A Mechanism for the Formation of Lower Bainite

G. SPANOS, H.S. FANG, and H.I. AARONSON

A diffusional mechanism for the formation of lower bainite is proposed based primarily on
transmission electron microscopy (TEM) observations of isothermally reacted specimens of Fe-
C-2 pct Mn alloys. The mechanism involves the initial precipitation of a nearly carbide-free
ferrite "spine," followed by sympathetic nucleation of "secondary (ferrite) plates" which lie at
an angle to the initial "spine." Carbide precipitation subsequently occurs in austenite at
ferrite: austenite boundaries located in small gaps between the "secondary plates." An "an-
nealing" process then occurs in which the gaps are filled in by further growth of ferrite and
additional carbide precipitation; the annealing out of fenite: ferrite boundaries between impinged
"secondary plates" completes this process. This annealing stage contributes to the final ap-
pearance of lower bainite sheaves as monolithic plates containing embedded carbides. The pres-
ent mechanism accounts for the single variant of carbides oriented at an angle to the sheaf axis
repeatedly reported in lower bainite; it is also consistent with the previous observation of one
"rough" side and one "smooth" side of lower bainite "plates."

I. INTRODUCTION the chemical free-energy change attending the trans-

A L T H O U G H there were many observations of the
microstructure now known as bainite in steel prior to 1930
(as summarized by Hultgren [q) and the morphological
features of this constituent, to the extent that they could
be characterized by optical microscopy, were well de-
scribed by Robertson t2j in 1929, it was not until the fa-
mous paper published in the following year by Davenport
and Bain I31 on the time-temperature-transformation (TTT)
diagram that the bainite reaction was taken seriously as
a major mode through which the decomposition of aus-
tenite can take place. Mehl t4] made an important contri-
bution to the nomenclature of this reaction in 1939 when
he divided the morphology of ferrous bainites into two
categories, "upper" and "lower" bainite, named in ac-
cordance with the temperature regions in which each ap-
pears. Although no such division of bainites formed in
substitutional nonferrous alloys has so far been reported,
this categorization has been widely accepted and utilized
in steels.
Lower bainite has occupied a position of particular
prominence with respect to evaluations of the basic
mechanism of the bainite reaction. These evaluations,
which have been undertaken repeatedly since 1930, largely
revolve around the issue of whether the unit atomic pro-
cess through which iron atoms are transported across
austenite:ferrite boundaries is one of shear or of diffu-
sional jumps. That the driving force for these jumps is
G. SPANOS, formerly Graduate Student, Carnegie Mellon
University, is with the Physical Metallurgy Branch of the Naval
Research Laboratory, Washington, DC 20375-5000. H.S. FANG,
formerly Visiting Professor, Carnegie Mellon University, is Professor,
Division of Metallic Materials, Department of Mechanical Engineering,
Tsinghua University, Beijing 100084, People's Republic of China.
H.I. AARONSON, R.F. Mehl Professor, is with the Department of
Metallurgical Engineering and Materials Science, Carnegie Mellon
University, Pittsburgh, PA 15213.
This paper is based on a presentation made in the symposium
"International Conference on Bainite" presented at the 1988 World
Materials Congress in Chicago, IL, on September 26 and 27, 1988,
under the auspices of the ASM INTERNATIONAL Phase Transfor-
mations Committee and the TMS Ferrous Metallurgy Committee.
formation of austenite to ferrite is rarely disputed. There
also seems to be general agreement that, at least in dis-
tinctly hypoeutectoid alloys, the first formed product of
the transformation is ferrite. I5] However, a broad spec-
trum of opinion about the relative contributions of these
9two fundamental mechanisms of atomic attachment is now
available, especially as a function of isothermal reaction
temperature. At one extreme, Widmanst~itten ferrite
(containing few if any carbides), upper bainite, and lower
bainite are all considered to form by high-velocity shear.
At the other extreme, all of these transformation prod-
ucts are taken to develop through diffusional mecha-
nisms, with only martensite forming by a shear process.
Probably the majority opinion is that the contribution of
shear increases with decreasing temperature. Numerous
investigators thus adhere to the view that the formation
of lower bainite, at the least, takes place by shear./6-91
Conversely, several criticisms against a shear mecha-
nism for bainite formation have been made previ-
ously; tl~ hence, only a few of the major topics of this
debate which are pertinent to the present model of lower
bainite formation will now be briefly considered.
One of the strongest arguments for the growth of lower
bainite plates by shear is the now "classic" presence of
elongated carbides lying at an angle of approximately 55
to 60 deg to the longitudinal axis of the plates. I8,16-~9]An
obvious mechanism through which these carbides form
is by the precipitation from the (necessarily highly super-
saturated) ferritic component of lower bainite. In view
of the much higher diffusivity of carbon in ferrite [2~ than
in austenite, [2q both acquisition and retention of suffi-
cient carbon supersaturation to account for the large
volume fractions of carbides often found within lower
bainite require that its ferritic component grow at rates
well above those allowed by the diffusion of carbon in
austenite away from the growing bainite plates, i.e., by
shear taking place at quite high velocities. [~,l~ How-
ever, there is still no reliable experimental evidence for
such high growth rates, as discussed in detail in another
paper in this symposium, u3] One of the most puzzling
features of this mechanism is the repeated observation

