You are on page 1of 12

Materials Science & Engineering A 571 (2013) 1–12

Contents lists available at SciVerse ScienceDirect

Materials Science & Engineering A


journal homepage: www.elsevier.com/locate/msea

Hot deformation behavior and microstructure evolution of a stabilized


high-Cr ferritic stainless steel
S.V. Mehtonen n, L.P. Karjalainen, D.A. Porter
Center for Advanced Steels Research, University of Oulu, P.O. Box 4200, Oulu 90014, Finland

a r t i c l e i n f o abstract

Article history: The hot deformation behavior and static microstructure evolution of a 21Cr stabilized ferritic stainless
Received 5 December 2012 steel was studied using axisymmetric hot compression tests on a Gleeble 1500 thermomechanical
Received in revised form simulator. The deformation was carried out at 950–1050 1C to strains of 0.2 to 0.6 using strain rates of
24 January 2013
0.01, 0.1 and 1 s  1. The compression was followed by a holding period of 0 to 180 s in order to study
Accepted 25 January 2013
Available online 9 February 2013
the static recrystallization kinetics. The electron backscatter diffraction (EBSD) technique was used in
analyzing the resultant microstructures. A constitutive equation that well describes the flow stress as a
Keywords: function of strain, strain rate and temperature was developed. The active dynamic restoration
Ferritic stainless steel mechanism was found to depend on the Zener–Hollomon parameter, such that continuous dynamic
Flow stress
recrystallization was observed under low Zener–Hollomon parameter conditions but under high
Constitutive analysis
Zener–Hollomon parameter microstructures were dynamically recovered, and no dynamic formation
Restoration mechanisms
Recrystallization of new grains occurred. Static recrystallization resulted in little or no grain refinement, and further,
EBSD strain did not have an accelerating effect on the static recrystallization kinetics beyond the strain of 0.4.
& 2013 Elsevier B.V. All rights reserved.

1. Introduction recovery is very intense in these steels. Dynamic recrystallization


can occur by three different mechanisms: discontinuous dynamic
The high and volatile prices of nickel and molybdenum have recrystallization, continuous dynamic recrystallization and geo-
made the high chromium (Cr 418 wt%) ferritic stainless steels an metric dynamic recrystallization [4]. Discontinuous dynamic
attractive alternative to austenitic Cr–Ni and Cr–Ni–Mo and recrystallization operates by nucleation and growth and is gen-
ferritic Cr–Mo grades in general stainless steel applications such erally considered not to take place in metals with high stacking
as in building materials or kitchen appliances. These applications fault energy because of the intense recovery, even though it has
often require good deep drawability and resistance to roping or been shown that its occurrence is dependent on the purity of the
ridging. These properties can be improved by increasing the material, e.g. [5], and on the deformation conditions, i.e. the
amount of the favorable recrystallization texture in the cold Zener–Hollomon parameter [6]. In continuous dynamic recrystal-
rolled and annealed sheet and by breaking up colonies of similarly lization, no nucleation phase is present and the new grains form
oriented grains [1,2]. Hot rolling and hot band annealing affect by the gradual increase of the misorientation of low-angle
both of these. Because of their composition, these steels are boundaries. Geometric dynamic recrystallization operates at high
ferritic at all temperatures below the solidification temperature, strains 5–10, where the grains are fragmented into new grains
and the austenite–ferrite phase transformation does not occur during the deformation [7]. In ferritic stainless steels, due to
during hot deformation. The detrimental textures inherited from intense dynamic recovery, conventional discontinuous dynamic
the solidification structure can be weakened in the hot rolling recrystallization is usually considered as unfeasible. Instead,
stage only by recrystallization. continuous dynamic recrystallization by the coalescence of
Recovery and recrystallization are competing restoration pro- subgrain boundaries has been shown by a number of authors to
cesses which take place during hot deformation and during the occur under certain conditions during hot deformation of carbon
inter-pass times of hot rolling leading to the softening of the steels in the ferrite range and of ferritic stainless steels [8–13].
material. The stacking fault energy of ferritic stainless steels is The dynamic restoration mechanisms depend on the deforma-
relatively high and hence dislocations can cross-slip readily which tion conditions, as described above, and thereby the relations
promotes recovery over recrystallization [3]. Hence, dynamic between flow behavior and dynamic microstructure evolution are
very important factors affecting the hot rolling loads as well as the
following static recrystallization and texture evolution during hot
n
Corresponding author. Tel.: þ358 294 482147; fax: þ 358 8 5532165. band annealing. The dislocation storage, i.e. the amount of stored
E-mail address: saara.mehtonen@oulu.fi (S.V. Mehtonen). energy, serves as the driving force for static recrystallization during

