H. J. Grabke, M. Schutze
Oxidation of Intermetallics
8WILEY-VCH
Further Reading from WILEY-VCH
Gerhard Sauthoff Intermetallics ISBN 3-527-29320-5
K. H. Matucha (Ed.) Structure and Properties of Nonferrous Alloys
ISBN 3-527-26821-9 Volume 8 of the series Materials Science and Technology R. W. Cahn, P. Haasen, E. J. Kramer (Eds.)
R. W. Cahn (Ed.) Processing of Metals and Alloys
ISBN 3-527-26828-6 Volume 15of the series Materials Science and Technology R . W. Cahn,P. Haasen, E. J. Kramer (Eds.)
G. Petzow, G. Effenberg (Eds.) Ternary Alloys A Comprehensive Compendium of Evaluated Constitutional Data and Phase Diagrams Please write for our comprehensive Prospectus
H. J. Grabke, M. Schutze
Oxidation of Intermetallics
@ WILEY-VCH
Weinheim . Berlin . New York . Chichester Brisbane Singapore - Toronto
Prof. Dr. H. J. Grabke Max-Planck-Institut fur Eisenforschung Max-Planck-StraBe 1 40237 Dusseldorf Germany
Dr. M. Schiitze Karl-Winnacker-Institut der DECHEMA e.V. Theodor-Heuss-Allee 25 60486 Frankfurt/Main Germany
This book was carefully produced. Nevertheless, authors, editors and publisher do not warrant the information contained therein to be free of errors. Readers are advised to keep in mind that statements, data, illustrations, procedural details or other items may inadvertently be inaccurate.
i
Editorial Director: Dr. Jorn Ritterbusch Production Manager: Hans-Jochen Schmitt Cover picture: Metallographic section of the oxide scale and the subsurface zone of a twophase TiAl alloy after oxidation in air at 1350 "C for 1 h (interference layer metallography) Every effort has been made to trace the owners of copyrighted material; however, in some cases this has proved impossible. We take this opportunity to offer our apologies to any copyright holders whose rights we may have unwittingly infringed. Library of'Congress Card No. applied for. A catalogue record for this book is available from the British Library.
Deutsche Bibliothek Cataloguing-in-Publication Data: Oxidation of intemetallics / H. J. Grabke ; M. Schiitze. - Weinheim ; Berlin ; New York ; Chichester ; Brisbane ; Singapore ; Toronto : Wiley-VCH, 1997 ISBN 3-527-29509-7 0 WILEY-VCH Verlag GmbH, D-69469 Weinheim (Federal Republic of Germany), 1998 Printed on acid-free and chlorine-free paper All rights reserved (including those of translation into other languages). No part of this book may be reproduced in any form - by photoprinting, microfilm, or any other means nor transmitted or translated into a machine language without written permission from the publishers. Registered names, trademarks, etc. used in this book, wen when not specifically marked as such, are not to be considered unprotected by law. Composition: Mitterweger Werksatz, D-68723 Plankstadt Printing: Strauss Offsetdruck, D-69509 Morlenbach Bookbinding: Groabuchbinderei J . Schaffer, D-67269 Griinstadt Printed in the Federal Republic of Germany
Preface
About 10 years ago, many research activities were started in Europe, USA and Japan on the mechanical properties and oxidation resistance of intermetallic alloys at high temperatures, striving for new structural materials for application in engines and turbines. Especially aluminides and silicides appeared to be promising, which show high melting point, low density and high strength, and possibly good oxidation resistance. Vast efforts were made in research and development to attain improved high temperature strength and low temperature ductility, much less research and testing was performed on the resistance of intermetallics against oxidation and high temperature corrosion. Oxidation and corrosion resistance, however, are as important for actual application of materials as mechanical properties. Aluminides and silicides were expected to form protective alumina or silica layers and especially for the compounds with high A1 or Si content no great problems were foreseen. But in fact many phenomena were observed, which are deleterious for corrosion resistance: void formation beneath the scales, cracking and spalling, growth of unprotective oxides such as TiO, and Nb,O, on Ti-and Nb-aluminides, effects of nitride formation, even internal and accelerated oxidation, intergranular oxidation and ‘pesting’, i. e. disintegration of materials ... so a wide field opened for the study of kinetics and mechanisms of oxidation and corrosion attack. But also means of improving the oxidation resistance were found and from the relatively susceptible intermetallic compounds more resistant multi-component and multi-phase alloys were developed. After now, the possibilities and limitations of intermetallic compounds have been widely explored, and the research activities begin to decrease, it was time to try to get an overview on the present knowledge. The call for contributions brought 27 papers from Germany, Great Britain, Italy, and the Netherlands. Additionally, three keynote speakers were invited, G. Suuthoff to give an overview on the present state of intermetallics development and G. H. Meier and S. Tunigicchi to give keynote lectures on the research activities in the USA and Japan, on oxidation of intermetallics. In the workshop (held in Frankfurt, Germany, January 18-19,1996 and attended by 65 participants) the presentations and vivid discussions gave a rather complete view, especially of the oxidation and corrosion of the Ti-, Ni- and Fe-aluminides. Now this and the two next volumes of ‘Materials and Corrosion’ present a part of the papers in full length and another part as extended abstracts with references on where to find the complete work. The organizers of the workshop would like to extend sincere thanks to authors for their great effort in contributing to a fine compilation of work on the ‘Oxidation of Intermetallics’.
April 1996,H. J. Grubke
List of Contributors
P. Andrews Rolls Royce plc Derby United Kingdom
W. Auer Institute of Materials Science Chair for Corrosion and Surface Technology University of Erlangen-Nurnberg Martensstr. 7 91058 Erlangen Germany
M. J. Benett Coatings and Interfaces Section AEA Technology Harwell Laboratory Didcot Oxfordshire OX11 9AH United Kingdom M. Bobeth Max-Planck-Gesellschaft Research Group on Mechanics of Heterogeneous Solids Hallwachsstr. 3 01069 Dresden G e r rna n y
L. B. Bradley Corrosion and Protection Center University of Manchester Institute of Science and Technology Manchester M60 1QD United Kingdom
M. W. Brumm Max-Planck-Institut fur Eisenforschung GmbH Postfach 14 04 44 40074 Dusseldorf Germany S. J. Bull Coatings and Interfaces Section AEA Technology Harwell Laboratory Didcot Oxfordshire OX1 1 9AH United Kingdom
K. T. Chuah Corrosion and Protection Center University of Manchester Institute of Science and Technology Manchester M60 1QD United Kingdom F. Dettenwanger Max-Planck-Institut fur Metallforschung Institut fur Werkstoffwissenschaft Seestr. 92 70174 Stuttgart Germany M. Eckert Research Center Julich Institute of Materials in Energy Systems P.O. Box 1913 52425 Jiilich Germany
VlII
List of Contribicfors
B. Eltester Max-Planck-Institut fur Eisenforschung GmbH Postfach 14 04 44 40074 Dusseldorf Germany
A. H. H. Janssen Netherlands Energy Research Foundation ECN PO Box 1 1755 ZG Petten The Netherlands
A. Gil University of Mining and Metallurgy 30059 Krakow Poland
J. Klower KruppVDM CimbH Plettenberger Str. 2 58791 Werdohl Germany
H. J. Grabke Max-Planck-Institut fur Eisenforschung GmbH Postfach 14 04 44 40074 Diisseldorf Germany M. J. Graham National Research Council Institute for Microstructural Sciences Ottawa, K1A OR6 Canada V. A. C. Haanappel Institute for Advanced Materials Joint Research Center of the European Commission 21020 Ispra (Va) Italy
K. Hilpert Research Center Julich Institute of Materials in Energy Systems PO. Box 1913 52425 Julich Germany M. Hollatz Max-Planck-Gesellschaft Research Group on Mechanics of Heterogeneous Solids Hallwachsstr. 3 01069 Dresden Germany
R. Klumpes Delft University of Technology Lab. for Materials Science Div. of Corrosion Technology and Electrochemistry Rotterdamseweg 137 2628 AL Delft The Netherlands V. Kolarik Fraunhofer-Institut fur Chemische Technologie 76327 Pfinztal Germany R. Krajak Max-Planck-Institut fur Eisenforschung GmbH Postfach 14 04 44 40074 Diisseldorf Germany R. Kremer Institute of Materials Science Chair for Corrosion and Surface Technology University of Erlangen-Niirnberg Martensstr. 7 91058 Erlangen Germany
List of Contributors
C. Lang Karl-Winnacker-Institut der DECHEMA e.V. Theodor-Heuss- Allee 25 60486 Frankfurt/M. Germany J. Leggett Cranfield University Cranfield Bed f ord United Kingdom C. H. M. Maree University of Utrecht Department of Atomic and Interface Physics 3508 TA Utrecht The Netherlands
IX
W. Pompe Max-Planck-Gesellschaft Research Group on Mechanics of Heterogeneous Solids Hallwachsstr. 3 01069 Dresden Germany W. J. Quadakkers Forschungszentrum Julich GmbH IWE 1 52425 Jiilich Germany
J. Rakowski Department of Materials Science and Engineering University of Pittsburgh Pittsburgh, PA 15261 USA
G. H. Meier Department of Materials Science and Engineering University of Pittsburgh Pittsburgh, PA 15261 USA
I. Rommerskirchen Krupp VDM GmbH Plettenberger Str. 2 58791 Werdohl Germany
J. R. Nicholls Cranfield University Cranfield Bedford United Kingdom
M. Ruhle Max-Ylanck-Institut fur Metallforschung Institut fur Werkstoffwissenschaft Seestr. 92 70174 Stuttgart Germany
M. Palm
Max-Planck-Institut fur Eisenforschung GmbH Postfach 14 04 44 40074 Dusseldorf Germany
G. Sauthoff Max-Planck-Institut fur Eisenforschung GmbH Postfach 14 04 44 40074 Dusseldorf Germany
B. A. Pint Oak Ridge National Laboratory PO. Box 2008 Oak Ridge,TN 37831-6156 USA
P. Schaaf Universitat Gottingen 37073 Gottingen Germany
X
List of Conlributors
Institute for Advanced Materials Joint Research Center of the European Commission 21020 Ispra (Va) ltaly
S. StrauB Max-Planck-Institut fur Eisenforschung GmbH Postfach 14 04 44 40074 Dusseldorf Germany
B. Schramm
M. F. Stroosnijder
H. J. Schmutzler
Institute of Materials Science Chair for Corrosion and Surface Technology University of Erlangen-Nurnberg Martcnsstr. 7 91058 Erlangen Germany E. Schramm Delft University of Technology Lab. for Materials Science Div. of Corrosion Technology and Electrochemistry Rotterdamseweg 137 2628 A L Delft ‘The Netherlands E. Schumann Max-Planck-Institut fur Metallforschung Institut fur Werkstoffwissenschaft Seestr. 92 70174 Stuttgart Germany
M. Schiitze
Karl-Winnacker-Institut der DECHEMA e.V. Theodor-Heuss- Allee 25 60486 FrankfurtIM. Germany
F. H. Stott Corrosion and Protection Center University of Manchester Institute of Science and Technology Manchester M60 1QD United Kingdom
Institute for Advanced Materials Joint Research Center of the European Commission 21020 Ispra (Va) Italy
J. D. SunderkGtter Institute for Advanced Materials Joint Research Center of the European Commission 21020 lspra (Va) Italy S.Taniguchi Department of Materials Science and Processing Faculty of Engineering Osaka University’ 2-1 Yamadaoka Suita, Osaka 565 Japan P. ETortorelli Oak Ridge National Laboratory PO. Box 2008 Oak Ridge,TN 37831-6156 USA
J. P. 7’.Vossen Netherlands Energy Research Foundation ECN PO Box 1 1755 ZG Petten The Netherlands
B. Wagemann Max-Planck-lnstitut fur Eisenforschung GmbH Postfach 14 04 44 40074 Dusseldorf Germany
E. Wallura Forschungszentrum Julich GmbH IWE 1 52425 Julich Germany J. H. W. de Wit Netherlands Energy Research Foundation ECN PO Box 1 1755 ZG Petten The Netherlands J. H. W. de Witt Delft University of Technology Lab. for Materials Science Div. of Corrosion Iechnology and Electrochemistry Rotterdamseweg 137 2628 AL Delft The Netherlands
I. G. Wright Oak Ridge National Laboratory P.O. Box 2008 Oak Ridge.TN 37831-6156 USA J. C. Yang Max-Planck-Institut fur Mctallforschung Institut fur Werkstoffwissenschaft Seestr. 92 70174 Stuttgart Germany
N. Zheng Forschungszentrum Julich GmbH IWE I 52425 Julich Germany
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
V
List of Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
VII
Part I Introduction 1
2 3
State of Intermetallics Development .......................... G. Sauthoff Research on Oxidation and Embrittlement of Intermetallic Compounds in the U.S. ...................................... G, H. Meier Oxidation of Intermetallics - Japanese Activity . . . . . . . . . . . . . . . . S. Taniguchi
3
15 59
Part I1 Ni-Aluminides 4 5 6
7
8 9
10
The Oxidation of NiAl and FeAl .............................. 79 H. J. Grabke, M. W Brumm, B. Wagemann Sulfidation Behaviour of Nickel Aluminides . . . . . . . . . . . . . . . . . . . 85 B. Schramm, W Auer The Influence of Chromium on the Oxidation 99 of P-NiAI at 1000 "C ........................................ R. Klumpes, C. H. M. Marie, E, Schramm, J. H. W de Wit Oxidation of P-NIAI, Undoped and Doped with Ce,Y, Hf . . . . . . . 109 I. Rommerskirchen, I/: Kolarik TEM Investigations on the Oxidation of NiAl . . . . . . . . . . . . . . . . . . 121 E. Schumann, J. C. Yang, M. J. Graham, M. Ruhle Failure of Alumina Scales on NiAl Under Graded Scale Loading .............................................. 135 M. Hollatz, M. Bobeth, W Pompe The Corrosion Behaviour of NiAl in Molten Carbonate 161 at 650 "C ................................................... J. I? T Vossen,A. H. H. Janssen, J. H. W de Wit
XIV
Contents
Part 111 Fe-Aluminides 11 12
13 14
15
Oxidation of P-FeAl and Fe-AI Alloys ......................... 175 I. Rommerskirchen, B. Eltester, H. J. Grabke The Oxidation Behaviour of ODS Iron Aluminides . . . . . . . . . . . . . 1S3 B. A. Pint, F! F: Tortorelli and I. G. Wright High Temperature Corrosion Behaviour of Iron Aluminides and Iron-Aluminium-Chromium Alloys ....................... 203 J. Klower 221 Oxidation-Sulphidation of Iron Aluminides .................... E H. Stott, K. 7: Chuah, L. B. Bradley Metal Dusting of Fe,AI and (Fe, Ni),AI ........................ 233 S. Strauo., R. Krajak, M . Palm, H. J. Grabke
Part IV Ti-Aluminides 16
17 18
19 20
21
22
23
Determination of Thermodynamic Activities in the Alloys of the Ti-AI System and Prediction of the Oxidation Behaviour of the Alloys ................................................ M , Eckert, K. Hilpert The Initial Stages in the Oxidation of TiAl ..................... C. Lang, M. Schiitze Development and Microstructure of the Al-Depleted Layer of Oxidized TiAl ............................................. E Dettenwanger, E. Schumann, J. Rakowski, G. H. Meier, M. Riihle Beneficial and Detrimental Effects of Nitrogen on the Oxidation Behaviour of TiAl-Based Intermetallics ....................... M! J. Quadakkers, I? SchaaJ N Zheng,A. Gil, E. Wallura Influence of Moisture on the Oxidation of y-XAl . . . . . . . . . . . . . . . R. Kremer, M!Auer Ion Implantation as a Tool to Study the Oxidation Behaviour on TiAl-Based Intermetallics ................................. M. E Stroosnijder, H. J. Schmutzler,V A. C, Haanappel, J. D. Sunderkotter Protection of Titanium Aluminides by FeCrAlY Coatings . . . . . . . M. J. Bennett, S. J. Bull Hot Salt Corrosion of Titanium Aluminides .................... J. R. Nicholls, J. Leggett, F! Andrews
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
239 245 265 275 289 299 313 329 345
Part I Introduction
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
1 State of Intermetallics Development G. Sauthoff
1.1 Introduction
It is known since long that intermetallic phases excel by special physical and mechanical properties, the latter being high strength at high temperatures and low deformability at lower temperatures [l-31. Since the fifties it has been tried to develop new high-temperature materials on the basis of intermetallic phases without resulting in applications in the past because of processing problems [4,5]. Only more recent developments, which were initiated by the “ductilization” of Ni,AI by the addition of boron [6], have progressed successfully with centres of gravity in USA, Japan and Germany and are at the brink of commercialisation [7-lo]. The developments have been paralleled by extended fundamental research which, however, has not yet led to a clear physical understanding of the observed phenomena [2, 11-13]. In the following the state of development and the prospects of the various candidate phases for high-temperature applications are overviewed.
1.2 Titanium Aluminides 1.2.1 Ti,AI In the USA the a2alloys and super-a, alloys have been developed by alloying Ti,Al with large amounts of Nb which show sufficient ductility at room temperature and aim at applications in aircraft industry [14-161. The materials development is practically complete, production and processing technologies have been developed, and alloys are commercially available. However, these alloys are not yet applied because of insufficient oxidation resistance, higher demands for strength and high economic barriers [17-191. Corresponding developments have not been done in Germany because of the limited potential of these alloys, but possibilities for alloy optimization by microstructure control have been studied successfully 1201.
4
G. Southoff
1.2.2 TiAl Compared with Ti,AI, the Al-richer so-called y titanium aluminide TiAl is regarded as more promising because of lower density, higher strength at higher temperatures and higher oxidation resistance [18]. Thus the world-wide efforts to develop a lightweight alloy for application as high-temperature structural alloys are concentrated presently on ‘IiAI. The intensive developments in USA have led to the y titanium aluminide alloys which contain 48 at.% Al and about 2 at.% Nb, Cr and/or other transition metals and which are two-phase with Ti,AI as second phase [21-241. The mechanical behaviour as characterized by strength, ductility and toughness can be optimized by careful control of processing, i. e. microstructure which is lamellar, duplex or mixed. These developments aim at aerospace applications, whereas a TiAl turbocharger rotor has been developed in Japan for application in car engines [25].Further Japanese developments are subject of a national materials program [26]. The enormous potential of TiAl has been recognized early in Germany, Austria and Switzerland which gave rise to various ongoing developments [27-361. Development aims are applications as valves in car engines, turbocharger rotors, gasturbine blades and lightweight sheet material.
1.3 Nickel Aluminides 1.3.1 Ni,AI Ni,AI is known since long as the strcngthening y-phase in superalloys and has been described repeatedly in review papers [37-391. Characteristic is its anomalous temperature dependence of flow stress, i. e. increasing flow stress with rising temperature up to temperatures of about 700°C and normal softening at higher temperatures. Polycrystalline Ni,Al is practically not deformable at room temperature, whereas singlecrystalline Ni,AI is ductile. Polycrystalline Ni,AI with reduced Al content can be “ductilized” by microalloying with B. Ni,Al suffers from environmental embrittlement at room temperature, which is claimed to be caused by moisture in the atmosphere, and from high-tcmperature embrittlement at about 700°C which is reduced by Cr addition. On the basis of “ductilized” Ni,Al the so-called Advanced Aluminides were developed at the Oak Ridge National Laboratory which are now ready for application [38, 401. These alloys are recommended for applications in water turbines, steam turbines and gas turbines where high strength and high resistance against fatigue, wear, erosion, cavitation and oxidation is important. A problem is the limited hot-forming capability. Ni,Al-base alloys are also promising for applications in combustion engines as is exemplified by a development aiming at turbocharger rotors in large Diesel engines [41]. The potential of Ni,Al-base alloys is regarded as limited since it does not differ sufficiently from that of the Ni-base superalloys [42]. German development work concentrates on production technology and aims at making use of the high corrosion resistance of Ni,Al-base alloys [43].
I Slate of Internwtallics Development
5
1.3.2 NiAl Compared with NiAI, and conventional superalloys, NiAl is regarded as more promising for high-temperature structural applications because of higher melting temperature, lower density and higher thermal conductivity [4446]. Furthermore, NiAl shows an excellent oxidation resistance which excels that of most other intermetallics [47]. Problems are the low deformability at room temperature and the insufficient strength and creep resistance above 1000°C. It is noted that Al-deficient NiAl, which is instable, can transform martensitically, which is used for developing shape-memory alloys [48--501.Intensive developments in USA are aimed at applications as blades in flying gas turbines and have already led to components which have been tested successfully under service conditions 1511.The world-wide development efforts have been paralleled by cxtended fundamental research which has been reviewed repeatedly [52-551. In Germany the potential of NiAl-base alloys has been recognized early which initiated respective developments [56-591. These developments rely on the formation of second phases by alloying with third elements which improves high-temperature strength and creep resistance sufficiently with still tolerable room-temperature brittleness. A special development has been based on eutectic NiAl alloys with a ductile Cr, Mo or Re phase and aims at applications in flying gas turbines and car engines [60, 611. Another development by mechanical alloying makes use of oxide dispersions for increasing the creep resistance and of ductile particles for increasing the toughness [62,63]. A third development results from alloying NiAl with Ta and/or Nb to produce a strengthening Laves phase of type TaNiAl or NbNiAl [57,64-67]. Such alloys show not only an advantageous combination of high creep resistance at 1200°C and acceptable toughness at room temperature, but also an excellent thermoshock resistance and hot-gas corrosion resistance which makes them particularly promising for applications in car engines. A further development makes use of alloying with Fe to produce a second ductile particulate phase, and the resulting alloy with additional Cr shows an advantageous combination of strength, ductility and corrosion resistance against oxidation, carburization and sulphidation with a high potential for applications in coal conversion plants, petrochemical plants and industrial furnaces [68,69].
1.4 Iron Aluminides 1.4.1 Fe,Al The particular magnetic properties of Fe,AI resulted in the development of the magnetic head material Sendust, which is based on Fe,(AI,Si) and is applied world-wide in huge quantities [70).Presently Fe,AI is regarded as promising for structural applications, and respective developments have been started in USA [71-731. Fe,Al-base alloys show higher strength with anomalous temperature dependence in comparison to conventional iron alloys, and they excel by their corrosion resistance in oxidizing and sulphidizing atmospheres. Problems are the little deformability at room temperature and the strength decrease above about 500°C because of lattice transformation [74].
6
G. Sauthoff
The mechanical behaviour is improved by reduction of Al content, alloying with further elements and thermomechanical treatments, which leads to alloys for eventual applications in conventional power plants - in particular in steam turbines - or in coal conversion plants. Exhaust units in cars and electric heating elements are near application. In Germany Fe,AI has found only little interest up to now - apart from fundamental research [75,76] - and the little interest aims at applications in the petrocheniical industry [77].
1.4.2 FeAl FeAl shows the same crystal structure as NiAl, and the behaviour of FeAl is indeed similar to that of NiAI. A ductility of few percent at room temperature could be achieved by reducing the A1 content and appropriate processing [78]. FeAl is regarded as highly promising for high-temperature applications in view of the advantageous deformation and corrosion properties and the comparatively low density, and respective materials developments are on the way in IJSA, including the development of intermetallic matrix composites (IMC) with FeAl matrix [74,79-811.
1.5 Chromides Already in the past the Laves phase TiCr,, which crystallizes with the hexagonal C14 or C36 structure or the cubic C15 structure depending o n temperature and composition, was regarded as promising for high-temperature applications because of high strength and oxidation resistance 182,831. However, its high brittleness at room temperature has precluded any direct application. Present development efforts in IJSA are based on multiphase TiCr,-Ti-Nb alloys with the hard Laves phase in a ductile matrix, which offer an acceptable room-temperature toughness [84,85]. A rather similar Laves phase is NbCr,, which has been selected as candidate phase for high-tcmperature applications because of its high melting temperature, high strength and creep resistance, low density and potential oxidation resistance [86-88]. Again this phase is combined with ductile Cr-rich or Nb-rich phases as matrix or particles to obtain multiphase alloys with acceptable toughness.
1.6 Silicides Silicides represent the transition from intermetallics with predominantly metallic bonding to non-metallic compounds since-silicon is no longer a metal, but a semiconductor. Nevertheless the silicides are often comprised within the intermetallics. Silicides were selected for high-temperature applications already in the past because of their potentially high oxidation resistance at highest temperatures [89-911. Presently
1 Slate of Inrermetallics Developtnent
7
new developments are in progress [92].The general interest is concentrated on MoSi,, which has been applied since long as electric-heating material [93-951. Various other alloy systems have been selected, in particular Nb,Si,-Nb and Cr,Si-Cr, which allow ductile phase toughening [96-991. In Germany various silicides have been selected for materials developments. Ti,Si, excels by its high-temperature capabilities with melting temperature of about 2130°C and its low density of about 4g/cm3 [loo, 1011. However, this phase is brittle at lower temperatures, and its potential can be used only by combining this phase with more ductile phases. It has indeed been shown that Ti$,-Ti,Al alloys on the one hand [102, 1031 and Ti$,-Ti alloys on the other hand [loo, 1041 are promising as high-temperature lightweight materials. The high oxidation resistance of the disilicides MoSi, and TiSiz has initiated additional developments [105-1071. The magnesium-rich silicide Mg,Si has been selected as basis for the development of two-phase Mg,Si-Mg and Mg,Si-A1 lightweight alloys for application as piston material in car engines because of the outstandingly low density of only 1.88g/cm3of this phase [108-110].
1.7 Conclusions and Prospects New structural intermetallic alloys for high-temperature applications are in the centre of the present interest in intermetallics which is still growing. Some few developments, which are based on the classic phases Ni,AI, Ti,AI and TiA1, are on the brink of commercialization, but even these developments are still in an early stage compared with other developments of advanced materials, e. g. the modern engineering ceramics. Much more experimental and theoretical work is necessary for solving the processing problems and for adjusting the property spectra to the specific applications. The much advanced nickel aluminides and titanium aluminides can be used only up to about 1000°C because of limited strength or oxidation resistance or both at higher temperatures as has been stated before [3]. For applications significantly above 1000°C other less-common phases with higher melting temperatures have to be used. Such phases are available, and examples are shown in Fig. 1 [ lll] . In comparison to the nickel aluminides and titanium aluminides, the less-common phases are stronger and more brittle, their crystal structures are complex, their handling is difficult, and thus they are regarded as exotic. However, these exotic phases may fill the gap between metallic high-temperature alloys and ceramics, as is visible in Fig. 1.The brittleness of the less-common phases can be alleviated by combining them with softer phases to form multiphase alloys with adequate microstructures. Even strengthening hard phases may improve the toughness by impeding crack growth. The mechanical behaviour can be optimized by optimizing the microstructure, which requires a careful control of processing. However, it has to be emphasized that one cannot expect to obtain new intermetallic materials with properties similar to existing conventional metallic alloys. The “ductilizations” that have been achieved in few cases - in particular Ni,AI and (Fe,Co,Ni),V - rely on rather specific mechanisms and cannot be expected for other intermetallic phases. Thus intermetallic materials have to be regarded as a materials
8
G. Smithoff
25 E
Y
.s 20 v) v)
a L
t
v,
9 .-a
15
)5
U .-
y-
lo
!!i d 5
v)
0
500
1000 1500 temperature in O C
Fig.1. Specific yield strength (0.2 % proof stress in compression per unit weight density at 10-4s-1strain rate in compression) as a function of temperature for the DO,, phase AI,Nb (112. 1131,the Heusler-type phase Co,TiAl[67], the Laves phasesTiCr, sSi,, andTaFeAl[67,114]. the two-phase alloy NbNiAl-NiAl with 15vol.% NiAl in the Laves phase NbNiAl[67,114], and the hexagonal D8, phaseTi,Si, [lo01in comparison to the superalloy MA 6000 (in tension) [I 151 and the hot-pressed silicon rutride HPSN (upper limit of flexural strength) [116].
class of its own with property spectra which differ significantly from those of other materials and which can be varied within broad limits corresponding to metals on one side and non-metals on the other side. This offers enormous possibilities for manifold developments which are exciting with respect to;ooth practical applications and materials science. Finally it is noted that much interest is concentrated on the development of intermetallic alloys for application as blades in flying gas turbines. This application is most demanding and it is not clear when all the problems with strength, ductility, toughness and corrosion resistance can be solved at economic costs. Less-high-technology applications may be more rewarding presently for the introduction of new intermetallic materials. An example may be the car engine where strong light components with sufficient corrosion resistance are needed. Here brittleness is not the problem since designers have learnt to use ceramic materials for e. g. valves. However, new materials must be compatible with the metallic engine with respect to the physical properties, in
I State of Intermetallics Development
9
particular thermal expansion and thermal conductivity. This requirement corresponds to the characteristics of intermetallic materials which are hard and brittle with mainly metallic atomic bonding, i. e. metallic physical properties. Thus new intermetallic materials are expected to play an important role in the manufacture of car engines and similar applications.
1.8 References [l] G. Suuthoffi Intermetallics. In: K. H. Matucha (ed.) Structure and Properties of Nonferrous Alloys. Verlag Chemie, Weinheim (1995) 643-803. [2] G. Sauthoff:1ntermetallics.Verlag Chemie, Weinheim, (1995). [3] G. Suurhoff:Plastic Deformation. In: J.H. Westbrook and R. L. Fleischer (eds.) Intermetallic Compounds: Principles and Practice. John Wiley & Sons, Chichester (1995) 911-934. [4] J. H. Westbrook:Mechanical Properties of Intermetallic Compounds - A Review of the Literature. In: J. H. Westbrook (eds) Mechanical Properties of Intermetallic Compounds. J. Wiley. New York (1960) 1-70. [5] J. H. Westbrook: Silicides, Borides, Aluminides, Intermetallics and Other Unique Refractories. In: High Temperature Technology. McGraw Hill, New York (1960) 113-128. [6] K.Aoki, 0. Zzumi:J. Jap. Inst. Metals 43 (1979) 1190. [7] G. Sauthoff: Neue Strukturwerkstoffe auf der Basis intermetallischer Phasen - Stand und Perspektiven. BMFT-Symposium Materialforschung - Neue Werkstoffe, KFA-PLR, Jiilich (1994) 309-322. [8] H.-J. Engell, A . Von Keitz, G. Sauthoff: Intermetallics - Fundamentals and Prospects. In: W. Bunk (ed.) Advanced Structural and Functional Materials. Springer Verlag, Berlin (1991) 91-132. 191 G. Sauthoff: Internationaler Stand der Werkstoffentwicklungen auf der Basis intermetallischer Phasen. In: F. J. Bremer (ed.) Intermetallische Phasen als Strukturwerkstoffe fur hohe Temperaturen. Forschungszentrum Julich GmbH, Jiilich (1991) 1-13. (lo] G. Suuthoffi Z . Metallkde. 81 (1990) 855. Ill] High-Temperature Ordered Intermetallic Alloys V (MRS Symp. Proc. Vol. 288). Materials Research Society, Pittsburgh, (1993). [12] Structural Intermetallics. TMS, WarrendalelPA, (1993). [13] Ordered Intermetallics - Physical Metallurgy and Mechanical Behaviour. Kluwer Acad. Publ. Dordrecht, (1992). [14] D. Banerjee:Ti,Al and its Alloys. In: J. H. Westbrook and R. L. Fleischer (eds.) Intermetallic Compounds: Principles and Practice, Vol. 2. John Wiley & Sons, Chichester (1995) 91-132. [15]H. A . Lipsitt: Mat. Res. SOC.Symp. Proc.288 (1993) 119. [16] H. A. Lipsitt: Titanium Aluminides - An Overview. In: C. C. Koch, C. T. Liu, and N. S. Stoloff (eds.) High-Temperature Ordered Intermetallic Alloys. MRS, Pittsburgh (1985) 351-364. [17] C. M.Ward-Close,E H. Froes:J. Metals 46 (1994) 28. [18] f? H. Froes, C. Suryanarayana,D. Eliezer: J. MateT. Sci. 27 (1992) 5113. [19] E H. Froes, C. Suryanarayana,D. Eliezer: ISIJ Intl. 31 (1991) 1235. [20] G. Proske, G. Liitjering,J.Albrecht, D. Helm, M . Daeubler: Mater. Sci. Eng. A152 (1992) 310. [21] Y - W Kim: J. Metals 47 (1995) 39. [22] Y - W Kim: J. Metals 46 (1994) 30. [23] S. C. Huang, J. C. Chesnutt: Gamma TiAl and Its Alloys. In: J. H. Westbrook and R. L. Fleischer (eds.) Intermetallic Compounds: Principles and Practice, Vol. 2. John Wiley & Sons, Chichester (1994) 73-90. [24] D. M. Dimiduk, D. B. Miracle, Y - W Kim, M . G. Mendiratta:ISIJ Intl.31 (1991) 1223. [25] Y Nishiyama, Z Miyushita, S. Zsobe, Z Noda: Development of Titanium Aluminide TurboCharger Rotors. In: s.H. Whang, C. T. Liu, D. F! Pope et al. High-Temperature Aluminides and IntermetallicsTMS, Warrendale (1990) 557-584.
10
G. Sairthoff
[26] M . Yamaguchi, H. Inui: TiAl Compounds for Structural Applications. In: R. Darolia. J. J. Lewandowski, C. T. Liu et al. Structural Intermetallics. TMS, Warrendale/PA (1993) 127-142. [27] H. Clemens: Z . Metallkde. 86 (1995) 814. [28] C. Koeppe, A . Bartels, H. Clemens, I? Schretter, W Glatz:Mater. Sci. Eng.,A201 (1995) 182. (291 R. Wagner, E Appel, B. Dogan, I? J. Ennis, U. Loreni, J. Miillauer, H. I? Nicolai, W Qiradakkers, L. Singheiser, W Stnarsly, W Vaidya, K. Wirrzwallner: Investment Casting of Gamma TiAl Based Alloys: Microstructure and Data Base for GasTurbine Applications. 1n:Y.-W. Kim, R. Wagner, and M. Yamaguchi (eds.) Gamma Titanium Aluminides (ISGTA ‘95).TMS, Warrendale/PA (1995) 387-404. [30] R. Wagner: Intermetallische Gamma-Titan-Aluminide - Von den Grundlagen zur Bauteilanwendung -. BMFT-Symposium Materialforschung Neue Werkstoffe. KFA-PLR, Jiilich (1994) 753-756. (311 M . Nazmy, M. Staubli: Scripta Metall. Mater. 31 (1994) 829. [32] A. Bartels, C. Koeppe, K. Wurzwallner, I? Schretter, H. Clemens: Microstructure and Mechanical Properties of TiAI Alloys after Thermomechanical Processing. In: H. Bildstein and R. Eck (eds.) Plansee Proceedings - Proc. 3‘h Interntl. Plansee Seminar. Vol. 3. Plansee Metall AG, Reutte (1993) 564-577. [33] H. Clemens, I? Schretter, K. Wurzwallner, A . Bariels, C. Koeppe: Forging and Rolling of Ti48A12Cr on Industrial Scale. In: R. Darolia, J. J. Lewandowski, C.T. Liu et al. Structural Intermetallics.TMS, WarrendalelPA (1993) 205-214. [34] (1. Herold-Schmidt, H . Kiihnle, S.Schwantes: Oxidation and Versprodung von TitanaluminidBlechen. In: H. Bildstein and R. Eck (eds.) Plansee Proceedings - Proc. 3Ih Interntl. Plansee Seminar. Vol. 3. Plansee Metall AG, Reutte (1993) 607421. [35] I? A. Beaven, EAppel, B. Dogan, R. Wagner:Fracture and Ductilization of Gamma-Titanium Aluminides. In: C. T. Liu, R. W. Cahn, and G.Sauthoff (eds.) Ordered Intermetallics - Physical Metallurgy and Mechanical Behaviour. Kluwer Acad. Publ. Dordrecht (1992) 413432. [36] G. Frommeyer, W Wunderlich, Z Kremser, Z . C. Liu:Mater. Sci. Eng. A152 (1992) 166. [37] D. L. Anton: Ni,AI in Nickel-Based Superalloys In: J. H. Westbrook and R. L. Fleischer (eds.) Intermetallic Compounds: Principles and Practice, Vol. 2. John Wiley & Sons, Chichester (1995) 3-16. [38] C. Z Liu, D. I? Pope: Ni,AI and its Alloys. In: J. H. Westbrook and R. L. Fleischer (eds.) Intermetallic Compounds: Principles and Practice,Vol. 2. J. Wiley, Chichester (1995) 17-52. [39] A! S. Stofoff:Intl. Metals Rev.29 (1984) 123. [40]C.Z Liu, J. 0.Stiegler, E H. Froes: Ordered Intermetallics. Metals Handbook Vol. 2: Properties and Selection: Non-Ferrous Alloys and Special Purpose Materials. ASM, Materials Park (1990) 913-942. [41] J. W Panen: Nickel Aluminides for Diesel Engines. In: S. H. Whang, C. T. Liu, D. P. Pope et al. High-Temperature Aluminides and Intermetallics. TMS, Warrendale (1990) 493-503. [42],D. M. Dimiduk, D. B. Miracle, C. H. Ward:Mater. Sci.Technol.8 (1992) 367. [43] U.Brill, J. Klower: High Temperature Corrosion of Intermetallic Phases Based on Ni,AI. In: L. A. Johnson, D. I? Pope, and J. 0. Stiegler (eds.) High Temperature Ordered Intermetallic AUoys IV. MRS, Pittsburgh (1991) 963-968. [44] D. B. Miracle, R. Darolia: NiAl and its Alloys. In: J. H. Westbrook and R. L. Fleischer (eds.) Intermetallic Compounds: Principles and Practice, Vol. 2. John Wiley & Sons, Chichester (1995) 53-72. [45] R. Darolia: NiAl for Turbine Airfoil Applications. In: R. Darolia. J. J. Lewandowski, C. 7:Liu et al. Structural Interrnetallics. TMS, Warrendale/PA (1993) 495-504. [46]R. Darolia: J. Metals 43(3) (1991) 44. [47] G. H. Meier, A! Birks, E S. Petiit, R. A . Perkins, H. J. Crabke: Environmental Behavior of Intermetallic Materials. In:R. Darolia, J. J. Lewandowski, C.T. Liu et al. Structural Intermetallics. TMS, Warrendale/PA (1993) 861-877. [48] R. Kainuma, K. Ishida, Z Nishizawa: Metall.Trans. 23A (1992) 1147. [49] R. Kainuma, H. Nakano, K. Oikawa, K. Ishida, 7:Nishizawa: High Temperature Shape Memory Alloys of Ni-AI Base Systems. In: C.T. Liu, M. Wuttig, K. Otsuka et al. Shape-Memory Materials and Phenomena - Fundamental Aspects and Applications MRS, Pittsburgh (1992) 403-408. [50] S. Furukawa, A . Inoue, Z Masumoto: Mater. Sci. Eng. 98 (1988) 515. ~
I Slate of Inrernietallics Development
11
[51] R. Daro1ia:Acta Met. Sin. (Engl. Lett.) 8 (1995) 625. [52] 1. Baker: Mater. Sci. Eng. A193 (1995) 1. [53] D.B. Mirac1e:Acta Metall. Mater. 41 (1993) 649. [54] R. D. Noebe, R. R. Bowman, M . k! Nathal: Intl. Mater. Rev.38 (1993) 193. [55] I. Baker, P R. Munroe: Properties of B2 Compounds. In: S. H. Whang, C. T. Liu, D. P. Pope et al. High-Temperature Aluminides and Intermetallics. TMS, Warrendale (1990) 425-452. [56] G. Sauthofi Neue Strukturwerkstoffe auf der Basis intermetallischer Phasen. 2. Symposium Materialforschung des BMFT. KFA-PLR, Jiilich (1991) 877-898. [57] G. Sauthoff: Intermetallische Phasen. Symposium Materialforschung 1988. KFA-PLR, Jiilich (1958) 399-414. [58] M.Rudy, I. Jung, G. Saurhoffi Ferritic Fe-Ni-A1 Alloys for High Temperature Applications. In: J.B. Marriott, M. Merz, J. Nihoul et al. High Temperature Alloys -Their Exploitable Potential. Elsevier Appl. Sci., London (1987) 29-37. [59] M. Rudy, G. Sauthofl: Creep Behaviour of the Ordered Intermetallic (Fe,Ni)Al Phase. In: C. C. Koch, C.T. Liu, and N. S. Stoloff (eds.) High-Temperature Ordered Intermetallic Alloys. MRS, Pittsburgh (1985) 327-333. [60] W Kowalski: Mikrostruktur und mechanische Eigenschaften intermetallischer NiAl-Cr-Legierungen. VDI Verlag, Diisseldorf, (1994). [61] W Kleinekathofer, A. Donne6 H. Meinhardt, M. Hengerer, G. Sauthofif B. Zeumer, G. Frommeyer, H. J. Schiifer: Entwicklung von intermetallischen NiAl-Basislegierungen fur Motorenkomponenten. BMFT-Symposium Materialforschung - Neue Werkstoffe. KFAPLR,Jiilich (1994) 1014-1015. [62] E. Arzt, P Grahle: Mat. R e s SOC.Symp. Proc.364 (1995) 525. [63] E. Arzr, E. Gohring, P Grahle: Mat. Res. SOC.Symp. Proc. 288 (1993) 861. [64]G. Sauthofj W Kleinekathofer: Hochstwarmfeste NiAl-Basis-Legierungen fur Strukturbauteile im Verbrennungsmotor. In: Effizienzsteigerung durch innovative Werkstofftechnik. VDI Verlag, Dusseldorf (1995) 647-654. [65] B. Zeumer, W.Wunnike-Sanders, G. Sauthoffi Mater. Sci. Eng. A 192/193 (1995) 817. [66] G. Saurhoff: High Temperature Deformation and Creep Behaviour of BCC Based Intermetallics. In: 0.Izumi (ed.) Proceedings of the International Symposium on Intermetallic Compounds - Structure and Mechanical Properties - (JIMIS-6). The Japan Institute of Metals, Sendai (1991) 371-378. [67] G. Sauthoffi Mechanical Properties of Intermetallics at High Temperatures. In: S. H. Whang, C. T. Liu, D. P. Pope et al. High-Temperature Aluminides and Intermetallics.TMS, Warrendale (1990) 329-352. [68] J. Klower, G. SauthoB D. Letzig:Alloy 10 A1 - A New Sulphidation and Carburization Resistant Alloy for Fuel Combustion and Conversion. In: Corrosion 96. NACE International, HoustorvTexas (1996) 144/1-144/15. [69] D. Letrig: Zur Entwicklung kaltumformbarer NiAl-Basis-Legierungen. Dr. rer. nat. Dissertation, RWTH Aachen (1995). [70] T. Yamamoto:On the Discovery of the High Permeability Alloy “Sendust” and the Progress of its Industrialization. In: T. Yamamoto (ed.) The Development of Sendust and Other Ferromagnetic Alloys. Committee of Academic Achievements, Chiba (1980) 1-6. [71] S. C. Deevi, I/:K. Sikka: Intermetallics 4 (1996) 357. [72] K K. Sikka: Processing and Applications of Iron Aluminides In: J. H. Schneibel and M. A. Crimp (eds.) Processing, Properties, and Applications of Iron Aluminides. TMS, Warrendale (1994) 3-18. [73] k! K . Sikka, S.Vkwanathan, C. G. McKamey: Development and Commercialization Status of Fe,Al-Based Intermetallic Alloys. In: R. Darolia, J. J. Lewandowski, C. T. Liu et al. Structural Intermetallics. TMS, Warrendale/PA (1993) 483-491. [74] K. Vedula:FeAl and Fe,AI. In: J. H. Westbrook and R. L. Fleischer (eds.) Intermetallic Compounds: Principles and Practice,Vol. 2. John Wiley & Sons, Chichester (1995) 199-210. [75] H. Rosner, E. Nembach: Mater. Sci. Eng. A 196 (1995) L 1. [76] W Schroer, C. Hartig, H. Mecking: Z. Metallkde. 84 (1993) 294. [77] J. Klower: Stand der Entwicklung intermetallischer Phasen auf Basiss der Eisenaluminide Eigenschaften und Herstellung (VDM - internal report Nr. 10/92(1992)).
12
G. Sauthoff
[78] 1. Baker, l? Nagpal: A Review of the Flow and Fracture of FeAI. In: R. Darolia. J. J. Lewandowski, C.T. Liu et al. Structural IntermetallicsTMS,WarrendalelPA (1993) 463-473. [79] C. T. Liu. K. S. Kumar: J. Metals 45 (5) (1993) 38. (801K. S. Kutnnr: ISIJ Intl. 31 (1991) 1249. [81] K. Vedula:Strength, Ductility and Toughness of Intermetallic Matrix Composites. In: 0.Izumi (ed.) Proc. Int. Symp. Intermetallic Compounds - Structure and Mechanical Properties (JIMIS-6).The Japan Institute of Metals, Sendai (1991) 901-926. [82] R. D. Grinthnl: WADCTechnical Rep.53-190 (1958) 1. [83] WArbiter:WADCTechnical Rep.53-190 (1953) 1. [84] K. C. Chen, S. M . Alleti, J. D. Livingston: Mat. Res. SOC.Symp. Proc. 364 (1995) 1401. [85] R. L. Fleischer. R. J. Znbala: Metall.Trans.2lA (1990) 2149. [86] D. L. Antoti, D. M . Shah: Mat. Res. Soc. Symp. Proc. 288 (1993) 141. [87] D. L. Anton, D. M . Shah: Ternary Alloying of Refractory Intermetallics. In: L. A. Johnson. D. P. Pope, and J. 0.Stiegler (eds.) High Temperature Ordered Intermetallic Alloys IV.MRS. Pittsburgh (1991) 6348. 1881 M. Takeyamn, C. T Liu: Mater. Sci. Eng. A132 (1991) 61. 1891 E. Firzer: Hochsttemperaturbestandige Werkstoffe durch Silizieren von Wolfrani und Molybdan. In: E Benesovsky (ed.) Pulvermetallurgie (1. Plansee-Seminar. Reutte, Tirol. 1952). Komm.-Verlag Springer, Wien (1953) 244-258. [90] l? Schwarzkopi R. Kieffer: Refractory Hard Metals. Macmillan, New York. (1953). [91] R. Lowrie:Trans.AIME 194 (1952) 1093. [92] K. S. Kumar: Silicides: Science, Technology and Applications. In: J. H.Westbrook and R. L. Fleischer (eds) Intermetallic Compounds: Principles and Practice,Vol. 2. John Wiley & Sons. Chichester (1995) 211-236. [93] J. J. fetrovic: Mater. Sci. Eng. A 193 (1995) 31. [94] D. A. Hardwick, l? L. Martin, S. N. Patankar, J. J. Lewandowski: Processing Mircostructure Property Relationships in Polycrystalline MoSi,. In: R. Darolia. J. J. Lewandowski. C.T. Liu et al. Structural Intermetallics.TMS, WarrendalelPA (1993) 665-674. [95] 11 fetrowic: MRS Bulletin 18(7) (1993) 35. (961 M.R. Jackson, B. E Bewlay, R. G. Rowe, D. W Skelly, H. A. Lipsitt: J. Metals 18(1996) 3Y. (971 I! R. Subramanian, M. G. Mendiratta, D. M . Dimiduk: J. Metals 48 (1996)33. (981 M. Nazmy, C. Noseda, G. Sauthoff; B. Zeumer, D. Anton: Mat. Res. SOC.Symp. Proc. 364 (1995) 1333. [99] S. Mazdiyasni, D. B. Miracle: Survey of Eutectic Systems as Potential Intermetallic Matrix Composites for HighTemperature Application. In: D. L.Anton, P. L. Martin. D. B. Miracle et al. Intermetallic Matrix Composites MRS. Pittsburgh (1990) 155-162. [ 1001 G. Frommeyer, R. Rosenkranz, C. Liidecke: Z . Metallkde. 81 (1990) 307. [ 1011 P A . Beaven, 1 S. Wu, B. Dogan, C. Hartig, J. Seeger, R. Wagner:GKSS-Jahresbericht (1989) 49. [lo21 M. Es-Souni,R. Wagner,I! A. Benven, F-I! Schimansky. R. Gerling: Scripta Metall. Mater. 26 (1992) 1845. [lo31 J. S. Wu, D. Chen, l? A. Beaven, R. Wagner, J. Seeger: Microstructure and Properties of Two Phase Alloys Based on (Ti,Nb),(AI,Si) and (li,Nb)s(Si.Al)3. In: M. Kong and L. Huang (eds.) Structural Materials (Proc. C-MRS International ‘90. Beijing, Vol. 2). Elsevier Sci. Publ. Amsterdam (1991) [lo41 R. Rosenkranz, G. Frommeyer, W Smnrsly: Mater. Sci. Eng. A152 (1992) 288. [lo51 F Jansen, E. Ltcgscheider: Neue Struktur- und Beschichtungswerkstoffe auf MoSi,-Basis fur Hochtemperaturanwendungen. In: Efliienzsteigerung durch innovative Werkstofftechnik. VDI-Verlag, Dusseldorf (1995) 147-1 50. [lo61 R. Rosenkranz, G. Frommeyer: Z . Metallkde. 83 (1992) 685. [lo71 E. Ludetischeider, U. Westermann, J. Wonka, H. Meinhnrdt, H. Neisiris, R. Arnold: Investigations on Molybdenum and Titanium Disilicide as Structural Materials for Highest Temperatures. In: 0. Izumi (ed.) Proc. Int. Symp. Intermetallic Compounds - Structure and Mechanical Properties (JIMIS-6).The Japan Institute of Metals,Sendai (1991) 621425. [lo81 S. Beer, G. Frommeyer, E. Schmid, H. He1big:VDI-Berichte 1080 (1994) 89. [lo91 E. E. Schmid, K. von Oldenburg, G. Frommeyer: Z . Metallkde. Sf (1990) 809. [110] K. von Oldenburg, G. Frommeyer, E. Schmid, W Henning: Microstructures and Mechanical Properties of As Cast and Mechanically Alloyed Mg,Si-A1 Alloys. In: T. Khan and G. Effen-
1 State of Intermetallics Development
13
berg (eds.) Advanced Aluminium and Magnesium Alloys. ASM International, Materials Park (1990) 477-484. [ill] G. Sauthoff: Creep Behaviour and Creep Mechanisms in Ordered Intermetallics. In: C.T. Liu, R. W. Cahn, and G. Sauthoff (eds.) Ordered Intermetallics - Physical Metallurgy and Mechanical Behaviour. Kluwer Acad. Publ. Dordrecht (1992) 525-539. [112] C.-I! Reip: Untersuchung des Verformungsverhaltens der DO,-geordneten intermetallischen Phase A1,Nb. Dr. rer. nat. Dissertation, RWTH Aachen (1991). [113] C I ! Reip, G. Sauthoffi Intermetallics 1 (1993) 159. [1141 L. Machon: Untersuchung des Verformungsverhaltens hochschmelzender hexagonaler Laves-Phasen. Dr. rer. nat. thesis. RWTH Aachen, Aachen, (1992). [115] Inco pamphlet (1982). [116] E Porz, G. Grathwohl: KfK-Nachr. I 6 (1984) 94.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
2 Research on Oxidation and Embrittlement of Intermetallic Compounds in the U.S. G. H. Meier
2.1 Introduction The general aspects of the oxidation of intermetallics are contained in a number of reviews [l-S].lhe purpose of the present paper is to describe current work on this subject being done in the U.S.,for the systems the author feels are significant, and to indicate how the results of this work enhance our understanding. Clearly, previous work in the US., Europe, and Japan has laid the foundations for the current work and, where appropriate, will be described. The fundamentals of intermetallic oxidation will be briefly described followed by discussion of work on the Ni- and Fe aluminides, Ti aluminides, and refractory metal compounds. It will become clear that work is now focussed in two general areas: (i) adding to the fundamental knowledge of oxidation mechanisms and (ii) developing approaches to allow intermetallics to reach the application stage e. g. coating, composite development, and life predictions.
2.2 Background Important features of the selective oxidation process are shown schematically in Figure 1.The slow growth rates of alumina and silica, illustrated in the plot of parabolic rate constants versus temperature at lower right, makes the formation of one of these oxides as a continuous surface layer necessary for long term oxidation protection. This requires that the protective oxide be more stable thermodynamically than the more rapidly growing oxides. The plot of standard free energy of formation as a function of temperature at lower left shows that the Ni-AI system satisfies this condition. Alumina is stable, relative to NiO, even when the activity of aluminum in the alloy is very low. However, when the A1 concentration is low the alumina forms as internal oxide precipitates and is non-protective allowing an external layer of NiO to form (illustrated in the cartoon at top). Therefore, a critical concentration of Al exists above which out-
16
G. H. Meier
Fig. 1. Important aspects of the selective oxidation process: Relative magnitudes of free energies of formation of various oxides, relative magnitudes of the parabolic rate constant for the growth of various oxides, and schematic diagrams showing internal formation of alumina in a dilute NiAl alloy and external alumina on a concentrated alloy such as NiAI.
ward Al diffusion creates a continuous, protective alumina layer. The oxidation morphology also contains transient Ni oxides which form before the alumina becomes continuous. The selective oxidation process is somewhat different for many of the intermetallics of interest. Figure 1 indicates that Ni-, Fe-, and Mo-base systems generally satisfy the thermodynamic condition that alumina or silica is the most stable oxide in the system. However, this is not the case for Nb- and Ti-base systems where the base-metal oxides are of comparable stability to alumina and silica. In these cases the activity of A1 or Si must be maintained at a high level in the alloy to maintain stability of the protective oxide. Furthermore, the potential transient oxides, e. g. TiO,, have much higher growth rates than NiO which makes development of a continuous alumina o r silica layer difficult even when the thermodynamic conditions are clearly favorable. This is particularly the case for the refractory metals Mo, W, Nb, and Ta whose oxides d o not exhibit parabolic kinetics and grow rapidly inward with linear kinetics such that the nuclei of the protective oxide are removed before they can grow together to form a continuous layer. A n additional complication arises because some intermetallic compounds, e. g. MoSi, and NbAI,, have narrow ranges of stoichometry so that a new phase forms beneath the oxide in the zone which is depleted of A1 or Si by formation of the external layer. The newly-formed phase then influences the subsequent stability of the protective oxide as well as its adherence.
2 Research on Oxidation und Embritilement of Intermetallic Compounds h the U S .
17
T
4
3
Fig. 2. Schematic diagrams showing the cracking and spalling of an oxide layer, which often occurs on cooling from the oxidation temperature and comparing the isothermal and cyclic oxidation kinetics.
In order for an alumina or silica layer to remain protective it must remain adherent to the substrate. The spalling of a protective oxide during thermal cycling and the resultant oxidation kinetics are illustrated schematically in Figure 2. The damage to the oxide is generally the result of the thermal stresses arising during cooling because of thermal expansion mismatch between the metal and oxide. Table 1 shows the linear coefficients of thermal expansion for selected materials [&lo]. The data indicate large thermal expansion mismatches between alumina and the aluminides of Ni and Fe whereas the mismatch between alumina and TiAl is small. These data arc also pertinent to the behavior of composites. For example, Sic, a commonly used strengthening phase, has large mismatch with potential matrices such as MoSi, and Ti,AI. The major interest in the intermetallics is for developing systems with higher specific strengths at high temperature. Thus the density is an important property which is the most attractive feature of the titanium aluminides which have densities less than half that of the Ni-base superalloys (Table 1).Additionally important are the ductility and fracture toughness of the internietallics.The latter is seen in Figure 3 to be inferior to that of a typical Ni-base superalloy and of the same magnitude as that of grey cast iron. The mechanical properties of intermetallics are generally further degraded, sometimes catastrophically, by exposure to oxidizing environments.
2.3 Oxidation of Selected Compounds 2.3.1 Nickel and Iron Aluminides The binary phase diagram for the Ni-A1 system is presented in Figure 4. Superposed on this diagram are the oxidation data of Pettit [ll].The crosshatched region indicates
18
G. H. Meier
Table 1. Selected physical property data CTE X106 (“C-’)
Material FeAl Fe,Al NIAl Ni,AI MarM-246 All03 TiAl Ti,AI TiAI, Sic XB, MoSi, Mo Nb
16.5 16.5 15 12.5 16 9 11 9 13 5.5 7.8 8.1 5 7.3
Density (g/cm3)
T, (“C)
-
5.56 6.70 5.86 7.65 8.44
1337 1540 1647 1390 1317
3.91 4.20 3.30
1462 1602 1352
6.30
2030
~
Fracture Toughness of Selected lntermetallics
1
IN738LC
Fig.3. Comparison of the room temperature fracture toughness of selected intermetallic compound and that for a nickel-base supcralloy and cast iron.
that NiAl should form and maintain protective alumina under all conditions while Ni,Al is a “marginal” alumina former at temperatures below 1200°C. The most successful structura! application of any intermetallic has clearly been Ni,AI as the major phase in modern Ni-base superalloys. Research on the oxidation behavior of superalloys is continuing with major emphasis on the adherence of the alumina scales.The influence of reactive elements (Y,Hf) in improving adherence and of impurities (S) in degrading it suggest that further improvement in the oxidation resistance of superalloys is still possible. Figure 5 shows the effect of reducing the sulfur content of two single crystal alloys from about IOppmw to less than 1 ppmw [12].’fie reduction of the sulfur content has clearly improved the cyclic oxidation resistance, particularly for PWA 1484 which contains 0.1 wt% Hf. Such data indicate the distance
2 Research on Oxidation and Enibrifllement of lntertnetnllic Conipoirnds in the U.S.
19
1800
F - l6O0 1400 s!
3 2
1200
g 1000
I-
800
600
Ni 10
20
30
40
Al
60 70
50
(at
80
90
Al
Oh)
Fig.4. Binary phase diagram for the Ni-A1 system showing the compositions which form protective external alumina scales, after Doychuk [5). 10 cu
B
-
w
LOW
Low s 1480
I 0
E
s 1484
0
a
gI-
I
-10
-
-20
Fig.5. Plot of mass change versus time indicating the effect of sulfur content on the cyclic oxidation kinetics of two single-crystal, nickel-base superalloys at 1100°C[12].
the monolithic intermetallics must travel to obtain oxidation resistance sufficient to make them competitive as alternative materials to the superalloys. The areas concerning monolithic intermetallics which have been studied in recent years are (i) the formation of metastable aluminas, and their transformation to stable a-alumina, (ii) the formation of interfacial voids and scale adherence and how these are influenced by reactive elements and sulfur, and (iii) accelerated oxidation at intermediate temperatures. Additionally the applications oriented areas of (iv) coatings, (v) oxidation of composites, and (vi) life predictions have received attention.
2.3.1.1 Formation of a-Alumina from Transient Aluminas This subject has been reviewed extensively by Doychak [5] and will only be treated briefly here, primarily, with regard to recent work.'l%e oxide scales formed on Ni,Al at
20
G. H. Meier
oxygen partial pressures on the order of 1 atm in the temperature range 950-1200”C consist mainly of Ni-containing transient oxides, NiO and NiAI2O4,over a layer of columnar a-alumina. Schumann [13] has recently studied the very early stages of transient oxidation of the (001) faces of Ni,AI single crystals in air at 950°C using crcsssection TEM. After 1min. oxidation simultaneous formation of an external NiO scale and internal oxidation of y’ were observed. The internal oxide particles were identified as y-Al,03, which possesses a cube-on-cube orientation with respect to the Ni matrix. After 6 min. oxidation a continuous y-Al,O, had formed between the internal oxidation zone (IOZ) and the y’ single crystal. Oxidation for 30 min. resulted in a microstructure similar to the 6min. oxidation but the Ni in the two-phase zone was oxidized to NiO. Oxidation for 50 h resulted in a scale consisting of an outer layer of NiO, an intermediate layer of NiAI,O,, and an inner layer of y-Al,O, in which a-A1203grains had nucleated at the alloy/oxide interface.The spinel was presumed to have formed by a solid state reaction between NiO and y-Al,O,. A crystallographic orientation relationship was found between the a and y alumina whereby (OOOl)[lTOO], I1 (11I)[lTO],. i. e. close packed planes and close packed directions of c1 arc parallel to close packed planes and directions in y. Pint and Hobbs [14] studied the oxidation of yttria (2~01%)-dispersed Ni,AI at 1000 and 1200°C and did not report transient alumina formation for oxidation times as short as 1hour. The difference in the results of [13] and [14] may stem from slightly different oxidation temperatures or stabilization of the y-Al,O, on the specific orientation or the single crystal because of surface energy considerations. The transient oxidation of Fe,Al has not been extensively studied, however, the formation of 8-A120, has been invoked to explain anomalies in the oxidation rate at 900°C [ 151. The oxidation of NiAl is somewhat unique in that, at temperatures of 1000°C and above, there are negligible amounts of Ni-containing transient oxides. The transient oxides are all metastable phases of AI,O, (y, 6, and/or 0) [ S ] . The transition of these metastable phases to the stable a-Al,O, results in significant decreases in the scale growth rate and a “ridged” oxide morphology which is distinct from the columnar morphology observed on Ni,AI [5].The transition aluminas have been shown to grow primarily by outward migration of cations while a grows primarily by inward transport of oxygen [5].The effect of oxidation time and temperature on the phases present in the scales has been studied by several authors. Rybicki and Smialek [16] identified t) as the transient oxide on Zr-doped NiAl and found that the transition to a occurred at longer times at lower oxidation temperatures e. g. scales consisted entirely of 8 after 100hours at 800°C while it transformed to a in about 8hours at 1000°C. Only a was observed at 1100°C and 1200°C.Pint and Hobbs [17] observed only a on undoped NiA1 after 160 seconds at 1500°C. Brumm and Grabke [IS] observed two transforniations at 900°C for undoped NiA1.The scale consisted initially of y which transformed to 8 after approximately 10hoursThe transformation of 8 to a occurred at much longer times but accelerated as the oxidation temperature was increased. The transition to a on NiAl has been reported to initiate at the scale/gas interface 1191 which is in contrast to the observations on Ni,AI where it is reported to initiate at the scale/alloy interface [13].The scale transformations have also been found to be sensitive to the presence of third elements in the alloy.Additions of Cr accelerate the transformation to a [18] by a proposed mechanism involving transient Cr,O,, which is isostructural with a , providing nucleation sites. This results in a finer-grained a which grows somewhat faster
2 Research on Oxidolion nnd Etnhrittlenienr of Inlermerallic Coniportnds in the U.S.
21
than the a formed on binary NiA1.A study of oxide dispersed NiAl found that the 0 to transformation was slowed by Y, Zr, La, and Hf and accelerated by Ti [20]. The proposed explanation of these results is that the transformation is slowed by large ions which can enter the more open lattices of the transition aluminas.This result is in contrast to the observation that no transition aluminas were observed on Y,O,-dispersed Ni,A1 [ 141. These phenomena have not been well studied for FeAl but limited data suggest a similar set of transformations in the aluminas formed on FeAJ [21]. (Y
2.3.1.2 The Formation of Interfacial Voids and Scale Adherence and how these are Influenced by Reactive Elements and Sulfur The integrity of the alloy/oxide interface and the adherence of the alumina scale to the alloy are critical issues for the application of aluminides or aluminide c0atings.A common feature of the oxidation of the Ni- and Fe-aluminides is the formation of voids at the alloy/oxide interface.'I'his is illustrated, schematically, in Figure 6 for NiAl[3]. The voids result from the diffusion of Ni into the alloy, as A1 is consumed to grow the oxide, which brings vacancies to the interface where they condense. Brumm and Grabke [22] have shown that void formation is negligible on Al-rich NiAl where A1 diffusion to the interface predominates as compared to Ni-rich NiAl where Ni-diffusion away from the interface predominates. The presence of the interface voids has long been known to be a source of poor adherence [23].'I'he interface voids are greatly suppressed by additions of reactive elements [24]. This generally results in improved alumina adherence and better cyclic oxidation resistance. The effect of reactive element additions as oxide dispersions on the cyclic oxidation resistance of NiAl at 1200°C is illus-
Fig. 6. Schematic diagrams demonstrating the fluxes of A1 and Ni caused by oxidation of NiAI, leading to vacancy formation and condensation as voids beneath the oxide scale.
G. H.Meier
22 7 ,
0.1 IZ r
I I
0
-10
,
0
Ref.: B.Pint and L. Hobbs. M R S Symp. Proc.. Vol. 364. 1995 I
I
2
4
I
6
I
I
8 10 12 Cycles (2 hwdcycle)
14
I
16
1
18
1
Fig.7. Cyclic oxidation kinetics of NiAI, containing various oxide dispersions, in air at 1200°C [XI.
Mass Change/Area (rng/cm2) ......... ..
-...a-,.-*.y,--
-
_-0
200
400
600
t (hours)
800
lo00
1200
Fig.8. Effect of sulfur content on the cyclic oxidation behavior of NiAI.The as-received material contained 20ppmw S and the sulfur content of the hydrogen annealed material was calculated to be less than 1ppmw.
trated in Figure 7 [25].Lowering the sulfur content of NiAl also suppresses interface void formation [26] and, as illustrated in Figure 8, can greatly improve scale adherence. The addition of reactive elements has been shown to produce similar improvements in the scale adherence to Ni,AI [14],Fe,Al[15], and FeAl [21].
2 Research on Oxidation and Einhrittlernent of Internietallic Compounds in the U.S.
23
A number of studies have introduced the reactive element into the aluminides by ion implantation. The author considers this approach to be of limited value since, as discussed by Pint and IIobOs [27], the high local concentration of reactive element can stabilize the transition aluminas and, at high temperatures, has a short-lived effect. More importantly. from a fundamental standpoint, the implantation process can have a profound effect on the nature of the exposed surface. For example, Schumann [28] has shown that Y implantation into single-crystal NiAl results in a 45 nm thick, finegrained crystalline region which is disordered.
2.3.1.3 Accelerated Oxidation at Intermediate Temperatures A number of intermetallic compounds, which form protective alumina or silica scales at high temperature, undergo accelerated degradation at intermediate temperatures. This subject has been recently reviewed [29].The observations actually involve several different, but related, phenomena which may be subdivided into “accelerated oxidation”, “internal oxidation”, “intergranular oxidation” and “disintegration”. The following definitions will be used throughout this paper.
Protective Oxidation - formation of a continuous alumina or silica surface film with no internal oxidation and minimal penetration of oxygen into the substrate. The overall oxidation rate is determined by transport through this film. Accelereated Oxidation - the alumina or silica is not continuous and significant amounts of the other component(s) of the intermetallic are present in the surface film. The overall oxidation rate is substantially faster than that for the growth of alumina or silica. Internal Oxidation - precipitation of oxides rich in A1 or Si within the intermetallic. Intergranular Oxidation - special case of internal oxidation in which oxides form along grain boundaries within the intermetallic. Pesting - disintegration (fracture) of the intermetallic into smaller particles at the oxidation temperature. The occurrence of protective oxidation precludes the other phenomena from occurring as long as the oxide is not damaged. However, in the absence of protective oxidation, any or all of the other phenomena may occur together. In the case of NiAl this degradation occurs in the temperature range 700 to 1000°C and can take the form of intergranular oxidation at reduced oxygen partial pressures in the range lo-” to 10-’atm. and internal oxidation in the oxygen pressure range to 10 22atm.[30].A particularly complex form of degradation occurs when exposures are carried out in evacuated silica ampoules containing buffer mixtures, such as Cu/Cu,O. This attack, termed “pocks” [30] is illustrated in Figure 9 [31].The oxidation morphology consists of an outer zone of silicides and silicates (Si is the result of SiO vapor transport from the capsule walls) over an internal oxidation zone of alumina in virtually pure Ni. A substantial Al-depleted zone, which has transformed to Ni,Al, is observed beneath the internal oxidation zone. The detailed mechanism for this process is not completely understood but is known to require conditions that prevent the formation of a protective alumina layer (temperatures which favor formation of transition aluminas and possi-
24
G. H. Meier
Porous Al-oxides
Fig.9. Example of pock formation of NiAl oxidized at llOOK in a silica ampoule containing a mixture of Cu and Cu,O powders.
bly contaminants such as S) and a process which rapidly consumes Al (the internal oxidation). Pocks do not form when the buffer mixture is thoroughly dried prior to specimen exposure [30,31], when the NiAl is alloyed with Cr [30], or for aluminide coatings on Ni-base superalloys [31]. These observations are possibily all the result of conditions which favor the rapid establishment of a-alumina rather than transition aluminas. Similar studies have, apparently, not been performed for FeAI. 2.3.1.4 Coatings The aluminizing of Ni-base alloys via the pack-cementation process has been a conimercially viable process for many years [32]. The typical coating morphologies for the
2 Research on Oxidation and Ernbritllrnirnt of Iniernietnllic Compounds in the U.S.
25
Fig. 10. Schematic diagram showing the as-deposited structurc o f a “low activity” (left) and a “high activity” (right) diffusion aluniinide coating on a nickel-base superalloy. The high activity coating would receive a heat treatment to convert the Ni,AI, to NiAI.
so-called “low activity” and “high activity” processes are illustrated schematically in Figure 10.These coatings oxidize in a similar manner to bulk NiAl except that,superposed on the oxidation process. the coating loses A1 by interdiffusion with the substrate. In recent years the plating of Pt on the substrate prior to aluminizing has been used to produce “platinum-aluminide” coatings [33]. This type of coating, shown schematically in Figure 11, has resulted in improved cyclic oxidation rcsistance and reduced interdiffusion with the substrate. Currently, the improvement in hot corrosion resistance imparted to Ni-AI alloys by Cr additions has resulted in research efforts aimed at codepositing Cr and A1 in aluminide coatings. The simulatenous codeposition of Al and Cr via the pack cemcntation process, using pure elemental powders, is, however, difficult. The large difference in the thermodynamic stabilities of the A1 and Cr halides causes Al-halide species to predominate in thc pack atmosphere [34]. However, by employing binary chrorniumaluminum (Cr-A1) master alloys, the high relative A1 halide vapor pressures can be moderated.This is the result of the fact that chromium-rich master alloys exhibit negative deviations from idcality and the activity of A1 in the master alloy can be reduced by several orders of magnitude. The reduced thermodynamic activity of A1 results in generation of lower vapor pressurcs for the otherwise favored halide species (e. g., AICI, AICI, etc.). Therefore comparable A1 and Cr vapor pressures result. Thus, provided a suitable activator and binary master alloy is chosen, the codeposition of Cr and Al into Ni base materials is possible. Rupp and Bianco [35-371 have used this approach to form coatings containing as much as 13 at% Cr in P-NiAI on Ni-base superalloys and have shown them to have greater hot corrosion resistance at 900°C than aluminide coatings without the Cr modification [37]. Da Costa et al. [38, 391 have achieved Cr concentrations as high as 40 at%, using Cr-rich master alloys and multiple activators (NaCl+NH,CI). Stinner et al. 1401 arc currently studying how the pack
26
G. H. Meier
variables,such as the amount and composition of the source alloy and the activator. a f fect the composition and morphology of Cr-aluminide coatings on Ni-base alloys.
2.3.1.5 Composites The lack of toughness at low temperatures and low creep strength a t high temperatures suffered by many monolithic intermetallics has led to effords to use them in composites to achieve the required toughness and strength. The presence of the second phase not only affects the mechanical properties but also the oxidation behavior, usually, in a detrimental fashion.This has occurred even for NiAI. Doychuk et al. [41] found that incorporation of alumina fibers in a NiAl matrix resulted in oxidation along the fibers, in 1200 and 1300°C exposures, and cracking of the matrix. Perkins [42] found, in air exposures at 800 and llOO°C,that addition of 10~ 01%TiB, to NiAl resulted in an incubation period followed by rapid growth of TiO,. The addition of 20 ~0 1 %TiB, resulted in rapid oxidation from the beginning of the exposure. One strengthening phase which, apparently, does not degrade the oxidation resistance of NiAl is AIN [43].The oxidation of AIN-dispersed NiAl was similar to undoped. monolithic NiAl and, when small amounts of Y,O, were added, showed cyclic oxidation resistance comparable to Zr-doped NiAI.
2.3.1.6 Life Prediction The possibility of using intermetallics as structural materials has increased interest in predicting how long a component will last under a given set of exposure conditions. Nesbitt and coworkers [44-46] have treated this problem for the case of NiA1.The factor which controls life is the loss of A1 in forming alumina on the surface and through spallation during thermal cycling. The procedure involves identifying a failure criterion which can be a critical amount of surface recession, reaching a predetermined minimum Al content in the alloy, or the appearance of spinels in the oxide scale. After selecting the failure criterion the rate of Al loss is needed. This can be obtained by isothermal measurements of the oxide growth rate coupled with measurements of the amount of oxide which spalls on each cycle during cyclic oxidation. However. the measurement of the amount of spalled oxide is difficult, particularly at low temperatures. Therefore, an approximate procedure for estimating the A1 loss, which is good at all but very short times, is to use the linear portion of a cyclic oxidation plot such as that in Figure 2.The time dependence of the A1 content in the specimen can be calculated by solving the diffusion equations in the alloy. A simpler, approximate method may be used for compounds, such as NiAI, which have high interdiffusion coefficients by assuming the A1 content decreases uniformly. Either way the time at which the Al content falls below a certain value can be calculated. If desired the A1 loss can be converted to surface recession using the partial molar volumes for A1 in NiAI. Figure 12 shows the results of predicted life for various intermetallics as a function of specimen thickness using a failure criterion of a critical surface recession of 10% of the original specimen
2 Researcli on Oxidation and Etnbritrlenzenr of Intenneiallic Compounds in rhe U.S.
27
1200 "C NiAl-O.1Zr
0.10
0.15 0.20 Thickness (crn)
0.25
_I I
0.1 0
NiAI-0.1 Zr 1400 OC
0.1 5 0.20 Thickness (cm)
I
0.25
Fig. 12. Predicted oxidation lifetime for various intermetallic compounds as a function of specimen thickness using the time to produce a surface recession equivalcnt to 10 % of the original specimen thickness [45].
thickness [45].The longer lives predicted for NiAl-O.lZr, relative to NiAI, is the result of greatly reduced oxide spallation from the Zr-doped alloy.
2.3.2 Titanium Aluminides Alloys in the Ti-A1 system are of interest for high temperature systems such as aircraft engines because they have low density and maintain strength at high temperature. However, their resistance to oxidation and interstitial embrittlernent is a concern. Those alloys which form alumina scalcs have excellent resistance to surface re-
G. H. Meier
28
cession while those which form titania-rich scales oxidize at much higher rates. 'fie modification of the microstructure immediately beneath the oxide layer, resulting from interstitial dissolution and/or selective removal of one or more alloy component. is of particular concern since it can cause a substantial loss of mechanical properties. The discussion in this section covers (i) the thermodynamics of titanium-aluminide oxidation, (ii) oxidation of specific compounds. (iii) embrittlement of titanium aluminides, (iv) the effect of complex environments on the oxidation of titanium aluminides, (v) coatings on titanium aluminides, and (vi) oxidation of titanium aluniinidematrix composites.
2.3.2.1 Thermodynamics A n important aspect of the oxidation of Ti-aluminides, compared to the aluminides of Ni and Fc, is the small difference in standard free energy of formation between alumina and the oxides of titanium, Fig. 1,which is accentuated by a negative deviation from ideal solution behavior in the Ti-A1 system. The aluniinum activity is much smaller than unity in Ti,AI and TiAI. In fact, combining the activties with standard free energy data for the oxides indicates that 'Ti0 is more stable in contact with the alloy than is AI,O, for atom fractions of A1 much less than about 0.5 [47, 481. Thus, AI,O, is unstable in contact with binary a2 and is only marginally more stable than T i 0 in contact with y-TiAl. Figure 13 presents the Ti-AI phasc diagram with the compositions which have been abserved to form continuous alumina scales indicated by crosshatching.
p
1500
v
5 6 !?
I-
loo0
500
Ti
XAl
Al
Fig. 13. Binary phase diagram for thc Ti-A1 system showing regions (crosshatched) where continuous alumina films have been observed to form.
2 Reserirch on Oxidntion trnci Enibritrlmimt of Inrernietcillic Compounds in the U.S.
29
2.3.2.2 Oxidation of Specific Compounds Oxidation of 5,AI (a2) The oxidation of Ti,AI alloys would not be expected, in light of the above thermodynamic considerations, to form continuous alumina scalcs. Instead they form mixed rutile-alumina scales [49].The oxidation kinetics of Ti,AI between 600 and 950°C are reported to be essentially those expected for rutile growth [SO, 511. These kinetics result from the development of a complex oxide layer which contains continuous paths of TiO, through which rapid transport occurs. Typical oxide scalcs are shown schematically in Figure 14. The oxidation of Ti,Al is generally more rapid in oxygen than in air, as indicated in the 900°C oxidation data in Figure 14 [ 5 2 , 5 3 ] .This appears to be the result of a layer of TIN, which forms at the scaleialloy interface during air exposures, providing a diffusion barrier. Alloying of Ti,A1 with P-stabilizing elements, particularly Nb, at lcvcls less than about 10at% reduces the oxidation rate [53,54,55] as indicated in Figurc 14. Generally more complete nitride layers form on the Nb-containing alloys which accounts for part of the rate reduction. Wallace et al. [53] found a layer of y-'IiAl beneath a layer of TiN on ?'i-24AI-l1Nb exposed in air at temperatures between 700 and 1000°C. How-
Fig.14. Oxidation rates for a2 and a,+Nb alloys in air and oxygen at 900°C and schematic diagrams of the scales formed under thc various conditions.
30
G. H. Meier
ever, the effect of Nb also occurs in pure oxygen which indicates an additional effect. Niobium has been detected in the scales [52,55] leading to the proposal that the Nb ions produce a doping effect which decreases the concentrations of oxide ion vacancies and/or titanium ion interstitials in the rutile lattice [52]. The scales developed on cx2 alloys containing multiple additions of P-stabilizers have been described by Wallace et al. [56] and Schaeffer [57]. An additional aspect of the oxidation of Ti,AI alloys is dissolution of oxygen into the alloy at the scale/alloy interface. The embrittlement associated with this phenomenon can be more damaging to the mechanical properties than the surface recession caused by scale formation in the temperature range where Ti,A1 will likely be used (< 700°C) [58].This subject will be discussed in a separate section. Oxidation of orthorhombic alloys There are relatively few oxidation data available on alloys in the Ti-Al-Nb system with significant volume fractions of the orthorhombic phase. Howevcr, data are becoming available which indicate that the amount of Nb necessary to stabilize the orthorhvmbic phase results in more rapid oxidation than the a2alloys because of the formation of discrete Nb-oxides in the reaction products [59].The effect of N b on the parabolic rate constant for the oxidation of Ti-25 at% A1 at 800°C [52,54] is presented at right in Figure 15. The rate constant decreases with increasing Nb content up to the 5-10% range but increase with further additions. Additions of ? a have been found to reduce the oxidation rate slightly at a given Nb content [59].The rate at 700°C has been reported to decrease with Nb additions up to 10at% [55]. Additionally, alloys with Nb contents typical of the “orthorhombic alloys” undergo breakaway oxidation following an initial parabolic period as seen at left in Figurc 15.This phenomenon which is more prominent at the higher temperatures has been shown, using acoustic emission mea-
Oxidation Kinetics of Ti-22AI-23Nb at 500 900T in Air
-
Variation of Initial Parabolic Oxidation Rate at
800’C in Air of (Ti -25AI) + Nb
4 d C
-10
’ I
OTi-25AI b)
a)
I
21Nb-2Ta
20 OhNb ( a t . l )
I
40
Time (hours)
Fig. 15. Oxidation kinetics for Ti-22A1-23Nb“orthorhombic alloys” in air at temperatures in the range 500-90O0C(left) and the effect of Nb content on the parabolic rate constant for Ti-25 at% A1 alloys at 800°C (right).
2 Research on Oxidation rrnd Etnhrittlenient of Intermetallic Compounds in the U S .
31
surements, to result from oxide cracking as the scale is enriched in Nb-oxides [59]. During oxidation the alloy microstructure is modified beneath the oxide layer causing embritt1ement.This will be discussed in a subsequent section. Oxidation of y-TiAl Choridhitry et al., in a classic paper, [50]studied the oxidation ofTiAl (50at% Al) in 0, and air over the temperature range 800-1200°C. In 0, cast XAl which was abraded through 120 grit S i c formed alumina and exhibited k, values = 10-9g2cm-4hr-’ at 950 “C but polished specimens formed Ti0,-rich scales and exhibited k, values of about 10- to 10-6g2cm-4hr-’. Extruded TiAl formed alumina scales regardlcss of surface preparation. Choudhury et al. explained the effects of extrusion in terms of the absence of macroinhomogeneities which existed in the cast structure. The “surface finish effect”, observed for the cast material was presumed to result from coarse grinding homogenizing the alloy at the surface. A %,A1 layer was reported to form between the oxidc and the alloy for all the exposure. Oxidation behavior in air at 950°C was independent of specimen preparation or fabrication method with titania-forming kinetics (k, about 10-5-10-6g2cm-4hr-’) observed in all cases. The scales were similar to those formed on polished specimens in 0, at 950 “C. Experiments to determine the species responsible for the difference between exposures in 0, and air indicated that CO, CO, and €1,0 impurities were not responsible nor was the difference in poz. It was, therefore, concluded that the increased rate of oxidation in air was a “nitrogen effect” although no N,-containing phases were identified in the scale or substratc. The oxidation of ?>A1also exhibited a “temperature effect” in that kinetics for all alloys at 1100 and 1200”C,regardless of surface preparation or exposure atmosphere, indicated XO, formation. Internal oxidation was also observed at the two higher temperatures. There has recently been a substantial amount of work on the oxidation of TiAl with seemingly contradictory results. Unfortunately, much of the confusion has arisen because the authors are unaware that temperature, surface preparation, and composition of the exposure atmosphere all play a critical role in determining the nature of the rcaction products and the kinetics, as illustrated by Choudhury et al. [50] more than twenty years ago. The “temperature effect” is illustrated for TiAl with a ground (600 grit) surface [60] in Figure 16. For specimens exposed in pure 0, at temperatures below about 800°C a continuous alumina layer, shown schematically at left in Figure 17a, covers the entire specimen surface. At temperatures between 800 and lo00 “C small nodules, containing both alumina and titania, form at various locations, as seen at right in Figure 17a. The area density of the nodules increases as the oxidation temperature is increased but the nodules do not grow with time after their initial formation. At temperatures above 1000°C the nodules grow continuously and develop a continuous mixed oxide layer, similar to that which forms on a2with greatly accelerated kinetics, Fig. 16. Mendiratta and Choudhury [51] reported that varying the Al-content of TiAl(50,53, and 54 at%) did not affect the oxidation behavior. Appalonia et al. [61,62], however found that increasing A1 content, in the range 50 to 56 atyo,increased the temperature at which the accelerated kinetics began. The “surface finish effect” for oxidation temperatures below 1000°C is illustrated in the mass gain vs. time plot in Figure 18 for Ti-50at% A1 oxidized at 900”C.The ground
G. H. Meier
32
A
20,
700
750
800
850 900 Temperature (“C)
950
1000
1 50
Fig. 16. Mass change versus temperature for TiAl oxidized for 58 hours in air, oxygen, and oxygen which is contaminated with a small amount of nitrogen.Air and O?(N?)curves are for 54 at% A1 and 0, curvc is for 52 at% Al.
