Acra merall. Vol. 32, No. I, pp. 137-146. 1984 0@%6160/8433.00 +o.

oo
Printed in Great Britain. All rights ~~~tved Copyright Q 1984Pergamon Press Ltd

A NEW KINETIC MODEL FOR ANISOTHERMAL

METALLURGICAL TRANSFORMATIONS
IN STEELS INCLUDING EFFECT OF
AUSTENITE GRAIN SIZE

J. B. LEBLONDf and J. DEVAUX

Centre de Calcul de Framatome, Etablissement de Chalon sur Saone, 71380 Saint Marcel, France

(Received 18 May 1983)

Abstract-In this paper, we propose a new kinetic model for anisothermal metallurgical transformations,
valid for ditYusion controlled as well as martensitic transformations. It is first described in the case of two
phases and a unique transformation, then extended to the case of n phases and several (possibly
simultaneous and unsymmetrical) transformations. The effect of austenite grain sixe can also be inchrti,
an evolution equation for austenite grain sixe is then needed. We show that the classical equation is limited
to the case of constant au&mite proportion, and propose a reasonable extension of this classical equation
to the general case.

R&mm&Dam cet article, nous proposons un nouveau modele cinCtique pour la description des
transformations m&alhsrgiques anisothermes dans les aciers, valable aussi bien pour les transformations
contr&es par la diffusion que pour les transformationsmartensitiques.II est d’abord d&tit darts le cas
de deux phases et d’uae transformation unique, puis etendu au cas de n phases et de plusieurs
transformations (6ventuellement simultan&s et non sp&triques). On peut aussi tenir compte de l’influence
de la taille du grain aust&itique; en ce cas, on a besoin dune equation d’evolution pour cette taille de
grain. Nous montrons que I’Cquation classique est limit&e au cas dune proportion d’austtnite constame,
et proposons une extension raisonnable de cette Equation classique au cas g&&al.

Zusammenfassrmg-In dieser Arbeit wird ein neues kinetisches Model1 fiir nicht-isothenne metallurgische
Umwandlungen vorgeschlager. Es wird xuerst fiir den Fall von xwei Phasen und einer einxigen
Umwandhmg beschrieben. danach erweitert fir den Fall von n Phasen und mehreren (m6glichenveise
gleichxeitigen und nichtsymetrischen) Umwandlungen. Der EinfluB der KomgriiBe des Au&nits kann
ebenfallseingeschlossenwe&n; dams ist jedoch eine Gleichungfur die Entwicklung der KomgriiSe n6tig.
Es wird gexeigt, daB die klassische Gleichung auf den Fall eines konstanten Austemtanteils besch&tkt
ist. Diese klassische G&hung wird auf den allgemeinen Fall erweitert.

