Materials perspective

ClassiŽ cation and quantiŽ cation of

microstructures in steels
G. Thewlis

The International Institute of Welding (IIW) microstructure classiŽ cation scheme for ferrous weld metals has been
investigated as a basis for the quantiŽ cation of complex microstructures in steels. The aim has been to cover the full
range of microstructures observed in plain carbon and low alloy steel products, as well as ferritic weld metals and
parent plate heat affected zones. The mechanisms of formation of the principal structures and the characteristic
ferrite morphologies produced in the reconstructive and displacive transformation regimes of ferrous materials have
been brie y reviewed. The classiŽ cation and terminology used for intragranular as well as austenite grain boundary
microstructural constituents have been considered, and also the way in which transformation products are orientated
in space. Problems encountered in relating microstructural constituents to principal structures have been discussed
in detail and solutions proposed. The microstructure classiŽ cation and terminology used in the IIW scheme have
been built upon and new terminology incorporated into a table providing descriptions of the principal structures and
sub-category components. A new classiŽ cation scheme has been deŽ ned in the form of  ow charts with guidelines for
identifying the principal structures. Evaluation exercises have been carried out with the new scheme. These have
shown that a reasonable degree of consistency may be obtained between operators in identifying primary ferrite,
pearlite, martensite and the transformation products constituting ferrite sideplate and acicular ferrite structures,
notably Widmanstätten ferrite and bainite. A means is thus provided of obtaining database information for
developing microstructure – property relationships, or generating data for calibrating physical models, which have
the principal structures as their output. MST/5675

Keywords: Steel microstructures, Low alloy steels, Ferrite, Bainite, Martensite, Metallography, Microstructure classiŽ cation,
Phase transformation products

The author is with Corus Research, Development and Technology, Swinden Technology Centre, Moorgate, Rotherham,
S60 3AR, UK (graham.thewlis@corusgroup.com). Manuscript received 17 October 2002; accepted 22 September 2003.
# 2004 IoM Communications Ltd. Published by Maney for the Institute of Materials, Minerals and Mining.

structures in the reconstructive (diffusion controlled with

Introduction
With the advances in computer power in recent years, there
has been increasing interest in the development of models
to predict the microstructure of steel products, particularly
those processed by thermomechanical treatments1 – 4 or
fabricated by welding.5 – 1 1 The driving force for much of
the modelling work has been the need for computer based
systems to control and optimise microstructure and mecha-
nical properties. Linear regression analysis on large data-
bases of information has proved a useful tool in generating
composition– structure– property relationships.More recently,
neural network techniques9 have enabled prediction of situ-
ations too complex for simple analytical models or multiple
regression techniques. Regression and neural network
models often Ž nd use as online models for process control.
However, they are restricted to well deŽ ned data over limited
ranges of composition and process parameters. Such con-
straints can to a large extent be circumvented by the
development of physical models based on fundamental
metallurgical principles. The major advantage of physical
models is their general applicability. They can be used as
design tools for a wide variety of new materials and pro-
ducts. Classical nucleation and growth theory may, for
example, be used to predict the microstructure of steels
processed to a given austenite grain size and cooled at
different rates through the austenite to ferrite transforma-
tion temperature range.4 ,6 – 8 The transformation sequence
during cooling and the phase proportions of allotrio-
morphic ferrite, pearlite, Widmanstätten ferrite, bainite and
martensite may be the outputs. The latter are the principal
slow rates of reaction) and displacive (shear dominated with
rapid rates of reaction) transformation regimes of contin-
uous cooling transformation (CCT) diagrams.1 2 ,1 3
While the development of regression and neural network
models requires good quality database information, the
development of sophisticated physical models for micro-
structure prediction in steels has led to a need for accurate
calibration data. However, the microstructures observed in
steel products are complex. A variety of reaction products
may form at austenite grain boundary sites in thermo-
mechanically processed or heat treated steels. In the fusion
zone of welds, the simultaneous and competitive formation
of a variety of phases from both austenite grain boundary
and intragranular sites may occur while, in the parent plate
heat affected zone (HAZ), steep thermal gradients may give
rise to a wide range of transformation products. A scheme is
thus required for classifying and quantifying complex steel
microstructures.
Classifying and quantifying the microstructures of steels
has long been a contentious issue.1 4 – 1 8 Depending on the
plane of observation, constituents that are part of the same
principal structure may appear morphologically different
giving rise to sub-category components. Furthermore
some structures may have similar morphological or generic
features but be mechanistically different. A scheme for
identifying the various ferrite morphologies in isothermally
transformed steels was Ž rst used by Dubé et al.1 7 and later
extended by Aaronson.1 8 However, the effect of continuous
cooling was to render the distinguishing morphological
features much less distinct. Allotriomorphic ferrite mor-
phologies were readily identiŽ ed and also various sideplate

