Scripta Materialia 47 (2002) 181–185
www.actamat-journals.com
Bainite: defect signatures
Gary R. Purdy *
Department of Materials Science and Engineering, McMaster University, 1280 Main Street West, Hamilton, Ont., Canada L8S 4L7
Abstract
The defect signatures of bainite and martensite in steels are reviewed, with the object of highlighting similarities and
differences between the two transformation products. It is concluded that the weight of evidence is strongly in favour of
different local interfacial migration modes for each of the two products of transformation.
2002 Acta Materialia Inc. Published by Elsevier Science Ltd. All rights reserved.
Keywords: Bainite; Martensite; Phase transformations
1. Introduction nite. Because of the latter effect, much, if not all, of
the most relevant experimental evidence deals with
This contribution will focus on crystal defects alloyed steels.
and their implications for mechanisms of growth As noted by Hillert [1], a number of authors
of bainite. Bainite is defined here as a dispersion of have considered the possibility that bainitic ferrite
carbide and ferrite resulting from the non-coop- forms by a mechanism identical to that for mar-
erative (or weakly cooperative) decomposition of tensite, and that the substitutional lattice trans-
ferrous austenite. Other contributors may choose formation is followed by carbon diffusion away
do deal with nucleation, or with the thermody- from the ferritic phase. This concept has been es-
namic and kinetic aspects of this reaction; the poused and enlarged upon by Bhadeshia and Ed-
former seems to this writer quite elusive, the latter monds [2] and Bhadeshia [3]. Others, notably
two promising but complicated by the necessary Aaronson et al. [4,5], have taken the view that the
extrapolations from higher-temperature thermo- interstitial carbon must be rejected from the fer-
dynamic and kinetic data. However, all of these rite as the bainitic ferrite/austenite interface ad-
are clearly valuable avenues of inquiry, each ca- vances, leaving in its wake a lower-carbon (but not
pable of adding to the total weight of evidence. necessarily a full equilibrium) product. The two
Since this contribution is focused on issues of opposing points-of-view have been summarized
structure, the main roles of alloying elements several times, for example in back-to-back contri-
would seem to be: alteration of the lattice param- butions to Phase Transformations’87 [5,6].
eter ratios of ferrite and austenite; suppression of That the debate has lasted for more than a half-
carbide precipitation; and stabilization of auste- century hinges in part on the microstructural sim-
ilarity of lath martensite and bainitic ferrite, and in
part on the fact that the host lattice structural
*
Fax: +1-905-528-9295. transformation from f.c.c. to b.c.c. (here termed
1359-6462/02/$ - see front matter 2002 Acta Materialia Inc. Published by Elsevier Science Ltd. All rights reserved.
PII: S 1 3 5 9 - 6 4 6 2 ( 0 2 ) 0 0 1 2 6 - 4
182 G.R. Purdy / Scripta Materialia 47 (2002) 181–185
the ‘‘iron transformation’’) can be decoupled, reactions in steels, the unambiguous identifica-
conceptually, from the interstitial diffusion of tion of all aspects of interfacial defects requires
carbon. If one were to suppose that the iron that the parent phase be retained, at least in the
transformation mechanism for bainite is identical interface region, and this usually requires a cer-
to that for martensite, the interface would have a tain degree of alloying of the material studied.
very high intrinsic mobility, and at least two dis- • Trailing defects: In some cases, defects are ob-
tinct transformation paths could be postulated. In served to be left in the product phase, and to
one, the iron structural transformation would be be connected to the interface. These can also
limited and paced by the rate of carbon diffusion yield useful information about the nature of the
away from the b.c.c. phase; in the other, the struc- migration process. They can be dislocations,
tural transformation would locally precede the twins, APBs etc.
rejection of carbon from the ferritic phase, as en- • Inherited defects: If defects exist in the parent
visaged by Bhadeshia [3]. phase, and if they are found to be inherited by
If on the other hand the local processes of in- the product phase, either in original or in
terfacial migration for martensite and for bainite altered form, they can also be brought into evi-
were different, such that the iron structural trans- dence concerning the mechanism of transforma-
formation for bainite is limited in part by the in- tion.
trinsic mobility of the interface (as believed to be
the case for Widmanst€ atten ferrite), then the pos-
sibility of a martensitic mechanism of bainite 1.1. Interfacial defects
growth could be excluded.
