phenomena, as, for example,
in CUO,1 CaO,2 PbS,3 is considerably more complicated. All of these
compounds show intrinsic electron defects in a definite temperature and
nonmetal partial pressure region which are characterized by an equality
of the concentration of free electrons and holes:
Xull ........... e+e (2.24)
However, this does not give any information about the ionic lattice defects
(in contrast to the n- and p-conducting crystals). Cupric oxide, with the
largest intrinsic semiconducting region in the group (from Po. = 10-3 to
greater than 1 atm), is still the compound about which least is known with
respect to ionic lattice defects. In contrast to the lattice defect types treated
above, in this case one cannot obtain any information about ionic lattice
defects from changes in the electrical conductivity resulting from the
addition of ions of other valences. For example, in CuO, changes resulting
from the introduction of trivalent chromium ions will be counteracted by
the production of free electrons and the annihilation of holes (== Cu3 +) in
accordance with the relations
(2.25a)
and
(2.25b)
The inverse change in concentration of free electrons and holes results
from the introduction of Li 2 0. Since it may be assumed that the mobilities
of the free electrons and the holes are not more than an order of magnitude
different, a higher conductivity may be expected with lower-valent as well
as with higher-valent additions. Experiments have confirmed this (Fig. 14).
However, in order to obtain information about ionic lattice defects and the
migration mechanism of the ions in CuO, diffusion experiments should be
carried out with radioactive Cu 2+ ions and 180 isotopes.
In the case of amphoteric compounds without added impurities (e.g.,
PbS with only a narrow intrinsic semiconducting region), one can assume

1 Baumbach, H. H., H. Diinwald, and C. Wagner: Z. physik. Chern. (B) 22, 226 (1933);
K. Hauffe and H. Grunewald: Z. physik. Chern. 198, 248 (1951).
2 Hauffe, K., and G. Tranckler: Z. Physik 136, 166 (1953).
3 Eisenmann, L.: Ann. Physik (5) 38,121 (1940); H. Hintenberger: Z. Physik 119,1
(1942); C. Wagner: J. Chem. Phys. 18,62 (1950).
2.2. In Nonstoichiometric Ionic Crystals 23

t -!5 t---+--t---lt-t-
~

Fig. 14. The electrical conductivity of in- Jr-~5t---+--t-~~~~


trinsic semiconducting CuO as a function of
additions of LiaO and CrgOS at 100 and 200°C -J.5r-~---+--~~+-~---1
in air, according to Hauft'e and Griinewald.
(Here the conductivity increases with ions of
lower as well as higher valence.)

that in a lead-excess sample there will be a formation of lead ions in inter-


stitial lattice positions PbO", a formation of sulfur ion vacancies SO",
and an appearance of free electrons. In a sulfur-excess sample, there will be
formation of an equivalent number of lead ion vacancies PbO" and the
appearance of holes.! Self-diffusion measurements with lead and sulfur
isotopes, recently carried out by Wagner and co-workers 2 in PbS as well as
PbSe, have extended our knowledge of the lattice defect structure and the
diffusivityofions via lattice defects. In both compounds, the lead ion diffusion
is predominantly via Frenkel defects, where interstitial diffusion is faster.
For the diffusion mechanism of the nonmetal ions, some questions still
remain unanswered. The poT-x phase diagram of the lead-sulfur system was
investigated by Bloem and Kroger. 3
Because a special conduction mechanism (intrinsic spinel semiconduc-
tion)4 prevails in normal as well as in inverse spinels with the general formula
M2+N:+O~-, the kind and extent of the ionic lattice defects in these cases
is still unknown. For an understanding of this state of affairs, we consider
the structure of the spinel lattice (Fig. 15) as it is taken from the works of
Machatschki,o and Barth and Posnjak,6 as well as Verwey7 and Kordes. 8
Here two kinds of interstices are formed from the face-centered cubic packing
arrangement of the oxygen ions. One kind is made up of four oxygen ions
1 Wagner, C.: J. Ohem. Phys. 18, 62 (1950).
a Seltzer, M. S., and J. B. Wagner: J. Ohem. Phys. 36, 130 (1962); G. Simkovich, and
J. B. Wagner: J. Ohem. Phys., 38, 1368 (1963) M. S. Seltzer, and J. B. Wagner, Tech.
Rept. No.3, Office: Naval Res., Contract NONR 609(34), Oct. 1961.
3 Bloem, J., and F. A. Kroger: Z. physik. Ohem. [NF] 7, 1 (1956).
4 Hauffe, K.: Ergeb. wakt. Naturw. 25, "274J! (1951).
5 Machatschki, F.: Z. Krist. 82, 348 (1932).
8 Barth, T. F. W., and E. Posnjak: Z. Krist. 82, 325 (1932).
7 Verwey, E. J. W., and J. H. De Boer: Ree. trav. ehim. 55, 531 (1936); E. J. W. Verwey
and E. L. Heilmann: J. Ohem. PhYB. 15, 174 (1947); E. J. W. Verwey, F. De Boer, and
J. H. van Santen: J. Ohem. PhYB. 16, 1091 (1948); 18, 1032 (1950).
8 Kordes, E.: Z. Krist. 92,139 (1935).
24 2. Lattice Defect Phenomena and Diffusion Processes

Fig. 15. Spinel lattice according to Machatschki, Barth,


and Posnjak. The large white spheres represent the
oxide ions; the dotted spheres, metal ions in octahedral
positions; and the small black spheres, metal ions in
tetrahedral positions.

(tetrahedral interstice) and the other of six oxygen ions (octahedral inter-
stice). The octahedral interstices are somewhat larger than the tetrahedral
ones. If one considers an elementary cell with 32 oxygen ions, then one finds
32 interstices between the oxygen ions which are bounded by octahedra
of oxide ions, and 64 interstices which are surrounded by tetrahedra of
oxide ions. In a normal spinel, of these interstices, 8 tetrahedral ones are occu-
pied by divalent metal ions and 16 octahedral ones by trivalent metal ions.
This geometry, however, does not always prevail, as was shown with NiAl 204
and CoAl 20 4 by Schmalzried. 1
The following combinations of occupancy are frequently realized in
spinels of the MN204 type, which are important in the formation of
scaling layers:
M2+N~+O~­ 2-3-S Pinel}
M4+N~+O~- · I Normal
4- 2-Spme
N3+(M2+N3+)O~- Inverse spinel (for example, Fe304)
It is obvious that a large number of spinels can be obtained between
the limits of the normal spinels and the inverse spinels discussed by Barth
and Posnjak. These inverse spinels have only half of the trivalent metal ions
in octahedral positions, which fact is responsible for their extremely high
conductivity.
If the above-mentioned interstices were responsible for the migration
of ions through the lattice, then one would hardly obtain an increase or
decrease in the diffusion rate of ions by initiating a nonstoichiometry with
foreign ions of higher or lower valence than the ions in the spinel lattice.
Recently, Schmalzried and Wagner 2 showed that the lattice defects in
ternary ionic crystals are dependent on the thermodynamic activities of the
1 Schmalzried, H.: Z. phY8. Chern. [NF] 28, 203 (1961).
2 Schmalzried, H., and C. Wagner: Z. physik. Chern. [NFl 31, 198 (1962).
2.3. Near the Surface of Nonstoichiometric Ionic Crystals 25

single oxides forming the spinel. Because of the dependence of the activity
on the oxygen partial pressure, the lattice defects in the spinel must also
be a function of the oxygen partial pressure. This conclusion can be proved
by self-diffusion measurements with iron isotopes using Fe304, as the diffusion
coefficient is proportional to the iron ion vacancy concentration'! Experi.
ments have confirmed this.
Specifically, the iron ion vacancy concentration in magnetite is given
by
d lognFeD" = - 4d logaFeo
and
d lognFeD'" = - 4d logaFeo
where nFeD" and nFeD'" denote the concentrations of Fe 2 + and Fe3+ vacancies,
respectively, and aFeO is the thermodynamic activity ofFeO in Fe304. From
the governing equilibrium
Fe30~) ~ 3FeO(g) + lO~)
with the equilibrium condition
3 J.
aFeO'Po, = const
we obtain
d lognFeD" = id logpo,
and
d lognFeD'" = fd logpo.
and since, D Fe ,..., nFeD" or nFeD''', we get for the diffusion coefficient D Fe
d log DFe = id logpo.
Measurements of spinel formation in various gas atmospheres have
provided considerable insight into the diffusion mechanism of ions through
spinellattices. 2

2.3. Lattice Defect Phenomena in Regions Near the Surface of


Nonstoichiometric Ionic [rystals
The above treatment of lattice defect phenomena in nonstoichiometric
ionic crystals as dependent on temperature and gas atmosphere is based on
the assumption that the ion lattice defects and the free electrons and holes
appearing in the whole crystal are uniformly distributed and are in thermo·
dynamic equilibrium with the surrounding gas atmosphere. This equilibrium
1 Schmalzried, H.: Z. physik. Chem. [NFl 31, 184 (1962).
2 Schmalzried, H.: Z. phys. Ch~m. [NFl 33,111 (1962).
26 2. Lattice Defect Phenomena and Diffusion Processes

can frequently be attained by choosing a sufficiently high experimental


temperature. The temperature required for the realization of this condition
depends only on the migration rates of the ionic lattice defects and of the
electrons and holes in the crystal. If the diffusion rates of the ion lattice
defects are sufficiently high, then the effect of oxygen on the oxide is to produce
an equal number of ionic lattice defects and electrons and holes distributed
uniformly throughout the crystal, thus maintaining electroneutrality in the
entire crystal.
If the temperature is lowered so that the migration rate is not high
enough to attain equilibrium of the ionic lattice defects in the entire crystal
in a finite time, complicated phenomena appear. Under such conditions, there
is a large change in the electron and hole concentrations because of a
preferred chemisorption of the oxygen on the oxide surface. A simultaneous
change in the concentration of the ionic lattice defects with decreasing
temperature is always insignificant. Oxygen chemisorption alone, in contrast
to penetration into the lattice, causes only an electron displacement, or
rather a transfer of free or lattice electrons from the oxide to chemisorbed
oxygen. The flow of electrons either to or from the semiconductor (oxide)
without a simultaneous significant ion diffusion (whose direction is given
by the sign of the charge on the ions capable of migration) must result in a
difference in charge between the semiconductor interior and its surface.
This charge effect, which is associated with the appearance of a particular
electrical field strength, limits the inward or outward movement of the elec-
trons in the semiconductor. In other words, during chemisorption, electron
exchange always becomes energetically less favorable, and at a certain layer
depth (which can lie between 100 and a few 1000 A, depending on the
conditions of temperature and gas pressure) practically ceases. The excess
charges become localized either in the chemisorption layer, where they form
a plane phase, (cr-phase) or they group as a space charge in a region close to the
surface of the semiconductor, the space-charge boundary layer. 1 Thus, the
phenomena at the oxide/chemisorption interface are similar to those which
appear at the metal/semiconductor phase boundary, and result in the for-
mation of barrier layers. Practical application for these barrier layers is
found in crystal rectifiers. Certain approaches from crystal-rectifier theory-
especially those in the presentation by Schottky and Spenke 2-have been
found useful for a quantitative treatmpnt of these phenomena.

1 Hauffe, K., and H. J. Engell: Z. Elektrochem. 56, 366 (1952); 57,762 (1953); K. Hauffe:
Z. Electrochem. 65,321 (1961); P. B. Wei~z: J. Chern. Phys. 21,1531 (1953).
2 Schottky, W.: Naturw. 26,843 (1938); Z. Physik 113, 367 (1939); Z. Physik llS, 539
(1942); W. Schottky and E. Spenke: Wiss. Veroffentl. Siemens-Werken IS, 25 (1939);
E. Spenke: Z. Physik 126,67 (1947); Elektroni~che Halbleiter, Springer-Verlag, Berlin,
1955.
2.3. Near the Surface of NODstoichiometric Ionic Crystals 27

At temperatures where oxidation still occurs, pure chemisorption


accompanied only by changes in the free electron and hole arrangement in
the space-charge boundary layer is rarely the case, but there is a complete
equilibrium at the oxide/gas phase boundary. Hauffe and IIschnerl were
able to demonstrate that there must be a change in the ion defect concentra-
tion in the boundary layer, since the appearance of space-charge boundary
layers is a result not only of the low mobility of the ion lattice defects, but
is also a consequence of the independently achieved equilibrium of ions and
electrons with the neighboring phase. If an electron affinity exists, the separate
equilibria lead to a difference in concentration between ionic and electronic
defects in the boundary layer of the crystal. Under these conditions, one
obtains for the electrostatic potential difference between the quasi-neutral
inner phase and the surface, e.g., in NiO,
VD = V(R) _ r(H) = kT In z~~)
(z + l)e nG) (2.26)
where z is the valence of the nickel ion and VD is the diffusion potential. The
concentration changes and the energy relationships for the boundary layer
described by this equation are schematically illustrated in Fig. 16.
Free Energy
cal/Mol

Neighboring
Phase Boundary Layer Inner Phase

Fig. 16. The variation of the chemical and electrochemical


potentials of the electrons, holes, and lattice defects in an
equilibrium space-charge boundary layer, according to
HaufIe and Ilschner.
It is evident that electric fields which are set up in these space-charge
boundary layers cause an ion transport which is practically unobservable
1 Hade, K., and B. Ilschner: Z. Elektrochem. 58, 467 (1954).
28 2. Lattice Defect Phenomena and Diffusion Processes

when only a chemical potential gradient is present. The results of this


motion are of basic significance for low-temperature oxidation as well as for
the formation of passive layers, and will be discussed in detail later.

2.4. Lattice Defect Phenomena and Dillusion Mechanisms in Metals


The lattice of a metal or an alloy is similar to that of a "disturbed"
ionic crystal, especially at elevated temperatures. Depending on the spatial
energetic conditions, we encounter a configuration in which either vacancies
or metal atoms in interstitial lattice positions predominate. The appearance
of an approximately equal number of vacancies and interstitial lattice
positions (quasi-Frenkel lattice defects in metals) is also a possibility. The
diffusion of metal atoms in the presence of one or more chemical potential
gradients is decisively influenced by these defects. A chemical potential
gradient can be set up experimentally either by bringing two alloys with the
same base metal and different concentrations of alloying metal into contact,
or by allowing a chemical reaction to occur (e.g., an oxidation) in an alloy
made up of a noble and a non-noble component, where the reaction, for
practical purposes, is limited to the non-noble element. While the latter
case is more important for subsequent considerations of the oxidation
mechanism in alloys, the first method is used when the problem is the
determination of diffusion coefficients.
On basis of the appearance of such lattice defects in metals, the sub-
division of metal diffusion into vacancy and interstitial diffusion is natural-
analogous to diffusion investigations in ionic crystals. Because of the special
conditions in metals (lattice defects and diffusion of atoms), there still exists
the additional possibility of a direct exchange of two neighboring atoms,
which does not require that a lattice defect act as "intermediary." This
mechanism, which we shall call position interchange in metals, was first
discusscd by Bernall and checked by Huntington and Seitz 2 for the special
case of self-diffusion in copper. Experimental evidence failed to support
the Bernal mechanism: the calculated activation enthalpy t1H was four
times as large as the experimental value. On basis of similar calculations,
Zener 3 "howed that a simultaneous exchange of four neighboring atoms in a
circle-ring diffusion-is energetically more favorable (Fig. 17). However, this
mechanism requires that both metals in a binary-alloy diffusion system have
the same diffusion coefficients, and this condition frequently is not fulfilled.

1 Bernal, J. D.: Trans. Faraday Soc. 34,837 (1938).


2 Huntington, H. B., and F. Soitz: Pltys. Rev. 61, 315 (1942); H. B. Huntington: Pltys.
Rev. 61, 325 (1942); F. Seitz: Phys. Rev. 74,1513 (1948).
3 Zener, C.: Acta Cryst. 3, 346 (1950).
2.4. Defect Phenomena and DitJusion Mechanisms in Metals 29

Fig. 17. Ring diffusion according to Zener.


Here four atoms go through a position inter-
change in one elementary act. (a-in a face-
centered cubic; b-in a body-centered cubic
§------
.,....
-.
---

lattice). a b

The existence of different diffusion coefficients can result in a larger mass


flow in one direction than in the other, so that we observe an increase in
mass on one side of the diffusion interface and a decrease. on the other side.
In fact, Smigelskas and Kirkendall l were the first to be certain of this
effect. Figure 18 represents the classical experimental arrangement for the

Fig. 18. Experimental arrangement for the detec-


tion of the migration of the interface between oc-brass
(70--30) and copper, according to Smigelskas and
Kirkendall. After the diffusion experiment the
distance d had become smaller.

determination of this phenomenon. A block consisting of 70-30 brass was


surrounded with markers of thin molybdenum wires and plated with copper.
During the heating period there was more diffusion of zinc into the plated
copper layer than there was of copper into the brass block, owing to the
more rapid zinc diffusion rate. Thus, mass flows toward the outside, and
the marker is displaced in the direction of the brass block. This phenomenon
is called the Kirkendall effect. It was confirmed later in the copper-a.-brass
system for various compounds and temperatures by Correa da Silva and
Mehl,2 by Buckle and Blin,3 and Barnes.4 The Kirkendall effect was also
observed by these authors as well as by Seith and co-workers in the following
metal and alloy systems: Cu-Ni,2.4-6 Cu-Au,6 Ag-Au,2.5.6 Ag-Pd,5.6 Ni-Co,5.6
Ni-Au,5.6 Fe-Ni,6 and in the a.-region of Sn-Cu,2 and AI-Cu. 2.3
1 Smigelskas, A. D., and E. O. Kirkendall: Traw. AIME 171, 130 (1947).
2 Correa da Silva, L. C., and R. F. Mehl: Trans. AIME 191, 155 (1951).
3 Buckle, H., and H. Blin: J. Inst. Metals 80,385 (1951/52).
4 Barnes, R. S.: Proc. Phys. Soc. (B) 65, 512 (1952).
5 Seith, W., and A. Kottmann: Naturw. 39,40 (1952); W. Seith and R. Ludwig: Z.
Metallk. 45, 401 (1954).
6 Seith, W., and A. Kottmann: Angew. Chern. 64, 376 (1952).
30 2. Lattice Defect Phenomena and Ditlusion Processes

As Seith and Kottmann1 were able to demonstrate, the ideal Kirkendall


effect-for example, in the Ni-Cu system (Fig. 19)-is often significantly
disturbed by the appearance of pores in the copper in the vicinity of the phase
boundary and by the appearance of a "swelling" in the nickel. If we ten-
tatively assume a vacancy mechanism for diffusion, then the difference
between the diffusion currents has a further consequence. For example,
the greater the relative diffusion from the copper side of the system to the
nickel side (i.e., more atoms diffusing from copper to nickel than the reverse),

Swelling Phase Boundary


}fter the Experiment

Hi Cu
Phase Boundary
Before the Experiment

Fig. 19. Schematic representation of the


Kirkendall effect with "cavity and swelling
formation" appearing in the neighborhood of
the phase boundary in the nickel-copper
system, according to Seith and Kottmann.

the greater the deficiency of atoms in copper and the larger the excess in
nickel. If we consider the vacancies 0 as components of a ternary alloy,
e.g., Cu-Ni-D, then this is equivalent to a vacancy flow through the marker
plane in the direction of the weaker material flow. One may say, then, that
the sum of the material and vacancy flows is equal in any direction.
Originally,2 it was thought that new net planes were built up or destroyed
by the movement of dislocations in the lattice rapidly enough to maintain
the vacancy concentration in equilibrium. Balluffi and Alexander3 have
discussed the recrystallization processes in this "disturbed" diffusion
zone.
Seith and Wever4 were able to produce a Kirkendall effect by means
of electrolysis of fi-Cu-AI alloys. The results indicate a migration of both
metals in the direction of the anode. However, the aluminum moves more
slowly and therefore a relative aluminum-enrichment occurs on the cathode
side. The relative transference numbers were calculated and compared with
the partial diffusion coefficients that were obtained. In a continuation of
1 Seith, W., and A. Kottmann: Angew. Chem. 64, 376 (1952).
2 Seitz, F.: Acta Cryst. 3, 355 (1950).
3 Balluffi, R. W.: J. Appl. Phys. 23,1407 (1952); R. W. Balluffi and B. H. Alexander:
J. Appl. Phys. 23, 1237 (1952).
4 Seith, W., and H. Wever: Z. Elektrochem. 57, 891 (1953).
2.4. Defect Phenomena and Diffusion Mechanisms in Metals 31

earlier work,l Seith and Ludwig 2 investigated the mechanism of the forma-
tion of cracks or fissures in the diffusion pair Ni-Cu. Fissure formation was
observed in the vicinity of the weld on the copper side, as well as the usual
swelling on the nickel side and the constriction on the copper side (Fig. 19).
The portion of the surface on the fissure was about 80% of the total cross-
sectional area of the cylinder. The additional pores appearing in the interior
have an octahedral form, as described earlier by Barnes. 3 In recent
experiments, Seith was unable to reproduce the second row of cavities on
the nickel side observed by Seith and Kottmann1 in their first experiments.
The diffusion barrier caused by the large fissure formation appearing on the
copper side was reduced by grain-boundary and surface diffusion, accom-
panied by transport processes in the gas phase.
Balluffi and Seigle4 have also reported on the displacement of the
Matano plane and pore formation from vacancies. Of the metal pairs which
were investigated, CU-IX-brass, Cu-Ni, and Ag-Au, porosity was observed only
in the last system. Further informative investigations on the displacement
of the Matano plane in IX-brass and the dependence of the partial diffusion
coefficients on the composition were carried out by Horne and Meh1. 5 The
following equations were used for the calculation of the partial diffusion
coefficients Dzn and Deli:

D = XZn Deu + xeu D zn


and
~Jf
2t =
(D
Zn -
D
eu)
(ddf"
Xzn) 211

Here gM is the marker coordinate and (dxzn/dg)M is the tangent to the


zinc-concentration curve at the marker site. A few values from the summary
by Horne and Mehl are given in Table 1.
Reviews of progress in the field of metal diffusion have been given by
Birchena1l6 and recently by Shewmon and Love. 7 Since metal diffusion is
assuming an ever-increasing role in the explanation of the oxidation mechan-
ism in many alloys, a short presentation on vacancy and interstitial diffusion

1 Seith, W., and A. Kottmann: Angew. Chern. 64, 376 (1952).


2 Seith, W., and R. Ludwig: Z. Metallk. 45, 401 (1954).
3 Barnes, R. S.: Nature 166, 1032 (1950).
4 BaHuffi, R. W., and L. L. Seigle: Naturw. 40, 524 (1953); J. Appl. PhY8. 25, 607,
1380 (1954); R. W. BaHuffi: Acta. Mat. 2, 194 (1954).
S Horne, G. T., and R. F. Mehl: J. Metal8 7,88 (1955).
6 BirchenaH. C. E.: Ind. Eng. Chern. 47, 604 (1955).
7 Shewmon, P. G., and G. R. Love: Ind. Eng. Chern. 53, 325 (1961).
32 2. Lattice Defect Phenomena and Diffusion Processes

Table 1. Integral and Partial Diffusion Coefficients in IX-Brass of Different


Compositions at 855°C, According to Horne and Mehl

At. % Zn D X 10 9 , D zn x 10 9 , Dcu X 109 , (D zn - DCu)


cm 2 (sec cm 2(sec cm 2 Jsec X 109 ,
cm 2 Jsec

5.2 1.15 1.20 0.203 0.965


15.2 3.70 4.225 0.770 3.45
12.0 2.35 2.60 0.490 2.11
20.5 7.80 9.30 2.00 7.30
22.5 10.3 12.3 3.50 8.78
25.3 16.0 19.1 6.90 12.2

has been included in the following. Detailed discussions of the diffusion


mechanisms and the mathematical interpretation were given by Jost,l
Hauffe,2 Seith,3 Lazarus,4 Van Bueren,5 and Lomer.6 In a later chapter
we will tabulate the most important experimental data, so that the reader
may have them available, to use in conjunction with Wagner's theory,7
for future calculations of the oxidation rate of alloys for metal-diffusion-
controlled oxidation processes.

2.4.1. Vacancy Diffusion in Metals and Alloys


The appearance of the Kirkendall effect confirms the existence of a diffusion
via both vacancies and interstitial positions. Although it is assumed that a
vacancy mechanism predominates in diffusion in face-centered cubic lattices,S
the interstitial mechanism may be important in crystals with other lattice
structures. For this reason, we will treat both mechanisms briefly in the following.
It can be concluded from calculations by Huntington and Seitz 9 on the
temperature dependence of self-diffusion in copper (which has an activation
energy of 48,000 cal/g-atom or 2.1 eV) that there is a predominant vacancy
diffusion in this element. 10 Furthermore, the increase in the diffusion coefficient

1 Jost, W.: Diffusion in Solids, Liquids, and Gases, Academic Press, New York, 1952;
"Platzwechsel in Kristallen," in Halbleiterprobleme 2, edited by W. Schottky, Braun-
schweig, 1955, p. 145.
2 Hauffe, K.: Reaktionen in und an jesten Stoffen, p. 259j, Springer, BerlinJGi:ittingenJ
Heidelberg, 1955.
3 Seith, W.: Diffusion in Metallen, 2nd ed., Springer, BerlinJGi:ittingenfHeidclberg, 1955.
4 Lazarus, D.: Solid State Physics 10, 71 (1960).
5 Van Bucren, H. G.: Imperfections in Crystals, Intersciencc, Ncw York, 1960.
6 Lamer, W. M.: Progr. Metal Phys. 8, 255 (1959).
7 Wagncr, C.: J. Electrochem. Soc. 99, 369 (1952).
8 Le Claire, A. D.: Progr. Metal Phys. 4, 265ff (1953).
9 Huntington, H. B., and F. Seitz: Phys. Rev. 61,315 (1942).
10 Maier, 1\1. S., and H. R. Nelson: Trans. AIAIE 147, 39 (1942).
2.4. Defect Phenomena and Diffusion Mechanisms in Metals 33

of cobalt in Co-AI alloys with increasing aluminum content above 50 at. % Al


observed by Nix and Jaumotl is understandable if an increasing vacancy con-
centration is assumed. The investigations of Brinkman 2 concerning the tempera-
ture-dependence of the diffusion mechanism in CuaAu and Cu are worthy of note
in this connection. At higher temperatures (> 200°C), the vacancy mechanism was
apparently favored in both materials, while in the lower-temperature regions
(- 30 to 100°C), interstitial diffusion was evidently the preferred mechanism.
The work of Bardeen,a together with the supplementary formulations of
Darken4 and Seitz 5 which treat vacancy diffusion, gives us the following expressions
for the rate at which A atoms migrate through 1 cm 2 of the phase boundary
from 1 -+ 2 or from 2 -+ 1

(2.27a)

(dnA.)
\ dt 2-+1 = kA ano (an'l)
n,t + a----ar (2.27b)

where nA and no are the concentrations of the atoms and vacancies, respectively,
a is the distance between crystallographic planes 1 and 2, and ~ is the position
coordinate. For the net atom transport we have the following:
dnA an,t
dt - - k,t a2 n0---ar (2.28)

The diffusion coefficient of the atoms is given by

(2.29)

The existence of a chemical diffusion requires that we also consider the con-
centration gradient of the vacancies eno/eg. Then equations (2.27a) and (2.27b)
assume the following forms:

(2.30a)

or

(2.30b)

Here kv and kR are proportional to the jump frequency in both directions

1 Nix, F. C., and F. E. Jaurnot, Jr.: Phys. Rev. 83,1275 (1951).


2 Brinkman, J. A., C. E. Dixon, and J. C. Meechan: Acta Met. 2, 38 (1954).
a Bardeen, J.: Phys. Rev. 76, 1403 (1949).
4 Darken, L. S.: Metals Technology, Tech. Pub!. No. 2311 (1948).
5 Seitz, F.: Phys. Rev. 74,1513 (1948); "Fundamental Aspects of Diffusion in Solids, in
Phase Transformation in Solids, ASM, New York, 1951, p. 77.
34 2. Lattice Defect Phenomena and Diffusion Processes

along the ~ axis. In general kv will be different from kR and kA, since the thermo-
dynamic potential of the atoms in the individual planes is different; this difference
is the cause of the ultimate asymmetry in the potential barrier.
Assuming ideal behavior of an alloy system, kA = kv = kR, we obtain for
the net atom transport

(2.31)

or

(2.32)

where fL is the chemical potential and

DA = kA a2 no/kT (2.33)

In spite of the considerable amount of work that has been done on the
formation and annealing kinetics of vacancies in metals, no absolute value of
the concentration of vacancies was available. Simmons and Balluffil carried
out the first measurements of the absolute value of no by simultaneously
measuring the lattice parameter aL and the macroscopic length l. The maximum
values near the melting point for no were about 10-3 for aluminum and 10-4
for silver. The entropy of formation of a vacancy in aluminum was found to
be 2.4k (k = Boltzmann's constant). Further calculations were undertaken by
Seeger. 2
It is generally accepted that vacancy pairs are highly mobile and can
contribute appreciably to the annealing of vacancies. A trivacancy mechanism
in a face·centered cubic lattice was proposed by Dienes. 3

2.4.2. Interstitial Diffusion in Metals


In interstitial solid solutions (e.g., C in y·Fe and Ta), one must always
deal with a predominant interstitial diffusion. This mechanism was investigated
in detail both experimentally and theoretically by Wert and Zener. 4 Careful
note should be taken of the theoretical significance of the important quantity Do
in the general diffusion formula,

D = Do exp (- LJUjRT) (2.34)

The theoretical treatment of the transition·state method is based on the work of

1 Simmons, R. D., and R. W. Halluffi: Phys. Rev. 117,52 (1960); 119,980 (1960).
2 Seeger, A., and E. Mann: J. Phys. Chem. Solids 12, 326 (1960); A. Seeger, P. Schiller,
and H. Kronmiiller: Phil. Mag. 5, 853 (1960).
3 Damask, A., G. Dienes, and V. Weizer: Phys. Rev. 113,781 (1959).
4 Wert, C. A., and C. Zener: Phys. Rev. 76,1169 (1949); C. A. Wert: Phys. Rev. 79, 601
(1950); J. Appl. Phys. 21,1196 (1950).
2.4. Defect Phenomena and Diffusion Mechanisms in Metals 35

Wigner and Eyring,1 and the brief summaries referred to here were given by
Seitz,2 Jost,3 and Hauffe. 4
From theoretical considerations the following expression is obtained for Do:

(2.34a)

where the resonance frequency II = (H/2md)112 may be calculated from the


activation enthalpy H, the mass m of the diffusing particles, and the distance d
between two neighboring interstitial lattice positions. The accompanying value
of the entropy LIs ~ - H(d In s/dT) may be determined from the temperature
dependence of the shear modulus s (for body-centered cubic crystals: <X = rr
and n = 4, for face·centered cubic crystals: <X = ,'. and n = 12). From self-
diffusion measurements in ,B·tin,Dowas found to be between 1 and 10 in agreement
with Zener's theory.s Snoek 6 and Zener7 have shown that the relaxation phenom-
enon which they observed in metallic substitution solid solutions is directly
connected with the diffusion of the foreign atoms which were introduced. Upon
application of a mechanical stress, e.g., pressure and tension, the different
interstitial positions cease to be energetically equivalent. With a mechanical
stress, e.g., a periodic pressure-tension deformation or torsion, the carbon or
nitrogen atoms dissolved in the metal jump from interstitial lattice positions
of higher energy to lower energy levels. It seems reasonable that as a result of
these atomic processes, such a periodic deformation would undergo particularly
strong damping when the frequency of the wave is in resonance with the jump
frequency of the atoms that change position in the lattice. On basis of this
relationship it is possible to determine the jump frequency II of the atoms from
the resonance frequency.
Thomas and LeakS reported on this type of measurement on carbonized iron
wires between 24 and 74°C. Using the high-temperature diffusion data ofStanley,9
they obtained an equation for carbon diffusion in <x-iron which agrees remarkably
well with the equation formulated by Wert and Zener. Similarly good agreement
was obtained by the same authors for the diffusion of nitrogen in iron wires.
Fast and Verrij p lO were able to determine the nitrogen content in iron between
- 20 and + 60°C by the damping method and obtained good agreement with the

1 Cf. S. Glasstone, K. J. Laidler, and H. Eyring: The Theory of Rate Processes, McGraw-
Hill, New York, 1941.
2 Seitz, F.: "Fundamental Aspects of Diffusion in Solids," in Phase Transformation in
Solids, ASM, New York, 1951.
3 Jost, W.: Diffusion in Solids, Liquids, and Gases, Academic Press, New York, 1952.
4 Hauffe, K.: Reaktionen in und an festen Stoffen, p. 323ff, Springer, BerlinJGottingenJ
Heidelberg, 1955.
S Meakin, J. D., and E. Klokholm: Trans. Am. Inst. Mining, Met., Petrol. Eng. 218.
463 (1960).
6 Snoek, J. L.: Physica 8.711 (1941).
7 Zener, C.: Phys. Rev. 71.34 (1947).
8 Thomas, W. R., and G. M. Leak: Phil. Mag. 45. 656, 986 (1954).
9 Stanley, J. K.: Trans. AIME 185. 752 (1949).
10 Fast, J. D., and M. B. Verrijp: J. Iron Steel Inst. 176. 24 (1954).
36 2. Lattice Defect Phenomena and Diffusion Processes

Wert-Zener equation. A few experimental values for the diffusion of some


nonmetals in Fe, Ti, Zr, Nb, and Ta are found in Tables 2,3,4, and 5.
For the diffusion of foreign metals in nickel and silver, Swalinl proposed
a model where the solute ion is thought of as behaving as an elastic sphere, and
thus part of the enthalpy of jumping is assumed to result from the two-dimensional
hydrostatic compression of the solute ion and part from dilatation of the con-
striction which represents the saddle point. With the equation derived on
basis of this model, the activation energy can be calculated.

2.4.3. Summary of Some Diffusion Data in Metals and Ionic


Crystals Which Are Important for the Oxidation
of Alloys
The multiplicity of diffusion phenomena in metals, alloys, and ionic crystals
has given rise to copious reports on various experimental and industrial methods,
many of which have been discussed in the books of Jost, Hauffe, and Seith cited
earlier. In addition to the diffusion experiments with single crystals, there is
extensive material on polycrystalline solids. Generally, special attention is
given to grain-boundary and surface diffusion in the polycrystalline materials,
especially at intermediate temperatures. A critical discussion of these diffusion
phenomena was given by Le Claire 2 with references to the literature. In this book
we will limit ourselves to a summary of some of the more important diffusion
data necessary to account for the competition between the diffusion of the metal
atoms in the alloy phase and the ions in the scaling layer during the oxidation of
metal alloys and "internal oxidation," especially of nickel, silver, and copper
alloys. This information is requisite for an understanding of oxidation
processes and the formation of the scaling layer.
Finally we shall consider use of the diffusion equations to evaluate diffusion
experiments. Fick's second diffusion equation in the form

(2.35)

which is used for the treatment of diffusion in alloys should be considered only
as a limiting case for extremely dilute solutions. At higher concentrations, the
concentration dependence of D is no longer negligible. For the latter case, we
obtain the following equation for linear diffusion:

Since the anisotropy of a crystal lattice oan strongly influence diffusion, the
diffusion coefficient is written as a function of the crystallographic direction;
D is then no longer a scalar quantity but rather has the geometric properties
of a so-called tensor. A rectilinear coordinate system may be chosen such that
1 Swalin, R. A.: Acta .'.Iet. 5, 443 (1957).
2 Le Claire, A. D.: Progr. Metal Phys. 4, 280ff (1953).
2.4. Defect Phenomena and Diffusion Mechanisms in Metals 37

only the diagonal elements D,n;, D yy, and D .. of these tensors do not disappear
(transformation of the principal axis). In this special coordinate system, we obtain
a diffusion equation analogous to equation (2.35):
(2.36)
In an experimental approach, one generally determines the content of a particular
layer of the diffusion sample rather than c(O.
The solution and practical use of the diffusion equation, with applioation
to special case'>, has been discussed in detail by several authors. 1

2.4.4. Summary of Self-Diffusion Coefficients in


Metals and Inorganic Solid Materials
A few self-diffusion coefficients are tabulated in this chapter, but an extensive
description of the self-diffusion mechanism and of the calculation of Do will not
be given. In addition to the presentations of Zener,2 Huntington and Seitz,3
and Buffington and Cohen,4 we may especially refer to a review by Le Claire 5
in which Zener's theory was extended. Relationships for the self-diffusion
constants Do were derived by considering (1) vacancy diffusion; (2) ring diffu-
sion; (3) interstitial diffusion; and (4) interstitialcy.
Essentially, it may be shown that the experimentally obtained values of
Do and LI U [in the equation D = Do exp( -LI UjRT)] may be explained for face-
centered metals only with a vacancy mechanism and for body-centered cubic
metals only with a ring mechanism. The relationship developed by Le Claire
for D, which contains the Zener relationship for Do as a limiting case, can be
used to estimate D values of metals whose self-diffusion coefficients have not
yet been measured. The equations developed for the calculation of LI U and Do
are

and

or

Do = a 2 vexp{LlS 1 - (LlU - k1LS) G: -~J}/R


Here kl and k2 are constants which were calculated for the following metals as:
kl (Pb, Ag, Au, Cu, y-Fe, Co) = 0.125 and k2 (Pb, Ag, Au, Cu) = 0.215. The
positive entropy LlS1 can frequently be calculated; M is the atomic weight,
e the density, Ls the latent heat of fusion, and 8 the shear modulus, where 8'
1 See for instance K. Hauffe: Reaktionen in und an festen Stoffen, Springer,
BerlinJGottingenJHeidelberg, 1955, pp. 259ff.
2 Zener, C.: J. Appl. Phys. 22,372 (1951).
3 Huntington, H. B., and F. Seitz: Phys. Rev. 61, 315, 325 (1942).
4 Buffington, F. S., and M. Cohen: Acta Met. 2, 660 (1954).
5 Le Claire, A. D.: Acta Met. 1,438 (1953).
38 2. Lattice Defect Phenomena and Diffusion Processes

and e, are differentiated with respect to the temperature, and 80 and eo represent
the extrapolated values of 8 and e for T = OOK; a is the lattice constant and "
is the vibration frequency of a lattice atom.
Le Claire calculated the self-diffusion coefficients of aluminum and molyb-
denum, for example, by use of the above relationships

D!~ = 0.72 exp ( - 42,600jRT)

D:~ = 16 exp ( - 120,OOOjRT) cm 2/sec

A few self-diffusion coefficients of metals in metals and alloys and of ions in ionic
and covalent crystals are summarized in Tables 18 and 19. Some data concerning
the effects of grain boundaries on self-diffusion will be found in the following
section.

2.4.5 Grain-Boundary and Surface Diffusion in


Metal and Ionic Crystals
It is frequently observed in diffusion measurements on polycrystalline
materials, in metals as well as ionic crystals, that Do is found to be several
orders of magnitude smaller than would be expected from theory. These
anomalously low values of Do are always accompanied by low LJH values,
and the deviations become greater with decreasing temperature. These facts
indicate that with decreasing temperature the lattice diffusion (the diffusion
through defects in the lattice) is replaced to an increasing extent by a more
rapid diffusion. This more rapid process takes place through special channels
in the material-predominantly through grain boundaries. While at high
temperatures this "short-circuit" diffusion does not playa significant role,
at lower temperatures it is the lattice diffusion which becomes insignificant
owing to its considerably smaller temperature coefficient (corresponding
to a smaller LJH).1 This results from the situation presented schematically
in Fig. 20.

- - - r r - - - - (=0

Fig. 20. Schematic representation of the


concentration of diffusing atoms in the
vicinity of a grain boundary of a bicrystal.
---Ird g is the primary diffusion direction and 7J is
the lateral diffusion direction, which becomes
I r--Grain Boundary noticeable in the neighborhood of the grain
II boundary.

1 Zener, C.: Imperfection8 in Nearly Perfect CrY8tal8, New York, 1952; A. S. Nowick:
J. Appl. PhY8. 22, 74 (1951).
2.4. Defect Phenomena and Diffusion Mechanisms in Metals 39

At a critical temperature, the effective rate of grain-boundary diffusion


is equal to the rate of lattice diffusion. This critical temperature not only
differs from system to system, but depends to a large extent on the total
structure of the material [crystal structure, form, and number of the grain
boundaries per square centimeter of effective diffusion cross-sectional area,
and on other lattice disturbances, e.g., dislocations caused by deformation
and mechanical stresses (pressure, tension, torsion)]. For this reason, grain-
boundary or, more generally, short-circuit diffusion, in contrast to lattice
diffusion, is very much dependent upon the structure of the material, and is
therefore frequently referred to as structure-sensitive diffusion.
For the quantitative evaluation of the relationship of grain-boundary
and lattice diffusion Fisherl and Turnbull 2 formulated diffusion equations
which essentially represent the actual physical situation. Without going
into the mathematical derivations of these equations, which are summarized
elsewhere,3 we will simply clarify the relationships in the formulation of the
model which represents concentration changes caused by lattice and grain-
boundary diffusion of a diffusing material in a quasi-bicrystal with grain
boundaries. In addition to the primary direction of diffusion shown in Fig.
20 (from top to bottom), a lateral diffusion, which is the cause of the funnel-
shaped concentration region expanding along the grain boundaries, should
also be considered. The preferred grain-boundary diffusion of chromium in
Fe-Ni alloys can be seen in Fig. 2l.4
Turnbull was able to show quite impressively, using a formula not
reproduced here, that in the case of diffusion where no grain-boundary
effects are noticeable, and where the grain-boundary diffusion area qK is
equal to 0.05% of the total area, DK is not approximately equal to Dv, but
DK/Dv = 105 , where the subscripts K and V indicate grain-boundary and
lattice diffusion, respectively.
Using the formulation developed by Fisher, Le Claire derived a simple
relation between the diffusion ratio DK/Dv and angle oc, obtained experi-
mentally (Fig. 20):
DK 1
- = -2(7TDvt)! cot2 oc
Dv L1
Here L1 denotes the average width of the grain boundary and t the diffusion
time. If we use this relationship in diffusion experiments for copper in
1 Fisher, J. C.: J. Appl. Phys. 22, 74 (1951).
2 Turnbull, D.: Atom Movements, ASM, Cleveland, 1951, pp. 129ff; J. Metals 3, 661
(1951); D. Turnbull and R. E. Hoffman: Acta Met. 2,419 (1954).
3 Hauffe, K.: Reaktionen in und an festen Stoffen, Springer, Berlin/Gottingen/Heidelberg,
1955, pp. 401ff.
4 Hoffman, R. E.: Tracers and Other Techniques of Diffusion Measurement8 in Atom
Movements, ASM, Cleveland, 1951, pp. 51ff.
40 2. Lattice Defect Phenomena and Diffusion Processes

Fig. 21. Preferred diffusion of chromium along the grain


boundaries of an iron-nickel alloy, according to Hoffman
(Enlarged 250 times.)

polycrystalline nickel1 and choose tX = 30°, Dv ~ 10-5 cm 2jday, and


Ll = 5 X 10-8 cm as the grain-boundary thickness, we obtain for the diffusion
ratio at 1000°C
DK
- = 8 X 10 5
Dv
from which we find for the constant of grain-boundary diffusion
DK ~ 8 cm2/day
This value is about six orders of magnitude larger than that for volume or
lattice diffusion.
Achter and Smoluchowski 2 investigated grain-boundary diffusion as a
function of the relative orientation of crystallites and grain boundaries.
The measured results for silver at about 700°C show that the grain-boundary
diffusion is clearly more rapid than lattice diffusion only if the angle between
the adjacent crystallites is greater than 20° but less than 70°. A maximum
in the rate of grain-boundary diffusion was observed in the vicinity of 45°.
On the other hand, according to investigations by Whipple 3 grain-boundary
diffusion was also observed at very small angles.
Turnbull and Hoffman 4 investigated the course of grain-boundary

1 Barnes, R S.: Nature 166, 1032 (1950).


~ Achtcr, M. R, and R Smoluchowski: J. Appl. Phys. 22, 1260 (1951).
3 Whipple, R. T. P.: Phil. Mag. 45, 1225 (1954).
4 Turnbull, D., and R. E. Hoffman: Acta Met. 2, 551 (1954).
2.4. Defect Phenomena and Diffusion Mechanisms in Metals 41

diffusion in silver bicrystals prepared with initially parallel (100) axes which
wp,re shifted out of their parallel orientation by rotation through the angle
1X(9° ;;;; IX ;;;; 28°). The data can be evaluated if it is assumed that the self-
diffusion coefficient DK in the "dislocation tubes" of the grain boundaries
is independent of IX and that DK is several orders of magnitude larger than
the lattice self-diffusion coefficient (at 100°C: DK ~ 4 x 106Dv). The
activation energy in the expression
DK = 0.14exp(-19,700jRT) cm2Jsec
agrees well with the value of 20,000 cal/g-atom which was obtained earlier
from self-diffusion measurements on polycrystalline silver.
Couling and Smoluchowski1 reported an anisotropy of the grain-boundary
diffusion of silver in copper bicrystals which only appeared with an angular
orientation between 8 0 and 35 0 and between 50 0 and 78 0 • With "soft,,,
materials, such as lead bicrystals, Okkerse 2 observed similar anisotropy of the
grain-boundary self-diffusion at 220°C.
Additional experimental material on grain-boundary diffusion in
alloys was reported by Wajda 3 (zinc between 75 0 and 200°C), Langmuir4
(thorium in tungsten), Bugakow and Rybalko 5 (zinc in IX-brass), and Zwikker6
as well as by Pirani and Sandor7 (in polycrystalline tungsten-steel). A
decisive role may be played by grain-boundary diffusion in ionic crystals
in the region of intermediate and lower temperatures, as, for example, Pfeiffer,
Hauffe, and Jaenicke 8 were able to demonstrate at 20°C in solidified AgBr
melts and AgBr layers built up on silver. At that time there was still no
material suitable for a quantitative evaluation.
Apart from the difficulties in developing formulas to permit quantitative
evaluation of data on the overall diffusion coefficients, one may attempt to
find suitable models of grain boundaries and dislocations. Bragg and Nye 9
succeeded in doing this very clearly with the creation of a soap-bubble
model (Fig. 22).
Burgers lO and Read and Shockleyll have concerned themselves with
special grain-boundary and dislocation models and have calculated the

1 Couiing, S. R. L., and R. Smoluchowski: J. Appl. Phys. 25, 1538 (1954).


2 Okkerse, B.: Acta Met. 2, 551 (1954).
3 Wajda, E. S.: Acta Met. 2, 184 (1954).
4 Langmuir, 1.: .T. Franklin Inst. 217, 543 (1934).
5 Bugakow, W., and F. Rybalko: Zhur. Tekh. Fiz. 2, 617 (1935).
6 Zwikkcr, C.: Physica 7, 189 (1927).
7 Pirani, M., and J. Sandor: J. Inst. Metal8 73,385 (1947).
8 Pfeiffer, 1., K. Hauffe, and W. Jaenicke: Z. Elektrochem. 56, 728 (1952).
9 Bragg, L., and J. F. Nye: Proc. Roy. Soc. London (A) 190,474 (1947).
10 Burgers, J. M.: Proc. PhY8. Soc. (B) 52, 23 (1940).
11 Read, W. T., and W. Shockley: Phys. Rev. 78, 275 (1950).
42 2. Lattice Defect Phenomena and Diffusion Processes

Fig. 22. Soap bubble model according to Bragg and Nye.


Both the grain boundaries and the statistical distribution of
the vacancies (black circles) can be seen clearly.

dependence of the relative grain-boundary energy on orientation. The


numerical values resulting from their theoretical expressions agree quite
well with the measured values which were obtained later by various authors
for iron,l tin,2 and lead. 3 An excellent review on the theory of dislocations
has been given by Cottrell,4 and Gjostein 5 has published a stimulating report
concerning the measurement of the surface self-diffusion coefficient of copper
using a thermal-grooving technique. An extension of these models to de-
scribe the diffusion in fine-grained polycrystalline samples was given by
Levine and MacCallum. 6
Expressions can be obtained for surface diffusion which are completely
identical with those derived by Fisher for grain-boundary diffusion, but
experimental data on this subject is still very meager. Nickerson and Parker7
measured the surface self-diffusion of silver in the temperature region between
225 and 350°C and found it to be
Do = 0.16 exp( -1O,300jRT) cm2 jsec
The surface diffusion of radioactive copper atoms (Cu 64) on silver was deter-
mined in a similar manner by Fraunfelder8 at 750°C to be 7 x 10- 2 cm 2/sec.
1 Dunn, C. G., and F. Lionetti: Trans. A/ME 185,125 (1949); 188, 1245 (1950).
2 Aust, K. T., and B. Chalmers: Proc. Roy. Soc. London (A) 201, 210 (1950).
3 Aust, K. T., and B. Chalmers: Proc. Roy. Soc. London (A) 204, 359 (1950).
4 Cottrell, A. H.: Progr. Metal Phys. 4, 205 (1953).
5 Gjostein, N. A.: Trans. A/ME 221,1039 (1961).
6 Levine, H. S., and C. J. MacCallum: J. Appl. Phys. 31,595 (1960).
7 Nickerson, R. A., and E. R. Parker: Trans. Am. Soc. Metals 42, 376 (1950).
8 Fraunfelder, H.: Helv. Phys. Acta 23,347 (1950).
2.4. Defect Phenomena and Diffusion Mechanisms in Metals 43

Using the electron field emission microscope,l Drechsler2 succeeded in


explaining the processes involved in specially prepared single crystals.
Barbour, Bettler, and Charbonnier3 published a detailed investigation of the
surface self-diffusion of tungsten using the field emission microscope. Upon
the application of very high fields, individual atomic layers in the surface
could be made visible. The diffusion expression obtained in this manner was
Dw = 4 exp( -72,000/RT) cm2/sec
1 Stranski, 1. N., and R. Suhrman: Ann. Physik (6) I, 153, 169 (1947); A. Eisenloefi'el
and 1. N. Stranski: Z. MetaUk. 41,10 (1950); E. W. Muller: Ergeb. exakt. Naturw. 27,
290 (1953).
2 Drechsler, M.: Z. Elektrochem. 58, 327, 334,340 (1954)-extended the investigations of
Stranski and Miiller.l
3 Barbour, J. P., and F. M. Charbonnier: PhY8. Rev. 117, 1452 (1960); P. C. Bettler
and F. M. Charbonnier: PhY8. Rev. 119, 85 (1960).
Table 2. Diffusion Coefficients of a Few Elements in (X- and y-Iron

Approximate atomic Probably occurs in (X-Fe at 800°C, y-Fe at 1l00°C, Refer-


Element radius, A Fe as Din cm2 /sec Din cm 2 /sec ence

Hydrogen 0.46 H+ 2.75 x 10-4 (774°C) m 1.9 X 10-40 •


2.7 X 10- 4 a
Boron 0.97 B+b 6.1 x 10- 7
0.77 C+b 1.7 x 10- 6 d 6.7 X 10- 7 e
Carbon
Nitrogen 0.71 N-b 7.3 x 10- 7 c 3.8 X 10- 7
Oxygen 0.60 1 x 10- 9 g
1.28 3 x 10- 12 h
Iron
Body centered ...0
Self-diffusion 1.26 9 x 10-12 b
Face centered ~
CIJ
Cobalt 1.26 1.5 x 10-12 I 2.5 X 10- 12
Face centered
g
-
Nickel 1.25 8 x 10-12 n
Q
1.18 k (D
Manganese 2 x 10-11
(Cubic) !3l
n
Molybdenum 7 x 10-12 k 4 X 10-11
CIJ
-a
a Sykes, C., H. H. Burton, and C. C. Gegg: J. Iron Steel Inst. 156, 155 (1947).
b Seith, W., and T. Daur: Z. Elektrochem. 44, 256 (1938).
C Wells, C., according to C. E. Birchenall: "Volume Diffusion," in Atom Movements, ASM Cleveland, 1951, pp. 112f!.

d Stanley, J. K.: Trans. AIME 185, 752 (1949).


e Wells, C., W. Batz, and R. F. Mehl: Trans. AIME 188, 553 (1950).
r Bramley, A., and G. Turner: Oarnegie Scholar8hip Memoirs 17, 23 (1938).
g Brower, T. E., B. M. Larsen, and W. E. Schenk: Trans. AIME 61, 113 (1934).
b Birchenall, C. E., and R. F. Mehl: Trans. AIME 188, 144 (1950).
I Ruder, R. C.: Thesis, Carnegie Institute of Technology, 1950.
j Wells, C., and R. F. Mehl: Trans. AIME 145, 329 (1941).
k Ham, J. L.: Trans. Am. Soc. Metals 35, 331 (1945).
1 Wells, C., and R. F. Mehl: Trans. AIME 145, 315 (1941).
m Eichenauer, W., H. Kiinzig, and A. Pebler: Z. MetaUk. 49, 220 (1958).

....
Q1
~

!W

Table 3. The Dependence of the Rate of Diffusion of Carbon on Carbon Content in the Austenite Phase,
According to Wells, Batz, and Mehla I
1:1
Diffusion coefficients in D X 107 , cm 2 /sec ~
T,oC :4-
Carbon in at. %
2 3 4 5 6 7 ~
750 0.18
802 0.30 0.41 0.63
848 0.27 0.36 0.47 0.67
905 0.82 1.03 1.33 1.54 1.95
910 0.87 0.98 1.23 1.54 2.26
Ii
950 1.74 2.15 2.66 3.27 4.09 1:1
1000 2.87 3.48 4.40 5.63 7.17 10.03
1042 4.92 5.54 6.36 7.49 8.92 11.08 ~
1128 10.81 12.35 13.89 15.95 19.04 26.45 40.67
i
'"II
a Wells, C., W. Batz, and R. F. Mehl: Tran8. AIME 188, 553 (1950). a
r:
III
III
III
Table 4. Diffusion Coefficients of Carbon in Iron and Iron Alloys at Various Concentrations and Temperatures

Concentration of the Dc, Au, Refer-


Diffusing diffusing material, T, "C cm 2/sec kcal/g-atom Remarks ence
material wt.%

Carbon 1.1 900-1250 Do = 4.86 X 10-1 36.6 Welded combinations of high- a


0.1-1.0 750-1250 Do = 0.12 + 0.07 32.0 + 1.0 and low-alloyed steel b

0.4 1050 3.9 x 10 ? Composition of the steel: c


0.25% Mn, 3.80% Si, 0.011 % P,
0.31% Cr
0.5 1050 5.7 x 10-7 0.88% Mn, 0.05% Si, 0.020% P, 52
0.008% S if
0.6 1050 5.4 x 10-7 0.45% MD, 0.14% Si, 0.035% P,
0.008% S t
1.1 1050 6.7 x 10-7 0.28% MD, 0.20% Si C'l
2.2at.% C lOOO 3.7 X 10-7 Diffusion in two-phase system d 0
1190 1.88 x 10-6
X 10-7
2.2 at. % C+ 1.98 Co 1000 4.3
58n
1190 2.19 x 10-6
2.2 at.% C + 3.91 Co 1000 4.7 X 10-7

1.0 (1.5)
1190
950 1.17
2.67
(1.3)
x 10-6
x 10-7 Decarburization of gray cast e
I
1.0 (1.5) 1000 2.83 (2.88) x 10-7 iron in CO-CO. mixtures
1.0 (1.5) 1050 4.54 (5.27) x 10-7 Composition of the steel:
1.0 (1.5) 1100 8.3 (7.1) x 10-7 0.33-0.45% Si, 0.31H1.52% MD,
in IX-Fe <800 Do = 7.9 X 10-3 18.1 0.067-0.153% P

.. Paschke, M., and A. Hauttmann: Arch. EiBenhu.ttenw. 9, 305 (1935).


b Wells, C., and R. F. Meh1: Metala Tech. 7, Tech. Publ. No. 1180 (1940).
o Darken, L. S.: Trans. AIME, Metals Tech. 15, Tech. Publ. No. 2311 and 2443 (1948).
d Smoluchowski, R.: Phya. Rev. 62, 539 (1942).
e Baukloh, W., F. Schulte, and H. Friedrichs: Arch. Eiaenhu.ttenw. 16, 341 (1943).
r Thomas, W. R., and G. M. Leak: Phil. Mag. 45, 986 (1954).

~
Table 5. Diffusion Coefficients of Hydrogen, Deuterium, Oxygen, and Nitrogen t
in Titanium, Zirconium, Niobium, and Tantalum

Diffusing gas Diffusion medium Temperature range, °C Do, cm 2 /sec L1 U, kcal/g- Additional information Refer-
atom ence
~
Hydrogen ",-Ti 500-824 1.8 x 10 2 12.4 ± 0.7 pH2 = 760 mm Hg a
t"'
{3-Ti 800-1000 1.95 x 10-3 6.64 ± 0.5
Zr 60-250 1.09 X 10- 3 11.4 pH2 = 50 mm Hg b ~
n-/II
Deuterium Zr 100-225 0.73 x 10- 3 DH/DD = 1.5 b o
/II
;-
(')

Nitrogen ",·Ti 900-1600 1.2 x 10- 2 45.3 ± 2.3 Measured above the c
...
'"CI
{3-Ti 900-1600 3.5 x 10- 2 33.8 ± 1.4 '" ~ {3 transition point ;-
TiN 1000-1450 5.4 x 10- 3 52.0 ± 3.5 1 atmN2
{3-Zr 920-1640 1.5 x 10- 2 30.7 d 15
Ta 1.23 x 10- 2 39.8 e
50-2000 x 10- 2 ~
Nb 9.8 38.6
~
Oxygen 1.6 48.2 According to measured [
",·Ti 700 and 800 D ~ 10-10 and D ~ 10- 9 data of Jenkins
Ta 1.9 x 10- 2 27.3 e o
50-2000 1.47 x 10- 2
Nb 27.6 ~
Zr 390-620 9.4 51.78 , fIJ

:I
Carbon Ti 5.6 43.5 ",-phase h '"CI
736-1150
108 48.4 {3-phase a
(')
/II
fIJ
a Wasilewski, R. J., and G. L. Kehl: Metallurgia 46,225 (1955). fIJ
/II
b Gulbransen, E. A., and K. F. Andrew: J. Electrochem. Soc. 101, 560 (1954). fIJ
C Wasilewski, R. J., and G. L. Kehl: J. Inst. Metals 83,94 (1954/55).

d Mallett, M. W., J. Belle, and B. B. Cleland: J. Electrochem. Soc. 101, 1 (1954).


e Ang, C. Y.: Acta Met. I, 123 (1953).
r Jenkins, A. E.: J. Inst. Metals 82, 213 (1953/54).
g Pemsler: J. Electrochem. Soc. 105,760 (1958).
h Bucur, E., and F. C. Wagner: Final Tech. Rep., Contract DA-36-034-0RD-1157-(NP-5502), Sept. 1954.
Ta.ble 6. Diffusion Data for a Few Metals in Copper and Copper Alloys at Different Concentrations
and Tempera.tures (Cu: Cubic-Hexoctahedral)

Concentration of .du,
Diffusing Cu alloy, T,OC D, cm2 /sec Do, cm 2/sec kcal/g- Measurement method Refer-
metal at.% atom ence

Radioactive indicators
Cu (self-diffusion) 685-1062 0.27 47 (a) Measurement of surface •
activity
685-1062 0.20 47.12 (b) Polishing off of very thin b.o
layers t;:j

630 5.6 x 10- 5 0.011 9.2 d


-
~
~
H (266-651°C) Volumetric
8
C'l
Ag 1 710 0.012 35.6 0 0
ID
860
!:II
Al 0.8 800 1.75 X 10- 2 37.7 Analysis of thin pieces e
fIJ
4 800 4.55
3.75
X
X
10-2
10-1
39.5
I
8 800 43
12 800 6.76 48
16 800 253 54

Au 750-1000 5 x 10- 20 (20°C) 0.1 44.9 Radioactive Au, cut into thin
pieces

Be 0 800 2.32 x 10-4 28 Analysis of thin pieces e


4 800 7.13 X 10-4 30
8 800 3.18 X 10- 2 37

(continued)
-0

Ot
0
Table 6 continued

Concentration of LlU,
Diffusing Cu alloy, T,oC D, cm 2 /sec Do, cm 2 /sec kcal/g- Measurement method Refer-
metal at.% atom ence
~
e t"
Cd 0 800 1.97 X 10- 9 8· Analysis of thin pieces III
0.5 800 1.97 x 10-8 10 an
1 710-860 0.0034 29.2 ID
t::I
ID
Mn 8-11A 400 2.0 x 10-13 0.72 X 10- 5 23.2 X-ray measurement of the lattice g :;-
650 3.7 x 10-11 constants ~
850 1.3 x 10-10 "1:1
1:1"
ID
g
Ni 7.5-11.8 550 7.1 x 10-13 6.5 X 10- 5 29.8 X-ray measurement of the lattice g
700 104 x 10-11 constants
e
ID
950 2.1 x 10-10 g

Pd 4.3-6.2 490 9.0 x 10-13 0.16 X 10- 5 21.9 X-ray measurement of the lattice g
!t::I
700 1.3 X 10-11 constants
950 2.5 x 10-10 ~
GIl

g
8
Pt 2.4-3.5 490 5.8 x 10-13 1.02 X 10-4 21.9 X-ray measurement of the lattice "1:1
700 1.3 X 10-11 0041 52.5 constants C!n
960 1.10-2.32 x 10-10 0.41 52.5 a ID
GIl
GIl
ID
GIl
Si 0 800 3.7 X 10- 2 40.0 Analysis of thin pieces e
4 800 4.06 X 10-1 48.2
8 800 18.6 53.8

(continued)
Table 6 continued

Sn o 800 11.3 45 Analysis of thin pieces e


4 800 3.25 x 10-3 30.5
5.6-8.0 475 1.8 x lO-l1 26 b
600 2.1-2.8 X lO-10
708 8.7 x 10-10

Zn 0--18 800 5.8 X lO-4 42.0 Analysis of thin pieces e


0--9.25 641 5.1 x lO-14 5.8 X lO-4 42.0
884 6.4 x lO-12 5.8 X lO-4 42.0
0--28.6 641 2.3 x lO-13 3.2 X lO-3 42.0 Vaporization of liquid Zn out of
884 3.4 x 10-11 3.2 X lO-3 42.0 the alloy
X lO-7
~
27.5-35.4 700 0.6 24.5
950 1.44 x 10- 6 24.5 ~
Brass 800 1.2 -3.2 x 10-8 k
Brass 800 1.16--1.39 X 10- 7 ~
1 710--860 2.4 x lO-3 30.2 k

a Nowick, A. S.: J. Appl. Phys. 22, 1182 (1951); values calculated by the author.
b Kuper, A., H. Letaw, Jr., L. Slifkin, E. Sonder, and C. T. Tomizuka: Phys. Rev. 96, 1224 (1954).
C Kubaschewski, 0.: Trans. Faraday Soc. 46, 713 (1950).
d Eichenauer, W., and A. Pebler: Z. MetaUk. 48, 373 (1957).
t
a
GIl

e Rhines, F. N., and R. F. Mehl: Trans. AIME 128, 185 (1938).


r Martin, A. B., and F. Asaro: Phys. Rev. 80, 123 (1950); A. B. Martin, R. D. Johnson, and F. Asaro: J. Appl. Phys. 25,364 (1954).
Strong grain boundary diffusion is noticeable below 700°C.
g Matano, C.: J. Phys. (Japan) 9, 41 (1934).
h Vero, J. A.: Kgl. ungar. Palatin-Joseph.Univ. tech. u. Wirtschaftswiss. Sopron, Mitt. berg-u. hUttenmann. Abt. 12, 141 (1940).
I Dunn, J. S.: J. Chem. Soc. 129,2973 (1926).
j Bugakow, W., and W. Neskutschaw: Zhur. Tekh. Fiz. 4, 1342 (1934); 9, 1767 (1939); Bugakow, W., and F. Rybalko: Zhur. Tekh. Fiz.
5, 1729 (1935).
k Seith, W., and W. Krauss: Z. Electrochem. 44, 98 (1938).
1 Thomas, D. E., and C. E. Birchenall: J. Metals 4,867 (1952); R. W. Balluffi and B. H. Alexander: J. Metals 4,1315 (1952).

CII
...
CI1
tw
Table 7. Diffusion Data for a Few Metals in Gold and Gold Alloys at Different Concentrations
and Temperatures (Au: Cubic-Hexoctahedral)

Concentration of
Diffusing the foreign metal T,oC D, cm2jsec Do, cm 2jsec .dU, Measurement method Refer-
metal in Au, at.% kcaljg-atom ence
~
Au 198 800 1.8 X 10-10 2.04 X 10- 2 45.0 By means of radioactive isotopes a t'"
II)
(self-diffusion) 900 1.6 X 10- 9 (a) Sandwich diffusion and de- b
1000 4.2 x 10- 9 termination by autoradiogram
a
~
721 3.7 x 10- 11 2.04 X 10- 2 51.0 (b) Measurement of the surface "0
869 1.3 x 10- 9 activity ~
966 4.7 x 10- 9 ;0
702-899 1.17 x 10-1 42.1 "...'tI
1:1'
Ag 8.77 806 7.6 x 10-10 4 X 10- 2 37.0 Spectral analysis of sliced layers d
858 1.6 x 10-9
B
0
891 2.9 x 10-9 e
931 4.4 x 10-9
966 6.7 x 10-9 i
II)
1003 7.5 x 10-9 1:1
1017 1.4 x 10- 8 Q.

Cu From pure Cu 301 1.5 x 10-13 1.06 x 10-3 27.4 X-ray measurement of the lattice e
III
560 9.4 x 10- 11 constants
0
616 2.2 x 10-10 1:1
-~
25.6 443 2.4 x 10-12 5.8 X 10- 3 27.4
648 1.7 x 10-10 .,'tI
0
740 9.3 x 10-10
~
"
III
III
~
Fe 18.3 753 5.4 x 10-10 1.16 X 10-4 24.4 X-ray measurement of the lattice g III
903 3.1 x 10- 9 constants
701-948 8.2 x 10- 2 41.6

(continued)
(Table 7 continued)

Co 702-948 6.8 X 10- 2 41.6 c

Ni 15.0 800 7.7 x 10-10 1.74 X 10-3 31.2 X-ray measurement of the lattice g
902 2.3 x 10- 9 constants
1003 6.9 x 10-9
702-988 3.4 X 10- 2 42.0

Pd 17.1 727 5.8 x 10-12 1.13 X 10-3 37.4 X-ray measurement of the lattice h
820 5.2 x 10-11 constants
970 3.2 x 10-10

Pt 20.1 740 4.7 x 10-12 1.24 X 10-3 39.0 X-ray measurement of the lattice h
824 2.2 x 10-11 constants
986 1.7-2.8 x 10-10
r
n
Po 470 1.2 x 10-14

Ra (B + C) 470 3.9 x 10-12 CIl


I
a Gatos, H. C., and A. Azzam: J. Meta18 4, 407 (1952).
b McKay, H. A. C.: Trans. Faraday Soc. 34,845 (1938); see also A. M. Sagrubsky: Bull. acado 8ci. URSS (1937),903: Physik. Z. Sow-
jetunion 12, 1I8 (1937).
C Duhl, D., K. Hirano, and M. Cohen: Acta_ Met. 11, I (1963).
d Ebert, H., and G. Trommsdorf: Z. Elecktrochem. 54, 294 (1950).
e Jost, W.: Z. phY8ik. Chem. (B) 16, 123 (1932).
fLiempt, J. A. M. van: Rec. trav. chim. 60, 634 (1941).
g Kubaschewski, 0., and H. Ebert: Z. Elektrochem. 50, 138 (1944).
h Jost, W.: Z. phY8ik. Chem. (B) 21, 158 (1933).
1 Wertenstein, M. L., and H. Dobrowolska: J. phY8. radium 4, 324 (1923).

g:
U"I
Table 8. Diffusion Data for a Few Metals in Silver and Silver Alloys at Different Concentra.tions ...
and Temperatures (Ag: Cubic Hexoctahedra.l)

Concentration of
Diffusing the foreign metal T,oC D, cm2 jsec Do, cm 2 jsec ..1u, Measurement method Refer-
metal in Ag, at.% kcaljg-atom ence ~
t"'
Ag105 725 8.0 x 10-11 0.835 44.9 Radioactive indicators and cut- •
(self·diffusion) 800 4.0 X 10- 10 0.895 45.9 ting into thin pieces b a
c,d n
(D
875 1.5 x 10-9 0.89 46
903.5 2.55 x 10- 9 Measurement of the radioactivity t:I
(D
950 5.5 x 10-9 on both sides of the 30-110 p. ;-
800-900 1.8 47.4 thick samples e
~
."
H 388-600 2.82 x 10-3 7.5 Pressure measurement =-
g
Al 7.5-14.2 500 10-10 Microhardness testing I
a
m
b III
Au 49.2 1.2 x 10-1 44.1 Radioactive indicators
From pure Au 218 2.6-6.6 x 10-17 8-
456 4.9 x 10-13 5.3 X 10-4 29.8 Metallographic methods t:I
601 1.1 x 10-11 ;
870 4.3 x 10-10 GIl
18.4 767 3.2 x 10-10 1.1 X 10-4 26.6 Analysis of thin pieces 0
1:1
847
916
6.4 x 10-10
1.5 x 10-9
-
."
.,0
n
(D
Cd 2.0 592-937 0.44 41.7 Spectral analysis of thin pieces k GIl
GIl
650 2.6 x 10-10 4.85 X 10- 6 22.35 (D
GIl
800 1.4 X 10-9
895 1.3 x 10-8
5.0 800 1.3 X 10- 9 Vaporization of liquid Cd
900 6.1 x 10- 9

(continued)
(Table 8 continued)

",.phase 627 2.8 x 10- 8 20.25 Vaporization of liquid Cd m


717 8.1 x 10-8
827 2.0 x 10- 7
,a.phase 400 3.2 x 10- 9 9.0 Vaporization of liquid Cd n
500 1.6 X 10-8
600 3.4 X 10-8 41.0

Cu 2.0 650 8.8 x 10-11 5.95 X 10- 5 24.8 Spectral analysis of thin pieces
760 3.6 x 10-10
800 5.9 X 10-10
895 9.4 x 10-10
t:;j

In 0.1-2 500-900
613-936
0.46
0.41
40 Calculated
40.63 Cut into thin pieces k
-
~
fIJ
Q
-=
C'l
Pd 20.2 444 1.3 X 10-12 0.64 X 10- 5 20.2 X.ray determination of the lattice 0 Q
ID
571 3.7 x 10-11 constants
637 7.0 x 10-11 Ell
()
742 1.2 x 10-10 ;-
917 1.2 x 10-9 ...
fIJ
=
Ra (C + B) 470 4.4 x 10-12 P

Sb 2.0 650 3.8 x 10-10 0.73 44 Spectral analysis of thin pieces q


760 1.5 x 10-9 5.31 X 10- 5 21.7
895 4.3 x 10- 9 21.7
468-942 0.169 38.32

Sn 2.0 592-937 0.25 39.3 Spectral analysis of thin pieces k


650 6.2 x 10-10 0.63 42.5 c
760 1.4 x 10- 9 7.82 X 10- 5 21.4 q
895 7.3 x 10-9

(continued) u.
u.
CIt
(Table 8 continued) CI'

Concentration of
Diffusing the foreign metal T, DC D, cm 2 Jsec Do cm 2 Jsec .dU, Measurement method Refer-
metal in Ag, at. % kcalJg-atom ence
~
Zn 2.0 650 2.9 x 10-10 Spectral analysis of thin pieces t"'
760 1.2 x 10- 9
800 1.9 x 10-9 7.3 X 10- 5 24.4
n
840 4.8 x 10- 9
a
~
895 1.3 x 10-8 t:l
m ~
8.3 700 4.6 x 10- 9 Vaporization of liquid Zn ;-
800 7.0 X 10-9 n
850 1.2 x 10-8 ...
."
16.6 700 4.6 x 10-9 Vaporization of liquid Zn ~
=-
800 9.3 X 10- 9
850 2.3 x 10- 8
g
24.9 650 3.5 x 10-9 Vaporization of liquid Zn
750 1.3 x 10-8 gi
800 3.9 X 10-8
ex-phase 400 ~1 X 10-10 Microhardness measurement g ~
Q.
627 5.8 x 10- 8
727 1.3 x 10- 7 Vaporization of liquid Zn t:l
827 2.6 x 10- 7 14.1
(II
,B-phase 410 1.2 x 10- 8
500 1.6 X 10- 8 -8~
650 2.8 x 10-8 4.9 Vacuum vaporization ."
u '"I
Ag + 7wt.% Zn 400-500 2.8 x 10-3 27.15 0
3.75 X 10- 3 27.7 n
Ag + Zn + Au ~
(II
Ag + Zn + Ga 0.67 x 10-3 24.8 !:J.S= - 6.8 calJg-atom (II
Ag + Zn + Sn 0.38 X 10-3 24.0 ~
- 6.24 (II
Ag + Zn + Sb 0.11 x 10-3 22.0 - 9.46
Ag + Zn + Al 2.26 x 10-3 28.15 - 10.44
Ag + Zn + In 0.31 x 10-3 23.6 - 12.8
7.29
-
- 10.88
(continued)
(Table 8 continued)

Fe 600-850 9.4 X 10- 7 29.6 v

Co 600--850 3.0 X 10- 7 29.7 v

Ni 600--850 8.5 X 10- 7 28.7 v

o 400--800 3.7 X 10-3 11.0 Pressure measurement w

a Krueger, H., and H. N. Hersh: J. Metals 7, 125 (1955).


b Johnson, W. A.: Trans. AIME 143, 107 (1941).
C Nowick, A. S.: J. Appl. Phys. 22, 1182 (1951); values calculated by the author.
d Tomizuka, C. T., and D. Lazarus: J. Appl. Phys. 25, 1443 (1954). ~
e Kryukov, S. N., and A. A. Zhukovitsky: Dokl. Akad. Nauk. SSSR 90, 739 (1953).
! Eichenauer, W., H. Kunzig, and A. Pebler: Z. Metallk. 49, 220 (1958).
S-l:I
g Buckle, H.: Metallforschung 1, 47, 175 (1946). oo
b Johnson, W. A.: Trans. AIME 147, 331 (1942).
I Jost, W.: Z. physik. Ohem. (B) 9, 73 (1930).
j Braune, H.: Z. physik. Ohem. 100, 147 (1924).
k Tomizuka, C. T., and L. M. Slifkin: Phys. Rev. 96, 610 (1954).
I Birchenall, C. E.: "Volume Diffusion," in Atom Movements, ASM, Cleveland, 1951, pp. 112ff.
m Bugakow, W., and B. Sirotkin: Zhur. Tekh. Fiz. 7, 1577 (1937).
n Tomizuka, C. T., L. M. Slifkin, and D. Lazarus: Bull. Am. Phys. Soc. 28, No.2, Paper Z-10 (1953).
o Jost, W.: Z. physik. Ohem. (B) 21, 158 (1933).
I
P Wertenstein, M. L., and H. Dobrowolska: J. phys. radium 4,324 (1923).
q Seith, W., and E. Peretti: Z. Electrochem. 42, 570 (1936).
r Sonder, E., L. M. Slifkin, and C. T. Tomizuka: Phys. Rev. 93, 970 (1954).
S Rollin, B. V.: Phys. Rev. 55, 231 (1939).
t Bugakow, W., and F. Rybalko: Zhur. Tekh. Fiz. 5,1729 (1935).
U Brutzyk, M., and S. Gerzriken: Zhur. Tekh. Fiz. 20, 428 (1950).

v Hirano, K., M. Cohen, and B. L. Averbach, Acta Met. 11, 463 (1963).
W Eichenauer, W., and G. Muller: Z. Metallk. 53, 321 (1962).

~
U1
go
Table 9. Diffusion Data for a Few Metals in Aluminum and Aluminum Alloys at Different Concentrations
and Temperatures (AI: Face-Centered Cubic)

Concentration of Lfu,
Diffusing the foreign metal, T,oC D, cm2 /sec Do, cm2 /sec kcal/g- Measurement method Refer-
metal at.% atom ence ~
t""
Ag 1.26 465 1.9 x 10-10 1.1 X 106 32.6 Spectral analysis of thin pieces
1.26 573 3.5 x 10- 9 "b a
n
~
2.8-5.5 500 2.0-3.1 x 10-9 38.5
" 0~
:;-
Cu Cu-AI-eutectic 440 5.0 x 10-11 2.3 34.9 Metallographic investigation b
490 2.4 x 10-10
...n
~
540 1.4 x 10-9 ;'
0.85 457 8.0 x 10-10 8.4 X 10-2 32.6
3.05 500 1.5-5.8 x 10-10 Spectral analysis of thin pieces "b g
0.17 515 5.1 x 10-10
0.17 565 1.3-1.4 x 10- 9 "
c
I
Mg AI-Mg-eutectic 365 8.6 x 10-12 Metallographic investigation E.
400 6.4 X 10-11 1.52 X 10 2 38.5 0
440 3.3 x 10-10 0.32 30 d
1.32 450 1.9 x 10-9 Spectral analysis of thin pieces e
III
5.5-11.0 395 5.5-6.7 x 10-11 1.17 X 10-1
i...
0
447 2.6 x 10-10 28.6
577 4.4 x 10-9 " =
~
14.9 420 6.6-7.6 x 10-11
7.3 x 10-10 Chemical analysis of thin pieces
an
~
5.3-8.0 475 III
g III
520 8.7 x 10-9 38.0 ~
410 6.5 x 10-11 III
15.9

H 470-590 0.21 10.9 Pressure measurement b

(continued)
(Table 9 cantinued)

Mn 0.1 600 5 X 10-11 79.3 Microhardness testing


625 1.5 X 10-10
650 4.5 X 10-10

Si 0.50 465 3.4 X 10-10 0.36 30.5 Spectral analysis of thin pieces d
500 9.9 X 10-10 0.90 30.55
600 9.3 X 10- 9 "
1.88 510 2.0 X 10- 9 31.5 Chemical analysis of thin pieces e
0.70 550 b

Zn 0.84 415 2.5 X 10- 10 0.32 30 Spectral analysis of thin pieces d 0


473 5.3 X 10-10 11.6 27.8
555 5.0 X 10- 9 " III
4.4 450 7 X 10-10 b Q
l::I
9.4 9.5 X 10- 10 25.8 Chemical analysis of thin pieces
-~
0Q
15.1 1.2 X 10- 9
4.5-15.0 400 3.0-3.5 X 10-10 ID
Ell
;.
a Baerwald,A. H.: Z. Elektrochem. 45, 789 (1939). a
b Mehl, R. F., F. N. Rhines, and K. A. von den Steinen: Metals and Alloys 13, 41 (1941). III
• Brick, R. M., and A. Philips: Trans. AIME 124, 331 (1937).
d Nowick, A. S.: J. Appl. Phys. 22, 1182 (1951); values calculated by the author.
e Freche, H. R.: Trans. AIME 122, 326 (1936).
r Bungardt, W., and F. BoIIenrath: Z. MetaUk. 30, 377 (1938).
g Bungardt, W., and H. Cornelius: Z. MetaUk. 34, 360 (1942).
h Eichenauer, W., and A. Pebler: Z. MetaUk. 48, 373 (1957).
I Buckle, H.: Z. Elektrochem. 49, 238 (1943).

~
Table 10. Diffusion Data for a Few Metals in Lead and Lead Alioys at Different Concentrations :
and Temperatures (Pb: Face-Centered Cubic)

Concentration of Llu,
Diffusing the foreign metal, T,oC D, cm 2 /sec Do, cm 2 /sec kcal/g- Measurement method Refer-
metal at.% atom ence
~
t"4
Pb Pb (ThB) 106 1.7 x 10-16 2.7 27 Radioactive isotope; measure- ..
182 4.7 x 10-13 ment of the surface activity a
258 1.3 x 10-11 6.56 27.9 b ~
324 5.5 x 10-10 t:I
(D
;0
Ag 0.12 220 1.5 x 10- 8 7.4 X 10- 2 15.2 Spectral analysis of thin pieces c n
...
265 5.4 x 10-8 d "Ij
285 9.1 x 10-8 ::r
B
Au 0.03-0.09 100 2.3 x 10-9 3.43 X 10-1 14.0 Chemical analysis of thin pieces e
150 5.0 x 10-8
200 8.6 X 10-8
256 3.2 x 10- 7
300 1.5 X 10-6
Ii
t:I
Bi 2.0 220 4.8 x 10-11 1.5 24.5 Spectral analysis of thin pieces ..
265 2.3 x 10-10 1.83 X 10-2 18.4 g
285 4.4 x 10-10

Cd 1.0 167 4.6 x 10-11 0.55 21.0 Spectral analysis of thin pieces ..
I
..."Ij0
200 1.3 X 10-10 1.83 X 10-3 15.4 I g
8.6 x 10-10 GIl
252 GIl
(D
GIl

Hg 4.0 150 12.3


225 Vaporization of liquid Hg h
10-15 150 10.1
225
(continued)
(Table 10 continued)

In <1.0 300 1.1 X 10- 9 1.3 24 Spectral analysis of thin pieces I, a

Mg 2 220 1.2 X 10-10


270 1.3 X 10- 9 0.81 22 Spectral analysis of thin pieces a
0.26 250 2.4-3.7 X 10- 10
1.0 6.9 X 10-10
2.0 8.6 X 10-10
3.0 1.1 X 10- 9

Ni 1.0 285 2.3 X 10-10 6.6 X 10- 1 25.3 Spectral analysis of thin pieces g
3.0 252 3.5 X 10-11 a
320 3.5 X 10-10 2.7 27 0

Po 310 1.5 X 10-11 Radioactive isotope; as above k


~
fIl
Q
first series 1::1
-
C')
Sn 2.0 245 3.1 X 10-11 2.1 26 Spectral analysis of thin pieces Q
" (D
265 7.0 X 10-11 3.96 26.2 d
285 1.6 X 10-10 Ell
n
(D
TI 2.0 220 2.8 X 10- 11 2.51 X 10- 2 19.4 Spectral analysis of thin pieces a-
3.1 X 10-10 d fIl
285
8.5-53.0 270 1.1 X 10-10 1.7 25 a
300 3.5 X 10-10 1.03 24.6
315 5.8 X 10-10
"Nowick, A. S.: J. Appl. Phys. 22,1182 (1951); values calculated by the author.
b Hevesy, G. von, W. Seith, and A. Keil: Z. Physik 79,197 (1932); Z. MetaUk. 25,104 (1933).
C Seith, W., and A. Keil: Z. physik. Chem. (B) 22, 350 (1933).

d Seith, W., and J. G. Laird: Z. MetaUk. 24, 193 (1932).


e Roberts.Austen, W. C.: Phil. Trans. Roy. Soc. London (A) 187,404 (1896).
r Seith, W., and H. Etzold: Z. Elekrochem. 40, 829 (1934); 41, 122 (1935).
II Seith, W., E. Hofer, and H. Etzold: Z. Elektrochem. 40,322 (1934).
h Hertzriicken, S. D., M. Butsik, and E. Golubenko: Mem. phys. Ukr. 8, 55 (1939).
1 Seith, W.: Z. Elektrochem. 39, 538 (1933); 41, 872 (1935).
j Seith, W., and J. Herrmann: Z. Elektrochem. 46, 213 (1940).
k Hevesy, G. von, and A. Obrutschewa: Nature 115, 674 (1925). t::'o
...
Table 11. Diffusion Data for a Few Metals in Iron and Iron Alloys at Different Concentrations
~
and Temperatures (ct-Fe: Body-Centered Cubic; y-Fe: Face-Centered Cubic)

Concentration of Au,
Diffusing the foreign metal, T,oC D, cm 2/sec Do, cm2/sec kcal/g- Measurement method Refer-
metal at.%, atom ence ~

Al Al powder, pure 900 3.S x 10-9 44.0 Metallographic .. ~


n
1050 2.0 x 10- 8 ~
t::;j
~

Co In austenite 90 SO b ;-
c n
....
1140-1340 1.25 72.9 With Co60 tracers
"1:1
Cr Pure 1150 6.8 x 10-10 13.5 Chemical analysis of thin pieces d ff
1300 2.2-5.3 x 10- 8
g
Iron-Cr powder 1200 1.7-S.1 x 10-8 X-ray lattice constant e
10 b
~
in austenite 75 i
>10 at.%, SOO-1400 I.4S 54.9 Microanalysis
-----~~
;
Q.
Mn About 27 960 3.0 x 10-10 g
3 1400 9.6 x 10-8 b
0-20 950-1450 Approx. formula i
~
0
Mo 0-3.1 1200 2.3-3.0 x 10-9 Chemical analysis of thin pieces
=
"1:1
g 0
Ni About 22 1200 0.9 x 10-10 Chemical analysis of thin pieces n
'"
k ~
0-20 1050-1450 Approx. formula GI
1 GI
Pure 600 0.77 67.0 Iny-Fe ~
1050 1.33 56.0 In paramagnetic ct-Fe above GI

SOO°C
1.4 5S.7 In ferromagnetic ct-Fe below
6S0°C

(continued)
(Table 11 continued)
960 7.5 x 10-9 g
Si About 35
1150 0.45 x 10- 7 m
3-4 1435 ± 5 1.1 x 10- 7

Sn 950 9.7 x 10-10 46.0 Metallographic n


Pure 1000 2.0 x 10-9
1050 3.9 x 10-9
1100 7.6 X 10-9

W 0-1.3 1280 3.7 x 10-10 Chemical analysis of thin pieces


0-1.2 1330 2.4 x 10- 9 o
0-3.4 1.0 x 10-8 b
In austenite 13 75
a Agew, N. W., and O. J. Vher: J. IMt. Metals 44, 83 (1930).
b Gruzin, P. L.: Dokl. Akad. Nauk. SSSR 94, 681 (1954), where further information on D(Co, Cr, W) in ferrites is given.
i
~
C Suzuoka, T.: Trans. Japan. Inst. Metals 2, 176 (1961).
d Bardenheuer, P., and R. Mueller: Mitt. Kaiser- Wilhelm-Inst. Eisenforsch. DUsseldorf 14,295 (1932).
~
~

e Hicks, L. C.: Trans. AIME 118, 163 (1934). More recently Gerzriken and Deghtjar: Zhur. Tekh. Fiz. 20, 1005 (1950), investigated the
effect of Ni, Be, Ti, W, Si, Sn, and Nb on the diffusion of Cr in y-Fe. They developed formal physical relationships, such as the dependence
of the activation energy of the diffusion on the valence of the impurities and their atomic numbers.
r Heumann, T., and H. Bohmer: Arch. Eisenhuttenw. 31,749 (1960).
g Fry, A.: Stahl u. Eisen 43, 1039 (1923).
h Owen, E. A.: J. Inst. Metals 73,471 (1947).
i
I Wells, C., and R. F. Mehl: Trans. AIME 145, 315 (1941)-the empirical diffusion formula for Mn in y-Fe.
DMn y-Fe = (0.486 + 0.011 wt. % Mn) exp( - 66,000jRT)
was given, which is valid from 0-20 wt. % Mn between 950 and 1450°C, with an accuracy of ± 15%. The influence of C was accounted
for in the equation Dc = Dc (1 + 2.53 wt. % C), which is valid from 0-1.5% C, where Dc is the diffusion coefficient for C = O.
l Grube, G., and F. Liebenwihh: Z. anorg. u. allgem. Chem. 188,274 (1930). 0
k Wells, C.: Trans. AIME 145, 329 (1941)-the empirical diffusion formula for the diffusion coefficients of Ni in y-Fe
DNly-Fe = (0.344 + 0.012 wt. % Ni) exp( - 67,500JRT)
was given, which is valid from 0-20 wt. % Ni between 1050 and 1450°C, with an accuracy of ± 20 %. The influence of C was accounted
for in the equation Dc = Dco(l + 2.3 wt.% C).
1 Hirano, K., M. Cohen, and B. L. Averbach: Acta Met. 9, 440 (1961).
m Bradshaw, F. J., G. Hoyle, and K. Speight: Nature 171, 488 (1953).
n Bannister, C. N., and W. D. Jones: J. Iron Steel Inst. 124, 71 (1931).
o Grube, G., and K. Schneider: Z. anarg. u. allgem. Chem. 168, 17 (1927). 0-
~
t
Table 12. Diffusion Data for a Few Metals in Platinum and Platinum Alloys at Different Concentrations
and Temperatures (Pt: Face-Centered Cubic)

Concentration of .du, ~
Diffusing the foreign metal, T,oC D, cm2jsec Do, cm 2jsec kcaljg- Measurement method Refer- ~
metal at.% atom ence
~
II>
Au 900 3 X 10-11 Chemical analysis of pieces a t::j
b II>
;-
l4.
Cu 13.9 1041 2.2-2.5 x 10- 11 4.78 X 10- 2 X-ray determination of lattice "tj
55.7 t:r
1152 1.6 x 10-10 constants
1241 6.6 x 10-10
1350 1.4-1.5 x 10- 9
1401 1.7 x 10- 9
I~
Ir Ir and Pt Powder d
&
t::j
Ni 14.9 1043 5.2 x 10-11 7.75 X 10-4 43.1 X-ray determination of lattice
1241 4.8 x 10-10 constants ~
(Ij
1401 1.5 x 10- 9
--------
g
"tj
Ra (B + C) 470 9.3 x 10-12 Measurement of surface activity e
a~
(Ij
(Ij
• Jedele, A.: Z. Elektrochem. 39, 691 (1933). II>
(Ij
b Matano, C.: Proc. Phys.-Math. Soc. Japan 15, 405 (1933).
c Kubaschewski, 0., and H. Ebert: Z. Elektrochem. 50, 138 (1944).
d Zogagintsev, J.: Appl. Chem. USSR 17, 22 (1944).
e Wertenstein, M. L., and H. Dobrowolska: J. phys. radium 4, 324 (1923).
Table 13. Diffusion Data for a Few Metals in Nickel at Different Temperatures
(Ni; IX ~ {3; IX = Face-Centered Cubic and {3 = Hexagonal)

Concentration of ..::lu,
Diffusing the foreign metal, T, DC D, cm 2 /sec Do, cm 2 /sec kcal/g- Measurement method Refer-
metal at.% atom ence

Al 1100-1300 1.87 64.0 Alloy/Ni of diffusion couples &

Au 900 ~7 X 10-11 Chemical analysis of pieces b t:J


~
fIl
0
Cu 650 3.9-4.3 x 10-12 1.01 X 10- 3 35.5 X-ray determination of lattice c 1:1
-
890 1.8-2.4 x 10-10 constants d 0
0
~
$
n
In Pure Ni 1000 1.2 x 10-10 X-ray determination of lattice e ~
Ni + 0.5% Mn 0.3 X 10-10 constants 1:1
...-
fIl

Mn 970 9.2 x 10-11 g


1100-1300 7.5 67.1 Alloy /Ni diffusion couples &

Mg Pure Ni 1050-1300 0.44 56 Mg-Ni alloy/pure Ni h

Mo Pure Ni 1150-1400 3.0 68.9 Mo-Ni alloy/pure Ni h

(continued)
~
Qo
a-
a-
(Table 13 continued)

Concentration of .dU,
Diffusing the foreign metal T, DC D, cm 2 /sec Do, cm 2 /sec kcal/g- Measurement method Refer-
metal at.% atom ence
~
Si Pure Ni II 10-1300 1.5 61.7 Si-Ni alloy/pure Ni b t"'

n
Ti Pure Ni 1l00-1300 0.86 61.4 Alloy/Ni diffusion couples &
a
(D

I:'
(D

W Pure Ni 1l00-1300 11.1 76.8 Alloy/Ni diffusion couples & ;-


...n
'tI
Co 800-1000 1.46 68.3 Measurement of surface activity ::r
ID
l:I
Q

Mg 720 ~10-9 Vaporization of Mg a


ID
i
a Swalin, R. A., and A. Martin: J. Metals, Trans. AIME 206,567 (1956).
b Jedele, A.: Z. Elektrochem. 39, 691 (1933). Q.
C Matano, C.: Proc. Phys.-Math. Soc. Japan 15, 405 (1933).
=
I:'
d Matano, C.: Mem. Coll. Sci., Kyoto Imp. Univ. 15, 351 (1932).
e Grube, G., and A. Jedele: Z. Elektrochem. 38, 799 (1932). III
f Matano, C.: J. Phys. Soc. Japan 8, 109 (1933). Q
g Smithells, C. J.: Metals Reference Book, London 1949, p. 406. l:I
h Swalin, R. A., A. Martin, and R. Olson: J. Metals, Trans. AIME 207,936 (1957).
-~
i Ruder, R. C., and C. E. Birchenall: J. Metals, Trans. AIME 191, 142 (1951).
.,'tI
Q
j Rouse, G. F., and R. Forman: Phys. Rev. (2) 82, 574 (1951). n
ID
III
III
ID
III
Table 14. Diffusion Coefficients of Co-60 in a Few 3-AI-Ni Alloys and Mo as a. Function
of Tempera.ture and Composition, According to Berkowitz, Jaumot, and Nixa

Concentration of D, cm 2 /sec AU,


of Ni in AI, 1050°C 1150°C 1250°C 1350°C Do, cm 2/sec kcaI/g- Refer-
at.% atom ence
GIl
i
5"
47.3 2.5 x 10-11 8.85 X 10-11 3.68 X lO-10 1.05 X 10- 9 4.7 X 10-2 56.6 a
48.5 l.66 x 10-11 6.00 X 10-11 2.30 X 10-10 8.32 X 10-10 9.3 X lO-2 59.9 C')
"
49.4 3.96 X 10-11 l.14 X lO-10 3.96 X 10- 10 4.4 X lO-3 52.5 0
ID
50.7 80.6 2.46 X 10-11 l.52 X lO-1O 8.36 X 10-10
53.1 2.44 X 10-11 1.03 X 10-10 5.66 X lO-10 l.97 X 10- 9 2.6 67.6
Ell
55.5 7.84 X 10-11 4.00 X 10-10 1.38 X lO-9 3.11 X lO-9 7.2 X 10-3 47.1 ~
a
GIl

Mo Between 2.82 X lO-6 34.8 b


900 and 1700°C

a Berkowitz, A. E., F. E. Jaumot, Jr., and F. C. Nix: Phys. Rev. 95, 1185 (1954).
b Byron, E. S., and V. E. Lambert: J. Electrochem. Soc. 102, 38 (1955).

cr-
':I
~
00

~
Table 15. Diffusion Coefficients of Foreign Ions of Other Valences in Heterotype Solid Solutions t""

Diffusing ion Diffusion medium Diffusion coefficient D x 1()4, cm 2/sec Refer- i


ence ~
ID
2 3 4 ;-
l4.
&
Cd 2 + AgBr {O -0.1 mol.% 0.083} 0.03} 0.013}
concentration interval 0.6-1 mol.% 0.24 [672°K] 0.13 [622°K] 0.1 [572°K] 0.026\[5220K]
2 -4 mol.% 0.9 0.7 0.11 0.028)

Pb 2 + AgCl 2.4 x 1O-3[542°K, and PbCl2 (0.63 mol. %) saturated Agel] b

AgBr {O -0.1 mo\.%


concentration interval 0.6--1 mol.:YO
2 -4 mol. Yo
2 x
0.1
0.12
1O- 2 }
[622°K],
1 x
1 x
1 x
1O- 3 }
10-3
10- 3
[472°K] a
I
~
Q.

SbH AgSbS 2 4 X 10- 7 [673°K] self-diffusion


~
III
Ag3SbS3 > 10- 9 [673°K] self-diffusion c 0-
::s
a Schone, E., O. Stasiw, and J. Teltow: Z. physik. Chem. 197, 145 (1951).
b Wagner, C.: J. Chem. Phys. 18, 1227 (1950).
III
C Rickert, H., and C. Wagner: Z. Elektrochem. 64, 793 (1960).
1
III
ID
III
Table 16. Diffusion Coefficients of a Few Foreign Ions in Germanium, Silicon, and Quartz

Diffusion Diffusing foreign Refer.


medium atom T, °0 D, cm 2 /sec Do, cm 2 /sec .dU, kcal/g·atom ence

Self·diffusion 700-900 87 73.5 &

Vacancy diffusion 3.9 23.5


t::l
Ge Ou 400-900 6.2 2.94 b
c
~
-
C'Il
700-900 1.9 x 10-4 4.1
Q
B 800 6 X 10-13 ::s
-
0Q
(p
Sb124 837 5.5 x 10-11 !3l
900 2.1 X 10-10 10 57.0 d n
(p
Ga 800 3 x 10-12 e -a
As 700-900 1 x 10-11 3 C'Il

Ni 800 5 x 10- 9 0.8 21 g


In 900 2 X 10-13 e
As 800 x 10-1L5 X 10-12 2.1 55 h
Li 1 x 10- 9 2.5 X 10- 3 ll.8
P 900 8 X 10-11 e
Fe 775 1 x 10- 6 25.0
930 4.3 x 10- 6 k

(continued)

t;..
-=
(Table 16 continued)
~

Diffusion Diffusing foreign Refer.


medium atom T,oC D, cm 2 /sec Do, cm 2 /sec L1 U, kcal/g·atom ence
!-I
Si B,P HOO 4 X 10-13 10.5 85
B 1000-1300 1 x 10-3 58.0 m
Li HOO 2 x 10- 5 9.4 x 10 18.1 n
Au 700-1300 O.OOH 25.8 i
Fe 900-1400 0.0062 20.0 t:I
~
Cu 200 5 X 10- 5 ;-
~
Si0 2 Li+ 300-500 6.9 x 10-3 20.5 0

Na+ Parallel to the c·axis 3.6 X 10-3 24.0


K+ 0.18 31.7

a Letaw, R., Jr., M. L. Slifkin, and W. M. Portnoy: Phys. Rev. 93, 892 (1954).
b Penning, P.: Phys. Rev. 110, 586 (1958).
c Fuller, C. S., J. D. Struthers, J. A. Ditzenberger, and K. B. Wolfstirn: Phys. Rev. 93, H82 (1954).
d Dunlap, W. C., Jr.: PhY8. Rev. 94, 1531 (1954).
i
[
t:I
e Dunlap, W. C., Jr.: Phys. Rev. 86, 615 (1952).
f Albers, W.: Solid State Electronics 2, 85 (1961).
g van der Maesen, F., and J. A. Brenkman: Philips Res. Rept. 9, 225 (1954).
h Mcafee, K. B., W. Shockley, and M. Sparks: Phys. Rev. 86,137 (1952); W. Bosenberg: Z. Naturfor8ch. lOa, 285 (1955).
I Fuller, C. S., and J. C. Severiens: Phys. Rev. 96, 21 (1954).
j Fuller, C. S., and J. A. Ditzenberger: J. Appl. Phys. 27, 544 (1956).
i
"C
~
k Bugai, A. A., V. E. Rosenko, and E. G. Miselyuk: Zhur. Tekh. Fiz. 27, 207 (1957).
I Fuller, C. S., and J. A. Ditzenberger: J. Appl. PhY8. 25, 1439 (1954). i
CD
m Fuller, C. S., and J. A. Ditzenberger: Phys. Rev. 91, 193 (1953).
n Struthers, J. D.: J. Appl. Phys. 27, 1560 (1956).
I:
o Verhoogen, J.: Am. Mineralogist 37,637 (1952).
Table 17. Additional Diffusion Data at Different Concentrations and Temperatures

.dU,
Base Diffusing kcaljg- Measurement Refer-
metal metal at.% T,oC D, cm 2 jsec Do, cm 2/sec atom method ence

Cd Hg 4 156 2.7 x 10-10 2.6 19.6 Vaporization a


202 2.5 x 10- 9
Ph 2 252 8 X 10-12 Spectral analysis of ..
thin pieces

0
Cr Co 0-40 1000-1360 0.443 63.6 b
Fe Dilute solution 1104-1406 0.21 62.7
~
UJ

::I
Q
In Tl204 50-155 0.049 15.5 d 0
~

Eln l
~
Mo B 900-1300 8.84 X 10- 6 12.2
Th 1615 3.6 x 10-10 Thermionic emission
-
...::I
UJ
2000 1.0 X 10- 6

W B 900-1300 1.3 x 10- 5 17.2 e


C 1702-1727 0.31 59.0 Thermionic emission g
Fe 0.13 1927-2527 11.5 140.0 h
Mo 1533-2260 5 x 10-3 80.5 Chemical analysis of
thin pieces
U 1727 1.30 x 10-11 1.14 100.0
Y 1.82 X 10-8 0.11 62.0 Thermionic emission
Zr 3.24 X 10- 9 1.1 78.0

(continued) ...:J
...
.....
t.-
(Table 17 continued)

LlU,
Base Diffusing kcal/g- Measurement Refer-
metal metal at.% T, DC D, cm 2 /sec Do, cm 2 /sec atom method ence
~
k t'"
u Zr 12.5 950 Du = 7.7 X 10- 9 ~
Dzr = 6.5 X 10- 10 Du/Dzr = 11.8 ..........
n
U.5 1000 Du = 1.6 X 10-8 LlUu = 36 ~

Dzr = 1.0 X 10- 9 Du/Dzr = 16 LlUzr 40


= t:I
~
9.5 1075 Du = 3.7 X 10- 8 ~
Dzr = 2.9 X 10- 9 Du/Dzr = 12.8 n
...
Ti 18.0 950 Du = 4.7 X 10- 9 'tI
Du/DTi = 3.9 ::r
~
DTi = 1.2 X 10- 9
:;3
0
18.0 1000 Du = 9.5 X 10- 9 LlU u = 38.5
Du/DTi = 3.3 S
DTi = 2.9 X 10- 9 LlUT! = 40.0 ~
:;3
16.5 1075 Du = 2.2 X 10-8 ~

DTi = 5.8 X 10- 9 Du/DTi = 3.8 ~


:;3
Co
t:I
8 Seith, W., E. Hofer, and H. Etzold: Z. Elektrochem. 40, 322 (1934).
b Weeton, E. W.: Nat. Advis. Comm. Aeronaut. Rept. 1951, 1.
C Mead, H. W., and C. E. Birchenall: J. Metals 7,994 (1955).
-~
Ul

dEckert, R. E., and H. G. Drickamer: J. Chem. Phys. 20, 13 (1952).



:;3
e Samsonov, G. V.: Dokl. Akad. Nauk SSSR 93, 859 (1953). 'tI
f Nelting, H.: Z. Physik 115, 469 (1940). '"I
0
g Pirani, M., and J. Sandor: J. Inst. Metals 73, 385 (1947). n
~
b Liempt, I. A. M. van: Rec. trav. chim. 64, 239 (1945). Ul
Ul
I Liempt, I. A. M. van: Rec. trav. chim. 51, U4 (1932). ~
Ul
j Dushman, S., D. Denissen, and N. B. Reynolds: Phys. Rev. 29,903 (1927).
k Adda, Y., J. Philibert, and H. Farraggi: Rev. Met. 54, 597 (1957).
1 Adda, Y., and J. Philibert: Acta Met. 8, 700 (1960).
Table 18. Some Self-Diffusion Coefficients in Metals

Temperature Diffusing Diffusion Diffusion formula, cm 2 jsec Additional experimental Refer-


range metal medium conditions ence

1104-1406 D = 0.S3 exp( -67,700jRT) a


1000-1250 C0 60 Co D = 0.032 exp( - 61,900jRT) b
1050--1250 D = 0.367 exp( -67,000jRT)
725-950 A g l05 Ag D = 0.S95 exp( - 49,950jRT) d
Dk = 0.03 exp( - 20,200jRT)
640-903 D = 0.11 exp( -40,SOOjRT) Subscript k = grain boun- e
dary diffusion rJl
SOO-1000 AU l9S Au D = Do exp( -45,000jRT) !Eo
721-966 D = 2.04 X 10- 2 exp( - 51,000jRT) g 'i"
tl
50-100 Cd Cd D"II = 0,05 exp( -IS,200jRT) h
D"J. = 0.10 exp( -19,100jRT) ~
Dk = 1.1 exp( -13,000jRT) ...-
0'"
6S5-1062 Cu 65 Cu D = 0.20exp(-47,120jRT) ;:s
720--900 Fe 59 <x-Fe D = 2.3 X 10-3 exp( -73,200fRT) 0
970-1357 y-Fe D = 5.S exp( -74,200jRT) 0
(D
<155 In 114 In D = 1.02 exp( -17 ,900/RT) k
46S-627 Mg2S Mg D = 1.0 exp( - 32,000/RT) Sl
m
O.
(D
0--95 Na 22 Na D = 0.242 exp( -10,450jRT)
0-95 D = 0.176 exp( -12,060/RT) At SOOO kg/cm 2 pressure n ...;:s
200--300 Pb(ThB) Pb D = 6.56 exp( - 27,900jRT) 0

IS0--223 Sn l21 Sn DII = 1.2 X 10- 5 exp( -10,500jRT) Self·diffusion parallel and p
'"
DJ. = 3.7 x 10- S exp(-5900/RT) perpendicular to the tetra-
gonal c axis
253-393 Zn 65 Zn Do = I.S40 exp( -19,600/RT) Parallel and perpendicular q
DJ. = 1.3S x 10-5 exp(-25,900jRT) to the c axis at 1 atm and
D, = 4.S X 10-4 exp( - 25,000jRT) SOOO atm
D 1. = I.S X 10- 7 exp( - 32,000jRT)
Pt Pt D = 0.33 exp( - 6S,200jRT)
Ni Ni D = 1.27 exp( - 66,SOOjRT)
750--1025 Au l9S Au+ o at.% Ni D = 0.26 exp( -45,300jRT)
795-950 Au + 20 at.% Ni D = 0.05 exp( - 40,200/RT)

(continued) "'I
~
(Table 18 continued) ~
Temperature Diffusing Diffusion Diffusion formula, cm 2 jsec Additional experimental Refer·
range metal medium conditions ence
850-925 Au + 35 at.% Ni D = 0.06exp(-42,700jRT)
805-940 Au + 50 at.% Ni D = 0.09 exp( - 43,400jRT) ~
850-925 Au + 65 at.% Ni D = 0.51 exp( -48,800jRT)
t"'
850-1000 Au + 80 at.% Ni D = 1.1 exp( - 60,500jRT)
900-1100 Ni D = 2.0 exp( - 65,000jRT) ~
900-1250 C0 60 Ni D = 1.46 exp( - 68,300jRT) a n'
/I)
150-275 Ti 204 Ti u
Due.axls = 0.4 exp( - 22,900jRT) tj
D.L = 0.4 exp( - 22,600jRT) /I)

~
a Mead, H. W., and C. E. Birchenall: J. Metals 7,994 (1955). g.
bRuder, R. C., and C. E. Birchenall: J. Metals 3,142 (1951). '"C
C Nix, F. C., and F. E. Jaumot, Jr.: Phys. Rev. (2) 80, 119 (1950); 82, 72 (1951); see also G. W. Callendine, Jr., V. C. Ridolfo, and ::r
/I)
M. L. Pool: Phys. Rev. 86, 642 (1952).
d Hoffman, R. E., and D. Turnbull: J. Appl. Phys. 22, 634 (1951).
e Johnson, R. D., and A. B. Martin: Phys. Rev. 86, 642 (1952). /I)
~
f Gatos, H. C., and A. Azzam: J. Metals 4, 407 (1952). ;
g McKay, H. A. C.: Trans. Faraday Soc. 34, 845 (1938); see also A. M. Sagrubsky: Physik. Z. Sowjetunion 12, 118 (1937).
h Wajda, E. S., G. A. Shirn, and H. B. Huntington: Acta Met. 3,39 (1955). §
I Kuper, A., H. Letaw, Jr., L. Slifkin, E. Sonder, and C. T. Tomizuka: Phys. Rev. 96, 1224 (1954). c:o
j Birchenall, C. E., and R. F. Mehl: Trans. AIME 188, 144 (1950); P. L. Gruzin, J. W. Kornew, and G. W. Kurdjumow: Dokl- Akad. tj
Nauk. SSSR 80,49 (1951) studied the influence on the self·diffusion of iron.
kEckert, R. E., and H. G. Drickamer: J. Chem. Phys. 20,13 (1952). ~
I Shewman, P. G., and F. N. Rhines: J. Metals 6, 1021 (1954). '"....o
mNachtrieb, N. H., E. Catalano, and J. A. Weil: J. Chem. Phys. 20, 1185 (1952); R. E. Meyer, and N. H. Nachtrieb: J. Chem. Phys. =
23, 405 (1955). '"C
n Nachtrieb, N. H., J. A. Weil, E. Catalano, and A. W. Lawson: J. Chem. Phys. 20, 1289 (1952). 8n
o Hevesy, G. von, and A. Obrutscheva: Nature 115, 674 (1925); W. Seith, and A. Keil: Z. Physik 79,197 (1932); B. Okkerse: Acta Met. /I)

2,551 (1954) found grain boundary diffusion in lead below 260°C.


/I)
P Fensham, P. J.: Austrian J. Sci. Res. (A) 3,105 (1950).
'"'"
q Liu, T., and H. G. Drickamer: J. Chem. Phys. 22, 312 (1954); see also G. A. Shirn, E. S. Wajda, and H. B. Huntington: Acta Met. '"
1,513 (1953); P. H. Miller, Jr., and F. R. Banks: Phys. Rev. 61, 648 (1942); H. C. Gatos, and A. D. Kurtz: J. Metals 6,616 (1954).
r Kidson, G. V., and R. Ross: Proceedings International Conference on One Use of Radioisotopes in Science Research, Paris 1957.
S Hoffman, R. E., F. W. Picus, and R. A. Ward: Trans. AIME 206, 483 (1956).
t Kurtz, A. D., B. L. Averbach, and M. Cohen: Wright Air Development Center under Contract AF 33 (038)-23281; see also A. D.
Kurtz, Thesis, M.LT., Cambridge, 1952.
U Shirn, G. A.: Acta Met. 3, 87 (1955).
Table 19. Self-Diffusion Coefficients of a Few Ions in Solid Inorganic Compounds

Diffusing Diffusion LlU, Refer-


ion medium T,oC D, cm2/sec Do, cm 2/sec kcal/g-atom ence

AgI06 IX-Ag 2S04 430-700 2.4 26.7 &

p-Ag2S04 250-430 6.7 x lO-5 13.4


AgBr 300 1.02 x lO-7 b
20 _10-15
P-Ag2S 179 3.8 x lO-7 1.4 X lO-2 11.1
IX-Ag2S 200-400 2.8 x lO-4 3.45 d
IX-Agl 2 x 10-5 e rn
(I>
300
190 -1 x lO-5 f
IX·Ag zHg14 127 (extrapolated)
IF
0
2 X lO-6 g

Ba140 BaO 1080-1220 lO- L I0- 11 As neutral particles,


r--g
about a factor of 20 C'l
smaller via vacancies b 0
(I>
BaO-SrO-Cathode <lOOO -96 I
>lOOO -9.5
!31
n
Ba131 a, J (I>
BaSi03 500 80
BaTi03 0.8 89
-a
GIl

BiS2 NaBr 500-700 47.5 k


50

Ca45 CaFeZ04 800-1200 30 86.0


IX·CaSi03 Preliminary measured 7 x 104 112 m
p-CaSiOa data without knowledge 0.2 78
CaaSi z07 ofT! lO-2 73
IX-Ca zSi04 2 x 10- 2 55
IX'-Ca zSi04 1200-1500 3.6 x 10-2 65
CaO 900-1300 0.4 81

(continued) ~
U1
~
(Tahle 19 continued) 0-

Diffusing Diffusion .dU, Refer-


ion medium T,oC D, cm 2/sec Do, cm 2/sec kcal/g-atom ence

Ni NiO(po, = 1 atm) 1000-1400 4.4 X 10-3 44.2 n


~

C0 60 0 I:"'
CoO(po, = 1 atm) 800-1350 2.15 X 10-3 34.5
n
Cu 64 CU20 800-1050 0.0436 36.1 p
a
It
q t:)
0 18 1030-1120 6.5 X 10-3 39 ± 4
m.
n
Fe 55 CaFe204 800-1200 0.4 72 m ...
'ij
FeO 699-983 0.118 29.7 ::r
Fea04 799-987 5.2 55.0
Fe20a 900-1300 4 x 105 112.0 r, • B
Q
ZnFe204 800-1300 8.5 x 10 2 82
FeS(ps, = 100 mm Hg) 700 3.2 X 10-8
800 8.1 X 10-8 i
900 1.94 X 10-7
&
Hg20a ex-Ag 2HgI4 127 5 x 10- 8 g
S!
u
~
(II
Pb(ThB) PbCl 2 1.06 x 10 7 38.12 Q
Pbl 2 3.43 X 104 30.0 1:1
-
ex-PbO in air and on Pb 400-600 105 66 v
'ij
without air .,
Q
PbSiOa below the melting point 85 59 m n
w It
Pb 2Si0 4 up to 150 8.2 47 (II
(II
PbS (sulfurization) 580 2.3 x 10-11 It
(II
PbS (in vacuum) 7.9 x 10-12

Pb 210 PbSe 400-800 4.98 x 10- 6 19.1 x


PbSe + 0.5% Bi2Sea 400-800 4.28 x 10-2 37.0
PbSe + 0.5% Ag2S 400-800 4.41 x 10-7 12.7

(continued)
(Table 19 continued)

Se PbSe 400-800 2.1 x 10- 6 27.6

PbS (stoichiom.) x lO- s 35.2 y


Pb 210 400-800 8.6
PbS n·type 400-800 2.6 x lO- s 31.0
PbS p.type 400-800 ~2 x 10-4 ~23
PbS + 0.5% Bi2S3 400-800 4.8 x 10-8 24.6

Na22 NaCI >550 3.13 41.4 z


<550 1.6 x lO- s 17.7 rIl
.... ID
Nao.78W03 664-832 0.87 51.8
IFt:l
S3S fl· A g2S 179 3.6 x 10-13 e
bb
Sulfur >100 2.80 x 1013 46,800
,
III
0-
I'
Zn 65 ZnO(in 02) 800-1400 1.3 73.3 cc 0
ZnO in equilibrium 400 10-16 m 0
ID
with Zn 59
ZnFe204 900-1350 8.8 x 10 2 86 cc n
ZnO dd ID
900-1025 4.8 73 I'
-...
III

Mg MgO in air 1400-1600 0.249 79 ee


0 CdO 640-820 8 x 10 6 93 ±5 ff
Ti0 2 860-1030 1.1 73 II
U02.002 550-800 1.2 x 103 65.3 ±5 hh
U0 2 .063 320-500 2.1 x 10-3 29.7 ± 2.5 hh
Zro.8sCao.1s01.8S 680-900 5.1 x 10-3 29.8 II
CU20 1030-1120 at Po, = 135 mm Hg 6.5 X 10-3 39.3 ± 4.5 q

Cu CU2S 400 1 X lO- s JJ

(continued) ...:a
...:a
(Table 19 continued) ..:J
go
a Johansson, G., and R. Lindner: Acta Chem. Scand. 4,782 (1952); R. Lindner: J. Chem. Phys. 23, 410 (1955).
b Murin, A., and Ju. Tausch: Dokl. Akad. Nauk SSSR 80,579 (1951); see for example J. Teltow: Z. Elektrochem. 56, 767 (1952).
C Pfeiffer, I., K. Hauffe, and W. Jaenicke: Z. Elektrochem. 56, 728 (1952).

d Allen, R. L., and W. J. Moore: J. Phys. Chem. 63, 223 (1959).


e Peschanski, D.: J. chim. phys. 47, 933 (1950).
f Tubandt, C., H. Reinhold, and W. Jost: Z. phys. Chem. 129, 69 (1927).
g Zimen, K. E., G. Johansson, and M. Hillert: J. Chem. Soc. (London) 392 (1949).
~
b Redington, R. W.: Phys. Rev. 82, 574 (1951).
I Bever, R. S.: J. Appl. Phys. 24,1008 (1953). i
j Garcia-Verduch, A., and R. Lindner: Arkiv Kemi 5,313 (1952). t)
/D
k Schamp, H. W., and E. Katz: Phys. Rev. 94, 828 (1954).
I Hedvall, J. A., C. Brisi, and R. Lindner: Arkiv Kemi 5,377 (1952).
t::I
m Lindner, R.: J. Chem. Phys. 23, 410 (1955). ;.
n Shim, M. T., and W. J. Moore: J. Chem. Phys. 26, 802 (1957). rl
o Carter, R. E., and F. D. Richardson: J. Metals 6,1244 (1954): "CI
D Co 9 9
CoO -_ 26
. X 10- pO.35
o. cm 2 /sec (for 1000°C) and D Co CoO -- 90
. X 10- pO.30
o. cm 2/sec (for 1150°C) .
P Moore, W. J., and B. ~elikson: J. Chem. Phys. 19, 1539 (1951); 20, 927 (1952).
q Moore, W. J., Y. Ebisuzaki, and.r. A. Sluss: J. Phys. Chem. 62, 1438 (1958).
f
r Himmel, L., R. F. Mehl, and C. E. Birchenall: J. Metals 5, 827 (1953). ~
I\)
S Lindner, R.: Arkiv Kemi 4, 381 (1952).

t Meussncr, R. A., and C. E. Birchenall: Corrosion 13, 677 (1957).


U Hevesy, G. von, and W. Smith: Z. Physik 56, 790 (1929).
&
v Lindner, R.: Arkiv Kemi 4, 385 (1952).
W Anderson, J. S., and J. R. Richards: J. Chem. Soc. (London) 537 (1946).

x Seltzer, M. S., and J. B. Wagner: J. Chem. Phys. 36, 130 (1962).


