You are on page 1of 11

Available online at www.sciencedirect.

com

Acta Materialia 59 (2011) 738–748


www.elsevier.com/locate/actamat

Martensite ? austenite phase transformation kinetics in an


ultrafine-grained metastable austenitic stainless steel
S. Rajasekhara, P.J. Ferreira ⇑
Materials Science and Engineering Program, The University of Texas at Austin, TX 78712, USA

Received 17 August 2010; accepted 5 October 2010


Available online 31 October 2010

Abstract

A generalized phase transformation kinetics model is used to understand the martensite to austenite transformation in a cold-rolled
and annealed metastable AISI 301LN ultrafine-grained austenitic stainless steel. The model shows that the presence of interstitial nitro-
gen and heavy cold-rolling is important to promote fast transformation kinetics, through rapid nitrogen-diffusion and austenite nucle-
ation at austenite/martensite phase boundaries.
Ó 2010 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Austenite; Martensite; Stainless steel; Phase transformation kinetics

1. Introduction which takes place during the annealing process. Krauss


and Cohen [10] demonstrated for the first time, the reverse
The process of heavy cold-rolling and subsequent short- martensite transformation (a0 ? c) in a Fe–33.5% Ni alloy.
time annealing of metastable austenitic stainless steels (SS) Subsequently, Breedis [11] showed that in Fe–Cr–Ni alloys
has generated considerable interest in the research commu- the reversion products consisted of thin austenitic twins
nity and industry because SS with ultrafine austenite grains formed in a martensitic matrix. Jana and Wayman [12],
may be achieved. The presence of ultrafine grains results in Kessler and Pitsch [13] and Guy et al. [14] showed
SS with high strength and ductility, making these materials experimentally that, depending on the SS composition and
desirable for use in structural and automotive industries. heating rates, the a0 ? c transformation may be a time-
This concept has been demonstrated in specialty non- dependent phenomenon where the austenite phase fraction
commercial SS grades [1,2], and more recently in commer- increases as a function of annealing time. Montaneri [15]
cial AISI 301, 301LN and 304 SS grades [3–9]. In fact, performed internal friction and dynamic modulus experi-
excellent mechanical properties with yield strengths exceed- ments on a 95% cold-rolled and annealed AISI 304 SS,
ing 700 MPa and elongation of approximately 35% were and observed that an incubation time preceded the start of
reported in cold-rolled and annealed AISI 301LN SS [8,9]. the a0 ? c transformation. The occurrence of this incuba-
The basis for these properties relies on initially cold rolling tion period was attributed to the stress relaxation of the
a metastable austenitic SS to form lath- and dislocation- parent martensite phase and the diffusion of interstitial
cell-type martensite (a0 ) which, upon annealing, transform atoms to the a0 /a0 interface during annealing.
into ultrafine austenite (c) grains. More recently, Tomimura et al. [16] demonstrated that,
In the past, several studies have qualitatively discussed depending on the alloy composition, austenite nucleates at
the a0 ? c phase transformation in various SS grades, martensite grain boundaries and grows in size when heated
for long annealing durations. According to Tomimura
⇑ Corresponding author. Tel.: +1 512 471 3244; fax: +1 512 471 7681. et al. [16], the initial shape of the austenite grains depends
E-mail address: ferreira@mail.utexas.edu (P.J. Ferreira). on the morphology of the parent martensite phase. If the

1359-6454/$36.00 Ó 2010 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2010.10.012
S. Rajasekhara, P.J. Ferreira / Acta Materialia 59 (2011) 738–748 739

cold-rolled martensite structure is lath-type, the austenite where they assume that the embryos of a new phase nucle-
phase nucleates at the lath boundaries and intersections, ate randomly from the time-dependent parent phase. In the
and grows from these sites into the martensite lath. On the case of the a0 ? c phase transformation, this relationship
other hand, if the parent martensite morphology is disloca- may be expressed as:
tion-cell-type (comprising dislocation forests and ultrafine Z
4p t 3 3
lath martensite), equiaxed ultrafine austenite grains nucleate lnð1  nc ðT ; tÞÞ ¼  v ðT ; tÞJ ðT ; tÞðt  sÞ ds ð1Þ
3 0
at martensite grain boundaries. The present authors and
other research groups have consistently observed these find- where nc(T, t) is the time (t) and temperature (T) dependent
ings in several cold-rolled and annealed commercial and spe- austenite phase volume fraction, v(T, t) the austenite grain
cialty SS grades [17–20]. Qualitatively, this time-dependent growth velocity, J(T, t) the austenite phase nucleation rate
“diffusion-type” mechanism is characterized by: (i) a wide and s the incubation time for the formation of the first aus-
annealing temperature range; (ii) the formation of defect- tenite nuclei. For simplicity, we assume the nucleation rate
free equiaxed austenitic grains which grow in size with time; to be independent of annealing duration. Eq. (1) thus re-
(iii) a wide austenite grain size distribution; and (iv) possible duces to:
formation of secondary phase precipitates [18]. Z
4pJ ðT Þ t 3 3
Clearly, there is experimental and phenomenological lnð1  nc ðT ; tÞÞ ¼  v ðT ; tÞðt  sÞ ds ð2Þ
3 0
evidence that seem to confirm the diffusion-type phase
transformation mechanism in cold-rolled and annealed While the term J(T) will be discussed later, the integral in
ultrafine-grained austenitic SS. However, despite the avail- Eq. (2) may be solved by using an explicit time and temper-
able work, there is a lack of quantitative understanding ature dependence relationship for v(T, t), which is given by
regarding the kinetics of the a0 ? c phase transformation. [20]:
In addition, several fundamental concepts involved in this 1

