You are on page 1of 10

Deformation-Induced Transformation of Retained Austenite

in Transformation Induced Plasticity-Aided Steels:


A Thermodynamic Model
MONIDEEPA MUKHERJEE, SHIV BRAT SINGH, and OMKAR NATH MOHANTY

A model based on thermodynamics of deformation-induced transformation of retained aus-


tenite to martensite has been developed, so that the amount of martensite or the amount of
austenite left untransformed after a certain amount of strain can be predicted. The model takes
into account the mechanical driving force or the energy contribution of the deformation strain,
which has been determined empirically from reported experimental data. The model was used to
predict the volume fraction of untransformed austenite as a function of strain for a wide range
of alloys. The predictions were reasonably accurate in most of the cases.

DOI: 10.1007/s11661-008-9599-x
 The Minerals, Metals & Materials Society and ASM International 2008

I. INTRODUCTION isothermal bainitic transformation. As a result, the


microstructure at the beginning of intercritical annealing
LOW-ALLOY transformation induced plasticity is martensite, which is different from that in the
(TRIP)[1]–aided steels offer an excellent combination of conventional treatment where the starting microstruc-
high strength and ductility at affordable cost.[2] These ture before intercritical annealing is ferrite + pearlite.
unique properties can be attributed to the complex The resultant lath type of matrix microstructure has
microstructure of these steels. Conventional polygonal been named annealed martensite by Sugimoto et al.[5]
ferrite (PF) matrix TRIP-aided steel (PF steel) consists They found that intercritical annealing for 1200 seconds
of polygonal ferrite, carbide free bainite, retained resulted only in recovery and decrease in the dislocation
austenite, and in some cases, a small fraction of density of the martensite formed during prior heat
martensite. These steels are produced by a two-step treatment but did not result in the nucleation of
heat-treatment process that involves intercritical anneal- polygonal ferrite.
ing followed by isothermal bainitic transformation The most important constituent of the complex
treatment.[2,3] microstructure of the TRIP-aided steels is 10 to
Two more varieties of TRIP-aided steels have been 20 vol pct of retained austenite. Deformation causes
developed in recent years in response to the complex progressive transformation of retained austenite to
formability requirements of the present day autobody martenite, which increases the strain hardening rate
manufacturers.[4,5] These steels are based on bainitic and delays the onset of necking.
ferrite (BF) and annealed martensite (AM) matrices and A large amount of work has been dedicated to study
are termed as BF steel and AM steel, respectively.[4–6] the deformation-induced transformation of austenite to
The TRIP-aided steels with bainitic ferrite as the matrix martensite. As a result, a number of emperical as well as
phase and retained austenite as the second phase (BF semiemperical models have been formulated to predict
steels) can be produced by a two-step heat-treatment the kinetics of this transformation.[3,7–14]
process, where complete austenitization is followed by The aim of the present work was to develop a
an isothermal bainitic transformation treatment.[4] The thermodynamic-based model for the deformation-
third variety of TRIP-aided steels called AM steels induced transformation of austenite to martensite so
consists of ‘‘annealed martensite’’ as the matrix phase that the amount of martensite or the amount of
and carbide free bainite and retained austenite as the austenite left untransformed after a certain amount of
other phases.[5] In this case, the steel is subjected to an strain can be predicted.
additional austenitizing and oil quenching step prior to
the conventional treatment of intercritical annealing and
II. THEORETICAL BACKGROUND
MONIDEEPA MUKHERJEE, Researcher, is with Research and
Development, Tata Steel, Jamshedpur-831001, India. Contact e-mail:
Martensitic transformation is a first order phase
monideepa13a@gmail.com SHIV BRAT SINGH, Associate Professor, transformation and occurs by nucleation and
is with Department of Metallurgical and Materials Engineering, Indian growth.[15,16] For the transformation to occur it is
Institute of Technology, Kharagpur-721302, India. OMKAR NATH necessary that (1) the free energy of the system decreases
MOHANTY, Vice Chancellor, is with Biju Patnaik University of and (2) heterogeneous nucleation sites are present.[9] The
Technology, Orissa, Ghatikia, Bhubaneswar -751 003, India.
Manuscript submitted September 19, 2007. energetics of transformation can be understood from
Article published online July 15, 2008 Figure 1.[8] At the T0 temperature, the chemical free

