THE LAVES PHASE EMBRITTLEMENT OF FERRITIC STAINLESS STEEL TYPE AISI 441 M

THE LAVES PHASE EMBRITTLEMENT OF FERRITIC STAINLESS STEEL
TYPE AISI 441
M AITSE P. SELLO
THE LAVES PHASE EMBRITTLEMENT OF FERRITIC STAINLESS STEEL
TYPE AISI 441
BY
M AITSE P. SELLO
SUPERVISED BY
PROFESSOR W ALDO E. STUMPF
Submitted in fulfilment of the requirements for the degree
Philosophiae Doctor, PhD (Metallurgy)
in the
Department of Materials Science and Metallurgical Engineering
Faculty of Engineering, Building Environment and Information Technology
University of Pretoria
REPUBLIC OF SOUTH AFRICA
2009, June 15
© University of Pretoria
PREFACE
This dissertation is submitted for the degree of Doctor of Philosophy at the University of
Pretoria.
The research described herein was conducted under the supervision of
Professor W. Stumpf in the department of Materials Science and Metallurgical
Engineering, University of Pretoria.
Except where acknowledgements and references are made to previous work, this work
is, to the best of my knowledge, original. This dissertation is the result of my own work
and includes nothing of which is the outcome of work done in collaboration with others
except where specifically indicated in the text. Neither this, nor any substantially similar
dissertation has been, or is being submitted to my knowledge for any other degree,
diploma, or other qualification at any other university.
_______________________
Maitse P. Sello
P a g e | iii
ACKNOWLEDGEMENT
I would like to thank God for his guidance in my life, and also for being my strength and
guidance in my life.
I would like to express my sincere thanks to my academic supervisor, Prof. Waldo E.
Stumpf for his invaluable guidance, encouragement and input in this research project. I
have benefited a lot from his experience and insight during this research project.
I would like to thank Mrs Sarah Havenga, Mrs Lillian Barlow and Mrs Elsie Snyman –
Ferreira for their help with the administrative work.
I would like to thank Prof L. Cornish (currently at the University of Witswatersrand) and
Mrs L. H. Chown for their training and guidance during thermodynamic modelling using
Thermo-Calc® software at Mintek
I would like to thank Mrs A. Tuling and Mr C. van der Merwe for their help with TEM
work, Dr N. van der Berg for his expertise in the electron diffraction indexing and Mr C.
Coetzee for his help with the SEM.
I would also like to thank Dr S. Verryne and Prof. J de Villiers for their help in XRD
analysis and also their expertise, they have always been available for help whenever
needed, and for that I am very grateful.
I would like to thank everyone at the University for company and support during my time
there. There are many interesting, encouraging and happy moments to remember with
my friends during the course of my studies.
Finally I would like to thank my parents and siblings for supporting me in every decision
that I have taken in my life, thank you for your support and patience.
P a g e | iv
THE LAVES PHASE EMBRITTLEMENT OF FERRITIC STAINLESS STEEL
TYPE AISI 441
Author:
Maitse P. Sello
Supervisor: Professor Waldo E. Stumpf
Department of Materials Science and Metallurgical Engineering
University of Pretoria
Philosophiae Doctor (Materials Science and Metallurgical Engineering )
Synopsis
The effect of Laves phase (Fe2Nb) formation on the Charpy impact toughness of the
ferritic stainless steel type AISI 441 was investigated.
The steel exhibits good
toughness after solution treatment at 850 °C, but above and below this treatment
temperature the impact toughness decreases sharply. With heat treatment below 850
°C the presence of the Laves phase on grain boundaries and dislocations plays a
significant role in embrittlement of the steel whereas above that temperature, an
increase in the grain size from grain growth plays a major role in the impact
embrittlement of this alloy. The toughness results agree with the phase equilibrium
calculations made using Thermo–Calc® whereby it was observed that a decrease in the
Laves phase volume fraction with increasing temperature corresponds to an increase in
the impact toughness of the steel. Annealing above 900 °C where no Laves phase
exists, grain growth is found which similarly has a very negative influence on the steel’s
impact properties. Where both a large grain size as well as Laves phase is present, it
appears that the grain size may be the dominant embrittlement mechanism. Both the
Laves phase and grain growth, therefore, have a significant influence on the impact
properties of the steel, while the Laves phase’s precipitation behaviour has also been
investigated with reference to the plant’s manufacturing process, particularly the cooling
rate after a solution treatment.
The microstructural analysis of the grain size shows that there is a steady increase in
grain size up to about 950 °C, but between 950 °C and 1000 °C there is a sudden and
rapid 60 % increase in the grain size.
The TEM analysis of the sample that was
annealed at 900 °C shows that the Laves phase had already completely dissolved and
cannot, therefore, be responsible for “unpinning of grain boundaries” at temperatures of
900 °C and higher where this “sudden” increase in grain size was found. The most
Page |v
plausible explanation appears to be one of Nb solute drag that loses its effectiveness
within this temperature range, but this probably requires some further study to fully prove
this effect.
During isothermal annealing within the temperature range of 600 to 850 °C, the time –
temperature – precipitation (TTP) diagram for the Laves phase as determined from the
transformation kinetic curves, shows two classical C noses on the transformation curves.
The first one occurring at the higher temperatures of about 750 to 825 °C and the
second one at much lower temperatures, estimated to possibly be in the range of about
650 to 675 °C. The transmission electron microscopy (TEM) analyses show that there
are two independent nucleation mechanisms that are occurring within these two
temperature ranges. At lower temperatures of about 600 °C, the pertaining nucleation
mechanism is on dislocations and as the temperature is increased to above 750 °C,
grain boundary nucleation becomes more dominant.
Also, the morphology of the
particles and the misorientation with the matrix changes with temperature. At lower
temperatures the particles are more needle-like in shape, but as the temperature is
increased the shape becomes more spheroidal.
The effect of the steel’s composition on the Laves phase transformation kinetics shows
that by lowering the Nb content in these type 441 stainless steels, had no significance
effect on the kinetics on precipitation of the Laves phase. However, a Mo addition and a
larger grain size of the steel, retard the formation of the Laves phase, although the
optimum values of both parameters still need further quantification.
The calculation made for the transformation kinetics of the Laves phase, using the
number density of nucleation sites No and the interfacial energy
γ
as the fitting
parameters in this work, demonstrated a reasonable agreement with experimental
results.
Keywords: Laves phase (Fe2Nb), titanium niobium carbonitrides (Ti,Nb)(C,N), impact
embrittlement, grain size, ductile-to brittle transition temperature (DBTT), Laves phase
transformation kinetics, Cottrell approach to grain size, Smith model of brittle grain
boundary phases, Thermo- Calc®.
P a g e | vi
TABLE OF CONTENT
PREFACE
III
ACKNOWLEDGEMENT
IV
TABLE OF CONTENT
VII
TABLE OF FIGURES
XIII
NOMENCLATURE
XIX
CHAPTER ONE
1
GENERAL INTRODUCTION
1
1.1
Introduction
1
1.2
Problem Statement
3
1.3
Objectives
3
CHAPTER TWO
5
LITERATURE REVIEW
5
2.1
Introduction
5
2.2
Classification of Stainless Steels
5
2.2.1
2.2.2
2.2.3
2.2.4
2.3
Ferritic Stainless Steel
Austenitic Stainless Steel
Martensitic Stainless Steel
Duplex Stainless Steel
7
7
7
8
Composition of Stainless Steels
8
2.3.1
2.4
10
Effect of Grain Size on Brittle Behaviour
Embrittlement at 475°C
Precipitation of the Secondary Phases in Stainless Steels
Notch Sensitivity
Weldability of Ferritic Stainless Steel
Effect of Niobium and Titanium Additions to Ferritic Stainless Steels
10
12
13
14
15
15
Theories of Brittle Fracture
2.5.1
2.5.2
2.5.3
2.5.4
2.6
8
Toughness of Ferritic Stainless Steels
2.4.1
2.4.2
2.4.3
2.4.4
2.4.5
2.4.6
2.5
Structure of Ferritic Stainless Steel
16
Zener’s/Stroh’s Theory
Cottrell’s Theory
Smith’s Theory
Cleavage Fracture Resistance
17
18
21
23
Thermomechanical Processing
23
P a g e | vii
2.6.1
2.6.2
2.6.3
2.6.4
Cold-Rolling
Hot-Rolling
Cooling Rate
Heat Treatment
24
24
24
25
2.7
Applications of Stainless Steels in Automobile Exhaust System
25
2.8
Heat Resistant Ferritic Stainless Steels
27
2.9
Stabilisation
27
2.9.1
2.9.2
2.9.3
2.9.4
2.9.5
Stabilisation with Titanium
Stabilisation with Niobium
Solid Solution Hardening and Solute Drag by Niobium
Effects of Temperature on Solute Drag
Dual Stabilisation with Titanium and Niobium
28
28
29
31
31
2.10
AISI Type 441 Stainless Steels
32
2.11
Calphad Methods
33
2.11.1
2.12
Thermodynamic Softwares
35
Intermetallic Laves Phase
2.12.1
2.12.2
2.12.3
36
Crystallographic Structure
Occurrence
Orientation Relationship
36
37
38
CHAPTER THREE
39
THEORY OF PRECIPITATION REACTIONS IN STEELS
39
3.1
Introduction
39
3.2
Classical Theory of Nucleation
39
3.2.1
Activation Energy for Nucleation within the Matrix
3.2.2
Activation Energy for Nucleation on the Grain Boundary
3.2.3
Misfit Strain Energy Around the Particle
3.2.4
Interfacial Energy
3.2.4.1
Fully Coherent Precipitates
3.2.4.2
Incoherent Precipitates
45
3.2.4.3
Semi-Coherent Precipitates
46
Nucleation Rate
The Time-Dependent Nucleation Rate
Chemical Driving Force
46
47
48
3.2.5
3.2.6
3.2.7
3.3
40
40
42
43
44
Growth by Supersaturation
3.3.1
3.3.2
49
Diffusion Controlled Growth Rate
Multicomponent Diffusion Growth
50
51
3.4
Transformation Kinetics
54
3.5
Overall Transformation Kinetics
55
3.5.1
3.5.2
The Robson and Bhadeshia Model
Fujita And Bhadeshia Model
55
56
P a g e | viii
3.6
Capillarity
57
3.7
Dissolution of the Metastable Phase
58
3.8
Particle Coarsening
58
3.8.1
3.8.2
3.8.3
3.9
Diffusion Controlled Coarsening of the Particles within Matrix
Diffusion Controlled Coarsening of the Particles on Grain Boundary
Diffusion Controlled Coarsening of the Particles on Subgrain Boundaries
Summary
58
59
60
60
CHAPTER FOUR
62
EXPERIMENTAL PROCEDURES
62
4.1
Materials
62
4.2
Thermodynamic Modelling
64
4.3
Heat Treatments
64
4.3.1
4.3.2
4.3.3
4.3.4
4.4
64
64
66
66
Mechanical Testing
4.4.1
4.4.2
4.4.3
4.5
Laves Phase Dissolution/Precipitation Temperatures
Heat Treatment for the Embrittling Effect
Hot-Rolling of Experimental Alloys
Laves phase Kinetic Study
67
Tensile Tests
Notched Charpy Impact Test
Hardness Tests
67
68
68
Microanalysis of Specimens
68
4.5.1
Optical Microscopy
4.5.2
Transmission Electron Microscopy (TEM)
4.5.2.1
Preparation of TEM Specimens
69
69
69
4.5.3
70
4.6
Scanning Electron Microscopy (SEM)
Identification of Precipitates
70
4.6.1
XRD Study
4.6.1.1
Specimen Preparation
70
70
4.6.1.2
4.6.2
4.7
Analysis
71
Electron Diffraction Patterns
76
The Orientation Relationship Between the Laves Phase and the Matrix
77
CHAPTER FIVE
78
THERMODYNAMIC MODELLING
78
5.1
Introduction
78
5.2
Description of Thermo-Calc® Software
78
5.3
Experimental Alloys
79
5.4
Possible Stable Phases at Equilibrium
81
5.5
Phase Diagrams
82
P a g e | ix
5.6
Property Diagrams
84
5.7
Relative Phase Stabilities
84
5.8
Equilibrium Chemical Composition of the Laves Phase
89
5.9
Driving Force for Nucleation
94
5.10
Summary
96
CHAPTER SIX
100
EXPERIMENTAL RESULTS
100
6.1
Introduction
100
6.2
Microstructural Analysis of an AISI Type 441 Ferritic Stainless Steel
100
6.2.1
6.3
Precipitate’s Identification
101
Effect of Annealing Treatment on the Microstructural and Mechanical Properties
6.3.1
6.3.2
6.3.3
6.3.4
107
Microstructural Analysis
Precipitate’s Morphology
Mechanical Properties
Effect of Grain Size on the Mechanical Properties of Steel A
107
112
113
116
6.4
Effect of Annealing Treatment on the Charpy Impact Energy and DBTT
118
6.5
Effect of Re –embrittlement treatment on The Room Temperature Charpy Impact
Energy
119
6.5.1
6.5.2
Effect Of Cooling Rate
Effect of the Reheating Treatment
119
122
CHAPTER SEVEN
125
EXPERIMENTAL RESULTS
125
EFFECT OF THE STEEL’S COMPOSITION
125
7.1
Effect of Annealing Treatment on Steel B
125
7.2
Effect of the Equilibrium Laves Phase Volume Fraction on the Room Temperature
Charpy Impact Energy
126
Effect of Annealing Treatment on the Embrittlement of the Experimental Stainless
Steels C to E
128
7.3
CHAPTER EIGHT
131
EXPERIMENTAL RESULTS
131
LAVES PHASE KINETICS STUDY
131
8.1
Introduction
131
8.2
Equilibrium Laves Phase Fraction
131
8.3
Laves Phase Transformation Kinetics
133
8.4
Temperature Effect on Isothermal Transformations
134
Page |x
8.5
Effect of the Grain Size on the Transformation Kinetics of Laves Phase
135
8.6
Effect of the Steel’s Composition on the Laves Phase’s Transformation Kinetics
136
8.7
Microstructural Analysis of the Transformation Kinetics
137
8.8
Orientation Relationship Between the Laves Phase and the Ferrite Matrix
141
CHAPTER NINE
144
DISCUSSIONS
144
LAVES PHASE EMBRITTLEMENT
144
9.1
Introduction
144
9.2
Precipitates Found in AISI 441 Ferritic Stainless Steel
144
9.2.1
9.2.2
9.3
Effect of the Steel’s Composition on the Precipitate’s Solvus Temperature
Effect of the Steel’s Composition on the Precipitate’s Composition
Embrittlement of Type 441 Ferritic Stainless Steel
148
9.3.1
Effect of Grain Size on Flow Stress: the Hall-Petch Relationship
9.3.2
Crack Nucleation
9.3.3
Effect of Precipitates in the Embrittlement of this Steel
9.3.3.1
Embrittlement and the Cottrell’s Approach
9.4
145
147
148
149
151
151
9.3.3.2
Embrittlement by Grain Boundary Precipitates (The Smith’s Model)
154
9.3.3.3
Effect of Cooling Rate
157
Recrystallisation and Grain Growth
159
CHAPTER TEN
163
DISCUSSIONS
163
TRANSFORMATION KINETICS MODELLING
163
10.1
Introduction
163
10.2
Modelling in Kinetics of Laves Phase Precipitation
163
10.2.1
10.2.2
10.2.3
10.2.4
Nucleation
Growth
Coarsening
Diffusion Coefficients
163
164
165
166
10.3
Parameters Required for Calculations
166
10.4
Calculations
169
10.4.1
10.5
Volume Fraction and Particle Size
Summary
170
172
CHAPTER ELEVEN
173
CONCLUSIONS AND SUGGESTIONS FOR FURTHER WORK
173
11.1
Conclusions
173
P a g e | xi
11.2
Suggestions for the Further Work
176
APPENDIX A
177
APPENDIX B
179
REFERENCE
183
P a g e | xii
TABLE OF FIGURES
Figure 1.1. Catalytic converters growth industry in South Africa [].
Figure 2.1. The effect of nickel and chromium content on the structure of the main stainless
steels. Note in particular that Ni is a strong austenite former whereas Cr is a strong ferrite
former [].
Figure 2.2. Effect of carbon and chromium contents on the structure of some of the main
stainless steels. ELI is extra low interstitial steel [].
Figure 2.3. Fe-Cr equilibrium phase diagram [].
Figure 2.4. Shifting of the boundary line (α +γ)/γ in the Fe-Cr system through additions of C + N
[].
Figure 2.5. The effect of the grain size on the impact toughness for several Fe – 25Cr ferritic
stainless steels. Steel 59 and 68 are the alloy numbers [].
Figure 2.6. The effect of the interstitial content on the impact transition temperature for the Fe–
18Cr–2Mo and Fe–25Cr steels with the grain sizes within the range 35 to 75 µm. The
numbers given correspond to the alloy number [].
Figure 2.7. A schematic diagram of the grain boundary showing carbides in the chromiumdepleted zone near the grain boundaries.
Figure 2.8. Effect of niobium and titanium additions on the impact toughness of 18%Cr-2%Mo
ferritic stainless steels [].
Figure 2.9. Zener’s model for cleavage fracture.
Figure 2.10. Stroh’s model for cleavage fracture.
Figure 2.11. Cottrell’s model for cleavage fracture.
Figure 2.12. Smith’s model for cleavage fracture.
Figure 2.13. Automobile exhaust system components.
Figure 2.14. Effect of alloying element additions on the 0.2% proof strength at 900 °C of a
13%Cr ferritic steel [].
Figure 2.15. The solvus temperatures of the precipitates found in stabilised ferritic stainless
steels [].
Figure 2.16. Schematic flow diagram showing the Calphad approach used to obtain a
thermodynamic description of a multicomponent system.
Figure 2.17. The three polytypes of the Laves phase structure in a hexagonal setting.
Figure 3.1 The free energy change associated with the formation of a stable nucleus with the
radius r.
Figure 3.2. The ratio of the free energy required to form a nucleus on various types of grain
boundary sites to that required to form a nucleus in the grain matrix, is plotted as a
function of the contact angle parameter cos θ.
P a g e | xiii
Figure 3.3. Different possibilities of the precipitate’s interface on grain boundaries.
Figure 3.4. Illustration of the variation of the function f(c/a) of an incoherent nucleus with its
shape.
Figure 3.5. Fully coherent precipitates, with no broken inter-atom bonds and with δ=0. The
interface is indicated by the circle.
Figure 3.6. Coherent precipitate with different lattice parameters only in the vertical direction.
The volume influenced by the lattice misfit, ε is marked by the dotted line.
Figure 3.7. The solute concentration profile during diffusion - controlled growth of β from α.
cαβ and cβα are concentrations at the interface α/β in the matrix α and the precipitate
β, respectively.
Figure 3.8. A schematic isothermal section through the Fe-C-M phase diagram, showing the
ferrite matrix α and alloy carbide β fields. The alloy composition is plotted as point a
[4].
Figure 3.9. Distribution of the solute when (a) both (β) and (γ) are precipitating, and (b) where
the precipitation of (β) has been completed. Note that c′ is the instantaneous solute
concentration in the matrix (α) [99].
Figure 3.10. The kinetics of the precipitation sequence in 9Cr-0.8Nb ferritic stainless steel [101].
Figure 4.1. Experimental plan.
Figure 4.2. Embrittlement through reheating to determine the effect of the Laves phase reprecipitation on the DBTT and upper shelf energy of steel A.
Figure 4.3. Embrittlement through cooling to determine the effect of the Laves phase reprecipitation on the Charpy impact toughness.
Figure 4.4. The furnace used for the precipitation kinetic study. (A) tube furnace; (B)
temperature controller; (d) data logger; (D) type k thermocouple; (E) recording computer.
Figure 4.5. The temperature gradient of the Charpy impact specimen inside the furnace.
Figure 4.6. Schematic diagram of the subsize tensile test specimen.
Figure 4.7. The XRD powder pattern of the phases that were expected to be present in type 441
stainless steel as generated using a PowderCell software.
Figure 4.8. The XRD powder pattern showing the peak’s positions of the carbide and nitrides of
titanium and niobium. Notice the position of the (Ti,Nb)(C,N).
Figure 4.9. A typical XRD scan of the precipitate’s residue from Steel A showing the presence of
the Laves phase peaks (indicated by the lines in the top figure). The remaining peaks are
the carbides and nitrides, indicated by (∗). Note the good residual difference between
the calculated and the measured spectrum as is shown by the spectrum below.
Figure 4.10. The single crystal electron diffraction pattern
Figure 5.1. Thermo-Calc® calculation of the isopleth diagram for type 441 stainless steel with a
constant amount of alloying elements and 0 to 0.5 wt.% of carbon. Below any line, these
represents the stable region for the phase.
P a g e | xiv
Figure 5.2. Thermo-Calc® calculation of the isopleth diagram for the high Mo-containing type
444 ferritic stainless steel E with a constant amount of alloying elements and 0 to 0.5wt.%
of carbon.
Figure 5.3. The property diagram that shows the dependence of phase proportion on
temperature; (a) mole fraction of stable phase and (b) weight fraction of stable phase.
Figure 5.4. Thermo-calc® plots of weight fraction of the stable phases as a function of the
temperature in the Steel A with composition 0.444Nb-0.153Ti; (a) Laves phase and (b)
(Ti,Nb)(CN).
Figure 5.5. Thermo-calc® plots of weight fraction of the stable phases as a function of the
temperature in the Steel B with composition 0.445Nb-0.149Ti; (a) Laves phase and (b)
(Ti,Nb)(CN)).
Figure 5.6. Thermo-calc® plots of weight fraction of the stable phases as a function of the
temperature in the Steel C with composition 0.36Nb-0.171Ti; (a) Laves phase and (b)
(Ti,Nb)(C,N).
Figure 5.7. Thermo-calc® plots of weight fraction of the stable phases as a function of the
temperature in the Steel D with composition 0.36Nb-0.171Ti-0.54Mo; (a) Laves phase
and (b) (Ti,Nb)(C,N).
Figure 5.8. Thermo-calc® plots of the weight fraction of the stable phases as a function of the
temperature in the Steel E with composition 0.251Nb-0.106Ti-1.942Mo; (a) Laves phase
and (b) (Ti,Nb)(C,N).
Figure 5.9. The normalised chemical composition of the Laves phase in Steel A: (a) is the mole
fraction and (b) is the weight fraction of a component in the phase.
Figure 5.10. The normalised chemical composition of the Laves phase in Steel B: (a) is mole
fraction and (b) is a weight fraction of a component in a phase.
Figure 5.11. The normalised chemical composition of the Laves phase in Steel C: (a) is the
mole fraction and (b) is the weight fraction of a component in the phase.
Figure 5.12. The normalised chemical composition of the Laves phase in Steel D: (a) is the
mole fraction and (b) is the weight fraction of the component in the Laves phase.
Figure 5.13. The normalised chemical composition of the Laves phase in Steel E: (a) is the
mole fraction and (b) is the weight fraction of a component in the Laves phase.
Figure 5.14. The free energy change ∆G for the precipitation reaction of Laves phase in ferrite
with temperature for : (a) Steel A; (b) Steel B; (c) Steel C; (d) Steel D and (e) Steel E,
calculated using Thermo-Calc®, (G = J/mol).
Figure 6.1. Micrographs from Steel A in the as received hot rolled condition, showing the grain
structure. (a) optical microscopy image and (b) SEM images. Note the large difference in
magnification with figure (a) showing the “particle decorated” grain structure while figure
(b) shows primarily the “particle decorated” subgrain structure.
Figure 6.2. Micrographs of the as received hot rolled Steel A showing its grain structure. (a) An
optical microscopy image and (b) a SEM image.
Figure 6.3. SEM – EDS micrograph showing a precipitate consisting of a central cubic core of a
mainly titanium containing particle surrounded by a cluster of niobium precipitates.
P a g e | xv
Figure 6.4. Transmission electron micrographs of particles from extraction replicas and their
analyses by electron diffraction and EDS of the as-received hot rolled Steel A showing
different particle morphologies.
Figure 6.5. A typical XRD scan of the precipitate residue after electrolytic extraction from Steel
A, i.e. the as received material, showing the presence of the Laves phase peaks
(indicated by the lines in the top figure). The remaining peaks are the carbides and
nitrides, indicated by (∗) and the α - Fe matrix, indicated by (♣). Note the good residual
difference between the calculated and the measured spectrum as is shown by the
spectrum below.
Figure 6.6. Optical micrographs of the specimens from Steel A after annealing at different
temperatures for 30 minutes followed by water quenching (In comparing the
microstructures, note the differences in magnifications).
Figure 6.7. SEM micrographs of Steel A showing the effect of annealing temperature on the
morphology of the second phase.
Figure 6.8. TEM micrographs from Steel A showing the presence of the fine Laves phase
precipitates on the subgrain boundaries of the specimens that were annealed at the
shown different temperatures for 1 hour and then water quenched.
Figure 6.9. Thin foil electron transmission micrographs from steel A, annealed at 700 °C for 1
hour and then water quenched. The micrographs show (a) the nucleation of the Laves
phase precipitates on grain boundaries and dislocations and (b) some fine matrix
precipitates surrounded by a strain halo as well as dislocation nucleated precipitates.
Figure 6.10. Effect of annealing temperature on the room temperature Charpy impact energy of
the as hot rolled and annealed AISI 441 stainless Steel A. The samples were annealed
for 30 minutes and then water quenched.
Figure 6.11. Examples of the Charpy fracture surfaces at different magnifications of steel A (a &
b) from the as received specimen; (c & d) after annealing at 850 °C; and (e & f) after
annealing at 900°C.
Figure 6.12. Effect of annealing temperature above 850 °C on the grain size and Vickers
hardness for the AISI type 441 ferritic stainless Steel A.
Figure 6.13. TEM micrograph showing the presence of a dislocation substructure and some fine
Laves precipitates in the as received hot rolled specimen of Steel A, indicating a lack of
full dynamic recrystallisation during the last stage of hot rolling.
Figure 6.14. Effect of annealing temperature at 850 °C and above on the tensile strength and
elongation of the 441 stainless steel A.
Figure 6.15. Charpy impact energy of the 441 ferritic stainless steel A as a function of the test
temperature from specimens that were annealed at four different temperatures, both
within and outside the Lave phase formation region.
Figure 6.16. Effect of linear cooling rate in °C/s on the room temperature impact toughness of
the specimens from Steel A that were cooled at linear cooling rates from 850 °C and 950
°C, respectively.
Figure 6.17. TEM micrographs of the samples of Steel A that were solution annealed at 850 °C
and 950 °C for 5 min then cooled at 60 °C/sec. (a & b) solution treated at 850 °C; (c & d)
solution treated at 950 °C. Note the differences in the microstructures from both
samples.
P a g e | xvi
Figure 6.18. TEM micrographs of the samples from Steel A after being cooled at 1 °C/sec from:
(a) solution annealed at 850 °C and (b) 950 °C for 5 min before cooling.
Figure 6.19. Effect of the cooling rate on the volume fraction of the Laves phase in Steel A after
cooling at different rates from annealing at 850°C.
Figure 6.20. Charpy impact energy of Steel A as a function of the test temperature of specimens
first solution annealed at 950°C and then re-annealed at different temperatures.
Figure 6.21. Optical microscopy micrographs showing microstructural evolution in Steel A
during re – heating treatments after an original solution treatment at 950°C.
Figure 6.22. Effect of the Laves phase re-precipitation in Steel A on the hardness of the material
during embrittlement treatment after an original solution treatment at 950°C.
Figure 7.1. Effect of annealing treatment on the Laves phase’s % volume fraction, grain size
and the Charpy impact toughness of the 441 ferritic stainless steel, Steel B.
Figure 7.2. Effect of the Laves phase precipitation kinetics on the Charpy impact toughness of
Steel B.
Figure 7.3. Optical micrographs of the specimens from steel B in the (a) as received plant hot
rolled condition and (b) to (d) after being annealed at different temperatures from 850 to
950°C for 30 minutes followed by water quenching.
Figure 7.4. Effect of annealing temperature on the room temperature Charpy impact energy of
the laboratory hot rolled materials. The samples were annealed for 30 minutes at
different temperatures and then water quenched: Steel C (Nb-Ti alloy); Steel D (Nb-TiMo alloy) and Steel E (Type 444 alloy).
Figure 7.5. The microstructure of the laboratory hot-rolled experimental steels, showing different
grain size distributions if compared to those of the commercial Steels A and B: (a) Steel
C; (c) Steel D; and (d) Steel E.
Figure 8.1. The Laves phase volume fraction – temperature/time curves during isothermal
annealing in the temperature range 600 °C to 850 °C.
Figure 8.2. The Laves phase transformation curves according to the Johnson–Mehl–Avrami–
Kolmogorov (JMAK) type of equation.
Figure 8.3 A time – temperature – precipitation (TTP) diagram for the Laves phase formation in
Steel A.
Figure 8.4. Effect of the grain size on the Laves phase kinetics transformation in Steel A. The
specimens were annealed first at 850 and 950°C respectively to set different grain sizes
and were then annealed both at 750 °C for different annealing periods.
Figure 8.5. Effect of the steel’s composition on the Laves phase transformation kinetics. The
specimens from these steels were all annealed at 750 °C for different annealing periods.
Figure 8.6. TEM micrographs of the specimen of Steel A annealed at 600 °C; (a) a low
magnification micrograph shows coarse grain boundary Laves phase precipitates, and (b)
the same area but at a high magnification, showing Laves phase precipitates nucleated
on subgrain boundaries and dislocations.
Figure 8.7. TEM micrographs of the specimen of Steel A annealed at 750 °C; (a) a low
magnification micrograph showing grain and subgrain boundary Laves phase
P a g e | xvii
precipitates, and (b) at a high magnification, showing Laves phase precipitates nucleated
on the subgrain boundaries.
Figure 8.8. TEM micrographs of the specimen annealed at 750 °C; (a) at a low magnification,
showing grain boundary Laves phase precipitates, and (b) at a higher magnification
showing Laves phase precipitates nucleated on the subgrain boundaries.
Figure 8.9. Transmission electron micrographs and corresponding selected area diffraction
(SAD) pattern from Steel A annealed at 600 °C.
Figure 8.10. Transmission electron micrographs and corresponding selected area diffraction
(SAD) pattern from Steel A annealed at 750 °C.
Figure 8.11. Transmission electron micrographs and corresponding selected area diffraction
(SAD) pattern from Steel A annealed at 800°C.
Figure 9.1. TEM micrograph shows the presence of the M6C or (Fe3Nb3C) type carbide in the
subgrain structure from Steel A. Note that the specimen was annealed at 700 °C for 30
minutes and other fine particles were determined to be Fe2Nb Laves phase particles.
Figure 9.2. Comparison between experimental and Thermo-Calc® calculated weight fractions of
Laves phase in Steel A. The points and dotted line represent the experimental results
while the full line is as predicted by Thermo-Calc® for this steel.
Figure 9.3. The effect of grain size on the yield strength of Steel A.
Figure 9.4. A room temperature tensile test of the specimen of Steel A that was annealed at 850
°C for 30 minutes and then water quenched.
Figure 9.5. High resolution field emission scanning microscope image showing the cracking of
(Ti,Nb)(C,N) particles after impact testing the specimen at room temperature. This
specimen of Steel A was annealed at 850 °C followed by quenching in water.
Figure 9.6. The plot of transition temperature versus {ln d1/2} of 441 ferritic stainless steel, Steel
A.
Figure 9.7. Effect of annealing temperature above 850 °C on the grain size for the AISI type 441
stainless steel, Steel A.
Figure 9.8. TEM micrographs of the microstructures of the specimens from Steel A that were
annealed at (a) 850 °C and (b) 900 °C. Note that with the specimen that was annealed at
900 °C, there were no grain boundary Laves phase precipitates.
αβ
Figure 10.1. The relationship between ln x Nb
and T-1 for AISI type 441 ferritic stainless steel.
Figure 10.2.
Comparison between the experimental data and calculated isothermal
transformation curves for the Laves phase’s precipitation at 700 °C in the AISI type 441
ferritic stainless, with No = 4.3 x 1014 m-3 and γ = 0.331 Jm -2.
Figure 10.3.
Comparison between the experimental data and calculated isothermal
transformation curves for the Laves phase precipitation at 800 °C in the AISI type 441
ferritic stainless, with No = 2.9 x 1013 m-3 and γ = 0.331 Jm -2.
P a g e | xviii
NOMENCLATURE
α3
is the three-dimensional parabolic
rate constant
β*
atomic impingement rate
δ
volume misfit of the precipitate in the
matrix
crαβ
solute concentration in the α matrix
that is in equilibrium with a spherical
particle of β and r is the radius of
curvature
cαβ
equilibrium solute concentration in
the α matrix at which r→∞
cβα
corresponding concentration in the β
which is in equilibrium with α;
δdisl
effective diameter of dislocation
δgb
width of the grain boundary
γ
interfacial surface energy per unit
area associated with the interface of
the two phases
ci
mole fraction of species i
cj
mole fraction of species j
γf
effective surface energy of ferrite
d
grain size
γs
surface energy of the exposed crack
surface
D
diffusion coefficient of the rate
controlling solute atoms in the matrix
γT
true surface energy
Ddisl
σi
friction stress
diffusion
coefficient
dislocation
σy
yield strength
Dgb
diffusion coefficient along the grain
boundary
ν
down
a
Poisson’s ratio
E
Young’s elastic modulus
α
lattice spacing of the matrix
fGB
β
lattice spacing of the precipitate
phase
fraction of potential grain boundary
sites filled by solute
Gm
shear modulus of the matrix
νb
mobility rate
Gr
growth rate
τ
incubation time
∆G
τe
effective shear stress
molar free energy change of the
precipitate reaction
τi
lattice friction shear stress
∆Gv
τN
Gibbs chemical free energy released
per unit volume of new phase
shear stress for crack nucleation
τy
∆Gε
misfit strain energy per unit volume
yield shear stress
υβ
molar volume of the phase β,
∆G*
known as the activation energy
G°
τs
shear stress
Φ
extent of the reaction parameter
Gibbs energy due to the mechanical
mixing of the constituents of the
phase
θ
contact angle
a
mean atomic lattice distance of the
matrix phase
b
Burgers vector
ν
ν
c
cα
average concentration of the solute
in the matrix alone
equilibrium solute composition within
the matrix
id
Gmix ideal mixing contribution
xs
Gmix excess Gibbs energy of mix (the
non-ideal mixing contribution)
∆Gε
Gm − H
Gibbs energy relative to a standard
element reference state (SER)
Planck constants
h
H
strain energy
SER
m
SER
m
enthalpy of the element in its stable
state
P a g e | xix
k
Boltzman constant
∆pppt
retarding force exercised by particles
on the grain boundary
k ys
Hall – Petch constant for shear
Lgb
length of grain boundary per unit
volume
Q
activation energy for diffusion
r*
critical radius
Lki , j
binary
interaction
parameter
between species i and j
r0
initial average particle radius
R
gas constant
M0
intrinsic grain boundary mobility in
pure material
Sgb
surface area of grain boundary per
unit volume
MT
overall mobility due to intrinsic plus
solute drag
t
time
T
absolute temperature
MB
mobility in the presence of solute
drag elements
V’
instantaneous volume fractions of
alloy precipitates
n
number of dislocations in the pileup
Veq
N&
nucleation rate
equilibrium volume fractions of alloy
precipitates
N′
number of dislocations that meet
each particle
Vβ
instantaneous fraction
Vβα
maximum fraction of a given phase
N*
concentration of critical – sized
nuclei
V
Nc
density of
corners
Xs
atom fraction of solute in the bulk
metal
N0
initial number density of nucleation
sites per unit volume
z
coordinate normal to the interface
with the value z*
pd
driving force for the grain boundary
mobility
the
grain
boundary
iα
maximum volume fraction of the ith
phase
P a g e | xx
CHAPTER ONE
GENERAL INTRODUCTION
1
1.1
INTRODUCTION
Columbus Stainless is the primary manufacturer of flat cast and wrought stainless steel
products in Southern Africa. One of the growth sectors in the use of stainless steels is in
the automotive components industry and more particularly, in catalytic converters in the
automotive industry. The manufacture of automobile emission control systems in South
Africa is one of the fastest growing industry sectors in the world. Founded on the growth
and development of catalytic converters, South Africa supplies in excess of 10% of the
world’s production, mainly stemming from its dominance in PGM (Platinum Group
Metals) production.
Catalytic converters are the largest of the auto component
groupings being exported from South Africa and now amount to $500 million/year. The
growth of the local catalytic converter industry has been spectacular, see Figure 1.1.
18
16
Converters (millions)
14
12
10
8
6
4
2
0
1992
1994
1996
1998
2000
2002
2004
2006
Figure 1.1. Catalytic converters growth industry in South Africa [1].
The operating temperatures for catalytic converters are in the region of 900°C but are
associated with a frequent temperature variation as automobiles are used intermittently.
Thus, the material for this application requires excellent thermal fatigue resistance and
high temperature strength.
The primary steel used in this application is type 441
stainless steel, (which is equivalent to DIN 1.4509). This steel is fully ferritic over a wide
range of temperatures. Type 441 is a dual stabilised (titanium and niobium) ferritic
stainless steel with 18 wt% chromium, and the typical composition of this steel is shown
in the Table 1.1 below. Titanium and Niobium carbides are more stable than chromium
carbides and prevent the formation of chromium carbides on grain boundaries (which
Page |1
results in sensitisation of the alloy in near grain boundary regions).
The dual
stabilisation by both Ti and Nb imparts beneficial corrosion resistance, oxidation
resistance, high temperature strength and formability to the steel [2].
Table 1.1. A typical chemical composition (wt%) of type 441 ferritic stainless steels.
C
N
Mn
Si
Min
Max
Cr
Ni
17.50
0.03
0.045
1.00
1.00
18.50
Ti
Nb
S
P
Remarks
1.00
0.015
0.04
0.10
0.50
0.60
%Nb≥ 3xC + 0.3
Many researchers have investigated the effects of niobium stabilisation on ferritic
stainless steels for their use in automotive exhaust systems [3,4,5,6], and have found
that niobium additions can improve the high temperature strength of ferritic stainless
steels by solid solution strengthening, which allows its operation at such high
temperatures. It is, therefore, important to know how much of the niobium precipitates
out as carbides or carbo-nitrides and which fraction remains in solution.
Niobium
stabilised ferritic stainless steels form the precipitates of Nb(C,N), M6C (Fe3Nb3C) and
the Laves phase type Fe2M (Fe2Nb).
Because C and N remain in the steels, Nb(C,N) and Ti(C,N) are easily formed in some
processing stages, e.g. in hot–rolling and annealing processes. These carbo – nitrides
increase the strength and decrease both the toughness and ductility. In the work by
Fujita et al. [5], on 13Cr – 0.5Nb, the authors have observed that the high temperature
strength of Nb added steels decreases during high temperature ageing, caused by
coarse M6C (Fe3Nb3C) formation in ageing.
The volume fraction of the Laves phase was observed to reach a maximum at 700°C
and its dissolution occurs at temperatures over 900°C [7,8]. The same observation has
been made by the use of thermodynamic software such as Thermo-calc® [3, 9]. The
Laves phase precipitates firstly at the grain boundaries as a fine precipitate and as the
steel is slowly cooled from a high temperature of about 900°C, the amount of this Laves
phase increases inside the grains and they then coarsen.
This intermetallic Laves
phase is known to affect both the mechanical properties and corrosion resistance of
ferritic stainless steel, mostly negatively. In one instance, it has been found that the fine
precipitates of Laves phase at the grain boundaries improve the high temperature
strength when still fine [5]. However, the rapid coarsening of the Laves phase at high
temperatures reduced the high temperature strength [10]. The exact mechanism of
strength reduction is still not clear and requires clarification.
Sawatani et al [8],
researched the effect of Laves phase on the properties of dual stabilised low carbon
Page |2
stainless steels as related to the manufacturing process of Ti- and Nb-stabilised low C,
N-19%Cr-2%Mo stainless steel sheets, and they have found that the Laves phase has a
significant influence on the mechanical properties of the steel. It was found that Laves
phase on the grain boundaries shifts the brittle to ductile transition temperature to higher
temperatures, and that large amounts of Laves phase degrade the room temperature
ductility of cold rolled and annealed sheet and greatly enhance its strength. It was also
observed that after a 20% cold rolled reduction (of a sheet that had been cold rolled and
annealed after reductions between 0 and 92%) that there was a peculiarly rapid
precipitation of Laves phase, which caused a severe degradation of the mechanical
properties [11].
1.2
PROBLEM STATEMENT
Columbus Stainless has experienced an embrittlement problem at times during the
manufacturing process of type 441 stainless steel. This problem is considered generic
because it appears from the hot band material prior to annealing, whereby the materials
become embrittled after hot rolling and coiling. It was assumed that the embrittlement
might be attributed to the formation of an intermetallic Laves phase type. The reason for
this assumption is that after the hot finishing mill (Steckel mill), the temperature of the
steel strip is approximately 850°C, and the strip is then rapidly cooled (to avoid the
Laves phase transformation temperature range) to approximately 650°C by means of
laminar cooling with water sprays on the run-out table and is then coiled. The coil is left
to cool for 3 to 5 days to approximately 50°C before further processing. For instance,
one particular hot rolled coil was found to be brittle and could not be processed further.
It appears furthermore that rapid cooling (by water sprays) after hot rolling alleviates the
embrittlement problem, and that slow cooling of this particular hot rolled coil might have
allowed the precipitation of the Laves phase below 650ºC and that this resulted in the
embrittlement of the material. Also, the preliminary evaluation by Columbus Stainless of
this steel indicated that this embrittlement is not related to a coarse ferritic grain size
effect, as the degree of embrittlement was far higher than expected from this source.
1.3
OBJECTIVES
Laves phase precipitation is believed to be detrimental to mechanical properties, in
particular leading to a low toughness in this type of steel. The following two objectives
have been identified:
Page |3
1. Firstly, by making use of this reject material, to determine those process variables
that may affect the toughness of the material after processing, principally
annealing temperatures and cooling rates from typical hot rolling temperatures;
and
2. To model and experimentally determine the kinetics of precipitation of this phase,
since it would then be possible to predict what volume fraction of Laves phase
forms firstly, during processing and secondly, what forms possibly throughout the
component’s lifetime in exhaust systems.
The type 441 steel (composition shown in Table 1.1) is susceptible to Laves phase
formation over a relatively wide temperature range. A number of studies of Laves phase
precipitation in this steel provide suitable experimental data with which to test the
precipitation model [12].
Page |4
CHAPTER TWO
LITERATURE REVIEW
2
2.1
INTRODUCTION
Stainless steels are iron-base alloys containing a minimum of 11wt.% chromium content
for adequate corrosion resistance. This chromium content is the minimum that prevents
the formation of “rust” in air or in polluted atmospheres by forming a very thin surface
film of chromium oxides known as the “passive film”, which is self-healing in a wide
variety of environments. Today, the chromium content in stainless steels approaches
30wt.% in some alloys and other elements are often added to provide specific properties
or ease of fabrication.
Some of these elements are nickel (Ni), nitrogen (N) and
molybdenum (Mo) which are added for corrosion resistance; carbon (C), Mo, N, titanium
(Ti), aluminium (Al) and copper (Cu) which are added for strength; sulphur (S) and
selenium (Se) are added for machinability; and Ni is added for formability and
toughness, particularly to obtain an austenitic microstructure which is far less prone to
the loss of toughness from a large grain size.
2.2
CLASSIFICATION OF STAINLESS STEELS
Stainless steels are divided into three groups according to their crystal structures:
austenitic (face-centred cubic, fcc), ferritic (body-centred cubic, bcc) and martensitic
(body-centred tetragonal or cubic, bct). Stainless steels containing both austenite and
ferrite usually in roughly equal amounts are known as “duplex stainless steels”.
The general considerations for the choice of the base metal are that it should have the
following properties: (a) since these alloys are used at high temperatures or under
demanding conditions, they should have adequate corrosion resistance. This implies
that either Cr or Al at a level of about 15% or higher, should be added to the alloy. (b)
The room temperature structure should be austenitic, primarily to avoid the formation of
martensite during cooling to room temperature and secondly, to prevent a ferritic
structure which has a lower solubility for carbon and favours the formation of
intermetallic precipitates rather than carbides. The addition of austenite formers (mainly
Ni, Mn and N) is, therefore, necessary in austenitic stainless steels.
Page |5
The Fe-Cr-Ni system as the base alloy is by far the most suitable as large quantities of
Cr can be taken into solution and maintained in solution down to room temperature.
Secondly, Ni is also a strong austenite former and Cr lowers the martensite start (Ms)
temperature sufficiently to avoid the formation of martensite, see Figure 2.1.
Both nitrogen and carbon are strong austenite formers, with nitrogen being increasingly
used to provide certain attractive properties such as good fracture strength and it also
improves the corrosion resistance [13,14]. From Figure 2.2, it is noted that carbon is a
very strong austenite former, and if the carbon content is very low, slightly more Ni may
have to be added to compensate for the loss of the austenitic properties of the carbon.
Figure 2.1. The effect of nickel and chromium content on the structure of the main
stainless steels. Note in particular that Ni is a strong austenite former whereas Cr
is a strong ferrite former [15].
Figure 2.2. Effect of carbon and chromium contents on the structure of some of the main
stainless steels. ELI is extra low interstitial steel [15].
Page |6
2.2.1 FERRITIC STAINLESS STEEL
This stainless steel derives its name from the bcc crystallographic structure that is
generally stable from room temperature up to the liquidus temperature. Typically, ferritic
stainless steel contains approximately 11 to 30wt.% chromium and small amounts of
other alloying elements. The chromium additions give the steel its corrosion resistance
and it further stabilises the bcc crystal structure. Recently, a very low (C+N) content has
been specified in the so-called super-ferritic stainless steels.
The higher alloy
compositions can also include up to 4%Ni, provided this does not alter their fully ferritic
structure. Due to their adequate corrosion resistance and lower cost, ferritic stainless
steels are chosen over austenitic stainless steels in less severe applications such as a
replacement to mild carbon steels in automobile exhaust systems.
However, poor
weldability, which leads to low toughness and is also associated with grain growth in the
HAZ, limits their use even with very low carbon levels.
2.2.2 AUSTENITIC STAINLESS STEEL
As with ferritic steel, the austenitic stainless steel’s name originates from its fcc
crystallographic structure. These steels contain 16 to 25wt% chromium and 7 to 10%
nickel. Austenitic stainless steel has a high nickel content to stabilise the austenite fcc
structure at room temperature. The increase in alloy content creates a higher cost of
production but the fcc structure exhibits very high ductility, resulting in material with good
formability and very good corrosion resistance. Another advantage of these steels is the
relative ease of recrystallisation, which allows for better control of the mechanical
properties.
2.2.3 MARTENSITIC STAINLESS STEEL
Martensitic stainless steels contain 12 to 17% chromium for good corrosion resistance.
However, since chromium is a strong ferritic stabiliser, austenite stabilisers are added so
that the necessary austenite can be formed during solution treatment for the subsequent
martensite formation. Therefore, these steels have a high carbon content to stabilise the
austenite at higher temperatures. The high carbon content will increase the strength
through solid solution strengthening and the precipitation of a large number of (Fe, Cr)
carbides. These steels use the quench and temper process to achieve a very high
strength with reasonable ductility. Because of the high alloy content, these steel have a
superior hardenability.
The disadvantage of the high hardenability often leads to
Page |7
degradation of the corrosion resistance when compared with ferritic and austenitic
stainless steels.
2.2.4 DUPLEX STAINLESS STEEL
These steels contain a mixture of ferrite and austenite phases at room temperature in
order to combine the beneficial properties of both components. These steels typically
contain 18 to 30% chromium and an intermediate amount of nickel (3-9%) that is not
enough for the formation of a fully austenitic structure at room temperature. Duplex
stainless steels have an intermediate level of high mechanical strength and corrosion
resistance properties lying between those of austenitic and ferritic products.
2.3
COMPOSITION OF STAINLESS STEELS
The composition of stainless steel can be related to its non-equilibrium metallurgical
structure by means of a Schaeffler diagram [16], which shows the microstructure
obtained after a rapid cooling from 1050°C to room temperature. It is, therefore, not an
equilibrium diagram and is often used in welding phase analysis. This diagram was
originally established to estimate the amount of delta ferrite (that is, ferrite formed on
solidification, as opposed to alpha ferrite, which is a transformation product of austenite
or martensite) content of welds in austenitic stainless steels. The alloying elements
commonly found in stainless steels are regarded either as austenite stabilisers or as
delta ferrite stabilisers.
The relative “potency” of each element is conveniently
expressed in terms of an empirical equivalence to either nickel (austenite stabiliser) or
chromium (ferrite stabiliser) on a weight percentage basis. The nickel and chromium
equivalents, which form the two axes of the Schaeffler diagram, can be calculated as
follows:
%Ni equivalent = %Ni + %Co + 30%C + 25%N + 0.5%Mn +0.3%Cu
%Cr equivalent = %Cr + 2%Si + 1.5%Mo + 5%V + 5.5%Al + 1.75%Nb + 1.5%Ti + 0.75%W
2.3.1 STRUCTURE OF FERRITIC STAINLESS STEEL
Ferritic stainless steels at room temperature consist of alpha (α) solid solution having a
body centred cubic (bcc) crystal structure. The alloy contains very little interstitial carbon
and nitrogen in solution; most of the interstitial elements appear as finely distributed
carbides and nitrides. A typical phase diagram of the iron-chromium system is shown in
Figure 2.3.
Page |8
Chromium is a ferrite stabiliser and it extends the alpha-phase field and suppresses the
gamma-phase field. This results in the formation of the so-called gamma loop as seen
in Figure 2.3, which, in the absence of carbon and nitrogen, extends to chromium
contents of about 12 – 13wt % [17]. At the higher chromium contents, transformation to
austenite is no longer possible and the metal will remain ferritic up to its melting
temperature. This constitutes an entirely different class of stainless steels in which grain
refinement can no longer be brought about by transformation through heat treatment.
Figure 2.3. Fe-Cr equilibrium phase diagram [17].
With carbon and nitrogen present in these alloys the diagram is modified in certain
respects. The effect of carbon and nitrogen is to shift the limits of the gamma loop to
higher chromium contents and widens the duplex (α + γ) phase area [18]. Figure 2.4
shows the changes in this part of the diagram. However, the solubility levels of the
interstitials in the ferrite matrix are sufficiently low so that it is rarely possible to
distinguish between solute embrittling effects and the effects of second-phase
precipitates. The precipitates, in fact, become more important than the solute when the
amount of interstitial elements significantly exceeds the solubility limit. The presences of
carbon or nitrogen, in amounts in excess of the solubility limit, serve to increase the
ductile to brittle transition temperature (DBTT). This embrittling effect is closely linked to
the amount or the number and size of carbides and nitrides formed on the grain
boundaries but also to the ferrite grain size. Precipitate films act as strong barriers to
slip propagation across the grain boundaries and are also often inherently brittle by
themselves. Grain boundary precipitates are suppressed by quenching from above the
solution temperature when the interstitial content is low enough.
Page |9
(b)
(a)
Figure 2.4. Shifting of the boundary line (α
α +γγ)/γγ in the Fe-Cr system through additions of
C + N [18].
2.4
TOUGHNESS OF FERRITIC STAINLESS STEELS
Some of the metallurgical problems that are encountered in the ferritic stainless steels
are too large a grain size, “475°C” embrittlement, secondary phase precipitation, hightemperature embrittlement and notch sensitivity. They all have been shown to some
extent to affect the ductile–to–brittle transition temperature of the ferritic stainless steels,
mostly negatively by raising it.
2.4.1 EFFECT OF GRAIN SIZE ON BRITTLE BEHAVIOUR
The effect of grain size on the impact toughness of ferritic stainless steels has been well
documented over the years even though it is not always clearly understood. It has been
proven previously that the DBTT tends to increase with increasing grain size, see Figure
2.5 [18,19]. The work done by Ohashi et al [20] has shown that while the transition
temperature increases with increasing grain size, the upper shelf energy is largely
independent of the grain size. They have also reported that in V–notched samples that
the effect of grain size is very noticeable; but in brittle welded samples this grain size
dependency of the upper shelf energy and the DBTT is very small but still observable. It
can be concluded that the coarse grain size tends to promote crack initiation even in a
blunt–notched specimen, but the grain size effect contributes mainly to the resistance of
initiation of brittle fracture and only slightly to the propagation of brittle fracture [20].
P a g e | 10
Figure 2.5. The effect of the grain size on the impact toughness for several Fe – 25Cr
ferritic stainless steels. Steel 59 and 68 are the alloy numbers [18].
Figure 2.6 shows the impact transition temperature as a function of the interstitial
content for two different ferritic stainless steels [21]. For a given interstitial content, both
the DBTT and the upper shelf energies were not affected by the presence of 2% Mo or
the change of Cr content from 18% to 25%. These findings agree with the results of
Woods [22], who showed that for 18% Cr in Ti stabilised stainless steels, that 2% Mo
has no effect on the impact transition temperature.
P a g e | 11
Figure 2.6. The effect of the interstitial content on the impact transition temperature for
the Fe–18Cr–2Mo and Fe–25Cr steels with the grain sizes within the range 35 to 75
µm. The numbers given correspond to the alloy number [21].
For a desirable low DBTT and a high shelf energy, the combined carbon and nitrogen
content should be kept below 150 ppm, below which further decreases have no further
effect.
Above150 ppm, however, increases in interstitial content cause a marked
increase in DBTT, though once above a certain level (C + N ≈ 600 ppm) further
increases are not significant. For these stainless steels a room temperature Charpy
Impact Energy (CIE) value of between 14 to 70 J, and above the DBTT a value of 160J
in the Upper Shelf Energy region, would be considered typical [21].
2.4.2 EMBRITTLEMENT AT 475°C
When ferritic stainless steels containing at least 12wt.% chromium are subjected to
prolonged exposure to temperatures between 400 and 500°C, the notch ductility is
considerably reduced while the tensile strength and hardness increase considerably.
However, the increase in the tensile strength is of no practical significance since the
alloy is extremely brittle [18,19]. This phenomenon is known as “475°C” embrittlement
because the maximum effect occurs at about 475°C [23]. Brittle fracture is transgranular
in nature and is similar in appearance to the low temperature cleavage of unembrittled
P a g e | 12
steels [24]. Transgranular fracture is observed to initiate at the interaction of slip bands
with grain boundaries. However, the precise mechanism of the crack nucleation by slip
bands has not been resolved experimentally [24].
The recognised mechanism of
“475°C” embrittlement is the precipitation through spinodal “unmixing” of Cr-rich alpha
prime (α′) and Fe-rich α phases, arising from the miscibility gap in the Fe-Cr equilibrium
phase diagram, Figure 2.3. The kinetics of α′ precipitation increase with increasing Cr
content. A spinodal phase mixture exists within the miscibility gap, and the precipitation
of the α′ is thought to occur by a nucleation and growth mode outside the spinodal
boundaries and by spinodal decomposition within [19,23].
The phenomenon of “475°C” embrittlement can be removed by heating the embrittled
alloy to a temperature above 550°C for a sufficient amount of time to dissolve the α′
phase followed by rapid cooling to room temperature.
2.4.3 PRECIPITATION OF THE SECONDARY PHASES IN STAINLESS STEELS
Ferritic stainless steel in the temperature range of 500 to 900 °C can also precipitate
intermetallic phases such as sigma (σ), chi (χ) and Laves phases. From the Fe-Cr
equilibrium phase diagram in Figure 2.3, it can be seen that sigma phase (σ) is an
equilibrium phase at temperatures up to 820 °C. The sigma (σ) phase is an intermetallic
compound with approximate composition of FeCr. It is hard and brittle and can result in
a severe harmful influence on the toughness properties of the alloys. Although the
stability range of this phase in the Fe-Cr binary system varies from 519 to 820°C,
several substitutionally dissolved elements can markedly modify this temperature range.
Substitutional elements such as Mo, Si and Ni shift the σ-phase boundary to a lower
chromium range [18,19]. On the other hand, the precipitation of sigma phase can be
strongly accelerated by the pre-existence of carbides in the alloy, particularly the M23C6
type [25,26,27,28]. In the lower Cr content ferritic stainless steels, sigma phase forms
very slowly and is usually a service problem only after a long exposure at elevated
temperatures.
Cold work enhances the precipitation rate of the sigma phase
considerably, and in very high chromium containing steels sigma phase has been found
in an air cooled as-cast structure [18,23]. Increasing the chromium and molybdenum
contents favours the formation of sigma, chi and Laves phases.
The relationship between the toughness and the Fe2Mo Laves phase precipitates in a
9%Cr – 2%Mo ferritic martensitic steel was investigated by Hosoi et al. [29]. They
P a g e | 13
observed that the DBTT increases and the upper shelf energy decreases when the
Laves phase begins to precipitate during ageing.
2.4.4 NOTCH SENSITIVITY
Sensitisation in ferritic stainless steels occurs whenever the steel is heated to a
sensitising temperature (usually above 900°C), such as during thermo-mechanical
processing and/or welding. The precipitation of M23C6, an incoherent carbide with a
complex cubic structure, usually occurs on the grain boundaries, which then form Crdepleted zones along the grain boundary and places the alloy in the sensitised state,
resulting in intergranular corrosion adjacent to the grain boundaries when exposed to
corrosive environments. A sketch showing the carbides and the associated chromiumdepleted zones along the grain boundaries is shown in Figure 2.7. Precipitation of
chromium nitrides such as Cr2N requires slow cooling through a temperature range of
500 to 700°C. Above 700°C, the diffusion of chromium is fast enough to replenish the
chromium at the grain boundaries and below 500°C, the diffusion of chromium is too
slow for the precipitates to form in nitrogen containing alloys. One place where this slow
cooling can occur is in the heat-affected zone (HAZ) in the vicinity of a weld joint.
Figure 2.7. A schematic diagram of the grain boundary showing carbides in the
chromium-depleted zone near the grain boundaries.
Sensitisation can be overcome by one of several methods, all of which increase
production cost:
•
Annealing alloys at temperatures between 950 and 1100°C in the austenite
regions, allowing Cr to rediffuse back into the Cr-depleted zones.
P a g e | 14
•
Retard the kinetics of sensitisation through molybdenum (Mo) additions, which
lengthens the time required for sensitisation by slower diffusion of the Mo which is
then thermodynamically built into the M part of M23C6.
•
Reduce interstitial levels by better control during steelmaking processes.
•
“Free” carbon and nitrogen levels can be further reduced by the addition of strong
carbide and nitride forming elements such as the stabilisers Ti, Nb and Zr,
thereby reducing the formation of the Cr-rich M23C6.
2.4.5 WELDABILITY OF FERRITIC STAINLESS STEEL
The problem faced in joining ferritic stainless steels is grain growth to coarse grain sizes
in the weld zone and the heat-affected zone of fusion welds, and consequently low
toughness and ductility due to the absence of any phase transformations during which
grain refinement can occur. In general, austenitic stainless steels are easily weldable.
When austenitic stainless steel joints are employed in cryogenic and corrosive
environments, the quantity of delta-ferrite in the weld must be minimised or controlled to
avoid degradation during service [30].
It has been shown that the addition of niobium and titanium to stabilise the steel does
not adversely affect the weldability of the steel. The addition may in fact increase the
toughness in titanium stabilised steel [31,32]. In addition dual stabilisation was found to
produce tough, clean weld lines during high frequency welding when compared to only
single titanium stabilised steels. An optimum balance or ratio normally exists between
the titanium and niobium content (with the niobium being the greater i.e. Nb:Ti shown to
be 2:1 for a 17% chromium ferritic steel) for optimum toughness and ductility of welds
[33].
Finally it has been shown that dual stabilised steels can be susceptible to
intergranular hot cracking in the fusion and heat-affected zones. It has been shown that
niobium is in fact the most deleterious element and care should be taken to minimise
this stabilising element as far as possible [23].
2.4.6 EFFECT OF NIOBIUM AND TITANIUM ADDITIONS TO FERRITIC STAINLESS STEELS
Redmond [34] has investigated the effects of residual and stabilising elements on the
toughness of a series of 18%Cr-2%Mo steels which contained 0.015%C and 0.015%N.
The residual elements that were varied were sulphur, manganese and silicon and they
all were shown to have only a minor effect on the impact toughness. The effects of
P a g e | 15
niobium and titanium dual stabilisation on the impact toughness are shown in Figure 2.8.
From this figure it can be seen that as for base metal materials, increasing the titanium
content resulted in a constant increase in the transition temperature. For the alloys with
0.34 and 0.44%Nb the maximum for both occurs at about 30°C for 0.2%Ti. The trend is
less clear when the niobium content was varied and the titanium content was kept
constant. In the welded condition, the trend still does not correlate well with the niobium
content. The niobium stabilised steel had a better impact toughness than the titanium
stabilised steel, with a mixed niobium-titanium stabilisation producing results between
the two [34].
Figure 2.8. Effect of niobium and titanium additions on the impact toughness of 18%Cr2%Mo ferritic stainless steels [34].
2.5
THEORIES OF BRITTLE FRACTURE
Brittle fracture of steel structures has been a matter of considerable concern to both
engineers and metallurgists for many years, it now being generally recognised that the
cleavage mode of failure occupies a central position with respect to the problem. A
physical model of the process has thereby been constructed by others [35], and in so
doing, previous theories were critically examined. The first of these contributions was
the identification of the crack initiators from which cleavage nucleates and triggers a
propagating crack in a ferritic matrix, together with the criterion for this to occur. Stroh
P a g e | 16
[36,37] proposed a wedge dislocation pile-up at a grain boundary as the initiator, Cottrell
[38] a sessile wedge dislocation pile-up resulting from dislocation coalescence, and
Smith [39] the combination of a wedge dislocation pile-up at a grain boundary with a
contiguous brittle second-phase particle or inclusion, broken by the pile-up. All these
initiators are able to nucleate a crack at a neighbouring ferrite grain if assisted by plastic
flow, while the last two are also able to propagate the nucleus of the crack once formed
if assisted by the stress state. An energy balance involving the effective surface energy
of the metallic matrix and the initiator size (the ferritic grain, the second-phase particle,
etc.) allows the tensile stress required for propagating the nucleus to be derived, thus
predicting the maximum principal stress that triggers cleavage and provides a cleavage
fracture criterion.
2.5.1 ZENER’S/STROH’S THEORY
Zener [ref. by Chell, 40] suggested that the local stress concentration produced at the
head of a dislocation pile – up could lead to cleavage fracture when the leading
dislocations were squeezed together to generate a stress concentration leading to a
crack nucleus.
The model shows that the crack nucleation of length 2c (see Figure
2.9) occurs when the shear stress τs created by a pile – up of n dislocations of
Burger’s vector b each at the grain boundary, reaches the value of:
 2γ s 