phase transformation, such as: (i) the type of the diffusing ðktÞn
vðT ; tÞ ¼ ð3Þ
species driving the phase transformation; and (ii) the nat- nt
ure of austenite/martensite grain boundary junctions where where k is the austenite grain growth parameter, n is the
the transformation occurs are not yet known. An under- austenite grain growth exponent and t is the annealing
standing of these factors is critical for tailoring the compo- time. As has been discussed elsewhere [20,23–26], the grain
sition of metastable austenitic SS grades, as well as for growth parameter k is a measure of the material’s grain
optimizing the cold-rolling procedure and subsequent boundary mobility, which depends on annealing tempera-
annealing to promote an efficient a0 ? c transformation, ture and grain growth activation energy, while the grain
which can lead to the production of ultrafine-grained SS. growth exponent n depends on the annealing temperature,
In this work, we apply the general phase kinetics rela- grain orientation, texture, residual strain and soluble impu-
tionship proposed by Erukhimovitch and Baram [21,22], rities. By combining Eqs. (2) and (3), the generalized phase
and quantitatively address the a0 ? c phase transformation transformation kinetics relationship can be rewritten as:
in a cold-rolled and annealed metastable ultrafine-grained Z 1
!3
austenitic AISI 301LN SS grade, the composition of which 4pJ ðT Þ t ðktÞn 3
lnð1  nc ðt; T ÞÞ ¼  ðt  sÞ ds ð4Þ
is shown in Table 1. The AISI 301LN SS is an ideal candi- 3 0 nt
date for this study because: (i) it has a wide commercial use;
(ii) it exhibits a very rapid (1–100 s) diffusion-driven a0 ? c Solving Eq. (4) for nc(T, t), we obtain the expression
transformation within a wide annealing temperature range 3 3
!
(700–1000 °C) [8,18]; and (iii) there is experimental phase 4pJ ðT Þk n tnþ4
nc ðt; T Þ ¼ 1  exp  ð5Þ
fraction and grain size data against which our analysis 3n3
may be compared [8,18,20].
Eq. (5) gives the austenite phase fraction nc as a function of
2. Kinetics model grain growth kinetic parameters k, n and the nucleation
rate J(T). Not surprisingly, Eq. (5) has a form similar to
2.1. Basis for the model the kinetics relationships developed by Avrami [27,28]
and Johnson and Mehl [29]. However, Eq. (5) is written
Erukhimovitch and Baram [21,22] proposed a general in terms of experimental parameters, namely the nucleation
kinetics relationship for solid-state phase transformations rate J(T) and the grain growth parameters k and n.

Table 1
Weight percent and mole fractions of the various elements present in AISI 301LN SS.
C N Ni Cr Mn Si Cu Mo Fe
wt.% 0.017 0.15 6.5 17.3 1.29 0.52 0.2 0.15 Balance
Mole fraction 0.0008 0.006 0.063 0.186 0.013 0.005 0.002 0.002 0.744
740 S. Rajasekhara, P.J. Ferreira / Acta Materialia 59 (2011) 738–748

According to classical nucleation theory, the time- grain boundary area of the eliminated a0 /a0 interface and
dependent nucleation rate J(T, t) in solid-state transforma- the volume free energy change DGv(T). Thus, DG(T)
tions is given by [30,31]: may be expressed as [33]:
   s
DG 4ðbra0 =c  ara0 =a0 Þ
3
J ðT ; tÞ ¼ ZbC 1 exp  exp  ð6Þ DG ðT Þ ¼ ð10Þ
kBT t 27c2 ðDGv ðT ÞÞ
2

where Z is the Zeldovich non-equilibrium factor that ac- where a is the eliminated martensite grain boundary area
counts for the nuclei that exceed the critical nucleus size, during the a0 ? c phase transformation, b is the new aus-
b is the rate at which atoms are added to the critical nu- tenite/martensite grain boundary area, c is the new austen-
cleus, C1 is the concentration of available nucleation sites, ite nucleus volume [33] and DGv(T) is the volume free
DG is the activation energy for heterogeneous austenite energy change due to the martensite to austenite transfor-
nucleation and s is the incubation time. As mentioned pre- mation. The parameters a, b and c depend on the geometry
viously, we initially assume the nucleation rate to be con- of the nucleating grain (to be discussed later). On the basis
stant within 1–100 s of annealing. Later, in Section 3, we of Eqs. 7, 8a, 8b, 9, 10, Eq. (5) may be rewritten as:
will show that this assumption is justified. Thus, the time-   3 1
0 DG ðT Þ
independent nucleation rate may be written as: 4pZðT ÞbðT ÞC gb exp 
3
k n tnþ4
1
  @
nc ðt; T Þ ¼ 1  exp 
kB T
A
DG 3n3
J ðT Þ ¼ ZbC 1 exp  ð7Þ
kBT
ð11Þ
The relationships for Z and b have been developed by
Johnson et al. [31], and modified for the specific case of where all the terms retain their previous definitions. At this
austenite nucleation as shown below: point, to calculate nc(t, T) we still need to determine
DG(T). Therefore, the values for the a0 /c interfacial
2
V a0 Fe ½DGv ðT Þ energy, the a0 /a0 interfacial energy, the volume free energy
ZðT Þ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð8aÞ
8p k B TK j ra0 =c change DGv(T), the martensite grain boundary width d gb a0 ,
the grain size da0 and the magnitudes of a, b and c for dif-
16pr2a0 =c DðT ÞLj ferent austenite/martensite grain junctions in Eq. (10) must
bðT Þ ¼ 2
ð8bÞ
a4a0 ðDGv ðT ÞÞ be found. We estimate these parameters in the subsequent
sections.
where Va 0 Fe is the average volume of an iron atom in the
martensite phase, DGv(T) is the Gibbs free energy change
2.2. Estimation of the interfacial energies ra0 /a 0 and ra/c
per unit volume of martensite to austenite phase transfor-
mation, Kj and Lj are geometric parameters that depend
When an austenite grain nucleates within the martensite
on the a0 -phase and c-phase interfacial energies, D(T) is
matrix, the martensite grain boundary is eliminated and a0 /
the temperature-dependent diffusion coefficient of the spe-
c interface is created. Prior research has shown that
cies that govern the martensite to austenite transformation,
h1 1 0i-type austenite (fcc) boundaries form at h1 1 1i-type
ra 0 /c is the martensite–austenite interfacial energy, and aa 0 is
martensite (bcc) boundaries [34,35]. Keeping this in mind,
the martensite lattice parameter.
and based on theoretical grain boundary calculations per-
To calculate J(T) in Eq. (7), we further need to estimate
formed by Shibuta et al. [36] for a specialty Fe–Cr alloy,
C1, the concentration of nucleation sites available for aus-
the bcc phase h1 1 1i-type grain boundary energy ra0 /a 0 is
tenite nucleation, which depends on the type of nucleation
assumed to be approximately 1.3 J m2 [35]. Furthermore,
sites considered. On the basis of several studies on cold-
because the interfaces are crystallographically related
rolled SS grades and Fe–X alloys [7,8,16,17,32], we have
through the Nishiyama–Wasserman (½110c ==½11 1a0 ) rela-
assumed that the austenite nucleates on martensite grain
tionship [34,35,37], we assume that the energy of the new
boundaries. The concentration of martensite grain bound-
a0 /c interface is equal to the eliminated a0 /a0 interface. This
aries C gb 1 available for austenite nucleation is estimated
is a reasonable assumption, as discussed by Gjostein et al.
according to [26]:
[38], who have experimentally determined the bcc/fcc inter-
dgb facial energy ratio for Fe–C alloys to be approximately 1.
a0
C gb
1 ¼ C 0  ; ð9Þ
da0
2.3. Estimation of the volume free energy change DGv(T)
where C0 is the total equilibrium concentration of nucle-
ation sites available, d gb
a0 is the martensite grain boundary DGv(T) is the free energy change per unit volume
width and da0 is the average martensite grain size. between the product-phase austenite and the parent-phase
Finally, to determine J(T) in Eq. (7), the activation martensite. It can be expressed (in J m3) as:
energy for heterogeneous nucleation DG(T) must be calcu-
0
lated. DG(T) depends on the volume of the austenite Gc ðT Þ Ga ðT Þ
DGv ðT Þ ¼  ð12Þ
nucleus, the grain boundary area of the a0 /c interface, the Vc Va
0
S. Rajasekhara, P.J. Ferreira / Acta Materialia 59 (2011) 738–748 741