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 39A, OCTOBER 2008—2319


Fig. 1—Schematic representation of the energetics of martensite
transformation.[8]
Fig. 2—Schematic illustration of the influence of temperature and
applied stress on martensite transformation mechanisms.[20] Solid
line indicates the critical stress required to initiate martensite trans-
energy of austenite is equal to that of martensite and formation at temperatures above Ms.
they are in thermodynamic equilibrium. To bring about
the transformation of austenite to martensite, it is
essential that the free energy change of the system be The critical stress in this temperature region is lower
large enough to compensate for the surface energy and than the yield stress (for slip) of stable austenite and the
the elastic strain energy associated with martensite. transformation occurs by heterogeneous nucleation on
Thus, martensite cannot form spontaneously at T0. the same nucleation sites that are responsible for
Transformation is possible only after a certain amount transformation on cooling below the Ms temperature.
of undercooling to a temperature designated as Ms This mode of transformation is called stress-assisted
(<T0) at which a critical chemical driving force, DGcrit transformation.[21] The yield stress for slip in austenite
becomes available for initiating the transformation.[17] decreases with increasing temperature and above MrS,
In the absence of any other energy, the magnitude of the the stress needed to initiate martensite transformation is
chemical free energy difference must be greater than or large enough to cause plastic deformation (by slip) in
equal to |DGcrit| for the transformation to take place. austenite.[22] At temperatures higher than MrS, martens-
Martensite transformation involves an invariant- ite transformation takes places in a plastically deformed
plane strain shape deformation with a large shear austenite on new nucleation sites generated by plastic
component. The transformation can therefore be deformation and at stresses that are considerably lower
regarded as a mode of deformation with a change in than is expected from a simple extrapolation of the
crystal structure also. It follows that the transformation critical stress–temperature line between MS and MrS
can be influenced by external deformation.[16] The (Figure 2).[20] This mode of transformation is termed
deformation or strain energy can be regarded as an strain-induced transformation.[21] At temperatures above
additional mechanical driving force, which can aid the Md (>Mrs ), the chemical driving force is so small that
chemical driving force in accomplishing the transforma- deformation is unable to trigger martensite transforma-
tion.[8,18,19] External deformation and the associated tion and the transformation is not possible at all.
mechanical driving force can thus trigger martensite
transformation above the Ms temperature, even though
chemical driving force is not large enough to initiate the
III. MODEL DEVELOPMENT
transformation at such temperatures (>Ms). In other
words, deformation-induced martensitic transformation Considering the initial volume of austenite to be Vc0 , a
can be viewed as dissipation of strain energy accumu- volume Va¢ of martensite is assumed to be formed when
lated during deformation.[8] a true strain e is applied at a constant temperature T.
The magnitude of chemical driving force decreases The volume of austenite remaining untransformed is
linearly with an increase in temperature above Ms (Vc0  Va0 ) and the free energy of the system is given by
(Figure 1).[8] Therefore, a proportionally higher  
mechanical driving force is required to trigger martens- G ¼ Vc0  Va0 gc þ Va0 ga0 þ Va0 A0 þ Vc0  Va0 Ffeg
ite transformation as the temperature is increased ½1
beyond Ms. The critical stress required to cause trans-
formation then increases linearly with temperature where gc is chemical free energy of austenite per unit
(Figure 2)[20] in the temperature range of Ms to Mrs . volume at temperature T, ga¢ is chemical free energy of

2320—VOLUME 39A, OCTOBER 2008 METALLURGICAL AND MATERIALS TRANSACTIONS A


martensite per unit volume at temperature T, A¢ is Equation [6] can therefore be rewritten as
elastic strain energy per unit volume of martensite due
to shape deformation, and F{e} is the energy contribu- df dFfeg
¼ ½8
tion of applied strain e per unit volume of austenite or 1  f ðDGc!a0 þ A0  FfegÞ
the mechanical driving force.
The surface energy term has been neglected in Eq. [1], When the temperature is maintained constant, at true
because it is quite small compared with the elastic strain strain e = 0, f = 0 and at any true strain e = e, f = f.
energy term.[16] Integrating Eq. [8] between e = 0 to e = e and consid-
When the strain increases by an infinitesimal amount ering F{e} = 0 at e = 0
De, the amount of martensite increases by DVa¢ and the
 c!a0 
volume of austenite decreases by DVa¢. The correspond- 1 DG þ A0  Ffeg
ing free energy change between the two strains (e and ln ¼ ln ½9
1f DGc!a0 þ A0
e + De) is given by
 
DG ¼DVa0 ga0  gc þ A0 þ Vc0  Va0 DFfeg The term f can be expressed in terms of the initial
 DVa0 Ffe þ Deg ½2 volume fraction of retained austenite at zero strain, say
fc0 , and the volume fraction of retained austenite that
where DF(e) is the increase in strain energy from strain e remains untransformed at a strain e, say fce , in the
to e + De. following way
Adding and subtracting D Va¢F{e}, Eq. [2] can be
rewritten as Va0 Vc0  Vce fc0  fce
f¼ ¼ ¼ ½10
  Vc0 V c0 fc0
DG ¼DVa0 ga0  gc þ A0  Ffeg þ Vc0  Va0 DFfeg
 DVa0 DFfeg ½3 where Vce is the volume of retained austenite that
remains untransformed at a strain e (all other terms have
been defined earlier).
Ignoring DVa0 DFfeg, as it is a very small quantity, and Hence, Eq. [9] can be rewritten in terms of the
simplifying we have measured retained austenite volume fraction as follows
    0 
0
DG ¼ DVa0 DGc!a þ A0  Ffeg þ Vc0  Va0 DFfeg DGc!a þ A0
fce ¼ fc0 ½11
DGc!a0 þ A0  Ffeg
½4
where DGcfia¢ is the chemical free energy change per unit The term F{e} can be expressed in terms of e as
volume associated with the transformation of austenite
to martensite at temperature T. F fe g ¼ k 0 e q ½12
Dividing throughout by Vc0 , we have
where q is the strain exponent and k¢ is a constant of