DOI 10.1179/026708304225010325 Materials Science and Technology February 2004 Vol. 20 143
144 Thewlis Classi®cation and quanti®cation of microstructures in steels
morphologies (often classed as bainite). Widmanstätten
ferrite was difŽ cult to place but was regarded as a generically
similar structure to bainite. Intragranular component phases
such as acicular ferrite posed a much greater degree of
difŽ culty. Much effort was made by the welding fraternity in
the 1980s to develop an overall microstructure quantiŽ cation
scheme for weld metals incorporating both prior austenite
grain boundary and intragranular nucleated constituents,
and addressing stereological effects, i.e. the way constituents
are orientated in space.1 9 ,20 A scheme was devised which
became recognised as the International Institute of Welding
(IIW) classiŽ cation.1 9 Most of the constituents deŽ ned in the
IIW scheme were relatively easily identiŽ ed. Furthermore,
the scheme could just as readily be applied to steels where
austenite grain boundary transformations dominate, as to
weld metals where intragranulartransformationsare the rule.
However, identiŽ cation of the actual transformation pro-
ducts constituting component structures such as ferrite
sideplate and acicular ferrite has proved difŽ cult. Anelli
and Di Nunzio2 1 recently devised a scheme providing
guidance on identifying transformation products associated
with sideplate structures which has had some success, but
stereological effects and intragranular constituents were not
treated in depth.
The objective in the current work has therefore been to
investigate the IIW microstructure classiŽ cation scheme as a
basis for quantiŽ cation of complex microstructures in steels.
The overall aim has been to develop a scheme that, although
requiring a basic knowledge as to the mechanism of for-
mation of the principal structures, will be relatively easy to
use given optical microscopy, standard specimen polishing
and etching techniques and appropriate guidance. The
approach has been to review microstructural constituents
in the IIW scheme in the context of the development of
principal structures found in the reconstructive and dis-
placive transformation regimes of steels. Detailed intragra-
nular as well as austenite grain boundary transformation
products have been considered and also stereological effects.
Problems relating microstructural constituents to principal
structures in the IIW scheme have been investigated
together with possible solutions so that a new quantiŽ cation
scheme may be developed with a much broader application
range. The intention has been to cover microstructures
observed in carbon (up to abount 0.8%) and low alloy (up to
approximately 5%) steels, as well as weld metals (up to
0.10%C and 5% alloy) and weld HAZs.
ClassiŽ cation of microstructures and
terminology
In this section, the mechanisms of formation of the principal
structures and the characteristic ferrite morphologies pro-
duced in the reconstructive and displacive transformation
regimes of ferrous materials are brie y reviewed. The
classiŽ cation and terminology used in the IIW scheme are
described, together with that of Dubé et al.1 7 to provide a
link with the early work on classiŽ cation of prior austenite
grain boundary ferrite morphologies. Terminology used in
recent work by the present author and co-workers2 2 is also
included to provide a contemporary view of complex intra-
granular transformations, including those generating the
microstructure commonly known as acicular ferrite.
RECONSTRUCTIVE TRANSFORMATION
REGIME
In the high temperature, reconstructive transformation
regime, a change from the austenite to ferrite crystal struc-
ture occurs by a reconstruction process involving movement
of atoms across the c/a transformationinterface. The principal
Materials Science and Technology February 2004 Vol. 20
1 Allotriomorphic and idiomorphic primary ferrite
phases are ferrite and pearlite. Reactions tend to be
diffusion controlled with slow rates.
Ferrite
In low hardenability materials, the Ž rst phase usually
forming on prior austenite grain boundaries during cooling
below the Ae 3 temperature is classically referred to as
allotriomorphic ferrite, as shown schematically in Fig. 1.
The ferrite nuclei have a Kurdjumov– Sachs (K – S) orien-
tation relationship with one austenite grain and grow into
the adjacent austenite grain with which they should normally
have a random orientation relationship.2 3 At some lower
temperature, ferrite may begin to nucleate on inclusions
inside the austenite grains2 2 ,2 4 and this is termed idiomor-
phic ferrite (see Fig. 1). The indications are that ferrite
idiomorphs do not have a Ž xed orientation relationship
with the matrix grains into which they grow.2 5
Growth at reconstructive transformation temperatures
tends to be controlled by substitutional element diffusion
away from the c/a interface at low undercooling and carbon
diffusion at high undercooling. Various growth modes
are recognised, in order of decreasing transformation
temperature:2 6
(i) local equilibrium with bulk partition of substitu-
tion alloying elements (PLE)
(ii) local equilibrium with negligible partition of sub-
stitutional alloying elements (NPLE)
(iii) paraequilibrium, where only the interstitial carbon
atoms diffuse.
The diffusion rate of carbon in austenite may be many
orders of magnitude greater than that of substitutional
atoms at reconstructive transformation temperatures. True
equilibrium segregation during phase transformations at
migrating interfaces is therefore unlikely to be achieved with
regard to all components. Growth under diffusion control
with local equilibrium at the interface is then envisaged.
Two phases may differ either signiŽ cantly (PLE) or
negligibly (NPLE) in terms of substitutional alloy content.
Element concentration or depletion spikes are invoked to
satisfy the thermodynamic constraints. In many cases, as
the transformation temperature is decreased, the relative
rates at which elements are able to diffuse negate the
assumption of local equilibrium, since the interface com-
position spike would be only several atomic layers thick. In
such cases, the concept of paraequilibrium is applied, i.e.
there is no redistribution of iron or substitution atoms at the
interface between the phases and only the interstitial carbon
atoms diffuse The different growth modes described above
may result in signiŽ cant changes in ferrite growth mor-
phology from equiax grains towards a plate shape (see
below).
Dubé et al.1 7 refer to prior austenite grain boundary
allotriomorphic ferrite as GBF. The IIW classiŽ cation
scheme refers to the Ž rst phase forming at reconstructive
transformation temperatures as primary ferrite, termed PF.
Prior austenite grain boundary primary ferrite allotrio-
morphs are termed PF(G) in the IIW classiŽ cation scheme
 Thewlis Classi®cation and quanti®cation of microstructures in steels
1: intragranular ferrite idiomorphs; 2: grain boundary ferrite
allotriomorphs
2 Morphologies of ferrite at prior austenite grain bound-
ary and intragranular sites in 0.06%C, 1.46%Mn sub-
merged arc weld metal, continuously cooled, iced
brine quenched from 670°C22
and are usually observed in the form of polygonal grains or
veins, as shown schematically in Fig. 1. Reference is made
in the IIW scheme to polygonal ferrite grains in the intra-
granular regions (see Fig. 1) of a size approximately three
times greater than those of the surrounding ferrite laths or
grains. These ferrite grains in reality may be cross-sections
of ferrite allotriomorphs that have grown from prior
austenite grain boundaries beneath the plane of observation
and have a wide range of sizes. They are termed PF(I) in the
IIW scheme. The present author and co-workers2 2 have
referred to the different forms of prior austenite grain
boundary primary ferrite as GB(PF), so that a distinction
may be made with idiomorphic primary ferrite as described
below.
In weld metals, stable particle dispersed steels and some
microalloyed steels, ferrite may nucleate not only at the
austenite grain boundaries but also on particles inside
the austenite grains2 2 ,2 7 (see Fig. 2). The author and co-
workers2 2 have termed these intragranular ferrite idio-
morphs I(PF). Depending on the temperature in the
reconstructive regime, the intragranular ferrite morpholo-
gies2 2 may take the form of blocks, loops, ellipses, rose
petals or wedges. The IIW classiŽ cation scheme does not
have a terminology for these primary ferrite idiomorphs.
Pearlite
Classically, pearlite transformation may occur at austenite
grain boundaries or an inhomogeneitysuch as an inclusion.2 3
Ferrite or cementite nucleation may initiate the pearlite
transformation depending on whether the steel is hypo- or
hyper-eutectoid in composition. Growth of a pearlite nodule
into an austenite grain proceeds with the formation of
alternate ferrite and cementite plates or lamellae. Both the
cementite and ferrite possess unique crystallographic orien-
tations within the pearlite nodule.2 3 Edgewise growth of the
plates may occur and also branching of the cementite
lamellae. The rate controlling process in the growth of
pearlite is the diffusion of carbon. As the transformation
temperature is lowered, the driving force for the reaction is
increased but the diffusivity of carbon is decreased so that the
pearlite interlamellar spacing is decreased.
At high transformation temperatures, pearlite is generally
observed as nodules of alternate ferrite and cementite
lamellae that may be quite coarse and degenerate. When
viewed in cross-section, the lamellae may appear as a
ferrite – carbide aggregate. As the transformation tempera-
ture is lowered, the lamellae become increasingly Ž ne
until the structure becomes irresolvable under the light
145
1: alternate ferrite/cementite lamellae; 2: ®ne ferrite ± carbide
aggregate; 3: irresolvable pearlite
3 Resolvable and irresolvable pearlite in 0.83%C,
0.50%Mn, as rolled rod
microscope (see Fig. 3). The pearlite may then have a light
etching response. Alternatively, the lamellae may become
subjected to distortion and bending, appearing as a dark
etching, ferrite – carbide aggregate or barely resolvable,
somewhat non-lamellar pearlite, often described in older
nomenclature as primary troostite.2 8 ,2 9
In the IIW scheme, FC(P) is used to describe lamellar
pearlite, degenerate or coarse pearlite, and Ž ne colony
or irresolvable pearlite. The term FC is used to describe
ferrite – carbide aggregate. At reconstructive transforma-
tion temperatures, large islands of pearlite or ferrite –
carbide aggregate may be interspersed with prior austenite
grain boundary primary ferrite PF(G). A similar situation
may occur with idiomorphic primary ferrite I(PF) (see
Fig. 4).2 7 In some cases pearlite may be present as micro-
phase (see below).
DISPLACIVE TRANSFORMATION REGIME
In the low temperature, displacive transformation regime, a
change from the austenite to ferrite crystal lattice occurs by
an invariant plane strain shape change with a large shear
component. Diffusion of interstitial carbon atoms may
accompany the shear transformation. For a purely dis-
placive transformation there is no movement of atoms
across the c/a interface. Reactions in the displacive trans-
formation regime tend to be rapid. The principal phases are
Widmanstätten ferrite, bainite and martensite.
1: idiomorphic ferrite; 2: ferrite ± carbide aggregate; 3: irresolva-
ble pearlite
4 Intragranular primary ferrite and pearlite in as cast,
0.13%C, 2.0%Mn, cerium sulphide particle dispersed
steel27
Materials Science and Technology February 2004 Vol. 20
146 Thewlis Classi®cation and quanti®cation of microstructures in steels
5 Primary and secondary Widmanstätten ferrite
WidmanstaÈtten ferrite
A classic feature of Widmanstätten ferrite formation is that
it may occur at relatively low undercooling.2 3 The growth
mechanism is thought to involve the simultaneous forma-
tion of pairs of mutually accommodating plates so that less
driving force is required for transformation than with
bainite or martensite.3 0 The ferrite plates grow rapidly with
a high aspect ratio (~10 : 1), resulting in parallel arrays.
Widmanstätten ferrite is not the result of a purely displacive
transformation but forms by a paraequilibrium mechan-
ism,3 0 ,3 1 involving the rapid diffusion of interstitial carbon
atoms across the advancing interface into the remaining
austenite during the shear transformation. At the relatively
low undercooling required for Widmanstätten ferrite for-
mation, microphases of retained austenite, martensite or
ferrite/carbide aggregate (pearlite) may be formed between
the growing ferrite plates.
Widmanstätten ferrite can easily be confused with bainite.
Dubé et al.1 7 describe both prior austenite grain boundary
Widmanstätten ferrite and bainite as ferrite sideplate FS but
reference is also made to intragranular plates IP. The IIW
classiŽ cation scheme refers to all forms of Widmanstätten
ferrite and bainite as ferrite with second phase FS, although
a distinction may be made in the terminology when
Widmanstätten ferrite can be positively identiŽ ed, e.g.
FS(SP).
Characteristically, primary Widmanstätten ferrite plates
grow directly from a prior austenite grain boundary, whereas
secondary Widmanstätten ferrite plates grow from allo-
trimorphic ferrite at the grain boundaries, as shown sche-
matically in Fig. 5. Primary Widmanstätten ferrite plates
may also grow from inclusions, while secondary Widman-
stätten ferrite plates grow from intragranular idiomorphic
ferrite.2 2 ,3 2
Widmanstätten ferrite that grows from prior austenite
grain boundary sites is usually seen as colonies of coarse
sideplates with aligned microphase (see Fig. 6), which are
termed FS(A) in the IIW scheme. The individual plates
within an array are separated by low angle boundaries that
are difŽ cult to resolve under the light microscope, although
careful specimen polishing and etching may reveal them.
Depending on the plane of observation, the microphases
may appear non-aligned. When viewing a cross-section of
ferrite laths that have grown from prior austenite grain
boundaries beneath the plane of observation, all that may
be seen are islands of microphase in a matrix of ferrite
within the prior austenite grains (see Fig. 6). The Widman-
stätten ferrite is then classiŽ ed as FS(NA). The present
author and co-workers2 2 have referred to the different