(600 grit) surface results in very slow kinetics in 0, as the result of continuous alumina formation as shown schematically in Figure 17a.The polished (1 pm diamond) results in greatly accelerated kinetics as the result of the formation of a continuous layer of a mixed oxide, Figure 17b. Part of the beneficial effect of grinding may be in homogenizing the surface, as proposed originally by Choudhury et al. [50].Indeed, the distribution of the minor a, phase, generally present in TiAI, has been shown to influence the composition of the external scale [63,64]. However, an additional important effect is the recrystallization of the ground surface as illustrated in the transmission electron micrographs in Figure 18 [64]. The deformed layer, produced by grinding, recrystallizes during the initial oxidation exposure to produce y grains with a grain size on the order of 1 pm from a starting y grain size on the order of 100 pm. No recrystallized layer forms on the polished surface. It has been suggested [64] that enhanced Al diffusion in the fine-grained layers promotes the fo5mation of continuous alumina in the same manner as that reported by Giggins a$# Pettit [65] for the development of chromia films on Ni-Cr alloys. The “nitrogen effect” is also illustrated in Figure 16 and the mass gain vs. time plot in Figure 18.The kinetics in air are essentially independent of surface perparation and are much faster than those for a ground surface exposed in pure 02.Thescales formed in air for ground and polished surfaces are shown schcmatically in Figures 17c and 17d, respectively. Similar observations have been made by a number of investigators [50,66,67]. In a study to clarify the “nitrogen effect” [60] protective alumina scales were observed to form on TiAl (52 at% Al), exposed in 0,,up to 1000°C. However, the same exposures conducted in air resulted in the formation of Ti0,-rich scales which grew at rates orders of magnitude faster than pure alumina scales and even trace amounts of N, influenced the oxidation morphology, Fig. 16.The rate of oxidation increased continually as increasing amounts of N, were added to pure 0, at 900”C.The addition of
2 Research oti Oxidation and Etiibrittlemetit of Intermetallic Compounds in the US.
33
Mixed oxide nodule
1
Recrystallized ylayrr
1
a) Oxygen exposure, 600 grit surface preparation
I
Recivstallized Ylaver
c) Air exposure, 600 grit surface preparation
b) Oxygen exposure, 1 micron surface preparation
I
d) Air exposure, 1 micron surface preparation
Fig. 17.Schematic diagrams of the scales formed on y-TiAl in the temperature range 800-900°C in air or oxygen and with two different surface finishes.
10 % N, to 0, resulted in the formation of nodules of intermixed TiO, and AI,O, interspersed with thin areas of protective AI,O, which would cover the entire surface in the absence of N,. The area density of these nodules increased as the concentration of N, increased, until the surface was completely covered with the mixed oxides when the gas contained 90 % N,. One effect of N, appears to involve the nucleation and initial growth of the scale since preoxidation in a nitrogen-free gas develops an alumina scale which remains protective during subsequent exposures in air [60] even under cyclic conditions [68]. The influence of N, on the initial scale development has been investigated using Auger spectra collected while sputtering through scales formed for short times in 0, and air [60].The scales on specimens which were heated to 900°C and immediately cooled were 2000-3000p\ thick with AES profiles which were consistet with intermixed transient oxides of A1 and Ti. However, specimens exposed in air were found to have a N,rich layer at the scale/alloy interface.This layer was postulated to contain TiN but was
34
G. H. Meier
Time (h)
600 grit
1 Pm
Fig.18. Oxidation rates for y-TiAl at 900°C in air and oxygen with two surface finishes and crosssection TEM micrographs of specimens with two different surface finishes exposed in oxygen for 1 hour showing the recrystallized layer which formes on the ground specimen and the cubic “Xphase” which forms in the Al-depleted zone on both specimens.
not unequivocally identified. The scales on specimens which were held in O2at 900°C for 15 minutes were quite different from those in air.The AES profiles indicated that a continuous layer of aluminum oxide was forming below the mixed transient oxide in 0, while in air the scale consisted only of intermixed Ti- and Al-oxides and was almost ten times as thick as that in 0,. It was suggested that a major effect of the N, was in forming a nitride layer which prevented the alumina from developing continuity. Nitrogen has also been observed at the scale/alloy interface in SNMS profiles through scales formed in air for 1.5h on Ti-5OAl at 800°C [69]. The proposal that the “nitrogen effect” involved formation of nitrides which prevented the alumina from bqoming continuous has been verified by elemental spectroscopic imaging (ESI) in a transmission electron microscope 1701. An ESI map for the elements Ti, Al, 0, and N from a cross-section of the scale/alloy interface of Ti50 at% A1 oxidized in air for one hour at 900°C is presented in Figure 19.These maps indicate that the alumina is broken up by islands of TiN. The micrographs in Figure 18 reveal an intermcdiate layer between the oxide and the y phase.?his zone is depleted in A1 and was originally thought to be a2(501.However, an
2 Research on Oxidation and Embrirtlernent of Intermetallic Compounds in the U.S.
35
Fig. 19. Maps formed using electron spectroscopic imaging in a transmission electron microscope from the cross-section of a specimen of y-TiAl oxidized for 1 hour in air at 900°C showing intermixed regions of TiN and alumina.
important observation was made by Dowling and Donlon in 1992 [71] when they identified a cubic phase with a lattice parameter of 0.69nm and a’WAl ratio of approximately 24. These observations were confirmed [70] and the cubic phase was found to have 432 point group symmetry [72]. Extensive TEM investigation of this phase using convergent beam electron diffraction (CBED) [73] showed it to belong to one of two space groups, P432 or P4,32, and EDS analysis indicated an approximate composition of 57at% Ti33at%AI-lOat %O [74]. Increased exposure times and/or higher oxidation temperatures result in the Al-depleted zone becoming two-phase as the result of nucleation of a2 at the interface between the cubic phase and the parent y 1741.This observation confirms earlier suggestions of a two-phase depletion layer based on WDS measurements [66], Auger electron spectroscopy [75], and X-ray diffraction [76]. The influence of the depleted zone on the oxidation behavior of y alloys is not yet clear, however, it is crucial to the mechanical properties, as will be discussed in a subsequent section.
G. H. Meiet
36
The effects of alloying elements on the oxidation behavior of y-TiAl have been tabulated by Doychuk [5] with regard to their positive or negative effects and the subject has been recently addressed by Rahmel et al. [77]. The early work by Choirdhirry et al. [50] indicated that Nb and W additions promoted the formation of continuous alumina on y at 950°C independent of surface preparation or exposure environment (air or 0,)while Hf had a minimal effect and Y and Ga were detrimental. Unpublished activity data are quoted as indicating that Nb and W additions increase the a,,/a, ratio and promote alumina stability. Unfortunately, comparison of much of the more recent work is difficult because experiments have been performed at different temperatures and in different atmospheres and various surface finishes (which are sometimes not clearly specified) have been used. However, the following generalizations can be made. Additions of Nb [SO, 78-81], W [SO, 78,791, Ta [Sl], and Si [79,82] are generally beneficial. Additions of V and Mo promote alumina formation at temperatures above 900°C [78,79] but actually accelerate scale growth at the lower temperatures (= 800°C) where y alloys are of practical interest. Additions of Mn are generally detrimental [83]. Somewhat surprisingly, small additions of P and CI are reported to improve oxidation resistance [79]. Chromium is an alloying element of fundamental and practical significance. Small additions of Cr to y ( 5 5 at%) accelerates the growth of the mixed oxide by doping the TiO, to increase the concentration of oxide ion vacancies and/or titanium ion interstitials [84]. However, at higher concentrations the Cr promotes the formation of a continuous external layer of alumina [82].These effects are illustrated for 900°C oxidation in air in Figure 20 [31]. The micrographs in Figure 21 show the different scale morphologies produced by the different Cr additions. The binary alloy formed a mixed oxide layer approximately 10 p,m thick while the layer on the 4 Cr alloy is more than twice as thick. The scale on the 12Cr alloy is continuous a-alumina and is less than 1p,m thick (Note this micrograph is a taper section which makes the layer appear thicker.) The oxide map in Figure 22 [82] indicates the combinations of Cr and Al contents that result in the formation of external alumina in air. The alloying levels required increase as the oxidation temperature is decreased. The mechanism whereby 12 -,?
rj . z 57
v
108-
d
{ G!$
r2
2
4-
M ."
$ d
Ti-48A1 2-
0
Ti-49AI- 12Cr
0
I
I
I
I
I
I
Time (Hours)
I
I
I
Fig.20. Effect of Cr content on the oxidation rate ofTi-48 at% Al at 900°C in air.
2 Research on Oxidation and Embrilllrrnerii of Intermetallic Compounds in the (J.S.
Fig.21. Cross-section SEM micrographs of the thrce alloys from Figure 20. (a) X-48AI. b) Ti48A1-4Cr,and c) Ti-48A1-12Cr)Note that the micrograph of the 12Cr alloy is from a taper section which magnifies the scale thickness by about a factor of 10.
37
38
G. H. Meier
TiCrAl
Ti
-
1000°C Cyclic
8"
I
0
A1
1
03
0.8
0.7
0.6
05
0.4
03
0.2
0.1
0
Cr
-
500 1000 1500 2000 2500 3000 3 Time (Hours)
IC
Fig.22. Oxide map showing the regions of Cr and A1 contents which form external alumina scales in air at various temperatures (left) and cyclic oxidation kinetics for Ti-Cr-AI alloys over a range of compositions for exposures in air at 1OOO"C(right).
Cr promotes alumina formation is still unclear. The compositions which form alumina all fall outside of the single-phase y field [31,85,86]. See, for example, Figure 21c. Figure 23 from the work of Brudy et al. [85]indicates the stable phases in the Ti-Cr-A1 system at 1000°C.The alloys along the boundary for alumina formation, except those with very high Cr contents, consist of two phases, y and a ternary Laves phase TiCrAl. One likely effect the Cr has on forming A1 is by increasing the activity ratio a,&. Brudy et al. [87] observed that binary alloys with A1 contents between 49 and 53 at% reacted with alumina in diffusion couples at 1000°C. However, a Ti-42A1-27Cr alloy did not react under the same conditions. An additional effect is that the "nitrogen effect" apparently does not operate in the presence of sufficient amounts of Cr even though nitrides form beneath the oxide during the early stages of scale formation [31]. The cyclic oxidation plot in Figure 22 indicates an additional feature of the TiCrAl alloys, the excellent adherence of the alumina layers formed on them. The alloys for which the data are plotted include a wide range of compositions, as indicated by the 1073K Alumina Boundary
A1
*O Cr
Fig.23. Section of the Ti-Cr-A1 phase diagram at 1000°C indicating the equilibrium phases for the alloy compositions which form continuous alumina scales (after Bra& et a].).
2 Research on Oxidation arid Embrittlement of Intermetallic Compounds in the U.S.
39
stars on the oxide map. The spalling resistance of the alumina on these alloys is consistent with the good thermal expansion match between y-TiAl and a-alumina (Table 1). Oxidation of TiAI, The oxidation of TiAl, results in the exclusive formation of external alumina [62] and the parabolic rate constants are essentially the same as those observed for NiAl [88]. However,TiAI, is a line compound and, as such, is difficult to prepare by melting and casting as a single phase material. It has been found [88] that excess Al results in rapid transient oxidation prior to the attainment of the slow steady-state rate.
2.3.2.3 Embrittlement of Titanium Aluminides The effects on mechanical properties of exposure to oxidizing atmosphere constitutes what is currently the most important area of study regarding the use of intermetallics as high temperature structural materials. The creep resistance and ductility of a2 at high temperature are much lower in air than in vacuum [58] and room temperature fracture toughness is severely degraded by exposure to oxygen at elevated temperatures [58,89-911. Ward et al. [89] found that embrittlement of Ti-25Al-lONb-3Mo-lV during tensile testing in air at 550 and 650°C was strain-rate sensitive and proposed a dynamic process whereby oxygen embrittled the alloy in the region of a propagating crack. Rakowski et al. [52] found that room temperature ductility of Ti-21Al-llNb was severely degraded by exposures to oxidizing atmospheres at temperatures from 600 to 900°C and that the embrittlement was sensitive to the composition of the exposure environment,Table 2. Oxygen caused severe embrittlement whereas high-purity nitroTable2. Embrittlement of’I’i-21Al-llNb Exposure Conditions
Elongation (YO) 24 h - 900°C 20 h - 900°C Purged System Evacuate and Backfill
Argon < 1ppm H,O < 10 ppm H,O
3.60
Nitrogen < 1 ppm H,O < 32 ppm H,O
1.55
Oxygen < 3 ppm H,O < 50 ppm H,O
3.06
Air < 3 pprn H,O < 50 ppm H,O
6.87
Air - 46 YO H,O Oxygen - 37 YoH,O Hz - 39 Yo HZO
1.30 1.91 1.11
> 20.3
17.0
3.60
2.70
40
G. H. Mrier Cracks
-
6 -6-
&*Mixed
TI, Nb. Al Oxide
2
10
0
500 c T c 800 C in Air
200
400
600 800 1000
Time (hours)
Fig.24. Schematic diagram of the scale and interstitial affected zone (IAZ) which forms on “orthorhombic alloys” in the temperature range 500-800°C (left) and the time dependence of the IAZ thickness (right).
gen or argon did not. However, the presence of water vapor in the nitrogen or argon caused embrittlement and analysis of the fracture behavior suggested a synergistic effect between hydrogen and oxygen. The detailed mechanisms of these phenomena have not been studied and the effect of alloying additions have not been extensively investigated. The orthorhombic alloys have been reported to be less susceptible to environmental embrittlement than the a2alloys [92], however, more recent work [59,93] has indicated that this is not generally the case.The microstructure of the alloy is substantially modified by the oxidation process, as indicated schematically in Figure 24. An “interstitial affected zone” (IAZ) in which the volume fraction of (x2 is increased, forms in the alloy below the scale. A very small angular phase also develops in this
Variation of RT Ductility After 100 hr. Preoxidation in Air
0
400
(degrees
800
Fig. 25. Room temperature ductility (measured in 3-point bending) of “orthorhombic alloys” after 100 hours exposure in air at various temperatures,
2 Resetrrch on Oxidation nnd Ernbrirtlenient of lnternietallic Compounds in the U.S.
41
zone but it has not yet been identified. The hardness of the (IAZ) has been determined to decrease from a high value just beneath the scale to values typical of the base alloy at its deepest penetration. The hardness profiles have been used to measure the growth of the IAZ, Figure 24. The growth is parabolic as long as the overall oxidation kinetics are parabolic, Figure 15, but establishes a constant thickness when breakaway occurs. The formation of the IAZ greatly decreases the resultant room temperature ductility as indicated in Figure 25. Exposures at temperatures as low as 500°C reduce the plastic strain at fracture to near zero. 'I'he fracture toughncss [94] and ductility [71,95-961 of y alloys are also reduced by exposures in air at elevated temperatures as a result of the layers which form beneath the oxide, as described above. Exposures at temperatures as low as 315°C and for times as short as six minutes at 650°C werc found to embrittle Ti-48A1-2Cr-2Nb [95], Figure 26. An additional factor in the embrittlement is the effect of temperature and atmosphere during the mechanical test. Shrouding the test in argon, Figure 26, or raising the test temperature to 150"C,Figure 27, negate most of the embrittling effects of the prior exposure at elevated temperature [95]. Clearly there is a need for a detailed understanding of the mechanisms for influence of environmental exposures on the mechanical properties of the titanium aluminides.
3
2.5
+
+
A 650°C
X 650T (Argon Test)
unexposed
2
0
X
0
+
'3
1 6
a?
g 1.5
0
.-
X
I
M m
0"
Ei
As-Machined (Argon Test)
0 315°C
1
0
A
A
A
0.5
0
A
8
A Ref.: C.Austin and T.Kelly, Structural Intermetallics.TMS. 1993.p.143. I
0.01
I
0.1
0
0
41
A
A
1
I
I 10 Exposure Time (h)
I
100
I
loo0
Fig.26. Effects of air exposure and test atmosphere on the tensile ductility of y-Ti-48AI-2Cr-2Nb at room temperature (after Austin and Kelly).
G. H. Mrier
A A
0 0 Ref.: C. Austin and T. Kelly, Structural Intermetallics. TMS. 1993. p.143. 0
50
100 Test Temperature ("C)
Fig. 27. Effects of air exposure and test temperature on the tensile ductility of y-Ti-48AI-2Cr2Nb (after Austin and Kelly).
2.3.2.4 Effect of Complex Environments on the Oxidation of Titanium Aluminides There has been relatively little work published on the reaction of titanium aluminides in atmospheres other than air or oxygen. Niu et al. [96] studied the reaction of Ti25A1-11Nb in a simulated combustion atmosphere (N,+1%0,+ O.5%SO2)with and without surface deposits of Na,SO,+ NaCl at temperatures between 600 and 800T. Exposures in the absence of surface deposits resulted in reaction rates similar to those described above for simple oxidation.The rates in the presence of the deposits at 600 and 700°C were initially rapid and then slowed markedly after 25 to 50 hours exposure. The rate at 800°C remained rapid with the kinetics being essentially linear. The major difference in the corrosion morphology at 800°C was the presence of copious amounts of sulfides below the oxide scales. The authors postulate a mechanism of attack involving a combination of sulfidation-oxidation and scale-fluxing. Schaeffer et al. [97] have compared the behavior of the y alloy Ti-48A1-2Cr-2Nb with the nickel-base superalloy RenC 80 in a high velocity oxidation test in a burner rig burning Jet A fuel and a hot corrosion test in which sea salt was injected into the burner rig. The cross-sections of the two alloys after the high velocity oxidation test are compared in Figure 28.The scale on the y alloy is thinner and more compact than that on the superalloy. Figure 29 presents macroscopic photographs of the two alloys following the hot corrosion tests which indicate the hot corrosion resistance of y is equivalent to, or better than, that of RenC 80.The y alloy formed a compact oxide that
2 Research otz Oxidatiotz and Enibrittleinerit of bilernietallic Compounds in the U S . a)
Ti-48AI-2Cr-2Nb
Rent 80
50p.m
Fig.28. Cross-sections of Ti-48AI-2Cr-2Nb and Rene 80 after 200 hours of high velocity oxidation testing at 871 "C (after Schaeffer et al.).
43
44
G. H. Meirr
Gamma (Ti-48A1-2Cr-2Nb) Ren6 80 Fig.29. Macroscopic photographs ofTi-48A1-2Cr-2Nh and 1icnC 80 aftcr 5 0 hours o f h o t coi-rosion testing (after Schrreffer ct al.).
spalled in the region of maximum tcrnperature but did not form liquid corrosion products. RenC 80 underwent oxidation and sulfidation t o ii substantial depth with the formation of liquid corrosion products. It is anticipated that studies, such as those above. will become more numerous as titanium aluminides approach the application stage.
2.3.2.5 Coatings on Titanium Aluminides The rapid oxidation kinetics of the titanium aluminides and. particularly, the embrittlement has created a substantial interest in the possibility of forming a protective coating on these alloys. Coatings of TiAI, have been formed by pack cementation on cx2 [99,100] and y [101].These coatings form continuous alumina scales but TiAI; i\ an extremely brittle compound and tends to crack, particularly for thicker coatings [ 1001. It is also likely that the presence of a layer of TiAl, on the surface of a2or y will be as embrittling as a high temperature oxidation exposure. The ability of TiCrAl alloys to form protective alumina scales raises the possibility of applying them as protective coatings [102]. Coatings have been successfully applied to y substrates by sputtering, low pressure plasma spraying, high velocity oxygen fuel spraying and slurry fusion [102]. Figure 30 shows thuxyclic oxidation kinetics for a sputtered coating and a cross-section of the coating after the oxidation exposure. The oxidation kinetics were similar to those observed for bulk’TiCrA1 alloys and there appears to be niinimal degradation of the coating by either oxidation or interdiffusion with the substrate. Cockerum and Rupp have evaluated the kinetics of silicide coatings on ’Ti [ 1031 and have used a halide-activated pack-cementation method to form boron- and germanium-doped silicide coatings on orthorhombic alloy substrates [104].The coatings greatly decreased the cyclic oxidation kinetics and microhardness measurements did not indicate diffusion of oxygen into the substrate.
2 Research on Oxidatiori m i d Embriitlement of Intermetallic Compounds in the US. h
y
45
0.9 ,
0.2 > , .
0
I
Sputtered Coating on TiAl (28.7Ti-44.5AI-26.8Cr) Cyclic Oxidation at 900°C in Air
500
I
,
.
1
1
1
,
loo0
r
I
1500
,
1
l
.
2Ooo
I
I
,
21
x)
Time (Hours)
Fig.30. Cyclic oxidation kinetics for a sputteredTiCrA1 coating on a y-substrate at 900°C in air and a cross-section of the coating after exposure.
The important aspect of coating any of the titanium aluminides is that thc coating prevent interstitial embrittlemcnt and that thc coating does not function as an embrittlement layer. Additional studies are necessary in this area.
2.3.2.6 Oxidation of Titanium Aluminide-matrix Composites The use of fibres, such as Sic, is being investigated as a way to strengthen a2 and orthorhombic alloys. These can be fabricated by pressing the fibers between thin foils of the titanium aluminide. The presence of these fibers can degrade the oxidation resistance and penetration of oxygen along the fibers can degrade the interfacc and embrittle the surrounding matrix. Additionally, the substantial thermal expansion mismatch between S i c and the matrix (Table I ) can result in matrix cracking during cooling. These phenomena are shown schematically in Figure 31.
Fig.31. Schematic diagram of the oxidation morphology and cracking observed in “orthorhombic
alloy/SiC composites (left) and the oxidation rates of the composites over the temperature range 500-900°C (right).
G. H. Meier
46
2.3.3 Refractory Metal Compounds Refractory metal silicides have been used for many years as furnace heating elements and protective coatings on refractory metals. In recent years there has been interest in using refractory metal intermetallics as structural materials and in composites which have the potential of exceeding the temperature capability of the nickel-base superalloys. However, the oxidation resistance of most of these compounds is poor despite their high contents of Si or A1.The following section describes: (i) the oxidation of specific silicides with emphasis on MoSi2,(ii) silicide coatings. (iii) silicide-based composites, and (iv) oxidation of compounds in the Nb-AI system.
2.3.3.1 Oxidation of Specific Silicides Oxidation of MoSi, Molybdenum disilicide is an intermetallic compound which has been extensively used for high temperature applications, particularly furnace heating elements. The oxides of Mo (MOO,, MOO,) are much less stable than SiO, (Fig. 1) so that silica should be the stable oxide for any but the most dilute Mo-Si alloys [105]. In fact, the nature of the external scale formed is a strong function of temperature. Figure 32 shows the rates of oxidation of MoSi, over the temperature range 500 to 1400°C.The cross-hatched region represents a large amount of data in the range 600-1400°C where the mass changes are small. However, at 500°C the rates are much faster.The oxidation mechanisms can actually be broken into three temperature regimes. 7.8 6.8
5.8
%
h
4.8
\
Z’ a8
3.8
W
2.8
3 m
2
b0 .*
s” Q
1.8
0.8
-0.2
-1 2 0
20
40
60
80
Time, h
100
120
140
160
1
‘O
Fig.32. Oxidation data for MoSi, as a function of temperature.
2 Resenrch ori Oxirlntiori and Emhrittlenient of Interriiernllic Compounds iri the U.S.
47
1. Regime I(lOOO”<:to 1400°C) All studies report protective oxidation in this temperature range resulting from the formation of a continuous silica layer.
2. Regime I1 (600°C to 1000°C) This temperature regime has not been as extensively studied as the high- and low-temperature ranges. I Iowever, an important transition between protective and non-protective oxidation occurs. A silica layer forms but does not completely seal the surface since oxidation in microcracks under the scale has been observed [lo51 and Mo oxides have been detected in the scale [106].However, the presence of the Mo in this temperature range does not greatly accelerate t h e oxidation kinetics. This is consistent with reports that high temperature oxidation kinetics can be extrapolated as far down as 600°C and that the activation energies for the oxidation of Si and MoSi, are virtually identical in the range between 600 and 1400°C [107].
3. Regime 111 (300°C to 550°C) In this temperature range a protective silica film does not form.This leads to the phenomenon of “accelerated oxidation” [105,108] which involves the formation of scales containing crystalline oxides of Mo and vitreous silica and rates which are orders of magnitude faster than those extrapolated from higher temperatures. Accelerated oxidation is a material property of MoSi,, including single crystals. It has been proposed that the transition between Regimes I11 and I1 is associated with the increase in volatility of the Mooxidcs as temperature increases [105,108]. This is shown schematically in Figure 33. At high temperatures the transient Mo-oxides evaporate and allow the silica regions to grow laterally into a continuous layer. At low temperatures the reduccd volatility of the Mo oxides and slower growth of the silica prevent the development of a continuous silica layer. The rapid inward growth of MOO, results in continuous removal of the silica nuclei from the surface and produces an intermixed layer of MOO, and SiO,. Direct measurements of the oxide evaporation rates from MoSi, have not been performed but the above scheme is supported by the following: i. Water vapor in the oxidizing gas, which accelerates the evaporation of the Mo-oxides as hydrated species, aids the development of continuous silica [log]. ii. Incorporation of Na into the MoSi, to accelerate the lateral growth of silicates and react with MOO, has been reported to alleviate accelerated oxidation, Figures 34 and 35 [109]. iii. Compounds such as NbSi, and TaSi,, which have transient oxides with very low vapor pressures, undergo accelerated oxidation to temperatures well above 1000°C [105]. The results in Figures 34 and 35 were produced by a novel approach to controlling accelerated oxidation [lo91 in which alkali salts were applied to the surface of MoSi, prior to oxidation.‘Ihe mechanisms, for the case of NaF, is shown schematically in Figure 35. The MOO, reacted to form Na-molybdate precipitates and the silicon reacted to form Na-silicate, which grows more rapidly than silica, to provide a continuous, protective layer. A limited number of experiments at 1200°C indicated that the presence of the Na did not degrade the high temperature oxidation resistance. This result sug-
G. H. Meier
48
Fig.33. Schematic diagrams of the development of the oxidation morphology on MoSi, at temperatures of 600°C and higher (top) and the temperature range around 500°C (bottom).
..
0.8
0.6
-5
0.4
P
E
20
0.2
v
Treated MoSi,
0
-0.2
-0.4 600
Time (h)
1200
I
Fig.34. Oxidation data for MoSi, a t 500°C with and without application of NaF (after Cockerorn and Rupp).
2 Reseurch on Oxidofion und Embrittlemerii o fIntermetallic Compounds in [he U.S.
49
Fig.35. Oxidation morphology development for MoSi, at 500°C with application of NaF (after Cockeram and Rapp).
gests that the presence of alkali salts do not always catastrophically degrade silica formers, as is generally believed. However, weight losses associated with evaporation were reported and it is possible that the amount of Na remaining was insufficient to impede silica formation. If the MoSi, contains defects, such as microcracks or pores, the occurrence of accelerated oxidation within these defects can result in crack propagation and fragmentation of the MoSi,. Fitzer [110] first described the phenomenon and named it “pest” in 1955. Since that time, many researchers have attempted to describe pesting of MoSi, and determine under what circumstances it occurs. Berkowitz-Mattuck et al. [lll,1121 studied zone-refined material of less than 75 % theoretical density and found that pesting occurs in oxygen, but not nitrogen, carbon dioxide, carbon monoxide or argon, and that the rate of oxidation is very sensitive to the partial pressure of oxygen. Furthermore, as there is little or no lattice parameter change when MoSi, is equilibrated in oxygen, the solubility of oxygen in MoSi, is assumed to be very 1ow.They concluded that, as a result of the high residual stresses introduced on cooling of the anisotropic material from the melt, a stress-enhanced oxidation could occur at the tips of Griffith flaws, eventually leading to brittle fracture. The disappearance of pesting at T > 600°C was explained as the result of plastic deformation of the matrix near the flaw accommodating the stresses. The cracks were found to be mostly transcrystalline. Westbrook and Wood [113] proposed that the catastrophic nature of the “pest” mechanism was the result of preferential intergranular diffusion of a gaseous element (most likely oxygen or nitrogen), coupled with a temperature dependent hardening reaction. Fitzer et al. [llO, 114,1141 and Schlichting [116] describe pesting of MoSi, as intercrystalline attack whereby each individual grain is enveloped by reaction product.They note that most oxidation occurs in pores or internally along pore canals, and failure occurs as the result of a wedging effect from oxide growth in the defect. Fitzer and Schlichting give no evidence that attack is intergranular and may only assume so, as pores tend to form predominantly along grain boundaries. This is inconsistent with Berkowitz-Mat-
50
G. H. Meier
tuck et al. who found fracture from pesting to be predominantly transcrystalline. Rccent work [lo51 has shown that, while accelerated oxidation is generic to all f o r m s of MoSi,, grain boundaries alone do not result in pesting since dense HIPetl MoSi, did not pest even though it was polycrystalline. Only cast material, which contained prcexisting microcracks, was observed to undergo pesting. It was concluded that pesting was the result of the occurrence of accelerated oxidation within the microcracks. There is a large change in volume going from Mo to MOO; (= 340%), along w i t h the volume change of forming SiO, from Si (= 180 YO).These processes enhance the widening of the pre-existing cracks leading to pesting (i. e. turning to powder).These phenomena are illustrated schematically in Figure 33. 'This mechanism is supported by the observation of oxidation-induced growth of cracks formed in HIPed MoSi, by a microhardness indenter which ultimately resulted in pesting of a material which did not undergo pesting in simple oxidation exposures [117]. McKnrncy et al. [ 1 181 have also concluded that pesting occurs as the result of oxidation in preexisting cracks and pores. There are few studies of alloying effects on the oxidation behavior of MoSi, in the published literature. Recently Yunagihara et al. [ 1191 have reported on alloys in which 15% of the Si was replaced by A1 or 10 % of the Si was replaced by Ta.Ti, Y or Zr.l'he MoSi, had the C l l b crystal structure, the Al-containing alloy the C40 structure. and the others were two-phase Cllb+C40. The alloys were oxidized in air over the temperature range 1435 to 1685"C.Theoxidation kinetics for MoSi, and the Al-;ra-, and --containing alloys were parabolic with MoSi, and MoSi,+-Ta having the slowest rates and MoSi,+Al having a rate constant more than a factor of ten larger. The MoSi,+Ti had rates similar to MoSi, at temperatures below the eutectic temperature in the'I'i0,SiO, system (1550'C) and rates similar to MoSiz+Al above this temperature. The alloys containing Y and Zr exhibited non-parabolic kinetics and oxidation rates which were much greater than for the other alloys. The scales formed on MoSi, and MoSi,+Ta were silica, that on the Al-containing alloy was alumina or a liquid solution. and the scales on the other alloys were complex mixtures of silica and other oxides.
Oxidation of other silicides Accelerated oxidation phenomena have not been studied in detail for refractory metal silicides other than MoSi,. It is generally observed that WSi, has oxidation resistance roughly comparable to that of MoSi,, while TaSi, and NbSi, have considerably poorer resistance, even at very high temperatures [105,120].The Me,Si,-type silicides are only observed to form protective silica films at extremely high temperatures [ 1211. In all cases the rapid oxidation results from the rapid formation of a refractory metal oxide which prevents the formation of a continuous silica layer. Virtually all of the refractory metal silicides have been observad to undergo pesting.
2.3.3.2 Silicide Coatings The refractory metal silicides have been used for many years to protect refractory metals from oxidation in very high temperature, but short duration applications [121]. These coatings have been highly successful but their use in applications which require long term stability have been limited by problems with accelerated oxidation and
2 Research on Oxidation mid Embrirrlement of Intermetallic Compounds in the U.S.
51
pesting, evaporation of SiO at low oxygen partial pressures, interdiffusion with the substrate, and cracking because of thermal expansion mismatch between the coating and substrate.These fractors have been reviewed in detail by Packer [122] and Kircher and Courtright [123].Rccent work by Rnpp and coworkers [124,125] has been directed at improving the resistance of MoSi,-based coatings for Nb-base alloys, for which there is a good thermal expansion match, and Mo, for which there is a relativcly poor thermal expansion match (Table 1).The coatings on Nb, formed by pack cementation, consisted of W additions to the MoSi, to strengthen it and Ge additions to increase the thermal expansion of the protective SiOz layer to improve thc cyclic oxidation resistance and to lower the viscosity to reduce accelerated oxidation at low tempcratures. The coatings were reported to provide cyclic oxidation resistance on Nb for 200 hours at 1370°C [124].The coatings on Mo have made use of Ge doping and also the abovedescribed Na-doping [109], by means of a NaF activator in the coating pack, to successfully limit accelerated oxidation and pesting at low temperatures 11251. Data with regard to coating cracking during thermal cycling from high temperatures, which would be expected to be severe because of the poor match between the thermal expansion coefficients of Mo and MoSi,, have not yet been reported.
2.3.3.3 Silicide-Based Composites The compounds, such as MoSi,, are intrinsically brittle at low temperatures and weak at elevated tempcratures. Composites have been developed to toughen and strengthen the matrix. However, the presence of the second phase also influences the oxidation behavior, usually in a detrimental manner. Figure 36 [126] shows thc effects of 30 vol% of various reinforcing phases on the oxidation of MoSi, in air at 1200°C. The presence of TiB, and HfB, result in an increase in scale growth rates because of the incorporation of TiO, and HfO, into the scales (see schematic in Figure 37). The
Tim (hr)
Fig. 36. Effect of various second phases on the oxidation rate of MoSi, at 1200°C.
52
G. H. Meirr
Fig.37. Schematic diagram of the oxidation morphology developed on a MoSi,-HfB, compo~itt.at 1200°C.
presence of Sic, which is also a silica-former, has relatively little effect on the oxidation rates at this temperature [l26,127].The reinforcing phases have little effect on the low temperature oxidation kinetics under isothermal conditions, i. e. the composites undergo accelerated oxidation at 500°C similar to that shown in Figure 33. However, under thermal cycling conditions, the thermal expansion mismatch between the MoSi, matrix and reinforcing phases, such as S i c (Table l ) , causes matrix cracking. Cracks resulting from discontinuous S i c [127] and from S i c fibers (1281 have been shown to cause complete pesting of MoSi,-Sic composites during oxidation at 500°C.
2.3.3.4 Oxidation of Compounds in the Nb-AI System The oxidation behavior of compounds in this system was first extensively studied by Svedberg [129] who found that the only binary compound which formed a continuous alumina scale was NbA1,. The lower compounds formed scales consisting of NbAIO, and Nb,O, and oxidized nearly as fast as Nb. However. NbAI, is a "line compound" and the A1 depletion caused by the formation of the alumina scale results i n the immediate formation of the lower aluminide Nb,AI below the scale. Thus, after rupture of the initial alumina layer, rapidly growing NbAIO, and Nb,O, form until the Nb,AI layer is consumed and alumina can again be formed o n theNbA1,. Repetition of this process results in a layered scale and nearly linear oxidation kinetics [2,130-1321. Excess Al prevents the formation of the layered scale but degrades the mechanical properties and the long term oxidation resistance because of Al evaporation and alumina growth in the grain boundaries.
2 Rcseardi on Oudiitiori and Ernbrrrrl~meiitof Itirermrrullrc Compoitnds in the U S .
53
The oxidation of NbA1, at intermediate temperatures is a striking example of pesting. It has been observed [130] that NbAI, is susceptile between 550°C and 950°C with a maximum between 650°C and 850°C. Gruhke and coworkers have intensively studied NbAI, of stoichiometric composition [ 132-1341 applying thermogravimetry and Auger electron spectroscopy (AES). The maximum rates were observed at 750°C and reduced pressures between and 0.1 bar 0,.At 750°C and pressures between 0.1 and lO-"bar 0, (He-0, mixtures) a stepwise disintegration was observed. At 2xlO-'O to 7 ~ 1 0 - ' ~ ) b O2 a r (in ampoules with C U - C U ,or ~ Ni-NiO) the disintegration was abrupt and after 24 h the specimens were converted completely to powder. At bar 0, (Nb0,-Nb,O,, Cr-Cr203)no attack or disintegration occurred. However, after oxidation at 750°C under all the conditions listed, AES investigations of the specimens fractured in a IJHV system showed oxygen penetration into the grain boundaries. At bar 0, there was no AI,O, formation in the grain boundaries, but at the higher pressures after the preceding grain boundary saturation with oxygen, oxide formation was observed progressing from the surface to the interior. The very rapid disintegration was ascribed to the wedging effect of the inwardly growing A1,0,. To explain the very fast ingress of oxygen into the NbAI, at temperatures about 750"C, corresponding to a diffusivity of about 10-('cm2/sec,a mechanism was put forward in which formation of fissures is assumed at the grain boundaries, where oxygen can enter by gaseous or surface diffusion [135].The pesting of NbAl, was explained as follows: (i) selective oxidation of aluminum with AI,O, scale formation results in Aldepletion of the NbAI, phase, preferentially along grain boundaries, (ii) the Al-depletion leads to a phase transformation to Nb,AI at the grain boundaries, NbAI, transformation causes fissure formation at the grain boundaries and cracking of the outer scale. (iii) oxygen from the atmosphere penetrates into the fissures and A&O, is formed on the surfaces of the scparated grains within the material. (iv) dislocations and low-angle boundaries can also act as short-circuit paths for aluminum diffusion to the grain surfaces, then Al-depletion and Nb,A1 formation also opens cracks into the grains. Doychak and co-workers have also performed extensive studies on accelerated oxidation of binary NbAI, and material alloyed with Cr and Y [136-138].The high temperature oxidation is improved by the alloying additions [137]. The Cr additions resulted in a layer of AlNbCr forming beneath the scale.This compound maintained the stability of the alumina scale better than the Nb,AI which forms on the binary compound. The Y additions reduced convolutions in the scale. The alloyed NbAI, also undergoes pesting [136, 1381 but the presence of Cr apparently prevents grain boundary oxidation. A careful TEM study of the initial stages of accelerated oxidation, which leads to pesting, revealed an external scale covering an internal oxidation zone consisting of A120, and Nb.'The rupture of the external scale and rapid oxidation of the Nb was the cause of the accelerated oxidation. The authors suggest that a similar phenomenon may be occuring in the grain boundary region of binary NbAI,. A number of studies have evaluated the effects of alloying elements on the oxidation behaviour of Nb-A1 alloys [132,136-1411. In some cases the alloys formed alumina scales but none of the alloys were resistant over a broad enough range of exposure conditions to be considered for extensive application. This is particularly the case when the compitions are restricted to those which provide even marginal mechanical
54
G. H. Meier
properties. The author believes that, despite the existence of ongoing programs, attempts to protect Nb-base compounds by the selective oxidation of A1 or Si are fruitless.
2.4 Concluding Remarks The above discussion indicates that there are relatively few intermetallic compounds that form protective alumina or silica scales over wide ranges of exposure conditions. 'The use of high-volume fraction y' (Ni,AI) Ni-base superalloys and aluminide coatings on superalloys remain the most successful technological applications of intermetallic compounds in structural materials. The titanium aluminides are approaching the application stage. However, exposure to oxidizing environments degrades the mechanical properties of essentially all the titanium aluminides. The author believes that developing an understanding of the mechanisms associated with this phenomenon and finding solutions are the most important problems currently in intermetallics oxidation. The possibilities of other intermetallics reaching widespread application as structural materials seems unlikely because of poor fracture toughness and/or poor oxidation resistance. Additionally, it is clear that addition of second phases for strengthening or toughening generally degrades the oxidation resistance of those compounds which are oxidation resistant. The compounds Ni,Al, NiAI, Fe,AI, and FeAl have sufficient oxidation resistance for high temperature applications if their mechanical property shortcomings can be resolved. The only refractory metal compound with adequate oxidation resistance is MoSi,.
2.5 Acknowledgements The author gratefully acknowledges the contributions of D. Berztiss, R. Cerchiara, 1. Rakowski, C. Sarioglu, and C. Stinner for generating many of the results discussed in this paper and for preparation of the illustrations. The author wishes to thank colleagues M.P Brady, J. Doychak, J. A. Nesbitt, R. A. Perkins, B. A. Pint, R. A . Rapp, J. C. Schaefer, and J. L. Smialek who provided copies of their papers, in some cases prior to publication. Finally, the author acknowledges the continued, valuable collaboration with N. Birks and E S.Pettit at the University of Pittsburg.
2.6 References [l] E. A . Aitken in: Intermetallic Compounds, J. H. Westbrook el., p. 491, Wiley (1967). (21 G. H. Meier in: Oxidation of High Temperature Intermetallics,T. Grohstein and J. Doychak
eds., p. 1,TMS (1989). (31 G. H. Meier, E S. Pettit: Mater. Sci. and Eng. A153 (1 992) 548.
2 Research on Oxidation and Embrittlernerit of Intermetallic Compounds in the [J.S.
55
[4] G. f1. Meier, N. Birks, E S. Pettit, R. A. Perkins, H. J. Grabke: in Structural Intermetallics. R. Darolia J. J. Lewandowski. C. T. Liu. P. L. Martin. D. B. Miracle. M. V. Nathal eds., p. 861,TMS (1993). [5] J. Doychak in: Intermetallic Compounds. J. H. Westbrook. R. L. Fleischer eds., p. 977, Wiley (1994). (61 A. K. Vasudevan, J.J. Petrovic: Mater. Sci. and Eng.AZ55 (1992) 1. 171 R. Gibala, A. K. Ghosh, D. C. Van Aken, D. J. Srolovitz, A. Basu, H. Chang, D. P Mason, W. Yang: Mater. Sci. and Eng. A155 (1992) 47. [8] J. Cook, A. Khan, E. Lee. R. Mahapatra: Mater. Sci. and Elng. A255 (1992) 183. 191 R. 7:DeHoff:Thermodynamics in Materials Science, McGraw-Hill, (1993) p. 492. [lo] C. Dior,P Choquet, R. Mevrel in: Proc. Internat. Conf. on Residual Stresses (ICRS-2) G. Beck, S. Denis. A Simon eds., Elsevier, London, (1988) p. 273. [ l l ] ES.Petrit:Trans. Mct.Soc.AIME239 (1967) 1296. 1121 G. H. Meier, E S. Pettit, J. L. Smialek: Materials and Corrosion. 46 (1995) 232. 1131 E. Schumann, M. Riih1e:Acta Metall. Mater. 42 (1994) 1481. 1141 B. A. Pint, L. W. Hobbs: “The Oxidation Behavior of Y,O,-Dispersed Ni,AI”. in: Oxide films on Metals and Alloys, B. R. McDougall et al. eds.. Electrochem. SOC.Proc., Vol. 92-22,1992, p. 92. [15] B. A. Pint, K. B. Alexander, F? 1: Tortorelli: “The Effect of Various Oxide Dispersions on the Oxidation Resistance of Fe,AI”, Mat. Res. SOC.Symp. Proc.,Vol. 364.1995, p. 1315. [16] G. C. Rybicki,J. L. Smialek: Oxid. Metals31 (1989) 275. [17] B. A. Pint, L. W Hobbs: Oxid. Metals 41 (1994) 203. [18] M. M. W Brumm, H.J. Grabke: Corr. Sci.33 (1992) 1677. [19] J. Doychak, M. Riih1e:Oxid. Metals32 (1989) 431. [20] B. A. Pint, M. Treska, L. W Hobbs: submitted to Oxid. Metals. [21] L. Smialek, J. Doychuk, D. J. Gaydosh: Oxid. Metals 34 (1990) 259. [22] M. W Brumm, 11.J. Grabke: Corr. Sci.34 (1993) 547. [23] J. L. Smialek: Me1.Trans.A 9A (1978) 309. [24] E. Schumann, J. C. Yang, M. J. Graham, M. Riihle: Materials and Corrosion 46 (1995) 218. [25] B. A. Pint, L. W Hobbs: “The Cyclic Oxidation Behavior of Oxide-Dispersed P-NiAI”, Mat. Res. SOC.Symp. Proc., Vol. 364,1995, p. 987. 1261 C. Stinner, F;S. Pettit, G. H. Meier: unpublished research, Univ. of Pittsburgh, 1994. [27] B. A. Pint, L. W Hobbs: J. Electrochem. SOC.141 (1994) 2443. [28] E. Schumann: Oxid. Metals 43 (1995) 157. [29] H.J. Grabke, G. H. Meier: Oxid. Metals 44 (1995) 147. 1301 M. W Brumm, H. J. Grabke, B. Wagemann: Corr. Sci. 36 (1994) 37. 1311 D. A. Berztiss: Ph. D Thesis, Univ. of Pittsburgh, 1996. [32] G. W Coward, L. W Cannon: “Pack Cementation Coatings for Superalloys: A review of History,Theory, and Practice”, Paper 87-GT-50,Gas Turbine Conf., ASME, 1987. [33] J. S. Smith, D. H. Boone: “Platinum Modified Alurninides-Present Status”, Paper 90-G‘r-319, Gas Turbine and Aeroengine Congress, ASME, 1990. 1341 S. C. Kung, R. A. Rapp: Oxid. Metals 32 (1989) 89. [35] R. Bianco, R. A. Rapp: J. Electrochem. SOC.140 (1993) 1181. 136) K. Bianco, R. A. Rapp: in High Temperature Materials Chemistry-V, W. B. Johnson and R. A. Rapp Eds.,The Electrochem. SOC.1990 p. 211. 1371 R. Bianco, R. A. Rapp, J. L. Smialek: J. Electrochem. SOC.140 (1993) 1191. [38] W Da Costa, B.Gleeson, D. J. Young: J. Electrochem. SOC.141 (1994) 1464. [39] W Da Costa, B. Gleeson, D. J. Young: J. Electrochem. SOC.241 (1994) 2690. [40] C. Stinner, E S. Pettit, G. H. Meier: unpublished research, IJniv. of Pittsburgh, 1995. [41] J. Doychak, J. A. Nesbitt, R. D. Noebe, R. R. Bowman: Oxid. Metals 38 (1992) 45. [42] R. A. Perkins: unpublished research, Lockheed Palo Alto Res. Lab., 1993. 1431 J. Whittenberger in: Structural Intermetallics, R. Darolia, J. J. Lewandowski, C.T. Liu, P. L. Martin, 11.B. Miracle, M. V. Nathal eds.,p. 819,TMS (1993). [44] J. A. Nesbitt, C. A. Barrett in: Structural Intermetallics, R. Darolia, J. J. Lewandowski, C.T. Liu, P.L. Martin, D. B. Miracle, M. V. Nathal eds.. 1.’ 601,TMS (1993). [4S] J.A. Nesbitt, C. E. Lowel1:“Prediction of the High‘remperature Oxidative Life of Intermetallics”, Mat. Res. SOC.Symp. Proc.,Vol. 288,1993, p. 107.
56
G. 11. Meier
(46) J. I,. Smialek,J. A.Nesbitt, W J. Brindley, M . P Brariy, J. D ~ ~ h nR.kM. . L)ickcwm.U R. l l ~ t l l : “Service Limitations for Oxidation Resistant Intermetallic Compounds”. Mat. Rcs. Soc. Symp. Proc.,Vol. 364,1995, p. 1273. [47] A. Rahmel, PJ.Spencer: Oxid. Metals35 (1991) 53. [48] K. L. Luthra: Oxid. Metals 36 (1991) 475. [49] J. M . Rakowski: Senior Thesis, Univ. of Pittsburgh, 1992. [50] N S. Choudhurv, H. C. Graham, J. W Hinze: “Oxidation Behavior o f?’itanitmi Aluminides”. in Properties of High Temperature Alloys. Z . A. Foroulis and F.S. Pettit etls., The E1ectr.ochcm SOC..1976,p. 668. [Sl] M. G. Mendiratta, N. S. Choudhurv: “Properties and Microstructure o f High-Tempcraturc Materials”. AFMLTR-78-112, Contract No. F33615-7.5-C-l0OS,(Systems Research Lnlioratories, Inc., 0hio.August 1978). [52] J. Rakowski. G. H. Meier. R. A . Perkins:The Oxidation and Embrittlement of CY: (Ti;AI)l i t a nium Aluminides, in Microscopy of Oxidation 2. Inst. of Materials. 1993.p. 376. (531 7:A. Wallace,R. K. Clark, K. E. Wiedemann:Oxid. Metals 42 (1994) 4.51. 1541 M. Khobnib, E W. Vahldiek: “High Temperature Oxidation Behaviour o f Ti;AI Alloys”, Second International SAMPE Metals Conference. F. H. Froes and R. A. Cull etls.. C’OVIIIX Ca, 1988,pp. 262-270, [55] C. If. Koo, J. W Evans, K. I.:Song, 7:I I . Yu:Oxid. Metals -12 (1Y94) 529. [56] 7:A. Wallace,R. K. Clark. K. E. Wiedetnann, S. K. Sankaran: Oxid. Metals 37 (1992) 1 I 1. [57] J. C. Schaeffer: Scripta Met.28 (1993) 791. [58] S. J. Balsone: “The Effect of Elevated Temperature Exposure on the Tensile a n d C’rrep Properties of Ti-24AI-1 INb” In Oxidation of High Temperature Interinetallics.‘T.Grobstcin and J. Doychak eds.,?’MS, 1989,p. 219. [59] R. Cerchiara, J. Rakowski, E S . Pettit, nnd G. H. Meier: Unpublished Research, I!nivrrsity oS Pittsburgh, 1993. [60] G. H. Meier, E S.Pettit, S. l f u :“Oxidation Behavior of Titanium Aluminides” J. de Physiclue IV, Colloque C9.1993, p. 395. 1611 D. S. Appolonia: “Thc Oxidation of Gamma-Titanium Aluminides”, Bachelor of Science Thesis,University of Pittsburgh, Pittsburgh, PA (1988). [62] G. H. Meier, D. S Appalonia, R. A. Perkins, K. 7: Chiang in: Oxidation o f Iligh-Tempcraturc Interrnetallics,T. Grobstein and J. Doychak. eds.. (TMS, 1989) p. 185. (631 A. Gil, H. Hoven, E. Watlura, and W J . Quadakkers: Corr. Sci.34 (1993) 615. [64]J. M. Rakowski,E Dettenwanger, E. Schitmann, G. ?I.Meier: E S.Pcttit. ill. Riililc: ”The Effect of Surface Preparation on the Oxidation Behavior of Gamma TiAl Base Intermetallic Alloys’’, submitted to Scripta Mater. [65] C. S. Giggins, E S. Pettit:Trans. Met. SOC.AIME 245 (1969) 2509. [66] S. Becker,A. Rahmel, M. Schorr, M. Schiitze: Oxid. Mctals 38 (1992) 425. [67] N. Zheng, W J. Quadakkers,A. Gill, H. Nickel: Oxid. Metals 14 (1995) 417. [68] E. Kobnyashi, M. Yoshihara, R. Tanaka: H$h Tcmp.Tech. 8 (1990) 179. [69] CL Figge, A. Elschner, N Zheng, H. Schitsfk( W J . Quadakkers: Fresenius J. Anal. Cheni. 3.46 3 (1993) 75. [70] J. M. Rakowski, E S. Pettit, G. H. Meier. E Dettenwnnger, E. Schumnnn. ill. Riihle: Scripla Met. et Mater.3.7 (1995) 997. [71] W E. Dowling Jr., W.7:Donlon: Scripta Met. et Mater. 27 (1992) 1663. [72] R. Field, J. Schaeffer, C. Austin, 7:Kelly: unpublished, GE Aircraft Engines. 1993. [73] I! Cheng, f?Dettenwanger, J. Mayer, E. Schumnnn, M . Riihle: Scripta Met. et Matcr. 3-1 ( I 9%) 707. [74] E Dettenwanger, E. Schumann, J. Rakowski, G. H. Meier, M. Riihle: submitted to Materials and Corrosion. [75] K. W Beye, R. Gronsky:Acta Met. et Mater. 42 ( I 994) 1373. [76] N. Zheng, W Fischer. H. Griibmeier. I/:Shemet, W J. Quadakkers: Scripta Met. et Matcr. .?.? (1995) 47. [77] A. R a h m , W J. Quadakkers, M. Schiitze: Materials and Corrosion 46 (199s) 271. [78] R. A&.??kins. K. 7:Chiang, G. H. Meier, R. A. Miller: “Formation o f Alumina on Niohium a n d Titanium Alloys”, in: Oxidation of IIigh-Temperature Intermetallics, 71 Grohstein and J. Doychak, eds.,The Min., Met., and Materials SOC.,1989 p. 1.57.
2 Reseurch on Oxitlotion and Enibritrlenzent oflriternzetallic Conipounds in the U S .
57
[79] Y Shida, H. Anntlu: Mater.Trans. JIM 35 (1994) 623. [SO] H . Nickel, N. Zheng,A. Elschner, W J. Quodnkkers:Mikrochim. Acta. 119 (1995) 23. [81] T A . Wnlluce, R. K. Clark, S. N. Sarikarnn, K. E. Wiederrrinnri:In Environmental Effects on Advanced Materials, R. H. Jones and R. E. Ricker eds.,TMS 1991,p. 79. [82] R. A. Perkins. G. H. Meier: “Oxidation Resistant Aluminides for Metal Matrix Composites”, in Advanced Materials Confcrence 11,F. W. Smith ed.. Advanced Materials Institute, 1989.p. 92. [S3] Y Shirlo, H . Anudn: Corr. Sci. 35 (1993) 945. [84] C. Stinner:M. S.Thesis, University of Pittsburgh, 1992. [85] M . P Brndy, J. L. Sminfek, F: Terepkn:Scripta Met. et Mater. 32 (1995) 1659. [86] M. R Brndy, J. L. Sniinfek,D. L. llumphrey: Mat. Res. SOC.Symp. Proc..Vol. 364.1995, p. 1309. [87] M . P Hrod-y, J. L. Sniiolek, D. L. Ilunzphrey: “Mechanism of Alumina Formation in li-Cr-A1 Alloys“, Extended Abstract. Fall Meeting.The Electrochem. SOC.,1995. [88] J. L. Sniialek, D. L. llumphrey: Scripta Met. et Mater.26 (1992) 1763. [89] C. H . Word,J.C. Williarns,A. W Thompson: Scripta Met.28 (1993) 1017. [90] C. H. Wurk 1nternat.Mater. Rev.38 (1993) 79. [91] Y Saifohand K . Mino: Mater.Trans. JIM34 (1993) 393. [92] P R. Smith, J. A. Groves, C. G. Rhodes: “Preliminary Mechanical Property Assessment of a SiC/OrthorhombicTitanium Aluminide Composite’‘ in Structural Intermetallics, R. Ilarolia, J. J. Lewandowski.
58
G. H. Meier
[116] J. Schlichting: HighTemperatures-High Pressures 10 (1978) 241. [117] D. R. Rishel, ES.Pettit, G. H. Meier: Univ. of Pittsburgh, Unpublished Research. 1992. [ 1181 C. G. McKumey,f? E Torforelli,J. H. Devan, C. A. Curmichael: J. Mater. Res. 7 (1992) 2747. [ 1191 K . Yunugihara,7: Mnruynmu, K. Nuguta: Intermetallics 3 (1995) 243. [120] R. E. Regari. W A. Baginski, C.A. Krier: Ceram. Bull. 46 (1967) 502. [121] H. Lovendel, R. A. Perkins, A. G. Elliot, J. Ong: “Investigation o f Modified Silicide Coatings for Refractory Metal Alloys with Improved Low pressure Oxidation Behavior” AFML-TR65-344.1965. [122] C. M. Pncker in: Oxidation of High-Temperature 1ntermetallics.l’. Grobstein and J. Ihychnk, eds.,The Min., Met., and Materials SOC.,1989,p. 235. [123] T A .Kircher, E. L. Courtright: Mater. Sci. and Eng.AI.55 (1992) 67. (1241 A . Miteller, G. Wung,R. A. Rapp. E. L. Courtright: J. Electrochem. SOC.139 (1992) 1266. (1251 B. V Cockeram, G. Wung,R. A. Rapp: Materials and Corrosion 46 (19%) 207. [ 1261 C. Sarioglu: “Oxidation Behavior o f MoSi, and MoSi,-Based Composites at 500-1200”C”, M. S.Thesis, Univ. of Pittsburgh. 1993. [127] J. Cook,A.Khan, E. Lee, R. Muhupnfru:Mater.Sci. and Eng.Al5.5 (1992) 183. [128] M. J. Maloney, R. J. Hecht: Mater. Sci. and Eng. A155 (1992) 19. [129] R. Svedberg: “Oxides Associated with the Improved Air Oxidation Performance of Some Niobium Interrnetallics and Alloys”, in Properties of High Temperature Alloys. Z. A. Foroulis and F. S. Pettit eds.,The Electrochem SOC.,1976,p. 331. [130] G. Rnisson, A.Vignes:Revue o f Physique Appl. 5 (1970) 535. [131] M . Steinhorsf, H. J. Grubke:J. Mater. Sci. A220 (1989) 55. [132] H. J. Gruhke, M. Steinhomf,M. W. Brumm, D. Wiemer:Oxid. Metals 35 (1991) 199. [133] M. Sfeinhorsf,H. J. Grubke: Z. Metallkunde 81 (1990) 732. [134] H. J. Grubke, M. W.Brumm, M. Steinhorst: Mat. Sci. &Tech. 8 (1992) 339. [135] V K . Tolpygo,H.J. Crubke: Scripta Met. et Mater.29 (1993) 747. [136] J. Doychuk, S. V Ruj, I. E. Locci, M. G. Hebsur: “Accelerated Oxidation of NbA1,-Base Alloys at Intermediate Temperatures”, NASA HITEMP Review 1991, NASA Conf. Publ. 10082,1991. 11371 J. Doychuk, M. C. Hebsur: Oxid. Metals 36 (1991) 113. 11381 S. V Raj, M. G. Hebsur, 1. E. Locci, J. Doychuk:J. Mater. Res. 7 (1992) 3219. 11391 R. A. Perkins, C. lf. Meier: JOM 42 (1990) 17. 11401 M. I! Brudy, R.J. Hunrahun Jr., S. I! Elder Rundull, E. D. Vering: Scripta Met. et Mater. 28 (1993) 115. 11411 H. J. Crubke, M. W. Brumm, M. Steinhorsf:Mater. Sci. and Tech. 8 (1992)399.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
3 Oxidation of Intermetallics - Japanese Activity S. Taniguchi
3.1 Introduction During the last ten years the research on the high-temperature oxidation of intermetallic compounds became very active in developed countries. It is therefore worthwhile to surnmarise and review the recent Japanese activity on such researches. A brief review of this kind is useful to compare the trends among foreign countries where researchers and engineers may have different views and concepts in the development and application of potential intermetallic compounds. Of course, the scientific intcrest should also be emphasized where the discussions among researchers of various origins and senses are very fruitful. There are nearly ten rescarch groups in universities, two at a national institute and nearly tcn groups in industries in Japan, active on oxidation of intermetallics. Intermetallic compounds have often been the main subject of international and domestic meetings and symposia [l-51. In particular, the materials based on y-TiAl have been receiving considerable attention, because they arc referred to as structural materials having high specific strengths [6,7] at elevated temperatures in near future. With the above background an cxtensive research has been carried out over their mechanical properties, microstructures, crystallography, phase stability, production processing, and behaviour in oxidising and corrosive environments. The possibility of their practical application is widening in recent years. The production technology showed a significant progress in the formation of turbocharger rotors, engine valves and so on. Their service performance is under study. Although the strength of TiAl-base materials is a little lower than that of the conventional nickel-base supcralloys at present, they are promising under the circumstances where their high specific strength becomes a great advantage. The research activity on the oxidation and corrosion in Japan is characterised by the fact that the major effort has been expended on TiAl-base materials. The research on MoSi, follows that on TiAl-base materials with much small percentage. Other aluminides and silicides are being dealt with. There seems to be no research on beryllides. Throughout this paper AI,O, means a-alumina and TiO, rutile, and the specimen cornposition is expressed in mass per cent unless otherwise stated.
60
S. Taniguclii
3.2 TiAl-base Materials 3.2.1 Fundamentals of Oxidation It seems helpful to provide fundamental knowledge at the beginning briefly. More details are dealt with in recent reviews [8-121. As'TiAl has a large fraction of Al, one will expect to form a protective AI,O, scale on it. However, this is not the case in air or oxygen under atmospheric pressure at temperatures of interest (1 100 to 1400 K ) . I n general, the oxide scale during steady-state oxidation consists mainly of two layers [13-171 as shown in Fig. 1:an outer layer consisting mainly of TiO, grains and a porous inner layer which is a mixture ofTiO, and A1,0, grains.The inner layer contains many small pores almost uniformly distributed. Al,O, grains are enriched near the interface between the two layers. Although they cannot be sufficiently continuous, they can act as a barrier to some degree.The rate controlling step for the steady-state oxidation is ionic diffusion through the scale, provided that the outer TiO, layer is dense. The mass transport through TiO, is much faster than Al,O,.Therefore, it is effective to suppress the growth of TiO,, when a decrease in the oxidation rate is required. It was shown that the oxidation curve can be divided into three stages according to the oxidation rate and the scale structure 112,171. During stage I the scale is very rich in AI,O, and correspondingly the oxidation is slow. This is experienced during heating to a test temperature or during the initial period of oxidation at relatively low temper-
Fig. 1. SEM micrograph of a fractured section of n TiAl specimen oxidised at l3OOK for 100 ks in purified oxygen. showing the scale structure.
aturcs. However, there is the first breakaway to stage I1 where the oxidation is fast with a scale having a two-layer structure as shown in Fig. 1. The transition from stage I1 to stage IT1 takes place after a long period at high temperatures, where the Al,O, grains enriched near the interface dissolve in the outer TiO, layer and rcprccipitate near its outer surface.
3.2.2 Alloying Addition After an examination of a considerable number of relevant studies a conclusion was reached that in inany cases the selection of the additional element and its amount was based on vague foundation. The establishment of guiding principles is, therefore, now necessary for an efficient survey of more effective additives. The following three mechanisms can be derived as guidelines from previous studies and we can propose the fourth mechanism. i) Valence-control rule ii) Wagner’s scaling model iii) Formation of a barrier layer in the scale, and iv) Modification of the initially formed scale
3.2.2.1 Valence-Control Rule If the formation of TiO, was suppressed or minimized, then the situation becomes more favourable to form an Al,O,-rich or A120, scale. For this purpose the valencecontrol rule (VCR), or Wagner-Hauffc rule [lS], is applicable. As TiO, grows mainly by the diffusion of oxygen via oxygen vacancy in it [19], additives which can decrease the oxygen vacancy are very effective to decrease the overall oxidation rate. Of course, their solubility in ’MI2and their valence state should be known.
3.2.2.2 Wagner’s Scaling Model The suppression of internal oxidation of Al to form discrete AI,O, platelets in the substrate is also effective to make a continuous Al,O, layer on the substratc.Thc criterion for the transition from internal to external scaling was given by Wagner 1201 and was fundamentally based on the competition in diffusion between oxygen and A1 in the substrate. The important parameters concerned are the solubility and diffusivity of oxygen in the substrate, and the content and the diffusivity of Al in the substrate. However, the reported data for these parameters are few. The diffusivity measurement for intermetallic compounds is also relatively few [21,22].Shida and Anadu [23-251 tried to assess the oxygen solubility experimentally. A study by Perkins et al. [26] provides some knowledge for finding suitable elements on the basis of this criterion. I Iowevcr, the verification of this model is limited at present.
62
S. Tariiguchi
3.2.2.3 Formation of a Barrier Layer A small addition of Si was reported to be effective to decrease the oxidation rate by forming discrete SiO, aggregates in the scale near the scalchbstrate interface [27-291. The SiO,-rich layer can work as a barrier to some extent. As a result of this, AI,O, is enriched near the interface.This will further contribute to the decrease in oxidation rate. The addition of other elements such as Nb 1241 and M o [23] leads to a similar enrichment of their oxides after a certain time of oxidation. However. before this mechanism becomes effective a considerable amount of constituent elements is consumed. The enrichment of the additional element in the substrate also takes place [17,23,24,30] and this has also been claimed to decrease the oxygen solubility [23,24].
3.2.2.4 Modification of the Initially Formed Scale The addition of 0.2Zr [31] or 0.2Hf [32] to TiAl resulted in the formation of virtually A1203scales at and below 1300 K.The effect of these additives cannot be explained by the above three mechanisms. We propose the modification of the nature or stabilisation of the initially formed AI,O, scale as the fourth mechanism on the basis of our results and consideration. Another possible mechanism is the enhancement of the nucleation and growth of A1,0, grains by the addition, because these elements have large affinities to oxygen than Al and Ti. The resultant scale would be very rich in Al,03. Of course, a further study is necessary to examine these views. Figure 2 shows isothermal oxidation curves of TiAl and TiAl-0.2Zr in oxygen under atmospheric pressure [31]. Here, Am.A-' means the mass gain due to oxidation per surface area of the original specimen. It can be understood that the addition of 0.2Zr is very effective to decrease the oxidation rate at 1200 and 1300K. However, this effect disappears at 1350 and 1400 K. Thus, it can be said that the effect of Zr addition is limited to temperature up to around 1300 K. This temperature is, however, much higher than the maximum application temperature considered which is limited by the mechanical strength of TiAl-base materials at present.
Time , t I ks
Fig. 2. Isothermal oxidation curves of TiAl and Ti Al-0.2Zr in purified oxvgen under atmospheric pressure. .
3 Oxidation of Intermetullics - Jupanese Activiiy
63
Fig.3. SEM micrograph of a fractured section of a TiAl-0.2Zr specimen oxidised at 1300K for 100ks in purified oxygen, showing a very thin AI,O, scale.
The scales formed on TiAl-0.2Zr were very thin at temperatures up to 1300 K, as shown in Fig. 3. The figure shows an uniform surface layer of about 3.8 pm thickness. This layer consists of two parts; the outer layer of about 0.7 p,m thickness is virtually an Al,O, scale and the inner layer of about 3.1 pm thickness is the substrate where A1 is slightly depleted (so-called diffusion layer), seemingly for forming the A1,0, scale. A very small amount of TiO, is present on the Al,O, scale. Figure 4 shows the mass gain due to oxidation plotted against the Hf content. The addition of 0.2Hf is very effective to decrease the oxidation rate. However, further additions of Hf slightly enhance the oxidation at 1200K and increase the mass gain rapidly toward that of TiAl at 1300K. The scale on XAl-0.2Hf is virtually A1,0, and its structure is very similar to that shown in Fig. 3. Further additions of Hf increased the amount of TiO, in the scale and finally the scale structure became similar to that shown in Fig. 1. Judging from Fig. 4 there might be an optimal amount for the additive. It may mean that the excess amount of Hf has an adverse effect. For instance, they may be present as HfO, particles in the scale and enhance the oxidation, because HfO, allows relatively fast diffusion of oxygen. The addition of 0.2Zr and 0.2Hf improved the oxidation resistance significantly even under thermal cycle conditions with temperature varying between room temperature and 1300K for at least 100 cycle (100h). Figure 5 shows a summary of the influence of additional elements on the mass gain due to oxidation. Here, the ratio of the mass gain of ‘I‘iAl-X (X means additive) to that of TiAl is plotted against the group of the periodic table of element.The solid signs are based on other Japanese researchers’ [29,33,34], while open circles on our results. The results shown in other papers were also reviewed, but they gave similar knowledge.
S.Taniguchi
64
. .
f
100
a
0
2
4
Hf content , C / mass%
0.22
00.2H1
Group
Fig.4. Mass gain duc to oxidation plotted against the Hf content ofTiAl-IIf specimens oxidised for 100 ks in purified oxygen.
= I
5OOppmCI
* : mol ppm
Fig.5. Summary of the influence of additives on the mass gains due to oxidation ofTiAl-base specimens.
The addition of 2Cu, lY, 1.5Cr or 1.5Mn enhances thc oxidation.This can be partly explained in terms of VCR, if thesc elements are in solution in TiO,. If elements of group IV have a valence of fouf in X02,they may have little influence. In accordance with this view 2Sn, 2Zr and 2Hf showed a little influence. However. the explanation for the latter two should be different such that the excess amount of Zr and Hf may enhance the oxidation by forming ZrO, and HfO, particles in the scale and thus cancel the effect given by small additions of Zr and Hf. On the other hand, the effect of 1Si was attributed to the formation of a barrier layer as already discusscd. Regarding group V and VI elements, minor additions of P, Se and Te are effective to decrease thc oxidation rate and thc resultant scales are rich in A1,0, [33,35].This was explained again by VCR [33,35].There are optimal amounts for these elements. On the other hand, the influence of V, Nb and Ta belonging to group VA varies very wide-
3 Osidaliotz ofbirernietallics Jupmese Activiiy ~
40 N
E
I
0
i
I
I
I
I
1
1
65
I
Nb-free
6,
Y 0
. 0 7
k
E 20 a
.-C
(d
0)
rn
m
z
0' I
I
I
200
I
400
i
600
~
I
800
Fig.& Cyclic oxidation curves of TiAl and Nb-ion implanted TiAl (35 mole% Nb in a thin surface layer) with temperature varying between room temperature and 1200K in purified oxygen.
ly such that 2V enhances oxidation significantly, 2Nb decelerates it and 2Ta enhances it slightly. If V had a valence of two or three inTiO,, then its influence is attributable to VCR. The addition of Nb has been studied extensively [17,24,28,30,36-383. I h e important findings are i) decrease in the oxidation rate with an increase in the Nb content in a range 2 to 20%, ii) enrichment of Nb in the substrate and iii) formation of TiN [17, 30, 37, 391 and Ti,AlN 117, 391 in the scale near the scalehbstrate interface in air. When an excess amount of Nb was added,l'iNb,O, was formed 1381,which did not improve the oxidation resistance. Term i) was explained partly by VCR and by terms ii) and iii). On the basis of the above survcy surface application of Nb of a high concentration may be effective to enhance the formation of A1,0,. This hypothesis was confirmed by using an ion-implantation technique. The results are shown in Fig. 6 [40], where the Nb-implanted specimen with a maximum surface concentration of 35 mole% Nb shows an excellent oxidation resistance during thermal cycle with temperature varying between room temperature and 1200K for at least up to 10 cycle (200 h).The scale was virtually A1,0, and its structure was similar to that shown in Fig. 3, and did not show any spallation during the cyclic oxidation. The influence of Ta is not explained. The experimental results regarding the addition of Ta are very limited, so we need further study before coming to a conclusion. Anada and Shida reported that the influence of Mo [23] and W [24,25] is similar to that of Nb, but they put more importance on a decrease in the oxygen solubility for Mo addition and the completion of a continuous A1,03 layer in the scale for W addition. A surprising finding is that the addition of 500ppm CI to TiAl-2.5Mn remarkably decreased the oxidation rate as shown in Fig. 7 [34]. Here, IM means specimens produced by ingot metallurgy, while PM those by extrusion and subsequent reactive sin-
66
N
S. 7aniguchi 300 T I - 4 7 . 3 A I - 1 .7Mn 1223Kx86.4ks
"0
0.01
0.02
CI content
0.03
0.04
0.05
mass%
Pig.7.Variation in the mass gain due to oxidation ofTiA1-1.7 mole% Mn with the C1 content [34].
tering. The resultant scale was virtually AI,O,. If CI ion replaced oxygen ion in the TiO, lattice having valence of - 1, the oxygen vacancy will be decreased according to VCR.'I'his will finally decrease the oxidation rate. The vapourisation of titanium chloride was suspected as another mechanism, because it will enrich Al in the surface layer. F and Br were reported to have a similar effect as C1[34].
3.2.3 Surface Treatments 3.2.3.1 Pack Cementation Traditional aluminising using a pack-cementation process has been performed for Ti [41], Ti alloys [41] and TiAl [4143] to form a TiAI, layer on the substrate. The scales formed on the TiAl, layer, however, showed limited protection against oxidation. Breakaway oxidation starts after certain protective periods, which vary according to the nature of the Al-rich layer formed. The main disadvantages arising from the aluminising are (1) relatively rapid diffusion of A1 and Ti, and ( 2 ) crack formation in the TiAI, layer. The formation of a brittle TiAI, phase during the treatment or annealing after it was also reported [4143]. The rapid diffusion causes impoverishment of Al in the TiAI, layer leading to breakaway oxidation, because the Al,O, scale formed is not maintained anymore. I t also causes Kirkendall voids near the coating/substrate interface during the treatment and/or oxidation.This will reduce the adherence of coating to the substrate. In order to prevent the rapid diffusion a thin Ni layer was electroplated on a TiAl specimen before aluminising [44]. However, this trial was not successful, because the diffusion of Ti through the nickel aluminide layer formed was rather fast. Brittleness of the TiAI, and TiAl? layers was pointed out [43] as a reason for thcir mechanical failure. Large cracks were formed at specimen corners owing to the development of internal stress by aluminising [43].
3 Oxidation of Intennetallics -Japanese Activity
67
Application of chromising is very few and the study on siliconising cannot be found. Chroniising developed a diffusion zone in the substrate, which was porous and thus not protective [45]. However, since TiAl-base alloys containing large amounts of Cr showed good oxidation resistance [26], the enrichment of Cr in a surface layer seems very effective and hence further studies are expected in this respect. A similar thing can be said to siliconising, because Si is an effective additional element [27,28]. Electroplating of Cr of 20pm thickness was found effective at 1173K for at least up to 360 ks [46]. A combination of aluminising and preoxidation was also tried and found that this combination showed some success [43].
3.2.3.2 MCrAlY Coating Plasma spray coating of NiCrAlY [45] of 100 and 200 pm thickness, and CoNiCrAlY [46] of 50 to 60km thickness on TiAl decreased the oxidation rate. This effect was slightly increased when the treatment was performed under a reduced pressure. The effect was, however, not so large and the mass gain due to oxidation of the coated specimens was still larger than that of Inconel 713C. It is uncertain whether a definite A1,0, scale was formed on the coating and maintained for a certain period. Contrary to the above, the application of a fine-grain Co-30Cr-4Al film of about 30 Fm thickness to TiAl coupons by magnetron sputtering resulted in Al,O, scales on the coating [47].The oxidation rate was much more decreased as expected.Three oxidation stages can be recognised in the oxidation curves: initial transient, parabolic and accelerated stages. The acceleration is attributable to the local growth of TiO, grains on the AI,O, scales. The TiO, grains were connected to the coating through cracks in the scale like pegs. Only exception is that parabolic oxidation continued for more than 800 ks at 1100K with a sound A1,0, scale on top of the coating. This temperature may be the maximum use temperature for TiAl-base materials. The coating consisted of columns with the long axis almost normal to the coating/ substrate interface. Each column consisted of grains of about 0.2 pm in size. There were micropores in each column. The recrystallisation of the coating during oxidation led to the accumulation of micropores at the scale/coating and coatinghubstrate interfaces. Kirkendall effect arising from preferential diffusion of Co into the substrate was thought to be additionally contributing to the void formation at the coating/substrate interface. A small addition of Y to the coating resulted in flat A1,0, scales [48] as has often been discussed in terms of the so-called reactive element effect. However, the protectiveness of the coating was not improved rather it was slightly decreased. A variation in the nature of the coating by the Y addition may be a reason for this.
3.2.3.3 Ceramic Coating An excellent oxidation resistance was obtained when TiAl specimens were coated with an Si,N, film of 0.5 pm thickness prepared by ion-beam enhanced deposition as shown in Fig. 8 [49].Thc mass gain due to 30 cycle (600 h) oxidation with temperature
S. Tarriguchi
68
7
$6, 100 1
Y
5
0
E
-0-
0
I
I
5
10
Cycle
\
15
'
3
Fig.8. Cyclic oxidation curves o f X A l coupons coated with Si,N, films by ion-beam-enhanced deposition, film thicknesses (1) 0.5, (2) 1 and (3) 2 km. (4) is for an uncoated TiAl specimen.
varying between room temperature and 1300K is quite small and its change is also very small. However, this effect decreases as the film thickness increases up to 2 pm. The excellent oxidation resistance obtained is attributable to the formation of a layer rich in AI,O, and silicon compound beneath a thin outer TiO, layer in the scale during an early oxidation period.The nitride film was incorporated in the scale, but it was not confirmed whether it was oxidised to SiO, or remained as nitride fragments. The nitride film formed by this method was found to be amorphous or microcrystalline [50].Therefore, crystallisation or recrystallisation would take place during heating to test temperature or the initial period of oxidation. This structural change induced partial fracture of the film during an early oxidation period. The degree of fracture was found to be larger for the thicker film [49]. The application of sol-derived SiO, coating of around 0.2 pm thickness resulted in good oxidation resistance up to 1200K [Sl].At 1300K the coating was protective only yer was formed beneath the for short periods. Again in this cas coating which is incorporated in the scale a t oxidation period. An AI,O, film of 3 p m thickness formed a1 vapor deposition was also very protective at hmperatures between 10 for at least l00ks. during which mass gains were negligible [52].The sp was almost intact, except that at 1273K irregular-shaped small Al,O, c Applicatmn of films of CaTiO,, SrTiO,, BaTiO, and AI,TiO, of several p m thicknesses prepared by hydrothermal process or hydrothermal-electrochemical process improved the oxidation resistance of TiAl to some degree [S3]. In general, ceramic coatings having structural instability experience structural changes owing to crystallisation or recrystallisation during heating or early stages of oxidation. These structural changes induce small cracks in the coatings. However. as far as these cracks are small, the access of oxygen to the substrate is slow enough to
3 Oxidnrion of Interrnetallics - Jnpunese Activiiy
69
allow enrichment of ALO, beneath thc coating by relatively fast diffusion of Al in the substrate. This prevents the internal oxidation of Al, favouring the formation of a continuous AI,O, layer beneath the coating.
3.2.3.4 Preoxidation The fact that the difference in the affinity to oxygen between A1 and Ti is small is thought to be a reason for that no A1,0, scale is formed on TiAI. Selective oxidation of A1 would take place under a very low oxygen potential. Preoxidation of TiAl specimens under such a potential resulted in excellent oxidation resistance. A vacuum pump was used in one case [43,54], while usage of oxide powder pack [55-581 gave much better results. I n the latter method TiAl specimens were packed with oxide powder and the whole was encapsulated in a silica tube under a vacuum of 1.3 X lo-, Pa and heated at 1200K for 100ks. Chemical reagents Cr203,SiO, (a-cristobalite) and TiO, powders were used for packing.The scale thickness after prcoxidation and after 20 cycle (400 h) oxidation at 1300K are summarized in Fig. 9. It is noteworthy that the scale thickness decreases as the dissociation pressure of the oxide powder decreases. The preoxidation in S O , or TiO, powder pack resulted in very thin scales, virtually A1,0, scales, while that in Cr,O, powder pack thicker scales but very rich in A1,0,. All the scales grew very slightly during the subsequent cyclic oxidation. A combination of a CaTiO, coating and subsequent preoxidation under a low oxygen partial pressure resulted in very protective scales [59].
after preoxidation after cyclic oxidation
0 Cr,O,
SIO,
Ti02
Pack composition
Fig.9. Change in the scale thickness before and after 20 cycle oxidation at 1300 K, varying with oxide powder used for preoxidation.
70
s.Tanigudzi
3.2.3.5 Presulfidation It was quite interesting to observe the sulfidation behaviour of TiAI. Nrrriru et al. [60] found that the sulfidation of XAl in a gaseous mixture of H,S and H, results in the formation of titanium su1fides.A~the sulfide scale grew by the outward diffusion ofX. A1 was enriched on the substrate surface as TiAI, with a small amount of TiAI, as shown in Fig. 10 1601. After removing the sulfide scale they oxidised the specimen to show a very excellent oxidation resistance. A similar idea may apply to nitridation of TiA1, if preferential nitridation of Ti takes place. TIAI, (+TiAlz)
inner scale
outer scale
=:4.5 ._ v1
5 3 J+ f 1.5
Distance
Fig. 10. A cross section of a TiAl specimen sulfidised at 1173K for 86.4 ks and EPMA profiles for the relevant elements, showing the formation of aTiAI, layer on the substrate [60].
:.y 3.2.3.6 Mechanical Polishing and Grain & k i n g Mechanical polishing, sand blasting and grain refining are techniques sometimes used to enhance the diffusion of the scale forming element. The oxidation behaviour of TiA1 specimens subjected to electropolishing, mechanical polishing and sand blasting was studied [61]. At lOOOK the mass gain due to oxidation was much smaller for the mechanically polished and sand blasted specimens than the electropolished. However, the difference became small as the oxidation temperature was raised and at 1173 K there was no appreciable difference among them. The maximum effective temperature may be limited by the recrystallisation behaviour of the materials by which the lattice defects introduced by the mechanical polishing and the sand blasting disappear. The influence of microstructure was also studied [62,63]. Grain refining was found effective to decrease the oxidation rate to some degree below lOOOK, however above which the effect became very small.
3 Oxithion of Intermetallics Jopanese Ac1ivit.y ~
71
3.3 Nickel Aluminides The oxidation study on p-NiAI is very few in Japan in the last ten years or so. The influence of a small addition of Ti, Zr or Hf on the oxidation behaviour of Ni,AI and Ni,AI-O.lB was studied. It was found that these additives are effective to improve the resistance to cyclic oxidation by improving the adherence of the scale to the substrate as shown in Fig. 11 [64].The scale had a two-layer structure: the outer layer is NiO and the inner layer A1,0, which acted as a harrier against the oxidation. After a long period at high temperatures NiA1,0, appeared by the reaction of the both 1ayers.The addition of B was to improve the ductility of the materials and had a little influence on the oxidation bchaviour [64,65].
rn
.-L P)
40
-
-20
-
-40-
-1ooL
Time/ k s
Fig. 11. Cyclic oxidation curves of Ni,Al-O.lB specimens containing (1) non, (2) 1.6Ti, (3) 2.2Zr and (4) 1.9Hf, in purified oxygen. Duplicate runs for showing repeatability. Inset is the temperature pattern for one cycle
3.4 Iron Aluminides Since a general understanding is that the strength of iron aluminides is lower than that of other aluminides of interest, the relevant research is relatively few. However, an excellent resistance to oxidation and corrosion was reported for iron aluminide of p-FeA1 type by Okada et al. [66] together with their mechanical properties at high tempcratures. Fe-40A1-0.01B,Fe-47A1, Fe-49A1 and Fe-53A1 (all in mole %) specimens were subjected to a wear test where they were rotated in a bed of AI,O, powder at temperatures between 1023 K and 1323K for 3600 ks.Their mass losses due to wear and oxida-
s. TtrlligLlChi
72
tion in air were very small and more than one order smaller than those of SUS304 and SlJS310 steels. Their corrosion losses in a molten salt of 85 YO V,O, and 15 YO Na,SO, at 1123K were found also very small [66].