DG DVa0  c!a0  Vc0  Va0 proportionality.
¼ DG þ A0  Ffeg þ DFfeg Equation [11] can thus be rewritten as
V c0 V c0 V c0
 0 
½5 DGc!a þ A0
fce ¼ fc0 ½13
DGc!a0 þ A0  k0 eq
Considering that DG = 0 at equilibrium, and approx-
imating DV to dV and DF to dF we have The two empirical constants (k¢ and q) can be
 estimated by fitting Eq. [13] to the experimental data
dVa0  c!a0 0
 Vc0  Va0 (Appendix).
DG þ A  Ffeg þ dFfeg ¼ 0
V c0 Vc0
½6
IV. MODEL APPLICATION
Va0
The term Vc0 ,
represented here by f, is the fraction of Mukherjee et al.[6] have carried out a systematic study
the initial volume of austenite that transforms to of the deformation-induced transformation behavior of
martensite. It can also be interpreted as the ratio of retained austenite in PF, BF, and AM steels at different
the volume fraction of martensite to the initial volume strain rates and temperatures. The details of heat treat-
fraction of austenite in the microstructure. If the starting ment and microstructure can be found in Reference 6.
microstructure is completely austenitic then f is simply Their experimental results are shown in Figure 3.
the volume fraction of martensite. From the preceding Figure 3 also shows the values of fce as a function of
definition of f deformation strain, calculated from Eq. [13] for the best
fit values of ln (k¢) and q, as continuous solid lines. For
dVa0
¼ df ½7 each dataset, DGcfia¢ was calculated as the difference
Vc0 between the chemical free energies of ferrite and

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 39A, OCTOBER 2008—2321


 
p 2m p
A¼ ls2 þ ld2 ½15
8 1m 4
where s and d are the shear and dilatational components
of the invariant plane strain shape deformation, m is the
Poisson’s ratio, and l is the shear modulus. With values
typical of martensite transformation in ferrous alloys,
A = 2.4 9 103 MJ m-3.[16]
Similar fce – e data were collected from various
published literature[3,4,6,25,27–31] and analyzed exactly as
described previously with the value of A¢ quoted
previously and DGcfia¢ calculated using the Thermo-
Calc program.[23] Table I lists the type of matrix
microstructure, deformation conditions like deforma-
tion temperature (T) and strain rate and retained
austenite characteristics like fc0 (the initial volume
fraction of retained austenite at e = 0 and measured
at room temperature), Cc0 (carbon concentration of
retained austenite at e = 0), and Ms temperature of
retained austenite for each experimental dataset consid-
ered. The Ms temperature of the retained austenite was
calculated using Andrews’s formula,[32] taking the
chemical composition of the retained austenite into
account (as was done in the case of DGcfia¢ calculations).
Fig. 3—Variation of untransformed retained austenite fraction fce
The best-fit values of ln (k¢) and q along with the R2
with e in PF, BF, and AM steels tested under different conditions.[6] value for the individual experiments are listed in
The solid continuous lines correspond to the calculated value of fce Table II. In most of the cases R2 is close to 1 indicating
using Eq. [13] and the best-fit values of ln (k¢) and q (RT: room tem- a good fit.
perature). To formulate a general equation for F{e}, the average
of the individual best-fit values of q and ln (k¢) was
calculated and found to be 1.03 and 8.99, respectively.
austenite phases of the same composition, using the All the data were then analyzed together and fce was
Thermo-Calc computational thermodynamics pro- calculated for each dataset using these average values of
gram[23] and taking the chemical composition of q and ln (k¢). Figure 4 shows the agreement between the
retained austenite and the deformation temperature as calculated values of fce and the corresponding experi-
inputs. It is generally assumed that paraequilibrium mental/measured values. Considering the range of alloys
conditions exist during the prior heat treatments (inter- analyzed, the agreement appears reasonably good with
critical annealing and isothermal bainite transforma- most of the points lying within ±0.05.
tion) used to produce TRIP-aided steels.[24–26] However, there are certain cases where the calculated
Therefore, only carbon atoms were assumed to partition values of fce do not agree with the measured values. For
between austenite and ferrite. The carbon concentration example, Figure 5 compares the calculated and mea-
of austenite as estimated by X-ray diffraction measure- sured fce for experiment numbers 1, 8 through 13, and 17
ments was used in all calculations. In certain cases where of Table I. For all of these experiments, the deformation
the intercritical annealing treatment was carried out for temperature (i.e., room temperature) is less than the Ms
very long durations (‡1000 seconds), manganese content temperature of the corresponding retained austenite.
of austenite was assumed to be 1.5 times that of the bulk Majority of the measured fce values are £0.06 and the
manganese concentration. This is in accordance with calculated values of fce are higher than the correspond-
earlier work.[3,4,6,27] Accordingly, depending on the ing measured values. The difference between the calcu-
processing conditions, the austenite chemical composi- lated and actual fce is greater than 0.05 for experiments 1
tion was adjusted before making free energy calcula- and 17 and therefore, they have been shown separately
tions. in Figure 5(a). On re-examining the data, it is found that
The term A¢, which represents the elastic strain energy experiments 1 and 17 correspond to PF steel samples
per unit volume due to shape deformation, was calcu- subjected to tensile tests at room temperature.[6] In these
lated using the relationship[15] cases, stress-assisted transformation is reported to have
 c taken place; even the shoulder section of the tensile
A0 ¼ A ½14 tested samples showed martensite transformation.[6]
r
Similarly, for experiments 8 through 13,[28] the defor-
where cr is the aspect ratio of martensitic plates with plate mation temperature T < Ms and the difference between
thickness equal to 2c and plate length equal to 2r. calculated and measured values of fce are in general
Typically cr is taken to be equal to 0.05.[15] directly related to the difference between the Ms tem-
The term A in Eq. [14] is given by[16] perature and the deformation temperature (Figure 5(b)