0921-5093/$ - see front matter & 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.msea.2013.01.077
2 S.V. Mehtonen et al. / Materials Science & Engineering A 571 (2013) 1–12

Nomenclature R2 correlation coefficient


R Pearson0 s correlation coefficient
A1 phase transformation temperature for austenite– E experimentally achieved value
ferrite phase transformation E mean value of E
e strain P predicted value obtained from the equation
e_ strain rate ( s  1) P mean value of P
Q def activation energy for deformation (kJ mol  1) N number of the dataset
R universal gas constant (8.314 J mol  1 K  1) Xrex isothermally recrystallized fraction
T deformation temperature (K) t time (s)
A0 material constant t0.5 time for 50% recrystallization (s)
A00 material constant n Avrami exponent
A material constant C material constant
0
nc material constant d grain size
nc material constant Qapp apparent activation energy for static recrystallization
b material constant (kJ mol-1)
a material constant p material constant, ‘‘strain exponent’’
sp peak stress (MPa) q material constant, ‘‘strain rate exponent’’
sss steady state stress (MPa) s material constant, ‘‘grain size exponent’’
se true stress at true strain of e (MPa)

the post hot rolling annealing. The amount of stored energy of a were machined from the hot-rolled bands. The size of the speci-
bulk material can be measured quantitatively by thermal analysis mens was chosen so that the height to diameter ratio is within the
methods, such as differential scanning calorimetry (DSC) e.g. acceptable limits of 1–2 [24]. Axisymmetric compression tests
[14–16]. Electron backscatter diffraction (EBSD) and transmission were performed employing a Gleeble 1500 thermomechanical
electron microscopy (TEM) [16,17], microhardness e.g. [16,18] and simulator. The specimens were reheated at the rate of 20 1C/s to
the line broadening of X-ray diffraction (XRD) [19] are methods for 1100 1C for solution annealing, held for 2 min and then cooled to
qualitatively analyzing the dislocation density and thereby the the deformation temperature (950, 1000 or 1050 1C) at the cool-
local stored energy. Further, the amount of stored energy can be ing rate of 10 1C/s. The specimens were compressed in a single hit
estimated qualitatively from the amount of work hardening in the to a true strain of 0.2–0.6 at a constant true strain rate of 0.01,
flow stress curves [20]. The dynamic restoration processes greatly 0.1 or 1 s  1. After the compression, the specimens were either
affect the amount of stored energy and hence the static restoration water quenched to room temperature immediately after the
mechanisms and kinetics. The flow behavior and the microstruc- deformation or held at the deformation temperature for 10 to
ture evolution of austenite in static recrystallization process during 180 s, and subsequently air cooled to room temperature. The
hot deformation are extensively studied subjects. Also the defor- deformed and annealed microstructures were characterized using
mation of ferrite in interstitial free and low-carbon steels has been a field emission scanning electron microscope FEG-SEM (Carl
investigated by many authors in warm working temperature range, Zeiss Ultra Plus) with an electron backscatter diffraction detector
e.g. [21,22], but in these studies, the deformation took place below (EBSD) (HKL Nordlys). The recrystallized fractions were deter-
the austenite–ferrite phase transformation temperature A1. There- mined metallographically from the deformed and annealed speci-
fore, the restoration mechanisms may differ from those experi- mens using the SEM-EBSD. For metallographic examinations,
enced in the hot deformation temperature range of 950–1050 1C. cross-sections of the cylindrical compression specimens halved
Zhang et al. [23] investigated the effect of shear bands formed parallel to the cylinder axis were used. One specimen was
during hot rolling on static recrystallization rate in a 21%Cr ferritic quenched immediately after the 2 min of solution annealing at
stainless steel, however they did not pay attention to the flow 1100 1C in order to measure the initial grain size. This was found
behavior of the steel. to be 157 mm using the mean linear intercept method.
In this study, the hot deformation behavior of stabilized high-
Cr ferritic stainless steel was investigated by hot compression
tests with the final aim of clarifying the hot rolling schedules, 3. Results and discussion
which might improve the final product quality. The purpose was
to identify the dynamic restoration processes taking place during 3.1. Flow behavior during hot deformation
hot deformation and to develop constitutive equations to predict
the flow stress under various Zener–Hollomon parameter condi- The measured flow stress curves during compression tests are
tions. Also, the effect of Zener–Hollomon parameter on the static presented in Fig. 1. The shape of flow curves is similar to those
restoration kinetics was studied in order to explore ways to previously measured for high-Cr ferritic stainless steel [12]. (Note
promote recrystallization after the hot rolling stage. that the drop in the flow stress curves at the higher strain rates at
the strain of about 0.06 is caused by a machine effect and has no