0
where Gc(T) and Ga (T) are the free energies per mole of the parameters describing the interaction between the ith and
0
austenite and martensite phases respectively, and Vc, Va the jth element in the respective phases. Miettinen has
the molar volumes of austenite and martensite phases tabulated the first-order interaction parameters Lij, for
respectively. The free energy per mole contribution of an the bcc and the fcc phases in commercial SS, to determine
element j in the martensite and austenite phases can be the excess free energy in the respective phases [40,42].
written as [39,40]: Finally, we ignore the magnetic ordering contribution to
Xn X
n the free energy because, as has been discussed elsewhere
0 0 0 0
Ga ðT Þ ¼ xj Gaj ðT Þ þ xj RT lnðxj Þ þ Gaxs ðT Þ þ Gam ðT Þ [40], their values are close to zero above room temperature.
j¼1 j¼1 Thus Eq. (11) may be rewritten as:
ð13aÞ Pn Pn P Pn
c
j¼1 xj Gj ðT Þ þ lnðxj Þ þ n1
j¼1 xj RT i¼1
c
j¼iþ1 xi xj Lij
DGv ðT Þ ¼
X
n X
n V c

Gc ðT Þ ¼ xj Gcj ðT Þ þ xj RT lnðxj Þ þ Gcxs ðT Þ þ Gcm ðT Þ Pn a0 Pn Pn1 Pn a0


j¼1 xj Gj ðT Þ þ j¼1 xj RT lnðxj Þ þ i¼1 j¼iþ1 xi xj Lij
j¼1 j¼1  0 ð17Þ
Va
ð13bÞ The mole fractions of the elements (xi and xj) present in
where xj is the mole fraction of the jth element present in AISI 301LN SS are shown in Table 1. The relationships for
0
the SS grade, Gaj ðT Þ and Gcj ðT Þ are thePstandard free energy the free energy contributions by individual elements to the
n
of the jth element in a0 and c phases, j¼1 xj RT lnðxj Þ is the a0 and c phases, as well as various interaction parameters,
free energy contribution due to the entropy of mixing in the are shown in Appendix A. Finally, since the lattice param-
0
respective phases, Gaxz ðT Þ and Gcxs ðT Þ are the excess free en- eters of the c and a0 phases are ac  3.56 Å and aa0  2.85 Å
ergy due to the interaction between the various elements respectively [7], the molar volumes of austenite and
0 0
present in SS with each other, and Gam ðT Þ and Gcm ðT Þ are martensite phases, Vc and Va were determined to be 5 
6 6 3 1
the magnetic ordering contributions by the various ele- 10 and 4.7  10 m mol , respectively.
ments to the respective phases. According to Dinsdale On this basis, DGv(T), given by Eq. (17), was determined
[41], the free energy of any ith phase as a function of tem- as a function of temperature. Not surprisingly, its value is
perature, is given by the power series expansion in terms of negative and decreases with temperature (Fig. 1a), which
temperature (T): implies that the a0 ? c transformation is favored as the
X annealing temperature is increased. Aaronson et al.
Gij ðT Þ ¼ A þ BT þ CT lnðT Þ þ DT n þ Gpres ð14Þ [43,44] have observed a similar trend for nucleation kinetics
in Fe–X alloys. To validate whether the DGv(T) values
where A, B, C and D are coefficients of the power series shown in Fig. 1a were realistic, we have calculated the typ-
expansion, n represents a set of integers 2, 3 and 1, and ical austenite nuclei radius formed during the a0 ? c phase
Gpres the pressure dependent contribution to the Gibbs free transformation according to [31]:
energy respectively. Dinsdale [41] has determined the mag-
nitudes for constants A, B, C and D for the bcc and fcc 2ra0 =c
rðT Þ ¼  ð18Þ
phases due to individual elements.
P We have ignored the DGv ðT Þ
higher order terms given by DTn because calculations re-
The results are shown in Fig. 1b. The austenite nuclei radii
veal that their contribution is negligible. Furthermore, we
for the temperature range 700–1000 °C is estimated to lie
have ignored Gpres because its contribution is negligible at
between 9 and 10.5 Å, which is typical for diffusion-driven
atmospheric pressure conditions under which the cold-
phase transformations [26,45].
rolling and annealing of the SS grades was performed.
Thus, Eq. (14) may be rewritten as follows:
2.4. Estimation of Z(T) and b(T)
Gij ðT Þ ¼ A þ BT þ CT lnðT Þ ð15Þ
Knowing ra0 /a 0 , ra 0 /c and DGv(T), we are now in a posi-
The free energy contributions from individual elements
tion to determine the Zeldovich factor Z(T) and the rate at
present in AISI 301LN SS for the fcc and bcc phase are
which atoms get added to the critical nucleus, b(T). As
reproduced from Dinsdale [41] in Appendix A. The excess
mentioned earlier in Eqs. (8a) and (8b), the geometric
free energy contribution, as described in Eq. (13), for the
factors Kj and Lj depend on the interfacial energies, and
martensite and austenite phases are given by [40]:
are given by [31]:
0
n1 X
X n
0
Gaxs ¼ xi xj Laij ðT Þ ð16aÞ 1 
Kj ¼ 3  3 cosðhÞ þ cos3 ðhÞ ð19aÞ
i¼1 j¼iþ1 2
n1 X
X n Lj ¼ 1  cosðhÞ ð19bÞ
Gcxs ¼ xi xj Lcij ðT Þ ð16bÞ
i¼1 j¼iþ1 r 0 0
where cosðhÞ ¼ 2ra =a0 and h is the dihedral angle for three
a =c
where xi and xj are the mole fractions of elements present in typical nuclei shapes (2-, 3- and 4-grain junctions with
0
a multi-component alloy, Laij ðT Þ and Lcij ðT Þ the interaction the parent martensite phase) shown in Fig. 2. As shown
742 S. Rajasekhara, P.J. Ferreira / Acta Materialia 59 (2011) 738–748

