CDNA08652ENC_001.pdf

HIGH TEMPERATURE

MATERIALS CORROSION IN

COAL GASIFICATION ATMOSPHERES

A Publication of the CEC High Temperature Materials

Information Centre, Petten (N.H.), The Netherlands

Edited by

J. F. NORTON

ELSEVIER APPLIED SCIENCE PUBLISHERS

HIGH TEMPERATURE MATERIALS CORROSION IN COAL GASIFICATION ATMOSPHERES

Proceedings of a seminar held at the JOINT RESEARCH CENTRE of the COMMISSION OF THE EUROPEAN COMMUNITIES Petten Establishment, The Netherlands

and organised by the

CEC High Temperature Materials Information Centre, Petten, The Netherlands

HIGH TEMPERATURE MATERIALS CORROSION IN

COAL GASIFICATION ATMOSPHERES

Edited by

J. F. NORTON Joint Research Centre, Petten Establishment, The Netherlands

PARL EUROP. Biblioth.

N. C.

eJ& S65Z EA) PUBLISHERS

ELSEVIER APPLIED SCIENCE PUBLISHERS LONDON and NEW YORK

ELSEVIER APPLIED SCIENCE PUBLISHERS LTD Ripple Road, Barking, Essex, England

Sole Distributor in the USA and Canada ELSEVIER SCIENCE PUBLISHING CO., INC.

52 Vanderbilt Avenue, New York, NY 10017, USA

British Library Cataloguing in Publication Data

High temperature materials corrosion in coal gasification atmospheres 1. Materials at high temperatures 2. Corrosion and anti-corrosives I. Norton, J. F. 620.1217 TA407 ISBN 0853342415

WITH 57 ILLUSTRATIONS ECSC, EEC, EAEC, BRUSSELS AND LUXEMBOURG, 1984

Publication arrangements by: Commission of the European Communities, DirectorateGeneral for the Information Market and Innovation, Luxembourg

EUR 8652 EN LEGAL NOTICE

Neither the Commission of the European Communities nor any person acting on behalf of the Commission is responsible for the use which might be made of the following information.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording,

or otherwise, without the prior written permission of the copyright owner. Printed in Great Britain by Galliard (Printere) Ltd, Great Yarmouth

Preface

This book contains the proceedings of a seminar organised by the Commission of the European Communities at the Joint Research Centre, Petten Establishment, in The Netherlands. It is part of a continuing series of events which the 'Information Centre' of the High Temperature Materials Programme organizes with an overall objective of encouraging the exchange and dissemination of scientific knowledge at a European level.

The study of high temperature corrosion in aggressive gaseous atmospheres has intensified over the past decade as the emerging and established industrial nations of the world make increasing demands upon the more efficient use of natural resources such as coal, oil and gas. This informal seminar concentrated upon the high temperature corrosion of materials in coal gasification-type environments with particular emphasis on the reactants carbon, oxygen and sulphur.

A foundation is laid by the first two contributions which cover the theoretical considerations of high temperature gaseous corrosion of metals and alloys in single and multi-reactant atmospheres. The following two presentations build on this foundation and American and European authors consider the modes and extent of the corrosive degradation of metallic components used in the con-struction of coal conversion plant. Finally, an overview is given concentrating on the promise and potential offered by ceramics for use in significant regions of gasification systems.

VI PREFACE

As scientific co-ordinator of this seminar (J.F.N.) and as 'Information Centre' project leader (M.M.) we would like to record our sincere thanks to P. Kofstad, H. J. Grabke, J. Stringer, H. J. Schrter, W. Schendler and H. Weber for their efforts in preparing and presenting their contributions, as well as the delegates to the seminar for their active participation. These were essential elements and ensured the success of the seminar. We also acknowledge the valuable assistance of F. Van der Smaal and Mrs S. Haesen-Sommer with the administrative arrangements.

J. F. NORTON M. MERZ

Contents

Preface .

Introduction ix M. VAN DE VOORDE (Division Head, High Temperature

Materials Programme, Joint Research Centre, Petten Establishment, 1755 ZG Petten, The Netherlands)

1 High Temperature Corrosion in Single-Reactant Gaseous Environments 1

P. KOFSTAD (Department of Chemistry, University of Oslo, PB 1033, Blindem, Oslo 3, Norway)

2 High Temperature Corrosion in Complex, Multi-Reactant Gaseous Environments 59

H. J. GRABKE (Max-Planck-Institut fr Eisenforschung, 4000 Dsseldorf, Max-Planck-Strasse 1, Postfach 140 260, Federal Republic of Germany)

3 Materials of Construction in Coal Gasification Systems: US Experience, Part IMetallic Materials . . 83

J. STRINGER (Electric Power Research Institute, 3412 Hillview Avenue, PO Box 10412, Palo Alto, California 94304, USA)

Vill CONTENTS

4 High Temperature Corrosion of Metallic Heat Exchanger Materials in a Fluidized Bed Gasifier . . . 1 0 3

H.-J. SCHRTER (Bergbau-Forschung GmbH, Franz-Fischer Weg 61, D4300 Essen 13, Federal Republic of Germany)

W. SCHENDLER and H. WEBER (Mannesmann-Forschungsinstitut GmbH, Postfach 251167, 4100 Duisburg 25, Federal Republic of Germany)

5 Materials of Construction in Coal Gasification Systems: US Experience, Part IIRefractory and Ceramic Materials 117

J. STRINGER (Electric Power Research Institute, 3412 Hillview Avenue, PO Box 10412, Palo Alto, California 94304, USA)

Index 133

Introduction

M . V A N D E V O O R D E

Division Head, High Temperature Materials Programme, Joint Research Centre, Petten Establishment,

The Netherlands

Dependable energy sources for the future are a major concern of the European Communities. Their development requires the applica-tion of high technologies which are strongly dependent on materials availability. One of these developing technologies is coal gasification and, as will become apparent, corrosion in such atmospheres poses severe materials problems, making it a very appropriate subject for this seminar.

The energy sources available to the industrial world are mainly drawn from three sourcesfossil fuels, water power and nuclear power. Water power resources are largely exploited already, at least within reasonable distances of the points of energy requirement; nuclear power is subject to strong opposition on environmental grounds; and mineral oil is a diminishing source, as well as being a political pawn. The so-called renewable sourcessolar, wind, wave, tidal, etc. only offer modest contributions in the foreseeable future. Major attention is therefore being paid, at the present time, to expanding power output from coal, of which extensive reserves are available, and which are being increased from time to time as new exploitable coal fields are discovered.

Revival of interest in coal is leading to studies of the efficiency of conversion and utilization processes and to the development of more advanced techniques, particularly to convert raw coal into more convenient forms of energyeasier to transport and cleaner to use.

X INTRODUCTION

The conversion processes include liquefaction, fluidized bed combustion (FBC) and gasification.

Coal derived liquids have been available for many years though there are still important development programmes both in terms of process and materials parameters which aim to improve the efficiency and reliability of the operation. Atmospheric FBC is growing steadily as a commercially available process but the development of pressurised FBC for power generation has been demanding considerable research input. Equally demanding in research problem areas are the gasification processes which include processes using fluidised bed, fixed bed or entrained flow principles. The materials and engineering problems are also dependent on whether the process results in dry ash or liquid ash, i.e. slagging.

In gasification processes, which are the subject of this seminar, the reacting feed gas may be air, steam, hydrogen or a mixture of these gases, and the process may operate at pressures up to 100 atmospheres. The resultant atmosphere in the reaction zone varies widely, depending on the process itself and on the nature of the coal, but in all cases will contain gaseous corrosive species such as carbon monoxide, carbon dioxide, methane, hydrogen and steam, as well as gaseous compounds of sulphur, nitrogen and chlorine. In addition there can be significant quantities of entrained solids which consist of ashes and char. Temperatures in the reactor are about 900-1000 C for the fluidised bed; temperatures up to 1800C being recorded for other gasification processes. The reactions between the various fossil fuels, gases and sulphur retaining minerals result in the production of salts which lead to hot corrosion problems.

Plant reliability, overall efficiency and safety are among the critical factors for designers and operators of gasification systems. All of the exposed materials of construction of the plant whether they be metallic or ceramic will suffer corrosion by the gaseous atmosphere and hot salts, and also erosion by the entrained solids, as well as being subject to the stresses, steady or fluctuating, imposed by the working pressure and the transported reactants. Therefore, in order to reach optimum levels for all the operating requirements, it is necessary to have improved knowledge on materials science and how it affects engineering behaviour.

INTRODUCTION XI

Because of the high sensible heat in the gas, combined with the need to control bed temperature, heat exchangers are almost invariably found somewhere in the system, along with cyclones, ducts, valves etc. In some of the critical areas the metallic materials will be protected by refractories, but in others they will be exposed at high temperature to atmospheres with low oxygen and high sulphur partial pressures, and high carbon activity. Hence complex reactions involving oxidation, sulphidation, carburization and hydrogen diffusion may take place. This corrosion may take place in the flowing gas stream, where stagnant gas collects, or even under a solid or liquid salt deposit.

There is world wide interest in these materials problems which appear in some form in all the gasification systems that are being explored. In this way the experiences described at conferences and symposia organized in the USA almost on an annual basis are very pertinent. In Europe most work has been aimed at material selection under simulated or reproduced process conditions. The Commission of the European Communities is now playing a role in increasing these efforts.

It is therefore opportune that European workers should review the present position, starting from an understanding, as complete as possible, of the underlying basic phenomena and carrying the examination of the topic right through to the behaviour of materials in coal gasification practice.

