Transformations in TRIP-assisted Steels: Microstructure and Properties

Transformations in TRIP-assisted Steels:
Microstructure and Properties

A thesis submitted for the degree of
Doctor of Philosophy
Sourabh Chatterjee
Darwin College, University of Cambridge
November, 2006
. . . in memory of my father
i
Preface
This dissertation is submitted for the degree of Doctor of Philosophy at
the University of Cambridge. The work reported herein was carried out under
the supervision of Professor H. K. D. H. Bhadeshia in the Department of
Materials Science and Metallurgy, University of Cambridge between January
2004 and October 2006.
To the best of my knowledge, this work is original, except where suit-
able references are made to previous work. Neither this, nor any substantially
similar dissertation has been submitted for any degree, diploma or qualica-
tion at any other university or institution. This dissertation does not exceed
60,000 words in length.
The following publications have been made:
Chatterjee S. and Bhadeshia H. K. D. H., TRIP-assisted Steels: Crack-
ing of High-carbon Martensite, Materials Science and Technology 22 (2006)
645
Chatterjee S., Wang H. S., Yang J. R. and Bhadeshia H. K. D. H.,
Mechanical Stabilisation of Austenite, Materials Science and Technology 22
(2006) 641
Maalekian M., Kozeschnik E., Chatterjee S. and Bhadeshia H. K. D.
H., Mechanical Stabilisation of Eutectoid Steel, Materials Science and Tech-
nology, accepted
Sourabh Chatterjee
November, 2006
ii
Acknowledgments
I am indebted to Prof. H. K. D. H. Bhadeshia for all his support, advice,
inspiration and encouragement during the work and my stay here. I would
also like to thank Prof. A. L. Greer for the facilities and the atmosphere
in the department. I am sincerely grateful to the Management, Tata Steel,
India for providing me with the opportunity to experience Cambridge.
Dr M. Murugananth, a past member of the PT-group, deserves especial
thanks for the optimisation calculations used in my research work. I am also
grateful to Dr D. Bhattacharjee and Dr N. Gope, Tata Steel for arranging
the making of some alloys in the Research and Development laboratories.
I would like to thank every member of the department for being very
kind to me. I wonder if this work would have been at all possible without
Mr Kevin Roberts, Mr Michael Brand, Mr David Vowles, Mr Bryan Adams,
Mr Andrew Rayment, Mr Frank Clarke and Mr Andrew Moss. I also wish
to express my sincere thanks to Mr David Green, Engineering Department
and his team for timely machining some test specimens used in my work.
I shall cherish for long the memory of being with the PT-group and
I thank every member of the group for being so friendly with me. I would
like to thank especially Dr Mohammed Y. Sherif and Mr Mathew J. Peet
for being with me in every dicult situations. All the help from Dr Thomas
Sourmail, Dr Richard Kemp, Mr Saurabh Kundu and other group members
is also gratefully acknowledged.
The support from Darwin College, including the wonderful accommo-
dation and the good food at the cafeteria, made my life a really enjoyable
iii
one. The association with many friends in the College and the Department
has been very benecial. I would also like to thank Mrs Felicity Higginson
and her family for allowing me to exploit the ambience of their home in the
later days.
Finally, I wish to record my deep sense of appreciation for my mother
and my two sisters and their families for relentlessly supporting me during
the entire period, in spite of all their hardship.
iv
Abstract
Despite the presence of high-carbon martensite, TRIP-assisted steels
possess large uniform elongation. High-carbon martensite is normally brittle.
In this thesis, it has been demonstrated that this apparent anomaly is due
to the ne size of the martensite plates.
The mechanical properties of these steels are due to the transformation
of retained austenite into martensite during deformation and hence appear to
be dominated by the volume fraction and carbon content of retained austen-
ite. These parameters have been related to the chemical composition and
heat treatment of the steels with neural networks, using published data.
An optimum alloy was formulated by combining the neural network
with a genetic algorithm, to minimise the silicon addition whilst maximising
the retained austenite fraction. This resulted in the creation of a radically
dierent microstructure, designated -TRIP.
Transformation of austenite into martensite during deformation ceases
beyond a critical strain. A theory has been developed to predict this limit.
Calculations using the theory indicate that the high-carbon austenite in these
steels may transform into martensite due to stress, rather strain.
These materials are often tested for stretch-angeability, a measure of
formability. Neural network analysis of the published data revealed the ulti-
mate tensile strength to be the most important tensile parameter inuencing
stretch-angeability.
Contents
1 Introduction 1
1.1 Scope of the research . . . . . . . . . . . . . . . . . . . . . . . 2
2 TRIP-assisted steels 4
2.1 Martensite and TRIP steels . . . . . . . . . . . . . . . . . . . 5
2.2 Modern TRIP-assisted steels . . . . . . . . . . . . . . . . . . . 8
2.3 Microstructural evolution . . . . . . . . . . . . . . . . . . . . . 9
2.4 Alloying elements in TRIP-assisted steels . . . . . . . . . . . . 16
2.5 Mechanical performance . . . . . . . . . . . . . . . . . . . . . 18
2.6 Factors aecting performance . . . . . . . . . . . . . . . . . . 20
2.6.1 Proportion of phases . . . . . . . . . . . . . . . . . . . 21
2.6.2 Stability of retained austenite . . . . . . . . . . . . . . 22
2.6.3 Test parameters . . . . . . . . . . . . . . . . . . . . . . 26
2.6.4 State of stress or strain . . . . . . . . . . . . . . . . . . 29
2.7 Strain-induced martensite formation . . . . . . . . . . . . . . 30
2.8 Special properties . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.8.1 Formability . . . . . . . . . . . . . . . . . . . . . . . . 33
2.8.2 Crash-worthiness . . . . . . . . . . . . . . . . . . . . . 36
2.8.3 Fatigue resistance . . . . . . . . . . . . . . . . . . . . . 36
v
vi CONTENTS
2.8.4 Bake hardening . . . . . . . . . . . . . . . . . . . . . . 38
2.9 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.10 Other variants . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.11 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3 Brittle martensite 45
3.1 Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.2 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . 48
3.4 Simulating microstructural evolution . . . . . . . . . . . . . . 55
3.5 Martensite in TRIP-assisted steels . . . . . . . . . . . . . . . . 60
3.6 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4 Microstructural Modelling 71
4.1 Neural network modelling . . . . . . . . . . . . . . . . . . . . 71
4.2 Model for retained austenite fraction . . . . . . . . . . . . . . 75
4.2.1 Database . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.2.2 Model characteristics . . . . . . . . . . . . . . . . . . . 76
4.2.3 Model predictions . . . . . . . . . . . . . . . . . . . . . 78
4.3 Model for carbon in retained austenite . . . . . . . . . . . . . 85
4.3.1 Database . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.3.2 Model characteristics . . . . . . . . . . . . . . . . . . . 85
4.3.3 Model predictions . . . . . . . . . . . . . . . . . . . . . 86
4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
vii CONTENTS
5 -TRIP steel 95
5.1 Optimisation . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.2 Optimised TRIP-assisted steel . . . . . . . . . . . . . . . . . . 98
5.3 Thermodynamic calculations . . . . . . . . . . . . . . . . . . . 99
5.4 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6 Mechanical Stabilisation 125
6.1 Mechanical driving force . . . . . . . . . . . . . . . . . . . . . 125
6.2 Role of strain . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.3 Mathematical formulation . . . . . . . . . . . . . . . . . . . . 129
6.4.1 Austenitic stainless steels . . . . . . . . . . . . . . . . . 133
6.4.2 TRIP-assisted steels . . . . . . . . . . . . . . . . . . . 137
6.4.3 Bainitic steels . . . . . . . . . . . . . . . . . . . . . . . 143
6.4.4 Athermal martensite . . . . . . . . . . . . . . . . . . . 145
6.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
7 Formability 149
7.1 Stretch-angeability . . . . . . . . . . . . . . . . . . . . . . . . 149
7.2 Neural networks . . . . . . . . . . . . . . . . . . . . . . . . . . 150
7.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
8 Conclusions 164
Chapter 1
Introduction
The worldwide demand for a reduction in greenhouse gas emissions, better
fuel economy and safety in automobiles has led to the development of a vari-
ety of steels. In addition to the high strength, these materials can be formed
into complex shapes. The steels rely on the transformation of austenite into
martensite during deformation for achieving their mechanical properties and
hence are known as transformation-induced plasticity (TRIP) steels.
There are two types of such steels. Those having a fully austenitic mi-
crostructure are called TRIP steels [Zackay et al., 1967]. These steels tend
to be rich in nickel and other expensive austenite stabilising elements. By
contrast, austenite is only a minor phase in the overall microstructures of
TRIP-assisted steels [Matsumura et al., 1987a; Takechi et al., 1987]. Allotri-
omorphic ferrite comprises about 50-60 vol.% of the microstructures of these
materials, the remainder being a mixture of bainite and carbon-enriched re-
tained austenite. TRIP-assisted steels are generally lean in solute content
with only about 0.2 wt% carbon, 1.5 wt% manganese and some 1-2 wt%
silicon. It is nevertheless possible to retain copious amounts of austenite in
1
2 1.1 Scope of the research
the microstructure. This is due to the fact that silicon retards cementite pre-
cipitation from the untransformed austenite during bainite formation. The
carbon that is partitioned into the untransformed austenite following the
formation of bainite, stabilises it and allows it to be retained at ambient
temperature.
1.1 Scope of the research
As will be evident from the next Chapter, there is a huge amount of published
work on TRIP-assisted steels, hardly surprising given their technological im-
plications. At the same time, much of this research is repetitive and has
ignored some puzzling issues which are the subject of this thesis.
There is, for example, no theory to describe whether the transforma-
tion of the austenite during deformation is a consequence of stress, plastic
strain or a combination of these. This is not a trivial issue because the eect
of stress is purely thermodynamic whereas plastic strain has an ambivalent
inuence on martensitic transformation. Defects introduced by deformation
may provide heterogeneous nucleation sites for martensite, thereby acceler-
ating the decomposition of retained austenite. In contrast, martensitic trans-
formations involve the displacement of glissile interfaces, which are hindered
by dislocation debris created by plastic strain. Indeed, it has been known for
a long time that excessive plastic strain can stop martensite from forming
by a process of mechanical stabilisation. There is, however, no quantitative
theory capable of predicting this phenomenon to achieve such a description
was one aim of the work.
There is much research on the roles of silicon in TRIP-assisted steels,
3 1.1 Scope of the research
all of which aims to reduce its concentration to a minimum. This is in order
to improve the all-important surface quality, a critical factor in automobile
steels. However, attempts at doing this or to replace silicon with aluminium
frequently fall foul of the need to retard the formation of cementite during
the course of the bainite reaction. The concentrations of silicon-substitutes
that can be added are limited by the fact that elements such as aluminium
form -loops on phase diagrams. At a large enough concentration, it becomes
impossible to fully austenitise the steel because of the stability of -ferrite.
It was felt at the outset that this fear of -ferrite is unjustied since
TRIP-assisted steels contain a large amount of allotriomorphic ferrite. The
latter could simply be substituted by -ferrite, to form a new class of steels,
here designated -TRIP. This concept was another aim of the work.
A major puzzle which is neglected in published work, is why the forma-
tion of untempered, high-carbon strain-induced martensite does not embrittle
the TRIP-assisted steels. The solution to this problem is an outcome of this
thesis.
Finally, there are many measures of formability quoted in the literature,
but development work seems to focus on an empirical product of the ultimate
tensile strength and elongation. This makes it very dicult to dene how
good steel microstructures, which perform well during forming operations,
can be developed. The justication for the formability criteria is another
feature of the present work.
Chapter 2
TRIP-assisted steels
An overview of TRIP-assisted steels is presented in this chapter. Typical
mechanical properties such as proof strength (YS), ultimate tensile strength
(UTS), total elongation (TEL), strain-hardening exponent (n) and plastic
anisotropy ( r) of various automobile steels are shown in Table 2.1. As can be
seen, any increase in strength is in general associated with a loss of ductility.
The exceptions are the TRIP-assisted steels, where considerable ductility is
obtained in spite of the strength. These promising properties are thought to
be due to transformation-induced plasticity.
4
5 2.1 Martensite and TRIP steels
Steel type YS / MPa UTS / MPa TEL / % n r
Mild 140/270 140 270 38-44 0.05-0.15 1.8
BH 210/340 210 340 34-39 0.23 1.8
BH 260/370 260 370 29-34 0.18 1.6
IF 260/410 260 410 34-38 0.13 1.7
DP 280/600 280 600 30-34 0.2 1.0
IF 300/420 300 420 29-36 0.21 1.6
DP 300/500 300 500 30-34 0.2 1.0
HSLA 350/450 350 450 23-27 0.16 1.0
DP 350/600 350 600 24-30 0.22 1.1
DP 400/700 400 700 19-25 0.14 1.0
TRIP 450/800 450 800 26-32 0.14 0.9
HSLA 490/600 490 600 21-26 0.24 1.0
DP 500/800 500 800 14-20 0.13 1.0
SF 570/640 570 640 20-24 0.14 1.0
CP 700/800 700 800 10-15 0.08 1.0
DP 700/1000 700 1000 12-17 0.13 0.9
Mart 950/1200 950 1200 5-7 0.09 0.9
Table 2.1: Typical mechanical properties of steels for automotive applications
[ULSAB-AVC].
2.1 Martensite and TRIP steels
In steels, austenite can transform to ferrite either by a reconstructive or
a displacive mechanism. In reconstructive transformation, all the atomic
bonds are broken and the atoms are rearranged into the new form while
minimising the overall strain energy. A displacive transformation occurs by
a homogeneous deformation of the original crystal structure. Martensitic
transformation is displacive and can occur at temperatures where diusion
is inconceivable within the time scale of the process. Transformation starts
only after cooling to a particular temperature called martensite-start tem-
perature or M
S
. The fraction transformed increases with the undercooling
below M
S
. A martensite-nish temperature or M
f
is usually dened as the
temperature where 95% of the austenite has decomposed. Unlike M
S
, M
f
has no fundamental signicance.
Martensite can also be induced to form by stress or strain at tem-
peratures above M
S
[Scheil, 1932]. The work done by an external stress
compensates for the shortfall in the driving force for transformation to occur
above M
S
[Patel and Cohen, 1953]. The higher the temperature above M
S
,
the greater is the magnitude of the stress required. However, the strength
of the austenite is lower at high temperatures. When the stress required for
transformation exceeds the strength of the austenite, plastic strain precedes
transformation. This lowers the stress requirement for transformation, as
illustrated in Figure 2.1. However, martensite can only form upto a cer-
tain maximum temperature limit termed as M
d
temperature, above which
austenite only deforms plastically.
In steels containing austenite in the temperature range M
S
-M
d
, the
formation of martensite during plastic deformation helps to maintain strain-
hardening. This delays the onset of necking, resulting a large uniform elon-
gation. This is why TRIP steels are useful.
The rst practical exploitation of TRIP came from Zackay and co-
workers who developed steels with dramatically improved elongation, as a
consequence of deformation-induced martensitic transformation [Zackay et
al., 1967]. However, these alloys contained large concentrations of expensive
solutes which restricted their popularity (Table 2.2). The alloys were fully
austenitic at ambient temperature.
The elongation obtained can vary with the tensile test temperature
Figure 2.1: Stress requirement for martensitic transformation above M
S
in-
creases at higher temperatures; formation of martensite is not possible be-
yond M
d
.
C Si Mn Cr Ni Mo
0.31 1.92 2.02 8.89 8.31 3.8
0.25 1.96 2.08 8.88 7.60 4.04
0.25 1.90 0.92 8.80 7.80 4.00
0.25 - - - 24.4 4.10
0.23 - 1.48 - 22.0 4.00
0.24 - 1.48 - 20.97 3.57
Table 2.2: Typical chemical compositions (wt%) of early TRIP steels [Zackay
et al., 1967].
8 2.2 Modern TRIP-assisted steels
(Figure 2.2) due to changes in the stability of austenite to martensitic trans-
formation; this is one of the key design issues.
Figure 2.2: Mechanical properties of TRIP steels depend strongly on the
test temperature. The elongation shows a maximum at an intermediate
temperature between M
S
and M
d
[Tamura et al., 1970].
2.2 Modern TRIP-assisted steels
In the 1980s, the TRIP eect was demonstrated in low-alloy steels made
with 0.2 C, 1-2 Mn and 1-2 Si (wt%) [Matsumura et al., 1987a; Takechi
et al., 1987]. The microstructures consisted of 50-60 vol.% allotriomorphic
9 2.3 Microstructural evolution
ferrite, 20-30 vol.% carbide-free bainite, the remainder being high-carbon
retained austenite with some martensite, Figure 2.3 [Jacques et al., 2001b].
Such steels are henceforth referred to as TRIP-assisted to distinguish them
from fully austenitic TRIP steels.
The TRIP-assisted alloys are lean and hence aordable (Table 2.3).
They typically contain only 10-30 vol.% austenite [Hanzaki, 1994; Matsumura
et al., 1992; Sakuma et al., 1991a; Sugimoto et al., 1993, 1992a]. The mi-
crostructural evolution of these materials is described in the following section.
Figure 2.3: A typical multiphase microstructure of a modern TRIP-assisted
steel, with allotriomorphic ferrite (F), carbide-free bainite (B) and retained
austenite (A) [courtesy of Jacques].
2.3 Microstructural evolution
The microstructure of TRIP-assisted steel can be generated using both hot
rolling and cold rolling. Hot rolling is generally carried out at temperatures
where the steel is fully austenitic. After rolling, the material is cooled to
C Si Mn Al P Nb Mo Cu
0.38 1.53 0.83 0.007
0.18 2.0 1.5 0.037 0.015
0.19 2.48 1.49 0.036 0.014
0.11 0.59 1.55 1.5 0.012
0.14 0.53 1.57 0.204
0.22 1.55 1.55 0.028 0.035
0.20 1.47 1.51 0.028 0.004 0.047 0.2
0.20 1.6 1.6 0.028 0.041 0.3
0.21 1.49 1.49 0.028 0.005 0.017 0.1
0.14 1.49 1.51 0.04 0.0012 0.51
Table 2.3: Typical chemical compositions (wt%) of TRIP-assisted
steels [Chen et al., 2002; Hanzaki et al., 1995; Hashimoto et al., 2004; Jacques
et al., 2001b; Jiao et al., 2002; Kim et al., 2002; Matsumura et al., 1987b;
Sugimoto et al., 1992a].
ambient temperature. The cooling rate is controlled so that austenite rst
transforms to allotriomorphic ferrite and then to bainite. However, a two-
stage annealing treatment is required to produce the desired microstructure
in cold rolled TRIP-assisted steels (Figure 2.4). The material is initially
heated to a temperature in the (+)
1
phase region generating a mixture of
ferrite and austenite, which subsequently decomposes to bainite at a lower
temperature.
Austenite is retained in TRIP-assisted steels due to the incomplete reac-
tion phenomenon associated with bainitic transformation [Bhadeshia, 2001].
Bainitic ferrite grows without diusion but excess carbon subsequently par-
titions into the residual austenite (Figure 2.5). Diusionless growth can only
be sustained at temperatures below T
0
, at which ferrite and austenite of iden-
tical chemical composition have the same free energy. It follows that in the
1
The terms and represent ferrite and austenite respectively.
Figure 2.4: Schematic illustration of the two routes to generate the mi-
crostructure of TRIP-assisted steel, with typical temperature and time in-
dicated. Curves 1 and 2 stand for the transformation from fully austenitic
state after hot rolling and intercritical annealing after cold rolling respec-
tively. The terms , ,
b
and
represent austenite, allotriomorphic ferrite,

bainitic ferrite and martensite respectively [Bhadeshia, 2001].
absence of carbide precipitation the bainite reaction stops when the carbon
content in austenite equals the T
0
limit (considering the stored energy of
bainite), as illustrated in Figure 2.6.
Some elements like silicon, aluminium and phosphorous retard ce-
mentite precipitation allowing the carbon to remain dissolved in austenite.
The carbon-enriched residual austenite becomes stable against any further
martensitic transformation during subsequent cooling [Andrews, 1965], and
hence is retained. The pronounced eect of carbon in stabilising austenite is
therefore exploited in retaining austenite, thus permitting the concentration
of other solutes to be kept to a minimum.
Figure 2.5: Schematic illustration of bainite reaction mechanism [Bhadeshia,
2001].
Microstructural evolution during the isothermal formation of bainite is
identical for both hot and cold rolled TRIP-assisted steels. A typical trans-
formation map is shown in Figure 2.7 [Jacques et al., 2001b]. As the bainite
reaction progresses, the residual austenite becomes enriched with carbon and
gains stability against martensitic transformation during cooling. When the
enrichment is inadequate at low holding times, the austenite may decompose
partly into martensite during cooling. On the other hand, amount of un-
Figure 2.6: The incomplete reaction phenomenon: bainite reaction stops as
carbon in austenite approaches T
0
limit, x is the bulk carbon content of
the steel and A
e
3
represents the paraequilibrium ( + )- phase bound-

