INFORMATION TO USERS

This reproduction was made from a copy of a document sent to us for microfilming. While the most advanced technology has been used to photograph and reproduce this document, the quality of the reproduction is heavily dependent upon the quality of the material submitted.

The following explanation of techniques is provided to help clarify markings or notations which may appear on this reproduction.

1.The sign or “ target” for pages apparently lacking from the document photographed is “ Missing Page(s)” . I f it was possible to obtain the missing page(s) or section, they are spliced into the film along w ith adjacent pages. This may have necessitated cutting through an image and duplicating adjacent pages to assure complete continuity.

2. When an image on the film is obliterated with a round black mark, it is an indication of either blurred copy because of movement during exposure, duplicate copy, or copyrighted materials that should not have been filmed. For blurred pages, a good image of the page can be found in the adjacent frame. If copyrighted materials were deleted, a target note will appear listing the pages in the adjacent frame.

3. When a map, drawing or chart, etc., is part of the material being photographed, a definite method of “sectioning” the material has been followed. It is customary to begin film ing at the upper left hand comer o f a large sheet and to continue from left to right in equal sections with small overlaps. I f necessary, sectioning is continued again-beginning below the first row and continuing on until complete.

4. For illustrations that cannot be satisfactorily reproduced by xerographic means, photographic prints can be purchased at additional cost and inserted into your xerographic copy. These prints are available upon request from the Dissertations Customer Services Department.

5. Some pages in any document may have indistinct print. In all cases the best available copy has been filmed.

University M iaorilm s International 300 N. Zeeb Road Ann Arbor, Ml 48106 8305383

Raynaud, Guy-Michel

THE IN-SITU HIGH TEMPERATURE OXIDATION OF OFHC COPPER AND NICKEL

The Ohio State University PH.D. 1982

University Microfilms

International 300 N. Zeeb Road, Ann Arbor, M I 48106

Copyright 1983 by Raynaud, Guy-Michel All Rights Reserved PLEASE NOTE:

In all cases this material has been filmed in the best possible way from the available copy. Problems encountered with this document have been identified here with a check mark V .

/ 1. Glossy photographs or pages ^

2. Colored illustrations, paper or print_____

3. Photographs with dark background

4. Illustrations are poor copy______

5. Pages with black marks, not original copy__

6. Print shows through as there is text on both sides of page______

7. Indistinct, broken or small print on several pages__

8. Print exceeds margin requirements______

9. Tightly bound copy with print lost in spine______

10. Computer printout pages with indistinct print______

11. Page(s)______lacking when material received, and not available from school or author.

12. Page(s)______seem to be missing in numbering only as text follows.

13. Two pages numbered______. Text follows.

14. Curling and wrinkled pages______

15. Other______

University Microfilms International THE IN-SITU HIGH-TEMPERATURE OXIDATION

OF OFHC COPPER AND NICKEL

DISSERTATION

Presented in Partial Fulfillment of the Requirements for

the Degree Doctor of Philosophy in the Graduate

School of The Ohio State University

By

Guy-Michel Raynaud

******

The Ohio State University 1982

Reading Committee: Approved By

Professor R. A. Rapp

Professor J. P. Hirth

Professor W. A. T. Clark Department of Metallurgical Engineering ACKNOWLEDGMENTS

I would like to thank Professors W.A.T. Clark and

J.P. Hirth, Drs. A.N. Khodan and S.K. Verma , Tom

Jungling, R. Farrar and J. Ross, for their help during

this past period.

Three years ago, when I first stepped in Professor

R.A. Rapp's office, I noticed a little card perched on his

shelf from one of his previous students. It was dedicated

to the "world's best adviser." My first impression was one

of surprise but now I fully agree and if I were not afraid

it would disturb the orderliness of his office, I would

gladly give him another one.

Financial support of the research by the Department of Energy is grateful ■ acknowledged. VITA

June 20, 1957 ...... Born - Montlu^on, Allier, France

1979 ...... B.S. Electrochemistry. Ecole Nationale Superieure d'Electro- chimie et d 'Electrometallurgie de Grenoble, Isere, France

1979-1982 ...... Graduate Research Assistant, The Ohio State University. Columbus, Ohio TABLE OF CONTENTS

Page

ACKNOWLEDGMENTS ...... ii

VITA ...... iii

LIST OF T A B L E S ...... vi

LIST OF F I G U R E S ...... vii

I. INTRODUCTION AND LITERATURE SURVEY ...... 1

A. High Temperature Oxidation of . . 1 B. High Temperature Oxidation of Copper . . 13 C. High Temperature Oxidation of Nickel . . 17

II. EXPERIMENTAL EQUIPMENT, MATERIALS AND PROCEDURES ...... 25

A. Literature Survey ...... 25 B. Modification of the S E M ...... 28 C. Post Oxidation A n a l y s i s ...... 46

III. OXIDATION OF OFHC C O P P E R ...... 49

A. Introduction ...... 49 B. Specimen Preparation ...... 49 C. Results and D i s c u s s i o n ...... 55

IV. OXIDATION OF N I C K E L ...... 159

A. Introduction ...... 159 B. Experimental Procedure ...... 159 C. Results and D i s c u s s i o n ...... 163

V. IRREGULAR OXIDE GROWTH, MECHANISMS AND THE EVOLUTION OF SCALE MORPHOLOGY ...... 212

A. Introduction ...... 212 B. Morphologies and Growth Mechanisms . . . 213 C. Influence of the T e m p e r a t u r e ...... 228 D. Influence of Water V a p o r ...... 232

iv Page

VI.SUMMARY AND CONCLUSIONS...... 23 7

A. Oxidation of OFHC Copper in Pure Oxygen . 2 38 B. Oxidation of Ni c k e l ...... 240

APPENDIXES

A ...... 243

B ...... 247

C...... 252

D ...... 258

E ...... 260

LIST OF R E F E R E N C E S ...... 262

i

v LIST OF TABLES

Table Page

1. Equilibrium Partial Pressures for Vapor Species Over Cu and Cu20 at 927°C ...... 125

2. Number of Successful Collisions for Different Values of C for Modified Copper Flux (Z ) 2 ^ ^ c and for Unmodified Copper Flux (ttR BJ„xh) . . . 137 nij 3. Chemical Analysis of the 99.999% (Marz grade) Ni Polycrystalline ...... 160

4. Lattice,Grain Boundaries and Surface Diffusion Coefficients of Ni in NiO^-^, 177,178, 269 230

vi LIST OF FIGURES

Figure Page

1. Photograph Showing the Modified Stage ...... 29

2. Photograph Show'ng the Heating Stage and Its Parts ...... 30

3. Photograph Showing the Specimen, Heater, Shields, and Cylinder ...... 31

4. Heating Stage with the Cylinder, Shields, Heater and Specimen in Cross-section ...... 32

5. Calibration of the HSESEM Thermocouple . . . . 38

6. Apparatus for Monitoring and Calibrating Gas Flow into Microscope ...... 40

7. Calibration of the Flowmeter ...... 42

8. Signal Enhancement Versus Applied Potential on the Top Shield ...... 45

9. Schematic Diagram of the Hot Stage and EDAX System ...... 47

10. Temperature/Pressure Regions of Formation of

CU2 O and CuO (x x after Rapp and S h o r e s , 1^5 o o after Honjol^®) ...... 50

11. Contamination during a 25 hr Run at 950°C from In-situ EDAX Analysis Inside the HSESEM. . 53

12. Contamination Detected with an Auger Micro­ scope after Oxidation of Copper at 800°C Inside the HSESEM ...... 54

13. Oxide Film Formed on Electropolished OFHC Copper at 300°C in P_ = 3x10“^ atm 2 after 20 hours ...... 59

vi i Figure Page

14. Schematic Growth of CU2 O on OFHC Copper at 300°C and P = 3xl0-4 at m ...... 59 2 15. Oxidation of Electropolished OFHC Copper at 500°C, P = 3xl0~4 a t m ...... 60 2 2 16. Plot of -log(l-f) versus t for Three Differ­ ent Grains of OFHC Copper Oxidized at 500°C in Pn = 3xl0“4 a t m ...... 64 2 17. Oxidation of Electropolished OFHC Copper at 700°C, P = 3x 10“4 a t m ...... 65 Uo

18. Photomicrographs at 1000X showing Cu Specimen Surface when Heated in SEM to 850°C in P = 2x10-5 atm. (1,OOOX)...... U2 . . . 67

19. Oxidation of OFHC Copper at 820°C for 30 s (2, 00 0 X ) ...... 70

20. Morphology of OFHC Copper at 820°C (2,000X) . . 72

21. Oxidation of Electropolished OFHC Copper at 500°C in P = 3xl0“4 a t m ...... 73 2 22. Oxidation of Electropolished OFHC Copper at 500°C in Pn = 3x10“4 a t m ...... 77 2 23. SEM Photographs of In-situ Oxidation of OFHC Cu Oxidized at 850°C for 4.5 and 4.75 hr. in P_ = 2x10“^ atm (2, 000X) 78 2 24. Heating Curves for 3 Different Experiments on Electropolished OFHC Copper in Vacuum. Leak­ ing of O 2 Took Place at t]_, t2 and t3 , respectively. Corresponding Morphologies are S h o w n ...... 81

25. Oxidation of Cold Worked OFHC Copper at 520°C in P_ = 3xl0“4 atm (2 000X) 84 2

vi i i Figure Page

26. Copper Oxide Grown on OFHC Copper at 550°C in P = 3x 10“4 at m ...... 89 2 27. Copper Oxide Formed on OFHC Copper at 550°C in P = 3x 10"4 after 11 hrs and 20 m i n ...... 90 2 28. Copper Oxide Formed on OFHC Copper at 550°C in P = 3x10 atm at 1 day and 8 h o u r s ...... 91 2 29. Copper Oxide Formed on OFHC Copper at 550°C in P = 3x 10“4 a t m ...... 93 2 30. Copper Oxide Formed on OFHC Copper at 550°C in P q „ = 3x10-4 atm after 4 days (a) 5,000X, (bf 2, 0 0 0 X ...... 95 31. Oxide Film Grown on OFHC Copper at 920°C in P = 3x10-4 at m ...... 96 2 32. Oxide .Film Formed on OFHC Copper at 930°C in P = 3x 10“4 a t m ...... 98 2 33. Oxide Film Formed on Electropolished OFHC Copper at 950°C in Pn = 3xl0“4 atm (1,000X) . 101 2 34. Oxide Film Formed on Electropolished OFHC Copper (1000X) (a) after Oxidation at 970°C for 25 minutes; (b) after Temperature Reduced to 500°C (Pn = 3x 10“4 atm)...... 102 2 35. SEM Micrograph Showing Surface of OFHC Copper Oxided at 930°C in Pq2 = 3xl0“4 atm for 300 min (800X). Micrograph Taken after Cooling to Room T e m p e r a t u r e ...... 103

36. TEM Grid and OFHC Copper Holder after 30 min of Oxidation at 950°C in Pn = 3xl0”4 atm . . . 105 2 37. Morphological Changes with Thermal Cycles for TEM Grid Oxidized in P = 3x10"^ atm (1000X) ...... 2...... 106 Figure Page

38. Oxidation Morphology of OFHC Copper Holder at 940°C in p = 3x 10”3 atm for 2 hours after 2 Cooling to Room Temperature ...... 108

39. SEM Picture of Oxide Formed on OFHC Copper at 940°C in P = 3x 10"4 atm ...... 109 2 40. Dark Field TEM Picture of Oxide Particle Formed on OFHC Copper at 94 0°C in P = -4 2 3x10 atm...... 110

41. STEM Micrograph of Fume Particles Attached to the Substrate (500,000X) ...... 112

42. EELS Spectra of the Substrate and Particles Shown in Fig. 41...... 114

43. Microdiffraction Patterns (a) of the Particles Shown in Fig. 41; (b) of the Underlying Oxidized Substrate ...... 116

44. STEM Micrograph of the Substrate (5,000,000X) . 118

45. Oxide Film Formed on Electropolished at 1280°C after 10 Minutes 1,000X; 15 minutes 1,000X ...... 121

46. Schematic Cu20/gas Interface and Equivalent Electrical Circuit in HSESEM ...... 122

47. Geometry of the Sample and the Gas Pipe ...... 129

48. Number of Cu-02 Collisions as a Function of x : Z (x) given by Eq. (37) Using a Nondepleted Copper Flux ...... 135

49. Relative Velocity of an O 2 with Respect to a Cu(v) ...... 139

50. OFHC Copper Heated Rapidly in Hydrogen and Oxided at 920°C in PQ = 3x10"^ atm for 10s, 20s, 30s, 40s (5?000X) ...... 146

x Figure Page

51. OFHC Copper Oxidized at 920°C in Pq ^ = 3x10 ^ atm for 25 min and then Subjected to Scale Spallation by Thermal Shock ...... 149

52. OFHC Copper Oxidized at 950°C in Pq2 = 3x10 ^ atm for 35 min and then Reduced in Hydrogen at 960°C (1,000X), after 31 min of Reduction (1,OOOX), after 39 min (1,000X), after 49 min (1, 000X) 153 -4 53. OFHC Copper Oxidized in P0 2 = 3x10 atm for 15 min at 980°C (1,000X), for 20s at 920°C (2,000X), for 14 min at 920°C (1,000X) .... 156 -4 54. OFHC Copper Oxidized at 860°C in Pq2 = 3x10 atm for 50s (1,000X), 4 min (2,OOOX)...... 157

55. Specimen and Specimen Holder Arrangement used during the Oxidation of Marz Grade N i ...... 161

56. Oxidation of Electropolished 99.996% Nickel at 680°C in a Pq2 = 3xl0“ 4 a t m ...... 164

57. Oxidation of Mechanically Polished 99.999% Ni at 680°C in P0 2 = 3xl0”4 a t m ...... 168

58. Schematic Growth of NiO at 680°C in Pq _ = 3x10“ 4 a t p ...... 7 ... . 170

59. Oxidation of Mechanically Polished, Hydrogen heated Polycrystalline 99.999% Ni at 1110°C in P0 2 = 3xl0-4 a t m ...... 173

60. Oxidation Morphology of Mechanically Polished Hydrogen Heated Polycrystalline 99.999% Nickel after 6 hours at 1110°C in P0 = 3xl0"4 a t m ...... ? ...... 178

61. Oxidation of 99.999% Ni in Superdry Oxygen at 1100°C in Pqo = 3xl0“4 atm after 1 min (2,000X), 3 hours (2,000X), 5 hours (20,000X) . 179 Figure Page

62. Oxidation Morphology Observed in 99.999% Nickel after 5 hours of Oxidation at 1100°C in Wet Oxygen (P0 2 “ 3x10“^ atm) (4,720X) . . . 181

63. Oxidation Morphologies of 99.999% Ni Oxidized in Superdry Oxygen Immediately after Hydrogen Heating at 1100°C ...... 183

64. Oxidation of 99.999% Ni in Wet Oxygen Imme­ diately after Hydrogen Heating at 1100°C . . . 185

65. 99.999% Ni Heated in Hydrogen for 3 hours at 1 1 0 0 ° C ...... 186

66. Oxidation of 99.999% Nickel at 100°C in Extra Dry Oxygen P0 2 = 1x10“® a t m ...... 188

67. Oxidation of 99.999% Nickel at 1000°C in Wet Oxygen Pq^ - 1x10“^ a t m ...... 190

68. Reduction of a Ni Oxide Scale Grown at 680°C (i.e. Figure 57) and Reduced at 950°C in Dry Hydrogen ...... 192

69. (a) Thin Oxide Film Formed During Heating of 99.999% Ni at 1100°C in Vacuum (2,000X), (b) Metallic Substrate Obtained by Reduction of the Above Oxide Scale in H 2 at 1100°C for 35 min (5, 000X)...... 195

70. Oxidation 99.999% Ni at 1100°C (the Oxide Formed during Heating was Reduced by H 2 , i.e., Fig. 69) 197 -4 71. Oxidation of 99.999% Nickel in P0 2 = 3x10 atm at 1100°C (20, 00 0 X ) ...... 199

72. Oxidation 99.999% Nickel in P0 2 = 3x10“^ atm at 1350°C after 15 m i n ...... 201

73. Stereo Pairs of 99.999% Ni Oxidized for 35 min in P0 2 = 3x10"^ atm and at 1300°C 20,000X . . . 202

74. Oxidation of 99.999% Nickel Single at 1100°C in Pq 2 = 3xl0“4 atm (Superdry)...... 204 Figure Page

75. Oxidation Morphologies Obtained when a 99.99% Nickel Grid is Oxidized in Superdry Oxygen at 1000°C in Pc>2 = 3 x 1 0 “^ at m ...... 208

76. Oxidation Morphologies Obtained when a 99.99% Nickel Grid is Oxidized in Superdry Oxygen at 1000°C in Pq2 = 3xl0“4 a t m ...... 210

77. Schematic Model for Growth of Surface Oxide by Screw (after V e r m a ^ ) ...... 216

78. Schematic Model for Whisker Growth (Low Temperature) and Pyramid Growth (High Temper­ ature) ...... 218

79. Model of Whisker Growth during Oxidation of Iron, 400°C, in dry 02* From Tallman and Gulbransen250 ...... 220

80. Spiral Arrangement of Atomic Planes Near a Screw Dislocation of Burgers Vector b, |b| = 4C0 , with the Core Removed. One of the Four Connected Surfaces is Shaded from Alexander et al.263 225

81. Process of Formation of a Dislocation Having a Large Burgers Vector on the Surface. From Gotoh and Komatsu2$5 ...... 226

82. Ratio of Free Energy Function for M 2 X and MX as a Function of Temperature from Ref. 214 and 2 1 5 ...... 249 I . INTRODUCTION AND LITERATURE SURVEY

A. High Temperature Oxidation of Metals

Since the classical work of Tamman/ the subject of the high temperature oxidation of metals has been the subject of numerous research efforts. Much of this work 2—7 8—17 has been reviewed in a number of books and articles.

Although the overall reaction of a with oxygen written as:

XM + ? 0o ■> M 0 (1) 2 2 x y appears simple, a more detailed analysis shows that, in most cases the progress of the reaction involves many phenomena, i.e. phase boundary reactions and mass and charge diffusion processes.

1. Initial Stages of Oxidation

The first step in the oxidation of metals involves the adsorption of the oxygen on the metal surface. This is usually considered as a chemisorption followed by a process of dissociation and partial ionization of the oxygen. However, the discovery of MO2 with

M = (Ni,Pd,Pt,C u ) '19 SUggests that oxygen-metal complexes may be involved in the very initial stages. For example, 20 the oxidation of cadmium gives CdC^ rather than CdO 2 and there is some evidence for nondissociative chemi- 21 sorption m the oxidation of thallium.

As oxygen is added to the metal surface, 2D and 3D nuclei are formed. Modern surface analysis (L.E.E.D,

H.E.E.D., A.E.S. ESCA and work function analysis, have helped to distinguish the 3D nuclei which start after the adsorbed layer (2D nucleation) covers the entire metal surface. Since the first observations of isolated 22 oxide nuclei on iron at 850°C by Bardolle and Benard, the influence of temperature, pressure, crystalline orientation and have been extensively 4 23_26 studied. ' However, few experiments give information about high temperature adsorption. In addition, suf­ ficiently clean surfaces are seldom attained to obtain 27 28 meaningful data. ' Finally, the structure of the adsorbed layer is a strong function of the temperature.

The origin of the 2D nuclei for reaction product compounds is commonly the dislocation emerging from the metal base (for example see reference 29 for the nucleation of sulfide or ). The nucleation of the 3D or bulk compounds has been related to the crystalline defects of the 2D or adsorbed layer. These defects favor metal diffusion. For example, the impingement of several 2D islands produce such defects on the simple c(2x2) struc­ ture formed on the {100} face of an f.c.c. or b.c.c. 3

30 metal. Consequently, the defect density m the 2D

adsorbed layer increases with the supersaturation,

thereby increasing the 3D compound nucleation. For high

temperature and low supersaturations the 2d layer has time

to recrystallize before 3D nucleation and the sites for 24 bulk compound nucleation are the metal dislocations. in

1956, Cabrera already suggested that oxide nuclei should 31 form at points where dislocations intersect. However, 11 23 32 experimental studies ' ' showed no definite correla­

tions between nuclei and dislocations except for the

formation of many nuclei where stacking faults intersect 33 -3 -4 the surface on copper oxidized at 525°C in Pn ^10 -10 2 Rather, steps and terraces, often associated with 34 stacking faults, provide preferential nucleation 3 5 36 sites. ' Small angle grain boundaries have been shown

to act as heterogeneous nucleation sites during the 37 oxidation of copper at 950°C.

2. Nucleus Growth

After the nucleation stages, the nuclei start to 3 8 grow and Neugebauer has described the different stages of film growth. A number of models have been developed 39 for growth kinetics of the nuclei. Orr divides the process into three successive steps: adsorption of oxygen, surface diffusion of oxygen and finally capture of oxygen by the growing nucleus. Different growth laws are expected depending on which one of these three

processes is the rate limiting step. These processes stop

when nuclei grow together to form a continuous film. 40 41 Experimental studies on copper oxidation ' showed a

complicated behavior, but generally a parabolic rate law

was found corresponding to surface diffusion as the rate 37 42 limiting step. However, Benard et al. proposed that

at high temperatures and reduced oxygen pressure, oxygen

adsorption on the metal surface is the rate limiting step

since a linear rate law is found experimentally. Whereas

the previous models assumed that the underlying substrate

for surface diffusion was a ready source of metal , 43 Rhead, developed a model whereby outside the oxide nu­

clei a thin layer intermediate between a chemisorbed layer

and an oxide layer exists. The surface diffusion of metal 44 atoms is significant above 500°C and is enhanced by the 4 3 4 5 presence of a chemisorbed layer. ' According to this model, the area of the particles increases linearly with 2 time. Using a statistical method, Evans showed that the

fraction f of total surface area covered at time t for a constant nucleation density is

f = 1-exp (—najv^t^) (2)

where id is the density of nuclei and v is the radial

growth velocity. 46 This was experimentally verified for tin oxide and 47 aluminum oxide. In the latter case, distinction was

made between the growth rate of crystalline Al^O^, and the

amorphous film between nuclei.

Examination of growing films in the electron micro­

scope has also shown that recrystallization by grain 48 boundary migration occurs during growth. Coalescence

of metallic clusters to form a new island occupying an

area smaller than the sum of the original two, exposes

some bare surface where "secondary" nucleation occurs.

The growth of thin oxide film is not well documented but

growth of metallic clusters is possibly applicable to oxide

growth. During film growth, dislocations can be formed by

several different mechanisms:

- When two islands impinge, whose lattices are 4 9 rotated, they form a subboundary.

- Because the substrate and the film have different 50 lattice parameters, there will be misfit dislocations.

- Stresses can generate dislocations at the edges

of holes.

- Dislocations from the substrate can continue in

the oxide film.

It is known, that the screw component of a dislocation can provide steps for further growth at low supersatura- 52 53 tion. Very recently Bauser and Strunk showed that edge dislocations can act as persistent sources of

monomolecular steps for crystal growth.

3. Thin Film Growth

As soon as a thin continuous oxide film has formed

on a metal surface, the reactants are separated by a

barrier and the reaction proceeds by the transport of

cations, anions, or both and electrons through the oxide.

A large number of theories have been proposed for the th.in

oxide film formation. The logarithmic type rate equation

has been explained by not less than five rate limiting

mechanism. 54 55 -Quantum mechanical tunnelling of electrons ' 8 56 -Cation diffusion in high space charge fields '

-Space charge^ 5 8 -Place exchange

-Deactivation of the surface for chemisorption with

increasing thickness.

Experimentally, most metals indeed grow with a

logarithmic rate law in the thin film range. However,

the formation of thin oxide films has been shown to be

influenced greatly by the metal orientation, surface

pretreatment, and the oxidation procedure.^

The transition between a thin film and a thick film 61 has been studied by Wagner who showed that no simple rate laws exist between the limiting cases of the transport of ions across a thin oxide film by virtue of

an electrical field and of the diffusion of ions and

electrons across a thick oxide layer obeying the classi­

cal diffusion equations.

4. Thick Oxide Growth

a. Lattice diffusion 62 63 The theory of Wagner ' covers the ideal case where

transport through the scale is limited by diffusional

transport in the lattice only. This behavior leads to

the parabolic rate law in terms of the electrical con­ ductivity and the electrical transference numbers of the migrating ions and the electrons or also by the self­ diffusion coefficients of the ions (for an electronically conducting scale). A study of the point defects in oxide

scale is necessary to confirm the validity of Wagner's

theory. For example, if the predominant defects in an

MO oxide are singly charged metal vacancies, the electrical conductivity and the metal self-diffusion coefficient vary 1/4 as P . Deviation from the predicted oxygen pressure 2 dependency may arise from nonideal diffusion behavior or from a complex defect structure. For example, clusters of defects in FeO are known to be highly stable.in addition, the defect structure may change across the oxide scale.^ Finally, the effect of impurities on electrical conductivity measurements is tremendous (doping effect).^ In conclusion, our knowledge of defect structures in oxides is often not sufficient to distinguish between a nonideal behavior caused by complex defect structures or else by nonlattice diffusion,

b. "Short-Circuit" Diffusion

It is well known that nonlattice transport may play 70 an important role during oxide scale growth. Particularly, diffusion in oxide scales has a lower 70 activation energy than lattice diffusion. Consequently, grain boundary diffusion becomes more important at lower temperatures or for smaller grain sizes. In the scale growth models, the overall transport takes place both 71-75 along the grain boundaries and through the lattice.

An effective diffusion coefficient can then be described by:

D = D. (1-f) + D, f (3) ef f S b where and are the diffusion coefficients for lattice and grain boundary diffusion, respectively, f is the fraction of diffusion sites in the boundaries. The use of'Eq. (3) leads to a modified parabolic rate law and an effective rate constant. More generally, the influence of dislocations during scale growth is a continuing subject of interest. It was already mentioned that screw and 52 53 76 edge dislocations provide growth steps for the oxide. ’ ' '

The importance of surface diffusion along the pyramid 7 8 steps has now been modelled. Equivalently dislocations provide growth paths for blades or whiskers formed on 79 oxide scales. Voss et al. concluded that the primary

mechanism for hematite blade growth is surface diffusion

along a central tunnel. It is very likely that the tunnel

is centered on one dislocation or a bundle of parallel 8 0 dislocations. Frank showed that a dislocation whose

Burgers vectors exceeds a critical value is in equilibrium

with an empty tube at its core. The rate of surface

diffusion along the surface of the tunnel is lowered with

time as the surface area available for Fe transport

decreases by narrowing of the tunnel radius. Ultimately

the tunnel is sealed and the blade stops growing.

5. Faulted Scales

If the oxide scale grows by outward migration of metal ions, voids will develop due t6 the condensation of 6 7 vacancies at the scale/metal interface. ' Most of the

potential void volume created by the outward migration of

cations is filled by the plastic flow of the scale which 81 maintains contact at the metal/oxide interface.

Furthermore, some vacancies precipitate at grain bound- 82 — 85 aries in the . Other factors may also influence 8 6 cavity formation. For example, Caplan et al. showed

that carbon in nickel promoted cavity formation at the metal grain boundaries during oxidation. For chromium oxidation, significant wrinkling of the scale has been 87 — 89 observed above the grain boundaries of the metal. 10

One interpretation of this phenomenon is that the scale grows by counterdiffusion of metal and oxygen and that new oxide is formed within the scale. When voids are formed at the metal/scale interface, the process of outward dif­ fusion of the metal is perturbed. Diffusion of the reactants may be maintained where the scale is still attached to the metal but three different kinds of trans­ port can be created across the voids.

a. Metal Evaporation

At high temperatures, the vapor pressure of some metals (chromium, for example) is high enough to supply the 87 88 scale growth above the void. ' The voids may be filled with gases used during the preoxidation treatment, for 8 9 example, hydrogen or gases resulting from impurities m 90 the metal (CO and CO2 )• In the case of chromium oxida­ tion, the metal vapor pressure is so high that the rate limiting step is still the diffusion through the scale above the void.

b. Dissociation Mechanism

As the metal ions move outward, the chemical potential of oxygen at the void/scale interface is increased. Conse­ quently, dissociation of the oxide can occur and oxygen is liberated within the void. A new internal scale is formed 6 7 83 at the metal/void interface. ' ' This mechanism is responsible for the formation of a fine grained and porous , 7,91-97 inner scale. 11

c. Development of Microchannels

The model for development of microchannels has been 7 developed by Mrowec et al. During scale growth, oxide grain boundaries serve as preferential sites for vacancy condensation; channels are therefore created along the oxide grain boundaries. Eventually, the grain boundary fissure ultimately reaches the surface to permit the inward migration of oxygen molecules. Consequently, the oxidation may take place directly at the metal/scale 9 8 interface.

6. Stresses in Oxide Scales

The creation of dislocations in thin oxide films has already been discussed and more generally, stress gener- 84 ation has been reviewed by Hancock and Hurst and 99 Stringer. When the oxide scale grows at the metal/oxide interface (for example, by inward diffusion of oxygen), compressive stresses are established in the scale.

Pilling and Bedworth'*'0^ related the sign and magnitude of the stresses in the scale to the ratio of the molar volume of the oxide to that for the metal (P.B.R.). In the case of scales growing by outward metal migration, stresses arise as a result of adherence at the metal/oxide inter­ face.^'"*"^ Stresses may also arise if the oxide would grow by counterdiffusion of metal and oxygen, such that new oxide is formed within the scale. However, 12

109 Speight and Harris demonstrated the impossibility of

concurrent diffusion of both anions and cation point

defects. The problem has been resolved by Atkinson'*'^ who

showed that new oxide is formed within the scale only if

the defects are metal and oxygen interstitials or charged

vacancies and uncharged interstitials. Finally stresses

arise during thermal cycling since the linear coefficients

of expansion are often very different for a metal and its . , 84 oxide.

It has already been mentioned that formation of voids at the metal/scale interface can account for stress relief during the oxidation of a metal. Cracking of the scale constitutes another way to relieve stresses. For example, during the oxidation of chromium, successive layers of compact oxide layers are formed, the critical thickness 105 being the thickness at which cracking occurs. Break­ down of a protective film often leads to a linear rate law.^"^ As mentioned above plastic deformation, i.e., dislocation glide, dislocation climb, grain boundary sliding, mechanical twinning, or more generally high, temperature creep, will relieve stresses in the oxide scale.

It is clear from the previous general literature review that metals behave differently when exposed to an oxidant gas at high temperature. There are no general rules by which the oxidation behavior of a metal can be pre­ dicted from its atomic and physical properties, for example. 13

Consequently, a literature review of high temperature oxidation for copper and nickel is necessary.

B. High Temperature Oxidation of Copper

This review is restricted to the formation of CU2 O on

Cu since our experimental conditions guarantee the exclusive formation of only one oxide, CU2 O.

Many investigators have studied the reaction of copper with oxygen at high temperatures and most of the results, obtained before the seventies, are conveniently 4 14 114 summarized in three reviews. ' ' Different purity and specimen preparation were used but there has been a tendency to use electropolished OFHC copper. The adsorp­ tion and nucleation of cuprous oxide has already been mentioned. In addition, most of the adsorption studies were made at room temperature or lower and consequently are not of interest here. On the contrary, nucleation of copper oxide at elevated temperature was extensively studied (see, for example, 42) and oxidation rate studies of single crystals of copper showed that surface orienta­ tion plays an important role in determining oxidation behavior. There was good agreement between the different authors on the order of decreasing oxidation rate, i.e.,

{100} > {110} > {111} > {311} at temperature below 500°C.