The objective of the seminar is to explore both the fundamental aspects of high temperature gaseous corrosion and the practical application of such knowledge to coal conversion technology. Although it will not be formally presented, the effect of corrosion on mechanical properties cannot be ignored.

A further objective is to bring together the views and needs of scientists concerned with fundamental corrosion problems, ma-terials selection etc. and engineers involved with the problems of design and efficient operation of gasification plant. The panel of lecturers represents acknowledged experts from this wide spectrum and will ensure that there is this interchange of ideas.

High Temperature Corrosion in Single-Reactant Gaseous Environments

P. KOFSTAD Department of Chemistry, University of Oslo, Norway

1. INTRODUCTION

When the total chemical equation for the reaction between a metal and a single oxidant gas such as oxygen is written:

M(s) + | o 2 = Ma06 (1)

high temperature corrosion of metals may seem to be among the simplest chemical reactions. However, the reaction path and the reaction behaviour may involve a large number of phenomena and processes and depend on a variety of factors, and the reaction mechanisms may, as a result, be complex.1-6

Considering some of the main phenomena and processes encountered (Fig. 1), starting with a truly clean metal surface, the initial step in the interaction is the adsorption of the oxidant on the surface. On continued exposure the initial nuclei of the reaction product are formed. These often grow laterally to produce a continuous film which covers the whole surface. At the same time the oxidant dissolves in the metal substrate to an extent determined by the solubility and diffusivi ty of the oxidant in the metal. The continuous film separates the metal and the oxidant gas, and further reaction is determined by the availability of the oxidant in the ambient atmosphere and the rate of transport of the reactants or electrons through the film.

P. KOFSTAD

02(g)

0 0 ' 0 0 Adsorption

mf^nij. Oxide nucleotion*growth //^/// Oxygen dissolution

S-n Cavities ^ Porosity ^ ^ Microcracks

Macrocracks . Possible molten oxide phases, oxide evap.

FIG. 1. Schematic illustration of main phenomena and partprocesses taking place in the reaction of metals with single oxidant gases, e.g. oxygen.

After Ref. 6. At low temperatures and in dry, single oxidant gases the corrosion

processes are usually slow and often of no practical consequence. Under these conditions the oxidation of many metals follows logarithmictype rate equations for which the reaction comes to a virtual standstill after extended exposure. But at high temperatures the corrosion reactions proceed through thermally activated processes; the reaction rates increase with increasing temperature and materials degradation becomes of great concern.

If the layer of reaction products on the metal surface is dense and continuous, reactions are governed by diffusional transport of the reactants or electrons through the layer of reaction products. The thin layers grow into thicker scales. As the length of the diffusion paths increase with increasing scale thickness, the reaction rate decreases with time. Under many conditions the reaction can be approximately described by a parabolic rate equation.

This solid diffusional transport takes place by various mechanisms, i.e. diffusion through the lattice and along grain

CORROSION IN SINGLE-REACTANT GASEOUS ENVIRONMENTS 3

boundaries and other easy diffusion paths. During the scale growth many additional phenomena and secondary processes may occur. Depending upon the growth mechanism, cavities and closed pores may develop in the scale and the metal. These features will, in turn, affect the transport processes through the scale.

Large stresses may also build up in the scales and the underlying metal. The stresses may cause plastic deformation or lead to cracking of the scales. If repeated cracking takes place, the scales lose their protective ability. In such cases reactions may be governed by diffusion through a thin reaction product layer of approximately constant thickness next to the metal or by phase boundary reactions. Oxidation which is linear with time is commonly observed for this type of reaction behaviour. Depending upon the systems under investigation and the reaction conditions, other processes may take place. These may involve the formation of reaction products which are liquid or continuously evaporate.

With these many aspects to consider the discussion has been limited by placing the major emphasis on formation and growth of continuous scales. These provide the protective properties against high temperature corrosion, and from a practical point of view their growth and properties are the most important aspect of the oxidation of metals and alloys.

2. THERMODYNAMICS OF GAS-METAL REACTIONS

Initially it is appropriate to briefly recapitulate on the basic thermodynamics of such reactions, i.e. to consider the conditions necessary for the reactions to take place and the thermodynamic stability of possible reaction products.

Consider then the formation of an oxide Ma06 from the metal M and oxygen gas: aM + b/202 = MaOb. The metal M can only be oxidized to the oxide (Ma06) if the ambient partial pressure of oxygen is larger than the dissociation pressure of the oxide in equilibrium with the metal. The dissociation pressure is given by

2 ' 2

. KOFSTAD

IO"8

'. > >>"-

_ Pco/PcOj-2/2-02 latm)

TEMPERATURE, C 500 1000 1500

_ l I 1 2000 10'-

.10!

IO4

)0-ioo io-3

,10

FiG. 2. EllinghamRichardson diagram illustrating the stability of various binary metal oxides as a function of temperature.

where AG(MaO,,) is the standard free energy of formation of the oxide at temperature and R is the gas constant. Such free energy data and the dissociation pressures of oxides as a function of temperature are conveniently summarized through the well known EllinghamRichardson diagram. This is illustrated in Fig. 2, with data for a few oxides of importance in high temperature oxidation of metals and alloys. Corresponding diagrams may also be prepared for sulphides, carbides, halides, etc.

Constant partial pressures of oxygen are given by straight lines radiating from the common zero point, AG = 0 and T = 0K, in


the upper left hand corner of the diagram. As example, the line corresponding to p(02) = 10~20 atm. is shown on the diagram.

From this diagram it may be seen, for instance, that the dissociation pressure of Ni O at 1000C is 10 10atm. while that of CoO is 10~12atm., for Cr203 10~22atm., and for A1203 10"35 atm. Partial pressures of oxygen higher than these values are necessary in order that the respective metals are to be oxidized to the oxides. Such data are, for instance, of considerable practical consequence in oxidation of alloys.

These low partial pressures (activities) of oxygen in ambient gases are in practice never realized by means of vacuum systems. Rather, they are established in mixtures of gases in which oxygen is a component. In this respect the most important pairs of gases are in practice represented by C 0 2 + C O and H 2 0 + H2. The partial pressures of oxygen are established through, the relation

C 0 2 = C O + ^ 0 2 (3) and

H 2 0 = H 2 + 0 2 (4) The partial pressures of oxygen are given by the respective

equilibria , = 2 (5)

Peo

Po2 = K2(P^)2 (6) and

where Kt and K2 are the equilibrium constants. For constant ratios f PcoJPco ( /2/2)> t n e Partial pressure of oxygen will be independent of the total pressure of the system.

The partial pressures of oxygen corresponding to various ratios of C0 2 + CO and H 2 0 + H2 may also be read from Fig. 2 by means of the values for pCo/Pco2 anc* PH2/PH2O F r instance, a line for a constant pCo/Pco2 r a t i ' s constructed by connecting point 'C' on the left-hand scale with the desired number on thepCo/Pco2 scale. This is

6 P. KOFSTAD

illustrated for a pCo/Pco2 r a tio of ^ 2 m Fig 2- At the temperature where the 2 = ^~20 a t m - l i n e crosses the pCo/Pco2 = 102 line, this PcJPco2 mixture has a partial pressure of 10 ~20 atm. 02This occurs at 900 C. Furthermore, for the metal/oxide systems which lie above the/?C0/pC02 = 102 line (e.g. Ni/NiO, Co/CoO,Fe (Fel _yO)), the corresponding metals will not be oxidized, but for the systems below the line, the metals (e.g. Cr, , Si, Ti, Al) will, in principle, be oxidized in this mixture.

Corresponding lines for constant pHJpH2o ratios are similarly constructed by connecting the point 'H' on the left-hand scale and the number for the desired ratio on the pHJpH2o scale.

It is important to recognize that dissolution of oxygen in the metals will of course proceed at partial pressures of oxygen below that of the decomposition pressures of the corresponding oxides. The extent of the dissolution will be determined by Henry's law, of which Sievert's law is a particular case which often applies to dissolution of small amounts of diatomic gas molecules in metals.

From a knowledge of these metal/oxide equilibria and of phase diagrams the final reaction products in the oxidation reactions can be predicted. But it is also important to note that metastable compounds or phases may be important intermediate reaction products in gas-metal reactions.

3. GROWTH OF SCALES ON UNALLOYED METALS BY LATTICE DIFFUSION

3.1. The Wagner Oxidation Theory As already mentioned, the solid state diffusional transport

through continuous scales will involve lattice diffusion and transport along grain boundaries and other easy diffusion paths.

Consider the ideal model where the scale is completely dense and where the transport takes place by lattice (volume) diffusion only. Any possible formation of cavities and pores is neglected. Such an ideal case is covered by the Wagner oxidation theory.1-8 In the following the theory is considered in terms of metal-oxygen reactions, but it applies equally well to other gas-metal reactions.

CORROSION IN SINGLE-REACTANT GASEOUS ENVIRONMENTS

Metal i ^ ^

Oxide ^

Metal ionsis; /. ^Oxygen ione s

; Electrons 1

Oxidant gas e.g.02.

m //Interst. metal iohs

S ^1

b. / / ^ f l e t r o n s j S

'/ Metal vacancies m sa FIG. 3. Transport processes through dense, singlephase scales growing by lattice diffusion. Transport of reacting atoms or ions or of electrons is ratedetermining. The overall reaction follows a parabolic rate,

dx/dt = k'jjx. After Ref. 6. The basic assumption of the theory is that the lattice diffusion of

the reacting atoms or ions or transport of electrons through the scale is the ratedetermining process in the oxidation reaction. These transport processes are illustrated in Fig. 3(a) for a dense, singlephase oxide scale. It is further assumed that thermodynamic equilibria are established at the metal/oxide and oxide/oxidant interfaces. The 'driving energy' of the reaction is the free energy change of the reaction between the metal and the oxidant to form the reaction product. The migrating species may alternatively be considered to constitute lattice and electronic defects, i.e. vacancies and interstitial ions and electron holes and electrons, respectively (Fig. 3(b)).