2322—VOLUME 39A, OCTOBER 2008 METALLURGICAL AND MATERIALS TRANSACTIONS A


Table I. Experimental Details

Experiment Number Matrix Morphology T (C) Strain Rate (s-1) fc0 Cc0 Ms (C) Reference
-5
1 PF 20 3.3 9 10 0.25 0.77 138.4 6
2 PF 20 2.8 9 10-4 0.12 1.15 -54.2 27
3 PF 20 2.8 9 10-4 0.19 1.40 -138.1 4
4 PF 20 5.6 9 10-4 0.12 1.60 -189.3 25
5 PF 20 5.6 9 10-4 0.19 1.58 -179.8 29
6 PF 20 5.6 9 10-4 0.13 1.48 -142.2 29
7 PF 20 5.6 9 10-4 0.16 1.45 -128.8 29
8 PF 20 6.7 9 10-4 0.09 0.61 237.3 28
9 PF 20 6.7 9 10-4 0.08 0.68 207.7 28
10 PF 20 6.7 9 10-4 0.06 0.73 186.5 28
11 PF 20 6.7 9 10-4 0.08 0.85 119.8 28
12 PF 20 6.7 9 10-4 0.09 0.93 85.9 28
13 PF 20 6.7 9 10-4 0.08 0.97 69.0 28
14 PF 20 1.6 9 10-3 0.20 1.59 -182.6 30
15 PF 20 1.6 9 10-3 0.26 1.41 -117.9 30
16 PF 20 1.6 9 10-3 0.22 1.27 -59.3 30
17 PF 20 3.3 9 10-2 0.25 0.77 138.4 6
18 PF 25 1.0 9 10-3 0.14 1.08 -35.3 31
19 PF 27 2.8 9 10-4 0.15 1.47 -168.9 3
20 AM 20 3.3 9 10-5 0.20 1.08 10.5 6
21 AM 20 2.8 9 10-4 0.13 1.26 -100.7 27
22 AM 20 3.3 9 10-2 0.20 1.08 10.5 6
23 BF 20 3.3 9 10-5 0.15 1.12 15.6 6
24 BF 20 2.8 9 10-4 0.07 1.34 -134.5 27
25 BF 20 2.8 9 10-4 0.13 1.36 -121.2 4
26 BF 20 3.3 9 10-2 0.15 1.12 15.6 6
27 PF 55 1.0 9 10-3 0.14 1.08 -35.3 31
28 PF 75 1.0 9 10-3 0.14 1.08 -35.3 31
29 PF 100 2.8 9 10-4 0.12 1.15 -54.16 27
30 PF 150 3.3 9 10-5 0.25 0.77 138.4 6
31 PF 150 3.3 9 10-3 0.25 0.77 138.4 6
32 PF 150 3.3 9 10-2 0.25 0.77 138.4 6
33 AM 100 2.8 9 10-4 0.13 1.26 -100.7 27
34 AM 150 3.3 9 10-5 0.20 1.08 10.5 6
35 AM 150 3.3 9 10-4 0.20 1.08 10.5 6
36 AM 150 3.3 9 10-2 0.20 1.08 10.5 6
37 BF 100 2.8 9 10-4 0.07 1.34 -134.5 27
38 BF 150 3.3 9 10-5 0.15 1.12 15.6 6
39 BF 150 3.3 9 10-2 0.15 1.12 15.6 6
Note: T = deformation temperature; fc0 = initial volume fraction of retained austenite measured at room temperature; Cc0 = carbon con-
centration of retained austenite at e = 0; and Ms = temperature of retained austenite calculated using Andrews’s formula.[32]

and Table I), with the difference being larger for higher where stress-assisted martensite transformation appears
Ms temperatures. to be the dominant mode of transformation, F{e}
The nature of the critical stress-temperature plot with estimated from the average values of ln (k¢) and q may
a discontinuity at Mrs in the schematic illustration shown not be accurate. Referring once again to Figure 2, the
in Figure 2 implies that the stress and hence deforma- average values of F{e} calculated previously is an
tion energy or mechanical driving force required for the underestimation when applied to stress-assisted trans-
stress-assisted and strain-induced modes of transforma- formation. This results in over prediction in the values
tion are different. As discussed earlier, the mechanical of fce (Figure 5). If F{e} values appropriate to stress-
driving force required for strain-induced transformation assisted transformation mechanism is used, the predic-
is smaller when austenite undergoes plastic deformation tions would improve. Indeed Figures 3(a) and (b)
by slip. Considering that most of the experiments listed representing experiments 17 and 1, respectively, confirm
in Table I were conducted at temperatures well above the same. Even for experiments 8 through 13, the
the Ms temperature of the corresponding retained agreement between the calculated and actual fce
austenite, it is possible that the average values of ln improves when the Ms temperature is closer to the
(k¢) and q calculated are biased toward the strain- deformation temperature. The possible reason for this
induced mechanism of martensite transformation. could be that as the Ms temperature reduces, the
Therefore, for experiments 1, 8 through 13, and 17, retained austenite becomes more stable and, hence, the