 nb 
τs ≈ τi + 
Equation 2.1
where τi is the lattice friction stress in the slip plane and γs is the surface energy of the
exposed crack surface.
Figure 2.9. Zener’s model for cleavage fracture.
P a g e | 17
Stroh [36] included the effect of the grain size d in a model, suggesting the condition for
the shear stress created by a dislocation pile – up of length
d/2
to nucleate a
microcrack as follows (Figure 2.10):
τe = τ y − τi ≥
Eπγ
4(1 − ν 2 )d
Equation 2.2
where τe = effective shear stress, τy = the yield stress, τi = lattice friction stress, ν is
Poisson’s ratio and
E
is Young’s elastic modulus.
This model indicates that the
fracture of the material should depend primarily on the shear stress acting on the slip
band but is also grain size dependent through a d-1/2 relationship .
Figure 2.10. Stroh’s model for cleavage fracture.
2.5.2 COTTRELL’S THEORY
Cottrell [38] proposed that a dislocation mechanism for the cleavage fracture process
should be controlled by the critical crack growth stage under the applied tensile stress.
This model showed that the crack’s nucleation stress can be small if the microcrack is
initiated by the intersection of two low energy slip dislocations to provide a preferable
cleavage plane, Figure 2.11.
[ ]
a
a
111 + [111] → a[001]
2
2
Equation 2.3
This results in a wedge cleavage crack on the (001) plane which is the usual cleavage
plane in ferritic materials. Further propagation of the crack is then controlled by the
applied tensile stress. As the dislocation reaction in Equation 2.3 is accompanied by a
P a g e | 18
decrease in dislocation energy, crack nucleation will be easier than if it occurs by the
Zener/Stroh mechanism.
Figure 2.11. Cottrell’s model for cleavage fracture.
Cleavage fracture will now be propagation controlled with the cleavage fracture stress,
σf given by:
1
σf ≥
2γG m − 2
d
k ys
Equation 2.4
where d is the grain size and k ys is the Hall – Petch constant for shear. The value of
γ calculated was about 20 Jm-2, i.e. about an order of magnitude greater than the
surface energy of the lattice. Cottrell attributed this large value to extra work done in
producing river lines or transversing grain boundaries [41]. The critical conditions for
crack nucleation at the yield stress (σf = σy) can be expressed by the Cottrell equation
[38] as:
1
2
1
σ y k ys d 2 = k ys +σ i k ys d 2 ≥ C 1G m γ
Equation 2.5
where σy is yield stress, σi is friction stress, Gm is shear modulus, γ
is effective
surface energy of an implied crack, C1 is constant related to the stress state (∼4/3 for a
notched specimen and 4 for a plain or unnotched specimen). Cottrell’s model therefore
emphasises the role of tensile stress and explains effects of grain size and yielding
parameters on fracture. Hardening, other than by decreasing grain size, is predicted to
promote brittle fracture by raising the value of tensile stress at the yield point.
Equation 2.5 expresses the conditions for plastically induced crack nucleation at a given
temperature. Any factor that increases σi, k ys , or d increases the tendency for brittle
P a g e | 19
fracture. In ferritic steel, a marked increase in σi with decreasing temperature is usually
experienced. The parameter k ys may depend upon alloy content, test temperature or
heat treatment. Generally, k ys will increase with the stacking fault energy, hence it will
be sensitive to alloying. It is obvious from Equation 2.5 that grain size has a direct effect
on the transition temperature, and thus, the expression for DBTT at which the fracture
stress (σf ) and yield strength (σy) are equal for only one grain size d* could be
derived. Thus
DBTT =
1
*
2 
1 
lnBk y d 2 /(C1Gm γ − k ys )
β 

Equation 2.6
For a given material, the relation between transition temperature and grain size may,
therefore, be reduced to:
DBTT = D + 1 / β ln d
*
1
2
Equation 2.7
where D is a constant. Thus it can be predicted from Equation 2.7 that the temperature
at which a DBTT behaviour occurs, decreases with smaller grain size. Plumtree and
Gullberg [21] have shown that the DBTT increased linearly with grain size and that the
DBTT of lower purity alloys tended to be less affected by grain size changes.
Sometimes the presence of precipitates at grain boundaries camouflages the effect of
grain size on the impact toughness of the ferritic stainless steel. The heat treatment that
accelerates the precipitation of the carbide and nitride, Laves or sigma phases
decreases the resistance to crack initiation significantly during dynamic loading. The V–
notched samples embrittled by these precipitates were found to show a toughness
behaviour similar to that of the solution treated and sharply notched specimens [42].
Therefore, the precipitates assist greatly in the initiation of brittle cracks at the time of
dynamic loading [20]. The DBTT increases with the volume fraction of grain boundary
precipitates, thereby facilitating fracture by decreasing the surface energy for fracture
(γ). Such grain boundary precipitates can act as starting or initiation points for fracture
causing a marked increase in the transition temperature.
Brittle failures are typical transgranular cleavage fractures which occur in the body –
centred cubic (BCC) metal at low temperatures and high strain rates. Plumtree and
Gullberg [21] proposed the following model which applies only to the initiation of
P a g e | 20
cleavage cracks. Crack nucleation occurs when the concentrated stresses at the tip of a
blocked dislocation band equal the cohesive stress and is given by:
(τN - τi) nb = 2γT
Equation 2.8
Where τN = shear stress for crack nucleation, τi = friction shear stress, n = number of
dislocations in the pileup, b = burgers vector, and γT = true surface energy.
Equation 2.8 indicates that a crack will form when the work done by the applied shear
stress (τN nb) in producing a displacement nb equals the combined work done in
moving the dislocations against the friction stress (τi nb) and the work done in creating
the new fracture surface (2γT). In most metals where some relaxation occurs around
the blocked dislocation band, the term γ is used rather than γT as grain boundaries,
hard particles and of course, grain boundary particles act as barriers to dislocation
motion, favouring crack initiation.
When the second phases are inhomogeneously
distributed the value of γ is reduced [43]. Subsequently the amount of work done in
crack nucleation is reduced. Transgranular cleavage cracks form more easily and the
toughness is reduced, compared with those alloys where γ remains high due to the
second phase distribution.
2.5.3 SMITH’S THEORY
This is an alternative model that provides the starting point for growth – controlled
cleavage fracture, to incorporate the effect of brittle second phase particles on grain
boundaries, see Figure 2.12.
Here, a brittle particle of thickness
Co
at the grain
boundary dividing adjacent grains, is subjected to the concentrated stress ahead of a
dislocation pile – up of length D. A microcrack is initiated when a sufficiently high
applied stress causes local plastic strain within the ferrite grains to nucleate a
microcrack in the brittle grain boundary particle of thickness
Co.
Applying Stroh’s
analysis, the particle will crack under the influence of the resulting dislocation pile – up if:
1
 4Eγ e  2
τe ≥ 
2 
πd 1 − ν 
(
)
Equation 2.9
where γe is the effective surface energy of the particle. Similarly, nucleation controlled
cleavage of the ferrite will occur (at the yield point) if:
P a g e | 21
1
 4Eγ f  2
τe ≥ 
2 
πd 1 − ν 
(
Equation 2.10
)
where γf is the effective surface energy of ferrite. If, however, τe lies between these
two limits and γf is greater than γe, then propagation controlled cleavage fracture is
predicted with the particle’s microcrack propagating into the ferrite under the combined
influence of the dislocation pile – up and the applied stress. By examining the change in
energy with crack length, the cleavage fracture stress, σf, is given by the following
expression leading to an inequality expression predicting growth or propagation of the
initiated crack:
2
 4  Co 1/ 2 τ i 
4Eγ f
 Co 
σ   + τ e2 1 +  
 ≥
πd 1 − ν 2
d 
 π  d  τ e 
2
f
(
)
Equation 2.11
Figure 2.12. Smith’s model for cleavage fracture.
In the absence of a contribution by the dislocations (the second term in Equation 2.11)
the inequality expression is reduced to:
1
 4E γ f
2
σf > 