1. INTRODUCTION
The amount of work devoted to the description of the
kinetics of isothermal metallurgical transformations
in steels is enormous. From an industrial point of
view, however, anisothermal transformations are
much more important; models are needed to describe
such transformations, for instance to evaluate numer-
ically the metallurgical structure as a function of
position and time during quenching or welding oper-
ations. Such models are necessarily empirical as the
problem of anisothermal transformations seems to be
too complex to be theoretically tractable. In the
present paper, we describe a new model which is both
simple and general (for instance it applies to diffusion
controlled as well as to martensitic transformations).
It is given under a form suitable for computer
calculations.
tAlso with the Laboratoire de Mecanique des Solides,
Ecole Polytechnique, route de Saclay, 91128 Palaiseau,
France.
137
We tirst consider the case where there are only two
phases 1 and 2 (the proportion of phase 2 is denoted
2.) and a unique transformation 1-* 2. We then show
that the use of existing models [l-7) raises some
difficulties due to the fact that they do not take into
account the existence (for each temperature T) of an
“equilibrium proportion z,(T)” (to be defmed later).
We propose a very simple model integrating this fact,
which is shown to solve most difficulties.
We then generalize this model to the case of n
phases and several (possibly simultaneous) trans-
formations; attention is paid to the fact that most
transformations are “unsymmetricaT’ (i.e. the kinet-
ics of the reverse transformation is not the same than
that of the direct one).
The model is next extended to the case when the
austenite grain size D hasan effect on the kinetics.
The evolution equation for the proportions of the
different phases must then be coupled with an evo-
lution equation for D.
It is shown that the commonly used equation for
D [8-lo] is limited to the case where the austenite
138 LEBLOND and DEVAUX: A KINETIC MODEL FOR ANISOTHERMAL
proportion is constant; this equation is extended to
the case where the austenite proportion varies.
The described models have been integrated to the
TITUS finite element program. A numerical com-
putation of metallurgical structure and stresses dur-
ing the stainless strip-cladding of a low-alloy plate is
reported elsewhere [I 11.
2.THECASEOF-lWOPHASESlAND2AND
A UNIQUE TRANSFORMATION I-, 2
Let us reason first for simplicity in the case
of a (ferrite + pearlite) 3 (austenite) transformation;
phases 1 and 2 refer thus respectively to ferrite +
pearlite and austenite, the austenite proportion is
denoted z.
The existing models for the evolution of z can be
divided into three classes:
(a) The “additivity rule” [ 1,2] states that if r (z, T)
is the time which is necessary to reach the proportion
z at constant temperature T (starting from z = 0), one
reaches the proportion z in an anisothermal process
after a time r such that
(b) Some authors [3,4,6] suppose that i can be
written in a factorized form
i =/(T)g 0).
In that case the transformation is called isokinetic in
Cahn’ sense [3]. It is shown in [3] (p. 494) that the
additivity rule (a) then holds.
(c) Some authors [7] suppose that i depends only
on z and T, but in a more general form
i =f(T, z).
It is relatively easy to show that (b) cannot be of a
genera1 applicability (i.e. one can determine functions
f and g for a given thermal history T(f), but these
functions will soon be found to become inadequate if
one changes the thermal history). Indeed, if that was
the case, equation (b) should, in particular, hold (at
least qualitatively!) for isothermal processes. But
there exists for each temperature T an “equilibrium
proportion z,(T)“; this is the proportion which is
reached at constant temperature after an infinitely
long time; z,[T(r)] is also the austenite proportion in
the case of an extremely slow heating. It is determined
by the fact that the free enthalpy of the
(ferrite + pearlite)-(austenite) mixture is a minimum
at the temperature T for this particular proportion
z,(T). It can be read from an iron-carbon diagram
(incorporating effects of alloying elements). This
tThat (b) is not satisfactory can also be deduced from the
fact that (b) implies (a), which is itself not satisfactory
(see below).
TRANSFORMATIONS
means that at constant temperature T
z =z,(T)(=cst)
must be a stationary solution of equation (b), which
means that
Let us restrict to the temperature interval where the
transformation occurs; then
f(T)+0
(otherwise no transformation can occur). This implies
g [zcpWI = 0.
But Z~ (T) takes every value comprised between 0 and
1 when T varies; so g is identically zero, which is
absurd (it means that no transformation can occur).?
A very simple example will show that the additivity
rule (a) suffers from the same kind of drawback. Let
us suppose that i depends only on T and z and that
the isothermal kinetics at temperatures T, and T2 are
of the kind indicated in Fig. 1. The time needed to
reach z = 1 is equal to r at T2 and is infinite at T,.
Let the material undergo the thermal history indi-
cated in Fig. 2. Then the time evolution of z is the
same than that at constant temperature T2; z = 1 is
reached after a time r. The additivity rule would give
elapsed time at T,
time necessary to reach z = 1 at T,
elapsed time at T2
+
time necessary to reach z = 1 at T2 = ’
r/2 r/2 1
i.e.
m+-=i=l
t
which is false. The difficulty comes from the fact that
the transformation is uncomplete at T,; it is therefore
connected with the existence of a z,(T) # 1 at
T, tz,(T,) = l/21.
It appears thus that approaches (a) and (b) are of
limited applicability (i.e. they cannot be valid, even
qualitatively, for widely varying thermal histories),
T/2 T
t
Fig. I. Supposed isothermal kinetics at temperatures r, and
r,. The time needed to reach z = I is equal to T at 7’*and
infinite at T, (the transformation is uncomplete at this
temperature).
 LEBLONDand DEVAUX: A KINETIC MODEL FOR ANISOTH~RMALTRANSFO~ATIONS 139

solution at constant temperature T {i.e.

f[T, z,(T)] = 0 for each Tf. This is obtained by
supposing that z is not too far from z,(T), so that

:JT) -z
x (2 -z,(T)] =
r(T)
where
1
- = -g-t*, z,(T)].
Fig. 2. Supposedthermal history. The transformation kin- * (T)
etics is the same than if the tcmpcratun was kept equal to
‘I’*;the additivity rule is not valid. The evolution equation is then

z=: z,(T) - z
=(T)
because they do not take into account the existence
of z,(T). On the other hand, approach (c) is corn- it involves two functions of T: z,(T) and t(T). The
patible with the existence of z,(T) iff(T, z) verifies former is comprised between 0 and 1. The latter is
necessarily positive [otherwise z would not tend to-
f&7 z,(T)1 = 0 wards zg(T) but depart from it]; it represents the
for each T [so that z = z,(T) is a stationary solution characteristic time of the transformation [at constant
of the evolution equation at constant temperature T]. temperature T, z tends exponentially towards z,(T)
Some authors [7Jhave used approach (c), but their with a time constant equal to 7 (T)].
function f (T, z) is given by The predictions of this equation are q~~tively
described in Fig. 3. For very slow heating rates, z has
f(T,z)= K(T)z”(J’(l -z>p(n
the time to almost reach z,(T) for each temperature
consequentlyf (T, z) is zero only for z = 0 and z = 1 T (“quasistatic” evolution); for increasing heating
(~su~ng 151(T)# 0, otherwise there is no trans- rates this becomes more and more false so that z
formation). Other authors [q have used the following follows z,(T) with an increasing retardation effect.
generalisation of Johnson and Mehl’s kinetics This is well in accordance with physical reality.
To determine parameters z,(T) and r (T) one can
z=l-exp - ‘f[T(t’)](t - t’)‘dr’ proceed as follows. First, z,(T) can be deduced
IS 0 1 either from an iron-carbon diagram incorporating
effects of alloying elements or from dilatometry
which gives upon derivation
curves at very slow heating rates, assuming that the
length I(T, z) of the specimen is (1 -z) t,(T) -i-
i =exp - ‘f(r(Q](t - r’r dr’ z 12(T), where f,(T) and f,(T) are the lengths (at
II 0 I temperature T) of specimens composed exclusively of
phase 1 or of phase 2. Next, an order of magnitude
x 'f[T(f)]3(t - t’)2dt’ of z can be determined by using the fact that the raise
s0 of the AC, (or AC,) temperature due to the non-
=(l -2) 'f[Z-(i')] 3(1 - r’)2dt’
I0
so that i depends on all previous temperatures T(r’),