2. Experimental
Table 1
The experimental steel used in the study was laboratory cast Chemical composition of the investigated
ingot of 70 kg hot rolled at 1100 1C from 48 mm down to the steel (in wt%).
thickness of 15 mm with 3 passes. The chemical composition (in
weight percent) of the steel is listed in Table 1. Cylindrical C N Cr Ti Nb

specimens 10 mm in diameter and 12 mm high (height to 0.004 0.0077 21 0.33 0.11


diameter ratio 1.2) cut with their axis normal to the rolling plane,
S.V. Mehtonen et al. / Materials Science & Engineering A 571 (2013) 1–12 3

Fig. 1. Flow stress curves at strain rates of 0.01–1 s  1 at (a) 950 1C, (b) 1000 1C and (c) 1050 1C.

metallurgical origin.) The flow stress increases with decreasing 3.2. Constitutive equations
deformation temperature and increasing strain rate. However, it
is seen that at 1000 and 1050 1C at low strain rates of 0.01 and Constitutive equations describing flow stress in terms of
0.1 s  1 there is hardly any work hardening, except at very small deformation temperature and strain rate can be expressed by
strains, indicating the occurrence of pronounced dynamic recov- the Zener–Hollomon parameter (Z), which is a temperature
ery (DRV). However, a real steady-state flow condition where the modified strain rate, defined by the equation [26]:
flow stress is constant is only reached at 1050 1C at the strain rate  
of 0.01 s  1, i.e. a low Zener–Hollomon parameter (Z) condition Q def
Z ¼ e_ exp ð1Þ
(see Eq. 1 below). At the higher strain rate of 1 s  1, the flow stress RT
increases with strain indicating dislocation storage in the micro-
where e_ is the strain rate, Q def the activation energy for deforma-
structure, particularly at 950 1C. This indicates that the deforma-
tion, R the gas constant (8.3145 kJ/mol), and T the deformation
tion mechanisms operating are different under low and high Z
temperature (K). Constitutive equations based on Z can be applied
conditions. The high Z condition is not sufficient for DRV to
not only on homogeneous materials but also on multi-phase
overcome the hardening produced by dislocation storage in the
materials such as functionally graded materials [27].
microstructure, which is observed as work hardening and conse-
The relationship between Z and the flow stress can be
quently as higher flow stress. Since the stored energy of deforma-
expressed by different experimental equations, such as 2–4; i.e.
tion consists of the energy of dislocations and point defects, such
the power function, the exponential function and the hyperbolic
as vacancies, created during the deformation [3], higher amount
sine function, respectively.
of work hardening leads to higher amount of stored energy. The
low Z condition is favorable for DRV, and therefore the flow stress  
Q def 0
curves can reach a steady state. Z ¼ e_ exp ¼ A0 snc ð2Þ
RT
In the instance of dynamic recrystallization (DRX), the flow
stress curves typically exhibit a single maximum, which is  
Q def
followed by a softening stage until eventually, a steady stage in Z ¼ e_ exp ¼ A00 expðbsÞ ð3Þ
RT
the flow stress is reached. This is also typical of continuous
dynamic recrystallization (CDRX) [25]. However, it can be seen  
from Fig. 1 that there are no such signs of DRX in the flow stress Q def
Z ¼ e_ exp ¼ A½sinhðasÞnc ð4Þ
curves. RT
4 S.V. Mehtonen et al. / Materials Science & Engineering A 571 (2013) 1–12

where A0 , A00 , A, nc, n0c, b and a are material constants. a is a stress Table 2
multiplier and an adjustable constant which brings as into the The calculated values for n0c, b, a, and nc at strains of 0.1–0.35.
right range, and can be described as
True strain Power law Exponential law Hyperbolic sine law
b
a ð5Þ nc0 b a nc
n0c