However, this order can be changed when the oxidation temperature is higher than 500°C. The orientation of the 14 oxide films has also been examined in great detail. The degree of scale/metal epitaxy varies with P , temperature °2 and film thickness, but the orientation relationships between the oxide and the base metal are always the same.

A number of different rate laws have been reported for the oxidation of copper. At low temperatures (T < 200°C) the direct and inverse logarithmic laws have been observed.

The cubic law predicted by Mott, Cabrera and Hauffe for intermediate temperature (170°C < T < 720°C) was seldom observed. On the contrary, it is generally agreed that the growth of cuprous oxide occurs by outward diffusion of the metal in the high temperature range where only parabolic behavior has been reported. However, in two studies at 1000°C, it was concluded that the reaction at the CU2 O/O2 interface may compete with diffu­ sion through the oxide layer in the rate control at short times. Irregular scale (i.e., variable scale thickness) was already observed before the seventies. At low tempera­ ture (300°C) Halliday and Hirst^^ revealed the presence of pronounced mounds on the oxide surfaces. The observa­ tion of whiskers have been particularly numerous during the oxidation of copper and it is commonly agreed that they consist of CuO, grow along the <110> direction and contain 118 an axial screw dislocation. Hardy suggested that dislocations ending in the metal surface served as nucleation points; the dislocation would be continued into the oxide layer, where it would form a preferred diffusion path and an oxide whisker could grow. Appleby and 8 5 Tylecotte showed more generally that the formation of whiskers of CuO is a consequence of stresses arising during 119 the oxidation process. In addition, Pfefferkorn showed that CuO whiskers or platelets grow at their tips, and 120 Jaenicke and Albert concluded that whisker growth occurs by surface diffusion with a parabolic rate law. 121 However, Gulbransen et al. found a cubic rate law between 250 and 450°C and showed that whiskers have no appreciable effects on the kinetics of the overall oxida­ tion process.

Since the seventies, emphasis in the field of initial oxidation has been put on the oxidation of single crystals -3 at very low Pn (< 5*10 torr) and intermediate tempera- 2 122—12 9 ture (< 700°C). The new techniques used, T.E.M.,

L.E.E.D., R.H.E.E.D., in situ U.H.V.-R.H.E.E.D. and medium energy electron diffraction guarantee ultra-high vacuum and permit reproducible results. Since these experimental conditions are far from those of the current study, there is very little need to expand on the new results obtained.

However, from these studies, it is very clear that the nucleus morphology is very strongly influenced by pn . The 2 formation of individual islands containing misfit disloca­ tions is frequently preceded by the growth of an oxide 16

128 superlattice up to several monolayers thick. However, this has not been confirmed by medium energy electron diffraction where the copper surfaces appear to be 120 unoxidized between the oxide crystallites.

Whereas the ideas on the adsorption and nucleation phenomena before the seventies seem to be confirmed by the modern techniques, the older, simple ionic defect models are currently criticized.

As discussed above, cuprous oxide is believed to be a p-type containing metal vacancies as pre­ vailing defects^^ but Tretyakov et al.^"*" have reported a rather large oxygen deficiency at low oxygen activities and 1000°c. Measurements of deviation from stoichiometry,^'*' electrical conductivity of cuprous . , 134, 136-138 . ,* .... . * 139 ... . oxide a n d self-diffusion of copper indicate the presence of neutral vacancies as predominate 14 0 defects. On tljie contrary, Wagner favored a model with singly charged vacancies and electron holes as prevailing defects. A more complex model considers also the presence of oxygen vacancies and their complexes with charged . . .141 metal vacancies. 142 Maluenda showed that the electrical conductivity of Cu~0 varies as '*'//n with n=8 for P close to 10 ^ 2 2 2 atm and n=6 for the higher P corresponding to Cu„0/Cu0 °2 equilibrium. This variation of electrical conductivity was modeled by an ideal solution including the new charged 17 species (V^u • Finally, the behavior of the electrical conductivity at low oxygen activities shows deviation from that observed at higher a values . 2

C. High Temperature Oxidation of Nickel

There have been many studies of both high- and low- temperature oxidation of nickel, because of the use of nickel-base alloys and because the system Ni-NiO is rela­ tively simple with only one stable oxide. Much of this work is concerned with the growth of thick oxide scales and the observed deviations from Wagner's classical theory.

However, the modern techniques of surface science 143-147 148 A.E.S., L.E.E.D., R.H.E.E.D. and X-ray 149 150 151-155 fluorescence spectroscopy, ' , X.P.S. and UPS „156, 157 _ , T _ 158,159 , ISS, Secondary Ion Mass Spectroscopy and 143 154 work function change ' are currently supplying data on initial oxidation of nickel. Many studies have been performed at room temperature following oxidation but more recently, emphasis has been put to the high temperature range (see, for example, 147) . However, there do not seem to be great differences between the low and high temperature regime. For example, the {100} face of a nickel undergoes the same chemisorption of oxygen in two steps (p(2x2) structure followed by c(2x2) structure) at any temperature. The chemisorption structure for other faces is generally more 18 complicated. Within the chemisorbed film there appears to exist a number of nucleation sites dependent on temperature and step and dislocation density.

There is a delay before nucleation whose duration depends greatly on the surface control. In addition, the epitaxial relationships between Ni and NiO vary with temperature, , ... . 63, 143,148-150,161 oxygen pressure, and oxide thickness. '

The oxygen dissolution in the metal and then the length of the induction period is also a function of crystallographic orientation. The lateral growth of NiO is by 14 9 oxygen capture at the periphery of the nucleation sites.

In some instances, at temperatures below 200°C NiO does not thicken after the oxide islands are approximately three 149 151 154 oxygen planes thick. ' ' On the contrary, at temperature above 200°C, the film thickens with additional oxygen exposure. At temperature below 600°C, the reaction kinetics are logarithmic^"^ or parabolic"*"*^ with time. Varying epitaxial relationships between nickel and ^11, 75, 149,150, 161, 164,165 . , t .. NiO may lead to changes m the oxide microstructure and its kinetic behavior. In addi­ tion, there is a strong dependence of oxide growth upon crystalline orientation,^^ the rate of NiO growth for Ni decreases in the sequence {110} > {111} > {100} >

{113} > {115}. This could be due to the induction period in the initial stages as mentioned above, although

Graham^ observes increasing differences during the later 19

stages. Since NiO has the NaCl structure, lattice diffu­

sion and a lattice diffusion controlled growth rate should be isotropic. Consequently it is commonly advanced that

the rate of oxidation is controlled by transport along defects in the oxide (dislocations or grain boundaries

Mutual diffusion of oxygen and nickel through the oxide 107 scale has also been proposed for NiO growth. However, nickel self-diffusion in NiO is faster than that for oxygen by about two orders of magnitude at the same temper- 130 ature and . At high temperature large differences

• 4. w 4- 2 4-k • j . • . . . 91, 92, 95-97, 166-172 exist between the oxidation reaction rates.

Above 1200°C, it is generally agreed that the nickel vacancies have an average effective charge of one and that lattice diffusion is the important transport 92, 96, 97, 166 . 173 . , mechanism. Atkinson and Taylor have measured the nickel tracer lattice diffusion coefficient in single crystalline NiO at temperatures from 522 to 1400°C to give

Dn^ = 2.210 ^ exp (-24 7, 000 (Joules)/RT)cm^sec ^ (4)

The activation energy for the parabolic rate constant is accordingly 24 7 kJ/mole-°K“l. However, the observed activation energies range from 145 to 170 kj/mole in the temperature range 500-1100°C (see, for example, ref.

174). This changing activation energy could result from either a transport via short-circuit diffusion or a gradual change in the lattice defect structure with temperature.

Several effects have been studied to solve the problem. 96 - Effect of Cold Work. Caplan et al. showed that

the oxide scale formed on cold worked nickel was finer

grained than on annealed nickel and grew faster initially.

Such growth is interpreted as controlled by diffusional

transport by both lattice diffusion and along easy diffu­

sion paths. The more rapid initial oxidation of the cold

worked metal results from diffusion along additional easy 96,97 diffusion paths.

- Effect of Metal Orientation. As mentioned above, 74 Herchl et al. oxidized differently oriented nickel

single crystals in 400 torr Pn at 500-800°C. 2 Graham et al. ' observed different decreasing oxida­

tion rates, i.e., {100} > {111} > {112}. Again the oxida­

tion rates were related to the density of short-circuit diffusion paths. Oxidation rates are also related to different surface preparations that change the density of 176 easy diffusion paths. 9 8 63 - Tracer Studies. Atkinson et al, using Ni 18 tracers and o tracers showed that whereas lattice diffu­ sion of nickel vacancies prevails at 1000°C, nickel is predominantly transported along easy diffusion paths at lower temperatures (500-800°C). During the initial oxidation of polycrystalline nickel at 600 and 1000°C, there was no inward oxygen penetration in the scale. But 1 8 for thicker scales (> lp at 600°C and 9y at 1000°C) 0

was found to penetrate to the metal/scale interface,

developing a duplex scale structure consisting of a porous

inner layer with fine grained oxide and an outer layer

with larger columnar grains.

It is now commonly agreed that deviation from parabolic

NiO growth within the intermediate temperature range

results from transport over easy diffusion paths. Using 63 177178 the tracer Ni, Atkinson and Taylor ' were able to distinguish between the contributions from dislocation diffusion in low angle boundary arrays and from diffusion

along high angle grain boundaries. The diffusivity for

nickel along a dislocation at 1 atm P in the temperature 2 range 522-800°C is

D,. . =0.26 exp (-192000/RT) cm^ sec ^ (5) disl.

_ s Assuming a boundary width of 7x10 cm, the value of the diffusion coefficient for grain boundary diffusion at

1 atm P and 500-800°C is 2 2 -1 D , =0.43 exp (-171000/RT) cm sec (6) CJ • D • ’

At 500°C Dg k f°ur orders of magnitude greater than that for lattice diffusion. Most of the experimental data between 300 and 1000°C confirm an activation energy, given by Eq. (6). Above 650°C, grain growth takes place 22

and lowers the parabolic rate constant. Graham

et al.16^'176 found that the parabolic rate constant was 1/6 proportional to P ' and doubly charged vacancies were Op 69 the predominant defects. Atkinson and Taylor found

this PQ dependence for nickel diffusion in dislocations

and along grain boundaries. Consequently the defects at

the grain boundaries seem to be the same as in the bulk if

dislocation-point defects interactions are neglected.

The model for grain boundary diffusion presented above (3)

can be successfully applied to nickel oxidation at 500-

800°C.75

As mentioned above the NiO scale becomes double layered

and consists of an outer compact layer and an inner porous

layer.91,92,95,98,108,166"172 The dissociation mechanism

and the development of microchannels have been advanced

to explain the formation of such duplex scales. It was

already mentioned that carbon dissolved in nickel results 8 6 in the formation of cavities. Again scale adhesion is maintained by the deformation of the oxide. Hales showed

that at 1000°C such deformation occurs by the glide of a 17 9 <110> dislocations on {110} planes.

Irregular scale growth has been observed during NiO 180 growth at the scale/gas interface. Armanet et al. observed a very flat oxide scale when oxygen was carefully dried at 750°C-1250°C. On the contrary, in wet oxygen faceting occurred. Furthermore, the scale/gas interface was covered with whiskers when cold worked nickel was oxidized in air. Statement of Purpose

In the past fifteen years, the introduction and improvement of surface observation and analysis equipment has led to great advances in our understanding of the oxidation (sulfidation, etc.) of metals. A number of previously accepted mechanistic beliefs have been disproven and replaced by more sophisticated models.

Nevertheless, important fundamental questions still exist, for which the new, excellent, post-oxidation analysis equipment is not ideally suited. After the cool­ ing of any specimen from the reaction temperature, one can only hope to obtain a structural, morphological, and chemi­ cal analysis of one steady-state view in the evolution of the products, and even this picture may be clouded by mechanical or chemical changes induced by the cooling itself. In particular, one loses the opportunity to view the evolution of morphology prior to the attainment of a given steady-state condition. In the research described here, a scanning electron microscope (SEM) was adapted for inclusion of a hot-stage and an environmental chamber, so that the evolution of morphologies of product phases at the scale/gase interface could be studied "in situ."

Specifically, the oxidation of pure copper and nickel were studied over a range of temperature.

24 II. EXPERIMENTAL EQUIPMENT, MATERIALS

AND PROCEDURES

A. Literature Survey

In April 1969, a meeting entitled "The Use of the SEM

in Corrosion Science" was organized by J.E. Castle and held

at the Central Electricity Generating Board in Leather-

head, England. The following passage is quoted from

Castle's summary of the Proceedings:

Many workers with access to an SEM were sur­ prisingly hesitant about tieing the machine up for several days so that metal oxidation can be carried out in situ. This is however such a power­ ful method that the author feels that the instru­ ment will be used extensively for the purpose. A second SEM at CERL is now being prepared for long­ term studies of this type of metal oxidation."

While a multitude of authors have used the SEM for post-oxidation examination of corrosion products, only three publications have used a Hot Stage Environmental

Scanning Electron Microscope (HSESEM) for in-situ oxida- 18 2“18 4 tion. There have been three previous designs of a 185-188 hot stage for SEM, but none of these seemed ade­ quate for oxidation studies. However, these authors described a number of factors which are important for any hot-stage design. In particular, nonmagnetic materials were recommended as materials of construction, and a clean DC power supply is required because any ripples

25 26 present in the power supply would produce an alternating magnetic field to disturb the electron beam. Degassing of the specimen stage at high temperature was found to be 183 188 a problem. ' Mostly the use of in the construction of the hot stage contributed to degassing at high temperature. The correct positioning of a thermo­ couple was found to be important in all cases. 183 The paper by Castle and Hunt provided details for the solution of experimental and interpretation problems.

For example, a calculation shows that the gas jet can establish an effective pressure at the specimen surface which is about 3000 times higher than the average background pressure in the SEM chamber. In this way, i.e., by using a capillary leak to the specimen surface, the chamber -4 pressure can be maintained at and below the 10 torr above which corona discharges (gas ionization) would ruin the

SEM image, while the oxidizing atmosphere at the analyzed spot can be held at about 10 ^ torr which is sufficient for the study of surface oxidation. Although heat from the hot stage caused the rapid deterioration and subsequent removal of the Al films covering the scintillator, a replacement by an evaporated carbon film proved completely successful. In view of this problem, the use of radiation shields around the specimen heater to reduce the heat loss and protect the secondary electron detection systems from 27

is discussed later.

In this first "in-situ" oxidation study, Castle and

Hunt observed the oxidation of a polycrystalline -5 hypoeutectoid steel at 3X10 atm C>2 and 360°C. The dif­

fering scaling behavior of the three phases (alpha, Fe^C,

and MnS) became obvious at about 50,000X magnification. -4 The oxidation of an iron single crystal at 10 atm C>2 and

350, 400 and 450°C showed the nucleation of Fe2C>2 crystal­

lites on the initial oxide film and initiation of cracking

in the magnetite layer. Obviously, this promising initial

study leaves dozens of interesting and controversial

mechanistic aspects of oxidation open to study and

resolution.

The need to use temperatures greater than 450°C should

be easily satisfied by construction of a radation-shielded 185 186 furnace resembling that of Fulrath, et al. ' They

used a biased grid between the detector and the heater to minimize the noise in the picture introduced by thermally excited electrons emitted from the sample or the heater. 183 With this hot stage the authors studied the sintering of many materials (in vacuum) at temperature up to 1750°C and magnifications of 5000X. In fact, a mixture of the two designs should permit entry of the environment with the use of a 1000°C stage without radiation damage to the scintillator. 28

With these factors in mind, the specimen stage of a

Cambridge Stereoscan S 4 SEM was modified for high temper­

ature studies in the present work.

B. Modification of the SEM

1. Specimen Stage

The original SEM specimen holder plate was replaced by a type-304 stainless steel plate of 0.6 cm thickness which supported the cylinder (3.17 cm ID, 3.81 cm OD) containing the specimen, heater, and the radiation shields.

The whole assembly is shown in detail in Figures 1-4. In the case of copper, the specimen consisted of a 1 cm dia.,

3 cm tall cylinder. In the case of nickel, the specimen

(5 mm dia, 2 mm thick) was mounted on a specimen holder made from nickel rod.

The specimen was surrounded by radiation shields made of molybdenum sheets of 0.04 cm. thickness. These two or j three shields (shown in Fig. 3) of diameters 1.27, 1.90 and 2.54 cm were insulated from each other at their bases by rings cut from alumina tubes. These shields also have similar slip-over ceramic covers which are posi­ tioned after the specimen is placed in the shield assembly.

Molybdenum was chosen for its low emissivity even at high 18 9 temperature (see section on Heater and Power Supply).

Nickel and platinum shields were also tried but showed a catastrophic increase in 29

* I H

Figure 1. Photograph Showing the Modified Stage, st - Specimen stage mv - Microvalve G - Gas pipe R - Radiation Shield (R1 for top and R" for bottom shield E - Ground wire of the stage T - Thermocouple wire H - Heater wire or power connection C - Cylinder 30

Figure 2. Photograph Showing the Heating Stage and Its Parts. C - Cylinder C' - Ceramic ring E - Ground wire R - Radiation shields S' - Specimen 31

Figure 3. Photograph Showing the Specimen, Heater, Shields, and Cylinder. E - Ground wire C' - Ceramic ring for specimen insulation H - Heater wire R" - Radiation shield for bottom R' - Radiation shield for top R - Radiation shields Molybdenum Shields Double Tungsten Wire

Thermocouple

□ Ceramic

Sample

Figure 4. Heating Stage with the Cylinder, Shields, Heater and Specimen in tJ Cross-section. to temperature. However, during very high temperature experi­

ments, radiation shielding was improved by using an outer­

most aluminum oxide-strengthened silver radiation shield

having a very low emissivity at all temperatures. The

heater wire was a tungsten wire of 0.05 cm of diameter which

was insulated from the specimen by fine ceramic alumina

tube segments as shown in Figure 3. The heater coil was

doubly wound in order to minimize any inducing electro­

magnetic fields and to improve its rigidity. Depending on

the size of the specimen and the desired temperature, the

heater wire was positioned either outside or inside the

cylindrical specimen (holder). In the latter case, the

emission of thermally excited electrons collected by the

scintillator and the power input to achieve a given temper­

ature were greatly reduced.

The specimen or holder cylinder was insulated and

electrically isolated at its top and bottom by thin slices

of alumina tube. The specimen rested on a molybdenum

circular disc supported by a spring within the first radi­

ation shield. The initial vertical positioning of the

specimen was fixed by the adjustment of the spring. The

specimen and the three radiation shields contained within

the stainless cylinder were mounted by a set of screws positioned at the bottom of the cylinder shown in Figure

3. A Pt/Pt-Rh thermocouple was inserted into the hole at the bottom of the specimen and was mounted in such a way that the tip of the thermocouple covered with a thin layer of alumina, touched the specimen near the oxidizing sur­ face. The solubility between nickel and platinum 190 could influence the temperature reading. To minimize the heat loss from the top and bottom of the assembly, additional radiation shields were used. Only one Mo shield of 0.04 cm thickness was used at the bottom with three shields at the top. One of the top shields was mounted on the cold part of the stage. One of these shields was held at a variable bias with respect to an electrical ground to minimize the noise resulting from thermally excited electrons.

The X-ray detector (EDAX 711) must be in line-of-sight of the sample, so the top Mo shields contained holes to allow X-rays to reach the detector without obstruction.

The detector was provided with another shield of beryllium or aluminum over the Be-window to keep the detector cool and to avoid contamination by oil or metal vapors.

The use of a liquid nitrogen trap between the diffu­ sion pump and the specimen chamber has proved to minimize contamination of the specimen. In particular, post­ oxidation Auger analysis excluded the contamination of the specimen by carbon, and in-situ X-ray mapping showed no contamination by Mo or Ag from the radiation shields. 35

The oxidizing gas was leaked to the top surface of

the sample by a 0.10 cm ID 304 stainless steel or platinum

pipe. Helical bellows in the gas pipe and a setscrew on

the cylinder (shown in Fig. 1) helped to accommodate the

movement of the stage in x-y-z direction without changing

the position of the gas pipe on the surface.

As shown in Figure 2, the ceramic ring positioned at

the top of the sample was used to fix an electrical ground

wire to the specimen. This grounding was very important

to drain any electrostatic charge built up on the surface, 19 without which good resolution could not be obtained.

2. Heater and Power Supply

The heater was tungsten wire of 0.05 cm diameter and was double-wound around the cylindrical sample to minimize

any magnetic field effects. A diameter of 0.05 cm was

found to be suitable from preliminary calculations of the power requirements for the furnace. A calculated current requirement which considers reflections by two Mo radia­ tion shields is given by

4 4 irae (Ts -T. ) (7)

where

D = Tungsten wire diameter

o = Stefan-Boltzmann constant 36

eQ = Emissivity of molybdenum

Tg = Temperature of the specimen

T^ = Temperature of the outermost stainless steel cylinder

A2,A1 = Surface area of the first and second shield, respectively

A£ = Surface area of the cylindrical specimen or specimen holder.

= Surface area of the outside cylinder. 189 192 p = Tungsten resistivity ' at 20°C 0 (5.6xl0-8 fim)

1 = Length of the heating wire 192 -3 -1 a = Resistivity coefficient (4.5 x 10 °C )

T^ = Temperature of the heater

T = Ambient temperature (20°C) o ' From Eq. (7), the current requirement calculated and

experimentally needed for heating the specimen to

800-1000°C was found to be in the range 8-10 amperes, in

the early experimentation the current was drawn from a

12 volt battery of 500 ampere-hour strength, but long-term experiments required the use of a very clean Hewlett-

Packard 0-20 V, 0-20 A 6264 B DC power supply. The fact that the tungsten wire resistance increases sharply with increasing temperature assisted in the maintenance of a 192 constant temperature. Any temperature drop decreased the resistance which at fixed voltage increased the cur­ rent automatically when other parameters were held constant. 37

An in-situ calibration of the specimen thermocouple

was performed by observing (SEM image) the melting of Au,

Cu and Al in a hydrogen atmosphere. The thermocouple

reading exactly matched tabular values for the 660°C

melting temperature of A l . For the melting temperature of

copper (1083°C) and gold (1063.5°C), the in-situ thermo­

couple read low by 10°C (Fig. 5). The temperature readings

used in the experimental results are all corrected from

Fig. 5.

3. Gas Pressure at the Sample Surface

A good vacuum is required for the electron gun and

electron collection in the normal operation of the

scanning electron microscope. But locally a higher gas

pressure can be directed at the specimen surface by leaking

gas at a known rate through a gas pipe near the surface of

the sample. From the kinetic theory of gases, the Hertz-

Langmuir equation given by Eq. (8) was used to evaluate

the effective oxygen pressure at the surface of the

sample.

J = AP . / (2ttRTM. )1/2 (8) max 1 1 where J is the maximum (or gross) molecular flow rate max ^ 2 in mole/sec, P^ is the pressure in dynes/cm , the molecular weight of gas (for O ^ r 32 gm/mole), R the gas 7 2-2-1 -1 constant (8.314x10 g cm sec mol K ), T the 2 temperature (K) , A the area (cm ) of exposed sample. Figure 5. Calibration of the HSESEM Thermocouple. HSESEM the of Calibration 5. Figure Real Temperature, 800 1200 700 900 100 ,00 lbain SSM Thermocouple HSESEM f o alibration C 700 etn o N/u Eutectic Ni/Au of Melting Melting of of Melting 800 t A n Difference no hroope Reading Thermocouple 0 1300 900 etn o Cu of Melting 1000 HSESEM 1100 efc Thermocouple Perfect 200 38 In order to use this equation for calculation of an

effective pressure P., the flux J must be known. if ^ i max the gas flow at the surface is known, e.g., 3 cc/min,

-i j t -i /c n 10-3 273 mol _ , . max would eI3ual 3/60 x 2271 X 300 S 5 5 7 ' For thls value 2 J , P. was calculated to equal 40.4 dynes/cm or max x . ^ 2 -4 3x10 atmospheres for a typical exposed area of -2 2 1.5x10 cm . The maximum local equivalent oxygen pres- -4 sure usable in the HSESEM is then about 3x10 atm and

this pressure or lower pressures were used in the experi­ ments .

To calibrate the gas flow rate, a setup shown in

Fig. 5 was used. Oxygen was passed from the gas cylinder

so that part of it bubbled out in the bubbler and part passed through, the capillary of the manometer. In this way, atmospheric pressure of oxygen was maintained on the higher pressure side of the manometer and U-tube. The oxygen flow through the capillary of the U-tube created a pressure difference in the two arms of the U-tube and was shown by the fluid height difference. The gas flow rate was calibrated against the difference in fluid height of the manometer. A graduated cylinder (shown by dotted lines in Fig. 6) was incorporated between the leak microvalve of the microscope and the manometer U-tube as shown. Oxygen was passed from one side of the U-tube and the microvalve of the microscope was opened at the other side. While oxygen was flowing through the system, a Capillary Tube Stopcocks

Oxygen Cylinder To Microscope

i

I -=\

^ S o ap Solution

Atmospheric Bubbler

Figure 6. Apparatus for Monitoring and Calibrating Gas Flow into Microscope. little soap solution was introduced to form a soap film in

the graduated tube. Under the pressure differential, the

flat soap film rose in the graduated tube and its measured

volume displacement with time equalled the rate of oxygen

flow into the microscope. The height difference in the

U-tube was recorded for various flow rates. A plot of the

flow rate vs the fluid height difference in U-tube then

represented the calibration of the oxygen flow rate.

Obviously, the graduated cylinder was removed and the set

up shown by solid lines in Fig. 6 was used during actual

experimental runs. The flow rate was directly measured

from the AH reading of the flowmeter and Fig. 7. Also

the oxygen pressure at the sample surface could be evalu­

ated by the reading of the Penning specimen chamber

vacuum gage, Fig. 7. A localized P at the specimen -4 2 surface equal to 3x10 atm (or a flux of 3 cc/min) -3 corresponded to a 0.18x10 torr pressure in the specimen

chamber, a value 1270 times lower.

191 193 4. Collection of Secondary Electrons ' The radiation shields could obstruct the direct flow of secondary electrons such that low-energy electrons

(0-50 ev) could be deflected or lost, increasing the noise of the image. To avoid this, the collector was kept at maximum value of 300V (positive relative to ground) to

increase the signal/noise ratio. Further improvement in A H ,cm 20 gur . i on of he Fo eter. m Flow e th f o n io t a r lib a C 7. re u ig F 0.06 enn Gg, 0‘3 Torr 3 ‘ I0 Gage, Penning Flow Flow Rate ,cc/mn Rate 0.18 0.24 42 43

the image, especially for high-temperature runs, was made

by decreasing the beam voltage while retaining the high- 191 beam current. This reduced the beam penetration in the

sample surface, so that secondary electrons from only the

top few layers were collected to improve the signal/noise ratio. Increasing the spct size by lowering the final

lens current was also found to increase the signal/noise ratio.

The electrical grounding of the specimen was found to 191 192 be very important. ' The specimen was therefore insulated from the rest of the stage and electrically grounded. This helped to drain any electrostatic charge accumulated on the surface of the specimen. Otherwise, surface charge tended to deflect the beam from the chosen spot and no resolution was achieved.

Interference from the thermal electrons and photons 18 5 186 reported by other authors ' was not found to be significant in ourexperiments conducted up to a tempera­ ture of 900°C. Masking of the secondary electron image by the photons may occur if experimental runs are conducted above 1000°C. For very high temperature runs (up to

1350°C) one of the top shields was biased with respect to the ground. This negative potential applied at the top of the furnace minimized the obstruction of secondary electrons by the radiation shields and thus increased the ratio of secondary electrons to all types of electrons at 44 a constant temperature. Figure 8 is a plot of the signal versus the voltage on the radiation shield at room temper­ ature. At low temperatures, the loss of thermal electrons was negligible but a similar plot could be obtained at high temperature where, however, the signal increase was somewhat smaller. As shown in Fig. 8, the best results were obtained with a potential of -10V on the top shield and at intermediate magnification

(200X to 2K).

Two types of scintillators were used, a scintillator tip in Perspex B coated with aluminum coupled with a plastic light pipe for low and intermediate temperature experiments (300°C to 900°C), and a quartz light pipe coated with Perspex B and aluminum for higher temperatures

(T > 900°C). Frequent recoating was needed, but the recoating of the low temperature scintillator could be easily performed in a regular vacuum evaporator.

Phosphor scintillators showed no resistance to the heat radiated from the furnace and were therefore not used.

5. Recording of the Image

A 35 mm camera with a motordrive was available to take time-lapse photographs for events which occurred rather slowly. Initially, a 16 mm movie camera was adopted for the events that might take place during the start of the in-situ oxidation. But later, a video system Figure 8. Signal Enhancement Versus Applied Potential on Potential Applied Versus Enhancement Signal 8. Figure Signal -4 the Top Shield. Top the X -20 20x 50x 20k -24 200x -28 500 500 x

45 46

consisting of a TV monitor coupled with a videotape recorder

was installed. The TV monitor greatly assisted the viewing

and recording of rapid high-temperature oxidation phenomena

where significant changes occur in a time span of seconds.

In addition, operating at the TV scan rate reduced specimen 194 charging effects as discussed by Wells. ' However, the

use of the SEM at a TV scan rate is restricted to electron O beam diameters greater than 1000 A and therefore to rela­

tively low resolution images. In addition, the micrographs

obtained from the TV screen are generally quite noisy.

6. Conclusions and Performances

An overall schematic view of the HSESEM is shown in

Fig. 9. The ensemble has proved to be versatile and rela­

tively easy to use. It has been successfully used to -4 temperatures up to 1350°C, to pressures up to 3x10 atm

and with magnifications up to 20,000X. The Cambridge

Stereoscan S4 has proved to sustain very high temperatures and relatively high pressures without major repairs.

However, very frequent cleaning of the chamber and the column was necessary to maintain a maximum resolution.

C. Post Oxidation Analyses

Frequently, parallel experiments were performed in a conventional furnace at the same oxygen pressure and temperature as the HSESEM. Some samples oxidized in Figure 9. Schematic Diagram of the Hot Stage and EDAX System. EDAX and Stage Hot the of Diagram Schematic 9. Figure Cathode Ray Tube Recorder Display Camera 35 mm 35 Video Video I Videotape Monitor T.V. Grounded or Sample Either Biased rScintillator/- Photomultiplier

Excitation

Heater to Wire Power Supply Pole Piecev » 7“1 r7

Mo-Radiation Shields ( Thermocouple Be-Window n r Al-Window —L_ J Gas Pipe Mtetector High Voltage Power Supply Power Supi

U

r-m ->fc>—»Pre-Amp Computer Analyser ^ Liquid Tank N 2 Amp. Linear Video - Teleprinter o vj either the conventional furnace or inside the HSESEM were

further studied with post-oxidation equipment such as a

high resolution Jeol JXA 35 elec-tron probe, an X-ray micro­ analyzer, a PHI 595 Auger microscope, an Hitachi-H600 AEM

and a Vacuum Generators HB501 STEM with EELS capability. III. OXIDATION OF OFHC COPPER

A. Introduction 196 The early electron diffraction work of Honjo showed

a marked discrepancy between the theoretical and experi­

mental P ,T diagram for the Cu/CuO/CU2 O system. Figure 10

shows the regions of theoretical stability for Cu, CuO 195 and CU2 O given by (dotted lines) and the experimental

boundaries found by Honjo. This discrepancy shows that

the chemical equilibrium at the oxide/gas interface

assumed in Wagner's theory of oxidation, does not hold

especially at lower temperatures. However, this assumption

is fairly good- at high temperature.