METALLURGICAL TRANSACTIONS A VOLUME 21A, JUNE 1990-- 1381

that nearly all carbide plates or rods precipitate parallel
to only one crystallographic variant of their habit plane
in the ferritic component of a given lower bainite
plate. I8,~6-191 This is in contrast to tempered martensite,
where carbide precipitation parallel to multiple members
of the same crystallographic form is repeatedly ob-
served. [23-261Insofar as the present authors are aware, the
only explanation offered for this phenomenon on the basis
of.precipitation wholly within ferrite is that the carbides
form along parallel twins created in the bainitic ferrite
plates while these plates grow by a shear mechanism. [271
However, twins have not been observed in lower (or in
upper) bainite. 17'271
A substantial body of experimental evidence has now
been accumulated suggesting that this view of lower
bainite formation is incorrect. Smith et a1.128~ have thus
observed with hot-stage optical microscopy the forma-
tion of martensite occurring with great rapidity, while
that of lower bainite is simultaneously taking place very
slowly at reaction temperatures below Ma. Speich I291
subsequently reported measurements of both the length-
ening and thickening kinetics of lower bainite plates with
the same technique in hypereutectoid Fe-C and Fe-C-X
alloys. He satisfactorily rationalized these measurements
in terms of diffusional growth. A more recent report based
upon photoemission electron microscopy, that lower
bainite plates in an Fe-0.34 wt pct C-2.02 wt pet Si-3.0 wt
pct Mn alloy lengthen much more rapidly than carbon
diffusion in austenite will allow, [9J may be due instead
to bainite-stimulated martensite formation occurring well
above the Ms temperature, as observed by Smith e t a / . [281
and recently characterized in detail by Okamoto and
Oka. [30]
Crystallographic arguments in favor of carbide nucle-
ation predominantly within the ferritic component of lower
bainite [81 are often ambiguous. I31'32J In particular,
Shackleton and Kelly I311 showed that the Bagaryatski [331
orientation relationship between ferrite and cementite
observed in lower bainite, [3~1 which is the same as that
found in tempered martensite, [33'34'351 can be explained
just as well in terms of precipitation of certain variants
of both ferrite and cementite directly from austenite.~27~
Orientation relationship arguments that the ferrite forms
b y shear [6,36] are diminished by the finding that lower
bainite and martensite formed in the same alloy have quite
different crystallographies. [371 The appearance of tent-
shaped rather than invariant plane strain surface relief
effects in association with lower bainite plates t29] further
detracts from arguments in favor of a shear mechanism
for the formation of lower bainite, since it has been dem-
onstrated that diffusional transformation products can yield
tent-shaped reliefs I38,39'4~ wheareas martens[tic products
should not do so.
On the basis of this background, the question of the
mechanisms through which lower bainite develops is en-
gaged in this paper. Instead of shear, purely diffusional
ledgewise migration of partially coherent ferrite: austenite
boundaries and sympathetic nucleation [411of ferrite crys-
tals at the ferrite: austenite boundaries of previously formed
ferrite crystals will be utilized to develop a new trans-
formation mechanism capable of explaining the distri-
bution and morphology of carbides within lower bainite.
This mechanism will be based primarily on an optical
1382--VOLUME 21A, JUNE 1990
and transmission electron microscopy (TEM) investiga-
tion of lower bainite morphology in both hypo- and
hypereutectoid Fe-C-Mn alloys.
II. EXPERIMENTAL PROCEDURE
Six high-purity Fe-C- - 2 pct Mn alloys* and a single
*All compositions are expressed in weight percent; the exact man-
ganese content in each of these alloys is actually within -+0.2 pct
of 2 pct.
Fe-0.1 pct C-2.9 pct Mn alloy were employed in this in-
vestigation. In the former alloys, the carbon concentra-
tion ranged from 0.34 to 1.37 pet. The composition and
estimated Ms temperature [42J of all alloys used are listed
in Table I. The TEM results presented here are centered
on one hypoeutectoid alloy (0.34 pet C) and one hyper-
eutectoid alloy (0.95 pet C). The heat treatment and TEM
procedures utilized are described in Reference 4 3 . [431
III. RESULTS AND PROPOSED MECHANISM
Before presenting the findings of the present investi-
gation, it is necessary to note that the term "sheaf" was
originally employed to describe a packet of precipitate
plates, particularly of ferrite, formed by successive "face-
to-face" sympathetic nucleation, [41,441 as shown sche-
matically in Figure 1. Sympathetic nucleation has been
defined as the nucleation of a precipitate crystal at the
interphase boundary of another precipitate crystal of the
same phase when the matrix and precipitate differ con-
tinuously in composition. [41,441 The dashed lines across
these plates drawn perpendicular to their broad faces in-
dicate that "edge-to-edge" sympathetic nucleation 144,45]
has also occurred and has participated in their length-
ening. Oblak and Hehemann [TJ have also reported this
type of microstructure but interpreted it differently, as is
considered in Section IV. Published optical and trans-
mission electron micrographs [1~ indicate that the fer-
ritic component of upper bainite consists primarily of
single intragranular plates (or laths [271) and closely spaced
sideplates (or laths I271) at higher temperatures but in-
creasingly of sheaves of sideplates and of intragranular
plates at lower temperatures. These sheaves are predom-
inantly of the type sketched in Figure 1. In the lower
bainite region, we shall now demonstrate that the ferritic
component also consists of sheaves of crystals but with
a different and considerably more complex geometry.
Unlike upper bainite, many of the individual plates in a
sheaf will be shown to make an appreciable angle with
respect to the axis of the sheaf as a whole. Given this
background, the proposed mechanism for lower bainite
formation will now be presented concurrently with the
experimental observations.
On the present model, the initial element of a lower
bainite sheaf* is evidently a single, nearly carbide-free
*For reasons which will become apparent, this paper uses the term
lower bainite "sheaf" to refer to the microstructure which has in the
past been termed a lower bainite "plate." [8,27,471
ferrite plate. Figures 2(a) and (b) are bright- and dark-
field TEM micrographs of a lower bainite sheaf formed
at 250 ~ in an Fe-0.95 pct C-1.93 pet Mn alloy. The
METALLURGICAL TRANSACTIONS A
 Table I. Alloy Compositions and Estimated M,* Temperatures
Wt Pct C Wt Pct Mn Wt Pct Si Wt Pct P Wt Pct S Ms (~
0.10 2.90 0.02 0.004 0.005 448
0.34 1.90 0.003 0.004 0.005 337
0.57 1.99 0.007 0.003 0.003 243
0.77 1.94 0.003 0.002 0.003 185
0.95 1.93 <0.01 0.003 0.003 146
1.12 1.82 0.003 0.003 0.004 131
1.37 1.99 0.005 0.003 0.002 104
*Estimates were made from formulas compiled by Andrews. [42j