The defect signature of a phase transformation Interfacial defects intrinsic to the interface
can often provide important evidence about the structure may take the form of misfit-compensat-
nature of interface migration. In certain cases, ing dislocations, structural ledges [10], or discon-
such as the growth of c from b Cu–Zn, where the nections (defects that combine the properties of
misfit between cube-oriented cubic phases is taken ledges and dislocations [11]).
up by edge components of interfacial dislocations, The interfacial structure of the broad faces of
and where antiphase boundaries (APBs) are ob- Widmanst€atten ferrite plates formed at 450–475
served to trail the interfacial misfit dislocations in C, (a temperature range that would ordinarily
the ordered product, this evidence is quite clear [7]: encompass upper bainite) as they meet parent (re-
at lower driving forces, the interfaces move by a tained) austenite crystals was the subject of a de-
lateral mechanism; at higher driving forces, a finitive electron microscopical study by Rigsbee
transition to a normal mechanism occurs. In other, and Aaronson [10]. In that work, it was found that
more complex situations, the conclusions to be these interfaces are each composed of a single set
drawn are less obvious, but the method still has of misfit-compensating edge dislocations, spaced
value. 4 nm, and a (near-orthogonal) set of structural
For the present purposes, the possible defects ledges. These latter defects serve to displace low
associated with a transformation are classified into energy interfacial microfacets by a few atomic
three categories: distances. They are considered intrinsic compo-
nents of the structure, and serve to optimize the
• Interfacial defects: These are intrinsic to and lo- matching between parent and product structures
calized to the interface. They may be associated (which in the cases studied shared a Kurdjumov–
with the mismatch between parent and product Sachs, Nishiyama–Wasserman or similar orien-
phases [8]; they may also be associated with the tation relationship). Similar results have been
migration of the interface [9]. The nature of obtained for the cases of b.c.c./h.c.p. interfaces, for
these defects can yield important information example in some zirconium based alloys [12], which
about the possible interfacial migration pro- clearly undergo transformations involving the dif-
cesses. In the case of austenite decomposition fusional redistribution of substitutional alloying
G.R. Purdy / Scripta Materialia 47 (2002) 181–185 183
elements. In each case, the interfacial defects ob- This kind of microstructural evidence and in-
served are not capable of motion by simple glide. ference suggests strongly that the defect natures of
The observed habit plane facets have been ra- the habit plane interfaces of ferrous bainite and
tionalized in many cases by application of O-Lat- martensite are fundamentally different: one is ca-
tice theory [8]. In f.c.c./b.c.c. and f.c.c./h.c.p. pable of motion by glide (martensite) and the other
diffusional transformations, it now appears rather is not (bainite).
general that the habit plane contains an invariant
line of the transformation, and that the major axis 1.2. Trailing defects
of intergranular precipitates lies very close to this
direction [13,14]. Implicit in application of O- The defects left behind a transformation inter-
Lattice or similar concepts is the notion that the face are important clues to the mechanism of dis-
interface locally optimizes structural matching placement. In the case of ferrous martensites, some
(which very likely simultaneously minimizes the of these defects are considered to relate to the ‘‘lat-
specific interfacial energy). It is clear to this writer tice-invariant deformation’’, required to maintain
that interfaces defined by energy minimization will the transformation interface plane as one of zero––
not in general be capable of motion by glide. or at least extremely small––net distortion. This
These optimally matched interfaces joining dis- deformation stage is particularly important in fer-
similar structures have been found consistently to rous martensitic interface propagation, since this
contain variably spaced growth ledges, whose riser process admits of no thermally activated relax-
height is typically many times that of any struc- ation.