1’ 6 t; but i is zero only if z = 1.

Difficulties encountered in using approaches (a)
and (b), or (c) without imposing the condition
f [T,z,(T)] = 0 are not only theoretical but also Z

practical: for instance Giusti [6] showed that it is

necessary to add an empirical multip~~~ve factor 1
(dependent on the previous thermal history) to the‘
equation i = f (T) g(z) to obtain an agreement
between experimental and theoretical results. Figure At 43

2 of Ref. (51exhibits also a rather poor agreement T

between experimental and theoretical results. Fig. 3. Qualitativepredictions of the proposed model. At
We now look for the simplest evolution equation very low heating rates, r(l) follows the I [T(r)] curve. At
of type (c) ensuring that z = z,(T) is a stationary higher heating rates, a retardation c8xt appears.
140 LEBLOND and DEVAUX: A KINETIC MODEL FOR ANISOTHERMAL TRANSFORMATIONS

Table 1. Experimental and theoretical AC, and AC, temperatures Table 2. Parameters z,(T) and r(T) for the austcnitic trans-
as functions of the heating rate for the A508 cl. 3 steel formation of the A508 cl. 3 steel (n.b.: 4(T) and r(T) must be
linearly interpolated between the mdicated points)
Heating
rate AC, exp AC, th. AC, cxp AC, th. Temperature Temperature
_j”C s-v (“C) (“c) (‘C) (“0 (“C) % (“C) (:)
I 730 735 840 835 0 0 0 I
I5 750 745 845 840 730 0 730 1
30 755 750 850 845 840 I 830 0.2
70 760 760 870 855 1000 I 900 0.05
110 765 765 870 Kltm on5
130 770 770 880
500 775 790 540

pcrature A,); for instance the transformation stops

zero 7’ is roughly L%:indeed, roughly speaking, the
transformation does not start instantaneously when
T reaches A, = AC,[(? = O)] but after a time T;
at that time the temperature is Ai + h so that
AC, (f’) N A, + i%. A complete determination of
r(T) can be achieved by trying to reproduce with
various r(T) functions the experimental values of
AC, and AC, for several heating rates.
Table 1 exhibits the correlation which was ob
tamed between experimental and theoretical AC, and
AC, temperatures in the case of the A508 cl. 3 steel,
with the parameters shown in Table 2t.
It appears that this very simple model is able to
reproduce satisfactorily experimental results for a
wide range of heating rates. Another point of interest
is the fact that the r(T) function shown in Table 2
is a decreasing function of T (this is because numer-
ical experience shows that if 7 did not decrease with
increasing T, retardation effects would be much too
marked for temperatures around AC’,, i.e. AC’,(?)
would be much too dependent on 0, this is in
accordance with the fact that the transformation is
thermally activated.
Let us now consider the case of the (aus-
tenite) -P (ferrite + pea&e) transformation; phases 1
and 2 now refer respectively to austenite and
ferrite + pear&e; z is the proportion of ferrite +
pea&e. The same reasoning as before lead8 to the
same equation, where z4(T) is now the equilibrium
proportion of ferrite + pea&e.
Figure 4 shows a XT-diagram for the A508 cl. 3
steel according to Le Creusot laboratory. It appears
that the start temperature for the ferritic-pearl&
transformation is much more sensitive to ?’than AC,
and AC,; this means that one will have to use much
larger r-values than for the austenitic transformation.
This comes physically from the fact that the
ferritic-pearlitic transformation is much slower be-
cause it implies carbon diffusion. What is more, Fig.
4 shows that even for very slow cooling rates
(f = - 10°Ch-l), the transformation is so slow that
z does not follow the theoretical z,(T) (i.e. the one
corresponding to minimum free enthalpy, i.e. the one
increasing from 0 at temperature A, to I at tem-
tA precise definitionis needed for AC, and AC, as the
transformation is
in theory never complete (z tends
asymptotically towards I). It was assumed that the AC,
point corresponds to z = 0.05 and .4C, to ,- = 0.95.
only at 615°C and not at A, N 730°C.This means that
the model must be used with an “effective” z,(T)
(determined from dilatometry curves or CCI dia-
grams for very slow cooling rates), which does not
coincide with the theoretical one [because the kinetics
is so slow that the theoretical z,(T) can never be
approached].
Table 3 displays the correlation between experi-
mental and calculated values of start and finish
temperatures and final proportions of
ferrite+pearlite, for the ferritic-pearl& trans-
formation of the A508 cl. 3 steel, using parameters
given in Table 4. The quasi-inIinite value of r at
600°C is necessary because Fig. 4 shows that the
transformation stops always at 615”C, even when it
is very uncomplete. This sudden rise of r as T
decreases is of course schematic and could be reI!ned;
it is nevertheless physically plausible that T increases
when T decreases.
The proposed model is also easily applied to the
austenite -+ martensite transformation. For most
steels, the martensite proportion depends only on T
and is given for instance by Koistinen and Mar-
burger’s formula (121. This is obtained by putting
z,(T) equal to Koistinen and Marburger’s formula
(here z = martensite proportion) and using a very
600
200 -
100 I I I I J
IO5 IO' d 102 IO
Cooling rote ot 700% t’Ch-‘J
Fig. 4. CCT diagram for the AS08 cl. 3 steel (austenitisation
conditions: 900°C. I h).
 LEBLOND and DEVAUX: A KINETIC MODEL FOR ANISOTHERMAL TRANSFORMATIONS 141

