International Journal of Plasticity, Vol..~, pp. 573-591. 1989 0749-6419/89 $3.00 + .

00
Printed in the U.S.A. Copyright © 1989 Pergamon Press pie

MATHEMATICAL MODELLING OF
TRANSFORMATION PLASTICITY IN STEELS
Ih COUPLING WITH STRAIN HARDENING PHENOMENA

J . B . LEBLOND

Ecole Pol~echnique and Centre de Caicul de Framatome

(Communicated by George Weng, Rutgers University)

Abstract-The models for the plastic behaviour of steels during phase transformations proposed
in Part I and in a previous paper (LEsLOr~Det aL [1986bl) for the case of ideal-plastic phases
are extended to include strain-hardening effects (isotropic or kinematic hardening). An expression
for the transformation plastic strain rate is obtained by modifying the treatment of Part I in a
suitable manner. The classical plastic strain rate is also studied in a similar way. Complemen-
tary evolution equations for the hardening parameters are finally given, taking into account the
possible "recovery" of strain hardening during transformations (i.e., the fact that the newly
formed phase can "forget," partially or totally, the previous hardening).

!. INTRODUCTION

Pan I studied both theoretically and numerically the response of phase-transforming

steels to variations of the proportions of the phases (transformation plasticity), and also
incidentally to variations of the temperature (classical plasticity at constant applied
stress), in the case of ideal-plastic phases. A previous paper (Lr~sLONDet al. [1986b]) had
been devoted to the response to variations of the stress applied (classical plasticity at con-
stant temperature), also for ideal-plastic phases. This led to models describing accurately
the plastic behaviour during phase transformations in the absence of strain hardening
effects. The purpose of this paper is to extend these results by including these effects.
This generalization is quite desirable since at the temperatures of interest, the stress-
strain curves of individual phases are rather steep and exhibit generally no sharply de-
fined yield point.
Introduction of strain hardening phenomena makes the physical situation quite com-
plex because of couplings between them and transformation plasticity. Indeed strain
hardening will affect the microscopic plastic strain of the weaker phase and therefore
also the (macroscopic) transformation plastic strain, which arises at least partly, and of-
ten entirely, from this microplasticity (G~ENWOOD ~ JOHnsoN'S [1965] mechanism).
Conversely, occurrence of transformation plasticity implies microscopic plastic straining
of the weaker phase and will therefore induce an evolution of the hardening parameters.
To reach the desired result, two kinds of evolution equations are needed: for the (mac-
roscopie) plastic strain and for the hardening parameters. In defining the latter, atten-
tion must be paid to the possible "recovery" of strain hardening during transformations
(i.e., the fact that the newly formed phase can have only a partial memory, or even no
memory at all, of the previous hardening). This effect results from the fact that strain
hardening is due to the arrangement of dislocations in the material, which can be greatly
affected by the structural change (FRmDEL, [private communication]).

573
574 J . B . LEBLOND

To our best knowledge, SJt~TROM [1984] was the only author who addressed this prob-
lem. He supposed that strain hardening phenomena took place once a "global yield
stress" S yt was reached. He introduced then separate hardening parameters for each
phase, scalar for isotropic hardening (E~ ff and E~fe), tensorial for kinematic hardening
(Btij and Bz~j), with evolution equations given by (with our notations):

E ~ f f = ]F_eq., E e f f .~ E e q - - (1 -- O ) z f ~ ff (1)

Blij = Eii,
" p" B2ij = Ei~ - (1 - 0)zB2~j. (2)

In these equations z denotes the proportion of newly formed phase (phase 2), E p the
(macroscopic) plastic strain tensor, ~7eq the (macroscopic) equivalent uniaxial plastic
strain rate defined by:

• . P 1/2
(3)

and 0 a "memory coefficient" comprised between 0 and 1. 0 = 0 corresponds to no

memory at all (full recovery); in that case the evolution equations for the hardening
parameters of the new phase (phase 2) involve a "correction term" [-(1 - 0)zE~ ff or
-(1 - O)isB2iy] which tends to reduce these parameters down to zero, thus describing
recovery. 0 = 1 corresponds to a complete memory (no recovery at all); the equations
are then the same for both phases.
For stresses less than 27y, Sj6str6m assumed that no classical plasticity took place
(disregarding thus the fact that the stress-strain curve is affected, even for vanishing ex-
ternal stresses, by the microplasticity arising from volume differences between the
phases: see LEBLOtCDet al. [1986b]). Furthermore he assumed that transformation
plasticity could be described by eqn (6) of Part I. This means neglecting implicitly strain-
hardening effects. Indeed the quantity K is considered as a constant in this model; since
it is inversely proportional to o( in a first approximation (eqns (3), (4), or (t9) of Part
I), this means that ÜYis constant and therefore that no isotropic hardening takes place.
Furthermore eqn (6) of Part I involves only the stress deviator and not this deviator mi-
nus some tensorial hardening parameter, which implies that no kinematic hardening is
present either.
Section II of the present paper is devoted to the problem of the evolution equation
of the (macroscopic) plastic strain. Considering first the case of stresses less than S y,
we extend the treatment of Part I to incorporate strain hardening effects in the expres-
sion of the transformation plastic strain rate. For this, some of the physical hypothe-
ses made in Part I have to be modified or completed in a specific way according to the
type (isotropic or kinematic) of hardening considered. The response to variations of the
stress applied and the temperature (classical plasticity at constant temperature or con-
stant stress) is investigated in a similar way. We examine next the problem of the expres-
sions of the "global yield stress B y" and of the plastic strain rate in the case where the
stress applied is equal to 27y.

tThis "globalyieldstress"was supposedto be givenin terms of the yieldstressesof, o~ of individualphases

simplelinear mixturelaw:
a n d their respective p r o p o r t i o n s 1 - z, z b y a

~Y = (1 -- Z ) e ( + ZoO.
 Transformation plasticity in steels-I[ 575

In Section III, evolution equations are finally derived for the hardening parameters
introduced. They are obtained by homogenization, starting from the evolution equations
of the corresponding microscopic parameters (including the effect of recovery). The re-
sults are different from those of Sj6str6m (eqns (1) and (2)), even in the case considered
by this author where the stress applied is equal to Z7y. The differences are analyzed in
detail.