2
 π (1 − ν )C o 
Equation 2.12
This model, therefore, emphases the importance of the precipitate’s thickness C0 and
indicates clearly that the coarser particles give rise to lower fracture stresses.
If,
however, the effective shear stress τe is written as k ys d −1 / 2 , the equation predicts that σf
is independent of grain size, other factors being equal. Indirectly, however, the grain
size will determine the value of C0 for a given volume fraction of the grain boundary
phase and, therefore, does exert an indirect influence on the fracture strength. In
P a g e | 22
practice, therefore, a fine grain size with a large grain boundary area per unit volume, is
associated with thin precipitates, and values of σf are usually expected to be high with
smaller grain sizes.
2.5.4 CLEAVAGE FRACTURE RESISTANCE
Analyses of the factors controlling the cleavage fracture stress have been made
previously by different authors [44]. Although the fracture process in the steel does not,
in general, correspond to the condition for which these expressions were derived (that is,
fracture is frequently initiated at the tip of a stopped microcrack or by the joining of
several microcracks by tearing), directly analogous expressions should exist for these
conditions. From Equation 2.13, it should be noted that the fracture stress depends on
the grain size, as well as yield strength, and that refining the grain size increases both of
them.
σ f = (8G m γ m / k y )d
−
1
2
(clean material)
Equation 2.13
(less pure material)
Equation 2.14
1
 2Eγ b  2 − 21
σf = 
d
2 
 πα 1 − ν 
(
)
where Gm and E are the shear and Young’s moduli, respectively, γm and γb are the
appropriate surface energy terms, ν is Poisson’s ratio and α is a constant.
The most probable important environmental factor that affects the failure of the materials
is the service temperature. Although many different criteria exist for the conditions of a
DBTT, they all effectively point to the temperature at which
σf = σy. It has been
observed that grain refinement increases both of these stresses.
In the case of solid solution and precipitation strengthening, the ductile to brittle
transition temperature is usually raised.
This results from these strengthening
mechanisms that do not increase the fracture stress, as grain refinement does.
2.6
THERMOMECHANICAL PROCESSING
In the previous section it was mentioned that the toughness of ferritic stainless steel
depends on the grain size, and this grain size can be refined by thermomechanical
processing.
However, a thermomechanical process also affects the recrystallisation
parameters and precipitation of carbides, nitrides, sigma and Laves phases [18,45]. The
P a g e | 23
thermomechanical processes that affect the changes in microstructure of the ferritic
stainless steel will be considered in this section.
2.6.1 COLD-ROLLING
Cold rolling was found to raise the DBTT, but the effect is not consistent nor is it very
great [18].
Differences in cold rolling temperature and the extent of preferred grain
orientation make the effects more difficult to quantify. Sawatani et al [8] have observed
the effects of a reduction by cold rolling on the elongation and the amount of Laves
phase precipitates present in a Ti and Nb stabilised low C, N-19%Cr-2%Mo alloy. Their
results show that at a 20% cold reduction a very large amount of Laves phase is
precipitated that increases the strength and decreases the elongation of the steel. Cold
rolling is considered to give a high enough dislocation density to nucleate these Laves
particles at dislocations and not only at grain boundaries. The optimum properties of
cold rolled and annealed steel of this composition are obtained by cold reduction of more
than 80%, followed by annealing at 920 °C, which will dissolve the Laves phase,
followed by rapid cooling.
2.6.2 HOT-ROLLING
Ferritic stainless steel can be hot-rolled without difficulty. However, grain refinement
during the hot-rolling process does not occur readily and the grain structure can be quite
coarse, particularly after relatively high finishing temperatures.
Hot-rolling may also
accelerate the precipitation of the Laves phase. It is thus imperative not to hot work in
the temperature range of the stable Laves phase precipitation in these steels. After hot
rolling, the steel is usually rapidly cooled from the temperature range of 900 to 950 °C to
prevent the formation of the Laves phase during cooling. Care must also be exercised
when working with high interstitial content alloys, since hot-rolling at high temperatures
and quenching can lead to high temperature embrittlement [46].
2.6.3 COOLING RATE
The cooling rate affects the intensity of the precipitate’s formation by altering its
nucleation rate. Fast cooling rates can prevent precipitation; intermediate cooling rates
cause maximum age-hardening, while slow cooling rates give over-ageing which
produces low strength. If the precipitation has been suppressed during cooling, it can be
induced during the ageing process [47,48].
P a g e | 24
2.6.4 HEAT TREATMENT
Heat treatment is a major factor in controlling the properties of stainless steels. The
ferritic steels are annealed at temperatures from 750 to 900 °C, and the upper limit
should not be exceeded substantially since grain growth will induce a decrease in
toughness [49]. However, by keeping (C + N) levels very low, concern about grain –
size related DBTT increases can be prevented [18]. Annealing must be followed by
rapid cooling to avoid prolonged exposure to the temperatures at which Laves phase will
form or “475 °C” brittleness will develop.
With total (C + N) content below approximately 500 ppm, quenching generally produces
optimum toughness by preventing sigma or Laves phase precipitation as well as carbide
and nitride precipitation. With higher (C + N) levels, nothing is gained by quenching and
if a high temperature anneal is used, rapid quenching can cause embrittlement by
severe carbide and nitride precipitation.
This rapid cooling embrittlement can be
reduced substantially by the addition of carbide and nitride stabilisers, such as Nb and
Ti.
2.7
APPLICATIONS OF STAINLESS STEELS IN AUTOMOBILE EXHAUST SYSTEM
Recently, environmental pressures have required the reduction of dangerous exhaust
gas emissions from motor cars, and there has been an effort by automobile
manufacturers to raise the temperature of exhaust gases from automobile engines to
about 900°C in order to give a cleaner exhaust gas [50, 51]. To improve the efficiency of
the automobile engines and to reduce their weight, more conventional stainless steel
sheets and tubes in the application of exhaust systems are now replacing the traditional
cast iron, particularly in near-engine applications. The exhaust system is divided into
two main parts, the hot and the cold ends. The components of the hot end consist of the
exhaust manifold, front pipe, flexible pipe and catalytic converter and the components of
the cold end consist of a centre pipe main muffler and tail end pipe, see Figure 2.13.
Typical operating temperatures and the current materials used for these different
components are shown in Table 2.1.
P a g e | 25
Figure 2.13. Automobile exhaust system components.
Table 2.1. Component of automobile exhaust system and their typical operating
temperatures [52,53,54].
Component
Service temp
(°C)
Exhaust manifold
Required properties
Current materials
950 – 750
•
•
•
•
High temperature strength
Thermal fatigue life
Oxidation resistance
Workability
409, 441, 304, 321,
309
304, 321,309, 409
304, 321, 309 and
316Ti
800 – 600
•
•
High temperature strength
High temperature salt damage
resistance
Workability
Oxidation resistance
Thermal shock resistance
Salt damage resistance
Corrosion resistance at inner
surface (condensate)
Corrosion resistance at outer
surface (salt damage)
Front pipe
Flexible pipe
Catalytic
converter
Shell
Catalyst
carrier∗
Centre pipe
1000 – 1200
600 – 400
Main muffler
400 - 100
Tail end pipe
•
•
•
•
•
•
441, 409, 321, 309
304, 409, 441
409, 434, 436,
430Ti, 321, 304
304, 316
Thus, the materials for automobile exhaust systems require excellent thermal fatigue
resistance and higher temperature strength. Ferritic stainless steels are the commonly
used materials in exhaust systems as a compromise between inexpensive carbon steels
and the higher cost of higher alloyed ferritic or austenitic steels. Ferritic stainless steels
have better corrosion resistance and thus a longer life than low carbon steels and have
lower cost than the more highly alloyed stainless steels.
However, ferritic stainless
steels have relatively low strength at elevated temperatures compared to austenitic
steels. Therefore, efforts have been made to create new ferritic stainless steels with
∗
The catalyst carrier are usually made out of ceramic, there has been a recent development in metal carrier made out of ferritic
stainless steel foils because they have good thermal shock properties and a small heat capacity [53].
P a g e | 26
excellent thermal fatigue resistance and high strengths at elevated temperatures. This
has been achieved by the use of stabilising elements such as the addition of niobium
and titanium, as will be discussed in the next section.
2.8
HEAT RESISTANT FERRITIC STAINLESS STEELS
Niobium-containing ferritic stainless steels are being used in automotive exhaust
systems because of their excellent heat resistant properties, especially their thermal
fatigue resistance which is very important for materials of exhaust systems because of
the frequent heating and cooling cycles.
When in solid solution, the Nb addition
increases the initial high temperature strength compared with other alloying elements,
see Figure 2.14 below [4].
However, Nb forms several types of precipitates during
service, which can cause degradation in high temperature strength and thermal fatigue
resistance [3].
In the work by Fujita et al. [5], the effect of Nb additions in 0.01C – 0.01N – 13Cr steel
on the 0.2% proof strength at 900 °C is observed to be effective above 0.2 %wt Nb. On
the other hand, Mo additions increase the high temperature strength approximately in
proportion to the alloy addition content.
Figure 2.14. Effect of alloying element additions on the 0.2% proof strength at 900 °C of a
13%Cr ferritic steel [4].
2.9
STABILISATION
To combat the problem of sensitisation, the carbon in stainless steels is reduced through
a combination of steelmaking practices and through the addition of elements with strong
P a g e | 27
affinity for carbon and nitrogen. Firstly, these steels are processed by an “argon-oxygen
decarburisation (AOD)” or “vacuum-oxygen decarburisation (VOD)” process to reduce
the carbon content to very low levels (<0.02%C). The carbon and nitrogen in solution
are further reduced through the addition of stabilising elements such as titanium (Ti) and
niobium (Nb), tantalum (Ta), vanadium (V) and molybdenum (Mo). These elements
form carbides, nitrides or carbo-nitrides and precipitate at higher temperatures and
shorter times than chromium carbides, thus removing most of the carbon and nitrogen
from solution before the chromium carbides form. Because of the high cost of Ta, it is
normally not used in the stabilisation of stainless steels. Either Nb or Ti individually
and/or Ti plus Nb additions have been used in stainless steels for automobile exhaust
systems, and the two stabilisation methods will be compared in the next section.
2.9.1 STABILISATION WITH TITANIUM
This is the most widely used stabilising addition to stainless steels. The highly reactive
element forms highly stable titanium nitride (TiN) precipitates in the presence of N and
titanium carbide (TiC) in the presence of C. Whilst in the presence of both nitrogen and
carbon, titanium carbonitride (Ti(C,N)) is formed owing to the mutual solubility of TiN and
TiC [55]. The ratio of TiC/TiN in the Ti(C,N) is dependent upon the ratio of N/C in the
alloy. In Ti stabilisation practice it is generally accepted that the level of Ti required to
fully-stabilise a stainless steel is directly dependent on the C and N content, and can be
calculated empirically from [55]:
Ti=0.2 + 4(C+N)
The disadvantage of Ti stabilisation is that Ti-stabilised stainless steel suffers from
surface defects that require surface grinding and these result in an increase in the
overall production cost. This can be overcome by combining Ti additions with other
stabilising elements such as aluminium (Al), niobium (Nb) or vanadium (V).
2.9.2 STABILISATION WITH NIOBIUM
Many researchers have investigated the effects of niobium stabilisation on ferritic
stainless steels for their use in automotive exhaust systems [3,4,5,6] and have found
that niobium additions can improve the high temperature strength of ferritic stainless
steels by solid solution strengthening, which allows its operation at such high
temperatures. Generally, solute Nb readily precipitates out as a carbo-nitride when the
steel is used at high temperatures of about 900°C for long times [3,4]. In addition to its
P a g e | 28
stabilisation effects, small amounts of Nb in the stainless steel have been known to
affect both mechanical and microstructural properties [6]. It is, therefore, important to
know how much of the niobium precipitates out as carbides or as carbo-nitrides and
which fraction remains in solution. Niobium stabilised ferritic stainless steels form the
carbide precipitates NbC, M6C (Fe3Nb3C) and the Laves phase type Fe2M (Fe2Nb). The
main benefit of the Nb is its ability to suppress recovery, recrystallisation and grain
growth in ferritic steels [4].
2.9.3 SOLID SOLUTION HARDENING AND SOLUTE DRAG BY NIOBIUM
Niobium addition is the most frequently used microalloying element because it has a
significant effect on the microstructure and the mechanical properties of ferritic stainless
steels. Depending on the state of Nb (in solid solution or in precipitates) as determined
by the heat treatment, Nb has a significant effect on the recrystallisation and the grain
growth. The effect of Nb on the recrystallisation and the grain growth of the austenite
has been widely studied [56,57,58]. Nb addition, even in small amounts, can lead to a
significant decrease in grain boundary mobility, as well as an increase in the
recrystallisation temperature.
Both the decrease in the grain boundary mobility and
increase in the recrystallisation temperature might be caused by a solute drag effect of
Nb in solid solution and the pinning effect of fine precipitates such as NbC and Fe2Nb.
Suehiro [57] has shown that Nb retards the migration of grain boundaries during
recrystallisation due to the solute drag effect.
The quantitative theory of the solute drag effect on a moving grain boundary during
recrystallisation was originally formulated by Lücke and Detert [59]. It was later modified
by Cahn [60] and then by Lücke and Stüwe [61]. Since then, this theory has been
further refined by several authors [57,58,62]. The equation proposed by Cahn on the
rate of grain boundary movement as it is affected by solute drag is [63,64]:
pd
νb
=
1
1
αX s
=
+
MT M 0 1 + β 2ν b2
(
)
Equation 2.15
where pd is the driving force for the grain boundary mobility and νb is the mobility rate,
MT is the overall mobility due to intrinsic plus solute drag, M0 is the intrinsic grain
boundary mobility in pure material, Xs the atom fraction of solute in the bulk metal, α is
a term related to the binding energy of solute to the grain boundary and β is a term
related to the diffusivity of the solute in the vicinity of the grain boundary.
P a g e | 29
In the work by Le Gall and Jonas [63], the authors have determined the effect of solute
concentration and temperature on the grain boundary mobility against driving force using
the Cahn model. Their results show that; at high concentrations and low temperatures,
the mobility of the grain boundary is low and if the driving force or the temperature is
increased, the increase in mobility is very significant and tends towards that of pure
metals. The overall results are described by curves with approximately two slopes; M0
representing the intrinsic mobility in the pure metal and MB which refers to the mobility
in the presence of solute drag elements.
Hillert and Sundman [65,66] developed a theory that is most general because the
validity of the theory is not limited to the dilute solution case and it can be applied to both
grain boundary and phase interface migrations.
Suehiro et al.[58]
developed the
simplified model to calculate the solute drag effect on a moving phase interface, but it
can also be applied to both grain boundaries and phase interfaces. They applied this
model to the phase transformation in an Fe-Nb system [57].
In this model,
recrystallisation is described as a phase transformation for which the driving force
results from the energy stored as dislocations. It should also be mentioned that the
stored energy in a plastically deformed material may give rise to the driving force for
grain boundary migration in recrystallisation through the Strain Induced Boundary
Migration (SIBM) mechanism.
It was shown that the Cahn equation (that is, Equation 2.15) can be simplified when the
driving forces are low [63]. The overall mobility due to intrinsic plus solute drag is then
given by:
1
1
1
=
+
MT M0 MB
Equation 2.16
where M0 is the intrinsic mobility in a pure metal and MB is the grain boundary mobility
under solute drag conditions. The solute drag equation is then reduced to:
ν b = MT pd
Equation 2.17
This equation is only valid for low driving force phenomena such as grain growth,
whereas with higher driving forces as with the recrystallisation, Le Gall and Jonas [63]
proposed a “law of mixtures”:
MT = M0 (1 − fGB ) + M BfGB
Equation 2.18
P a g e | 30
where M0 and MB are defined as above, fGB is a function of potential grain boundary
sites filled by solute and is temperature dependent.
2.9.4 EFFECTS OF TEMPERATURE ON SOLUTE DRAG
The effect of temperature on solute drag during recrystallisation has been investigated in
an Fe-Nb alloy [57].
These results show that Nb decreases the velocity of the
recrystallising grain boundary at the composition covering the critical composition and
this depends on the temperature and total driving force. This retardation was found to
be caused by the solute drag effect of Nb. In work by Suehiro et al. [58], the author
studied the effect of Nb on the austenite to ferrite transformation in ultra low carbon
steel.
Their results indicate that there is a critical temperature where the rate of
transformation changes drastically. The transformation that occurs above and below the
critical temperature are both partitionless massive transformations.
The critical
temperature was found to be composition dependent, and for the 0.25% Nb alloy it was
found to be 760 °C and for 0.75%Nb alloy it was 720 °C
From Equation 2.15, however, it can be predicted that by increasing the temperature to
the point where the mobility of solute atoms becomes high, solute drag becomes less
effective and the overall mobility parameter approaches the intrinsic mobility factor, that
is MT ≈ M0. . Le Gall and Jonas [63] have observed solute drag by sulphur atoms in
pure nickel, while observing that the transition from solute drag to purely intrinsic mobility
of the grain boundary is not a gradual one but that it occurs at a critical temperature that
provides a “break” in the mobility versus inverse temperature relation.
2.9.5 DUAL STABILISATION WITH TITANIUM AND NIOBIUM
Research conducted over the last few decades has revealed the benefits of stabilising
carbon and nitrogen by using both titanium and niobium. These observations led to the
development of the modern generation of ferritic stainless steels, i.e., dual-stabilised
ferritic stainless steel. The dual stabilisation imparts beneficial corrosion resistance,
oxidation resistance, high temperature strength and formability to the steel [2,31]. At
present, it appears to be well accepted that to achieve full stabilisation of ferritic
stainless steels a ratio of Nb:Ti = 2:1 is preferred [31]. The optimum limits of dual
stabilisation have been established experimentally, and are given by:
Ti + Nb = 0.2 + (C + N)
at min
Ti + Nb = 0.8
at max
P a g e | 31
To reduce grain growth during high temperature annealing extra niobium is added,
leading to intermetallic precipitation. The addition of titanium and niobium is made in
such a way that an extra content of niobium (∆Nb) is kept in solid solution after the
carbo-nitride precipitation [67]:
∆Nb = [Nb] – 7([C] + [N])
in case of Nb stabilisation
∆Nb = [Nb] – 7/2 [C]
in case of (Ti + Nb) stabilisation
At temperatures between 600 and 950 °C, isothermal treatment has shown that the
residual niobium precipitates out to form intermetallic Fe2Nb. Schmitt [67] has observed
that depending on the temperature, a part of the precipitation occurs on the grain
boundaries combined with a fine intergranular precipitation, and this strongly slows down
the grain boundary mobility.
The nature of the precipitates, the size and location
depends also on the alloy composition.
The value of the excess ∆Nb content is
significant, as well as the silicon and molybdenum content. Both the Si and Mo have an
effect to enhance the precipitation of the Laves phase [29].
2.10 AISI TYPE 441 STAINLESS STEELS
Type 441 alloy is a low carbon, Ti and Nb stabilised, heat resisting ferritic stainless steel
providing good oxidation and corrosion resistance for applications such as automotive
exhaust system components. The high chromium content makes type 441 steel far
more corrosion resistant than its counterpart type 409 [52].
The 441 steel is dual
stabilised with niobium and titanium to provide good weld ductility and resistance to
intergranular corrosion in the weld’s heat affected zone. Type 441 is manufactured
according to the requirements of EN 10088-2 and certified as 1.4509, see Table 2.2 for
its chemical composition.
The structure of type 441 is a completely ferritic alloy, i.e. the matrix has a body centred
cubic crystal structure up to its liquidus temperature. Angular carbo-nitrides of titanium
and niobium that have precipitated from the melt are randomly dispersed throughout the
structure. The presence of the titanium nitrides and carbides in the steel tends to lower
the melting point of the steel, as shown by Gordon [55], see Figure 2.15.
Excess
niobium is taken into solid solution during high temperature annealing and precipitates
as very fine particles of Laves phase (Fe2Nb) upon either slow cooling or upon holding
P a g e | 32
at intermediate temperatures of 600 – 950 °C [68]. Strengthening by this dispersion is
responsible for improved elevated temperature strength [69].
Table 2.2. Chemical composition of type 441 stainless steel in accordance with EN 100882 [70].
C
N
Mn
Si
Min
Max
Cr
Ni
17.50
0.03
0.045
1.00
1.00
18.50
Ti
Nb
S
P
1.00
0.015
0.04
0.10
0.50
0.60
Remarks
%Nb≥ 3xC + 0.3
Figure 2.15. The solvus temperatures of the precipitates found in stabilised ferritic
stainless steels [55].
2.11 CALPHAD METHODS
The CALPHAD (CALculations of PHAse Diagram) method primarily uses numerical
techniques based on thermodynamic principles for addressing complex problems like
stable and metastable phase equilibria in the multicomponent systems for equilibrium
conditions.
The essence of the Calphad approach is to obtain the parameters of
thermodynamic models for the Gibbs energies of the constituent phases in terms of
known thermodynamic and phase equilibrium data in the lower order systems, binaries
and ternaries. The Gibbs energies of multicomponent alloy phases can be obtained
from those of lower order systems via an extrapolation method [71].
These Gibbs
energy values enable engineers and scientists to calculate reliable multicomponent
phase diagrams in many instances.
Experimental work is then only required for
P a g e | 33
confirmatory purposes and not for the determination of the entire phase diagram. The
Calphad approach in calculating phase diagrams of multicomponent alloy systems is
shown schematically in Figure 2.16 [71].
Figure 2.16. Schematic flow diagram showing the Calphad approach used to obtain a
thermodynamic description of a multicomponent system.
Thermodynamic descriptions of the constituent lower order systems, normally binaries
and ternaries, are obtained through experimental and “first principles” of total free
energy information and phase equilibrium data [72]. However, once descriptions for the
lower order systems are known, it is possible in many cases to obtain thermodynamic
descriptions of the higher order systems by using an extrapolation method so that more
complete phase diagrams of the system can be calculated.
The Gibbs energy of a phase is described by a model that contains a relatively small
number of experimentally optimised variable coefficients.
Examples of experimental
information used include melting and other transformation temperatures, solubilities, as
well as thermodynamic properties such as heat capacities, enthalpies of formation and
chemical potentials.
For pure elements and stoichiometric compounds, the following model is commonly
used:
G m − H mSER = a + b ⋅ T + c ⋅ T ⋅ ln(T) +
∑ d i ⋅T i
Equation 2.19
where, Gm − H mSER is the Gibbs energy relative to a standard element reference state
(SER), HmSER
is the enthalpy of the element in its stable state at the temperature of
P a g e | 34
298.15 Kelvin and a pressure of 105 Pascal (1 bar), and a, b, c, and di are model
parameters.
For multi-component solution phases, the following expression for the Gibbs energy is
used:
G = G°+ id Gmix + xs Gmix
Equation 2.20
where G° is the Gibbs free energy due to the mechanical mixing of the constituents of
the phase,
id
Gmix is the ideal mixing contribution, and
xs
Gmix is the excess Gibbs energy
of mix (the non-ideal mixing contribution).
If a phase in a multi-component solution is described with a single sub-lattice model,
then the G°,
xs
Gmix and
xs
Gmix contributions to the Gibbs energy can be expressed as
follows:
Equation 2.21
G ° = ∑ ic i ⋅ G oj
id
Gmix = R ⋅ T ⋅ ∑ic i ⋅ ln( c i )
Equation 2.22
G mix = ∑ i∑ j > i c i ⋅ c j ⋅ ∑ k L ki , j ⋅ ( c i − c j ) k
Equation 2.23
xs
where ci and cj are the mole fraction of species i and j respectively, and L ki , j is a
binary interaction parameter between species
i
and
j.
The binary interaction
parameter L ki , j is dependent on the value of k. When the value of k equals zero or
one, the equation for
xs
Gmix becomes regular or sub-regular, respectively.
2.11.1 THERMODYNAMIC SOFTWARES
During the last few decades, the rapid development of thermodynamic and kinetic
software packages, such as Thermo-Calc®, DICTRA [73,74], ChemSage, FactSage and
MTDATA have made it possible to calculate complex phase equilibria in multicomponent
systems. It is evident that gradually the accuracy of these calculations will become more
and more dependent on the databases behind the packages. So far, a great number of
binary, ternary and even quaternary systems have been assessed to obtain a sufficiently
sound database for multicomponent steels [75,76,77]. However, there are still some
subsystems not assessed earlier or assessed but not shown in the open literature.
P a g e | 35
2.12 INTERMETALLIC LAVES PHASE
2.12.1 CRYSTALLOGRAPHIC STRUCTURE
Many intermetallic phases belong to the group of the Laves phase with AB2
compositions of topologically close packed (TCP) structures. Furthermore, these Laves
phases are capable of dissolving considerable amounts of ternary alloying elements
[78,79,80]. Laves phases are generally stabilised by the size factor principles, that is,
the atomic size ratio rA/rB, which is ideally 1.225 with a range of 1.05 - 1.68 that is
usually observed [81,82]. Three types of Laves phase (see Figure 2.17 [85]) and the
Pearson symbols and their crystallographic structures are given below [83,84,85,86]:
Table 2.3. Crystallographic structure and the space groups of the three types of the
Laves phase.
Typical
composition
C14, MgZn2
C15, MgCu2
C36, MgNi2
Structure
Pearson symbol
Space group
hexagonal
fcc
hexagonal
hP12
cF24
hP24
P63/mmc
Fd 3 m
P63/mmc
Space group
number
194
227
194
Figure 2.17. The three polytypes of the Laves phase structure in a hexagonal setting.
The Fe2Nb Laves phase (Mg2Zn type, C14) is a hexagonal close packed phase with the
space group P63/mmc, and the lattice parameters of a=0.473 nm and c=0.772 nm with
c/a = 1.633. The main factor determining its formation is the relative atomic size of the
constituent atoms, with the ranges of composition quite small.
This Laves phase
probably does not form at its exact equilibrium composition [8,87]. In Nb and/or Ti
stabilised stainless steel grades, Fe2Nb or Fe2Ti (more seldom) can form.
P a g e | 36
2.12.2 OCCURRENCE
The precipitation rate of the Laves phase in a Ti and Nb stabilised steel was observed to
reach a maximum at 700°C and its dissolution occurs at temperatures over 900°C [88].
The same observation has been made by the use of thermodynamic software such as
Thermo-Calc® [3] and its growth and coarsening rates have been simulated by Bjärbo
[9] using DICTRA software. The Laves phase precipitates firstly on sub- and grain
boundaries as a fine precipitate and as the steel is slowly cooled from a high
temperature of about 900°C, the amount of Laves phase increases but now inside the
grains and the particles then coarsen. In the work by Murata et al, [89] they have shown
that the Laves phase’s solvus temperature depends on the content of the alloying
elements, that is, the solvus temperature increases with increasing alloying content.
The intermetallic Laves phase is known to affect the mechanical properties and
corrosion resistance of ferritic stainless steel. It has been found that a fine precipitate of
Laves phase at grain boundaries improves the high temperature strength when still fine
[6]. However, rapid coarsening of the Laves phase at high temperatures reduced the
high temperature strength [10] although the exact mechanism is still not clear and still
requires clarification. In research done by Fujita et al.[5] to determine the solubility
products in a niobium stabilised ferritic steel it was found that a Mo addition enhances
the precipitation of the Laves phase even if the addition is as small as 0.5mass%Mo.
Sawatani et al [11], studied the effect of Laves phase on the properties of dual stabilised
low carbon stainless steels as related to the manufacturing process of Ti- and Nbstabilised low (C, N)-19%Cr-2%Mo stainless steel sheets, and have found that the
Laves phase has a significant influence on the mechanical properties of the steel. It was
found that Laves phase on the grain boundaries shifts the brittle to ductile transition
temperature to a higher temperature, and large amounts of Laves phase degrade the
room temperature ductility of cold rolled and annealed sheet and greatly enhances its
strength. It was also observed that after a 20% cold rolled reduction (of the sheet that
was cold rolled and annealed after 0 to 92% reductions) that there was a peculiarly
rapid precipitation of Laves phase, which caused a severe degradation of the
mechanical properties [8].
The reason for the rapid increase in Laves phase
precipitation was not clear, but it was assumed that there is possibility that an
autocatalytic reaction occurs to substantially increase the precipitation rate of Laves
phase. The rate of autocatalytic reaction under which the nucleation and growth of
P a g e | 37
particles occur at dislocations will greatly depend on the formation rate of new
dislocations from the cold work which then act as new precipitation sites.
2.12.3 ORIENTATION RELATIONSHIP
Cocks and Borland [90] have investigated the orientation relationship and morphology of
Fe2Nb precipitates within the ferrite matrix in 0.6 at%Nb and 1.9 at%Nb alloys
respectively, using electron diffraction techniques. They had found that there exists a
single orientation relationship:
{112 0 }
Fe 2 Nb
// {111 }α : 0001
Fe 2 Nb
// 112
α
Equation 2.24
and the morphology of these particles are disc-shaped lying on the {111}α plane. Lathshaped particles were found to have developed in overaged alloys and most of these
particles tend to be elongated in the 〈112〉 matrix direction. The orientation relationship
between the rod-like Laves phase particles and the matrix was found to be [11,91]:
{112 0 }
Fe 2 Nb
// {111 }α
: 0001
Fe 2 Nb
// 110
α
Equation 2.25
The analysis suggests that the habit plane, if any, must be {110}α and the preferred
growth direction must be 〈110〉 α [92]. The two orientation relationships are completely
different from one another and the habit planes are also different. In work by Miyahara
et al. [93] on Fe-10%Cr ferritic alloys, they have observed very small disk–like Laves
phase (Fe2Mo) precipitates that had formed on the {100}α plane and which have a
coherent strain field in the matrix. Therefore, it may be reasonable to consider that there
can be several orientation relationships rather than only one orientation relationship.
None of the above authors tried to relate the orientation relationship of the Laves phase
to its precipitation morphologies.
In the work done by Murata et al. [94], the authors determined the surface interfacial
energy γSF of the coherent and incoherent Laves phases in the Fe-Cr-W- C quaternary
system.
They observed that the coherent fine Laves phase has a lower interfacial
energy than the incoherent granular Laves phase, and their estimated values were 0.1
J/m2 and 0.468 J/m2, respectively. Also, their results show that there is a morphological
change of the Laves phase from the fine coherent precipitates to the granular ones.
This morphological change occurs in a regular ageing sequence in the steel if it contains
more than 4 wt.% W.
P a g e | 38
CHAPTER THREE
THEORY OF PRECIPITATION REACTIONS IN STEELS
3
3.1
INTRODUCTION
The formation and subsequent behaviour of individual particles in any precipitation
process involve its nucleation, growth and coarsening but the overall precipitation
process must account for solute concentration impingement effects. In some cases the
nucleation of second phases in steel might involve the dissolution of metastable
precipitates and the coarsening of stable precipitates. These stages of nucleation and
growth of the precipitates will be discussed in this chapter.
3.2
CLASSICAL THEORY OF NUCLEATION
Most particle-sized second phases in steels precipitate through a nucleation and growth
mechanism.
During the nucleation process, the associated free energy change is
dominated by the Gibbs chemical free energy, the surface free energy and the misfit free
energy terms. The total free energy change can be represented by summing all of these
contributions:
∆G =
4 3
4
πr ∆Gv + 4πr 2γ + πr 3 ∆Gε , (where ∆Gv ≤ 0)
3
3
Equation 3.1
where ∆Gv is the Gibbs chemical free energy released per unit volume of the new
phase and has a negative value,
γ is the interfacial surface energy per unit area
associated with the interface of the two phases and ∆Gε is the misfit strain energy per
unit volume. Plotting this equation demonstrates the dependency of the stability of the
new embryo on its size, see Figure 3.1.
P a g e | 39
1/Z
Figure 3.1 The free energy change associated with the formation of a stable nucleus with
the radius r.
3.2.1 ACTIVATION ENERGY FOR NUCLEATION WITHIN THE MATRIX
From Figure 3.1, the maximum total free energy occurs at a critical radius r* when the
free energy has a value ∆G*, known as the activation energy. An embryo with a radius
larger than r* will tend to grow spontaneously rather than dissolve since its growth
leads to a decrease in free energy and then the embryo becomes a stable nucleus. At
r =r*, d∆G/dr equals zero, so that ∆G* is given by:
r* =
− 2γ 2
(∆Gν + ∆Gε )
∆G * =
Equation 3.2
16πγ 3
3(∆Gν + ∆Gε )
2
Equation 3.3
3.2.2 ACTIVATION ENERGY FOR NUCLEATION ON THE GRAIN BOUNDARY
In the preceding section, the formation of a nucleus has been regarded as a
homogeneous process occurring with equal probability in all parts of the assembly. In
practice, this is unlikely to happen unless the assembly is extremely pure, and also
contains (if in the solid state) very few structural defects. More usually, the presence of
impurity particles and structural defects (dislocations and sub- and grain boundaries)
P a g e | 40
that enable nuclei to be formed with a much smaller free energy of activation than that of
the homogeneous nuclei, occurs in most industrial alloys.
If we consider nucleation that takes place on a grain boundary, a certain surface area A
of the grain boundary is removed from the system. This forms an additional driving force
for nucleation and must, therefore, have a negative sign in the basic free energy
equation for nucleation (see the last term in Equation 3.4 below).
∆G = K1{∆Gν + ∆Gε } + K 2γ ppt Appt + K 3γ ppt / gb Appt / gb − γ gb Agb
Equation 3.4
From this equation it is clear that the system will preferentially select those higher
energy grain boundary sites first where it will gain the most energy as an additional
driving force.
The following relationship between the activation energy for grain boundary nucleation
for the various possible grain boundary sites and the contact angle θ should, therefore,
be expected, Figure 3.2. The ratio of the free energy required to form a grain boundary
nucleus to that needed to form a homogeneous nucleus obviously decreases as the
ratio of the grain boundary energy to the interphase boundary energy increases through
the contact angle parameter cos θ.
From Figure 3.2 it may be seen that at a given contact angle θ, nucleation on grain
corners requires the lowest activation energy, then grain edges and lastly grain
boundaries. Grain boundaries can also lower the retarding force arising from the strain
energy ∆Gε
and, thereby, create a new driving force. In cases where the surface
energy and the strain energy are lowered enough through nucleation on grain
boundaries, the grain boundaries may become the preferred sites fully and no nucleation
within the grains will occur. This is what is typically found in Al - Mg alloys where the βphase nucleates only on grain boundaries and embrittles the alloy [95].
P a g e | 41
Figure 3.2. The ratio of the free energy required to form a nucleus on various types of
grain boundary sites to that required to form a nucleus in the grain matrix, is
plotted as a function of the contact angle parameter cos θ.
Precipitates on grain boundaries represent a special case because they are located on
the junction of two or more different crystal orientations and may also possess different
interface lattice parameters in the case of interphase boundaries. Figure 3.3 shows
different possibilities of interfaces on a boundary.
Figure 3.3. Different possibilities of the precipitate’s interface on grain boundaries.
3.2.3 MISFIT STRAIN ENERGY AROUND THE PARTICLE
The structural misfit across a matrix / precipitate interface plays a prominent role in the
strain energy around a coherent or semi coherent precipitate. This misfit very often
leads to interfacial dislocations or ledges and steps. For incoherent particles the strain
energy arises mainly out of the differences in the lattice spacing and the strain in the
matrix volume because of differences in densities.
Nabarro [96] derived the elastic
strain energy for an ellipsoidal nucleus of β phase having semi-axes a, a and c in an
isotropic α matrix as [108]:
∆Gε =
2
Gmδ 2ν β f (c / a )
3
Equation 3.5
P a g e | 42
where Gm is the shear modulus of the matrix, δ is the volume misfit of the precipitate
in the matrix and is defined as:
δ =
ν β −ν α
=
νβ
Equation 3.6
where νβ and να are the lattice spacings of the precipitate phase and the matrix,
respectively. Thus the elastic strain energy is proportional to the square of the volume
misfit, δ2. The function f(c/a) is a factor that takes into account the shape effect and is
shown in Figure 3.4.
Figure 3.4. Illustration of the variation of the function f(c/a) of an incoherent nucleus with
its shape.
A flat disc, therefore, provides a lower value of ∆Gε but also has a higher surface area
per unit volume and this results in a higher total surface energy.
Therefore, a
compromise has to be reached which very often leads to ellipsoidal shapes.
3.2.4 INTERFACIAL ENERGY
Boundaries between different solid phases can be classified into coherent, semicoherent and incoherent interfaces. For nucleation in crystalline solids the interfacial
energy γ can vary widely from very low values for a coherent interface to high values for
an incoherent interface.
The interfacial energy is a vital parameter for kinetic
simulations. Even small variations of this parameter can have a massive impact on the
nucleation and also strongly influences growth and coarsening of precipitates.
The
P a g e | 43
value of this energy is not often known, since it depends on the crystallography of the
precipitate and matrix, on the chemical composition of involved phases, on their
misorientation, the degree of coherency and finally, any segregation effects of other
minor elements to the interface. Commonly, a mean value of the interfacial energy is
used in numeric simulations, which is denoted as an “effective interfacial energy” γ. It is
of great importance, therefore, to evaluate possibilities to predict these parameters from
existing data (e.g. thermodynamic databases). To treat the problem of multi-component,
multi-phase, multi-particle precipitation kinetics on a more physical basis it is necessary
to predict the interfacial energies of the precipitates depending on the actual system
state.
3.2.4.1 FULLY COHERENT PRECIPITATES
A coherent interface arises when all the lattice planes of the two phases are continuous
across the interface. An interface is said to be fully coherent if each atomic plane in one
crystal that intersects the interface, is matched by another plane on the opposite side of
the interface.
Figure 3.5. Fully coherent precipitates, with no broken inter-atom bonds and with δ=0.
The interface is indicated by the circle.
The interface energy in the case of the two lattice spacings with no difference (δ = 0) is
only of a chemical nature through the bond-energy of atoms A and B, because there is
no structural difference between their lattice spacings. Fully coherent precipitates occur
whenever there is no or only insignificant lattice mismatch, at least in one direction.
P a g e | 44
Especially small precipitates often meet this condition; for example Guinier-Preston1
(GP) zones in many Al-alloys. If there is no lattice mismatch at all these interfaces show
identical, usually very low, interfacial energy values in every direction (see Figure 3.5).
Consequently, these precipitates are usually spheres, for example in the Al-4at% Ag
model alloy [97].
If the lattice parameters of the precipitate and matrix differ, considerable strain energy
can be the result. Because of the different lattice structures of precipitate and matrix,
the lattice will be deformed elastically, depending on the Young’s modulus and also
depending on the orientation of the particle in the lattice. For example, a situation where
a particle has a similar lattice parameter in a horizontal direction but not in a vertical
direction, is illustrated in Figure 3.6. To keep the misfit energy to a minimum these
precipitates arrange themselves in the horizontal direction and develop plate or needle
like shapes. Similar misfits in every direction will also lead to a spherical particle.
Figure 3.6. Coherent precipitate with different lattice parameters only in the vertical
direction. The volume influenced by the lattice misfit, ε is marked by the dotted
line.
3.2.4.2 INCOHERENT PRECIPITATES
When the interfacial plane has a very different atomic configuration in the two adjoining
planes, or even if it is similar but the inter-atomic distances of the two phases differ by
more than 25 %, there will not be good matching across the interface and the interface
will most probably be incoherent [98]. During the evolution of a precipitate, it is very
1
e.g.: Al-Cu system with monatomic, disc shaped, Cu layer in (100) plain. Fully coherent with matrix, typical diameter
of 3-10nm
P a g e | 45
likely that the effective interfacial energy changes depending on the actual size and
composition of the particle. A small precipitate when the ΣA/ΣV of the second phase is
still large, will probably start with fully coherent interfaces and introduce only low elastic
strains, despite a possible lattice misfit. During growth as the ΣA/ΣV decreases, some
interfaces can transform themselves into semi coherent interfaces to reduce the elastic
strain energy. This coherency loss is accompanied by the formation of vacancies and/or
misfit dislocations in the interface.
3.2.4.3 SEMI-COHERENT PRECIPITATES
Semi coherent precipitates share some coherent and some incoherent interfaces with
the matrix, depending on the crystal orientation. The strain associated with a coherent
interface raises the total free energy of the system, and for a sufficiently large atomic
misfit, it would be energetically more favourable to replace the coherent interface with a
semi-coherent interface in which the misfit is periodically taken up by misfit dislocations.
Summarising these findings, the interfacial energy is dependent on crystallographic and
chemical parameters and is not a single scalar value in general. For practical reasons
an effective interfacial energy is often introduced in computer simulations.
3.2.5 NUCLEATION RATE
The probability of the barrier to nucleation being overcome by a potential nucleus is
given by a Boltzman expression, and so the concentration of critical – sized nuclei N*:
 − ∆G * 
N * = N0 exp

 kT 
Equation 3.7
where N0 is the initial number density of nucleation sites per unit volume, and k is the
Boltzman constant. Each critical sized embryo can be made supercritical and become a
nucleus by transferring an atom in contact with the embryo into it.
In the absence of the soft impingement, whereby the mean solute concentration
decreases as the reaction proceeds, the nucleation rate is assumed to occur at a
constant rate, and this is difficult to justify experimentally because of the different
precipitation reactions that occur simultaneously [5,99,100,101,103,104].
Classical
nucleation theory is used to estimate the nucleation rate for each type of precipitate.
There are several formulae for the nucleation rate per unit volume that are based on the
P a g e | 46
Turnbull and Fisher nucleation model [95]. Following this model the nucleation rate can
be written as:

V'
N& =  1 −
V eq

*



N kT exp  − ∆ G + Q 
 o h


RT



Equation 3.8
where Q is the activation energy for diffusion, k and h are the Boltzmann and Planck
constants respectively, T is the absolute temperature, V’ and Veq are instantaneous
and equilibrium volume fractions of alloy precipitates, respectively. The term (1 - V′/Veq)
is commonly used in the Avrami theory to account for the matrix which can no longer
contribute to the transformation and that has been incorporated in the equation to
account for the fact that nucleation sites are consumed as transformation proceeds
[104].
If long-range diffusion occurs as in carbide precipitation, the expression exp(-Q/RT) will
be multiplied by the mole fraction
c
of the slowest moving solute (in this case niobium)
[5,104,100]. Then the nucleation rate per unit volume in this instance is given by:

V'
N& = c  1 −
V eq

The calculation of
nucleation
*



N kT exp  − ∆ G + Q 
 o h


RT



∆G*
Equation 3.9
requires a knowledge of the chemical driving force for
∆Gν, which depends on the chemical composition of the alloy and the
equilibrium concentrations (which determine the initial supersaturation) or alternatively,
by the undercooling ∆T below the equilibrium transformation temperature. There are
two unknowns that cannot be determined experimentally for a complete reaction kinetic
calculation of precipitates to be made, that is, the interfacial energy (γ) and the initial
number of nucleation site (N0). These parameters are treated as fitting parameters and
the sensitivity of the results in the context of M6C carbide precipitation in Fe–C–Nb steel,
has been made previously by Fujita et al. [103]. Their results show that the nucleation
rate is affected more by the interfacial energy than by the number density.
3.2.6 THE TIME-DEPENDENT NUCLEATION RATE
The above equilibrium model did not include the likelihood of the particle’s size greater
than the critical embryo, r* ever decaying. According to the Becker-Döring theory a
decay of nuclei with r >r* is still likely. This is accounted for by the Zeldovich factor Z.
P a g e | 47
A nucleus is supercritical when the size is larger than the size range indicated by 1/Z in
Figure 3.1.
According to the classical theory of nucleation, a general time-dependent equation for
calculating the rate of isothermal nucleation is given by:
 ∆G * 
τ 
 exp 
N& = Zβ *No exp −
t 
 kT 
Equation 3.10
where β* is the atomic impingement rate (which includes the temperature dependent
diffusion rate), No the number of available nucleation sites. The simulation time is t, k
is the Boltzmann constant, T is the absolute temperature, and τ is the incubation time.
The gradient of the driving force ΔG within the region 1/Z is rather small and the cluster
will move across this region predominantly by random walk with the jump frequency β*.
The expected time to cover the distance 1/Z is identified with the incubation time τ:
τ =
1
2β *Z 2
β* =
Equation 3.11
4πr *2Dc α
a4
Equation 3.12
where a is the mean atomic lattice distance of the matrix phase, D is the diffusion
coefficient of the rate controlling solute atoms in the matrix and cα is the equilibrium
solute composition within the matrix. The atomic impingement rate β* is the effective
rate, or probability, with which the atoms change from the matrix to the nucleus surface
and is, therefore, temperature dependent. Equation 3.12 holds true for spherical nuclei
and a binary system.
3.2.7 CHEMICAL DRIVING FORCE
One of the critical parameters that are needed for the calculation, that is, ∆Gv the
chemical free energy change per unit volume of precipitate, is given by:
∆Gν =
∆G
υ iV iα
Equation 3.13
where υ is the molar volume of the ith phase and ∆G is the molar free energy change
of the precipitate reaction, Viα is the maximum volume fraction of the ith phase. ∆G
P a g e | 48
for the formation of some of the niobium carbides can be obtained with a CALPHAD
method via explicit equations (see Section 2.11, Equations 2.16 – 2.20, which are based
on minimising the Gibbs free energy relative to the standard element reference state)
[102,103,104,105].
In multi-component systems of the type Fe-C-Nb-Ti, there are thermodynamic data
available for the carbides and nitrides, but unfortunately not for the Laves phase (Fe2Nb)
and M6C (Fe3Nb3C) type carbides. However, recently the solubility products for these
phases that can be used to estimate ∆G have been determined [6,106]. Table 1 below
shows the solubility products of the precipitates within the ferrite matrix.
Table 4. Solubility products of the precipitates within a ferrite matrix.
System
TiN
TiC
NbN
NbC
Fe3Nb3C
Fe2Nb
Product
–
log[Ti][C] = 4.4 – 9575/T
log[Nb][N] = 4.96 – 12230/T
log[Nb][C] = 5.43 – 10960/T
3
log[Nb] [C] = 5.2178 – 11613/T
log[Nb] = 2.4646 – 3780.3/T
Soluble phase
Ferrite
Ferrite
Ferrite
Ferrite
Ferrite∗
Ferrite∗
∗These solubility products were determined by Fujita and co-workers [6]
Fujita et al [6] have attempted to find thermodynamic parameters using the solubility
) for the free energy
products, and these expressions with the mole fractions (e.g. x αβ
Nb
changes of the precipitation reactions from a niobium – supersaturated ferrite matrix, are
given by:
αβ 3
Fe3Nb3C (β); ∆ G β = − 222509 − RT { 6 .423 + ln( x Nb
) ( x Cαβ )}
αγ
Fe2Nb (γ); ∆ G γ = − 72334 − RT { 0 . 5469 + ln x Nb
}
3.3
GROWTH BY SUPERSATURATION
Attempts to model particle growth by a decrease in supersaturation in a ternary or a
higher system have been made by assuming a binary approach in which only solutes
are considered in the growth equations; thereby violating local equilibrium at the
interface. Bhadeshia stated that the procedure for modelling particle growth is falsely
justified by stating that the solute considered is the one that controls growth, despite the
fact that the theory of diffusion in multi-component systems has been well established
[105].
P a g e | 49
3.3.1 DIFFUSION CONTROLLED GROWTH RATE
A reasonable approximation for isothermal diffusion controlled growth in a binary alloy is
that the compositions of the phases in contact at the interface are locally in equilibrium.
It follows that the concentrations are given by a tie–line on the equilibrium phase
diagram. For a binary system, the tie–line is unique and passes through c , which is
the average concentration of the solute in the matrix alone. The concentration profile
that develops during the precipitation of a solute–rich phase such as carbides, is shown
in Figure 3.7, where cαβ is the concentration of the solute in the ferrite (α) matrix which
is in equilibrium with the precipitate (β) and cβα is the corresponding concentration in
the β which is in equilibrium with α; whereas both are obtained from the phase
diagram.
t3
t2
t1
r1
r2
r3
Figure 3.7. The solute concentration profile during diffusion - controlled growth of β
from α. cαβ and cβ α are concentrations at the interface α/β in the matrix α and
the precipitate β, respectively.
Solute is removed from the matrix as the precipitate grows but since the temperature is
fixed during isothermal growth, the interface compositions must remain fixed at cαβ and
cβα if local equilibrium is to be maintained. Note that the concentration cαβ is affected
by the Thomson-Freundlich or the Gibbs-Thomson equation and is in quasi-equilibrium
with the surface energy γ and is, strictly speaking, not an equilibrium value. In order to
maintain a constant concentration at the interface, the diffusion flux of the solute at the
α/β interface must equal the rate at which solute is partitioned from the matrix to the
precipitate so that:
ν(cβα − c αβ ) = −D
∂c
∂z z = z *
Equation 3.14
P a g e | 50
where ν is the growth rate, z is a coordinate normal to the interface with the value z*
and D is the solute diffusivity. Note that the concentration gradient is evaluated at the
position of the interface, z = z*.
The prominent feature of precipitate growth is that after some incubation time and the
formation of a critical nucleus (taken to have a radius, r = 0), the precipitate grows by
depleting the matrix solute immediately ahead of the advancing interface.
At any
position z in the matrix phase (i.e. z > r(t)) the solute concentration is a monotonically
decreasing function of time [107].
3.3.2 MULTICOMPONENT DIFFUSION GROWTH
In the work done by Robson and Bhadeshia [77], they treated the growth of carbides
using a binary approximation, i.e. in terms of the diffusion of the substitutional element
alone. This is incorrect because the mass balance equation will not be satisfied for the
interstitial solute. For a ternary system like Fe–C–M, where M stands for a substitutional
solute, the tie–line will not in general pass through
c
because it is necessary to
simultaneously satisfy two conservation equations at the interface, one each for the
substitutional element (M = Nb) and carbon, which diffuse at different rates:
βα
αβ
ν(c M
− cM
) = −DM
ν(c βα
− c Cαβ ) = −DC
C
∂c M
∂z
Equation 3.15
z=z*
∂c C
∂z
Equation 3.16
z=z*
Because DC >> DM, these equations cannot in general be simultaneously satisfied for
the tie–line passing through
c , apparently implying that growth cannot occur with
equilibrium at the interface. However, in a ternary alloy there are many tie–lines to
choose from at any given temperature because of the extra degree of freedom given by
the phase rule. The local equilibrium condition can be maintained by choosing a tie–line
which either minimises the concentration gradient of carbon (thus, allowing substitutional
solute flux to keep pace) or maximises the gradient of the substitutional solute to
compensate for its slower diffusivity. The mass conservation equation can be satisfied
simultaneously in two ways. The first is to choose the tie–line which greatly increases
the concentration gradient of M to compensate for its lower diffusivity.
This would
require the carbide to have virtually the same niobium (M) concentration as the matrix
with very little partitioning of Nb, but with a sharp concentration spike at the interface in
P a g e | 51
order to maintain local equilibrium. This is only possible with a very large driving force
and hence it is not applicable to niobium carbide precipitation in microalloyed steels
[103,100]. The alternative is to select a tie–line which reduces the gradient of carbon to
such an extent that the flux of carbon is reduced to a level consistent with that of
niobium.
The intersection of the vertical line with the α / (α + β) phase field defines the tie–line
completely which fixes the interface compositions in a manner which satisfies the
conservation conditions because the large diffusion coefficient of carbon is compensated
for by the very small concentration gradient of carbon. All this assumes that the far field
concentration
c
does not change during transformation, i.e. there is no ‘soft–
impingement’ of the diffusion of different particles and very short annealing times that
prevail.
Note that this also maintains local equilibrium at the interface since the
compositions at the interface (given by the points c and d for α and β respectively)
are connected by a tie–line on the phase diagram. The locus of the matrix composition
due to solute depletion during precipitation is along the direction b → e (Figure 3.8).
The change in the matrix composition leads to a different choice of tie–line, the locus of
cαβ being along c → f. This tie–line shifting continues until the reaction stops when the
tie–line intersects the average composition a and cαβ = f.
Figure 3.8. A schematic isothermal section through the Fe-C-M phase diagram, showing
the ferrite matrix α and alloy carbide β fields. The alloy composition is plotted
as point a [4].
P a g e | 52
Figure 3.9. Distribution of the solute when (a) both (β) and (γ) are precipitating, and (b)
Note that c′ is the
where the precipitation of (β) has been completed.
instantaneous solute concentration in the matrix (α
α) [99].
The mean field approximation can be used to calculate the change in
c
as
precipitation proceeds. The instantaneous value of the matrix composition c′ , is given
by:
c′ = c −
V β (c βα − c )
Equation 3.17
1−V β
Once the interface compositions are defined as described previously, established theory
for diffusion-controlled growth can be applied to estimate the particle radius r as a
function of time:
r = α3
D M t with α3 ≈ 2
c − cαβ
c
βα
Equation 3.18
−c
where α3 is the three-dimensional parabolic rate constant and this equation can only be
used in the absence of the knowledge of interfacial energy or the shape of the nucleus.
The driving force for nucleation must also be affected by the soft impingement. To deal
with this, the extent of the reaction parameter Φ is defined as follows:
Φ=
Vβ
V
βα
with V βα =
c − c αβ
c βα − cαβ
Equation 3.19
P a g e | 53
where Vβ is the instantaneous fraction, and Vβα is the maximum fraction of a given
phase. The function Φ ranges from 0 to 1 and represents the fraction of excess solute
remaining in the matrix relative to the equilibrium composition of the precipitate. It is
assumed that the driving force (∆Gv in Equation 3.9) for the precipitation is linearly
related to Φ:
∆Gν = (1 − Φ )∆Gνo
Equation 3.20
where ∆Gv and ∆Gvo are the driving forces for precipitation at an arbitrary instant and at
t=0, respectively.
3.4
TRANSFORMATION KINETICS
The evolution of volume fraction of the second phase during transformation can be
described using the well known Johnson-Mehl-Avrami-Kolmogorov (JMAK) equation
[95], which, for spherical particles and isothermal conditions, can be expressed as
 1

Vv = 1 − exp − πN& Gr3t 4 
 3

where N&
Equation 3.21
is the nucleation rate, Gr is the growth rate and t is time. Equation 3.17
above is for the special case of homogeneous nucleation. However, most nucleation in
steels occurs heterogeneously, for instance, with the possibility of nucleation on a grain
boundary that may be higher than that on other nucleation sites. There are three types
of nucleation sites on grain boundaries for the nucleation to take place: planes, edges
and corners [100].
In practice, the volume fraction
Vv
of the second phase is
determined as a function of time and this produces a very frequently found sigmoidal or
S-shaped curve. A more general form of the JMAK equation usually used in practice,
takes the form of:
(
Vv = 1 − exp − kt n
)
Equation 3.22
where k is a rate constant, subsuming effects associated with nucleation and growth,
and is usually empirically evaluated for each temperature, n is the time exponent, often
called the Avrami exponent. For this equation to be valid, a plot of {lnln [1/(1-Vv)]} vs {ln
t} should be linear with a slope n and intercept ln k. Table 5 shows the theoretical
values of n and k.
P a g e | 54
Table 5. Values of n and k in the general form of Equation 3.19, Sgb is the area of grain
boundary per unit volume, Lgb is the length of grain boundary edges per unit
volume and Nc is the density of grain boundary corners. All nucleation on the
grain boundaries is assumed to occur before growth [108].
Nucleation site
Time exponent, n
k – value
Homogeneous nucleation
4
π & 3
NGr
3
A plane on grain boundary
1
2 S gb Lgb
An edge on the grain boundary
2
πL gb G r2
A corner on grain boundary
3
4
πNcGr3
3
3.5
OVERALL TRANSFORMATION KINETICS
The Johnson-Mehl-Avrami-Kolmogorov (JMAK) approach treats the precipitation kinetic
problem as an overall kinetic theory. Robson and Bhadeshia developed a simultaneous
precipitation reaction model by extending the classical JMAK concept of extended space
to many phases [109]. Fujita and Bhadeshia improved that model to deal with carbide
size and to account for the capillarity effect [100]. In their work, the JMAK theory has
been applied to describe the kinetics of the single phase in a power plant steel, utilising
MTDATA software to predict the driving forces.
3.5.1 THE ROBSON AND BHADESHIA MODEL
In a simple simultaneous reaction in which both β and θ precipitate at the same time
from the parent α-phase, it is assumed that the nucleation and growth rate do not
change with time and that the particles grow isotropically. If only β is formed, the
untransformed α will contribute to the real volume of β. On average, a fraction
[1 – (Vβ + Vθ)/V] of the extended volume will be in the previously untransformed material.
It follows that the increase in real volume of β is given by the change in extended
volume dV βe :
 Vβ + Vθ
dVβ =  1 −
V

 e
dVβ

Equation 3.23
 Vβ + Vθ  e
dVθ = 1 −
dVθ
V 

Equation 3.24
and, similarly for θ
P a g e | 55
In general, Vβ is a complicated function of Vθ and it is not possible to analytically
integrate these equations to find the relationship between the actual and extended
volumes. However, in certain simple cases, it is possible to relate Vβ to Vθ
by
multiplication with a suitable constant K:
Vθ = KVβ
Equation 3.25
Equations 3.20 and 3.21 can then be rewritten as:
 Vβ + KVβ  e
dVβ = 1 −
dVβ
V


Equation 3.26
 V + KVθ  e
dVβ = 1 − θ
dVθ
KV 

Equation 3.27
Accordingly, the final expressions of the volume fractions of β and θ phases are:
 1 
 1
3 4 
Vβ = 
1 − exp− (1 + K )πN& β Gr ,β t  
 1 + K 
 3

Equation 3.28
 1  1 + K  & 3 4 
 K 
Vθ = 
1 − exp− 
πNθ Gr ,θ t  
 1 + K 
 3 K 

Equation 3.29
In practice, the multiple reactions found in most industrial steels have important
complications not included in the model above. Precipitation reactions may affect each
other by removing solute atoms from the matrix. Any change in the matrix composition
must alter the nucleation and growth rates of the phases. Therefore, there are no simple
constants linking the volume fraction of all the phases and a different approach is
needed.
3.5.2 FUJITA AND BHADESHIA MODEL
The Robson and Bhadeshia model can be used to estimate the volume fraction of
carbides but it would be useful also to treat particle sizes and coarsening after growth.
This has been achieved by Fujita and Bhadeshia, who attempted to take better account
of multicomponent diffusion and capillarity [100]. Given the small equilibrium volume
fraction of carbides in most ferritic steels, they also relaxed the extended volume
concept to permit particle sizes to be calculated approximately.
P a g e | 56
3.6
CAPILLARITY
The state of equilibrium between two phases changes with the curvature of the interface
separating them. This is the well established Gibbs–Thompson capillary effect and is
due to the curvature of the interface that then influences the change in equilibrium
compositions at the particle/matrix boundary. The free energy change of the particle
phase varies relatively sharply with a deviation from the stoichiometric composition so it
can be assumed that the particle composition is insensitive to the curvature. However,
the equilibrium composition of the matrix changes as follows [99,101,103,104,110]:

γ υ β 1 − c αβ
c rαβ =  1 +

kT r c βα − c αβ

where
crαβ

c αβ


Equation 3.30
is the solute concentration in the α matrix that is in equilibrium with a
spherical particle of β and r is the radius of curvature, which in this case also defines
the instantaneous particle size.
The term
cαβ = crαβ
when
r = ∞.
The modified
composition crαβ is, therefore, relatively easy to estimate for each particle. At some
critical value where r = rc (where rc is a critical radius) and c rαβ = c , growth ceases and
coarsening starts.
Note in particular in the case of the determination of the particle density measurements,
that the number of particles per unit volume Nv first increases, as would be expected
during the nucleation stage, but that soon thereafter it starts to decrease, as would be
expected for the coarsening stage. This transition generally takes place after only a few
minutes of annealing [111].
For a ternary alloy, capillarity is approximated by calculating the α / (α + β) phase
boundary on an isothermal section of the phase diagram, as a function of r, using
Equation 3.30. Equation 3.30 is used to calculate c rαβ, C and c rαβ, M for a fixed value of
interface radius of curvature r. The growth velocity can then be calculated using the
curvature-modified phase boundary. Embryo particles that are smaller than the size of
the critical nucleus obviously cannot grow and will dissolve again. Nucleation occurs
statistically by random fluctuations in composition so that the growth part in the
P a g e | 57
computational scheme must start beyond the nucleation stage. Particles nucleate at
different times during the course of the reaction, giving rise to a distribution of sizes.
At any given stage of precipitation, the smaller particles will grow at a slower rate than a
larger particle because the capillary effect reduces the supersaturation at the interface
for small particles.
Capillarity has the consequence that large particles have lower
solute concentrations at the interface
crαβ
than small particles [112].
This drives
coarsening, which becomes a natural consequence of the precipitation theory, since
changes including the dissolution of particles, continue to happen as long as there are
solute concentration gradients.
3.7
DISSOLUTION OF THE METASTABLE PHASE
With soft-impingement being considered, the mean solute concentration within the
matrix alone decreases as the reaction proceeds.
Each precipitating phase will
consume or reject atomic species into the untransformed matrix, whose subsequent
transformation behaviour will be altered. Consider the precipitation of β and γ from an
α ferrite matrix, which initially has a uniform composition.
Schematic composition
profiles are shown in the earlier Figure 3.9 (a) of how the distribution of the solute might
change during precipitation at the intermediate stage when both the metastable
β
phase and the equilibrium γ phase are precipitating in the matrix of α simultaneously.
Also from Figure 3.9 (b) when the solute concentration in the matrix reaches equilibrium
with β; precipitation of β has been completed and as γ precipitates further removing
solute, β will start to dissolve. For the situation illustrated, the maximum fraction Vβα, of
each phase is given by a lever rule (see Equation 3.19).
3.8
PARTICLE COARSENING
3.8.1 DIFFUSION CONTROLLED COARSENING OF THE PARTICLES WITHIN M ATRIX
The coarsening of particles occurs (even though the transformation is said to be
completed) when there is no significant change in the precipitate volume fraction over a
period of time.
Note that it is traditional to separate transformation growth and
coarsening but the two processes are in fact both fulfilled with capillarity effects. The
coarsening rate of precipitates can be calculated using the classical theory of the
Ostwald ripening equation that is due to Lifshitz and Slyozov [113] and Wagner [114]
P a g e | 58
and is often called LSW coarsening. The LSW coarsening rate equation for diffusion
controlled coarsening is given by:
r n − ron =
8γυ β Dc αβ
t
9RT
Equation 3.31
where r is the average particle radius, r0 the initial average particle radius (a fictitious
value as no particles exist at t = 0), γ is the interfacial energy, D is the diffusion
coefficient of the rate controlling species, υβ is the molar volume of the phase β, cαβ is
the equilibrium solute concentration in the α matrix at which r→∞, R is the gas
constant, T is the absolute temperature, t is the holding time at the isothermal heat
treatment temperature.
For intragranular particles it can be assumed that the
coarsening of the particles is controlled by bulk diffusion, therefore n =3.
3.8.2 DIFFUSION CONTROLLED COARSENING OF THE PARTICLES ON GRAIN BOUNDARY
The above Equation 3.31 is only valid for diffusion controlled coarsening of the particles
within the matrix.
But there are some scenarios whereby the particles in the grain
boundaries coarsen through the diffusion of the solute atoms down the grain boundary.
Here, there is a two dimensional diffusion down the grain boundary from the smaller
particle to the larger particle. The theory is basically the same as for matrix diffusion and
the coarsening rate equation changes to:
r 4 − ro4 =
K1γυ β Dgb c αβ δ gb
t
kT
Equation 3.32
where K1 is a constant, Dgb = diffusion coefficient down the grain boundary, δgb =
width of the grain boundary and
t = time at temperature.
The above coarsening
Equation 3.32 is also only fully valid if all of the precipitates were situated on grain
boundaries and this, of course, is quite difficult to achieve. It may be approximated,
however, by introducing a subgrain structure into the matrix in which, very often, most of
the particles are “captured” by the subgrain boundaries.
In the work by Kostka et
al.[115] and Mukherjee et al.[116], the authors observed that the grain boundary
carbides coarsen faster than the matrix carbides.
P a g e | 59
3.8.3 DIFFUSION CONTROLLED COARSENING OF THE PARTICLES ON SUBGRAIN BOUNDARIES
This mechanism may be found in cases where high temperature creep at a service
temperature takes place and all of the precipitates become entangled and are, in fact,
interconnected by dislocations in subgrain boundaries:
r 5 − ro5 =
where
K1γυ β Ddislcαβδ dislN′
t
kT
K = constant,
1
Equation 3.33
δ = effective diameter of a dislocation,
disl
N′ = number of
dislocations that meet each particle and D = diffusion coefficient down a dislocation.
disl
Direct experimental evidence of this power law has not been obtained although many
cases of direct and indirect evidence do exist that precipitates that are situated on
subgrain boundaries, grow at a faster rate than others.
3.9
SUMMARY
Many researchers have modelled and calculated the values for the interfacial energy of
the M6C carbide and Laves phase in Fe–C–Nb systems and have estimated them to be
0.286 Jm-2 [103] and 0.331 Jm-2 [77,117,118], respectively.
From these results,
because of a complex unit structure of the Laves phase, its interfacial energy seems to
be larger than that of M6C carbide as would be expected. In other work also done on
the Fe–C–Nb systems by Sim et al. [3], they have assumed a much higher value of 1.0
Jm-2 for the Fe2Nb Laves phase because of its incoherency with the α-matrix.
Fujita et al., [99] using the MTDATA thermodynamic software, have modelled the
precipitation kinetics in the Fe-Nb-C system for a 9Cr-0.8Nb steel at 950°C.
The
precipitation sequence was found to be as follows:
α → α + Nb(C,N) + Fe2Nb + Fe3Nb3C → α + NbN + Fe3Nb3C
and this shows that in this steel the equilibrium phases are NbN and Fe3Nb3C, and the
Laves phase is a meta-stable phase. Figure 3.10 shows the comparison between the
experimental and the calculated phases in obtaining the precipitation sequencing.
P a g e | 60
Figure 3.10. The kinetics of the precipitation sequence in 9Cr-0.8Nb ferritic stainless steel
[101].
P a g e | 61
CHAPTER FOUR
EXPERIMENTAL PROCEDURES
4
4.1
MATERIALS
Type 441 ferritic stainless steel which is a niobium and titanium dual stabilised steel,
was studied in this work. The main emphasis was on understanding the formation of the
intermetallic Laves phase and its impact on the toughness of this steel. Steels A and B
were supplied from Columbus Stainless in the hot rolled condition with Steel A having
not received any water cooling on the hot mill’s run out table due to a system failure.
This coil was found to be too brittle to be processed further and was, therefore, made
available for this study while specimens of Steel B were as processed normally with
adequate laminar flow cooling after hot rolling.
The effect of the alloy’s chemical composition was also investigated through two
experimental alloys that were within the type 441 ferritic stainless steel’s specification.
The first experimental alloy identified as Steel C, i.e. a Nb–Ti alloy, is similar to Steel A
but has a higher carbon content (almost double that of Steel A’s 0.012%C) and lower Nb
content of only 0.36%Nb if compared to the Steel A’s Nb-content of 0.44%. The second
experimental alloy identified as Steel D, is a Nb–Ti–Mo alloy very similar to Steel A but
contains an additional ∼0.5 wt.% Mo.
The effect of even higher molybdenum (Mo)
additions was studied on AISI type 444 ferritic stainless steel, that is Steel E, which is a
high Mo containing steel with the nominal composition of 18Cr–2Mo; this alloy is also a
dual stabilised ferritic steel with a slightly lower content of niobium and titanium. Table
4.1 shows the chemical compositions of the alloys used in the present study while
Figure 4.1 shows the detailed flow chart of the research plan to study the nucleation and
kinetic models for the Laves phase precipitation and its effects on the impact toughness
of the AISI type 441 stainless steel.
P a g e | 62
Table 4.1. Chemical composition (in %wt) of ferritic stainless steel studied in this work.
Elements
C
Mn
Co
Cr
B
V
S
Si
Ti
Ni
N
Al
P
Cu
Nb
O
Mo
Supplied by Columbus
Steel A
Steel B
441 SS†
441 SS‡
0.012
0.015
0.51
0.54
0.03
0.02
17.89
17.9
0.0004
0.0006
0.12
0.14
0.001
0.002
0.5
0.153
0.149
0.19
0.19
0.0085
0.021
0.009
0.009
0.025
0.023
0.08
0.09
0.444
0.445
0.0076
0.008
Hot rolled experimental alloys
Steel C
Steel D
Steel E
Nb-Ti
Nb-Ti-Mo
444 SS§
0.023
0.012
0.014
0.46
0.35
0.45
0.02
17.9
17.6
18.3
0.0006
0.12
0.0073
0.0018
0.003
0.33
0.31
0.47
0.171
0.171
0.106
0.13
0.12
0.15
0.024
0.026
0.016
0.011
0.013
0.01
0.024
0.032
0.024
0.07
0.06
0.12
0.36
0.39
0.251
< 0.01
0.54
1.942
Figure 4.1. Experimental plan.
†
Columbus Stainless MPO Number 3533603- failed during production
Columbus Stainless MPO Number 3658671 – didn’t fail
§
Columbus Stainless MPO Number 3631171
‡
P a g e | 63
4.2
THERMODYNAMIC MODELLING
Modelling precipitation in AISI type 441 ferritic stainless steel requires knowledge of both
thermodynamic and kinetic parameters for a variety of possible phases. To obtain these
parameters, the computer package Thermo-Calc® version Q (TCFE3 database [119])
thermodynamic software was used to make these thermodynamic calculations. All of
these experimental alloys were initially modelled to determine the precipitation
temperatures of the phases present, in particular the Laves phase. This has assisted in
designing a suitable heat treatment for this alloy without causing major grain growth. A
detailed modelling procedure is given in Chapter 5.
4.3
HEAT TREATMENTS
The results obtained from the Thermo-Calc® modelling were used as guidance for
designing an appropriate heat treatment for the alloys. The effect of the Laves phase
embrittlement was studied by subjecting the specimens to different heat treatments in an
inert argon atmosphere.
4.3.1 LAVES PHASE DISSOLUTION/PRECIPITATION TEMPERATURES
The first step was to determine the annealing temperatures at which the Laves phase
dissolves and its impact on the mechanical properties of the steel during its reprecipitation.
Most of the annealing treatments were designed around the solvus
temperature of the Laves phase, which is within the range of 780 to 950 °C depending
on the composition of the alloy, as reported in the literature and calculated by ThermoCalc® predictions. Steel A was used for this part of the study and the specimens were
annealed within the temperature range of 600 to 1100 °C for 30 min. followed by water
quenching.
4.3.2 HEAT TREATMENT FOR THE EMBRITTLING EFFECT
Heat treatments for embrittlement of the specimens from AISI type 441: Steel A were
performed on:
(1) Charpy impact specimens that had first been solution annealed at 950 °C for
1hr to dissolve the Laves phase and then quenched in water.
These
specimens were then reheated within the temperature range of 600 to 900°C
for 30 min to re-precipitate the Laves phase and were then water quenched,
see Figure 4.2.
From these specimens the effect of the Laves phase reP a g e | 64
precipitation at different temperatures on the upper – shelf energy and DBTT in
Steel A could be studied.
(2) A second set of experiments was used to determine the effect of cooling rate
from typical hot rolling temperatures on the embrittlement of these materials
through the application of programmed linear cooling rates on the specimens
from the solution treating temperature. The programmed cooling rates were
applied to specimens in a Gleeble® 1500D Thermal Simulator using subsize
Charpy specimens that were annealed at 850 and 950 °C respectively for 5
minutes in an inert argon atmosphere followed directly by forced cooling with
helium at different linear cooling rates ranging from 1 to 60 °C/sec, see Figure
4.3.
Temperature (°C)
Solution annealing at 950
°C for 1hr
900 °C
800 °C
700 °C
WQ
600 °C
WQ
Time (min)
Figure 4.2. Embrittlement through reheating to determine the effect of the Laves phase
re-precipitation on the DBTT and upper shelf energy of steel A.
annealing at 850 / 950 °C
for 5 min
Figure 4.3. Embrittlement through cooling to determine the effect of the Laves phase reprecipitation on the Charpy impact toughness.
P a g e | 65
4.3.3 HOT-ROLLING OF EXPERIMENTAL ALLOYS
The experimental alloys C and D were prepared by Mintek in a 5 kg vacuum induction
melting furnace (VIM) using scrap from Steel A as base scrap material. The as-cast
alloys were homogenised in an argon atmosphere at 1200 °C and hot forged to bars of
about 30 mm thick. Steel E was supplied by Columbus Stainless as a 28 mm thick hotband material. Subsequently, these three steels, i.e. C, D and E were solution treated
at 1200 °C for 1hr followed by hot rolling to a final thickness gauge of 5.5 to 6.0 mm and
quenched in water.
The average last pass temperature before quenching was
approximately 950°C, which is above the calculated Laves phase formation temperature.
4.3.4 LAVES PHASE KINETIC STUDY
The study of the Laves phase kinetics was conducted using a small tubular furnace that
provided a suitable small temperature gradient, as shown in Figure 4.4.
Both the
furnace and specimen temperature were monitored and controlled by the data logger
model DT500 dataTaker® and PID temperature controller, respectively.
A type K
thermocouple was spot welded to the specimens to monitor the annealing temperature
inside the furnace. Figure 4.5 shows the time- temperature profile of the specimen
inside the small tubular furnace.
Note that the start of the annealing time was
considered from the point whereby the specimen had reached a desired temperature
and had stabilised.
After annealing for a required period of 1 – 1000 minutes, the
specimens were subsequently quenched in water.
These specimens where then
subjected to the Charpy impact test and also the precipitates were extracted
electrolytically from them, see Section 4.6.1.
E
C
A
B
D
Figure 4.4. The furnace used for the precipitation kinetic study. (A) tube furnace; (B)
temperature controller; (d) data logger; (D) type k thermocouple; (E) recording
computer.
P a g e | 66
Figure 4.5. The temperature gradient of the Charpy impact specimen inside the furnace.
4.4
MECHANICAL TESTING
4.4.1 TENSILE TESTS
Tensile tests were carried out at a cross head speed of 5 mm/min at room temperature
using two subsized specimens prepared according to the ASTM E8** standard for each
measurement.
The longitudinal axes of the specimens were parallel to the rolling
direction of the plate throughout. The schematic drawing of the tensile specimen and its
dimensions are shown in Figure 4.6 and Table 4.2, respectively.
Figure 4.6. Schematic diagram of the subsize tensile test specimen.
Table 4.2. The actual dimensions for the subsize tensile test specimen.
G
W
T
R
L
A
B
C
**
Parameters
Gauge length
Width
Thickness
Radius of fillet
Overall length
Length of a reduced section
Length of grip section
Width of grip section
Dimensions (mm)
25
6
5
6
100
32
32
10
ASTM E8 – 08: Standard Test Methods for Tension Testing of Metallic Materials
P a g e | 67
4.4.2 NOTCHED CHARPY IMPACT TEST
Charpy impact test values were obtained as an average from three subsized specimens
of 5 x 10x 55 mm3 prepared according to the ASTM E23†† standard. A standard 2mm Vnotch was made in the specimens after the heat treatment to avoid oxidation effects
from the heat treatment at the notch tip. The longitudinal axes of the specimens were
parallel to the rolling direction of the plates throughout the study.
The specimens for DBTT Charpy tests were prepared by immersing them in different
liquid media to obtain the desired test temperature. For the temperatures higher than 25
°C, heated water controlled by a thermal regulator was used. To achieve a 0 °C testing
temperature, icy water was used and below 0 °C, a mixture of dry ice‡‡ and ethanol or
liquid nitrogen§§ and ethanol were used, depending on the desired lower temperature
one wants to reach. The mixtures were prepared inside a thermo flask and different
mixing ratios of dry ice to ethanol or liquid nitrogen to ethanol were used to achieve the
desired testing temperature. Once the testing temperature was reached the specimens
were allowed to settle for about 5 minutes in order to normalise the temperatures before
they were subjected to immediate Charpy impact testing.
4.4.3 HARDNESS TESTS
Vickers microhardness measurements were made using a load of 30 kgf to give a
relatively large indentation to offset any effects due to localised variation in structure.
The average of at least five readings was taken for each result.
4.5
MICROANALYSIS OF SPECIMENS
Several types of microscopy were performed to locate, identify, and quantify the
microstructural components in the steels after their different treatments, ranging from
optical microscopy, scanning electron microscopy to transmission electron microscopy.
Analysis of small particles can often prove to be difficult but through a combination of
these different techniques coupled with energy dispersive spectroscopy, one can
reasonably gain insight into the precipitation behaviour.
††
ASTM E23 – 02a: Standard Test Methods for Notched Bar Impact Testing of Metallic Materials
Dry ice sublimes at -78.5 °C
§§
Liquid nitrogen boils at -196 °C
‡‡
P a g e | 68
4.5.1 OPTICAL MICROSCOPY
The metallographic specimens were prepared by standard metallographic techniques on
a polishing system to a 3µm finish. The specimens were then etched electrolytically in
60% nitric acid in water at a potential of 1.5 V dc for a period of 30 to 120 sec., to
optimise the phase contrast and grain size effects. Microstructural analysis was carried
out using Olympus PGM and Nikon Eclipse ME600 optical microscopes, both equipped
with AnalysisTM image software.
The grain size was determined using the linear
intercept method in accordance to the ASTM E 112 - 96*** standard.
4.5.2 TRANSMISSION ELECTRON MICROSCOPY (TEM)
TEM observations were carried out using a Philips CM 200 TEM operating at an
accelerating voltage of 160kV and a JOEL JEM – 2100F field emission TEM operating at
an accelerating voltage of 200kV.
4.5.2.1 PREPARATION OF TEM SPECIMENS
Two types of specimens were examined using TEM: thin foils and carbon extraction
replicas. From these, the types of precipitates were identified using energy dispersive
X-ray analysis spectroscopy (EDS) and small angle electron diffraction patterns (SAED).
Thin foils were cut from the bulk materials as 3mm diameter discs of about 1 mm
thickness using an electrospark wire cutting device. After cutting, the specimens were
attached to a large steel block using hot wax. These specimens were mechanically
ground from both sides with silicon carbide paper to less than 70 µm thickness before
the wax was dissolved in acetone to release the thinned specimens. Electropolishing
was conducted until perforation of the disc occurred using a twin jet electropolisher in a
10 % perchloric acid in 90% acetic acid solution at room temperature at a voltage of 55
dc. Finally, the specimens were cleaned using a four step ethanol cleaning process to
remove all residues from the specimens.
Carbon replica specimens were firstly mounted as for metallographic examination and
etched the same way. A carbon film was applied by vacuum evaporating carbon onto
the etched specimen surface. The carbon film was electrolytically detached in a solution
of 5 vol.% hydrochloric acid in methanol at a constant potential of 1.5 V dc after which
***
ASTM E112 – 96: Standard Test Method for Determining Average Grain Size
P a g e | 69
the replicas were washed using ethanol and then floated off onto distilled water, from
where the replicas could be collected on 3 mm diameter copper grids.
4.5.3 SCANNING ELECTRON MICROSCOPY (SEM)
The scanning electron microscopy (SEM) investigations were carried out on a Jeol JSM
– 6300 and also on a high resolution field emission scanning microscope model Joel
JSM – 6000F (SEM). The accelerating voltage was set at 15 kV and the probe current
was 7 x 10-9 A in both electron microscopes. Both the thin foils and the metallographic
specimens were used for the particle’s analyses. The advantages of using TEM thin
foils in a SEM is that: (1) the specimens are very clean which helps to resolve small
particles; and (2) the electropolishing tends to leave fine precipitates on top of the
surface, facilitating a good image and analysis of the particles.
4.6
IDENTIFICATION OF PRECIPITATES
4.6.1 XRD STUDY
XRD analysis was carried out on a PANalytical X’Pert Pro powder diffractometer with
X’Celerator detector and variable divergence and receiving slits equipped with an Fefiltered Co-Kα (λ = 1.789 Å) anode as a source of X-rays. The scanning was carried out
at an angular range of 20 to 90 (°2 Theta) employing a step size of 0.017° 2 Theta. The
generator setting was at 50 mA and 35 kV.
4.6.1.1 SPECIMEN PREPARATION
X-ray analysis on a larger representative volume of each steel, not only supports the
TEM observations from small volumes of materials, but also provides information on the
quantity of each phase with a reasonably good accuracy. Electrochemically extracted
residues of second phase particles were examined by X-ray powder diffraction to identify
the types and volumes of precipitates formed in these alloys. To obtain residues, the
precipitates were extracted by means of potentiostatic electrolysis at a constant potential
chosen between 100 and 200 mV against a silver chloride electrode in a solution of 10
vol.% acetyl acetone and 1 vol.% tetramethyl ammonium chloride in methanol (TMAC),
in which only the iron matrix dissolves [8], leaving a mixture of the second phases
behind as a powder residue. Also, the use of 60% nitric acid in water at a potential of 5
V dc to electrolytically extract the particles was found effective in extracting larger
volumes of precipitate residue in a short space of time.
The residue was vacuum
P a g e | 70
suction filtered using a sub-micron glass filter paper (≤ 0.7 µm) to trap the fine particles
with the filter paper pre-weighed to account for the very fine precipitates that might
remain trapped in it during the filtration process. The extracted residues where rinsed
with ethanol, dried and weighed to obtain the weight fraction from the previously
weighed steel specimen.
4.6.1.2 ANALYSIS
Phase identification was carried out using X’Pert Highscore Plus software, using an
integrated interface and fast retrieval software for the International Centre for Diffraction
Data (ICDD) reference database, flat file format and relational databases. Any number
of user reference patterns that have been modified can be added into the reference
databases.
Figure 4.7 shows the calculations for user reference patterns for the phases that were
expected to be present in these steels.
These patterns were generated using the
PowderCell 2.4 software and can be read in ICDD structure files by exporting them in a
variety of forms into the X’Pert Highscore Plus software.
The generation of these
patterns requires the knowledge of the symmetry (translational and space group), unit
cell data and atomic position parameters, see Appendix A.
The complexity in quantification of the carbides and nitrides expected in these steels
arises from the structural phases in which these carbides and or nitrides may be formed,
i.e. whether the carbides are present as NbC/ TiC and /or NbN / TiN or Nb(C,N) or
Ti(C,N) or as (Ti,Nb)(C,N). All of these phases have similar peaks, but the differences
arise from the peak’s position of each phase, see Figure 4.7. From this figure it can be
seen that there is not much of a difference in the peak’s position for the carbide and
nitride phases of niobium or titanium, but the combination of niobium and titanium, as in
(Ti,Nb)(C,N), makes a significant difference in the positions of the peaks.
Note that, in Figure 4.7 and Figure 4.8, both figures show the simulated phase patterns
and these are, therefore, not the actual results. PowderCell predictions were based on
the assumption of a pure specimen that has equal amounts of phases. Therefore, the
results from the precipitate residues will be slightly different from these, and they
required a further refinement using Autoquan/BGMN software to obtain an accurate
volume fraction for each phase.
P a g e | 71
2728
822
311
20
25
30
35
40
45
55
2 theta / deg
60
400
75
006
644
300
301
204
642 105
211
70
114
210
444
65
555
302
213
733
222
50
203
533
620
104
202
004
102
400
200
551
331
222
101
0
002
100
220
311
110
422
200
201
103
1364
511
111
220
200
I/ rel.
112
ALPHA- FE 25.0%
FE2N B-LAVES PH ASE 25.0%
FE3N B3C 25.0%
TIN B(CN ) 25.0%
80
85
90
Figure 4.7. The XRD powder pattern of the phases that were expected to be present in type 441 stainless steel as generated
using a PowderCell software.
P a g e | 72
311
400
400
222
222
311
222
220
111
111
220
200
4126
311
220
200
111
200
I/ rel.
NB(C,N) 16.7%
N BC 16.7%
N BN 16.7%
TIC 16.7%
TIN 16.7%
TIN B(CN ) 16.7%
111
8253
0
25
30
35
40
45
50
55
60
2 theta / deg
65
70
75
80
85
90
Figure 4.8. The XRD powder pattern showing the peak’s positions of the carbide and nitrides of titanium and niobium. Notice
the position of the (Ti,Nb)(C,N).
P a g e | 73
Quantification of the various phases’ weight fractions in the powder residue was carried
using Autoquan/BGMN software (GE Inspection Technologies) employing a Rietveld
refinement approach [120,121]. The quantification was done using the phases’ reference
data from the ICDD databases or from the user reference database.
A typical XRD
spectrum from the analysed powder residue after quantification is shown in Figure 4.9.
Notice how good is the residual differences (i.e. bottom spectrum) between the calculated
and the measured spectrums.
The quantified results from the X-ray analysis are given as weight fraction, therefore for the
volume fraction calculation; the densities for each phase are needed. Table 4.3 shows the
densities of all phases present in these alloys [122].
Table 4.3. The density of the phase that were used in the quantifications.
Phase
Fe2Nb
Fe3Nb3C
α-Fe
Nb,TiC
NbC
NbN
TiC,N
TiC
TiN
3
Density (g/cm )
8.63
8.44
7.87
6.39
7.80
8.40
7.82
4.94
4.83
P a g e | 74
500
FE2NB-LAVES
450
400
350
I / cps
300
∗
250
200
150
∗
∗
∗
∗
∗
∗
100
50
0
25
30
35
40
45
50
55
2 Theta / °
60
65
70
75
80
85
60
65
70
75
80
85
Differences
200
150
I / cps
100
50
0
-50
-100
-150
-200
25
30
35
40
45
50
55
2 Theta / °
Figure 4.9. A typical XRD scan of the precipitate’s residue from Steel A showing the presence of the Laves phase peaks
(indicated by the lines in the top figure). The remaining peaks are the carbides and nitrides, indicated by (∗
∗). Note the
good residual difference between the calculated and the measured spectrum as is shown by the spectrum below.
P a g e | 75
4.6.2 ELECTRON DIFFRACTION PATTERNS
Selected area electron diffraction (SAED) patterns of second phases using carbon
extraction replicas, is a reliable method to identify unknown precipitates because there is
no interference from the matrix. The intended precipitates are chosen using a selected
area aperture and a corresponding diffraction pattern is obtained. Figure 4.10 shows the
schematic drawing of the geometry of diffraction.
Figure 4.10. The single crystal electron diffraction pattern
Rhkl is the distance between the transmitted spot and the diffracted spot from the hkl
plane. θB is the Bragg angle corresponding to the hkl plane and L is the camera
length.
The relationship between these parameters can be described as follows
[123,124]:
Rhkldhkl = Lλ
Equation 4.1
where dhkl is the calculated spacing of hkl planes and λ is wavelength given by:
h
λ=
λ=
Å
2meV
12.236
V
Å
Equation 4.2
Equation 4.3
Electrons accelerated by a potential difference of V volts have a kinetic energy of ½
mν2, where:
½ mν2 = Ve
Equation 4.4
P a g e | 76
e being the standard for an electron charge.
The wavelength λ is associated with an electron of mass m grams travelling with a
velocity ν cm/sec. A relative correction is needed for the actual conditions involving the
voltages used, so that the actual formula is slightly more complicated, as follows:
λ=
12.236
[V (1 + 0.9788x10 −6 V )]
The camera constant
(Lλ)
Å
Equation 4.5
was calibrated using a gold film when the electron
microscope is operating at 160 kV. From equation 4.5 above, the wavelength (λ) was
calculated as 2.84 x 10-12 m, and the camera constant (Lλ) was determined to be 2.70 x
10-12 m2 when using a camera length L of 950 mm.
4.7
THE ORIENTATION RELATIONSHIP BETWEEN THE LAVES PHASE AND THE MATRIX
The habit plane and the morphology of the Laves phase precipitates were determined
from the thin foil specimens.
For simplicity, because of a very small size of the
precipitates to be analysed without interference from the matrix, the orientation
relationship was determined firstly, by obtaining the pattern from the adjacent matrix
followed by the pattern from both the matrix and the precipitates. From the two patterns,
one is able to determine the pattern of the precipitate by correlating the two patterns.
P a g e | 77
CHAPTER FIVE
THERMODYNAMIC MODELLING
5
5.1
INTRODUCTION
Modelling precipitation in a ferritic stainless steel AISI type 441, requires knowledge of
both thermodynamic and kinetic parameters for a variety of possible phases. To obtain
these parameters, experimental data are required.
The prerequisite thermodynamic
information for modelling consists of:
1. the stable phases corresponding to the annealing temperature;
2. the phase fractions corresponding to the annealing temperature;
3. the relative phase stabilities; and
4. the equilibrium chemical compositions of possible phases.
The use of the computer package such as Thermo-Calc® version Q (TCFE3 database
[125]) thermodynamic software makes these thermodynamic calculations possible.
5.2
DESCRIPTION OF THERMO-CALC® SOFTWARE
Thermo-Calc® software is a powerful and flexible software package that can perform
various kinds of equilibrium thermodynamic and phase diagram calculations by
minimising the total Gibbs energy of the system specified [126]. It can handle complex
problems involving the interaction of many elements and phases.
It is specially
designed for the systems and phases that exhibit highly non-ideal behaviour [127,128].
Without an accurate and validated database, any thermodynamic software is useless
and misleading.
Therefore, a high quality thermodynamic database is essential for
conducting reliable thermochemical calculations and simulations.
The TCFE3 TCS
Steels/Fe-alloys database which covers a complete and critical assessment of binary
and some ternary systems as well as the iron rich corner of some higher order systems,
was employed in these calculations. It was pointed out that with TCFE1 TCS Steels/Fealloys database (version 1.1), data for intermetallics such as σ, µ and Laves phases are
less reliable and there is nothing more being said about improvements to the TCFE2
P a g e | 78
and TCFE3 databases for these phases [129].
This might be the reason why the
calculations show the presence of the σ-phase in AISI type 441 stainless steel, whereas
it is not expected to be formed in practice in this alloy.
Also the recommended
temperature range in using the database is from 700 to 2000 °C for the TCFE3.
Calculations slightly outside these limits may give reasonable results but it requires
experience and skill to correctly extrapolate data and interpret the calculational results.
This is particularly so at temperatures of 500 °C and lower where thermodynamic
equilibrium is highly unlikely in industrial processes due to slow diffusion. A lower limit of
600°C was, therefore, adopted in the thermodynamic modelling for this study. Note the
use of the following phase names in the database; the number symbol # is used to
denote different composition sets of the same phase, Table 5.1.
Table 5.1. The phase names as used in the TCFE3 Steels/Fe-alloys database.
Phase
Austenite
Ferrite
M(C,N)
5.3
database
FCC_A1#1
BCC_A2
FCC_A1#2
EXPERIMENTAL ALLOYS
The purpose of this research is to study the precipitation behaviour of the Laves phase
in the ferritic stainless steel AISI type 441. The ternary and quaternary systems of the
Fe-Nb-Ti and Fe-Nb-Ti-Mo were therefore, chosen, the latter being more of an
experimental alloy.
The experimental alloys’ compositions were chosen within and
outside the recommended Columbus Stainless composition specification for the ferritic
stainless steel AISI type 441 (see Steel A & B††† in Table 4.1) in order to determine the
effect of the Nb, Ti and Mo on the Laves phase formation and its embrittlement effect.
Thermo–Calc® software was used in the design of these experimental alloys, with the
main emphasis on determining the effect of composition change on the volume fraction
and solvus temperature of the Laves phase.
The following criteria were used in
designing these alloys:
1. lowering the Nb content while keeping Ti content constant;
2. keeping both the Nb and Ti constant and adding 0.5% Mo; and
3. increasing the Nb content to 0.6 wt% and adding about 0.25%Mo.
†††
Note that, both steels were supplied by Columbus Stainless for a research purposes. Steel A is the
rejected material because it was brittle for further processing, while Steel B was acceptable.
P a g e | 79
A summary of the Thermo–Calc® predictions is given in Table 5.3, in which the results
were obtained in a similar manner as those that are shown in the following sections.
Note that a Mo addition increases both the volume fraction and the solvus temperature
of the Laves phase precipitates, whilst reducing the Nb content lowers both the volume
fraction and the solvus temperature of the Laves phase in the alloy. From this prediction
it was then decided to replace 0.6Nb – 0.15Ti – 0.25Mo with a ferritic stainless steel AISI
type 444, and cast the other two experimental alloys using AISI type 441 as basis from
scrap metal. The chemical composition of the experimental alloys that were physically
tested in this study are shown in Table 4.1.
Table 5.2. Chemical composition (in %wt) of the ferritic stainless steels studied.
Elements
C
Mn
Co
Cr
B
V
S
Si
Ti
Ni
N
Al
P
Cu
Nb
O
Mo
Supplied by Columbus
Steel A
Steel B
‡‡‡
§§§
441 SS
441 SS
0.012
0.015
0.51
0.54
0.03
0.02
17.89
17.9
0.0004
0.0006
0.12
0.14
0.001
0.002
0.5
0.153
0.149
0.19
0.19
0.0085
0.021
0.009
0.009
0.025
0.023
0.08
0.09
0.444
0.445
0.0076
~0
0.008
Hot rolled experimental alloys
Steel C
Steel D
Steel E
****
Nb-Ti
Nb-Ti-Mo
444 SS
0.023
0.012
0.014
0.46
0.35
0.45
0.02
17.9
17.6
18.3
0.0006
0.12
0.0073
0.0018
0.003
0.33
0.31
0.47
0.171
0.171
0.106
0.13
0.12
0.15
0.024
0.026
0.016
0.011
0.013
0.01
0.024
0.032
0.024
0.07
0.06
0.12
0.36
0.39
0.251
< 0.01
0.54
1.942
Table 5.3. Laves phase’s compositions, volume fractions and solvus temperatures used
in designing the optimum chemical composition.
Alloy
composition
% mole fraction
0.35Nb – 0.15Ti
0.444Nb – 0.15Ti – 0.5Mo
0.6 Nb – 0.15Ti – 0.25Mo
(FeCr)2(Nb,Ti)
(FeCr)2(Nb,Ti,Mo)
(FeCr)2(Nb,Ti,Mo)
0.62
2.40
1.70
Solvus temperature
(°C)
780
950
900
‡‡‡
Columbus Stainless MPO Number 3533603- failed during production
Columbus Stainless MPO Number 3658671 – didn’t fail
****
Columbus Stainless MPO Number 3631171
§§§
P a g e | 80
5.4
POSSIBLE STABLE PHASES AT EQUILIBRIUM
The steels in Table 4.1, apart from the primary Fe and Cr alloying elements, also contain
carbon, nitrogen, titanium and niobium as common secondary alloying elements.
Therefore, the precipitates of Nb(CN) and Ti(CN) and the Laves phase can, in principle,
be formed.
Steels A, B and C contain mainly titanium and niobium as the secondary alloying
elements; therefore, it is expected that the stable second phases will include the Laves
phase Fe2M (with M representing mainly Nb) and the carbonitrides Nb(CN)/Ti(CN). The
M6C carbide (in which, ‘M’ stands for metallic elements) is not expected to be formed or,
alternatively, if present its quantity will be negligible. In steels C and E that contain
titanium, niobium and molybdenum as the secondary alloying elements, therefore, it is
expected that the stable second phases will be Laves phase, the Nb(CN)/Ti(CN)
carbonitrides and possibly the M6C carbide. The composition of the Laves phase will
change from one alloy to another.
The phases allowed for in these calculations are listed in Table 5.4. The calculations
were performed for the temperature range of 600 to 1500 °C. Also, the data for the
intermetallic Laves phase are less reliable due to a scarcity of experimentally
determined data. It is necessary to consider the existence of the sigma phase even
though it is not expected to be formed in practice at the early stages of processing, that
is, sigma phase if it happens to be formed in this alloy, is only expected to nucleate at
higher service temperatures and after a long service period.
The possibility of a
miscibility gap for the Cr-rich phases (α+α′) was also not introduced in the calculations.
Table 5.4. Prospective phases for equilibrium calculations using Thermo-Calc® with a
TCFE3 database for steels.
Steel
Alloy
Phases
A
Type 441
Liquid, ferrite, carbo-nitrides, sigma, Laves phase, M6C
B
Type 441
Liquid, ferrite, carbo-nitrides, sigma, Laves phase, M6C
C
Nb–Ti
Liquid, ferrite, carbo-nitrides, sigma, Laves phase, M6C
D
Nb–Ti– Mo
Liquid, ferrite, carbo-nitrides, sigma, Laves phase, M6C
E
Type 444
Liquid, ferrite, carbo-nitrides, sigma, Laves phase, M6C
P a g e | 81
5.5
PHASE DIAGRAMS
The isopleth diagrams for the phases that were expected to be stable and at equilibrium
over a wide range of carbon content in these types of materials, are shown in Figure 5.1.
These calculations are based on the fixed chemical composition of the steels studied in
this work. A line illustrates where a new phase appears or disappears, and a number
representing a phase is located on the side of the line where the phase is expected to be
unstable.
With type 441 stainless steel (marked with an arrow in Figure 5.1); the equilibrium
phases that are formed over a wide range of temperatures are the Laves phase, titanium
and niobium carbo-nitrides (FCC_A1#2) and the sigma phase. The sigma phase should
not be given much consideration in this particular work since it has often been reported
that this phase only precipitates after a very long period at a high service temperature
[25,26,27]. Note that the carbonitrides Ti(CN)/Nb(CN) are precipitated from the melt,
that is their solvus temperatures are higher than the liquidus temperature in these alloys.
This result predicts that the steel is fully ferritic (BCC_A1#1) up to the liquidus
temperature (line 6, Liquid). The austenitic phase will only be formed at carbon contents
above 0.09 wt%C (FCC_A1 #1). Both the solvus temperature and the volume fraction of
each phase in these alloys are dependent on their respective chemical compositions.
Thermo-Calc® calculations show that the formation of Laves phase is more favourable
at a lower carbon content, that is below 0.1 wt.%; and that above this level of carbon and
based on the fixed chemical composition of these alloys, the Laves phase will not be
formed. Also, note that the M23C6 carbide is not formed in the presence of the Laves
phase, that is, in these alloys the precipitation of the M23C6 carbide is only limited to an
alloy with a high carbon content where the Laves phase does not exist.
P a g e | 82
1600
Laves phase
Sigma phase
M23C6
BCC_A2
FCC_A1#1
Liquid
FCC_A1#2
Liquid
BCC, α-Fe
1400
Temperature (°C)
AISI type 441
1200
FCC, γ-Fe
(Ti,Nb)(C,N)
1000
M23C6
800
Laves phase
Sigma phase
600
0.0
0.1
0.2
0.3
0.4
0.5
Weight percent C
Figure 5.1. Thermo-Calc® calculation of the isopleth diagram for type 441 stainless steel
with a constant amount of alloying elements and 0 to 0.5 wt.% of carbon. Below
any line, these represents the stable region for the phase.
The isopleth diagram for the stable phases at equilibrium in the high Mo-containing AISI
type 444 ferritic stainless steel is shown in Figure 5.2. According to Fujita et al [5], they
have observed that Mo additions to these alloys do slow down the diffusivity of Nb.
Therefore, the precipitation kinetics of Fe2Nb will be expected to be slower if
molybdenum is present.
There are certain aspects from these calculations that do not make sense, that is,
Thermo-Calc® calculations predict the presence of austenite phase even at room
temperature, but using the Fe-Cr phase diagram, and this steel is fully ferritic up to the
liquidus temperature. The more useful information about phases present in this alloy
can be obtained from the Thermo-Calc® “property diagram”, see Section 5.6 below.
1600
Laves phase
FCC_A1#1
M23C6
Liquid
Sigma phase
Liquid
Temperature (°C)
1400
FCC_A1#1
1200
M23C6
1000
800
Laves phase
Sigma phase
600
0.0
0.1
0.2
0.3
0.4
0.5
Weight percent C
Figure 5.2. Thermo-Calc® calculation of the isopleth diagram for the high Mo-containing
type 444 ferritic stainless steel E with a constant amount of alloying elements and
0 to 0.5wt.% of carbon.
P a g e | 83
5.6
PROPERTY DIAGRAMS
Some of the results obtained from these calculations are shown in Figure 5.3. in the
form of a “property diagram”, which shows the dependence of the phase proportions on
temperature. This diagram shows which phase has nucleated and grown at a specific
temperature under equilibrium conditions.
In multicomponent systems, property
diagrams are often more useful than phase diagrams, as they give information within the
phase region whereas phase diagrams give only the information of when the set of
stable phases changes. One of the good examples of the problems of Thermo-Calc® at
low temperature is that it predicts a weight faction of 25% for sigma phase, which is just
impossible.
These diagrams are similar for Steels A – E, the only difference is the phase proportion
as a function of temperature in these different steels. From this diagram the composition
of each phase over its respective existence temperature range can be calculated.
1.0
1.0
Mole fraction
0.8
0.6
0.4
0.2
Sigma phase
Laves phase
FCC_A1#2
BCC_A2
Liquid
(b)
0.8
Weight fraction
Sigma phase
Laves phase
FCC_A1#2
BCC_A2
Liquid
(a)
0.6
0.4
0.2
0.0
600
800
1000
1200
1400
Temperature (°C)
0.0
600
800
1000
1200
1400
Temperature (°C)
Figure 5.3. The property diagram that shows the dependence of phase proportion on
temperature; (a) mole fraction of stable phase and (b) weight fraction of stable
phase.
5.7
RELATIVE PHASE STABILITIES
The relative stability of a phase can be deduced by removing existing phases one by
one in the equilibrium calculations. Thermo-Calc® predicts that the phase proportion of
each existing secondary phase in type 441 stainless steel is independent from one to
another, that is, removing one phase from the calculation won’t affect the amount of the
other phase. Therefore, the removal of the sigma phase from the calculations will not
affect the proportion of the carbo-nitrides or Laves phases. The temperature range of
P a g e | 84
interest will be from 600°C to 950 °C, whereby it was previously determined that this is
the stable region for Laves phase precipitation in these steels.
AISI Type 441: Steel A
Figure 5.4 shows the weight fraction of the equilibrium phases in steel A with
composition 0.444Nb-0.153Ti. Figure 5.4 (a) shows the Laves phase weight fraction,
which is estimated to be about 0.92 wt% at 600 °C and this fraction decreases with
increasing temperature. The solvus temperature of Laves phase is estimated to be
about 825 °C in this steel, and this is in broad agreement with results from the literature
for this type of 441 stainless steel [3,9,99]. Figure 5.4 (b) is the weight fraction of the
(Ti,Nb)(C,N), which is estimated to be less than 0.12 wt%. The solvus temperature of
the (Ti,Nb)(NC) is predicted to be about 1480 °C. Note that the weight fraction of the
(Ti,Nb)(NC) first increases with temperature up to about 825 °C before it decreases
down to zero at 1480 °C. This increase below 825 °C is most likely due to the release of
Nb from the dissolving Laves phase with the released Nb increasing the thermodynamic
driving force for more (Ti,Nb)(C,N) to form. In this sense, there seems to be an indirect
dependence between the two separate phases.
(a)
0.010
(b)
0.0014
0.0012
Weight fraction
Weight fraction
0.008
0.006
0.004
0.0010
0.0008
0.0006
0.0004
0.002
0.0002
0.000
600
650
700
750
Temperature(C)
800
850
0.0000
600
800
1000
1200
1400
Temperature (°C)
Figure 5.4. Thermo-calc® plots of weight fraction of the stable phases as a function of
the temperature in the Steel A with composition 0.444Nb-0.153Ti; (a) Laves phase
and (b) (Ti,Nb)(CN).
AISI type 441: Steel B
Figure 5.5 shows the weight fraction of the equilibrium phases in Steel B with
composition 0.445Nb-0.149Ti. This steel is very similar to Steel A except for more than
P a g e | 85
double nitrogen content, i.e. 0.021%N versus the lower 0.0085%N in Steel A. Figure 5.5
(a) shows the Laves phase’s weight fraction, which is estimated to be about 0.84 wt% at
lower temperatures and this decreases with increasing temperature.
The solvus
temperature of the Laves phase in this steel is about 800 °C, i.e. about 25°C lower than
in Steel A.
Figure 5.5 (b) shows the weight fraction of the (Ti,Nb)(C,N), and it is
estimated to reach a maximum of 0.18 wt% at about 800 – 850 °C. This weight fraction
is higher than the 0.12 wt% of steel A for this carbo-nitride, most likely due to the higher
nitrogen content in steel B. The solvus temperature of the TiNb (NC) is about 1550 °C.
According to Thermo-Calc®, there is not much difference in the Laves phase weight
fractions between Steels A and B, but the major difference is in the weight fraction of
(Ti,Nb)(C,N), which is dictated by the carbon and increased nitrogen content level.
(a)
0.010
0.0020
(b)
0.008
Weight fraction
Weight fraction
0.0015
0.006
0.004
0.0005
0.002
0.000
600
0.0010
650
700
Temperature(C)
750
800
0.0000
600
800
1000
1200
1400
Temperature (°C)
Figure 5.5. Thermo-calc® plots of weight fraction of the stable phases as a function of
the temperature in the Steel B with composition 0.445Nb-0.149Ti; (a) Laves phase
and (b) (Ti,Nb)(CN)).
Nb-Ti Alloy: Steel C
Figure 5.6 shows the weight fraction of the equilibrium phases in steel C with a
composition of 0.36Nb-0.171Ti. Figure 5.6 (a) shows the Laves phase weight fraction,
which is estimated to be about 0.57 wt% at 600 °C, and the weight proportion decreases
with increasing temperature. The solvus temperature of the Laves phase is estimated to
be about 760°C in this case. This predicts that the phase proportion and the solvus
temperature of the Laves phase are dependent on the niobium content, that is, by
lowering the niobium content to 0.36 %wt as compared to 0.444 %wtNb in Steel A, both
P a g e | 86
the solvus temperature and the weight fraction will decrease, with the Laves phase’s
solvus temperature now as low as about 760°C.
Figure 5.6 (b) shows the weight fraction of the (Ti,Nb)(C,N), which is estimated to be
about 0.24 wt% maximum at the temperatures of about 800 – 850 °C and its solvus
temperature is predicted to be about 1525°C. From Table 4.1, it can be seen that this
steel has a higher carbon and nitrogen content of 0.023wt% and 0.024wt% respectively
than for any of the other steels studied.
This means that most of the carbon and
nitrogen had precipitated out as carbo-nitrides, resulting in a high content of
(Ti,Nb)(C,N). Both carbon and nitrogen have a significant impact on the fraction of
(Ti,Nb)(C,N), and this reduces the amount of Nb available to have contributed to the
solid solution strengthening of the steel at high temperatures and also the weight fraction
of Laves phase at lower temperatures. Comparing Figure 5.4(a) and Figure 5.5(a), it
can be seen that by lowering the Nb content from 0.444wt% as in Steel A to 0.36wt% as
in Steel C, this also lowers both the Laves phase content and its solvus temperature.
This shows that the Nb content in this steel plays a more significant role in the
precipitation of Laves phase than the Ti does.
(a)
0.006
(b)
0.0025
0.005
Weight fraction
0.0020
Weight fraction
0.004
0.003
0.002
0.0010
0.0005
0.001
0.000
600
0.0015
620
640
660
680
700
720
740
760
Temperature (C)
0.0000
600
800
1000
1200
1400
Temperature (°C)
Figure 5.6. Thermo-calc® plots of weight fraction of the stable phases as a function of
the temperature in the Steel C with composition 0.36Nb-0.171Ti; (a) Laves phase
and (b) (Ti,Nb)(C,N).
Nb-Ti-Mo Alloy: Steel D
Figure 5.7 shows the weight fractions of the equilibrium phases in steel D with
composition 0.39Nb-0.171Ti-0.54Mo.
Figure 5.7 (a) shows the Laves phase weight
fraction, which is estimated to be about 1.11 wt% at 600 °C, and this weight fraction
P a g e | 87
decreases with increasing temperature. The solvus temperature of the Laves phase in
this steel is estimated to be about 810 °C. The addition of 0.54%wt molybdenum has a
significant impact on the predicted weight fraction of the Laves phase, that is, Mo
additions increase the weight fraction of the Laves phase, but do not have a significant
impact on the solvus temperature of the Laves phase.
Figure 5.7 (b) shows the weight fraction of the (Ti,Nb)(C,N) phase, which is estimated to
reach a maximum of 0.18 wt% at about 800 to 850 °C, and its solvus temperature is
about 1520 °C. Also, the increase in both carbon and nitrogen levels has a significant
impact on the (T,N)(CN) weight fraction whereas the level of titanium and niobium have
no major impact on this value.
(a)
(b)
0.012
0.0020
0.010
Weight fraction
0.0015
Weight fraction
0.008
0.006
0.004
0.0010
0.0005
0.002
0.000
600
650
700
Temperature(C)
750
800
0.0000
600
800
1000
1200
1400
Temperature(°C)
Figure 5.7. Thermo-calc® plots of weight fraction of the stable phases as a function of
the temperature in the Steel D with composition 0.36Nb-0.171Ti-0.54Mo; (a) Laves
phase and (b) (Ti,Nb)(C,N).
AISI type 444: Steel E
Figure 5.8 shows the weight fraction of the equilibrium phases in steel E with
composition 0.251Nb-0.106Ti-1.942Mo and this steel falls under the category of a steel
commonly referred to as a high molybdenum, high chromium steel; usually with a
composition of 19Cr-2Mo. Figure 5.8 (a) shows the weight fraction of Laves phase
expected to be about 1.54 wt% at 600 °C, and this also decreases with increasing
temperature. The solvus temperature of the Laves phase in this steel is estimated to be
about 780 °C. The addition of 1.942%wt molybdenum is predicted to have a significant
impact on the weight fraction of the Laves phase, that is, Mo additions increase the
weight fraction of the Laves phase, but at the same time they do not increase its solvus
P a g e | 88
temperature. This shows that only niobium additions have a significant impact on both
the weight fraction and the solvus temperature of the Laves phase.
Figure 5.8 (b) shows the weight fraction of the (Ti,Nb)(C,N), which is estimated to be
about 0.155 wt% maximum at about 800 °C. The solvus temperature of the TiNb (NC) is
about 1500 °C.
(a)
(b)
0.018
0.0014
0.016
0.0012
Weight fraction
Weight fraction
0.014
0.012
0.010
0.008
0.006
0.0010
0.0008
0.0006
0.0004
0.004
0.0002
0.002
0.000
600
0.0016
650
700
750
0.0000
600
Temperature (C)
800
1000
1200
1400
Temperature (°C)
Figure 5.8. Thermo-calc® plots of the weight fraction of the stable phases as a function
of the temperature in the Steel E with composition 0.251Nb-0.106Ti-1.942Mo; (a)
Laves phase and (b) (Ti,Nb)(C,N).
5.8
EQUILIBRIUM CHEMICAL COMPOSITION OF THE LAVES PHASE
The chemical composition of a precipitate shifts towards the equilibrium composition
during growth in which local equilibrium is maintained at the interface. Coarsening may
then occur at a slower rate. The chemical compositions of the equilibrium Laves phase
as a function of temperature are shown in Figure 5.9 – 5.13. The summary of the
normalised weight fractions of the elements that form the Laves phase’s composition at
their respective temperatures, are shown in Table 5.6.
AISI type 441: Steel A
Figure 5.9 (a) shows the composition of the Laves phase (in mole fractions) which is
mainly composed of Fe and Nb in the ratio of about 2:1, with less than 0.1% mole
fraction of other alloying elements present within its composition, Ti being the major one
at lower temperatures and Cr entering this phase at higher temperatures. Therefore, it
can be concluded that the composition of the Laves phase is given by (Fe,Cr)2(Nb,Ti).
P a g e | 89
(a)
Fe
Nb
Cr
Ti
Mole fraction
0.8
0.6
0.4
0.2
0.0
600
650
700
750
800
Temperature (°C)
(b)
Weight fraction
0.8
Fe
Nb
Cr
Ti
0.6
0.4
0.2
0.0
600
650
700
750
800
Temperature (°C)
Figure 5.9. The normalised chemical composition of the Laves phase in Steel A: (a) is the
mole fraction and (b) is the weight fraction of a component in the phase.
AISI type 441: Steel B
Figure 5.10 shows the composition of the Laves phase which consists mainly of Fe and
Nb in the ratio of about 2:1, with less than 0.1% mole fraction of other alloying elements
present within its composition, Cr being the major third element in this steel and with the
Ti presence that is minimal in this steel as compared to Steel A. Therefore, it can be
concluded that the composition of the Laves phase in this steel is also given by
(Fe,Cr)2(Nb,Ti).
P a g e | 90
(a)
Fe
Nb
Cr
Ti
Mole fraction
0.8
0.6
0.4
0.2
0.0
600
650
700
750
800
Temperature (°C)
(b)
Fe
Nb
Cr
Ti
Weight fraction
0.8
0.6
0.4
0.2
0.0
600
650
700
750
800
Temperature (°C)
Figure 5.10. The normalised chemical composition of the Laves phase in Steel B: (a) is
mole fraction and (b) is a weight fraction of a component in a phase.
Nb – Ti alloy: Steel C
Figure 5.11 (a) shows the composition of the Laves phase in this steel, which mainly
consists of Fe and Nb in the ratio of 2:1, with less than 0.05% mole fraction of other
alloying elements present within its composition. Ti does not play a significant role in the
composition of the Laves phase as was also the case in Steel A, but the Cr does plays a
significant role. Therefore, it can be concluded that the composition of the Laves phase
is predicted to be (Fe,Cr)2Nb in this steel.
P a g e | 91
Mole fraction
(a) 0.8
Fe
Nb
Cr
Ti
0.6
0.4
0.2
0.0
600
620
640
660
680
700
720
740
760
Temperature (°C)
(b)
Fe
Nb
Cr
Ti
Weight fraction
0.8
0.6
0.4
0.2
0.0
600
620
640
660
680
700
720
740
760
Temperature (°C)
Figure 5.11. The normalised chemical composition of the Laves phase in Steel C: (a) is
the mole fraction and (b) is the weight fraction of a component in the phase.
Nb – Ti – Mo alloy: Steel D
Figure 5.12 shows the composition of the Laves phase in steel D which consists mainly
of Fe, Mo and Nb.
Note that Mo enters the Laves phase composition at lower
temperatures, and its content within the Laves phase decreases gradually with
increasing temperature whilst the Nb content, on the other hand, increases.
This
indicates that an addition of about 0.5wt.% Mo to an AISI type 441 stainless steel has a
significant impact on both the composition and the volume fraction of the Laves phase.
The composition of Laves phase changes from about 600 °C and higher as the Nb starts
to replace the Mo to some degree. Also, Cr enters this phase at higher temperatures,
therefore, it can be concluded that the composition of the Laves phase is predicted to be
(Fe,Cr)2(Nb,Mo,Ti) in this steel.