Table 3. Experimental and theoretical values of the start and finish tcmpxaums and of the firmi
&rite-pcartite proportion for the f~riti~~~it~ t~n~o~ation of the A508 d. 3 st& as fusions of the
coolina rate (at 700°C)
Start Start Finish Finish Final Final
Caoling tcmpcrdlure temperature temperrture temperature J-P f-P
rate (cxp) (th.) (cxp) (th.) proportion proportion
(“C h-‘) (“C) (“C), (“C) (“C) (cnp) (th.)
-10 ..-
800 795 615 - 626 1.00 1.00
-100 775 780 615 620 0.85 0.83
-lMKl 74s 740 615 620 0.45 0.40

Table 4. Parameters z_,(T) and t(T) for the ferriticpcarlitic trans- 0 for the 24 1 transformation:
formation of the A508 cl. 3 steel (n.b.: +JT) and r(T) must be
linearly interpolated bctwccn tbc mdieatcd points)
Temperature Temperature
(“0 L.. (“C) i,
where p&(T) and z 12(T) are the equilibrium propor-
0 I 0 10,000
boo I boo 10,000
tion of phase 2 and the characteristic time, for
800 0 620 500 transformation 1-t 2, andp:&(T) and ~~‘(T)are the
1000 0 1000 500
equilibrium proportion of phase 1 and the character-
istic time, for transfo~ation Z-P 1.. The lack of
small value for T ensuring that there are almost no symmetry means that z”(T) # TV’. Moreover, in
retardation effects (this small 7 value is compatible theory, the equilibrium phase proportions are deter-
with the fact that the transformation is very quick). mined by thermodynamic considerations and should
The kinetics depends also on t for certain steels: this not depend on the direction of the transformation,
case can be dealt with by using larger t-values. One i.e. one should have
sees thus that although the definition of z,(T) raises
some theoretical problems for the austenite--, P::(T) + P:&(T) = 1.
martensite transformation (because the martensitic Nevertheless it has been shown in Section 2, that one
phase is very far from thermodynamic equilibrium), may have to use “effective” equilibrium proportions
the model can nevertheless be successfully applied in different from the theoretical, thermodynamically
this case too by using approp~ate vahms for z,(T) determined ones. In that case, the effective equi-
and r(T). The same formalism applies thus to very librium proportions are determined by kinetics, so
different transformations: austenitic, ferritic-pearlitic that they depend on the direction of the trans-
and martensitic [the case of the bainitic trans- formation; i.e. the above relation does not hold.
formation will be considered in Section 3c)]. The problem is to give a criterion for using one
evolution equation or the other, i.e. for deciding
whether the transfo~ation takes place in the I + 2
3. THE CASE OF II PHASES AND SEYRRAL
(POSSIRLY !3MuLTANEOus AND direction, or the opposite one, or does not occur. It
UNSYMMRTRICAL) TRANSFORMATIONS is easily derived by noting that equation (1) is valid
only for the l--r 2 transformation, i.e. for fi2 > 0, i.e.
We will proceed in three steps: generalization to the for p2 <p&,(T); similarly equation (2) holds only for
case of 2 phases, 1 and 2, and 2 unsymmetrical pt <p&(T), i.e. for p2 > 1 -p&(T). One is thus
transfo~ations, 1 -B 2 and 2-+ 1; rewriting the lead to the idea of distin~shing between 3 regions
equations under a form more convenient for further as shown in Fig. 5
generalization; generalization to the n phases case.
0 in region I, where p2 > 1 -p&(T), the evo-
lution equation is
(a) The case of 2phases, 1 and 2 and 2 unsymmetrical
t~~forrnat~o~~ 1 --, 2 imd 2 -+ I p.t =P:&(T)-PI
Let pi and p2 denote the proportions of phases 1 x2’(T)
and 2 (p, +p2 = 1). The evolution equation can be (the transformation occurs in the 2 -+ 1 direction)
written: 0 in region III, where p2 <p&(T), the evolution
0 for the 13 2 transformation: eouation is