For such a reaction mechanism the growth of the scale (e.g. the scale thickness) is parabolic with time. The differential and integral forms of the parabolic rate equation are given by

dt px X2=2k't + r kpt + cp

(7)

(8) where is a measure of the oxide thickness which may be variously expressed as the thickness of the scale or as the oxidant uptake (e.g.

8 P. KOFSTAD

weight gain per unit surface area of the metal, etc), t denotes the time, cp the integration constant and k'p (or k'p = \kp) is the parabolic rate constant.

Wagner7,8 derived the expression for the growth of the scale (and thereby for k'p) by starting with the expression for the particle current density or the flux of the migrating reacting ions through the scale. Further, by use of the GibbsDuhem relation and the fact that equivalent amounts of oppositely charged particles are transported through the scale, k'p is expressed in terms of the electrical conductivity, , and the transport number of ions () and electrons (te) in the oxide:

kT fP02 *>-Jb?\> ^ d l n ^ (9)

where/?o2 and/?2 are respectively the partial pressures of oxygen at the oxide/gas (outer) and metal/oxide (inner) interfaces of the scale, k is the Boltzmann constant, the absolute temperature, and e the electronic charge. The letter b represents the number of gram atoms of oxygen per mole of 06.

In order to evaluate the parabolic rate constant and perform the proper integration, it is necessary to know , , and te as a function of oxygen pressure from 2 to 2. It can also be noted that if the scale is an electronic conductor, then the scale growth is governed by the diffusion of reacting ions through the scale, but if the scale is an ionic conductor, then scale growth is determined by the electron transport through the scale.

When the scale is an electronic conductor (te = 1) the parabolic rate constant can be rewritten in terms of the selfdiffusion coefficients of the metal and oxygen ions in MflO, DM and D0, by the use of the NernstEinstein relation.8 This yields

* ; = C o ' ( z , o + a > ) d l n ^ (10) where c0 represents the concentration of oxygen in the oxide, zc and za are the valencies of the cations and anions, respectively.

Equation 10 can be considerably simplified for certain limiting


conditions. For instance, consider the case where the scale growth takes place by metal transport diffusing via vacancies having an effective charge, a. If

i) Po2 ^ Po2> i-e- the ambient oxygen pressure is very much larger than the oxygen pressure at the inner scale boundary,

ii) DM >Z)0, i.e. oxygen diffusion is insignificant compared to metal diffusion,

k'p is given by

k'p = ikp = (a + l)Du (11) where >, is the metal self-diffusion coefficient in the oxide at the oxide/gas interface.

If the metal defects are interstitial metal ions with an effective charge a, k'p is given by

^ = ^ p = (a + l ) ^ (12) where D'M is the diffusion coefficient of the metal ions in the oxide at the metal/oxide interface.

It is important to note that the parabolic rate constant is determined by the self-diffusion coefficient of the metal in the oxide at the gas/oxide interface in the case of metal vacancy diffusion. For interstitial metal diffusion the parabolic rate constant is determined by the self-diffusion coefficient in the oxide at the metal/oxide interface. This difference is not always sufficiently emphasized or understood in considerations of scale growth by various diffusion mechanisms.

Similar expressions for k'p apply to limiting cases where oxygen diffusion by vacancies or interstitials is rate-determining for the scale growth.

It may also be noted that parabolic scale growth by metal vacancy or interstitial oxygen diffusion is dependent upon the ambient partial pressure of oxygen, k'p ccp^{a +1}, while for metal interstitial or oxygen vacancy diffusion the scale growth is independent of the ambient oxygen pressure (as the important parameter is the partial pressure of oxygen at the metal/scale interface, which is not affected by the ambient gas pressure).

10 P. KOFSTAD

3.2. Comparison of Directly Measured and Calculated Values of kp It is of interest to test the applicability of the ideal model by

comparing values of the parabolic rate constant measured from oxidation studies with values calculated from independent measure-ments of self-diffusion or electrical conductivity. This may, for instance, be done for oxidation of cobalt to Co O in oxygen atmospheres. A number of studies have been made of both oxidation of high-purity cobalt and of Co-tracer diffusion in CoO.8-21

Cobalt diffusion is much faster than the oxygen diffusion in CoO, and accordingly DCo>D0.9 Studies of nonstoichiometry and electrical conductivity of CoO suggest that the predominating defects in CoO are singly charged cobalt vacancies.9-21 Thus in calculating values of kp, is approximately equal to 1. When tracer diffusion coefficients, D^, are used in the calculations, a correction must be made for the correlation effects.9 For Co-vacancy diffusion in CoO (NaCl-structure) the correlation factor is = 0-78 and thus >co = cV0-78.

If cobalt diffusion is the rate-determining process in oxidation of cobalt, the parabolic rate constant and the self-diffusion coefficient should exhibit the same dependence on the ambient partial pressure of oxygen. Such a comparison is shown in Fig. 4. The results confirm the main aspects of the model.

Following Equation 11, the ratio of k'p (in cm2/s) to >0 (in cm2/s) should equal J/>*o = ( + 1)/078. A comparison of k'p and D%0 is shown in Fig. 5. Both values exhibit the same temperature dependence ( ~ 160 kJ/mole), and the ratio is close to 2-5. This yields a value of of ~ 1 , in accord with the nonstoichiometry and electrical conductivity studies. This again confirms the main assumptions of the oxidation model.

This comparison is based on a simplified model assuming that the same type of defect (singly charged cobalt vacancies) prevails through the entire scale. In more detailed considerations of the transport processes and the reaction mechanism one could take into account possible changes in the value of through the oxide scale22 and whether other types of defects, e.g. interstitial cobalt ions and/or complex ions (vacancy-interstitial defects) contribute to the overall

CORROSION IN SINGLE-REACTANT GASEOUS ENVIRONMENTS 11 ' in

CM E LU o o g Z3

10" Oxidot ion of Co:

12 P. KOFSTAD TEMPERATURE . 'C

o s S S c to if - 2

* M I0"'[r

=> < li- K o: o iu m o < < fr

F L-^-' ""

V a O

.. \

;

- 1 - ! " ; Padassi et al . Krger et al . Snide et al . Bridges et al (interpol.

val.)

xidation of Co in air.

/ \ .. ; ; Co-tracer diffusion in \^_ * - CoO. air. V* :

Carter & Richardson \ g 7 o Chen et al .

I \ ]

1 , 1 ' 5.0 6.0 7.0 8.0

10' .

FIG. 5. Comparison of the parabolic rate constant for oxidation of cobalt to CoO, k'p, with the cobalt tracer diffusion coefficient in CoO, D0, in air. The activation energy is 160 kJ/mole. The ratio of k'p to D0 is ~ 25, which yields =: 1. Cobalt tracer diffusion studies of Carter and Richardson10 and Chen et al.21; oxidation studies of Bridges et al.,11 Snide et al.,12 Krger et

al.,13 and Paidassi et al.1* After Ref. 6. of grain boundary diffusion, etc. and makes it comparatively easy to study defectdependent properties of CoO.

In order to make corresponding comparisons for scales growing by oxygen vacancy diffusion, it is necessary to have available data for the oxygen selfdiffusion coefficient in the oxide at the metal/ oxide interface. Such data are, as yet, not available, as oxygen diffusion studies have commonly been carried out at nearatmospheric oxygen pressures. Comparisons of calculated and measured values of kp can thus not be made at this stage.

3.3. Variations in the Defect Structure and Transport Properties across the Scale

The preceding comparisons make use of the simplifying


approximation that the same defect structure prevails across the entire scale. This may be an oversimplifiedor even incorrect model for many systems, particularly those where very stable reaction products existing over large oxygen pressure ranges are formed. Examples of very stable oxides formed as scales in oxidation of metals are A1203, Cr203 , Zr0 2 and Ta205 .8 For Zr0 2 at 1000 C, for instance, the partial pressure of oxygen at the Zr 02/Zr-O solid solution interface is of the order 10~40 atm. Thus when the corresponding metals are oxidized in atmospheric oxygen, the difference in partial pressure of oxygen between the outer and inner interfaces of the scale may cover up to 40 orders of magnitude. It would not be surprising that the defect structure situation in such cases may vary over the scale through changes in the intrinsic disorder or through effects of impurities (extrinsic behaviour). Consequently, the predominating point defects may differ in different layers of the scale, the scale may be an ionic conductor in one layer scale and an electronic conductor in another, etc.

If the Wagner oxidation theory is applied to such situations, one may subdivide the total scale into layers having different defect structure situations and transport properties, and estimate the parabolic constant for each layer. At this time there is generally insufficient knowledge to map out the defect structure of very stable oxides over their entire existence range, and detailed analyses can not be made of the changes of the defect structure through such scales.

3.4. Electron Transport as a Rate-Limiting Factor In interpretations of reaction mechanisms of high-temperature

corrosion in the literature it is generally assumed that the scales of reaction products are electronic conductors and that the transport of the atoms or ions of the reactants is the rate-determining process. However, the possibility that electron transport can be a rate-determining factor should also be considered. This particularly applies to certain oxide scales.