forms of prior austenite grain boundary Widmanstätten
ferrite as GB(WF) so that a distinction may be made with
intragranular Widmanstätten ferrite as described below.
In the intragranular regions of weld metals and in
some steels,2 2 ,2 7 ,3 2 multiple large plates (aspect ratio>
4 : 1) of Widmanstätten ferrite with aligned microphase
Materials Science and Technology February 2004 Vol. 20
1: idiomorphic ferrite; 2: prior austenite grain boundary Widman-
staÈtten ferrite with aligned microphase; 3: prior austenite grain
boundary WidmanstaÈtten ferrite with non-aligned microphase
6 Interlocking colonies of Widmanstätten ferrite in 0.05%C,
1.35%Mn, HSLA steel, submerged arc weld HAZ
may be observed that grow from inclusions (primary
Widmanstätten ferrite) or from idiomorphic ferrite
(secondary Widmanstätten ferrite) as shown in Fig. 7.
The IIW classiŽ cation scheme does not have a terminology
for these plates. However, they have been designated
intragranular ferrite sideplates FS(I) in recent work by
the present author.3 2 In many cases, individual plates may
be observed that have grown relatively unimpeded from
intragranular inclusions (see Fig. 8). These plates do not
have aligned microphase and may be interspersed with
bainite or martensite.2 2 ,2 7 ,3 2 The inclusions from which the
plates grow may not be viewed since they may be under the
plane of observation. These plates have been designated IFP
by the present author,3 2 who summed FS(I) and IFP to give
a total quantity of intragranular Widmanstätten ferrite,
referred to as I(WF). Where there is a high density of
inclusions, multiple hard impingements of individual
Widmanstätten ferrite plates growing from inclusions2 2 ,3 2
may produce a Ž ne interlocking structure (see schematic
diagram, Fig. 5). The IIW classiŽ cation scheme refers
generally to this type of structure as acicular ferrite AF
(see below).
Bainite
Bainite is generally recognised as forming at temperatures
where diffusion controlled transformations are sluggish and
has features in common with low temperature martensitic
7 Intragranular Widmanstätten ferrite sideplates in as
deposited, 0.08%C, 2.87%Mn, 0.35%Mo, 0.0027%B,
0.019%Ti, submerged arc weld metal:32 arrow indicates
multiple plates of Widmanstätten ferrite with aligned
microphase nucleated on large intragranular inclusions
 Thewlis Classi®cation and quanti®cation of microstructures in steels
1: idiomorphic ferrite; 2: individual plate of WidmanstaÈtten fer-
rite nucleated on large intragranular inclusions
8 Growth of intragranular Widmanstätten ferrite plates
in 0.06%C, 1.37%Mn, 0.17%Mo, 0.0028%B, 0.027%Ti
submerged arc weld metal, continuously cooled,
helium quenched from 620°C22
transformations.2 6 It grows as individual plates or sub-units
to form parallel arrays or sheaves. The growth of each sub-
unit is accompanied by an invariant plane strain shape
change with a large shear component. There is no
redistribution of iron or substitutional solute atoms at the
transformation interface. Classically, bainite has been cate-
gorised into two component structures, notably upper and
lower bainite, depending on the transformation tempera-
ture. Carbon partitions into the residual austenite in upper
bainite, and precipitates as cementite between the bainitic
ferrite plates. In lower bainite, the ferrite becomes super-
saturated with carbon and some carbide precipitation
occurs within the ferrite sub-units as well as between them.
The exact growth mechanism of bainite is still the subject
of much debate.3 3 – 3 5 A paraequilibrium mechanism in
upper bainite involving a shear transformation accompa-
nied by the rapid diffusion of interstitial carbon atoms
across the c/a interface would mean that bainitic growth was
in part similar to Widmanstätten ferrite. However, a purely
displacive transformation would require no redistribution
of atoms across the c/a interface. A temperature curve To
may be identiŽ ed on the Fe – C phase diagram deŽ ning
thermodynamically where austenite and ferrite of the same
composition have identical free energy.2 6 ,3 3 At the To
temperature there is no driving force for transformation.
The To curve has a negative slope with carbon concentra-
tion, lying between the Ae 1 and Ae 3 lines of the Fe – C phase
diagram. In a steel with a carbon concentration lower than
that deŽ ned by the To curve, bainitic ferrite plates may
begin to grow without diffusion at an appropriate hold
temperature, then partition excess carbon into the residual
austenite. Further diffusionless growth of plates may take
place from the carbon enriched austenite, and the process
continues until such transformation becomes thermodyna-
mically impossible at the To curve. This is termed the
incomplete reaction phenomenon. Continuous undercool-
ing of the steel below To will cause the bainite reaction to be
maintained. Carbide precipitation occurs when the trans-
formation conditions are kinetically favourable. For a
purely displacive transformation, therefore, a rapid redis-
tribution of carbon atoms is envisaged after the diffusion-
less growth of bainitic ferrite sub-units.2 6
Bainite can easily be confused with Widmanstätten ferrite,
as noted above. Both structures are referred to as ferrite
with second phase, FS in the IIW classiŽ cation scheme,
although a distinction may be made in the terminology where
bainite can be clearly identiŽ ed, e.g. FS(B). A further dis-
tinction may be made where upper and lower bainite can be
positively identiŽ ed, e.g. FS(UB) and FS(LB) respectively.
147
9 Bainite sheaves and sub-units
Characteristically, bainite may grow directly from a prior
austenite grain boundary2 6 or an intragranular inclusion,3 6
as shown schematically in Fig. 9. Sympathetic nucleation of
bainite plates from existing sheaves is a common feature.
Bainite that grows from prior austenite grain boundaries
is commonly observed in the form of interlocking sheaves of
very Ž ne plates with aligned cementite particles (see
Fig. 10), which are designated FS(A) in the IIW scheme.
In upper bainite, FS(UB), carbide particles are observed
between the plates, while in lower bainite, FS(LB), the
carbides are within as well as between the plates and the
structure tends to have a darker etching response. The
individual plates within a sheaf are separated by low angle
boundaries that are virtually irresolvable under the light
microscope. The sheaves are shown in the process of growth
in Fig. 11. Extensive sympathetic nucleation is evident.
Depending on the plane of observation, cementite particles
may appear non-aligned. When viewing a cross-section of
ferrite laths that have grown from prior austenite grain
boundaries beneath the plane of observation, all that may
be seen are carbide particles in a matrix of ferrite within the
prior austenite grains (see Fig. 10). The bainite is then
classiŽ ed as FS(NA). The present author and co-workers2 2
have referred to the different forms of prior austenite grain
boundary bainitic ferrite as GB(B) so that a distinction may
be made with intragranular bainite as described below.
In some steels and weld metals,2 6 ,3 2 ,3 6 bainite sheaves may
be seen to grow from intragranular inclusions (see Fig. 12).
Individual Ž ne plates of bainitic ferrite may also be
observed that grow relatively unimpeded from intragranu-
lar inclusions (see Fig. 13). The latter plates do not have
aligned carbide particles and may be difŽ cult to distinguish
from Widmanstätten ferrite plates IFP (see above). The
inclusions from which the plates grow may not be observed
1: lower bainite with carbide particles between as well
as within subunits; 2: upper bainite with aligned carbide;
3: bainitic ferrite with non-aligned carbide
10 Interlocking sheaves of upper and lower bainite in
0.17%C, 1.0%Mn steel, laser weld HAZ
Materials Science and Technology February 2004 Vol. 20
148 Thewlis Classi®cation and quanti®cation of microstructures in steels
11 Growth of bainite sheaves and (arrowed) sympathetic
nucleation of laths in 0.38%C, 1.39%Mn, 0.039%S,
0.09%V steel, isothermally transformed, 45 s at 400°C
12 Growth of bainite sheaves from intragranular inclu-
sions in 0.38%C, 1.39%Mn, 0.039%S, 0.09%V,
0.013%N steel, isothermally transformed, 38 s at
450°C: arrow indicates multiple laths of bainite with
carbide particles between as well as within subunits
13 Growth of intragranular bainite plates in 0.38%C,
1.39%Mn, 0.039%S, 0.09%V, 0.013%N steel, isother-
mally transformed, 38 s at 500°C: arrows indicate
individual plates of bainitic ferrite nucleated on small
intragranular inclusions
since they are under the plane of observation. The IIW
classiŽ cation scheme does not have a terminology for the
different forms of intragranular bainite, but the author and
co-workers2 2 have termed them I(B). Where there is a high
density of inclusions, multiple hard impingements of
individual bainitic plates growing from the inclusions may
Materials Science and Technology February 2004 Vol. 20
14 Lath martensite in 0.13%C laser weld metal: arrow
indicates martensite laths with highly dislocated sub-
structure
15 Plate or twin martensite in 0.27%C laser weld metal:
arrow indicates lenticular martensite with twinned
substructure
result in a very Ž ne interlocking structure2 6 ,3 2 (see schematic
diagram, Fig. 9). The IIW classiŽ cation scheme refers
generally to this type of structure as acicular ferrite AF
(see below).
Martensite
Martensite is classically an extremely rapid, diffusionless
transformation where carbon is retained in solution.3 7 As
the austenite lattice changes from fcc to the required mar-
tensite bcc or bct lattice, strain energy considerations
dominate and the martensite is constrained to be in the form
of thin plates.
In low carbon steels (less than ~0.2%C) lath martensite
with a bcc crystal structure is the commonly occurring
form3 7 and is designated M or M(L) in the IIW scheme. The
martensite units are formed in the shape of laths that
are grouped into larger sheaves or packets (see Fig. 14).
The sub-structure consists of a high density of dislocations
arranged in cells; each martensite lath is composed of many
dislocation cells. As the steel carbon content increases signi-
Ž cantly above about 0.2%C, plate martensite tends to form
with either a bct or bcc crystal structure.3 7 The martensite
units form as individual lenticular plates (see Fig. 15) with a
substructure consisting of very Ž ne twins. This form of
martensite is termed twinned martensite in the IIW scheme
and is designated M or M(T). Martensite, whether in plates
or lath form, is generally irresolvable under the light micro-
scope and tends to have a slow etching response.
 Thewlis Classi®cation and quanti®cation of microstructures in steels
16 Nature of acicular ferrite
Acicular ferrite
Conventionally,2 6 acicular ferrite is recognised as an intra-
granular nucleated morphology of ferrite in which there are
multiple impingements between grains. The acicular ferrite
nucleates on inclusions inside the prior austenite grains
during the cda transformation. Provided there is a high
density of inclusions, a Ž ne interlocking structure (generally
<5 mm) can be produced.
In the IIW scheme, acicular ferrite is designated AF. For
a long time acicular ferrite was thought to be a single
transformation product. Early work3 8 suggested that it
was intragranularly nucleated Widmanstätten ferrite. Later
research2 6 provided evidence for intragranularly nucleated
bainite. However, recent research by the author and co-
workers2 2 has demonstrated that the nature of acicular
ferrite may be as shown schematically in Fig. 16. Different
reaction products may nucleate on intragranular inclusions
at reconstructive and displacive transformation tempera-
a idiomorphic ferrite (arrowed) nucleated on large inclusions;
b WidmanstaÈtten ferrite plates (arrowed) nucleated on small
inclusions
17 Acicular ferrite development in 0.06%C, 1.37%Mn,
0.17%Mo, 0.0028%B, 0.027%Ti submerged arc weld
metal, continuously cooled, iced brine quenched from
615°C22
149
tures during continuous cooling depending on the nature,
size and amount of inclusions (see Figs. 2 and 17). Acicular
ferrite results from multiple hard impingements of the
different transformation products. The sequence of
transformations is consistent with the theoretical activation
energy barrier to nucleation of the different sites. Acicular
ferrite development may thus be deŽ ned in terms of con-
ventional steel transformation products and CCT diagrams
incorporating both intragranular and grain boundary
transformations.
Under continuous cooling transformation conditions
AF~I(PF)zI(WF)zI(B)
This leads to acicular ferrite that may have a variety of
forms depending on steel composition, cooling rate and
inclusion characteristics. Acicular ferrite may consist of
mixtures of different intragranular transformation products
(see Fig. 18).2 2 ,3 2 Alternatively, Widmanstätten acicular
ferrite or bainitic acicular ferrite may form per se.2 6 ,3 8
However, if reactions are completed at purely reconstruc-
tive transformation temperatures, it may be preferable to
use the term idiomorphic primary ferrite instead of acicular
ferrite to describe the microstructure, since intragranular
primary ferrite is likely to be coarse and non-acicular in
morphology (see Fig. 4).
Acicular ferrite is usually observed as a Ž ne interlocking
ferrite structure interspersed with microphases (see Fig. 18).
The shape of the ferrite plates may not appear to be needle-
like as the use of the term ‘acicular’ would imply. This is
because the different ferrite morphologies cannot grow very
far before mutual hard impingement. It is evident from
Fig. 18 that the degree of reŽ nement of the acicular ferrite is
dependent on the nature of the transformation products
inherent in its formation.
b
a intragranular primary ferrite ± WidmanstaÈtten ferrite in C ± Mn
weld metal;22 b intragranular WidmanstaÈtten ferrite ± bainite
in Ti ± Mo ± B alloyed weld metal32
18 Forms of acicular ferrite
Materials Science and Technology February 2004 Vol. 20
150 Thewlis Classi®cation and quanti®cation of microstructures in steels
Microphases
The different ferrite growth modes of the principal struc-
tures described above result in carbon enrichment of the
remaining austenite, leading to associated second phases of
retained austenite, martensite, bainite or ferrite – carbide
aggregate (pearlite), depending on the degree of carbon