ary [Bhadeshia, 2001].
transformed austenite decreases, as more bainite forms with greater holding
time. It follows that the quantity of retained austenite becomes maximum
at an intermediate holding time during bainite reaction.
One fascinating aspect of the bainite reaction is the sub-unit mecha-
nism of growth [Bhadeshia, 2001]. In steels containing sucient silicon or
aluminium, bainite forms as an aggregate of ferrite plates separated by un-
transformed austenite. The individual plates are called sub-units while the
aggregate is known as a sheaf (Figure 2.8). A sheaf of bainite therefore
evolves by repeated nucleation of new sub-units predominantly at the tip of
the pre-existing ones, rather on their sides.
But such a mechanism may not occur in TRIP-assisted steels because
Figure 2.7: Typical microstructural evolution map during the bainite reac-
tion. Retained austenite peaks at an intermediate holding time [Jacques et
al., 2001b].
of the size of the available austenite [Jacques, 2003]. Bainite in these steels
consists of adjacent platelets that span the whole of the austenite grain (Fig-
ure 2.9). This is more likely to be observed in the cold rolled steels subjected
to intercritical annealing than in hot rolled materials. Intercritical anneal-
ing produces very ne regions of austenite (about 1 m) so that the space
available for the sub-unit mechanism to operate is minimised.
Epitaxial ferrite is also more likely to be present in the microstructures
of cold rolled steels than the hot rolled products. After intercritical annealing
some ferrite may form during cooling to the bainite transformation tempera-
ture. This occurs by the movement of an existing austenite-ferrite boundary
i.e., epitaxial growth [Zaeerer et al., 2004].
Figure 2.8: Schematic illustration of sub-units and sheaf of bai-
nite [Bhadeshia, 2001].
Figure 2.9: Bainite consisting of adjacent ferrite platelets in TRIP-assisted
steels [courtesy of Jacques].
16 2.4 Alloying elements in TRIP-assisted steels
2.4 Alloying elements in TRIP-assisted steels
Solutes other than carbon are added for the following reasons:
to optimise the fraction of retained austenite,
to control cementite precipitation,
to increase the hardness of ferrite and
to increase the hardenability so that pearlite formation can be avoided
before bainite reaction.
Carbon, manganese and silicon
Like carbon, manganese is an austenite stabiliser. Silicon although a ferrite
stabiliser, helps to retain carbon-enriched austenite by suppressing cementite
precipitation from austenite. Silicon also solid-solution strengthens ferrite
and hence can enhance the overall strength of the steel.
An increase in manganese may compensate for any reduction in sili-
con [Sakuma et al., 1991a; Shi et al., 2002], but this shifts the T
0
curve to
lower carbon concentrations, thereby limiting the amount of bainite that can
form. Additionally, pronounced banding may occur in steels containing a
large manganese concentration [Kim et al., 2001].
Aluminium and phosphorous
Both aluminium and phosphorous inhibit cementite precipitation and hence
can substitute for silicon [Mintz, 2003]. However, unlike silicon, aluminium
does not strengthen ferrite. Steels in which silicon is replaced by aluminium
may therefore be weaker [Girault et al., 2001; Jacques et al., 2001a].
17 2.4 Alloying elements in TRIP-assisted steels
Phosphorous, on the other hand, strengthens ferrite [Pichler et al.,
1998]. An addition of 0.1 wt% phosphorous leads to an increase of about
75 MPa in the strength of ferrite [Pickering, 1978].
Niobium and Molybdenum
Niobium in solid solution is reported to increase the quantity of retained
austenite [Hanzaki et al., 1995; Pereloma et al., 1999]. The reason is not
clear, as niobium is known to be a ferrite stabiliser. Fine precipitates of
niobium carbides and complex NbMoC precipitates can also strengthen the
ferrite [Hashimoto et al., 2004].
Molybdenum is a solid solution strengthener of ferrite and retards
pearlite formation. In low-silicon steels, its addition can lead to mechani-
cal properties comparable to high-silicon steels viz. ultimate tensile strength
in excess of 1000 MPa with a total elongation of about 36% [Bouet et al.,
1998a,b].
Copper
Copper, being an austenite stabiliser, helps to retain austenite [Im et al.,
2000a; Kim and Lee, 1999]. Besides solid solution strengthening, -Cu pre-
cipitates in ferrite can boost the overall strength [Kim et al., 2002]. Thus
copper can be thought to replace silicon in both the roles of retaining austen-
ite and increasing strength of ferrite.
Boron
Small concentrations of boron are known to signicantly improve hardenabil-
ity. Boron-containing low-silicon steels have been found to contain sucient
18 2.5 Mechanical performance
retained austenite and display a sensitive TRIP eect [Sadhukhan et al.,
2001].
2.5 Mechanical performance
TRIP-assisted steels are reputed for the superior uniform elongation com-
pared to other similarly strong steels like dual phase steels. The latter with
microstructures of 10-15 vol.% low carbon (C < 0.3 wt%) martensite dis-
persed in ferrite have high strength [Rashid, 1979a,b]. But these steels lacked
formability. Traces of untransformed retained austenite in the microstruc-
tures of these steels were indeed thought to enhance the ductility through
TRIP eect [Goel et al., 1985; Kim, 1988; Rao and Rashid, 1983; Sangal et
al., 1985].
As noted by Hassani and Yue [1999], TRIP-assisted steel can be imag-
ined to be a modication of the dual phase steel. Substantially greater quan-
tities of retained austenite in TRIP-assisted steels yielded even greater elon-
gation (Figure 2.10), as compared to dual phase steels.
However, the TRIP-assisted steel does not seem to exhibit continuous
yielding unlike the dual phase steel (Figure 2.10). Continuous yielding is
advantageous in forming operations as this helps to avoid stretcher strain or
L uder bands. The gradual yielding of dual phase steels is mainly due to the
free dislocations present in the ferrite grains. Formation of bainite, instead
of martensite, in TRIP-assisted steels is not very eective to induce free dis-
locations in ferrite, resulting discontinuous yielding [Sakuma et al., 1992]. It
may be possible to avoid this by allowing some martensite to form during
cooling after the isothermal formation of bainite [Choi et al., 1988]. The dis-
19 2.5 Mechanical performance
continuous yielding of the intercritically annealed cold rolled TRIP-assisted
steels could also be due to the strain ageing of ferrite at the temperature
where bainite forms [Choi et al., 1988]. The ageing of ferrite can occur due
to the dierence in the solubility of the interstitial solutes at the intercritical
annealing temperature and the bainite formation temperature.
Figure 2.10: Superior elongation observed in a TRIP-assisted steel compared
to a dual phase steel with similar strength level [Hassani and Yue, 1999].
The large uniform elongation in TRIP-assisted steels is mainly at-
tributed to the strain-induced martensitic transformation of retained austen-
ite. However, the transformation strains themselves can contribute at most
2% to the observed elongation given the small fraction of austenite present
in these materials [Bhadeshia, 2002]. Other phases in the microstructure
must also inuence the overall mechanical behaviour. An interesting conse-
quence of the strain-induced martensitic transformation of austenite has been
20 2.6 Factors aecting performance
revealed by Jacques et al. [2001c]. The volume expansion due to marten-
sitic transformation generates dislocations in adjacent ferrite (Figure 2.11).
Freshly produced dislocations in the ferrite phase can then take part in the
deformation process.
Martensite that forms in these steels inherits the high carbon content of
the austenite. It is surprising that the freshly produced high-carbon marten-
site does not impair the ductility. Carbon embrittles martensite, a fact that
seems to have been ignored in explaining the properties of these materials.
Figure 2.11: Dislocations generated in ferrite due to transformation of austen-
ite during deformation [courtesy of Jacques].
2.6 Factors aecting performance
There are many factors that have been thought to control the mechanical
properties of these steels, as summarised in this section.
2.6.1 Proportion of phases
TRIP-assisted steels consist of allotriomorphic ferrite, bainitic ferrite, re-
tained austenite with or without traces of martensite. As reported by Choi
et al. [2002], a high volume fraction of retained austenite improves elongation
as well as the ultimate tensile strength of the steel (Figure 2.12).
Figure 2.12: Higher UTS and uniform elongation obtained with H-type sam-
ple as compared to L-type, H-type sample contains higher austenite volume
fraction than the L-type [Choi et al., 2002].
However, a high retained austenite volume fraction implies less bainite
which may adversely aect the strength. Similarly, high ferrite fraction dur-
ing intercritical annealing nally results in less bainite, leading to a reduction
in strength [Imai et al., 1992].
Aluminium-containing steels contain more ferrite as compared to steels
relying on silicon [Monohar et al., 2002, 2003]. This is because in aluminium-
containing steels, austenite starts transforming into allotriomorphic ferrite
at a higher temperature and with a faster kinetics than in steels made with
silicon [Manohar et al., 2002]. A high ferrite fraction in the microstructure
can reduce the overall strength of a steel made with aluminium.
Steels with reduced silicon content have been shown to possess prop-
erties similar to high-silicon steels by Jacques et al. [2001b]. This is due to
the presence of some martensite in the low-silicon material.
It seems, phases other than retained austenite can inuence the me-
chanical behaviour of these steels. However, most of the research has been
focussed on the TRIP eect due to the strain-induced transformation of
austenite.
2.6.2 Stability of retained austenite
The rate at which austenite transforms during deformation appears to be
the highly emphasised factor aecting the properties. The austenite stability
should be such that it transforms progressively during deformation, so that
damage can be accommodated at all stages of deformation.
Increasing carbon content of austenite lowers the driving force for
martensitic transformation. It is known that after isothermal bainite
transformation, retained austenite is not homogeneously enriched with car-
bon [Schrader and Wever, 1952]. Regions in the vicinity of the bainite sheaves
or the lms of austenite trapped between bainite are rich in carbon. This may
even lead to the observation of two dierent lattice parameters of austenite
formed in the same sample [Matas and Hehemann, 1961]. So austenite areas
that are depleted of carbon, transform rst during deformation. It follows,
with increasing plastic strain, untransformed austenite may be observed to
have a greater lattice parameter, because of the larger carbon content as
shown in Figure 2.13 [Itami et al., 1995].
Besides the carbon content, austenite stability can be also enhanced by
a ne grain size. High stability of the retained austenite with ne grain size,
dispersed in a ferritic microstructure obtained after intercritical annealing,
was reported long back by Rigsbee [1979]. This is due to the absence of
substructures in the austenite, for example stacking faults and other defects,
that provide nucleation sites for martensitic transformation. Austenite in
TRIP-assisted steels exists in two forms, isolated austenite grains in ferrite
away from the bainitic phase and lms of austenite in-between bainitic ferrite
plates. The latter type of austenite is generally observed to be more stable
during deformation than the former types [Basuki and Aernoudt, 1999; Koh
et al., 1998; Sugimoto et al., 1993]. It is thus thought that isolated small
austenite grains dispersed in ferrite actually contribute to the TRIP eect.
This is illustrated in Figure 2.14.
By micromechanical modelling, Reisner et al. [1998] showed that the
tendency for strain-induced transformation depends on the orientation of the
austenite grains. Austenite with a random texture behaves similarly to the
average of the sharp texture components.
The mechanical stability of retained austenite during deformation also
seems to depend on the hardness of the other phases. The hardness in-
creases in the order ferrite, bainite, austenite and martensite [Furnemont et
(a)
(b)
Figure 2.13: With increasing plastic strain, volume fraction of austenite de-
creases along with the increase in carbon content, which is apparent from
the shift in the positions of the austenite,
R
, peaks [Itami et al., 1995].
Figure 2.14: Eect of grain size and morphology of austenite on its stability,
retained austenite with an aspect ratio larger than 2.5 and grain size ner
than 1.2 m remains unchanged with increasing tensile deformation [Basuki
and Aernoudt, 1999].
al., 2002]. An increase in the hardness of ferrite will obviously delay the
transfer of stress to the austenite and hence the onset of TRIP. This is the
basis of the composite eect as proposed by Jacques et al. [2001d]. Steels
with reduced silicon content are generally low in retained austenite fraction.
Nevertheless, it is possible to achieve favourable properties in these steels
due to the presence of martensite strengthening the matrix hence delaying
the onset of TRIP eect.
It follows that any increase in the hardness of the matrix phase can be
benecial for the mechanical properties. This is consistent with the promising
properties of the steels containing phosphorus or copper, that strengthen the
ferrite matrix. However, any increment of matrix strength may not always be
benecial. Sakuma et al. [1991b] showed that increasing carbon content from
0.1 wt% to 0.4 wt%, increases the volume fraction and stability of retained
austenite. But this reduces the ferrite content, thereby increasing strength
with an associated loss in ductility. This, of course, removes any advantage
associated with TRIP steels.
Similarly, aluminium-containing steels may be expected to have inferior
properties as compared to silicon containing steels due to softer ferrite matrix.
However, due to higher carbon enrichment in the retained austenite, steels
based on aluminium instead of silicon achieve better properties [De Meyer et
al., 1999a,b].
2.6.3 Test parameters
Temperature
Higher temperature increases the stability of austenite against martensitic
transformation due to reduced driving force for transformation. Tempera-
ture can therefore exert a signicant inuence on the mechanical properties
(Figure 2.15).
The ultimate tensile strength and total elongation, unlike the yield
Figure 2.15: Eect of test temperature on the ow curve of TRIP-assisted
steels [Sugimoto et al., 1992b].
strength, appear to vary much more strongly with temperature. However, the
properties seem to improve only upto a certain maximum temperature [Sug-
imoto et al., 1993, 1992a]. This is consistent with the observed behaviour of
the classical TRIP steels with a fully austenitic microstructure, as mentioned
earlier.
Modern TRIP-assisted steels possess a multiphase microstructure, only
10-30 vol.% of which is austenite. The strength and elongation of the individ-
ual phases are also very likely to depend on the test temperature. These may
all inuence the temperature dependence of the overall mechanical proper-
ties. This aspect appears to have not been considered in understanding the
mechanical behaviour of these steels.
Strain rate
Figure 2.16 shows the inuence of the strain rate on the nature of the ow
curves at dierent temperatures. An intermediate strain rate appears to
maximise elongation, consistently at all the testing temperatures.
Figure 2.16: Eect of strain rate on the nature of the ow curve at dierent
test temperatures [Sugimoto et al., 1992b].
High strain rate tensile testing is often referred to as dynamic me-
chanical testing. The volume fraction, size, distribution and morphology of
retained austenite have all been shown to aect the dynamic mechanical
properties [Wei et al., 2003]. Increasing strain rate from 10
3
to 2.510
2
s
1
raises both yield strength and ultimate tensile strength while the strain-
hardening behaviour remains almost the same, as reported by Choi et al.
[2002].
2.6.4 State of stress or strain
Unlike uniaxial tension or compression, deformation-induced martensitic
transformation is always known to be suppressed by hydrostatic pressure.
The stability of the austenite can therefore be enhanced by applying hydro-
static pressure during tensile test, resulting higher uniform elongation [Pysh-
mintsev et al., 2002]. This may be important in hydroforming operations.
Retained austenite in such a case would not be fully consumed during form-
ing, thereby allowing the TRIP eect to be utilised for crash-worthiness in
nished components.
Strain-induced martensitic transformation is retarded under plain-
strain conditions relative to uniaxial loading [Im et al., 2002]. This is due
to lower applied stress and strain under plane-strain condition. By contrast,
under biaxial stretching, the transformation is promoted relative to uniaxial
testing, although the results vary with the sample orientation relative to the
rolling direction [Streicher et al., 2002]. This may be related to the texture
of austenite grains [Reisner et al., 1998].
To summarise, the strain-induced transformation of retained austenite
has been considered to be the major factor inuencing the mechanical per-
formance. This is despite the fact that these steels possess multiphase mi-
crostructures, some 10-30 vol.% of which is only retained austenite. A mod-
erately stable austenite appears to be the key to optimise ductility. Among
30 2.7 Strain-induced martensite formation
many factors, carbon content of the retained austenite inuences its stabil-
ity. Therefore not only the fraction of austenite, but also its carbon content
appears to be a salient microstructural feature controlling the properties. It
is best to have the austenite transforming progressively during deformation.
The kinetics of this transformation has been a subject of much research, as
described in the following section.
2.7 Strain-induced martensite formation
The martensitic transformation of austenite during deformation may occur
due to stress or strain (Figure 2.1). The progress of strain-induced transfor-
mation is illustrated in Figure 2.17. Rapid transformation at rst is due to
the nucleation sites introduced by the deformation. However, an excessive
strain can retard the transformation, eventually bringing it to a halt. This
latter phenomenon is known as mechanical stabilisation, a characteristic fea-
ture of displacive transformations. By contrast, the eect of stress is purely
thermodynamic.
There are several empirical models for the strain-induced transforma-
tion. One of these was proposed by Olson and Cohen [1975] to model the
experimental data reported by Angel [1954],
f
= 1 exp[{1 }
n
] (2.1)
where f
is the volume fraction of martensite obtained at a strain . The

model contains two temperature dependent parameters, and . The former
depends on the stacking fault energy of the steel, which in turn is aected by
Figure 2.17: Kinetics of the strain-induced martensite formation, each curve
represents a steel with dierent carbon content of the retained austen-
ite [Jacques et al., 2001c].
temperature. denes the probability that a shear-band intersection creates
a nucleation site, and therefore can be thought to depend on the chemical
driving force, hence temperature. However, the exponent n in the model
comes from tting to experimental data. A value of 4.5 is appropriate for
the steels studied by Angel.
Guimaraes [1972] studied a group of Fe-Ni-C alloys and proposed a
similar exponential t:
f
= 1 exp(k
Z
) (2.2)
with empirical constants k = 28 and Z = 3.7.
A dierent parabolic type of dependence was proposed by Gerberich et
al. [1970]:
f
= A
2
1/2
(2.3)
where A
2
is an empirical constant.
As can be seen, none of these models include volume fraction of austen-
ite and therefore exclude any possible interaction between the parent and the
transformed phases. Angel [1954] and Ludwigson and Berger [1969] consid-
ered this and proposed the following model:
f
= A
1
B
V
(2.4)
where A
1
and B are constants, with B = 3 for stainless steels.
All these models are formulated on the basis of the experimental data
for fully austenitic steels. The strain-induced transformation of retained
austenite in TRIP-assisted steel has been expressed by Sugimoto et al.
[1992b] as
ln(V
0
) ln(V
) = k (2.5)
where V
0
is the original retained austenite volume fraction and V
is the
volume fraction of the untransformed austenite at a plastic strain . The
empirical constant k must be measured for each steel.
A novel way of describing strain-induced transformation of retained
austenite in these TRIP-assisted steels has been proposed by Sherif et al.
[2004], a development of the model of Sugimoto et al. [1992b]. The frac-
tion of untransformed austenite is expressed as a function of plastic strain,
33 2.8 Special properties
deformation temperature, chemical composition and the original amount of
austenite. The nal equation is given as,
ln(V
0
) ln(V
) = k
1
G
(2.6)
where k
1
takes a value of 0.002017 J mol
1
, independent of the steel compo-
sition.
All these models explain adequately the kinetics of strain-induced for-
mation of martensite. But none takes into account the possibility of mechan-
ical stabilisation at large plastic strains.
2.8 Special properties
2.8.1 Formability
Formability is dened as the ease with which an object having a complex
geometry can be manufactured using the material. This can approximately
be related to plasticity or ductility which tends to deteriorate with increasing
strength. TRIP-assisted steels are unique in this respect, they exhibit better
formability than several other steels of comparable strength [Konieczny, 2003;
Sugimoto et al., 1996].
Formability is not assessed by a single attribute but expressed as a com-
bination of properties such as deep drawability, stretch formability, stretch
angeability and bendability. The state of stress is dierent in each of these
forming operations, as illustrated schematically in Figure 2.18.
Figure 2.18: State of stress existing during dierent modes of forming oper-
ations [Takahashi, 2003].
Deep drawability
This is expressed by the Lankford parameter or r value, the ratio of true
strain in the width direction to that in thickness direction. r is the average
of r values measured on specimens obtained at dierent directions, namely at
0
, 45
and 90
to the rolling direction. A higher r value leads to better deep

drawability. TRIP-assisted steels exhibit good deep drawability [Hiwatashi
et al., 1994; Nagasaka et al., 1999a,b], due to the retarding inuence of the
hydrostatic pressure generated during the drawing operation on the strain-
induced transformation of austenite.
Stretch formability
In this test, rectangular blanks marked with circular grids on the surface
by electrochemical etching method are deformed under a punch until local
fracture occurs. This is also sometimes referred to as Limiting Dome Height
test [Im et al., 2000b; Lee et al., 2004, 2002a]. After the test, the circular
grids turn into oval shape. The strain values on the major and the minor
axes are measured and used to construct a Forming Limit Curve.
The maximum stretch height and the maximum stretch load are also
measured. Good stretch formability is expressed by a high stretch height and
a low stretching load [Sugimoto et al., 1995; Sugimoto and Kobayashi, 1994].
Microstructures with isolated retained austenite islands in ferrite away from
bainite have better stretch formability than networks of austenite.
Stretch angeability
Stretch angeability is a more complex property assessed by a two step pro-
cess viz. (a) hole punching and (b) hole expanding [Sugimoto et al., 1999].
A hole is rst punched on the steel blank and then allowed to expand until
cracks appear on the surface. The ratio of the increase in diameter of the
hole to the original diameter is measured in this test.
The key to superior stretch angeability is to maintain a high level
of untransformed retained austenite on hole punching, with a large ductil-
ity through TRIP eect on hole expanding. Thus, stretch angeability is
mainly controlled by the stability of the retained austenite rather its volume
fraction [Sugimoto et al., 1999]. Carbon content and morphology of retained
austenite as well as the temperature of operation are reported to have signif-
icant inuence on stretch angeability [Nagasaka et al., 1996, 1998].
2.8.2 Crash-worthiness
One of the key requirements for automobile steels is crash-resistance. In a
crash scenario, the strain rate may well exceed 250 s
1
and the prevailing
state of stress or strain may be much more complex than uniaxial tension.
Static and quasi-static tensile tests performed with a slow strain rate of about
0.005 s
1
are therefore not sucient. Servo-hydraulic testing system, Split
Hopkinson Bar, pneumatic tensile impact tester, dynamical tensile tests on
impact loading, crash tests on welded double hat-shaped specimens are some
examples of testing procedures for crash-resistance assessments [Wei et al.,
2003; Yan and Xu, 2003, 2002].
TRIP-assisted steels, like all other steels, exhibit a positive strain-rate
sensitivity index, which means that an increase in strain rate increases the
ow stress but elongation is reduced. However, superior elongation has been
reported for TRIP-assisted steels even at high strain rate [Uenishi et al., 2000;
Wei et al., 2002a,b], presumably due to the TRIP eect. A comparison on
crash-worthiness of dierent types of steels can be found in Figure 2.19 [Uen-
ishi et al., 2000]. With crash-worthiness justied properly, these steels can
be used as front or rear door side-impact bars in a passenger car.
2.8.3 Fatigue resistance
The importance of fatigue resistance is obvious for a steel to be used in auto-
mobile components that are routinely subjected to cyclic loading [Yan, 2002].
Prestraining has been identied to be the most eective factor increasing fa-
tigue strength.
Fatigue strength increases signicantly in steels having an allotriomor-
Figure 2.19: Superior crash-worthiness of TRIP-assisted steels compared to
other types, the steels are tested with a prestrain of 0% and 5% in each
case [Uenishi et al., 2000].
phic ferrite matrix when compared to bainitic steels with retained austen-
ite [Song et al., 2003, 2001]. This is because of the higher compressive stress
in the matrix of the former type of materials due to the strain-induced trans-
formation of the retained austenite during prestraining.
Dual phase steels are known to exhibit high fatigue strength. Fig-
ure 2.20 illustrates comparable fatigue limit of TRIP-assisted steels with dual
phase steels [Takahashi, 2003]. Besides solid solution hardening of ferrite
with silicon, compressive residual stress due to the martensitic transforma-
tion during cyclic loading can act as an extra advantage in TRIP-assisted
steels. These materials therefore appear to play a signicant role in modern
automobiles with lighter body weight but adequate safety.
Figure 2.20: Eect of cyclic yield stress on fatigue limit of dual phase and
TRIP-assisted steels [Takahashi, 2003].
2.8.4 Bake hardening
Bake hardening refers to the increase in strength of a steel sheet as a result
of thermal ageing at an elevated temperature after a small amount of pre-
straining. This oers a practical advantage of boosting the strength level of
a formed component during the process of paint baking.
A nominal increase of about 25 MPa strength has been reported in each
of two TRIP-assisted steel grades of 0.2C-1.5Si-1.5Mn (all in wt%) with and
without 0.5 wt% addition of copper [Wang et al., 2004]. After the baking
treatment at a temperature of 170
C for 1200 s, some retained austenite

transforms to bainite with simultaneous increase in carbon content of the
39 2.9 Limitations
untransformed austenite. Retained austenite in these steels has indeed been
shown to undergo decomposition into a mixture of ferrite and cementite
when exposed to high temperature [Sugimoto et al., 1992b]. The increase
in strength expected from the increase in carbon content of retained austen-
ite has been observed to negate the decrease in strength due to the loss of
austenite forming bainite. The bake hardening has thus been ascribed to the
ferritic matrix of these steels.
2.9 Limitations
The attractive mechanical properties of TRIP-assisted steels make them suit-
able for the manufacture of new-generation passenger cars. However, there
are many problems encountered during the processing of these materials due
to the high silicon content.
Use of thinner gauge sheets of high-strength steels in automobile com-
ponents necessitates protection against corrosion using zinc coatings through
hot-dip galvanising process [Mintz, 2001]. This consists of dipping the steel
in a molten zinc bath maintained at a temperature of around 450
C, at which
iron and zinc share great anity and allow an alloy to form, whereas pure
zinc prevails at the surface. The nal product is a steel surface protected
with a zinc coating.
The cold rolled sheets are initially heated to a temperature in the (+)
phase region. This is followed by another dwelling at a lower temperature
of some 400
C. In between these two stages, the sheet is passed through