In order to illustrate this aspect consider the electrical conductivity of Ta205 . Ta205 is an oxygen-deficient oxide, and the growth of continuous Ta205 scales takes place by oxygen diffusion.26-30 When the electrical conductivity of Ta205 is

14 P. KOFSTAD

10-3

.c o

Q o o

CORROSION IN SINGLE-REACTANT GASEOUS ENVIRONMENTS

To205/02

15 /25

-0 sol. sol'n

TQ2 05-scale Intrinsic

O-diffusion rate-determin

-X

Exlrinsi_ Electron transp rate-determ.

h*

! -X i -

o2(g)

p/n trans.' I

o 2> 2 2s> ^ . 0.8 o S, 0.6 $ S, a*

s s 0.2 "* 10

1 1 1000'C P /n ,- rans

k"2 / '

.1/2 ' : 1/2 /

J L /

I io- 10"

OXYGEN PRESSURE, atm. FIG. 7. Subdivision of Ta205 scales in an outer extrinsic and an inner intrinsic layer. Growth of an outer part of the extrinsic layer is assumed to be governed by electron hole transport with fc as the parabolic rate constant. Electron transport is ratedetermining for growth of an inner part of the extrinsic layer (fcn) and oxygen diffusion for the intrinsic layer ().

After Ref. 6. Similar situations are to be expected for other oxidation reactions.

The defect structure of Zr0 2 is qualitatively similar to that of Ta205 , although the extrinsic range covers a larger oxygen pressure range.9 A large part of Zr0 2 scales may accordingly have extrinsic properties. The growth of this extrinsic layer is therefore expected to be governed by the electronic transport. These considerations are in accord with results of Bradhurst et al.30 and Opara et al.31 who made e.m.f. measurements across growing Zr02 scales. From their

16 P. KOFSTAD EO

140

120

100

80

60

40

20

Oxidation of Zr f 750 "C ' /

Eriksen & Hufte / o / / oxidation with

J electronic short 8 circuit across

/ oxide scale /

i / / / JfT

' sf s normal oxidation

fs*

_

10 15 T I M E , h

20 25

FIG. 8. Oxidation of zirconium at 750 C under 'normal' conditions and when the surface of the Zr02 scale is electronically shortcircuited to the underlying zirconium metal with a platinum wire. Results after Eriksen and

Hauffe.32 Previously published in Ref. 25.

results it may be concluded that electron transport is partially rate controlling.

In this respect it is also of interest to quote studies by Eriksen and Hauffe.32 They electronically shortcircuited the surface of Zr0 2 scales to the zirconium metal by means of a platinum wire. The oxidation under these shortcircuit conditions was faster than under 'normal' conditions (Fig. 8). Furthermore, an interesting effect is that the 'normal' oxidation follows an overall cubic rate, while oxidation with electronic shortcircuiting was parabolic. This may suggest that the 'normal' cubic oxidation may at least partially be associated with a ratedetermining electron transport. The deviation from parabolic behaviour may further reflect changes in the electronic transport properties during the course of oxidation. Despite these results it is commonly assumed in the extensive literature on zirconium oxidation that oxygen transport is the rate determining process in growth of Zr02 scales. Clearly it would be of


interest to evaluate the possible role of rate-determining electron transport through part of Zr02 scales.

Similar studies of electrical properties of growing alumina scales on Pt-22wt %A1 also indicate that the growth of an outer layer of the scale is governed by electron transport while an inner layer is governed by ionic transport.33,34

4. EFFECTS OF NON-LATTICE DIFFUSION

Although the growth of scales by lattice diffusion and systems where this type of transport predominates are important, the transport processes can be more complex, and non-lattice diffusion may be significant or play a dominant role in the growth of continuous scales. The oxidation behaviour of nickel can serve to illustrate this.

In a similar way to the oxidation of cobalt, it is of interest to compare measured values of the parabolic rate constant for high-purity nickel with values calculated from independently measured values of the nickel self-diffusion coefficient in NiO.4 8 - 5 0 Such a comparison is shown in Fig. 9. At very high temperatures, above about 1200C, the activation energies for oxidation of nickel and nickel diffusion in NiO are closely similar. Under these conditions the ratio of k'p to D^ is ~25. This suggestsas for the cobalt-oxygen systemthat the oxidation of nickel at very high temperatures is governed by lattice diffusion and that the nickel vacancies in NiO are singly charged.6'51'52

However, at lower temperatures the observed activation energy for nickel oxidation becomes smaller. While the activation energy for lattice-diffusion controlled growth is 230-250 kJ/mole, the activation energy at lower temperature ( < ~ 1000C) is of the order of 130-145 kJ/mole.6

A number of other interesting effects are also observed for oxidation of nickel at and below 1200-1000C.53-61 Cold-worked nickel oxidizes faster than annealed nickel (Fig. 10). Furthermore, oxidation of cold-worked nickel does not follow an ideal parabolic rate, the rate decreases faster with time than for the ideal case.44

18 P. KOFSTAD TEMPERATURE C

. ICT

o

< cc m < ce. < a.

IO"'

ce LU CJ < ce

10-

10"

Oxid, of Ni. l a tm. 0 2 -D Fueki Wagner Caplan et al. Sorteli * L

Ni - tracer diff. Coeff ic ient :

Volpe Reddy o Atkinson Taylor

Lindner kerstrm

6.0 8.0 7.0 10 4 / T .K

FIG. 9. Comparison of parabolic rate constant for oxidation of highpurity nickel (k'p in cm2/s) with the tracer diffusion coefficient of nickel in NiO (Dft in cm2/s) at latm. 0 2 . At high temperatures (>1200C) k'JDfacz 25. Nickel tracer diffusion studies of Volpe and Reddy4950 and Lindner and kerstrm48; oxidation studies of Sartell and Li,35 Fueki and

Wagner,40 and Caplan et al.AAt After Ref. 6. Oxidation rates vary significantly with metal orientation,54,59 and different types of surface preparation significantly influence the reaction behaviour.5458 From all these studies it has been concluded that the relatively rapid reaction rates can be correlated with the microstructure and grain size in the oxide.6 Oxidation rates are concluded to be rapid when oxide scales contain small grains and


OXIDATION OF NICKEL

5 10 15 TIME.h

FIG. 10. Oxidation of cold-worked and annealed, zone-refined nickel at 900, 1100 and 1270 C. Results of Caplan et al." Previously published in

Ref. 6.

boundaries available for short-circuit diffusion. The general conclusion is that non-lattice diffusion becomes increasingly important with decreasing temperature.56

These results are supported by nickel and oxygen isotope tracer studies in growing NiO scales. Atkinson et al.61 deposited a thin layer of 63Ni on the surface of nickel specimens prior to oxidation, and after oxidation studied the distribution of the tracer in the scale normal to the metal. Similar distribution studies were done with 180-tracers, but in these cases the first half of oxidation was carried out in natural oxygen (1602) and the second half in oxygen enriched in 1 8 0 2 .

The distribution of the tracers in the scale reflects the transport mechanisms in the scale. Scales grown on (100) faces at 1000 C were compact and adhered well to the substrate. The tracer distributions were consistent with that expected for lattice diffusion of nickel vacancies. This is illustrated in Fig. II.61 For scales grown on (100) faces at 500 and 800 C the 63Ni distribution was completely different from that at 1000C. The 63Ni distribution showed a peak close to the metal/oxide interface as illustrated in Fig. 12. This strongly suggests that under these conditions nickel is pre-dominantly transported along easy diffusion paths.

20 P. KOFSTAD

< i "0 0.5 1.0 NORMALIZED SCALE THICKNESS, x/X

FIG. 11. 63Ni distribution in NiO scales grown on (100) Ni crystals at 1000C and 66 102atm. 0 2 for 104h. The scale had a thickness of

38 . After Atkinson et al.61

The 1 8 0 distribution curves give interesting information on other transport processes in the growth of NiO scales. On polycrystalline nickel specimens at 610 and 1000C there was no inward oxygen penetration during the initial scale formation. But for thicker scales (>l /an at 600C and >9 at 1000C) 1 8 0 was found to penetrate to the region of the scale next to the metal/oxide

E Metal/Scale interlace

Si 1.0

gO.5 o < cc o

_ Ni metal

o, o

o _ o

o

o

. "1 .

1 o . "

% N0 Scale .

N i ^ ' - o o " o V

B00C,1atm.O2 500C, s

1.0 0.5 0 0.5

NORMALIZED DISTANCE, / FIG. 12. Distribution of 63Ni tracers in NiO scales on (100) crystals after oxidation at 500 and 800 C in 50 torr 0 2 . The total oxide thickness after oxidation at 500 C was 0 3 and at 800 C 10 . After Atkinson et al.61

Previously published in Ref. 6.


S S 1 < ce

z0.5h o IM I cc o

Metal/scale Scale/gas interface interface

' L e r

NiO scole f "

Ni-concentr. "*

S ^ ^

*^wOdlstrb.

-OXIDE THICKNESS-

FIG. 13. 1 8 0 distribution in NiO scale formed during oxidation of polycrystalline nickel at 1000 C and 50 torr 0 2 . The specimen was oxidized for 12 h in 1 6 0 2 and then for 12 h in 1802enriched atmosphere. After Atkinson et al.61 Previously published in Ref. 6.

interface (Fig. 13). This penetration seems to be correlated with the development of a duplex scale structure with an inner, porous layer and an outer, compact layer. Such duplex layers are well known in numerous studies of high temperature nickel oxidation and will be further discussed.

As 1 8 0 penetration only takes place after more extended oxidation, Atkinson et al. conclude that the most probable mechanism involves the development of microcracks as a result of growth stresses in the scale. It is interesting to note that similar penetrations of the oxidant to the inner part of the scale have been observed in oxidation of Co,24 Cr,62 Fe35%Cr,63 reaction of FeCr alloys with C02 ,6 4 6 5 and the reaction of nickel with S02 ,6 6 etc.