Fig. 1. (a) Calculated volume free energy change as a function of


annealing temperature in cold-rolled and annealed AISI 301LN SS grade
and (b) calculated austenite nuclei size as a function of annealing Fig. 2. Austenite nucleation at (a) 2-grain, (b) 3-grain and (c) 4-grain
temperature. austenite/martensite junctions (adapted from Clemm and Fisher [33]).

by Eq. (8a), Z(T) has a strong power dependence on


DGv(T) and Va 0 Fe, but a weak dependence on the type of 2.5. Estimation of the concentration of martensite nucleation
austenite/martensite nucleation sites (2-, 3- or 4-grain junc- sites, C1
tions) available. Since Va0 Fe is constant and DGv(T) changes
gradually with temperature (Fig. 1a), Z(T) is also expected As discussed in Eq. (9), the nucleation rate parameter
to change gradually with annealing temperature and with C gb
1 depends on the equilibrium nucleation sites C0, the
the shape of austenite nuclei. Not surprisingly, the Z(T) martensite grain boundary width d gb a0 and the average mar-
values calculated between 700 and 1000 °C for 2-, 3- and  0 . For AISI 301LN SS, C0 is determined
tensite grain size da
4-grain junction-type nucleation sites range between to be approximately 8:8  1025 atoms m3. The martensite
approximately 0.015 and 0.02, which are comparable to grain boundary width d gb a0 is assumed to be approximately
the values obtained by Johnson et al. [31]. On the other 5 Å [26] and the grain size of martensite grains da0 is deter-
hand b(T), given by Eq. (8b), exhibits a strong dependence mined by analyzing the transmission electron microscopy
on DGv(T) and D(T) but a weak dependence on the type of (TEM) images of heavily cold-rolled AISI 301LN SS
available nucleation sites. (Fig. 3). Image J software (National Institute of Health,
In this work, we assume nitrogen to be the interstitial ele- Bethesda, MD) was used on the scanned TEM negatives
ment driving the a0 ? c transformation. However, for com- of the cold-rolled sample to estimate the martensite grain
parison, we also perform calculations based on chromium size. Keeping in mind that the morphology of lath-type
(substitutional element with the highest concentration and martensite (comprising martensite platelets) is significantly
relatively large size) diffusion as the rate-limiting factor for different from dislocation-cell-type martensite (ultrafine
the a0 ? c transformation. With these assumptions, and martensite), area measurements were obtained from a total
knowing that the diffusivities of nitrogen DN(T) and of 200 grains comprising both lath-type and dislocation-
chromium DCr(T) (m2 s1) in the bcc phase of a commercial cell-type martensite. Subsequently, based on the average
SS grade are approximately 8  106 expð 0:82 kB T
Þ and area from 200 grains, the average martensite grain size
3 2:48
2:4  10 expð kB T Þ respectively [40], where k B ¼ 8:62 da0 was estimated to be approximately 0.17 lm. By employ-
105 eV =K, bN(T) and bCr(T) are determined to be approxi- ing the average area to determine da0 instead of directly
mately between 1010 and 1011 s1 and between approxi- measuring the size of individual martensite grains, we min-
mately 2.5  103 and 5  107 s1, respectively, for the imized any bias towards grains of a particular morphology
temperature range of 700–1000 °C. (lath-type or dislocation-cell-type). On this basis, the
S. Rajasekhara, P.J. Ferreira / Acta Materialia 59 (2011) 738–748 743

Fig. 4. Activation energy for heterogeneous austenite nucleation as a


function of annealing temperature for 2-, 3- and 4-grain austenite/
martensite junctions.

Table 2
Kinetic parameters for grain growth n and k at different annealing
temperatures [22].
Annealing Kinetic parameters
temperature (°C)
n k

Fig. 3. TEM bright-field image of cold-rolled AISI 301LN SS [18]. The 800 2.8 0.1
lath martensite size da0 was determined to be approximately 0.17 lm. 900 2.8 1.6
1000 2.4 3.6

available sites for austenite nucleation C gb


1 were determined
to be approximately 2.6  1023 m3.