heavy arrow in each of these figures points out the ini- boundaries (partially) enclosing the gaps. More time is
tiating ferrite plate. thus available for carbide nucleation at the austenite: ferrite
Once the initiating ferrite plate of a sheaf, which may interface.
be termed its "spine," has appeared, other ferrite crystals The fourth and final stage in the development of a
are then sympathetically nucleated at it but often on only lower bainite sheaf is further ferrite growth around the
one of its two broad faces, as in Figure 2. These crys- carbides, and perhaps additional carbide precipitation,
tals, whose morphology often appears to approximate that until the small amount of austenite remaining in the gaps
of a thick plate, frequently develop at a marked angle to has been decomposed. If a significant proportion of Si
the spine. Figure 3(a) shows another example of this is present in the alloy, though, both the third and fourth
structure near the tip of a lower bainite sheaf. In steps of this sequence can be greatly inhibited.
Figure 3(b), the higher magnification employed permits The overall external shape of the sheaf will thus quickly
observation of serrations at the sides of the sheaf, de- lose its initial serrated appearance and will soon more
veloped where the individual "secondary plates" have nearly resemble that of a lenticular ferrite monocrystal
lengthened unequally, perhaps because they nucleated at
different times. Figure 4 illustrates this type of structure
in the 0.34 pct C alloy; gaps between adjacent "second-
ary plates" are evident (indicated by arrows in Figure 4).
The third step in this process is the precipitation of
carbides in the small gaps between "secondary plates"
(which may have already impinged along other portions
of their interphase boundaries). In the present alloys,
electron diffraction studies indicated that the carbide was
cementite. A typical example of the diffraction analysis
is shown in Figure 5. (Also note the ferrite spine
in Figure 5(a).*) These carbides were presumably nu-
*The appearance of a few carbides in the region of the spine shown
in Figure 5(a) is most likely due to a stereological effect in which the
carbides associated with sympathetically nucleated ferrite plates over-
lap the spine, either above or below the spine, within the T E M foil.
However, it is also possible that some carbide precipitation occurred
at austenite:ferrite boundaries of spines at boundary orientations where (a)
their growth was markedly slowed by a low density of growth ledges
and later resumed, or reinitiated, by sympathetic nucleation, t44,451

cleated at the austenite: ferrite boundaries forming the sides

of the gaps in the same manner that carbide precipitation
occurs between ferrite laths in upper bainite. Extensive
overlap within the austenite gaps of carbon diffusion fields
associated with the adjoining ferrite must markedly in-
crease the carbon concentration in the gaps and thereby
diminish the migration kinetics of the ferrite:austenite

( i )
(- i )
(b)

C
I
"
!
I
'
I
!
' ) Fig. 2 - - A lower bainite sheaf formed in Fe-0.95 wt pct C-1.93 wt
pct Mn reacted at 250 ~ for 15,000 s: (a) bright-field T E M micro-
Fig. 1 - Schematic representation of an upper bainite "sheaf." graph and (b) corresponding dark-field micrograph.