The power function (Eq. 2) is used for lower stress values than 0.1 5.504 0.122 0.022 3.959
0.15 5.070 0.115 0.023 3.770
the exponential function (Eq. 3), whereas the hyperbolic sine
0.2 4.934 0.109 0.022 3.673
function (Eq. 4) can be applied through the whole stress range [28]. 0.25 4.899 0.105 0.021 3.637
The peak stress sp , e.g. [29] or sometimes the steady state 0.3 4.828 0.101 0.021 3.587
stress sss , e.g. [30] is used in defining the constitutive equations 0.35 4.835 0.101 0.021 3.585
and for calculating the deformation activation energy, especially Average 5.012 0.109 0.022 3.702

for austenitic steels. However, in the current study, there is no


well-defined peak stress in the flow-stress curves nor is a steady
state flow stress reached under most deformation conditions, and
therefore, the constants for the constitutive equation were deter-
mined for various strain increments (0.1, 0.15, 0.2, 0.25, 0.3 and
0.35) by using se which is the measured true stress at each strain
increment, and an average value for Qdef was calculated.
Writing Eqs. 2–4 as follows,
 
Q def 1
lne_ þ ¼ lnA0 þ nc0 lnse ð6Þ
R T
 
Q def 1
lne_ þ ¼ lnA00 þ bse ð7Þ
R T
 
Q def 1  
lne_ þ ¼ lnA þnc ln sinhðase Þ ð8Þ
R T

shows that the values for n0c, b and nc can be obtained by plotting ln e_
vs. ln se , lne_ vs. se , and lne_ vs. ln sinhðase Þ , respectively. We have:
 
@lne_
n0c ¼ ð9Þ
@lnse T
Fig. 3. The plot used for calculating Qdef at 0.2 strain using the hyperbolic sine law.
 
@lne_
b¼ ð10Þ
@se T
  Table 3
@lne_ The calculated Qdef as a function of strain.
nc ¼ ð11Þ
@ln½sinhðase Þ T
Power law Qdef Exponential law Qdef Hyperbolic sine law Qdef
The plot used for determining nc at the strain of 0.2 are shown in Strain
[kJ/mol] [kJ/mol] [kJ/mol]
Fig. 2. s0:2 represents the measured true stress at the true strain of
0.2. An average value for a was calculated to be 0.022 using Eq. 5. The 0.1 417.956 390.824 392.076
0.15 387.461 382.116 385.750
values for n0c, b and nc at different strains are presented in Table 2.
0.2 393.318 387.564 390.687
The activation energy for deformation can be calculated using 0.25 405.465 392.333 399.982
plots (in Fig. 3 for hyperbolic sine equation) together with the 0.3 407.447 390.628 400.826
Eqs. 12–13 for the power equation, exponential equation, and the 0.35 418.931 404.959 412.287
Average 405 391 397

hyperbolic sine equation as follows:


" #
@lnse
Q def ¼ Rn0c  ð12Þ
@ 1=T _
e
 
@se
Q def ¼ Rb ð13Þ
@ð1=TÞ e_
 
@ln sinhðase Þ
Q def ¼ Rnc ð14Þ
@ð1=TÞ e_
The calculated values for Qdef at various strains are presented in
Table 3. Some variation in the calculated values of Qdef at different
strains is apparent. The average deformation activation energy was
calculated to be 405, 391 or 397 kJ/mol by using the power
equation, exponential equation or the hyperbolic sine equation,
respectively. Comparing the plots in Figs. 4–6 shows that the
Fig. 2. Evaluating the value of nc at 0.2 strain by plotting lnðe_ Þ vs. ln½sinhðas0:2 Þ. hyperbolic sine equation together with a Qdef of 397 kJ/mol gives
S.V. Mehtonen et al. / Materials Science & Engineering A 571 (2013) 1–12 5

previously determined values of 385 kJ/mol for a high purity 17%Cr


ferritic stainless steel [31], 375 kJ/mol for AISI 430 (0.05% C, 16.65%
Cr) in the dual phase region [32] or 372 kJ/mol for Ti containing
interstitial free (IF) steel deformed in the ferrite region [21].
However, some different values have also been reported previously;
for instance 83.54 kJ/mol for AISI 430 (0.018% C, 16.5% Cr) [8],
280 kJ/mol for ferritic iron [33] and 315 kJ/mol for 17%Cr ferritic
stainless steel [34].
The self-diffusion activation energy in ferrite in the pure
ferritic iron is 239 kJ/mol [35]. However, the activation energies
for deformation in hot working can be much higher than that
of self-diffusion. In general, Qdef is a function of alloying,
being higher in highly alloyed steels, as is the case with the
present alloy.