tural ledges, and whose lateral motion accounts In lower carbon martensites, (less than about
for the displacement and reconstruction of the 0.4 mass% C) the lattice-invariant deformation
low-energy habit plane interfaces. takes place mainly via glide, which in principle
Returning to the specific interface: bainitic fer- need not, but could, leave trailing defects. (High
rite/austenite, it is noted that studies of bainite densities of dislocations are commonly found in
laths formed at 350–375 C in Fe–Ni–C alloys and lath martensite.) Twins have been reported in fer-
Fe–Mn–Si–Mo–C alloys [15] revealed interface rous martensites with carbon contents as low as
characteristics similar to those found by Rigsbee 0.3 mass% (and as low as 0.1 mass% C for alloys
and Aaronson [10]. Sets of closely spaced (2 nm) transformed at high pressures) [18].
interfacial defects were found consistently to pos- Although bainitic ferrite is often found to be
sess strain field contrast of edge character, with heavily dislocated, this writer is not aware of any
Burgers vector [1 1 0]c lying in the interface plane evidence for trailing twins, regardless of carbon
(near (1 1 1)c and (0 1 1)a). content of the austenite. Again, this observation
In contrast to these results for bainites, Sandvik militates against a martensitic mechanism of bai-
and Wayman [16,17] demonstrated that the defect nite growth. It would be of interest to examine
structure of interfaces between lath martensite and bainites formed from austenites subjected to high
(heavily alloyed) austenite was consistent with an hydrostatic pressure during transformation. High
athermal propagation mechanism. The disloca- pressures are believed to promote the formation of
tions and the interfacial steps observed experi- twinned martensite by depressing the temperature
mentally were shown to have strain fields in accord of martensite formation.
with a model of an interface structure capable of
accomplishing the transformation from f.c.c to 1.3. Inherited defects
b.c.c. by glide. They also noted that the martensite
transformation interface, as modeled, was not one A further possibility relates to the incorporation
of minimum energy, and that a ‘‘non-conserva- in the product phase of defects which have been
tive’’ transformation interface (one incapable of introduced into the parent phase, for example by
glide) between similarly oriented phases would prior deformation. If a martensitic process is to
have a lower specific energy. occur, such that the lattice of the parent phase is
184 G.R. Purdy / Scripta Materialia 47 (2002) 181–185
simply deformed into that of the product, then an interface were capable of moving in an ather-
defects may also be subject to the same deforma- mal fashion. In contrast, martensite/austenite
tion, and be found in the product as signatures of interfaces are required to be glissile. Further,
the transformation mechanism. Embury [19] has more detailed, evidence concerning the nature
noted that the martensite in ausformed steels may of these interfaces for different levels of carbon
be strengthened by the presence of sessile disloca- and alloying elements for both bainite and mar-
tions of the type ah1 0 0ia0 , derived from a=2h1 1 0ic tensite would of course be welcome. However, it
parent phase dislocations by the transformation seems that the basic characteristics of the inter-
strain. (Two thirds of the possible a=2h1 1 0ic faces, as inferred from transmission electron mi-
dislocations will become a=2h1 1 1ia0 , and the croscopical studies [10,16], are now rather well
remaining third transform to ah1 0 0ia0 .) In addi- understood.
tion, the {1 1 1}c glide planes become {1 1 0}a0 on • Trailing defects have received less attention. It is
transformation. This type of structural inheritance suggested here that if bainites formed from high
has been demonstrated for subboundaries [20]. It carbon austenites were the result of a marten-
also has been shown to occur in (analogous) de- sitic mechanism, one would also expect to see
formation twinning systems, where again, the twins in the product phase, originating at and
transformation strain acts not only on the parent trailing the interface, as is found in higher-car-
lattice, but also on dislocations and related defects bon martensites.
which survive the lattice deformation but are al- • Inherited defects have been characterized in de-
tered by it [21]. In this case, detailed evidence has tail in products of mechanical twinning trans-
been presented for the existence of a½0 1 0 seg- formations, but not, to the writer’s knowledge,
ments resulting from the operation of the twinning in martensites or bainites. It would seem that
transformation strain on dislocations whose orig- a ‘‘martensitic’’ interfacial migration mecha-
inal Burgers vectors were a=2½1 0 1. nism for growth should result in the inheritance
The internal defect structure of bainite formed of some (now sessile) dislocations with Burgers
from previously deformed austenite has not re- vectors transformed from those typical of de-
ceived the same level of scrutiny as that of the formed austenite. A search for defects of this
martensite or the mechanical twinning trans- kind has the potential to add to our knowledge
formation described above. Sandvik [22] showed of both bainite and martensite transformations.