METALLURGICAL TRANSACTIONS A VOLUME 21 A, JUNE 1990-- 1383

 (a)
(b)
Fig. 3 - - Leading tip of a lower bainite sheaf in Fe-0.95 wt pct C- 1.93
wt pct Mn reacted at 250 ~ for 15,000 s: (a) lower magnification
view and (b) higher magnification image showing serrations at the
sides of the sheaf.
containing embedded carbides. As illustrated in Figure 6,
this "smoothing" process occurs with sufficient rapidity
so that the polycrystalline nature of the sheaf can be de-
tected only near its leading edge, where this process was
terminated by quenching.* At the same time, migration
*Note that the long boundaries discerned near the center and lower
right-hand portions of the sheaf in Figure 6 (and which lie parallel to
the sheaf axis) are not the ferrite:ferrite boundaries between second-
ary sideplates discussed throughout the text. The latter boundaries are
only visible near the leading edge (e.g., see the higher magnification
micrographs of Figures 3 and 4). The former boundaries apparently
result from impingement of multiple ferrite "spines." Such spines have
been "caught" at an earlier stage (before impingement) in the micro-
graphs in Figures 2 and 8.
of small-angle grain boundaries formed by impingement
between adjacent ferrite crystals within a sheaf will soon
remove much of the internal evidence that many crystals
were involved in the development of the sheaf, much in
the same manner as an annealing or recovery process,
e.g., during the tempering of martensite. I23~ Addition-
ally, after complete impingement has occurred along the
broad faces of ferrite sideplates, since their relative mis-
orientation should be very small, such boundaries ex-
hibit very weak c o n t r a s t - - and may even be nonexistent.
A large portion of the sheaf shown in Figure 6 thus ap-
pears, except upon closest inspection, to be simply the
product of the precipitation of carbide plates parallel to
but a single habit plane within a large single crystal of
ferrite.
Recapitulating this mechanism through the sequence
1384--VOLUME 21A, JUNE 1990
Fig. 4 - - G a p s (indicated by arrows) between ~secondary ferrite plates"
near the tip of a lower bainite sheaf in an Fe-0.34 wt pct C-1.90 wt
pct Mn alloy reacted at 350 ~ for 40 s.
of sketches (which are greatly simplified for the pur-
poses of illustration) in Figure 7, a carbide precipitation
process emerges which differs significantly from that of
precipitation of a single variant of carbides entirely from
ferrite. Carbide precipitation in lower bainite is now sug-
gested to occur at the austenite: ferrite boundaries of the
"secondary plates," predominantly from austenite (as will
be considered further in Section IV). In effect, then, the
sequence proposed in Figure 7 is essentially the same as
the one which is widely accepted for upper bainite, i.e.,
precipitation of carbides at t~: y boundaries forming nar-
row "pockets" or "gaps" of carbon-enriched austenite
between upper bainite plates, t8,46~ Thus, the major dif-
ference between the development patterns of sheaves of
upper bainite plates and those of lower bainite plates ap-
pears to be that the ferritic component of lower bainite
sheaves is composed of one or a few long "spines" plus
a number of shorter plates making a substantial angle
with respect to the spines, while the ferrite plates (or
laths) in upper bainite form by face-to-face sympathetic
nucleation.
The fine scale on which the sympathetic nucleation
and the growth of ferrite plates occur in the lower bainite
region and the swift obscuration of the initial micro-
structure are probably the principal reasons why pre-
vious investigators have not discerned the present
mechanism for the formation of lower bainite. Addi-
tionally, the thin ferrite spines are often difficult to de-
tect even by TEM unless the lower bainite sheaf is prop-
erly oriented in the microscope such that the secondary
sideplates (and the corresponding carbides which form
between them) do not overlap the spine in the field of
view. Nevertheless, some common features between the
current mechanism and previous models will be consid-
ered in Section IV.
Two complications in the morphology of ferrite within
lower bainite sheaves must now be briefly noted. Figure 8,
an enlargement of a portion of Figure 2(a), demonstrates
that multiple ferrite "spines" (pointed out by arrow-
heads) are present in some sheaves. These are often both
METALLURGICAL TRANSACTIONS A
Fig. 5--Fe-0.95 wt pct C-1.93 wt pct Mn reacted at 250 ~ for 15,000 s: (a) bright-field TEM, (b) dark-field TEM taken from a 113 carbide
reflection, and (c) corresponding diffraction pattern. Small circles in (c) represent cementite reflections.

shorter and thinner than the spine first formed. Length-
ening rapidly, such spines would physically prevent fur-
ther lengthening of large numbers of secondary plates
(formed earlier), as shown in Figure 8. Several ferrite
"spines" can thus be discerned in the lower bainite sheaf
shown in Figure 5 of Hehemann's tS] well-known review
of the bainite reaction, reproduced here in Figure 9.
The second complication in the morphology of ferrite
in lower bainite sheaves to be noted is the strong ten-
dency for individual ferrite crystals within these sheaves
to form facets diagonally across their leading edge. This
feature is, of course, most readily observed at the lead-
ing tip of a sheaf, as illustrated in Figure 10; arrowheads
point out prominent examples of such facets. Ohmori
e t al. t48] also drew attention to facets of this type in their
sketches of bainite sheaves, emphasizing their potential
for serving as nucleation sites for bainitic carbides. Car-
bides growing "allotriomorphically" or as plates along

Fig. 6 - - L o w e r magnification bright-field TEM image of the lower

bainite sheaf corresponding to Fig. 4, illustrating the smooth character
of the sheaf boundary further from the tip. Fig. 7 - - Sketch of the proposed mechanism for lower bainite formation.