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 39A, OCTOBER 2008—2323


Table II. Fitting Parameters for Experiments Listed
in Table I

Experiment
Number q Ln (k¢) R2 Reference
1 0.67 10.29 0.99 6
2 0.59 10.19 0.97 27
3 0.86 7.74 0.96 4
4 1.64 8.16 0.96 25
5 1.22 9.51 0.98 29
6 1.34 9.22 0.99 29
7 1.55 10.16 0.97 29
8 0.85 10.97 0.96 28
9 1.46 12.20 0.95 28
10 0.68 10.52 0.92 28
11 0.60 8.94 0.97 28
12 0.97 9.45 0.98 28
13 0.94 8.85 0.98 28
14 1.03 8.03 0.99 30
15 0.95 8.71 0.97 30
16 1.15 9.89 0.99 30
17 0.89 11.11 0.86 6 Fig. 4—Comparison of calculated and actual values of fce for experi-
18 0.99 9.29 0.99 31 ments listed in Table I. Solid line represents complete agreement be-
19 1.34 9.46 0.99 3 tween calculated and actual fce , whereas the dotted lines represent a
deviation by ±0.05.
20 1.16 9.43 0.99 6
21 1.29 13.11 0.96 27
22 0.79 8.36 0.98 6
23 1.39 9.97 0.99 6
24 0.92 9.76 0.90 27
25 0.57 7.14 0.88 4
26 1.29 9.83 0.99 6
27 0.84 8.18 0.99 31
28 0.81 7.46 0.99 31
29 0.71 7.02 0.96 27
30 1.47 9.89 0.88 6
31 1.07 8.49 0.82 6
32 0.47 6.75 0.76 6
33 0.80 7.12 0.89 27
34 1.43 8.82 0.92 6
35 1.11 7.79 0.89 6
36 0.31 5.43 0.94 6
37 1.07 6.51 0.94 27 Fig. 5—Comparison of calculated and actual values of fce for experi-
38 1.68 9.33 0.96 6 ments (a) 1, 17 and (b) 8 through 13 of Table I. Solid line represents
complete agreement between calculated and actual fce , whereas dot-
39 1.19 7.48 0.94 6
ted lines represent a deviation by ±0.05.

available for triggering martensitic transformation is


extent of stress-assisted transformation is lower. There- low. The magnitude of mechanical driving force F{e}
fore, the F{e} calculated using the average values for ln should thus depend on the deformation temperature
(k¢) and q could better represent the actual transforma- also. Because this factor has not been taken into account
tion mechanism leading to higher accuracy of predic- in calculating fce at high temperatures, the calculated fce
tion. This is clear from experiment 13, for which the values are lower than the actual values.
calculated fce matches the measured values. The agreement between calculated and measured
Figure 6 compares the calculated and measured values of fce for room temperature deformation (exper-
values of fce for experiments carried out at temperatures iments 2 through 7, 14 through 16, and 18 through 26 of
greater than room temperature (experiments 27 through Table I) is shown in Figure 7. The Ms temperature of
39 of Table I). The calculation seems to generally the retained austenite in these cases is lower than the
underestimate the value of fce . This effect is most deformation temperature. Barring a few exceptions,
prominent for tests carried out at relatively high strain particularly when the measured fce is small, the estima-
rates (open circles). The chemical free energy difference tions of the model are reasonably accurate, considering
between austenite and martensite is lower at high the inherent experimental error expected in the mea-
deformation temperatures (Figure 1), and it decreases surement of fce using X-ray diffraction, and the simpli-
further if adiabatic heating takes place during high fying assumptions discussed subsequently. It should also
strain rate experiments. Because slip is easier at higher be noted that measurement of the volume fraction of
temperatures, a large proportion of the deformation retained austenite using X-ray diffraction is particularly
energy is consumed by slip.[9] As a result, the energy unreliable when volume fraction is less than 0.05.[33]