3.5 Nb,Al-base Materials Materials based on Nb,Al are referred to as a candidate for structural materials for extremely-high-temperature application because of their excellent high-temperature strengths at around 1800K. However, their oxidation resistance is a matter of concern, because no A1,0, scales were formed. Under rising temperature conditions the oxidation of Nb,Al starts at about 1200K and the oxidation products are Nb,O,, NbO and AINbO, [67]. In order to cope with the poor oxidation resistance various alloying additions were tried and a part of the results is shown in Fig. 12 [67].The additions of Re and Mo are effective to decrease the mass gain due to oxidation at 1173 and 1473K to some degree, however no A1,0, scales were formed. At 1773K the specimens resulted in the complete oxidation. Accordingly, various surface coatings were also tried [67]. There was a slight improvement in the oxidation resistance, however it was quite difficult to provide a sufficient oxidation resistance around 2000 K. There is a design in which an A1,0, coating is to be applied with an insert layer to mitigate the thermal stress. The measurement of thermal expansion coefficient of relevant materials is of considerable value for the development of a protective coating system. Results of the measurements of the coefficient are summarized in Fig. 13 [67]. However, successful coating system was not developed yet.
4 E
6,
Nx
15
-
w
I900°C
Binary Nb-xat.oiAl
Ternary Nb-25at.VA-5at.’%X
3 0-tidation o,f 1nrc.rrrretallics- Japanese Activity
-
73
11 10
X ‘ 0 9
2
‘ 8
.3
.E 7
2
g 6
w
5 4
I
1000
.
1200
,
1400
.
,
,
1600
Temperature/C
,
1800
F‘ig.13. Thermal expansion coefficients of Nbb ase materials [67].
3.6 Silicides MoSi, has a high melting point (2293 K) and a relatively low density (6.24 X lo3kg m-’), and exhibits a brittle-to-ductile transition at about 1200 K. It has a long-time carrier as a heating element because of its outstanding oxidation resistance. Its wide application as a high-temperature structural material has been discussed. For this purpose it is necessary to improve its ductility and oxidation resistance including the prevention or minimisation of pest phenomenon. One of the disadvantages accompanying the usage of MoSi, is the partial spallation of an SiO, scale due to its structural change during cooling. In order to circumvent this, the influence of a small addition of Al to MoSi, was studied by Yanagihara et at. [68].Their major results are as follows. The oxidation rate is increased by the addition of Al, however the overall oxidation resistancc is improved by preventing the spallation of the scale under thermal cycle conditions. Below the eutectic temperature, 1868K, in the SO,-mullite system AI,O, scales are formed.The oxidation rate is determined by the diffusion of Al through the Mo,Si, layer which is formed beneath the scale. Above the eutectic temperature a liquid scale consisting of SiO, and AI,O, is formed. The oxidation rate is controlled by the inward diffusipn of oxygen molccules. The oxidation behaviour of refractory metal disilicides, NbSi,, TaSi,, MoSi, and WSi,, produced by hot isostatic pressing was assessed at 1773 K in purified oxygen with a very rapid heating rate to the test temperature by Kurokawa et al. [69]. MoSi, and WSi, showed an outstanding oxidation resistance such as n o measurable mass change for at least about 180 ks, bccause of the formation of a protective glass-like SiO, layer. However, no protective scales were formed on the other disilicides. During the oxidation at a heating rate of 10 K.min-’ to 1473 K all thc specimcns except MoSi, showed rapid mass gains due to oxidation as shown in Fig. 14 [69]. Further, isothermal oxidation of solid solution of MoSi,-50 mole% WSi, at 1173 K followed a parabolic rate law. The oxidation behaviour was not much influenced by
P'
E
. + Do
2 c
€ a
0.01
.-mc
[5)
v)
I = o
I
I
1000
I
I
I
I
I
I
I
I
1200 1400 Temperature / K
I
~
1
Fig.14.Oxidation kineticsofrefractorysilicides during heating to 1473K at a rate of 0.17 K.s-' (lOK.min-') [69].
the presence of WSi, which was added to strengthen the material. However, when the oxidation was performed at a lower heating rate vapourisation of W 0 3 took place. There have been several studies where reinforcements were added to MoSi,. however a suitable balance between the mechanical property and the oxidation resistance has not been attained yet.
3.7 Concluding Remarks The main part of this review was devoted to the influence of additives on the oxidation behaviour of'TiAl-base materials, since they are most extensively studied in recent years. Four guidelines have been explained and discussed on the basis of experimental results for selecting the useful additives. However, it seems that more than one mechanisms are operating simultaneously for a few additives and their relative significance varies as the oxidation proceed. For example, a certain enrichment of an additive is necessary before it becomes effective. It is quite difficult to assess the contribution from each mechanism. A further study is necessary with a different view or technique to obtain a new information or concept. In addition, it is clear that fundamental parameters are lacking such as the oxygen solubility and diffusivity of relevant elements in the substrate as well as in the scale. A considerable effort should be paid for assessing the activities of A1 and Ti, and the influence of additives on them. The compatibility with the mechanical property and production requirement should be studied toward more efficient application of TiAl-base materials. Some of the coating methods shown here are effective to provide sufficient oxidation resistance to TiAl-base materials. Further studies should, however, be carried out before practical application, because the experimental conditions so far studied are
3 Oxidation of Iritermetnllics - Jupanese Activity
75
far from practical ones. For instance, structural components are subject to varying mechanical load and thermal cycles with rapid temperature changes simultaniously. In addition. practical environment may contain more corrosive constituents. Our preliminary experiments revealed that the presence of water vapour or carbon dioxide in the oxidising gas enhances oxidation significantly. Therefore, the influence of these gases must be clarified and a suitable measure should be taken to avoid their influence. Study on the potential silicides is relatively few. Therefore, much more effort should be paid on it.The useful knowledged obtained for the oxidation of TiAl should be examined and applied to save the experimental labour.
3.8 Acknowledgements ‘The author is very grateful to Prof. H.J. Grubke of Max-Planck Institut fur Eisenforschung, and the organisers and sponsors of the EFC Workshop on “Oxidation of Intermetallics” for the invitation. He is also very grateful to Japanese colleagues who kindly sent copies of their papers and allowed the reproduction of a few of their figures.
3.9 References [l]Proc.TMS Workshop “Oxidation of High-temperature Intermetallics”, Ed. by T. Grobstein, J. Doychak, (1989). [2] Proc. JSCE Symp. “High-temperature Oxidation of Intermetallics” Ed. by S. ‘laniguchi, (1991). [3] Proc. “Intermetallic Compounds” JIMIS-6, Ed. by 0.Izumi, (1991). [4] Proc. 3rd Japan Int. SAMPE Symp. “Intermetallic Compounds for High-Temperature Structural Applications” Ed. by M. Yamaguchi, H. Fukutomi, (1993). [5] Proc. JSCE Symp. “Corrosion Resistance of Intermetallic Compounds at High-temperature Environment” Ed. by A Takei, (1994). [6] H.A. Lipsitt, D. Shechfman, R.E. Schafrik: Mct.Trans 6A (1975) 1991. [7] K . Kawabata, 7:Kanai, 0.Izumi:Acta Met.33 (1985) 1355. [8] G.H. Meier: Ref. [l],p.l. [9] S. Taniguchi: Bull. Japan Inst. Metals 31 (1992) 497. [lo] G.H. Meier, ES. Pettif, S.J. Hu: J. de Phys. IV Colloq. C9 (1993) 395. 1111 S.Taniguchi: MRS I3ull.15 (1994) 32. [12] A . Rahmel, WJ. Quadakkers, M. Schiitze: Mater. Corros. 46 (1995) 271. [13] K . Kasahara, K . Hashirnoto, H. Doi, T Tsujirnoto:J. Japan Inst. Metals 53 (1989) 58. [14] Y Urnakoshi, M . Yarnaguchi, 7:Sakagami, 7:Yamane:J. Mater. Sci. 24 (1989) 1599. 1151 S. Taniguchi, 7:Shibata, S. Itoh: Mater.‘Trans. JIM 32 (1991) 151. [16] Y Shida, H.Anada:J. Japan Inst. Metals 55 (1991) 690. [17] S. Becker, A. Rahmel, M. Schorr, M . Schiitze: Oxid. Met. 35 (1992) 425. [18] K . Hauffe: Prog. Metal Physics 4 (1953) 71. [19] I? Kofstad “Nonstoichiometry, Diffusion, and Electrical Conductivity in Binary Metal Oxides”, R. Krieger Pub.Co., Florida (1983). [20] C. Wagner:Electrochem. 63 (1959) 772. [21] M. Koiwa, H. Nakajirna, K . Ifoh:Bull. Japan Inst. Metals 28 (1989) 723. (221 S. Kroll, H. Mehrer, N. Stolwijk, C. Herzig, H. Rosenkranz, G. Frornmeyer: Z.Metallkde. 83 (1992) 591. [23] H. Anuda. Y Shida: J. Japan Inst. Metals 58 (1994) 746.
76
S.Tuniguchi
[24] Y.Shida, H. Anadn: J. Japan Inst. Metals 58 (1994) 754. [25] H.Anada, Y.Shida: J. Japan Inst. Metals 58 (1994) 1036. [26] R.A.Perkins. K. 7:Chiurig, G.H. Meier: Scr. Metall.21 (1987) 1505. [27] K. Kosnhura, K. Hushimoto, H. Doi, 7:Tsiijimoto:J. Japan Inst. Metals 54 (1990) 948. [28] K. Maki, M. Shioda, M. Sayashi, 7:Shimizzi,S.Isobe: Mater. Sci. Eng. A153 (1992) S91. [29] Y.Shida, H. Anadn: Corros. Sci. 35 (1993) 945. [30] U Figge, A. Elschner, N. Zheng, H. Schuster, WJ. Qiiadakkers: Fresenius J. Anal. Chem. 346 (1993) 75. [31] S.Taniguchi, 7:Shibatu, fl. hiso: Proc. 8th Intern. Met. Mater. Congr. (1995). p. 435. [32] S. Taniguchi, H. Juso, 7: Shibatu: Mater.Trans. JIM 37 (1996) 245. [33] 7:Hunarnuru. Y.Ikeniutsu, If. Morikawa, M. 7ariino. J. Takamura: Ref. (31, p. 179. 1341 M. Kiimagai, K. Shibiie, M. Kim, M. Yonetnirsu: to be published. [35] Y.Ikernatsii, 7: Hanamiira, If. Morikawa, M. Tanino, J. Takumurn: Ref. (31.p. IY1. [36] D.W McKee. S.C. Hunrig: Corros. Sci.33 (1993) 1899. [37] H. Nickel, N. Zheng. A . Elschner, WJ. Quarfakkers: Mikrochim. Acta. 199 (1995) 23. [38] M. Yoshihara. K. Miura: Intermetallics 3 (1995) 357. [39] 7:A. Wallace, R.K. Clark, S.N. Sankarun, K. E. Wiedemann: Proc. “Environmental Effects on Advanced Materials” Ed. by R.H. Jones, R.E. Ricker: (1991) p. 79. [40] S.Tuniguchi, 7:Shibata, 7:Saeki, H.-X. Zhang, X.-H. Liu: Mater.Trans. JIM 37(1996) 998. [41] A. Takei, A. Ishida: Rep.The 123rd Committee of JSPS,31 (1990) 327. [42] H. Mabuchi, 7:Asui, Y Nakayunza: Scrip. Met. 23 (1989) 685. [43] M. Yoshiharu, ’I:Suzuki, R. Tunaka:Tetsu-to-HaganC 77 (1991) 274. [44] K. Kusahara, A. Rikei, 7:Yamashita:J. Surface Finis. Society Japan 45 (1994) 428. [45] H. Fztrukawa: Kef. [2], p. 54. 1461 7:Shimizu, 7:likubo, S.Isobe: Mater. Eng. A153 (1992) 602. [47] S. Taniguchi, 7:Shibata, A? Ausanunza, H. Lou, E Wang,W Wu:Oxid. Met. 39 (1993) 457. [48] S. Taniguchi, N. Aasanrrma, 7:Shibata, E Wang, H. L,ou. W Wit:J. Japan Inst. Metals 57 (1993) 781. [49] S. Taniguchi, 7:Shibatu, 7:Yamadu, X . Liu, S. Zou: ISIJ Intern. 33 (1993) 869. [50] X . Liu, Y.Yu, Z. Zhen, W Huang, S. Zou, Z. Jin, M. Chang, S.Xu, S.Tuniguchi, 7:Shibatu, K. Nukamuru: Surface CoatingTechnol. 46 (1991) 227. [51] S. Taniguchi, 7:Shibata, A? Katoh. J. Japan Inst. Metals 57 (1993) 666. [52] S.Taniguchi, 7:Shibatu, K. Tukeuchi:Mater.Trans. JIM32 (1991) 299. [53] M. Yoshimuru,W Urrtshihnra, M. Yushima,M . Kakihanu: Ref. [4] p. 1471. [54] E. Kobuyashi, M. Yoshihura, R. Tanaku: J. Japan Inst. Metals 53 (1989) 251. [55] S.Taniguchi, 7:Shibata, S.Sakon: Zairyo-to-Kankyo 41 (1992) 453. [56] S. Taniguchi, T Shibata: Ref. [4],p. 1461. [57] S. Taniguchi, 7:Shibnta, A. Murukami: Oxid. Met. 41 (1994) 103. [58] S. Taniguchi, 7:Shibata, A. Murakami, K . Chihara: Oxid. Met. 42 (1994) 17. [59] M. Yoshimuru,W Urushihara, M. Yahima, M. Kakihana: Intermetallics, to be published. [60] 7: Yoshioka,7:Narita: Kep.The 123rd Committee of JSPS 35 (1994) 221. [61] K. Kasuhura, M. 7akeyaniu: J. Japan Inst. Metals 57 (1993) 1288. (621 Y Togawu, H. Umehuru: Ref. [2],p. 1. [63] K. Kasuhara, A. Tukei:J. Japan Inst. Metals 57 (1993) 544. [64] S. Taniguchi, I: Shibata: Oxid. Met. 25 (2986) 201. [6S] S.Tuniguchi, ?: Shibata, H. Tsuruoka: Oxid. Met.26 (1986) 1. [66] I. Okada, H. Kawai, K. Matsuntoto, 7:Hirota, K. Watanabe, K. Kuho: Ref. [ S ] ,p. 35. [67] 7:Fujiwara, ’I: Obuna: Kep.l’he 123rd Coinmittee of JSPS 36 (1995) 73. [68] K. Yanngiharn, T Maruyama, K. Nagata: Mater.Trans. JIM 34 (1993) 1200. [69] K. Kurokawa, 11. Matsuoka, 7:Nagai: Trans. Mat. Res. Soc. Jpn. 14A (1994) 255.
Part I1 Ni-Aluminides
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
4 The Oxidation of NiAl and FeAl H.J. Grabke, n4.W Brumm and B. Wagemann
In the earlier studies [l]the phase P-NiAl was found to be very oxidation resistant, forming a protective alumina scale in a wide range of temperature and oxygen pressure. However, in more recent studies [2-51 many critical features .were detected in this alumina formation and even accelerated attack by intergranular and internal oxidation was observed under special conditions. The kinetics of alumina growth on P-NiAI was investigated in the temperature range 700-1 400 "C [2], generally parabolic diffusion-controlled kinetics prevail but with time- and temperature-dependent changes in the values of the parabolic constant k,, caused by formation, transformation and growth of different alumina modifications. Therefore, an Arrhenius diagram results (Fig. l a ) with different lines for the growth of the various modifications,which were identified by X-ray diffraction (Guinier camera). In the temperature range 700-850°C y-Al,03 is growing at a relatively high rate. Between 875-925 "C k, increases during the first 10-20 h, indicating a change from y-Al,O, to B-Al,O, growth, which shows a typical needle-like morphology.The growth of 0-A1,0, is in the temperature range 850--1050°C followed by a transformation to a-Al,O, [2], accompanied by a distinct decrease of the k,-values. The a-phase forms the slow growing, dense layer, necessary for materials protection. Its formation can be favored by the presence of certain alloying additions to NiAI, e.g. Cr, Y and Ce, as described in other contributions in the workshop [6,7].The effect of Cr is also shown in Fig. lb, an Arrhenius plot which summarizes the results on NiAl, pure and alloyed with different concentrations of Cr. The data points for the NiAl-Cr alloys demonstrate a fast transformation of 8- to a-Al,03 already at 1000°C, however, the k, values for a-Al,O, are somewhat enhanced by the presence of Cr.This behaviour is explained by formation of hexagonal Cr,O, in the transient state, which favors a-A1203nucleation and fast 8- to a-A1203transformation. However, due to the favored nucleation, the a-Al,O, is more fine grained than on pure NiAl and the growth by grain boundary diffusion enhances the rate of a-Al,O, growth. The responsible Cr content was detected near the outer surface in the a-A1203layer [6,8].Thus, the formation of protective a-A1203can be steered by favorable alloying additions, however, its adherence is still a problem, since an indigenous mechanism of pore and void formation beneath the scale tends to decrease the area of coherence. Especially, in NiAl with Ni excess the Al
H.J. Grabke, M. W Brunzm und B. Wugernanti
80
temperature
-
1400
FC]
1000 900
1200
800
700
Y)
's
'' 2 2
lo-"
, ,
1~-16(
,
,
,
I
I
, I
5.5 6 6.5 7 7.5 8 8.5 9 9.5 10 10.5
a1
reciprocal temperature [104/K]
temperature
-
1400
1200
Pc] 800
1000 900
700
v)
Y
3 "
5 c
.I! -
v)
c
8 8 o w 5a
'
c
b)
-
-.
- T NiAI-34Cr A
NiAl-15Cr NiAl-7Cr
10-16 - Ni50A150, 5.5 6 6.5 7 7.5 8 8.5 9 915 10 10.5 I
8
reciprocal temperature [104/K]
Fig. 1. Arrhenius diagrams of log l/T.a) forpureNiA'. bj lines for pure NiAI. symbols for NiAl-Cr.
depletion by A1,0, growth causes rapid inward diffusion of Ni into the material [3,9]. The loss of Ni and A1 beneath the scale leads to void nucleation and growth at the oxide/metal interface (Fig. 2). This mechanism can be further favored by the presence of S in the NiAl which would decrease the surface energy of the void, stabilizing it and accelerating cavity formation [10,11].The A1 depletion was calculated and demonstrated [12].The question was tackled, how the A1 is supplied to the growing A120, in the case of a scale widely undermi&d by cavities and it was shown [3] that at high temperatures of about 1200°C the supply by evaporation through the cavity is sufficient. whereas at lower temperature, e. g. 900°C other additional transport mechanism must contribute, probably surface diffusion in the cavity. The void formation most probably occurs not only below the scale but also at other heterogeneities, e. g. grain boundaries and this effect may allow oxygen ingress at grain boundaries and intergranular oxidation of NiAl [4]. 'The ingress of oxygen is possible
4 The Oxidation of NiAl and Fen1
b)
NI-AI
A1203
NI-AI
81
A1203
Fig. 2. Void and cavity formation beneath the A1,0, scale on NiAI. a) SEM photo of the surface of Ni50Alsoafter 40 h oxidation at 1200"C, partly covered with oxide (lower part) and partly spalled area with smooth cavities and with ridges, which had been connected to the oxide; b) schematics of the mechanism of void and cavity formation.
when due to impurities in the gas phase and on the NiAl n o protective scale is formed, e.g. in the presence of SiO(g) in quartz ampoules at low PO, pressures established by Cu-Cu,O, Fe,O,-Fe,O, or else [4,5] and in the presence of H,S in oxidizing H,-H,O atmospheres [13]. By Auger electron spectroscopy the ingress of oxygen in grain boundaries of NiAl could be followed, fracturing specimens after different periods of exposure under the conditions described [4]. The increase of oxygen concentration at the grain boundaries is uniform over the specimen cross section and linearly time dependent, which indicates that the entering of oxygen into the grain boundary is rate controlling. The intergranular oxidation leads to precipitation of A1,0, along the grain boundaries but not to 'pesting', i.e. disintegration of the material as in the case of NbAI, [ 14.15,16]. obviously since NiAl is less brittle and can tolerate the stresses caused by the precipitates at the grain boundaries.
82
H.J. Grabke. M . W Brunim and B. Wagernann
The intergranular oxidation can be followed or accompanied by internal oxidation of NiAI. The internal oxidation must be initiated by a rapid oxidation causing considerable Al-depletion but without formation of a protective scale [5].Then oxygen can diffuse into the NiAl matrix and reacts under formation of inward growing A1,0, precipitates. I n the ampoule experiments with various rnetal/oxide or oxide/oxide mixtures the attack is mostly localized and inward and outward growth of pocks or cones is observed on the material (Fig. 3), but also general attack can occur, especially in the
Fig. 3. Local internal oxidation of NiA1, starting froni a contaminated spot and leading to inward and outward growth of a double-cone (pocks). a) SEM micrograph of a single pock,grown on NiAl at 827°C and 2 X 10 bar 0, (Fe,O,-Fe,O,); b) schematics of the mechanism of pock growth, cross section showing the phases formed and concentration profiles of 0 and Al.
3 The Osidation of NiAl and FeAl
83
presence of H,S [4,5,13].These observations demonstrate that NiAl can be subject to accelerated oxidation under special conditions and is not oxidation resistant under any conditions. Rut it can be noted that NiAl with high A1 content and also multiphase alloys, e. g. NiAl-Cr, NiAl-NbNi Al and NiAl-NbNiAl-NbAl, are much lcss susceptible to the intergranular and internal oxidation attack described [17].This is caused for the alloys from the system Ni-Nb-A1 by the fact that in high temperature equilibrium the component NiAl always contains the maximum Al content, which prevents most mechanisms of oxidation described. Thus. NiAl is still an intcresting and important base material, for application as structural high-temperature material or as a protective coating. I n fact, many coatings such as Nicralloy, are based on the phase p-NiAI and the oxidation resistance of these multiphase alloys is well established.
4.1 References [l] I? S Pettit:Trans. Met. Soc.AIME239(1967) 1296. 121 M. W Brumm, H. J. Grabke:Corros. Sci. 33 (1 992) 1677. 131 M . W Brumm, H. J. Grabke:Corros.Sci.34 (1993) 547. (41 M. W Brumm, H. J. Grabke, B. Wagemann:Corros. Sci. 36 (1994) 37. (51 H. J. Grabke, G. H. Meier: Oxid. Metals 44 (1995) 147. [6] J. H. W de Wif,R. Klumpes, E. Schramm, 7:Marie: Materials and Corrosion, im Druck. 171 I. Rommerskirchen, I/: Kolarik: Materials and Corrosion, im Druck. [8] M. Riihle, E. Schumann, (1. Salzberger: unveroffentlicht. 191 S. Shankar, L. L. Seigle: MetalLTrans. 9.4 (1978) 1467. [lo] H. J. Grabke, D.Wiemer, H. Viefhaus:Appl. Surf. Sci. 47 (1991) 243. [ l l ] H. J. Schmutzler, €I. Viefhaus, H. J. Grabke:Surf. Interface Analysis 18 (1992) 581. [12] M. Boberh, E. Bischoff; E. Schzimann, M. Rockstroh, M . Riihle: Corros. Sci. 37 (1995) 657. (131 B. Schramm, W.h e r : Materials and Corrosion, im Druck. 1141 H. J. Grabke, M. W.Brumm, M. Steinhorst: Fresenius Z.Anal. Chem. 341 (1991) 378. [l5j H. J. Grabke, M. W. Brumm, M . Steinhorst, B. Wagemann: J. Physique 111, Volume 3 (1993) 385-393. [16] M. Steinhorsr, H. J. Grabke: %. Metallkde 81 (1990) 732. [17] H. J. Grabke, M. W. Brumm, M. Steinhorst: Mater. Sci.Technol.8 (1992) 339.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
5 Sulfidation Behaviour of Nickel Aluminides B. Schramm and W Auer
5.1 Introduction Nickel aluminides are of great interest in high-temperature applications because they combine high tensile strength with oxidation resistance even at high temperatures. Rencwed interest has arisen because of improvements in their mechanical properties in the past. Therefore many investigations have been published on the oxidation behaviour of the intermetallic phases y’-Ni,Al and P-NiAI during the last years [1-4]. In atmospheres with sufficiently high oxygen pressures these compounds are well protected by a layer of A1,0,. If, however, the oxygen pressure is low, then as reported by Grubke et al. [3],rapid inner oxidation occurs, similar to the so called “pest” which was observed, for example, in the intermetallic compound NbAI, [5]. In many engineering applications, such as coal gasification and coal liquification systems, structural components are exposed to mixed atmospheres, at very low oxygen activity but appreciable sulfur activity. In such environments sulfidation is often thc predominant mode of corrosion attack. There are, however, only few investigations dealing with the high temperature corrosion of nickel aluminides in sulfidizing or mixed atmospheres [6-91. Godfewsku et al. [6] investigated the sulfidation of pseudobinary P-NiAl-a-Cr alloys in sulfur vapor at 900 and 1000°C. Under thcse conditions sulfide scales formed on P-NiAI consist of an outer layer of liquid Ni,S, and an inner solid AI,S,-rich layer.Tlle scales are not protective and the corrosion rates are high. He and.Dougl@s.~.[S] studied the corrosion behaviour of y’-Ni,Al and P-NiAI in H,H,S-II,O gas mixtures over the temperature range 600 - 1000°C.y’-Ni,Al forms thin, non continous outer scales of nickel sulfide and a mixed inner layer of A1,0, and Al,S, containing some Ni. The P-NiAl forms thin protective Al,O, scales in this environment. In consequence of the sparsc information it is the purpose of the present work to contribute to the understanding of the processes involved in the sulfidation of intermetallic Ni-A1 alloys. Further we want to clarify the influence of low oxygen pressure on thc sulfidation process.
86
B. Schrarnni and W.Auer
5.2 Experimental Procedures 5.2.1 Materials Four Ni-AI alloys with different Al contents (25,30,36 and 45 at.% Al) were investigated in the present work.The alloys with 25 and 45 at.% Al are single phasc y'-Ni,Al or P-NiA1, while the other alloys contained both intermetallic phases. 'Ihc nickel a l u m nides containing 25.30 and 45 at.% Al were made of Ni (99.9 %) and Al (99.9 YO),arccast under an argon atmosphcre and homogenized for 36h at 950°C in vacuum. Ni36A1 was a hot isostatically pressed powder alloy. Coupons of 20*10*2 nim' were cut from the buttons by electro-discharge machining and then polished using 1200 grit Sic-paper. Before the samples were inserted into the system they were cleaned in ethyl alcohol and aceton.
5.2.2 Sulfidation Tests The sulfidation tests were performed measuring the continous mass gain in flowing gas mixtures of H,-H,S (0,l-10 vol.%) with and without water vapor at tcmperatures between 750 and 950 "C. The gas components were dried using phosphorous oxide before introduced into the systcm in order to lower the oxygen partial pressure since earlier tests had shown that sulfidation was often prevented by growth of A1,0, scales. At the gas outlet a zirconia probe was used to control the oxygen partial pressure. Sample heating was not started before the oxygen partial pressure of thc gas mixture bar at 900°C. respecbar at 800°C or reached its minimum value, which was tively. For tests in moist atmospheres, the hydrogen was passed through gas washers containing distilled water at 15°C in order to obtain water vapor saturation.
5.2.3 Analytical Techniques After sulfidation the scales were examined by X-ray diffraction analysis. In some cases the scale was detached from the metallic,Substrate, so that the inner scales as well as the metal-scale interface could be examined. Other specimens were mounted and metallographically polished in cross sections without using lubricant because A& tends to hydrolysis. These specimens were examined then and inspected using optical and scanning electron (SEM) microscopy as well as X-ray energy dispcrsive analysis (EDX).
5 Sitlfidntiori Behaviour of Nickel Aluminides
87
5.3 Results 5.3.1 Sulfidation in H,-H,S 5.3.1.1 Kinetics Sulfidation tests of the different alloys were carried out over the temperature range 750-950°C. Plots of the square of mass gain vs. time in different H,-H,S atmospheres at 900°C are shown in Fig. la,b,c.'Ihe kinetics of Ni36Al and Ni45A1 follow the parabolic rate law with rather high corrosion rates. In gas mixtures containing less than 0.5 vol.% H,S no sulfidation but alumina growth was observed. The kinetics on the Ni25Al alloy deviates from the parabolic rate law at 900"C, in atmosphercs containing less than 10 vol.% H,S a fast initial stage is followed by a slower final stage (Fig. la). Fig. 2 shows that the aluminium content of the alloy has only a slight influence on parabolic sulfidation rate. The influence of temperature on the sulfidation kinetics of Ni36A1 at a constant sulfur partial pressure of 6.4x1k7 bar is shown in Fig. 3a.The sulfidation kinetic follows the parabolic rate law over the whole temperature range and the sulfidation rate increases with increasing temperature. From the Arrhenius plot of the parabolic rate constants for sulfidation of Ni36AI in Fig. 3b we calculate an apparent activation energy of 58 kJ/mol.
5.3.1.2 Morphology of Sulfide Scales Fig. 4 shows the surface of a Ni36A1 specimen after 96 h sulfidation at 800°C in a gas mixture containing 5 vol.7'0 H,S.The surface is covered by globular Ni,S, particles solidified from an originally liquid layer. Both XRD and EDX analysis show that these spheres are Ni3S,. Tests at temperatures between 800 and 950°C in gas mixtures containing more than about 1 vol.% H,S produced the same type of attack. A cross-section of the scale formed on Ni36Al after 24 h sulfidation at 900°C in H,10 vol.% H,S is shown in Fig. 5.The bright particles at the surface are Ni,S,. Under this outer scale a thick dark inner scale has formed consisting of a mixed Ni-Al-sulfide, with only about 10 at.% Ni. XRD analysis of this inner scale detached from the specimen, did not indicate either AI,S, or Ni,S,. EDX-analysis revealed a homogenous composition of NiA13,5Ss.sconfirming the presence of a single phase mixed sulfide. The metal surface zone is Al-depleted by inner sulfidation and consists of the 7'-Ni,Al phase and Al-rich sulfide precipitations. The cross-section in Fig. 6 shows a scale formed in H,-1 vol.% H,S.'This mixture corresponds to a sulfur partial pressure which is not sufficient to form an outer Ni,S, scale. Consequently the scale consists of NiA13,5S5,5 only. The Al-depleted metal surface zone is thicker now probably due to Ni-diffusion backwards into the matrix. The scale formed on Ni25Al after 35 h at 900°C in gas mixture containing 10 vol.% H,S (Fig. 7) differs markedly from that formed on the alloys with a higher A1 content.
U. Scliramiir and W.h e r
88 3000 2500
2000 1500
1000 500 0 0
15
10
5
20
25
Fig. la. Parabolic plot of mass gain vs. time for Ni2SAl;T = 900°C
200
Fig. lb. Parabolic plot of mass gain vs. time for Ni36AI:T - YOO'Y:
so
Fig. lc. Parabolic plot of mass gain vs. time for Ki4SAl;'l - 900 '('
time ( h )
a) h
"E
x ; W
c . E
N
1500
1OO0
h
500
Q
W
0
time ( h )
b)
0
c)
150
100
50
0
10
20
30
time ( h )
5 Siilfidation Beliavioicr of Nickel Aluminides
c
3
100
80 -
Y
40
-
20
-
n
0
-
E i?
N
89
CI 0
60-
0 ,1
30
20
50
40 0
at%*'
10 Vol.%,9Oo"C 5 vol.%,9oooc
Fig. 2 Influence of Al-content on sulfidation rate of NiAl at 900°C
h
"E
Y
N
8
v N
n
4
E
\
d
W
0
10
20
30
40
50
60
70
80
90
100
time ( h )
Fig. 3a. Parabolic plot of mass gain vs. time for Ni36Al at constant ps2 = 6.4X lo-' bar and different temperatures
T in "C h
E
"E0 \
N
M I
E
W
Ni36AI apparent activation energy: 58 kJ / mol W-4
9.oe-4
1/T in 1/K
io,oe-4
Fig. 3b. Arrhenius plot of parabolic rate constants for Ni36Al
90
D.Sclzranim and W Auer
Fig. 4. Ni,S2-spheres on the surface of Ni36AI after 96 h suIfidation:T=8UOoC: 5 Val.% H2S
70
60 50
40
30 20 10 0
100
200
300
distance / pm Fig. 5. Cross-section of the scale formed on Ni36Al after 24 h sulfidationT-900°C. 10 Val.% H,S and composition profile; region 1:matrix; 2:Al-depleted zonc with Al-rich sulfides: 3: Ni-Al mixed sulfide; 4:nickel sulfide;
5 Siilfidntion Behnviotrr of Nickel Aluminides
91
20 10
0 0
100
200
distance / pm
300
Fig. 6. Cross-section of the scale formed on Ni45Alafter 45 h sulfidationT=900"C, 1 Vol.% H,S and composition profile; region i:Ni-A1 mixed sulfide;2:Al-depleted zone with Al-rich sulfides; 3: P-NiAI matrix
Fig. 7. Cross-section of the scale formed on Ni25AI after 35 h sulfidation;T=900"C; 10 Vol.% H,S; region 1:y-Ni,Al matrix; 2: Ni-A1 mixed sulfide; 3: nickel sulfide
92
B. Schratnm and W Auer
The outermost layer now is thick and relatively compact Ni,S, and the inner layer consists of mixed sulfide, containing large Ni,S, precipitates. Below the scale in the metal no Al-depletion and no internal sulfidation was observed.
5.3.2 Investigations in H,-H,S-H,O Atmospheres 5.3.2.1 Corrosion Morphology Water vapor in the testing atmosphere changes the corrosion attack dramatically. This is seen in Fig. 8 showing a cross-section of a Ni36A1 specimen after 500 h at 900°C in a gas mixture with 1 vol.% H,S and at an oxygen partial pressure of only about lo-,’ bar (900°C). Only a thin alumina layer forms at the surface of the specimen but further sulfidation is completely inhibited. In atmospheres containing water vapor up to a partial pressure of 17.5mbar, corresponding to an oxygen partial pressure of bar at 9OO0C,the corrosion morphology changed to a severe attack by rapid internal oxidation of Al. The Ni36A1 specimen in Fig. Ya was completely destroyed by this process.The cross-section in Fig. 9b s h o w large amounts of alumina (dark phase) between the remaining y-Ni,Al (bright phase). As illustrated in Figs. 10a and b the attack on Ni45A1 is quite similar. except for the formation of even more a-Al,O,. ‘The cross-section in Fig. 11 shows a Ni25AI specimen after 500h under the samc conditions. However, the dark inclusions are not a-A1203,but Al-rich sulfides, this was repeatedly verified. The earlier stages of the attack on Ni36A1 are shown in Fig. 12. Local pock like attack occurs from the surface of the specimen. On top of these pocks, spherical particles of nickel sulfide are often found. The cross-section in Fig. 12 shows the dark alumina phase growing into the matrix. At the front of attack the 0-NiAI phase of the alloy changed to y’-Ni3Al because of the A1 consumption. In this area small Al-rich sulfides were found.
Fig. 8. Cross-section of Ni36AI after 50011 at 900°C; 1 VO~.?’O II$; po2 lo-?’ bar;
5 Sirlfidation Behaviour of Nickel Ahtminides
93
Fig. 9a. Survey of Ni36AI after 500 h at 900°C; 1 vol.% H,S; po2 = bar;
Fig. 9b. Cross-section of specimen in Fig. 9a;
The cross-section of a Ni30AI specimen in Fig. 13a shows a morc uniform distribution of this type of attack and smaller amounts of alumina. As more closely shown in Fig. 13b, the attack is quite similar to the case of Ni36Al.The aluminium consumption at the front of the corrosion attack causes a phase transition from P-NiAI to y’-Ni,Al and again small Al-rich sulfides were found here.
5.3.2.2 Kinetic Measurements The influence of H,S content in the H,-H,S-H,O mixture on the mass gain of Ni36Al is shown in Fig. 14.After an incubation period of about 50 h the mass gain begins in the gas mixture with thc higher H,S concentration and the rate accelerates rapidly with time. In the atmosphere with 1 vol.% H,S the acceleration is much less. Considering the localized character of attack, the break away type may be explained either by an increase of the true local reaction rate, or by an increase of the numbcr of places of local
94
B. Schrcirnrn arid W Aicrr
Fig. 10a. Survey of Ni4SAI after SO0 h at 900°C: 1 V O ~ . % H2S po2 = lo-*"bar;
Fig. lob. Cross-section of specimen in Fig. 10a;
Fig. 11. Cross-section of Ni2SAI sample after 500 h: T=900"C: 1 Val.% H2S; poz = lo-'" bar:
5 Szilfidntion Behnviour of Nickel Aluminides
95
Fig. l2a Surface of Ni36A1 sample after 100h at 900°C; 1Vol.% H2S;po2 = bar
Fig. 12b. Cross-section of specimen in Fig. 12a;region 1:A1,0,: 2: y’-Ni,Al; 3: P-NiAl; 4: Al-rich sulfides
attack with increasing H,S concentration. From our data we are not able as yet, to decide which of these possibilities applies o r if both occur. Fig. 15 shows the mass gain vs. time curve in a test where the H,S flow was changed during the testing time. After 100h, when H,S flow was switched off, the mass gain almost immediately stopped, but restarted when H,S is switched on again showing that H,S plays an essential role in the mechanism of accelerated internal oxidation of nickel aluminides. From Fig. 16 it is seen that the mass gain of Ni30Al is much lower than in the case of Ni36A1 apparently due to the lower A1 content.Also there is no incubation period and the mass gain follows an approximately linear rate law, nearly independent of the H,S concentration. Again, however, the reaction rate drops off wherever the H,S flow is switched off (Fig. 17).
96
B. Schramm arid W Aiier
Fig. 13a. Cross-section of Ni30AI after 100 h at 900°C; 1 Vol.% I-12s;po2= 10 bar
Fig. 13b. Detail of Fig. 1.h; region 1:A1,03; 2: y'-Ni,Al: 3: P-NiAI;4:Al-rich sulfides
0
50
100
150
time ( h )
200
250
Fig. 14. Influence of H,S-content on the kinetic of internal oxidation of Ni36Al at 900°C; po2 = 10 2'1 bar
5 Sitlfidrition Brhavioztr of Nickel Aluminides 25
E
u
3
W
E
B €
97
~
20 15
5
Fig. 15. Influence of changing H,S content on mass
0
50
0
100
200
150
250
time ( h )
gain caused by inner oxidation of Ni36AI;T 900°C; po2 = bar; 7
n
E
CI
Y
P
W
+
2Vol.SbH2S
3
2
1
Fig. 16. Influencc of
0 lo
2o
30
50
6o
time ( h )
'O
loo
H,S-content on the kineticsofinternaloxidationof Ni30A1 at 900 "C; po2 = bar;
5.4 Conclusions 5.4.1 Sulfidation of Nickel Aluminides in H,-H,S Atmospheres 1) The sulfidation kinetics of the investigated nickel aluminides except Ni25A1 approximately follows,the parabolic rate law in H,-H,S atmospheres between 750 and 950°C. 2) In these atmospheres on the nickel aluminides double layered sulfide scales are growing composed of an outer Ni,S, scale and an inner mixed sulfide with an composition of NiA13,sSs,s.At low H,S contents (51 vol.%) no outer Ni,S, is formed. Nickel aluminides containing more than 25 at.% A1 form an aluminium depleted zone consisting of y'-Ni,Al with inner Al-rich sulfides at the scale/ metal interface. 3) Variations in the Al-content of the alloy between 25 and 45 at.% A1 have nearly no effect on the parabolic sulfidation rate.
B. Schrrrmm und M! Auer
98
0
50
100
150
time ( h )
200
250
Fig. 17. Influence of changing H2Scontent on mass gain caused by internal oGdation of NijOAl: T = 900 "C; poz = 1O-'" bar;
5.4.2 Internal Oxidation of Nickel Aluminides in H,-H,S-H,O Atmospheres 1) Even a small amount of water vapor suppresses the growth of sulfide scales described above. 2) Nickcl aluminides containing more than 25 at.% A1 undergo severe attack by rapid internal oxidation of aluminium. y'-Ni,Al shows no internal oxidation but internal sulfides are formed. 3) H,S plays an essential role in the mechanism of the rapid inncr oxidation of A1 in nickel aluminides. H2S is necessary to start this oxidation mechanism as well as to propagate it.
5.5 Acknowledgement Financial support of this study by the Deutschc Forschungsgemeinschaft in gratefully acknowledged.
5.6 References M. W Brumm, 11. J. Grabke:Corros. Sci. 33 (1992) 1677. M. W Brumm, 11. J. Grabke:Con'& Sci. 34 (1993) 574. M . W Brumm, 11. J. Grabke, B. Wagemann:Corros. Sci. 36 (1993) 37. S. Tuniguchi,I: Shibatu:Oxid. Met.28 (1987) 155 M.'Steinhorst,H. J. Grabke:%. Metallkde.81 (1990) 732. E. Godlewska el at.: Mat. Sci. Engineering 87 (1987) 183. K . Natesan: Oxid. Met. 30 (1988) 53. Y R. he, D.L Douglass: Oxid. Met. 40 (1993) 337. 1/: S. Bhide, M! W Smeltzer: J. of Electrochem. SOC.128 (1981) 903.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
6 The Influence of Chromium on the Oxidation of P-NiAI at 1000°C R. Klumpes, C.H.M. Marbe, E. Schramm and J.H. W de Wit
6.1 Introduction Thc intermetallic compound P-NiAI is of great interest for protection of nickel-basc alloys at high temperatures as e. g. in turbine blades. It is applied as a surface alloy laycr and protects the base alloy by the formation of an aluminium oxide layer in an oxygen containing atmosphere. This oxide layer grows slowly and is selfhealing. Several types of A1,0, have been reported for alloys forming alumina layers [1-4]. At high temperatures the metastable (transition) oxides y, 6, and 8-A1,03 will transform to the a-AI,O, according to the following sequence [3,5]: y - Al,O, (cubic spinel)
L 6- A1,0, (tetragonal)
1 8 - AI,O, (monoclinic)
1 a-Al,O, (rhombohedral)
At 1000°C y- and S-AI,O, are rapidly converted into O-AI,O,, leading eventually to formation of an a-layer. The 8 + a-transformation is accompanied by a changc in transport properties of the layer [5].A 8-layer grows mainly by outward cation diffusion. In this way voids can be produced at the oxide-substrate interface, resulting in a deterioration of the adherence of the layer on the substrate. By contrast, the growth of an a-layer is determined by anion-diffusion, so that the previously formed voids arc fillcd with oxide. This improves the adhesion of the layer to the substrate. Thercforc, thc formation of an a-layer is preferred. The work of Rybicki and Srnialek [2]implies that the 8 + a-Al,O, transformation is a thermally activated process. According to Brumm and Grabke [6] chromium incorporated in the oxide layer, originating from the alloy, can initiate the nucleation of a-Al,O,. However, until now the presencc of chromium in the oxide layer has never been shown. In this study several P-NiAl alloys with different chromium contents were investigated on their oxidation behaviour at 1000"C using electron microscopy (SEM), Rutherford Backscattering (RBS) and X-ray diffraction measurerncnts
(XKD).
100
R. Klutnpes, C. H .M . Mar&, E. Schmnini and J.H. W dr Wit
6.2 Experimental The alloys were prepared by arc melting (at Philips Laboratory of Physics) and annealed in argon for 100 hours at 1000°C.Using a diamond wafering blade, samples were cut into 1mm thick slices. The diameter of the slices was 12mm. Subsequently the samples were ground and polished to 1 diamond paste. The oxidation experiments were performed at 1000°C in pure oxygen, which had been dried and purified with P,O, and inolsieve A4. The y-NiAl samples contained 0, 1 , 5 and 10 at % chromium respectively. ‘The degree of homogeneity was investigated with EPMA and XRF. The oxidized samples were investigated with X-ray diffraction (XRD) and Rutherford Backscattering (RBS).The RBS measurements were performed using a-particles with a kinetic energy of 2.87 MeV.The back scattered a-particles were detected with an S. S. B. detector at an angle of 10”. From this RBS spectrum information on the composition and thickness of the oxide scale can be obtained. For more details we refer Chu et al. [7]. In Table 1 the samples, used for the measurements, are shown. Both normal (perpendicular) as well as glancing angle beam angles were used.The advantage of a glancing angle (50°C) incidence is the spreading of the depth profiles, leading to higher depth resolution as can be seen in Fig. 1.The oxide layer on the sample, oxidized during 100 hours, is too thick to be analysed fully with a glancing angle beam, because the a-particles do not reach the substrate/ oxide interface. On the spectra curve fits were calculated with RUMP (RBS software developed at the Cornell University in Ithaca (N.Y.) [13]) rendering the composition and the thickness of the oxide layer. Also the variation in the thickness can be estimated with this procedure. Two possible situations are schematically given in Fig. 2a, b, but also intermediate situations can exist. These configurations result in a broadening of the interface edge of nickel. In order to calculate the thickness of the oxide layer and its variation “integrated Gaussian fitting” is used in RUMP for the spreading values of the nickel edge energy. For a meaningful procedure the following assumptions were made: - ‘fie composition of the oxide was independent of the depth. - ‘fie aluminium oxide was stoichiometric Al,O,. The morphology of the oxide layers was investigated with SEM.
Table 1. Survey of the samples oxidized at 1OOO”C investigated with RBS -
sample
A B C D
-__-
chromium content
oxidation time
(at.?”)
(h)
1
10 1 10 100
10 10 10
_
-
radiation beam perpendicular (0”)
_
+ +
i
~
+ + + ~
~
radiation beam glancing angle (50°C)
i-
_
_
_
~.
~
_
_
_
_
~
6 ?he Itiflicerice of Chroniiutn on h e Oxidation of P-NiAl at 1000°C
glnncing angle benm
perpendicular benm
L = (l+l/COSIOo)
x = 2x
101
L = (I/cosSO" +l/cos6O0) x = 3.6 x
sample
o . . . . . .
L = pathlength of a-particle in sample Fig. 1. Schematical RBS measurement: difference between a normal perpendicular and glancing angle beam.
substrate
substrate
Fig. 2. Waving interfaces leading to local variations in thickness.
6.3 Results According to the ternary phase diagrams of aluminium-chromium-nickel [7] chromium is soluble in P-NiAI up to about 4 at.% at 1000°C. This is in agreement with the EPMA results for these alloys. The 0 at 1 at.% Cr alloys were homogeneous; whereas, segregation was observed in the 5 and 10 at.% Cr alloys. In Table 2 the composition of the alloys, determined with XRF, is reported. In the EPMA-micrographs in Fig. 3 chromium enrichment (a-Cr) can be seen at the grain boundaries. The morphology of the oxide layer on the samples was investigated with SEM. It appeared that there were no major differences between the samples. In the early stage of the oxidation process the surface was covered with small nodules (Fig. 4a). After a
102
R. Klitnipes, C. H. M. Maree, E.Schranini and J.H. W.de Wit
Fig. 3. a.: 5 YOCr sample.The Cr Enrichment (a-Cr phase) at the grain boundaries is visible in the SEM picture (lower part) and the EPMA Cr mapping (upper part). @-phase:NiAI. b.: 10 % Cr sample.The CR Enrichment (wCr phase) at the grain boundaries is visible in the SEM picture (lower part) and the EPMA Cr mapping (upper part). P-phase: NiAL.
Fig. 4. a,: NiAl surface morphology after 1 hour oxidation at 1000°C, SEM taken at 45 degrees. b.: NiAl surface morphology after 100 hours at 100O"C, SEM taken at 45 degrees.
6 The Itiflurncr of Chrorniicni on the Oxidation of p-NiAl at 1000°C
103
l'able 2. Composition of the alloys (at.%) determined with XRF alloy NiSOA150 NiAI-1 %Cr KiAI-5 %Cr NiAl-1O%Cr
nickel
aluminium
chromium
phase
47.9 49.4 50.0 46.1
52.1 49.4 44.7 43.0
-
P
P
1.2 5.3 10.9
a, P a. P
longer period the surface was smoothened whilc ridges evolved resulting in a lacey structure as can be seen in Fig. 4b.This has been observed for pure NiAI [4,5,9,11]. The crystallographic structure of the oxide scale was investigated with XRD. An example of an X-ray diffractogram is shown in Fig. 5. The only phases identified in all these plots were P-NiAI, O-AI,O, and a-Al,O, [lo]. The 0-peaks are most difficult to distinguish because they overlap with the substrate @)-peaks or have a low yield. At lower temperatures (e. g. 800°C) the initial scale consists primarily of 0-A120,. In this way we were able to discriminate between the two phases reliably by using more diffraction lines [12]. Samples with comparable preferential orientation (110) were taken.Thus we could use the peak at 28= 33 degree as the base for our calculations. 'The oxide-peak heights were measured and the 8-peak-height fraction was calculated as a function of thc oxidation time. This only gives qualitative information, but provides an idea about the transformation rate. In Fig. 6 the results of these calculations have been given. In this plot four curves corresponding to the four alloys can be observed. It can clearly be seen that in the 5 and 10 YO Cr containing alloys already after short oxidation times (e. g. 2 hours) no 0-phase can be found anymore while the oxide layer on the pure P-NiAI alloy contains quite some 0-oxide during several hours. The 1 YOCr alloy lies somewhere in between these curves.
20
30
40 50 2*theta (")
60
70
Fig. 5. X-Ray diffractogram of a p-NiAI-1 YOCr sample oxidized for 5 hours at 1OOO"C
R. Kltcmpes, C.H. M. Marie, E. Sclirainm and J.H. W. rle Wit
104
0
2
6
4
oxidation time (h)
8
10
Fig.6.8-peak fraction as a function of the oxidation time at 1000°C on alloys with different chromium contents.
For the RBS measurements several samples of two alloys (1 and 10% Cr) were prepared with variable oxidation times (Table 1).The chromium content of sample A is very small. Therefore the chromium in the resulting oxide could not well be detected with RBS. In Fig. 7a a chromium edge can hardly be observed. It is rcmarkable that not only oxygcn and aluminium edges arc found at their theoretical surface energy level after oxidation, but a nickel edge as wcll.This is also the case with sample C. (Fig. 7b). On these samples small oxide free regions were obscrved with light microscopy. (These regions did not spread over more than 10% of the surface.) Both samples have a similar oxidation time (10 h). For samples with 10 YOchromium (B, C, D corresponding to Fig. 7c, b, d) all RBS spectra show clear chromium edges at the theoretical energy level for surface-chromium. For B and D this proves the presence of chromium in the oxide 1aycr.For C the oxide free regions may have obscured the information. Using a glancing angle a higher depth resolution can be obtained. An example of such a measurement for sample B has been given in Fig. 7e.The results show that there is no significant change in the chromium yield indicating that the chromium is distributed homogcneously in the oxide. Table 3.Thickncss (d) and variation (Ad) of the oxide layer on the samples sample A B C D
perpendicular beam d(pm) Ad (pm) 0.60 0.34 0.59 1.25
. ...
0.27 0.10 0.10 0.27
glancing anglc beam d (k) Ad (km) Cr (at.%)
0.54 0.31 0.59 -
0.25 0.11 0.12
-
8.9 ? 1.7 16.9 2 1.2 9.7 2 0.6
6 The Influence of Clzrotniiir17on the Oxidation of P-NiAl at 1000°C Energy [MeV) 0.5
1.0
1.5
2.0
I.o
50';~ -lox
c,.
Energy (MeV) 1 .s
I
'op
10s
hOY.
(0)
2.0 :,
i
I.
, . : I
, .
10
d
-
50
100
150
200
250
Channel Energy (MeV) 40
-
05
50
15
10
,
100
150
200
Channel
250
360
20
300
350
50
100
150
2w
250
Chonnel
1
1 .o
I
b
350
..
Energy (MeV)
m
Energy (MeV) 1.5
2.0
350
Je 1
3%
Chon n eI
C
Legend: measureddata -- - - simulation (calculated with RUMP) calculated element specific yield contributing to the simdation : theomtical element specific surface energy level
-
M
1w
150
200
Channel
2%
Ixx)
354
Fig. 7. RBS spectra, with calculated element contributions, and elemental surface edge energies: a: sample A, b: sample C. c: sample B, d: sample D, e: sample B, glancing angle.
From the RSB-results the thickness of the layers was calculated. The results of the integrated Gaussian procedure are given in Table 3. Also values of the chromium content are given assuming the presence of a homogeneous oxide layer.
106
R. Klurnpes, C.H.M.Marfie, E. Schrurnm und J.H. W de Wil
6.4 Discussion Despite the fact that some samples were homogeneous and others were enriched in chromium on the grain boundaries, the morphology of the oxide layers didn't vary very much. 'The nodules indicated outward growth, generally ascribed to the 0-phase. However, the formation of the lacey structure is more difficult to understand. although several papers gave models for the development of these structures [9,11,13]. Commonly this morphology is associated with the a-phase. This is in agreement with our previous studies 141. From Fig. 6 can be concluded that an increasing Cr-content in the alloy enhances the rate of formation of a-oxide from 0-oxide. The oxide free regions observed with light microscopy on the 10 hours oxidized samples imply a bad adherence of the scale of the substrate.The oxide layer had spalled off during cooling.The bad adherence is probably due to the formation of voids on the oxide / substrate interface [4]. On the RBS measurements these bare regions caused a nickel- and chromium yield at energy levels corresponding to the levels of surface atoms (samples A and C), but the chromium yield cannot fully be explained by these regions. The chromium yield is too high compared with the nickel yield (1:l instead of the ca. 1:4.5 in the substrate), so a part of it must be caused by chromium in the oxide layer. The other RBS measurements show more clearly that chromium is present in the oxide layer. They also show that the chromium has been distributed homogeneously throughout the scale.The variation of the thickness in Table 3 can be explained by the nodules at the surface in Fig. 4a. However, it cannot be excluded that a waving substrate/oxide interface contributed to this variation. These oxide free regions on sample C can also explain its higher chromium content as shown in Table 3. The values of B and D are more in agreement which each other. This can lead to the conclusion that the chromium content in newly formed oxide is similar. This explains the homogeneous distribution of the chromium in the scale. The combination of the RBS and XRD-results supports the theory of Rriinznz and Grabke [6]: chromium oxide acts as a nucleus for the 0 + a-transformation. It has been clearly shown that chromium is present in the scale and that the 0 + a-transformation rate was enhanced with increasing chromium content in the alloy also leading to an increasing chromium content of the scale.
6.5 Conclusion The oxidation behaviour of several P-NiAI alloys with a different chromium content (resp. 0,1,5 and 10 at. %) was investigated. RBS showed the presence of chromium in the oxide layer. The XRD results affirm a relationship between the presence of chromium and the rate of the 0 + a-phase transformation. The results support the model of Brumm and Grabke: chromia particles in the 0-oxide act as nuclei for this transformation. Chromium does not influence directly the morphology of the scale.
6 The Influence of Chroniii~ri~ o n the Oriclntiori of P-NiAl at 1000°C
107
6.6 Acknowledgement The authors wish t o thank N.J.M van de Pers for performing the XRD-measurements and his help with the interpretation. Many thanks to T. Kleinlzout and L. Korbijn for the determination of the composition of the alloys with XRF.
6.7 References [ I ] t? 7: Moseley, K. R. Hyiie. U.A Rellrirny. G Tappin:Corros. Sci. 24 (1984) 547 [2] G.C. Rybicki, J. L. Sniinlek:Oxid. Met. 31 (1988) 275 [3] J. Doychrik, J. L. Srnialek, 7: E. Mitchell:MetaLTrans. A20 (1989) 499 [4] P A . van Manen, G.WR. Leibbrandt, R. Kliirnpes, J.H.W de Wit:J. de Physique I V proc. 3rd Int. Symp. on High Temp. Corr. and Protecting of Mat Les Embiez (Fr.)ed. K.Streiff et al p. 123,1992 [ 5 ] P A . van Marien. C.W van der Wekken. D.Scha1koord J.H.W de Wit:Surf. int. anal. 12 (1988) 391 [6] M.W Urunzrn,H.J. Grabke:Corros.Sc.33 (1992) 1677 [7] W K . Chu, .I. W. Mayer, M.A. Nicolet: Backscattcring Spectrometry. Acad. Press. New York 1978 [S] P Rog1:Ternary Alloys. (4)ed. Petzow, Effenberg, Weinheim, 1991 [Y] R. Prescott, D.F. Mitchel, M.J. Graham: Corros. Sc. 50 (1994) 62 [ 101 JCPDS powder diffraction files,ASTM, Philadelphia, 1969 [Ill J. K. Doychak: Masters Thesis, 1984, Case Western Reserve University Cleveland (Ohio), alSO N.AS.A. CR-174756 [12] P A . van Manen: Ph.Il.Thesis, Delft University of Technology, 1991 [13] H.M. Hindurn, W W Srnelzer:J. Electrochem. Soc. 127 (1980) 1631
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
7 Oxidation of P-NiAl, Undoped and Doped with Ce, Y, Hf I. Rommerskirchen and K Kolarik
7.1 Introduction Since many years the intermetallic phases based on aluminides have been an important topic for research and developmcnt, because of their high melting points, low densities and excelent corrosion resistance at high temperatures Especially, nickel aluminides havc been of great interest as coating materials for several high temperature applications. The high temperature oxidation of P-NiAI and 7’-Ni,Al and its oxidation mechanism under various conditions as well as their oxidation kinetics have been the matter of various investigations and are quitc understood and described [l-91. Also the positive influence of rare earth elements on the corrosion resistance of nickel aluminides was studied and discussed extensively [l,3,8-13]. Rut there is still a need of more detailed information about the diffusion paths and the solubility of oxygen in P-NiAI. The purpose of this research work, therefore, was a contribution to a better understanding of these processes studying P-NiAI undoped and dopcd with Ce, Y and Hf. The rare earth element concentrations were chosen high enough that the material consisted of two phases, the binary matrix and a ternary phase, precipitated along thc grain boundaries and within the grains. The ternary phase precipitations were observed as a detector for the oxygen diffusion paths. Furthermore there is a need of detailed information about the formation of metastable alumina modifications during the isothermal oxidation of P-NiAI and about the influence of reactive elements o n the alumina modifications. At temperatures below 900°C and in the initial state the formation of y-Al,O, was reported by Doychak [14] on Zr-doped P-NiAI and on undoped P-NiAI by Brumm and Grabke [4].The initial formation of @-AI,O, on P-NiAI and its transformation to a-Al,O, at temperatures higher than 950°C was reported by Brumm and Grahke [4], Rybicki and Smialek [S] and Doychak 1141. A novel approach to the investigation of the formation of metastable A120, phases during the high temperature oxidation of P-NiAI was undcrtakcn using the time and temperatvsc resolved X-ray diffraction as an in situ method.The method allows the in situ identification of the oxides and thcir modifications, kinetic evalution of each reaction and it detects the lattice changes continuously and in situ. The high temperature X-ray diffraction is a very suitable method for the in situ investigation of corrosion processes and has been successfully applied in several studies [lS-201.
110
I. Rornrnerskirchen anti L! Kolarik 1
I
0
1
TI!---
1.P,O,-column 2. capillary flowmeter 3: two-way valve 4:thermocouple
5: sample 6: furnace 7: microbalance 8: writer
02
3
Fig. 1. Experimental set-up for thermogravimetry
7.2 Experimental The high temperature oxidation of P-NiAI, undoped and doped with Ce, Y and Hf was studied in situ by thermogravimetry in He with p(0,) = 5 . 10-hbar at 1000°C and by high temperature X-ray diffraction at 950 and 1000°C in air. After the in sitii experiments the samples were analysed metallographically by optical microscopy and by scanning electron microscopy (SEM) with energy dispersive analysis (EDX). The experimental set-up for thermogravimetry is shown in Fig. 1 . The sample is hanging o n a microbalance in the temperature constant zone of a furnace. The furnace temperature is controlled by means of a Pt-Rhl8 thermocouple. The oxidation gas consisted of commercial helium gas with a partial oxygen pressure of p(0,) = 5 . 1 0 P b a r . The gas was dried through a P,O, column and then lead to the furnace with a flowing velocity of 2 ml SKI controlled by a capillary flowmeter. The measuring system for the time and temperature resolved X-ray diffraction consists of an X-ray diffractometer and a high temperature device with a programmable temperature controller. Isothermal measurements and free selectable temperature programs can be performed between room temperature and 1600°C under oxidizing conditions. Series of X-ray diffraction'p'atterns with defined time intervals o r temperature steps are recorded in situ yielding the structural changes in the sample as a function of the time or temperature. Phase changes, formation of new products and dilatation as well as contraction of the lattice are detected continuously during the experiment. For kinetic evaluation of the oxide formation from a series of in sifir X-ray diffraction patterns the intensities of the oxide peaks are determined as a function of time by the summing method [19]. This procedure calculates the peak intensity summing the counts of each channel in the range of the peak. The underground is subtracted to eliminate the influence of changing underground intensity. The resulting curves i z ( t )
7 Oxidarion of P-NiAl, IJndoped and Doped with Ce, I:H f
111
for each oxide and its modification show their formation as a function of time taking into account also the absorption of the X-ray beam in the growing layer.
7.3 Results and Discussion 7.3.1 Characterization of the Samples Before the Oxidation The composition and microstructure of the samples was studied by SEM and EDX before the oxidation experiments. The compositions and concentrations of the studied nickel aluminides are listed in Table 1. The micrographs show that the concentrations of the dopant elements are high enough for a two phase structure, the binary matrix and a ternary phase. 'Ihc ternary phase precipitatcd along the grain bounaries as well as within the grains (Figs. 2 to 4). It was found by EDX analysis that the ternary phases consisted of Ni and the rare earth component with 10 to 20wt.% A1 (Tab.2).
7.3.2 Thermogravimetry and Post Oxidation Studies The mass gain curves from the thermogravimctry experiments at 1000°C are represented in Fig. 5. All investigated systems show a very fast initial mass gain corresponding to the transitional fast growth of metastable, non-protective alumina modifica-
Table 1. Composition of the P-nickcl aluminides ltd. chemical analysis in wt.%
p-NiAI P-NiAI+Hf P-NiAI+Y P-NiAI+Ce
A1
Fe
Ni
C
31.3 30.3 31.4 29.9
0.04 0.12 0.07 0.05
68.5 65.3 68.0 69.4
0.0034 0.0095 0.0097 0.0084
IIf
Y
Ce
0.0006 0.0008 3.9 0.0006 0.0006 -
-
-
-
0.85
S
0.39
-
Table2. Composition of the ternary phase Itd. EDX analysis in at.% before oxidation
along the grain boundaries
within the grains
48 '70Ni 20 Yo A1 29% Y 44% Ni 20% Al 36% CC
40 % Ni 10 YOA1 50% Hf 50 YONi 19 O h Al 30% Y 43 YONi 18YoA1 37 YoCe
p-Ni A1+ Hf P-NiAI +Y P-NiAI +Ce
112
1. Rommerskirchen and I/: Kolurik
Fig. 2. NiAl + Hf before oxidation, the ternary phase is visualized by the white spots
Fig.3. NiAl+Y before oxidation; the ternary phase is visualized by the white spots
Fig. 4. Ni Al t Ce before oxidation: the ternary phase precipitations are found along the grain boundaries and within the grains
7 Oxidiition of p-NiAl, Undoped and Doped with Ce, Y
Hf
113
tions. After longer oxidation times thc mass gain decreases in consequence to the transformation of thc metastable alumina phase to the slow growing a-Al,O,. The system P-NiAI+Hf shows a lowcr mass gain than the other studied materials. Thc mass gains for undoped P-NiA1 and P-NiAI+Y are comparable, but P-NiAl+Y shows the fast oxide growth in the initial state indicating the formation of a metastable alumina modification. For P-NiAI+Ce a significantly higher mass gain was observed indicating a faster oxide formation than in the other materials. The formation and conversion of thc metastable alumina modification is also indicated by the time dependent changes of the parabolic rate constant k,. From initially high values corresponding to the fast growth of the metastable alumina modification, thc k , values decrease to a lower level corresponding to the slow growth of the aA120, (Fig. 6). From Fig. 6 it is evident that the dopant elements have an influence on thc transformation period of the metastable alumina phase to a-Al,O,. The longest transformation period is observed for P-NiAl+Ce with 150h, followcd by P-NiAl+Y with lOOh and undoped P-NiAI with 50 h. The shortest transformation period is observed for P-NiAI+Hf with 25 h. The samples from the thermogravimctry experiments were analysed metallographically in order to explain the observed results. The optical micrographs from metallographic cross sections show that the ternary phase precipitations along the grain boundaries are oxidized in an increasing extend from P-NiAI+Hf over P-NiAI+Y to P-NiAI+Ce (Figs. 7 to 9). In the same order the oxidation rate increases in the thermogravimetric experiments (Fig. 5). In p-NiAI+Y and P-NiAI+Ce the ternary phase precipitations were oxidized along the grain boundaries only, and no oxidation within the grains was observed (Figs. 8 and 9). An element analysis of the oxides along the grain boundaries by SEM (EDX) showed that the rare earth oxides HfO,, Y,O, and CcO, were formed (Tab. 3).
x-xx7x-
x-x-x
n N
X
U
X-X
X
NIRI-C~
10
t
0 0
50
100
150
time Chl
200
250
Fig. 5. Mass gain curves of the thermogravinietry experiments in He with 5 . bar oxygen partial pressure at 1000°C
I. Rornnierskircheri arid V Kolarik
114
-0.5
0
50
100
150
time [hl
200
2 9
~0
R
Fig. 6. Time de endence of the parabolic rate constant k , with 5 . 10- bar oxygen partial pressure at 1000°C
Fig,7. P-NiAI+Hf after oxidation at 1OOO"C in He with 5 .
bar oxygen partial pressure
No internal oxidation of the rare earth additions was observed. indicating that no oxygen was dissolved in the material. Obviously, the oxygen enters into the material only along the grain boundaries and not through the bulk.
7.3.3 High Temperature X-ray Diffraction During the oxidation time of 70 h series of X-ray diffraction patterns were recorded at time intervals of 1h.The formation of each alumina modification is observed simultaneously and in situ (Fig. 10). The peak intensities of each oxide were determined as a function of time and plotted as i z ( t ) curves showing the growth of each observed modification.
7 Oxidation o f P-NiAl, Undoped iind Doped with Ce, Y, Hf
Fig.8. P-NiAI+Y after oxidation at 1000°Cin He with 5 .
115
bar oxygen partial pressure
Undoped P-NiAI forms at 950 and 1000°C in the initial state of the oxidation process @-AI,O, besides t h e stable a-AI,O,. The amount of @-A1,0, increases at 950°C up to 15h and then it decreases until the phase disappears after 40 h.The same effect is observed at 1000°C with 0-A120, diappearing after 5 h. During the existence of 0AI,O, the intensity curve i z ( t )for a-Al,O, shows a fast linear growth. The slopc of the lincar growth is at 950°C lower than at 100O"C,but due to the longer existencc period of @-AI,O, more a-Al,O, is formed. At 950°C the linear section of the iz(t) curve for a-A1,03 shows a change in the slope at the maximum presence of 0-AI,O, at 15 h. After the @-phase disappeared, the a-Al,O, shows a slow growth, like expected for a protective oxide scale (Fig. 11). P-NiAI+Hf shows a similar oxidation behaviour like undopcd P-NiAI forming also 0-A1,0, at 950 and 1000°C in the initial state of the oxidation process besides of a-
Fig.9. P-NiAI+Ce after oxidation at 1000°C in He with 5 .
bar oxygen partial pressure
116
I. Romrnerskirchen and V Kolarik
Fable3. Composition of the ternary phase Itd. EDX analysis in at.% after the oxidation along the grain boundaries
within the grains
4 % Ni 0 % A1 83 Yo Hf 15 % Ni 6 % A1 42 % Y 5 % Ni 1 o/o Al 76 % Ce
40% Ni 9 % A1 50% Hf 55 YONi 15 YoAl 29% Y 42% Ni 10 YoAl 44% Ce
P-NiAI+Hf P-NiAI+Y P-NiAI+Ce
AI,O,. At 950°C the amount of the O-phase increases until 9 h and then it decreases disappearing after 30h. At 1000°C the @-phase is observed during the first 7 h of the oxidation. During the existence of the O-phase the a-Al,O, shows a linear growth changing into parabolic after the disappearance of the 0-AI,O, (Fig. 12). At both tcmpcratures the formation of HfO, was observed during the whole experiment. P-NiAI+Y forms 0-AI,O, at 950°C in the first 5 h of the oxidation process and during its existence the a-Al,O, grows fast and with a linear time law. No @-A1,0, was
NiAl
+ Hf
950°C
2 0 [Deg.] Fig. 10.Time dependent series of X-ray diffraction patterns recorded during the oxidation of P-NiAI+Hf at 950"Cin air
7 Oxidation of P-NiAl, Undoped ntid Doped iviih Ce, ..
3CWO
Hf
117
-..
NiAl (undoped)
25000
.y -
2WW
a
15000
J
low0
5wo
0 0
10
20
30
40 t
50
M
w
70
Fig.ll. Growth of a-and @-A1,0, on undoped P-NiAI at 950 and 1000°C 2oooo 18000 16000 14oW
-a
12000
10000 BM)o
Mxxl
4WO 2oW
0
0
10
20
30
40 1 m1
50
60
70
Fig. 12. Growth of a-and 0-AI,O, on P-NiAl+Hf at 950 and 1000°C
detected at 1000°C and the growth of the stable a-phase indicates the formation of a protective layer. At 950°C the formation of a-alumina is stronger than at 1000°C due to the linear growth during thc coexistence with the @-phase (Fig. 13). N o yttrium oxide was detected by X-ray diffraction. P-NiAI+Ce shows an opposite behaviour to the system with yttrium. At 950°C no @-AI,O, was detected and a slow growth of a-Al,O, was observed. At 1OOO"C the 0phase was formed in the first 5 hours and during its formation a strong linear growth of the a-Al,O, was observed. After the disappearance of the @-phase the a-alumina shows a slow growth (Fig. 14).The formation of CeO, was observed during the whole experiment at 950 and 1000"C.The results are summerized in Table 4. For comparison the growth of the a-Al,O, in the different systems was plottcd together in one diagram. At 1000°C the p-NiAI+Hf and p-NiAl+Y show a comparable growth of a-alumina and on undoped P-NiA1 the growth curve lies slightly higher. An
I. Rommerskircheti and V Kolnrik
118
Table4. Periods o f presence of @-AI,O, on the invcstigated P-nickel aluminides at 950 and 1OOO"C Material
Temperature ["CI
p-NiAI illidoped
Presence period for O-Al,O, 40 h 5h 30h 7h 5h no @-Al2O3 no C-)-AI,O, 5h
950 I000 950 1000 950 1000 950 1000
@-NiAI+Hf
p-NiAIAY P-NiAI+Ce
Other oxides
~~
30000 25oW
-
2owo
p -1 P
-
1
m
ww WW 0 0
10
20
30
40 t
50
50
70
lhl
Fig.13. Growth of a- and @-AI,O, on P-NiAI+Y at 950 and 1000°C
-.
NiAl + Ce
0-yo, 9 M ' C
0
1
10
e-m,o, 950'c
20
30 t
m1
40
50
70
Fig. 14. Growth of a- and @-AI,O, on @-NiAl+Ceat 950 and 1000°C
7 Oxidation ofP-NiAl, Undoped and Doped wilh Ce, X a-A120,
119
1000°C
P 0
Hf
NiAl (""doped)
10
20
30
40
50
60
70
1 lhl
Fig. 15. Growth of a-Al20, on p-NiAI undoped and doped with Hf,Y, and Ce at 1000°C
0
10
m
30
40 t
50
€0
70
M
Fig. 16. Growth of a-A120,on P-NiAI undoped and doped with Hf,Y. and Ce at 950°C
enhanced formation of a-Al,O, is observed on P-NiAI+Ce, especially in the first 5 h of the oxidation,during the existence period of @-AI20,(Fig. 15). Fig. 16 shows the intensity curves i z ( t ) for the growth of a-Al,O, at 950°C. It is observed that the level on which the slow growth after the disappearance of @-AI,O, occurs, strongly depends on the first state with the linear growth of the a-phase. Undoped P-NiAI and P-NiAI+Y reach the same level for the slow growth period, but the duration of the fast growth period is different. Less a-alumina is formed on pNiAl+Hf and P-NiAI+Ce shows the lowest formation of a-A1,03, corresponding to the observation from Fig. 14 that no @-phase is formed. The intensity curves iz(t) for a-A120, from Fig.15 correspond to the mass gain curves for the total oxide scale in Fig. 8. Both thermogravimetry and in situ X-ray diffraction lead to the result that P-NiAI+Ce forms a thicker oxide scale than the other
120
1. Rommerskirchen and V Kolarik
investigated systems.The results from in situ X-ray diffraction, however, show that relatively small amounts of @-AI,O, are formed, and that a-Al,03 is growing fast during the existence of @-AI,O, (Figs. 11 to 14). Obviously, a-Al2O3is formed simultaneously by oxidation and transformation of 0-to a-Al,O, in the initial state of the oxidation. A fast formation of the metastable @-phase transforming quickly into a-Al,O, would explain its enhanced increase during the existence of 0-AI2O,.
7.4 References [1] J. Jedlinski, M.J. Graham, G. 1. Sproule, D. E Mitchell, G. Borchardt, A. Bernasik: Materials and Corrosion 46 (1995) 297. [2] M. W Brunirn, H. J. Grubke, B. Wagemann:Corros. Sci. 36 (1994) 37. (31 B. A. Pinf,L. W Hobbs: Oxid. Metals 41 (1994) 203. (41 M. W.Brurnrn, H. J. Grabke: Corros. Sci.33 (1992) 1677. [5] H. J. Schmutzler, H. Viefhaur,H. J. Grubke:Surf. Interface Anal. 18 (1992) 581. [6] H. J. Grubke, M. Steinhorst, M. W Brumm, D. Wiener:Oxid. Metals 35 (1991) 199. [7] J. Jedlinski, G. Borchurd Oxid. Metals 36 (1991) 317. [8] G. C. Rybicki,J. L. Srnialek: Oxid. Metals 31 (1989) 275. [9] J. Jedlinski, S. Mrowec: Matcr. Sci. Eng. 87 (1987) 281. [lo] E. W A. Young,J. H. W de Witt:Oxid Metals26 (1986). [11] H. M. Hindum, W W Srnebzer: Oxid. Metals 14 (1980) 337. [12] A.Rahmel, M. Schiitze: Oxid. Metals 38 (1992) 255. [13] C. A. Barrett: Oxid. Metals 30 (1988) 361. [14] J. K. Doychuk: Doctoral Thesis, Case Western Reserve University, Cleveland, Ohio USA (1986). [15] 1.: Kolarik, M. Juez-Lorenzo, W. Engel, N. Eisenreich: Proceed. UK Corrosion & EUROCORR 94 (The Institute of Materials, no 269275),Vol3 (1994) 378-387. [16] M . Juez-Lorenzo,V Kolurik, N. Eisenreich, W Engel, A. J. Criatio: Proceed. 3rd Europ. Powder Diffraction Conf. EPDIC-3, Mat. Sci. For. 166-169 (1994) 361-366. [17] M . Juez-Lorenzo, V Kolurik, N. Eisenreich, W Engel, A. J. Criado: Progress in Understanding and Prevention of Corrosion (The Institute of Materials),Vol2 (1993) 1129-1 135. [18] 1.: Kolarik, M. Juez-Lorenzo, W Engel, N Eisenreich: Fresenius J. Anal. Chem.346 (1993) 252. [19] 1.: Kolarik, M. Juez-Lorenzo, N. Eisenreich, W. Engel: Proceed. 3rd Int. Symp. on High Temperaturc Corrosion, J. Physique IV (Colloque C9) 3 (1992) 447-452. [20] M. Juez-Lorenzo, V Kolarik, N. Eisenreich, W Engel, A.J. Criado, E. Otero: Proceed. 9th Europ. Congress on Corrosion (Royal Netherlands Industries Fair. 1989) FU-273.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
8 TEM Investigations on the Oxidation of NiAl E. Schumann, J. C. Yang, M. J. Graham and M. Riihle
8.1 Introduction NiAI belongs to a group of intermetallics which show promise for high temperature applications due to a combination of desirable properties, such as low density, high melting point and good thermal conductivity. NiAl has a large potential for oxidation resistance, due to its ability to form a protective a-AI2O3scale and the large amount of Al present in the alloy. NiAl is the base metal for the most advanced oxidation coatings currently in use. In the present paper experimental results from two important aspects of the oxidation of NiAl are presented: (i) the transient oxidation of NiAl and (ii) the segregation of reactive elements (Y, Zr) during oxide growth. Thc paper summarizes work which has been published recently [l - 31. Transient oxidation of NiAl has been studied by Doychak et al. [4]using planar TEM. The authors report the growth of transition alumina phases, like y-Al,O,, 6-A120, and O-AI,O, in the early stages of oxidation. These metastable phases transform to a-A1203during longer oxidation periods. This phase transformation is reflected in a decrease of the parabolic rate constant kp by two orders of magnitude, as has been found by Rybicki and Smialek [5]as well as by Brumm and Grabke [6]. An important aspect of the oxidation resistance is the maintenance of a-Al,O, scales during long term oxidation especially under thermal cycling conditions. During cooling large thermal stresses arise which eventually leads to scale spallation. It is well known that the adhesion of a-Al2O3scales can be dramatically improved if alloys are doped with reactive elements (e. g.Y, Zr). However the mechanism of the reactive element effect is still under discussion. In order to further study this effect the segregation of reactive elements at the metal/oxide interface as well as within the oxide scale has been investigated using analytical TEM. In this paper results from two sets of experiments are described. In the first part results from a study of the early stages of oxidation are presented. These experiments have been performed using single-crystal NiAI, because of the epitaxial aspects, which are important for the growth of transition aluminas. In the second part results from segregation measurements on Y and Zr doped polycrystalline NiAl are reported.
122
8.2
E. Schumann, J. C. Ynng,M. J. Grrihnrn and M. Hiihle
Experimental
NiAl single crystals were cut by spark erosion with thc surface normal aligned parallel to the (001) pole within 1" as measured by Laue diffraction. The composition of the NiAl single crystal is 50.0 at 70Al doped with 0.01 wt YOY. Wet chemistry revealed 6 ppni S impurity in the sample.The single crystal NiAl samples were oxidized at 950 "C for 0.1 h or 50 h respectively in air. The polycrystalline NiAl samples contained 0.1 wt% Y or 0.2 wt% Zr.The S level in the alloy was 10 ppm. These two samples were oxidized at 1200 "C, first in 1 6 0 2 and then in '*02. The Zr-doped alloy was oxidized for 100 h in "02followed by 100 h in I8O2.The Y-doped alloy was oxidized for 12 h in "02and then 12 h in "02.SlMS investigation were performed in a Perkin-Elmer, Physical Electronics Industries Model 5500 system (see [7] for details). The preparation of TEM cross-sections includes mechanical thinning followed by ion thinning using a Bal-Tec ion mill. A detailed description of the TEM preparation can be found in [8]. Conventional TEM studies were performed on a JEOL 2000 FX operating at 200 kV. The analytical studies were performed on a VG HBSO1-STEM equipped with a field emission gun operating at 100 kV and an EDS ultra-thin window Ge-detector. High resolution investigations werc pcrformed using a JEOL JEMARM1250 operated at 1.25 MV,
8.3 Results In this section the experimental results are presented, according to the two sets of NiAl samples. First the results from the NiAl single crystal oxidized at 950 "C are shown.
8.3.1 Oxidation of Single-Crystal (001)NiAl at 950 OC 8.3.1.1 0.1 h Oxidation Figure 1 represents a bright field TEM micrograph of a cross-section of the NiAl/oxide interface after the NiAl was oxidized in air at 950 "C for 0.1 h.The oxide scale was observed to be a 40 - 50 nm thick layer all along thc NiAl surface. The scale contains many planar defects, possibly twin boundaries. Faceted 200 - 300 nni voids were observed with their facets parallel to (001)NiAI and (01 1)NiAI planes. A thin oxide layer was found at the facetted metal surfaces of some of these voids. EDS measurements revealed that the oxide scale consists of AI,O, containing less then 1 at % Ni.The NiAl was doped with a small amount of Y (0.01%). EDS measurements in the STEM have been performed however no Y could be detected within the oxide scale or at the interface.The detection limit has been calculatcd t o bc 0.05 monolayer. Figure 2 is an electron diffraction pattern taken from thc NiAlloxide interface. The main spots are due to the [loo] zone axis of NiAI. Extra spots can be indexed as the
Fig. 1. Cross-sectional ' E M image of the NiAl/y-Al,O, interface showing facetted voids after oxidation for 0.1 h at 950 "C.
Pig. 2. Electron diffraction pattern of the NiAlly-Al,O, interface (figure 1). Some of the diffraction spots of y-Al,O, are indexed.
124
E. Schurnann, J. C.Yang, M. J. Graham and M. Ruhle
Fig. 3. HKEM image from the (01l)NiAI/(11 l)y-A1203 interface (figure 1).
[110] zone axis of y-Al,O,. Some of the diffraction spots have been indexed in figure 2.The overall orientation relationship of the y-Al,O,, with respect to thc NiAl was dctermined to be:
(001) (100lNiAI I1(001) [llO]y-Al,O, The presence of long facets and the high degree of epitaxy between the NiAl and the oxide scale suggested high resolution investigations. Figure 3 is a HREM micrograph of the interface between a (O11)NiAl-facet and the oxide.The two crystals have an excellent relative orientation since all lattice planes, including the (440)y-A120, with a lattice spacing of d = 0.139 nm are resolved. The following orientation relationship could be revealed for this (011)NiAI facet: (011) [100]NiAI II (111) [llO]y-Al,O, The interface is atomically flat. The lattice mismatch of the (112)y-AI20, and the (O11)NiAI planes is calculatet+to'be 17%. However no misfit dislocations or stand-off distance were observed at this interface.
8.3.1.2 50 h Oxidation Figure 4 shows a SEM image of the oxide scale after 50 h of oxidation at 950 "C,revealing the platelet-shape morphology on the oxide surface. Figure 5 represents a crosssectional TEM image of the oxide scale and the interface. A high number of planar defects was observed in the oxide scale. Electron diffraction revealed that the orienta-
8 T E M Investigations on the Oxidarion of NiAl
125
Fig. 4. SEM image of the oxide surface after oxidation for 50 h at 950 "C showing the platelet morphology.