METALLURGICAL TRANSACTIONS A VOLUME 21A, JUNE 1990-- 1385

 I
Fig. 8--Enlargement of Fig. 2(a) revealing multiple ferrite spines.
such facets probably contribute to the carbide population
whose long axis lies at about 55 to 60 deg with respect
to the longitudinal axis of the sheaves.
In regard to the carbide morphology within lower bainite
sheaves, the dark-field micrographs of Figures 1 l(a) and
(b) indicate that the aligned carbides are not monolithic
plates but, rather, are individual carbide crystals nu-
cleated side by side, presumably along austenite:ferrite
boundaries. Some have probably grown laterally into
contact, while others may have been sympathetically nu-
cleated at the edges of carbides formed a little earlier
along these boundaries. The morphology of an aggregate
Fig. 9 - - T E M micrograph of lower bainite in 4360 steel reacted at
275 ~ taken from Ref. 8 (magnification was not reported).
1386--VOLUME 21A, JUNE 1990
Fig. 10--Facets at crystals at the leading tip of a lower bainite sheaf
in Fe-0.34 wt pct C-1.90 wt pet Mn reacted for 40 s at 350 ~
of many such individual carbide crystals seems better
described as interphase boundary allotriomorphs than as
plates.
IV. DISCUSSION
A. Origin of the Transition from Upper to Lower
Bainite Sheaves
The most likely explanation of why edge-to-face sym-
pathetic nucleation of secondary plates against a broad
face of the ferrite spine appears in the lower but not in
the upper bainite region is as follows. The higher driving
force for ferrite nucleation available in the lower bainite
regime simply makes the higher work of formation (AF*)
of edge-to-face critical nuclei for sympathetic nucleation
of ferrite, relative to the AF* for face-to-face sympa-
thetic nucleation, t441 a competitively surmountable bar-
rier at these temperatures. For example, calculations of
the driving force for nucleation of ferrite have shown
that a decrease in temperature from 500 ~ (in the upper
bainite range) to 350 ~ (within the lower bainite range)
in the Fe-0.34 wt pct C-1.90 wt pct Mn alloy can lead
to a 50 pct reduction in AF* for a given critical nucleus
geometry and interfacial energy. 143j
B. Some Questions Associated with the Present
Mechanism for Formation of Lower Bainite Sheaves
Speich t291 has found that lower bainite plates formed
in an Fe-0.66 pct C-3.32 pct Cr steel yield a tent-shaped
surface relief effect. This appears inconsistent with the
present view that lower bainite consists of sheaves of
secondary plates sympathetically nucleated at one broad
face of (usually) a single or a few ferrite spine(s).* How-
* Speich also concluded that sympathetic nucleation does not appear
to be operative during the formation of the lower bainite structures he
studied. However, the lesser resolving power of the observational
techniques he employed, optical and replication electron microscopy,
may have prevented observation of the fine (and easily obscured)
structural features which led the present authors to the opposite
conclusion.
METALLURGICAL TRANSACTIONS A