a molten zinc bath for galvanising. A typical process cycle is depicted in
Figure 2.21.
40 2.9 Limitations
Figure 2.21: Schematic illustration of a typical continuous galvanising line.
Silicon causes problems during the galvanising operation [Mahieu et al.,
2002b; Mintz, 2001; Pichler et al., 2003, 2002]. The intercritical annealing is
carried out in a furnace maintained with low oxygen partial pressure before
hot-dip galvanising. Under this condition, iron does not get oxidised but
silicon can still be oxidised preferentially. This gives rise to a sheet surface
with poor wetability in molten zinc.
Silicon also aects the nature of the surface scale formed in a hot rolled
material. During hot rolling, the steel is exposed to air at high temperature.
This leads to surface oxidation. Normally the oxide layer consists mainly of
FeO. But in TRIP-assisted steels, the surface scale changes to complex oxides
such as Fayalite (Fe
x
SiO
y
). This reduces the surface quality and aects the
rolling properties adversely by changing the friction coecient.
41 2.10 Other variants
Silicon is added in these steels mainly to inhibit cementite precipitation
during bainite formation. This can also be achieved by adding aluminium
or phosphorous instead of silicon. Materials have therefore been developed
replacing silicon partly or fully by aluminium or phosphorous [De Meyer
et al., 1999a,b; Girault et al., 2001; Jacques et al., 2001a; Pichler et al.,
1998]. Phosphorous however imposes the risk of embrittlement due to grain-
boundary segregation. This probably restricts its addition to some 0.2 wt%
maximum in these steels.
Substitution of silicon with aluminium helps by modifying the surface
oxide [Mahieu et al., 2002b, 2001a; Maki et al., 2001]. Aluminium-containing
steels exhibit better wetabilitiy than those made with silicon after processing
in a furnace with low dew point prior to galvanising [Maki et al., 2003].
The dew point of the annealing atmosphere aects the rate of oxidation of
the elements segregated to the surface and thereby exerts an inuence on
wetability.
2.10 Other variants
Silicon-containing steels with fully bainitic microstructures possess retained
austenite which transforms to martensite during deformation. Therefore
these alloys also rely on TRIP for the properties. The properties are ex-
cellent despite the presence of austenite mainly as thin lms interspersed
with bainitic ferrite [Bouet et al., 2000]. This is not in accordance with the
advantage associated with the ne granular austenite in contrast to the thin
lms of austenite as observed for the conventional TRIP-assisted steels [Ba-
suki and Aernoudt, 1999; Koh et al., 1998; Sugimoto et al., 1993].
42 2.11 Summary
Exciting combination of mechanical properties has also been reported
by Sugimoto et al. [2004] in fully bainitic steels containing carbon in the
range of 0.1-0.6 wt% and 1.5 wt% each of silicon and manganese. Mechani-
cal behaviour of these steels were investigated at dierent temperatures. In
the temperature range 0
C to 75
C, development of mechanical properties

has been mainly ascribed to the TRIP eect. However, in the tempera-
ture range 200
C to 300
C, steels containing carbon in excess of 0.4 wt%

showed strain-induced bainitic transformation and dynamic strain-ageing,
rather than strain-induced formation of martensite.
Steels with fully bainitic microstructures have been found to display
better stretch angeability than the conventional TRIP-assisted steels [Sug-
imoto et al., 2002a, 2000]. Further enhancement of the properties has also
been observed in steels with annealed martensite as the matrix phase [Sugi-
moto et al., 2002b].
2.11 Summary
Modern TRIP-assisted steels are lean in chemical composition. Nevertheless,
it is possible to retain some 10-30 vol.% of austenite in ferrite-rich microstruc-
tures of these alloys. This is achieved by exploiting the fact that addition
of silicon or aluminium helps inhibiting the precipitation of cementite dur-
ing bainite formation and hence allows carbon to remain dissolved in the
untransformed austenite. Silicon, however, degrades the surface quality of
the product and also makes the processing dicult, due to the formation
of adherent surface scale. Replacing silicon by aluminium is thus generally
thought as a practical solution to make these steels.
43 2.11 Summary
The steels record superior uniform elongation, compared to dual phase
steels of similar strength, and hence are suitable for the manufacture of cars
with lighter body weight yet adequate safety. However, TRIP-assisted steels
are generally found to suer from discontinuous yielding unlike dual phase
steels.
The large uniform elongation of TRIP-assisted steels is mainly at-
tributed to the strain-induced transformation of the retained austenite into
martensite. Retention of austenite in these alloys is due to the carbon en-
richment. Martensite that forms also inherits the high carbon content of the
austenite and should therefore be brittle. This does not reconcile with the
remarkable elongation observed in these steels.
The rate of strain-induced transformation of austenite appears to be
the key to optimise ductility of these materials. The stability of austenite
should be such that it transforms into martensite progressively during defor-
mation. Carbon content of the retained austenite, among many other factors,
inuences its stability. Therefore, not only the volume fraction of retained
austenite but also its carbon content appears to be salient in controlling the
properties. It is therefore important to understand the variation of these pa-
rameters as function of the processing conditions and chemical compositions
of these steels.
Role of plastic strain in austenite on subsequent transformation into
martensite is ambivalent. Strain creates defects that can act as additional
nucleation sites for transformation. However, an excessive defect density due
to large plastic strains can inhibit the growth of martensite plates, ultimately
leading the transformation to a halt. This is called mechanical stabilisation.
44 2.11 Summary
The properties of TRIP-assisted steels are generally thought to rely on the
strain-induced transformation of austenite. Hence it is essential to under-
stand if there is any contention of mechanical stabilisation in these steels.
However, there exists no quantitative theory for predicting the onset of me-
chanical stabilisation.
These steels are often subjected to dierent complex tests for assess-
ing formability. The results are often interpreted in terms of the strength-
elongation product but without proper justication. There has been no
proper attempt in understanding the relationship between the forming prop-
erties and the tensile properties.
Chapter 3
Brittle martensite
Modern TRIP-assisted steels are generally believed to rely on stress or strain-
induced martensitic transformation of retained austenite to achieve proper-
ties. The carbon concentration of the retained austenite is usually in excess
of 1 wt%. Martensite that forms in these materials inherits the high carbon
content of austenite and remains in freshly produced untempered condition.
This is therefore expected to be brittle, but the mechanical properties, espe-
cially formability, of these materials are nevertheless impressive. This does
not reconcile with the apparent brittleness of the high-carbon martensite.
The purpose of work presented in this chapter is to explain this contradic-
tory behaviour.
3.1 Hypothesis
Martensite that forms from the high-carbon austenite in the TRIP-assisted
steels actually exists in a composite microstructure. This can be imagined
to be similar to a uniaxially aligned, discontinuous, brittle-bre composite
45
46 3.2 Experiments
material, with the brittle martensite plates embedded in a ductile austenite
matrix. During loading, axial stress is transmitted from the matrix into the
bre via shear loading of the cylindrical surface of the bre, , [Cox, 1952].
This load transfer causes the tensile stress in the bre to rise to a
maximum value,
P
, half way along the length of the bre. If
P
exceeds the
fracture strength,
F
, then the bre breaks. However, if the bre is short,
end eects prevent the maximum stress in the bre from reaching
F
. The
bre then remains intact. The critical bre length below which it does not
fracture is designated l
C
. Fibres longer than l
C
tend to break into segments
of length l
C
.
It is proposed here that load transfer into the hard martensite becomes
similarly dicult as the scale of the martensite in a relatively soft matrix
decreases. In other words, the martensite will not crack easily if its size
is reduced below some critical value. Naturally, work-hardening comes into
this scenario so that sucient load should eventually be transferred into the
martensite to cause fracture. However, the stage at which this occurs ought
to be delayed when the martensite is ne.
3.2 Experiments
Material and heat treatment
A steel with the chemical composition of Fe-0.98C-1.46Si-1.89Mn-1.26Cr-
0.26Mo wt% was selected. The bulk composition of the alloy represents the
high-carbon austenite, hence the martensite, that exists in the modern TRIP-
assisted steels. The alloy was homogenised at 1200
C for 48 h. Cylindrical
47 3.2 Experiments
samples of length 12 mm and diameter 8 mm were made out of this steel
and sealed in quartz tubes containing a partial pressure of argon for heat
treatment in an air furnace.
The sealed samples were austenitised at 1200
C, 1100
C, 1000
C and
950
C for 10 minutes in order to generate a range of austenite grain sizes for

subsequent transformation into martensite during quenching in water. An-
other set of samples was cooled slowly by switching o the furnace following
the same austenitising conditions. This was done in order to obtain a clear
denition of the austenite grain boundaries. One more sample was heated to
1200
C for 10 minutes followed by soaking at 950
C for another 10 minutes

before nally quenching to room temperature.
Metallography
Cross sections of the quenched samples were polished using standard met-
allographic techniques. They were then examined in unetched states using
optical microscopy to characterise any cracks. The distances between ad-
jacent cracks in the individual plates of martensite were measured using a
graticule tted to the microscope.
The quenched samples were etched using 10 wt% aqueous sodium
metabisulte solution to reveal the martensitic microstructures. The ap-
parent martensite plate lengths measured on two-dimensional sections were
determined in a similar way as the distance between cracks.
The furnace cooled samples were etched with boiling alkaline sodium
picrate solution (made with 2 g picric acid, 25 g NaOH and 100 mL distilled
water) to reveal the austenite grain size as a function of the austenitising
temperature; the size was characterised using the lineal intercept (L) method.
48 3.3 Results and discussion
3.3 Results and discussion
Etched surfaces of the quenched samples revealed the dark-etched lenticular
plates of martensite with untransformed regions of austenite, as presented
in Figures 3.1-3.4. A signicant amount of untransformed austenite can be
seen in each sample. This is due to the high carbon content of the alloy.
Figure 3.1: Micrograph of the sample quenched after austenitising at 1200
C.
C.
C.
C.
Martensite plates exist with a distribution of plate lengths in each of the
samples. Growth of the initial plates is restricted only by the austenite grain
boundaries, while those formed subsequently are also stopped by the pre-
existing plates. New plates forming during the progress of transformation
therefore tend to be ner than the earlier ones, the coarsest ones in each
sample being given by the prior austenite grain size. The microstructures of
these samples therefore get rened at lower austenitising temperatures that
lead to ner prior austenite grain sizes (Table 3.1).
Austenitising temperature Mean lineal intercept % Error
(
C) (m)
1200 410 13
1100 124 20
1000 65 18
950 25 26
Table 3.1: Variation of austenite grain size with austenitising temperature,
error represents 1.
Quenched samples in the polished but unetched conditions were exam-
ined using an optical microscope to observe any cracks. Samples austenitised
at 1200
C and 1100
C revealed cracks across the coarser plates in periodic

arrays. In general, the ner plates were found to be less susceptible to crack-
ing. Only a few isolated cracks were observed in the sample quenched from
1000
C and the plates, in general, were observed to be singly-cracked. The

sample quenched from 950
C was found to be free of any cracks. Micrographs

obtained from the polished and the etched surfaces of the samples quenched
from 1200
C and 1000
C are presented in Figures 3.5 and 3.6 respectively.

The cracking of the martensite plates observed in these experiments
is presumably due to the quench stresses. These stresses might be thought
to increase with higher austenitising temperatures and thus could aect the
observation. To eliminate any possible role of this dierential quench stresses,
one sample was austenitised at 1200
C for 10 minutes and soaked at 950
C for
another 10 minutes before nally quenching in water. Similar microcracks as
the sample directly quenched from 1200
C were also observed in this sample,

as depicted in Figure 3.7.
(a)
(b)
Figure 3.5: Microcracking in the sample austenitised at 1200
C: (a) etched
sample; (b) periodically cracked plate in unetched sample.
The apparent distances between two adjacent cracks in individual frag-
mented plates of martensite in the samples austenitised at 1200
C and 1100
C
were measured using optical metallography. These measurements follow a
distribution as depicted in Figure 3.8. The distances are designated l
a
to
(a)
(b)
C: (a) etched
sample where cracks are dicult to detect; (b) occasional cracks visible in
unetched sample.
emphasise that they are apparent measures because stereological eects are
not taken into account. The mean values for these measurements were found
to be l
a
= 58 m and l
a
= 38 m for the samples austenitised at 1200
C and
C and soaked
at 950
C before quenching.
1100
C respectively.
The minimum value observed in all the measurements, within experi-
mental error, was found to be 10 m. This is more relevant than the mean
value in the present context. If it is assumed that the critical stress transfer
length l
C
= 10 m, then an austenite grain size which is somewhat less than
this value would avoid quench cracking of the type explored in this work.
One diculty is that cracks were not detected in the T
=950
C sample with
a prior austenite grain size of 25 m, which is greater than l
C
. It is possible
that the cracks exist but have been missed because the sample contains few
plates of length greater than l
C
. Therefore, the plate length distributions
were also measured for each sample.
The apparent lengths of approximately 180 martensite plates from each
sample were measured using optical metallography. In all cases the plates
were chosen at random by arbitrarily translating the stage across a cursor
which identied the plate to be measured. The size distributions are illus-
Figure 3.8: The frequency of observation versus the apparent distance be-
tween cracks on individual martensite plates in the samples with the coarsest
microstructures (T
= 1200 and 1100
C).
trated in Figure 3.9, plotted on identical scales to facilitate comparisons.
It is evident that a coarse grain size tends to have a more uniform dis-
tribution of plate sizes (Figure 3.9(a)) whereas the smallest plates dominate
when the austenite grain size is reduced (Figure 3.9(d)). This is consistent
with the work by Guimaraes and co-workers on the heterogeneity of marten-
sitic transformation when the austenite grain size is small [Guimaraes and
55 3.4 Simulating microstructural evolution
Saavedra, 1984; Guimaraes and Gomes, 1978]. Some austenite grains trans-
form via a burst of martensite, while others remain fully austenitic. Such
bursts can be expected to lead to a greater preponderance of ne plates. By
contrast, coarse grained microstructures transform uniformly with the ma-
jority of austenite grains participating, thus allowing a gradual evolution of
microstructure as the sample is cooled.
3.4 Simulating microstructural evolution
The objective in this section is to theoretically compute the evolution of
martensite plate size distribution. As mentioned earlier, formation of marten-
site plates partitions the parent austenite grain into several compartments.
Untransformed regions of austenite naturally decrease in size with more
plates of martensite forming. Thus with the progress of transformation, new
martensite plates form with reduced plate lengths generating a distribution
of plate sizes. This concept, as proposed originally by Fischer [1949], has
been used here to calculate the theoretical distribution of martensite plate
lengths:
At n
th
instant (n varying as 1, 2, 3, . . .), the number of austenite pockets
available for transformation, N, is given by
N = 2
n1
(3.1)
This is taken as the number of martensite plates forming at that instant,
assuming only one martensite plate forms from a given pocket.
The volume of each of the austenite pockets at the n
th
instant, V
n
, is
Figure 3.9: The distribution of apparent martensite plate lengths (T
and L
values stated). The distributions are all on the same scale to allow compar-
isons to be made. The bin size on the horizontal scale is 5 m.
given by
V
n
=
V
0
2
n1
(3.2)
where V
0
is the original volume of austenite grain. The volume of austenite
pockets can be converted into mean lineal intercept, invoking the relation-
ships L = 1.69a and V = 11.314a
3
for truncated octahedral grains, where L
is the mean lineal intercept, V is the volume and a is the edge length of the
unit cell [Mark, 1956].
This results in the following equation relating L
n
to V
n
at n
th
instant:
L
n
= (0.4266V
n
)
1
3
(3.3)
This can be considered to be the length of any martensite plate forming at
that instant.
There is, however, one limitation of the above formulation. Volume
of the pre-existing martensite plates has not been taken into account while
calculating the volume of austenite available for transformation, V
n
, at the
n
th
instant. Shape of the martensite plates are generally considered to be
thin circular discs with a nite volume given by r
2
c; 2r being the length of
the plate and c, the semi-thickness, which is assumed to be 0.05r. The total
martensite plate volume is thus subtracted from the instantaneous austenite
pocket volume in order to calculate the eective austenite volume.
Knowing the initial mean lineal intercept of the austenite grain size,
it is now possible to obtain a theoretical variation in the martensite plate
length using the above methodology. The theoretical and the experimental
plate length distributions obtained with dierent prior austenite grain sizes
are presented in Figure 3.10.
The calculated and measured curves compare well, in terms of the in-
uence of the initial austenite grain size and with respect to their general
shapes. However, the absolute reduction in length is much greater with the
calculated curves. This is because equation 3.1 implies that the number of
plates N that form in each generation n is given by N = a
n1
with a = 2.
This leads to a rapid increase in N and a correspondingly rapid decrease in
the plate length as n increases. Thus, whereas each point in the calculated
curve can represent N plates of identical length, every point in the measured
curve is a dierent plate. The gradual variation in the measured plate length
distribution implies that a < 2.
(a)
(b)
Figure 3.10: (a) The experimentally measured apparent lengths of martensite
plates arranged in the sequence in which they are assumed to form. (b) The
corresponding curves for the calculated length versus the order in which the
plates form.
60 3.5 Martensite in TRIP-assisted steels
3.5 Martensite in TRIP-assisted steels
In TRIP-assisted steels, austenite is retained because of its carbon enrich-
ment following the bainite reaction. Partial decomposition of the untrans-
formed austenite into martensite during cooling after bainite reaction is not
unlikely. This is because the distribution of carbon in untransformed austen-
ite after the bainite reaction is mostly heterogeneous [Matas and Hehemann,
1961]. Areas near the bainitic ferrite-austenite interface are richer in carbon
than those away from it. The austenite away from the interfaces may trans-
form into martensite on cooling to ambient temperature. Indeed there is
evidence of martensite, albeit traces, formed during heat treatment of these
materials [Jacques et al., 2001d; Zaeerer et al., 2004]. The austenite in
TRIP-assisted steels is in a nely divided state with a grain size of about
1 m [Jacques, 2003]. It therefore follows that martensite forming in these
ne regions of austenite is likely to be devoid of microcracks.
Similarly, the high-carbon martensite produced during plastic deforma-
tion of TRIP-assisted steels can also be expected to resist cracking because
of the ne plate size. However, the cracking of martensite plates discussed
in Section 3.3 results from the quench stresses. The nature of these stresses
is arbitrary and has not been characterised.
Further studies were thus carried out with the following objectives:
To investigate the cracking tendency of the martensite plates during
controlled straining.
To study the nature of the martensite formed during deformation of
austenite.
61 3.6 Experiments
3.6 Experiments
Material
The alloy used for the experiments described in Section 3.3 was modied
with nickel to depress the M
S
temperature below room temperature. A
model developed by Capdevila et al. [2003] was used to study the eect
of adding nickel on the M
S
temperature of Fe-1C-1.5Mn-1.5Si (wt%) alloy
(Figure 3.11). From this, it appears that about 8 wt% nickel can depress
the M
S
temperature of the alloy suciently below room temperature. The
alloy can therefore be expected to be fully austenitic when quenched to room
temperature after austenitising at a higher temperature.
-60
-50
-40
-30
-20
-10
0
7 8 9 10 11
M
s

t
e
m
p
e
r
a
t
u
r
e

/

o
C
Ni content / wt%
Figure 3.11: Calculated eect of Ni addition on the M
S
temperature of Fe-
1C-1.5Mn-1.5Si (wt%) alloy.
The chemical composition of the modied alloy was analysed to be Fe-
0.91C-8.42Ni-1.58Mn-1.55Si-1.15Cr-0.23Mo (wt%). Calculations using MT-
DATA [2005] with the SGTE database reveal austenite to exist as a single
62 3.6 Experiments
phase at 1100
C (Figure 3.12). Austenite, ferrite, cementite and liquid phases

were allowed to exist in equilibrium for the calculations. The alloy was there-
fore homogenised at 1100
C for 48 h before further experiments.

0
20
40
60
80
100
800 1000 1200 1400 1600
a
m
o
u
n
t

o
f

p
h
a
s
e

/

w
t
%
Temperature /
o
C
Liquid
Austenite
Cementite
Figure 3.12: Diagram showing the calculated temperature range at which
austenite exists as the single phase in the modied alloy.
Heat treatment
Cylindrical samples of length 10 mm and diameter 5 mm were made out of
the Fe-C-Ni alloy. These samples were then sealed in quartz tubes containing
a partial pressure of argon for heat treatment in an air furnace. The sealed
samples were austenitised at 1100
C for 10 minutes before quenching in water