~
Y Simkovich, G., and J. B. Wagner: J. Chem. Phys. 38, 1368 (1963). S-=:I
Z Mapother, D. E., H. N. Crooks, and R. J. Maurer: J. Chem. Phys. 18,1231 (1950); see also A. Murin, and B. Lure: Dokl. Akad. Nauk

SSSR 73,933 (1950). "CI


aa Smith, J. F., and G. C. Danielson: J. Chem. Phys. 22, 266 (1954). ~
n
bb Haissinsky, M. M., and D. Peschanski: J. chim. phys. 47, 191 (1950). /D
cc Lindner, R.: Acta Chem. Scand. 6,457 (1952). ~
/D
dd Secco, E. A., and W. J. Moore: J. Chem. Phys. 26, 942 (1957). Ul
ee Lindner, R., and G. D. Parfitt: J. Chem. Phys. 26,182 (1957).
If Haul, R., and D. Just: Z. Elektrochem. 62, 1124 (1958).
gg Haul, R., D. Just, and G. Diimbgen: Proc. Fourth Intern. Symp. on the Reactivity of Solids, Amsterdam, 1960, p. 65.
bh Auskern, A. B., and J. Belle: J. Chem. Phys. 28, 171 (1958); Belle, J., A. B. Auskern, W. A. Bostrom, and F. S. Susko: Proc. Fourth
Intern. Symp. on the Reactivity of Solids, Amsterdam, 1960, p. 83.
ii Kingery, W. D.: J. Am. Ceram. Soc. 42, 293 (1959).
jj Wehefritz, V.: Z. physik. Chem. [NF] 26, 339 (1960).
3. The Mechanism of Oxidation of
Metals-Theory

In Chapter 1 we sketched a general-if not an exhaustive-picture ofthe


characteristics of metal oxidation. We now wish to determine quantitatively
which steps are rate-determining for the oxidation process and relate and
explain the experimental results by a generally valid theory. The existence
of sufficiently reproducible results permits the creation of a model, and a
hypothesis can be propounded, the validity of which can then be proved
through further investigations. In such a situation, it is always reasonable to
first make theoretical assumptions for oxidation processes which can be
described by a simple rate law, where additional phenomena-as for example,
nucleus formation, crystal growth, microscopic crystal disturbances (e.g.,
pores), and phase-boundary reactions-can for the most part be neglected.
Simple relationships of this kind can frequently be obtained by exposing pure
metals to an oxygen atmosphere at high temperatures. Under these experi-
mental conditions one generally observes a parabolic rate law, which indicates
that a diffusion process is rate-determining.
To successfully isolate the mechanism of this rate-determining diffusion
process, one must first explain the transport of reactants through the scaling
layer which is formed on the metal, since the reactants are spatially separated
from one another. Deviations from stoichiometry cause the inorganic com-
pounds constituting the scaling layer to exhibit different kinds of lattice
defects in discernible concentrations. This being the case, it is obvious that
these various lattice defects must be responsible for the difference in trans-
port rates of the several reactants-metal or nonmetal or both-in the scaling
layer, where the reactants, of course, are present not as atoms but as ions
and electrons.
Thus a reasonable approach is brvught into focus which suggests
formulation of valid transport equations for particles capable of migration in
"solid electrolytes," specifically in these inorganic scaling layers, just as has
been done for ions in aqueous electrolytes for a long time. In 1933 Wagner
published a theory of high-temperature oxidation of metals which for the
first time made it possible not only to explain earlier oxidation experiments
79
80 3. The Mechanism of Oxidation of Metals-Theory

and to partially evaluate them quantitatively but-which was still more


important-also to plan new experiments from which we could obtain further
clarification of the reaction mechanism.
Before we concern ourselves with the presentation of a general theory
of diffusion- and transport-controlled metal oxidation, in which, of course,
the Wagner theory is contained as an essential segment, we shall mention a
few experimental results which cannot be explained by the following theory
and are therefore still stumbling blocks in the way of theoreticians. Pfeffer-
korn l has published electron-microscopic investigations of metal oxide
layers, where a "crystal needle forest" was formed on the surfaces at inter-
mediate temperatures. A few pictures of oxidized metals are given in Fig. 23
for different temperatures and times. The appearance of these crystal needles
and lamellae was frequently due to a preferred direction of crystal growth
and surface diffusion. Paidassi 2 reported similar formation of iron oxide
needles during the oxidation of iron. Recently, Fischmeister3 and Rickert 4
published experimental results on the formation of whiskers on Ag 2S during
sulfidation of silver under special conditions. Albert and Jaenicke 5 have
investigated the mechanism of the oxide-needle formation on copper between
300 and 500°C at 1 atm air pressure. The length distribution of the growing
oxide needles can be expressed by the following exponential equation
dN
- = k exp ( - k'\)
NdA
where A is the needle length and k decreases with increasing reaction time.
The influence of the oxide-layer growth on the distribution of needle length
was also studied. 6 It is important to note, concerning this needle growth
phenomenon, that the oxide needles do not form on the bare metal surface
but grow out on an oxide layer from one to several microns thick, and at high
temperatures they virtually disappear. These facts support the assumption
that, in most cases, the oxidation rate law is not influenced by formation of
needles, since the transport of matter through the compact oxide layer prob-
ably remains the slowest process, and is, therefore, the rate-determining
step.
In the following arguments, we will assume an ideal parallel layering of
the growing oxide on the metal surface, and with this in mind we turn now to
the diffusion and transport processes in these scaling and tarnishing layers.
1 Pfefferkorn, G.: Naturwiss. 40, 551 (1953); Z. MetaUk. 46,204 (1955).
2 Paidassi, J.: Trans. AIME 197, 1570 (1953).
3 Fischmeister, H., and J. Drott: Acta Meta. 7,777 (1959).
4 Rickert, H.: Z. physik. Chern. [NFJ 23, 23 (1960).
5 Albert, L., and W. Jaenicke: Z. Naturforsch. 14a, 1040 (1959).
6 Albert, L., and W. Jaenicke: Z. Naturforsch. 15a, 59 (1960).
Fig. 23. Growth of oxide needles
on the oxide layer which is
formed during oxidation, accord-
ing to Pfefferkorn. Enlarged
12,000 times: (a) copper oxide
on phosphorus bronze, oxidized
1 hr at 430°0; (b) tantalum
oxide, oxidized by heating at
560°0 ; (c) zinc oxide from
oxidation for 15.5 hr at 340°0.
a

b
82 3. The Mechanism of Oxidation of Metals-Theory

3.1. Dillusion and Transport Processes in Scaling and


Tarnishing Layers
The particle current density of a lattice component which is capable of
migration (via lattice defects) to any location g in the scaling or tarnishing
layer of an oxidizing metal may be calculated with certain simplifying
assumptions and the help of statistical mechanics (theory of absolute rates).1
We will make use of certain results of this theory later. Here we consider first
the generally valid expression for a particle current density which is produced
between two phase boundaries I and II by an electrochemical potential
gradient. This reads as follows: Particle current density = mobility x particle
density x electrochemical potential gradient. If for a lattice defect particle of
type J, we define the mobility as UJ = DJ/RT, where DJ is the correspond-
ing diffusion coefficient, nJ (g) the particle density at the point g, and grad l)J (g)
the electrochemical potential gradient at the point g, then the general
expression for the particle current density is
DJ
jJ(g) = - -nJ(g).grad T}J(g) (3.1)
RT

If we split the electrochemical potentiall) into the chemical potential /L and


the electrical potential V, and use the above definitions,
(3.2)

we then obtain in place of (3.1) the expression

(3.3)

where N L is the Loschmidt (or Avogadro) number; ZJ is the valence of the


J-type lattice defect (including free electrons and holes); (f (g) = grad V (g).
Equation (3.3) is completely valid for any set of conditions-even for higher
lattice defect concentrations.
Equation (3.3) can be derived in general from the theory of saddle
transitions for lattice defect transport. We consider as a separate process
either the motion of a particle A or B from a certain interstitial position i
to one of a nearest neighbor position k or a transition of a particle A or B
from a lattice site into a neighboring A or B vacancy. Under the assumption
that in its new rest position the particle is always in equilibrium with its
lattice neighbors, the overall system will have the same energy after

1Glasstone, S., K. J. Laidler, and H. Eyring: The Theory of Rate Processes, McGraw-
Hill, New York, 1941; C. Zener: J. Appl. Physics 22,372 (1953).
3.1. Ditlusion and Transport Processes 83

each jump. In the presence of an electric field, the initial and final states will
differ in mean electrostatic potential by the amount ze V (z ~ 0 is the charge
on the particle). In all cases the initial and final position is fixed by a suffi-
ciently deep potential well, but between i and k the particle must necessarily
overcome a potential barrier. Since it will prefer those positions where the
potential is a minimum along the direction perpendicular to the motion,
although it is a maximum along the direction of motion of the particle, we
may speak of a potential saddle which must be crossed. A particle transition
between two valleys i and k is called a saddle jump.
According to Schottky, I the time sequence of any kinetic process is
determined if the particle population of certain valleys i in the crystal is
known for the initial state, and if the elementary transition probabilities
Wik (transition per unit time) into a valley k, adjacent to i, can be given.
Retaining the assumption that defect interactions are to be excluded, we
shall limit ourselves here to i -+ k transitions of elementary particles located
a sufficient distance away from all other defects. In formulating the expres-
sion for Wik we can make use of the treatment by Jost 2 and Schottky.1 It is
found that Wik can be represented as the product of a frequency factor,
kT/h ~ 1014 sec-I, having the dimension of reciprocal time, and an activation
term containing the difference in Gibbs free energy AG between the saddle
and the rest position at i:

kT
W'k = -exp( - AGio) (3.4)
h

where u represents the voltage equivalent of temperature, kT/e, and h is


Planck's constant. Here, as in all subsequent expressions, the free energies
and chemical potentials are correspondingly expressed in eV units, and the
electrostatic term assumes the simple form z V. Lidiard3 has recently pub-
lished a particularly clear derivation of the Wi" expression under a simplifying
classical assumption.
In the following, it is expedient to express AG not as a difference between
the free energies of the total crystal in the transition and ground states, but
rather to subtract from both quantities the Go values for a defect-free crystal.
Then, for a crystal with interstitial defects, G, - Go represents the basic
component YJ' of the electrochemical potential for an excess particle at an
interstitial site (without the concentration term) and Gm - Go represents
the corresponding electrochemical potential YJm for the same excess particle
1 Hauffe, K., and W. Schottky: "Deckschichtbildung auf Metallen," in Halbleiter-
probleme, Vol. V, Braunschweig, 1960, pp. 203jJ.
2 Jost, W., in Halbleiterprobleme, Vol. II, Braunschweig, 1955, p. 145.
3 Lidiard, A. B.: Handbuch der PhyaikXX, pp. 246jJ, Springer Verlag, Berlin, 1957.
84 3. The Mechanism of Oxidation of Metals-Theory

located in the saddle. Instead of (3.4), we then obtain the following relation
for the transition i -7 k:

(3.5)

A corresponding expression is obtained for a transition k -7 i in the opposite


direction.
It is well known that the particle current density jik in the direction
i -7 k, and the corresponding diffusion coefficient, is given by the product of
the homogeneous volume concentration n of the migrating particles and their
mean velocity caused by the i -+ k transitions. This velocity, however, is
equal to the number of Wik transitions per second multiplied by the path rik
traversed during each transition. For the reverse transitions k -+ i one must
substitute Wki. The resultant particle current density jik in the direction
i -7 k is thus
jik = nrik(wik - Wki)
(3.6)
jik = nrikwik(l - exp Ll1)ik/t»
where LlYjik = Yjk - Yji. This difference involves only the electrochemical
potentials of the particles in the ground states i and k; if z ~ 0 represents the
effective excess charge caused by the presence of the particular defect in
question, then by definition
(3.7a)
where fL is the basic component of the chemical potential and V is the
electrostatic potential caused by the presence of the electric field (f which
changes from ito k by an amount -rik(f cos (E, rik). Here E and r are vectors
of (f and r. If LlYj ~ tl, then in equation (3.7a)
1 - exp Ll1)ikjt> = - Ll1)ik/t> (3.7b)
However, in view of (3.7b),
Ll1)ik = zLl Vik = - z(frik cos(E, rik)
and thus (3.6) assumes the form

jik = [nzr;kwik cos (E, rik)/t>](f (3.Sa)


In order to determine the current distribution of the i -7 k transitions in an
arbitrary direction x, we must also multiply by cos (x, rik) and we thus
obtain
jx = (nzjt»2,wikr;k cos (E, rik) cos (x, rik)(f (3.8b)
k
3.1. Diffusion and Transport Processes 8S

Thus, in the general case, if Ll1)ik ~ l), j:.r; is related to (f by a tensorlike factor
which depends both upon the direction of E with respect to the various Wik
and upon the angle between the chosen x direction and rik' Thus, equation
(3.8b) is also applicable for transport processes in an irregular lattice.
In order to relate the diffusion coefficient D of the migrating particles
under consideration to the Wik values it is only necessary to establish a direct
relationship between the diffusion coefficient and the mobility. Neglecting
the relation derived elsewhere,! we obtain as an intermediate result for the
resultant particle current in the i ~ k direction
ilk = ( - nr:kwlk/U) gradlk'1} (3.9)
where the gradient of the electrochemical potential 1) acts on the particles
under consideration as a general driving force. Under our assumption 1)
contains no grad ft component but is given only by grad ~ + z grad V, with
the statistical position component ~ = l) In n/N, where N is the total number
of equivalent lattice sites for the rest position of n jumping particles. If we
now rewrite equation (3.9) as
ilk = ( - rl~Wtkgradn - nzr:kWtk!v) grad V (3.10)

it or in more usual notation


(3.1Oa)
where U is the mobility (cm2jV-sec), then the first term of (3.lOa) may be
regarded as diffusion current and the second as field current. From this
it follows that
U
Dik = -U£k (3.11)
Z

With these definitions, we may thus regard equation (3.11) as being generally
valid within the framework of saddle-transition theory, for both a scalar and
tensor character of U and D. This is in fact the well-known Einstein relation.
In applying this relation to transport processes in oxide crystals, we
denote all types of ion defects by J and the free electrons or holes by a
subscript e and $, respectively, i.e.,DJ, UJ, De, ue,etc. In the DJ notation,
using equations (3.10a) and (3.11) and writing 1)J = ftJ + ZJ V as well as
grad V = - (f, we obtain the following transport equations, with CJ (in
moles/cm 3 ) in place of nJ:
jJ = (- DJ/u)cJ grad '1}J (3.12a)
jJ = ( - DJ/u)cJ (grad ftJ - zJ(f) (3.12b)
1Hauffe, K., and W. Schottky: "Deckschichtbildung auf Metallen," in Halbleiter-
probleme, Vol. V, Braunschweig, 1960, pp. 203JJ.
86 3. The Mechanism of Oxidation of Metals-Theory

and
ie = ( - De/v)ce grad'T]e (3.12c)

ie = ( - De/v)ce (grad /Le + <f), etc. (3.12d)

We shall apply these relations in the following sections.


The transport processes through the chemical as well as through the
electrical potential gradients in the region of intermediate and lower tem-
peratures in thin tarnishing layers lead to rather complicated reaction
mechanisms. These mechanisms may be proven experimentally only with
certain-and frequently justified-simplifications. The description of the
mechanism of high-temperature oxidation is considerably simpler however.
This is because of the fact that at high temperatures and with relatively
thick scaling layers, the effects of the electric field transport regions are
negligible, since their contribution under these experimental conditions is
no longer detectable. The migration processes in thick scaling layers-which
are beyond the field transport region-are diffusion processes involving two
types of lattice defects: interstitial cations or anion vacancies and free elec-
trons in the conduction band or cation vacancies with holes in the valence
band. Both types of lattice defects generally have substantially different I
mobilities or diffusion coefficients. For the description of these types of
processes at high temperatures we use equation (3.3), where we cancel the
second member in the parenthetical expression for the above-mentioned
reasons. A special coupling mechanism must come into play during the
diffusion because of the absence of space charges and external electric fields.
This coupling hinders the rapid movement of the considerably more mobile
electrons or holes in the "diffusion field" produced by the chemical potential
difference and thus accounts for the fact that equal numbers of electronic
and ionic defects pass through per unit cross-sectional area at right angles
to the diffusion direction, with consequent maintenance of electroneutrality.
This coupling mechanism between ionic and electronic defect migration is
designated as ambipolar diffusion. In am bipolar diffusion the rapidly moving
electrons or holes are sharply decelerated in the chemical potential field.
The ion lattice defects are not noticeably accelerated because their mass
is considerably greater than that of the electrons.
Having noted these preliminary considerations, we shall now proceed to
a derivation of the rate law for high-temperature oxidation of metals, which
it should be noted is very close to Wagner's presentation. If the conditions
established by Wagner, which we have discussed above, are maintained,
the derived oxidation formulas are always valid. This last statement
deserves consideration since here and there in the literature one encounters
3.2. The Wa~ner Theory of Oxidation 87

attempts to "squeeze" misunderstood data into the Wagner theory, which


are then followed by the "discovery" that that theory is inapplicable to
"this case" or by other less-than-sensible conclusions.

3.2. The Wagner Theory of Oxidation


If the diffusion coefficient DJ is substituted for the partial electrical
conductivity ~J, then we obtain for the transport rate jl, j2, and ja in moles
cations, anions, and electrons per cross-section and time units:
~l 07]1
jl = ----- (3.13a)
Z~F2 og
~2 07]2
j2 = ----- (3.13b)
z~F2 og
~a 07]a
ja = ----- (3.13c)
z~F2 og
By using "f)J = jLJ + zJFV and Za = -1 we get for the transport rates J l ,
J 2 , and Ja in equivalents per cross-section and time units:

J l = Zdl = (3.14a)

~2 OjL2 ~2 oV
J2 = IZ21j2 = - - - - - +-- (3.14b)
IZ21F2 og F og
~3OjLa ~3 oV
J 3 =ja= ---+-- (3.14c)
F2 og F og
Since often the condition IJll ~ IJ2 1 or IJ21 ~ IJll is fulfilled we obtain
for the first case on basis of electroneutrality
Jl = J3
After dividing equation (3.14a) by ~l and equation (3.14c) by ~3 and ad-
ding, we get

1 1) 1 (OjLl OjL3) (3.15)


J l ( ~l + ~3 = - zl F2 .8g" + Zl8g"
If we consider that the chemical potential jLMe of the metal may be expressed
by
jLMe = jLl + ZljL3 (3.16)
88 3. The Mechanism of Oxidation of Metals-Theory

then from equation (3.15),

1 "1'''3
J 1dg = - -- dJLMe (3.17)
ZlF2 "1 + "3
In order to make this equation suitable for a practical evaluation, we
integrate from 0 to Llg, and simultaneously integrate the variables JLMe
between the limits JL~~ (oxide layer/metal phase boundary) and JL~~~
(oxide layer/gas phase boundary). In view of the frequently present depen-
dence of the partial conductivity on JLMe we place "1 = " . t1-with t1 as
transference number of the cations-and "3 under the integral sign and
obtain the following expression:

(3.18)

The total amount of the equivalents of the oxidation product d(n/q)/dt


which is formed per cross section q per unit time is determined by the sum
of the absolute values of the transport rates of the cations and anions:
d(n/q)
- - =.11 + IJ21 (3.19)
dt
and finally we obtain

(3.20a)

or

(3.20b)

where JLX is the chemical potential of the nonmetal X (e.g., oxygen, sulfur,
halogen). If we split the partial conductivity into the total conductivity "
and the transference numbers h, t2, and t3, then we get from equation (3.20b):

(3.21)

The cross-section, or the surface q, which is being oxidized is constant-


3.2. The Wagner Theory of Oxidation 89

and its placement in front of the integral sign justified-only in the case of
oxidation of sheetlike samples. When wires are involved, the surface which
is exposed to the attacking gas changes with progressive scaling, which
requires that q be placed under the integral sign. The expression in the braces
is constant with time under the previously stated conditions and is called the
ratwnalscaling constant, k, in equivalents/em-sec. This is equivalent to Tam-
mann's parabolic tarnishing constant.
We will postpone a detailed discussion of the scaling expression to a
later chapter, but will mention here a few points concerning the applicability
of this equation. The scaling constant (the expression in parentheses) depends
on the transference numbers of the individual particles capable of migration,
i.e., cations, anions, and electrons, and further, on the total conductivity"
of the scaling layer as well as on the chemical potential gradients of the
nonmetal or metal in the layer. The quantities just mentioned are also
closely related to the lattice defect state and the lattice defect type, as we
discussed earlier in Chapter 2. If one varies either the transference number tJ
or the partial conductivity "J, whichever has the smallest value, one will
also simultaneously change the scaling rate. In the crystals which make up
the scaling layer, the lattice defect is a function of the chemical potential,
/Lx = RT In px, so the oxidation rate will be a function of the partial pres-
sure of the nonmetal X and can indicate the relationship between the lattice
defects and partial pressure of the surrounding atmosphere. A few oxidation
systems are listed in Table 20, where calculation of the scaling constants was
possible from electrical data. As can be seen, the agreement between cal-
culated and experimentally obtained values is quite good.

Table 20. Calculated and Measured Scaling Constants of a Few Scaling Systems,
According to Wagner

Rational scaling constant,


equivalents/cm -sec
Metal Nonmetal partial Reaction T,oC Refer-
pressure p, atm product Observed Calculated ence

Ag Sulfur Ag2S 220 1.6 x 10- 6 2.4 X 10- 6 .


(liquid)
Cu PI. = 0.06 CuI 195 3.4 x 10-10 3.8 X 10-10 b
Ag PBr, = 0.23 AgBr 200 3.8 x 10-1 2.7 X 10-11 c
Cu Po. = 8.3 X 10- 2 Cu2 0 1000 6.2 X 10- 9 6.6 X 10- 9 d
= 1.5 X 10- 2 4.5 X 10- 9 4.8 X 10- 9
= 2.3 X 10- 3 3.1 X 10- 9 3.4 X 10- 9
= 3.0 X 10-4 2.2 X 10-9 2.1 X 10- 9

a Wagner, C.: Z. physik. Chern. (B) 21, 25 (1933).


b Nagel, K., and C. Wagner: Z. physik. Chern. (B) 25, 71 (1934).
C Wagner, C.: Z. physik. Chern. (B) 32, 447 (1936).
d Wagner, C., and K. Grunewald: Z. physik. Chern. (B) 40, 455 (1938).
90 3. The Mechanism of Oxidation of Metals-Theory

The lattice defect concentration and the transference numbers or partial


conductivities can be influenced not only by a change in the nonmetal
partial pressure, but also-and generally to a much higher degree-by the
introduction of ions of other valences into the scaling layer. This introduction
can be effected either through vaporization of the foreign oxide during the
oxidation or by alloying with suitable additives.
Advances in the techniques of manufacturing and handling of radio-
active isotopes make it possible, in many cases, to determine the self-diffusion
coefficients-which can be equal to the component diffusion coejficientsl-of the
ionic species responsible for the scaling rate, and these may be used in place of
the electrical data. A few points stilI remain to be considered for the proper
execution of such measurements but we will postpone discussion of these
until later.
If one sets j1+2 = j = (l/q)(dn/dt) for fJ-J = RT In aJ in the initial
equation (3.3) for the overall current of cations and anions (designated by the
indices I and 2) and also considers that Dj is a function of aJ, as was dis-
cussed earlier, after some intermediate calculations one obtains the expres-
sion also derived by Wagner 2 for the oxidation rate:

(3.22a)

or

dn
- = -
q {
CequU
af~!(Dr + -IZ21)
D; dlnaMe} (3.22b)
dt Jg a(a)
~
>te

Here Dr and D; are the self-diffusion coefficients of the metal ions and anions
and ax and aMe the thermodynamic activity of the nonmetal and the metal,
respectively. Frequently, in the case of smaller lattice defect concentrations
one can assume ideal conditions and use the corresponding partial pressures,
fJ- = RT lnp (3.23)
which are more easily measured, rather than the activities.
How can one apply equation (3.22) correctly? This is perhaps not always
clear in the literature, and since in some cases equation (3.22) was actually
1 Hauffe, K., and W. Schottky: "Deckschichtbildung auf Metallen," Halbeiterprobleme
Vol. V, Braunschweig, 1960, p. 208.
2 Wagner, C.: "Diffusion and High Temperature Oxidation of Metals," in: Atom Move.
ments, ASM Cleveland, 1951, pp. 153ff.
3.3. DUJusion and Scaling Coefficients 91

used with the wrong component diffusion coefficients-even though Wagner's


presentation should have been clear enough-Hauffe and Ilschnerl have
discussed the connection between scaling constants and the tracer and self-
diffusion coefficients in even greater detail. Before we analyze these relation-
ships more closely, the following point should be emphasized. In Wagner's
scaling formula, cations and anions are initially considered equally likely
diffusing species. Whether cation or anion migration will take place pre-
ferentially is determined solely by the lattice defect structure of the scaling
layer. We mention this point again only because errors are still being made.

3.3. Dillusion and Scaling (oeBicients


In the use of diffusion and self-diffusion coefficients for the calculation
of scaling constants we should consider the fact that DJ in (3.3) contains the
difference between the free energies of the equilibrium and saddle-point
positions of the migrating ions in an exponential form. This energy difference
is generally designated as the free activation energy ofthe position interchange
of the lattice defect position. 2 Here the DJ used in the transport equations
is not to be confused with the self-diffusion coefficients of the corresponding
type of ion in the lattice. In dealing with radioisotopes, we must consider two
separate probability factors for migration in the lattice: first, the probability
of motion of a lattice defect corresponding to the DJ, and second, the prob-
ability that the isotope is either adjacent to a vacancy position, into which
it may jump, or that it is in an interstitial lattice position. Unlike lattice
defects, which are "self-drivers," the radioactive lattice ion must, figuratively
speaking, wait for a "taxi." Thus in the exponents of the self-diffusion
coefficients,
D~ = const· exp { - (ED + EA)/RT} (3.24)
we must have the activation energy EA for the formation of a lattice defect
within the crystal as well as the activation energy to surmount the saddle
height ED, so that

Thus, there is a difference in the temperature dependence of the diffusion


coefficients DJ and Dj. In the last analysis, however, the product D~J('),
which appears in the description of solid reactions corresponds in its tem-
perature dependence to the transport coefficient tJ. The scaling layers which
are formed in such oxidation systems are distinguished by a "chemical"
1Hauffe, K., and B. Ilschner: Z. Elektrochem. 58, 478 (1954).
2See, for example, N. F. Mott and R. W. Gurney: Electronic Processe8 in Ionic Orystals,
Oxford, 1948, pp. 33ff.
92 3. The Mechanism of Oxidation of Metals-Theory

lattice defect if the lattice defects in the ionic crystals which make up the
scaling layer are changed by the presence of either the reaction gases or the
metal or alloy. Hauffe and Ilschner were able to show that the energy
required for lattice-defect formation in equilibrium with a neighboring
phase-whether metal or reacting gas at different partial pressures-depends
on whether the neighboring phase has an excess or deficiency of the com-
ponent Me or X in the ionic crystal MexXy (Fig. 24).

(a)
Cu

Fig. 24a. Schematic representation of the con·


centration of copper ion vacancies, CuD', through
the CU20 layer during oxidation (vacancy current
(J f- and metal ion current proceed in opposite directions).

(i) (a)
Zn

Fig. 24b. Schematic representation of the concen·


tration of zinc ions in interstitial lattice positions,
ZnO", through the ZnO layer during the oxidation
(lattice defect current and metal ion current proceed
(J
f-- in the same direction).

If we now consider that


D* ~ exp ( - EA/RT)
then the self-diffusion coefficient must also change, and we will obtain quite
different values for the self. diffusion coefficients, e.g., in CU20 and ZnO,
depending upon whether we measure the self. diffusion of the metal ions in
the oxide in equilibrium with oxygen or with the metal. Wagner called
attention to these facts in the case of oxidation of copper to CU201 and derived
appropriate equations. For the parabolic scaling constant k (equivalent/cm-
sec), we obtain the following expression in the present case from equation
1 '.Vagncr, C.: "Diffusion and High Temperature Oxidation of Metals," in: Atom JYIove·
ments, ASM, Cleveland, 1951, pp. 153ff.
3.3. Diffusion and Scaling Coefficients 93

(3.22a) after integration, when D~ :::::: 0 and if, to a first approximation, the
concentrations of the vacancies in both phase boundaries can be substituted
for the activities:
C(i)}
*(a) ( CuD'
k = (1 + ZCu)cequilDcu 1 - <a>- (3.25)
ccuo'

As one may recognize from (3.25), the determining factor in the calculation
of k is the self-diffusion coefficient D~~) of the copper ions in the Cu 2 0
which must remain in equilibrium with the oxygen pressure prevailing during
the oxidation. For high oxygen pressures where cg~o' ~ c~~o' equation
(3.25) simplifies to
,...., 2Cequil D*(a)
k ,...., cu (3.25a)
or when k can be replaced with k' (cm2/sec) and D~~) == D* is used
k'/D* :::::: 2 (3.25b)
The experimental evidence for the validity of the relationship (3.25b) was
given by Moore and Selikson,l whose results are summarized in Table 25
in Section 4.2.l.l.
The self-diffusion of copper ions determined by Bardeen, Brattain, and
Shockley2 was measured during copper oxidation in a CU20 layer where
the lattice defects were not in equilibrium with the oxygen pressure ruling
during oxidation but were governed by a concentration gradient of copper
ion vacancies. For this reason, the evaluation of D~u and its conversion to
k values is rather complicated.
In the case of zinc oxidation the behavior is just the opposite. Since
ZnO is an n-type conductor, with zinc ions in interstitial lattice positions, the
self-diffusion coefficient D~~) is rate-determining for the zinc ions in ZnO,
which is in equilibrium with the zinc and not with the oxygen. The self-
diffusion coefficient agrees in the region of the highest lattice defect concen-
tration-which here is at the Zn/ZnO phase boundary-in the calculation
(Fig. 24b). The expression derived by Wagner is identical to (3.25), and for
zinc oxidation reads

k = (1 + ZZn) Ce.quil D~~)i { 1 _ c~a~o


J
.. } (3.26)
cz(i ..
no
or .h
WIt ZZn =
2 an d (i)
CZnO " ~
( )
ctnO'.:
,...., 3CequIl
k.,...., . D*(tJ
Zn (3.26a)

1 Moore, W. J., and B. Selikson: J. Chern. Phys. 19, 1539 (1951); 20, 927 (1952).
2 Bardeen, J., W. H. Brattain, and W. Shockley: J. Chern. Phys. 14, 714 (1946).
94 3. The Mechanism of Oxidation of Metals-Theory

If one does not consider the relationships sketched here and instead deter-
mines D;n
in ZnO after annealing in air, then no kind of agreement should
be expected.
In conclusion, we note as a rule for the comparison of D* and k values:
1. For oxidation systems with n-type conducting protective layers-
regardless of whether there are anion vacancies or cations in interstitial
lattice positions-the self-diffusion coefficients of the anions or cations must
always be determined in the oxide crystal which is in equilibrium with the
metal phase.
2. For scaling systems with p-type conducting protective layers, the
self-diffusion coefficient of the cations-and also that of the anions in anion
interstitial lattice positions-must always be determined in the crystal which
forms the scaling layer, and which is in equilibrium with the oxygen partial
pressure prevailing during the oxidation.
Before we leave pure diffusion-controlled oxidation processes, we will
discuss the calculation of the absolute oxidation rates of metals, and for this
section we will draw upon a presentation by Gulbransen.!

3.4. Calculation of the Absolute Oxidation Rate Constants of


Metals for the Parabolic Rate Law
As we have shown in earlier chapters, a parabolic rate law for metal oxidation
is to be expected only if the phase. boundary processes take place sufficiently fast
and if the rate· determining factor is the diffusion of only one type of ion (cations
or anions) or of electrons through lattice defects. Under these conditions we
obtain the parabolic rate law (1.2), which in integrated form reads

(LI ~)2 = 2k' t (3.27)

with

(3.27a)

where Q represents the volume of the oxide per metal ion and n(a) and n(i) denote,
respectively, the number of vacancies or interstitial ions per cubic centimeter at
the oxide/gas or oxide/metal phase boundary.
This equation can be used in the case of nickel oxidation, where p.conducting
NiO is the scaling layer formed. By neglecting n(i) and using only n(a) == nNiD" =
N/(4 1 / 3)plJ:
exp( - LlFo/3RT), which is only the mass action law (2.21) with an
additional term for the standard free energy of lattice defect formation, we
obtain the following expression for the parabolic scaling constant k':

k'-- fJD 1/6


41 / 3 Npo,exp(-LlFo/3RT) (3.28)

1 Gulbransen, E. A., and K. F. Andrew: J. Electrochem. Soc. 101, 128 (1954).


3.4. Absolute Oxidation Rate Constants for the Parabolic Rate Law 95

Here N denotes the total number of lattice positions per cubic centimeter and
AFo = - RT In K, where K is the mass action constant of the lattice defect
equation.
Using the diffusion theory as modified by Zener,l we obtain for the diffusion
coefficients appearing in (3.27a) or (3.28)

D = ya2 vexp-(LlF*/RT) (3.29)

where v denotes the vibration frequency in the direction of diffusion, a is the


average jump distance, and AF* is the isothermal work of diffusion, and y is a
constant which depends on the nature of the diffusing ions. By consideration of
the well-known thermodynamic relationships
and LlF* = LlB* - T LI S*

we finally obtain from (3.28) and (3.29) the expression

k' = y a2 n v N
41 /3
p1/6
0,
exp {(,180 /3 + ,1S*)/R} exp { - (,1Ho/3 + ,1H*)/RT}

(3.30)

Here ASo is the standard entropy for vacancy formation, LJHo the corresponding
heat of formation, and AS* and LJH* the activation entropy and activation heat
of diffusion, respectively_
The calculation of the parabolic rate constants may be demonstrated using
nickel oxidation as an example. The quantities v, Ct, y, and N may be determined
from the structure ofNiO. (v = 0D k/h, where 0D is the Debye temperature, k is
the Boltzmann constant, and h is Planck's constant.) The quantity AHo + LJH*
was obtained from temperature dependence considerations as 41,200 cal/mole
NiO. The quantity ASo may be calculated from thermodynamic data using the
techniques of Jost 2 and Mott. 3 The entropy change upon introduction of oxygen
into the lattice can be estimated under the two following conditions:
I. Entropy change without disturbance of the NiO lattice.
2. Entropy change under consideration of a disturbance in the NiO lattice
caused by a cation vacancy and two holes per oxygen ion introduced.
Gulbransen and Andrew4 reported a similar type of calculation. For NiO at
1000oK, we obtain:
S of !O~) = 29.10 cal/deg
S of 0 2 - in NiO = 9.04 cal/deg
AS (lattice defect) = 2.4 cal/deg

1 Zener, C.: J. Appl. Phys. 22, 372 (1951).


2 Jost, W.: J. Ohern. Phys. 1, 466 (1933); Physik. Z. 36, 757 (1935).
3 Mott, N. F., and M. J. Littleton: Trans. Faraday Soc. 34, 485 (1938).
4 Gulbransen, E. A., and K. F. Andrew: J. Electrochem. Soc. 101,560 (1954).
96 3. The Mechanism of Oxidation of Metals-Theory

Thus it follows that LlSo = 22.5 cal/deg or, since three lattice defects are
formed per oxygen ion introduced, L1So = -7.5 cal/deg per gram atom of lattice
defects.
The activation entropy of diffusion L1S* has a positive sign and, according to
Zener, reads

(3.31 )

where ,\ represents that portion of the free energy of activation caused by the
lattice disturbance, T m is the melting point of NiO, and
d(s/so)
(3 = - d(T/Tm)
where s is the modulus of elasticity and So is the value of s at T = O. {3 can be
calculated from Op using statistical thermodynamics,

o =~{RT2(a1nQ.)}
p aT aT p
with Q.=kT/h
from which, by differentiation, we obtain
o p
=R~{T
aT _ p2 dlnv}
dT (3.32)

From the expression given by Einstein for the relationship between the com-
pressibility X of a solid body and its characteristic frequency VE ~ 1/X1/2 or
VE ~ SI/2, since X ~ l/s it follows from (3.32) that

o = R ~{T ,_ T2 dInS} (3.33)


p aT 2 dT
From this one finally obtains from equation (3.33) for three degrees of freedom

Op = 3R {1 {3
}
+ -T -So (3.34)
Tm s
Since s = So at T = 0 and s = 0 at T m, then s/so = 0.6 at Tm = 10000K.
Thus (3 = 0.42. Under the assumption that L1H* is t(L1H* + L1Ho) for NiO,
Gulbransen calculated L1H* = 20,600 cal/mole and ..18* = 2 cal/deg-mole. Using
these data, we finally obtain the parabolic scaling constant at 700°C, k' =
1.35 X 10-13 cm 2 /sec, in good agreement with the experimentally obtained
value, k' = 1.02 X 10-13 cm 2 /sec.
The conclusions we have reached regarding oxidation properties do not apply to
cobalt at400°C,as Gulbransen has shown mathematically, but this is not surprising
when we consider that here we are dealing with formation of an additional C0 3 04
layer with field transport phenomena. These must be considered and significantly
complicate the mechanism, so that the above-mentioned assumptions are no
longer valid.
After these supplementary considerations to \Vagner's scaling theory, we
will now proceed to a discussion of oxidation where an additional effect due to
electric fields set up in the oxide layer during the oxidation process must be
taken into account.
3.5. Effect of Electric Fields on Metal Oxidation 97

3.5. The Etlect of Electric Fields on Metal Oxidation


3.5.1. Space-Charge Layers at Phase Boundaries
In order to get the clearest possible picture of the rather complicated
mechanism that characterizes the growth of the tarnishing layer at inter-
mediate or lower temperatures, we shall first deal with the chemisorption of
the reacting gas, which introduces the overall process of oxidation. In this
we shall as a first approximation neglect all changes due to the growth
process. Thus we are considering an ionic crystal which is large in comparison
to the region close to the surface in which electric space charges arise as a
result of the different thermodynamic potentials of the electrons and ions,
which occur when the crystal is in equilibrium with its neighboring
phase. This region close to the surface, in which electric space charges are
present, following the terminology of semiconductor physics, is designated
as a space-coorge boundary layer or space-coorge layer.1
The differences in the thermodynamic potential of electrons and ions
in the crystal which is in equilibrium with a neighboring phase, e.g., oxygen,
is a sufficient condition for the concentration of the ion lattice defects at the
phase boundary or in the space-charge layer to be different from that of the
electron lattice defects:
ne (0) l' no (0) and ne (0) l' no (0)
or
and

Here the coordinate g = 0 marks the phase boundary at the surface and the
superscript R indicates the space-charge layer. The feature that distinguishes
the ensuing discussion from the space-charge-Iayer theory of chemisorption
according to Engel and Hauffe 2 is that we will also consider the mobility of
vacancies and interstitial ions, which after all is responsible for the formation
of tarnishing layers, whereas in the Engell-Hauffe theory, the process of
chemisorption is limited to an electron exchange, and the ion lattice defects
are as a first approximation, because of the low temperature, considered
immobile.
The starting point of a discussion of these space-charge-Iayer effects
must be a treatment of the combined effects of electrons- and ions of the
tarnishing layer on the impinging gas and the adjacent metal. While the
action of the gas on the solid body has ·been investigated very thoroughly,

1 Schottky, W.: Naturwiss. 26,843 (1938); Z. Physik 113, 367 (1939); 118, 539 (1942);
N. F. Mott: Proc. Roy. Soc. London (A) 171, 944 (1939); B. Davidov: Z. Physik SSSR 1,
2 (1939).
2 Hauffe, K., and H. J. Engell: Z. Elektrochem. 56, 366 (1952); H. J. Engell and K.
Hauffe: Z. Elektrochem. 57, 762 (1953).
98 3. The Mechanism of Oxidation of Metals-Theory

the space-charge-Iayer effects at the tarnishing layer/metal phase boundary


during oxidation have not yet been successfully studied. Although this
ignorance of space-charge-Iayer effects on the metal side limits the general
validity of oxidation theory, nevertheless the following considerations help
to advance our understanding of the mechanism of the space-charge oxidation
processes.
The inequality of the lattice defect concentrations described earlier can
be interpreted in the following way: Because of the large electron affinity of
the chemisorbed gas, e.g., oxygen, electrons are withdrawn from the oxide
with a simultaneous formation of ions (e.g., 0-). While the concentration of
free electrons in an n-type conducting oxide is reduced in this way relative
to the concentration of cations in interstitial lattice positions 0, so that

(3.35a)

additional holes are created in a p-type conducting oxide by chemisorption,


which results in

ne > no (3.35b)

Since the space charges appear in this way and set up an electric field-
which counteracts the chemisorption process-the region of the oxide crystal
which acts as an electron supplier is relatively narrow (only a few hundred
Angstroms). Corresponding to the electron depletion in the first case, we
speak of depleted space-charge layers and in the second case, because of the
enrichment of holes, of enriched space-charge boundary layers. In any case,
we obtain a positive space charge, which is compensated by the negative
surface charge of the chemisorbed oxygen (see Fig. 25). The process of
chemisorption of oxygen is written symbolically in the following way:1

(for example, in ZnO) (3.36a)

(for example, in NiO) (3.36b)

1 Further information on the mechanism of chemisorption can be found in the article by


K. Hauffe and W. Schottky, "Deckschichtbildung auf Metallen," in Halbleiterprobleme,
Vol. 5, p. 287, Berlin, 1960. According to new investigations by W. Doerffler and K.
Hauffe, J. Catalysis, 3, 156, 171 (1964), we are forced to assume that in addition to its
role in the mechanism (3.36a), the chemisorption of oxygen at ZnO surfaces is involved
significantly by the following reactions:
O(g)
2
+ e(H) --' 0-2 chern
0;;--

and
02' chern + ZnO' + 6(H) ~ ZnO + O;;bem
3.5. Effect of Electric Fields on Metal Oxidation 99

Wagner Zone

Fig. 25. Schematic representation of a space·


charge boundary layer with positive space
charge and local variation of the defect
concentration ne or nO and nO or ne in this
boundary layer and in the quasi.neutral
Wagner zone. l is the maximal boundary
layer width that appeared (z = 1). 0.. ~ l ;-
Space·Charge Layer Widtli

Here the superscript H denotes the neutral bulk of the semiconductor. As


already indicated, besides the depletion of free electrons or enrichment of
holes in the vicinity of the surface, there is a simultaneous enrichment of
interstitial lattice ions or a depletion of vacancies, which is schematically
represented in Fig. 25. The space· charge densities e, which are caused by
the difference in concentration between the ion and electron lattice defects
in the space· charge layers, are given, for example, for a p.type conducting
oxide at the location ~ by
(3.37)
where Zo gives the "valence" of the cation vacancy relative to the lattice.
These space charges, which have their origins at the surface, also extend a
certain distance into the crystal. Their motion against the field is affected by
the scattering tendency of the thermal energy. Just as the molecular distri-
bution in the atmosphere is determined by the barometric altitude formula,
so does the Boltzmann law govern here. For a space-charge layer in a p.type
conducting oxide it is written in the following way:
no(~) = nH exp [ - Zo V(g}/o] (3.38a)
n(fM) = Zo nH exp [ + V(~)/o] (3.38b)
Here V is calculated so that the potential disappears in the bulk of the
crystal [V(~ ~ (0) ~ 0],
ne(oo) = Zo 11. 0 (00) = Zo nH

and, according to equation (3.37), e(oo) = O. Wt: may make a statement con-
cerning nH if we raise (3.38b) to the zth power and multiply by equation
(3.38a). This yields for all ~
(3.39)
100 3. The Mechanism of Oxidation of Metals-Theory

or especially for g= 0
(3.40)
If we introduce the Boltzmann distribution with activation energies Uo
and u(j) for the ion and electron lattice defects at the phase boundary, where
U = Uo + Z u(j) is the activation energy discussed by Hauffe and Ilschner
for the creation of a neutral lattice defect, then by considering the effect of
the space charge, we obtain for the temperature dependence of the lattice
defect state

(3.41)

The term Z + 1 appearing in the denominator of the exponent is a conse-


quence of the assumption that the ion and electron equilibria are attained
separately at the phase boundary. The general formula corresponding to
equation (3.39)
'fl 0 (~) n~ (~) = const (~) for T and px, = const

is a statement of the chemical mass action law.


From (3.38), we obtain the electric potential difference between the
surface with the chemisorb ate and the crystal interior directly:

V(O) - V(oo) == VD =
o (zno(O)}
In - - (3.42)
(z + 1) n(j)(O)

If we now introduce a suitable Boltzmann distribution into (3.42), as was


done earlier in (3.40) for no(O) and n 8 (0), then we see that the potential
difference across the space-charge layer is essentially proportional to the
difference in free energy, f = U - Ts, between ions or electrons in the ionic
crystal and in the neighboring phase, which is generally known as the electron
affinity (Fig. 26). The difference in the free energies or thermodynamic
potentials of electrons, on the one hand, and ions, on the other, of a sur-
rounding phase VS. an ionic crystal is thus the confirmed cause of the space-
charge layer.
Inside the crystal, thermodynamic equilibrium is produced in such a
way that the electrostatic potential V(g) is added to the chemical potential.
We thus obtain
'Y)3 = flG - eV = 'Y)o = fI 0 + Zo e V (3.43)

While the chemical potentials fL for ionic defects and hol('s are different
functions of location, the electrochemical potentials '1) remain constant
(i.e., they are independent of location).
3.5. Effect of Electric Fields on Metal Oxidation 101
Free Energy
cal/Mol

I
I
I
I
I
I
I
I
"!!l ::.1Jjg~II~ ____ _3~+I..;..-a...;I.;;,;on,-=,-,u..;;el,-=,-.-a_(II_~_

Neighboring
Phase Boundary Layer Inner Phase

o
Fig. 26. The energy relationships in the equilibrium space·
charge boundary layer, according to Hauffe and Ilschner
(f-L(H) =0 f-LH - log nH).

A calculation of the potential as well as the spatial distribution of the


lattice defects is then possible by integration of the Poisson equation

(3.44a)

where e(g} = - e[no(g) - n$(g}] and z = 1. From this we obtain for mono-
valent ionic lattice defects the relationship, first given by Cabrera and Mott,l
d2V 87Te 2
- - = --nHsinh(Vjl1} (3.44)
dg2 e

which in appropriate form becomes


2
-d '" = 4
-sinh", (3.45)
dg2 g5
where", == V/o and go = (eo/27TenH}1/2 represents a constant Debye-like
length (e is the dielectric constant of the ionic crystal). From the usual
relation ",' = f(",) one then finds
exp (g/2go) +B
V(g) = 211 In - - - - -
exp (g/2go) - B
1 Cabrera, N., and N. F. Mott: Repts. Progr. in Physic8 12, 163 (1949); Mott, N. F.: J.
chim. phys. 44, 172 (1947).
102 3. The Mechanism of Oxidation of Metals-Theory

with
exp (VD/2o) - 1
B=------
exp(VD/2o) +1
Thus, the space-charge effect disappears when g ~ go. Completely analogous
expressions can be obtained for n-type conducting crystals. The mechanism
of formation of the space-charge layers is now apparent and we must concern
ourselves with the transport processes in boundary-layer fields and space-
charge zones.

3.5.2. Transport Processes in Space-Charge Layers


Before we discuss transport phenomena of ionic and electronic defects,
we will again make use of the Wagner theory to deal with the ambipolar
diffusion of these defects in the lattice. Generally speaking, both types of
defects have essentially different mobilities or diffusion coefficients. In the
case of a joint diffusion of both kinds of defects-for example cation vacancies
and holes-caused by a chemical potential gradient of the metal or nonmetal
in the crystal, in the absence of space charges and electrical currents, a special
coupling mechanism must operate which hinders the movement of the more
mobile type of defects. The coupling is provided through an electric field set
up between the lattice defect pair.
For a quantitative evaluation of the effect of this local field, one superim-
poses an auxiliary electrical potential V' on the chemical potential in such a
way that the auxiliary potential opposes the chemical potential for the more
mobile type of defects but reinforces the chemical potential for the less
mobile type, which of course has the opposite sign on the charge. This
auxiliary field then represents the interaction between the ion and electron
defects. It does not correspond to a space charge in the sense of the Poisson
equation since it just serves to stabilize the quasi-neutrality of the "lattice
plasma." The relationships for a p-conducting oxide with cation vacancies,
such as CU20 or NiO, are presented in Fig. 27.

Fig. 27. Ambipolar diffusion in an oxide


protective layer; presentation with applica-
tion of an auxiliary potential V'(g) without
consideration of a boundary layer. The
variation of the auxiliary potential V' is
MeO M~ plotted with short dashes (z = 1).
3.5. Effect of Electric Fields on Metal Oxidation 103

Since according to the assumption that DJ grad/'l7J should be equal for


both defects 1 and 2, where 'T]J = p.J + zJe V' represents the effective electro-
chemical potential for the defect species J with ZJ as its valence, we obtain V'
according to (3.3) in the space-charge-free and currentless case as!
(3.46)

Since a solution of (3.46) must always have the form V' (g) = - f3 p.(g) we
find the proportionality factor to be
p DI-D2
= ± NLe(Dl + zD2 ) (3.47)

where Z is the valence of the ionic defect. We can eliminate the auxiliary
field from (3.46), and finally, by introducing a common ambipolar diffusion
coefficient Da for both types of defects, we can write Da grad p. instead of
D J grad "1)J. Depending on the sign of the defect type, we then have
Dlgrad1]l = Dl(1 + NLefJ) gradfL:= Da grad p.
D2 grad 1]2 = D 2(1 + N L efJ) grad fL:= Da grad fL
Substituting (3.47), we obtain from either equation the well-known relation-
ship
DID2
Da = (1 + z)--- (3.48)
Dl + ZD2
In crystals which are electronic conductors, Dl ~ D 2, and in ionic con-
ductors, D2 ~ D1 , as has already been mentioned above. From this follows,
for the first case, which is frequently encountered, a relationship that has
been derived elsewhere by Wagner: 2
Da ~ (1 + z)DJ (3.49)

where DJ is always the smaller diffusion coefficient. This expression is valid


for both cation and anion defects. For a NiO crystal we therefore obtain
Da ~ 3DNi; or for Fea04, if we consider only the trivalent iron ions to be
mobile, Da ~ 4DFe .
It is generally true, as was shown in Section 3.5.1, that in thin tarnishing
layers-even at higher temperatures-there is always interaction among
electrical fields in the space-charge layers formed during oxidation. Thus, that
these electric fields do not playa significant role even at higher temperatures
is due to the fact that the thin space-charge layers are "overrun" very
shortly after the onset of oxidation; i.e., thicker tarnishing layers are formed
1Hauffe, K., and B. Ilsclmer: Z. Elektrochem. 58, 478 (1954).
2Wagner, C.: "Diffusion and High Temperature Oxidation of Metals," in: Atom Move-
ments, ASM, Cleveland, 1951, pp. 153.ff.
104 3. The Mechanism of Oxidation of Metals-Theory

where the slower ambipolar diffusion in the "hinterland" of the space-charge


layer rather than the transport process in the space-charge layer is rate-
determining.
However, when only space charge layers are formed, the concentration
gradient of lattice defects compared to the space-charge induced electric fields
cannot be neglected. This consideration is quite well known from the treat-
ment of the processes in the space-charge fields near the electrodes in gas
discharges. 1 These relationships can be made clear in the following way: in
Fig. 28 the strong electric field repels only the negatively charged defects and

,
I
.. I
0- .. + :
n J~,: J--+
0-"0 .. o,;~ :
]
""E c:
0- + + i ~ I~'~
n.(t)=no(~)
+ I I Fig. 28. Local variation of the
·e~ ~~
10
0-
0- ..
. +
(~)
no .. :
I
:
I
defect concentration ne and no
(z = 1) in the space-charge boun-
U Boundary: dary layer and in the quasi-neutral
Layer ,
Positive: Quasi-Neutral Inner Phase Metal Wagner zone of a p-conducting
0/""0- Space: Wagner Zone tarnishing layer during oxidation:
0- Charge: l is the stationary boundary layer
breadth induced during the oxida-
o tion. The arrows indicate the
direction of the 0 and EB current.

attracts the defects with positive charges. Now, why don't these defects follow
this action of force since they do have a sufficient mobility? The reason for
this phenomenon is understandable only if we consider that the oppositely
directed concentration gradient exactly compensates this force effect. Thus,
its neglect would lead to a physically unreasonable decomposition.
The treatment of transport processes in the space-charge layers
formed during oxidation is pointed out in a consideration of the complete
equation (3.3). For a p-conducting tarnishing layer with holes EB and cation
vacancies 0 with the valence zo, we set up the following equations:

{
- Do grad no
(i;W}
+ zono -u- (3.50a)

j® = _ D®{grad n® _ n® (i;~g)} (3.50b)

1\Veizel, W., and R. Rompe: Theorie der elektrischen Lichtbogen und Funken, Leipzig,
1949, pp. 97JJ.
3.5. Effect of Electric Fields on Metal Oxidation 105

with the additional condition j!J) = zojo = const. This condition states that
per unit time, equivalent quantities of negative and positive defects shall
pass through a square centimeter of an assumed boundary plane in the
tarnishing layer and that the material flow shall be "divergence-free" in
the sense that no matter is accumulated in or withdrawn from the channel of
flow. The last requirement is often fulfilled only to a first approximation. In
solving equation (3.50), in order to eliminate the field strength <i(t), which
is generally unknown, we use the first integral of the Poisson equation, which
relates the field strength to the space-charge density:

(3.51)

Inserting (3.51) into (3.50), we obtain after some simplification

d ne
-d~
.-
+ ----.-0
2 n(!)
Xo n(!)
J(
~

n(!) - Zo no) d ,,+


I: J (!) = 0
o
(3.52)

Here nS and n~ are the defect concentrations at the surface where t = o.


Further,

and
(3.54)

Once again we transform these integro-differential equations, which are


very difficult to evaluate, by using a relationship based on the Boltzmann
expression for j(!) = zojo = 0, viz.,
n(!) (;) = n~ exp {- "'I'(;)} .
no (;) = n~ exp {_ x(;)} wIth "'1'(0) = X(O) = 0 (3.55)

Inserting these expressions into (3.52), we obtain after repeated differen.


tiation:

(3.56)
106 3. The Mechanism of Oxidation of Metals-Theory

If one considers the defect concentrations, n(j) and 1/0, which are expressed by
ifi and X in the parenthetical expressions in (3.55), then one has in (3.56) a
system of two simultaneous differential equations for the desired quantities
n(j)(~") and no(g). But even these equations are still too complicated to
be useful for the treatment of concrete examples and it frequently becomes
necessary to seek approximate solutions based on justifiable simplifying as-
sumptions. Still, the system of equations (3.56) is genemlly valid and in-
cludes all the special cases of oxidation that remain to be discussed.
In addition, it may be noted that in the cases J (j) = 0 or J o = 0 equation
(3.56) assumes the form of the usual Poisson equation. This is completely
understandable if one considers that in the currentless, or equilibrium, state,
the variation in defect concentration is described by the usual Poisson
equation. It may be shown, in fact, that in this case equation (3.56) leads to
equation (3.44a), which was discussed earlier. These relationships permit us
to approach the system of equations (3.56) as simply an extension of the
Poisson equation in which the transport current flows through the space-
charge layer.
From this presentation, which deals with the interaction of the electric
fields and concentration gradients (diffusion forces) for particle transport, it
is clear that one must always consider both influences, a fact that must be
observed for oxidation processes with thin tarnishing layers, i.e., those for
which 0 < g < l (critical boundary-layer thickness).
In growing oxide films, space-charge layers are created both by the
different affinities of the individual phases and the different mobilities of the
electrically charged defects migrating through the oxide layer. Ilschnerl has
been able to show that space-charge layers caused solely by defect migra-
tion (ambipolar diffusion) can be of approximately the same order of
magnitude as those caused by the electron affinity of oxygen.

In order to simplify the presentation of the quantitative relationships, we


use the following assumptions, which are valid for many cases:

n_ =n+ =n ~nln~ 1 (I)

(n_ 0= n9 or no' and n+ = n(j) or no) and obtain with (3.47), where z 1, for
the concentration in the space-charge layer ~n from the Poisson equation:

(II)

Now, in order to answer the question of when the space charge-the expression
~n/n-can be neglected, more detailed information concerning the variation of
fL(O must be available.

1 Ilschner, B.: Z. Elektrochem. 59, 542 (1955).


3.5. Effect of Electric Fields on Metal Oxidation 107

Since the lattice defect concentration is generally very small, we can express
,.,.Was
I'm = R TIn (n(,)/N) + const
where N is the total number of all lattice or interstitial lattice positions in a
unit of volume of the material. Accepting the validity of equations (I), we can
set n(,) =n + ex, o
ex = (n('1) - )/'1
nO
(III)

where ~1 is the thickness of the oxide layer. In p-conducting oxides, n(6) ~ no,
from which it follows that
n(,) = ex,
(III')
Finally, it is generally true that D_ ~ D+ or D+ ~ D_. For the former condition
we have
1
p=-- (3.4 7')
Nr,e

and for the latter


1
p =Nr,e
-- (3.47")

Combining (3.47') and (III') with (II), we arrive at the following approximation
for the space charge:

(Positive space charge) (IV)

The interesting quantity here is '8n/n which is given by

I I
Sn(~) eo
- n - = 41Ten(~)
I
Ii (V)

The condition '8n ~ n yields the following inequality for the quantity ~ with the
Debye length, discussed above, if we include the factor t under the radical:

(VI)

In order to obtain the characteristic length d, which is necessary for the evalua-
tion, we must replace the function n(~) in (VI) by n(6) == n max with the help
of (III'). We then obtain
(VII)
Thus d depends on the layer thickness of the oxide film 6 and the maximal
defect concentration nmax-therefore for p-conducting oxides it depends on the
defect concentration at the oxide/oxygen phase boundary. In a region 6d < ~ < ~h
the space charge On produced by the mechanism of ambipolar diffusion is less
108 3. The Mechanism of Oxidation of Metais-Theory

than 1% of the defect density n(~). The boundary-layer thickness defined in this
way is presented in Fig. 29.for different values of n max in percentages of the oxide-
layer thickness.

t
.c

1
~ jfQI--+f---H--++--+--+--t-;

!
~ Fig. 29. Percent of the boundary layer as a
-8 function of the thickness of the entire oxide
§ to ----- ---- layer, calculated by Ilschner. T = 1000oK,
~ mL?~.~~~~~~~~~~-m~-·~~m £ = 10£0' Curve 1: n max = 1020 cm- 3 ; curve
2: n max = 1018 cm- 3 ; curve 3: n max =
1015 cm-3 •

It is apparent from this that the spread of the space-charge zone at the
edge of the oxide layer is largest just where the initial defect concentration
is least. Thus, in the case of an n-conducting oxide layer, e.g., the Zn/ZnO/02,
system the influence of the space-charge zone caused by ambipolar diffusion is
virtually limited to the vicinity of the ZnO/02 phase boundary. With p-con-
ducting oxides reverse relationships exist. Here the space-charge zone at the
phase boundary, e.g., CU20/CU, should be significant as long as there is no
lattice defect inversion. 1 Very little is known about that at present. However,
even in the absence of a defect inversion at the inner phase boundary the
space-charge zone at the outer phase boundary, CU20/02, may also be sig-
nificant, since the electron affinity of the oxygen causes a static space-charge
layer to appear, with a thickness characterized by the inequality
g }> [€u/(87TenH)]1I2

where nH is the equilibrium defect concentration in the interior of the oxide


layer. Accordingly, if we use this equation as an approximation for growing
oxide layers, then we have to use the quantity no for the quantity nH (at
the boundary with the smaller defect density). Since n(g) in (VI) is larger
than no (for p-conducting oxides), for normal values of the electron affinity
in the neighboring phase, the influence of n(g) on the electronic and ionic
defect distribution in the layer reaches somewhat further into the interior
than the effects of the ambipolar diffusion mechanism.
The above discussion leads to the conclusion that an oxidation system
with an n-conducting oxide layer, for all practical purposes, we deal with a
1 A lattice defect inversion is a transfer from a p-type to an n-type disorder in the
crystal and vice versa.
3.5. Effect of Electric Fields on Metal Oxidation 109

single space· charge zone-i.e., in the vicinity of the surface, for example
ZnO/02-while in the case of an oxidation system with a p-conducting oxide
layer, as for example NiOjNi, a quantitative description always involves two
space.charge layers, the outer one, which is very small, being caused by the
electron affinity of the attacking oxygen and the inner one being due almost
entirely to ambipolar diffusion, except that in the latter case electron affinity
of the NiO and Ni phases may playa significant part.

3.5.3. Formation of Thin Tarnishin~ Layers-


The Parabolic and Cubic Rate Law Caused by
Participation of Field Transports
The space· charge layers discussed in detail earlier generally cause field
transport of ionic defects in the formation of thin tarnishing layers. The
electric field set up during oxidation has its origin in the negative surface
charge, which owes its existence to the chemisorption of oxygen or of another
oxidizing gas, and the corresponding positive space charge present in the
tarnishing layer to a depth of several thousand Angstroms. Cabrera and
Mottl have introduced several notable simplifications in their well-known
theory of metal oxidation for thin tarnjshing layers. However, these simplifi-
cations are not always valid, so that we shall defer a discussion of practical
applications and confine ourselves to general observations.
It was first assumed that in layers not thicker than a few hundred
Angstroms the space-charge effects discussed above could be neglected. This
assumption is justified, to a first approximation, in the case of oxidation
which produces n.conducting tarnishing layers with small concentrations of
free electrons and cations in interstitial lattice positions or anion vacancies-
as is the case, for example, with aluminum oxide as a tarnishing layer in
aluminum oxidation. In n-conducting oxides, all free electrons in the space-
charge layer are drawn off by the chemisorption of oxygen, and a depleted
space-charge layer is formed with a very small concentration offree electrons,
which yields a very small space· charge density with a small ion defect con-
centration. When n-conducting tarnishing layers occur with high defect
concentrations as well, e.g., in Ag 2S or NiS, then Mott's assumption is of
dubious validity, and a detailed inspection must be undertaken.
On the other hand, in metal deficit or p-type conductors, e.g., CU20 and
NiO, which,becauseofthe chemisorption of oxygen, form enriched space-charge
layers with large concentrations of holes, the space· charge effect cannot be
neglected. Only if the layers are extraordinarily thin-about 10 A-can
the space charge in the tarnishing layer be neglected, even to a first

1 Cabrera, N., and N. F. Mott.: Rept. Progr. in Physics 12, 163 (1949).
110 3. The Mechanism of Oxidation of Metals-Theory

approximation, as was shown elsewhere.! The cubic rate law of oxidation,


which leads to tarnishing layers with space-charge-Iayer thicknesses up to
and greater than lOOO A, is based on this space-charge effect.
Secondly, Mott's theory contains the assumption, even though it is not
stated explicitly, that the defect concentration within the thin oxide layer is
constant (Fig. 30a). This assumption gives rise to doubts concerning the
D
-e · ~ -- +
0/9"') - r-- --:----.----¢+ Met al
=I ~ no'~ +
+
-? .. +
_ ? Very Small +-
Space Charge ..
- ? +
- 2 MeO +
r;:=~ / +
o~ _
\
Strong Surf:Ce
Cooting
e,
d: Mott's Case

Ot (gos)
I
.. Metal
-e·(f - -
.. .
:
Fig. 30. Local variation of the concentration
of the ions in interstitial lattice positions
.. +- +-.. n O(ne ;:; 0) in an n-type oxide layer with a
noftJ Lo+,.ge
Charg~ + ..
.. + homogeneous electrochemical field (field

I
-
.. .. ."Space . . strength (l; as a vector): (a) without con-

.. . ."MeO+ .. +-.. . ... .


+
sideration of a concentration gradient of the
I:! m etal ions in interstitial lattice positions
and the space charge, according t o Cabrera
- + + +
and Mott; (b) with consideration of this
~
!> -
Weak
Surface Coating
t concentration gradient and the space charge,
according to Hauffe and Ilschner.
b: Houffe-llschner Case
validity of the theory, since it was assumed in the model that the reactions at
both phase boundaries proceed rapidly and that it is the transport through
the layer that determines the rate of the total process. This has the conse-
quence that at the two phase boundaries unchanging lattice defect concentra-
tions are produced which can have values of different orders of magnitude.
This difference effects a strong concentration gradient between the phase
boundaries in the tarnishing layer, which the Mott theory does not consider.
What we consider to be the generally valid variation of the defect concentra-
tion as it occurs during oxidation is given in Fig. 30b.
If one agrees that the conditions neglected by Mott are indeed negligible,
then one can readily accept the explanations in the work of Cabrera and
Mott. As shown for the t arnishing system pictured in Fig. 31 , electrons leave
the metal as a result of thermal emission and enter the oxide layer, migrating

1 Hauffe, K.: Reaktionen in und an festen Stoffen, Springer, Beriin/G6ttingen/Heidelberg,


1955, pp. 55 Iff.
3.5. Effect of Electric Fields on Metal Oxidation 111

toward the outer boundary layer to the energetically more favorable chemi-
sorbed oxygen until the homogeneous field thus set up produces a stationary
equilibrium (Fig. 3Ib). Since supposedly no significant space charges are

Fig. 31. Schematic representa-


tion of the positions of electron
levels in the metal, oxide, and
adsorbed or chemisorbed oxygen,
according to Cabrera and Mott:
(a) before electron transfer;
(b) at steady state.

present, an equal number of charges should be found on both sides of the


oxide layer-negative exterior and positive interior (simplified model of a
condenser, see Fig. 32). The potential difference V prevailing between the
phase boundaries is just equal to the difference between the two energy
terms for the metal and the chemisorbed oxygen, and these in turn are
determined by the quantity of charge.
OICide + Metal
~----1+
+
+
+
Fig. 32. Approximately space-charge-free oxide
+
layer with only surface charges (analogous to a +
condenser) formed during oxidation. (f = V/6, +
where V represents the potential difference
between the charges at the interfaces, which is
High Interface Charge +
determined by t.he energy-level difference o
0-( chem) - Zn. t-

Figure 3la represents the electron levels in the metal, in the oxide, and
in the adsorbed oxygen before the electron transfer. According to the calcula-
tion by Bardeen,l the energy level of adsorbed oxygen lies about I eV below
the Fermi level in the metal. On the other hand, Fig. 3lb qualitatively shows
the variation of the energy terms in the stationary state after setting up an
electron equilibrium without including transport processes for ions. Since
V generally attains a value between I and 2 eV, the temperature influence on
the potential may be neglected at lower temperatures.
Under the assumption of a potential difference V which is independent
of the layer thickness 6, there is, according to the first assumption, no signifi-
cant space charge in the tarnishing layer, so that in the total layer a constant
field (f = V /g (Fig. 32), independent of the coordinate g, influences any defect.
Further, if the second assumption proves correct, then, e.g., for A1 20 a in the
1 Bardeen, J.: Phya. Rev. 71, 374 (1947).
112 3. The Mechanism of Oxidation of Metals-Theory

case of aluminum oxidation, a cation current j flows through the oxide layer
between interstitial lattice positions:
. I V
Jo = - uono-
6
where U o is the mobility of the cations in interstitial lattice positions, and
n~ = noW for the corresponding concentration at the metal/metal-oxide
phase boundary. This ion current causes further growth of the layer, so that
k'
(3.57)
dt
We also obtain a parabolic rate law in this case; for high-temperature
oxidation, however, the parabolic law is based on a different mechanism with
other constants. Hauffe and Ilschner were able to show that one can obtain
the same results as Mott but without the second assumption [n~ = no (g)].
For the case of n-conducting protective layers with relatively small
lattice defect concentrations we will-following Mott-also neglect the space
charges in the tarnishing layer with respect to the surface charges at the
Me/MeO and MeO/02 phase boundaries. This is evidently valid in a first ap-
proximation for zinc oxidation, as can be concluded from the investiga-
tions of Moore and Lee. l
According to the discussions of the preceding section, not only the
field strength <r but also the concentration gradient [no (g) - no (0)]/6 acts
as a driving force, where no refers to the zinc ions in interstitial lattice
positions in ZnO. Accordingly we obtain an identical relationship which is
adapted for the present case from (3.50):

jo = - Do{gradno + zno !::. 2.-} (3.58)


tl gl

In the stationary state jo must be constant for all g,


that is, div jo = O.
From this, the following differential equation is obtained:
d2no 1 V dno
--+--z-=O
dg 2 gl tl dg
with the solution
(3.59)

n6 and ~ are constants of integration which are determined by the following


boundary conditions:
no (0) = ng and no (6) = n~
1 Moore, W. J., and J. K. Lee: Tran8. Faraday Soc. 47, 501 (1951).
3.5. Effect of Electric Fields on Metal Oxidation 113

We then obtain

« = --------------
1 - exp ( - Vln)
I II
no - no
n6 = n0II + -----------"----
1 - exp ( - Vln)
Within the framework of the approximations which were used, one can
neglect ng compared to n~ as well as V = kTle compared to V in the
argument of the exponential function. Accordiugly, we have
n*
o '"
'" n 0I
independent of 6.
Insertion of (3.59) with the physically reasonable simplifications in (3.58)
yields the following simple expression for the ion current:
I V 1
- Dozno -- --
jo ~
n el
since the exponential term for the diffusion and field component is eliminated,
but the constant term remains. Now that the growth of the oxide film, d6/dt,
is proportional to the transport current jo' a parabolic tarnishing law for
thin layers follows with the constant k' = Q Dozn~ Vlv, where Q is the
oxide volume per metal ion,
(3.60)
The fact that the tarnishing constant k' appearing here for thin ZnO layers is
basically different from that in the Wagner description of the appearance of
thick oxide layers is also apparent from the oxygen pressure dependence of
the zinc oxidation. While Wagner and Griinewaldl found no dependence on
oxygen pressure, as is to be expected at high temperatures and thick oxide
layers from Wagner's oxidation theory (since n~ is independent of oxygen
pressure) an oxygen-pressure dependence does exist in the case of formation
of thin oxide layers, as Engell and Hauffe 2 were able to show using thermo-
dynamic reasoning, since the V contained in the constant k' is a function of
the oxygen pressure.
In the chemisorption of oxygen, electrons are removed from the zinc
through the oxide layer according to
1
2" o~g) + e (Me)..- O-(u) (3.61a)

where e(Me) is a metal electron. In order to eliminate the term for the
1 Wagner, C., and K. Griinewald: Z. phyBik. Ohern. (B) 40, 458 (1938).
2 Engell, H. J., and K. Hauffe: Metall 6, 285, (1952).
114 3. The Mechanism of Oxidation of Metals-Theory

concentration of the conduction electrons in the metal from the final expres-
sion, we describe the chemisorption process in the following symbolic equa-
tion:

(3.61b)

where Ef> (Me) is a positive charge in the form of a hole ( = absent electron)
in the metal surface.
From electrochemical thermodynamics, the equilibrium equation
(3.61 b) leads to

- -12 f1 o. + rO- + re +
lI.(a) II.(Me) e V= 0

or
n~le)n(u)
---exp
(
+-
V) = K (3.62)
pl/2 l)
o.
From our first assumption, n<:e) = n(u), it follows that
(n(u»4
--exp
(
+ 2V)
- = K2
Po. l)

On the other hand, the potential in the "condenser" formed in the system
is related to n(u):

V = -
27Te 6
n(u) (3.63)

Therefore
2V)
V4 exp ( + -;- =
(27Te)4
6 -e- K2PO• (3.64)

Since in the temperature region in question V is always much greater than 1)


(V ~ 1 volt), the function V, which is a linear factor in front of the exponen-
tial expression, can be approximately replaced by a constant V, which cor-
responds to an average value. Then directly from (3.64):

V = 2" In {(27Te)4
l)
Elf 6 K2p02
}
or with (3.60)

or
(3.65)
3.5. Effect of Electric Fields on Metal Oxidation 115

This is in accord with experimental results of Moore and Lee. In addition to


the parabolic rate law (Fig. 33a), an oxygen pressure dependence was
found in the oxidation rate which may be described satisfactorily by the
logarithmic relationship of equation (3.65) (Fig. 33b). This oxygen pressure

~
I l' 16'
Sjo/)()mm
0100"
flg
11

... P.oso·»
.
~
..:,
.20" u
>.
,-0 " g. ~
~-;:: 8 '" 5" •
~:'2
t: ><

Fig. 33a. The growth of the


'><0
.!::!.. * 1-----o~7""~.<!>5.....= - - - _ I _ - - - _ l
thickness of the ZnO layer (para- ~'5
bolic rate law) at 400°C for
different oxygen pressures, ac-
cording to Moore and Lee. Time-

cmo/sec
1.8~---4----+---~~~~~---1

1iI
~ ~r-----or~~~~----~~----r----1

Fig. 33b. The logarithmic


variations of the oxygen- ""'
pressure dependence of the
parabolic rate constants
evaluated by Engell and
Hauffe from oxidation experi- 1.5 2,0
ments by Moore and Lee. logpo. [mmHg]-

dependence, which the Wagner theory does not explain, is an impressive


indication of the effects of boundary-layer processes on the kinetics of
oxidation. Moreover, from the second term in (3.65) it follows that the rate
law for zinc oxidation with thin ZnO layers cannot be strictly parabolic; this
is more obvious in other tarnishing processes similar to zinc oxidation.
Cabrera and Mott used the same physically-based assumptions that led
to the derivation of the parabolic rate law for thin layers for the development
of the rate law for p-conducting oxide layers which appear, e.g., in copper or
nickel oxidation. A relationship completely identical with equation (3.57) was
formulated, except that the symbol 0 for interstitial-lattice-position ions
was replaced by the symbol 0 for vacancies. If the assumptions are valid-
no space charges in the ox:ide layer and a practically constant vacancy
concentration in the oxide layer-then the vacancy density no near the oxide
surface is proportional to the surface density of the chemisorbed oxygen ions.
116 3. The Mechanism of Oxidation of Metals-Theory

This in turn is inversely proportional to the layer thickness according to el


(3.63)-at least as long as the space charges are negligible-thus
1
no = const 1;'
and by substitution in (3.57) we obtain

del = const~ (3.66)


dt ~i

The cubic rate law follows from integration :


;l(t) = (A t + B)l/3 (3.67)
where A and B are constants which may be calculated from the Cabrera-
Mott theory. At this point we depart from their interpretation, since the
physical assumptions introduced for oxidation systems with p-conducting
protective layers cannot be retained.
A later investigation1 dealt with whether a similar cubic tarnishing law
would be found under the assumption of space-charge layers and the applica-
tion of transport phenomena in these layers in p-conducting oxides. With
these considerations, it turns out that the whole oxide layer behaves as a
space-charge layer. This corresponds to consequences of the assumption that
the hole concentration in the layer is generally large compared to the cation
vacancy concentration, ne ~ no; no as well as ne should decrease con-
e e
tinuously from = 0 to = 6, since chemical equilibrium was assumed at
both phase boundaries.

+
n.
+ +

+
teO>. + + +
+ oj.
+
n + .. oj.
+
12 0

oz (9" ) MeO Fig. 34. L ocal variation of the concentration of


electron holes and vacancies in the space.charge
boundary layer of a thin p.conducting oxide film
o
e- with the layer thickness 6, according to Engell,
Hauffe, and Ilschner.

A divergence-free metal-ion current through a layer of that type is then


possible, as can be shown theoretically, if the defects are present in the
distribution shown schematically in Fig. 34. Finally, in any instant of layer
growth, that is, for all 6, the hole concentration at the phase boundary
1 Engell, H. J., K. Hauffe, and B. I1schner: Z . Elektroohem. 58, 478 (1954).
3.5. Effect of Electric Fields on Metal Oxidation 117

MeOjMe is kept constant by the comparably "inexhaustible" electron supply


of the metal.
As the calculation shows, electronic equilibrium with a constant con-
centration n* then occurs when, by analogy to (3.53),
(3.68)

where x~ is again a constant-the Debye length. A further consequence of this


constant concentration is that the hole concentration ne(O) at the MeOj02
surface becomes a function of n* and the layer thickness:
U
ne (0) = - (xo - 6)-2 (3.69)
47Te
Therefore ne(O) becomes greater with increasing 6, as long as 6 remains
smaller than the stationary thickness of the space charge layer. This is, in
accordance with (3.53), of the same order of magnitude as [euj(27Ten~)1/2 =
Xo < x~], since the equilibrium concentration at the surface of an extended
crystal, ne, is larger than n,*.
Using Mott's assumptions, we set the concentration of nickel vacancies
at the surface proportional to n(u), which is related to the electric field
<f(0), and obtain
(3.70)

One also obtains this functional dependence from the theory of the space-
charge-free boundary layer according to Mott. With the assumption that only
the holes contribute significantly to the electric field, one obtains from the
condition that the vacancy current is free of divergence,
Do<f(g)}
o= div { - Do grad no + zno --u-
and the relationship

Q;(;) =- (4n-e/e) Jn(f]W d;


o

the following equation for the vacancy current:


jo =- Do {grad no - z n O/(;1 + xo)} = ± Do {:J (3.71)

Since the vacancy current flows toward the metal, a positive sign is to be
chosen for {3. One then obtains for the vacancy density no(O) at the MeOj02
phase boundary:
no (0) = {:J Xo - n6
= (:J(x~ -$1) - n6 (3.72)
118 3. The Mechanism of Oxidation of Metals-Theory

This expression fulfills the condition of equation (3.70) in a particularly simple


form when

{3 = K' (x~ - ';1)-2


n6 = K"(x~ - ~1)-1

For the vacancy current this means


. DoK'
10 = (x~ - ~1)2 (3.73)

Here it should be noted that Do is not the self-diffusion coefficient of the


metal ions in the oxide lattice, but the diffusion coefficient of the vacancies.
Since the layer growth d6/dt is proportional to this current, it follows after
suitably combining all the constants that
d6 const
dt (x~ - 6)2
and by integration of the rate law,
~(t) = x~ + const'(t + to)1/3 (3.74)
This leads to a cubic tarnishing law, but we know that the so-called cubic
law represents only a possible limiting case which is completely valid only
when the simplifications made here are actually fulfilled in the oxidation
system. Thus, if one recognizes the extent of the conditions we have con-
sidered negligible, one can understand the deviations from one-third in the
exponent of (3.74) which are frequently observed in experiments, and can
then accept the transition to the parabolic rate law, which can follow when
layer thicknesses are greater than 1000 A.
The cubic rate law mentioned above results from the space-charge-
induced variation of field strength in the boundary layer. A typical example
for an oxide layer formation with a cubic rate law where the space charge
does not seem to playa significant role is the 100°C oxidation of copper.1
The oxygen pressure dependence of the oxidation rate is given by
~ pl/4
k c~ 0

The assumption of an existing regulating equilibrium of oxygen chemisorp-
tion at 100°C is convincing. In spite of the fact that the oxygen pressure lies
in the region of the CuO phase, no CuO formation is observed, probably
because of the considerable retardation of CuO nucleus formation. The rate-
determining process of the CU20 formation is known to be field transport via
copper ion vacancies.
1 Grimley, T. B., and B. M. W. Trapnell: Proc. Roy. Soc. (A) 234 (1956).
3.5. Effect of Electric Fields on Metal Oxidation 119

On basis of the relation for copper oxidation derived by Grimley,l


where t/J = V /'0,

g(dg) = Dc
t/J cos t/J
Qn(a) , - - -
u CuD exp t/J _ 1

where (a) represents the equilibrium at the outer surface (CU20/02),


i O(~)+ 2e(Cu) +t CU20 + 2CuO'
n~~D' = p~~ exp ( - .:1p-/'o) exp (t/J)
we get under special conditions in a first approximation a cubic rate law of
the following form:

dg = ~{DcuQP~4 ellt/J:) exp ( - .:11l-/11 )}


dt g2 • 47Teno~K
In Figs. 35 and 36, the cubic course of copper oxidation at intermediate
temperatures is presented as it was observed by Campbell and Thomas 2 as

pg/em

I f

Fig. 35. Oxidation of copper at EOI\)..


intermediate temperatures at 150 mm ~
Hg oxygen pressure (or 150 mm Hg
oxygen + 15 mm Hg water) in a log-
log plot, according to Campbell and
Thomas. The oxidation curves with
water·free oxygen largely obey the
cubic oxidation law with an exponent
which lies between 0.32 and 0.38.
(The filled·in circles indicate the values
with water vapor, which always tlfJJ,L-----"±'tJ-----~~'l1O:::-------:;tOa
~'O
reduced the rate.) Minutes -

well as by Tylecote. 3 A cubic rate law was also found by Waber 4 for the
oxidation of titanium at 216°C (Fig. 37) and for tantalum at 350°C (Fig. 38).
1 Grimley, T. B.: Chemistry of Solid State, London, 1955, Chapter 14.
2 Campbell, W. E., and U. B. Thomas: Trans. Electrochem. Soc. 91, 345 (1947).
3 Tylecote, R. F.: J. Inst. Metals 81, 681 (1952/53).
4 Waber, J. T.: J. Ohern. Phys. 20,734 (1952).
120 3. The Mechanism of Oxidation of Metals-Theory

rngfcmZ
w~---------+----------~~~~--~

Fig. 36. Oxidation of heated


copper sheets in air at 1 atm
f,1------"'l~=-----I-~=------< containing water vapor (PH,O ~
15 mm Hg), according to Tyle-
~~ cote. The initial cubic course of
the 520°C curve changes over to
a parabolic course after about
5 hr. The transition takes place
after about 1.5 hr in the 725°C
Hours_ curve.

Belle and Mallett! reported a cubic rate law for the oxidation of zirconium
in 0.1 or 1 atm of oxygen between 300 and 425°0 as well as between 575 and
950°0 (see Fig. 102 in Section 4.4.2.3). Hauffe and co-workers 2 determined

Fig. 37. Oxidation of titanium


in oxygen at 216°C in a log-log
JO plot, according to Waber. The
g.cm-2
80 exponent of the cubic tarnishing
l/V law, t = const t 1 / 3 , amounts to

---
0.39 here. In a later work,
J. T. Waber, G. E. Sturdy, and
~
~~ E. N. Wise: J. Am. Chern. Soc.
75, 2269 (1953), found a log-
J
J--'> arithmic rate law at 250 and
316°C which could not be
8 explained by field transport
5 :J(J 30 100 phenomena, since layer thick-
Time nesses of 4000-7000 A are in-
volved.

~--"

I ~··
~
L----ft

L--- V--
~

J Fig. 38. Oxidation of tantalum


in oxygen at 350°C in a log-log
plot, according to Waber. The
h cubic tarnishing law resulting
Time - - from this reads: t = const t 1/ 3 ,05.

1 Belle, J., and M. W. Mallett: J. Electrochem. Soc. 101, 339 (1954).


2 Engell, H. J., K. Hauffe, and B. Ilschner: Z. Elektrochem. 58, 478 (1954).
3.5. Effect of Electric Fields on Metal Oxidation 121

that in the oxidation of nickel in oxygen at different pressures at 400°0, a


rate law of the form g '" t l / 3 . 7 prevails, but that at increased layer thicknesses
(> 800 ~ 1000 A) there is a transition to a parabolic rate law (Figs. 39a and
39b). In this case it was observed that the point of transition to the para-
bolic rate law was displaced to greater layer thicknesses with increasing
2000r----.-----,---r-,-.-----r----,---,-,-~~~

A
110M~--~----~~~~~~~~~~~~~~~
~~---+----4r~~~+.~~~--~~~~4---~
~
c: 6W t-,------=01___=-"'---I-c.t-""'f=--z-P'''''-----:--1-
-"
u •
i§ 'lQ1l 1-""2~--+=-----:~--.=-t---j".....+<=-:c-----j4 :P02= 30 mm Hg:-'-i;-r;-:;:-:o-I
.... J IS :Oxidation of Pure icke/.

!
$-- :P02= 240mmHg According to Gulbransen
9 2: ". = 120" and Andrew. at 475°C and
&W~--~----~--~3~:~.. -==~6~0~-
.. ~-P~O~2~=-7~6~m-=m~H~g~.._--~
1()' 1U~ fU J
Minutes _ _

Fig. 39a. Oxidation of nickel at 400°C and different oxygen pressures


in a log-log representation, according to Engel!, Hauffe, and
Ilschner. The exponent here was 1/3.3 to 1/3.6.

Fig. 39b. Parabolic part of the oxidation curve


of nickel at 400°C at Po, = 30 mm Hg, according
to Engel!, Hauffe, ana Ilschner. As is shown
in Fig. 39a, a predominantly parabolic oxidation
begins under these experimental conditions
at a NiO thickness of about 500 A. 41l 50 min
v'f/me

temperature as well as with increasing oxygen pressure. A similar observation


was also made by Tylecote in the oxidation of copper in air containing water
vapor at temperatures around 500°0; the transition to the parabolic rate law
occurred after an oxidation time of about 10 hr. We do not yet have an
explanation for the initial parabolic course of the 370 and 420°0 curves (Fig.
36), which shows a decrease in slope after 20 hr.
The continual changes in the parabolic oxidation constants with oxida-
tion time, observed by Gulbransen and Andrew,l are due to the fact that at
smaller NiO-Iayer thicknesses, oxidation does not follow a parabolic rate law,
but in the sense of the above discussion follows a cubic law, where space-
charge effects are responsible for the progress of the oxidation. In Fig. 39a,
the 475°0 curve of Gulbransen and Andrew is evaluated according to equation
(3.74). The fact that the measurements of Hauffe and co-workers yielded
approximately the rate constants at 400°0 which Gulbransen and Andrew
1 Gulbransen, E. A., and K. F. Andrew: J. Electrochem. Soc. 101, 128 (1954).
122 3. The Mechanism of Oxidation of Metals-Theory

found at 475°C is probably due to the greater purity of nickel in the latter
case.
Spurred on by these results, Uhlig and co-workersl studied the initial
oxidation rate of nickel around the Curie temperature. By means of a 24-hr
oxidation period within the temperature range 307 to 442°C, NiO films up to
3400 A maximum were produced on nickel foils. A discontinuity in oxide
growth was observed at 353°C, the Curie temperature. Furthermore, a two-
stage oxide growth rate could be identified-a higher rate follows an initially
lower rate. Both stages could be represented by a linear relation between
oxide thickness and log time. Uhlig could show that the experimental
results in Fig. 39a also follow the two-stage logarithmic equation.
Finally, it is thought that the change of work function of nickel above
the Curie temperature might be responsible for the change in the growth rate.
For this concept, however, further experiments are necessary. The chemi-
sorption and initial oxidation rates of atomically clean nickel crystal faces
were investigated by Farnsworth. 2
The oxidation of Cu 20 sheets between 800 and 1OOO°C at different oxygen
pressures yielded a surprising result: a cubic rate law was found which
exhibited fair agreement with an exponent of! (Fig. 40). We concluded from
this that there is a rate-determining field transport of ions through the CuO

1000
QJl:m 6 .!: 810°C up 02=1 Atm
r--- x 810 or uPOa =46mmHg /
/..
.900
° 810°C uPOz = 5mmHg
17 81f8 °G U PO> =f Atm C·1,8{{}13
800
~y
t I /
700
V Y1c.t5510-fJ
~
/ ./
..,~600
.,/ V
?Ib-- /
~500

"00 f - --
V/ V
/'
V
v l / kc
17~
-9.3-fO"·
./ I
300

{{ Vx-
x

I~
0

~/
200 ...- Fig. 40. Oxidation of CU20
~
.--r
~/ ~ ~t10'" at 810 and 848°C at different
fOO .-Jl- f--" I oxygen pressures, according
to Hauffe and Kofstad. The
~~
~/"
oxidation follows the cubic
o 600 1200 !SOO 2~OO 3000 3600 9200 1fIJ00 5.,JO rate law with exponents of
Seconds_ 1/2.9 to 1/3.1.

1 Uhlig, J. Pickett, and J. MacNairn: Acta Met. 7, 111 (1959).


2 Farnsworth, H. E., and J. Tuul: J. Phys. Chern. Solids 9, 48 (1959).
3.5. Effect of Electric Fields on Metal Oxidation 123

protective layer which is formed. 1 This finding is an impressive indication


that the temperature region in which field-transport processes become rate-
determining is strongly dependent on the kind of lattice defect and the
height of the saddle barrier which must be overcome in the migration through
the oxide layer formed. The logarithmic oxygen pressure dependence of the
oxidation rate is to be regarded as further evidence toward confirming that
field transport in the CuO layer is rate-determining (Fig. 41).2
cO
1. -,
g'cm ·see V
0/°

c/V.
,/0
Fig. 41. Dependence of the oxidation
rate of CU20 at 810°C on the oxygen
pressure with similar pretrcatment of
the CU20 samples (annealed for 1 hr

V,
in high vacuo), according to Hauffe
and Kofstad. The cubic tarnishing
constants are plotted against the
logarithm of the oxygen pressure in o .J
mmHg.

Meijering and Verheijke 3 confirmed the experimental results concerning


CU20 oxidation. They believe, however, that the cubic rate law is caused by
an aging proccss in the CuO scale during oxidation. Hence, they use the
following equation:
(3.74a)
where k is not constant but is a function of the aging time,
k = B{J + (1 - I) exp ( - k't)} (3.74b)
Integrating (3.74a) and introducing (3.74b) into (3.74a), we obtain

~ = 2(1 - f) B{1 - exp (-k't) + _I_kIt}


k' 1- I
(3.74c)

If the values of the parameter 1/(1-1) are between about 0.3 and 0.5 then
the cubic rate law is satisfactorily fulfilled with mean values of k't. However,
if the oxidation time is either very short or very long, then A = B and we
get a parabolic rate law. The small temperature coefficient of the oxidation
1 The relatively great layer thickness of about 20,000 A may have been made up of an
inner compact layer about 2000 A thick (field transport) and an outer porous layer
(grain-boundary diffusion).
2 Hauffe, K., and P. Kofstad: Z. Elektrochem. 59, 399 (1955).
3 Meijering, J. L., and M. L. Verheijke: Acta Met. 7, 331 (1959).
124 3. The Mechanism of Oxidation of Metals-Theory

rate is proof of this mechanism. Specifically, with increasing temperature,


the aging rate increases and the oxidation· rate will thus be decreased.
Oxidation of aluminum appears to be similarly governed, and Smeltzer1
has found that under special conditions, the Al 2 03 formation at 500°C can
become more extensive than that at 600°C.
Still, the extent to which a cubic rate law is the result of field transport
cannot always be determined. We may note in this connection the investiga-
tions of Waber and Sturdy. 2 A cubic rate law was also found in the attack of
moist oxygen on thallium at 38°C. Further investigations into the structure
and semiconductor characteristics of the tarnishing layer could be carried
out in order to ascertain the validity of the mechanism discussed above. In
any case, from previous investigations, in practically all metals and alloys at
intermediate and lower temperatures, deviation from the parabolic oxidation
law seems to be the rule.
As a supplement to the above discussions it may be noted that it has
been experimentally verified that the cubic rate law which is observed in
titanium and zirconium oxidations is not caused by field transport phenom-
ena, but by other additional processes-in this case by the presence of an
oxygen solution in the metal or by aging processes. We therefore have to
take into consideration, in addition to an oxide layer formation, the process
involving solution and diffusion of oxygen in the metal. After a discussion
with Carl Wagner, we believe that the quasi-cubic rate law is caused by a
superposition of two parabolic rate steps, whereby the rate constant of the
gas uptake into the metal becomes smaller as the oxidation time increases.
The initial rapid oxygen uptake indicated in Fig. 42a is due to the large

I I
1
Ti
Fig. 42a. Schematic representation of the oxida-
tion of titanium under consideration of a finite
oxygen solubility in titanium in equilibrium with
the oxide: (a) at the beginning of oxidation;
(b) during the oxidation; and (c) solubility
equilibrium.
Smeltzer, W. W.: J. Electrochem. Soc. 103,209 (1956).
2 "Taber, J. T., and G. E. Sturdy: J. Electrochem. Soc. 101, 538 (1954).
3.5. Effect of Electric Fields on Metal Oxidation 125

concentration gradient of oxygen in the metal. However, during the course


of oxidation the concentration gradient becomes smaller and smaller and
thus the rate of oxygen absorption decreases as well. During this oxidation
period we get a continuous decrease of the rate of total oxygen consumption,
which is the reason for the cubic rate law observed. The change from the cubic
to the parabolic rate law for the oxidation of titanium between 438 and 592°C
is shown in Fig. 42b. 1

1()2

..
N

v
E
......
<>
o
536°

~~:
x
-; 10'
\!)

~43"C
10 1 1()2 1()3 10'
Ox idation Time in min
Fig. 42b. The cubic time law of the oxidation of Kroll titanium
in dry air of 1 atm, according to Kofstad, Hauffe, and Kjollesdal.

3.5.4. Formation of Very Thin Tarnishin~ Layers-


The Lo~arithmic and Reciprocal-Lo~arithmic
Rate Laws
If we now go on to a discussion of very thin tarnishing layers (of the
order of 20 to 80 A) as they appear in the oxidation of metals and alloys and
in the formation of passive layers, e.g., on iron in nitric acid, at low tempera-
tures (0-200°C), we must begin to consider other phenomena responsible for
the reaction mechanism for this kind of thin layer. In the ensuing discussion
we will draw upon the Cabrera-Mott theory, and development ofthe theoret-
ical relationships begins with the following considerations:
1. The layers are so thin that the space charges contained in them may
be neglected to a first approximation, and this is valid even in p-conducting
tarnishing layers. It is thus sufficient to consider Mott's model, which is
limited to surface charges of equal magnitude but opposite signs on the two
sides of the tarnishing layer.
2. The very thin tarnishing layers that appear permit an electron
transfer from the metal to the chemisorbed oxygen via the quantum mechan-
ical tunneling effect, up to about 40 A, even when thermal emission of
1 Kofstad, P., K. Hauffe, and H. Kjollesdal: Acta Chern. Scand. 12, 239 (1958).
126 3. The Mechanism of Oxidation of Metals-Theory

Fig. 43. Schematic representation of the


saddle jump of the metal ions for vacancy
migration in the presence of an electric
field in the tarnishing layer (a 9= a*;
U1> U2); j ~jo'exp{ -(U ± tUE)/kT}
with UE = zea (f [see (3.75)].

electrons into the defects of the valence band or into the conduction band of
the oxide, or vice versa, is precluded.
3. For the thin layers under discussion here, the energy UE which a
defect with a z-fold charge within a lattice distance a can obtain from the
field is comparable to its thermal energy Uth, especially at low temperatures
(a is defined in Fig. 43):

zeaV
kT = Uth ~ UE = zea(f = - - (3.75)
g
In the absence of a concentration gradient, however, the rate j at which a
charge carrier moves through a constant field in the lattice is given byl

j = -kT a exp ( - Uth/kT) { exp ( - - U


E ) - exp (U 11 )}
--- (3.76)
h 2kT 2kT
since the field increases the potential barrier in one direction and decreases it
in the other. If U E < kT-which is frequently the case at higher tempera-
tures-only terms of the first power in UE/kT in the series expansion of the
exponential function in equation (3.76) need be considered, and one obtains
the linear dependence of the field current on (f that was used in the preceding
section. However, if UE > kT-which is usually the case at lower tempera-
tures-the rate of migration j of the particles in the tarnishing layer is
exponentially dependent on the field strength, which in turn is inversely
proportional to the thickness of the layer. This leads one to expect a rapid
ion current through the layer.
4. On basis of the foregoing, the further assumption, introduced by
Cabrera and Mott, of a rate-determining transfer of metal ions from the
metal into the oxide phase or transfer from the oxide into the chemisorption
1 Mott, N. F.: J. chim. phys. 44, 172 (1947).
3.5. Etlect of Electric Fields on Metal Oxidation 127

layer is open to challenge. In that case a phase-boundary reaction will be


rate-determining. Such an assumption is, however, not clearly a consequence
of electric fields, as Mott stated. A more detailed consideration of these
relationships was given by Hauffe and Ilschner,I but the latter has been
subjected to certain limitations by the recent investigations of Dewald (see
8.7.2.1). The other case-rate-determining field transport through the tar-
nishing layer-occurs whenever the rate-determining activation energy VI
for the transfer of a metal ion from the metal phase into the tarnishing layer
is smaller than the activation energy V2 for migration within the layer. In
Fig. 43, VI > V 2 , so that the transfer is rate-determining.
If we assume a sufficiently rapid ion and electron transport through the
tarnishing layer, the field influences the rate-determining transfer of ionic
defects in the lattice in such a way that the determinative activation energy
VI is reduced to about VE (Fig. 43). If we consider both (3.75) and our earlier
discussion of the number of defects passing per unit time and surface,
11,1 = Al exp (- Vi/kT)
is finally obtained as the expression for the growth of the layer thickness:

(lZaV)
-d6 = const.exp - - - = const.exp (go/6) (3.77)
dt 6 2u
from which

go=-
zaV
2u

Strictly speaking, the distance a* of the first lattice plane of the oxide
from the metal surface should be used rather than the lattice constant a of
the oxide (Fig. 43). However, a ~ a* gives a sufficiently good approximation.
In order to write the differential equation (3.77) corresponding to the time
law of oxidation, we may consider the case where go > 6 (go;;;; 26 would
already be sufficient; a closed integration is not possible for go ~ gil. We
may then write instead of (3.77)

d ~l ~0 (I: /1: ) const ~o (I: /1: ) (3.78)


at = const ~- exp '>0 '>1 """ ~o + 2~1 exp '>0 '>1

By separating variables and integrating, we obtain

with

1 Hauffe, K.: Reaktionen in und an jesten Stoffen, pp. 565 ff. BerlinjGottingenjHeide1berg,
Springer, 1955, pp. 565ff.
128 3. The Mechanism of Oxidation of Metals-Theory

or

(3.79)

to is a constant of integration. Equation (3.79) is called the reciprocallogarith-


mic tarnishing law. In the literature it is frequently referred to, incorrectly,
as the "logarithmic tarnishing law," and this misnomer can lead to its pos-
sibly being confused with the actual logarithmic time law. Equation (3.79) is
generally fuund in the form

:1 = A' - B'lnt (3.80)

An expression of this form does not lead of course-as one can easily show-
to (3.77), which was derived by Mott. For the range of values in question
here, it can show that the logarithmic member of (3.79) containing 6 changes
only very slightly. If we set 6 equal to an average value g :::; 20 A, then
we are in a position to give the constants in the approximation (3.80) by
means of solution (3.79), as follows:

(3.81)

According to (3.79), at a certain stage, oxide layer growth decreases so


sharply that a critical layer thickness g* can be defined, above which the
increase in layer thickness with time is so slight that it is practically unob-
servable. 1 In this connection Cabrera and Mott compare their theoretical
treatment with the experimental results of Giinterschulze and Betz. 2 These
authors found the exponential dependence of the current on the field strength
(f discussed above in the anodic oxidation of aluminum in suitable aqueous
electrolytes:
i = ex exp(fHj;) (3.82)

A similar time law was discussed earlier by Verwey.3 With the help of (3.78)
and (3.81), the constants a and f3 in (3.82) may be related to those in (3.80).
If V = 2 volts, we obtain a critical layer thickness of about 20 A which, in
fact, was found by Cabrera, Terrien, and Hamon 4 in oxidation experiments
on transparent vaporized aluminum layers. (Light transmission and reflec-
tion were measured optically as a function of time.)

1 Definition by Cabrera and Mott: The reaction will practically stand still when 10 5 sec
is required for the building up of a new lattice plano.
2 Giinterschulze, A., and H. TIetz: Z. Physik 92,367 (1934).
3 Verwey, E. J. W.: Physica 2,1059 (1935).
4 Cabrera, N., J. Terrien, and J. Hamon: Compl. rend. 224,1558 (1947).
3.5. Effect of Electric Fields on Metal Oxidation 129

One must evidently assume similar relationships for the oxidation of


copper at low temperatures down to about 30°C, where Evans,l Rhodin,2 and
Lustman and Mehl 3 have found a reciprocal logarithmic rate law. Above 50°C
a transition to the cubic rate law is frequently observed. Furthermore, an
oxygen pressure dependence of the oxidation rate was found, which will be
discussed later. The finding reported by Campbell and Thomas4 of a decrease
in the oxidation rate (about 10% between 100 and 160°C) in the presence of
water vapor is understandable, if one considers that through the chemisorp-
tion of water according to
H 2 0(g) + EBIR) ___ H20+IO)

the concentration of electron holes in the boundary layer nG) (R) as well as the
contained space charge were reduced, causing a slowing dQwn of the field
transport. A detailed and precise explanation of the mechanism of oxide-
layer formation is generally made more difficult by the fact that at lower
temperatures the CuO layer covers the Cu 20 layer more and more with
decreasing temperature. Thus, it was observed by Tylecote 5 that at 700°C
the protective layer consists of about 35% CuO, and at 300°C approximately
95%.
If we discard the supposition of a constant potential difference intro-
duced by Mott and assume a constant electric field strength independent of
the layer thickness, according to Grimley and Trapnell,6 we obtain in the
case of aluminum oxidation 7 a linear rate law, with a rate-determining field
transport of aluminum ions via interstitial lattice positions
dg U
- = UAI Q - nAI exp (3a/,\) (3.83)
dt 3a
where UAI and nAI are, respectively, the mobility and the concentration of
aluminum ions in interstitial positions and Q is the oxide volume per metal
ion. Further,
IOU
,\=--
27Ten(<T)

1 Evans, U. R., and H. A. Miley: Nature (London) 139,283 (1937).


2 Rhodin, T. N.: J. Am. Chem. Soc. 72, 5102 (1950).
3 Lustman, B., and R. F. Mehl: Trans. AIME 143, 246 (1941).
4 Campbell, W. E., and U. B. Thomas: Trans. Electrochem. Soc. 91, 623 (1947).
5 Tylecote, R. F.: J. Inst. Metals 78,301 (1950/51).
6 Grimley, T. B., and B. M. W. Trapnell: Proc. Roy. Soc. (A) 234, 405 (1956).
7 New measurements by means of radioactive tracers (125Xe) as markers during the
anodic oxidation of aluminum make it probable that both metal and oxygen ions migrate
through the oxide layer, but that at 0.1 mA/cm 2 the film grows mainly by oxygen ion
migration [J. A. Davies, J .. P. S. Pringle, R. L. Graham, and F. Brown: J. Electrochem.
Soc. 109, 999 (1962)].
130 3. The Mechanism of Oxidation of Metals-Theory

so that in every case a saturation of chemisorption is supposed, e.g., n(u) ~


N(u) [N(u) is equal to the number of adsorption sites per square centimeter
of surface.] Further experiments in this area are desirable.
We are confronted with a reciprocal logarithmic rate law in the oxide
film or passive layer formation on iron in suitable electrolytes-especially
HN03-HN02 solutions'! About ten years ago in a work on the dependence of
ionic transport in such passive layers on an electric field set up in a HN0 3-
HN02 solution or produced in a 1 N H2S04 solution by anodic polarization,
Vetter2 was able to produce experimental evidence for the validity of the
assumption of a similar mechanism. As Miley and Evans3 were able to show
for the attack of moist air on iron at 18°C, and Bonhoeffer and Vetter4 for
the attack of dilute HN0 3 on iron, the main part of the passive layer consists
of an oxidized material with trivalent iron ions, probably Fe 203. Electron
diffraction measurements and film stripping experiments of passive iron
carried out by Cohen 5 indicate that the passive layer is indeed composed
mainly of y-Fe203. While above 200°C oc-Fe 203 appears preferentially, it
seems, according to Miley and Evans, that below 200°C the formation of
y-Fe203 is preferred. Since Fe 203 is apparently an electron-excess conductor6
with a very small equivalent concentration of free electrons and oxygen ion
vacancies, we can conclude agreement with Mott's first assumption. The
thickness of the passive layer was found to be from 40 to 80 A.7 Figure 44

-4f-----I---------,i'-----hd----\P'--------I

t
~
-<::
'";; -51-----'-"-7IF'==-=--=--=-=7'1= -;I'''---i------.-r------j-----------j
::::
:::. I
.§ i 0

-~~~--~JO~--t~.2~--~1.~---1~.8~V~·~
Fig. 44. Experimentally obtained de.
pendence of the current density on
the potential of passive iron for differ-
ent stationary initial conditions E s ,
Iron Potential _ according to Vetter.
1 Hauffe, K.: Z. MetaUk. 44, 576 (1953).
2 Vetter, K. J.: Z. Elektrochem. 58, 230 (1954).
3 Miley, H. A., and U. R. Evans: J. Chern. Soc. 1295 (1937); U. R. Evans, Metallic
Corrosion, Passivity, and Protection, Longmans, London, 1948.
4 Bonhoeffer, K. F.: Z. Metallk. 44, 77 (1953); K. J. Vetter, Z. Elektrochem. 55, 274,
675 (1951).
5 Cohen, M.: Can. J. Chern. 37,286 (1959); see also J. E. D. Mayne and M. J. Pryor:
J. Chern. Soc. 1831 (1949).
6 Verwey, E. J. W., P. W. Haayman, and F. C. Romeyn: Chem. Weekblad 44,705 (1948);
K. Hauffe, Metalloberflache, A. 8, 97 (1954).
7 Vetter, K. J.: Z. Elektrochem. 55,121 (1951).
3.5. Effect of Electric Fields on Metal Oxidation 131

shows the experimentally obtained dependence of current density i on the


potential Vh of the passive layer for different stationary initial conditions.
Since in these thin passive layers-just as in the electrolyte system AljAl20a
-Ohm's law is no longer valid, and an exponential dependence on the
field strength analogous to (3.82) prevails, Vetterl formulated the following
equation, for the transfer (Durchtritt) potential 2 :

i=io{exp(~~z ,:) -exp ( - (l~iF z ~)} (3.84)

The expression for the current density when zVjm ~ RTjF is then
. = to. exp (IXF
~
V)·
RT Zm (3.84a)

However, these expressions may be altered in the same way that (3.82) was
obtained from (3.78) if we seV the factor at ~ t and the activation barriers
m = 6ja. A question still remains as to which of the two possible phase-
boundary reactions-entry of a chemisorbed oxygen ion into a vacancy in
the Fe20a at the Fe 20ajelectrolyte phase boundary or formation of an oxygen
ion vacancy in the Fe203 at the Fe203jFe phase boundary-is rate-deter-
mining or whether the field transport does not primarily determine the rate.
Vetter'sl investigations seem to climinate the first possibility.
Ifindeed VI < V 2 (Fig. 43), then-always assuming a sufficiently rapid
electron transport through the passive layer which is formed-the phase-
boundary reaction proceeds more quickly, leaving the field transport of
the ionic defects as the slowest and, therefore, rate-determining step. Also
under these conditions we obtain a rate law of the form of (3.80). Further
investigations could be suggested to determine whether the experimental
results by Vetter can be better described by this mechanism. (A detailed
discussion will be found in Chapter 6.)
In some cases the oxide-film formation with thin p-type layers can be
described by a logarithmic rate law of the form
dg
- = const. exp (-ggo) (3.85)
dt
and as an example, we will discuss the Cu 20 layer formation on copper. Under
the assumption of a prevailing field transport of the copper ions through the
layer via vacancies we obtain for the rate of growth of the Cu 20 layer the
following expression, identical to (3.83):
dg 0
- = UCuO' Q - ncuo,(g) exp (a/A) (3.86)
dt a
1 Vetter, K. J.: Z. Elektrochem. 58,230 (1954).
2 Vetter, K. J.: Z. physik. Chem. [NF] 4, 165 (1955).
132 3. The Mechanism of Oxidation of Metals-Theory

where ncuo,(g) and UCuo' are the concentration of copper ion vacancies at
the CU20j02 phase boundary and the mobility of the vacancies, respectively.
>. is the same factor as in (3.83). The production of the vacancies is deter-
mined by the following equilibrium

!02(gas) + 2e(Cu) ~ CU20 + 2CuD'W (3.86a)


with the corresponding law of mass action

ncuo,(g) = PlJ; exp (-JpIn) exp (Vjn)


Jp. is equal to half the standard free energy for the above reaction (3.86a).
When the diffusion coefficient D cllo ' is substituted for ucuo', we obtain a
logarithmic rate law for copper oxidation:
dg
- = k exp (-2gp,) (3.86b)
dt
with
Dcuo'
k = -- SJplJ4 exp (-LJp.jo) exp (ajA)
a '
The dependence of the oxidation rate on the oxygen pressure under field
transport conditions noted before now becomes understandable.
A rate-determining electron supply is another possible cause for the
appearance of the logarithmic rate law. In 1939 Mottl tried this approach
in order to explain the logarithmic course of zinc oxidation below 225°C
observed by Vernon and co-workers.2 Measurements, in 1953, on the effect
of oxygen on nickel between 0 and 200°C-carefully carried out experi-
mentally by Scheuble3-provided Hauffe and Ilschner4 with support for
Mott's approach, and they presented a further explanation of the logarith-
mic time law. Scheuble found a logarithmic rate law for the consumption of
oxygen, which is not caused, as he assumed, by a diffusion of oxygen in a
compact metal lattice, but rather by formation of a NiO film of the type
6 = goln(t + to) - const (3.87)
corresponding to the differential equation (3.85). The explanation of this rate
law is difficult if one does not apply the Grimley-Trapnell relationship
(3.86).
Evans 5 has shown that a law of the type of (3.85) is also obtained if one

ll\Iott, N. F.: J. Inst. Metals 65, 333 (1939).


2 Vernon, \'\'. H. J., E. J. Akeroyd, and E. G. Stroud: J. Inst. Metals 65, 301 (1939).
3 Scheuble, W.: Z. Physik 135,125 (1953).
4 Hauffe, K., and B. Ilschner: Z. Elektrochem. 58,382 (1954).
5 Evans, U. R.: Trans. Electrochem. Soc. 91, 547 (1947).
3.5. Effect of Electric Fields on Metal Oxidation 133

assumes a decisive influence from the blisters and fissures which are statisti.
cally distributed in the oxide, but these porosity phenomena, which are
understandable in thick oxide layers, are hardly appropriate in the oxide
films treated here, which are only a few atom layers thick. As we have seen
above, generally the number of electrons entering through the surface per
unit of time and surface is greater than the number of ions. This is also
understandable according to earlier experiments on semiconductors. If we
turn now to very thin tarnishing layers ( < 30 A), we note that ionic transport
through the oxide layer must take place quickly because of the very high
electric fields-greater than 10 7 volts/cm. The question may now be asked
whether the electron transport required on basis of electroneutrality is
always able to keep up with this process. The results indicated by Scheuble
seem to show that this is not always the case with nickel oxidation. Let us
assume that in very thin oxide layers-and only there-electron transport
is the slowest step. That means that in this case chemisorption equilibrium
never sets in; it may then be that the external oxygen pressure is extremely
low. Under these conditions the oxidation rate must be independent of oxygen
pressure, which is also inferred from the experimental results of Scheuble.
The high resistance of the oxide film to electron flow which is assumed
here would be caused by the impediments to electron transfer generally
present at the oxide/gas phase boundary as well as the special characteristics
of the "band structure" of very thin semiconducting layers in the presence of
high electric fields. In Fig. 45 the relationships are represented schematically.

Transition unknown

..
Fig. 45. Local dependence of -=
the ion and electron current t
in very thin tarnishing layers,
according to Hauffe and
Schottky. The rate is deter-
mined in region I by elec-
tron transport, in region II II
by ion transport. The tran-
sition region is still unknown. -+ Thickness of oxide layer

One can see the possibility that with very small layer thicknesses not as
many electrons succeed in passing through the layer (region I) as ions, which
migrate under the influence of the fields, assuming that there are sufficient
electrons present. The appearance of these thin layers evidently involves the
134 3. The Mechanism of Oxidation of Metals-Theory

influence of an entirely new effect: electron transport from the conduction


band of the metal to the chemisorbed oxygen by means of the wave mech-
anical tunnel effect. Since here the number of the electrons "tunneling"
through the oxide film per unit time and surface decreases exponentiallyl
with g, the previously mentioned assumption leads directly to an equation
of type (3.87) and thus to the experimentally observed rate law (3.85). From
a suitable graphic plot of the data for nickel oxidation at, for example, 200°0
~r--------'--------------'---,

~~--------~--------~---+--~

t..
~

W~------~L-------------+---~

Fig. 46. Oxidation of nickel at 200°0


in air, evaluated by Hauffe and Ilschner
from the experimental results of
mL--------=~------------~~~
Scheuble (logarithmic time law).

(Fig. 46), the constants on the right side of (3.87) may be empirically deter-
mined. Here go is obtained as about 1-2 A, and is related in a known manner
to the "height" <l> of the potential wall, which we assume to be perpendicular
for the sake of simplicity, and the electron mass m:
h ,/--
$0 = ---
4n
v2m <l>
With <l> ~ 1 eV we obtain a value for go which is also about 1 A in good
agreement with the value obtained empirically from the experimental
results.
Since sometimes in earlier investigations, ion and electron inhibition
was introduced in order to interpret the natural and reciprocal logarithmic
rate laws, it appeared desirable to subject the suppositions and limitations of
these inhibition phenomena to critical examination. 2 Under certain assump-
tions, the relation Idsel/dgl < Idsion/dgl for region I in Fig. 45 and the sub-
sequent calculation of the electric current of Sel is quite reasonable.
For region II in Fig. 45, we obtain for the course of the ionic current
Sion with constant V:

(sionhr = C exp (,8 ;)


1 See appropriate text books in theoretical physics.
2 Hauffe, K., and W. Schottky: Halbleiterprobleme, Vol. 5, Berlin, 1960, pp. 316-319.
3.5. Effect of Electric Fields on Metal Oxidation 135

Under these conditions the electrons rush on ahead and give rise to a moving
electric field Vjg = <f which satisfies <faj2 > 1). The corresponding relation
for (selhI has to be calculated with V ~ 0 and <I> = const. Finally, we get
for (Sion)n the following expression:

W + LJp
(sionhI = 0 exp {J ---
g
where W is the energy difference between the trap level and the metallic
Fermi level and AV == In njn° = const, with the surface trap concentration
n° occupied by electrons.
It was found that the extrapolation from region II into region I, pre-
viously performed by US,l is not justified. Strictly speaking, therefore, it is
not even certain that a region with predominant electron inhibition is
involved at all. The counter potential which forces the electrons backward
in region II, in the case of small g values, might decrease (and possibly even
reverse) to such an extent that the tunnel current becomes faster than the
ion current, so that the ion inhibition becomes determinative even in the
case of small g values. The observations in Fig. 46 certainly do not represent
any decisive counterargument in this regard. On the one hand, at small g
with constant V the reciprocal logarithmic curve has a region which can be
represented approximately by a logarithmic curve, and on the other hand
the V variation with g could change the reciprocal logarithmic curve in the
required sense. From these points of view, the space-charge theory with a
field-dependent mobility of ions discussed earlier can be applied to the
representation of Sion and dgjdt.
The existence of a precisely valid time law is doubtful, however, since
in the case of such thin layers inhomogeneities due to lateral extension in
"repeated steps," discussed by Stranski,2 and the formation of new nuclei
on the oxide already formed have to be taken into account. Under these
conditions the application of equation (3.85) is only a very crude approxima-
tion to the actual relationships, but because of the inherent irregularity of
the oxide layer, the deviations from the logarithmic rate law which actually
envelops the true course of the reactions are averaged out within the accuracy
of measurement.
In this connection, a work of Hirschberg and Lange 3 is of interest.
These authors have oxidized fresh oxide-free zinc surfaces at temperatures
between 20 and 407°C for 20 min, and'thereby obtained layers of different

1 Hauffe, K., and B. Ilschner: Z. Elektrochem. 58, 382 (1954).


2 See for example O. Knacke, and I. N. Stranski: "Die Theorie des Kristallwach-
stums," Ergeb. exakt. Naturw. 26, 383 (1952).
3 Hirschberg, R., and E. Lange: Naturwiaa. 39, 187 (1952).
136 3. The Mechanism of Oxidation of Metals-Theory

thicknesses, on which they carried out Volta potential measurements. It was


shown that at low temperatures, i.e., very small layer thicknesses, positive
Volta potentials exist which always decrease with increasing layer thickness
(because of increasing oxidation temperature) and finally reverse their signs.
Since the value of the Volta potential denotes the "sign" of one of the two
types of charge carriers, the results of Hirschberg and Lange can be explained
as a transition from "region I" to "region II" in Fig. 45.
As mentioned before, a logarithmic rate law seems to rule in the Cuj02
system under certain conditions, as can be concluded from the measurements
of the rate of oxidation by Evans l and Rhodin 2 (Figs. 47 and 48). In these

feo r
f
~ ~ - ~

100

80 r- -r- SOC

....]I
80

20

o Fig. 47. Oxidation of copper at low


temperatures, according to Evans and
Time Miley.

,
0
,
,,
0
, ,-6'
/
0
,
V" Fig. 48. Oxidation of copper between

20
V
.-P' ~~
78 and 353°K with a logarithmic time
plot, according to Rhodin. While the
plots in 1 to 5 obey the logarithmic
~o- ~
,pi>
.-~~---
J ---- tarnishing law from 10 sec to 30 hr, the
2 ---- oxidation at 353°K (or 80°C) had already
~ 1--r'---- changed over to a cubic law after ! hr.
o10,-J 10 -,;> 10 -f
t 10 tOO (Curve 1: 78°K; 2: 273°K; 3: 298°K;
Hours_ 5: 323°K; 6: 353°K).

1Evans, U. R., and A. H. Miley: Nature (London) 139, 283 (1937).


2Rhodin, T. N., J. Am. Chem. Soc. 72, 5102 (1950); see also W. E. Campbell and U. B.
Thomas: Trans. Electrochem. Soc. 91, 623 (1947).
3.5. Effect of Electric Fields on Metal Oxidation 137

measurements a natural as well as a reciprocal logarithmic tarnishing law


was found.
White and Germerl found a logarithmic rate law at 20°0 and P0 2 = 20
mm Hg of the form
g = 4 + 6.5 log t
where g is calculated in Angstroms and t is in minutes. The appearance of the
two rate laws can have a decisive influence on the composition of the tarnish-
ing layer. Where the rate law is logarithmic, the system does not attain the
oxygen chemisorption equilibrium on the surface of the tarnishing layer
because of an "electron deficiency" so that although the oxygen pressure
may be sufficiently high for OuO formation the chemical potential of the
chemisorbed oxygen required for the OuO formation is not reached. Rapid
field transport of copper ions through vacancies keeps the stationary surface
concentration of the oxygen low. A predominant CuO formation does not
take place until larger layer thicknesses-which are favored by the increases
in temperature and in oxygen pressure-are produced, because of the larger
electron supply as compared with the number of electrons available at the
ionic chemisorption equilibrium with a sufficiently high surface concentration
of chemisorbed oxygen ions. The growth process of the tarnishing layer will
be different depending on the experimental conditions, which give rise to
either a natural or a reciprocal logarithmic rate law, even though, in terms of
oxygen pressure, we are always in a region where the OuO phase is stable.
These considerations are of general validity and are significant in other
oxidation systems as well, where oxides of different valences can appear,
e.g., oxides of iron, cobalt, and titanium.
A logarithmic rate law can also be expected if the chemisorption of the
oxidizing gases is regarded as the rate-determining step in the total reaction.
Landsberg 2 called our attention to this in a theoretical treatment which is
in agreement with our chemisorption theory.3 However, this mechanism-
especially in very thin oxide layers-is open to discussion.

3.5.5. The Nonparabolic Tarnishing Law as a Result of Blocking


and Cavity Formation at the Metal/Oxide
Phase Boundary
Until now our discussion has included only oxidation during which a
compact and pore-free scaling layer is formed on the metal, that is, the case
1 White, A. H., and L. H. Germer: Trans. Electrochem. Soc. 81, 305 (1942).
2 Landsberg, P. T.: J. Chem. Phys. 23,1079 (1955).
3 Hauffe, K., and H. J. Engel!: Z. Elektrochem. 56, 366 (1952); H. J. Engel! and K.
Hauffe: Z. Elektrochem. 57, 762 (1953).
138 3. The Mechanism of Oxidation of Metals-Theory

where the vacancies migrating to the metal/metal oxide phase boundary


during oxidation do not precipitate as large cavities, but are destroyed by
the plastic flow of both the protective layer and the metal. At intermediate
and lower temperatures, these conditions are frequently no longer operative,
so that there is pore formation, and frequently the protective layer bursts.
Brasunas l reported on the mechanism of pore formation and blocking. In
Fig. 49 pore formation can be seen in an Inconel alloy directly beneath the

- .. ..
'

•.- • •
.'