Fig. 5. Cross-sectional 1 E M image showing the presence of y-Al,O, and a-AI,O3after 50 h of oxidation at 950 "C.
tion relationship between y-Al,O, and NiAl was the same as that found after 0.1 h of oxidation. Oxide grains which gave bright uniform contrast were observed at the interface. Electron diffraction from these grains revealed that their identity is a-Al20,. The a-Al,O, grains appear bright since these grains are not in a pole direction. No preferred orientation has been found for the a-Al,O, grains. The a-Al,O, grains were observed to be irregular and growing into the transient oxide scale, the bulk NiAl and along the interface.The oxide next to the a-Al,O, grains is y-Al,O, as determined from electron diffraction.'I'he size of these grains is of the order of 200 nm.
126
E. Schumann, J. C. Ynrzg, M . J. Grahani nriif M. Riihle
Fig. 6. HREM image of an oxide platelet showing the presence of (lll)y-A1203twin boundaries.
Figure 6 represents a high resolution image from a typical platelet from the surface of the oxide scale. The electron diffraction pattern can be indexed as [I lO]y-Al,O,. The HREM image shows the presence of twins within the platelets.llis is reflected in the streaks along [lllly-Al,O, in the diffraction patten (figure 7). Approximately 10 platelets in two TEM samples were examined by NRLM, and all showed twinning along ( 1 l l ) y - AI,O,. The twin boundaries were always observed to be parallel to the growth direction of the platelets.
8.3.2 Oxidation of Polycrystalline NiAl at 1200 "C The oxidized NiAl samples doped with Y or Zr have been first investigated by SIMS and then by TEM. The results of the SIMS measurements can be found in [7]. In the following the results of the microstructural and analytical studies with TEM are presented.
8 TEM Itwestigations on the Oxidation of NiAI
127
Fig. 7. Electron diffraction pattern from an oxide platelet (figure 6) showing the presence of streaks along [lllly-Al,O,.
8.3.2.1 NiAl
-
0.2 wt YOZr
A TEM cross-section of the NiAUoxide interface is shown in figure 8. The oxide scale consists of a-AI,O,, which was determined by electron diffraction. The thickness of the oxide layer is approximately 3 - 4 pm. Near the oxide/gas interface small Zr02 particle were observed. N o Zr-rich particles were found at the metal/oxide interface. Also no voids are present at the metal/oxide interface. The concentration of Zr as a function of the distance from the metal/oxide interface was measured with STEMEDS.The dark line in figure 8 is a schematic drawing of the EDS line profile. Segregation of Zr at the metal/oxide interface can clearly be seen (figure 9). Using a quantification technique proposed by Ikeda et al. [9]. the amount of segregated Zr was calculated to be 0.15 - 0.1 monolayers. Figure 10 represents a bright field TEM image from a grain boundary within the aA1,0, scale. STEM-EDS measuremcnts were performed across the grain boundary as shown by the dark line in figure 10. Segregation of Zr at thc a-A1203grain boundary
E. Schumann, J. C. Yung, M. J. Graham and M. Riililv
128
can be seen (figure 11).The amount of Z r was determined to he 0.2 - 0.1 monolayer. Z r was found at all the oxide grain boundaries examined (about 30 grain boundaries). N o sulfur was detected at the metal/oxide interface or in the alloy. The detection limit of S is 0.1 monolayer.
Fig.8.Cross-sectionalTEM image from NiAl - 0.2 wt% Zr after 200 h oxidation at 1200 O C .
L
N
E
3 0-0I-"-"-"-"
z
Interface L
NiAl
c
.a ul
c 0
, 200
0
150
A'2°3 i
4
-40 - 3 0 - 2 0 - 1 0 0 Dlstance
from
10 Interface
20
30
40
(nm)
Fig. 9. STEM-EDS measurement across the NiAl/a-AI20, interface (NiAI-0.2 73).
S TEM Investigations on the Oxidation of NiAl
129
Fig. 10. TEM image of an a-A120, grain boundary (NiA1-0.2 Zr).
Grain Boundarv
2 0.08 i
N
E
0.07 0.06
0)
E
0
0 h
0.05
0.04
x
0.03
.u
0.02
m
0.01
al
5
L
0 -30
-20
-10
0
10
20
Distance from Grain Boundary (nm)
30
Fig. 11. STEM-EDS measurement from an a-Al,O, grain boundary (NiAI-0.2 Zr).
8.3.2.2 NiAI - 0.1 wt % Y Figure 12 is a bright field TEM image of the Y-doped NiAl sample oxidized at 1200 "C. The thickness of the oxide is 1 pm. No Y-rich particles or voids were observed at the metal/oxide interface.The concentration of Y at the NiAl/ol-Al,O, interface and at the oxide grain boundaries was measured and analyzed by the same methods as the for Zr-doped alloy. Y was detected at the metal/oxide interface (figure 13) as well as at the oxide grain boundaries (figure 14).The calculated amount of Y was 0.07 0.04 mono-
E. Schiitnnnn, J. C. Yang, M. J. Grahnm arid M. Ruhle
130
Fig. 12. Cross-sectionalTEM image of NiAI-O.1Y after 24 h oxidation at 1200 "C showing the presence of A12Y409 particles (indicatcd by arrows) in thc alloy.The particle indicated by S contains sulfur.
Interface > E
!!
c
01 c
2
0
0 >.
E >i
NiAl
250 200
1
I
4
t c
I
150{
T I
- 4 0 -30 - 2 0 - 1 0
0
Distance from
10
20
30
40
Interface (nm)
Fig. 13. STEM-EDS measurement ay'oss the NiAl/a-Al,O, interface (NiAI-0.lY). 1% 5
layer at the metauoxide interface and 0.2 0.1 monolayer at the oxide grain boundaries. No sulfur was detected at the metal/oxide interface or at the oxide grain boundaries.The detcction limit for S is 0.1 monolayer. TheTEM cross-section image from NiAl - 0.1 wt% Y shows the presence of precipitates within the alloy, which are pointed out by arrows in figure 13.STEM-EDS analysis showed that the precipitates consist of Y, Al and 0.Using electron diffraction the
8 TEM Investigations on the Oxidition of NiAl
- 4 0 -30 - 2 0 - 1 0
0
10
20
Distance from Grain Boundary
30
131
40
(nm)
Fig. 14. STEM-EDS measurement from an a-A120, grain boundary (NiAI-O.1Y).
2500
-
) .
I
el
I
2000 1500
C
2
1000
X
500
0
0
2
4 Energy
6
8
10
(KeV)
Fig. 15. STEM-EDS spectrum taken from the S-rich region in figure 12.
phase of the particles has been determined to be AI,Y,O,. During further LDS analysis sulfur was found in some of the particles. One example is shown in figure 12, where S is detected in the particle indicated with the letter S. The S was only found in the small dark region within the particle. The corresponding EDS spectrum is shown in figure 15.
132
E. Schrrniam, J. C. Yung, M . J. Gruham und M . Riihlt.
8.4 Discussion 8.4.1 Transient Oxidation of NiAl The present results support earlier observations by Doychak et al. [4] concerning the epitaxial growth of metastable A120, phases in the transient oxidation stage of Xi Al. An interesting result is the presence of ( lll)y-A120, twin boundaries within the oxide platelets. The observation that these twin boundaries have been always found to be parallel to the growth direction of the platelets indicates that the twin boundaries play an important role as fast diffusion paths during the growth of the platelets. Similar observations have been made by Doychak et al. [4]. who identified oxide platelets as twinned 8-A1,03.The reason for the formation of 8-Al,O, instead of yAl,O, could be the higher oxidation temperature of 1100 "C compared to 950 'C of the present experiment. Regarding the nucleation of a-A120,, other authors have noted randomly oriented a-A1203grains, but the literature is contradictory concerning the nucleation sites of aA1,03. Doychak [lo] suggested that nucleation of a-A1203occurs at the oxiddgas interface. Studies on the oxidation of NiCrAl by Smialek and Gibala [ll] using planar ' E M indicated the presence of a-Al,O, grains next to the metal. The in-depth localization of specific structures is ambiguous by planar TEM. The present observations on cross-sectional specimens clearly show that the nucleation of a-AI2O33occurs at the metal/oxide interface and not at the oxide surface. Furthermore the microstructure present in the TEM cross-section indicates that the nucleation of a-A1203occurs by the formation of new oxide underneath the y-A120, layer rather than by a transformation from the yAl,03. The nucleation of a-A120, is followed by its further growth into the alloy as well as along the interface and by transformation of the metastable y-Al,O,. Concerning the transformation path from the metastable A120, to the thermodynamically stable a-Al2O3Doychak et al. [4] found the following sequence of phases: y+s+s+a
The present results indicate that the direct transformation from y-Al,O, to a-Al,O, is also possible: -Y+a
An important aspect is the observation of voids underneath the oxide layer in the initial stage of oxidation (0.1 h, 950 T ) . The formation of interfacial voids is a characteristic phenomenon during the oxidation of NiAl as has been shown by Brumni and Grabke [12]. However after 50 h of oxidation no voids have been found at the interface. A possible explanation could be the formation of inward growing a-A1,03,which could have filled preexisting voids.
8.4.2 Segregation of Zr andY The experimental results concerning the segregation of Y and Zr at oxide grain boundaries are consistent with the observations of Pint et a]. [13] from NiAl - 0.2 Zr.
8 TI;M Itivestigatioizs on thP Oxidation o f NiAl
133
However in the present study the first quantitative results of reactive element segrcgation have been obtained. A comparison with the results of the SIMS investigations [7] indicates that the segregated reactive elements (Zr, Y) reduce the cation diffusion along the a-Al,O; grain boundaries, as has been proposed by Przybylski ct al. [ 141. In contrast to grain boundary segregation studies the segregation of reactive elements at the mctal/oxidc interface has not yet been measured. Segregation at the interface is important to scvcral models. such as the ..poisoned interface model" [lS]. This model proposes that rcactivc clcmcnts scgrcgate to the interface where they intcract with defects and thereby influence the oxide growth mechanism. However further high spatial resolution experiments are necessary in order to further investigate the metal/oxide interface. An important result of the present analytical study is the detection of S in Y-rich precipitates within the alloy. This observation supports the ,,sulfur effect model", which proposes that reactive elcnients getter the S [16,17].The presence of S reduces the surface energy of the mctal and thereby promotes the formation of interfacial voids, as has been shown by Grabkc ct al. [18].Although no S-rich particles have been found in NiAl - 0.2 wt% Zr. it is possible that the Zr can lower thc activity of S by solute - solute interaction. Further work is necessary in order to study the interaction between reactive elements and S.
8.5 Summary TEM invcstigations havc been performed in order to study the oxidation of NiAl.The transient oxidation of NiAl is characterized by the growth of metastable phases. IJndcr the conditions of the prcscnt cxperiments ((001)NiAl, 950 T) y-Al,O, was the predominant oxide phase in the initial stage of oxidation. (1 11)y-Al,O, twin boundaries seem to play an important role during the formation of the typical platelet morphology. After longer oxidation the formation of a-Al,O, initiated at the mctal/oxidc in terfacc. After oxidation at 1200 T a-AI,O, was the only AI,O, phase present in the oxide scale. Analytical studies revealed the scgrcgation of reactive elements (Zr, Y) at the oxide grain boundaries as well as at the mctal/oxidc interface. Analytical studies on NiAI-O.1Y further more indicate that Y can getter the S in the alloy.
8.6 Acknowledgement The authors thank Ute Salzbcrgcr for the preparation of the excellent 'I'EM samples and Jirg Thomas for his assistance with the S'I'EM. Igor Levin and Ilarald Mullcjans is thanked for hclpful discussions. The work at the Max-Planck-Institut fur Metallforschung was funded by the Dcutsche Forschungsgemeinschaft. J. Y. gratefully acknowledges the support by the National Science Foundation.
I34
E. Schiirrrann, J. C. Yang,M. J. Graham arid M. Riihle
8.7 References [l] E. Schunzarirr, J. C. Yang, M. J. Grnhnm nnd M. Riihlc. Miitrrials anti Corrosiorr, -lh ( 1 9 Y 5 ) 218 - 222. (21 J. C. Yang, K. Nadarzinski. E. Scluinicinn and M. Riihle. Scriprrr Mrtrtll. Mtrrc~r.,.U (1995) 104.7 - 1048. [3] J. C. Yang, E. Schunzann. I. Levirr, I% Midlejnrrs and M. Riihle, Mot. Res. S w ~ r pProc., (1996) ill press. [4] J. Doychak, J. L.Smialek arid 7: E. Mitchell, Met. Turns. A ,-7OA (1989) 499. [ S ] G. C’. Rybicki and J. L. Smialek, Oxid. Met., 31 (1989) 275. [ 6 ] M . W Brrtmin and If.J. Grabke, Corros. Sci., 3-3 (1992) 1677. [7] R. Prescott, D. E Mircliell and M. J. Grahnm. Proc. 2nd Inr. Cotif on Microsc.opy of ~l.~-iilu~iorr. (1993) 455. 181 A. Strecker, U. Salzberger and J. Mayer, I’racticrrl Metullogrirphy, 30 (1993) 48-7. 191 J. A. S.Ikeda. I! M. Chiang. A. J. Garatt-Reed and J. B. I/: Sande. J. Artwr. Ccsr Soc., 76 (1993) 2447. [ 101 J. Doychak, Doctord thesis. Cnse Western Reserve Univer.yity.Clcvelnrrd Ohio, (1986). (111 J. 1,. Smialek and R. Gibala. Met. Trans. A, 1 4 A (198.7) 2143. [ 121 M. W Brumm and II. J. Grabke. Corr. Sci., 34 (1993) 547. [13] B. A. Pint, A . J. Gnrntt-Reed and L. W Hohhs, Proc. 2nd Itit. Corrf Microscol>.ycif Osidatiotr, (I 99.3) 463. [14] K. Przybylski, A. J. Garatt-Reed, B. A. Pint and G. J. Yurek,J. Elecrrochcrri. Soc., 134 (1987) 3207. [lS] B. Pieraggi and R. A. Rapp, J. Electrochem. Soc., 140 (1993) 2844. [16] J. G. Srneggil, N. S.Bornstein and M. A. DeCrescente, Oxid. Met., 30 (1988) 259. [17] D. R. Sigler, Oxid. Met.,29 (1988) 23. [I81 H. J. Grabke, D. Wiemer and H. Viefinus,Applied Surf Sci.,47 (1991) 243.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
9 Failure of Alumina Scales on NiAl Under Graded Scale Loading M. Hollatz, M. Bobeth and W Pompe
9.1 Introduction The investigation of thc fracture-mechanical behaviour of protective oxide scales is important in order to understand typical failure mechanisms of the scale and to be able to diminish scale failure by an appropriate materials preparation. Mechanical stresses within scales can lead to their cracking, delamination and spalling. The stresses are caused by various mechanisms: (i) comprcssive stresses during oxidation owing to lateral oxide growth in the scale, (ii) stresses due to oxide phase transformations, (iii) stresses resulting from temperature changes due to the thermal expansion misfit between the oxide and the substrate, and (iv) stresses owing to a deformation of the underlying substrate under service conditions. Typical failure mechanisms of oxide scales have been reviewed comprehensively in the literature [ 1-51. In particular, scale failure under compressive load has been studied extensively [6,7] because it can lead to a rapid destruction of the material by a spallation of scales. Spallation by buckling and wedging has been discussed in detail in [S] as a function of the scale thickness and the temperature drop upon cooling of specimens. Tensile loading of the scale affects the formation of through-scale cracks which can be followed by a delamination and spalling. The crack initiation and propagation within thin films and along the film-substrate interface is well understood theoretically for the case of elastic films on elastic substrates [9]. However, for the present study of alumina scales on NiAI, the ductility of the metallic substrate plays an important role. To the authors’ knowledge, there arc only a few theoretical studies of this problem which take into account the substrate ductility [lo, 111. The mechanical bchaviour of oxide scales has been investigated by a complex of appropriate techniques (cf. for example [12]), where the attention has been focused on the analysis of the stress development in the scale, as wcll as the measurement of the scale adherence. In the case of weak scale adherence, spontaneous scale failure is often observed during oxidation or cooling of specimens. For systematic investigations of the fracture-mechanical properties of oxide scales, scale failure is induced by a controlled loading of the scale which is produced by an appropriate dcformation of the whole specimen. To this end, different bend tests in combination with acoustic emission analysis are widely applied (131.For example, tensile failure of oxide scales in dependence on the
136
M. Hollritz, M. Bobeth mid W Poriipe
temperature and strain rate has been studied recently by Nag1 et al. 114,151 by means of the four-point bend test. Acoustic emission was employed to monitor the formation of tensile cracks and subsequent spallations. Additionally, the fracture process was observed in situ in a scanning electron microscope. By using standard tensilc and creep tests. the influence of temperature and strain rate on the spacing of tensile cracks in oxide scales has been investigated in detail by Hancock et al. [16]. The bend test has the advantage of testing the scale simultaneously under tensile and compressive load. Surprisingly, in connection with the bend test, experimental studics deal mostly with scale failure under tension whereas defect patterns on the side face of bending bars are less analysed although they provide a lot of additional information as will be shown below. The application of bend tests at lower temperatures is limited to materials which exhibit a sufficient ductility. Otherwise, thc specimens fracture before damage in the oxide scale occurs. In the present study. this was observed for thc intermetallic NiAl at room temperature (RT). In such cases, indentation can serve as alternative method to induce scale failure in a simple reproducible way. Indentation tests have been applied recently to test the adherencc of brittle films. Hou and Stringer [17] used a Vickers indenter and measured the area of the delaminated oxide scale around an indent as a function of the indentation load and thc oxidation conditions. Drory and H I I ~ C I ~ I ' I I S O I I [18] induced a large-scale delamination of diamond coatings on titanium alloys by means of a cone-shaped indenter. From the extent of the delaminated area, the layer adherence was estimated. Another method to evaluate the film adherence was applied by Sigler [19]. Tcnsile stresses in the scale were induced by forming macroscopic domes in oxidised specimens using a rounded punch and collar. For well-adherent films, the formation of patterns of nearly equidistant parallel cracks were observed whereas for weak adherence extensive spallations occurred. In order to simplify the theoretical interpretation of fracture-mechanical experiments based on controlled specimen deformations, the geometry of the deformation should be as simple as possible. Accordingly, four-point bending and spherical indentation were chosen as deformation methods in the present study.The aim was to elucidate typical failure mechanisms and to cvaluate the fracture-mcchanical properties of the scale from ex-situ observations of the arising defect patterns in the scale. The essential idea was to analyse those scale regions on the deformed specimens where the stress field in the scale varies spatially. In the case of the bend test, such a graded scale loading is present on the side face of the bending bar where the scale is under tension and compression on the top and bottom side, respectively (Fig. 1). In the case of the indentation test, the scale region outside the indent has been proven to be most suited for an analysis of crack patterns (Fig. 2). Onc of the difficulties of a quantitative analysis of scale failure is the determination of the stress state in the scaleas it results from the oxide growth process and from additional substrate deformations. The application of X-ray diffraction provides usually only a limited latcral spatial resolution. For alumina scales consisting of the a-Al20, phase, Lipkin and Clarke [20] have shown recently that stresses can be measured at KT also by means of optical fluorescence spectroscopy (OFS) employing the piezospectroscopic effcct [21]. This method pcrmits a high spatial resolution of stress measurements in the micrometre region and is thereforc especially useful for the analysis
9 Eiiilure ofAlumiria Scales on NiAl Under Graded Scale Loading
137
I Fig. 1. Schematic of cracking and spalling of the oxide scale on the side face of an oxidised bending bar under a graded scale loading, a,,(y),at room temperature (RT) and oxidation temperature (OT):C - cracks, S - spallations. N F - neutral fibre. Rockwell-Indenter
Random cracks
Circumferentialcracks
Radial cracks
Fig.2. Schematic of crack formation in oxide scales due to spherical indentation ( R = 200 km, h = l ~ m , a = 3 0 ° , 1 0 a dF=20tolOON).
of a graded scale loading. In connection with the present study, OFS has been applied to measure stresses in as-grown alumina scales on NiAl after cooling to RT [22]. The paper is outlined as follows. After a brief description of the specimen preparation and the stress measurement by OFS, crack and spallation patterns on oxidised NiA1 specimens are shown, which have been produced by bend and indentation tests at room or oxidation temperature. In the second part of the paper, first attempts to an interpretation of the crack patterns are presented. In particular, the possibility to derive the oxide fracture toughness and tensile strength from the crack patterns is discussed.
9.2 Experimental Specimens were manufactured from single crystalline P-NiAl with nearly stoichiometric composition (Al-content from 49 to 50at.%).The single crystals were prepared by the Bridgman method. Specimens for bend tests at RT were doped with about 0.1 at.% Fe and exhibited an enhanced RT ductility [23]. The specimens were mechanically polished to a 0.25 p.m diamond finish. The amplitude of the resulting roughness, controlled by laser scanning and scanning tunnelling
138
M. Ifollritz, M. Bobeth and W.Ponipe
microscopy, was less than 200nm. The specimens were oxidised in air for clit'ferent duration t,,, = 20 to 100 hours at temperatures T,, = 1223 and 1373 K. Stresses in the oxide scale after cooling to RT were measured by means of OFS (cf.Sect. 5). For comparison, also an X-ray measurement using the siiz' y-method was performed. Oxidised specimens were deformed by spherical indentation and by four-point bending with typical strain rates ranging from to lO-?s-'. In a few ciises, also three-point bending was used. The size of the bending specimens for 17T tests was 30 X 3 X 2mm3,and for high temperature tests 45 X 4 X 4mm'. For the spherical indentations, a diamond indenter (Rockwell-Indenter) with sphere radius of 200 p i was used. After deformation, the crack patterns i n the oxide scales were examined b y laser scanning microscopy (LSM) and scanning electron microscopy (SEM).
9.3 Scale Morphology The morphology of alumina scales on NiAl depends on the oxidation temperature and oxidation time. At lower temperatures (<1400 K). the metastable oxide phases yand 8-A1,0, are formed initially. These metastable phases transform into a-A120., during a transient time which decreases with higher temperature (> 1 200 K ) [24--261. 'The scales consisting of the metastable alumina phases show a needle- or plate-like oxide surface. The transformation to cu-AlzO, is accompanied by a large shrinking of the specific volume, which can cause tensile stresses in the scale. As a consequence. a transient appearance of microcracks was observed (221. A characteristic feature of the oxide transformation is the growth of a network of so-called oxide ridges (cf. Figs. Xc and d).The ridges grow along grain boundaries in the oxide scale [24] and probably nlso at microcracks. After cooling of the specimens to R'T, the alumina scales were mostly well adherent. Spontaneous spallations were observed only occasionally. In those cases, the forniation of large interfacial voids or high densities of smaller voids was found to be the reason for the weak adherence.
9.4 Scale Loading The development of stresses in the scale is caused by various mechanisms which are briefly considered in the following. The relation between the stress. u, and the elastic strain, within the alumina scale is given by the Hooke's law.'The elastic properties of the polycrystalline scale are assumed to be isotropic with E, as Young's modulus and I'(, as Poisson's ratio. Because of the free surface of the scale. a plane stress state in the scale is supposed with uD= 0. z is the direction perpendicular to the film plane, and x,y are the in-plane coordinates.The x-component of the stress tensor is then given by
9 Failitre of Aluiiiinu Scales on NiAl Under Graded Scale 1,oathg
139
The elastic strain tensor in the scale. and analogous for uyy. results as sum of different contributions (for brevity, the tensor indices are suppressed in the following)
denotes the tangential substrate strain at the oxide-metal interface due to an additional deformation of the specimen. In writing (2), it is supposed that the scale is flat and adherent to the substrate so that the substrate deformation is transferred to the scale. The remaining part of the strain, E ~ is. called residual strain and summarises all mechanisms during oxidation and cooling which influence the stress state in the scale. E~ denotes the elastic strains which result from thc transformation of the metastable oxide phases to a-Al,O,. Because of the large specific volume shrinkage of about 10% during this transformation, rather high tensile strains can arise which cause localised scale cracking [22]. The term E~ is due to lateral oxide growth in oxide grain boundaries. This term grows with oxidation time and would lead to large compressive stresses in the scale. However, these stresses relieve partly by oxide creep at elevated temperatures, or by a wrinkling of the scale [27].The corresponding relief in the elastic strain is denoted by &. Commonly, the sum, + &, which results as competition between lateral oxide growth and stress relief, is called growth strain. At room temperature, the main contribution in (2) is the strain due to the thermal expansion mismatch between alumina and NiAl
where thc indices ‘m’ and ‘0’refer to NiAl and alumina, respectively. Using the data of the thermal expansion of NiAl and polycrystalline alumina quoted in [28] and [29],respectively, the value E~ = -0.006 is obtained for cooling from 1373K to RT.The corresponding stress is given by uT = & E ~ / ( 1 - v,,) = -3.2 GPa ( E = 400 GPa, v = 0.24 [30]). A certain porosity in the scale would lead to a diminished value of the Young’s modulus. SEM observations revealed however a compact polycrystalline structure.
9.5 Stress Measurement Chromium impurities in the a-alumina phase fluoresce under laser irradiation of appropriate wavelength. For materials exhibiting the piezo-spectroscopic effect, the wave number, v , of the fluorescence line depends on the stress state. In the case of alumina, the line shift, Av, is approximately proportional to the stress for stresses up to a few GPa
(sum convention) where the Hi, are the piezo-spectroscopic coefficients and the aij are the components of the stress tensor [21]. New measurements of the piezo-spectroscopic coefficients of alumina, have been reported recently by He and Clarke [31].
140
M. Hollati, M. Boberh and W Pomye
Within a coordinate system suitably matched to the crystallographic directions of alumina, the non-orthogonal piezo-spectroscopic coefficients vanish (nij= 0 for i =I= j). For polycrystalline specimens without texture, the line shift is proportional to the trace of the stress tensor. In the case of untextured thin films with free surface, the stress component perpcndicular to the film plane vanishes.Thus, the line shift is given by
Formula (5) rcvcals that in polycrystalline films only the trace of the stress tensor,
+
uxx uyy, can be measured by OFS.
OFS has been applied in [22] to examine the stress development in alumina scales on (111) oriented NiAl. After oxidation and cooling to RT, the stress state in the polycrystalline alumina scale is isotropic within the film plane. Test measurements by OFS and X-ray diffraction gave no evidence for a texture. The stress analysis of scales grown at 1373K in dependence on the oxidation time rcvcaled large compressive stresses at RT which were mainly caused by the thermal expansion mismatch between alumina and NiAl. For oxidation times smaller than two hours, mean stress values of about 2.9 GPa were found, and for larger oxidation times of about 3.5 GPa. The latter value was satisfactorily confirmed by an X-ray measurement which yielded a value of 3.4GPa. The increase of the stress with oxidation time was presumably affected by lateral oxide growth in oxide grain boundaries. By comparison with the stress of 2.9 to 3.3 GPa resulting from the thermal expansion mismatch (thc interval corresponds to the uncertainty in the knowledge of the elastic constants and thermal expansion), a compressive growth stress during oxidation of 200 to 500 MPa seems to be likely. X-ray measurements usually yield a mean stress averaged over a scale area of few mm2.This area is large compared to characteristic features of the scale morphology. Measurements by OFS with a laser spot of about 3 ym suggested the presence of considerable stress variations in scales grown at 1373K. One rcason for these variations could be the growth of oxide ridges which leads to an inhomogeneous stress state as a result of scale thickness variations. Furthermore, localised delaminations can considerably diminish the stress.This is indicated in Fig. 3 by the different line shifts of fluorescence signals obtained from adherent, delaminated and spalled pieces of the oxide scale. The line shifts correspond to compressive stresses of 3.3, 1.6 and OGPa. respectively. Further details on the potential of OFS for the analysis of stresses in alumina scales havc been outlined in [20,22].
9.6 Analysis of Scale Failure 9.6.1 Bend Test Oxidised single crystalline NiAl specimens were deformed by four-point bending a t elevated temperatures. Bend tests at RT could be performed only with Fe-doped NiAl because of its higher ductility. Typical scale defects which were generated on the side
9 I+iiliire ofAluniiria Scales on NiAl IJnder Graded Scale Loading
141
0.15
0.10
0.05
0.00
14460 14440 14420 14400 14380
14360 14340
Wavenumber [cm-']
14 20
Pig. 3. Ihorescencc signals obtained from adherent, delaminated and spalled regions of the oxidc scale rcvcaling different line shifts due to a complete or partial stress relaxation.
face of bending bars are shown schematically in Fig. 1. In the following, the observed defect patterns under tensile and compressive load will be discussed.
9.6.1.1 Tensile Load Under tensile load, nearly equidistant parallel cracks propagate from the edge of the bending bar to the neutral fibre according to the increasing bending deformation. Examples of parallel cracks observed at RT after bending at elevated temperature are shown in Fig.4. The spacing between the cracks was larger for lower strain rate (cf. Fig. 4a and b). The cracks presumably do not penetrate into the metallic substrate because of its higher fracture toughness (cf. Fig. 5). For a given bending deformation, the cracks propagate up to a critical distance to the neutral fibre (Fig. 1). Below this distance, the tensile load in the scale is too small to cause crack initiation or propagation. For weak scale adherence and low substrate ductility, through-scale cracking is accompanied by a delamination and spalling of the scale. This was observed to a large extent on Fe-doped NiAl specimens after bending at KT (Fig. 6).'l'hc weak adherence in this case was caused by a pronounced formation of interfacial voids.Void formation at the oxide-metal intcrfacc has often been observed for NiAI. It was studied in more detail by Brurnnz and Grubke in [32]. For well adherent scales, delamination can occur subsequent to through-scale cracking at a higher substrate deformation [15]. However, in the present bending cxpcrimcnts at 1223K, delamination under tensile load was not observed. The high ductility of NiAl at elevated temperatures enabled presumably a sliding of the oxide scale along the oxide-metal interface. In this way, stress concentrations at the oxide-metal interface at t h e bottom of through-scale cracks, which try the scale to delaminate, were avoided obviously.
142
M . Hollaiz, M. Bobeth and W.Potnpe
Pig.4. Parallel cracks under tensile load on the side face of bending bars after bending at osidas- ': b: iox = I00 11, tion temperature ( T , , = 1223 K) and cooling to RT (a: to, = 20 h; E = & = 10-3s-1).
9 Faihrre ofAliminaScale7 OIZNiAl Uizder Graded Scale Loading
143
Fig.5. Schematic of the lateral propagation of a through-scale crack in thc scale.
Fig.6. SEM micrograph of the side face of a bending bar after 3-point-bending at RT showing regions of spalled scale (bright spots) under tensile (top) and compressive (bottom) load (Fe-doped NiAI single crystal, I,, = 21 h, To, = 1243 K, NF - neutral fibre).
9.6.1.2 Compressive Load Under compressive load. extensive spallations have been found on Fe-doped NiAl after RT bending (Fig. 6), which were affected by a high intcrfacial void density. On the contrary, despite a rather large deformation (cf. Fig. 7a), no scale defects were detected
144
M. Hollatz. M. Boberh and W Pompe
on the specimen shown in Fig.7b, which was deformed at 1223K with a strain rate of 10-5s-'. The specimen shown in F i g . 7 ~was deformed at the same temperature, but with a strain rate of 10 3s-3.Localised spallations are visible on this bending bar also very near to the neutral fibre. Within the spalled regions, exceptionally large voids with diameters up to 30 pm were found (Fig. 7d). The scale failed obviously by scale buckling and subsequent delamination and spalling IS]. Surprisingly, also small cracks. running parallel to the specimen edge, were found on the compressive side as indicated in Fig. 7d. The cracks were caused by the perpendicular tensile deformation, E?,. acconipanying the compressive deformation in x-direction.
Fig.7. Scales on the side face of single crystalline NiAl bending bars after bending at oxidation temperature (Tax = 1223 K) and cooling to RT (a) deformed bending bar, (b) adherent scale aftcr slow deformation ( t = 20 h,E = 10-ss-l), (c) isolated spallations at large interfacial voids (d) magnification of spallations together with tensile cracks o n the (lox100h,E = 10-3s-1y compressive side of the specimen.
9 Foilitre ofAlicniino Scales on NiAl Under Graded Scale Looding
145
146
M. Hollarz, M.Bobeth and W Pompe
9.6.1.3 Substrate Deformation and Scale Loading An accurate calculation of the strain on the side face of bending bars for the present experiments is very complicated. The reason is that the specimens consisted of single crystalline NiAl which exhibits strongly anisotropic plastic material properties. Furthermore, the bending bars were not specially oriented. Nevertheless. in order to demonstrate the specific mechanism of crack pattern formation on the side face of bending bars, the substrate deformation is calculated in the following under the simplifying assumption of isotropic material properties. As a consequence, the results derived below do not apply rigorously to the present experiments. The x-component of the strain on the side-face of a thin bending bar subjected to a four-point bending is approximately given by (cf. Fig. 1)
where y denotes the distance from the neutral fibre and R is the radius of curvature of the bending bar. The perpendicular strain component is related to eyxs by E ~ =~ -v,,E,,~. S By inserting this relation and (6) into (1) and (2),one obtains
where a laterally isotropic rcsidual strain E~ was supposed. In the limiting case of ideal plasticity, the Poisson's ratio of the metal v,, = 0.5. For bending at oxidation temperature, the strain contribution E~ is relatively small since E~ = 0.The spatial dependence a,,(y) for cR = 0 is plotted in Fig. 1. At room temperature, the residual strain eR in (7) represents a large negative contribution which has to be overcompensated by the term proportional toylR in order to induce tensile stresses in the scale. A more accurate description of the elastic-plastic deformation of the bending bar, which takes into account large deformations, can be achieved by a numerical simulation based on the finite element method. The scales on the side face of bending bars exhibit usually a defect-free region in the vicinity of the neutral fibre. The border lines between the defect-free region and the cracked regions on the tensile and compressive side of the bending bar define critical distances y:" and y,'"", respe&ly. Insertion of these values into (6) yields critical substrate deformations E$:;;ari&:;.;". The corresponding critical scale loads, cT!&!',and &om xx,c, are obtained by inserting y:" and yFminto formula (7). It should be noted that applications of formulas (2) and (7) for determining the actual scale loading is justified only when the scale is not able to relieve the additional stresses induced by the substrate deformation. For slow deformations a t clevated temperatures, this supposition is not fulfilled under compressive load when oxide creep or a slight scale wrinkling during the deformation affect a stress relief. In such cases. the scale-substrate composite can be characterised by the critical substrate deformation beyond which the scale fails.
9 fiiilure ofAlurnina Scales on NiAl Under Graded Scale Loading
147
For a quantitative illustration, the critical substrate strains, which characterise the appearance of scale spalling, have been estimated for the specimen shown in Fig.6. The critical substrate strains are 6 :;= yF"/R = 0.011 and = yYm/R= -0.014. By inserting thcse values into (7) and assuming E~ = E ~the , corresponding stresses result as cry;c= 1.3 GPa (OF = -4.0GYa) on the tension side and a::; = -8.0GPa ( U y = - 1.2 GPa) on the compression side. The corresponding strains in the oxide are obtained as E$$" = 0.006 (&;:,'en = -0.011) and E $ F=~- - 0.019 ( & y e =p 0.002). Note the strongly anisotropic stress states indicated by the large differences between the xand y-components of the stress tensor. It has to be emphasised that the value of the critical tensile stress derived from (7) results as the difference of two large terms - the tensile substrate deformation and the compressive strain due to the thermal expansion mismatch.Thus, the relative error of aE.;is rather large. For example, a change in the strain E~ of 0.001 leads to a change of aEcof about 0.5 GPa. When the border line ycte"is defined by the absence of through-scale cracks in the region y < y,"", then the value a$'= a,,(y~")represents a lower bound of the oxide tensile strength. As outlined below, the critical strcss is in general not equal to the tensile strength of the scale. Only in those cases when new cracks are initiated on the side face of the bending bar, uR,$haracterises the tensile strength.
e:P,m
9.6.2 Spherical Indentation 9.6.2.1 Anisotropic Crack Patterns The surface strains on oriented NiAl surfaces which are produced by spherical indentation depend sensitively on the surface orientation and the crystal direction within the surface. This orientation dependence is caused by the strongly anisotropic plastic material behaviour of single crystalline NiAl. Correspondingly, the crack pattcrns in oxide scales resulting from spherical indentation of oxidised NiAl single crystals exhibited a pronounced anisotropy depending on the surface orientation [33].The basic mechanisms of crack formation owing to spherical indentation are illustrated in Fig. 2 for the case of isotropic plastic material properties. Similar to the bend test, the cracks in the alumina scale are caused by large tensile strains of the underlying NiAl substrate which are transferred to the scale provided the scale is well adherent. The alumina scales tested at RT exhibited usually a good adherence. Localised spallations were found onlypccasionally (Fig. 8c). Through-scale cracks inside a spherical indent showea a rather random arrangement. Outside the indent, relatively regular arrangements of radial and circumferential-like cracks were observed (Figs. 8 and 9). The length of radial cracks was up to 40% of the radius of the indent. Note that the radial cracks in Fig. 8a propagate only radially outward during the indentation process, whereas in Fig. 8c also new radial cracks arc initiated in some distance from the periphery of the indent. This difference is presumably connected with the different scale morphologies. The scales grown at 1223K seem to be relatively homogeneous despite the needle-like oxide surface. On the other hand, the scales grown at 1373 K exhibit a network of oxide ridges which is probably connected with relatively large scale defects.
148
M. tlollatz, M. Bobeth and W Ponipe
Fig. 8. SEM micrographs of crack patterns outside spherical indents on (1 11) NiAl showing radial (a and c) and circumferential-like (b and d) cracks (a and b: To, = 1223 K, to, = 20 h, c and d: To,= 1373K, to, = 100 h; RT indentation, BI -boundary of indent).
The arrangement of the radi,al and circumferential-like cracks around spherical indents depends strongly on the surface orientation [33]. A schematic representation of the crack pattern on (111) NiAl is shown in Fig. 9 together with surface profiles in different crystal direction in the surface plane.
9.6.2.2 Substrate Deformation and Scale Loading In order to get a quantitative estimate for the indentation-induced substrate deformation, a numerical simulation was performed by employing the finite element method (FEM). To simplify matter, isotropic material properties were supposed. The presence of the oxide film was neglected in a first approximation because of its small thickness
9 Failure ofAlicrnina Scales on NiAl Under Graded Scale Loading
149
'.,
[1?11
100 V r n
Fig.9. Arrangement of cracks around a spherical indent on (111) Ni A1 toghether with surface profiles measured by LSM.
compared to thc indenter radius. The diamond indenter was approximated as a rigid body and the friction between indenter and specimen was neglected. The substrate material was supposed to obey the von Mises yield criterion with yield stress Y and a piecewise linear work hardening characterised by a tangent modulus ET. In accordance with studies in the literature [34-361, the FEM calculations showed that the w h c e profile produced by spherical indentation is closely connected with the hardening of the material. In the limiting case of weak work hardening, a piling-up of the surface around the indent was obtained whereas in the opposite case a sinkingin of the surface occurred (cf. Fig. 10a). In order to illustrate this effect, two special examples are shown here. Thc plastic material parameters for the example exhibiting a piling-up are similar to data given in 137,381. In order to simulate also a sinking-in of the surface, the high hardening rate given in [38] for deformations <2 % was extrapolated to large deformatiomThe clastic material parameters used are similar to that of polycrystalline NiAl[23]. The results of the FEM calculations for the surface displacement and surface strain outside the indent are shown in Fig. 10.The penetration depth of the indenter was cho-
M. Hollatz, M. Bobeth and W Ponzpe
150 0.05
0
&
-0.10 -0.15l' 0.0
n
6\"
2.0
.-s
0.0
'
'
'
'
'
'
"
'
'
"
'
I
0.4 0.6 0.8 Radial distance r / R
0.2
"
'
1.0
z -2.0 ; o
5
r
v,
v-
5
(n
-4.0
-6.0 -8.0
0.0
/
r
-1.0 0.0
c "
-
I ' ' . "' . I ' ' ' ' 0.5 1.0 1.5 2.0 Radial distance r / a g - 1
'
.
"
Fig. 10. Results of FEM calculations of the spherical indentation: surface profile (a), radial (b) and circuinlerential (c) surface strains versus radial distance from the centre of the indent for a piecewise linear work hardening (tangent modulus E7. 0.1 E for E ~<, 0.02 and E, = 0.0044 E for E ~>,0.02 (full line): ET= 0.1 E (dashed line); E 180 GPa, u = 0.3, Y = 0.2 GPa, a,lR = 0.5).
sen as in the present experiments.The plots in Fig. 10 were normalised by introducing a geometrical contact radius ug = Jd(2R-d). where R is the indenter radius and rl the indentation depth. The calculated surface profiles (Fig. 10a) are similar to those observed in different crystal directions on (111) NiAl (Fig. 9).'The plots of the surface strains (Figs. 10b and c) show that for weak work hardening the piling-up of the surface is connected with compressive radial strains, E,: and tensile circumferential strains, E&, For strong work hardening it is opposite. By inserting the calculated substrate strain, E ~into , formulas (1) and (2), one finds a decrease of the stress in the scale with distance from the periphery of the indent according to the decrease of E ~For . the example of the piling-up of the surface, a circumferential tensile stress arises, and for the example of the sinking-in a radial t e n d e stress. Correspondingly, in the former case radial cracks are expected, and in the latter case circumferenti?! cracks. The tensile stresses calculated at positions, which correspond to the observed tip of radial cracks and of the outermost circumferential-like
9 Failure ofAlurnina Scales 011 NiAl Under Graded Scale Loading
151
crack. amount to about 1 to 2GPa. These values are compatible with known tensile strengths of alumina. A more detailed analysis of the resulting scale stresses, which includes also the effect of friction between indenter and spccimen, has bcen given in L331.
A comparison of the observed crack patterns with the predictions, derived above for isotropic material properties, revealed that the strongly anisotropic plastic propertics of NiAl lead to crack patterns which can considerably differ from thosc expected for an isotropic model. On the (1 11) NiAl surface (Fig.9), the surface displacement u,(r) (i. e. sinking-in or piling-up) and the crack direction correlate in the same manner as for an isotropic model. However, on the (100) surface, radial cracks have been observed also in surface regions were a sinking-in of the surface occurred [33].This peculiar plastic deformation is presumably affected by the small number of operative slip systems for dislocation gliding in NiAl. The calculation of the complex stress state in scalcs on NiAl single crystals, which is produced by spherical indentation, seems to be impractical at present. An experimental analysis is eventually possible by applying techniques of strain measurements bascd on image analysis.
9.7 Tensile Strength and Fracture Toughness In the following, a brief mechanical analysis is given which is to provide a first qualitative understanding of the relationship between the crack patterns and the fracturemechanical oxide properties. It will be pointed out that appropriatc scale loading generates crack patterns which characterise either the tensile strength or the fracture toughness of the oxide scale.
9.7.1 Cracking Under Tension Through-scale cracks are initiated when the tensile stress in the scale exceeds the oxide tensile strength. As it is well known, the tensile strength is determined by the oxide fracture toughness and the size and arrangement of initial flaws in the scale. The flaws start to grow when the stress inten-sity factor, K , at the crack tip of the flaws exceeds the fracture toughness, K,c. The stress intensity factor is given by K = K~u jTf where K f is a geometry factor of the order of unity, u is the tensile stress, and a, is the flaw size. (In cases, where the elastic interaction of flaws is important, a, has to be identified with a so-called composite defect size [39].) As a consequence, the tensile strength, u,, and fracture toughness are connected by the equation K,, = K~(T,/%. However, the factor K f and the flaw size are not known often. In those cases, one cannot derive, for example, the fracture toughness from the tensile strength, but both quantities have to be deterrnincd experimentally. After initiation of a crack, the further propagation is determined by the stress field in the scale and the fracture toughness of the material. Because of the higher fracture toughness of the metallic substrate, the cracks are presumably arrested at the oxidemetal interface. When the oxide-metal adhesion is not too wcak, the cracks do not de-
152
M. Hollatz. M. Bobeth and W Pompe
flect into the oxide-metal interface which would lead to a partial delamination of the scale (cf. [lo]). Consequently, the cracks propagate only laterally as shown in Fig. 5. A theoretical analysis of this crack mode for elastic films on elastic substrates, supposing isotropic material properties, has been given in [9]. Extensions of the analysis to ductile substrates have been reported in [lo, 111. As a peculiarity, it was found that the stress intensity factor for thc lateral crack propagation, K,, does not depend on the crack length, but on the scale thickness (provided L>>h, Fig. 5). Furthermore, it is a function of the ratio of the Young’s moduli of the film and the substrate. EJE,,. and of the yield stress Y of the substrate.Thus, K, can bc expressed as [40] K I = K I ( E ~ / E ,o/Y)ov%. ,,
(8)
By taking into account thc yielding of the substrate near the film-substrate interface, H u and Evans [lo] derived within a so-called ‘shear lag approximation’ the following formula
where K depends on E,IE, (K = 1.5 for polycrystalline alumina on polycrystalline NiAl). Formula (9) concerns the propagation of a single crack.The additional effect of a mutual unloading of parallel cracks has been considered in [ 9,11,41]. Cracks propagate laterally when the stress intensity factor, K,, exceeds the fracture toughness, Klc,of the scale. In a nonuniform stress field, also the stress intensity factor varies spatially according to formula (9).The position of the crack tip in a nonuniform stress field is determined by the equation K, = K,,.The last equation together with formula (9) permits, in principle, to derive the fracture toughness from the position of the crack tip provided thc tensile stress is known as a function of the position (cf. Fig. 11). (In an accurate description, formula (9) has to be replaced by a corresponding relation K , ( o )which takes into account the interaction of parallel cracks.)
Fig. 11. Schematic of crack initiation and propagation under graded scale loading as a test of the fracture toughness and tensile strength of the scale.
9 Fditre oJAlurniria Scales on NiAl Under Graded Scale Loading
153
For example, on the side face of a bending bar the tensile stress decreases with distance from the specimen edge. With increasing curvature of the bending bar, the cracks are driven inward to the neutral fibre.The position of the crack tip,y,,p,at maximum bending deformation has to be inserted into (7) in order to calculate the tensile stress at the crack tip ut,p= uxx(ytlp) (Fig. 11).The fracture toughness follows then from thc equation K,, = K,(utlp). For the specimen shown in Fig. 4b, a value Y , ~ J=R0.007 was derived by optical microscopy. This value has to be inserted into (7) to obtain utlP. The residual stress in the scale could not be measured by OFS since the scale consisted mainly of metastable alumina phases. Consequently, also the growth stress could not be determined. The bending temperature was equal to the oxidation temperature so that E~ = 0. Neglecting the remaining contributions to E ~a, fracture toughness of 1 2 M P a G at a temperature of 1223K follows from formulas (7) and (9). The corresponding high temperature values of the yield strength and the Young’s modulus were assumed as 50MPa [23] and 340GPa 13, 421, respectively. According to measurements in [43], the scale thickness was chosen as 1.1 km. The estimate of the fracture toughness of 12 M P a G is relatively large compared to data for oxide scales reported in [3]. One reason for this large value is probably the neglect of the interaction of parallel cracks when formula (9) is applied.Also, the presence of growth stresses would diminish the estimate of the toughness. Furthermore, the ratio y,,JR was possibly overestimated by our measurement. The detection of the crack tip is rather difficult since the crack opening width becomes very small upon cooling to RT. In particular, when the failure strain &,,,,e’Jen is smaller than the thermal expansion mismatch E ~ a, complete crack closure could occur. The largc value of the estimated fracture toughness could be also effected by the small yield stress of 50 MPa obtained from bulk measurements of polycrystalline NiAl [23]. Because of a considerably enhanced dislocation density in NiAl beneath the scale, a strong work hardening could lead to an increase of the effective yield stress. This is also suggested by the small crack spacings observed (Fig. 4b), with a minimum I,, = 8 p m . If one estimates the yield strength roughly from the relation Y = 1.7uh/In,,”[lo], which was derived for film cracking under uniform tensile stress (i. e. it does not apply strictly to the present situation), Y = 500MPa is obtained. This large value would lead to a fracture toughness of 5 M P a G . The above consideration concerning the determination of the fracture toughness is quite general. It applies also to the propagation of the radial cracks during spherical indentation. An analogous method can be used in order to determine the tensile strength of the scale. In this case, the direction of the spatial variation of the tensile stress field, i. e. the direction of the stress gradient, Va,has to be parallel to the direction of the tensile stress. This situation is realised, for example, in the case of the circumferential-like cracks observed near spherical indents. A schematic of the basic relationships is shown in Fig. 11. Without any defect formation, the scale loading would increase with substrate deformation as indicated by the plot of the imaginary stress in Fig. 11. Actually, the stress relieves by scale cracking. The following two mechanisms can be distinguished. (i) When Va is parallel to the direction of the tensile stress, a subsequent initiation of new cracks occurs during the increase of the scale loading at the position y, with a,,(yc) = uc.(ii) When Va is perpen-
154
M. Hollatz. M. Bobeth arid W Ponipe
dicular to the direction of the tensile stress, the initially formed cracks propagate with increasing substrate deformation according to the equation K,(v,J = K,, ivirh utlP = uxx(yrlp). However, new cracks are not further initiated provided the size of initial flaws is smaller than the scale thickness. In both cases, further cracks between earlier parallel cracks can be formed, of course, at a considerable higher substrate deformation since also the stress between cracks can increase with deformation. For a finite yield strength of the substrate, a minimum crack spacing is however reached [lo, 111. In some cases (cf. Figs. 4b and 8c), the initiation of new cracks was found also when Vu was perpendicular to the direction of the tensile stress. This phenomenon can he explained by the presence of spatial stress variations in the scale or by the presence of large scale defects of the order of the scale thickness so that the stress intensity factor to initiate these large flaws is roughly equal to the stress intensity factor for the cracks propagating from the specimen edge. The crack pattern in Fig. 4b shows that new cracks are preferentially formed at inhomogeneities in the scale (dark spots) which correspond most likely to regions where the transformation of the metastable alumina phases to the a-phase starts. In these regions, intrinsic tensile stresses owing to the oxide phase transformation are expected. These additional tensile stresses are obviously responsible for the observed initiation of new cracks.
9.7.2 Cracking and Spalling Under Compression The compressive failure of oxide scales has been investigated comprehensivcly in [6-81.According to the analysis by Evans and Lohb [6], there are two main routes to spalling - scale buckling and wedging. The alumina scale shown in Figs.7~and d failed certainly by the buckling mechanism. Within the spalled regions on the specimen surface, large interfacial voids were detected which represent interfacial flaws where the scale buckles during scale loading. A mechanical stability analysis yields the critical equi-biaxial stress, uh.when buckling occurs [7]
where h is the scale thickness, a the radius of a circular interfacial crack (a >>h), and K~ = 1.22. For u/h = 10, et critical buckling stress of 5 GPa results from (10). The characteristic siFe of single voids observed on different specimens ranged from 3 up to about 30 p m karge voids were found, for example, on the specimen shown in Figs.7~and d. The specimen exhibited also a few spontaneous spallations. If one inserts the thermal expansion mismatch stress of 2.7 GPa for cooling from 1223K to RT into the Ihs of formula (lo), a value of 15 pm is obtained for the critical radius where the scale buckles (h = 1.1 pm). This value is compatible with the observed size of the largest voids and explains therefore the appearance of a few spontaneous spallations. After buckling, at a critical compressive stress in the scale. an interfacial crack starts from the periphery of the buckled area which leads to an increase of the delaminated area. The spalling of the scale occurs by the deflection of the interface crack toward
9 Fnilure ofAlloiiinn Stoles 011 NiAl Under Grnded Scale Loading
155
the scale surface. According to the calculations in [7], it is most likely that the corresponding radius of the spalled regions, o, is given by asp = 1.9JEO/ah.
Inserting, for example, the thermal expansion mismatch stress of 2.7 GPa into formula ( l l ) , one obtains a spalling radius of 25 km. Typical spalling radii observed on the specimen in Fig. 7c were in the range from 10 to 25 Fm. The somewhat smaller radii observed can be explained by a higher stress value due to the bending deformation, which initiates spalling. SEM observations on a series of NiAl specimens showed that the formation of large interfacial voids at 1223K is more the exception than the rule. Typically, small faceted voids of only a few micrometre size were found. Larger voids are formed by a coalescence of small voids which leads to a complicated shape of the voids. Such a void structure with high void density was found on the specimen shown in Fig. 6. The high density of voids diminished the effective scale adherence and affected the large-scale delaminations observed. On the specimen deformed at 1223K with a low strain rate of s-' (Fig. 7b), spallations or cracks were not detected under compression despite the large specimen deformation (Fig. 7a) which would yield compressive stresses > 10GPa when formula (7) is applied. Presumably, the stresses relieve during the relatively slow bending deformation by Coble creep of the oxide or by a slight scale wrinkling [27].This suggestion was confirmed by RT stress measurements on the side face of bending bars using OFS. The measured stress on the compressive side of the bending bar was considerably smaller than that expected from formula (7). For high temperature applications of materials, spallation of oxide scales has to be prevented. A high oxide-metal adhesion is, of course, most important in order to avoid a large-scale delamination of the protective oxide. However, also for high oxide-metal adhesion, the scale can fail by buckling under compression when large voids are present. After scale buckling, spalling occurs by the propagation of cracks through the scale. Thus, a high fracture toughness of the oxide is also important in order to suppress compressive scale failure. The fracture toughness is probably easier to evaluate by tensile tests as discussed above.
9.8 Experimental Difficulties and Theoretical Problems The investigation of defect patterns in oxide scales formed under a graded scale loading seems to be promising since critical load parameters could be analysed ex situ. However, there are a series of experimental difficulties and open theoretical problems in realising this approach. Within the proposed approach, the fracture toughness is determined from the position of the crack tip in a nonuniform stress field. Because of the small crack opening width near the crack tip, the tip position can be detected with high precision probably only by electron microscopy. A difficulty arises in measuring the distance of the tip
156
M . Hollntz, M . Bobeth arid W.Ponipe
from the specimen edge. In the present work. optical microscopy was applied t o nieasure this distance where features of the scale were used to identify the tip position at lower magnification. When cracks are produced at elevated temperatures and analysed at KT. there is an uncertainty in the detection of the original crack tip position since the cracks can partly close during cooling of the specimen. In particular. this could happen when the failure strain of the oxide is smaller than the strain due to the thermal expansion mismatch. In-situ observation at high temperature avoids this problem. A further difficulty is to determine the stress field in the scale with sufficient accuracy. This concerns especially the determination of the tensile strength or fracture toughness from RT experiments on scales which exhibit high residual stress (> 1 GPa). Since the total strain in the scale results as the small difference of two large terms, cS- kRl(cf. formula (2), E ~ 0< ) ,both terms have to be known with high accuracy. This problem is not specific to the present approach. It applies also to other tests of tensile failure.The problem does not arise for small residual strains, i. e. for materials with low thermal expansion mismatch, and for tests at elevated temperatures. The best way to determine the stresses in the scale would be a direct measurement. However, X-ray methods have usually a limited spatial resolution which makes it difficult to measure nonuniform stress fields. 'The application of OFS for scales consisting of a-A1203provides a sufficient spatial resolution and permits, i n principle, to examine stress variations in the scale. However, only the trace of the stress tensor can be nieasured for an untextured polycrystalline scale. Thus, anisotropic stress states have to be analysed in combination with a mechanical modelling of the scale loading in order to deduce the stress components from the trace of the stress tensor. The intent in the present work was to determine the scale loading by measuring the isotropic residual strains in the scale at RT and by calculating the additional strains due to a controlled substrate deformation. However, the calculation of the substrate deformation in the case of single crystalline NiAl is very difficult because of its strongly anisotropic plastic behaviour. To simplify matter, isotropic material properties were supposed in the above analysis. For the case of spherical indentation, it was found that isotropic models are insufficient to describe the observations. Thus, in the case of NiAI, indentation can serve only as a semi-quantitative method for comparative studies. In the case of the bending deformation, the deformed specimens gave no evidence for strong deviations of their shape from that expected for isotropic properties. Nevertheless, a more detailed analysis of the scale deformation due to specimen bending is necessary. The above estimates of fracture-mechanical parameters by assuming isotropic material properties represent only rough approximations. The fracture analysis given above concerned the propagation of a single crack. The observed crack patterns suggest however the propagation of a front of parallel cracks. An accurate analysis has to take into account the mutual unloading of parallel cracks. Crack interaction effects have been studied in [9,41] for the case of an elastic substrate, and in [lo, 111 the ductility of the substrate has been taken into account.'The crack cl&sity can increase with substrate deformation due to the formation of new cracks between earlier cracks. However, due to the ductility of the substrate, a maximum density will be reached as outlined in [lo].At high tensile deformation, also scale delamination can occur by the deflection of trough-scale cracks into the oxide-metal
9 Failure ofAluminn Scales on NiAl Under Graded Scale Loading
157
interface [15]. The onset of delamination can be used to characterise the scale adherence. Related problems have been investigated in [9, lo].
9.9 Summary The intent of the present paper was to focus the attention on the analysis of crack patterns in scales caused by a graded loading. In this way, the scale can be tested simultaneously under different loads which should permit to derive critical loads or failure strains by ex-situ observations provided the scale loading is known. In the present work, the graded loading was established by four-point bending (observing the side face of bending bars) and spherical indentation of oxidised specimens. Spherical indentation is easily to perform. However, a determination of the resulting stress state in the scale seems to be too difficult.Thus, this test can be used mainly for comparative studies. The bend test is presumably more appropriate for determining fracture-mechanical properties. Other specimen deformation can be more suitable for special purposes. An ex-situ analysis of crack patterns formed under a nonuniform scale loading can provide independent information on the oxide fracture toughness and tensile strength. Whether the crack pattern characterises the toughness or strength depends on the choice of the direction of the stress gradient relative to the direction of the tensile stress. In this way, the fracture toughness could be determined, in principle, without knowledge of the size of initial flaws in the scale. A difficulty of the proposed approach is the determination of the graded scale loading. For scales consisting of a-A1203,optical fluorescence spectroscopy permits to measure stresses with high lateral resolution. A disadvantage is that only the trace of the stress tensor is obtained. Thus, an interpretation of spectroscopic data has to be complemented by a mechanical modelling of the scale loading. In the present analysis, large-scale spallations of alumina scales on NiAl were always connected with the presence of interfacial voids. In the absence of large voids and for small strain rate, it was found that the scale was well adherent under compressive substrate deformation even for large specimen deformations (cf. Fig. 7a).This suggests the presence of an effective stress relief mechanism by a slight scale wrinkling or by oxide Coble creep. In summary, the establishment of a graded scale loading in combination with spatially resolved stress or strain measurements could be an effective method for an ex-situ investigation of failure mechanisms and fracture-mechanical parameters.
9.10 Acknowledgements This work was supported by the Deutsche Forschungsgemeinschaft. The authors would like to thank Th. Hutzler, W Loser and G. Vuerst for providing NiAl single crystals and performing initial bend tests, and J. Edelmunn for his assistance in the
158
M. Hollatz, M. Bobeth and W Pompe
SEM analysk’lle authors are very grateful to D. R. Clarke and D. M . Lipkin for the support in measuring stresses by OFS, and to 11. Balke and V: Marx for helpful discussions concerning the FEM modelling.
9.1 1 References [l] A. G Evans, R. M. Cannon: Mater. Sci. Forum 43 (1989) 243. [2] J. Robertson, M. I. Manning: Mat. Sci.Technol. 6 (1990) 81. [3] M. Schiifze:Die Korrosionsschutzwirkung oxidischer Deckschichten unter thermisch-chemisch-mechanischer Werkstoffbeanspruchung, Gebriider Borntraeger Berlin. Stuttgart (1991). [4] M. M. Nagl, W i? Evans: J. Mater. Sci. 28 (1993) 6247. [5] H. E. Evans: Matcr. Sci. Rev. 40 (1995) 1. [6] H. 6 Evans, R. C. Lobb: Corr. Sci.24 (1984) 209. [7] A. G. Evans, J. W.Hutchinson: Int. J. Solids Structures20 (1984) 455. [8] H. E. Evans, G. Z? Mitchell, R. C. Lobb, D. R. J. Owen:Proc. R. SOC.L0nd.A 440 (1993) 1. (91 J. W.Hutchinson, Z. Suo: Mixed mode cracking in layered materials, in: Advances in Applied Mechanics (eds. J. W. Hutchinson and T. Y. Wu) vol. 29, p. 64, Academic Press, New York (1991). 1 [lo] M. S Hu, A. G. Evans: Acta metall. 37 (1989) 917. [l 11 F Delannuy, fl Warren: Acta metall. mater. 39 (1991) 1061. [12] S.R. J. Saunders, H. E. Evans, J. A. Stringer (eds.): Mechanical properties of protective oxide scales,Workshop,Teddington, UK, June 1994,in: Mater. High Temp. 12 (1994) 83. [13] S. R. J. Saunders, M. M. Nagl, M. Schiitze:Mater. High Temper. 12 (1994) 103. [14] M. M. Nagl, S.R. J. Saunders, W I: Evans, D. J. Hall Corr. Sci.35 (1993) 965. [15] M. M. Nugl, W I: Evans, D. J. Hall, S.R. J. Saunders: Oxid. Met. 42 (1994) 431. [16] Z? Huncock, J. R. Nicholls, K. Mahmood Corr. Sci. 35 (1993) 979. 1171 I? I!Hou, J. Sfringer:Oxid. Met. 38 (1992) 323. [18] M. D.Drory, J. W Hutchinson: Science 263 (1994) 1753. [19] D. R. Sigler: Oxid. Met. 40 (1993) 555. [20] D. M. Lipkin, D. R. Clarke: Oxid Met., submitted [21] Q. Ma, D. R. Clurke:J. Am. Cer. SOC.76 (1993) 1433 [22] D. M. Lipkin, D. R. Clarke, M. Hollatz, M. Bobeth, W Pompe: Corr. Sci.. submitted. (231 D. B. Mirac1e:Acta metall. mater. 41 (1993) 649. [24] J. K. Doychak, i? E. Mitchell,J. L. Smialek: Mat. Res. SOC.Symp. Proc. Vol. 39 (1985) 475. [25] M. W Brumm, H. J. Grabke: Corr. Sci. 33 (1992) 1677. [26] E; Schumann: Oxid. Met. 43 (1995) 157. [27] Z. SUO:J. Mech. Phys. Solids 43 (1995) 829. [28] R. D. Noebe, R. R. Bowman,M. K Nathal: Inter. Mater. Rev.38 (1993) 193. [29] Thermophysical Properties of Matter, The TPRC Data Series, Ed. Y.S. Touloukian, Vol. 13, IFIIPlenum, New York (1977). [30] M. Schiitze: Oxid. MFt. 44 (1995) 29. [31] J. He, D. R. Clarke: J.Am. Cer. SOC.78 (1995) 1347. [32] M. W Brumm, H. J. e a b k e : Corr. Sci.35 (1993) 547. [33] M. ZIollatz, M. Bobefh,W Pompe, K Marx: Acta metall. mater., submitted. [34] K. L. Johnson: Contact Mechanics, Cambridge University Press, Cambridge (1985). [35] R. Hill, E R. S., B. Storakers, A. B. Zdunek: Proc. R. SOC.Lond. A 423 (1989) 301. [36] S.Biwa, B. Storakers: J. Mech. Phys. Solids 43 (1995) 1303. [37] R. D. Noebe, R. Gibala: Scr. metall. 20 (1986) 1635. [38] Z? Lipetzky,A. Wanner, B. Schietinger, E. Arzt: in: Numerical Predictions of Deformation Processes and the Behaviour of Real Materials, Proc. 15th Riso Int. Symp. on Materials Science, eds. S. I. Andersen, J. B. Bilde-Sorensen. T. Lorentzen, 0.B. Pedersen and N. J. Sorensen. Roskilde, Denmark 1994,p. 399.
9 Failure ofAlunzina Scules on NiAl Under Graded Scale Loading [39] l? Hancock, J. R. Nicholls: Mater. Sci.Technol.4 (1988) 398. [40] M . S. HLL: Mater. Res. Soc. Symp. Proc.Vol.130 (19x9) 213. [41J M. D. Thotiless. E. Olsson. A. Gupta:Acta rnetall. rnater. 40 (1992) 1287. [42] J. R. Nicholls, C. Mendes, l? Hancock: Mater. HighTemp. 12 (1994) 85. [43] M. Boberh, E. Bischoff; E. Schurnann, M. Rocksrroh, M. Riihle: Corr. Sci. 37 (1995) 657.
159
162
J. P 7:Vossen,A .H. H. Junsseri mid J. H. W. du Wir
Although originally AISI 347 stainless steel seemed a reasonable choice, it was soon shown that it is not suited for long term operation over several years. The best corrosion protection would be provided by materials that can passivate. AISI 316 or AISI 310 stainless steel would in principle provide such a relatively stable and protective oxide scale in the presence of molten carbonate at least under cathodic conditions. IJnder fuel gas conditions AISI 316 is unsuitable since no protective layer is present [1,21. A systematic evaluation of various possible materials classes showed that the choice had to be made from nickel alloys as Inconels, Incoloys and Hastelloys, or austenitic stainless steels such as the already mentioned AISI 316L or AISI 310s. At the same time it became clear that no single material would be able to perform well under both reducing and oxidising conditions. An extra problem was created by the conflicting demand for corrosion resistance - as provided by a ceramic layer under anodic conditions, for example - and the necessary good electrical conduction.l'he conclusion from experimental work in many different laboratories was that different coatings would be necessary for different parts of the separator plates [2]. A nickel or copper coating would be appropriate for the active part at the anode, while aluminium would be a good choice for the wet seal area, where a LiAIO, oxide layer, which would be formed, could provide the necessary high ohmic resistance. Aluminides or aluniinised diffusion coatings have shown to be effective. Ni and Al both being interesting materials for the MCFC, we have investigated pure Ni and various Ni alloys including the intermetallic NiAl for their corrosion resistance under fuel cell conditions [l,3-61. The objective of this paper is to compare the corrosion behaviour of NiAl in molten carbonate with that of metallic nickel.
10.2 Experimental The experiments were performed in a so-called pot-cell which is schematically given in Fig. 1. The inlet gas composition is the MCFC anode inlet gas composition: 64'26 hydrogen, 16% carbon dioxide and 20% water vapour; the carbonate used is an eutectic mixture of 62 Yo Li,CO, and 38 % K,CO,. The cell temperature was kept constant at 923 K. The outlet gas was led through a water lock to prevent air inlet. The reference electrode consisted of an alumina tube filled with carbonate and with a small hole (app. 0.1 mm diameter) in the bottom. The function of the hole is to realise contact between the melt in the reference electrode compartment and in the working electrode compartment. The reference gas was a mixture of 14 % oxygen and 30 YOcarbon dioxide, balanced with nitrogen. The nickel used was of 99.98 % purity as purchased from Goodfellow Metals Ltd., Cambridge, England. The specimens had dimensions of 5 x5x 1mm3 and were ground to grit 1000.The counter electrode consisted of 99.99 YOpure gold foil that was bent in a cylindrical shape. The nickel electrode and the counter electrode were spot-welded to a 99.99 % pure gold wire of 0.5 mm diameter.
10 The Corrosion Behaviour of NiAl in Molten Carbonate at 650°C
163
Fig. 1. Schematic set-up of a pot-cell typc set-up. 1.Gas inlet reference electrode; 2. Gas inlet; 3. Gas outlet; 4. Glass flange glued to the jackct; 5. Outer alumina jacket; 6. Catalyst in gas inlet; 7. Alumina crucible; 8. Reference electrode gas outlet;9. Silicon rubber stops; 10. Glass lid; 11.Flag electrode. CE: Counter electrode. WE: Working electrode.TC: Thermocouple. RE: Reference electrode
The nickel-aluminium was supplied as a small bar, made upon special order by Philips Research Laboratories, Eindhoven, The Netherlands, by mixing the proper amounts of nickel and aluminium and subsequent arc-melting. The bars were cut into slices. The dimensions of the specimens used for the electrochcmical measurements were 6 x 6 ~ mm3.The 1 surface of the specimcns was not pre-treated because this alloy was considered to be too brittle to apply abrasive techniques without damage to the specimens. Specimens of the NiAl intermetallic were laser welded to a nickel-wire because gold and aluminium can form an eutecticum with a liquidus temperature below 873 K [7]. The polarisation curves werc recorded immediately after immersion with an initial potential of - 1700mV (Ni) or - 1100mV (NiAI) and a final potential of + 100mV. Potential steps of 5 mV werc applied, the current was measured after 50 s. To investigate the steady-state corrosion layers, quenching experiments were performed. Shortly after immersion, the working electrode potential was fixed at a chosen value for 4 hours. After this period the specimens were taken out of the melt while still being under potential control and cooled down to room temperaturc in air as fast as possible.The carbonate adhering to the spccimen prevented contact and interaction
164
J. l? 7: Vossen.A . H.H. Janssen and J. H. W de Wit
between the specimen and air. Sometimes longer polarisation times of 16 or 24 hours were applied to enable the formation of more oxide. The carbonate was removed by rinsing in water (pure nickel) or in a mixture of acetic acid and acetic acid anhydride (NiAl). This mixture was chosen to prevent the dissolution of the aluminium oxide scale in water. Quenching experiments were performed at potentials of 1 I00 niV 01' more anodic, since these potentials are of interest for MCFC operation. The impedance measurements were recorded in the potentiostatic mode w i t h ;I Zahner IM5d-system and IMSd-software. The amplitude of the applied potential for the impedance measurements was 10 mV (peak-peak).For our system this value gave a good signalhoke ratio and was sufficiently small to avoid clcviation from linearity [4].111e impedance was recorded in the range of 50mHz to 10 kHz.The measurements were started immediately after applying the dc-potential, from the lowest frequency, going up to higher frequencies and finally back to the lowest frequency. The impedance spectrum was measured 5 times at each frequency. The reproducibility was good. i. e. in a Nyquist plot, the results of duplo measurements were within 5 '%deviation of the average. For more experimental details we refer to earlier publications (3-61. The surface and the cross section of the quenched specimens were investigated with SEM/EDX, the corrosion products were investigated with ex-situ X-ray powder diffraction (Guinier-de Wolff method, Du-K,,.,). ~
10.3 Results and Discussion 10.3.1 Quenching Experiments Metallic nickel is stable at - 1100mV and -900 rnV. Therefore no oxide scale is formed on the surface at these potentials. Analysis of NiAl quenched after 4 hours polarisation at - 1lOOmV or -900mV shows that the surface is covered with small crystals with an octahedral and cubic morphology. as can be seen in Fig. 2a. SEMEDX-analysis showed that the corrosion product contains aluminium and oxygen (lithium can not be detected by EDX). The cross section of this specimen (Fig. 2b) shows that the corrosion product is present as a thin oxide layer. The only phase that could be detected by X-ray diffraction is a-LiAlO,. After longer oxidation times (up to 24 hours) the dimensions of the a-LiA102-crystals increased slightly,but t M i c k n e s s of the oxide scale did not increase significantly. a-LiAIO, could alobserved on aluminised stainless steel exposed under MCFC fuel and oxidant conditions for 5400 hours [8]. Approximated phase diagrams [9] and data of other groups [ l O , l l ] indicate that y-LiAlO, is probably the stable phase in molten carbonate of the chosen composition at 650"C.Thc a- and P-LiAIO, phases may also exist, but are not expected to be stable on the long term. Therefore a-LiA102is expected to be formed as an intermediate for the y-LiAlO,, which is formed slowly. aLiAIO, may be formed preferentially a consequence of epitaxial growth, while nucleation problems for the phase transition from a-LiAlO, to y-LiAlO, may occur. After quenching Ni specimens from potentials of - 775 mV and more anodic, oxide crystals can be observed also on the surface of these specimens as can be seen in
10 The Corrosion Behavioiir of NiAl irr Molten Carbonate at 650°C
165
Fig. 2. Surface (a) and cross section of NiAl (b) after 4 hours polarisation at .1100mV in molten carbonate
Fig. 3. SEM/EDX analysis confirms the presence of oxygen on the surface. X-ray diffraction analysis shows that the corrosion product is NiO. Whcn NiAl is quenched after 4 hours polarisation at potentials of -700mV or more anodic, the surface morphology and the cross section are similar to that of the specimcn quenched at -1100mV (Fig. 2), even though the potential is sufficiently anodic for the nickel oxidation to take place.This indicates that a protcctive aluminium oxide layer has been formed before inserting the specimen into the mclt or in the period just after inserting the specimen into the mclt and before applying the potential. The protective properties of the lithium aluminate scale seem very good.
166
J. f? 7:Vossen.A .H. H. Janssen and J. H . W de Wit
Fig. 3. Surface of nickel after 4 hours polarisation at -750mV in molten carbonate.
10.3.2 Polarisation Measurements 4
The potential limits for performing electrochemical experiments are limited by the stability range of the carbonate mixture. The anodic limit is given by reaction 10 while the cathodic limit is given by the decomposition of carbonate to carbon not given here. Rut already at less cathodic potentials we are dealing with the reduction of water to hydrogen, according to reaction 4. The quasi-stationary polarisation curves of nickel and NiAl are given in Figs. 4 and 5. The polarisation curves of these materials are similar.Therefore we will discuss the plot for NiAl by close comparison with Ni. The melt reactions and the oxidation reactions of nickel that correspond with the potential regions in the polarisation curves are:
H,O + CO, + 2e- + H, f C02(-1600 < E < -1070mV) H, t CO,,- -+ H,O + CO, (-1070 < E < -800mV) Ni + NC;, + 2eaccompanied by
+ 2e-
Ni + C0:- -+ NiO + CO, + 2e (-800 < E < -700mV)
(4)
(5)
(6) (7)
Ni + Ni;:, + 2e(8) (dissolution through passive layer; -700 < E < -525mV)
10 The Corrosiori Bekavioiir of NiAl in Molten Carbonate at 650°C
Nf?,:, + NiCi0 + e(-525 < E < 0 mV) accompanied by reaction (6) and (7). CO:- + CO, (E 2 50 mV)
167
(9)
+ Vz02 + 2e-
Since these reactions have been discussed in previous publications, these reactions will only be discussed here briefly [3-61.
2.00 h
N
E
aE 0
0.00
v
.-
-2.00
-4.00I
I
-1.50
-1.oo
0.00
-0.50
E (V vs. 14% 02,30% C02, bal. N2)
Fig. 4. Quasi-stationary polarisation curves of nickel in molten carbonate. Initial potential - 1700 mV, final potential I- 100 mV, 5 mV/step, 50 s/step.
3.00 -
E aE v
.-
Ni-50 Al
2.00 -
1.00 -
0.00
1
-1.oo
-0.50
0.00
E (V vs. 14% 02,30% C02, bal. N2)
Fig. 5. Quasi-stationary polarisation curves of NiAl in molten carbonate. Recorded immediately after immersion. Initial potential: OCP, final potential + 100mV, 5 mV/step, 25 s/step.
168
J. I? 7:Vossen.A.H. H. Jnnssen and J. H. W. de Wit
For a discussion of the stability of Ni the potential region of -800mV to 0 m V is most relevant. From -800mV to -525 mV Ni is oxidised and a NiO layer is formed, while from -525 to 0mV some of the Ni'+ is oxidised to Nij' accompanied by a further growth of the oxide layer due to increased ionic and electronic conductivity. Impedance measurements led to the following reaction mechanisms for the formation of the NiO scale and the partial oxidation of Ni'' to Ni3+in the NiO scale respectively [6]. The reaction mechanisms are given schematically in Figs. 6 and Fig. 7. NiO formation
accompanied by the further oxidation of nickel: Ni + C0:-
+ NiO + CO, + 2e
(22)
The NiO layer is not very protective due to the restricted potential region where it is stable. It can be easily reduced cathodically, which is accompanied with detachment of the oxide layer [4]. Because we already know from the quenching experiments that an oxide layer is formed on NiAl at -1100mV, where Ni is not yet oxidised, the anodic current at this potential is ascribed to the oxidation of aluminium. For Ni the current &nearly zero at this potential, due to near-equilibrium of the reactions 4 and 5. PossIMe reactions for the oxidation of aluminium are: 2AI
+ 302- + 2A120, + 6e-
c0;- -+ 0,- + co, Al + Lit + 2CO:- + LiAIO, + 2C0, + 3e-
(23a) (23b) (24)
co:
I
"p' Ni'-O2-Y-l NiZ*
NiO
Fig. 6. Schematic representation of the nickel oxidation and passivation mechanism in molten carbonate.
o
e
e.
Fig. 7. Schematic representation of the reaction mechanism for the transformation of bivalent nickel ions to trivalent nickel within the oxide scale.
The formation of AI,O, can only take place as an intermediate or as a thin interphase because it is not stable in contact with molten carbonate. As a matter of fact the electrolyte tile matrix (LiAIO,) is made in situ by reaction of the carbonate phase with A1,0,. Direct formation of LiA10, according to equation 24 seems possible too. Thermodynamic calculations [12] show that aluminium oxide/lithium aluminate is stable at much more cathodic potentials than -1100mV.Therefore it seems likely that the oxidation reactions proceed at all potentials in the carbonate stability range. Thus some oxide will probably be formed before the polarisation mcasurements are started when the electrode is dipped into the molten carbonate. Thc constant current from -1000mV to -800mV reflects some further but limited growth of the oxide layer, which reaches a thickness of 1 km after 24 hours (see Fig. 2).The crystallite size of the oxide also increases with exposition time especially during the first few hours. After about 4 hours the morphology hardly changes anymore. In this potential region up to -800mV no clear hydrogen oxidation (on Ni from -1070 to -800mV) either dissolved in the metal or dissolved in melt proceeds on the LiAIO, layer. The current is much lower than on Ni and no peak current is observed, which is typical for the hydrogen peak (see also Fig. 4) on Ni due to the limited amount available in the metal [4].This reflects the early protection of the base alloy by LiAlO,, so that no hydrogen is formed. The formation, onrNiO takes place between -800mV and -700mV, as is clearly visible in Fig. 4, while further oxidation to Ni3+takes place around -500mV. Onc would expect that these reactions would not be visible in the polarisation diagram of NiAl, if the oxide layer would be impervious for Ni ions. Judging from the relatively high current density around -1OOOmV, where only LiAIO, has been formed, which is a poor conductor (see Fig. 5 ) it seems justified to conclude that the layer is not blocking ion
170
.I.E! 7:Vossen,A. H. N. Jarissen and J. H. W de Wit
transport completely, possibly due to grain boundary diffusion or transport through macroscopic defects. Nevertheless, the fact that no NiO phase can be found by X-ray diffraction on top of the LiAIO, after further anodic polarisation leads to the conclusion that the anodic currents in Fig. 5 above -800mV also reflect the oxidation of the thin Ni wire to which the NiAl electrode was attached and not only the oxidation of Ni ions transported through defects in the oxide layer. Long term behaviour of NiAl in molten carbonate is probably much better than suggested by the short term electrochemical measurements. This is also suggested by the good behaviour of Kanthal-A1 (with only 5.8% of aluminium) at long term i n mersion, which was ascribed also to the formation of an aluminium oxide layer, in that case between the outer Li(Fe,Cr)O layer and the base alloy. Small amounts of Al were found to be beneficial also by Uchida [13]. Polarisation plots of preoxidised Kanthal did not show any anodic peak currents, due to this protective layer. For more details we refer to reference [l].Due to the problems with the unavoidable Ni wire no such experiments could be done on NiAI.
10.4 Conclusions The oxidation product in molten carbonate on Ni is NiO and on NiAl is a-LiAIO,.The latter layer with a thickness of about 1 pm after a few hours exposure either at the OCP or under potential control in a wide range of potentials, provides good protection against further attack, due to its low conductivity. The protection is much better than provided by NiO, which is quite insufficient due to the limited potential stability region. Long term protection by LiAIO, is probably much better than suggested by the current density around -1000mV, due to the further blocking of the fast transport paths as has been shown before for other alloys with a very much lower Al content.
10.5 Acknowledgements Financial support by the Netherlands Agency for Energy and the Environment (NOVEM) and the Netherlands Ministry of Economic Affairs is gratefully acknowledged.
10.6 References 111J.I?T Vossen, L. Plomp, J. H.W de Wit, G. Rietveld J. Electrochem. SOC.142 (1995) 3327.
[2] J.H.W de Wit:in: “Case Studies in Manufacturing with Advancerd materials”, Vol 11. Eds. A. Demaid and J.H.W. de Wit, North Holland, Amsterdam, 1995,pp 137-177. (31 J.E!T Vossen: Ph. D. Thesis: “Corrosion of Separator Plate Constituents in Molten Carbonate”, 1994,Delft University of Technology,The Netherlands. [4] J.PT Vossen, L. Plomp, J. H. W de Wit J. Electrochem. SOC.141 (1994) 3040.
I 0 The Corrosion Behaviour of NiAl in Molten Carbonate
at
650°C
171
[ S ] PC.H. Anzenf:ECN-report number: 93-R-93-027.1993. [6] J.PT Vossen, P C H . Amenr, J.H. M! rfe Wit:“Mechanisms for the Oxidation and Passive Behaviour of Nickel in Molten Carbonate‘., accepted by J. Electrochem. SOC..March 1996. [7] T B . Mussalski (Editor in Chief): “Binary Alloy Phase diagrams”,Vol. I., Am. SOC.for Metals. October. p. 90,142,1986. [S] G. Rielveld: ECN, priv. comm., March 1995. 191 J.R. Selman, H.C. Maru: in: “Advances in Molten Salt Chemistry”, G. Mamantov, ed., Plenum Press, New York, p. 159.1981. 101 K.K. Ghosh: Bull. Electrochem. 7(1991) 512. 111 G.H. Kucerrz: Argonne Natl. Lab., CONF-811014-4. 121 D.A. Shores, P Sitigh: Proc. Electrochem. SOC.84-13 (1984) 271. 131 I. Uchida, 7:Nishinu: 1992 Fuel Cell Seminar, Nov. 29-Dec.2 Tucson,AZ, p 550. Courtesy Associates., Washington. DC. 1992.