(the transformation occurs in the 1 -+ 2 direction)t

Wtese first two eases are mutually incompatible because l in region II where p&(T)Sp26 I -p:&(T),
p*“(T) is necessarily inferior or equal to 1 -p,*‘(T) the transformation does not occur (p, =a = 0).
(o&erwise physical absurdities would arise: in Fig.? for
instance, line GE would be under line BD so that the Figure 5 illustrates q~li~tiveIy the predictions of
ferritic-peariitic transformation would take place at that model for the austenitic and ferritic-pearlitic
higher temperatures than the austenitic transformation). transformations, during heating, cooling and re-
142 LEBLOND and DEVAUX: A KINETIC MODEL FOR ANISOTHERMAL TRANSFORMATIONS

heating (phase 1 = ferrite-t pearhte, phase

2 = austenite).
Several points are noteworthy:
0 The austenitic (I -3 2) transfo~ation occurs
not only during heating (portion BC) but also during
the beginning of cooling (portion CD). This is be-
cause at the end of heating (point C) p2 is still inferior
to p&,(T) (the equilibrium is not reached); the
austenitic transformation stops at point D where pz
reaches p&(T). This effect is notable only if the
heating rate is quite quick, otherwise p2 is only little T

inferior to p&(T) so that the austenitic trans- Fig. 5. Qualitative predictions of the proposed model. Three
formation stops quickly after the end of heating (D regions must be distin~sh~ I, where a 24 1 trans
is very near C). formation occurs G2 x0); II. where there is no trans-
0 During portion DE, no transformation occurs formation & = 0); III, where a 1-D 2 transformation occurs
@ > 0).
~hysi~ly this is because the t~~~tu~ is too high
for the ferritio-pearlitic transformation to take
place). This causes a sort of “hysteresis loop” in the [the determination of z,(T) and t (T) is thus com-
T-p2 curve. pletely equivalent to that of k(T) and I(T)].
0 If the austenitic transformation is not complete If both I-+ 2 and 2 -+ 1 transformations occur, let
(p* # 1 at point D), the ferritio-pearlitic trans- us denote k,2(T) and I,,(T) the data relative to the
formation starts at a lower temperature (point E) 1+ 2 ~sfo~ation, and k&T) and /a,(T) those
than in the ease of a complete austenitic trans- relative to the 2-r 1 transformation, i.e.
formation (dotted curve on Fig. 5; the ferritic-
pea&tic notation starts then at point ~33. k,,(T) -P:&(Wft2t0
0 Similarly the ferritio-pea&c transformation
k,(T) -P:&(Ws2’(~)
occurs not only during cooling (portion EF) but also
during the beginning of reheating (portion FG); there h,(T) = fl -P:2,vN~~‘2(~)
is no transformation during portion GH; and if the
12,(n = 11 - P:!Bqml172’m.
ferritic-pearlitic transformation is not complete
(pa#O at point G) the austenitic transformation The evolution equations of section 2(a) are equiv--
starts at a higher temperature (pomt H) than in the alent to
case of a complete ferritic-pearlitic transformation
lit = - k,*(TlP, + 42cnP2
(point B).
iti = h2WPI - 42cw2
All these predictions are physically satisfactory, cx-
cept the lower starting temperature for the i if k,2(T)p, - I,2(T)pr > 0 (14 2 transformation)
fe~ti~~~tic notation in the case of an
uncomplete austenitic transformation; this point will li, = k2, mp, - 4, COP,

be discussed later (Section 4). Jj2”= - k2, (UP2 + ~2,mPI

(b) Rewriting the equations in M equivalent form more i if ka,(T)p2 - Z2,(T)p, > 0 (2 --+ 1 transfo~ation)
suitable for further generalization dl =O
Let us consider first the case of a unique I--, 2 Ij2=0
transformation. The evolution equation
if h2Wp1 - h2Vlp2 d 0