P a g e | 92
(a)
Fe
Nb
Cr
Ti
Mo
Mole fraction
0.8
0.6
0.4
0.2
0.0
600
650
700
750
800
Temperature (°C)
(b)
Fe
Nb
Cr
Ti
Mo
Weight fraction
0.8
0.6
0.4
0.2
0.0
600
650
700
750
800
Temperature (°C)
Figure 5.12. The normalised chemical composition of the Laves phase in Steel D: (a) is
the mole fraction and (b) is the weight fraction of the component in the Laves
phase.
AISI type 444: Steel E
Figure 5.13 shows that the composition of the Laves phase in Steel E consists mainly of
Fe and Mo, with Nb entering this phase at higher temperatures. Cr forms a minor
content of this phase. Therefore, it can be concluded that the composition of the Laves
phase in this steel is predicted to be (Fe,Cr)2(Nb,Mo) with Nb and Fe being the major
elements up to about 700°C and Fe and Mo the major elements above that temperature.
Also, from this steel it is observed that although a Mo addition increases the Laves
phase’s content it also lowers it solvus temperature. Fujita et al. [99] have observed this
enhancement of the Laves phase precipitation by a Mo addition, and they have
P a g e | 93
suggested that a Mo addition can slow the precipitation rate of the Laves phase by
lowering the diffusivity of Nb.
(a)
Fe
Nb
Cr
Mo
Mole fraction
0.8
0.6
0.4
0.2
0.0
600
620
640
660
680
700
720
740
760
780
Temperature (°C)
(b)
Fe
Nb
Cr
Mo
Weight fraction
0.8
0.6
0.4
0.2
0.0
600
620
640
660
680
700
720
740
760
780
Temperature (°C)
Figure 5.13. The normalised chemical composition of the Laves phase in Steel E: (a) is
the mole fraction and (b) is the weight fraction of a component in the Laves phase.
5.9
DRIVING FORCE FOR NUCLEATION
The Gibbs free energy changes for the Laves phase formation in the five steels were
also calculated using Thermo-Calc®. Figure 5.14 shows the relationship of the free
energy change for the Laves phase reaction in these steels as a function of temperature
and for equilibrium states. The free energy change increases (becomes less negative)
as the temperature increases from 600°C towards the solvus temperature, and its values
are dependent on the composition of the specific alloys.
P a g e | 94
The calculated values are not the driving force for nucleation ∆GV but rather the overall
Gibbs free energy change ∆G for the reaction. With these results, the driving force for
nucleation ∆GV can be obtained from:
∆G =
ν
∆G
υV
Equation 5.1
where V is the equilibrium volume fraction of the Laves phase and υ is the molar volume
of the Laves phase.
P a g e | 95
(a)
0
0
(b)
-100
G (J/mol)
G (J/mol)
-100
-200
-300
-200
-300
-400
-500
600
650
700
750
-400
600
800
650
Temperature (°C)
(c)
700
750
800
Temperature (°C)
0
0
(d)
-50
-100
G (J/mol)
G (J/mol)
-100
-150
-200
-300
-200
-400
-250
-300
600
620
640
660
680
700
720
740
-500
600
760
650
Temperature (°C)
(e)
700
750
800
Temperature (°C)
0
-100
G (J/mol)
-200
-300
-400
-500
-600
600
650
700
750
Temperature (°C)
Figure 5.14. The free energy change ∆G for the precipitation reaction of Laves phase in ferrite
with temperature for : (a) Steel A; (b) Steel B; (c) Steel C; (d) Steel D and (e) Steel E,
calculated using Thermo-Calc®, (G = J/mol).
5.10 SUMMARY
The thermodynamic equilibrium data have been generated for AISI type 441 ferritic
stainless through Thermo – Calc® software, see Table 5.5 for a summary table and
Table 5.6 for a detailed comparison.
These data will be used to determine the
P a g e | 96
nucleation kinetics in the alloys. The results obtained from this model will be compared
directly to the experimental results.
Hopefully, a better understanding of the Laves
phase precipitation will be possible, and the effects of composition on its kinetics will be
predicted.
P a g e | 97
Table 5.5. A summarised results from the Thermo-Calc® predictions.
Steel
Main secondary
alloying elements
Composition differences
with reference to steel A
0.012C-0.0085N0.444Nb-0.153Ti∼0 Mo
Laves phase
(Ti,Nb)(C,N)
%Wt frac.
at 600 °C
Solvus
temp °C
Molar composition
%Wt frac
at 600 °C
Solvus
temp °C
Main reference steel
0.920
825
(Fe,Cr)2(Nb,Ti)
0.108
1480
Increased N content but
C, Nb and Ti relatively
unchanged
0.840
800
(Fe,Cr)2(Nb,Ti)
0.173
1550
Increased (C+N), lower
Nb and higher Ti content
0.573
764
(Fe,Cr)2(Nb)
0.225
1525
D
0.012C-0.026N0.39Nb-0.171Ti0.54Mo
Lower C ad Nb content, N
and Ti relatively
unchanged, and Mo
addition
1.111
815
(Fe,Cr)2(Nb,Mo,Ti)
0.176
1520
E
0.014C-0.016N0.251Nb-0.106Ti1.942Mo
C relatively unchanged, N
increased, lower Nb + Ti
contents, high Mo content
addition
1.542
780
(Fe,Cr)2(Nb,Mo)
0.150
1500
A
B
C
0.015C-0.021N0.445Nb-0.149Ti0.008Mo
0.023C-0.024N0.36Nb-0.171Ti<0.01Mo
Conclusions
Increased weight fraction
of (Ti,Nb)(C,N) from
increased N content
Laves phase solvus
decreases with lower Nb
content
Laves phase content
increased with the
addition of Mo.
(Ti,Nb)(C,N) content and
solvus temp increased
from increased N content
Laves phase content
increases with the Mo
addition. Its solvus temp
decreased with lower Nb
content
P a g e | 98
Table 5.6. Detailed results of thermodynamic calculations (Thermo-Calc®). The normalised weight and mole fractions of the Laves
phase composition components are shown.
Steel
Solvus temp
(°C)
A
825
B
800
C
760
D
810
D
780
Temp
(°C)
600
650
700
750
775
800
600
650
700
750
775
600
650
700
750
600
650
700
750
800
600
650
700
750
∆G
J/mol
-425
-395
-325
-230
-125
-90
-373
-330
-270
-150
-80
-259
-218
-150
-46
-475
-405
-325
-175
-50
-555
-463
-350
-150
Wt frac.
(%)
0.920
0.817
0.617
0.413
0.283
0.142
0.840
0.607
0.520
0.273
0.160
0.573
0.464
0.300
0.100
1.111
0.861
0.639
0.333
0.056
1.542
1.231
0.807
0.292
Concentration in Laves phase ppt (wt)
Fe
Nb
Cr
Ti
Mo
0.549
0.398
0.020
0.033
0.542
0.411
0.022
0.022
0.533
0.420
0.024
0.022
0.522
0.424
0.024
0.010
0.522
0.426
0.024
0.010
0.522
0.433
00.24
0.009
0.533
0.439
0.044
0.022
0.524
0.440
0.044
0.044
0.524
0.444
0.044
0.044
0.522
0.444
0.044
0.044
0.522
0.444
0.044
0.044
0.524
0.444
0.020
0.524
0.444
0.020
0.524
0.444
0.027
0.524
0.444
0.027
0.539
0.320
0.020
0.013
0.122
0.533
0.333
0.020
0.011
0.107
0.528
0.364
0.020
0.009
0.078
0.524
0.378
0.020
0.009
0.056
0.520
0.394
0.020
0.009
0.044
0.522
0.139
0.320
0.522
0.162
0.300
0.518
0.183
0.267
0.518
0.242
0.213
Concentration in Laves phase ppt (mole)
Fe
Nb
Cr
Ti
Mo
0.642
0.285
0.029
0.050
0.637
0.292
0.033
0.055
0.633
0.300
0.033
0.055
0.633
0.308
0.033
0.055
0.633
0.308
0.033
0.055
0.633
0.310
0.033
0.055
0.633
0.317
0.030
0.008
0.633
0.317
0.030
0.008
0.632
0.325
0.033
0.008
0.632
0.325
0.033
0.008
0.632
0.325
0.033
0.008
0.633
0.325
0.029
0.633
0.325
0.029
0.630
0.325
0.033
0.630
0.325
0.033
0.646
0.233
0.017
0.017
0.079
0.642
0.242
0.032
0.013
0.067
0.642
0.267
0.033
0.008
0.054
0.633
0.283
0.033
0.008
0.042
0.633
0.292
0.033
0.008
0.030
0.650
0.100
0.025
0.232
0.646
0.117
0.025
0.217
0.642
0.138
0.029
0.200
0.638
0.175
0.029
0.154
P a g e | 99
CHAPTER SIX
EXPERIMENTAL RESULTS
6
6.1
INTRODUCTION
This chapter examines collectively the experimental results. The main focus is on the
Laves phase formation and its dissolution and how it affects the mechanical properties
of the AISI type 441 ferritic stainless steel.
To gain a better understanding of the
embrittlement effect from the Laves phase, the results are divided into three parts:
1. microstructural analysis from the failure of materials to understand the source of
this embrittlement;
2. effect of the annealing treatment on the Laves phase precipitation and, therefore,
its embrittling effect; and
3. the kinetics of the Laves phase formation, and the parameters (such as,
temperature and composition) that affect its formation.
6.2
MICROSTRUCTURAL ANALYSIS OF AN AISI TYPE 441 FERRITIC STAINLESS STEEL
In order to understand the source of embrittlement in this steel, microstructural analyses
were carried out.
These analyses range from simple optical microscopy to more
complex transmission electron microscopy analysis. The analyses were carried out on
the ferritic stainless steel AISI type 441, and particularly on Steel A that had failed during
processing. As it has previously been determined by Columbus Stainless, the source of
embrittlement was apparently not grain size related, as the grain size was measured to
be 24.8 ± 4.2 µm which is still considered to be fine enough not to have embrittled this
material, see Figure 6.1(a). The general grain microstructure had a pan cake structure
with the grains elongated in the rolling direction. The scanning electron microscopy
(SEM) micrographs show the presence of the grain boundary and matrix precipitates,
see Figure 6.1(b).
P a g e | 100
(a)
(b)
Figure 6.1. Micrographs from Steel A in the as received hot rolled condition, showing the
grain structure. (a) optical microscopy image and (b) SEM images. Note the large
difference in magnification with figure (a) showing the “particle decorated” grain
structure while figure (b) shows primarily the “particle decorated” subgrain
structure.
6.2.1 PRECIPITATE’S IDENTIFICATION
The morphology of the various precipitates in this Steel A changes with the choice of
annealing treatment, as revealed by the two types of precipitates (that is, by relative
size) that was observed from the SEM micrographs in Figure 6.2.
Only the large
particles could be analysed successfully using the SEM energy dispersive X-ray
technique while the small precipitates were too fine for even a representative semiquantitative analysis. The large precipitates were found to be most likely a titanium
carbo–nitride, which had varying degrees of smaller niobium carbo–nitride precipitates
clustered around it and the central precipitate’s size was in the region of 2 to 10 µm, see
Figure 6.3. According to Gordon and van Bennekom [55], they have reported that the
colour of the carbo – nitride changes with the ratio of N/C in the steel. A yellow/orange
colour indicates a Ti(C,N) as found in this steel that contains about 10 at. % TiC, see
Figure 6.2(a).
P a g e | 101
(a)
(b)
Figure 6.2. Micrographs of the as received hot rolled Steel A showing its grain structure.
(a) An optical microscopy image and (b) a SEM image.
Figure 6.3. SEM – EDS micrograph showing a precipitate consisting of a central cubic
core of a mainly titanium containing particle surrounded by a cluster of niobium
precipitates.
The SEM results confirm the predictions made by Thermo-Calc®, in that both nitrides
and carbides are formed as complex FCC (Ti,Nb)(C,N) precipitates with the Ti(C,N)
nucleating first and then the Nb(C,N) nucleating on its surface as a shell or alternatively
as a cluster of loose particles around the Ti(C,N). This is not surprising as Ti(C,N) is
P a g e | 102
known to have a lower solubility in these steels than Nb(CN) and will, therefore, form
first during cooling from the melt while the Nb(CN) subsequently forms at slightly lower
temperatures by heterogeneous nucleation on the existing Ti(CN). Furthermore, in the
work done by Craven et al.[130], they also observed that the (Ti,Nb)(C,N) complex has a
least soluble core consisting of TiN that precipitates at higher temperatures during
cooling from the melt, and the Nb/Ti ratio in the core is markedly affected by the cooling
rates during solidification and later thermomechanical processing.
The small precipitates, however, were analysed using transmission electron microscopy
(TEM) and XRD techniques. To be able to analyse the types of precipitates present in
this steel without any interference from the matrix, carbon extraction replicas of the
precipitates were used and TEM energy dispersive X-ray spectrometer (TEM-EDS) and
electron diffraction patterns were used in identifying these precipitates.
The
transmission electron microscopy (TEM) micrographs for the as received hot rolled Steel
A are shown in Figure 6.3.
Three types of precipitates were analysed from this
specimen;
1. large plate–like precipitates (analysis 1);
2. needle – like precipitates (analysis 2); and
3. small plate-like precipitates (analysis 3),
while there were some fine globular precipitates that could not be accurately analysed
using both of these techniques because of their small size. In the work done by Fujita et
al [6] it was suggested that the intensity ratio of niobium to iron, Nb/Fe in EDS analysis
for the M6C (Fe3Nb3C) type carbide is larger than that for Laves phase (Fe2Nb). This
suggests that large plate – like precipitates (analysis 1) are the M6C carbide and both
the needle–like and small plate–like particles (analyses 2 and 3) are the Laves phase
Fe2Nb. Both M6C type carbide and Laves phase particles include small amounts of
titanium and chromium.
However, the amounts of titanium and chromium in both
precipitates are much less than those of niobium and iron. From these observations it
can be concluded that the composition of the M6C and Laves phase precipitates are
most likely to be (Fe,Cr)3(Nb,Ti)3C and (Fe,Cr)2(Nb,Ti), respectively.
P a g e | 103
Analyses 1
Analyses 2
Analyses 3
Analyses 1
(a) 2
Analyses
Analyses 3
Fe3Nb3C [ 1 11]
Fe2Nb [ 1 00]
Fe2Nb [10 2 ]
(b)
Figure 6.4. Transmission electron micrographs of particles from extraction replicas and
their analyses by electron diffraction and EDS of the as-received hot rolled Steel A
showing different particle morphologies.
The diffraction patterns of the large globular precipitates (analysis 1) show a cubic
crystal structure with a lattice parameter ao = 1.11 nm. Therefore, these precipitates are
identified as most likely to be cubic M6C carbides. On the other hand, the diffraction
patterns of the needle-like or elongated phase (analysis 2) and plate-like precipitates
(analysis 3) show the crystal structure to be most likely identified as hexagonal. The
lattice parameters were measured to be ao = 0.48 nm and co = 0.79 nm, with co/ ao =
1.64, which is close to the MgZn2 C14-type of the Laves phase.
P a g e | 104
Figure 6.5 shows the results of the XRD analyses of the precipitate residues after
electrolytic extraction from the as received material, Steel A.
Two main types of
precipitates where detected, i.e., the Laves phase and the carbo-nitrides, but there was
a small amount of M6C carbides also present, as was also found in the TEM analysis.
From Figure 6.5 , the peaks for the M6C phase is not clearly visible, but when the
Rietveld refinement approach was used to quantify these phases, its presence was
detected as a small amount. Note, however, that it is not always possible to get a fully
pure precipitate extraction of precipitates without the interference of the matrix material,
the α-Fe matrix phase is considered to be also present in this measurement. Table 6.1
shows the weight fractions of the precipitates present in this steel, with the error
calculated as a 3 sigma error (meaning three times the standard deviation). The volume
was calculated using the densities of each phase and the total mass of precipitates’
residue extracted from the steel by weighing. The results indicate that the normalised
weight fractions for the Laves phase and (Ti,Nb)(C,N) to be about 1.14% and 0.33%,
respectively.
Using the Rietveld refinement approach, the lattice parameters of the Laves phase were
determined to be ao= 0.481 nm and co = 0.784 nm and these values are close to the
ones determined above from the TEM diffraction patterns.
Table 6.1. The measured weight fractions of the precipitate phases found in the as
received hot – rolled material, Steel A.
Phase
Fe2Nb
M6C
α - Fe
(Ti,Nb (C,N)
XRD analysis
Weight
3σ
fraction (%)
56.36
2.04
2.02
0.66
0.84
16.24
25.39
2.52
Weight
phase (g)
0.037
0.001
0.011
0.017
Calculated values
Normalised
Density
weight frac. (%)
(g/cm3)
1.143
8.63
0.041
8.44
0.329
7.87
0.515
7.82
Vol.
( x 10-3cm3)
1.32
0.05
0.42
0.80
P a g e | 105
400
FE2NB-LAVES
350
300
♣
I / cps
250
∗
200
♣
150
∗
∗
∗
∗
100
∗
50
0
25
30
35
40
45
50
55
2 Theta / °
60
65
70
75
80
85
Differences
200
150
I / cps
100
50
0
-50
-100
-150
-200
25
30
35
40
45
50
55
2 Theta / °
60
65
70
75
80
85
Figure 6.5. A typical XRD scan of the precipitate residue after electrolytic extraction from Steel A, i.e. the as received material,
showing the presence of the Laves phase peaks (indicated by the lines in the top figure). The remaining peaks are the
carbides and nitrides, indicated by (∗
∗) and the α - Fe matrix, indicated by (♣
♣). Note the good residual difference
between the calculated and the measured spectrum as is shown by the spectrum below.
P a g e | 106
Fujita et al., [101] using the MTDATA thermodynamic software, have modelled the
precipitation kinetics in the Fe-Nb-C system for a 9Cr-0.8Nb steel at 950 °C. They have
observed that the atomic ratio of Fe to Nb in the extracted residue in their steel
approaches unity with increasing ageing time, and this indicates that the Laves phase
(Fe2Nb), which contains more iron than required by stoichiometry in Fe3Nb3C, dissolves
to give way to Fe3Nb3C. This meant that the Laves phase was a metastable phase that
gave way to the precipitation of Fe3Nb3C, and they have suggested that the precipitation
sequence towards the equilibrium phases in their high niobium content stabilised ferritic
stainless steel to be NbN and Fe3Nb3C:
α → α + Nb(C,N) + Fe2Nb + Fe3Nb3C → α + NbN + Fe3Nb3C
where α represents ferrite. In this study on the embrittled Steel A with a lower niobium
content reported here, the Thermo-Calc® prediction did not show this reaction sequence
of Fujita et al., [101] because this sequence is a time dependent reaction, and ThermoCalc® calculations only apply to the phases at full equilibrium. Secondly, the relatively
large difference in niobium content between the two steels should also be noted. In
order to prevent a great loss in strength during service, it is necessary to control the
precipitation sequence during high temperature aging.
This can be achieved by
reducing the carbon content or by adding some other alloying elements that have a
stronger affinity for carbon, and this will impede precipitation of the coarse M6C carbides.
6.3
EFFECT
OF
ANNEALING TREATMENT
ON THE
MICROSTRUCTURAL
AND
MECHANICAL
PROPERTIES
6.3.1 MICROSTRUCTURAL ANALYSIS
Figure 6.6 shows the effect of annealing heat treatment on the grain structure evolution
from the as-received hot-rolled material, Steel A. During this process the grain size was
measured by the linear intercept method.
The microstructural analysis of the as
received material shows that full dynamic or static recrystallisation had indeed not
occurred during and after hot rolling and, therefore, there was no measurable grain
refinement from dynamic recrystallisation during hot rolling. In fact, full recrystallisation
was observed to have occurred only at temperatures above 1000 °C. With annealing at
P a g e | 107
temperatures below 1000 °C, the grain size evolution shows that only recovery had
taken place and only a minimal level of grain growth was observed.
(a)
(b)
850 °C
As received
(c)
(d)
875 °C
(e)
900 °C
(f)
950 °C
1000 °C
P a g e | 108
(g)
(h)
1050 °C
1100 °C
Figure 6.6. Optical micrographs of the specimens from Steel A after annealing at different
temperatures for 30 minutes followed by water quenching (In comparing the
microstructures, note the differences in magnifications).
The SEM and TEM results also indicate that the volume fraction of fine Laves phase
precipitates tends to decrease as the annealing temperature increases towards the
solvus (see Figure 6.7 and 6.9, respectively), as would be expected also from the
Thermo-Calc® predictions.
Both micrographs show that the Laves phase has
precipitated on the sub-grain and grain boundaries and only small amounts are found
within the grains. This indicates that the Laves phase possibly nucleates firstly at the
sub-grain and grain boundaries and then, only after site saturation, within the grain
matrix. With annealing temperatures above 900 °C, complete dissolution of the Laves
phase is observed, and the grain’s microstructure has fully recrystallised. At annealing
temperatures between 700 °C and 850 °C, the TEM micrographs show that there is a
higher volume fraction of the grain boundary Laves phase, and these precipitates have
coalesced to form stringers. It is these grain boundary precipitates that would later
result in embrittlement of this material. In the study done by Sawatani et al [8] the same
observation was also made, and they have found that the quantity of the precipitates in
their as-hot rolled material, reaches a maximum after being annealed and slowly cooled
from about 700 °C but that the volume fraction then decreases with slow cooling from
above 900 °C. They have also observed that the Laves phase dissolves into a solid
solution at an annealing temperature of over 900 °C, and this finding coincides with the
results of the XRD and TEM analyses, where Laves phase was detected only at
temperatures up to 850 °C.
P a g e | 109
(a)
(b)
850 °C
As received
(c)
(d)
875 °C
900 °C
(f)
(e)
950 °C
1000 °C
P a g e | 110
(g)
(h)
1050 °C
1100 °C
Figure 6.7. SEM micrographs of Steel A showing the effect of annealing temperature on
the morphology of the second phase.
(a)
(b)
As received
600 °C
(d)
(c)
650 °C
700 °C
P a g e | 111
(e)
(f)
750 °C
(g)
800 °C
(h)
850 °C
900 °C
Figure 6.8. TEM micrographs from Steel A showing the presence of the fine Laves phase
precipitates on the subgrain boundaries of the specimens that were annealed at
the shown different temperatures for 1 hour and then water quenched.
6.3.2 PRECIPITATE’S MORPHOLOGY
Transmission electron microscopy micrographs (TEM) of the dislocation structures in the
specimen annealed at 700 °C and then water quenched, are shown in Figure 6.9. The
precipitation of Fe2Nb particles on the dislocations and subgrain boundaries are shown
in Figure 6.9 (a & b). The halo surrounding these particles within the grains is believed
to be associated with the strain fields surrounding these particles [92]. Nucleation of the
Fe2Nb occurs firstly on grain boundaries, then at dislocations, and at a later stage within
the matrix. In the work by Li [131] on the precipitation of Fe2W Laves phase in a 12Cr2W alloy, the author observed that Laves phase particles on grain boundaries are
coherent with one grain but grow into the adjacent grain with which they do not have a
rational orientation relationship. This is due to the higher mobility of the incoherent
interface if compared to the semi- or coherent one. This may also be clearly seen in this
P a g e | 112
Steel A on the “horizontal” grain boundary in the middle of Figure 6.9 (a) where all the
Laves phase particles appear to be growing only into the top grain.
(a)
(b)
Figure 6.9. Thin foil electron transmission micrographs from steel A, annealed at 700 °C
for 1 hour and then water quenched. The micrographs show (a) the nucleation of
the Laves phase precipitates on grain boundaries and dislocations and (b) some
fine matrix precipitates surrounded by a strain halo as well as dislocation
nucleated precipitates.
6.3.3 MECHANICAL PROPERTIES
The results from the V-notch room temperature Charpy impact tests on specimens
solution treated at different temperatures and water quenched of the process embrittled
Steel A, are shown in Figure 6.10. The Charpy Impact Energy (CIE) shows a maximum
of about 60 J for the specimens that had been annealed at 850 °C, which coincides with
the predicted Laves phase solvus temperature, but with a decreasing CIE with annealing
on both sides of this temperature. In the specimens that had been annealed at 900 °C
and above where no Laves phase is present, the impact energy averaged only about 10
J. It can be concluded that above 850 °C, a grain size effect from grain growth plays a
major role in lowering the impact energy, but below 850 °C, the precipitation of the
intermetallic Laves phase contributes to the lowering of the impact toughness.
Comparing these results with the prediction from Thermo-Calc (Section 5.7), it can be
seen that as the volume fraction of the Laves phase decreases in the steel with a rise in
the annealing temperature towards 850 °C, the impact toughness also increases to its
maximum of 60 J at 850 °C.
Beyond 850 °C grain growth plays a major role in
embrittling this ferritic steel, a problem that is well known in most ferritic stainless steels.
P a g e | 113
Figure 6.10. Effect of annealing temperature on the room temperature Charpy impact
energy of the as hot rolled and annealed AISI 441 stainless Steel A. The samples
were annealed for 30 minutes and then water quenched.
Typical fracture surfaces from specimens after annealing at different temperatures are
shown in Figure 6.11. In Figure 6.11 (a) the micrograph is shown of the as received and
process embrittled Steel A with a CIE of only slightly above 10 J, with the fracture
surface showing both macro- and micro-voids in a dimpled surface, which is
characteristic of a ductile failure although the CIE was actually relatively low.
The
(Ti,Nb)(C,N) particles are readily identifiable as “blocky” precipitates, as clearly seen in
Figure Figure 6.12 (b). These macro-voids appear to be formed by the presence of
these large cuboidal particles which indicates a low cohesive force between these
carbo-nitride precipitates and the steel matrix.
These precipitates are, therefore,
possibly not the direct cause of the brittleness of the alloy. The presence of the Laves
phase on the grain boundaries (see Figure 6.7 and Figure 6.8 (as received)) must then
have contributed to the brittleness of these materials, but the exact mechanism could not
be revealed from the fractured surface analysis.
Figure 6.11 (c & d) show the fractographs of the specimen that had been annealed at
850 °C where the maximum CIE was found, indicating that fracture had occurred in a
ductile manner. Noticeable, however, are the large numbers of voids situated on a few
of the grain boundaries in Figure 6.11 (d), indicating the presence of particles on some
of these grain boundaries, most likely the last remaining Laves phase.
There are,
however, no further significant or obvious differences in the observed failure mode
between the hot rolled as received material with a CIE of only about 10 J and the sample
annealed at 850 °C with a CIE of 60 J, except that there are also a few Laves phase
precipitates present on the grain boundaries of the as received steel (see Figure 6.7(a))
P a g e | 114
whereas the steel annealed at 850 °C, according to Thermo-Calc® should have had no
Laves phase in its microstructure. Previous studies have shown that the presence of the
Laves phase degrades the mechanical properties of this high chromium ferritic steel
where ever this phase is formed within the temperature range of 500 – 750 °C [132].
Figure 6.11 (e & f) show the fracture surfaces of the specimen that had been annealed
at 900 °C; which shows transgranular cleavage fracture with individual grains identified
by changes in the directions of the river markings. Figure 6.11(f) shows the specimen
displaying microcracks that are clearly visible on the fracture surface.
Evidence
suggesting cleavage or grain boundary microcracking was occasionally observed, but
little role of microstructure in the crack initiation process could be directly observed.
(T,iNb)(C,N)
(a)
(c)
(b)
(d)
P a g e | 115
(e)
(f)
Figure 6.11. Examples of the Charpy fracture surfaces at different magnifications of steel
A (a & b) from the as received specimen; (c & d) after annealing at 850 °C; and (e &
f) after annealing at 900°°C.
6.3.4 EFFECT OF GRAIN SIZE ON THE MECHANICAL PROPERTIES OF STEEL A
Figure 6.12 shows the variation of the hardness and grain size as a function of the
solution annealing temperature above 850 °C, i.e. where there should not be any Laves
phase present in the Steel A but where grain growth is expected. The results indicate
that there is a steady decrease in the Vickers hardness whereas the as-received and
brittle hot-rolled steel shows a much higher hardness than that of the specimens that
had been annealed in the grain growth region. This can not be attributed purely to the
grain growth effect since there is only a relatively small difference in grain size between
the as-received hot rolled material and the specimen that had been laboratory annealed
at 850 °C after hot rolling.
This difference in hardness may, however, arise from
accumulated strains induced during the hot rolling process in the as received steel,
particularly if full dynamic recrystallisation had not occurred.
The microstructural
analysis of the as received material shows that full dynamic recrystallisation had indeed
not occurred, and therefore, there was no measurable grain refinement from dynamic
recrystallisation, see Figure 6.13. The presence of the fine grain size and the Laves
phase could have contributed to a much higher hardness value of the as received
material; but as the steel is further annealed, recrystallisation starts to occur and thereby
softens the steel.
P a g e | 116
220
20 0
V icke rs H ardnes s
G rain s iz e
18 0
16 0
200
12 0
180
10 0
80
Grain size (µm)
Vickers Hardness
14 0
60
160
40
20
140
0
A s rec.
8 50
87 5
900
92 5
950
97 5
1000
10 25
T em pe rature (°C )
105 0
10 75
110 0
.
Figure 6.12. Effect of annealing temperature above 850 °C on the grain size and Vickers
hardness for the AISI type 441 ferritic stainless Steel A.
Figure 6.13. TEM micrograph showing the presence of a dislocation substructure and
some fine Laves precipitates in the as received hot rolled specimen of Steel A,
indicating a lack of full dynamic recrystallisation during the last stage of hot
rolling.
The results also show that there is a steady increase in grain size up to about 950 °C,
but between 950 °C and 1000 °C there is a sudden and rapid 60 % increase in the grain
size. The TEM micrographs of the specimens that were annealed at 850 °C and 900 °C
respectively, with the presence of some remnants of Laves phase in the specimen that
was annealed at 850 °C, were shown in Figure 6.8.
At 900 °C this phase had
completely dissolved, see Figure 6.8 for the microstructural evolution during annealing at
these temperatures.
Figure 6.14 shows the effect of annealing temperature at and above 850 °C on the
tensile properties of the Steel A. Both the 0.2 % offset yield strength and ultimate tensile
strength (UTS) decrease whilst the percentage elongation increases with increasing
annealing temperature. Significantly, however, the specimen that had been heat treated
P a g e | 117
at 850 °C shows a relatively poor elongation whereas at 900 °C the elongation had
already increased significantly, confirming that at 850 °C some last remaining and
embrittling Laves phase was still present while at 900 °C it had completely dissolved.
The effect of grain growth on the 0.2% yield strength in this steel was tested according
to the Hall-Petch relationship, and was found that it only applies in the temperature
range of 850 °C to 950 °C but beyond 950 °C, the relationship did not hold.
52
600
UTS
% Elongation
0.2% Yield Strength
50
46
500
44
42
400
40
% Elongation
Strength (MPa)
48
38
300
36
34
200
32
As Rec 850
875
900
925
950
975
1000
1025
1050
1075
1100
Temperature (°C)
Figure 6.14. Effect of annealing temperature at 850 °C and above on the tensile strength
and elongation of the 441 stainless steel A.
6.4
EFFECT OF ANNEALING TREATMENT ON THE CHARPY IMPACT ENERGY AND DBTT
The results of Charpy impact tests at various testing temperatures on the heat treated
specimens of this 441 ferritic stainless steel (Steel A) are shown in Figure 6.15. The
specimens were initially annealed in the temperature range of 600 to 950 °C for 30
minutes and subsequently quenched in water. The specimen that was quenched from
850 °C exhibits a maximum toughness of about 60 J at 25 °C and a Ductile Brittle
Transition Temperature (DBTT) of about 5 °C at about 30 J. Solution treatment at 950
°C resulted in the DBTT to be as high as 40 °C, in comparison with the steel treated at
850 °C while all these specimens fractured in a brittle manner. On the other hand, the
upper shelf energy for the specimen solution treated at 950 °C was found to be much
higher, averaging about 90 J.
The grain size of the heat treated samples varied from about 22 to 60 µm as the
annealing temperature was increased from 850 to 950 °C (see Figure 6.15). The DBTT
was observed to rise with increasing grain size from both specimens, but the upper shelf
energy seemed to be independent of the grain size.
This could be caused by the
P a g e | 118
presence of some Laves phase on the grain boundaries still present in the specimen
that was annealed at 850 °C, resulting in lowering the upper shelf energy to about 60 J
compared to 90 J in the specimen that was annealed at 950 °C.
100
60 0 °C
70 0 °C
85 0 °C
95 0 °C
80
CVN (J)
60
40
20
0
-8 0
-6 0
-40
-2 0
0
20
40
60
80
1 00
T em p eratu re (°C )
Figure 6.15. Charpy impact energy of the 441 ferritic stainless steel A as a function of the
test temperature from specimens that were annealed at four different
temperatures, both within and outside the Lave phase formation region.
6.5
EFFECT
OF
RE –EMBRITTLEMENT
TREATMENT ON
THE ROOM TEMPERATURE CHARPY
IMPACT ENERGY
6.5.1 EFFECT OF COOLING RATE
The effect of cooling rate on the room temperature impact toughness of the steel after
annealing at 850 °C and 950 °C respectively for 5 min in the Gleeble simulator followed
by forced helium cooling at different linear cooling rates, is shown in Figure 6.16. The
aim of this exercise was to simulate the effect of the post-hot rolling cooling rate on the
final grain size and on the precipitation of Laves phase and hence their impact on the
Charpy impact energy to provide process guidance on the importance of rapid cooling
rates after the final stage of hot rolling. As observed before, rapid cooling from 850 °C
gives a higher impact energy than after cooling from 950 °C, and this is caused initially
by a difference in grain size.
This can be observed from the specimens that were
subject to cooling rates of about 50 °C/s, where there was a difference of 34 J in the
upper shelf energy between the specimens annealed at 850 °C and 950 °C respectively,
see Figure 6.16, with results that are comparable to those in Figure 6.10.
Note
furthermore, that the impact strength after cooling from 950 °C is only marginally
affected by differences in cooling rate while this is not so after cooling from 850 °C
where faster cooling results in smaller volumes of Laves phase forming during cooling
P a g e | 119
and, hence, in higher impact strengths. This difference is significant as it proves that the
embrittling effects of a large grain size introduced by cooling from 950 °C, overrides any
further embrittlement by Laves phase caused by slow cooling. Embrittlement in these
ferritic stainless steels from a larger grain size, therefore, appears to be even more
deleterious than that of Laves phase if both are present together.
9 5 0 °C
8 5 0 °C
50
CVN (J)
40
30
Laves phase
effect
20
10
Grain size
effect
0
0
10
20
30
40
50
C o o lin g ra te (°C /s )
Figure 6.16. Effect of linear cooling rate in °C/s on the room temperature impact
toughness of the specimens from Steel A that were cooled at linear cooling rates
from 850 °C and 950 °C, respectively.
In relation to the observed toughness behaviour, the Laves phase precipitation
behaviour was considered. The metallurgical observation from the two samples that
were solution treated at these temperatures and cooled at 60 °C/s, shows two
completely different microstructures, see Figure 6.17. After solution treatment at 850 °C,
the microstructure has not recrystallised and the electron micrograph shows the
presence of a subgrain structure in Figure 6.17 (a & b). Solution treatment at 950 °C,
however, shows that the microstructure has now fully recrystallised and is without any
subgrain structure in Figure 6.17 (c & d). In both cases, there are few large particles still
present that were analysed as niobium carbo-nitride, Nb(C,N) using TEM EDX.
Figure 6.18 shows the effect of the cooling rate on the precipitation of the Fe2Nb Laves
phase after cooling at a rate of 1 °C/s.
Figure 6.18 (a & b) show a band of fine
precipitates from the samples that were solution treated at 850 °C and 950 °C and then
cooled slowly, respectively. Note that these fine precipitates are found on the grain
boundaries in the sample that was solution annealed at 850 °C, whereas in the sample
that was solution treated at 950 °C, these precipitates where observed within the grains.
P a g e | 120
This suggests that the Laves phase can precipitate homogeneously in a coarse grain
microstructure. The volume fraction of the Laves phase in the steel decreases with
increasing cooling rates, as seen in Figure 6.17 and Figure 6.18.
From this
investigation, the quantitative work on the amount of the Laves phase precipitated from
the two solution annealing temperatures was then carried forward.
(a)
(c)
(b)
(d)
Figure 6.17. TEM micrographs of the samples of Steel A that were solution annealed at
850 °C and 950 °C for 5 min then cooled at 60 °C/sec. (a & b) solution treated at
850 °C; (c & d) solution treated at 950 °C. Note the differences in the
microstructures from both samples.
(a)
(b)
Figure 6.18. TEM micrographs of the samples from Steel A after being cooled at 1 °C/sec
from: (a) solution annealed at 850 °C and (b) 950 °C for 5 min before cooling.
The effect of the cooling rate on the amount of Laves phase precipitated from the
specimen annealed at 850 °C is shown in Figure 6.19. As the cooling rate increases,
the amount of the Laves phase formed decreases and the Charpy impact value
P a g e | 121
increases (see Figure 6.16). This serves as evidence showing that the volume fraction
of the Laves phase plays an important role in embrittling this steel. In the work by
Sawatani et al.[8], they have also found that a large amount of fine Laves phase
nucleates first at the grain boundaries and then within the grains during cooling. As the
cooling rate becomes lower, the amount of Laves phase that forms increases, as would
be expected. These large amounts of precipitates were found to increase the strength
and decrease the elongation [8,69], whilst also decreasing the toughness of the steel.
The toughness and ductility can be recovered by heating the material again to above
900 °C to dissolve the Laves phase, followed by rapid cooling.
0.0035
0.0030
Volume fraction
0.0025
0.0020
0.0015
0.0010
0.0005
0.0000
0
10
20
30
40
50
Cooling rate (°C/s)
Figure 6.19. Effect of the cooling rate on the volume fraction of the Laves phase in Steel
A after cooling at different rates from annealing at 850°°C.
6.5.2 EFFECT OF THE REHEATING TREATMENT
The results of the Steel A samples that were reheated at 600 °C to 900 °C for 30
minutes after being solution treated at 950 °C for 1 hour, are shown in Figure 6.20. The
aim of these tests was to dissolve the Laves phase completely and then re–precipitate
them at the different annealing temperatures.
After the reheating treatment, all the
samples show a significant decrease in both ductile-to-brittle transition and upper-shelf
energy.
All the samples that were supposed to be embrittled have a transition
temperature of about 15 to 18 °C. Solution treatment at 950 °C resulted in the DBTT to
be as high as 40 °C, in comparison with the steel treated at 800 °C while all these
specimens fractured in a brittle manner, due to a large grain size that originated from
annealing at 950 °C. The upper shelf energy seemed to be independent of the Laves
phase precipitation.
Figure 6.21 shows the microstructural evolution during the
P a g e | 122
reheating treatment; and from this optical micrograph the grain boundary Laves phase
precipitates could be observed. At both 600 and 900 °C, there are few precipitates
present as compared to the specimens annealed at 700 and 800 °C. The hardness
results in Figure 6.22, show the effect of this Laves phase re-precipitation on the
hardness values of the Steel A. The result shows a slight increase in the hardness
value at the annealing temperature of 700 °C, then the hardness decreases with
annealing at higher temperatures. Comparing these results with those from Figure 6.22,
it can be concluded that there is no clear correlation between the Laves phase
precipitation and the Charpy upper shelf energy.
10 0
600
700
800
900
950
80
°C
°C
°C
°C
°C
CVN (J)
60
40
20
0
-80
-60
-40
-2 0
0
20
40
60
80
10 0
12 0
T em p era tu re (°C )
Figure 6.20. Charpy impact energy of Steel A as a function of the test temperature of
specimens first solution annealed at 950°°C and then re-annealed at different
temperatures.
P a g e | 123
600 °C
700 °C
800 °C
900 °C
Figure 6.21. Optical microscopy micrographs showing microstructural evolution in Steel
A during re – heating treatments after an original solution treatment at 950°°C.
220
Hardness (HV300g)
200
180
160
140
120
100
500
600
700
800
900
1000
Tem perature (°C)
Figure 6.22. Effect of the Laves phase re-precipitation in Steel A on the hardness of the
material during embrittlement treatment after an original solution treatment at
950°°C.
P a g e | 124
CHAPTER SEVEN
EXPERIMENTAL RESULTS
EFFECT OF THE STEEL’S COMPOSITION
7
7.1 EFFECT OF ANNEALING TREATMENT ON STEEL B
In order to understand the precipitation behaviour of the Laves phase more precisely, an
AISI type 441 (Steel B (0.149%Ti-0.445Nb-0.008%Mo)) that did not fail during
processing was used. The Charpy impact specimens from this steel were annealed at
temperatures between 850 °C and 950 °C for 30 minutes followed by quenching into
water. The quantity of precipitates in the steel after annealing was then electrolytically
extracted and the volume fraction of Laves phase was measured and calculated through
powder XRD. The results in Figure 7.1 show that after annealing at 850 °C, the amount
of Laves phase has decreased to about 0.13% from the 0.2% in the as received
condition, and annealing at 950 °C, a further decrease to about 0.05% occurred. A
small remnant of Laves phase of about 0.05% volume fraction was still found in this
steel, even after annealing at the temperature of 950 °C.
This contrasts with the
calculated solvus temperature of about 800 °C as was found by Thermo-Calc® for this
particular steel.
The Charpy impact tests, however, reveal that after annealing at 850 °C, the remaining
quantity of Laves phase of about 0.15% does not reduce the CIE as significantly as did
the larger Laves phase volume fraction of 0.2% on the grain boundaries upon hot rolling
in the as received material. In this particular steel it, therefore, appears that a critical
maximum volume fraction of Laves phase that may avoid significant impact
embrittlement is about 0.15% although this particular value may also depend on the size
of the Laves phase particles on the grain boundaries and on the precise alloy
composition. With annealing at 950 °C, grain growth starts to affect the CIE value and
the very small volume fraction of about 0.05% of remaining Laves phase is probably
insignificant.
P a g e | 125
90
0 .0 0 3 0
30
0 .0 0 2 5
25
70
CVN (J)
0 .0 0 2 0
60
0 .0 0 1 5
50
vol. fraction
40
30
20
a s re c
850
900
0 .0 0 1 0
20
15
10
0 .0 0 0 5
5
0 .0 0 0 0
0
grain size (µm)
Grain size
Laves phase % vol. fraction
80
S te e l B
V o lu m e (c m 3 )
g r a in s iz e
950
T e m p e ra tu re (° C )
Figure 7.1. Effect of annealing treatment on the Laves phase’s % volume fraction, grain
size and the Charpy impact toughness of the 441 ferritic stainless steel, Steel B.
7.2
EFFECT
OF THE
EQUILIBRIUM LAVES PHASE VOLUME FRACTION
ON THE
ROOM
TEMPERATURE CHARPY IMPACT ENERGY
In the above experiment discussed in Section 7.1, a constant annealing time of 30
minutes was used at various temperatures to vary the volume fraction of the Laves
phase, i.e. at lower temperatures it may be that equilibrium conditions had not been fully
achieved. A second set of Charpy specimens was then used with annealing at a
constant temperature of 800 °C but with now varying the time to study any possible
effects of achieving or not achieving equilibrium conditions in the Laves phase formation.
The Charpy impact specimens were, therefore, heated at 800 °C for a period of between
5 minutes and 300 minutes followed by quenching into water and then the quantity of the
Laves phase precipitates was measured. The results are given in Figure 7.2, which
show that in the as received hot rolled material which had a relatively high volume
fraction of about 0.2% of Laves phase, this had resulted in lowering the CIE to about 25
J, and after annealing at 800 °C for 5 minutes, 33% of the Laves phase had already
dissolved in this steel and this resulted in a large improvement in its CIE to above 50 J.
Further annealing periods did improve the CIE only marginally to about 60 J.
It,
therefore, appears that equilibrium conditions are neared relatively quickly in the first few
minutes of annealing at 800 °C with only a small time dependence after longer annealing
times.
P a g e | 126
Considering the change in volume fraction in Figure 7.2 it appears that the dissolution of
the slightly more than 0.2% Laves phase in the starting or as received material, takes
place very rapidly with only about 0.07% left after 5 minutes at 800 °C. A very small
degree of “overshoot” then seems to take place in the dissolution rate before equilibrium
is established after 30 to 60 minutes at 800 °C. The reasons for such an “overshoot” are
not quite clear but it would clearly fall into a non-equilibrium transient process. The very
rapid dissolution rate in the first 5 minutes is even more remarkable if one considers that
the heating up time within these first few minutes of the 5 mm thick specimens, should
really be discounted from these first 5 minutes.
70
0.0025
CVN
Vol frac
0.0020
50
0.0015
40
0.0010
30
0.0005
20
Vol. fraction
CVN (J)
60
0.0000
as
5
10
30
60
120
180
240
300
1200
ageing time (min)
Figure 7.2. Effect of the Laves phase precipitation kinetics on the Charpy impact
toughness of Steel B.
Comparing the CIE values in this Steel B with the earlier reported ones for Steel A in
Chapter 6, it is surprising that there is only about 14 J difference in the CIE between the
two as received hot rolled materials as processed by Columbus with Steel A rejected by
Columbus due its CIE of only about 10 to 11 J and Steel B’s 25 J which was accepted
by Columbus. Microstructural analysis, however, revealed that Steel B had a much finer
grain size than Steel A, and this may have resulted in the higher CIE for Steel B.
Figure 7.3 shows the optical microstructural evolution of Steel B in the as received hot
rolled condition and after annealing treatments at different temperatures ranging from
850 to 950 °C. The as received microstructure in Figure 7.3(a) appears to be a partly
dynamically recrystallised microstructure or alternatively, a post-rolling statically
recrystallised microstructure with some recrystallised grains in an overwhelmingly
unrecrystallised microstructure. Note furthermore, that recovery appears to have taken
place primarily during static annealing at 850 °C with full recrystallisation only occurring
at temperatures of 900 to 950 °C.
P a g e | 127
(a)
(b)
850 °C
As received
(c)
(d)
900 °C
950 °C
Figure 7.3. Optical micrographs of the specimens from steel B in the (a) as received plant
hot rolled condition and (b) to (d) after being annealed at different temperatures
from 850 to 950°°C for 30 minutes followed by water quenching.
7.3
EFFECT
OF
ANNEALING TREATMENT
ON THE
EMBRITTLEMENT
OF THE
EXPERIMENTAL
STAINLESS STEELS C TO E
Figure 7.4 shows the Charpy impact toughness behaviours of Steels C to E as a
function of annealing treatment.
These alloys are brittle over a very wide range of
temperatures with the toughness’s averaging below 5 J. It should be noted that some of
these results from Steel C (0.36%Nb – 0.171%Ti - <0.01%Mo alloy, i.e. higher in Ti and
lower in Nb than the commercial Steels A and B) are scatted between roughly two
extremes; i.e. specimens of this steel are either entirely brittle or entirely ductile. Steels
D (0.54%Mo) and E (1.942%Mo) are the molybdenum containing steels and they have
shown very similar brittle behaviour. This suggests that any Mo additions to a AISI 441
type stainless steel has a negative impact on the toughness of this alloys, no doubt due
to the expected higher volume fractions of Laves phase as predicted earlier in Chapter 5
by Thermo – Calc®. Comparing the CIEs of Steel C (Mo-free) to those of the MoP a g e | 128
containing Steels D and E, it can be seen that Steel C had a much higher CIE at a
number of annealing temperatures than the other two steels, although at some
temperatures this was not the case where Steel C was equally brittle to Steels D and E.
Again, this can be related to the quantity of Laves phase in Steel C, as it was previously
shown by Thermo-Calc® predictions (see, Chapter 5.) that the Laves phase volume
fraction in this steel should be much lower than in Steels D and E. Any Mo additions to
a type 441 ferritic stainless steel should, therefore, be avoided.
Secondly, comparing the somewhat erratic CIE behaviour of Steel C (lower Nb and
higher Ti contents) with annealing to the more consistent behaviour of the two
commercial Steels A and B, one may conclude that lowering the Nb content and raising
the Ti content in the dual stabilisation from the standard dual stabilisation, should be
carefully considered in practice as it may have negative consequences. This tentative
conclusion is, however, somewhat uncertain as these experimental steels also had a
much larger grain size (see below) than the two commercial steels and some further
studies could be considered to fully prove or disprove this tentative conclusion. The
initial observation of the microstructure of the laboratory hot rolled experimental Steels
C, D and E prior to any annealing treatment, showed them to have much larger grain
sizes, typically more than 150 µm if compared to the plant hot-rolled materials in Steels
A and B, see Figure 7.5. The effect of grain size on the toughness of ferritic stainless
steels has been well documented over the years, and it is known that a large grain size
decreases the toughness of the steel significantly. Because of the larger grain sizes
than expected, no further work was done on the brittle behaviour of these three steels
although they were still included in the studies to determine the effect of composition on
the nucleation and kinetic precipitation behaviour of the Laves phase in these alloys.
P a g e | 129
35
Steel C (Nb-Ti Alloy)
Steel D (Ni-Ti-Mo Alloy)
Steel E (Type 444 Alloy)
30
25
CVN (J)
20
15
10
5
0
As HR
600
700
800
850
900
950
Temperature (°C)
Figure 7.4. Effect of annealing temperature on the room temperature Charpy impact
energy of the laboratory hot rolled materials. The samples were annealed for 30
minutes at different temperatures and then water quenched: Steel C (Nb-Ti alloy);
Steel D (Nb-Ti-Mo alloy) and Steel E (Type 444 alloy).
(a)
(b)
(c)
Figure 7.5. The microstructure of the laboratory hot-rolled experimental steels, showing
different grain size distributions if compared to those of the commercial Steels A
and B: (a) Steel C; (c) Steel D; and (d) Steel E.
P a g e | 130
CHAPTER EIGHT
EXPERIMENTAL RESULTS
LAVES PHASE KINETICS STUDY
8
8.1
INTRODUCTION
One of the most important aspects of this research work was to study the transformation
kinetics of the Laves phase, in particular its nucleation and the overall volume fraction
evolution. The research, therefore, focused on determining the Laves phase volume
fraction as a function of time during annealing below 850 °C and this produced a
frequently found sigmoidal or S–shaped curve. The transformation kinetics are then
described using the well known Johnson–Mehl–Avrami–Kolmogorov (JMAK) type of
equations [95]. The following parameters that affect the kinetics of the Laves phase
precipitation were considered: annealing temperature and time, the alloy’s and Laves
phase’s composition and the steel’s grain size.
8.2
EQUILIBRIUM LAVES PHASE FRACTION
Thermo-Calc® software was first used to estimate the equilibrium Laves phase fraction
in Steel A, as a function of temperature, see Chapter 5.
All the possible phases,
including the carbo-nitrides, were included in the calculations. The equilibrium fraction
of the Laves phase as it was formed after long time annealing was then determined
experimentally from the Steel A, i.e. the steel which had failed during production. The
materials were first annealed at 850 °C for 2 hours in order to dissolve and minimise the
Laves phase content and at the same time making sure that there was minimal grain
growth, and the material was then quenched into water. The specimens where then
annealed at temperatures between 600 and 850 °C for periods ranging from 5 to 1000
minutes and then water quenched.
The precipitate residues where electrolytically
extracted and analysed from these annealed specimens using XRD techniques. The
physical quantity of the precipitates was determined from the relative intensity of XRD
lines using a Rietveld technique, which enables the volume fraction – time relationship to
be found with fair accuracy. Figure 8.1 shows the typical S-shaped curves plotted from
these analyses. Note that the XRD analyses give the fractions in weight fraction, and
these values had to be converted to volume fractions using the densities of each phase
P a g e | 131
and using the overall weighed amount of the precipitate residues collected. An example
of such a calculation was shown in Section 6.2.1 and the overall results are shown in
Figure 8.1.
0.05
600
825
700
750
800
775
725
850
0.04
°C
°C
°C
°C
°C
°C
°C
°C
Vol. fraction
0.03
0.02
0.01
0.00
-0.01
1
10
time (min)
100
1000
Figure 8.1.
The Laves phase volume fraction – temperature/time curves during
isothermal annealing in the temperature range 600 °C to 850 °C.
From Figure 8.1, it was observed that for the specimens that were heated at 600 °C,
even after annealing for 1000 minutes, the Laves phase transformation had not reached
equilibrium and was still continuing. This indicates that although the chemical driving
force for Laves phase nucleation should be relatively high at this temperature due to a
high undercooling, the lowered diffusivity of Nb atoms at this temperature led to a low
overall nucleation rate. Also, it was observed that at 700 °C, the transformation kinetics
of the Laves phase are much higher than those of the specimens annealed at higher
temperatures of 750 °C and above, i.e. the volume fraction of Laves phase was
observed to be much higher at 700 °C than with any other specimen annealed at higher
or lower temperatures.
This already hints at the possibility of a second lower
temperature nose in a TTT diagram for this phase. The same observation was made by
Sawatani et al [8], on the Ti and Nb stabilised low C, N – 19%Cr – 2%Mo stainless steel,
where they have observed the Laves phase precipitates to be in a far larger quantity
after annealing at about 700 °C. At 850 °C in Figure 8.1, the volume fraction of the
Laves phase precipitates in Steel A reached a maximum of only 0.005 %, and a proper
S-curve could not be established. However, this already shows a discrepancy between
the experimental evidence of 0.005% volume fraction remaining at 850 °C versus the
Thermo-Calc® prediction that no Laves phase should be present at and above about
825 °C in this Steel A.
P a g e | 132
8.3
LAVES PHASE TRANSFORMATION KINETICS
The kinetics of an isothermal transformation is usually expressed by the modified
Johnson–Mehl–Avrami–Kolmogorov (JMAK) type of equation:
(
Vv = 1− exp − kt n
)
Equation 8.1
where Vv is the Laves phase volume fraction, k is the reaction rate constant, and n is
the time exponent. From Figure 8.1, the points that fall between 5 and 95 percent on the
S- curves (on the specimens annealed in the temperature range of 700 – 825 °C) were
used to determined the evolution of the Laves phase by plotting them in the {lnln [1/(1Vv)]} vs {ln t} relationship, see Figure 8.2. The results produced a linear relationship with
the slope n and intercept ln k, and the calculated values are shown in Table 8.1. The
time exponent values of n were found to range between 1.48 and 1.54, with n at 800
°C the lowest with n = 1.37. The results suggest that Laves phase nucleation takes
place on either plane or edge grain boundaries [108].
Figure 8.2. The Laves phase transformation curves according to the Johnson–Mehl–
Avrami–Kolmogorov (JMAK) type of equation.
Table 8.1. The measure values of the time exponent n and the rate constant k, obtained
from the best fit equation of the plots in Figure 8.2.
Temp (°C)
700
725
750
775
800
825
Best fit eqn
y=1.54x – 10.54
y=1.48x – 11.42
y=1.48x – 11.23
y=1.48x – 10.40
y=1.37x – 9.64
y=1.54x – 9.61
R2
0.97
0.99
0.99
0.91
0.98
0.96
n
1.54
1.48
1.48
1.48
1.37
1.54
ln k
-10.54
-11.42
-11.23
-10.40
-9.64
-9.61
k
2.65E-05
1.10E-05
1.33E-05
3.04E-05
8.54E-05
6.69E-05
P a g e | 133
The activation energy for the Laves phase was estimated from the reaction rate constant
k which is temperature dependent and can be expressed as :
 −Q 
k = A exp