All these results clearly demonstrate that the oxidation rate is larger the finer the grain size and that nonlattice transport of both nickel and oxygen contributes to or may dominate the transport processes in the overall oxidation. At reduced temperatures the observed oxidation rates may be several orders of magnitude larger than the value estimated for oxidation by lattice diffusion.

This situation has prompted the development of models for growth of scales by simultaneous lattice and grain boundary

22 P. KOFSTAD

diffusion.53-5567'68 The relative contributions to growth by lattice and grain boundary diffusion, respectively, have been estimated. The estimates suggest that grain boundary diffusion is the predominant transport process at reduced temperatures.6

Results of this type are not limited to the oxidation of nickel. Through extensive studies it has been concluded that the growth of Cr203 scales on pure chromium always takes place through non-lattice diffusion.69

5. FORMATION OF VOIDS IN SCALES AND METAL SUBSTRATES

Oxide scales that grow by outward metal transport always adhere well to the metal during the initial stages of the reaction. But after more extended reaction it is a general phenomenon that voids and cavities develop in the scale, at the metal/scale interface, and within the metal itself promoting the formation of double-layered scales consisting of an outer, compact layer and an inner, porous layer. The outer layer usually consists of relatively large, columnar grains, while the inner layer is much more fine-grained.35 _ 4 7 , 6 0 '6 1 , 7 0 - 8 0

The principal reason for the void formation is the outward migration of metal through the scale. Thus if we consider that the oxide forms a completely rigid envelope around a metal specimen, the total volume of the voids should be equal to the volume of the metal that is converted to oxide. But only a fraction of this volume develops as voids, cavities or porosity. This is due to a tendency for the scale and the metal to maintain contact and adhesion. This may occur through two major processes:

i) plastic deformation of the oxide scale. As the metal surface recedes and changes shape and morphology, the oxide deforms to maintain contact between the metal and the scale;

ii) vacancy injection and void formation in the metal. When a metal atom enters the oxide at the metal/scale interface, a metal vacancy is left behind in the metal lattice. As oxidation proceeds, the metal lattice vacancies will to some extent form


a supersaturated solid solution of vacancies in the metal. Furthermore, the vacancies diffuse into the metal and precipitate out as voids at-favourable sites, e.g. at grain boundaries.7274-76

It may be noted that excess vacancies in the metal may also condense out as planar defects (vacancy discs). These may collapse and produce dislocation loops resulting in shrinkage of the metal.

The tendency for void formation is a function of the shape of specimens and their surface-to-volume ratio (radius of curva-ture).72,74 For a metal with an infinite plane surface oxidizing uniformly along the surface, the oxide can in principle continuously collapse on the retreating metal surface without any deformation of the scale. If non-uniform oxidation takes place, strains are produced in the scale. For specimens and components with finite dimensions constraints are imposed on the system, and this produces voids and cavities.

The effect of variations in the surface-to-volume ratio and in the surface geometry has, for instance, been demonstrated by Hales and Hill.72 They oxidized high-purity nickel rods with different diameters (5, 2-5 and 1-25 mm diameter) and studied the void formation in the various samples after oxidation in 1 atm. 0 2 at 1100C for times ranging from 240 to 720 h. Prior to oxidation the material was vacuum annealed for 12 h at 900 C to develop a large stable grain size.

Voids developed at the grain boundaries in the metal as a result of oxidation, first in small numbers in a narrow band close to the metal/ oxide interface, but on extended oxidation also towards the centre of the specimen. Oxide was also formed along the voids in the interior of the metal. A main finding was that the density of voids increased with decreasing diameter of the rods and with increasing oxidation time. The effect of this surface-to-volume ratio is due to the fact that the oxide envelopes (with equal thicknesses) have a reduced ability to deform the smaller the diameter of the rod.

Similar effects were found in the oxidation of flat slabs made from 1 mm thick foil.72 Voids were found in the metal at the edges and corners of the specimens rather than at the centre of the major surfaces.

24 P. KOFSTAD

5.1. Effects of Void Formation on Oxidation Behaviour and Transport Mechanisms

The overall void formation affects the oxidation behaviour and mechanism. Considering injection of metal vacancies in the metal substrate, if these are annihilated at vacancy sinks such as dislocations, interfaces and grain boundaries, the metal surface recedes. For contact to be maintained at the metal/oxide interface, the scale will be strained. As the oxide grows in thickness, the oxide scale will not be able to deform sufficiently and maintain contact. Eventuallyand at a critical oxide thickness depending upon the geometrythere will be a loss of adherence at the metal/oxide interface.

If the injected metal vacancies partially condense as three-dimensional voids at grain boundaries in the metal, the metal surface will not recede to the same extent and contact between the metal and the scale can be more easily maintained. However, an accumulation of voids along grain boundaries in the metal may severely affect the mechanical integrity and properties of the metal substrate.

When voids are formed at the metal/oxide interface or there is a loss of scale adherence, the path for the outward solid state diffusion of the metal is broken. This could lead to the conclusion that the oxidation should slow down or be interrupted. However, numerous studies of oxidation of nickel have, for instance, shown that oxidation continues and there is no significant decrease in the activation energy of the oxidation. The scale growth continues in the form of a double-layered scale consisting of an apparently dense, outer layer with large, columnar Ni O crystals, and an inner porous, more fine-grained, oxide layer. This means that transport of one or both of the reactants continues. There appears to be two main reasons for this continued oxidation:5,7074

i) dissociative transport across voids; ii) the development of microchannels in the outer layer which

permits penetration of the oxidant to the inner layer. i) Dissociative transport across voids. When a void has been

formed at the metal/oxide interface, metal ions will continue to


Oxygen

FIG. 14. Schematic illustration of dissociative transport across pores and voids in continuous scales.

move outwards from the void surface facing the outer oxide surface. This will decrease the chemical potential of the metal at the outer void surface and correspondingly increase the partial pressure of the non-metal (oxygen) in the void. Thus as the metal ions move outwards, the non-metal is liberated as gas in the void. The non-metal is transported in the gas phase across the void, and new oxide is formed on the metal (Fig. 14).

When the oxidizing metal contains carbon, CO and C0 2 are formed, and this gas mixture fills the voids. This has been demonstrated by Graham and Caplan.73 If an approximate equilibrium is established, the C02/CO ratio will be determined by the partial pressure of oxygen in the void. It has been concluded that the presence of the C02/CO gas in the voids will act as a carrier gas and assist in the gaseous transfer of oxygen across the voids. Gas mixtures of H20/H2 may similarly be formed and act as a carrier gas in the oxidation of hydrogen-containing metal. If this dissociative mechanism is the important one in transferring oxygen across the voids, the voids will move outward in the scale as oxidation proceeds. Concurrent with this movement, the partial pressure in the voids also gradually increases.79

ii) Development of microchannels. As previously discussed 1 8 0-tracer studies have demonstrated that oxygen is also transported from the ambient atmosphere to the inner porous layer after more extended reaction. It appears that the inward transport of oxygen is

26 P. KOFSTAD 02(g)

O x i d e M j n . e . p ^ 'y U\ " 3 ^

I Me"* e- R o P i d metal diffusion along grain boundary

02(g) Constant

02 = potential

Opening of ^ grain boundary

Microcrack, inward O2 transport "mM,

02(g) Stabilization of micro-crack, inward O2 transport, outward metal transp.

FIG. 15. Model for development of microchannels above cavities at the metal/oxide interface or in the oxide scale. Solid state diffusion in the oxide takes place by outward metal diffusion in the lattice and along grain boundaries. When grain boundary diffusion is much faster than lattice diffusion, the grain boundary opens into a microchannel. This in turn permits inward transport of oxygen.5,70'74 Previously published in Ref. 6.

associated with, or is a result of, the duplex scale formation. It is also concluded that this oxygen transport takes place along microcracks that develop in the outer layer.

The model for the development of such microchannels has been advanced by Mrowec and coworkers570 and Gibbs and Hales.74 Consider a newly formed void at the metal/oxide interface with the scale above the void containing an oxide grain boundary normal to the metal surface as illustrated schematically in Fig. 15.74 For the sake of simplicity the grain boundary is considered to extend from the void to the oxide/gas interface. If the outward diffusion of metal is faster along the grain boundary than through the lattice, the grain boundary area begins to 'open' up next to the void. This process continues and eventually opens the grain boundary as a channel which penetrates the scale. The channel then permits the oxidant to


migrate as molecules down the channel to the metal and contributes to the oxidation of the void surface and formation of the inner porous scale. The overall oxidation will, under these conditions, involve the inward transport of oxidant molecules along the microchannels and the outward transport of metal in scale regions with continuous solid state diffusion paths.

This mechanism for the opening of microchannels in the scale requires that diffusion along grain boundaries is much faster than lattice diffusion. As grain boundary diffusion has a lower activation energy than lattice diffusion, it is reasonable to conclude that at sufficiently high temperatures, where lattice diffusion becomes of the same order of magnitude as grain boundary diffusion, the mechanism of channel opening no longer operates. Conversely, formation of microchannels is expected to be of increasing relative importance at reduced temperatures. This qualitatively explains the 180-tracer studies in growing Ni O scales which show that inward oxygen penetration becomes increasingly important at lower temperatures.61

The opening of microchannels in oxide scales not only provides an explanation for inward transport of oxygen molecules from the ambient atmosphere, but the mechanism may also explain the occurrence of sulphidation at the metal/scale interface beneath an outer oxide layer in reactions of nickel with S02 ,66 and similarly carburization in C0 2 atmospheres,82 etc.