and k2, (7’1~2 - It, (TIP, Q 0

is equivalent to Ic(no transformation).
* = - k (TIP, +
Pi OPT
(c) generalization to the n phases case
d* = kV)p, - [COP2
Let p,, . . . p,, denote the proportions of the
where p, and p2 are the proportions of phases 1 and different phases (Xp, = 1). Let A, denote the algebraic
2 (PI = 1 -z,p2=z), and proportion of phase i which is transformed into phase
 LEBLOND and DEVAUX: A KINETIC
Fig. 6. Supposed thermal history. The transformation of
austenite during cooling CD is complete so that the state of
the material is the same at points A and E. Furthermore
heatings AB and EF are identical; the austenite grain size is
therefore the same at points B and F.
j per unit of time (A, > 0 for a i +j transformation,
A, c 0 for a j ---) i transformation; Ali = - A,/).
We then have
pi = - c A, (i = 1,2, . . . n).
j#i
The problem is to give an expression for A,. The
simplest assumption consists in supposing that the
i +j transformation is not affected by phases k, k # i
and k #j, and is given in terms of pi and p, by an
expression of the same kind as before
c k,/(Tlp,- W-lfi
if k#‘lp, - &“lpj > 0
(i + j transf.);
-k,i(T)P/ + !/t(T)P,
Au=< if k,dT)p,-fJT)pt>O
(j ---*i transf.);
0 if k&Q, - f&‘Yp; G 0
and k,dT)p, - I/(T)p, G 0
I (no transformation be@een phases i and j).
This applies to simultaneous transformations (for
instance those of ferrite + pearlite, bainite and mar-
tensite into austenite) as well as to successive ones
(for instance the ferritic-pearlitic and the bainitic
transformations). Table 5 gives the k&T) and Iv(T)
functions for the austenitic, ferritic-pearlitic and
martensitic transformations of the A508 cl. 3 steel [as
deduced from the Z~ (T) and T (T) functions for these
transformations by the formulas given in Section
3 (b)l.
The case of the bainitic transformation was not
considered in Section 2 because it cannot be treated
by assuming that there are only two phases (the
ferrite+ pearlite proportion is not zero during the
bainitic transformation; what is more, the martensitic
transformation can interfere with the bainitic trans-
formation). It can be handled with the present for-
MODEL FOR ANISOTHERMAL TRANSFORMATIONS 143
malism. Nevertheless, for the A508 cl. 3 steel, we
found it difficult to obtain a good agreement between
experimental results and theoretical predictions for
such a wide range of cooling rates as shown in Fig.
4; this is very probably due to the fact that the
mechanisms of bainite formation are quite different
atlow(zSOCh-‘)andhigh(=5.104”Ch-’)cooling
rates. A much better agreement (shown on Table 6)
was obtained when assuming that the k, and 1,
functions depend not only on T but also on f, under
the following factorized form
k,(T, f) =Jj(T) II,(~)
fv( T, ?‘) = gv( T) h,(t).
Table 7 displays theJj( T), go(T) and h,(T) functions
which were used.
This “extended” model reproduces thus quan-
titatively all features of the CCTdiagram in Fig. 4;
it also gives predictions (which should be at least
qualitatively correct) for arbitrary thermal histories,
even quite different from those considered in the
(XT-diagram (abruptly varying heating or cooling
rates, uncomplete heating or cooling etc. . . ).
As an example of such predictions, let us consider
the case where the material is quickly cooled down to
375”C, and then kept at that temperature. The ba-
initic transformation is negligible during the quick
cooling; the martensitic transformation starts at
400°C and stops quasi-instantly when the cooling
stops (because the martensitic transformation is so
quick that the “equilibrium proportion” is reached at
every instant). Then the remaining austenite is (much
more slowly) converted into bainite.
It is quite probable that the proposed model can
also be successfully applied to annealing and liq-
uefaction.
4. THE INFLUENCE OF THE
AUSTENITEGRAIN SIZE
Thisinfluence has been disregarded up to the
present point [all transformation parameters (for
cooling) were determined for the same austenitisation
conditions: 9OO”C,1 h]. It is necessary to include it to
obtain such effects as the well known fact that the
temperature range of the bainitic transformation gets
higher and higher when the austenitisation gets less
and less pronounced. Indeed, we have seen in Section
3(a) that if the austenite grain size influence is
disregarded, an opposite effect is obtained: the ba-
initic transformation starts at lower temperatures in
the case of an uncomplete austenitic transformation
than in the case of a complete one.
For that purpose, we note that the d,‘s are always
given in terms of the A,,3 by the same formula as in
Section 3(c); and we assume that the A,‘s depend on
thepis in the same manner as in Section 3(c), but that
the ku and fq constants depend not only on T and
(possibly) on 1”(case of the bainitic transformation in
144 LEBLOND and DEVALJX: A KINETIC MODEL FOR ANISOTHERMAL TRANSFORMATIONS

Tabk 5. Parameters k,,(T) and k,(T) for the austcnitic, ferritic-pearlitic and martensitic
transformations of the A508 cl. 3 steel (ferritcpcarlite = phase I; bainite = phase 2.
martensite = phase 3. austenite = phase 4)

Austenitic transformation’
-rcmperaturc
t-3 0 730 750 770 790 810 830 840 860 880 900 1000
k
$s-'1 0 0 0.22 0.53 1.05 2.02 4.55 5.6 7.37 10.77 20 20

I
&s-l) 1 1 0.97 0.94 0.87 0.76 0.45 0 0 0 0 0
1%
‘It is assumed that the three transformations: ferrite + pearlitc-raustcnitc. bainitc-r
austcnitc. martcnsitc+aurtcnite an idcntic91; thus k,,. k, and k,, arc identical and
so ara I,,. I,. I,.
Ferritic-peariitic transformation

Temperature
(‘c) 0 600 620 800 1000
kr, (s-l) 0.0001 O.OcKtl 0.0018 0 0
I,, (s-l) 0 0 0.0002 0.002 0.002

Martcnsitic transformation*

-rcmpcrature
(“c) 0 250 270 290 315 340 370 400
50 48.5 45 38.5 30 16.5 0
1.5 5 11.5 20 33.5 50
%e k,, and I,, valuer were obtained by taking a very short chara&&ic time 7 (7 = 0.02 I),
ensuring that the martcnsitic proportion I follows z,(T) with ahnost no retardation
affects even for cooling rates as high as lo”“Ch-‘.