Part 111 Fe-Aluminides
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
11 Oxidation of P-FeAl and Fe-A1 Alloys I. Rommerskirchen, B. Eltester and H. J. Grabke
11.1 Introduction
Whereas the oxidation behaviour of P-Ni AI has been investigated under various conditions [l-51, the oxidation bchaviour of the related P-FeAI phase is entirely unknown. Therefore, as a comparison, the oxidation of P-FeAI was studied by means of thermogravimetrical experimcnts in the temperature range from 800°C to 1100°C. Similar phenomena as for NiAl have been observed concerning formation and conversion of A1,0, modifications and occurrence of porosity beneath the scale. At temperatures higher than 650°C the P-FeAI phase is connected to the wFe(A1) phase field without phase boundary. To elucidate the effect of decreasing order with decreasing aluminium content on the oxidation behaviour, samples from the whole range of this phase with 5,10,15,20,25,30,40 and 48 at.% A1 were investigatcd in exposure experiments. The isothermal oxidation behaviour was studied in situ by thermogravimetrical experiments in hclium with 1 vol.% 0, in the temperature range between 800°C and 1100°C (Fig. 1). The p-FeAI samples investigated had the compositions Fe,,AI,, and Fe,Al,. To analyse the transition from external to internal oxidation, exposure experiments in H,/H,O mixtures at very low oxygen partial pressures have been carried out using samples with varying aluminium content between 5 and 48 at.%.The oxygen partial pressurc was controlled with an oxalic acid water saturator to a value below the thermodynamic stability range of iron oxide (Fig. 2). In the thermogravimetrical cxperimcnts the sample is hanging from a microbalance in the temperature constant zone of a vertical furnace (Fig. 1).The microbalance records the mass gain due to oxidation and gives this information to the connected x-twriter. The temperature of the furnace is controlled by means of a Pt-Rhl8 thermocouple. The oxidizing gas is a mixture of commercial helium and oxygen which are dried by P,Os columns and then controlled by a capillary flowmeter to a total flow velocity of 2 ml/s. To analyze the transition from external to internal oxidation dependency on aluminium content, it is important to establish a small oxygen partial pressure which allows only aluminium oxide growth and not iron oxide growth. Therefore, commercial hydrogen gas was cleaned from rest oxygen by a Cu catalyst and then dried by a P,O, co-
176
I. Rornmerskirctieii, B. Eltester und H. J. Grabke 1
exhaustpipe
7
mnba Row
Fig. 1. Flow apparatus for thermogravimetrical investigations 1.P,O, column, 2. capillary flowmeter, 3. three step cock, 4. thermocouple, 5. sample holder with sample, 6. resistance furnace, 7. electronic microbalance, 8. x-t recorder
Fig. 2. Experimental equipment for exposure experiments at very low oxygen partial pressure 1.active Cu catalyst, 2. P,O, column, 3. oxalic acid water saturator, 4. capillary flowmeter, 5. fur. nace
lumn (Fig.2). A defined concentration H 2 0 is added to the H, gas while flowing through an oxalic acid water saturator containing a mixture of 90 % C,H20, . 2 H,O and 10% C,H,O,. Corresponding to the dissociation equilibrium of H,O at the given furnace temperature, a very low oxygen partial pressure in the range lo-" to 10 bar is attained in the reaction atmosphere (Table 2).
I I Oxidation of P-FeAI and Fe-AI Alloys
177
Table 1. Composition of the investigated alloys and iron aluminides Itd. chemical analysis A1 wt.O/,
Fe wt.%
2.47 5.12 7.86 10.9 14.0 16.6 23.7 30.3
98.1 95.4 92.4 89.1 86.0 83.0 76.2 69.2
Ni
C
S
0
wt.%
wt.%
wt.%
ppm
0.0018
49 81 81 81 100 92 127 34
<0.01 <0.01 <0.01
<0.01
<0.005 <0.005 <0.01 <0.005 ~~
0.0053 0.0044 0.0046 0.0036 0.0050 0.0033 0.0051 0.0056
0.0017 0.0020 0.0012 0.0014 0.0011 0.0013 0.0006 ~
Table 2. Oxygen partial pressures in the reaction atmosphere of exposure experiments
T
Po2
I"C1
b l
900 1000 1100
8.9. 5.0. 10-20 1.5 '10-19
11.2 Thermogravimetrical Investigations The mass gain versus time curves measured on P-FeAI in oxidation experiments in He-l%O, at various temperatures are shown in Fig. 3a [6]. All samples show initially linear curves corresponding to the rapid growth of metastable, disordered modifications of alumina. Figure3b illustrates the temperature region from 800 to 900°C.The increase of diffusion coefficients through the initially formed aluminium oxide scale leads to increasing mass gain with higher temperatures. At 800 and 850 "C, no a-alumina but traces of 8-A1,0, were detected after the end of the oxidation experiments by X-ray diffraction.The curves for the temperature range between 9OCL1050"C are plotted in Figure 3c. After some hours the metastable modifications transform to protective a-alumina and the curves become parabolic. The rate of mass gain of P-FeAl decreases from 900°C over 950°C and 1000°C to 1050°Ccorresponding to the accelerated conversion to slow growing a-alumina with increasing temperature. At 1100°C (Fig. 3a) an increase in mass gain was observed in consequence to the accelerated diffusion of thc oxide forming components, aluminium and oxygen, through the scale. The increased tendency to oxide spallation at 1100°C also attributes to the high mass gain rate at this temperature. Values of the oxidation rate constants k, were determined from the slope of plots of the square of mass gain per area versus time. It is possible to separate the regions for the growth of metastable alumina modifications (I) in the first hours of the experiments and for protective a-alumina growth (11) after longer oxidation times. The
1. Komnwrskirchen, B. Eltester utid H. J. Gruhke
0.35 -I
0,lO 0.00
,. . , I~
0 . -
. ,2'
. .__ __ ..-. t'-._._._ .,;/
,,., .
*
.
,
,
,
.
,
.. ..
.
. .. ..
_c
.-.
950°C . . . . . . . . 1OOo"C 1050°C
1 ...
....
,1...-
.r-
""
I
0
1
I
1
100
200
300
time [h] Fig.3. P-FeAI. He-l%O,, mass gain per area as a function of time: a) examples from the hole range of investigated temperatures, b) 800 to 900°C and c ) 900 to 1050°C.Explanations see text
values of k, are listed in Table3. The temperature dependence of the k, v'cl I ues of p-FeAl in comparison to P-NiAl is shown in the Arrhenius plot in Figure 4. It becomes evident that there are no large differences in the oxidation rate constants for the two related aluminides. The therniogravimetrical experiments with Fe,,AI,, yielded corresponding results within the experimental error, so they are not presented in detail.
I 1 Oxidution of p-FeAl and Fe-AlAlloys
179
Table3. Parabolic rate constant k, for oxidation of E~e,,AI,, from linear regression of (Am/A)? versus t T I"C1
I
800 850 900 950 1000 1050 1100
2.6.10-14 7.9 ' 10-14 2.3. lo-" 3.2 ' 10-13 4.4 ' 10-13
k, [ g ? ~ r n - ~ s - ' ] I1
2.4
'
-_
lOr'5
3.5.10-'~ 5.5.10-15 5.8 lo-'' ~ - ~ _ _ _ _ _ _ _ _ ~ _ _ _ _ _ _ _ _ ~
11.3 Exposure Experiments In the three scanning electron micrographs (Figs. 5-7) the oxide morphology at various temperatures after oxidation for 312 h is shown. At 900°C (Fig. 5) the oxide is characterized by the typical whisker morphology within oxide troughs of 8-A1,0,. At 1000°C the oxide whiskers are shorter and more compact (Fig. 6). Oxide ridges begin to form and the oxide scale shows a tendency to spall. Pits in the metal matrix can be observed in spalled areas of the scale. At 1100°C the oxide shows the typical network of oxide ridges of a-AI,O, (Fig. 7). Large cavities in the matrix are found which are a
I
7x10"
I
I
8x' 1'0 9x19 reciprocal temperature [IK]
1
1xw3
Fig.4. Arrhcnius plot for the parabolic rate constants of P-FeAI in comparison to P-NiAI
180
1.Rornmerskirchen, B. Eltester arid H. J. Grabke
Fig.5. SEM-picture of P-FeAI. 312h oxidized in He-l%O, at 900°C oxide whiskers,typical for 0-Al203.
Fig.6. SEM-picture of P-FeAI,312 h oxidized in He-1 %O, at 1000°C, oxide ridges begin to form, pits grow in the matrix as is visible in spalled areas
Fig. 7. SEM-picture of p-FeAl, 312 h oxidized in He-l%O, at 1 lOO"C, network of oxide ridges, typical for a-Al,O,
1 I Oxidution of p-FeAI arid Fe-A1Alloys
181
Fig.3. Optical micrograph of a cross section of Fc,,,AI!,, oxidized in H,/H,O at 900°C. (A Ni coating is applied for the preparation of the cross section). Internal oxidation (black intrusions), iron protrusions on top of oxide scale (bright half circles on top of oxide scale)
consequence of aluminium depletion. From the analysis of the samples after exposure experiments it can be concluded, that there are more pits and cavities at higher aluminium content of the alloys and at higher temperatures. Samples with aluminium contents less than 10 at.% show a planar surface without any pits or cavities. Internal oxidation was observed only at 900°C in samples with less than 10 at.% aluminium, whereas at 1000°C and 1100°C even 5 at.% aluminium was still enough to form a complete alumina outer layer. The internal oxides generate stresses, which cause formation of outer iron protrusions (Fig. 8). The pore formation beneath the oxide scales which is caused by Al-consumption and inward diffusion of Fe, is more marked for the ordered phases near 50 at.% and less for the disordered alloys with low A1 content.These, however, at 900°C show a tendency to internal oxidation. A1,0, is growing inward while Fe protrusions are pressed outward in the process [6,7].
11.4 References [l] ES. Pettit:Trans. Metall. Soc.AIME 239 (1967) 1296. [2] M. W Brumm, If.J. Grubke: Corros. Sci. 33 (1992) 1677. [3] G. C. Rybici, J. L. Smiulek: Oxid. Metals 31 (1993) 275. [4]M. W Brumm, If.J. Grabke: Corros. Sci. 34 (1993) 547. [5] M. W.Brumm, If. J. Grubke: Corros.Sci. 36 (1994) 37. [6] I. Rommerskirchen: Ph. D.Thesis. Universitat Dortmund (1995). [7] E. Schumann, G. Schnotz, K. fl Trumble, M. Riih1e:Acta metall. rnater.40 (1992) 1311.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
12 The Oxidation Behaviour of ODS Iron Aluminides B.A. Pint, El? Tortorelli and I.G. Wright
12.1 Introduction Interest in advanced power generation cycles that involve indirectly-fired gas turbines, in which coal- or gas-fired high-temperature heat exchangcrs are used to heat a working fluid in a closed system, has led to interest in materials for heat exchangers capable of operation at temperatures of the order of 1200 to 1300°C. The candidate materials are ceramics and, possibly, oxide dispersion-strengthened (ODS) alloys. An ODS FeCrAl alloy met the strength requirements for such an application, in which the working fluid at 0.9 MPa was heated from 800 to 1100°C over a tube length of 4m [l]. The oxidation life of ODS FeCrAl alloys is determined by their ability to form or reform a protective alumina scale, and can be related to the time required for the aluminum content of the alloy to be depleted to some minimum level [2-51. As a result, the service life is a function of the available aluminum content of the alloys and the minimum aluminum level at which breakaway oxidation occurs. Therefore, there is a minimum cross sectional thickness which can be safely employed at tempcraturcs above 1200"C.The major factors that result in depletion of the reservoir of aluminum in the alloy are the inherent growth rate of the aluminum oxide; and the tendency for the scale to spall, which results in a (temporary) increase in oxide growth rate in the area affected by spallation. Because of their significantly higher aluminum content 2 2 8 at% compared to -9 at%), alloys based on Fe,Al afford a potentially larger rcservoir of aluminum to sustain oxidation resistance at higher temperatures and, therefore, offer a possible improvement over the currently-available ODS FeCrAl alloys [61. Iron aluminides possess excellent high temperature (800-1300 "C) corrosion resistance [7-111 but at these temperatures they do not have sufficient strength for structural applications [9,12,13].Two possible solutions are to use them as coatings [14,15] and to develop a means of strengthening the alloy at high temperatures [12]. Conventional alloying approaches have not been successful in retaining the strength levels necessary for high temperature applications, and thus current efforts are focused on oxide dispersion strengthening to improve the creep strength at 800-1200°C [13]. Oxide dispersions offer an excellent opportunity to improve both the mechanical properties and the oxidation resistance by the addition of a reactive element (RE) oxide such as Y,O,. (Commercial ODS alloys typically contain a mixed Y,03-A1,0,
184
B.A. Pint, R E Torlorelli and 1.G. Wrighl
dispersion [16-18]). Previous work [10,11.19] has shown that RE alloy additions (Y or Zr) are effective in improving the oxidation resistance of iron aluniinides at temperatures above 900°C and a similar improvement is expected for the addition of an RE oxide dispersion. The use of RE oxide dispersions has improved the oxidation resistance of various A120,-forming alloys such as FeCrAl [ 1-5,18.20,21], NiCrAl [22.23] and NiAl[24,25].'The present paper examines the effects of several different oxide additions on the oxidation behaviour of a base Fe-28atY0A1-2%Cr (FAS) alloy. A coinparison is made among cast, RE-doped FAS. A120,-dispersed FAS and RE oxide-dispersed FAS. Initial comparisons of ODS Fe,AI to cast Fe,Al showed that there was little or no difference in performance at temperatures below 1000°C [26]. T h u s this work has focused on higher temperatures, particularly 1200" and 1300°C.
12.2 Experimental Procedure Gas-atomized Fe-28 %A]-2 %Cr [27] powder and sub-micron oxide powder were mechanically blended in a flowing Ar atmosphere using a water-cooled, high-speed attritor and stainless steel balls. Standard dopant additions of 0.2 cation% were made. FAS powder was milled without any addition to create an A1,0, dispersion. The blended powder was canned, degassed, and extruded at 1100°C. For comparison, a FAS powder extrusion without an oxide addition, cast FAS, cast Fe-28 %A1-5 %Cr-0.1 %Zr (FAL), cast P-NiAI (with and without a 0.04 YO Zr addition), and a commercial Zr0,dispersed (0.06 %Zr) Fe-20.4 %Cr-lO.6 %A1 alloy (Kanthal alloy APM) were also tested. Chemical compositions are given in Table 1. Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) of the as-extruded alloys consistently showed a = 1pm grain size and a bimodal particle distribution of 2 6 5 0 nm particles within grains and 0.2-1 pm particles at grain boundaries. 'fie number of larger particles generally increased with oxygen content. Oxidation coupons (10-15 mm diameter X 1 mm thick) were polished to a 0.3 pm finish with alumina powder and ultrasonically cleaned in acetone and methanol prior to oxidation. Isothermal oxidation experiments were conducted in 1atm dry, flowing 0, with weight gains measured by a Cahn model 1000 microbalance. Cyclic oxidation experiments were conducted at 1200°C and 1300°C in air and in dry, flowing 02.In short-term cyclic testing (2 h cycles), specimens were hung in a furnace with dry, flowing 0, and weight changes were measured continuously using a Cahn model 1000 microbalance to check for any indication of isothermal spallation. (In these tests, spallation was only observed upon cooling.) During long-term testing (100 h cycles), specimens were placed in the furnace in pre-annealed alumina crucibles so that spalled oxide could be collected and weighed. In both types of cyclic tests, the specimen weight changes were measured before and after oxidation using a Mettler modcl AE240 balance. After oxidation, specimens were examined using scanning electron microscopy with energy dispersive X-ray analysis (SEM/EDXA). Selected samples were analyzed using TEM equipped with EDXA.
12 The Oxidation Behavioitr 0.f O D s Iron Alitrninides
185
Table 1. Chemical analysis of the as-extruded alloys. Concentrations (in atomic percent) were determined by inductively coupled plasma analysis. Sulfur, oxygen and nitrogen contents wcre determined by combustion analysis. FAS: Cast E;xtruded
A120,-1
AI203-2 AI2O;-3 Ce Ce Hf La La Nd
sc Y Y
Y-low 0 Y Y Y Yb Zr APMt NiAl NiAl FAL MA956? ODM031
dopant (at%)
A1 (at%)
Cr (at%)
0 (at%)
N (at%)
S
(at%)
Si (at%)
Ti (at%)
<0.01 Y <0.01 Y <0.01 Y <0.01 Y CO.01 Y 0.19 0.10 0.18 0.16 0.04 0.17 0.19 0.05 0.09 0.17 0.19 0.28 0.46 0.29 0.17 0.06Zr <0.01 Zr 0.04 Zr 0.10Zr 0.30 Y 0.24 Y
28.40 24.81 25.9 1 27.34 29.84 26.60 28.88 29.77 26.49 25.68 27.39 29.16 26.17 25.75 29.22 26.92 26.84 26.60 27.04 27.21 10.65 50.23 49.65 27.23 8.40 6.65
2.00 2.04 1.94 2.21 2.81 1.98 2.63 2.21 1.98 1.89 2.13 2.76 2.13 2.19 2.74 2.39 2.33 2.26 2.06 2.16 20.37
0.01 0.10 6.33 2.34 1.32 4.88 3.31 1.63 4.23 6.57 2.39 1.90 5.06 5.33 2.03 3.03 3.35 3.90 3.67 2.73 0.17 n.d. n.d. 0.04 0.66 1.16
0.001 0.24 2.89 0.67 0.10 3.24 0.40 0.13 2.37 3.20 0.65 0.19 0.42 0.24 0.30 0.11 0.16 0.23 1.27 0.32 n.d. n.d. 0.02 0.01 n.d. 0.08
29 48 61 62 24 53 24 24 62 60 60 23 54 51 21 49 54 50 63 70 10 20 35 53 110 71
<0.01 n.d. 0.13 0.20 0.32 0.19 0.60 0.71 0.10 0.11 0.05 0.63 0.18 n.d. 0.66 n.d. n.d. n.d. 0.10 0.05 0.43 <0.01 n.d. <0.01 n.d. 0.09
<0.01 0.08 0.07 0.08
<0.01
<0.01 5.14 19.39 13.29
c0.01
0.08
<0.01 <0.01
0.06 0.07 0.08 <0.01 0.07 0.07 <0.01 0.07 0.08 0.08 0.07 0.07 0.03 n.d. <0.01 <0.01 0.39 0.67
12.3 Results 12.3.1 Oxidation Behaviour of Cast Fe,Al In order to fully characterize the effect of an oxide dispersion on the oxidation behaviour of Fe,Al, it is first necessary to illustrate the behaviour of the cast material. Above -20at%Al, Fe,Al is a primary alumina-former with no significant transient formation of Fe-rich oxide [8,27]. Like most alumina-formers, there is poor adhesion of the undoped alumina scale. However, the scale formed on Fe,Al is more highly convoluted than scales observed on MCrAl alloys and P-NiAl, Figure la. This results in small pieces of spalled oxide which tend to stick to the substrate, but can be brushed off easily. Reactive element alloy additions of Y and Zr have been shown to improve the scale adhesion [10,11,19],but compared to similar additions to P-NiAl, these additions are not nearly as effective. Figure 2 shows a comparison of total weight gains for Fe,Al and P-NiAl with and without a Zr addition, during 10 X lOOh cycles at 1200°C. While there is an improvement for both alloys with the addition of Zr, the total weight
B.A. Pint, l?E Tortorelli arid I.G. Wright
186
Fig. 1. SEM secondary electron images of the scale surface after 2 h at 120Oo<'in O,, the substrates were made from (a) unmilled, extruded Fe-28Al-2Cr powder, (b) milled, extruded powder which contains an A$O1 dispersion and (c) milled, extruded powder which contains a YzO, dispersion.The addition of an oxide dispersion flattens the a-A1,03 scale of Fe,AI.
0
100 200 300 400 5 0 0 6 0 0 700 8 0 0 9 0 0 1000
Time (hr)
0
1 ' 2
3
4
5
6
7
8
Square Root Time ( K r )
9
10
Fig. 2. Total weight change (sample + spalled oxide) during 100 h cycles at 1200°C for P-NiAI and Fe,Al,with and without a Zr alloy addition. P-NiAI t Zr shows almost no scale spallation during 10 cycles, while the other alloys show significant spallation. The Zr addition is not nearly as effective in improving scale adhesion in Fe,AI.
Fig. 3. Parabolic plot of weight gain versus square root of time for Zr-doped P-NiAI and Fe,AI alloys at 1200°Cin 1 atm 02.
12 The Oxidation Behnviour of ODs Iron Aluminides
187
gain on FAL, is higher due to scale spallation. Isothcrrnal kinetic data in Figure 3 and Table 2 indicate that the parabolic growth rate for Fe,AI + Zr is not significantly different than that for P-NiAI + Zr. However, after just one 100h cycle. a clear difference in scale adhesion can be observed, Figure 4. At higher magnification, it appears that the substrate deforms and at "peaks" in the substrate, the scale spalls upon cooling. Thus these observations indicate that RE alloy additions are less effective in improving scale adhesion in iron aluminides. Table 2. Parabolic oxidation rate constants at 1200°C for isothermal exposures in dry. flowing 0, (except where noted). Alloy
Rate Constant (g2/cm4s) 1200°C
Extruded FAS FAS + A1203-2 FAS + AI,O,-l FAS + 0.05Y FAS + 0.1Y FAS + 0.2Y FAS + 0.3Y FAS + 0.5Y FAS + 0.2Ce FAS + O.1Ce FAS + 0.2Hf FAS + 0.2La FAS + 0.2Nd FAS + 0.2Sc FAS + 0.2Yb FAS + 0.2Zr FAL (Fe,AI + Zr) P-NiAI + Zr Undoped FeCrAl FeCrAl + ZrO,
1.3 X 10 I") 1.1-1.5 X lo-" 8.8-9.6 X 5.6 X 7.4 x 10-12 2.5-3.5 X lo-'* 3.9 x 10-12 5.7-6.0 x lo-'* breakaway 6.9 X lo-'* 1.1 x 10 I I **) 5.0 X lo-'* 1.3 X 10 I I 5.1 X 10 5.3 x 10 l 2 6 . w . 4 x 10-17 2.74.5 X 1.8 X 10 II"*') 3 4 x 10 I***')
_
_
_
_
~
--__ #
of tests 1 2 3 1 1
3 1 2 1 1 1 1
1
1 1 1 2 5 2 3
~
*) from reference [26], cxperiments performed in dry, flowing air ") accelerated oxidation rate, not parabolic *") from reference [18]
12.3.2 Effects of an A120, Dispersion Several authors have concluded that any stable oxide dispersion, including AI,O,, can improve alumina scale adhesion [28-301. During powder blending, some oxygen uptake is unavoidable and this creates a fine A1,0, dispersion in the consolidated alloy, with a volume fraction proportional to the 0 content. Several Al,O,-dispersed FAS compositions were created in order to study the effect of an A1,0, dispersion,Table 1. The most striking effect of the AI,O, dispersion was that it initially produced a flatter, more adherent a-Al,O, scale at 1200"C,Figure lb. However, on closer examination, it appeared that flattening the scale was the only effect of the A1,0, dispersion. Unlike a typical RE addition, the Al,O, dispersion did not change the scale surface morpholo-
188
B.A. Pint, P F Tortorelli and I.G. Wright
Fig. 4. SEM secondary electron plan views of the a-AI2O3scale formed after oxidation in 1 atm 0, for 100 h at 1200°C on (a) Fe-28AI-5Cr-O.lZr (FAL) and (b) Ni-SOAl-O.O4at%Zr.Some scale spallation is observed on FAL but not on Zr-doped p-NiAl.
Fig. 5. SEM secondary electron images of the scale surface after 2 h at 1200°C in 0,,the substrates were rnadc from (a) unmilled, extruded Fe-28AI-2Cr powder, (b) milled. extruded powder which contains an A1,0, dispersion and (c) milled, extruded powder which contains a Y,O, dispersion. At high magnification, there is little effect of the A1,0, dispersion on the scale morphology compared to a Y,O, dispersion.
gy or grain structure, Figures 5 and 6 respectively. The fine surface ridges o n cast FAS and Al,O,-dispersed FAS are indicative of outward A1 transport (Figure 5) as are the whiskers at the gas interface (Figure 6a) [18]. Finc grains at the scale surface are only observed when a RE addition is present, Figure 5c.The cross-sectional grain structure (Figure 6a) is typical of undoped a-A1203[18], which grows by the transport of both Al and 0 along grain boundaries [31-331. In isothermal experiments at 12OO0C, the
12 The Oxitlarion Behaviour of O D s Iron Aluriiinides
189
Fig. 6. SEM secondary electron images of fracture sections of the a-Al,03 scale formed after oxidation in 1 atm O2for 100 h at 1200°C on Fe-28A1-2Cr (FAS) dispersed with (a) AI,O, and (b) Y,O,.’Ihe grains in (a) are somewhat elongated but do not have the same columnar structure as with Y,O,-doping.1he Y,O, addition also eliminates the whisker formation shown in (a).
parabolic rate constant of Al,O,-dispersed FAS was similar to that of cast FAS and an undoped FeCrAl alloy,Table 2. In short-term cyclic testing (20 X 2 h) at 12OO0C,the AI,O, dispersion was somewhat effective in improving scale adhesion, Figure 7. As mentioned previously, the fine spa11 on the non-dispersed, extruded FAS tended to stick to the sample. However, during the 20 cycles, the sample began to lose weight. FAS with the highest A1,0, addition, Al2O3-1,showed little spallation during the 20 cycles. The flatter scale (Figure 1) was mostly adherent during this shortterm testing. FAS with lower AI,O, additions, A1,0,-2 and A1,0,-3, also produced a flat scale after 2 h, but spalled readily after a few 2 h cycles, Figure 7.
I .2 h
NE
y
5 0
00
9 c u
1.0
FAS AI,O,- 1
08
APM (FeCrAl
0.4
0.2 00
.zM
-0.2
-
-0.6
.-
d 0
; p.
+ Zr)
06
-0.4
-0.8 -1.0
-1.2 0
5
10
15
Number of 2h Cycles
20
Fig. 7. Sample weight change of several alloys cycled from room temperature to 1200°C with a cycle time at temperature of 2 h. Zr0,-dispersed FeCrAl exhibits virtually no spallation during this test.
B.A. Pin/, Pl? Tortorelli and I.G. Wrighl
190
s + A12q-3 1.3at% 0
= 1.32mm
0
100 2 0 0 300 400 5 0 0 600 700 8 0 0 9001000
Time (h)
Fig. 8. Total weight change (sample + spalled oxide) during 100 h cycles at 1200°C for Kanthal APM and undoped Fe,AI with various 0 contents. In long-term testing, Zr0,-dispersed FeCr A1 showed excellent scale adhesion but A1,0, dispersions did not improve scale adhesion and illstead shortened the time to breakaway compared to a cast alloy.The extruded FAS appeared to “slump” and rapidly oxidize.
As Figure 8 demonstrates, for longer-term testing (10 X 100 h), the AI,O, dispersion was not effective in improving scale adhesion. The almost linear total weight gain indicates that there was nearly complete spallation after each lOOh cycle. Examination of the samples by SEM showed scale spallation at the metal-scale interface and the formation of large interfacial voids. Furthermore, when comparing cast and Al,O,-dispersed FAS of similar thickness, the A1,0, dispersion significantly shortened the time to breakaway oxidation at 1200“C, Figure 8. Thus, an A1,0, dispersion alone was not effective in improving the oxidation behaviour of Fe,AI.
12.3.3 Reactive Element Oxide Dispersions The other oxide additions were made at a standard 0.2 cation% level in order to determine the relative effectiveness of the various additions. An initial screening test of 20 X 2 h cycles at 1200°C was used in order to assess the performance of the various dopants. Sample weight change data are shown in Figure 9. As a baseline, Kanthal APM shows almost no spallation during this test with a weight change almost identical to that measured isothermally [IS]. As shown in Figure 9, oxide additions of Ce, La and Sc accelerated the oxidation rate and led to FeO formation in the casc of the Ce addition [26].The scale on the Laand Sc-doped alloys was highly convoluted but adherent, e. g. Figure 10. In each case, the addition appeared to cause an acceleration in the oxidation rate (Table 2).This detrimental influence has been reported for La,O, additions to P-NiA1 [24] and for CeO, additions to FeCrAl [34], and is attributed to “ovcr-doping” of the RE addition. In order to test this premise, lower additions of La,O, (0.05 % La) and CeO, (0.1 YO)
12 The Oxidalion Behaviour of ODS Iron Aluminides Ce
E
2.0 I .8 1.6
v
1.4 1.2
5Cr-O.1Zr
0.8
A P M (FeCrAl
. ? h
N
I .o 06
e,
+
Zr)
Y
0.4
Nd Yb
0.2
-a
191
0.0
-0.2 -0.4
Zr
!l-0.6 -0.8
vl
-1.0 0
4
8
12
16
20
Number of 2h Cycles
Fig. 9. Sample weight change of Fe-28A1-2Cr (FAS) with various cation oxide dispersion (0.2 at%) cycled from room temperature to 1200°C with a cycle time at temperature of 2 h. None of the alloys perform as well as Zr0,-dispersed FeCrAl.
Fig. 10. SEM secondary electron image of the alumina scale surface formed on La@,-dispersed (0.2 %La) FAS after 50 h at 1200°C in O,.The highly convoluted scale results in an accelerated oxidation rate.
were added to FAS. In short-term testing (Figure ll), this appeared to reduce thc negative effects, but, particularly in the case of La, not eliminate them. During longerterm testing (10 X 100h cycles, Figure 12), these lower-doped alloys did not perform well, exhibiting accelerated and breakaway oxidation. Oxide additions of Y, Nd, Yb, Hf or Z r produced a flat alumina scale after a 2 h exposure, similar to that observed with an A1,0, dispersion, Figure 1.However, the scale on each of these alloys spalled to some degree during the 20 X 2 h cycles. The similar behaviour of these alloys did not sufficiently differentiate the performance of the various dopants. Longer-term testing (Figure 12) showed that Y,O,-dispersed FAS per-
B.A. Pint. R E Tortorelli arid I. G. Wright
192
. 2 h
N
E
1 6 1 4
0.05La
1 2
v
10
0.05Y
0 8
0 6 0 4
0.ZY
0 2 0 0
0 2 04
0
5
10
15
20
Number of 2h Cycles
Fig. 11. Samplc weight change of Fe-28AI-2Cr (FAS) with various cation oxide dispersions after 2 h cycles at 1200°C. By reducing the initial 0.2 cation% dopant leve1,improved performance was observed for additions of CeO, and La,O,. Similar improvements may be possible for reduced levels of HfO, and ZrO,. Various Y,O, contents showed little change in performance.
0
1 0 0 200 300 4 0 0 5 0 0 600 700 8 0 0 9 0 0 1 0 0 0
Time (h) Fig. 12. Total wcight change (sample + spalled oxide) during 100 h cycles at 1200°C for Kanthal APM and various oxide dispersions in PAS. La, Ce and Sc additions resulted in accelerated oxidation. None of the other additions performed as well in FAS as Y,O,.
formed better than any of the othcr additions. In isothermal testing, the alloy with a 0.2 % Y addition had the lowest parabolic rate (Table 2), but the alloy with other dopants did not have significantly higher rates. Rather than a slower scale growth rate, the lower rate of weight gain of Y,O,-dispered FAS reflects a lower degree of scale spallation relative to the other alloys. The Hf, Nd, Yb and Zr additions did produce some beneficial effect compared to an AI,O, dispersion, for which breakaway oxidation was observed (Figures 3 and 12).
12 The Oxidation Behavioiir of O D s Iron Aluminides
193
Because of the apparently superior behaviour of the Y,O, dispersion, an attempt was made to optimize this addition. However, higher and lower levels of Y in FAS (Table 1)did not produce better results than the initial 0.2 % addition, Figure 13.It is interesting to note that of the various Y,O, additions, the alloy with a 0.2% Y addition showed slightly more spallation during 20 X 2 h cycles (Figure 11 and Reference [26]), but the least spallation during 10 X 100h cycles (Figures 13), and the lowest isothermal rate,Table 2. Comparison of the performance of FAS with a 0.2 % Y addition to that of cast, Zralloyed Fe,A1 indicates that Y,O,-dispersed FAS had a lower weight gain during 10 X 100 h cycles, Figure 13. This behaviour was also observed during 100h cycles at 13OO0C,where Y,O,-dispersed FAS again showed a lower total weight gain than the cast material, Figure 14. Thus, there was some improvement in oxidation behaviour of Fe,AI by the addition of an optimal oxide dispersion. However, in a comparison of Y,O,-dispersed FAS to Kanthal APM or P-NiAI+Zr at 12W0C and 1300°C (Figures
5 d
.3
100 2 0 0 300 4 0 0 5 0 0 600 7 0 0 800 9 0 0 1000
Time (h)
Kanthal APM (FeCrAl t-=A)
0.04at96Zr 0
I00
200
300
400
I
,
,
500
600
700
Time (h)
800
Fig. 13. Total weight change (sample + spalled oxide) during 100h cycles at 1200°C for cast FAL (O.lZr), P-NiAI+Zr and various oxide dispersions to FAS. A 0.2 cation% Y addition to FAS resulted in the best Fe,Al performance. Different Y contents did not show improved performance.
Fig. 14. Total weight change (sample + spalled oxide) during 100h cycles at 1300°C for several materia1s.A~at 12OO0C,the ODS Fe,AI outperformed cast Fe,AI + Zr but showed more scale spallation than ODS FeCrAl or P-NiAI +Zr.
194
B.A. Pint, RE: Tortorelli arid I.G. Wriglil
12-14). the RE dispersion appeared to be less effective in FAS.The higher total weight gain for Y,O,-dispersed FAS reflects a greater amount of scale spallation. The ODs FeCrAl and NiAl+Zr showed almost no spallation at either temperature.
12.3.4 Increased Spallation on ODS FAS The increased degree of spallation on ODS FAS appears to be related to the forniation of a large fraction of interfacial voids which grow with time and limit contact between the metal and oxide. Examination of Y,O,-dispersed (0.2%)Y ) FAS aftcr isothermal exposures for 100, 200 and 400 h showed an increased degree of spaliarion with oxidation time. Spallation occurred mainly a t the metal-scale interface and n o t within the scale. Figure 15 shows an increasing fraction of smooth dimpled areas on the exposed metal surface with time. These dimpled regions on the metal surface reflect a loss of contact between the substrate and the a-A1,03 scale, and the formation of an interfacial void. Both the number and size of the voids increased with time. These voids were also studied o n Y,O,-dispersed FAS after 2 h at 1200°C before scale spallation occurred, by I’EM [35].Voids were observed along the metal-scale interface, Figure 16. Small (20-50nm) “perturbation” voids were also observed by I’EM at the metal-scale interface of ODS FeCrAl alloys [18]. In Y,O,-dispersed FAS, the size and frequency of these voids appeared to be greatly increased. If the voids grow with time, the string of voids in Figure 16a might be expected to coalesce to form a much larger void.The alloys with dispersions of Nd, Zr,Yb and Hf were also observed to spa11 in the same manner and to form a significant volume fraction of interfacial voids. One possible reason for the increased void formation of FAS is that the 0 and N contents (thus larger dispersoid fraction) in the ODS FAS alloys might provide an increase in the number of nucleation sites compared to commercial ODS FeCrAl
Fig. 15. SEM secondary electron images of the Y,O,-dispered FAS metal surface after oxidation at 1200°Cin 0, for (a) 100 h (b) 200 h and (c) 400 h.The void fraction (smooth areas) increased with oxidation time.
12 The Oxidatiori Behnvioiu of O D s Iron Aluminides
195
Fig. 16. TEM bright ficld image of the metal-scale interface of Y70,-dispersed Fc-28AI-2Cr after 2 h a t 1200°C in O,.The arrows mark interfacial voids [35].
. 5 El h
FAS (0 In!% 0) unmilled powder
1.2
FAS (6.3ar46 0)
N
0 8
$,
0.4 0.4
8
0.0 0.0
B
s
FAS+O.IY (5.3ats 0) FAS+O.OSY (S.Oal% 0) FAS+O.ZY ( 3 Oat% 0)
F A S 4 . 2 Y (2.0a146 0)
M
d
-0.4
0
-0.8 v1
-1.2
0
2
4
6
8
10
12
14
16
18
20
Number of 2hr Cycles
Fig. 17. Sample weight change of Fe-28A1-2Cr (FAS) with various cation oxide dispersions cycled from room temperature to 1200°C with a cycle time at temperature of 2 h.A larger amount of spallation was observed when thc 0 content was reduced.
(APM, Inco’s MA956 and Dour Metal’s ODM031,Table 1). Using more careful processing techniques, several alloys were produced with lower 0 and N contents,Table 1. However, these ODS alloys instead showed an increased amount of scale spallation, Figure 17. In general, the dopant type and level appeared to be a more critical factor than the 0 and N contents in determining performance.
12.4 Discussion The results described above and previously reported findings lead to several interesting hypotheses about the evolution of scale and intcrface morphologics at very high temperatures, and the influence of the substrate on this evolution. The present
B.A. Pint,P E Tortorelli and I.G. Wright
196
conclusions regarding the development of scale damage relate to processes that occur isothermally at the oxidation temperature. However, their effects are then manifested in terms of the proclivity for scale spallation during cooling. As such, scale characteristics resulting from cooling stresses must be distinguished from those that develop during high-temperature cxposure.
12.4.1 Stress and Substrate Effects on Scale Morphology There are a number of mechanisms by which stresses can develop in the scale and at its surface with the alloy during oxide growth [36J.If scale adherence is maintained, these stresses can be of sufficient magnitude to deform weak substrates and influence the morphology of the growing oxide. Interfacial voids can play an important rule in the stress development and the associated evolution of scale structure. Onc way [his can occur is schematically depicted in Figure 18. Given a typical compressive stress in the plane of the alumina scale, a perturbation caused by a defect, such as an interfacial void (Figure 16) can result in a tensile stress component normal to the metal-scale interface, Figure 18a. With a strong substrate, a void may grow in size without any scale buckling as the oxide layer thickens, Figure 18b. In contrast, as shown in Figure lSc, a weak substrate may accommodate the growth stress by deformation and scale huckling [37].With additional oxidation, the buckle increases in size and there is further deformation of the substrate, Figure 18d. Variations in scale thickness and cation diffusion distance in the vicinity of the buckle may result in additional local stresses. Thus, buckle formation and growth can be aided by a weak substrate and, with all other factors being equal, at sufficiently high temperature, weaker alloys would be expected to have more convoluted scales upon exposure to an oxidizing environment.
t
- Void
/zi
Gas
Gas
Oxide Metal
Oxide P Metal (b)
possiple crack path
Void
Metal
Fig. 18. A scale buckling model (a) with compressive, in-plane growth stresses in the oxide scale. a defect such as an interfacial void results in a normal tensile stress; (b) with a strong substrate. the void may continue to grow but no deformation occurs; (c) with a weak substrate, the metal is deformed by the tensile stress and the scale begins to buckle (dashed line marks the original metal interface), the buckle may crack and spall during cooling; (d) the buckle can continue to grow. leading to increased growth stresses due to the geometry and the range of diffusion distance for Al.
12 The Oxidation Behaviour of O D s Iron Alrirninides
197
Based on this scale buckling model, any mechanism which results in strengthening of the substrate can affect scale morphology under the appropriate conditions. Oxide dispersions can improve the creep strength of ODS alloys by several orders of magnitude by dislocation pinning (provided the alloy grain size is large) [38].Therefore, it is not unreasonable to expect that one effect of a dispersed oxide phase on high-temperature oxidation behaviour would be to promote a flatter scale by minimizing deformation of the substrate at temperatures where the dispersion-free alloy is quite weak. This mechanism can explain the present observations of a flatter scale on ODS FAS (whether dispersed with only AlzO, or A120, and a RE oxide). Figure 1.The oxidedispersoid strengthening effect has been observed for FAS with Y,O, [13], but has not been evaluated for Al,O,-dispersed FAS because of an inability to significantly increase the grain size. All of the oxidation tests in this study used alloys in the asextruded, fine (= 1 pm) grain condition. Nevertheless, it is possible that some high temperature strengthening of this alloy has also occurred. Similar strength-scale morphology considerations can be applied to a comparison of the oxidation behaviour of cast FAS, FeCrAl, and P-NiAl. Because these alloys typically would not be considered for structural applications at the oxidation temperatures used in the present study, the relevant high temperature strength values are rarely measured. However, a qualitative estimation of the relative strengths is possible from simple visual examination of as-oxidized specimens.These indicate that NiAl is able to support its own weight in the crucible tests at 1200°C and 1300"C, and does not deform as readily as cast FeCrAl or Fe,AI compositions. A similar comparison indicates that cast FeCrAl is somewhat stronger than cast Fe,Al. Based on the above model of the effect of substrate strength on scale morphology, these observations explain why the alumina scale on FAS is so highly convoluted while the scale on NiAl after the same exposure (2 h at 1200°C) is virtually flat [39]; the stronger NiAl resists deformation.The alumina scale on cast, undoped FeCrAl[18] has fewer convolutions than that observed on Fe,AI (Figure l a ) consistent with the observation that FeCrAl may be somewhat stronger.
12.4.2 The Effect of an A1,0, Dispersion The R E effect is generally described as an improvement in scale adhesion, a change in the scale growth mechanism (resulting in a reduction in the growth rate), and a modification of the scale microstructure [40].The results presented here for Al,O,-dispersed FAS indicate that this oxide did not produce any of the typical R E effects. The aAl,O, scale formed at the same rate (Table 2) and had the same scale morphology and grain structure (Figures 5 and 6) as a scale formed on an undoped alloy. Thus, there is no reason to believe there was a reduction in growth stresses, yet a flatter scale was formed. Clearly, the present observations with Al,O,-dispersed FAS indicate that, while a R E addition might inhibit scale buckling by reducing growth stresses [41], RE doping is not a necessary condition in achieving a flat scale. However, due to the lack of mechanical properties data on these materials, it is not possible at this time to conclude that strengthening the substrate is sufficient to prevent scale buckling. This issue will require more study, (for example references 42 and 43).
198
D.A. Pint, PI? Torrorelli and
I.G. Wright
Previously reported beneficial effects for Al,O,-dispersed alumina-formers [28,29] may be explained by the strengthening mechanism described above. A flatter scale would remain in better contact with the substrate than a convoluted scale. During longer-term testing, the flat scale fails as a result of the growth of interfacial voids, which limit contact between the scale and substrate. When the void fraction reaches a critical level, the cooling stresses are sufficient to spa11 the scale. Models have been presented which suggest that indigenous sulfur accelerates the growth of interfacial voids [40,44-451. R E additions are able to inhibit this detrimental role of S, but AI,O,dispersed FAS with no R E addition may not be able to do so. It has been suggested that an A120, dispersion could getter S at the dispersoid-matrix interface [46]. If that mechanism was operative, it was not sufficient to improve scale adhesion in this case. As shown in Figure 8, the Al,O,-dispersed FAS exhibited a shorter lifetime than that of cast, undoped FAS. The lifetime was reduced as the A1,0, content increased. This may be a result of particle coarsening. Large AI,O, particles in the substrate may allow rapid transport of 0 into the substrate when these particles come into contact with thc scale, as has been observed when alumina particles or fibers are introduced into alumina-forming matrices [47,48].
12.4.3 Effects of RE Dopants Based on this study, it appears that when added in the proper amount, RE oxide dispersions in FAS produce all of the RE effects described above (to some degree). Focusing on Y,O,-dispersed FAS scale adhesion was improved (Figures 13 and 14), a microstructure with fine grains at the gas interface (Figure 5c) and columnar grains in cross-section (Figure 6b) was observed, and there was a minor reduction in the scale growth rate (Table 2). This scale microstructure and reduction in scale growth rate is similar to that observed for ODS FeCrAl [18]. Based on lROtracer experiments, the addition of Y,O, or ZrO, to FeCrAl inhibits the outward transport of Al. resulting in growth primarily by 0 inward transport 131-33,391. The reduced growth rate and columnar microstructure on Y,O,-dispersed FAS is consistent with a similar growth mechanism modification in this case. For each cation addition, it is expected that therc is an optimum dopant level. Therefore, it is not possible at this time to conclude that any of the oxides examined, except A1,0,, are ineffective additions. The use of 0.2 cation% appeared to be “overdoping” for additions of La, Ce and Sc.The scale convolutions produced by the La addition (Figure 10) are similar to those observed for scales grown on Y-implanted FeCrAl [49-511. In both cases, an excessive amount of RE in or near the scale may result in detrimental effects. For example, an excess quantity of segregated cations on scale grain boundaries may allow rapid transport of 0.Alternatively, the coarsening of oxide particles in the alloy may allow accelerated oxidation, similar to that suggested for an A1,0, addition. If present in lower concentrations, these dopants may well be effective in improving scale adhesion. For the case of Y,O, additions, 0.2 YOY appeared to be the optimum level. Higher Y,O, contents in ODS FeCrAl have been observed to slightly accelerate oxidation [21]. For the other additions, lower dopant levels may also be more effective for
12 The Oxidution Behnviour of O D s Iron Aluminides
199
reasons similar to those discussed above. Assuming that the commercial Kanthal APM alloy has an optimized ZrO, content of O.O6at% Zr, the 0.2 YOZr addition used in this study may not have produced the best effect that could be achieved with Zr-doping. Lower (0.05 '%) dopant levels of Zr and I-If are currently being investigated that may improve performance (Figure 11). Although lower R E oxide additions may produce the optimum doping effect on the a-A1203scale, the R E addition cannot be reduced below the oxide volume fraction required for strengthening. Conversely, if the R E oxide content needed for strength improvement reduces the oxidation resistance, then the use of the alloy in high-temperature environments will be problematical. Yttrium may be unique due to its low solubility in these alloys. A fine Y,O,-AI,O, oxide dispersion coarsens very slowly and seems to provide the necessary requirements for an optimal RE effect. Elements which are more soluble in the matrix are expected to show faster particle coarsening and diffuse faster into the scale [40]. Based on the results with a Zr alloy addition and R E oxide dispersions, it appears that RE additions to Fe,Al are not as effective in improving scale adhesion as similar R E additions to FeCrAl and NiAI. (Previous work on ODS FeCrAl and P-NiAI alloys [18,39], has demonstrated that Y and Zr dopants are equally effective in improving alumina scale adhesion.) In Zr-doped NiAl and Zr0,-dispersed FeCrAl, the R E addition results in a flat adherent scale with minimal interfacial void formation. In the case of cast FAL (0.1 YOZr), it appeared that the substrate was deformed and the resulting convoluted scale was subject to spallation, Figure 4a. As suggested earlier, this could be a result of the weak substrate. Although the Zr addition may improve adhesion to some degree by gettering S [52-541 or by decreasing growth stresses [36,41], etc., there may still be sufficient stress at temperature to deform the relatively weak Fe,Al substrate. The Zr addition would not likely provide much strengthening effect, compared to the cast, undoped FAS, and in fact deformed significantly during crucible oxidation tests at 1200°C and 1300°C. The inability of a Zr alloy addition to prevent scale spallation on Fe,AI at the temperatures used in this study suggests that there are additional first-order factors, besides sulfur gettering [52-541, which affect scale adhesion. (Zr does have a beneficial effect on scale adhesion at lower oxidation temperatures, where the substrate is stronger and the scale is thinner [10,19].) There is sufficient Zr present to getter the S present in this alloy and yet there is significant scale spallation. However, many of the results on the effect of S were conducted with Ni-base superalloys, which are much stronger at these temperatures [42]. Indigenous S may be less of a factor for a weak substrate. Smialek et al. [55] also observed scale spallation on Fe-40 %A1 with Hf and Zr alloy additions. They concluded that the ina6ility of these RE additions to prevent scale spallation was related to a larger difference in the coefficient of thermal expansion between a-Al,O, and iron aluminides than other alumina-formers. The present results suggest that, rather than excessive damage during cooling, scale spallation on RE-alloyed iron aluminides was a result of scale buckling which occurred isothermally. Void formation might have been suppressed by the Zr addition, but scale buckling appeared to lead to cracking and spallation (shown schematically in Figure 18c) during cooling. In the case of Y,O,-dispersed FAS, the substrate may be sufficiently strengthened that there is some improvement over a Zr addition to the cast alloy. The deformation
200
U.A. Pint, t?K Tortorelli mid
I.G Wright
mechanism may no longer lead to a spallation-prone scale morphology. However. a second mechanism, the growth of interfacial voids, appeared to limit performance relative to ODS FeCrALThe excessive formation of interfacial voids may be similar to that observed on P-NiAI [56]. Due to the diffusion characteristics of the ordered Fe,Al matrix (unlike a FeCrAl alloy), these materials may be inherently more susceptible to Kirkendall-type voids at the oxidation front. This idea and others are currently being investigated. The higher spallation rate of ODS FAS means an increased A1 consumption rate. l h i s reduces the potential lifetime benefits of the high Al content of Fe,AI 16,111. l h u s , attempts to improve scale adhesion are a critical step in the development of these materials.
12.5 Conclusions 1. At 1200°C and 1300"C,incorporation of a reactive element in Fe,A1 as an oxide dispersion is more effective in improving scale adhesion than a comparable elemental addition. It is proposed that this is due to strengthening of the substrate. 2. A1,0, dispersions also strengthen the Fe,AI substrate and reduce the amount of scale buckling. However, the scale grain structure is not modified and no long-term benefit to oxidation resistance is achieved. 3. Currently, the best ODS Fe,AI spalls more readily than commercial ODS FeCrAl alloys.This is attributed to more rapid interfacial void nucleation and growth in the case of ODS Fe,Al. 4. Comparison of the oxidation behaviour of Fe,Al with various oxide dispersions at 1200°C indicates that n o addition was superior to Y , 0 3 in conferring a beneficial effect.
12.6 Acknowledgements The authors would like to thank J.R. DiStefano, J. H. DeVan, D. E Wilson and K. B. Alexander at ORNL for their comments on the manuscript and 11. Rohrig for the German translations; R. K . Williams for assistance with preparing the oxide powders; K . B. Alexander and P .J. Maziasz at ORNL and R. N. Wright at INEL for TEM contributions; and K. S. Blakely, L. D. Chitwood, M. Howell, J. Weaver and J. W Jones for assistance with the experimental work. This research was sponsored by the US. Department of Energy, Fossil Energy AK&TD Materials Program under contract DE-ACOS960R22464 with Lockheed Martin Energy Research Corporation. BAP is supported by the U.S. Department of Energy Distinguished Postdoctoral Research Program administered by the Oak Ridge Institute for Science and Education.
12 The Oxidation Behriviour of O D s Iron Aluminides
201
12.7 References [ 11 E Starr, A. R. White, B. Kazirnierzak: in: Materials for Advanced Power Engineering 1994, Eds. D. Coutsouradis, et al.; Kluwer Academic Publishers, Dordrecht, 1994 p. 1393. [2] W.J. Quadakkers, K. Bongartz, E Schubert, H. Schirster: in: Materials for Advanced Power Engineering 1994. Eds. D. Coutsouradis. et al.; Kluwer Academic Publishers, Dordrecht, 1994 p. 1533. [3] W J. Quadakkers, M .J. Bennett: Mat. High‘renip. 10 (1994) 126. [4] W J. Quadakkers, K. Bongurtz:Werkst. Korros. 4S (1994) 232. [ 5 ] M. J. Bennett, R. Perkins, J. B. Price, N. Starr: in: Materials for Advanced Power Engineering 1994, Eds. D. Coutsou-radis. et al.: Kluwer Academic Publishers. Dordrecht, 1994 p. 1553. [6] I. G. Wright, B. A. Pint, C. S. Simpson, I? E Tortorelli: “Hizh Temperature Oxidation Life Characteristics of ODS Fe,AI” Materials Science Forum, in press (1996). [7] C. Sykes.J. W Bampfylde:J. Iron and Steel Inst. 130 (1934) 389. [8] I? 7brnaszewicz,G. R. Wallwork:Rev. High Temp. Mat. 4 (1978) 75. [9] C. G. McKamey,J. H. UeVan,I? E Tortorelli, V K. Sikka:J. Mater. Res. 6 (1991) 1779. [ 101 I? E Torrorelli, J. H. DeVan: in: Processing, Properties and Applications of Iron Aluminides, Eds. J. H. Schneibel, M. A. Crimp; TMS,Warrendale, PA, 1994,p. 257. [lI] J. H. DeVan, I? E Tortorelli, M. J. Bennett: in: Proc. Eighth Annual Cont Fossil Energy Materials. CONF-9405143, Eds. N. C. Cole. R. R. Judkins: U.S. Department of Energy. Washington,D.C., 1994,p. 316. [12] C. G. McKamey,P .J. Mazimz, J. W Jones: J. Mater. Res. 7 (1992) 2089. [13] I. G. Wright, C. G. McKamey, B. A. Pint in: Proc. Ninth Annual Conf. Fossil Energy Materials. CONF-9504204, Eds. N.C. Cole, R.R. Judkins; U. S. Department of Energy. Washington, D.C.. 1995,p. 355. [I41 I? I:’ Tortorelli, J. H. DeVan, G. M. Goodwin, M. Ilowell in: Elevated’remperaturc Coatings: Science and Technology I, Eds. N.B. Dahorte, J. M. Hampikian, J. J. Stiglich; TMS, Warrendale, PA, 1995,p. 203. [15] I? I:‘ Tortorelli, I. G. Wright, G. M. Goodwin. M. Howell in: High Temperature Coatings 11, Eds, N.B. Dahorte, J. M. Hampikian;TMS, Warrendale, PA, 1996,p. 175. [16] 7:Raghavan, L. Steeds, R. Petkovic-I,uton: Met.Trans. 13A (1982) 953. [17] 1. S. Polkin, E. K Ivanova, B. I? Matyhin: in: Structural Applications of Mechanical Alloying, Eds, F. H. Froes, J. J. deBarbadillo,ASM International, Materials Park, OH, 1990, p.131. [18] B. A. Pint, A. J. Garraft-Reed,L. W.Hobbs: Mat. HighTemp. 13 (1995) 3. [ 191 I? E Tortorelli, J. H. DeVan, U. K. Abdali:in: Corrosion 93, NACE, Houston,TX, 1993,p. 25811-9. [20] M. J. Bennett, M. R. Houlton: in: High Temperature Materials for Power Engineering, Eds. E. Bachelet, et al.; Kluwer Academic Publishers, Amsterdam, 1990,p. 227 [21] W J. Quadakkers, K. Schmidt, H. Grubrneier, E. Wallura: Mat. High Temp. 10 (1992) 23. [22] H. T Michels: Met.Trans. 9A (1978) 873. [23] K. I-. Lcithra, E. L.Hall Oxid. Met.26 (1986) 385. [24] B. A. Pint, L. W Hobbs: in: High Temperature Ordered Intermetallics VI, Eds. J. Horton, S. Hanada, I. Baker, R. D. Noebe, D. Schwartz; Symp. Proc. Vo1.364, Materials Research Society, Pittsburgh, PA, 1995,p. 987. [25] B. A. Pint:“The Oxidation Behaviour of Oxide-Dispersed P-NiAI: I. Short-Term Cyclic Data and Scale Morphology,” submitted to Oxidation of Metals. [26] B. A. Pint, K. B. Alexander, I? I;: Tortorelli: in: High Temperature Ordered Intermetallics VI, Eds. J. Horton. S. Hanada, I. Baker, R. D. Noebe, D. Schwartz; Symp. Proc. Vol. 364, Material Research Society, Pittsburgh, PA, 1995,p. 1315. [27] J. H. DeVan, F1 E Tortorelli: Corr. Sci.35 (1993) 1065. [28] J. K. Tien, E S. Pettit: Met.Trans.3 (1972) 1587. [29] L. M. Kingsley,J. Stringer: Oxid. Met. 32 (1 989) 371. (301 D. l? Whittle,J. Stringer: PhiLTrans. Koyal SOC.London, Serial A295 (1980) 309. [31] W J. Quadakkers, H. Holzbrecher, K. G. Briefs, H. Beske: Oxid. Met.32 (1989) 67. [32] B. A. Pint, J. R. Martin, L. W llobbs: Oxid. Met.39 (1993) 167. [33] D. Clemens, K. Bongartz, W Speier, R. J. Hussey, W.J. Quadakkers: Fresenius’ J. Anal. Chem. 346 (1993) 318.
202
B.A. Pint, PI? Tortorelli and 1.G Wright
[34] B. A. Pint: .*Study of the Keactive Element Effect in ODS Iron-Base Alumina-Formers,” iMaterials Science Forum. in press (1996). [3S] K. B. Ale.rander, B. A. Pint: Oak Ridge National Laboratory, unpublished rescarch 1996. [36] E: H . Sfoft,A.Atkinson:Mat. High Temp. 12 (1994) 195. [37] H. E. Evans: Int. Mat. Rev. 40 (1995) 1. [38] E: H. Froes, J. J. deBarbadillo: in: Structural Applications of Mechanical Alloying, Eids. F. H. Froes, J. J. deBarbadillo, ASM International, Materials Park. OH, 1990,p. 131. 1391 B. A. Pint: Ph. D Thesis. Massachusetts Institute of lechnology. Cambridge, MA. 1Y92. [40] B. A. Pint: Oxid. Met. 45 (1996) 1. (411 E A. Golightfy,I;: H. Stott, G. C. Wood:Oxid. Met. I 0 (1976) 163. 1421 W J. Britzd/ey,J. D. Whittenberger: Mat. Sci. and Eng.A163 (1993) 33. (431 A. Strawbridge. H. E. Evans: “Spallation of oxide scales from NiCrAlY overlay coatings,” Materials Science Forum. in press (1996). 1441 H. J. Grabke, D. Weimer,H. Viefhaus:App. Surf. Sci. 47 (1991) 243. [45] B. A. Pint: “On the Formation of Interfacial and Internal Voids in a-Al,O, Scales”, submitted to Oxidation of Metals. [46] D. G. Lees: Oxid. Met.27(1987) 75. [47] J. Doychnk,J. A. Nesbitl, R. D. Noebe, R. R. Bowman:Oxid. Met. 38 (1992) 45. [48] I? E Tortorelli, J. H. DeVan, C. G. McKarney, M. Howell: Ceram Trans.29 (1991) 961. [49] J. G. Stneggil,A.J. Shuskus:J.Vac. Sci.Tech. A4 (1986) 2577. 1501 J. Jedlinski, G. Borchardt, S. Mrowec:Werkst. Korros. 41 (1990) 701. [51] W J. Quadakkers, J. Jerllinski, K. Schmidt, M. Krasovec, G. Borchardt, H. Nickel App. Surf. Sci.47(1991) 261. 1521 A. W Funkenbllsh, J. G. Smeggil, N. S. Bornstein:Met.Trans. 16A (1985) 1164. [53] J. L. Smialek: Met.li.ans.22A (1991) 739. (541 D. R. Sigler: Oxid. Met.32 (1989) 337 1551 J. L. Smialek,J. Doyclzak, D. J. Gaydosh: Oxid. Met. 34 (1990) 259. [56] M. W Brumni, H. J. Grnbke: Corr. Sci. 34 (1993) 547.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
13 High Temperature Corrosion Behaviour of Iron Aluminides and Iron-AluminiumChromium Alloys J. Klower
13.1 Introduction Aluminides based on the intermetallic phases Ni3A1and Fe,AI are considered both as structural materials and as coatings for high temperature applications [1-6]. Their excellent corrosion resistance is due to their forming a dense, protective alumina scale. Alumina, especially a-A1,03, shows low rate constants even at temperatures above 1000°C [7]. Unlike chromia, which is formed on conventional stainless steels and nickel base alloys, alumina does not evaporate above 1000°C [8] and it is even stable in oxygen deficient atmospheres. Table 1 shows the equilibrium oxygen partial pressure of several metal oxides. i.e. the minimum oxygen partial pressure which is required for oxide formation. A comparison with Table 2, in which typical oxygen partial pressures of industrial processes are listed, clearly demonstrates, that in coal gasification processes and petrochemical plants stable oxide scales can only be expected on alumina formers. Chromia becomes unstable at the low oxygen partial pressures encountered in such processes. There should be no significant difference between alumina scales formed on nickel aluminide and those formed on iron aluminide. Compared to Ni,AI, however, Fe3AI has some major advantages:
(1) The metal activity of iron aluminides is about two orders of magnitude higher than that of nickel aluminide, as can be seen in Fig. 1.The higher mctal activity results in a higher thermodynamical stability of iron aluminides. (2) The diffusion coefficient of aluminium in a ferritic iron-aluminium matrix is some orders of magnitude higher than in austenitic Ni,AI. Because of the low diffusion coefficient, diffusion in nickel aluminides is slow and aluminium depletion beneath the alumina scale and formation of nonprotective nickel oxides has been observed in Ni& [4]. (3) It is well-known that nickel base alloys are susceptible to the formation of liquid Ni-Ni3S, eutectics (melting point: 635°C). If once formed, the liquid phases can cause rapid wastage of the material. Fe-S compounds, on the other hand, become liquid not below 985 "C [7].
J. KIoiver
204
Table 1. Equilibrium partial pressure of several metal oxides. At oxygen partial pressures below those given in the table, the oxides are thermodynamically unstable (calculated with thermodynamical data after Barin [32]) Oxide
Po~11'002 ~
~
1.6 x 1 1 ~ 3 4 3.4 x 1@2? 9 x lo-" 4.9 x 10-15 9.5 x 10-14
213 Al, 0, 2/3 Cr20, 2NiO 112 Fe, O., 213 Fez 0,
Table 2. Typical oxygen partial pressures of high temperature processes -.
Typical oxygen partial pressurein bar
Process ~
Ref.
~
10-22-1030 10-25-10 30 Fluidised bed combustion 10"-10-5 Coal gasification z10-2'
(331
Petrochemical plants Gas cooled reactors
P31 1331 (121
Because of their excellent sulphidation resistance iron aluminides are considered for application in coal gasification and other processes, were high concentrations of sulphur in combination with low oxygen partial pressures are encountered. It is generally found that the corrosion resistance increase with increasing aluminium concentrations [2,10].High aluminium concentrations. however. decrease the ductility of iron aluminium alloys [ll] and can cause severe problems during casting,cold working and
- 4.0
- 3.0
-a
B
rn
-0
- zn
- 1.0
Fig. 1. Calculated and experimental activity o f
0
20
40
ATOMIC *A/ A1
60
aluminium in solid Ni-A1 and Fe-AI alloys at 1000°C (after Steiner and Komurek [Y]): dashed lines indicate measured values
13 High Temperature Corrosion Belzaviortr of Iron Aliinzinides
20s
welding. That means that for reasons of fabricability it is of interest to keep the aluminium concentration as low as possible. The present work has been undertaken to determine the minimum aluminium concentration which is required to provide sufficient corrosion resistance in a variety of corrosive environments and to find out whether the minimum aluminium concentration can be influenced by application of chromium.
13.2 Experimental Details 13.2.1 Alloys The compositions of the alloys investigated are listed in Table 3. The aluminium concentration was varied between 6 and 17 wt%, the chromium concentration between 2 and 10 wt%. All alloys contained 0.1 wt% hafnium to improve the scale adherence. Alloys #2, #5 and #6 additionally contained 0.13 to 0.18 wt% mischmetal (which is mainly consisting of cerium and lanthanum). The carbon concentration was normally between 300-500ppm; extra carbon was added to alloys #2,3,4,5 and 6 to increase the mechanical strength. The concentrations of sulphur and phosphorus were below 40ppm in all alloys investigated. Alloy 800H (1.4958) was included in the test programme for comparison purposes; its composition is listed in Table 3, too. Samples of the alloys to be investigated were cut from 10-kg laboratory size ingots that were vacuum induction melted. All alloys were hot rolled at temperatures between 1000 and 1100°C to sheets of about 4mm thickness. Coupons for corrosion testing with dimensions of about 15X25X3 mm3 were made from the sheets by machining. Prior to corrosion testing all specimens were ground to a 120-grit surface finish, ultrasonically cleaned and degreased in acetone. At least 3 specimens of each alloy were used in the tests. Table 3. Composition of iron-aluminium-chromium alloys investigated; MM: mischmetal, a mixture of reactive elements (mainly cerium and lanthanum); * composition of alloy 800H: nominal composition No. Alloy 1
2 3 4 5 6 7 8 9 10 I1 12
FeAl6 Cr2 FeAl8 Cr2 FeAl 10 Cr2 FeAl10 Cr2 FeAl12 Cr2 FeAl 15 Cr2 FeAll7 Cr2 FeAl6 Cr3 FeAl6 Cr8 FeAl6 CrlO FeAl8 CrlO FeAl10 CrlO 800 H *
Fe
Al
Composition in wt% Cr Hf
bal.
5.8 7.9 10.6 9.9 12.1 15.5 16.2 6.2 5.9 5.8 8.0 9.8 0.3
1.9 2.0 1.9 2.0 2.0 2.0 2.2 2.7 7.5 9.6 9.7 9.7 20
0.13 0.1 0.06 0.15 0.11 0.10 0.15 0.10 0.10 0.11 0.10 ~
c 0.003 0.02 0.013 0.01 0.017 0.022 0.067 0.003 0.004 0.004 0.004 0.005 0.07
Others 0.13 MM 0.18MM 0.18MM
31 Ni
206
J. KIower
13.2.2 Procedure of Corrosion Testing The compositions of the corrosive gases are listed in'rable 4.All test gases were selected with regard to practical applications. Cyclic oxidation tests were conducted i n dry air to simulate the conditions of industrial furnaces and electrical appliances. The carburisation tests were carried out at 85O"C, 1000°C and 1100°C in oxygen deficient CH,/H, gas, which was selectcd with regard to applications in petrochemical plants. The carbon activity of the gas was 0.8. Since metal dusting conditions would require a carbon activity of at least 1 , metal dusting was not expected, although the gas is highly carburising. Table 4. Composition, temperature range and oxygen partial pressure of the test gases
_________
__
Gas (inVol%)
T
air CH,/H,; ac
1100°C 85OOC 1000°C 1100°C 850°C 650-850°C 550°C
= 0.8
0.25% C1, / 20 %O, I A r 1 and 5 and 10 %SO, in air 1%H2S/ lOYoC0, / H2
1% Po, -28
-25
-24 -26
Sulphidation tests were carried out in both SO,/air and H,S-containing environments. Sulphidation is a frequent problem in coal conversion plants. Oxidising SO,containing gases are encountered in conventional power plants, while oxygen deficient H,S gases are typical for coal gasification plants. Since process gases may additionally contain chlorides or chlorine, the resistance of iron-chromium-aluminium alloys to chlorine gases was tested, too. The oxidation tests were carried out in a chamber furnace at 1100°C.All other tests were performed in tube furnaces with the tube being sintered alumina and the crucibles being made from heat-resistant, alumina forming Fe-20wt%Cr-5wt%Al alloys to avoid contamination of the specimens by evaporating chromium oxide. Premixed gases were used for these corrosion tests. The oxygen deficient gases (CH, and H,S) were passed over silica to keep them dry.The oxygen partial pressure of the gases was constantly controlled by determining the dewpoint of the gas. The corrosion tests were performed as cyclic test. One cycle comprised 1.5 hours heating, 16 hours at test temperature and slow cooling to room temperature. After at least 3 cycles the mass change of the specimens was determined by weighing. The mass change data were automatically divided by the surface of the sample. Each value represents the average of three specimens with the scatter band being less than ? 15 %. The entire testing time was 1008 hours (42 cycles) in air, CI, and CH, and 2016 hours (84 cycles) in sulphur containing gases. After the corrosion testing two specimens of each alloy were investigated by optical microscopy. Both depth of internal corrosion and metal loss were measured at 5 locations and the maximum values of both specimens were determined. The "metal loss (A-B)/2" comprised both adherent scale and spalled scale, the latter being calculated
13 High Temperntitre Corrosion Rehnviour of Iron Alicrninides
---- - _ _ - - _ _ _ _ _ _ _ _
207
Fig. 2. Metallographicdetermination of corrosion attack
according to Fig. 2 by measuring the thickness of unaffected material and subtracting it from the original thickness of the coupon. Therefore the total corrosion attack Ca is a theoretical maximum value representing the sum of maximum internal corrosion and maximum metal loss. Microprobe analysis was carried out at the Metallabor of the Metallgesellschaft, Frankfurt, on selected specimens.
13.3 Results 13.3.1 Results of Oxidation Tests The oxidation behaviour was tested in dry air at 1100"C.The mass change versus time curves of the alloys without mischmetal are shown in Fig. 3. It can be seen that the mass changes are low and the differences between the alloys are small. The values ranging between 7 and 14g/m2 after 1008 hours (42 cycles) exposure are very similar to those found for commercial Fe-20Cr-5AI heating element alloys [ 131. No significant spalling was observed during testing.Taking into account that 42 cycles with a holding time of 16 hours each cycle result in 672 hours total exposure at test temperature and assuming parabolic behaviour, the parabolic rate constants were calculated and found to be between 2.7X g2/m4s(Fe-AllO-Cr2) and 8 X 10-13g2/m'%(Fe-A16-Cr2). The average mass changes for all alloys investigated after 42 cycles (1008 hours total exposure) and the results of the microscopical investigations are listed in Table 5. The results of alloys #2, #5 and #6 with additions of mischmetal and those of alloy 800H are included in Table 5, too. It can be seen that all Fe-Al-Cr alloys without mischmetal show low corrosion rates if compared to alloy 800H. Alloys containing additions of 0.13 t o 0.18 wt% mischmetal, however, show mass changes twice to three times as high as the alloys which contain only 0.1 wt% hafnium. Mischmetal does not effect the scale thickness of the Fe-Al-Cr alloys, but it obviously increases the depth of internal oxidation. While alloys without mischmetal show a depth of internal corrosion of 40mm and less, those alloys containing mischmetal were severly attacked by the formation of internal oxides. The effect of internal oxidation caused by mischnietal becomes more pronounced, as the aluminium concentration of the alloy is increased. Alloys containing 15 wt% aluminium show internal oxides over the entire cross section after 42 cycles of exposure.
J. Klower
208
Table 5. Mass change, maximum depth of internal oxidation and maximum scale thickness '5"of iron-aluminium-chromium alloys after 1008 hours of cyclic exposure in air at 1100°C ')These alloys contain 0.15- 0.18 wt% mischmetal 2, Average values of three specimens.The scatter hand is always less than 2 10% No.
Alloy
1 2 3 S 6 8 9 10
FeAl6 Cr2 FeAl 8 Cr2 2, FeAl 10 Cr2 FeAl 12 Cr2 I ) FeAl 15 Cr2 I ) FeAl6 Cr3 FeAl6 Cr8 WAI 6 CrlO FeAl8 CrlO FeAl10 CrlO 800H
Average 2, mass change in e/m'
~~~~
11
12
~~
Depth of internal corrosion in pm
Scale thickness "S" in pm
~~
13.8 2 0.3 26 8.1 ? 0.2 32
~~
6
40 72 20 123 >2000
45 12.9 f 0.1 8.8 -c 0.5 9.6 -C 0.3 11.7 2 0.3 0.7 2 0.6 - 320
10
6 8
10 8 8 6 8 6 43
30 40 40
40 35 200
.."
..........Q
+8 (FeAI 0 C r 3 )
IFeAl 10 C r l o l
0 0
200
400
600
time i n h
000
iooo
Fig. 3. Mass change versus time plots of different iron-aluminium-chromium alloys after cyclic exposure in air at 1100°C.
1.7 High Tempernture Corrosion Behnviour of Iron Aluminides
209
Fig. 4. Qpical oxide scales on iron-aluminium-chromium alloys after 1008 hours of cyclic exposure in air at 1100°C. a) Fe-AI10-Cr2, b) Fe-A110-Cr10, c) Fe-AIlS-Cr2 with 0.18 wt% mischmetal
Typical micrographs of Fe-Al-Cr-alloys after oxidation testing are shown in Fig. 4. All alloys without mischmetal show a similar shape of the oxide scale.The scale is thin, with very few protrusions and a slightly wavy morphology. The scale appearance was affected neither by chromium nor by aluminium within the composition range investigated. None of the alloys showed significant void formation beneath the scale (Fig. 4a,b). A typical micrograph of an alloy containing 0.18 wt% mischmetal is shown in Fig. 4c. Large internal oxides are visible in the cross section.The scale had partially spalled off. Void formation, however, was not detected in any of the mischmetal containing alloys either.
J. Klower
210
13.3.2 Results of the Carburisation and Chlorination Tests The carburisation behaviour was tested in an oxygen deficient CH,/H2 atmosphere. Mass changes of Fe-1OAl-2Cr and Fe-15Al-2Cr after exposure at 1000°C are plotted in Fig. 5 in comparison to alloy SOOH, and the average mass changes after carburisation testing at 850°C and 1100°C are summarised in Table 6. Literally n o mass change was measured for the Fe-Al-Cr- alloys investigated, while alloy XOOH suffered severe carbon pick-up in CII,/H2 at 850,1000 and 1100°C. The results of metallographic investigations are in agreement with the gravimetric measurements. It can be seen in Fig. 6 that alloy 800H shows significant internal carburisation, while the Fe-Al-Cr alloys are entirely unaffected by carbon. ' h e oxide scales of the Fe-Al-Cr alloys were too thin to be investigated by microprobc analysis after exposure in CHJH,; it is however, very likely that they consist of pure alumina. Table 6. Mass changes due to carbon pick-up after 1008 hours of cyclic exposure in carburising CH,/H, gas with a carbon activity of a,=0.8 ~~
Mass change in g/m2 after 1008 h at T = 850 "C 1000°C
No.
Alloy
3 6 7
FeAl10 Cr2 FeAl15 Cr2 FeAll5 Cr2 (Alloy 800 H)
11W"C
2.4 i 0.3 2.8 2 0.3
1.7 ?- 1.2 257 ? 10
1025 Spalllng
329 t 8.8
~
300--
N
E \
D
5
200--
al
[II
C
m
l z 0
n m
II 8
100-
-
04 0
m -
E
- - -
200
400
'"
-
- .
600
" .
I
000
"
-
.
"
.
1000
time in h
Fig. 5. Mass change versus time plots of two diffcrent iron-aluminium-chromium alloys in comparison to alloy 800H after cyclic exposure in carburising CH,/H, gas atmosphere at 1000°C
13 High Ternperuliire Corrosion Behuvioiir of Iron Alioizinides
21 1
b)
Fig. 6. Typical corrosion attack of an iron-aluminium-chromium alloy and 800H after 1008 hours of cyclic exposure in carburising CH,/II, gas atmosphere at 1000°C a ) Fe-Al10-CR, b) Alloy 800H
Fig. 7. Corrosion attack of Fe-AllO-Cr2 after 140 hours of cyclic exposure in chlorinating 0.25 Cl2I2O%0,IAr gas mixture at 850°C
The tests of Fe-AllO-Cr2 in chlorinating CI,/O,/Ar gas at 850°C had to be finished after 140 hours because severe cvaporation of aluminium chloride took place. ‘fie mass change was 1138g/m2 after 140 hours, the scale thickness was about 400pm (spalled material not included). After the chlorination test the scale consists of alternating layers of different corrosion products, as can be seen in Fig. 7.
J. KIower
212
13.3.3 Results of Sulphidation Tests in SO,/air Sulphidation/oxidation tests were carried out on alloy Fe-AI 10-Cr2 and alloy 800H in 1,5 and 10% SO,/air at 650"C,in 1 YOSO,/air at 750°C and in 1 % ' SO,/air at 850"C.A typical mass change versus time plot at 650°C (10 % SO,) is shown in Fig. &The mass change of the Fe-Al-Cr alloy was only 3 g/m', while alloy 800H suffered sevcre mass loss due to spallation of corrosion products.The total corrosion attack was determined according to Fig. 2.The average mass changes after 2016 hours total exposure time (84 cycles), the depths of internal corrosion and the metal losses ((A-B)/2) are summariscd in Table 7 for all sulphidation tests performed in SO,. As expected, the corrosion attack increases with increasing SO,-concentration in the atmosphere. While both alloy 800H and Fe-AllO-Cr2 show reasonably low corrosion rates in l % SO,, tested in 10% SO, alloy 800H suffered a total corrosion attack of 205 pm after 1008 hours resulting in a corrosion rate of 0.9mdyear (linearly extrapolated).The corrosion rate of Fe-AllO-Cr2, on the other hand, was calculated to be less than 0,l mm/year. The increase of corrosion attack with increasing SO,-concentration is correlated with internal oxidation in Fe-AI10-Cr2 and an increased spallation in alloy 800H. The corrosion attack increases with increasing temperature, too. In Fe-AllO-Cr2 the dcpth of internal corrosion after 82 cyclcs was increased from 2 pni at 650°C to about 15 pm at 850°C. Significant scale spallation was not detected under any of the test con-
# 3 ( F e A l 10 C r 2
1
0-N
E
\
m
T=
2 -200--
650 q: / 10% S02/ a i r
m m m C
K
m m a
-400-
r
-600-
-
0
8
0
500
0
1 : 1000
1500
2000
t i m e in h
Fig. 8. Mass change versus time plot of Fe-AllO-Cr2 in comparison to alloy 800H after cyclic exposure in sulphidising 10 %SO,/air gas atmosphere at 650°C
1.3 High Temperature Corrosion Behaviour of Iron Aliiniinides
213
Table 7. Mass change, maximum depth of internal oxidation and metal loss ((A -B)/2) of Fe-AllO-Cr2 and alloy 800H after 2016 hours of cyclic exposure in SO,/air at 650,750,850"C
values of 3 specimens.The scatter band is always less than -k 10% Maximum value of 2 specimens.The value includes adherent and spalled scale (see Fig. 2)
I) Average
2,
_
~
Alloy T in "C
650 650 750 750 850 850
FeAIlOCr2 800 H FeAIlOCr2 800H FeAIlOCr2 800H
_
_
_
Mass change in g/m' after 2016 hours in SOJair 1% 5% 10%
so,
so,
soz
1 2 2 -513 3 -1
-3 -34
3 -650
~~
~
Depth of internal corrosion in pm after 2016 hours in SO,/air 1% 5% 10%
so:
so,
soz
Metal loss in pm after 2016 hours in SO,/air 1% 5% 10% SO? so, so,
2 6 0 40 15 70
10 40
20 50
0 0 0 130 0 50
0 0
0 155
ditions. Alloy 800H, on the other hand, suffered severe spallation of -573g/m2 even at 1 % SO,, when the temperature was increased from 650 to 750°C. Surprisingly spalling was reduced, when the temperature was further increased to 850°C. Typical micrographs after corrosion testing in SO,/air at 650°C are shown in Fig. 9. The reference alloy 800H shows a thick scale, which had partially spalled off. It can be seen that corrosion attack of the metal occurs mainly via gain boundary attack. In FeAIlO-Cr2, on the other hand, literally n o corrosion attack could be detected. As after the carburisation tests, the scale was found to be very thin and it could not be investigated by microprobe analysis.
Fig. 9. Typical corrosion attack of an iron-aluminium-chromium alloy and 800II after 2016 hours of cyclic exposure in sulphidising 10 O/oS02/airgas atmosphere at 650°C a) Fe-AIIO-Cr2, b) Alloy 800H
J. Klower
214
13.3.4 Results of Sulphidation Tests H,S The mass change vs. time curves after exposure at 550°C in 1 %,H,S/1O0/~CO2/Hz for alloys containing different concentrations of aluminium and chromium are plotted in Fig. 10. It can be seen that the alloys Fe-A115-Cr2, Fe-Al17-Cr2 and Fe-A116-CrlO showed very slight mass gain after 2016 hours exposure. Alloys Fe-AI6-Cr2. Fe-A1SCr2, and Fe-AllO-Cr2, on the other hand, suffered substantial mass losses due to spallation of corrosion products. The results of microscopical investigations are summarised inTable &The depth of internal corrosion was not influenced by either the aluminium or the chromium concentration. The alloy composition, however, was found to have a significant impact on the scale thickness. Microscopical observations revealed thick multilayered scales on all alloys containing 2 wt% chromium and less than 10% aluminium, as can be seen in Fig. 11.The thickness of the scale was found to decrease with increasing aluminium concentration and alloys containing 15 wt% aluminium show only local corrosion attack with a depth of less than 20 pm. The beneficial effect of aluminium can also be seen in Fig. 12, where the depth of corrosion attack is plotted versus the aluminium concentration. Chromium obviously has a n effcct on the scale appearance, too. Compared to alloy Fe-A16-Cr2 alloy Fe-A16-CrlO shows much lower corrosion rates and a thinner corrosion scale.The scale appearance, however, is similar to that found on Fe-A16-CrlO; the scale consists of an external scale with some cracks and a multi-component subscale which adheres to the base metal.
s\
m C
.r)
0
m C m
K
m
-200 -400 IFeAl 6 Cr21
-600
-800
t
-10004 0
2 (FeAl B C r
t
I
500
I
I000
1500
2000
time i n h
Fig. 10. Mass change versus time plot of different iron-aluminium-chroniiunialloys after cyclic exposure in sulphidising 1%H,S/ 10%CO,/H, gas atmosphere at 550°C
13 High Temperature Corrosion Behavioicr of Iron Aluminides
215
The scale on alloy Fe-AllO-Cr2 was investigated by microprobe analysis. The outer scale was found to consist of iron sulphide, while the inner scale consists of mixed alumina and chromium sulphide. Table 8. Mass change, maximum dcpth of internal oxidation and metal loss ((A-B)/2) of Fe-Cr-A1 alloys after 2016 hours of cyclic exposure in 1 %H2S/10%C0,1H2 at 550°C I ) Average values of three specimens.The scatter hand is always less than 2 10% *) Maximum value of 2 specimens.The value includes adherent and spalled scale (see Fig. 2) ~
_
_
_
~
No.
Alloy
Mass change in g/m?
Depth of internal corrosion in pm
Metal loss in pm
1 2 4 10 6 7
FeAl6 Cr2 FeAl8 Cr2 FeAl 10 Cr2 FeAl6 CrlO FeAll5 Cr2 FeAll7 Cr2
- 2685 - 522 - 352 + 53 +1 +1
20 20 30 40 50 10
670 280 225 120 45 1
Fig. 11. Typical corrosion attack of iron-aluminium-chromium alloys after 2016 hours of cyclic exposure in sulphidising 1%H,S/ 10%C02/H,gas atmosphere at 550°C a) Fe-AI6-Cr2, b) Fc-AllO-Cr2, c) Fe-A115-Cr2, d) Fe-A16-CrlO
J. Klower
216
T = 55OOC
0
4
8
12
A1 - concentration in wt %
16
20
Fig. lZ.'Total corrosion attack of different Fe-Al-Cr2 allovs as a function of aluminium concentrati& after 2016 hours of cyclic exposure in sulphidising i %H,S/lO %CO,/H, gas atmosphere at 550°C
13.4 Discussion 13.4.1 Oxidation Behaviour Compared t o alloy 800H all iron-aluminium-chromium alloys without mischmetal show superior resistance to oxidation and scale spallation. The oxidation behaviour of the iron-aluminium-chromium alloys investigated in this study is very similar to that found in commercial alumina forming Fe-20wtY0Cr-5 %A1 heating element alloys [13]. The wavy structure of the oxide scale is typical for iron-chromium-aluminium alloys containing other reactive elements than yttrium, whilc yttrium-doped alloys normally show smooth scales [14].The low parabolic rate constants of 2.7X10-'3-8X10-'7 g2/m4s,which are close to those measured by Snzialek for a-Al,O, - formation on Hfdoped FeAl at 1100°C (5X10-I3 -3X10-12g2/m4s)[6],clearly indicate that all Fe-Al-Cr alloys investigated in this study had predominantly formed slowly growing a-Al,03 scales.The slightly higher rate constants found by Smialek are probably caused by the higher concentration of reactive elements (zirconium and/or hafnium) in his study. It is evident from the results of the present work that "overdoping" with reactive elements has 5 detrimental effect on the oxidation behaviour and may increase the oxidation rate by an order of magnitude (4X10-'* g2/rn4s in Fe-Al12-Cr2 containing mischmetal compared to 2.7X g2/mJs in Fe-AllO-Cr2 without mischmetal). Increased oxidation rates, substantial internal oxidation and reduced life time caused by overdoping have been observed in yttrium-doped Fe-Cr20-AI5 alloys [13.15].in yttriacontaining ODs-alloys [16], and in Fe-Cr-Al-alloys containing cerium or mischmetal [17], too. While additions of mischmctal have a huge impact on the oxidation behaviour of iron-aluminium-chromium alloys, changes in the concentration of aluminium or chromium d o not effect the corrosion behaviour significantly. The differences between the
13 High Trniperarure Corrosion Behaviour of Iron Aluminides
217
oxidation rates of the mischnietal-free alloys are to small to derive any beneficial effect of chromium or aluminium within the investigated composition range. It should, however, be notcd that the tests wcre carried out on relatively thick specimens. If the material thickncss is decreased to foil dimensions high aluminium conccntrations may be beneficial to prevent the so-called “breakaway corrosion” i. e. corrosion caused by aluminium depletion of the substrate and subsequent formation of nonprotective oxides [18,19].