 RT 
Equation 8.2
where A is the pre-exponentials factor, Q is the activation energy, R is the gas
constant, and T is the temperature. By plotting a linear relationship of ln k vs 1/T, the
activation energy for the Laves phase precipitation was determined as the slope using
the following relationship:


 ∂ ln k 

Q = −R 
 ∂1 


 T 
Equation 8.3
and Q was estimated to be 211.3 kJ/mol, a value that is somewhat lower than the
activation energy for the diffusion of Nb in a ferrite matrix, which was estimated to be
about 240 kJ/mol [103]. It should be noted that this measurement of the activation
energy was carried out using only data from the temperature range of 750 to 825 °C as it
appears that the precipitation mechanism may be different at lower temperatures than
that in this temperature range. If all the data points from all temperatures were used the
activation energy was estimated to be much lower, only about 107.5 kJ/mol, possibly
suggesting the introduction of enhanced diffusion along grain boundaries into the
kinetics.
8.4
TEMPERATURE EFFECT ON ISOTHERMAL TRANSFORMATIONS
The kinetics of the Laves phase transformation were found experimentally at a number
of different constant temperatures, and this led to a complete isothermal transformation
diagram being drawn. Figure 8.3 is a time – temperature – precipitation (TTP) diagram
for the Laves phase formation, and it gives the relation between the temperature and the
time for the fixed fractions of transformation to be attained. Three of these curves are
given in the TTP diagram, for the measuring times (t5% and t95%) for the beginning and
end of transformation and for 50% transformation, i.e. t50%. From Figure 8.3, it should
be noted that the points on the solid line were determined experimentally and the dash
lines are extrapolated estimated lines.
P a g e | 134
900
850
5%
50%
95%
Temperature (°C)
Thermo-Calc® limit
800
750
700
650
600
103
104
105
Time (sec)
Figure 8.3 A time – temperature – precipitation (TTP) diagram for the Laves phase
formation in Steel A.
The results show two classical C noses on the transformation curves, the first one
occurring at a higher temperature of about 825 °C and the second one at much lower
temperatures, estimated to possibly be in the range of about 650 to 675 °C. Because of
extraordinary long annealing times due to the much slower diffusion, it was found to be
impractical to find the exact point of this lower temperature nose. The available results,
however, are sufficient to show that there are probably two types of the Laves phase
transformations controlled by different nucleation mechanisms that are taking place in
this steel. In the work done by Silva et al. [133] on the AISI 444 ferritic stainless steel
(that is, similar to Steel E from this work, but slightly different in composition), the
authors have also estimated the C nose of the transformation curve to be around 800 to
850 °C.
Thermo-Calc® predictions on this Steel A have estimated the solvus
temperature of the Laves phase to be 825 °C, but the experimental results indicate that
the Laves phase still exists up to the temperature of 850 °C or even more.
8.5
EFFECT OF THE GRAIN SIZE ON THE TRANSFORMATION KINETICS OF LAVES PHASE
The transformation kinetics of the Laves phase were compared in two specimens from
Steel A with different grain sizes, with the objective to evaluate the effect of grain size on
the formation of the Laves phase precipitates in this AISI type 441 ferritic stainless steel.
The materials were first annealed at the temperatures of 850 °C and 950 °C for 2 hours
and then water quenched.
The corresponding linear intercept grain sizes were
determined to be 49.9 µm and 152.1 µm, respectively from which the grain boundary
surface areas per unit volume Sv were calculated to be 4.008x104 and 1.315x104
m2/m3 respectively, i.e. a difference of 67% in the potential nucleation site availability
P a g e | 135
for grain boundary nucleation. Subsequently, the specimens were annealed at 750 °C
for different periods in order to determine the Laves phase transformation kinetics.
Figure 8.4 show the two typical S – shaped transformation curves obtained from the
XRD analyses. The results show that the precipitation kinetics of the Laves phase is
retarded by the large grain size, and this is due to a smaller number of nucleation sites.
However, it was observed that an equal level of the maximum volume fraction could be
achieved from both grain sizes, although at different annealing times.
Within the
specimen with a larger grain size the same volume fraction of Laves phase precipitates
could be achieved but by annealing for approximately an extra 400 minutes.
It is
revealing to note that the difference in time to achieve the 50% transformation level in
Figure 8.4 is reasonably close to the earlier 67% difference in the grain boundary
nucleation site availability. These results confirm that grain boundary nucleation of the
Laves phase’s nuclei is of overriding significance. In a similar study done by Pardal et
al. [134] on a superduplex stainless steel UNS S32750, the authors have made similar
qualitative observations, but they did not determine the overall effect of the grain size on
the volume fraction as was done here.
0.04
49.9 µm
152.1 µm
Vol. fraction
0.03
0.02
0.01
0.00
1
10
100
1000
Time (min)
Figure 8.4. Effect of the grain size on the Laves phase kinetics transformation in Steel A.
The specimens were annealed first at 850 and 950°°C respectively to set different
grain sizes and were then annealed both at 750 °C for different annealing periods.
8.6
EFFECT
OF THE
STEEL’S COMPOSITION
ON THE
LAVES PHASE’S TRANSFORMATION
KINETICS
The effect of the steel’s composition on the Laves phase kinetics was investigated
comparing Steel A (0.444Nb – 0.153 Ti – ~ 0Mo), with the Steel C (0.36Nb – 0.171Ti –<
P a g e | 136
0.01Mo) and Steel D (0.251Nb – 0.106Ti – 1.942Mo). The grain size of these materials
was kept constant and the transformation kinetics was investigated by annealing the
specimens at 750 °C over different periods of time. Figure 8.5 shows the results of the
effect of composition on the kinetics of the Laves phase precipitation, which indicates
that at 750 °C, Steel A has a much higher volume fraction of Laves phase precipitates
than Steels C and E, and this coincides with the Thermo-Calc® predictions. Note that
from Figure 8.5, in order to be able to compare the volume fractions of Laves phase of
Steels C and E, their smaller values are shown on the secondary axis while that of Steel
A is shown on the primary axis.
These results also indicate that by reducing the Nb and Ti contents as with Steel C (in
comparison to Steel A), this did not have any significant impact on the kinetics of the
Laves phase precipitation but only on the final volume fraction.
But the addition of
1.942% Mo has a measurable impact by retarding the precipitation rate of the Laves
phase. In the work published by Ahn et al. [135] the authors have observed that the
precipitation of Fe2Nb Laves phase in 0.01C–0.38Nb–1.2Mo steel was slower than in an
0.01C-0.38Nb steel, and this was due to Mo retarding the rate of Nb diffusion.
0.035
Vol. fraction
0.0016
Steel A (0.444Nb-0.153Ti-~Mo)
Steel C (0.36Nb-0.171Ti-<0.01Mo)
Steel E (0.251Nb-0.106Ti-<1.942Mo)
0.0014
0.030
0.0012
0.025
0.0010
0.020
0.0008
0.015
0.0006
0.010
0.0004
0.005
0.0002
0.000
1
10
100
Vol. fraction
0.040
0.0000
1000
Time (min)
Figure 8.5. Effect of the steel’s composition on the Laves phase transformation kinetics.
The specimens from these steels were all annealed at 750 °C for different
annealing periods.
8.7
MICROSTRUCTURAL ANALYSIS OF THE TRANSFORMATION KINETICS
The nucleation mechanisms of the Laves phase precipitates during transformation was
investigated from the specimens that were annealed at 600 °C, 750 °C and 800 °C for
30 minutes, which allowed for a suitable time of precipitate growth, so that they could be
analysed using TEM – EDX.
P a g e | 137
Figure 8.6, shows the microstructure of the specimen annealed at 600 °C. At the lower
magnification (Figure 8.6(a)) the remnant of the grain boundary Laves phase
precipitates could be found from the solution treatment at 850 °C for 2 hours. At a
higher magnification (Figure 8.6(b), the same area analysed from Figure 8.4 (a) is
indicated by a circle, and the micrograph indicates that the nucleation site for the Laves
precipitation is mainly at dislocations and subgrain boundaries, with only a very few
precipitates on grain boundaries and almost none within the grains themselves. Laves
phase precipitates that have nucleated on the grain boundaries, are possibly coherent or
alternatively have a low mismatch with one grain but grow into the adjacent grain with
which they do not have a rational orientation relationship because of that interface’s
higher mobility. Similar observations were made by Li [131] on 12Cr – 2W steel, where
it was found that with the nucleation and growth of the Fe2W Laves phase precipitates,
the coherent or low mismatch interfaces have a low mobility while the incoherent ones
have a higher mobility, and therefore, the incoherent interface will grow into the grain
with which there is no rational orientation relationship.
(a)
(b)
Figure 8.6. TEM micrographs of the specimen of Steel A annealed at 600 °C; (a) a low
magnification micrograph shows coarse grain boundary Laves phase precipitates,
and (b) the same area but at a high magnification, showing Laves phase
precipitates nucleated on subgrain boundaries and dislocations.
P a g e | 138
The microstructure of the specimen of Steel A that was annealed at 700 °C is shown in
Figure 8.7. Comparing this specimen with the one that was annealed at 600 °C, it can
be seen that even after only 30 minutes of annealing, the volume fraction of the Laves
phase is higher, also suggesting a high nucleation rate at 700 °C in this specimen. Also,
there is more of the grain boundary Laves phase precipitation than the dislocation
precipitates, see Figure 8.7 (a). This suggests that the most preferred nucleation site for
the Laves phase are the grain and subgrain boundaries, unlike as in the specimen
annealed at 600 °C, where more precipitates on dislocations were observed.
(a)
(b)
Figure 8.7. TEM micrographs of the specimen of Steel A annealed at 750 °C; (a) a low
magnification micrograph showing grain and subgrain boundary Laves phase
precipitates, and (b) at a high magnification, showing Laves phase precipitates
nucleated on the subgrain boundaries.
Figure 8.8 shows the micrographs of the specimen that was annealed at 800 °C, which
is heavily “decorated” with the Laves phase precipitates. At the higher magnification, it
was observed that the preferred sites for the Laves phase nucleation are grain
boundaries, with very few precipitates on dislocations compared to the specimen
annealed at 600 °C.
Also, it can be observed that the presence of the dislocation
P a g e | 139
density did not assist as much as preferred nucleation sites for the Laves phase
precipitation as was the case at lower temperatures.
Comparing all of the microstructures at these three different temperatures, it can be
concluded that dislocations, subgrain and grain boundaries act as preferred nucleation
sites for Laves phase precipitation. Analyses of the microstructures and also on the
basis of the classical heterogeneous nucleation theory, demonstrates that nucleation on
the grain boundaries is dominant at the higher testing temperatures of 750 °C and
above, see Figure 8.7 and Figure 8.8, where the undercooling and hence the driving
forces for nucleation are relatively low and the system then lowers its retarding forces
through grain boundary nucleation.
As the temperature is decreased and the
undercooling and hence the driving forces are higher, however, heterogeneous
nucleation on dislocations becomes more significant.
(a)
(b)
Figure 8.8. TEM micrographs of the specimen annealed at 750 °C; (a) at a low
magnification, showing grain boundary Laves phase precipitates, and (b) at a
higher magnification showing Laves phase precipitates nucleated on the subgrain
boundaries.
P a g e | 140
8.8
ORIENTATION RELATIONSHIP BETWEEN THE LAVES PHASE AND THE FERRITE M ATRIX
Figure 8.9 to Figure 8.11 show bright field images and the corresponding selected area
diffraction patterns (SADP) from the Steel A specimens annealed at 600 to 800 °C, for
30 minutes. The microstructure consists of a number of differently shaped particles
ranging from needle-shape at 600 °C, globular-shaped at 750 °C and plate–like at 800
°C. These precipitates were all identified as Fe2Nb Laves phase with a C14 crystal
structure.
Figure 8.9 shows the specimen of Steel A that was annealed at 600 °C and the
crystallographic orientation relationship between the needle-like Laves phase and the
matrix was analysed using the SAD pattern as:
(21 1)α
− Fe
(
)
// 1 101 Fe2Nb
and [011]α − Fe // [1 1 02 ]Fe 2Nb .
(1)
According to Cocks and Borland [90], by rotating the above orientation relationship (1) to
within ±5° around the habit plane {111}α-Fe and a direction in this plane, the following
orientation relationship is found;
(1 1 1)α −Fe // (11 2 0 )Fe2Nb
and [121 ]α − Fe // [0001 ]Fe 2 Nb .
(2)
The needle-like Laves phase particles found in this specimen are, therefore, elongated
in the 112 matrix directions.
Figure 8.10 shows the specimen that was annealed at 750 °C, in which the
crystallographic orientation relationship between the globular-like Laves phase and the
ferrite matrix was analysed to be the following orientation relationship:
(211 )α − Fe // (0001 )Fe 2 Nb and
[1 11]α
− Fe
[
]
// 01 1 0 Fe 2 Nb
(3)
According to Yamamoto et al.[91], using the stereographic projection of the orientation
relationship (1) between the Laves phase and the ferrite matrix, they have found that this
orientation relationship is equivalent to the following expression by the lower mirror
indices;
(211 )α − Fe // (0001 )Fe 2 Nb , and
[0 1 1]α
− Fe
[
// 2 110
]
Fe 2 Nb
(4)
P a g e | 141
Figure 8.11 shows the specimen that was annealed at 800 °C, in which the
crystallographic orientation between the plate-like Laves phase and the ferrite matrix
was analysed to follow the orientation relationship:
(110 )α − Fe // (0001 )Fe 2 Nb and
[2 33 ]α
− Fe
[
]
// 11 2 0 Fe 2 Nb
(5)
If the orientation relationship (5) is rotated by ±5° from the above mentioned relationship,
it will follow:
(110 )α − Fe // (0001 )Fe 2 Nb and [001]α − Fe
[
// 1 2 10
]
Fe 2 Nb
(6)
The habit plane from this orientation relationship is {110}α-Fe, and the preferred growth
orientation must be in the 001 matrix directions.
Comparing all three of these SAD patterns, it can be seen that the orientation
relationship between the Laves phase particle and the ferrite matrix may have a
significant effect on the shape of the particles. Also, the orientation relationships for the
particles analysed are completely different from one another, and the habit planes are
also different.
[011]α − Fe // [1 1 02 ]Fe 2Nb
Figure 8.9.
Transmission electron micrographs and corresponding selected area
diffraction (SAD) pattern from Steel A annealed at 600 °C.
P a g e | 142
[1 11]α
− Fe
[
]
// 01 1 0 Fe2Nb
Figure 8.10. Transmission electron micrographs and corresponding selected area
diffraction (SAD) pattern from Steel A annealed at 750 °C.
[2 33 ]α
− Fe
[
// 11 2 0
]
Fe 2Nb
Figure 8.11. Transmission electron micrographs and corresponding selected area
diffraction (SAD) pattern from Steel A annealed at 800°°C.
P a g e | 143
CHAPTER NINE
DISCUSSIONS
LAVES PHASE EMBRITTLEMENT
9
9.1
INTRODUCTION
In this chapter the parameters that affect embrittlement of the type 441 ferritic stainless
steel are discussed. Parameters such as the grain size, grain boundary Laves phase
and the cooling rate were tested against existing fracture theories of Stroh, Cottrell and
Smith. Also, the effect of possible niobium solute drag during the recrystallisation and
grain growth was considered qualitatively.
9.2
PRECIPITATES FOUND IN AISI 441 FERRITIC STAINLESS STEEL
Thermo-Calc® calculations predict that the phases that are stable over a wide range of
temperatures in these steels are the Laves phase, carbo–nitrides and the sigma phase
in a structure that remains completely ferritic from its initial solidification. The presence
of the M6C or (Fe3Nb3C) type carbides were detected experimentally using XRD and
TEM analyses. But this precipitate seems to appear infrequently in these steels and its
volume fraction could not be quantified due to its low value. In the TEM image they
appear as coarse grain boundary precipitates, see Figure 9.1. Such coarse precipitates
are known not to cause any significant strength after high temperature ageing [5]. But
the presence of the coarse M6C type carbides can be prevented by adding elements
which have a stronger affinity for carbon than the Nb in these steels, that is Ti. The
amount of Nb addition depends on the carbon and nitrogen content, and it was found
that an addition of 0.1wt.%Ti is normally enough to stabilise 0.02 wt.%C and N as
Ti(C,N) [5].
Angular carbo-nitrides of titanium and niobium that have precipitated from or soon after
the melt are randomly dispersed throughout the structure.
The presences of these
second phases in these steels tend to lower the melting point of the steel [55]. Excess
niobium is taken into solid solution during high temperature annealing and is reprecipitated as very fine particles of the Laves phase (Fe2Nb) upon either slow cooling
or upon holding at intermediate temperatures of 600 – 950 °C [136]. Strengthening by
P a g e | 144
this dispersion was found to be responsible for improved elevated temperature strength
[8,69]. Although Thermo-Calc® calculations show the presence of the sigma phase as
an equilibrium phase, others have indicated that this phase is not supposed to form in
this type of alloy and neither was it encountered in this study. It has often been reported
that sigma phase precipitates in high chromium ferritic stainless steels but only after very
long periods at high service temperatures [25,26,27].
Figure 9.1. TEM micrograph shows the presence of the M6C or (Fe3Nb3C) type carbide in
the subgrain structure from Steel A. Note that the specimen was annealed at 700
°C for 30 minutes and other fine particles were determined to be Fe2Nb Laves
phase particles.
9.2.1 EFFECT OF THE STEEL’S COMPOSITION ON THE PRECIPITATE’S SOLVUS TEMPERATURE
The solvus temperature and volume fraction for each phase is alloy composition
dependent. From the Figures 5.4 to 5.8 in Chapter 5; the minimum temperature for the
dissolution of the Laves phase and the carbonitrides were estimated. These dissolution
temperatures for the Laves phase were used as a guide to avoid excessive grain growth
during higher annealing temperatures. Using Steel A with a nominal composition of
0.012C-0.0085N-0.444Nb-0.153Ti-∼0Mo as this study’s reference steel, it was
established that the C and N contents have a significant impact on both the solvus
temperature and composition of the (Ti,Nb)(C,N), with increasing one or both of them
resulting in increments of both solvus temperature and volume fraction.
The Laves phase’s formation was found to be more dependent on the Nb content than
on the Ti content. Decreasing the Nb content of the steel, decreased both the solvus
temperature and volume fraction of the Laves phase. For instance, by reducing the Nb
content to 0.36wt % in Steel C (0.023C-0.024N-0.36Nb-0.171Ti-<0.01Mo) as compared
to 0.444wt % in Steel A, this resulted in lowering the solvus temperature to 765 °C (as
P a g e | 145
compared to 825 °C in Steel A) and the weight fraction determined at 600 °C was
reduced to 0.573 wt% (compared to 0.92 wt% in Steel A). An increase in Ti content in
Steel C, to 0.171 wt% (compared to 0.153 wt % Ti in Steel A) did not have any
significant impact on the both solvus temperature and weight fraction of the Laves
phase. A Mo addition was found to increase the volume fraction of the Laves phase,
whilst at the same time lowering its solvus temperature.
For instance, it has been
observed in this study that a Mo addition from practically zero in Steel A to 1.94 wt % in
Steel E enhanced the formation of the Laves phase in the Nb-Ti-Mo steel (see Figure
5.8 and Table 5.6), while at the same time slowing its precipitation kinetics (see Figure
8.5). Some authors have observed that this is due to the Mo retarding the diffusivity of
Nb to form Laves phase precipitates [5,6,135]. A more detailed analysis of the kinetics
of nucleation of the Laves and other second phases in this and other similar Nbcontaining ferritic stainless steels will be reported in Chapter 10.
The Thermo-Calc® results of Steel A where compared with the experimental results of
obtained from the XRD analysis, see Figure 9.2. The results from Thermo-Calc® predict
that at 825 °C, the weight fraction of the Laves phase should be zero, whereas the
experimental results found that there was still about 0.031wt% present and a further
remnant still existed even at 850 °C and above. The presence of the Laves phase up to
a temperature of 950 °C in a similar ferritic stainless steel to the one studied in this work,
was observed by other researchers, where they have observed that the Laves phase
improved the high temperature strength of their steel [69,136].
0.010
Weight fraction
0.008
0.006
0.004
0.002
0.000
600
650
700
750
Temperature (°C)
800
850
.
Figure 9.2. Comparison between experimental and Thermo-Calc® calculated weight
fractions of Laves phase in Steel A. The points and dotted line represent the
experimental results while the full line is as predicted by Thermo-Calc® for this
steel.
P a g e | 146
9.2.2 EFFECT OF THE STEEL’S COMPOSITION ON THE PRECIPITATE’S COMPOSITION
It was established from the Thermo-Calc® predictions that the composition of the
titanium and niobium carbonitrides (Ti,Nb)(C,N) is not affected by a change in the steel’s
composition, but only the stoichiometric composition is affected. On the other hand, the
Laves phase’s composition is highly affected by the change in the steel’s composition.
Thermo-Calc® predictions show that the composition of the Laves phase in Steels A and
B contains the following elements: Fe, Cr, Ti and Nb, but no C or N, and this led to the
conclusion that the Laves phase is in the form of (Fe,Cr)2(Nb,Ti). The absence of C or N
in the Laves phase was also observed by Sawatani et al. [8], using an electron probe
microanalysis technique, but in the work of Lundin [137] on a 10.6Cr-1.01W-1.02Mo0.04Nb ferritic stainless steel and using an atom-probe field-ion microscopy (APFIM)
analysis of the Laves phase, the author found carbon to be soluble in the Laves phase.
It needs to be noted, however, that the respective carbon contents of the two steels from
the work by Sawatani [8] and Lundin [137] were 0.008%wt. and 0.11%wt, respectively.
Therefore, the solubility of the carbon in the Laves phase appears to be directly
dependent on the carbon content of the steel.
When the Nb content is reduced and the Ti content is increased (as in the Steel C
(0.36Nb-0.171Ti)) or only the Ti is decreased (as in Steel E (0.251Nb-0.106Ti)), ThermoCalc® predictions show that Ti is not taken up into the Laves phase, but resulted in an
increase in the Ti-content of the Ti(C,N) in this steel. The discrepancy caused by the Ti
solubility in these alloys as compared to other alloys could not be established from this
work. Therefore, a further study to gain a better understanding on the solubility of Ti in
the Laves phase in these steels is needed.
In the case where Mo additions were made, such as in Steels D and E, it was found that
Mo is taken up into the Laves phase as ((Fe,Cr)2(Nb,Ti,Mo) for Steel D and
(Fe,Cr)2(Nb,Mo) for Steel E respectively. It was, therefore, concluded that the nominal
composition of the Laves phase should be reported according to the major alloying
elements, i.e. for Steels A – D as Fe2Nb and for Steel E as Fe2(Nb,Mo) respectively.
P a g e | 147
9.3
EMBRITTLEMENT OF TYPE 441 FERRITIC STAINLESS STEEL
9.3.1 EFFECT OF GRAIN SIZE ON FLOW STRESS: THE HALL-PETCH RELATIONSHIP
The change in grain size of Steel A was introduced by annealing the as received and
plant-embrittled material at temperatures of 850 to 1100 °C, i.e. above the estimated
Laves phase solvus temperature. From Section 6.3.4 in Chapter 6, the results show a
steady increase in grain size up to about 950 °C, but between 950 °C and 1000 °C there
is a sudden and rapid 60 % increase in the mean linear intercept grain size. When the
effect of the grain growth on the 0.2% yield strength in this steel was tested according to
the Hall-Petch relationship, it was found that it only applies in the temperature range of
850 °C to 950 °C (see Figure 9.3) but beyond 950 °C, the relationship did not hold. The
Hall – Petch relationship is given by:
σ y = σ 0 + k ys d
−
1
2
Equation 9.1
where σ0 is the friction stress due to dislocation obstacles on the slip plane in the
materials, k ys is the locking parameter representing the grain boundary as an obstacle
to propagation of deformation, and d is the grain size. According this Equation 9.1,
plastic deformation leads to a pile-up of dislocations at the head of a slip plane at a grain
boundary.
This pile-up causes stress concentrations at the boundary and in the
adjacent grain. When the stress concentration is high enough and reaches a critical
value, a dislocation source in the adjacent grain may be activated and plastic
deformation can spread out into this next grain. The values for σ0 and k ys
in this
material were determined experimentally as 229.03 MPa and 469.8 MPa .µm1/2,
respectively from Figure 9.2, although the relatively few data points and their scatter
(see the relatively low value of the regression coefficient R2) make these values
somewhat approximate. In the work by Miyahara et al. [138], working on two types of
ferritic steels with a fine and coarse grain size, the authors have estimated a locking
parameter of 260.0 MPa .µm1/2 which was in good agreement with the conventional
results for a low – carbon steel from Pickering’s [47].
Comparing this value with the value obtained from this work, it can be concluded that
grain boundary strengthening at temperatures ranging up to 950 °C was sufficient to
prevent the embrittlement of this steel and, therefore, the source of embrittlement found
in this study is most likely from the presence of the Laves phase precipitates on suband grain boundaries.
P a g e | 148
330
Yield strength, σ y (MPa)
325
320
315
y = 469.8x + 229.03
R2 = 0.8863
310
305
300
295
290
285
0.10
0.12
0.14
0.16
d
-1/2
(µ
µ m)
0.18
0.20
0.22
-1/2
Figure 9.3. The effect of grain size on the yield strength of Steel A.
9.3.2 CRACK NUCLEATION
According to Stroh [36], the value for the shear stress created by a dislocation pile – up
of length d/2 to nucleate a microcrack is as follows:
1
 πGm γ f  2
τe ≥ 