enrichment of the austenite and the prevailing cooling
conditions. The second phases associated with Widman-
stätten ferrite and acicular ferrite are generally quite small
(2 – 5 mm), and are termed microphases.
IIW classiŽ cation scheme problem areas
and solutions
The objective in the present work was to investigate the IIW
microstructure classiŽ cation scheme for weld metals as a
basis for quantifying the full range of microstructures found
in plain carbon and low alloy steels, as well as ferritic weld
metals and parent plate heat affected zones. A means may
thus be provided of obtaining database information for
developing microstructure– property relationships, or gen-
erating data for calibrating physical models that have the
principal structures primary ferrite, pearlite, Widmanstät-
ten ferrite, bainite and martensite as output.
It is clear from the above review that while the IIW
scheme provides a sound structure for quantifying complex
microstructures in steels, the classiŽ cation of constituents
such as ferrite sideplate and acicular ferrite is incompatible
with the principal structures found in the reconstructive and
displacive transformation regimes of ferrous materials.
Knowledge of the actual transformation products consti-
tuting ferrite sideplate and acicular ferrite structures is
required. ClassiŽ cation is also needed of idiomorphic ferrite
and ferrite sideplate structures growing relatively unim-
peded from intragranular inclusions.
Problems that may be encountered in relating sub-
category microstructural components to principal struc-
tures at prior austenite grain boundary and intragranular
sites are discussed below together with possible solutions.
The ways in which transformation products associated with
ferrite sideplate and acicular ferrite structures may be
identiŽ ed will be addressed. The use of optical microscopy
with specimens polished to a 0.25 mm Ž nish and etched in
2% nital is assumed as standard. However, instances will be
given where different instruments and techniques may be
needed to solve problems. Where possible, the effects of
steel composition and heat treatment will be highlighted,
but detailed examples are outside the scope of the present
paper.
PRIMARY FERRITE
In low alloy weld metals, care has to be taken in identifying
primary ferrite due to stereological effects. Ferrite allo-
triomorphs growing from prior austenite grain boundaries
beneath the plane of observation may appear as polygonal
ferrite grains in the intragranular regions (see Fig. 1). If
these ferrite allotriomorphs are of a size approximately
three times greater than those of surrounding acicular
ferrite laths or grains, it is likely that they are the constituent
PF(I) described in the IIW scheme. It is unlikely that such
large grains are idiomorphic ferrite, I(PF), nucleated on
inclusions as referenced in the literature,2 2 since the latter
tend to nucleate at lower temperatures with relatively little
time for growth (see Fig. 2).
PEARLITE
Problems may arise in classifying pearlite when it is present
along with displacive transformation products.
Materials Science and Technology February 2004 Vol. 20
Lamellar pearlite, FC(P) in the IIW classiŽ cation
scheme, may be confused with martensite if the ferrite/
cementite plates are irresolvable under the light microscope.
A distinguishing feature is the generally rapid etching
response and lower hardness of the pearlite.
The dark etching, non-lamellar pearlite known as ferrite –
carbide aggregate, FC in the IIW classiŽ cation scheme, may
sometimes be confused with bainite. The nodular appear-
ance of pearlite as opposed to the sheaf appearance of
bainite may provide a distinguishing feature. The carbon
content of the steel may also give an indication as to how
much pearlite may be expected; high volume fractions
should not be present in low carbon steels. Ultimately,
however, knowledge of the thermal history and transforma-
tion conditions of the steel may be needed to provide a check
on classiŽ cation (see below). The reconstructive pearlite
transformation should take place slowly at high tempera-
tures and over a wide temperature range. A displacive
transformation to bainite should take place rapidly at lower
temperatures and over a relatively small temperature range.
It is notable that in bainitic steels, prolonged holding at a
given temperature may result in the incomplete reaction
phenomenon (see above). Continued isothermal treatment
can result in pearlite formation from the remaining carbon
enriched austenite.2 6
DifŽ culties in identiŽ cation of pearlite may be com-
poundedby a eutectoid transformationthat has been noted in
continuously cooled plain carbon steel (0.11%C, 0.5%Mn).
This involves ferrite growing in conjunction with repeated
nucleation of alloy carbides on the moving c/a interphase
boundary.3 9 The reaction has been termed interphase pre-
cipitation of cementite. Dark etching equiaxed ferrite grains
containing a Ž ne dispersion of carbides are observed under
the light microscope while, under the transmission electron
microscope, the cementite is seen in sheets.
FERRITE SIDEPLATE
Bainite and Widmanstätten ferrite may be present in
signiŽ cant amounts in heat treated steels and the coarse
grained HAZ of welds but they are difŽ cult to classify
individually so that both structures have been generally
referred to as ferrite sideplate.
WidmanstaÈtten ferrite
ClassiŽ cation of Widmanstätten ferrite can prove difŽ cult
because of its similarity to upper bainite, but certain
guidelines may be followed to avoid confusion.
The free energy requirement or driving force would be
expected to be lower for Widmanstätten ferrite formation
than for the upper bainite transformation, since the former
is thought to grow by the mutual accommodation of plates
and the latter by sub-units (see above). All else being equal,
therefore, Widmanstätten ferrite may be expected to occur
at higher temperatures than upper bainite and exhibit a
generally coarser structure with a lower dislocation density.
Furthermore,the microphasesbetween Widmanstätten ferrite
laths may be expected to be a mixture of pearlite, bainite,
martensite or retained austenite, whereas the nature of
bainite formation (see above) means that cementite particles
may generally be observed between the bainitic ferrite
plates.2 6 Microphases may be revealed by the use of dif-
ferent chemical etchants (see below).
The identiŽ cation of secondary Widmanstätten ferrite
with aligned microphase, FS(A) in the IIW scheme, is
relatively easy since it grows from existing allotriomorphic
ferrite, but care has to be taken in distinguishing the
boundary between the two structures. IdentiŽ cation of
primary Widmanstätten ferrite is signiŽ cantly more difŽ -
cult; it grows directly from prior austenite grain boundaries
and may be more easily confused with upper bainite. The
use of colour etching methods4 0 ,4 1 in conjunction with
 Thewlis Classi®cation and quanti®cation of microstructures in steels
optical microscopy may prove helpful in distinguishing
Widmanstätten ferrite from bainite. These techniquesinvolve
complex electrochemical reactions and require careful
experimentation, but can provide a means of distinguishing
various phases by their colouring response. Nanohardness
measurements may also prove useful; these are obtained
using a modiŽ ed scanning force microscope (SFM).4 2 The
nanoindentation technique allows very small regions of
grains to be investigated and different phases to be dis-
tinguished. All else being equal, Widmanstätten ferrite
should exhibit a lower hardness than bainite.
Although Widmanstätten ferrite may be distinguished
from upper bainite using the above guidelines, care has to
be taken with stereological effects. Widmanstätten ferrite
plates within a colony tend to grow in a common crystal-
lographic orientation. They are therefore generally sepa-
rated by low angle boundaries. When prior austenite grain
boundary Widmanstätten ferrite is seen end-on with non-
aligned microphase, FS(NA) in the IIW scheme, the plates
can give the appearance of ferrite grains interspersed with
microphase, thereby creating confusion with regions of
intragranular acicular ferrite, AF. In the case of acicular
ferrite, hard impingements of the different ferrite morpho-
logies growing from inclusions results in high angle
boundaries, which are signiŽ cantly more distinct than the
low angle boundaries of Widmanstätten ferrite. Careful
specimen polishing and etching may be required to dis-
tinguish the two structures.
In the intragranular regions of welds, it may be relatively
straightforward to identify multiple plates of Widmanstät-
ten ferrite with aligned microphase growing unimpeded
from large inclusions, described as FS(I) in the literature.3 2
Recognising single plates of Widmanstätten ferrite without
aligned microphase, designated IFP, may be more difŽ cult,
but these plates are likely to be quite coarse and grow from
large inclusions. Formation of the latter may appear con-
tradictory from a mechanistic viewpoint. It is possible that
the second plate is beneath the plane of observation (see
Fig. 8). Alternatively, the absence of aligned microphase
may be because, during plate growth, carbon is rejected into
the remaining austenite, which then undergoes a secondary
transformation at lower temperatures to bainite, martensite
or Ž ne acicular ferrite nucleated on small inclusions.
Bainite
The effects of steel composition may compound many of the
problems associated with distinguishing Widmanstätten
ferrite from upper bainite described above.
Low carbon content in bainitic steels can increase the
transformation temperature and result in a coarse lath size
so that bainitic ferrite with aligned second phase, FS(A) in
the IIW scheme, appears similar to Widmanstätten ferrite.
High silicon content in bainitic steels (generally >1%) can
retard the precipitation of carbide from austenite2 6 and
result in martensite or retained austenite microphases
between the bainitic ferrite laths, thereby creating confusion
with Widmanstätten ferrite. Granular bainite, which tends
to form in continuously cooled, low carbon bainitic steels,
poses a similar problem.2 6 This structure appears as a
relatively coarse aggregate of bainitic ferrite and retained
austenite or martensite islands; the bainitic sub-units have
very thin regions of austenite between them, which cannot
be resolved under the light microscope.2 6 Ultimately, high
resolution SEM, TEM or electron back-scattering diffrac-
tion (EBSD) techniques4 3 ,4 4 may be needed to distinguish
these forms of bainite from Widmanstätten ferrite by
revealing the crystallographic sub-structure and thereby the
mechanism of formation, but some electron metallographic
techniques are time consuming and often difŽ cult.
When trying to distinguish upper, FS(UB), and lower,
FS(LB), bainite in the IIW scheme, stereological effects may
151
cause confusion. Cross-sections of upper and lower bainite
sheaves may appear similar. In general, however, the carbides
are likely to be Ž ner and the etching response darker in the
lower bainite.
In weld metals, individual plates of bainitic ferrite, I(B),
growing unimpeded from intragranular inclusions may be
difŽ cult to separate from Widmanstätten ferrite plates, IFP.
However, the former are likely to be signiŽ cantly Ž ner than
the latter and the nucleating inclusions may be smaller.
Colour etching methods4 0 ,4 1 may be helpful for identiŽ ca-
tion but, ultimately, electron metallographic techniques may
be required to determine the nature of the plates.
MARTENSITE
Martensite is often present together with bainite in the HAZ
of laser welds and to some extent electron beam welds; these
phases also occur in high strength weld metals.3 2 Most low
carbon steels have martensite start temperatures above
room temperature so that, at slower cooling rates, carbon
atoms can redistribute and precipitate, i.e. autotempering
can take place. It is then difŽ cult to distinguish between
autotempered martensite, M, and lower bainite, FS(LB), in
the IIW scheme. The carbides precipitated inside the laths in
lower bainite are, however, likely to be coarser and some
interlath carbide should be evident (see above).
Colour etching methods4 0 ,4 1 may be investigatedas a means
of distinguishing between bainite and martensite. Com-
paratively simple nanohardness measurements4 2 may also
prove useful in separating martensite from other principal
structures,and in distinguishingthe different forms of marten-
site. Since carbon content generally governs the martensitic
hardness, twinned martensite, M(T), may be expected to
exhibit a much higher hardness than lath martensite, M(L).
ACICULAR FERRITE
Distinguishingthe intragranulartransformationproducts that
compose acicular ferrite, AF in the IIW scheme, is likely to be
very difŽ cult comparedwith identifying the structure itself. It is
recommended, therefore, that for the purposes of calibrating
models, a pragmatic solution be adopted. Thus measured
volume fractions of acicular ferrite should be compared with
the sum of the intragranularconstituents I(PF)zI(WF)zI(B)
predicted by modelling. However, care should be taken to
distinguish between acicular ferrite, AF, where multiple
impingementoccurs between the different intragranularferrite
morphologies, and the intragranular transformation products
I(PF), I(WF) and I(B), which may grow relatively unimpeded
and may be identiŽ ed in their own right.
MICROPHASES
Microphases are normally revealed using a standard etch
polish technique with a 2% nital etch. However, problems
may arise in distinguishing martensite and retained
austenite, which often occur together as MA phase. TEM
techniques may be employed to separate the phases but are
time consuming and difŽ cult. The proportion of austenite in
the MA phase may be determined using X-ray diffraction
techniques. In some cases, etching in picral can reveal the
nature of the microphases. Thus cementite may appear
black; a light brown coloration indicates lath martensite; a
yellow-brown colour is likely to be twin martensite while a
grey-white colour is indicative of retained austenite.
New classiŽ cation scheme
In the previous section, problems in the IIW microstructure
classiŽ cation scheme were discussed and guidelines pro-
posed for identifying the principal structures associated
Materials Science and Technology February 2004 Vol. 20
 152