/ ~ .•
\
I ••
0
f •·

Fig. 49. Photomicrograph of an Inconel alloy oxidized for 210 hr in air at


1250°0, according to Brasunas.

oxide layer which was obtained by oxidation in air at 1250°0 for 210 hr.2 In
this case the vacancies which were formed at the metal/metal oxide phase
boundary have migrated into the metal phase. At lower temperatures,
however, pore formation takes place predominantly in the oxide layer. The
forces favoring plastic flows and thus "healing" of the pores and removal of
blocking are due to the stresses in the interior limiting the growth of oxides
on metals with different molar or atomic volumes. These internal stresses,
which at higher temperatures-since they are naturally not too large-exert
a b eneficial influence on the formation of compact scaling layers, frequently
cause fracture in the oxide layer at lower temperatures, where the "plasticity"
of the crystal in the scaling layer has become very small. This rupturing
1Brasunas, A. de S.: Metal Progr. 62, 88 (1952).
2According to new investigations this pore formation seems to be caused by a preferential
oxidation of chromium to OrO which evaporates into the gas phase.
3.5. Effect of Electric Fields on Metal Oxidation 139

effect was first observed above 100 A. Below this layer thickness neither
vacancy production nor a mass accumulation seems to be effective.
Evans l as well as Dankov and Churaev 2 deserve credit for the experi-
mental detection of internal stresses in oxide films growing on metals. A
metal film about 200 A thick was vaporized on a thin mica foil of about 15 f"
thickness and finally oxidized at an oxygen pressure of 10-12 mm Hg. It was
observed that in the case of an oxidized iron or nickel film, the mica foil
curved in such a way that the oxide film was located on the convex side,
while in the case of an oxidized magnesium film the oxide layer was observed
only on the concave side. These experimental results show that in the first
case the growing oxide produces a spatial extension while in the second case
the growing MgO causes a contraction. This can be explained on basis of the
volume ratios: for FeO 1.76, for NiO 1.65, and for MgO 0.81. The deformation
rate was always the greatest at the beginning. The presence of water vapor
increased the deformation effect in the oxidation of nickel and iron, but
decreased it in the case of magnesium oxidation.
On basis of the equations for the tension (j and the relative extension e

and

(where h is the thickness of the foil, g the thickness of the oxide layer, ro the
radius of curvature of the foil, and E or Eo the modulus of elasticity of the
foil or the oxide layer), Dankov and Churaev concluded, from the fact
that the value for ro obtained from the experiments was too low, that
there was a premature rupturing of the oxide film, so that the maximal
tension and extension did not show the effect expected from the above
expression. This kind of behavior is evidently responsible for the continuous
rupturing of the scaling layer in sulfide formation on nickel (see Chapter 5)
and in the oxidation of both titanium and titanium alloys and of zirconium
and zirconium alloys (see Section 4.2.2). Jaenicke3 has reinvestigated the
existence of mechanical stresses in CU20 layers during the oxidation of copper.
Pore formation and cavity formation have frequently been reported in
the scaling layer even at high temperatures. Baukloh and Thiel 4 and others
found extensive pore formation in the neighborhood of the metal/scaling
layer phase boundary in the oxidation of ARMCO-Fe in C02 and water vapor

1 Evans, U. R.: In8t. Metal8 Sympo8ium on Internal Stre88e8 in Metal8 and AlloY8, p. 291
(1947).
2 Dankov, P. D., and P. V. Churaev: Dokl. Akad. Nauk. SSSR 73, 1221 (1950).
3 Jaenicke, W., and S. Leistikow: Z. phY8. Chem. [NF] IS, 175 (1958).
4 Baukloh, W., and G. Thiel: Korro8ion u. Meta1l8chtz. 16, 121 (1940).
140 3. The Mechanism of Oxidation of Metals-Theory

above 1000°C (Fig. 50a). However, the metal surface below the pores
and blisters was covered with a thin adherent oxide film. Such cavity for-
mation was also observed by other authors 1 in the oxidation of ARMCO-Fe
and steels in air. Birchenall and co-workers 2 have studied oxidation of
iron producing thick scales at high temperatures in an oxygen atmosphere.
Large cavities are always found in specimens whose total oxygen content
has been brought to that of FeO (see Fig. 50b). The scale thickness of such
specimens, along with other evidence, implies that FeO is plastic in the
temperature range in which it is stable, while one or both of the higher oxides
is relatively rigid. Evidence is presented which indicates that iron is trans-
ported to its surrounding scale with little or no bulk diffusion path. The
overwhelming process seems to be transport of iron and oxygen via pores and
grain boundaries.3

COz or HzO

Fig. 50a. Schematic representation of blister forma-


tion at the phase boundary metal· scaling layer in the
oxidation of ARMCO·Fe in CO 2 and water vapor
above 1000°C, according to Baukloh and Thiel.

Fig. 50b. Box·shaped cavity from the


oxidation of iron at 850°C, according to
MacKenzie and Birchenall. The tube in
the center is caused by a hole drilled to
receive the platinum suspension wire.

It is evident that such cavity formation can be responsible for the


deviations from the parabolic law. Evans4 was the first to undertake an
explanation of nonparabolie rate laws caused by structure defects and cavity
1 Heindlhofer, K., and B. 1\1. Larsen: Trans. Am. Soc. Steel Treating 21, 865 (1933);
C. Upthegrove and D. W. Murphy: Trans. Am. Soc . Steel Treating 21,73 (1933); L. B.
Pfeil: J. Iron Steel Inst. 119,501 (1929); B. W. Dunnington, F. H. Beck, and M. G.
Fontana: Corrosion 8,2 (1952).
2 Juenker, D. 'V., R. A. Meussner, and C. E. Birchenall, Corrosion 14, 57 (1958); J. D.
MacKenzie and C. E. Birchenall, Corrosion 13, 783 (1957).
3 See also H. J. Engell: Z. Elektrochem. 63, 835 (1959).
4 Evans, U. R . : Trans. Electrochem. Soc. 91, 547 (1947).
3.5. Effect of Electric Fields on Metal Oxidation 141

formation in the scaling layer. In a subsequent work Evansl extended the


relationships, and described the logarithmic oxidation law for thick oxide
films. This cannot be explained in terms of the boundary layer theory of
oxidation. Following Evans' line of reasoning, we consider, with simplifying
assumptions, the effect of a vacancy current from the surface of the scale to
the metal. This vacancy current either can cause an accumulation of vacancies
(cavity formation) in the neighborhood of the metal/scale phase boundary
or can penetrate into the metal, where it can be captured by dislocations
or produce pores by sufficiently rapid diffusion of these vacancies further
into the interior of the metal to the defects which were already present
in the crystal, as Brasunas has shown in Fig. 49. Since cavity formation has
often been observed around the metal/scaling layer phase boundary, it shall
be treated here in some detail.
By the vacancy accumulation in the neighborhood or in the metal/scale
boundary, the metal surface was reduced to a fraction e
= qt!qo < 1. Here
qo is the metal surface before the beginning of the experiment and qt the
surface which is still free of pores and blisters after an oxidation time t. We
obtain for vacancy formation at the phase boundary or for the decrease of
the effective (i.e., pore-free) surface
(3.88a)
or
(3.88b)
where dm is the weight increase. Furthermore, if we designate the layer
thickness growth by dg/dt in a pore-free scaling layer then it follows that

a:t
dm
=
k0 e Cit
de = k 3 t-1/2 exp ( - k1 m) (3.89)

Integration and solution of (3.89) then yields, after combining the constants,
a logarithmic rate law of the form
(3.90)
Without doubt the simplifications introduced here are frequently not
realized; however, this does not reduce the value of the Evans presentation
in any way. The passage of dissociated oxygen through the cavities, which
was discussed by Dravnieks and McDonald 2 and recently by Maak and
Wagner:! for the oxidation of copper-berryllium alloys helps us to understand
the progress of oxidation even in the parts of the phase boundary with pores
1 Evans, U. R.: Rev. Pure and Appl. Chem. 5,1 (1955).
2 Dravnieks, A., and H. H. McDonald: J. Electrochem. Soc. 94, 139 (1948).
3 Maak, F., and C. Wagner: Werkstoffe u. Korro8wn 12, 273 (1961).
142 3. The Mechanism of Oxidation of Meta.1s-Theory

and blisters. Both diffusion of oxygen molecules through the pores and surface
diffusion in the pore or blister must be taken into account when considering
subsequent progress of the reaction. By the latter mechanism metal and
oxygen ions continually diffuse preferentially from the inner surface of the
blister to the side of the blister which lies closest to the metal surface so
that the blister is continually filled on the metal side and extended upward,
as is indicated schematically in Fig. 51. If this mechanism were to become

Oxygen Oxygen

a,

Fig. 51. Schematic representation of the "outward bubbling" of cavities from the
scaling layer at higher temperatures: (a) cavity formation in the neighborhood of
the metal/oxide phase boundary; (b) migration of the cavities to the surface
through diffusion of oxide molecules on the inner surface of the cavities and
grain boundaries.

effective one would observe an "outward bubbling" of blisters and pores from
the scaling layer. Since the surface diffusion upon which the bubble effect is
based is one order of magnitude more rapid than the lattice diffusion ,the
effect is always in evidence, especially at intermediate and lower tempera-
tures, as long as the scaling layer is not too thick and porous.
BirchenalJ1 has recently reported on the change of the parabolic and
linear rate law with time as well as a reciprocal exponential oxidation law on
basis of a pore-formation mechanism. The mathematical relationships, which
form a valuable foundation for further quantitative considerations in this
direction will not be presented here since the experimental material has not
yet provided adequate proof of these relationships. Wagner and co-workers
were able to describe the relationships quantitatively for the case of tungsten
oxidation (see 4.5.1).
Evans also reported on a further oxidation process with a logarithmic
rate law. Since normal lattice diffusion is very slow at lower temperatures,
only the grain boundaries and crystal defects are considered as diffusion
channels. As Kofstad and co-workers 2 could show by diffusion of niobium-95

1 Birchenall, C. E.: Metallurgical Rep. 1, Princeton Univ., Rep. Control Nr. OSR·TN-
54-286.
2 Kofstad, P., O. Kalvenes, P. B. Anderson, and G. Lunde, ASTrA Document No:
AD Contract AF 61(052)-460, TN No.3 (1961).
3.5. Effect of Electric Fields on Metal Oxidation 143

in sintered specimens of Nb 2 0s and in the high-temperature range of 800-


1200°C, the reaction is predominantly a grain-boundary diffusion. As a result
of such an oxidation one does not find any smooth phase boundaries between
the oxide and metal. On basis of stripping experiments on oxide films Evans
and Stockdale1 were able to detect a zone in a gently heated iron sample
which consisted of a conglomerate of oxide and metal. Vernon and co-workers 2
found an oxide film formation on a polished iron surface following the
parabolic rate law above 200°C. Furthermore, the weight of the dissolved
oxide film was in agreement with the weight increase determined during the
oxidation. Below 200°C, however, no agreement was found and a logarithmic
rate law prevailed.
Davies, Evans, and Agar3 repeated the experiment under somewhat
different conditions. They used an iron sheet with an unpolished surface
upon which the oxide had been reduced with hydrogen before the
oxidation experiment. This treatment yielded a rough surface, and oxidation
experiments which were carried out under these conditions yielded a transi-
tion at 300°C from the parabolic to the logarithmic rate law. On basis of
these results Evans concluded that there was a preferred oxidation at these
roughened places which he designated as "weak" places. This process appears
to extend preferentially along the surface and not toward the interior. During
the course of the oxidation, the number of places which were preferentially
attacked by oxygen always became smaller, and decreased exponentially.
Under consideration of this mechanism Evans found a further possibility for
obtaining a plausible explanation of the observed logarithmic rate law.
Additional experiments should give further information on the refined
mechanism of the destruction and eventual recreation of these "weak"
spots.
1 Evans, U. R., and J. Stockdale: J. Chern. Soc. (London) 2651 (1929).
2 Vernon, W. H. J., E. A. Calnan, O. J. B. Clews, and T. J. Nurse: Proc. Roy. Soc. (A)
216, 375 (1953).
3 Davies, D. E., U. R. Evans, and J. N. Agar: Proc. Roy. Soc. (A) 225, 443 (1954).
4. Scaling Processes in Metals and Alloys
with Formation of Thick Protective
Layers

We have been concerned until now with the fundamental phenomena


involved in oxidation processes in and on both thick and thin protective
layers. In the present chapter we will extend and apply the mechanisms
discussed earlier. We start with the oxidation processes which lead to
compact and thick oxide layers, which generally appear at high temperatures
after long oxidation intervals. At the present time, research is being pushed-
particularly studies of oxide systems with thick oxide layers as opposed to
those with thin tarnishing layers-so that a detailed discussion of the
mechanisms of these processes is especially useful. We can take Wagner's
scaling formulas (3.21) and (3.22) as our starting point. (We must also consider
the dependence of the self-diffusion coefficients on the chemical potential of
the metal or nonmetal.) Unlike (3.21), which is generally valid, (3.22) is
limited since it is only applicable to oxidation systems with electron con-
ducting protective layers.
Equation (3;21) lends itself to a classification of oxidation systems
according to their physical behavior, e.g., a consideration of conductivity
and transference numbers, rather than on basis of chemical designations,
e.g., oxide, halide, sulfide protective layers. While equation (3.21) expresses
the relationship between oxidation rate and the electrical parameters in
the scaling layer (conductivity and transference number) and their dependence
on the chemical potential or the partial pressures of the metal and nonmetal,
equation (3.22) shows the manner in which the rate of oxidation can be
determined from the self-diffusion coefficients of the ions that form the oxide
crystal. \Ve will in the following subdivide oxidation systems according to
whether they have ion- or electron-conducting protective layers. In the
former system, i.e., the case of ion-conducting protective layers-regardless
of whether the cations or anions take over the current transport through the
protective layer-the scaling rate will be determined by the electron partial
conductivity t3x = x 3, because of the fact that tl R::< 1 and t2 R::< 0 or
144
4. Scalinl1 Processes with Fonnation of Thick Protective Layers 145

t2 I=:::! 1 and tl I=:::! O. In the latter case, the cation or anion partial con-
ductivity, a function of the mobility of the cations or anions, represents the
rate-determining factor. An additional distinction between electron- and
hole-conducting protective layers must also be considered here as it is of
decisive significance in the oxidation of alloys and is a critical factor in the
choice of the correct self-diffusion coefficients for the crystal of the tarnishing
layer in atmospheres similar to ambients during oxidation. This will come
up again in a later section where we deal with individual scaling systems.
Before proceeding to a discussion of individual scaling systems, we will
set down a few useful expressions for the evaluation of oxidation experiments,
which of course will be formulated without taking account of phase-boundary
reactions. For this purpose the reader is reminded of the parabolic Tammann
tarnishing formula l given at the beginning of this book, which when inte-
grated reads as follows:
(,1 ~)2 = 2k' t (4.1)
where Llg is the thickness of the scaling layer, usually in centimeters, and t
is the oxidation time in seconds, hours, or days. We designate the constant
k', for example, in cm 2/sec, as the Tammann scaling constant. Pilling and
Bedworth 2 formulated the same rate law independently of Tammann. To
avoid confusion in the use of the equation, it may be noted once more that
(4.1) is only applicable in oxidation experiments which involve protective
layers more than 5000 A thick, where no scale-aging effects and no space-
charge layers or phase-boundary reactions noticeably affect the results.
Frequently, one must calculate the Tammann scaling constant k' from
the mtionalscaling constant k, using the units determined by the experimental
conditions and vice versa. This is done by the introduction of equivalent
quantities such as the equivalent volume ii, (equal to equivalent
weight/density) and the equivalent number n. These quantities are related
to the layer thickness as follows:

- L1~q (4.2)
n=-_-
v
Differentiation with respect to the time t yields
dn _ q d(L1;)
(ft - iJ- d t (4.3)

From (4.3) and the differentiation of (4.1) it follows that


dn q k'
(ft=-:;rrv (4.4)

1 Tammann, G.: Z. anorg. u. allgem. Chem. 111, 78 (1920).


2 Pilling, N. B., and R. E. Bedworth: J. Inst. Meta18 29, 529 (1923).
146 4. Scalin~ Processes with Formation of Thick Protective Layers

where q is the cross-section or the metal surface in square centimeters.


From this we obtain the conversion formula for the rational scaling
constant:

k'
k = ~ in equivalents/cm-sec (4.5)

Similar conversion formulas have already been used by other authors,l


who followed the layer thickness growth as a function of time using optical
measuring procedures. 2 Detailed and critical presentations of methods were
given by Evans,3 Masing,4 and Kubaschewski. 5 A critical analysis of
optical measuring procedures with a discussion on the theory of interference
colors is given by Winterbottom 6 in the appendix of the well-known book by
Evans. However, these optical methods are only suited to the measurement
of thin tarnishing layers, where transport processes in space-charge layers
are frequently rate-determining.
For measurements on thicker oxide layers, which we designate as scaling
layers to distinguish them from the thinner ones, the increase in layer thick-
ness is generally determined from the weight increase, LIm, in grams, for the
oxidizing sample, using a suitable balance. This is done, for example, with
an appropriate experimental arrangement using a quartz balance 7 ,8
or a magnetic balance. 8 ,9 The balance used by Gulbransen8 is especially
recommended for very sensitive measurements. (A brief presentation of
simple measuring procedures is given in Chapter 8.)
Analogous to (4.1), the practical scaling constant kIf may be obtained from
the mass increase per unit surface, Llm/g, in grams per square centimeter

1 Evans, D. R.: J. Soc. Chem. Ind. (Trans.) 45, 211 (1926); J. S. Dunn: Proc. Roy.
Soc. (A) 111, 210 (1926); F. H. Constable: Proc. Roy. Soc. (A) 115, 570 (1927); D. R.
Evans and L. C. Bannister: Proc. Roy. Soc. (A) 125, 370 (1929).
2 Tammann, G., and G. Siebel: Z. anorg. u. allgem. Chem.152, 149 (1926); F. H. Constable:
Proc. Roy. Soc. (A) 117, 376 (1928); 125,630 (1929); K. Fischbeck: Z. Elektrochem. 37,
593 (1931).
3 Evans, D. R.: Kolloid·Z. 69,129 (1934).
4 Masing, G.: Korrosion metallischer Werkstoffe, Leipzig, 1936, p. 97.
5 Kubaschewski, 0., and B. E. Hopkins: Oxidation of l'fIetals and Alloys, London, 1953,
pp.80ff·
6 \Vinterbottom, A. B., and D. R. Evans: 2\Jetallic Corrosion, Passivity and Protection,
London, 1948, pp. 802ff.
7 Wagner, C., and K. Grunewald: Z. physik. Chem. (B) 40, 455 (1938); K. Hauffe and
C. Genseh: Z. physik. Chem. 195, 116 (1950).
8 Gulbransen, E. A.: Rev. Sci. Ins/r. 15, 201 (1944).
9 See for example H. Dunwald and C. 'Vagner: Z. anorg. u. allgem. Chem. 199, 321
(1931); R. J. Maurer: J. Chem. Physics 13, 321 (1945); see also C. Duval: Inorganic
Thermogravimetric A nalysis, Amsterdam, 1953.
4. Scaling Processes with Formation of Thick Protective Layers 147

per unit time as

(4.6)

where k" is usually in units of g2/cm 4-sec or g2/cm4-hr. The weight increase
Llm is equal to the product of the equivalent number ii and equivalent
weight of the nonmetal X (equal to atomic weight Ax divided by the valence
IZ2j). If we solve (4.2) for Lle, we then obtain
Ll$ - nv - Llmvlz2! (4.7)
--q--q~

By inserting (4.7) into (4.1) we obtain


k' = ~ ( Ll m
t q
)2 .!.2 (IZ21..AxV)2 (4.8)
and from this the important conversion equations follow directly:

k' = _~ (IZ2! V)2k" (4.9)


2 .Ax

1.- = ~v (~)2 k" (4.lO)


2 .Ax

Equation (4.lO) can also be used for the determination of oxidation constants
from volumetric gas measurements.
Another method for the determination of the rational scaling constant k
is based on the use of the difference in conductivity of the metal and its
scaling layers. At the moment the last-remaining metallic phase disappears,
there is often a relatively rapid increase in resistance, which represents the
electrical conductivity of the scaling layer. This conductivity is generally
several orders of magnitude less than that of the metal. If we designate the
time between the beginning of the reaction and the resistance change by T
and the thickness of the metal sheet before the beginning of the experiment
by 0, and if we consider further that at the end of the reaction the quantity
o v
has grown in relation to the equivalent volumetric (scaling layer) and
VMe (metal), then the following relationship is obtained:

L1 $ = -21 v
J -...;;--
VMe

Using (4.1) and (4.5), we finally get


1 15 2 V
k=--- (4.11)
8 T vi.
This method is poorly suited for the determination of the course of the
oxidation with time at a slow scaling rate since the change in resistance is
not sharp. However, for rapidly oxidizing metal foils the last method is
148 4. Scaling Processes with Formation of Thick Protective Layers

especially convenient because one needs only to determine the time T, since
all other quantities occurring in (4.11) are known. 1
Following the above classification of scaling processes according to
Wagner's scaling expression, we will first deal with the few known oxidation
systems with ionically conducting protective layers.

4.1. Scaling Systems with Ion-Conducting Protective Layers


The bromination and chlorination of silver are examples which have
been thoroughly investigated. Wagner 2 studied the rates of bromination and
chlorination of silver foils (99.99% chemically pure) 0.1 mm thick, in the
temperature region between 200 and 400°0 at different halogen partial
pressures, using the gravimetric method. Earlier measurements in a more
empirical way have been carried out by Tammann and Koster3 as well as
Kohlschiitter and Krahenbiih1. 4 The sheets in these experiments were
heated from t to 2 hr in a chlorine or bromine stream or in a nitrogen stream
laden with chlorine or bromine at the given temperatures and finally weighed.
Hauffe and Gensch 5 improved the experimental technique by use of a quartz
spiral balance which permitted continuous observation of the mass increase.
As can be seen in Fig. 52 the course of halogenation is determined by a
parabolic rate law. The experimental results obtained by Wagner are
summarized in Tables 21 and 22. Equation (4.10) was used for the calculation
of the rational scaling constants k from the practical scaling constants k",
which were obtained directly from the experiments.
Transference measurements carried out by Tubandt and co-workers 6
show that in silver halide crystals at high temperatures, above 250°0,
practically only the silver ions take part in the current transport, which in
itself is sufficient to indicate a preferred silver ion lattice defect and mobility.
The lattice defect model formulated by FrenkeF (the same concentration
of silver ion vacancies and silver ions at interstitial lattice positions,
XAgO' = XAgO') was confirmed by Wagner and BeyerS on basis of measure-
ments of the density and the lattice constants of AgBr at 4lO00. Further,

1 Palmer, W. G.: Proc. Roy. Soc. (A) 103,444 (1923); H. Reinhold and H. Mohring:
Z. physik. Chern. (B) 28,178 (1935).
2 Wagner, C.: Z. physik. Chern. (B) 32, 447 (1936).
3 Tammann, G., and W. Koster: Z. anorg. u. allgem. Chern. 123, 196 (1923).
4 Kohlschutter, V., and E. Krahenbuhl: Z. Elektrochem. 29, 570 (1923).
5 Hauffe, K., and C. Gensch: Z. physik. Chern. 195, 116 (1950).
6 Tubandt, C., H. Reinhold, and W. Jost: Z. physik. Chern. (A) 129, 69 (1927); Z. anorg.
u. allgem. Chern. 177,253 (1928).
7 Frenkel, J.: Z. Physik 35, 652 (1926).
8 Wagner, C., and J. Beyer: Z. physik. Chern. (B) 32, 113 (1936).
4.1. Scalin~ Systems with Ion-Conductin~ Protective Layers 149

50 I--+---l----=-R--i--j--+----i

Fig. 52. Parabolic course of the weight


increase during the bromination of
silver and silver-cadmium alloys at
330°C and 170 mm Hg bromine partial
pressure, according to Hauffe and
Gensch (L1mjq in gjcm 2 ; the numbers
on the straight lines denote at. % Cd).

Koch and Wagner1 as well as Teltow 2 concluded from the variation in


electrical conductivity in AgCI-CdCI2 and AgBr-CdBr2 solid solutions with
increasing content of CdCh or CdBr2 that the silver ions in interstitial lattice
positions have a higher mobility.
For evaluation of the experimental results, equation (3.21) was trans-
formed by replacing the chemical potential /Lx by the halogen partial pressure,
for example, PBr. :
(4.12)

The expression (t1 + t2)tax may be simplified to tax = Xa, where xa is the
partial electron conductivity since t1 ~ 1, t2 ~ 0, and ta ~ 1. Although,
in general, the electron flow is three or more orders of magnitude greater
than the ion flow, the electron partial conductivity is the rate-determining
quantity in the thermal bromination of silver because of the special lattice
defect behavior. Since the partial conductivity of a charge carrier is given
by the product of its mobility and concentration, the concentration of
electrons or holes-in this case holes in AgBr-must be extraordinarily
small compared to the concentration of silver ion defects. In order to find
a relationship between the bromination rate and the bromine partial pressure,
1 Koch, E., and C. Wagner: Z. physik. Chern. (B) 38, 295 (1937).
2 Teltow, J.: Ann. Physik (6) 5, 71 (1949).
150 4. Scaling Processes with Formation of Thick Protective Layers

Table 2l. Scaling Rate of Silver in Chlorine, According to Wagner

k X 1010 [equivalents/em-sec] Relative values


k(PCl. = 0.04) k(PCl. = 0.17)
T,oC PCl, = 0.04 PCl. = 0.17 PCl, = 1 atm k(PCl. = 1.0) k(PCl. = 1.0)
300 0.19
350 0.16 0.34 0.82 0.20 0.41
400 0.35 0.70 2.06 0.17 0.34

Table 22. Scaling Rate of Silver in Bromine, According


to Wagner

k X 1010 [equivalents/em-sec]
k(PBr, = 0.09)
T,oC PBr. = 0.09 PBr, = 0.23 atm k(PBr, = 0.23)
200 0.23 0.38 0.61
250 0.53 0.91 0.58
300 0.96 1.78 0.54
350 1.21 2.32 0.52
400 1.15 2.27 0.50

we consider the AgBr crystal as a solid solution in which the defects act as
they would in an ideal dilute solution. The effect of bromine can now be
described by the following symbolic reaction equation:
~ Brkg) --- AgBr + Ag 0' + EB (4.13)
According to (4.13) equal numbers of vacancies and holes are produced.
However, since the mass action relationship (2.2) given earlier is always
valid, in addition to (4.13) the following reaction must also come into play:
~ Br~g) + Ag 0 . --- AgBr + EB (4.13a)
The ion lattice defects originally present are relatively more numerous and
their concentration is not noticeably changed by the effect of bromine;
the change in hole concentration (x(!) <{ XAgD') follows from the mass action
equation resulting from (4.13):
X()) = const pit;, (4.14)
where the quantity XAgD' is incorporated into the constant since it remains
relatively unchanged. By introduction of x(!)(PBr,=l) as a reference value,
we write (4.14) as
(4.14a)
4.1. Scaling Systems with Ion-Conducting Protective Layers 151

and corresponding to the partial conductivity of the electron holes needed


in (3.21)
p1/2
"'e - "'e (p Br, ~ 1) Br.
U _ OJ

(4.15)
After insertion of (4.12) and (4.15) into (3.31), it follows that

dnAgBr
dt
= ~{RT
LJ, F2 " $(PBr,-I)
_ (Vv§""
PBr,
_v9E)}
PBr, (4.16)

p~~, and p~~, are the corresponding bromine partial pressures at the AgBrJBr2
and AgBrJAg phase boundaries and F is the Faraday number. Since the
diffusion is rate-determining thermodynamic equilibrium prevails at both
phase boundaries and p~~I. is the decomposition pressure of AgBr. However,
since p~~. ~ P~:, is valid for the parenthetical expression in (4.16):
RT
k -- -~~Eil(Pnr =)
Ipl/2
Br (4.17)
F2 • •
(p:~, is designated for simplicity as PBr,).
The proportionality to be expected between the rational scaling constant
k and the square root of the bromine partial pressure given earlier was
experimentally confirmed by Wagner (see Table 22). Analogous equations
can be obtained for the chlorination of silver.
Since the partial conductivity of the holes in silver halides is determined
only with difficulty from other measurements, it is expedient to calculate
the partial conductivity "Eil from the experimentally obtained bromination
rate and the transference number of the electrons by division of "$ by the
total conductivity". At 400°0 and PBr, = 0.23 atm, "Eil in AgBr amounts
to only 3.8 x 10-4 compared to "ion = 3.8 X 10-1 ohm-1-cm-1 and we
obtain for tEil == ta = 0.001; at 200°0 under otherwise similar experimental
conditions we get the considerable value of 17%(ta = 0.17). By the addition
of OdBr2 the value of "Eil will of course be reduced again, since the AgO'
concentration is increased and that of the holes is reduced through the OdBr2
addition-as will be seen from the following arguments.
If the partial conductivity of the free electrons e or of the holes $ is
rate-determining for the total reaction in scaling systems with ionic con-
ducting protective layers, then with decreasing values, for example, of "$
in the case of the AgJAgBrJBr2 system, the scaling rate decreases. This
assumption is quite generally valid: if the phase-boundary reaction proceeds
rapidly enough, then the scaling rate of a metal or of an alloy changes when
the partial conductivity for the rate-determining step is changed. In scaling
systems with ionic conducting protective layers, therefore, one must observe
a. decrease in the scaling rate if one increases the ionic partia.l conductivity
152 4. Scalin~ Processes with Formation of TWck Protective Layers

and reduces the partial conductivity of the electrons. On the other hand,
one will increase or decrease the scaling rate in scaling systems with electron
conducting protective layers according to whether one increases or decreases
the ionic partial conductivity. Wagner and Hauffe,l were the first to take
note of these relationships.
Hit is assumed that the mobility of the ion lattice defects, free electrons,
and holes is practically independent of the defect concentration-and this is
often allowed in a first approximation-then changes in the partial conduc-
tivity of a defect-type charge carrier must be caused by changes in its con-
centration. Such regulating of the def~ct concentration can be realized by
the incorporation of foreign ions with different valences, as has been shown
by Wagner,2 Verwey,3 and Hauffe. 4 Accordingly, one must alloy the metal
with a small concentration of a second metal with a different valence-whose
ions will be incorporated into the scaling layer. This mixed crystal (solid
solution) is called a heterotype mixed phase. The solubility of the impurity
cation generally increases with increasing similarity of the cation radii.
In view of these facts, cadmium is a particularly suitable scale-reducing
alloy addition for the AgBr protective layer on silver. First of all, it has a
higher valence (divalent) than silver and, secondly, it possesses an ionic
radius close to that of the silver ion, if one may accept the "ionic radii"
given by Goldschmidt, which is permissible in the case of ionic crystals with
a predominant heteropolar character:
Ag+ : 1.13 A, Cd 2 +: 1.03 A, Pb 2+: 1.32 A, Zn 2+: 0.83 A
In agreement with the ionic radii, it can be seen from the solubility diagram
of a few metal bromides in AgBr according to Teltow5 (Fig. 53) that at
330°C, up to 25 mol. % CdBr2 dissolves, while the solubility strongly de-
creases to 8 mol. % for PbBr2 and to about 0.5 mol. % for ZnBr2. Thus,
introduction of Cd 2 + ions into the AgBr lattice during the bromination of
a silver-cadmium alloy leads to the greatest drop in bromination rate,
which decreases with increasing cadmium content, as has been shown by
bromination eXlJeriments in silver alloys at 330°C and a bromine partial
pressure of 170 mm Hg (Fig. 54).6
1 See for example K. Hauffe: Z. Metallk. 42, 34 (1951); "The Mechanism of Oxidation
of Metals and Alloys at High Temperatures," in Prog. in Metal Physics 4, 71 (1953).
2 Koch, E., and C. Wagner: Z. physik. Chem. (B) 38,295 (1937); C. Wagner: J. Chem.
Physics 18, 62 (1950).
3 Verwey, E. J. W., P. W. Haayman, and F. C. Romeyn: Chem. Weekblad 44,705 (1948).
4 Hauffe, K.: Ann. Physik. (6) 8, 201 (1950); "Fehlordnungserscheinungen und Leitungs.
vorgange in ionen· und elektronenleitenden festen Stoffen," in Ergeb. exakt. Naturw.
25, 193 (1951).
5 Teltow, J.: Ann. Physik (6) 5, 63,71 (1949).
6 Gcnsch, C., and K. Hauffe: Z. physik. Chern. 195, 386 (1950).
4.1. Scalin~ Systems with Ion-Conductin~ Protective Layers 153

~- I /~
V
lA9.~ Cd)Br
/
V
[7
lOll

Mo/%
Jf) Zf) Mi% JO Cd Brz '10 11 11 Mol % 10 Ca Brz 11 Zn Br2 .f

Fig. 53. Silver bromide side of the phase diagram AgBr-CdBr2, AgBr-PbBr2' AgBr-
CaBr2, and AgBr-ZnBr2' according to Teltow.

z
I

~.Ib I I

.
T-3.JO°C
\~ P8rz .170mm Hg
't,
"
'-. .......
~ f<Ag.P/i
. I
l-"·~
1>

'l - I
$!
Fig. 54. The bromination rate of ""
Ag-Cd, Ag-Pb, and Ag-Zn
alloys with increasing content
of the alloying partner at 330°C ~r--~ >---- --- Ag-ru t-~-c
and 170 mm Hg bromine vapor
pressure, according to Gensch 0 am o.oa aOJ aOli a05 aoo 0.07 aM
and Hauffe.

As discussed earlier, on basis of electroneutraJity, the introduction of a


divalent cadmium ion into a silver lattice position in a AgBr lattice causes
a Ag+ ion to be displaced from its lattice position because of its positive excess
charge, and a silver ion vacancy AgO' thus appears, which represents a
negative excess charge:
CdBr 2 ..---'" Cd.' (Ag) + Ag 0' + 2AgBr (4.18)
(see also Fig. 3). At constant bromine partial pressure and constant tempera-
ture, it follows from the mass-action law (4.13):
(PBr. = const) (4.19)
that there is a simultaneous decrease in the hole concentration X(j).
The relationship between scaling rate and electrical conductivity in the
alloy and the heterotype mixed phase on one hand and in the pure AgBr
154 4. Scaling Processes with Formation of Thick Protective Layers

on the other can be calculated by the combination of the mass-action


condition for the lattice defect equilibrium of the AgBr crystal,
(4.20)
and the electroneutrality relationship which is valid for the solid solutionl :

(4.21)
From (4.20). and (4.21) the concentration of silver ion vacancies with the
substitution Xed == XCd.·(Ag) is obtained as

X Ago '
1
= 2' Xed + {( XCd)2
2 + K }1/2 (4.22)

Since the concentration of the silver ion vacancies according to (4.19) is


inversely proportional to the hole concentration, the ratio of these quantities
in the solid solution and in pure AgBr (superscript 0) is

(4.23)

However, since the electron partial conductivity is directly proportional to


the hole concentration as well as to the scaling constant, according to (4.17),
k x,'!) X~gO'
J;O=-o-=
X0 XAgo'

we obtain the ratio of the scaling constants for the alloys k and the pure
silver kO by combination with (4.22) as

---+
XCd
2x~g
[ (-Xed)2
- +1 ] 1/2
2x~g
0' 0'
(4.24)

Figure 55 shows the satisfactory agreement 2 for the kjk O ratio between the
1.0 Fig. 55. The dependence of the ratio
I
of the scaling constants klk o on the
i i cadmium content of the alloy at

0.8 1\ !
I
I 330°C and at a bromine vapor pressure
of 170 mm Hg, according to Hauffe
\~~
1\\
I
i
and Gensch. Curve 1 was obtained

)( -
from experimental data; curve 2
I\·~,
"
I
represents the theoretical course cal-
culated according to (4.24). The tangent
obtained at XCd = 0 intersects the
tJ' '---J{=~ ~: abscissa at XCd = X~gD' = 2.9 X 10-3,
i
0.08 0.08 0.08 in agreement with the value obtained
by Teltow from conductivity measure-
ments.
1 Gensch, C., and K. Hauffe: Z. physik. Chem. 195,386 (1950).
2 Hauffe, K., and C. Gensch: Z. physik. Chem. 195, 116 (1950).
4.1. Scaling Systems with Ion-Conducting Protective Layers 155

measured values (Curve 1) and the values calculated according to (4.24)


(Curve 2) on basis of the value X~gO' = 2.9 X 10-3 , given by Teltow for
330°C. The deviations may be traced back to the fact that expression (4.24)
was derived assuming an ideal mass action law. This assumption is evidently
not always valid in defect concentrations greater than 0.5 mol. % because
of the appearance of Debye-Huckel effects.1
Corresponding to equation (4.17), we obtain the following expression
for the dependence of the bromination rate on bromine partial pressure
for higher alloyed Ag-Cd alloys, where

(-Xed
- -)2 ~1
2x~.o·
or

is a valid approximation:

(4.25)

In contrast to equation (4.17), here xG:l(PBr,~l) denotes the electron


partial conductivity of the AgBr-CdBr2 mixed phase at PBr, = 1 atm.
In addition it should be noted that in the present scaling system XCd (alloy
phase) = XCd (scaling layer = AgBr-CdBr2 solid solution), as shown in
the analysis. As investigations have indicated, this condition does not always
have an analog in other scaling systems. This will be discussed further later.
Relationships which are completely identical to (4.17), (4.24), and (4.25)
are also to be expected for the chlorination and iodination of silver and silver
alloys. Himmler2 tried to reduce the chlorination rate of silver by saturating
the chlorine atmosphere with PbCl 2 or CdC1 2. However, measurements
have shown that the effects obtained here are very small and in no case do
they approach the effect of a cadmium alloy addition. The two temperature
regions for the curves in Fig. 56, which represent the temperature dependence
of the bromination rate of silver and silver alloys with about 0.5 mol. %
cadmium, lead, and zinc, still cannot be explained at this time. The tempera-
ture independence of the halogenation rate of silver with iodine between
15 and 140°C is especially peculiar and unclear.

1 See especially the review article by J. Teltow: "Assoziation und Wechselwirkung


von Stiirstellen in Ionenkristallen und Halbleitern," in Halbleiterprobleme, Vol. 3, edited
by W. Schottky, Braunschweig, 1956, pp. 26-58.
2 Himmler, W.: Z. physik. Chem. 195, 129 (1950). See also J. Schatz, Werkstoffe u.
Korrosion 1, 248 (1950).
156 4. Scalin~ Processes with Formation of Thick Protective Layers

-O,----r----~---.----,_--_,----,

I -10

1-----'------1
"'" .. 4g (/1--/2&101)
~ -11 + Ag(p._aa>alJ I---+---l"-~-+---"l
• Ag/Zn Fig. 56. The temperature depen-
M Ag/ Pb dence of the bromination rate of
o Ag/ Cd
pure silver and silver alloys with
0.5 at. % Pb, 0.5 at. % Zn, and
-12t5 U; 1.7 1.IJ 1.0 2.0 2.1 0.6 at. % Cd at a bromine vapor
pressure of 170 mm Hg, according
1
r - to Gensch and Hauffe.

Evans and Bannisterl conducted corrosion experiments on silver in


different iodine-containing solvents, such as hexane, ether, and CCI 4 , between
o and 35°C, and observed that a parabolic rate law prevails, but in thick
layers this is caused by a predominant diffusion of iodine through pores
and along grain boundaries. Reinhold and Seidel 2 have also investigated
iodination of silver at high temperatures in the oc.-AgI phase region. The
experimental results are shown in Fig. 57. Jost and Weiss 3 have calculated
the transference number of electrons te- as a function of the iodine partial

-~7 r-~---~-----r----,

- &4~+-~~-~~----+----~

-4~ ~+----~1-----+~~--~

Fig. 57. Temperature dependence of


the iodination rate of silver at PI. =
10 mm Hg, according to Reinhold and
- ~~1.~
.•~---~f.4'--I--~~~O~---~
~C
Seidel. 0 are experimental and + are
calculated values of the rational
-P -- scaling constantk in equivalents/em.soo.

1 Evans, U. R., and L. C. Bannister: Proc. Roy. Soc'. (A) 125, 370 (1929).
2 Reinhold, H ., and H . Seidel: Z. Elektrochem. 41, 499 (1935).
3 Jost, W., and K. Weiss: Z. physik. Chem. [NF] 2, 112 (1954).
4.1. Scaling Systems with Ion-Conducting Protective Layers 157

pressure from the tarnishing constants in the formation of f3-AgI and find
for 140°C:
te- = 0.02 for PI. = 1.5 mm Hg
te- = 0.18 for PI. = 198 mm Hg

in good agreement with the observed value of te- ~ 0.14. This result is
comparable with that obtained by Wagner for AgBr.
If ka. and kfl are the rational scaling constants in the iodination of silver
in the ex and f3 regions of AgI, then, according to Wagner's theory we obtain

if to a first approximation the electromotive force of the cell AgIAgIII2(gas)


corresponding to the formation energy of both phases is about equal. If for
a temperature of 140°C and an iodine partial pressure of lO mm Hg one sets
t e- ~ 0.03 and considers this value applicable even at the transformation
point, then for lea. = 1.6 x lO-12 and kfl = 6 x lO-13, one obtains for the elec-
tron transference number in ex-AgI at 179°C and PI, = 10 mmHg, t~ ~ lO-5.
For pure ex-AgI at 179°C and PI, = 3 mm Hg, a value of te- ~ lO-7 has been
estimated.
The lead halide protective layers formed in the halogenation of lead also
show practically pure ionic conduction. As transference measurements by
Tubandtl have shown, the current transport in PbCl 2 and PbBr2 at higher
temperatures is predominantly due to halogen ions, while in PbI 2 approxi-
mately equal contributions by the lead and iodine ions to the current trans-
port were observed. Under the assumption of predominant intrinsic lattice
defects according to the Schottky model, we have the following:
Null~PbO" + 2CIO'
or with association 2 :
Null ~ (PbDCID), + CIO'
Therefore, the halogenation rate can perhaps be decreased in the presence
of oxygen or of PbO on lead. As a result of the introduction of oxygen ions
into the lead halide lattice, for example,
PbO = O.'(CI) + CIO' + PbCh (4.26)
or
PbO + (PbDCIO), = O.'(Cl)
1 Tubandt, C.: Leitfahigkeit und trberfiihrungszahlen in festen Elektrolyten, Hab.
Exp. Physik XII, I, Leipzig, 1932, p. 38!.
2 Simkovich, G.: J. Phys. Chem. Solids 24,213 (1963).
158 4. Scaling Processes with Formation of Thick Protective Layers

the number of halogen ion vacancies increases and therefore the hole con-
centration, which determines the halogenation rate, decreases. Experiments
with this system from these points of view have not been undertaken.
Dravnieks and McDonaldl were able to demonstrate a predominant migra-
tion of halogen ions through the PbCl 2 or PbBr2 protective layer in the
direction of the metal using "marker" experiments.

4.2. Scaling Systems with Electron..[ondueting


Protective Layers
If current transport in the scaling layer is almost exclusively due to
free electrons and holes, then the dependence of the scaling rate on the partial
pressure of the oxidizing gases and on the addition of ions of other valences
is dependent to a large extent on. whether free electrons or holes are available
for current transport in the scaling layer lattice. In this case t3 ~ 1 in the
Wagner scaling equation (3.21), so that when t2 ~ 0 the partial conductivity
of the cations (h" = "1) is the rate-determining factor for the oxidation,
and when h ~ 0 the partial conductivity of the anions (t2" = "2) is the rate-
determining factor. If, to a first approximation, one regards the mobility of
the defects as constant at constant temperature and independent of the
concentration of the defects, then the ion partial conductivity and thus the
scaling rate are direct functions of the ion lattice defect concentration.
While in scaling systems with p-type conducting protective layers the
oxidation rate increases with increasing partial pressure of the nonmetal
in the ambient atmosphere, the oxidation rate in scaling systems with
n-type conducting protective layers is independent of the external nonmetal
partial pressure, as will be shown in detail. Further, it may generally be
established that alloying additions whose cations are of higher valence than
the cations of the host crystal cause the oxidation rate to increase in the
case of a p-type oxide layer and to decrease in the case of an n-type layer.
The opposite influence on the oxidation rate is to be expected for an alloying
metal with lower-valent cations.
We will subdivide the oxidation of metals and alloys as far as possible
from this point of view. A strict division will of course not always be possible,
since additional phenomena which ff~qu~re separate treatment can appear,
and furthermore, necessary data are still lacking on many scaling systems,
which makes grouping difficult. We begin with the treatment of scaling
systems which form a p-type conducting protective layer.

1 Dravnieks, A., and H. J. McDonald: J. Electrochem. Soc. 93,177 (1948).


4.2. Scalinl1 Systems with Electron-Conductinl1 Protective Layers 159

4.2.1. Scaling Systems with p- Type Conducting


Protective Layers
The oxidation of copper and nickel and their alloys, which was in-
vestigated in detail, and the oxidation of cobalt and chromium, are discussed
in this section. The oxidation of manganese and iron, whose lower-valent
oxides MnO and FeO are pronounced p-type conductors, will not be dealt
with until later, since at higher oxygen pressures complicated protective
layers are formed which consist of several oxides. Furthermore, in the case
of oxidation of iron to FeO in CO 2-CO atmospheres above 900°C, phase-
boundary reactions determine the rate of oxidation. Investigations on
exclusive formation of MnO protective layers have not been reported.

4.2.1.1. The Rate of Oxidation of Copper and Copper Alloys


The first detailed investigations of the rate of oxidation of copper
sheets and wires in oxygen between 400 and 1000°C and in air between 800
and 1000°C were made by Pilling and Bedworth. 1 While above 800°C the
oxidation may be described by a parabolic rate law (diffusion processes are
rate-determining), considerable complications appear, especially at lower
temperatures. A few of these difficulties have been noted earlier. Thus, for
example, after 25 min of parabolic oxidation at 500°C an increase in the
oxidation rate appeared to be due to rupturing of the oxide film. Hudson
and co-workers 2 found that the oxidation of arsenic-containing copper
(0.47 wt.%), between 300 and 600°C can be represented by an undistorted
parabolic law for 6 hr. All of the copper samples which were already oxidized
had a thin oxide film from having been stored in air. Such oxide films will
also be present on other metals unless they are removed before the experi-
ment. Valensi3 investigated the influence on oxidation of these oxide films.
The mechanism of the low-temperature oxidation of copper which has already
been discussed (Sections 3.5.3 and 3.5.4) leads to a reciprocal logarithmic
law and to a cubic rate law. Based on structural investigations and studies
of the physical characteristics of the oxide layers formed on copper under the
most varied experimental conditions, Tylecote 4 showed that below 600°C
the oxide film is not expected to have sufficient "ductility," which is an
indispensable prerequisite for adherent, noncracking, protective layers.
At 1000°C, because of the greater ductility, a sufficiently adherent and
compact Cu2 0 layer will always be produced, provided the oxygen pressure
is kept below 70 mm Hg so that the oxidation remains in the Cu 2 0-phase
1 Pilling, N. B., and R. E. Bedworth: J. Inst. Metals 29, 529 (1923).
2 Hudson, D. F., T. M. Herbert, F. E. Ball, and E. H. Bucknall: J. Inst. Metal842, 221
(1929).
3 Valensi, G.: Proc. Pitt8burgh Intern. Conf. on Surface Reactions, 1948, p. 156.
4 Tylecote, R. F.: J. Inst. Metals 78,301 (1950/51).
160 4. Scaling Processes with Formation of Thick Protective Layers

region. Such measurements were carried out by Wagner and Grunewald. 1


According to the lattice defect model of Cu 20 a predominant migration of
copper ions and electrons via vacancies and holes, respectively, is to be
expected as a result of the formation of Cu+ vacancies, CuD'; and holes,
EB, (:; Cu++, see Fig. 58) when a chemical potential gradient of copper or

Cu" c"" Cu' Cu+ Cu'


0= 0= 0= 0=
Cu· 0 Cu' Cu" Co
0= 0= 0= 0= Fig. 58. Schematic representation of the lattice defect
arrangement in CU20, according to Wagner. (Holes are
Cu" Cu· Cu' Cu' 0 represented in CU20 by divalent copper ions.)

oxygen exists in Cu 20. At higher temperatures this migration or diffusion


will be the rate-determining step, since after attaining a certain layer thick-
ness the phase-boundary reactions proceed sufficiently fast and field transport
processes become insignificant. As Fig. 59 shows, the lattice defect concentra-
tion at the Cu 20jCu phase boundary in the case of thick Cu 20 layers is small

Position of the Cu CUeD Ck


Phase Boundaries
Phase EqUIlIbrium I
I I CuICUsO I
CUeD De Fig. 59. Scaling scheme for the
oxidation of copper to CU20,
according to Wagner. While the
CUZO+02 ((Jos) J'-- ----- - ----------------- concentration of the Cu+ vacancies
Xcuo' at the CU20/CU phase bound-

Cu-Deficit
of the CU20-Phase
I ary is negligible, it assumes con-
siderable values at the CU20/02
phase boundary (e.g., at 1000°C and
Po. = 33 mm Hg, xCuO' = 1.14 X
Phase Equilibrium L 10-3 , according to Wagner and
Cu(l1elul),Cu.O J Hammen). This governs the varia-
Position Coordinate-- tion of holes in the oxide layer.

compared to that at the CU20j02 phase boundary and therefore is to be


neglected. Thus, only the change in the lattice defect concentration with
varying oxygen pressure at the external phase boundary is rate-determining
for the oxidation. The copper vacancy concentration in CU20, which is in
equilibrium with oxygen, can assume considerable values here, as has been
shown by analytical determinations of the excess oxygen content or the
copper deficit. Thus Wagner and Hammen 2 have found, for example, at
1000°C and Po, = 33 mm Hg that XCuO' = X® = 2.2 x 10- 3 • With the
symbolic lattice defect equation,

! O~)""---CU20 + 2CuO'+ 2EB (4.27)

1 Wagner, C., and K. Grunewald: Z. physik. Chern. (B) 40, 455 (1938).
2 Wagner, C., and H. Hammen: Z. physik. Chern. (13) 40, 197 (1938).
4.2. Scaling Systems with Electron-Conducting Protective Layers 161

with the mass action law,


2 2 _ K 1/2
XCUD' X® - Po.
and taking account of
XCuD' = X(j)

the oxygen pressure dependence of the defect concentration is obtained as


XCUD' = Xe = K P~~ (4.28)
A dependence of both the scaling constant, which is proportional to the
vacancy concentration, and the electrical conductivity, which is proportional
to the hole concentration, on the oxygen pressure can be expected from the
eighth root of the oxygen partial pressure. Wagner and co-workers found,
in good agreement with the theory, that both the electrical conductivityl
(Fig. 60) and the oxidation rate 2 (Fig" 61) are proportional to the seventh
root of the oxygen pressure.
a.
J~!--v:
+ :

.... --::;:::. ~k-


~
V- :-
f-"-
,/ ,,/~
<-;:::::; :::---
,//'
/// ,/ '~7a/
/ ,/'

-a. 1--.&// //

<,
//

-a!
/ /
/ /
- 1.0 1-.
~

"
9""-
-.J -if -1 {} m.m,Hg 1
10

Fig. 60. Dependence of the electrical conductivity of a Cu 20


foil on oxygen pressure between 700 and 900°C, according to
Gundermann, Hauffe, and Wagner. EEl-Foil I; x-Foil II;
+-FoilIII; x ~ ptj7. (At lower oxygen pressures considerable
deviations appear b'elow 800 0 e which have not yet been
explained.)
If in place of the oxygen chemical potential in (3.21), we introduce the
corresponding oxygen pressure
dILx = tdILo , = t RTd In Po ,
1 Dunwald, H., and C. Wagner: Z. physik. Chern. (B) 22, 212 (1933); J. Gundermann,
K. Hauffe, and e. Wagner: Z. physik. Chern. (B) 37,148 (1937); R. S. Toth, R. Kilkson,
and D. Trivich: Phys. Rev. 122,482 (1961); J. Appl. Phys. 31, 1117 (1960); see also
J. Bloem: Philips Res. Rept. 13, 167 (1958).
2 Wagner, C., and K. Grunewald: Z. physik. Chern. (B) 40, 455 (1938).
162 4. Scaling Processes with Formation of Thick Protective Layers

and if we consider the oxygen pressure dependence of the total conductivity


~ _ " pl/7 (4.29)
- (PO.=l) 0,

and introduce the transference numbers of the copper ions tcu+ :::::: 5 x lO-
at lOOO°C, which was determined by Gundermann and Wagner,! the following
rational scaling constant in the oxidation copper is obtained:

(4.30)

where P~~ is the oxygen equilibrium pressure over Cu 2 0 + Cu. The validity
of relationship (4.30) was confirmed by Wagner and Grunewald 2 on basis
of oxidation experiments at lOOO°C and oxygen pressures between 3.0 x lO-4
and 8.3 x 10- 2 atm (see Fig. 61). Furthermore, it can be seen in Table 23

uiv./cm-sec /
I ~,
/
.. /
o
. / Fig. 61. Rational scaling constant k for the oxidation
of copper to CU20 at 1000°C as a function of the seventh
az 0.0 atmo.a root of the oxygen pressure, according to Wagner and
Grunewald. (The intercept of the extrapolated straight
Equilibrium line with the abscissa gives the oxygen equilibrium
Cu+ Cu2 0 pressure over CujCu20.)

that the scaling constants calculated from (4.30) agree well with those which
were obtained experimentally.
As can be seen from Fig. 62, the parabolic rate law is not strictly obeyed
at the beginning of the copper oxidation, and this is understandable if one
considers that in the first stage of the reaction a predominant dissolution
of oxygen in copper and finally a CU20 nucleation take place. This mechanism
and the crystal growth associated with it which forms a compact Cu 2 0 layer
are probably responsible for the initial course of the oxidation, while the
reaction for the CU20 formation and the transfer of ions from the metal to
the scaling layer proceeds sufficiently rapidly.
If the oxygen equilibrium pressure (Cu20/CUO) is exceeded, a thin
CuO layer forms on the CU20 layer. This causes the oxidation rate to become
approximately independent of the oxygen pressure, which was also observed

1 Gundermann, J., and C. Wagner: Z. physik. Chern. (B) 37, 155 (1937).
2 Wagner, C., and K. Grunewald: Z. physik. Chern. (B) 40, 455 (1938).
f'-
~
r:Jl
e.e-
(Jo.

~
~
Table 23. Oxidation of Copper to Cu 2 0 at 1000°C, According to 'Wagner and Grunewald a
fIJ

~
....
Copper deficiency of the CU20 Rational scaling constant k,
Concentration Rate constant I e-
phase, g-atomsjO.5 mole CU20 difference at phase equiv.jcm-sec
of the phase- t'l
CU20 in CU20 in boundary boundary ;'
Po., equilibrium with equilibrium with equilibria reaction.
atm Cu, y(O 02, y(a) y(a)_y(i) obs calc equiv.jcm2~sec

x 10~} 0.40 X 10- 3 0.28 X 10- 3 2.0 X 10- 9 2.1 X 10- 9 0.6 X 10- 6
3.0
2.25
1.45
X 10- 3
X 10- 2
0.12 X 10- 3 0.62
0.91
X
X
10- 3
10- 3
0.50
0.79
X
X
10- 3
10- 3
3.1
4.5
X
X
10- 9
10- 9
3.4
4.8
X
X
10- 9
10- 9
1.4
2.1
X
X
10- 6
10- 6
l
g
go
/')
8.3 X 10- 2 1.29 X 10- 3 1.17 X 10- 3 6.2 X 10- 9 6.6 X 10- 9 3.4 X 10- 6
~
(Jo.

"C

.,
Ii
fIJ

...
g;
164 4. Scaling Processes with Formation of Thick Protective Layers

sec ·lcrnd f/
/

It ! ~

.
5· .v
~
",/
!
, . /'
111 ./
/•
/ /
/ ~
vj
./~
10 5
/!'
.
3

.
0// /'"
1\
2· 105 i'"
~/ •/ ' r"~ Fig. 62. Representation of the experi-
....~4'""'/
///,..A _.f'" ,/ mental results by use of the relationship
tj(L1mjq) = 1jk"(L1mjq) + 1jl" (l" is the
1· 1v rate constant of the phase boundary
~r' reaction in gjcm 2-sec) in the detection of
phase boundary reactions at the onset
o of oxidation, according to Wagner and
510
. -J
10 to -.I g/cm8 15tO-J Grunewald. 1: Po. = 3.0 X 10-4 ;
dm_ 2: Po. = 2.25 x 10- 3 ; 3: Po. = 1.45 X
rr 10- 2 ; 4: Po. = 8.3 X 10- 2 atm.

in the investigations by Pilling and Bedworth,l Feitknecht,2 and Frohlich. 3


Even at relatively high oxygen pressures, up to 30 atm, the oxidation rate
of copper is independent of the oxygen pressure, as is shown in the results
obtained by McKewan and Fassel1. 4 A few of these are summarized in Table
24.

Table 24. Oxygen Pressure Independence of the


Oxidation Rate of Copper at Higher Oxygen
Pressure, According to McKewan and Fassell

P02' Average k",


atm g2jcm4-sec

600 1 3.3 X 10- 10


20.4 3.2 x 10-10
800 1 7.6 X 10- 9
3.4 7.8 x 10- 9
6.8 7.8 x 10- 9
13.6 7.7 x 10- 9
27.2 7.9 x 10- 9
900 1 3.0 X 10- 8
20.4 2.8 x 10- 8

1 Pilling, N. B., and R. E. Bedworth: J. Inst. llfetals 29, 529 (1923).


2 Feitknecht, W.: Z. Elektrochem. 35, 142 (1929).
3 Frohlich, K. W.: Z. ",fetaUk. 28,368 (1936).
4 McKewan, W., and iV. 1\1. Fassell, Jr.: J. Metals 5,1127 (1953).
4.2. Scaling Systems with Electron-Conducting Protective Layers 165

Shockley et aU and Moore et al. 2 studied the oxidation of copper by use


of the radioactive copper isotope, copper-64, and determined the diffusion
rate of the copper ions through the'Cu 20 layer in the temperature region be-
tween 800 and lOOO°C as Dcu+ = 0.0358. exp( -37,000/RT) cm 2/sec. The
activation energy found here agrees quite well with that for copper oxidation
of 38,000 ± 2000 cal/mole CU20-with the parabolic rate law-and can be
regarded as further evidence for a preferred copper ion diffusion through the
scaling layer. (The activation energy in the cubic rate law is only
28,000 cal/mole: kc = 0.025.exp(-28,300/RT) g3/cm6-hr).3 As was men-
tioned in Section 3.3, and had been indicated earlier by Wagner,4 a quanti-
tative calculation of the scaling constants k or k' from the self-diffusion
coefficients obtained on basis of these data is difficult because here D~~)­
the self-diffusion of copper ions in Cu 20 which is in equilibrium with the
oxygen atmosphere prevailing during the oxidation-was not measured;
rather D~u in Cu 2 0 with a concentration gradient of copper ion vacancies
was determined. However, the D~~) values, which were later determined
by Moore and Selikson,5 are required for quantitative evaluation. Correspond-
ing to (3.25), we obtain, with ZCu+ = 1 and division by ceq(k/ceq = k'), the
relationship which was used by Moore and Selikson:

(4.31)

A comparison of the k' and D~~) values in Table 25 shows that with the
above assumption (4.31) is obeyed quite well.

Table 25. Diffusion and Oxidation Rate Constants in


Copper Oxidation at 0.1 mm Hg Oxygen, According
to Moore and Selikson

k', D*(a)
cu '
T,cC cm 2 /sec cm 2 /sec k'/D~~a)

800 2.1 X 10- 9 1.9 X 10- 9 (1.1)?


850 8.4 x 10- 9 4.0 X 10- 9 2.1
900 1.2 x 10- 8 7.7 X 10- 9 1.6
950 3.7 x 10- 8 1.4 X 10-8 2.6
1000 5.8 x 10- 8 3.2 X 10- 8 1.8

1 Bardeen, J., W. H. Brattain, and W. Shockley: J. Ohem. PhY8. 14, 714 (1946).
2 Castellan, G. W., and W. J. Moore: J. Ohem. PhY8. 17,41 (1949).
3 Tylecote, R. F.: J. Inst. Metal8 81, 681 (1952/53).
4 Wagner, C.: "Diffusion and High Temperature Oxidation of Metals," in Atom Move-
ment8, ASM, Cleveland, 1951, p. 153.
5 Moore, W. J., and B. Selikson: J. Ohem. PhY8. 19, 1539 (1951); 20, 927 (1952).
166 4. Scaling Processes with Formation of Thick Protective Layers

As was discussed earlier, a cubic rate law prevails in the intermediate-


temperature region between 200 and 700°C in at least the initial reaction
period of about 5-10 hr. Considerable complications appear because of the
physical characteristics and the structure of the protective layers, as Tylecote 1
was able to show on basis of his extensive investigations. The higher oxida-
tion rate observed for heated copper compared to hard-rolled copper can be
explained if one considers that the (110) planes 2 appearing in the hard-rolled
copper oxidizil more slowly than that (100) planes, which predominate after
heating (by about a factor of two). The influence of crystal orientation on
the oxidation of metals was extensively investigated by Gwathmey and
co-workers. 3 In the case of the oxidation of copper careful measurements
of the oxidation rate of electropolished {100}, {Ill}, {1I0}, and {31I} faces
of copper at 178°C were carried out by Gwathmey 4 and Rhodin 5 with the
aid of elliptically polarized light and a vacuum microbalance (Fig. 63).
The ratio of the thickness of the oxide on the {100} face to that on the {31I}

1000
(100)
o~
;§. 800 ~-----l,.L----t-----+--i
~
c::
ti 600
~

u: 100
t
200
(110)
(31 I)
Fig. 63. Oxidation of four faces
of a copper single crystal at
a 10 20 30 40 50 60 178°C, according to Gwathmey
->- Time (in min) and co-workers.

1 Tylecote, R. F.: J. Inst. Metals 81,681 (1952/53).


2 Barrett, C. S.: The Structure of Metals, New York, 1943, pp. 407 and 426.
3 Gwathmey, A. T., and K. R. Lawless: "The Influence of Crystal Orientation on the
Oxidation of Metals," in The Surface Chemistry of Metals and Semiconductors, ed. by
Harry C. Gatos, John Wiley & Sons, New York, 1960, p. 483.ff; F. W. Young, Jr., J. V.
Cathcart, and A. T. Gwathmey, Acta j'flet. 4, 145 (1956).
4 Gwathmey, A. T., and K. R. Lawless: "The Influence of Crystal Orientation on the
Oxidation of Metals," in The Surface Chemistry of Aleta Is and Semiconductors, ed. by
Harry C. Gatos, John Wiley & Sons, New York, 1960, p. 483.ff; F. W. Young, Jr.,
J. V. Cathcart, and A. T. Gwathmey, Acta ll!Iet. 4,145 (1956); see also J. Bougnut and
N. Nifontoff: Compt. rend. acado sci. Paris 248, 1683 (1959).
5 Rhodin, T. N., in Advances in Catalysis, cd. by W. G. Frankenburg, V. 1. Komarewsky,
and E. K. Rideal, Academic Press, New York, 1953, Vol. 5, p. 39.
4.2. Scaling Systems with Electron-Conducting Protective Layers 167

face at the end of 10 min was approximately 12. Gremlund and Benard!
found that under the same experimental conditions, the extent and the
size of the CU20 nucleus increases from the {loo} plane over the {1l0} to
the {Ill} and {311} planes. Not only the nucleus growth, but also the oxide
layer growth in a later stage of the oxidation, is dependent on the crystallo-
graphic plane.
Single crystals of copper oxidized in air of 1 mm Hg at 300°C exhibit
oxide films with alternating interference colors according to Menzel and
St6sse1. 2 At 500°C very small nuclei are formed in an atmosphere of
5 x 10-3 mm Hg air at the very beginning of oxidation, but in a few
minutes they grow to larger crystals with a defined structure. Obviously,
between these crystals only a monomolecular coverage of oxygen is present,
since the rate of crystal growth is faster than the rate of nucleus formation.
Similar phenomena have been observed during the oxidation of single-
crystal spheres of nicke1. 3 It might be expected that the differences among
the rates of oxidation on different faces would decrease at the higher temper-
atures. Actually Benard and Talbot4 reported that the rate of oxidation
of copper differed at 900°C with face by only a few percent.
It was further observed that, especially at intermediate temperatures
blister formation was pronounced in the oxide layer, which is evidently
responsible for the "sealing off" of the oxide layer from the copper. This
blister formation, which frequently leads to scaling of the oxide layer, is
caused by the high mechanical stresses existing during oxidation. Dankov
and Churaev 5 were able to demonstrate internal stresses of the order of
magnitude of 2-4 tons/cm 2. The high mechanical strains imposed on the
oxide film cannot be relieved to any great extent until above 600-700°C,
when the oxide becomes sufficiently plastic from the change of the ionic
distance in the scaling layer. There are extensive experimental data on the
rate of oxidation of copper alloys as a function of temperature and gas
atmosphere which were critically examined and supplemented with several
measurements by Tylecote. While larger quantities of nickel in copper
decrease the oxidation rate, the alloys containing only 0.1 wt. % nickel oxidize
significantly faster than pure copper. 6 The mechanism is rather complicated.
In the following discussion we will disregard low-alloyed coppers since they
1 Gr0nlund, F., and J. Benard: Compt. rend. 240, 624 (1955); F. Gr0nlund: J. chim. phys.
53, 660 (1956).
2 Menzel, E., and W. Stossel: 41, 302 (1954).
3 Menzel, E., and M. Otter: Naturw. 46, 66 (1959); M. Otter: Z. Naturforsch. 14b,
355 (1959).
4 Benard, J., and J. Talbot: Compt. rend. acado sci. Paris 225,411 (1948); J. Morean and
J. Benard: Compt. rend. acado sci. Paris 248, 1658 (1959).
5 Dankov, D., and P. V. Churaev: Dokl. Aka4. Nauk SSSR 73, 1221 (1950).
6 Bouillon, F. S., and J. Stevens: Acta Met. 7, 774 (1959).
168 4. Scaling Processes with Fonnation of Thick Protective Layers

involve a new oxidation phenomenon, internal oxidation, which shall be


treated later, separately. The copper-zinc, copper-beryllium, and copper-
palladium alloys will not be discussed here, since the diffusion rate in the
alloy phase for these scaling systems must be considered with certain
assumptions, as will be discussed later in detail in support of Wagner's
explanation (see Section 4.7).
From a consideration of the defect model of CU20, and by analogy to the
case of the halogenation of silver-cadmium alloys discussed earlier, we see
that a lowering of the CuD' concentration-and thus of the oxidation rate
of Cu to Cu 20-is not possible since it would require the introduction of
lower-valent cations, which is not feasible. Therefore it is understandable
if one prefers to use alloying partners which are considerably less noble
than copper and which under certain experimental conditions preferentially
"oxidize out" with formation of a foreign metal oxide layer which is often
pore-free and adherent. Such selective oxidation takes place with copper-
beryllium alloys.l,2 As Hubrecht 3 was able to show, a BeO film which protects
against further oxidation appears in copper-beryllium (2%) alloys only
when the oxidation temperature chosen is above 600°C and the oxygen
partial pressure is less than 0.02 mm Hg. A BeO layer produced in this way
hinders any significant oxidation in the temperature region from 25 to
500°C. Copper-beryllium alloys not treated in this way oxidized considerably,
for example, between 400 and 500°C, where the largest part of the scaling
layer consisted of CuO with only a small amount of BeO. The observations
of Hessenbruch4 are similar in that he could increase the working temperature
from 600 to 800°C for constantan ('" 45% Ni + 55% Cu) by the addition
of 1 % Be + 2% Si or 1% Be + 1 % Si. We shall discuss the influence of
beryllium in the oxidation of copper more extensively later.
Aluminum as an alloying partner was preferentially oxidized under
certain experimental conditions, resulting in an Al 20 3 layer which strongly
retarded the oxidation rate. Hall owes and Voce 5 succeeded in obtaining-
by selective oxidation of Cu + 5% AI-protection against atmospheric
oxidation up to 800°C, which, however, was not effective in atmospheres
containing S02 and HCI. On basis of oxidation experiments with copper-
aluminum alloys with aluminum contents of 0.2 to 20 wt. % at different
temperatures Spinedi 6 found the greatest scaling stability with an aluminum
1 Frohlich, K. W.: Z. Metallk. 28, 368 (1936); L. E. Price, and G. J. Thomas, Metal
Ind. (London) 54, 189 (1939).
2l\Iaak, F.: Z. Metallk. 52, 538 (1961).
3 de Br6uckere, L., and L. Hubrccht: Bull. soc. chill!. Belges 61,101 (1952).
4 Hesscnbruch, \V.: l1Ietalle und Legierungen fiir Iwlle Temperaturen, Springcr, Berlin,
1940.
5 Hallowes, A. P., and E. Voce: .Metallurgia 34,95 (1946).
6 Spinedi, P.: Met. ital. 45, 457 (1953).
4.2. Scaling Systems with Electron-Conducting Protective Layers 169

content of about 9%. As already indicated above, temperature plays a


decisive role in determining the structure of the scaling layer. Thus, for
example, at lower oxidation temperatures the structure of the scaling layer
of an aluminum-bronze (7% Al + 4% Mn + remainder Cu) Was completely
different than at higher temperatures, as was shown by electron diffraction
patterns.! While a predominant CU20 formation was observed in oxygen at
183°C, the scaling layer consisted almost exclusively of Mn02 after oxidation
at 400°C. No crystallized Ah03 was found in the scale, in agreement with
the observations by Hubrecht under these experimental conditions.
Both Nishimura 2 and Fueki3 investigated the effect of small alloying
additions of nickel, iron, titanium, chromium, and arsenic on the rate of
oxidation of aluminum-bronzes with aluminum contents from 8-10% at
700°C in air. While additions of up to 1% As, 1 % Cr, 4% Ni, and 6% Fe
were practically without effect, an addition of titanium up to 6% caused a
slight decrease in the resistance to oxidation.
Miyake 4 carried out electron diffraction experiments on copper-nickel
alloys with 7% Ni, 40% Ni, and 70% Ni + 1.4% Fe + 1 % Mn (Monel). The
external scaling layer appearing on the copper alloy containing 7% Ni after
oxidation between 300 and 700°C, for example, consisted entirely of copper
oxides. After removal of this outer scaling layer, the brown layer lying under
the surface could be identified as NiO on basis of the lattice constants which
were obtained. Sartell et al. 5 as well as Levin and Wagner 6 have also in-
vestigated the oxidation of copper-nickel alloys. Besides kinetic experiments
and micrographic studies, they could conclude from marker experiments
that an oxygen diffusion through the Cu 20 + NiO porous zone must take
place. The scale has the following composition:

unoxidized alloy ICu + NiOlCu20 + NiO + poresICu20ICUOI02(gas)


We shall discuss the interesting experimental results in detail later in
Section 4.8. At higher nickel concentrations a preferred NiO formation was
observed which was explained later on basis of the oxidation theory of metal
alloys treated in Section 4.7, which was developed from experiments on
copper-nickel alloys with 12-90% Ni, according to Hickman and
Gulbransen. 7

1 Preston, G. D., and L. L. Bircumshaw: Phil. Mag. (7) 20, 706 (1935).
2 Nishimura, H.: Suiyokwai Shi 9, 655 (1938).
3 Fueki, K., and H. Ishibashi: J. Electrochem. Soc. 108, 306 (1961).
4l\iiyake, S.: Sci. Papers Inst. Phys. Chem. Research (Tokyo) 31, 161 (1937).
5 Sartell, J. A., S. Bendel, T. L. Johnston, and C. H. Li: Trans. Am. Soc. Metals 50,
1047 (1958).
6 Levin, R. L., and J. B. Wagner: J. Electrochem. Soc. 108, 954 (1961).
7 Hickman, J. W., and E. A. Gulbransen: Trans. ALME 180, 534 (1949).
170 4. Scalin~ Processes with Formation of Thick Protective Layers

Noble-metal additions to copper cause a different situation than the


alloying additions which were discussed earlier, since they do not enter
into the oxide layer. This causes large concentration changes in the alloy
phase close to the alloy/oxide phase boundary. Leroux and Raub l reported
on oxidation experiments on copper-silver alloys in oxygen and air between
600 and 700°0. In later measurements 2 the parabolic rate law in copper
alloys with 10-80% Ag was confirmed. Furthermore, it was observed that the
oxidation rate was fastest for pure copper and the same for copper-silver
alloy with 80% Ag. Above 90% Ag the oxidation rate was very rapid
initially and later even assumed "negative" values (mass loss). The internal
oxidation which was particularly marked in these alloys will be discussed
in detail in Section 4.8. Raub and EngeP investigated the oxidation of
copper-gold, copper-palladium, and ternary alloys. In all these cases an
"inner" and "outer" oxidation zone appeared. While no adherent oxide layer
was obtained in nickel-gold alloys,3 Kubaschewski4 found the parabolic oxida-
tion law in copper-gold alloys with 5-10% Ou, owing to the good adherence
of the Ou 20 layer that was formed. This could be traced back to the smaller
volume ratio of Ou 20 and gold-copper alloys compared to OU/OU20
[Au + 5% Ou between 500 and 900°0: k" = 675· exp(-11,200/RT) and
Au + 10% Ou between 400 and 870°0: k" = 11 X 104 • exp(-11,500/RT)
g2/cmL sec] .
The relatively good plastic properties of OU20 layers at higher tempera-
tures, where compact and pore-free protective layers continually arise,
are quite impressively demonstrated by the following crucial experiment
by Moore. 5 A sphere of polycrystalline copper was heated in air until the
oxide layer was about 1-3 mm thick. Since the oxide layer remains poly-
crystalline and free of cavities and cracks, and since the density of the inner
copper nucleus does not change, the oxide covering must contract uniformly
as much as the diameter of the copper sphere decreases. In a further experi-
ment a copper tube (outer diameter 9.6 mm and wall thickness 0.58 mm)
was heated in air at 1000°0 while argon was passed through the interior.
After 20 hr, the inside diameter had decreased from 8.4 to 8.3 mm while on
the outside an oxide covering had formed which was 0.9 mm thick. This
experiment was impressive confirmation that the mass deficit at the copper/
oxide phase boundary was removed due to plastic flow. Mackenzie and
Birchenall 6 have studied the plastic deformation of wustite (FeO) under a

1 Leroux, J. A., and E. Raub: Z. anorg. u. allgem. Chem. 188, 205 (1930).
2 Raub, E., and lVr. Engel: Vortriige d. Hauptver8. Deut8ch Ge8. Metallkunde, 1938, p. 83.
3 Wagner, C., and K. Grunewald: Z. phY8ik. Chem. (B) 40, 455 (1938).
4 Kubaschewski, 0.: Z. Elektrochem. 49, 446 (1943).
S Moore, W. J.: J. Chem. PhY8. 21, 1117 (1953j.
6 Mackenzie, J. D., and C. E. Birchenall: Corro8ion 13, 283 (1957).
4.2. Scalin~ Systems with Electron-Conductin~ Protective Layers 171

load of 15 and 25 kg at 800-1000°C. This is clearly shown in parts a and b of


Fig. 64, where the strain rate drops rapidly within a relatively short period.
z.o
. I
I
-0- IS Ka'LoGd

-.. 15 ~I.oad

a.

o o
~ <:
IV-O<
~ ~---
do

20 60 80
Time Under Streu. Hours
Fig. 64. Effect of plastic deformation of wustite at 900°C in a
hydrogen-water vapor atmosphere under varying loads, ac-
cording to Mackenzie and Birchenall. The Armco iron paral-
lelepipeds were 2 x 0.7 x 0.2 cm.

When the applied stress is increased after a constant creep rate has been
reached for a particular stress, further sintering may occur, as evident from
part b of the graph, where a load of 15 kg has been suddenly increased to
25 kg. This plasticity is only observable when the formation of magnetite
and hematite is excluded. It is well known that plastic flow occurs much more
easily in metal oxides having a cubic close-packed structure than in those
with a hexagonal structure. Thus the spinel Fe304 should be less plastic
than FeO, which has the ~aCI-type structure. Analogous to A1 20 3, which
was investigated by Wygant,l the hexagonal ot-Fe 203 has a more rigid
structure than Fe304 and is therefore the least plastic of the three oxides
of iron. The mechanism of these plastic flows has been partially described
by Schottky.2
4.2.1.2. The Rate of Oxidation of Nickel and Nickel Alloys
Nickel is attacked by oxygen at high as well as at low temperatures.
It becomes covered with a somewhat thick, practically pore-free oxide layer,
in which the reacting materials are spatially separated one from another
so that one of the two reactants, nickel or oxygen, or both, diffuses through
the scaling layer. The oxidation rate of nickel is slower than that of most
other metals. Corresponding to the defect structure discussed in Section 2.2,
the NiO formed on nickel is a p-type conductor with nickel ion vacancies
1 Wygant, J. F.: J. Am. Ceram. Soc. 34,374 (1951).
2 Schottky, W.: Z. Elektrochem. 63, 784 (1959).
172 4. Scaling Processes with Formation of Thick Protective Layers

NiDI! and holes EE> (= Ni3+). For this case, a preferred diffusion of nickel
in the form of ions and electrons via vacancies and holes, which is rate-
determining at high temperatures and causes the parabolic rate law, is to
be expected and was found.
One would expect from the exclusive diffusion of nickel ions and electrons
through the NiO layer that platinum markers in the form of very thin
chemically inert wires placed on the surface of the nickel would remain
there after the oxidation. In the Ilschner and Pfeifferl oxidation experiments
carried out at lOOO°C (see Fig. 65) the platinum wires lay in the middle of

Pc

Fig. 65. Photomicrograph of nickel-manganese alloy (0.1 wt. % Mn)


oxidized for 48 hI' in air at 1000°C, according to Ilschner and
Pfeiffer. The platinum wires (0.3 em diameter) used as "markers"
lie in the middle of the oxide layer, contrary to expectations (en-
largement 150 times).

the scaling layer after the experiment. These findings are difficult to reconcile
with the mechanism of diffusion via lattice defects mentioned above. To
what extent a "marker displacement" can be caused by an inward plastic
flow in the NiO layer, as was demonstrated for copper oxidation by Moore,
should be tested by suitable experiments. A critical discussion on the
usability of marker experiments was given by Mrowec. 2 Birchenall3 explains
the anomalous position of the markers in the oxide layer by an "under-
cutting" mechanism.

1 Ilschner, B., and H. Pfeiffer: Naturw. 40, 603 (1953).


2 Mrowec, S.: Z. physik. Chern. [NF] 29, 47 (1961).
3 Meussner, R ., and C. Birchenall: Corrosion 13, 677 (1957).
4.2. Scaling Systems with Electron-Conducting Protective Layers 173

Corresponding to (2.21a) in Section 2.2, the nickel ion vacancy con-


centration XNiO" or the nickel ion partial conductivity UNi is proportional
to the sixth root of the oxygen pressure:
"Ni = const . p~,6

From a derivation completely analogous to that for copper oxidation,


we determine that the oxidation rate of nickel is proportional to the sixth
root of the oxygen pressure. Therefore,
k = const {W - 6v'Pffi} (4.32)

as has been shown in oxidation experiments by Wagner and Grunewald. 1


The experimental results obtained at 1000°C and different oxygen pressures
are summarized in Table 26. At high oxygen pressures p~), the value for
P~~ (equilibrium pressure over Ni + NiO), for example, 7.4 'x 10-11 atm at
1000°C, and even the sixth root of the value, can be neglected. Since the
relative partial conductivity of the Ni 2 + ions through vacancies (hu = Ul)
is unknown, a preliminary calculation of the scaling constants, as in the case
of copper oxidation, cannot be carried out.

Table 26. Oxidation of Nickel at 1000°0, According to Wagner and Grunewald

Rate constant I of the phase·


Po" Rational scaling constant k, boundary reaction,
atm equiv./cm-sec equiv./cm2 -sec

3.0 X 10- 4 0.6 X 10- 11 0.7 X 10-8


2.25 X 10- 3 1.1 X 10- 11 3.1 X 10- 8
1.45 X 10- 2 1.4 X 10- 11 4.6 X 10- 8
8.3 X 10- 2 1.8 X 10- 11 5.3 X 10- 8
1.00 2.8 X 10- 11 6.2 X 10- 8

Kubaschewski and von Goldbeck 2 and Gulbransen and Andrew3 have


concerned themselves with the kinetics of the oxidation of high-purity
nickel and have compared their later results with those of earlier work. The
temperature dependence of the scaling constants of a few nickel samples
from different authors are plotted in Fig. 66. As can be seen, there are
differences in the scaling constants (up to 104 ) for different types of nickel,
and if we consider the defect behavior in NiO, it appears obvious4 that these
differences can be traced back to metal impurities such as iron, chromium,
etc., which are introduced during the oxidation as higher. valent cations in
1 Wagner, C., and K. Grunewald: Z. physik. Chern. (B) 40, 455 (1938).
2 Kubaschewski, 0., and O. von Goldbeck: Z. MetaUk. 39, 158 (1948).
3 Gulbransen, E. A., and K. F. Andrew: J. Electrochem. Soc. 101, 128 (1954).
4 See the review by K. Hauffe: Ergeb. exakt. Naturw. 25, 193 (1951).
174 4. Scaling Processes with Formation of Thick Protective Layers

the NiO layer. According to (2.23b) in Section 2.2, the small NiH vacancy
concentration and thus the oxidation rate can change by about one order of
magnitude. On this basis we may regard as most probable the data which
give the smallest scaling constant for pure nickel. Just as for copper oxidation,

~~r-------r-------r-----~

Fig. 66. Summary of a few results of the


temperature dependence of nickel oxida-
tion, by Gulbransen and Andrew. A: von
Goldbeck: Diplomarbeit, Stuttgart, 1944
(carbonyl nickel melted in vacuo); B: Pilling
and Bedworth (electrolytic nickel); C:
Matsunaga: Japan. Nickel Rev. 1, 347 (1933)
(electrolytic nickel); D: Kubaschewski and
von Goldbeck: Diplomarbeit, Stuttgart, 1944
(carbonyl and pure nickel); E: Moore: J.
Chem. Phys. 19, 255 (1951). (99.9% pure
nickel); F: Gulbransen and Andrew (especially
Q7
pure nickel with only 0.0002% Fe, 0.0005% Si,
and 0.001 % Cu).

participation of phase-boundary reactions was noted for nickel oxidation


even at 1000°C (Table 26); an interpretation of these reactions is still out-
standing. Recently, oxidation experiments with nickel have been extended
to higher temperatures (1000-1200°C) and higher oxygen pressures (from
6.5 x 10-3 to 20.4 atm).1 In agreement with earlier experiments, the rate of
oxidation below 1 atm follows the PlJ.4 dependence. However, the rates
above 1 atm appeared insensitive to pressure. NiO oriented in a preferred
manner with {100} planes parallel to the polycrystalline nickel base.
Moore and Lee 2 studied the temperature dependence of the oxidation
rate of nickel at Po. = 100 mm Hg between 400 and 900°C (Fig. 67). The
activation enthalpy iJH obtained was 34.7 ± 0.8 kcaljmole, in good agree-
ment with the value of 34 kcaljmole according to Kubaschewski and von
Goldbeck. The value for the activation entropy iJS was found to be -17.3
± 0.9 caljmole-deg.
Wagner and Zimens 3 as well as Gulbransen and Andrew 4 were able to
1 Baur, J. P., R. W. Bartlett, J. N. Ong, Jr., and W. M. Fassell, Jr.: J. Electrochem.
Soc. 110, 185 (1963).
2 Moore, W. J., and J. K. Lee: Trans. Faraday Soc. 48,916 (1952).
3 Wagner, C., and K. E. Zimens: Acta Chem. Scand. 1, 547 (1947).
4 Gulbransen, E. A., and K. F. Andrew: J. Electrochem. Soc. 101, 128 (1954).
4.2. Scalin~ Systems with Electron-Coliductin~ Protective Layers 175

-etl

loI: V
°6
~-1d'
rI /
J
I
E
u
I-:
~f6'

Fig. 67. Temperature dependence of the rate of oxidation .f'f


I
as 1.tl.] 1.2 1.1' 1.&
of nickel between 400 and 900°C at an oxygen pressure
of 100 mm Hg, according to Moore and Lee. f-
detect a decrease in the rate of oxidation with increasing oxide layer thickness
at 1000°0 and 1 atm oxygen as well as at 475°0 and Po. = 76 mm Hg.
While this process is not well understood at high temperatures, at 475°0
space-charge effects evidently simulate decrease of the oxidation rate, as
Hauffe and co-workersl were able to show, since, by plotting the data as a
cubic rate law, a straight line was obtained (Fig. 68), with a scaling constant
ztJtJtJ

r
-- - -
A
o: '"
,...,,-
o6'IKJ

Fig. 68. Cubic course of oxidation of ~ ,...--


~ 4I7t7
nickel at 475°C and Po. = 76 mm Hg
from data of Gulbransen and Andrew,
evaluated by Engell, Hauffe, and i
Ilschner. i'tJ(/ IW 6W 1000 ?6I7Q
Minutes ___

independent of the layer thickness. The point of view taken by Gulbransen


that the parabolic scaling constant, which becomes gradually smaller with
time, is to be traced back to an "oxidizing out" ofthe less-noble "impurities"
with higher valence than the nickel ions appears to be quite a plausible
explanation for the process found at 1000°0. Recently, Kalvenes and Piene 2
were able to demonstrate the mass loss due to evaporation of chromium
oxide, probably as OrO, during the oxidation of Nimonic 75 (on 80% Ni-
20% Or basis) at 900-1000°0 in a 00 2 atmosphere. As is shown in Fig. 69,
the oxidation at low 00 2 pressures causes an initial mass loss, but after a
longer oxidation time a weight increase is found, which is related to formation
of an oxide layer. Nevertheless, in experiments at 475°0 the author is more
inclined to the viewpoint that the space-charge effect is responsible for the
course of the oxidation with time.
1 Engell, H. J., K. Hauffe, and B. Ilschner: Z. Elektrochem. 58, 478 (1954).
2 Kalvenes, 0., and K. Piene: private communication.
176 4. Scaling Processes with Formation of Thick Protective Layers

If higher-valent metal ions, for example Cr3+ or Mn3+ or Mn4+, penetrate


the NiO layer during oxidation of nickel or of nickel alloys, then the number of
the nickel ion vacancies NiO" is increased (Fig. 11) according to (2.23b)
in Section 2.2, as was discussed in detail earlier. According to (2.22b),

0.7

0.6

0.5

0.4

0.3
10 .2
0.1

Fig. 69. Oxidation rate of Nimonic 75 at 900°0 and various OOz pressures, according
to Kalvenes and Piene.

introduction of monovalent metal ions, as for example Li+ or Ag+, into the
NiO lattice should decrease the number of nickel vacancies and slow down
the rate of oxidation. Since lithium was not considered suitable for alloying
with nickel because of its low melting point, Pfeiffer and Hauffe 1 used silver
as an alloying metal. The oxidation rate of nickel with additions of 0.1 to
1 at.% Ag is practically unchanged as the experiments have shown (Fig. 70).
These findings are reasonable if one considers that due to the relatively large
ionic radius of Ag+ (1.1 A compared to 0.78 A for NiH) there is no appreciable
solubility. However, there is still the possibility of introduction of divalent
silver ions into the NiO lattice, according to the equation
(4.33)
where lattice defects are neither additionally formed nor annihilated because
the silver ions have the same charge as the nickel ions. Here Ag. X(Ni)
represents a divalent silver ion situated in a NiH lattice position and the
cross ( x ) denotes no excess charge. Under these conditions the solubility of
Ag 2+ ions could be high. The appearance of divalent silver ions has already
1 Pfeiffer, H., and K. Hauffe: Z. MetaUk. 43, 364 (1952).
4.2. Scalin~ Systems with Electron-Conductin~ Protective Layers 177
$~----------~----------~------------~--~~----~--------,
ge cm-4 I : Ni. 1.0 Atom % Cr
n : Ni + 0.5 " " "
m: Ni + 0.1 " " "
IY:Ni.O.1 " " " If
14
f: Pure Nickel
1'I: Ni + 1.0 Atom % Ag
I'D: Ni .0.7" " "
1'111: Ni +03 "
10 IX: Ni • 0.1 "

o
Time,hr-
Fig. 70. Oxidation of nickel, nickel-chromium, and nickel-silver alloys at 1000°C in
oxygen at 1 atm, according to Pfeiffer and Hauffe.

been described by different authors.! Unlike silver, Li+ ions can be introduced
in considerable quantity into the NiO lattice because of their ionic radius
of 0.78 A, as conductivity measurements on the NiO + Li 2 0 system by
Verwey and co-workers 2 have shown (Fig. 12). By use of the mass action
law in (2.20) at constant oxygen partial pressure:
XNiD""X;=K (po 2 =const) (4.34)
and the electroneutrality relationship
X@ = 2xNiD" + XLI (4.35)
where XLI is used for XLi.'(Ni) for brevity, and with XLI ~ x~ = 2x~ ON'
we obtain for the ratio of the scaling constants in the formation of a pure
NiO layer k O and a layer of a NiO-Li 2 0 solid solution k
k x'!v
!CO = XLI
(4.36)

The symbol x~ denotes the concentration of holes in the pure NiO phase
at the corresponding temperature and the applied oxygen partial pressure.
x~ can be determined from conductivity and Hall-effect measurements.
1 See for example J. H. De Boer and J. van Ormondt: Proceedings of the International
Symposium on the Reactivity of Solids, G6teborg, 1952, p. 557.
2 Verwey, E. J. W., P. W. Haayman, and F. C. Romeyn, Chem. Weekblad44, 705 (1948).
178 4. Scalin~ Processes with Formation of Thick Protective Layers

In order to check the validity of expression (4.36), at least qualitatively,


Pfeiffer and Hauffe1 determined the oxidation rate of nickel at lOOO°C in a
pure oxygen atmosphere and then in an oxygen atmosphere containing
Li 20 vapor. Figure 71 shows the experimental results. As previous experi-
ments indicated, the degree to which the effects are obtained depends on the

~~----+-----+-----4-----~~~~

~~---4-----+-----r~--~--~

J
~w~----+-----+~---I------I-----\
E1l'
~
Fig. 71. Decrease in the rate of
oxidation of nickel at 1000°0 in an
oxygen atmosphere at 760 mm Hg
containing Li20 vapor, according
to Pfeiffer and Hauffe. I: nickel
o~~~~ ....~w~----~~........~~~....-~= sheet in pure oxygen; II: nickel
sheet in oxygen + Li2 0 vapor (kN
Time, hr-_ is in g2/cm4 -hr).

vapor pressure, that is, on the concentration of the lithium ions which are
incorporated, as was to be expected. Possibly, under suitable conditions, the
scaling constant could be reduced still further. Since nickel-lithium alloys
are not commercially available, experimental difficulties are encountered
with the introduction of Li 2 0 from the outside into the scaling layer which is
formed because quartz, as well as metal in direct contact with Li2 0, is
vigorously attacked, and suitable methods for the production of Li20 vapor
still have to be developed. The initially larger scaling constant (Curve IIa
in Fig. 71) indicates that under the experimental conditions, the oxidation
first proceeds too rapidly for Li 2 0 incorporation in suitable concentration
in the oxide layer which is formed. The following experiment shows this
assumption to be true. Nickel was heated in a Li 2 0-vapor-containing
oxygen atmosphere of about I.mm Hg for 5 hr at lOOO°C. Under these con-
ditions, the initial oxidation proceeded slowly enough to produce the expected
effect of the incorporation of Li+ ions in a sufficient concentration. Of course,
during this pretreatment in vacuum, the Li2 0 which was placed in the reaction
tube vaporized extensively and was not available in sufficient quantity to
affect the later course of the oxidation. Thus, it is understandable that
curve IIb in Fig. 71 cannot display the maximum reduction in the scaling
constants. Experimentally, it is much simpler to incorporate Li 20 into the
1 Pfeiffer, H., and K. Hauffe: Z. Metallk. 43, 364 (1952).
4.2. Scaling Systems with Electron-Conducting Protective Layers 179

oxide layer which is formed by placing the nickel sample which is to be


oxidized loosely in a Li 20-MgO mixture (for example, 1 to 10 mol. %
Li20 + remainder MgO) rather than by heating in a Li20 vapor.
In the same way, Brauns and Rahmel1 tried to decrease the oxidation
rate of iron by heating it in a Li 20-containing air atmosphere of 0.5 mm Hg
at 850°C. The presence of an Fe203 layer causes the effect to be rather small.
Because of the inherent very high concentrations of FeD" and EB in
FeO, the degree of incorporation of Li + ions is not sufficient to achieve a
significant lowering of the oxidation rate. This has been confirmed by the
experiments.
The relationships in the oxidation of nickel-chromium alloys are not
quite so clear, as the kinetic measurements by Wagner and Zimens 2 and the
electron optical investigations of this scaling layer by Gulbransen3 have
shown. At lower chromium contents in the nickel-chromium alloys, with
the assumption of an equally probable appearance of NiH and Cr 3 +
ions in the scaling layer [xcr(alloy) = xcr(scaling layer)], we obtain from
(4.34) in a similar manner as above and from the electroneutrality relation-
ship
(4.37)

where XCr == XCr.·(Ni) with XCr ~ x~ = 2x~iO' for the ratio of the scaling
constants of the alloy k, and the pure nickel kO :

k XCr
F=XO- (4.38)
o
Oxidation experiments with nickel alloys with small additions of chromium
and other metals which form ions with a valence greater than two confirm
the expected increase in the oxidation rate according to (4.38) (see Table 27
and Fig. 72).4 The decrease in the scaling rate for higher chromium additions
(greater than 6 at. %) is evidently associated with the ever-increasing
formation of Cr203 or spinel, which starts out at the grain boundaries of the
NiO-Cr203 solid solutions, and represents the main product for chromium
additions in excess of 15 at. %. A more extensive discussion follows later.
The assumption of an equal concentration of the metal ions in the scaling
layer and in the alloy phase for the nickel---chromium alloys is generally not
fulfilled, as is evident from the experimental results of Wagner and Zimens.
This can often be attributed to the different mobilities of the NiH and Cr3+ ions

1 Brauns, E., and A. Ralunel: Wer~toffe u. Korroaion 7, 448 (1956).


2 Wagner, C., and K. E. Zimens: Acta Chern. Scand. 1,547 (1947).
3 Hickman, J. W., and E. A. Gulbransen: Trans. AIME, Techn. Public. No. 2069 (1946).
4 See also L. Horn: Z. Metallk. 40, 73 (1949).
180 4. Scalin~ Processes with Formation of Thick Protective Layers

Table 27a. Oxidation Rate of Nickel-Chromium Alloys at 1000°C and po,


1 atm, According to Wagner, Zimens, Pfeiffer, and Hauffe

Wt.% Oxidation time,


LIm
q' k" = ~ (LI; )2,
Cr sec g/cm 2 g2/cm4-sec
Wagner and Wagner and Pfeiffer and
Zimens Zimens Hauffe

o 16,080 (14,400) 2.5 X 10-3 3.8 X 10-10 2.4 X 10-10


0.1 14,400 8.4 X 10-10
0.3 14,400 4.7 X 10-3 15 X 10-10 11 X 10-10
0.5 14,400 14 X 10-10
1.0 14,400 6.3 X 10- 3 28 X 10-10 28 X 10-10
3.0 14,400 7.2 X 10- 3 36 X 10-10
10.0 14,400 2.7 X 10- 3 5.0 X 10-10

Table 27b. Temperature Dependence of the Parabolic Scaling Constants of an


80 Ni-20Cr alloy with Information on the Activation Enthalpy LlH, the Activation
Entropy LIS, and the Free Activation Energy LlF, According to Gulbransen and
Andrew

k", g2(cm4 -sec, LlH, LIS, LlF,


T,oC average value cal/mole call (mole-deg) cal/mole

650 2.32 x 10-15 - 15.7 52,650


750 1.46 x 10-14 16.2 54,750
38,150
850 9.49 x 10-14 15.8 55,900
875 1.81 x 10-13 - 15.1 55,450

in the scaling layer. By analogy to (3.9), we obtain for the migrating quantity
of nickel ions plus electrons in equivalents per second:

dnNi 1 {d/LNi2+ dV}


-----;u- = !qF2)(Ni - - ~ - 2Fdf (4.39)

where F is the Faraday number. On basis of the dissociation equations, we obtain


for the chemical potentials /LNi and /LCr of the neutral metal atoms
/-lNi2+ -I- 2/-l e- = /-lNi (4.40)

Jlcr 3 + -I- 3/-le- = /-lCr (4.41 )

In the case of solid solutions, we can relate the chemical potential of the nickel
or of the chromium, /LNi or /LCr, to that of the pure phases /L(Ni) or /L(cr) and the
activities aNi or aCr according to 1
Il)<j = /-l(Ni) -I- R T In aNi
(4.42)
/-lCr = /-l(cr) -I- RTlnacr

1For example, \V. Schottky, H. Ulich, and C. Wagner: Thermodynamik, Springer,


Berlin, 1929, pp. 357ff.
4.2. Scaling Systems with Electron-Conducting Protective Layers 181

Using equations (4.40) to (4.42) we obtain the following equations from (4.39)
after conversion in the manner shown in Section 3.2.
1 RT d In aNi
- q"2 F2 "Ni dg (4.43)

1 RT RT d In aCr
- q"3 }fi"cr N LB dg (4.44)
dt
Further, the ratio of the equivalents of nickel and chromium migrating through
the scaling layer (independent of the layer thickness) is
dncr 3 + / dnNi 2 + _ ZCr _ 3 (4.45)
dt dt - z;;-. XCr - 2"XCr
Substituting (4.43) and (4.44) in (4.45) we find the ratio of the transported ions
to be
dlnacr
dlnaNI (4.46)

The ratio of the conductivities is proportional to the number of the individual


jumps of the two types of ions. Because of the different charges, the ratio is
proportional to the factor 3/2 and, moreover, there is an additional proportion-
ality to 3/2 due to the ratio of the number of the individual jumps in an electric
field of given strength or in a prevailing, constant, chemical-potentil1>l gradient.

Fig. 72. Oxidation scheme of a nickel-


chromium alloy. While in the case of
pure NiO protective layer the concentra-
tion XNiD greatly increases from phase
H

boundary I to phase boundary II, in the


case of simultaneous introduction of
chromium ions, the concentration XNiD"
assumes high values at phase boundary I
and, depending on the mobility ratios of

I
the chromium and nickel ions in the
~as Phas~ O~ scaling layer, either does not change
~oy HIO-J1l(iij
+ Cr"-lons: notably (above representation) or even
L--Njs+ - I decreases toward phase boundary II.
Diffusion { ~Cr!+--l Thus, it shows just the opposite course as
process ~lJor38- -1 with pure nickel protective layers.

When chromium is present in very small amounts, the ratio of jumps is given by
the ratio of the concentrations (Xcr), while with larger amounts the jump ratio
of the chromium ions, fJxcr, is greater by about a factor fJ because of the electro-
static attraction of the oppositely charged. chromium ions and nickel ion vacancies.
By introduction of the mobilities, UCr and UN;' we then obtain
3 3 UCr
= -XcrfJ-- (4.4 7)
2 2 UNi
and from (4.45), (4.46), and (4.47)
din aCr 1 UNi
(4.48)
d In aNi fJ UCr
182 4. Scaling Processes with Formation of Thick Protective Layers

The relationships between activities and lattice defect concentrations are


obtained from the equations for the incorporation of the metals in the oxide
Ni + NiD" + 2EB = (NiH + 2e-) in the lattice position,
or (4.49)
Cr + NiO" + 3$ = er.·(Ni)

and the equilibrium conditions which follow [Xcr == XCr.·(Nl)]

aNi· XNi~" • x~.v = KI (4.50)


and
aCr • XNI 0" • X~ = K 2 • XCr (4.51 )
On basis of the reaction equation Ni + !02 = NiO, the following is valid:

fLNI + ~ /1o, = /lNiO (4.52)


Since fLNiO is to be regarded as a constant, it follows from (4.52) that
d/lNI = RTdlnaNi = -td/lO, = -tRTdlnpo, (4.53)
By introduction of (4.53) into (4.48), we finally obtain

dIn XCr = ~(l _ ~ _ ~ UNi ) (4.54)


dlnp0 2 2 3 {3ucr
If in addition, as Wagner and Zimens emphasize, it is not possible to make a
simple prediction about the numerical factor (1/ {3)(UNi/Ucr), then the factor {3,
which is significant in further evaluation as the measure of the electrostatically
favored stopping probability, can be calculated according to (4.54) from addi·
tional measurements for the determination of the mobilities UN! and UCr within
certain concentration regions of the NiO-Cr 2 03 solid solution.
In the following we introduce the symbol y for the expression in (4.54), i.e.,

y = ~(1
2
_ ~3 ~{3 UNi)
UCr
(4.55)

Integrating (4.54) between the limits I and II (see Fig. 71) and taking anti·
logarithms, we obtain
xII (pH)
-;:-1
Or
= -pI
0, Y
(4.56)
or 0,

From (4.54) and (4.55), with d In po, = (l/y)d In XCr, and the relationships

XCr = 2xNio"

and

(4.57)
we obtain for the rate of transport of the Ni 2+ ions, according to (4.32), (4.53),
and (4.54):
dnNi RT K dXcr
(4.58 )
dt = qFz8y dg
4.2. Scalin~ Systems with Electron-Conductin~ Protective Layers 183

In the steady state, the nickel transport rate may be regarded as a constant,
independent of the location coordinate g and the concentration gradient dXcrjdg,
so that we can deal with an average chromimn content xcr:
(4.59)

By means of (4.56) and (4.59), we can calculate the chromimn contents at the
phase boundaries, if the average chromimn content Xcr and the corresponding
oxygen pressures, P~2 and p~., at phase boundaries I and II (Fig. 72) are known.
We then obtain

(4.60)
and

(4.61)

Inserting the difference of the values from (4.60) and (4.61) in (4.58) and dividing
by Jg, we obtain after consideration of (4.57), corresponding to XCr '" XNI. the
following expression for the transport velocity:

dnN! q RT 2 (pIJ./pb.)y 1
(4.62)
dt = --:1f 4F2 "N!; (p~./p~.) Y + 1

Since a general application and solution of (4.62) is not always possible, Wagner
and Zimens discuss the following limiting cases:
1. If y --+ 0, that is, if (3(Ucr/UN!) ~ 2/3, then (4.62) simplifies to

dnN! q RT I II / I
~ = ---:1f 4F2 "N! n(p 0, po.) (4.63)

Therefore the transport speed and, consequently, the scaling rate is only slightly
dependent on the oxygen pressure because of the logarithm of the oxygen-pressure
ratio. Further, the diffusion gradient of the Cr 3 + ions is practically nonexistent,
and because of pair formation, Cr 3 +NiD", and accompanying electrostatic
effects, the concentration gradient of the NiO" particles is negligibly small.
2. For the case where y > 0, so (3(UcrjUNi) > 2/3, there is an enrichment
of chromium on the side with the higher oxygen partial pressure. Further,
if p~. ~ p~" then the final reaction in (4.62) is equal to one, which yields an
oxygen-pressure independence for the oxidation rate.
3. For the case where y < 0, so (3(Ucr/UNl) < 2/3, the right side of (4.55)
becomes negative, with a related preferential enrichment of chromimn in loca-
tions of lower oxygen pressure. The concentration gradient will be oppositely
directed to the migration direction of the cation vacancies, so that a migration
to the oxide/oxygen phase boundary can be produced only by an additional
diffusion potential. This case appeared during the course of structure investiga-
tions of the scaling layer by Moreau and Benard (see Fig_ 73).
184 4. Scalin~ Processes with Formation of Thick Protective Layers

Further studies on structure and composition of scaling layers in


nickel-chromium alloys are required for evaluation of these theoretical
relationships. Experiments of this kind were published by Gulbransen and
Andrew.! According to these and earlier results,2 NiO formation predominates
in the first stage of oxidation (2 hr) between 500 and 1000°C and po. = 76 mm
Hg, while for experimental timcs over 30 hr a Cr203 enrichment was observed.

Fig. 73. Structure of the inner and


outer oxidation zones of a nickel-
chromium alloy, according to Moreau
and Benard. While the right part of
the outer zone consists principally of
almost pure NiO or a NiO-Cr20a solid
solution, a quantity of NiO and
NiCr204 is contained in the part lying
N i+ O ~ N i O in the interior. (The chromium content
NiO+CrzO, - NiCr2o,. of the alloy here amounts to only 5%.)

Under the same experimental conditions as above, predominant Cr203


formation in the scale was observed in a nickel-chromium alloy which
had the following composition in weight percent: 0.08 C; 0.01 Mn; 1.39 or
0.30 Si; 19.91 or 19.98 Cr; 0.34 or 0.32 Fe; 0.10 or 0.05 Zr; 0.024 or 0.029 Ca;
and 0.07 or 0.08 AI. The experimental results on the scaling rate are presented
in Table 27b with information on the activation enthalpy, entropy, and the
free energy of activation.
Similar results have been obtained by Birks and Rickert3 who studied
the oxidation of nickel-chromium alloys containing 30, 60, and 80 wt. % Cr
in the t emperature range of 850 and 1l00°C. The composition of the oxide
scale was practically the same as represented by Moreau and Benard in
Fig. 73 . It is proposed that oxygen diffuses into the bulk alloy, where
chromium is preferentially oxidized to Cr203, which acts as a marker in a
matrix of pure nickel. This matrix itself is oxidized as pure nickel. Con-
sequently, by the simultaneous existence of Cr203 and NiO, a spinel forma-
tion could become possible.

1 Gulbransen, E. A., and K. F. Andrew: J. Electrochem. Soc. 101, 163 (1954).


2 Gulbransen, E. A., and W. R. McMillan: Ind. Eng. Chem . 45,1734 (1953).
3 Birks, N. , and H. Rickert : J. Inst. Metals 91, 308 (1962/63).
4.2. Scaling Systems with Electron-Conducting Protective Layers 185

For the proof of an overwhelming Cr20a formation the authorsl used


the following expression, derived by Wagner 2 (see p. 355):
+ 0
Xcr - Xcr
----- = 7T1! 2U expu2 • erfc u
XCr - XCr,O.
with

and the oxidation constant ke:


ke = (LlYmetad 2
2t
where x~r and x~r is the mole fraction of chromium in the oxide and in the
alloy, respectively. XCr,O, is the mole fraction of Cr20a in the oxide layer,
D is the interdiffusion coefficient in the alloy, and LlYmetal is the displacement
of the metal surface at time t. Using lee = 8 X 10-14 cm 2/sec for 1000°C,
according to Gulbransen and Andrew,3 and the corresponding diffusion
coefficient of chromium in Ni-20% Cr alloys, D = 2 X 10-11 cm 2/sec at
xg
1000°C,4 and with u = 4.5 X lO- a, we then obtain Xdr = 0.27 for r = 0.33.
Thus, even in the case of the alloy with the lowest chromium content (30%),
the chromium content at the metal/oxide interface is not expected to be
reduced to a level where NiCr204 would become the thermodynamically
stable phase. In all cases, therefore, only Cr20a should be expected,
as was in fact observed. Thermodynamical considerations confirm this
result. 5
On the other hand, a preferred spinel formation of the composition
MnCr204 occurs if one increases the manganese content to 1.70 wt. %.
The oxidation stability obtained at 1175°C was four to six times smaller
here. Gulbransen assumes a preferred diffusion of the chromium ions through
vacancies. A spinel formation in nickel-manganese alloys is to be expected
only if chromium is simultaneously present. In the absence of chromium
in nickel-manganese alloys, a constant increase in the oxidation rate appears
with increasing manganese content, as Wagner and Zimens have demon-
strated. Unfortunately, we do not have at present the kinetic and thermo-
dynamic data needed to determine wh.y in one case a preferred NiO or
Cr20a formation occurs and in another case a predominant spinel formation
1 Birks, N., and H. Rickert: J. Inst. Metals 91, 308 (1962/63).
2 Wagner, C.: Z. Elektrochem. 63, 772 (1959).
3 Gulbransen, E. A., and K. F. Andrew: J. Electrochem. Soc. 101, 163 (1954).
4 Wagner, C.: Z. Elektrochem. 63, 772 (1959).
:; Gruzin, P. L., and G. B. Fedorov: Doklady Akad. Nauk SSSR 105,264 (1955); A. D.
Tyutyunnik and G. V. Estulin: Fizika Metallov i Metallovedenie 4, 558 (1957).
186 4. Scaling Processes with Forlllation of Thick Protective Layers

occurs. On basis of electron diffraction procedures Jitaka and Miyakel were


able to show in oxidized 80% Ni-20% Cr alloys that the largest part of the
scaling layer consisted of NiCr204. The remarkably high resistance to oxida-
tion of the alloys investigated by Fontana,2 which consisted of 25 wt. % Cr,
50 wt. % Ni, 0.50 wt. % C, and the remainder Fe, may probably be attributed
to a spinel formation.
It appears, on basis of the lattice defect structure and measurements
of the diffusion rate of Crs+ ions in Cr20S carried out by Gulbransen,a that
the spinel phase generally exhibits the greatest resistance to diffusion.
Recently Hagel and Seybolt4 have conducted more extensive investigations
on the diffusion of chromium in Cr2 0a. On basis of the previous experimental
data, we may assume that the scaling layer which is formed on an 80% Ni-
20% Cr alloy from the attack of oxygen at 1l00°C consists of a spinel of the
composition NiCr2 04. Because Cr20a is vaporized probably via CrO with
relative ease, a slight deficiency of CrO will occur at the phase boundary II,
so that the possibility of a preferred migration of chromium ions discussed
by Wagner and Zimens is probable, since the additional concentration
gradient appearing because of vaporization favors the transport. On basis of
these considerations too, the appearance of a large outer Cr20a layer is
improbable, at least at 1100°C. At lower temperatures, where the rate of
vaporization of Cr20s is negligibly small, we have to deal with the appearance
of a Cr20a layer on the outer edge of the scaling layer, as Hickman and
Gulbransen 5 have demonstrated from electron optical investigations.
As investigations by Moreau and Benard6 showed, the composition of
the scaling layer on nickel-chromium alloys (0-10 wt. % Cr) can vary owing
to different experimental conditions. This can be seen, for example, after
long oxidation times in air between 800 and 1300°C in the structure that
is shown schematically in Fig. 73. In addition to the exterior scaling layer,
which consists of a heterogeneous mixture of NiO and NiCr204 in the interior
and of pure NiO in the region of the scaling layer/air phase boundary, an
inner oxidation zone appeared, in which the alloying metal was oxidized
to Cr20a in a fine distribution by the oxygen which penetrated into the alloy.
Thus, the inner oxidation zone consisted of statistically distributed, small
Cr2 0a particles embedded in the pure nickel matrix (for the mechanism of
internal oxidation see Section 4.8).

1 Jitaka, J., and S. Miyake: Nature (London) 137, 457 (1936).


2 Fontana, M. G.: Ind. Eng. Ohem. 45, 95 A, No.9 (1953).
3 Gulbransen, E. A., and K. F. Andrew: J. Electrochem. Soc. 99, 402 (1952).
4 Hagel, W. C., and A. U. Seybolt: J. Electrochem. Soc. 108, 1146 (1961).
5 Hickman, J. W., II-nd E. A. Gulbransen: Trans. AIME, Techn. Pub!. No. 2069 (1946).
6 Moreau, J., and J. Benard: Oompt. rend. 237, 1417 (1953); J. Inst. Metals 83, 87
(1954/55).
4.2. Scalin~ Systems with Electron-Conductin~ Protective Layers 187

As the oxygen partial pressure is reduced, for example, by use of


hydrogen-water-vapor mixtures, an increasing degree of selective oxidation
is observed. In this case chromium was preferentially oxidized, and appeared
as Cr2 03 on the alloy. Under these conditions, no significant internal oxidation
took place. On basis of oxidation experiments with nickel-chromium-alloy
samples of 4.6 wt. % Cr between 800 and 1250°C with a H 20jH2 ratio
between 6.5 x 10-3 and 5.9 x 10- 2, five oxidation periods can be observed,
which are shown in Fig. 74. In the first oxidation period, a varying, large
number of Cr203 crystallites with an average dimension of less than 1 f1,

Fig. 74. The five oxidation periods appearing


in the oxidation of a nickel-chromium alloy,
according to Moreau and Benard. (The periods
correspond to the progress of the nucleus
formation with time and to the growth of the
oxide crystals as a function of temperature and
gas composition.)

Fig. 75. Period I according to Fig. 74. Primary nucleus formation at


900°0 and PH 2 0/PH 2 = 6.5 X 10-3 • (Enlarged 135 x.)
188 4. Scaling Processes with ForIllation of Thick Protective Layers

appear on the various crystal surfaces (Fig. 75). In the second oxidation
period the H 20 partial pressure is increased, which leads to larger Cr203
crystallites. In both cases a zone free of oxide crystallites appears along the
grain boundaries of the crystallite, always lying on that side of the crystallite
which exhibits the smaller number of oxide crystals (see Fig. 76). The third
oxidation period is characterized by the fact that at higher temperatures
and still higher H 20 vapor pressures, the growing oxide crystals are assembled
into oriented bands (Fig. 77). Furthermore, it is significant that in any
individual crystal only definite parallel layering of the growing small oxide
crystals is permitted. If the reaction is biased in favor of this band formation
by a further increase in temperature and in the H20 partial pressure (oxida-
tion period IV), then at greater magnifications one can observe (Fig. 78),
that the bands in individual crystallites in the neighborhood of the grain
boundary change direction and fit into one another over the grain boundaries.
Treatment in hydrogen or in argon (in the latter case, of course, it is not so
pronounced) causes the band formation to practically disappear. After long
oxidation times (period V) a compact Cr203 layer is finally evident.
This band formation during selective oxidation can be pictured (in
terms of present empirical evidence) as a selective-oxidation-caused periodic
modification in the alloy surface covered with a thin Cr203 film, as Moreau
and Benard have schematically indicated in Fig. 79. The surface was almost

Fig. 76. Period II according to Fig. 74. Nucleus formation


and recrystallization of the oxide crystals at 1040°C withpH,o!PH. =
6.5 x 10- 3 (enlarged 500 x). (Depleted zones in the oxide crystals
appear at the grain boundaries.)
4.2. Scalin~ Systems with Electron-Conductin~ Protective Layers 189

Fig. 77. Period III according to Fig. 74. Assembly of growing oxide
crystals into oriented bands at 1200°0 with PH.O/PH. =
2.4 X 10-2 •
(Enlarged 850 x.)

Fig. 78. Electron microscope picture of oxidized nickel-chromium


alloys in the fourth oxidation period. (Enlarged 2500 x , reduced 10% for
r eproduction.) Here one can recognize the transition in the individual
oxide bands from one crystallite in the grain boundary to the other.
190 4. Scaling Processes with Formation of Thick Protective Layers

entirely cleared of the oxide film by treatment with hydrogen. Because of


the high surface tension of the oxide-free alloy, any surface projections tend
to disappear during this treatment, that is, a leveling of the surface takes
place.! Iron-chromium with 5% Cr produces the same phenomenon.
Similar observations of the oxidation of copper have also been reported
by other authors.2 Results, especially those in nickel-chromium oxidation,
permit the conclusion that the structure of the oxide layer in the oxidation of
mctal alloys depends decidedly upon the experimental conditions that are
significant in the oxidative rate law and adherence of the scaling layers.

Profile in
Oxidizing
Atmosphere
Oxide Film Fig. 79. Schematic representation of the profile
\
of the surface of the nickel-chromium alloy which
is formed in oxidizing and reducing atmospheres,
according to Moreau and Benard.

As a supplement to the above investigations, we may mention the


results of electron-diffraction experiments by Miyake3 on scaling layers on
iron-nickel (3-36%) alloys, on chromium steel, chromium-nickel steel
(18% Cr and 8% Ni), and nickel-chromium alloys. While a scaling layer
forms on iron-nickel alloys and iron-aluminum alloys which consists
predominantly of cx-Fe203, either spinel layers together with Cr 203 or solid
solutions of Cr203 and Fe 203, depending on the experimental conditions,
were observed on chromium-nickel steels and nickel-chromium alloys.
At highcr temperatures oxidized nickel-chromium alloys always exhibit
an cxterior scaling layer of practically pure NiCr204.4 The experimental
results of Quarre1l 5 on high-temperature-resistant chromium-nickel steels
are similar. Reference was made to the general significance of the protective
effect of such spinel layers in oxidizing furnace atmospheres. Further, the
author has given a qualitative discussion on the dependence of spinel forma-
tion on the composition of both the steel and the furnace atmosphere.
Russian authors 6 have measured the diffusion rate and activation energy
of chromium-51 in NiCr204 and NiAh04 as 2.0 x 10-5 and 1.2 x 10-3 cm 2/sec

1 Benard, J., J. Moreau, and J. Plateau: Z. Elektrochem. 61, 59 (1957).


2Elam, C. F.: Trans. Faraday Soc. 32, 1604 (1946); A. T. Gwathmey and A. F. Benton:
J. Chem. Phys. 8, 431, 569 (1940); P. Jaquet, Recherche aeronaut. 1951, 35.
3 Miyake, S.: Sci. Papers Inst. Phys. Chem. Res. (Tokyo) 31, 161 (1937).
4 See among others A. Grunert, W. Hessenbruch, and K. Schichtel: Elektrowarme 5,
2, 131 (1935); B. Lustman, Iron & Coal Trades Rev. 154, 889 (1947).
5 Quarrell, A. G.: Heat Treating and Forging 27,345 (1941); Iron & Coal Trades Rev.
142,703, 709 (1941).
6 Belokurova, I. N., and D. V. Ignatov: Atomnaya Energiya 4, 399 (1958).
4.2. Scaling Systems with Electron-Conducting Protective Layers 191

and 44,800 and 50,000 caljmole, respectively. For iron· 59 in these spinels,
values of 1.4 x lO-3 cm2/sec and 61,000 cal/mole were obtained. The
diffusion coefficients were compared with the oxidation rates of an 80%
Ni-20% Cr and of a 73% Ni-20% Cr-7% Al alloy.
Extensive data on the rate of oxidation of nickel alloys, up to 1939,
were summarized by Hessenbruch1 and now appear in a new edition by
Pfeiffer and Thomas. 2 The influence of oxidic embedding material on the
alloys, resistance to oxidation, and the adherence of the oxide layers which
are formed on them was investigated from the special viewpoint of the use
of nickel-chromium alloys as resistance wires at high temperatures. A
suitable mechanism for the corrosion of such wires in wet imbedding material
is proposed by Pfeiffer. 3 The following behavior can generally be noted:
an increase in the scaling rate of the resistance wires will always be observed
if a portion of the oxidic embedding material diffuses into the scale and
increases the concentration of the ion lattice defects, or if lower· melting
reaction products result with the oxides and spinels which are formed on the
wires. In the last case an especially rapid destruction of the heat conductor
takes place. We refer in such cases to a "catastrophic" oxidation (see Section
4.3).
By means of electron-optical investigations on oxidized nickel-chromium
and nickel-chromium-iron alloys of various compositions, Pfeiffer was able
to demonstrate that the composition of the scale depends on its thickness.
This can be seen from Table 28. 4 The remarkable change in the composition
is caused by displacement reactions. The percentage change in length of such
nickel-chromium and nickel-chromium-iron wires due to oxidation in air
at 1200°C5 is particularly significant.
Preece and Lucas 6 reported on the rate of oxidation of nickel-tungsten
and nickel-molybdenum alloys (with 5, lO, and 15 wt.% W or Mo) between
800 and 1200°C; there was no detectable spinel formation in the scaling
layer. The oxidation rate of the alloys becomes higher than the rate for
pure nickel only at temperatures above lOOO°C. The scaling layer formed on
nickel-molybdenum alloys consists of an outer layer of NiO and an inner
one of a mixture of NiO and molybdate. No vaporization of molybdenum
oxides was observed.
A further explanation of the oxidation mechanism of nickel-molybdenum

1 Hessenbruch, W.: Metalle und Legierungen fur hohe Temperaturen, Springer-Verlag,

Berlin, 1940.
2 Pfeiffer, H., and H. Thomas: Zunderfeste Legierungen, Springer-Verlag, Berlin, 1963.
3 Pfeiffer, H.: Werkstoffe u. Korrosion 12, 669 (1961).
4 Pfeiffer, 1.: Z. Metallk. 51, 322 (1960).
5 Pfeiffer, H.: Z. Metallk. 52, 481 (1961).
6 Preece, A., and G. Lucas: J. Inst. Metals 81,219 (1952/53).
192 4. Scaling Processes with Formation of Thick Protective Layers

Table 28. Composition of Oxide Layers on Nickel-Chromium and Nickel-


Chromium-Iron Alloys and Its Dependence on the Thickness of the Oxide Layer,
According to Pfeiffer

Oxide layer thickness


Composition, wt. %
300-500A 1000 A 10-30 p.
Alloy Ni Cr Fe Si Mn (at 1000°C) (at 800°C) (at 1l00°C)

Ni-Cr 77.6 21 1 1.4 0.2 NiCr 204 Cr20a Cr20a, NiCr 204
Cr 20a* NiCr204* MnCr204
FeCr204t
Ni-Cr-Fe 61.4 18 17.5 1.2 0.5 FeCr204 Cr20a, NiCr204
MnCr 204 MnCr204
NiCr 204 FeCr 204
Ni-Cr-Fe 33 21 43 2.3 0.8 MnCr204 Cr20a, MnCr204
FeCr204t MnFe204
Fe20at NiCr 204*

* Trace amount.
t Small amount.

alloys is provided by Brenner's experiments. l The influence of molybdenum


content on the oxidation rate as well as on the structure of the scaling layer
was investigated. As can be seen from Fig. 80, at 10000e the rate of oxidation

Fig. 80. Dependence of the rate of oxidation of


nickel-molybdenum alloys on molybdenum-tungsten
content at 1000°C in air, according to Brenner.

increases for nickel-molybdenum-tungsten alloys up to 4 at. % Mo + W,


and then remains constant with further additions up to about 14 at. % Mo.
The rate of oxidation decreases if the molybdenum content exceeds 15 at. %,
and reaches the value for pure nickel at about 20 at. %. Probably, the
initial increase in the scaling rate with increasing molybdenum content,
which was observed earlier by Horn,2 can be attributed to the formation of
a heterotype mixed phase with increased vacancy formation, as is indicated
by the symbolic equation
Mo0 2 .......... .:\10." (Nil + ~i 0" + 2NiO
The scaling layers formed on the alloys may always be divided into three

1 Brenner, S. S.: J. Electrochem. Soc. 102, 7 (1955).


2 Horn, L.: Z. Metallk. 40, 73 (1949).
4.2. Scaling Systems with Electron-Conducting Protective Layers 193

regions, as shown schematically in Fig. 81. As in the nickel-chromium


alloys, there is an inner oxidation zone in addition to the outer scaling layer
consisting of NiMo0 4 and NiO. A partial oxidation ofMo to Mo0 2 occurs in the
inner oxidation zone from the oxygen diffusing into the alloy. Little is known
about NiMo04, which appears in the outer scaling layer, but it appears to

Fig. 81. Schematic representation of the struc· Ni-Mo


ture of the scaling layer on a nickel-molybdenum
alloy at 1000°C, according to Brenner. (Finely
divided M002 crystals are present in the inner
oxidation zone abo The outer scaling layer
be + cd consists of two layers, NiMo04 and
Alloy '---y-.J Outer
Inner Oxidation Zone
NiO.) Oxidation Zone

be stable up to 1150°0. According to Brenner, the decrease in the oxidation


rate of nickel-molybdenum alloys by molybdenum additions exceeding
15% is effected by the increase of the Mo0 2 particles in the inner oxidation
zone, which retards the transport of new nickel atoms to the reaction front.
The growth rate of the inner oxidation zone is given by the difference between
the diffusion rates of the metal ions through the scaling layer and of the
oxygen through the alloy phase or through the inner oxidation zone. The
relationships may be expressed quantitatively, as will be shown later in
Section 4.8 for the internal oxidation of copper alloys, when the diffusion
coefficients concerned may be determined individually. Further investigations
on the influence of molybdenum on the oxidation rate of iron will be dis-
cussed later in connection with iron alloys.