II. EVOLUTION EQUATIONS FOR THE TRANSFORMATION

AND CLASSICAL PLASTIC STRAINS

The same conventions as in Part I are adopted: we restrict to transformations occur-

ring on cooling; phase 1 denotes the old ~ phase and phase 2 the new a (or ~') phase.

II.l. lsotropic hardening

The usual hardening parameter for isotropic hardening is the equivalent uniaxial plas-
tic strain e eq defined by

.p.p
~eq = eijeij • (4)

However this parameter cannot be used to describe isotropic hardening in the presence
of transformations. Indeed its very definition implies that it can only increase; therefore
it cannot account for the recovery effect. We assume therefore that the amount of (iso-
tropic) hardening can be measured on a microscopic scale by an "effective plastic strain"
egfr, A = 1 or 2 (the index depending on the phase the point considered belongs to); egfr
is equal to e~q only in the absence of transformations. The evolution equation for this
parameter will be discussed later on. The yield stress o~ of phase A depends on egff ex-
actly as it depends on e,(q in the absence of transformations.

II. 1.a. Case o f stresses less than 27Y. To derive an equation for the transformation
plastic strain rate/~tp we start from eqn (8) of Part I. Restricting again to the case
where the Magee mechanism is negligible and assuming that phases 1 and 2 are essen-
tially, respectively plastic and elastic (hypothesis H.2), we discard the two last terms of
the right-hand side of this expression and obtain again eqn (9) of Part I. Assuming as
in hypothesis H.3 that the phases obey the Von Mises criterion and the associated flow
rule, we write 8ef/6z under the form:

6Z = 2at q 8-~sl = 2o.~V(efff---------

~ 8Z s l ' 2

The expression of Etp becomes then:

¢7. I'-~ff,~ $I Z" (5)

IVt

"el depends of course not only on e~ ¢ but also on the temperature 7". This dependence is omitted for sim-
plicity in the notation.
576 J.B. LEBLOND

Rigorously speaking, a ( (el ff) cannot be extracted from the average value symbol as
in the case of ideal-plastic phases, since it is no longer constant. However no tractable
model can be obtained if £-tp is kept under that form; indeed it depends then on an in-
finity of internal parameters (the values of e~ff at all the points of the austenitic phase).
We make therefore an approximation which consists in replacing a( (El ff) in eqn (5) by
some kind of average value, namely the value of aiv corresponding to the average hard-
ening parameter in phase 1. or( can then be extracted from the average value symbol
and eqn (5) becomes

F_tp =
3 ( 1 - z ) [tSe~q
2cr--~Y(-E~f--~) \ tSZ SI
I (6)
/VI

where

E,~ff --- (E;~ff)vA (A = 1,2). (7)

Etp depends then on a single internal parameter E~ ft.

Following the same procedure as in Part I, eqn (6) is further transformed by neglecting
correlations between 5efq/~z and s, (hypothesis H.4) and approximating (s,)v~ by the
overall deviator S (hypothesis H.5). The result is

3(1- Z)[~efq I
1~tp = 2o.iV(E~ff) \--~zlvSZ (8)

which is identical to eqn (15) of Part I, except that a y is replaced by ai~ (E~ff). Finally,
using the estimate of (&fU/~z)v, provided by eqn (17) of part I (which is readily verified
to be independent of whether phase 1 exhibits strain hardening or not), eqn (8) becomes

£,p = 3A~f~2 S (In z)L (9)

triV(E~ff)

This expression is again identical to eqn (18) of Part I except for the replacement of a y
by aY(E[ff).
In fact the previous treatment is limited to small applied stresses, like for ideal-plastic
phases (in particular because the estimate of (t~efq/~z)r, given by eqn (17) of Part I is
valid only in this case: see Part I). For high stresses, and for ideal-plastic phases, the nu-
merical simulation of transformation plasticity of Part I led us to introduce a correct-
ing factor h(F-,ea/,~u) where 2~eq is the macroscopic Von Mises stress:

seq = (3 SijSij) t/2 (10)

and 2; u the ultimate stress. We adopt here the same correcting factor, replacing the ul-
timate stress 2?u (which does not exist if strain hardening takes place, since the stresses
can increase indefinitely) by the "global yield stress" E y, the expression of which will be
discussed below. Furthermore eqn (9) overestimates F tp for very low values of z, like
eqn (18) of Part I, because phase 1 is not immediately entirely plastic as assumed in the
treatment. Like in Part I, we will take this fact into account by "cutting off" the effect
 Transformation plasticity in steels- II 577

of transformation plasticity below z = 3%. The expression of/~t, in the most general
case is then:

0 if z - 0.03
EtP
=

f th
3Ael_2 h/Eeq\
oY(E~ ff)
S [~-7)(lnz)z ifz>0.03.
(11)

This equation describes the effect o f strain hardening on transformation plasticity,

through the term o'lY(E[ff). Anticipating that austenite will harden during the transfor-
mation due to microplasticity, one sees that for a specimen undergoing a transforma-
tion under a constant uniaxial positive stress in the x direction, the slope of the curve
E ~ = f ( z ) will decrease more quickly than in the case of ideal-plastic phases (Fig. 2 of
Part I).
We now study the classical plastic strain rate at constant temperature, proportional
to the stress rate (E/f'). The general expression of E[t, is (LEBLOr¢Oet al. [1986a]):

(12)

This equation means that the macroscopic plastic strain due to variations of the stress
applied is simply obtained by taking the average value of the corresponding microscopic
plastic strain. This result is analogous to that concerning the (macroscopic) plastic strain
due to temperature variations and relies on the same hypothesis H. 1 (see Part I).
Following the same procedure as for F tp, 5/rz and z being replaced by 6/rF,ij and
~,j, we transform (12) into the following equation, analogous to (8):

3(l-z) (/tefq / •

We introduce then the following hypothesis, identical to that made in LEBLONDet al.
[1986b] in the case of ideal-plastic phases:
Hypothesis 1-1.6."[~u is nonzero only if 2~~q varies, i.e.