Table 1 ClassiŽ cation scheme for microstructural constituents

Category terminology
Principal structure
classi®cation Overall Main Sub Component structure description Comments

Reconstructive transformations (diffusion controlled, with slow rates of reaction)

Ferrite PF* PF(GB) PF(G)* Grain boundary primary ferrite Ferrite veins or polygonal grains aligned with prior austenite grain boundaries
Allotriomorphic ferrite
Polygonal ferrite
Ferrite veins

Materials Science and Technology

PF(NA) Polygonal primary ferrite non- Polygonal ferrite grains within the prior austenite grains, of a size approximately three times greater
aligned than the surrounding ferrite laths or grains; cross-sections of ferrite allotriomorphs that have grown
from prior austenite grain boundaries below the plane of observation
PF(I) PF(I) Idiomorphic ferrite Ferrite idiomorphs associated with intragranular nucleation sites (large oxide/sulphide inclusions)
in weld metals and particle dispersed steels
Pearlite P* P* FC(P)* Lamellar pearlite Nodules of alternate ferrite/cementite lamellae, which are often dif®cult to resolve under the optical
Degenerate pearlite microscope. The structure has a rapid etching response in 2% nital and a generally low hardness.
Fine colony pearlite Pearlite may be present as a microphase
FC* Ferrite ± carbide aggregate Pearlite lamellae viewed in cross-section. Distorted pearlite lamellae may appear as a dark etching
virtually irresolvable ferrite/carbide aggregate known as primary troostite. Dif®cult to distinguish
ferrite/carbide aggregate from bainite
Displacive transformations (shear dominated, with rapid rates of reaction)
WidmanstaÈtten ferrite WF WF(GB) FS(A)* WidmanstaÈtten ferrite with Colonies of parallel ferrite laths (or sideplates) with microphases aligned between the laths ranging

February 2004 Vol. 20

aligned microphase
WidmanstaÈtten ferrite sideplates
FS(NA)* WidmanstaÈtten ferrite with
non-aligned microphase
WF(I) FS(I) Intragranular WidmanstaÈ tten
from pearlite to martensite. Lath boundaries are dif®cult to resolve. Primary WidmanstaÈtten ferrite
grows from the prior austenite grain boundaries, whereas secondary WidmanstaÈ tten ferrite
grows from allotriomorphic ferrite at the boundary
Aggregate of microphase islands and WidmanstaÈtten ferrite within the prior austenite grains;
cross-sections of WidmanstaÈtten ferrite sideplates that grow from prior austenite grain boundaries below
the plane of observation
Multiple coarse WidmanstaÈtten ferrite plates (aspect ratio greater than 4 : 1) with aligned microphases,
Thewlis Classi®cation and quanti®cation of microstructures in steels

ferrite sideplates which grow from intragranular inclusions. Primary intragranular ferrite sideplates grow from
inclusions, whereas secondary sideplates grow from ferrite idiomorphs associated with inclusions
FP(I) Intragranular WidmanstaÈ tten Individual coarse plates of WidmanstaÈ tten ferrite that grow relatively unimpeded from intragranular
ferrite plates inclusions
AF* WidmanstaÈtten acicular ferrite Fine interlocking structure formed by multiple impingements of individual WidmanstaÈ tten ferrite
plates growing from intragranular inclusions
Bainite B B(GB) FS(A)* Bainitic ferrite with aligned carbide Sheaves of parallel ferrite laths (or sub-units) with cementite particles aligned between the laths. Lath
Bainite sheaves boundaries are generally irresolvable under the light microscope. Sheaves grow from prior austenite
grain boundaries; sympathetic nucleation of laths from existing sheaves is a common feature
FS(NA)* Bainitic ferrite with non-aligned Aggregate of coarse carbides and bainitic ferrite within the prior austenite grains; cross-sections of
carbide bainite sheaves that grow from prior austenite grain boundaries (or existing sheaves) below the plane
of observation
FS(UB)* Upper Bainite Carbide particles are precipitated between the bainite sub-units. Upper bainite has a higher dislocation
density than primary WidmanstaÈtten ferrite. Bainite may appear as a microphase between WidmanstaÈ tten
ferrite sideplates
FS(LB)* Lower bainite Fine cementite particles precipitated within as well as between bainitic ferrite plates. Lower bainite
has a generally darker etching response than upper bainite. Dif®cult to distinguish lower bainite from
autotempered martensite
 Thewlis Classi®cation and quanti®cation of microstructures in steels 153

with prior austenite grain boundary and intragranular sites,

taking into account stereological effects. In this section, the
information gained has been used to develop a new classi-
Ž cation scheme. The application and accuracy of the new