Tabk 6. Experimental and theoretical values of the start and finish temperatures and of the final
bainita proportion, for the bainitic transformation of the A508 cl. 3 steel, as functions of tbe cooling
rate’
Start Start Finish Finish Final Final
cooling tanpcnture tempcraturc temperature temperature f-P f-P
(CXP) proportion proportion
IOEi) (“Cl (CXU) (th.)
-70 490 470 380 390 0.15 0.11
-700 540 530 355 350 0.55 0.50
-1500 550 535 Et 360 0.75 0.72
-7200 550 535 400 1 1
-15,000 520 520 -2 0.76
-43NtO 435 450 1 :::5 0.10
‘We mean the cooling rate at z 5OOT, where the transformation ooxrs and not that at 700°C like
on Fig. 4; to determine the above mentioned quantities as functions of f (5OOT). we used the
CCT diagram of Fig. 4 by assuming that f (5OOT) was for the thermal histories used in the
diagram, about 0.67 times f (7OOT).
‘At these cooling rata the bainitic transformation is interrupted by the martasitic transformation,
and its 6&h temperature is not known with a good precision.

A508 cl. 3 steel), but &o on the austenite grain size
D; parameters kt axtd I, must then be determined for
different CXXdiagrams (corresponding to different
austenite grain sizes).
The evolution equation for the p,‘s must then be
coupled with an evolution equation for D; the prob-
lem is to tind that equation. Several authors, for
instance Alberry and Jones [S] and Ikawa et al. (91
have used the following very plausible equation
$ (Da) =C exp -$
( >
where a, C and Q are positive constants and R is the
gas constant.
Ashby and Easterling [IO] have proposed a more
general equation
6 =fWev(-Q/W.
They have also extended this expression to include
effects of carbides and nitrides which prevent grain
growth.
It is easy to see that these equations are limited to
the case where the austenite proportion does not
vary. Let us consider for instance the thermal history
shown in Fig. 6.
Application of either one of the prccceding equa-
tions yields the following unsatisfactory result: D
increases always so that D is greater at point F than
at point B, whereas it should have the same value at
both points; what is more, the value of D at point F
 LEBLOND and DEVAUX: A KINETIC MODEL FOR ANISOTHERMAL TRANSFORMATIONS 145

Table 7. Parameters f&t. g&t, II,@) for the bainitic trans- where d V(C) is the volume occupied at i + di by the
for~tion of the A508 cl. 3 steel(bainite = phase 2, austcnite = austenite formed between f and I +dr.
phase 4)
Assuming that this newly formed austenite has a
Temperature
(“C) 0 340 350 450 550 1000 zero initial grain size
___.
0 0 0.014 0.067 0 0
0 0 0 0 0.067 0.067

Cooling
rate
(“C h-‘) -43,000
--__
-I5,000 -7200
-..._-
-1500 -700 -70
=(I-Gg)( D”(t) +; (D”(r)) dr
),,,
h,z 0.2 1 1.5 0.22 0.1 0.0044

is in&need by that at point E which has no real

physical meaning (there is no austenite at that Point).
The problem comes from the fact that the austenite
at point F is some newfyfirmedausteaite whose grain
size has nothing to do with that of the (hypothetical)
“old” au&mite at point E.
Q
We thus look for a general&i evolution equation =Cexp -- -ZIP(r)
for the grain size, valid even if the austenite propor- ( RW) )
tion z varies. The considered grain size can only be or equivalently
an uneruge grain size because there is no unique grain
size in the general case where z varies. Indeed, let us
consider for instance the case where the material is
re-austenitized after an uncomplete martensitic trans-
formation (i.e. there remains some “residual’” aus- (6) Austenire is disappearing(V decreases with time,
tenite at the beginning of reheating). At the end of i < 0)
reheating the material is a mixture of two different Now b“(r) can be expressed as
kinds of austenite: “aid” austenite which existed
already at the beginning of reheating, and whose p(r) = <WDvta
grain size has kept growing, and “new” austenite
formed during reheating, with smaller grain size.
We assume that on a microscopic scale, b is given
by the above equations; we use the Alberry-Ioaes- +z(r)-z(t +dr)
Ikawa formula for ~rnp~~ty, but the present reason- <~‘@))dY(r)
z(Q
ing could easily be extended to the case of Ashby and
Easterling’s more general formulation (including where dV(r) is the volume occupied at z by the
effects of pinning carbides and nitrides). We consider austenite disappearing between r and r + dr. As-
the following average grain size suming that this disappearing austenite has the same
ti = (IY)? average grain size than the rest of the austenite
where ( )” denotes volume averaging over the vol-
ume V occupied by austenite, and we look for the
expression of #4
+2(r)-z(t +df)
<~4(~)>vct+drl
z(t)
Two cases must be considered
= <Da~rh’o+df)*
(a) Austenite is being formed (V increases with time,
i 3 0) On the other hand we have
At time t, P(t) is given by D’(t + dt) = (D”(f + drDYf,+~,)
dV(t) = <M&, whence
At time t + dr, D(t + dt) is given by
B’(t + dt) = (D”(t + dr)) fl,+d6
Thus there is no difference in that case between
= -$$&<W + W,