_,.. I .
"- b- - Lf
This hypothesis is reasonable, though certainly not rigorously satisfied. Note for instance
that a variation of the mean stress ]Ei~, which does not imply any variation of ,~'~,
does not induce any variation of plastic strain (the response of the material being then
purely elastic because of hypothesis H.1).
The expression of/~I:p becomes then

3(1 - Z) [ 6e~ \ S2~eq. (13)

It was shown in LEm.or~ et al. [1986b] that for z ~ 0.5, a reasonable estimate of
(&~/6~"e)vl is given by I/zE where E is Young's modulus. Numerical simulations of
578 J.B. LEatO~D

the overall stress-strain curve for fixed values of T and z led us to generalize this for-
mula into g ( z ) / E for all values of z; the function g is given in L~BLoND et al. [1986b].
We obtain therefore the final expression of ~7~t' under the form:

/~p = 3(1 - z) g(z) s£eq" (14)

2aY(E~ rr) E

We consider finally the classical plastic strain rate at constant applied stress, propor-
tional to the temperature rate (~7~-P). Proceeding again as for Etp, ~/~z and ~ being
replaced by 6 / 6 T a n d T, we transform eqn (21) of Part I into:

3(1 - Z ) /tSefa~
E~'P = 2"~a~(f~) \ 5T /z~ S~"

which is analogous to eqns (8) and (13). Using the estimate of (sefq/ST)r~ given by
eqn (23) of Part I, the preceding equation becomes:

E~.p = 3(oq - t~2) z (In z)S7r. (15)

oY(E~ ff)

II.1 .b. Case o f stresses equal to ,S y. The first problem which arises is that of the def-
inition and expression of the "global yield stress" B y. This stress is not related to the
first appearance of plasticity, which is present even for vanishing external stresses be-
cause of internal stresses due to the volume difference between the phases; it is connected
with general yield. Indeed two regimes of plastic flow can be qualitatively distinguished:
in the first one (studied above), the stress applied is low enough for plasticity to remain
confined in the austenitic phase, whereas in the second one, it is so large that plastic
deformations are present everywhere and limited only by strain hardening (if the behav-
iour was ideal-plastic, the ultimate stress S u would be reached and unlimited flow
would occur). By definition, we will say that the "global yield stress" S y is reached in
that second case.
In the case of ideal-plastic phases, several authors (ToMorA et ai. [1976]; GnssrI [1981];
SJ0STR01d [1984]; Dtmohs et al. [1984]) assumed that/?u was given in terms of the yield
stresses of the individual phases by a linear mixture rule:

S u = (1 - z ) a ( + za~

However numerical studies of the overall stress-strain curve (LEBLONDet al. [1986b]) led
us to propose a more accurate nonlinear mixture rule:

Z7~ = [1 - f ( z ) l a ~ +f(z)a~v;

the f u n c t i o n f i s given in L~aLOrCDet al. [1986b]. We will assume that in the case of iso-
tropic hardening, _ry is given by the same expression except that the yield stresses of the
phases are no longer constant:

~Y = [ 1 - f ( z ) ] o'((E~ ff) + f ( z ) f f Y ( E [ f f ) • (16)

 Transformation plasticity in s t e e l s - I I 579

In this formulation the influence of strain hardening is introduced only through the aver-
age hardening parameters of the phases, as in the case of stresses less than ,l~y (see
§II. l.a above).
If the "global yield stress" is reached, it is no longer necessary to distinguish between
classical and transformation plasticity to obtain the flow rule of the material, since plas-
tic deformation is then no longer determined by the detailed influences of ,~, 7", 2 on the
microscopic plastic strain rate in austenite but only by strain hardening, as mentioned
above. We will only make the (reasonable) supposition that the material obeys a mac-
roscopic Mises-type criterion (2~eq _< 2~y) and the associated flow rule:

S if 2~eq = E y (17)
2 2~eq

where/~'~ is defined by (3). The expression of ~,eq in this equation is to be determined

from the condition ~_eq= 2~y (T,z, EIeff ,E2eff) plus the evolution equations for the hard-
ening parameters E~ ff,/~ = 1 or 2 (which are discussed below). (This is the mathemati-
cal translation of the fact that the amount of plastic deformation is determined by strain
hardening.)

II.2. Kinematic hardening

The centre of the microscopic yield surface in phase A will be denoted a~. The evo-
lution law for this parameter, taking into account the recovery phenomenon, will be dis-
cussed later on.

II.2.a. Case of stresses less than F,~. We consider first the transformation plastic
strain rate. Equations (8) and (9) of Part I remain unchanged, as in the case of isotro-
pic hardening (§II. 1.a). We write then (assuming again Von Mises' criterion and the as-
sociated flow rule to be obeyed on a microscopic scale):

6el 3 6elq 3 ~efq

6Z - 2~f q 6Z (sl--al)=2~---~ 6--Z--( s l - - a l )

where

~ff = (sl~j - a~j)(s~ U - aw) j .