at room temperature. One set of samples was further quenched in liquid
nitrogen to produce martensite.
63 3.6 Experiments
Compression tests
The heat-treated samples were subjected to compression experiments. In
these tests, samples with initial diameter d
0
and height h
0
were compressed
to a height h. Compression strains were then measured as the percentage
reductions in the height of the samples. From the constancy of volume crite-
rion, the diameter of the compressed samples, d, is expected to be larger than
the initial diameter and is given by the relation d
2
h = d
2
0
h
0
. The portion of
the sample in the mid-height region is free to ow outward. However, friction
at the contact surfaces of the samples with the anvils restricts the material
ow at the ends.
The dierential ow of material leads to a barrelled prole of the com-
pressed specimen with non-uniform strain distribution across the sample di-
mensions. This can be minimised with the use of lubricants at the contact
surfaces of the samples with the anvils. The lubricant grease used in the
present work consists of silicone oil and lithium soap. The chance of bar-
relling is also reduced in samples with a higher h
0
/d
0
ratio [Dieter, 1988].
But there is a practical limit of the h
0
/d
0
2 (which is used for the samples
in this work), beyond which the samples may buckle during deformation.
Tests were performed at room temperature as well as at -501
C. The
crosshead velocity used was 0.0017 mm s
1
. The martensitic samples were
compressed at room temperature to two dierent strain levels viz. 3% and
7%. A strain of about 30% was applied in case of the austenitic sample
compressed at a temperature of -501
C.
Metallography
The compressed samples were sectioned along the vertical direction. These
sections were then polished, etched with 2% nital and observed using optical
microscopy.
Figure 3.13 depicts the microstructure of the sample quenched in liquid ni-
trogen. The micrograph reveals dark-etching lenticular plates of martensite
along with untransformed austenite. This is similar to the microstructures
presented in Figures 3.1-3.4. However, microcracking of the martensite plates
of the type illustrated in Figures 3.5-3.6, was not observed in Figure 3.13.
This is despite the comparable scales of the microstructures. It is possible
that quenching in liquid nitrogen from room temperature does not induce
enough stress to crack the martensite plates, unlike quenching in water after
soaking at high temperatures. The addition of nickel might also be thought
to have reduced the characteristic brittleness of the high-carbon martensite
formed in the present alloy.
Martensitic samples produced by quenching in liquid nitrogen were fur-
ther subjected to compression tests. Figure 3.14 presents the microstructures
of the compressed samples. The plates of martensite are seen to contain
arrays of parallel cracks across the plate length. This is similar to the mi-
crocracks observed in Figure 3.5. Furthermore, the larger plates are found
to be preferentially cracked over the smaller ones (Figure 3.14(a)). This is
consistent with the earlier quenching experiments that revealed a greater
Figure 3.13: Plates of athermal martensite formed without any microcracks
in the present alloy.
propensity of the coarser martensite plates to crack (Section 3.4). Cracks
were more frequent following higher plastic deformation (Figure 3.14(b)).
Sometimes cracks were also visible along some portions of the martensite-
austenite interfaces in the 7% strained sample. The apparent distance be-
tween adjacent cracks was also measured in these samples, Figure 3.15. This
is found to follow a similar distribution as that observed in Figure 3.8.
Figure 3.16 presents the micrographs obtained from the sample
quenched at room temperature and then compressed by 30% at -501
C.
Figure 3.16(a) reveals clearly the dark-etching small lens shaped martensite
plates formed during straining of the austenite. Strain-induced martensite
retained the lenticular shape but the individual plates are much ner in size
than expected from the prior austenite grain size. Martensite plates cannot
traverse across strong obstacles like grain boundaries, twin boundaries or
other pre-existing martensite plates. In the present situation, growth of the
(a)
(b)
Figure 3.14: Cracking of martensite plates after (a) 3% and (b) 7% reduction
in height of the martensitic samples.
martensite plates appears to be stied by the defects produced due to plastic
strain in austenite. Hence the plates tend to grow only upto limited sizes.
(a)
(b)
Figure 3.15: The frequency of observing cracks versus the apparent distance
between cracks on martensite plates after (a) 3% and (b) 7% reduction in
height.
The plates were also observed to be free of cracks.
A notable feature of the microstructure is the particular alignment of
the martensite plates along the two preferred directions (Figure 3.16). This is
due to the eect of stress on transformation [Patel and Cohen, 1953]. Marten-
site is known to form on some specic crystallographic planes of austenite
known as habit planes. Without any external stress, transformation takes
place on all the possible habit planes. This therefore does not lead to any
bias in the microstructure. By contrast, martensite tends to form on only
some specic habit planes under the inuence of stress. This explains the
particular alignment of the martensite plates along some preferred directions
in the micrographs presented in Figure 3.16.
Figure 3.16(b) depicts that martensite is formed in one of the two grains
visible in the micrograph, while there is no sign of transformation in the other
grain. This may be due to the role of stress as described above. The resolved
shear component of the external stress on the habit planes actually leads to
transformation in steels, where the volume expansion due to transformation
is small. Transformation becomes possible only above some critical magni-
tude of the resolved shear stress. The orientation of the individual austenite
grains must inuence the magnitude of the resolved shear stress on the crys-
tallographic planes. Therefore it is not unlikely that some grains may not
be favourably oriented for transformation under the inuence of stress, as
observed in Figure 3.16(b). The other possibility could be the heterogeneous
nature of the compressive strain. All the grains might not be strained to
the same extent under compression. This may also be responsible for no
transformation to occur in one of the two grains visible in Figure 3.16(b).
69 3.8 Summary
3.8 Summary
High-carbon martensite is known to be a brittle phase. Yet it does not ap-
pear to be a liability in the TRIP-assisted steels reputed for good mechanical
properties, especially formability. This apparent anomaly can be explained
considering the role of martensite plate size on the brittleness. It has been
demonstrated that the ner martensite plates are better able to resist crack-
ing under stress or strain. It is the ne grain size of austenite that allows the
high-carbon brittle martensite to form in these materials without endanger-
ing the mechanical properties.
It has been also shown that the plates of martensite formed during
straining of austenite remain ne in size. This is probably due to the defects,
created in austenite due to strain, that stie the growth of martensite. The
small plates of strain-induced martensite are observed to form only along
some preferred directions. This is due to the eect of the accompanying
stress. The ne-sized ordered martensite plates that form during the strain-
ing of austenite can be expected to be resistant to cracking. Untempered
martensite that forms in the TRIP-assisted steels is therefore not expected
to have any adverse eect on the mechanical performance, despite its high
carbon concentration.
70 3.8 Summary
(a)
(b)
Figure 3.16: Martensite plates formed during straining of the austenitic sam-
ple; (a) showing preferred alignments of the plates, (b) revealing the prefer-
ential transformation in one of the two adjacent grains of austenite.
Chapter 4
Microstructural Modelling
Allotriomorphic ferrite, carbide-free bainite and retained austenite are con-
stituents of the microstructure of TRIP-assisted steels. These phases can be
produced beginning either with hot rolled or cold rolled steel. The austen-
ite after hot rolling transforms to allotriomorphic ferrite and subsequently
to bainite. However, a two-stage annealing process is necessary after cold
rolling to produce the desired microstructure. During bainite formation, un-
transformed austenite gets enriched with carbon and is therefore retained at
ambient temperature. Processing conditions as well as the chemical compo-
sition of the steel inuences both the volume fraction and carbon content of
retained austenite. The present chapter deals with an attempt to model the
microstructure of TRIP-assisted steels using neural network techniques.
4.1 Neural network modelling
Neural network analysis represents a general method of regression that ts
a non-linear function to experimental data. This can help quantify complex
71
72 4.1 Neural network modelling
phenomena involving many variables. The subject has been reviewed by
Bhadeshia [1999a]. In linear regression, the product of each input variable
x
i
and a weight w
i
is added to a constant to obtain the output y,
y =
i
w
i
x
i
+ (4.1)
The weights and the constants can be imagined to be varied systematically
until a best-t expression of the output is reached as a function of the input
variables. The diculties with such a regression are:
1. A specic relation has to be chosen prior to the analysis.
2. The regression equation thus derived applies across the entire input
space.
These diculties can be overcome by making the network non-linear,
setting the sum of the products to the argument of a hyperbolic tangent
function,
y =
j
W
j
h
j
+
, with h
j
= tanh(
i
w
ij
x
i
+
j
) (4.2)
where W
j
is a weight and
is a constant. The hyperbolic tangent function

introduces exibility, non-linearity and the form of the function also varies
in the input space. The strength of the function depends on the weights w
ij
.
One major limitation of the non-linear regression method is the possi-
bility of overtting the data, as illustrated in Figure 4.1. This can be veried
easily by testing the model with some unseen data. The experimental dataset
can be divided randomly into two parts viz. training dataset and test dataset.
After the model is created using the training dataset, its performance can be
checked using the test dataset.
Figure 4.1: Modelling noisy data in a case where y should vary with x
3
. The
lled points represent training data, and the circles the testing data. (a) A
simple linear function. (b) A cubic polynomial with optimum representation
of both the training and test data. (c) A fth order polynomial which does
not generalise properly. (d) Variation in the test and training errors as a
function of the model complexity [Bhadeshia, 1999a].
Figure 4.1(a)-(c) illustrates an imaginary case of modelling a dataset
where y varies as a function of x
3
. A linear function is too simple whereas an
overcomplex function fails to generalise. The optimum behaviour is described
in Figure 4.1(b). The best model is chosen by the minimum in the test error,
as the training error tends to decrease continuously with increasing model
complexity (Figure 4.1(d)).
Modelling uncertainty
An important aspect of neural network modelling is the uncertainties asso-
ciated with the predictions. The overall error in the predicted values of the
output, E
D
, can be calculated by comparing the predicted values (y
n
) against
the dataset values (t
n
) as,
E
D

n
(t
n
y
n
)
2
(4.3)
If all the important input variables are not considered in the analysis,
E
D
tends to increase. E
D
represents only an overall perceived level of noise
in the output and does not describe the uncertainties of the modelling pro-
cess itself. The best-t functions with the most probable weights may not
necessarily describe the uncertainties in regions of the input space with noisy
or sparse data.
Neural networks used in Bayesian framework allow the calculation of
error bars representing the uncertainty in the tting parameters [MacKay,
2003]. This invokes the possibility of tting or extrapolating many functions
into uncertain regions of the input space, still maintaining an optimum t
in regions having good quality data. A probability distribution of sets of
weights is then used to dene the tting uncertainty. It follows that the
error bars tend to increase in magnitude when data are noisy (A) or sparse
(B), as illustrated in Figure 4.2.
75 4.2 Model for retained austenite fraction
Figure 4.2: The uncertainty in dening a tting function when data are
noisy (A) or sparse (B). The thinner lines represent error bounds due to
uncertainties in determining the weights [Bhadeshia, 1999a].
4.2 Model for retained austenite fraction
4.2.1 Database
The cold rolled steel is initially heated to a temperature in the ( + )
phase region (intercritical annealing), to form ferrite and austenite. The
material is then cooled rapidly to a lower temperature at which austenite
transforms to bainite. Parameters of this two-step heat treatment along
with the chemical composition were taken as input variables to the model, as
shown in Table 4.1
1
. Figure 4.3 presents the distribution of the data points
in the input space for each variable.
1
Data used are due to Chen et al. [2002]; De Meyer et al. [1999a]; Itami et al. [1995];
Jacques et al. [2001a,b,c, 1999]; Kim et al. [2002, 2001]; Lee et al. [2002b]; Matsumura et
al. [1992, 1987b]; Nakagaito et al. [2003]; Pichler et al. [1998]; Sakuma et al. [1993, 1991a];
Traint et al. [2002, 2000]
Inputs Min. Max. Average S.D.
Carbon / wt% 0.0950 0.3920 0.2102 0.1020
Manganese / wt% 0.6000 1.9900 1.4057 0.2796
Silicon / wt% 0.0400 2.1000 0.9619 0.5369
Aluminium / wt% 0.0000 2.0000 0.2979 0.5236
Phosphorus / wt% 0.0000 0.2040 0.0223 0.0401
Molybdenum / wt% 0.0000 0.1400 0.0043 0.0218
Copper / wt% 0.0000 0.5100 0.0065 0.0573
IA temperature /

C 730 840 778.2653 25.7829
IA time / s 60 450 213.2143 110.8414
IT temperature /

C 300 525 404.2347 39.5554
IT time / s 0 5400 441.3673 588.7228
Output Min. Max. Average S.D.
Retained austenite / vol.% 0.1267 27.0710 10.3805 5.8054
Table 4.1: Data used for the retained austenite fraction model, IA: intercrit-
ical annealing and IT: isothermal transformation to bainite, S.D.: standard
deviation.
4.2.2 Model characteristics
The model was trained using 80% of the dataset values. The remainder
was used to test the performance of the model with unseen data. Figure 4.4
compares the normalised predicted values with the corresponding normalised
database value during training and testing steps respectively. Normalisation
of the variables is carried out as follows
x
n
=
x x
min
x
max
x
min
0.5 (4.4)
where x
n
is the normalised value of the variable x, with x
min
and x
max
as
the minimum and the maximum values respectively. The performance of the
model can be seen to be satisfactory both during the training and testing
stages.
0
0.1
0.2
0.3
0.4
0 10 20 30
c
a
r
b
o
n

/

w
t
%
retained austenite / vol.%
(a)
0
0.5
1
1.5
2
0 10 20 30
m
a
n
g
a
n
e
s
e

/

w
t
%
(b)
0
0.5
1
1.5
2
0 10 20 30
s
i
l
i
c
o
n

/

w
t
%
(c)
0
0.5
1
1.5
2
0 10 20 30
a
l
u
m
i
n
i
u
m

/

w
t
%
(d)
0
0.05
0.1
0.15
0.2
0 10 20 30
p
h
o
s
p
h
o
r
o
u
s

/

w
t
%
(e)
0
0.05
0.1
0.15
0 10 20 30
m
o
l
y
b
d
e
n
u
m

/

w
t
%
(f)
0
0.25
0.5
0 10 20 30
c
o
p
p
e
r

/

w
t
%
(g)
720
760
800
840
0 10 20 30
I
A

t
e
m
p
e
r
a
t
u
r
e

/

o
C
(h)
0
100
200
300
400
500
0 10 20 30
I
A

t
i
m
e

/

s
(i)
250
350
450
550
0 10 20 30
I
T

t
e
m
p
e
r
a
t
u
r
e

/

o
C
(j)
0
2000
4000
6000
0 10 20 30
I
T

t
i
m
e

/

s
(k)
Figure 4.3: Distribution of data in the input space for the retained austenite
fraction model
Figure 4.5(a) illustrates the reduction in noise with increasing number
of hidden units. The nal model is formed as a committee of models, which is
selected on the basis of the observed minimum in the test error on combining
increasing numbers of networks (Figure 4.5(b)).
Figure 4.6 presents the signicances of each of the input variables as
perceived by the models in the nal committee. Bainite reaction time appears
to be the most important factor controlling the retained austenite content in
the microstructure.
4.2.3 Model predictions
Predictions were made by varying input variables over a base steel compo-
sition of 0.2C-1.5Mn-1.0Si-0.5Al-0.02P-0Mo-0Cu (all in wt%), intercritical
annealing parameters of 780
C, 300 s and bainite transformation conditions

as 400
C, 500 s. Both carbon and manganese are known to be austenite

stabilisers [Pickering, 1978]. Therefore as expected, both carbon and man-
ganese are benecial for increasing austenite fraction (Figure 4.7(a)-(b)).
Similarly, the addition of copper to these steels can also increase austenite
content [Kim et al., 2002] (Figure 4.7(f)). This is probably because copper
reduces the driving force, G, for the transformation of austenite to fer-
rite. Therefore these elements seem to aect the retained austenite content
through the thermodynamic stability of austenite over ferrite.
Additionally, silicon, aluminium and phosphorous are known to help
retain austenite by suppressing cementite formation during the bainite reac-
tion. Figure 4.7(c) depicts this role of silicon on increasing retained austenite
fraction. However, with 1 wt% silicon, the addition of aluminium up to about
0.6
0.4
0.2
0
0.2
0.4
0.6
0.6 0.4 0.2 0 0.2 0.4 0.6
Database values, normalised
P
r
e
d
i
c
t
e
d

v
a
l
u
e
s
,

n
o
r
m
a
l
i
s
e
d
(a)
0.6
0.4
0.2
0
0.2
0.4
0.6
0.6 0.4 0.2 0 0.2 0.4 0.6
P
r
e
d
i
c
t
e
d

v
a
l
u
e
s
,

n
o
r
m
a
l
i
s
e
d
(b)
Figure 4.4: Model predictions compared with the actual database values
during (a) training and (b) test.
1 wt% does not appear to have any marked eect on the retained austenite
fraction (Figure 4.7(d)). Predictions with higher aluminium content are as-
N
o
i
s
e
0.06
0.08
0.1
0.12
0 4 8 12 16 20
Number of hidden units
(a)
0.2
0.3
0.4
0.5
0 4 8 12 16 20
C
o
m
b
i
n
e
d

T
e
s
t

E
r
r
o
r
Number of models
(b)
Figure 4.5: (a) Variation of noise () with increasing hidden units. (b)
Variation of the combined test error with number of models in committee.
sociated with large uncertainties suggesting more data are necessary in this
range of the input space. Molybdenum increases hardenability and therefore
Figure 4.6: Signicances of the input variables (A-K) for the retained austen-
ite fraction model, A: carbon, B: manganese, C: silicon, D: aluminium, E:
phosphorous, F: molybdenum, G: copper, H: intercritical annealing temper-
ature, I: intercritical annealing time, J: bainite transformation temperature
and K: bainite transformation time.
is likely to increase retained austenite fraction. Figure 4.7(e) depicts only a
very subtle eect of molybdenum on increasing retained austenite fraction,
with large uncertainties.
Figure 4.8(a) depicts the variation of retained austenite fraction with
intercritical annealing temperature. It is generally believed that higher inter-
critical annealing temperature produces more austenite but with less stability
(due to lower carbon content) whilst stability increases but the amount of
austenite decreases at lower temperatures [Gallagher et al., 2003]. There-
fore, an intermediate annealing temperature is generally considered to max-
0
5
10
15
20
25
30
0 0.1 0.2 0.3 0.4 0.5
r
e
t
a
i
n
e
d

a
u
s
t
e
n
i
t
e

/

v
o
l
.
%
carbon / wt%
(a)
0
5
10
15
20
25
30
35
0.5 1 1.5 2 2.5
r
e
t
a
i
n
e
d

a
u
s
t
e
n
i
t
e

/

v
o
l
.
%
manganese / wt%
(b)
0
4
8
12
16
0 0.5 1 1.5 2 2.5
r
e
t
a
i
n
e
d

a
u
s
t
e
n
i
t
e

/

v
o
l
.
%
silicon / wt%
(c)
0
5
10
15
20
0 0.5 1 1.5 2 2.5
r
e
t
a
i
n
e
d

a
u
s
t
e
n
i
t
e

/

v
o
l
.
%
aluminium / wt%
(d)
2
4
6
8
10
12
14
0 0.04 0.08 0.12 0.16 0.2
r
e
t
a
i
n
e
d

a
u
s
t
e
n
i
t
e

/

v
o
l
.
%
molybdenum / wt%
(e)
0
5
10
15
20
25
0 0.2 0.4 0.6 0.8 1
r
e
t
a
i
n
e
d

a
u
s
t
e
n
i
t
e

/

v
o
l
.
%
copper / wt%
(f)
Figure 4.7: Eect of the alloying elements on retained austenite fraction: (a)
carbon, (b) manganese, (c) silicon, (d) aluminium, (e) molybdenum and (f)
copper.
imise retained austenite content in the nal microstructure. However, Kim
et al. [2001] reported higher retained austenite content with an intercritical
2
6
10
14
18
22
720 760 800 840 880
r
e
t
a
i
n
e
d

a
u
s
t
e
n
i
t
e

/

v
o
l
.
%
intercritical annealing temperature /
o
C
(a)
0
5
10
15
20
25
30
35
0 100 200 300 400 500
r
e
t
a
i
n
e
d

a
u
s
t
e
n
i
t
e

/

v
o
l
.
%
intercritical annealing time / s
(b)
0
5
10
15
20
25
30
250 300 350 400 450
r
e
t
a
i
n
e
d

a
u
s
t
e
n
i
t
e

/

v
o
l
.
%
bainite reaction temperature /
o
C
(c)
0
4
8
12
16
0 500 1000 1500 2000 2500
r
e
t
a
i
n
e
d

a
u
s
t
e
n
i
t
e

/

v
o
l
.
%
bainite reaction time / s
(d)
Figure 4.8: Eect of heat treatment parameters on retained austenite frac-
tion: (a) intercritical annealing temperature and (b) time; (c) bainite trans-
formation temperature and (d) time.
annealing temperature lower than the intermediate. This was ascribed to
the greater stability of the austenite produced at a lower intercritical anneal-
ing temperature. Predictions from the present model suggest, either a high
or a low intercritical annealing temperature for obtaining greater retained
austenite content (Figure 4.8(a)).
Unlike interstitial carbon, the diusion of substitutional solutes needs
longer times. Increasing intercritical annealing time favours more ecient
partitioning of the substitutional solutes like manganese in austenite. Man-
ganese, like carbon, stabilises the austenite. This can therefore increase the
retained austenite content in the nal microstructure (Figure 4.8(b)).
The eect of bainite reaction temperature and time on retaining austen-
ite is depicted in Figure 4.8(c)-(d). Lower bainite transformation tempera-
tures mean a slower rate of bainite formation. For a given transformation
time, the amount of bainite formed is less at lower transformation temper-
atures. This will leave more untransformed austenite. Also, transformation
ceases with higher carbon enrichment in the untransformed austenite at lower
temperatures (T
0
criterion of the incomplete reaction phenomenon). Thus
austenite becomes more stable at lower transformation temperatures. But
this cannot be expected at an intermediate time. It is therefore likely that
the increasing trend of retained austenite content with lower bainite forma-
tion temperatures arises from the larger amount of untransformed austenite
due to the slower bainite reaction kinetics.
As the bainite transformation progresses, the residual austenite gets
enriched with carbon. This imparts stability against subsequent marten-
sitic transformation, thus increasing the retained austenite content (Fig-
ure 4.8(d)). On the other hand, the amount of untransformed austenite de-
creases with increasing time. It follows that the amount of retained austenite
will be maximum at an intermediate time depending on transformation tem-
perature [Jacques et al., 2001b; Nakagaito et al., 2003]. Figure 4.9 illustrates
this eect in a 0.2C-1.5Mn-1.5Si (wt%) steel with intercritical annealing pa-
rameters 780
C and 300 s. Retained austenite fraction maximises at an

intermediate bainite reaction time for both the transformation temperatures
of 350
C and 375
C.
85 4.3 Model for carbon in retained austenite
2
3
4
5
6
7
8
9
10
11
12
0 300 600 900
r
e
t
a
i
n
e
d

a
u
s
t
e
n
i
t
e

/

v
o
l
.
%
(a)
2
3
4
5
6
7
8
9
10
11
12
0 300 600 900
r
e
t
a
i
n
e
d

a
u
s
t
e
n
i
t
e

/

v
o
l
.
%
(b)
Figure 4.9: Variation in retained austenite volume fraction with bainite re-
action time at transformation temperature of (a) 350
C and (b) 375
C.
4.3 Model for carbon in retained austenite
4.3.1 Database
Data used for the model made for predicting the carbon content of retained
austenite are presented in Table 4.2
2
. Figure 4.10 presents the distribution
of each of the input variables against the output.
4.3.2 Model characteristics
As expected, increasing the number of hidden units reduces the noise level
(Figure 4.11(a)). However, the test error does not tend to decrease contin-
uously with increasing model complexity (Figure 4.11(b)). The optimum
number of models in the committee was found to be seven, when the test
error is minimum. The model appears to perform reasonably well, during
2
Data used are due to De Meyer et al. [1999a]; Itami et al. [1995]; Jacques et al.
[2001a,b,c, 1999]; Kim et al. [2002]; Matsumura et al. [1992, 1987b]; Nakagaito et al.
[2003]; Wang et al. [2004]; Zaeerer et al. [2004]
Inputs Min. Max. Average S.D.
Carbon / wt% 0.1100 0.3920 0.2676 0.1150
Manganese / wt% 0.8300 1.6600 1.3019 0.1813
Silicon / wt% 0.3400 1.9900 1.2150 0.5284
Aluminium / wt% 0.0000 1.5000 0.1879 0.3982
Phosphorus / wt% 0.0000 0.0400 0.0025 0.0058
Copper / wt% 0.0000 0.5100 0.0051 0.0506
IA temperature /

C 730 830 773.2778 25.9773
IA time / s 60 360 286.9697 50.8797
IT temperature /

C 350 450 392.6263 28.4305
IT time / s 0 9725 578.7222 991.3905
Output Min. Max. Average S.D.
Carbon in retained austenite / wt% 0.49 1.8793 1.1949 0.3836
Table 4.2: Data used to model the carbon content of retained austenite, IA:
intercritical annealing and IT: isothermal transformation to bainite, S.D.:
standard deviation.
both the training and testing step (Figure 4.12(a)-(b)).
Figure 4.13 shows the model-perceived signicance of each of the input
parameters. Bainite reaction time appears to exert the strongest inuence
on the carbon content of retained austenite.
4.3.3 Model predictions
Predictions were made varying each input parameter over a base composition
of 0.2C-1.5Mn-1.0Si-0.5Al-0.02P-0Cu (wt%); intercritical annealing parame-
ters of 780
C, 300 s and bainite transformation conditions of 400
C, 500 s.
Figures 4.14(a)-(d) demonstrate that the results for carbon, aluminium and
silicon are uncertain while manganese shows an inverse trend. Manganese
shifts the T
0
curve to lower carbon contents and hence reduces the carbon
content of retained austenite.
0.1
0.2
0.3
0.4
0.4 0.8 1.2 1.6 2
c
a
r
b
o
n

/

w
t
%
carbon in ret. aust. / wt%
(a)
0.9
1.2
1.5
1.8
0.4 0.8 1.2 1.6 2
m
a
n
g
a
n
e
s
e

/

w
t
%
(b)
0.3
0.9
1.5
2.1
0.4 0.8 1.2 1.6 2
s
i
l
i
c
o
n

/

w
t
%
(c)
0
0.4
0.8
1.2
1.6
0.4 0.8 1.2 1.6 2
a
l
u
m
i
n
i
u
m

/

w
t
%
(d)
0
0.02
0.04
0.4 0.8 1.2 1.6 2
p
h
o
s
p
h
o
r
o
u
s

/

w
t
%
(e)
0
0.25
0.5
0.4 0.8 1.2 1.6 2
c
o
p
p
e
r

/

w
t
%
(f)
720
760
800
840
0.4 0.8 1.2 1.6 2
I
A

t
e
m
p
e
r
a
t
u
r
e

/

o
C
(g)
0
100
200
300
400
0.4 0.8 1.2 1.6 2
I
A

t
i
m
e

/

s
(h)
350
400
450
0.4 0.8 1.2 1.6 2
I
T

t
e
m
p
e
r
a
t
u
r
e

/

o
C
(i)
0
2500
5000
7500
10000
0.4 0.8 1.2 1.6 2
I
T

t
i
m
e

/

s
(j)
Figure 4.10: Distribution of data in the input space for the carbon content
of retained austenite model.
N
o
i
s
e
0.04
0.08
0.12
0.16
0 4 8 12 16 20
(a)
0.4
0.8
1.2
1.6
0 4 8 12 16 20
Number of models
C
o
m
b
i
n
e
d