2.6. Estimation of the activation energy for heterogeneous


nucleation, DG*(T)

The heterogeneous nucleation activation energy given


by Eq. (9) may now be determined. As shown in Eq.
(10), the magnitude of DG*(T) depends on the a0 /a0 and
a0 /c interfacial energies (determined in Section 2.2), the vol-
ume free energy change due to the a0 ? c phase transfor-
mation and the geometric parameters a, b and c. Clemm
and Fisher [33] have developed relationships that relate
these geometric parameters to the dihedral angle (h) for
the three typical nuclei shapes with 2-, 3- and 4-grain junc-
tions with the parent martensite phase (Fig. 2), which are
listed in Appendix B.
On the basis of previously calculated interface energies
(Section 2.2) and the volume free energy change, DG(T)
for different austenite nuclei shapes has been determined
as a function of annealing temperature (Fig. 4). Clearly,
the activation energy for 4- and 3-grain austenite/martens-
ite junctions is significantly lower than that for the 2-grain
austenite/martensite junction.

2.7. Determination of the kinetic parameters – k and n Fig. 5. (a) Exponential regression analysis to determine k value at 700 °C;
(b) linear regression analysis to determine n value at 700 °C.
Finally, to solve Eq. (11), the magnitudes of the grain
growth kinetic parameters k and n are also required. For information and solving the grain growth law for samples
the present work, the grain growth parameters have been annealed at 800, 900 and 1000 °C (Table 2) [20]. The grain
determined by using experimentally known grain size growth law was applied to only these samples because they
744 S. Rajasekhara, P.J. Ferreira / Acta Materialia 59 (2011) 738–748

are assumed to comprise of only austenite phase [20]. determine its values at 700 °C (Fig. 5a). Determining n val-
According to Table 2, k1000 °C > k900 °C > k800 °C, which is ues is less trivial because of challenges associated with
expected because k is directly proportional to the probabil- modeling grain growth parameter as a function of temper-
ity of atoms moving from one grain to the next [26]. We ature and material system. Burke and Turnbull [46], who
also note that n800 °C  n900 °C > n1000 °C. The higher n have evaluated grain growth in various metallic systems
value at 800 and 900 °C is likely due to residual strains (microalloyed carbon steels, aluminum, brass), have sug-
and the high texture from the parent martensite phase. gested that n values increase with decreasing temperature.
Additionally, these samples contain chromium nitride pre- Atkinson [47] reviewed that n values depend on the grain
cipitates [18], which inhibit grain growth and contribute to growth mechanism; specifically n-values for grain growth
a higher n value. On the other hand, annealing at higher in an impure system by lattice or grain boundary diffusion
temperatures relieves the residual strains, dissolves the pre- were 3 or 4, respectively. Vandemeer and Hu [48] tabulated
cipitates and increases the atomic migration along the grain that grain growth exponents in zone-refined Fe alloys to
boundaries, which promote the loss of grain curvature and range between 2 and 4. Gavard et al. [49] experimentally
texture; all of which lead to a decreased n value. Indeed, determined n values in ultra-pure stainless steels to range
samples annealed at 1000 °C exhibit a lower n value. between 2 and 6. In the present context, we have obtained
The samples annealed at 700 °C contain significant n-values from linear regression (Fig. 5b). This is a reason-
amounts of martensite, and therefore reliable k and n val- able first assumption because it empirically takes into
ues cannot be obtained by applying the grain growth law. account experimental observations of increasing n-values
Therefore, k and n values for samples annealed at 700 °C with decreasing annealing temperatures. Under these con-
have been obtained by applying regression analysis to the ditions, n and k values at 700 °C are found to be approxi-
k and n kinetic parameters at 800, 900 and 1000 °C. k is mately 3.1 and 0.007 respectively.
shown to have an exponential dependence on temperature Naturally, a possible error in calculating the austenite
[26], and thus an exponential regression is employed to phase fraction is introduced by using k and n values for

Fig. 6. Calculated (curves) and experimental (scatter points) austenite phase percentage as a function of annealing time and temperature. (a–c) show the
phase percentage when nitrogen-diffusion is assumed to be the rate-limiting step, whereas (d–f) show the phase percentage when chromium diffusion is
assumed to be the rate-limiting step.
S. Rajasekhara, P.J. Ferreira / Acta Materialia 59 (2011) 738–748 745

700 °C obtained from regression analysis. However,


detailed calculations (not shown here) reveal that changing
the values of k and n (obtained from regression analysis) by
as much as ±50% does not significantly influence the phase
fractions of austenite.