We obtain thus:

/~tp= 3 ( 1 - z ) / & f q )
(18)

Note that here the extraction of ai~ from the average value symbol is a rigorous opera-
tion (unlike in the case of isotropic hardening), since it is constant within the represen-
tative volume ~ as in the ideal-plastic case.
Hypotheses H.4 and H.5 of Part I are now introduced with some appropriate modi-
580 J . B . LEBLOND

fication: for H.4 the quantities supposed to be uncorrelated are 5efq/Sz and s~ - a~ (in-
stead of Sl). Furthermore we define the following quantities:

A;~=(ax)va(A=I,2);A=(a)v=(I-z)AI+zAz(V=VILJV2). (19)

Equation (18) becomes then:

3(1- z)[6efq I
E"= -2o (20)

which is analogous to eqn (15) of Part I, S being replaced by S - At.

Finally, using eqn (17) of Part I, and introducing the correcting factor h and the "cut-
off" below z = 3°70 to account for high applied stresses and for the very beginning of the
transformation, we transform eqn (20) into:

0 if z -< 0.03

f ~ry (S - AI)h \zy! (ln z):~ if z > 0.03.

In this expression, S ~q is the macroscopic Von Mises stress given here by:
(21)

E eq = (S~j - A~j)(S~j - A , j ) (22)

and S y the "global yield stress" (to be defined precisely below).

For a specimen transformed under a constant uniaxial positive load in the x direction,
eqn (21) predicts that the slope d E ~ / d z will decrease more quickly than in the ideal-
plastic case, because A ~= will increase during the transformation due to the hardening
of austenite (see below). This prediction is qualitatively the same as in the case of iso-
tropic hardening (see §II.l.a).
We now study ~?~P. Following the same procedure as for Etp ( ~ / 6 z and z being
replaced by ~/6S, ij and ~0), eqn (12) is transformed into:

20( ~ ~Siy Iv~

which is analogous to (20). Hypothesis H.6 of §II. l.a is then introduced in a modified
form, because the analog, for kinematic hardening, of the quantity ,¢.eq intervening in
isotropic hardening is not ~eq itself, but rather the time-derivative of Beq obtained by
considering A ~ as constant, i.e. 3

~q = 3_~ (Sij _ A l i j ) a i j ' z~,'eq = ( S O - Alij)(Siy - AIO) (23)

2E'eq

3 0 eq
In classical plasticity models (for a single phase), the flow rule is ~i~ = 2--M -o~ sjj for isotropic hardening but
3 o~" 3
e'JP = --2M on - ( so - aij) for kinematic hardening (where M is the hardening slope and O~ = 2 - ~ ( so -
a°)s°)"
Also, in a uniaxial tension test in the x direction, o~, = o "° for isotropi¢ hardening but = off for kinematic
hardening. Thus 0~ in isotropic hardeningcorresponds to off in kinematichardening.
 Transformation plasticityin steels- II 581

In its modified form, hypothesis H.6 reads:

Hypothesis H.6:

/ /l'S •

The expression of E~' becomes then:

£.~p = 3(1 - Z) [ 6er q ~ ( S - A , ) , ~ q. (24)

2a y \ ~Seq /v,

We use then the estimate of (~efq/~,eq)vl given in LEBLONDet al. [1986b] to obtain E~P
under its final form:

/~p = 3(1 - z) g ( z ) (S - A t ) £ ~ q. (25)

2a~' E

We consider finally E~-P. Proceeding again as for Etp (a/az and ~ being replaced by
/f/~Tand ~h), we transform eqn (21) of Part I into:

3(1- /
~o-? \ ~7"/r, ( s - A,)7"

(which is analogous to (20) and (24)), then into:

E~p = 3(cq - ~2) z In z ( S - A i ) 7r (26)

using eqn (23) of Part I.

II.2.b. Case o f stresses equal to 2: ~. For kinematic hardening (unlike for isotropic
hardening: see II. 1.b), the "global yield stress" can be given a rigorous definition, as the
maximum possible value o f the Von Mises stress F,eq defined by eqn (22). Indeed the
Von Mises stress a ¢q being a norm (in the usual mathematical sense) of the stress devi-
a t o r s , one has (denoting [[[ this norm and Vthe total volume I/! U I/2):

27eq = ~ S - A ~ = I ( s - a)v[ - ( [ ] s - aU)z = (1 - z)(~s, - al~)vi + Z(]s2-- a2~)v2

_< (1 - z)<a~>v, + zOI>v2 = (I - z ) ~ + z ~ .

This inequality, which is exactly the same as in the ideal-plastic case (LEBLOND el
al. [1986b]), implies that the upper bound 27y of 2: '~ is finite and not greater than
(1 - z) a[ + za~. (Of course, this does not mean that S is bounded; only S - A is.)
We will adopt the same expression for 27y as that for 2:" in the ideal-plastic case (Le-
aLor~ et al. [1986b]):

2: y = [1 - f t z ) ] a~ 4- f ( z ) a ~ (27)
582 J.B. LEBLOND

which is consistent with the above inequality since a( < al' (phase 1 is the weaker one)
and f ( z ) _< z (see Table 2 of LEBr.OSI) et al. [1986b]).
If the "global yield stress" is reached, we make no more distinction between classical
and transformation plasticity (plastic flow being then determined only by strain hard-
ening) and write simply the flow rule associated with the Mises-type criterion S eq < EY:

3 /~q
E~( = E,,, + E~; + E ~ - p ) = _ (S - A) (28)
2 _S_ eq
Eeq being defined by (3). The expression of ~7eq in this equation is to be deduced from
the condition 27eq = 27Y(T,z) and the evolution laws for the hardening parameters A1
and A2, as in the case of isotropic hardening (see §II. l.b).