+z(r +dr)-z(t) The dissymmetry between the two cases comes

<Mr + dt)),U,j
Z(I + dr) from the assumptions that the average value of D’

A.M. m--I
146 LEBLOND and DEVAUX: A KINETIC MODEL FOR ANISOTHERMAL TRANSFORMATIONS

over volume dV is equal to zero if i 3 0 (newly denotes the algebraic proportion of phase i which is
formed austenite has a zero initial grain size), and to transformed into phase j Per unit of time)
(De), if i d 0 (disappearing austenite has the same
average grain size than the rest of the austenite). iii= - ;+
It is easy to see that the use of this new evolution
equation resolves the above mentioned problem. In-
c kgpf- I~pj if kept - l,~, > 0
deed, the evolution equation during the austenitic
transformation is (i-j transf.);
- kj@,+ lgi if kg, - I’,#~
>0
-jpP)=zCexp
( 1 -zT ,
A,= (j 4 i transf,);

i.e. Ir;=2 ~~+~~z(~)cexP(-~)d~ 0 if kgt--Zvp,,,<O

zr
where subscripts i and f indicate initial and final
values. Let us neglect for simplicity the austenite
grain growth during the austenitic transformation
(because the tem~rat~e is too low to allow for a
significant growth), then
Thus, if z,=O, d is equal to 0 at the end of the
a~te~tic ~sfo~ation whatever its initial value;
the subsequent values of d are thus independent of
6,. This means that the grain size is exactly the same
at points B and F on Fig. 6, which is physically
satisfactory.
Let us consider now the case where the material is
reaustenized after an uncomptete ~ansfo~a~on of
austenite during cooling (z, # 0).
The above equation then says that the value of 6’
at the end of the austenitic transformation is the
and kgj - 4~3,< 0
\.(no transformation between phases i and j).
For most transformations the ky and t, functions
can be assumed to depend only on T, but it was
found that for the bainitic transformation of A$08 ci.
3 steel an extra ?‘dependence must also be included.
This model was then extended to the case where the
austenite grain size has an influence on trans-
formation kinetics, by making the k,‘s and 1,‘sdepend
also on the grain size. The evolution equations for the
p;s must then be coupled with an evolution equation
for the gram size. We showed that the classical
evolution equation [&IO] is limited to the case where
the austenite proportion z is constanS that in the
general ease only an ayerrrge gram sixe d can be
calculattd; and that a reasonable generalization of
the classical equation (valid in all cases) is as follows

(
Cexp --ST -5 ‘6”
average of the ath power of the grain size of the “01P ( >

11
austenite which existed already at ii (i.e. dg) and the
ath power of the grain size of the newly formed d(& *
i.e. $(zP)=zCexp -& if i 30,
dt E (
austenite (i.e 0), with coefficientsequal to the voltic
proportions of these two kinds of austenite (i.e. z,/zr
C exp - +r if i < 0.
and (z,- z,)/z,). As the preceeding one, this prediction ( 1
is physically satisfactory.
5. CONCLUSION RJZPERENCES

J. W. Cahn, Acta metall. 4, 572 (1956).

In the case of 2 phases 1 and 2 and a unique
transformation 1 + 2, we have shown in Section 2
that the classical approaches [l-7] raise some the-
ore&at and practica1 difhculties; this had led us to
propose the following improved evolution equation
(where z denotes the proportion of phase 2)
z-(T) -z
‘= -i(T)
which takes into account the existence of the “equi-
librium proportion” z,(t). This equation can be
applied to diKusion-controlled transformations (such
as the austenitic transformation) as well as to mar-
tensitic transformations.
This equation was next extended to the case of n
phases and several {possibly simultaneous and un-
symmetrical) transformations in the following way
(where pi denotes the proportion of phase i and ‘A,
:: W. L. Bradley and R. Mattes. Mater. Sci. Enmz
__ 21,227
(1975). -
3. J. W. Christian, The Theory of Trapformations in
MetatS and Alloys. Pmtamon Press. oxford (1965).
4. T. Inoue and B. kanie&i, J. Me&. khys. So& 24,. 187
(1978).
5. T. Inoue, S. Nagaki, T. Kishino and M. Monkawa,
Ingenieur Arch. 50, 3 IS (I 98 1).
6. J.Giusti, These dc do&rat &tat, Paris (1981).
I. Centre de Recherches Metallurgiques de Lilge, Con-
vention C.C.E./CRM 7210, EAf2f205 (1977).
8. P. J. Alberry and W. K. C. Jones, Marchwood En-
aineerina Lab.. Internal Reuort No. R/M/R282 (1979).
9. H. Ikawa, S. Shin, H. Gshige and Y. ‘Mekuchi, .Tran;.
J.W.S. 8, 46 (1977).
IO. M. F. Ashby and K. E. Easterling, Acta metall. 30, 1969
(1982).
II. J. Devaux. D. Dubois and J. 8. Leblond, presented at:
7ih SMIRT Co&, Chicago (1983).
12. D. P. Koistinen and R. E. Marburger, Actu merall. 7,
59 (1959).