13.4.2 Corrosion in Hot Process Gases As expected. no carburisation attack at all was detectcd on iron-aluminium-chromium alloys after 1000 hours exposure in CH,/H, environments at 850”C, 1000°C and 1100°C. Since the formation of chromia and iron requires relatively high oxygen partial pressures, alumina is the only stable phase at the low partial pressure of the used gas. If once formed, alumina is impcrvious to carbon, provided the scale remains intact [20]. Excellent resistance to carburisation was also found for other alumina forming alloys like nickel aluminides [21] and Ni-Al-Cr alloys [22].The results of the present work show that 10 wt% aluminium are sufficient to prevent carburisation. It is expected, that the minimum aluminium concentration is evcn lower than 10 wt%. Iron-aluminium-chromium alloys should, however, not be applied in conditions where carbon activities above 1 are encountered. In CO-C0,-H, gas atmospheres, which were oversaturated with carbon (a, >> l), rapid material wastage by so-called “metal dusting” was observed [23,24]. As long as the carbon activity is below 1,however, excellent resistance of iron-aluminium-chromium alloys to carburisation can be expected even in oxygen deficient atmospheres. If the oxygen partial pressure is increased and/or components like sulphur or chlorine are added to the gas, reactions which are competitive to the formation of alumina become likely. One possible reaction is the formation of aluminium chloride in chlorinating environments. Although aluminium oxide has a more negative energy of formation than aluminium chloride, AlCI, may form in oxygen deficient areas beneath the surface if chlorine penetrates the oxide scale via cracks or grain boundaries in the A1,0, scale. Aluminium chloride AlC1, becomes volatile at temperatures above 173“C and tends to evaporate [25]. When moving to the oxygen rich surface. AICI, is transformed to A1,0,, which is thermodynamically more stable. Chlorine is released for further reactions, thus acting as a catalyst for oxidation. This process of chlorine enhanced oxidation is very similar to that described by Grubke for the chlorination of steels via formation of volatile iron chlorides [26].The process results in thick, multilayered, non adherent scales containing predominantly metal oxides. A microscopical investigation of the Fe-AllO-Cr2 specimens exposed to a chlorine-containing gas indeed reveals such thick, multilayered scales and gives evidence that the mechanism described above applies. In sulphidising SO,/air environments between 650 “C and 850 “C the iron-aluminium-chromium alloys did not suffer any significant sulphidation attack. Although formation of a-A1203cannot be expccted at temperatures as low as 650 to 850”C, the scales formed were protective enough to prevent internal and external sulphidation.
218
J. Klower
10 wt% aluminium and 2 wt% chromium in iron-aluminium-chromium alloys are obviously sufficient to cause a sulphidation resistance, which is much better than that of conventional chromia forming high temperature materials. If once formed, alumina scales should be protective in oxygen deficient sulphidising H,S-environments at intermediate temperatures, too [27]. 10 wt% aluminium in Fe-Al(x)-Cr2, however, are not sufficicnt to form alumina and prevent sulphidation in oxygcn deficient HzS-gasesat 55O"C.'lo create protective alumina scales, at least three methods can be imagined: 1. Preoxidation of the material at high temperatures ( > l O O O T ) and high oxygen partial pressures. 2. An increase in the aluminium activity in the respective alloy. 3. Additions of alloying elements that favour the formation of alumina. Preoxidation, i. e. the formation of protective alumina prior to exposure in H,S-containing gases clearly decreases the corrosion rates in H,S-containing coal gasification environments, as was shown by several investigators. Weber and Hocking [28] and Santorelli et at. [29] found that the corrosion rates of iron-chromium-aluminium alloys in H,S could bc dramatically decreased, if preoxidised specimens were exposed to H,S, while matcrials without any preoxidation suffered rapid sulphidation. Preoxidation prior to exposure, however, is oftcn not applicable. Scalcs obtained by preoxidation may be removed by erosion or spalling during in-plant service. If healing of the eroded or cracked scale is not possible, rapid sulphidation of the base material occurs after an incubation period. An increase in aluminium activity would be a more reliable method to increase sulphidation resistance in oxygen deficient, sulphidising environments. Thc results of the tests in H,S havc clearly shown, that an increase of the aluminium concentration increases the sulphidation resistance of iron-aluminium-chromium alloys dramatically. The results of the present work are in agreement with those found by Notesun and Johnson [lo], who determined a "threshold aluminium concentration" to prevcnt sulphidation in H,S. Alloys containing 12 wt% aluminium and more did not show any evidence of corrosion attack in H,S (pS2=9X1t7;p02'4xl0-") at 875 "C.The aluminium concentration in materials, which have to be manufactured as semi finished products (sheets, wires or tubes), however, is limited to about 10 wt% aluminium because of the manufacturing problcms (welding, cold rolling, casting) to be expected. For all these reasons an increase in chromium concentration may be the best solution to increasc the sulphidation resistance of iron-aluminium alloys. The present work has shown that chromium has a very beneficial effect on the sulphidation rcsistance in oxygen deficient gases. Such improvement of sulphidation resistance in H,S environments has also been found in nickel-aluminium alloys, where the corrosion rate decreased significantly, when 8 wt% chromium were addcd [22].Whether chromium increases the rate of alumina formation or contributes to the formation of protcctive spinell type oxides Fe(Cr,Al),O,, cannot be decided from the available results. It is also possible that chromium catalyses the transition from a less to a more protective type of A1,0,. It was for example found that chromium enhances the transition from @-A1,0, to a-A120, in Ni-A1 at temperatures above 1000°C [30]. The present results on the chromium effect are contradictory to those found by 'htorelli and &Van, who measured a detrimental effect of 2 - 5 wt% chromium in iron
1.7 High Temperature Corrosion Behnvioitr of Iron Ahtminides
21 9
aluminides at 800°C in H,S-containing gases [4,31]. It is possible, that the beneficial effect of chromium is limited to certain gas conditions and/or temperature ranges. Further investigations are necessary to explain the effect of chromium in alumina forming alloys at intermediate temperatures.
13.5 Summary and Conclusions The high temperature corrosion behaviour of different iron aluminides and iron-aluminium-chromium alloys containing 6- 17 wt% aluminium, 2-10 wt % chromium and additions of mischmetal has been investigated in both air and hot process gases. In air, carburising atmospheres and sulphidising/oxidising atmospheres all alloys showed excellent corrosion resistance due to the formation of protective oxide scales. It was found that the concentration of aluminium or chromium does not significantly effect the corrosion behaviour in any of these environments. 10 wt% aluminium, probably even less, are sufficient to enable the formation of protective Al,O,-scales even at temperatures as low as 650°C. Overdoping with reactive elements (mischmetal), however, causes high oxidation rates in air and should be avoided. In oxygen deficient H,S-environments, at least 12 wt% aluminium are required in alloys containing 2 wt% chromium.The sulphidation resistance in H,S is improved by chromium. If the chromium concentration is increased to 10 wt%, the minimum content of aluminium is decreased to 10 wt%. Since Fe-Al-Cr alloys with high concentrations of chromium and aluminium cannot easily be manufactured by conventional metal working, the application of “composite materials” should be considered. An experiment carried out at VOEST-ALPINE GmbH, Linz, has shown, that conventional boiler steels (17CrMo44) can successfully be cladded with Fe-AllO-Cr2 by hot rolling.
13.6 Acknowledgements The author wishes to thank Dr. U. Heubner and Dr. L! Brill for reading the manuscript and making many valuable suggestions for corrections and improvements. Thanks are also due to Mr. H.-G.Schlutt and Mr. If. Niecke for carefully carrying out the high temperature corrosion tests. The financial support of the “Ministerium fur Wirtschaft, Mittelstand und‘rechnologie des Landes Nordrhein-Westfalcn” is gratefully acknowledged.
13.7 References [l] G. Saurhofjc “Intermetallics”, VCH Weinheim, 1995. [2] J.V Cathcart: in: C.C. Koch et al. (eds.): “High temperature ordered intermetallic alloys I” MRS Symp. Proc. Vol39 (1985) 445.
220
J. Klower
[3] R E Tortorelli,J.H. deVnn, U.K. Abdnli: Corrosion 1993, Paper No. 258, NACE International. Houston,TX, (1993). [4] P E Tortorelli, J.H. deVan: in: S.H. Whang et al. (eds.) “High temperature aluminides and intermetallics”, Elsevier Applied Science, Essec (1992) 573. [S] VK. Sikka, S.C. Deevi: in: K. Natesan, P. Ganesan and G. Lai (eds.): “Heat resistant materials II”ASM, Metals Park, Ohio, (1995) 567. [6] J. L. Smialek, J. Doychak, D.J. Gaydosh:Oxid. Metals 34 (1990) 259. [7] I? Kofstad: “High temperature corrosion”, Elsevier Appl. Sci. Publ., London, New York. 1988. [S] S. Leistikov: in: A. Rahmel (ed.): Aufbau von Oxidschichten und ihre technische Bedeutung. DGM, Oberursel, Germany, (1983) 33. 191 A . Steiner, K . L . Komarek:Trans,AIME230 (1964) 786. (101 K. Natesnn, R.N. Johnson: in: K . Natesan, P. Ganesan and G. Lai (eds.): “Heat resistant materials II”ASM,Metals Park, Ohio, (1995) 591. [ l l ] M.J. Marcinkowsky, M.E. Taylor:Journal of Mat. Sci. 10 (1995) 406. [ 121 W Bnkker: “Mixed Oxidant Corrosion in Nonequilihrium Syngas at 540°C” Doctoral thesis. Delft, NL. 1995. [13] J. Klower, G. Li: accepted for publication in “Materials and Corrosion” [14] G.H. Meier, ES. Pettir, J.L. Smialek: Materials and Corrosion 46 (1995) 232. [15] E Clementdot, J.M. Gas, J.C. van Duysen: in: R. Streiff at al. (eds.) “High Temperature Corrosion and Protection of Materials” 1,es editions de physique, Les Ulis Cedex, France. 1992. p. C9-291. 1161 WJ. Quadakkers, K. Bongartz:Werkst. Korros.45 (1994) 232. [17] J. Deakin, G.C. Wood, EH. Stott, V Prunier: Report EDF-DER 1995, Electricit6 de Francc, Moret sur Loing, France, 1995. [18] WJ. Quaddakers, M.J. Benneft:Mat. Sci. and Tech. 10 (1994) 126. [19] M.J. Bennett, R. Perkins, J.B. Price, E Staur: in: D. Coutsouradis et al. (eds.): “Materials for Advanced Power Engineering” Kluwer Academic Publ. 2 (1994) 1553. [20] H.J. Grabke, A. Schnaas: Oxid. Metals 12 (1978) 387. [21] J. Klower, U. Brill: Z. Metall,Heft 9 (1992) 1. [22] J. Klower, G. Sauthofi D. Lefzig:Corrosion 1996, Paper No. 144, NACE Int., Houston.TX. 1996. 1231 H.J. Grabke: MPI fur Eisenforschung, Dusseldorf, Germany; private communication. [24] H.J. Grabke, R. Krajak, J.C. Nava Paz: Corrosion Sci.35 (1993) 1141. [25] R. Bianco, R.A. Rapp, N.S.Jacobsen: Oxid. Metals 38 (1992) 33. [26] E. Reese, H.J. Grabke:Werkst. Korros. 43 (1993) 547. [27] J.E Norton, T R Levy: Materials and Corrosion 46 (1995) 286. [28] J.K.R. Weber, M.G. Hocking: Oxid.Metals32 (1989) 1. [29] R. L. Santorelli, J.E Norton, E Bregani:Werkst. Korros. 41 (1990)669. [30] M . W Rrumm, H.J. Grabke:Corrosion Sci. 33 (1992) 1677. [31] J. H. deVan:T.Grobstein and J. Doychak (eds.): “Oxidation of f Iigh Temperature Interrnetal1ics”The Metallurgical Society,Warrendale, PA, (1989) 107. 1321 I. Barin:‘Thermodynamical data of pure substances”VCH Weinheim, 1993. 1331 V L . Hill, H.S. Meyer: in: M.F. Rothman (ed.): “High temperature corrosion in energy systems”, AIME, Warrendale PA, (1986) 29.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
14 Oxidation-Sulphidation of Iron Aluminides EH. Stott, K.T. Chuah and L.B. Bradley
14.1 Introduction There is considerable interest [l] in the high-tcmperature corrosion behaviour of alloys based on iron aluminides (FeAI and Fe,AI) for applications in gaseous environments of high sulphur and low oxygen activity, such as those pertaining in coal-conversion or catalytic-cracker plant; it is well established that Al,O,-forming alloys are more resistant than Cr,O,-forming alloys in such environments [2,3]. Indeed, the development of aluminized surfaces by pack aluminizing conventional alloys can result in considerable improvements in degradation resistance [4,5]. An important conclusion from this earlier research was that protection against sulphidation is not merely due to the presence of an effcctive barrier, the A1,0, scale, but also to the preferential re-establishment of AI,O, scale, rather than basis metal sulphides, following breakdown of the scale and penetration of environmental specics to the alloy substrate. For aluminized alloys, protcction is maintained as long as the aluminium concentration at the interface remains above some critical value [5]. Aluminized alloys have a finite supply of aluminium and, eventually, loss of this element by back diffusion into the alloy or by spallation of the AI,O, scale results in the aluminium concentrations decreasing below the critical value; this enables basis metal sulphides to become established and the rate of metal loss to increase considerably [4,5].However, for components that are made entirely from an aluminium-containing alloy, such as an aluminide, loss of aluminium by back diffusion should not be a factor. Nonetheless, it has been widely reported that, for Al,O,-forming iron-chromium-aluminium alloys containing typically 5 wt% Al, periods of protection by an A1,0, scale can be followed by breakdown and the development of iron-chromium-rich sulphides [3]. It is also known [6,7] that the composition of the steady-state scale, Al,O,-rich or basis metal sulphide, depends critically on the composition of the gas as well as the concentration of aluminium in the alloy. Preoxidation in a sulphur-free gas prior to exposure in a sulphur-containing gas can assist in giving protection for longer pcriods since such a treatment facilitates establishment of an A1,0, scale [S]. However, in environments that arc essentially sulphidizing to the basis mctals, the preformed oxide scale breaks down after somc time and less protective sulphides are established. For iron-chromium-aluminium-base alloys, the period during which the protective scale is maintained depends on the period of preoxidation and the chromium concentration in the alloy,for a given aluminium content [9].
222
EH. Stoit, K. 7:Chuah and L.B. Bradley
In a view of the potential benefits of an increased alloy aluminium concentration on the establishment and maintenance of an Al,O,-rich scale, there has been significant research into the properties of iron aluminides, including corrosion resistance in oxidizinghlphidizing gases [lo] and the effects of alloying elements on mechanical properties. It has been shown that 2 to 5 a t % Cr increases the room temperature ductility of Fe,Al-based alloys while decreasing the aluminium content below 20 at% improves significantly the ductility and fracture toughness, even in the absence of chromium D11. In the present research, investigations have becn made into the degradation resistance of a series of novel iron aluminides in gaseous envrionments of high sulphur and low oxygen activities, with emphasis on the effect of aluminium concentration in the aluminide maxtrix. In particular, previous work has shown that an alloy containing 5 at% Cr and 16 at% A1 is less resistant to H,S-containing environments than alloys containing > 18 at% Al, but superior to conventional iron-chromium-aluminium alloys [lo]. Hence, in this study, the effects of yttrium additions to iron aluminides containing 5 wt% Cr and aluminium concentrations in the region of 8 to 12 wt% (16 to 24 at Yo)on oxidatiodsulphidation resistance have becn examined and compared with that for a basic iron aluminide containing 16 wt% Al. This paper presents some data on the extents of corrosion and describes some of the scale morphologies and compositions. As the possible advantages of increased aluminium concentration for degradation resistance have t o be balanced against the possible disadvantages in terms of materials properties, particular emphasis has been placed on the effects of aluminium concentrations on the maintenance of protection.
14.2 Experimental A set of four iron aluminides (FA 49, FA 56, FA 57 and FA 58) was supplied by Shell Research, with compositions given in Table 1.These contained about 5 wt% Cr and aluminium concentrations from about 8 YOto 16 wt%. They all included a small amount of zirconium while three of them also contained yttrium, to assist in maintenance and retention of the protective A1,0, scale. For comparison, two other Al,O,-forming alloys and two Cr,O,-forming alloys were included in the programme. The former alloys contained much more chromium than the aluminides but much less aluminium; MA Table 1. Compositions of the alloys, wt%
Alloy Designation Fe
Cr
Al
FA 49 FA 56 FA 57 FA 58 MA 956 Fecralloy 310SS Alloy 800
5.5 5.0 5.0 5.1 20.0 22.0 25.2 20.0
15.8 8.1 9.8 11.9 4.5 4.8
Bal. Bal. Bal. Bal. Bal. Bal. Bal. Bal.
-
0.2
.-.
C
Y
Zr
Ti
Si
Ni
-
-
0.20 0.20 0.19 0.17
-
-
-
-
-
-
-
-
-
-
-
-
0.5
-
-
<0.01 0.008 0.017 0.06
0.08
0.22 0.31 0.32 0.50 0.30
-
-
-
-
-
0.2
0.4 0.8
-
19.3 2.0
14 Oxidation-Sulphidation of Iron Aluminides
223
956 is an yttrium-dispersed mechanically alloyed material while Fecralloy contains yttrium as an alloying addition. The aluminide specimens were discs, 10 mm diametcr and 3mni thickness, while the others were rectangles, (10 X 5 X 2)mm. Each was ground to 1200 grit finish on silicon carbide paper, degreased, measured accurately and weighed prior to exposure in the furnace. Tests were carried out in a horizontal system. Three mixed gas environments were used in the programme: Environment 1: H,-0.2 Y0H~s-4.2 %H,O at 700°C Environment 2: H,-1.5 %H,S-4.2 %H,O at 700°C Environment 3: H,-1.5 y0H~S-4.2 %H,O at 500°C Calculations of the equilibria at the reaction temperatures give the following: Environment 1: Po, = 2.8 X atm;Ps, = 1.2 X atm Environment 2: P02 = 2.8 X atm; P,, = 6.8 X lC9atm Environment 3: Po, = 3.9 X lo-?' atm; Ps2 = 2.2 X lC1'atm The mixed gases were obtained by passing H, at a known flow rate through distilled water held in a water bath set at the relevant temperature, determined by calibration experiments, and mixing in with H,/2 %H,S at a flow rate that would give the required mixture of H,/H,O/H,S. Heating tapes between the water bath and the reaction tube prevented condensation of water from the gas. A refractory holder containing four specimens was located in a double-walled silica reaction tube, with a 6 mm cavity, containing alumina granules, between the two walls (Fig. 1).During an experiment, the mixed gas entered the reaction tube and flowed through the granules in the hot zone of the furnace, set at either 500 (23)"C or 700 (t3)"C. This enabled the gas to approach equilibrium prior to passing into the inner tube and over the specimens. After sealing the reaction tube, the heating tapes and water bath were switched on and the system was purged with argon (five gas changes) before the relevant mixed gas was introduced. After a further five gas changes, the furnace was switched on and the exposure period was commenced once the relevant temperature had been attain-
Specimen location
.Porous mesh
granules
Gas inlet
Fig. 1. Schematic diagram of the mixed-gas reaction tube
EII. Stott, K . 7:Chuah arid L.B. Bradley
224
ed (about 1h). Some experiments involved a preoxidation period prior to exposure to the mixed gas. Here. the furnace was brought to temperature under a hydrogen atmosphere; thereafter, the specimens were exposed for 16 h to H2/ 4.2'26H20 before the flow rates were adjusted to enable the relevant mixed gas to pass through the system. After the required exposure period, the furnace was switched off and, once the temperature had cooled to 300"C, the mixed gas was replaced by an argon flow. The cooling rate was such that it took about 2 h to cool from 700" to 400°C. Following the tests, the specimens were weighed and examined/analyzed by various electronoptical and surface analysis techniques.
14.3 Results Following exposures for 50 h, most specimens showed only small weight changes, as indicated inTable 2. Each datum point is the average value for at least three spccimens. In general, the reproducibility was good, with the recorded values for a given alloy under a given set of conditions being within 25 to 30% of the quoted mean for small weight gains and 10 to 20 YOfor larger weight gains. Although there were small differences in the mean values for the four aluminides in the three environments, overall, the weight gains were always consistent with good protection. This was also the case for Fecralloy and MA 956 in most environments, although the weight gains in environment 2 for the non-preoxidized specimens were large, particularly for the latter alloy. The Cr,O,-forming alloys, 310 stainless steel and Alloy 800H, gave relatively small weight increases in environment 1, but very large increases in the more sulphur-rich environment 2, even following preoxidation. Examination of the specimens after exposure confirmed that thin protective scales had developed and were retained on the iron aluminides and on the other alloys, except where large weight gains were recorded, in all three environments. The scales on the iron aluminides at 700°C were essentially A1,0,, as confirmed by X-ray diffraction Table 2. Weight gain data after 50 h exposure in the various envi-
ronments
FA 49 FA 56 FA 57 FA 58 Fecralloy MA956 310SS Alloy800
-
_
~-~ Weight Gain, rng c m 2 Env 1 Env 2 Env 3 N P N P N -___-
_
0.03 0.06 0.06 0.04 0.06 0.07 0.09 0.49
~
N - non-preoxidized ND -not determined
0.02 0.04 0.04 0.04 0.03 0.01 0.07 0.28
---
0.03 0.11 0.08 0.04 1.40 15.50 27.5 54.0 P
0.03 0.04 0.04 0.04 0.03 0.02 6.91 5.10
-__
- Preoxidized
0.01 0.27 0.06 0.02 0.05 0.53 ND ND
14 0.ridntion-Sulphidalion of Iron Aluminides
225
and by in-depth Auger electron spectroscopcial analysis following ion beam sputtering, for both preoxidized and non-preoxidized specimens, in region away from the nodules described later (Figures 2(a) and (b)). Thus, after 50h exposure in environment 2, the scale on FA 49 was about 200nm thick, with no iron or chromium being detected in it. The preoxidation treatment also produced a thin alumina scale which had thickened to about 300 nm after a further 50 h exposure in environment 1.Again, only aluminium and oxygen were detected in thc scale. Several interesting features were apparent when the aluminide specimens were observed in plan after exposure at 700°C (Figure 3). In all cases, there was little evidence for loss of scale by spallation. However, second-phase particles were observed on all the surfaces, particularly of the lower aluminium-containing alloys, FA 56 and FA 57, while the number and size of such particles for a given alloy were less for preoxidized specimens than for those exposed directly in the mixed gas. Detailed examination and EDX analysis of the particles (e. g. labelled, C, A and B respectively in Figures 3(c), 3(d) and 3(f)) indicated them to be rich in iron and sulphur; although too small for precise analysis, they were undoubtedly sulphide nodules. These nodules were always small and discrete, with no evidence that extensive sulphidation of the substrate had
Depth, nm
v)
c
C
0
0
0
I
100
I
200
300
400
Depth, nm
500
600
Fig. 2. Auger electron spectroscopy depth profiles for FA 49; a) exposed for 50 h to environment 2 at 700°C,b) preoxidized for 16 h and exposed for 50 h to environment 1 at 7ob"C
226
EH. Stott. K. 7: Chunh arid L.B. Bradley
Fig. 3. Scanning electron micrographs of surfaces of specimens after exposure for 50 ti at 700°C; a) FA 49,environment 1,b) FA 58,environment 1.c) FA 56, environment 1. d) FA 4Y. preoxidized, environment 2, e) FA 56, environment 2, f) FA 56, preoxidized, environment 2
ensued. Although difficult to confirm, the nodules were apparently separated from the substrate by an A1,0, scale. In-depth Auger analyses through regions of nodules showed iron and sulphur enrichments at the surface, with aluminium and oxygen enrichments below the nodules. However, the sputtering technique was not sensitive enough to allow sulphur levels in the underlying oxide scale to be determined. Occasionally there were oxide nodules on the specimen surfaces (marked C in Figure 3(b)): these were rich in zirconium. Also, a few dark patches were observed on the surface of the yttrium-containing alloys (marked A in Figure 3(b)); these were rich in yttrium, iron and aluminium, consistent with second-phase yttrium-rich phase from the substrate. In both cases, alumina scales had become, at least partially, established over these second-phase regions. Although the scales were too thin to be observed usefully in cross section, such examinations revealed extensive grain-boundary precipitation in the substrate, often penetrating through much of the specimen thickness, for FA 57 and FA 58 (Figures 4(a) and (b)); no such precipitation was observed for FA 49 and FA 56. EDX analysis
14 Oxidation-Siilpliidation of Iron Aluminides
227
Fig. 4. Scanning electron micrographs of specimens in section after exposure for SO h at 700°C; a) FA 57. environment 1. b) FA 58. preoxidized. environment 2
of the precipitates indicated the presence of zirconium and yttrium, although they were too small for accurate measurements. After exposurc to environment 3 at 500"C, FA 49 and FA 58 developed vcry thin scales with no obvious signs of sulphide nodules on their surfaces. However, significant numbers of sulphide nodules, up to 5 pm thick, were observed on FA 56, while a few small sulphide nodulcs were detected on FA 57. Exposure of Fecralloy and MA 956 directly to environment 1 at 700°C also resulted in the development of thin scales. However, these were not entirely alumina. Figure 5(a) shows a typical patch of dark oxide scale, rich in aluminium, on Fecralloy, surrounded by a more granular oxide scale, rich in chromium. The nodule (B) was rich in iron and sulphur only.
Fig. 5. Scanning electron micrographs of specimens after exposure for 50 h at 700°C; a) Fecral loy, environment 1,in plan, b) MA 956, environment 1,in plan, c) Fecralloy, environment 2, in section, d) MA 956, environment 2, in section
228
F H . Sio~, K. 7:Chuah arid L.B. Bradley
Similarly. MA 956 developed two types of oxide scale (Figure 5(b)), a dark aluminium-rich surface and a lighter, more granular, chromium-rich surface. These maintained protection and only a few discrete sulphide nodules were detected. In environment 2, this was no longer the case. Thick sulphide scales developed on MA 956; these consisted essentially of an outer iron-rich sulphide and an inner iron-chromium-aluminium-rich sulphide (Figure 5(c)). Likewise, although not as thick, substantial sulphide scales had formed on Fecralloy after 50 h (Figure 5(d)), with significant internal sulphidation of the underlying alloy. Preoxidation was an effective treatment for these alloys in the tests since both maintained the protective oxide scale, even after Soh in environment 2. Nonetheless, as observed in Figure 6(a), relatively large sulphide nodules had developed by the end of the exposure period, indicating an approach to breakdown of the protective scale. After exposure for 50 h to environment 3 at 500"C, MA 956 had developed a relatively uniform sulphide scale, approximately 5 pm thick, with no evidence for any retained oxide scale; however, Fecralloy showed better resistance and the essentially oxide scale was very thin. As expected, the Cr,O,-forming alloys were not very resistant to degradation, particularly in environment 2.The scale formed on 310 SS in environment 1 was essentially Cr,O, and spalled significantly on cooling (Figure 6(b)). A similar oxide scale was formed on Alloy 800, although it was thicker, more nodular and much more adherent than that on 310 SS. Extensive sulphidation of these two alloys was observed in environment 2, even after preoxidation. The large nodules shown in Figure 5(c) and the thick scales in Figure 6(d) are typical sulphides, rich in iron in the outer layer and ironchromium in the inner layer.
Fig. 6. Scanning electron micrographs of specimens after exposure for 50 h at 700°C; a) Fecralloy, preoxidized, environment 2, in plan, b) 310SS, environment 1, in plan, c) 310SS,preoxidized, cnvironment 2, in plan, d) Alloy 800,preoxidized, environment 2, in section
14 O.~idation-SuIphidnriono,f Iron Aluminides
229
14.4 Discussion The four iron aluminides developed and maintained protective oxide scales in the mixed-gas environments at 700 "C, although discrete sulphide nodules were formed on the surfaces. There was some correlation betwecn the aluminium content of the alloy and the weight gains (and number and size of nodules on the surface) in environments 1 and 2, with the weight gains increasing with decreasing aluminium content, from FA 49. FA 58, FA 57 to FA 56. This is consistent with the results of DeVnri and Tortorelli [lo] who reported that an alloy containing 16 atyo Al and 5 YOCr was less resistant than alloys containing >18 at% Al to H,S-H,-H,O mixed gases, but more resistant than conventional iron-chromium-aluminium alloys. Others have found that the corrosion rates of iron-chromium-aluminium alloys in sulphur-rich gases decrease with increasing alloy aluminium concentration [ 121. However, is has also been reported that additions of only 2 % Cr to iron aluminides containing 26 to 28 at% Al can be detrimental to sulphidation resistance [13], although tests in that research involved higher temperatures and gases of higher H,S concentrations than the present work. One aim of this research has been to investigate the effects of yttrium on the effectiveness of the A1,0, scales in giving protection, based on the well-known beneficial effccts of this element in improving the mechanical integrity and spallation resistance of such scales during high-temperature oxidation. However, in thc present tests, thc thickness of the A1,0, scales were only of the order of 200 to 300 nm and scale spallation was not a factor for any of the iron aluminides. Nonethelcss, the two alloys containing the most yttrium, FA 57 and FA 58, developed extensive intergranular precipitates in the substrate. These precipitates were rich in both zirconium and yttrium. The causes of such prccipitation are unclear at present, but its occurrence is much more damaging than the small extents of sulphide formation in the tests. Although M A 956 and Fecralloy are essentially Al,O,-forming alloys at high temperatures, it is interesting that the scales developed during preoxidation and during exposure in environment 1 at 700°C were not entirely AI,O,; some areas were covered by a more granular Cr,O, scale. Presumably this is because the rates of growth of the initial AI,O, nodules are very low at the relatively low temperature and low oxygen potentials of the tests while the rate of aluminium diffusion to the surface is also low at this temperature, enabling a Cr,O,-rich scale rather than an Al,O,-rich scale to become established in local regions on the surface. O n exposurc to environment 2, it is also difficult for an Al,O,-rich scale to develop and faster growing sulphides are able to become established rapidly, particularly on M A 956. The relatively poor performance of 310 SS and Alloy 800 in the mixed gases is consistent with the widely-reported susceptibility of Cr,O,-forming alloy t o sulphidation in such environments. The importance of a high temperaturc for establishment of a protective oxide scale is also illustrated by the results in environment 3 at 500°C. Mere, an effective A1,0, scale was not able to be established fully on either FA 56 (containing 8.1 YOA1 and 5.0 YoCr) or M A 956 (containing 4.5 YOAl and 20.0 YOCr) and the weight gains were relatively large, consistent with formation of significant amounts of sulphide. More effective scales were established on the other iron aluminides, containing greater concentrations of aluminium, and the weight gains were relatively low.
230
E H. Stott, K . 7:Chuah and L.B. Bradley
Generally, this research has shown that iron aluminides can have good resistance to mixed gases of high sulphur and low oxygen potential. However, the results also indicate that, for alloys containing 5 YOCr, aluminium concentrations above 10 wt% give better protection than those between 8 and 10 wt%.As alloys containing less than 8 or 9 wt% A1 have much higher room temperature ductilities than those containing over 10 wt% Al,it is apparent that further development of such alloys could involve a trade off between mechanical properties and sulphidation resistance, with emphasis on the effects of other minor alloying elements, such as the yttrium used in the work, on overall performance.
14.5 Conclusions 1. Iron aluminides containing 5 wt% Cr, 0.2 wt% Zr and 8 to 16 wt% A1 show better resistance to a mixed-gas, HJ1.5 %H,S/4.2 %H,O environment at 700°C than conventional A1,0,- and Cr,O,-forming alloys, with the extent of sulphide formation for the aluminides increasing with decreasing aluminium concentration in the substrate. 2. Under these conditions, the presence of 0.2 to 0.3 wt% Y in the iron aluminides has little influence on the effectiveness of the thin A1,0, scale in giving protection; however, extensive intergranular precipitates, rich in yttrium and zirconium, are formed in alloys containing 0.3 wt% Y. 3. It is more difficult to establish a protective A1,0, scale at lower temperatures than at higher temperatures, particularly on alloys of relatively low aluminium concentrations; thus, an iron aluminide containing 8 wt% A1 formed more extensive sulphides at 500°C than at 700"C, although a protective oxide scale prevented significant sulphidation of an alloy containing 12 wt% Al at either temperature.
14.6 Acknowledgements The experimental work was undertaken in MSc dissertation programmes (by KTC and LMB), sponsored by Shell Research, Arnhem. The authors are also grateful to Shell Research, particularly ?: Wolfert,for providing the iron aluminides and for useful discussions.
14.7 References [l] C. G. McKarney, J. H. DeVan, I! E Tortatelli, V K . Sikka: J. Mater. Res.6'(1991) 1779. [2] C. M . Packer, R. A. Perkins: Roc. Conf. on High-Temperature Alloys in Aggressive Environments, 813,The Metals Society, London (1980). [3] E H. Stott, E M. E Chong, C. A. Stirling: Materials Science Forum 43 (1989) 327. [4] S. W Green, E H. Stotf:Corros Sci. 33 (1992) 345.
14 Oxidation-Sulphidationof Iron Aluminides
[S] [6] [7] [8] [9) [lo] [Ill 1121 [13]
S. W.Green, E H. Stott: Oxidation Metals 36 (1991) 239.
231
M. H. La Branche,A. Garratt-Reed, G.J. Yurek:J. Electrochem. SOC.130 (1983) 2405. C. R. Wang, %. B. Zhao, S. K . Xia, W Q. Zhang: Oxidation Metals 32 (1989) 24. E H. Stott, E M . E Chong, C. A . Stirling:High Temperature Corrosion in Energy Systems, ed. M. Rothman, 253, Met. SOC.AIME, New York (1985). P A. Mari, J. M. Chaix,J. I? Larpin: Oxidation Metals 17(1982) 315. J. M.DeVan, P E Tortorelli:Corros. Sci. 35 (1993) 1065. b! K. Sikka: Proc. 6th Annual Fossil Fuel Energy Conf. ORNI.IFMP-92Il. 195, Oak Ridge Nat Lab,Oak Ridge.TN (1992). S. Mrowec, M . Wedrychowska: Oxidation Metals 13 (1979) 481. J. If. DeVan: Oxidation of High-Temperature Intermetallics, cd. T. Grobstein and J. Doychack, 107.Minerals, Metals and Materials SOC.(1989).
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
15 Metal Dusting of Fe,AI and (Fe, Ni),AI S. Straufl, R. Krajak, M. Palm and H.J. Grabke
15.1 Introduction
Sevcral new materials have been dcveloped on the basis of the aluminides Fe,AI, Ni,AI (Fe, Ni),AI, alloyed with Cr and doped with small additions of Zr or Hf [14]. These materials proved to be rather resistant against high temperature corrosion in oxidizing, sulfidizing and carburizing environmcnts. The carburization resistance was tested in CH,-H, atmospheres at high temperatures and carbon activities a, < 1 [4]. However, at intermediate temperatures 400-900°C and carbon activities a, > 1 there is another dangerous corrosion phenomenon ‘metal dusting’ which leads to disintegration of Fe-, Ni- or Co-base materials to a dust of fine metal particles and carbon (graphite) [5-141. Low alloy steels show general metal wastage and high-alloy materials show pitting and holc formation. It was hoped that the M,Al based materials would be resistant against metal dusting and might be used as coatings or construction materials under ‘metal dusting’ conditions. Thus, their metal dusting resistance was testcd under standard conditions.
15.2 Experimental The M,Al based materials were exposed in flowing CO-H,-H,O mixtures at 650°C for different periods of time. Corresponding atmospheres are used in many processes attained from methane conversion for use in synthesis of ammonia, hydrocarbons, alcohols etc. or in the direct reduction of iron ores. The materials tested were (concentration in wt%): 1. 2. 3. 4. 5.
Fe,AI Fe,Al-2.2% Cr Fe-16 %AI-4.8%Cr-0.15 %Zr Ni,Al Ni-26 %Fe-10 %A1-8 %Cr-0.1 %Hf
The mass of thc samples was determined before and after the exposure, and after removal of the corrosion product, the ‘coke’ which had grown on the specimens, locally from pits or on the whole surface.
S. StrutiJ3,R. Krujuk, M . Pulni and H.J. Grabke
234
15.3 Results All M,AI aluminides containing Fe were attacked by metal dusting, most severe attack was observed on Fe,AI (Fig. 1).Vast protrusions of coke were growing from pits on Fe,AI, from the material (2) with 2.2% Cr only a few protrusions grew from the unpolished edges, and on the material ( 3 ) only a thin coke layer was observed without pitting. In contrast, the pure Ni,AI showed only a thin carbon layer on the unpolished edges. Some data on the mass changes by coke growth (removable), metal wastage resp. oxidation and carburization (remaining weight change) are compiled in Fig. 2.
Fig. 1. Metal dusting attack on Fe,AI, vast outgrowth of coke after 7 d exposure at 650°C and a, = 10,the arrow indicates the sample from which the coke has grown.
1000
time: Id, 2d
Mo
g
10
2 .-C
%
1 0.1 0.01
Y) VI
E 0.001 -0.01 -0.1
-1
Fig. 2. Mass changes of four alloys in exposures at 600°C and a, = 41.5for one day and a second day. black columns: removable coke and hatched columns: remaining mass change by metal wastage (loss), carburization and oxidation (gain).
I5 Meral Ditsrirrg of Fe.,Al and (Fe. Ni).,Al
235
The metal dusting of Fe and low alloy steel proceeds via intermediate formation of the instable carbide Fe,C [6,7,8] which then decomposes according to Fe,C -+ 3 Fe + C. For Fe,Al an intermediate formation of the K-phase Fe,AIC [lS] might occur, which also would be instable if formed at a, > 1.Therefore. specimens of Fe,AI were investigated after metal dusting for 48h at 650°C: and a, = 1.5, using metallography, scanning electron microscopy with EDAX and X-ray structure analysis to search for the K-phase. However, no indication of its presence was found but clear evidence of the presence of Fe,C. This instable carbide was observed as a thin layer on thc Fe,AI, obviously growing from the inside and decomposing to the outside, as in the case of metal dusting of iron and steels [6,7,8].The A1 obviously is oxidized under formation of A1,0, but this oxidation does not lead to a protective scale. Protection appears to be improved by alloying with Cr, whereby the metal dusting attack is decreased. Ni,AI is not attacked by metal dusting, but materials containing Fe show pitting after prolonged exposures (see Fig. 3).
Fig. 3. Pitting by metal dusting of a Ni-base aluniinide, after 330 h at 650°C. and a, = 20.
236
S.StrauJ3, R. Krajuk, M. Palm and H.J. Grabke
15.4 Conclusions Fe,AI based materials are susceptible to metal dusting. In CO-H,-H,O atmospheres such materials disintegrated to a ‘dust’ of iron, carbon and alumina. By alloying with Cr the attack is decreased somewhat.The disintegration proceeds via intermediate formation of the instablc carbide Fe,C, as in the case of iron and low alloy steels Ni,AI is not susceptible to metal dusting, however, aluminides (Ni,Fe),AI containing Fe are subject to metal dusting. After long-tcrm exposures such materials show pitting such as high alloy steels [7].
15.5 References [l] J. H. DeVan, C. A. Hippsley: in: Oxidation o f High Temperature Internietallics. Ed.T. Grohstein. J. Doychak,The Minerals, Metals and Matcrials SOC.1989,p. 31 and p. 107. [2] J. H. DeVan: in: Heat-Resistant Materials, Proceedings of the First Int. Conference, Fontana. Wisconsin, Sept. 1991. Eds. K. Natesan, D.J.Tillack ASM 1991, p. 235. [3] J. H. DeVan, P E Tortorelli: Materials at HighTemperature 11 (1993) 30. [4] J. KMwer:Workshop ‘Oxidation of Intermctallics’, in press [5] J. C. Nava Paz, H. J. Grabke: Oxid. Metals 39 (1993) 437. 161 H . J. Grabke, R. Krajak, J. C. Nava Paz: Corrosion Sci.35 (1993) 1141. [7] H. J. Grabke, R. Krajak, E. M. Miiller-Lorenz:Werkst. Korros. 44 (1993) 89. [ 8 ] H. J. GraDke, C. B. Bracho-Troconis, E. M. Miiller-Lorenz:Werkst. Korros. 45 (1994) 215. [9] H. J. Grabke, R. Krajak: Harterei-Techn. Mitteilungen 49 (1994) 150. [lo] H. J. Grabke, E. M.Miiller-Lorenz, 1;. Pippel, S. StrauJ3: Conf. Papers U K Corrosion & EUROCORR ‘94,Vol.3, Chameleon Prcss Ltd. London 1994,p. 361. [ l l ] H. J. Grabke: Corrosion NACE 51 (1995) 711 [12] E. Pippel, H.J. Grabke, S. StrauJI, J. Woltersdorf: steel res. 66 (1995) 217 [13] H. J. Grabke: Solid State Phenomena 41 (1995) 3. [14] H. J. Grabke, E. M. Muller-Lorenz: steel res. 66 (1995) 254. I151 M. Palm, G. Inden: Intermetallics3 (1995) 443.
Part IV Ti-Aluminides
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
16 Determination of Thermodynamic Activities in the Alloys of the Ti-A1 System and Prediction of the Oxidation Behaviour of the Alloys M. Eckert and K. Hilpert
Titanium aluminide alloys based on Ti,Al and TiAl are of interest as construction material for high temperature components particularly in aerospace industry. Good mechanical properties can be attained with alloys consisting of y-TiAl with 3 to 15 vol% a2-’fi3A1.The disadvantages are the low ductility and the inadequate oxidation resistance at service temperatures of 700-900°C [l].A fundamental understanding of the oxidation behaviour is necessary in order to improve the corrosion resistance. The formation of the oxides on the alloy surface depends on the temperature, the oxygen partial pressure of the corrosive atmosphere, and the thermodynamic activities of Ti and A1 in the alloys. The measurement of the thermodynamic activities of Ti and Al in the alloys of the Ti-AI system is the aim of this study in order to render possible an improved understanding of the formation of the various oxides. The measurements of the thermodynamic activities of Ti and Al in Ti/AI alloys carried out so far are insufficient.They were essentially conducted by Hoch and Usell [2] using Knudsen effusion mass spectrometry and by Samokhval et al. [3] using e. m. f. measurements. Hoch and Usell have measured the thermodynamic activities of Ti and Al at 1780K in the @-Tiphase for the composition range 0 < xA, 5 0.4. Samokhval et al. measured the A1 activities at 960 K in the composition range 0.05 5 xA, 5 0.5. Recently,Jacobsen et al. [4] reported A1 pressures for three samples of the phase compositions %,A1 + TiAI, Ti-0.52A1, and TiAl + TiAI, obtained by Knudsen effusion mass spectrometry. The thermodynamic studies by Hoch and Usell as well as Samokhval in addition to many other investigations [5-131 are the basis for the recent calculation of the phase diagram of the Ti-A1 system by Katfneret al. [14]. The present study was carried out to determine for the first time the thermodynamic activities of Ti and Al over the complete composition range 0 < xA, < 1 in the temperature range from 1200 to 1600K. A detailed presentation of the results obtained will be reported in Ref. [15].The study reported here is part of our systematic investigations on the influence of ternary addiditives to the Ti/AI alloys. Vaporization studies by Knudsen effusion mass spectrometry are used for the thermodynamic investigations. The results obtained by this method for many alloy systems are summarized in a review article by Hilpert [ 161. ‘The measurements reported in this work complement
M . Eckert and ti.Hilpert
240
our investigations on the alloys of the Ni-A1 system [ 17,181 including the intermetallic phase Ni,AI [17,19]. Nineteen Ti/AI alloy samples with different compositions covering the complete composition range of the Ti-AI system were prepared by inductive melting in a purified argon atmosphere using the levitation technique (Fig. 1). The oxygen content of the alloys was additionally determined after the vaporization measurements, because of the known high oxygen solubility of Ti. The 0, concentrations in ppm given in parentheses were obtained in the alloys with Al concentrations of 0.05 (759 i 98). 0.10 (1119 -C 54), 0.15 (2766 2 31), 0.20 (2340 -C 162), 0.25 (708 % 59), 0.30 (861 75). 0.35 (17610 2 1550),0.40 (571 % 11),0.45 (957 % 168),0.47 (461 % 48),0.50 (673 ? 28),0.55 (165 ? 2), 0.65 (65 2 15), 0.69 (32 2 15), 0.725 (31 -C l l ) , 0.80 (39670 -? 26680), and 0.90 (329 % 183). The measurements were carried out with the substantially modified Knudsen cell mass spectrometer system described in Ref. [20].The Knudsen cell mashined from Mo was completely lined with yttria.The geometry of the Knudson cell with lining was the same as described and depicted in Ref. [17]. Ti and A1 partial pressures were determined in the temperature range of the mcasurements (cf. Fig. 1) over the different samples. Thermodynamic activities of Ti and Al resulted according to the relation a, = p, / pv (i = Ti,AI) by comparing the pressures over the alloy samples, p,, with those of pure Ti(s) and AI(I), py.The results obtained at the mean temperature of 1473K are given in Fig. 2. The thermodynamic activity of Ti was, additionally, evaluated from the measurement results by the ion intensity ratio integration method. The results obtained by this method are also shown in Fig. 2. The two different evaluation methods agree in general very well. Knowledge of the thermodynamic activities of Ti and A1 in the alloys is essential for predictions on the stability of oxide scales. Such predictions were carried out so far by -f
T
-
I
-
I
-
t 2000
1800 1600 Y
;1400 1200 1000 800 0 Ti
02
0.4
06 xAl
08
1 Al
Fig. 1. Phase diagram by Kattner et al. [14] with the composition of the prepared samples and the temperature ranges of the vaporization measurements
16 Determinniiori of ThermodyaniicAciiviiies in the Alloys of the Ti-A1System
241
1
0.9 0.8
0.7
c ‘
< .,”
G
0.6
05 0.4
0.3 0.2 0.1 U I
XAl
L
*
Fig.2.Thermodynamic activities of X and A1 at 1473 K in solidTi/Al alloys of differcnt compositions obtained by the ion intensity ratio method ( 0 )and by the equation a, = pi / pi0 (+,0)
Rahmel and Spencer [21] as well as Lurhra [22]. Only estimated thermodynamic activities were, however, used by these authors. The computations by Rahmel and Spencer as well as Lurhru are repeated in the following by employing the thermodynamic activities determined in this work. The oxygen partial pressures of the reactions 3 2A1 + 2 O,(g) w A1,0, (s) and (1) 1 Ti + 2O,(g) w TiO(s) (2)
were computed.The computation was carried out at 1373K in order to render possible a comparison with the results obtained by Rahmel and Spencer [21]. The result of this computation is shown in Fig. 3. It is obvious, that the oxygen partial pressures belonging to Eqs. (1) and (2) intersect at an A1 concentration of about xAl= 0.54. Rahmel and Spencer [21] determined the Al concentration for this intersection as xAl = 0.61, while Lirthra [22] obtained xA, = 0.55 for this intersection at 1073 K. This means, that the results of this work lead to the prediction of a stable A1,0, scale for lower A1 concentrations than by Rahmel and Spencer [21]. Moreover, the data obtained by us show that this A1 concentration decreases with decreasing temperature. Dissolution of oxygen in the alloy was neglected in our computations and in the predictions by Rahmel and Spencer. Work is, therefore, in progress to elucidate the influence of oxygen dissolution o n the thermodynamic activities ofTi and Al.
242
M. Eckerf and K. Hilpert
Fig.3. Variation of the metal/oxide equilibrium pressures in the Ti-A1-0 system at 1373 K
Acknowledgements The authors thank the Deutsche Forschungsgemeinschaft for financial support within the program ,,High Temperature Corrosion“.
16.1 References [I] A. Rahmel, WJ. Quadakkers, M. Schiitze: Materials and Corrosion 46 (1995) 271. 121 M. Hoch, R.J. Usell:Met.Tfans. 2 (1971) 2627. [3] K V Samokhval, RA. Poleshchuk,A.A. Vecher: Russ. J. of Phys. Chem. 45 (1971) 1174. [4] N S . Jucobsen, M.I? Bra$y, G.M. Mekrota: 188th Meeting of The Electrochemical Society. Chicago, Oct. 1995. Symp. on “High Temp. Materials/Corrosion” - High Temp. Mat. Chemistry 7. exten&&bstract. (51 K. Shibata, T Sato, d Dhira: J. Cryst. Growth 27 (1978) 1329. [6] E. W Collings: MetaKTrans. IOA (1979) 463. [7] A. Abdel-Humid, C,H.Alliberf, F: Durand Z. Metallkd. 75 (1984) 455. IS] R.D. Shul1,A.J. Me-Alisfer,R.C. Reno:Titanium Sci.Tech.3 (1985) 1459. [9] R.M. Waterstrat: NISlIR 88-3856, United States Department of Commerce, (1988). [lo] K. Kalfenbach, S. Gama, D.G. Pinatti, K . Schu1ze:Z. Metallkd.BO(1989) 511. [ l l ] C. McCullorcgh,J.J. Vulenciu, C.G. Levi, R. Mehrabian: Acta Metall. 37 (1989) 1321. [12] J.C. Schuster, H. Ipser: Z. Metallkd.81 (1990) 389. [13] J.C. Mishurda, Perepezko: in: “High-TemperatureAluminides and Intermetallics”, edited by S.H. Whang et al. (TMS, Warrendale, Pennsylvania, 1990). [14] U R .Katfner, J.C. Lin, Y A . Chang: Met.Trans.23A (1992) 2081.
16 Delerniinntion of Tlzerniori),namicActivilies in the Alloys of the Ti-A1 System
243
115) M. Eckerr. L. Bericze, D. Knth. H. Nickel, K . Hilperl: Ber. Ihnsenges. Phys. Chem., in press. [ 161 K . Ifilpert:Structure and Bonding 73 (1990);Springer.Heidelberg. [17] K. Hilput, D. Kobertz, V: Veriugopnl, M. Miller, H. Gerads9 EJ. Bremer, H. Nickel: Z. Naturforsch.42a (1987) 1327. [18] K . Hilpert, M. Miller, H. Gerads, H. Nickel Ber. Bunsenges. Phys. Chem. 94 (1990) 40. [19] M. Albers, M. Sai Baba, D. Kath, M. Miller, K . Hilpert: Ber. Bunsenges. Phys. Chem. 96 (1992) 1663. [20] K. Hilpert: Habilitationsschrift,Technische Hochschule Darmstadt 1981; Report from the KFA Jiilich. Jul-1744,p. 28. [21] A. Ruhnrel, PJ. Spencer: Oxid. Met.35 (1991) 53. [22] K.Z.. Lirflira:Oxid. Met.36 (1991) 475.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
17 The Initial Stages in the Oxidation of TiAl C. Lang and M. Schiitze
17.1 Introduction The high temperature oxidation behaviour of titanium aluminides has been investigated by several authors in recent years [l-51. It is well known from these studies that in most cases the titanium aluminides based on y-TiAl do not form an protective A1,0, oxide layer and therefore exhibit an insufficient oxidation behaviour at temperatures above 800°C. Instead of a dense Al,O, layer the oxide scale usually exists of a layered structure TiO2//AI20,//TiO2 + A120, mixture//Al depletion zone from the outside to the inside.The A1,03 in the second layer from the outside provides protection as a barrier layer during a certain time of the oxidation process, but is not stable for longer oxidation times [4]. Often a better oxidation resistance of y-TiAI based titanium aluminides is observed if the oxidation tests are carried out in pure oxygen instead of air, since the absence of nitrogen in the oxidizing environment has a beneficial effect on oxidation kinetics in most cases [S-71. Also thc addition of ternary elements can improve t h e oxidation behaviour of TiAl [8]. One of the most effective elements to decrease the oxidation rate is niobium, as several investigations have shown, e. g. [l, 9-11]. In spite of this extensive number of results on TiAl oxidation the mechanisms are not yet understood in detail. In particular a lack of information exists on the processes taking place in the early stages of oxidation. Because it is to be expected that transient oxidation at the beginning influences the oxidation behaviour in the later stages it is the aim of this paper to elucidate the mechanisms of the initial stages of TiAl oxidation with respect to long-term behaviour.The influence of nitrogen and the effect of niobium are considered in particular.
17.2 Experimental Procedures The experiments were carried out with 'K36A1 (mass-%) and Ti35A15Nb (mass-%). The materials were produced by vacuum melting and casting, leading to a microstructure consisting of a y-TiAl matrix with embedded two phase colonies of lamellar.a,Ti,A1 and y-TiAl.The overall content of a2-Ti3AIwas very small. In Table 1 the chemical composition of the investigated materials is given.
246
C. Lung and M.Schiirze
Table 1. Chemical composition of thc materials investigated
(in mass-%)
Material
Ti
Ti36A1 Ti3SA15Nb
62.30 37.25 59.25 35.65 4.32
-.
A1
Nb ~
C
H
N
0
0.009 0.003 0.011 0.31 0.039 0.004 0.008 0.34
Specimens with the dimensions of 10 X 1.5 X 1mm were cut from cast rods and wcre ground with S i c paper to a 4000 grit surface finish. Beforc the oxidation tests the samples were cleaned ultrasonically in ethanol. Two kinds of oxidation experiments were carried out: isothermal exposure tests at tcmperatures of 800 to 1000°C:for 0.5 h up to 4 h and continuous mass change measurements at 900°C, both in laboratory air. To ensure rapid heating up the exposure tests were started by introducing the samples directly into the hot furnace. The same test conditions were used for the continuous thcrmogravimetric measurements by raising the hot furnacc around the specimen chamber. Because of the expected small dimensions of the oxidation products and of the oxide scale after short oxidation times it was necessary to use electron microscopic methods to characterize the samples after oxidation. Besides scanning electron microscopy (SEM) transmission electron microscopy (TEM) was mainly used to describe the morphology of the oxide scale and to identify the oxidation products by energy dispersive X-ray analysis (EDX) as well as electron diffraction. A detailed description of the preparation of TEM cross-sections and of the experimental procedure is given in [ 121.
17.3 Results 17.3.1 Thermogravimetric Measurements In Fig. 1 the weight gain curves of Ti36A1 and Ti35A15Nb obscrved in the continuous thermogravimetric measurements at 900°C in air are presented. As it is known very well in the literature (see beforc) the oxidation resistance of niobium containing titanium aluminides is much better compared to that of niobium free TiA1. This is also found in the early stages of o3idation. After 25 h of oxidation the relative mass gain of Ti36A1 is distinctly higher thah that of Ti35AlSNb. The oxidation behaviour of both materials is characterized by a strong increase in weight during the very beginning of oxidation.This stage with a very high oxidation rate lasts up to about 1 to 2 h and is referred to as stage I. After stage I the oxidation rate decreases strongly and an almost linear oxidation behaviour is observed. This second stage of oxidation ends after about 10h for Ti36Al and after about 20 h for Ti35A15Nb when the oxidation kinetics start to obey a parabolic rate law [12].The different oxidation rates in the early stages of oxidation can be clearly observed in a parabolic plot of the mass gain which is shown in Fig. 2. Due to the absence of a linear dependence of the square of mass gain on oxidation time at the beginning it is evident that TiAl oxidation does not follow parabolic kinetics in the early stages.
I7 The Initial Stages in the Oxidation of TiAl
2,0r
9OO0C, air
t
Ti36A1
r
u,u
0
247
I I 10
5
20
15
25
tinh
Fig. 1. Mass gain vs. oxidation time of Ti36AI and Ti35A15Nb at 900°C in air.
2.0
1 m
1.5-
6 . t
N
0,
1.0 -
E
.-c N
a
E
0.5
0.0
0
1
2
3
4
5
6
7
8
9
101112 131415
t in h
Fig.2. Parabolic plot of the mass gain of Ti36AI and Ti35A15Nb at 900°C in air.
17.3.2 Characterization of the Oxide Scale and the Subsurface Layer 17.3.2.1 Ti36A1, 9OO0C, 0.5 h In Fig. 3 a TEM micrograph and a schematic illustration of the complete oxide scale and of the subsurface layer of Ti36AI after oxidation at 900°C in air for 0.5 h is presented. The oxide scale consists of an oxide mixture of AI,O, and TiO, in the outer part of the scale. Beneath this oxide mixture a porous partial layer rich in TiO, is found. At
248
C. Lang and M . Schiitze
Fig.3.TEM micrograph and schematic illustration of the oxide scale and of the subsurface layer of Ti36A1 after oxidation at 900°C in air for 0.5 h.
the metal/oxide interface AI,O, and Ti-nitrides are formed, which penetrate into the A1 depleted metal subsurface 1ayer.This is also shown at a higher magnification in Fig. 4. Because of the small dimensions of the oxides and nitrides at the metalioxide i n k t face an unambiguous identification of the phases is not free of problems. Several electron diffraction patterns indicate, however, the existance of a metastable AI,O, or an aluminium oxynitride similar in structure to Al,,O,,N [13]. Furthermore TIN is found at the metal/oxide interface.The analysis of the metal subsurface layer by electron diffraction (Selected Area Diffraction (SAD)) and energy dispersive X-ray analysis (EDX) revealed that the Ti-rich phase of this zone, which contains approximately 62 at.-% Ti and 38 at.-% Al, is neither a,-Ti,Al nor y-XAl. Although there is no additional phase known from the binary phase diagram of the system TiAl[14] several authors reported of a new phase rich in titanium [4, 151. The composition of this new phase which was identified as a cubic structure (161 was notified to be similar in composition to Ti,Al, [17], Ti,AI [18, 191 or Ti,A1,0, [20] and the lattice parameter was given in the range of 0.66 to 0.69nm. The indices of this new cubic phase (NCP) fit very well to different SAD patterns of the same grain of the Ti-rich subsurface layer. An example of a SAD pattern of the new cubic phase is given in Fig. 4.
17.3.2.2 Ti36AI, 9OO0C,4 h Figure 5 shows the backscattered electron (BSE) image and element distribution maps of the oxide scale on Ti36A1 after a 4 h exposure at 900°C in air.'The structure of the oxide scale is already similar to that observed after long-term exposure, cf. e. g. [4]. Beneath an outer TiO, layer a partial layer enriched in A1,0, had formed which can be taken as a preceeding step to the formation of the Al,O,-barrier observed later on [4]. The inner part of the scale consists of a mixture of A1,0, and TiO,. An extensive
I7 The Initid Stciges in the Oxidation of TiAl
249
Fig.4. a)'TEM micrograph of the metal/oxide interface and metal subsurface layer of Tl36A1 after oxidation at 900°C in air for 0.5 h. b) SAD pattern of the new cubic phase.
depletion zone in the metal subsurface is not observed. This is confirmed by TEM analysis of the metal subsurface zone, the melal/oxide interface and of the inner oxide scale. As can be seen in Fig. 6 beneath a porous, Ti0,-rich partial layer Tknitrides and aluminium oxide are observed at the metauoxide interface. Only small grains of a Tirich metal phase similar in composition to Ti,AI can be detected by EDX [12].A more detailed analysis of the aluminium depleted layer by electron diffraction revealed that the new cubic phase yields a satisfactory indexing of the SAD patterns. The nitrides which form at the metal/oxide interface are found to be Ti,AIN and TiN, whereas the aluminium oxide at this part of the oxide scale (contrary to the outer part of the scale) is not a-Al,O, [12]. A series of SAD patterns of the same aluminium oxide grain at various sample orientations shows that AI,,O,,N is formed (see Fig. 6) at the mctal/oxide interface.
250
C. Lung and M . Schutze
Fig.5. BSE image (a) and element distribution (b: oxygen, c: aluminium. d: titanium) of the oxide scale and mctal subsurface layer ofTi36A1 after oxidation at 900°C in air for 4h.
17.3.2.3 Ti36A1, 8OO0C,4 h After 4 h oxidation of Ti36Al at 800°C in air an outer oxide scale consisting of a mixture of X O , and A1,0, is also found as in the cases describcd before. Beneath this outer mixture a porous titania rich partial layer is observed (see schematic illustration in Fig. 7). At the metal/oxide interface aluminium oxide and Ti-rich phases are present similar to the sample oxidized at 900°C.From Fig. 8 which shows the metal/oxide intcrface it is obvious that due to the small grain size and the irrcgular appearance it is difficult to identify the phases developed at this site. In some regions where a high Ti content is found Ti,AIN is present according to the SAD patterns analysed. For SAD patterns, which cannot be assigned to Ti,AlN, other Ti-rich phases like Ti,AI or the new cubic phase may occur, but these could not be ascertained. The aluminium oxide at the metal/oxide is found to be A127039Nas at 900°C.
I7 The Initid Stages in the Oxidation of TiAl
25 1
Fig.6. a) TUM micrograph of the rnetal/oxide interface ofTi36A1 after oxidation at 900°C in air for 4 h. b) SAD patterns of AI,,O,,&
In some parts of the sample a certain featurc of thc metal/oxide interface is striking. As it is shown in Fig. 8 a necdle shaped grain is pcnetrating through the metal/oxide interface layer into the base metal. The analysis of this phase shows that Ti,O, is formed in this area of thc metal/oxide interfacc.
252
C. Lung and M. Schiitze
TiAl
Pig.7. Schematic illustration of the oxide scale and of the subsurface layer of'fi36AI after oxidation at 800°C in air for 4 h.
17.3.2.4 Ti36A1, 1000°C, 4 h Due to the higher temperature a much thicker oxide scale is formed on'I'i36Al during a 4 h oxidation at 1000°C compared to 800°C and 900°C. In Fig. 9 a BSE imagc of the cross-section of the oxide scale and of the base metal is presented. While the layered structure of the oxide scale is in principle the same as after oxidation at 900°C for 4 h the metal subsurface zone shows a significant difference. Contrary to an oxidation temperature of 900°C the bright layer beneath the oxide scale indicates the cxistance of an aluminium depletion zone. This is confirmed by the clement distribution images (see Fig. 9). In addition in the BSE imagc the depletion layer appears not to be continuous, but interrupted by a second phase. A more detailed analyses by TEM revealed the presence of single a,-Ti,AI grains which occur at equal distances at the interface (see Fig. 10) [12].They are surrounded by titanium nitrides and aluminium oxide with the latter also growing along the grain boundaries into the metal subsurface layer.'I'he phase identification of the nitrides yields the presence of TiN, whereas the SAD patterns obtained for the aluminium oxide indicate the formation of Al,,O,,N. Beneath the Ti,AI grains a layer with an aluminium content between X3AI and TiAl is obscrved.The analysis of this layer by electron diffraction yields the same result as after 0.5 h of oxidation at 900"C.The suggested new cubic phase fits very well to the diffraction patterns.The res'yks are summarized in a schematic illustration in Fig. 11.
17.3.2.5 Ti35A15Nb, 9OO0C,4 h In Fig. 12 a TEM micrograph and a schematic illustration of the oxide scale and o f thc metal subsurface zone of Ti35A15Nb after 4 h of oxidation at 900°C is shown. Beneath an outer oxide mixture of AI,O, and XO, a layer of coarse-grained XO, is observed.
I7 The Initial Striges in the Oxidation of 7iAl
253
Fig.8.TEM bright field image of the metal/oxide interface of Ti36A1 after oxidation at 800°C in
air for 4 h.
Towards the inner oxide scale a porous layer containing 30, and titanium nitrides (TiN, Ti2AIN) follows. At the metal/oxidc interface Al2,0,N is predominantly formed as results from clcctron diffraction indicatc. Contrary to the observations with Ti36A1
254
C. Lung rind M . Schiirze
Fig.9. BSE image (a) and element distribution (b: oxygen, c: aluminium, d: titanium) of the oxide scale and metal subsurface layer of'fi36A1 after oxidation at 1000°C in air for 4h.
n o depletion in aluminium can be detected in the metal subsurface zone. Even an enrichment in aluminium is locally ascertained. In addition internal oxidation occurs leading to the formaticm'hf internal aluminium oxide (see Fig. 12). Neithcr in the oxide scale nor at the metalhxide interface niobium oxides are found. Except from the base metal a signifkant niobium content only exists in the coarse-grained 'IiO? layer where the niobium is dissolved in the'I'iO,.
I7 The Initial Srages in the Oxidation of TiAl
255
Fig. 10.'IEM bright field image of the metal/oxide interface of Ti36AI after oxidation at 1000°C in air for 4h.
H
1w
TiAl
/
TuAl
b03'
NCP
AION
Fig. 11. Schematic illustration of the oxide scale and of the subsurface layer of'I'i36AI after oxidation at 1000°C in air for 4 h.
256
C. Long a d M. Schiitze
Pig. 12. TEM micrograph (a) and schematic illustration (b) o f the oxide scale and metal subsurfacr layer ofTi3SA15Nb alter oxidation at 900°C in air lor 4 h.
17 The IniiialStapes in the Oxidation of 1iAI
257
17.4 Discussion 17.4.1 Oxidation Progress 17.4.1.1 Stage I As can be seen from the weight gain curves the oxidation process can be divided into several stages. This is also shown in the schematic illustration in Fig. 13. Stage I of the oxidation process represents the initial stage with a very high oxidation rate. At the very beginning TiO, and A1,0, form due to the high oxygen partial pressure at the metal surface which is higher than the equilibrium oxygen partial pressure of the oxides [12,21]. During further oxidation aluminium is preferentially oxidized because of the higher thermodynamic stability of A1,0, compared to TiOiTiO, in equilibrium with the y-?'iAl at the metal/oxide interface [21].This leads to the formation of an aluminium depletion layer in the metal subsurface zone [4,22,23]. From investigations of the first minutes of oxidation [12] and from the results presented in this paper it can be concluded that at the metal/oxide interface no a-Al,O, is formed but either a metastable AI,O, or an aluminium oxynitride similar in structure to AI,,O,,N. ?'he AI,,O,,N has a crystal structure which is comparable to a metastable, intermediate stage of the Mg- or Ni-spinel [24,25]. It is conceivable that Al,,O,,N may be stabilized on the one hand by the presence of nitrogen in combination with the low oxygen par-
A
1
j
I1
III
- t
I:
n:
111:
Iv:
Gas
TiN,AlON NCP TiAl
Pig. 13. Schematic illustration of oxidation mechanisms at the mctal/oxidc intcrfacc during stage 11.
258
C. Larig nnd M. Scliiiize
tial pressure at the metal/oxidc interface and on the other hand by the dissolution of titanium. In contradiction to the binary phase diagram of TiAl the depletion layer in the metal subsurface zone after a 0.5 h oxidation at 900°C consists of a phase with a composition between y-TiAl und a2-Ti3AI.The analysis of this phase by electron diffraction reveals that this phase is equal to the recently reported new cubic phase (NCI’) which is similar in compositions toTi,Al,,?’i,AI or TiSAl3O2[ 17,18,19,20]. By the depletion of the metal subsurface zone in aluminium due to rapid oxidation during stage I the activity of titanium a.fiincreases. The oxygen partial pressure at the metalbxide interface is sufficiently high to form aluminium oxide but too low for titanium oxide formation. In some way nitrogen can penetrate the scale and the nitrogen partial pressure allows the observed formation of titanium nitrides at the metal/oxide interface when a certain value of the titanium activity is reached. ‘The high oxidation rate is caused by the fast formation of nitrides (which later become oxidized to TiO,) in addition to the slowly growing discrete aluminium oxide particles without significant barrier effect. ‘The aluminium depleted layer is consumed by the rapid formation of further nitrides as the absence of a depletion layer after 4 h oxidation at 900°C indicates. Dettenwanger et al. [26] who also report of titanium nitride formation at the mctal/oxide interface did not observe a depletion zone after 1 h oxidation, either. Because of the thin and non-protective oxide scale at this point of time a sufficient supply of nitrogen to the metalhxide interface is presumed to promote this process of nitride formation. Figge et al. [22] also found the formation of nitrides after 1.5 h of oxidation by SNMS investigations.
17.4.1.2 Transition from Stage I to Stage I1 Because of the observed microstructures of the metal/ oxide interface and of the metal subsurface zone after 0.5 h and 4 h of oxidation it is most likely that the transition of stage I to stage I1 is characterized by the consumption of the aluminium depletion layer due to nitride formation and the enrichment of Al,O,-particles in an outer layer of the scale which represents a preceeding step in the formation of a temporary A1,0,barrier. The consequence of the consumption of the aluminium depletion layer is a decrease in titanium activity a, at the metal subsurface zone. ‘llerefore further titanium nitride should be suppressed. However, due to the formation of A127039Nat the metal/oxide interface a local depletion of the metal subsurface layer in aluminium occurs again. This again leads to the formation of titanium nitrides. However, the ratedetermining step for nitride formation is the formation of AI,,O,,N, because Al2,0,,N formation is the basic requirement for aluminium depletion and subsequent titanium nitride formation. In contrast to stage I where an extensive aluminium depletion layer already exists when nitride formation starts the nitride formation during stage I1 is triggered by the preceeding A1,,0,9N formation.
17 The Initial Stuges in the Oxidation oJTiAl
259
17.4.1.3 Stage 11 During stage I1 linear oxidation kinetics are observcd. It is assumed that this linear oxidation behaviour is caused by the repeated cycle of aluminium oxide formation, subsequent local depletion of thc metal subsurface zone in Al, consumption of the Tirich metal phase by nitride formation since nitrogen is still present at the interface. 'fie nitrogen which is needed for this process is set free by the oxidation of the titanium nitrides to T i 0 2 at increasing oxygen partial pressure when the oxide scale grows inwards. 'I'he suggested mechanisms which hamper the renewed formation of an aluminium depletion zone and which lead to the formation of the mixed inner oxide scale are schematically illustrated in Fig. 14 [12]. It shows the situation at the mctal/oxide interface at three successive points of time. At the beginning of stage I1 (t = 1,) A12,0,,N (AION) and titanium nitrides are present in the oxide close to the metal/oxide interface within the metal subsurface zone beneath the interface, a Ti-rich
metastable Ah03 /
A1ox
metal / oxide interface
'l'i-richmetal
'lil
(NW
j
1 11
0
TiN
metastable A1201 I AlON
'Ti-rich metal
(NCP)
I
dissolution of metastable A1203 / AION and outward diffusion of A1 oxidation ofTiN to T i 0 2
I
t =t2
interface
EN rii
t-
Ti01
pore
t3
......... .........
POX
AION
%-rich metal
(NCV
TiAl
b 1: ......
Ti&
Ti02
metastable Ah01 I /\ION
Ti-rich metal
(NCP)
mctastable A l 2 a I AlON TiAl
......
Ti-rich metal
(NW
TiAl
Fig. 14. Schcrnatic illustration of thc stagcs ofTiAl oxidation by means of the weight gain curve and the corresponding microstructures of the oxide scale.
260
C Long and M. Schiitze
metal phase designated as new cubic phasc (NCP) being in contact with AlON (due to aluminium depletion by AlON formation) and y-TiAl being in contact with TiN (due to aluminium enrichment by TiN formation). As thc oxide scale grows inwards (t = tr) at the sites of high ’I’i activity a,,, titanium nitrides form. whereas aluminium oxide is formed at sites of sufficient aluminium activity aA,(TiAl).’I’he nitrides of the previous interface layer are oxidized to ’Ii02 when they become incorporated into thc oxide scale where a higher oxygen partial pressure exists. Furthermore it is assumed that the AION-phase becomes unstable in the vicinity of high’li activity and of nitride forming conditions and dissolves in the titania of the adjacent oxide scale. ‘I’hc dissolved aluminium diffuses outwards and reprecipitates in the form of a-Al,O, in the outer oxide scale at a higher oxygen partial pressure pO,.‘Ihc PO, dependend dissolution of AI,O, in TiO, is known from investigations of the long-term oxidation behaviour 141 and may explain the observed formation of a titania rich and porous partial layer adjacent to the metal/oxide. The transition of stage I1 to stage 111 (parabolic growth), which was not investigated in this study, occurs when in the outer oxide scale the outward diffusing dissolved aluminium forms a dense AI,O, barrier. The breakaway which takes place after 300-500h of oxidation (stage IV) was found to be caused by the dissolution o f this outer A1,03 barrier 141.
17.4.2 Temperature Dependence The observed differences in scale thickness and microstructure between the oxide scales and subsurface zones at the various oxidation temperatures seem to be mainly attributed to the different diffusion rates at the respective temperatures. Since the oxidation products formed do not show any differences in the temperature range of 800°C to 1000°C it is concluded that no significant effect of the thermodynamic stability on the composition and structure of the oxidation products occurs. From the calculations of Rahmel and Spencer [21] it is known, however, that the activity of A1 and Ti in the system Ti-AI varies depending on the tempcrature. Thus it has to be taken into account that the temperature may have an influence on the expansion of the phasc fields of some importar$ phases in the system ’li-Al-N-0. Nevertheless it is evidently the temperature which-’mainly influences the kinetics because the structure of the mctaYoxide interface, the formation of titanium nitrides, AI,,O,,N and an aluminium depleted metal phase is on principle always very similar. In this way the effect of different temperatures can, to a certain degree, be interpreted as that of a shift in the different stages of the oxidation process. ’bus, at 800°C an earlier state should be expected for the same oxidation time compared to oxidation at 900°C. However, an oxide scale which exhibits the same characteristics is found at 800°C indicating that at the investigated point of time stage I1 of oxidation has already been reached. One feature only found at 800°C is the formation of needle-shaped Ti,O, grains at the metWoxide interface (see Fig. 8). Because of the crack-like appearance one might speculate that by the growth of the oxide scale tensile stresses arisc in the metal subsurface layer which lead to fracture in this area due to embrittlcment of the metal subsurface layer by dissolved oxygen 118, 19, 27, 281.
I7 The Initial S~agesin rhe Oxidation of TiAl
261
Thus a conncction to rcgions of higher oxygen partial pressure is obtained and instead of nitrides titanium oxide is formcd at the mctal/oxide intcrfacc. As a consequence local temporary regions of high oxidation ratc arc formcd within the scale which may lead to nodules observed on thc surface of the oxidc scale. At highcr temperatures such cracks are not found possibly because of an increased plasticity of the aluminium depleted subsurface zonc or because of a rapid healing rate duc to enhanccd diffusion ~71. In contrast to the situation at 800°C the stage of oxidation invcstigatcd at 1000°C corresponds to a latcr s t a g cornpared to 900"C.'fie main difference conccrning the other samples investigated (i. c. at 800°C and 900°C) can be seen in the occurance of a two-phasc depletion zonc consisting of a?-Ti;Al and the new cubic phase. Evidcntly after stage I1 of oxidation an aluminium depletion zone is formed again as it is known from studies of thc long-term oxidation bchaviour, cf. e. g. [4]. It is also known that this zone is a two-phase dcpletion laycr aftcr longer oxidation times [4,20J.Thus, it is concluded that because of the higher temperature this stage is already rcachcd. The progressive formation of AlON at the metal/oxide interface leads to further depletion of the new cubic phasc in aluminium. As a consequcncc the formation of local a2-X3A1 grains at the mctal oxide interface is observed.The titanium activity increases at these sites and in the further oxidation process the formation of Ti oxidcs or nitridcs is favoured again thcrc. In this way the two-phasc subsurface laycr with diffcrent activities of titanium and aluminium leads to an inner oxidc scale mixcd of'TiO, and Al,O,.'Ihe differcnces found in the form that at 1000°C only AlON is found in contact with NCP while at 900°C it is AlON and TiN can be cxplained by thc effect of the higher scale thickness at 1000°C at thc same oxidation timc.Thc partial pressures at the interface are influenced by the thickness and the quality of the growing oxide scale. Assuming that the nitrogen partial pressure at the interface is defined by thc transport proccsses through the scale it is expected that with increasing oxide thickness the nitrogen partial pressure pN, at the interface decreases. By this means the nitrogen partial pressure pN, may drop below the cquilibrium partial prcssurc of nitride formation and thercforc a higher titanium activity may bc necded to form titanium nitrides.This can be an explanation why at a later stage of oxidation A1,0, is thc favoured phasc in contact with the aluminium depleted ncw cubic phase compared to the nitrides in the carlier stages. Quadakkers et al. 129,221 and Meier et al. 161 observed the disappearancc of nitrides aftcr somc timc which formed in the early stagcs.
17.4.3 Effect of Niobium Addition The striking features after oxidation of Ti35A15Nb for 4 h at 900°C are the slight enrichmcnt of aluminium in the mctal subsurface zone instcad of aluminium depletion, the preferred formation of AlON in wide parts of the metal/oxidc interface and the development of a rathcr densc, coarse-grained partial layer consisting of titania in the oxide scale. Several rcasons for the beneficial effect of niobium addition on the oxidation behaviour are discusscd in the litcraturc [ S . 10,111. Bcsidc the influcnce of niobium on thc a.JaA, ratio and expansion of the y-TiAI phasc field thc effect of doping of titania by niobium is often discusscd. By doping of titania with niobium the concentra-
262
C. I,ang and M . Schiitze
tion of oxygen vacancies and titanium intcrstitials should be reduced and thus the transport of oxygen inwards and titanium outwards may be slowed down [30]. If a doping mechanism is assumed it is also cxpectcd that in the same scnsc thc solubility of A1,0, in X 0 2 is decreascd, since the aluminium is dissolved as an interstitial [4]. Thus thc dissolution of AlON in titania and thc outward diffusion of aluminium is reduced. This may stabilize the AlON at the metalloxide intcrfacc leading to a dense AlON barrier there and, thus, to slow oxidation kinetics.
17.5 Conclusions From the prcsent investigations it is concluded that the dccisivc processcs of TiAl oxidation take place at the mctal/oxidc interface. First of all no a-A120, is formed at thc mctal/oxide interface, but a metastable AI,O, or aluminium oxynitride similar in structure to Al,,O,,N which dissolves in the surrounding titania whcn during oxide scalc growth the titanium activity incrcases around the AlON particles and titanium nitride formation takes placc. Thc titanium nitridcs occur at the metal/oxide interface when critical values of the pO,/pN, ratio and of the titanium activity duc to aluminium depletion of the metal subsurface zone are rcachcd. Thc aluminium deplcted metal subsurface zone consists of a ncw cubic phase in thc initial stages of oxidation and in addition of a,-Ti,Al aftcr further aluminium depletion of the metal subsurface layer. Thc intcrrnediate formation of titanium nitrides which are oxidizcd to TiO, during further oxidation prevents the formation of a continuous A1,0, layer, resulting in cnhanccd oxidation. 'lhe alternating processes of A1,0, formation, subsequent aluminium depletion and titanium nitride/oxide formation lead to an inncr oxidc scalc mixed of AI,O, and TiO,. In the niobium containing alloy which shows a bettcr oxidation rcsistance the doping of titania with niobium may reduce the dissolution of AION. By this means a thin layer AlON is formed at the interface lcading to a reduced oxidation rate.'Thus it is assumed that the oxidation behaviour of titanium aluminidcs could bc improvcd by stabilizing the aluminium oxidc at the mctal/oxidc interface cither by prevcntion of aluminium depletion of the metal subsurface zone or by reduction of AI,O, dissolution in TiO,.
17.6 Acknowledgements This work has been financially supported by the Dcutsche Forschungsgcmcinschaft (DFG), which is gratefully acknowledgcd.