Ill. EVOLUTION EQUATIONS FOR THE HARDENING PARAMETERS

The classical and transformation plastic strain rates, such as expressed in the preceding
section, depend only on the average hardening parameters of phases 1 and 2 (EF" and
E~ ff for isotropic hardening, At and A2 for kinematic hardening). What we are inter-
ested in is therefore only the evolution laws for these average parameters. It will nev-
ertheless be necessary to define the evolution equations for the microscopic hardening
parameters (e;~ff or a~, h = 1 and 2) in order to derive those for their average values by
homogenization. The possible "recovery" phenomenon (see the Introduction) will be
taken into account by introducing "natural" assumptions concerning hardening on the
microscopic scale.

III. 1. Isotropic hardening

III.l.a. General formulae. We want first to express ~?;~frin terms of ~fr (A = 1 and
2). This will be done by differentiating eqns (7) with respect to time, paying attention
to the fact that volumes V~ and I/2 are moving because of the progress of the transfor-
mation. Using the classical formula for time-differentiation of integrals over a moving
domain (GERMAL~[1986]), we get:

£dt f n ~ ( f dv= fn~fff d v - fF~fffUndS

where F is the transformation front (surface separating the volumes Vt and V2 occupied
by the phases) and U. its normal velocity, considered as positive if directed towards V~
(this convention is logical since the phase whose proportion is increasing is phase 2).
Therefore:

/~ff ~ "~[(1-Z)VoI(V)Le~ffdv]
= (1 - Z)TvoI(V) ~.~ffdo + (1 - Z)voI(V)

-- 1 fF~ff UndS
(1 - Z)voI(V)
 Transformation plasticity in steels-11 583

where vo1(12) denotes the volume of a region 12. Taking into account the obvious rela-
tion ~ = [ U~dS/vol(V), we transform the preceding equation into:
Jr

Ef ff = <~rf)v, "4" 1 - Z E~ff - l-"-~Z

z .~'eff~.
tF(U.) (29)

where (e~ff)F{Un) denotes the average value of e~ff over the transformation front F,
weighted by the normal velocity U.:

(~ eff\,'FtU.) = fj ~" U~dS

(A = 1,2). (30)

Replacing the index 1 by 2 and I - z by z in eqn (29), we get similarly:

~'~ff =
(~ff)v2 - Z E~ff + -Z/~eff,
- x 2 ,'Fry,,). (31)
Z Z

The microscopic hardening laws must now be stipulated. This means modelling the
interactions between phase transformations and strain hardening phenomena. Our
description will be based on the following elements:
1. At a point which does not undergo a transformation (i.e., which does not belong
to F) at the instant considered, the evolution law for the hardening parameter will be
taken quite naturally as the usual one: ~fr = ~q, where ~eq denotes the equivalent
uniaxial plastic strain defined by eqn (4) and the index A = 1 or 2 depends on the phase
the point considered belongs to. This implies that:

(~.~ff)v,~ = (~:~q)vA (A = 1,2). (32)

2. Any influence of the hardening parameter e~ff of the disappearing austenite on the
position and velocity of the transformation front will be neglected; in other words, the
transformation will be supposed to be unaffected by previous strain hardening. This ap-
proximation might be criticized but is unavoidable in the absence of any quantitative in-
formation concerning such an influence. One has then:

.effxI F ( U n ) =
¢1 (e~ff)v, = E~ ff (33)

Indeed, if the partial average value (~ff)Fcu,) was not equal to the global average value
(~ffr)z,, and was for instance greater, this would imply either that F is located in a re-
gion where efn is statistically greater than its (global) average value, or that U, takes
statistically its greatest values in the portions of F where efff is also greatest; this would
mean existence of correlations between ~ff and the position or velocity of F.
3. The recovery effect will be described by means of a "memory coefficient" 0 (0 _
0 < 1) such that the new hardening parameter ~ff at a point of F will be equal to the
old one ~fff multiplied by 0. The value of 0 depends on the transformation: according
to FInDEr [private communication], 0 = 0 (no memory) is a good value for ferritic and
584 J.B. LEBLOND

bainitic transformations, which imply large displacements of the atoms by diffusion and
therefore important rearrangements of the lattice leading to annihilation of the dislo-
cations; on the other hand martensitic transformations, which are diffusionless and in-
volve only very small displacements of the atoms, are better described by 0 = 1 (full
memory). The average value of e~fr on the transformation front is then

(~eff)
. F~U,,~ = O ( ~ f f ) F ( U n ) : OE~ff (by eqn (33)). (34)

Inserting eqns (32-34) into (29) and (31), we get:

Efff = <~fq>/,"l }
(35)
J ~ f f : ( ~ q > v , -- -~ E~ ff q- 0 ~ E~ ff.
- Z Z

These equations are to be compared with Sj6str6m's eqns (l). Anticipating (see
§III.l.c) that Sj6strOm's results correspond to the case where ~fq and ~q are uniform
within V~ and I:2 and equal to Eeu (defined by eqn (3)), it appears that the b t e r m of
J~ff is not the same in eqns (l) and (35). To analyze the difference, we will consider the
simple case where Eeq = (~fq)v~ = (~q)v2 = 0; this means that there is no strain hard-
ening and that only recovery a n d / o r memory phenomena are present. 4
Consider first the case where 0 = 0 (full recovery). Equation (1)2 reads then ~'~rf =
~,E elf which is readily integrated into E~ rr - Cst x e-Z. Thus, if the initial proportion
zi of phase 2 is 0 and the final proportion z: is 1, the initial [(E~ff)i] and final [(E~ff)f]
values of E~ ff a r e related by the equation (E~ff)f = ( E ~ f f ) i / e . This is clearly incorrect,
because the final value of E~If should be independent of its initial value (which has no
real physical meaning since there is no phase 2 initially) and equal to 0, the recovery be-
ing complete. On the other hand, eqn (35)2 reads/~fr = __z E ~ f f which is equivalent
Z
tO zE~ ff = Cst. Thus, with the same notations as above:

LTeffx
,-.2 Zi
J: = -- (E~ff)i-
z:

If zi = 0, (E~ff)f = 0 independently of the value of (E~ff)i, as desired. If zi :~ 0, this for-

mula is easily verified to be still satisfactory. Indeed (E~fr)f can be evaluated directly,
without appealing to eqns (1)2 or (35)2, as follows. At the final instant, two kinds of
phase 2 can be distinguished: phase 2 which existed prior to the transformation (propor-
tion zi), with average hardening parameter equal to (E~ff)i; and new phase 2 (propor-
tion zI - zi), formed with a zero hardening parameter (full recovery). Thus the average
final hardening parameter of phase 2 (total proportion z f ) is

(~'2
lg'eff'IfL = -Zi- (E~ff)i .at- Z f -- Zi X 0 = -Zi- (E~ff)i.
z: z: z:

The direct reasoning leads therefore to the same result as eqn (35)2.