4.2.1.3. The Rate of Oxidation of Other Metals and Alloys with


p- Type Conducting Protective Layers
Iron, cobalt, manganese, and chromium are examples of other metals
which form p·type conducting oxides, at least in the lowest oxidation state.
The oxidation mechanism of iron and manganese will be discussed later from
another viewpoint (see Sections 4.4 and 4.5). Since cobalt is known to assume
several oxidation states, discussion of some of the experimental results will
be deferred until later.
On basis of the intrinsic1 and p-conducting character of Or20a, the
oxidation of chromium belongs in this section, even though the mechanism
of defect formation is different 2 from that previously discussed for normal
1 Fischer, W. A., and G. Lorenz: Arch. Eisenhiltfenw. 28, 497 (1957); Z. physik. Chern.
[NF] 18, 265, 308 (1958).
2 Hauffe, K., and J. Block: Z. physik. Chern. 198,232 (1951).
194 4. Scaling Processes with Formation of Thick Protective Layers

p-conducting oxides (e.g. NiO). It is not possible to determine the ion defect
mechanism involved solely on basis of electrical conductivity and thermal
power measurements-we are supplied with information only on the electronic
disorder:
Null..--> e
+ Ef> (4.64)

Presumably the concentration of chromium-ion vacancies CrO'" and


chromium ions in interstitial lattice positions CrO" and oxygen ion vacan-
cies 00" which can determine the oxidation rate is small.
Gulbransen and Andrew1 studied the rate of oxidation of chromium
in an oxygen atmosphere at 76 mm Hg between 700 and 900°C. The experi-
mental results given in Table 29, which are evaluated by use of the transition
state theory, may be described by the parabolic rate law. 2 Carbon present
in the chromium reacts with the oxide or oxygen to give carbon monoxide.
As can be seen from the numerical values, the oxidation rate of chromium

Table 29. Scaling Constants and Activation Quantities for the Oxidation of
Chromium, According to Gulbransen and Andrew

k", LIS, LlH, LlF,


T,oC g2/cm4- sec cal/mole-deg cal/mole cal/mole

700 2.38 X 10- 14 13.0 53,680


800
850
900
2.2 x 10-13
8.9 x 10-13
1.32 x 10- 11
10.8
10.7
13.5
) 66,300
54,700
54,280
50,450

is very small and is of the same order of magnitude as that of the nickel-
chromium alloys.
So far no oxidation experiments with chromium alloys have been
reported in the literature. Unfortunately, in view of the defect equilibrium
(4.64), it is not possible to predict the effect on oxidation rate of alloying
elements that form di- or tetravalent metal ions in the Cr203 protective layer.
Presumably, an addition of titanium would reduce the oxidation rate of the
chromium.
The oxidation of cobalt is more easily implemented. According to
Valensi3 the oxidation of cobalt in air above 700°C proceeds with a pre-
dominant CoO formation, and below 700°C with C030 4 formation. Arkharov
and Lomakin 4 determined the critical temperature of the C030 4 formation
1 Gulbransen, E. A., and K. F. Andrew: J. Electrochem. Soc. 99, 402 (1952).
2 Arkharov, V. 1., V. N. Konev, 1. S. Traktenbsrg, and S. V. Shumilina: Fiz. :Metal
i Metalloved., Akad. Nauk. SSSR, Ural Filial 5, 190 (1958).
3 Valensi, G.: Met. ital. 42, 77 (1950).
4 Arkharov; V. 1., and G. D. Lomakin: Zhur. Tekh. Fiz. SSSR 14, 155 (1944).
4.2. Scaling Systems with Electron-Conducting Protective Layers 195

in air as 890°0. Further, Arkharov1 demonstrated by X-ray investigations


that the thickness of the 00304 layer formed on the 000 layer increases with
decreasing oxidation temperature in the oxidation of cobalt in air between
385 and 800°0. The higher rate of oxidation of 000 to 00304 which was
observed by Ohauvenet 2 is not understandable, and it could not be confirmed
by Johns and Baldwin3 from a check of the experimental results (Fig. 82).

C020, ~

C030~ ~~~~~~~~~~~;
[I
CoO~ffh'MW'/$W/.0W.&0W.0W
M-
t~J~--~--~--+---+---1--,H
g~cm:"·h·1 For C o Oxidation
"DI/nn I
10-* Cllol/vene/ -+----1-----l'1li"---1
II V D
o kilns & Baldwin
o Carler&RicllardSQfl

fO-~I--+--+--+---I---7+---i

tO~r---r-~~~+---;----r--~ Fig. 82. Compilation of the data of several


1.8 1.6' f.II J 1.2 1.0 as os authors on the temperature dependence
-;-if, of the oxidation rate of Co to CoO to
J{)() WlO .fOO GOO Coa04 at I atm oxygen.

We shall regard the rate of diffusion of the cobalt ions via vacancies in the
scaling layer as the rate-determining step above 900°0, where only the 000
phase is stable in air, as well as below this temperature, since 000 is a
p-type semiconductor with cobalt ion vacancies, as Wagner and Koch4
were able to demonstrate on basis of the oxygen-pressure dependencies of
the electrical conductivity. The increase in the conductivity effected by
incorporation of Li 2 0 into 000 confirmed the p-type character.5
Gulbransen and Andrew 6 investigated the oxidation rate of cobalt in the

1 Arkharov, V . 1., and Z . A. Vorsohilova: Zhur . Tekh. Fiz. SSSR 6, 781 (1936); V. I.
Arkharov and K. M. Graevskii, Zhur. Tehk. Fiz. SSSR 14, 132 (1944) .
2 Chauvenet, G.: Diss. Univ. Calin, No . 34 (1942).
a Johns, C. R., and W . M. Baldwin, Jr. : Metals Trans. 185, 720 (1949).
4 W agner, C., and E . K och: Z. physi k. Chem. (B) 32, 439 (1936).
5 Verwey, E. J . W., P. W. Haayman, and F. C. Romeyn: Chem . Weekblad 44,705 (1948).
6 Gulbransen, E. A., a nd K. F . Andrew: J. Electrochem . Soc. 98, 241 (1951).
196 4. Scaling Processes with Formation of Thick Protective Layers

temperature region between 200 and 700°0 with a different pretreatment


of the cobalt samples. The experimental finding is remarkable in that
below 400°0 and even at 200°0 a parabolic rate law was observed. Selected

Table 30. Scaling Constants of Cobalt at po. = 76 mm Hg, According to


Gulbransen and Andrew

T, °0 Pretreatment of the cobalt sheet

200 4.2 X 10-16


300 4.0 X 10-14
400 1.2 X 10-12
Cold worked and rolled
500 2.8 X 10-12
600 8.1 X 10-12
700 1.3 X 10-11

400 8.9 X 10-14


500 1.6 X 10-12
Six hour heat treatment at 885°C
600 5.8 X 10-12
700 2.2 X 10-11

experimental results are given in Table 30. Moore and Leel found the
following expression for the rate constants in the oxide film formation on
nickel between 400 and 960°0 and on cobalt between 500 and 800°0:
k' = 3.09 X 10-5 exp( - 38,400JRT) cm 2Jsec (for Ni)
and
k' = 7.60 X 10-4 exp(38,400jRT) cm2jsec (for 00)
Since the activation energy of nickel diffusion in NiO is the same as that of
cobalt diffusion in 000, we can attribute the 25-fold increase in the oxidation
rate of the cobalt to the cation vacancy concentration in the 000, which is
25 times greater than in NiO. Tichenor2 theorized that the greater diffusion
rate of the respective cations in 00304 compared to Ni o.995 0 is responsible
for the higher rate of oxidation of cobalt, and this does appear reasonable
for preponderant 00304 formation. The formation of an additional exterior
00 20 3 layer in oxidation in air is possible only below 350°0. Further investiga-
tions may be required in order to explain the oxidation mechanism in the
00304-phase region. Vallee and Paidassi 3 studied the oxidation rate of 000
to 00304 in ail' between 700 and 900°0 and found that the kinetics is rather
complex. Further clarification of the mechanism of the oxidation of 000
to 0030 4 may be desirable.
1l\Ioorc, W. J., and J. K. Lee: J. Chern. Phyy. 19, 255 (1951).
2 Tichenor, R. L.: J. Chern. Phys. 19, 796 (1951).
3 Vallee, M. G., and M. J. Paidassi: Corrosion et anti· corrosion 10, 132 (1962).
4.2. Scaling Systems with Electron-Conducting Protective Layers 197

The current results of the previously mentioned authors, as well as


the experimental results of Dunn 1 and those of Preece discussed at the close
of this section, still do not permit any statement about the mechanism of the
oxidation. The oxygen diffusion through the CoO layer proposed by Preece,
Valensi, and Arkharov appears slightly probable on basis of the p-type
lattice defect structure, insofar as there are no pores in the layer. The most
complete experimental results at this time on the mechanism of the C02+
diffusion through CoO and on cobalt oxidation were published by Carter and
Richardson. 2 Since these experimental results are of general significance and
an entirely new problem-that of the distribution of the cobalt-ion vacancies
in the oxide layer-can be discussed, a somewhat detailed consideration is
justified.
The data of Carter and Richardson 3 plotted in Fig. 82 for the temperature
dependence of the oxidation rate are supplemented with data from measure-
ments between 1000 and 1350°C by other authors. The scatter of the measured
values at lower temperatures is probably caused by the simultaneous
appearance of a C03 0 4 phase in the outer part of the scaling layer. According
to Fig. 83, the oxygen-pressure dependence of the oxidation rate of cobalt

----
I

-
"
2
V I--
~

Fig. 83. Oxygen·pressure dependence for '~


the rate of oxidation of Co to CoO at
1148°C, according to Carter and Richard-
son, kH '" Po. 0.29.

at 1148°C follows the 0.29th power of the oxygen pressure, which is in good
agreement with the value of 0.30 for the rate of diffusion of the cobalt-60
ions, thus confirming the applicability of the Wagner scaling formula
for this case. On basis of a practically unique diffusion ofCo2+ ions (D~ = D~~)
and D; = D~ ~ 0) the oxidation rate constant can be calculated from the
self-diffusion coefficients for the present case from the scaling formula con-
verted from (4.10) and (3.22)

(4.65)
1 Dunn, J. S., and F. Wilkins: "Review 9fthe Oxidation and Scaling of Heated Metals,"
II, The Oxidation oj Nonferrous Metals, London, 1936, p. 67.
2 Carter, R. E., and F. D. Richardson: J. Metals 7,336 (191i5), with a theoretical supple-
ment by C. Wagner.
3 Carter, R. E., and F. D. Richardson: J. Metals 6, 1244 (1954).
198 4. Scaling Processes with Formation of Thick Protective Layers

at various temperatures and an oxygen pressure of 1 atm. The good agree-


ment between the calculated and experimentally obtained practical scaling
constant can be seen in Table 31.

Table 31. Comparison of the Practical Scaling Constants in the Oxidation of


Cobalt to CoO in Oxygen at 1 atm Which Were Obtained Experimentally and
Calculated from Self-Diffusion Coefficients from Equation (4.65), According to
Carter and Richardson

k" X 108 , % deviation


g2jcm4 -sec of the calc
k X 10 9 ,
T,oC equiv/cm-sec calc obs and Ob8 k"

1000 1.25 2.72 2.43 +11


1148 5.15 10.5 9.3 +11
1350 31.25 68.2 78.2 -13

This agreement can always be expected when we are dealing with pure
diffusion-controlled scaling processes. The Wagner formula makes no state-
ment-and it does not need to-about the distribution of the lattice defects
or the diffusion of ions through the oxide layer which is formed. As was
explained in Section 3.2, the Wagner formula is not concerned with the
special conditions which are always present in the boundary layers in the
neighborhood of the phase boundaries, even at high temperatures, in which
we must always deal with rapid field transport processes and diverging
concentrations of lattice defects, since the diffusion processes appearing
simultaneously in the neutral Wagner zone (which here makes up by far
the largest part of the scaling layer) determine the rate of the total reaction.
Information on this special behavior in the boundary layers at the
Co/CoO and CoO/0 2 phase boundaries can be expected from the analysis of
the isotopic distribution of cobalt-60, which appears in the oxide layer
during oxidation. The variation in concentration of the isotope obtained by
Carter and Richardson using a refined layer removal method is given in
Fig. 84. As can be seen from the curve in Fig. 84, the concentration of the
isotope is too large in the neighborhood of the CoO/0 2 phase boundary and
too small in the neighborhood of the other phase boundary. However, a course
of just this type is to be expected, if one considers the special conditions in
these space-charge layers.
As was already discussed in detail above, the oxidation process starts with
a chcmisorption of oxygen. Then it works together with the am bipolar diffu-
sion to increase the hole concentration and decrease the concentration of metal
ion vacancies (see Figs. 25 and 28). However, this reduction in the vacancy
concentration produced by the space charge can only be effected by an
4.2. Scaling Systems with Electron-Conducting Protective Layers 199

increased flow of cobalt ions, in this case, primarily isotopes. Thus, we must
have an enrichment of cobalt isotopes in the neighborhood of the surface,
which is in agreement with experimental conditions.

Boundary
Layer wich l
Posilive I
Space I
Charge I
9 r-----~--+_~~~~~-+--------r-_+~
I
I
I
I
I
I
I
I
2~----+_-1------~~~-------+--+-4

Fig. 84. Concentration of C0 60


ions in the CoO layer, ac·
cording to Carter and Richard·
son. (One can divide the oxide
layer into three zones according
to concentration.)

In the neighborhood of the Co/CoO phase boundary, the conditions are


complicated insofar as we presently cannot determine whether the too-low
concentration value of cobalt-60 is caused by an increased CoD" formation,
as is indicated in Fig. 85, or by complete filling of the vacancies, connected with

Fig. 85. Schematic representation of the concentra-


tion of cobalt ion vacancies ncoo" and electron
holes in the CoO layer during the oxidation of
cobalt. While a positive or negative space charge
appears in the boundary layers, the Wagner zone
is electrically neutral. (Therefore the diffusion
coefficient is constant, i.e., independent of location,
only in the Wagner zone.)

a lattice defect inversion (CoO·· and 8) with the regular cobalt ions, which
are practically the only ones present after long oxidation times. No matter how
the space-charge layer mechanism at the inner phase boundary may appear
in detail, a further decrease in the isotope concentration (dilution effect)
occurs. If the arguments above are accurate, the space-charge layer (although
narrow) should appear spread out with decreasing oxidation temperature,
which probably causes a broadening of the Wagner zone (Fig. 85), on one
200 4. Scalin~ Processes with Formation of Thick Protective Layers

hand, and large concentration enrichments and depletions, that is, larger
deviations in the concentration gradients of the isotope in the space-charge
layers, on the other. With increasing temperature, especially above 1200°C,
the space-charge layer effects should always become insignificant, so that
we finally have a practically location-independent concentration gradient
(constant D) through the entire oxide layer. The approach to this ideal
state will vary from oxide to oxide and could perhaps be attained in a simpler
way in the CU/CU20/02 and Fe/FeO/02 systems.
Only the concentration gradient in the interior of the oxide layer-
after "cutting away" the two space-charge layers-is a determinant in the
Wagner theory, and this gradient is linear with the distance, as was assumed
in earlier calculations. The further observation that the cobalt isotope in a
CoO layer on cobalt metal diffuses inward much more slowly than in a CoO
layer which is isolated from the metal, and which is in equilibrium with the
respective oxygen atmosphere, is evident from the above discussion, since
under these experimental conditions (total equilibrium through the entire
crystal) no space-charge layers are found.
It is apparent from these considerations that such a concentration-
distribution analysis of the isotope in scaling and tarnishing layers is a new
and useful method to further experimental investigations of the breadth
and intensity of space-charge boundary layers. Further work in this direction
-with cobalt as well as with other metals-appears to be desirable for
clarification of the oxidation mechanism, since it is of general significance.
Pcttit and Wagner l have investigated the oxidation of cobalt in pure
CO 2 and in CO/C0 2 mixtures at a total pressure of 1 atm between 920 and
1200°C. After a short incubation period, the oxidation follows a linear rate
law. When the thickness of the CoO layer has reached about 5 x lO-3 cm,
a parabolic rate law is observed. The linear increase of the oxide layer is
proportional to the mole fraction of CO 2 in the gas phase, and the rate-
determining step is the dissociation of the C02 molecules in CO and chemi-
sorbed oxygen atoms. Besides a compact outer layer, a porous inner layer
was observed. Obviously, the oxidation is influenced by the pores in the
inner zone of the oxide layer, which shall be discussed further in Section 4.4. 2
The rate of oxidation in air of the purest available nickel, cobalt, and
niekcl-cobalt alloys was investigated between 800 and 1400°C by Frederick
and Cornet. 3 As can be seen from Fig. 86,the rate of oxidation hardly changes
up to 10% cobalt. The oxidation rate increases only with an increase in the
cobalt content, and then continuously approaches the value for pure cobalt.
1 Pettit, F. S., and J. B. Wagner, Jr.: J. Metals 13, 673 (1961).
2 Nizhel'skii, V. F., and M. G. Vladimirova: Trudy Leningrad. Lesotekh. Akad. im S.M.
Kirova 92, 3 (1961).
3 Frederick, S. F., and 1. Cornet: J. Electrochem. Soc. 102, 285 (1955).
4.2. Scaling Systems with Electron-Conducting Protective Layers 201

The lack of influence of small additions of cobalt on the oxidation rate of


nickel is understandable if one considers that most of the cobalt ions entering
the scaling layer are divalent, but nevertheless, the vacancy concentration
in the oxide crystals, which determines the rate of oxidation, remains
unchanged. ~
g2· cm.i sec1
f. I
fOOl"C
1 Atm Air
I
I
/
V
Fig. 86. Dependence of the rate of oxidation of nickel- .--c V
cobalt alloys on composition at 1000°C in air at 1 o Zfl I/(J IilJ Btl 100
atm, according to Frederick and Cornet. %Co-
Finally, a few investigations into the oxidation rates of cobalt and cobalt
alloys by Preece and Lucas 1 may be mentioned. These authors observed an
oxidation rate that decreased (by about a factor of four) in the range of 950
to 1000°0, and then increased with increasing temperature. The scaling
layer which was formed in air consisted of an exterior hard, adherent 000
layer 'with an interior brittle 000 layer. Further, the oxidation was in-
vestigated with a few other cobalt alloys. The addition of 5 or 10 wt. % Mo
caused a decrease of the oxidation rate between 800 and HOOoe by about
a factor of two. At 800 e a 00-40% Ni alloy exhibits an oxidation
0

rate about 20 times smaller than that of cobalt; this increases again with
increasing temperature. At llOOoO the oxidation rate of the alloy is smaller
by only a factor of three. A 00-5% Al alloy displays similar behavior. The
scaling layer which is formed consists externally of 000 and internally of a
blue spinel, probably of the composition 00A1 2 0 4 . The scaling layers formed
above 1000°0 remain adherent after cooling. The resistance to oxidation
of a cobalt-chromium alloy with 25-30 wt. % Or is noteworthy, and addi-
tions up to 22% Or do not effect any increase in the oxidation rate (Fig. 87).

4.2.2. Scaling Systems with n- Type Conducting


Protective Layers
Systems with n-conducting protective layers are formed during oxida-
tion on zinc, cadmium, titanium, zirconium, and their alloys. In addition,
1 Preece, A., and G. Lucas: J. Inst. Metals 81,219 (1952/53).
202 4. Scaling Processes with Formation of Thick Protective Layers

we probably have to deal with the formation of an n-conducting protective


layer in the oxidation of the following metals: molybdenum, tungsten,
vanadium, tantalum, niobium, uranium, cerium, and thorium. The scaling
systems with n-conducting protective sulfide layers, which must likewise
be grouped together here , shall be treated in a special chapter, since owing
to the presencc of additional influences, the mechanism can be more com-
plicated.

o.~O I-I-II----+-I-- 1

OA? ~~--------~---------r----1
Fig. 87. The influence of the chromium
content on the rate of oxidation of
cobalt-chromium alloys at different
temperatures in air, according to
Preece and Lucas. (The amount of
oxide formed after 50 hr is plotted on
the ordinate.)

4.2.2.1. Rate of Oxidation of Zinc and Zinc Alloys


If zinc is oxidized at 400°0 and various oxygen pressures, then one
observes first of all the appearance of a compact ZnO protective layer and,
secondly, that the oxidation rate of thicker ZnO layers (more than 5000 A)
is independent of oxygen pressure. According to oxidation experiments by
Wagner and Grunewald 1 on zinc at 390°0 the scaling constants k" at an
oxygen pressure of 1 atm or at 0.022 atm are on the average 0.72 x 10-10
to 0.75 X 10- 10 g2jcm 4 -hr. At sufficiently high temperatures and for thick
ZnO protective layers, where according to the Wagner scaling theory only
diffusion processes due to the prevailing gradient of the chemical potential
of zinc are significant, the observed oxygen-pressure independence is under-
standable (Fig. 88) . Using (3.21) and (~.16a) or (2.16b), we obtain the
following expression after a conversion, as was demonstrated in the deriva-
tion of the rate formula (4.30) for copper oxidation:

k 6RT
= --tznx(p
4 F2 °2
~1) (:(J;1 YJdl)
-
pet)
- -
pIa)
(4.66)
02 02

1 Wagner, C., and K . Grunewald: Z. physik. Chem. (B) 40, 455 (1938).
4.2. Scalin~ Systems with Electron-Conductin~ Protective Layers 203

where according to earlier notation tZn is the transference number of the


zinc ions, and X(Po.~l) denotes the electrical conductivity of ZnO at
Po. = 1 atm. As can be seen from Fig. 88 and the lattice defect structure
of ZnO, the concentration of zinc ions in interstitial positions at the ZnjZnO

Zn.

Fig. 88. Concentration of zinc ions in


interstitial lattice positions through the
ZnO layer, according to Wagner. Lines
1 and 2 show the concentration decrease
of the zinc ions in the interstitial lattice
positions at low and high oxygen pressures.
As can be seen, (dczno-;d~h;:::; (dczno-;d~l2. Position Coordinate g__

phase boundary is considerably greater than at the ZnOj02 phase boundary.


Furthermore, it follows from the heat of formation of ZnO that Po(i) ~ p(oa)
.) • I.
and thus Ijpg, ~ Ijp~; (also the sixth root), so that the second term in the
parentheses of (4.66) can be neglected. Therefore, equation (4.66) simplifies
to

k = ~ R_T XZn(Po ~l) 6 1_.__1_ _ (4.67)


2 F2 ,...j p 3~ = const
In (4.67) the radical at constant temperature is a constant value and thus
k is independent of oxygen pressure.
Since the experimental determination of zinc ion conductivity required
for the calculation of k is not simple, it appears to be expedient to draw
upon the self·diffusion coefficients of zinc in ZnO in the calculation of the
scaling constants. The experimental determination is more easily accessible
owing to the progress in diffusion measurement techniques utilizing radio-
active isotopes. For this purpose we use the expression given earlier (3.26),
which after rewriting becomes
c~G~o.)
_
k - (1 + zZn) zZn cznDzn
*(i> (
1- -(1)-
CZn o'
(4.68)

where ZZn is the valence or the number of equivalents, CZn the total concen-
tration of the zinc ions in ZnO, and DZ*!i)
,n the self-diffusion coefficient of the
zinc ions in ZnO, which occur in equilibrium with zinc. For higher oxygen
pressures, the parenthetical expression equals one because ci~~o' ~ c~anb ;
204 4. Scaling Processes with Form.ation of Thick Protective Layers

as was mentioned above the self-diffusion coefficient of zinc in ZnO m


equilibrium with zinc is necessary for the evaluation. Moore and Williams,!
however, have demonstrated that the diffusion of zinc in ZnO is not con-
trolled by defects associated with excess zinc because the self-diffusion
coefficient is of the same order of magnitude in atmospheres of zinc and in
atmospheres of oxygen. The following expression for the self-diffusion
coefficient of zinc may be used:
D~n = 1.3 X 10- 5 exp( -43,500 ± 1l,000/RT) cm 2 jsec
(for 1000 to 1200°C)
if an extrapolation is valid from 1000°C to 400°C. Th'is is, of course, open
to question. A notable feature of the results is the roughly constant value of
the self-diffusion coefficien·t of zinc, independent of the nature of the ambient
atmosphere. If the excess zinc concentration in ZnO determined by equi-
librium (2.13a) is present interstitially, then at 1000°C the value of the
mole fmction Xzno' for zinc oxide in air should be about 104 times smaller
than for zinc oxide at a constant zinc vapor pressure of 1 atm. Furthermore,
one should expect a lower activation energy associated with the self-diffusion
coefficient of zinc in ZnO in equilibrium with zinc vapor. Since these effects
are not observed in marker experiments, the diffusion of radioactive zinc
cannot be controlled by excess zinc present interstitially. From the very
low diffusion coefficient of oxygen, a mechanism based on thermally produced
Frenkel defects
Null~ZnO' + ZnO'
IS suggested by Moore. More experiments are necessary to explain this
phenomenon. Mathematical evaluation 2 of the self-diffusion measurements
carried out by Moore 3 during zinc oxidation is considerably more complicated.
As Moore and Lee 4 were able to show, the oxidation rate of zinc in
thinner oxide layers (less than 1000 A) is proportional to the logarithm of
the oxygen pressure (see Fig. 33). This dependence, however, was
observed only as long as the rate of oxidation depended on the field transport
of the zinc ions through the oxide layer. Vernon et al. 5 investigated the
oxidation rate of zinc at lower temperatures (25-400°C), and found that
below 300°C the parabolic rate law was replaced by a logarithmic one.
Figure 89 shows the oxidation with time at lower temperatures. With very

1 Mooro, W. J., and E. L. 'Williams: Discussions Faraday Soc. 28, 86 (1959).


2 Wagner, C.: "Diffusion and High Temperature Oxidation of Metals," in Atom Move-
ments, ASM, Cleveland, 1951, p. 153.
3 Moore, W. J.: J. Electrochem. Soc. 100, 302 (1953).
4 Mooro, 'V. J., and J. K. Loe: Trans. Faraday Soc. 47, 501 (1951).
5 Vernon, 'V. H. J., E. J. Akeroyd, and E. G. Stroud: J. Inst. ~Metal8 65, 301 (1939).
4.2. Scalin~ Systems with Electron-Conductinl1 Protective Layers 205

thin layers of 20--40 A, electron tunneling can be regarded as the


slowest step, just as in the low-temperature oxidation of nickel discussed
earlier, and explored by Mott in 1939.1 Since the logarithmic rate law was
a$ rr----,-----r---~-----r--~
g/cm 2
a8 ~----1_----~--_+----~~~

'"~
. O.~~----~~~~~_+~~~~~

~It>-
a2 ~~~<b~~~----+_----~~~

Fig. 89. Oxidation of zinc in air


between 25 and 275°0, according to
Vernon and co·workers. (Here the
weight increase is proportional to the 1.0 f.~ 1.8 i'.2 2.0
logarithm of the oxidation time.) log t ( t In hours) ___

also observed with considerably thicker ZnO layers (greater than 1000 A)
another mechanism must be the determining factor for the layer growth.
Cavity formation and recrystallization phenomena, which were noted in
Section 3.5.5, are the first possibilities to be considered.
As can be seen from equations (2.17a) and (2.1Sa), the concentration
of mobile zinc ions in interstitial positions increases with the introduction
of Li+ ions into the ZnO layer and correspondingly decreases with the
introduction of A13+ ions. This fact is of tremendous significance for the
oxidation rate of zinc alloys. On the assumption that the alloying metal
enters the ZnO layer in ionic form in sufficient quantities, the oxidation
rate must decrease in a zinc-aluminum alloy and increase in a zinc-lithium
alloy.
By combination of the mass-action law following for constant oxygen
partial pressure from (2.15)
xz no ' xe = K (Po, = const)
with the corresponding electroneutrality relationships
xe = xzno ' +
xAle'(Znl for (ZnO + A1 20 3 )
and
XZno · = xe + XLie'(Znl for
we obtain by completely identical derivations as were given earlier for the
halogenation of silver alloys, the ratio of the oxidation rate constants of a
zinc alloy, k, and of pure zinc, kO:
1 Mott, N. F.: J. Inst. Metals 65, 333 (1939).
206 4. Scaling Processes with Formation of Thick Protective Layers

(a) For the case of a zinc-aluminum alloy:


1

- -+ -
XAI
2x~no'
- +1 ]1/2
[(
2x~no'
XAl)2 (4.69)

(b.) For the case of a zinc-lithium alloy:

-=--+
k
kO
XLi
2x~no'
- - + 1] 1/2
[ (
2x zno '
XLi)2
(4.70)

where XLi == xLi.'(Zn) and XAI == XAl.·(Zn). The concentration of the zinc
ions in interstitial lattice positions x~nO' in pure ZnO may be estimated
from conductivity measurements by Hauffe and Vierk1 at 400°C as
4 X 1O-5(x~no' = x~). Accordingly, at smaller alloying additions of 0.1 at. %
XAI or XLi ~ x~nO" From the above conditions the corresponding approximate
expressions follow from (4.69) and (4.70):

(4.69a)

(4.70a)

Corresponding to (4.67)

(4.71)

The above relationships were confirmed from oxidation experiments with


zinc-aluminum and zinc-lithium alloys at 390°C in air (Fig. 90).2
If we regard the scaling constants for spectroscopically pure zinc found
by Gensch and Hauffe 2 with kO = 8 X 10-10 g2jcm4 -hr as correct and

8J
QI· l, I
(QHt%l
,
70
I
1M
... $(I
.... \;!

~
, 'l1
_ 30

ZIi Fig. 90. Oxidation of zinc alloys as a function


10
of the alloying partners (lithium, thallium,
and aluminum) and their concentrations at
Q
i/O
390°C in 1 atm of air, according to Gensch
'0
and Hauffe.
t-

1 Hauffe, K, and A. L. Vierk: Z. physik. Chern. 196, 160 (1950).


2 Gensch, C., and K. Hauffe: Z. physik. Chern. 196, 427 (1950).
4.2. Scalinl1 Systems with Electron-Conductinl1 Protective Layers 207

calculate to a first approximation (because of the poor reproducibility of


the measured results with pure zinc) with k"o ~ 109, then we must obtain
for the scaling constant k" of a zinc-aluminum alloy with 0.1 or 1.0 at. % Al
xO 4 x 10-5
k" = k"o znO· ~ 10-9 = 4 X 10-11
XAI 10-3
or
4 X 10-5
k" ~ 10-9 - - - =4 X 10-12 g2Jcm4 .hr
10-2
The experimental value of k" was found to be about 1 x 10-11 • Oxidation
experiments on zinc-lithium alloys with 0.4 atm. % lithium yield
k" ~ 2 X 10-7 g2Jcm4 -hr
and k" was calculated at about 1 x 10-7• These values of the scaling constants
are clearly greater than the values given by other authors. These values,
unlike other meMls which form p-conducting protective layers, are not a
sign of impurity, but rather an indication of higher purity, as will be shown
in more detail below.
Oxidation experiments carried out simultaneously on zinc-thallium
alloys yielded an increase in the oxidation rate at low-thallium content
and a decrease at high-thallium content. Whether introduction of mono-
and trivalent ions causes this course of the oxidation cannot be determined
at this time. In any case, the numerical data for the oxidation of zinc alloys
have to be used with caution, since the impurity oxide content of the oxide
layers was not known when they were compiled. According to Wagner,1
the decrease of the oxidation rate of a zinc-aluminum alloy may be caused
by formation of a scale of Ah03 rather than by a decrease in the concentra-
tion of interstitial zinc ions in ZnO, since ZnO in a zinc-aluminum alloy is
thermodynamically unstable because of the highly negative L1F value of
the displacement reaction (see Section 4.5.2):
3ZnO + 2AI (soluted in Zn) = Al 20 3
Only ZnO could be detected by X-ray diffraction measurements of the scale
scraped off the surface of the oxidized zinc-lithium-aluminum alloy with
2 at. % Li and 0.52 at.% Al.2
Finally, we shall consider why the scaling constants for zinc found by
Gensch and Hauffe permit us to draw conclusions about especially pure
zinc. While, for example, in nickel, cobalt, and iron the usual metallic
impurities, such as aluminum, iron, chromium, etc., cause an increase in the
1 Wagner, C.: J. Electrochem. Soc. (in press).
2 Hauffe, K.: unpublished results.
208 4. Scaling Processes with Fonnation of Thick Protective Layers

oxidation rate, with zinc and certainly with cadmium these same "impurities"
cause the scaling rate to decrease. This experimental observation which was
previously not understood may now be interpreted from the above con-
siderations. While the "impurity" metal enters the oxide layer as a trivalent
ion during the oxidation of the first group of the metals (nickel, cobalt,
copper, iron) and increases the concentration of the cation vacancies and
thus the oxidation rate, in the second group of metals (zinc, cadmium) the
introduction of these higher-valent impurity metal ions decreases the
concentration of metal ions in interstitial lattice positions or of oxygen ion
vacancies, and thus the oxidation rate.
Recently, the kinetics of the oxidation of molten zinc were measured
by Copel for the temperature range 440-700°C at an oxygen pressure of
200 mm Hg (see Fig. 91). The measurements conform to a parabolic law
and the rate constant is given by
k = 2.2 X 10-5 exp( -1.08jkT) cm2 jsec
where the activation energy is expressed in electron volts. This result is
in fair agreement with the calculated values of the rate constant obtained

-10

H
1
U

f'" -II
E
"u
~
~
]' -12
t
Fig. 91. Temperature dependence
-13~~--~~~~~ of the oxidation rate of liquid zinc
1.04 1.14 1.24 1.34 1.44 at 200 mm Hg oxygen between 440
->- 103jTOK and 700°C, according to Cope.

by Thomas 2 from measurements of the solubility and diffusion coefficients


for excess zinc
k = 0.44 X 1O~5 exp( -1.20jkT)
although experimental errors must be taken into account. An extrapolation
of the results of Cope's measurements to 390°C gives a value of
3.1 x 10-13 g2jcm 4 -sec, which agrees 'with Hauffe's result of 2.2 x 10-13
1 Cope, J. 0.: Trans. Faraday Soc. 57, 493 (1961).
2 Thomas, D. G.: J. Phys. Chern. Solids 3,229 (1957).
4.2. Scaling Systems with Electron-Conducting Protective Layers 209

g2jcm 4-sec. The oxidation process for liquid zinc is therefore consistent with
the solid-zinc oxidation in which transport is controlled by the diffusion
of defects associated with excess zinc.
The same mechanism should apply to the oxidation of cadmium,
becAuse OdO is also an n-type conductor probably with oxygen ion vacancies.
Bouillon and Jardinier- Offergeld 1 recently investigated the nucleation of
oxide particles during oxidation.

4.2.2.2. The Rate of Oxidation of Titanium and Titanium Alloys


Titanium and its alloys must be included in the group of metals which,
like zinc, form n-conducting oxide layers. It has good mechanical and
technological characteristics even at higher temperatures (up to about
550°0) and a relatively small specific weight. Alloying techniques and
mechanical-technological characteristics are being investigated as a function
of temperature and gas atmosphere. The study of the effect of the gas
atmosphere is of special significance because of a characteristic of titanium,
namely, its capacity to dissolve gases, especially hydrogen, oxygen, and
nitrogen. Gas solubility severely limits the use of titanium and titanium
alloys at higher temperatures if one does not succeed, through some techno-
logical device, in destroying this undesirable property, which is characterized
by an increase in hardness and a decrease in ductility.
In the following we will concern ourselves in some detail with the
mechanism of the effect of oxygen and nitrogen on titanium and a few of
its alloys at higher temperatures. 2 Right at the beginning it may be
emphasized that the greatest part of earlier work on the attack by oxygen
represents experiments of short duration only, which frequently are worth-
less for the evaluation of titanium's resistance to oxidation, as we shall see
below. Further, at this time there are still no investigations on the "internal
oxidation" which can take place especially with titanium alloys, for which
the oxide of the alloying metal possesses a high negative heat of formation.
The thermodynamic data of individual oxides-TiO, Ti 20 3, Ti30s, and
Ti0 2-which can appear during oxidation of titanium were reported by
Kubaschewski and Dench. 3 X-ray measurements on Ot-Ti and TiO were
carried out by Rostoker.4 Further thermodynamic data were published
1Bouillon, F., and M. Jardinier-Offergeld: Compt. rend. 252, 1470, 2566 (1961).
2 An unfortunate, but significant, explosion in the corrosion laboratory of the College
Park, Maryland, station of the Federal Bureau of Mines, U.S.A., with titanium sheets
which were treated with fuming nitric acid, indicated that titanium and its alloys are
potentially dangerous with red, fuming nitric acid or with other strongly oxidizing fluids
and vapors. Therefore, the necessary security measures must be taken with such fluids
[for details see Corrosion 11, 86 (General News) 1955].
3 Kubaschewski, 0., and W. A. Dench: J. Inst. Metals 82,87 (1953/54).
4 Rostoker, W.: J. Metals 4, 981 (1952).
210 4. Scaling Processes with Formation of Thick Protective Layers

on the titanium-oxygen system by Groves, Roch, and Johnston} The


Ti-Ti0 2 phase diagram is given for a general orientation in Fig. 92 as it was
constructed from the data by DeVries and Roy.2 Phase analysis studies on
the titanium-oxygen system have been carried out by Magneli et al. 3

~r---------------------------------~--------~
°C

.x-:Jl
solid ~
solut ion
Mo 55 '" ],
+ '"
11055
0
F '"'"
~

~
0- nZ D] 55
""N ~

riDl 55

.
(rutile)
0
(Z-TI 55

00
1
F

I \
()
Ti
'fU (J.¥ (J.6' 1M 1.()
Atom Rotio OITi
1.Z

Fig. 92. Possible phase diagram for the Ti-Ti02 system, represented by DeVries
-- 1.iI 1.G

and Roy from data available in the literature at the time (88 denotes solid
solutions) .

Experimental results on the kinetics of oxidation and on the structure


of the scaling layer lead us to believe that the oxidation mechanism is quite
complicated. While at intermediate temperatures (up to 800°C) the scaling
layer appears to consist almost entirely of Ti0 2 ,4 above 800°C-especially
after long oxidation times-TiO formation in the immediate neighborhood
of the metal phase, as well as a Ti 20 3 layer in the middle region of the scaling
layer, were observed in addition to Ti0 2 . 5 This oxide series, as it was also
found in iron oxidation, is clear and to be expected on thermodynamic
1 Groves, W.O., M. Hoch, and H. L. Johnston: J. PhY8. Chem. 59, 127 (1955).
2 DeVries, R. C., and R. Roy: Arnl. Ceram. Soc. Bull. 33, 370 (1954).
3 Andersson, S., B. Collen, U. Kuylendstierna, and A. Magneli: Acta Chem. Scand. 11,
1641 (1957).
4 Hickman, J. \V., and E. A. Gulbransen: J. Anal. Chem. 20, 158 (1948); A. E. Jenkins:

J. Inst. Metals 82, 213 (1953/54).


5 Arkharov, V. 1., and G. P. Luschkin: Ber. Akad. Wi8S. SSSR 83, 837 (1952).
4.2. Scalin~ Systems with Electron-Conductin~ Protective Layers 211

grounds. The thickness of this oxide region depends on the rate of formation
of the individual oxides-that is, on their lattice defect structure and
associat ed i onic diffusion . Morton and Baldwin! divided the structure of
the scaling layer into three temperature regions on basis of microscopic
and X-ray investigations (Fig. 93):

%7&/,/&11
Tt
m , ,Ro';!"
- Bronze Colored

7/V---0tl? 'c

iiiiiil::~
~ ScalingBrown)
' --Rutlle (Yellow Layer
(Light Blue)
(Dark Blue )
1..'1~<:!"- II ",;/A
+ some Ti (Metallic)
.. no - Powder
&!S··-BS() °C (Long Time)

~~~~~22t-
;;
Rutile (Dark Blue)
Ti ~ 03 (fine Purple-Colored Powder)

Fig. 93. Diagram of the scaling - TiO l Metallic


- TiO+ Some Ti J Appearance
layer which is formed on titan- Ti + TiO - Powder
ium at different tempera tures
and oxidation times, according
to Morton and Baldwin. 87S··· 7()SfJ °C (Long Time)

1. An exclusive Ti0 2 formation with a blue coloring was observed up


to 800°C with oxidation times that were not too long.
2. Between 825 and 850°C a blue-colored Ti0 2 layer also formed after
a long period of oxidation in the exterior part of the scaling layer.
A TiO layer, which could be identified with relative certainty due
to its NaCI structure (lattice parameter a = 4.19 A), formed between
this blue layer and the metal. Because of the considerable mutual
solubility of the TiO and Ti phases, this part of the scaling layer is
relatively loosely attached to the titanium and thus the initially
good adherence of the oxide layer disappears.
3. Between 875 and 1050°C, a Ti 2 0 3 formation, which was relatively
thin compared t o the Ti0 2 and TiO layers, was observed after long
oxidation times. Again the TiO layer is frequently primarily respon-
sible for the poor adherence of the scaling layer.
The dark-blue color of the exterior Ti0 2 layer was probably caused by
captured conduction electrons in oxygen ion vacancies 00·· on cooling

1 Morton, P. R., and W. M. Baldwin: Trans. ASM 44,1004 (1952).


212 4. Scaling Processes with Formation of Thick Protective Layers

to room temperature, similar to NaCI crystals turning blue, according to


Pohl and Schottky.1 Based on the lattice defect arrangement in Ti0 2
Null..------ 00·· + 28 + !02 (4.72)
we can formulate the capture process in the following way:
00·· + e..-- 00· (4.73)
blue F -center
To what extent the blue coloration can also be caused by the capture process
(4.73a)
is at present difficult to determine. 2
Buessem and Butler 3 as well as Kofstad 4 have investigated the defect
structure of rutile (Ti0 2 ) by a gravimetric method, in the temperature
range 1200-1500 K. The relative, isothermal weight change of rutile equi-
0

librated in mixtures of CO and CO 2 has been found proportional to PO;/6


at oxygen vacancy concentrations larger than approximately 1019 cm-3 ,
in agreement with equation (4.72). Below this vacancy concentration the
weight change is approximately proportional to PO!/2. This transition takes
place at values of Xoo·· of6 x 10-4 to 10-3 (see Fig. 94). Close to stoichiometry,
intrinsic ionization
Null~ e + EB
will predominate thus giving Xoo·· '" p~~/2. Forland 5 found xoo .. '" PO;/6
above llOO°C in the oxygen-pressure region of 10- 7 to 10-16 atm.
The present difficulties in the explanation of the oxidation mechanism
are understandable in view of the complicated structure of the scaling
layer and the complete lack of quantitative diffusion data for the titanium
and oxygen ions in the individual oxide layers. It can be added that different
rate laws were found for oxidation in different temperature regions. Alexander
and Pidgeon 6 reported that titanium oxidation may be described between
25 and 330°C at oxygen pressures of from 20 to 200 mm Hg by a logarithmic
rate law of the form
In(t +
3) = kl V +
k2

where t is the experimental time, v is the total oxygen volume which was
consumed, and 3 is an arbitrary numerical factor which probably depends
1 According to K. Hauffe: Ergeb. exakt. Naturw. 25,193 (1951).
2 See for example T. Hurlen: Acta Chem. Scand. 13, 365 (1959).
3 Bucssem, W. R., and S. R. Butler: "Defect Reactions in Ti0 2 ," in Kinetics of High.
Temperature Processes, Endicott-House Conference, MIT, Cambridge, 1959, pp. 13ff.
4 Kofstad, P.: J. Phys. Chem. Solids 23, 1579 (1962).
5 Fl1lrland, K. S.: Acta Chem. Scand. 18, 1267 (1964).
6 Alexander, W. A., and L. M. Pidgeon: Canad. J. Res. (B) 28,60 (1950).
4.2. Scaling Systems with Electron-Conducting Protective Layers 213

on pretreatment of the titanium sample, but, in contrast to kl and k2, is


independent of the experimental temperature. Waber,l on the other hand,
found a cubic rate law for oxidation of pure titanium sheets at 216°0, which
has been noted earlier (Section 3.5.3) and recently confirmed by other
authors.2

I1U ' (

·l

'Io~( ,'n·c
.
\

~ \

,
~

\ \
\ \ \
'If)1-,
\
.~ /-
"12 \
--lor -P'z \ 911"
"

' 18 ' I~ ' 10

Fig. 94. The logarithm of the mole fraction of the oxygen


vacancy concentration in Ti02, xoo", as a function of
log Po" according to Kofstad.

Gulbransen and Andrew 3 evaluated their results obtained from oxida-


tion experiments between 250 and 600°0 according to the parabolic rate
law. However, the curves presented in Fig. 95 show-especially in the
initial oxidation period-considerable deviations from this. Kofstad and
Hauffe 4 were able to demonstrate that uniform straight lines can be obtained
with a slope which corresponds to the cubic rate law on a double logarithmic
plot of the weight increase V8. time. The break in the log k" - liT lines
observed by Gulbransen and Andrew at 350°0 frequently characterizes the
transition from the cubic to the logarithmic rate law. This brings the results
of Gulbransen and Andrew into good agreement with those of other authors.5
However, after longer oxidation times a transition to theparabolicratelawwas
observed (similar to nickel oxidation, see Section 3.5.3). The oxygen-pressure
1 Waber, J. T.: J. Chem. Phys. 20,734 (1952). In a report published later, Waber finds
a logarithmic rate law for the oxidation at 250 and 316°C. After 1200 hr, the oxide layer
had attained a thickness of 5000 A at 250°C.
2 Kofstad, P., K. Hauffe, and H. Kj611esdal: Aeta Chem. Seand. 12, 239 (1958).
3 Gulbransen, E. A., and K. F. Andrew: Metals Tran8. 185, 741 (1949).
4 Kofstad, P., and K. Hauffe: Werkstoffe u. Korrosion 7,642 (1956).
5 Morton, P. H., and W. M. Baldwin: Trans. ASM 44, 1004 (1952).
214 4. Scalin~ Processes with Formation of Thick Protective Layers

dependence of the oxidation rate of titanium at 550°C found by Gulbransen


-especially with thin oxide films-permits conclusions to be drawn
about a field transport mechanism, as was discussed in Section 3.5.3 for

/Fe
:/
/' >so··
;;;;-

/ V
l---

7V
--
~
V
500°

y/l---
~
~
llSO'
lHJOo
~
~

V/ XO· Fig. 95. Oxidation of titanium between


o 6'IJ eo Wm,n 120
250 and 600°0 in oxygen at 76 mm Hg,
according to Gulbransen and Andrew.
Time_ (Oubic rate law below 400°0.)

the oxygen pressure dependence of the oxidation rate of zinc. (Possibly


concentration-dependent oxygen diffusion in titanium?) On basis of the
n-conduction of the Ti0 2 such a dependence should not be observed in
diffusion-controlled oxidation.
One must conclude from the experimental results of Davies and
Birchenall1 in Fig. 96 that the course of titanium oxidation with time above

g·cm .$
20
/"
%O°C [;it Fig. 96. Oxidation of titanium
between 820 and 950°0 at an oxygen

7 V
pressure of 1 atm, according to
Davies and Birchenall (linear rate

V I~
law). A. E. Jenkins [J. Inst. Metal8

---
880' 82, 213 (1953/54)] found in agree-

/ V O l--- 820'
ment with this, above 800°0 in
oxygen at 1 atm, a linear rate law
with approximately the same values

o~
~ for the linear rate constant I" in
IHJO Q(J()
mg/cm 2 .min: I" = 2.4 X 104 X
exp( - 47,500/RT), according to
Time Davies and Birchenall.

820°C may be described by a linear rate law. Below this temperature, down
to about 590°C, the course of the oxidation with time may be described by
a parabolic rate law in agreement with the experimental results of Pfeiffer
1 Davies, M. H., and O. E. Birchenall: J. Metal8 3, 877 (1951).
4.2. Scalin~ Systems with Electron-Conductin~ Protective Layers 215

and Hauffe1 (Fig. 97), Jenkins,2 Richardson and Grant,3 and Kofstad,
Hauffe, and Kjollesdal 4 (Fig. 98a and b). While Richardson and Grant
found a parabolic rate law also up to 1000°0 by short-time experiments

, /
v
L
to

I
I
s /
Fig. 97. Oxidation of titanium at 800°C and
an oxygen pressure of 1 atm with and without
the presence of W0 3 , according to Pfeiffer
and Hauffe (parabolic rate law). Curve I: in
pure oxygen; Curve II: in oxygen and in
loose contact with W0 3 •
o
! 8 *
/r".&S3.f(}"

Time -
~
6 h

(Fig. 98a), long-time oxidation experiments at temperatures above 900°0


produced a linear rate law after a parabolic start (Fig. 98b). The experi-
mental findings of the latter three groups are understandable on basis
of the above qualitative structural consideration of the scaling layer. As
we saw in Fig. 93, below 800°C a blue-colored Ti0 2 layer appeared only after
oxidation times that were not too long. This layer was relatively compact,
so that under these conditions an ionic diffusion results over oxygen-ion
vacancies through the oxide layer to the metal, and with a rate-determining
diffusion leads to a parabolic rate law. Actually, here the oxygen ions diffuse
through the layer, in contrast to the metal ions which diffuse over interstitial
lattice positions, as is the case in zinc oxidation. This was expected from the
lattice defect equilibrium (4.72)5 and can be proven on the basis of marker
experiments with radioactive silver by Davies and Birchenall. 6
1 Pfeiffer, H., and K. Hauffe: Z. MetaUk. 43., 364 (1952).
2 Jenkins, A. E.: J. Inst. Metals 84,1 (1955/56).
3 Richardson, L. S., and N. J. Grant: J. Metals 6,69 (1954).
4 Kofstad, P., K. Hauffe, and H. Kjiillesdal: Acta Chem. Scand. 12,239 (1958).
5 Earle, M. D.: Phys. Rev. 61, 56 (1942); K. Hauffe, H. Grunewald, and R. Tranckler-
Greese: Z. Elektrochem. 56, 937 (1952).
6 Davies, M. H., and C. E. Birchenall: J. Metals 3,877 (1951).
216 4. Scalin~ Processes with Formation of Thick Protective Layers

~O r-----~-------'-------'-------'

mmHg

?27°C

Fig. 98a. Oxygen pressure decrease


during the oxidation of titanium in
oxygen at 0.5 atm pressure between
o 3 6 !J mIn. 12 679 and 1012°e, according to Richard.
VT- son and Grant (surface area: 13.52 cm 2 ).

105~------~--------T--------'--------~

/
~e
~

S!
. ./'
Parabolic
X
c:
~
.c ..
--
~

Time l min

Fig. 98b. Oxidation of melted and annealed (30 min at 800 0 e


and Po, < lO-4 mm Hg) van Arkel titanium in oxygen of
1 atm, according to Kofstad, Hauffe, and Kjiillesdal.
4.2. Scaling Systems with Electron-Conducting Protective Layers 217

In order to elucidate the diffusion mechanism, Kofstad et aU studied


the oxidation mechanism of titanium between 800 and 1200°0 by means
of X-ray diffraction and micro-indentation hardness studies on one hand,
and metallographic and marker experiments together with kinetic measure-
ments on the other. Two significant results were obtained.
1. In agreement with Wallwork and Jenkins 2 the initial parabolic rate
law is a result of the rate.determining diffusion of oxygen in titanium.
In terms of this interpretation, the activation energy measured during
the initial parabolic rate law (ilEA = 51 kcal/mole) may be considered
an average activation energy for oxygen diffusion in titanium-
oxygen solid solutions (Do = 0.5 exp( -51,000/RT) cm 2/sec). The
outer metal layer reaches a composition of TiO o.35 which is associated
with important structural changes. 3
2. In the marker studies between 800 and 900°0, the platinum markers
were always found on the surface of the oxide scale. At higher
temperatures, where a two-layered Ti0 2 scale was seen, the platinum
markers were invariably found at or close to the border between the
two Ti0 2 layers. In this case the burial of the markers appears to be
directly associated with recrystallization and plastic flow in the
outer layer of the scale. Besides this uncertainty in the marker
experiments, on basis of all experimental results we may conclude
that oxygen diffusion is the mechanism responsible for oxidation.

On the other hand, above 820°0 we no longer obtain adhering oxide


layers on pure titanium, since TiO and Ti 2 0 3 appear, which form as loose-
partly light powdery-layers. Then the oxygen has available sufficiently
broad grain boundaries and pores for rapid transport to the reaction front,
so that phase-boundary reactions become rate-determining. Since the rate
of oxidation is zero with respect to oxygen, a rate-determining process can
hardly come into question at the oxide/oxygen phase boundary. On the
other hand, with rate-determining reactions at the Ti/TiO phase boundary,
the oc-,B-titanium conversion4 at about 880°0 should become noticeable,
which is not to be inferred from previous experimental results. The absence
of this effect is understandable if one considers that the transition point
can be displaced to over 1000°0 by the solution of oxygen in titanium. The

1 Kofstad, P., P. B. Anderson, and O. J. Krudtaa: J. Les8-Common Metal8 3,89 (1961).


2 Wallwork, G. R., and A. E. Jenkins: J. Electrochem. Soc. 106, 10 (1959).
3 Andersson, S., B. Collen, V. Kuylenstierna, and A. Magneli: Acta Chem. Scand. 11,
1641 (1957).
4 Jaffee, R. I., H. R. Ogden, and D. J. Maykuth: TraM. AIME 188, 1261 (1950);
A. J. Williams, R. W. Cahn, and C. S. Barrett: Acta Met. 2,117 (1954).
218 4. Scaling Processes with Formation of Thick Protective Layers

explanation of all these observations is made still more difficult by the


following fact: with increasing temperature-especially above 700°C--
oxygen dissolves in titanium in considerable quantities. Furthermore, on
cooling the scaling layer, a decomposition into TiO and TiOo.35 appears in
part in the exterior, oxygen-rich, scaling layer. The TiO preferentially separ-
ates at the grain boundaries, which leads to the destruction of the initially
compact protective layer. Jenkins l succeeded in demonstrating by a vacuum
treatment between 900 and lOOO°C that the oxide layers produced on
titanium were absorbed up to equilibrium by the metal phase by the overall
reaction:
Ti02 + Ti ~ 20 (dissolved in Ti)
In agreement with Wallwork and Jenkins 2 the initial approximate parabolic
oxidation rate has been shown to be primarily associated with oxygen
dissolution in the metal. When the oxygen concentration in the outer layer
of the metal reaches a composition TiO o.35, heavy oxide formation begins to
take place and the oxidation eventually follows a linear law 3 (nucleation and
recrystallization). Morton and Baldwin found an irregular course of oxida-
tion with time between 900 and 1l00°C, where in period changes after a
certain oxidation time the rate of oxidation accelerated and then gradually
decreased again to its earlier value. On basis of experiments of long duration
(lasting for several days) Hauffe and Kofstad4 were able to confirm the
fluctuations of the oxidation time curves, which probably can be attributed
to a periodic cracking of the oxide layer. The quasi-linear rate law may also
be summarized by small parabolic branches. On this basis, nonlimiting long-
time experiments are recommended for the correct evaluation of the
resistance to oxidation of titanium and titanium alloys. Earlier investigations
which used short oxidation times should not be used for an evaluation of
this resistance.
The parabolic oxidation of titanium in oxygen at 0.5 atm which was
found by Richardson and Grant 5 between 700 and lO50°C is a typical result
of a short-time experiment. As can be seen from the positions of the points
in Fig. 9Sa, the parabolic course is evidently adhered to quite exactly,
if short oxidation times are chosen between 25 and 150 min. Frequently
in these short oxidation times the loose TiO + Ti and Ti 20 3 layers which
are probably decisive for the linear scaling are not formed. Also in these
short oxidation times one can recognize a parabolic rate law between 700 and
1 Jenkins, A. E.: J. Inst. Metals 82,213 (1953/54); see also L. G. Carpenter and W. N.
Nair: J. Inst. Metals 88,38 (1959).
2 Wallwork, G. R., and A. E. Jenkins: J. Electrochem. Soc. 106, 10 (1959).
3 Kofstad, P., P. B. Anderson, and O. J. Krudtaa: J. Les8·Common Metals 3,89 (1961).
4 Hauffe, K., and P. Kofstad: unpublished measurements.
5 Richardson, L. S., and N. J. Grant: J. Metals 6,69 (1954).
4.2. Scalin~ Systems with Electron-Conductin~ Protective Layers 219

900°C in the experimental results of Davies and Birchenall. Furthermore,


the value of the activation energy of 47.4 kcalJmole (Fig. 99) found by
Grant 1 above 700°C is almost twice as great as the value of 26 kcaljmole
.1
'sec-' V
1000
0/
Va
J
/0 /
I
/v Az
V
/
Fig. 99. Temperature dependence of the parabolic
oxidation and nitridation constants, according to
/
lW S(}(J ai'J 1tW 0 C
Richardson and Grant. Temperature_

below 600°C obtained by Gulbransen and Andrew. 2 These differing values


of the activation energies are explained by the fact that above 700°C diffusion
processes (even into the titanium phase) predominate-thus the parabolic
course-and of course below 600°C field transport of the ions through the
oxide layer or gas diffusion in the metal take over-thus the quasi-parabolic
and cubic rate laws.
Lohberg and Schleicher 3 have studied the oxidation rate of titanium
in water vapor between 600 and 1000°C. The water consumption with time
is slower than it should be according to the parabolic rate law. Figure 100
shows the water consumption after 60 and 1000 min as a function of the
temperature. Rutile was the only oxide in the layer which could be detected
by X-ray measurements.
Arkharov and Luschkin4 report that the rate of oxidation of titanium
in air above 1150°C is more rapid than in oxygen. X-ray investigations
carried out simultaneously on the oxide layers which were formed on Ti02
indicated that the lattice constants of Ti0 2 between 1100 and 1200°C which
were obtained in one case from oxidation in air and in the other from oxygen
are different from one another, and below these temperatures are equal.
The deviation of the lattice constants of Ti0 2 from oxidation experiments
1 Richardson, L. S., and' N. J. Grant: J. Metals 6, 69 (1954).
2 Gulbransen, E. A., and K. F. Andrew: J. Metals 1, 741 (1949).
3 L6hberg, K., and H. W. Schleicher: Z. physik. Chem. [NF] 15,223 (1958).
4 Arkharov, V. 1., and G. P. Luschkin: Dokl. Akad. Nauk SSSR 83,837 (1952).
220 4. Scalin~ Processes with Formation of Thick Protective Layers

in air and oxygen at 1200°0 increases with the thickness of the oxide layer.
This change in the lattice constants was also attained when Ti02 produced
in pure oxygen at 1200°0 was heated to the same temperature in nitrogen.!

10000 900 0 800 0 700 0 600 0

~
E
~
~
c:
.2 10- 2
Q.
E
."
c
13...
'"
~ 10- 3
t

10-4~ ______ ~______~________~______~~


7.5 8.5 9.5 10.5 11.5
-+ 104 fT

Fig. 100. The rate of oxidation of titanium in water vapor,


represented by the water consumption after 60 and
lOOO min as a function of the reciprocal absolute tempera-
ture, according to Liihberg and Schleicher.

Jaffee and co-workers 2 consider the introduction of nitrogen into the Ti0 2
lattice responsible for this, which according to the symbolic incorporation
equation:
(4.74)
should have caused an increase in the vacancy concentration. The increase in
the lattice constants and the rate of oxidation is understandable in terms of
this additional vacancy production. However, further experiments are still
needed to demonstrate the validity of equation (4.74).
Carpenter and Reave1l 3 compared the rate of oxidation with that of
nitride formation for titanium between 700 and 1000°C. In all cases the rate
of oxidation was greater than that of nitridation; we will discuss the nitride
formation later. The rate laws which were observed in various temperature
regions and oxidation times may be summarized in the following scheme:

1 Ehrlich, P.: Angew. Chern. 59,163 (1947).


2 Jaffee, R. I., H. R. Ogden, and D. J. Maykuth: Trans. AIME 188, 1261 (1950).
3 Carpenter, L. G., and F. R. Reavell: Metallurgia 39,63 (1949).
4.2. Scaling Systems with Electron-Conducting Protective Layers 221

Rate Law

1 hr oxidation logarithmic cubic parabolic parabolic parabolic

longer parabolic parabolic


oxidation cubic parabolic with linear 2 and
jumpsl linear 3
I I I I I
300 400500600 7bo sbo 900 10~0 uJooc

Under the assumption of a preferred Ti0 2 formation during the attack of


oxygen on titanium-which appears to be the case in short-time experiments
-and lattice defects according to (4.72), one should expect a decrease in the
oxidation rate from small alloying additions of tungsten or niobium to
titanium or from embedment of titanium samples to be oxidized in W03 or
Nb 2 0 5 , since the introduction of hexavalent tungsten ions or pentavalent
niobium ions into the Ti0 2 lattice according to the incorporation equation:
WO a + 00" <-----" W.··(Ti) + Ti0 2 (4.75)
or
(4.76)
is associated with a reduction in the 00" concentration. The increase in
the concentration of free electrons e is equivalent to (4.75), according to

WOs .....--" W."(Ti) + 2 e + Ti0 2 + ~ o~g) (4.77)

which was confirmed 4 from the observed decrease in the electrical conductivity
(Fig. 101). Hauffe and Pfeiffer were able to confirm the correctness of the
conclusions by oxidation experiments on titanium with a loose covering of
W03 (Fig. 97)5 and on titanium-tungsten or titanium-niobium alloys be-
tween 600 and 1000°C. The decrease in the rate of oxidation from a niobium
addition of 1 at. % can be seen in Fig. 102.6 The parabolic course of the
oxidation plotted here, which was obtained for longer times remains un-
disturbed by a periodic bursting of the protective layer. In contrast to
1 By cracking of the scaling layer.
2 Porous protective layer; dependent on the quality of the titanium.
3 With higher temperatures the parabolic course predominates again, since the Tammann
temperature was attained for Ti02, therefore better sintering.
4 Hauffe, K., H. Grunewald, and R. Tranckler·Greese: Z. Elektrochem. 56, 937 (1952).
5 Pfeiffer, H., and K. Hauffe: Z. MetaUk. 43, 364 (1952).
6 Kofstad, P., and K. Hauffe: unpublished results.
222 4. Scaling Processes with Formation of Thick Protective Layers

GOOoe

250~----~--~----~----~

200~----~--~--r--+~~~

Fig. 101. Variation of the conductivity


ratios u/u o of Ti0 2-WOa mixed oxides as
a function of the WOa content in 1 atm of
air, according to Hauffe, Grunewald, and
Tranckler-Greese. u is the electrical con-
ductivity of the mixed phase and UO is that
of the pure Ti0 2 phase.

titanium a greater adherence of the protective layer can frequently be


observed with titanium alloys. As X-ray and electron-optical investigations
show,l the compact scaling layer consists of a rutile or a solid solution of
Ti0 2-Nb 20 5 . TiO and Ti 20 3 formation is not observable. The brittleness
of the alloy produced in the oxidation of titanium-niobium alloys is so
complete that one can shatter the alloy sheet with ease. In order to reduce
this brittleness one may use metallic or ceramic protective layers, for example,
as was done by Toler.2
Since the oxygen-pressure dependence of the conductivity of
Ti02-Cr203 mixed oxides below 900°C with oxygen pressures of 1 atm and
higher indicates an inversion from n- to p-type, then,3 under these experi-
mental conditions, the incorporation equations (4.75) and (4.76) are no
longer valid. Under these conditions, an addition of a metal with valence
lower than four would be advisable. Therefore, an alloying of small amounts
of chromium below 800°C should cause a decrease in the oxidation rate.
Here we must consider, as an investigation of the titanium-chromium phase
diagram 4 has shown, that the solubility of chromium in titanium below
885°C is less than 0.5 wt.% (IX phase). Not until this temperature is exceeded
is the solubility of chromium in the {3-1 phase sufficiently great--then
1 Kofstad, P., and K. Hauffe: unpublished results.
2 Toler, H. R., Jr.: Am. Ceram. Soc. Bull. 34,4 (1955).
a Hauffe, K.: in Transition 1vletal Compounds, Informal Proc. Buhl Intern. Conf. on
Materials, Pittsburgh, Penn. 1963, p. 37.
4 Cuff, F. B., N. J. Grant, and C. F. Floe: J. ]!'letals 4,848 (1952).
4.2. Scaling Systems with Electron-Conducting Protective Layers 223

generally at least up to 20 wt. % Or. This temperature region is, however,


uninteresting for the oxidation of titanium-chromium alloys according to the
above explana.tions, since above 885°0 the introduction of chromium ions
in the Ti0 2 scaling layer increases the vacancy concentration of the oxygen
ions and the oxidation rate. Indeed, the oxidation rate of titanium is increased
by the addition of a few percent of chromium as McPherson and Fontana!
were able to demonstrate. Not until more than 17 wt. % Or is added does a
certain improvement in the resistance to oxidation occur, but this cannot be
explained on basis of the lattice defect mechanism in a Ti0 2 mixed phase.

g2/cm.4
,
fTirvanAr;e// ! /TifvonArl<elJ
~Qr---~~-------,--~---+--~,:>00

,,
+ I /
,
,
I .I
I~
./
'()()(J

Fig. 102. Rate of oxidation of


van Arkel titanium and a
titanium-niobium alloy with
1 at. % Nb in oxygen at 1 atm
at 895°0, according to Kofstad
and Hauffe. The rate of oxida-
tion of the alloy was about
1.5 orders of magnitude less
than that of titanium.
The structural investigations 2 in the system titanium-chromium (up to
20 wt.%)-oxygen (up to 10 wt.%) carried out by Grant and co-workers
are useful for the explanation of the influence of chromium on the rate of
oxidation. Finally, the structural investigations and the phase diagrams
of titanium-tungsten-oxygen,3 titanium-niobium-oxygen,4 and titanium-
chromium-oxygenS alloys may be referred to and are of general interest.
1 McPherson, D. J., and M. G. Fontana: Trans. ASM 43, 1098 (1951).
2 Wang, Oh.-Oh., and N. J. Grant: J. Metals 6,200 (1954).
3 Maykuth, D. J., H. R. Ogden, and R. I. Jaffee: J. Metals 5,231 (1953).
4 Ames, S. L., and A. D. McQuillan: Acta Met. 2, 831 (1954).
5 Rostoker, W.: J. Metals 7,113 (1955).
224 4. Scalin~ Processes with Formation of Thick Protective Layers

Maynor, Barrett, and Swiftl studied the oxidation rate in air of a large
experimental series of titanium and titanium alloys (binary and ternary)
with small additions between 0.5 and 5 wt. % of the following alloying metals
at 650, 760, 871, and 982°C: aluminum, chromium, copper, iron, manganese,
molybdenum, niobium, nickel, silicon, tantalum, vanadium, and tungsten.
The results ofthese experiments, shown in Fig. 103, are clear proof that in the
light of the above consideration a decrease in the oxidation rate is caused
only by those alloying additions which are introduced into the scaling layer
predominantly as ions with a valence greater than four, e.g., niobium,
tantalum, tungsten (silicon is an exception). The effect of molybdenum and
manganese is difficult to understand, since these two metals also form
stable trivalent and tetravalent ions in the form of Mo0 2 and Mn02 or
Mn20a or Mna04. According to the experimental results shown in Fig. 103,
the oxidation-limiting effect of niobium, tantalum, and tungsten is increased
by a silicon addition of about 1 wt.%. However, the addition of 1% Si by
itself causes a considerable decrease in the oxidation rate. The cause of this
effect can scarcely be attributed to a reduction in the oxygen ion vacancy
concentration, since silicon can only appear as a tetravalent ion in the
scaling layer. Probably the beneficial effect of the silicon is due to an improve-
ment of the epitaxy and to a silicate layer formation which forms as a
"barrier layer" between the alloy and the Ti0 2- or Ti0 2-mixed oxide layer.
With higher silicon contents (3 wt. %) in titanium an increase in the oxidation
rate was observed again, especially at higher temperatures.
The considerable increase noted in the oxidation rate of titanium with
small additions of vanadium (0.84%) is significant, particularly since with
an addition of about 4 wt.% V the increase is only about half as great.
Systematic investigations of these alloying additions are required for an
explanation of the cause of this effect. Oxidation experiments in the presence
of V205 appear to be desirable for an investigation of the usefulness of
titanium alloys as industrial materials suitable for exposure to furnace
gases containing vanadium oxide.
The explanation of the mechanism of oxidation of titanium carbide,
which will be discussed in Chapter 6, is quite difficult since today we do not
know enough about carbon influence on oxidation. Kinna and Riidiger 2
found a parabolic rate law for the oxidation oftitanium- and titanium-cobalt
carbidcs between 600 and 800°C (Fig. 104). Oxidation experiments carried
out simultaneously on a hard metal alloy with 82 wt. % TiC and 18% Co at
1000°C yielded a somewhat higher scaling constant than pure titanium
ll\Iaynor, H. W., Jr., 13. R. Barrett, andR. E. Swift: WADC Tech. Rep. 54·109, Contract
No. AF 18 (600)-60, Project No. 7351, ::\Iarz, 1955. See also A. E. Jenkins: J. Ins!.
lVIetals 84, 1 (1955/56).
2 Kinna, 'Y., and O. Rudiger: Arch. Eisenhiittenw. 24, 535 (1953).
IIJI7
(If
9/",,1
IVl

(¥JI

r (lQ3

.,..
~~
-5"
.cq~
.~

~ 4ttY

m'e w'e

" K»
7fI fI(J 'lU
Hours -
Fig. 103. Oxidation of titanium (with different heat treatments), titanium alloys and
steels at (a) 760°C, (b) 871 °C, and (c) 982°C in air at 1 atm, according to Maynor, Barrett,
and Swift. The following numbers give the composition of the titanium alloys in weight
per cent:
1 : Titanium (No. 7 7: 4.37% Ta 15 : 1.03% Si 19 1.0 % Nb
and 21) 8: 302-Type Steel 16: 2.28% Fe + 2.55% 20 0.54% Ta
2: Titanium (No. 115) (Republic Steel) V (Mallory-Sharon- 21 0.84% V
3: Titanium (Stand- 9 4% W + 1% Mo Titanium) 22 3.88% V
ard) 10 6% Al + 1 % Si 17: 1.93% Al + 5.38% 23 4.03% Cr
4: 2.95% W 11 4% W + 1% Si Cr (Mallory-Sharon- 24 1.05% Cr
5: 302-TypeSteel(Uni- 12 4% Ta + 1 % Si Titanium) 25 3.96% Mo
ted States Steel) 13 4% Ta + 1% W 18: RS-70 (Titanium 26 3.69% Si
6: 4.45% Nb 14 4% Ta + 1% Nb of Republic Steel)
226 4. Scaling Processes with Formation of Thick Protective Layers

carbide. After 40-60 hr below 900°C the scaling layer consisted predominantly
of Ti0 2 with C0304 inclusions and, at 1000°C, predominantly of Ti0 2 with
embedded Co and CoO. These investigations are in accord with the oxidation
experiments on hard metals by Greenhouse and co-workers l with a composi-
tion of 80 wt. % TiC and 20% Co. In agreement with the above explanation

7r----,r----,
go/cm;

5J___---+--+-__I

t
~
~JJ----~4-----I

~~

Fig. 104. Oxidation of an alloy consisting of 82 wt. % TiC


![f[f '?[f[f and 18 wt. % Co in air at I atm, according to Kinna and
Hours_ Rudiger (parabolic rate law).

of the mechanism of titanium oxidation, the diffusion of oxygen through the


scaling layer was regarded as the rate-determining process in the hard
metals.
In this connection a work by Roach 2 on the improvement of the resistance
to oxidation of titanium carbide by additions of chromium should be men-
tioned. A chromium content of 5 wt. % clearly causes a decrease in the oxida-
tion rate between 600 and 1400°C. Besides this work, still another investiga-
tion on the beneficial influence of small Cr, TiN, Mo, and MoSi additions can
be specially noted. 3 Further, in the ternary carbide systems of titanium,
silicon, and boron an oxidation-retarding influence of impurity carbide
additions was reported. 4 At this time nothing definitive may be said about
the mechanism of this influence. Basic knowledge necessary for the explana-
tion is still lacking (see also Chapter 6). The phase diagram of the titanium-
carbon system has been investigated by Jaffee and co-workers. 5 Duwez

1 McBride, C. C., H. M. Greenhouse, and T. S. Shevlin: J. Am. Ceram. Soc. 35,28 (1952).
See also O. E. Accountius, H. S. Sisler, T. S. Shevlin, and G. A. Bole: J. Am. Ceram.
Soc. 37, 173 (1954).
2 Roach, J. D.: J. Electrochem. Soc. 98,160 (1951).
3 Tinklepaugh, J. R., E. W. Holman, and R. E. Wilson: Rep. ATI-I08738.
4 Accountius, O. E., H. S. Sisler, T. S. Shevlin, and G. A. Bole: J. Am. Ceram. Soc. 37,
173 (1954).
5 Jaffee, R. I., H. R. Ogden, and D. J. Maykuth: J. Metals 2, 1261 (1950).
4.2. Scalin~ Systems with Electron-Conductin~ Protective Layers 227

and Odell1 reported on structure investigation in binary carbide systems of


zirconium, niobium, titanium, and vanadium. 2
Gulbransen and Andrew 3 studied the attack of nitrogen on titanium
between 550 and 850°C. The nitride formation follows an approximately
parabolic rate law in this temperature region with an activation energy of
36.3 kcaljmole (Fig. 105). The experimental results for nitride formation

o.@~----+-----~~

Fig. 105. Temperature dependence of the rate of nitride


formation on titanium in nitrogen at 76 mm Hg between
550 and 850 0 0, according to Gulbransen and Andrew.

on titanium carried out at 76 mm Hg of nitrogen are summarized in Table


32.4 As can be seen from Fig. 106, the rate of nitride formation is dependent
upon the nitrogen pressure.
Thermodynamic and structural investigations on TiN layers were con-
ducted by Munster and co-workers 5 in connection with other work. The lattice

do.-----,------,------,--,
gjcm 2
f~O~---4----~=-~~

~ 90~----~~~~~~~~~
~ I
~1t:>-i<Oi-;~~==_+------r__--_+___I
Fig. 106. Influence of the nitrogen pressure
on the rate of nitride formation on titanium °O~----~~O~----~8~O.---~~~
at 800 0 0, according to Gulbransen and Nitride Formation Time_
Andrew.

1 Duwez, P., and F. Odell: J. Electrochem. Soc. 97, 299 (1950).


2 See also A. E. Kovalskii and Y. S. Umanskii: Zhur. Fiz. Khim. SSSR 20,769 (1946);
H. Nowotny and R. Kieffer: Metallforschung 2,257 (1947).
3 Gulbransen, E. A., and K. F. Andrew: Metals Trans. 185, 741 (1949).
4 Wyant, J. L., and N. J. Grant: Preprint No.3, ASM (1953).
5 Miinster, A., and W. Ruppert: Z. Elektrochem. 57, 558, 564 (1953); A. Munster and
K. Sagel: Z. Elektrochem. 57, 571 (1953).
228 4. Scalin~ Processes with Formation of Thick Protective Layers

Table 32. Nitridation of Titanium in Ammonia, According


to Wyant and Grant, and in Nitrogen, According to
Gulbransen and Andrew

PNH, = I atm PN, = 0.1 atm

k'l, k'l,
T, °0 g2(cm4 -sec T, °0 g2(cm4 -sec

743 1.5 x 10-11 550 5.3 x 10-15


780 3.2 x 10-11 600 3.0 X 10-14
800 7.0 X 10-11 700 2.8 X 10-13
827 8.8 x 10-11 775 1.1 x 10-12
854 9.6 x 10-11 800 1.5 X 10-12
883 1.1 x 10-10 850 3.2 x 10-12
910 1.7 x 10-10

defect conditions in TiN crystals are still not well known. According to
Munster,l TiN layers can exhibit metallic (better than titanium) and
semiconducting properties, which depend on the material below these thin
layers (see also the electrical conductivity measurements on chromium-
titanium nitrides).2 Unfortunately because of the short half-life of radioactive
titanium isotopes, self-diffusion measurements cannot be carried out, so that
an important experimental method is not available for the clarification of the
diffusion mechanism of the ions in TiN crystals. Marker experiments should
permit a decision between the two diffusion possibilities (whether titanium
or nitrogen). Munster and Sage13 have investigated the oxidation of TiN
in pure oxygen under normal pressure between 625 and 1075°C, and their
experimental data could be evaluated with the following rate equation:
t = aLlm 2 + bLlm, which contains both a parabolic and a linear term. The
experimental data are affected with an uncertainty because of the application
of TiN spirals for the oxidation experiments (see Fig. 107).

4.2.2.3. The Oxidation Rate of Zirconium


Gulbransen,4 and Cubicciotti,5 studied the oxidation rate of zirconium
between 200 and 425°C in 76 mm Hg oxygen pressure and between 600 and
920°C at oxygen pressures of from 0.1 to 202 mm Hg. The results of the
former can be described by a cubic rate law with an activation energy of
about 26.2 kcal/mole. The parabolic course of the oxidation found by
ll\Iunster, A., K. Sagel, and G. Schlamp: Sature (London) 174,1154 (1954); A.Munster
and K. Sagel: Z. Physik. 144, 139 (1956).
2 Olson, E. R., E. H. Layer, and A. E. Middleton: J. Electrochem. Soc. 102, 73 (1955).
3 :i\Iiinstel', A., and G. Sehlamp: Z. physik. Chem. [XF] 13, 59, 76 (1957).
4 Gulbransen, E. A., and K. F. Andrew: J. Metals 1, 515 (1949).
5 Cubiceiotti, D.: J. Am. Chem. Soc. 72, 4138 (1950).
4.2. Scaling Systems with Electron-Conducting Protective Layers 229

2.5

2.0

~
E
:s
u

1.5
"0
E
<1

i 1.0

0.5

o 100 200 300


-+ Time (min)
Fig. 107. The rate of oxidation of TiN spirals in oxygen of
1 atm, according to Munster and Schlamp.

Cubic ciotti at higher temperatures could not be confirmed by Belle and


Mallett.1 As can be seen from the experimental results in Fig. 108 the course
of oxidation of zirconium in 1 atm of oxygen at 900°C also follows a cubic
tarnishing law. The temperature dependence of the rate constants kc (in
cm 9jcm6 -sec or cm3jsec) may be given by the following expression:
kc = 3.9 X 106 exp( -47,200/RT)
Little can be said at this time about the oxidation mechanism, since
the lattice defect conditions in Zr02, which is formed as the primary product
in the oxidation, are not sufficiently well known. Under the conditions
existing here a preferred n-conduction is probable. Through analogy with
experimental observations on the current conduction in Zr02-Y203 solid
solutions, especially in the Nernst material (85% Zr02 + 15% Y203)2 one
would surmise that there were predominant Schottky defects with ZrO'"

1 Belle, J., and M. W. Mallett: J. Electrochem. Soc. 101, 339 (1954).


2 Wagner, C.: Naturwi88. 31,265 (1943).
230 4. Scaling Processes with Formation of Thick Protective Layers

and 00" vacancies in the Zr02 lattice, where obviously the oxygen ion
vacancies are due to the greater mobility of the charge carriers,! as must be
assumed from measurements of the emf of suitable electrochemical cells
(CO/C0 2[Zr02 + Y203[02). Because of the extensive intrinsic ionic lattice

t t.Ot---:::;..='-t---+--+---+----=""
~
]O£~---+----~~~-4----~--~
E
"~a"~~£T----~-4~~~~
(l
t:
OJ
~a2~~-T----~-4~-4----~--~ Fig. 108. Oxygen consumption during
><
o the oxidation of zirconium in 1 atm
oxygen between 690 and 900°C, according
C.~~'--~2~O----.~O~~60~~~~O---2~~~;---.~OC to Belle and Mallet [(cubic rate law:
Minutes_ gl(t) = (At + B)l/n)].

defects it is difficult to see in which way additions of higher- and lower-


valence ions can affect the oxidation rate.
Charlesby 2 also found indications for a cubic rate law in the electrolytic
film formation in anodic oxidation. However, it is still not possible to relate
this observation to those at higher temperatures. Hayes, Roberson, and
Robertson 3 have reported on the resistance to scaling of zirconium in various
attacking media. In the presence of gases other than oxygen the situation is
considerably more complicated, as was impressively shown in the Phalnikar-
Baldwin4 oxidation experiments with zirconium in the presence of nitrogen
(air) .
Mallet and Albrecht 5 investigated the oxidation rate of zirconium-tin
alloys with 1.5 and 2,5 wt.% Sn in 1 atm oxygen between 600 and 900°C.
The alloy with the lmver tin content oxidized according to a cubic rate law,
while the oxidation of the other could be described by a parabolic rate law.
In the first case two oxidation regions were obtained:
1. Between 600 and 800°C
kc = 5.34 X 10 4 exp( -38,400jRT)(cm3jcm2)3/sec
1 'Wcininger, J, L" and P. D. Zemany: J. Chern. Phys, 22,1469 (1954),
2 Charlesby, A.: Proc, Phys, Soc, (B) 66, 317 (1953); Acta Met. 1, 340, 348 (1953).
3 Hayes, E, T., A, H, Roberson. and R, H. Robertson:J. Electrochem, Soc, 97, 316 (1950).
4 Phalnikar, C, A" and \y,::\1. Baldwin: Proc. Am. Soc, Testing l11aterials 51,1038 (1951).
5 Mallott, ::\1. VV" and \Y. ::\1, Albrecht: J. Electrochem. Soc. 102,407 (1955),
4.2. Scalin~ Systems with Electron-Conductin~ Protective Layers 231

2. Between 825 and 900 0 e:


kc = 8.71 X 101 exp( -22,600jRT)(cm3jcm2)3jsec
The temperature dependence of the parabolic scaling constants in the
case of the second alloy may be given between 550 and 900 0 e by the follow-
ing expression:
k = 2.63 X 103 exp( -32,400jRT)(cm3fcm 2)2fsec
Thermoelectric measurements in these Zr02 layers yield a negative sign,
which indicates free electrons and an n-conducting character of the protective
layer, and is compatible with the appearance of oxygen ion vacancies.
However, recent studies by Rudolph1 of electrical conductivity of Zr02 as
a function of oxygen pressure suggest that Zr02 is a p-type conductor. This
behavior at least excludes a defect structure in terms of oxygen vacancies.
Kofstad 2 studied the electrical conductivity of Zr02 as a function of oxygen

.
!
m onocUnlc 917 '1;

-14 .12 -to .,


L09 PO l Iotm)

Fig. 109. The logarithm of the resistance (R in ohms) of Zr02 specimen as a function of
the logarithm of the partial pressure of oxygen, according to Kofstad.

pressure between 927 and 1330°C (Fig. 109) and attempted to carry out
thermogravimetric measurements under the same conditions to explain
these contradictory results. According to his measurements, the weight loss
at 1450 0 K in going from po, = 1 to 10-15 atm corresponds to a change
in the value of x of < 0.001(Zr02-x). This indicates that Zr02, in contrast
to Ti0 2, has a very small homogeneous n-type range and that the free energy
of formation of defects is large : .Mi ;::: 300 kcaljmole at 1450 K and at 0

1 Rudolph, J.: Z. Naturforsch . 14a, 727 (1959).


2 Kofstad, P. : private communication.
232 4. Scaling Processes with Formation of Thick Protective Layers

x = 0.001. This is more than three times the value estimated by Aronson. 1
On basis of electrical conductivity measurements, the monoclinic Zr02
shows a p-type conductivity between 10-4 and 10-8 atm oxygen in agreement
with the Rudolph result (~ ~ p~,5 --+ p~,5.5). At smaller oxygen pressures
( < 10-8 atm) the conductivity has an n-type character. A p-type conductivity
may either indicate a defect structure involving cation vacancies or inter-
stitial oxygen, while an n-type conductivity may reflect a defect structure
either with interstitial cations or oxygen vacancies. On basis of transference
experiments of oxygen ions through Zr02 by means of emf measurements
with small and larger additions of CaO as solid electrolyte 2 we may conclude
that in both cases the defect structure is represented by interstitial oxygen
and oxygen vacancies, respectively. In a recent publication Schmalzried3
discussed the quantitative relations of the electronic and ionic conductivity
dependence upon temperature and oxygen pressure.
The observation of Mallet and Albrecht that the cubic rate law observed
with these thick layers cannot be explained in terms of the theory of Cabrera
and Mott is convincing. They noted that the protective layers in their
experiments were considerably thicker than the ::; 2 x 10-4 cm thicknesses
permitted in the Mott theory and that this makes it probable that the cubic
rate law is caused by aging processes or by the simultaneous run of oxygen
diffusion in the metal phase and of layer formation as it was described for
the titanium oxidation. After an initial period, variations in the kinetics
were observed, and for alloys this has in many cases been found to result in a
"breakaway" oxidation. 4 It has been suggested that this change in kinetics
is a result of a mechanical breakdown of the scale, a change in Zr02 modifica-
tion, or oxygen dissolution in the metal. No clear correlation has yet been
established.
Additional experiments concerning the influence of oxygen content in
zirconium on the oxide-layer growth have been carried out recently by
Osthagen and Kofstad. 5 The oxidation rate was found to increase with
increasing oxygen content up to 12-15 at. %, above which no further increase
in oxidation rate was observed. The fact that the time lag decreases until
the oxygen concentration of the zirconium-oxygen alloys reaches 12-15 at. %
suggests that the "saturation" limit as it was observed for titanium corres-
ponds to this concentration range. Zirconium may dissolve up to 29 at. %
oxygen and it is not altogether clear why the "saturation" value does not

1 Aronson, S.: J. Electrochem. Soc. 108, 312 (1961).


2 Kuikkola, K., and C. Wagner: J. Electrochem. Soc. 104, 308, 379 (1957).
3 Schmalzried, H.: Z. physik. Chem. [NF] 38, 87 (1963); 25, 178 (1960).
4 Wallwork, G. R., and A. E. Jenkins: J. Electrochem. Soc. 106, 10 (1959); H. A. Porte,
.J. G. Schnizlein, R. C. Vogel, and D. F. Fisher: J. Electrochem. Soc. 107, 506 (1960).
5 l>sthagen, K., and P. Kofstad: J. Electrochem. Soc. 109, 204 (1962).
4.2. Scalin~ Systems with Electron-Conductin~ Protective Layers 233

correspond to this solubility limit. Corresponding effects were obtained by


X-ray diffraction measurements. The interplanar (205) spacing increased
from 0.83 A without oxygen to 0.86 A with 15 at.% oxygen. At higher
oxygen content no further increase could be observed. The rate of oxygen
dissolution is influenced by these large structural changes and further
dissolution of oxygen may lead to large stresses and strains which partly
result in cracking of this layer. The increased rate of oxidation with oxygen
content shown in Fig. llO may reflect a competition and interplay between
oxygen dissolution and oxide formation . The latter is governed by the rate
of nucleation and rate of growth of oxide nuclei during breakaway oxidation.

24
22

20

18
16

14

12

10 ~
4 .lop. "'0

8
6

4
2

20 40 60 80 100 120 140 160 180

Fig. 110. Oxidation of zirconium and zirconium-oxygen alloys


at 800°0 and 1 atm oxygen, according to Osthagen and Kofstad.

Microscopic and X-ray investigations of zirconium samples oxidized


in air at 900°C yield, besides a white exterior, a thin dark Zr02 layer in the
interior of the scale. At the same time, the surface of the oxidized sample
enlarged about 200% after 4 hr, an effect not observed to the same extent
with other metals. As can be seen from the photomicrographs, the scale/zir-
conium boundary surface shows a strikingly irregular pattern. This large
dimensional change. cannot be explained by the solubility of oxygen and
nitrogen in zirconium (volume effect). Samples suspended in the reaction
furnace exhibit a greater broadening at the lower end than at the upper
end after the oxidation experiment. Figure lIla shows the percentage
234 4. Scaling Processes with Formation of Thick Protective Layers

increase of the surface of the zirconium sheet with time in air at various
temperatures. It is significant that sheets which scaled in pure oxygen or
pure nitrogen at 900°C for 4 hr showed no increase in surface. It thus seems
to be necessary that both gases be present to cause such a phenomenon.

% ~
zoo
I I
I, II
wo°CV I

I l s
/ O'Oo0C V I!
I I sl
0
/
I/o / rool
't12
/ Of 06" J
/
"Ii 1/l
/
2/l 'fO 5/l1t»
Fig. IlIa. Percentage increase of the
surface with time during oxidation in
air, according to Phalnikar and
Hours- Baldwin.

The application of zirconium alloys as atomic reactor materials, has


caused a tremendous increase in the amount of literature published in
journals and special unclassified reports concerning the behavior of zirconium
and zirconium alloys in water, steam, and oxygen between 300 and 700°C.
Most of these investigations were carried out from an empirical viewpoint.
Only a few, e.g., the contribution of Porte et aZ.,l were dealt with under
a theoretical concept. Except for some alloys of zirconium-niobium base 2
the oxidation rate of all investigated zirconium alloys may be described in
the first stage of oxidation by a parabolic and sometimes by a cubic rate law.
Then a sudden change to higher oxidation rates associated with a transition
to a linear rate law is observed. This transition time is dependent on the
temperature and on the composition of the alloy and its heat treatment.
In the pretransition period the oxide layer is black; after the transition the
growing oxide first becomes spotted with white and then turns completely
white owing to a breakaway process and an oxidation of the initially sub-
stoichiometric zirconium oxide to a stoichiometrically composed one.
On basis of the present results, it is assumed that the black pretransition
oxide is apparently protective while the white post-transition oxide is
1 Porte, H. A., J. G. Schnitzlein, R. C. Vogel, and D. F. Fischer: J. Electrochem. Soc.
107,506 (1960); see also B. Cox: J. Electrochem. Soc. 108, 24 (1961).
2 Kisclev, A. A., V. A. Myshkin, A. V. Kozhevnikov, S. r. Korolev, and E. G. Shorina:
Proc. Conf. Reactor 1'rlaterials, Vol. 2, pp. 67-104, Salzburg 1962, translated by the
Atomic Energy of Canada, AECL-1724.
4.2. Scaling Systems with Electron-Conducting Protective Layers 235

stoichiometric and nonprotective. Thus, transition is the onset of a second


oxidation process characterized by a more rapid oxidation. Therefore, it is
the aim of all developments for corrosion-resistant zirconium alloys to prevent
this transition and to preserve the existence of the black protective oxide
layer.
Besides the oxide layer formation and its growth decrease an additional
problem is posed by hydriding in a water-cooled reactor, which causes
undefined changes in mechanical properties of the zirconium alloys by
precipitation of a hydride phase. In its commonest form, this phenomenon
occurs in tubular fuel sheaths or hot pressure tubes in which a thermal
gradient causes a rim of hydride-rich material to form on the outer, cooler
surface of the tube. Schwarz and Vaughan l have produced evidence to show
that the hydrogen causes damage to the surface and increases the rate of
oxidation. The damage was roughly proportional to the integrated flux of
hydrogen atoms through the oxide layer. These effects are particularly
important relative to the corrosion of zirconium alloys in water or steam
above 300°C.2
Under the presupposition of a compact and coherent oxide layer the
rate of oxidation is determined either by the ionic or by the electronic
conductivity in the growing oxide layer. Therefore, one must first investigate
the transference number of the ions and electrons in the oxide layer under
growing conditions. This has been done with the following electrochemical
cell
Au, O2 (1 atm)IZr02lZr

0.8

0.6 V
0.4
t to= /
Fig. Hlb. Temperature dependence of the
transference number to~ of oxygen ions in a
0.2
V
V
growing Zr02 layer on pure zirconium in oxygen -+T(°C)
at 1 atm, according to Kirkbride, Thomas,
Dalgaard and Hauffe. 300 400 500 600 700

1 Schwartz, C. M., and D. A. Vaughan: EMI·H20, 1956.


2 Wanklyn, J. N., and B. E. Hopkinson: AERE, M/R·2390, 1957; see also J. N.
'Vanklyn, C. F. Britton, D. R. Silvester, and N. J. M. Wilkins: J. Electrochem. Soc. 110,
856 (1963).
3 Kirkbride, L. D., and D. E. Thomas: WAPD·T-308; S. B. Dalgaard, and K. Hauffe:
unpublished results.
236 4. Scalin~ Processes with ForInation of Thick Protective Layers

The experimental results are represented in Fig. llIb. While at 300°C the
oxide layer is an overwhelming electronic conductor, at 600°C the layer
becomes overwhelmingly ionic-conducting. These measurements agree with
those of Misch and !seler.1 These results, however, are not striking, because
no measurements have been carried out under irradiation in the reactor.
This motivating concept for a reasonable investigation and develop-
ment of highly resistant zirconium alloys is missing in many investigations.
Therefore, it is not astonishing that many investigations were carried out
without success. 2 Rosner's investigations3 of zirconium-based alloys with
small additions of niobium and copper or calcium were more successful.
A good corrosion-resistant alloy is zircaloy-2 developed by Lustman and
co-workers. Asher and Cox4 could demonstrate that high fluxes of fast
neutrons (> 1013 nfv) have a significant effect in both the pretransition and
post-transition regions of oxide film growth. In contrast to these results,
Dalgaard could not find an accelerating effect. 5

4.2.2.4. The Oxidation Rate of Niobium


Investigations into the oxidation rate of niobium are especially interest-
ing, since it has a high melting point (2415°C), good mechanical properties
at high temperatures, and, in contrast to, e.g., the molybdenum oxides,
upon appearing in the scaling layer, Nb 20 5 exhibits only a slight
vaporizability. Measurements of the rate of reaction with oxygen, nitrogen,
and hydrogen at lower temperatures (250-375°C) were carried out by
Gulbransen and Andrew. 6 They found a parabolic rate law with an activation
energy of 22.8 kcal/mole. Since according to Brauer7 three oxide phases
(NbO, Nb203, and Nb 20 5 ) exist, both the structure of the scaling layer and
the kinetic measurements must be considered. Furthermore, the three modi-
fications of Nb 2 05 may be significant for the scaling layer structure. Brauer
found that the "T" modification is stable between 500 and 900°C, the "M"
modification between 1000 and 1l00°C, and the "H" modification above
1100°C.
In the initial stages of oxidation before the onset of oxide nucleation, the
oxygen concentration may exceed its equilibrium value. Two sub oxides
1 Misch, R. D., and G. W. !seier: ANL·6434, 1961.
2 See for instance, J. N. \Vanklyn, J. T. Demant, and D. Jones: AERE-R 3655, Harwell,
1961; H. Klepper, E. L. Dunn, R. E. Blood, D. L. Douglas, and J. S. Armijo: USAEC,
Contr. AT (04·3)·189 GEAP.4211, 1963.
3 Rosner, U.: USAEC Symposium on Zirconium Alloy Development, Pleasanton,
Cal., Nov. 1962.
4 Asher, R. C., and B. Cox: LA.E.A. Conference, Salzburg, June 1962.
5 Dalgaard, S. B.: private communication.
6 Gulbransen, E. A., and K. F. Andrew: Trans. AIME 188, 586 (1950).
7 Brauer, G.: Z. anorg. u. allgem. Chern. 248,1 (1941).
4.2. Scaling Systems with Electron-Conducting Protective Layers 237

of niobium, NbO x and NbO z, have been found, both with tetragonal structures
closely related to that of the metal.1 The statement by Brauer and Miiller 2
that NbO x decomposes in vacuo above 350°0 into metal and amorphous
pentoxide Nb205 may be due to their lack of knowledge of the Nb02 phase,
which apart from a relatively strong reflection for d = 2.74 A, gives the same
X-ray pattern as the metal. Nb0 2 could only be found when niobium was
oxidized at pressures up to 100 mm Hg. Both suboxides have a metallic
appearance and show an epitaxial growth. The present results suggest that
Nb0 2, which forms light-grey plates embedded in the niobium matrix, grows
parallel to the basal planes of the niobium matrix. 3 Little is known about the
stability of the suboxide phases.
We infer from a work by Inouye 4 on the rate of oxidation of niobium
in 1 atm air, with and without water vapor, between 400 and 1200°0, that
the course of the reaction with time above 600°0 obeys the linear rate law
after a brief parabolic course (Fig. 112). While at 600°0 a water vapor

~r---~---r,-.--r'----.

g/cm 2
~ ~--~~~~~~--~

Fig. 112. Linear course of the oxidation of


niobium in air at 1 atm between 600 and 1200°C,
according to Inouye.

pressure of 18.6 mm Hg yielded about a 50% decrease in the oxidation rate,


at 800°0 this had no effect. At 900°0 the temperature dependence of the
linear oxidation rate constant shows a "crack" which can be correlated
with the disappearance of the "T" modification. Below 900°0 the activation
energy is 13,400 and above 900°0 only 4350 cal/mole. The extreme brittleness

1 Norman, N.: J. Less·Common Metals 4,52 (1962).


2 Brauer, G o, and Ho Muller : Mineral Chemi8try (Proceed. IUPAC's 16th Congress),
Butterworth, London, 1958, p . 63.
3 Norman, No, Po Kofstad, and O. J. Krudtaa: J. Les8-Common Metal8 4, 124 (1962).
4 Inouye, Ho: Document of the Oak Ridge Natiohal Lab., Tennessee, ORNL 1565
(1954).
238 4. Scaling Processes with Formation of Thick Protective Layers

of niobium sheets oxidized above SOO°C is noteworthy, although the oxygen


solubility is relatively small.
Recently, Kofstad and Kj611esdal1 have reported on kinetic studies of
the oxidation of niobium and on structural investigations of the oxidized
specimens by means of X-ray diffraction, electron diffraction, electron
microscopy, and metallographic techniques. The oxidation behavior of
niobium has been investigated in the temperature range 500 to 1200°C
and at oxygen pressures of 760, lOO, lO, 1, and 0.1 mm Hg. At the beginning
of the oxidation, where the oxygen penetration into the metal is governed
by the diffusion of oxygen, a parabolic rate law is observed. Approximately
20 min later a linear rate law rules (Fig. 113a). The activation energy for

I ~OOOC
80 V I

E
u
.. 60
5
~/-Lk
;l ,f.fl / I 9921

I /'/r/l/~ I~
A--:?8JO b

o--'cyyo--b 902"
I
.1f}~;r.>1 ~a-o 708 0

t!~ ~-I L
fZ;~ ~- ~_ 600°C
~<-~-,~~~,\..o_o~ .508° Fig. 113a. Oxidation of
niobium in oxygen at
o 60 120 180 240 300 10 mm Hg, according to
Time (min}-->- Kofstad and Kjiillesdal.

oxygen diffusion in niobium was determined as 26.9 kcaljmole and is larger


than the average activation energy for the total oxidation reaction. Estimates
show that at 1l00°C a specimen of 2 mm thickness should be at least
95% saturated after 90 min of oxidation. The authors could also illustrate
the relationship between oxide formation and oxygen dissolution in the
metal by microhardness studies at 600 and 700°C. In spite of the observed
linear kinetics, it is believed that the rate-determining process must be a
transport of oxygen ions through a thin compact oxide layer at the metal
surface, while the outer oxide layer is porous. The depth of this thin layer
is expected to be approximately constant. Due to variations in oxide
plasticity, the thickness of the barrier is thought to be a function of tempera-
ture, oxygen pressure, and Nb 20 5 modifications being formed. The strong
1 Kofstad, P., and H. Kjiillesdal: Trans. AIME 221,285 (1961).
4.2. Sca1in~ Systems with Electron-Conductin~ Protective Layers 239

oxygen-pressure dependence of the oxidation rate should be noted. Evalu-


ating the rate constant 1for the linear part of the oxidation curves, one finds
that 1" '" Po. at higher oxygen pressures and 1" '" p~; at lower oxygen
pressures. The results at 700°0 suggest the latter oxygen-pressure dependence
over the whole pressure region. At 900°0 the relationship was 1" '" p~.2.
The dependence of the electrical conductivity of Nb 20 5 on the oxygen
pressure confirms the assumption of a prevailing oxygen diffusion via
oxygen ion vacancies. The conductivity of ex-Nb 20 5 at 750 to 1200°0 has
been found to be proportional to PO~/4.2 in the oxygen-pressure range 1 to
0.001 atm.l Oorresponding studies in mixtures of 00 and 00 2 suggest that the
conductivity" ,..., PO~/6.
Ong and Fasse1l2 have used an Arrehenius plot for linear-rate data of
several investigations on niobium. 3 To fit the curves in the 625 to 700°0
region, the weight gain due to the two-phase boundary processes together
with their appropriate area change functions were summed and the best
straight line was chosen as the rate (Fig. 113b). With increasing oxygen
pressure the rapid decrease from a maximum at 625°0 to zero at 700°0,
shown by the dotted line, is increased. Similar behavior was also recognized
by other authors.4
So far as predictions are at all possible here, additions of 0.1 to 2 at. %
tungsten to niobium should increase the resistance to oxidation.

4.2.2.5. Other Scaling Systems with Probable n-Conducting


Protective Layers
Levesque and Oubicciotti5 reported on the rate of oxidation of thorium.
The experimental results in Figs. 114a and 114b show that the time-
dependent course of the oxidation between 250 and 350°0 is determined by a
parabolic rate law and between 350 and 450°0 by a linear one. The respective
activation energies are 31 and 22 kcal/mole. This change in the rate law is
related to the changing of the color of the Th02 protective layer from black
to white. The protective layer becomes porous and phase-boundary reactions
become rate-determining after this transition in a manner similar to that
discussed above. A similar mechanism is followed in tungsten oxidation,
which will be discussed in detail in Section 4.5.1. In an experiment at 500°0,
1 Kofstad, P.: AD AF61 (052}-460, TN 4 (1962); E. H. Greener, D. W. Whitmore, and
M. E. Fine: J. Chern. Phys. 34,1017 (1961).
2 Ong, J. N., Jr., and W. M. Fassel, Jr.: Corrosion 18, 392 (1962).
3 Hurlen, T.: J. Inst. Metals 89, 273 (1960/1); P. Kofstad and H. KjiiUesdal: Trans.
AIME 221,285 (1961); B. B. Argent and B. Phelps: J. Inst. Metals 88,301 (1959/60);
D. W. Bridges and W. M. FasseII, Jr.: J. Electrochem. Soc. 103, 326 (1956); C. T. Sims,
W. D. Klopp, and R. 1. Jaffee: Trans. ASM 51,256 (1959).
4 McLintock, C. H., and J. Stringer: J. Less-Common Metals 5, 278 (1963).
5 Levesque, P., and D. Cubicciotti: J. Am. Chern. Soc. 73, 2028 (1951).
240 4. Scaling Processes with Formation of Thick Protective Layers

--

to '-z
- ID'/T'!(

O.
Fig. H3b. Arrhenius plot for linear oxidation rate data of several investigations on
niobium, according to Ong and Fassell.
Rurlen, all pressures.
0.12] Bridges and Fassell, 1, 6, 8, 27.2 atm.
x Argent and Phelps, 1 atm.
f'::,A Kofstad and Kj6llesdal, all pressures.
+ Sims, Klopp, and Jaffee, 1 atm.

in which the reaction temperature was observed immediately at the thorium


sample, a steep increase in the temperature to over 700°0 occurred after
about 6 min of reaction time. In the subsequent course of the oxidation the
temperature fell gradually and after one hour practically equalled the
furnace temperature. This steep increase was not observed in experiments
below 450°0.
q5'~--------~----~----~

g/crn~

J5'~---4----~--~~----~

..tZ5'~---+----~----r---~
~

~I\:"
~~--~~---4-7~-h~--~

Fig. H4a. Parabolic course of the oxidation


of thorium between 250 and 350°0 in oxygen
of 760 mm Rg, according to Levesque and
Oubicciotti.
4.2. Scaling Systems with Electron-Conducting Protective Layers 241

At higher temperatures oxidation experiments with thorium by


Oubicciottil between 500 and 650°0 and by Gerds and MaIlett 2 between
850 and 1415°0 again yield a parabolic rate law. The oxidation of thorium
in the last temperature interval can be represented by the following relation-
ship:
k" = 5.5 X 10 7 exp( -62,800jRT)(cm 3 jcm2)2jsec
(oxygen consumption in cm3jcm 2 surface). Obviously at high temperatures a
compact and adherent Th0 2 protective layer forms again. At this time our
knowledge about the structure of the high. temperature protective layer is
very limited.

O~r------r------~-----'----,
g·cm···10 3

Fig. 114b. Linear course of the oxidation


of thorium between 350 and 450°C in
oxygen at 760 mm Hg, according to
Levesque and Cubicciotti.

Oerium shows a similar behavior at intermediate temperatures on


attack by oxygen. 3 Between 30 and 125°0 in an observation period of 100 min
a parabolic rate law was found, and above 125°0 (up to 190°0) after an
initial parabolic course a linear rate law was determined. 4 The duration of
the parabolic phase always became shorter with increasing temperature.
The experimental results obtained by Oubicciotti are reproduced in Figs.
115a and 115b. It is noteworthy that here the parabolic rate law is valid
in the lower temperature region. At room temperature for example, the
parabolic oxidation constant is about 2 x 10-15 g2jcm 4 -sec. Furthermore,
the small activation energy of 12 kcaljmole is significant, and this value
also appears in the linear-rate-Iaw region. Here also the appearance of the
linear rate law is related to the lack of a compact and pore-free protective
layer. Quite similar behavior was found in the oxidation of uranium,S where
a parabolic rate law was observed up to about 170°0 and above that a linear
type of oxidation was found. Here too, one is inclined to regard the special
kinetic behavior as a consequence of a porous oxide layer formation.
1 Levesque, P., and D. Cubicciotti: J. Am. Chem. Soc. 73, 2028 (1951).
2 Gerds, A. F., and M. W. Mallett: J. Electrochem. Soc. 101, 171 (1954).
3 Loriers, J.: Compt. rend. 229, 547 (1949).
4 Cubicciotti, D.: J. Am. Chem. Soc. 74, 1200 (1952).
5 Cubic ciotti, D.: J. Am. Chem. Soc. 74, 1079 (1952).
242 4. Scalin~ Processes with Formation of Thick Protective Layers

The activation energies are notably greater here. They are 31 kcalJmole
in the parabolic-rate-law region and 22 kcalJmole in the linear-law region.
A new investigation yielded two linear rate laws between 125 and 250°0. 1
Belle and Auskern 2 have investigated oxygen ion self-diffusion in uranium

Fig. 115a. Course of the oxidation of cerium


with time in oxygen at 760 mm Hg between
30 and 125°C, according to Cubic ciotti. The
parabolic rate law prevails in this temperature-
Time [minJ time interval.

dioxide considering the deviation from the stoichiometry. It was found that
the diffusion of oxygen at 450°0 in U0 2 . 00 4 with D = 4 X 10-14 is about one
order of magnitude smaller than in U0 2 .06 4 (D = 6 x 10-13 cm 2 Jsec). In
essentially stoichiometric U0 2 , it is likely that defects are predominantly
of the anti-Frenkel disorder type with oxygen ions in interstitial positions

Fig. 115b. Course of the oxidation of cerium


with time in oxygen at 760 mm Hg between 125
~~~~;b~~~~~1.~25~O~CJ and 190°C, according to Cubicciotti. The linear
o 1::: 20 *0 tJ(J 80 fOD rate law prevails in this temperature-time
Time [min.] interval.

together with an equal number of anion vacancies. Based on the experimental


results, it is probable that in U0 2 oxygen interstitials are more mobile than
1 Leibowitz, L., J. G. Schnizlein, J. D. Bingle, and R. C. Vogel: J. Electrochem. Soc.
108, 1155 (1961).
2 Belle, J., and A. B. Auskern: Kinetics of High-Temperature Processes, Endicott House
Conference, 1959, p. 44; J. Nucl. Mater. 3, 267 (1961).
4.2. Scaling Systems with Electron-Conducting Protective Layers 243

oxygen vacancies because of the tendency of U02 to become nonstoichio-


metric with excess oxygen.
A considerably different situation is found for the attack of nitrogen
(1 atm) on uranium. Here in the temperature region of 550-900°C a parabolic
rate law was found-at least in the time interval up to 300 min, which was
as long as it was measured-by Mallett and Gerds.l According to X-ray
diffraction in the temperature region from 550-750°C a preponderant
formation of UN2 with only a small amount of U 2N3 was evident, while
between 775 and 900°C all three nitrides, UN, U 2N 3, and UN2 appear in the
scaling layer. Also, in agreement with this finding, the temperature depen-
dence of the parabolic scaling constants was different. For the first tempera-
ture region it was
k = 202 exp( -25,500JRT)(cm 3Jcm2)2Jsec
and for the second temperature region
k = 3.95 exp( -15,100JRT)(cm 3Jcm 2 )2Jsec
The activation energies are accurate to ± 2000 cal/mole.
Preliminary experiments on the kinetics of oxidation were carried out
for tantalum,2 vanadium,3 molybdenum,4 and tungsten. s Detailed investiga-
tions have been published on the oxidation rate of tungsten 6 and molyb-
denum 7 (for the oxidation of tungsten see Section 4.5.1).
The oxidation behavior of tantalum has been intensively studied by
Kofstad in the temperature ranges 300-550°C8 and 500-1000°C,9 at oxygen
pressures ranging from 760 to 0.1 mm Hg. The work comprises oxidation
rate measurements as well as structural investigations of oxidized specimens
by means of X-ray diffraction, electron diffraction, electron microscopy, and
metallographic techniques. After an initial incubation period the oxidation
obeys a linear rate law which at 500°C is independent of the oxygen pressure
and above 600°C proportional to plJ,2 (Fig. 116). The linear oxidation is a

1 Mallett, M. W., and A. F. Gerds: J. Electrochem. Soc. 102,292 (1955).


2 Gulbransen, E. A., and K. F. Andrew: Trans. AIME 188, 586 (1950); J. T. Waber:
J. Chem. Phys. 20,734 (1952); J. T. Waber, G. E. Sturdy, E. M. Wise, and C. R. Tipton,
Jr.: J. Electrochem. Soc. 99, 121 (1952).
3 Gulbransen, E. A., and K. F. Andrew: J. Electrochem. Soc. 97, 396 (1950).
4 Gulbransen, E. A., and W. S. Wysong: Trans. AIME 175,611,628 (1948); R. M. Park:
Metal Progr. 60, 81 (1951); E. Nachtigal: Z. MetaUk. 43, 23 (1952).
5 Gulbransen, E. A.: Ind. Eng. Chem. 41, 1385 (1949).
6 Webb, W. W., J. T. Norton, and C. Wagner: J. Electrochem. Soc. 103, 107 (1956).
7 Simnad, M., and A. Spilners: J. Metals 7, 1011 (1955).
8 Kofstad, P.: J. Inst. Metals 91, 209 (1962/3).
9 Kofstad, P.: J. Inst. Metals 90, 253 (1961/2); 91, 411 (1962/3); J. Electrochem. Soc.
110, 491 (1963).
244 4. Scaling Processes with Formation of Thick Protective Layers

result of heavy Ta205 formation. Contrary to previous suggestions1 the


initial oxidation does not involve a protective stage in the usual sense.
The linear oxidation is interpreted in terms of an oxygen chemisorption
equilibrium followed by a rate-determining step governed by nucleation

0 I
// ;'700":
/-/ I
/0650"C

V
/

V2 A3/
k ..... r~ ///" ",0600"C
n
/ /
X///
/)1 /
O 0575"(
I
I
~// / I
/ /if.t. / 0 __ ..0550"(
/~-(r _---- I
.0

,
/ -/
0
f ( / / ./
X./ /
01525'C

-1_r-~~t-
.-
·f /

I
-3 o500 C('

I
0475"(
Fig. 116. The linear rate constants
l" of oxidation of tantalum as a
I function of oxygen pressure in the
0.1 1 10 100 160 Torr
temperature range 500 to 700°C,
o 1 2 3
according to Kofstad. The dotted
logpOz(Poz,i7mmHg) lines represent calculated values.

and growth of Ta 205 nuclei. Ong 2 developed a rate equation which repro-
duces within a factor of two all reported experimental rates in the tempera-
ture range 475-1400°C and pressure range 2.6 x 10-5 to 40.8 atm. oxygen.
According to measurements of the electrical conductivity, the thermo-
electric force, and of the weight change of Ta 2 05 as a function of the oxygen
pressure, Ta 205 exhibits p-type disorder at pressures close to 1 atm oxygen
and n-type disorder at low oxygen pressures. 3 In agreement with the above-
mentioned results, the interstitial oxygen ions prevail in the p-type Ta 205
and the oxygen vacancies in the n-type Ta205. Recently the reaction be-
tween tantalum and nitrogen at 800-1300°C has been studied. 4 From the
experimental results it may be concluded that the total nitride formation
involves a parabolic growth rate of Ta2N and linear growth of higher nitrides,
e.g., TaN o.s-o.9 and TaN. From X-ray diffraction studies, it would be assumed
that the linear growth of TaN occurs on top of Ta2N.
Molybdenum oxidation obeys a parabolic rate law in the intermediate
temperature region from 250-450°C at oxygen pressures of 0.75-76 mm Hg.
1 Pavel, R. E., J. V. Cathcart, and J. J. Campbell: J. Electrochem. Soc. 107, 956 (1960).
2 Ong, J. N., Jr.: Trans. MSA 224, 991 (1962).
3 Kofstad, P.: J. Electrochem. Soc. 109, 776 (1962).
4 Osthagon, K., and P. Kofstad: J. Less-Common Metals 5,7 (1963).
4.2. Scaling Systems with Electron-Conducting Protective Layers 245

-,
~~C-1
-/I
(
I
-" -'
. A/~kca~rl
J.fIH.f(J°C

/
""'10

Q
I
Fig. 117. Temperature dependence of the parabolic ,H
oxidation rate constants of molybdenum between f7 .. ,../5
-!fI7l
250 and 450°C, according to Gulbransen.

The temperature dependence of the parabolic scaling constants is plotted


in Fig. 117. The activation energy obtained is 36.5 kcalJmole. A continual
vaporization of the oxide layer which is formed appears at temperatures
over 800 C. D

Simnad and Spilnersl reported on the oxidation rate of molybdenum


between 500 and 770 D C in oxygen at 1 atm. The quantitative description
of the experiments above 600 D C was made more difficult by the high vaporiza-
tion rate of the Mo03, which appears in the exterior scaling layer during

~\Joc
-JfJ
\p

1\ \

;ifJ
I\~oc

Fig. USa. Temperature dependence of the rate of


1\ 55tre
\.
I ;fl5
vaporization of MoOa in oxygen at 1 atm, according to
Simnad and Spilners. f-
the oxidation. A separate determination of the vaporization rate of M003
was obtained in this temperature region, which gives an activation energy
of 89,600 above 650 D C and a value of 53,000 calJmole below this temperature,
as can be inferred from Fig. 1I8a. The latter value is in accord with the
50,800 obtained by Gulbransen in the temperature region 474-523°C.
1 Simnad, M., and A. Spilners: J. Metals 7, 1011 (1955).
246 4. Scalin~ Processes with Formation of Thick Protective Layers

While the oxidation rate below 600°0 can be determined almost


solely from the weight increase of the molybdenum sheet, above this temper-
ature the vaporized quantity of Mo0 3 must be added to this weight. In
these cases the authors determined (by methods stated):
1. The vaporized quantity of Mo0 3.
2. The quantity of Mo03 present in the exterior part of the scaling layer
(by solution in ammonia).
3. The residual part of the scaling layer, which consists of Mo0 2 (by a
2-hr reduction at 700°0 with hydrogen).
In Fig. 118b the ratio of the vaporized quantity of Mo0 3 to the quantity
of Mo0 3 in the scaling layer at different temperatures is plotted against time.