17 7he Initinl Stages in the Oxidation of TiAl
263
17.7 References N.S. C‘houdhury,I I . C Graham, J. W.Hinze: in “Proc. of the Symp. on Properties of IIigh Temperature Alloys”, Iilectrochemical Society, Princeton, N.J. (1977) 668. R.A.I’erkins, K.7: Chiang, G.II. Meier: Script. Met.21 (1987) 150.5. G. Welsch.A.I. Kahveci: in “Oxidation of high-temperature intermctallics”, T. Grobstein, J. Doychak (cds.), the Minerals. Metals and Materials Society (1989) 207. S. Becker,A. Rahmel, M. Schorr, M. Schiitze: Oxidation of Metals 38 (1992) 425. A. Rahmel, WJ. Qiiadakkers. M. Schiitze: Materials and Corrosion 46 (1995) 271. G. I I . Meiw I,.’S. Pettii, S. Hid: J. Physique 1V 3 (1993) 395. N. Zheng, W.J. Qundukker.s,A. G I , I I . Nickel Oxidation o f Metals 44 (1995) 477. K Shidn, H.Anadn: Oxidation of Metals 45 (1996) 197. K. Maki, M. Shioda, M. Sayashi, 7: Shimizu,S. Isobe: Matcr. Sci. and Eng. A153 (1992) 591. II. Nickel, N. Zheng, A. Elschner, WJ. Quadakkers: Mikrochimica Acta 119 (1995) 23. WJ. Qundakkers, A . Llschrier, N. Zheng, H. Schuster, H. Nickel in “Microscopy o f Oxidation 2”; S.B. Ncwcomb, M.J. Bennett (eds.).’Ihe Institute of Materials London (1993) 488. C. I,ang, M. Schiitze:Oxidation o f Metals 46 (1996) 255. Joint Committee on Powder Diffraction Standards, File No. 2633. G. t’etzow, G. Effenberg: “Ternary Alloys vol. 7”,VCH Weinheim (1993) 386. R.W.Beye, I<. Grorrsky:Acta Metall. Mater. 42 (1994) 1373. Y E Cheng, I< Dettenwanger, J. Mayer, E. Schunzann, M. Riihle: Scripta Materialia 34 (1996) 707. S. Nonrbakhsh, 0.Sahin, H. Margolin: Acta Metall. Mater. 43 (1995) 3035. W.E. Dowling, W T Ihnlon: Scripta Metallurgica et Materialia 27 (1992) 1663. W 7: Donlon, W E .Dowling: Mat. Res. Soc. Proc. 288 (1993) 629. N . Zheng, W Fischer, H. Griibmeier, V Shernet, WJ. Quadakker.s: Scripta Metallurgia et Matcrialia 3.3 (1995) 47. A . Rahmel, PJ..Spencer: Oxidation o f Metals35 (1991) 53. U.Figge,A . Elschner, N Zheng, I I . Schuster, WJ. Quadakkers: Fresenius J. Anal. Chem. 346 (1993) 75. A. Gil, H. Hoven, E. Wallura, WJ. Quadakkers: Corrosion Science 34 (1993) 615. D. Michel Rev. Int. HautesTemper. et Refract. 9 (1972) 225. H. Saalfeld, H. Jagodzinski: Zeitschrift f u r Kristallographie 109 (1957) 87. E Dettenwanger, E. Schumann, M. Kiihle, J. Rakowski, G.H. Meier: Proc. MRS Fall Meeting (1994). M. Schiitze, M. Schmitz:Trans. Mat. Kes. SOC.Jpn. 14A (1994) 219. J. Kameda, C.R. Gold, E.S. I,ee, T E . Bloomer, M. Yamaguchi:Mat. Res. SOC.Symp. Proc. 364 (1995) 1297. WJ. Quadakkers, A. Elschner, N. Zheng, H. Nickel: 12th Int. Corrosion Congr., 19-24 Sept. 1993,Houston, Proc. NACE Houston (1993) 3842. P Kofytad “High Temperature Corrosion”, Elsevier Applied Science London (1988) 295.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
18 Development and Microstructure of the Al-Depleted Layer of Oxidized TiAl E Dettenwanger, E. Schumann, J. Rakowski, G.H. Meier and M. Riihle
18.1 Introduction Alloys based on y-TiAl bccome important due to their combination of good mechanical properties at high temperaturcs and low density comparcd to superalloys. The use of y-TiAl at elevated temperatures is, however, limited through the insufficient oxidation behavior. The formation of an aluminium depletcd layer between the oxide and thc alloy has been observed by many investigators [1-4].The importancc of the aluminium depleted subsurface layer which forms during the high temperature oxidation of y-TiAl seems to play an important role for both oxidation and mcchanical properties of the alloy. Investigations of y-TiAl oxidized in oxygen by different investigators [3,S, 61 showed that the Al-dcpletcd region develops from a single phase to a two phase region consisting of cw,-Ti,Al and a simple cubic-phase not known in the ternary system Ti-AI-0. This study focuses on the microstructural time development of the aluminium depleted region during the oxidation process for different exposure times in air.
18.2 Experimental Thc alloy investigatcd was Ti-50(at%)AI which had becn produced by drop casting on a copper chill and then isostatically hot pressing at 1300°C.The resulting y-'fiAl grains had a sizc of hundred to scveral hundrcd microns in diameter. Slices cut from the alloy ingot were mechanically polished down to a 1pm diamond finish, ultrasonically cleaned in acetone and alcohol and isothermally oxidized at 900°C in air for 1h, 100h, 140 h, 200 h, and SO0 h.'The specimens were investigatcd in cross scction with scanning electron microscopy (SEM) and transmission electron microscopy (TEM). The preparation method for the TEM specimens was a face to face preparation technique. A sandwich of two oxidized specimens was glued togethcr and embedded in a thin brass cylinder of 3 mm diameter. Disks were mechanically cut from the tube, polishcd, dimpled, and argon ion milled.Thc cut disks from the tube served also as the starting material for the metallographic cross sections investigatcd with SEM and light microscopy. Bright field imaging, sclectcd area electron diffraction (SAD), energy dispersive Xray spectroscopy (EDS), electron energy loss spectroscopy (EELS), and electron
266
E Detienwanger, E. Schurnann, J. Rnkoivski, G. H . Meier and M.Riihle
spcctroscopic imaging (ESI) were performed with a JEOL 2000 FX, a dedicated VG S I Y M IIB 501. and a ZEISS 912 OMEGA. Vickcrs microhardness measurements (VIIN) were pcrfornied using a Fischcrscopc HlOO using an indent load of 20mN. SEM cross section studies wcrc carricd out with a JEOL JXA 6400 electron probe microanalyzer (EPMA) using secondary electron image modc (SE) and backscattcring clcctron image mode (BSE). The RSE contrast diffcrcnce between thc various phascs beneath the oxide scale was weak due to thc small composition variation and could be greatly improved using metallographic etching and light microscopy. Chemical compositions of the aluminum dcplctcd phascs were measurcd using analytical TEM and the electron probc microanalyzcr. ‘I’he EPMA nieasurements wcrc pcrformed with wavelength dispersive X-ray spectroscopy (WDS) using purc Ti, A1 and
Fig. 1. Bright field TEM cross section of y-TiAl oxidized for 1 h and corresponding ESI maps.
18 Developmenr arid Microstricctiire of the Al-Depleled Luyer 0.f Oxidized TiAI
267
Fe#, as standards. EDS measurcnients with the TEM cross scctions were performed for the Ti and Al contcnt using the base metal y-TiAl as standard and EELS was used for mcasuring thc 0 contcnt because of the ovcrlapping of theTi-I,
18.3 Experimental Results Figure 1 shows the bright field TEM image of the sample oxidized for 1h and the corresponding elemental maps. The oxide scale had a thickness of 5 pn and showed the typical structure of an outer rutil layer followed by a thick mixed oxide layer. ‘Ihe microstructure of the metal/scale intcrfacc has developed to a characteristic feature of titanium nitrides interrupting the alumina as reported in [7].The formation of the aluminum depleted phase beneath the oxide can be seen from the ESI maps but has not developed to a continuous layer at this stage as reported recently [8]. Selected area electron diffraction patterns of the aluminum depleted grains show the cubic structure of this phase which will be called X-phase in this study. The X-phase is well developed after an oxidation time of l00h and forms now a continuous layer beneath the oxide as can be seen from the BSE image (Figure 2). The bright field image of Figure 3 shows the grains of the X-phase and a corresponding SAD pattern indicating the cubic structurc of thc phase with a lattice parameter of 0.692 L 0.005 nm.The spacegroup of the X-phase was determined to P432 or P4,32 which is described in detail in [9]. No special crystallographic relationship between the X-phase and the y-TiAl base metal could be found. The metallographic cross section of a sample oxidized for 140 h is shown in Figure 4. A second phase appears in the depletion zone locally but most of the depletion zone consists of single phasc X-phase.The new grains arc located only at the y-TiAIIX-phase interface. Such a region is shown in the ‘ E M cross section (Fig-
Fig.2. SBM image of y-’IiAl oxidized for 100 h showing the single phase Al depleted layer.
268
II' Dettenwnnger, 1;. Schumnnn, J. Hakowski, G. H. Meier and M . Kiihle
Fig.3. Bright field'I'EM cross section of y-TiAl oxidized for 100 h.'lbe SAD pattern rcvcals the cubic structure of the Al depleted grains.
Fig.4. Metallographic cross section showing the Al depleted layer after oxidation for 140 h.
18 Drvelopnirrii arid Microsfriicturr o,f the Al- Depleted Layer o f Oxidized TiAl
269
Fig.5. Bright field TEM cross section of y-'IiAl oxidized for 140 h showing thc a2-'Ii3Algrain (SAD) in bctwccn the X-phase and the base metal.
ure 5) where cw,-Ti,Al was identified with electron diffraction in between the X-phase and the base metal. After 200 h the depletion zone shows clearly a two phase structure with most of the a,-Ti,Al grains not reaching the X-phadscale interface (Figure 6). After oxidation for 500 h the subscale has developed completely into a two phase region with columnar grains (Figure 7). 'Ihe chemical composition of the different phases measured with two different methods are given in Tablc l.'I'he content of N was below the detection limit of about 2 at.%. The thickness of the oxide and the Al-depleted region varies with time as given in'pdble 2. 'Ihe values showed a scattering of several microns due to the rough metauscale interface. The microstructural difference between the two phases of the depletion zone is shown in the bright field image of Figure 8. The amount of defects in the X-phase is much less than in the a2-TilAIas can be seen from the resulting extinction contours of the bright field'rEM image.
E Dettetiwangi>r,E. Schitmnnn, J. Hnkowski, G.H. Meier and M. Riihle
270
The difference in thc microstructure was also reflected by microhardness-measurements done on the various phases. The X-phase showed the highest VI IN followed by a2-Ti3A1and y-TiAl (lable 3). ‘Ihc brittleness of the phase was manifested also during preparation of the spccimens which resulted at some places in transcrystallinc cracks r u n n i n g through the grains especially of the X-phase. Table 1. Chemical composition (at.%) of the aluminum depleted two phase region. -
-
-
Phase
Element
Microanalyser (WDS)
TEM (EDSOXLS)
X
Ti A1 0 Ti A1 0
54 2 31 1 15+2 65.5 I- 4 27.5 5 2 71-3
57 5 6 33 5 5 10 ? 8 65 f 5 29 4 655
a2
+
+
*
~~
Table 2. Thickness determination of oxide scale and subscale. Xrne
lh
l00h
140h
200h
500h
Scale Ikplcted Layer
5p.m -
20p.m 3p.m
25p.m 4pm
50p.m 9pm
70pm 15ym
.
-
~-~
Table3. VHN of the various subscale phases (20mN, 30 sec). Phase
VHN
y-TiAl a,-Ti,Al X-Phase
487 5 139 1470 5 241 2770 f 522
~
-
-
Fig.6. Metallographic cross section showing the Al depleted layer after oxidation for 200 h.
18 Developmen: and Microstructure of the Al-Depleted 1,ayer of Oxidized TiAl
27 1
Fig. 7. Metallographic cross section showing the two phase structure of the A1 depleted layer after an oxidation timc of 500 h.
Fig.8. A1 depleted zone in TEM cross section oxidation for 500 h consisting of X-phasc and txz-li,Al.
272
I;: Dettenwcitiger, L. Schrinzuiin,J. Knkowski. G.11.Meier und M. Kiihle
18.4 Discussion Figurc 9 shows thc schcmatic cvolution of the subsurfacc scale for the oxidation of yTiAl in air at 900°C resulting from thc investigated spccirncns in this study.The aluminum depleted layer fornicd during thc oxidation in air of y-'fiAl develops from a single phase region consisting of thc sirnplc cubic X-phase to a two phasc region consisting of a2-Ti3AIand the X-phase. The formation of the X-phase starts aftcr 1 h and transforms after about 140 h into the two phasc region. Chemical analyses showcd that both phases contain oxygen. The measured oxygcn content of the a,-Ti,Al of 7 at.% is in good agrccmcnt with rcccntly published data of the oxygcn solubility in a2-Ti3A1of 8.6 at.% [lo]. If this valuc is the solubility limit is questionable due to other measurements giving a higher solubility limit for a2-Ti3AIof 12 to 15 at.% 1111. The formation of an A1 dcplctcd phasc during oxidation of y'fiAI occurs also in oxygen and develops after longer exposure times in a two phasc region [2,3,6]. FIlie grains of the single phase depletion zone show also a simple cubic structurc with lattice parameter a = 0.69nm and seem to belong to the same spacegroup [12].Howevcr, the oxygcn contcnt was determined to about 20 at.%" [3] and is therefore higher than for thc X-phasc in this study. This suggests that in both c a m oxidation in air and oxygen, thc first forming phasc dcplctcd in A1 is structural identical but may vary in cheniical composition. This phasc forms whcthcr thc oxidation product is the rapid growing mixed oxide - typical for thc air casc - or a passivating alumina scalc formcd in oxygen. At least for the early stages of oxidation thc subsurface zone does not dctcrminc thc oxidation mode and is only a result of aluminum dcplction of the base metal due to thc oxidation process. Concerning the oxidation in oxygcn it is possible that the formation of a2-X3A1in thc subsurfacc zone can lead to a breakdown of the alumina scale but this must bc provcd by futhcr investigations. The presented microstructural dcvclopmcnt of the A1 depleted region suggcsts that after longer oxidation time a phase transformation from the X-phase to the a2-Ti3Al occurs with a heterogenous nucleation occurring at the X-phasely-TiAl interface. The a2-'fi3AI grains grow rapidly into the X-phase grains with a direction mainly perpen-
I
lOOh
lh I
Scale
Scale
Fig.9. Schematic time development o f the depletion layer for 900°C in air.
18 Development and Microstrricture of the Al-Depleted I,ayer o,f Oxidized 7iAI
273
dicular to the surface. The growth process has to bc accompanied by diffusion due to the chemical differences between the two phases. The X-phasc shows a high value of hardness and is therefore one reason for thc observed reduction of the ductility of oxidized y-‘IiAl specimens [5,13]. The cY,-Ti,Al is significantly harder than the y-‘IiAl although a comparison with data from the literature indicate an cqual hardness of both phases [14].’1’he increase in the hardness of Ti,AI can be explained through the solution of oxygen which leads to the embrittlcmcnt of this phase.
18.5 Acknowledgement This work was supported by the Deutsche Forschungsgemeinschaft. The authors wish to thank K . Steinfein for the excellent metallographic work.
18.6 References [l] N.S. Choudhury, H.C. Graham. J. W Hinze: Oxidation Ikhavior o f Titanium Aluminides, Z.A. Foroulis, F.S. Pettit, Eds., Proccedings of the Symposium on I’ropcrties of High ?‘cmperaturc Alloys (?he Electrochemical Society, I’rinccton, N.J.), (1976), pp. 668. 121 S. Decker, A. Hahmel, M. Schorr, M. Schiitze:Oxid. Met. 38 (1992) 425. [3j h! %hen;, W Fischei II. Griibmeier, V! Shemet, WJ. Quadakkers: Scripta Metall. Mater. 33 (19951 47. 141 k Ilerold-Schmidt, R. Opolka, S. Schwantes: Prakt. Mctallogr. 30 (1993) 344. [5] W E .Dowling Jr., W 7:Donlon: Scripta Metall. Mater. 27 (1992) 1663. [6] R. W Reye, R. Gronsky:Acta metall. mater. 42 (1994) 1373. 171 J.M. Hakowski, E S Pettit, <;.I{. Meier; h: Dettenwanger, E. Schumann, M. Hiihle: Scripta Metall. Matcr. 33 (1995) 997. [XI E Dettenwanger, E. Schumann, M. Riihle, J.M. Hakowski, G. H. Meier: MRS Symp. Proc. 364 (1995) 981. [Y] k Cheng, I:‘ Dettenwanger, J. Mayer, E. Schumann, M. Kiihle: Scripta Metall, Matcr.34 (1996) 707. [lo] k: Shida, II. Anada: Matcrials ‘Iransactions JIM35 (1 994) 623. [ I l l M.-X. Zhung, K.-C Ilsieh,J. DeKock, Y A .Chang: Scripta Mctall.Mater.27 (1992) 1361. [I21 R. Beye, M. Verwerft,J.7:M. DeHosson, R. Gronsky: to appear in Acta Metall. Mater.. [13] G. II. Meiel-:Personal Communication. [14] M.P Hrady, J.L. Smialek, D.L. Humphrey: MRS Symp. Proc.364 (1995) 1309.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
19 Beneficial and Detrimental Effects of Nitrogen on the Oxidation Behaviour of TiAl-Based Intermetallics WJ. Quadakkers, l? Schaaf, N. Zheng, A. Gil and E. Wallura
19.1 Introduction With respect to oxidation rcsistancc, y-titanium aluminides differ significantly from conventional Ni- and Co-bascd superalloys which are commonly used in highly stressed turbine components [l, 21: in spite of the high aluminium contcnts, titanium aluminidcs do not easily form protectivc alumina scalcs upon high-tcmperature exposure. Scvcral authors have found thc oxidation resistance of titanium aluminides to differ significantly in air and oxygen [3,4]. In spite of a considerablc number of publications on this effect the undcrlying mechanisms have not yet bccn fully clarified. Choudury ct al. [3] and Meier et a1 [5] observcd for binary alloys containing 50-54 At. YO A1 higher oxidation rates in air than in oxygen. Becker ct al. [4] observed for Ti-50Al and 'I'i-49Al-lV at 900°C a highcr oxidation rate in oxygen than in air; howcvcr, at 1000°C they found forTi50AI the oppositc effect. Zheng et al. [6]found forTi-50AI at 900°C higher, for Ti-48Al-5Nb lower oxidation ratcs in air then in Ar/O,. In the present study the oxidation bchaviour of three binary titanium aluminidcs, 'Ii45AI,Ti48hl and'Ii50AI (additions given in at.%), as wcll as a number of ternary alloys with Nb contents bctween 2 and 10 YOwere studied at 900°C in ordcr to clarify the apparent contradictory effect of nitrogcn on thc oxidation resistance of titanium aluminides.
19.2 Experimental The following alloys were produced by induction melting in an argon atmosphere (additions in At.-'%): binary alloys containing 45,48 or 50 at.-% Al and the ternary alloys, 'li45A12Nb, Ti45A15Nb,'Ii4SAllONb, Ti48A12Nb,Ti48A15Nb. Disk shape specimcns of l0mm diamcter and 2mm thickness wcre prepared from the cast ingots and then ground to a 1200 grit surfacc finish.Thc oxidation kinctics at 900°C in air and in argon + 20 YO oxygen were invcstigated using a KORAL thermobalance. Spccimens were examined before and after oxidation using optical microscopy, scanning electron microscopy (SEM), energy dispcrsive X-ray analysis (EDX) and X-ray diffraction (XRD). For studying the nitrogen distribution in scalcs and sub-surface layers, selected spccimens wcre analyscd by nuclear reaction analysis (NRA).
276
WJ. Qicadakkers, I? Sckaaf; N. Zherig, A. Gil and E. Walhtra
19.3 Results 19.3.1 Binary Titanium Aluminides Fig. 1 shows the results of the “3-analyses during the isothermal oxidation of thc three binary alloys at 900°C in Ar/02.It is obvious, that the weight changes decrease with increasing aluminium content. A more detailed analysis of the data reveals that the weight change curves do not obey a parabolic time dependancc. In all cases. the rate constant K, (defined as [Am]? = K;t) increases with increasing oxidation time. A reason for this result can at least partially be derived from the metnllographic cross sections: the alloys tend to oxidize with different rates on different areas of the spccimen, an effect which becomes more apparent with increasing Al-content (Figs. 2. 3). After a given oxidation time the specimen is partially covered by a thin alumina based scale, whereas in other places, a breakdown of this protective scale is observed leading to localized rapid growth of a mixed alumina/titania scale (Fig. 3). As the number of mixed scale protrusions increases upon continuing exposure. the overall oxidation rate measured during the TG analysis increases with increasing oxidation time. The relative surface coverage by the mixed scale increases with decreasing Al-content (Figs. 2,3). Beneath the initially formed, slowly growing alumina based scale (Fig. 3a) the suhsurface depletion layer is single-phase [7,8]. It consists of a ternary ‘1.1-AI-0 compound of approximate composition Ti,A1,02 designated as %-Phase by Z h n g et al. [ 71 and recently as X-Phase by Rakowsky et al. 191 and as NCP by f,ang and Schiitze 1101. It is probably the same compound as described by Doivling and Donlon of given compositionTi,AI [ll]. Beneath the mixed scale the depletion layer is two-phase (Fig. 3 ) . consisting o f %Phase and oxygen containing a,-?‘i,Al [7, ~ ] . T Wdifferent O morphologies can he tlistinguished beneath the mixed scale. In the one case a,-fingers are growing from the alloy/deplction layer interface intq,.the Z-Phase, which is in contact with the inner part of the oxide scale (Fig. 4). In the other case, a2 and Z-Phase are parallclly arranged (Fig. 3c) and are both in contact with the oxide scale [8].’lhis last mentioned morpho35 z30-
.~
1 -
Ti 45AL
I
6 25-
fl
m v
20fl
0
F150 c
N
Z
Z X
X
M
z
x*
,--
Ti 48AI
I9 Beneficial and Detriniental Eflecrs of Nitrogen on the Oxidation Rehaviortr
277
Mg.2. Metallographic cross
sections of Ti4SA1 after oxidation in A r /Q at 9 0 0 T a) 25 h, b) 65 h, c) 100h, enlarged view near scale/ alloy interface.
278
WJ. Quudakkers. I! Schauj N. Zheng, A. Gil and E. Wulliiru
7 mixed scale
c) E’ig.3. As Fig. 2 for X48AI. Arrow in Fig. c indicates start of internal oxidation.
logy is accompanied by the occurence of internal oxidation (Fig. 2c, 3c). The internal oxidation occurs after longer oxidation times whcreby the onset time incrcases with decreasing AI content (Figs. 2,3). Fig. 5 compares the oxidation behaviour of the three binary alloys in air and Ar/O,. For Ti45AI the oxidation rates appear to be similar in both atmospheres. For ’I’i48AI the oxidation rates in air are initially slightly higher, later slower than in Ado, . For TiSOAl the weight changes in air are higher than in Ar/O,. The sub-surface layeraformed beneath the mixed scales after short oxidation times in air and Ar/O, slea?ly differ. For TiSOAI no clear depletion layer can be observed after air oxidation for 24 h (Fig. 6 ) . Instead, optical metallograhpy clearly revealed thc presence of a very thin,Ti-rich nitride layer beneath the oxide (Fig. 6). Similar ohservations were made for Ti48AI whereas Ti45A1 formed after this oxidation time already a clear depletion layer similar to that in Ar/O,. After 100 h oxidation in air, also for TiSOAI the two-phase depletion layer was found, whereas for ‘I‘i48A1, still a thin nitride layer was observed (Fig. 7).
19 Beneficirrl crnrl Derrirnenral Effects of'Nitrogen on the Oxidation Behaviour
279
Fig. 4. Ti50Al after 100 h oxidation at 900°C in Ar/O,,a) Metallographic cross section in area where mixed scale has formed, b) tapercd section.
19.3.2 Nb-Containing Alloys Figs. 8 and 9 compare the oxidation rates of ternary Nb-containing Ti-aluminides with 45% A1 at 900°C in Arlo, and air. As described previously [6, 121 in all cases the weight changes initially tend to be smaller in Ar/O, then in air. During air oxidation, the alloys show approximately parabolic oxidation rates after an initial rapid oxidation as shown in Fig. 8, derived from data in reference [12]. TG-data up to l00h in Ar/O, shown in reference [12] indicated that thc oxidation rates of high Nb alloys increased after longer times. This is confirmed by the data up to 300 h presented in Fig. 9. The metallographic examinations show, for the alloy containing 5 or 10 YONb after exposure in Ar/O, a clear subsurface depletion layer (Fig. 10). After longer oxidation times internal oxidation occurs (Fig. lo), preferentially at the boundaries between the two-phases in the depletion layer (Fig. l l ) , in agreement with findings in reference [13].This is in contrast to the oxidation in air where no subsurface depletion layer or internal oxidation is observed by optical metallography (Fig. 10). The metallographic observations revealed formation of a thin nitride layer at the scale/alloy interface (Fig. 10). This was in agreement with prcvious SNMS-analyses 1141 and new NKA-studies after oxidation at 800°C (Fig. 12).
W.J. Quudukkers, F! S c h u f ;N. Zheng, A. Gil and E. Wdliira
280
0
40
20
80
60
Time ( h )
10
I Ti 50AI, 9OO'C I
,'
/*z argon/oxygen
F**+++*ctt-*-**1
20
-
I
40
.
,
60
Time ( h )
a
,
80
,
-
1
I0
Fig.S. Wcight change data during isothcrmal oxidation o f Ti45Al.'l'i38AI,m d TiSOAl ;it 900°C' in Arlo, and air.
19 Rerwficial und Detrirntwtal Iiffecls of Nitrogen on (he Oxidation Behaviour
281
Fig.6. K50A1 after 24 h oxidation at 900°C; Metallographiccross sections in specimen areas where a mixed scale has formed. a) Ar/O, and b) Air.
Fig.7. Metallographiccross sections after 100 h oxidation at 900°C in air, a)'fi50AI, b)Ti48AI.
b E
v
54-1
Y
a
3-\
g -
2 - \
'\
'-. - _- _- - - - - Ti45A15Nb - - _ _ _ _-_ _ T i 4 5 A l l ONb
------.
'
-.-O rEZ-l
0
20
40
60
Time ( h )
I
80
'
1 I0
Fig.8. K -values as function o f time forTi45A1 with various Nb contents during 100 h isothermal oxidatid: in air at 9 0 0 [12j. ~
W.J. Qiradakkers, P Scharij N. Zheng, A . Gil urid E. Wallurn
282
,,
3.50
50
0
4
150 200 Time ( h )
100
4
.-I
-
L
2
250
350
300
L900C ' inAr/On]
3
1E-01=
Ti45AIlONb
.,,'
.-
-______-----
8: 4
,
,
,
,
,
,
,
Ti45A12Nb r
,
.
Fig.9. Isothermal oxidation of Ti45A12Nb and'Ii45AllONb at 900°C in Ar/O, up to 300 h. a )
Fig. 10. Metallographic cross sections of 'Ii48A15Nb after oxidation at 9 0 ° C in Ar/O, (a, c) and air (b, d) for 100 h (a,b)and3(H)h (c.d).
19 Beneficial and Detrimentnl Effects o,f Nitrogen on the Oxidation Behaviour
283
Composition in At-% Position
50,O
36,5
48,4
Nb
Fig. 11. SEM-image and EDX-analysis (excluding oxygen) ofli48A15Nb after lOOh oxidation in A d o , at 900°C.
0
200
400
600
Depth ( nrn )
800
1
w
Fig. 12. Nitrogen distribution inTi45A15Nb after oxidation at 800°C in air (NKA data).
284
W.J. Qiinclnkkers, I! Schnaf;N. Zheng,A. G I and E. Wallura
19.4 Discussion For rationalizing thc discussion of the cxpcrimental results presented, wc rcfcr to thc schematical reprcscntation of the various stages of oxidation of titanium aluminides shown in Fig. 13. 'I'his schematic diagram describing the oxidation bchaviour in pure oxygcn is uscd, although the authors arc well aware of the fact that the phase equilibria in the depletion layer occuring during air oxidation might be more complex [lo]. Fig. 13 is based on considerations in Ref. [13] and expcrimcntal results in Refs. [7] and [SJ. In oxidation stage I an alumina based scalc and a single-phase depletion laycr cotisisting of Z-Phase is formed. In stage I1 a mixed titaniahlumina scalc is forming whereby ci2 is growing in a fingerlikc manncr from the alloy side into the Z-Phase depletion laycr 181, only the Z-Phasc bcing in contact with the inncr part of the oxide scale (Fig. 4). In stage 111 the ci2- and Z-Phase are parallely arranged in the depletion layer whcreby both phases are in contact with the oxide.This is accompanied by intcrnal oxidation of aluminium and a destruction of thc outer alumina-rich barricr layer [4, 131. 'Ihese three stages of oxidation can clearly be seen after different tiincs of oxidation of Ti48Al in Ado, (Fig. 3): after 24 h oxidation (Fig. 3a) the main part of the specimen exhibits oxidation stage I. After 65 h only few areas exhibit stagc I whereas the other areas show oxidation stage 11. After 100 hours (Fig. 3c), the oxide scalc shows a morphology corresponding to a change from stagc I1 into 111. For alloy Ti4SAl after 100 h, oxidation stage 111 is prcscnt on the wholc spccimcn, accompanied by high scale growth rates (Fig. 2c). Presence of nitrogen in thc atmosphere has onc, frequently described cffcct: it shifts the switch-over from oxidation stage I into stagc I1 to shorter times [5, 6, 131. This might be related to the formation of Ti-rich nitride in the early stages of oxidation [S,
/. /
/'
,/-A
time
______*
I
Fig. 13. Schematic illustration of the oxidation stages of y-titanium aluminides during oxidation in oxygen. Diagram based on considerations in reference [13] and data in referencc [7] and [8].
19 Beneficid rozd Detritrientnl Effects 0f Ni~rugerion the Oxidntion Rehaviour
285
12,141 thereby hampering the development of the dense alumina-rich scale. It was also proposed, that the destruction of the initial protective scale is related to oxidation of the initially formed Ti-rich nitride into Ti-oxidc [14].The destruction of the protective aluminium rich scale causes for a given oxidation time the oxidation rates in air to be larger than those in Ar/O,. This effect will be more pronounced for alloys with a higher tendency to form in Ar/O, the initial protective scale over the whole specimen surface, i. e. for alloys with high Al-contents. as shown in Fig. Sc. At 900°C for Ti50Al the oxidation rates are much higher than in Ar/O, up to 100 h, for Ti48A1, however, up to only around 20 hours (Fig. 5). For Ti4SA1 oxidation stage I in A r / 0 2is so short, that hardly any difference is observed between air and Ar/O, oxidation. After the switch-over from oxidation stage I into stage I1 has occurred, the effect of nitrogen is less easily to be explained. For l’i48Al the presence of nitrogen appears at least for some time to be beneficial (Fig. 5). Metallographic analyses revealed, that during those time periods the specimens oxidixd in air showed a thin, nitride-rich layer beneath the oxide scale (Fig. 7) whereas in Ar/O, a two-phase depletion layer together with internal oxidation occurred (Fig. 3c). Apparently, in the time periods studied, the presence of nitrogen retards the shift from oxidation stage I1 into stage I l l in the case of Ti48AI, whereas for TiSOAI this does not seem to be the case. Thc retardation of the oxidation rates by the presence of nitrogen in oxidation stage 11, which could only be detected in some eases for the binary alloys, was observed in all cases for the Nb-containing alloys studied for times up to 300 hours. For the Nb-containing alloys the time period for oxidation stage I in Ar/O, is relatively short [12]. For instance in the case of Ti48A15Nb after 24 hours, already oxidation stage I1 prevails [12]. After longer times a clear shift to oxidation stage I11 occurs, accompanied by internal oxidation. This change from oxidation stage I1 into stage 111 can be observed in Fig. 10a (stage I1 in the left part, stage I l l in the right part of the figure). In all cases studied, the presence of nitrogen in the atmospheres prevented thc change from oxidation stage I1 into stage I11 for the Nb containing alloys: a near-continuous Ti-rich nitride layer rather than internal oxidation was found in the sub-surface layer (Figs. 1Ob, d). The prevention of internal oxidation by the formation of a Ti-rich nitride layer can be explained in the following way:?he formation of the nitride will decrease the Ti/AIactivity ratio in the alloy beneath the surface scale thereby promoting alumina to precipitate in the inner part of the scale rather than as internal oxide. It is, however, not easy to clarify, why the presence of nitrogen in the atmosphere retards the shift from oxidation stage I1 into stage 111 in all cases studied for the Nb-containing alloys, but only in some cases for the binary alloys. In reference [14] it was proposed, that the Tirich nitride can initially be present beneath the mixed scale as a near-continuous layer. This is confirmed by the recent TEM studies of Lung and Schiitze [lo] although these authors also found more complex nitrides and oxynitridcs along with the main nitride ’IiN. Depending on the growth mechanisms of the oxide scale, the pN,/pO,-ratio at the oxide/alloy interface, can after longer times change in such a way that the ncar-continuous nitride layer is transformed into oxide [lo, 141. The results of the NRA analyses of ‘Ii-4SA1specimens oxidiLed at 800 “C might confirm this mechanism (Fig. 14), although the limited penetration depth of the NRA method does not allow a definite conclusion.
W..1.Qiiadakkers, i? Sclzaaf;N. Zheng, A . Gil arid E. Wallura
286 -20
,
Fig. 14. Nitrogen distribution in Ti45A1 after various oxidation times ofTi45Al at 800°C in air (NRA data).
Assuming this mechanism to occur it is obvious that the stability and the time for dissolution of the "continuous"Ti-rich nitride layer will dcpend on various paramctcrs such as alloy composition, growth rate and growth mechanism of thc scale. as well as on the phases present in the alloy depletion layer at a given time. Let us consider the two phases, which, apart from oxides and/or nitrides can be prcscnt in the depletion layer: a*- and Z-phase. According to thc findings in rcfercnce [7], t h e AI/Ti-activity ratio in thc Z-phase is expected to be higher than in az.As the formation of a near-continuous 'Ti-rich nitridc layer at the oxide/alloy interfacc will tend to increase thc Al/Ii-activity ratio, the presence of the Ti-rich nitride layer is expected to stabilize thc Z-phase at the cxpensc of a,, which secms to be in agrecment with thc TEM studies in reference [lo].As soon as the near-continuous nitride layer is dcstroycd, e. g. due to a change of thc pO,/pN,-ratio at the alloy/oxide interface, the AITTi-activity ratio will decrease leading to the formation of a,. Bccausc of thc high solubility of oxygcn in a,, this eventually causes thc aluminium to precipate as internal oxide rather than to be mcapoated in the inner part of the scalc and/or thc outer barrier layer. Consequcntly, dmsolution of the ncar-continuous nitridc laycr beneath the mixed scale causes the oxidation mechanism to change from oxidation stage I1 into stage 111 (Fig. 13) as observed for Ti48AI after around lOOh (Fig. 3c) and for'TiSOAl betwcen 24 and 65 hours (Figs. 6,7). In the Nb-containing alloys the near-continuous nitride layer was found under all conditions studied after air oxidation (Fig. lob, d).lhis can be explained by the better protective propertics of the oxide scale formed on thesc alloys, which rcsults in a lower pO,/pN,-ratio at the oxide/ alloy interface in case of the binary alloys. Conscquently the near-continuous nitride layer remains stable up to very long oxidation times. A shift from oxidation stage I1 into 111, which occurs f o r the Nb-containing alloys in a nitrogen-free environment (Fig. lOa, c), docs thcrefore not occur during air oxidation.
I 9 Beneficial arid Detrimental Lffects of Nitrogen on the Oxidution Behaviour
287
19.5 Conclusions Referring to Figure 13, the effect of nitrogen on the oxidation behaviour of binary and ternary, Nb-containing y-TiAl based alloys can be summarized in the following way: - In oxidation stage I, the presence of nitrogen always seems to be detrimental as it destroys the protective alumina-rich scale, perhaps due to oxidation of the initially formed Ti-rich nitrides. ‘fie enhanced oxidation induced by nitrogen is therefore more pronounced for alloys which possess a larger tendency to formation of an initial protective Al-rich scale. i. e. Ti48AI and TiSOAI whereas for Ti45A1, nitrogen has hardly any effect. - In oxidation stage 11, nitrogen is beneficial as long as a near-continuous nitride layer, which is probably in equilibrium with the Z-phase, is stable beneath the oxide layer. In this way the formation of a wide, a,-containing sub-surface depletion layer, accompanied by internal oxidation, destruction of the outer alumina barrier layer and high oxide growth rates (shift from stage I1 into stage 111) is prevented. - In Nb containing alloys this positive effect of nitrogen in oxidation stage I1 prevailed under all experimental conditions studied after longer exposure times, i. e. presence of nitrogen tends to decrease the oxidation rates. For the binary alloys, however, the stability of a near-continuous nitride layer seems to depend on alloy composition, oxidation time and probably temperature. Consequently, for the binary alloys nitrogen can in oxidation stage I1 be beneficial as well as detrimental.
19.6 References [11 A. Rahmel, RJ. Spencer: Oxid. Met.35 (1991) 53. [2] G.H. Meier, FS.Pettit: High Temperature Intermetallics.London, 30 April - 1 May 1991,The
Royal Society, p. 66. [3] N.S. Choudhury, 1I.C. tiraham, J.W Hinze: Properties of High Temperature Alloys, Z.A. Foroulis and F.S. Pettit, eds. ( l h e Electrochemical Society, 1976),p. 668. [4] S. Becker,A. Hahmel, M. Schorr, M . Schiitze: Oxid. Met.38 (1992) 425. [S] <;./I. Meier, FS. Pettit, S.Hu: High Temperature Corrosion (Les Embiez, France, 1992), Proc. J. Phys. IV 3 (1993) 395. [6] N. Zheng, WJ. Quadukkers,A. Gil, I f . Nickel Oxid. Met. 44 (199.5) 477. [7] N. Zhetig, W Fischer, I I . tiriibmeier, C: Shemet. WJ. Quadakkers: Script. Metall. et Mater. 33 (1995) 47. [8] WJ. Quadakken, N. Zheng, A. ( 2 1 , R. Wallurn, I I . Hoven: High Temp. Corro., Les Embiez, 20-24 May 1996, proceedings in print. [9] J.M. Rakowski, E Pettit, G Meier, F Dettenwanger, L. Schumann, M . Hiihle: Scripta Metall. et Mater.33 (1995) 997. 101 C. I,ang, M. Schiitze: I’rocecdings “Wcrkstoffwoche”Stuttgart, 1996,in print. 111 W. Ihwling, W Donfon:Scripta Met.27(1992) 1663. 121 I I . Nickel, N. %heng,A. Elschner, WJ. Quadakkers: Microchimica Acta 19Y (1995) 23. 131 A. Rahmel, WJ. Quadakkers, M . Schiitze: Materials and Corrosion 46 (1995) 271. 141 U. Figge, A. Elschner, N. Zheng, 11. Schuster, WJ. Qundakkers: Arbcitstagung “Angewandtc Oberflachenanalytik”, 22-25 June 1992, Julich, FRG, Proceedings in Rescnius J. Anal. Chem.346 (1993) 75.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
20 Influence of Moisture on the Oxidation of y-TiAl R. Kremer and FZ Auer
20.1 Introduction In industrial applications the environments usually contain more than one reactant. For cxample high temperaturc oxidation occurs in air by thc combined attack of oxygen, nitrogen and quite frequently water vapour. However, most of the studics concerning thc oxidation rcsistance are performcd in dry oxygcn or dry air. The oxidation behaviour of the intermctallic phases of the ‘Ti-Al system has recently received considcrable attention. The influence of water vapour on the oxidation of titanium aluminides has not been studied intensivcly. There are only a few studics of the high tcrnperature corrosion of titanium and its alloys. Liihherg and Schleicher examincd the behaviour of titanium in water vapour in the temperature range of 600°C to 1000°C [lJ.The oxidation in watcr vapour did not follow the cxpected parabolic time depcndcnce. A slightly increased oxidation rate was attributed cither to an incrcasc of oxygen vacancy concentration in rutil or to a change of the interfacc oxiddmetal due to solution of hydrogen in the metal phase. The first mechanism was also postulated by Moroishi and Shida for titanium in the tcmperature range of 450°C to 650°C [2]. Motre ct al. observed an accelerated oxidation attack of Ti and Ti-6A1-4V in moist atmosphcrcs [3].In the casc of Ti-6A1-4V they observed a two layercd rutil scale with an intermediatc alumina laycr instead of thc multilayered rutil scale obtained in oxygcn. Mofte et al. proposed that stress relaxation leads to recrystallisation of the outer part of the scale whereas in dry oxygen the scalc detaches, if a critical thickness is reachcd. The objcctive of this paper is to present thc results of an investigation on the influcncc of watcr vapour in the oxidation of y-TiAl at 900°C.
20.2 Experimental Procedure Ti-42 at.% A1 and Ti-50 at.% Al were obtaincd from Hauncr Metallischc Werkstoffe, Gcrmany. The materials werc used in the as-cast condition. Spccimens werc cut from the cast rod to approximately 16 x 5 X 1 mm. The surface was ground to 1200 S i c paper and cleaned ultrasonically with ethanol prior to oxidation.
290
K.Kteniet arid W Aurr
The thcrmogravimetric measurcmcnts were performed at 900°C in 0, or 07-€120gas mixtures. Saturation with water vapour was obtained by passing the gas through a washing bottle fillcd with deionized water. The water reservoir was held at 15°C. which corresponds t o a water vapour pressure of 19.3mbar. Lower water vapour pressures were established by saturating only a part of the reaction gas with water vapour, reduced oxygen pressurcs by adding helium. Prior to the experiments the reaction chambcr was flushed with the reaction pas for at lcast 40h. 'Ihc flow rate was 120 ml/min. The sample mass gain was rccorded continuously during cxposurc using a microbalance with an accuracy of I 15.g (Sartorius microbalance type 4410). All experiments werc startcd by raising the heatcd furnace to the specimen chamber. The mass gain during heating of thc spccimcn was 0.2 mg/cm?. After exposure the specimen was cooled to room temperature in thc reaction gas. Duplicate tcsts were carried out in cvcry case. Additionally experiments were pcrformcd in Ar-H20- and H1-Il,O-gas mixturcs. The oxygen pressurc was measured using an oxygen probc in the exhaust gas, consisting of a ZrO,/CaO-tube in contact with two platinum wircs. The surface of thc specimen was investigated by means of scanning clectron microscopy (SEM, Jeol 6400) in combination with energy-dispersive X-ray analysis (EDX, Tracor 2), X-ray diffraction (XKD, Siemens Kristalloflex) and Auger electron microscopy (AES, PHI 670). Typical AES working parameters were an electron beam voltage of l 0 k V with a current of 10nA.The lateral resolution is about 20-30nni. Augcr depth profiling was carried out with the following sputter pararncters: electron beam voltage: 5 kV, sputter gas: Ar, sputter time: 2 min and sputter area: 4 mm X 4 mm.
20.3 Experimental Results The oxidation of Ti-50 at.% Al is always faster in moist than in dry oxygen. Fig. 1 shows that in oxygen saturated with water vapour the oxidation rate steeply increases after a few hours while no such breakaway occurred in dry oxygen.'I'hc duration of the transient oxidation prior to the accclerated attack depends on thc water partial pressure pllZo.Saturation of oxygen with water at 19.3mbar pHZ0leads to faster oxidation after 5 h (Fig. 1). A p,170of 2 mbar accelerates oxidation after about 50 h (Fig. 2). More detailed information concerning the influcnce of water vapour on the reaction mcchanism was obtained by two-stage oxidation experiments. As can be seen in Fig. 3 a change of the atmosphere from moist to dry oxygen results in an immediate deceleration of the oxidation ratc.&hc deceleration indicates role of surface reaction control. As also shown in this fig reoxidation in dry oxygen inhibits subsequent rapid oxidation only for a few hours. A change from moist to dry oxygen shortly bcfore breakaway leads to a low oxidation rate as observed in dry oxygen. The rapid oxidation can be approximately described by a linear rate law. An influence of the flow rate was not observed. The effect of watcr vapour pressure pH2(]can be scen in Fig. 4 showing a decrease of thc oxidation rate with decreasing pl170.Thc oxidation attack is uniform in moist oxygen at pIIz02 4mbar. At lower water vapour pressures no uniform scales arc obtained. Hence a proper calculation of the rate constant is difficult.
20 It!pitetzcr of Moistrrre
I,
011
h e Oxidaiiori of y-TiAl
291
I
dry w
w n
P,
Fig. 1. Effcct of moisture (p(II,O)=
4
.
N
5
3
.e .-
2
B 00 m v)
-
1
1
Fig.2. Effcct of water vapor pressurc on the
0
Hm i s t oxygen
0
20
xo
60
40
time in h
0
10
20
30
time in h
40
50
60
Fig.3. Effect of moisture (p(H,O)= 19.3 mbar) on the oxidation of'Ii-50 at.%, Al at 900°C and p(0,) = 0.2 bar. Changc of the gas composition during the experiment.
Ag.4. Lffcct of water vapor pressurc on the oxidation kinetics of'fi-50 at.% AI at 900°C in moist oxygen.
R. Kremer nrrd W.h e r
292
For pllzo2 4 thc linear rate constant is approximately proportional l o pII . , l I ' 2 . As can bc sccn in Fig. 5, in moist oxygen the oxidation rate increases with decreasing o x y gcn prcssure po, proportional to p0,-"l. At low po, thc reaction rate begins to dcci-casc with time after extended lincar oxidation. 'I'hc duration of the linear stagc clecrcases with decreasing oxygcn prcssure. Experiments in Ar-H,O and I [?-I 1 2 0 gas mixtures enable measurements a t verv low values o f po,. In moist argon the oxygen pressure is 10 I' har, in moist hydrogen 10 '"bar. Undcr thcse conditions the oxidation kinetics can he approximated by par;ibolic ratc law without any initial transition (Fig. 6). I n thcsc cxpcriments oxidation is faster at higher po:,. In dry argon thc weight incrcasc was ncgligible. The oxygcn in dry argon results from thc dccomposition o f rcsidual water (po, varied bctwccn 10 Is bar to 10-I.I bar). The very thin oxide scale consists only of Al?O.,, as rcvealcd by AES depth profiles. The morphology of the scales grown in dry oxygen is typically complex i n the way that a thin alumina layer as well as oxide nodules cocxist on the surface (Fig. 7). A E S depth profilcs reveal that the thin alumina laycr also contains traces of TiOl. Thc cxternal growth of alumina rcsults in a n Al depletion in thc subsurface scale, which in addition is oxygen rich. In thc case of oxide nodules, the scale is multilaycrcd, with an outer TiO, layer followed by an intermcdiate alumina layer and a coarsc mixture of both, AI,O, and TiO, (Fig. 8). The intcrmediate alumina layer is oftcn referred to as barrier layer. AES depth profiles show additional peaks for 'I'i in this alumina laycr. The subsurfacc zone is two-phascd, Al-depleted and oxygcn enriched. The most marked differcnce between thc oxidation of TiAl in dry and in moist oxygen is the growth
0
time in h
I
0
40
30
20
10
10
20
w
time in h
50
'I0
M)
50
Fig.5. Oxidation kinetics of'ri-50 at.'%Al at 900°C in moist oxygcn, p(H,O)- 19.3mbar.
Fig.6. Effect of oxygen pressure on the oxidation kinetics ofTi-SO at.% Al at YOO"C, reaction gas saturated with H,O.
20 Inf7iretic.r ofMoi.vtitre oti the 0.ririutinti of y-7i’Al
293
Fig. 7. SIiM micrograph of the surface after oxidation of K-50 at.% Al at 900°C in dry oxygen with ~ ( 0 =~0.2 ) bar aftcr 100 h.
Fig.8. Micrograph of the oxide scale aftcr oxidation of%-SO at.% Al at 900°C in dry oxygcn with p(0,) - 1 bar aftcr 10Oh.
o f thick, continuous scales in moist oxygen, except if pH,()is less than 4 mbar. A s can bc seen in Fig. 9 the morphology of the continuous oxidc scalc obtained after exposure in moist oxygen resembles that of the oxide nodulcs. Aftcr exposure in moist oxygen A E S spectra of the surface show pcaks for ‘Ii and 0 and small characteristic LMM transitions for alumina, which vanish aftcr two sputter cycles, leaving pure ‘TiO,. Between the outer Ti02layer and the inner oxide mixture, alumina has prccipitated as separate and quitc large particles in the TiO,.This part of the scale also contains large pores. Beneath the external scale internal A1,0, has precipitated in the two-phased subsurface zone. X-ray diffraction pattern show some TiH, diffraction lines. In gcncral, scale adherence was good. Despite the thickness o f the oxide no spallation was observed.
294
H.Kretner and W Airer
"lie scalcs grown in H2-H,0 or Ar-H20 are different from the scales obtained in 0,-I I@. In Hz-H20and Ar-H,O blades and whiskers covcr the scale surface. Thcse blades and whiskers consists of TiO, with traccs of AI,O,. Next to the outer, comparably thick TiO, layer is an cxtrcmely thin intermediate Al,O,-layer (Fig. 10). 'Ibis alumina layer is followed by a very fine grained, Al,O,-rich oxide mixturc. Thc subsurfacc zone shows extensive internal oxidation of A120,. Continuous, inultilaycred scales are also obtaincd in dry oxygcn, if thc Al-contcnt is lowered to 42 at.%,. The morphology of thc scalcs corresponds to that of the oxide
Fig.9. SEM micrograph of the surface after oxidation ofx-50 at.% Al at 900°C in 0 , / 1 1 2 0 with p(0,) = 0.2 bar after 80 h.
Pig. 10. SEM micrograph of the oxidc scale after oxidation of Ti-50 at.% Al at 900°C in H2/11,0 with p(0,) - 10 2o bar after 20 h.
20 Infliience of Moisture on the Oxidntion of y-TiAl
295
n
E
PE .C
15
-
10
i
s
Ti-50 ar.% Al.
%
0 0
5
10
15
20
time in h
25
30
35
40
Fig. 11. Oxidation kinetics of titanium aluminide at 900°C and p ( 0 J = 0.2 bar.
nodules. As can bc seen in Fig. 11 the oxidation is slower than for 3 - 5 0 at.% Al in moist oxygen despite the much lower Al-concentration.
20.4 Discussion It is apparent from the data prescnted that moisture cnhanccs significantly the oxidation rate of y-TiALThc oxidation rates observed in moist oxygen are generally higher than in dry oxygcn duc to the formation of different scalc morphologies in both the media. In dry oxygen, an A170, scalc devclops containing additionally a small numbcr of mixed TiO,-Ai,O, oxide nodules. The corrosion rate, measured by therrnogravimetry, results from thc growth of the fast growing,TiO,-rich oxidc nodules and the slower growing AI,O, laycr. In moist oxygen, TiAI shows disastrous breakaway oxidation aftcr an initial protectivc oxidation pcriod. During brcakaway oxidation continuous, mixedTi0,-Al,O, scalcs arc formed. Likewisc thick, continuous scales are formcd during oxidation of 'Ti-42 at.% A1 in dry oxygcn. However, dcspitc thc lowcr Al-activity of the aluminidc, oxidation is slower than that of'ri-50 at.% A1 in moist oxygen.Thc morphology of the continuous oxide scales on Ti-42 at.% Al and of thc oxide nodulcs on T-50 at.% A1 formed in dry atmosphercs arc similar to that formcd in moist atmospheres. ' h e most markcd differcnce concerns thc alumina cnrichment between the outer KO,-layer and thc inncr oxide mixture. In dry oxygen the oxide scale cstablish an intcrmediate, compact AI,O, barrier laycr. In 0,-H,O-gas mixtures n o compact barricr laycr is formcd, instead separate Al,O,-particles are formed embeddcd in the outer 'IiO, layer. An A170, barrier layer would restrict thc transport of oxygcn into the oxide and the lowcr oxidation rates in dry oxygcn can possibly be associated to thc formation of that barrier layer. Additionally acccleration of the oxidation in moist atmosphere might be caused by doping of TiO, or AI,O, with hydrogcn. However, no clear evidence has been obtained for hydrogcn dissolved in 730, or AI,O, yct. Breakaway oxidation was also rcportcd by Becker ct al. after prolongcd exposure of XAI in dry oxygcn [4J. This breakaway was attributed to a change in the scale structure.l'he A1?0, barricr laycr was supposed to have no long-term stability. Aftcr a critical thickness is reached the alumina layer begins to dissolve and precipitates again as
separate particles in thc outer 'fiO,-laycr. As a consequence the oxidation is accelcrsted. Oxide scale and oxidation behaviour are similar to that observed in moist oxygen. Howevcr, in moist oxygen no thick. continuous scales with a n intermediate A1,O barrier laycr arc cstablished before breakaway. Before breakaway, thc oxidation ratc in moist oxygen is just as low as in dry oxygen.'Ihe reason for the loss ofthe protective behaviour in moist oxygen may be a changc in the subsurface scale. According t o Kakowski et al. cxposurc to oxygen with > lOppm 11:O enibrittles'Ti;Al-alloys more scverc than exposure in dry oxygen at 70O"C-90OcC[S].It is known t h a t titanium aluminidcs are susceptiblc to hydrogen embrittlcment 16-81. As indicated by the XRD-pattern hydrogen dissolvc in thc alloy. 'lhe breakdown of the protective scalc may be associatcd to the dissolution of hydrogen in the aluminidc. Ilowcver. oxidation was not accelerated if the gas was changed from moist to dry oxygen shortly before brcakaway. Additionally no transicnt oxidation is observed in H,-H,O-mixtures. Thcreforc the reason for breakaway in moist oxygen is not clear yet. The kinetics of thc breakaway oxidation can bc dcscribcd with a linear rate law. A linear relationship bctwecn mass gain and time suggcsts, that both. a phase boundary reaction or the transport of gaseous species arc rate detcrniining. Also a surfacc rcaction control is indicated by the fast decclcration after changing from moist to dry oxygen. Since the flow rate does not influence the oxidation ratc, the transport of gas molccules from the gas phasc to thc scale surfacc is not ratc controlling.Thc oxidation ratc increases with increasing pl,,o and decreasing po,. Considcring the rcactions of the surface, several steps are possible including adsorption and dccomposition of water and oxygen. I t may be assumed, that the decomposition of water on the surfacc is rate determining. The decomposition of water can only take place on reactivc sites, which arc free of adsorbed oxygcn.'I'hercfore the reaction rate is proportional to the activity of adsorbed oxygen aO(ad) and hence decreases with increasing pO,. If dccomposition of water is ratc dctcrmining, the rcaction ratc should be proportional to pk,20. Gencrally, further invcstigations on the mechanism initiating rapid breakaway oxidation in 0,-H,O appear to be nccessary.
20.5 Summary After initial protective oxidation of y-TiAl rapid breakaway oxidation follows in moist oxygen. In general, the oxidation is fastcr in moist than in dry oxygen. ? h e oxidation rate increases with increasing p,l,o and decreasing po2,which may indicate surface reaction control by 13, dissociation, retarded by adsorbed oxygen. After breakaway, continuous mixed ?'iO,-Al,O, scales are obtained in moist oxygen. Thc top laycr contains mostly TiO, and some discrete AI,O, particlcs, beneath this layer there is a fine graincd Al,O,-rich mixture of both oxides. An Al-depletion layer is formed bencatli the surfacc. In this subsurface zonc, internal oxidation of A1 to AI20, occurs. Thc scale adhercnce is good.
20 In.fliience of Moisrure oti the Oxidation of y-?’iAl
297
20.6 Acknowledgements ‘Ihe research dcscribed in this paper was supported by Dcutschc ~orschungs-C;emeinschaft. ‘I’he authors wish to thank Prof. IZ. Krresche and Prof. H.J. Grahke, MPI Dusseldorf for helpful discussions.
20.7 References K. LBhherg, 1l.W Schleicher: Z.Physik. Chem. 15 (1958) 223. 7: Morishi, Y Shida: Met.‘Tians. 9A (1978) 2773. I.: Morfe, C. Coddet, P Sarrazin, M.Azzopurdi, J. Bessoti: Oxid. Metals I0 (1976) 113. S Recker, A. Rahtnel, M . Schorr, M. Schiitze: Oxid. Metals 38 (1992) 425. J. Hakowski, D. Monceau, E S Pettit, G.H. Meier: Microscopy o f Oxidation 2 (1993) 476 D. I,egzditzu, I.M. Rohertsotr, K.11. Rirnbautn: J. Mater. Kes.6 (1991) 1230. W - YChu, A.W 7’honzpson:Scripta Mctall. Matcr.25 (1991) 2133. 81 C.?:Liii, Y - WKim:Scripta Mctall. MateI.27(1992) 599.
I] 21 31 41 51 61 71
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
21 Ion Implantation as a Tool to Study the Oxidation Behaviour of TiAl-Based Intermetallics M. E Stroosnijder, H.J. Schmutzler, KA. C. Haanappel, and J.D. Sunderkotter
21.1 Introduction TiAl-based intcrmctallic alloys have attracted considerablc intcrcst as structural materials for high-temperature applications such as in acrospacc industry due to their low dcnsity and substantial mcchanical strength at high tcmpcratures [l-51. I Iowcvcr, major drawbacks hindering industrial application arisc from lack of ductility at low tcniperatures and, particularly, from insufficient oxidation resistance at thc cnvisaged service ternperaturcs bcyond 700°C [l-81. A great deal of research work has been directed towards improving these properties c. g. by adding ternary and quaternary alloying clcmcnts and by developing suitablc thermomechanical treatments and forming tcchniques [l-111. Alloying TiAl with ternary and quatcrnary elements such as niobium, silicon, tantalum, and tungstcn has been rcported to improve thc oxidation rcsistance of the alloy significantly, but these additions can also impedc thc mechanical properties [ 1,6].A proper balance of mcchanical and oxidation propcrties has not been achicvcd yet. Additionally, publishcd corrosion data show significant scatter and could lcad to completely different conclusions [ 8 ] .Discrepancics in corrosion data obtaincd in different laboratorics arc possibly partially duc to differences in alloy purity since several alloying clcmcnts are known to altcr the corrosion propcrtics cven if they are prcsent in small quantities. I Iencc, thcre is need for a fast scrccning procedure of possible alloying elements with rcspect to their impact on the corrosion propcrtics of materials. The aim of thc present contribution is to emphasize that ion implantion can serve as such a rcscarch tool. Ion implantation offcrs the advantage of adding an elcmcnt to thc alloy in a nearsurfacc laycr in a wcll controlled and rcproducible manncr 112, 13].'I'he major drawbacks of ion implantation, howevcr, arc l ) the necessity of working under vacuum conditions and 2) relatively high cquipment and processing costs [12,13]. Thc impact of ion-implantation technology as an indispensable tool has alrcady been dcmonstrated in proccss dcvclopmcnt and manufacturing in semiconductor industrics; a comprehcnsivc rcview is given e. g. in [12].Ion implantation has been applicd on a morc basic research level t o othcr ficlds of materials scicnce and engineering such as tribology,
300
M. E Siroosnijder, 1I.J. Sclimiitzler, V A .C. Ilnnnnppel arid J. D.Simderkiitier
wear, and corrosion [12.13] (and references therein).To mention only one example on the field of high-temperature oxidation phenomena, ion implantation has successfully served as a research tool in investigations dealing with the elucidation of the so-called reactive element effcct [13] (and references therein). The present contribution focuses on the effcct of various elements added by ion implantation on the isothermal and cyclic oxidation bchaviour o f the y-Ti Al based intermetallic alloys Ti-48AI-2Cr and ‘fi-48A1-2Cr-2Nbat 800°C in air. These particular materials were selected since the ternary chromium addition improves the mechanical properties especially room temperature ductility [9-111 and the quaternary niobium addition improves the oxidation resistance [S, 14, 151. Comparison will be niadc bctween materials modified by ion implantation and alloys in which clcmcnts were added by alloying techniques.
21.2 Experimental Ti-48Al-20, Ti-48A1-2Cr-2Nb, and ‘Ti-47A1-2Cr-0.2Sisheet materials (the compositions are given in at.-%) were manufactured and provided by Plansee AG. The latter two materials were cmployed for comparative studies. Coupons (10 x 10 X 1 mm or 20 x 10 x 1mm) were cut from the sheets which were thermoniechanically processed in order to obtain a fine-grained near-gamma microstructure of the niaterial [9-11 ].The near-surface depletion zoncs formed during heat treatment of the specimens were mechanically removed. The spccimcns wcre subsequently ground to a 600grit surface-finish, ultrasonically cleaned in acetone and freon and transferred into a preparation chambcr for ion implantation. Both large faces of the specimens were implanted which accounts for 80 to 90% of thc total surface area. Ion implantation was undertaken at 180 to 200 keV acceleration potential and at a nominal dose of lo’’ ions/cm2.The elements implanted were Ti, AI, Cr, Mo, Y, Mn, Pt, Nb, and Si in casc of ‘li-48A1-2Cr and Nb in case of ’l’i-48A1-2Cr2Nb. The element distribution profile was determined experimentally by Rutherford backscattering spectroscopy (RRS) and, for instance, it was found for Nb implantation a maximum concentration of 20 at.-% at 60nm depth from the surface. IWS indicated the retained dosis to cqual the nominal one.’I’he Nb distribution profile was in good accordance with the prediction made by a depth-profile modelling program [l6]. ’lhc implantation depth was about 200nm. The implanted spccimcns were isothermally oxidized a t 800°C for 150 h in flowing, synthetic air having a dew point of 8:’C which corresponds to a water vapour partial pressure of 0.011 bar. The mass change of the spccimcn upon oxidation was continuously recorded using a commercial thermobalance (Cahn TG-l71).The specimens wcrc suspended from the microbalance and located in the temperature-constant zone of the vertical mounted furnacc.Thc mass change data were not corrected for the nonimplanted surface of the spccimcns. ‘Ihercfore,the mass change data werc not cvaluatcd in terms of parabolic rate constants for oxide growth. Cyclic oxidation tests were conducted at 800°C in static laboratory air for up to 1500 cycles. Each temperature cycle consisted of exposure for one hour at reaction
21 l o t i Irnpliintritioti ns (7 7boi 10 Study thi. Oxidation Llehaviour
30 1
tcmpcrature a nd twelve minutes at room tcmpcraturc which was sufficient to cool thc spccinicns bclow 50°C. Mass changes of thc specimcns were recorded after certain numbcrs of t hcrmal cyclcs. 'Ihc structurc, composition. and morphology of the oxidation products were invcstigatcd by X-ray diffraction (XKL)). cnergy-dispersive X-ray spcctroscopy (ELIS), and scanning electron microscopy (SEM).
21.3 Results and Discussion 21.3.1 Implantation Ti-48A1-2Cr with Ti,Al, and Cr Ions 'Ihc mass gain of non-implanted Ti-48AI-2Cr recorded during isothermal oxidation (Fig. 1) revcalcd a slow transient oxidation during the initial 10 hours of exposurc followed by a morc rapid oxidation. Reproducibilty was checked with other spcciinens of the same batch. The change in oxidation rate is caused by a change in thc oxidation mechanism from predominant. protective Al,O,-rich to non-protective T i 0 2 and mixed-oxidc growth. 'This is illustrated by a typical micrograph of a cross-section through thc scale (Fig. 2a) which revealed that a multi-layered oxide scale was formed. 'Ihc outcr laycr consistcd of large TiO, crystals followed towards the metal substrate by a more Al,O,-rich layer which was not continuous and showed some porosity. Fur-
10
10
u)O"C, synth. air
-
8 a -
Q V
a2
E 6
6
Y
a2
m
2 0
Non-implanted
. E
€
I
1017ions/cm2
8
ol C r m
4
0
m
m
3
4
I" 2
0
2
0
40
80
Time [h]
120
160
0 0
5
10
15
Time [h0.5]
Fig. 1. Course of mass change for non-implanted andTi-, Cr-, and Al-ion implanted (10'' ions/cm2) Ti-48A1-2Cr upon isothermal oxidation at 800Y: i n air: (a) mass gain versus time, (b) mass gain versus square root of time. T h c mass gain data for the ion-implanted material was not corrected for the non-implanted part of the specimen surface.
302
M . i:' Stroosnijclrr; II.J. Schmuizler, V A .C. Huannppel rind J.I).SirntltJrkorter
10 pm
10 prn
Fig.2. Morphology of the corrosion layer formed onT-48A1-7Cr upon isothermal oxidation for 150 h at 8 O O T in air: (a) backscattercd elcctron imagc of metallographic cross-scction. (b) top view of the oxide scale.
21 Ion Itnplrititntion ( i s n Tool to Stiidy the Osidation Hehrn~iorcr
303
thcr towards the metal, an Al- and Ti-containing mixed-oxidc laycr prcsumably consisting of a mixturc of A1,0, and Ti02 followed. Within this mixed oxidc laycr, pores were obscrved as well as macroscopic cracks which were presumably formcd during preparation of the cross-section.‘fie oxidc next to the mctal phasc adhered very well to thc substratc.‘fihe micrograph also givcs evidence of a sub-surface zonc within thc metal phase which was cnrichcd in ‘Ti and dcpletcd in Al. This zonc was presumably formed due to Al-consumption during the course of oxidation. Fig. 2b gives an impression of the morphology of rutile columns of the scale surfacc. The overall oxidation bchaviour o f and the scqucnce of thc oxide layers within the scale formed on X-48A1-2Cr arc generally in accordance with the observations made on y-l’iAl based matcrials as revicwed in ref. [8]. Implantation of l’i-48A1-2Cr with Ti ions did not alter the oxidation behaviour significantly (Fig. la); implantation with A1 ions, however, resulted in a markedly decreased mass gain upon oxidation (Fig. la). Implantation with Cr ions resulted in a slow transicnt oxidation but at later stages the oxidation rate increascd and became almost the same as for non-implanted or Ti-implanted matcrial as indicated by comparablc slopcs of the curves in the mass gain versus square root of timc diagram shown in Fig. l b 1171. Implantation with A1 ions increases the Al Concentration in a near-surface region within the base mctal, hcncc promoting thc formation of a protcctivc alumina layer. Howcver, thc increase of Al concentration near the surface is tcchnologically not dcsired for loss of ductility by the formation of brittle TiAl,-bascd compounds. The addition of chromium to the base metal could lcad to a decreased transient oxidation rate. It is known that chromium prescnt in thc alloy promotes the formation on alpha alumina on e.g. NiAl-Cr intermetallic compounds [7,18,19]. Chromia which is likely to form initially on the metal surface can serve as nuclci for alumina formation since both compounds have the same crystal structurc and are completely miscible [7, 18,191. For longer exposure times, the mass gain data in Figs. l a and b show that Cr addition by ion implantation obviously was not sufficient to lead to a n improved longterm oxidation bchaviour. Ti addition by ion implantation did not change the oxidation bchaviour of ‘l‘i-48A12Cr which further suggested that radiation damage which might have occurred in a ncar-surfacc laycr within thc alloy during ion implantation only had a minor cffect. This is in accordancc with previous obscrvations that radiation damage is rapidly healed at modcrate to high temperatures 1131 and does not have an effcct on the oxidation bchaviour.
21.3.2 Implantation of Ti-48A1-2Cr with Mo,Y, Mn, and Pt Ions The mass gain rccordcd for ‘Ti-48AI-2Cr spccimcns implanted with molybdenum, yttrium, inanganesc and platinum (Fig. 3a) reveals that thesc implanted clemcnts did only affect the initial, transient stage o f oxidation. For longer cxposurc timcs the observed oxidation rates bccamc comparably high as indicated by similar slopes of thc mass gain versus square root of time curvcs in Fig. 3b. Hence, adding Mo,Y, Mn, and Pt by ion implantation did not improvc thc long-term oxidation behaviour of Ti-48A1-2Cr.
M . E Stroosriijdrr. 1I.J. Schniutzkr. V A .C. Hmticippel and J.L). Siiriderkiitter
304 10
10
. F 6 a
-
"E
E .
0
0
E 6
u
al
a
I
al
a m
m
r
6
z
a
4
S
0
m
r
4
z
2
2 0
0
80
40
160
120
Time [h]
0
0
5
10
15
Time [h0.5]
Fig. 3. Course of mass change for non-implanted and Mo-, Y-, Mn-, and Pt-ion implanted ( I O l 7 ions/cm2)Ti-4SA1-2Cr upon isothermal oxidation at 800°C in air: (a) mass gain versus time, (b) inass gain versus square root of time. 'Ihe mass gain data for the ion-implanted material was not corrected for the non-implanted part of thc specimcn surface.
21.3.3 Implantation of Ti-48A1-2Cr with Si and Nb Ions Implantation of Ti-48A1-2Cr with N b and Si ions led to a markedly dccrcased mass gain and oxidation ratc upon isothermal oxidation (Figs. 4a and 4b). The mass change data presented in Fig. 4a arc not corrected for the oxidc growth on the non-implanted portion of the spccimen surface. Since the implanted surface amounts about 80 %, o f the total surface, thc mass gain pcr unit area duc to oxidc growth on thc iniplantcd surface should be even lower. Tablc 1 provides a summary of mass gain data upon isothermal oxidation of 150 h and subscqucnt cooling to room tcmperaturc of non-irnplanted versus Nb- and Si-implanted materials. I t should be noted that after cooling to room tcmperaturc a mass loss of about 22 % of the mass gained upon oxidation for
Table 1. Mass change of non-implanted and 10'' ions/cm2implanted'l'i-48AI-2Cr upon isother-
mal oxidation for 150h at 800°C in air and subscquent cooling to room temperature. .
Ion -
Nb Si ~~
.. .
-
~.
isothermal
Am/A (mg/cm7) after cooling
8.57 1.23 1.22
s.57 1.23 0.96
.~
E (Yo)
0 0 22
~.
I': relative mass change upon cooling from 800°C to room temperature.
~
~
21 loti Irnpbntntiori as t i Tool 10 Stri(ly the Oxidation Rehaviour
30.5
B O O T , synth. air 10” ions/cm2 Non-implanted
~
0
40
80
Time [h]
120
160
0
0
5
10
15
Time [hO.q
Fig.4. Course of mass change for non-implanted and Si- and Nb-ion implanted (10’’ ions/cm2) ‘li-48AI-2Cr upon isothermal oxidation at 800°C in air: (a) mass gain versus time, (b) mass gain versus square root of time. The mass gain data for the ion-implanted material was not corrected for the non-implanted part of the specimen surface.
150h was obtaincd for Si-implanted Ti-48AI-2Cr whereas no significant mass loss was observcd for non-implanted and Nb-implanted Ti-48A1-2Cr. ‘This indicates that the oxide layer formed on Si-implanted ‘l’i-48A1-2Cr was prone to large-scale spallation during cooling to room tcmpcrature. Examination of the specinicns by an optical microscope (Figs. 5 and 6a) confirmed this finding. The morphology of the oxide layer formed on Nb-implanted ‘l’i-48A1-2Cr is shown in Fig. 6a and in morc detail in Fig. 6b; thcsc figures reveal thc boundary 701163 betwccn the non-implanted part of the spccimcn surface (causcd by a clip of the sample holder which covered this part of thc surface during ion implantation) and the implanted one. In contrast to the non-implanted part of thc ‘Ii-48A1-2Cr spccimcn, large rutile columns were abscnt on thc oxide layer grown on the Nb-implanted part of the specimen surfacc. However, XRD analysis and cxarnination of a mctallographic cross scction of Nb-implanted Ti-48A1-2Cr rcvcaled a similar scquence of the oxide layers formed upon oxidation for 150h at 800°C as obscrvcd for non-implanted material. This comprises of an outcr titania layer, a thin intermediate alumina-rich layer, and an inner mixed alumina and titania layer. Only the growth of outer and inner titania-rich scales seemed to be retarded which led to a reduced oxidation rate (Fig. 4b). Surprisingly, niobium-bcaring compounds were not detected, ncithcr by XKD analyscs nor by clcctron microscopy, although the Nb concentration amountcd to maximum 20 at.-‘% at a depth of 60nm from thc surfacc.‘To date, thc detailed mechanism for the effcct of niobium on thc oxidation behaviour has not been elucidated (sec scction 3.4).
306
M . I:‘ Slroostiijder, 1I.J. Schmutilet; 1/:A. C. Iluunuppel and I D . SutzderkoiicJr
Fig.5. Morphology of the oxide laycr formed on the surface ot Si-implanted ( I O ” SIionsicni’) ‘li-48A1-2Crupon oxidation for 150 h at 800°C in air.
21.3.4 Comparison with Oxidation Behaviour of Conventionally Alloyed Ti-47A1-2Cr-0.2Si and Ti-48A1-2Cr-2Nb The mass gain of Ti-48AI-2Cr-2Nb and Ti-47A1-2Cr-0.2Si recorded during isothermal oxidation (Fig. 7a) rcvealed a significantly rcduced oxide growth ratc (Fig. 7b) compared to ‘Ti-48A1-2Crwithout any quaternary alloying additions. Thc mass gain curve lor Ti-47A1-2Cr-0.2Si showed discontinuities; similar features in thc oxidation behaviour were observed in control cxperimcnts with other Ti-47AI-2Cr-0.2Si specimens. As determincd by XRD and SEM/EDS, the gencral morphological scqucnce o f thc oxidc laycrs within the scales formcd on Ti-48A1-2Cr-2Nb and Ti-47A1-2Cr-0.2Si was not changed compared to Ti-48A1-2Cr; again the scale consisted of an outer titania-rich layer, an intermediate alumina-rich layer and an inner mixed alumina and titania layer. Only thc growth of the outer titania laycr and the inncr mixed alumina and titania laycr sccmed to bc rctarded, particularly for the Nb-containing matcrial. Inspcction of the surface of the oxide covering Ti-47AI-2Cr-0.2Si upon isothermal oxidation for 150 h at 800°C rcvealed that parts of the oxidc grown werc spalled off aftcr cooling to room tcmpcrature. Thc discontinuities in thc corresponding mass gain curve (Fig. 7a) further suggest that the oxide laycr may already bccn damaged by continuous spallation and acceleratcd re-growth during isothermal oxidation. Hcnce, qualitatively Nband Si-implantcd ’Ti-48A1-2Cr showcd a comparable isothcrmal oxidation behaviour to Nb- and Si-alloycd material.
21 Ion Irnplanrcriion as a Tool io Siic(lv ihe 0-ridaiion Uehnviour
307
Fig.6. Morphology of the oxide layer formed on the surface of Nb-ion implanted (10’’ Nb ions/cm’)‘ri-48AI-2Crupon oxidation for 150 h at 800°C in air: (a) overview,(b) detail of the boundary zone of non-implantcd and implanted part of the specinicn surface.
A great deal of research effort has been directed towards explaining the effect of niobium alloying on the oxidation behaviour of the titanium aluminides y-TiAl and aZ-l’i3AIin air. A comprehensive survey is given in [14). Among the possible mechanisms currently discussed arc: formation of a thin Ti-rich nitride layer at the scale/alloy interface which acts as a diffusion barrier,
308 10
1
800°C
Synthetic B O OAirF - l
/
/
8
r7
E
e
E 6
Y
C
dm
I
u)
/
4
Ti-47AI-ZCr-0.2Si
2
2 TMA1-2Cr-2Nb
0
4
0 0
50
loo
Time [h]
0
150
b)
5
10
15
Time [h'.']
Fig. 7. Course of mass change for 'li-48A1-2Cr,l'i-47AI-2Cr-0.2Si, and Ti-48AC2Cr-2Nb upon isothermal oxidation at X00"C in air: (a) mass gain versus time, (b) mass gain versus square root of time.
increase of A1 activity relative to Ti in the alloy, thereby favouring the formation of an alumina-rich scale, doping of the titania latticc with Nb ions, thercby decreasing the number of oxygen vacancies and/or titanium interstitials, formation of Nb,O, which forms mixed oxides with TiO, and AI,O,, thereby acting as a link between otherwise nearly immiscible oxides and enabling the formation of a more coherent scalc with blockage of fast diffusion paths, decrease of oxygen solubility in the alloy. It was cmphasizcd in [14] that the above-mentioned mechanisms wcrc based on observations which have becn made in studies covering a wide variety of experimental conditions, e. g., alloy composition, tempcrature and corrosive cnvironmcnts which impedes a detailed analysis of the published literature in terms of the prcdominant mechanism. Most likcly not nccessarily one particular mechanism is rcsponsiblc for the effcct of niobium, rather a combination of various mechanisms would be expccted to bc active [14]. Additional analyscs such as cross-scctional TGM 1201 of the corros e a y e r and corrosion layer / metal interface region are needcd for further investions and to pinpoint the action of niobium. Another point which necds to be considered is the long-tcrm stability of thc oxide laycr under thcrmal cycling conditions. The results of cyclic oxidation tests (Fig. 8) rcvealed a dramatic mass loss of the Si-bearing material and, thus, further confirmed the earlier obscrvation of poor adhcrence of the oxidc layer on X-47A1-2Cr-0.2Si.?'he spallation resistancc of Ti-48A1-2Cr appeared to be better compared to the scalc on Ti-47A1-2Cr-0.2Si,although no apprcciable mass gain was detected after about 200 cycles indicating that mass loss due to spallation equals mass gain due to oxidc growth and rc-hcaling of thc oxide layer. Thcre was no clear indication of mass loss for Ti-48A1-2Cr-2Nb.
21 lor1 Irrrplnntatioii ns ( r Tool to Stridy the Oxidation Behaviour 15
309
I
10
-5
-10 0
500
lo00
Number of Cycles
1500
Fig. 8. Cyclic oxidation bchaviour of Ti-48AI-2Cr, Ti-47AI-2Cr-0.2Si, and Ti-48AI-2Cr-2Nb in air. Each temperature cycle consistcd of cxposure for 111at 800°C in air with a hold time of 12 min at room tcmpcraturc.
Nb-implantcd 'ri-48A1-2Cr has been exposed to cyclic oxidation for up to 600 temperature cycles [14]. Inspection of the oxide layer grown on the spccimens revealed that spallation occurred cxclusively on the non-implanted part of thc spccimcn surface (Fig. 9) indicating that Nb-ion implanted Ti-48A1-2Cr possessed a substantial rcsistancc against spallation upon thermal cycling.
Fig.9. Morphology of thc oxidc laycr formcd on the surface of Nb-ion implanted Nb ionsicn?) 'li-48A1-2Crupon cyclic oxidation for 600 one hour cycles at 800°C in air (1 cycle: 80O:'Cil h, RTi12 min).
M. ?I Stroosriijder; I I J Schtnutzler, FA. C. Ilaciricippel and .I. D. Sunderktmer
310
21.3.5 N b-Implanted and Non-Implanted ‘Ii-48AI-2Cr-2Nb Adding N b by ion implantation to already Nb-alloyed Ti-48Al-2Cr resulted in a further decreased mass gain upon isothermal oxidation (Fig. 10a), however, the longtcrm oxidation rate did not change (Fig. 10b). Addition of N b by ion implantation improved the short-term oxidation behaviour of Ti-48A1-2Cr-2Nb but this additional niobium amount might bc consumed during the course of oxidation so that the longterm oxidation behaviour was not affected. N b addition by ion implantation led to an improvcd short-term oxidation hehaviour of Ti-48A1-2Cr-2Nb which is in accordance with a rcccnt study [21] in which an increased Nb content of y-XAI based intermetallics has been found to slightly improve the oxidation resistance.
21.4 Conclusions and Outlook The high-temperaturc oxidation behaviour of y-TiAl bascd compounds implanted with various elements was studied and compared with matcrials in which some of these elemcnts were added by alloying.‘Ihe results presented have shown that
8wc
10
10
Synthetic Air
8
Ti-48ACZCr
-
z.
N -
E
,o
Z 6
E 6
Y
m C
I
Ti-48AI-ZCr
Y
8
5
0
a,
0 m C
5 4
4
m
f
2 2
0
2
Ti-48AI-ZCr-ZNb Ti-48AI-ZCr-ZNb + Nb 0
40
80
Time [h]
120
Tl-48AI-ZCr-ZNb
0 160
0
5
10
15
Time [ho.S]
Fig. 10. Coursc of mass change for non-implanted Ti-48A1-2C:r and Ti-48Al-ZCr-ZNb and Nb-ion implanted (10’’ Nb ions/cm2)Ti-48AI-2Cr-2Nb upon isothermal oxidation at 800°C in air: (a) mass gain versus time. (b) mass gain versus square root of time. The mass gain data for the ion-implanted material was not corrected for thc non-implanted part of the specimen surfacc.
21 Ion Implaiitatioii as n Tool t o Stirdy the Oxidation Hehaviow
311
ion implantation can serve as a rescarch tool to study the influence of various elements on the corrosion behaviour of ‘I’iAl-bascd alloys, ion implantation can be employed in thc frame of ‘screcning’ tests for evaluating the effect o f possible alloying additions, ion implantation could further servc to obtain a better undcrstanding of underlying corrosion mechanisms.
21.5 Acknowledgements The authors would likc to thank Dr. H. Clernens, Plansee AG. Reutte, Austria, for providing TiAl-bascd shect material and for many stimulating discussions. They are grateful to the staff of the Institute for Advanced Materials, Ispra, Italy, for technical assistance, in particular Mcssrs. A. Fusari and fl. Willers for ion implantation, Messrs. G Mncchi and G. Pisoni for oxidation experiments, Dr. ?I Dos Suntos for electron microscopy, and Dr. U Wufjen,JRC-Geel, Belgium, for Rutherford backscattering spectroscopy measurements.
2 1.6 References 111 K. Dariola, J.J. Lewaizdowski, C.7:Liu, P L . Martin, D.B. Miracle, M.V Nathal (cd tural Intermctallics” The Minerals, Metals & Materials Society, Warrendale, Pennsylvania, USA, 1993. 121 J.A. Ilorton, I. Raker, S. Haizada, R. I). Noebe, D.S. Schwartz (eds.): “High-Temperature Ordered Intcrmetallic Alloys 1V” MRS Symposium Proceedings Volume 364, Materials Kesearch Society, Pittsburgh, Pcnnsylvania, LISA, 1995. [3] C.M.Austiiz,TJ. Kelly: in ref. [l],pp.143. [4] Y - WKim: Journal of Metals 42 (7) (1989) 24. [ 5 ] Y - WKim:Journal of Metals 46 (7) (1994) 30. [6] S.-C. Huang: in ref. 111,pp. 299. [7] G H . Meiei; N Rirks, ES. Pettit, R.A. Perkins, H.J. Grubke:in rcf. [l], pp. 861. [SI A. Rahmel, WJ. Quadukkers, M.Schiirze: Materials and Corrosion 46 (1995) 271. [9] I l . Clemens, I? Schretter, K. Wurzwallner,A. Rartels, C Koeppe. in ref. [l],pp.205. [lo] II. Clemens, J. Rumherg, I? Schretter, S. Schwantes: Intermctallics 2 (1994) 179. 1111 I I . Clemens, I? Schretter, W Glatz:Prakt. Metallogr.33 (1996) 1. 1121 M. Iwaki: CRC Critical Reviews in Solid Statc and Materials Scienccs 15 (5) (1989) 473. [13] M . E Stroosizijder, K.Mivrel, M.J. Bennett: Materials at High’kmperatures 12 (I)(1994) 53. [14] M.E Stroosnijder, N. Zheng, WJ. Quadakkers, R. Hofmann, A. Gil, I:‘ Lanza: Oxid. Met. 46 (1996) 19. [lSJ 1I.J. Sdimutzlei; N. Zheng, WJ. Quadakkers, M.E Stroosnijder: Surfacc & Coatingsl’echnology 8.1 (1996) 212. [16] PROFIIJ?, version 3.18, Implant Sciences Corporation, 1991. [17J 11. Pieruggi: Oxid.Mct.27(1987) 177. [ 181 M. W Rriimm, H.J. Grabke: Corros. Sci. 3.7 (1992) 1677. 1191 M. W. Rrumrn, H.J. Grahke: Corros. Sci. 34 (1993) 547. [20] K. Wheeler, I I J . Schmutzler, I1.L. Frasei; M.E Stroosnijder: to be published. [21] II. Nickel, N Zheng,A . Elschner, WJ. Quadakkers: Microchimica Acta, I Y Y (19%) 23.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
22 Protection of Titanium Aluminides by FeCrAlY Coatings M.J. Bennett and S.J. Bull
22.1 Introduction Titanium aluminide intermctallics have attracted increasing technological interest in recent ycars on account of dcsirablc properties, such as high strength to mass ratios, low density ctc. Onc of thc principal aspccts of concern for industrial cxploitation, however, is inadequate oxidation resistance at envisaged servicc tempcraturcs of 700-900°C. The basic reason is thc inability for a protective,slow growing,AI,O, scalc to form and not be reduced by the TiAl alloy to a less protective faster growing Ti oxide. As a consequence substantial rcsearch activity has bccn focused on understanding the oxidation behaviour of thesc intermctallics (as rcviewed recently, e.g. references 1 and 2) and on mcans of improvement, particularly by facilitating the formation and long term stability of a protcctivc Al,O, scale. Several approaches have bccn pursued to achieve these objectives, including ternary and in some instances quaternary alloying (c.g. references 2 and 3), increasing the surfacc aluminium activity by aluminising (e.g. rcferences 4 and 5) and preforming an alumina scalc by controllcd preoxidation (c.g. refcrenccs 6 and 8).Thcse have proved beneficial to varying extents but in no ease were completely successful. Another approach would be to overcoat the titanium-aluminide intermetallic with an alumina forming coating, such as FeCrAlY and the influence of such a coating on the oxidation in air of’Ii,Al at 800°C and ofTiAl at 900 and 1000°C is reported in this paper.