Low carbon martensite with a lath structure and heavily dislocated sub-structure. Lath martensite has

within the prior austenite grains. Smaller colonies may be treated as microphases. Microphases may
a slow etching response in 2% nital and a generally high hardness. Colonies of martensite may form
Sheaves of ®ne bainitic ferrite plates with aligned carbide, which grow from intragranular inclusions
Individual ®ne plates of bainitic ferrite that grow relatively unimpeded from intragranular inclusions.
Very ®ne interlocking structure formed by multiple impingements of individual bainitic ferrite plates
scheme have been addressed, and consideration given to its
evolution.

DEFINITION
Using the information gained above, the traditional IIW
classiŽ cation scheme has been modiŽ ed and new termino-
logy deŽ ned as in Table 1. The main and sub-categories of
microstructural constituents of the table re ect the mechan-

High carbon martensite with a plate structure and twinned sub-structure

isms of formation of the principal structures and the
characteristic ferrite morphologies produced in the recon-
structive and displacive transformation regimes of steels.
Traditionally, the IIW classiŽ cation scheme terminology
places the transformation product Ž rst and the location
second, whereas the reverse is often the case in the wider
published literature.1 7 ,2 2 ,3 2 For consistency, therefore, the
terminology described in Table 1 follows the traditional
consist of martensite with retained austenite (MA)

IIW notation. Thus, the constituents GB(PF), I(PF),

GB(WF), I(WF), GB(B), I(B) described in the literature2 2
are replaced by PF(GB), PF(I), WF(GB), WF(I), B(GB),
B(I) as main category terms in Table 1. Likewise, the
growing from intragranular inclusions

constituent IFP in the literature3 2 is replaced by the sub-

category constituent FP(I) in Table 1.
To avoid con ict in Table 1 between the terminology
adopted for idiomorphic primary ferrite PF(I) and that for
cross-sections of ferrite allotriomorphs growing from prior
austenite grain boundaries below the plane of observation,
the latter terminology has been changed from PF(I) to
PF(NA), i.e. primary ferrite not aligned with prior austenite
grain boundaries. PF(NA) may be added together with
Comments

PF(G) to give an overall quantity of reconstructive prior

austenite grain boundary nucleated ferrite PF(GB).
It should be noted in Table 1 that the new sub-category
component terminology automatically deŽ nes its location
either at prior austenite grain boundaries or in intragranular
regions. In practice, therefore, an identiŽ cation system may
be employed which directly links a sub-category component
Component structure description

to the principal structure, e.g. B-FS(A) and WF-FS(A).

Intragranular bainite sheaves

Flow charts that incorporate the classiŽ cation and

Intragranular bainite plates

terminology of Table 1 but provide detailed guidance on

Bainitic acicular ferrite

identifying principal structures are shown in Fig. 19. The

key to the  ow charts is given in Fig. 20. Separate charts are
Twin martensite

provided for austenite grain boundary and intragranular

Lath martensite

microstructural components. Progression through the charts

from sub-category component structures to the principal
structures is dependent on answering a number of boxed
questions on a yes/no basis. The questions are derived from
the considerations made in this paper. If the answer to a
question is ‘yes’, progression is made to the right of the
chart towards the principal structure. If the answer is ‘no’
M(T)*
M(L)*

a move vertically downwards is needed to obtain more

FS(I)
FP(I)
Sub

AF*

information before, eventually, progress is made to the right

again. The  ow charts thus potentially provide a means of
Category terminology

quantifying complex steel microstructures in terms of the

Main

principal structures, thereby enabling the generation of either

B(I)

M*

database information or data for calibration of theoretical

models.
Overall

M*

*Retained IIW terminology.

APPLICATION
(Continued )

To assess the accuracy of the new classiŽ cation scheme and

identify discrepancies between operators, exercises were
Principal structure

carried out to quantify widely different microstructures.

The microstructures were obtained by thermally cycling
classi®cation

steels of compositions 0.051 – 0.17%C, 0.51 – 1.46%Mn in a

Martensite
Table 1

dilatometer to peak temperatures of 900 – 1300°C and

cooling at rates between 2 and 200 K s2 1 . Full details of the
quantiŽ cation exercises, including a complete statistical

Materials Science and Technology February 2004 Vol. 20

154 Thewlis Classi®cation and quanti®cation of microstructures in steels

a prior austenite grain boundary constituents; b intragranular constituents

19 Guidelines and terminology for identiŽ cation of principal structures

Materials Science and Technology February 2004 Vol. 20

 Thewlis Classi®cation and quanti®cation of microstructures in steels
20 Key to  ow charts
analysis, are outside the scope of the present paper. However,
the results for selected steels are summarised below.
Six dilatometer sample microstructures covering a wide
transformation temperature range were photographed using
an appropriate magniŽ cation. The resulting microstructural
Ž elds are shown in Fig. 21. A mesh grid inscribed on trans-
parent acetate paper was overlaid in a Ž xed position on the
photographs so that those microstructural constituents
under or just touching the grid cross-lines could be quanti-
Ž ed. Each cross-line was identiŽ ed from the grid scale, e.g.
A1, A2, A3, …, B1, B2, B3, … A total of 500 points was
counted of each Ž eld. Because the grid points were Ž xed,
results from different operators could be compared and the
constituents that were most difŽ cult to quantify could be
relatively easily identiŽ ed.
Initially, a single operator was employed to point count
the volume percentages of microstructural constituents in
the six microstructural Ž elds using the traditional IIW
and the new classiŽ cation schemes. The results (Table 2)
demonstrate the advantages of the new scheme in being able
to rationalise the principal structures associated with ferrite
sideplate. Ultimately the microstructural output is reduced
to the Ž ve principal constituents.
Following the above exercise, different operators were
employed to determine the volume percentages of the prin-
cipal structures in the six microstructural Ž elds using the
new scheme per se. The results are shown in the form of
histograms in Fig. 22. Most operators chose to identify the
major transformation products directly, although some
operators chose to classify subcategories and thereby the
major components. In all cases, microphases associated
with primary ferrite and Widmanstätten ferrite were treated
separately, while bainitic ferrite was quantiŽ ed together
with the carbide. Because of the Ž xed position of the point
counting grid, the variations in phase proportions in Fig. 22
155
are due to differences in microstructural interpretation by
the individual operators, rather than point counting errors
that would emerge between operators from random reposi-
tioning of the grid in the dilatometer sample microstruc-
ture. When quantifying the volume fraction of secondary
Widmanstätten ferrite, some discrepancy occurred between
operators owing to the need to distinguish the boundary
between allotriomorphic ferrite and Widmanstätten ferrite
(see Fig. 22a). Further differences occurred because of
the need to distinguish between ferrite carbide aggregate
(pearlite) and bainite (see Fig. 22b and c), and to some
extent lower bainite and autotempered martensite (see
Figs. 22d and f). These difŽ culties were compounded by the
low resolution of the photographic images.
A signiŽ cant improvement in the consistency between
operators was achieved, after appropriate training, when
quantifying phase proportions randomly over a relatively
large area in actual steel samples. In this case different
magniŽ cations could be used to reveal difŽ cult features. A
light microscope with a Swift point counting stage was
employed to count 500 points of various dilatometer sample
microstructures, again covering a wide transformation tem-
perature range. The statistical errors in point counting4 5 – 4 7
were determined using the formula according to Gladman
and Woodhead4 7
svf =Vf ~‰(1{Vf )=Pa Š1=2
where sv f is one standard deviation, Pa the fraction of
counts in the a phase and Vf the volume fraction of a phase.
The phase proportions obtained by two operators on six
steels are shown in Fig. 23. The 95% conŽ dence limits (2sv f )
are superimposed. The results show that the phase pro-
portions obtained by the individual operators were in many
cases within the statistical error deŽ ned in the point count-
ing exercise. However, to obtain a sensible statistical analysis
Materials Science and Technology February 2004 Vol. 20
156 Thewlis Classi®cation and quanti®cation of microstructures in steels

a b

c d

e f

a 0. 051%C, 0. 51%Mn, 1200°C, 10 K s2 1; b 0.17%C, 0.52%Mn, 1300°C, 10 K s2 1; c 0.13%C, 1.02%Mn, 1300°C, 2 K s2 1; d 0.13%C,
1.02%Mn, 1200°C, 10 K s2 1; e 0. 13%C, 1.02%Mn, 1300°C, 50 K s2 1; f 0.13%C, 1.02%Mn, 1300°C, 200 K s2 1
21 Microstructural Ž elds of steels thermally cycled in dilatometer to temperatures of 1200 or 1300°C and cooled at
rates between 2 and 200 K s2 1 (800 – 500°C)