3. Results and discussion



The values of C gb1 , DG (T), Z(T), bCr(T) and bN(T) have
been used in Eq. (11) to determine the percentage of
austenite phase at the annealing temperatures of 700–
1000 °C for the rate-limiting cases of nitrogen and
chromium diffusion, when austenite nucleates at 2-, 3-
and 4-grain austenite/martensite junctions (Fig. 6). The
experimental data is also plotted for comparison.
Based on the calculations performed, when the a0 ? c
phase transformation is driven via the expected nitrogen-
diffusion, two distinct observations may be made. For the
annealing temperatures of 900 and 1000 °C, calculations
show that complete a0 ? c transformation occurs irrespec-
tive of the type of the a0 /c junction geometry, which agrees
with the available experimental data (Fig. 6a–c). However,
for the lower annealing temperatures of 700 and 800 °C,
the results show that the nature of the a0 /c geometry plays
an important role in the a0 ? c transformation kinetics.
The following observations can be made: (i) the phase
transformation is sluggish relative to the experimental data
when austenite nucleates at 2-grain junctions (Fig. 6a); (ii)
the a0 ? c phase transformation is faster and agrees
reasonably well with the experimental data when austenite
nucleates at 3-grain austenite/martensite junctions
(Fig. 6b); and finally (iii) for the case of austenite nucle-
ation at 4-grain martensite junctions, our calculations
predict the a0 ? c transformation to be considerably faster
than the experimental observations (Fig. 6c).
For comparison, calculations were also performed for
the case of chromium driven diffusion. In this case, the
phase transformation is considerably sluggish relative to
experimental data when austenite nucleates at 2- and 3- Fig. 7. (a) Incubation times preceding the a0 ? c transformation when
grain a0 /c junction geometries (Fig. 6d and e). On the other nitrogen and chromium diffusion are the rate-limiting factors. (b, c)
hand, the phase transformation kinetics is considerably fas- Austenite nucleation rates, normalized with respect to a maximum
ter relative to the experimental observations when austenite nucleation rate for a given temperature, as function of annealing time
and temperature assuming nitrogen and chromium diffusion, respectively,
nucleates at 4-grain junctions (Fig. 6f). Clearly, in all the
as the rate-limiting factors.
three cases just discussed, the calculated phase transforma-
tion data with chromium diffusion as the rate-limiting fac-
tor does not agree with the experimental observations. nitrogen-diffusion-controlled austenite nucleation at 3-
The results also show that the a0 ? c transformation, grain martensite junctions appears to agree reasonably well
regardless of the nature of the diffusing species, strongly with the experimental data (Fig. 6b).
depends on the geometry of available nucleation sites. In So far, in order to determine the phase concentration
particular, it is sluggish when austenite nucleates at 2-grain nc(t, T), we have assumed the austenite nucleation rate
martensite junctions and faster when it nucleates at 3- or 4- J(T) to be constant. However, if a time-dependent nucle-
grain martensite junctions (Fig. 6). This is expected because ation rate were assumed in our calculations, the incubation
the austenite phase fraction, given by Eq. (10), depends time in Eq. (6) would need to be determined. This can be
strongly on the activation energy for heterogeneous nucle- done by solving by the following expression [32]:
ation DG(T), which is large for 2-grain martensite junc-  
tions relative to that for 3- or 4-grain martensite 8k B T ra0 =c a4a0 K j
s¼ 2 ð20Þ
junctions (Fig. 2). In summary, our calculations show that V a0 DG2v Di xi Lj
746 S. Rajasekhara, P.J. Ferreira / Acta Materialia 59 (2011) 738–748

where Di and xi are the diffusivity and the mole fraction of Appendix A
the diffusing species, and all the terms retain their previous
definitions. The incubation times for the rate-limiting cases The free energy of pure elements that are used in this
of nitrogen and chromium diffusion as a function of work have been obtained from the work of Dinsdale [41]
annealing temperatures (700–1000°C) were calculated and Miettinen [42], and are shown below. Gibbs free ener-
(Fig. 7a). The incubation time with nitrogen-diffusion as gies of pure elements in the bcc phase:
the rate-limiting factor is approximately 5–6 ls, while it is
Gbcc o
C ðT Þ  GC ¼ 107; 350 þ 35:76T
much higher (several hundred milliseconds) with chromium
diffusion as the rate-limiting factor. This should not be sur- 1 o
Gbcc
N ðT Þ  GN 2 ¼ 31; 187 þ 59:8T
prising because of the inverse dependence of the incubation 2
time on the diffusivities (Eq. (20)). bcc
Gnickel ðT Þ ¼ 3535 þ 114T  22 lnðT Þ
Based on the incubation times for the above-mentioned
annealing conditions, the nucleation rates for austenite Gbcc
chromium ðT Þ ¼ 8856 þ 157T  27 lnðT Þ

nucleation at 3-grain austenite/martensite grain boundary Gbcc


manganese ðT Þ ¼ 8115 þ 130T  23 lnðT Þ
junctions were determined. For each annealing tempera-
ture and diffusing species scenario, the nucleation rate Gbcc
iron ðT Þ ¼ 1226 þ 124T  23 lnðT Þ
was normalized with respect to the maximum possible Gbcc
copper ðT Þ ¼ 3753 þ 129T  24 lnðT Þ
value for that case. The results are shown in Fig. 7b and
c. Clearly, the normalized nucleation rate is constant for Gbcc
molybdenum ðT Þ ¼ 7746 þ 132T  24 lnðT Þ
all annealing temperatures and times when nitrogen- Gbcc
silicon ðT Þ ¼ 38; 857 þ 114T  23 lnðT Þ
diffusion is the rate-limiting factor (Fig. 7b). Thus, our
assumption of constant nucleation rate for the case of Gibbs free energies of pure elements in the fcc phase:
nitrogen-diffusion is valid. Gfcc  GoC ¼ 155; 006 þ 13:7T
C ðT Þ
For the case of chromium diffusion, calculations indi-
cate that the normalized nucleation rate is not constant 1 o
Gfcc
N ðT Þ  GN 2 ¼ 23; 066 þ 274T  20:15T lnðT Þ
for the time of interest (1–100 s) at 700 °C annealing tem- 2
perature (Fig. 7c). At 1 s of annealing duration, the nor- Gfcc
nickel ðT Þ ¼ 5179 þ 117T  22 lnðT Þ
malized nucleation rate is approximately 80% of its
Gfcc
chromium ðT Þ ¼ 1573 þ 157T  27 lnðT Þ
maximum value. In the context of the analysis presented,
this means that we overestimated the austenite phase con- Gfcc
manganese ðT Þ ¼ 3429 þ 132T  25 lnðT Þ
centration at 700 °C. In other words, if a time-varying
Gfcc
iron ðT Þ ¼ 237 þ 132T  24 lnðT Þ
nucleation rate were employed to estimate austenite phase
concentration at 700 °C, the predicted nucleation kinetics Gfcc
copper ðT Þ ¼ 7770 þ 131T  24 lnðT Þ
would be even more sluggish than the one shown in
Fig. 6d. Therefore, for the case of chromium diffusion as Gfcc
molybdenum ðT Þ ¼ 7454 þ 133T  24 lnðT Þ