In addition, Fig. 10 shows our experimental Pn 2 window. According to Honjo, our experimental conditions

guarantee the formation of cuprous oxide only.

B. Specimen Preparation

1. Pretreatments

Well-annealed OFHC copper cylinders (1 cm dia., 3 cm

tall), with a tilted cross-section for exposure to oxygen, were mechanically polished up to 600 SiC grade. Differ­

ent methods were subsequently used to obtain a clean

surface with varying amount of cold work. Figure 10. Temperature/Pressure Regions of Formation of Cu_0 and CuO (x x after x (x CuO and Cu_0 of Formation of Regions Temperature/Pressure 10. Figure

atm c6 n -8 200 Rapp and Shores,^95 0 Shores,^95 and Rapp P O 2 max 2 O P 300 400 -o- --- o a- a -o- 0 600 500 --- 0 after Honjo^^°). after 0 T° C 700 -o CItJD/CtlO 800 C h /C.X) 900 1000 O tn 51

- After mechanical polishing, the specimens were electropolished in an aqueous 70 percent phosphoric acid

solution at 20°C. The specimen were then inserted inside the chamber and heated to temperature either in the

HSESEM vacuum or in a beam of hydrogen (P„ - 1x10 ^ atm) 2 - Alternatively, after electropolishing, the specimen could be heated in an oxidizing atmosphere up to 1065°C.

Subsequent heating below the melting point of copper would liquefy the metal/oxide interface because of a eutectic in the copper-oxygen system at 1.54 weight % oxygen and 190 1065°C. This treatment produced extremely clean copper surfaces for subsequent oxidation.

- Equivalently any initial cuprous oxide scale could be reduced either in vacuum or by hydrogen before oxida­ tion of the underlying copper.

- After mechanical polishing, the specimen could be shot-peened. A controlled stream of 25 y dia. plastic spheres in a stream of dry air at high velocity impinged against the surface of the specimen and produced a highly cold worked structure of 20 y depth. The specimen was then inserted inside the specimen chamber.

2. Contamination during Oxidation

Contamination of the specimen during oxidation can result either from the evaporation of volatile species from the radiation shields (MoO^fv) or Ag(v)) or else 5 2 from the deposition of carbon and sulfur produced by the cracking of the diffusion pump oil. But these impurities could be detected either by in-situ EDAX analysis or post­ oxidation Auger analysis.

X-ray maps did not reveal any contamination from MO or Ag where the sample was at a temperature below 600°C.

Above 600°C contamination of the surface by Moo^ and Ag occurred only during the cooling of the specimen (Fig. 11).

During the oxidation, the volatile species emitted from the radiation shields should only condense on the colder parts of the stage, rather than on the hot specimen sur­ face. On the contrary, these species could condense on the sample during fast cooling (in the case of Fig. 11, the oxidation run lasted 24 hours). Figure 12 shows the contamination detected with a post-oxidation Auger analy­ sis on a copper sample oxidized at 800°C for 16 minutes.

The use of a liquid nitrogen trap has considerably lowered the amount of sulfur, chlorine and carbon detected on the sample. Furthermore, by annealing the specimen in vacuum for 1/2 hour at 1000°C in the HSESEM before oxidation, the contamination of the first oxide monolayers by sulfur and carbon has been supressed. As discussed previously, the contamination by molybdenum has occurred during the cooling of the specimen. In Fig. 12, a sputtering time of O 10 mn corresponds to an average material removal of 100 A. Fast Cooling

T-950° C— >*— Cooling

o Fast Cooling o

Slow Cooling

Slow Cooling

20 2 4 2 4 h Time(mn)

Figure 11. Contamination during a 25 hr Run at 950°C from In-situ EDAX Analysis Inside the HSESEM. o Atomic 70 ’ Figure Figure 100 40 30 20 50 90 80 ) b ( 12 . . 12 Sputtering Time (mn) irgnTapn ad nta Pretreatment Liquid Initial with (c) and but Trapping Nitrogen Trapping Pretreatment; Initial Nitrogen without Liquid (b)with Traps; After Oxidation of Copper at at Copper of Oxidation After tion. Nitrogen Liquid of use the (a)without HSESEM otmnto Dtce ih n ue Microscope Auger an with Detected Contamination (annealing at at (annealing Cu 0 0 0 0 0 0 70 60 50 40 30 20 10 SputteringTime (mn) 1000°C 1000°C < 0^ o £ O/ 120 40 50 30 20; 60 80 90 70 10 4 for for - 0 0 0 0 0 0 0 80 70 60 50 40 30 20 10 (c) 30 30 Sputtering Time (mn) i eoe Oxida­ before min 0° 800°C nie the Inside Cu s,c

54

55

In conclusion, clean substrate conditions could be

obtained by using a cold trap close to the specimen and by

evaporating away several layers of the specimen surface

with a pre-oxidation anneal at 1000°C in the HSESEM.

C. Results and Discussion

1. In-Situ Oxidation of Copper at 300°C and Pn = 3x 10“ 4 atm 2 No features could be resolved before 15 minutes of

oxidation; at that time oxide ridges of about 0.1 micron

thickness were seen. The initial oxidation corresponds

certainly to the "invisible" oxide film proposed by 4 Benard.

Very little change in the oxide surface was seen even

after six hours of oxidation. After 20 hours, the scale

appeared to be compact with circular oxide grains of about

0.2 microns diameter, as shown at 5,000 and 20,000X magnification in Figs. 13a and 13b. The dark depression

in the center of these grains suggests that grain boundary diffusion may be predominant in the growth of a surface ridge around each small grain. Similar morphologies were observed by Halliday and Hirst"^^ and Jaenicke et al.^^

The dissociation pressure of cuprous oxide at this -25 195 temperature is 1.0x10 atm. Thus the experimental 21 pressure was 3x10 larger than the equilibrium pressure.

Such large supersaturations produced a large defect density (fj 1 ym

Figure 13 Oxide Film Formed on Electropolished OFHC Copper at 300°C in P0 ~ = 3x10“^ atm after 20 hours (a) 5,000x; \b) 20,000x; after 3 days 2.5 hours, (c) 20,000x. in the 2d adsorbed layer and a large number of 3D oxide nuclei randomly distributed over the surface without specific epitaxial relationships with the metal sub- 197 strate . There are no data available for grain boundary copper diffusion in cuprous oxide but it is very likely that it exhibits a lower activation energy than the lattice diffusion, as does Ni in NiO. More generally, the initial growth of cuprous oxide grains is certainly related to dislocation diffusion along the defect network created by the formation of the 2D adsorbed layer. Grain boundary diffusion, creating the dimple structure shown in Fig. 13 occurs only when the 3D nuclei impinge together

(see Part V).

Small oxide whiskers were seen after thirty hours of oxidation (Fig. 13c). These whiskers apparently nucleated at random sites in the underlying oxide film. The whiskers were found to grow with time at their tips, and to develop lateral projections at their tips. When the temperature was increased by 50°C, such a whisker would not increase in length, but the protrusion at the tip of the whisker grew. Similarly, upon a removal of the oxygen flux (evacuation of the system), the whisker tip was seen to shrink.

A general discussion on nonlayered growth will appear in Part V, but in short, it is believed that whisker growth occurs by surface and lattice diffusion along the 58 screw component of a dislocation. The dislocations can either be created during the oxide film growth, or else 30 come from the metal substrate (see literature review).

In the case of the oxidation of copper at 300°C and -4 3x10 p , no whiskers were observed before 30 hours and 2 they are certainly produced by dislocations created during oxide growth. 198 As discussed by Sartell et al, cooling from below

700°C causes the oxide on copper to spall extensively, because of the large disparity between the expansion . . 84 coefficients of Cu and CU2 O. Consequently, the oxides grown on copper at relatively low temperature could not be studied with post-oxidation techniques. The processes of -4 oxidation of OFHC copper at 300°C and P = 3x10 atm 2 are sketched in Fig. 14.

2. In-situ Oxidation of OFHC Copper at Intermediate Temperatures (500°C to 850°C)

a. Nucleation of the 3D layer

As discussed above, the formation of the 2D adsorbed layer which corresponds to the initial induction time, could not be resolved in the HSESEM. However, the obser­ vation of discrete nuclei on top of this 2D layer could be observed at temperatures higher- than 400°C where the reaction supersaturations are lower.

Figure 15a shows the surface of electropolished OFHC copper at 500°C prior to oxidation. Thermal etching 59

0

Cu20 Single Crystals

IS

G.B. Diffusion

Ih < t< 3 0 h Growth of Cu20

t> 3 0 h - Whisker Growth Crystal Defects

Figure 14. Schematic Growth of Cu20 on OFHC Copper at 300°C and P = 3tL0- ^ atm. 2 60

Figure 15. Oxidation of Electropolished OFHC Copper at 500°C, PQ2 = 3xl0~4 atm. (a) Initial Surface 5000X; (b) oxide nuclei after 40 seconds 5000X; (c) 80 Seconds 10000X; (d) 3 minutes, 10 seconds 10000X. 61

revealed a triple point of grain boundaries, and two twins

(lower right and center left), which are indicated by arrows. Unless otherwise stated, the specimen was heated

in the vacuum of the HSESEM, its temperature raised to 900°C and then cooled down to the reaction temperature. The equilibrium PQ for Cu/C^O equilibrium at 500°C is “16 ^ 7.4x10 atm and consequently, a thin film of oxide was certainly formed during heating in the HSESEM pressure “ 8 regime (P = 1.0x10 atm). 2 Furthermore, the presence of a thin oxide film would explain the appearance of grain boundaries on electro­ polished OFHC copper after heating in vacuum. Grain boundaries pn electropolished OFHC copper cannot be resolved by a secondary electron emission at room tempera­ ture. Oxide nuclei were formed differentially on these grains as soon as oxygen was leaked into the system to -4 provide P = 3x10 atm locally. A similar type of o2 behavior on copper has been explained (Mitchell and 199 Lawless ) on the basis that the incubation period was the result of oxygen going into solution until the oxygen concentration near the surface reached a critical value which varied with crystal face at which time oxide nuclei suddenly formed by a precipitation process. After 4 0 seconds, the elongated oxide nuclei of Fig. 15b exhibit a preferred orientation, and are aligned differently on each grain. Likewise, from grain to grain, differing oxide grain densities and growth rates are observed. The area of each oxide grain was measured from photos as a function of time and was found to vary linearly with time. Conse­ quently, on the surface where oxide nuclei have not yet formed, there seems to exist a thin layer intermediate 43 between a chemisorbed layer and an oxide layer. 43 Rhead's model, considering the growth of oxide particles to be controlled by surface diffusion was developed for circular discs. For small oxide nuclei the concentration field around the oxide possesses approximately a circular symmetry. Consequently, the linear rate law between the area of the particle and the time still holds. The nature 37 of the diffusing species is still a matter of conjecture, ' but, in view of the high supersaturation of oxygen, copper seems to be the best candidate in our case. The elongated shape of the nuclei probably results from the epitaxial relationships between the oxide and the metal and the anisotropy for Dg, the surface diffusion coefficient (see for example ref. 200). It is shown later that at these relatively high supersaturations, the nuclei locations are related to the defects of the 2d layer rather than directly to the underlying metal structure. Consequently, the long straight edges of oxide nuclei probably correspond to low index surfaces for the oxide with minimum surface energies and a shortage of growth ledges. For example, such grain 201 shapes were found. on iron. A plot of -log(l-f) versus the square of the time

where f is the fractional area of oxide nuclei is shown

in Fig. 16 for 3 different grain orientations. The

linearity confirms the statistical model given by (2) for

constant nucleus density. According to this model, all

the nuclei appear at the moment when the exposure starts

without any subsequent appearance of further nuclei.

The disc shaped nuclei are also assumed to expand on the

surface at a constant radial growth velocity. For a + 7 2 given nucleation density of 3.9x10 per cm , which was

found to be more or less the same for different orienta­

tions, the radial growth velocity varied from

1x10 ^ cm s ^ to 4x10 ^ cm s ^ .

Figure 17. shows the oxidation of electropolished -4 OFHC copper at 700°C in Pn = 3x10 atm. The oxida- 2 tion at this temperature shows a morphology similar to

that at 500°C.i However, the presence of facets on the

initial surface is now clearly seen (Fig. 17a). Thermal

facets, along certain crystallographic directions changing 202 from g r a m to grain represent high index planes.

Evaporation of copper surfaces in oxygen is known to produce a stepped structure related to the emergence of 203 dislocations lines. Whereas the step structure is difficult to observe in an inert atmosphere, oxygen modi­ fies the surface energy anisotropy and diffusion produces 204 the faceted surface of minimum surface free energy. gur Pl of es t f he Differ e f f i D Three r fo t s versu ) f - l ( g o l - f o t lo P . 6 1 re u ig F -lo g (l-f) 0.2 0.4 0.3 0.5 0.6 -o 200 Gri of FC opr di 500°C t a d e iz id x O Copper OFHC f o s rain G t n e n P in 02 30 am. atm ^ 340” = 2 1200

64 65

Figure 17. Oxidation of Electropolished OFHC Copper at 700°C, P0 2 = 3x'10-4 atm, (a) Initial Surface 3800X; (b) Oxide Nuclei after 1 min 1520X. As seen on Fig. 17, there is some tendency for the oxide to grow along the directions of the facets and at right angle to them giving an epitaxial initial scale.

Figure 18 shows the oxidation of OFHC copper at -5 . . . 850°C in P = 2x10 atm. Thermal faceting is again 2 clearly seen. As soon as oxygen was leaked into the microscope,oriented, elongated oxide nuclei were seen spontaneously. Photograph 18(b) was taken after 0.5 min oxidation, and random nucleation of oxides was evident.

Again, there is some tendency for the oxide to grow along the direction of the existing facets as described earlier

After 3-5 minutes, preferential nucleation both at grain boundaries and along the thermal facets became apparent

(see Fig. 18 b,c,d). These oxide mounds grew, along with a uniform oxide film between them (see Fig. 18e). Such nuclei have been seen by other authors working with thin films of copper where preferred nucleation and growth of 129 oxide on certain metal planes were observed.

At this temperature, the supersaturation ratio is only 3.8xl0+^ and nucleation at the metal defects

(e.g., at grain boundaries) is relatively more favorable than for the lower temperatures. Furthermore, the 2D adsorbed layer may recrystallize at this relatively high temperature.24

To assess the nucleation phenomena in this range of temperature and pressure, OFHC copper was oxidized and Figure 18. Photomicrographs at 1000X showing Cu Specimen Surface when Heated in SEM to 850°C in Po = 2 2x10 atm. (1,000X); (a) After .5 hr in Vacuum; (b) after 0.5 min in 0~, (c) after 3 min in C^. 68

Figure 18 (continued). (d) after 7 min in C^; (6) after 40 min. in (f) after 2.5 hr. in C>2 • 69 reduced at 820°C in different oxygen pressures. Figure 19 shows the same surface spot, identified by a twin boundary of 5 y width going from the lower left corner to the upper right corner of the photos. The initial oxide of Fig. 19a was reduced in vacuum and the surface was reoxidized. Although the oxide nuclei retained their same orientations on their respective metal grains, the exact sites for renucleation were different. Consequently, the crystallographic defects responsible for nucleation belong to the 2D adsorbed layer rather than to the metal sub­ strate. When the temperature of oxidation is raised at constant P (lowering the supersaturation) , the metal 2 defects become more and more important to the oxide nucleation process. Furthermore, the oxide nuclei are clearly stopped at a neutral grain boundary on the surface

(see point P on Fig. 19b). As mentioned above, the preference of nuclei for preferred orientations and growth directions results from a differing availability for growth ledges and the anisotropy of D . Very similar b morphologies were observed by Tomlinson and Y a t e s ^ ^ ' ^ ^ during the oxidation of copper in P = 1 atm, although no 2 metastable dendrites were seen in the HSESEM prior to the formation of CU2 O crystallites.

The same spot was oxidized at different P (Fig. -4 19d at P =6x10 atm was taken after partial removal of 2 70 ■

v Ojj m

Figure 19. Oxidation of OFHC Copper at 820°C for 30 s (2,000X); (a) and (b) in Pq2 = 3x10”^ atm;

(c) in Pq = lxl0~4 atm; (d) in P0 = 6x10“^ atm. ^ 2 oxygen.) According to Fig. 19, the nucleation density is an increasing function of the oxygen pressure. However, the growth rate for a single nucleus does not seem to in­ crease with increasing P The nuclei on the lower portion

2 * of Fig. 19c has developed a dendritic morphology where secondary dendrites are attached to a primary growth dendrite. This bidirectional growth is clearly evident in the next figure. OFHC copper was heated to 820°C in vacuum and faceting occurred on one grain of Fig. 20a.

Oxygen was subsequently leaked into the HSESEM chamber and cuprous oxide nucleated perpendicular to the facet direction (Fig. 20b). These nuclei were then reduced in vacuum and the facets reappeared (Fig. 20c). Finally, the reintroduction of oxygen produced new oxide nuclei again parallel to the facet directions. This suggests that Dg, the surface diffusion coefficient , is more influenced by the close packed directions than by profile of the surface.

b. Scale Growth

After nucleation and lateral growth of the oxide grains through their impingement, the surface is completely covered with 3D oxide. Figures 21a and b show the same spot as in Fig. 15 after 20 min. and 3 hours of oxidation, respectively. After 20 minutes, the surface in Fig. 21a is apparently completely covered by oxide; the twinned metal grain in the lower left avoided the nucleation of Figure 20. Morphology of OFHC Copper at 820°C (2,000X) (a) Heated in vacuum; (b) oxidized for 20 s in Pc>2 = 3xl0-4 atm; (c) reduced in vacuum, (d) Oxidized for 30 s in Pn = 3xl0“4 atm. 2 Figure 21. Oxidation of Electropolished OFHC Copper at 500°C in Pq 2 = 3xl0”4 atm; (a) after 20 min (b) after 3 hrs. Figure 21 (continued) (10,000X) after (c) 20 hrs 50 min.; (d) 1 day 53 min; (£) 1 day, 1 hr and 37 min. large initial crystals and developed a morphology similar

to that for 300°C-20 hr of Fig. 13b. Other features of

Fig. 21a can be related with difficulty to the earlier

morphology of Fig. 15d; the brighter areas indicate areas

of thinner oxide. The large initial oxide grains of

Fig. 15d have been obscured by further oxidation. After

3 hours of oxidation, Fig. 21b, a rough surface is evident;

as the locations of grain boundaries for the initial oxide grains are'well known, the morphology of Figs. 21a

and 21b does not suggest that these boundaries served as

sites for preferential growth.

After 20 hours of oxidation the oxide scale in another location developed pyramids of relatively large grain size

(~ 3 pm) composed of relatively flat faces, except for the existence of macroscopic spiral ledges seen in Fig. 21c.

In addition, the top of the pyramid is clearly charging up under the electron beam. Such charging could be produced by a pit or a tunnel at the top of the pyramid as shown 207 by Motojima and Sugiyama on hollow Cr^Si^ whiskers.

After 25 hours, a new grain appears in the grain boundary (center left) of Fig. 21d adjacent to a grain with horizontal ledges. However, about a half hour later,

Fig. 21c, the new grain is covered (or incorporated) by the pyramid on the right. A discussion of pyramid growth will appear in Part V. After 44 hours of oxidation the scale is relatively

flat and uniform. At still later times, oxide cracking,

whisker, and platelet formation were observed. Figure 22

shows such whiskers observed after five days of oxidation.

Figure 22 a,b and c show the same whisker

rotating around a vertical axis. These latter observa­

tions are consistent with the presence of large strains

during the growth of the oxide scale at this temperature.

The temperature is too low to accommodate these strains

by a plastic flow of material or creep by a diffusion mechanism.

As shown in Fig. 18 at 850°C the oxide grains grew

laterally, consuming the featureless uniform oxide film between them (see Fig. 18f) . During later growth stages,

fissures appeared at the scale/gas interface (Fig. 22a).

Such perforation of the scale has been explained by Mrowec 7 and Werber for sulfide scales (Fig. 23c). The subsequent

filling of gaps between oxide grains and of fissures nucleated during scale growth is one of the most striking phenomena observed in the high temperature range (see

Fig. 23b) . The white featureless oxide seen at the grain boundaries and filling voids is an oxide fume formed in the gas phase. Consequently, the term "oxide fume" is used to distinguish this electrically charged oxide in the SEM from the larger cuprous oxide grains (see Part III C3). 77

Figure 22. Oxidation of Electropolished OFHC Copper at 500°C in P0 2 = 3xl0”4 atm after (a) 5 days, 1 hr and 15 min; (b) 5 days, 2 hrs and 50 min; (c) 5 days, 6 hrs and 10 min; (d) 5 days, 12 hrs and 5 min, (5,000X). 78

Figure 23. SEM Photographs of In-situ Oxidation of OFHC Cu Oxidized at 850°C for 4.5 and 4.75 hr. in Pq ^ = 2x 10“5 atm (2000X) (a) Shows voids at grain boundaries (arrows) (b) Shows that some voids appearing in (a) have been filled by growing oxide Oz 79

Cu e

Dissociation Fissures

Figure 23 (continued) (c) model for evolution of scale fissures across cation conducting scale after Mrowec and W e rb e r.16 80

c. Influence of Surface Pretreatments

1. Electropolished specimen heated in vacuum

As shown during the oxidation of electropolished

OFHC copper at 850°C, there is a featureless uniform oxide film between the cuprous oxide grains. This extremely thin oxide develops during heating in the vacuum of the 196 HSESEM chamber. According to the results of Honjo shown in Fig. 10, the specimens are oxidized between 20°C and 750°C during heating in vacuum. Figure 24 shows the different morphologies obtained on electropolished OFHC copper for different heating profiles. The mass of oxide formed during heating, m, is:

t 750 r (9) '0 dm where the oxygen impingement rate (^pjr)T is a function of the temperature which is itself a linear function of time

(see Fig. 24).

T = at (10) where a is the slope of the heating profile. The time t^j-g needed to reach 750°C is given by Fig. 24. From the 114 data collected by Ronnquist and Fischmeister, an approximation of (g^)T as a function of temperature is given by: 0 50 100 150 t(mn)

Figure 24. Heating Curves for 3 Different Experiments on Electropolished OFHC Copper in Vacuum. Leak­ ing of O 2 Took Place at t^, t2 and t^, respec­ tively. Corresponding Morphologies are Shown. -2 where m is in yg cm , t in minutes and T m degree

Celsius. Consequently, the total thickness of oxide h

formed during heating between 20° and 750°C is:

h = 3.78xl0-11 x a4 x | ^ s q 5! x 145.9 A (12)

0 2 where 14 5.9 A corresponds to 1 yg/cm for a compact 114 CU2 O film. Finally, the thickness of the initial

oxide layer for the 3 experiments were 0.15 ym, 0.6 ym

and 2.30 ym, respectively. Upon leaking the oxygen at the

reaction temperature, cuprous oxide grains are nucleated

and grow but generally never achieve intimate contact

because of the initial thin oxide film (see for example

experiment #3 of Figure 24). If, however, for exactly the

same experimental oxidation conditions, the heating time i

reduced, then the thickness of the initial oxide is reduced

and CU2 O grains grow laterally more rapidly (experiment

#2). If the heating time is short enough, complete cover­

age can occur and grain growth is possible (experiment #1).

2. Effect of cold work on the oxidation of OFHC Copper at 520°C in P = 3x10" atm 2

It was already mentioned that dislocations affect greatly the oxidation behavior at intermediate temperatures

Most of the morphological features described earlier for low to intermediate temperatures are related to disloca­ tions created during growth of the thin oxide film. But heterogeneous nucleation occurring at substrate disloca­ tions (metal grain boundaries) was observed at high tem­ peratures (Fig. 18 b,c, and d ) . For further clarification, the effect of initial cold working of the metal substrate was investigated. The specimen was subjected to a shot- peening surface treatment and a highly cold worked struc­ ture, 20 p thick, was produced at the surface (Fig. 25a).

Upon oxidation, tiny crystals of C ^ O nucleated randomly on the surface without any obvious habit relations with the substrate (Fig. 25 b,c and d). After 3 hours, the cuprous oxide crystallites had an average diameter of

1 pm (Fig. 25e). Subsequently, the nuclei grew in height with small steps on their faces (Fig. 25 f and g). After

1 day and 5 hours of oxidation, deep holes were clearly seen at the top of the crystallites (Fig. 25h).

Overgrowths were observed on certain sites of the sample surface (Fig. 25i) . After 1 day and 10 hours

(Fig. 25j) the particles were clearly growing by a step advance on pyramid faces very similar to the growth mechanism observed on electropolished copper at 500°C

(Fig. 21 c,d,e). After 2 days the oxide scale is rather flat for high Pn (Fia. 25k) except for the presence of 2 whiskers (Fig. 251), very similar to the one grown on 84

Figure 25. Oxidation of Cold Worked OFHC Copper at 520°C in Pq 2 = 3x 10“4 atm (2000X) after (a) 0 min; (b) 6 min; (c) 40 min; (d) 3 hrs, 14 min. 85

Figure 25 (continued) after (e) 3 hrs 14 min (5,000k ); (f) 23 hrs and 55 min (5,000X); (g) 1 day, 1 hr and 10 min (5,000X) (h) 1 day, 5 hrs and 40 min (5,000X). 86

Figure 25 (continued) after (i) 1 day, 6 hrs and 10 min (2,000X); (j) 1 day, 10 hrs and 20 min (20,000X); (k) 2 days, 10 hr and 10 min (500X); (1) 2 days and 18 min (10,000X). 87 electropolished OFHC copper at 300°C (Fig. 13c). These morphological changes completely rule out any predominance of grain boundary diffusion through the outside scale since the crystals grew first at their tip and then along the steps created by their initial growth.

In conclusion, the effect of preliminary cold work of the metal greatly changes the oxidation morphology of

OFHC copper. Instead of an initial epitaxial layer nucleated on the defects of the 2D adsorbed layer, the dislocations of the metal substrate acted directly as nucleation sites for the oxide. In this intermediate temperature regime, predominant grain boundary diffusion has never been clearly observed. On the contrary, the oxide grains grew at their tips (see part V ) .

3. Oxidation of Electropolished, Hydrogen Heat-- treated OFHC Copper at 550°C and in P_ = 3x 10“4 atm 2

Another way to suppress formation of the initial 2D adsorbed layer consists of using hydrogen during heating in the HSESEM.

Electropolished OFHC copper specimens were heated in the HSESEM in a hydrogen atmosphere for one hour. The hydrogen treatment was then stopped and entry of oxygen immediately followed. About five minutes is needed to purge H 2 from the gas pipe and the connections outside the specimen chamber. During hydrogen treatment, the cop­ per surface became very rough and perforated by randomly distributed pores at grain boundaries as well as within metal grains. Hydrogen diffuses very rapidly into the bulk of the specimen and complicates tremendously its subse­ quent oxidation behavior. Consequently, no simple oxide nucleation was observed, but instead, a very anisotropic oxidation with porous morphologies was seen. Figure 26a-d shows the oxide scale formed after 5 hr 20 min and after

10 hr. The underlying copper grains are still visible after 10 hours of oxidation (Fig. 26d). In addition, the oxidation is greatly anisotropic; some copper grains show a relatively slow growing smooth oxide (grain I in Fig.

26b or d). Some other grains exhibit a very porous oxide scale (II in Fig. 26b) which grows quite rapidly (II in

Fig. 26d). The oxide grains in either case are strongly oriented and certainly epitaxial. For example, the oxide nuclei form relatively smooth grains aligned in a specific direction. After 11 hr. 20 min. of oxidation, oxide nuclei between 0.1 ym and 1 ym in diameter are distributed on the oxide surface (Fig. 27a and b). These new nuclei grow very slowly and after 1 day and 8 hours of oxidation they appear to be uniform dodecahedra with a 1 ym edge (Fig.

28a-d). These oxide grains nucleated preferentially on the porous initial oxide (Fig. 28a). The similarity 89

Figure 26. Copper Oxide Grown on OFHC Copper at 550°C in p02 = 3xl0“4 (a) (b) after 5 hrs 20 min; (c) (d) after lo hrs; (a) 2,000X, (b) 500X; (c) 2,000X (d) 1,000X. Figure 27. Copper Oxide Formed on OFHC Copper at 550

in P q 2 = 3 x 1 0 “ 4 after 1 1 hrs and 2 0 min. (a) 10,000X, (b) 2,OOOX. 91

Figure 28. Copper Oxide Formed on OFHC Copper at 550°C in PQ2 = 3x10“^ atm at 1 day and 8 hours. (a) 2,000X, (b) 5,000X, (c) 10,0O0X (d) 20,000X. 92 between these oxide grains and the grains observed during oxidation of cold worked OFHC copper and their regular disposition on the oxide surface (Fig. 28b) suggests that they may be related to the metal grains or subgrain boundaries.

During initial growth, the oxide grains nucleate where dislocations end on the metal surface instead of a nucleation produced by the 2D adsorbed layer. However, once these dislocations slip out of the crystals they cannot provide any more growth steps. A process of secondary nucleation such as just described is due to the fact that "newer" dislocations are seen to provide growth at the oxide surface (see part V). Obviously, the grain or subgrain boundaries are almost infinite sources of dislocations. For example, the secondary nucleation observed in Figs. 27 and 28 can result from dislocations emerging at metal grain or subgrain boundaries.

After 2 days of oxidation, the grains have grown sufficiently to connect to each other Fig. 29a. Ultimately these secondary oxide nuclei cover the entire surface

(Fig. 29b). After 2 days and 7 hours of preferential growth, oxide overgrowths were observed (Fig. 29c). These morphologies are similar to the ones observed previously and are consistent with the presence of large growth strains in the oxide scale at this temperature. Finally, Figure 29. a large grain size (4 0 ym) was observed after four days of oxidation (Fig. 30a,b). The oxide growth occurs by the translation of macroscopic surface ledges very similar to the ones observed previously. Consequently, for the quite late stages of oxidation (several days), the features of strain accommodation and growth via screw dislocation ledges are conserved during the oxidation of hydrogen-treated OFHC copper. However, oxide nucleation and the early stages of the oxidation are different from vacuum-treated specimens, and undoubtedly complicated by the diffusion of hydrogen into the copper during heating in the HSESEM. As will be shown for the high-temperature regime, the effect of hydrogen is greatly reduced for rapid heating and consequently moderate diffusion and damage by hydrogen inside the metal.

3. Oxidation of OFHC at high-temperatures

a. Eiectropolished OFHC copper heated in vacuum

1. Results

Figure 31a shows the typical morphology of the oxide -4 formed on OFHC copper oxidized at 920°C in Pn = 3x10 2 atm for 15 minutes. The scale was composed of relatively flat cuprous oxide grains 10 ym in diameter; the grain boundaries were clearly delineated by the oxide fume. At this high temperature, two simultaneous mechanisms are observed: the initial growth of solid cuprous oxide 95

l(3) 2pm (b) 10pm

Figure 30. Copper Oxide Formed on OFHC Copper at 550°C in Pn = 3xl0-4 atm after 4 days (a) 5,000X; 2 (b) 2,0 OOX. 96

Figure 31. Oxide Film Grown on OFHC Copper at 920°C in PO2 = 3xl0“4 atm (a) after 15 min (5,000X); (b) after Argon Has Been Introduced for 1 min (5,000X); (c) after Oxygen was Introduced again for 2 min (5,000X). crystals and the formation of a solid amorphous copper

oxide fume ("smoke" or "fog") analogous to that formed 208 from liquid melts. At 930°C, the vapor pressure of

copper is high enough^^^ (1.3x10-^ atm.) so that some cop­

per atoms can collide with oxygen molecules (at -4 3x10 atm.) above the metal surface to form a solid

smoke of C ^ O . When a high flux argon beam of about 2 15 cc/mm -min was used to strike the oxide scale for 30s,

the oxide fume' was immediately mechanically blown away

from the scale and the residual oxide seen in Fig. 31b

consisted of the solid cuprous oxide grains. Upon rein­

troduction of oxygen, the oxide fume again covered the

oxide scale as seen in Fig. 31c. During this time, oxide

grain growth took place and the unstable grain marked

1 eventually disappeared.