3.3. Modification of the constitutive equation

Fig. 4. The plot of lnZ vs. ln s (power law). During hot deformation, the strain rate is maintained constant
while the stress value needed to achieve this is being measured.
At each value of strain, flow curves are in different stages of
deformation and may feel different deformation or softening
mechanisms. The variation of flow stress as a function of strain
was incorporated into the constitutive equations by Slooff et al.
[36] by adding a strain-dependent parameter. The method has
been later used successfully by many authors, e.g. [37–40]. Fig. 7
shows the parameters of the constitutive equations as a function
of strain and some variation can be seen. In the current study a
linear fit (Eq. 15) was imposed on the parameters of the
constitutive equation (Fig. 7).
a ¼ 0:02330:0073e
nc ¼ 4:12261:4414e
Q def ¼ 376:93 þ 88:903e
ln A ¼ 32:823 þ 8:0137e ð15Þ

Fig. 5. The plot of lnZ vs. s (exponential law).

Fig. 6. Plot of lnZ vs. ln½sinhðasÞ (hyperbolic sine law with a ¼0.022).

the best fit, i.e. the highest value for the correlation coefficient R2
(Fig. 6).
The values for Qdef for ferrite given in the literature are quite
scattered. The present value of 397 kJ/mol is in line with the Fig. 7. Variation of Qdef, ln A, nc and a as a function of strain.
6 S.V. Mehtonen et al. / Materials Science & Engineering A 571 (2013) 1–12

Fig. 8. Comparison between the measured and predicted flow stress considering the strain compensation in the constitutive equation at (a) 950 1C, (b) 1000 1C and
(c) 1050 1C.

The comparison between the measured flow stress curves and


the predicted flow stress values using the strain modified hyper-
bolic sine equation are presented in Fig. 8. The predicted values
are very close to the experimental values except at 950 1C and
1000 1C at the highest strain rate of 1 s  1. At 950 1C, the predic-
tion for 1 s  1 is higher than the actual values whereas at 1000 1C
the predicted values are lower than the measured experimental
flow stress.
Pearson0 s correlation coefficient R (Eq. 16) was used to esti-
mate the success of the prediction.
PN
i ¼ 1 ðEi EÞðP i PÞ
R ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
PN
ffi ð16Þ
2 PN 2
i ¼ 1 ðEi EÞ i ¼ 1 ðP i PÞ

where E is the experimentally achieved value and P is the


predicted value obtained from the constitutive equation, and E
and P represent the mean values of E and P, respectively. N is the
number of the dataset. It can be seen in Fig. 9 that the correlation
between the experimental and predicted flow stress values is Fig. 9. Comparison between the experimental and predicted flow stress values.
good with a Pearson0 s correlation coefficient value of 0.995.
specimens, small grains were detected, which were apparently
3.4. Dynamic microstructure evolution formed dynamically during the deformation.
The amount of substructure and hence the amount of stored
The dynamically formed microstructures were examined using energy depended on the deformation conditions being highest
the SEM-EBSD method, and their evolution at 950–1050 1C at the under the high Z deformation conditions. High deformation
strain rates of 0.01–1 s  1 is presented in Fig. 10. Generally, the temperatures and low strain rates (low Z) allow more time for
deformed microstructures consisted of dynamically recovered dislocation climb and cross-slip and hence DRV is very efficient.
grains with low-angle grain boundaries. However, in some The dislocations are able to migrate in order to form lower energy
S.V. Mehtonen et al. / Materials Science & Engineering A 571 (2013) 1–12 7

Fig. 10. Dynamic microstructure evolution at 950–1050 1C at strain rates of 0.01–1 s  1 (0.4 strain). The white and yellow lines represent small angle grain boundaries with
misorientation of 21–51 and 51–151, respectively. The black colored boundaries have the misorientation greater than 151. New dynamically formed grains at 1000 1C are
pointed out by yellow arrows. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