05
° 700 or
• 650°C ...- ~
---
a~
o 600°C?
t 0.3
'"
~
:g.a2
o/(
0.1 I~Y:
III Fig. USb. Influence of time and temperature
on the ratio MoOa vapor/MoOa scaling layer =
8 12 16' nD/nz, formed during oxidation, according
Hours_ to Simnad and Spilners.

As can be seen, below 650°0 the ratio assumes a constant value after an hour,
which is not the case at 700°0. The Mo0 2 layer which formed below the
exterior scaling layer attained a constant thickness, independent of the
temperature, between 600 and 725°0 after one-half hour. Obviously the
rate of formation of Mo0 2 is equal to the rate of further oxidation to Mo03.l
Furthermore, the oxidation rate proceeded according to a parabolic rate
law below 700°0, while above this temperature it followed a linear rate law
with variations. The course of the oxidation between 500 and 770°0 is
presented in Fig. 119.
Markers of radioactive silver placed on the metal surface before the
beginning of the experiment were always found on the surface of the scaling
layer after the oxidation. This result precludes a molybdenum ion diffusion
under these experimental conditions. The question of whether it is a case of
oxygen diffusion via vacancies or through pores under these conditions
cannot be decided at this time.
A parabolic course was also observed for the oxidation of tungsten. A
few scaling constants for the region between 700 and 1000°0 2 are summarized
1 The quantitative calculation for the oxidation of tungsten will be found in Section 4.5.1.
2 Dunn, J. S.: J. Chern. Soc. (London) 1929, U49.
4.2. Scaling Systems with Electron-Conducting Protective Layers 247
1.0 ,-----.-----,-----.-----,

700·C

o 8
Hours _
12 16
b
Fig. 119. Course of the oxidation of molybdenum in oxygen at 1 atm at
various temperatures, according to Simnad and Spilners. (a) A parabolic course
is followed below 700· C in the formation of the total amount of MoO a. (Below
600°C the heat ·of activation of the MoOa formation is 48,900 cal/mole.)
(b) Above 700°C the increase of the total amount of MoO a follows a linear rate
law.

in Table 33. New investigations have been carried out by Ong,l who found
that a good approximate form for the rate of tungsten oxidation above
700°0 is
d(m jq)
- - = 5.89 X 106 exp( -12,170/T)p~2 mgjcm 2 -hr
dt •
where Po. is expressed in atmospheres of oxygen. The evaluation of the

Table 33. Temperature Dependence of the


Oxidation Rate of Tungsten in Air at 1 atm,
According to DUIUl

k",
T,oC g2jcm4 -sec

700 4.5 x 10- 9


800 5.5 X 10-8
900 4 X 10-8
1000 1.3 X 10- 7

lOng, J. N., Jr.: J . Electrochem . Soc . 109, 284 (1962); see also E",A. Gulbransen and
K. F. Andrew: J. Electrochem. Soc. 107,610 (1960); W. B. Jepson and D. W. Aylmore:
J. Electrochem. Soc. 108, 942 (1961).
248 4. Scaling Processes with Formation of Thick Protective Layers

experimental data is complicated because of the volatility of WOa. Oismary1


reported on the oxidation in 00 2.
Only a few investigations of the oxidation rates of molybdenum and
tungsten alloys were undertaken. Kessler and Hansen 2 reported on the
attack by air and oxygen on molybdenum-chromium alloys containing
up to 40 wt. % Or. The rate follows a linear law and the oxide layer which
is formed is porous and powdery. Electron diffraction investigations by
Hickman 3 on molybdenum and tungsten alloys showed that with small
alloying additions of 7% Ni, 5% 00, 5% Or, the scaling layer consisted ex-
clusively of M002 + MoO a or W0 2 + W0 3 . Because of the technological
significance of silicide formation on metal surfaces, Kieffer and Oerwenka4
have concerned themselves with the oxidation rate of molybdenum-silicon
and tungsten-silicon alloys in air up to 1500°0. The greatest oxidation
stability was found in the region of 22-40 wt. % Si. Fitzer5 was able to show
that MoSi 2 exhibits remarkably good scaling stability up to 1700°0. An
explanation of the mechanism has not yet been given.
The oxidation rate of aluminum should be especially small, with the
formation of a compact and pore-free Al 20 3 protective layer, since the
diffusion of ions through AbOa is very limited, even at high temperatures.
Pilling and Bedworth6 found, as expected, the parabolic oxidation constant
at 600°C to be k" = 8.5 X 10-16 g2jcm4 -sec. Makolkin 7 and Smeltzer8
repeated the oxidation experiments on aluminum between 460 and 600°C.
As seen in Fig. 120 and Table 34, values found in oxidation experiments
at 600°C were about two orders of magnitude higher. The dashed curved
line in Fig. 120 cannot be attributed to a parabolic rate law. Evidently,
these deviations are caused by ionic transport in the high electrical fields
within the Al 20 a film. At 460°C, after attaining a certain thickness, the
AbOa film apparently stops growing. Nevertheless, as earlier investigations
by Gulbransen 9 showed, the oxidation rate and mechanism (rate law)
depend substantially on the state of the aluminum surface and its purity.

1 Cismary, D., and G. D. Cismary: Acad. rep. populare Roml,ne, Studii cercetari chim.
7, 35 (1959).
2 Kessler, H. D., and M. Hansen: Trans. Am. Soc. Metals 42, 1008 (1950).
3 Hickman, J. W.: Trans, AIME 180, 547 (1949).
4 Kieffer, R., and E. Cerwenka: Z. lYletaUk. 43, 101 (1952); R. Kieffer, F. Benesovsky,
and E. Gallisti: Z. MetaUk. 43,284 (1952)-Comparison of the Resistance to Oxidation of
MoSi 2, WSi2, CrSi2, TaSi2, TiSi 2, ZrSi2, NbSi2, VSi2, ThSi2.
5 Fitzer, H., in: Passivierende Filme ttnd Deckschichten, edited by H. Fischer, K.
Hauffe, and \V. Wiederholt, Springer.Verlag, BerlinfGottingenfHeidelberg, 1956.
6 Pilling, N. B., and R. E. Bedworth: J. Inst. Metals 29, 529 (1923).
7 Makolkin, 1. A.: Zhur. prikl. khim. 24, 460 (1951).
8 Smeltzer, W. W.: J. Electrochem. Soc. 103, 209 (1956).
9 Gulbransen, E. A.: Trans. Electrochem. Soc. 91, 537 (1947).
4.2. Scalin~ Systems with Electron-Conductin~ Protective Layers 249

This hypothesis was justified by the alternate appearance of the linear and
parabolic rate laws.

V.
/.
I
/
,/
/
I

I
1
I
I
I
I
~
I ,~

,,,t ,,'
Fig. 120. Course of the oxidation of
aluminum with time in air between /
/
460 and 600°C in a parabolic plot, / ~(1°C
according to Makolkin. (After an '1
v,....... p,--
initially rapid oxidation for 4 hr
experimental time, the oxidation can 8 12 18
be described according to a parabolic Hours_
rate law.)

Table 34. Parabolic Rate Constants k", Energies LIE, and Entropies
.dB of Activation for the Oxidation of Metallographically Polished
Aluminum Degassed for 30 min at 500°C, According to Smeltzer"

kif, LIE, LIS,


T,oC (g/cm 2 )2/ sec cal/mole cal/mole·deg

400 7.5 X 10-16 40,500 - 3.2


450 1.4 x 10-14 40,500 - 1.7
500 1.0 x 10-13 40,500 - 1.5
8.9 x 10-14 40,500 - 1.8
550 2.4 x 10-13 40,500 - 3.2
2.3 x 10-13 40,500 - 3.2
600 1.1 x 10-12 40,500 - 3.2

a Smeltzer, W. W.: J. Electrochem. Soc. 103, 209 (1956).

Cabrera and co-workers 1 studied the oxidation rate of aluminum


layers vaporized onto carriers in the temperature region 1O-450°C in high
vacuum in air, with and without water vapor, and in ozone (Fig. 121).
1 Cabrera, N.: Rev. llfetallurgie 45,86 (1948); N. Cabrera andJ. Hamon: Compt. rend.
224,1713 (1947); 225, 59 (1947); N. Cabrera, J. Terrien, and J. Hamon: Compt. rend.
224, 1558 (1947).
250 4. Scaling Processes with Formation of Thick Protective Layers

The reaction-accelerating effect of ozone and water vapor was considerable,


especially at low temperatures, while in the absence of ozone the presence
of water vapor produced only an approximate doubling of the oxidation
rate.

" i
rA7~%

'If -- {J!."Io
'~/~ 25%
Fig. 121. Influence of the water vapor content on the
2 0%- rate of oxidation of aluminum in ozone at 250°C,
according to Cabrera and Hamon. (The percentages
o
v.
2 9 G 8 to indicated on the curve refer to the saturation of the
Days_ ozone with water vapor at 10°C).

4.3. tatastrophic Oxidation


It has been variously reported that steels and other alloys with large
additions of certain metals are inclined to an especially rapid oxidation,
which can lead to a rapid destruction of a portion of the alloy. Leslie and
Fontanal found an unusually high oxidation rate for steels with a relatively
high molybdenum content, which they attributed to the low-melting M003
(melting point 795°0) and named catastrophic oxidation. Similarly, steels
with a high vanadium content also show a catastrophic oxidation, since the
V2 05 that is produced (melting point 658°0) forms a low-melting eutectic
with the other oxides produced in the oxidation. Rathenau and Meijering 2
were able to show through further experiments, especially with copper, silver,
and chromium-nickel steel in contact with Mo0 3, that catastrophic oxidation
sets in at the eutectic temperature of Mo0 3 with the metal oxides which are
formed. Thus one can recognize, for example in Fig. 122, that a rapid oxida-
tion sets in at 500°0 on a silver sheet which is placed in Mo03-containing air.
The cause of the sudden speed-up of the reaction was considered to be the
rapid diffusion of silver ions in a eutectic melt of Ag 2Mo04 + Mo03 appearing
at 495°0. Brasunas and Grant 3 obtained the same results. On this basis, we
can understand the catastrophic oxidation of resistance wires in furnaces
1 Leslie, ,,y. C., and M. G. Fontana: Trans. ASJJI 41,1213.(1949).
2Meijering, J. L., and G. W. Rathenau: Nature (London) 165, 240 (1950); Metallurgia
42, 167 (1950).
3 Brasunas, A., and N. J. Grant: Iron Age 85, 17 (1950).
4.3. Catastrophic Oxidation 251

in the presence of oxide insulators which contain some low-melting oxides,


e.g., PbO, observed a long time ago by Hessenbruch.1 Thus, for example,
a small amount of asbestos in the embedding mass will suffice to cause a
resistance wire to burn through within a few minutes if it is heated to about
1300°C (the alloy consists of 73% Fe, 20% Cr, 6% AI, and 1% Co), while
the usual lifetime at this temperature in air is several hundred hours.

1M
)1
120

(0
teo
.,
o
.§ so _ Heating I
~
~., *0
Duration
in hours
xIS; )
1 Sf
.;
:
<>..
) /

~- .......
20
Fig. 122. Effective penetration depth in microns in ~
the attack on silver by air containing MoOa as a o'tOO 520 '1:
function of the heating temperature, according to *W WJO 58fl
Heat ing Temperature_
Meijering and Rathenau.

The mechanism of the molybdenum effect on the oxidation of alloys


containing molybdenum was explained in detail in a work by Brenner.2
At first it was shown that niobium-molybdenum alloys containing up to
30% Mo as well as iron-molybdenum alloys containing up to 20% Mo were
not oxidized very rapidly in air at 1000°0. Addition of niobium or chromium
or both to iron-molybdenum alloys caused catastrophic oxidation. To study
this, Brenner proceeded from a 20% Mo-80% Fe alloy and stepwise replaced
the iron with niobium. Here he found from 2-hr experiments at 1000°0
in air (Fig. 123) that up to 10% Ni a slight decrease in the oxidation rate.
occurred, then with higher niobium additions the rate suddenly increased
by about a factor of 15 through a maximum in a relatively short alloying
interval. In Fig. 124 the alloying regions which are inclined toward cata-
strophic oxidation are designated by shading. According to these results
chromium appears to be especially dangerous in an iron-molybdenum alloy
in terms of susceptibility to catastrophic oxidation. The appearance of the
scaling layer leads to the conclusion that an addition of chromium will
increase the porosity of the scale. Since oxygen thus can penetrate freely,
virtually no more Mo0 2 appears, but only easily melting MoOa, which can
1 Hessenbruch, \V.: .Metalle und Legierungen fur hohe Temperaturen, Berlin, Springer,
1940.
2 Brenner, S. S.: J. Electrochem. Soc. 102, 16 (1955).
252 4. Scaling Processes with Formation of Thick Protective Layers

• f\
I

/' Fig. 123. Rate of oxidation of a 20% Mo-Fe--N

I \ alloy at 1000°C in oxygen at 1 atm (21iters/min)


as a function of nickel content, according to Brenner.

o
o
~V
J)
Weigh t t Ni -
"---
I(J
t--
fj
(The initial alloy consisted of 20 wt. % Mo and
80 wt. % Fe, which was replaced by an increasing
content of nickel.) The weight loss of metal in
mg/cm 2 after an oxidation time of 2 hr was plotted.

react with the chromium and iron along the alloy/scale phase boundary
in the following way:
Mo0 3 + 2Fe~Mo + Fe 0 2 3

Mo0 3 + 2Cr ~ Mo + Cr203

This reaction was promoted by the fact that liquid Mo0 3 probably dissolves
in the iron and chromium oxide which is formed.
The formation of chromium-rich oxide layers was also observed by
Fontana.l As mentioned above, in molybdenum-free alloys the Mo03 or
V20 5 which cause the catastrophic oxidation can be introduced from the
outside via the gas atmosphere, as was seen, for example, in the case of an
8% Al-92% eu alloy by Meijering and Rathenau. These experimental results

Weight% Ni

Fig. 124. Oxidation diagram of iron-


molybdenum-nickel-chromium alloys,
according to Brenner. The numbers
indicate the weight loss in mg/cm 2 after
an oxidation time of 2 hr at 1000°C.
The shaded areas indicate the region of
catastrophic oxidation.
We ight'YoCr

1 McCullough, H. M., M. G. Fontana, and F. H. Beck: Trans. ASM 43, 404 (1951).
4.3. Catastrophic Oxidation 253

are now of the greatest technological significance, because some fuel oils
contain vanadium oxides, which appear in the combustion gases in the form of
V2 05 and can produce destructive scaling by the mechanism discussed
above. In agreement with the above results, Preece and co-workers! found
that alloys whose heat stability depends on chromium are very susceptible
to V 20 5 attack at high temperatures. Therefore on applications where the
alloy will be exposed to gases containing V 205, chromium should be replaced
by some other metal which is insensitive to V 205. In selecting the substitute
metal, a high melting point with V 20 5 is to be regarded as the prime criterion.
In gas turbines for example, phenomena of this type were described by
Schlapfer, Amgwerd, and Preis 2 as well as by Sykes and Shirley.3 On the
other hand, is seems that titanium and titanium alloys should be plating
materials that, at least at the outset, would be insensitive to the V2 05 in
combustion gases since, first of all, the V 20 5 would probably "cement"
the nonadhering, partially porous protective layer that appears at high
temperatures, whereby the diffusion via pores is prevented, and thus
the oxidation rate is lowered through a decrease in the number of vacancies
(in this case 0 2- vacancies). Experiments in this direction have not yet been
carried out. However, it has been shown that alloying additions of vanadium
to titanium produce a slight increase in the oxidation rate (see Fig. 103).
These considerations yield important suggestions for the construction
of industrial furnaces. By choosing suitable oxides as insulators for resistance
wires, catastrophic oxidation can be reduced, and in addition the resistance
to oxidation and the lifetime of the wires can be substantially improved.
The cause of catastrophic oxidation described above is only one-even
though especially pronounced-of the possible causes.

4.4. Scaling Systems with Rate-Determining Phase-Boundary


Reactions
When the rate-determining factor in the oxidation of a metal or an
alloy is a phase-boundary reaction a linear rate law is in effect, i.e., the
amount of oxygen consumed-or the increase in thickness of the scaling
layer-is proportional to time. This rate law holds for the formation of
porous protective layers up to any desired thickness since, in this case, the
~ttacking gas can diffuse rapidly to the reaction front. However, for the

1 Lucas, G., M. Weddle, and A. Preece: J. Iron. Steel Inst. 179, 342 (1955).
2 Schlapfer, P., P. Amgwerd, and H. Preis, Schweiz. Arch. angew. Wiss. u. Tech. 15, 291
(1949).
3 Sykes, C., and H. Shirley: Symposium on High-Temperature Steels and Alloys Jor Gds
Turbines, Iron Steel lnst., 1951, p. 153.
254 4. Scaling Processes with Formation of Thick Protective Layers

formation of compact, i.e., nonporous, protective layers, the linear rate law
is applicable only to the point where the diffusion rate, which decreases with
increasing layer thickness according to a parabolic rate law, becomes com-
parable with the rate of the phase. boundary reaction. With further growth
of the protective layer, the diffusion rate becomes the rate-determining
factor as can be seen by the "pure" parabolic rate law that begins to operate.
At lower temperatures, (e.g., 20°C) a parabolic rate law has been ob-
served on occasion with protective porous layers. In bromination experiments
with silver in bromine-containing bromide solutions a parabolic law was
found by Jaenicke l and Pfeiffer. 2 On basis of simultaneous conductivity
and potential measurements on suitable electrochemical cells, this can
be satisfactorily explained only if one assumes a rate-determining diffusion
of silver ions along grain boundaries and through pores of the AgBr layer.
Evans 3 and Birchena1l4 were able to show that, in certain cases of a
seemingly compact protective layer the parabolic law can be replaced by a
linear one. This is true especially at low temperatures, where because of the
lack of a sufficiently rapid plastic flow an accumulation of vacancies takes
place. Details of this mechanism have been discussed in the previous chapter,
but, in the following we will not consider this phenomenon, whose explana-
tion has still not been sufficiently well developed.
If onc follows a general discussion by Fischbeck5 on the temperature
dependence of the diffusion resistance WD and the reaction resistance WR
(see Fig. 125), or their equivalents, the diffusion coefficient and the rate

Fig. 125. Schematic representation of the


superposition of diffusion resistance W D and
1 reaction resistance W R in logarithmic represen-
r - tation.

1 Jaenicke, 'V.: Z. Elektrochem. 55, 186 (1951).


2 Pfeiffer, r., K. Hauffc, and W. Jaenicke: Z. Elektrochem. 56,728 (1952).
3 Evans, U. R.: Trans. Electrochem. Soc. 91, 547 (1947).
4 Birchenall, C. E.: jictallurgical Rep. I, Princeton Univ. Rep. Control No. OSR-TN-
54-286.
5 Fischbeck, K., L. Neundeubel, and F. Salzer: Z. Elektrochem. 40, 517 (1934); K.
Fischbeck and F. Salzer: l~Ietallwirtschaft 14, 733 (1935).
4.4. Scaling Systems with Phase-Boundary Reactions 255

of the phase-boundary reactions, it immediately becomes clear that in a


certain range of layer thicknesses the decisive influence is the temperature-
regardless of whether the diffusion or phase-boundary reaction is rate-
determining. Furthermore, it can easily be seen that the parabolic rate law
loses its physical sense when dealing with very small layer thicknesses-
.dg -+ O-and sufficiently high temperatures should be chosen, so that the
influence of electrical space-charge fields and their transport-promoting
effect can be neglected to a first approximation.
This critical temperature region-whether the linear or the parabolic
rate law is rate-determining-differs from one scaling system to another.
Frequently the linear rate law is replaced by the parabolic one so near
the very beginning of the reaction that it is not observable with normal
experimental methods. For the oxidation of Fe to FeO in a CO-C0 2 atmos-
phere below 900°C, for example, Hauffe and Pfeiffer1 found a parabolic
rate law, and above 900°C a linear one, in agreement with the qualitative
scheme in Fig. 125. While this explanation of the reaction mechanism of
diffusion-controlled solid body reaction was possible in many cases, the
explanation of the mechanism is made much more difficult in phase-boundary
processes. Jost 2 and Wagner3 noted this situation some time ago. The
explanation of the mechanism of the metal oxidation with a rate-determining
phase-boundary reaction is satisfactory only in isolated cases. The cause
for this is to be sought simply in the fact that the phase-boundary reaction
can be determined by any of the following steps:
1. Splitting up of the oxygen molecule and chemisorption of the atoms
at the MeOj02 boundary surface.
2. Introduction of the chemisorbed oxygen ions into the MeOj02 phase
boundary.
3. Passage of metal ions from the metal or alloy phase into the oxide
lattice and also, if necessary, the reaction with the oxygen arriving
there in the case of predominant oxygen diffusion.
4. If several oxide phases make up the protective layer, the transfer
of metal or oxygen ions from one oxide phase into the other or the
change in charge of the ions upon transfer into a neighboring phase.
5. Nucleus formation and crystal growth.
If in the course of the reaction no new phases appear, which is the case
in heterogeneous catalyzed reactions and in the processes that were used,
for example, in the carburization and decarburization of metals in CO-C02
1 Hauffe, K., and H. Pfeiffer: Z. Metallk. 44, 27 (1953).
2 Jost, W.: Diffusion und chemische Reaktion in feeten Stoffen, Dresden, 1937, p. 133.
3 Wagner, C.: "Chemische Reaktion der Metalle," in Handbuch Metallphysik, Vol. I,
2, Leipzig, 1940, p. 139.
256 4. Scaling Processes with Formation of Thick Protective Layers

gas mixtures or the solution of gases in metals, then the explanation of the
decisive phase-boundary reaction is considerably simpler and frequently
given.
Of the older works that first considered the total phase-boundary
reaction, that of Wilkins and Rideal 1 may be noted, in which phase-boundary
reactions were recognized as decisive on basis of the oxidation of copper
with time at 200°0 and different oxygen pressures. From this view Benard
and Talbot2 tried to see how far the parabolic rate law is valid in the oxida-
tion of iron in the first minutes of the reaction. For this purpose an iron foil
of 0.3 mm thickness and 14 cm 2 surface was brought into the reaction zone
by means of a rotating cylinder (rotation rate: 0.4 mm/sec). The curves ob-
tained under these conditions between 850 and 1050°0 in air at 1 atm can be
divided into the following three periods (Fig. 126):

8
V-
I
ng !cm2
V
I
6

I 8

2
1/
V
o

o
--r
20 1;0
I
80 BOsec m Fig. 126. Course of the oxidation of iron at 950°C in the
first stage of the reaction, according to Benard and
Time- Talbot.

1. Period OA where thermal equilibrium sets in. With the sheets which
were used, it is attained after about 20 sec.
2. The linear period AB, also with a duration of about 20 sec.
3. A period which essentially obeys the parabolic rate law.
Therefore there exists a transition period during which the oxidation
rate is independent of any diffusion processes. This situation remains in
force as long as the rate of the processes at the phase boundary is slower
than the initially very rapid diffusion rate, which decreases with the increase
in thickness of the oxide layer, and finally after a definite reaction time or a
critical layer thickness becomes the slowest and therefore the rate-determining
1 Wilkins, F. J., and E. K. Rideal: Proc. Roy. Soc. (A) 128,394 (1930); F. J. Wilkins:
Proc. Roy. Soc. (A) 128,407 (1930).
2 Benard, J., and J. Talbot: Compt. rend. 226, 912 (1948).
4.4. Scaling Systems with Phase-Boundary Reactions 257

step. Benard and Talbot found the transition from a pure phase-boundary
reaction to diffusion of iron ions at 850°0 with a thickness of about 1.3 I-'
and 950°0 at about 7 1-" The slope ofthe linear part of the curve as a function
of temperature is a measure of the true activation energy, which corresponds
to the oxidation of oc-iron and is displaced by a limiting diffusion. Its value
of about 59 kcaljmole is about the same as the enthalpy offormation of FeO,
which is 63 kcaljmole. On the other hand, the energy increment obtained
from oxidation experiments with rate-determining diffusion was only
36.6 kcaljmole,l but as a consequence of these considerations, this large
difference between the two values is not surprising, since in the first case
we are dealing with a real phase-boundary reaction and in the second with a
diffusion.
In the theoretical consideration of the competition between diffusion and
phase-boundary reactions-each of which can be rate-determining under
certain experimental conditions as was indicated for the oxidation of iron
and copper-a rate law of the following kind is to be used, according to
Evans,2 Fischbeck,3 Jost,4 Wagner and Grunewald,5 and Noldge: 6

Llm.~+(Llm)2._1_. =t (4.78)
q l" q kIf

Here l" (gjcm 2-sec) is the reaction constant of the phase-boundary reaction
and kIf the often-mentioned, practical scaling constant. Because of its
character this rate law can naturally be used only as a guiding principle.
For the direct evaluation of the measured results we divide equation
(4.78) by LJmjq and obtain a linear relationship between LJmjq and tj(LJmjq):
1 LIm 1 t
F + -q-7l' = Llm/q (4.79)

Therefore, from a plot of Jmjq VB. tj(Jmjq), we obtain the reciprocal


value of the practical scaling constant kIf (in g2jcm4-sec) and the reciprocal
value of l" from the slope of the straight line and from the intercept of the
straight line with the ordinate axis (Jg = 0), respectively.
This is how the rate constants l" recorded in Tables 23 and 26 were
calculated by Wagner and Grunewald for the phase-boundary reaction in the
oxidation of copper and nickel. Even without a lengthy discussion of the
mechanism one can recognize from the values in the tables, as well as from
the curves in Fig. 62, that the rate constants l" or l, where l = l" jAo (Ao = 8,
1 Benard, J., and O. CoquelIe: Compt. rend. 222, 796 (1946).
2 Evans, U. R.: Trans. Electrochem. Soc. 46, 247 (1924).
3 Fischbeck, K.: Z. Elektrochem. 39, 316 (1934).
4 Jost, W.: Diffusion und chemische Reaktion in festen Stoffen, Dresden, 1937, p. 31.
5 Wagner, C., and K. Griinewald: Z. physik. Chern. (B) 40, 455 (1938).
6 Niildge, R.: Physik. Z. 39, 546 (1938).
258 4. Scaling Processes with Formation of Thick Protective Layers

the equivalent weight of oxygen), increase with increasing oxygen pres-


sure. Wagner suspected that with high oxygen pressures the transition
of metal ions from the metal into the oxide phase is to be regarded as the
slowest of the phase-boundary reactions mentioned above, while at low
oxygen pressures the reaction at the exterior metal oxide/oxygen phase
boundary is to be regarded as the rate-determining process because of
chemisorption and splitting up of oxygen molecules into atoms and intro-
duction of these as ions into the oxide lattice. In the latter case the reaction
rate must also depend on the oxygen pressure, as will be discussed in greater
detail for the oxidation of iron.
Oxidation experiments carried out on iron by Fischbeck and co-workers 1
between 850 and 1000°0 in the Fe 2 03 phase region show that in the neighbor-
hood of the oc-y transition point (about 900°C) a definite break in the oxida-
tion time curve appears, which can be seen in Fig. 127. Here oc-iron oxidizes

20
~.cm-J.miri1 !
V
~

«-Iron! \rlron
5 I
I
I I
:
o 800 goO 1000 C
0 Fig. 127. Change of the rate of oxidation of iron in the
transition from the Q( to the y phase, according to Fischbeck
Temperature - and Benard.

more rapidly than y-iron. These results were confirmed by Benard and
Talbot,2 who found, moreover, that the discontinuity in the oxidation curve
of oc- and y-iron does not coincide with the crystallographic transition point
of iron, but is somewhat higher (about 905-910°C).
The discontinuity appearing around 900°C in Fig. 128 can be recognized
in the break in the oxidation-time curve, as it was obtained from oxidation
experiments with iron sheets in a CO-C02 mixture (30 vol. % CO) in the
FeO-phase region by Hauffe and Pfeiffer. 3 In contrast to the parabolic
oxidation in air, a linear rate law prevails here. The explanation given by
Fischbeck and co-workers based on a rate-determining phase-boundary
reaction at the Fe/FeO phase boundary-whereby the transfer of iron ions
plus electrons from oc~iron into the FeO phase should proceed more rapidly
1 Fisehbeck, K., L. Neundeubc1 and F. Salzer: Z. Elektrochem. 40, 517 (1934); K.
Fisehbeck and F. Salzer: Jletallwirtschaft 14, 733 (1935).
2 Denard, J., and J. Talbot: Compt. rend. 226, 912 (1948).
3 Hauffc, K., and H. Pfeiffer: Z. clletaUk. 44, 27 (1953).
4.4. Scalin~ Systems with Phase-Boundary Reactions 259

than from y-iron-can only be considered if a linear rate law of oxidation


prevails, as was found, for example, for FeO formation on iron. If a parabolic
rate law is observed, this explanation cannot be supported, since, as has
been shown, a parabolic law precludes a rate-determining phase-boundary
reaction. Nevertheless we will show that the discontinuity appearing at


""
]'
-~5r-----~----~~--~----~
Fig. 128. Temperature dependence of the
rate of oxidation of iron at cohstant
oxygen partial pressure (70 vol. % C02
+ 30 vol.% CO), according to Hauffe and
Pfeiffer. A bend in the curve appears at
about 900°C.

about 900°C cannot be caused by the transfer of iron ions from the phase
even with the existence of a linear rate law. For this purpose we will acquaint
ourselves in greater detail with phase-boundary reactions, using the oxida-
tion of iron to FeO as a case in point.
If a linear rate law is observed then, as already noted, one of the phase-
boundary reactions is generally rate-determining for the total reaction.
In the presence of a p-type oxide protective layer, e.g., FeO, the oxidation
of a metal must always be oxygen-pressure-dependent (as long as nucleus
formation and crystal gro-w-th are not determinative). It is immaterial at which
of the two phase boundaries the reaction takes place. On the other hand,
the oxidation of a metal whose oxide is an n-type semiconductor is only
oxygen-pressure-dependent when the rate-determining step proceeds at the
oxide/gas phase boundary. Since a parabolic rate law prevails for the oxida-
tion of iron in the presence of an FeO layer covered with an Fea04layer, where
the highest possible oxygen pressure for the FeO formation (i.e., the decom-
position pressure of the Fea04) is rate-determining, but a linear rate law
prevails for the oxidation in the absence of the covering Fea04layer, although
the oxygen pressure remains approximately equal to the decomposition
pressure of Fea04, FeO formation in CO-C02 mixtures is given by the
following chemisorption:
(4.80)

As the curve in Fig. 129 shows, the oxidation rate in the presence of an
Fea04 layer (oxidation experiment at 1 atm) is considerably more rapid
260 4. Scaling Processes with Formation of Thick Protective Layers

than the oxidation rate without an Fe304 layer at approximately the same
effective oxygen pressure (about the decomposition pressure of the Fe304).
This characteristic behavior in the case of the oxidation of iron to FeO
according to a linear rate law-which is the case above 900°C-precludes
the possibility of the rate-determining transfer of iron to the oxide, since
otherwise with practically the same oxygen pressure at the FeO/Fe304 or

60
g,'cm' fo,=1.1at

f Fig. 129. Course of the oxida-


tion of iron at 1000°C, according
to Hauffe and Pfeiffer. Parabolic
15 course of the oxidation in pure
1'0, .e.21·1O-"at oxygen at 1 atm and linear
course in a C02-CO mixture
with an oxygen partial pressure
o 45. 5.0 7.5 10,0 approximately equal to the
Duration of Oxidation decomposition pressure of Fe a04.

FeO/C02-CO phase boundaries approximately equal oxidation rates would


be found and, furthermore, the oxidation rates for the higher oxides that
follow FeO would obey the linear rate Jaw as well. The reaction-limiting
chemisorption according to (4.80) should be stopped by the Fe304 layer
which is simultaneously formed at high oxygen pressures.
After this consideration only the hypothesis of the rate-determining
step given in equation (4.80) remains. The activation energy which is rate-
determining for this process is, in the case of oxidation in air or oxygen
at 1 atm, replaced by another, lower value. Further experiments have shown
that no transition from a linear to a parabolic rate law takes place when the
oxygen equilibrium pressure Fe304/FeO (decomposition pressure of the
Fe304) is attained. The formation of the Fe304 phase does not begin until a
considerably higher oxygen pressure is used, as oxidation experiments at
lOOO°C with CO-poor gas mixtures (lO and 1 vol.% CO) have shown. The
formation of Fe304 under the previously mentioned oxygen partial pressures
was not observed until the iron sheet was completely oxidized to wustite
(FeO). It follows that in the presence of iron, because of the relatively high
diffusion rate of iron ions through the FeO layer to the FeO/gas phase
boundary, the chemisorbed oxygen formed according to equation (4.80)
is continuously consumed by iron ions arriving at the surface, so that under
these experimental conditions the chemical potential of the chemisorbed
oxygen ions required for the formation ofFe304, which is determinative for
the Fe304 formation, was not attained, even though the chemical potential
4.4. Scalin~ Systems with Phase-Boundary Reactions 261

or the partial pressure of the oxygen in the gas phase was quite sufficient.
The molecular oxygen present in the C02-CO mixtures certainly has
no influence on the oxidation rate, since its concentration on the FeO surface
is vanishingly small compared with that of the CO 2. Furthermore, the high
dissociation energy of 5.05 eV compared with 2.9 eV for CO 2 is an argument
against any such influence. As can be inferred from the oxidation experi-
ments depicted in Fig. 130, the oxidation rate follows the third root of the
oxygen pressure:
peo
1" = const ( --'- )2/3 = const*. /
p~ 3 (4.81)
Peo '

The empirically determined relationship (4.81) can be approximately


valid, of course, only within certain limits, since l" will necessarily equal
zero at an oxygen pressure which corresponds to the equilibrium pressure
at the Fe/FeO phase boundary. According to Fig. 130 the relationship iR
valid in the oxygen-pressure region of po, = 1.6 X 10-13 to 9.8 X 10-15 atm,
corresponding to the CO mole fraction of Xco = 0.2-0.5.

140

V'
110

./

./
/ '"
50
,;<1'
Fig. 130. Dependence of the reciprocal
value of the rate constants Ill" on the
reciprocal value of the C02-CO ratio 20 0
V
[1/(pco Ipco)] at 1000°C, according to
Hauffe 'and Pfeiffer: h = 0.042 and
"2 = 0.29.

Under the assumption of a rate-determining chemisorption of the C02,


whose slow rate is understandable because of the high concentration of the
holes in FeO(xe ~ 0.1), one can assume no significant space-charge layers
here because of the high mobility of the iron ions over vacancies and the
high hole concentration. The following relationship was obtained:
I I +~
1" 2 -;--(p-c-o',!'pco)
-;k- l'l (4.82)

where k1 and k2 (= k1ka) are constants. In Fig. 130 the experimental results
are plotted according to (4.82). The constants obtained were k1 = 0.042
and k2 = 0.29. The FeO formation investigated by Hauffe and Pfeiffer was
262 4. Scaling Processes with Formation of Thick Protective Layers

probably the first case where phase-boundary processes-especially the


chemisorption of CO 2 and the introduction of the chemisorbed oxygen into
the FeO lattice-rather than diffusion procf:)sses, are rate-determining above
900°C with a detectable compact protective layer. This finding is under-
standable on basis of the high iron ion vacancy concentration from 7-11 at. %
in the FeO lattice at 1000°C. The FeO lattice can be thought of as a sieve,
with many vacancies over which the transport of the iron ions is relatively
rapid. Since such a high lattice-defect concentration has not been observed in
other oxides, it is evident that this type of phenomenon is not operative in
the formation of a compact protective layer in other scaling systems.
Further considerations and additional experiments with this reaction
have been carried out by Wagner and co-workers'! According to these
authors we obtain the following expression for the oxidation rate:

l" = k1Pe0 2 - k2Peo


or
l" = k1P(1 + K) [xeo. - xeo.(equilibriuffi)]
where K = k2/kl = peo,/peo with l" = 0 and P = Pe0 2 + Peo with the
mole fraction x of CO 2 in the gas phase Peo. = PXe02 and peo = P(l - xeo.).
In agreement with this equation, it was observed that l" is proportional to
the sum of the partial pressures,P = Pe0 2 + Peo, in a pressure range between
0.4 and 1 atm.
Pfeiffer and Laubmeyer 2 investigated the oxygen-pressure dependence
of the oxidation of Fe to FeO in the oxygen pressure region from 10-3 to
1 mm Hg (here only FeO is formed due to the kinetics). They found the
following pressure dependence for the linear oxidation rate l", which, more-
over, was of the same order of magnitude as that with C02/CO mixtures:
l" = const. pO.O7
2

This finding is further evidence for the earlier assumption that only the
oxygen obtained from the chemisorption of C02 according to (4.80) is rate-
determining, and not the oxygen partial pressure. As soon as oxides higher
than FeO are formed, the oxidation follows a parabolic rate law and is
independent of the gas pressure.
The observation of the increase in thickness of the individual oxide
layers FeO, Fe304, and Fe203 by Benard and Coquelle 3 is noteworthy. It
was observed that the Fe304 and Fe203 layers grew according to a linear
rate law during oxidation of iron at 1 atm oxygen or air between 700 and
1 Wagner, C.: Mimeographed Notes, Course 83.23, Kinetics in ::\Ietallurgy, MIT, Cam-
bridge, 1955; F. S. Pettit, R. Yinger and J. B. Wagner: Acta M:et. 8, 617 (1960).
2 I)fciffcr, R., and C. Laubmcyer: Z. Elektrochem. 59, 579 (1955).
3 Benard, J., and O. Coquelle: Compt. rend. 222, 884 (1946).
4.4. Scaling Systems with Phase-Boundary Reactions 263

1000°C, while the FeO formation could be described by a parabolic rate law.
The finding of a linear growth of the Fe304 and Fe203 layers with time is
then understandable only when both of these oxide layers are porous.
A linear growth rate for these layers could not be confirmed by Himmel,
Mehl, and Birchenall.l
As indicated above, we also find a transition from the parabolic to the
linear rate law and vice versa in other scaling systems, for example, in the
oxidation of titanium (see Section 4.2.2),2 germanium,3 silicon,4 cerium,5
thorium,6 and in sulfide formation on nickel.? In all these cases, except that
of germanium oxidation, we still lack an adequate explanation for the
mechanism of the phase-boundary reaction which is determinative for the
linear rate law. The oxidation of germanium between 575 and 705°C, which
was investigated by Bernstein and Cubicciotti,3 is quite interesting because
the GeO which is predominantly formed at these temperatures is relatively
volatile, and the oxidation rate can be determined by the weight decrease
with time. From this we obtain the fact that the vaporization of GeO is the
rate-determining step for the oxidation. The rate law reads
no, = nv[l - exp( -kvt)]
where no, is the number of moles of oxygen consumed per square centimeter
of metal surface and kv is a constant. The quantity nv is approximately
proportional to the reciprocal value of the oxygen pressure.
It has been found that arsenic is oxidized according to a linear rate
law. 8 A total vaporization of the Mo0 3 layer formed in the oxidation of
molybdenum was observed when the oxidation temperature was above the
melting point of Mo03 (795°C).9 However, since the linear rate law of the
oxidation is also found below the melting point, the vaporization cannot
determine the kinetics over the entire range of possible temperatures.
Nothing can be said at this time about the mechanism at lower temperatures.
Frequently a linear oxidation law is found to result from formation of
quite porous coatings. Porous coatings preferentially appear where the molar
volume of the reaction product being formed, e.g., an oxide, is smaller than
the atomic volume of the underlying metal (or the molar volume of the

1 Himmel, L., R. F. Mehl, and C. E. Birchenall: J. Metals 5,827 (1953).


2 Davies, M. H., and C. E. Birchenall: J. Metals 3, 877 (1951).
3 Bernstein, R. B., and D. Cubicciotti: J. Am. Chem. Soc. 73, 4112 (1951); Law, J. T.,
and P. S. Meigs: J. Electrochem. Soc. 104, 154 (1957).
4 Brodsky, M. B., and D. Cubicciotti: J. Am. Chem. Soc. 73, 3497 (1951).
5 Cubicciotti, D.: J. Am. Chem. Soc. 74, 1200 (1952).
6 Levesque, P., and D. Cubicciotti: J. Am. Chem. Soc. 73, 2028 (1951).
7 Hauffe, K., and A. Rahmel: Z. physik. Chem. 199, 152 (1952).
8 Kalman, L.: Magy. Kem. Folyoirat. 57, 65 (1951).
9 Lustman, B.: Metal Progr. 57, 629 (1950).
264 4. Scalin~ Processes with Formation of Thick Protective Layers

underlying compound). This kind of ratio of the atomic and molar volumes
of the initial and fin al materials is found with the alkali and alkaline earth
metals and with magnesium. On this basis it is also understandable that the
oxidation of barium (Fig. 131)1 and of magnesium (Fig. 132)2 follows a linear

1'2 3
9-em ·10 /
V
20

15 .I

/V
V
.#
'0
/1 Fig . 131. Course of the oxidation of barium at
0
17 °C in air at 71.5 % relative humidity, according
o 100 2tJ(J IIUn to Pilling and Bedworth.

rate law . However, at lower temperatures magnesium behaves differently


than one would expect on basis of the volume ratio . In spite of the high
enthalpy of formation ofMgO of -146.1 kcal/mole, magnesium is remarkably
resistant to oxidizing atmospheres at room temperatures. Finch and Quarre1l3

500

O H----+--/'9iF--/-t-l
Fig . 132. The temperature dependence of the
......
~
linear scaling constant 1"(mg/cm 2 ·hr) in the
F oxida tion of magnesium, according to Leontis
-1 rr--~~--~-r---1-1 and Rhines.
o = Rolled magnesium
oxidized in pure
oxygen
• = Initial rate
-2 t7
"O*""""''--~
1.32~L..L-:f.2/I:7.-..L.J.-t:-':'8~-'
o = Rate toward the } Magnesium, oxidized
end of the in dry a ir
f ·tg1- oxidation

attempt to base this behavior on the oriented growth of MgO on the metal
in thin layers in an unstable modification, the eventual formation of compact
layers. Evidently, when a critical thickness of the protective layer is reached,
1 Pilling , N. B., a nd R. E. Bedworth: J . Inst. Metals 29,529 (1923).
2 Leontis, T. E. , and F . N . Rhines: Trans. A/ME 166, 256 (1946); Svec, H. J., and
D. S. Gibbs : J. Electrochem. Soc. 104,434 (1957).
3 Finch, G. J ., and A. G. QuarreIl: Nature (London) 131, 877 (1933); Proc. Roy. Soc. (A)
141, 398 (1933).
4.4. Scaling Systems with Phase-Boundary Reactions 265

the forces in the MgO lattice are sufficient to produce the normal cubic
structure, then the protective layer breaks open. Leontis and Rhines l find
an activation energy of 50.5 kcal/mole for the pure chemical reaction which
set in between 475 and 575°C. Above 600°C the reaction proceeds very
rapidly and leads to self.ignition because of the relatively high vaporization
rate of magnesium. The compact coating appearing at the beginning of the
oxidation below 450°C is determined by a parabolic oxidation and has also
been described in the literature. 2 Paidassi et al. 3 was able to confirm these
results. A linear rate law governs between both the 350-550° interval with
an activation energy of 26.2 kcal/mole and the 550-600° interval with an
activation energy of 45.2 kcal/mole.
It follows, then, that even the crystal orientation of the coatings growing
on a metal or an alloy can decisively influence the oxidation rate, even when
we no longer observe a linear rate law, as is the case, for example, in the
oxidation of copper at intermediate temperatures. Here Benard and Talbot 4
found a dependence of the oxidation rate on the crystallographic directions
of the Cu 20 crystals growing on the copper. Similarly, Benard and Bardolle,5
in their study of the oxidation of nickel alloys, and Gwathmey and Benton,6
Lustman and Mehl, 7 and Gulbransen and Ruka 8 were also able to demonstrate
the dependence of the oxidation rate on crystal orientation for other systems.
A review article on this subject has been published by Gwathmey.9
The flow velocity of the attacking gas and its influence on the oxidation
rate is another experimental criterion for the recognition of phase. boundary
processes. Baukloh and Reif lo report that the oxidation rate of ARMCO· Fe
depends on the flow rate. Murphy, Wood, and Jominyll observed a certain
critical flow velocity in the oxidation of soft steel up to which the oxidation
rate increases and above which it remains constant. Schroederl2 was able to
confirm this observation. Transport phenomena in the gas phase evidently
playa decisive role in this case.

1 Leontis, T. E., and F. N. Rhines: Trans. AIME 166, 256 (1946).


2 Gulbransen, E. A.: Trans. Electrochem. Soc. 87, 589 (1945).
3 Darras, R., J. Paidassi, and F. Leroy: Compt. rend. 254, 869 (1962).
4 Benard, J., and J. Talbot: Compt. rend. 225, 411 (1947).
5 Benard, J., and J. Bardolle: Compt. rend. 232, 231 (1951); 239, 706 (1954).
6 Gwathmey, A. T., and A. F. Benton: J. Phys. Chem. 46, 969 (1942).
7 Lustman, B., and R. F. Mehl: Trans. AIME, Techn. Pub!. No. 1317 (1941).
8 Gulbransen, E. A., and R. Ruka: J. Electrochem. Soc. 99, 360 (1952).
9 Gwathmey, A. T., and K. R. Lawless: in The Surface Chemistry of Metals and Semi·
conductors, Wiley and Son, New York, 1960, .p. 483.
10 Baukloh, W., and O. Reif: Metallwirtshaft 14, 1055 (1935).
11 Murphy, D. W., W. P. Wood, and W. E. Jominy: Trans. Am. Soc. Steel Treating 19,
193 (1931).
12 Schroeder, W.: Arch. Eisenhuttenw. 6, 47 (1932).
266 4. Scaling Processes with Formation of Thick Protective Layers

In sulfidation experiments with copper and silver, Wagner and


co-workers1 found a noticeable participation of phase-boundary reactions
as rate-determining steps. During the reaction of silver with sulfur at 400°0,
the diffusion of silver ions through the Ag2S layer determines the rate of the
layer growth, while between 200 and 300°0 under the same experimental
conditions the rate of sufidation (also with formation of thick and compact
layers), is determined by the transfer of silver ions across the AgjAg 2S
boundary. This rate-determining transfer could be detected by the deviation
of the silver-ion concentration from the thermodynamic equilibrium in
Ag 2S at the AgjAg 2S phase boundary with the aid of a suitable electro-
chemical cell. However, at the Ag 2SjS phase boundary, equilibrium is
approximately established, and with a formula derived from the elementary
steps of the Ag 2S formation, the rate of the sulfidation up to very large
thicknesses (~5 em) can be calculated quantitatively (see p. 374).
Finally we may add still another interesting example, which is drawn
from the field of heterogeneous catalysis-the catalytic N 20 decomposition
with simultaneous oxidation of copper. As Dell, Stone, and Tiley2 have
determined, the reaction sequence on copper sheets covered with Ou 20 in
the N 20 decomposition is:

N 20(g) ---+ O-;;-hem + EEl + N~g) slow


O-;;-hem ---+ OU2 0 + 20uD' + EEl
Ou diffusion via vacancies
Ou(metal) + OuD' + EEl ---+ Null

in which the chemisorption is the slow step. Furthermore, the rate of decom-
position of N 20 is independent of the layer thickness of the OU20. On the
other hand, no gaseous oxygen was formed, that is, the desorption rate of
oxygen or the desorption reaction is negligible compared to the oxidation
rate or the rate of diffusion of the copper ions in OU20. As in the oxidation
of Fe to FeO in 00 2-00 mixtures, the rate of oxidation of copper is solely
dependent on the decomposition rate of the N 20. An increase in the hole
concentration in the Ou 20 (for example, through an oxygen pretreatment)
should increase the oxidation rate, since the N 20 decomposition was
increased. 3

1 Rickert, H.: Z. physik. Chern. [NFJ 23, 355 (1960): H. Rickert and C. Wagner: Z.
physik. Chern. [NFJ 31, 32 (1961).
2 Dell, R. M., F. S. Stone, and P. F. Tiley: Trans. Faraday Soc. 49, 201 (1953).
3 See also the review article by J. Block: "Phasengrenzreaktionen bei der "~1 etalloxydation,"
in: Passivierende Filme und Deckschichten, edited by H. Fischer, K. Hauffe, and W.
Wiederholt, Springer, Berlin 1956, pp. 129 ff.
4.5. Scaling Systems with Protective Layers Containing Several Phases 267

4.5. Scaling Systems with Protective Layers (ontaining


Several Phases
Theoretically it is not difficult to produce a one-phase homogeneous
layer on metal in the laboratory through a suitable choice of the partial
pressure of the scaling atmosphere (e.g., the formation of FeO on Fe and
Cu 20 on Cu or CoO on Co) but it is frequently not possible in industrial
processes, since the oxygen partial pressure, for example, in firing processes
and other reactions with an oxygen excess in the reaction gas, can often
reach 1 atm and higher. However, under these conditions we are in the
stable region of the highest oxidation state of the scaling product (for example,
the stable region of the Fe203 phase in the case of iron oxidation). Investiga-
tions have shown that under these conditions scaling layer consists of
several oxide phases either in an irregular arrangement or in a layering parallel
to the metal.
Under the assumption of a preferred parallel layering of the individual
oxides and sulfides in the scaling layer, in which the lowest oxidation state
is located next to the metal and the highest in equilibrium with the oxidiz-
ing atmosphere, the mechanism of the protective layer formation in a
few cases may be quantitatively described to a first approximation as far
as the oxidation rates of the individual oxides-for example Cu -+ Cu20
and CU20 -+ CuO or Fe -+ FeO, FeO -+ Fe304, and Fe304 -+ Fe203-are
known. Such investigations have been carried out for the oxidation of copper,
iron, and manganese, and also for sulfidation of iron, which will be described
in greater detail as examples of other scaling systems with multiphase
coatings.
If, for example, we introduce a copper sample at 1000°C into a gas
atmosphere whose oxygen partial pressure is greater than 100 mm Hg,
then we observe that first of all the oxidation rate may be described by a
parabolic rate law, and that secondly the oxide layer consists primarily of
CU20 and exhibits only a very thin exterior layer of CuO, although the oxygen
pressure prevailing during the oxidation lies in the stable region of the CuO
phase. The cause of a preferred Cu 2 0 formation lies in the kinetic processes
and has been qualitatively described by Wagner. l As seen in Fig. 133, a
very thin CuO layer was formed at the beginning of the oxidation by the
effect of oxygen on copper in the stable region of the CuO phase, which in
the progress of the oxidation was dissolved in the Cu 20 layer on the metal
side because of the large supply of copper ions and electrons. The consider-
ably greater diffusion rate of the copper ions via copper ion vacancies in
1Wagner, C.: "Reaktionen mit Metallen," in Handbuch der Metallphysik, Vol. I, 2,
Leipzig, 1940, p. 144.
268 4. Scaling Processes with Formation of Thick Protective Layers

the CU20 causes the slowly forming CuO layer to be continually consumed,
leaving only a very thin layer of constant thickness. The CuO layer can-
not grow further until all the copper is oxidized to Cu 20, that is, until no
more copper is present. The lack of dependence of the oxidation rate of
copper on the oxygen pressure found by Pilling and Bedworth,1 Feitknecht,2
and Frohlich3-when oxidation is in the CuO stable region-can be under-
stood because in an oxide layer series of CUjCU20jCUOj02, just as in the
Fig. 133. Scaling scheme for the
oxidation of copper to CU20
+ CuO in the stable region of
the CuO phase: (a) shows the
beginning of the oxidation,
a Field b Field where only a thin CuO layer
Transport O!!'uslon Tra!!!.port
%--
(~I/L) is present on the copper;
Positive (b) represents the structure of
Space the scaling layer after longer
Charges oxidation times. While the con-
centration of the copper ion
CU20 O/g) vacancies xCuo' at the CU20/Cu
phase boundary is n\lgligible in
CU20, it has the highest value
at the CU2o/euo phase boundary
(e.g., at 1000°C xCuo' > 1.2
x 10-3 ). The concentration of
the electron holes in the CU20
o layer follows the same course.

simple CU20 formation, the diffusion in the Cu 20 phase is solely rate-deter-


mining and, therefore, due to the prevailing constant CuO decomposition
pressure [po,(CuOjCu20)], the diffusion must be independent of the external
oxygen pressure. The concentration of the copper ion vacancies and holes in
the Cu 20 phase at the CujCu20 and CuOjCu20 phase boundaries is clearly
determined by the intermediate equilibria Cu + CU20 and Cu 20 + CuO.
Thus the chemical potential gradient of the copper or the concentration
gradient of the copper ion vacancies in the CU20 layer is constant, and this is
what is ultimately responsible for the diffusion rate of the copper ions in CU20.
Dravnieks 4 was able to demonstrate that the oxide layer which formed
during the oxidation of copper at 500°C at an oxygen pressure of 2 mm Hg
consisted of pure CU20 up to about 4 x 10-4 cm and above this thickness
was gradually covered with a growing CuO film. This finding does not directly
contradict the above results, if one considers that in this temperature region
we have already demonstrated the existence of space-charge layers and
field-transport phenomena, and we noted earlier that these can change the
structure of the scaling layer considerably.
1 Pilling, N. B., and R. E. Bedworth:J. Inst. Metals 29,529 (1923).
2 Feitknecht, W.: Z. Elektrochem. 35,142 (1929).
3 Frohlich, K. W.: Z. Metallk. 28,368 (1936).
4 Dravnieks, A.: Private communication.
4.5. Scalin~ Systems with Protective Layers Containin~ Several Phases 269

Under the assumption of a parabolic growth for the OU20 layer as well
as for the OuO layer, Wagner! and Valensi 2 calculated the constant thick-
nesses of the OU20 and OuO layers which form on copper during oxidation.
These formulas, which are valid for all scaling systems with parallel layering
of the individual oxide phases and parabolic oxidation, are not applicable
for the layer thickness calculation in the copper-oxygen system in that here
the oxidation of Ou 20 to OuO is not governed by a parabolic but by a cubic
rate law. 3 This rate law as weil as the logarithmic oxygen pressure dependence
of the oxidation rate seems to indicate a field transport of the ions or an
aging effect in the OuO layer, as was discussed in Section 3.5.3. The tempera-
ture dependence of the cubic scaling constants in the OuO formation at
1 atm oxygen between 500 and 1000°0 is reproduced in Fig. 134.

IiJ

t-
- 0 k...

U
~
....

~ .......
?-
.''5-?£
Fig. 134. Temperature dependence "b. " ~
of the rate of oxidation of CU20 to
CuO in oxygen at 1 atm between ' ........,
500 and 1000°C, according to Hauffe 16
and Kofstad. The sample was
annealed for 1 hr in a high vacuum
prior to oxidation-about 10-3 to
10-4 mm Hg; k is the cubic tarnish- 111, flO
r
ing constant in g 3 /cm6 .sec. -f-
If the assumption of a parabolic rate law for all oxide phases is fulfilled,
then a simple relation may be derived in somewhat more modified and
generalized form 4 than was done by Jost 5 and Valensi. 2 These relations,
given below, are based on the similarity of the defect and the /LA distributions
in each individual oxide layer with variable ~ values and on the assumption
that at the boundary of two partial layers values of /LA and /LB prevail
which are independent of the overall ~ value-for example, it is given by the
equilibrium oxygen pressure (decomposition pressure) of the coexisting
oxide phases. This relation is not only fulfilled if the metal, on the one hand,
1 Wagner, C.: "Reaktionen mit Metallen," in Handbuch der MetallphY8ik, Vol. 1,2.
Leipzig, 1940, p. 144.
2 Valensi, G.: International Oonference on Surface ReactionB, 1948, p. 156.
3 Hauffe, K., and P. Kofstad: Z. Elektrochem. 59, 399 (1955).
4 Hauffe, K., and W. Schottky: Halbleiterprobleme, Vol. V, Braunschweig, 1960. pp.
258-267.
5 Jost, W.: DiffUBion und chem. Reaktion in futen Stoffen, Dresden, 1937, pp. 162 and
167.
270 4. Scalin~ Processes with Formation of Thick Protective Layers

and the oxidizing gas with a constant pressure, on the other, are given,
but also in the multilayer case if the oxide layer, is composed of different
oxides. If this condition is fulfilled, similarity relationships are also valid for
the relative thicknesses of the partial layers in the steady state. The transport
currents 8A and 8B flowing through the individual layers are inversely pro-
portional to the layer thickness in the steady state and thus also inversely
proportional to the number of nM of lattice molecules per unit area in the
corresponding layer. The A and B currents are generally not equal in the
individual layers ; this difference in the currents can be used to convert lat.tice
molecules of one of the neighboring phases into lattice molecules of the other.
As a simple example of a multilayer process, we shall consider an
arrangement
I II III
A (metal) I AB I AB21 B2 (gas)
I 2

such as is realized, for example, in the action of sulfur vapor on iron within
certain ranges of pressure and temperature. If the sulfur transport is assumed
to be negligibly small, i.e., SB ~ 0 in both phases, then there can be no AB
formation at the AjAB phase boundary but only at boundary II. At the
gas boundary III the AB2 lattice growth is completely determined by the
A current 8A(2) through layer 2. At phase boundary II the formation of AB
independently of the migration mechanism of the A particles involves only
the A particles which do not migrate noticeably further because of the
displacement reaction
A(l) + AB2 (2) --+ 2AB(1) (4.83)
Since the number of A particles becoming available is determined by
8A(I) -8A(2), we obtain the total rate of formation of AB particles according
to equation (4.83):

dn M1I ) )
(- -- = 2(8A(I) - 8A(2») (4.84a)
dt total

dnM(2»)
(- -- = 2SA(2) - SA(I) (4.84b)
dt total

where dnM(l) and dnM(2) are the changes in the number of molecules per unit
area in layers 1 and 2, respectively, which are formed or created or dis-
integrated. Then we get for the ratio of the SA currents

with
4.5. Scaling SysteIlls with Protective Layers Containing Several Phases 271

In the steady state the ratio dnM(2)/dnM(1) = nM(2)/nM(1) = const := rp and


we obtain from equations (4.84)
U12/ rp-l/2 U12 - rp/2
rp= (4.85b)
1 - U12/rp rp - U12
This is a quadratic equation in rp; for U12 ~ 1, the expression simplifies to
(4.85c)
From this it follows that the quantitative ratio of lattice molecules in the
two phases is determined to within a numerical factor of their Tammann
scaling constants k'; the ratio of the partial layer thicknesses is, according to
equation (4.85c),
(4.85d)
This relation could be verified experimentally if, under a given external
sulfur vapor pressure, one measured the thickness ratio of the two partial
layers on the one hand and k~ and k~ on the other with the /LFe or /LS values
which are valid for the double-oxide-Iayer system. The latter can be deter-
mined for the partial layers by taking dgjdt measurements on the individual
layers. Such measurements have in fact been carried out by Hauffe and
Rahmel,l and Meussner and Birchenall.2 Since in equation (4.85d) N2 = N 1 ,
we obtain the ratio of the FeS-FeS2layer:

gFeS,/gFeS = 2kFeS,/kFes = 4.2 X 10-4


with the actual result kFeS, = 0.7 X 10-11 for ps, = 1 atm and kFeS =
3.3 X 10-8 equiv/cm-sec for Ps, = 100 mm Hg and T = 670°C. Very thin
layers of FeS2 have in fact been observed. No quantitative verification is as
yet available, however.
Gurnick and Baldwin 3 studied the oxidation rate of manganese in air
between 400 and 1000°C, which may be represented by a parabolic rate law.
In spite of the thermodynamic possibility of formation of MnO, Mn a04, and
Mn20a, the oxide layer at 900°C consisted almost exclusively of Mn a04
(see Fig. 135). The portion of MnO in the scaling layer did not increase with
increasing temperature until above 900°C. By HOO°C it was already about
20%. New investigations have shown that between 400 and 850°C the scale
contains MnO, Mna04, and Mn02 and between 900 and 1200°C only MnO
and Mna04.4
1 Hauffe, K., and A. Rahmel: Z. physik. Chern. 199, 152 (1952).
2 Meussner, R. A., and C. E. Birchenall: Corrosion 13, 677 (1957).
3 Gurnick, R. S., and W. M. Baldwin, Jr.: Trans. ASM 42, 308 (1950); E. B. Evans,
C. A. Phalnikar, and W. M. Baldwin, Jr.: J. Electrochem. Soc. 103, 367 (1956).
4 Paidassi, J., and A. Echeverria: Acta Met. 7, 293 (1959).
272 4. Scaling Processes with Formation of Thick Protective Layers

The oxidation of solid and liquid lead involves another scaling system
where two oxide layers-PbO and Pb 3 0 4-can appear below 540°0 at 1 atm
oxygen, as Gruhll and Baldwin 2 were able to show. Below 400°0 one should

~
Mn,04 '" Fig. 135. Relative layer thicknesses of
I'
I
,
' )'. the MnO and Mna04 layer formed
during the oxidation of manganese
I ,-
" between 400 and 1200°C in air at
1 atm, according to Gurnick and
~' Baldwin. (The breadths of the region
of stability of the manganese oxides
'100 6aJ 8(j(j 1(}(J(j lztlQ"C are plotted above.)

also expect an additional Pb 2 0 3 formation, about which nothing is known at


this time. Besides these possibilities, below 486°0 a PbO layer (tetragonal
,a-PbO) was observed by both authors which can form from yellow (ortho-
rhombic IX-PbO) which is always formed at the beginning of the oxidation.
As can be seen from Fig. 136, Weber and Baldwin 2 found that the course

'I~
I
~

~/ ./
,..
u/1 .&- V '"
I V Fig. 136. Course of the oxidation of

u
u
11/ tu 8J W
liquid lead with a smooth surface, at
500°C in oxygen at 1 atm, according to
Weber and Baldwin (three parabolic
Hours_ rate laws).

of the oxidation of liquid lead with time at 500°0 in 1 atm air can be described
by three different parabolic scaling constants. X-ray investigations on these
oxide layers in the region where the first parabolic scaling constant is valid
unfortunately gave no clear picture of the presence of Pb304 besides PbO.
In the region where the second parabolic scaling constant applies (below
540°0), yellow (above 486°0) or red (below 486°0) PbO could be plainly
determined in addition to small quantities of Pb304, which could be definitely
excluded from the region of the third parabolic scaling constant. This

1 Gruhl, W.: Z. MetaUk. 40, 225 (1949).


2 Weber, E., and W.l\f. Baldwin, Jr.: J. Metals 4,3 (1952).
4.5. Scaling Systems with Protective Layers Containing Several Phases 273

finding is significant because it reinforces the assumption, in agreement


with Baldwin, that the initially slow oxidation rate is probably due to a
predominant Pb304 formation at the beginning of the oxidation, while
Gruhl assumes a predominant PbO formation, followed by a Pb304 forma-
tion, which is then held responsible for the decrease in the oxidation rate.
When cobalt is oxidized between 200 and 1000°C in air or oxygen at
1 atm, then besides CoO one also observes a more or less large amount of
C0304 in the scaling layer. However, no C0 20 3 layer has ever been found on
the outer edge of the scale. With decreasing oxidation temperature the
proportion of C0304 becomes greater. By means of electron diffraction
experiments, Gulbransen and Hickmanl showed that only a CoO layer
existed on cobalt sheets oxidized for a short time between 300 and 400°C
at 1 mm Hg oxygen and that this CoO was covered by a perceptible C0304
layer only after prolonged oxidation. In all cases the CoO part predominated
in the scaling layer. 2 The rate of the further oxidation of CoO to C0304,
which may be described by a parabolic rate law, was determined by
Chauvenet. 3 It should proceed about 10 times more rapidly than the oxida-
tion of cobalt under the same conditions, but other experimental results
described earlier (Section 4.2.1) have shown that this is not the case.
Since a flaking of the CoO layer from the metal was frequently observed
on cooling, Arkharov4 investigated the oxide growth conditions at various
temperatures and mechanical pretreatments of the cobalt sheet. Edwards
and Lipson 5 demonstrated on basis of X-ray experiments that the cubic
structure of CoO is stable above 500°C and that significant quantities of
hexagonal CoO were not evident until 300°C was attained, which on further
cooling to room temperature amounted to 50%. These two structures must
frequently be responsible for the varying adherence of the CoO coatings
on the metal below. On basis of the measured values for the cobalt oxidation
by Johns and Baldwin6 and especially the most recent value for the cobalt
oxidation to CoO by Carter and Richardson,7 one can calculate the thickness
of the CoO and C030 4 layer from (4.84) and (4.85).
One of the most interesting examples of scaling systems with several
oxide layers is oxidized iron, where not only was the existence of all three
oxide layers-FeO, Fe304, and Fe203-shown, but their rates of formation

1 Gulbransen, E. A., and J. W. Hickman: Trans. AIME 171,306 (1947).


2 Arkharov, V. 1., and Z. A. Voroshilova: Zhur. Tech. Fiz. 6, 781 (1936).
3 Chauvenet, G.: Diss. Univ. Caen (1942), No. 34.
4 Arkharov, V. 1., and G. D. Lomakin: Zhur. Tekh. Fiz. 14, 155 (1944).
5 Edwards, O. S., and H. Lipson: J. Inst. Metals 69,177 (1943).
6 Johns, C. R., and W. M. Baldwin, Jr.: Metals Trans. 185, 720 (1949).
7 Carter, R. E., and F. D. Richardson: J. Metals 7,336 (1955), with a theoretical appen-
dix by C. Wagner.
274 4. Scalin~ Processes with Formation of Thick Protective Layers

were quantitatively measured as well. The oxidation mechanism derived


for this was confirmed through self-diffusion measurements in the individual
oxides. Because of the outstanding technological importance of iron and
its alloys this considerable expenditure of effort is understandable. Older
review works on the oxidation rate of iron were written by Fischbeck and
Salzer,l Pfeil and Winterbottom,2 Hudson and Rooney,3 and others. Even
though these works were concerned at first only with the determination of
the rate of the total process of the oxidation and the microscopic investiga-
tionof the scaling layer, a qualitative description of the extent of the
sublayers and the conditions of their formation in a heterogeneous-
Fe/FeO/Fe304/Fe203/02-scaling layer could be given even at that time,
although no measurements existed either on the FeO formation on Fe or
the Fe304 formation on FeO, etc., or on the diffusion rates of ions through
the individual oxide phases. Fischbeck discussed the structure of the scaling
layer in terms of the oxygen solubility in iron, as is given, for example, in a
work by Paidassi4 and by Rahmel in Fig. 137. Again and again subsequent
investigations have confirmed this structure of the scaling layer. 5

Fig. 137. Results of the oxidation of a pure iron sample oxidized in


air at 600°C for 16 hr, according to Rahmel. (Enlarged 500 times.)
Analysis of the iron in wt. %: 0.003 C, 0.002 Si, 0.002 Mn,0.007 P,
0.005 S, 0.004 N, 0.007 AI, 0,075 O.

1 Fischbeck, K., L. Neundeubel, and F. Salzer: Z. Elektrochem. 40, 517 (1934); K.


Fischbeck and F. Salzer: MetaUwirtschaft 14, 733 (1935).
2 Pfeil, L. B., and A. B. Winterbottom in: Review of Oxidation and Scaling of Heated
Solid .Metals, H.M. Stationery Office, London, 1935.
3 Hudson, J. C., and E. E. Rooney in: Review of Oxidation and Scaling of Heated Solid
l'.:Ietals, H.M. Stationery Office, London, 1935.
4 Paidassi, J.: J. Metals 4, 536 (1952).
5 Paidassi, J.: Bol. soc. chilena quim. 3, 55 (1951); 5, 46 (1953); 6 (1954).
4.5. Scalin~ Systems with Protective Layers Containin~ Several Phases 275

Using the approach to the calculation of layer thicknesses given above,


Jost,l and Hauffe and Schottky 2 generalized the formulation to more than
two layers, so that a calculation of the layer thicknesses of the individual
oxides is possible, and this will be discussed later. These relationships, how-
ever, are only applicable when all phase-boundary reactions-including
nucleus formation and crystal growth-proceed sufficiently rapidly and the
rate of formation of the individual oxide layers may be described by a
common rate law, e.g., by the parabolic rate law. These assumptions are also
to be considered here, since a calculation of the individual oxide layer
thicknesses is not possible if they are not fulfilled.
On basis of detailed experimental results on the oxidation of iron 3 and
self-diffusion measurements of iron in individual iron oxides by Himmel,
Mehl, and Birchena1l4 as well as structure investigations into the oxide
layers growing on different compounds of iron, especially those by Benard
and co-worker,5 we are today in a position to quantitatively describe the
oxidation mechanism and the structure of the scaling layer. Since these
results serve as a basis for the explanation of the scaling mechanism of the
technologically interesting iron alloys and steels, they shall be discussed in
considerable detail. 6
The phase diagram of the iron-oxygen system 7 (Fig. 138) shows that
we have to distinguish two temperature regions in the oxidation. In agree-
ment with the phase diagram, below 570°0 by far the greatest part of the
scaling layer consists of Fea04, which is covered by a thin layer of y- and
IX-Fe 2 03. The oxidation rate is due only to the growth rate of the Fe304
layer. 8 Above 570°0 the greatest part of the scaling layer consists of FeO,
with only a thin outer layer consisting of Fe304 and Fe20a, as can be seen
from Fig. 139. Of these two exterior layers the Fe 2 03 layer is the thinner.
Birchenall and co-workers investigated the temperature dependence of the
composition of the scaling layer. As can be seen from the experimental

1 Jost, W.: DiJJusion und chem. Reaktion in festen StoJJen, Dresden, 1937, pp. 162 and
167.
2 Hauffe, K., and W. Schottky: Deckchichtbildung auf Metallen in Halbleiterprobleme 5,
Braunschweig, 1960, p. 203.
3 Davies, M. H., M. T. Simnad and C. E. Birchenall: J. Metals 3,889 (1951); 5, 1250
(1953); K. Hauffe and H. Pfeiffer: Z. Elektrochem. 56,390 (1952); Z. Metallk. 44,27
(1953).
4 Himmel, L., R. F. Mehl and C. E. Birchenall: J. Metal8 5,827 (1953).
5 Bardolle, J., and J. Benard: Rev. met. 49,613 (1952); Compt. rend. acado sci. 239,706
(1954).
6 Hauffe, K.: Metalloberflache (A) 8, 97 (1954).
7 Darken, L. S., and R. W. Gurry: J. Am. Chem. Soc. 68, 798 (1946).
S Benard, J., and O. Coquelle: Rev. met. 43, 113 (1946); O. A. Tesche: Tran8. AIME
142, 641 (1950).
276 4. Scaling Processes with Formation of Thick Protective Layers

Oxygen at.% -
f2()()
,
50 51. 58
Ite,O. FeA
°C Fe?
y - fe ' ,, feD Hematite
,, Wustite Magnetite" Hematite '" Oxygen
1(}(}(. Wust ite
,, ,
,,
I
,, Wustite
Magnetite
Magnet ite &

flO() \, ,
!
I
Hematite

{X- Fe- ,, ,:
Wustite ,,
, !
GOO \
{X- Fe' Magnetite

t;l){J
24 26 28 J(}

Oxygen wt.%- -

Fig. 138. Phase diagram of the iron-oxygen system, accord-


ing to Darken and Gurry. The boundary lines of the FeO
phase indicated by dashes are taken from measurements by
Benard.

results shown in Fig. 139, above 750°0, significant contents of Fea04 and
Fe20a can no longer be detected. The parabolic rate law found between 650
and 975°0 thus described the growth rate of the FeO layer under the special
conditions of a constant oxygen partial pressure [po.(FeOjFea04}] owing to
the presence of a thin Fea04 layer.

100
"
/
% feD
eo
I
20
\ Fig. 139. Temperature dependence of the

o
\ Fe,O.
--IlOO----WO [eA
composition of the scaling layer of iron sheets
oxidized in pure oxygen at 1 atm, according to
600 700 8()() C1f,'00 Davies, Simnad, and Birchenall.
Q

Hand in hand with kinetic investigations which aim toward an explana-


tion of the rate law of oxidation and the determination of the absolute
reaction rate, there were studies concerning the influence on the oxidation
rate of both the orientation of the iron crystal on the metal surface and the
4.5. Scaling Systems with Protective Layers Containing Several Phases 277

structure of the oxide film. Benard1 and Gulbransen and co-workers 2 were
particularly concerned with the latter problem. Microscopic investigations
show that on an iron crystal oxidized at 850°C under an oxygen pressure of
10-3 to 10- 2 mm Hg, the number of oxide nuclei forming per unit surface and
time is dependent on the crystal orientation, as can be clearly seen from
Fig. 140. Figure 141 presents a stereographic diagram according to Bardolle


• •
•,

.
~

• ,'.
Fig. 140. Photomicrograph (enlarged
1200 times) of an iron surface oxidized
"
at 580 0 in oxygen of 1O- L I0- 3 mm Hg
for a short time, according to Bardolle
and Benard. (The number of oxide
nuclei varies from one crystal face to
another.)

Fig. 141. Storeographic diagram of the


number of oxide nuclei as a function
of the crystallographic planes (double
circle), according to Bardolle and
Benard. (The iron was oxidized as in
Fig. 140.) The numbers in the indivi-
dual circles give the number of the
oxide nuclei per 10-4 mm 2 crystal
surface.

1 Bardolle, J., and J. Benard: Rev. met. 49, 613 (1952); J. Moreau and J. Benard:
Compt. rend. 248, 1658 (1959).
2 Gulbransen, E. A., and R. Ruka: J. Electrochem. Soc. 99, 360 (1952); E. A. Gulbransen,
W. R. McMillan, and K. F. Andrew: J. Metals 6,3 (1954).
278 4. Scaling Processes with Formation of Thick Protective Layers

and Benard,l where the corners of the triangle are determined by the
crystallographic planes (lOO), (Ill), and (IlO). The average number of
nuclei per 0.0001 mm 2 which were obtained under the experimental con-
ditions indicated are recorded on the face of the triangle. Since the orienta-
tion of the IX-FeO on the (100) plane of the iron crystal is most favorable,
it is not surprising that most of the FeO nuclei were formed on the (100)
planes (see the left corner of Fig. 141).

Appearance of the
Growth of the Second Loyer
Oxjde Crystol
Chemjsorptjon Loyer

A B o D

Fig. 142. Schematic representation of the growth of the oxide layer on iron as it is
obtained according to electron microscopic investigations by Gulbransen: (A) chemi-
sorption of oxygen; (B) layer of about 100 A with fine mosaic structure; (C) growth of
the oxide crystals to 500 A; (D) appearance of the second layer.

According to Gulbransen the course of the oxidation can be divided


into four reaction sections, as is shown schematically in Fig. 142. Diagram A
characterizes the initial rate of an oxide-free iron surface. The oxidation
which sets in relatively rapidly at 850°C produces a compact FeO layer
lOO A thick (reaction section B), upon which the FeO crystallites present
in the photomicrograph (Fig. 140) grow in a definite crystallographic arrange-
ment in subsequent stages of the oxidation. Reaction section D shows the
state of thicker oxide layers, where the cavities arising in reaction section C
are filled up by the further growth of FeO crystallites. In an extension of these
experiments the following decomposition reactions of the higher oxides to
FeO were studied by Gulbransen:
and
and
Fe304 + Fe ~ 4FeO
Also, the FeO formed in the reduction shows oriented layers. Sato's2 experi-
mental results on oriented overgrowth of oxides forming on IX-iron at lower
temperatures (250°C) in air arc similar. Under these conditions a preferred

1 Bardolle, J., and J. Benard: Compt. rend. 239,706 (1954).


2 Sato, R.: J. Phys. Soc. Japan 8, 758 (1953).
4.5. Scaling Systems with Protective Layers Containing Several Phases 279

y-Fe20a formation on IX-iron was determined by electron diffraction proce-


dures, in agreement with earlier investigations by Iimori.1 However, because
of the structural similarity of the y-Fe20a and Fea04 and the mutual
solubility,2 no comprehensive statements can be made at this time. Earlier
results by Mehl and McCandless a have been confirmed by current investiga-
tions on the epitaxy ofthe oxide layers on iron.4 As we will see in the follow-
ing high-temperature mechanism of iron oxidation is basically different
from the low-temperature mechanism, which is to be expected from the
structure investigations.
The parabolic course of the oxidation of Fe to FeO, FeO to Fea04, and
Fea04 to Fe20a between 850 and 1100°C is presented in Figs. 143 and 144

10 1-----t------t.-,/,:::"""'"----4-~
g/cm Z
8 1----f----rY'-'-------t---P".~--+____j

J Qt----~~~~~~
~

Fig. 143. Parabolic course of the oxida- o 10 EO


tion of FeO to Fea04, according to
Davies, Simnad, and Birchenall. 5 W [min] -

and in Table 35. In contrast to the oxidation of the Cu 20, the increase in
layer thickness of the higher iron oxides also obeys a parabolic rate law.

1-.2
g-cm ·10

GtO°C L
t8
----
..--r

~6 ~
~ " / 850°C
Fig. 144. Course of the oxidation of
Fea04 to F e20a in oxygen at 1 atm,
according to Davies, Simnad, and
2
P
Birchenall. 5 o 1000 2(}()O 3()()O ¥(J(J() mrn

lIimori, T.: Nature (London) 140,278 (1937); Sci. Papers Inst. Phys. Ohem. Research
(Tokyo) 34, 60 (1937) .
2 Wyckoff, R . W. G. : Ory8tal Structures, Vols. I and II, Interscience, New York, 1951.
a M ehl, R. F ., and E. L. McCandless: Trans. AIME 125, 531 (1937).
4 Haase, 0.: Z. NaturJorsch. lla, 46 (1956).
5 Davies, M. H ., M. T . Simnad, and C. E. Birchenall: J. Metals 3,889 (1951).
280 4. Scaling Processes with Formation of Thick Protective Layers

The parabolic course of the oxidation of Fe to FeO in an H 2-H 20 atmosphere


above 950 0 is noteworthy. At these temperatures Hauffe and Pfeiffer1
found a linear rate law in a CO-C0 2 atmosphere. Evidently the correspond.
ing processes with H 20 proceed rapidly enough compared with the chemi.
sorption of C02 and its dissociation so that the transport of the iron ions
through the FeO layer is the slow process and thus determines the rate law.

Table 35. Comparison of the Experimentally Obtained Scaling Constants with


Those Calculated According to Wagner's Scaling Formula by Use of the Experi-
mentally Obtained Self-Diffusion Coefficients of Iron Ions in FeO, Fe304, and
Fe203, According to Himmel, Mehl, and Birchenall

k" ,
k, g2/cm4-sec
Reactions T,oC equiv/cm-sec calc obs

l02(g) + Fe = FeO 983 2.8 X 10- 8 5.9 X 10- 7 6.7 X 10- 7


897 1.1 X 10- 8 2.3 X 10- 7 2.5 X 10-7
800 2.5 X 10- 9 5.3 X 10- 8 5.3 X 10- 8
3FeO + l02(g) = Fe304 1100 9.2 X 10- 9 2.9 X 10- 8 3.2 X 10- 8
1050 4.1 x 10- 9 1.2 X 10- 8 1.7 X 10- 8
1000 1.4 X 10- 9 4.5 X 10- 9 8.1 X 10- 9
2Fe304 + l02(g) = 3Fe203 1100 1.7 x 10-12 4.8 X 10- 12 1.0 X 10-8
1000 2.1 X 10-14 5.8 X 10-14 2.3 X 10- 9

As can be deduced from the self-diffusion coefficients in FeO, Fea04,


and Fe203 derived from self-diffusion measurements with ions of radioactive
iron Fe-55 and from the scaling constants calculated from them (see Table 35),
a predominant diffusion of iron ions takes place through the FeO and the
Fe304 layers, while in the Fe203 layer, because of the large difference in the
values so obtained and those calculated from diffusion measurements, one
has to assume a preferential oxygen diffusion. The measured values of the
self-diffusion coefficients of iron ions in FeO can be inferred from Fig. 145.

100
CIne·seC·'·10 9
/0
I
.983°C
80

80 ? 0
Fig. 145. Dependence of the self·diffusion
coefficients of Fe in wustite on the com-
,,~

<:::i w / 891"CJ position or on concentration of lattice


----- ~ ~ defects in the FeO phase, according to

.- 3 5
"
o8(J()°C
8
Himmel, Mehl, and Birchenall. The short
vertical dashes give the boundaries of
stability of the FeO phase at the corres-
ponding temperatures, according to
Number of feo "/cm 3. 10-31 Darken and Gurry.

1 Hauffe, K., and H. Pfeiffer: Z. MetaUk. 44, 27 (1953).


4.5. Scalin~ Systems with Protective Layers Containin~ Several Phases 281

Above 900°C the self-diffusion coefficients depend on the iron ion vacancy
concentration and thus on the oxygen pressure in the surrounding gas
atmosphere during the diffusion. Corresponding to the finding of Hauffe
and Pfeiffer, which is reproduced in Fig. 146,

xFeD" = const· p~:

we obtain the following relationship for the self-diffusion coefficients (D;e)Feo


of the iron ions in FeO at 1000°C when we set const == XFeo"[po.(FeOj
Fe304)] and 1) = kTje
* )FeO -- XFeO"[po.(FeO/Fe.o.»UFeO" 1) Po.1/6
(D Fe

Here xFeo"[po.(FeOjFe304}] is the concentration of the iron-ion vacancies at


the oxygen equilibrium pressure in the reaction 3FeO + ! O2 ~ Fe304.
With UFeO ~ 1.3 X 10-4 cm 2jsec-volt, the self-diffusion coefficient at
N

1000°C at Po. ~ 10-13 atm was obtained as about 2 x 10-7 cm 2 jsec, in


satisfactory agreement with the measured values of Himmel, Mehl, and
Birchenall.
-aD
/'
T=.Ys.
~ T- woy ~
V /
,Y

/ L
V /V
,/

-a.9 1/ ... t

15 13

Fig. 146. Dependence of the iron ion vacancy concentration


XFeO"( = lX(j)in FeO on the oxygen partial pressure at 950
and 1000°C, according to Hauffe and Pfeiffer. The slope of
the straight line yields n = !, so that it follows that:
XFeO" = const po. 1/6.

The self-diffusion coefficients of the iron ions in Fe304 were determined


in an argon atmosphere containing water vapor between 799 and 987°C
and that of the iron ions in Fe20a in pure oxygen between 1000°C and
1217°C. While the experimental condition is correct for Fe304, it appears
to be questionable in Fe203. As was noted hi Section 3.3, the self-diffusion
282 4. Scaling Processes with Formation of Thick Protective Layers

coefficient of the iron ion in Fe20a determined at a certain oxygen pressure


would only be useful for the calculation of scaling constants if Fe20a were
a p-type conductor, which is improbable according to the earlier investiga-
tions by Wagner l and Morin. 2 At intermediate and lower temperatures
Fe 20a is an n-type conductor; however, at higher temperatures it exhibits
an increasing amount of intrinsic semiconduction and finally in strongly
oxidizing atmospheres can show a predominant p-type conduction with
iron ion vacancies. Since this inversion was not definitely determined, it
would be useful to repeat the self-diffusion measurements with iron ions in
an Fe 20a which is in equilibrium with Fea04 in an atmosphere where the
oxygen partial pressure approximately corresponds to the Fe 20a decomposi-
tion pressure. The self-diffusion coefficient will, of course, take on a greater
value although it is doubtful that the difference in the measured and cal-
culated scaling constant found in Table 35 will be removed. Even though a
definite conclusion will be possible only upon completion of the above-men-
tioned experiments, we, nevertheless, consider a preferential diffusion of
oxygen ions through the Fe20a layer likely, in agreement with Himmel,
Mehl, and Birchenall.
The self-diffusion coefficients found in the Fe 20a single crystals are about
one-half to one order of magnitude lower than the values found by Lindnera
in polycrystalline Fe 2 0a tablets (Fig. 147), which indicates a predominant
grain boundary diffusion in the latter case. The activation energies obtained
from the experimental results in the two figures
(D;e)Fe,o. = 5.2 exp( -55,000/RT)
and
(D;e)Fe 2 o, = 4 x 105 exp( -1l2,000/RT) cm2 /sec

are greater in the first case than the activation energy of 45,000 cal/mole
obtained from oxidation experiments of Fe to Fea04 below 570°C, and in
the second case greater than the activation energy of 53,000 cal/mole obtained
in the further oxidation of Fea04 to Fe20a. By use of the Wagner scaling
formula

where Zl and Z2 are the valences of the iron and oxygen ions, C2 the concentra-
tion of the oxygen in g-atoms/cm 3 and ((~il and a~a) are the thermodynamic
1 Wagner, C., and E. Koch: Z. physik. Chem. (B) 32, 439 (1936).
21\1orin, F. J.: Phys. Rev. 83,1005 (1951); 93,1195 (1954).
3 Lindner, R.: Arkiv. Kemi. 4, 381 (1952).
4.5. Scaling Systems with Protective Layers Containing Several Phases 283

taJ() !JOO ;0-10 1800


~ y~ ~
"\
;I I 1\
°\0
0\
c~

\\
-~
10 8.2 7.0 7.8 8.8
a b
Fig. 147. Temperature dependence for the self-diffusion of Fe ions in
magnetite (Fe304) and hematite (Fe203) between 890 and 1200°C,
according to Himmel, Mehl, and Birchenall. (a) Measurements on
magnetite with an average composition of Fe2.99304 in argon atmospheres
containing water vapor; (b) measurements on Fe203 single crystals . ,
and on sintered Fe203 tablets (0, according to Lindner) in oxygen.

activities of the oxygen at the iron/oxide or oxide/oxide and oxide/oxygen


phase boundaries, respectively, and for D:
~ 0, we obtain the following
relationship, which can be evaluated:

(4.86)

Further, since we can set In ao = 2.3 log(PH 2 o/PH,), it follows for the
oxidation constants for iron and the individual oxides:

a(a)

f D;e ~ d
o
k FeO = 0.383 (log PRIO) (Fe-+ FeO) (4.87a)
Xpo PRO
a(i)
o
a(a) )
kFe•O• = 0.551 D"Fe log ( a~1) (4.87b)

(4.87c)

In Table 35 the scaling constants calculated by means of (4.87a-4.87c)


from the self-diffusion measurements on iron are compared with those
obtained from oxidation experiments. While use of (4.87a) and
(4.87b) is suitable for the calculation of scaling constants, the use of
284 4. Scaling Processes with Formation of Thick Protective Layers

(4.87c) leads to results which are about three or four orders of magnitude
too low relative to those obtained from oxidation experiments, and when
viewed in light of the above discussion, this fact is not at all surprising.
Although a preferred oxygen diffusion through the Fe20s layer appears to be
a reasonable assumption, it would still be advisable to conduct further
experiments-in addition to the marker experiments of Davies, Simnad, and
Birchenall, according to which ~ D: D:-in
order to ascertain without
doubt that oxygen diffusion is indeed preferred. For (4.87c) the following
relationship may then be used:

(4.87d)

So·far no self-diffusion measurements on oxygen in Fe 20s single crystals


have been carried out. However, equation (4.87d) can also be used to cal-
culate the self-diffusion coefficients of oxygen at 1000 or llOO°C from the
scaling constants which have been obtained.
In addition to demonstrating the applicability of Wagner's theory
to scaling systems with several oxide layers, the work of Benard, Birchenall,
Gulbransen, Hauffe, and Mehl serves to a large extent to explain the mechan-
ism of iron oxidation at high temperatures. Finally a schematic representa-
tion of the diffusion and phase-boundary processes which can participate
in iron oxidation at oxygen pressures in the stable region of the Fe20s
phase is given in Fig. 148. 1
Independent of the mass transport through the Fe20s layer, the iron
oxidation shows that the calculations of the layer thickness ratio, derived
earlier for the system A(gas)jABjAB2jB2(gas) must be completed in two
directions. For one thing we have to take into consideration three layers
and for another the assumption that 8B = 0 in all partial layers must be
eliminated because of the fact that in the Fe 20s layer the oxygen ions seem
to diffuse preferentially. In spite of this, a general calculation of the g ratios
from the DA and DB values is also possible. Here, we have to express the
changes with time of the values of 6, g2, 6 of the individual layers by the
currents 8Fe and 80 (8A and 8B in general notation) through these individual
layers. According to equation (4.84) we use dnM(3)jdt determined by the
sum (dnM(3)jdthu + (dnM(3)/dthv. (The numbers I, II, III, and IV denote
the phase boundaries in Fig. 148 and the numbers 1, 2, 3 the individual
phases FeO, Fe304, Fe203.) While (dnM(1)/dth and (dnM(3)/dthv are un-
equivocally determined by the (8A(1)h and (8B(3)hv values, we have to consider
for the phase boundaries II and III that here the difference of 8A or 8B in
every two neighboring phases will be determinative for the corresponding
dnM/dt, as for example (dnM(2) /dthu and (dnM(3)/dt)UI' In analogy to equations

1 Hauffe, K.: Metalloberflache (A) 8, 97 (1954).


4.5. Scaling Systems with Protective Layers Containing Several Phases 285

(4.84) these 8 differences supply the numbers of the reacting particles which
then are available for the displacement reactions at the corresponding
phase boundary; at the III boundary, for instance, the reactions (4.86a)
occur. The relationships between the (dnM(3)/dt)UI and (dnM(2)/dthn values
and the 8Fe and 80 differences can be derived only from these displacement
reactions.
n m If

Iron wustitj;
Magnetite
. Hematite Oxygen
Fe ~ ~~ ~~ o.
2
1 - - - - Fe 2+ Fe +
Fe'+ - I-- O·-=-----
-:-: 2 )

"
Over Tetrahedral ..
Over FeO~ctahedral OverDo D,'ffius,'on
1/ Positions
1 - - - - - e- e- - - e--
Over al
____L -__________- L_ _ _ _
feralande Over e
~L_ __ ~ __ ~

Fig. 148. Schematic representation of the diffusion


processes and phase boundary reactions during the oxida-
tion of iron in oxygen, according to Hauffe.
Phase Boundary Reactions
I: Fe (metal phase) + FeD" + 2EElhn FeO '=7 Null
II: Fe304'=7 (4FeO + FeD" + 2EJ:));n FeO(Fe304 decom-
position). Some ofthe Fe 2+ ions arriving through the FeO
phase and the Fe 3+ ions ("" EEl) arising at the
FeO/Fe304 phase boundary pass over into the
magnetic phase:
EEl in FeO + (Fe 0" + FeD"') vacant tetrahedral positions in Fe,O.
'=7 Filled lattice positions + (2FeO") in FeO
III: 12Fe203 '=7 9Fe304 + (FeD" + 2FeO'" 8$) +
(Fe203 decomposition)
2Fe304 '=7 3Fe203 + (00" + 28hn Fe.O,
(Fe203 formation)
IV: t02(g) + (OON + 28hn FezO, '=7 Null

In the same manner we can derive the relations for the II boundary.
For carrying out the calculations we have to consider only the effective
diffusion coefficients D A and DB of the individual phases without information
about the lattice defect migration and the electronic currents. In this way,
a desired universality of the derivations results which represents an important
case, especially for the calculation of the Fe304 formation where both Fe 0"
and Fe 0'" vacancies must be considered.
The conditions are considerably different if iron is oxidized below 570°C.
Here the overall oxidation rate is given by the growth rate of Fe304, while
an FeO layer is formed only as a very thin film at the iron/tarnishing layer
phase boundary and is difficult to observe directly. In connection with
investigations by Chaudron and Benard,l Gulbransen and Ruka 2 studied the
1 Chaudron, G., and J. Benard: Bull. soc. chim. France (1949), 88.
2 Gulbransen, E. A., and R. Ruka: J. Metals 2, 1500 (1950).
286 4. Scaling Processes with Formation of Thick Protective Layers

structure of the coating after oxidation experiments in the neighborhood


of the equilibrium point (570°0) in the reaction
(4.88)
On basis of electron diffraction procedures it was determined that
equilibrium from both sides is attained very slowly and a good measurable
rate is in evidence only in the presence of thin films. With short oxidation
times with thin oxide films, an FeO formation could be observed below
570°0 down to 400°0. This temperature reduction for the FeO-Fea04
conversion is characteristic only for thin tarnishing layers and is largely
cancelled by an ensuing vacuum oxidation (Po. ~ 10-2 to lO-a mm Hg).
The larger lattice parameter of the growing oxide layer appearing in the
first oxidation state indicates that the oxide layer is formed under mechanical
stresses. The conclusion of the authors that the temperature reduction of the
conversion point is due to the stress state of the Fea04 layer growing on
iron appears justified if one considers that FeO layers do not cause that type
of phenomenon.
Kinetic investigations on the initial rate of reaction (4.88) between
570 and 630°0 always yield higher rates with thin oxide films than with thick
ones. The kinetics of the change were studied as a function of time, tempera-
ture, and composition of the oxide. The nucleus formation or the crystal
growth of Fe and Fea04 crystals in the scaling layer are regarded as rate-
determining steps in the course of the reaction from right to left in reaction
(4.88). The expressions for the reaction rate are based on the presentations
of Volmer and Weber1 and Becker.2
Evans a and Vernon and co-workers 4 have explained the Imy-temperature
oxidation mechanism, and in this connection, a work by Oaule and Oohen5
is noteworthy. Vernon used low-alloy steels (0: 0.09; Si: 0.019; S: 0.016;
P: 0.024; Mn: 0.33; Ni: 0.048; Or: 0.038; and Ou: 0.10wt.%), which he
oxidized between 180 and 225°0 after various surface treatments. While
a parabolic rate law dominated the oxidation above 200°0, the oxidation rate
followed a logarithmic rate law below 200°0 of the form (see Fig. 149a)
L1m/q = log (O.lt + 1)
In the same way-although with a displaced transition (parabolic ~
logarithmic) region at higher temperatures (about 325°0)-Evans also
1 Volmer, M., and A. Weber: Z. physik. Ohern. 119, 277 (1926).
2 Becker, R.: Ann. Physik (5) 32, 128 (1938).
3 Davies, D. E., U. R. Evans, and J. N. Agar: Proc. Roy. Soc. (A) 225, 443 (1954).
4 Vernon, W. H. J., E. A. Calnan, C. J. B. Clews, and T. J. Nurse: Proc. Roy. Soc. (A)
216, 375 (1953) .
.5 Caule, E. J., and M. Cohen: Oan. J. Ohern. 3, 298 (1955).
4.5. Scalin~ Systems with Protective Layers Containin~ Several Phases 287

found a logarithmic rate law below a certain temperature (here below 325°C),
as can be seen in Fig. 149b. The Swedish iron used by Evans (C: 0.034;
Si: 0.010; Mn: 0.145; Ni: 0.030; Cr: 0.007; and Cu: 0.010 wt.%) was some-
what purer, which perhaps could have caused the displacement of the
transition region to higher temperatures. To what extent the oxidation
course at 300°C in Fig. 149b may be better represented by cubic rate law-
as a "transition rate law" to the logarithmic-must be decided through

c~z V
~
!7u
2'
/
/'
/
/'
Fig. 149a. Logarithmic course of the oxida-
tion of iron in oxygen at 180°C, according to
Vernon and co-workers. 0 equals mechanically
o
V o,s 1.0 1.5 3.0
worked surface, L chemically etched surface,
D surface annealed in vacuum. log(O,1·t;-1) (t in h) -

t 3OI---+--+--+------r---'':-7'f'--+---i
~
~201--+--t--7r_-+_::;;;>"'~-!---j
Q
~
Fig. 149b. Logarithmic course of the oxida- fOf--;J.D""",,!=---j---f'-'---"'':'';-..ECo.r--l
tion of pure iron in oxygen at 250 and 300°C,
according to Davies, Evans, and Agar. (W
represents the oxygen content of the 0(-Fe20a
in /Lg/cm 2 .)

further experiments. It was clearly established that below a certain tempera-


ture a logarithmic rate law was found which could not be explained for
oxide-layer thicknesses greater than or equal to 1000 A by the theory of a
rate-determining electron supply discussed by Mott and noted in Section
3.5.4. This fact and the experimental demonstration of growing oxide nuclei
on different faces of the iron crystal in different numbers and sizes, according
to Bardolle and Benard! lends full credence to the theory first developed by
Evans 2 of the logarithmic growth law and its extensions,3 discussed in
Section 3.5.5.
1 Bardolle, J., and J. Benard: Rev. met. 49,613 (1952).
2 Evans, U. R.: Nature (London) 164, 909 (1949).
3 Evans, U. R.: Rev. Pure Appl. Chem. 5, 1 (1955).
288 4. Scaling Processes with Formation of Thick Protective Layers

Through the discriminating use of different measuring methods-


gravimetric, electrochemical reduction,l microchemical analysis, and electron
diffraction-Evans and Vernon have furnished a valuable contribution to
the explanation of the oxidation mechanism at lower temperatures. Inci.
dentally, it should be noted that the outer zone of the tarnishing layer
always consists of Ot-Fe203 which has been separated from the metal by a
zone of y-Fe203, and which with increasing time is converted into Fe304.
According to Vernon, the Fe304layer forms relatively rapidly above 225°C;
at 180°C more than 450 hr are necessary to guarantee its formation. Whether
the parabolic rate law at relatively low temperatures is caused solely by a
chemical potential gradient of oxygen or iron (Wagner's theory) or by a
predominant field transport of iron or oxygen ions due to an electric field
created by the chemisorption of oxygen cannot be determined at this time.
Here the oxygen-pressure dependence of the oxidation rate between 180
and 250°C can give certain information (rate independent of oxygen pressure
indicates diffusion; and logarithmic dependence indicates field transport).

4.5.1. Joint Parabolic and Linear Growth of


Two Adjacent Oxide Layers
As we have shown in the case of copper and iron oxidation in the stable
range of the highest oxidation state, on both cases, even after long oxidation
times, a parabolic oxidation was found which was caused by the diffusion
of the metal ions through the "lowest" oxide layer (CU20 or FeO). A simple
quantitative calculation of the growth of the individual layers is possible
only when the growth of all layers is determined by the same rate law,
i.e., by related mechanisms.
Conditions become more complicated when after a rather long oxidation
time the parabolic rate law changes into a linear one, as was described, for
example, for titanium, cerium, and thorium oxidations (see Section 4.2.2).
The change of the course of oxidation with time may now be due to a quite
different cause. If the oxides of lower valence arising below those of higher
valence (as for example in the case of the titanium oxidation the TiO and
Ti 203 under Ti0 2) tend to have porous structures because of the lack of a
good epitaxy, then during the oxidation, the total oxide layer, including
the initially compact segment, becomes porous so that phase-boundary
reactions become rate-determining rather than diffusion and transport
processes. In such cases a preliminary calculation of the course of the oxida-
tion with time is not possible.
Conditions are considerably simpler when the compact layer grows
directly on the metal without formation of a porous layer until later in the
1 Evans, U. R., and J. Stockdale: J. Ohern. Soc. (1929),2651.
4.5. Scalin~ Systems with Protective Layers Containin~ Several Phases 289

course of the oxidation. In this case, the maximum thickness of the compact
oxide layer may be calculated mathematically and the point in time at which
the parabolic rate law changes into the linear one may be determined.
These types of calculations were carried out by Webb, Norton, and Wagner!
in the evaluation of their experimental results in tungsten oxidation. Since
the relationships obtained by these authors are applicable to other oxida-
tion systems with similar mechanisms for formation of protective layers,
we will, for illustrative purposes, discuss the quantitative relations for
tungsten oxidation.
Tungsten sheets (0.05 cm thick and with a surface of 4 to 10 cm 2) of99.9%
purity (0.002% Fe; 0.005% Si; 0.01 % AI; 0.033% C; and 0.004% wt. % S)
were oxidized in oxygen at 1 atm between 700 and lOoo°C. In the first
oxidation period the oxidation proceeds according to a parabolic rate law
with the formation of an adherent pore-free protective layer of a dark-blue
oxide with the composition of W0 2. 7S • However, quite soon a loose yellow
oxide layer of W03 is noted on the outside. The dark-blue oxide layer grows
according to the parabolic rate law and at 700°C attains a thickness of about
2 x 10-3 cm. At 900c C the layer is about three times as thick. The outer
yellow oxide layer grows with a constant rate. Thus, we have to consider
the following two rate equations:
d11l1. = ~ _ b (4.88a)
dt ~

dm2 = fb (4.88b)
dt
where ml or m2 is the mass of the oxygen in the adherent inner or porous
outer layer per unit surface, and f is the ratio of the oxygen content per
gram-atom in the porous and in the compact oxide layer; a and b are constants.
The total quantity of oxygen which is taken up per unit surface is then

(4.88c)

After short oxidation times when the lower layer is still very thin and
therefore ml is also very small, the first member on the right-hand side of
(4.88a) predominates. Hence, the constant a can be calculated as follows:

a = lim J~ d (..1 m)2}. = ~ k" (4.88d)


1-+0 l2 dt 2

and is found to be just half of the practical scaling constant kIf in g2jcm4-sec.
The rate law (4.88b), which is determinative for the rate of the con-
version of the compact oxide (W 4011 + W03) into the porous yellow oxide

1 Webb, W. W., J. T. Norton, and C. Wagner: J. Electrochem. Soc. 103, 107 (1956).
290 4. Scaling Processes with Formation of Thick Protective Layers

(WOs) , becomes more important with increasing reaction time, as can be


seen from the experiment by the change from the parabolic into the linear
rate law. Finally the total oxygen consumed with increasing layer thickness
which contributes only to the growth of the outer oxide layer, that is, the
transport of tungsten ions to the inner/outer oxide layer phase boundary
is just as great as the consumption of tungsten in the further growth of the
outer oxide layer. In this way ml reaches a limiting value, ml(max), and then
dml/dt = O. From this it follows that

m 1 (max) = alb (4.88e)


The rate of the mass increase per unit of surface is then dm2/dt, and we
obtain from (4.88b) under consideration of (4.88c) (LIm = m2):

Ib = lim{d(Lfm)} = l" (4.88f)


1-+00 dt
where l" (in g/cm 2-sec) is the rate constant of the linear rate law for long
oxidation times.
Integration of (4.88a) and (4.88b) yields the following expressions:
In(l - bm1/a)-1 - bm1/a = b2 t/a (4.88g)
m 2 = bit (4.88h)
whereby the values of ml and m2 in (4.88c) are determined as functions of
time. By use of the substitutions
(4.88i)
and
Y = b2 tja (4.88j)
we can rewrite (4.88g) in the following form:
X - [1- exp(-X)] = Y (4.88k)
and log Y-calculated from (4.88k)-is plotted against log X. With the
help of this plot we obtain the values of X for a previously given Y = b2t/a.
Through substitution of (4.88g) and (4.88h) into (4.88c), we finally
obtain
L1 m = ~ In (1 - b mIla) -1 + b (f - 1) t
(4.881)
=
(/
7) [X + Y(f-l)]
From this equation the corresponding value of LIm may be calculated
for any experimental time t and compared with the experimental data
(f = 1.09 for W02.75)' This was calculated with! ~ 1, since the composition
4.5. Scaling Systems with Protective Layers Containing Several Phases 291

of the layer is not sufficiently well-known and with a = lk w and b = I".


The good agreement between the values calculated according to (4.881)
and those obtained experimentally can be seen in Fig. 150.
10 -fc----r-,-.."."""'..,.,...-,--,-,--..,.-,.-.-,--,--,-,-,-.-..,
g/cm 2

1fl-31:-'r------&4-~"-----l_",C-.--____:l

1fl'~---r~~-----+-_-_-_-m-oo-su-r-ed~
Fig. 150. Course of the oxida- - - - - calculated
tion of tungsten sheets with
time in oxygen at I atm in a J1fl-'mJ3-~~~~~-~~~~1fl~,---~~gi
log-log plot, according to Webb,
Norton, and Wagner. Seconds_

The maximal thickness, Lffmax, of the compact blue oxide layer may be
calculated from the following relationship:
Lf ~max = m 1 (max)M/(16N o e) (4.88m)
where M is the molecular weight and e the density of the blue oxide. Further-
more, No characterizes the number of oxygen atoms per tungsten atom.
Under the assumption, toa first approximation, that No ~ 3, the following
are obtained for Lffmax and ml(max) :
700 800 900 1000°0
Lffmax 2.0 x 10-3 3.1 X 10-3 5.5 X 10-3 1.35 X 10-2 cm
ml(max) 3_1 x 10-3 4.7 X 10-3 8.4 X 10-3 2.0 X 10-2 g/cm2
to.5 1490 280 264 214 sec
The "half-life" to.5, the time required for ml = -!mlmax, i_e., for half the
maximum obtainable thickness of the protecting blue oxide layer to be
reached, is calculated from (4.88e) and (4.88g) as
to_5 = ;2 (In 2 - 0_5) = 0_19 ~ (4.88n)
It can be seen that above 700°0 only a short oxidation time is required to
reach the half-thickness value of this oxide layer.
From this discussion of tungsten oxidation it becomes clear that for an
understanding of the oxidation mechanism in the terms discussed above,
292 4. Scaling Processes with Formation of Thick Protective Layers

further investigations of the oxidation of cerium, uranium, thorium, and


other metals which show a transition from a parabolic to a linear rate law
are desirable.
Fassell and co-workers 1 investigated the oxidation rate of tungsten
between 600 and 850°C in an oxygen pressure region of 1.4 to 34 atm. Apart
from an initially short parabolic oxidation, which evidently is dependent
upon the pretreatment of the sample prior to oxidation, the course of the
oxidation with time in the temperature and pressure region which was
investigated can be expressed in terms of a linear rate law. Figure 151 shows

30,1--------,--------r--------,-~
g·crriZ·sec-1

20r-------~----_?~--------r__4

J
~

Fig. 151. Logarithmic oxygen pressure


dependence of the linear scaling con-
x stants for 750, 800, and 850°C with
750°C tungsten, according to Baur, Bridges,
°O~------~O'~5------~1.~O--------1.~.,--~ and Fassell. (Circles and crosses
correspond to tungsten sheets of
log POz(P02in Atm)- different origin.)

the oxygen pressure dependence of the linear scaling constants for 750, 800,
and 850°C. Below 750°C the oxidation rate was practically independent
of the oxygen pressure (l~oo = 0.6 x 10-7, l~50 = 2.5 X 10- 7 , and
l;oo = 9 x 10- 7 gjcm 2 -sec). The activation energy of the tungsten oxidation
between 600 and 850°C is 48 kcaljmole. The modified Langmuir equation
used by the authors for the explanation of the oxygen pressure dependence
in the oxidation rate and the use of the "absolute rate theory" of Eyring
still do not permit a description of the reaction mechanism, in spite of the
apparent agreement.
In the same manner Peterson and FasseIl2 investigated the oxidation
rate of molybdenum between 525 and 700°C in the oxygen pressure region
from 1-47.6 atm. The oxidation obeys a linear rate law here in agreement
with the investigations by Simnad noted in Section 2.2.2.4. Between 550

1 Baur, J. P., D. 'V. Bridges, and 'V. 1\1. Fasscll, Jr.: J. Electrochem. Soc. 103, 266
(1956) .
2 Peterson, H. C., and 'V. 1\1. Fassell, Jr.: Tech. Rep. No.6, September, 1954, Army
Ordnance Contract DA·04-495-0HD-237.
4.5. Scaling Systems with Protective Layers Containing Several Phases 293

and 650°0 the oxygen pressure dependence of the oxidation rate may be
expressed as
l" = const. K1Co2/(1 +klC02)
where Kl is the equilibrium constant in the adsorption of oxygen at the
reaction sites and co, the oxygen concentration in the ambient gas.

4.5.2. Passivity and Inhibition During


High- Temperature Oxidation
According to Wagner1,2 a metal or alloy is called passive when the
amount of at least one of the metallic components consumed by a chemical
or electrochemical reaction in a given time is significantly lower at a higher
affinity than at a lower affinity. This phenomenon is completely identical
to the passivity of iron in nitric acid first introduced by Schonbein.
A phenomenon closely related to passivity is inhibition. In contra-
distinction to passivity, we define inhibition as the decrease of the rate of
reaction at a constant affinity, if the decrease is caused by the change of the
concentration of a component which is not involved in the reaction under
consideration.
Such a passivity is indeed realized in the oxidation of molten silicon.
Kaiser and Breslin3 found that the steady. state oxygen content of molten
silicon at 1410°0 in an oxygen-helium stream with Po, < 0.01 atm was
proportional to the oxygen pressure, but constant if Po, > 0.01 atm. This
experimental result is understandable if one considers that at low oxygen
concentrations in the oxidizing atmosphere no Si02 layer formation is observ-
able because all oxidized silicon evaporates as SiO:
Si(l) + t02(gas) = SiO(gas)
At higher oxygen pressures, however, when PSiO reaches PSiO(eq) a protective
Si02 layer may be formed.
As it is shown in Fig. 152 the rate of SiO formation at low oxygen
pressures, jSiO,
DsiOPSiO
jsiO = - - - - (4.89)
oSiO RT

is proportional to Po,. If the critical value po,(crit), which was calculated


by Wagner,l is reached,
po,(crit) = t(DsiOjDO.)1/2psiO(eq) = 0.006 atm
then a rapid drop in the oxidation rate is observed.
1 Wagner, c.: J. Appl. Phys. 29, 1295 (1958).
2 Wagner, C.: J. Electrochem. Soc., in press.
3 Kaiser, W., and J. Breslin: J. Appl. Phys. 29, 1292 (1958).
294 4. Scalin~ Processes with Formation of Thick Protective Layers

If passivity is established and the oxygen partial pressure is gradually


decreased, then Wagner could demonstrate that the Si0 2 layer should be
stable down to Po. = 3 X 10-8 atm in the gas stream. This hysteresis loop
for the rate of SiO volatilization on increasing and decreasing oxygen partial
pressure is shown in Fig. 152.

PSIO = 1.5' 10-' atm

Q
~I""'"
"'co:0
-"'''''
...
o -

.~ "
PSIO = 2· 10- 1• atm
.... I

PO: (min) = 3' 10-8 atm


t
po~ (max) = 6.1'10- 3 atm

Fig. 152. Rate of attack jSl of silicon in oxygen-helium


mixtures of constant total pressure at 1410°C as a function
of oxygen partial pressure p8" according to Wagner.
DslO is the diffusion coefficient of 8iO molecules, and
PSIO/RT = GSIO is their concentration at the surface
and IlslO is the effective thickness of the boundary layer
for diffusion of 8iO.

Darken and co-workers1 could realize conditions corresponding to an


active and passive state during the oxidation of copper, iron, nickel, and
other metals at temperatures where the vapor pressure of the metal is
significant while the vapor pressure of the corresponding oxide is insignificant.
According to Wagner,2 one obtains for the rate of metal vaporization which is
equal to the rate of oxide film formation

. 2D O ,PO,( vDMePMe)
JMe = 1 +- ---
vRTLl 2D o ,po,
or
2Do,po,
jMe ~ if DMe ~ Do, and Po, ~ PMe
vRTLl
where v is the number of oxygen atoms reacting with one metal atom and
D Me and Do, are the diffusivities of metal atoms and oxygen molecules,
respectively; Ll is the effective thickness of the total diffusion boundary
layer.
1 Turkdogan, E. T., P. Grieveson and L. 8. Darken: J. Metals 14, 521 (1962).
2 Wagner, C.: J. Electrochem. Soc., in press.
4.5. Scaling Systems with Protective Layers Containing Several Phases 295

At a critical oxygen pressure po.(crit), which is determined by the


relation

vLlPMe( RT )1/2
po.(crit) = - - ---
2Do• 27TMMe

with M Me the atomic weight of the metal, the vaporization mechanism breaks
down, i.e., a protective layer of metal oxide is formed and the rate of oxide
vaporization is strongly decreased.
The decrease in the rate of metal oxidation after formation of a solid
oxide layer is especially high for zinc oxidation. If one considers that the
rate of ZnO film formation at 400°C with pZn = 1.0 X 10-4 atm is
2.5 x 10-5 mole Zn/cm 2 -sec when the metal surface is free of oxide and other
impurities, then the oxidation rate in the presence of an oxide layer with
2.5 x 10-10 mole Zn/cm 2 -sec after 1 hr1 represents a strong decrease.
The oxidation of copper-beryllium alloys is another typical example of
an oxidation system with passivity. While the oxidation rate of copper
with beryllium smaller than 6.6 at. % is about the same as for pure copper,
a copper-beryllium alloy containing 12.6 at. % Be is oxidized at a much
lower rate under formation of a very thin, dense outer layer of BeO. 2 A similar
mechanism acts in the oxidation of copper-zinc and copper-aluminum
alloys. If the aluminum content is high enough, aluminum atoms migrate
toward the surface and react preferentially in comparison to copper and
zinc in view of the highly negative free energy of AbOa formation. If Cu 2 0
is initially formed then by the displacement reaction

AbOa is formed. Thus the scale is supposed to consist exclusively of Al 2 0a


which is responsible for the passivity. According to Wagner the excellent
resistance to oxidation of iron-chromium-aluminum alloys may be inter-
preted in the same way.
A typical example of inhibition according to Wagner is the reduction
of the oxidation rate of nickel in presence of Li 2 0 vapor (see Fig. 71) which
does not cause a significant change in the affinity of oxidation of nickel.
Another example of inhibition is the decrease in the rate of chlorination or
bromination of silver due to addition of small amounts of cadmium as an
alloying element. A further example is the decrease in the oxidation rate of
a metal in a gas phase involving two oxidizing components. When nickel

1 Gensch, C., and K. Hauffe: Z. phys. Chern. 196, 427 (1951).


2 Maak, F.: Z. Metallk. 52, 538, 545 (1961).
296 4. Scalin~ Processes with Formation of Thick Protective Layers

is heated in an argon-iodine stream with small amounts both of iodine and


oxygen, then instead of the reaction
Ni(s) + I2(gas) = NiI2(gas) (4.90a)
due to NiO formation the following reaction is established:
(4.90b)
The rate of the last reaction is determined by the equilibrium pressure
PNiI.( eq) and the rate of diffusion of NiI2(gas) from the surface of the sample
into the bulk stream, according to an equation analogous to equation (4.89):

jNn. = DNiI.PNiI.(eq)/(oRT)
The standard free energy change of reaction (4.90b) is about 37 kcal at
700°C.1,2 Thus the ratio PNiI.(eq)/PI. is about 10-7 for Po. = 0.01 atm.
Hence the rate jNiI. in the presence of a NiO layer is about 107 times lower
than at a bare nickel surface. In agreement with the definition of an inhibi-
tion this lowering effect due to presence of oxygen is a typical inhibition
since the oxygen does not alter the affinity of the basic reaction (4.90a).
The phenomena discussed in this section are considered in order to
stimulate the development of appropriate methods for increasing corrosion
resistance at elevated temperatures.