For the slightly higher temperature of 930°C, Fig. 32a

shows the initial formation of cuprous oxide on OFHC copper

heated in vacuum. The metal surface was covered with

oxide crystals immediately as oxygen was leaked into the

system, and after 30s the average grain size was about

2.5 ym (see Fig. 32a). At 930°C the scale was very flat

and the amount of oxide fume was slightly increased and obvious at the oxide grain boundaries. Grain growth was very rapid after three hours (Fig. 32b to e). 98

Figure 32. Oxide Film Formed on OFHC Copper at 930°C in Pq 2 = 3x10“^ atm after (a) 30 s (2,000X), (b) 195 min (1,000) , (c) 210 min (1,000X). Figure 32 (continued) (1,000X) after (d) 242 min, (e) 244 min, (f) 271 min. 100

-4 When OFHC copper was oxidized in P = 3x10 atm at 2 950°C, the amount of oxide fume increased (Fig. 33a). Upon decreasing the oxygen pressure to 1x10 ^ atm at 950°C, the amount of fume observed on the surface decreased as oxide grain growth became predominant (Fig. 33). Similarly, when the temperature of the sample was reduced, thus lowering the vapor pressure of copper, the formation of oxide fume in the gas phase stopped (Fig. 33c), and the growth of the oxide grains engulfed the fume existing on the surface.

At a still higher temperature (T = 970°C), the vapor pressure of copper is so high that a considerable amount of fume is formed. The oxide grains can even be completely covered by the oxide fume formed in the gas phase

(Fig. 34a) .

Upon decreasing the temperature of the sample, the oxide fume undergoes a quite reversible structural and morphological transformation around 519°C (Fig. 34b).

Figure 35 shows the morphology of the copper oxide scale cooled from 930°C to room temperature. At temperatures below its 519°C transition temperature, the deposited oxide fume separates from the grain boundaries. Indeed, it would be very difficult for a usual post-oxidation analysis using a conventional oxidation procedure to imagine that the "debris" seen in Fig. 35 was comprised of oxide fume formed in the gas phase. In order to charac­ terize better these different oxidation products, the 101

Figure 33. Oxide Film Formed on Electropolished OFHC Copper at 950°C in Pq ^ = 3xl0-4 atm (1,000X)

after 65 min; (b) 700 min after Pq ^ Has been Reduced to P 0:> = lxlO-5 atm; (c) 135 min at 740°C. Figure 34. Oxide Film Formed on Electropolished OFHC Copper (1000X) (a) after Oxidation at 970°C for 25 minutes; (b) after Temperature Reduced to 500°C (P = 3x10“^ atm). 2 103

Figure 35. SEM Micrograph Showing Surface of OFHC Copper Oxidized at 850°C in Pq2 = 3x10”^ atm for 300 min (800X). Micrograph Taken after Cool­ ing to Room Temperature. 104

oxidation of a 20 vim TEM grid OFHC copper grid was per­

formed in the HSESEM.

A TEM grid sample at 940°C inside the HSESEM is shown

in Fig. 36a, after 100 minutes of oxidation in -4 P_ = 3x10 atm. The holder marked H is an electro- 2 polished OFHC copper cylindrical specimen. Upon leaking

oxygen into the specimen chamber, oxide nucleated on the

OFHC copper holder in an epitaxial manner as observed in

Fig. 18. However, oxide formation on the TEM grid was much finer grained and uniform in morphology. After about

one hour of oxidation, the boundaries between the oxide

grains of the Cu holder were brighter and appeared white on the surface; the same phenomenon was observed on the

TEM grid. After one hundred minutes of oxidation, the oxide scale morphologies on the TEM grid and on the holder were similar. Figures 36a and b show that in each case the cuprous oxide grains are covered with a smoke or fume.

The oxidized TEM specimens were subsequently cooled in the oxygen atmosphere of the HSESEM. As previously mentioned, the fume residing on the specimen surface undergoes a transformation at 519°C. The transformation is quite reversible at this particular temperature, and the white

(high-temperature) oxide recovers its featureless appear­ ance upon heating, Figs. 37a-d. However, as seen in comparing Figs. 37b and d, the details of the fume 105

Figure 36. TEM Grid and OFHC Copper Holder after 100 min of Oxidation at 950°C in Pn = 3xl0-4 atm 2 (a) 500X, (b) 1;000X. 106

Figure 37. Morphological Changes with Thermal Cycles for TEM Grid Oxidized in Pq2 = 3xl0"4 atm (1000X) . (a) Oxidized at 742°C for 2 hrs; (b) cooled to 500°C; (c) reheated to 583°C; (d) Cooled to 459°C. 107 morphology is changed by thermal cycling. The OFHC copper holder underwent the same morphological transition,

Figs. 38a and b. As rationalized in the Discussion sec­

tion, this transition corresponds to the transformation of -4 cuprous oxide to cupric oxide in 3x10 atm of oxygen pres­ sure. The observed transition temperature (519°C) cor­ responds remarkably well with the calculated value for the equilibrium temperature between CU2 O and CuO in -4 P = 3x10 atm taking into account surface energy effects. 2 To confirm that the morphologies seen in the HSESEM were not artifacts caused by contamination inside the specimen chamber, parallel experiments were performed on

OFHC copper TEM grid samples in a separate furnace (SF) at the same oxygen pressure and temperature. These TEM grids were subsequently observed in a Hitachi H-600 high resolu­ tion analytical electron microscope (AEM). Figure 39 shows the OFHC copper grid oxidized separately (in SF) for 30 minutes at 940°C and P = 10 4 . At 40,000X, spherical 2 particles are seen to cover the oxide surface. In the dark field TEM mode, as seen in Fig. 40, the shape of these particles can be established; in fact, they are nearly equiaxed'with an elongated neck where they are attached to the substrate. The thickness fringes seen in Fig. 40b indicate that these particles are facetted. The underly­ ing oxide substrate in Fig. 4 0a seems to be amorphous but in fact is analogous to the one shown later to consist of 108

Figure 38. Oxidation Morphology of OFHC Copper Holder at 940°C in P0 2 = 4xl0~3 atm for 2 hours after Cooling to Room Temperature, (a) 100X; (b) 1000X. 109

Figure 39. SEM Picture of Oxide Formed on OFHC Copper at 940°C in I* = 3xl0“4 atm. (a) 500X; (b) 40,OOOX. 2 110

(4) $oo h

Figure 40. Dark Field TEM Picture of _ . Particle Formed on OFHC Copper at 94 0°C in Pq = 3 x

10-4 atm. (a) 300,000k , (b) 1,000,000X. Ill o 25 A diameter oxide grains. The shape of the particles in

Fig. 40 and the fact that the underlying substrate consists

of 25 A diameter grains indicate that these oxidation products have been formed in the gas phase and have there­

after impinged on the surface. The particular shape of the particles shows that they have sintered to the sub­

strate .

From the observations of Figs.39 and 40 carried out in the Hitachi H-600 AEM, both the TEM image contrast and microdiffraction indicate that the particles and the sub­ strate are quite different in character. The oxidation morphologies were examined more closely on a Vacuum

Generators HB-501 STEM. A typical region containing particles attached to the substrate is shown in Fig. 41.

Examination of the larger particle shows that it is not spherical but bounded by a number of planar surfaces, any two of which meet in a sharp line, and four of which join at a single vertex. This morphology is typical of small 209 metal nuclei which condense onto substrates, the initial nucleus being a tetrahedron bounded by four {111} planes for fee metals, and the observed particle being the agglomeration of six such tetrahedra into an icosohedron.

The composition of the particles and substrate shown in

Fig. 41 were more closely examined; in particular, the oxygen distributions were determined by electron energy 112

Figure 41. STEM Micrograph of Fume Particles Attached to the Substrate (500,000X). loss spectroscopy (EELS). Although EELS does not give

reliable quantitative results, except under certain 210 experimental conditions, it is very sensitive for

detecting the presence or absence of light elements. The

EELS spectrum shown in Fig. 42a was taken from the large

particle in Fig. 41 using a 20 & diameter beam contained

within the particle. The spectrum shows the presence of

copper, but it was not possible to distinguish any oxygen

peak at 75 ev above the spectrum noise, although a small

perturbation, due to silver contamination, at 370 ev

could be discerned. For comparison, a spectrum taken in

the substrate adjacent to the particle, Fig. 42b (mathe­

matically smoothed to enhance the peak visiblity), shows

both oxygen and carbon, in addition to the copper peak.

In addition to the EELS spectra, microdiffraction patterns

of both the particle and the substrate were made. The

pattern for thp particle showed clear Kikuchi lines

expected in a convergent beam diffraction (CBD) pattern

from a single crystal (Fig. 43a). For the underlying sub­

strate, while not containing the sharp detail of the particle, the pattern still showed clear diffraction spots

Fig. 43b, in a pattern characteristic of a cubic structure

The presence of many additional spots showed that more than one crystallite had contributed to the pattern, although the regularity of their distribution suggests Intensity ,N(E) gur 2. ES Spectrum EELS 42a. re u ig F I28K 100 200 t r e Son n Fi 41. . ig F in Shown te tra s b u S e th f o Energy Loss, eV Loss, Energy 300 2K 400 500 600

700 114 Intensity ,N(E) 10 0 30 0 50 600 500 400 300 200 100 0 gur 2. ES pcrm t i es Son n Fi 4 . 41 . ig F in Shown s le tic r a P e th f o Spectrum EELS 42b. re u ig F I28K 8K Energy Loss, eV Loss, Energy 116

Figure 43. Micro Diffraction Patterns (a) of the Particles Shown in Fig. 41; (b) of the Underlying Oxidized Substrate . 117 that all the crystallites have the same structure. There­ fore, the substrate is not amorphous, but consists of small crystallites.

For further characterization of the particles and the substrate, a lattice fringe image of one of the particles was formed. A 2 & periodicity was revealed for the particle (Fig. 44a) as were sets of fringes for the substrate. In agreement with the diffraction observations, the oxide substrate was revealed to consist of crystallites O typically 25 A in diameter, such as shown at A in Fig. 44b.

In conclusion, the evidence of high resolution STEM imaging, microdiffraction, and EELS is that the oxide fume is in fact composed of a microcrystalline copper oxide substrate and relatively few tiny single crystals of copper.

2. Discussion

The investigation of the in-situ oxidation of OFHC copper inside the HSESEM has revealed interesting mor­ phologies, most of which cannot be observed by a conven­ tional post-oxidation examination of a specimen after extended oxidation and cooling. In this study, three different phenomena have been particularly examined inside the HSESEM. Firstly, rapid grain growth of the oxide scale was observed (Figs. 31 and 32). Because of the large grain size, the rapid boundary migration and the small oxide 118

Figure 44. STEM Micrograph of the Substrate (5,000,000X). 119

grain thickness, oxide grain growth was driven predominant­

ly by strain energy introduced by the oxidation and not by

the small capillarity forces.

For the high-temperature oxidation of copper, the

formation of an oxide fume on the top of the cuprous oxide

scale was observed. Indeed, two simultaneous and. competi­

tive mechanisms were observed: the growth of dense cuprous

oxide crystals and the formation of a microcrystalline

fume analogous to that formed from liquid melts exposed to 208 dilute mixtures of oxygen and an inert gas. Similar 211 studies performed for Na, K, and Zn vapors have shown

that the density of the fume particles increases while

their size decreases with increasing oxygen pressure. 208 211 However, in both cases ' the oxygen was doped into

an inert gas carrier stream, and the models for nucleation

of fume were based on counterdiffusion of metal vapor and 208 212 oxygen within boundary layers of inert gases. ' 213 Kaito et al. studied the microscopic morphology of metal

oxide fume particles, including C ^ O , prepared by burning metals in an Ar-C>2 gas. Cuprous oxide fume was found to possess an octahedral or more complicated polyhedral habits.

Our experimental arrangement does not involve dilute oxygen in an inert gas, so that the modeling of fume

formation involves a different mechanistic approach which must consist of calculating the number of collisions in the gas phase. Before presenting this model it should be pointed out that two kinds of fumes are present. At relatively low temperatures (T = 850°C to T = 900°C) oxide fumes are seen filling gaps and voids in the gas scale

(Fig. 23). The evaporation of copper from these fissures accounts for the localized formation of the oxide fume.

In this case, the oxide fume is only filling the cavities.

The oxidized area in the HSESEM is very small compared to the total surface area of the specimen, so that the unoxidized part of the sample contributes most of the fume formation, and condensed fume should indeed cover the entire surface. Then a relatively large amount of fume formed in the HSESEM is a consequence of the particular geometry of the system. However, fume formation was also detected upon oxidation of copper in a conventional furnace, where the entire sample is oxidized. In this case, the fume is only a consequence of evaporation from gaps, cracks, voids or grain boundaries between oxide grains. The same phenomena were observed during the oxidation of iron at 1280°C (Fig. 45). At 1280°C, the -7 195 vapor pressure of iron (3.9x10 atm) was high enough to form an oxide smoke by a mechanism identical to the forma­ tion of cuprous oxide smoke on pure copper at 850°C

(Fig. 23).

A question arises as to why the fume at the scale grain boundaries is more highly charged in the HSESEM than Figure 45. Oxide Film Formed on Electropolished Iron 1280°C after (a) 10 Minutes 1,000X; (b) 15 minutes 1/000X. 122

e

Fume ft/- vi'sol v 7

& Y

Oxide Fume

Oxide Scale

Figure 46. Schematic CUjO/gas Interface and Equivalent Electrical Circuit in HSESEM. 123 the fume deposited on the top of the grains. This peculiar electrical effect is consistent with the model outlined in Fig. 46 which sketches the surface morphology and an equivalent electric circuit. The voltage driving the electron beam of the microscope should impose a constant voltage drop across the total oxide product. However, since the electronic electrical resistance of the imperfect

CU2 O grain boundary regions should be greater than that for the grain interiors, i.e., Rg>R^, the fume particles located above grain boundaries should be more highly charged, with the potential at B more negative than at G.

A locally more negative potential at some site on the specimen surface can reduce the secondary electron emis- 194 sion from neighboring points.

Accordingly, as can be seen in Fig. 33b, the fume residing on the grain centers at the external surface will also appear somewhat brighter (by emitting secondary electrons because this area is more distant from the grain boundaries).

Theory of Fume Formation

The in-situ observation of OFHC copper oxidation revealed a C ^ O fume residue with some small Cu particles at the C^O/gas interface. The particular geometry of the

HSESEM resulted in a small oxidized spot (about 2 mm dia.) on an unoxidized copper plane of about 1 cm dia. In 124 the absence of an inert gas in the HSESEM or with pure 02 of reduced pressure in a conventional furnace (as reported) the fume formation is believed to initiate from bimolecu- lar collisions of copper atoms and oxygen molecules in the gas phase.

The thermodynamically stable bulk oxide in the Cu/O system at T > 800°C in the HSESEM is Cu20. An evaporation contribution from molecules such as Cu2 (g) from copper metal can be neglected. The vapor pressures for

Cu20, CuO and Cu species over Cu20(s) or Cu(s) have been calculated using the dissociation energies of Cu20(v) and

CuO(v) and the free energy functions derived from refer­ ences 215,216* These values for 927°C and P = 1x10 ^ 2 atm. are listed in Table 1. (see Appendix B). According to Table 1, the predominant vapor species prior to any collision are Cu(v) from solid copper and C>2(g) in the gas -4 beam. The equivalent P of about 3x10 atm in the 2 HSESEM corresponds to the transitional flow regime, i.e., the calculated mean free path of 0.026 cm for oxygen molecules is somewhat smaller than the dimensions (about

1 mm) of the collision system. The evaporation of copper follows a cosine law directional distribution, i.e., evaporation of atoms in random directions.

If fume formation arises from the collision of copper atoms and oxygen molecules in the gas phase, the first step of the reaction must be: Table 1. Equilibrium Partial Pressures for Vapor Species Over Cu and CU2O at 927°C.

PCu O OVGr Cu20(s) PCuO °ver Cu20 (s ) Pcu OVer Cu20(s) Pcu °V6r Cu*s) 2 * j for P = 1x10 atm for P_ = 1x10 atm 2 2

8.69x10 ^ atm 2.35x10 ^ atm 2.33x10 ^ atm 1.34x10 ® atm 125 126

Cu(v) + 02 (g) -»■ Cu02*(v) .(13)

since large numbers of collisons by three or more bodies

are a statistical impossibility. The existence of the 18 217 binary dioxygen vapor complexes has been demonstrated. ' 19 Lever et al. have confirmed the existence of a C u 0 2 18 molecule. Darling et al. when condensing copper vapor

at low temperatures in argon containing 10% oxygen, proved

the existence of a Cu(C>2)2 molecule. Furthermore, the

reaction:

Cu + 02 + CuO + 0 (14)

218 is highly endothermic. On the contrary, the reaction

(13) seems more favorable energetically. Finally an

energetic CuC>2 molecule could reduce its kinetic energy

by colliding with a previously formed fume particle (M) in

the gas phase:

Cu02* + M -* Cu02 + M (15)

By the same argument, a molecule such as Cu02 should

only be needed for the initiation of the process. The

propagation could result from further collisions of

Cu(g) atoms and 02 (g) molecules with early particles of

stable fume.

Nevertheless, it is significant to calculate the num­ ber of collisions for reaction (13) to provide an approxi­ mation for the relative (minimum) amount of copper vapor 127 involved in fume formation. The following reaction steps are proposed:

(.1) Cu(v) + 02 (g) -*■ Cu02 (13)

(2) Cu02*(v) + M + Cu02 (v) + M (15)

(3) Cu02 (v) + Cu(v) -> Cu20(v) + 0(g) (16)

The asterix * emphasizes the fact that Cu02 is relatively energetic before any further collisions.

From classical collision theory, the number of copper/ oxygen collisions per second is:

where nCu and nQ are the densities of copper atoms and oxygen molecules, VCu is the average relative velocity of copper atoms to oxygen molecules and a is the effective cross-section for collision defined as:

(18)

oxygen molecules, respectively.

The calculation of is performed in Appendix c taking into account the temperature difference between the two gases.

The evaporation of copper from the specimen is assumed to occur uniformly and follow a cosine law distribution. 128

Oxygen molecules are assumed to be contained in the cylin- 2 der ttR h where R is the inner radius of the leak pipe and

h the distance between the end of the pipe and the speci­

men surface. The oxygen density in the beam is a constant

n02° (see Appendix D). Using the copper flux,

nCu vCu J^,u = --- Eq. (17) can be rewritten as:

Z = 2/2 o 5Cu n0^° Jcu/?cu (18)

where vCu is the average velocity of Cu(v). 219 Following Ruth and Hirth, the rate of copper

evaporation originating from a differential area dA^ and

arriving at a parallel planar area dA of coordinates x

and b (see Fig. 47) is given by

9 Jfj 1 d j (b,x) = — cos8 dwdA^ (19) where J is the Hertz-Langmuir flux given by nlj

JHL ' PCu/(2™ C u kT)1/2 <20> and 0 is the angle between the normal to dA^ and the di­ rection of the beam, w is the solid angle of the beam, so that:

, dA cos0 , > dw = — (21) d

The total copper flux at x and b is then: 129

dA

dw

dA,

Figure 47. Geometry of the Sample and the Gas Pipe. 130

'2u J 2 H L c o s ( J (b, x) = ~ jd A '~X'~ = J rdr j d

2 2 2 s = b + r = 2br cos <{> (23)

COS A 0 = XJ (24) a 2 2 2 and d = x + s (25)

Equation (22) becomes

P 2tt 2 ,-m r rdrd({) J x J(b,x) = / / ---2— 2— 2------2 (26) J J u(x +b +r -2br cosij))

J(b,x) is calculated in Appendix E and equals:

1 2 / 2 _Lu 2 \ J(b,x, = -4£ J" - <* +b )------+ , (27) v—A-----3-- 5-- ^-----5-- o J A +2a A x -bA + (xAb ) m m 219 As treated by Ruth and Hirth, in the case of I << x, m Eq. (27) reduces to

2 2 J UT £ x t /u ir\ — HL m (28) ' - 2 ” 272' (x -b )

For the present case of fi. >> x and b is small

(Eq. (28) reduces to:

2 J (b / x) = JRL (1- 2-j) (29) £m which is only a function of x. This flux does not take into

account the copper loss due to collisions below x. Conse­

quently G(x) is defined as the flux at x, modified by

depletion through collisions; we have:

x G (x) = J (x) C I Z(x) dx (30) o

where C is a positive constant less than or equal to unity

which equals the fraction of Cu-C^ collisions that are

successful and produce CuC^. As written in Eq. (30) G(x)

is smaller than the real flux at x since Cu-C^ collisions would not always occur over the entire distance between O

and x. The real flux is somewhere between J(x) and G(x).

Equation (30) can be replaced by the following differential

equation:

From the appropriate substitutions of Eqs. (18) and (23),

Eq. (3 0) becomes:

dG (x) o 2/2C a n G (x) (32) dx Cu 2

The general solution of E q . (32) is: and k is the constant of integration. Because G(o) =

JHL: .

k = (35) (BC)

Finally, the modified flux at x is given by:

2 - 2Jhl r 1 # 1 . ^ m 1 . -BCX, 2 tBC (BC x) + 2 ^ (36) £m (BC) “BCX From the series expansion of e , G(x) = J(x) for C=0,

and the modified flux, in the case of totally unsuccessful

collisions, is equal to J(x) given by E q . (29) .

From Eqs. (18) and (36), the number of collisions per

unit area and height at x is:

. 2BJHL . 1 ,1 . ^ / m 1 , -BCX. ,,,, z(x) = -JT- BC (BC - x> + (T - - r ^ )e 1 (37) m i

Z(x) is a decreasing function of x and there exists a

value x for which: o

Z(x ) = 0 (38) o

If x q £ h, the flux of copper would be totally con- 2 sumed by collisions in the cylinder of volume ttR h, and the total number of collisions in the cylinder per unit time is: 133 X

Z = / 2iTbdb J Z(x)dx (39) 0 0

Neglecting the oxidized area we integrate from 0 and R.

From Eq. (37)

2 2 n 2 2JHL rXo Xo ^ / m 1 w . -BCX Z = ttR — n— [— - + (-7 ------~) (1-e o) ] (40) I C B (BC) m

Substituting Eqs. (29), (37) and (38) in (40)

Z = ^ J ( Xo) (41)

The condition x q << prevents C from being close to zero and the case x q < h is possible only for C close to one.

The term J(xo ) in Eq. (29) also depends strongly on the collision efficiency C.

The total number of successful collisions is then

Z = 7iR2J (x ) (42) sue o

However if x q > h, the flux of copper in the cylinder 2 ttR h is not all consumed and the total number of success­ ful collisions per unit time is:

R h Z = C f 2-nbdb f Z(x)dx (43) O Uw J J 0 0 or The total number of collisions is in reality larger than this since the real flux of copper is greater than G(x).

In fact, the total number of collisions is larger than the 2 value given by E g . (44) but lower than ttR Bh JHL which is the total number of collisions without considering copper depletion. ^

The function Z(x) given by E q . (37) is plotted in

Fig. 48. The value for C corresponding to a linear dependence of Z(x) with x is called Co

For (-— ^ -) = 0 , C o = 0 - (45) dx Co m

Figure 48 shows a plot of Z(x) versus x for the value of

Ch when the copper flux is totally consumed at x = h

Z (h) = 0 for Ch (46)

Since the value x for which Z(x) = 0 with C = Co is equal £m to —= and always much greater than h: /2

l>Ch >>Co (47)

The dotted lines in Fig. 48 show the number of col­ lisions for a nondepleted flux given by Eq. (29) and 135

C

X 'n

Figure 48. Number of Cu-C>2 Collisions as a Function of x : Z(x) given by Eq. (37) Using a Nondepleted Copper Flux. 136

As mentioned above, Eqs. (37) and (48) are lower and

higher bounds, respectively, for the real number of

collisions.

For a typical flux of oxygen of 2 cc STP/mn, an

oxygen temperature of 2 98 K and a specimen temperature of 1223 K, Eqs. (34), (C31)and (C33) give:

B = 3.77 cm ^

I = .5 cm Co = .75 from E q . (4 5) m ^ h = .1 cm Ch = 11> 1 from E q . (4 6)

Because Ch is greater than 1, the copper flux is not 2 entirely consumed in the cylinder ttR h for these condi­ tions. Consequently Eq. (44) should be used to calculate the lower bound for nonsuccessful collisions per unit 2 time. The higher bound is calculated using Z = ttR Bh J„T H L For T = 1223 K JUT = 2.13x1c*16cm“2s"1 CU nJL

T0o = 298 K J_ 0 , . 2 C>2 = 2 c c / m m _2 h = 0.1 cm 2 = . 5 cm and R = 5x10 cm m

The number of successful collisions for different values of C is listed in Table 2. The real number of collisions 2 is between the value given by Z and ttR BJ„t . SUC Hli Since the reaction CuC^fv) + Cu(v) Cu20(v) + Cu(v) is extremely favorable energetically, we can assume that most of the CuC^ particles formed during the first collision will react to produce the stable C ^ O molecule.

Consequently, the amount of fume formed depends only on 137

Table 2. Number of Successful Collisions for Different Value of C for Modified Copper Flux (zsuc)

and for Unmodified Copper Flux (ttR^BJ.,-. h) H Lj

c z S~1 ^R2BJOTh sue. n Li

1 5 .18 x 1 01 3 6fi .3x3 1013

12 12 io-1 5.57x10 6.3x10 CM 1 o 6 . 2 0 X 1 0 11 6 . 3 x 1 0 11 138

the number of successful collisions as listed in Table 2. 13 -1 If we consider Z = 1x10 s and a 50% factor of sue loose packing for Cu20 fumes we calculate a thickness of

.3 ym of fumes deposited on the surface after 1 hour.

From Fig. 39, C f 1 and some collisions are elastic such

that no reaction occurs (thereby permitting condensation of copper particles). If V is the relative velocity of an oxygen molecule with respect to a copper atom (see Fig. 43), the collision produces a Cu02 molecule only if the com­ ponent of relative velocity R = vcose parallel to the line of centers is greater than some value R . If R < R o —o no reaction should occur between Cu (g) and 02 (g) because the necessary activation energy is lacking. However, both species would be slowed down and since the square of the average velocity is linearly related to the temperature of the gas, the Cu (g) atom involved in such an elastic col­ lision (conservation of kinetic energy and momentum) would be slowed down. A later condensation or reaction could then be expected for such a copper atom. These nonreactive elastic collisions provide an explanation for the nuclea- tion of tiny solid copper spheres above the specimen sur­ face. Obviously, the propagation of this condensation mechanism occurs by collisions of Cu (g) with already formed solid copper.

During cooling of the specimen, condensed copper particles sinter to the surface in the same way as the Figure 49. Relative Velocity of an 2 O Molecule with Respect to a Cu(v) Atom. 140

oxide fume. Consequently the morphology of Fig. 39 should

be expected and it is interesting to note that the con­

densed copper particles are uniformly distributed on the

specimen surface. The fume particles grow by coalescence

of the gas evaporation. The particles collide with each

other and then unite with definite orientation so as to

minimize their interface energies and get to the shapes

shown in Figs. 40 and 41.

This model could be refined by calculating the col­

lisions between CuC^ and Cu and by taking into account the

formation of a cuprous oxide scale decreasing the area of

copper evaporation. However, these refinements do not

seem significant since the propagation of the fume forma­

tion is guaranteed by collisions of Cu(g) and 0 2 (g) with

already formed fume particles and since the fume formed

comes primarily from the evaporation of copper outside the

oxidized area.)

A more important contribution to the model consists

of taking into account nonelectrostatic interactions such as 220 221 the Lifshitz-Van der Waals interaction. ' Wagner

emphasized the importance of such interactions of particles with each other particularly in the free molecular 222 regime. The Lifshitz-Van der Waals interaction, which

is an interaction potential energy resulting from the sum

over the fluctuation-induced fluctuations of the particles 223 electrons, enhances the collision rate. Marlow 141 calculated a collision rate enhancement factor in the case of free molecular regime for silver equal to 4.86 and

2.82 for 1 nm and 100 nm particles, respectively. During the cooling of the specimen the chemically active cuprous oxide fume transforms to cupric oxide at an experimentally measured temperature of 519°C (Figs. 34,37,38). If the surface energy for the tiny particles is taken into effect, this transition temperature can be calculated as follows.

For a liquid drop of radius r, surface tension y, and molecular volume Q, the chemical potential p for a species 2vfi obeys the relation p = p + — !— with p = p° + RT In a, u oo 2T oo 9 where pro and p° are the molar chemical potentials for the bulk species and for the species in its bulk standard 224 state. Defay et al. have shown that this relation remains true for a crystalline solid with a shape obeying the Curie-Wulff equation = const., where y^ is the surface tension of the face i of the crystal and h^, the distance between that face and a central point. The

"drop model" is assumed to be valid for cubic C ^ O and monoclinic CuO. Thus the Gibbs energy change for the . reaction:

Cu20(s) + 1/2 C>2 = 2 CuO(s) is written as

AG = AG° - - m (49) YCuO Yc u 2° 142

The values of v_ _ and v_ _ are not available, and the data 'CuO 'Cu„0 225 226 of Hondros and McLean (confirmed by Bauer ) are used

to estimate these surface tensions. These investigators

reported a decreasing value for Yq u with increasing

until cuprous oxide was formed on the metal. The value of

y was found to equal y at low P . LU 2U 2

Then for the present evaluation,

y c u 2o “y Cu O ” YCu = Y* = 1 -4 Jm 2 at 1027°C - (50>

22 7 228 The dependency of y* on temperature is given by: '

dv* -2-1 -4-2-1 (5t } ~ _0*5 ergS Cm C 05x10 Jm °C . (51)

The average value of r^u ^ and rCu0 are estimated at 2_io rCu„ 0 = rCuO = r° = 25x10 m. Equation (49) can then be rewritten as

AG = AG» - ™ In P02 + §- (2fiCu0 - a ^ ) h * ♦ < | f ) (T-1300) J

(52) where

AG° = - 130,930 + 94.5 T J/mole Cu20.195 (53)

-4 For P = 3x10 atm the calculated equilibrium temper- 2 ature between Cu20 and CuO is T = 512°C, a value in remark­ able agreement with the experimentally observed 519°C, considering the numerous approximations used in the calcu­ lation . 143

The shape of copper particles corresponds well with 22 9 the summary given by Friedel on the atomic structures

which vary with the number of atoms:

- Small close-packed molecules leading to ico-

sahedra of 13 atoms.

- Amorphous structures of interpenetrating icosohedra,

stabilized by a surface tension effect up to about 50

atoms.

- Multilayer icosohedra with five-fold symmetry, up

to somewhat less than 100 atoms.

- Irregularly twinned fee crystals.

The shape of the cupric oxide particles is not as

easily resolvable because of their .small sizes. has

a cubic structure with less symmetry than the fee structure

and the surface tension effects mentioned above are weaker.

Furthermore, it is possible that it is not a 3-dimen­

sional compact configuration as suggested by Halicioglu 2 30 and White. A two-dimensional configuration may be energetically preferred if one considers three-body inter­ actions. In addition, it is impossible to rely on a lattice parameter measurement (for particle A, for example, Fig.