~This hypothesis is made only momentarily, to simplify the comparison between eqns (1) and (35), and is
in fact very unrealistic. Indeed, as mentioned below (§lII. 1.b), the austenific phase hardens always during the
transformation (even in the absence of any external load), because of the microscopic plasticity arising from
volume incompatibilities between the phases.
 Transformation plasticity in steels-ll 585

Consider now the case where 0 , 0 , 0 = 1 for instance (full memory). Even if the
missing l/z factor is introduced into eqn (1)2, it is still different from eqn (35)2. Equa-
tion (1)2 reads simply E ~ f f = 0 , that is ~/ 2l~eff~ - (Efff)i • On the other hand, eqn (35)2
i f __
reads ~[ff -- - - Zz E~ ff + ~ reff ' that is ~d ( z E [ ff) = ZE~ if, which yields u p o n integra-
Z Jt'~l
tion (using the fact that E f n = Cst by eqn (35)3: z E ~ ff - z E f ff -- - tEeff~
Cst. Thus ,~/~ 2 Jf -
Z f E ~ ff = z i ( E ~ f f ) i - z i E e (f , which implies that:

c'effxIf = -Zi- (E:~ff)i + Zf -- Zi E~ff

z:'2
zf zf

Now (E~rf)f can again be calculated directly. Indeed the final phase 2 (proportion zf)
is composed of old phase 2 (proportion z3 with average hardening parameter (E[ff)t,
and new phase 2 (proportion zf - zi) with average hardening parameter E~ ff (since the
hardening of phase 1 is transferred to phase 2 without any loss of memory). Hence its
average hardening parameter is:

(E[ff)f = z i ( E ~ f f ) i + z f - - Zi E~ff"
Zf Zf

Therefore the direct reasoning shows that the correct equation is again (35)2 .5
It is somewhat surprising that Sj6str6m's equations are different from ours, because
the reasoning leading to them, though presented in a much less formal and explicit way
than ours, relies on the same basic ideas. The differences may well be simply due to
inadvertance.
We must now distinguish between the cases 27eq < 27y and ,~eq = S~.

III.l.b. Case where ~_,eq< 27y. (~q)v I can then be calculated using the equation:

(~fq)v, = \ 8z Ivt~ + \ ~ ' ~ / v , \ 5T /vl ~

plus the evaluations of (~efq/iiz)vl, (~e~/6F, eq)v~, (6efq/6T)v~ provided in LEntor,,v

et al. [1986b] and in Part I. Equation (35)1 becomes then:

E~f = - 2 Ae[h-'2 h 27"q) (In z)z + g(z) ,~ev + 2(ai - a2)z In z ~. (36)
1-z T; -T l-z

where the first term of the right-hand side is to be replaced by 0 if z -< 0.03.
Equation (36) describes the influence of transformation plasticity on the hardening of
anstenite (through the first term of the right-hand side), and more generally the hard-
ening due to micro-plastic strains arising from variations of z, 27, or 7". Note that even

SNote that even in the absence of recovery (0 -- I), the expressions of ~ and ~ n given by eqns (35) are
not symmetrical (unlike those given by eqns (I)).Though perhaps surprisingat firstsibht,thisdissymmetry
can eusflybe shown to be reasonable: indeed existenceof memory implies that the hardening rate~tr of the
new phase must be influenced by the hardening parameter E~af of the old phase, whereas there is no reason
why ~ffr should be influenced by E~ fr,
586 J . B . LEBLOND

if the material is transformed without any external load (27 = 0) and does not therefore
undergo any macroscopic plastic strain, austenite is hardened during the transformation.
This is because even if the average value of the microscopic plastic strain rate in austenite
is zero, it is not zero locally, due to the microplasticity arising from the volume differ-
ence between the phases. Thus there is no connection between the macroscopic equiv-
alent uniaxial plastic strain r a t e / ~ q (defined by (3)) and the hardening rate of austenite
J~ff: if S = 0 , / ~ q = 0 but ~"~ff ::;/:0.
Furthermore, phase 2 being supposed to remain elastic (hypothesis H.2: see Part l),
~]q = 0 so that eqn. (35)2 reads:

~?{ef = _ _2 E~ff + 0 _2 E [ ff. (37)

Z Z

This equation expresses the fact that no strain hardening takes place in the harder phase
2; the evolution of E~ rr arises only from recovery a n d / o r memory phenomena.

llI.l.c. Case where ~eq = Sy. The intuitive idea that plasticity is then no longer con-
fined in the austenitic phase but spread everywhere (see §II. l.b) will be formalized and
completed by the following hypothesis:
Hypothesis H. 7: When the "global yield stress" is reached, the microscopic plastic
strain rate is nearly uniform within the whole representative volume V.
This is the assumption made usually (and notably by SJOSTROM [1984]) for large
stresses; it means that in these conditions the material behaves rather like bars in par-
allel than like bars in series. This is confirmed by Fig. 9 of LEntoNt) et al. [1986b] which
shows that except for very low proportions of the harder phase, the ultimate stress of
a mixture of two ideal-plastic phases is closer to that of bars in parallel [(1 - z)a~ +
zo~] than to that of bars in series (ai~). (In other words, the f u n c t i o n f ( z ) intervening
in the expression (16) of S e is closer to z than to 0.)
The general expression of ~'~' is (LEBtOt~D et al. [1986a]):