Fig. 1 1 - - L o w e r bainite formed in 0.95 wt pct C-1.93 wt pct Mn re-
acted at 200 ~ for 130,000 s: (a) dark-field micrograph taken from
a carbide reflection and (b) lower magnification dark-field image.
ever, since the spines have a thickness of about 0.2/xm
or less and may often lie below the plane of polish, they
are not resolvable with optical microscopy. Conse-
quently, the reliefs observed should be those produced
by the array of parallel secondary plates. On the other
hand, such an array ought to introduce a considerable
substructure into the surface relief effect, again on a very
fine scale, and perhaps also not detectable with optical
interference microscopy.
A further question about lower bainite sheaf mor-
phology is why do secondary plates often form on only
one side of many of the ferrite spines? Hehemann has
long emphasized that lower bainite "plates," which the
present authors refer to as sheaves, often have one smooth
side and one rough, with growth evidently occurring only
on the rough side. tS~This observation, made mainly with
METALLURGICAL TRANSACTIONS A
replication electron and optical microscopy, is com-
pletely consistent with the present TEM observations.
The "smooth" side is now shown to correspond to an
initiating ferrite spine, and the "rough" side is formed
by the leading edges of the secondary plates. Hence, this
question becomes one of explaining the asymmetry of
sympathetic nucleation on opposite sides of the initiating
spine. Although the current experimental findings do not
provide a direct answer to this question, a possible ex-
planation will now be offered. On a recent analysis of
sympathetic nucleation, t44~ it may be postulated that the
nominally smooth side of the spine actually has a higher
density of growth ledges than the side at which second-
ary plates are sympathetically nucleated. The larger the
average intededge distance, the longer is the time avail-
able for sympathetic nucleation on the terraces and, thus,
the smaller is the probability that the newly formed crys-
tals will be overrun by an advancing ledge. ~44jHowever,
the origin of the postulated unequal densities of growth
ledges on opposite broad faces of a "spine" remains un-
c e r t a i n - - a somewhat unsurprising situation in view of
the difficulties being encountered in explaining the for-
mation kinetics of growth ledges.j491
C. On the Crystallography of Secondary
Sideplate Formation
An obvious question which arises is why the broad
faces of "secondary sideplates" lie at a characteristic angle
of approximately 55 to 60 deg [47,5~ to the "spine" when
the {111}~ atomic habit planes of ferrite plates tS~j form
70.5 deg angles with respect to one another (within a
given matrix austenite grain). A survey of published thin
foil and replication TEM micrographs [7,27'46,48,521of lower
bainite sheaves showed that the apparent angle (i.e., not
corrected for stereological effects) between the carbide
direction and the sheaf axis actually varies between 46
and 67 deg, with 70 pct of the angles lying between 55
and 61 deg; a determination of the true angle by trace
analysis procedures from a statistically acceptable num-
ber of plates has yet to be reported. Nevertheless, the
often quoted angle of 55 tS~ or 60 deg [471 appears to be
somewhat restrictive.
Previous optical microscopy studies of lower bainite
"plates" have shown considerable scatter in the habit plane
at a given temperature as well as a change in the range
of habit planes observed with decreasing tempera-
ture. t4,531 The habit planes measured by optical micros-
copy are now seen to be those averaged over all the
crystals comprising a given lower bainite sheaf; both
structural and growth ledges at ferrite:austenite inter-
faces can cause even further deviation of the atomic habit
plane from that observed by optical microscopy, rS1} As
far as the present authors are aware, the only study in
which the atomic habit plane between ferrite and aus-
tenite has been deduced is that of Rigsbee and Aaronson tSl~
for ferrite plates formed at 450 ~ to 475 ~ in an Fe-
0.62 pct C-2.0 pct Si alloy. They reported triatomic
structural ledges spaced between 2.2 and 3.6 nm apart,
which cause the apparent habit plane, observed by con-
ventional TEM techniques, to deviate by as much as
18 deg from the atomic habit plane, which was found to
be {111}~/{110}~. Thus, if the atomic habit plane of the
VOLUME 21A, JUNE 1990--1387
spine and the sideplates in lower bainite sheaves are dif-
ferent variants of {11 1}J{1 10}~, the presence of such
ledges could account for the presence of angles smaller
than 70.5 deg between the apparent habit planes of these
variants.* The variations of the apparent habit plane re-
ported by Rigsbee and Aaronson could thus explain the
range of apparent angles between the carbides and the
overall sheaf axis observed in the literature.**
*The almost invariable presence of martensite above and below
the bainite sheaves, especially in the vicinity of the tip and the thin
initiating "spine," and the high volume fraction of carbides resulted
in the appearance of so many superimposed spots in the selected area
diffraction patterns that the misorientation between the thin initiating
spine and the "sideplates" could not be determined with any confidence.
**The precise magnitude of the t r u e angle is not critical to this
explanation; the pertinent issue is that a single variant of carbides
(lying at an appreciable angle to the sheaf axis) has been repeatedly
observed in lower bainite. 17,~7,'z-45.5~
D. Mechanism and Kinetics of "Subunit" Growth and
the Origins of Bainitic Carbides
As noted in the Introduction, the present mechanism
for lower bainite formation is based upon diffusional
growth of ferrite (and, implicitly, also of carbides). Also,
as previously remarked, Oblak and Hehemann tT] have
postulated that individual ferrite plates (which they termed
"subunits') within sheaves grow at much higher than
diffusion-limited velocities, sufficient to provide the high
carbon supersaturation in ferrite which is a prerequisite
to extensive carbide precipitation within ferrite plates.
Kinsman and Aaronson employed thermionic emission
electron microscopy to demonstrate that individual fer-
rite plates (a term the present authors prefer to "sub-
unit," with its connotation of subboundaries formed by
polygonization) within upper bainite sheaves grow at ap-