configurations such as low-angle grain boundaries, i.e. polygoni-


zation occurs. Also, dislocations annihilate and the total disloca-
tion density is decreased. Subgrain coarsening, which was
observed especially at 1050 1C at 0.01 and 0.1 s  1, lowers further
the amount of stored energy. In addition to the differences of the
subgrain structure as a function of the deformation conditions,
the amount of substructure varied between the grains.
The average intercept length between boundaries with 21
misorientation in the normal direction was determined from the
EBSD data, and it is presented as a function of Z for the strain of
0.4 in Fig. 11. This intercept length gives an estimation of the
subgrain size. It is clear that for this strain the subgrain size is a
function of Z. The intercept length and therefore also the subgrain
size increase with increasing deformation temperature and
decreasing strain rate, i.e. decreasing Z. Under low Z deformation
conditions, DRV is extensive because dislocations can climb more
and low-angle grain boundaries can move further than under the
high Z conditions [4].
After 0.4 strain at 950 1C, a distinct subgrain structure has not Fig. 11. The average intercept length between boundaries with 21 misorientation
yet formed at the strain rate of 1 s  1, as shown in Fig. 12. in the normal direction as a function of the Zener–Hollomon parameter at
Dislocations and subgrain boundaries are clustered in tangles in 0.4 strain.
the vicinity of the original high-angle grain boundaries, and it is
evident that there is less substructure towards the grain interiors. be detected (arrowed in Fig. 10). The grains are surrounded by
This is probably a result from multislip, which especially takes high-angle grain boundaries for the most part, i.e. they have a
place close to grain boundaries in order to accommodate the different orientation compared to the neighboring original grains.
differing deformation behavior of the neighboring grains. It seems It is typical for newly formed grains by CDRX that they are not
that subgrain formation begins close to the original grain bound- completely surrounded by a high-angle grain boundary in the
aries where the amount of deformation is highest and hence the early stages of CDRX, since the CDRX involves the gradual
formation of subgrain boundaries easiest. After 0.4 strain at transformation of low-angle grain boundaries into high-angle
0.01 s  1, the subgrains are clearly formed throughout the micro- grain boundaries [4].
structure (Fig. 12). At the original grain boundaries, grains, which Like with deformation at 950 1C, at the strain rate of 1 s  1 at
seemed to have formed dynamically during the deformation, can 1000 1C, a clear subgrain structure was not formed. However,
8 S.V. Mehtonen et al. / Materials Science & Engineering A 571 (2013) 1–12

Fig. 12. Dynamic microstructure evolution at 950 1C at (a) 1 s  1 and (b) 0.01 s  1 (0.4 strain). The white and yellow lines represent low-angle grain boundaries with
misorientation of 21–51 and 51–151, respectively. The black colored boundaries have the misorientation greater than 151. (For interpretation of the references to color in
this figure legend, the reader is referred to the web version of this article.)

normal direction between 21 subgrain boundaries was 51 mm at


0.01 s  1, 27 mm at 0.1 s  1 strain rate and 17 mm at 1 s  1 strain
rate. The subgrains were also larger than after deformation at
950 1C (22, 14 and 13 mm at strain rates of 0.01 s  1, 0.1 s  1 and
1 s  1, respectively) or at 1000 1C (33, 20 and 10 mm at strain rates
of 0.01 s  1, 0.1 s  1 and 1 s  1, respectively). At 0.01 s  1 strain rate
at 1050 1C, the subgrain size is significantly larger than at any
other deformation conditions—this can be seen clearly in Figs. 10
and 11.