6. STRESSES AND STRAINS IN GROWING OXIDE SCALES

In the preceding discussion brief references have already been made to the development of stresses and strains in growing scales. In general, growth stresses and their resultant effects can be significant or of decisive importance in the corrosion behaviour of metals and alloys. In this context a few considerations will be made regarding the corrosion of pure metals in single oxidant gases forming continuous scales.

28 P. KOFSTAD

Consider initially the situation where lattice and grain boundary diffusion of the oxidant (e.g. oxygen) is the predominant mode of transport through the continuous scale. In such cases fresh reaction products are formed at the metal/scale interface. If the molar volume of the reaction products per metal atom is larger than that of the metal atom, the volume increase associated with the formation of reaction products at the metal/scale interface will lead to compressive stresses in the scale. After reaching a critical thickness the scale will crack or fracture, and the protective properties are lost. Examples of this type of behaviour are the high-temperature oxidation of niobium and tantalum.4 This situation corresponds to the classical Pilling-Bedworth model for oxidation of metal taking place by inward migration of oxygen. Pilling and Bedworth related the sign and magnitude of the stress in the scale to the ratio of the volume of the oxide per metal ions to the volume of the metal per metal ion. This is now commonly termed the Pilling-Bedworth Ratio, i.e. PBR.

With basis in the Pilling-Bedworth model it was originally suggested that there is no reason for the development of growth stresses if the scale grows by outward metal ion lattice diffusion. Fresh oxide is then formed in a supposedly stress-free manner at the scale/gas interface. But as discussed earlier, stresses arise in such scales when adherence is maintained between the metal and the scale and the metal surface simultaneously recedes during oxidation. In such cases the magnitude of the stress and the stress pattern are also dependent on the geometry of the metal samples.

Large stresses may also arise if grain boundary diffusion, and particularly counter-current boundary diffusion of both metal and the oxidant, is important. In such cases a significant amount of oxide is formed at the grain boundaries. This produces compressions in the scale. This type of mechanism has been proposed to be important in the oxidation of nickel60,80 and chromium.69,83,84

In the oxidation of chromium this type of growth mechanism can lead to extensive bulging, deformation and cracking of the scales. An example of this is illustrated in Fig. 16, which shows a scanning electron micrograph of the Cr 20 3 scale formed on chromium oxidized in latm. 0 2 at 1200 C. This type of bulging is not an


FIG. 16. Scanning electron micrograph of chromium specimens oxidized at latm. 0 2 and 1200C.

uncommon phenomenon for C r 2 0 3 scales grown on chromia-forming alloys.

6.1. Relief of Growth Stresses Growth stresses may be relieved through various mechanisms.

These may be divided into four main categories:75-78

i) plastic deformation of the scale; ii) deformation of substrate metal; iii) detachment at the metal/scale interface; iv) elastic failure (cracking) of the scale.

Various features of these categories have been mentioned earlier. In this context some aspects of plastic deformation of oxide scales will be briefly mentioned.

6.1.1. Plastic Deformation of Scales In principle various deformation processes may take place:

dislocation glide, grain boundary sliding, mechanical twinning, high-temperature creep, etc.7677 It appears that high-temperature creep through the Herring-Nabarro mechanism and dislocation climb processes are the most important mechanisms.76-77

30 P. KOFSTAD

Creep in ceramics is normally controlled by diffusion of the slower moving species, i.e. metal or oxygen ions. Thus while oxidation is governed by the faster moving reactant (majority defects), plastic deformation of the scales is governed by the slower moving reactant. In order to interpret the overall oxidation behaviour and properties of scales it is necessary to have a detailed understanding of the defect structures of the oxides both as regards the metal and oxygen defects. Generally there is an incomplete knowledge of the defect structure related to the slower moving species (minority defects).

In metal-deficient oxides such as Co O and Fe1_JC0, high temperature creep is a function of non-stoichiometry, and creep rates increase with increasing non-stoichiometry.9'76,7784-87 On this basis it has been suggested that creep in these oxides is related to the metal vacancy concentration. But in CoO, for instance, the creep varies with p\*, which is not the expected oxygen pressure dependence if transport of cobalt vacancies is a rate-determining factor in the creep process. Rather the results suggest that the slower moving species are oxygen interstitials either as neutral interstitials or as a complex interstitial defect. In both cases the rate-determining oxygen self-diffusion may vary as pj*.9

When applied to growth of metal-deficient oxide scales this means that high-temperature creep is greater at the scale/gas interface than at the metal/scale interface. Thus stresses are more easily alleviated at the scale/gas interface. This leads to the speculation as to whether these differences in creep rates across the scale also contribute to development of duplex layers in the growth of Ni O and CoO scales.

For oxygen-deficient oxides where interstitial metal ions and oxygen vacancies are the important point defects in the metal and oxygen ion lattices, respectively, creep is expected to increase with decreasing partial pressure of oxygen. This is the proposed situation for Cr203 at reduced oxygen pressures.69'83'84 Thus the ability of Cr203 scales to deform increases dramatically with decreasing partial pressure of oxygen. This is illustrated in Fig. 17 which shows low magnification scanning electron micrographs (SEM) of chromium specimens oxidized at 1075 C at 1, 1-3 10 ~3, and 7 10 ~7 atm. 0 2 , respectively. At 1 atm. 0 2 the Cr203 scales are formed plane-parallel to the metal surface, but at decreasing

CORROSION IN SINGLEREACTANT GASEOUS ENVIRONMENTS 31

A'!'.' i: vi

FIG. 17. Scanning electron micrographs of chromium specimens oxidized at 1075C: (a) 1 atm. O, for 25 h; (b) 13 "3 atm. 0 2 for 43 h; (c) 7 IO"7 atm. O, for 72h.8384

pressure they begin to deform extensively and to bulge away from the surface. Figure 18 shows the appearance of a chromium specimen oxidized at 1075C and 7 10"7atm. 0 2 . The scale has ballooned away from the surface of the originally rectangular specimen and has also sagged to yield a pearshaped envelope of scale.

These examples clearly illustrate the importance of stresses and strains in oxide scales and their ability to deform under different

32 P. KOFSTAD

FIG. 18. Appearance of chromium specimen oxidized for 26 h at 1075 C and 7 10~7 atm. 0 2 . The scale is completely detached from the metal, and the oxide has deformed and sagged into a pear-shaped form.83 '84

conditions. In order to improve our understanding of the oxidation of metals it is highly desirable to increase our knowledge of these aspects.

7. HIGH TEMPERATURE CORROSION OF ALLOYS Oxidation of alloys involves the same general phenomena and processes described and discussed for unalloyed metals. But as alloys in general contain two or more oxidizable constituents, a number of additional factors and parameters must be taken into


account if a comprehensive description of the reaction behaviour of alloys is to be given. As a result, reaction mechanisms are more complex for alloys than for pure metals.1-5

The components of alloys will have different affinities for the oxidants, and reacting atoms do not diffuse at the same rate either in the scale or the alloy phases. As a result, scales on alloys will not contain the same relative amounts of the alloy constituents as does the alloy phase. Furthermore, the composition and structure of scales on alloys will often change as the reaction proceeds, and the reaction kinetics, in turn, often markedly deviate from ideal and simple rate equations. If the oxidant dissolves in the alloy phase, the least noble alloy component also reacts within the alloy to form particles of oxide, sulphide, carbide, etc. (internal oxidation). As a result of this, it is often difficult to give an overall quantitative description of high temperature corrosion of alloys.

A comprehensive treatment of the oxidation of alloys comprises a large number of phenomena which cannot be adequately covered here. Instead of discussing a more general case, important features may be illustrated by specific alloy systems. As such it is of interest to look at the behaviour of alloys of nickel and cobalt with additions of chromium and aluminium. In addition to illustrating important fundamental aspects, these systems are of great practical interest since they constitute the basis for a large number of important commercial super-alloys.

7.1. Oxidation of Co-Cr and Ni-Cr Alloys Consider first the oxide phases which can exist in the Co-Cr-0

and Ni-Cr-0 systems. These can be summarized as follows: Co-Cr-O: CoO, Co304 (stable < 970C in latm. 02), Cr203 ,

CoCr204 Ni-Cr-O: NiO, Cr203 , NiCr204 Chromium oxide is more stable than both NiO and CoO (Fig. 2), and on purely thermodynamic grounds it would thus be expected that Cr203 would be the reaction product in the oxidation of Co-Cr and Ni-Cr alloys. However, this would require a continuous supply

34 P. KOFSTAD

Cont Cr203 scale with assolv. Co or with CoCr2C\

Co-10Cr. 1200C 100torr 02. 3hrs

Co-35Cr, 1300C 2 torr 02. 22hrs

FIG. 19. Schematic illustration of processes taking place in binary Co-Cr alloys.88~90 (a) Initial, transient oxidation involving formalion of both cobalt and chromium oxide, oxygen dissolution in the alloy, diffusion of chromium to the metalscale interface, internal oxidation of chromium beneath the CoO: (b) insufficient concentration of chromium in the alloy ( < ~ 30 " Cr) ; the reaction products consist of a two-layered surface scale and internal oxidation of chromium; the outer layer of the surface scale consists of CoO ( + C o 3 0 4 at 7" ~ 3 0 % Cr); selective oxidation of chromium; (d) metallographic cross-section of a Co-Cr alloy oxidized for 3 h at 1200C and 100 torr 0 2 ; (e) metallographic cross-section of a Co-Cr alloy oxidized for 22 h at 1300C and 2 torr O,. After

Ref. 91.


of chromium from the inner part of the alloy to the metal/oxide phase boundary, i.e. it requires that both the concentration and diffusivity of chromium in the alloy are sufficiently high that chromium is always in ample supply at the metal/oxide interface. If this condition is not fulfilled, the other alloy component will be oxidized, and the reaction mechanism becomes a function of several factors which include the composition of the alloy and rates of transport in both the alloy phase and the oxide scale.