proximately the rates permitted by the diffusion of car-
bon in austenite, tl~ Equivalent experiments are now
required on lower bainite. A useful deduction can be
made, though, from the failure of lower bainite to grow
at high velocities below Ma. [28] Strain energy is consid-
ered to be the factor limiting the growth of individual
ferrite plates or "subunits" to small fractions of the ex-
ternal dimensions of a sheaf. [7a~ However, sufficient
driving force should be available below Ma to permit
these plates to grow swiftly to the much greater lengths
characteristic of martensite plates, which are also ob-
served at these temperatures. [28]The failure of the overall
growth kinetics of lower bainite sheaves to increase greatly
below Ma, while martensite plates simultaneously grow
at much greater velocities, t28] strongly implies that their
component plates do not grow more rapidly than diffu-
sion control allows, even for a short distance (e.g., the
length of a "subunit').
On the diffusional growth mechanism, ferrite should
contain no more carbon than that corresponding to the
extrapolated ferrite/(ferrite + austenite) phase bound-
ary. [l~ Analysis of X-ray diffraction data on the com-
position of retained austenite associated with partial
transformation to both upper and lower bainite indicates
that the average carbon concentration in bainitic ferrite
is even smaller than this. [~~ While these carbon concen-
trations are higher than those of the extrapolated ferrite/
1388--VOLUME 21A, JUNE 1990
(ferrite + carbide) phase boundary, the portion of car-
bide which can precipitate from within ferrite must, ac-
cording to the Lever Rule, be very small under this
circumstance. Most of the bainitic carbides should thus
precipitate from austenite at ferrite:austenite interfaces.
Perhaps the most striking outcome of the present
mechanism is its ability to explain the "single variant"
of carbides oriented at an angle with respect to the lon-
gitudinal sheaf axis [8,~6-19] as a direct and natural con-
sequence of (interphase boundary) carbide precipitation
in gaps between the broad faces of the secondary (fer-
rite) plates (and, to a lesser extent, along facets of ferrite
crystals at the edges of the sheaf). The single "habit plane"
of these carbides is thus a consequence of the predom-
inance of one orientation of secondary ferrite plates in a
bainite sheaf.
E. Comparisons with Some Previous Proposals for
Lower Bainite Formation Mechanisms
Most of the previous proposals are based upon a shear
mode for the formation of the ferritic component of bain-
ite, followed by carbide precipitation within the ferrite.
(Some authors have proposed that even the carbides in
lower bainite form by shear. [54]) Attention will now be
restricted to those mechanisms containing one or more
elements of the formation sequence proposed as a result
of the present investigation.
Huang and Thomas [18] used crystallographic and mor-
phological observations to conclude that carbides formed
in association with lower bainite precipitate predomi-
nantly from austenite at austenite:ferrite boundaries, in
agreement with the present model. They particularly em-
phasized that the single habit plane operative is char-
acteristic of interphase boundary precipitation [47,55]rather
than of carbide precipitation wholly within the ferrite
phase, though their mechanism, if based upon precipi-
tation at a single, essentially planar a : y boundary, is
necessarily incomplete, as will now be described. Ex-
plaining the single carbide "habit plane" as a result of
precipitation at the broad faces of a monolithic lower
bainite plate, tl8] as represented schematically in Figure 12,
/
- ~' '~ "
Fig. 1 2 - - S k e t c h illustrating carbide precipitation at the broad face of
a monolithic bainite plate.156]
METALLURGICAL TRANSACTIONS A
fails to account for the long, semicontinuous sheets of
such carbides making a "55 to 60 deg" angle with re-
spect to the interface, rather than lying parallel to the
interface, as occurs during interphase boundary carbide
precipitation at the broad faces of (nearly) monolithic
grain boundary ferrite allotriomorphs. [47] As was briefly
mentioned in Section III, Ohmori e t al. t4sl also sug-
gested that the carbides precipitate from the austenite phase
at a: y boundaries. In particular, they proposed that car-
bide precipitation may occur at facets on these bound-
aries (at the edges and faces of ferrite plates) which make
a marked angle with respect to the overall sheaf axis
(Figure 13). Both investigations t~8,481were apparently made
on heavily transformed specimens in which many bainite
sheaves had impinged; hence, detailed observations were
not reported on the morphology of the leading edges of
the sheaves.
Insofar as the present authors are aware, the only pre-
vious investigation in which the tips of lower bainite
sheaves were imaged by TEM at reasonably high mag-
nifications is that of Yada and Ooka. t571 Their micro-
graphs also reveal one smooth side and one serrated side
of lower bainite "plates" (at magnifications -<20,000
times); they indicated that a "substructure" of thin ferrite
plates appears to be present. However, they essentially
adhered to Hehemann's tsl proposal that the carbides pre-
cipitate solely within ferrite which has formed by shear.
V. CONCLUSIONS
In this paper, a diffusional mechanism for the for-
mation of lower bainite sheaves has been proposed based
on TEM observations on specimens of Fe-C- ~2 pct Mn
alloys which usually contained a relatively low (less than
about 30 pct) volume fraction of bainite. The mecha-
nism can be summarized in four steps (depicted
schematically in Figure 7): (1) precipitation of a nearly
carbide-free ferrite "spine"; (2) sympathetic nucleation
of "secondary plates" of ferrite, usually on only one side
of and at an angle of approximately 55 to 60 deg to the
initiating "spine"; (3)precipitation of carbides in aus-
tenite at a: y boundaries forming gaps between adjacent
"secondary (ferrite) plates"; and (4) an "annealing" pro-
cess in which the gaps are filled in with further growth
of ferrite and additional carbide precipitation, causing
the sheaf to lose its initial serrated appearance not far
behind its leading edges. Annealing out of ferrite: ferrite
boundaries formed by lateral impingement of adjacent
"secondary plates" further contributes to the appearance
of lower bainite sheaves as monolithic ferrite plates