4.6. Scaling of Iron Alloys


Because of the industrial significance of iron alloys and steels and
especially because they are used at higher temperatures in different corroding
atmospheres, a special section is devoted to the scaling processes on iron
alloys. Without considering the mechanical properties of the alloys, we will
discuss in the light of the oxidation mechanism, possible ways of reducing
the oxidation rates of iron and steel, and compare our conclusions with
experimental results reported in the literature.
As was shown above, at temperatures over 570°C the oxide layer consists
of FeO, which is a p-type conductor with a relatively high concentration
of holes EEl and iron ion vacancies, FeD". It can be seen immediately from
the Wagner scaling formula of Section 3.2 that at a constant oxygen pressure
Po. or chemical potential /1-0, the oxidation rate is governed by the partial
conductivity of the iron ions "tFe(" = total conductivity and tFe = trans-
ference number of the iron ions), since to'- is about equal to zero, and because
of the predominant electron conduction, the transference number of the
1 Schafer, R., R. Jacob and K. Etzel: Z. anorg. u. allgem. Chern. 286, 42 (1956).
2 Kiukkola, K., and C. Wagner: J. Electrochem. Soc. 104, 379 (1957).
4.6. Scaling of Iron Alloys 297

electrons te- is about equal to 1. Thus, the possibility exists that the oxida-
tion rate of iron can be decreased by decreasing the partial conductivity of
the iron ions, which is the rate-determining step. If one further considers this
partial conductivity

and assumes that the mobility UFeO" of the iron ions or iron. ion vacancies
to a first approximation is independent of the concentration of the impurity
ions in the FeO layer, which is certainly permissible with small foreign
ion contents, the partial conductivity of the iron ions can only be decreased
by decreasing the concentration of the iron ion vacancies XFeO" This can
be achieved by alloying the iron or steel with a metal whose ions enter the
scaling layer during the oxidation process in monovalent form, or by per-
mitting oxides with monovalent cations to diffuse from the surroundings
into the forming scaling layer. Here the ionic radius of the impurity metal
ion must approximately correspond to that of the iron ion.
As we have seen in preceding chapters, upon introduction of mono-
valent metal ions, for example in the form of Li20, on basis of electro-
neutrality the number of iron ion vacancies must decrease and the number of
electron holes must increase. The corresponding lattice defect equations read
+ FeO" ....---.2Li./(Fe) + FeO
Li 2 0
or
io~g) + Li 0....---.2Li./(Fe) + 2(B + 2FeO
2

and from the defect equation

the mass-action law obtained at constant oxygen pressure reads


XFeonX; = K (Po, = const ) (4.91)
For the introduction of Li 2 0 into FeO-in exactly the same way as was done
for NiO-the ratio of the FeO" and EB concentrations is controlled by the
mass-action equation (4.91). Since the heterotype mixed phases formed
frequently have only a narrow stable range, suitable additions would be
no more than 3 at. % or mol. %. On the other hand, since FeO can exhibit iron
ion vacancies up to lO mol. %, this method of decreasing the oxidation rate
is not particularly promising. This reasoning is supported by experiments
on the influence of Li 2 0 on the FeO formation rate by Brauns and RahmeI.I
Furthermore, from the standpoint of the lattice-defect phenomena in
oxides, the introduction of foreign metal ions of higher valence is excluded,
since it increases the number of the iron ion vacancies as well as the oxidation
1 Brauns, E., and A. Rahmel: Werkstoffe u. Korrosion 7,448 (1956).
298 4. Scalin~ Processes with Formation of Thick Protective Layers

rate, as is to be expected, for example, in the case of small chromium addi-


tions (::; 1 at. %) on the basis of the associated defect equation,
(4.92)
Similar oxidation accelerating effects should be observed even with small
additions of titanium, tungsten, molybdenum, aluminum, etc. to unalloyed
iron and steel, provided no other mechanism comes into play.
Conditions are considerably different when the alloying additions are
so high that the formation of a heterotype FeO mixed phase fades into the
background and new oxide phases with better protecting properties appear,
as is the case with a spinel formation-for example, chromium and chromium-
nickel steels-or when the alloying additions are so much less noble that
they are preferentially oxidized out of the steel and form a compact, oxide
barrier layer. We will concern ourselves at first with the last case and will
consider the mechanism of the spinel formation later.
As we saw from the preceding presentation, the high oxidation rate of
iron and unalloyed steel above 570°C can be attributed to the rapid rate of
formation of the FeO phase. However, this "dangerous" FeO formation
can be hindered by alloying to the iron a metal whose oxide preferentially
forms on the iron surface and eliminates the contact between the iron and
iron oxide. As a consequence of this barrier layer of impurity oxide, the
accumulation of iron ions is greatly retarded, and because the FeO phase is
unstable, Fe3 04 is formed as the lowest oxide. The oxidation experiments on
iron-molybdenum alloys by Brenner! shed some light on the approach.
As can be seen in Fig. 153, the oxidation rate of iron in oxygen at 1 atm
Wr---'-~~---r---'---'---'

2'1 --H----+----+-JL--

18 ----1-f------6"