This may be considered in more detail starting with a clean Co-Cr alloy surface. When exposed to oxygen gas, both the surface atoms of cobalt and chromium are oxidized and initially form Co O (and Co304 at r

36 P. KOFSTAD

FIG. 20. Metallographic cross-section of Co-10%Cr alloy specimen oxidized for 193 h in air at 900C. The outer layer consists of CoO and the inner layer of CoO and CoCr 20 4 . In addition chromium is internally

oxidized.88"90 100 magnification.

FIG. 21. Metallographic cross-section of Co-25%Cr alloy specimen oxidized for 236 h in air at 1000C. The scale contains the same phases as that for the Co-10%Cr alloy (Fig. 20), but the inner layer contains a proportionately larger concentration of the spinel.88"90 100

magnification.


conditions continuous, protective scales of Cr203 are built up at and above a critical concentration of chromium: this critical con-centration is, in turn, a function of temperature and ambient partial pressure of oxygen.

At lower chromium concentrations a continuous layer of Cr203 fails to develop and the oxidation reaches a steady state and involves formation of Co O and Cr203 ; these also undergo a solid state reaction to yield the CoCr204 spinel. A schematic diagram of the composition of the external oxide scale and the internally oxidized zone and of the reaction mechanism is shown in Fig. 19. An outer layer of the scale consists of Co O (under conditions where Co304 is not formed), an inner layer consists of Co O + CoCr204, while the internally oxidized zone consists of Cr203 particles in cobalt metal. Particularly the inner layer of the external scale (consisting of CoO + CoCr204) also develops relatively large amounts of voids and closed porosity. The rate of reaction is governed by solid state diffusion of cobalt through the Co O part of the inner and outer layers; in addition oxygen can move across the closed pores and this results in an inward growth of the CoO layer. Finally oxygen dissolves in the alloy substrate forming internal particles of Cr203 . Because of the inward growth of CoO (caused by oxygen transport across the pores), the internally formed Cr203 particles become embedded in CoO, and these then react to form the CoCr204 spinel.88 ~91 More detailed illustrations of cross-sections of oxidized Co-Cr specimens are given in Figs. 20 and 21. Figure 20 refers to a Co-10%Cr specimen oxidized for 193 h in air at 900 C, while Fig. 21 shows a Co-25 %Cr specimen oxidized for 236 h at 1000 C.

As the concentration of chromium in Co-Cr alloys is raised, an increasingly larger fraction of the inner layer will consist of the CoCr204 spinel. Diffusion through this oxide is slow compared to that in the CoO phase, and essentially the CoCr204 particles block part of the area for diffusional transport through the scale, and as a consequence the rate of oxidation of Co-Cr alloys gradually decreases with increasing chromium concentration until the critical Cr concentration is reached where a continuous Cr203 layer is formed. This dependence of the parabolic rate constant on Cr concentration is illustrated in Fig. 22 for oxidation at 1100C.

38 P. KOFSTAD 1000

100

<

-r^-r

\ \ \ \ I 1 t I 1 atm. 02 I I

1100C

Co-Cr alloys

10 torr 02 I I I I I I I I I Cr. 1 atm 02

~^-> ^ 0 10 20 30 60 50 90 100

WEIGHT PERCENT Cr

FIG. 22. The parabolic rate constant for oxidation of cobaltchromium alloys at 1100C as a function of chromium content. Results of Kofstad

and Hed.8 8"9 0 After Ref. 91.

V *'....^ ''I.W'.ii . . . . ^ "S*

" ..i . i . ".''.%.'^.' ;V'.'.V : f ' ;^ >'. .'ty.

2

FIG. 23. Metallographic crosssection of Co10 %Cr specimen oxidized in air for 140h at 1000C. Section near scale/metal interface. Note the

preferential internal oxidation along the grain boundaries.8890

CORROSION IN SINGLEREACTANT GASEOUS ENVIRONMENTS 3 9

At chromium concentrations lower than 10 %, the oxidation rate increases with increasing Cr concentration. This is primarily believed to be associated with effects of voids and porosity developing in the scale, and thus rapid gaseous transport across these pores overrides the effect of diffusional blockage by the CoCr 2 0 4 particles in the inner scale.

The mode of internal oxide formation will depend on the mechanism of oxygen diffusion in the alloy, e.g. on the relative importance.of grain boundary and volume diffusion of oxygen in the substrate. As a general rule grain boundary diffusion becomes increasingly important the lower the temperature, and correspondingly internal oxidation along grain boundaries will dominate at low temperatures, while homogeneous internal oxidation resulting from volume diffusion is most important at high temperatures. An example of such temperature dependence on the mode of internal oxidation is shown in Figs. 23 and 24 which refer to Co10%Cr specimens reacted at 1000 and 1200C, respectively.8890 The 1000C specimen is clearly characterized by preferential internal oxidation along grain boundaries, while a more homogeneous nucleation is observed at 1200C.

It is probable that internal oxidation along grain boundaries will

..;; . sBy.'*4k . . Vii,. :*W. \ .

40 P. KOFSTAD

be more detrimental as regards mechanical properties. On the other hand, if the external scale is directly connected with preferentially formed oxide spikes along grain boundaries at the scale/substrate interface, the scale can be more resistant to loss of scale adherence and oxide spallation.

The main features of this reaction mechanism for CoCr alloys also apply to a large number of other alloys such as NiCr, CoAl, Ni, Fe alloys, etc.1 5

Although these are the general aspects it may be noted that additional factors can influence the reaction behaviour. It is, for instance, well established that the transition to selective oxidation of chromium takes place at lower chromium contents for coldworked alloys and alloys with small grain size.92 It is generally concluded that this is due to a relatively rapid diffusion of chromium along grain boundaries and shortcircuit diffusion paths. The transition to selective oxidation is also a function of specimen geometry. Thus for Co35%Cr specimens the flat surfaces of rectangularly shaped specimens will exhibit selective oxidation of chromium, while edges and corners will mostly have doublelayered scales.90 This is related to the surfacetovolume ratio at the edges and corners.

7.2. Effects of Aluminium Additions to NiCr Alloys As the complexity of the alloy compositions is increased the same

main principles apply. Consider briefly a ternary alloy such as NiCrAl. Apart from the good protective properties of possible A1203 scales which may be formed, alloy additions of aluminium are important due to the y'(Ni3Al)strengthening mechanism.

Alumina, A1203, is thermodynamically more stable than Cr203 (Fig. 2). Furthermore, diffusion in A1203 is considerably slower than in Cr203 ,9 and thus continuous scales of A1203 formed through selective oxidation of the Alconstituent provide considerably better oxidation resistance than corresponding Cr203 scales. On the other hand, under conditions where aluminium does not become selectively oxidized, the aluminium constituent in the alloy will promote internal oxidation. Because of the high stability of A1203, aluminium will, for instance, be internally oxidized in the substrate below continuous, compact scales of Cr203 .


7.3. Oxide Maps A certain critical concentration again applies where the reaction

during steady state conditions changes from one mechanism to another. For some alloy systems sufficient data are available to permit the construction of 'oxide maps' which delineate the composition ranges for formation of different types of oxide scales and reaction behaviour. An example of this is illustrated in Fig. 25 for the oxidation of Ni-Cr-Al alloys at 1000 C; the three different types of oxides formed and the reaction behaviour are delineated.93-95

In region I the scale consists of NiO, Ni-Cr and Ni- spinels, and in addition the alloy is internally oxidized. The rate of oxidation is primarily determined by the Ni-diffusion in NiO in the scale. In region II the scale consists of Cr203 and in addition aluminium is internally oxidized. The oxidation is primarily determined by the transport through the Cr203 scale. In region III selective oxidation of aluminium to form a continuous surface scale of -1203 occurs. The oxidation rate is determined by diffusion through the A1203 scale.

Similar maps may be constructed for other alloy systems. In general the ease of formation of protective -1203 scales is in the order of Fe-Cr-Al > Ni-Cr-Al > Co-Cr-Al.96

WEIGHT PERCENT NI FIG. 25. 'Oxide map' for alloys in the Ni-Cr-Al system delineating the composition ranges for formation of different types of oxide scales. In region I the scale consists of NiO, NiCr204 and NiAl204 spinels and in addition the alloy is internally oxidized; in region II the scale consists of Cr203 and in addition aluminium is internally oxidized; in region III aluminium is selectively oxidized.93-95 Previously published in Ref. 91.

42 P. KOFSTAD

The width of the internal zone or whether selective oxidation will occur, will in every case be determined by the relative rates of diffusion in the alloy of oxygen and the reactive metal ions (e.g., Al). The delineations between the three different regions should not be considered absolute values. As previously noted the transitions to selective oxidation will be dependent on various factors such as pretreatment, grain size, specimen geometry, etc.

It should be emphasized that such oxide maps today only represent a summary of empirical, experimental data. A great challenge for investigators in the field is to be able to derive such maps from more basic data involving the interplay of thermodynamics, diffusion data and interface reactions. A most valuable approach in this respect is through the 'diffusion composition path method'.96

7.4. Effects of Temperature on Selective Oxidation The delineation between the different regions in the oxide maps

will also change with temperature. The reason for this is that the diffusion rates of oxygen and the different components in the alloy will have different temperature dependencies, i.e. their activation energies are generally different.