(38)

disregarding the Magee mechanism as in Part I, we drop the last term of the right-hand
side. Hypothesis H.7 implies then that:

~'P = ~P = i f . (39)

By eqns (3) and (4), we get then/~eq = ~[q = ~q so that eqns (35) read

(40)
/~ff = / ~ q - -z e ~ " + 0 -2 e f fr "
Z Z

Thus the uniformity of strain rates implies that there is a direct connection between
the/~,rf's (A = 1,2) and Eeq unlike in the ease where 2?eq < ,~Y. This makes it possible
to calculate/~eq in the flow rule (17), by replacing ~Tfrf and ~7~¢fin terms of ~7eq in the
equation obtained by differentiating the equality _req = ~ Y (, T ,, -~.,E Ielf , E 2eff~J with respect
to time.
 Transformation plasticityin steels-I1 587

III.2. Kinematic hardening

We will restrict for simplicity to linear kinematic hardening (i.e., constant hardening
slope for a given temperature). The extension to nonlinear kinematic hardening does not
raise any particular problem, once the hardening law has been stipulated. It will there-
fore be supposed that in the absence of transformations, the centre a of the microscopic
yield surface is given by the classical expression for linear kinematic hardening in the
presence of temperature changes (see e.g., GIOSTI[1981]; SJ{~STROM[1984]): a = H(T)e p,
where H ( T ) is a material parameter. In the presence of transformations, we will have
similarly ax = Hx(T)bx (A = 1 or 2 according to the phase the point considered belongs
to), where the (microscopic) parameter bx will be different in general from e~', since it
must for instance be zero just after the passage of the transformation front if complete
recovery occurs. Taking the average value of the preceding expression and using the def-
inition (19)1 of Ax, we get:

Ax = Hx(T)Bx where B~ = (bA)v~ (A = 1,2). (4l)

We look for evolution equations for the A~'s, or equivalently for the Bx's.

III.2.a. General formulae. Following a procedure analogous to that of §III.l.a (e:~fr,

E,{fr, and e,{q (A = 1,2) being replaced by b~, B~, and eft), one obtains:

B, = <~f>v, 1
B2 + 0 ~ BI I (42)

where 0 is the "memory coefficient." The expression of/~2 differs from that given by
Sj~Sstr~im (eqn (2)2), just as eqn (35)2 differed from eqn (1)2.

III.2.b. Case where 27eq < ,!7j'. Assuming phase 2 to remain elastic (hypothesis H.2),
we get (.i~')vz = 0. Furthermore, the Magee mechanism being neglected, eqn (38) be-
comes E a = (1 - z)(dP)v,. Thus eqns (42) read:

/~'= 1-Z
(43)

.z z
t
where EO = Etp + E~p + Efp is given by eqns (21), (25), (26).
Equation (43)1 describes the influence of transformation plasticity on the hardening
of austenite, like eqn (36) in the case of isotropic hardening. Note however that unlike
in the isotropic case, there is no evolution of the hardening parameter B, of austenite
if no external load is applied during the transformation (and if there is no previous hard-
ening) (indeed ~-t, =/~to + E~O + Efp is zero if S - A is zero: see eqns (21), (25), (26)).
This is due to physical differences in the formulations of kinematic and isotropic hard-
ening: in the kinematic model, hardening is sensitive to the plastic strain rate tensor ~P
(the average value of which is zero if S - A is zero) whereas in the isotropic model, it
is sensitive to the norm ~eq of this tensor (the average value of which is ~: 0 even if
S = 0).
588 J . B . LEBLOND

III.2.b. Case where seq = ,~y. Making the assumption of uniformity of strain rates
as in the case of isotropic hardening (hypothesis H.7), we insert eqns (39) into eqns (42)
and obtain:

" (44)
BI = £" Z Z
1
The quantity/~eq in the flow rule (28) can now be calculated by differentiating the con-
dition ,~eq = By with respect to time and expressing, in the equation o b t a i n e d , / i =
d
dt [(1 - z)Al + zA2] in terms of B~ and B2 using (41)t, then or £ P using (44), then
of [?eq using (28).

IV. SUMMARY AND CONCLUSIONS

For each of reference, we give here a summary of all the models which have been
proposed.
Ideal plasticity:

•eq = SijSij) 1/2; S u = [1 - f ( z ) ] a y + f(z)a~.

• If S eq < Su:

th
E tp 3A~1~2 h/ Eeq\
ay S [-~-7) (In z ) L or 0 if r < 0.03;

E~p = 3(1 - z ) g(z) s~eq;

2a y E

£~p = 3 ( o q - a2) Z (In z)S7 r.

• IfE~q=s":

~'P = l i S (where ,,i is indeterminate).

Isotropic hardening:

P p~l/2

1J)' = [ 1 - f ( z ) ] o~'(Ef ff) + f(z)o~(E~ff).