22.2 Experimental 22.2.1 Titanium Aluminides Titanium aluminide alloys with two compositions, and differing structures, were uscd as substrates. The a-2 X,Al was a commercial alloy (Heat T8991), fabricated as 1mm thick sheet, by the Titanium Mctals Corporation of Amcrica and contained (as ”/,) 14.2 Al, 19.4 Nb, 3.2 V, 2.0 Mo, 0.10 Fe, 0.13 N, 0.08 0 and 0.023 C. The y-TiAl alloy, containing 48% Al, was produced as, buttons, by melting at AEA Technology, Iiarwcll “,I. Spccimcns, 20X10X1 mm, wcre cut from both matcrials and ground to an acceptable finish using 600 grade S i c paper.
314
M.J. Bennett ond S.J. Riill
22.2.2 Coating Procedures Aftcr outgassing at 100-200°C and ion cleaning, sputter ion plating 1101 from a Fe t lS'%Cr + 5%A1 + 0.3%Y source was undcrtaken using a 50 V AC substratc bias. Thc DC voltage applied to the source plates was changcd in the sequcnce 2 kV for 2 h, 3 kV for 2 h and finally 2 kV for 10 h.Thc substratc tempcrature incrcased to 500-700"C in this process to facilitate coating bonding by interdiffusion with the substrate. Thc rcsulting coatings wcre betwcen 45 and 135 pm thick depending on position i n thc coating chamber and containcd marginally lcss A1 and Y than thc feedstock alloy. In the as coatcd condition the FeCrAIY coating had a zonc 1 microstructure [ I I]. consisting of columnar grains with opcn boundaries. 'Ii, cxaminc the influcnce of microstructural changes the coatings wcre peencd using glass beads transportcd in air at pressures of cither 2.5 or 50 psi (173 and 364 kPa rcspectivcly).Thc critical cffects of pccning were first t o dcnsify thc coatings by squashing up the columns thercby closing thc gaps betwcen thcm and sccond, to introduce mechanical bonding bctween the columns of thc coating and to the substrate. Increasing thc air prcssure incrcased both the extent and depth of damage. With 25 psi prcssure the consolidation depth was -25 pm, while with the higher air pressure the coating was dcnsificd throughout its entirc depth. TiN and AlN barrier layers wcre also depositcd by sputtcr ion plating from titanium and aluminium sourcc plates. After outgassing at 100-200°C the samples were ion clcaned prior to coating (-500V, 0.025 m A ~ m -30 ~ ,min). A 150nm metal interlayer was dcpositcd by sputtering in pure argon to promote adhesion prior to admittance of nitrogen. For theI'iN coating a main power of 0.8 kW was uscd with a substratc bias of -60 V, to producc a dense coating -4 p m thick on the titanium aluminidc samples. The sample temperature was 600°C during coating. In the case of AIN a lower main power was uscd (0.5 kW) to prevent melting of the source plates. A 2-3 km AIN coating was dcpositcd at -60V bias in thc same time as the thicker TiN layer was produccd. Thc tempcrature during AIN coating ncver excceded 500°C.
22.2.3 Oxidation Conditions and Assessment of Oxidation Serics of as coatcd, coatcd/peencd, as well as uncoatcd, specimcns were oxidised in flowing laboratory air at threc tempcratures. 'fie cffectivcness of FeCrAlY coatings on the oxidation behaviour of ?'i,Al was examined at 800°C. while the corresponding influence on TiAl oxidation was investigatcd at both 900°C and 3 000°C. All thc specimens werc contained in alumina crucibles and wcre cycled to room tcniperature after each succcssivc oxidation period. Exposures were continucd for cumulativc timcs of l0OOh at 8oc)"C, 775 h at 900°C and 75 h at 1000°C.Aftcr defined times the extent of the oxygen uptakcs, and whcre appropriate, also of spallation were dctcrmincd gravimetrically. & After oxidhtion the chcmical and physical natures of the reaction products resulting from oxidation and also from interactions bctwecn the coating and thc substratc alloys were cxamincd by a range of surfacc analytical techniqucs. Thcsc includcd scanning electron microscopy (SEM), energy dispersive X-ray analysis (EDAX), electron
22 I'rorection
of
Titanium Aluniinides by I K r A l Y Coriting.7
315
probc microanalysis (EPMA), X-ray diffraction (XKD) and secondary ion inass spectrometry (SIMS). Transverse sections wcrc prcparcd by standard metallographic polishing techniques with one addition involving as a first step, coating the surface with a gold layer by sputtcring. This enablcd identification of t h e outer gas interface during subsequent analysis. An elemcnt which could havc bccn involved in the various chemical reaction sequcnccs was nitrogen and so was invcstigatcd. IJnfortunatcly sincc thc substratc alloys all contained titanium, N analysis by EPMA presented problcms duc to ovcrlap of thc 'I'i and N peaks. In certain circumstances those can be resolved and whcn prcscntcd in section 3.2 the genuineness of specific results arc cornrncnted upon.
22.3 Results 22.3.1 Gravimetric Assessment of Coating Effectiveness The cffcctiveness of the FcCrAIY coatings upoii the oxidation behaviour of Ti,AI at 800°C and of TiAl at 900°C and 1000°C was evaluated by gravirnetric measurements, which are shown in Figures 1-3 respectively.The attack ofTi,Al at 800°C was characterised by a decreasing oxidation rate with incrcasing tirnc (Figure 1) and ovcr the 1000h cxposurc no scalc spallation was apparcnt. Thc corresponding oxygcn uptakc by as coatcd %,A1 was significantly lowcr than on X3A1 with the coating reducing the attack by a factor of eight after 1000h. Oxidation resistance was enhanced further when the coating was peened using the lower (25 psi) air transport pressure. More extensive peening, using 50 psi air pressure, proved detrirncntal as the oxygen uptakcs
1
LOO
-
L
600
TIME [HOURS)
l
1
800
-
1
-
1000
Fig. 1. Improvcmcnt by a FeCrAlY coating of the cyclic oxidation behaviour at Ti,AI in air at 800°C
316
M.J. Bennett orid X J . Bull
were considerably higher than on the as coated and the 25 psi peened E'cCrAIY coated Ti& Also, after extended exposures they began to approach those o f uncoated Ti& indicating a gradual deterioration in coating performance. The oxidation behaviour of'I'iAl at 900°C and 1000°C (Figs. 2 and 3) was characteriscd by increasing attack with increasing temperature, as expected by protective type kinetics (rate decreasing with increasing time) initially but then after 50h at 900°C and 30 h at 1000"C, coincident with the onset of substantial scalc spallation, by an increasing oxidation rate. 'I'hc kinetics of cumulative spallation were similar to those of oxidation. As at 800°C the extent of oxidation of FeCrAlY coated TiAl at both
0
Fig.2. Improvement by a FcCrAlY coating of the cyclic oxidation behaviour of'l'iAl in air at 900"C -
v-
E
U
&LO-€ z
3 30-
(3
;
20-
I : -
10 -
20
:
-
I
/
-
LO TIME ( H O U R S )
I
60
1
A ~-
I"ig.3. Irnproverncnt by a 12ccCrA1Ycoating of the cyclic oxidation behaviour ofTiAl in air at 1000°C
'I'able 1. XliD Analyses of uncoated and IxCrAIY coatcdTi,Al 'ilter 10Ot) h oxidation in air at 800°C Spccinien
Phases present
X3AI FcCrAIY coated
'l'iOz (Rutile) [aA120,] a-Al,O, . T i 0 2 (Kutilc) . [Cr,O,] [O-Al20,J[y-Al,O;] cu-Al?O,.Fe,03. [B-A120,]
FcCrAIY coated pccncd. 25 psi FcCr AlY coated pccncd, 50 psi
a-Al?O; .'Ti02 (Rutile) . [?iO?(Anatase)l [Cr,O,] [Fe,03J[AI(OH), Baycrite] ~-
-
-
temperatures were significantly lower, e.g. by a factor of 10 after 75 h at 1000"C, and were reduced further by peening with the lower air pressure. The cffcct of more substantial peening using the higher air pressure was not examined at either temperature. The attack of both as coated and peencdkoated TiAl increased with temperature, while only aftcr 75 h exposure at 1000°Cwas any spallation apparent.
22.3.2 Scale Characterisation XKD analyses of uncoated and F~CI-AIY coated (with and without peening) X3AI after 1000 h oxidation in air at 800°C are summarised in Table 1. The phascs identified are listed in order of prominence and those within brackets were present at minor concentrations. Only an outer layer of the scale of Ti,AI or of the three coatings were examined, as thc X-ray penetration depth was of the order of 5 pm.'fie main constituent of the li3Al scale was 'l'i20 (rutile) but it also contained a-Al,O,. In contrast the principal oxidation product on the three FeCrAlY coated specimens was a-Al,O,. Ti,O (rutile) was apparent on both the uncoated and 50 psi pressure peened but not on the 25 psi peened FeCrAlY coated surface. Additionally the X-ray traces of all three coatings indicated the presence of other minor constituents, including transition (0, y) Al,O,, Cr,O,, Fe,O, and AI(OH),. A scanning electron micrograph of a cross-section of the scale formed on 'Ii3Al is shown in Figure 4. On the basis sf associated EDAX analyses and the XKD results (Table l ) , the scale comprised of alternate layers of TiO, (light, thicker) and a-Al,O; (darker, thinner), with some contained voids. Below the scale a defined Ti,AI zone was enriched in oxygen. Electron and associated Fe, Cr, A1 and Ti X-ray images of a transverse section through a FeCrAlY coated Ti,AI after 1000h oxidation in air at 800"C, are shown in Figure 5. An a-Al,O, layer formed at the surface, while both A1 and Ti were oxidised within the open columnar microstructure of the FeCrAlY coating from the gas interface. Ti was enriched also as a band within the coating near the interface with alloy substrate. EPMA was unable to resolve categorically whether the Ti was associated with nitrogen or present in a metallic state. Reactions had occurred at the Ti,AIFeCrAlY coating interface. which resulted in diffusion of l i and to a lesser extent possibly of A1 from the substrate alloy into the coating, as evidenced by denuded zones in
318
M.J. Retinett rind S J . Rid1
Fig. 4. Scanning electron niicrograph o f a transverse CI-ossseetion o f thc attack of.I'i,AI after 1000 h oxidation in air at X 0 0 " C
Fig.5. Electron image (A) and Fe (B), Cr (C),AI (D) a n d 3 (E) X-ray images of a transvcrse cross-section through FcCrAIY coated'I'i,Al after 1000h oxidation in air at 800°C
22 I'rorectiotr of 7i'tutiiumAluminides b!l FeCrAIY ('onrings
31 9
the'Ii,AI.'Ibis was accompanied by a counter diffusion of Fe and <7r from the coating into'I'i,Al, as evidcnccd by Fc and Cr cnriched zones at the substrate alloy surface. Associated with thcsc reactions voidage developed at thc coating-alloy intcrface. Peening with thc lower air transport air prcssurc had oblitcrated or sealed most of the columnar boundary nctwork, at least through the outer 60-70% of the FeCrAlY coating but this was still apparent in the inncr rcgion.This was evidenced by the climination of internal coating oxidation (Figure 6a) and the absence of TiO, in the reaction products (Table 1). Othcr than this the same rcactions occurred as notcd already on a FeCrAlY coatcd Ti,AI, namely the formation of an a-Al,O, surfacc oxide layer and interdiffusion at the substrate-coating intcrfacc. 'The extents of the latter were similar whether or not the FcCrAlY coating was peened with 25 psi air pressure. Thc resulting distinct Fe and'I'i rich layers arc shown in greater detail by the electron and associated Fe and Ti X-ray images in Figurcs 6b, c and d. Voidage was apparent both abovc thc Ti rich and bclow the Fc rich laycrs at the coatingPli,Al interfacc. The main effects of enhanced pccning with the highcr air pressure wcre inore extensivc intcrfacial reactions leading to a 50 pm thick Ti rich honeycomb layer, which also
(A) Electron image
(13) Electron image
(C) Fc X-ray image of (B)
(I))'Ii X-ray image of (B)
Fig.6. Electron and X-ray image of a transverse cross-seetion through 25 psi peened FeCrAlY coated'Ii,Al after lOOOh oxidation in air at 800°C
320
M.J. Retinett and S.J. Rull
containcd Fc, Cr and Al and intcrnal oxidation within the E'cCrAIY coating involving Cr,O, particlc formation (Rgurc 7 ) . Additionally EPMA suggestcd thc possibility of bands of AIN and 'GN being formed both within the coating and the substrate alloy o n either side of thc honeycomb laycr. The nct rcsult was that protcction afforded by the coating gradually detcrioratcd (Figurc 1) with cventual coating dctachmcnt. Most of the scale formed on y'l'iA1 during oxidation in air for 775 h at 900°C and 75 h at 1000°Cspalled (Figures 2 and 3).Thc intact scales remaining at the completion of thcse cxposures (e.g. after 75 h at 1000"C,Figurc 8) compriscd of a mixture of Ti and Al rich oxides, consistent with prcvious more detailed obscrvations [l. 21. On FeCrAlY coated and coated-pecned y'GA1 alloys similar reactions had occurred at 900°C and 1000°C.as reportcd abovc for FeCrAlY coated a-2'I'i3Al at the lowcr temperaturc (800°C). A continuous a-Al,O, oxide layer formed at the surfacc (Figurcs 9a and d), whilc pecning with 25 psi transport air pressure eliminated any intcrnal l'i oxidation apparcnt within the as fabricated FeCrAIY coating (c.g. after 75 h oxidation at 900°C (Figurc 10)). lnterdiffusion again occurrcd at thc coating-alloy intcrfaces involving the same clements as at the lowcr ternpcrature (Figurcs 9 and 10): Fe and Cr diffuscd from the coating intoTiAl, while therc was countcrflow of'l'i and possibly also of Al, to a lesscr extent, in the revcrse direction.Thc ratcs and, as a conscquence, cxtcnt of thesc rcactions incrcased with tcmperature (Figures 10 and 9). 'Ihe Fe rich laycr
Fig.7. Electron imagc (A) and Fe (B), and Cr (D) a n d 5 (E) X-ray images of a transverse crosssection through 50 psi pcencd FcClAlY coated Ti,AI after 1000h oxidation in air 800°C
22 Prorecrion o,f Tiiariium Aluminides by FeCrAlY Courings
32 1
Pig. 8. Scanning electron micrograph of a transverse cross-section of the residual scale formed on TiAl after 75 h oxidation in air at 1000°C
formed within y-TiAl proved to be brittle (Figure lo), while these coating-substrate interactions again resulted in extreme void formation, either at or near this interface. Although initially the FcCrAlY coating afforded remarkable oxidation protection to the y-XAl and again peening proved beneficial, eventually its cffectiveness broke down at times decreasing with increasing temperature. The cause was coating decohesion arising from the thermal strains on cycling from temperature to room temperature and emanating from the mismatsch of the thermal expansion coefficients of thc different layers at the coating substrate interfaces. The propensity for coating failure was exacerbated by the brittleness of the Fc rich layer formed in T A l , (Figure 10) the defective nature of the interface arising from voidage (Figures 5 , 6 , 7 , 9 and 10); and through coating cracking leading to air ingrcss to and subsequent rapid oxidation of the substrate alloy (Figure 9).
22.4 Discussion The basic concepts underlying the choice of an FeCrAlY coating as a means of providing oxidation protection to the titanium aluminide intermetallics, Ti,AI and TiAl at 800-1000°C were substantiated by the observed results (Figures 1-3). [Jncoated alloys underwent significant oxidation, resulting in the formation of scales consisting of both
322
M.J. Benrier~onrl S.I. Ritll
Rg.9. Electron image (A) and Fe (U), Cr (C),AI (D) and Ti (E) X-ray images of a transverse cross-section through FeCrAlY coated'IiAl after 75 h oxidation in air at 1000°C
Ti and Al rich oxidcs, which on TiAl at 900 and 1000°C spalled cxtcnsively and also caused oxygen embrittlcment of sub-scale regions within thc alloys (Figures 4 and 8). Oxidation of the alloys was prevented completely by the coating, and more specifically by the formation of u-AI,O, protective surface oxide layers (Figures 5-7.9 and 10). Protection was afforded to both intermetallics for extended periods of up to lOO0h duration at temperatures, 800°C for Ti,AI, 900°C and 1000°C for TiAI, well in exccss of those envisaged for the current technological use of both matcrials. It is highly prohable, therefore, that even longer periods of protection would ensue at temperatures below those of the current tests. For both substrates coating dcnsification by peening proved to be beneficial. This closed or at best, eliminated the gaps between thc columns making up the as fabricated coating and as a conscqucnce prcvcnted its intcrnal oxidation (Figures 5 and 6). However the peening proccss needcd careful optiniisation. Too severe mechanical deformation, extending beyond the coating into the substrate alloy, proved detrimental (Figure 7). The oxidation tests in air were cyclic and eventually after cumulative exposurcs of loo0 h at 800°C for Ti,AI, 775 h at 900°C for TiAl and 75 h at 1000°C again for TiAl visual deterioration (bowing, cracking, or spallation) was apparent in the condition of most coatings, which either did or subsequently would have undcrmined thc protection they affordcd. In somc instances breakdown was observed directly gravimetrical-
22 Protection of'Tilnniirt?zAlrrmiriides by FeCrAlY Contifigs
323
Fig. 10. Eletron image (A) and Fe (R), Cr (C),Al (D) and'ri (E) X-ray images of a transverse cross-section through 25 psi peened FcCrAlY eoatedl'iAl after 775 h oxidation in air at 900°C
ly. Thc causc was not chcmical failurc of thc FeCrAIY coating, by enhanced Al consumption leading to non-protective attack. Coating failure was always essentially mechanical with decohesion arising from strains induced on thermal cycling and emanating from mismatch of the thcrmal expansion cocfficients of thc various intcrnal layers adjacent to thc coating-substratc boundary. Many of these arose from chcmical interactions at this intcrface (Figures 5-7,9 and lo), which involved diffusion of iron and chromium from the coating into the alloy substrate, together with titanium, and possibly also aluminium, counter movcment in the reverse direction, i.e. from substrate to coating. Thesc internal-intcrfacial chemical reactions were accompanied by substantial void dcvelopment at or near this interfacc. This voidage, togcther with thc brittlcness of thc Fe rich layer formed in %A], would have facilitated decohcsion, while mechanical failurc was exacerbated by through-thickness coating cracking leading to air ingress to thc underlying 13 aluminidc intermctallic and its subscquent rapid oxidation. Whilc this study was in progress, related investigations wcre being pursued in the US [12] and in China [13] which substantiated the main conclusions. Namely, that MCrAI(Y) type coatings CoCrAl and FcCrAlY are cffective in providing oxidation protcction to titanium intcrmctallics at 815-1000°C but that coating intcgrity was o r could be cventually undermined by coating-substrate reactions and the associated Kirkendall void formation.
324
M.J. Bennett mid S J . Bicll
Coating-substratc interactions wcrc clearly undesirable and the obvious way to improve coating performancc would bc by the insertion of an additional coating layer at the interface with the intcrmctallic to prcvent rcactions and interdiffusion. The validity of this approach has been demonstrated in numerous systems, e.g. where a TIN interlaycr prevented interaction between silica and Incoloy 800H [14]. For the prcsent systcm TiN and AIN were identificd as potential barriers and thin (5-8 pm) layers were dcposited, again by a sputter ion plating route, onto Ti,AI prior to overcoating with a 60 pm thick FeCrAIY layer. 'Ihc effect of peening the FeCrAlY coating using the lower air transport prcssurc was again asscsscd. Cyclic oxidation testing in air a t 800°C was undcrtaken as prcviously and the resulting gravimetric measurcnicnts arc shown in Figure 11.The FeCrAlY + AIN coating, whether unpccncd o r peened, failed by splitting away from the substrate aftcr S0h at thc first thcrmal cyclc. 'I'hc FeCrAlY + TiN coatings, by contrast, affordcd protection over thc cxposure period with thc kinetics and extents of oxidation of the unpeencd duplex coating being comparable to those for unpcened FeCrAlY coated Ti,AI. Throughout the exposure period attack of both coatings was reduced by a factor of two by pccning. Dctailcd surface analysis (Figure 12) confirmcd that protcction of the FeCrAlY + TIN coating again derived from the formation of an a-Al20,surfacc oxide layer and that peening, by densifying the FcCrAlY coating, rcduced its internal oxidation and thereby the mcasured oxygen uptakcs. As oxidation protection was afforded solely by the external FeCrAIY layer the extent of oxidation of an intact coating should not have bccn influenced by the buried ceramic layer, as observed. Raekscattercd electron and Fe, Cr,Al and Ti X-ray images of a transverse cross-scction through pecned FeCrAlY + TiN coated Ti,AI, following 1000 h oxidation in air at 800"C, shown in Figure 12, also clearly demonstrated that the 'I'IN layer was continuous and of uniform thickness. It completcly preventcd the counter movement of elements between the coating and substrate as did a thin (1-2 pm) and Cr metallic layer [12]. No oxygen had pcnetcated bencath the outcr protective oxide layer. Howcver,
i 4
TIME [HOURS)
Fig. 11. Effect of'l'iN and AIN intcrlayers upon the improvcment by a FcCrAIY coating of thc cyclic oxidation behaviour ofTi,Al in air at 800°C
22 Proimion of Tiianiiirn Alunlinides by FeCrAlY
Coatings
325
at the completion of the exposures both FcCrAIY i TiN samples had bowed, indicating delamination of the duplex coating. Analysis (Figure 12) revealed some Al bclow thc 7’iN layer suggesting that the plane of decohesion was within the intermetallic. Again this was a mechanical failure and originated from thermal strains arising from mismatch o f thermal expansion coefficients of the respective layers.
I”ig.12.Backscattered electron image ( A ) and Fe (B),Cr (C),AI (U) and‘li (E) X-ray images of a transverse cross-section through 25 psi peened FeCrAlY + TiN coatedTi,AI after 1000h oxidation in air at 800°C
326
M.J. Rennetl mnrl S J . Bull
Throughout the cxposurc the duplex coating functioned chcniically in providing cxtcrnal oxidation protection and an internal diffusion barrier. However, once again the coating delaminated and esscntially had bccomc ineffective. Thc next objective, therefore, would be to modify the coating chemistry at the ‘TiN substrate interface to improve adhesion. Such an approach has bcen pursued to a succcssful conclusion in several other systems and given the immenscly flexible capability of sputter ion plating processing therc is no reason to suppose this could not be achieved also with the prcsent systems but does require further innovative rcsearch activity.
22.5 Conclusions E‘cCrAIY coatings deposited by sputter ion plating were found to improve the oxidation resistance of both LY-2‘li3Al and y-TiAl alloy matcrials. The oxidation protection was due to the alumina scalc formation on the coating surface but the lifc of the coating was limitcd by chcmical rcactions at the coatinglsubstrate intcrface. A TiN diffusion barrier was effective in reducing thcse intcractions for ‘Ii3Al. However, thc coating/ substrate system is still susceptible to mechanical failure in rcsponsc to stresses generatcd on thermal cycling by thermal expansion mismatch between thc coating layers. Improvements in FcCrAlY/TiN and particularly TiN/li,Al interfaces are necessary to incrcase lifc. Even so therc is cvidencc that failurc has occurred within the Ti,Al and so minimisation of thermal expansion mismatch stresses will be equally important. It may be for this reason that a protective coating with a lower expansion coefficient than FeCrAlY is ncccssary for better protection in these titanium aluminidc systems.
22.6 Acknowledgements This work was partly supported by the Corporatc Research Programme of A E A ’Icchnology. The authors would like to thank Joe Desport, Ron Perkins and Jennie Cullen for experimental assistance with oxidation and coating and Hugh Bishop and Brian Beffamyfor assistance with chemical and structural analysis.
22 Protection o f 7i’tuniutnAliiniiriides by FeCrAlY Coaiing
327
22.7 References [ I ] G H. Meirr: Oxidation of IIigh Tcnipcraturc Intcrmetallics (cditors T. (irobstein and J. Doychack) ‘lhc Mincrals, Metals and Matcrials Socicty,Warrendalc, (1988) 1. [2] A . Rahmel. WJ. Quadakkers. &I. Schiirze: Matcrials and Corrosion 46 (1995) 271. [3] 1). W.McKee, S.C. Ifirong:Coros. Sci 33 (1992) 1899. 141 H . Srrieft S. Poize: High Temperaturc Corrosion (Editor 1i.A. Kapp) NACE (1983) SY I . [S] SC. Kutig: Oxid. Met. 3.1 (1990) 217. [6] S. 7iuiigudii, 7:Shibafo,A . Mirraknnzi: Oxid. Met. 42 (1994) 103. [7] S. Tanigrrchi. 7:.Shihatcr,A. Murnknnzi, K . Cliiharrc. Oxid. Met. 42 (1994) 17. IS] S. Tanigrrchi. 7:Shihatrr, A. Murnkarni, K . Chiharo:Matcrials’1rans.TIM 35 (1994) 616. 191 M . Strangwood, M.J. Renrielc, C A . Hrppslev, J. H. 7iwetl: I’rocecdings o f the Euromat 91 Conference, Institutc o f Mctals. London (1991),224. 101 D.S. Rickerby. M.I. Wood J.Vac. Sci.Technol.A.1 (1986) 2557. 111 D.S. Rickerbp S.J. Hull:Surfacc and Coating’tcchnology 39/40 (1989) 315. 121 D.W McKre. K.1,. /,uthra: Surfacc and Coatings’kchnology 56 (1993) 109. 131 E Wang, H. I,ou, W Lou:Acta Mctall. Sinica 29 (1995) B1 IS. .I41 S.R.J. Snunders, K . S Coley, M.J. Bennett. L).S Hickerby, CEAyres.A.7: Tuson, J.A. Desporf: Mat. Sci. and Eng. A139 (1991) 91.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
23 Hot Salt Corrosion of Titanium Aluminides J.R. Nicholls, J. Leggett and I? Andrews
23.1 Introduction Although y-TiAl was first identifield as a potential high temperaturc structural alloy as early as 1956 111, serious interest in this alloy for gas turbinc applications has only occurred ovcr the last 10-15 years.'Ihis increased interest over the last decade reflccts the attractive properties of this alloy which includc low density, high stiffness, good retention of strength to high tcmperatures, good oxidation behaviour and burn resistance. A number of alloy compositions have been developed that meet thc high temperature mechanical requirements for service up to 750°C and arc therefore suitable for use in gas turbine engine rotational components. Most of thcse alloys are of the form Ti-(4548)AI-( l-lO)X, where X is one of vanadium, chromium, manganese, niobium, tantalum, tungsten or molybdenum [2,3J. These alloys are generally two phased, containing a small amount of a2,as well as the y-phase. Heat treatmcnt of these two phased alloys permit a widc range of microstructures to be achieved, with an associated variation in propertics. Duplex microstructures give greater strcngth and ductility, while fully lamellar microstructures offcr better creep strength and fracture toughncss. In rcccnt years a new type of y-alloy, containing a dispersion of TiB, (thc XDTMprocess) [4], has been thc subjcct of much intcrcst. The addition of 0.8v/o TiB, leads to grain refinement of cast y-alloy microstructures. The y-alloy uscd in this study is a Ti-45A1-2Nb-2Mn alloy, with a 0.8 v/o TiR, addition. This alloy is a y + a, alloy with a lamellar microstructure. It is evident from the forcgoing that considerable effort has gone into improving the high temperaturc properties and microstructure of the alloys. However, if y-TiAl alloys are to find service as high temperature componcnls within the gas turbinc, then data on the oxidation and corrosion data of the alloy is required. A considerable body of data on the oxidation ofTiAl is availablc, albeit for relatively short times at clcvated temperatures, 800-1400°C [S-191. Oxidation data at potential operating tcmperatures within the turbine (650-750°C) is much more scarce, 1171, whilc data on the hot salt corrosion of y-TiAl does not exist. This paper therefore concentrates on a study of the hot salt corrosion of y-TiAl over the tempcrature range 500-700"C. Limited oxidation data at 700°C is also presented by way of comparison with the hot salt corrosion data, and published high tcmpcrature oxidation data.
330
J. R. Nichol1.s. J. Legget! oncl ?i Anciruws
23.2 Experimental Procedure ?'hc oxidation and hot salt corrosion of a l'i-4SA1-2Nb-2Mn alloy, containing 0.8 vol.% TiR, has been studied thcrmogravimetrically. Continuous inicro-thcrniogravimetry was used to monitor thc corrosion processes for temperatures between 600-700°C and exposures of 100h. Corrosion proecsses at lowcr temperatures, SOU and 550°C wcrc monitored discontinuously with specinicns removed and weighed in accordance with the guidelines published for discontinuous corrosion tcsting (201.Samples werc removed and rewcighed at 1,2,5.10,20. SO,100,200,500 and 1000 hours. Test coupons were disk shaped, nominally 10.0mm diameter X 1.Smni thick although thickness varied between l .4 mm and 1.9 nim from specimen to spccimcn. All specimens wcrc cross hatch polished to a 1200 grit surface finish using 240 and 1200 grit SIC abrasive papers. Following polishing, specimens were meticulously cleaned using volasic 344, then isopropanol (IPA) bcforc vapour degreasing using Analar TPA in a reflux condenser for 1 hour. Coupons were then stored in a dcssicator until required. 'The potential of using y-TiAl to replace nickel based alloys within the high pressurc compressor and fourth stage turbine requires that the y-TiAl be resistant, not only to oxidation but hot salt corrosion as well. The level of salt expccted at these location is low, typically 0.34.5 p,g/cm2/h,but ncverthcless may have a significant effect on oxidation processes. Corrosion coupons were salted in a custom designcd salt e.g. apparatus, shown schematically in Figure 1.'Ihc salt solution was mixed to BS3900 pt 4.The salting procedurc consisted of preheating samples to 200"C, using the furnacc attached to the apparatus, than raising the salting chamber and injecting thc salt spray fog, whilst rotating the samples. Thc level of salt deposited (see Figure 1b) is determined by the time in the salt spray cabinet which has previously been calibrated. IJp to six samples can be salt coated at any one timc. One sample from each batch is used'to undertake a salt analysis, the remainder were stored in glass phials in a dessicator until use. The salt analysis consisted of soaking the reference sample in deionised water for 8 h, to redissolve all the salt, then determining the salt concentration in the dcionised water by titration with silver nitrate ( 0 4 . 5 molar) using an Orion EA940 expandable ion analyser link to a single junction reference electrode. Reproducibility of thc salt measurement using this equipment was 0.001 mg/cm2. Oxidation and salt corrosion continuous thermogravimctric studies wcrc undertaken using separate environmental cells attached to a C.I. electronics microbalance with Robal control unit. The balance head was flushed with argon at 35-40cc/min to prevent tbLt+head being contaminated and therefore corroded. The reaction gas was circ~Iated'~around the tcst coupon at a flow rate of 100cc/min. The test coupon was suspended within the environmental cell and furnace using a platinum wire attached to the microbalance head. Mass change data from the microbalance was acquired using a 16-bit data acquisition card attached to a personal computer using a custom built software package. Each data point recorded was the average of 120 repeat rcadings. Using this approach the repeatability of measurement was typically rt 5 p,g/cm2. When very low mass gains werc to be expected, i. c. less than 100 yg/cni' at the end of a 100 h exposure, mass gains were evaluated discontinuously.'I%is applied to oxida-
23 I l o t Salt Corrosioti o f TitaniumAIiiniiriide.7 Steel frame
33 1
. -< ; +
I
Furnace
Locking
nut
Spraying nozzle
. .~
I \
I
\
solution
_ _-.
Coupon Turntable Spraying chamber
Peristaltic
Pig. 1. a) Schematic diagram o f the Salting Kig, b) Salted test piece
tion data at 700°C and hot salt corrosion at 500 and 550°C. Samplcs wcre prepared and salt coatcd as pcr thc continuous thcrmogravimctric studics. During the period of oxidation thcsc samplcs saw a pcriodic cyclc to room tcmperature, where they were weighed. Mass gains wcrc measured after 1,2,5,10,20,50,100,200,500 and 1000h, dcpending on the test in progress, with the specimens reinserted for a further corrosion pcriod. No spallation was noted duc to this cycle for any of thcsc low mass gain samplcs.
23.3 Results 23.3.1 Oxidation of y-TiAl at 7OOOC As statcd in the introduction a considerable body of data exists on the oxidation of yXAI. Much of this rclatcs to studics ovcr thc tcmperature range 800-1400"C, for relatively short exposurc times. These types of experiments result in measurable. casily characterised scales, aiding the understanding of oxidation mechanisms; but do not
J.R. Nicholls, J. Lrggetr and [?
332
Aiidrews
dircctly relate to expected scrvicc tcmperaturcs within a gas turbinc (650 -750°C). 'I'hus a limited oxidation study has bccn complctcd (1000 h at 700°C) to confirm that the oxidation rates on this Ti-45A1-2Nb-2Mn i 0.8v/o TiB, alloy is consistcnt with publishcd data [17], and that similar scalc morphologics result. This data would also provide the rcfcrcncc against which the hot salt corrosion data would be compared. Figure 2 illustratcs the discontinuous mass-gain measured at 700"C, for exposure out to 1000 h. The insert within this diagram is an expansion of thc first 100 h of data. 'l'hc data follows logarithmic-kinctics for the first l00h exposure, Figure 3a, prior to the cstablishment of parabolic kinetics, Figure 3b. At l00h the mass gain is typically 60-80 p&m7, with 230 kg/cm2 gained at 1000 h. Scalc thickncsscs are less than 1.5 bm after 100 h at 700°C and only 0.4-0.5 p n after 100 hours cxposure. XRD analysis of the surface of oxidc scalcs indicated the prcscnce of TiO, and A1,0, in addition to yXAl. Figure 4, a SEM micrograph of the oxidc scale after 100 h at 700"C, shows the presence of AI20, whiskers within the 0.4 pm thick mixed TiO, + AI,O, scale. This data agrees well with the studies of Becker ct al. [17] at 700"C, who have shown that a rnultilayered TiO, + AI,O, scalc forms on y-XAl alloys over thc temperature rangc 700-900°C. The A1,0, laycr acts as a diffusion barrier and is stabiliscd by the presencc of niobium, but may breakdown and reprecipitate in the outer rutilc scalc after extended exposures at higher temperaturcs [17]. No whisker growth was reported at 700°C in the study of Becker et al.. The prcsence of A1,0, whiskcrs suggests the cxistence of short circuit diffusion paths through the rutile scale and implics that the continuous AI,O, barrier laycr has not formed by 100h at 700°C.
23.3.2 Hot Salt Corrosion of y-TiAl, Between 500-7OO0C Tablc 1 and Figure 5 , summarisc thc thcrmogravimctric studies of hot salt corrosion for y-"iAl ovcr the temperature rangc 500-700°C. 0.25
6
N
2 -
0.2
0.15.
c ._ m
;0.1..
=2
0.05:.
, 0.
i
./;"
m
= 0.01 0
20
40
60
80
100 x)
Fig. 2.1000 h discontinuous oxidation of y-?'iAl at 700°C
2.1 llot Salt Corrosion of fitnriiiirii Alurninides
0.05
0.03
y
s
0.02 0.01
--
0, 0.5
0.7
1.1
0.9
13
15
1.7
1.9
2.1
l a g Time (hrs)
Fig.3a. 1,ogarithniic oxidation kinetics of y-'fiAI for first lo() h at 700°C
0
5
10
15
20
25
30
35
Sqrt Time (h)
Fig.3b. Parabolic kinctics of y-'IiAI exposure times exceeding 100 h a1 7 0 0 T
Fig. 4. SEM micrograph of thc surfacc of the oxidc formcd after 100h at 700°C
333
J.I<. Nichvlls, J. I,eggc.tt orid I? Atidr-rLzs
334
Table 1. A summary of the results for discontinuous and continuous hot salt corrosion thcrnogravimetric tests
_____-_-------
lcmpcraturc ("(2)
Exposure Time (hrs)
500 550 550 600 650 700 700
I00 100 100 100 100 100
loo
______---___---
----------____ Salt Lcvcl Final Weight (Kglcni') gain ( p.g/cm2)
35.0 40.6 35.0 33.0 42.3 42.3 33.0 -
62 63 65 345 660 700 1100
--------_---
The first thing to note is thc diffcrcnce in thc magnitude of mass gain from a small amount of salt deposition, prior to oxidation. At 700°C, the mass gain measurcd after 100 h oxidation is 60-80 pn/cm2,compared to 700-1000 mg/cm2 for a similar exposure when coated with salt. Figure 6 illustratcs this behaviour. A second point to notc is the shape of the hot salt corrosion curve compared to that undcr isothermal oxidation. It can be scen that a major part of this mass gain, some 30-35 YO,occurs within the first 10h of exposure; where upon the corrosion rate plateaus, increasing again aftcr some 40 h of cxposurc.
20
0
60
80
100
Corrosion time (b)
a)
0
b)
40
20
40
60
80
100
Coirosion time (h)
Fig. 5.'lhcrmomicrobalance studies, under hot salt corrosion conditions, a ) Data over the tcmperaturc range 500-700"C, b) Data at 500 and 5 5 0 T
23 Hot Salr Corrosion of Tilonium Alrrminides
335
As for thc early stages of oxidation at 700"C, the carly stages of hot salt corrosion were also found to follow logarithmic kinetics, see Figurc 7 as an example.'I'his behaviour was observed for all the tcst tcmpcratures examined between 500°C and 700°C. A t 500 and 550°C the logarithmic kinetics pcrsisted out to 100 h, with mass gains after 100 h at thcse ternperaturcs similar to that following 100 h oxidation at 700°C. At 600°C the initial logarithmic kinetics gave way t o a more rapid form of attack after 17 h, which became asymptotic by 35-40 h. N o further breakaway transitions were observed out to 1.00h. At 650°C a similar bchaviour to that at 600°C was obscrvcd, although thc transition to more rapid attack now occurred al'tcr 4 h. Thc sccond plateau again was formed, giving rise to parabolic kinetics after some 65 h of cxposurc. At 700"C, the carly stages o f corrosion again followed logarithmic kinetics.Thc transition to more rapid attack (at around 4 h at 650°C) was not observed at this tcmpcrature. It could have possibly occurrcd at much shortcr timcs, say bclow 1h, in which case it would not have been detected using the measurement methods in this study. Instead two types of behaviour were observed: a more gradual transition from logarithmic to
000
2000
4000
6000
8000
10000
Corrosion time (h)
Fig.6. Comparison of mass gains at 700°C in the prescnce of a 37 Fg/crn2 salt deposit, with that under isothermal oxidation conditions
-1
-05
0
0.5
1
15
2
Log time (h)
Fig.7.1,ogarithmic corrosion kinctics for thc early stage of hot salt corrosion attack of y-IiAl, at 600°C
336
J. K . Nicholls, J. Legget1 ond I? Andreus
parabolic kinctics for onc samplc with the transition around 60 h and a rapid brcakdown for the second after 43 h. For the former the mass gain at 100 h was 700 pglcni’ while for the lattcr it was I100 pgkm’. ’Ihe scalc morphologies observed during the hot salt corrosion of y-Ti A l arc illustratcd in Figure 8, for l00h exposures at tcmperaturc bctwcen 500 and 700°C. Retween 500 and 600°C isolated arcas of increased corrosion are observed, associated with the individual salt droplets that had becn spraycd onto thc saniplcs surface (Figure 8a). N o whisker growth was obscrvcd (Figure 8b) and the surface oxidc was predominantly rutilc. At 650°C: and after 100 h exposure, N205 whiskers were present on top of the rutile scalc (Figurc 8c). For tcmpcraturcs o f 650°C and above, the alumina whisker growth was found to spread across the surface with time at tcniperaturc. For example, at 650°C after 2 h no whisker formation was obscrvcd, with the scale morphology similar to that after l 0 0 h at 600°C (Figure 8b). By 11 h a few isolate regions of alumina whiskers wcrc forming, although these wcrc often difficult to find. After
Fig.8. Surface micrographs of hot salt corrosion of y-TiAl a ) 1Ot)h at 550°C b) lOOh at 600°C c) 100h at 650°C d) lOOh at 700°C
2.3 IIot .Tall Corrosion of Titanium Aluminides
337
100h, many arcas of alumina whisker growth wcre evident (Figure 8c). Raising the tempcrature to 700"C, resulted in a profusion of alumina whiskers forming, which now totally covcrcd the surfacc after 100 h (Figure 8d). 'liansvcrsc sections wcre taken through salt corroded samples at all temperatures, following l0Oh exposurc. At 500 and 550°C a vcry thin scale was formed, typically 0.5 pm thick. Aftcr 100 h at 550"C, prefercntial attack o f the a,-phasc in the Ti-45Al2Nb-2Mn + O.8v/o XB, was observed. ?'he dcpth of internal attack was only a few microns and of a similar lcvcl to that aftcr 11 h at 650°C. By lOOh at 650"C, the internal attack of the a, lathes were well established, extcnding somc 6-10 pm into thc two phase y-TiAl alloy, as can be seen in Figurc 9a. Following 100 h at 700"C, it was n o longer necessary to use the back scattered detector on the SEM to identify the depth of intcrnal attack. It was clcarly visible by optical microscopy (Figure 9b), with the depth of internal attack penetrating somc 45-60 pm. Thc associated scale thickness was 13-20 pm. increasing locally to over 30 km, Figure 9c. EPMA and XRD of the thicker hot salt corrosion scales, shows that the majority of the scale is rutile. Alumina is present as a discrete phasc within the scalc, but does not form a continuous layer. Furthcr alumina is present as whiskers on the scalc surface. Aluminium is also detected within the rutile rich regions, at a low level, inferring that it is either in solution or distributcd as micro particles bclow thc resolution of the SEM.
a)
Back scattered SEM micrograph: 1Wh at 650%
b)
Optical micrograph following
100h at 7OO0C
c)
SEM micrograph in an area
of local attack: 1OOh at 700%
Fig.9.Tranversc sections of hot salt corrosion of y-l'iAl
23.4 Discussion 23.4.1 The Role of NaCl in Accelerating the Oxidation of y-TiAl It is clear from the results scction that the small level of salt dcposited i n this study (37 pg/cm2) had a significant cffcct on the ratc of oxide growth on a two phase y I a?. Ti-45AI-2Nb-2Mn alloy. At 700“C, the corrosion rate in a salt ladencd atmosphere was upto 20x that of the cquivalent oxidation, following a 100 h cxposure. One must ask, thcrcfore “Are thcsc lcvels of salt to be expected in scrvicc?” for if they arc hot salt corrosion will be a significant factor in the adoption of y-‘FiAl for turbine applications; and secondly “What is the mechanism by which small salt dcposits accelerate the oxidation ratc of y-TiAl‘?”. To address thc first qucstion, calculations have been made by Hnncock [20] of the salt flux likely to be ingested in to a mariniscd gas turbinc (obviously a severe cxample). For 0.01 pppm salt ingestion into the enginc, the contaminant flux is expected to be 270 pg/cm2/h.Assuming that only 10 % of thc ingested salt may deposit and that the atmosphcric aerosol rcduces by 2 ordcrs of niagnitude from 6 m abovc sca lcvel to 600m abovc sea level [21], then salt deposition rates within operating civil aircraft could be expcctcd to be 0.27 pg/cm2/h.Hencc, 37 pg/cm2 per 100 h corrosion test is not unrcalistic. At such low salt Icvels, the salt will deposit as local salt ladened droplets which then evaporate, or as microscopic salt particles. This mcans that the salt concentration secn across thc surface will vary on a micron scale as can be seen for an as spraycd sample in Figure 1b, and for a corroded samplc in Figure 8a. Titanium alloys have bccn know to be susceptible to hot salt corrosion, and more particularly hot salt stress corrosion (HSSC) for over four decades. Bauer in 1955 1221 had noticed sevcrc surface cracking of a Ti-6A1-4V alloy, during a basic crccp test. Thc local attack, which lead to premature failure, originated in an area of corrosion in the pattern of a finger print. Thus Bauer attributed thc onset of failure to contamination and corrosion by salt prcscnt in perspiration. This obscrvation lcads to considcrablc work ‘in the sixtics and seventics into the hot salt stress corrosion of titanium alloys. During such studics Ashbrook 1231 undertook ficld survcys of the actual salt deposhed onto compressor components, during engine overhaul. Avcrage conccntrations of 155 Fg/cm2 were measurcd, consistent with the abovc salt deposition calculations if thc periods bctwcen overhaul wcre of the ordcr of 500 flight hours. A rcsult of thcsc hot salt corrosion and hot salt stress corrosion studics of titanium alloys was the importance of alloy composition in dctcrmining a titanium alloy susceptibility to attack. It was shown that alloys high in aluminium wcre prefercntially attackcd by hot salt [24,25], with both A1 and Zr being incorporatcd in the internal and extcrnal oxidc scales [26].l’his early work raised the importance of aluminium in thc hot salt corrosion of titanium alloys. ‘I’hc observation of A120, whisker formation, the lack of formation o f a continuous alumina layer within the oxidc, and the discrctc precipitation of alumina particles con-
23 Hor Solt Corrosiori of Titnriium Aluminides
339
firms that an interaction bctwcen the salt contamination and aluminium within the alloy is a significant factor i n thc acceleratcd oxidation of this Ti-45A1-2Nb-2Mn. y + a 2 two phasc alloy.
23.4.2 A Mechanism for Salt Accelerated Oxidation of y-TiAl NaCl has bccn shown to have a major influencc on the early stages o f scale formation. Evcry corrosion test, when salt coated, suffcred a high initial rate of mass gain, over thc first 5-10 h of exposure.This bchaviour can be seen in Figurc 6, where at 700°C the initial mass gain is increased by an order of magnitudc in the presence of salt; and is evident in the hot salt corrosion tests at 500 and 550°C which have mass gains similar to that obscrvcd at 700°C undcr isothermal oxidation conditions. From thcsc 500 and 550°C test thc first stage in salt accelerated oxidation can be deduced. As cvident in Figurc 8a this consists of localised corrosion in areas of salt deposition. Figure 10 is a schematic diagram of the three stages of scale growth observed for a two phasc, y tcu,TiAl alloy. Surface before exposure Large salt deposits
Stage 1: Short time, high T or long time, low T Liquid salt content increases. and spreads over wrface
++
Small deposits converge Thin oxide
Stage 2: I,:,C I
j
w'
dWhisker 1
3
1
growth w
Thin outer rutilc layer builds up around high [salt1
v L
salt incorporated in the oxidc Surface oxide thickening
Salt altack along alpha 2
Stage 3:
n - y U F a ! & r
growth over whole
Outcr & inner oxide Iaycrs meet as salt 'used up'
Severe chloride altacl( along alpha 2 plates
Fig. 10. Salt corrosion mechanism for Ti-4s-2-2
340
J. I<. Nichollr, J. I,cggctt atid P Atidreus
23.4.2.1 Stage 1 Growth During stage 1 growth. the localiscd salt deposits react with thc t h i n rutilc surface scale, formed on thc alloy surfacc. Drawing on the cxpericnce of hot salt corrosion o f titanium alloys, thcn sodium titanate [27] and sodium hydroxide [30] should be cxpected to form. Possible rcactions include: NaCl -1- H,O 2NaOH + TiO, ’I’i -1- 2NaCI + ’/202 1 H?O ‘l‘iC12+ 0,
+ -+ + +
HCl + NaOH (1) Na,TiO, t H?O (2) TiClz I 2NaOII (3) TiO, t CI, (4)
Thus thc rcaction of sodium chloride with titanium from the alloy, or rutilc from the scalc lcads to the formation of TiCI,, Na,TiO,, HCI and CI,. This list of rcactions is by no mcans exhaustivc, and may also involve reaction with other alloying additions which may be substituted for thc titanium metal. Particularly, the rolc of aluminium must be of importance. A thcrmodynarnic analysis by ‘lkuvkin et al. [32]confirms that complex oxidc scales are formcd when titanium and its alloys react with NaCI. Particularly that thc formation of volatile mctallic chloridcs arc thermodynamically favourable, especially for alloying additions Zr, Mo and Al. The subsequent pyrohydrolysis of thcsc mctal chloridcs results in thc formation of I-ICI gas, particularly with MoCI,, and AICI,. Furthcrmore, such pyrohydrolysis of halidc salts may be accclcrated by thc presence of alumina within the scales, which acts as a catalyst [25]. Thus during stage 1 growth, reaction of the local salt deposits with the thin rutile scale, and alloy surface, lcads to the formation of sodium hydroxide, sodium titanate and furthcr rutile through pyrohydrolysis of TiCI2. Similarly, alumina is formcd within the scalc through pyrohydrolysis of AICI,, hcncc,
2A1 + 6NaCI + 1’/,0,+ 31H20 2A1C13 + 3H,O
+ +
2AIC1, + 6NaOH ( 5 ) AI,O, + 6HCI (6)
whcrcupon the prescnce of a finc distribution of alumina furthcr catalyscs thc pyrohydrolysis reactions. A result of thcse reactions is that HCl and C1, gascs arc released to diffuse through the scale and react with the alloy furthcr, at the alloy/oxide interface. The formation of TiCI, and AICI,, results in thc formation of mixed chloride salt deposits, lowcring the mclting point of the salt, whcreupon thc liquid contcnt of the salt dcposit increases and the salt spreads latcrally across thc surface of thc sample, increasing the arca open to salt acceleratcd attack. A nct result of this complex serics of reactions is that the corrosion rate initially incrcases rapidly, until the complctc surface is covcred in salt, whereupon transport through the now thickened scalc controls thc corrosion ratc. During this first s t a g of attack corrosion rates arc observed to fdllow logarithmic kinetics.
23 Hot Solt Corrosion of 7itcmirrm Aliirninidrs
341
23.4.2.2 Stage 2 Growth Stage 2 growth is charactcrised by the appcarance of alumina whiskers on the oxidc surface, thc onsct of chloride attack of thc a2 lathcs and the formation of a porous rutile scalc in areas of high salt concentration. This position is reached after some 100h at 550"C, 17 h at 600°C and 4 h at 650T.The existcnce of stage 1 growth at 700°C has not bccn identified in this study. but is believed to occur within thc first hour of cxposure at 700"C, thus the Ti-45A1-2Nb-2Mn alloy appears to cntcr stagc 2 growth directly when exposcd to salt at 700°C. Stage 2 growth is thought to involve thc transport of HCI and CI, through pores and channcls i n the rutile scale, formcd during stage 1 growth, down to thc metal-oxidc interface. Here, the halide gases rcact with the metal surface forming volatile chlorides, viz: Ti Al
i
21IC1
+ 31IC1
+ -+
TiCl, t 211 AICI,+ 3H
' h e formation of AICI, is thermodynamically more favoured. Thc AICI, and ?'iC12that are formed diffusc outward through the same porcs and channcls, where thcy rcact in thc outer part of thc scale, or at the scalc surfacc. Any hydrogen that is formcd diffuses away into the y'I'iA1 alloy. As was identified for high aluminium containing titanium alloys, the a,-phase is preferentially attacked 127-291. At, or ncar, the surfacc the mctal halides rcoxidise, for example. TiCI, i 0, 2AICl, + 3H,O
-+ +
Ti0, + C12 AI,O, + 6I1CI
(9) (10)
'I'hus, thc halide containing gascs are regcnerated, rcady to rediffuse in through the scale and into the alloy, and a counter flux of gascous halidc spccies are cstablished in thcsc channels and pores with HCI or C1, diffusing inward and AICI, or 'liC12,diffusing outward. Through this mechanism a self sustaining corrosion process is established, that requircs only small quantities of chloridc phases bc present. This vapour transport mechanism along short circuit diffusion paths accounts for the growth of thc alumina whiskers. Thcsc grow preferential where the short circuit path through thc oxide reach the surfacc.'fie AICI,, that is formed within the alloy or at the mctal oxide intcrface, rapidly oxidises to Al2O, and producing further HCl. 'Thus thc alumina grows at the top of these channels forming hollow whiskers. Transport can also take placc through the porous oxide scalc that is formcd, thus a fine dispersion of alumina is found throughout the scalc and no continuous alumina inner oxidc can be formcd. The prcfcrential attack of the a, lathcs is evident as dcep channcls penetrating into the Ti-45A1-2Nb-2Mn alloy, see Figure 9a, which may penetrate up to 40-50 pm after 100h at 700°C. Stage 2 also follows logarithmic kinctics, reflecting competition bctwcen parabolic oxidc growth and short circuit diffusion down prefcrred channcls. Initially, the short circuit paths account for the early observcd rapid scale growth. A transition is later observed to parabolic kinetics, which marks the onsct of thc third stage of scalc growth in hot salt accclcrated oxidation of y-TiAl.
342
J.K. Nicholls, J. Leggeti and P Andrews
23.4.2.3 Stage 3 Growth Stage 3 growth is charactcrised by a surface covered with a profusion of alumina whiskers, thick two layered scales and scvere internal attack along a2 lathcs. Figures 8d and 9b illustrate this mode of attack. The fact that parabolic kinctics are now obscrvcd, suggest that diffusion transport through the scalc now dominates. It is thcreforc likcly that either the short circuit diffusion channels have closed, or that the path length (a7 lathe depth of attack + scale thickncss + alumina whiskcrs length) is so long that thc short circuit flux matches that transported through the scale. A further contributing factor to the transition to stagc 3 growth in thcse studies is the observation that little salt remains at the intcrfacc betwccn the inncr and outcr scale. Thus thc available vapour halide species that contributc to stage 2 growth must have been lost to the environment. In reality, under turbine service conditions, the available chloridc species would be continuously replaced and this would have thc effcct of extending the region of stage 2 growth.
23.5 Conclusions 1. Thc presence of small amounts of salt, typically 37 Fg/cm*,rcsult in accelerated oxidation rates for a y + ci2TiAl alloy ovcr thc temperature range 50&700"C. 2. At 500°C and 550"C,scaIe growth follows logarithmic kinctics and no transition to parabolic growth is observed in 100h exposure studied. 3. The scale growth rates at 500 and 550°C in a salt ladencd environment were similar at 62-65 Fg/cm2/100h.This rate of attack was obscrved at 700°C for tcsts in a salt free environment. 4. Bctwecn 600 and 700"C, scale growth was characteriscd by the formation of AI2O, whiskcrs and preferential internal attack of the a,-phase. Both the amount of whiskers formed and thc depth of internal attack increased with time at tempcrature. 5. At 700"C, after 100h exposure, the corrosion rate in the presence of salt is up to 20x that of the equivalent oxidation exposure. 6. A mechanish based on the formation of volatile mctal halides is prcsentcd. The corrosionlprocess proceeds in three stages: i) Enhanccd oxidation is observed in areas of salt deposition. Keactions between the salt, thc rutile scalc and the metal surface lead to thc formation of volatilc metal halides, sodium titanate and gaseous halide species. Solution of the metal halides into the salt deposits, lower thc salt melting point such that molten salt deposits are formed that can spread laterally across the mctal surfacc, accclerating the rate of scale formation. ii) Pcnetration of the gaseous halide specics through the porous scale allows dircct rcaction with the TiAl substrate at thc metal oxidc intcrfacc. Prefcrential attack of a2occurs, leading to deep channels into the substratc, and the formation of vapour phase metal-halide species. Outward diffusion of these gaseous mctal ha-
23 H o t S d t Corrosion of7i’laniirm Aluminides
343
lides results in the growth of a porous scale and the formation of alumina whiskcrs. Keaction at the oxide surface results in the regeneration of the gaseous acid halide species and hence the corrosion mechanism becomes self sustaining. iii) A transition to parabolic kinetics is observed after extended attack of the substrate. This is bclicved to reflect thc loss of available gaseous halides to thc environment. removing the transport species for stage 2 attack.
23.6 Acknowledgements The authors are grateful to Kolls Royce plc and the Defcncc Iiesearch Agcncy (DRA) for providing the research funds to completc this work. One of the authors (J.L.) wishes to thank Kolls Royce plc for the research bursary to study for his PhD. Particular thanks go to Dr. Sreve Kerry. at the DRA, for his knowledgeable discussion throughout the progress of this work.
23.7 References
151
161 I71
1131
J.B. McAndrew, 1I.D. Kessler:Journal of Metals (1956) Oct. 1’1348. Y - W .Kim: Journal o f Metals 43 (1991) 40. Y - WKim: Journal o f Metals 46 (1994) 30. I,. Christodoulou, R A . Parrish, C R . Crowe:in: *‘High?‘emperature/High Performance Composites” I’roc. MRS symp. 120 (1990) 889. A. Rahmel, RJ. Spencer: Oxid. Met.35 (1990) 53. D. W. McKee, S C. Huntig: in: “High ‘kmpcrature Ordered Intermetallic Alloy IV”, Proc. MKS Symp.213 (19%))937. M . Yoshihnra et al.: in: “High Temperature Ordered Intermetallic Alloy IV”, Proc. MRS Symp.213 (1990) 97.5. K.J. Liithra: Oxid. Met.36 (1991) 47.5. D. W McKee, S. C. Ifuang: Corrosion Science 33 (1991) 1899. U.Ilerold-Schmidt,R. Opolka, S Schwantes: Prakt. Metallogr. 30 (1993) 344. A. Takasaki, K . Ojima, Y Taneda, 7:Hoshiya.A. Milsuhashi:J. Mater. Sci. 28 (1993) 1067. G. Welsch,A. Kaveci: in: “Oxidation of High Temperature Intermetallics” (eds. T. Grobstein, J. Doychak).’Ihe Minerals, Metals and Materials Society,Warrendale, PA (199S), p. 207. K. W! Wiedemann. S N. Sankaran, H.K. Clark, ?:A. Wullace: in: “Oxidation of IIigh Tcmperaturc Intermetallics”, (eds.‘r. Grobstcin and J. Doychak),Thc Minerals, Metals and Materials Society,Warrendale PA (1988). p. 195 7: Shida, I/.Anada: Corrosion Science 35 (1993) 945. C; Chen, %. Sun, X . Zhou: Corrosion 48 (1992) 939. 1% Figge, A. Elschner, N. Zheng, H. Schuster, WJ. Quudakkers: Frescnius J. Anal. Chcm. 346 (1993) 75.
S. lkcker,A. /
[24] R.S. Ondryjcirr. M . R. 1,oLirban: “Role of Chloride i n Hot-Salt Strcss-Corrosion Crackin? o f ‘l‘itaniuni-Aluminium Alloys”, NASA ClCl133, (1968). [2S] S. /? liirleour. S. /? I,ourhori, C. I<. Selby: “Basic mechanisms o f stress-corrosion cracking o f titanium”. in Sti-ess-Corrosion Cracking of’l‘itanium,ASI‘M STP397. pp. 137 1.51.(1966). [26] F! Dunzrrs, C. .Sr.John: Oxid Met. 10 (1976) 127. 1271 VC.Prrer.yeti, H.H. l~ornbrrge,.:“’Ihc mechanism o f salt attack on Titanium Alloys”, in Stress Corrosion Cracking of Titanium. ASIM STP397, pp. 80-94, (1966). I281 H.K. Gray:“Hot-Salt Stress-Corrosion of Titanium Alloys as related t o turbinc cngine opcration; in Titanium Science and ‘l’cchnolo_py.Plenum I’rcss. New York, (1973), pp. 2627-2638. 1291 J.R. Myers, J.A. IInll: Corrosion 33 (1977) 252. 1301 G. Mnrrin: “Investigation of long-term exposure effects under stress o f two titaniuni structural alloys”. in Stress (’orrosion Cracking ofl’itanium, ASTM. SI‘P396, pp. 95-120. (1966). 1311 A.J. Ilnrch. 1I.W Kosenherg, 1i.E Erhiti: “Effects of cnvironment on cracking of titanium alloys”, in Stress Corrosion Cracking in Titanium, ASTV ST1’397. pp. 122-1 36 (1 9%) [32] V V Tmvkin, V F fshirkov, B.A. Kolashev: Sov. Mater. Sci. 15 (1979) 134.
Oxidation of Intermetallics by H. J. Grabke, M. Schiitze 0 WILEY-VCH Verlag GmbH, 1998
Index
a,-alloys, titanium aluminides 3 a-alumina formation 19 a-crystobalite 69 accelerated oxidation - intermediate temperatures 23 - refractory metal compounds 47 - titanium aluminides 338 acoustic emission analysis, nickel aluminides 135 advanced aluminides 4 air, exposure conditions 39 AISI stainless stccls, corrosion behavior 162 A1201dispersion, iron aluminides 187, 197 AI,O, whisker formation, titanium aluminides 338 alloy 800H - carbunzing 210 - chemical composition 222 - iron aluminides 205 - oxidatiodsulfidation 2 16 alloy compositions, nickel aluminides 103 alloying additions - iron aluminides 190, 218, 233 - Nb-A1 system 53,72 - TiAl-base materials 61,63, 299, 340 alloying element effects 36 alloys, iron aluminides 205 alumina growth rates 15 alumina modifications - iron aluminides 177 - nickel aluminides 113 alumina scale failure, NiAl 135 ff alumina whiskers 341 aluminides see individual types aluminium content - iron aluminides 218 - nickel aluminides 86 - sulfidation 89 aluminium depletion layer - TiAl 265 IT - titanium aluminides 257 aluminium implantation, 'l'i-48AI-2Cr 30 I
aluminizing - TiAl-base materials
66 titanium aluminides 313 analytical techniques, nickel aluminides anisotropic crack pattern, nickel aluminides 147 anncaling, nickel aluminides 100 arc melting NiAl 163 - nickel aluminides 100 argon, exposure conditions 39 argon moisture, y-TiAl 292 Arrhcnius diagram - iron aluminides 178 - NiAl/FeAl 79f - nickel aluminides 87 Auger electron spectroscopy (AES) - iron aluminides 225 - Nb-A1 system 53 - nickel aluminides 81 - y-TiAl 290 austenitic stainlcss steels, corrosion behavior 162 -
kFeA1, iron aluminides 175 ff PNiAl oxidation, CeNIHf doped 109 ff kstabilizing elcments 29 backscattered electron imaging (BSE) - titanium aluminide coatings 325 - titanium aluminides 248 barrier layer formation, TiAl-base materials 62 bend tests, nickel aluminides 136, 140 binary titanium aluminides 275 ff blockage, diffusion paths 308 bowing, titanium aluminides 322 breakaway oxidation - iron aluminides 192 - TiAl-base materials 66 - y-TiAI 293 Bridgman method, nickel aluminides 137
86
346
Index
brittle-to-ductile transition, silicides 73 brittlcness, TiAl-base materials 66 carburization, iron aluminides 210, 235 cast Fe,Al 185 casting, titanium aluminides 265 catalytic cracker plant, iron aluminides 22 I cation additions, iron aluminides 198 cation diffusion, nickcl aluminides 133 cavity formation, nickel aluminidcs 81 ceramic coating, TiAl-base materials 67 ceramic layers, corrosion behavior 162 cerium - iron aluminides 190, 205 - nickcl aluminides 109 f chemical compositions - iron aluminides 177, 185, 205, 222, 233 - nickel aluminides 103, 111 - titanium aluminides 246, 270 chlorination, iron aluminides 2 10 chromia, iron aluminides 203 chromides 6 chromium, alloying element 36 chromium additions - nickel aluminides 100 - TiAl-base materials 64 - titanium aluminides 329 chromium diffusion, titanium aluminide coatings 320 chromium doping, metal dusting 233 chromium implantation, Ti-48A1-2Cr 30 I chromium impurities, nickel aluminides 139 chromium influence, 0-NiAI 99 ff ehromizing, TiAl-base materials 67 coal conversion, iron aluminides 22 I coal gas 204 coating failure, titanium aluminides 323 coatings 24 - ODS iron aluminides 183 - TiAl-base materials 67 - titanium aluminides 44,3 13 ff Coble creep I55 coke vast protrusions, iron aluminidcs 234 columnar grains, titanium aluminides 269, 314 composite defect size, nickel aluminides 151 composites 26 compounds oxidation, Nb-AI system 52
comprcssive load, nickcl aluminides 143 convergent beam electron diffraction (CBED) 35 copper addition, TiAl-base materials 64 corrosion - itastelloys 162 Incoloy/Inconel 162 - iron aluminides 203 A; 217 NiAl in molten carbonate 161 ff nickel aluminides 85, 92 - TiAl 300 - titanium aluminides 329 ff corrosion layer, Ti-48Al-2Cr 302 corrosion rncchanism, salts 339 cracking - binary titanium aluminides 277 - iron aluminidcs 196 .- nickel aluminides 135, 142, 146 - - under compression 154 - - under tension 151 - orthorhombic alloys 45 - refractory metal compounds 49 - TiAl-base materials 66 - titanium aluminides 322 creep - - iron aluminides 183 -- nickel aluminides 155 cyclic oxidation kinetics - NiAl 22 -. TiAl 300 - titanium aluminides 324 defects refractory metal compounds 49 - TiAI-base materials 70 deformation, titanium aluminides 322 deformation methods, nickcl aluminides 136 delamination - nickel aluminides 135 - titanium aluminide coatings 325 depletion layer, titanium aluminides 249, 257,265 R - binary 276 diffusion - rcfractory metal compounds 49 - titanium aluminides 66, 341 diffusion barrier, titanium aluminides 307 diffusion cocfficicnt, iron aluminidcs 203 -
Index dislocation pinning, iron aluminides 197 dispersion, iron aluminides 187, 197 dissolution - nickel aluminides 166 titanium aluminides 259 doping - iron aluminides 198 - nickel aluminides I09 ff drop casting, titanium aluminides 265 ductility, orthorhombic alloys 40 ductilization 3, 7 duplex coating delamination 325 duplex microstructures 329
EDSIEELS, TiAl 265 electron probe microanalysis (EPMA) 3 15 electropolishing, TiAl-base materials 70 elemental spectroscopic imaging (ESI) 34 - TiAl 266 embrittlement - titanium aluminides 39 -- U S . research 15 ff energy dispersive X-ray analysis (EDX) - binary titanium aluminides 275 - iron aluminides 225 - nickel aluminides 86, 110, 164 - titanium aluminides 246, 3 I4 environment effects - iron aluminides 223 - titanium aluminides 42 evaporation, Mo oxides 47 exposure conditions - embrittlement 39 - iron aluminides 176 extmded Fe,AI 185 FA 49/56-58, iron aluminides 222 failure, alumina scale 135 ff failure criterion 26 FAS iron aluminides 187 Fe-A1 alloys, iron aluminides 175 ff FeAl oxidation 79 f Fe-GI-Cr alloys 203 ff - oxidatiodsulfidation 216 Fc-A110-Cr2, chlorinating 2 11 Fecralloy, chemical composition 222 FeCrAlY coatings 3 I3 ff
347
finite element method (PEM) 148 flaw size, nickel aluminides 15 I flow apparatus, iron aluminides 176 fluidized bed combustion 204 four-point bending, nickel aluminides 140 fracture analysis, nickel aluminides I56 fracture touchness 18 - - y-alloys 41 - nickel aluminides 15 1 y-alloys 36, 41 y-phases, superalloys 4 y-titanium aluminides 4 gas cooled reactors 204 gas turbine applications 329 graded scale loading, nickel aluminides 135 ff grain boundaries - nickel aluminides 102, 111 - Zr segregation 128 grains - iron aluminides 188 - layers 60 - titanium aluminides 260, 269, 3 I4 gravimetry, titanium aluminide coating 3 15 growth, a-Al,O, 117 growth rates, ahmindsilica 15 Guinier camera 79 Guinicr-Wolff method 164
€1,-I1,S sulfidation, nickel aluminides 87, 97 €J2-H2S-H20,nickel aluminides 92 hafnium addition - iron aluminides 191 - metal dusting 233 - nickel aluminides 109 f - reactive elements 18 hafnium content, TiAl-base materials 63 halide activated pack-cementation 44 halide containing gases 341 Hastelloys, corrosion behavior 162 Heat T899 I , titanium aluminides 3 13 helium thennogravimctry 175 Heusler-type phases 8 high-activity proccsses 25 high-temperature applications, nickel aluminides 85, 121
348
Index
high-tcmpcrature corrosion 203 ff high-temperature corrosion resistance 183 high-temperature X ray diffraction 1 14 high-velocity oxygen fuel spraying 44 hot-process gas corrosion, iron aluminides 2 I7 hot-salt corrosion 329 ff, 338 hydrogen, y-TiAI 292 impurities 18 IncoloyiInconcl, corrosion behavior 162 increased spallation, ODS FAS 194 inlet gas composition, pot-cell 162 inner layer, grains 60 interdiffusion -~ refractory metal compounds 51 titanium aluminidc coatings 3 19 intcrfacc kinetics, titanium aluminides 259 intcrfaeial voids 2 1 intergranular diffusion 49 intergranular oxidation 23 NiAl 82 intermetallic matrix composites (IMC) 6 internal oxidation 23 binary titanium aluminides 278 coatings 319 Fchl 181 . - iron aluminides 208, 2 13 - nickel aluminides 98 - titanium aluminides 254 internal oxidation zone (IOZ) 20 interstitial affected zone (IAZ) 40 interstitials, titanium aluminides 308 ion beam enhanced deposition 68 ion implantation, TiAl-based intermetallics 299 ff ion intensity ratio, titanium aluminides 241 iron aluminides 5 , 173-236 - Jwanese research 71 - U.S. research 15 ff iron-aluminium-chromium see; Fe-Al-Cr iron diffusion 320 isopropanol 330 isothermal oxidation, binary titanium aluminides 280 Japanese research
59 ff
K phase formation, metal dusting 235 Kanthal iron aluminidcs 190, 199 - nickel aluminides 170 kinetics, nickel aluminides 93 Kirkendall cffcct, TiAI-basc materials 66 Kirkendall voids - iron aluminides 200 - titanium aluminides 323 Knudscn cell 239 lanthanum, iron aluminides f90, 205 laser welding, NiAl 163 lattice defects, TiAl-base niaterials 70 lattice parameters, titanium aluminides 267 Laue diffraction 122 Laves phases - chromides 6 - titanium aluminides 38 layers - aluinindsilica 16 - titanium aluminides 60, 247 life prediction 26 long-term behavior, titanium aluminides 245 low-activity processes 25 magncton sputtering, TiAI-basc matcrials 67 manganese addition - TiAl-base materials 64 - titanium aluminides 329 manganese implantation, Ti-48A1-2Cr 303 mass change -- alloy 800H 213 carburizing 2 10 iron aluminides 208 mass gain - hot-salt corrosion 332 - ion implantation 301 - iron aluminides I86 - Nb,Al-base materials 72 - nickel aluminides 88, 113 - Ti36AUTi35A15Nb 247 MCFC anode, nickel aluminides 161 MCrAlY coating, TiAl-base materials 67 mechanical behavior, oxide scales 135 mechanical deformation, titanium aluminides 322
Index melting, titanium aluminides 275 metal activity, iron aluminides 203 metal dusting - Fc,AI/(Fe,l\ii),Al 233 ff - iron aluminides 217 metal halides, titanium aluminidcs 341 metal loss, iron aluminides 2 13 metal wastage, iron aluminides 234 metastable alumina modifications 177 metastable phascs, nickel aluminides 12 1 microcracks 49 microhardness measurements 270 mischmetal, iron aluminides 207, 2 19 mixed gas tube, schematic diagram 223 moisture, y-TiAl 289 ff molten carbonate corrosion, NiAl 161 ff molybdenum - refractory metals 16 - titanium aluminides 329, 340 molybdenum disilicides, refractory metal compounds 46 molybdenum implantation, Ti-48A1-2Cr 303 morphology, sulfide scales 87 NaCl accelerated oxidation 338 Nb,Al-base materials 72 needle shaped grains, titanium aluminides 260 neodymium, iron aluminides 191 new cubic phase (NCP), titanium aluminides 258 NiAl oxidation 79 f nickel alloys, corrosion behavior 162 nickel aluminides 4, 77- 172 - Japanese research 71 - U.S.research 15 f nickel base superalloys 17 nickel in molten carbonate, polarization curves 167 niobium addition - p-stabilizing clement 29 - refractory metals 16 - TiAl-base materials 65 - titanium aluminides 261, 299, 329 niobium containing alloys 279 niobium implantation, Ti-48AI-2Cr 304 nitride incorporation, TiAl-base materials 68 nitrogcn content, iron aluminides 185
349
nitrogen distribution, Ti45A15Nb 283 nitrogcn effect, titanium aluminides 32 1; 258,275 ff nodulcs - y-TiAl 292 - nickel aluminides 101 NRA, binary titanium aluminides 275 nucleation, a-Al,O, 132 optical fluorcscencc spectroscopy (OFS) 136 orthorhombic alloys 30 outcr layer, grains 60 outgassing, titanium aluminides 3 14 oxidation life times 27 oxidation stagcs - NiAl single crystal 122 - titanium aluminides 245 ff, 257 oxidation-sulfidation, iron aluminides 22 I ff oxidation times, nickel aluminides 100 oxidc addition, iron aluminides 191 oxide dispersion strengthened (ODS) iron aluminides 183 ff oxidc layer variations, nickel aluminides 104 oxide scale, titanium aluminides 247 oxygen, exposure conditions 39 oxygen content, iron aluminides 185 oxygen partial pressure - iron aluminides 177, 203 f - nickel aluminides 115 - titanium aluminides 257 pack-aluminizing, iron aluminides 22 1 pack-cementation - process 24 - TiAl-basc matcrials 66 - titanium aluminides 44 parabolic growth, titanium aluminides 260 parabolic kinetics, titanium aluminide corrosion 333 parabolic rate constants 178, 187 passivation, NiAl 169 peening effects, titanium aluminides 3 14 pesting - intermediate temperatures 23 - molybdenum disilicides 49 - NiAI/FeAI oxidation 81 - refractory metal compounds 5 1
350
Index
petrochcinical plants 204 phase diagram - Ni-A1 system 19 - Ti-AI system 28,240 phase transformations, nickel aluminides 105, 132 phascs, titanium aluminides 317 piezo-spectroscopic effect, nickel aluminides 139 plasma spraying, titanium aluminides 44 plating, titanium aluminides 3 14 platinum implantation, Ti-48Al-2Cr 303 platinum-aluminide coatings 25 pocks intermediate tcmperatures 23 nickel aluminides 82, 92 Poisson ratio, nickel aluminidcs 138 polarization, nickel aluminidcs 163, 166 polycrystallinc NiAl oxidation 126 pores, refractory metal compounds 49 post oxidation studies, nickcl aluminides 111 pot cell, schcmatic set-up 162 preoxidation - iron aluminides 2 18, 221 TiAl-base materials 69 titanium aluminides 3 13 presulfidation, TiAl-base materials 70 process gas corrosion, iron aluminides 217 protective oxidation 23 quaternary elements, TiAl alloying 299 quenching experiments, NiAl 164 rate constants, nickel aluminides 89 reaction atmospheres, iron aluminides 177 reactivc element oxide dispersion, iron aluminides 190 reactive elements 18 - iron aluminides 198. 205 - nickel aluminides 121 f - ODS iron aluminides 183 - TiAL 300 - TiAI-basc materials 67 - void formation 21 refractory metal compounds 15,46 refractory metals 16 Rene 80 superalloy 42
residual stress, nickel aluminidcs 156 ridges, FeA1 180 Ruthcrford backscattcring (RRS) - nickcl aluminidcs 99 - TiAl 300 rutile layer, titanium aluminides 267 rutilc surfacc scale, titanium aluminides rutilc Ti,O 3 17
340
salt accelerated oxidation 329 ff, 339 salting rig 33 1 samples characterization, nickcl aluminidcs 11 1 sand blasting, TiAl-base materials 70 scale adherence 21 scale buckling model, iron aluminides 196 scale failurc, nickel aluminides 135, 140 scale formation, binary titanium aluminides 285 scale loading, nickel aluminides 138, 146 f scale modification, TiAl-base materials 62 scale morphology - nickel aluminidcs 138 - stress effccts 196 - titanium aluminide corrosion 336 scale thickncss, iron aluminides 208 scale wrinkling, nickel aluminides 155 scales, titanium aluminides 247, 3 17 scaling model, TiAl-base matcrials 6 I scandium, iron aluminides 190 scanning clcctron microscopy (SEM) 86 - p-FeAI 180 y-TiAl 290 binary titanium aluminides 275 - FA alloys 226 - Fc-doped NiAl single crystal 143 - Fccralloy 227 - iron aluminides 184 - nickel aluminides 86, 99, 110, 164 - oxide surface 125 - Ti-50 293 - titanium aluminide corrosion 333 - titanium aluminidcs 37, 60, 246, 265, 301, 314 segregation, zirconium in NiAl 127 selected area diffraction (SAD) - TiAl 265 - titanium aluminides 248
lndex servicc temperatures, titanium aluminides 239, 3 13 shear lag approximation 152 silica growth rates 15 silicide based composites 5 1 silicide coatings 50 silicides 6 - Japanese research 73 - rcfractory metal compounds 46 silicon implantation, Ti-48Al-2Cr 304 silicon, TiAl alloying 299 SIMS, titanium aluminides 315 single crack propagation 156 single-crystal NiAl 121 slurry fusion, titanium aluminidcs 44 spalling - iron aluminides 180, 207, 216, 221 - nickel aluminides 135, 144, 154 - ODS iron aluminidcs 183, 194 - oxide layer 105 - titanium aluminides 3 14, 322 spherical indentation, nickel aluminides 147 spinels 257 spraying, titanium aluminides 44 sputter ion plating 3 I4 sputtering, TiAl-base materials 44, 67 stabilizing elements 29 stages, titanium aluminidcs 245 ff stainless steel 224 stress - iron aluminides 196 - nickel aluminides 135 f subsurface layer - binary titanium aluminides 278 - titanium aluminides 247 subsurface zone, titanium aluminides 257 sulfidation 206, 21 2 - iron aluminides in H,S 214 - nickel aluminides 85 ff - Rene 80 44 sulfide nodulcs, iron aluminides 229 sulfur - void formation 21 Y-doped NiAl 130 sulfur contents, iron aluminides 185 sulfur effect model, nickcl aluminides 133 super-a, alloys, titanium aluminides 3 supcralloy Rcne 80 42 surface finish effect, titanium aluminides 3 1
351
surface morphology, NiAl 102 surface treatments, TiAl-base materials 66 surfaces - aluminized 221 sphcrical indentation 147 tantalum refractory metals 16 titanium aluminidcs 299, 329 temperature dependence - parabolic rate constants 178 - titanium aluminides 3 1, 260 temperature ranges - accclcrated oxidation 23 - alumina formation 20 - y-TiAI 289 - iron aluminides 175 - NiAIReAI oxidations 79 - nickel aluminides sulfidation 87 - orthorhombic alloys 30 - refractory metal compounds 47 - titanium aluminides 239, 245, 329 tensile load, nickel aluminides 141 tensilc strength, nickel aluminides 151 ternary elements, TiAl alloying 245,299 tcmary phases, nickel aluminides 1 12 test gases, iron aluminides 206,223 testing atmosphere, nickel aluminides 92 TG analysis, binary titanium aluminidcs 276 thermal cycling conditions, nickel aluminides 121 thermal expansion, Nb,Al-base materials 73 thermal expansion coefficients, titanium aluminids 32 1 thermodynamic activities, titanium aluminides 28,239 ff thermodynamic calculations, NiAl 169 thcrmogravimetry - y-TiAl 289 - iron aluminides 175 - nickel aluminides 110 - refractory metal compounds 53 - titanium aluminides 246, 330 Ti-A1 system alloys 239 ff titanium aluminide-matrix composites 45 titanium aluminides 3, 237-344 - Japanese research 60 - U.S. research 15, 27 -
352
lndcx
titanium implantation, Ti-48AI-2Cr 301 transient aluminas, a-alumina formation I9 transient oxidation, NiAl 132 transition oxides. nickel aluminides 99 transition phases, nickel aluminides 12 1 transition stages, titanium aluminides 258 transmission electron microscopy (TEM) - iron aluminides 184 - nickel aluminides 121 ff - Ti35A15Nb 256 - titanium aluminidcs 246, 265 tribology, ion implantation 299 trivalent nickel ions 169 tungsten - refractory mctals 16 - TiAl alloying 299 titanium aluminidcs 329 U.S. rcsearch
15 ff
vacancies, titanium aluminides 308 vacuum conditions, ion implantation 299 valence control rule, TiAl-base materials 61 vanadium, titanium aluminides 329 vapor pressure, y-TiAl 29 1 vapor transport mcchanism, titanium aluminides 341 vast protrusions, iron aluminides 234 Vickers hardness, titanium aluminides 270 voids 21 - iron aluminides 200, 209 - nickel aluminides 80, 132, 141, 155 - ODS FAS 194 - TiAI-base materials 66 - titanium aluminide coatings 3 19 - titanium aluminides 323 Wagner’s scaling model water vapor 92, 289
61
wavelength dispersive X-ray spectroscopy, TiAl 266 weight change - binary titanium aluminides 276 - FAS 191 - iron aluminidcs 186, 224 whiskers 180, 336 X-ray diffraction ( X m ) 35 binary titanium aluminidcs 275 y-TiAl 290 iron aluminidcs 177 NiAl 165 nickcl aluminidcs 79, 86, 99 TiAl 301 - titanium aluminidcs 3 15
-
yield stress, nickcl aluminidcs 152 Young modulus, nickcl aluminides 138, 153 ytterbium, iron aluminidcs 191 yttrium addition - iron aluminidcs 191 - ODSalloys 184 - reactive element 18, 121 - TiAl-base materials 64 yttrium doping, nickel aluminides 109 f, 129 yttrium implantation, Ti-48A1-2Cr 303 ytkiurn seggregation, nickel aluminides 132
Z phase formation, titanium aluminides 284 zirconium addition - iron aluminides 191, 199 - metal dusting 233 - ODS iron aluminides 184 ~- rcactivc element 121 TiAl-basc materials 63 titanium aluminides 340 zirconium doped NiAl 127 zirconium scgrcgation 132