The extent to which aluminium becomes selectively oxidized in a given NiCrAl alloy will depend on the temperature. An example of this is illustrated in Fig. 26, which shows microprobe scans and metallographic crosssections of Ni9Cr6Al specimens oxidized in 1 atm. 0 2 at temperatures ranging from 900 to 1200 C.95 At 900 C the oxide is double or multilayered and consists of NiO, Cr203 , NiAl204 and, in addition, aluminium is internally oxidized; at 1200 C the oxide scale only consists of A1203 after extended oxidation, i.e. aluminium is selectively oxidized.

This change in reaction behaviour and the transition in the composition of the oxide scale markedly affects the rate of oxidation as illustrated in Fig. 27.95 While the oxidation rate increases from 900 to 1100C, when multilayered scales are formed and internal oxidation takes place, the oxidation rate drops with further temperature increase due to the formation of a continuous 1203 scale with better protective properties. With still further temperature


20 0 20 0 20 0 20 DISTANCE FROM ALLOY/OXIDE INTERFACE,microns

FIG. 26. Microprobe scans and Al x-ray images for Ni-9Cr-6Al specimens oxidized in 1 atm oxygen in the temperature range 900-1200C. Aluminium becomes more and more preferentially oxidized the higher the

temperature.95

2.0

1.6

E 12 < 0 8

04

Ni ~ Air

-

~/&

V w/o Cr I

-5.8W/OAI

I

1 1

i o o * c ^ "

1300'C

IOOO'C 1200'C Dou'c

1 1

-

-

-

20 30 40 50 TIME,h

FIG. 27. Oxidation of Ni-9Cr-6Al in the temperature range 900-1300C. Weight gain of specimen as a function of time"5

44 P. KOFSTAD increase, the oxidation rate again increases as the oxidation now is controlled by diffusional transport through 1203. The overall effect is thus a maximum in oxidation at about 1100 C where the oxidation mechanism changes. 7.5. Effect of Pretreatment/Experimental Procedure

It has already been noted that different pretreatments may significantly affect the reaction of metals and alloys. By using different experimental procedures one may in fact subject the metal specimens to different pretreatments before the start of the recorded experiment. For instance, marked differences may be observed depending on whether a specimen is heated in the oxidizing atmosphere up to the test temperature, whether it is exposed immediately to test conditions, or whether it is brought to temperature in good vacuum before being exposed to the oxidizing atmosphere. An example is shown in Fig. 28 for comparative runs at 1000 C on a Ni9Cr6Al0 1Y alloy.98 In one case an annealed and subsequently polished specimen was introduced directly into the oxidation apparatus at 1000C and 1 atm. 0 2 . In the second case the

E i < E 3 o

<

Ni

y / I I

-

r-

1

-9Cr-

i

6 Al

1

-0.1 Y

1 1 1

1000C 1atm. 02

- Annealed polished immediately before oxidation

Anneal polish as above 1 hr. vac. anneal (10-5torr)

I

_

-

-

-

20 40 60 TIME, h

60 100 120

FIG. 28. Effect of pretreatment on oxidation of Ni9Cr6Al01Y in 1 atm. 0 2 at 1000C: specimen (a) was annealed for 1 h at 1000C in 10~6 torr and then polished immediately prior to exposure under oxidizing conditions; specimen (b) annealed and repolished as specimen (a), but was in addition treated for 1 h at 10" 5 torr prior to exposure in oxygen. During this vacuumanneal a continuous, protective film of 12 0 3 was formed.98


specimen was held in 10~5 torr at 1000C for 1 h before 1 atm. 0 2 was introduced. In the latter case the oxidation is lower by about a factor of 10. This is due to the fact that the specimen developed a virtually pure, continuous -1203 film with excellent protective properties during the vacuum treatment. When the specimen was introduced directly into the furnace at test conditions, the initial transport oxidation involved formation of the different oxides NiO, Cr203 , A1203 and corresponding spinels, and considerable oxidation was needed to establish a steady-state continuous -1203 layer.

Such pre-annealing and formation of highly protective films by selective oxidation may have some merit in improving oxidation resistance. However, if subsequent spalling of the oxide scale takes place, such initial beneficial effects will be lost. 7.6. Oxide Evaporation

Vaporization losses from oxide scales may in cases prove detrimental in the oxidation of high-temperature alloys. In dry oxygen atmospheres, losses from Si02 and Cr203 scales are of particular concern, but in atmospheres containing water vapour other oxides may also evaporate significantly through the formation of volatile hydroxide species.

Silicon and suicides form highly protective scales of Si02. But at reduced oxygen pressures Si02 scales are broken down or are not formed due to formation of Si O :

Si+|02SiO(g) The transition from oxidation involving SiO-evaporation to protective Si02-formation takes place at critical partial pressures of oxygen; these are illustrated graphically in Fig. 29 for varying temperatures. This evaporation means that silicon and suicides, which form highly protective Si02 scales at near-atmospheric pressures, do not form protective scales and cannot be used at low partial pressures of oxygen." In steam of high pressures, Si02 also evaporates as the hydroxide and leaves the silicon surface free from oxide (Fig. 29).100

Cr203 does not evaporate as such, but as Cr03 . In the presence of

46 P. KOFSTAD TEMPERATURE/C

UOO 1200 1000 700 600 10 1 1 1 . 1 L

S

Iff'

4 -

gio- HT

\ h h 1 ' l ' I ' ! Oxid* evaporation presene

___ of HjO*'. No surface oxidefc

PHD

I ,Melting point of Si HIO'C

Si 02-tormation

500

FIG. 29. The partial pressures of oxygen at which oxidation of silicon changes from protective (passive) Si02formation to nonprotective oxidation (active) involving SiOevaporation. The figure further illustrates the critical pressures of water vapour above which Si02 is lost through

vaporization.99100

both oxygen and water vapour volatile species of Cr02(OH) are also formed. The reactive vaporization of Cr203 takes place through the reaction:

i C r 2 0 3 + f 0 2 ^ C r 0 3 ( g ) and accordingly the rate of vaporization is proportional to/?0/4. Due to this large dependence on the partial pressures of oxygen, Cr03 formation is not important at reduced oxygen pressures.

Due to the simultaneous oxide formation and reactive vaporization of Cr203 , the rate of Cr203formation can be described as the sum of a parabolic, diffusioncontrolled scale thickening and a timedependent vaporization loss:

dt Kb where kb is the rate constant for the vaporization loss. After


TIME, h

FIG. 30. Comparison of oxidation of Ni-30Cr and TD NiCr (Ni-20Cr-2 vol.% Th02) at 1200C.37 Weight change as a function of time. Both alloys form an external Cr203 scale. In oxidation of Ni-30Cr the marker is located at the metal/oxide interface, while for TD NiCr the marker is located at the surface of the scale. Results of Wallwork and Hed94 and

Giggins and Pettit.93

extended oxidation the two terms on the right-hand side of the equation become more and more equal and a steady-state scale thickness is approached. In thermogravimetric measurement of oxidation rates this leads under the steady state conditions to a continuous loss of the specimens. This behaviour is illustrated in Fig. 30.9394

The extent of oxide vaporization losses is markedly dependent on the rate of gas flow past the specimens. In static gas a large fraction of the evaporating molecules are reflected back on the surface. The oxide loss increases with gas flow, and Lowell and co-workers101 report that the weight loss may be 100 times larger in 1 Mach gas streams at 1200C than under static conditions. The high velocity values are also close to the maximum possible theoretically calculated Hertz-Langmuir rates for free evaporation.

48 P. KOFSTAD

The vaporization losses may be reduced by the presence of other oxides such as NiO or spinels on top of the Cr203 layer; surface layers of A1203 effectively inhibit the evaporative loss from Cr03 .

7.7. Alloys with Oxide Dispersions During recent years interesting and highly beneficial effects of

oxide dispersions on the oxidation behaviour of alloys have been found and studied.102 Th02 , A1203, Y203 , etc. have been used as oxide dispersions.

In the case of chromia-forming alloys the oxide dispersions produce four main effects:

1) Selective oxidation of chromium is enhanced, i.e. con-tinuous, protective layers of Cr203 are achieved at lower chromium contents in the alloys.

2) The rate of growth of the Cr203 layer is reduced. 3) The growth mechanism of the Cr203 layer can be changed.

Cr203 layers appear to grow inward (oxygen diffusion) on alloys with oxide dispersions.

4) The adherence of Cr203 scales to the alloy substrate is greatly improved.

The two first features are illustrated in Fig. 30, which shows typical oxidation data for TD NiCr(Ni-20wt %Cr-2vol. %Th02) and for Ni-30wt%Cr at 1200 C. As regards the change in growth mechanism platinum markers are found at the metal/oxide interface for alloys without oxide dispersions and at the oxide/scale surface for oxide-dispersed alloys. Th02 particles were not observed after short time (< 1 h) oxidation, while numerous Th02 particles were found in the scale after extended oxidation.

The enhanced selective oxidation of chromium has been interpreted as being due to the presence of a fine grain substructure and associated dislocation density, stabilized by the oxide particles, and that this results in rapid short-circuit diffusion of chromium in the alloy and correspondingly rapid development of protective Cr203 . However, chromium diffusion studies on TDNiCr and Ni-20Cr do not show differences in diffusivities in the two types of materials for a given grain size

CDNA08652ENC_001.pdf

Documents

high temperatures

norton joint research

norton elsevier

european authors

european level

following information

science publishers london

aggressive gaseous atmospheres