Fig. 153. Parabolic course of the oxida-


8 tion of iron and iron-molybdenum alloys
~~~L-~~~~--~__~~ at 1000°0 and in flowing oxygen (about
1 liter/min) at 1 atm, according to
Brenner.

and lOOO°C decreases with small additions of molybdenum (sec also Table 36).
This finding indicates primarily that the scaling layer that is formed cannot
be composed of an FeO-Mo0 2 heterotype mixed phase, since introduction of
1\100 2 into the FeO lattice would eau"e an increase in the number ofthe FeD"
1 Bronnor, S. S.: J. Electrochem. Soc. 102, 7 (UJ55).
4.6. Scaling of Iron Alloys 299
Table 36. Decrease in the Oxidation Rate of Iron-Molybdenum Alloys with
Increasing Molybdenum Content Between 800 and 1OOO°C, According to Brenner
Mo, k" X 108, A, LJE,
at.% T,oC g2/cm4·sec g2/cm4.sec cal/mole
1000 80.0 1.2 36,000 ± 1000
0.64 1000 5.8
1.83 800 0.019 1.0 x 103 63,000 ± 3000
850 0.048
900 0.18
950 0.75
1000 1.60
2.61 1000 1.22
5.29 1000 0.72
6.16 800 0.23 23.0 54,000 ± 2000
850 0.56
900 0.15
950 0.50
1000 1.15
6.51 1000 0.66
12.5 800 0.022 1.2 x 10-3 33,000 ± 1000
850 0.053
900 0.087
950 0.16
1000 0.26
k" = A exp (- ILJE/RT)

positions as well as in the oxidation rate. Photomicrographs and X-ray


diffraction experiments have confirmed that the scaling layer consists of
three distinct layers (see Fig. 154). Of these, the outer layer is Fe20a, the

Fig. 154. Structure of the scaling layer of an iron-molybdenum aHoy with


0.64 at. % Mo after 350 min of oxidation in oxygen at 1 atm at 1000°C,
according to Brenner (enlarged 250 times). 1: F8203; 2: Fe304; 3: MoOz
+ (FezMoy)O; 4: alloy.
300 4. Scaling Processes with Formation of Thick Protective Layers

middle one Fe304, and the inner layer is an oxide mixture of unkown com-
position with most of the Mo0 2. No FeO layer could be detected.
This heterogeneous structure of the scaling layer can be explained in
terms of the Wagner theory of oxidation of metal alloys, which will be
discussed later, in the following way: On basis of the low dissociation pressure
of FeO and Fe 304, for example, 1.69 x 10-15 and 6.08 x 10-15 atm at
1000 o e, compared to that of Mo0 2 with 2.99 x 10-14 atm, in the first stage
of the oxidation, FeO and Fe304 are preferentially formed with a small
amount of Fe203, without detectable Mo0 2 formation. Since iron is con-
tinually oxidized out of the regions near the surface of the alloy and the
molybdenum enrichment effected in this way was only slightly decreased
by diffusion to the interior of the alloy, a critical concentration of molyb-
denum ultimately arose which led to Mo0 2 formation. With consideration
of the two reaction equations
2Fe + O 2FeO
2 ,,----

Mo + 02-------Mo02

we obtain the following relationship (see Section 4.7.1):

where 7TFeO and 7TMoO, represent the dissociation pressures of FeO and MoOg
and X(i)Fe is the atom fraction of iron in the alloy at the alloy-scaling layer
phase boundary. Using the above values for the dissociation pressures we
conclude that under ideal conditions at 1000 o e, an Mo0 2 formation must
appear with X(i)Fe = 0.21. Therefore, as soon as the iron content
becomes smaller than 0.21, FeO will be replaced by Mo0 2 along the alloy/
scaling layer phase boundary. In this way the FeO layer is separated from
the iron below it, and this causes the FeO to become thermodynamically
unstable during the continued oxidation, and it is then further oxidized
to Fe 304. The considerable decrease in the oxidation rate of the iron-molyb-
denum alloys is understandable from the disappearance of the FeO phase.
No catastrophic oxidation appeared through molybdenum additions as
was observed in copper and nickel alloys by Meijering and Rathenau,1
Figure 155 is a schematic representation of the growth mechanism based on
the development of the scaling layer. The extent to which a migration of
0 2- and Mo4+ ions through the Mo0 2-(Mo x Fey)0 layer takes place cannot
be determined at this time. The mechanism of formation of the two outer
layers, Fe 304 and Fe203, is identical to that of the oxidation of pure iron,
which was discussed in Section 4.5.

ll\Icijering, J. L., and G. W. Rathenau: Nature (London) 165, 240 (1950).


4.6. Scalin~ of Iron Alloys 301

Additional investigations on the oxidation mechanism of iron-molyb-


denum alloys with molybdenum contents up to 5.6 wt. % have been carried
out by Rahmel, Jager, and Becker.1 They observed that 2 wt.% Mo caused
the strongest decrease in the oxidation rate. The scaling layer of iron-
molybdenum alloys oxidized between 500 and 620°0 is composed of three
layers. Under the outer layers of hematite and magnetite, there exists a
thin layer on., the alloy with the composition Fe4Mo6016. An oxidation be-
tween 620 and 700°0 also produces three different oxide layers, but now the
inner layer is a two-phase layer with wustite and the spinel Fe2Mo04.
Above 700°0, however, the scale consists. of four layers where a closed
wustite layer is formed between the magnetite and the inner two-phase
layer. These results differ from the oxidation experiments at 1000°0 (see
Fig. 153).

Fe-Mo MoOz+ Fe304 Fez 03 oz'3)


(Moxf~)O

Fig. 155. Schematic representation of the Enric hment f+-Mo4+


(J2- ..-
diffusion processes through the individual inM a FeZt
layers of the scale forming on an iron- __ Mo4+ lFe 3+ Q2=-
molybdenum alloy at 1000°0 in oxygen.

With alloying additions such as chromium and aluminum, which are


inclined to spinel formation in oxidation with divalent metal ions, the
scaling layer: consists of a chromium-iron spinel or a chromium-aluminum-
iron mixed spinel. Nevertheless, in chromium-containing steels one also
has to deal with a predominant Or20a formation, if the oxidation tempera-
ture is not too high, as indicated by Lustman 2 and others. On basis of electron
diffraction procedures, Hickman and Gulbransena were able to determine
a predominant Or20a formation in iron-chromium itlloys with 1-13 wt. % Or
contents when they were oxidized in an oxygen atmosphere of 1 mm Hg
at 600°0 for 1 hr. The alloy with 13 wt. % Or also shows spinel formation;
this was not observed with the other alloys. Unlike the case of pure iron,
an Fea04 formation appears in a 13% Or-Fe alloy far above 5700 0-at
700 0 0-which is only possible in the presence of chromium. Mahla and
Nielsen 4 observed the same results when they found a tarnishing layer
composed of IX-Fe 20a and Fea04 on chromium-iron alloys with 12 and
27 wt. % Or after an oxidation time of 10 min at 675°0. The composition
1 Rahmel, A., W. Jager and K. Becker: A~h. Eiaenhilttenw. 30,351 435 (1959).
2 Lustman, B.: Iron & Coal Trades Rev. 154,889 (1947).
3 Hickman, J. W., and E. A. Gulbransen: Trans. AIME, Tech. Pub!. No. 2069 (1946).
4 Mahla, E. M., and N. A. Nielsen: The Electrochem. Soc. Preprint 87-27 (1947).
302 4. Scaling Processes with Formation of Thick Protective Layers

of the scales on iron alloys containing from 0.063 to 4.69 wt. % Al at 950°0
was determined by Schmahl, Baumann, and Schenck. 1 These investigators
showed that a higher diffusion rate of silicon in the O(.alloy phase containing
silicon accounts for the formation of a better protective scale. 2
The oxidation mechanism of iron-carbon alloys are in Section 6. a
The oxidation mechanism of the chromium steels appears to be quite com-
plicated in that besides the three iron oxides FeO, Fea04, and Fe20a, the
chromium oxide also appears in the scaling layer, and this leads to addi-
tional spinel reactions in the scaling layer. As microscopic investigations
by Moreau4 have shown, FeOr204 "islands" are formed in the FeO phase
neighboring the alloy, while an iron-chromium mixed spinel appears in the
Fe304 layer which reaches a high concentration at the FeO/Fea04 phase
boundary and decreases with increasing distance from the FeO/Fea04
phase boundary until finally it is no longer detectable. The exterior part of
the scaling layer in any case consists almost entirely of pure 0(.Fe 203. After
longer oxidation times Moreau could no longer detect Fea04, since the Fe304
layer which was originally present had been almost completely converted
into a mixed spinel layer. Furthermore, in the exterior part of the scaling
layer an iron-chromium oxide mixed phase appeared in place of the 0(-Fe 20a.
The results of these investigations, which are schematically presented in
Fig. 156, confirmed and enlarged upon earlier observations, and may be
~Ir--------------'
a

-Fe3 o..
+
Mixed Spinel
3' C 3+,.z+IO"
[ Fez- x rJ; . . J 4

FeD
Fig. 156. Schematic representation of the
structure of the scaling layer of an iron-
chromium alloy oxidized between 800
and 1250°0, according to Moreau.

analyzed in the following way: In the first stage of the reaction the supply
of iron is relatively great because of the easy FeO formation, so that a
preferred FeO formation occurs at high temperatures. The increasing
1 Schmahl, N. G., H. Baumann, and H. Schenck: Arch. EisenhuUenw. 30, 345 (1959).
2 Schmahl, N. G., H. Baumann, and H. Schenck: Arch. Eisenhuuenw. 30,267,415 (1959).
3 Bohnenkamp, K., and H. J. Engell: Arch. Eisenhuuenw. 33,359 (1959).
4 Moreau, J.: Campt. rend. 236,85 (1953); These Paris 1953; J.Moreau and J. Benard:
Campt. rend. 237, 1417 (1953).
4.6. Scalin~ of Iron Alloys 303

enrichment of chromium in the alloy at the alloy/oxide phase boundary


which occurs with progressive oxidation is caused as a result of chromium
oxidation with an increasing supply of Cr203 at this phase boundary. Because
of the low solubility of Cr203 in FeO, we find, on the one hand, the formation
of FeCr spinel nuclei, which grow in the course of time, and on the other
hand, a diffusion of chromium into the outer, initially thin, Fe304 layer,
where conditions are favorable for mixed-spinel formation. Through the
continuous consumption of chromium in the spinel phase of the scale a
relatively larger gradient of the chemical potential of the chromium ions
in the FeO phase is introduced, and the preferred diffusion of the chromium
into the spinel phase thus becomes understandable. On the other hand, in
the mixed-spinel zone the diffusion rate of the metal ions is decreased, and
thus the scaling stability of the alloy is finally increased.
Recently, Benard and co-workers1 reported on new investigations into
the oxidation kinetics of austenitic stainless steels with 18% Cr-8% Ni at
1050°C in air. On basis of these results, which are shown in Fig. 157, we have

30
~
, ~
,V,
iK I
B1
I
___c
··
,
I
I
I
,I
r-sV'"
·,,·•· ·, ,
I

A
!-! -. ··•
I A".A_..... _.i!
o ---- 25 50 75 700 "fUrs

Fig. 157. The reaction periods during the oxidation of stainless


steel at 1050°0 in air, according to Benard and co·workers.

to distinguish among three oxidation periods. The first one from 0 to A


is characterized by a slow oxidation, the second one from A to B by a rapid
oxidation, and the third one from B to C and beyond by a slow oxidation.
With difficult specimens under the same experimental conditions it was
not possible to reproduce the initial oxidation period OA, which can vary
from several to lOO and more hours (OA', OA "). The oxidation rate obeys a
parabolic rate law. After microscopic, X-ray, and electron diffraction
measurements at the various oxidation periods it was found that in the
first oxidation period the oxide layer is composed of FeCr204 and Cr203,
1 Benard, J., J. Hertz, Y. Jeannin, and J. Moreau: Oompt. rend. 248, 2095 (1959).
304 4. Scaling Processes with Formation of Thick Protective Layers

whereas in the second period a predominant Il(-Fe203 formation sets in, which
seems to be responsible for the rapid acceleration of the oxidation.
Unlike the mechanisms for face-centered cubic and body-centered cubic
oxide lattices, the diffusion mechanism in the spinel lattice is at present
not particularly well understood, and we are not in a position today to pro-
pose a model concerning the capability of influencing the diffusion rate of the
ions in the spincllattice. Besides, whether the spinel lattice conforms to the
same rule in the influencing of the ionic lattice defects as do the cubic oxides
is subject to question. Recently Schmalzried and Wagner l discussed the
lattice-defect mechanism in ternary ionic crystals, for instance, spinels. They
were able to demonstrate convincingly the decisive importance of the
determination of the thermodynamically independent variables which are
necessary for the evaluation of diffusion and conductivity measurements
in ternary ionic crystals. Thus, for example, the statement of self-diffusion
coefficients in the spinel phase is worthless unless the thermodynamic
activity of one of the oxides forming the spinel, as well as the temperature
and the total pressure, are known. Schmalzried 2 was able to confirm this
expcrimentally by self-diffusion measurements of iron through Fe304 as a
function of the logarithm of the pco 2 /Pco ratio (Fig. 158). The increase

-8

FeO

-10

Pco.,
2 4 log-p..
3 co

Fig. 158. The dependence of the self-diffusion coefficient in Fe3 04 on


the oxygen partial pressure at 1115°C, according to Schmalzried.

in the diffusion coefficient with increasing oxygen partial pressure, i.e., with
increasing iron deficiency, points to a vacancy diffusion mechanism. This
indicates that in the formation of scaling layers of the spinel type there
are jump possibilities for the diffusing ions in the basic lattice, dependent
on the oxygen pressure and on the thermodynamic activity of one of the
1 Schmalzried, H., and C. Wagner: Z. physik. Chem. [NF] 31, 198 (1962).
2 Schmalzried, H.: Z. physik. Chem. [NF] 31, 184 (1962).
4.6. Scaling of Iron Alloys 305

oxides forming the spinel. However, the height of the "jump barriers," which
manifests itself primarily in the magnitude of the activation energy of the
diffusion, largely determines the extent of the diffusion, while the change in
the number of "jump places" due to the change in the number of vacancies
only affects the pre-exponential factor. On this basis the change in the diffusion
rate of the ions appearing in the mixed spinel phase, which is indirectly
measurable through a change of the oxidation rate of a corresponding alloy,
can be attributed to the change both in the vacancy concentration and in the
height of these barriers appearing in the mixed· spinel formation. Investiga-
tions in this direction would be desirable and informative for the further
explanation of this mechanism.
If a scaling layer that consists primarily of a chromium spinel does not
fulfill the functions of a protective scale to a sufficient degree, a suitable
amount of a metal which forms a spinel with chromium, iron, or some other
alloying element during oxidation and has a lower vacancy concentration
or a higher barrier and thus lower diffusion rates than the original iron-
chromium or nickel-chromium spinel may be added. Aluminum is obviously
such a metal. The addition of aluminum can, depending on the concentra-
tion and conditions of oxidation, exert a favorable influence in one of two
ways. Either an iron-aluminum spinel will be formed preferentially in the
course of oxidation, or an Al 20 a layer (or a layer of a ruby-type AI 20 a-
Cr20a mixed oxide) will form preferentially between the steel and the scaling
proper, as has been shown by experiment) Such aluminum additions seem
to have a retarding influence on the oxidation rate similar to that found by
Hessenbruch 2 for the addition of cerium and thorium to resistance wires.
In both cases the diffusion rate of the ions is lowered by the appearance of a
barrier layer (of spinel or AhOa in the one case and of Ce02 or Th02 in the
other), which finally leads to a reduction in the oxidation rate. Another
promising possibility is the joint addition of aluminum and beryllium or
titanium and magnesium in such concentrations as to favor a preferential
formation of beryllium-aluminum spinel or magnesium-titanium spinel
(perovskite). Experiments on this-but mostly in copper-based alloys-will
be discussed in Section 4.7.
The appearance of a barrier layer has also been observed at low silicon
contents in chromium steeP As Portevin, Pretet, and Jolivet4 demonstrated

1 Kornilow, J. J., and A. J. Schpikelmann: Dokl. Akad. Nauk SSSR 53, 813 (1946).
2 Hessenbruch, W.: Metalle und Legierungen fur hohe Temperaturen, Springer, Berlin,
1940.
3 Kornilow, J. J., and A. J. Schpikelmann: Dokl. Akad. Nauk SSSR 54, 5ll (1946); D, M.
Dovey and J. Jenkins: J. Inst. Metals 76,581 (1950); B. Lustman: Trans. AIME 188,
995 (1950).
4 Portevin, A., E. Pretet, and E. Jolivet: Rev. met. 31,219 (1934).
306 4. Scaling Processes with Formation of Thick Protective Layers

on basis of oxidation experiments with silicon steels, no silicon can be found


in the iron oxide layers. All the oxidized silicon is in the form of Si0 2 at the
steel/scale phase boundary. Further investigations by Caplan and Cohenl
confirmed these observations. Evidently this phenomenon is the result of the
large heat of formation of the Si02-similar to the Al 20a-at high tempera-
tures compared to that of the other oxides which appear. A reduction in the
chromium or iron oxides with separation of Si02 is effected by the
presence of silicon especially in chromium steel at the steel-oxide layer phase
boundary, which in the further course of the scaling process can cause the
oxide to crack off owing to excessive enrichment of Si02. The alloy exposed
in this way re-covers itself with an oxide layer, which also cracks open
after attaining a critical Si02 concentration. The course of the oxidation
curve is thus stepped, the stepped form periods being relatively long, up
to 200 hr, as is shown schematically in Fig. 159. In such cases it is important

:§ A
~ .---- Fig. 159. Schematic representation of the oxidation
course of an iron-chromium alloy (0.20% C; 0.40% Mn;
0.44% Si; 0.32% Ni, and 26.54 wt.% Cr) between 800
200 Hours - and 1l00°C, according to Caplan and Cohen.

that the experiments proceed through as long a period as possible, since


insufficiently long observation times will produce only the flat AB part of
the oxidation curve, which of course, would indicate a much too favorable
scaling stability.
On the other hand, there are a number of molybdenum-silicon-chro-
mium, molybdenum-chromium-titanium, and 18-8 steels which obey the
parabolic rate law without discontinuities in the oxidation-rate curve after
an oxidation time of 15,000 hr, for example, results from oxidation experi-
ments in water vapor atmospheres at 495°C in Fig. 160, according to Rohrig,
van Duzer, and Fellows. 2 According to these experiments the chromium
steels with about 12 wt. % Cr without nickel are more resistant to water-
vapor attack than the 18-8 steel. Solberg, Hawkins, and Potter3 came to
essentially the same conclusions, finding that the 18-8 steels stabilized with
niobium showed an especially high resistance to scaling. The decrease in

1 Caplan, D., and M. Cohen: J. Metals 4, 1057 (1952).


2 Rohrig, I. A., R. M. van Duzer, and C. H. Fellows: Trans. ASME 66,277 (1944).
3 Solberg, H. L., G. A. Hawkins, and A. A. Potter: Trans. ASME 64, 303 (1942).
4.6. Scalin~ of Iron Alloys 307

Inchr----r-.--.--.-,----,----r----r----,----,

J
Dim

r 5
8
9
f
j

....
"...
0
-.I

fJ

11
516
3 I 6 IJ KJ 12 N' ,pi
Hours _

Fig. 160. Course with time of a few stools in water vapor at 595°C, according to Rohrig,
van Duzer, and F ellows. The composition of the steel is as follows:

No. Steel type C Si 1I1n P S Cr Ni Mo Al V Ti wt.%

1 SAE 1010 0.11


2 SAE 1035 0.34
3 Cr-Mo-Si 0.10 1.40 0.41 0.01 0.013 1.24 0.58
4 8i-Mo 0.11 1.35 0.19 0.01 0.012 0.50
5 SAE 6120 0.24 0.28 0.65 0.02 0.021 0.94 0.18
6 5 Cr-1II0- 8i 0.10 1.55 0.30 0.01 0.016 4.83 0.51
7 C-1IIo 0.15 0.34 0.40 0.023 0.Ql5 0.49
8 1.25 Cr-Mo-8i 0.09 0.78 0.41 0.01 0.013 1.21 0.58
9 5 Cr-Mo 0.13 0.32 0.45 0.01 0.016 4.96 0.52
10 C-1IIo 0.18 0.35 0.70 0.51
11 5 Cr-Mo-Ti 0.06 0.41 0.36 0.01 O.OOS 5.18 0.19 5.58 0.46
12 Nitralloy 0.35 0.25 0.40 0.02 0.02 1.25 0.20 1.20
13 12 Cr-2 Mo 0.05 0.30 0.13 0.024 0.01112.18 0.10 2.01
14 18 Cr-8 Ni
Type 304
15 12 Cr,
0.07 0.16 0.14 - 17.16 10.20
Type 420 0.32 0.22 0.36 0.014 0.00813.57
16 12 Cr,
Type 416 0.11 0.25 0.35 0.016 0.30812.27
17 12 Cr,
Type 403 0.10 0.20 0.35 0.017 0.02212.33 0.18
308 4. Scaling Processes with Formation of Thick Protective Layers

the oxidation rate of a chromium steel with increasing chromium content


is plotted in Fig. 161.
Cornelius and Bungardtl investigated the scaling behavior of chromium-
nickel steels alloyed with vanadium. While a chromium-nickel steel with
17 wt. % Cr and 20 wt. % Ni shows good resistance to scaling at
930°C in air at 1 atm, with the addition of 2 wt. % V, the rate was increased
by about two orders of magnitude. This (dominating) influence of vanadium
can evidently be curtailed by decreasing the nickel content to about 8 wt. %
and the chromium content to about 12%. The detailed discussion of Section
4.3 leads to the conclusion that there is a preferred V 205 formation in
vanadium-containing steel during oxidation, which forms a low-melting
eutectic oxide mixture with the other oxides in the scaling layer, and thus
the protective formation of the spinel or of the Cr 203 is prevented. The
experimental results of scaling experiments with low-alloy steels below
900°C-especially at 650°C-are in accord with these considerations. The
(!(}2(J

'nch

\ I
\
"*-.
.5
Weight
~
10
%Cr-
-
1.5 2(}
Fig. 161. Influence of chromium content on the
thickness of the scale formed after 1300 hr in
water vapor at 650°C on steel, in inches, accord-
ing to Solberg, Hawkins, and Potter.

protective effect here is due to the fact that the melting point of the eutectic
oxide mixture has not been reached and that the V205 together with the iron
oxides forms an adherent pore-free protective layer. Molybdenum admixtures
to steel can cause a similar effect, as has already been pointed out. At high
temperatures (lOOO°C or higher), not only the partial melting of the scale,
but also the high evaporation rate of Mo03 is responsible for the high rate of
oxidation. 2
A series of cast irons based on chromium-nickel-cobalt-iron-tungsten
alloys with 18 to 20 wt. % Cr and varying amounts of nickel, cobalt, iron,
and tungsten with 0-1.31 wt. % B were oxidized in air between 900 and
1l00°C by Binder and Weisert. 3 Iron and boron and to a lesser extent cobalt
1 Cornelius, H., and W. Bungardt: Arch. Eisenhuttenw. 15, 107 (1942).
2 Kubaschewski, 0., and O. von Goldbeck: Metalloberflache (A) 7, 113 (1953); O.
Kubaschewski and A. Schneider: J. Inst. Metals 75, 403 (1949).
3 Binder, W.O., and E. D. Weisert: Corrosion 9,329 (1953).
4.6. Scalinl1 of Iron Alloys 309

adversely affect the resistance to oxidation. Because of the relatively high


chromium content of the alloys, molybdenum-in agreement with the
experimental results by Brenner1-effects a strong degeneration of the
scaling stability. X-ray diffraction experiments have shown that the Cr20a
or spinel protective layer cannot form in the presence of higher boron
concentration unless the iron content is chosen at less than 6 wt. % Mo
and is replaced by tungsten as an alloying element for resistance to deforma-
tion at high temperatures, and unless cobalt is largely replaced by nickel.
Uhlig and Brasunas 2 investigated the influence of the magnetic trans-
formation of chromium steels at the Curie point on the oxidation rate of these
alloys. Steels with 0.5 to 1 wt. % C, 0.2 to 0.3 Si, 0.3 to 0.4 Mn and the
following Cr contents: 9.22,10.59,12.22,13.84,19.12, and 24.08 wt.% were
investigated. A definite decrease of the oxidation rate was always observed
below the Curie point, for the Curie points (which decrease from 750 to 615°C
with increasing chromium content) of the above-mentioned steels. In Fig.
mg/cm 2 -day - 1
to I
8
.'\.
I

8
'\ ~- 1.J.8'1~Cr- I---
~
\ .'b
I::

\ ~
~
1
'" ~ "-.....0
Fig. 162. Temperature dependence of the
0.8 9.2 ;tZ8 11.2 11.8
rate of oxidation of an iron-chromium alloy 9.8 ;tZ~--'!
with 13.84 wt.% Or in oxygen at 1 atm, ! T I

according to Uhlig and Brasunas. 800 75{} :w

162, e. , the course of oxidation with time of a chromium steel with


13.84 wt. % Cr, is plotted as a function of the temperature. Earlier observa-
tions by Tammann and Siebe13 were confirmed and extended.
When the attacking gas consists not only of oxygen, but also contains
S02 and C02, the scaling mechanism is still more complicated. According to
Preece,4 in the absence of oxygen, combustion gases with 0.15% S02 attack
a low-alloy steel at 850°C with considerable formation of sulfide in the scale.
Experiments show that with increasing oxygen content the S02 influence
in combustion gases is reduced, and at 4% oxygen is eliminated completely.
1 Brenner, S. S.: J. Electrochem. Soc. 102, 7, 16 (1955).
2 Uhlig, H. H., and A. de S. Brasunas: J. Electrochem. Soc. 97, 448 (1950).
3 Tammann, G., and G. Siebel: Z. anorg. u. allgem. Chem. 148,297 (1925).
4 Preece, A., J. Iron Steel In8t. 139, 149 (1951).
310 4. Scalini1 Processes with Formation of Thick Protective Layers

Thus with 4 %oxygen no sulfide formation occurred, as shown schematically


in Fig. 163. One may therefore conclude that with more accurate feeding

Fig. 163. Schematic representation of the rate of


oxidation of lower alloyed steels between 600 and
1200°C as a function of the composition of the
combustion gases, according to Preece. Curve (a)
S02 free, curve (b) atmosphere containing 0.15%
'1321 : ' 2 3 ' 1 S02. (While the attack of S02 in the presence of
%CO of the :%02ofthe free oxygen is very small, S02 in oxygen-free com-
Combustion Gases Combustion Gases bustion gases effects a strong scaling.)

of the gas or oil firing with an oxygen excess, a sulfur content in the fired
gas-if it appears as a single impurity-is without influence on the oxidation
rate of the steel. In this connection an investigation on the influence of
K 2S04 on the oxidation rate of stainless steels in a furnace atmosphere
containing sulfur oxides is worth mentioning.! Above 625°C an intensified
oxidation is observed because of the formation of potassium-iron (III)
sulfate [KaFe(S04h] which with surplus K 2S04 forms an eutectic mixture
melting at 627°C_ However, the rapid oxidation disappears above 800°C
because KaFe(S04)a is not stable in this temperature region in the presence
of sulfur (VI) oxide.
While a chromium or aluminum addition to steels is generally prefer-
entially oxidized, which is observable from the appearance of a Cr20a or
Al 20 a or spinel layer, nickel in steel in the absence of chromium and aluminum
is oxidized out to a much smaller extent, as Moreau's2 investigations im-
pressively demonstrated. Corresponding to the phase diagram of the mixed
oxide FeO + NiO in Fig. 164, for example, at 800°C only about 2 wt. %NiO
is soluble in FeO. Higher nickel contents must favor an Fea04 formation
according to

which is of significance, for example, for the displacement of the FeOjFea04


point toward higher temperatures. In this connection the oxidation experi-
ments with iron-nickel alloys with 42 wt. %Ni between 600 and 900°C in

1 Rahmel, A.: Arch. Eisenhitltenw. 31,59 (1960).


2 Moreau, J.: "Recherches sur l'oxydation des alliages binaires Fe-Ni et Fe-Cr aux
temperatures elevees," These Paris (1953).
4.6. Scaling of Iron Alloys 311

air by Foley! are quite informative. On basis of electron diffraction experi-


ments and chemical analyses, which yield besides Fe20a a ferrite structure
and a relatively high nickel content, the formation of NiFe204 appears to
be more probable than that of Fea04. The rate-determining step is the
diffusion through the nickel-ferrite layer, which reduces the oxidation rate.

V
,,
I
[ Fe Ni]O

,,
1//
POll

,,
, ,,
, Fe,O; +Ni+FeD

, ,
Fig. 164. Stability diagram of the FeO-NiO solid
,,
solution, according to Moreau. The dashed line
corresponds approximately to the equilibrium JIJIJ I
temperature line of the reaction D 2 ,;
3FeO + NiOdlssolved In Feo..->-Fe304 + Ni.
I J "
Weight % NiQ -

Brabers and Birchena1l2 found a preferential subscale formation at grain


boundaries during the oxidation of iron-nickel alloys at 1050°C. An equilibra-
tion under an inert atmosphere occurred mainly by the dissolution and
diffusion of oxygen in the alloy phase. Porosity developed in the subscale
region during equilibration, then began to sinter out on further heating.
By reference to the relative rate of scaling layer growth on pure iron, it is
anticipated that the elimination of wustite as a stable phase should lead to a
decrease in the oxidation rate of the alloys by a factor of about 20 due to
the formation of a spinel phase. In agreement with investigations by Quarrell,3
nickel as well as chromium stabilized the spinel which was formed. The
oxidation rate was

k" = 0.129 exp( -44,200jRT) (alloy)


k" = 0.37 exp( -33,OOOjRT)g2jcm4 -sec (iron)
The above is significant for the use of low-alloy ferritic steels for steam-
boiler construction. Here one wants to shift the temperature of FeO forma-
tion-which for pure iron is 570°C-to higher values in order to raise the
operating temperature of the boiler above 570°C. This can be done at these
higher temperatures only if, by the addition of alloying elements that enter
into the scaling layer during oxidation, one can induce the formation of
1 Foley, R. T.: J. Electrochem. Soc. 102, 440 (1955).
2 Brabers, M. J., and C. E. BirchenaH: Corrosion 14, 33 (1958).
3 QuarreH, A. G.: Trans. AIME 171, 341 (1947).
312 4. Scalin~ Processes with Formation of Thick Protective Layers

Table 37. Chemical Composition of the Scaling Layers Which Form on Ferritic
Steels After 20 hr of Oxidation at 980°C in Air, According to Radavich and Yearian

Alloy composition, wt. % Composition of scaling layer, wt. %


Type C Mn Si N Cr Si Si02 Cr Cr20a Fe Fe20a MnO

A 0.077 0.40 0.23 13.72 12.4 18.1 50.8 72.6


15.2 22.2 44.0 63.0
4.3 9.2 15.2 22.2 46.2 66.0 0.41
B 0.14 0.68 0.74 0.1l5 26.31 51.5 75.2 16.1 23.0
5.1 10.9 50.9 74.5 17.9 25.6 3.84
C 0.05 0.38 0.54 27.88 44.7 65.3 5.1 7.3
49.3 72.0 4.8 6.9
7.3 15.6 50.6 74.0 5.2 7.5 2.40
D 0.022 0.12 0.16 0.016 26.01 57.8 84.5 6.3 9.0
2.2 4.6 58.3 85.6 6.5 9.2 0.79
E 0.009 0.01 0.12 0.018 25.47 65.1 95.0 4.8 6.9
7.3 15.6 65.0 95.0 6.1 8.7 0.73

Fe304, which offers improved scaling protection. Thermodynamic considera-


tions show that this is possible only if one adds a metal with a high oxide
heat of formation, for example, chromium. The higher the heat of formation
of the added oxide, the smaller the alloying addition necessary to achieve
the desired effect. Detailed investigations on the oxidation of ferritic and
austenitic steels were carried out by Yearian and co-workersl at higher
temperatures. An intensive study was devoted to the explanation of the
composition of the scaling layer. From the many experimental data, we have
extracted the analytical results for the scaling layers of a few ferritic steels
from 13 to 25.5 wt.% Cr and summarize them in Table 37. In all cases
electron diffraction experiments established the existence of a scaling layer
which consisted of a solid solution of 0-50 wt. % Cr203 and CX;-Fe203. Proof
of the presence of a spinel phase was complicated by the fact that its lattice
parameter was difficult to determine. As can be inferred from Table 37,
the chromium content in the scaling layer (except for alloy A) was consider-
ably greater than in the metal phase. This phenomenon is further evidence
for the preferential oxidizing out of chromium from the alloy according to
the Wagner theory. Furthermore, it is noteworthy that manganese was
also preferentially oxidized out of the alloy. X-ray investigations and
chemical analyses yield values which agree. In spite of the appearance of
1 Yearian, H. J.: "Investigations of the Oxidation of Chromium and Nickel-Chromium
Steels," Summary Tech. Rep. 1954, Contr. N. 7onr-39419; J. F. Radavich and H. J.
Yearian: High Temperature Oxidation of Some Ferritic Stainless Steels, unpublished
report.
4.6. Scaling of Iron Alloys 313

MnCr204 determined by Longo l and Kortright,2 the quantities of manganese


determined in the scaling layers of alloys D and E were present not as spinels
but predominantly as MnO.
The appearance of Cr2N in the initially formed thin oxide film as
observed by Dulis and Smith3 in creep-rupture experiments on austenitic
steels is significant. However, this nitride formation appeared only on alloy B,
with the highest nitrogen content. In alloys with higher silicon content,
an Si0 2formation at the inner phase boundary was noted. These experimental
results are extraordinarily informative and are particularly impressive
indications of how difficult predictions on resistance to oxidation and its
influence on multi component iron alloys are at this time.
The very few-but very important-examples that have been selected
from among the multiplicity of phenomena that act in the scaling mechanism
of steels are sufficient to indicate how complicated the questions are that
are being raised here and how frequently the absence of sufficient experi-
mental results makes it impossible to furnish definitive answers. Under
present conditions it seems desirable to use alloying additions which either
form a new spinel covering layer with the iron ions or, in the case of pre-
dominant oxidation of the added material, a compact protective layer made
up of its oxide. Either type of layer will drastically retard metal or oxygen
diffusion, as for example the compact Al 20 a or Ce02 formed at the steel/scaling
layer phase boundary. Another possibility is the addition of metals which,
without any appreciable participation by the iron, form adhering impurity
spinel layers, as for instance aluminum together with beryllium, cobalt,
or calcium. Whether rare-earth metals or other alkaline-earth metals are
suitable has not yet been determined. In any case the basic possibilities
for improvement of the scaling stability of steels seem to have been exhausted
except for iron alloys with high tantalum, niobium, or chromium content,
which regardless of the theoretical possibilities they offer, have to be dis-
missed for economic reasons, and besides oxidation experiments on such
alloys are not particularly attractive even from a scientific point of view
Isince the elementary processes involved are not yet clearly understood.
The above conclusions are supported by a critical consideration of high-
temperature steels and alloys by Wetternik. 4 No doubt the further develop-
ment of high-temperature steels and alloys with a view toward increasing
the scaling stability at high temperatures is limited both by the melting
points of the alloys and by the nature of the oxides that make up the
scaling layer. Similarly, the drop-off in the resistance to deformation
1 Longo, T. A.: M. S. Thesis in Physics, Purdue Univ. (1953).
2 Kortright, J. W.: M. S. Thesis in Physics, Purdue Univ. (1953).
3 Dulis, E. J., and G. V. Smith: J. Metals 4, 1083 (1952) .
.4 Wette=ik, L.: Radex Rundschau 1954, 87.
314 4. Scaling Processes with Formation of Thick Protective Layers

at high temperatures and the resistance to creep stress limits the temperature
at which an alloy can be worked. In the development of technologically
valuable highly-scaling-resistant alloys the mechanical properties at the
working temperatures will thus always have to be considered. 1
The scaling of cast iron will not be treated in this chapter since at this
time no reliable results exist on the participation of carbon, which is present
in relatively high concentration. As Scheil2 was able to demonstrate a long
time ago, cast iron can expand by about 13% on heating in air. According to
Benedicks and L6fquist,3 after several heatings and coolings a reversible
volume increase was no longer observed. Thyssen4 concluded that one can
experimentally assume a preferential oxidation of graphite in cast iron and
of the iron in proximity to the graphite. However, this large volume change
cannot be attributed solely to an oxidation process. Rather it seems that the
cementite decomposition which sets in simultaneously plays a determinative
role. An addition of 4-10 wt. % Si effects a considerable lessening of this
effect. Furthermore, just as in normal steel, chromium and nickel additions
effect a further improvement in the resistance to oxidation and the mechanical
technological properties, which are required, for example, for rapid heating
and cooling.

4.7. The InJluence of Metal Diffusion in the Alloy Phase on the


Scaling Rate
In previous discussions of the oxidation mechanism we considered only
the diffusion and transport processes in the scaling layer, assuming that metal
diffusion in the alloy phase proceeds rapidly enough to preclude any possibility
of its being a rate-determining factor. However, as we did note in individual
alloy systems (nickel-molybdenum, iron-molybdenum, nickel-chromium,
etc.) and as we will show in the following, this assumption is frequently not
valid. In the preceding sections we have established that very different
conditions are always found when the alloying metal exhibits a considerably
greater heat of formation than the base metal during oxidation. This then
results in the preferred oxidation of the less-noble alloying partner and with
it a strong enrichment of this oxide in the scale. The following factors will
determine whether in the oxidation of a binary alloy preferentially one oxide
or an oxide mixed phase with approximately the samc metal ratio as the
1 A more technical report concerning technologically useful steels and superalloys for
high temperatures was published by Clauser (H. R. Clauser: l11aterials and :Methods
1954, llS). Sec also C. L. Clark: High-Temperature Alloys, Pitman Pub!. Corp., 1953.
2 Scheil, E.: Arch. Eisenhiltlenw. 6, 66 (1932/33).
3 Benedicks, C., and H. LOfquist: J. Iron Steel Inst. 115, 603 (1927).
4 Thyssen, 1\1. H.: J. Iron Steel Inst. 130, 153 (1934).
4.7. Influence of Metal Diffusion in the Alloy Phase 315

alloy phase appears in the scale, or whether two oxides are formed
simultaneously or sequentially in a heterogeneous mixture:
1. The free energy of formation of the oxide under consideration.
2. The mutual solubility of both oxides.
3. The per cent composition of the alloy.
4. The ratio of the oxidation rate constants of the pure metals.
5. The ratio of the diffusion constants within the alloy to the oxidation
constants of one of the pure metals.
One factor important of and by itself-the solubility and diffusion rate
of oxygen, or generally the attacking gas, in the alloy-which leads to the
phenomenon of "internal oxidation," shall not be considered here, but will
be treated separately in the closing section.
In order to simplify the somewhat complex phenomena mentioned in
points 1-5, we next consider the oxidation mechanism of a nickel-platinum
or nickel-gold alloy at higher temperatures, since here one has to deal only
with a scaling layer consistingofNiO. In this waypoints 2 and 4 are eliminated.
While for nickel-rich, noble-metal alloys the diffusion rate of the nickel ions
through the NiO layer is the rate-determining step, this is no longer the
case with small nickel content. Here, as will be shown in the following,
the diffusion of nickel atoms in the alloy phase out of the interior to the
alloy/oxide phase boundary is rate-determining. The problems to betreated
here therefore are based on the concurrence of the diffusion processes in the
alloy and in the oxide phase. For the following consideration thermodynamic
equilibrium at the phase boundaries and diffusion processes as the only
rate-determining processes were assumed; the discussion is based upon the
work by Wagner, "Theoretical Analysis of the Diffusion Processes Deter-
mining the Oxidation Rate of Alloys."l

4.7.1. Formation of a Smooth Alloy/Scale


Phase Boundary
While Wagner and Griinewald 2 found in the case of the oxidation of
nickel-gold alloys at 900°C that the scale consisted of a porous conglomerate
of nickel oxide and gold without a "smooth" alloy/scale phase boundary,
Kubaschewski and von Goldbeck 3 were able to demonstrate on basis of
oxidation experiments with nickel-platinum alloys between 800 and llOO°C
that a compact, adherent NiO layer forms. No parabolic rate law was observed
in the first case (nickel-gold), while in the second case such a rate was
1 Wagner, C.: J. Electrochem. Soc. 99, 369 (1952).
2 Wagner, C., and K. Griinewald: Z. physik. Ohern. B. 40, 455 (1938).
3 Kubaschewski, 0., and O. von Goldbeck: J. Inst. Metala 76, 255 (1949).
316 4. Scaling Processes with Formation of Thick Protective Layers

clearly present. Furthermore, in contrast to the nickel-platinum system,


an increase in the oxidation rate was found with increasing gold additions.
As developed in Section 4.2.1.2 the oxidation rate of pure nickelis given as
k = const [(p(a»)1/6 _ (p(t»)1/6]
o. 0,
where p~~ is the oxygen pressure of the ambient gas and Pg!
is the equi-
librium pressure at the Pt-Ni/NiO interface, which in the case of the "pure"
Ni/NiO system is equal to the dissociation pressure ?TO. of NiO coexisting
with pure nickel. As is to be seen from Fig. 165, where the concentration

NiO OJ
Ni Ni C'
2e-

Fig. 165. Schematic representation of diffusion


processes and the chemical potential of nickel in the

e- alloy and oxide phases during the oxidation of a.


nickel-platinum alloy.

changes in an oxidized nickel-platinum alloy are schematically represented,


the conditions become considerably more complicated in a nickel-noble
metal alloy since during oxidation the nickel concentration in the alloy will
be decreased at the Ni-alloy/NiO phase boundary and, therefore becomes
considerably smaller than in the interior of the alloy as long as the alloy
can be regarded from the "mathematical diffusion" standpoint as "infinitely"
thick.l Corresponding to the progressive consumption of nickel during the
oxidation, a concentration gradient of nickel and platinum (or another
noble metal) is produced which leads to a diffusion of the noble metal into
the interior of the alloy and to a diffusion of nickel to the oxidation front.
Generally we write the following for the oxidation constants or cation flows
So and S through the NiO layer forming on pure nickel and a nickel-noble metal
alloy (for a nickel alloy the exponent n in the oxygen pressure is also approxi-
mately 6):
So = const{(p(oa»l/ n - 1T~n}/L1goxide (4.93)
2 2

and
( 4.94)

In the case of the oxidation of a nickel alloy we consider the following equi-
librium condition:
2Ni a ll OY + O~).----' 2NiO (4.95)

1See also K. Hauffe: Reaktionen in und an festen l:3toffen, Springer, Berlin/Gottingenj


Heidelberg, 1955, pp. 269f].
4.7. Influence of Metal Diffusion in the Alloy Phase 317

with the mass·action law

(4.96)
where K has the dimension of a pressure, and z is the valence of the metal ion;
for nickel, z = 2. If we substitute the atom fraction of the alloy XNi for the
thermodynamic activity, then the equilibrium between an alloy with the atom
fraction of nickel, x~: and the appropriate oxygen pressure at equilibrium
Pg! yields
(4.97)
and that between the nickel atom fraction, x~:, in an alloy, which is itself in
equilibrium with NiO and the oxygen partial pressure of the surrounding gas is

(4.98)
From this it follows the "equilibrium atom fraction" of nickel, xj;1, for an
oxygen partial pressure of the ambient gas is

(a) _ (:'TO, )Z/4 (4.99)


X Ni - (a)
Po,
If ex is the ratio of the scaling constants of an alloy and of pure nickel, then it
follows for a given thickness of the oxide layer and by division of (4.94) by (4.93),
and elimination of Pg; and p<';: by use of the expressions (4.97) and (4.98)

S 1 - (xi://x~D4/'n
ex = -So = ----'---';---cc-~-
1 - (xi:i)4/,n
(4.100)

With reference to the diffusion in the alloy phase, it is assumed in a first approxi-
mation that the diffusion coefficient D is independent of the composition of the
alloy, so that we are able to use Fick's second law:

where XNi is the atom fraction of nickel at the distance g from the surface of
the alloy. The initial conditions reads
XNi = XNI(alloy) for t = 0 and g> 0 (4.101)
where XNj(alloy) is the atom fraction of nickel far in the interior of the alloy.
If 0 is the concentration of the metal in gram-atoms for a unit volume of the
alloy, then 0(1 - x~:)dLlgmetal is the amount of platinum in gram-atoms in the
alloy per volume element which diffuses per unit cross section and thickness
dLl gmetal into the interior of the alloy, which can be formulated by Fick's first
law as follows:

(4.102)
318 4. Scaling Processes with Fonnation of Thick Protective Layers

From the Tammann tarnishing law and the relationship (4.100) it follows for
the thickness decrease in the alloy:

1_ L1~metall = Ct~ (4.103)


dt L1~metal
Since frequently k and XNl at a given oxygen partial pressure and D from diffusion
measurements can be sufficiently exactly determined and the dissociation
pressure "ITo, of NiO is known, a calculation of the other quantities, especially
of x~l and ex can be carried out according to (4.97) and (4.100). By use of the
dimensionless characteristic quantity y = Djko, Wagner finally obtained the
following relationship

XNi(allOy) - XN1(I) = F{(-1 ~)1/2} (4.104)


1 - x~~ 2 y
where the value of the subsidiary function F(u) with u = (!exjy)1/2 can be read
off Fig. 166.

!--

0.0 V V
/
/
/
/

o
( Fig. 166. Graphical representation of the
function F(u) = 7T1 / 6u(1 - CPu) exp u 2 as
0.0 as til t.3 a function of u = (ICt/y)l/2, according to
u- Wagner.

By use of equations (4.lO0) and (4.lO4) Wagner was able to calculate


ex as a function of the nickel atom fraction of the nickel-platinum system,
where the dissociation pressure "ITO, of Fricke and Weitbrechtl and the
diffusion coefficients from measurements by Kubaschewski and Ebert 2
were used. The result of the calculations is presented in Fig. 167. While
the points obtained by Kubaschewski and von Goldbeck in experiments at
850°0 fit the theoretical curve calculated by Wagner, at 118000 the points
are rather scattered.
At low temperatures x<:1 ~ 1, so that in (4.100) the expressions
(x~l!x~D4/zn and (x<:I)4/Zn are to be neglected except if x~~ is also very small,
about equal to 0.01. 8 therefore becomes 80, ex = 1, and thus the diffusion
in the oxide, or the oxidation rate, is virtually the sole rate-determining
factor for scaling of the alloy-and the composition does not affect the

1 Fricke, R., and G. Weitbrecht: z. Elektrochtm. 48,87 (1942).


2 Kubaschewski, 0., and H. Ebert: Z. Elektrochem. 50, 138 (1944).
4.7. Influence of Metal Diffusion in the Alloy Phase 319

oxidation rate as long as x~: > 0.01. For x~l < 0.01 we have corresponding
to (4.104)

as well as rx < 1, and thus the diffusion of nickel in the alloy in the direction
of the alloy/oxide phase boundary coupled with the diffusion of the platinum
in the reverse direction is virtually rate-determining for the total reaction.
o

~
~ -1.5 f----f-tf---+---+---+-----!

Fig. 167. Ratio of the oxidation rates of


nickel-platinum alloys to pure nickel as a
function of nickel content XNi of the alloy.
x and 0: points measured at 850 and 1l00°C,
according to Kubaschewski and von Goldbeck. 1.0
The curves were calculated from equations
X N; -
(4.100) and (4.104) by Wagner.

Thomas 1 has shown that with increasing nickel content there is a


transition from rx > 1 to rx ~ 1 in oxidation of nickel-platinum alloys which
is not observable during oxidation of copper-platinum, copper-gold, and
copper-palladium alloys. The oxidation rate of these alloys was determined at
850, 925, and 1000°C in oxygen at 1 atm. If one uses low oxygen pressures-
up to 100 mm Hg-at 1000°C, the oxide layer consists of CU20, which at
higher oxygen pressures is covered with a CuO layer (see Section 4.5). Since
the decomposition pressure of Cu 20 is considerably greater than that of NiO,
x~~ in (4.99) must also be greater (under the indicated conditions, about
0.045), where x~l is replaced by x~~. Accordingly curves calculated according
to (4.100) and (4.104) (rx VS. XCu) show no sharp bend as in the oxidation
curves for the nickel alloys, but the oxidation rate decreases continually
with decreasing copper content. In Fig. 168 two series of measurements are
plotted. The simultaneously present "internal oxidation" observed by
Thomas, which we shall go into later, complicates the mechanism and the
descriptive equations, but the complications are purely practical, and
present no further theoretical difficulties. The interaction of external and
internal oxidation was studied, especially in the copper-palladium alloys.
1 Thomas, D. E.: J. Metals 3,926 (1951).
320 4. Scaling Processes with Formation of Thick Protective Layers

In the temperature region investigated internal oxidation in these alloys


proceeded much more slowly than the external oxidation with palladium
contents of less than 5 at. %, but with increasing palladium content the

~/.
!
!, ,
I

I
I
I /
V
-1
I

/
I
I
I

/:
2

b
t
Ol
.3
/

II
3

I,
° 1000°C

11
I< 850°C
x I
*,WO'C- 2.15'10'7 cm2fsec
Ik850'C -2.B7·1O-8 cm2fsec Fig. 168. Ratio of the rate of oxidation
for fure eu I ex of copper-palladium alloys to pure
I copper as a function of the copper content
'd az al< aB as eu XCu of the alloy in oxygen at 1 atm at
850 and 1000°C, according to Thomas.

relative rate of the internal oxidation continuously increased. The quantita-


tive values of the external and internal oxidation rates can be taken from
Figs 169a and b, where the temperature dependence of the two rates is
plotted for different palladium-copper alloys between 850 and 1000°C.
As can be seen from the figures, e.g., for a palladium-copper alloy with
50% Pd, the rate of internal oxidation at 1000°C and 1 atm oxygen is about
two orders of magnitude greater than that of the external oxidation.
Earlier we considered noble-metal alloys whose less noble alloy com-
ponent, e.g., copper or nickel, forms a p-type oxide with cation vacancies
during oxidation. In analogy with the above, the oxidation of similar alloys,
whose less noble metal, e.g., zinc, cadmium, or titanium, forms an n-type
oxide with metal ions in interstitial lattice positions or with oxygen-ion
vacancies, presents the following picture:
According to equation (4.66) from Section 4.2.2.1 the expression for the
cation or anion current reads:
(4.105)
As was shown in the case of zinc oxidation, (p~;)-l/n can be neglected
to a first approximation. From consideration of expressions equivalent
4.7. Influence of Metal Diffusion in the Alloy Phase 321

to·c,----.,----,-----.----.
Crrf/m

10·

~ffr---t---+---4--~

Fig. 169a. Temperature dependence of the


parabolic growth constant ka of the outer to a75 aao fO J 0.85
oxide layer for a few copper-palladium alloys T~
in oxygen at 1 atm, according to Thomas. 1000 gZS Mo de

m~I ------~-----+--~~~

Fig. 169b. Temperature dependence of the


parabolic growth constant kl of the internal
oxidation zone of a few copper-palladium alloys
in the presence of an external layer, according
to Thomas.

to (4.95)-(4.99) but adapted for a zinc-noble-metal alloy, we obtain for oc:

(4.106)

from which we obtain with x~~ = x~~ from (4.97):


tX = (x~~)4/zn (4.107)
322 4. Scalin~ Processes with Formation of Thick Protective Layers

By combination of an expression analogous to (4.lO4) with equation


(4.107) the ratio of the scaling constants may be represented as a function
of the alloy composition, that is, IX as a function of XZn, if we set 4/zn = 1/3,
which corresponds to the system Zn/ZnO (see Fig. 170). Here again one dOles
not find the knee in the curves that is characteristic of the nickel-platinum
alloys.

1.0.---,---,----,---,--~

Fig. 170. Theoretically calculated curves for


the rate of oxidation of noble metal alloys
whose less noble alloy component forms an
oxide with metal excess, according to Wagner.
o 0.2 O.'f o.(} 0.8 1.0 The ratio of the oxidation rate ex, according to
(4.106), is plotted against the atom fraction
x Zn, - - Xz n . (1) y = 10; (2) y = 0.1.

The phenomena treated here were observed not only in alloys where one
component is a noble metal, but also under certain conditions in alloys in
which both alloying partners are oxidizable. Extensive experimental data
exist on the oxidation behavior of this type alloy, e.g., copper-nickel,
copper-zinc, copper-beryllium, nickel-chromium, nickel-molybdenum, and
iron-molybdenum, noted earlier in several sections. Dunn,l Scheil and
Schultz,2 Scheil and Kiwit,3 Scheil,4 Rhines and Nelson,5 Brenner,6 Moreau
and Benard,7 Gulbransen and Hickman, Boettcher, de Brouckere, and
Hubrecht,8 and Bouillon and Stevens 9 may be cited as investigators whose
results can be used for a direct proof of the theoretical relationships derived
above. On the assumption that the oxides that are formed are practically

1 Dunn, J. S.: J. Inst. Metals 46, 25 (1931).


2 Scheil, E., and E. H. Schulz: Arch. Eisenhutlenw. 6,155 (1932/33).
3 Scheil, E., and K. Kiwit: Arch. Eisenhutlenw. 9, 405 (1935/36).
4 Scheil, E.: Z. Metallk. 29, 209 (1937).
5 Rhines, F. N., and B. J. Nelson: Trans. AIME 156,171 (1944).
6 Brenner, S. S.: J. Electrochem. Soc. 102, ':l (1955).

7 Moreau, J., and J. Benard: J. Inst. Metals 83,87 (1954/55).


8 Hickman, J. W., and E. A. Gulbransen: Trans. AIME 171, 344 (1946); 180, 519,
534 (1949); J. Phys. Chem. 52,1186 (1948); Anal. Chem-. 20,158 (1948); J. W. Hickman:
Trans. AIME 180, 547 (1948); A. Boettcher: Z. angew. Physik 2, 249 (1950); L. de
Brouckere and L. Hubrecht: Physik. Ber. 32, 63 (1953).
9 Bouillon, F., and J. Stevens: Ind. Chim. Beige 24, 1335 (1959).
4.7. Influence of Metal Diffusion in the Alloy Phase 323

insoluble in one another, but that the metal in the alloy phase form an
ideal solid solution, the equilibrium conditions are
(x1»)41Z.p~). = 7TO. for Oxide AO (4.108)
(x~»)4IZ .pg~ = 7TO. for Oxide BO (4.109)
If the diffusion processes in the alloy proceed with sufficient rapidity
to preclude lo\~al enrichment or depletion of the components A or B at the
alloy/scaling layer phase boundary, the equilibrium pressure p~). for ideal
solutions can readily be obtained as a function of the bulk composition of
the alloy from equations (4.108) and (4.109). As can be seen from Fig. 171

I Z5
r--->
<3
:3 1.0
'1'-Z
}/
c'"
~
e~~O.5
Fig. 171. Oxygen equilibrium pressures over the two .§'
solids AB alloy + AO oxide (curve 1) and AB alloy
'----'
V
o .../
+ BO oxide (curve 2) with 7TO.(DO) = I07To.(AO) and 0.8 1,0
D ~ k~o and k~o ,respectively, according to Wagner.
In region I there is exelusive formation of BO.

the oxygen pressure for A-rich alloys for the equilibrium between the alloy
and AO is lower than that for the equilibrium between alloy and BO, so
that in the A-rich alloy region BO is unstable in comparison to AO. That the
reverse is true for B-rich alloys can also be seen from the figure. Generally
the greater the difference in the dissociation pressures of the two oxides, the
smaller the stable region of the oxide with the higher dissociation pressure.
A suitable example would be the oxidation curve of copper-palladium alloys
as a function of composition. As can be seen from the experimental results
of Thomas, a definite PdO formation does not appear until above 90 at. % Pd
in the temperature region 850 to 1000°C. A second example, the oxidation
curve of the copper-platinum alloys as a function of composition, is
shown in Fig. 172.
On the other hand, if the diffusion rate in the alloy phase is not sufficiently
great, but DA ~ k10 and DB ~ k~o' then local concentration differences
appear and the oxygen equilibrium pressure at these phase-boundary surfaces
is considerably changed. These conditions are considered in Fig. 173.
We will now use the above considerations to evaluate the oxidation
mechanism of nickel-copper and copper-zinc alloys, and attempt to arrive
at the previously unexplained dependence of the oxidation rate on alloy
composition for these two series. Pilling and Bedworth1 studied the oxidation
1 Pilling, N. G., and R. E.Bedworth: J. Inst. Metals 29, 529 (1923).
324 4. Scaling Processes with Formation of Thick Protective Layers
10"$,---..,.----,-------,--.-----.---,
cm2/sec -<r Externa l Oxidat ion
- ~-Internal Oxidation

m7~--~~+_--+_--+_--+___i

Fig. 172. Oxidation curve of the copper-


,. ...... platinum alloys as a function of composition,
according to Thomas. With increasing
platinum content the portion of the internal
mtl~_~_~-~~-~-~-~
! Itl 1j i'(l i:f J(l oxidation increases to the extent that the
A/om-% Pf - rate of the external oxidation decreases.

rate of nickel-copper alloys as a function of alloy composition. They found


that at 950°C the parabolic rate law governed only in the region of 0-20 and
80-100% Ni. The experimental results are presented in Fig. 174. For the
reactions which are determinative here,
2Nialloy + o~g) ~ 2NiO, (4.110)
4CUa ll oy + o~g) ~ 2Cu20 (4.111)
the equilibrium conditions are

(x~l)2 . Po, = :7l:'O,(NiO) (4.112)


(xg~)4 . Po, = :7l:'o,(Cu,O) (4.113)
where 7To,(NiO) and 7TO,(CU20) are the corresponding oxygen equilibrium

Fig. 173. Oxygen equilibrium pressures


over the two solids AB alloy + AO and BO
for 7TO,(BO)= 107To,(AO) and DA ::::; kio
and DB ::::; k1o' respectively, according to
Wagner. Curve (1) oxygen equilibrium

1
~%O l---~~--+---+---r~t-~
pressure at the alloy/growing oxide AO
phase boundary; curve (2) oxygen pres-
sure of the virtual equilibrium between
alloy and oxide BO at the alloy/AO
~ phase boundary; curve (3) oxygen
~ equilibrium pressure at the alloy/growing
d'
oxide BO phase boundary; curve (4)
~
~
a51--~~---4---~---r--~
. oxygen pressure of the virtual equilibrium
b etween alloy and oxide AO at the
alloy /BO phase boundary: region (I)
exclusive formation of AO; region (II)
o~--~~--a~*----~
a6n---A
a~8--~ exclusive formation of BO; and region
X B --- (III) formation of a conglomerate of
f-- [ - 4------=-- m - -- - .,1 D both oxides.
4.7. Influence of Metal Diffusion in the Alloy Phase 325

pressures of NiO and CU20 with coexisting nickel and copper. As Wagner
was able to show on basis of mathematical approximations, above XNi = 0.75
we have at 950°C exclusive NiO formation and below this value we expect
formation of a heterogeneous conglomerate of NiO and CU2 0.
-1.5
000

'"
0

-8.0 - l'-..
~
1\
-3.0
1\
Fig. 174. Decrease of the metal loss of ~
nickel-copper alloys with increasing nickel
content (indicated as Ll~metal V8. XN1) after 4~ D.G 48 to
1-hr oxidation in air at 950°C, according
to Pilling and Bedworth. trNj-

As can be inferred from the experimental results of Pilling and Bedworth,


the oxidation rate in the nickel-copper alloys with less than 25 at. % Cu is
somewhat higher than that of pure nickel. Wagner tried to explain this
finding by assuming that at the beginning of the oxidation islands of CU20
form in various places on the surface, which then react with nickel according
to the following displacement reaction
(4.114)
According to the schematic presentation in Fig. 175 a continual "removal"
of the oxygen bound to the copper occurs, which is plausible on basis of
thermodynamic relationships. Owing to their greater mobility or diffusion

Fig. 175. Concurrence of the


rate of diffusion of the Cu+
and Ni 2 + ions through the
CU20 island in the scale,
according to Wagner. (a) The
CU20 "island" grows, since
it is in direct contact with the
alloy. (b) The amount of
this contact diminishes with b
time and here the contact has
a
Oxygen
been completely interrupted,
so that the CU20 "island" NiO
can no longer grow, since NiZ< 2e- Slow Diffusion
now the nickel diffusion in orNi
NiO controls the total oxida-
tion process. Ni-Cu-Alloy
326 4. Scalin~ Processes with Formation of Thick Protective Layers

rate through the CU20 lattice compared to nickel ions, the copper ions which
are liberated from oxygen in this way reach the surface where the chemi-
sorbed oxygen ions are located quite rapidly and form "new" Cu 20. In
spite of the thermodynamic instability, the CU20 phase will continue to
grow because of its more rapid formation compared to NiO, and thus the
total progress of the oxidation is accelerated. However, in the later course
of the oxidation when the Cu 20 islands are cut off from the alloy phase, as
is indicated in Fig. 175b, the rate of oxidation is dependent only on the
diffusion rate of the nickel ions through the ever-spreading NiO layer. In
the absence of a direct "copper source," as is the case for the contact between
metal and Cu20 islands, a copper ion can only diffuse through the CU20
island to the surface, when according to the following reactions

(CU20 + 2CuD')cu,o ~ (NiO + NiD")NiOj


Ni(metal) + NiD" + 2EEl ~ Null (4.115)

0-(0-) ~ CU20 + 2CuD' + EEl

a copper ion liberated from the oxygen is released in the Cu 20 island. For
sufficiently long oxidation times the oxidation rate of a copper-nickel alloy
with less than 25 at. % Cu would then have to equal that of pure nickel,
even if the CU20 islands once formed-on the assumption of a mutual
insolubility of the oxides, which is difficult to maintain for a large NiO
excess-were preserved "forever" immediately at the surface.
The oxidation mechanism of copper-zinc alloys was discussed by Wagner
as another interesting system. Oxidation experiments at 800°C with copper-
zinc alloys with 0.1 to 10 at. % Zn yield a parabolic rate law with a scaling
constant which approximately corresponds to that of pure copper. The
scaling layer consists primarily of CU20 with small inclusions of ZnO.l
As can be seen from Fig. 176, the oxidation rate of the alloy decreases be-
tween 10 and 20 at. %, and, furthermore, shows considerable deviations
from the parabolic rate law, which evidently must be attributed to the
displacement reaction
(4.116)
since both oxides appear in practically the same proportion. In order to
investigate the particular significance of the displacement reaction (4.116)
Levin and \Vagncr 2 repeated the oxidation experiments of some copper-
zinc alloys in pure oxygen at 700°C. The displacement reaction was manifested
as a change in the parabolic rate constant, which was shown to decrease
1 Dunn, J. S.: J. Inst. 11;Jetais 46,25 (1931); F. N. Rhines and B. J. Nelson: Trans.
AIME 156, 171 (1944).
2 Levin, R. L., and J. B. Wagner: J. Electrochem. Soc. 108,954 (1961).
4.7. Influence of Metal Diffusion in the Alloy Phase 327

:\\
g-cm2
10-2 \
880°C
0

BOO°C
Fig. 176. The weight increase of
copper-zinc alloys (in g/cm 2 ) after 73SoC
5·hr oxidation in oxygen as a C>
function of zinc content, according
to Dunn. o a3
XZn - -

as the more impermeable ZnO layer was formed. These authors could
show that by interrupting the oxidation and by allowing the displacement
reaction to occur during an isothermal annealing in an inert atmosphere, the
rate of subsequent oxidation was greatly diminished, as is shown in Fig.
177. For alloys with more than 20 at. % Zn the oxidation rate is again
16

<1

:;" 12 '" ,,..,3.2.\0- 9 "...


I \<.:;.,.....; 0_-
E ~ ~~a_-
~
'" <b OOJ~O-
~ 8 ,..,. /'-
« 9 TWO HOUR ANNEAL IN ARGON
~"d

Time (minutes)

Fig. 177. Oxidation rate curve for a 90% Ou-lO% Zn alloy oxidized in pure
oxygen at 700°0, interrupted by a 2-hr anneal in argon at the same temperature,
according to Levin and Wagner.

approximately independent of the alloy composition. Furthermore, the


parabolic rate law is again valid and the scaling layer consists predominantly
of ZnO. This observation by Dunn is in agreement with Wagner's theoretical
treatment, with the resulting equation

2 ( k" )112 1 - 0.177 (3 XZn(min) (4.117)


XZn(min) = 16zzn c D · 1.128

where k" is the parabolic scaling constant, f3 = Ii In Djdxzn, and XZn(min) is


the critical zinc concentration in brass-above this concentration a hetero-
geneous mixture of two oxides, Cu 20 and ZnO, appears in the scaling layer.
The values for XZn(min) are summarized in Table 38 for three temperatures.
328 4. Scalin~ Processes with Formation of Thick Protective Layers

Table 38. Summary of the Calculated Values for Critical Zinc


Concentration in Brass from Oxidation and Diffusion Data,
According to Wagner

kif, D,
T,oC g2/cm4_sec cm 2/sec f3 XZn (min)

725 0_35 X 10-10 0.8 X 10-10 7.4 0.14


800 1.4 X 10-10 3.0 X 10-10 6.5 0.15
880 6.4 X 10-10 1.3 X 10- 9 5.5 0.16

The oxidation experiments published earlier by Froehlich1 on copper


alloys with noble and non-noble alloying partners can be explained by this
mechanism. In the case of a copper-aluminum alloy with more than
3 at. % AI, the appearance of only an Ah03 layer, which is responsible for
the great reduction in the oxidation rate, is understandable on basis of the
high heat of formation of A1 203. This is equally true for the copper-beryllium
alloys, which were investigated by.Bro.uckere and Hubrecht 2 and Maak. 3 The
experimental results of Honj o4 on the selective oxidation of copper alloys
with 7 wt.% Mn and 7% Ni as well as iron alloys with 13 wt.% Al may be
explained in the same way. The formation of the oxide layer from the oxide
with the higher heat of formation was dependent upon the oxidation con-
ditions. The higher the temperature and the lower the oxygen pressure,
the more decided the selective oxidizing out ofthe less noble alloy component.
The results of the copper-nickel oxidation experiments carried out by Sartell5
and co-workers will be discussed later.
Without doubt the theoretical considerations developed by Wagner
in the oxidation mechanism of binary alloys are of the greatest significance
for the explanation of the oxidation mechanism of commercial alloys, as, for
example, the nickel-chromium, nickel-chromium-molybdenum, and nickel-
chromium-aluminum steels. Frequently in these steels a good protective
scale is observed if the scaling layer is not a mixed oxide phase, but rather a
spinel phase or an individual oxide, e.g., Cr203, A1 203, etc., where an ionic
diffusion can proceed only extraordinarily slowly. Using expressions similar
to those formulated by Wagner, one can calculate the least concentration
of valuable-alloy metals which is required in order to obtain the desired
protective oxide layer in a certain temperature region. From this point of
view the electron diffraction procedures of Gulbransen and co-workers which
were cited above in the explanation of the structure of the scaling layer in
1 Froehlich, K. W.: Z. Metall1c. 28, 368 (1936).
2 Brouckere, L. de, and L. Hubrccht: Physik. Ber. 32, 63 (1953).
3 J\Iaak, F.: Z. Metallk. 52, 538, 545 (1961).
4 Honjo, G.: J. Phys. Soc. Japan 8,113 (1953).
5 Sartell, J. A., S. Bendel, T. L. Johnston, and C. H. Li: Trans. ASM 50, 1047 (1958).
4.7. Influence of Metal Diffusion in the Alloy Phase 329

oxidized alloy samples are of great value. All these works have immediate
technological consequences since a theoretical approach not only explains
for the first time the so-called "phenomena"-e.g., the green-red tarnishing
of nickel-chromium base heat conductorsl-but also indicates reasonable
countermeasures without the costly necessity of a series of empirical investi-
gations.

4.7.2. Formation of a Rugged Alloy/Scale


Phase Boundary
A further possibility in the oxidation of a noble-metal alloy is the forma-
tion of a two-phase scaling layer, where the oxide is embedded in a matrix
of the noble alloying partner. Wagner 2 was able to work out the conditions
under which a uniform oxide layer is stable and where a two-phase scaling
layer with a rugged alloy/scale phase boundary appears. Under simplifying
assumptions (for example, the alloy/scale phase boundary exhibits a wavy
profile which may be approximated by a sine function) a characteristic
quanity q is developed
xl D'/V'
q= I - xl D;./V"

whose numerical value is of decisive significance for the appearance of a


plane phase boundary. In the above formula x~ denotes the mole fraction
of the alloy partner A in the oxide at the intermediate phase boundary,
D is the interdiffusion coefficient of the alloy,3 D1 is the self-diffusion co-
efficient of A in the oxide, and V' and V" are the molar volumes of the alloy
and the oxide, respectively (to a first approximation assumed independent
of the composition).
Wagner found that with q > 1, the plane phase boundary is stable and
with q < 1 an irregular phase boundary is possible. However, the latter
condition (q < 1) is probably a necessary, but in no way sufficient, condition
for a rugged phase boundary. The quantity q in the above equation is,
by definition, smaller than 1 when the concentration of A in the interior
of the alloy and at the phase boundary is small.
Recently Lichter and Wagner4 extended the theoretical treatment with
respect to the formation of a rugged scale and have tested their formula
by sulfidation experiments on copper-gold, silver-gold, silver-copper, and
1 Spooner, N., J. M. Thomas, and L. Thomassen: J. Metals 5,844 (1953).
2 Wagner, C.: J. Electrochem. Soc. 103, 627 (1956).
3 Interdiffusion coefficients are understood to be the average diffusion coefficients of
the metal atoms A and B diffusing toward one another in an alloy AB.
4 Lichter, B. D., and C. Wagner: J. Electrochem. Soc. 107, 168 (1960).
330 4. Scalin~ Processes with Formation of Thick Protective Layers

nickel-copper alloys. A sulfur attack on low.gold-content copper-gold and


silver-gold alloys yields a composite scale consisting of an outer homogeneous
sulfide layer and an inner two-phase layer involving gold-rich alloy and
sulfide. This is in accordance with theoretical considerations by which a
plane alloy/sulfide interface is not stable because the interdiffusion coefficients
D' in the alloy are much smaller than those of Ag,D!g' through the sulfide
layer. The tendency to form a rugged alloy/sulfide interface decreases with
higher gold contents, especially at lower temperatures. In this case the sulfida-
tion rate decreases accordingly. Finally, gold-rich alloys may yield a nearly
planar interface with negligible irregularities. In spite of the large difference
in the free energies of formation of Ag 2 S and Cu 2 S, the sulfidation of silver-
copper alloys delivers a sulfide layer of uniform thickness with a component
ratio the same as that of the alloy. The oxidation constant is about as high
as that of pure silver or copper. The sulfidation mechanism of nickel-copper
alloys is rather complex: two sulfide layers are formed. The outer layer is
supposedly digenite, CU1.8S, and the inner layer a solid solution with nickel
sulfide as the solvent and copper sulfide as the solute.
However, other conditions can also appear as shown in the oxidation
experiments in copper-gold alloys by Raub and EngeP Such an alloy with
0.14 at. % Cu gave a CuO film of uniform thickness, while alloys with higher
copper contents (0.50, 0.60, and 0.87 at. %) gave oxide films with varying
local thicknesses. In these alloys with q < 1, planar oxide films of uniform
thickness appear. By setting V' equal to V" assuming plastic flow in only
the oxide layer to explain the movement of the oxygen, Wagner was able to
describc the mechanisms of all these phenomena.
In a further work Wagner 2 refined the considerations sketched in
Section 4.7.1 on the oxidation of metal alloys, where in the main both alloy
partners react to form oxides ,vith different free energies of formation.
Figure 178 illustrates the following: oxide AO grows more rapidly than oxide
BO, until finally the latter becomes buried and the outer scale is found to

Oxygen

Fig. 178. Schematic cross section


of a scaling layer of an alloy AB
with an outer oxide layer AO and
an inner layer BO, according to
\Vagner. The oxide BO can grow
at the AO/BO phase boundary
through the exchange reaction
B2+ + 2e- + AO = A2+ + 2e- + BO
although it is not in contact with the
oxidizing atmosphere.

1 Raub, E., and M. Engel: Fortriige der Hauptversammlung d. Deutsch. Ges. Aletallkunde
(1938), p. 83, Berlin, VDI·Verlag (1938).
2 'Wagner, C.: J. Electrochem. Soc. 103, 627 (1956).
4.7. Influence of Metal Diffusion in the Alloy Phase 331

consist entirely of AO. This arrangement, observed for example by Rhines


and Nelson l in copper-zinc alloys, is stable only when oxide BO (for example,
ZnO) exhibits the higher free energy of formation. In that case BO can
continue to grow even when it is not in direct contact with oxygen. However,
when BO has a lower free energy of formation than AO, then, in the course
of the oxidation,when it is covered by AO,it will disappear. Wagner has, now
formulated the mathematical analysis for the case in which BO is stable.

4.7.3. The Mechanism of the Formation of an


Oxide Layer Consisting of Two Oxides
The first quantitative description of this oxidation mechanism was
given by Wagner in the work noted earlier. In this it was found exp~dient
to introduce certain simplifications to make an evaluation of the equations
possible, at least for certain limiting cases. As the first simplification, it was
assumed that the molar volumes of the oxides AO and BO are practically
equal (VAO = YEO) and that plastic flow in both phases can be neglected.
It was further assumed that each volume element of the oxides and of the
alloy contained the same number of metal atoms (mathematically V =
VAO = YEO), where V is the volume of the alloy). On basis of the scheme
given in Fig. 17S it was further assumed that the size of the BO particles
and the breadth of the AO channels between BO particles is small compared
to the total thickness of the scaling layer. Finally the diffusion was con-
sidered to be perpendicular to the surface (one-dimensional problem).
In the following the activity of the atomic oxygen, a, is calculated as a
characteristic variable, which should be equal to 1 at the outer surface,
g = o. Furthermore, we use 6 and 6 for the location of the BOjAO and
AO/alloy phase boundaries. The self-diffusion coefficients D! and D~ of
the metal ions in the oxides AO and BO, which are dependent on the defect
concentration or on the activity of the oxygen, can be measured and then
used in the equations which were formulated. If we designate the volume
fraction of the oxide BO in the scaling layer at a distance g from the surface
as 0/ and set D! = D~ • a IX or D; = D~.aP, where at and f3 are constants
(positive for p-type oxides and negative for n-type oxides), then the following
expressions are obtained for the transport rates in the g-direction:

jA = (1 - 'II') (D~/V)aa(alna/a~) + u(1 - "P)/V (4.11Sa)


jB = "P(D~/V)af3(alna/a~) + u"P/ V (4.11Sb)
U = dg 2 /dt is the drift velocity of the oxide in the direction toward the alloy
due to its decrease in mass.
1 Rhines, F. N., and B. J. Nelson: Trans. A/ME 156,171 (1944).
332 4. Scaling Processes with Formation of Thick Protective Layers

On basis of the simplifications introduced above

iA + iB = 0 (4.119)

From (4.118) and (4.119), we obtain

u = d~2/dt = -[(1 - tp)D1. al% + tpD~a/i](olna/o~) (4.120)

Since no further limiting assumptions were made, equations (4.118) to


(4.120) are valid for the outer scaling layer, where ifi = 0 and only the oxide
AO appears, as well as for the inner region of the scaling layer where ifi > 0
and both oxides AO and BO are present.
After a few intermediate calculations Wagner obtained for the rate of
the change of the concentration of B in the scaling layer

01jJ __ d';2 0 {1jJ(1-1jJ)[D~al%-D~a/i]\


(4.121)
fit - -a;t fiT (1 -1jJ) D~al% +
1jJD~a/i J

In the calculation of the constants zl = 6/2(D1t)1/2 and Z2 = t2/2(D1t)1/2


the following two equations were derived:
xB = x~ + (xB(2) - x'1) f/> (zl/rl/2)/f/> (z2/rl/2)
qat (4.122)

Here XB is the local mole fraction of B in the alloy (0 characterizes the


initial state), XB(2) = XB at t = t2, f/> characterizes the error function,
y = [3 - (1., q = D~/D1, and r = D/D~.
Using the latter two equations, we can express the oxygen activity a
and the volume fraction ifi in the two.phase region of the scaling layer as a
function of z, when the values of the parameters (1., [3, q, r, a2, XB(2), and x~
are given.
In spite of the relatively large number of parameters, a few general
conclusions may be drawn. Under the often present condition that the free
energy of formation of the oxides LJFAO and LJF BO is very much greater
than RT and LJFAO < LJF BO , and that a2 for the coexistence of the alloy/AO/
BO phases is very small compared to a2 at the phase boundary, the following
can be noted:
When r ~ 1, that is, D ~ D~, the concentration differences in the alloy
are very small, so that the composition of the alloy in the interior is approx-
imately equal to that at the alloy/scaling layer phase boundary. Further-
more, only the oxide AO or BO is formed when x~ < XB(2) or x~ > XB(2).
On the other hand, when r ~ 1 the scaling layer contains two oxides
in a certain alloy region. Under this assumption and with the hypothesis of a
metal deficit of the two oxides (for example, FeO and NiO) with (1. = [3,
4.7. Influence of Metal Diffusion in the Alloy Phase 333

that is, y = 0, Wagner is able to give a quantitative description of the oxida-


tion mechanism.
As is to be inferred from the earlier sections describing the experimental
results, in agreement with Wagner's theories, the oxide of the base metal is
often found in the outer part of the scaling layer, while the oxide of the alloy
itself is found in the interior of the scaling layer. The conditions become still
more complicated when, besides the two oxides, there is additional spinel
formation, as, for example, in the oxidation of iron-nickel alloys.1
An additional complication must be taken into consideration, if the
oxidized alloy delivers a volatile oxide or if the growing oxide layer contains
pores. As an example of the first phenomenon, we can discuss the oxidation
of the intermetallic compound, InSb, investigated by Rosenberg and Lavine. 2
The oxidation of single crystals of InSb was studied in the temperature range
between 212 and 494°C at an oxygen pressure of 0.3-0.4 mm Hg. Initially
antimony is preferentially attacked and evaporates from the surface as
(Sb20ak In20a is simultaneously formed and produces a compact layer.
As one can see from Fig. 179, it is not possible to classify the oxidation as a
parabolic, cubic, or higher-order reaction. Both a doping with tellurium
(n-type) and with cadmium (p-type) causes an increase in the oxidation
rate. Elemental antimony, which is accumulated at the In20a/lnSb interface,
10 19

1016 ~------4-----~~~-----+------~~

10 15 L-______.....L..______--L_ _ _ _" -_ _ _ -L..~

1 10 100 1000 10000


->- Minutes
Fig. 179. Oxidation of [110] InSb surfaces at po, = 0.3 to
0.4 mm Hg, according to Rosenberg and Lavine.
1 Foley, R. T., J. :U. Druck, and R. E. Fryell: J. Electrochem. Soc. 102, 440 (1955).
2 Rosenberg, A. J., and 1\1. C. Lavine: J. Phys. Chern. 64,1135,1143 (1960).
334 4. Scaling Processes with Formation of Thick Protective Layers

can partly diffuse through the In203layer and vaporize as (Sb 203)2, especially
if oxidation is interrupted. The ratio of the rate constants for the oxidation
of InSb and of indium is quantitatively predicted on the premise that both
reactions are controlled by the diffusion of indium ions via interstitial
positions in In 2 03.
For the purpose of testing of the following relation for the BO formation

x~ = ~(7Tk"(XB > X~) )1/2


ZBAo D

which can be derived from the equations of Section 4.7.1, where x~ is the
bulk concentration of metal B in the alloy, Ao is the atomic weight of
oxygen, and 7T is the dissociation pressure of BO, Maak and Wagner1 oxidized
copper-beryllium alloys with XB = XBe = 0.008, 0.016, 0.027, and 0.067 at
850°C in air of 1 atm. All these alloys oxidize approximately with the same
rate of lO-8 g2jcm4 -sec. Only at a much larger beryllium addition,
XBe = 0.13, could the desirable BeO protective layer with kif of about
lO-12 g2jcm 4 -sec be observed. According to the above equation, the minimum
content of beryllium, x~ = XBe, amounts 0.018, e.g., 1.8 at. %, with
V = 7.1 cm3jg-atom copper, ZBe = 2, D = lO-9 cm 2jsec, and kif = 2 X lO-12.
This large deviation between the calculated and the experimentally ascer-
tained value, can be caused by the following mechanism.
From photomicrographs of oxidized copper-beryllium alloys, one can
observe the existence of pores, in addition to a Cu 20 and BeO formation with
enrichment of the latter nearest the metal. Accordingly, nearest the metal
phase a three-phase zone with CU20, BeO, and oxygen gas in pores is present.
Obviously, this structure appears because there is an internal oxidation of
beryllium in the alloy as well as the Cu 20 formation. At this zone, compara-
tively long BeO needles are formed representing a kind of scaffold in the three-
phase zone (Fig. 180). According to this reaction scheme, the formation
of BeO takes place preferentially within the alloy and a significant displace-
ment reaction can be excluded. In addition to a migration of oxygen molecules
within the pores of the three-phase zone, one might also conceive of a trans-
port of oxygen as CuO. At present, however, no pertinent information is
available.
The same mechanism, seems to rule in some other systems. In agreement
with Rhines and Nelson,2 Mrowec and co-workers3 have found during the
oxidation of copper-zinc alloys at 900°C, with XZn = 0.09 and 0.15, an

1 Maak, F., and C. Wagner: Werkstoffe u. Korrosion 12, 273 (1961); F. Maak: Z. Metallk.
52, 538 (1961).
2 Rhines, F. N., and B. J. Nelson: Trans. AIME 156,171 (1941).
3 Czerski, L., S. Mrowec, and T. Werber: Arch. Hutnictwa 3,37, 113 (1958).
4.8. Mechanism of Internal Oxidation of Alloys 335

exterior compact Cu 20 layer and a zone of Cu 20 + ZnO + pores beneath,


where oxygen diffuses preferentially. The same mechanism was observed
for the oxidation of copper-nickel alloys by Sartell and co-workers,l who
were able to confirm the oxygen migration by marker experiments.
Internal Oxidation Scaling layer
~'~---B-e-O--~A--------~

' -0
,- . . - 02 (Pore gas) .
02 (gas)

~
CuBe
...-
Cu ~ O 2 (gas)
\
,
CuO

h7TTTTn.~n7"':
I'-'-..............+-'".............~ I
I
Fig. 180. Scheme of the mass I
transport in a three-phase zone
CU20 + BeO + gas in the oxida- Formation '/, O 2 (gas)
t
CU20 = 2Cu+ Formation
tion of copper-beryllium alloys,
according to ::\Iaak and "'agner.
ofBeO = 0 (in Cu) + 2e- + '/2 O 2 OfCU20
(gas)

Such a pore mechanism in the absence of a three-phase zone has also been
identified for the attack by sulfur on pure silver, where sulfur molecules
are present in the pores from the dissociation of Ag 2S.2

4.8. The 'Iechanism of Internal Oxidation of Alloys


As was shown in detail in the preceding sections, if a compact oxide
protective layer is formed, the rate of oxidation of many metals and aUoys
at high temperatures (500 to 1000°C, depending on the mobilities of the ion
lattice defects in which the oxide crystals make up the protective layer)
can be described by the parabolic Tammann-Pilling-Bedworth rate law.
In all these cases the scaling layer grows in proportion to the square root
of time. Furthermore, within a certain oxidation period and a definite
temperature region the logarithm of the squared value of the scale thickness
is proportional to the reciprocal value of the absolute oxidation tempera-
ture. These relationships are valid as long as the diffusion of metal ions and
electrons or oxygen ions can be regarded as the determining process solely
on basis of a chemical potential gradient of the species capable of migration.
1Sartell, J. A.., S. Bendel, T. L. Johnston, and C. H. Li: Trans. ASM 50,\1047 (1958).
2::\Irowec, S., and T. "'erber: Acta .lIet. 7, 696 (1959); H.Rickert: Z.physik. Chern.
[NFl 23, 355 (1960).
336 4. Scaling Processes with Fonnation of Thick Protective Layers

In contrast to the oxidation of a pure metal, where the mechanism of the


appearance of only one or even several distinct oxide layers can be easily
described, the mechanism for the oxidation of alloys is considerably more
complex. Two frequently encountered mechanisms have been discussed in
the preceding sections:
1. We have become acquainted with alloy oxidation processes where an
oxide solid solution is preferentially formed, which can be described
in terms of Wagner's scaling theory and the theory of lattice defects
for heterotype mixed phases. For small alloying additions, this is
frequently the case.
2. We have also noted many alloys (noble-metal alloys fund alloys whose
components exhibit widely varying free energies of formation of the
oxide) for which, owing to differences in the thermodynamic condi-
tions of formation of the individual oxides, the structure and growth
of the scaling layer are determined by the competition between the
diffusion processes in the alloy and in the scaling layer. In the
preceding chapter we showed that, at least in principle, this mech-
anism can be explained.
However, yet another phenomenon appears which makes the oxidation
lllechanism still more complex and the explanation even more difficult:
oxygen solubility in different metals, e.g., silver, copper, nickel, titanium,
zirconium, etc., and the oxidation of the less noble alloy partner-even when
it appears in a smaller concentration-in the alloy phase. Rhines l was
first to appreciate the full significance of this phenomenon and he offered
the following explanation, based on detailed investigations. Oxidation of,
e.g., copper alloys produces not only a scaling layer on the alloy, which
consists of Cu 20 or of CU20 with an exterior thin CuO layer, depending on
the oxygen partial pressures, called the "external scale," but also an oxida-
tion zone in the interior of the alloy, which we call the subscale. This internal
oxidation zone is due primarily to the solubility of oxygen in the alloy phase
and the higher free energies of formation of the oxides of the less noble
metals in the alloy. Their continual consumption of oxygen, which causes
a high oxygen chemical potential gradient (which, to a good approximation,
is frequently identical with its concentration gradient) between the alloy/
external scaling layer and alloy/internal oxidation zone phase boundaries
effects a continuous replenishing of the supply of oxygen, which is made
available through the following dissociation:

CU20(external oxide layer) ~ 2Cu + O(diSSOl ved in alloy)


1Rhines, F. N.: Trans. AIME 137, 246 (1940); J. Corrosion 4,15 (1947); F. N. Rhines,
W. A. Johnson, and W. A. Anderson: Trans. AIME, Tech. Pub!. No. 1368 (1941).
4.8. Mechanism of Internal Oxidation of Alloys 337

Meijering and Druyvesteynl included internal oxidation in their explanation


of the hardening of ductile metals or metal alloys. Since the work of Rhines
et al. was not known to the former, certain problems, e.g., the penetration
of oxygen into weakly alloyed metals, were also investigated. However,
this in no way detracts from the value of the previous work since it arose
from a different approach to the problem, and the thermodynamic considera-
tions form a useful complement to the kinetic representation.
By new experiments, Meijering 2 was able to show that the change in
hardness is mainly determined by the change in velocity of the oxidation
boundary, which affects the dispersion of the oxide. Informative measure-
ments on the growth of the internal oxidation zone in copper-palladium
alloys were carried out by Thomas,3 who also used high alloying additions
for the first time (see Figs 169a and b). Here the less noble alloying partner
is copper itself, which was oxidized to CU20 in the alloy, as was noted earlier.
As a prelude to our discussion, it may be noted that if the rate of diffusion
of the oxygen to the alloy/outer scale phase boundary becomes smaller than
that in the alloying metal, the degree of internal oxidation should decrease
with decreasing temperature and ultimately become negligible. Especially
precise investigations into the temperature dependence of the internal and
external oxidation of copper-palladium and copper-platinum alloys were
published by Thomas. It would be desirable to carry out these kinds of
measurements on other alloying systems (e.g., copper and nickel alloys),
since the appearance of an internal oxidation which is significant for the
mechanical properties of all alloys could then be predicted with certainty,
which in turn would be very useful when dealing with the application of
construction materials. Thus, a knowledge of the mechanism of the internal
oxidation and of the factors that tend to suppress it is indispensible for an
evaluation of the technological usefulness of this type of alloy as a high-
temperature industrial material.
The first quantitative investigations on this type of oxidation were
undertaken by Rhines and co-workers, who at first investigated the internal
oxidation of both alloys with a copper base and small impurity metal con-
centrations in the IX-solid solution region at different temperatures in air and
in oxygen, and samples embedded in CU20 in the presence of inert gases
or in vacuo. The latter experimental arrangement was chosen in order to
study internal oxidation in the absence of an external scaling layer. In Fig.
181, the photomicrographs of a pure copper sample and a copper-silicon alloy
with 0.5 wt. % Si after a 2-hr oxidation in air at lOOO°C are reproduced.
1 Meijering, J. L., and M. J. Druyvesteyn: Philip8 Res. Rep. 2, 81, 260 (1947); see also
J. W. Martin and G. C. Smith: J. Inst. Metal8 83, 153 (1954/55).
2 Meijering, J. L.: Tran8. AIME 218, 968 (1960).
3 Thomas, D. E.: J. Metal8 3,926 (1951).
338 4. Scaling Processes with Formation of Thick Protective Layers

a b
Fig. 181. Photomicrograph of a copper and a copper-silicon alloy sample heated
for 2 hr in air at lOOO°C, according to Rhines (enlarged 150 times). (a) CU20
layer on pure copper. (b) Outer and inner scaling layers of a copper-silicon alloy.

Although the two samples were oxidized under the same experimental con-
ditions, we recognize on the scaled alloy, in contrast to pure copper, two
oxidation zones which are clearly distinguishable from one another. The
external (darker) zone, which consists of a quantity of Cu 20 and Si02, has
practically the same layer thickness as the scaling layer of the pure copper.
The adjacent internal oxidation zone contains Si0 2 crystals which are embed-
ded in pure copper, and no copper oxide could be observed. Subscale, which
was coined by Rhines, is the same as zone of internal oxidation. If we choose
a sufficiently low oxygen partial pressure, po, :s; 7TO. (over Cu20/CU in
equilibrium), then only an internal oxidation zone forms. As can be seen
from Fig. 182, under these special experimental conditions, the surface of
the copper-silicon alloy remains completely free of any oxide layer, apart
from the silicon atoms occupying the surface, which are oxidized to Si02,
while the internal oxidation proceeds to a considerable depth in the alloy.
These experimental conditions permit a separation of the two processes, the
external and the internal oxidation, and thus a study of the individual types
of oxidation. On basis of this possibility, Rhines and co-workers could
quantitatively formulate the partial processes that were determinative for
4.8. Mechanism of Internal Oxidation of Alloys 339

Fig. 182. Internal oxidation of a copper alloy with 0.103% Si, according
to Rhines (enlarged 50 times). Sample was oxidized at an oxygen pressure
which was so low that no exterior layer was observed. The internal oxida·
tion zone is recognizable by a dark.gray, sharply bounded shading.

external and internal oxidation by compiling a greater quantity of experi-


mental data.
For this type of experiment the base metal must be of high purity.
For example, the copper used in the copper alloys for the oxidation experi-
ments contained less than 0.05 wt. % total impurity. In the same way the
alloying partner consisted of metals as pure as could be obtained. The alloys
were melted in a graphite crucible under a pure borax melt and were free
of oxide inclusions. All alloys to be oxidized were either heated in air or
embedded in a mixture of equal parts of Cu 20 and copper powder
in a copper-coated iron tube. In the last experimental arrangement the
oxygen partial pressure never exceeded the oxygen equilibrium pressure
of the reaction, CU20 ~ 2Cu + i02. In general only two experimental
temperatures, 600 and 1000°C, were chosen. Corresponding to earlier investi-
gations on the solubility of oxygen in copper,l the oxygen first penetrates
the copper alloy, without causing an oxide formation. The internal oxidation
shown in Fig. 182 does not occur until the oxygen concentration required
for the impurity oxide formation is attained. In order to determine the
progress of the internal oxidation zone with time, small pieces were removed
from the experimental samples and polished and etched, and finally the
1 Rhines, F. N., and C. H. Mathewson: Trans . AIME 111, 342 (1934).
Table 39. Oxidation Rates in the External and Internal Oxidation Zones in Copper Alloys, According to Rhines, Johnson,
and Anderson c.>
....
0

Constants of Values for ....


Constants of Constants of [I (t.ti)2C (dt;j2 External oxidation zone
d] In
[ t2 ~2Cl[ a') og--- = - + n
10 - = - 2 t T Oxida·
a+ b][ 10 - - " = - + b' 2-
gt T g t T Values for [ for an internal . for an tion S·
for internal for internal t oxidation zone mternal tem- iJO.
oxidation oxidation for internal forming under the oxidation pera- 'tI
Dis- alone alone oxidation external one zone under an Values for >;
alone at ture 0
solved Concentra- external one (Llte)2 n
metal tion, \vt. ~~ a b a' b' 600°C c d at 600°C -- °c ~
Ul
Ul
~
Ul
Al 0.03 -12,110 4.110 -10,350 1.035 -11,890 3.614 2.02 x 10-10 4.25 X 10- 9 750
Al 0.06 -11,580 3.261 -10,350 1.035 2.45 x 10- 12 -11,680 3.185 1.05 x 10- 10 ~.
....
AI 0.08 -11,540 3.142 -10,350 1.035 -11,710 3.053 1.46 x 10-10 ::r
-10,350 1.035 6.52 x 10-11 -12,020 2.912 8.51 x 10-11 { 2.54 X 10- 9 750 "1
AI 0.17 -12,160 3.142 3.90 x 10- 8 875 0
>;
{ 6.35 X 10- 9 750 S
AI 0.45 -11,970 2.983 -10,350 1.035 -10,800 1.215 3.16 x 10-8 875 Pl
1.69 X 10- 7
....
1000 o·
Al 0.72 -12,140 2.753 -10,350 1.035 -12,140 2.015 ::s
B 0.05 -10,420 2.882 -10,400 1.582 5.76 x 10-11 -10,420 2.762 1.13 x 10- 7 1000 0
,...,
Ba 0.10 -13,480 5.228 -13,700 4.432 1.82 x 10-10 -13,480 4.248 7.10 x 10-11 ~
Be 0.018 -11,150 3.330 -10,420 0.542 2.42 x 10-10 -11,150 3.492 2.96 x 10-10 1.72 X 10- 7 1000 ::r
Be 0.054 -14,290 4.830 -10,420 0.542 6.13 x 10-11 -14,290 4.487 1.39 x 10-10 n·
7.74 x 10-11 ~
Be 0.101 -11,980 2.652 -10,420 0.542 -11,980 2.572 2.44 x 10-11
Ca 0.01 -10,080 2.121 -11,150 0.974 5.18 x 10-10 -10,080 1.955 2.26 x 10-10 1.16 X 10- 7 1000 'tI
>;
Nb 0.04 -10,250 2.742 -10,210 1.328 8.39 x 10-11 -10,250 2.681 3.20 x 10-10 ....0
-10,870 2.984 -10,880 1.009 5.05 x 10-11 -10,870 2.984 1.28 x 10-10 1.83 X 10- 7 ~
Ce 0.01 1000 n
Co 0.14 -10,500 2.842 -11,750 2.932 7.10 x 10-13 -10,500 2.812 2.68 x 10-10 1.31 X 10- 7 1000 ....
:;;.
Cr 0.08 -11,910 3.910 -11,450 2.557 2.45 x 10-12 -11,910 3.824 1.45 x 10-12 1. 79 X 10- 7 1000 ~
Fe 0.037 -11,220 3.944 -10,260 1.594 t"'
Fe 0.10 -10,390 2.509 -10,260 1.594 4.19 x 10-12 -10,390 2.620 1.75 x 10-12 2.07 X 10- 7 1000
0.56 -10,730 2.393 -10,260 1.594 9.08 x 10-11 -10,730 2.072 1.40 x 10- 7
~~
Fe 1000 >;
Fe 1.52 -10,630 1.885 -10,260 1.594 3.68 x 10-11 -10,630 1.533 1.11 x 10- 7 1000 Ul
Fe 2.65 -10,310 1.288 -10,260 1.594 4.70 x 10-12 -10,310 0.878 1.35 x 10- 7 1000
Ga 0.03 -15,880 7.357 -15,930 5.885 -15,880 7.135 4.80 x 10-10

continued
Table 39 continued
Ge 0.02 -10,210 1.996 -11,470 1.543 -10,210 1.790 1.59 x 10- 7 1000
In 0.25 -11,300 3.376 -10,720 2.206 3.93 x 10- 11 -11,300 3.274 1.60 x 10- 10 1.78 X 10- 7 1000
Li 0.02 -10,330 2.094 -11,450 1.290 6.13 x 10- 11 -10,330 2.011 1.07 x 10- 10
Mg 0.10 -11,830 3.453 -11,130 1.773 4.24 x 10- 11 -11,830 3.362 1.05 x 10- 10 1.66 X 10- 7 1000
Mn 0.033 -10,980 3.452 -10,620 1.640 4.71 x 10-12 -10,710 2.765 2.50 x 10- 11
Mn 0.084 -10,620 1.640 -10,610 2.708 1.16 x 10-10
Mn 0.22 -10,620 1.640 -10,840 2.316 6.16 x 10-11
Mn 0.42 -10,470 1.778 -10,620 1.640 6.12 x 10- 12 -10,570 1.648 1.04 x 10- 11 4.46 X 10- 8 875
Mn 1.55 -10,500 1.452 -10,620 1.640 1.78 x 10-11 -10,570 1.186 3.27 x 10-11 "'"
00
Ni 0.115 -11,110 3.695 -12,840 2.306 1.20 x 10-10 -11,110 3.660 3.48 x 10-10 2.74 X 10- 7 1000
Ni 5.00 -13,710 4.032 -13,710 3.510 1. 78 x 10- 7 1000 s::
P 0.03 -12,050 3.835 -11,870 2.142 3.62 x 10-11 -12,050 3.703 ...,tI>
P 0.07 -11,700 3.274 -11,870 2.142 -11,700 2.955 ~
=-
P 0.24 -11,050 2.050 -11,870 2.142 2.32 x 10-11 -1l,050 1.749 e.rn
Ph 0.03 -10,610 1.115 -11,290 0.150 1.84 x 10- 8 1000
Si 0.045 -11,320 3.275 -11,110 1.660 1.48 x
10-10 -10,800 2.614 1.57 x 10-10 i3
Si 0.076 -11,110 2.790 -11,1l0 1.660 1.73 x
10-10 -10,680 2.181 3.91 x 10-10 2.21 X 10- 9 750 0
...,
Si 0.103 -11,010 2.481 -11,1l0 1.660 2.79 10-10
x -10,720 2.048 1.53 x 10-10 ...::s
Si 0.180 -10,310 1.723 -1l,110 1.660 3.38 x
10- 11 -10,730 1.914 1.19 x 10-10 ...
tI>
Si 0.30 9,725 0.919 -11,110 1.660 6.12 x 10-11 -10,050 8 58 10-11 r 2.84 x 10- 9 750 '1
0.952 . x } 3.41 X 10-8 ::s
875 ~

Si 0.59 9,345 0.407 -11,1l0 1.660 5.36 x 10- 11 8,950 -0.331 8 16 X 10-11 2.30 X 10- 9 750
. L 1.77 X 10-8 875 0
~
Si 0.858 7,785 1.205 -11,110 1.660 1.33 x 10-10 9.24 X 10- 11
-s:
Si 1.93 6,570 2.709 -11,1l0 1.660 1.27 x 10-10 7,010 2.667 7.37 x 10-11 I'>
Sn 0.31 -11,820 3.665 -13,550 4.521 3.21 x 10-11 -1l,820 3.536 7.51 x 10-11 ...o·
Sr 0.10 -11,140 3.602 -13,510 4.492 -11,140 3.444 1.69 x 10- 7 1000 ::s
Ta 0.04 -10,810 3.208 -10,820 1.843 6.50 x 10-11 -10,810 3.000 1.69 x 10-10 0
...,
Ti 0.05 -10,430 3.010 -10,220 2.849 2.25 x 10- 7 1000
V 0.09 -13,000 4.633 -13,200 3.767 2.45 x 10- 10 -13,000 4.492 7.72 x 10-11 >
:::
Zn 0.16 -10,840 3.185 -12,750 3.855 3.58 x 10-10 1.80 X 10- 7 1000 0
Zn 0.21 -12,180 4.049 -12,750 3.855 2.12 x 10-10 -12,180 3.861 5.22 x 10-11 '<
rn
Zr 0.16 -11,110 2.955 -10,430 1.511 -11,110 2.809 2.64 x 10-10 1.73 X 10- 7 1000
Al+Be 0.049+0.003 -13,600 4.534 3.30 x 10-10 3.91 X 10-11
AI+Sn 0.06 +5.43 -12,150 1.976 9.51 x 10-11 -12,150 2.132 1.76 x 10-11
AI+Zn 0.13 +9.29 -10,500 0.991 3.62 x 10-13 -10,500 0.510 2.81 x 10-11
Be+Sn 0.02 +0.30 -13,500 4.357 7.92 x 10-11 -13,500 3.703 2.67 x 10-11
Be+Sn 0.006+4.93 -12,380 2.440 7.52 x 10- 10 -12,380 2.555 5.22 x 10-12
Be+Zn 0.03 +9.52 -1l,040 1.307 5.21 x 10-13 -11,040 1.046 1.76 x 10-12 CH
Si+Sn 0.085+5.02 -12,000 2.342 1.66 x 10-10 -12,000 2.300 2.45 x 10-12 ..."'"
Si+Zn 0.085+9.81 -11,120 1.337 3.34 x 10-11 -11,200 1.140 2.45 x 10-12
342 4. Scalin~ Processes with Formation of Thick Protective Layers

breadth of the zone was measured at ten different places on a ground section
by means of a sensitive micro comparator. About one thousand such samples
were studied using this procedure. A summary of the alloys which were
used with the evaluated data for the internal and external oxidation is found
in Table 39. The following expression was used to evaluate the temperature
dependence of the layer-thickness growth:
Liei
log - = -+
a
b (4.123)
t T

The oxidation obeyed a parabolic rate law, according to Tammann. Here


Llgi is the thickness of the internal oxidation zone in centimeters, t is the time
in seconds, T is the absolute temperature, and a and b are constants which
are characteristic for a given alloy system. As experiments show, equation

80

Fig. 183. Parabolic course of the


Whh'--+--~--b~-+-~-~
increase of the internal oxidation
zone with time between 750 and
ZO H~--+-~~---t~-+----~~ 1000°C in copper-silicon and copper-
manganese alloys, according to Rhines:
o (I) 0.42% Mn at 750°C; (2) 1.003%
o ~~~~~~~~~
aOO2 aoo¢ aOO6 aooe aOfOcm a018 Si at 750°C; (3) 0.033 % Mn at 750°C;
(4) 0.103% Si at 875°C; (5) 1.55% Mn
(!l if - at 1000°C; (6) 0.103% Si at 1000°C.

(4.123), which is of the type previously used for the thickness increase with
time of the external oxidation zone, is also valid for the growth with time
of the internal oxidation zone. This finds application, for example, in the
growth of the internal and external oxidation zones of copper-silicon and
copper-manganese alloys, as is presented in Figs. 183 and 184. (A parabolic

5r-----.----~---...----,
h.

3 1--I-- -,.f-- r--+--7"'e::.....--/---l Fig. 184. Parabolic course of the increase


with time of the internal oxidation zone and

I
2Hb--+-*--;P'~+-----/---l the exterior scaling layer of a copper-silicon
and a copper-manganese alloy at 875°C,
according to Rhines, Johnson, and Anderson.
(l) Internal oxidation 0 59 t o/c S'
(2) External oxidation . w. 0 1
aooe aOOl; aOOG an' (3) Internal oxidation 042 t o/c M
Mo yer Thickness/-- (4) External oxidation . w. 0 n
4.8. Mechanism of Internal Oxidation of Alloys 343

rate law was always observed for these alloys between 750 and 1000°0.)
Below 600°0 greater deviations from the parabolic law appear, whose
explanation was hitherto not attempted (for example, field transport
phenomena in the internal oxidation zone).
In this connection we will concern ourselves with the mechanism and
the quantitative course of the internal oxidation and the transition from
internal to external oxidation on basis of the presentation by Wagner;l
we will determine the concentration change in oxygen, the base metal
(copper or silver) and impurity metal in the oxidation zone and in the alloy
phase, and the corresponding transport processes. According to the schematic
representation in Fig. 185, a copper alloy should form a compact scaling

Internal
External Scaling
S~~. Unoxidize.d Alloy
I I
I
I
I

Fig. 185. Schematic representation of the


concentration of the alloying metal CM (in
the interior of g ---;. ro, CM = c~.r'''), of the
oxygen Co and of the Ceu in the unoxidized
alloy, the internal oxidation zone, and the
exterior scaling layer, according to Rhines.

layer at the external oxide/gas phase boundary, 0 to I, of OU20 or a hetero-


geneous quantity of OU20 plus impurity oxide. A growth of this external
scaling layer can only take place if Ou+ ions and electrons from the alloy
phase diffuse through the OU20 layer and form an extended lattice on the
outside of Ou 20 with the oxygen chemisorbed at the surface, according to
the mechanism discussed in Section 4.2.1. We are not yet in a position to de-
scribe the course of the migration of silicon and the formation of Si0 2
islands in the external Ou 20 layer.
The formation of an internal oxidation zone, I to II, is now based on
the following hypotheses:
1. Oopper shows a sufficiently great solubility for oxygen. 2 Furthermore,
the maximum saturation concentration of the oxygen c~ in copper
is a function of the oxygen decomposition pressure of Ou 20 at the
corresponding experimental temperature,
c~ = j[7TO.(OU20/OU)]
1'Vagner, C.: Z. Elektrochern. 63, 7i2 (1959).
2Vogel, R., and W. Pocher: Z. Metallk. 21, 333 (1929); F. N. Rhines, and C. H. Mathew-
son: Trans. AIME 111, 342 (1934).
344 4. Scaling Processes with Formation of Thick Protective Layers

2. The mobility of the oxygen atoms in silver or copper or in the


O(-copper alloy must be greater than the mobility of the atoms of
the alloying metal in the silver or copper. In the presence of a con-
centration gradient, the capability for diffusion of the alloying
metal may be considerable, which is always to be considered in
the quantitative evaluation, and was already discussed in Section 4.7.
3. The standard free energy, LlF, of the oxide formation from the
alloying metal must always possess a larger negative value than the
free energy of formation for Cu 20 or CuO (for example, LlFv.o.
= - 397.1 kcal, LlFer20. = - 248.3 kcal, LlFsi02 = - 190.5 kcal,
LlF MgO = - 137.9 kcal and so forth, compared with LlFeuo
= - 25.0 kcal or LlF eu •o = - n.8 kcal).

Figure 185 presents the idealized change of the concentration of the


oxygen and of the alloying metal in the internal oxidation zone and in the
alloy phase. The coincident concentration changes in the external scaling
layer will not be considered at first. If c~ is the oxygen concentration limited
by the decomposition pressure of CU20 in the presence of copper in the alloy
at the external scaling layer/internal oxidation zone phase boundary and
cg is the oxygen concentration at the internal oxidation zone/alloy phase
boundary, whose value is determined by the decomposition pressure of the
oxide of the alloy metal, then under the assumption of a straight-line con-
centration change, (c~ - c~I)/Llg is the concentration gradient in the internal
oxidation zone. If, as noted earlier, the oxygen partial pressure is kept
sufficiently low [Po. = 7T02(CU20/CU)], then only the less noble alloying
metal will oxidize. The concentration of the alloying metal will decrease
owing to consumption in the neighborhood of the internal oxidation
zone/alloy phase boundary. In the stationary state the concentration c~
should be small compared to the starting composition CM that still prevails
in the interior of the alloy. Correspondingly, the concentration gradient is
(CM - c~)/Llg, where Llg is the diffusion distance, which extends from the
phase boundary II to the interior of the alloy at the place where the con-
centration is about 95% of the initial concentration eM. While subsequent
change in the concentration of the impurity metal in the interior of the alloy
is a significant factor in the progress of the internal oxidation, the further
course of the oxygen concentration within the alloy phase can be ignored
because of its small value and insignificant effect.
Thc rate of diffusion of oxygen through copper was measured at dif-
ferent temperatures by Ransley.l It seems justified as a first approxima-
tion to assume the same rate for oxygen diffusion in the internal oxidation
zone of low-alloy copper, as long as the oxides that are formed in this zone
1 Ransley, C. E.: J. Inst. Metals 65,147 (1939).
4.8. Mechanism of Internal Oxidation of Alloys 345

do not reduce the diffusion cross section significantly.1 In weakly alloyed


copper alloys this assumption appears to be fulfilled.
In the same way the diffusion rate of a number of metals in IX-copper
alloys was measured by Rhines and Mehl,2 to whose data we have referred
in the evaluation of the measured results in internal oxidation. In order to
obtain a pure diffusion problem, we further assume that the oxygen supply
at phase boundary I (Fig. 185) according to the reaction equation

Cu 2 0 ........ 2CU(alloy) + O(dissolved in copper alloy)

and the oxygen consumption at phase boundary II according to the reaction


equation, e.g.,
2 O(diSSOl ved in copper alloy) + Si(diSSOl ved in copper) ........ Si 02(embedded in copper)
are relatively high compared to the diffusion rate of oxygen and alloy metals
or silicon.
The above-discussed diffusion possibility permits the derivation of an
equation based on the Fick diffusion law, which makes possible a quantitative
calculation of the time-dependent growth of the internal oxidation zone from
diffusion and concentration data which are available in the literature. 3
A detailed derivation and discussion has been presented elsewhere,4 so we
may skip immediately to the final equations for calculation of the parabolic
scaling constants k' for internal oxidation. Using Rhines' symbols, we
obtain for k'

(4.124a)

or
2Doc~ - 1.68cMDM(OjM)
(4.124b)
cM(OjM) + (c~j3)
where Do and DM are the diffusion coefficients of the oxygen and of the
impurity metal in the copper alloy, L1gt is the thickness of the internal
oxidation zone, c~ is the concentration of the oxygen at the phase boundary I,
and CM the concentration ofthe alloy metal before the oxidation. Furthermore,
O/M is the weight ratio of oxygen to metal in the impurity metal oxide in
the subscale. According to Rhines the error which is obtained from the
1 Meijering, J. L., and M. J. Druyvesteyn: Philips Research Rept. 2,81,260 (1947).
2 Rhines, F. N., and R. F. Mehl: Trans. AIME 128, 185 (1938).
3 Rhines, F. N., W. A. Johnson, and W. A. Anderson: Published as an appendix to
Document No. 1588 of American Documentation Institute, Office of Science Service,
2101 Constitution Ave., Washington, D.C.
4 Hauffe, K. : Reaktionen in und an featen StojJen, Springer, Berlin/Gottingen/Heidelberg,
1955, pp. 522jJ.
346 4. Scaling Processes with Formation of Thick Protective Layers

calculation of the course of the internal oxidation with time from equation
(4.124) is not greater than 1% when
Doc~,O
>5
DMCM,M

However, if this factor should be smaller than 5, then the applicable equa-
tions must be derived more exactly, and, of course, are more complex. Such
derivations appear in the literature}
Table 39 contains a comparison of the experimental data and those
calculated according to equation (4.124b). The agreement can be considered
quite good, even though considerable deviations occur in a few systems.
Generally, as can be seen from equation (4.124), the internal oxidation rate
must decrease with increasing percentages of the alloying metal.
The following relationship, identical to equation (4.123), was used
to describe the temperature dependence of the subscale growth rate:
Llg2 a'
log_i CM = - + b' (4.125)
t T
It is generally fulfilled within the alloying region of 0.1 to 1 wt. % alloying
metal to an accuracy of ±5%. The corresponding values for a and bare
noted in Table 39, column 4.
The mechanism of internal oxidation sketched here is valid between
750 and 1000°C. Below 750°C significant deviations in the calculation of the
time-dependent course of internal oxidation were observed, and the devia-
tions became greater with decreasing temperature. Furthermore, it was
determined that generally at higher temperatures (greater than 800°C)
a separation of the foreign oxide in the interior of the copper crystal results in a
homogeneous statistical distribution, while at lower temperatures the oxide
formation takes place predominantly at the grain boundaries and disloca-
tions. Occasionally, considerable disturbances of the oxygen diffusion and a
decrease in the internal oxidation rate can occur through such a preferred
oxide formation along the grain boundaries because of the covering of the
copper crystal by the oxide layer. A rupture of the oxide covering of the
copper crystal can appear at somewhat higher concentrations of the alloy
metal because of the large amount of oxide formation at the grain boundaries.
However, the diffusion of oxygen was facilitated here because of "internal
cavities," which are associated with an increase in the internal oxidation rate.
Further deviations from relationships (4.124) can also appear if a liquid
phase forms during the oxidation, as is the case with copper alloys with
molybdenum (Section 4.3), and may be suspected when dealing with lead,
cadmium, and higher contents of phosphorus.
1 As references 3 and 4 on page 345.
4.8. Mechanism of Internal Oxidation of Alloys 347

As mentioned at the outset, the appearance of an internal oxidation


zone has a rather considerable influence on the mechanical properties of these
alloys at higher temperatures. Martin and Smith! investigated the influence
of internal oxidation on the fatigue phenomena of copper alloys with
0.3 wt. % Si and 0.05 and 0.25 wt. % Al in polycrystalline material as well
as in single crystals. They determined that in polycrystalline copper alloys
fatigue resistance decreased with internal oxidation, while in single crystals
the opposite was true, that is, there was an increase in the fatigue resistance.
The fatigue phenomena in a material with alternate stresses, which is
connected with the plastic flows and also with the movement of dislocations,
were correlate with the nature of the separation of the oxide particles formed
in the internal oxidation zone by Martin and Smith on basis of the presenta-
tion of Fisher, Hart, and Pry.2 Here the "free" dislocations in the sub scale
of an alloy single crystal, which are determinative for the fatigue phenomena,
are "captured" by the impurity oxide particles, and thus are by and large
eliminated from the plastic flow. The extent of this capture process 'will
depend essentially on the size of these oxide particles and their distances
from one another.
Meijering and Druyvesteyn3 and Smith and Dewhirst 4 have concerned
themselves with the influence of internal oxidation on the hardness of copper
and silver alloys.S
In the above we have dealt exclusively with the mechanism of internal
oxidation, without considering the formation of the external scaling layer
and its influence on the formation of the subscale. Since the attacking
atmosphere often effects an external oxidation, this aspect of the process
must be considered as well. Measurements of the internal oxidation rate in
the absence and presence of an external scaling layer have shown that the
scaling layer generally retards the internal oxidation rate. In analogy to
relationship (4.123) the follo"\\<-:ing expression is valid for the temperature
dependence of the internal oxidation rate in the presence of an external
scaling layer:

I og-t--P
Ll~i_c+d (4.126)

The validity of this relationship is illustrated by Figs. 186 and 187. The
values of the constants c and d are summarized for a few alloys in Table 39.

1 ~Iartin, J. W., and G. C. Smith: J. Inst. Metals 83, 153 (1954/55).


2 Fisher, J. E., E. W. Hart, and R. H. Pry: Acta 31et. 1, 336 (1953).
3l\Ieijering, J. L., and l\I. J. Druyvesteyn: Philips Research Rept. 2, 81, 260 (1947).
4 Smith, G. C., and D. ,Yo Dewhirst: Trans. AU8tralian Inst ..Metals 3,71 (1950); see
also J. J. de Jong: Ingenieur 64, 0.92 (1952).
6 l\Ieijering, J. L.: Tran8. AIME 218, 968 (1960).
348 4. Scalin~ Processes with Formation of Thick Protective Layers

1!71-.----"'-"--+!----l--+-----l
i
em'

11!7'S
'"
03 Fig. 186. Temperature dependence of the
.::s! 1!7,0I---+----+----"''R-c-'''*-=-----1
square of the thickness of the internal oxida,
tion zone of a few copper alloys, according to
Rhines, Johnson, and Anderson: (1) 0,02%
Ge; (2) 0.108% Si; (3) 0.10% Fe; (4) 1.55wt. %
1!7.,~--~=_--~--~=_--~~~
alJ08 1 aOOD
7- Mn.

As can be seen from the figures, constants a and c practically agree on basis
of the same slopes of the straight lines I and 3 or 2 and 4, although the exact
calculation yields certain deviations (see Table 39). The same is true for the
constants band ii. Here, however, the deviations are somewhat greater and
more significant.

10" 1

r ! ma
~ 1J·51-----'-'''''k'''-2''-c--i~--+-------I--------I

~ Fig. 187. Temperature dependence of the


'"~ 1!7'0I----+---+-----"""'l'-..2"-2'k---j square of the thickness of the interior (Lf~il2
and the external oxidation zone (Lf~e)2 of a
~ copper alloy with 0.33% Mn (1 and 2) and
one with 0.45 wt.% Si (3 and 4), according to
Rhines, Johnson, and Anderson: (1 and 3)
internal oxidation alone; (2 and 4) internal
and external oxidation.

The mechanism of combined internal and external oxidation is under-


standably more complex than that of internal oxidation alone. The schematic
representation of the diffusion or concentration behavior is reproduced in
Fig, 185. In order to formulate a convenient mathematical relationship
between the oxidation rate and the determinative diffusion coefficients
and concentrations for this oxidation mechanism, it is necessary to intro-
duce certain simplifications, which must then be experimentally realizable
to a first approximation.

1. The oxygen concentration should be practically zero in the unoxidized


part of the alloy. (This can be attained approximately if one chooses
an alloy metal whose oxide possesses a highly negative free energy
of formation.)
2. The presence of the foreign oxide in the internal oxidation zone
should have no influence on the course of the external oxidation,
4.8. Mechanism of Internal Oxidation of Alloys 349

which is actually possible and observable at very low and non-


noble alloy additions. However, as we have shown in Sections 4.2.1
and 4.2.2, frequently the inward migration of impurity ions into
the external oxide layer even at small alloy additions cannot be
avoided, so that considerable changes can occur in the oxidation rate
in the external oxidation zone, which in return can affect the time-
dependent course of the internal oxidation. If, for example, alloying
additions in the external scaling layer are oxidized out causing the
external oxidation rate to increase by about one order of magnitude,
then the formation can be almost entirely repressed, in spite of the
fulfillment of the thermodynamic condition for the foreign oxide
formation in the internal oxidation zone.
3. The presence of a CuO layer in the external scaling layer of a copper
alloy (always very small, as long as a metal phase exists, see Section
4.5) should be negligible.
4. In the internal oxidation zone as well as in the external scaling layer,
the concentration gradient of oxygen and alloying metal should be
considered as linear to a first approximation.
The general procedure is similar to the discussion above. In the deriva-
tion, for simplicity, c~, c;,i, and c~I were assumed to be equal to ;zero and
further approximations were introduced as before. The following expressions
are the final equations obtained for the internal and external oxidation:

(Internal oxidation) (4.127)

and

(External oxidation) (4.128)

with the expressions for ()(. and f3:

LX = V )(1 + 1.68 D; )
~ cM (1 + 0 ~:
where k = (a-v'ke + ~)2 and a is the quotient of the product of the
thickness of the Cu 20 and the doubled atomic weight of copper divided
by the product of the thickness of the copper times the molecular weight of
the CU20, and
R =
I-' (1 - cM) 0 VIC;
+ 2-LX -
M ke

In addition to the symbols already defined, we may add (c~u - cg u ) and


Dgu as the concentration gradient and the diffusion coefficient of the copper
350 4. Scalin~ Processes with Formation of Thick Protective Layers

in the external scaling layer, and k i and ke as scaling constants of internal


and external oxidation. Unlike the equation for conditions involving only
internal oxidation, these equations cannot be brought into a simple form.
To calculate the value of (C~ll - C~ll) • Dell in order to evaluate equa-
tion (4.128), we can use the vacancy concentration c~llD' obtained by Wagner
and Hammenl in pure CU20 at the phase boundary 0 (see Fig. 185), if we
consider that

(Note that Rhines used dimensions of weight percent.) Furthermore, Dell


in OU20 is known from self-diffusion measurements by Moore and Selikson,2
and is calculable from the experimental results of the oxidation of Ou to
Cu 20 according to Wagner and Griinewald 3 (see Section 4.2.1.1). Rhines
et al. have used the older data according to Pilling and Bedworth4 for their
calculations, which do not have the desired accuracy. In spite of the use of
these old data, the calculated scaling rates are in satisfactory agreement
with the measured ones. In Table 40 a few data are summarized which were
obtained on copper-aluminum, copper-beryllium, and copper-silicon alloys.
In certain copper alloys with beryllium and aluminum as alloying
partners, Rhines 5 found within the internal oxidation zone an intermediate
formation of narrow oxidation zones, which alternated at regular intervals

Table 40. Comparison of Calculated and Observed Oxidation Rates in the Ex-
ternal and Internal Oxidation Zones, According to Rhines, Johnson, and Anderson

Thickness of the external LIte or internal


Llgi oxidation zone after 100 hr, em
Dis-
solved c M' ('CI-C2)·D cu Llge Llgi Cor-
metal wt.% T,oC x 10 7 calc obs calc obs rected

Al 0.03 750 l.43 0.042 0.038 0.093 0.061 0.095


Al 0.45 750 l.43 0.041 0.047 0.017 0.0129 0.016
Al 0.45 875 14.0 0.126 0.107 0.072 0.097 0.076
Al 0.45 1000 87.4 0.314 0.247 0.235 0.260 0.249
Be 0.018 1000 87.4 0.323 0.249 l.031 0.923 l.051
Si 0.076 750 l.43 0.042 0.028 0.048 0.032 0.051
Si 0.18 750 l.43 0.041 0.032 0.028 0.028 0.030
Si 0.18 875 14.0 0.128 0.111 0.110 0.126 0.114
Si 0.59 750 l.43 0.040 0.029 0.012 0.0142 0.014
Si 0.59 875 14.0 0.124 0.080 0.048 0.068 0.052

1 Wagner, C., and H. Hammen: Z. physik. Chern. (B) 40,197 (1938).


2 Moore, W. J., and B. Selikson: J. Chern. Phys. 19,1539 (1951); 20,927 (1952).
3 Wagner, C., and K. Grunewald: Z. physik Chern. (B) 40, 455 (1938).
4 Pilling, N. B., and R. E. Bedworth: J. lnst. Metals 29, 529 (1923).
5 Rhines, F. N.: Trans. AlielE 137, 246 (1940).
4.8. Mechanism of Internal Oxidation of Alloys 351

with narrow metallic zones which were not oxidized, in a manner similar
to the well-known phenomena of Liesegang rings in the rhythmic precipita-
tion of insoluble compounds in gels (see Figs. 188 and 189). These phenomena
were described earlier by Smith! and recently by Rapp.2 The mechanism

a b
Fig. 188. Copper alloy with 0.106 wt.% Be, oxidized for 2 hr at lOOO°C, shows
Liesegang bands in the internal oxidation zone [(a) enlarged 50 times, (b) enlarged
500 times], according to Rhines and co·workers.

of this phenomenon will be discussed at the end of this section. In this


connection, we may note a work by Wagner3 on the precipitation of phases
during diffusion.
Information on the varying influence of the alloying partner on the
relative rates of the internal and external oxidation is obtained from Fig.
190. Here the scaling constants k; and k~ are plotted as special functions of
the silicon content by Rhines and co-workers for a copper-silicon alloy.
It can be seen that the growth rate of the external oxidation zone decreases
rather slowly with increasing silicon content, while the growth rate of the
internal oxidation zone initially decreases rapidly with increasing silicon
1 Smith, C. S.: Mining and Met. 11, 213 (1930); 13, 481 (1932). J. Inat. Metals 46, 49
(1931).
2 Rapp, R. A.: Acta Met. 9, 730 (1961).
3 Wagner, C.: J. Metals 6, 154 (1954). See also J. H. Hollomon and D. Turnbull,
"Nucleation," in Progr. in Metal Phys. 4, 333 (1953).
352 4. Scalin~ Processes with Formation of Thick Protective Layers

Fig. 189. Copper alloy with 0.72 wt. % Al oxidized for 2 hr at 1000°C, shows
perlite structure with Liesegang bands, which represent a rhythmic separation
of Ah03 in copper (enlarged 230 times), according to Rhines and co·workers.

content; however, at higher concentrations (greater than 0.8 wt. %) it too


changes only slightly. Examples of the photomicrographs of a few copper
alloys oxidized at lOOO°C for 2 hr are reproduced in Fig. 191. In all cases
the internal oxidation zone which was sharply bounded by the upper alloy
could be clearly recognized.

\
_~i li
\
Fig. 190. Dependence of the rate of growth of

~ r-- the internal (ki') and external (ke') oxidation


zone of a few copper-silicon alloys as a
function of silicon content in wt. %,
o 0.8 0,,, a8 0.8 according to Rhines and co·workers.
%Si - [ki' = (iJti)2/t and ke' = (iJte)2/t in cm 2 /seo.]

It was Wagner1 who developed the conditions and the quantitative


formulations for the transition from internal to external oxidation. From
existing experimental data it can be concluded that a small concentration
1 Wagner, c.: Z. Elektrochem. 63, 772 (1959).
4.8. Mechanism of Internal Oxidation of Alloys 353

Fig. 191. Copper-silicon alloys oxidized for 2 hr at 1000°C


in air (enlarged 50 times), according to Rhines. Decrease of
the rate growth of the internal oxidation zone with increas·
ing silicon content [silicon in wt.%: (a) 0.045; (b) 0.076;
(c) 0.103; (d) 0.180; (e) 0.30; (f) 0.59; (g) 0.85%].

of a less noble alloying metal and a large oxygen solubility of the base
metal favor internal oxidation without oxide layer formation, whereas a
very low oxygen concentration of the ambient or a small oxygen solubility
and a large concentration of the less noble metal advance the formation of a
354 4. Scaling Processes with Formation of Thick Protective Layers

coherent external oxide layer without a noticeable internal oxidation.


Owing to the relatively high solubility of oxygen in the above-mentioned
metals, silver, copper, etc., the oxygen diffuses rapidly into the alloy and
reacts with the less noble metal causing a precipitation of oxide particles
of the less noble metal in a metallic matrix. The depth g of the internal
oxidation zone is a parabolic function of the time t if diffusion control prevails.
Thus we may let

g= 2y(Dot)1/2 (4.129)

where Do, as mentioned above, is the diffusivity of oxygen in the base metal,
and y is a dimensionless parameter which is calculated below.!
Wagner has proposed that the reduction of the internal oxidation and
at least the transition from internal to external oxidation is due to a reduced
cross section of diffusion for oxygen caused by the precipitated oxide particles
in the alloy. The diffusion of oxygen or metal ions within these oxide particles
in the matrix themselves is negligible, and the reactant atoms are brought
together only by diffusion around the particles. When the volume fraction of
the oxide particles exceeds a critical mole fraction, further precipitation
occurs in the same zone as the prior precipitation, and ultimately a compact
coherent oxide layer results. In order to obtain relations suitable for graphing,
Wagner introduced various simplifying assumptions. Primarily, we assume
that the molar volume of the alloy shall be approximately independent
of the composition of the alloy and the volume increase due to oxygen
dissolution shall be negligible. Furthermore, the specimen shall be sufficiently
thick that the center of the alloy remains unchanged. The free energy of
formation of the oxide AO v must be sufficiently negative so that the solubility
product is very small for the equilibrium:

(A + vO) dissolved in the base metal ~ AO v


Both the mole fraction of oxygen and of the less noble metal A, xo, and XA,
is very small at the precipitation point g compared to the corresponding
saturation mole fractions x~) and xc:.) at the surface and in the initial alloy,
respectively. Under consideration of the following initial and limiting
conditions
X -
0-
xiS)
0 for x = 0, t > 0
Xo ~ 0 for x ~ g, t > 0
xA = x~~) for x ~ g, t = 0

xA ~ 0 for x ~ g, t > 0

1 Wagner, C.: Z. Elektrochem. 63, 772 (1959).


4.8. MechanislIl of Internal Oxidation of Alloys 355

where x is the local coordinate, the equations read


erf[x/2(Dot)1/2] }
Xo = x~){1 - ------- (4.130)
erf I'
erfc[x/2(D At)1/2] }
(4.131)
erfc( yepl/2)
where erf and erfc are the error function and the complementary error
function, respectively, and ep = Do/D A.
At the precipitation point x = g, the flux of the oxygen must be equiva-
lent to the flux of the A metal. Applying Fick's first law, we obtain

lim [_ vDo(OXO) = DA(OXA) ] (4.132)


e .... O . ax X=s-e ax X=s+e
where v is the number of oxygen atoms per A atom in the oxide AO v • Sub-
stitution of equations (4.130) and (4.131) into equation (4.132) yields
VX~) exp 1'2 •erf I'
(4.133)
X~) epl/2 exp(y2ep) erfc(yepl/2)
from which both the parameter I' introduced in equation (4.129) and there-
fore the thickness g of the internal oxidation zone are determined. The
following limiting cases are realizable:
1. If I' ~ 1 and yep1/2 ~ 1, then we obtain for
erf I' ~ 21'/ y;; (4.134)
and for
oJ ( 1/2) exp( _y2ep)
enc yep ~ ---- (4.135)
v'1iYcpl/2
and it follows from equation (4.133) with exp 1'2 ~ 1,

1''''
'"
B (S)

0
2vx(O)
A
(4.136)

Substitution of equation (4.136) in equation (4.129) yields


g= [2x~IDot/vx~)]1/2 (4.137)
Under these conditions the progress of the precipitation front g into the
interior of the alloy is determined only by the diffusion of oxygen. The
validity of equation (4.137) has been tested by oxidation experiments with
silver-indium alloys at 550°0 in air of 1 atm by Rapp.l
1 Rapp, R. A.: Acta Met. 9, 730 (1961).
356 4. Scaling Processes with Formation of Thick Protective Layers

2. If y ~ 1 and yrpl/2 ~ 1, then y is calculable with the following


relation:

y ;:::; y:;Trpl/2x~) = y:;T(DO)I/2 x~) (4.138)


2vx~ 2 DA 2vx~~)
In regard to the equation (4.138) the conditions y ~ 1 and yrpl/2 ~ 1 are
fulfilled, if

and (4.139)
and if

x~)/x<:,) ~ DAIDo ~ 1 (4.140)


From equations (4.129) and (4.138) we obtain instead of equation (4.137)
the following expression for the precipitation front g:
vx(s) Dot 1 / 2
g= 7Tl/2_0_. _ __ (4.141)
x(O) Dl/2
A A

Under these conditions the progress of the precipitation front g is determined


by Do as well as by D A . Correspondingly, more oxide AO y is in the internal
oxidation zone than metal A was present in the initial alloy.
If the mole fraction of AO y in the subscale is denoted by J and the mole
volume by V, then the concentration AO y in moles per unit volume is
JI V. If q represents the area of the sample, then the number of moles of AO v
in a volume element qdg is (f/V)qdg, which must be equal to the number of
moles of A arriving at g in the time dt via diffusion; hence

Jq dg . [ qD A
- - = hm - - - -
(OXA) ] dt (4.142)
V e~O V ox x~g+e
Applying equations (4.129) and (4.131) with equation (4.142), we obtain for
the ratio (1. of the mole fraction of In 203 formed during the oxidation of
silver-indium alloys to the mole fraction of indium in the bulk alloy accord-
ing to Rapp:
J 1
(1. = - ----------- (4.143)
xi~ Y7T1 / 2rpl/2 exp(y2rp) erfc(yrp)1/2
In view of the values of y and rp reported by Rapp, the conditions
y ~ 1 and yrpl/2 ~ 1 arc valid for silver-indium alloys oxidized in air at
550 c C corresponding to (1. ~ 1 according to equation (4.143), i.e., a negligible
enrichment of indium. Under these conditions, the transition from internal
to cxternal oxidation is reachcd at J = xi~ = 0.15. Thc gcnerally valid
4.8. Mechanism of Internal Oxidation of Alloys 357

relation for the transition from exclusive internal to external oxidation


reads:
(0) [71' (s) ] 1/2
XA < 2fmaxvxo DojDA (4.144)

where fmax denotes the maximum mole fraction of A0 y in the two-phase


oxidation zone where an external oxidation will just be prevented. This
relation has been checked by Rapp for the oxidation of silver-indium alloys.
Fair agreement can be seen in Fig. 192.
0.16
0.14
0
0.12
0.10 0
Fig. 192. Transition from internal
to external oxidation of Ag-In
0.08 alloys oxidized at 550°C and various
0
0.06 0 oxygen pressures, according to
0
0
Rapp.-Curve calculated for oxida.
0.04 tion at constant pressure; X alloy
exhibiting local perturbation to the
0.02 penetration of internal oxidation,
i.e., the start of the transition;
OL-~-r-.--.-~-.~ o alloy exhibiting virtually no
-7 -6 -5 -4 -3 -2 -I 0 internal oxidation.

Recently, Maak1 was able to simultaneously extend Wagner's calqula-


tion for internal oxidation with simultaneous external oxidation. Using
equations (4.129) to (4.131), he obtained the following expression
X~)D~2 exp( _y2) VX<J.)D1J.2 exp( _y2rp)
(4.145)
erfy - erf(kj2Do)1/2 erfc(yrp1/2)
In the case of internal oxidation without external oxidation, for instance,
for the silver-beryllium alloy oxidation, generally x~) ~ x~) is valid and
therefore y ~ 1. Accordingly, it is assumed for the problem
y = ~j2(Dot)1/2 ~ 1
Since ~e < ~ with ~e as the thickness of the external oxide layer,
~ej(2Do)1/2 = (kj2Do)1/2 ~ 1
where k is the parabolic rate constant. Then we get
exp( _y2) ~ 1 )
erf y ~ (2j71' 1/2)y (4.146)
erf (kj2Do)1/2 ~ (2j71'1/2)(kj2Do)1/2
with an error smaller than 10%, if y < 0.3.
1 Maak, F.: Z. Metallk. 52, 545 (1961).
358 4. Scaling Processes with Formation of Thick Protective Layers

After substitution of rp = DojDA and equations (4.146) and elimination


of y and k, we obtain
(S)D g(g - gel
(0) 1
x 0 0 = vX (4.147)
2t
A F[gj2(D At)1/2]
if gj2(Dot)1/2 < 0.3 where the auxiliary function F is defined through the
equation 2
F(u) = TTl/2U exp(u2 ) erfcu
1 3
= 1 - - + - - ... (4.148)
2u2 4u 4
The product x~)Do and (g - gel can be determined by photomicrographs of
the oxidized specimens. Equation (4.147) is also applicable to internal
oxidation systems without an external oxide layer (ge = 0). In Table 41
these equations are used for oxidation results for copper-beryllium alloys at
850°C with DA = DEe = 1 X 10-9 cm 2jsec and with Do = 4 X 10-7 cm2jsec
according to Rhines et aU

Table 41. Product of x~) Do in Copper from Measurements of the Thickness of


the Internal Oxidation Zone in Copper-Beryllium Alloys at 850°C, Calculated by
Maak [y = gj2(Dot)I/2 has Values Between 0.1 and 0.2 in Agreement with the Pos-
tulation y < 0.3]'

Mole fraction, g - ge, x~) Do,


Xne t, 104 sec 10- 2 cm g, 10- 2 cm 10-10 cm 2 /sec

0.0077 0.72 1.51 2.27 1.9 (± 0.2)


0.015 1.44 1.46 2.36 1.8 (± 0.2)
0.027 1.44 1.17 2.24 2.6 (±0.2)
0.027 2.16 1.46 2.74 2.6 (±0.6)
0.066 1.48 0.86 1.64 4.3 (± 0.4)
0.060 4.32 1.22 3.10 3.1 (±0.6)

• The value of x~)Do at 875°C communicated by Rhines 2 is 2.4 x 10- 10 cm 2 jsec.

Oxidation experiments on ternary copper alloys were also carried out


with a view toward significance of internal oxidation for alloys of industrial
importance which represent binary alloy systems. Figures 193 and 194 are
photomicrographs of copper-zinc and copper-tin alloys with additions of
beryllium, aluminum, and silicon which were oxidized at 1000 or 900°C.
1 Rhines, F. N.: Trans. AIME 137, 246 (1940); J. Corrosion 4,15 (1947); F. N. Rhines,
'V. A. Johnson, and W. A. Anderson: Trans. AnvIE, Tech. Pub!. No. 1368 (1941);
published as an appendix in the American Documentation Institute, Office of Science
Service, 2101 Constitution Ave., 'Vashington, D.C. in Document No. 1588.
2 Rhines, F. N.: Trans. AIME 137, 246 (1940); J. Corrosion 4,15 (1947); F. N. Rhines,
'V. A. Johnson, and 'V. A. Anderson: Trans. AljIE, Tech. Pub!. No. 1368 (1941).
4.8. Mechanism of Internal Oxidation of Alloy's 359

a b c

Fig. 193. Photomicrographs of a few ternary copper alloys oxidized at


1000°C for 2 hr, according to Rhines (enlarged 500 times). The dark·colored
internal oxidation zone contains a ternary mixture of ZnO, the oxide of the
third alloying partner, and metallic copper. The second narrow zone contains
a finely divided oxide precipitate of only the least noble oxides. The white
zone at the alloy/internal oxidation zone phase boundary is the result of
a "dezincing." (a) Contains 9.52 wt.% Zn + 0.03% Be; (b) contains 9.25% Zn
+ 0.13% AI; (c) contains 9.81 % Zn + 0.085% Si.
360 4. Scaling Processes with Formation of Thick Protective Layers

In all these alloy systems two clearly visible internal oxidation zones appear.
The oxides of both alloy metals are contained in the oxidation region border.
ing the external scaling layer, while the narrow internal oxidation region
only exhibits one oxide (that with the largest negative free energy of forma.
tion) bounded by the alloy. The relative expansion of this oxidation region
is determined by (1) the concentration ratio of both alloy metals, (2) the
temperature, and (3) the oxygen concentration and the gradients in the

a b c

Fig. 194. Photomicrographs of ternary copper alloys oxidized at 900°C for


3 hr, according to Rhines (enlarged 500 times). Also, above the first dark-
colored internal oxidation zone is a narrow second zone with a finely divided
oxidation separation of the most stable oxide. (a) Contains 4.93 wt.% Sn
+ 0.006% Be; (b) contains 5.43% Sn + 0.06% AI; (c) contains 5.02% Sn
+ 0.085% Si.
second region of the internal oxidation zone. Finally, the difference in the
energy of formation of the two impurity oxides is to be noted as a self.
evident assumption. The greater this difference is (for example, LlFAI.O.
= - 364.1 and LlFzno = - 70 kcal), the more clearly distinguishable the
double zone will be. In ternary alloys with the same order of magnitude for
the energies of oxide formation of the alloy metal, only a single zone within
the subscale is observed, as is also the case with binary alloys.
In order to be able to estimate the growth rate in the subscale in ternary
alloys to at least an order of magnitude, an approximation formula was given
4.8. Mechanism of Internal Oxidation of Alloys 361

by Rhines and co-workers for the case in which uniform internal oxidation
zones are formed. The equation has the following form:

(.1 ~i)2
t
= 2Doc~
y
(1 + _k_)
12Do
(4.149)

where

Y = 2c~ + 0
1\fCM
(1 + 1.68-DMk- ) + W
0 CN
(1 + 1.68 TDN) + etc.

Here the characters M and N denote the alloy metals M and N, and k is the
average parabolic oxidation constant in the total internal oxidation (i.e.,
when two metal oxides form).
According to this equation, the increase in thickness of the internal
oxidation zone for the scaling of a copper alloy with 0.02% Be and 0.30 wt. %
Sn at lOOooO was calculated as 5.67 x lO-3 cm/hr, while the measured
value was 4.68 x lO-3 em/hr. Because the boundary of the internal oxida-
tion zone in ternary alloys frequently is not exactly determinable, the for-
mulation of an exact expression does not appear to be urgent at the present
time. A summary of the results with ternary alloys is found at the end of
Table 39.
The internal oxidation of silver and white-metal alloys was investigated
in the same way by Rhines and Grobe l and recently by Rapp.2 A summary
of a few experimental results is found in Table 42. In numerous alloys the
formation of the internal oxidation zone is identical with that in copper
alloy. This is particularly apparent in Fig. 195. The internal oxidation zone
here also has a sharp boundary, frequently parallel to the external phase
boundary. However, at higher alloying additions considerable deviations

Table 42. Rate of Internal Oxidation of Silver Alloys at 850° C, According


to Rhines and Grobe

Thickness
Dis· of the internal
solved oxidation zone
in CM, Co, Do X 10 6 , DM X 10 9 , Oxide after 3 hr at 850°C,
Ag wt.% wt.% cm2jsec cm 2jsec cm
calc obs

Cd 2.48 0.0024 1.6 2.16 CdO 0.0139 0.0276


Cu 1.10 0.0024 1.6 0.86 Cu20 0.0242 0.0945
In 0.73 0.0024 1.6 1.28 In20 0.0414 0.0809
Sn 2.89 0.0024 1.6 5.13 SnO 0.0180 0.0294
Sb 2.79 0.0024 1.6 3.10 8 b203 0.0106 0.0235

1 Rhines, F. N., and A. H. Grobe: Trans. AIME 147, 318 (1942).


2 Rapp, R. A.: Acta Met. 9, 730 (1961).
362 4. Scaling Processes with Formation of Thick Protective Layers

Fig. 195. Photomicrographs of a few silver alloys oxidized at 850°C in air,


according to Rhines and Grobe. The four lower pictures show the oxidation
sepa ration in the internal oxidation zone (enlarged 500 times).
Ag with 0.41 wt.% AI (3 hr oxidized, enlarged 75 times)
Ag with 1.99 wt.% Mg (3 hr oxidized, enlarged 75 times)
Ag with 0.04 wt. % Ti (i hr oxidized, enlarged 50 times)
Ag with 0 .04 wt. % Fe (i hr oxidized, enlarged 50 times)
4.8. Mechanism of Internal Oxidation of Alloys 363

can occur because of a preferential impurity oxide formation at the grain


boundaries, as can be seen from photomicrographs, where an irregular
formation of the internal oxidation zone without a smooth boundary in the
oxide layer can be observed.
Moreau and Benard! recently demonstrated in the oxidation of nickel-
chromium alloys between 800 and 1300°C in air at 1 atm the presence of
the external scaling layer, in addition to the internal oxidation zone, which
was referred to in Section 4.2.1.2 and Fig. 74. However, the formation of the
internal oxidation zone stops when we go over to H 2-H 2 0 mixtures. Evidently,
the chemisorption of H 2 0 and the dissociation according to the overall
reaction equation
H 2 0(g) F O(diSSOlved in alloy) + H~)
is significantly slower than the diffusion of chromium from the alloy to the
alloy/scaling layer phase boundary.
Finally, it may be noted that internal oxidation, not only plays a
decisive role for scaling stability and mechanical behavior (heat resistance,
hardness, fatigue phenomena on alternate loading, etc.) in copper and silver
alloys, but also is significant in the oxidation of other alloys (for example,
steels, nickel, titanium, and zirconium alloys) with alloying partners of
higher negative heat of oxide formation in small concentration, where the
base metal shows a capability for dissolving oxygen. In many cases a very
small oxygen solubility, which is scarcely detectable analytically, will suffice
to cause an internal oxidation on long heat treatments of alloys in air.
The direct measurement of the zone thickness in the photomicrograph
has been used in the determination of the growth of the internal oxidation
zone 2-a procedure which furnishes relatively exact results, but is quite
cumbersome. Raether3 tried to determine the growth of the internal oxida-
tion zone by the change in the electrical conductivity. Quite apart from the
experimental arrangement, which was not very simple, surmounting the
source of disturbances presented considerable difficulties, since the smallest
alloy impurity caused a resistance change which frequently was of the same
order of magnitude as the variation caused by the internal oxidation.
Further comparative investigations appear to be desirable.
In this connection the investigations published by Pawlek 4 on the
electrical conductivity of "purest" copper are very interesting and in-
formative. It was observed that "purest" copper which was heat-treated
1 Moreau, J., and J. Benard: Compt. rend. 237,1417 (1953).
2 Rhines, F. N.: J. Corrosion 4, 15 (1947).
3 Raether: Diss. University Greifswald (1955); S. Raether and K. Hauffe: in Passi-
vierende Filme und Deckschichten, Anlaufschichten, Springer-Verlag, Berlin, 1956, p. 106.
4 Pawlek, F.: Z. MetaUk. 43, 351 (1952).
364 4. Scaling Processes with Formation of Thick Protective Layers

in air showed a higher electrical conductivity than the same copper when
it was heat-treated in high vacuum. This different electrical behavior of the
two copper samples can, in line with the above exposition, be explained only
by the fact that in the first case the traces of less noble alloy components
remaining in the copper-which would cause a lessening of the electrical
conductivity-are "extracted" by internal oxidation by the intruding
oxygen, while, in the case of heat treatment in high vacuum or in an inert
gas, they remain in the alloy.
. Finally it may be noted that one probably has to deal with the appearance
of internal oxidation primarily in titanium and zirconium alloys, since both
metals show a relatively high oxygen uptake, through which the mechanical
technological characteristics are decisively changed. Suitable experiments
should check the increase in the hardness and the brittleness of so-called
pure titanium to determine whether it is to be attributed to the solubility
of oxygen alone or to an internal oxidation of the enriched impurity metal,
which always occurs at the grain boundary (e.g., niobium, tantalum, tungsten,
etc.) with high heats of oxide formation. Even after a short oxidation period
titanium alloys with small additions of tungsten and niobium (about 1 at. %)
became so brittle that samples of sheet 1 to 2 mm thick could be casily
broken.
5. The Mechanism of the Attack of Sulfur
and Sulfur Compounds on Metals and Alloys
Sulfide formation in metals for rate-determining diffusion and transport
processes obeys the same rate laws that were described earlier for oxygen
attack on metals. Wagner1 described the process for the case of sulfide
formation with silver and liquid sulfur at 220°C. However, since the reaction
mechanism is often considerably complicated by rate-determining phase-
boundary reactions, nucleus formation, and pore diffusion, it appears
advisable to consider these reactions separately in order not to cloud the
clarity of the previous presentation.
It has generally been observed that sulfide coatings on metals are
frequently much less adherent than corresponding oxides. This obviously
has to do with the fact that the volume quotient (that is, the ratio of the
volume of the sulfide divided by the volume of the equivalent quantity of
metal), which should always be larger than unity, assumes excessively high
values, so that stresses at the metal/scaling layer phase boundary lead to
cracks and fractures of the sulfide layer. The volume quotients of the sulfides
which are predominantly formed (FeS, CoS, NiS, MuS, Cr2 Sa, etc.) on alloyed
steels are frequently relatively high, and those of the higher sulfides are
still larger (greater than 2.5), so that adherent sulfide coatings are precluded.
Through the appearance of porous layers, the direct contact between the
alloy and sulfur-containing atmosphere is maintained so that scaling can
proceed at an unlimited rate. In Table 43 some data on the common sulfides
are collected. These are taken from a summary on sulfur reactions by
Kubaschewski and von Goldbeck. 2 A further complexity is the considerably
lower melting point of the sulfide compared with the oxide, which may be
further reduced by the formation of an eutectic with the neighboring metal.
The eutectic of the Ni-NiS system lies especially low, at about 645°C. This
behavior of the sulfides, which differs from that of the oxides, makes it ob-
vious that the sulfide coatings frequently do not show favorable protecting
characteristics.
1 Wagner, C.: Z. physik. Chem. (B) 21, 25 (1933).
2 Kubaschewski, 0., and O. von Goldbeck: Metalloberflache (A) 8, 33 (1954).
365
366 5. Mechanism of the Attack of Sulfur and Sulfur Compounds

Table 43. Physicochemical Data for Sulfide Formation for a Few Sulfides,
Compiled by Kubaschewski and von Goldbeck

Heat
Sulfide Stable region Transition Melting Apparent Volume of for-
point, °C point, vapor pressure, quotient mation,
°C mm Hg (900°C) kcal/g-
atom

Ag2S Ag2.00-2.002S 178 842 1.67 7.5


AhSa AhSa-x(950°C) 1100 3.7 64.3
C04Sa 787 - 932° 17.5
C04S2.8-a.2
COgS 8 (835) 24.5
CoS CoS!+x decomposed (10- 4) 2.37 21.0
CoS 2 COS2±x decomposed 4.0 17.0
CrS CrSl.O-1.17 2.5
Cr2Sa CrS1.22-1.48 (56)
Cu 2S CU1.65-2.00S 103/350 1130 1.95 19.6
CuS decomposed 2.9 12.1
FeS FeSl.o-l.a 138 1195 0.8 X 10- 5 2.57 22.8
FeS2 FeS1.95-2.05 (450) decomposed 3.4 20.8
MgS 1.4 84.0
MnS 1530 3 X 10- 6 2.95 49.0
MnS2 decomposed 4.7 (49)
MoS2 1185 7.5 X 10-6 3.5 28.1
NiaS2 Ni3Sl. 95_2.4(700°) 555 (810) 24.8
Ni 7S6 NiSo.82-0.87 390 (560) 21.9
NiS NiSl.o-l.06(5000) 396 decomposed (10- 2 ) 2.5 20.2
NiS 2 NiS 2+x decomposed 4.25 15.9
PbS PbSl±x 1114 0.7 1.75 22.9
ZnS 1020 (1850) 1.2 X lO- a 2.6 48.2

Experiments on the attack of sulfur and H 2 S on copper have often been


reported. As Fischbeck1 was able to show, freshly reduced copper powder
mixed with sulfur and pressed into tablets instantly gave an almost complete
reaction, even at room temperature and below.
A similarly rapid reaction rate was found by Fischbeck and Jellinghaus 2
under the same experimental conditions in the silver-sulfur system although
at somewhat higher temperatures (50-110°C). Mole and Hocart 3 repeated
the experiments of Fischbeck in the Cu 2S formation at room temperature
under pressures to a few thousand atmospheres and found an increase in
the reaction rate with increasing pressure. Even though the increase in the
reaction rate with pressure is only partially understood at this time, there
appears to be a relationship between the great increase in the electrical
conductivity of CU2S with increasing pressure, as was observed by Arseneva-
Gei1. 4 By use of Tubandt's pellet method with a suitably revised experimental
1 Fischbeck, K.: Z. anorg. u. allgem. Chern. 154, 261 (1926).
2 Fischbeck, K., and \V. Jellinghaus: Z. anorg. u. allgem. Chern. 165, 55 (1927).
a Mole. R., and R. Hocart: Compt. rend. 229, 424 (1949).
4 Arseneva.Geil, A. N.: Zhur. Tekh. Fiz. SSSR 17, 903 (1947).
5. Mechanism of the Attack of Sulfur and Sulfur Compounds 367

arrangement, both Tubandt and co-workers! and Fischbeck2 were able to


establish that in sulfide formation on silver and copper only metal ions diffuse
through the sulfide layer from the metal to the sulfur. Of course, the mechan-
ism, as we shall see below, is quite complicated. As an instructive confirma-
tion of the preferred metal ion migration Fischbeck and Dorner3 obtained
tube-shaped formations of Cu 2S, or Cu 2S with a rather thick CuS outer layer
in sulfide formation on copper wires. The concentration equalization experi-
ment described by Fischbeck 4 using two Cu 2S cylinders with different
sulfur contents may be noted as a further indication of a preferred copper
ion migration through the Cu 2 S layer. Further, Dravnieks and Neymark 5
reported on the course of sulfide formation with time in H 2S-H2 atmospheres
and liquid sulfur in the temperature regions between 86 and 224°C and be-
tween 126 and 385°C. While the reaction product from the attack of H 2S
consists of Cu 2S, according to X-ray investigations, and the rate of the
sulfide formation can be represented by a rate law of the form:

dn/dt = AtB (with B > 1)

(see Fig. 196), the sulfide formation in liquid sulfur follows the parabolic
rate law 6 (Fig. 197). The rate of the H 2S reaction, which increases with time
permits us to conclude that the rate-determining process is a phase-boundary
reaction, where evidently the Cu 2S nuclei, whose number increases with
time, are responsible for the accelerated sulfide formation. In spite of the
parabolic course ofthe reaction with liquid sulfur and the evident assumption
of a simple diffusion, the true mechanism appears to be considerably com-
plicated, as one must conclude from the temperature dependence of the
sulfur formation rate in Fig. 198. The relatively rapid increase in the rate of
sulfide formation with decreasing temperature may be attributed to a rapid
pitting below 200°C. However, this initially rapid pitting soon comes to a stand-
still, and in the neighborhood of the sulfur melting point the duration of the
attack is about 10 min. Above 280°C no more pitting was observed, and the
attack proceeded slowly and uniformly. Furthermore, it was determined that
with decreasing temperature the CuS formation increased at the expense of
the Cu 2S formation. Below 210°C practically no Cu 2S was detectable in the
scaling layer. Evidently in this temperature region the diffusion rate

1 Tubandt, C., H. Reinhold, and A. Neumann: Z. Elektrochem. 39, 227 (1933); H.


Reinhold and H. Mohring: Z. physik. Chern (B) 38, 221 (1937).
2 Fischbeck, K.: Z. Elektrochem. 37,593 (1931).
3 Fischbeck, K., and O. Dorner: Z. anorg. u. aUgem. Chern. 181, 372 (1929); 184, 167
(1929).
4 Fischbeck, K.: Z. MetaUk. 24,313 (1932).
5 Dravnieks, A., R. S. Neymark, and C. H. Samans: J. Electrochem. Soc. 105, 183 (1958).
6 Rickert, H.: Z. phys. Chern. [NFJ 23, 355 (1960).
368 5. Mechanism of the Attack of Sulfur and Sulfur Compounds
10tl

t
em ~
O'c?
O'6'°V
, 50
~
.v.#IJ /'
'" 3(J
/
~
c?c? .xl
Hours-
?tltl
em

J2arc
22. o cb
Lc
I 178°C /
III 15rC
/
7 7 /
U / II Fig. 196. Course of sulfide for-
mation of electrolytic copper
HI / V in H2S at different temperatures,
according to· Dravnieks and
ilJJ ~
V MOC Neymark. (An increase in the
scaling constant appears with
IJ 13 $
Hours_ time: dn/dt = const t B > 1).

in CuS (probably formed from a porous layer) was considerably greater


than in Cu 2S, which was obvious through the observation of an increase
in the absence of a Cu 2S layer. The rapid sulfide formation will not
be retarded until a CU2S layer forms between the copper and the CuS. Above
2lOoC the rate of sulfide formation can be described by a parabolic rate law
with an activation energy of about lO kcal/mole. At the same time the
contribution from Cu 2S and CU1. sS increases with increasing temperature.
Recently, Rickert! dealt with new experimental data on the kinetics of the

2fO°C

Fig. 197. Parabolic course of the sulfide


025 a50 1.0 1.5 20 2.5 .1.0 formation on copper in molten sulfur,
Hours ___
according to Dravnieks ami Ncymark.
1 Rickert, H.: Z. physik. Chem. [NF] 23, 355 (1960).
5. Mechanism of the Attack of Sulfur and Sulfur Compounds 369

Fig. 198. Temperature dependence of the weight


loss of copper during the sulfide formation in
liquid sulfur after different experimental times,
as a result of sulfide formation, according to
Dravnieks and N eymark.

CU2S-layer formation and especially on the interaction of the copper ion


diffusion and the phase-boundary reaction. The experimental conditions
and the quantitative relation will be dealt with at the end of this section.
In this case the formation of a porous layer was prevented. For the purpose
of the determination of the phase diagram of the copper-sulfur system,
Wehefritzl and Wagner 2 performed X-ray, conductivity, and emf measure-
ments. In particular, the activity of copper in cuprous sulfide as a function
of the copper-sulfur ratio has been determined with the help of coulometric
titrations. The homogeneity range of Cu 2S at 400°C extends from a Cu/S
ratio of l.9996 ± 0.0002 for samples coexisting with the cubic phase digenite
(CuuS). For hexagonal CU2S in equilibrium with copper no measurable
copper deficit has been observed. By means of emf measurements at the
cell Cu/CuBr/CuxS/C, where CuxS is a local part of the layer, the chemical
potential of the copper in a thicker CU2S layer as a function of the distance y
from the Cu/Cu2S interface could be determined (Fig. 199). If no relative
motion between the metal and the sulfide layer is possible, one always obtains

500r-----r-----.------r-----.---~~~~-~

emf
(mY)

Fig. 199. Chemical potential


of copper in a 24·mm·thick
t400
CU2S layer as a function of
the distance y from the
CUjCU2S interface, according
to Wehefritz. 5 10 15 20
- y(mm)
1 Wehefritz, V.: Z. physik. Chern. [NFl 26, 339 (1960).
2 Wagner, J. B., and C. Wagner: J. Chern. Phys. 26,1602 (1957); J. Electrochem. Soc. 104,
.509 (1957).
370 5. Mechanism of the Attack of Sulfur and Sulfur Compounds

a partially porous layer due to migration of copper ions and electrons from
the CU/CU2S phase boundary to the outside.
Hoar and Tuckerl studied the course of sulfide formation on copper
with sulfur dissolved in benzene and in aqueous ammonium polysulfide in
solution at room temperature. In the first solvent the attack proceeded
unevenly. The sulfide formation rate in the polysulfide solution could be
described at the beginning of the reaction by a linear rate law and was
proportional to the overall concentration of polysulfide. However, in the
later course of .the sulfide formation a transition to a parabolic rate law
was observed, and the dependence of the sulfide-formation rate on the poly-
sulfide concentration immediately became considerably weaker. The reaction
rate increased with increasing pH.
The parabolic rate law appearing for the formation of an electrolytic
coating cannot be interpreted as an ionic diffusion over lattice defects, in
terms of Wagner's scaling theory, since an ionic migration due to only a
chemical potential decrease in the crystal is unlikely at such low temperatures.
We have more reason to assume a predominant pore diffusion through the
coating. Apparently a similar situation occurs in the bromination 2 and
sulfidation3 of silver in bromine- and sulfide-containing solutions, where
on basis of tarnishing and potential measurements a predominant pore
diffusion could be demonstrated for the parabolic rate law.
Extensive experimental data also exist concerning sulfide formation
on silver and silver alloys by sulfur and sulfur compounds at high and low
temperatures, which is understandable because of its technological
significance, since silver and silver alloys have technical applications (electric
contacts, chemical apparatus) as well as uses in dental techniques and for
other commodities (silverware industry). Aside from the large number of
works which rest on a more empirical basis, probably the first systematic
scientific approaches to the kinetics of sulfide formation on silver were carried
out by Reinhold and co-workers4 and by Wagner. 5 Figure 200 shows the
experimental arrangement by means of which Wagner determined the scaling
rate of silver in liquid sulfur at 220°0. According to Tubandt's method 6 for
transference-number measurements a silver cylinder with two thin Ag 2S
cylinders and a glass tube were placed between two brass plates, which
served to contain the liquid sulfur, and were pressed together by means of

1 Hoar, T. P., and A. J. P. Tucker: J. 1Mt. Metals 81, 665 (1952/53).


2 Pfeiffer, I., K. Hauffe, and W. Jaenicke: Z. Elektrochem. 56, 728 (1952).
3 Jaenicke, W.: Z. Elektrochem. 55, 186 (1951).
4 Reinhold, H., and H. Mohring: Z. physik. Chern. (B) 38, 221 (1937); H. Reinhold and
H. Seidel: Z. physik. Chern. (B) 38,245 (1937).
5 Wagner, C.: Z. physik. Chern. (B) 21, 25 (1933).
6 Tubandt, C.: Handbuch Exp ..PhysikXII, Part 1, Leipzig, 1932, pp. 381ff.
5. Mechanism of the Attack of Sulfur and Sulfur Compounds 371

Glass Tube

Fig. 200. Experimental arrangement for the


measurement of the rate of sulfide formation on
silver in liquid sulfur at 220°C, according to
Wagner.

three spiral springs. The apparatus was heated in an aluminum block furnace
in a stream of nitrogen. After the experiment it was found that a Ag 2S
plug of 1-2 mm thickness had grown into the glass tube from the upper
Ag 2S cylinder. Table 44 summarizes the experimental results. If certain

Table 44. Scaling Experiments in the Silver/Sulfur System in the Stable Range
of the IX.Ag2 Phase, According to Wagner (Experimental arrangement accord·
ing to Fig. 200, effective cross sections q = 0.12 cm 2)

Weight change of the Height


Experi- cylinder, mg of Ag2S
mental cylinder, Sealing constant k,
No. time, sec Ag Ag2 S(I) Ag2 S(II) em equiv.jcm-sec
1
2
- 108
- 137
+3
+ 1
+
1+
117
135
0.61
0.77 Ux 1O-,}
2.3 x 10- 6
3 } 3600 - 84 + 1 + 96 0.65 1.2 x 10- 6 1.6 X 10- 6
4 - 108 +2 + 126 0.60 1.4 x 10- 6
5 - 121 +5 + 131 0.58 1.5 x 10- 6

side reactions which are caused by the vapor pressure of the sulfur (reaction
in the gas phase) which becomes detectable at 220°0 are neglected, the
migration of silver ions and electrons through the Ag 2S cylinder from the
silver to the liquid sulfur could be demonstrated clearly and quantitatively.
Marker experiments confirm this mechanism. l The experimentally obtained
scaling constant k was 1.6 x 10-6 equivalent/em-sec for 220°0. This value
satisfactorily agrees with that of 2 or 3 x 10-6 , which was calculated from
(3.21) or (5.4). These investigations were extended by Rickert,2 who studied
the kinetics of the Ag 2 S layer formation at higher temperatures (300 and
400°C), and by special techniques, he also studied the role of the phase-
boundary reaction. 3 As is shown in Fig. 201, in no case is the parabolic
rate law fulfilled exactly, e.g., the total process is not determined solely by
the ion diffusion. The same is true for the sulfidation of copper. 4
1 Mrowec, S., and T. Werber: Acta Met. 8, 819 (1960).
2 Rickert, H.: Z. physik. Chern. [NFl 23, 335 (1960).
3 Rickert, H., and C. D. O'Briain: Z. physik. Chern. [NF] 31,71 (1962).
4 Mrowec, S., and T. Werber: Acta Met. 7, 696 (1959); J. Mikulski, S. Mrowec, I. Stronski,
and T. Werber: Z. physik. Chern. [NFl 22, 20 (1959).
372 5. Mechanism of the Attack of Sulfur and Sulfur Compounds

Because it is generally true that phase-boundary reactions participate


in a tarnishing process, we dealt with some results of Rickert's investigations.
Generally, the rate of diffusion is inversely proportional to the thickness x
of the layer and, furthermore, is a function of the chemical potential.
em

3.0

2.5

2.0

1.5

1.0
Fig. 201. The temporal forma-
0.5 tion of a compact sulfide layer
on silver in liquid sulfur at
200, 300, and 400°C, according
0 see 25000 50000 75000 100000 125000 to Rickert.

Since at sufficiently high temperatures thermodynamic equilibrium is


established at the Ag 2S/Ag phase boundary, fL~g - fL<t~ equals the negative
half-value of the free energy of formation - iLlG = E*F. E* was measured
by Kiukkola and Wagner1 with the following electrochemical cell:
(i) (a)

Ag I AgJ I Ag2S I Pt, S (liquid)


According to Rickert and Wagner2 we obtain for the silver ion flow through
the AgjAg 2S phase boundary and through the sulfide layer:
(5.1a)
and

(5.1b)

Since the electron transfer through the phase boundary is not restrained,
it follows for the electrochemical potential "Ie of the electrons in both
phases:

or
fLe(Ag) - FV(Ag) = fLe(Ag 2S) - FV(Ag 2S)
and therefore
fLe(Ag) - fLe(Ag 2S) = F[V(Ag) - V(Ag 2S)]
1 Kiukkola, K., and C. Wagner: J. Electrochem. Soc. 104, 379 (1957).
2 Rickert, R., and C. Wagner: Z. physik. Chem. [NFJ 31, 32 (1961).
5. MechanisIn of the Attack of Sulfur and Sulfur COInpounds 373

with generally valid relations


/L~g = /LAg+(Ag) + /Le(Ag)
/L~)g = /LAg+(Ag2S) + /Le(Ag2S)
where /LAg+(Ag) and fLAg+(Ag 2S) are constant because of the pure silver and
the high silver ion defect concentration in Ag 2S, respectively, we obtain
from equation (5.la):
jAg+ '" eXP{;[V(Ag) - V(Ag 2S)]}
By taking logarithms an equation is obtained which is analogous to the
Tafel equation for electron reactions in aqueous solutions
B
10gjAg+ = const + -[V(Ag) - V(Ag 2S)]
F
To calculate the unknown chemical potential /LAg in equation (5.la) or
(5.lb) we consider that the expressions of the right-hand side of equations
(5.la) and (5.lb) are equal:
O (i) _ XAg+ ( (i) (a»)
(5.lc)
A exp[B(II'Ag
r - r"'Ag)] - LJgF2 rAg
II, - rAg /J.

With the calculated value for fL<fg substituted in equation (5.la) or (5.lb)
we get the silver ion flow jAg+ or the partial current density i Ag+ of the silver
ions for a given thickness of the Ag 2 S layer. The change of g with the time
at unidimensional growth may be expressed as
(5.ld)
where v is the molar volume of Ag 2S.
By means of the derived equations the growth rate is calculable. For this
purpose, we consider the following two limiting cases:
Case 1: For thin Ag 2S layers or short reaction times, the denominator
on the right-hand side of equation (5.lc) becomes very small, so that with
LJ' g -+ 0 the left-hand side of equation (5.lc) receives a maximum value and
we get /L~)g ;;::: /L~~. Under these conditions, the difference of the chemical
potentials disappears and the transfer of the silver ions through the Ag/Ag 2 S
phase boundary becomes rate-determining:
jAg+ = A exp[B(/L~g - /L~~)]
This time law, independent of the thickness of the Ag 2S layer, rules in the
temperature range between 200 and 300°C for a reaction time of about
I min.l The calculated and measured rates of the Ag 2S growth are compiled
in Table 45.
1 Czerski, L., S. Mrowec, K. Wallichow, and T. Werber: Arch. Hutnictwa 3, 49 (1958).
374 5. Mechanism of the Attack of Sulfur and Sulfur Compounds

Table45. Comparison of Measured and Calculated Ag2S


Formation Rates for Short Reaction Times, According
to Rickert and Wagner

dUdt, cm/sec
T,OC
calc obs

220 0.5 0.9 X 10-4 0.85 X 10-4


300 3.9 7.0 X 10-4 6.5 X 10-4

Case 2: For thicker Ag 2S layers or longer reaction times the linear rate
law is no longer valid, since I'-~)g - I'-~~ in the Ag 2 S layer becomes larger and
may no longer be neglected. Simultaneously, the difference I'-~g - I'-~~
becomes continuously smaller and approaches the limiting value EKF,
where EK is the barrier potential for silver ion transfer through the Ag/
Ag 2 S boundary:
o
I'-Ag -
w"'EF
I'-Ag '"
K

Finally, it follows from equations (5.lb) and (5.ld) that

(5.2)

The barrier potential EK is always present and is responsible for the


deviation from thermodynamic equilibrium at the Ag/Ag 2 S phase boundary;
hence we always have to consider its magnitude. For 300°C we obtain from
equation (5.2) with E* = 0.244 volt, EK = O.B volt, and UAg+ = 4.26
ohm-l-cm-I, the value for the sulfidation constant k' = 1.0 X 10-4 cm 2/sec,
in good agreement with the value 0.8 x 10-4 obtained experimentally.

To resolve the apparent contradiction for Ag2S diffusion processes described


by Tubandt,l which indicated pure cation' conduction, and the Hall-effect
measurements by Klaiber,2 which led to a conclusion of predominant electron-
excess conduction, Wagner3 undertook a study of the conduction and diffusion
mechanism in e<-Ag2S, since this disparity precluded a quantitative description
of the scaling mechanism. As X-ray measurements by Rahlfs 4 show, the sulfur ion
partial lattice in e<-Ag 2 S (stable above 179°C) is ordered, while the silver ion
partial lattice contains a large number of lattice defects. The deviation from the

1 Tubandt, C.:·Handbuch Exp.-PhysikXII, Part 1, Leipzig, 1932, pp. 404ff.


2 Klaiber, F.: Ann. Physik (5) 3, 229 (1929).
3 \Vagner, C.: Z. Elektrochem. 40, 364 (1934).
4 Rahlfs, P.: Z. physik. Chern. (B) 31, 157 (1935).
5. Mechanism of the Attack of Sulfur and Sulfur Compounds 375

stoichiometric composition was determined quantitatively by Wagner. l Kracek 2


demonstrated a sulfur excess on basis of his phase diagram, which was obtained
at lower and intermediate temperatures. However, this determination is in-
consistent with the findings of Wagner and with the electron excess conduction
demonstrated by the Hall-effect measurements, since a sulfur excess in the
Ag 2 S lattice can only appear through electron consumption in the lattice, from
which electron holes, not free electrons, originate. Hebb 3 tried to resolve this
contradiction by assuming that excess sulfur was present in the form of neutral
atoms rather than S 2 - or S - ions. This assumption would correspond to the Schottky
superoxidation semiconductor type. 4 A further possibility is that with the
assumption of intrinsic electronic disorder (0.,-' e + EB), which is independent
of ionic lattice defects, electron holes have a far greater mobility than free
electrons. On basis of later experimental data Wagner conclusively proved the
presence of a slight excess of silver ions and free electrons. s From potential
measurements on the following cell:

and the formula

the silver excess in Ag 2 S could be determined, although at 200°0 the composition


varied only between the narrow limits Ag 2 . 001O S and Ag 2 •000 S. Here P.M. denotes
the chemical potential of the silver in Ag 2 S and the 0 characterizes the pure
silver phase; Zl is the valence of the silver. This result is in agreement with the
Hall-effect measurements by Klaiber.
J ost and Riiter 6 as well as Wagner7 furnished definite proof, on basis of
transference and emf measurements on suitable cells, that in the Ag/Ag 2 S/S(IlQUld)
arrangement only about 1 % of the silver ions take part in current transport.
For 100 % ionic conduction in Ag 2 S, the emf of the cell

Pt(sulfur)/Ag 2 S/Ag

should be about 0.2 volts, corresponding to the reaction energy for the Ag 2 S

1 Wagner, C.: J. Chern. Phys. 21,1819 (1953); see also C. Wagner: "Galvanic Cells with
Solid Electrolytes Involving Ionic and Electronic Conduction," Proc. Comite into Ther-
modyn. and Cinetique Electrochim., London, 1957, p. 361.
2 Kracek, F. C.: Trans. Am. Geophys. Union 27,274 (1946).
3 Hebb, M. H.: J. Chern. Phys. 20, 185 (1952).
4 Schottky, W.: Z. Elektrochem. 45, 33 (1939).
5 Wagner, C.: J. Chern. PhYiJ. 21,1819 (1953); see also C. Wagner: "Galvanic Cells with
Solid Electrolytes Involving Ionic and Electronic Conduction," Proc. Comite into
Thermodyn. and Cinetique Electrochim., London, 1957, p. 361.
6 Jost, W., and H. Riiter: Z. physik. Chem. (B) 21, 48 (1933).
7 Wagner, C.: Z. physik. Chern. (B) 21, 42 (1933).
376 5. Mechanism of the Attack of Sulfur and Sulfur Compounds

formation. However, an emf of about only 2 to 5 X 10-3 volts was found. From
this the transference number of the silver ions,

Eobs
tAg+ = - -
Eca!c

was about 0.001 to 0.002.


Wagner was able to show in a detailed thermodynamic treatment that the
predominant conduction character of a mixed conducting ionic crystal, such as
Ag28, is decisively determined by the electrode material. If, for example, we
choose the following arrangement

i.e., Ag2S between two metallic conductors, then by application' of an electric


field only electrons are furnished at one side and removed at the other side,
unless the potential across the cell is so high (0.2 volts) that decomposition takes
place. Under steady·state conditions in Ag 2S a gradient of the metal-to-nonmetal
ratio is formed,! so that the ionic diffusion compensates the electrolytic migration.
On the other hand, in the experimental arrangement chosen by Tubandt
+ Ag/Ag28/AgI/Ag-
no electrons are furnished from the right to the left at the AgI/Ag28 phase
boundary because of the pure ionic conduction character of AgI. In the stationary
current flow a concentration gradient of electrons in Ag2S is introduced which is
just equal to the applied potential gradient, so that there is only a silver-ion
current, fulfilling Faraday's law. Just this was observed by Tubandt
Weight

Liquid sulfur

Glass tube

Nitrogen atmosphere

Ag
Fig. 202. Apparatus for determining the potential
difference at the Ag/ Ag2S in terface in the Ag/ Ag2S /Sllquld
tarnishing system.

1 Wagner, C.: Proc. Comitli Int. Thermodyn. and Cinetique Electrochim. 103,627 (1956).
5. Mechanism of the Attack of Sulfur and Sulfur Compounds 377

and co-workers'! It can be seen that the special experimental arrangement


forces the ion conduction in Ag 2 S.
This "electron blocking" mechanism can be made clear in the following way:
The chemical potential of sulfur in Ag 2 S at the Ag 2 SjAgI phase boundary is
increased by the passage of current, since the silver ions migrate from the Ag 2 S
boundary surface through the AgI much more quickly than through the AgaS
phase, so that we obtain a local gradient of the chemical potential in the Ag2 S
layer similar to that in the AgjAg 2 SjS tarnishing system. The silver ion transport
caused by the chemical potential gradient is independent of the electric field_
The validity of this conclusion can be confirmed in two ways_
1. In the experimental arrangement used by Wagner the anode was divided
in half (see Fig. 203); the halves were isolated from one another, and only one

Ag Cathode

AgI
Fig. 203. Experimental arrangement for the determina-
tion of the migration mechanism of the Ag+ ions through AgzS
Ag2 S in the current transference with two separated
silver anodes, according to Wagner. (Both silver halves
I and II show equal mass loss.)

half was connected to the positive pole of the potential source. After completion
of the transference experiment, the mass increase of the silver cylinder on the
cathode side was equal to the mass decrease of the two silver halves on the anode
side and losses for the two halves were equal. That is, the silver half without
the external potential showed the same mass loss as the connected half; this
can be explained only in terms of a superimposed transport process-such as a
tarnishing process.
2. We now consider an electrochemical cell with ionic conductors at both
sides of the Ag 2 S cylinders as grounding electrodes
(i) (a)
Ag I AgCl I Ag 2 S I AgCl I Ag
Ag Suo

and bring silver to one AgCljAg 2 S phase boundary and liquid sulfur to the other.
According to the methods of electrochemical thermodynamics the emf E of this
cell may be ca.lculated according to the following formula:

E = - -1
F
f
I-'s(a)
ta
-dl-'s
Zg
(5.3a)
I-'S(I)

By differentiation one obtains from this


ta(l-'s = I-'s(a)) = - 2F(dEjdl-'s(a)) (5.3b)
where I-'S denotes the chemical potential of the sulfur at the (a) and (i) phase
boundaries, Zs = 2 is the valence of the sulfur, and ta is the transference number
1 Tubandt, C., S. Eggert, and G. Schibbe: Z. anarg. u. allgem. Chern. 117, 1 (1921).
378 5. Mechanism of the Attack of Sulfur and Sulfur Compounds

of the electrons. The experimentally obtained value for E agrees with that
calculated from equation (5.2) with t3 ~ 1.
If we regard 0.2 % as the maximum deviation from the stoichiometry in
Ag2S, then we calculate the difference in the silver excess concentration at the
AgjAg2S and Ag2SjAg phase boundaries at about 6 x 10-4 g-atomjcm 3 , which
is the maximum value. Now, if we use this value in the usual Fick equation
for the calculation of the effective diffusion coefficient with the use of the rational
scaling constants noted in Table 44, then we obtain De!l= 2.7 X 10-3 cm2jsec
at 200°0 obtained by Tubandt, Reinhold, and Jost. 1 This was explained by
Wagner in the following way: by neglecting self-diffusion coefficients of sulfur,2
D~ ~ D1~, we obtain from equation (3.22) the expression

(5.4)

where CAg is the average concentration of the silver ions in Ag2S in g-atomjcm3
and a1)g and a~~ are the activities of the silver ions at the Ag2SjAg or Ag2SjS
phase boundaries; respectively. If for a first approximation we use the corre-
sponding concentrations CAg in place of the activities, we obtain from Fick's
formula
(5.5)

By neglecting the change of the lattice parameter with the AgjS ratio we use the
rela tionshi p

where V M = 35 cm 3 , which is the molar volume of Ag2S and r~i or r~~ is the
AgjS ratio in Ag 2S in equilibrium with silver or sulfur. By insertion in (5.5) it
follows that

(5.6)

Using k(220°C) = 1.6 x 10- 6 and actual measured values of rAg at 200 and
300°0 from potential measurements of suitable electrochemical cells
(AgjAgljAg 2SjPt), Wagner obtained the diffusion coefficient Derr = 2.8 X
10- 2 cm2jsec for 220°0 which is still about one order of magnitude higher than
the lower limiting value obtained by Hebb. This value, which on basis of later
meaSUI'e;nents by 'Vagner, can be regarded as quite exact, is therefore about 2.8 x
10 3 greater than the value D!g given above. 3 This difference, which must be con-
sidered for the calculation of scaling constants in self-diffusion measurements
when additional effects, e.g., electric field transport, appear, can be understood in
the silver-sulfur system if one considers that the electrons in Ag 2 S exit rapidly
1 Tubandt,C., H. Reinhold, and 'V. Jost: Z. anorg. u. al/gem. Chern. 177, 253 (1928).
2 Braune, H., and O. Kahn: Z. Elektrochem. 31, 576 (1925).
3 New self-diffusion measurements at 4-cm grown Ag 2 S cylinders have been carried out
by S. ~Irowec and H. Rickert: Z. physik. Chern. [NFl 32, 212 (1962), which are in
agreement with R. L. Allen and 'V. J. 1\Ioore: J. Phys. Chern. 63, 223 (1959).
5. Mechanism of the Attack of Sulfur and Sulfur Compounds 379

in large numbers, causing an additional electrical potential gradient in addi-


tion to the chemical potential gradients due to the silver ions which are already
present; this in turn causes an increase in the transport rate of the silver ions
through the tarnishing layer. In agreement with the above results for Deff we
obtain the following value for the ratio of the diffusion coefficients, according to
Wagner

(5.7)

where ro = 2 represents the stoichiometric ratio of Ag/S. In this connection we


may note an investigation into the structure of Ag 2 S: Rickert l proposed a
vacancy diffusion mechanism with a correlation factor between 0.3 and 0.4.

Reinhold and Seide1 2 found considerable deviations from the parabolic


time law in sulfide formation on silver with sulfur vapor or a mixture of
H 2-H 2 S, as was discussed above for sulfide formation on copper. Moreover,