of operator bias, a larger number of operators is needed.
Further work is required in the form of ‘round robin’
exercises to determine the statistical uncertainty between
operators when quantifying different types of microstruc-
ture, and to provide appropriate training measures for
widespread dissemination of the scheme.
The above studies were carried out without prior know-
ledge of the thermal history of the specimens examined.
However, transformation behaviour knowledge can provide
a useful check on results. The six microstructural Ž elds in
Fig. 21 were largely representative of the parent dilatometer
sample microstructures. The corresponding dilation curves,
percentage transformed versus temperature graphs and
peak rate transformation curves are shown in Fig. 24. The
dilatometer data in Fig. 24a show that for this particular
steel, transformation began at 793°C and took place over a
wide temperature range, Ž nishing at 628°C. As the trans-
formation proceeded, the rate of transformation increased
slowly to a peak at 715°C and then decreased slowly,
indicative of transformation controlled by diffusion. This
supports the operator classiŽ cation for the steel of about
70% primary ferrite and 5% pearlite, i.e. predominantly
reconstructive transformation (see Fig. 22a). By contrast,
the dilatometer data in Fig. 24c show that for this steel

Materials Science and Technology February 2004 Vol. 20

 Thewlis Classi®cation and quanti®cation of microstructures in steels 157

Table 2 Volume percentages of microstructural constituents obtained by single operator point counting microstructural
Ž elds (see Fig. 21) using traditional IIW and new classiŽ cation schemes
New scheme IIW scheme

Principal structure Phase, % Component structure Phase, %

0.051%C, 0.51%Mn, 1200°C, 10 K s2 1 (Fig. 21a)

PF 72.6 PF(G) 72.6
PF(NA) 0
P 6.4 FC(P) 0.7
FC 5.7
WF 21.0 FS(A)zFS(NA) 10.8z10.2
B 0
M 0 M 0
0.17%C, 0.52%Mn, 1300°C, 10 K s 1 (Fig. 21b)
2

PF 19.7 PF(G) 19.7

PF(NA) 0
P 54.7 FC(P) 2.4
FC 52.3
WF 25.6 FS(A)zFS(NA) 17.7z7.9
B 0
M 0 M 0
0.13%C, 1.02%Mn, 1300°C, 2 K s2 1 (Fig. 21c)
PF 28.4 PF(G) 28.4
PF(NA) 0
P 9.4 FC(P) 1.6
FC 7.8
WF 36.4 FS(A)zFS(NA)zFS(LB)zFS(I) 34.1z20.9z1.7z0.3
B 20 .6
M 5.2 M 5.2
0.13%C, 1.02%Mn, 1200°C, 10 K s2 1 (Fig. 21d)
PF 14.7 PF(G) 14.7
PF(NA) 0
P 2.6 FC(P) 2.6
FC 0
WF 55 .5 FS(A)zFS(NA)zFS(LB) 30.2z31.2z16.7
B 22 .6
M 4.6 M 4.6
0.13%C, 1.02%Mn, 1300°C, 50 K s2 1 (Fig. 21e)
PF 0 PF(G) 0
PF(NA) 0
P 0 FC(P) 0
FC 0
WF 0.5 FS(A)zFS(NA)zFS(LB) 25.1z14.6z43.2
B 82.4
M 17.1 M 17.1
0.13%C, 1.02%Mn, 1300°C, 200 K s2 1 (Fig. 21f)
PF 0 PF(G) 0
PF(NA) 0
P 0 FC(P) 0
FC 0
WF 0 FS(A)zFS(NA)zFS(LB) 0z0z0.3
B 0.3
M 99.7 M 99.7

transformation began at 717°C and Ž nished at 582°C. The
rate of transformation increased slowly at Ž rst, indicative of
reconstructive transformation, but then rose very rapidly to
a sharp peak at 646°C before falling steeply and then more
slowly in the Ž nal stages of transformation. The sharp peak
in the rate of transformation took place after around 40% of
reaction had occurred and was indicative of the beginning
of shear dominated transformation, which should account
for the remaining 60% of the transformation. The high
peak rate transformation temperature is indicative of
Widmanstätten ferrite formation rather than bainite. This is
broadly in agreement with the steel microstructure results in
Fig. 22c, where around 60% Widmanstätten ferrite was
quantiŽ ed by most of the operators. A similar analysis may
be carried out with respect to dilatometer data in Fig. 24d. In
this steel, the lower peak rate transformation temperature
(603°C) and lower Ž nishing temperature (532°C) are indica-
tive of bainite as well as Widmanstätten ferrite formation.
This again is re ected in the operator microstructure
classiŽ cation in Fig. 22d. It is notable that in the dilatometer
data of Fig. 24f, almost 50% of the steel transformation
occurred at one temperature (421°C). This extremely rapid
reaction rate and low transformation temperature are
indicative of martensite transformation, in agreement with
the operator classiŽ cation for the steel in Fig. 22f.
Overall, the above exercises show that a reasonable
degree of consistency may be obtained between operators
when using the new classiŽ cation scheme to identify the
principal structures (primary ferrite, pearlite, martensite)
and the transformation products constituting ferrite side-
plate structures, notably Widmanstätten ferrite and bainite.
EVOLUTION
The new classiŽ cation scheme deŽ ned above has attempted to
place knowledge of the classiŽ cation and quantiŽ cation of
steel microstructures on a Ž rm contemporarybasis. However,
it is of interest to consider possible future developments.
The guidelines proposed for phase recognition in the new
scheme are based on the mechanisms of formation of prin-
cipal structures, but there are still questions to be addressed
with respect to the kinetics of reactions, notably clariŽ ca-
tion of the growth mechanism of bainite. Improved know-
ledge in this area should result in greater accuracy in
distinguishing bainite from other phases. Overall, a better
understanding is needed of the dynamics of phase trans-
formations under continuous cooling transformation con-
ditions, where phases may form simultaneously and local
 uctuations in transformation conditions can make it
difŽ cult to recognise the transition between one phase