T
e
s
t

E
r
r
o
r
(b)
Figure 4.11: (a) Variation of noise () with increasing number of hidden
units. (b) Variation of test error with number of models in the committee.
Carbon may also be expected to have a similar eect as manganese.
But this is apparently missing in Figure 4.14(a). The concentration of car-
0.6
0.4
0.2
0
0.2
0.4
0.6
0.6 0.4 0.2 0 0.2 0.4 0.6
P
r
e
d
i
c
t
e
d

v
a
l
u
e
s
,

n
o
r
m
a
l
i
s
e
d
(a)
0.6
0.4
0.2
0
0.2
0.4
0.6
0.6 0.4 0.2 0 0.2 0.4 0.6
P
r
e
d
i
c
t
e
d

v
a
l
u
e
s
,

n
o
r
m
a
l
i
s
e
d
(b)
Figure 4.12: Model predictions compared with the actual database values
during (a) training and (b) test.
bon in austenite before bainite formation is not the same as that of the
bulk steel composition. During intercritical annealing, carbon in austenite
Figure 4.13: Signicances of the input parameters (A-J) to the models for
predicting carbon content of retained austenite, A: carbon, B: manganese,
C: silicon, D: aluminium, E: phosphorous, F: copper, G: intercritical anneal-
ing temperature, H: intercritical annealing time, I: bainite transformation
temperature, J: bainite transformation time.
increases due to ferrite formation. However, manganese in the austenite
formed during intercritical annealing is most likely to remain same as that
of the bulk composition. This is because partitioning of manganese requires
a longer time than partitioning of carbon. It seems that carbon enrichment
of retained austenite is aected by the bulk manganese concentration but
not bulk carbon content of the steel. It may therefore be inferred that car-
bon enrichment in retained austenite is aected by the factors that can alter
bainite reaction kinetics. This is also apparent from the eect of the heat
treatment parameters on the carbon content of retained austenite, as illus-
1
1.25
1.5
1.75
2
2.25
2.5
2.75
0.1 0.2 0.3 0.4 0.5
c
a
r
b
o
n

/

w
t
%
carbon / wt%
(a)
1
1.4
1.8
2.2
2.6
0 0.5 1 1.5 2 2.5
c
a
r
b
o
n

/

w
t
%
silicon / wt%
(b)
1.2
1.4
1.6
1.8
2
2.2
2.4
0 0.5 1 1.5 2
c
a
r
b
o
n

/

w
t
%
aluminium / wt%
(c)
0.5
1
1.5
2
2.5
3
3.5
0 0.5 1 1.5 2 2.5
c
a
r
b
o
n

/

w
t
%
manganese / wt%
(d)
Figure 4.14: Eect of alloying elements on the carbon content of retained
austenite: (a) carbon, (b) silicon, (c) aluminium and (d) manganese.
trated in Figure 4.15.
Austenite formed at a higher intercritical annealing temperatures is
leaner in carbon. This can accelerate the subsequent transformation to bai-
nite, resulting in rapid carbon enrichment of the residual austenite (Fig-
ure 4.15(a)). Similarly, with a longer holding period during the intercritical
annealing, austenite gets enriched more with carbon and manganese, hence
gaining stability. Subsequent transformation of austenite to bainite becomes
sluggish, which explains the lower carbon enrichment at longer holding times
of intercritical annealing (Figure 4.15(b)).
0.7
1.1
1.5
1.9
2.3
2.7
3.1
720 760 800 840 880
c
a
r
b
o
n

/

w
t
%
intercritical annealing temperature /
o
C
(a)
0.5
1
1.5
2
2.5
100 150 200 250 300 350
c
a
r
b
o
n

/

w
t
%
intercritical annealing time / s
(b)
0.2
0.6
1
1.4
1.8
2.2
300 350 400 450 500
c
a
r
b
o
n

/

w
t
%
bainite reaction temperature /
o
C
(c)
0.4
0.8
1.2
1.6
2
2.4
2.8
0 500 1000 1500 2000 2500
c
a
r
b
o
n

/

w
t
%
(d)
Figure 4.15: Eect of heat treatment conditions on the carbon content of
retained austenite: intercritical annealing (a) temperature and (b) time; bai-
nite transformation (c) temperature and (d) time.
Bainite transformation at a lower temperature should result in more
carbon content of the retained austenite (as expected from T
0
criterion).
This is observed up to a particular temperature, below which the carbon
content appears to fall, as can be seen in Figure 4.15(c). At low tempera-
ture, transformation kinetics can be slow. Therefore the progress of bainite
formation may not be sucient to enrich the retained austenite with car-
bon. With an increase in bainite reaction time, carbon in retained austenite
increases initially (Figure 4.15(d)). However, prolonged holding times may
93 4.4 Summary
result in some marginal loss of carbon, possibly due to cementite precipita-
tion. Predictions for longer holding times are however associated with large
uncertainties, limiting condence in the predicted behaviour. More experi-
ments are necessary in this region to conrm the predicted behaviour and its
cause.
Figure 4.16 depicts the variation of carbon in retained austenite with
progress of bainite reaction at two dierent temperatures for 0.2C-1.5Mn
(wt%) steel made with either 1.5 wt% silicon or aluminium. Carbon in re-
tained austenite appears to attain a maximum value at an intermediate time
for both the steels at both temperatures. The maximum carbon content is
higher at the lower transformation temperature for each steel. Furthermore,
the maximum carbon content is higher for the aluminium-containing steel
than that made with silicon. Both these predictions can be accounted for by
the T
0
criterion of bainite transformation theory and can therefore be judged
to be reasonable.
4.4 Summary
Neural network techniques have been adopted to create models for predict-
ing the retained austenite fraction and its carbon content as a function of
processing parameters and composition of TRIP-assisted steels.
Elements that tend to stabilise austenite over ferrite appear to increase
the nal retained austenite content. Retention of austenite is due to the car-
bon enrichment of the residual austenite after bainite formation. Therefore
the rate of bainite formation seems to have a strong inuence on retaining
higher austenite fraction. Any factor that alters the rate of bainite transfor-
94 4.4 Summary
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
0 300 600 900
c
a
r
b
o
n

c
o
n
t
e
n
t

/

w
t
%
(a)
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
0 300 600 900
c
a
r
b
o
n

c
o
n
t
e
n
t

/

w
t
%
(b)
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
0 300 600 900
c
a
r
b
o
n

c
o
n
t
e
n
t

/

w
t
%
(c)
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
0 300 600 900
c
a
r
b
o
n

c
o
n
t
e
n
t

/

w
t
%
(d)
Figure 4.16: Variation in carbon content of retained austenite with bainite re-
action time for 0.2C-1.5Mn-1.5Si (wt%) steel at transformation temperature
of (a) 375
C and (b) 425
C; for 0.2C-1.5Mn-1.5Al (wt%) steel at transfor-

mation temperature of (c) 375
C and (d) 425
C.
mation aects retained austenite content in these steels.
The carbon content of retained austenite is primarily controlled by the
extent of bainite transformation. However, the maximum carbon enrichment
of the retained austenite can also be inuenced by altering the T
0
point, with
suitable alloying additions or the bainite formation temperature.
The model for retained austenite fraction was further used to formulate
an optimum combination of silicon and retained austenite contents in these
steels, as explained in the following Chapter.
Chapter 5
-TRIP steel
In TRIP-assisted steels, silicon helps to retain austenite by suppressing the
precipitation of cementite. The carbon partitioned into the austenite, follow-
ing the formation of bainite, remains there and permits it to be retained at
room temperature. However, there is no quantitative theory to estimate the
exact concentration of silicon necessary to prevent cementite. The silicon
can be detrimental to the surface properties of automobile steels. So it is
important to optimise both the silicon and the retained austenite contents.
The work presented in this chapter began with the aim of nding an
optimum combination of silicon and retained austenite but the computa-
tional methods used gave surprising and interesting results. This stimulated
experiments which led to the discovery of a new concept in TRIP-assisted
steels.
To describe the story in a logical manner, the optimisation technique is
explained rst. This exploits the neural network model presented in Chapter
4. The nal part deals with the design and testing of a steel with a radically
dierent microstructure.
95
96 5.1 Optimisation
5.1 Optimisation
Numerical models can be used to optimise the chemical composition and
process parameters of a material in order to achieve the appropriate mi-
crostructure or properties [Murugananth et al., 2004]. This is done by the
repeated interrogation of the numerical model in order to attain a specic
output, in a process called optimisation (Figure 5.1).
The rst step is to identify the target, the design variables i.e., the
input variables of the numerical model that are to be varied and the param-
eters which should not be altered during the calculations. An initial set of
input variables is chosen and is used to evaluate the output parameter. The
dierence between the calculated output and the target is then used as a
tness for purpose parameter. If the tness is not satisfactory, the optimiser
algorithm decides the direction in which the design variables should be var-
ied and the whole process is then repeated. The process terminates once the
required degree of tness is attained.
Optimisation methods are classied as linear or non-linear, based on
the function relating the variables of the model [Nelder and Mead, 1965].
In practice, linear functions are seldom observed. Sequential quadratic pro-
gramming, downhill simplex methods, genetic algorithms represent some of
the non-linear methods of optimisation [Epogy, 2003].
Genetic algorithm
Genetic algorithms carry out the optimisation by mutation or recombination
and selection [Epogy, 2003]. A point is rst chosen in the design space.
A family of points is subsequently generated around the initial one. This
97 5.1 Optimisation
Figure 5.1: Flow diagram of the optimisation process.
is called mutation. By contrast, in the recombination process, values are
exchanged between a random number of points. These are done in such a
98 5.2 Optimised TRIP-assisted steel
way so that the points move in a direction allowing the objective function to
be optimised. The best of all the points is nally selected as the solution.
5.2 Optimised TRIP-assisted steel
A genetic algorithm was used in the present work to formulate an optimum
chemical composition of TRIP-assisted steel, given the objective of maximis-
ing the retained austenite fraction and keeping the silicon addition to a min-
imum. A neural network model was created relating the retained austenite
fraction to the concentrations of dierent alloying elements and the process
parameters of these materials, as described in Chapter 4. The model was
coupled with commercial software, Epogy [2003], to carry out the optimi-
sation process. The coupling and optimisation was carried out at Nanyang
Technological University, Singapore by Dr M. Murugananth.
Concentrations of all the alloying elements, except phosphorous, were
allowed to vary. The intercritical annealing temperature and the isothermal
bainite reaction temperature were also chosen as design variables. However,
the dwell time at these temperatures was kept constant. The uncertainty
in the model prediction was considered as a constraint, with a maximum
permitted value of 20%.
The algorithm converged to an alloy with the chemical composition
Fe-0.4C-2Mn-0.5Si-2Al-0.5Cu-0.02P (wt%), with a microstructure estimated
to contain 4120 vol.% of retained austenite, after intercritical annealing at
840
C followed by bainite transformation at 300
C.
The alloy is rich in carbon, manganese, aluminium and copper. Each of
these elements is known to help retain austenite [Girault et al., 2001; Jacques
99 5.3 Thermodynamic calculations
et al., 2001a; Kim and Lee, 1999; Sakuma et al., 1991a; Shi et al., 2002]. The
concentrations of these elements were therefore increased by the algorithm
while reducing the silicon content, as required by the objective. The result
obtained therefore appears to be physically meaningful.
5.3 Thermodynamic calculations
Amounts of various phases in equilibrium as a function of temperature were
calculated for the optimised alloy using MTDATA with the SGTE database.
Ferrite, austenite, cementite and liquid were allowed to exist. The results are
presented in Figure 5.2. The alloy starts to solidify by forming ferrite, which
reaches a maximum of about 85 wt%. This is called -ferrite. Austenite
starts to form thereafter with the amount of -ferrite decreasing. However,
the calculations indicate that austenite can never exist as a single phase in
the alloy. With cooling, the amount of austenite increases and attains a
maximum of about 80 wt% before it starts to decompose to allotriomorphic
ferrite.
Figure 5.2 also reveals the alloy to exist as a mixture of about 54 wt%
austenite and 46 wt% ferrite at 840
C, the predicted intercritical annealing

temperature for the optimum alloy. The chemical composition of the austen-
ite formed at 840
C was calculated to be Fe-0.67C-2.54Mn-0.53Si-1.35Al-

0.013P-0.51Cu (wt%). Temperatures at which the austenite can be expected
to start transforming into bainite and martensite were calculated using the
MAP STEEL MUCG73 program to be 354
C and 160
C respectively. Both
the predicted intercritical annealing and the bainite reaction temperature of
300
C therefore appear reasonable.

0
20
40
60
80
100
750 1000 1250 1500
a
m
o
u
n
t

o
f

p
h
a
s
e

/

w
t
%
Temperature /
o
C
Liquid
Ferrite
Austenite
Cementite
Figure 5.2: Calculated amounts of dierent phases as a function of temper-
ature for the optimised alloy.
Retention of austenite in these steels is due to its carbon enrichment
after the bainite reaction. Bainitic ferrite grows without any diusion but
excess carbon subsequently partitions to the residual austenite [Bhadeshia,
2001]. The reaction can be sustained till the free energy of the austenite
becomes equal to that of the ferrite. This happens when the mole frac-
tion of carbon in residual austenite becomes x
T 0
. This is calculated by the
MAP STEEL MUCG73 program to be 0.0497 (i.e., 1.115 wt%) at 300
C,
for the austenite formed during intercritical annealing at 840
C.
The initial mole fraction of carbon in a fully austenitic sample, x, can
be related to the nal volume fractions and carbon content of the bainitic
ferrite and austenite as,
x = v
+v
(5.1)
where v
and v
are the volume fractions of bainite and austenite in the

microstructure with mole fractions of carbon x
and x
respectively.
The mole fraction of carbon in the austenite formed at 840
C is cal-
culated to be 0.0312. Substituting x = 0.0312, x
= 0.0497 and assuming

x
= 0 in the above equation, the volume fraction of austenite remaining

untransformed after bainite reaction at 300
C is calculated to be 0.628. As-

suming the densities of bainite and austenite do not vary signicantly, the
calculations suggest the nal weight fraction of retained austenite to be 0.628.
However, in the present case, the initial weight fraction of austenite in the
material before the bainite reaction was only 0.54. Thus the nal retained
austenite weight fraction, and hence also the volume fraction, can be ex-
pected to be 0.339. This austenite is expected to contain 1.115 wt% carbon
and hence to be stable against further decomposition into martensite. The
thermodynamic calculation therefore seems to be in reasonable agreement
with the predictions made by the genetic algorithm.
Figure 5.2 indicates that the optimum alloy can never be fully
austenitic. This could be due to the aluminium which stabilises -ferrite
and forms -loop in the phase diagram. This is not generally desirable as it
prevents the steel from becoming fully austenitic. However, TRIP-assisted
steels in any case contain allotriomorphic ferrite which can, in principle,
be substituted with -ferrite. This denes a new class of steels, designated
-TRIP. Experiments were therefore carried out using alloys based on the
optimum composition with the objective of using the -ferrite to design a
102 5.4 Experiments
new microstructure and to test the properties.
5.4 Experiments
Materials
Based on the optimised composition, four alloys designated A, B, C and D
were made in laboratory with chemical compositions as shown in Table 5.1.
Alloy A was prepared in a vacuum melting furnace and in a small quantity
of some 50 g. This was made simply to establish proof of concept. The other
three alloys were made as 50 kg ingots in air-melting furnace. The three larger
melts were made because it was dicult to get the desired composition.
Alloy C Mn Si Al P Cu
A 0.41 2.42 0.46 1.57 0.26 0.42
B 0.36 1.96 0.73 2.22 0.022 0.52
C 0.42 1.91 0.62 2.30 0.022 0.53
D 0.43 1.90 0.63 2.16 0.022 0.52
Table 5.1: Chemical compositions of the alloys made for experiments.
Heat treatment
Samples of alloys A and B in the as-cast condition were sealed in quartz tubes
containing a partial pressure of argon for heat treatment in an air furnace.
One set of the alloy A samples was tempered at 300
C for 1 h, 2 h and 4 h
respectively. Samples of the alloys A and B were also subjected to two-step
heat treatment, involving initial heating to a temperature in the (+) phase
region followed by an isothermal dwell at a lower temperature, in order to
103 5.4 Experiments
obtain bainite. Details about the heat treatment parameters are described
in the following section.
Characterisation procedures
Microstructures of the samples in both as-cast and heat-treated conditions
were characterised using optical and eld emission gun scanning electron mi-
croscopy (FEGSEM). Samples were polished using standard metallographic
techniques and etched with 2% nital solution. The electron microscope was
operated at a 10 kV accelerating voltage.
X-ray diraction tests of the samples were carried out with a scan range
of 40
-120
at 0.5
intervals and a 5 s dwell at each step. Cu-K
radiation was
used and the operating voltage and current were set at 40 kV and 40 mA
respectively. Samples were polished, etched with 2% nital and then once
again just ne polished before nally putting into the diractometer.
Mechanical tests
The bulk hardnesses and the microhardness of the constituent phases were
recorded with 10 kg and 50 g load respectively.
Uniaxial tensile tests were carried out with the heat-treated samples
of the alloy B, at room temperature and 100
C. Specimens were machined

from the heat-treated blocks with circular cross-section of 5 mm diameter
and 30 mm gauge length. The samples were deformed with a strain rate
of 2.810
5
s
1
. An extensometer of 10 mm gauge length was attached to
the specimen tested at room temperature. However, this could not be used
for the high temperature test. Samples cut from near the fracture surfaces
were used for X-ray diraction studies in order to measure the fraction of
austenite remaining at fracture.
Alloy A
Figure 5.3 presents the equilibrium phase fractions as a function of temper-
ature for alloy A, as calculated using MTDATA with the SGTE database.
This predicts signicant quantities of -ferrite to be retained during solidi-
cation of the alloy. The microstructures of the alloy in the as-cast condition
are presented in Figure 5.4, showing light-etching elongated grains of ferrite.
The morphology is that of dendrites of -ferrite formed directly from the liq-
uid during solidication. The micrographs also reveal a dark-etching residual
phase, the nature of which is not clear from Figure 5.4. FEGSEM images,
presented in Figure 5.5, indicate this residual phase to be martensitic.
The microhardnesses of -ferrite and the residual phase in the as-cast
microstructure and the bulk hardness values are presented in Table 5.2. The
high hardness of the residual phase is consistent with that of the high-carbon
martensite [Bhadeshia and Honeycombe, 2006].
Phase Hardness / HV
-ferrite 29620
Residual phase 80157
Bulk 41225
Table 5.2: Microhardness of the constituent phases and the bulk hardness of
the as-cast sample. The values represent the average of 20 readings in each
case and the error bar represents one standard deviation.
The hardness data of Table 5.2 were used to estimate the volume frac-
tions of the constituent phases in the as-cast sample using a rule of mixtures.
Assuming the microhardnesses of the -ferrite and the residual martensitic
0
20
40
60
80
100
750 1000 1250 1500
a
m
o
u
n
t

o
f

p
h
a
s
e

/

w
t
%
Temperature /
o
C
Liquid
Ferrite
Austenite
Cementite
Figure 5.3: Calculated amounts of phases in equilibrium for alloy A.
phase are H
1
and H
2
respectively, the bulk hardness, H
b
, of the sample can
be expressed as:
H
b
= vH
1
+ (1 v)H
2
(5.2)
where, v is the volume fraction of -ferrite. A point count method was also
used to measure the phase fractions. Assuming that the densities of the two
phases are not signicantly dierent, the measurements eectively represent
the weight fraction of each phase, as presented in Table 5.3. Errors reported
in the point count measurements represent 1.
Figure 5.3 indicates a maximum of about 80 wt% -ferrite to form at
1400
C in equilibrium with liquid. Austenite forms on further cooling, lead-

ing to a decrease in the -ferrite content. Table 5.4 shows the calculated
(a)
(b)
Figure 5.4: Optical micrographs of alloy A in the as-cast condition.
phase fractions and the chemical composition of the -ferrite and the liquid
existing in equilibrium at 1400
C. The experimentally determined fraction

of -ferrite in the as-cast microstructure appears comparable with that cal-
(a)
(b)
Figure 5.5: FEGSEM images revealing the martensite-like plates in the resid-
ual phase. Some of the plates contain a midrib which is consistent with
high-carbon martensite.
Phase Weight fraction
from hardness from point count
-ferrite 0.77 0.740.04
Residual phase 0.23 0.260.01
Table 5.3: Weight fractions of the constituent phases in the as-cast mi-
crostructure.
culated, Table 5.4. It therefore appears that the alloy solidied by forming
-ferrite upto a temperature of around 1400
C.
Phase Amount / wt% Composition / wt%
C Mn Si Al P Cu
Ferrite 79.5 0.2 2.1 0.4 1.7 0.2 0.4
Liquid 20.5 1.2 3.7 0.7 1 0.2 0.4
Table 5.4: Calculated amounts and the chemical composition of the phases
in equilibrium for the alloy A at 1400
C.
The liquid coexisting with the -ferrite at 1400
C is rich in carbon (Ta-

ble 5.4). The austenite that will form from the liquid, and hence the marten-
site, can be expected to inherit the high carbon content. This supports the
high hardness of the residual martensitic phase in the as-cast microstructure
(Table 5.2). However, there must also be some austenite remaining untrans-
formed, due to its high carbon content. X-ray diraction was used to detect
the presence of austenite and the analysis was carried out with the help of
XPert Plus software which uses the Reitveld renement method due to Wiles
and Young [1981]. This gave about 7.80.7 wt% of retained austenite to be
present in the sample.
A set of as-cast samples of the alloy was tempered at 300
C for varying
time periods, mainly to conrm the high carbon content of the martensitic
residual phase. The microhardness of the phase was found to decrease drasti-
cally with increase in tempering time, as illustrated in Table 5.5. The average
values of 20 readings with the error as one standard deviation are reported
in Table 5.5.
Figure 5.6 shows the FEGSEM images obtained from a tempered sam-
ple. Compared to the as-cast condition (Figure 5.5), the residual phase after
tempering appears with better contrast (Figure 5.6). This might be due to
carbide precipitation from the martensite during tempering, increasing the
interfacial area per unit volume thereby resulting in more intense etching.
Tempering time / h Microhardness / HV
0 80157
1 65523
2 60630
4 52936
Table 5.5: Drop in microhardness of the residual martensitic phase with
tempering time at 300
C.
The two-step heat-treatment of the as-cast samples were carried out to
produce a microstructure comprised of -ferrite, bainite and austenite. Cal-
culations using MTDATA with SGTE database revealed alloy A to consist of
60 wt% austenite in equilibrium with 40 wt% ferrite at 810
C (Figure 5.3).
Samples of the alloy were therefore heated at 810
C for 12 h followed by a sec-

ond isothermal hold at 300
C for 6 h. The MAP STEEL MUCG73 program

was used to determine the time and temperature of the bainite reaction. The
chemical composition of the austenite formed at 810
C was calculated using

MTDATA and used as the input to the MAP STEEL MUCG73 program.
Figure 5.7 reveals elongated grains of -ferrite and a dark-etching sec-
(a)
(b)
Figure 5.6: FEGSEM images of the sample tempered at 300
C for 4 h.
ond phase, the nature of which is not clear. The FEGSEM images of the
sample are presented in Figure 5.8. It appears that the second phase is made
up of bainite and retained austenite. The presence of about 14.30.7 wt%
retained austenite is revealed by X-ray diraction analysis. This is, how-
ever, seen to be less than both the amount of retained austenite predicted by
the genetic algorithm and that calculated using the bainite transformation
theory. The carbon content of the retained austenite was estimated to be
1.350.02 wt% from the lattice parameter of austenite, using the equation
proposed by Dyson and Holmes [1979].
(a)
(b)
Figure 5.7: Optical micrographs of alloy A heat-treated to form bainite.
(a)
(b)
Figure 5.8: FEGSEM images of the heat-treated sample of alloy A revealing
bainite.
Alloy B
As expected from Figure 5.9, the microstructure of the alloy in the as-cast
condition reveals the presence of -ferrite, embedded in a dark-etching resid-
ual phase (Figure 5.10(a)). The microhardness of the residual phase in this
alloy was recorded as 38222 HV and that of the -ferrite as 2148 HV.
Unlike alloy A, the residual phase of this material does not therefore appear
to be martensitic. Figure 5.10(b) reveals the residual phase to be pearlite.
0
20
40
60
80
100
750 1000 1250 1500
a
m
o
u
n
t

o
f

p
h
a
s
e

/

w
t
%
Temperature /
o
C
Liquid
Ferrite
Austenite
Cementite
Figure 5.9: Amounts of phases in equilibrium at dierent temperatures for
the alloy B, calculated using MTDATA with SGTE database.
Samples of alloy B were subjected to the two-step heat-treatment. Fig-
ure 5.9 reveals the alloy to exist as a mixture of 45 wt% ferrite and 55 wt%
austenite in equilibrium at 940
C. Samples were therefore initially heated at