3.5. The mechanisms for dynamic microstructure evolution

Out of the three dynamic recrystallization mechanisms


(discontinuous dynamic recrystallization (DDRX), geometric
dynamic recrystallization (GDRX) and continuous dynamic
recrystallization (CDRX)), the ‘‘conventional’’ discontinuous
Fig. 13. Dynamically formed microstructure at 1000 1C at 1 s  1 after 0.6 strain
dynamic recrystallization, DDRX, occurs by the nucleation and
deformation. The white and yellow lines represent low-angle grain boundaries
with misorientation of 21–51 and 51–151, respectively. The black colored bound- growth of new grains. However, since the stacking fault energy of
aries have the misorientation greater than 151. Dynamically formed grains, which ferritic stainless steel is high and dislocations can cross-slip and
are either completely or partly surrounded by a high-angle boundary, are pointed climb easily, it is considered unlikely that DDRX would take place
out by yellow arrows. (For interpretation of the references to color in this figure during hot deformation. The second mechanisms, GDRX, would
legend, the reader is referred to the web version of this article.)
cause the original grains to split into new grains as a result of very
high strain deformation [41]. Because the strains used in this
subgrains have started to form in the vicinity of the original grain study were relatively low (0.2–0.6), GDRX could not have caused
boundaries, as seen in Fig. 10. At strains of 0.4 and 0.6, a few the formation of the new small grains.
relatively small grains ( 10–30 mm) surrounded by a high-angle CDRX includes a gradual change of the microstructure from
grain boundary were detected at the original grain boundaries deformed and recovered state into recrystallized structure. In the
and also inside the original grains (Fig. 13). The amount of first stages of CDRX the dislocations are created during the
deformation and substructure is highest close to the original deformation and they form low-angle grain boundaries because
grain boundaries because multislip in these regions generates of DRV. When the deformation continues, the gradual increase in
more dislocations enabling the low-angle grain boundaries close the misorientation of low-angle grain boundaries leads eventually
to the original grain boundaries to increase their misorientation. to the formation of high-angle grain boundaries, which are able to
Deformation to the strain of 0.2 (1000 1C, 1 s  1) was not sufficient migrate and cause the annihilation of dislocations within the
to produce dynamically formed new grains. newly formed grains [25]. The described mechanism for the
After deformation at 0.1 s  1 to 0.4 strain at 1000 1C, a clear formation of new grains is in line with the microstructural
subgrain structure was formed and the average intercept length observations of the current study.
between grain boundaries having misorientation of at least 21 Gourdet and Montheillet [4] concluded that the main features
was 20 mm. The lower strain rate of 0.01 s  1 resulted in a longer for CDRX are the decrease of the subgrain size and increase in the
average intercept length of 33 mm. The measured differences in misorientation of the subgrain boundaries during the transient.
the substructures are also visible in the EBSD maps (Fig. 10). The misorientation distribution of grain boundaries at strains of
At 0.01 s  1, new dynamically formed grains (30–70 mm) sur- 0.2, 0.4 and 0.6 is presented in Fig. 14. It can be seen that the
rounded by high-angle grain boundaries with no evident subgrain relative fraction of low-angle grain boundaries decreases with
structure had formed adjacent to the original grain boundaries, as increasing strain. Also, at 0.6 strain, the fraction of high-angle
can be seen in Fig. 10. grain boundaries is higher than at lower strains. This evolution of
At 1050 1C, similarly to lower deformation temperatures, the the misorientation distribution as a function of strain is a typical
amount of substructure is greater closer to the grain boundary feature for CDRX. Further, it was noted that the lower the strain
than towards the inside of the grain, and deformation to the strain rate is, the more subgrain boundaries have increased their
of 0.4 at 1 s  1 was not sufficient to initiate the dynamic formation misorientation, as seen in Fig. 10, and for example at 1050 1C at
of new grains. The strain rate had a pronounced effect on the size a strain rate of 0.01 s  1, only few low-angle grain boundaries are
of subgrains at 1050 1C; the average intercept length in the present.
S.V. Mehtonen et al. / Materials Science & Engineering A 571 (2013) 1–12 9

The dynamic formation of new ‘‘grains’’ under low Z deformation


conditions, as described above, is a result of the gradual increase
in the misorientation of low-angle grain boundaries leading
eventually to the formation of high-angle grain boundaries, and
is known as CDRX [4,25]. The operating dynamic restoration
mechanism depends on the deformation conditions. Under low
Z conditions, CDRX causes the formation of new grains but under
high Z conditions and high stresses, DRV is the main restoration
mechanism. The observed behavior is similar to that previously
observed in a-iron [33] or in 26Cr ferritic stainless steel [12].
Hence, even the high alloying by Cr does not change the
mechanism for dynamic microstructure evolution.

3.6. Static recrystallization kinetics

The isothermally recrystallized fraction, Xrex, as a function of


time, t, can be expressed by an Avrami-type equation [42]:
  n 
Fig. 14. The dependence of grain boundary misorientation distribution on strain t
at 1000 1C at 1 s  1. X rex ¼ 1exp 0:693 ð17Þ
t 0:5

where t 0:5 is the time for 50% recrystallization and n is known as


the Avrami exponent.
t 0:5 depends on deformation conditions and the composition as
follows:
 
s Q app
t 0:5 ¼ C ep e_ q d exp ð18Þ
RT

where C is a constant dependent on chemical composition, e is the


true strain, e_ is the strain rate, d grain size, and Q app is the
apparent activation energy for the static recrystallization.
The exponents p, q and s are material dependent constants
describing the powers of strain, strain rate and the grain size,
respectively [43]. They have been determined very extensively for
microalloyed steels by Somani and Karjalainen [44].
The data for the fraction of SRX can be fitted reasonably well
with a sigmoidal shape, as seen in Fig. 16. Values of the Avrami
exponent n and the experimental t 0:5 times are presented in

Fig. 15. Intercept length between 21 boundaries in the normal direction as a


function of strain at 1 s  1.