 2(1 − ν )d 
Equation 9.2
where τe is the effective shear stress; γf is the effective surface energy of the crack;
Gm is the shear modulus; ν is Poisson’s ratio; d is the grain size. The shear stresses
are comparable for the constant Gmb/γf ~ 8 [36], where b is the Burgers vector of the
slip plane. Assuming that Gmb/γf ≈ 8, with Gm = 8.6 x 1010 Pa and b = 2.48 x 10-10 m,
then γf ≈ 2.67 J/m2. It should be noted that this estimated effective surface energy of the
crack is much higher than that of the grain boundary energy in stainless steels which is
in the region of 0.5 to 0.7 J/m2 and that for the free surface about three times higher, i.e.
about 1.5 to 2.1 J/m2 [40,139]. According to Cottrell [38], an effective surface energy of
the crack greater than the true one appears to be a common feature of brittle fracture in
iron which is thought to be mainly due to irreversible work of tearing grain boundaries.
For the specimen annealed at 850 °C which had a grain size of 26.1 µm, and with ν =
0.28 and using Equation 9.2, produces an estimate of the effective shear stress of
approximately 438 MPa.
This value is significantly higher than the measured yield
strength of 323.5 MPa for this steel (see Figure 9.4) after annealing at 850 °C, leading to
a conclusion of ductile and not brittle failure at this grain size. From these results, it can
P a g e | 149
furthermore be concluded that if a crack nucleus can form, any increase in its
propagating length can lead to a decrease in the total system’s energy, provided that
there is no change in the surface energy encountered during the crack growth. The
critical number of dislocations that is sufficient to nucleate a crack at the end of the slipband is given by:
n=
π 2γ f
2τ e b
Equation 9.3
If the steel does show a tendency to fracture in a brittle manner, either because of a low
temperature or from brittle precipitates on the grain boundaries, then a smaller grain size
d would be helpful. The shear stress at the head of the dislocation pile up is given by
nτe, i.e. the smaller the grain size, the smaller the number of dislocations in the pile-up
when the slip band arrives at the grain boundary.
600
500
Stress (MPa)
400
300
200
100
0
0 .0
0 .1
0 .2
0 .3
0 .4
S tr a in (m m /m m )
Figure 9.4. A room temperature tensile test of the specimen of Steel A that was annealed
at 850 °C for 30 minutes and then water quenched.
When secondary precipitates are inhomogeneously distributed (such as the grain
boundary Laves phase precipitates) the value of the surface energy γf may be reduced
and the amount of work done in nucleating a crack decreases. Thus the crack initiates
more easily at grain boundaries, even though the resulting cracks probably propagate
transgranularly and not intergranularly. In the present study there is evidence of the
(Ti,Nb)(C,N) precipitates that have cracked after Charpy impacting the sample annealed
at 850 °C (see Figure 9.5). These carbo-nitride precipitates present on the fracture face
were found to be distributed within the grains, and their volume fraction
Vv
was
P a g e | 150
relatively much lower compared with that of the Laves phase. In this case, for the given
volume fraction of the (Ti,Nb)(C,N) precipitates γf will be higher for a distribution within
the grain’s interior.
Figure 9.5. High resolution field emission scanning microscope image showing the
cracking of (Ti,Nb)(C,N) particles after impact testing the specimen at room
temperature. This specimen of Steel A was annealed at 850 °C followed by
quenching in water.
The stress causing cleavage fracture is, therefore, predicted to be the effective shear
stress:
τ e = τ y − τ i = k ys d
−
1
2
Equation 9.4
where τy is the shear yield strength and τi is the lattice friction stress. If yield occurs at
the head of the slip band by nucleation of a crack, it can be assumed that the friction
stress for yield is approximately the same as for the dislocations in the piled up slip band
[24]. The above Equation 9.4, does not explain why cleavage fracture is predominant in
the presence of precipitates (where the yield strength is high, but the (τy - τi ) may be
little different from its value at room temperature).
9.3.3 EFFECT OF PRECIPITATES IN THE EMBRITTLEMENT OF THIS STEEL
9.3.3.1 EMBRITTLEMENT AND THE COTTRELL’S APPROACH
Based on the results of Steel A as shown in Chapter 6, it was observed from Figure 6.12
that the Charpy impact energy values indicate decreasing embrittlement with increasing
P a g e | 151
annealing temperature up to 850 °C, which correlates with the Thermo – Calc®
predictions of a decreasing volume fraction of Laves phase as 825 C is approached.
The same observations where made by Sawatani et al.[8] working on a Ti and Nb
stabilised low C, N – 19%Cr – 2%Mo stainless steel whereby they have observed that
an increase in the volume fraction of the Laves phase has a significant impact in
promoting embrittlement of their steel, but the authors did not make any detailed
analysis of why the presence of the Laves phase deteriorates the impact toughness. In
the work done by Grubb et al.[140], they have suggested that the embrittling effect
arises from the flow stress increase associated with the α’ – precipitation (i.e. 475 °C
embrittlement) and this can be readily understood using the approach of Cottrell. This
approach provided a useful basis for understanding the micromechanics of brittle
fracture even though, strictly speaking, the model does not encompass all the practical
modes of crack initiation [38]. The conditions for cleavage fracture can be expressed by
the Cottrell equation:
1
2
σ f k d = C1G M γ f
s
y
Equation 9.5
where σf is the fracture stress, C1 is a constant related to the stress (~4/3 for a notched
specimen and 4 for a plain specimen [21]), and the rest are as mentioned previously.
The ferritic stainless steels generally display a substantial increase in lattice friction
stress σ0, upon rapid cooling. This increase manifests itself in an increase in the
ductile to brittle transition temperature (DBTT), in accordance with the Cottrell approach,
which predicts that the DBTT or brittle fracture with grain size as a variable, will occur
when σf = σy which would lead to [18,19]:
1
2
1
σ y k ys d 2 = k ys +σ i k ys d 2 = C1GM γ f
1
Equation 9.6
1
Therefore, if σ y k ys d 2 > C1GM γ f , then brittle fracture occurs, and if σ y k ys d 2 < C1GM γ f , then
ductile fracture will take place. The yield strength σy is related to the grain size through
the Hall – Petch equation, that is Equation 9.1. At 850 °C, assuming that the yield
strength σy = 323.5 MPa, k ys is approximately 469.8 MPa .µm1/2 and with a grain size
of 26.1 µm, it can be seen qualitatively that according to Equation 9.6, this steel will be
brittle at the temperatures above 850 °C where the grain size increases due to grain
growth.
However, there is not much of a difference in grain size between the as
P a g e | 152
received steel and the specimen of Steel A annealed at
850 °C or at lower
temperatures. Therefore, it cannot be confidently assumed that at temperatures below
850 °C, that the grain size plays a significant role in embrittling this steel.
The presence of large pre-existing cracks is, therefore, not a necessary prerequisite to
cause a brittle fracture [43]. It is obvious from Equation 9.6 that grain size has a direct
effect on the ductile – to – brittle transition temperature (DBTT).
A reasonable
relationship expressing the DBTT at which the fracture stress (σf) and yield strength
(σy) are equal for a given grain size d* has been proposed, i.e. for a given material,
there is a theoretical and experimental justification of a relationship between the grain
size and transition temperature :
DBTT = D + 1/ β ln d
*
1
2
Equation 9.7
where D is a constant and the slope (1/β) should also be constant. These variations
of the DBTT with the grain size that turn out to be consistent with Equation 9.7 have
been noted previously by different authors [19,43,141].
The same relationship was
tested in the current work, and it was observed that there is a tendency for the DBTT to
increase linearly with ln d
*
1
2
(see Figure 9.6). For AISI type 441 stainless steel studied
here, it was observed that over the range of grain sizes from 25.2 to 55.9 µm, there is a
shift of 35 °C in the transition temperature. In the work of Plumtree and Gullberg [43]
on Fe-25Cr ferritic stainless steels with 0.005 wt%C and 0.15 wt%N, the authors noted a
positive shift of 26 °C in a grain size increase of 33 to 95 µm. These authors have also
observed that by increasing the interstitial element contents of the steel, this will reduce
the sensitivity of the transition temperature to the grain size. This indicates that by
changing the composition of the steel and making it less impure (such as with Steel C –
E), the dependence of the DBTT to the grain size becomes less sensitive. According to
some authors [43,19], they have observed that from a qualitative viewpoint, the effect of
grain size on the ferritic stainless steel’s transition temperature may be somewhat less
than projected by Equation 9.7. But from the current work it was observed that a shift of
even 10 °C is significant to be considered when using Equation 9.7.
P a g e | 153
40
y = 87.659x - 139.25
R2 = 0.9874
35
DBTT (°C)
30
25
20
15
10
5
0
1.5
1.6
1.7
1.8
1.9
2
2.1
1/2
ln d1/2 (µ
µ m)
Figure 9.6. The plot of transition temperature versus {ln d1/2} of 441 ferritic stainless
steel, Steel A.
Generally, the toughness of ferritic stainless steel can be assessed in terms of the
DBTT, or a temperature below which Equation 9.6 appears to be satisfied for a given
material by the flow stress also being the fracture stress, i.e. brittle fracture without any
significant plastic yielding. The satisfaction of Equation 9.6 with decreasing temperature
is generally associated with the tendency for the flow stress to increase with decreasing
temperature. High strain rates and constraints to plastic flow at the root of a notch have
the effect of raising the flow stress and lowering the value of C1 and thus, promoting
satisfaction for Equation 9.6. According to the Cottrell model this phenomenon may be
explained by the fact that the locking of dislocations by interstitials in bcc materials
causes the flow stress to increase.
However, the solubility level of the interstitials
carbon and nitrogen in ferritic stainless steel is sufficiently low that it is rarely possible to
distinguish between solute embrittling effects and the effect of second-phase
precipitates.
9.3.3.2 EMBRITTLEMENT BY GRAIN BOUNDARY PRECIPITATES (THE SMITH’S MODEL)
The precipitates, in fact, become more important than the solute when the amount of
interstitial elements significantly exceeds the solubility limit.
The presences of
interstitials in an amount in excess of the solubility limit, serves to increase the DBTT still
further. This embrittling effect is closely linked to the number and size of precipitates
formed on the grain boundaries. Thick precipitate films act as strong barriers to slip
propagation across the grain boundaries and raise
k ys .
In this case, the Smith
approach [39] can be used for a growth controlled cleavage fracture that incorporates a
brittle precipitate on grain boundaries. Here a Laves phase particle of thickness Co at
P a g e | 154
the grain boundary dividing adjacent grains, is cracked by the dislocation pile – up of
length d (which may be related to the grain size itself). The important stress in causing
fracture is predicted to be the effective shear stress, and it may be analysed in a manner
similar to that for the Cottrell model to obtain a failure criterion in the absence of plastic
flow by dislocations:
1
 4Eγ f
2
σf > 

2
 π 1 − ν Co 
(
Equation 9.8
)
where E is Young’s Modulus, σf
is the fracture stress; γf
is the effective surface
energy of the interface between the ferrite grain and the Laves particle, and
ν
is
Poisson’s ratio. Assuming, the earlier estimated effective surface energy of γf ≈ 2.67
J/m2, with Gm = 8.6 x 1010 Pa and E = 8/3 Gm = 2.29 x 1011 Pa. For the specimen
annealed at 850 °C, the largest Laves phase precipitates measured were about 513 nm
in thickness. With ν = 0.28 and using Equation 9.8, this gives a fracture stress of
approximately 787 MPa. This value is much higher than the experimentally determined
yield strength value of 323.5 MPa, once more not predicting brittle fracture to occur.
Although not agreeing quantitatively with Smith’s model, the above relationship clearly
indicates qualitatively that coarser Laves particles with a higher value of C0 give rise to
lower fracture stresses, and also predict that σf is independent (directly) of the grain
size. However, in practice a fine grain size is often associated with thinner particles due
to a higher nucleation rate upon forming the particles, and the value of σf
is then
expected to be higher. Grain boundary Laves phase precipitation can be suppressed by
quenching from above the solution temperature when the interstitial content is low.
However, the resulting fine intergranular precipitation may increase the DBTT by
increasing the lattice friction stress (σo).
This model appears to explain at least
qualitatively the cause of the embrittling effect associated with the Laves phase
precipitation in the fracture of AISI type 441 ferritic stainless steel.
In work done by Sim et al.[3] on a Nb stabilised ferritic stainless steel, they have shown
that the high temperature strength decreases with increasing ageing time at
temperatures between 600 °C and 950 °C, and this was mainly due to the loss of the
solid solution hardening effect as the Nb is transferred from the matrix into the Laves
phase. They have also observed that the rate of reduction in yield strength as a function
of the Nb content at an ageing temperature of 700 °C, showed a linear relationship with
the Nb content in solid solution. In their conclusion, they observed that during ageing at
700 °C, Nb(C,N) and Fe3Nb3C precipitate out and decrease the high temperature
P a g e | 155
strength slowly while the precipitation of the Fe2Nb Laves phase at 700 °C decreases
the high temperature strength more abruptly. This is reasoned to be because of the
Lifshitz-Slyosov-Wagner (LSW) quasi-equilibrium coarsening rate of the Laves phase
that appears to be much faster compared to that of Nb(C,N). This was believed to be
due to the incoherent high energy interface between the Fe2Nb and the matrix, with the
Fe2Nb having a higher surface energy than that of Nb(C,N). Thus, it can be concluded
that coarse Fe2Nb precipitates are very detrimental to the high temperature strength of
type 441 stainless steels.
In work done by other researchers on ferritic stainless steels it was shown that the
tensile strength increases with annealing temperature above 1000 °C whilst the
percentage elongation decreases [8]. Therefore, from such an observation it can be
postulated that as the annealing temperature increases the amount of niobium in
solution also increases due to dissolution of precipitates and this should, therefore,
increase the yield strength due to solid solution hardening. But in the present work the
same effect was not observed since the yield strength decreased with increasing
solution treatment temperature (see Figure 6.16), which is indicative of a grain
coarsening effect. Yamamoto et al.[69] have observed in a ferritic steel with composition
Fe – 10at.% Cr – Nb with 0.5 – 5 at.%Nb, that it is possible to design a steel
strengthened by Fe2Nb Laves phase to give an excellent high temperature strength.
They have suggested that the strength is insensitive to Nb content, but an optimum
concentration of about 1 – 1.5 at.% Nb would provide a proper volume fraction of the
Laves phase to ensure good room temperature ductility.
None of these authors,
however, had attempted to make any correlation between the Laves phase volume
fraction present and high strain rate testing such as the Charpy impact test. It is of
course well known that one may find good strength and even good ductility values at
slow strain rate testing without any sign of brittleness and the practicality of
strengthening these alloys by Laves phase needs to be approached with caution.
It should be noted that both the fracture and the yield strength depend on the grain size,
and refining the grain size will result in an increase of these strengths. According to the
above Cottrell model on the DBTT, the effective temperatures for the DBTT from an
increased grain size exists when the flow stress practically equals the fracture stress
with very little measurable plastic strain. In this case, that will be at 950 °C where there
is no grain boundary Laves phase. From Figure 9.7, the grain size increase between
850 and 950 °C is about 1.6 times or a 60% increase. According to Equation 9.1, it can
be concluded that by increasing the grain size, both the fracture stress and the flow
P a g e | 156
stresses will be decreased, and this will certainly shift the DBTT upwards. In work by
Kinoshita [42] on 18Cr-2Mo steel and 26Cr–4Mo steel, the author however, found no
significant grain size dependence of the DBTT where the grain size was varied through
strain annealing. However, many other authors have reported such a relationship as
was also found here. In cases where there are grain boundary precipitates present,
such as with the specimens annealed below 850 °C, the Laves phase will tend to lower
the cleavage fracture stress. In this case a crack nucleus is formed by the separation of
the particle–matrix and this interfacial crack initiates transgranular fracture, but that has
not been observed to cause grain separation.
200
180
160
Grain size (µm)
140
120
100
80
60
40
20
0
A s re c . 8 5 0
875
900
925
950
975
1000
1025
1050
1075
1100
T e m p e ra tu re (° C )
Figure 9.7. Effect of annealing temperature above 850 °C on the grain size for the AISI
type 441 stainless steel, Steel A.
9.3.3.3 EFFECT OF COOLING RATE
When the specimens were slowly cooled from the solution treating temperature above
the Laves phase solvus temperature, precipitation of the Laves phase occurred
afterwards during slow cooling. Steels embrittled from such a treatment behave very
similarly to those treated below the solvus temperature [7]. This points to the usefulness
of constructing a time – temperature – precipitation (TTP) curve for Laves phase
formation and such a TTP curve was determined experimentally for Steel A and was
reported in Chapter 8. Heat treatments that accelerate the precipitation of the Laves
phase (that is, below 850 °C) decrease the resistance to crack initiation significantly
during dynamic loading. It can be assumed that the volume fraction of the Laves phase,
particularly on the sub- and grain boundaries, plays a significant role in affecting both the
transition temperature and the upper-shelf energy and that the degree of embrittlement
should increase as the volume fraction increases at lower annealing temperatures (see
P a g e | 157
Figure 5.4), as also predicted by the Smith model of embrittlement by grain boundary
phases in ferritic steels. The width Co (according to Equation 9.8) of the embrittling
second phase is related firstly, to the volume fraction Vv of the Laves phase but also
secondly, to the grain size, the latter through the grain boundary nucleation site
concentration with a larger sub- and grain size leading to a thicker C0 through less
available Laves phase nucleation sites. Therefore, both a higher volume fraction as well
as a larger sub- and grain size will then indirectly decrease the effective fracture stress
of the steel through a higher value of C0. This was indeed found in Figure 6.12 with
annealing temperatures decreasing from 850 to 600°C, raising the volume fraction of the
Laves phase.
Other authors have shown that there is an optimum annealing
temperature at which the quantity of the Laves phase Fe2Nb reaches a maximum, and
this temperature depends critically on the Nb content in the steel [7,4,99], no doubt
through non-equilibrium volume fractions according to a relevant CCT type phase
transformation diagram. In the current work reported here, therefore, some remaining
Laves phase on grain boundaries at 850°C could have contributed to a lower upper shelf
energy if compared to the specimen annealed at 950 °C where dissolution should have
been complete.
Further analyses on the specimens that were annealed at 600, 700 and 850 °C, showed
from the Thermo-Calc® results (see Figure 5.4), that the predicted weight fraction of the
Laves phase at 600 °C is about 0.92 %wt, at 700 °C is about 0.67 %wt and at 800 °C is
about 0.14 %wt and that there should be no Laves phase present at 850 °C. Note,
however, that Thermo-Calc® only gives an estimated weight fraction at full equilibrium
conditions and it does not reveal anything about the nucleation rate and the kinetics of
the Laves phase formation before equilibrium has been achieved. In the current study
as reported in Chapter 8 on the kinetics of the formation of the Laves phase, it was
observed that the precipitation kinetics of the Laves phase are much higher at 700 °C
than at either 850 °C or 600 °C, and this resulted in a double nose TTP curve. From this
it follows that at 700 °C after equal times at temperature, the volume fraction of the
Laves phase would be much higher than either at 850 or at 600°C and this will result in
the specimen having a lower upper shelf energy at 700°C than the specimen annealed
at 600 °C.
From the measurement of the DBTT at room temperature and shown in Figure 6.17,
which falls within the ductile to brittle transitions but not at the same points for the two
temperatures, the specimen annealed at 950 °C is at the very bottom of the transition
while at 850 °C it is in the middle. A direct comparison of the absolute values on the two
P a g e | 158
curves is, therefore, not feasible as they fall within different parts of the transition curve.
At 950 °C the grain size effect is overwhelming (even more significant than the effect
from the Laves phase) and that the addition of Laves phase to an already brittle steel at
slow cooling rates, brings in no measurable further embrittlement. Therefore, the full
effect of the Laves phase precipitation on the Charpy impact toughness could be seen
only in the specimen annealed at 850 °C where the grain size effect was minimal.
Therefore, samples that were solution treated at 850°C show a significant decrease in
the impact toughness with decreasing cooling rates that allow Laves phase to form
during slow cooling. Specimens that were solution treated at 950 °C, however, show
only a slight increase in impact toughness after slow cooling of 10 °C/sec, and then the
toughness levels off with a further increase in cooling rate (see Figure 6.18). Again, as
the cooling rates decrease the amount of the Laves phase that precipitates increases,
resulting in deterioration in the impact strength. Therefore, comparing the curves for 850
and 950 °C annealing in Figure 6.18, it can be concluded that grain size plays a
measurable role in the nucleation of the Laves phase, as already stated above. The
results of this study on the effect of cooling rate on the Charpy impact toughness of
ferritic stainless steels, therefore, largely agree with the work of other researchers
[7,8,99], i.e. with the cooling rate having a pronounced influence on the impact energy
transition of low interstitial ferritic stainless steels, with the lower cooling rates showing a
higher transition temperature.
9.4
RECRYSTALLISATION AND GRAIN GROWTH
Although grain growth experienced in this alloy at annealing temperatures above 850°C
is generally expected, the “sudden” significant increase in grain size by about 60% within
the temperature region 950 to 1000 °C cannot be readily explained by normal grain
growth theories and needs some discussion. The most obvious explanation of “grain
boundary particle unpinning by dissolution” in this temperature range is the immediate
and obvious candidate. Particles that had nucleated on the recrystallisation front may
impede the motion of the grain boundary due to Zener drag [142,143]. The resulting
retarding force ∆pppt exercised by these particles on the grain boundary is given by:
 3  Vv 
∆p ppt =  γ  
 4  r 
Equation 9.9
where γ = grain boundary free energy, Vv = volume fraction of the precipitates and r =
radius of the precipitate. Note that a significant retarding force is only achieved with a
P a g e | 159
large volume fraction and small particles. At 850 °C the retarding force may have a
relatively small estimated value of ∼2 x 104 J/m3 (i.e. small volume fraction and large
radius of the Laves phase) and a condition will not easily be reached where it effectively
blocks firstly, full recrystallisation and subsequently grain growth to a large degree. A
number of authors [143,144,] however, have proposed and observed an alternative
process of “continuous recrystallisation” in aluminium alloys in which particles occupy
mainly subgrain boundaries and these only become “released” after some coarsening of
the particles, where after “full recrystallisation” appears to have occurred in a gradual
process. That such a “continuous recrystallisation” process may have occurred also
here is quite possible if the TEM microstructures after annealing at 850 and 900°C are
compared with each other in Figure 9.8 (a) and (b). After annealing at 850°C subgrains
are still very much present and are heavily decorated by the last remnants of the Laves
phase while at 900°C, the subgrain boundaries have completely disappeared and the
grain boundaries are clean of any Laves or other phases.
The effective rate of
continuous recrystallisation is, therefore, determined by the rate of coarsening (which is
accompanied by the rate of disappearance of the smaller particles), of the Laves phase
particles. At 850 °C, recovery has taken place but no recrystallisation while at 900 °C
where full recrystallisation has taken place. This shows that at 850 °C, the last remnants
of the Laves phase particles are still effective in locking subgrain boundaries but not at
900 °C. Therefore, it cannot be concluded that the sudden increase in grain size at the
higher temperature of between 950 °C and 1000 °C was due to an unpinning effect of
the grain boundaries by the Laves phase, which had already taken place earlier between
850 and 900 °C.
(a)
(b)
Figure 9.8. TEM micrographs of the microstructures of the specimens from Steel A that
were annealed at (a) 850 °C and (b) 900 °C. Note that with the specimen that was
annealed at 900 °C, there were no grain boundary Laves phase precipitates.
P a g e | 160
An alternative reasoning of “Nb solute drag of sub- and grain boundaries” may be
considered in explaining this “sudden” 60% increase in grain size at annealing
temperatures between 950 and 1000 °C.
The rate of movement
v b = pd M b
vb
of a recrystallisation boundary in austenite is given by
, where pd is the driving force for recrystallisation and Mb is the boundary’s
mobility [145]. An increase in recrystallisation rate and subsequent grain growth may,
therefore, result either from an increase in the driving force pd or the mobility Mb of the
moving sub- or grain boundary. The quantitative theory of the solute drag effect on a
moving grain boundary during recrystallisation was originally formulated by Lücke and
Detert [59]. It was later modified by Cahn [60] and then by Lücke and Stüwe [61]. Since
then, this theory has been further refined by several authors [57,58,62]. The equation
proposed by Cahn on the rate of grain boundary movement as it is affected by solute
drag is [63,64]:
αXs
pd
1
1
=
=
+
v b MT M0 (1 + β 2v b2 )
Equation 9.10
where pd and vb are defined as above, MT is the overall mobility due to intrinsic plus
solute drag, M0 is the intrinsic grain boundary mobility in pure material, Xs is the atom
concentration of solute in the bulk metal, α is a term related to the binding energy of
solute to the grain boundary and β is a term related to the diffusion rate of the solute
near the grain boundary. The product (αXs) primarily governs the overall mobility MT
because
(1 + β 2 v b2 ) > 1 , while if
(1 + β 2v b2 ) >> 1 , then
occurs. The apparent mobility factor MT
MT ≈ M0 and no solute drag
is critically dependent on the binding energy
factor α of solute to the boundary and the solute concentration Xs in the bulk metal.
The possibility needs to be considered that a fundamental change from “solute plus
intrinsic drag by Nb atoms” (i.e. controlled by the overall mobility parameter MT) to one
of only “intrinsic drag” and now controlled only by the mobility parameter M0, may occur
within the temperature range 950 to 1000°C in this alloy. The question, however,
immediately then arises whether such a transition from “solute plus intrinsic drag” at
lower temperatures to one of only “intrinsic drag” at about 950 to 1000 °C would not
rather be a gradual one than a “sudden” one? That such a “sudden” departure from
solute plus intrinsic drag to only intrinsic drag occurs “suddenly” at critical concentrations
of the solute or at critical temperatures, has indeed been found by some authors [58].
P a g e | 161
Firstly, that such a sharp “break” in the effect of low level concentrations of solute in
solute drag on grain boundaries (such as the Nb in solution in this alloy) is to be
expected, was shown by Suehiro [57].
The author had investigated the effect of
temperature on solute drag during recrystallisation in an Fe-Nb alloy. The theoretical
analysis predicted that small Nb-additions to iron would reveal a two orders of
magnitude sudden decrease in the grain boundary velocity vb at 700ºC at a critical Nb
concentration of 0.15 %wt and this depends on the temperature and total driving force.
This retardation was found to be caused by the solute drag effect of Nb on grain
boundaries. In further work by Suehiro et al. [58], the author studied the effect of Nb on
the austenite to ferrite transformation in an ultra low carbon steel. Their results indicate
that there is a critical temperature where the rate of transformation changes drastically.
The transformation that occurs above and below the critical temperature are both
partitionless massive transformations.
The critical temperature was found to be
composition dependent, and for the 0.25% Nb alloy it was found to be 760 °C and for
0.75%Nb alloy it was 720 °C. This is in line with the earlier conclusion that the overall
mobility parameter MT is critically dependent on the binding energy α of the solute to
the grain boundary. This segregation energy of solute to a grain boundary actually
determines the critical concentration of solute at which the mode of recrystallisation may
change from interfacial mobility control (or M0) to one of solute drag (or MB).
Secondly, the analysis by Le Gall and Jonas [63] of solute drag by sulphur atoms in pure
nickel also showed that this transition from solute plus intrinsic drag to purely intrinsic
mobility of a grain boundary is not a gradual one but occurs at a critical temperature that
provides a “break” in the mobility versus inverse temperature relationship. Furthermore,
this critical transition temperature was found to be highly dependent on the
concentration of the “dragging” sulphur atoms in solution [64] and may, therefore, vary
with different Nb concentrations in solution which, in turn, will be determined by the alloy
content and/or by the heat treatments that have led to the Nb in solution.
Although the likelihood that solute drag by Nb atoms was responsible for the sudden
increase in grain size between 950 and 1000 °C, could of course, not be fully proven in
this study, it remains currently as the only plausible mechanism. The possibility of Nb
solute drag affecting the full recrystallisation and later grain growth of type 441 stainless
steels may, therefore, be a useful avenue of further study to confirm the above
possibility of solute drag being responsible for the “sudden” 60% increase in grain size at
about 950 to 1000 °C in this alloy.
P a g e | 162
CHAPTER TEN
DISCUSSIONS
TRANSFORMATION KINETICS MODELLING
10
10.1 INTRODUCTION
The purpose of this chapter is to model the transformation kinetics of the Laves phase in
AISI type 441 ferritic stainless steel during the isothermal annealing as a typical case.
This model takes into account the diffusion of Nb, and is based on the assumption that
the precipitation of the Laves phase is the only one taking place. This has been done
within the framework of nucleation and growth theory which leads naturally to the
coarsening process.
10.2 MODELLING IN KINETICS OF LAVES PHASE PRECIPITATION
10.2.1 NUCLEATION
Classical nucleation theory is used to estimate the nucleation rate for the Laves phase
precipitation.
There are several formulae for the nucleation rate, for simplicity the
general equation for calculating the rate of isothermal heterogeneous nucleation is given
by [108]:
kT
 ∆G * +Q 
N& = cN o
exp −

h
RT 

∆G * =
where c
Equation 10.1
{
4πγ 3 2 − 3 cos θ + cos 3 θ
3( ∆Gν + ∆ G ε ) 2
}
Equation 10.2
is the amount of Nb additions in the steel, No is the number of available
nucleation sites, k and h are the Boltzmann’s and Planck’s constants respectively, T
is the absolute temperature, ∆G* is the activation energy for nucleation and γ is the
surface energy per unit area of the precipitate – matrix interface. θ is the contact angle
between the Laves phase particle and the grain boundary. The critical embryo size r*
for forming a nucleus is given by:
r* =
− 4γ
( ∆Gν + ∆Gε ) 2 − cos θ − cos 2 θ
{
}
Equation 10.3
P a g e | 163
where ∆Gν is the chemical free energy change per unit volume and ∆Gε is the misfit
strain energy around the particle, which is often relatively small compared to the driving
force ∆Gν and, hence, was neglected in the calculations for ∆G* and r*.
10.2.2 GROWTH
To simplify this model, it is assumed that all the Laves phase precipitation occurs along
the grain boundaries and the growth and nucleation of the precipitates takes place
simultaneously as it has often been observed that nucleation is superseded by growth
very early within the precipitation process. The growth of the precipitates is assumed to
be controlled by the diffusion of the Nb in the ferrite matrix. Assuming that the particle
size is large enough for the interface between the precipitate and the matrix to be
considered, the position of the interface under one dimensional parabolic growth, which
corresponds to the radius of large spherical particles, is given by [5,99,146,147]:
r = Ω D t with Ω ≈ 2
Nb
c − cαβ
Equation 10.4
c βα − c
where DNb is the Nb diffusion coefficient in the ferrite matrix, t is the reaction time, Ω
is the supersaturation, c is the alloying concentration, cαβ and cβα are the equilibrium
concentrations in the ferrite matrix α and in the precipitate β (i.e. the Laves phase,
Fe2Nb), respectively. Assuming that cαβ is approximated to be the solute concentration
after ageing for a long period of time and is obtained from the Thermo – Calc®
calculations (Table 5.6), the supersaturation of Nb for each steel can be calculated and
is shown in Table 10.1.
Table 10.1. Calculated equilibrium mole fractions at the interface Fe2Nb / ferrite for Nb
between 600 and 800 °C in AISI type 441 ferritic stainless steel.
Temperature (°C)
Nb addition
(Mole fract.)
600
650
700
750
800
2.83 x 10
-3
Nb in precipitates
(Mole fract.)
-3
2.25 x 10
-3
2.00 x 10
-3
1.50 x 10
-3
1.00 x 10
-4
4.20 x 10
Nb in solid solution
(Mole fract.)
-4
5.80 x 10
-4
8.30 x 10
-3
1.33 x 10
-3
1.83 x 10
-3
2.41 x 10
Supersaturation
Ω
1.410
1.411
1.414
1.416
1.420
From this model, it is further assumed that the nucleation rate N& for the Laves phase is
a constant during precipitation in the ferrite matrix, and that soft – impingement does not
occur. At time t the radius of the particle nucleated at time t1 (0< t1 <t) is expressed
P a g e | 164
by Equation 10.6. Assuming a spherical particle, the growth rate Gr at time t1 is given
by:
Gr = 4πr 2
dr
d (t − t1 )
Equation 10.5
Therefore, the number of nuclei precipitated between t1 and t1 + dt1 is N& dt1. The rate
of the increase in volume of all the particles formed at t1 is then given by:
3
αβ  2
3

dV 8 2
3/2 c − c
πDNb
=
 βα
 N& (t − t1 )2
dt
3
 c − c 
Equation 10.6
To convert Equation 10.6 to the volume fraction of the precipitates Vv, this equation is
multiplied by
(c − c ) / (c
αβ
βα
)
− c . By considering the composition factor and integrating
Equation 10.8, the following equation can be obtained [146,147]:
1

5
αβ  2

 16 2
3/2 c − c
& 2
Vv = 1 − exp−
πDNb  αβ
 Nt 
c − c 
 15



Equation 10.7
Equation 10.9 above, is the typical Johnson-Mehl-Avrami-Kolmogorov (JMAK) equation
applied to describe the kinetics of the single precipitating Laves phase in an AISI type
441 ferritic stainless less and utilising Thermo – Calc® software to predict the driving
forces.
10.2.3 COARSENING
The coarsening rate of precipitates can be calculated using the classical theory of the
Ostwald ripening equation that is due to Lifshitz and Slyozov [113] and Wagner [114]
and is often called LSW coarsening. The LSW coarsening rate equation for diffusion
controlled coarsening on a grain boundary is given by:
r 4 − r o4 =
K 1γυ β Dgb c αβ δ gb
t
kT
Equation 10.8
where r is the average particle radius, r0 the initial average particle radius, γ is the
interfacial energy, K1 is a constant, Dgb is the diffusion coefficient down the grain
boundary, δgb is the width of the grain boundary, υβ is the molar volume, T is the
absolute temperature and
t
is the holding time at the isothermal heat treatment
temperature. The above coarsening Equation 10.10 is also only fully valid if all of the
precipitates were situated on grain boundaries and this is, of course, quite difficult to
P a g e | 165
achieve. One factor that strongly affects the coarsening stage is the interfacial energy,
even though the equilibrium concentration cαβ is of a primary concern as this may vary
by many orders of magnitude from system to system.
The coarsening rate is also
influenced by the distribution of the particles before the coarsening stage, which is also
determined by the nucleation and growth stages. However, in this work there are not
enough experimental data to describe quantitatively all of the details of the coarsening
process. The specific data that are needed for the further discussion are the changes in
the particle size distribution during high temperature ageing.
10.2.4 DIFFUSION COEFFICIENTS
The knowledge of the appropriate diffusion coefficients is required for all the diffusion
controlled transformation equations. The overall diffusion coefficient is given by:
 −Q 
D = Do exp