(a)
(b)
Figure 5.10: Microstructures of the as-cast sample of alloy B, (a) optical
micrograph and (b) FEGSEM image.
940
C for 12 h and were subsequently allowed to transform into bainite at

300
C for 3 h.
Figure 5.11 depicts the microstructures of the heat-treated sample. The
distinct morphology of the -ferrite is seen to have been retained after the
heat-treatment, along with the formation of bainite. The presence of retained
austenite in the heat-treated sample was conrmed with X-ray diraction
analysis, which revealed 13.50.9 wt% of retained austenite containing about
1.170.03 wt% carbon.
The microhardnesses of -ferrite and bainite along with the bulk hard-
ness of the heat-treated sample are presented in Table 5.6. Hardness data
and point counts were used to estimate the fractions of -ferrite and bainite,
as described previously. Assuming that the densities of -ferrite and the
bainitic phase are approximately equal, the measurements eectively rep-
resent the weight fraction of each phase, as presented in Table 5.7. The
fraction of bainite is seen to be in reasonable agreement with that of the
austenite present at the intercritical annealing temperature, calculated using
MTDATA (Figure 5.9).
Phase Hardness / HV
-ferrite 2679
Bainite 51913
Bulk 38540
Table 5.6: Microhardnesses of the constituent phases and the bulk hardness
of the heat-treated sample of alloy B.
Heat-treated samples of the alloy B were mechanically tested using uni-
axial loading. The properties observed at room temperature are impressive
with 23% total elongation and an ultimate tensile strength of 1000 MPa (Fig-
(a)
(b)
Figure 5.11: Microstructures of the heat-treated sample of alloy B, (a) optical
micrograph and (b) FEGSEM image.
Phase Weight fraction
from hardness from point count
-ferrite 0.53 0.430.02
Bainite 0.47 0.570.03
Table 5.7: Weight fractions of -ferrite and bainite phases in the heat-treated
sample of alloy B.
ure 5.12). The 0.2% proof stress was recorded as 700 MPa. There has been
no signicant necking of the sample at fracture suggesting that the elongation
has mostly been uniform (Figure 5.13).
The true stress-strain plot in Figure 5.12 shows steady strain-hardening
continuing until fracture. This could be due to the transformation of re-
tained austenite into martensite during straining. The X-ray analysis of the
sample cut from near the fracture surface of the tensile specimen indicated
7.00.8 wt% of austenite to exist. During straining, the austenite content
was therefore reduced to half of that in the undeformed condition. The trans-
formation of austenite into martensite during straining might have played a
key role inuencing the properties.
Figure 5.14 depicts the stress-strain curve recorded for the sample
tested at 100
C. The more gradual slope of the elastic part of the curve

as compared to the room-temperature tested sample is simply due to the
absence of any extensometer. Compared to the properties at room temper-
ature, the strength of the material remained unaltered but total elongation
was reduced signicantly when tested at 100
C. The true stress-strain plot

in Figure 5.14 indicates that the strain-hardening has not been very eective
unlike the specimen tested at room temperature. This might be due to the
enhanced stability of austenite at the higher test temperature. The X-ray
0
400
800
1200
0 4 8 12 16 20 24
0 0.05 0.1 0.15 0.2
S
t
r
e
s
s

/

M
P
a
Engineering strain / %
True strain
Engineering
True
Figure 5.12: Stress-strain curve of the heat-treated material at room tem-
perature.
Figure 5.13: No necking is observed implying the elongation is mostly uni-
form.
analysis conrmed about 13.00.7 wt% retained austenite to be present in
the specimen after fracture. This is about the same as that of the heat-
treated state, suggesting that there has been no transformation of austenite
during straining.
0
400
800
1200
0 4 8 12 16 20 24
0 0.05 0.1 0.15 0.2
S
t
r
e
s
s

/

M
P
a
Engineering strain / %
True strain
Engineering
True
Figure 5.14: Stress-strain curve of the heat-treated material at 100
C.
121 5.6 Summary
Alloys C and D
Figure 5.15 indicates both the alloys C and D can be expected to have a sim-
ilar phase transformation behaviour as that of the alloys A and B. However,
it is surprising to observe that none of these materials revealed the presence
of -ferrite in the as-cast condition. Figure 5.16 depicts the light-etching
allotriomorphic ferrite formed along the prior austenite grain boundaries in
alloys C and D. The morphology of the ferrite grains in these materials is in
sharp contrast to that of the elongated -ferrite grains, as observed in alloys
A and B.
The absence of -ferrite in as-cast microstructures of alloys C and D
could be due to the cooling pattern during solidication. Figure 5.15 suggests
that -ferrite forms during solidication up to around 1400
C. At lower tem-
peratures, austenite starts gaining stability over ferrite. A fast cooling rate
during solidication can bring down the temperature rapidly below 1400
C,
thereby avoiding the chance of forming -ferrite. This might have led to the
solidication of the alloys C and D forming austenite, which has transformed
partly to allotriomorphic ferrite during subsequent cooling.
5.6 Summary
TRIP-assisted steels are made with silicon, which helps retaining austenite.
However, silicon leads to many diculties during processing and deteriorates
the surface quality of the steel. A genetic algorithm coupled with a neural
network model yielded an optimised chemical composition which is low in
silicon content but rich in aluminium, manganese and copper.
122 5.6 Summary
Experiments to conrm the calculations gave surprising results. Some
of the alloys demonstrated an enhanced tendency of -ferrite formation during
solidication, as expected from phase diagram calculations. However, there
was no evidence of the presence of -ferrite in the as-cast microstructures of
some other alloys, which could be due to the dierence in cooling pattern.
The as-cast samples of the alloys containing -ferrite were heat-treated
to form microstructures containing -ferrite, bainite and retained austen-
ite. The heat-treated samples were found to possess impressive mechanical
properties at room temperature. However, increased stability of the retained
austenite at higher temperature might have been responsible for impairing
the ductility of the specimen tested at 100
C.
123 5.6 Summary
0
20
40
60
80
100
750 1000 1250 1500
a
m
o
u
n
t

o
f

p
h
a
s
e

/

w
t
%
Temperature /
o
C
Liquid
Ferrite
Austenite
Cementite
(a)
0
20
40
60
80
100
750 1000 1250 1500
a
m
o
u
n
t

o
f

p
h
a
s
e

/

w
t
%
Temperature /
o
C
Liquid
Ferrite
Austenite
Cementite
(b)
Figure 5.15: Amounts of phases in equilibrium at dierent temperatures
for the (a) alloy C and (b) alloy D, calculated using MTDATA with SGTE
database.
124 5.6 Summary
(a)
(b)
Figure 5.16: Optical micrographs from the as-cast alloys C and D showing
allotriomorphic ferrite along the prior austenite grain boundary.
Chapter 6
Mechanical Stabilisation
Austenite can be induced to transform to martensite above the M
S
tem-
perature by stress or strain. This induced transition is widely believed to
be the cause of the excellent mechanical properties, especially the formabil-
ity, of TRIP-assisted steels. Stress adds to the thermodynamic driving force
whereas plastic strain aects the kinetics of the process. Defects created in
austenite due to strain provide additional nucleation sites. However, growth
of martensite plates can be hindered by defects in austenite and excessive
deformation can even halt the transformation, in a phenomenon known as
mechanical stabilisation. The purpose of the present chapter is to study the
role of stress and strain on martensitic transformations.
6.1 Mechanical driving force
Martensitic transformation during deformation of austenite was rst discov-
ered by Scheil [1932], the thermodynamics of which was explained by Patel
and Cohen [1953]. Stress makes transformation possible above the M
S
tem-
125
126 6.1 Mechanical driving force
perature by supplying the shortfall of the necessary thermodynamic driving
force. This is illustrated schematically in Figure 6.1. The free energy change,
G
1
, available at the M
S
temperature is the critical driving force necessary
for transformation. At a temperature T
1
above M
S
, the magnitude of the free
energy change due to transformation, G
2
, is less than the critical driving
force, G
1
. Martensite can only be formed at this temperature by supply-
ing additional energy of a magnitude U, so that U + G
2
= G
1
. This
can be done by applying stress. The work done by external stress adds to
the available chemical free energy change as a mechanical driving force for
transformation at any temperature.
Figure 6.1: A shortfall in driving force for martensitic transformation above
the M
S
temperature can be compensated for by the application of stress.
Transformation of austenite to martensite is accomplished by a perma-
nent shape deformation. This is characterised by an invariant-plane strain
deformation with and s as the dilatational and shear components of the
shape deformation respectively. The mechanical driving force due to an ex-
127 6.2 Role of strain
ternal stress, , can then be expressed as:
G
mechanical
=
N
+s (6.1)
where
N
is the normal component and is the shear component of the
applied stress resolved along the direction of the shear displacement of the
shape deformation. Both
N
and act on the crystallographic plane of
austenite on which the plate of martensite forms (Figure 6.2).
Work done by the shear component of the applied stress is always ex-
pected to be positive whereas that due to the normal component may be pos-
itive (in tension) or negative (compression). In steels, the shear component
of the shape deformation is much higher in magnitude than the dilatational
component. Thus the observed eect of stress arises mainly from the work
done by the shear component. However, hydrostatic stress interacts only
with the dilatational component and therefore is not very eective for in-
ducing transformation in steels. Hydrostatic compression, in fact, suppresses
transformation since in steels, there is a volume expansion on martensitic
transformation.
6.2 Role of strain
As is apparent from Figure 6.1, the higher the temperature above M
S
, the
greater the stress required to initiate transformation. It is possible therefore
that above a critical temperature austenite will be plastically deformed before
transformation, which is often referred as strain-induced transformation.
Plastic strain in austenite creates defects that can act as additional
128 6.2 Role of strain
Figure 6.2: Resolved normal and shear components of the external
stress [Bhadeshia, 2001].
nucleation sites for transformation. This is in contrast to the stress-assisted
transformation that starts from the same nucleation sites as athermal trans-
formation but with the thermodynamic assistance of external stress. Plastic
strain thus renders a kinetic advantage over the thermodynamic contribu-
tion of the accompanying stress [Lecroisey and Pineau, 1972; Manganon and
Thomas, 1970; Olson and Cohen, 1972; Venables, 1964].
Transformation from plastically strained austenite has been generally
observed to occur at a stress level lower than that expected from the contri-
bution of stress alone. There is a dierence in opinion regarding the cause of
such an observation:
Olson and Cohen [1979] ascribed this behaviour to the higher potency
of the nucleation sites created by plastic strain.
Onodera and Tamura [1979] explained the same behaviour in terms of
the stress concentration at defects, which may reach the critical stress
required for transformation, even when the magnitude of the applied
stress is less.
129 6.3 Mathematical formulation
In spite of the nucleation advantage, plastic strain in austenite may
actually act to retard transformation. Defects generated due to strain can
hinder the growth of martensite plates. This is generally explained in terms
of the martensite-austenite interface structure, which is made up of a set of
screw dislocations, as depicted schematically in Figure 6.3 [Christian, 2002].
Growth of a martensite plate involves translation of the glissile inter-
face, which can be made sessile by obstacles like grain boundaries. Plates of
martensite are therefore never observed to traverse austenite grain bound-
aries. Smaller defects like isolated dislocations resist the motion of the inter-
face but can often be incorporated into the lattice of martensite. However,
heavy dislocation debris in austenite generated by severe plastic deforma-
tion can halt the transformation. This is known as mechanical stabilisa-
tion [Bhadeshia, 1999b; Breedis and Robertson, 1963; Leslie and Miller, 1964;
Machlin and Cohen, 1951; Raghavan, 1992].
6.3 Mathematical formulation
The purpose here is to developed a theory to predict the critical strain nec-
essary for causing mechanical stabilisation of austenite. The problem is one
of balancing the force required to sustain the motion of the transformation
interface against the resistance due to the dislocation debris in austenite.
Dislocation theory
At low homologous temperatures, plastic deformation is caused by the trans-
lation of dislocations. The resultant plastic strain is given by [Honeycombe,
Figure 6.3: Model of austenite-martensite interface (after Frank [1953]), a
screw dislocation with a lattice Burgers vector is necessary at every sixth
atomic closed-packed plane to match the atomic rearrangements [Christian,
2002].
1984]:
= bL (6.2)
where is the dislocation density, b is the Burgers vector and L is the average
distance moved by the dislocations.
For a given dislocation density, the spacing between the dislocations
is given by l =
1
2
. Under these circumstances, the mean shear stress
required to move dislocations past each other is given as [Honeycombe,
1984]:
=
Gb
8(1 )l
=
Gb
1
2
8(1 )
(6.3)
where G is the shear modulus and the Poissons ratio. On combining with
equation 6.2 and noting that the force per unit length is b,
b =
Gb
3
2
8(1 )

L
(6.4)
The mean free distance L must decrease as the plastic strain increases
[Barlat et al., 2002]:
L =
D
+D
(6.5)
where D is the original grain size of austenite prior to straining and is a
coecient about equal to 1 m [Barlat et al., 2002].
Besides dislocation strengthening, solid-solution hardening adds to the
total resistance to interface motion [Olson and Cohen, 1976a-c; Ghosh and
Olson, 2001]. Solid solution hardening coecients for various solutes in
austenite have been reported for tensile strength by Irvine et al. [1969].
These values can be converted into shear stress equivalents
S
using the
Tresca criterion.
Chemical driving force
The stress
T
driving the motion of the interface originates from the chemical
free energy change G of transformation [Olson and Cohen, 1976a-c]:
T
= G (6.6)
where is a constant assumed to be equal to unity and G = G
i.e.,
the magnitude of the driving force which is a function of temperature and
the chemical composition of the alloy.
Lattice distortions due to the accommodation of the shape deforma-
tion accompanying transformation must be accounted for in terms of stored
energy. The magnitude of this is given by Christian [1979] to be about
600 J mol
1
. This has to be subtracted from the free energy change of trans-
formation (obtained using MTDATA), in order to determine the stress driv-
ing the interface motion. The magnitude of force per unit length available
for driving the austenite-martensite interface can be obtained by multiplying
T
with the Burgers vector b.
Model
Mechanical stabilisation occurs when the force driving the interface equals
that opposing its motion:
T
= +
S
(6.7)
bG =
1
8(1 )
Gb
3
2

L
+
S
b (6.8)
This equation can now be used to calculate the critical strain for mechanical
stabilisation.
The key input to the model for calculating the critical strain necessary for
mechanical stabilisation is the free energy change accompanying the transfor-
mation of austenite to martensite, G. The free energy change is a function
of temperature and the composition of the steel and can be calculated using
the software MTDATA. The software accesses the thermodynamic database
SGTE to calculate phase stabilities and free energies. The free energy values
obtained from MTDATA were reduced by 600 J mol
1
to account for the
stored energy of martensite. The magnitudes of dierent constants used in
calculations are listed in Table 6.1.
Property Value
Shear modulus of austenite 8 10
10
Pa
Poissons ratio 0.27
Burgers vector 2.52 10
10
m
Table 6.1: Values of constants used in the model.
6.4.1 Austenitic stainless steels
A range of chemical compositions of austenitic stainless steels are listed in
Table 6.2. The reason for varying the nickel concentration was to study the
eect of the chemical driving force for transformation on the stabilisation to
martensitic transformation. All the calculations were carried out allowing
for solid-solution strengthening and for a variety of temperatures.
Alloy C Mn Si Cr Ni Mo N
Alloy 1 0.05 1.5 0.75 18.0 6.0 0.0 0.03
Alloy 2 0.05 1.5 0.75 18.0 8.0 0.0 0.03
Alloy 3 0.05 1.5 0.75 18.0 10.0 0.0 0.03
Alloy 4 0.05 1.5 0.75 18.0 12.0 0.0 0.03
316L 0.01 0.53 0.75 17.1 11.9 2.0
Table 6.2: Chemical compositions (wt%) of dierent austenitic stainless
steels.
Figure 6.4 shows the chemical driving force (before the subtraction of
the 600 J mol
1
stored energy) as calculated using MTDATA. It is seen that
these are very large driving forces, especially for 300 K which is normally the
ambient deformation temperature. The original grain size was taken to be
40 m.
The critical strain calculated using the model for each alloy at dierent
temperatures is presented in Figure 6.5. It can be seen that the strain values
are very large, ranging from =2.94.5. The strain values increase with
decreasing nickel concentration i.e., increasing the free energy change, G,
of transformation.
Figure 6.4: Free energy change in dierent alloys as listed in Table 6.2 at
dierent temperatures.
In physical terms, this means that the mean free path L has to be very
small, less than 1 m, as illustrated in Figure 6.6. Huge plastic strains are
required to achieve the necessary defect density, so such a result may at rst
Figure 6.5: Calculated strain to initiate mechanical stabilisation at dierent
temperatures for various alloys.
sight seem unlikely. However, experiments carried out with 316L austenitic
stainless steel deformed to a true strain of = 6.3 conrms that severe
deformation is indeed necessary to mechanically stabilise the austenite [Wang
et al., 2005].
Figure 6.7 shows the curve obtained by tting to experimental data
from the 316L experiments [Wang et al., 2005]. It was demonstrated that
thermodynamically the alloy should be able to transform completely into
martensite, were it not for the fact that the deformed austenite becomes
mechanically stabilised. In fact, the amount of strain-induced martensite is
limited to 57%, with much of this formed at low strains before the onset
of stabilisation. On the basis of the tted curve (Figure 6.7), the onset of
mechanical stabilisation, i.e. the strain at which strain-induced martensite
ceases to signicantly increase with strain, is about = 2 3.
Figure 6.6: Variation of the mean free path with strain for mechanical sta-
bilisation, open circles for 400 K and lled circles for 300 K.
Figure 6.7: Experimental tted curve for strain-induced martensite formation
in 316L stainless steel [Wang et al., 2005].
The calculated data for 316L are presented as points in Figure 6.5;
they show that the critical strain for mechanical stabilisation is 3.45, 2.25
and 1.45 for the temperatures 300, 400 and 500 K respectively. The large
strain data of the experiments are based on wire drawing and it is conceivable
that the temperature rises by around 200 K during the deformation [Wang
et al., 2005]. The calculations are therefore consistent with the experimental
data. Furthermore, the analysis shows that large strains needed to achieve
stabilisation is due to the large values of G for transformation in such
stainless steels at the deformation temperature.
6.4.2 TRIP-assisted steels
In TRIP-assisted steels, high-carbon retained austenite transforms to
martensite during deformation. In order to calculate the strain required
to cause mechanical stabilisation in these steels, the chemical composition of
the retained austenite grains must be known precisely.
These steels are intercritically annealed in the (+) two phase region
prior to the bainite reaction. Retained austenite gets enriched with carbon
subsequent to the bainite reaction. The carbon content of retained austenite
is generally reported for these steels [Itami et al., 1995; Jacques et al., 2001c;
Sugimoto et al., 1992b; Tomita and Morioka, 1997]. However, substitutional
solutes do not partition during bainitic transformation. Therefore, the con-
centrations of these elements in the nal retained austenite remain the same
as that before the bainite reaction. The dwell time during intercritical anneal-
ing is very short and so equilibrium partitioning of the substitutional solutes
is very unlikely. By contrast, the carbon concentration can be expected to
attain equilibrium in austenite within the short intercritical annealing period.
The substitutional solute content of the retained austenite can be calculated
using MTDATA, assuming the existence of equilibrium and paraequilibrium
during intercritical annealing separately.
The thermodynamic driving force for transformation of austenite to
martensite was calculated using MTDATA at the temperature of deforma-
tion. Retained austenite grains in these steels are generally reported to be
very ne (about 1 m) [Jacques, 2003]. These values were used as inputs to
the model to determine the critical strain necessary for mechanical stabilisa-
tion (Figure 6.8).
0
1
2
3
4
0.5 0.7 0.9 1.1 1.3 1.5
C
r
i
t
i
c
a
l

s
t
r
a
i
n
Carbon / wt%
Equilibrium
Paraequilibrium
Figure 6.8: Critical strain necessary for onset of mechanical stabilisation in
high-carbon austenite of TRIP-assisted steels.
As seen from Figure 6.8, the critical strain for onset of mechanical sta-
bilisation in these steels decreases with increasing carbon content of retained
austenite. This is due to the lower chemical driving force for martensitic
transformation for higher carbon containing austenite. A range of critical
strain values are predicted for each carbon level: the upper and lower limits
arise from the existence of the equilibrium or paraequilibrium partitioning
of the substitututional solutes during intercritical annealing. However, the
range of predicted strain values is small in magnitude, possibly due to the
subtle dierence between the equilibrium and the paraequilibrium concen-
trations of the substitutional solutes in the austenite.
Figure 6.8 further demonstrates the occurrence of mechanical stabili-
sation at small plastic strain in austenite with carbon content higher than
about 1.1 wt%. It appears therefore that transformation of austenite with
such a carbon level can never be strain-induced. Interestingly, high-carbon
austenite is known to be the hardest phase in the microstructures of these
steels [Furnemont et al., 2002]. Austenite might thus be thought to remain
unstrained till the late stage of the deformation. It might be possible there-
fore that high-carbon austenite in these steels transforms to martensite due
to the eect of elastic stress rather plastic strain.
Stress-assisted transformation
Stress provides the shortfall in the chemical driving force for transformation
to occur. However, the magnitude of the assistance rendered by stress de-
pends on the nature of the stress state. For a simple uniaxial loading, the
stress coecient is reported by Olson [1982] to be 0.86 J mol
1
MPa
1
. With
this value, the total driving force (chemical + mechanical) can be calculated
as a function of temperature using MTDATA.
Figure 6.9 illustrates the total thermodynamic driving force for trans-
formation at dierent stress levels for a hypothetical composition of austenite
as Fe-1C-2Mn-2Si, all in wt%. The gure reveals the remarkable increase in
the total driving force of transformation due to stress. The diagram also
helps to understand the eect of a given stress level in terms of an equivalent
thermal undercooling necessary to achieve the same additional thermody-
namic driving force. For example, application of a stress of 400 MPa may
be considered as an equivalent undercooling of 55 K, resulting in additional
transformation.
-3100
-2800
-2500
-2200
-1900
273 293 313 333 353 373
G