the rate-limiting factor, our assumption of constant nucle- Gbcc


silicon ðT Þ ¼ 42; 837 þ 115T  23 lnðT Þ
ation rate is justified. This result also confirms that the
a0 ? c transformation is driven by the fast diffusion of The interaction parameters the interaction parameters
interstitial nitrogen atoms. describing the interaction between the ith and the jth ele-
ments in the bcc and fcc phases have been obtained from
4. Conclusions Miettinen [42], and are shown below. First-order interac-
tion parameters in the bcc phase:
We have applied a generalized phase transformation
Lbcc
FeCr ¼ 20; 500  9:68T
kinetics model to understand the a0 ? c transformation
in cold-rolled and annealed metastable austenitic stainless Lbcc
FeMn ¼ 2759 þ 1:237T
steel, in order to achieve ultrafine-grained austenite. The Lbcc
FeMo ¼ 36; 818  9:141T þ ð362  5:724T ÞðxFe  xMo Þ
classical nucleation theory [30,31,33,43] explains reason-
ably well the diffusion-driven a0 ? c transformation in Lbcc
FeNi ¼ 957  1:287T þ ð1789  1:929T ÞðxFe  xNi Þ
heavily cold-rolled and annealed metastable austenitic SS. Lbcc
FeSi ¼ 153; 141 þ 46:48T þ ð92; 352ÞðxFe  xSi Þ
In particular, we found that nitrogen-diffusion is responsi-
ble for the rapid a0 ? c transformation, while the process þ ð62; 240ÞðxFe  xSi Þ2
of heavy cold-rolling provides heterogeneous nucleation Lbcc
FeCu ¼ 41; 003  6:022T
sites for the formation of austenite. In this regard, the
geometry of the austenite/martensite grain junctions is a Lbcc
CrMn ¼ 20; 328 þ 18:734T þ ð9162 þ 4:418T ÞðxCr
significant parameter for promoting fast kinetics. The  xMn Þ
results indicate that the a0 ? c transformation appears to
occur at austenite/martensite 3-grain junctions. Lbcc
CrMo ¼ 28; 890  7:962T þ ð5974  2:428T ÞðxCr  xMo Þ
S. Rajasekhara, P.J. Ferreira / Acta Materialia 59 (2011) 738–748 747

Lbcc
CrNi ¼ 17; 170  11:82T þ ð34; 418  11:858T ÞðxCr  xNi Þ Lfcc
FeN ¼ 92; 779  56:509T þ ð28; 554 þ 24:734T ÞðxFe  xN Þ
2
þ ð3805  5:203T ÞðxFe  xN Þ
Lbcc
CrSi ¼ 102; 085 þ 9:543T þ ð49; 125 þ 14:11T ÞðxCr  xSi Þ

Lbcc
CrCu ¼ 77; 107 Appendix B
Lbcc
NiMn ¼ 51; 785 þ 3:5T þ ð6600ÞðxNi  xMn Þ
The geometric parameters a, b and c for different types of
Lbcc
NiMo ¼ 46; 422 martensite grain junctions as a function of the dihedral angle
Lbcc
NiSi ¼ 100; 000
between the surface bounding the austenite nucleus [33].
ra0 =a0
Lbcc
NiCu ¼ 8366 þ 2:802T k ¼ cosðhÞ ¼
2ra0 =c
Lbcc
MnMo ¼ 7500
2-grain junction:
Lbcc
MnSi ¼ 89; 621 þ 2:941T þ ð7500ÞðxMn  xSi Þ
a ¼ pð1  k 2 Þ
Lbcc
MoSi ¼ 70; 900
b ¼ 4pð1  kÞ
Lbcc
FeC ¼ 119:04T þ ð43:89T ÞðxFe  xC Þ þ ð7:89T ÞðxFe  xC Þ2 2p
c¼ ð2  3k þ k 3 Þ
Lbcc 2 3
FeN ¼ 19:12T þ ð6:22T ÞðxFe  xN Þ þ ð1:02ÞðxFe  xN Þ
3-grain junction:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
First-order interaction parameters in the fcc phase:
a ¼ 3dð1  k 2 Þ  k 3  4k 2
p 
Lfcc
FeCr ¼ 10; 833  7:477T þ ð1410ÞðxFe  xCr Þ
b ¼ 12  v  kd
12
Lfcc qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

FeMn ¼ 7762 þ 3:865T þ ð259ÞðxFe  xMn Þ k2 2 2


c ¼ 2 p  2v þ ð3  4k Þ  dkð3  k Þ
Lfcc
FeMo ¼ 28; 347  17:691T
3
Lfcc
FeNi ¼ 12; 054  3:274T þ ð11; 082  4:45T ÞðxFe  xNi Þ
where
2
 
þ ð726ÞðxFe  xNi Þ 1
v ¼ sin1 pffiffiffiffiffiffiffiffiffiffiffiffiffi
2 1  k2
Lfcc 0 1
FeSi ¼ 12; 5248 þ 41:116T þ ð142; 708ÞðxFe  xSi Þ
2 B k C
þ ð89; 907ÞðxFe  xSi Þ d ¼ cos1 @qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiA
2
Lfcc 3ð1  k Þ
FeCu ¼ 53; 360  12:626T þ ð11; 512 þ 7:095T ÞðxFe  xCu Þ

4-grain junction:
Lfcc
CrMn ¼ 19; 088 þ 17:542T 8 20sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 39
< 2 2 =
Lfcc
CrMo ¼ 28; 890  7:962T þ ð5974  2:428T ÞðxCr  xMo Þ
K K
a ¼ 3 2/ð1  k 2 Þ  K 4@ 1  k 2  A  pffiffiffi5
Lfcc : 4 8 ;
CrNi ¼ 8090  12:88T þ ð33; 080  16:036T ÞðxCr  xNi Þ
p 
Lfcc
CrSi ¼ 102; 085 þ 9:543T þ ð49; 125 þ 14:11T ÞðxCr  xSi Þ b ¼ 24  k/  u
3
8 20sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 3
Lfcc
CrCu ¼ 53; 196  3:312T < p  2 2
K K
Lbcc c ¼ 2 4  u þ kK 4@ 1  k 2  A  pffiffiffi5
NiMn ¼ 51; 785 þ 3:5T þ ð6600ÞðxNi  xMn Þ : 3 4 8
Lfcc
NiMo ¼ 4804  5:96T þ ð10; 800ÞðxNi  xMo Þ )
Lfcc
NiSi ¼ 161; 150 þ 43:799T þ ð118; 251 þ 43:33T ÞðxNi  xSi Þ
 2k/ð3  k 2 Þ