at higher temperatures (above 500°C) the reaction proceeds more slowly
than one would expect according to (3.21). Furthermore, the reaction rate
for a sulfide formation on silver at varying Ag 2S layer thicknesses in the
ratio 1 : 5 is practically independent of the layer thickness for a reaction
carried out at 196°C. This finding clearly precludes a rate-determining
diffusion and indicates it rate-determining phase-boundary reaction. Jost 3
suspected that the reaction at the S(vapor) /Ag 2S phase boundary should be the
limiting process. Investigations by Drott4 suggest that the rate-determining
step is diffusion on the surface of Ag 2 S. At 90°C the reaction is strongly
influenced by the H 2S pressure. The growth morphology was also studied.
If the sulfide formation on silver is diffusion-controlled, which should
always be the case at sufficiently high experimental temperatures with thick
sulfide layers, and if the silver ions in interstitial lattice positions are more
mobile than the silver ion vacancies, the rate of formation should be de-
creased by the introduction of sufficiently soluble metals with a valence
higher than one into the Ag 2 S layer since the concentration of silver ions in
interstitial lattice positions, which is decisive for diffusion as well as for the
scaling rate, is reduced by the addition of suitable divalent metal ions; for
example,
AgO" + CdS...---'- Cd.·(Ag) + Ag2S }
and (5.8)
CdS...---'- Cd.·(Ag) + AgO' + Ag2S
1 Rickert, R.: Z. physik. Chem. [NFJ 24, 418 (1960); 23, 355 (1960).
2 Reinhold, R., and R. Seidel: Z. physik. Chem. (B) 38, 245 (1937).
3 Jost, W.: Diffusion und chemische Reaktion in festen Stoffen, Dresden, 1937, p. 156.
4 Drott, J.: Ark. Kemi 16, 333 (1960).
380 5. Mechanism of the Attack of Sulfur and Sulfur Compounds

If, however, the vacancies are more mobile, then an inverse effect should be
detectable. Such investigations at high temperatures apparently have not
yet been performed but would be useful in an evaluation of the possibility
of development of sulfur-resistant alloys.
Foley and co-workersl conducted sulfide formation experiments on
silver alloys with additions of aluminum, cadmium, antimony, indium,
magnesium, t'hallium, and zinc in benzene and mineral oil which contained
sulfur, but only at lower temperatures (40-92°C). As spectroscopic measure-
ments have shown, the added metal-ion content in the Ag 2S layer is con-
siderably less than that which was present in the alloy phase. The experi-
mental results are summarized in Table 46.

Table 46. Impurity Metal Content in the Tarnishing Layer of Silver Alloys,
According to Foley

Al Sb Cd In Mg Mn TI TI Zn

Impurity metals in the} 4.0 1.8 30.0 U.8 0.4 4.3 2.9 5.9 9.6
alloy, wt.%

Impurity metals in the} 0015 0.45 0.9 0.95


tarnishing layer, wt. % . 7.0 0.4 0.04 O.oI 0.6 0.7
0.5

One may assume the widespread development of a heterotype mixed


phase in the scaling layer on basis of only the identification of j3-Ag 2S phase
by X-ray or electron-diffraction methods, since otherwise impurity sulfides
would also have been observed. Furthermore, the experimental results
indicate an enrichment of the alloy metal in the alloy phase at the alloy/tar-
nishing layer phase boundary. In all cases, a decrease of the sulfide-formation
rate after 100 hr of reaction time was observed if the foreign ions which were
introduced were at least divalent, according to (5.8). A thallium addition had
no influence on the rate of sulfide formation, since thallium apparently is
included as ThS. Lichter and Wagner 2 investigated the attack on copper-
gold, silver-gold, nickel-copper, and silver-copper alloys by sulfur at
elevated temperatures. Sulfidation of copper-gold and silver-gold alloys
involving low gold contents, e.g., 10 at. % Au, yields a composite scale
consisting of an outer homogeneous sulfide layer and an inner two-phase
layer involving gold-rich alloy and sulfide. This is in accord with theoretical
considerations in Section 4.7, where an alloy phase/sulfide interface is not
stable since the interdiffusion coefficients in the alloy are much lower than
1 Foley, R. T., M. J. Bolton, and W. Morill: J. Electrochem. Soc. 100, 538 (1953); H. O.
Spauschus, R. W. Hardt, and R. T. Foley: J. Electrochem. Soc. 101, 6 (1954).
2 Lichter, B. D., and C. 'Vagncr: J. Electroclum. Soc. 107, 168 (1960).
5. Mechanism of the Attack of Sulfur and Sulfur Compounds 381

the diffusivities in the sulfide phases. Silver-copper alloys yield a sulfide


layer of uniform thickness, in which the ratio of the components is the same
as in the alloy in spite of the large difference in the standard free energies
of formation of Ag 2S and Cu 2 S. This diffusion is the only process which is
significant in the sulfide layer, and accordingly, the rate of attack is about
as high as that on pure silver or copper. However, further experiments are
required for a deeper insight into the mechanism.
The scaling experiments on silver with selenium and telluriuml will be
discussed separately. While the reaction between silver and selenium fits
the reaction mechanism for sulfide reasonably well, complications appear
in the formation of Ag 2Te from strong limitations upon the reaction at the
AgjAgTe phase boundary. Thus, for example, at the outset of the reaction
the deficit of silver (represented by silver ion vacancies) near the Ag/AgTe
interface, caused by silver migration away from the Ag2Te phase is not
compensated by a transfer of silver ions from the metal phase. The silver
cylinder remains completely constant. The transfer of silver ions from the
silver into the Ag2Te phase did not begin until the silver deficit in Ag 2Te
exceeded a critical value of 0.03-0.04 grams Ag per gram Ag 2Te. A quantita-
tive explanation of the reaction mechanism is not possible at this time
because of inadequate knowledge of the lattice defect phenomena and of the
migration mechanism. 2 The influence of the state of the silver phase on the
reaction, which is significant for the transfer of the silver into the Ag 2Te
phase, is completely unexplained. Hard-rolled silver is much less reactive
than slightly heated material. Further complications occur in that the
migration rate of the silver ions and thus the electrical conductivity depend
on the composition of the Ag2Te phase. Further, a more rapid reaction was
obtained if a Ag 2S cylinder was placed between the Ag/Ag 2Te phase boun-
daries corresponding to the combination Ag/Ag 2S/Ag2Te. Evidently transfer
of silver ions into the Ag 2S phase and from there into the Ag2Te phase is
energetically more favorable than the direct transfer of silver from the
metal phase into the Ag 2Te layer. The insertion of Ag 2 S was regarded by
Wagner3 as a form of catalysis, despite the longer diffusion path.
Raub and co-workers 4 investigated the attack of sulfur on silver-
palladium alloys with one of the compositions usually used in dental work
(30 wt. % Pd), especially at 800°C. These particular alloy samples were
embedded in a gypsum-graphite mixture for this purpose and heated for

1 Reinhold, H., and H. Seidel: Z. physik. Chem. (B) 38, 245 (1937).
2 Electrical and optical properties of Ag2Te were reported by J. Appel: Dis8. Braunschweig,
1955; R. J. Ratchford and H. Rickert: Z. Elektrochem. 66, 497 (1962).
3 Wagner, C.: "Reaktionen mit Metallen," in Handbuch MetaUphyaik, Vol. 1, 2, Leipzig,
1940.
4 Raub, E., B. Wullhorts, and W. Plate: Z. MetaUk. 45, 533 (1954).
382 5. Mechanism of the Attack of Sulfur and Sulfur Compounds

various times (5-30 min) at 800°C. During the sulfide formation, besides the
main product, Ag 2S, some Ag 2Pd 3S appears, which forms an eutectic melt
at about 700°C with Ag 2 S. While a 20-I-'-thick gold coating forms a completely
protectiye layer against sulfur (Fig. 204), appreciable local scaling appears
with gold coatings between 5 and 10 I-' thick after 30 min of reaction time
at 800°C (Fig. 205). Reinacher and Wagner1 demonstrated a preferred

Fig. 205. Photomicrograph of a d ental alloy heated at 800°C for


30 min in a packing of gypsum and graphite with a 10'/L-thick
gold covering, according to Raub and co· workers (enlarged
160 times) . Note extensive corrosion.
1 Reinacher, G ., and E . Wagner: Zahnarztl. Wochenschr. 9, 126 (1954).
5. Mechanism of the Attack of Sulfur and Sulfur Compounds 383

movement of palladium into the Ag 2S phase, which causes depletion of


palladium in the alloy.
Arkharov and Mardeshev1 studied the growth structure of the copper
selenide layer which was formed on copper at 350, 450, and 500°C after a
selenide formation time of 2-10 hr. The same investigations were carried
out on copper telluride layers after a 2-hr attack by tellurium at 600°C.
A needle growth was observed in both cases, in a manner similar to that
described by Pfefferkorn 2 in the oxidation of copper and zinc. The Cu2Se
needles grew in the (1l0) direction and CU2Te needles grew in the (100)
direction. A similar needle growth was observed recently by Rickert during
the sulfidation of silver.3
Scaling experiments with sulfur and sulfur compounds have been carried
out with a few other metals besides copper and silver, especially iron and
steel. In this connection a work by Geld and Jessin4 should be mentioned
in which a similarity in terms of the model for the scaling of iron in oxygen
and in sulfur at higher temperatures was emphasized. As in oxidation,
in sulfidation a multiphase coating forms at higher sulfur-vapor pressures.
The primary part of the coating is made up of FeS phase, which is covered
by a thin FeS21ayer. As called for by the Wagner theory, a preferred diffusion
of iron ions and electrons from the metal to the sulfur side-i.e., to a region
of greater iron ion vacancy concentration-was assumed. This mechanism
leads one to expect a thickening of the sulfide layer toward the outside as
well as a formation of cavities at the Fe/FeS phase boundary. As the sulfide
formation progresses, cavity formation frequently becomes more extensive
so that contact between the scaling layer and the metal is interfered with
more and more. The iron transport is retarded to an increasing degree, so
that finally the reaction comes practically to a standstill, since a diffusion
of sulfur along the grain boundaries of the sulfide layer to the metal surface
or a plastic flow in the protective layer does not prevent a further loss of
contact between the metal and sulfide layers.
The experiments of Hauffe and Rahmel 5 and of Meussner and Birchena1l 6
contributed a further explanation of the mechanism of sulfide formation on
iron. In order to follow the progress of sulfide formation continually with
time the first authors used a quartz spiral balance, as can be seen in Fig. 206.
The sample for sulfide formation was placed on the balance in the reaction
bulb. The corresponding sulfur vapor pressure was produced at the bottom

1 Arkharov, V. 1., and S. Mardeshev: Dokl. Akad. Nauk SSSR 55, 517 (1954).
2 Pfefferkorn, G.: Naturwiss. 40, 551 (1953); Z. Metallk. 46, 204 (1955).
3 Rickert, H.: Z. physik. Chem. [NFl 21, 432 (1959).
4 Geld, P. W., and O. A. Jessin: Zhur. Priklad. Khim. 19, 678 (1946).
5 Hauffe, K., and A. Rahmel: Z. physik. Chem. 199, 152 (1952).
6 Meussner, R. A., and C. E. Birchenall: Corrosion 13, 677 (1957).
384 5. Mechanism of the Attack of Sulfur and Sulfur Compounds

of the reaction vessel, whose lower part was heated by means of a second
furnace which was placed close to the reaction furnace. The change in the
state of the quartz spiral balance with time could be followed without
interruption by means of a telescope sighted through a window situated in
the reaction furnace. Mrowec and Werber l carried out marker experiments
which indicate a compact inner sulfide layer and a porous outer layer.

Furnace

CirCUlar
Holder Fig. 206. Apparatus for the measurement of the rate
of sulfide formation of metals and alloys, according to
Hauffe, Neunhoeffer, and Rahmel.

Since FeS is a p ·t ype semiconductor, according to the model for the


production of Fe 2+ vacancies and electron holes in the FeS lattice due to
reaction of sulfur with FeS at higher temperatures,

~ s~g)..---- FeS + Fe D" + 2 (£) (5.9)

the rate of sulfide formation on iron was found to be proportional to the


sixth root of the sulfur vapor pressure, since sulfur appeared in the gas
phase as S2 molecules. 2 As can be inferred from the experimental results in
Fig. 207, the rate of sulfide formation on electrolytic iron or carbonyl iron
·VrT-----.,-----.------r~,

Fig. 207. Dependence of the rate of sulfide


formation (k" in g2/cm4·sec) of two kinds of iron
on the vapor pressure of sulfur in a log-log
representation, according to Hauffe and Rahmel.
(T = 760°C, x = first series of measurements,
., o , 2 o = second series of measurements, Ps. in
log ps!(pstin mm Hg ) - mmHg.)

1 Mrowec, S., and T. Werber: Naturwiss. 45, 335 (1958).


2 On the dissociation of sulfur, see H. Braune, S. Peter, and V. Neveling: Z. Naturforsch.
6a, 32 (1951).
5. Mechanism of the Attack of Sulfur and SuHur Compounds 385

at 760°C in the pressure region from 0.1 up to 100 mm Hg follows the fifth
or seventh root of the sulfur vapor pressure. The absolute value of the
scaling constant for 670°C at a sulfur vapor pressure of 100 mm Hg is
k" = 7 X 10-7 g2jcm4 -sec
For the rate of sulfide conversion of FeS to FeS2 at the same temperature
and a sulfur vapor pressure of 1 atm it was found that
k" = 7 X 10-10 g2jcm4 .sec
Accordingly, the Fe2S formation proceeds about 1000 times more slowly
than the FeS formation at the same temperature. Furthermore, both layers
grow with time according to the parabolic rate law. The temperature depen-
dence of the sulfide-formation rate in the reaction of iron to form FeS at a
sulfur vapor pressure of 10 mm Hg is presented in Fig. 208. The rate constants

-5.7
"'t
-6.0 ~ "'0..
'--0. I-

Fig. 208. Temperature dependence of the rate


of sulfide formation of electrolytic iron at a -tJ.(j~o
constant sulfur vapor pressure of 10 mm Hg,
according to Hauffe and Rahmel.

dealt with by Meussner and Birchenall are somewhat higher than those
reported in Fig. 207. Some data from these experiments are given in Table 47.
From the Wagner scaling formula rewritten for sulfide formation on iron

(5.10)

the value of k from the experimental data for the partial conductivity
of the iron ions, "Fe for P S 2 = 1 atm and T = 670°C, was calculated as

Table 47. Rational Rate Constants of


Sulfidation of Armco Iron

1
T, °0 k r , equiv. FeJcm 2 .sec

650 8
700 5.2
8.4 xx lO-
10-8
750 12.7 X lO-8 PS2 = 100 mm Hg
800 17.9 x lO-8
850 26.5 X lO-8
900 35.3 X lO-8
386 5. Mechanism of the Attack of Sulfur and Sulfur Compounds

2.9 x 1O- 2/ohm-cm. If we use the value obtained by Savelsberg1 for the
total conductivity of FeS at 670°C, " = 20/ohm-cm, then we obtain for
the transference number of the iron ion in FeS:
"Fe 2.9 X 10-2
t Fe = - = =1 X 10-3
" 29
In accordance with the considerations for the decrease in the rate of
oxidation of metals with a p-conducting coating (see Section 4.2.1), the rate
of sulfide formation on iron should be decreased by the addition of metals
which are introduced as monovalent ions in the sulfide layer, since in this
case the concentration of Fe 2+ ion vacancies is reduced. Since the alkali
metals are insoluble in iron, while copper and silver are only slightly soluble,
an improvement of the sulfide formation stability of iron cannot be achieved
in this way. This fact has already been noted by Kubaschewski and von
Goldbeck,2 who therefore proposed alloying with metals that have higher
heats of formation of their sulfides, which are thus formed preferentially,
and that further form compact coatings on the alloy with lower rates of
ion diffusion. On basis of empirical data, higher additions of chromium and
aluminum alloying partners appear to be suitable. Since aluminum is
relatively stable against sulfur, it is desirable to plate sulfur-sensitive alloys
with aluminum. In fact such coatings on steel protect quite well against
sulfur and also against H 2S,3 even above the melting point of aluminum.
Murakami and Shibata 4 have reported on optimal layer thicknesses. Accord-
ing to these authors, the good sulfur stability of aluminum-rich alloys and of
aluminum itself can be attributed to the formation of a protective film of
A1 2S3. To what extent the Al 20 3 film, which always exists on these alloys,
plays a principal role in scaling protection has to be established by suitable
experiments. As can be seen from the summary by West,5 the 18-8 steels
and steels with 17% Cr appear to be sufficiently sulfur-stable up to the
boiling point of sulfur. However, Dravnieks 6 was able to show that there is
a considerable attack of liquid sulfur on chromium-free steels. The attack
followed the parabolic rate law corresponding to the conductivity decrease
with time for thin steel sheets (0.6-0.8% C, 0.4% Mn, 0.03% S) in liquid
sulfur between 300 and 450°C. The temperature dependence of the parabolic
sulfide formation constants yields an activation energy of about 41 kcal/mole,

1 Savelsberg, W.: Z. Elektrochem. 46, 379 (1940).


2 Kubaschewski, 0., and O. von Goldbeck: Metalloberfliiche (A) 8, 33 (1954).
3 Ipavic, H.: Heraeus Vacuumschmelze 1923-1933, p. 290; F. Kayser: Foundry Trade J.
47,263 (1932).
4l\Iurakami, T., and N. Shibata: Sci. Rep. Tohoku Imp. Univ. 30, 252 (1942).
SWest, J. R.: Chem. Ind. Eng. 53, 223 (1946).
6 Dravnieks, A.: Ind. Eng. Chem. 43, 2897 (1951).
5. Mechanism of the Attack of Sulfur and Sulfur Compounds 387

which is shown in Fig. 209. The scaling layer consisted of a thin nonmagnetic
FeS layer directly on the steel, followed by a thicker magnetic FeS layer
and a very thin FeS21ayer. The phase directly neighboring the steel consisted
of a gray porous powder which recrystallized very quickly when the steel
lying underneath was consumed.
J

~
2

t .
~
~
(J
0",
t'-..0
Fig. 209. Temperature dependence of the para- t
bolic sulfide formation constant on thin steel
sheets (with 0.6-0.8 wt.% C, 0.4% Mn, 0.025% P,
"'
0.002% Si, 0.003-0.02% Cu, 0.004-0.06% Ni, and f.J 1.1(. t.7 1.8
0.033% S) in molten sulfur between 300 and 450°C,
according to Dravnieks. (Activation energy is about
41 kcaljmole.) '150 AA? JSO

The conditions for sulfide formation on nickel in sulfur vapor are essen-
tially different. As can be seen from Figs. 210 and 211, the rate of sulfide
formation at 630°0 proceeds according to the linear rate lawl and is pro-
portional to p~/2. This finding is not compatible with the behavior to be
2

Fig. 210. The course of sulfide


formation on nickel and nickel
alloys at 630°C and a sulfur vapor
pressure of 0.1 mm Hg, according
to Hauffe and Rahmel. (1) Ni
+ 1.0 at.% Ag; (2) Ni + 0.5 at.%
Ag; (3) Ni + 0.1 at.% Ag; (4) pure
Ni; (5) Ni + 0.1 at.% Cr; (6) Ni
+ 0.7 at.% Cr; (7) Ni + 1.0 at.% () ~----L---~~~O----~-----8~o~m
~in--
Cr.

expected from the lattice-defect picture of NiS. 2 As microscope investigations


have shown,l no internal cavities are created during sulfide formation on
nickel, which can always be regarded as a relatively certain sign of a prefer-
ential diffusion of metal ions plus electrons toward the exterior. Because
1 Hauffe, K., and A. Rahmel: Z. physik. Chem. 199, 152 (1952).
2 Hauffe, K., and H. G. Flint: Z. physik. Chem. 200,199 (1952).
388 5. Mechanism of the Attack of Sulfur and Sulfur Compounds

of the great difference in density of Ni and NiS (see Table 43) the assumption
of the appearance of fine pores and "expanded" grain boundaries and with it
a preferred rapid transport of sulfur to the nickel is probable.
If one assumes that the reaction
(5.11)

and not the diffusion is rate-determining, then, on basis of experimental


observations that k ~ p~/2 for both partial processes,
-k skg ) +e (Ni)..----'- S~em. an Ni (chemisorption) (5.12)

Scl;em. an:-li + ~i....-->-NiS + e(:-li) (incorporation into the surface) (5.13)


the splitting of the sulfur molecule into atoms during chemisorption is the
slowest and thus the rate-determining step.l (Here e(Ni) denotes an electron

---
2.*'10-8
21 _, I-
Molcm . sec
/'
V
L
/
/
c/
V
o t.2 Fig. 211. Rate of sulfide formation on pure
a* a8
w;;- nickel as a function of sulfur vapor pressure
at 630°C, according to Hauffe and Rahmel.

in the metal.) The absolute value of the rate constant in sulfide formation
at 630°C and a sulfur vapor pressure of 0.5 mm Hg is

l = 1.8 X 10-6 g-atomJcm2 -sec

From these relations, it is apparent that the behavior of the rates of


sulfide formation of nickel-silver and nickel-chromium alloys is inconclusive
(Fig. 210). As can be seen, the rate of sulfide formation becomes smaller
with increasing additions of chromium. New investigations on the sulfidation
of nickel alloys have been carried out by Pfeiffer 2 and Lichter and Wagner. 3
While the first author could confirm the linear rate law at 630°C and the
same dependence on the sulfur pressure, the other authors found that the
parabolic rate law is approximately obeyed at 400°C for nickel-copper

1 See also the short communication of S. Mrowec and H. Rickert: Z. physik. Chem. 28,
422 (1961); L. Czerski, S. Mrowec, and T. Werber: J. Electrochem. Soc. 109,781 (1962).
2 Pfeiffer, 1.: Z. Metallk. 49, 267 (1958).
3 Lichter, B. D., and C. Wagner: J. Electrochem. Soc. 107, 168 (1960).
5. Mechanism of the Attack of Sulfur and Sulfur Compounds 389

alloys. The overall Cu/Ni ratio in the scale is virtually the same as in the
alloy. Further investigations on the mechanism of sulfide formation and
especially on the state of the lattice defects in the NiS are very necessary
in order to explain the influence of alloying additions on the rate of sulfide
formation. The intrinsic semi conduction inferred from the sulfur pressure
independence of the electrical conductivity of NiS is not unreasonable
on basis of the phase diagram of the nickel-sulfur system (with NiaS2,
Ni7S6, NiS, Ni aS 4, and NiS 2), even though it has not yet been proved. The
explanation of the ionic lattice defects, which for intrinsic semiconductors
are not directly coupled with the free electrons and holes,! may be possible
through transference experiments and emf measurements with suitable
cells, e.g., those used by Wagner in the Ag 2 S system. Diffusion experiments
with radioactive sulfur and nickel ions in NiS single crystals would also
be very informative. Further investigations were published by Dravnieks 2
on the mechanism of formation of NiS and higher sulfides from the immersion
of nickel in liquid sulfur. Ni 6S5 was found to be a reaction product as well
as NiS and NiS 2 , but Ni 7S6 and Ni aS4 were not present. Between 205 and
445°C the rate of formation of NiS 2 was over a thousand times smaller than
that of NiS.
A few works of Gruber,3 White and Marek,4 and Ipavic5 are concerned
with the attack of dry and moist H 2S on iron and other metals. These in-
vestigations, as well as the sulfide experiments on iron alloys by Baukloh
and Spetzler6 in nitrogen or hydrogen atmospheres with 2% H 2S and with
various metals in helium with 5% H 2S by Farber and Ehrenberg,7 are of a
purely empirical nature. In these atmospheres silver, chromium, tantalum,
tungsten, molybdenum, and aluminum are relatively stable between 500
and 900°C, while weakly alloyed steels, copper, nickel, cobalt, and iron are
strongly attacked under these conditions. Alloyed steels occupy an inter-
mediate position. Chromium additions affect corrosion limitations in any
case; however, they are not fully effective until above 15%. The alloy
mentioned by Grunert, Hessenbruch, and Ruf,s 30% Cr-5% Al-65% Fe,
which shows a good scaling stability in moist H 2S is an especially instructive
example. It is shown from all of these experiments that a larger addition
of aluminum or chromium is especially effective in producing resistance to
1 Hauffe, K. : Reaktionen in und an feBten Stoffen, Berlin/Gottingen/Heidelberg, Springer,
1955, pp. 162ff.
2 Dravnieks, A.: J. Electrochem. Soc. 102,435 (1955).
3 Gruber, H.: Z. MetaUk. 23, 151 (1931).
4 White, A., and L. F. Marek: Ind. Eng. Chem. 24, 859 (1932).
5 Ipavic, H.: Heraeus VacuumBchmelze, 1933, p. 290.
6 Baukloh, W., and E. Spetzler: Korr08ion u. MetaU8chutz 16, 116 (1940).
7 Farber, M., and D. M. Ehrenberg: J. Electrochem. Soc. 99, 427 (1952).
8 Grunert, A., W. Hessenbruch, and K. Ruf: Heraeus VacuumBchmelze, 1933, p. 169.
390 5. Mechanism of the Attack of Sulfur and Sulfur Compounds

sulfidation. Even susceptible nickel and cobalt displayed resistance toward


sulfur attack after an addition of 15% aluminum.
According to the investigations of Vernon,! small quantities of S02
(0.01 %) in dry air do not influence the corrosion and scaling rate of numerous
metals and lower-alloyed steels. In moist air, on the other hand, strong
corrosion was observed even at room temperature. 2 The experiments by
Vernon 3 are quite informative on the cumulative effect of H 20 vapor and
S02 during the corrosion of nickel. Dry air containing S02 is only slightly
aggressive toward nickel. However, when the water vapor in the atmosphere
was added, a spontaneous corrosion occurred. On the other hand, if the
nickel sheet was first brought into an atmosphere containing water vapor
and then into a dilute S02 atmosphere, no change in the nickel sheet was
observed. Evidently the chemisorption and the attack of S02 must precede
that of H 20 in order to cause significant corrosion.
In contrast to H 2S and sulfur, S02 attacks magnesium more severely
than copper and copper-magnesium alloys.4 While additions of aluminum
increase the scaling stability of copper toward S02, 0.54% additions of other
metals (silver, cadmium, chromium, manganese, nickel, silicon, tin, zinc)
have practically no effect. 5 In addition to copper, iron and tantalum were
also severely attacked by S02. However, pure zirconium is extraordinarily
stable up to 500°C.6 Further investigations have been conducted on the
attack of S02 in flaming gases on cast iron? and gas turbine materials8
(e.g., steels with 14-20% Cr, 11-20% Ni, 6-10% Co, 2-4% Mo, 0-4% Cu,
0-0.08% Ti, and 0-3 wt.% Nb). In completely burned gases these steels
are not severely attacked by S02, which is present between 700 and 1l00°C-
in contrast to incompletely burned gases. Whittingham 9 reported on the
attack of S02 on lower-alloyed steels in flames containing CO as a function
of the surface temperature of the steel. At lower H20 vapor content of
0.60-0.70% the corrosion rate at 25°C increased to a maximum between
60 and 70°C and decreased with increasing temperature by about one order
of magnitude at 125-150°C. With increasing H 20 content the maximum was
displaced to higher temperatures (100-125°C).

1 Vernon, W. H. J.: Trans. Faraday Soc. 31,1668 (1935).


2 See also U. R. Evans: Metallic Corrosion, Passivity, and Protection, London, 1948,
pp. 155ff·
3 Vernon, W. H. J.: Nature (London) 167, 1037 (1951).
4 Baukloh, W., and W. W. G. Krysko: Metallwirtschaft 19, 169 (1940).

5 Hallowes, A. P. C., and E. Voce: Metallurgia 34,95 (1946).


6 Hayes, E. T., A. H. Roberson, and R. H. Robertson: J. Electrochem. Soc. 97,316 (1950).

7 Hallett, ::\1. 1\1.: J. Iron Steel Inst. 170, 321 (1952).


8 Sykes, C., and H. T. Shirley: Symposium on High-Temperature Steels and Alloys for
Gas Turbines, Iron and Stcel Inst., 1951, p. 53.
9 Whittingham, G.: J. Appl. Chem. 5, 316 (1955).
5. Mechanism of the Attack of Sulfur and Sulfur Compounds 391

Table 48, compiled by Kubaschewski and von Goldbeck, summarizes


the resistance to sulfide formation of a few metals and alloys.
As we have seen from the presentation in this section, unlike the oxida-
tion of metals and alloys, the mechanism of sulfide formation in sulfur vapor,
S02, and H 2S atmospheres is still only partially explained, so that much
still remains to be done in this field. Before one can begin with successful
experiments on sulfide formations, other than Ag 2S and Cu 2S, an intensive
study of the lattice defect mechanism of the basic sulfides which occur most
frequently (FeS, NiS, FeS2, CuS, AhS3, and MgS) must be carried out.

Table 48. Several Metals and Alloys That Display Resistance


to Attack by H 2S, S, and S02, According to Kubaschewski and
von Goldbeck

H 2S S S02

Ag Ag
Ni, Nb Ni, Cu Ni
Mild steel Mild steel
Fe Fe Fe
Fe-Mn Cu-1OMg
Inconel Fe-14Cr Fe-15Cr
Cu Cu-Mn Ta
Fe-15Cr 80Ni-13Cr-6.5Fe Cu, Brass Increasing
Mn AI-alloys stability
Fe-25Cr Cr Mo,W

j
Cr Fe-17Cr Fe-30Cr
Fe-18Cr-8Ni Fe-18Cr-8Ni Fe-18Cr-8Ni
Fe-22Cr-IOAI (Inconel)
Cu-1OMg Hastelloy C Cu-12AI
Fe-12AI, Ni-15AI Zr
Ta,Mo, W AI, Mg
AI, Mg
Au Au

Suitable electrochemical cells must be built in the manner proposed by


Wagner1 and the deviation from stoichiometry must be determined. Further,
self-diffusion measurements with radioactive isotopes for the determination
of the conduction character are indispensable for success in obtaining results
that can be evaluated.
1Wagner, c.: "Galvanic Cells with Solid Electrolytes Involving Ionic and Electronic
Conduction," in Proc. Comitli into Thermodyn. and Cinetique Electrochim., London,
1957, p. 361.
6. The Oxidation Mechanism of
Metal-Carbon Alloys and Carbides
As is well known, metal carbides are the bases of the hard metals, which
owing to their hardness and strength at high temperatures are of tremendous
technological significance. So that full advantage may be taken of this great
strength at high temperatures, continuing effort must be made to improve
the scaling stability of these hard metals. Since the explanation of the oxida-
tion mechanism is rather difficult due to the presence of carbon in the alloys,
it is not surprising that people have been concerned with this complex
question in a purely empirical way. Wagner and co-workersl have attempted
to determine experimentally the fundamental relationships describing the
effect of the carbon during the oxidation of carbon-containing metals and
carbides in terms of thermodynamic and kinetic considerations. Of great
importance for resistance to oxidation is awareness of the conditions under
which gaseous CO and CO 2 are formed, since these break up the protecting
oxide film and thus cause an increased oxidation rate.
For CO or CO 2 formation, which is affected by reaction of the carbon
present in the alloy with the oxide MeOy, we write
1 1
C (alloy) + - (MeOy) = CO(g) +- Me (alloy) (6.1)
y y
2 2
C (alloy) + - (MeO y) = CO 2(g) + - Me (alloy) (6.2)
y y
The equilibrium partial pressures of CO and CO 2 obtained from the mass
action law are given by
Peo = Klae/aU~ (6.3)

Peo, = K2ae/a2Jr~ (6.4)

where Kl and K2 are the equilibrium constants in (6.1) and (6.2) and ae
or a~Ie the activity ofthe carbon or of the metal, introduced during the course
of the reaction at the alloy/oxide phase boundary.
1 \Vebb, \V. \V., J. T. Norton, and C. \Vagner: J. Electrochem. Soc. 103, 112 (1956).
392
6. Oxidation Mechanism of Metal-Carbon Alloys and Carbides 393

In the following we will discuss two separate cases which are significant
for the explanation of the mechanism.
Case 1. If the metal has a relatively small affinity for oxygen, then a
preferred oxidation of the carbon will take place; the sum of the partial
pressures Peo + Peo, will thus be considerably greater than the externally
prevailing gas pressure p, so that the forming oxide film is ruptured con-
tinually. New alloy surfaces are thus constantly being exposed, and oxidation
of the alloy proceeds more rapidly than oxidation of the carbon-free metal,
as long as the oxide layer does not "heal" rapidly.
For the preferred oxidation of carbon, carbon activity at the alloy/oxide
phase boundary must be considerably smaller than that in the interior of the
alloy. Furthermore, from the porosity of the oxide film we get for the sum
of the partial pressures
Peo + Peo. = P (6.5 )
On basis of the reaction
1 1
- MeO y + CO ~ - Me + CO2 (6.6)
y Y
the CO/C0 2 ratio is given by the following expression:

Peo,/Peo = K3/aM~ (6.7)


where K3 is the mass action constant in reaction (6.6).
Carbon activity and the partial pressures of CO and CO 2 may be cal-
culated from (6.3), (6.4), (6.5), and (6.7).
The assumptions in Case 1 are realized in the oxidation of nickel-carbon
alloys with 2.3 wt. % C at 1000°C. Because of depletion of the carbon at
the alloy/oxide phase boundary, carbon diffusion occurs from the interior
to the phase boundary. As thermodynamic calculations using (6.3) and
(6.4) have shown, Peo + Peo, is about 106 atm at 1000°C. This high pressure
causes a continual rupturing of the oxide film and a great increase in the
oxidation rate, as is shown in Fig. 212.
Since at 1000°C the maximum solubility of carbon in nickel amounts to
only 0.27 wt. %, the excess carbon is present as graphite in needle-shaped
precipitates. After long oxidation times (about 17 hr), during which the
carbon is preferentially "oxidized out," cavities appear at the sites of the
graphite inclusions. The rate of carbon removal was somewhat higher than
that calculated using D~OOO = 3 X 10- 7 cm 2/sec.1 Since the weight decrease
with time due to the disappearing carbon was greater than the weight
increase caused by the NiO formation, the weight of the alloy sample
1 Lander, J. J., and A. L. Besch: J. Appl. Phys. 23, 1305 (1952).
394 6. Oxidation Mechanism of Metal-Carbon Alloys and Carbides

gJcm.' ----- ---- ---- ---- -


~--
6

_f
/
°0 3J/JCI'J WlOOO GOCW 8IJOOO !(JOQOO fi:'Qt1IXJsec lWlO(}()
Duration of Oxidation _
Fig. 212. Course of the oxidation of nickel and nickel-
carbon alloys with 2.3 wt. % C at IOOO°C in oxygen of
I atm pressure, according to Webb, Norton, and Wagner.
Total oxygen consumption L1molq for nickel (.6) and for
nickel-carbon alloy (- - -); CO 2 formation L1mco /q in
nickel-carbon alloy (0). '

decreased with time. For quantitative evaluation, separate determinations


of the formation of CO and CO 2 with time, on one hand, and of the oxide
film, on the other, are required. Thus, a few of the experimentally obtained
quantities will be noted in the following.
If we denote the oxygen uptake per unit surface by LJmo/q, then we
obtain:
,1 mo/q = (,1 m/q) + 1\-,1 mco.lq (6.8)
where iJm/q is the observed mass change in the alloy sample per unit surface
area and LJmco,/q is the quantity of CO 2 formed in the same time per unit
surface area; 3/11 (= 12/44) is the ratio of the atomic weight of carbon to the
molecular weight of CO 2 •
If the metal and the carbon are not preferentially oxidized, the total
consumption of oxygen is calculated from the following equation:
(1 - x) Me + xC + [t(1 - x)y + x] O2 = (1 - x) MeO y + xC0 2 (6.9)

where x is the mole fraction of carbon in the initial alloy or the initial carbide
and y the average number of oxygen atoms per metal atom in the oxide layer.
Then from (6.8) and (6.9), we get the following relationship :
iJ. mco.!q 44x
iJ.m/q 16(1 ~ x)y -12x (6.10)

If (6.10) is fulfilled (no preferred oxidation of the metal or carbon), then


the consumption of oxygen per unit surface may be calculated directly from
the weight change of the alloy sample. From (6.10) and (6.8) it follows that
iJ.mo 16(1 - x)y iJ.m
(6.11)
q 16(I-x)y-12x q
6. Oxidation Mechanism of Metal-Carbon Alloys and Carbides 395

By insertion of (6.10) into (6.11)we can also calculate L1mo/q from the quantity
of CO 2 that is formed:
LImo 16(1-x)y Llmeo,
(6.12)
q 44x q

Another possibility can be considered under Case 1. In spite of the


loss of carbon at the alloy/oxide phase boundary, the diffusion of carbon
can always be neglected if the regions which form in the metal-carbon
system in which each phase is homogeneous are very small and the oxide
formation proceeds very rapidly. This has been shown to be the case in the
tungsten-carbon system in which two carbides, W 2 C and WC, appear with
practically unchangeable composition. The solubility of carbon in tungsten
is very small. Kieffer and Kolbll did carry out investigations on the rate of
oxidation of these carbides, but they could not be quantitatively evaluated,
since only the weight change of the carbide was determined, but not the
simultaneous carbon loss. For this reason Wagner et a1. investigated the
oxidation of WC again. The experimental results for 700 and 1000°C are
presented in Fig. 213. At 700°C, WC oxidized according to a linear rate law
fOO
glcm.2

80
' 1Qtt7°C /
V
IIJ(7(J"C / ~OO°C
,
I
I

I
/
I
I

~
/
I
I
/
20 I
I
I
/
---- - ---- -
~------
700°C
---- - ---- ----
__ - - I

o0 ~OOO 8000 tZQtt7 lo(J(JQ i!lJCW 2fJtWsec28001J


Durat ion of Ox jdotjon _
Fig. 213. Course of the oxidation of tungsten and tungsten
carbide at 700 and 1000°C in oxygen of 1 atm pressure,
according to Webb, Norton, and Wagner [0 •• for pure W
and - - for WC; 0 (48/44) 6mco2/q and. (48/36) 6m/q].

with the ratc constant 1" = 4 X 10- 6 gjcm 2 .sec. At 1000°C, the accuracy
of the measured points was not very great because of the high rate. The
determination of the quantity of CO 2 which is formed simultaneously
corresponds to the total reaction
WC + 2.50 2 = W0 3 + CO 2
1 Kieffer, R., and F. K61bl: Z. anorg. u. allgem. Chem. 252, 229 (1950).
396 6. Oxidation Mechanism of Metal-Carbon Alloys and Carbides

and excludes a preferred oxidation of carbon. The higher rate of oxidation


of we compared to W is understandable if one considers that the oxide
layer is continually cracked open because of the formation of CO and C02.
In fact, the scaling layer consists of yellow porous W03. Unlike scaling on
pure tungsten, there is no compact lower blue oxide layer (see Section 4.5.1).
According to Newkirk,! tungsten carbide powder reacts relatively
slowly with oxygen below 500°C, but above this temperature a rapid oxida-
tion suddenly sets in which at 529°C leads to complete burning.
Case 2. If the absolute value of the free energy of formation of the oxide
per gram.atom of oxygen is very high, at relatively low temperatures the
sum of the partial pressures, Peo + Peo" according to the calculation from
(6.3) and (6.4), does not attain the value of the external prevailing total
pressure. Under these conditions the carbon can remain in the alloy and
not disturb the compact oxide film that is forming. This mechanism,
however, gives rise to other phenomena.
The carbon remaining behind at the alloy/oxide phase boundary can
diffuse into the alloy, increasing the carbon content of the alloy during the
oxidation. A mechanism of this type comes into play, according to Wagner
and co-workers, in the oxidation of manganese-carbon alloys. In accordance
with the above assumption, practically no CO or C02 formation could be
detected due to the high oxygen affinity of the manganese. After 90,000 sec
(25 hr) a weight increase in the manganese-carbon alloy with 1.33 wt. % C
of 0.08 g/cm 2 was found at 1000°C in oxygen, while the formed quantity
of CO 2 amounted to only 0.00027 g/cm 2 • This was about 2% of the expected
quantity in the nonpreferential oxidation. Under the above experimental
conditions the carbon analysis yielded a content of 2.15 wt. %, in agreement
with the calculated value, with the assumption that the carbon remains
in the alloy. The investigations of Vogel and Doring 2 and of Isobe 3 showed
that carbon is present-similar to the case in iron-in a y-manganese phase
at 1000°C. At higher concentrations a carbide forms which is stable at
1000°C in the region of 3-4 wt. % carbon.
On basis of the above considerations, it is expected that the oxidation
rates of pure manganese and manganese-carbon alloys will be practically
equal. The experimental results at 1000°C given in Fig. 214 confirm this.
In agreement with Gurnick and Baldwin,4 Wagner and co-workers found the
parabolic scaling constant for pure manganese to be kif = 5 X 10-6 g2/cm4-sec.
It was only insignificantly higher for the manganese-carbon alloy.

1 Newkirk, A. E.: J. Am. Chem. Soc. 77, 4521 (1955).


2 Vogel, R., and W. Doring: Arch. Eisenhilttenw. 9, 247 (1935).
3 Isobe, M.: Sci. Rep. Research Inst. Tohoku Univ. (A) 3, 468 (1951).
4 Gurnick, R. S., and M. W. Baldwin: Trans. ASM 42,308 (1950).
6. Oxidation Mechanism of Metal-Carbon Alloys and Carbides 397

Under Case 2 we still have to deal with the following possibility: if


the alloy is saturated with graphite, then no back-diffusion of carbon to
the interior of the alloy can take place because there is no carbon activity
gradient. Accordingly, a separation of carbon must occur, which leads to a
slight decrease in the oxidation rate for a finely divided precipitate since
the effective metal surface is somewhat reduced. With sufficiently rapid
metal diffusion, a decrease in the rate of oxidation is generally not observed
until higher carbon enrichments become noticeable at the alloy/oxide phase
boundary. At present no example of this exists.

!OOr----.----.----,----~--~----~

g/cm2
80~---+----+_--_1----1_~~--~~

.,t~~-+--~~~~r--+--~
~
~I~
~--~L-~+---~---+--~--__i

Fig. 214. Course of the oxidation


of manganese and a Mn-C alloy J~~-4----+_---4----+_--_+--~
with 1.33 wt. % C at lOOO°C in
oxygen of 1 atm pressure, according
to Webb, Norton, and Wagner.
(0: L,mjq for pure manganese, 200f}{/ tKJlJ(J(J G{/O{/{/ 801J(J(J Ifl(JIJ(J(J sec
0: L,mjq for Mn-C). Duration of Oxidation_

The diffusion of carbon through the oxide layer is to be regarded as a


further possibility under Case 2, if carbon is sufficiently soluble in the oxide.
The presence of carbon in the oxide can now reduce or increase the con-
centration of the ionic lattice defects, which can ultimately effect a reduction
or an increase in the oxidation rate. We have to deal with a mechanism of
this type in the oxidation of TiC, since TiC and TiO are soluble in one another
to a certain degree. At this time neither data nor information regarding
the mechanism of the solubility of carbon in TiO is known.
According to Schwarzkopf and Kieffer,l the titanium-carbide system
possesses a narrow region of solubility for carbon with a very stable carbide,
TiC. The titanium-oxygen system exhibits a broad IX-titanium region
below 900°C (Fig. 92) and above 900°C a narrow .s-titanium region, which
broadens with increasing temperature. It was mentioned earlier, in Section
4.2.2.2, that the course of the oxidation for short experimental times follows
a parabolic rate law with oxygen uptake by Ti and Ti0 2 formation between
1 Schwarzkopf, P., and R. Kieffer: Refractory Hard Metals, New York, 1953, p. 83.
398 6. Oxidation Mechanism of Metal-Carbon Alloys and Carbides

900 and 1000'°C in oxygen or air at 1 atm. Simnad and co-workers! have
determined the rate of oxygen uptake and particularly the rate of Ti02
formation.
For the testing of the above mechanism, titanium and TiC or a carbon-
poor carbide of the composition TiCo.63 was oxidized up to 50 hI' at 900 and
1000°C in oxygen. The course of the oxidation with time is given in Fig.
215. Neither titanium nor carbon was preferentially oxidized. The progress

10 10" lOS sec


Duration of Oxidation'_

Fig. 215. Course of the oxidation of titanium and titanium


carbide at 'OOO°C or 900°C in oxygen of 1 atm pressure in
a log-log pI ,t, according to Webb. Norton, and Wagner.
Results evah.llted according to equation (6.11) or (6.12).
Ti lOOO°C • D
TiCo.6a lOOO°C A t6.
TiC 1000°C • 0
TiC 900°C T V

of the oxidation as well as the weight change can be calculated from (6.11)
and from the CO 2 evolution according to (6.12). As can be seen from Fig. 215,
the values obtained in both ways agree. While the rates of oxidation of Ti
and TiC o.63 are equal, the oxidation of TiC proceeds more slowly. Deviations
from the parabolic rate law appeared in all experiments. The oxide layer
which was formed under these experimental conditions consisted only of
Ti0 2 , but with oxidation times of one week a TiO layer could also be ob-
served. After oxidation for 24 hI' at 1000°C, TiC forms two oxide layer
regions with the rutile structure-the outer region consisting of large crystals
and the inner region being made up of small crystallites with many pores.
The appearance of such a porous zone was not compatible with an exclusive
oxygen diffusion over vacancies in the Ti02. Obviously, a vacancy precipita-
tion takes place at the grain boundaries of the forming oxide layer.
The existing results do not yield a clear picture concerning the possibility

1 Simnad, M., A. Spilners, and O. Katz: J. Metals 7,645 (1955).


6. Oxidation Mechanism of Metal-Carbon Alloys and Carbides 399

of carbon diffusion through the oxide layer. Kinna and Rudigerl tried to
induce a detectable carbon diffusion via vacancies in Ti0 2 by increasing the
number of the vacancies with NiO additions. The results do not permit any
definite conclusions. The results of the chemical analysis, however, do
indicate a diffusion of carbon through the oxide layer because a carbon
content of 0.05% was discovered to be present in the outer zone of the Ti02
layer.
According to Kinna and Rudiger, at 1Ooo°C the rate of oxidation of TiC
is practically independent of the oxygen pressure, and nitrogen, which is
present at the same time, also has no effect. A parabolic oxidation was
observcd with pure TiC and with TiC-Co alloys with 18 wt.% Co (see Fig.
104). Structure investigations by means of X-ray and electron diffraction
procedures yielded a CoO or a C0304 formation on the oxidized hard-metal
alloy for oxidation times up to 5 hI' between 350 and 1000°C; Ti0 2 formation
could not bc observed until after longer oxidation times (greater than 20 hr).
Furthermore, at loo0°C the oxidation rate of the hard metal was somewhat
greater than that of pure TiC and Co. Even a hard-metal alloy with 20% Ni
oxidized at approximately the same rate as TiC.2 The commercial hard
metals, which are basically TiC with additions of tantalum and niobium,
yielded a decrease in the rate of oxidation of about a factor of 4 compared
to pure TiC at 1000°C. This decrease is understandable if one considers
(as was explained in Section 4.2.2.2 for titanium-niobium alloys), that the
pentavalent niobium ions entering into the oxide layer decrease the con-
centration of oxygen ion vacancies. Roach 3 found, on the basis of oxidation
experiments with TiC-Cr alloys between 650 and 1400°C, that an addition
of 5 wt. % Cr yielded a maximum reduction in the oxidation rate, although
only by a factor of about two. Roach also observed that the lowest oxidation
rate for the above-mentioned alloys was to be found at 800°C. Hinnuber and
Rudiger4 (Fig. 216) reported a similar observation for TiC-Cr3C2-Co alloys.
8,---~~--~-----,

gfcmZ

J 4<f---f---l--+-c1---l
~
~}:,.
2~----L---~+---~
Fig. 216. Temperature dependence of the rate of
oxidation of TiC-CraC2-Co alloys, according to Hin·
nuber and Rudiger. The weight increase after 2 hr of
oxidation of (I) TiC + 5% Cr2C2 + 6% Co and (2)
TiC + 7.5% CraC2 + 6% Co is plotted.
1 Kinna, W., and O. Rudiger: Arch. Eisenhutlenw. 24, 535 (1953).
2 Webb, W. W., J. T. Norton, and C. Wagner: J. Electrochem. Soc. 103, 112 (1956).
a Roach, J. D.: J. Electrochem. Soc. 98, 160 (1951).
4 Hinnuber, J., and O. Rudiger: Arch. Eisenhutlenw. 24, 267 (1953).
400 6. Oxidation Mechanism of Metal-Carbon Alloys and Carbides

The increase in the scaling stability of the hard metals with tungsten
carbide additions can be seen in Fig. 217. However, the additions should
be limited to 20% WC, since with higher contents the scaling stability

50

...,

~ '" -- .f K/
WC-Content .. r -
I Fig. 217. Rate of oxidation of20% Co +
(80-x)% TiC
+ x% WC alloys at lOOO°C as a function of WC
t.f "A> 20 content (x is in wt. %), according to Hinniiber and
Riidiger.

decreases again. As further experiments showed, the cobalt content in


TiC-CraC2-WC-Cq alloys should not be smaller than 10%. Vanadium carbide
as well as M0 2C additions to TiC-Co alloys considerably impair the resistance
to oxidation, as was indicated by experimental results on an alloy of TiC
plus 6% Co with addition of VC and M0 2C (Fig. 218).

'2.h-1 /"
4<

If /
J

I /
li
2
/
V
IjI 1
Fig. 218. Influence of VC and 1\-I02C additions on the rate
of oxidation of TiC-Co alloys at lOOO°C, according to
'() 1fl 20 .JO f070S0 Hinniiber and Riidiger: (1) 6% Co + (94 - x)% TiC
VC- or MozC-.Content ~- + x% VC and (2) 6% Co + (94 - x)% TiC + x% 1\-I02C,

According to the present state of the investigations and in agreement


with the theoretical considerations based on the titanium oxidation, addi-
tions of the carbides of tungsten, tantalum, and niobium to TiC or TiC-Ni
or Ti-Co alloys may produce the most effective decrease in the oxidation
rate of hard-metal alloys.
Bohnenkamp and Engell studied the mechanism of oxidation of carbon-
iron and carbon-nickel alloys.l After longer reaction times, the rate of carbon
oxidation prevails and the surface region of the alloy is decarburized. At
1 Bohnenkamp, K., and H. J. Engell: Arch. Eisenhuttenw. 33,359 (1962).
6. Oxidation Mechanism of Metal-Carbon Alloys and Carbides 401

950 and 1050°C, after a short oxidation period, the diffusion of carbon in the
alloy toward the surface becomes rate-determining.
Further investigations of more basic problems are needed for complete
explanations of these complicated processes which determine the course
of the oxidation.
7. The Mechanism of Oxide Layer
F ormation in Aqueous Electrolytes

If metals such as iron, nickel, zinc, and copper are placed in aqueous
electrolytes, then a variety of phenomena can be observed in the attack
by the electrolyte ions and the oxidizing gases which are dissolved in the
electrolytes. If the electrolyte contains only the ions of the metal immersed
in it, a metal ion potential E can be calculated from the standard potential
Eo (identical with the reaction energy under standard conditions) and the
respective metal ion concentration c by means of the Nernst equation:

where z is the valence of the metal ion and F = 96,500 coulombs. We obtain
an equilibrium state which is characterized by the fact that an equal number
of ions enter and leave the metal per unit time and surface area. Thus, in
the equilibrium state the metal is not affected by the electrolyte.
The situation is completely altered if an electric field is applied with the
positive pole on the metal side in the arrangement of an electrochemical
cell or if an electrolyte with "oxidizing properties" is chosen. In this case,
the equilibrium state is disturbed in the sense that for the attainment of
this equilibrium so many ions must be sent from the metal into the solution
that practically no more metal remains, i.e., the metal dissolves. This solution
process is called wet corrosion. The transition of the metal ions from the metal
phase into the electrolyte must be associated with the transport of electricity
because of the electric charges on the metal ions, so that an anodic current
must enter through the phase boundary. This anodic current, which depends
on the size of the surface and is thus given in amperes per square centimeter,
can be equal to the total current in a suitable arrangement (electrochemical
cell), in the absence of any other electronic change in charge process at the
metal surface (for example, in the presence of a redox-electrolyte). However,
since additional charge processes frequently appear, one has to deal rather
with a partial current density in corrosion. As is known from countless
experiments, this anodic current density is a function of the externally
402
7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes 403

applied potential. From the course of the current density-potential curve


one can study the extent of the metal dissolution. The externally applied
electric field, which depresses the electron level or the chemical potential
of the electrons in the metal, is removed and an oxidizing electrolyte intro-
duced. Because the oxidizing electrolyte has a high electron affinity, the
electrons are taken from the metal phase, thus causing the electron level
to decrease. Then dissolution of the metal will begin, which now is a function
of the redox-potential of the electrolyte.
On basis of thermodynamic considerations one should expect that those
metals whose standard potentials are displaced furthest to negative values
will dissolve most rapidly and most easily. As investigations have shown,
this is actually the case with the alkali and alkaline earth metals, but large
deviations appear with most of the other metals. For example, aluminum,
which has a much more negative (that is, less noble) standard potential, is
considerably more corrosion resistant than, for example, iron, which has a
considerably more noble standard potential. The reason for this behavior,
which on basis of only thermodynamic considerations is quite unexpected,
is to be sought in a secondary reaction-formation of a protective layer-
with which we will deal in detail since it seems to lead to possibilities of
combating the undesired corrosion of metals.
The first task is to determine the processes that lead to the formation
of a protective layer, as far as that can be done within the bounds of present
knowledge. Clearly, not all the phenomena that occur in the corrosion of
metals and alloys fall within the scope of this work. Rather, we shall con-
centrate on two problems which are of considerable theoretical significance
and form a useful starting point for the investigation of all other corrosion
phenomena. These are the mechanism of the formation of a pore-free pro-
tective layer (e.g., oxide film formation) and the mechanism of the solution
or corrosion current of a metal covered with an oxide layer. Not all processes
that lead to the formation of a porous coating and cause, e.g., the typical
rust process in iron, will be dealt with, since many are very complicated and
at present not sufficiently understood. These processes were all discussed in
Metallic Corrosion, Passivity, and Protection, a classic work by Evans,l
who must be credited with introducing, by his numerous contributions, a
certain order and system into the jungle of corrosion and rust processes. Uhlig's
work 2 may be singled out in the American literature and Masing's in the
German. 3
As noted above, we will discuss a few selected examples of the mechanism
1 Evans, U. R.: Metallic Corrosion, Passivity, and Protection, second ed., London, 1948.
2 Uhlig, H. H.: The Corrosion Handbook, New York, 1948.
3 Masing, G.: "Theorie der Korrosion," in Die Korrosion metallischer Werkstoffe, Vol. 1,
edited by O. Bauer, O. Krohnke, and G. Masing, Leipzig, 1936.
404 7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes

of formation of a coating on metals. For our special problem, we will refer


primarily to the works of Bonhoeffer1 and Evans and co-workers. 2 Apart
from the guiding works by Evans and the valuable contributions by Uhlig,
extensive experiments were carried out by Bonhoeffer and co-workers on
the formation and structure of the passive layer, especially on iron and on
the transport mechanism in passive layers. Schottky,3 in a joint discussion
with Vetter,4 was able to report remarkable results on the mechanism of the
solution current in iron covered with an oxide film.s
Obviously, we begin with the mechanism of the oxide film formation
on metals, which in most cases causes the passive state. For this reason we
will no longer talk of oxide films but of passive layers.

7.1. The Phenomenon of Passivity


If metals with coating.free surfaces are placed in an oxidizing electro-
lyte or an electrolyte containing a reacting gas, e.g., O2 or N0 2 , then one often
observes, after an initially rapid reaction, a substantial decrease (up to
several orders of magnitude) in the reaction rate. Metals that have attained
such a state are called passive, and the phenomenon itself is called passivity.
We shall now explore the ways in which this passive state can be attained
in a metal or alloy. In order to directly relate the results of low-temperature
oxidation in Section 3.5.4, we choose an oxidizing electrolyte, e.g., moderately
concentrated nitric acid, which possesses a sufficiently high oxygen chemical
potential for the passivation process. This reaction is initiated by chemi-
sorption of the oxygen, which we may write as
HNO~l) + e(Me)....-- 0-(0) + HNO~) (7.1 )
This chemisorption produces an electric double layer of high field strength
which the system tries to lower by either "pulling" metal ions from the
metal surface or "pushing" oxygen ions into the metal lattice. In both
1 Beinert, H., and K. F. Bonhoeffer: Z. Elektrochem. 47, 441, 536 (1941); K. F. Bon-
hoeffor: Naturwis8. 31, 270 (1943); K. F. Bonhoeffer, E. Brauer, and G. Langhammer:
Z. Elektrochem. 52, 29 (1948); K. F. Bonhoeffer, V. Haase, and G. Langhammer:
Z. Elektrochem. 52, 60 (1948); K. F. Bonhoeffer, and G. Langhammer: Z. Elektrochem.
52,67 (1948); K. F. Bonhoeffer, and K. J. Vetter: Z. physik. Chem. 196, 142 (1950);
K. F. Bonhoeffer: Z. Metallk. 44,77 (1953); and their further references.
2 Evans, U. R.: frletallic Corrosion, Passivity, and Protection, second ed., London, 1948.
3 Schottky, W.: "Passivitat und Losungsstrom," in Halbleiterprobleme, Vol. 2, p. 233,
edited by W. Schottky, Braurtschweig, 1955.
4 Vetter, K. J.: "trber den Mechanismus der Passivschichtbildung," in Pa8sivierungs.
und Anlaufvorgiinge an Metalloberfliichen, edited by H. Fischer, K. Hauffe, and W.
Wiederholt, BerlinJGottingenJHeidelberg, Springer, 1956.
5 Vetter, K. J., and K. Arnold: Z. Elektrochem. 64, 244 (1960).
7.1. The Phenomenon of Passivity 405

these cases the material transport leads to oxide formation until, because
of the newly chemisorbed oxygen ions, the surface is covered with a compact
oxide layer so that ions required for further oxide formation would have to
diffuse through the lattice of the oxide film. The potential which determines
the electric field remains practically constant during the course of the
reaction, but the transport distance (oxide film thickness) continually
increases. Consequently, the electric field strength (f = V/~ decreases
continually until it reaches a critical value at which, even with the kinetic
energy available at room temperature, significant transport of ions through
the oxide layer can no longer be induced. The reaction would practically
cease at this point, as was noted in Section 3.5.4, if it were not for the fact
that in the case of oxide formation in aqueous electrolytes another process
takes place-dissolution of the oxide film in the electrolyte. After the stable
passive layer thickness has been attained, this process determines the rate
of the further progress of the reaction, i.e., the corrosion rate. This corrosion
or solution current will be discussed later.
If iron or nickel is immersed in such an oxidizing electrolyte under
experimental conditions suitable to the mechanism sketched above, fairly
strong passivity is observed. This is related to the presence of a thin oxide layer
of 30-80 A thickness, as can be concluded from the work of Evans l and the
corroborating studies of Tronstad,2 Todt,3 and Bonhoeffer. 4 Since oxygen
produces oxide films at room temperature on all metals (except gold) there
is no reason why one should not expect oxygen-containing electrolytes to
cause oxide-film formation and thus passivity to some extent.
Before we discuss the formation of passivity-inducing oxide films in
detail, we shall take note of a second passivation process, which, unlike
the mechanism mentioned above, does not seem to be based on a protective
layer formation, but due rather to chemisorption of a suitable reactant
in the electrolyte. The passivity caused by chemisorption, which was described
by Uhlig,5 seems to be applicable in a few cases, and may be frequently
regarded as the decisive inhibiting process. 6 It is not our intention here to
debate the range of applicability of this mechanism; rather we will limit
ourselves to the description of this "chemisorption mechanism of passivity,"
using a particularly suitable example.
1 Evans, U. R.: J. Chem. Soc., (1930), 482; u. R. Evans, and 1. D. G. Berwick: J. Chem.
Soc. (1952), 432.
2 Tronstad, L., and C. W. Borgmann: Trans. Faraday Soc. 30, 349 (1934).
3 T6dt, F.: Z. Elektrochem. 55, 331 (1951).
4 Bonhoeffer, K. F.: Z. Metallk. 44, 77 (1953), including a bibliography covering the
work of U. F. Franck, K. J. Vetter, and others.
5 Uhlig, H. H.: Ann. N. Y. Acad. Sci. 58, 843 (1954).
6 Elze, J., and H. Fischer: Metalloberfliiche (A) 6, 177 (1952); T. P. Hoar and R. D.

Holliday: J. Appl. Chem. 3, 502 (1953).


406 7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes

Uhlig correctly traces the passivation effect of CO in HCI electrolytes


on 18-8 steels solely to the chemisorption of CO. As we know from Suhrmann's
work,! the chemisorption of CO on metals is accompanied by an electron
interaction. In contrast to the chemisorption of oxygen according to (7.1),
here there is only the formation of an electrical double layer, with the
further difference that in this case the electrolyte side is positively charged
and the metal side negatively charged. Nothing can be said at this time
about the magnitude of the electric potential. It can be concluded that the
chemisorption of CO becomes energetically more favorable, or the surface
concentration of the chemisorbed CO molecules becomes greater, as the Fermi
potential decreases. It should be possible to check these considerations with
chemisorption measurements of CO on palladium-gold alloys. The d-band
of palladium is short 0.6 electrons per atom and therefore a high chemi·
sorption of CO is to be expected. The chemisorption of CO should be depressed
more and more with increasing additions of gold, since gold with its filled
d·shell donates electrons to the palladium, filling the d-band in palladium
and thus changing the chemisorption equilibrium to desorption because of
the increase in the metal· electron concentration. The variation of the para·
magnetic susceptibility of palladium-gold alloys with increasing gold con-
centration described by Vogt 2 and the corresponding variation of the activa-
tion energy of the ortho-para hydrogen conversion noted by Couper and
Eley 3 can be regarded as indirect confirmation of these ideas. The results
of these investigations are presented in Fig. 219. From the experimental

~hl/MoL ~
~

l
\
\
\
\
\ ,
,, II Fig. 219. Activation energy of the ortho-para
hydrogen conversion on gold-palladium alloys as a
o ... function of the alloy composition, according to
2
0
......, Couper and Eley. The dashed curve represents
Jd 20 *0 60 80 Au the course of the paramagnetic susceptibility in
Atom-%Au- arbitrary units, according to Vogt.

results for magnetic saturation and specific heat, which is determined by the
electronic structure, as a function of the composition of copper-nickel
alloys4 (Fig. 220), we can conclude that the mechanism of chemisorption

1 Suhrmann, R.: Z. Elektrochem. 56, 351 (1952).


2 Vogt, E.: Ergeb. exakt. Naturw. 11, 321 (1932).
3 Couper, A., and D. Eley: Discussions Faraday Soc. 8,172 (1950).
4 Mott, N. F., and H. Jones: Theory of Properties of "lIetals and Alloys, Oxford, 1936,

p.199.
7.1. The Phenomenon of Passivity 407

and the passivation mechanism are the same for carbon monoxide as for
oxygen. Further experiments in this direction would be useful. Organic
molecules in the electrolytes should have a similar effect on the resistance
to corrosion of metals and alloys, if an electron interaction is possible, which
is actually the cause of chemisorption.

Fig. 220. Dependence of the magnetic


saturation moment and of the specific
heat of the electrons in copper-nickel
alloys on their composition, according
to Mott and Jones. Above 60 at.% eu
the d·band is completely occupied.

The question now arises whether, because of the high electric fields in
this case, ions are transported in a manner similar to the chemisorption
of oxygen, so that we have to reckon with a formation of a coating of a
carbonyl compound. The answer to this question is not simple, since there
are metals, for example, nickel and iron, which form quite stable carbonylsl
with CO, and others where a carbonyl formation has not been observed.
In the first casc, we have to consider primarily the formation of a carbonyl
layer, even if the actual existence of such coatings at room temperature is
still in question. The case of an ideal chemisorption can then be obtained
to a good approximation only if the chemisorbed molecule, for example, CO,
cannot penetrate the metal lattice and if at the same time the potential
difference in the electrical double layer is not high enough to cause a significant
transport of the metal ions in proximity to the chemisorbed ions. Here,
naturally, a monomolecular covering of the surface can occur in the most
favorable cases, so that we obtain an effective passivity as defined by
Uhlig. 2
In this connection, the question appears justified whether one should
consider the phenomenon of chemisorption of CO, a component which is
not involved in the reaction, which leads to corrosion limitation, as inhibi.
tion. As we know, in the case where chemisorption acts almost entirely
alone, without any subsequent reaction (ion transport or formation of a
protective layer), passivity and inhibition are indistinguishable. Uhlig
made particular reference to these processes in his explanation of passivity,
1See for example W. Hieber: Angew. Chern. 67, 211 (1955); see also the references.
2Uhlig. H. H.: Z. Elektrochem. 62. 626 (1958); in this connection the treatment of
N. Hackerman is of interest: N. Hackerman: Z. Elektrochem. 62. 632 (1958).
408 7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes

especially with transition metals, and so they will not be treated further here.
However, the fact that Uhlig! questions the chemisorption mechanism of
steel and titanium for the oxygen passivation cannot be passed without
question. The "adsorption measurements" as expressed by the Langmuir
equation referred to in support of his theory-quite aside from the fact that
here we certainly are dealing with predominant chemisorption and not with
adsorption 2-are not conclusive, as the steel and titanium which were used
evidently always showed a thin oxide film, so that the measurements by
Uhlig probably were carried out on the oxide film and not on the alloy or
titanium surface. In a work on oxygen consumption during the passivation
of 18-8 steel, Uhlig himself calculated a layer thickness of 17 A on Ithe
assumption of a Cr20a layer. a The author is of the opinion that the oxide
film and the chemisorption theory represented two limiting possibilities
in the explanation of the passivation mechanism of metals and alloys, and
that in any particular case one has to critically check which of the two
processes-oxide film formation or chemisorption-causes the passivity.
As Bonhoeffer and co-workers, among others, have shown, prior to
formation of the effective passive film, a porous coating is formed which
does not have to be chemically identical with the subsequent oxide film.
This primary protective layer occupies-in agreement with earlier ideas of
Miiller~the largest part of the surface and causes an increase in the true
current density in the pores of this coating. In this stage of the reaction
the iron still remains in its active state, and formation of the actual passive
oxide layer does not set in until after a critical value in the current density-
potential curve is attained, which is characterized by a new electrode process.
This compact oxide film then extends over the total metal surface replacing
the first porous protective layer. A quantitative consideration of this by
Franck5 furnishes the relationship between the passivation current density i
and the flow time 'Tp of this current required in the passivation:
(7.2)

where io is the equivalent current density in the corrosion of the isolated


first coating 6 (for the concept of the equivalent current density see Section
7.3). According to Franck, the symbol K indicates the quantity of electricity
passed through the first porous protective layer before the onset of passive
layer formation.

1 Uhlig, H. H., and A. Geary: J. Electrochem. Soc. 101, 215 (1954).


2 Hauffe, K.: Angew. Chem. 67,189 (1955).
3 Uhlig, H. H., and S. S. Lord, Jr.: J. Electrochem. Soc. 100, 216 (1953).
4 Muller, W. J.: Z. Elektrochem. 30,401 (1924); Wiener Monatsh. 48, 559 (1927).
5 Franck, U. F.: Z. Naturforsch. 4a, 378 (1949).
6 Bonhoeffer, K. F., and U. F. Franck: Z. Elektrochem. 55, 180 (1951).
7.2. Passive Layer Formation on Metals and Alloys 409

The phenomena discussed here, which by suitable experimental con-


ditions can be considerably suppressed in other metals, complicates the
kinetic treatment of passive layer growth on iron.

7.2. The Mechanism of Passive Layer Formation on Metals and Alloys


In addition to the detailed investigations of Bonhoeffer and co-workers
and of Evans and his students on the mechanism of the passivity of iron,
Giinterschulze and Betz1 as well as Verwey 2 described the phenomenon of
passivity on aluminum, which was evaluated by Cabrera and Mott. 3
Vermilyea,4 Torrisi,S and Dewald6 concerned themselves with the mechanism
of anodic oxide film formation on tantalum. 7 Dewald succeeded in explaining
oxide film formation by means of the theory of field transport in space-charge
boundary layers. Since these considerations, which were applied to Ta20s film
formation, are of general significance, we will discuss them in detail here.
In contrast to the attack of oxygen on metals, in the attack by neutral
and oxidizing electrolytes (aqueous or nonaqueous) we have to consider an
additional process-namely, the solution of the reaction products (e.g.,
oxides) in the electrolyte. Furthermore, there are positively and negatively
charged ions in the electrolyte, which exert an influence on the formation
and the dissolution of the passive layers and further complicate the reaction
mechanism. This is a phenomenon which does not occur in protective layer
formation with gases. For each case one has to determine whether the
structure of the electrolyte in the neighborhood of the passive layer sUliace
is significant for the kinetics of the passive layer formation-whether it is
anodically charged or in redox electrolytes. Incidentally, it may be noted
that the redox systems which frequently appear in the electrolytes are
basically comparable to the gaseous systems. This is especially true in the
treatment of the electronic equilibrium between the solid phase and the
electrolyte. In the absence of a redox system in the electrolyte, a passive
layer can only be produced when a sufficiently strong external electric
field is applied to the metal, which can be achieved with a suitable electro-
chemical cell with the plus pole at the metal.
1 Giinterschulze, A., and H. Betz: Z. PhY8ik. 91, 70 (1934). See also W. Ch. van Geel:
Phyaica 17, 761 (1951).
2Verwey, E. J. W.: Phyaica 2, 1059 (1935);.Afdel. Natuurk. Koninkl. Ned. Akad.
Wetenachap. No.7, p. 97 (1953).
3 Cabrera, N., and N. F. Mott: Rept. Progr. in PhY8. 12, 163 (1949).
4 Vermilyea, D. A.: Acta Met. 1,282 (1953); 2,482 (1954).
5 Torrisi, A. F.: J. Electrochem. Soc. 102, 176 (1955).
6 Dewald, J. F.: J. Electrochem. Soc. 102, 1 (1955).
7 Young, L.: Proc. Roy. Soc. (A) 244, 41 (1958); A. R. Bray, P. W. M. Jacobs, and L.
Young: Proc. PhY8. Soc. 71, 405 (1958).
410 7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes

Since the formation of a passive layer belongs to the domain of space-


charge layers near phase boundaries, where no quasi-neutrality of charge
carriers, i.e., ions or electrons and holes, can be expected either in the
layer itself or beyond it to a maximum depth of 10-6 to 10-3 cm, it seems
advisable to assume, in principle, that electrostatic fields and space charges
are present on both sides of the passive layer/electrolyte phase boundary.
These field regions-in the passive layer as well as in the electrolyte-in-
fluence each other in such a way that the reactions in one field region deter-
mine the position of the other. Therefore, in order to formulate a generally
valid theory of passive layer formation, we have to take into consideration,
in addition to the phase boundary reactions, the processes in both field
regions or space-charge zones. Suitable systems for study of the passive layer
formation include iron in oxidizing electrolytes, especially in concentrated
HN03; nickel in NiS04 or H 2S04 solutions; and tantalum in aqueous H 2S04
or H 3P04 or nonaqueous (in glycol) Na2B407 solutions. Of these three metals,
nickel and tantalum seem to be particularly suitable since, certainly in the
case of nickel and most probably in the case of tantalum, the passive layer
that appears consists of only a single oxide (NiO and Ta20s, respectively).
In the case of passive layer formation on iron, the structure of the passive
layer cannot be predicted. Here the passive layer can be formed primarily
of higher-valent (Fe203) as well as lower-valent oxides (Fe304, FeO), accord-
ing to whether the electron supply or the ion transport is the rate-determining
factor, as was cxplained for the low-temperature oxidation of copper in
Section 3.5.4.

7.2.1. The Theory of Passive Layer Formation on Metals


As we have noted earlier (in Section 3.5.4), Cabrera and Mott were the
first to have recognized that the reciprocal logarithmic rate law derived
by them for "dry oxidation" can also be utilized for an explanation of the
anodic film formation with time, as it applies to passivation processes_ They
simplified their theoretical development in that they assumed a space-charge-
free oxide layer. Furthermore, the film growth rate [the slope from Tafel's
plot for dE/d(ln i)] is then proportional to the absolute temperature. The
experimental data in the Ah03 film formation on aluminum in aqueous
solution may be adequately treated by means of this simplified theory, and
it is quite reasonable to apply these relations to the growth rate of the passive
layer on iron. A formal agreement between the theoretical expressions and
the experimental results obtained by Vetterl can also be demonstrated in this
case, as was shown in Section 3.5.4, but these and subsequent considerations
1Vetter, K. J.: Z. Elektrochem. 58, 230 (1954); see also K. J. Vetter: Z. Elektrochem.
62,642 (1958); N. D. Tomaschow: Z. Elektrochem. 62,717 (1958).
7.2. Passive Layer Formation on Metals and Alloys 411

admit no definite conclusions relative to the structure of the passive


layer on metals such as iron, which may have several oxide phases, since the
kinetic equations are inconclusive in this respect. Furthermore, at present
it is not possible to identify the rate-determining step from among the
following: (1) entry of a chemisorbed oxygen ion into a vacancy in the
lattice of the passive layer, (2) removal of a metal ion from the passive
layer into the chemisorption layer, (3) formation of an oxygen ion vacancy
by transfer of an oxygen ion onto the surface of the metal, (4) transfer
of a metal ion from the metal surface in an interstitial lattice position into
the passive layer lattice, or (5) ionic transport through the passive layer.
The conditions discussed here for an n-conducting passive layer are also
valid in a corresponding manner for a p-conducting passive layer, as, for
example, NiO on nickel.
A two-layer hypothesis for passive layer formation on iron was
proposed for discussion by Hauffe.1 We will discuss this later. The fact that
such a two-layer structure (e.g., Fe304/Fe203 or FeO/Fe304) is to be seriously
considered is also the result of a discussion by Weil and Bonhoeffer 2 on
reduction of the passive oxide. Here it was assumed on basis of the experi-
mental results that cathodic reduction generally leads not to a bare metal
but to a lower oxide (for example, FeO). Since, at present, our uncertainty
concerning tpe structure of the passive layer on iron is the primary obstacle
preventing a quantitative formulation of the reaction mechanism, we will
confine our considerations to the passive layer formation on tantalum and
aluminum. Quantitative relationships on the "interplay" of spaee charges in
an ionic crystal and in an electrolyte were probably given first by Grimley
and Mott. 3 The use of their considerations in the Ag/AgBr/electrolyte
system are based on the assumption that the silver halides formed on silver
even at 20°C are pure ionic conductors, which seems to be allowed, in agree-
ment with measurements by Klein and Matejek4 and theoretical considera-
tions by Wagner. 5
Under the assumption that the transport processes in the electrolyte
space-charge zone are rapid, we will discuss the mechanism of passive layer
formation, in support of the presentation by Dewald6 and by Young,7
which, unlike the theory of Cabrera and Mott, does take the influence of
the space charge into consideration. The temperature independence of the
1 Hauffe, K.: Werk8toffe u. Korrosion 6,117 (1955).
2 Weil, K. G., and K. F. Bonhoeffer: Z. physik. Chem. [NF] 4,175 (1955).
3'Grimley, T. B., and N. F. Mott: Discus8ion8 Faraday Soc. I, 3 (1947); T. B. Grimley:
Proc. Roy. Soc. (A) 201, 40 (1950).
4 Klein, E., and R. Matejek: Z. Elektrochem. 61, 1127 (1957).
5 Wagner, C.: Z. Elektrochem. 63, 1037 (1959).
6 Dewald, J. F.: J. Electrochem. Soc. 102, 1 (1955).
7 A. R. Bray, P. W. M. Jacobs, and L. Young: Proc. PhY8. Soc. 71, 405 (1958).
412 7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes

film growth rate found by Vermilyea, which could not bc explained by


the Cabrera-Mott theory, can be dealt with successfully using these extended
concepts.
For the rate.determining factor, discussed in Section 3.5.4,
exp -(Ut-ziea*fi)/kT (7.3)
where Zie is the charge of the transported ion and a* the distance from the
lattice position of a metal ion on the surface of the metal to the first inter-
stitiallattice position in the oxide film (see Fig. 43), we can determine the
following two limiting cases:
1. If U 1 - ziea*(f is greater than 40kT, then the reaction comes
practically to a standstill.
2. If, on the other hand, U1 - ziea*(f is about 20kT, then the penetra-
tion of ions takes place very rapidly.
Thus, in the first case a measurable reaction can only proceed when we
have very strong electric fields. However, if U1 is much smaller or, at most,
equal to 20kT, then weak electric fields are sufficient to effect a rapid penetra-
tionof the ions. It is obvious that an oxide film formation cannot appear if
ions which easily change valence exist in the electrolyte, because the electrical
potential required for oxide film formation cannot exist. In this case the metal
behaves as an unattackable electrode which causes only an electron exchange
of the ions in the electrolyte.
According to the Cabrera-Mott assumption-a limiting process at the
metal/passive layer phase boundary-one should assume that any ion which
has once entered the passive layer lattice immediately migrates to the passive
layer/electrode phase boundary. That is one of the possibilities which was
mentioned earlier. We can now show-independent of which is the effectively
slow process-that the migration of the io~ or~the transport flow ji in
particles/cm 2 -sec, which is composed of ji = ji + ji, is equal to

This relationship developed by Mott is valid for the case of~ a weak
electric field. At high electric fields the partial transport flow ji (back-
ward) is practically zero, and we obtain the expression

as was derived for passive layer formation on aluminum.! Here


rt.' = rt.vo exp( - [hkT) and f3 = Zia/O. (See Fig. 43 for a definition of U 2 .)
1 Hoar, T. P., and N. F. Mott: J. Phys. Chern. Solids 9,97 (1959) give a new mechanism
for tho corrosion rate of aluminum whieh provides plausible explanation for the experi-
mental results.
7.2. Passive Layer Formation on Metals and Alloys 413

We now consider the part ofthe film which grows at a constant rate. This
steady state requires that the ion concentration at any place remain con-
stant with time and therefore that

(7.6)

By use of equations (7.5) and (7.6) it follows that:

(on)
og
+ ntp ( Off)
t og t
= 0 (7.7)

where nt is the concentration of ions in interstitial lattice positions.


Furthermore it follows with aff/ag = 47Tzient/e,

(dn)
-
dg Stationary
= - pyn~ (7.8)

where y = 47TZte/e. Integration of (7.8) gives the concentration distribution


of the interstitial lattice ions in the passive layer (of course, only when this
is homogeneous):
no
n"=----
1 + Pynog
(7.9)

where no is the concentration of the interstitial lattice ions at the


metal/passive layer phase boundary (for example, Ta/Ta20s), therefore
at g = O. Inserting equation (7.9) into the Poisson equation for the electric
field at an arbitrary location g, we obtain

(7.10)

As can be seen from equation (7.10), the field at any location on the oxide film
is determined by two electrical components: by a contribution of the field
strength ffo, which is determined by surface charges, and by (/1ynog), where
the space charge is to be regarded as determinative. Equation (7.10), which
is of considerable interest, is to be inserted in (7.5) when the "surface-space-
charge mechanism" should be demonstrated in the passive layer formation.
However, as soon as we introduce the Cabrera-Mott assumption (ionic
transfer at the metal/passive layer phase boundary), the contribution of the
space charge to the electric-field effect ceases. Dewald thus was able to
illustrate that the existence of strong electric fields could very possibly
shift the rate-determining step from the phase boundary to the interior of
the passive layer. This is reasonable when one considers that in the presence
of higher electric fields the expressions (Ul - ztea*ff) and (U2 - ~ed) are
414 7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes

determinative for the transport rather than the quantities U I and U2 •


If now a* is greater than a, then the reduction of the saddle height UI is
greater than the ion transfer barrier U2 so that
(UI - Ziea*,f) < (U2 - ziea,f)
even though UI is greater than U2 in the absence of a field.
Further, when a* is less than a, the ionic passage does not seem to be
the unrestricted rate-determining step. With growing film thicknesses, that
is, with a growing number of barriers to be surmounted in the migration,
these can also be determinative compared with the high individual barriers,
U I , so that in such a case the transport through the oxide layer can become
the rate-determining step.
The component of the field strength,fo appearing in (7.10) and which is
determined only by the surface charge was obtained according to Cabrera
and Mott as
l) jo UI
,fo = - I n - - + - - (7.11)
zia* n(O)v(O)
i t
ziea*
where n(~) and v~o) are, respectively, the surface concentration and vibration
frequency of the metal ions. The constant no appearing in equation (7.9)
can be determined from the equation
(7.12)
if one sets the ionic current jo through the phase boundary equal to the
current jt in the film at the phase boundary g = 0 and if one introduces
(7.11) into (7.12). Solving for no then yields

n = (n;O) viOl) a/a· ,;(1- a/a') exp (U _ ~ U ) / k T (7.13 )


o a Vi 10 2 a* 1

In the steady state the quantity jo appearing in equations (7.11) to


(7.13) is equal to the external flow during the anodic formation of the passive
layer, and therefore is easily measurable.
An interesting relationship between the flowing material current jo
and the interstitial ion concentration no or ni follows from (7.13). Here the
stationary concentration no or ni must decrease with increasing current
density jo, when the "entrance distance" a* is smaller than the distance a
in the interior of the oxide film. However, when a* is greater than a,
ni or no must increase with increasing current density. This behavior is
conditioned by the fact that in the case where a is greater than a* an increase
in the electric field decreases the "saddle barriers" U 2 in the interior of the
oxide film to a greater degree than the "entrance" barrier U l . In this way,
with a sudden increase in the field more ions are removed from the oxide
7.2. Passive Layer Formation on Metals and Alloys '15

film than new ones can enter, so that a decreasein~occcurswith increasing


jo. In the other case, a < a*, conditions are just reversed,

(7.14)

If one uses the experimentally determinable average field strength,


where 6 is the thickness ofthe passive layer, and equation (7.10) is introduced
into !lquation (7.14) and no is replaced by expression (7.13), then we obtain
forC

~= Co + ~{(1 + ~) In (1 + 8) - I} (7.15)

where Co is given by (7.11), and S by the following expression:


(2)
4rrzte (n,vi )ala
S = flyno6 = - - -
(0) (0) *
exp
{u /
2 -(aa*)Ul}. *
jo(l-ala )6
Ell v kT
(7.16)
Equation (7.15) gives the dependence of the average stationary field
on the three variables: current density jt or jo (by Co, and no), film thickness
el. and temperature (by Co, fl, and S). The second term in (7.15) con-
siders the contribution of the space charge to the average field strength.
The magnitude of the space-charge effect is determined by the dimensionless
factor S. If S ~ 1, then only surface charges and the parameter in the
metal/passive layer phase boundary (Cabrera-Mott) are involved and
not space-charge effects. However, if S ~ 1, then both charge phenomena
are to be considered, and in the case where S ~ 1, only the space charge in
the oxide film is to be considered.
By evaluation of (7.15) for S ~ 1 and 8 ~ 1, to a good approximation
the following expressions result:
- II 4rrz~e U2
C ~ --In --eljo + -- for S~ 1 (7.17a)
z,a 27ev II Z(ea
- II jo Ul
C ~ -In - - +- for S ~ 1 (7.17b)
zta* n~O)v~Ol ztea*
A logarithmic current density dependence on the field strength is found
in both cases, which thus cannot be distinguished on this basis.
According to these relationships, developed by Dewald, we can explain
the temperature dependence of the slope of the Tafel lines reproduced in
Fig. 221 and found by Vermilyea, which is not possible according to the
416 7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes

"space. charge-free" theory of Cabrera and Mott. For this purpose we consider
again the quantity 0, which depends on the film thickness, the current
density, and the temperature. Of these, the temperature dependence is of
special significance. According to (7.16) the temperature influence depends
on the sign of the term U 2 - (a/a*) U!. If this term is negative, then 0 grows
with increasing temperature and the space charge also increases. With a
positive sign in this term the reverse is true, and with increasing temperature
the space charge becomes less significant relative to the surface charge.

Fig. 221. Temperature dependence of the


..-_A Tafel slope of a 1000-A-thick oxide film
B on tantalum at constant anode-current
density of 10-4 A/cm2. [O-measured
points by Vermilyea; curves A and B
drawn from calcluations by Dewald,
according to (7.16). For curve B, a/a*
= 1.35, a* = 3_1 and a/a*·Ul - U2 = 0_6
was chosen.]

The temperature influence on the slope of the Tafel lines may be recog-
nized from the differential form of (7.15):

(ooj_)
InJo T
= _1l_{1 + (; _ 1)ln(1 + OJ} = Tafel slope
Zia a" 0
(7.18)

The Tafel slope is influenced by the temperature in two ways: directly


through T and indirectly through a. These two influences generally work
in opposite directions, so that frequently they can virtually cancel each
other out. In light of this, the weak temperature effect found by Vermilyea
in the Ta/Ta205/electrolyte system is understandable. Young! found also
with niobium that the Tafel slope is substantially independent of tempera-
ture. These conditions are met when a = I, that is, just in the case where
surface and space charges contribute jointly to the field strength.
As we saw from the development of the equation, the magnitude of the
ratio a/a* p!n,ys a decisive role. Since this magnitude cannot always be
measured directly, a quantitative evaluation of the space-charge influence
in the formation of passive layers is still not possible. Dewald estimated
the ratio of a/a* at about 1.35 and U10r U 2 at about 1.5 or 1.4 eV from the
measurements by Vermilyea_ The distance (l = 3.1 A calculated for the
TazOs film was described by Dewald himself as being improbably high.
Young applied Dewald'" theory to the anodic oxide-film formation on
tantalum 2 and niobium. 3 The observed dependence of space charge on current
1 Young, L.: Trans. Faraday Soc. 52, 502 (1956).
2 Bray, A. H .• P. ,Yo ]\f. Jacobs, and L. Young: Proe. Phys. Soc. 71, 405 (1958).
3 Young, L.: Trans. Faraday Soc. 52, 515 (1956).
7.2. Passive Layer Formation on Metals and Alloys 417

density corresponded to the case in Dewald's theory in which the jump


distance for diffusion into the oxide is less than for diffusion through the
oxide. The data could be explained with reasonable values for all param-
eters except that the jump distances are too large (6 and 8 A). However,
it is possible that similar equations could be derived by assuming dual control
by the diffusion in the oxide and by the processes at the oxide/electrolyte
interface. The obscurity of these processes is probably the reason they have
not yet been seriously considered as a rate-determining step. It is not known,
for example, whether the oxygen for the film comes from H 20, OH-, or OH.
On basis of these and further results Young l developed a modified
theory. The reason for this modification is the experimental fact that the
transients-in particular, the "overshoot" in the field when the current is
increased-are inconsistent with the simple model involving a constant
background of immobile space charge (Dewald). The simplest explanation of
the transients is that the concentration of the mobile ions changes with a
change in the field. As postulated by Bean, Fisher, and Vermilyea 2 the
high-field production of Frenkel defects with immobile cation vacancies
and mobile interstitial ions in the oxide film give rise to "overshoot" in the
transients in the field strength which occur when the applied current is
suddenly changed. Because of amorphous oxide-film formation it is difficult
to maintain a distinction between lattice and interstitial ions, and therefore
to calculate the range of sit.e energies and jump distances.
Regardless of the magnitude of present difficulties in the use of the
Dewald or of the Young relationships, they do lead a step further in the
complicated individual processes in passive layer formation. To avoid all
misunderstanding, it should be a,dded that we have described only the forma-
tion of the passive layer with time and have not explored the corrosion
current, which is of primary interest to the corrosion chemist, and which,
in the last analysis, actually determines the rate of destruction of a metal
or alloy, in spite of a pore-free passive layer. In Section 7.3 we will explain
in detail the mechanism for the extent of the corrosion current. In the
following section, we will discuss some experimental results on passive layer
formation and a few ideas concerning the actual process.

7.2.2. Formation and Structure of the Passive Layer


In the attack of moist air on iron at 20°C Evans and Miley3 were able
to show that the major part of the passive layer which was finally formed
1 Young, L.: Can. J. Chem. 37,276 (1959); Proc. Roy. Soc. (A) 258, 496 (1960); Acta
Met. 5, 711 (1957).
2 Bean, C. R., J. C. Fisher, and D. A. Vermilyea: Phys. Rev. 101, 551 (1956).
3 Evans, U. R., and H. A. Miley: J. Chem. Soc. (1937) 1295.
418 7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes

consisted of an iron oxide with primarily trivalent ions, probably Fe20a,


about which Bonhoefl'er has already reported. Evans and Vetterl noted
the same conditions in the attack of moderately concentrated HNOa on
iron. The stationary thickness of the passive layer attainable here naturally
depends on the experimental conditions (temperature, applied field, and
electrolyte composition). This relationship has not been previously described
quantitatively for iron passive-layer formation. Only Tronstad,2 Bonhoefl'er
and Vetter,3 and Schwarz 4 made individual measurements, according to
which one must assume that the thickness of the passive layer is between
30 and lOO A. The influence of the temperature on the passive-layer thickness
was studied by Torrisi 5 in the tantalum/passive layer/electrolyte system.
The thickness of the passive layer as a function of the temperature (between
o and 200°C) at 100 volts applied potential is summarized in Table 49.
Even though the layer thicknesses vary depending on the individual methods
of determination, they still lie in the right region and show an increase with
increasing temperature. As can be seen from the data, the thickness of the
passive layer, expressed in angstroms per volt, decreases only slightly, from
about 20 to 15, with increasing potential between lOO and 500 volts. While
according to investigations by Vermilyea,6 the structure of the passive
layer on tantalum consists of a Ta205 layer directly on the tantalum, which

Table 49. Layer Thickness in Angstroms per Volt of the Ta 2 05 Film Formed on
Tantalum in Neutral Electrolytes After 1 hr of Anodic Polarization with
3.1 mA/cm 2 at Various Temperatures and Potentials, According to Torrisi

Layer thickness in angstroms per volt at formation


Measurement Voltage, temperature, °c
method V
0 25 95 195

Gravimetric 100 13 16 20 25
200 18.5
300 18
400 16
500 15
Optical 100 17 21 26 32
Spectrophotometric 100 18 22 28 35

1 Evans, U. R.: Trans. Faraday Soc. 40, 125 (1944); U. R. Evans, and 1. D. G. Berwick:
J. Chem. Soc. (1952),3432; K. J. Vetter: Z. Elektrochem. 55, 274, 675 (1951).
2 Tronstad, L., and C. W. Borgmann: Trans. Faraday Soc. 30, 349 (1934).
3 Bonhoeffcr, K. F., and K. J. Vetter: Z. physik. Chem. 196, 142 (1950); K. G. Wei I
and K. F. Bonhoeffer: Z. phY8ik. Chem. [NF] 4, 175 (1955); K. J. Vetter: Z. Elektrochem.
56, 16, 106 (1952).
4 Schwarz, W.: Z. Elektrochem. 55,170 (1951).
5 Torrisi, A. F.: J. Electrochem. Soc. 102, 176 (1955).
6 Vermilyea, D. A.: Acta j1,Iel. 1, 282 (1953).
7.2. Passive Layer Forlllation on Metals and Alloys 419

is covered by a thicker-probably porous-oxide layer of another composi-


tion, the structure of the passive layer on iron can still not be discussed with
certainty.
After the successful use of Dewald's space-charge theory and the modifi-
cation of this theory by Young in passive layer formation on tantalum,
it is natural to check the extent to which the relationships developed there
are also applicable to the mechanism of passivity in iron. What is at
stake here is whether the space-charge-free field transport through oxide
films according to Cabrera and Mott may be directly applied to the case
of passive layer formation in iron, as was proposed by Hauffe1 and hinted
at but not proved by the measurements of Vetter,2 or whether in this case
the space-charge theory yields a truer picture. In order to approach the
answer to this question, two possible cases for the structure of the passive
layer were discussed by Hauffe, and are reproduced in Figs. 222 and 223.

__ n _ _-,
~A,-

Fe

+ -
+
Fig. 222. Schematic representation of
the concentration of the free electrons
1
and ion defect positions (FeO·· == O'
and 00' == D·) in the homogeneously !':!
structured passive layer Fe20a with
space-charge inversion, according to
Hauffe. arge Surface Charge

As can be seen from the representation, large surface charges form in both
cases as well as large space charges, from which a negative space charge
appears on the metal side and a positive space charge on the electrolyte
side of the passive layer. 3 The concentration changes in the lattice defects
and the corresponding charge phenomena are the bases of the following
considerations:
1. Case of a Homogeneous Passive Layer of Fe 2 03. As can be seen from
Fig. 222, the surface charges which are formed are compensated by charges of
the opposite sign in space-charge zones set up in a steady state in the passive
1 Hauffe, K.: Z. Metallk. 44, 576 (1953); K. Hauffe, and I. Pfeiffer: Z. Metallk. 45, 554
(1954).
2 Vetter, K. J.: Z. Elektrochem. 58, 230 (1954).
a Hauffe, K.: Werkstoffe u. Korrosion 6, 117 (1955).
420 7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes

layer. Here the positive surface charge of the iron on the transfer of a
chemisorbed oxygen ion into a vacancy 00' (if we regard oxygen ion
vacancies 00' and free electrons e as determinative in Fe20a) and the
associated field transport are insignificant compared to the positive space
charge in the passive layer. However, these conditions change during the
migration of the oxygen ion as it approaches the iron/passive layer phase
boundary. As can easily be seen, the transport-aiding effect of the positive
space-charge zone on the transport of oxygen ions grows weaker asthe distance
from the passive layer/electrolyte phase boundary increases, and should
vanish at the space-charge-inversion boundary RI. It would find an in-
surmountable barrier in the narrower negative space-charge zone if it were
not for the fact that the positive surface charge-at least at the start of
passive layer formation-because of the relatively small number of negative
charges in the space-charge zone, exerts a transport-aiding effect through
zone I (Fig. 222). This case should be calculated according to the Dewald
theory for 0 ~ 1.

Fe
+
+ Large
+ - Space Charge
- --I ---.
+ Tho'
+ -
+
+
Fig. 223. Schematic representation of
the course of the concentration of
electrons and ion defects in a passive
F + layer composed of two phases (Fe304-
Fe203), according to Hauffe. (As in
Fig. 222, an opposing diffusion of iron
Large Surface Charge and oxygen ions is also assumed here.)

'When the system has reached maximum stationary passive-layer thick-


ness, however, we may encounter a situation where the surface charges
are compensated for to a large degree by the corresponding space charges.
As a consequence, the field effect may be neglected in the region of the
inversion zone (the width of RI is several interatomic distances), so that
arriving ions find an insurmountable barrier and the growth of the passive
layer comes to a standstill, even though the space-charge zones to the right
and left of the inversion zone RI contain electric fields stronger than necessary
to make field transport in these zones possible.
If the inversion zone RI as is indicated in Fig. 222 is only of the order
of magnitude of an atomic distance, then the further growth of the passive
layer takes place according to the "field compensation" depicted above,
7.2. Passive Layer Formation on Metals and Alloys 421

in that in zone II only oxygen ions and in zone I only iron ions are transported
to the inversion boundary RI and are united there, which condition is given
by the following reaction equation:
(7.19)
This reaction, however, is related to the further assumption that at the
iron/passive layer phase boundary, iron in the form of Fe3+ ions and electrons
e can transfer from the metal into the passive layer.
2. Case of a Diphasic Passive Layer. If a significant iron ion transport
via interstitial lattice positions is not possible in zone I, then this transport
can be achieved by the rebuilding or development of another iron oxide
with energetically more favorable migration conditions for the iron ion. As
we know from high-temperature measurements (see Section 4.5), there is a
slight possibility of an iron ion transport in FeO as well as in Fe304. Under
the assumption of preferential Fe304 formation, which was also observed
in oxidation experiments below 570°C-of course in considerably thicker
layers-zone I was either rebuilt by the initially present homogeneous
Fe 203 layer in Fe304, or formed directly from an Fe304 zone. An energetically
more favorable electron transport is assured by the special arrangement
of the di- and trivalent iron ions in octahedral positions in the inverse spinel
lattice of the Fe304, while the free tetrahedral and octahedral positions
which are present in sufficient number are available for the field transport
of the iron ions in the spinel. As can be seen, at the zone bOlmdary I/Il,
which also represents the Fe304/Fe203 phase boundary, abrupt changes in
the concentrations of electrons and holes as well as in ionic lattice defects
is to be expected (Fig. 223). The iron ion transport in the Fe304 layer over
"spinel vacancies" occurs alongside the field transport of the oxygen ions
in the Fe203 layer which was discussed earlier. These spinel vacancies can
be assumed to have a practically constant concentration through the entire
Fe304 layer because their number is fixed by the lattice structure.
Under the assumption that all phase boundary processes including
the reaction in the interior of the passive layer at the Fe304/Fe203 phase
boundary proceed rapidly,
(7.20)
the field transport of the iron ions in the Fe304 layer competes with that of
the oxygen ions in the Fe203 layer. If both field transports obey the same
rate law, the thickness of the zones (Fe304 and Fe203) may be calculated
from the ratio of the transport coefficients. This mechanism was experi-
mentally confirmed by Nagayama and Cohen.! Iron was anodically oxidized
in a borate-boric acid buffer solution of pH = 8.4 in the potential range
1 Nagayama, M., and M. Cohen: J. Electrochem. Soc. 109, 781 (1962).
422 7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes

extending from the active region to oxygen evolution. In the active region there
appeared to be some Fe 304 on the surface. In the passive region the iron was
covered with an oxide film 10-30 A thick, depending on the potential. The
structure of the film was studied by its cathodic behavior and electron
diffraction. In general, the phase next to the metal most closely resembles
Fe304 and that next to the solution a cation-vacancy cubic oxide containing
iron of a valency greater than three, with y-Fe203 between them. The
thickness of the layer and the concentration of defects were found to be
functions of the anodic potential.
In any case a thickness of the individual zones must be introduced in
such a way that both transport currents remain divergence-free. Because
of the very complicated mechanism of the formation of a passive layer on
iron, it seems advisable for the present, in spite of the large amount of work
that has been invested in this system, to deal first with the mechanism of the
formation of passive layers on nickel, aluminum, and tantalum, which all
have homogeneous structures. Landsberg and Hollnagel1 have reported on
the anodic behavior of nickel in dilute sulfuric acid. An interpretive experi-
ment on the passivity of nickel was undertaken ky Kolotyrkin 2 and by
Hauffe and Pfeiffer3 on basis of current density-potential curves (effect of
halogen ions on the passivity). Schwabe and Dietz,4 in particular, contributed
important results to this problem.
According to a more phenomenological theory by Bonhoeffer and Vetter 5
on the periodic activation and repassivation of iron, which falls in the field
of stability investigations of coatings on metals,6 and whose explanation
was promoted by additional experiments by Franck,7,8 there exists a desire
to extend the earlier results by means of a theory based on theoretical
molecular considerations. However, this will not be possible until we succeed
in providing a noncontradictory explanation for the synthesis and destruction
of the passive layer on iron. The activation-passivation processes for gold,
observed by Franck and co-workers,8,9 can be attributed to an alternative
mechanism, in spite of the similarity between the curves.

1 Landsberg, R., and M. Hollnagel: Z. Elektrochem. 58, 680 (1954).


2 Kolotyrkin, Y. M.: Z. Elektrochem. 62,664 (1958).
3 Hauffe, K., and 1. Pfeiffer: Z. MetaUk. 45, 554 (1954).
4 Schwabe, K., and G. Dietz: Z. Elektrochem. 62, 751 (1958); see also in this journal
the contribution of G. Okamoto, H. Kobayashi, N. Nagayama, and N. Sato, p. 775.
5 Bonhoeffer, K. F., and K. J. Vetter: Z. physik. Chem. 196, 127 (1950).
6 Bonhoeffer, K. F.: Angew. Chem. 67, 1 (1955); K. F. Bonhoeffer, and G. Vollheim:
Z. Naturforsch. 8b, 406 (1953).
7 Franck, U. F.: Habilitationsschr. Univ. G6ttingen, 1954; U. F. Franck and K. G.
'Veil: Z. Elektrochem. 56, 814 (1952).
8 Franck, U. F.: Z. Elektrochem. 62,649 (1958).
9 Franck, U. F.: Z. physik. Chem. [NF] 3, 183 (1955).
7.2. Passive Layer Fornlation on Metals and Alloys 423

Since a logarithmic rate law holds for low-temperature dry oxidation


of nickel, there is a question as to whether the reciprocal logarithmic
mechanism, arrived at using the theory of Dewald and Young, is a suitable
rate law for passive layer formation on nickel. If this should not be the
case, then we would try to propose an "electron-tunneling mechanism"
similar to that attempted for dry oxidation} That this mechanism can be
of positive significance in the synthesis of passive layers with time is apparent
from a work by Uhlig and Lord. 2 These authors followed the oxygen con-
sumption with time during the passivation of 18-8 steeis in distilled water
containing oxygen and acid solutions. A curve of this type is presented in
Fig. 224. The stationary thickness obtained in the passive layer was deter-
mined to be about 20 A, if Cr20a is assumed to be the passive layer material,
which appears reasonable on basis of the discussion of Section 4.6.

~z I---
..,/
0

Fig. 224. Course of oxygen consumption on an V


V 0

18-8 steel in water· containing air, according to


Uhlig and Lord. (The samples were treated
before the experiment at 35°C for 10 min in a /'
V
solution of 25 vol. % commercial concentrationHCI ! J !Il
and 25 vol.% commercial concentration H2S04.) Time of Reaction

As was explained in the introduction to this section, we have discussed


only that facet of the passivity of metals which is responsible for building
up the passive layer. However, as experience has taught, corrosion does not
stop with the complete covering of the metal or of the alloy with a passive
layer, even though its rate is subsequently decreased by about one order of
magnitude. The corrosion rate will be given in this case by the rate of
solution of this passive layer; however, the thickness of this stationary layer
will remain constant, since the loss caused by the departure of the lattice
ions from the passive layer into the electrolyte is counterbalanced by the
marshalling of new metal ions or oxygen ions at the metal surface by the
increased electric field in the passive layer. In the steady state the thickness
of the passive layer will depend on the relative magnitudes of these two
competing processes. Regardless of the details of the mechanism, the rate of
disintegration of the passive layer will in all cases determine the corrosion
rate or the solution current of a metal or alloy which forms a passive layer
in the corroding medium. Thus, it is useful to investigate the mechanisms
of the disintegration of passive layers. From a joint discussion with Vetter,
1 Hauffe, K., and B. Ilschner: Z. Elektrochem. 58, 382 (1954).
2 Uhlig, H. H., and S. S. Lord, Jr.: J. Electrochem. Soc. 100, 216 (1953).
424 7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes

Franck, and the author, Schottky! has dealt with the mechanism of the
solution current in detail. The following section is based on this exposition.

7.3. The Solution Current of Metals with Passive Layers


The iron/passive layer/electrolyte system seems to be especially suitable
for dealing with this question, since here the partial processes which are
decisive for the problems which exist and the determining quantities were
found quantitatively by Bonhoeffer and co-workers.2 A passive state of the
iron can be attained if the externally applied potential, which we designate
as Eft, is about +0.5 volts. The number of iron ions transferred into the
"neighborhood" of solution 3 per sec-cm 2 from the layer (region 2', Fig.
225) is to be regarded as a measure of the solution current ofthis wet corro8ion

Electrolyte
Passive ,-,------',-----~
Layer Quasi-Double Layer
r--"-v-'-.,

W i
m i l
i
1 1
1 1
1 1
1 I Fig. 225. Schematic representa-
1 I
iJl 31
i ql
I
tion Of the individual transport
and diffusion zones in the passive
layer and in the electrolyte.

proce88-at first without consideration of the charge state of the iron ions.
If 8(;:/3) is this solution current, and if the passive layer has the composition
Fe m On (H 2 0)p, where m and n can be considered as integral numbers because
of the considerable thickness of the passive layer in comparison to the size
of single ions, then the number N/cm 2 of the oxide particles in the layer will
be changed by the wet process according to the equation
1
_ _ 8(2'/3)
Fe
(7.21)
m
As discussed above, this loss will be compensated by field transport
and formation of new pa88ive-layer molecule8. We designate this process as
dry 'reaction and write

(dN)
dt dry
(7.22)

1 Schottky, W., Passivitiit und Losungsstrom: in Halbleiterprobleme, Vol. 2, p. 233,


Braunschweig, 1955.
2 See for instance the review by H. Gerischer: Angew. Chern. 70, 285 (1958).
7.3. The Solution Current of Metals with Passive Layers 425

While for the steady state

dN/dt = 0, s!J~2*) = 8!,j~/3) (7.23)

is valid, we obtain for the nonsteady state

dN
dt
= ~
m
(8Fe
(1/2*) _ (2'/3»)
8Fe (7.24)

Therefore the calculation of the two SFe quantities as functions of E h , pH,


and g is the theoretical problem requiring a solution.
In the solution of this problem the experimental results on s~~/3) in the
stationary-layer thickness by Vetterl are available. Further, it is known from
the same work that the iron ions appear predominantly in the trivalent
state in the electrolyte. Of course, these experiments do not consider a
possible entrainment of 0 2 - or OH- ions with the departing iron ions. 2
The decisive quantitative arguments for the stationary S~: /3) current,
however, were obtained through the direct current measurement in the
stationary state at a given Eh and pH, and thus the relationship between the
current flowing through the whole cell i/cm 2 and s~:/3) must be determined.
Here the nonstahonary state must also be treated. The electrolyte should
contain no impurity ions capable of a change in charge, so that no additional
current contributions need be considered. Moreover, the changes in Eh and
the increase in the layer thickness which may occur in the non stationary
state happen so slowly that the currents necessary for the rebuilding of the
space charge inside and near the boundaries of the passive layer may be
neglected compared to the currents that go through it. Then div i ~ 0 every-
where.
According to the stated definition 2 not only the 0 2 - penetration but
also the OH- penetration through the boundary layer (2'/3) was considered.
For the 0 2 - current we formulate
(7.25)

For the general nonstationary current and the fact that an sg'" /3) current
of -ndN/dt is associated with the decomposition lor synthesis of N layer
molecules, the following relationship was derived from (7.24):

S(2'/3) - _ ~ (S(I/2*) _ S(2'/3»)


(7.26)
o - m Fe Fe

1Vetter, K. J.: Z. physik. Chem. [NF] 4, 165 (1955).


2The introduction of 0 2 - ions in the electrolyte is useful for reasons of simplicity.
Regardless of whether oxygen ions are actually present in the solution the electro-
chemical potential T}O'- may be defined as T}H,O - 2 T}H+ or as T} OH- - T}H+, since
2T}H- + T}o'- = T}H,O'
426 7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes

From equations (7.25) and (7.26) and the sum of the two currents from the
di- and trivalent iron ions, 2es~~'~) + 3es~:'!) , it follows that the total non-
stationary current from the beginning of an immersion of an iron sheet
with a growing passive layer is generally expressed as

. _ (2 m2n)
~tot - -
(2'/3)
sFeH
+ (3 - m
2n) sFe'+ + m
2n sFe(112*)
(2'/3) (7.27)

which may be simplified in the presence of a homogeneous Fe 203 layer to


(rn = 2, n = 3):
(7.27a)
On the other hand, in the stationary state, where we always have
s~'/3) = 0 and dN /dt = 0, we obtain for the stationary current:
.
~stationary =
2esFez+
(2' /3)
+ 3eS (2' /3) (7.28)
Fe3 -t;

As can be seen from (7.28), there is no general equivalent relationship


between istationary and the total s~~/3) solution current. Thus the concept
of the equivalent current introduced by Bonhoeffer and Franck1 based on
the disappearance of one of the two partial currents is justified, since it is
shown that only Fe 2 + ions participate. In the same way the concept of the
"corrosion current" iK introduced by Vetter is equal to the corrosion
rate when only the iron ions leave in a charged state. In the passive layer,
therefore, where only Fe 3 + ions are transferred to the electrolyte, we obtain

iK = 3es~~{~) (7.29)
According to Vetter the measurable iK value can be written as a function
of the H + ion concentration:
iK = 5 X 10-5 X 1O-o.84 pH A/cm2 (7.30)
independent of Eh in the pure passive region (measured exactly between
Eh = 1 and 1.2 volts). ThepH dependence was determined in H 2 S04 solutions
between pH = 0.7 and 3.9. From equation (7.30) the iron disintegration is
calculated as
1 1019
s(2'/3) = -iK = iK ::::! 1014 X 1O-0.84 pH particles/cm2 -sec (7.31)
Fe 3e 3 x 1.6
If it is assumed that there are approximately 1015 iron atoms in the
lattice plane, then in the stationary corrosion process at a pH of 0 about
one-tenth of an atomic layer is decomposed per second (and replenished
by the iron), while at a pH of 4 only about 5 x 10-5 atomic layers are
1 Bonhoeffer, K. F., and U. F. Franck: Z. Elel,trochem. 55,180 (1951).
7.3. The Solution Current of Metals with Passive Layers 427

decomposed and then replenished [± 10% was indicated as the limit of


accuracy for the factor 0.84 in equations (7.30) or (7.31); according to
subsequent experiments by Weil and Bonhoeffer1 in the pH region 0.3 up
to 6 a notably weaker pH-dependent process was found; furthermore the
corrosion current is dependent on the type and concentration of the anions.]
We now wish to identify the rate-determining step for the solution
process in the passive layer. Occasioned by the further decisive observation
that iK in the passive region depends neither on the stirring process in the
solution nor on changes of the Fe3 +jFe 2 + ratio in the solution, Vetter believes
the rate-determining process is the transfer of the Fe 3+ ions from the passive
layer surface into the electrolyte. Schottky raised some enlightening objec-
tions to this assumption, which had already been given by Bonhoeffer.
It is questionable to ascribe such a fundamental effect as the independence
of the solution or corrosion current of the applied potential to a mechanism
according to which the effect is field-dependent but the field distribution
is assumed such that for a variable applied potential the effect of the field
on Vetter's iron ion transfer remains unchanged. Schottky succeeded in
finding a mechanism in which the solution process is generally independent
of the layer thickness and the mechanism of replenishing through the layer
and in which, moreover, the influence of the electrolyte (pH value and type of
anions) is automatically taken into consideration.
The goal was to maintain the assumption of an unimpeded equilibrium
between the electrolyte boundary 2' of the passive layer and the electrolyte
region 3 for the ions that determine the solution rate, and the rate-determin-
ing process in the removal of these ions is to be sought in the transfer of
the electrolytes from regions 3 to 4. Schottky was able to show that here it is
not a question of the Fe3 + ions themselves. Furthermore, it appears to be
only slightly probable that Fe3 + ions as such transfer in significant quantities
from 2' to 3. To account for the lack of dependence of the solution current
on the stirring rate, the particles are not permitted to transfer in the un-
changed state from 3 to 4, but must react with some constituent in the electro-
lyte or dissociate themselves into other constituents within the diffusion
distance, which is smaller than the boundary distances in the stirring (less
than 10-3 cm); in any case they must change through some kind of reaction.
This mechanism postulates a rapid removal of the reaction products so that
the "reaction diffusion" of the particles that leave the layer, which is directly
related to their equilibrium concentration, can be the rate-determining
process.
In order to formalize these requirements, it was assumed that in solutions
with anions which do not form complexes, the Fe3 + ions do not leave the

1 Weil, K. G., and K. F. Bonhoeffer: Z. physik. Chern. [NF] 4, 175 (1955).


428 7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes

passive layer from 2' to 3 alone, but rather in association with 0 2 - as FeO+
ions, and that their departure is unlimited, while the direct Fe3+ transfer
is strongly limited. For this unlimited transition, the equilibrium condition
for the FeO+ particles in electrolyte region 3 reads
(7.32)

Since Co- '" cH~' the following is valid for the equilibrium concentration:
c~lo+ = const cH + (7.33)

We can now assume that the effective diffusion distance .dg of the FeO+
is independent of pH, and the solution current thus obtained is proportional
to lO-pH in approximate agreement with the value observed by Vetter of
IO-O.8pH. With this, a basic discrepancy in the naive Fe3+ solution theory,
in which iK would be proportional to 1O-3pH, has been removed.
The same pH change in the equilibrium concentration would also appear
as in (7.33) if the hypothetical FeO+ ion were associated with one or more
H 2 0 molecules, whether in the form of a water sheath or in a closer chemical
bond.
In the case of ion transition in the active region Schottky proved that
a direct transfer of divalent iron ions from the passive layer into the electro-
lyte also cannot be involved, since the pH dependence of the Flade potential
EhF does not show the behavior observed by Bonhoeffer and co-workers,

EhF = const - O.058pH (7.34)

but rather the following: EhF = const - 2 x 0.058 x pH. Agreement is


also obtained with FeOH + ions rather than Fe 2 +. For this case the equilibrium
relationship reads
(7.35)

Because of the proportionality of cW~OH+ to cii+, equality of the iK


currents is reached for an EhF value, whose pH dependence corresponds
exactly to the observed relationship (7.34).
Therefore the theoretical study by Schottky generally proceeds from the
assumption that not all of the oxygen bonds to the iron are broken in the
transfer of the iron ions from the oxide into the electrolyte, but that the
transfer takes place in the form of "iron-oxygen complexes" which have
retained part of the bonding present in the oxide. The corrosion behavior
in the passive layer in strongly basic solutions with high pH values can be
mentioned as further support for the Schottky theory of complex dissolution.
Lossow and Kabalow 1 have reported on this where the complex ions FeO~-
1 Lossow, 'V. 'V., and B. N. Kabalow: Zhur. Fiz. Khim. SSSR 28, 824, 914 (1954),
7.3. The Solution Current of Metals with Passive Layers 429

and Fe20:-, which are transferred, have been determinative for the solution
process. The observed dependence of the solution current on the stirring
rate in an alkaline pH region indicates that the assumption introduced by
Schottky, that the complex ions leaving the passive layer disappear through
reaction diffusion within short diffusion distances, is evidently valid only
with lower pH values.
The question may now be raised as to whether the reaction in short
diffusion distances (less than 10-3 em), that is, the reaction diffusion, can be
explained. Schottky proposed for discussion the following reaction for the
lower-pH region:

(7.36)

where the equilibrium is largely shifted toward the right in strongly acid
electrolytes. Since a direct transfer of Fe3 + ions from the passive layer
should not be possible, Fe 3+ ions can only be formed by reaction (7.36).
However, a back reaction could only appear when the Fe3+ ions formed
were removed toward 4 too slowly (Fig. 225), which is certainly not the case
in the region of lower pH. We may thus assume that in the pH region from
o to about 3, reaction (7.36) proceeding from left to right is rate-determining
for the solution current. Here we deal with a dissociation process of FeO+
independent of the pH value which proceeds with a ratc constant dependent
only on the rate of thermal decomposition of FeO+ (in reaction with the
solvent).
This reaction mechanism is identical with the destruction mechanism
of minority carriers in semiconductors, where the minority carriers, i.e.,
e or EE> particles in a numerical minority, diffusing into a semiconducting
zone are caused to disappear in a "sea" of EE> or e particles by a reaction
which is independent of place and time. There exist a definite "lifetime"
-r and a diffusion length L = (D-r)1I2. The diffusion current which initially
flows in a plane is given by D.no/L = no(D/-r)1/2, where no denotes the
concentration of the particles in the initial face (here face 3). Since L is
comparable with the diffusion distance Ll g appearing in reactionless diffusion,
the pH independence is valid not only for L but also for Llg. However, this
signifies in turn that the pH change with 8Fe and iK is, according to (7.33),
uniquely determined by the pH change with CFeO+.
However, in a region of higher pH this removal mechanism encounters
difficulties since the Fe 3 + ion concentration formed according to (7.36) is
then limited to lower equilibrium concentrations C~~3+ as indicated above
in order to ensure a sufficiently great concentration gradient and thus a
sufficiently great Fe3 + diffusion current. According to Weil and Bonhoeffer,
this mechanism steps completely into the background at a pH of 5 since the
430 7. Mechanism of Oxide Layer Formation in Aqueous Electrolytes

solution in the entire electrolyte volume is filled in a fraction of a second to


saturation with Fe3+ ions.
The end product of the solution process in the region of intermediate
pH values is not the Fe 3+ ions diffusing into region 4, but rather the "rust"
FeOOH, which appears in colloidal form in the intermediate pH region
in a well-known way and is responsible for the brown coloration of the
electrolyte. Since in this pH region Fe3 + ions cannot appear as intermediate
products, Schottky has proposed the following two reactions in the reaction
diffusion of the FeO+ complex for discussion, according to which the neutral
free "rust molecule" was formed:

FeO+ + OH- -» FeOOH (7.37)

FeO+ + H 0 -» FeOOH + H +
2 (7.38)
As a calculation shows, the reaction rate of the hydrolytic reaction
(7.38), even with an activation energy of about 16 kcal ~ 0.7 eV, is still
large enough to give the required small Llg value in the reaction diffusion
of FeO+ required according to the theory, which is not so much the case for
(7.37). On this basis Schottky and Vetter prefer reaction (7.38).
Since most passivity layers on metals and alloys are represented by
oxides, and therefore by semiconductors, published investigations of the
behavior of semiconductors in various electrolytes seem to provide new
knowledge and prospects of this phenomenon. These studies of the semi-
conductor electrolyte interface date back to Brattain and Garrett's work1
on the germanium electrode. This and other works 2 have been summarized
in a review article by Dewald. 3 Although qualitative understanding of the
germanium electrode is fairly well advanced, complications arise because of
the formation of oxide films. The investigations were thus extended to oxides
where such a complication does not enter; zinc oxide was particularly suitable,
because, of all the thermodynamically stable semiconductors in contact
with aqueous electrolytes, its bulk properties (band structure, mobilities
and concentrations of the charge carriers, impurity ionization energies, etc.)
are presently the most thoroughly understood. Gerischer 4 and Gobrecht 5
have published valuable contributions to this problem. All these results
help to lead us toward greater understanding of the electronic and ionic
processes in passive layers .•
1 Brattain, W. R., and C. G. B. Garrett: Bell System Tech. J. 34, 129 (1955).
2 See for instance K. Bohnenkamp and R . .T. Engel!: Z. Elektrochem. 61, 1184 (1957).
3 Dewald, J. F.: in Semiconductors, Reinhold, New York, 1959, p. 727; Bell System Tech.

J. 39, 615 (1960).


4 Gerischer, H.: Z. physik. Chem. [NFJ 26, 223, 325 (1960); 27, 48 (1961); H. Gerischer
and F. Beck: Z. physik. Chem. [NFJ 24, 378 (1960).
5 Gobrecht, R., O. Meinhardt, and L. Schulz: Z. Elektrochem. 67, 142, 156 (1963).
8. A Few Approved Methods of
Measurement of Coating Growth

The measurement of the course of the oxide-layer growth with time,


especially at higher temperatures, is generally carried out from the deter-
mination of the weight increase of the metal or alloy sample with time,
caused, for example, by the quantity of oxygen taken up by the oxide forma-
tion. In this technique one frequently carries out the determination of the
weight increase as a measure of the rate of oxidation by placing several
samples in the furnace at a definite temperature and gas composition.
The samples are removed from the furnace at different times at suitable
time intervals and their weight increase is determined on an analytical
balance. With similar pretreatment of the samples the weight increases
which are found at different times permit the construction of a continuous
weight increase-time curve as long as no secondary disturbances, such as
bursting and spalling of the oxide layers or recrystallization processes,
appear. For further evaluation one can try to obtain a straight line with the
measured points, corresponding to a linear, parabolic, cubic, or logarithmic
rate law. As long as the material which is to be investigated, e.g., steel, is
available in large quantities, this "large-number measurement method"
is quite useful. Frequently, especially with nonadherent scaling layers, the
oxide layer is removed after oxidation and the weight loss of the metal
sample is determined. Here the scale is removed with a steel wire brush
or by a chemical or electrochemical pickling. This procedure is to be recom-
mended only with larger technological samples.
For valuable alloys other methods of measurement have come into use
in the laboratory and also for the control of operations in the course of time.
We will report on these in the following sections.

8.1. Use of the Microbalance in Oxidation Apparatus


In order to follow weight increase of small metal and alloy samples with
time, two types of balances have been successful: the quartz spiral balance
and the beam balance with a counterweight. Both balances are placed in an
431
432 8. Methods for Measurement of Coatin~ Growth

extended part of the reaction space of the apparatus and can operate under
the most varied conditions-in high vacuum as well as in different corroding
atmospheres, and even at higher temperatures in the reaction space itself.
The quartz spiral balanee has the simpler eonstruction and manipulation
and is more economical at higher sensitivity and larger measuring region.
The quartz spiral balance used by the author and his co-workers, made from
15 to 25 turns with a spiral diameter of 3 to 5 em, has a measuring region of
50,000 to 100,000 units; that is, the weight increase of a metal foil weighing
100 mg can be measured to exactly ± 1 x 10-3 mg during oxidation.
In order to avoid gas depletion in the neighborhood of the metal foil
in the oxidation space at low gas pressures, of 10-4 to 10-1 mm Hg, it is
advisable to provide for a sufficiently great flow rate of the reacting gases
in the furnace space througli a suitable system of inlet and outlet capillaries.
Thus, a sensitive quartz spiral is steadily lifted upward. This frequently
is accompanied by a vibration of the balance pointer so that a correct reading
is not possible. Moreover, under these conditions the determination of the
reference point causes difficulties in the reading. For this reason it has proven
necessary to interrupt the gas flow shortly before the reading, and to take the
reading after a certain waiting period (dependent on the oscillation time of
the spiral) and finally to turn on the gas stream again. A flow-type apparatus
with a quartz balance for oxidation experiments with suitable gas pressures
is shown in Fig. 226. (The quartz beam can be replaced by a quartz spiral.)

J e
1----.-~HgP

Fig. 226. Apparatus for the measurement of the


oxidation rate of metal foils by means of a quartz
balance, according to Wagner and Grunewald.
I-microscope; 2-quartz fiber as balance; 3-
reference mark; 4-platinum wire; 5-quartz tube;
6-furnace; 7-metal sample for the oxidation;
kIt and M 2-manometers; Kl-inlet capillary for
oxygen; K 2, K3, K4-outlet capillaries; HgP-
mercury diffusion pump; OP-oil pump; B V-buffer
volume.

For numerous experiments, as, for example, in sulfide formation on metals


in sulfur vapor, it is necessary to place the balance directly in the reaction
space, as was described earlier in Chapter 5 and shown in Fig. 206. In such
cases the quartz spiral balance is well suited because of its chemical resistance
and the low temperature coefficient of expansion. It represents, because of
its simple and economical construction, one of the best arrangements to con-
tinually follow the course of the weight increase with time with quite good
accuracy.
8.1. Use of the Microbalance in Oxidation Apparatus 433

Gulbransen developed a highly sensitive microbalance based on an


apparatus by Nernst and Donau with which one can determine the growth
of very thin oxide layers with time. The construction and manner of operation
of this balance can be understood from Fig. 227. A movable thin quartz

Pointer wit'h
~ Reference
~ "....., Mark I]-Quartz Tube
Reading Metal
Telescope Sample b
a @ Metal Sample

Fig. 227. Quartz microbalance, according to Gulbransen.


(a) Individual parts of the balance: beam length 15 em, beam
weight 1 g for samples of 0.7 g total weight. (b) Introduction
of the balance into the oxidation apparatus. The balance
part of the glass apparatus is placed on springs in order to
dampen the vibrations.

beam which carried a counterweight on the left side of about the same size
as the metal sample which is suspended on the right side rests on a very thin,
stretched tungsten wire, which is fastened to a quartz frame by means of
molten AgCl. A thin tungsten wire attached on the right end of the balance
beam serves as a pointer for the exact recording of the change in slope of
this quartz beam. The corresponding reference pointer is located on the
quartz frame. By means of a telescope with a suitable enlargement, the
position of the indicator was read off during the oxidation. Gulbransen!
indicates a sensitivity of 0.3 x 10-6 g for an oxidizing sample that has a
total weight of 0.6840 g. The pressure coefficient was determined to be less
than 0.3 x 10- 6 g per 1 atm pressure change. The temperature coefficient
was 0.8 x 10- 6 g;oC. The sensitivity of the balance was increased by about
a factor of six by Rhodin. 2
The balances equipped with electromagnetic adjustment operate on
the same principle. The counterweight on the left end is surrounded by
a current-carrying coil. The current is varied until the null point of the
1 Gulbransen, E. A.: Rev. Sci. Instr. 15, 201 (1944).
2 Rhodin, T. N.: J. Am. Chem. B06. 72, 4343 (1950).
434 8. Methods for Measurement of Coating Growth

balance-beam pointer, which is observed by means of a telescope, is


reached. This position is changed by the continuous weight increase during
the oxidation. Thus, the oxide-film growth is observed with time by means
of a sensitive current-measuring apparatus. These balances are available
commercially in different constructions.
A balance which is relatively simple to construct and which works
according to this principle has been described by Maurer.! The microbalance
is represented in Fig. 228. A quartz frame A comprises the balance mounting.

Fig. 228. Quartz microbalance, according to Maurer. Thedi-


mensions of the balance are such that it may be set directly
into the furnace space. In this case the pointer position can
be read by means of a telescope through a window in the
furnace. A-quartz-support rod; B-beam balance; C-
pointer; D-oxidizing sample; E-iron cylinder for electro-
magnetic regulation of the balance beam; F-counter-
weight; H-Helmholtz coils.

The balance beam B was made of six small 0.5-mm-thick quartz rods melted
together at right angles. Two of these rods form the balance beam and two
short rods located perpendicular to these serve for the adjustment to the
center of gravity. The remaining two were cut short and fused by means of
their tapered ends to the mounting provided on frame A. The end drawn
out to a fine point on the left part of the balance beam and the pointer G
serve for null-point indication, which was read off here by means of a tele-
scope. The iron cylinder necessary for the electromagnetic control, which is
about 5 mm long and 1 mm thick, was sealed into a thin quartz capillary
and bent at an angle of about 20° to the balance beam, as in Fig. 228. The
counterweight F consists of a platinum wire sealed into quartz, whose size
corresponds to the weight of the oxidizing metal sample. One can generally
manage with two or three such counterweights. As is indicated in Fig. 228,
two Helmholtz coils are located on the right end of the balance which are
powered by a continuously operating direct-current source (e.g., storage
batteries). The current now varies until the needle on the balance beam
1 Maurer, R. J.: J. Chern. Phys. 13, 321 (1945).
8.2. Gas Volumetric and Manometric Methods 435

moves to C on the null point. In this way a direct proportionality prevails


between the weight increase appearing in the oxidation (in milligrams) and
i2 - i~ (in amperes), where io is the current before the oxidation and i is that
during the oxidation (weight increase after a certain time) which is required
to reach the null point at C. If one works at different gas pressures, this slight
correction must also be determined in the calibration of the balance. In the
use of a balance with the masses given above, for example, in a pressure
region of 10 mm to 760 mm Hg, one has to deal with a subtractive term,
mo = 1 x 10-5 g, but this can be neglected in the weight increases of 0.1
to 50 mg which appear during oxidation.

8.2. Gas Volumetric and Manometric Methods for Measurement


of the Oxidation Rate
In the formation of thin oxide films on metals and alloys, gas volumetric
apparatus can frequently be used to advantage, where the gas volume
consumed in the reaction is measured in a microburette at constant tempera-
ture and gas pressure. The constant gas pressure was observed very accurately
here by a differential manometer or tensiometer, which serves as a null
instrument. One such apparatus is schematically represented in Fig. 229. 1

Fig. 229. Schematic representa..


tion of a gas volumetric oxida-
tion apparatus for the deter-
mination of the course of the
formation of thin oxide films Gas -f--r----r--r---+-.--_f_ To High-
Inlet /1" Vac u um
with time. (In front of the gas Pump
inlet there is the usual gas
purification train and on the Vtf'
way to the high vacuum there
is a trap with liquid air.)
H 1_4-connecting stopcocks; If
V R--experimcntal tube; 0 -
furnace; GB-ga.s burette; T- fl
differential tensiometer; K-
measuring capillary with milli-
meter graduations; M -mercury
manometer.

The actual measuring device is the differential tensiometer T in combination


with the gas microburette GB. The differential tensiometer consists essentially
of a horizontal capillary with a graduated scale, which lies between the
apparatus and a comparison vessel, in which the same gas pressure prevails
as is in the apparatus prior to the onset of oxidation. If the oxygen in the
apparatus decreases in the course of the oxidation, then a liquid drop
1 See for instance H. J. Engell, K. Haufi'e, and B. Ilschner: Z. Elektrochem. 58, 478 (1954).
436 8. Methods for Measurement of Coating Growth

located in the capillary is displaced. By raising the mercury level in the


gas burette the oxygen consumption can be compensated and the drops can
be returned again to the starting point. The electric furnace 0, which should
have not too small a heat capacity, generally had a temperature constant
to ± 1.0°0 in the temperature region from 200 to 800°0 by use of a sensitive
temperature regulator. Since periodic oscillations in the temperature result
in practically equal deviations, one can arrange the observation (apart from
the first minutes after the beginning of the reaction) so that it is made at
the moment the thermal potential of the thermocouple has just attained
the theoretical value set by the compensator. In this way the influence of the
temperature oscillation in the furnace on the gas pressure in the apparatus
can be approximately eliminated. The use of a thermostatted space with
automatic temperature regulation of at least ± 0.2°0 accuracy is required
in that type of sensitive measuring apparatus. (With an apparatus volume of
about 200 ml and the temperature fluctuation indicated above, the maximum
experimental error at 400°0 amounts to about 6 x 10-3 ml STP, which, for
example, at 30 mm Hg oxygen pressure in the apparatus and 20-cm 2 surface
of a nickel sheet corresponds to an oxide layer thickness of about 25 A.)
The measurement itself proceeds in the following way: the metal sample
to be oxidized is introduced into the quartz tube VR in a cold furnace; the
stopcock H2 is closed and the experimental tube above HI is pumped down
to 10- 6 mm Hg. Then the furnace is heated while the high-vacuum pump is
working and the metal sample heat-treated for a few hours at the oxidation
temperature. In the meantime oxygen is let into the previously evacuated
apparatus through H4 and into the left side of the apparatus by opening
stopcock H 3 . The drops of a suitable liquid with very low vapor pressure
(for example, diethyl phthalate) located in capillary K are introduced by a
gentle tilt in the tensiometer through a ground joint provided for it and
brought to the null mark on the graduation on the capillary. After closing HI
and admitting gas into the experimental tube by opening H 2 , it is advisable
to wait about 5-10 sec to obtain pressure equilibrium in the apparatus
closing stopcock H 3 . The drop in the capillary is now set in motion. However,
before the drop reaches the right end of the scale, it is pushed back again by
raising the level of the mercury in the gas burette GB. The value of the drop
position in the capillary is multiplied by an empirically determined conversion
factor and is used with the reading on the gas burette to calculate the gas
volume which was consumed. However, for exact measurements of the
initial course of the oxidation, the quantity of oxygen consumed by chemi.
sorption on the vessel walls in the furnace space is determined. This is done
in the same way as described above, only without a metal sample. This
apparatus is especially well-suited for the determination of the growth of
the oxide layer between 50 and 10,000 A thickness.
8.2. Gas Volumetric and Manometric Methods 437

For the measurement of greater layer thicknesses one can use either an
apparatus working according to the gravimetric principle, as discussed above,
or a simplified arrangement working on the gas volumetric principle with
pressure measurement, which shall be described briefly here because of its
simplicity (Fig. 230).1 In a horizontal furnace 0 with a metal core K for

o
K

Fig. 230. Representation of an oxidation apparatus with pressure


measurement for 1 atm total pressure. Ql and Q2-hollow
quartz glass containers; M-manometer; H l _4-connecting
stopcocks; V-metal or alloy sample; K-metal core; 0-
furnace; B-to bubble counter.

temperature homogenization, the experimental sample (this is also to be


considered in all other apparatus) is located in one of two quartz tubes,
one of which is connected with a mercury differential manometer M and can
be closed off from the other by a stopcock H 2 which is a direct connection
between the two quartz tubes. One quartz tube leads to the oxidizing metal
sample V, while the other serves as a comparison tube. In order to keep the
total volume of the apparatus as small as possible, two hollow sealed
quartz bodies Ql and Q2 of approximately the same size are located in the
two quartz tubes. After the metal sample was introduced, it was heated in
high vacuum to the oxidation temperature, and, by opening the stopcocks
HI and H2 and closing Ha and H4, oxygen or air up to 1 atm was rapidly
admitted. Directly after that, the stopcocks HI and H2 were closed and the
pressure change with time of the measuring tube compared with the com-
parison tube was observed by means of the differential manometer. The
calibration of the manometer was carried out by admitting or withdrawing
a definite gas volume in the apparatus (H2' Ha, and H4 closed) and the pressure
1Wagner, C., and K. Griinewald: Z. physik. Ohem. (B) 40, 473 (1938); K. Hauffe:
Z. anorg. u. allgem. Ohem. 257, 279 (1948).
438 8. Methods for Measurement of Coating Growth

difference read off in mm Hg. Only oxidation experiments in the region of


l-atm total pressure could be carried out in this apparatus. The sensitivity
of this apparatus may be easily increased by a factor of 10-15, if one uses
an organic liquid with a low vapor pressure as the manometer fluid in place
of the mercury.
A number of these and similar apparatuses have been described in the
literature. For more exact manometric measurements, we have to note
especially the apparatus of Campbell and Thomas. 1 Some time ago a highly
sensitive micromembrane manometer was made commercially available
and it can replace the above-mentioned differential manometer. With this
manometer the course of the oxidation with time is observable at the very
beginning of oxidation.

8.S. Further Methods for the Measurement of the Thickness of


Tarmdshiog Layers
When simultaneous structure investigations on tarnishing layers are
planned, the measurement of the layer thickness under the microscope
likewise offers a useful method. The microscopic measurement of the layer
thicknesses is at this time the most exact method (see Section 4.8) especially
for the separate determination of the progress of the internal and the
simultaneously appearing external oxidation zones.
A quite elegant procedure for the determination of thin tarnishing
layers was worked out by Evans and Bannister2 on an electrochemical
basis and especially by Miley3 and Thomas4 for the layer-thickness deter-
mination of corrosion and oxide layers on metals. Here the sample covered
with an oxide or sulfide or halide film was placed in a suitable electrolyte
and cathodically reduced. The sample was electrolyzed with a definite
and constant current density until the metal surface was completely free
of reaction product. Frequently, this was observable from a more or less
rapid potential increase. As shown in the experimental arrangement sche-
matically represented in Fig. 231, a sufficiently large anode was used on
both sides of the oxidized metal sheet. A Haber-Luggin electrode serves
as a reference electrode in the potential measurement, which is directly
adjacent to the cathode. From the current density i and the decomposition
1 Campbell, W. E., and U. B. Thomas: Trans. Electrochem. Soc. 91, 623 (1947).
2 Evans, U. R., and L. C. Bannister: Proc. Roy. Soc. (A) 125, 370 (1929).
3 Miley, H. A.: J. Am. Chern. Soc. 59, 2626 (1937); J. B. Dyess and H. A. Miley: Trans.
AIME 133, 239 (1939).
4 Price, L. E., and G. J. Thomas: Trans. Electrochem. Soc. 76, 329 (1939); J. Inst. Metals
63, 21, 29 (1938); W. E. Campbell and U. B. Thomas: Trans. Elecirochem. Soc. 76,
303 (1939).
8.3. Further Methods for Measurement of Thickness of Tarnishing Layers 439

time t it follows that for the layer thickness Llg of the tarnishing layer, if m
and e are the mass in grams and the density in grams per cm3 of the reaction
product, respectively,
itm x lO5
Llg = ----
ex 96,500
However, the potential increase is often insufficiently abrupt, especially
in the cathodic decomposition of oxide layers, so this procedure has only
limited application in layer-thickness measurement.

Fig. 231. Circuit diagram for the cathodic reduction of


oxide films, according to Campbell and Thomas. V-metal If {,
sample with oxide film: A-anodes in the electrolyte
(0.1 N KCI); RI-0.05 to 2.0 megohm; R2-suitable
resistance; L-reference electrode (for example AgjAgCI
in 0.1 N KCl) in Haber-Luggin capillary; Kl and K2-
switches; M1-milliammeter; Ma-microammeter.

The use of optical methods for film thickness measurement of oxidized


metals is generally recommended only when a special problem requires this
method, since the apparatus is quite expensive and the evaluation of the
results obtained from this method is not simple. Such an optical method
was worked out by Tronstad 1 and extended and improved by Winter-
bottom. 2
In oxidation experiments on metals, where one has to deal with a
considerable oxygen solubility in addition to the oxide layer formation,
as, for example, in the systems titanium-oxygen or zirconium-oxygen, the
determination of the true oxide-layer thickness is possible only by an optical
method, by means of which one can in combination with a volumetric method
separately determine which portions of the oxygen go into solution and which
are consumed in the formation of the oxide film.
1 Tronstad, L.: Trans. Faraday Soc. 29, 502 (1933) .
2 Winterbottom, A. B.: J. Sci. Instr. 14, 203 (1937); Tran8. Faraday Soc. 42, 487 (1946).
Author Index

Accountius, O. E., 226 Baldwin, W. M., 195, 211, 256, 257, 258, 262, 265,
Achter, M. R., 40 213, 230, 271, 272,273, 275, 277, 278, 285,287,
Adda, Y., 72 396 302, 303, 322, 363
Agar, J. N., 143, 286 Ball, F. E., 159 Bendel, S., 169, 328
Agew, N. W., 63 Balluffi, R. W., 30, 31, 34, Benedicks, C., 314
Akeroyd, E. J., 132, 204 51 Benesovsky, F., 248
Albers, W., 70 Banks, F. R., 15, 74 Benton, A. F .. 190, 265
Albert, L., 80 Bannister, C. N., 63 Berkowitz, A. E., 67
Albrecht, W. M., 230 Bannister, L. C., 146, 156, Bernal, J. D., 28
Alexander, B. H., 30, 51 438 Bernstein, R. B .. 263
Alexander, W. A .. 212 Barbour, J. P., 43 Berwick, I. D. G., 405, 418
Allen, R. L., 78, 378 Bardeen, J., 33, 93, III Besch, A. L., 393
Ames, S. L., 223 Bardenheuer, P., 63 Bettler, P. C., 43
Amgwerd, P., 253 Bardolle, J., 265, 275,277, Betz, H., 128, 409
Anderson, J. S., 21,78 278, 287 Bever, R. S., 78
Anderson, P. B., 142, 217, Barnes, R. S., 29, 31,40 Beyer, J., 148
218 Barrett, B. R., 224 Bienert, H., 404
Anderson, W. A., 336, 345, Barrett, C. S .. 166, 217 Binder, W.O., 308
358 Barth, T. F., 23 Bingle, J. D., 242
Andersson, S., 210, 217 Bartlett, R. W., 174 .Birchenall, C. E., 31, 45,
Andrew, K. F., 48, 94, 95, Batz, W., 45, 46 51, 57, 66, 72, 74, 78,
121, 173, 174, 184, 185, Bauer, G., 17, 236,403 140, 142, 170, 172, 214,
186, 194, 213, 219, 227, Baukloh, W., 47, 139, 265, 215, 254, 263, 271, 275,
228, 236, 243, 247, 277, 389, 390 279, 311, 383
Andrews, J. P., 17 Baumann, H., 302 Bircumsbaw, L. L., 169
Ang, C. Y., 48 Baumbach, H. H., 16, 17, Birks, N., 184
Argent, B. B., 239 19,22 Blin, H., 29
Arkharov, V. 1., 194, 195, Baur, J. P., 174, 292 Block, J., 21, 193, 266
210, 219, 273, 383 Bean, C. R., 417 Bloem, J., 23, 161
Armijo, J. S., 236 Beck, F. H., 140, 252,430 Blood, R. E .. 236
Arnold, K., 404 Becker, K., 301 Boettcher, A., 322
Aronson, S., 232 Becker, R., 286 Bohmer, H., 63
Arseneva-Geil, A. N., 366 Bedworth, R. E., 4, 145, Bohnenkamp, K., 302, 400,
Asaro, F .. 51 159, 164, 248, 264, 268, 430
Asher, R. C., 236 323, 350 Bole, G. A., 226
Auskern, A. B., 78, 242 Beerwald, A. H., 59 Bollenrath, F., 59
Aust, K. T., 42 Belle, J., 48, 78, 120,229, Bolton, M. J., 380
Averbach, B. L., 57, 63,74 242 Bonhoeffer, K. F., 130,404,
Aylmore, D. W., 217 Belokurova, 1. N., 191 405, 408, 411, 418,422,
Azzam, A., 53, 74 Benard, J., 167, 187,190, 426, 427

441
442 Author Index

Borgmann, C. W., 405, 418 Calnan, E. A., 143, 286 Darras, R., 265
Bosenberg, W., 70 Campbell, J. J., 244 Daur, T., 45
Bostrom, W. A., 78 Campbell, W. E., 119, 129, Davidenko, V. A., 18
Bougnut, J., 166 136,438 Davidov, B., 97
Bouillon, F., 209, 322 Caplan, D., 306 Davies, D. E., 143, 286
Brabers, M. J., 311 Carpenter, L. G., 218, 220 Davies, M. H., 214, 215,
Bradshaw, F. J., 63 Carter, R. E., 21, 78, 197, 263, 275, 279
Bragg, L., 41 273 DeBoer, J. H., 23, 177
Bramley, A., 45 Castellan, G. W., 165 deBrouckere, L., 168
Brasunas, A. De S., 138, Catalano, E., 74 deJong, J. J., 347
250, 309 Cathcart, J. V., 166,244 Dell, R. M., 266
Brattain, W. H., 93,430 Caule, E. J., 286 Demand, J. T., 236
Brauer, E., 404 Cerwenka, E., 248 Dench, W. A., 209
Brauer, G., 237 Chalmers, B., 42 Denissen, D., 72
Braune, H., 57, 378, 384 Charbonnier, F. M., 43 DeVries, R. C., 210
Brauns, E., 179, 297 Charlesby, A., 230 DeWald, J. F., 409, 411,
Bray, A. R., 409, 411, 416 Chaudron, G., 285 430
Brenkman, J. A., 70 Chauvenet, G., 195, 273 Dewhirst, D. W., 347
Brenner, S. S., 192, 251, Churaev, P. V., 139, 167 Dienes, G., 34
298, 309, 322 Cismary, D., 248 Dietz, G., 422
Breslin, J., 293 Cismary, G. D., 248 Ditzenberger, J. A., 70
Brick, R. M., 59 Clark, C. L., 314 Dixon, C. E., 33
Bridges, D. W., 239, 292 Clauser, H. R., 314 Dobrowolska, H., 53, 57,
Brinkman, J. A., 33 Cleland, B. B., 48 64
Brisi, C., 78 Clews, J. B., 143, 286 Doring, W., 396
Britton, C. F., 235 Cohen, M., 37, 53, 57, 63, Dorner, 0., 367
Brouckere, L. de, 322, 328 74, 130, 286, 306,421 Douglas, D. L., 236
Brower, T. E., 45 Collen, B., 210, 217 Dovey, D. M., 305
Brown, F., 129 Constable, F. H., 146 Dravnieks, A., 141, 158,
Brodsky, M. B., 263 Cope, J. 0., 208 268, 367, 386, 389
Bruno, M., 11 Coquelle, 0., 262, 275
Drechsler, M., 43
Brutzyk, M., 57 Cornelius, H., 59, 308
Drickamer, H. G., 72, 74
Buckle, H., 29, 57, 59 Cornet, I., 200
Drott, J., 80, 379
Bucknall, E. H.. 159 Correa da Silva, L. C., 29
Druck, J. U., 333
Bucur, E., 48 Cottrell, A. H., 42
Druyvesteyn, M. J., 337,
Buessem, W. R., 15, 212 Couling, S. R. L., 41
345, 347
Buffington, F. S., 37 Couper, A., 406
Bugakow, W., 41,51, 57 Duhl, D., 53
Cox, B., 234, 236
Bungardt, W., 59, 308 Duhlis, E. J., 313
Croatto, U., 11
Burgers, J. M., 41 Dumbgen, G .. 78
Crooks, H. N., 78
Dunlap, W. C., 70
Burns, R. p .. 18 Cubicciotti, D., 228, 239,
Dunn, C. G., 42
Burton, H. H., 45 241, 263
Dunn, E. L., 236
Busch, G.. 18 Cuff, F. B., 222
Dunn, J. S., 51, 146, 197,
Butler, S. R., 15,212 Czerski, L., 334, 373, 388
246, 322, 326
Butsik, M., 61
Byron, E. S., 67 Dalgaard, S. B., 235, 236 Dunnington, B. W., 140
Damask, A., 34 Dunwald, H., 146, 161
Cabrera, N., 7, 101, 109, Danielson, G. C., 78 Dushman, S., 72
128, 249, 409 Dankov, P. D., 139, 167 Duval, C., 146
Cahn, R. W., 217 Darken, L. S., 33,47, 275, Duwez, P., 227
Callendine, G. W., 74 294 Dyess, J. B .. 438
Author Index 443

Earle, M. D., 17, 215 Fitzer, H., 248 Greenwood, N. N., 21


Ebert, H., 53, 64, 318 Flint, H. G., 18, 387 Grieveson, P., 294
Ebisuzaki, Y., 78 Floe, C. F .. 222 Grimley, R. T., 18
Echeverria, A., 271 Foex, M., 17 Grimley, T. B., 118, 119,
Eckert, R. E., 72, 74 Foley, R. T., 311, 333, 380 129,411
Edwards, O. S., 273 Fontana, M. G., 140, 186, Grobe, A. H., 361
Eggert, S., 377 223, 250, 252 Gr~nlund, F., 167
Ehrenberg, D. M., 389 F~rland, T., 17, 212 Groves, W.O., 210
Ehrlich, p .. 15,220 Forman, R., 66 Grube, G., 63, 66
Eichenauer, W., 45,51,57, Franck, U. F., 405, 408, Gruber, H., 389
59 422,426 Gruhl, W., 272
Eisenloeffel, A., 43 Frankenburg, W. G.. 166 Grunert, A., 191, 389
Eisenmann, L., 22 Fraunfelder, H., 42 Grunewald, H., 17, 22,215,
Elam, C. F., 190 Freche, H. R., 59 221
Eley, D., 406 Frederick, S. F., 200 GrUnewald, K., 89, 113,
Elze, J., 405 Frenkel, J., 8, 148 146, 160, 161, 162, 170,
Engel, M., 170, 330 Fricke, R., 318 173, 202, 315, 350, 437
Engell, H. J., 26, 97, 113, Friedrichs, H., 47 Gruzin, P. L., 63, 74, 186
116, 120, 137, 140, 175, Frohlich, K. W., 168, 268, Gulbransen, E. A., 48, 94,
302,400,430,435 328 95, 121, 146, 169, 173,
Estulin, G. V., 186 Fry, A., 63 174, 179, 184, 185, 186,
Etzold, H., 61, 296 Fryell, R. E., 333 187, 194, 195, 210, 213,
Evans, E. B., 271 Fueki, K., 169 219, 227, 228, 236,247,
Evans, U. R., 129,130, 132, Fuller, C., 70 248, 265, 273, 277, 285,
136, 139, 140, 141, 143, 301, 322,433
146, 156, 254, 257,286, Gallisti, E., 248 Gundermann, J., 21, 161,
287, 288, 390, 403, 404, Garcia-Verduch, A., 78 162
405, 417, 418, 438 Garrett, C. G. B., 430 Gunterschulze, A.,128,409
Eyring, H., 35, 82 Gatos, H. C., 53, 74, 166 Gurney, R. W., 11,91
Geary, A., 408 Gurnick, R. S., 271, 396
Farber, M., 389 Gegg, C. C., 45 Gurry, R. W., 275
Farnsworth, H. E., 122 Geld, P. W., 383 Gwathmey, A. T., 166,190,
Farraggi, H., 72 Gensch, C., 146, 148, 152, 265
Fassell, W. M., 164, 174, 154, 206, 295
239, 292 Gerds, A: F .. 241, 243
Fast, J. D., 35 Gerischer, H., 424, 430 Haase, 0., 279
Fedorov, G. B., 186 Germer, L. H., 137 Haase, V., 404
Feitknecht, W.. 164, 268 Gerzriken, S., 57, 63 Haayman, P. W., 6, 18, 20,
Fellows, C. H., 306 Gibbs, D. S., 264 130, 152, 177, 195
Fensham, P. J., 74 Gjostein, N. A., 42 Hackerman, N., 407
Finch, G. J., 264 GIang, R., 17 Hagel, W. C., 186
Fine, M. E., 239 Glasstone, S., 35, 82 Haissinsky, M. M., 78
Fischbeck, K., 146, 254, Gobrecht, H.. 430 Hallet, M. M., 390
257, 258, 274, 366, 367 Golubenko, E., 61 Hallowes, A. p .. 168, 390
Fischer, H., 4, 248, 266, Graevskii, K. M., 195 Ham, J. L., 45
404,405 Graham, R. L., 129 Hammen, H.. 21, 160, 350
Fischer, W. A., 21, 193 Grant, F. A., 17 Hammon, J., 128, 249
Fischmeister, H., 80 Grant, N. J., 215, 218,219, Hansen, M., 248
Fisher, D. F., 232, 234 222, 223, 227, 250 Hardt, R. W., 380
Fisher, J. C., 39,417 Greener, E. H., 239 Hart, E. W., 347
Fisher, J. E., 347 Greenhouse, H. M., 226 Hartman, W., 18
Author Index

Hauffe, K., 4, 6, 7, 14, 17, Hogarth, C. A., 17 Johns, C. R., 195, 273
18,20,21,22,23,26,27, Holliday, R. D., 405 Johnson, G. H., 17, 78
32,35,37,39,41,78,83, Hollnagel, M., 422 Johnson, R. D., 51, 74
85, 90, 91,97,103,110, Hollomon, J. H., 351 Johnson, T. L., 169
113, 116, 120, 123, 125 Holman, E. W., 226 Johnson, W. A., 57, 336,
127, 130, 132, 134, 135, Honjo, G" 328 345, 358
137, 146, 148, 152, 154, Hopkins, B. E., 146 Johnston, H. L., 210
161, 173, 175, 178, 193, Hopkinson, B. E., 235 Johnston, T. L., 328
206, 207, 212, 213, 215, Horn, L., 179, 192 Jolivet, E., 305
218, 221, 222, 235, 248, Horne, G. T., 31 Jominy, W. E., 265
254, 255, 258, 263, 266, Hubrecht, L., 168, 322,328 Jones, H., 406
269, 271, 275, 280, 284, Hudson, D. F., 159 Jones, W. D., 63, 236
295, 316, 345, 363, 370, Hudson, J. C., 274 Jost, W., 8, II, 32, 35, 53,
383, 387, 389, 404, 408, Huntington, H. B., 28, 32, 57, 78, 83, 95, 148, 156,
411, 419, 422, 423,435, 37,74 255, 257, 269, 275, 375,
437 Hurlen, T., 17, 212, 239 378, 379
Haul, R., 78 Juenker, D. W., 140
Hauttmann, A., 47 Ignatov, D. V., 191 Just, D., 78
Hawkins, G. A., 306 limori, T., 279
Hayes, E. T., 230, 390 llschner, B., 27, 91, 103, Kabalow, B. N" 428
Hebb, M. H., 375 106, 116, 120, 132, 135, Kahn, 0., 378
Hedvall, J. A., 78 172, 175, 423, 435 Kaiser, W., 293
Heilmann, E. L., 23 Inghram, M. G., 18 Kalman, L., 263
Heindlhofer, K., 140 Inouye, H., 237 Kalvenes, 0" 142,175
Herbert, T. M., 159 Ipavic, H" 386, 389 Katz, E., 78
Herrmann, J., 61 Iseler, G. W., 236 Katz, 0., 398
Hersh, H. N., 57 Ishibashi, H., 169 Kayser, F., 386
Hertz, J., 303 Isobe, M., 396 Kehl, G. L., 48
Hertzrucken, S. D., 61 Keil, A., 61, 74
Hessenbruch, W., 168,191, JaCOb, H., 296 Kessler, H. D., 248
251, 305, 389 Jacobs, P. W. M.,409,411, Kidson, G. V., 74
Heumann, T., 63 416 Kieffer, R., 227, 248, 395,
Hevesy, G. Von, 61, 74, 78 Jaenicke, W., 4, 41, 78,80, 397
Hickman, J. W., 169,179, 139, 254, 370 Kilkson, R., 161
187, 210, 248, 273,301, Jaffee, R. I., 217, 220,223, Kingery, W. D., 78
322 226, 239 Kinna, W., 224, 399
Hicks, L. C., 63 Jager, W., 301 Kirkbride, L. D., 235
Hieber, W., 407 Jaquet, P., 190 Kirkendall, E. 0., 29
Hillert, M., 78 Jardinier-Offergeld, M., Kiselev, A. A., 234
Himmel, L., 78, 263, 275 209 Kiwit, K., 322
Himmler, W., 155 Jaumot, F. E" Jr., 33, 67, Kjollesdal, H., 125, 213,
Hinnuber, J., 399 74 215, 238, 239
Hintenberger, H" 22 Jeannin, Y., 303 Klaiber, F., 18, 374
Hirano, K., 53, 57, 63 Jedele, A., 64, 66 Klein, E., 411
Hirschberg, R., 135 Jellinghaus, W., 366 Klepper, H., 236
Hoar, T. P., 370,405,412 Jenkins, A.E.,48,210,215, Klokholm, E., 35
Hocart, R., 366 217, 218, 224, 232 Klopp, W. D., 239
Hoch, M., 210 Jenkins, J., 305 Knacke, 0., 135
Hochberg, B. M" 21 Jepson, W. B., 247 Kobayashi, H., 422
Hofer, E., 72 Jessin, O. A" 383 Koch, E., 9, 13, 21, 149,
Hoffman, R. E" 39, 40, 74 Jitaka, J., 186 152, 195, 282
Author Index 445

Kofstad, P., 123, 125, 142, Laubmeyer, C., 262 Maak, F., HI, 168, 295,
212, 213, 215, 217, 218, Laves, F., 13 328, 334, 357
221, 231, 232, 237,238, Lavine, M. C., 333 Machatschki, F., 23
239, 243, 244, 269 Law, J. T., 263 Maesen, F. van der, 70
Kohlschutter, V., 148 Lawless, K. R., 166, 265 Magneli, A., 210, 217
Kolbl, F., 395 Lawson, A. W., 74 Mahla, E. M., 301
Kolotyrkin, Y. M., 422 Layer, E. H., 228 Maier, M. S., 32
Komarewsky, V. I., 166 Lazarus, D., 32, 57 Makolkin, I. A., 248
Konev, V. N., 194 Leak, G. M., 35, 47 Mallet, M. W.,48,120,229,
Kordes, E., 23 LeClaire, A. D., 32, 36, 37 230, 241, 243
Kornew, J. W., 74 Lee, J. K., 112, 174, 195, Mann, E., 34
Kornilow, J. J., 305 204 Mansfield, R., 21
Korolev, S. I., 234 Leibowitz, L., 242 Mapother, D. E., 78
Kortright, J. W., 313 Leistikow, S., 139 Mardeshev, S .. 383
Koster, W., 148 Leontis, T. E., 264, 265 Marek, L. F., 389
Kottmann, A., 29, 31 Leroux, J. A., 170 Martin, A. B., 51, 66, 74
Kovalskii, A. E., 227 Leslie, W. C., 250 Martin, J. W., 337, 347
Kozhevenikov, A. V., 234 Letaw, Jr., H., 51, 70, 74 Martin, R. L., 21
Kracek, F. C., 375 Levesque, P., 239, 263 Masing, G., 146,403
Krahenbuhl, E., 148 Levin, R. L., 169, 326 Matano, C., 51, 64, 66
Kroger, F. A., 23 Levine, H. S., 42 Matejek, R., 411
Krohnke, 0., 403 Li, C. H., 169, 328 Mathewson, C. H., 339,343
Kronmuller, H., 34 Lichter, B. D., 329, 380, Maurer, R. J., 21, 78, 146,
Krudtaa, O. J., 217, 218, 388 434
237 Lidiard, A. B .. 12, 83 Mayer, A., 11
Krueger, H., 57 Liebenwirth, F., 63 Maykuth, D. J., 217, 220,
Krysko, W. W. G., 390 Liempt, I., 72 223,226
Kryukov, S. N., 57 Liempt, J., 53 Mayne, J. E. D., 130
Kubaschewski, 0., 51, 53, Lindner, B. R., 6, 15, 78, Maynor, H. W., Jr., 224
64, 146, 170, 173, 209, 282 Mead, H. W., 72, 74
308, 315, 318, 365, 386 Lionetti, F., 42 Meakin, J. D., 35
Kuikkola, K., 232,296,308, Lipson, H., 273 Meechan, J. C., 33
372 Littleton, M. J., 11, 95 Mehl, R. F., 29, 31, 45,46,
Kunzig, H .. 45, 57 Liu, T., 74 47, 51, 59, 63, 74, 78,
Kuper, A., 51, 74 Lofquist, H., 314 129, 263, 265, 275, 279
Kurdjumow, G. W., 74 Lohberg, K., 219 Meigs, P. S., 263
Kurtz, A. D., 74 Lomakin, G. D., 194, 273 Meijering, J. L., 123, 250,
Kuylendstierna, U., 210, Lomer, W. M., 32 300, 337, 345, 347
217 Longo, T. A., 313 Meinhardt, 0., 430
Lord, S. S., Jr., 408, 423 Menzel, E., 167
Laidler, K. J., 35, 82 Lorenz, G., 21, 193 Meussner, R. A., 78, 140,
Laird, J. G., 61 Loriers, J., 241 172, 271, 383
Lambert, V. E., 67 Lossow, W. W., 428 Meyer, R. E., 74
Lander, J. J., 393 Love, G. R., 31 Middleton, A. E., 228
Landsberg, P. T., 137 Lucas, G., 192, 201, 253 Mikulski, J., 371
Landsberg, R., 422 Ludwig, R., 29, 31 Miley, H. A., 129,130,136,
Lange, E., 135 Lunde, G., 142 417,438
Langhammer, G., 404 Lurin, B., 78 Miller, P, H., 15, 74
Langmuir, I., 41 Luschkin, G. P., 210, 219, Misch, R. D., 23"6
Larson, B. M., 45, 140 Lustman, B., 129,191,263, Mitoff, S. P., 19
Lashof, T. W., 18 265, 301, 305 Miyake, S., 169, 186, 190
446 Author Index

Mohring, H., 148, 367, 370 Neskutschaw, W., 51 Peterson, R. C., 292
Mole, R., 366 Neumann, A., 367 Pettit, F. S., 200, 262
Moore, W. J., 78, 93, 112, Neundeubel, K. L., 254, 274 Pfefferkorn, G., 80, 383
165, 170, 174, 196, 204, Neveling, V., 384 Pfeiffer, H., 21, 176, 178,
350, 378 Newkirk, A. E .. 396 191, 192, 215, 221, 255,
Moreau, J., 167, 187, 190, Neymark, R. S., 367 258, 262, 275, 280
277, 302, 303, 310, 322, Nickerson, R. A., 42 Pfeiffer, I., 41,78,192,254,
363 Nifontoff, N., 166 370, 388, 419, 422
Morill, W., 380 Nishimura, H., 169 Pfeil, L. B., 140, 274
Morin, F. J., 282 Nix, F. C., 33, 74 Phalnikar, C. A., 230, 271
Morton, M. C., 21 Nizhel'skii, V. F., 200 Phelps, B., 239
Morton; P. H., 211, 213 Noldge, H., 257 Philibert, J., 72
Mott, N. F., 7, II, 91, 95, Norman, N., 237 Philips, A., 59
97, 109, 126, 128, 132, Norton, J. T., 243, 289, 392, Pickett, J., 122
205,406,409,411,412 399 Pidgeon, L. M., 212
Mrowec, S., 172, 334, 335, Norwich, A. S., 38 Piene, K., 175
371, 373, 378, 384, 388 Norwick, A. S., 51, 57, 59, Pilling, N. B., 4, 145, 159,
Mueller, H., 237 61 164, 248, 264, 268, 323,
Muller, E. W., 43 Nowotny, H., 227 350
Muller, G., 57 Nurse, T. J., 143, 286 Pirani, M., 41, 72
Muller, W. J., 408 Nye, J. F., 41 Plate, W., 381
Munster, A., 227, 228 Plateau, J., 190
Murakami, T., 386 O'Briain, C. D., 371 Pocher, W., 343
Murin, A., 78 Obrutschewa, A., 61, 74 Pohl, R. W., 15
Murphy, D. W., 140, 265 Odell, F., 227 Pool, M. L., 74
Myschkin, V. A., 234 Ogden, H. R.,217,220,223, Portevin, A., 305
MacCallum, C. J., 42 226 Portnoy, W. M., 70
MacKenzie, J. D., 140, 170 Okamoto, G., 422 Posnjak, E., 23, 201
MacNairn, J., 122 Okkerse, B., 41, 74 Potter, A. A., 306
Mcafee, K. B., 70 Olson, R., 66, 228 Preece, A.. 192, 207, 253,
McBride, C. C., 226 Ong, J. N., Jr., 174, 239, 309
McCandless, E. L., 279 244, 247 Preis, H., 253
McCullough, H. M., 252 Osthagen, K., 232, 244 Preston, G. D., 169
McDonald, H. H., 141, 158 Otter, M., 167 Pretet, E., 305
McKay, H. A. C., 53, 74 Owen, E. A., 63 Price, L. E., 168, 438
McKewan, W., 164 Pringle, J. P. S., 129
McLintock, C. H., 239 Paidassi, J., 80, 196, 265, Prote, H. A., 232, 234
McMillan, W. R., 184, 277 271, 274 Pry, R. H., 347
McPherson, D. J., 223 Palmer, W. G., 148 Pryor, M. J., 130
Mcquillan, A. D., 223 Parfitt, G. D., 78
Park, R. M., 243 Quarrell, A. G., 191, 264,
Nachtigal, E., 243 Parker, E. R., 42 311
Nachtrieb, N. H., 74 Paschke, M., 47
Nagayama, M., 421, 422 Pavel, R. E., 244 Radavich, J. F., 312
Nagel, K., 21, 89 Pawlek, F., 363 Raether, S., 363
Nair, W. N., 218 Pebler, A., 45,51,57,59 Rahlfs, P., 374
Neilson, N. A., 301 Pemsler, J., 48 Rahmel, A., 179, 263, 271,
Nelson, B. J., 322, 326, 331, Penning, P., 70 297, 301, 310, 383, 387
334 Peretti, E., 57 Ransley, C. E., 344
Nelson, H. R., 32 Peschanski, D., 78 Rapp, R. A., 351, 355, 361
Nelting, H., 72 Peter, S., 384 Ratchford, R. J., 381
Author Index 447

Rathenau, G. W., 250, 300 Sagel, K., 227, 228 Seitz, F., 30, 32, 33, 35,37
Raub, E., 170, 330, 381 Sagrubsky, A. M., 53,74 Selikson, B., 78, 93, 165,
Read, W. T., 41 Salzer, F., 254, 258, 274 350
Reavell, F. R., 220 Samans, C. H., 367 Seltzer, M. S., 23, 78
Redington, R. W., 78 Samsonov, G. V., 72 Severiens, J. C., 70
Reif, 0., 265 Sandor, J., 41, 72 Seybolt, A. U., 186
Reinacher, G., 382 Sartell, J. A., 169, 328 Shevlin, T. S., 226
Reinhold, H., 78, 148, 156, Sato, N., 422 Shewmon, P. G., 31, 74
367, 370, 378, 379, 381 Sato, R., 278 Shibata, N., 386
Reynolds, N. B., 72 Savelsberg, W., 386 Shim, M. T., 78
Rhines, F. N., 51, 59, 74, Schafer, H., 296 Shirley, H., 253, 390
264, 265, 322, 326, 331, Schamp, H. W., 78 Shirn, G. A., 74
334, 336, 339, 343, 345, Schatz, J., 155 Shockley, W., 41, 70, 93,
350, 358, 361, 363 Scheil, E., 314, 322 165
Rhodin, T. N., 129, 136, Schenck, H., 302 Shorina, E. G., 234
166,433 Schenk, W. E., 45 Shumilina, S. V., 194
Richards, J. R., 78 Scheuble, W., 132 Siebel, G., 146, 309
Richardson, F. D., 21, 78, Schibbe, G., 377 Silvester, D. "R., 235
197,273 Schichtel, K., 191 Simkovich, G., 23, 78, 157
Richardson, L. S., 215, 218, Schiller, P., 34 Simmons, R. D., 34
219 Schlamp, G., 228 Simnad, M.. 243, 245, 275,
Rickert, H., 68, 80, 184, Schlapfer, P., 253 279, 398
266, 335, 367, 368, 371, Schleicher, H. W., 219 Sims, C. T., 239
372, 378, 379, 381, 383, Schlosser, E. G., 20 Sirotkin, B., 57
388 Schmahl, N. G., 302 Sisler, H. S., 226
Rideal, E. K., 166, 256 Schmalzried, H., 24, 25, Slifkin, L., 51, 57, 70, 74
Ridolfo, V. C., 74 232, 304 SlUSS, J. A., 78
Roach,J. D., 226, 399 Schneider, A., 308 Smeltzer, W. W., 124,248
Roberson, A. H., 230, 390 Schneider, K., 63 Smigelskas, A. D., 29
Roberts-Austen, W. C., 61 Schnitzlein, J. G., 232, 234, Smith, C. S., 351
Robertson, R. H., 230, 390 242 Smith, G. C., 337, 347
Rohrig, I. A., 306 Schone, E., 68 Smith, G. V., 313
Rollin, B. V., 57 Schottky, W., 4, 6, 9, 26, Smith, J. F., 78
Romeyn, F. C., 6, 18,20, 32, 83, 85, 90, 97, 134, Smith, W., 78
130, 152, 177, 195 155, 171, 180, 269, 375, Smithells, C. J., 66
Rompe, R., 104 404,424 Smoluchowski, R., 40, 41,
Rooney, E. E., 274 Schpikelmann, A. J., 305 47
Rosenber, A. J., 333 Schroeder, W., 265 Snoek, J. L., 35
Rosner, U., 236 Schulte," F., 47 Solberg, H. L., 306
Ross, R., 74 Schulz, E. H., 322, 430 Sominski, M. J., 21
Rostoker, W., 209, 223 Schwabe, K., 422 Sonder, E., 51, 57, 74
Rouse, G. F., 66 Schwartz, C. M., 235 Sparks, M., 70
Roy, R., 210 Schwarz, W., 418 Spauschus, H. 0., 380
Ruder, R. C., 45, 66, 74 Schwarzkopf, P., 397 Spenke, E., 26
Rudiger, 0., 224, 399 Secco, E. A., 78 Spetzler, E., 389
Rudolph, J., 231 Seeger, A., 34 Spilners, A., 243, 245, 398
Ruf, K., 389 Seidel, H., 156, 370, 379, Spinedi, P., 168
Ruka, R., 265, 277, 285 381 Spooner, N., 329
Ruppert, W., 227 Seigle, L. L., 31 Stahelin, P., 18
Ruter, H., 375 Seith, W., 29, 30, 31, 32, Stanley, J. K., 35,45
Rybalko, F., 41,51,57 45, 51, 57, 61, 72, 74 Stasiw, 0., 9, 68
448 Author Index

Stevens, J., 167, 322 Trapnell, B. M. W., 118, 129 Vogel, R. C., 232,234,242,
Stockdale, J., 288 Trivich, D., 161 343, 396
Stone, F. S., 266 Trommsdorf, G., 53 Vogt, E., 406
Stossel, W., 167 Tronstad, L., 405, 418, 439 Vollheim, G., 422
Stranski, I. N., 43, 135 Tubandt, C., 10, 78, 148 , Volmer, M., 286
Stringer, J., 239 157, 367, 370, 374, 377, vonGoldbeck, 0.,173,308,
Stronski, I., 371 378 315, 365, 386
Stroud, E. G., 132, 204 Tucker, J. P., 370 Vorsohilova, Z.A., 195, 273
Struthers, J. 0 .. 70 Turkdogan, E. T., 294
Sturdy, G. E., 124, 243 Turnbull, 0 .. 39, 40, 74, Waber, J. T.,1l9,124, 213,
Suhrman, R., 43, 406 351 243
Susko, F. S., 78 Turner, G., 45 Wagner, C., 5,6, 8, 9, 13,
Suzuoka, T., 63 Tuul, J., 122 16, 17, 18, 19, 21,22,23,
Svec, H. J., 264 Tylecote, R. F., 119, 129, 24, 32,48, 68, 89, 90,92,
Swalin, R. A., 36, 66 159, 165, 166 103, 113, 141, 146, 148,
Swift, R. E., 224 Tyutyunnik, A. D., 186 152, 160, 161, 162, 165,
Sykes, C., 45, 253, 390 170, 173, 174, 179, 180,
Uhlig, H. H., 122, 309, 403, 185, 195, 197, 202, 204,
Talbot, J., 167, 256, 258, 405,407,408,423 207, 229, 232, 243, 255,
265 Ulich, H., 180 257, 262, 266, 267, 269,
Tammann, G., 6, 145, 146, Umanskii, Y. S., 227 282, 289, 293, 294, 296,
148,309 Upthegrove, C., 140 304, 315, 329, 330, 334,
Tausch, J., 78 Ure, R. W., 11 343, 350, 351, 352, 354,
Teltow, J., 9, 68, 78 365, 369, 370, 372, 374,
Terrien, J., 128, 249 Valensi, G., 159, 194, 269 375, 376, 380, 381, 388,
Tetlow, J., 149, 152, 155 Vallee, M. G., 196 391, 392, 399, 411, 437,
Thiel, G.. 139 Van Bueren, H. G., 32 Wagner, E., 382
Thomas, D. E., 51, 208, van Duzer, R. M., 306 Wagner, J. B., 23, 78,169,
235, 319, 337 vanGeel, W. C., 409 200, 262, 326, 369
Thomas, H., 191 van Houten, S., 21 Wajda, E. S., 41, 74
Thomas, G. J., 168,438 van Ormondt, J., 177 Wallichow, K., 373
Thomas, J. M., 329 van Santen, J. H., 23 Wallwork, G. R., 217, 218,
Thomas, U. B., 119, 129, Vaughan, D. A., 235 232
136, 438 Verheijke, M. L., 123 Wang, Ch.-Ch., 223
Thomas, W. R., 35,47 Verhoogen, J., 70 Wanklyn, J. N., 235, 236
Thomassen, L., 329 Vernon, W. H. J., 132,143, Ward, R. A., 74
Thyssen, M. H., 314 286, 390 Wasilewski, R. J., 48
Tichenor, R. L., 196 Vero, J. A., 51 Webb, W. W., 243, 289, 392,
Tiley, P. F., 266 Verrijp, M. B., 35 399
Tinklepaugh, J. R., 226 Verwey, E. J. W., 6,18, 20, Weber, A., 286
Tipton, C. R., 243 23, 128. 130, 152, 177, Weber, E., 272
Todt, F., 405 195, 409 Weddle, M., 253
Toler, H. R., Jr., 222 Vetter, K. J., 130,131,404, Weeton, E. W., 72
Tomaschow, N. D., 410 405, 410, 418, 419, 422, Wehefritz, V., 78, 369
Tomizuka, C. T., 51, 57, 425 Weil, J. A., 74
74 Vher, O. J., 63 Weil, K. G., 411, 418, 422,
Torrisi, A. F., 409, 418 Vierk, A. L., 17, 206 427
Toth, R. S., 161 Virmilyea, D. A., 409,417, Weininger, J. L., 230
Traktenberg, 1. S., 194 418 Weisert, E. 0.,308
Tranckler-Greese, R., 17, Vladimirova, M. G., 200 Weisz, P. B., 26
22, 215, 221 Voce, E., 390 Weitbrecht, G., 318
Author Index 449
Weizel. W•• 104 Wiederholt. W.• 4. 248. 266. Wycoff. R. W. G•• 279
Weizer. V •• 34 404 Wygant. J. F .• 171
Wells. C •• 45. 46. 47. 63 Wiley. J .• 15 Wysong. W. S•• 243
Werber. T •• 334. 371. 373. Wilkins. F •• 197.256
384.388 Wilkins. N. J. M.• 235 Yearian. H. J .• 312
Wert. C. A•• 34 Williams. A. J .• 217 Yinger. R •• 262
Wertenstein. M. L •• 53. 57. Williams. E. L.. 204 Young. F. W.• 166
64 Wilson. R. E •• 226 Young. L •• 409. 411. 416.
West. J. R•• 386 Winterbottom. A. B .• 146. 417
Wetternik. L •• 313 274.439
Wever. H•• 30 Wise. E. M•• 243 Zemany. P. D •• 230
Weyl. W. A •• 17 Wittingham. G•• 390 Zener. C •• 28. 34. 35. 37.
Whipple. R. T. P .• 40 Wood. W. P •• 265 38. 82. 95
White. A. H.• 137. 389 Wullhorts. B •• 381 Zhukovitsky, A. A., 57
Whitmore. D. W,. 239 Wyant. J. L.. 227 Zimen, K. E., 78, 174, 179
Zogagintsev, J., 64
Zwikker, C., 41
Subject Index

Alum~num. 5. 128. 129. 248-250 -in ionic crystals. 68-72. 278-284


Ambipolar diffusion. 102-104 -in metals. 45-67
Diffusion tables. 32. 45-78
Barium. 264
Boundary layers. 25-28. 97-109. Electric fields in metal oxidation.
116. 199-200. 267 97-137
Bubble formation in oxide layers. Electrical conductivity. 13. 16. 17.
139-143 19-21. 161
Electrochemical solid cells. 370-
Carbide. 392-401 379
Catastrophic oxidation. 250-253 Emf measurements. 230. 371-377
Cavity formation in oxide layers. Energy of formation of oxides. 5. 344
137-143
Cerium. 241 FeO. 280. 283
Chemisorption. 26.41-102.113-114. Fe304. 280. 283
129-131. 259 Frenkel defects. 9
-and inhibition of corrosion. 407-
408 Grain boundary diffusion. 38-43
Chromium. 193-194
Cobalt. 194-201. 273 Halogenation of silver. 89. 148-157
Cobalt alloys. 200-201 Heat of formation of sulfides. 366
Copper. 5. 89. 92-93. 118-119. 136-
137. 159-170. 265-269 Internal oxidation. 184. 335-364
Copper alloys. 167-170. 336-361 Interstitial sites (in silver halides).
Copper-nickel alloys. 323-326 9ff
Copper-palladium alloys. 319-320 Ion-conducting oxide layers. 148-
Copper-zinc alloys. 326-327 158
CU20. 118-119. 267-269 Iron. 255-263. 267-288
Cubic rate law. 3.115-125.175.221. Iron-chromium alloys. 301-314
228 Iron-molybdenum alloys. 298-301
Curie temperature and oxidation. Iron-molybdenum-nickel alloys.
309 252
Iron-nickel alloys. 310-311
Diffusion. 3
-of interstitial sites inmetals.34-36 Kirkendall effect. 29
-of iron in iron oxides. 279-284
-in metals. 28ff Lattice defect phenomena. 8ff
Diffusion coefficients and tarnishing -in ionic crystals. 9-25. 148. 158.
processes. 91-94 160
-of gases. 45. 48 -in metals. 28ff

450
SUbject Index 451

-in space-charge layers, 25-28, Oxidation rates (see individual sub-


97-102 stances)
Lattice defect theory, 6, 9-25 Oxide layer, formation in an electro-
Lattice defects in AgBr, 9-13, 148- lyte, 402-430
150 Oxide needle formation, 80
-Cr203' 194 Oxygen pressure dependence of
-Cu20, 160, 266 oxidation, 90, 115, 121, 160-165,
-FeO, 280, 297 174-178,197-198,203
-NiO, 18-21, 176, 191
-PbCI 2, 157-158 Parabolic rate law, for diffusion,
-PbS, 23 3, 82-91, 148-149, 175, 206-208,
-spinels, 23-25 213-221
-Ti0 2,221 -, for field transport, 109-115
-ZnO,15-17 Passive layers, mechanism of for-
Lead, 272, 273 mation, 409-424
Linear rate law, 237, 239, 253-263, Passivity, 130-131,404-424
288-293 Phase boundary reactions, 3, 253-
Logarithmic rate law, 3, 130-137, 267
204-205, 212, 221, 286-287 Phase diagram,AgBr + MeBr2' 152-
153
MagneSium, 264-265 -, Fe + Ni + 0, 310-311
Manganese, 271-272 -, Fe + 0, 275-276
Manganese-carbon alloys, 396-397 -Ti + 0,210
Marker experiments, 29, 172, 284
Mass accumulation in oxidation, Radioactive isotopes, use of, 91-94,
137-143 165, 199, 278-284
Matano plane, 31 Reciprocal-logarithmic rate law, 3,
Measuring methods, 431-439 125-131
Metal-carbon alloys, 392-401 Ring diffusion, 26-27
Metal diffusion and scaling rate,
Scaling coefficient and diffusion, 91-
314-331
94, 280
Molybdenum, 244-246
Scaling constant, transformation of,
Mo0 2, effect of, 251
145-148
Multiphase protective layers, 187-
Scaling layer, formation of, 184-193,
193, 267 -288
210-211, 267-268, 272-276, 299-
Needle-shaped growth of oxides, 81 302
Nickel, 94-97, 121-122, 133-134, Scaling processes (definition), 7
171-193 Scaling theory (according to Wag-
Nickel-carbon alloys, 393-394 ner),87-91
Nickel-chromium alloys, 176-177, Schottky defects, 9, 229
179-192,363 Self-diffusion, 91-94, 165, 198
Nickel-molybdenum alloys, 192-193 Self-diffusion coefficients in ionic
Nickel-platinum alloys, 315-319 crystals, 75-78
Nickel-silver alloys, 176-177 -of metals, 73-74
Niobium, 236-239 Silver, 89, 148-157, 250, 251
Nitridation of titanium, 227-228 Silver alloys, 362
-uranium, 243 Silver-cadmium allOYS, 151-156
Silver-nickel alloys, 176
Oxidation apparatus, 431-438 Soap bubble model (of Bragg), 41-42
Oxidation rate, calculation of absa- Solution current of metals with pas-
lute, 94 sive layers, 424-430
452 Subject Index
Space charge. 25-28. 97-109. 110. Titanium. 119-120. 209-221
116. 199-200 Titanium-carbon alloys. 224-226.
Spinel lattice. 23-24 398-401
Spinels, 23-25, 186-191, 301-314 Titanium, nitridation of. 227-228
Steel. 298-314 Transport processes in space-
Sulfidation of alloys, 379-381. 386- charge layers. 102-109
391 Tungsten, 246-248. 289-292
-of metals. 365-379. 382-386 Tungsten carbide, 395-396
Surface charge. 110. 111
Uranium. nitridation of, 241-243
Tarnishing layers, formation of thin ,
109-125 Vacancies (in silver halides). 9ff
-,formation of very thin. 125-137, Vacancy diffusion in metals, 32-34
410-417 V 205. action of. 253
Tarnishing processes (definition). 7,
109 Zinc. 93-94. 115. 202-209
-theory. 102-137 Zinc alloys. 206-207
Thorium, 239-241 Zirconium. 228-236

You might also like