 Transformation plasticityin steels-If 589

• I f ~ e q < L'Y:

/~tp_._ 3A~fh--2 ( E'q )

o((E~ff) Sh ~ (In z)L or 0 if z-< 0.03;

3(1 - z) g(z) SE~q;

/~cp = 2o'iV(E~ff) E

E~p = 3(t~1 -t~2) z (In z)S7";

oiV(E[ff)

/~ff = 2A£[h2"i
=Z h(~'eq~,~,y
/ (In z)~. "1- Tg(z) £eq .~_ 2(°/1 -1 -Z0/2)Zlrl z

where the first term of the right-hand side is to be replaced by 0 if z -< 0.03;

~,[fr = _ _ZE~ff + O ~" E ~ ft.

z z

* If~r~q=.rY:

3 Eeq
EP = - ~ S;
2 ~q

E~ff = Eeq _ Z_ E~ff + O _ E~ ft.

z z

(Linear) kinematic hardening:

A~ =H~(T)B~ (A = 1,2);

A = (1 - z)A~ + zA2;

~'q = (s~/ - Ao)(S~ - A~j) ;

~q ----. 3 ( S i j -- A i i j ) S i y ' ~,eq . - (Sij - A l i j ) ( S i j - Alij)

2 F , eq

• P . p~l/2

, r , = [1 - f t z ) ] , r i ~ +ftz)aL
590 J.B. LEBLOND

• If S ~q < S~:

E'P-
, h
3Ael~2
7 (S-Al)h
(),.~eq
~ (lnz)z, or0ifz_<0.03;
oi

E}~_ 3(1 - z) g ( z )
(S -- A i ),~q;
2o( E

E~p _ 3 ( a l - 0~2)
z In z ( S -- Al)~";
oi

Bl : 1 ( E t p + E~p + E ~ p ) ;
1-z

B2=- -~Bz+o -~BI.

z z

" If S eq = S Y :

3 Eeq
L'P=- ~ (S-A);
2 S eq

BI = L'P;

B2 =EV _ ~
_ B2+ 0 ~ Bi.
z .7,

The functions f, g, h in these equations have been given in LEB~.ONDet al. [1986b] (Ta-
bles 2 and 3) and Part I (eqn (26)).
These models allow for a more realistic description of the plastic behaviour of phase-
transforming steels than previous ones (see e.g., GrtrsxI [ 1981]; SJ0STROM [ 1984]; Dt'aOlS
et al. [1984]; MIxxER [1987]). The improvements brought include notably:
1. The description of the influence of microplasticity due to volume differences be-
tween the phases on the response of the material to variations of the stress applied (and,
less importantly, of the temperature) (the classical plastic strain rate/~cp = ~.~p +/~fv
was always supposed up to now to be zero for stresses less than the "global yield stress").
2. The fact that the parameter K characterizing transformation plasticity (see Part I)
is given by an explicit formula in terms of the thermomechanical characteristics of the
material, instead of being considered as an independent material constant. This allows
for a more precise evaluation of this parameter; for instance the (notable) dependence
of K upon the austenitic grain size (due to the influence of this grain size on the yield
stress of austenite: see e.g., DEs~os [1981]) can be taken into account.
3. The description of transformation plasticity (and also of classical plasticity) in the
presence of strain hardening. The models developed include not only the couplings be-
tween transformation plasticity and strain hardening phenomena but also the influence
of recovery/memory effects during transformations on the evolution of the hardening
parameters.
 Transformation plasticity in steels-11 591

These models have been incorporated in the SYSTUS finite element code. Applica-
tions to numerical simulations of welding operations are now being performed (see e.g.,
DEv^ux [1986]).

REFERENCES

1965 GREENWOOD,G.W. and JoHssox. R.H., "The Deformation of Metals Under Small Stresses during
Phase Transformation," Proc. Roy. Sot., A 283, 403.
1976 TOYOTA,Y., KUROKi,K., MORI, T., and TAMURA,I., "Tensile Deformation of Two-Ductile-Phase Al-
loys: Flow Curves of c~-7 Fe-Cr-Ni Alloys," Mat. Sci. Eng., 24, 83.
1981 DESALOS,Y., "Comportement Dilatom6trique et M6canique de l'Aust6nite M6tastable d'un Acier
A.533," IRSID Report #95349401 MET 44.
1981 GrOSTI,J., Contraintes et Deformations R~siduelles d'Origine Thermique; Application au Soudage
et/t la Trempe des Aciers," Th~se d'Etat, Universit~ Paris VI (France).
1984 SJOSTR6M,S., "Interactions and Constitutive Models for the Calculation of Quench Stresses in Steel,"
in Proc. Int. Syrup. on Calc. Int. Str. in Heat Treat. Met. Mat., Link6jong (Sweden), pp. 221-246..
1984 Duaois, D., DEVAUX,J., and L~:BtO.,,'D,J. B., "Numerical Simulation of a Welding Operation: Cal-
culation of Residual Stresses and Hydrogen Diffusion," in Proc. 5th Int. Conf. on Press. Vess. Tech.,
San Francisco (U.S.A.), pp. 1210-1239.
1986 DEv^ux, J., "Proc(~d~ de R(~paration par la M(~thode/l Duret~ Limit~ sous Cordon-D~termination
Num~rique des Contraintes R~siduelles pour un D~p6t Monocouche de Quatre Cordons," Framatome
Internal Report #TM/CD DC/86.21.
1986 GERMAIS,P., M~canique, Ellipses, Paris (France).
1986a LEBI.OSD,J.B., MOTTET,G., and DEVAUX,J.C., "A Theoretical and Numerical Approach to the Plas-
tic Ikhaviour of Steels During Phase Transformations-I: Derivation of General Relations," J. Mech.
Phys. Solids, 34, 395.
1986b LEBLO.~D,J.B., MOTTET,G., and DEVAVX,J.C., "A Theoretical and Numerical Approach to the Plas-
tic Behaviour of Steels During Phase Transformations-II: Study of Classical Plasticity for Ideal-
Plastic Phases," J. Mech. Phys. Solids, 34, 411.
1987 MrrTER, W., "Umwandlungsplastizit~it und ihre Berucksichtigung bei der Berechnung yon Eigen-
spannunger," Borntr/iger-Verlag, Stuttgart (FRG).

Laboratoire de Mecanique des Solides

Ecole Polytechnique
91128 Palaiseau Cedex- France

(Received 4 November 1987; in final revisedform I I May 1988)