44b) to determine the exact structure of the particles.

There are some theoretical and experimental investigations suggesting that the lattice spacing will decrease with 2 31 decreasing particle size. Thermodynamically, this 144

232 effect is explained using the Gauss-Laplace formula. 232 Waltersdorf et al. not only found that the lattice parameter changed with the linear dimensions of those particles but observed that the lattice parameter within a single particle from center to border changed.

It was shown that the rate of fume formation is strongly dependent on the oxygen pressure. Consequently the oxygen dependency for the fume formation is much stronger than the oxygen dependency expected for parabolic cuprous oxide scaling (k^ a PQ or PQ depending on the defect model chosen). Previous kinetic studies of copper oxidation have only seldom been supplemented with post-oxidation microscopic examination. But Baur 116 et al. reported different morphologies for cuprous oxide formed at 1000°C, Pn = 80 torr, depending on how 2 the sample was cooled from the reaction temperature. When the system was evacuated before cooling, the oxide scale showed regular columnar cuprous oxide grains. When the system was kept in the oxygen atmosphere during cooling, particles of oxide were found on top of the oxide grains, to produce the morphology seen in Fig. 36. In light of the current evidence for smoke formation, cupric oxide must have formed from C ^ O fume on the top of the cuprous oxide scale. When their system was evacuated before cool­ ing, the C ^ O fume was probably reduced (as we have seen 145 in the HSESEM) so that no cupric oxide was formed. 116 Baur et al. also reported that the cuprous oxide scale showed "deep angular recesses between the crystals pos­ sibly extending to the base metal" which probably provided paths for copper evaporation. This particular scale morphology is a consequence of the specimen preparation.

b. Electropolish, Hydrogen-Treated OFHC Copper at 900°C in P = 3xl0-4 atm 2

An electropolished OFHC copper specimen was rapidly heated in a hydrogen atmosphere to 900°C. The heating and exposure to hydrogen did not exceed 10 minutes. Subse­ quently, oxygen was leaked into the specimen chamber. No hydrogen "damage", such as observed after long exposure in

H2 at 520°C, was seen on the OFHC copper surface during heating and prior to oxidation; the surface remained clean and smooth. Immediately after introducing oxygen, copper oxide nucleated on the surface (Figs. 50a-d) and rapid grain growth followed. The initial oxide morphology was different (more bulky) from the aligned, epitaxial grains observed previously for vacuum heatup. The kinetics of oxidation were faster in the case of hydrogen-treated sam­ ples. Apparently, the oxide film invisible in the SEM prior to oxygen entry for the case of vacuum-treated samples is important in establishing oxidation morphology and kinetics. Figures 50e-f shows the oxide for the 146

Figure 50. OFHC Copper Heated Rapidly in Hydrogen and Oxidized at 920°C in Pq2 = 3x10“^ atm for (a) 10s, (b) 20s, (c) 30s, (d) 40s. (5,000X). 147

Figure 50 (continued) OFHC Copper H2-treated and Oxidized at 920°C in = 3x10”^ atm for (e) 6 min

(20,000k ), (f) 9 min (10,000X), (g ) 11 min (5,000X), (h) 12 min. (2,000X). 148

I^-heated Cu after longer oxidation times; up to 10

minutes of oxidation there was no sign of either fume

formation or grain growth. But the scale morphology was

quite different from the scale observed for vacuum-treated

samples (Figs. 31 and 32). For a vacuum-treated specimen,

the oxide grains were flat with rather large gaps between

them after 15 minutes of oxidation. For hydrogen-treated

samples, the nucleation and impingement growth of oxide

grains occurred much faster. Consequently, the oxide scale

is composed of differently oriented grains and the surface

is very rough.

To investigate evaporation and subsequent fume forma­

tion, hydrogen-heated samples were further heated quickly

to 980°C. Because of thermal shock, the previously grown

oxide scale spalled off except at copper grain boundaries.

Figure 51a is a low magnification (200x) picture of the

resulting surface. The oxide adherent to the underlying

copper grain boundaries can be clearly seen in Figs. 51a-c.

Voids are seen at the prior metal/scale interface in

Fig. 51b; this morphology is similar to that observed by 232 Howes on Fe-Cr alloys. Metals such as copper which oxi­

dize by cation diffusion over vacancies through the oxide

film generate voids via cation vacancy condensation at the metal/oxide interface. These interfacial voids lead to

reduced scale adhesion, scale fissuring, and scale spalla­

tion, particularly upon thermal cycling. Secondly, the 149

Figure 51. OFHC Copper Oxidized at 920°C in Pq = 3x10“^ atm. for 25 min and then Subjected to Scale Spallation by Thermal Shock, (a) 200X, (b) 5,000X, (c) 2,000X. 150

linear expansion coefficients for Cu and of CU2 O are very 84 different; Hancock and Hurst report a ratio of linear

expansion coefficients for Cu and Cu20 of 4.32. As a

matter of fact, the system Cu/ Q ^ O has received consider- 82-84 198 able attention in the study of thermal stresses. '

However, the oxide scale had better adherence at the metal

grain boundaries. As shown in Fig. 51c, the oxide scale

did not spall, at the metal grain boundaries. The cation

vacancies transferred to the metal in the vicinity of

copper grain boundaries can probably diffuse more deeply

down the grain boundary and either support dislocation

climb or else precipitate in the boundary as a void, so

that void nucleation is avoided at the metal/oxide inter- 8 2—85 face. In this case, the oxide grains are better

anchored at metal grain boundaries and an improved local

response to thermal cycling was observed there. Accord­

ingly, an improved resistance to thermal cycling may be

expected for oxide scales grown on copper with a small grain

size. A preliminary surface deformation step such as shot peening or grit blasting which leads to surface recrystalli­

zation upon heating could be advantageous even for pure 233 metals. Giggins and Pettit have already demonstrated

such a favorable effect— for a different reason (forma­ tion of grain boundary films)--in the oxidation of

Ni-Cr alloys. 151

When the sample was heated more slowly to 980°C the

scale remained adherent to the metal; oxide fume was

indeed observed.

Hydrogen-treatment of electpolished copper prior to oxidation has proved to change the initial oxidation of copper dramatically. An extended hydrogen treatment upon slow heating damages the surface and leads to different initial oxide morphologies. Even if hydrogen is used for a short time, the initial stages differ slightly.

During heating in vacuum, a thin oxide film invisible in the HSESEM forms on the entire surface. Then upon increasing the oxygen pressure at the reaction temperature cuprous oxide grains grow but generally never achieve inti­ mate contact because of the initial thin oxide film. At -4 3x10 atm P , T > 920°C, and for rapid heating, contact 2 between C ^ O grains is achieved and permits subsequent grain growth. In this high temperature regime, atomic jumps between adjacent grains are sufficiently rapid to relieve the strain energy via rapid grain growth (Fig. 32). How­ ever, with slow heating there is no impingement of the

CU2 O crystals (Fig. 33). On the other hand, when the speci- ment is heated in hydrogen, no oxide film forms during heating and the initial cuprous oxide grains are able to impinge rapidly. Thus no gaps exist between cuprous oxide grains in the intermediate temperature regime (850°C to

900°C), and no fume is observed. 152

c. Electropolish OFHC Copper Previously Oxidized and then Reduced to Hydrogen

To minimize the diffusion of hydrogen into the metal,

reduction of the oxide by hydrogen was investigated, The

specimen was heated in vacuum to 950°C and then an oxide

scale was grown on the surface (Fig. 52a) for 35 min.

Then hydrogen was leaked into the specimen chamber and

this oxide was reduced. Upon complete reduction, hydrogen

flow ,yas stopped and oxygen was leaked into the HSESEM in

order to observe oxide formation on a very clean surface.

Figure 52a shows the morphology of an oxide formed at 950°C -4 for 35 m m in P = 3x10 atm. The subsequent reduction 2 of the oxide at 960°C is shown in Figs. 52b-h. The

reduction started at the boundaries between oxide grains 234 (Figs. 52b,c). Wagner rationalizes such a phenomenon and

Rao showed the same phenomenon during the reduction of 23 5 wustite. Because the reduction occurred at the metal/

oxide interface, copper nucleation was not observed. For

slow oxide reduction via a small driving force, the con­ version of C ^ O to Cu is more favorable at the metal/oxide

interface since no activation energy for copper nucleation

is needed there. As mentioned, Wagner has discussed this 234 mechanism. Figure 52h shows the reduced sample at a copper grain triple point. As soon as such the area was cleaned, oxygen was leaked into the HSESEM and oxidation was observed. Oxidation was carried out at 980°C, 920°, and 860°C following this procedure. After 30 min. of 153

. -4 Figure 52. OFHC Copper Oxidized at 950 C in Pn = 3x10 2 atm for 35 min, (a) and then Reduced in Hydrogen at 960 C (1,000X), (b) after 31 min of Reduction (1,000X), (c) after 39 min (1,000X), (d) after 49 min (1,000X). 154

Figure 52 (continued) (e) after 50 min, (f) after 51 min, (g) after 54 min, (h) after 60 min (1,000X). 155 oxidation in each case, the oxide scale was reduced prior to a new oxidation sequence on the same area.

As soon as oxygen was leaked at 980°C oxide nuclei appeared exclusively at grain boundaries of Fig. 52h.

This oxidation has been recorded on videotape. At this temperature the equilibrium oxygen pressure between Cu and -7 C ^ O is 3.2x10 atm. Since the working pressure m the -4 HSESEM is only 3x10 atm, the driving force for oxidation at this temperature is relatively small. Consequently, a heterogeneous nucleation of cuprous oxide along copper grain boundaries occurs. Figure 53a shows the oxide morphology after 15 min and preferential nucleation at metal grain boundaries is clear. At 920°C the driving force for copper oxidation is larger and Fig. 53b shows the oxide formed after 20 s. Preferential nucleation along grain boundaries is still visible but nucleation within copper grains has also occurred.

No such homogeneous nucleation was seen at 980°C after

5 min. Figure 53c shows the oxide morphology after 14 min of oxidation at 920°C.

Accordingly, the oxide nucleation at 860°C, Figure

54a is completely homogeneous, and after 10 min the scale is already uniform. Note that Figs. 52 through 54 repre­ sent the same spot on the specimen. Heterogeneous nucle­ ation at the metal grain boundaries in the very high temperature regime was also observed for samples where the Figure 53. OFHC Copper Oxidized in Pq 2 = 3x10 4 atm, (a) for 15 min at 980°C (1,000X), (b) for 20s at 920°C (2,000X), (c) for 14 min at 920°C (1,000X). 157

Figure 54. OFHC Copper Oxidized at 860°C in P0 2 = 3x10 ^ atm for (a) 50s (1,000X), (b) 4 min (2,000X). 158

initial oxide scale was removed by melting of the Cu/O

eutectic at the metal/oxide interface. Consequently

heterogeneous nucleation does not result because of the

oxidation of extremely reactive copper particles that have been previously reduced predominantly at the grain boundary.

In conclusion, if the initial oxide is removed, the nucleation sites are the sites of defects in the metal

instead defects in the 2D layers, and heterogeneous nucleation occurs especially at low supersaturations. IV. OXIDATION OF NICKEL

A. Introduction 195 From the data given by Rapp and Shores, the

dissociation oxygen pressure for the Ni/NiO equilibrium at — 8 •the high temperature of 1200°C is only 2.0x10 atm.

Consequently, . NiO is stable in all ranges of temperature

in the HSESEM vacuum. A clean metallic substrate requires

the use of a reducing gas such as dry hydrogen.

B . Experimental Materials

1. Material Used

Four different kinds and grades of nickel were used

for in-situ oxidation:

- A 99.996% polycrystalline rod from Williams and

Co. machined into a 1 cm dia., 3cm tall cylinder for direct oxidation inside the HSESEM.

- A 99.999% (Marz grade) polycrystalline rod pro­ duced by electron beam melting from Materials Research

Corp. The chemical analysis is listed in Table 3, and was performed by RMC using Mass spark spectrography (Leco method), conductometry (for carbon) and neutron activation.

The 1/4" dia. rod was cut into slices 2mm thick and these small discs were held in the HSESEM furnace with

99.99% polycrystalline Ni holders (see Fig. 55).

159 Table 3. Chemical Analysis of the 99.999% (Marz grade) Ni Polycrystalline

CHEMICAL ANALYSIS PPM

H Li Be B C N 0 F Na Mg Si P

<1.00 15.0 <5.00 12.0 0.15 2.00 17.0

Al s Cl K Ca Ti V Cr Mn Fe Co NI

0.90 0.25 1.00 0.20 0.45 <0.10 2.00 20.0 MAJOR

Cu Zn Gci Ge As Zr Nb Mo Rh Pd Ag CD

<0.10 <0.10 <0.10 0.65 <0.10 0.25 <0.10

In Sn Sb To Ta w Pi Au Pb Bl

1.20 0.50 <0.20 <0.10 <0.10

Others: <0.1 or

3 3 cm i he Oxi i Mr Gae i. N Grade Marz f o n tio a id x O e th g rin u d 2mm \ 7 7 T ^ hroope Tip Thermocouple Gas Pipe Gas

arzgrade M % 9 9 9 . 9 9 9. Ni % 9 .9 .9 9 161

162

- Nickel single crystals 99.99% from Monocrystals Co.

were cut into discs 6mm dia., 2mm thickness of random

orientation, mounted inside the HSESEM furnace as described

above.

-. A 99.99% Ni grid from Goodyear Atomic Corp.

2. Pretreatments

All nickel specimens, except for the Ni grid, were

initially mechanically polished with carbide paper

through 600 grade, with paste of 6 and 1 ym diameter

and finished with 0.3 ym alumina paste.

After ultrasonic cleaning, these specimens were either

inserted directly into the HSESEM chamber or else electro­

polished with a 60% sulfuric acid, 40% distilled water solu­

tion at 20°C and finally cleaned before insertion inside

the chamber.

Because of the stability of NiO in the HSESEM vacuum

at temperatures as high as 1400°C, the samples were either

heated inside the HSESEM under a hydrogen beam correspond- -5 ing to an equivalent hydrogen pressure of P - 1x10 a ^ 2 or else heated in vacuum before subsequent leaking of hydro­

gen to reduce the initial oxide scale. "Extra dry" hydrogen was used and was further purified by bubbling through a

liquid nitrogen trap before insertion into the HSESEM chamber.

As discussed for OFHC copper, contamination was minimized with the use of liquid nitrogen traps above the diffusion pumps. C. Results and Discussion

1. Oxidation of nickel at intermediate temperature (680°C) in P = 3xl0-4 atm 2 a) 99.996% nickel

Electropolished or mechanically polished 99.996% nickel heated in hydrogen was oxidized inside the HSESEM at 680°C. The morphological features were identical for both specimen preparations. However, oxide grains grew faster on an electropolished nickel substrate. Nickel oxide nucleation and growth at 680°C is shown in Fig. 56.

The discontinuous discrete nucleation of oxide observed for copper at 500°C was not seen; instead a thin poly­ crystalline film formed immediately upon exposure and uneven localized growth took place. The three grain boundaries shown in Fig. 56 did not provide any preferential nucleation sites for oxidation. On the contrary, no oxide nuclei are seen in the vicinity of the triple point (Fig.

56c). This effect could result from the more rapid dissolution of oxygen at the metal grain boundaries, an effect which would retard the achievement of oxygen super­ saturation locally. The dissociation pressure of nickel —17 195 oxide at 680°C is 1.83x10 atm. Thus the experimental 13 supersaturation ratio was 1.63x10 . Such a large super­ saturation ratio explains the random nucleation of a large number of small oxide grains over the surface, without Figure 56. Oxidation of Electropolished 99,996% Nickel at 680°C in a P0 2 = 3x10”^ atm. (a) Initial Un­ oxidized Surface 10,000X; (b) after Oxidation of 2 min.; (c) after 30 minutes 10,000X; (d) after 24 hours 10,000X. Figure 56 (continued), (e) after 5 days (20,000X), (f) after 1 day and 10 hours (10,000X). 166 obvious specific epitaxial relationships with the sub- 197 24 strate. As rationalized by Oudar, the nucleation sites for 3D oxide grains are either the dislocation lines of the 2D adsorbed layer or of the metal substrate. At this moderate temperature and high supersaturation, the 2D adsorbed layer has no time to recrystallize and provide the nucleation sites. In Fig. 56c, an irregular dislocation network is probably present, but a practically uniform 7 -2 oxide grain density slightly less than 10 cm is seen.

After 1 day of oxidation, metal grain boundaries are still visible at the scale/gas interface (Fig. 56d) and the metal defects did not seem to play a role in oxide nucleation. Even after 30 minutes of oxidation, the average grain size was only about 0.1 pm. At the latter stages of film growth the oxide grains size increases (Fig. 56c,d) .

This recrystallization reaction has been described by 73 Porrow et al. As mentioned by these investigators, these latter oxide films contain dislocations and incoherent boundaries. From Fig. 56e we again observe the dimple structure characteristic of grain boundary transport. In fact, the oxidation morphology for nickel at 680°C is quite similar to that observed on OFHC copper at 300°C (Fig. 13).

More generally, dislocations created during early growth would be expected to act as short-circuits paths for rapid diffusion of nickel. Figure 56c and f show overgrowth 167 or whisker growth (white dots) nucleated at these sites.

The new grains observed in the last two micrographs do not correspond to a new oxide nucleation but rather to con­ tinued growth of existing grains fed by dislocations (see

Part V ) . In conclusion, oxide grain boundaries and dislo­ cations act as short-circuit paths for the outward diffu­ sion of Ni in NiO at 680°C. Whereas these observations 73 236 confirm numerous previous works, ' there was no indi­ cation of inward diffusion of oxygen and subsequent swelling of the oxide grain boundaries as suggested by Rhines , 106—108 - , . , et al. for higher temperatures.

b) 99.999% nickel heated in hydrogen

Marz-grade nickel showed oxidation morphology identical to 99.996% nickel. Figure 57 shows the evolution of morphology for a mechanically polished specimen. As mentioned above, the oxidation of mechanically polished nickel is visually slower than on electropolished samples.

After 17 min (Fig. 57a) the oxide grain size is so small that the individual grains cannot be resolved at 10,000X.

After 1 hour, recrystallization has occurred and grain boundary diffusion leads to the now well known dimple structure (Figs. 57b and c). Similarly, metal grain boundaries could still be seen after 2 hours of oxidation.

Besides grain boundary diffusion, the observation after

1 day and 10 hours of growth related to dislocations created 168

Figure 57. Oxidation of Mechanically Polished 99,999% Ni at 680°C in Pq2 = 3xl0-4 atm after (a) 17 min (10,000X); (b) 1 hr and 17 min (10,0O0X); (c) 2 hrs (5,000X); (d) after 3 days, 5 hr and 38 min (20,000X). 169

in earlier stages explains the presence of whiskers and overgrowth at the scale/gas interface (Fig. 57d). From

the last micrograph, dislocations seem to arise mainly upon

grain impingement through growth since no whiskers are

observed in the middle of oxide grains. It is shown later,

that whisker growth is enhanced by the internal stresses

and by the presence of water vapor in oxygen. In conclu­

sion, the two grades of pure nickel showed similar oxidation behavior; the different processes are summarized in Fig.

58. However, oxidation kinetics were faster for electro­ polished nickel than for mechanically polished nickel.

2. Oxidation of nickel at high temperatures T > 950°C)

In the high temperature regime, the observed oxidation morphologies depended greatly on the degree of humidity contained into the oxygen gas. Three different types of oxygen were used: "extra dry" oxygen without further clean­ ing was used initially. Alternatively, the "extra dry" oxygen could be bubbled through a liquid nitrogen trap before insertion into the HSESEM chamber to eliminate any traces of water vapor. Finally, the extra dry oxygen could be passed through a water bubbler in order to satu­ rate the gas with water vapor. Here, these three different kinds of oxygen are called, extra dry oxygen, super dry oxygen, and wet oxygen, respectively. As usual, the nickel 170

Nucleation naoOflgftftQ (UwQOQBOQQlft ^jO 0 < t < I Hour 3 Disolution N 6.B.

Grain Growth rtonr-T-9 Ih < t < I Day

and

G.B. Grain Boundary Diffusion

Stresses Generation and Whisker Growth I Day

Figure 58. Schematic Growth of NiO at 680°C in P_ = 3x10“ ^ atm . 2 171

specimens were heated in a hydrogen atmosphere or their

initial oxide scale was reduced. After the reaction

temperature was reached the specimen chamber was evacuated

to remove the hydrogen , Oxygen at P = 3xl0-4 atm was 2 then leaked into the HSESEM either after some considerable

duration of evacuation or else immediately after hydrogen

removal. A 30 min long vacuum treatment of the specimen

in the HSESEM vacuum between hydrogen removal and oxygen

introduction, would allow the hydrogen to diffuse out of

the specimen before oxidation was started. Three kinds of

specimens were examined in this high temperature range.

The 99.999% Marz grade nickel, nickel single crystals, and

the 99.99%.’nickel grid. All the results reported below

deal with mechanically polished specimens. Electropolished

specimens showed the same oxidation morphology as those

mechanically polished for all types of nickel. However,

the kinetics were faster for the electropolished specimen

(as for the low temperature).

a. Oxidation of polycrystalline 99.999% nickel

1) Hydrogen-heated specimen

a) Oxidation at high pressure P - 3x10-4 atm 2 Figure 59 shows the oxidation at 1110°C of mechani­ cally polished, polycrystalline 99.999% nickel heated in hydrogen. This sample was held for 30 min at 1110°C in the HSESEM vacuum after hydrogen was removed. Then extra dry oxygen was leaked into the chamber. Upon the intro­ duction of oxygen at 1110°C oxide grains nucleated on the surface and grew extremely rapidly laterally. After 3 min

(Fig. 59a), the metal is completely covered with oxide grains of 0.2 ym average diameter. Figure 59a shows a triple point of 3 metal grain boundaries. The oxide grains are clearly oriented within the individual metal grains, very much as for the initial oxidation of OFHC copper at

500°C (Fig. 15). At this temperature the supersaturation 5 135 ratio is only 1.78x10 , and nucleation of the 3D oxide probably follows between the oxide grains and the metal substrate. The oxide scale is perforated by holes 0.1 ym 7 -2 average dia 1.4x10 cm density randomly distributed over the surface. These cavities are not obviously related to the metal grain boundaries. The creation of voids and cavities has already been discussed in the literature 6 7 86 review. ' ' In short, 3 different mechanisms have been

proposed: 237-239 - j j metal vacancies migrate to grain boundaries or the metal/scale interface where they nucleate voids, 2) internal oxidation of carbon at Ni grain boundaries generates CO-CC>2 gas pressures high enough to form cavities, 3) stresses accompany the growth of the oxide layer causes plastic deformation and conse­ quent creep cavitation. Mechanism (1) does not apply to our case since the voids observed on Fig. 59a,b are located in the middle of the scale or at the scale/gas 173

Figure 59. Oxidation of Mechanically Polished, Hydrogen heated Polycrystalline 99,999% Ni at 1110°C in Pq 2 = 3xl0“4 atm after (a) 3 min (5,000X); (b) 11 min (5,000X); (c) 27 min (5,000X), (d) 40 min (1,000X). 174

interface. Mechanism (2) cannot be advanced to explain the

morphologies^ of our micrographs since the carbon content

of our specimen is quite low (15 ppm, cf. Table 3) and

since the cavities are not related to the metal grain

boundaries. Finally a mechanism such as (3) could certainly

apply for thicker scale but certainly not for that after

3 min. of oxidation. In a recent article Atkinson"*"^

investigated the distribution and the supersaturation of

defects through a NiO scale using a model in which the

majority defects are metal vacancies (V„,and electron J 2 Nl holes and minority defects are charged oxygen vacancies

(Vq *-), i.e.,

0 t Ni... X + 0 x + V " + V ' * (54) ^ Ni o m o

The defect concentrations were calculated by two differ­

ent methods:

- Constant current method: At the inner and outer sur­

faces of the film the concentrations of metal and oxygen

defects take their equilibrium values and their fluxes are

independent of position x, i.e.,

^ = ■^2=0 (55) dx dx

It is assumed that there is no sources or sinks for lattice point defects such as dislocations or grain boundaries

- Equilibrium method: As in the original theory of

Wagner it is assumed that there are sufficient sources of 175

sinks for defects in the film for local equilibrium to be

attained through it, i.e., (from Eq. (54)),

K = C C (56) eq m o

Atkinson calculated that at 1000°C the constant cur­

rent and equilibrium methods give almost the same profile

for Ni vacancies. However, the oxygen profile is very

different for the two methods. In particular, the con­

stant current method predicted a very large supersaturation 3 of oxygen vacancies (by 'v 10 ) near the mid-point of the

oxide film. Such a large supersaturation could obviously

lead to void generation. As discussed by Atkinson, if

diffusion is mainly along grain boundaries, the same pro­

cedure can be used since grain boundaries are ordered 240 structures where point defects are stable. As a matter

of fact, the cation defect concentrations in the boundary 236 are identical to those of the lattice. 236 Atkinson and Taylor observed cavities in the middle of Ni oxide scales having a diameter as large as 1.5 ym. 241 242 Furthermore, Urban ' observed dislocation loops and and square voids in NiO. The formation of clusters (dislocation loops) was followed by the growth of cubic-like cavities with edges parallel to <100> direc­ tions . They showed the characteristic diffraction behavior 24 3 obtained from voids in the middle of a thick scale.

He concluded that the voids were probably stabilized by 176

gases during the growth process up to dimensions where a

collapse of the voids to form a dislocation loop became

impossible. The morphology shown in Fig. 59a corresponds

to such a mechanism where voids nucleated on oxygen

vacancy loops were stabilized by gases. It was already mentioned that neither CO nor CO2 can play a role for this

growth. The presence of sulfur is also known to contribute 144 to the formation of voids m NiO scales. However, our

99.999% Ni contained too little sulfur (0.25 ppm) to

account for void formation. Consequently, oxygen vacancy condensation leads to the formation of the initial voids by incorporation of oxygen. Initially these voids nucleate

in the middle of the scale and grow by vacancy incorporation.

Ultimately they open up to scale/gas interface and mor­ phologies such as in Fig. 59a are seen.

As the scale grew the cavities seen at the scale/gas interface progressively disappeared because of the oxide growth (Fig. 59 b and C).At latter stages, supersaturation of oxygen vacancies creates only dislocation loops within the scale but the absence of gases in contact with the nuclei prevents the growth of such cavities. At the same time, the growth of whiskers and platelets was observed after 10 min of oxidation (Fig. 59b,c and d ) . Initially, the whiskers nucleated above the top of the metal grain boundaries (Fig. 59b and c), but after 30 min. the oxide scale was covered almost uniformly with whiskers 177

(Fig. 59d). After 6 hours of growth at 1110°C, both the whiskers and the underlying oxide scale had grown exten­ sively and the average grain size was 5pm, Fig. 60a.

Fig. 60b shows a cross-section of the scale formed on

99.999% nickel at 1110°C for 6 hours, as seen in an SEM at room temperature. Platelets are clearly visible at the top of the scale (5pm average height), and it is clear that the scale itself is composed of an outer layer 20pm thick and an inner porous layer 4pm thick. It is shown later that whisker and platelet nucleation occurs at structural de­ fects such as dislocations or grain boundaries. Armanet 180 et al. observed the formation of whiskers in moist oxygen. In order to confirm this observation, it was decided to use super-dry oxygen instead of extra-dry oxygen.

Figure 61 shows the oxidation behavior of 99.999% Ni -4 oxidized in P super-dry oxygen at 1100°C in P = 3x10 2 2 atm. The initial formation of cavities seen at the scale/ gas interface was again observed (Fig. 61a), but no formation of whiskers was detected even after 5 hours of oxidation. The final morphology (Fig. 61c) was very similar to the morphology observed on Fig. 60 without the whiskers.

The grain in the lower left corner of Fig. 61c shows interesting pits attributed to dislocation lines emerging at the scale/gas interface (see later). In another experiment to confirm the effect of water vapor on the 178

porous Inner scale

Figure 60. Oxidation Morphology of Mechanically Polished Hydrogen Heated Polycrystalline 99*999% Nickel after 6 hours at 1110°C in Pn = 3x 1 0 ”^ atm 2 (a) (4,300X), (b) Cross Section (1,000X). 179

Figure 61. Oxidation of 99,999% Ni in Superdry Oxygen at 1100°C in P0 2 = 3x10“^ atm after (a) 1 min (2,000X); (b) 3 hours (2,000X), (c) 5 hours (20,OOOX). 180

oxidation of nickel, oxygen was passed through a water

bubbler to obtain moist oxygen.

Not only were whiskers randomly distributed on the

surface, but the oxide grains became facetted with prefer- 180 ential growth directions (Fig. 62). Armanet et al. observed similar facetting when nickel was oxidized in wet oxygen.

For all the experiments at high temperature described thus far, the specimen were heated in hydrogen, maintained in the HSESEM vacuum for 30 min to remove the hydrogen and then oxidized. Since water vapor has been shown to influ­ ence greatly the oxidation morphologies the avoidance of the intermediate annealing in vacuum and consequently the formation of water vapor from the reaction of oxygen with hydrogen in the metal should affect the nucleation of

NiO on Ni. The local vapor pressure of water vapor on the specimen surface can be estimated initially, knowing the standard Gibbs energy change for the reaction

H2 + l/202 = H20 (57)

The standard free energy of Eq. (57) is

AG57 = "RT ln (58) W where P is the equivalent partial pressure when the 2 -5 specimen was heated in hydrogen (P„ - 1x10 atm) 2 181

Figure 62. Oxidation Morphology Observed on 99,999% Nickel after 5 hours of Oxidation at 1100°C in Wet Oxygen (P = 3>i0_4 atm) (4,720X). 2 182

o 244 AGc-y is given by Schwedtfeger and Turkdogan

AG°7 = -249640 + 57.IT J. (59)

-4 At 1110°C for P = 3x10 atm, the initial water vapor 2 pressure from Eq. (58) is

P„ _ = 4.84 x 10-1 atm (60) 2

Consequently, the initial gas mixture before any scaling is composed of oxygen, hydrogen and water vapor with vapor -4 -5 -1 pressures of 3x10 atm, 1x10 atm and 4.84x10 atm.

Figure 63 shows the oxidation of 99.999% Ni oxidized in super-dry oxygen at 1100°C immediately after hydrogen heating. As in previous experiments (cf. Fig. 59a), oriented oxide nuclei grew on the Ni grains and after 3 min completely covered the surface. However, no cavities were seen at the oxide/gas interface. On the contrary, small 7 - 2 nuclei 0.1 ym average dia. with a density of 1x10 cm were seen on the top of the oxide grains after 3 min

(Fig. 63a and b). These nuclei grew as whiskers after

6 min and in Fig. 63b,c, showing a triple point of three grain boundaries, it is seen that metal grain boundaries were not preferential sites for whisker growth. After

1 hour (Fig. 63d) the whiskers entirely covered the initial oxide scale.

Contrary to Fig. 59 where the presence of oxygen stabilized the voids,nucleation from vacancy dislocation 183

Figure 63. Oxidation Morphologies of 99,999% Ni Oxidized in Superdry Oxygen Immediately after Hydrogen Heating at 1100°C (a) after 3 min ( 5,000X), (b) after 5 min (2,000k), (c ) after 22 min (2,000X), (d) after 1 hour 15 min (2,000X). 184

loops, the very high pressure of water vapor (4.83x10 ^

atm) suppressed the void stabilization and promoted

whisker growth. It was already mentioned that both water

vapor and dislocations have strong support for whisker

growth. The similarity between void density on a vacuum

annealed specimen and the whisker density of the last

experiment suggests that both types of features have a

common origin, i.e., dislocations created by vacancy conden­

sation in the oxide scale. The effect of water vapor is

not known but it obviously decreases the stability of the

{100} void walls and promotes the opening of dislocation

cores thereby promoting whisker growth (see Part V).