/

J
/
m
o
l
Temperature / K
0 MPa
400 MPa
600 MPa
Figure 6.9: Total driving force for austenite to martensite transformation for
Fe-1C-2Mn-2Si (wt%) austenite at dierent stress levels.
It can therefore be argued that the inuence of stress on the extent of
transformation can be computed in a similar way to athermal transformation,
using the empirical equation given by Koistinen and Marburger [1959]:
1 = exp{C(M
S
T
Q
)} (6.9)
where is the martensite volume fraction at temperature T
Q
and C is a
constant equal to 0.011 K
1
. Knowing the M
S
temperature of austenite and
the equivalent thermal undercooling produced by a given stress level, the
fraction of austenite transforming to martensite can be calculated using this
equation.
Table 6.3 presents data, reported by Jacques et al. [2001d], that have
been used to calculate the inuence of stress on transformation of austenite,
using the principle described above.
Sample Athermal
R
C in
R
M
S
(vol.%) (vol.%) (wt%) (K)
L30s 7 8.8 0.61 351 (calculated)
L60s 5 8.1 0.68 341 (calculated)
H360 0 7.9 0.85 298 (stated)
Table 6.3: Data for decomposition of residual austenite to martensite during
cooling to room temperature,
R
is retained austenite [Jacques et al., 2001d].
The M
S
temperature for sample H360 has been indicated by Jacques
et al. [2001d] to be close to room temperature i.e. 298 K, as only a small
amount of martensite was detected in this sample. On the other hand, the
M
S
temperatures for samples L30s and L60s were calculated using Koistinen-
Marburger equation, based on the reported amount of transformation at
298 K.
Stress levels corresponding to dierent strain values, at which the re-
tained austenite volume fractions were reported, were worked out from the
relevant true stress-strain curve. These values are presented in Table 6.4.
The table also shows the expected retained austenite fractions, as calculated
considering the thermodynamic eect of stress alone.
Sample Stress / MPa Calculated
R
vol.% Reported
R
vol.%
L30s 375 5.4 6.5
463 4.9 4.8
593 4.1 3.6
659 3.8 2.7
794 3.2 1.9
L60s 379 5.0 7.9
466 4.5 5.3
572 3.9 5.3
639 3.6 3.2
698 3.3 2.2
769 3.0 1.6
H360 500 4.0 6.0
710 3.0 4.8
830 2.6 3.6
914 2.3 3.4
Table 6.4: Calculated amounts of retained austenite (
R
) and the correspond-
ing amounts as reported by Jacques et al. [2001d] at each stress level.
Table 6.4 reveals the reported extent of transformation is higher than
that calculated for samples L30s and L60s. There is no contention of mechan-
ical stabilisation at the carbon levels of these samples (Figure 6.8). Plastic
strain in austenite creates defects that act as additional nucleation sites. This
can be considered to drive the transformation to a greater extent than that
expected from the eect of stress alone.
By contrast, transformation has been reported to a lower extent than
the computed ones for sample H360. Austenite in this sample contains about
0.85 wt% carbon. The high carbon level of austenite in H360 sample implies
a high hardness level [Furnemont et al., 2002]. Thus austenite in this sample
might not have strained plastically. It is possible therefore that transforma-
tion has been due to the eect of stress alone.
6.4.3 Bainitic steels
One of the fascinating aspects of the bainite reaction in steels is the sub-unit
mechanism of its growth. A platelet of ferrite grows to a limited size even
though there is no impingement with obstacles such as austenite grain bound-
aries. The transformation then propagates by the nucleation and growth of
another sub-unit, the collection of sub-units being known as a sheaf [Hehe-
mann, 1970]. The reason why each platelet only grows to a limited size is that
the shape deformation accompanying transformation is plastically accommo-
dated in the austenite adjacent to the plate [Bhadeshia and Edmonds, 1979].
This results in the creation of an intense dislocation debris which renders the
interface immobile, as illustrated in Figure 6.10.
Calculations were carried out for a Fe-0.4C-3Mn-2Si (wt%) steel, which
is also the material from which the micrograph in Figure 6.10 originates with
a transformation temperature of 623 K. As pointed out previously, the stored
energy has to be subtracted from the free energy change of transformation.
For bainite, the magnitude of this is given by Bhadeshia [1981] as 400 J mol
1
.
The results of calculations are presented in Table 6.5.
The shear strain associated with the growth of bainite is estimated to
be about 0.26 [Swallow and Bhadeshia, 1996]. This, in comparison with the
calculated data, can be expected to be sucient for onset of local mechanical
stabilisation, and hence a sub-unit mechanism of growth. Furthermore, the
Figure 6.10: Intense dislocation debris at, and in the vicinity of the bainite
(
ub
)-austenite () interface, the steel has a chemical composition Fe-0.43C-
3Mn-2.12Si wt% and is transformed at 623 K [Bhadeshia and Edmonds,
1979].
Temperature / K Critical Strain
500 2.1
600 0.6
650 0.06
Table 6.5: Strain necessary for mechanical stabilisation in the bainitic steel.
length of the sub-unit is expected to become larger and this is qualitatively
consistent with experimental data on bainite, the plates become more slender
as the transformation temperature is reduced (G increased) [Garcia-Mateo
et al., 2003]. It may be further noted that martensite forms in this steel
below about 500 K [Bhadeshia and Edmonds, 1979].
6.4.4 Athermal martensite
Austenite on cooling starts transforming to martensite only below a critical
temperature, M
S
. This temperature varies with chemical composition of the
steel. Most of the common alloying elements, except aluminium and cobalt,
are known to depress M
S
. Finer austenite grain sizes also lead to a lower
M
S
temperature [Raghavan, 1992]. This is generally believed to be due to
the hardening of austenite due to ner grain sizes, hindering the growth of
martensite. Transformation from plastically deformed austenite in a simi-
lar way can be expected to start at lower temperature, as reported in a steel
of composition Fe-0.75C-1.02Mn-0.28Si-0.11Cr-0.05Ni-0.015S-0.009P-0.08Cu
(all in wt%) [Maalekian et al., 2006]. This is becasue of the strengthening
of austenite due to a higher dislocation density than that in the undeformed
condition. A methodology is described in this section to calculate the de-
pression of the M
S
temperature caused by prior plastic strain in austenite.
The theory for mechanical stabilisation relies on balancing the force
required to move the austenite-martensite interface through a forest of dislo-
cations against the force due to the chemical free energy change (G) driving
the interface:
Gb
1
2
8(1 )
= |G| (6.10)
where b = 0.252 nm is the magnitude of the Burgers vector, G = 8 10
10
Pa
is the shear modulus of the austenite, = 0.27 is its Poissons ratio and
is the dislocation density. The extra driving force, (G), necessary for
transformation in a plastically deformed sample can be expressed as:
Gb
8(1 )
[
1
2
1
2
0
] = (G) (6.11)
where the dislocation density has a value
0
at zero plastic strain.
G for transformation of austenite to ferrite without a composition
change was calculated using MTDATA for the Fe-0.75C-1.02Mn-0.28Si-
0.11Cr-0.05Ni-0.015S-0.009P-0.08Cu (wt%) steel as a function of tempera-
ture:
G = 7.0337T 5218 J mol
1
or (G) 7.0337T (6.12)
where T is the absolute temperature. The equation applies over the temper-
ature range 450-510 K. The above expression therefore relates the magnitude
of the change in available driving force due to the presence of dislocations in
the austenite, (G), with the corresponding reduction in the M
S
temper-
ature, T.
It follows that the depression of the transformation temperature can
be calculated as a function of the change in the dislocation density of the
austenite. However, it would be useful to be able to estimate the change
in M
S
as a function of plastic strain rather than dislocation density. It
is necessary therefore to derive a relationship between plastic strain () and
dislocation density, which will necessarily be approximate. So two alternative
methods were attempted as follows:
Data from Vodopivec et al. [1999] are used to obtain an empirical rela-
tionship as:
= 2 10
13

0
+2 10
14
m
2
(6.13)
In the alternative approach, it is assumed that only 5% of the plastic
work is stored in the material, both due to changes in the austenite grain
surface per unit volume, S
V
, and due to the expected change in dislocation
density. By balancing the plastic work stored against the energy of defects
created, we obtain:
0.05
y

S
V
+ (6.14)
where
y
is the yield stress of austenite (measured to be 30 MPa at
1000
C [Maalekian et al., 2006]), 0.5b

2
is the dislocation energy per
unit length and
is the austenite grain boundary energy per unit area.

The magnitude of grain boundary energy must vary with the chemical com-
position of alloy and temperature. In absence of the actual data, this has
been taken to be 0.6 J m
2
[Murr et al., 1982].
For axisymmetric compression [Zhu et al., 2006],
S
V
S
V
0
=
[8e
+ 4e
2
]
1
2
+
1
3
[e
+ 2
]
1 + 2
3
(6.15)
The value of S
V
at zero deformation is given by S
V
0
= 2/d
where d
= 26
3 m is the mean lineal intercept dening the austenite grain size [Maalekian
et al., 2006].
The stored energy due to deformation is simply calculated using the
term on the left-hand side of equation 6.14. Removing the component due
to the change in S
V
(equation 6.15) from equation 6.14 gives the change in
dislocation density.
The predicted depression in the martensite-start temperature with in-
creasing prior plastic strain in austenite is shown in Figure 6.11. The calcu-
lated trends compare well with the measured value.
148 6.5 Summary
Figure 6.11: Calculated change in M
S
as a function of plastic strain in austen-
ite prior to its transformation, the points represent experimental data.
6.5 Summary
The role of stress and strain on martensitic transformation has been studied.
A theory has been developed to calculate the critical strain necessary for the
onset of mechanical stabilisation in steels. The chemical composition of the
alloys along with the transformation temperature are used as key inputs to
the model.
The theory has been used to demonstrate why large strain values are
necessary to stabilise certain stainless steels at ambient temperature. The
theory also helps to interpret the sub-unit mechanism of bainite growth and
the depression of M
S
temperature with prior plastic strain in austenite. The
theory indicates the possibility that high-carbon austenite in TRIP-assisted
steels might actually transform to martensite due to stress, rather strain,
during deformation.
Chapter 7
Formability
TRIP-assisted steels are expected to possess enhanced formability. This is the
ability of a material to be shaped, an ability related to plasticity. Formability
is characterised by dierent test methods. The purpose of the present chapter
is to study the inuence of simple mechanical properties such as the strength
and elongation on formability.
7.1 Stretch-angeability
Several methods have been designed to characterise the forming ability, one
of which is the hole expansion test. During this test, a hole is rst made
in the steel blank. The hole is then allowed to expand under a punch until
cracks appear on the surface. The ratio of the increase in diameter to the
initial diameter is called the hole expansion ratio. The higher the ratio, the
better is the stretch-angeability of the steel.
There are examples of strong steels tested for stretch-angeability [Mat-
sumura et al., 1992; Sudo et al., 1982]. Each of these materials possesses a
149
150 7.2 Neural networks
similar ferrite-rich microstructure with a dispersion of a hard phase which
could be martensite or bainite or both. The steels may also contain retained
austenite.
There has been much research in understanding the eect of microstruc-
ture on the stretch-angeability of these steels. For example, delayed strain-
induced transformation of the retained austenite into martensite is known to
improve the properties [Sugimoto et al., 1999]. Volume fraction, hardness and
distribution of the hard phase may also inuence the performance [Hasegawa
et al., 2004]. It has even been possible to achieve superior hole-expansion
ratio over the multiphase steels by using either fully bainitic or martensitic
steels [Sugimoto et al., 2002a,b, 2000].
However, there is no proper understanding on the inuence of the sim-
ple mechanical properties, such as strength and elongation, on the stretch-
angeability. During a tensile test, the specimen rst undergoes uniform
deformation. But plastic instability sets in when the work-hardening can
no longer compensate for the increased stress due to the reduction of cross
section. The resulting non-uniform strain is generally not considered to be
useful during forming, as this might degrade the aesthetic of the formed com-
ponent. To study the relationship between the tensile properties and the hole
expansion ratio forms the aim of the present work.
7.2 Neural networks
Neural network analysis was adopted in this study. This represents a general
method of regression that ts a non-linear function to experimental data.
Details of the method can be found in Chapter 4.
Analysis
The data presented in Table 7.1 are due to Fang et al. [2003a,b]; Hasegawa et
al. [2004]; Sudo et al. [1982]; Sugimoto et al. [1999]. They originate from steels
with microstructures of ferrite-martensite or ferrite-bainite, with or without
retained austenite. The data were used to create models for predicting the
hole expansion ratio () against various combinations of the tensile properties
viz. yield strength (YS), ultimate tensile strength (UTS), uniform elongation
(UEL), total elongation (TEL), yield ratio (YR) and the strength-elongation
product (UTS-UEL and UTS-TEL).
Each variable in the database was normalised as
x
n
=
x x
min
x
max
x
min
0.5 (7.1)
where x
n
is the normalised value of the variable x, with x
min
and x
max
as the
minimum and the maximum values respectively. The database was divided
randomly into two halves. Each model was trained using one half of the
dataset. The other half was used to test the performance of the model with
unseen data.
The rst model was created with the YS, UTS, UEL, YR and UTS-UEL
as the input parameters. Figure 7.1(a) illustrates the reduction in the noise
level with increasing number of hidden units. The nal model is formed
as a committee of models based on the minimum combined test error on
increasing the number of networks (Figure 7.1(b)). The nal committee in
Inputs Minimum Maximum Average S.D.
YS / MPa 335 783 444 92
UTS / MPa 484 1028 675 155
UEL / % 8 39 18 7
TEL / % 10 43 27 8
YR 0.49 0.82 0.67 0.09
UTS-UEL / MPa-% 6992 29715 12192 6113
UTS-TEL / MPa-% 9016 32379 18069 6894
Output Minimum Maximum Average S.D.
/ % 9 151 65 45
Table 7.1: Data used in the analysis, S.D. represents standard deviation.
this case was made of only one sub-model.
Figure 7.2 compares the measured hole expansion ratios with the cor-
responding model predictions made over the entire dataset. In general, the
predicted hole expansion ratios compare well with the corresponding mea-
sured values, but there are a few outliers. However, the uncertainty is seen to
be large. The model-perceived signicances of the input variables inuencing
the hole expansion ratio are presented in Figure 7.3. As can be seen, UTS
gained the maximum importance followed by the YS and the YR. Contrary
to general belief, UEL and UTS-UEL were perceived to be insignicant.
The apparent irrelevance of the UEL and the UTS-UEL might be real.
This is because, the attainment of UTS in the ow curve implies the end of
UEL. Therefore, UTS and UEL are actually not independent variables. To
test this hypothesis, two separate models were created. One of these had the
UTS with the UEL as the input variables while the other had the YS with
the UEL. Both these models were found to consist of only one sub-model in
the nal committee.
N
o
i
s
e
0.08
0.12
0.16
0.2
0.24
4 8 12 16 20
(a)
0.35
0.37
0.39
4 8 12 16 20
0.41
Number of models
C
o
m
b
i
n
e
d

T
e
s
t

E
r
r
o
r
(b)
Figure 7.1: Variation of (a) noise level, , with increasing number of hidden
units and (b) combined test error with number of models in committee.
0
40
80
120
160
0 40 80 120 160
p
r
e
d
i
c
t
e
d

/

%
measured / %
flange_1_result.dat using 4:5
Figure 7.2: Predicted hole expansion ratios compared with the measured
values using the model with YS, UTS, UEL, YR and UTS-UEL as the input
parameters.
Figure 7.3: Signicances of YS, UTS, UEL, YR and UTS-UEL as the input
variables.
(a)
(b)
Figure 7.4: Signicances of (a) UTS and UEL and (b) YS and UEL as the
input variables.
Figure 7.4 shows the signicances of the variables as perceived by the
model. The signicance of UEL was indeed boosted in the presence of the
YS, as compared to that with the UTS. However, its magnitude remained
small (Figure 7.4(b)). This indicates a possibility that the TEL could be a
better parameter than the UEL in describing the hole expansion ratio. A
model was therefore created with YS, UTS, TEL, YR and UTS-TEL as the
input variables. The nal committee of the model was found to have eight
sub-models.
Figure 7.5 presents the signicances of the input variables as perceived
by each sub-model in the nal committee. UTS, UTS-TEL and TEL were
found to exert considerable inuence on the hole expansion ratio, while YS
and YR did not gain any importance. However, the signicances for each
input parameters vary in dierent sub-models.
Figure 7.5: Signicances of YS, UTS, TEL, YR and UTS-TEL as the input
variables.
Figure 7.6 depicts the predictions made using this model over the en-
tire dataset. The predicted values compare well with the measured values.
Furthermore, the magnitudes of the error bars associated with the predic-
tions using this model are much reduced as compared to the rst model
(Figure 7.2). There are no signicant outliers in the predictions. All these
suggest, the model with the TEL as an input performs appreciably better
than the one with the UEL as an input.
0
40
80
120
160
0 40 80 120 160
p
r
e
d
i
c
t
e
d

/

%
measured / %
values using the model with YS, UTS, TEL, YR and UTS-TEL as the input
parameters.
The above results indicate the importance of UTS, TEL and UTS-
TEL in controlling the hole expansion ratio. This appears to be physically
reasonable, given that the hole is allowed to expand until cracks appear on the
surface. This therefore tests the ability of the material to deform plastically
until fracture, albeit locally. Two dierent models were therefore created
with the input parameters as UTS and UTS-TEL in one case while UTS and
TEL in the other.
From Figure 7.7, the signicances of both UTS-TEL and TEL can be
seen to be appreciable, although quantitatively these are less than that of the
UTS. In each case, the model consisted of only one sub-model. Figure 7.8
shows that the predictions made over the entire dataset compare well with
the corresponding measured values.
These models were used further to predict the variation in the hole
expansion ratio as function of UTS, TEL and UTS-TEL. Figure 7.9 and Fig-
ure 7.10 present the results as surface plots with the corresponding uncertain-
ties which were between 10-25. At constant UTS-TEL, the hole expansion
ratio seems to be higher at lower UTS (Figure 7.9(a)). This might be due to
the higher TEL (keeping UTS-TEL the same). However, at the same UTS,
the ratio is seen to be decreasing with increasing UTS-TEL, ultimately be-
coming constant with UTS-TEL exceeding 20000 MPa-%. This is therefore
not consistent with the benecial eect of the total elongation.
Figure 7.10(a) illustrates a more drastic reduction in the hole expansion
ratio with increasing UTS as compared to Figure 7.9(a). At constant UTS,
the ratio is however seen to remain almost constant with increasing total
elongation, although a positive correlation between the stretch-angeability
and TEL could be imagined ideally.
159 7.4 Summary
7.4 Summary
The present work identies the ultimate tensile strength to be the most
important tensile property inuencing the hole expansion ratio. The uniform
elongation loses its signicance as an input to the model, which also includes
the ultimate tensile strength as a variable. This is because both the properties
appear to be identical with respect to the ow curve.
The stretch-angeability deteriorates with increasing ultimate tensile
strength. Total elongation might be thought to be important, as hole ex-
pansion tests rely on the cracking of the surface while expanding the hole.
However, the present work suggests that increasing total elongation, keeping
the ultimate tensile strength same, might not be benecial.
160 7.4 Summary
(a)
(b)
Figure 7.7: Signicances of (a) UTS and UTS-TEL and (b) UTS and TEL
as the input variables.
161 7.4 Summary
0
40
80
120
160
0 40 80 120 160
p
r
e
d
i
c
t
e
d