2324—VOLUME 39A, OCTOBER 2008 METALLURGICAL AND MATERIALS TRANSACTIONS A


martensite c/r is known to vary with temperature,
driving force, austenite strength, extent of transforma-
tion, etc.[34] Strain hardening of austenite during defor-
mation can also lower the aspect ratio. Similarly, plastic
accommodation of martensite plates or formation of
self-accommodating variants alters the value of the
constant A¢ used in the model.[34]
The accuracy of the calculations depends on the
correct estimation of the chemical driving force for
martensite transformation, which in turn depends on the
thermodynamic database used. In the present work, the
chemical driving force for retained austenite to mar-
tensite transformation (DGcfia¢) for all the cases was
calculated using Thermo-Calc. For some alloys,
MTDATA was also used to cross-check the values
obtained by Thermo-Calc and the difference was not
significant. Moreover, the chemical driving force for
martensite transformation has been calculated using the
average concentration of the alloying elements in the
Fig. 6—Comparison of calculated and actual values of fce for experi-
carbon enriched retained austenite. This need not be
ments 27 through 39 of Table I, carried out at temperatures greater strictly correct in all the alloys analyzed because
than room temperature. Open circles correspond to experiments 32, elements like C and Mn are known to segregate.[2,24,35]
36, and 39, carried out at a strain rate of 3.3 9 10-2 s-1. Solid line Thus, regions with lower concentration of alloying
represents complete agreement between calculated and actual fce , elements would transform rather easily whereas rela-
whereas dotted lines represent a deviation by ±0.05.
tively enriched regions would be more resistant to
transformation.[36] Further complications arise if defor-
mation of a complex microstructure like the one in
TRIP-aided steel is considered. Stress and strain are not
expected to be uniform in all phases, and their distri-
bution amongst the various phases would depend on the
relative mechanical properties of the individual
phases.[37] Thus, estimation of a general value of F{e}
for all the steels based on the average macroscopic strain
may not be strictly accurate.
The surface energy term has not been considered here
as it is quite small compared with the elastic strain
energy term.[16] But when the austenite grain size is very
small (typically <1 lm), the surface/interfacial energy
term becomes significant.[38] In such cases, including this
term may improve the predictions.

VI. STRESS-ASSISTED vs STRAIN-INDUCED


TRANSFORMATION
The thermodynamics of stress-assisted martensitic
Fig. 7—Comparison of calculated and actual values of fce for experi- transformation was studied for the first time by Patel
ments 2 through 7; 14 through 16; and 18 through 26 of Table I, and Cohen.[18] Considering an invariant plane strain
where deformation was done at room temperature (20 C). The Ms shape deformation, the thermodynamic contribution of
temperature was less than the room temperature in all these cases so
that deformation was carried out at temperatures higher than the Ms
applied stress (F{r}) was estimated by calculating the
temperature. Solid line represents complete agreement between calcu- interaction work of the resolved stress acting through
lated and actual fce , whereas dotted lines represent a deviation by the shape strain of the most favorably oriented mar-
±0.05. tensite variant. For uniaxial tension, the magnitude of
the driving force contribution per unit value of applied
stress was calculated to be 0.86 J mol-1 MPa-1.[18,19]
V. SOURCES OF ERROR When this quantity is multiplied by the applied stress,
the value of F{r} (in J mol-1) is obtained.
The deviation of the calculated values of fce from the Figures 8, 9, and 10 compare F{r} with F{e} for PF,
measured values can be attributed to the simplifying AM, and BF steels tested under different deformation
assumptions of the model. conditions described in Reference 6. The F{e} was
The parameter A¢ = A(c/r) has been assumed to be calculated empirically using Eq. [13] with experimental
constant for all the cases. However, the aspect ratio of values of fce and fc0 (estimated using X-ray diffraction);

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 39A, OCTOBER 2008—2325


stress-assisted mechanism. Though stress-assisted trans-
formation is usually defined in the elastic regime, F{r}
has been calculated here for the entire strain range for
the sake of comparison, which may not be valid for
higher strains (beyond the elastic limit). Moreover, the
calculations by Patel and Cohen[18] were made for the
most favorably oriented nuclei, which surely get con-
sumed by the time higher strains are reached.
The plots for PF steel tested at 150 C (Figure 8(b))
and AM and BF steels tested at either temperature
(Figures 9 and 10), show that F{e} is lower than the
corresponding F{r} at low strains (0 to 0.1 or 0.2). In
other words, the driving force contribution of strain
Fig. 8—Comparison of F{r} and F{e} for PF steels tested at (a) RT
required for bringing about martensite transformation
and (b) 150 C. estimated empirically using Eq. [13] is lower than the
free energy contribution of applied stress as calculated
from Patel and Cohen’s formulation.[18] This is typical
of the strain-induced mechanism of transformation and
implies that the transformation of austenite is easier and
requires lower mechanical driving force when plastic
deformation of austenite by slip takes place first. The
observation is consistent with the schematics of mar-
tensite transformation shown in Figure 2; it is clear that
at temperatures greater than Mrs , strain-induced mar-
tensite transformation can be triggered at stresses that
are lower than is expected from a simple extrapolation
of the critical stress required to initiate stress-assisted
transformation at temperatures below Mrs . This is
because plastic deformation (by slip) introduces new
nucleation sites for martensite and thus makes the
transformation easier.[17]
Fig. 9—Comparison of F{r} and F{e} for AM steels tested at (a) However, in the case of PF steel tested at room
RT and (b) 150 C.
temperature (Figure 8(a)), F{r} is much lower than
F{e} indicating stress-assisted transformation. Results
reported in Reference 6 indeed show stress-assisted
transformation in this case.