If moist oxygen is used instead of super-dry oxygen

just after hydrogen heating, whisker growth is greatly

enhanced and the initial oxide scale is completely covered

with thick whiskers or platelets after only 30 min of

oxidation (Figu 64). Extensive exposure to hydrogen or

slow heating before oxidation would damage the metal sur­

face (Fig. 65a and b) in a fashion similar to OFHC copper

(see Fig. 26). Significant porosity appeared irregularaly on the metal grains (Fig. 65a) and hydrogen penetrated more deeply at the metal grain boundaries than elsewhere.

Leaking of super-dry oxygen on such a surface produced

immediate growth of oxide whiskers (Fig. 65c). There was no observation of oriented nucleation of oxide grains. If moist oxygen was impinged on such a damaged surface instead 185

Figure 64. Oxidation of 99,999% Ni in Wet Oxygen Imme­ diately after Hydrogen Heating at 1100°C (a) after 30s (2,000X), (b) after 1 min (5,000X), (c) after 2 min (5,000X5, (d) after 25 min (2,000X). 186

Figure 65. 99 , 9 9 9 % Ni Heated in Hydrogen for 3 hours at 1100°C, (a) 2,000X, (b) 10,000X, (c) Oxidized for 30s in Superdry Oxygen (2,000X), (d) Oxidized for 1 hr. 26 min in Wet Oxygen (1,000X). 187

of super-dry oxygen, whiskers appeared immediately as

before, but after 45 min of oxidation at 1100°C, a differ­

ent type of NiO oxide whiskers 30 tjm long nucleated and grew

on the top of the oxide scale (Fig. 65d). The initial growth of small whiskers during oxidation in super-dry oxygen has

already been observed (Figs. 63 and 64) and attributed to water vapor formation. After some time, all the hydrogen

is removed from the sample and no water vapor is present in the super-dry oxygen. On the contrary, when wet oxygen is used, further nucleation and growth of oxide whiskers such as in Figs. 59 and 60 is possible.

(b) Oxidation at low p 2 The formation of cavities in nickel oxide scales such as in Fig. 59 was a strong function of oxygen pressure.

Figure 66 shows the oxidation of 99.999% nickel heated in hydrogen, vacuum-annealed at 1100°C, and then oxidized in _ g super-dry oxygen at a pressure estimated to be 1x10 atm.

There was no formation of cavities upon leaking of the oxygen. However, the oxide scale had very strongly ori­ ented relationships with the substrate (Fig. 66a). Because extra-dry oxygen is not completely water vapor free, whisker growth was observed after 5 min of oxidation

(Fig. 66b) and whereas the initial grains did not nucleate at metal grain boundaries, these boundaries were preferen­ tial nucleation sites for whiskers (Fig. 66c). The increased tendency for epitaxy at low oxygen pressure can Figure 66. Oxidation of 99,999% Nickel at 1000°C in

Extra Dry Oxygen Pq2 1 x 1 0 “ ^ atm, (a) after 40s (5,000X), (b) after 18 min (5,000X), (c after 22 min (10,000X). 189

again be rationalized in terms of supersaturation

ratio : as the oxygen pressure decreases, the supersatur­

ation ratio and the driving force for oxidation both

decrease leading to a more heterogeneous nucleation. The

cavities did not form initially and there is obviously a

minimum gas pressure below which the cavities are not

stable.

Oxidizing 99.999% Ni in a low pressure of wet oxygen

after hydrogen heating and vacuum annealing showed similar

features (Fig. 67). However, the oxide whiskers or

platelets were definitely nucleated at metal grain boundaries (Fig. 67a) and at scratches made previously on the sample where dislocation areas were generated by cold work. In Fig. 67a scratches run from the lower left corner to the upper right corner and numerous dislocation- related overgrowths are generated.

For the high water vapor contents and low oxygen pressures, the oxide overgrowth resembled platelets more than whiskers (Fig. 67b).

2. Oxidation of 99.999% Ni where an initial oxide scale grown during heating is reduced in hydrogen.

In order to reduce hydrogen diffusion during heating, experiments were carried out in which an oxide scale was grown during fast heating inside the HSESEM chamber and then reduced in super-dry hydrogen. As soon as the under­ lying metal substrate was free of oxide and clean, 190

arain boundary

Figure 67. Oxidation of 99,999% Nickel at 1000°C in Wet Oxygen Pq 2 - 1x10-7 atm (a) after 20 min (1,000X), (b ) after 25 min (5,000X). 191

oxygen was leaked to study oxidation.

(a) Reduction of thick oxide scales

The reduction of thick oxide scales by dry hydrogen was investigated and as an example, the reduction at 950°C of oxide scale formed at 680°C on 99.999 Ni for 6 lays,

such as in Fig. 57d, is described. Figure 68a shows the initial oxide scale formed on 99.999% Ni after 6 days of -4 oxidation at 680°C m P = 3x10 atm. After an incuba- 2 tion time of about 10 min, small cavities were seen at oxide grain boundaries (Fig. 68b). As described above for the reduction of cuprous oxide, the reduction of NiO is more favorable at the metal/oxide interface and more precisely at the bottom of oxide grain boundaries since diffusion is faster along these short-circuit diffusion paths.

Later on, these cavities grew to be as wide as 1 pm

(Fig. 68c) and small nuclei are clearly seen on the top of the oxide grains and particularly at their boundaries.

These are metal nuclei formed from the reduction of NiO.

Initially hydrogen consumes the oxide at the metal/oxide interface, but ultimately depletes the outer portion of the scale. After 2 hrs and 35 min, all the oxide has been consumed (confirmed by post oxidation X-ray analysis), and the surface morphology is shown in Fig. 68d.

All four micrographs of Fig. 68 describe the same spot characterized by the metal twin boundary at the bottom. 192

Figure 68. Reduction of a Ni Oxide Scale Grown at 680°C (i.e. Figure 57) and reduced at 950°C in dry hydrogen, (a) initial oxide scale grown (5,000X) after 6 days and then reduced; (b) for 15 min (5,000X), (c) for 45 min (5,000X), (d) for 2 hrs 35 min (5,000X). 193

NjO 0

I0mn

t H H H Ni Reduced From NjO

30mn

Figure 68 (continued), (e) Schematic of the Reduction of a Thick NiO Scale. 194

The mechanism for thick scale reduction is described in

Fig. 68e.

Obviously such a metallic substrate (Fig. 68d) is not practical to use for further oxidation and consequently only the reduction of very thin oxide scales (formed during

fast heating in vacuum) was used for further oxidation.

(b) Reduction of thin oxide scales

The reduction in hydrogen of thin oxide scale such as shown in Fig. 69a formed during fast heating to 1100°C in vacuum on 99.999% Ni led to a clean metal substrate

(Fig. 69b). Such a starting surface was adequate for further oxidation studies. Obviously, the surface was covered by tiny pure metal particles not resolvable in the

HSESEM.

(c) Oxidation at 1100°C of clean metallic substrates obtained by reduction in H ^

As soon as super-dry oxygen was leaked to the metallic substrate of Fig. 69b, random nucleation of very fine oxide grain was observed, Fig. 70a. After 10s the average grain size was 0.4 ym. This nucleation morphology was obviously very different from Fig. 59, for example. The difference is attributed to the metallic particles randomly distributed on the surface and produced during reduction. Instead of an oriented nucleation of oxide on the metallic substrate (Fig. 59) the initial oxidation in this case occurred from the reaction between these highly 1 95

Figure 69. (a) Thin Oxide Film Formed During Heating of 99,999% Ni at 1100°C in Vacuum (2,000X); (b) Metallic Substrate Obtained by Reduction of the Above Oxide Scale in H- at 1100°C for 35 min (5,000X). 196 reactive metal particles and oxygen. Obviously after con­ sumption of these particles the oxide scale grew by outward diffusion of cations originating from the substrate. The consumption of the metal particles by the oxide lasted approximately 6 min and Fig. 7 0b shows the final stage of this process where the oxide grains observed on the sur­ face did not grow particularly at their tips. Later on, the oxide grain grew mainly at their tips and the cations consumed came from the substrate by diffusion (see, for example, grain A in Fig. 70a,b,c and d).

Figures 70e and f show an interesting phenomenon of apparent new nucleation (grain marked B) . it is shown in part V that new grains appear at the scale/gas interface if the growth of surrounding grains is stopped by disloca­ tion slip or tunnel heating. Consequently, the kind of sample pretreatments already used by several authors (see, 6 0 for example, Graham ) do not give give meaningful data on the growth of NiO on bulk metal. Instead, the tiny metal particles formed during reduction change the fol­ lowing growth morphology (for example, no oriented oxide grains are observed) and change the kinetics. Figure 71 shows the same oxidized spot after longer times and very high magnification (20,000X). The oxide grain in the cen­ ter of the micrographs grew mainly at its tip. This does not suggest either diffusion of the cations through the lattice or grain boundary diffusion, at least through Figure 70. Oxidation 99#999% Ni at 1100°C (the Oxide Formed during Heating was Reduced by H2, i.e. Fig. 69), (a) after 10s, (5,000X), (b) after 6 min (10,000X), (c) after 1 hr 8 min (10,000X), (d) after 2 hrs 11 min (10,000X). 198

4 *?

1. ' i

N gbrf ■* s t + - 4 ^ 4

fievMiuc 16d t i orfl

(l ) 2 p m

Figure 70 (continued) 5,000X; (e) 2 hrs 40 min; (f) 3 hrs 10 min. *VI

Figure 71. Oxidation of 99,999% Nickel in Pq 2 = 3x10 4 atm at 1100°C (20,000X) after (a) 2 hrs and 52 min; (b) 3 hrs and 20 min and (c) 4 hrs and 19 min. 200

the outer part of the scale. On the contrary, diffusion along dislocation pipes seems more relevant to explain such growth. In the high temperature regime, lattice diffusion through the scale is generally believed to lead to a flat scale. However, our results seem to contradict this general belief. As a matter of fact, dislocation pipe diffusion has been observed at temperatures as high as

1350°C during the growth of nickel oxide. Figure 72a shows the morphology of the surface after oxidation of 99.999% nickel in super-dry Pn at 1350C for 15 min. The metal 2 surface is immediately covered with oxide grains and after

15 min, the average oxide grain size is 5 ym. The scale is obviously not flat, and each grain shows surface irregularities-. Very high magnification photos taken in the SEM after cooling down showed either pyramids (Fig. 72b) or pits (Fig. 72c) at the top of the oxide grains. This was confirmed by taking stereopair photos with a 10° dif­ ference in tilt angle. Sometimes the pits or the pyramids were elongated (Fig. 72c). Both features (i.e., pits and pyramids) could be present within the same grain. For example, the growth features marked P in Fig. 73 is a pyra­ mid but the features at its top and left are pits. The two micrographs of Fig. 73 are stereopairs with a 10° tilt angle difference. These growth features are similar to the ones observed earlier (for example, the oxide grain at the bottom right of Fig. 61 shows 5 distinct pits on 201

Figure 72. Oxidation<^99,999% Nickel in Pq ^ = 3 * 1 atm at 1350°C after 15 min (a) 5,OOOX, (b) 23,000X, (c) 23,OOOX, (d) 18,OOOX. 202

Figure 73. Stereo Pairs of 99,999% Ni Oxidized for 35 min in Pq 2 = 3xl0“4 atm and at 1300°C 20,OOOX. 203

one of the faces). Obviously the kinetics above 1300°C

are much faster than at 1100°C and any growth features

are more distinct. As explained in detail in part V,

these growth morphologies arise because of the presence of

dislocations in the scale. The screw component of these

dislocations provide steps for scale growth initially.

After all the steps have been consumed, the scale becomes

flat and growth takes place by lattice diffusion through columnar grains. A flat scale, already observed by many authors studying nickel oxidation at high temperatures

is consequently obtained (see for example, ref. 180, 236 245 and 245). Stott et al. observed similar indentations or steps on oxide grains for nickel- and chromium-implanted polycrystalline nickel oxidized- for 24 hou-rs* in P = 2 1 atm in the range 950-1150°C. Part V of this thesis attempts to rationalize such irregular scale growths.

b) Oxidation of 99.99% nickel single cyrstals of~random orientaTtions

The oxidation behavior of nickel single crystals was studied for comparison with polycrystalline material.

Figure 74 shows the oxide morphology grown on a -4 99.99% nickel single crystal at 1100°C in -P = 3x10 atm. 2 The specimen was heated in hydrogen to 1100°C, vacuum annealed at this temperature for 30 min and then oxidized.

Figure 74a is very similar to Fig. 59a, both cavities and oxide growth were simultaneously observed. The oxide 204

Figure 74. Oxidation of 99,999% Nickel Single Crystals at 1100°C in P0 2 ^ 3xl0“4 atm (Superdry), (a) after 2 min (5,OOOX), (b) after 3 min (2,OOOX), (c) after 20 min (15,OOOX), (d) after 2 hrs (2,OOOX). 205

Figure 74 (continued). Water Vapor was Added to the Super­ dry Oxygen after 2 hours (Fig. 74d) of Oxidation (e) 15 min after addition (2,000X), (f) 15 min after addition (5,OOOX), (g) Cross section of the Scale at Room Temperature 20 min after Addition (1,OOOX). 206

grain size after 3 min was about the same for nickel

single crystals and for polycrystalline nickel. An orient­

ing of the oxide grains to the substrate was observed in both cases. However, the cavity density was slightly larger in the case of oxide grown on single crystals

(Fig. 74b). Dislocations and grain boundaries of both the metal substrate and its oxide act as vacancy sinks in the oxide scale. The crystallographic defect density is probably greater for a polycrystal than a single crystal.

Consequently there is a greater number of vacancy sinks in the oxide for polycrystalline nickel and thus a smaller cavity density.

After 20 min the oxide grains outgrew the initial cavities (Fig. 74c) and after 2 hours, the oxide scale shows a compact outer surface (Fig. 74d). Again each individual grain shows pits or pyramids at its surface before the scale becomes flat (Fig. 74d).

The effect of water vapor in the oxygen was the same for single crystals as for polycrystalline nickel. Water vapor was added to the oxygen after two hours of oxidation.

A scale similar to Fig. 74d was immediately covered by very long whiskers (Fig. 74e). After 2 min, whiskers could be

30 pm .long and stop growing. This corresponds to a

15 y mn ^ growth rate; surface diffusion along a central tunnel was certainly the rate determining step for the growth of a single whisker. A sealing of this central tunnel stopped the diffusion and the whisker stopped 207

.growing (see part V). When the whisker stopped growing,

lattice diffusion through the oxide can still widen the

whisker bottom (Fig. 74f). A cross-section of the scale

made at room temperature (Fig. 74g) clearly shows whiskers

at the scale/gas interface and voids at the metal/scale

interface. This void arose from the diffusion of cation 232 vacancies to the metal/scale interface.

In conclusion, the oxidation behavior of polycrystal­

line nickel and nickel single crystals is identical.

c) Oxidation of a 99.99% nickel grid

Oxidation of nickel single crystals were performed in

order to supplement the information concerning the effects

on oxidation of crystallographic defects such as disloca­

tion or grain boundaries. For the same reason, the oxida­

tion of a 99.99% nickel grid was investigated. A nickel

grid (screen) possesses apparently a high density of

defects resulting from the cold work in its fabrication.

As soon as super-dry oxygen was impinged on a nickel grid -4 at 1000°C in P = 3x10 atm, very small oxide nuclei 2 nucleated on the surface which could not be resolved with

the HSESEM. However, rapid grain growth occurred and after

25 min of oxidation the average oxide grain size was 1 ym

(Fig. 75a). After 55 min, the surface was uniformly

covered with pyramidal oxide grains 2 ym average dia

(Fig. 75b). Further grain growth did not occur and 208

Figure 75. Oxidation Morphologies Obtained when a 99#99% Nickel Grid is Oxidized in Superdry Oxygen at 1000°C in PQp = 3x10“4 atm (a) after 25 min (2,OOOX), ( b y after 55 min (2,OOOX), (c) after 2 hours 7 min (7,OOOX). 209

Fig. 75c shows the post-oxidation morphology on the nickel

grid. Each oxide grain consists of a pyramid; there do

not seem to be any orientation similarities between grains.

Most of the pyramid faces are facetted parallel to the basal plane. The large nucleation density is obviously

related to the large dislocation density of the metal sub­

strate. However, dislocations are also generated during oxide growth and these promote the growth of differently oriented grains.

Simultaneous to the growth of pyramidal grains, numerous whiskers were seen to grow at the edges of the grid (Fig. 76a). Since any hydrogen in the sample was removed by intermediate annealing and since super-dry oxygen was used, this whisker growth is not explained by the presence of water vapor at the surface. However, whisker growth is often associated with stresses in a material and the edges of the grid correspond indeed to a high stress region. After 1 hour of oxidation in super dry oxygen, whiskers were seen to nucleate at the top of the oxide grown on the grid (Fig. 76b). These whiskers are related to the stress buildup in the oxide scale.

These whiskers were quite dense with a rectangular or square cross-section (Fig. 76c). A typical longitudinal length was 30 ym and each side was about 2 ym wide. Their growth rate corresponded approximately to an increase in length of 2 ym nm ^ . 210

Figure 76. Oxidation Morphologies Obtained when a 99,99% Nickel Grid is Oxidized in Superdry Oxygen at 1000°C in Pq 2 = 3xl0-<* atm (a) after 10 min (50X), (b) after 1 hr, 45 min (2,000X), (c) 5 min after Water Vapor is Added to the Superdry Oxygen (1,000X), (d) (2,000X). When water vapor was added to the oxygen a different

type of whisker was seen (Fig. 76c and d). These

whiskers were very long (200 pm) and extremely small in

cross-section. Their growth rate could be as high as

50 pm min Figure 76c shows both types of whiskers.

Such whiskers were already seen previously (for example,

Fig. 74) and are associated with the presence of water vapor in the oxygen. Some whiskers showed intricate morphologies (Fig. 76d) and the growth mode for such cases

is difficult to imagine. An extensive discussion of whiskers growth will appear in the next section.

In conclusion, the morphologies observed on poly­ crystalline nickel were reproduced on a nickel grid and the effect of defects in both metals was again to provide ledges for growth of pyramidal grains or whiskers. The growth of whiskers was promoted both by dislocations in the absence of water vapor and by water vapor content. V. IRREGULAR OXIDE GROWTH MECHANISMS AND THE

EVOLUTION OF SCALE MORPHOLOGY

A. Introduction

It is generally agreed that, at high temperatures and large scale thicknesses, the rate limiting step for pure metal oxidation is the lattice diffusion of point defects through the scale. The crystallographic defects present in the scale (i.e., dislocations or dislocation arrays) should not influence the diffusion of the reacting species through thick scales at high temperatures. Consequently, an oxide scale grown at high temperature would be expected to exhibit a flat scale/gas interface and be composed of columnar grains. It is also generally agreed that at rather low temperatures, the oxide grains are typically shallow- dish shaped, w^th slight ridges at their boundaries because of predominant grain boundary diffusion.

From the observations of the oxidation of copper and nickel inside the HSESEM, these ideal morphologies were only seen after very long oxidation times and at film thicknesses well beyond those for which the thick scale growth mechanisms should prevail. For example, nickel exhibited an irregular oxide scale at temperatures as high as 1350°C and after 40 min of oxidation (Figs. 72,73).

212 213

Consequently, it is inferred that, although lattice

and grain boundary diffusion are involved during the

oxidation of a metal, the crystallographic defects (dis­

locations) present on the metal substrate or created during

oxide growth, lead to irregular growth morphologies.

B. Morphologies and Growth Mechanisms

Irregular growth morphologies can be classified into

four different categories:

- The grain surfaces tend to be concave and the grain boundaries are delineated by shallow grooves (Figs. 13,

56e, 57b and c). Faster grain boundary diffusion in the

intermediate range of temperature (= 300°C for Cu and

700°C for Ni) has already been used to rationalize this 2 2 6 2 4 3 morphology by several authors. ' Schematics of the processes appeared in Figs. 14 and 58.

- The growth of oxide pyramids was observed on copper at 500°C (Fig. 21c,d, and e, and Fig. 28), on nickel grown on polycrystalline Ni at 1300°C (Fig. 73) and on a nickel grid at 1100°C (Fig. 75). More generally,protuberances were observed at the scale/gas interface on OFHC copper oxidized at 500°C (Fig. 29) on Cu oxidized at 920°C after hydrogen heating (Fig. 50) and on nickel oxidized at

1100°C (Figs. 70 and 71) .

- Oxide growth around a pit was seen on nickel oxidized at 1100°C (Fig. 61a), at 1350°C (Fig. 72), and 214

at 1300°C (Fig. 73) .

- Whiskers and platelets were grown on OFHC copper

oxidized at 300°C (Fig. 13c), at 500°C (Fig. 22), and on

polycrystalline nickel oxidized at 680°C (Fig. 50e,f,

Fig. 57d), at 1100°C in extra-dry oxygen (Figs. 59,60), in wet oxygen (Fig. 67), in super-dry oxygen immediately after hydrogen heating (Figs. 63,64 and 65), on nickel single crystals after water vapor was mixed with super-dry oxygen at 1100°C (Fig. 74e,f, and g), and finally on a nickel grid oxidized at 1000°C.

The first type of irregular growth morphology

(grain boundary grooving) has already been explained, and attributed to faster cation diffusion at grain boundaries.

On the contrary, the last three irregular morphologies, particularly the whisker growth, have not received a clear mechanism in the literature.

Several mechanisms have been proposed to describe the growth of whiskers from the vapor or from a solution.

The two principal mechanisms for whiskers grown from the 246 247 vapor are the VLS (Vapor-Liquid-Solid) mechanism ' and the so-called Frank mechanism extended by Burton 52 248 et al. ' In the VLS mechanism impurites form a liquid alloy droplet of relatively low freezing temperature at the top of the nucleating whisker. The liquid droplet is a preferred site for deposition from the vapor (point effect of diffusion), which causes the liquid to become 215 supersaturated with the vapor species. The whiskers grow by precipitation of this species from the droplet. As a consequence, whiskers grown by the VLS mechanism often exhibit an extended tip. Since whiskers observed during metal oxidation grow by impingement of oxygen with an outer oxide scale whose growth is limited by cation dif­ fusion, this mechanism does not apply. 248 52 According to Frank and Burton et al. screw dis­ locations take part in the growth of a crystal. A screw dislocation emerging on the face of a crystal provides it with a reproducible surface step. This step is self- perpetuating because of the screw dislocation (Fig. 77a,b).

Hence the steps needed for the growth of a crystal are provided by the screw dislocations and the need of 2-dimen­ sional nucleation never arises. This mechanism explained the growth of crystals at low supersaturations. If the step height is large or the radius of curvature of the spiral is very small, a whisker will form on the surface.

On the contrary, for larger radii of curvature, a pyramid will grow (Fig. 77a and b). 8 0 Frank showed that in the case of dislocations hav­ ing large Burgers vectors, the reduction of strain energy obtained by cutting out material so as to form a hollow tube at the dislocation core would outweigh - the surface energy needed to create the tube. The equilibrium hole radius is: 216

(a)

I

/ ~ P ! (b)

(c)

(d) L ^ .

Figure 77. Schematic Model for Growth of Surface Oxide by Screw Dislocation (after Verma77) . where b is the Burgers vector of the dislocation, y the

shear modulus and y the surface energy per unit area.

Simple theoretical models relating surface energy to the

ideal strength of show that for most g structures y/y should be about 1/4x10 cm. It follows

that hollow tubes should form around screw dislocation O cores only when b > 10 A. Holes centered on screw disloca- O tions with b = 15 A on SiC platelets surfaces were ob­

served by Verma.^

For crystals grown from the vapor, the adsorbed

species are, directly available for further step growth. On

the contrary for oxidation, the cations must be supplied

somehow via diffusion through the oxide scale. The dif­

fusion of metal species through a whisker or a pyramid may

follow three distinct paths (Fig. 78):

- Surface diffusion along the sides of the central

tunnel centered on the dislocation core to its tip. This

step would be followed by surface diffusion of the metal

species down the outside of the whisker until incorpora­ tion into surface steps.

- Lattice diffusion through the overgrowth followed by surface diffusion on the steps created by the screw dislocation. 218

a Of M,

Closed

Closed

Low Temperature High Temperature

Figure 78. Schematic Model for Whisker Growth (Low Temperature) and Pyramid Growth (high Temper­ a tu re ) . 219

- Grain boundary diffusion through the overgrowth

followed by surface diffusion to steps maintained by screw

dislocations.

1. Surface diffusion along the surface of the central tunnel

This mechanism was first proposed by Tallman and 249 250 Gulbransen ' for the growth of a-Fe2 C>2 whiskers on

iron at 400°C in dry oxygen (Fig. 79). TEM studies

have confirmed the existence of a central hollow tunnel within the a-Fe2 0 2 blades or whiskers grown on iron from 79 600° to 800°C in 20 torr of flowing oxygen. However, the tunnels eventually became blocked and further growth was stopped. It is therefore extremely difficult to ob­ serve open tunnels with a post-oxidation TEM examination.

In our study, nickel oxide whiskers grown on the nickel grid were examined in a TEM and although the presence of a cen­ tral tunnel could sometimes be revealed, most of the whiskers looked compact because of tunnel healing.

However, charging up at pyramid tips were observed during OFHC copper oxidation (Fig. 21c). Such charging might result from a central tunnel as shown by Motjin.a and 207 Sugiyama on hollow Cr,. Si^ whiskers.

Another important feature of dislocations intersecting 8 0 free surfaces is the formation of pits. Frank developed a model for the equilibrium shape of a pit created by a dislocation intersecting a free surface. Anisotropic Tip growth due to internal diffusion

io2< i

Lateral growth due to R ~ 5 0 A or more lottice diffusion of iron outword Growth Site

Normol Oxide Metal

Figure 79. Model of Whisker Growth during Oxidation of Iron, 400°C. in dry 02. From Tallraan and Gulbransen.250 221 surface energies were later considered by McLean and 251 Hirth. According to these models, the shape of the pit minimizes the sum of line tension and surface energies.

The pits observed on NiO (Figs. 61a,72,73) are believed to be dislocations intersecting the free surface. As explained later, this phenomenon arises especially for dislocation bundles. If the screw component of the dislo­ cation provides oxide growth transport and steps, the pit will become a pyramid such as on Fig. 7 3 or a whisker.

These pits are frequently aligned and arise from disloca­ tion arrays such as oxide sub-boundaries (Fig. 72b). These dislocations were created during oxide growth and are not related to the metal substrate. The best example of dis­ locations aligned on specific growth directions and leading 252 to pyramidal growth is provided by the work of Jungling for the oxidation of iron at elevated temperatures.

On the coptrary, whiskers and platelets observed on polycrystalline nickel at low P_ (Figs. 66 and 67) were 2 associated v^ith crystallographic defects of the metal substrate (i.e., metal grain boundaries or scratches on the surface). Similarly, nickel oxide whiskers grown initially on the nickel grid were nucleated on substrate dislocations. 222

2. Lattice diffusion through the overgrowth followed by surface diffusion of the metal species to surface steps

At high temperatures lattice diffusion is believed to be the rate determining step for the oxidation of pure metals. However, screw dislocations probably still pro­ vide the surface steps necessary for growth by the mechanism shown in Fig. 77a,b. For example, the formation of flat pyramids (Figs. 21d and e, 73 and 75) cannot be explained by surface diffusion along the central tunnel.

Pyramidal growth could be initiated by such a mechanism

(Fig. 21c), but when the central tunnel is closed, further growth will be limited by lattice diffusion followed by surface diffusion to the growth steps maintained by the screw dislocation. The pyramids then grow (extend) side­ ways instead of at their tips (Fig. 21d,e). The disloca­ tion may slip out of the crystal (Fig. 77c) and leave a trace on the pyramid (Fig. 77d). As soon as the disloca­ tion slips out, no further steps are created, so further growth must occur by a lateral advance of the preexisting steps, and the top of the pyramid flattens (Fig. 21d,e).

These processes explain the appearance of "new" oxide grains at the' scale/gas interface. An oxide grain growing around a dislocation could be hidden by faster growing grains around it. If growth of these grains is stopped by either dislocation slip or tunnel healing further growth of the original grain will predominate and will look like 223 a new nucleation event (Fig. 27, 28 and 70e,f).

A flat oxide is obtained only when all short-circuit paths have been exhausted (Fig. 30). Obviously, a flat oxide scale composed of columnar grains occurs only after relatively long times of oxidation even at very high temperatures. However, if a thin oxide scale is previously grown during heating before the growth of a thick oxide scale, the oxide grains formed at high temperature become flat in a short time, and grain boundaries and screw dis­ locations do not provide growth steps.

3. Grain boundary diffusion

Other paths for the diffusion of metal species through the oxide scale are the oxide grain boundaries

(Fig. 78, lower right). This grain boundary diffusion would necessarily be followed by surface diffusion to permit attachment at the steps provided by screw disloca­ tions. In this case, the pyramid will grow uni­ formly along its faces: The pyramids observed on OFHC copper oxidized at 500°C grew initially at their tips by the surface diffusion along the central dislocation tunnel

(Fig. 21c). When the tunnels closed, the pyramid tip flat­ tened out (Fig. 21d) and growth took place uniformly on the pyramid faces (Fig. 21e).

The pyramids grown on oxides were not regular spirals 52 as described by Burton et al. with 2 24

r = 2p 0 (62) c where p is the radius of the critical nucleus, but rather c > polygonized spirals (Fig. 21c,d and e). Polygonized spirals were formed because the rate of advance of the steps is dependent upon the crystallographic direction.

The growth spirals become distorted so that they exhibit 253-257 the symmetry of the crystal face.

Monatomic steps could obviously not be resolved in the

HSESEM, but several effects can explain our observations of steps on pyramid faces (Fig. 21c,d,e). For example, the pile-up of monatomic steps to form macroscopic ledges through the presence of surface impurities is well known.

Macrosteps could also arise from dislocation bundles.

Each macrospiral consists of a group of cooperating single spirals originating from screw dislocations situated closely together in a small center with their separation less than 260 about 19/2 pc - Pyramids grown from the vapor phase or 261 from solutions from a bundle of dislocations exhibit numerous etch pits on their faces. Figure 7 2b is an example of such an oxide pyramid grown on nickel oxide at

1350°C. The pyramidal base grew around a bundle of dislo­ cations. Each of these dislocations produced an etch pit where they intersected the free surface. Some of these pits could correspond to the edge component of a disloca­ tion. The edge components of dislocations do not partici­ pate in oxide growth. Figure 7 3 shows similar morphologies F ig u re 80. 226

a (d)

( b ) e) ecrew dulocolion I edge (Allocation

nb 4J <:»

c) f)

Figure 81. Process of Formation of a Dislocation Having a Large Burgers Vector on the Crystal Surface. From Gotoh and Komatsu.265 227

to Fig. 7 2; in addition to the pits created at the top of

the base pyramid, there is a secondary pyramidal growth on

the lower right side of the grain.

Generally, only a small number of spirals are seen to

develop. This is obviously due to competition between

spiral layers originating from different sources.

In conclusion, macrosteps are generally supplied by

bundles of screw dislocations with a definite (+ or -)

sign. Most of the spiral growths observed in the litera­

ture corresponds to such bundles. However, monatomic step

heights were sometimes observed on oxide spirals grown from 26 2 a single screw dislocation. Sunagawa observed the O smallest step height (2.3 A) for a single molecular layer

on the basal plane of natural hematite Fe2 C>2 .