Materials Science and Technology February 2004 Vol. 20

158 Thewlis Classi®cation and quanti®cation of microstructures in steels
22 Volume percentages of principal structures obtained
by different operators point counting microstructural
Ž elds (see Fig. 21) using new classiŽ cation scheme:
PF~primary ferrite, P~pearlite, WF~Widmanstätten
ferrite, B~bainite, M~martensite
and another. In this respect an atlas of optical micrographs
with associated heat treatments and phase proportions
would be a useful accompaniment to the new classiŽ cation
scheme. Scanning electron images, with their greater
resolution, may be employed to describe local features. It
should be noted that a compendium of weld metal micro-
structures exists to accompany the traditional IIW classi-
Ž cation scheme.1 9
The classiŽ cation and quantiŽ cation of complex steel
microstructures by metallographic techniques is by nature
labour intensiveand it is appropriateto question the extent to
which computers may carry out such activities. There has
been a signiŽ cant amount of work done with regard to image
processing in recent years, driven by the advances in com-
puter technology.4 8 The appropriate processing steps depend
on the type of information required. The measurement of
images generally requires that features be well deŽ ned, by
edges, size, or unique brightness and colour. Image analysis
then attempts to Ž nd numeric descriptive parameters that
succinctly represent the information of importance in the
image. The new classiŽ cation scheme developed in the current
work provides guidelines on the important features for
phase recognition. It may thus be possible to train an image
analysis system to recognise these features. The question as
to how such information can be processed and analysed by
computeris a matter for further research.However, continued
rapid advances in computer power and image resolution may
make this type of activity tractable in the not too distant future.
Summary and conclusions
The International Institute of Welding (IIW) microstructure
classiŽ cation scheme for weld metals has been investigated
Materials Science and Technology February 2004 Vol. 20
a 0. 051%C, 0.51%Mn, 1200°C, 10 K s2 1; b 0.17%C, 0. 52%Mn,
1300°C, 10 K s2 1; c 0.13%C, 1. 02%Mn, 1300°C, 2 K s2 1; d
0.13%C, 1. 02%Mn, 1200°C, 10 K s2 1; e 0 .13%C, 1. 02%Mn,
1300°C, 50 K s2 1; f 0.13%C, 1.02%Mn, 1300°C, 200 K s2 1
23 Volume percentages of principal structures obtained
by two different operators point counting microstruc-
ture of steels under light microscope using new classiŽ -
cation scheme: PF~primary ferrite, P~pearlite, WF~
Widmanstätten ferrite, B~bainite, M~martensite
as a basis for quantifying the full range of microstructures
found in plain carbon and low alloy steels, as well as ferritic
weld metals and parent plate heat affected zones. The
following conclusions have been drawn:
1. The IIW scheme provides a sound structure for quanti-
fying complex microstructures in steels, but the classiŽ ca-
tion of constituents such as ferrite sideplate and acicular
ferrite is incompatible with the principal structures found in
the reconstructive and displacive transformation regimes of
steels. There is no classiŽ cation in the IIW scheme of
idiomorphic ferrite and ferrite sideplate structures growing
relatively unimpeded from intragranular inclusions.
2. There are problems in relating sub-category micro-
structural constituents in the IIW scheme to principal struc-
tures at prior austenite grain boundary and intragranular
sites owing to stereological and morphological effects. These
have been discussed in detail and solutions proposed. The
ways in which transformation products associated with
ferrite sideplate and acicular ferrite structures may be
identiŽ ed have been deŽ ned.
3. A new classiŽ cation scheme has been formulated. The
microstructure classiŽ cation and terminology used in the
IIW scheme have been built upon and new terminology
incorporated into a table providing descriptions of the
principal structures and sub-category components. Flow
charts have been devised with guidelines for identifying the
principal structures.
4. The new classiŽ cation scheme has been used to
quantify microstructures covering a wide transformation
temperature range. A difference in interpretation between
 Thewlis Classi®cation and quanti®cation of microstructures in steels
24 Transformation data obtained from thermally cycled steels in Fig. 21
individual operators has been identiŽ ed by point counting
micrographsusing a Ž xed grid. Some discrepancy occurred in
identifying the boundarybetween allotriomorphicferrite and
Widmanstätten ferrite, distinguishing between ferrite – car-
bide aggregate (pearlite) and bainite, and differentiating
between lower bainite and autotempered martensite. With
appropriate training, phase proportions obtained by two
individual operators point counting steel microstructures at
random using the light microscope were in many cases within
the statistical error deŽ ned in the point counting exercise.
5. Overall, a reasonable degree of consistency can be
obtained between operators when using the new scheme to
identify and quantify the principal structures (primary
ferrite, pearlite, martensite) and the actual transformation
products constituting ferrite sideplate structures, notably
Widmanstätten ferrite and bainite. Further work is required
in the form of ‘round robin’ exercises to determine the
statistical uncertainty between operators when quantifying
different types of microstructure, and to identify appro-
priate training measures for widespread dissemination of
the scheme.
6. A means has been provided of obtaining database
information for developing microstructure– property rela-
tionships, or generating data for calibrating physical models
that have the principal structures as their output.
159
Acknowledgements
The author would like to thank Dr S. V. Parker, Dr N. A.
Whittaker, Dr P. L. Harrison, Dr C. Wildash, Dr J. Butler,
Dr S. A. Butler, Professor A. A. Howe and I. W. Martin of
Corus RD&T, for helpful discussions and suggestions. The
author is also grateful to Professor R. C. Thomson,
Loughborough University, and Dr D. J. Abson, TWI, for
helpful comments. Thanks are Ž nally extended to ECSC
partners at TWI (UK), CSM (Italy), CEIT (Spain) and
IRSID (France) for support under ECSC steel research
programme 7210.PR/245(F5.01/00).
References
1. b. donnay, j. c. jerman, v. leroy, u. lotter, r. grossterlinden
and h. pircher: Proc. Int. Conf. on ‘Modelling of metal
rolling processes’, London, UK, December 1996, London, The
Institute of Materials.
2. j. k. lee and h. n. han: in ‘Thermomechanical processing of
steels’, Vol. 1, 245 – 254; 2000, London, The Institute of
Materials.
3. a. j. trowsdale, k. randerson, p. f. morris, z. husain and
Materials Science and Technology February 2004 Vol. 20
160 Thewlis Classi®cation and quanti®cation of microstructures in steels
d. n. crowther: in ‘Thermomechanical processing of steels’,
Vol. 1, 332 – 341; 2000, London, The Institute of Materials.
4. s. v. parker: ‘Modelling of phase transformations in hot rolled
steels’, PhD thesis, University of Cambridge, UK, 1997.
5. h. k. d. h. bhadeshia and l. e. svensson: in ‘Mathematical
modelling of weld phenomena’, 109 – 174; 1993, London, The
Institute of Materials.
6. s. j. jones: Modelling inclusion potency and simultaneous
transformation kinetics in steels’, PhD thesis, University of
Cambridge, UK, 1996.
7. s. j. jones and h. k. d. h. bhadhesia: Acta. Metall., 1997, 45,
(7), 2911 – 2820.
8. k. ichikawa and h. k. d. h. bhadhesia: in ‘Mathematical
modelling of weld phenomena 4’, 302 – 320; 1998, London, The
Institute of Materials.
9. d. j. c. mackay: in ‘Mathematical modelling of weld phe-
nomena 3’, 359 – 389; 1997, London, The Institute of Materials.
10. r. c. reed: ‘The characterisation and modelling of multipass
steel weld heat affected zones’, PhD thesis, University of
Cambridge, UK, 1990.
11. k. e. easterling: in ‘Mathematical modelling of weld phe-
nomena’, 183 – 200; 1993, London, The Institute of Materials.
12. m. atkins: ‘Atlas of continuous cooling transformation
diagrams for engineering steels’; 1977, Swinden Laboratories,
Rotherham, British Steel Corporation (ISBN 0 9500451 44).
13. z. zhang and r. a. farrar: ‘An atlas of continuous cooling
transformation diagrams applicable to low carbon, low alloy
weld metals’; 1995, London, The Institute of Materials.
14. b. l. bramfitt and j. g. speer: Metall. Trans., 1990, 21A, 817 – 829.
15. y. ohmori, h. ohtsubo, y. c. jung, s. okaguchi and h otani:
Metall. Trans., 1994, 25A, 1981 – 1989.
16. u. lotter and h. p. hougardy: Prakt. Metallogr., 1992, 29, (3),
151 – 157.
17. c. a. dubÉ, h. i. aaronson and r. f. mehl: Rev. Métall., 1958,
55, 201.
18. h. i. aaronson: ‘Decomposition of austenite by diffusional
processes’, 389; 1960, Philadelphia, PA, AIME.
19. ‘Compendium of weld metal microstructures and properties’;
1985, Abington, Woodhead Publishing.
20. ‘ClassiŽ cation of microstructures in low carbon – low alloy
steel weld metal and terminology’, Committee of Welding
Metallurgy of Japan Welding Society, IIW Doc IX – 1282 – 83.
21. e. anelli and p. e. di nunzio: ‘ClassiŽ cation of microstructures
of low carbon steels: preparation of a set of standard
micrographs’, ECSC Agreement 7210 – EC/405 (94 – D3.02a),
CSM, Rome, Italy, June 1996.
22. g. thewlis, j. a. whiteman and d. j. senogles: Mater. Sci.
Technol., 1997, 13, (3), 257 – 274.
23. r. w. k. honeycombe and h. k. d. h. bhadeshia: ‘Steels –
microstructure and properties’, 2nd edn, 35; 1995, London,
Edward Arnold.
Materials Science and Technology February 2004 Vol. 20
24. k. m. wu, t. yokomizo and m. enomoto: ISIJ Int., 2002, 42,
1144 – 1149.
25. g. myamoto, t. furuhara and t. maki: CAMP ISIJ, 2001, 14,
1172.
26. h. k. d. h. bhadeshia: ‘Bainite in steels’, 1st edn; 1992, London,
The Institute of Materials.
27. g. thewlis: ‘Stable sulphide particle dispersed steel’, Interna-
tional Patent Application 0105218.2, Corus UK Ltd, Mar, 2000.
28. r. m. brick and a. phillips: ‘Structure and properties of alloys’,
2nd edn, 334 – 337; 1949, New York, McGraw-Hill.
29. a. g. guy: ‘Elements of physical metallurgy’, 2nd edn,
474 – 476; 1960, Reading, MA, Addison-Wesley.
30. h. k. d. h. bhadeshia: Acta. Metall., 1981, 29, 1117 – 1130.
31. j. w. christian: ‘Military transformations – an introductory
survey’, 1 – 19; 1965, London, The Iron and Steel Institute.
32. g. thewlis: Sci. Technol. Weld. Joining, 2000, 5, (6), 365 –
377.
33. h. k. d. h. bhadeshia and j. w. christian: Metall. Trans. A,
1990, 21A, 767 – 797.
34. h. k. d. h. bhadeshia: Mater. Sci. Eng. A, 1999, A273 – A275,
58 – 66.
35. subra suresh (ed.): Scr. Mater., 2002, 47, (3) (Viewpoint Set on
‘Bainite’).
36. madariaga, i. gutierrez and h. k. d. h. bhadeshia: Metall.
Trans. A, Sept., 2001, 32A, 2187.
37. g. r. speich and w. c. leslie: Metall. Trans., 1972, 3, 1043 –
1054.
38. r. a. ricks, p. r. howell and g. s. barritte: J. Mater. Sci.,
1982, 17, 732.
39. a. t. davenport and p. c. becker: Mater. Trans., 1971, 2,
2962.
40. e. beraha and b. shpiglar: ‘Colour metallography’; 1977,
Metal Park, OH, American Society for Metals.
41. w. fin: ‘Basic principles for colour metallography’, 1983,
Beijing, Beijing Industry University.
42. p. maier, a. richter, r. g. faulkner and r. ries: Mater.
Charact., 2002, 48, 329 – 339.
43. i. m. watt: ‘The principles and practice of electron microscopy’,
2nd edn; 1997, Cambridge, Cambridge University Press.
44. a. j. schwartz, m. kumar and b. l. adams: ‘Electron
backscatter diffraction in materials science’; 2000, New York,
Kluwer/Plenum.
45. f. weinberg: ‘Tools and techniques in physical metallurgy’,
Vol. 1, 272 – 275; 1970, New York, Marcel Dekker.
46. b. pickering: ‘The basis of quantitative metallography’, 8 – 10;
1976, London, Metals and Metallurgy Trust for the Institute of
Metallurgical Technicians.
47. t. gladman and j. h. woodhead: J. Iron Steel Inst., 1960, 194,
189.
48. j. c. russ: ‘The image processing handbook’, 2nd edn; 1995,
Boca Raton, FL, CRC Press.