/

%
measured / %
(a)
0
40
80
120
160
0 40 80 120 160
p
r
e
d
i
c
t
e
d

/

%
measured / %
(b)
values using the models with (a) UTS and UTS-TEL and (b) UTS and TEL
as the input parameters.
162 7.4 Summary
600
800
1000
UTS / MPa
10000
20000
30000
UTS-TEL / MPa-%
40
80
120
160
predicted / %
(a)
600
800
1000
UTS / MPa
10000
20000
30000
UTS-TEL / MPa-%
10
15
20
uncertainty / %
(b)
Figure 7.9: Predicted variation in (a) the hole expansion ratio with UTS and
UTS-TEL and (b) the corresponding uncertainties.
163 7.4 Summary
600
800
1000
UTS / MPa
10
20
30
40
TEL / %
40
80
120
160
predicted / %
(a)
600
800
1000
UTS / MPa
10
20
30
40
TEL / -%
10
15
20
25
uncertainty / %
(b)
Figure 7.10: Predicted variation in (a) the hole expansion ratio with UTS
and TEL and (b) the corresponding uncertainties.
Chapter 8
Conclusions
TRIP-assisted steels have the advantage of large uniform elongation in spite
of the high strength. This makes them suitable for making cars with lighter
body weight but adequate safety. The steels contain some 10-30 vol.% high-
carbon retained austenite in ferrite-rich microstructures. The transformation
of austenite into martensite during deformation is generally believed to be
responsible for the properties of these steels. This is despite the fact that
the martensite that forms in these steels must inherit the carbon of austenite
and hence should be brittle.
It has been demonstrated here that the ner martensite plates are bet-
ter able to resist cracking. In TRIP-assisted steels, retained austenite exists
in a nely divided state with the grain size of around 1 m. The martensite
that forms from this austenite therefore remains with a ne plate size and
hence is not a liability for these steels, in spite of its high carbon content.
To maximise the properties, the stability of the retained austenite
should be such that it transforms progressively during deformation. Car-
bon content of the austenite inuences its stability. Retained austenite vol-
164
165
ume fraction and its carbon content are therefore important microstructural
parameters inuencing the properties of these steels. Using published data,
neural network models were created relating these parameters with the chem-
ical composition and the heat treatment conditions of these materials. These
models were found to be capable of illustrating the eects of various factors
on the retained austenite fraction and its carbon content, as expected from
the metallurgical principles.
The retained austenite content in steels transformed to bainite increases
with the addition of silicon. However, silicon leads to complications during
processing and deteriorates the surface condition of the steel. The neural
network model, along with a genetic algorithm, was used to formulate an al-
loy, lean in silicon concentration yet capable of retaining austenite in copious
amounts. The computational methods gave rise to a new type of material,
here designated -TRIP steel. The alloy shows an enhanced tendency for
-ferrite formation during solidication and cannot be made fully austenitic.
Usually the ferrite, that constitutes the microstructure of TRIP-assisted steel,
is generated out of the allotriomorphic transformation of austenite. By con-
trast, the present alloy gives an opportunity to substitute some of the allotri-
omorphic ferrite in these materials by -ferrite.
Experiments using alloys based on the optimisation calculation demon-
strated the formation of -ferrite during solidication. These alloys were
further heat-treated producing radically dierent microstructures containing
-ferrite, bainite and retained austenite. Specimens in heat-treated condi-
tions recorded impressive mechanical properties at room temperature. How-
ever, the ductility deteriorates when the tests are done at 100
C. This could
166
be due to the enhanced stability of the retained austenite against transfor-
mation into martensite at the high temperature.
The transformation of the retained austenite into martensite during
deformation could be either due to stress or strain. Stress generally adds to
the driving force while plastic strain retards the growth of martensite plates,
eventually causing the mechanical stabilisation of the austenite. A theory
capable of predicting the critical strain necessary for the onset of stabilisation
has been developed. Using the theory, it has been possible to explain the
requirement of the large strain for stabilising certain austenitic steels and
the sub-unit mechanism of bainite growth. Calculations involving the theory
indicate that the high-carbon retained austenite in TRIP-assisted steels may
actually transform due to stress rather strain. Further experimental work
is necessary to clarify the role of stress and strain on the transformation of
retained austenite in these steels.
There are various complex tests to assess the formability of these
materials. However, there has been no proper attempt in understanding
these results in terms of the tensile properties. Published data on stretch-
angeability of these steels were analysed using neural networks. It has been
shown that the ultimate tensile strength of the steel is the most important
criterion inuencing the stretch-angeability.
Bibliography
Andrews, K. W. Journal of The Iron and Steel Institute 203 (1965) 721
Angel, T. Journal of The Iron and Steel Institute 177 (1954) 165
Barlat, F., Glazov, M. V., Brem, J. C. and Lege, D. J. International Journal
of Plasticity 18 (2002) 919
Basuki, A. and Aernoudt, E. Journal of Materials Processing Technology
89-90 (1999) 37
Bhadeshia, H. K. D. H. and Honeycombe, R. W. K. Steels: Microstructure
and Properties, Elsevier, UK, (2006)
Bhadeshia, H. K. D. H. ISIJ International 42 (2002) 1059
Bhadeshia, H. K. D. H. Bainite in Steels, IOM Communications Ltd, UK
(2001)
Bhadeshia, H. K. D. H. ISIJ International 39 (1999) 966
Bhadeshia, H. K. D. H. Materials Science and Engineering A 273-275 (1999)
58
Bhadeshia, H. K. D. H. Acta Metallurgica 29 (1981) 1117
167
168 BIBLIOGRAPHY
Bhadeshia, H. K. D. H. and Edmonds, E. V. Metallurgical and Materials
Transactions A 10 (1979) 895
Bouet, M., Fillipone, R., Essadiqi, E., Root, J. and Yue, S. 42nd Mechani-
cal Working and Steel Processing Conference Proceedings, Iron and Steel
Society/AIME, USA, 38 (2000) 55
Bouet, M., Root, J., Es-Sadiqui, E. and Yue, S. 40th Mechanical Working and
Steel Processing Conference Proceedings, Iron and Steel Society/AIME,
USA, 36 (1998) 675
Bouet, M., Root, J., Es-Sadiqi, E. and Yue, S. Materials Science Forum
284-286 (1998) 319
Breedis, J. F. and Robertson, W. D. Acta Metallurgica 11 (1963) 547
Capdevila, C., Caballero, F. G. and Garcia de Andres, C. Journal de
Physique IV 112 (2003) 217
Chen, H. C., Era, H. and Shimizu, M., Metallurgical and Materials Transac-
tions A 20 (1989) 437
Choi, I. D., Bruce, D. M., Kim, S. J., Lee, C. G., Park, S. H., Matlock, D.
K. and Speer, J. G. ISIJ International 42 (2002) 1483
Choi, B. Y., Krauss, G and Matlock, D. K. Scripta Metallurgica 22 (1988)
1575
Christian, J. W. The Theory of Transformations in Metals and Alloys, Perg-
amon, UK (2002)
Christian, J. W. Proceedings of ICOMAT79, MIT Press, USA (1979) 230
169 BIBLIOGRAPHY
Cox, H. L. British Journal of Applied Physics 3 (1952) 72
Davies, R. G. and Magee, C. L. Metallurgical and Materials Transactions A
3 (1972) 307
De Meyer, M., Vanderschueren, D., De Cooman, B. C., ISIJ International
39 (1999) 813
De Meyer, M., Vanderschueren, D., De Cooman, B. C., 41st Mechanical
Working and Steel Processing Conference Proceedings, Iron and Steel So-
ciety/AIME, USA, (1999) 265
Dieter, G. E., Mechanical Metallurgy, McGraw-Hill, UK (1988)
Dyson, D. J. and Holmes, B Journal of The Iron and Steel Institute 208
(1979) 469
Epogy 2003 Users guide, Synaps Inc
Fang, X., Fan, Z., Ralph, B., Evans, P. and Underhill, R. Journal of Materials
Processing Technology 132 (2003) 215
Fang, X., Fan, Z., Ralph, B., Evans, P. and Underhill, R. Journal of Materials
Science 38 (2003) 3877
Fischer, J. C. Metals Transactions 185 (1949) 688
Frank, F. C. Acta Metallurgica 1 (1953) 15
Furnemont, Q., Kempf, M., Jacques, P. J., Goken, M. and Delannay, F.
Materials Science and Engineering A 328 (2002) 26
170 BIBLIOGRAPHY
Garcia-Mateo, C., Caballero, F. G. and Bhadeshia, H. K. D. H. ISIJ Inter-
national 43 (2003) 1238
Gallagher, M. F., Speer, J. G. and Matlock, D. K. Symposium on the Ther-
modynamics, Kinetics, Characterization and Modeling of: Austenite For-
mation and Decomposition, Minerals, Metals and Materials Society (TMS),
USA, (2003) 563
Gerberich, W. W., Thomas, G., Parker, E. R. and Zackay, V. F. Proceedings
of 2nd International Conference on Strength of Metals and Alloys, USA,
3 (1970)
Ghosh, G., and Olson, G. B. Journal of Phase Equilibrium 22 (2001) 199
Girault, E., Martens, A., Jacques, P., Houbaert, Y., Verlinden, B. and Van
Humbeeck, J. Scripta Materialia 44 (2001) 885
Goel, N. C., Sangal, S. and Tangri, K. Metallurgical and Materials Transac-
tions A 16 (1985) 2013
Guimaraes, J. R. C. and Saavedra, A. Materials Science and Engineering A
62 (1984) 11
Guimaraes, J. R. C. and Gomes, J. C. Acta Metallurgica 26 (1978) 1591
Guimaraes, J. R. C. Scripta Metallurgica 6 (1972) 795
Hanzaki, A. Z., Hodgson, P. D. and Yue, S. ISIJ International 35 (1995) 324
Hanzaki, A. Z., PhD Thesis, McGill University, (1994)
Hasegawa, K., Kawamura, K., Urabe, T. and Hosoya, Y. ISIJ International
44 (2004) 603
171 BIBLIOGRAPHY
Hashimoto, S., Ikeda, S., Sugimoto, K-I. and Miyake, S. ISIJ International
44 (2004) 1590
Hassani, F. and Yue, S. 41st Mechanical Working and Steel Processing Con-
ference Proceedings, Iron and Steel Society/AIME, USA, 37 (1999) 493
Hehemann, R. F. Phase Transformations, ASM, USA (1970) 397
Hiwatashi, S., Takahashi, M., Katayama, T. and Usuda, M., Journal of the
Japan Society for Technology of Plasticity 35 (1994) 1109
Honeycombe, R. W. K. The Plastic Deformation of Metals, Second Edition,
Edward Arnold, London, (1984)
Im, D. B., Lee, C. G., Kim, S. J. and Park, I. M. Journal of the Korean
Institute of Metals and Materials 40 (2002) 8
Im, D. B., Lee, C. G., Kim, S. J. and Park, I. M. Journal of the Korean
Institute of Metals and Materials 38 (2000) 447
Im, D. B., Lee, C. G., Song, B. H., Kim, S. J. and Park, I. M. Journal of the
Korean Institute of Metals and Materials 38 (2000) 1328
Imai, N., Komatsubara, N. and Kunishige, K. Japan Technical Information
Service (1992) 25
Irvine, K. J., Gladman, T. and Pickering, F. B. Journal of The Iron and
Steel Institute 207 (1969) 1017
Itami, A., Takahashi, M. and Ushioda, K. ISIJ International 35 (1995) 1121
Jacques, P. Journal de Physique IV 112 (2003) 297
172 BIBLIOGRAPHY
Jacques, P. J., Girault, E., Martens, A., Verlinden, B., Van Humbeeck, J,
and Delannay, F. ISIJ International 41 (2001) 1068
Jacques, P. J., Girault, E., Harlet, P. and Delannay, F. ISIJ International
41 (2001) 1061
Jacques, P., Furnemont, Q., Mertens, A. and Delannay, F. Philosophical
Magazine A 81 (2001) 1789
Jacques, P., Ladriere, J. and Delannay, F. Metallurgical and Materials Trans-
actions A 32 (2001) 2759
Jacques, P., Girault, E., Catlin, T., Geerlofs, N., Kop, T., Van der Zwaag,
S. and Delannay, F. Materials Science and Engineering A 273-275 (1999)
475
Jiao, S., Hassani, F., Donaberger, R. L., Essadiqi, E. and Yue, S. ISIJ Inter-
national 42 (2002) 299
Kim, S. J., Lee, C. G., Lee, T. H. and Oh, C. S. ISIJ International 42 (2002)
1452
Kim, S. J., Lee, C. G., Choi, I. and Lee, S. Metallurgical and Materials
Kim, C. Metallurgical and Materials Transactions A 19 (1988) 1263
Kim, S. J. and Lee, C. G. Journal of the Korean Institute of Metals and
Materials 37 (1999) 774
Koh, H. J., Lee, S. K., Park, S. H., Choi, S. J., Kwon, S. J. and Kim, N. J.
Scripta Materialia 38 (1998) 763
173 BIBLIOGRAPHY
Koistinen, P. P. and Marburger, R. E. Acta Metallurgica 7 (1959) 59
Konieczny, A. A. Processing and Fabrication of Advanced Materials XI, ASM
International, USA, (2003) 345
Lecroisey, F. and Pineau, A. Metallurgical and Materials Transactions A 3
(1972) 387
Lee, C. G., Kim, S. J., Lee, T. H. and Lee, S. Material Science and Engi-
neering A 371 (2004) 16
Lee, C. G., Kim, S. J., Song, B. H. and Lee, S. Metals and Materials 8 (2002)
435
Lee, C. G., Kim, S. J., OH, C. S. and Lee, S. ISIJ International 42 (2002)
1162
Leslie, W. C. and Miller, R. L. ASM Transactions Quaterly 57 (1964) 972
Ludwigson, D. C. and Berger, J. A., Journal of The Iron and Steel Institute
207 (1969) 63
Maalekian, M., Lendinez, M. L., Kozeschnik, E., Brantner, H. P., Cerjak,
H., Eect of hot plastic deformation of austenite on the transformation
characteristics of eutectoid carbon steel under fast heating and cooling
conditions, (2006, unpublished work)
Machlin, E. S. and Cohen, M. Trans. AIME 191 (1951) 746
MacKay, D. J. C., Information Theory, Inference and Learning Algorithms,
Cambridge University Press, UK, (2003)
174 BIBLIOGRAPHY
Mahieu, J., De Cooman, B. C., Maki, J., and Claessens, S. Iron and Steel-
maker 29 (2002) 29
Mahieu, J., Claessens, S. and De Cooman, B. C. Verlag Stahleisen GmbH,
Galvatech 2001, Germany, (2001) 644
Manganon, P. L. and Thomas, G. Metallurgical and Materials Transactions
A 1 (1970) 1577
Manohar, P. A., Kunishige, K., Chandra, T. and Ferry, M. Materials Science
and Technology 18 (2002) 856
Manohar, P. A., Kunishige, K. and Chandra, T., Materials Science Forum
426-432 (2003) 1127
MAP STEEL MUCG73 available freely in the world wide web,
http://www.msm.cam.ac.uk/map/steel/programs/mucg73-b.html
Maki, J., Mahieu, J., De Cooman, B. C. and Claessens, S. Journal of Mate-
rials Science and Technology 19(2003) 125
Maki, J., Mahieu, J., Claessens, S. and De Cooman, B. C. Verlag Stahleisen
GmbH, Galvatech 2001, Germany, (2001) 623
Mark, C. Proceedings of Cambridge Philosophical Society 52 (1956) 216
Matas, S. J. and Hehemann, R. F. Trans. Met. Soc. AIME 221 (1961) 179
Matsumura, O., Sakuma, Y. and Takechi, H. ISIJ International 32 (1992)
1014
Matsumura, O., Sakuma, Y. and Takechi, H. Transactions of the Iron and
Steel Institute 27 (1987) 570
175 BIBLIOGRAPHY
Matsumura, O., Sakuma, Y. and Takechi, H., Scripta Metallurgica 27 (1987)
1301
Mintz, B. Materials Science Forum 426-432 (2003) 1219
Mintz, B. International Materials Reviews 46 (2001) 169
MTDATA, National Physical Laboratory, Teddington, London, 2005
Murr, L. E. Interfacial phenomena in metals and alloys, Addison-Wesley
Publication Co., USA, (1975) 131
Murugananth, M., Babu, S. S. and David, S. A. Welding Journal, Supple-
ment, October, (2004) 267
Nagasaka, A., Sugimoto, K. and Kobayashi, M. ASM International, USA,
(1996) 557
Nakagaito, T., Shimizu, T., Furukimi, O. and Sakata, K., Tetsu-to-Hagane
89 (2003) 841
Nagasaka, A., Sugimoto, K-I., Kobayashi, M., Kobayashi, Y., Hashimoto, S.
Tetsu-to-Hagane 85 (1999) 885
Nagasaka, A., Sugimoto, K-I., Kobayashi, M. and Hashimoto, S. Tetsu-to-
Hagane 85 (1999) 552
Nagasaka, A., Sugimoto, K-I., Kobayashi, M., Shirasawa, H. Tetsu-to-Hagane
84 (1998) 218
Nelder J. A. and Mead R. Computer Journal 7 (1965) 308
176 BIBLIOGRAPHY
Olson, G. B. Deformation, Processing and Structure, American Society of
Metals, USA, (1982) 391
Olson, G. B. and Cohen, M. Proceedings of US-Japan Seminar on Mechan-
ical Behaviour of Metals and Alloys Associated with Displacive Transfor-
mation, Troy, NY, (1979) 7
Olson, G. B. and Cohen, M., Metallurgical and Materials Transactions A 7
(1976) 1897
Olson, G. B. and Cohen, M. Metallurgical and Materials Transactions A 7
(1976) 1905
Olson, G. B. and Cohen, M., Metallurgical and Materials Transactions A 7
(1976) 1915
Olson, G. B. and Cohen, M. Metallurgical and Materials Transactions A 6
(1975) 791
Olson, G. B. and Cohen, M. Journal of the Less-common Metals 28 (1972)
107
Onodera, H. and Tamura, I. Proceedings of US-Japan Seminar on Mechan-
ical Behaviour of Metals and Alloys Associated with Displacive Transfor-
mation, Troy, NY (1979) 24
Patel, J. R. and Cohen, M. Acta Metallurgica 1 (1953) 531
Pereloma, E. V., Timokhina, I. B. and Hodgson, P. D. Materials Science and
Engineering A 273-275 (1999) 448
177 BIBLIOGRAPHY
Pichler, A., Traint, S., Arnolder, G., Stiaszny, P., Blaimschein, M., and
Werner, E. A. Iron and Steelmaker, USA, 30, (2003) 21
Pichler, A., Traint, S., Arnoldner, G., Stiaszny, P., Blaimschein, M. and
Werner, E. A. 44th Mechanical Working and Steel Processing Conference
Proceedings, Iron and Steel Society/AIME, USA, 40 (2002) 121
Pichler, A., Stiaszny, P., Potzinger, R., Tikal, R. and Werner, E. 40th Me-
chanical Working and Steel Processing Conference Proceedings, Iron and
Steel Society/AIME, USA 36 (1998) 259
Pickering, F. B. Physical Metallurgy and the Design of Steels, Applied Science
Publishers Ltd, England, (1978)
Pyshmintsev, I. Y., De Meyer, M., De Cooman, B. C., Savray, R. A.,
Shveykin, V. P.and Vermeulen, M., Metallurgical and Materials Trans-
actions A 33 (2002) 1659
Raghavan, V. Martensite, A tribute to Morris Cohen, ASM International,
USA, (1992) 197
Rao, B. V. N. and Rashid, M. S. Metallography 16 (1983) 19
Rashid, M. S. Formable HSLA and Dual Phase Steels, USA, (1979) 1
Rashid, M. S. Dual Phase and Cold Pressing Vanadium Steels in the Auto-
mobile Industry, Berlin, (1979) 32
Reisner, G., Warner, E. A. and Fischer, F. D. International Journal of Solid
Structures 35 (1998) 2457
178 BIBLIOGRAPHY
Rigsbee, J. M. Proceedings of the ICOMAT79 Cambridge, MA, USA (1979)
381
Sadhukhan, S., Das, K. P., Bandyopadhyay, N. R. and Banerjee, M. K.
Journal of the Institution of Engineers, India, 82 (2001) 65
Sakuma, Y., Matlock, D. K., Krauss, G. Materials Science and Technology
9 (1993) 718
Sakuma, Y., Matlock, D.K., Krauss, G., Metallurgical and Materials Trans-
actions A 23 (1992) 1221
Sakuma, Y., Matsumara, O. and Takechi, H., Metallurgical and Materials
Sakuma, Y., Matsumura, O. and Akisue, O. ISIJ International 31 (1991)
1348
Sangal, S., Goel, N. C. and Tangri, K. Metallurgical and Materials Transac-
tions A 16 (1985) 2023
Scheil, E., Zeitschrift f ur anorganische and allgemeine, Chemie 207 (1932)
21
Schrader, A. and Wever, F. Arch. Eisenhittenwesen 23 (1952) 489
Sherif, M. Y., Garcia-Mateo, C., Sourmail, T. and Bhadeshia, H. K. D. H.
Materials Science and Technology 40 (2004) 319
Shi, W., Li, L., Zhou, Y., Fu, R. Y. and Wei, X. C. Heat Treatment of Metals,
China, 8 (2002) 9
179 BIBLIOGRAPHY
Song, S. M., Sugimoto, K-I., Kandaka, S., Futamura, A., Kobayashi, M. and
Masuda, S. Materials Science Research International, Japan, 9 (2003) 223
Song, S. M., Sugimoto, K-I., Kandaka, S., Futamura, A., Kobayashi, M. and
Masuda, S. Journal of the Society of Materials Science, Japan, 50 (2001)
1091
Streicher, A. M., Speer, J. G. and Matlock, D. K. Steel Research 73 (2002)
287
Sudo, M., Tsukatani, I. and Shibata, Z. Metallurgy of Continuous Annealed
Sheet Steel, AIME (1982) 310
Sugimoto, K-I., Tsunezawa, M., Hojo, T. and Ikeda, S. ISIJ International
44 (2004) 1608
Sugimoto, K-I., Nakano, K., Song, S-M. and Kashima, T. ISIJ International
42 (2002) 450
Sugimoto, K-I., Kanda, A., Kikuchi, R., Hashimoto, S. I., Kashima. T. and
Ikeda, S. ISIJ International 42 (2002) 910-915
Sugimoto, K-I., Sakaguchi, J., Iida, T. and Kashima, T. ISIJ International
40 (2000) 920
Sugimoto, K-I., Nagasaka, A., Kobayashi, M. and Hashimoto, S. I. ISIJ
International 39 (1999) 56
Sugimoto, K-I., Kobayashi, M. and Nagasaka, A. Canadian Institute of Min-
ing, Metallurgy and Petroleum (Canada) (1996) 503
180 BIBLIOGRAPHY
Sugimoto, K-I., Kobayashi, M. Nagasaka, A. and Hashimoto, S. I. ISIJ In-
ternational 35 (1995) 1407
Sugimoto, K-I. and Kobayashi, M., Iron and Steel Society/AIME, USA
(1994) 255
Sugimoto, K-I., Misu, M., Kobayashi, M. and Shirasawa, H. ISIJ Interna-
tional 33 (1993) 775
Sugimoto, K-I., Usui, N., Kobayashi, M. and Hashimoto, S. I. ISIJ Interna-
tional 32 (1992) 1311
Sugimoto, K-I., Kobayashi, M. and Hashimoto, S. I., Metallurgical and Ma-
terials Transactions A 23 (1992) 3085
Swallow, E. and Bhadeshia, H. K. D. H. Materials Science and Technology
12 (1996) 121
Takahashi, M. Nippon Steel Technical Report 88 (2003) 2-7
Takechi, H., Matsumura, O., Sakuma, Y. Japan Kokai Tokyo Koho Japan
Patent 62, 188, 729 (1987)
Tamura, I., Maki, T., Hato, H., Tomota, Y. and Okada, M. Proceedings of
2nd International Conference on Strength of Metals and Alloys, Asilomar,
CA, 3 (1970) 894
Tomita, Y. and Morioka, K. Materials Characterization 38 (1997) 243
Traint, S., Pichler, A., Stiaszny, P. and Wemer, E. A. 44th Mechanical
Working and Steel Processing Conference Proceedings, Iron and Steel So-
ciety/AIME, USA 40 (2002) 139
181 BIBLIOGRAPHY
Traint, S., Pichler, A., Tikal, R., Stiaszny, P., Wemer, E. A., 42nd Mechan-
ical Working and Steel Processing Conference Proceedings, Iron and Steel
Society/AIME, USA 38 (2000) 549
Uenishi, A., Kuriyama, Y. and Takahashi, M. Nippon Steel Technical Report
(Japan) 81 (2000) 17
ULSAB-AVC, http://www.worldautosteel.org/ulsab-avc
Venables, J. A. Philosophical Magazine 7 (1964) 35
Vodopivec, F., Reskovi, S. and Mamuzic, I. Materials Science and Technology
15 (1999) 1293
Wang, H. S., Yang, J. R., Bhadeshia, H. K. D. H. Materials Science and
Technology 21 (2005) 1323
Wang, Z. C., Kim, S. J., Lee, C. G. and Lee, T. H., Journal of Materials
Processing Technology 151 (2004) 141
Wei, X. C., Li, L., Fu, R. Y. and Shi, W. Journal of Iron and Steel Research
International, China, 10 (2003) 49
Wei, X., Fu, R., Li, L. and Shi, W., Shanghai Jinshu (Shanghai Metals),
China, 24 (2002) 32
Wei, X. C., Li, L., Fu, R. Y. and Shi. W. Acta Metallurgica Sinica (English
Letters), China 15 (2002) 285
Wiles, D. B. and Young, R. A. Journal of Applied Crystallography 14 (1981)
149
182 BIBLIOGRAPHY
Yan, B. 44th Mechanical Working and Steel Processing Conference Proceed-
ings, Iron and Steel Society, AIME, USA, 40 (2002) 509
Yan, B. and Xu, K. Iron and Steelmaker 30 (2003) 33
Yan, B. and Xu, K. 44th Mechanical Working and Steel Processing Confer-
ence Proceedings, Iron and Steel Society, AIME, USA 40 (2002) 493
Zackay, V. F., Parker, E. R., Fahr, D. and Bush, R., Trans. ASM 60 (1967)
25
Zaeerer, S., Ohlert, J. and Bleck, W. Acta Materialia 52 (2004) 2765
Zhu, C. M. S. Q., Sellars, C. M. and Bhadesiha, H. K. D. H. Materials Science
and Technology (2006) submitted

Transformations in TRIP-assisted Steels: Microstructure and Properties

Uploaded by

Copyright:

Available Formats

Transformations in TRIP-assisted Steels: Microstructure and Properties

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Transformations in TRIP-assisted Steels: Microstructure and Properties

Uploaded by

Copyright:

Available Formats

Transformations in TRIP-assisted Steels:

Microstructure and Properties

represent austenite, allotriomorphic ferrite,

represents the paraequilibrium ( + )- phase bound-

is the volume fraction of martensite obtained at a strain . The

is the original retained austenite volume fraction and V

to the rolling direction. A higher r value leads to better deep

C for 1200 s, some retained austenite

C. In between these two stages, the sheet is passed through

C, development of mechanical properties

C, steels containing carbon in excess of 0.4 wt%

C for 10 minutes in order to generate a range of austenite grain sizes for

C for 10 minutes followed by soaking at 950

C for another 10 minutes

C revealed cracks across the coarser plates in periodic

C and the plates, in general, were observed to be singly-cracked. The

C was found to be free of any cracks. Micrographs

C are presented in Figures 3.5 and 3.6 respectively.

C for 10 minutes and soaked at 950

C were also observed in this sample,

= 1200 and 1100

C (Figure 3.12). Austenite, ferrite, cementite and liquid phases

C for 48 h before further experiments.

C for 10 minutes before quenching in water

is a constant. The hyperbolic tangent function

C, 300 s and bainite transformation conditions

C, 500 s. Both carbon and manganese are known to be austenite

C and 300 s. Retained austenite fraction maximises at an

C and (b) 375

C, 300 s and bainite transformation conditions of 400

C and (b) 425

C; for 0.2C-1.5Mn-1.5Al (wt%) steel at transfor-

C and (d) 425

C followed by bainite transformation at 300

C, the predicted intercritical annealing

C was calculated to be Fe-0.67C-2.54Mn-0.53Si-1.35Al-

C therefore appear reasonable.

are the volume fractions of bainite and austenite in the

= 0.0497 and assuming

= 0 in the above equation, the volume fraction of austenite remaining

C is calculated to be 0.628. As-

intervals and a 5 s dwell at each step. Cu-K

C. Specimens were machined

C in equilibrium with liquid. Austenite forms on further cooling, lead-

C. The experimentally determined fraction

C is rich in carbon (Ta-

C for 12 h followed by a sec-

C for 6 h. The MAP STEEL MUCG73 program

C was calculated using

C. Samples were therefore initially heated at

C for 12 h and were subsequently allowed to transform into bainite at

C. The more gradual slope of the elastic part of the curve

C. The true stress-strain plot

C [Maalekian et al., 2006]), 0.5b

is the austenite grain boundary energy per unit area.

You might also like