Growth of pyramidal macrosteps could also be observed

if the growth takes place around a screw dislocation with a

Burgers vector n C where C is the interatomic distance o o , • j • j. j. 4-u i 263, 264 on the axis and r> is an integer greater than 1.

Figure 80 illustrates the spiral arrangements of atomic planes near a screw dislocation of Burgers vector |b| =

4C . A model for the formation of dislocations having o large Burgers vectors during crystal growth has been pro- 265 posed by Gotoh and Komatsu. According to this model, a dislocation having a large Burgers vector is introduced by the misarrangement in the combination of block crystals

(Fig. 81) . According to Eq. (61) , a screw dislocation 228

having a Burgers vector of n CQ has an equilibrium core 2 radius n times larger than a simple screw dislocation.

Dislocations of multiple strength consequently favor a

surface diffusion mechanism along the central tunnel and

were certainly responsible for the pyramids observed on

OFHC copper oxidized at 500°C (Fig. 21c,d, and e). These

pyramids did not grow around a dislocation bundle since no

etch pits were seen on their faces.

C. Influence of the Temperature

As explained above, lattice, grain boundary and sur­

face diffusion are responsible for pyramids and whisker 266 growth on oxides. As indicated by Turnbull and 267 268 LeClaire, ' if the rates of these three types of

diffusion in polycrystals are characterized by diffusion

coefficients, D^, D ^ and Dg and activation energies

Q,, Q , , and Q , then, D, < D , < D and Q, > Q , > Q . 1 gb s' 1 gb s W1 gb s This comparison is not possible for the diffusion of Cu

in CU2 O since no data are available for grain boundary and

surface diffusion. However, diffusion data for Ni in NiO

are more complete. Comparison of Eq. (4) and Eq. (6)

shows that D. < D , and Q. > Q , . There is not much data 1 gb 1 gb 269 for surface diffusion of Ni on NiO, but Dhallenne et al. were able to measure the lattice and surface diffusion coefficients of Ni on NiO using a grain boundary grooving

technique at 1400°C and 1520°C. Their values are given 229 in Table 4 and compared with those given by Eqs. (4) and ,,, 173,177,178 (6 ) . Since the agreement for lattice diffusion is excellent, it is possible to make a direct comparison between D , and D , with the result D , < D . This com- gb s gb s parison consequently holds for diffusion of Ni in/on NiO and in particular

(63)

177 178 It is interesting to note that Atkinson and Taylor ' measured a dislocation pipe diffusion coefficient for Ni in NiO given by Eq. (5) which is smaller than the grain boundary diffusion coefficient. Although pipe diffusion may be important in diffusion within oxides (for example, when surface diffusion is stopped by blocking of the tunnel), the model proposed here involves surface diffusion along the internal surface of a central tunnel centered on the dislocation core and not as pipe diffusion.

Since the' activation energy for lattice diffusion is larger than that for surface diffusion, surface diffusion will predominate as the temperature is decreased. The flux of metal species for both types of diffusion can be expressed:

(64) where Ac is the concentration difference across the scale and 6 is the diffusion distance. Although 6 is somewhat larger for whiskers than for bulk oxide, Eqs. (63) (64) Table 4. Lattice, Grain Boundaries and Surface Diffusion Coefficients^ • of * «•Ni m • «•NiO ^173, 177, 178, 269 173,177,178 269 Atkinson & Taylor Dhallenne et a l .

T°C/Dcm2g_1 D, D D„ D gb

1400° 4.2610"10 1.9710 6 4.510”10 4.510 6

-9 1520 1.410 4.4810 2.3410-9 1.3910-5 230 show that the flux associated with surface diffusion can be

several orders of magnitude larger than the flux associ­

ated with volume diffusion. However, the flux by surface

diffusion is limited by the size of the tunnel and the

sites for lattice diffusion are more numerous (Fig. 78).

The condition Q1 > Qg infers that as the temperature is

reduced, the ratio D^/Ds increases. Consequently, at low

temperatures, the growth associated with steps created by

a screw dislocation will lead to whisker growth (Fig. 78).

On the contrary, as the temperature is increased, lateral

lattice diffusion from the tunnel is favored, and oxide

growth leads to flatter morphologies such as pyramids

(Fig. 78). Accordingly for otherwise similar oxidation

conditions, whisker growth was observed during oxidation

of OFHC copper at low temperatures (T - 300°C, Fig. 13) whereas pyramids were observed at higher temperatures

(T > 400°C, Figs. 21 and 50). The same phenomena pre­ vailed for nickel although the morphologies were greatly

influenced by experimental conditions. For example, pyramid and whisker growth originated at dislocations was more predominant when the metal was initially cold-worked.

Whiskers were observed on a nickel grid oxidized at temperatures as high as 1000°C (Fig. 75). But only pyramids were observed at 1350°C.

As described above, the growth at the tip of whiskers or pyramids stops when the central tunnel is closed. Then 232 the growth proceeds mainly by lattice diffusion and flattening of the scale would occur. A flat scale was obtained when all the dislocation tunnels were consumed and thereafter growth was not related to dislocations but supported by lattice diffusion leading to flat and columnar oxide grains.

D. Influence of Water Vapor

The formation of whiskers was very sensitive to the water vapor content of the oxygen. When water vapor is introduced to oxygen either by direct mixing (Figs. 63,64) or by reaction of the hydrogen with oxygen initially con­ tained inside the specimen (Fig. 65), whisker growth is greatly enhanced. A flat oxide grown in super-dry oxygen would be covered with whiskers as soon as water vapor was added during the course of oxidation (Fig. 74e,f,g). This suggests that tunnels are always available. Similarly, although whiskers were grown at high temperature in super- dry oxygen on the nickel grid because of the high degree of deformation, new whiskers would nucleate and grow on single crystal Ni only after water vapor was mixed with oxygen (Fig. 76). The morphologies of the two types of whiskers were quite different (Fig. 75c). The whiskers grown in super-dry oxygen were short and had a relatively large cross-section similar to pyramids (Fig. 7 6b). On the contrary, the whiskers grown in oxygen saturated with water vapor were extremely long and thin (Fig. 76d) as if

water vapor promoted the growth along the axis. Armanet 270 et al. observed the same phenomena on a Ni-20Cr alloy

oxidized at 1000°C and 1200°C in air containing water

vapor. The growth of whiskers was directly attributed to

the moisture content of the oxidant gas. A related

phenomenon has also been observed here during the oxidation

of pure nickel in the temperature range of 1000°C to 18 0 1200°C. Armanet et al. proved that nickel was oxidized

more rapidly in wet atmospheres. Although this result was

not interpreted by these authors, the morphologies of NiO

grown in dry oxygen and in wet oxygen were very different. 180 The morphologies observed by Armanet et al. in wet. oxygen were similar to ours and dislocation-related growth was certainly more predominant in wet oxygen. A similar behavior was observed during the oxidation of mild steel 271 and of iron. In the latter case, the faster oxidation

in wet oxygen was attributed to better scale adherence.

The oxide possessed better plasticity because of the presence of hydrogen in the lattice. In our experiments, cross-sections of NiO scales did not reveal a difference in porosity between scales oxidized in different grades of oxygen (Figs. 60,79). Then faster oxidation is attributed to the greater tunnel fluxes for whiskers (or high 271 pyramids) in growth in wet atmospheres. Tuck et al. observed that this faster oxidation rate could not be 234

accounted for by accelerated diffusion of lattice vacancies

within the oxide. Similarly water vapor cannot influence

the surface diffusion coefficient along the central tunnel.

Consequently, the transport of metal species via surface

diffusion is not the rate limiting step for whisker growth,

but rather the break up of the oxidant molecules must limit

the kinetics. Lattice diffusion is still the rate limiting

step at high temperature for thick scales where the scale

does not grow by a tunnel mechanism.

In conclusion, because surface diffusion is so much

faster than lattice diffusion, the rate limiting step for

the high temperature oxidation of metals is the break up

of the oxidant gas molecules for whisker growth and lattice 272 diffusion otherwise. As a matter of fact Boreskov and 273 Wagner showed that dissociation of molecular was the

rate determining step for the oxidation of compact nickel

oxide at 250°C. In this case, dissociation of molecular

O 2 was envisioned as a step in series with lattice diffu­

sion. The transition temperature between diffusion as the

rate limiting step and the break up of the molecular

oxidant as the rate limiting step would be much higher for

surface diffusion (whisker growth) than for lattice diffusion (T = 250°C). Then at the tip of whisker centered on a dislocation tunnel, the rate limiting step in super- dry oxygen is: 235

02 (ads) * 20 (ads) (66)

and in wet oxygen there is a competitive reaction:

H20 (ads) £ H2 (ads) + 0 ads (67)

Whereas the forward rate constants for Eq. (67) are 274 available on some oxides (see, for example, ), there are

no rate data available for E q . (66) on NiO or Cu20 at high

temperatures. However, the break up of a H20 molecule is

generally found to be faster than that for most oxidant molecules. For example, the interfacial reaction in the

oxidation of iron to wustite in the temperature range 850-

1150°C is greater for water vapor-hydrogen mixtures than 275 for carbon dioxide-carbon monoxide gas mixtures.

Since the forward rate constant for (Eq. 67) is surely larger than that for Eq. (66), a metal species arriving to the tip of the tunnel will be more readily oxidized in moist oxygen than in pure oxygen (Fig. 78). The growth of oxide whiskers is consequently promoted by water vapor.

In conclusion, oxide whiskers grow predominantly by surface diffusion of metal species on the internal sur­ face of a tunnel centered on the core of a screw disloca­ tion or a bundle of dislocations. The rate limiting step for whisker growth is the break up of the adsorbed oxidant gas. Thus, the growth of whiskers is enhanced when water vapor is mixed with oxygen. This surface diffusion mechanism leads to the formation of pyramids when the 236 temperature is increased since competitive lateral lattice diffusion is enhanced (Fig. 78). Pyramidal growth is also obtained from lattice diffusion around a screw dislocation or a bundle of dislocations which provides the steps for growth (for example, when b is too small). Grain boundar­ ies may also provide diffusion paths before surface diffu­ sion to steps maintained by screw dislocations. When the screw dislocations have been consumed (slipping out of the crystal or healing of the central t u n n e l ) s c a l e growth proceeds by the consumption of existing steps. When these steps are consumed, the oxide scale becomes flat and diffusion through the lattice leads to columnar grains at high temperatures. Obviously, the oxide surface always has available sufficient ledges from screw dislocation to permit lattice extension. At moderate temperatures, grain boundary diffusion can be important and the oxide grains exhibit slight ridges at their boundaries. Finally, the extent of the transition time when dislocations provide growth steps is strongly influenced by specimen preparation. VI. SUMMARY AND CONCLUSIONS

The hot stage environmental scanning electron microscope (HSESEM) has proved to be a significant tool in the field of high temperature oxidation of metals.

With a minimum of maintenance, the HSESEM is able to run continuously and provides a tremendous amount of informa­ tion because, heating, oxidizing and analyzing are all performed in the same experimental setup. Consequently, several new mechanisms relevant to the high temperature oxidation of metals are proposed and some previous mechanisms are rejected or confirmed.

Although contamination of the specimen inside the

HSESEM chamber is greatly reduced with the use of liquid nitrogen traps, the HSESEM is not an ultra-high vacuum system and the initial metallic surfaces are often not free of oxides especially for reactive metals at low tempera­ tures. Despite the limitation for initial oxidation studies

(of clean surfaces), the HSESEM is a remarkable tool for the study of the growth of thick oxides.

Since both the surface pretreatment and the heating procedure greatly influence the oxidation behavior of OFHC copper and nickel, the conclusions drawn here are valid only for the experimental procedures used in this study.

237 238

A. Oxidation of OFHC Copper in Pure Oxygen

- At low temperature (T - 300°C), the high super­ saturation ratio for P leads to a very fine grain size 2 for the oxide scale. The scale seems to grow predom­ inantly by grain boundary diffusion of the cations and often exhibits whiskers at the scale/gas interface. These whiskers probably nucleate on screw dislocations ending at the metal surface or are initiated by plastic informa­ tion during scale growth. Whiskers may grow by surface diffusion of the metallic species along the inner surface of the tunnel situated around a dislocation core. The metallic species would then diffuse along the external whisker surface until it is attached to a surface ledge created and maintained by the dislocations.

- At intermediate temperatures, an oriented nuclea- tion of oxide grains is observed. These initial oxide grains are not related to the metal defects but rather to defects in the 2D adsorbed layer. However, where the metallic substrate is heavily cold worked before oxidation, the oxidation morphology exhibits very small grains grown at dislocations ending at the metal surface. These dislo­ cations or the dislocations created during oxide growth provide the atomic steps necessary for lattice extension and scale thickening. In this temperature range, pyramids rather than whiskers grow from these dislocations because lattice diffusion is more significant than at low 239

temperatures. However, so long as a tunnel at a disloca­

tion core is open, whiskers and pyramids result from a

similar mechanism. When the tunnel is ultimately closed, whisker growth is stopped and pyramid growth proceeds by

the consumption of initial steps. More importantly, the dislocations are a continuous source of growth steps so

long as they do not glide or climb out of the oxide crystal.

The scaling behavior is influenced by the heating and pretreatment procedures. The extent of the initial oxide grown during heating in vacuum influences the following oxide morphologies, e.g., the oxide grains do not impinge with each other when a thick initial oxide is grown during heating. When OFHC copper is heated for an extended period in hydrogen, the initial metallic surface is damaged and the resulting oxide morphologies relate directly to the microstructures of the metallic substrate.

- At high temperatures (T > 900°C), two simultaneous mechanisms are observed: the initial growth of solid cuprous oxide crystals exhibiting rapid grain growth and between and on the top of these grains, the formation of a solid copper oxide fume. A mathematical model shows that this oxide fume can initiate from bimolecular collisions between molecular oxygen and evaporated copper atoms in the gas phase. The oxide fume is comprised of tiny O (20 A dia.) cuprous oxide grains at high temperature which 2 40 transform to cupric oxide when the sample is cooled.

Consequently the post-oxidation morphology, previously observed by several authors, has not permitted a prior identification of this oxidation mechanism.

Cuprous oxide exhibits poor resistance to thermal cycling except at metal grain boundaries.

Reduction at high temperatures of a thick oxide scale take place initially at the metal/scale interface and at the oxide grain boundaries.

Finally, at very high temperatures, heterogeneous oxide nucleation at the metal grain boundaries is observed because of the small supersaturation ratio.

B. Oxidation of Nickel

Different grades of nickel were oxidized inside the

HSESEM and the morphological features of the oxide are rather similar for polycrystalline (99.996% and 99.999%) nickel and nickel single crystals. Electropolishing of the initial metallic surface does not influence the type of oxide morphologies but accelerates the kinetics of oxidation. However, the oxidation morphology of nickel is very dependent on the water vapor content of the oxygen.

- At low temperatures (T < 700°C), the oxide scale grown on nickel is comprised of very small oxide grains, randomly distributed over the initial metallic surface.

After recrystallization of the initial scale, the oxide 241

grain growth is quite slow and after an extended time of

exposure to oxygen, oxide whiskers grow on the top of

these oxide grains. Dislocations ending on the nickel

surface or created during oxide growth may provide ledges

for the growth of the initial nickel oxide grains and for

the growth of whiskers.

- At high temperatures, the oxidation of nickel is

greatly influenced by the water vapor content of the

oxygen. Oriented nickel oxide grains nucleate on nickel

when oxidized in super-dry oxygen. Simultaneously,

cavities appear at the scale/gas interface. These cavities

may nucleate on oxygen vacancies loops and are stabilized

by gases such as oxygen. After a short time, these cavi­

ties disappear during oxide growth. Even at very high

temperatures (1350°C), nickel oxide grains exhibit a

pyramidal shape, and growth steps are provided by screw

dislocations or bundles of dislocations created during

oxide growth or introduced into the metallic substrate by

cold work. These dislocations ending at the scale/gas

interface initially exhibit a pit because of the balance

between the dislocation line tension and the surface

energy. The screw component of the dislocations can then provide steps for oxide growth or open a large enough

tunnel at its core to permit significant surface diffusion of metal species. These metallic species reach the top of the tunnel by surface diffusion inside the tunnel, and 242

react immediately with the oxidant gas to extend the whisker or else diffuse down on the outside surface of the whisker before reacting to form pyramids. At the top of

the tunnel, the rate limiting step for oxidation is the breakup of the oxidant molecule. Far from the whiskers,

the rate limiting step is the lattice or grain boundary diffusion of metallic species through the scale. Since the breakup of a water molecule occurs more rapid than that for an oxygen molecule, whisker growth is greatly enhanced when nickel is oxidized in moist oxygen. When the tunnel is closed, the dislocation still provides steps for external oxide growth and the scale flattens out to form a pyramid. APPENDIX A

Between two surfaces 1 and 2 the net heat flux leaving

1 is :k ic E, ■■= E, + (1-ae

eQ = Absorptivity of the surfaces

The two surfaces are assumed to have the same e . 1- o represents the heat emitted by 2 on itself

E^ = heat emitted by 1

E2 = heat emitted by 2

We have the following system of equations

E* = E, + (1-e (f>) E* (A2) 1 1 o 2

E* = E 0 + (1-e )E? (A3) 2 2 o I

Solving for Ej and E2 we obtain

* E i + (1-e04>) E2 T7 _ ^______z.___ (A4\ bl “ l-(l-eo) (l-eo

* E 2 + (1-e0 4>) Ei Eo2 = A 1- i(l-eoT-----T (1-e 7 T --- TV (J>) (A5)

243 244

Consequently the net heat flux between 1 and 2 is

* * e0 — o ^ Q = Ei - e2 i- (l-eQ ) (l-eo

The factor 4> is independent of the temperatures and is

evaluated for T^=T2 .

If T1=T2, Q=0 s o that

E1 d> = (A7) 2

From heat transfer theory

E, = oe S, T, (A8) i = oeo “i n

E2 = oeo S2 T24 (A9) where a = Stefan-Boltzmann constant

S^,S2 = Surface areas of the surface 1 and 2, respectively

T^,T2 = Temperatures of the surface 1 and 2, respectively.

From (A7) and T^=T2

d> = ~ (AlO) 2

And (A6) becomes

2 4 4 oe S,(Tn -T- ) o = (A11)

For our geometry, the surfaces 1 and 4 represent the copper sample and the outside cylinder and the surfaces 245

2 and 3 represent the two molybdenum shields, Qi _2' ^2-3 and Q 3 _ 4 ' the net fluxes between these surfaces are written as (from All) 2 4 4 Q _ oeo S1

2 4 4 c e ^ S , (T- -T-. ) Qi1-2 t 1- i (1-e T-i ) T~/i(l-e *„ _) r- (A13) O O 2-3 2 4 4 ere S (T, -T. ) 0 = - -__-___ -______(Al4) Q3-4 1. (1-eo )(1-e0*3.ll)

We assume that the emmissivity for copper, molybdenum and stainless steel at their respective temperature are identical.4.- i 18 9

We define S.oe' 2 h i - 5. (A15) l-(l-e 0)(l-e0gi) 1

At steady state

Q12 = °23 = Q34 = Q (A16)

Using (A12), ... (A15), (A16) becomes:

B12^T1 _T2 * = e23(T2 _T3 * = ^34(T3 “T4 ) (A17)

(A17) consists of a system of 2 equalities and are solved 4 4 for T 2 or we obtain: 246

41 1 1 1 1 4 1 4 T22 B12 + B23(T“ + s~} 34 = [{F 34 ~ + ir-)T 23 1 + IT- 612 T4 4 ] (A18)

Q, the net heat loss given by (A16) is consequently equal

to:

1T14 ' T44] Q = -J----- h ---- ^ (A19)

612 623 334

Using (A15), (A19) becomes:

4 4>» oeo (Tl ‘T4 J 0 “ ° 2-e 2-e r ^ -

This heat loss- must be supplied by the current passing through the tungsten wire and

Q = P | I2 (A21) where p = Resistivity of Tungsten i Z = Length of the Tungsten wire

S = Cross section area of the Tungsten wire

I = Current passing through the Tungsten wire p is a function of the temperature:

p = Pq (1+ a) (Th-TQ ) (A22) where pQ = resistivity of Tungsten at room temperature

a = Tungsten resistivity coefficient

T^ = Heater temperature

Tq = Ambiant temperature

The combination of (A20), (A21) and (A22) gives (7). APPENDIX B

1) Evaporation of Ci^CMv) over (Ci^OCs)

Considering the reaction

Ci^CKs) * Ci^Ofv) (Bl)

The change in standard free energy is

^G G Cu20(v) G Cu20(s)

o o , g t "h o =

G°~H° 2) (B2) v T 'Cu20(s) where H° and H° are the atomization energies Ci^OCv) Cu20(s) of Ci^Ofv) and Cu2 0 (s) respectively.

Reaction Bl is the sum of four reactions of known atomization energies:

,215 CU20(v) - 2Cu(v) + 0(v) H° (Kcal/mole):135.673 (B3)

216 2Cu(s) + l/202 (g) = Cu20(s) H° (Kcal/mole); - 4 0. 309

(B4)

2Cu(v) = 2Cu(s) H°;-161.428216 (B5)

((v) = 1/2 02 (s) H° (Kcal/mole) --58.989216 (B 6)

Cu20(v) = Cu20(s) (Bl)

247 248

And

O o AH (Bl) = -125.053 Kcal/mole (B7) H0 "H0 Cu20(v) Cu20(s)

At 1200°K

O o gt "h o 215 - ( - V 2 ) = 33.358 Kcal/mole (B8) Cu20(s) and G°-H° - ( - V 2) = 72.817 Kcal/mole^^^ (B9) Cu20(v)

g t "h o Although the value of -(— -— ) is not explicitly Cu20(v) given in ref. 214 it can be deduced from the value of

GT~H0 - ( ?-) with the relationship CuO(v)

GT" H0 G>p-H0 x (.82)-1 at 1200 °K (BIO) Cu20 (v) CuO(v) derived from Figure 82 drawn using refs. 215 and 215.

From (B2), (B7 (B8) and (B9)

AG = 77702 = -2 (1200)In P (Bll) Cu20(V) choosing an unit activity for Cu20(s) .-15 Consequently Pcu20(v)/Cu20(s) (1200°K) = 8.6910 atm (B12) 0.98 H/O

0.94

0.92 L i/0

0.90 'MX(V)

0.88 - p r ^ ) M 2X(V) Q86

C/O 0 /S - - - B /0 0.84 At/O • • • • • Be/O F/O 0.82

Cs/O Cu/Se 0 8 0 Ag/Se

0.78 20 22

Figure 82. Ratio of Free Energy Function for and MX as a Function of 9 4 2 Temperature, from Ref. 214 and 215. 250

2) Evaporation of Cu(v) over Ci^CKs)

Cu20(s) = 2Cu(v) + 1/2C>2 (B13)

AG = 2U° + 1/2 U° - H° C u (v) U 2 UCu20(s)

G i ~ H o gt"hS gt"ho + T + 1/2 (— =—— ) - ( - V ^ ) (B14) C u (v) 02 * Cu20(s)

216 From the JANAF Tables AG is directly computed as:

AG = 106453 Kcal/mole = -2 (1200)In(P )2 (P )1/2 2 (Bl 5)

For P = 1*10 ^ atm 2

(Bl 6) PCu /Cu 2O(S) (1200°K> = 2 -33xl0-4

3) Evaporation of CuO(v) over Cu20(s)

Cu20(s) = CuO(v) + Cu(s) (Bl 7)

AG = Hg + HJ - HJ CuO(v) C u (s) uCu20(s)

GT"HS GT " Ho gt"ho 4- T (-TO) + ( T m ) “ (~m"“ ) (B18)

CU0(V) Cu(s) Cu2 ° (s) 216 From ref. 215 and JANAF Tables AG is computed as

AG = 69788 = -2 (l200)ln PCuQ(v) (Bl 9)

p \ //-< x (1200°K) = 2.35xl0~3 atm (B20) CuO(v)/Cu20(s) 251

Evaporation of Cu(v) over Cu(s)

Cu(s) * Cu(v) (B21)

G o — u o r* o _ tj o T O T O AG° = H°-H° + T (B22) v s T v 216 From the JANAF Tables

AG = 43502 = -2 (1200) In P (B23) Cu (v) and ,-8 P„ # / \ (1200°K = 1.34x10 atm (B24) Cu(v)/Cu(s)

(B12), (B16), (B20) and (B24) are summarized in Table 1. 'APPENDIX C

From the molecular theory of gases:

8 k T 0- 1/2 Vn = (-— -=■) (Cl) 2 02

8kT 1/2 and V (C2) Cu Trm _ Cu

277 If = T_. , Present showed the mean value of the 02 Cu relative velocity is

- 2 - 2 vo -

In our case, T / Tr and (C3) is not valid. The mean 2 value of the velocities of Cu atoms relative to the oxygen molecules is obtained by averaging VQ over the distribu­ tion functions of both gases. Thus

V_ = / / / dv dV„ dV_ / / / dVrt dV„ dV^ P_M . Cu HI Cux Cuy Cuz j j j 2y, 2 °2Z CU Cu V V Vo (C4) where

v o ■ f2+2] «C5, x y vcn ' Vp „ vr„ ' vn ' vn and are the components cux v' z 2 2 2 J x y z

252 253 of the velocities of copper and oxygen respectively.

P(V) = P(V , V , V„) is the probability that a species x y ^ has its velocity components between V and V + dV V J c x x x y and V + dv , V and V + dV . y y z z z

Using P(V)dVxd V d V 2 = I ® ! ^ ( V ^ t V ^ d V ^ dV (C4) becomes z

m +00 +00

5 ______) 3 / 2 (— i____)3/2 r dv r Cu l2TrkT„ 1 27TkT_ 1 / Cu ••• / dV Cu 2 J X J 2 z — OO —

, ,mCuVCu , m °2 V ° 2 „ (C6) exp * 2kTCu 2kT0

The computation is simplified if we transform V and 2 V„ to the relative velocity V_ and the velocity of the Cu J 0 J center of mass V^ M

Cr Vrs _ _ j _ _ mCu Cu 2 2 V^ = v„ -v v_ = -> u > u z— £ (Cl) 0 Cu 02 cM mcu+mo2

If we solve (C8) in component form

m 02 V = V +-(-- ) V Cux CM v V +mCu °x X ^ (C8) m r V = V - ---- ) V °2 <3. \ u +mo ’ °x X ^

Similar relationships apply for the y an d z component.

Finally 0o CU CU T 2nu m_ 0 2 2 = 2 [v - ——------V V + „ V (CIO) 2kT02 2kT02 CM mcu+m02 0 CM (m0 +mCu)2 0

MCuVCu , m °2V °2 ,mcu , m °2 ,„2 , m ° 2mCuV°VCM 2kTCu 2kT0, " 2kTCu 2kT0, CM k Z (Cll)

itu m_ , 0onu nu CU t + [i—Lm - -— rp J 1 — - C.H------n l [_^H m + m— -]v J v /■'j Cu C>2 2k (itIq +tti(-,u) O 2 Cu

The volume elements in the new velocity spaces are given by:

d V S, dVcM dV°M2 = VV V “ V V % (cl2> x y M M M M

dvo x dvo y dvo, 2 = vodvo sln eodeod*o (cl3)

Substitution of (Cll), (C12) and (C13) in (C7) gives

^Cu = (m02mCu) 3/2 (27Tk)~3 (TlT2)~3/4 (4tt)2 1 (C14) with

00 00 * ry A OI f r exp -(Ax +Cxy+Cy ) x y dxdy (C15) I = J1 J 0 0 and m o 111 „ _ 1 r u2 , Cu, A “ 7 k T T ^ (C16) T 0 2 C u 255

2 mru 1 1 B " ^ - S'

c = 2 Cu ( ^ u + _^2 J 2k(mo2+mcu' to2 TCU

Separation of variables gives

2 3 f°°2 2 1 = 1 exp (-Cy )y dy J x exp(-Ax -Bxy)dx (C19) J0 0

Moreover

2 „ - 2 /- 00 0 -yx-2Vx rr— , v / x2e dx = - - ^ 2 + V ^ 5 y [1-(t) = — f e dt = — e~ 21 -—5----- (C21) ^ JQ SS k=°- (2 k+l) ! !

OCl 2 /■» 1 ■" / x 2 ne'PX dx = 12" - U “ .-/j V o (C2 2 ) 0 2(2P)n and

2n+l -Px2 . _ ^ - r r (C23) J" xM ‘ e x ^ dx = 2pn+^

0 in our case

o ITl — g2 ^2mCu P = C - 4A = 2k(m0 2 T0 2 +mcuTCu)> 0 (C24)

Plugging (C20), (C21), (C22), (C23) and (C24) , (C19) becomes: 256

2 k + l .k-1 3B B______B . 1 I = - 2 2 T C ' 4 1 5 k+ 2 32A C A k=0 2/A /C 2(C) 2 (C-— ) ^ 4A

(2k+3) [B- ( 2 |;5I - 2A] (C25)

In the case of T C>2 = T£U ' B = 0 and (C25) reduces to (C3)

The series in C(25) can be reduced to an analytical form using

2 k _ (C26) X (1 -x) 3 0

UJ , k x (C27) Ek x = (1 -x)

(C28) x 1 -x 0

Substituting (C26), (C27) and (C28) with x fc (c25) becomes:

3B I = - 2 2 ' C 32A C ^ 4 V TV5 1 T.2 T n 2 9 ' 2 (C- ^— ) 2 (C- 5_) 4A 4A (C29)

a 2b______B i /i B 4 , B 2 , [3 3 3 2 2 AC ^ /AC 32(AC-B2)C AC AC

Taking TCu = 1223°K and TQ = 298°K from Cl.

— 4 -1 Vn = 4.44x10 cm sec (C30) 2 - 4 - 1 from C2 VCu = 6.41x10 cm sec (C31)

— 4 -1 from C3 = 7.78x10 cm sec (C32) from C14 and C29 = 3.25xl0+ “* cm sec ^ (C33) Cu APPENDIX D

The number of O 2 molecules lost by collisions with Cu in

dx is:

dn°2 = _n°2 ° "cudx (D1)

Integrated between x and h (Dl) becomes

f ° 2 d " ° 2 ( h J — - = - J cr n dx (D2 ) n 0 o x 0 2 2 nCu is not a simple function of x but (D2) is solved using the average density of copper atoms, n f between x and h:

n ° 2 = n 0 SXp ^Cu^h-X^

For all cases nV. < n_ (x=0) (D4) Cu — Cu n(2U (x=0 ) being the density of copper atoms at the specimen surface. Consequently:

exp [-onC u (0) (h-x) ] < (D5)

The minimum value for n_ 2 15:

no, . = no„ exp [-°nc u (0)(h_x)) 2 m m 2

Even at specimen temperatures as high as 950°C, the ratio

258 259

n ° 2 m in 5--- is extremely close to 1. Consequently, the

n ° 2 density of oxygen can be considered as a constant equal

i APPENDIX E

Calculation of

JL. r2l\ 2 j(b,~' _ / i rdrdtf) HL (26) ^ ^ , 2 ,,2 , 2 01_ O 0 (x +b +r -2 br cost}))

2 TT Using dx______2______. a > (El) J „ . ,, .n+ 1 „ ^ n+ 1 n /—*-— ^ O (a+bcosx) 2 , 2 .-5— / 2 , 2 (a -b ) A a -b

Pn being the Legendre polynomial with

Pn / \ = x (E2) 1 (x)

III 1 becomes

2 / m , . , 2,. 2, 2. x f 2 rdr(x +b +r ) (E3) J(b'X> = JHL — /7~2-J— 77T-ri~2^ /[(x +b +r ) -4b r ]

2 Changing variables with U=- r and using

f dy _ 2 (2 y + B) (E4) J (a+By+YY2 ) 3 7 2 A/a+By+Yy: and

ydy______2 g+gy (E5) /(a+ey+yy2)3/2 A/£+6v+A/a+By+yy 2‘

2 with A = 4ay - B (E6)

III 2 transforms to

260 261

JTTT i 2 , 2.2, HL r m - (x +b ) J (b, x) n I + 1 ] (27)

A A+2l_2 (x2 -b2) + (x2 +b2) 2 m m LIST OF REFERENCES

1. G. Tamman, Z. Anorg. Allgem. Chem., Ill, 78 (1920).

2. U.R. Evans, The Corrosion and Oxidation of Metals, Arnold Publ., London (1960).

3. 0. Kubaschwski and B.E. Evans, Oxidation of Metals and Alloys, Butterworths Publ., London (1962).