containing embedded carbides. The ferrite "spine"
Cementite
Fig. 13--Schematic representation of the mechanism of carbide pre-
cipitation suggested by Ohmori et al. 1481
METALLURGICAL TRANSACTIONS A
corresponds to the "smooth" side of lower bainite "plates"
first observed by Oblak and Hehemann t71 and sub-
sequently reported by others, t571 The present mech-
anism accounts for the single variant of carbides at an
angle to the sheaf axis observed repeatedly in lower
bainite, ts'16-191 as a direct consequence of carbide precip-
itation at the broad faces (and, to a lesser extent, on facets)
of "secondary ferrite plates" lying at a large angle to the
initiating ferrite spine.
ACKNOWLEDGMENTS
The authors express their appreciation to the National
Science Foundation for support of this work through Grant
No. DMR81-1907 to the Carnegie Mellon University
Materials Research Laboratory. Appreciation is also ex-
pressed to Professor W.T. Reynolds, Jr. for his careful
review of this manuscript and Professor Helio Goldenstein
for many helpful discussions.
REFERENCES
1. A. Hultgren: Trans. ASM, 1947, vol. 39, pp. 915-89.
2. J.M. Robertson: J. Iron Steellnst., 1929, vol. 119, pp. 391-419.
3. E.S. Davenport and E.C. Bain: Trans. AIME, 1930, vol. 90,
pp. 117-54.
4. R.F. Mehl: Hardenability of Alloy Steels, ASM, Cleveland, OH,
1939, p. 1.
5. G.V. Smith and R.F. Mehl: Trans. AIME, 1942, vol. 150,
pp. 211-26.
6. J.W. Christian and D.V. Edmonds: Phase Transformations in
Ferrous Alloys, A.R. Marder and J.I. Goldstein, eds., TMS-AIME,
Warrendale, PA, 1984, pp. 293-325.
7. J.M. Oblak and R.F. Hehemann: Transformation and Hardenability
in Steels, Climax Molybdenum Co., Ann Arbor, MI, 1967,
pp. 15-38.
8. R.F. Hehernann: Phase Transformations, ASM, Metals Park, OH,
1970, pp. 397-432.
9. H.K.D.H. Bhadeshia: Phase Transformations in Ferrous Alloys,
A.R. Marder and J.l. Goldstein, eds., TMS-AIME, Warrendale,
PA, 1984, pp. 335-39.
10. R.F. Hehemann, K.R. Kinsman, and H.I. Aaronson: Metall.
Trans., 1972, vol. 3, pp. 1077-94.
11. H.I. Aaronson: The Mechanism of Phase Transformations in
Crystalline Solids, Institute of Metals, London, 1969, pp. 270-81.
12. H.I. Aaronson and H.J. Lee: Scripta Metall., 1987, vol. 21,
pp. 1011-16.
13. H.I. Aaronson, W.T. Reynolds, Jr., G. Spanos, and G.J. Shiflet:
Metall. Trans. A, 1990, vol. 21A, pp. 1343-80.
14. H.I. Aaronson: in The Decomposition of Austenite by Diffusional
Processes, V.F. Zackay and H.I. Aaronson, eds., Interscience,
New York, NY, 1962, pp. 387-546.
15. H.I. Aaronson, C. Laird, and K.R. Kinsman: Phase
Transformations, ASM, Metals Park, OH, 1970, pp. 313-96.
16. Second Progress Report of Subcommittee XI of Committee E4:
Trans. ASTM, 1950, vol. 50, pp. 444-92.
17. H. Modin and S. Modin: Jernkontorets Ann., 1955, vol. 139,
pp. 480-515.
18. Der-Hung Huang and Gareth Thomas: Metall. Trans. A, 1977,
vol. 8A, pp. 1661-74.
19. G.Y. Lai: Metall. Trans. A, 1975, vol. 6A, pp. 1469-71.
20. D.N. Beshers: Diffusion, ASM, Metals Park, OH, 1972,
pp. 209-40.
21. C. Wells, W. Batz, and R.F. Mehl: Trans. A1ME, 1950, vol. 188,
pp. 553-60.
22. J.C. Fisher: Thermodynamics in Physical Metallurgy, ASM, Metals
Park, OH, 1950, p. 201.
23. G.R. Speich: Metals Handbook, 8th ed., ASM, Metals Park, OH,
1973, vol. 8, pp. 202-04.
24. G.R. Speich and W.C, Leslie: Metall. Trans., 1972, vol. 3,
pp. 1043-54.
VOLUME 21A, JUNE 1990--1389
25. M.G.H. Wells: Acta Metall., 1964, vol. 12, pp. 389-99.
26. H. Wagenblast and R.C. Glenn: Metall. Trans., 1970, vol. 1,
pp. 2299-2304.
27. F.B. Picketing: Transformation andHardenability in Steels, Climax
Molybdenum Co., Ann Arbor, MI, 1967, pp. 109-32.
28. M.F. Smith, G.R. Speich, and M. Cohen: Trans. AIME, 1959,
vol. 215, pp. 528-30.
29. G.R. Speich: in Decomposition of Austenite by Diffusional Pro-
cesses, V.F. Zackay and H,I. Aaronson, eds., Interscience, New
York, NY, 1962, pp. 353-70.
30. K. Okamoto and M. Oka: Metall. Trans. A, 1986, vol. 17A,
pp. 1113-20.
31. D.N. Shackleton and P.M. Kelly: Physical Properties of Martensite
and Bainite, Special Report 93, Iron and Steel Institute, London,
1965, pp. 126-34.
32. H.I. Aaronson, M.R. Plichta, G.W. Franti, and K.C. Russell:
Metall. Trans. A, 1978, vol. 9A, pp. 363-71.
33. Y.A. Bagaryatski: Dokl. Akad. NaukSSSR, 1950, vol. 73, p. 1161.
34. W. Pitsch and A. Schrader: Arch. Eisenhuettenwes., 1958, vol. 29,
pp. 485-88.
35. P.M. Kelly and J. Nutting: Proc. R. Soc., 1960, vol. 259,
pp. 45-58.
36. J.W. Christian: Decomposition of Austenite by Diffusional
Processes, Interscience, New York, NY, 1962, pp. 371-86.
37. G.R. Srinivasan and C.M. Wayman: Acta MetaU., 1968, vol. 16,
pp. 621-36.
38. K.R. Kinsman, E. Eichen, and H.I. Aaronson: Metall. Trans. A,
1975, vol. 6A, pp. 303-17.
39. M.G. Hall, H.I. Aaronson, and G.W. Lorimer: Scripta Metall.,
1975, vol. 9, pp. 533-42.
40. H.J. Lee and H.I. Aaronson: Acta Metall., 1988, vol. 36,
pp. 787-94.
1390--VOLUME 21A, JUNE 1990
41. H.I. Aaronson and C. Wells: Trans. AIME, 1956, vol. 206,
pp. 1216-23.
42. K.W. Andrews: J. Iron Steel Inst., 1965, vol. 203, pp. 721-26.
43. G. Spanos, H.S. Fang, D.S. Sarma, and H.I. Aaronson: Metall.
Trans. A, 1990, vol. 21A, pp. 1391-411.
44. E.S.K. Menon and H.1. Aaronson: Acta Metall., 1987, vol. 35,
pp. 549-63.
45. G. Spanos and H.I. Aaronson: Scripta Metall., 1988, vol. 22,
pp. 1537-42.
46. R.W.K. Honeycombe: Steels: Microstructure and Properties,
ASM, Metals Park, OH, 1982, pp. 106-20.
47. R.W.K. Honeycombe: Metall. Trans. A, 1976, vol. 7A,
pp. 915-36.
48. Y. Ohmoti, H. Ohtani, and T. Kunitake: Trans. Iron Steel Inst.
Jpn., 1971, vol. 11, pp. 250-59.
49. H.I. Aaronson: J. Microsc., 1974, vol. 102, pp. 275-300.
50. R.E. Reed-Hill: Physical Metallurgy Principles, Van Nostrand
Reinhold Company, New York, NY, 1973, p. 683.
51. J.M. Rigsbee and H.I. Aaronson: Acta Metall., 1979, vol. 27,
pp. 365-76.
52. M. Oka and K. Okamoto: Proc. Int. Conf. on Martensitic
Transformations, The Japan Inst. of Metals, 1986, pp. 271-75.
53. J.S. Bowles and C.S. Barrett: Prog. Met. Phys., 1952, vol. 3,
pp. 1-41.
54. B.P.J. Sandvik: Metall. Trans. A, 1982, vol. 13A, pp. 789-800.
55. M. Mannerkoski: Acta Polytech. Scand., 1964, ch. 26, p. 27.
56. D.A. Porter and K.E. Easterling: Phase Transformations in Metals
and Alloys, Van Nostrand Reinhold Co., New York, NY, 1981,
p. 339.
57. H. Yada and T. Ooka: J. Metall. Soc. Jpn., 1967, vol. 31,
pp. 771-76.
METALLURGICAL TRANSACTIONS A