VII. CONCLUSIONS AND SUMMARY


A model based on thermodynamics has been devel-
oped for the deformation-induced transformation of
retained austenite in TRIP-aided steels. The model takes
into account the mechanical driving force or the energy
contribution of the deformation strain, which has been
determined empirically from reported experimental
data. The predictions of the model are generally in
good agreement with the experimental measurements. It
Fig. 10—Comparison of F{r} and F{e} for BF steels tested at (a) has been argued that strain-induced transformation of
RT and (b) 150 C. austenite to martensite takes place in the case of AM
and BF steels deformed at room temperature and at
150 C and PF steel tested at 150 C. In all of these
DGcfia¢ and A¢ were calculated as in Section IV. The F{e} cases, it has been shown that the mechanical driving
so calculated thus represents the free energy contribu- force required to trigger strain-induced martensite
tion of strain required to bring about the observed transformation is much smaller than expected from the
transformation. For each strain level at which fce was theory of stress-assisted transformation. This can be
measured, the corresponding true stress value was attributed to new nucleation sites introduced by plastic
estimated from the respective flow curve[6] and used to deformation, which makes the transformation easier.
calculate the value of F{r} as discussed previously. For PF steel tested at room temperature, stress-assisted
These F{r} values correspond to the driving force mode appears to be the dominant mechanism of
contribution of applied stress to bring about transfor- martensite transformation, which corroborates the
mation of the most favorably oriented variant by the experimental results presented in Reference 6.

2326—VOLUME 39A, OCTOBER 2008 METALLURGICAL AND MATERIALS TRANSACTIONS A


APPENDIX k¢ constant of proportionality relating the free
energy contribution of deformation to true
This appendix explains how the best fit values of the
strain
two empirical constants (k¢ and q) in Eq. [13] were
Ms martensite start temperature
obtained. Equation [13] is as shown
Mrs maximum temperature up to which austenite
 0  can transform to martensite by the stress-
DGc!a þ A0
fce ¼ fc0 ½A1 assisted mechanism
DGc!a0 þ A0  k0 eq
Md highest temperature at which deformation is
able to trigger austenite to martensite
The preceding equation was rearranged as transformation
 0
 fc
 q strain exponent
DGc!a þ A0  1  0 ¼ k0 eq ½A2 r semilength of martensite plates
f ce R2 coefficient of determination
s shear component of invariant plane strain
The preceding equation was linearized as shape deformation associated with martensite
 formation
0
 fc

T temperature
ln DGc!a þ A0  1  0 ¼ lnðk0 Þ þ q lnðeÞ ½A3 T0 temperature at which austenite and ferrite of
fce
the same composition have the same free
energy
The left hand side (LHS) of Eq. [A3] was calculated as Umech driving force contribution of mechanical
a function of ln (e), and using least-squares analysis, the deformation
best-fit values of ln (k¢) and q were obtained. Va¢ volume of martensite formed after strain e
Vc0 volume of austenite at zero strain
Vce volume of austenite that remains
LIST OF SYMBOLS AND ABBREVIATIONS untransformed at strain e
a¢ martensite
SYMBOLS d normal component of the invariant plane
strain shape deformation associated with
A parameter associated with the strain energy martensite formation
for the formation of martensite e true strain
A¢ elastic strain energy per unit volume of c austenite
martensite due to shape deformation l shear modulus
c semithickness of martensite plates m Poisson’s ratio
C c0 carbon content of retained austenite at zero r true stress
strain
f fraction of the initial volume of austenite that
transforms to martensite
f c0 initial volume fraction of retained austenite ABBREVIATIONS
(measured at room temperature and at zero AM annealed martensite
strain) BF bainitic ferrite
f ce volume fraction of retained austenite that LHS left-hand side
remains untransformed after true strain e PF polygonal ferrite
F{e} free energy contribution of applied strain e RT room temperature
per unit volume of austenite TRIP transformation induced plasticity
F{r} free energy contribution of applied stress per
unit volume
ga¢ chemical free energy of martensite per unit
volume at temperature T REFERENCES
gc chemical free energy of austenite per unit 1. V.F. Zackay, E.R. Parker, D. Fahr, and R. Busch: Trans. ASM,
volume at temperature T 1967, vol. 60, pp. 252–59.
2. O. Matsumura, Y. Sakuma, and H. Takechi: Trans. ISIJ, 1987,
G free energy of the system vol. 21, pp. 570–79.
DGch chemical free energy difference of austenite 3. K. Sugimoto, M. Kobayashi, and S. Hashimoto: Metall. Trans. A,
and martensite or chemical driving force 1992, vol. 23A, pp. 3085–91.
DGcrit critical chemical driving force required for 4. K. Sugimoto, T. Iida, J. Sakaguchi, and T. Kashima: ISIJ Int.,
initiating the transformation of austenite to 2000, vol. 40, pp. 902–08.
5. K. Sugimoto, A. Kanda, R. Kikuchi, S. Hashimoto, T. Kashima,
martensite and S. Ikeda: ISIJ Int. 2002, vol. 42, pp. 910–15.
DGcfia¢ driving force for austenite to martensite 6. M. Mukherjee, O.N. Mohanty, S. Hashimoto, T. Hojo
transformation, calculated as the free energy K. Sugimoto: ISIJ Int., 2006, vol. 46, pp. 316–24.
difference between austenite and ferrite of the 7. G.B. Olson and M. Cohen: Metall. Trans. A, 1975, vol. 6A,
pp. 791–95.
same composition