4. J. Benard, l 1Oxidation des Metaux, vol. 1,2, Gauthier-Villiars Publ., Paris (1962, 1964).

5. K. Hauffe, Oxidation of Metals, Plenum Press Publ., New York (1965) .

6 . P. Kofstad, High Temperature Oxidation of Metals, John Wiley, New York (1966).

7. S. Mrowec and T. Werber, Gas Corrosion of Metals, Published for the National Bureau of Standards and the National Foundation of Science, by the Foreign Scientific Publications, Warsaw, Poland (1978).

8 . N. Cabrera and N.F. Mott, Rep. Prog. Phys., 12, 163 (1948,49).

9. K. Hauffe, The Surface Chemistry of Metals and Semi- Conductors, H.G. Gator, Ed., John Wiley, New York, 439 (1960).

10. M.W. Roberts, Quart. Rev., 16, 71 (1962).

11. K.R. Lawless, Ch. in Energetics in Met. Phenomena, Vol. 1, M.W. Mueller, Ed., Gordon and Breach, New York, 34 5 (1965).

12. F.P. Fehlner and N.F. Mott, Oxid. Met., 2, 59 (1970) .

13. R.C. Logan and W.W. Smeltzer, Can. Met. Quart., 10, 149 (1971).

14. K.R. Lawless, Rep. Prog. Phys., 37, 231 (1974).

15. G.C. Wood, Techniques in Metals Research, Vol. 4, R.A. Rapp, Ed., Interscience, New York (1970).

262 263

16. W.W. Smeltzer and D.J. Young, Prog, in Solid State Chem., 10, 17 (1975).

17. P. Kofstad,'Oxid. Mech. for Pure Metals in Single Oxidant Gases*in Proc. of the Int. Conf. on High Temperature Corrosion, R.A. Rapp, Ed., San Diego, CA, NACE (1981).

18. J.H. Darling, M.B. Garton-Sprenger and J.S. Ogden, Faraday Symp. of the Chemical Soc., n°8 , 75 (1973).

19. A.B.P. Lever, G.A. Ozin and H.B. Gray, Inorg. Chem., 19, 1823 (1980).

20. D. Michell and A.P. Smith, Phys. Stat. Sol., 27, 291 (1968).

21. J.S. Ogden, A.J. Hincheliffe and J.S. Anderson, Nature, 226, 940 (1970).

22. J. Bardolle and J. Benard, C.R. Acad. Sci. , 232, 231 (1951) .

23. E.D. Hondros, Coll. Int du CNRS no 122, Processus de Nucleation dans les Reactions Gaz sur Metaux, Paris, 57 (1965) .

24. J. Oudar, "First Stages of Gas-Metals Interactions," Proc. of the Int. Conf. on High Temperature Corrosion, R.A. Rapp, Ed., San Diego, CA, NACE (1981).

25. J.L. Domange and J. Oudar, Surf Sci., 11, 124 (1968).

26. J.L. Domange, J. Oudar and J. Benard, Molecular Process on Solid Surfaces, E. Drauglis, R.D. Gretz and R.i. jatfee, Eds., McGraw Hill Book Co., New York, 353 (1959).

27. D.O. Hayward and G.M.W. Trapnell, Chemisorption, Butterworths, London (1964) .

28. H.E. Furnsworth, The Surface Chemistry of Metals and Semi-Conductors, John Wiley, New York, 21 (1960).

29. J.M. Moison and J.L. Domange, Surf. S c i ., 97, 1 (1980).

30. J.J. Lander, Surf. S ci., 1, 125 (1964).

31. N. Cabrera, J. Chem. Phys., 53, 675 (1956). 264

32. L.B. Garmon and K.R. Lawless, Structures et Proprietes des Surfaces des Solides, Coll. Int. Cent. Nat. Res. Sci. No 187, 61 (1969) .

33. R.B. Marcus and L.O. Brockway, Processus de Nuclea- tion dans les Reactions des Gas sur les Metaux et Problemes Connexes, Coll. Int. Cent. Nat. Rech. Sci. Paris no 122, 61 (1965).

34. A.P. Rowe and L.O. Brockway, Proc. 22th Ann. Meeting Electron Microscope Society of America, C.J. Arceneaux Ed., Baton Rouge Claitors, 118 (1969).

35. V.A. Phillips, J . Appl. Phys., 3 3, 12 (1962).

36. B.K. Chakraverty and G.M. Pound, Acta Met., 12, 851 (1964).

37. T.G. Garbovitskaya and L.N. Paritskaya, Phys. Met. Metallu., 47 (3), 108 (1980).

38. C .A. Neugebauer, Handbook of Thin Film Technology. Chapter 8 , L.I. Maissel and R. Glang, McGraw Hill, New York (198?).

39. W.H. Orr, Thesis, Cornell University, Ann Arbor, Michigan (1962) .

40. W.W. Harris, F.W. Ball and A.T. Gwathmey, Acta Met., 5, 524 (1957).

41. A. Ronnguist, J\ Inst. Metals, 91, 89 (1962).

42. J. Benard, F. Gronlund, J. Ovudar and M. Duret, Z Elektrochem. 63, 799 (1959) .

43. G.E. Rhead, Trans. Faraday Soc., 61, 797 (1965).

44. J.M. Bakely, Prog. Mat. Sci., 10, 395 (1963).

45. F.J. Bradshaw, R.H. Brandon and C. Wheeler, Acta Met., 12, 1057 (1964).

46. W.F. Boggs, J . Electrochem. Soc., 108, 124 (1961).

47. A.F. Beck, M.A. Heine, E.J. Caule and M.J. Pryor, Corr. Sci, 1_, 1 (1967) .

48. G.A. Pashley, M.J. Stowell, M.H . Jacobs and T.J. Law, Phil M a g ., 10, 127 (1964). 265

49. G.A. Bassett, Proc. European Conf. Electron Micro­ scopy, Delft, p. 270 (1960).

50. D.W. Pashley, Thin film, American Soc. for Metals, Metals Park, Ohio, 59 (1963) .

51. V.A. Phillips, Phil. Mag., 5 (8 ), 571 (1960).

52. W.K. Burton, N. Cabrera and F.C. Frank, Phil. Trans. Roy. Soc., 243, 299 (1951).

53. E. Bauser and H. Strunk, J. Crystal Growth, 51, 362 (1981) .

54. N.F. Mott,Proc. Roy. Soc., A , 171, 344 (1939).

55. N.F. Mott, Trans. Faraday Soc., 36, 472 (1940).

56. N.F. Mott, Trans. Faraday Soc., 43, 429 (1947).

57. H.H. Uhlig, Acta Met., 5, 541 (1956).

58. D.D. Eley and P.R. Wilkinson, Proc. Roy. Soc. A, 254,327 (1960).

59. P.T. Landsberg, J . Chem. Phys., 23, 1079 (1955).

60. M.J. Graham, "Thin Oxide Film Formation on Metals," Proc. of the Int. Conf. on High Temperature Cor­ rosion, R.A. Rapp, Ed., San Diego, CA, NACE (1981).

61. C. Wagner, Corr. Sc., 13, 23 (1973). j 62. C. Wagner and K. Grunewald, Z. Physik, Chem. B, 21, 25 (1933).

63. C. Wagner, Atom Movement, Am. Soc. Met., Cleveland 153 (1951).

64. C.R.A. Catlow and B.E.F. Fender, J. Phys. Chem. Solid State Phys., 8 , 3267 (1975).

65. C.R.A. Catlow, W.C. Mackrodt, M.J. Norgett and A.M. Stoneham, Phil. Mag. A , 40, 161 (1979) .

66. J.E. Stroud, W.C. Tripp and J.M. Wimmer, J. Am. C e r . S o c ., 57, 172 (1974). 67. C. Wagner, J . Phys. Chem., 67, 738 (1953). 266

6 8 . C. Wagner, J . Electrochem. Soc., 99, 346 (1952).

69. G.H. Meier and R.A. Rapp., Z. Phys. Chem. 74, 168 (1971) .

70. G. Martin and B. Perraillon, J. Phys., Paris, 36, 14 (1975) .

71. E.W. Hart, Acta Met., 5, 597 (1957).

72. W.W. Smeltzer, R. R. Haering and J.S. Kirkaldy, Acta Met., 9, 880 (1961).

73. J.M. Perrow, W.W. Smeltzer and J.D. Embury, Acta Met., 16, 1209 (1968) .

74. R. Herchl, N.N. Khoi, T. Homma and W.W. Smeltzer, Oxid. Met., 4, 35 (1972).

75. N.N. Khoi, W.W. Smeltzer and J.J. Embury, j. Electrochem Soc., 122, 1495 (1975). ~

76. A.R. Lang, Ch. in Crystal Growth: an introduction, P. Hartman, Ed., North Holland, 444 (1973).

77. A.R. Verma, Crystal Growth and Dislocations. Butterworths, London (1953) .

78. G.H. Gilmer, R. Ghez and N. Cabrera, J. of Crystal Growth, 8 , 79 (1971) .

79. D.A. Voss, E.P. Butler and T.E. Mitchell, M e t . Trans. A ., 13, 929 (1982).

80. F.C. Frank, Acta Cryst., 4, 497 (1951).

81. W. Schottky, Z. Elektrochem., 63, 784 (1959).

82. R. Hales and A.C. Hill, Corr. Sci., 12, 843 (1972).

83. G.B. Gibbs and R. Hales, C orr. S c i ., 17, 487 (1977) .

84. P. Hancock and R.C. Hurst, Adv. in Corr. Sci and Tech., vol 4, M.G. Fontana and R.W. Staehle, Eds., Plenum, 1 (1974) .

85. W.K. Appleby and R.F. Tylecotte, Corr. Sci., 10, 325 (1970) . 267

8 6 . D. Caplan, R.J. Hussey, G.I. Sproule and M.G. Graham, Oxid. Met., 14, 279 (1980).

87. K.P. Lillerud and P. Kofstad, J. Electrochem. Soc., 127, 2397 (1980).

8 8 . P. Kofstad and K.P. Lillerud, J. Electrochem. Soc., 127, 2410 (1980) .

89. S.K. Bose and R.A. Rapp, to be published.

90. M.J. Graham and D. Caplan, J. Electrochem. Soc., 120, 769 (1973).

91. E .A. Gulbrarsen and K.F. Andrew, J. Electrochem. Soc., 105, 451 (1957).

92. K. Fueki and j.B. Wagner, Jr., J. Electrochem. Soc., 112, 384 (1965) .

93. A. Bruckman, Corr. Sci., 7, 51 (1967).

94. S. Mrowec, Corr. Sci., 7, 563 (1967).

95. G.C. Wood, J.M. Ferguson, B. Vaszko and D.P. Whittle, J. Electrochem. Soc., 114, 535 (1967).

96. D. Caplan, M.J. Graham and M. Cohen, J. Electrochem. Soc., 119, 1205 (1972) .

97. M.J. Graham, D. Caplan and M. Cohen, J. Electrochem. Soc., 119, 1265 (1972).

98. A. Atkinson, R.I. Taylor and P.D. Goode, Oxid. Met., 13, 519 (1979).

99. J. Stringer, Corr. Sci., 17, 487 (1977).

100. N.B. Pilling and R.E. Bedworth, J. Inst. Metals, 29, 529 (1923).

101. W. Jaenicke, S. Leislikow and A. Stadler, J . Electrochem. Soc., Ill, 1031 (1964).

105. K. P. Lillerud and P. Kofstad, Proc. of the Int. Conf. on High Temperature Corrosion, R.A. Rapp, Ed., San Diego, CA, NACE (1981).

106. F. N. Rhines and J.S. Wolf, Met. Trans., 1, 1701 (1970). 268

107. F.N. Rhines and R.G. Connell, Jr., Proc of symp. on Stress Effects and the Oxidation of Metals, Detroit, 94 (1977) .

108. F.N. Rhines and R.G. Connell, Jr., J. Electrochem. Soc., 124, 1123 (1977) .

109. M.V. Speight and J.E. Harris, Acta Met., 26, 1043 (1978).

110. A. Atkinson, Corr. Sci., 22, 347 (1982).

111. . P. Kofstad, J. Less Common Letals, 7, 241 (1964) .

112. P. Kofstad and S. Espevik, J. Electrochem. Soc., 112, 153 (1965).

113. J .V . Catheart, J.J. Campbell and G.P. Smith, J. Electrochem. Soc., 105, 442 (1958).

114. A. Ronnquist and H. Fischmeister, J. Inst. Metals., 89, 65 (1960-61).

115. D.W. Bridges, J.P. Baur, G.S. Baur and W.M. Fassell, J. Electrochem. Soc., 103, 475 (1956).

116. J.P. Baur, D.N. Bridges and W.M. Fassell, J. Electrochem. Soc., 103, 273 (1956).

117. J.S. Halliday and W. Hirst, Proc. Phys. Soc., B, 6 8 , 178 (1955).

118. H.K. Hardy, J . Inst. Metals, 79, 497 (1957).

119. G. Pfefferkorn, Proc. of the 3rd Int. Conf. on Electron Microscopy, London, 491 (1954).

120. W. Jaenicke and L. Albert, Naturwiss, 46, 491 (1959).

121. E .A. Gulbransen, T.P. Copar and K.F. Andrew, J. Electrochem. Soc., 108, 119 (1961).

122. A. Oustry, L. Lafourcade, A. Escaut, Surf. Sci., 40, 545 (1973).

123. K. Heinemann, D.B. Rao and D.L. Douglass, Oxid. Met., 9, 379 (1975).

124. D.A. Goulden, Phil. Mag., 33, 393 (1976). 269

125. F. Gr^nlurd and P.E. H0jlund Nielsen, Surf. Sci., 30, 388 (1972).

126. J.H. Ho and R.W. Vook, Phil. Mag., 36, 1051 (1977).

127. J.H. Ho and R.W. Vook, J. Crystal Growth, 44, 561 (1978).

128. R.W. Vook and J.H. Ho, Thin Solid Films, 58, 153 (1979).

129. G.G. Hembree, J.M. Cowley and M.A. Otooni, Oxid. Met., 13, (1979).

130. P. Kofstad, Non Stoichiometry Diffusion and Electri­ cal Conductivity of Binary Metals 0~xT3ei"^ W T T e y Interscience, New York C1972T7

131. Y.D. Tretyakov, V.F. Komarov, N .A. Prosvirnira and I .B. Kotserok, J. Solid State Chem., 5, 157 (1972).

132. S. Mrowec and A. Stoklosa, Oxid. Met., 3, 291 (1971).

133. H. Dunnwald and C. Wagner, Z. Phys. Chem. B, 22 212 (1933) .

134. M.O'Keefe and W.J. Moore, J. Chem. Phys., 36, 3009 (1962).

135. M. Yoshimura, A. Revcolevschi and J. Castaing, J . Mater. Sci., 11, 384 (1976).

136. R.S. Toth, R. Kilkson and D. Trivich, Phys. Rev., 122, 482 (1961) .

137. M. Tapiero, J.P. Zielinger and C. Nouguet, Ann. Phys., 7, 85 (1972).

138. J. Maluenda, R. Fahri and G. Petot Ervas, Ann. Chim. F r ., 4, 488 (1979) .

139. W.J. Moore and B. Selikson, J. Chem. Phys., 19, 1539 (1957).

140. C. Wagner, Progress in Solid State Chemistry, J. McCaldin and G. Somorjai, Eds., Pergaman, New York, 10, 3 (1970) .

141. F .A . Kroger, Chemistry of Imperfect Crystals, North Holland, Amsterdam, ch. 15 T19637 . * 270

142. J. Maluenda, R. Farhi and G. Petoterv, J. Phys. Chem. of Solids, 42, 911 (1981).

143. P.H. Holloway and J.B. Hudson, Surf. Sci., 43, 123 (1974).

144. R. Hales, A.C. Hill and P.K. Wild, Corr. Sci., 13, 325 (1973).

145. C. Benndorf, B. Egert, G. Keller, H. Seidel and F. Thieme, Surf. Sci., 80, 287 (1979).

146. H.M. Hindam and W.W. Smeltzer, Oxid. Met., 14, 337 (1980) .

147. H.J. Grabke and H. Viefhaus, Surf. Sci., 112, L719 (1981).

148. J.W. May, Advan. Catalysis, 21, 151 (1970).

149. D.F. Mitchell, P.B. Sewell and M. Cohen, Surf. Sci., 61, 355 (1976) .

150. D.F. Mitchell, P.B. Sewell and M. Cohen, Surf. Sci., 69, 310.(1977).

151. P.R. Norton, R.L. Trapping and J.W. Goodale, Surf. Sci., 65, 13 (1977) .

152. K.S. Kim and N. Winograd, Surf. Sci., 43, 625 (1974).

153. G.R. Brundle and A .F . Carley, Chem. Phys. Lett., 31, 423 (1975).

154. S. Evans, J. Pielaszck and J.M. Thomas, Surf. Sci., 56, 644 (1976) .

155. G.C. Allen, P.M. Tucker and R.K. Wild, Oxid. Met., 13, 223 (1979).

156. W. Heiland and E. Taglaves, Surf. Sci., 8 6 , 90 (1976).

157. H.H. Brongersma and J.B. Theeten, Surf. Sci., 54, 519 (1976).

158. A. Muller and A. Benninghoven, Surf. Sci., 41, 493 (1974) .

159. T. Fleisch, N. Winograd and N.N. Delglass, Surf. Sci., 78, 141 (1978) . 2 71

160. P.M. Marcus, J.E. Demuth and D.W. Jepsen, Surf. Sci., 53, 500 (1975).

161. J .V. Cathcart, G.F. Petersen and C.J. Sparks, Jr., J. Electrochem. Soc., 116, 664 (196.9).

162. S.H. Kulpa and R.P. Frunke.nthal, Corr. Sci., 124, 1588 (1977) .

163. K. Hauffe, L. Pethe, R. Schmidt and S.R. Morrison, J. Electrochem. Soc., 115, 456 (1968).

164. M.J. Graham and M. Cohen, J. Electrochem. Soc., 119, 879 (1972).

165. J.M. Perrow, W.W. Smeltzer and J.D. Embury, Acta Met., 16, 1209 (1968).

166. J.H. Sartell and C.H. Li, J. Inst. Metals, 90, 92 (1961-62) .

167. W.L. Phillips, Jr., J. Electrochem. Soc., 110, 1014 (1964).

■168. Y.L. Yao, J . Chem. P h y ., 33, 741 (1960).

169. K. Fueki-and H. Ishibashi, J. Electochem. Soc., 108, 306 (1961) .

170. G.C. Wood, I.G. Wright and J.M. Ferguson, Corr. Sci., 5, 645 (1965) .

171. J.J. Van^en Brock and J.J. Meijering, Acta Met., 16, 375 (1968) .

172. T. Ueno, Jap. J. A p p l . Phys., 13, 773 (1974).

173. A. Atkinson and R.I. Taylor, J. Mat. Science, 13, 427 (1978) .

174. B. Lesage, A.M. Huntz and P. Lacombe, J. Phys. Chem. Solids, 42, 705 (1981) .

175. M.J. Graham, R.J. Hussey and M. Cohen, J. Electrochem. Soc., 120, 1523 (1972).

176. M.J. Graham, G.I. Sproule, D. Caplan and M. Cohen, J. Electrochem. Soc., 119, 883 (1972). 2 72

177. A. Atkinson and R.I. Taylor, Phil. Mag. A., 39, 581 (1979) . '

178. A. Atkinson and R.I. Taylor, Phil. Mag. A., 43, 979 (1981).

179. R. Hales, Corr. Sci, 12, 555 (1972) .

180. F. Armanet, G. Beranger and D. David, Metallic Corr. Proc. 8 th Int. Cong, on Metallic Corr. Dechema, Ed., Frankfurt, Germany, 735 (1981).

181. J.E. Castle, Scanning Electron Microscopy, 471 (1969).

182. A.M. Brown and P.L. Surman, Surf Sci., 52, 85 (1975) .

183. J.E. Castle and M.R. Hunt, Corr. Sci., 16, 137 (1976).

184 . E. Metcalfe and A. Charalambous, Central Electricity Research Laboratories, VH102 (1981) .

185. R.M. Fulrath, Scanning Electron Microscopy, 18 (1972).

186. D.N.K. Wang, D.J. Miller and R.M. Fulrath, Scanning Electron Microscopy, I, 177 (1978).

187. M. Viestaro, J. Harkki and M.H. Kikkanen, Scandinavian J. Me t ., 8 , 5 (1978) .

188. E.K. Brandis, R. Hoover and G. Das, 3rd Annual Stereoscan SEM Coll., Kent Cambridge Scientific pubi r (1970) r ~ ..

189. CRC Handbook of Chem. and Phy. 4 9th ed. Chemical Rubber Co. Publ. Cleveland, Ohio (1968-69).

190. M. Hansen, Constitution of Binary Alloys, McGraw- Hill, New York (1958) .

191. S. Kimoto and H. Hashimoto, Proc. Symp. The Scanning Electron Microscope; The Instrument and Its Appli­ cations, Chicago (1968).

192. D. Halliday and R. Resnic, Fundamentals of Physics, John Wiley Publ., 510 (1970). 273

193. G.R. Booker, Proc. of Int. Summer Course on , "Modern Diffraction and Imaging Techniques in Materials Science," Antwerp, Belgium (1969). Publ. by North Holland Publishing Company (Amsterdam) (1970) .

194. 0. Wells, Scanninq Electron Microscopy, McGraw-Hill, New York, l50 (1974) .

195. R.A. Rapp and D.A. Shores, "Solid Electrolyte Galvanic Cells," Ch. in Techniques of Metals Research, vol. IV, part 2. R.A. Rapp, E d ., R.F. Bunshah series ed. Wiley Publ. Co., New York (1970).

196. G. Honjo, J . Phys. Soc. Japan, 4, 330 (1949).

197. T. Homma, N.N. Khoi, W.W. Smeltzer and J.D. Embury, Oxid. Met., 3 (5), 463 (1971). .

198. J.H. Sartell, R.J. Stokes, S.H. Bendel, T.L. Johnson and C.H. Li, Trans. Met. Soc. AIME, 215, 420 (1959).

199. D.F. Mitchell and K.C. Lawless, Proc. Paint. Res. Inst., 38, 575 (1966) .

200. R.T. Tung and W.R. Graham, Surf. Sci. , 97, 73 (1980).

201. J. Bardolle, Pub. Sci. Tech. Ministere de l'air (France) , 327 (1957]“; '

202. J.M. Molson and J.H. Domange, Surf. Sci, 67, 336 (1971).

203. E. Votava and A. Berghezan, Acta Met., 7, 392 (1959).

204. R.M. Lowe, Acta Met., 12, 1111 (1964).

205. W.J. Tomlinson and J. Yates, J. Cryst. Growth, 38, 265 (1977).

206. W.J. Tomlinson and J. Yates, J. Electrochem. Soc., 125, 803 (1978).

207. S. Motojima and K. Sugiyama, J. Cryst. Growth, 55, 611 (1981).

208. E.T. Turkdogan, P. Grieveson and L.S. Darken, J . Phy. Chem., 67, 1647 (1963). 274

209. J. Yacaman, S. Fuentes and J.H. Dominguez, Surf. Sci., 106, 472 (1981).

210. J.H. Hren, J*I. Goldstein and D.J. Joy, Introduction to Analytical Microscopy, Plenum Press, New York, 233' (1979)'.

211. J.Hecht, W.P. West and M.A. Norton, Surf. Sci., 106, 131 (1980).

212. D.E. Rosner, Oxid Metals, 4_, 1 (1972).

213. C. Kaito, K. Fujita, H. Shibahara and M. Shiojiri, Japan J. Appl. Phys., 17, 601 (1978).

214. K. Hilpert and K .A. Gingerich, Ber. Bunsenges. phys. Chem., 84, 739 (1980).

215. S. Smoes, F. Mandy, A. Von der Auwera-Mahieu and J. Drowart, Bull. Soc. Chim. Beiges., 81, 45 (1972).

216. JANAF, Thermochemical tables, Dow Chemical Corp., Looseleaf ed. (1980).

217. H. Huber and G .A . Ozin, Can. J. Chem., 50, 3746 (1976);

218. K.P. Huber and G. Herzberg, Molecular Spectra and Molecular Structure, Vol. 4, constants of diatomic molecules, Van Nostrand, New York, 208, 490 (1978) .

219. V. Ruth and J.P. Hirth, in Condensation and Evapora­ tion of Solids, E. Rutner, Pi Goldfinger and J.P. Hirth, Eds., 99 (1962).

220. D. Langbein, Theory of Van der Waals Attraction, Springer, Berlin ed. (1974).

221. S.F. Friedlander, Smoke, Dust and Haze, Wiley Publ., (1977).

222. P.E. Wagner, J. Colloid. Interface Sci., 56, 398 (1976).

223. W.H. Marlow, Surf. Sci., 106, 529 (1981).

224 . R. Defay, E. Prigognine, A. Bellemans and D.H. Everett, Surface Tension and Adsorption, Longsman- Green, London, 1966. 27 5

225. E.D. Hondros and M. McLean, J. Mat. Sci., 8 , 349 (1973) .

226. C.E. Bauer, M.S. Thesis, The Ohio State University, 1974 .

227. L.E. Murr, Interfacial Phenomena in Metals and Alloys, pp. 124-126Addison Wesley, London, T975.

228. E.D. Hondros, Techniques of Metal Research, vol. 4, Physiochemical Measurement in Metals Research, Part 2, pp. 293-348, R.A. Rapp, Ed., Wiley Publ., New York, 1970.

229. J. Friedel, Surf. Sci., 106, 582 (1981).

230. T. Halicioglu and P.J. White, Surf. Sci., 106, 45 (1981) .

231. G.W. Mays, F.S. Vermaak and D. Kuhlmann-Wilsdorf, Surf. Sci., 12, 124 (1968).

232. V.R. Howes, Corr. Sci., 10, 99 (1970).

233. C.S. Giggins and F.S. Pettit, Trans. Met. Soc. AIME, 245, '2509 (1969) .

234. C. Wagner, J . Metals, £, 214 (1952).

235. Y.K. Rao, Met. Trans. B., 10, 243 (1979).

236. A. Atkinson and R.I. Taylor, Phil. Mag. A, 43, 979 (1981).

237. R.H. Bricknell and D.A. Woodford, Acta Met, 30, 257 (1982).

238. D. Caplan, R.J. Hussey, G.I. Sproule and M.J. Graham, Scripta Met., 16, 759 (1982) .

239. R.H. Bricknell and D.A. Woodford, Scripta Met, 16, 761 (1982).

240. P.D. Brutowe, A. Brokman, F. Spaepen and R.W. Balluff, Scripta Met, 14, 943 (1980) .

241. K. Urban, Phys. Stat. Sol, (a), 3, k 167 (1970).

242. K. Urban, Phys. Stat. Sol, (a) 4, 761 (1971). 276

243. J. Van Landuyt, R. Gevers and S. Amelinckx, Phys. Stat. Sol., 10, 319 (1965).

244 . K. Schwerdtfeger and E.T. Turkdogan, Techniques of Metals Research, Vol. IV, Part I, R.a T Rapp, E d ., Wiley Publ. Co., New York, 324 (1970).

245. F.H. Stott, Zhou Peide, W.A. Gronk and R.P.M. Procker, Corros ♦ Sci., 22, 305 (1982).

246. R.S. Wagner and W.C. Ellis, Appl. Phys. Let., 4, 89 (1964).

247. R.S. Wagner and W.C. Ellis, Trans. TAIME, 233, 1053 (1965) .

248. F.C. Frank, Disc. Faraday Soc., 5, 48 (194 9).

249. E .A . Gulbransen, Mem. Sci. Rev. Met., 62, 253 (1965).

250. R.L. Tallman and E .A . Gulbransen, J. Electrochem. Soc., 114, 1227 (1967) .

251. M. McLean and J.P. Hirth, Surf. Sci., 11, 25 (1968).

252 . T. Jungling. M.S. Thesis, The Ohio State Univer­ sity (198 3) .

253. A.R. Verma , Phil. Mag., 42, 1005 (1951).

254 . R. Kaischew, E. Budevski and J. Malinowski, Z. Phys. Chem, 204, 348 (1955).

255. C. Chapon and A. Bonnisent, J. Cryst. Growth, 18, 103 (1973).

256 . N. Cabrera, Structure and Properties of Surfaces. Chicago Univ. Press, 295 (1953).

257. E. Budevsk, G. Staikov and V. Bostanov, J. Cryst. Growth, 29, 316 (1975).

258 . I. Suraqawa and P. Bennema, J. Cryst. Growth, 46, 451 (1979).

259. P. Bennema and G.H. Gilmer, in Crystal Growth, An Introduction, P. Hartman, Ed. North Holland, Amsterdam, 301 (1973) . 27 7

260. A. Kuhn, A. Chevy and E. Lendray, J. Cryst. Growth, 13/14, 380 (1972).

261. W.J.P. -Van Enckevork, R. Janssen-Van Rosmalen and W.H. Van der Linden, J. Cryst. Growth, 49, 502 (1980).

262. I. Sunagawa, J . Am. Miner., 46, 1216 (1961).

263. E. Alexander, Z. H. Kalman, S. Mardix and I.T. Steinberger, Phil. Mag., 21, 1237 (1970).

264 . S. Mardix, A.R. Lang, and I. Blech, Phil. Mag., 22, 683 (1971).

265. W. Gotoh and H. Komatsu, J. Cryst. Growth, 54, 163 (1981).

266. D. Turnbull, in "Atom Movements," A.S.M. Seminar, 129 (1950).

267. A .D . LeClaire, Prog. Met. Phys., 1, 306 (1949).

268 . A .D. LeClaire, Prog. Met. Phys., 4, 1263 (1953).

269. G. Dhallenne, A. Revcolevschiand G. Monty, Phys. Stat. Sol. (A), 5j>, 623 (1979) .

270. F. Armanet, A. Vejux, G. Johannesson and G. Beranger, Oxid. Met., 15, 3 (1981).

271. C.W. Tuck, M. Odgers and K. Sachs, Corr. Sci., 9, 271 (196p) .

272. G.K. Boreskov, Disc. Faraday Soc., 41, 263 (1966).

273. C. Wagner, Corr. Sci., 10, 641 (1970).

274. J.H. Grabke, K.J. Best and A. Gala, Werk. Korr., 21, 911 (1970).

275. E.T. Turkdogan, W.M. McKewan and L. Zwell, J. Phys. Chem., 69, 327 (1965).

276. R.B. Byrd, W.E. Stewart and E.N. Lightfoot, Trans- port Phenomena, John Wiley Publ., New York, 434 (1960) .

277. R.D. Present, Kinetic Theory of Gases. McGraw Hill, New York, 265 (1958).