that images of microtwinned regions in austenite
seemed to continue into bainitic ferrite, but was Defect signatures are effective indicators of
unable to determine whether the inherited regions mechanism of phase transformation; the evidence
had actually transformed to ferrite, or had re- reviewed here is strongly supportive of a mecha-
mained as platelets of retained austenite. It would nism of bainite formation that is quite different
seem that this is a promising area for further study. from that of martensite. In the author’s opinion,
the structure of the planar bainitic ferrite/austenite
interface is determined by energy minimization:
2. Conclusions the resulting interface is incapable of motion by
glide; its normal progress is accomplished by a
It has been helpful to classify transformation- lateral mechanism and accompanied by the si-
related defects into three distinct groups interfacial multaneous rejection of carbon.
defects; trailing defects; and survivor defects.
• Concerning interfacial defects, a certain amount Acknowledgements
of evidence is available to indicate that many
bainite/austenite habit plane interface structures This research was supported by a grant from
are determined mainly by optimal interfacial the Natural Sciences and Engineering Research
matching. It would be pure coincidence if such Council of Canada. The author is grateful to
G.R. Purdy / Scripta Materialia 47 (2002) 181–185 185
Messrs. H.I. Aaronson, Y.J.M. Brechet, J.D. [8] Bollman W. Crystal defects and crystalline interfaces. New
Embury, Mats Hillert and D. Quidort for valuable York: Springer-Verlag; 1970.
[9] Aaronson HI. In: Decomposition of austenite by diffu-
discussions. sional processes. New York: Interscience; 1962. p. 317.
[10] Rigsbee JM, Aaronson HI. Acta Metall 1979;27:351.
[11] Hirth JP. J Phys Chem Solids 1994;55:985.
[12] Zhang W-Z, Purdy GR. Acta Met et Metall 1993;41:
References 543.
[13] Dahmen U. Scripta Metall 1981;15:77.
[1] Hillert M. ISIJ Int 1995;35:1134. [14] Zhang W-Z, Purdy GR. Phil Mag A 1993;68A:291.
[2] Bhadeshia HKDH, Edmonds DV. Acta Metall 1980;28: [15] Li C, Perovic V, Purdy GR. In: Phase transformations’87.
1265. London: The Institute of Metals; 1988. p. 326.
[3] Bhadeshia HKDH. Bainite in steels. London: The Institute [16] Sandvik BPJ, Wayman CM. Metall Trans A 1983;14A:
of Metals; 1992. 823.
[4] Aaronson HI. In: The mechanism of phase transforma- [17] Sandvik BPJ, Wayman CM. Metall Trans A 1983;14A:
tions in crystalline solids. London: The Institute of Metals; 835.
1969. p. 270. [18] Vyknal RF, Radcliffe SV. Acta Metall 1967;15:1475.
[5] Aaronson HI, Reynolds Jr WT. In: Phase transforma- [19] Embury JD. In: Kelly A, Nicholson RB, editors. Strength-
tions’87. London: The Institute of Metals; 1988. p. 301. ening methods in crystals. New York: Wiley; 1971. p. 387.
[6] Bhadeshia HKDK. In: Phase transformations’87. London: [20] Shaller FW, Schmatz DJ. Acta Metall 1963;11:1193.
The Institute of Metals; 1988. p. 309. [21] Basinski ZS, Szczerba MS, Niewczas M, Embury JD,
[7] Stephens DE, Brechet YJM, Purdy GR. Scripta Metall Basinski SJ. Revue de Metallurgie 1997;76:1037.
1988;22:1897. [22] Sandvik BPJ. Metall Trans A 1982;13A:777.