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 39A, OCTOBER 2008—2327


8. H.C. Shin, T.K. Ha, and Y.W. Chang: Scripta Mater., 2001, 25. M. De Meyer, D. Vanderschueren, and B.C. De Cooman: Proc.
vol. 45, pp. 823–29. 41st Mechanical Working and Steel Processing Conf., Iron and
9. T. Angel: JISI, 1954, vol. 177, pp. 165–74. Steel Society, Baltimore, MD, 1999, pp. 265–76.
10. D.C. Ludwigson and J.A. Berger: JISI, 1969, vol. 192, pp. 63–69. 26. M. Takahashi and H.K.D.H. Bhadeshia: Trans. Jpn. Inst. Met.,
11. O. Matsumura, Y. Sakuma, and H. Takechi: Scripta Mater., 1987, 1991, vol. 32, pp. 689–96.
vol. 21, pp. 1301–06. 27. K. Sugimoto, M. Misu, M. Kobayashi, and H. Shirasawa: ISIJ
12. M.Y. Sherif, C. Garcia-Mateo, T. Sourmail, and H.K.D.H. Int., 1993, vol. 33, pp. 775–82.
Bhadeshia: Mater. Sci Technol., 2004, vol. 20, pp. 319–22. 28. P.J. Jacques, J. Ladrière, and F. Delannay: Metall. Mater. Trans.
13. M. Mukherjee, S.B. Singh, and O.N. Mohanty: Mater. Sci. Eng. A, 2001, vol. 32A, pp. 2759–68.
A, 2006, vol. 434, pp. 237–45. 29. D. Krizan, B.C. De Cooman, and J. Antonissen: Proc. Int. Conf.
14. M. Mukherjee, S.B. Singh, and O.N. Mohanty: Mater. Sci. on Advanced High Strength Sheet Steels for Automotive Applica-
Technol., 2007, vol. 23, pp. 338–46. tions, AIST, Winter Park, CO, 2004, pp. 205–16.
15. J.W. Christian: ICOMAT 79, Int. Conf. Martensitic Transfor- 30. A.K. De, R.S. Kircher, J.G. Speer, and D.K. Matlock: Proc. Int.
mations, Cambridge, MA, 1979, pp. 220–34. Conf. on Advanced High Strength Sheet Steels for Automotive
16. C.M. Wayman and H.K.D.H. Bhadeshia: in Physical Metallurgy, Applications, AIST, Winter Park, CO, 2004, pp. 337–47.
R.W. Cahn and P. Hassen, eds., Elsevier Science Publishers, 31. M. Radu, J. Valy, A.F. Gourgues, F. Le Strat, and A. Pineau:
Amsterdam, 1996, pp. 1507–54. Scripta Mater., 2005, vol. 52, pp. 525–30.
17. M. Cohen, E.S. Machlin, and V.G. Paranjpe: Thermodyamics in 32. K.W. Andrews: JISI, 1965, vol. 203, pp. 721–27.
Physical Metallurgy, ASM, Cleveland, OH, 1950, pp. 242–70. 33. B.D. Cullity: Elements of X-Ray Diffraction, 2nd ed., Addison-
18. J.R. Patel and M. Cohen: Acta Metall., 1953, vol. 1, pp. 531–38. Wesley Publishing Co., Inc., Reading, MA, 1978, pp. 411–15.
19. G.B. Olson and M. Cohen: Metall. Trans. A, 1982, vol. 13A, 34. G. Ghosh and V. Raghavan: Mater. Sci. Eng., 1986, vol. 79,
pp. 1907–14. pp. 223–31.
20. G.B. Olson and M. Azrin: Metall. Trans. A, 1978, vol. 9A, 35. J. Mahieu, D. Van Dooren, and B.C. De Cooman: in Proc. Int.
pp. 713–21. Symp. on Transformation and Deformation Mechanisms in
21. G.B. Olson and M. Cohen: J. Less Common Met., 1972, vol. 28, Advanced High Strength Steels, M. Militzer, W.J. Poole, and
pp. 107–18. E. Essadiqi, eds., CIM, Montreal, 2003, pp. 21–35.
22. G.F. Bolling and R.H. Richman: Acta Metall., 1970, vol. 18, 36. A. Itami, M. Takahashi, and K. Ushioda: ISIJ Int., 1995, vol. 35,
pp. 673–81. pp. 1121–27.
23. B. Sundman, B. Jansson, and J.O. Andersson: Calphad, 1985, 37. P.J. Jacques: Curr. Opin. Solid State Mater. Sci., 2004, vol. 8,
vol. 9, pp. 153–90. pp. 259–65.
24. G.R. Speich, V.A. Demarest, and R.L. Miller: Metall. Trans. A, 38. J. Wang and S. van der Zwaag: Metall. Mater. Trans. A, 2001,
1981, vol. 12A, pp. 1419–28. vol. 32A, pp. 1527–39.

2328—VOLUME 39A, OCTOBER 2008 METALLURGICAL AND MATERIALS TRANSACTIONS A

You might also like