Lfcc
NiCu ¼ 8366 þ 2:802T þ ð4360  1:812T ÞðxFe  xCu Þ where
Lfcc ¼ 22; 300 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

MnMo
4 3 2
Lfcc K¼  2k 2  k
MnSi ¼ 95; 600 þ 2:941T þ ð7500ÞðxMn  xSi Þ 3 2 3
Lfcc
MoSi ¼ 70; 900 K
/ ¼ sin1 pffiffiffiffiffiffiffiffiffiffiffiffi2ffi
Lfcc
FeC ¼ 162; 316  43:512T þ ð60; 803 þ 17:24T ÞðxFe  xC Þ 2 1k
pffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
þ ð10; 957  3:306T ÞðxFe  xC Þ2 1 2  k 3  K2
u ¼ cos pffiffiffiffiffiffiffiffiffiffiffiffiffi
K 1  k2
748 S. Rajasekhara, P.J. Ferreira / Acta Materialia 59 (2011) 738–748

References [24] Humphreys FJ, Hatherly M. Recrystallization and related annealing


phenomena. 2nd ed. Oxford: Elsevier; 2004. p. 333–9.
[1] Tomimura K, Takaki S, Tanimoto S, Tokunaga Y. ISIJ Int [25] Fullman P. Metal interfaces. Metals Park, OH: ASM; 1952. 179–205.
1991;31:721. [26] Porter DA, Easterling KE. Phase transformations in metals and
[2] Takaki S, Tomimura K, Ueda S. ISIJ Int 1994;34:522. alloys. 1st ed. London: Chapman & Hall; 1990.
[3] di Schino A, Salvatori I, Kenny JM. J Mater Sci 2002;37:4561. [27] Avrami M. J Chem Phys 1939;7:1103.
[4] di Schino A, Barteri M, Kenny JM. J Mater Sci Lett 2002;21:751. [28] Avrami M. J Chem Phys 1940;8:212.
[5] di Schino A, Barteri M, Kenny JM. J Mater Sci 2003;38:4725. [29] Johnson WA, Mehl RF. Trans Am Inst Metall Eng 1940;135:416.
[6] di Schino A, Kenny JM, Mecozzi MG, Barteri M. J Mater Sci [30] Chan KI, Lee JK, Shiflet GJ, Russell KC, Aaronson HI. Metall
2000;35:4803. Mater Trans 1978;9A:1016.
[7] Johannsen DL, Ferreira PJ, Kyröläinen A. Metall Mater Trans [31] Johnson WC, White CL, Marth PE, Ruf PK, Tuominen SM, Wade
2006;37A:2325. KD, et al. Metall Mater Trans 1975;6A:911.
[8] Rajasekhara S, Ferreira PJ, Karjalainen LP, Kyröläinen A. Metall [32] Huang W, Hillert M. Metall Mater Trans 1996;27A:480.
Mater Trans 2007;38A:1202. [33] Clemm PC, Fisher JC. Acta Metall 1955;3:70.
[9] Somani MC, Karjalainen LP, Kyröläinen A, Taulavuori T. Mater Sci [34] Bowkett MW, Keown SR, Harries DR. Met Sci 1982;16:499.
Forum 2007;539-543:4875. [35] Nishiyama Z. Martensitic transformations. New York: Academic
[10] Krauss Jr G, Cohen M. Trans Am Inst Metal Eng 1962;224:1212. Press; 1978.
[11] Breedis JF. Trans Am Inst Eng 1966;236:218. [36] Shibuta Y, Takamoto S, Suzuki T. Comput Mater Sci 2009;44:
[12] Jana S, Wayman CM. Trans Am Inst Eng 1967;239:1187. 1025.
[13] Kessler H, Pistch W. Acta Metall 1967;15:401. [37] Mangonon Jr PL, Thomas G. Metall Mater Trans 1970;1A:1587.
[14] Guy KB, Butler EP, West DRF. Met Sci 1983;17:167. [38] Gjostein NA, Domain HA, Aaronson HI, Eichen E. Acta Metall
[15] Montaneri R. Z für Metall 1990;81-H2:114. 1966;14:1637.
[16] Tomimura K, Takaki S, Tokunaga Y. ISIJ Int 1991;31:1431. [39] Chuang Y-Y, Chang YA. Metall Mater Trans 1987;18A:733.
[17] Ma Y, Jin J-E, Lee Y-K. Scripta Mater 2005;52:1311. [40] Miettinen J. Metall Mater Trans 1997;28B:281.
[18] Rajasekhara S, Karjalainen LP, Kyröläinen A, Ferreira PJ. Mater Sci [41] Dinsdale AT. Calphad 1991;15:317.
Eng A 2010;527:1986. [42] Miettinen J. Calphad 1998;22:275.
[19] Somani MC, Juntunen P, Karjalainen LP, Misra RDK, Kyröläinen [43] Aaronson HI, Kinsman KR, Russell KC. Scripta Metall 1970;4:101.
A. Metall Mater Trans 2009;40A:729. [44] Enomoto M, Aaronson HI. Metall Mater Trans 1986;17A:1385.
[20] Rajasekhara S, PhD dissertation, The University of Texas, Austin; [45] Cahn JW. Acta Metall 1957;5:169.
2007. [46] Burke JE, Turnbull D. Prog Met Phys 1952;3:220.
[21] Erukhimovitch V, Baram J. Phys Rev B 1994;50:5854. [47] Atkinson HV. Acta Metall 1988;36:469.
[22] Erukhimovitch V, Baram J. Phys Rev B 1995;51:6221. [48] Vandemeer RA, Hu H. Acta Mater 1994;42:3071.
[23] Hu H, Rath BB. Metall Mater Trans 1970;1A:3182. [49] Gavard L, Montheillet F, Coze JL. Scripta Mater 1998;39:1095.

You might also like