 RT 
Equation 10.9
where Do is the pre-exponent factor, Q is the activation energy for diffusion, R is the
gas constant and T is the absolute temperature. Fridberg et al. [148] reported the
intrinsic chemical coefficient D0 for the diffusion of several solutes in a ferritic matrix
and that were used in this work, see Table 10.2 below.
Table 10.2 Chemical diffusion coefficients and activation energies of elements in ferrite
[after Fridberg et al. 148].
Elements
Cr
Mo
Nb
2 -1
Do (m s )
-4
1.5 x 10
-4
1.5 x 10
-4
1.5 x 10
-1
Q (Jmol )
3
240 x 10
3
240 x 10
3
240 x 10
10.3 PARAMETERS REQUIRED FOR CALCULATIONS
The term ∆Gν, the chemical free energy change per unit volume of precipitate, is given
by:
∆G =
ν
∆G
υVv
Equation 10.10
where Vv is the equilibrium volume fraction of the Laves phase and υ is the molar
volume of the Laves phase, and ∆G is the molar free energy change of the precipitation
reaction. ∆G
can be obtained with from the thermodynamic calculations using the
P a g e | 166
Thermo-Calc ® software, see Figure 5.14 and Table 5.6.
However, it can also be
estimated from the precipitation reaction:
[Nb] + 2[Fe]
Fe2Nb
Equation 10.11
where [Nb] and [Fe] are the concentrations in solution in the ferrite. The activity of pure
solid Fe2Nb can be taken to be unity, assuming that the activities of the other elements
are equivalent to the concentrations. The solubility product is expressed as the solubility
of Nb with the following equation:
αβ
ln x Nb
=
with A =
A
+B
T
Equation 10.12
∆Go
and B = − ln[Fe] 2
R
and for the driving force ∆G:
αβ
∆G = ∆Go − RT ln x Nb
Equation 10.13
αβ
∆G can be estimated if the values of A and B are available. xNb
is the equilibrium
mole fraction of Nb in solution in ferrite.
Assuming that all of the carbon and nitrogen
precipitated as (Ti,Nb)(C,N) and the Fe concentration [Fe] is constant, it is possible to
calculate the volume fraction and concentrations in the ferrite at the interface of Nb.
Table 10.1 shows the calculated concentrations obtained from the Thermo-Calc®
αβ
and 1/T, the constants A, B
predictions. By plotting the relationship between ln xNb
and ∆Go in the Equation 10.14 were obtained, as illustrated in Figure 10.1 and the
results are given in Table 10.3.
P a g e | 167
-5.8
-6
-6.2
y = -6835.5x + 0.3614
R2 = 0.9962
ln xNbα β
-6.4
-6.6
-6.8
-7
-7.2
-7.4
-7.6
0.0009
0.00095
0.001
0.00105
0.0011
0.00115
0.0012
-1
1/T (K )
αβ
Figure 10.1. The relationship between ln x Nb
steel.
and T-1 for AISI type 441 ferritic stainless
Table 10.3. Calculated values of A, B and ∆Go for Fe2Nb in AISI type 441 ferritic
stainless steel.
-1
Values for mole fraction and
natural logarithm
Values for wt fraction
and natural logarithm
A (K )
B
-6835.5
0.3614
-2968.6
0.1569
-1
∆Go (Jmol )
-56830.35
The solubility product for the Laves phase Fe2Nb in the Steel A that was studied here
was obtained from Thermo-Calc® calculated data and the experimental data.
This
expression of the solubility product with weight fraction is given as:
log [Nb] = -2968.6/T + 0.1569
Equation 10.14
This expression is different from the one obtained by Fujita and co-workers [6] in their
recent work.
These authors were also working on niobium alloyed ferritic stainless
steels, and their expression is as follows:
log [Nb] = -3780.3/T + 2.4646
Equation 10.15
The thermodynamic expressions for the free energy change ∆G for the precipitation
reaction of Laves phase in type 441 ferritic steels were also calculated. This expression
with the mole fractions in the ferrite matrix, is given by:
αβ
∆G = – 56830.35 – RT{0.3614 + ln x Nb
}
Equation 10.16
P a g e | 168
From Equation 10.16, it was established that the calculated values for ∆G are about ten
times higher than the one obtained by Thermo-Calc® predictions. Therefore, it will be
reasonable to rather use the values from the Thermo-Calc® predictions, since they are
closer to the values from the literature [117].
10.4 CALCULATIONS
Before the calculations for the Laves phase transformation kinetics can be made, it
needs to be noted that the following assumptions were used; (i) no treatment of soft
impingement was introduced during the nucleation and growth stage, (ii) the particle –
size distribution at different nucleation times and growth rates remained constant and (iii)
that the precipitation proceeds in an Fe –Nb –C alloy system. Also, it was assumed that
Nb(C,N) does not form, and if it does, its volume fraction will be negligible and,
therefore, only the nucleation and growth of Laves phase is taking place. There are two
unknown parameters, the number density of nucleation sites No and the interfacial
energy γ. These parameters are treated as fitting parameters. The calculations were
carried out at the two temperatures of 700 and 800 °C. The diffusion coefficient and the
activation energy for the diffusion of Nb listed in Table 10.3 were used. The parameters
that were used for the Laves phase transformations are shown in Table 10.1.
Table 10.4. Parameters used in the calculations for Laves phase transformation.
Parameters
Boltzmann’s constant, h (J/K)
Planck’s constant, k (Js)
Gas constant, R (J/K/mol)
Avogadro constant, NA (mol-1)
3
Unit cell volume, Vcell (m )
Molar volume υm for Fe2Nb (m3/mol)
3
Density ρ of Fe2Nb (kg/cm )
Lattice parameter of bcc aFe (nm)
Contact angle, θ (°)
Values
1.38 x 10-23
6.626 x 10-34
8.314
6.022 x 1023
-28
1.57 x 10
-23
2.36 x 10
8.58
0.28
24
The activation energy for nucleation ∆G* and the critical particle size r* for the Laves
phase’s nucleation assuming the of the surface energy of the Laves phase
γ to be
0.331 Jm-2 [5], are shown in Table 10.5 below. It should be noted that both the values of
∆G* and r* are dependent on γ , and the values obtained at both temperatures are
realistic enough for the particle’s nucleation and growth.
P a g e | 169
Table 10.5. The calculated values of the activation energy for the nucleation ∆G*, and
the critical particle size r* for the Laves phase’s nucleation at the two
temperatures of 700 and 800 °C.
Temp. (°C)
700
800
∆G (J/mol)
-325
-90
∆Gν (J/mol)
-1.92 x 108
-2.30 x 108
∆G* (J)
8.28 x 10-18
5.72 x 10-18
r* (nm)
29.8
24.7
10.4.1 VOLUME FRACTION AND PARTICLE SIZE
Calculated volume fractions for the Laves phase during isothermal annealing at 700 and
800 °C are shown in Figure 10.2 and Figure 10.3 respectively, and compared with the
respective experimental observations. The results also show the calculated mean
particle’s radius with time, but there are no experimental values for comparison. One of
the draw backs of this approach is that it does not show an actual estimate of the
volume fraction within the steel, i.e the calculations show only the accumulated volume
fraction of the Laves phase during the transformation kinetics.
Therefore, for
comparison purposes the experimental values had to be normalised with respect to their
equilibrium volume fraction.
Figure 10.2 shows a similar comparison of the experimental and calculated values for
the change of volume fraction and the mean particle radius with time at 700 °C, using
reasonable values of No = 4.3 x 1014 m-3 and γ = 0.331 Jm-2. However, at 800 °C the
calculated results estimated that there are now ten time lower initial nucleation sites for
Laves phase than at 700 °C, that is No = 2.9 x 1013 m-3 and γ = 0.331 Jm-2. This
resulted in lowering the transformation kinetics of the Laves phase nucleation at 800 °C,
while the particle’s growth rate at 800 °C is much higher than at 700 °C, and this is
related to a higher diffusivity of the Nb.
Accordingly, a large particle, which has
nucleated within the early stages of precipitation, grows continuously even while the
volume fraction approaches equilibrium.
Because of the capillarity effect, the small
particles which nucleated late begin to dissolve from coarsening even though the large
one continues to coarsen. It has been demonstrated in the past that the mean particle
radius at first increases approximately parabolically with time as all the particles grow
1
from solid solution [105]. The mean radius changes at a rate of about
t 3 at longer
times as the number density of the particles decreases, and this is consistent with
expectations from coarsening theory.
P a g e | 170
When the experimentally determined activation energy (that is, Q = 211 kJ/mol within
the temperature range of 750 to 825 °C) for the Laves phase transformation kinetics is
employed in the model, this gives the initial nucleation site density No approximately
4.95 x 109 m-3 at 800 °C.
The sensitivity of the number density of the nucleation sites No and the interfacial
energy γ has been established in the niobium – alloyed ferritic stainless steel also for
M6C carbides by Fujita et al [5] and it was observed that the results are affected more by
the interfacial energy than by the number density. As an example, it was found that an
order of magnitude increase in No will lead to about a 20% decrease in mean particle
size and a corresponding increase in γ will lead to about a three fold increase in mean
particle size [5]. The parameters in this work were chosen as fitting parameters to
obtain a reasonable agreement with the experimental results.
7 x 1 0 -6
1 .2
C a lc u la te d v a lu e
E x p e rim e n ta l v a lu e s
M e a n p a rtic le s r a d iu s
6 x 1 0 -6
0 .8
5 x 1 0 -6
0 .6
4 x 1 0 -6
0 .4
3 x 1 0 -6
0 .2
2 x 1 0 -6
0 .0
1 x 1 0 -6
Mean particle radius (m)
Volume fraction
1 .0
0
102
103
104
105
T im e ( s e c )
Figure 10.2. Comparison between the experimental data and calculated isothermal
transformation curves for the Laves phase’s precipitation at 700 °C in the AISI type
441 ferritic stainless, with No = 4.3 x 1014 m-3 and γ = 0.331 Jm-2.
P a g e | 171
1 .2
C a lc u la te d v a lu e
E x p e rim e n ta l v a lu e s
M e a n p a rtic le s ra d iu s
1 2 x 1 0 -6
1 0 x 1 0 -6
0 .8
8 x 1 0 -6
0 .6
6 x 1 0 -6
0 .4
4 x 1 0 -6
0 .2
Mean particle radius (m)
Volume fraction
1 .0
1 4 x 1 0 -6
2 x 1 0 -6
0 .0
0
102
103
104
T im e (s e c )
Figure 10.3. Comparison between the experimental data and calculated isothermal
transformation curves for the Laves phase precipitation at 800 °C in the AISI type
441 ferritic stainless, with No = 2.9 x 1013 m-3 and γ = 0.331 Jm-2.
10.5 SUMMARY
This model agrees reasonably well with the experimental results and analyses of the
microstructures and also qualitatively with the basis of the classical heterogeneous
nucleation theory. For instance, it demonstrates that nucleation of the Laves phase on
grain boundaries (where the initial nucleation site density N0 is relatively much lower
than for homogeneous nucleation), is dominant at the higher annealing temperatures of
800 °C and above, where the undercooling ∆T and hence the driving forces ∆Gv for
nucleation are relatively low and the system then lowers its retarding forces through
grain boundary nucleation. As the temperature is decreased and the undercooling ∆T
and hence the driving forces are higher, however, heterogeneous nucleation on
dislocations becomes more significant, and hence, the initial number of the nucleation
sites N0 becomes higher. The initial number of nucleation sites No has a significant
impact on the transformation kinetics of the Laves phase, any increase in it results in a
decrease in the transformation kinetics, i.e. note the lower value in Figure 10.2 and 10.3
for t50% at 800°C with N0 = 2.9x1013 m-3 if compared to the higher value for t50% at
700°C with N0 = 4.3x1014m-3.
P a g e | 172
CHAPTER ELEVEN
CONCLUSIONS AND SUGGESTIONS FOR FURTHER WORK
11
11.1 CONCLUSIONS
The effects of some metallurgical and mechanical factors on the Charpy impact
toughness of AISI type 441 have been investigated. The following may be concluded:
•
The steel has an acceptable impact toughness of approximately 60 J after being
solution annealed at 850°C while below or above this heat treatment the impact
toughness decreases significantly, falling to values as low as only 10 J.
At
temperatures above 850 °C grain growth plays a dominant role in lowering the
impact toughness, but at temperatures below 850 °C, the low impact toughness is
associated with the precipitation of the intermetallic Laves phase on grain
boundaries. The results of the effect of annealing temperature on the Laves
phase precipitation agrees with the prediction made by Thermo – Calc®, whereby
an increase in Laves phase volume fraction resulted in lowering of the room
temperature impact toughness of this AISI type 441 ferritic stainless steel.
•
When comparing the results for Steel A from the Thermo-Calc® predictions and
the experimental values, Thermo-Calc® predicts that at 825 °C, the weight
fraction of the Laves phase should be zero, whereas the experimental results
prove that there is still about 0.031wt% of Laves phase present and this remnant
still exists even up at 850 °C and above.
•
Also, the Thermo-Calc® predictions and the experimental results show that the
volume fraction, the solvus temperature and the composition of the precipitates
are dependent on the composition of the steel. The stoichiometric composition,
also changes with the annealing temperatures.
•
The ductile – to – brittle transition temperature (DBTT) is dependent on both the
grain size and the presence of Laves phase precipitates on grain boundaries, but
the upper shelf energy appears to be only dependent on the presence of the
Laves phase precipitates. When a Cottrell approach was used to model the
effect of grain size on the embrittllement of the Steel A, it was observed that
according to the model, that at temperatures above and below 850 °C the steel
P a g e | 173
will still be brittle.
Therefore, it cannot be confidentially assumed that at
temperatures below 850 °C, the grain size plays a significant role in embrittling
this steel. It was also observed that over the range of grain size from 25.2 to 55.9
µm in Steel A, there is a positive shift of 35 °C in the transition temperature. It
was found that this sensitivity of the DBTT on grain size can be lowered by
increasing the alloy content of the steel, making it less pure (such as in Steels C
to E).
•
When applying the Smith model of brittle grain boundary carbides on the
embrittlement of Steel A, the predicted outcome and the experimental results do
not agree quantitatively although qualitatively some resemblance between what is
predicted and what was found was observed. For instance, the width Co of the
embrittling second phase is related, firstly, to the volume fraction Vv of the Laves
phase but also secondly, to the grain size, the latter through the grain boundary
nucleation site concentration with a larger sub- and grain size leading to a thicker
Co through less available Laves phase nucleation sites. Therefore, both a higher
volume fraction as well as a larger sub- and grain size will then decrease the
effective fracture stress of the steel though a higher value of Co. Due to the low
volume fractions of the (Ti,Nb)(C,N) and M6C type Fe3Nb3C carbides, it was
concluded that they don’t have any significance in the embrittllement in these
steels. In this case, for any given volume fraction of the (Ti,Nb)(C,N) precipitates,
the effective surface energy of the crack γf will be higher for a distribution within
the grain’s interior.
Therefore, the possibilities of transgranular or cleavage
fracture occurring are minimised.
•
The cooling rate has a pronounced influence on the severity of the Laves phase
embrittlement of these steels after annealing at 850 °C. After cooling from 950 °C
with an already embrittled structure from grain growth, the introduction of Laves
phase during slow cooling does not introduce a significant and measurable
additional embrittlement.
In steels of this type, the embrittlement caused by
excessive grain growth is, therefore, overriding.
•
Annealing at temperatures above 850 °C in this alloy, generally expected grain
growth was experienced, but, there was a “sudden” significant increase in grain
size of about 60% within the temperature region 950 to 1000 °C. As no grain
boundary pinning was evident at these temperatures, it was postulated that Nb
P a g e | 174
solute drag may be responsible for this “sudden release” of the grain boundaries,
affecting the full recrystallisation and later grain growth of type 441 stainless
steels at lower temperatures.
On the transformation kinetics of the Laves phase precipitates, the following could be
concluded:
•
A time – temperature – precipitation (TTP) diagram for the Laves phase that was
determined from the transformation kinetic curves appears to show two classical
C noses on the transformation curves, i.e. the first one occurring at higher
temperatures of about 750 to 825 °C and the second one at much lower
temperatures, estimated to possibly be in the range of about 650 to 675 °C. The
transmission electron microscopy (TEM) analyses show that there are two
independent nucleation mechanisms that are occurring at these temperatures. At
the lower temperatures of about 600 °C, the pertaining nucleation mechanism is
principally on dislocations and as the temperature is increased to above 750 °C,
grain boundary nucleation becomes more dominant.
•
The effects of grain size and Mo additions on the transformation kinetics of the
Laves phase, showed that, by increasing the grain size or adding the alloying
element Mo, that these lower the rate of formation of the Laves phase.
•
By assuming an interfacial energy of 0.331 Jm-2 for the Laves phase, the kinetic
model predicts that at 700 °C and 800 °C the initial number of nucleation sites No
is 4.3 x 1014 m-3 and 2.9 x 1013 m-3 , respectively. These parameters fit well with
the experimental results. When the experimentally determined activation energy
of Q = 211 kJ/mol for the Laves phase transformation kinetics is employed in the
model, this gives the initial nucleation site density No approximately as 4.95 x
109 m-3 at 800 °C, indicating that there might be fewer nucleation sites than
expected.
This would agree well with heterogeneous nucleation on grain
boundaries with a much lower nucleation site density instead of homogeneous
nucleation.
•
The solubility products of the Laves phase Fe2Nb in the AISI type 441 ferritic
stainless steel obtained from Thermo-Calc® calculated data and the experimental
data, is given as:
log [Nb] = -2968.6/T + 0.1569
P a g e | 175
11.2 SUGGESTIONS FOR THE FURTHER WORK
The effect of the steel’s composition on the toughness of AISI type 441 ferritic stainless
steel still needs to explored further, especially the optimum Mo content that is required to
lower both the solvus temperature and the volume fraction of the Laves phase
precipitation. Also, the minimum Nb content that will be optimum to stabilise the type
441 ferritic stainless steels, whilst at the same time providing the desired heat resistant
properties, and lowering the Laves phase content, also needs to be investigated further.
The possibility of Nb solute drag affecting the full recrystallisation and later grain growth
of type 441 stainless steels will be a useful avenue of further study to confirm possibility
of the disappearance of Nb solute drag being responsible for the “sudden” 60% increase
in grain size at about 950 to 1000 °C in this alloy.
Validation of the modelling of the transformation kinetics for the Laves phase that will
improve the precision of any predictions in future:
1. specific characteristics of the nucleation sites;
2. measurement of the interfacial energy; and
3. modelling of precipitation on grain boundaries.
P a g e | 176
APPENDIX A
SYMMETRY (TRANSLATIONAL AND SPACE GROUP), UNIT CELL DATA AND
ATOMIC POSITION PARAMETERS
P a g e | 177
P a g e | 178
APPENDIX B
THERMODYNAMIC MODELLING USING THERMO-CALC® SOFTWARE
Definitions of the Abbreviated Parameters Used In the Calculations
go data
sw-dat
tcfe3
def-sys
l-s
get
go p-3
s-c
l-c
c-e
s-a-v
t
s-d-a
set – lab
def-mat
– go to thermodynamic database module
– switch database
– Thermo-Calc® Steels/Fe-alloys database
– define system
– list system
– get data from the database
– go to module Poly 3
– set conditions
– list condition
– compute equilibrium
– set axis variable
– temperature
– set diagram axis
– set label
– define materials
TEMPLATE USED TO CALCULATE EQUILIBRIUM THERMODYNAMIC PARAMETER OF TYPE 441
go data
sw-dat
tcfe3
def-sys
fe c mn co cr b v s si ti ni n al p cu nb o
l-s
CONSTITUENT
reject phase*
restore phase liquid, fcc, bcc, hcp, laves_phase, m6c, m23c6
get
go p-3
s-c n=1, p=101325, t=1773
s-c w(c)=0.00012
s-c w(mn)=0.0051
s-c w(co)=0.0003
s-c w(cr)=0.1789
s-c w(b)=0.000004
s-c w(v)=0.0012
s-c w(s)=0.00001
s-c w(si)=0.005
s-c w(ti)=0.00153
s-c w(ni)=0.0019
s-c w(n)=0.000085
s-c w(al)=0.00009
s-c w(p)=0.00025
s-c w(cu)=0.0008
P a g e | 179
s-c w(nb)=0.00444
s-c w(o)=0.000076
l-c
@&
c-e
s-a-v
1
t
473
2000
3
save MPO_3533603
step
normal
post
s-d-a x t-c
s-d-a y np(*)
*
plot SCREEN
set-tit
Intermetallic phases
set-lab
d
s-t-m-s
x
s-s x n 450 1800
plot
SCREEN
set-inter
P a g e | 180
TEMPLATE USED TO CALCULATE THE ISOPLETHS OF TYPE 441
go p-3
def-mat
tcfe3
fe
Y
c
.012
mn .51
co .03
cr 17.89
v .12
si .5
ti .153
n .0085
nb .444
ni .19
mo .5
s .001
b .0004
al .009
p .0025
cu .08
o .0076
1000
*
liquid, bcc_a2, fcc_a1, laves_phase_c14, m23c6, m3c2, m6c, sigma,
NONE
Y
N
l-e
SCREEN
VWCS
s-a-v 1 w(c)
0
.03
2.5e-4
s-a-v 2 t
200
1800
40
save
map
post
s-p-f
1
plot
SCREEN
s-d-a w-p c
s-d-a x w-p c
s-d-a y t-c
plot
P a g e | 181
SCREEN
s-s x n 0 1
s-s y n 200 1600
plot
SCREEN
s-lab n
plot
SCREEN
s-lab b
plot
SCREEN
PLOT,,,,,;
P a g e | 182
REFERENCE
1. http://www.sassda.co.za/sectors/catcon.html/ visited on 19 June 2006
2. http://www.aksteel.com; Aksteel Product Data Bulletin 441 Stainless Steel
3. G-M. Sim, J. C. Ahn, S. C. Hong, K. J. Lee & K. S. Lee; Effect of Nb Precipitate on the High
Temperature Strength in Nb Containing Ferritic Stainless Steels; Materials Science and
Engineering, A396 (2005) pp. 159-165
4. N. Fujita, K. Ohmura, M. Kikuchi & T. Suzuki; Effect of Nb on High Temperature Properties for
Ferritic Stainless Steels; Scripta Materialia, vol. 35, No. 6, (1996) pp. 705-710
5. N. Fujita, K. Ohmura & A. Yamamoto; Changes of Microstructure and High Temperature
Properties During High Temperature Service of Niobium Added Ferritic Stainless Steels;
Materials Science and Engineering, A351 (2003) pp. 271-281
6. N. Fujita, K. Ohmura & M. Kikuchi; Expressions for Solubility Products of Fe3Nb3C carbide and
Fe2Nb Laves Phase in Niobium Alloyed Ferritic Stainless Steels; ISIJ International, vol. 43,
No. 12, (2003) pp. 1999-2006
7. A. Miyazaki, K. Takao & O. Furukimi; Effect of Nb on the Proof Strength of Ferritic Stainless
Steels at Elevated Temperature, ISIJ International 42 (2002) pp. 916 – 920
8 .T. Sawatani, S. Minamino & H. Morikawa; Effect of Laves Phase on the Properties of Ti and
Nb Stabilised Low C, N-19%Cr-2%Mo Stainless Steel Sheets; Transaction ISIJ, Vol. 22
(1982) pp.172-180
9. A. Bjärbo; Computer Simulation of Growth and Coarsening of Laves Phase in a Modified 12%
Chromium Steel; Scandinavian Journal Of Metallurgy; 32 (2003) pp. 94-99
10. D.G Morris, M. A. Muñoz-Morris & C. Baudin; The High-Temperature Strength of Some
Fe3Al Alloys; Acta Materialia, 52 (2004) pp. 2827 - 2836
11. T. Sawatani, S. Minamino & H. Morikawa; Effect of Laves Phase on the Properties of Ti and
Nb Stabilised Low C, N-19%Cr-2%Mo Stainless Steel Sheets; Tetsu-to-Hagane Vol. 65, No 8
(1979) pp. 1194-1203
12 G. Dimmler, P. Weinert, E. Kozeschnik & H. Cerjak; Quantification of the laves phase in
advanced 9–12% Cr steels using a standard SEM, Materials Characterization 51 (2003) pp.
341– 352
13 R. P. Reed; Nitrogen in austenitic stainless steels, Journal of Metals (1989) pp.16 – 21
14 J. W. Simmons; Overview: High nitrogen alloying of stainless steels, Materials Science &
Engineering. A207 2 (1996) pp.159 – 169
P a g e | 183
15 W. E. Stumpf; : “Course materials for NFM 700, Chpt 4- Precipitation of intermetallics
compounds in higher alloyed steels”, Phase transformation in metals and their alloys,
University of Pretoria
16 A. L. Schaeffler; Constitution diagrams for stainless steel weld metal -3a: Welding and
Joining, Metal Progr. Databook, 6 (1973) pp. 207
17 Computational Thermodynamics; Calculation of Phase Diagrams using the CALPHAD
method.
Iron-Chromium
(Fe-Cr)
Phase
Diagram.
http://www.calphad.com/iron-
chromium.html, visited on 4 July 2007
18 A. C. T. M. Van Zwieten & J. H Bulloch; Some considerations on the toughness properties of
ferritic stainless steels – A brief review, International Journal of Pressure Vessel & Piping 56
(1993) pp. 1 - 31
19 R. N. Wright; “Toughness of Ferritic Stainless Steels”, Toughness of ferritic stainless
steels, ASTM STP 706, R. A. Lula, Ed., American Society for Testing and Materials, (1980)
pp. 2 – 33
20 N. Ohashi, Y. Ono, N. Kinoshita & K. Yoshioka; “Effect of metallurgical and mechanical
factors on Charpy impact toughness of extra-low interstitial ferritic stainless steels,”
Toughness of Ferritic Stainless Steels, ASTM STP 706, R. A. Lula, Ed., American Society
for Testing and Materials, (1980) pp. 202 – 220
21 A. Plumtree & R. Gullberg; “Influence of interstitial and some substitutional alloying
elements,” Toughness of Ferritic Stainless Steels, ASTM STP 706, R. A. Lula, Ed.,
American Society for Testing And Materials, (1980) pp. 34 – 55
22 J. R. Wood; “Effect of residual elements and molybdenum addition on annealed and welded
mechanical properties of 18Cr ferritic stainless steels,” Toughness of Ferritic Stainless
Steels, ASTM STP 706, R. A. Lula, Ed., American Society for Testing And Materials, (1980)
pp. 145 – 160
23 R.D Campbell; Ferritic stainless steel welding metallurgy, Key Engineering Materials, 69 &70
(1992) , pp. 167 – 216
24 T. J. Marrow; The fracture mechanism in 475 °C embrittled ferritic stainless steels, Fatigue &
Fracture of Engineering Materials & Structure, Vol.19 no. 7 (1996) pp. 919 – 933
25 G. Restrepo Garcé, J. Le Coze, J. L. Garin, & R. L. Mannheim; σ-Phase precipitation in two
heat-resistant steel- Influence of carbides and microstructure, Scripta Materialia, 50 (2004)
pp. 651–654
P a g e | 184
26 T-H. Lee, C-S. Oh, C. G. Lee, S-J Kim & S. Takaki; Precipitation of σ-phase in high-nitrogen
austenitic 18Cr-18Mn-2Mo-0.9N stainless steel during isothermal aging, Scripta Materialia, 50
(2004) pp. 1325 – 1328
27 V. Kuzucu, M. Aksoy & M. H. Korkut; The Effect of strong carbide-forming elements such as
Mo, Ti, V And Nb on the microstructure of ferritic stainless steel, Journal of Materials
Processing Technology, 82 (1998) pp. 165 – 171
28 K. Kimura, M. Abe, M. Tendo & T. Senuma; Influences of heating and coiling temperatures
on recrystallisation during hot-rolling process in Ti added high-purity ferritic stainless steel,
International congress Stainless Steel ’99: Science and Market, 3rd European congress
Proceedings – Vol 2: Innovation in Processes and Products, Italy 6 -9 June 1999, pp 77 - 84
29 Y. Hosoi, N. Wade, S. Kunimitsu & T. Urita; Precipitation behavior of Laves phase and its
effect on toughness of 9Cr – 2Mo Ferritic – Martensitic Steel, Journal of Nuclear Materials,
141 – 143 (1986) pp. 461 – 467
30 V.V Satyanarayana, G. Madhusudhan Reddy & T. Mohandas; Dissimilar metal friction
welding of austenitic – ferritic stainless steels, Journal of Materials Processing Technology,
160 (2005) pp. 128 – 137
31 M. Hua, C. I. Garcia, G. Thither & A. J. DeArdo; Dual–stabilized ferritic stainless steels for
demanding applications such as automobile exhaust systems, 38th Mechanical Work and
Steel Processing Conference; Procedings,
Vol. XXXIV, Cleveland, Ohio, October 13 –
16,(1996) pp. 453 – 457
32 A.J. De Ardo, Influence of Niobium and Tantalum on Stainless Steel,
International
Symposium on Tantalum and Niobium, Orland, Florida; USA; 7 – 9 Nov 1998, pp. 435 - 465
33 S.R Keown; Niobium in Stainless Steels, 1946, Niobium Technical Report, NbTR – 09/86,
ISSN 0101 – 5936
34 J. D. Redmond; “Toughness of 18C-2Mo ferritic stainless steels,” Toughness of Ferritic
Stainless Steels, ASTM STP 706, R. A. Lula, Ed., American Society for Testing And
Materials, (1980) pp. 123 – 144
35 A. Valiente, J. Ruiz & M. Elices; A probabilistic model for pearlite – induced cleavage of plain
carbon structural steel, Engineering Fracture Mechanics, 72 (2005) pp. 709 – 728
36 A. N. Stroh; The formation of cracks as a results of plastic flow, Proceeding of Royal Society
A, 223 (1954) pp. 404 – 414
37 A. N. Stroh; A theory of the fracture of metals, Advanced in Physics, 6(24) (1957) pp. 418 –
465
P a g e | 185
38 A. H. Cottrell, Theory of brittle fracture in a steel and similar metals, Transactions of the
Metallurgical Society of AIME, (1958) pp. 192 39 E. Smith; The nucleation and growth of cleavage microcracks in mild steel. In: Proceedings of
the conference on physical basis of fracture. London: Institute of Physics and Physics
Society, (1966) pp. 34 – 46
40 G. G. Chell, Developments if fracture mechanics – 2: the mechanics and mechanisms of
fracture in metals, Applied Science publishers, London 1981, pp 104
41 J. F. Knott, Fundamentals of Fracture mechanics, Butterworth, London, 1973
42 N. Kinoshita; Brittle cracking in extra low interstitial ferritic stainless steels, Stainless Steel
’84, Chalmers University of Technology, Goteborg; Sweden, 1984 September 3 – 4 , pp. 43–
49
43 A. Plumtree & R. Gullberg; “Embrittlement of a continuously cooled Fe – 25Cr alloy”,
Metallurgical Transactions A, 7A (1976) pp. 1451 – 1458
44 H. Liebowitz; Fracture: An Advanced Treatise, Volume VI: Fracture for metal, Academic
Press , New York 1969
45 R. T. J. Whillock, R. A. Buckley & C. M. Sellar; The influence of thermomechanical
processing on recrystallisation and precipitation in austenitic alloys with particular reference to
the effects of deformation and ageing conditions, Materials Science and Engineering A276
(2000) pp. 124 – 132
46 L. Columbier & J. Hochmann; Stainless and Heat resisting steel, Paraguay : AsunciBon, 1968
47 F. B. Pickering; Physical metallurgy and the design of steels, Materials Science Series,
Applied Science Publishers Ltd, 1978, pp 78
48 J. E. G. Gonzalez; Study of the effect of hot rolling processing parameters on the variability of
HSLA steels, MSc Dissertation, University of Pittsburgh, 2002
49 H. F. G. de Abreu, A. D. S. Bruno, S. S. M. Tavares, R. P Santos & S. S. Carvalho; Effect of
high temperature annealing on texture and microstructure on an AISI – 444 ferritic stainless
steel, Materials Characterization, 57 (2006) pp. 342 – 347
50 Y. Koyama, A. Takahashi, T. Shimada, N. Fujita & S. Maeda; Ferritic stainless steel for
exhaust equipment of vehicle, United States Patent, Patent number: 5,843,370 1 December
(1998)
51 Y. Uematsu, N. Hiramatsu & S. Nakamura; Heat resisting ferritic steel excellent in low
temperature toughness, weldability and heat resistance, United State Patent, Patent number:
5,302,214, 12 April 1994
P a g e | 186
52 D. C. Oliver & M. Sephton; External corrosion resistance of steel and ferritic stainless steel
exhaust systems, The Journal of the South African Institute of Mining And Metallurgy, March
(2003) pp. 93 - 100
53 Y. Inoue & M. Kikuchi; Present and future of stainless steel for automotive exhaust system,
Nippon Steel Technical Report, No. 88 July (2003) pp. 62 – 69
54 A. Miyazaki, T. Yokota & F. Togashi; Hot – rolled ferritic steel for motor vehicle exhaust
members, United States Patent, Patent number: 5,792,285, 11 August 1998
55 W. Gordon & A. van Bennekom; Review of stabilisation of ferritic stainless steels, Materials
Science & Engineering, A351 (2003) pp. 126 – 131
56 C. Jian-chun, L. Qing-you, Y. Qi-long & S. Xin-jun; Effect of niobium on isothermal
transformation of austenite to ferrite in HSLA low – carbon steel, Journal of Iron and Steel
Research, International, 14(3) (2007) pp. 51 – 55
57 M. Suehiro; An analysis of the solute drag effect of Nb on recrystallisation of ultra low carbon
steel, ISIJ International, 38 (6) (1998) pp. 547 – 552
58 M. Suehiro, Z.-K. Lui & J. Ågren; Effect of niobium of massive transformation in ultra low
carbon steels: A solute drag treatment, Acta Materialia, 44 (1996) pp. 4241 – 4251
59 K. Lücke & K. Detert, A quantitative theory of grain boundary motion and recrystallisation in
metals in the presence of impurities, Acta Metallurgica, 5 (1957) pp. 628 – 637
60 J. W. Cahn; The impurity-drag effect in grain boundary motion, Acta Metallurgica, 10 (1962)
pp. 178 – 798
61 K. Lücke & H. P. Stüwe; Theory of impurity controlled grain boundary motion, Acta
Metallurgica, 19 (1971) pp. 1087 – 1099
62 M. Hillert; Solute drag in grain boundary migration and phase transformations, Acta
Materialia, 52 (2004) pp. 5289 – 5293
63 R. Le Gall & J. J. Jonas; Solute drag effect during the dynamic recrystallisation of Nickel,
Acta Materialia, 47 (1999) pp. 4365 – 4374
64 W. Stumpf & K. Banks; The hot working characteristics of boron bearing and a conventional
low carbon steel, Materials Science and Engineering A 418 (2006) pp. 26 - 94
65 M. Hillert & B. Sundman; A treatment of the solute drag on moving grain boundaries and
phase interfaces in binary alloys, Acta Metallurgica, 24 (1976) pp. 731 – 743
66 M. Hillert & B. Sundman; A solute drag of the transition from diffusion controlled to
diffusionless solidification, Acta Metallurgica, 25 (1977) pp. 11 – 18
P a g e | 187
67 J-H. Schmitt; Some examples of stainless steel use in the automotive industry, Key
Engineering Materials, 230 – 232 (2002) pp. 17 – 22
68 Technical Data Blue Sheet-Stainless Steel Type 441, AL 441HP Alloys. Allengheny Ludlum
Corporation
69 K. Yamamoto, Y. Kimura, F-G. Wei & Y. Mishima; Design of Laves phase strengthened
ferritic heat resisting steel in Fe-Cr-Nb(-Ni) system, Materials Science and Engineering A329
– 331 (2002) pp. 249 – 254
70 Columbus Stainless, Personal Communications
71 Y.A. Chang, S. Chen, F.Zhang, X. Yan, F. Xei, R. Schmid-Fetzer & W. A. Oates; Phase
diagram calculation: past, present and future; Progress in Materials Science; 49 (2004) pp.
313 - 345
72 C. Wolverton, X-Y. Yan, R. Vijayaraghavan & V. OzoliĦš; Incorporating first-principles
energetics in computational thermodynamics approaches, Acta Materialia, 50 (2002) pp. 2187
– 2197
73 A. Schneider & G. Inden; Simulation of the kinetics of precipitation reactions in ferritic steels,
Acta Materialia 53 (2005) pp. 519 – 531
74 J–O. Anderson, T. Helander, L. Höglund, P. Shi & B. Sundman; THEMO-CALC & DICTRA,
Computational Tools for materials science, Chalphad Vol. 26, no.2 (2002) pp. 273 – 312
75 Jyrki Miettinen; Thermodynamic description of solution phases of systems Fe-Cr-Si and FeNi-Si with low silicon contents and with application to stainless steels, Calphad, Vol 23, No2
(1999) pp. 249 – 262
76 Jyrki Miettinen; Approximate Thermodynamic solution phases data for steels, Calphad, Vol
22, No 2 (1998) pp 275 – 300
77 J. D. Robson & H. K. D. H. Bhadeshia; Kinetics of precipitation in power plant steels,
Calphad, Vol 20, No4 (1996) pp. 447 – 460
78 C.T. Liu, J. Stringer, J. N. Mundy, L. L. Horton & P. Angelini; Ordered intermetallic alloys: an
assessment, Intermetallics 5 (1997) pp. 579 - 596
79 J. H. Zhu, L. M. Pike, C. T. Liu & P. K. Liaw; Point defects in binary Laves phase alloys, Acta
Materialia Vol. 47 No. 7 (1999) pp. 2003 - 2018
80 K. S. Kumar & P. M. Hazzledine; Polytypic transformations in Laves phase, Intermetallics 12
(2004) pp. 763 – 770
81 J.H Westbrook & R.L Fleischer; Intermetallic compounds: principles and practice, Volume 1 –
principles, John Wiley & Sons, England, 1995, pp. 107 – 109
P a g e | 188
82 C. T. Liu, J. H. Zhu, M. P. Brady, C. G. McKamey & L. M. Pike; Physical metallurgy and
mechanical properties of transition-metal Laves phase alloys, Intermetallics 8 (2000) pp. 1119
– 1129
83 Pearson’s Handbook Of Crystallographic Data For Intermetallic Phases Edited by P. Villars &
L.D Calvert: vol. 1. American Society for Metals, 1985, pp. 568, 605, 760
84 W.B. Pearson: A Handbook of Lattice Spacing and Structure of Metals and Alloys, Chapter
III: Structure determination and lattice spacing in the theory of alloy formation, Pergamon
Press Ltd, London,1958, , p30
85 F. Stein, M. Palm, G. Sauthoff; Structure and stability of laves phases. Part I. Critical
assessment of factors controlling laves phase stability, Intermetallics 12 (2004) pp. 713–720
86 S. Hong & C. L. Fu; Phase stability and elastic moduli of Cr2Nb by first-principles calculations,
Intermetallics 7 (1999) pp. 5 – 9
87 T. F. de Andrade, A. M. Kliauga, R. L. Plaut & A. F. Padilha; Precipitation of Laves phase in a
28%Cr-4%Ni-2%Mo-Nb superferritic stainless steel, Materials Characterization 59 (2008) pp.
503-50
88 F. Gauzzi & B. Verdini; Analysis of precipitates in low-interstitial 18Cr-2Mo ferritic stainless
steels; Metallurgical Science and Technology 2 (1984) pp. 48 – 53
89 Y. Murata, M. Kamiya, T. Kuneida, A. M. Abdel-Daiem, T. Koyama, M. Morinaga & R.
Hashizume; Dependence of solvus temperature of the Laves Phase on (Mo + W + Re)
contents in high Cr ferritic steels, ISIJ International, 45 (2005) pp. 101 - 106
90 G. J. Cocks & D. W. Borland; The orientation relationship and morphology of Fe2Nb
precipitates in ferrite, Metal Science 9 (1975) pp. 384 – 389
91 K. Yamamoto, Y. Kimura & Y. Mishima; Effect of matrix substructures on precipitation of the
Laves phase in Fe – Cr – Nb – Ni system, ISIJ International 43 (2003) pp. 1253 – 1259
92 G. R. Speich; Precipitation of Laves phase from iron-niobium (columbium) and iron-titanium
solid solutions, Transaction of the Metallurgical Society of AIME 224 (1962) pp. 850 - 858
93 K. Miyahara, J-H. Hwang & Y. Shimoide, Ageing phenomena before the precipitation of the
bulky Laves phase in Fe-10%Cr ferritic alloys, Scripta Metallurgica et Materialia, Vol. 32 No.
12 (1995) pp. 1917 – 1921
94 Y. Murata, T. Koyama, M. Morinaga & T. Miyazaki; Prediction of the Laves phase Morphology
in Fe-Cr-W-C quaternary steel with the aid of system free energy concept, ISIJ International,
42 (12) (2002) pp. 1423 – 1429
P a g e | 189
95 W.E Stumpf, Phase Transformation in Metals and their Alloys, Course materials for NFM
700., Chapter 1: Nucleation and growth of Precipitates in a Supersaturated Solid Solution,
2006, University of Pretoria
96 F. R. N. Nabarro, The strain produced by the precipitation in alloys,
Proceedings of the
Royal Society of London. Series A, Mathematical and Physical Science, Vol. 175, No. 963
(Jul.18, 1940), pp.519 - 538
97 H.-J. Rajek; “Computer simulation of precipitation kinetics in solid materials and application to
the complex power plant steel CB8” PhD Thesis, Graz University of Technology, Austria,
2005
98 D. A. Porter & K. E. Easterling, Phase transformations in metals and alloys, 2nd edition,
London: Chapman and Hall (1992), p. 147.
99 N. Fujita, H. K. D. H. Bhadeshia & M. Kikuchi, Precipitation Sequence in Niobium-Alloyed
Ferritic Stainless Steel. Modeling and Simulation in Materials Science and Engineering 12
(2004) pp. 273 – 284
100 N. Fujita, Modeling Carbide Precipitation in Alloy Steels, PhD Thesis in Materials Science
and Engineering, University of Cambridge, 2000
101 N. Fujita & H. K. D. H. Bhadeshia, Modelling Simultaneous Alloy Carbide Sequence in
Power Plant Steels, ISIJ International 42 (2002) pp. 760 - 769
102 T. Sourmail & H. K. D. H. Bhadeshia, Modelling Simultaneous Precipitation Reactions in
Austenitic Stainless Steels, Computer Coupling of Phase Diagrams and Thermochemistry 27
(2003) pp.169 – 175
103 N. Fujita, H. K. D. H. Bhadeshia & M. Kikuchi, Modeling M6C Precipitation in NiobiumAlloyed Ferritic Stainless Steel, Metallurgical and Materials Transactions A 33A (2002) pp.
3339 – 3347
104 N. Fujita & H. K. D. H. Bhadeshia., Modelling Precipitation of Niobium Carbide in Austenite:
Multicomponent Diffusion, Capillarity, and Coarsening. Materials Science & Technology 17
(2001) pp. 403 – 408
105 H. K. D. H. Bhadeshia, Advances in the Kinetics Theory of Carbides Precipitation, Materials
Science Forum 426 – 432 (2003) pp. 35 – 42
106 R.G. Nooning, Jr., Effect of Stabilizing Elements of the Precipitation Behavior and Phase
Stability of Type 409 Ferritic Stainless Steels, MSc in Materials Science and Engineering
Thesis, University of Pittsburgh, 2002
107 H. B. Aaron & G. R. Kotler; Second phase dissolution, Metallurgical Transaction 2 (1971)
pp. 393 – 408
P a g e | 190
108 J. W. Christian; The theory of transformations in metals and alloys: an advanced textbook in
physical metallurgy.
Part 1: Equilibrium and general kinetic theory, 2nd Ed. New York:
Pergamon Press, 1975
109 J.D. Robson; “Modelling of precipitation in power plant steels”: PhD Thesis in Materials
Science and Engineering, University of Cambridge, (1996)
110 N. Fujita & H. K. D. H. Bhadeshia; Precipitation of Molybdenum Carbide in Steel:
Multicomponent Diffusion and Multicomponent Capillarity Effects, Materials Science &
Technology 15 (1999) pp. 627 – 634
111 W.E Stumpf, Phase Transformation in Metals and their Alloys, Course materials for NFM
700., Chapter 2: Coarsening of Precipitates, 2006, University of Pretoria
112 P. E. J. Rivera-Díaz-del-Castillo & H. K. D. H. Bhadeshia; Theory for Growth of Spherical
Precipitates with Capillary Effects, Materials Science & Technology 17 (2001) .30 – 32
113 I. M. Lifshitz & V. V. Slyozov; The Kinetics of Precipitation from Supersaturated Solid
Solutions, Journal of Physical Chemistry Solids 19 (1961) pp.35 – 50
114 C. Wagner; Theory of precipitate change by redissolution, Z Electrochem 35 (1961), pp. 581
– 591
115 A. Kostka, K. –G. Tak, R.J. Hellmig, Y. Estrin & G. Eggeler; On the contribution of carbides
and micrograin boundaries to the creep strength of tempered martensite ferritic steels, Acta
Materialia 55 (2007) pp. 539 – 550
116 T. Mukherjee, W. E. Stumpf, C. M. Sellar & W. J. McG. Tegart; Kinetics of coarsening of
carbides in Chromium steels at 700 °C, Journal of Iron and Steel Institute 207 (1969) pp. 621
– 631
117 J.D. Robson & H. K. D. H. Bhadeshia, Modeling Precipitation Sequences in Power Plant
Steels. Part 1 – Kinetic Theory, Materials Science & Technology 13 (1997) pp. 631 – 639
118 J.D. Robson & H. K. D. H. Bhadeshia, Modeling Precipitation Sequences in Power Plant
Steels. Part 2 – Application of Kinetic Theory, Materials Science & Technology 13 (1997) pp.
640 – 644
119 TCFE3– Thermo–Calc Database (Version 3.0), Royal Institute of Technology, Foundation
of Computational Thermo – dynamics, Stockholms/Sweden, 2002
120 L.B McCuster, R.B. Von Dreele, D. E. Cox, D. Louër & P. Scardi; Rietveld refinement
guidelines, Journal of Applied Crystallography 32 (1999) pp. 36 – 50
121 Course in Advanced X-ray powder diffraction and Rietveld Refinement, University of
Johannesburg, 29 Oct. – 02 Nov 2007
P a g e | 191
122 International Centre for Diffraction Data (ICDD) reference database
123 K. W. Andrews, D. J. Dyson, and S. R. Keown; (1971). Interpretation of electron diffraction
patterns, Hilger, London
124 J. W Edington (1975), Practical electron microscopy in materials science; monograph two:
electron diffraction in the electron microscope, The Macmillan Press Ltd, London
125 TCFE3– Thermo–Calc Database (Version 3.0), Royal Institute of Technology, Foundation
of Computational Thermo – dynamics, Stockholms/Sweden, 2002
126 J–O. Anderson, T. Helander, L. Höglund, P. Shi & B. Sundman; THEMO-CALC & DICTRA,
Computational Tools for materials science, Calphad Vol. 26, no.2 (2002) pp. 273 – 312
127 B. Jansson, B. Jönsson, B. Sundman & J. Ågren; The Thermo Calc Project, Thermochimica
Acta, 214 (1993) pp. 93 – 96
128 http://www.calphad.com
129 TCC TM Thermo-Calc® Software Users’ Guide Version Q
130 A. J. Craven, K. He, L. A. J. Gravie & T. N. Baker; Complex heterogeneous precipitation in titanium–
niobium microalloyed Al-killed HSLA steels—I. (Ti,Nb)(C,N) particles, Acta Materialia 48 (2000) 3857
131 Q. Li; Precipitation of Fe2W Laves phase and modelling of its direct influence on the
strength of the a 12Cr-2W steel, Metallurgical and Materials Transaction A 37A (2006) pp. 89
– 97
132 B.A. Senior; Materials Science and Engineering A119 (1989) L5 (letter)
133 C. C. Silva, J. P. Farias,H. C. Miranda, R. F. Guimares, J. W. A. Menezes & M. A. M. Neto;
Microstructural characterization of the HAZ in AISI 444 ferritic stainless steel welds, Materials
Characterization 59 (2008) pp. 528 – 533
134 J. M. Pardal, S. S. M. Tavares, M. Cindra Fonseca & J. A. de Souza; Influence of the grain
size on deleterious phase precipitation in superduplex stainless steel UNS S32750, Materials
Characterization 60 (2009) pp. 165 – 175
135 J. C. Ahn, G. M. Sim & K. S. Lee; Effect of ageing treatment on high temperature strength of
Nb added ferritic stainless steels, Materials Science Forum, 475 – 479 (2005) pp. 191 – 194
136 Technical Data Blue Sheet - Stainless Steel Type 441, AL 441HP Alloys, Allengheny
Ludlum Corporation
137 L. M. Lundin; Direct measurement of carbon solubility in the intermetallic (Fe,Cr)2(Mo,W)
Laves phase using atomic-probe field-ion microscopy, Scripta Materialia 34 (1996) pp. 741 –
747
P a g e | 192
138 K. Miyahara, S. Matsuoka & T. Hayashi; Nanoindentation as a strength probe – a study on
the hardness dependence of indent size for fine – grained and coarse – grained ferritic steel,
Metallurgical and Materials Transaction A 32A (2001) pp. 761 – 768
139 Fracture: An Advanced Treatise, ed. H. Liebowitz; Volume III: Engineering Fundamentals
and Environmental Effects, Academic Press , New York 1971
140 J. F. Grubb, R.N. Wright & P. Farrar, Jr.; Toughness of ferritic stainless steel, ASTM STP
706, R.A Lula, Ed., American Society for Testing and Materials (1980) pp. 56
141 M. Semchyshen, A. P. Bond & H.J. Dundas; In: Proceedings, symposium towards improved
ductility and toughness, Kyoto, Japan (1971) pp. 239 – 53
142 C. S. Smith, Grains; Phases and Interfaces: An Interpretation of Microstructure,
Transactions of the Metallurgical Society of AIME 175 (1948) pp. 15 – 51
143 R. Higginson & P. Bate; Substructure drag effects and recrystallisation textures in
aluminium, Acta Materialia. 47(1999), pp. 1079 – 1090
144 M.J. Jones & F.J. Humphreys; Interaction of recrystallization and precipitation: The effect of
Al3Sc on the recrystallization behaviour of deformed Aluminium Acta Materialia 51 (2003)
2149–2159
145 J.E. Bailey & P. B. Hirsch; The recrystallisation process in some polycrystalline metals,
Proceeding of the Royal Society A 267 (1962) pp. 11 – 30
146 H. Yu, Y. Kang, Z. Zhao & H. Sun; Morphology and precipitation kinetics of MnS in low –
carbon steel during thin slab continuous casting process, Journal of Iron and Steel Research,
International, 13 (2006) pp. 30 – 36
147 Y. Kang, H. Yu, J. Fu, K. Wang & Z. Wang; Morphology and precipitation kinetics of AlN in
hot strip of low carbon steel produced by compact strip production, Materials Science &
Engineering A351 (2003) pp. 265 – 271
148 J. Fridberg, L. Torndal & M. Hillert; Diffusion in iron, Jern-kontorets Annaler 153 (1969) pp.
263 -279
P a g e | 193

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project