When CDRX occurs, the size of the dynamically formed grains


decreases strongly up to a strain of  5 after which it decreases
slowly until a steady state is reached at very high strains ( 30)
[4]. In Fig. 15 is presented the average intercept length between
21 boundaries in the normal direction as a function of strain. It can
be seen that at 1050 and 950 1C the average intercept length, and
hence also the subgrain size, decreased with increasing strain as is
supposed to according to the CDRX theory. However, at 1000 1C a
slight increase was observed. In all cases, however, the differences
were relatively small. Because of the relatively low strains (0.2–
0.6) used in the present experiments, dynamically formed micro-
structures were mostly in recovered state while some boundaries
had increased their misorientation. A small number of dynami-
cally formed grains or crystallites were found in most of the Fig. 16. EBSD data for SRX fractions (strain 0.4, strain rate 1 s  1) and Avrami fits.
specimens, however they were not in all cases fully surrounded
by a high-angle grain boundary, which indicates that the CDRX
transformation and formation of new grains was incomplete.
Table 4
Further, the flow stress curves did not exhibit a softening stage Values of n and t 0:5 in the Avrami equation.
leading to a steady stage (Fig. 1), which is generally a common
feature for CDRX, however this does not mean that CDRX did 950 1C 1000 1C 1050 1C
not occur.
n 1.5 2.2 2.2
The effect of CDRX on the subgrain size, misorientation t 0:5 [s] 118 34 21
distribution and flow stress in the current study was very slight.
10 S.V. Mehtonen et al. / Materials Science & Engineering A 571 (2013) 1–12

Table 4. The Avrami exponent provides information about the stainless steel, in which n depended on the fraction recrystallized:
nucleation mechanism in the recrystallization event. It is gener- when the recrystallized fraction was below 70%, n was 2.85,
ally independent of temperature, but here some variation of n is whereas for recrystallized fraction exceeding 70%, n was 1.03. The
seen in the Table 4, presumably reflecting the contribution of limit of 70%, where the recrystallization rate slowed down, is
recovery in softening by reducing the driving force for recrystal- substantially lower than the 95% observed here and e.g. by
lization. Particularly at 950 1C, n is smaller than at higher Sinclair et al. [46]. However, it was concluded that the banded
temperatures, although values as low as 0.9 have been reported austenite present in the structure of the AISI430 was a reason to
[45] for dynamically recovered a-iron. However, a typical value retard the SRX kinetics [32].
for n in dynamically recovered material would be  2 [45]. Complex recrystallization kinetics of ferrite in an IF steel was
The dependence of t 0:5 on the deformation temperature is also observed by Akbari et al. [47]. In fact, in their study, SRX
presented in Fig. 17. The value for the apparent activation energy ceased after reaching a certain recrystallized fraction. The critical
for recrystallization Qapp determined from this data is 221 kJ/mol, fraction was found to be dependent on the deformation tempera-
which is slightly lower than the 280 kJ/mol reported for a-iron [33]. ture, varying between 500 and 800 1C after 50% of deformation:
The last 5% fraction of the grains seemed to recrystallize more the lower the deformation temperature the higher the recrystal-
slowly than expected from the Avrami fits (see Fig. 16). This lized fraction achieved. The authors concluded that this behavior
‘‘sluggish’’ recrystallization in ferritic stainless steels has also was caused by precipitation interacting with the recrystallization.
been noticed by Sinclair et al. [46], who concluded that it was Precipitation might have retarded the recrystallization kinetics
caused by heterogeneity in the microstructure and texture, which also in the present study, since the investigated steel is highly
was ultimately a result of two different types of recovered grains. alloyed with precipitate forming elements. Because of the stabi-
Also small precipitates were found to play a major role by lization, at least TiN and NbC precipitates are present in the
affecting the movement of grain boundaries through Zener drag microstructure. Further, other precipitates, such as chromium
and ultimately cause the ‘‘sluggish’’ recrystallization behavior carbides, might have formed during the deformation and holding
towards the end of the recrystallization. The first type of recov- periods and thereby affected the recrystallization kinetics.
ered grain had a heterogeneous subgrain structure, but the grain The recrystallization rate is generally strongly dependent on
type was prone to abnormal subgrain growth. The abnormally the degree of deformation. Here, the effect of strain is shown in
large subgrains were able to act as nuclei for static recrystalliza- Fig. 18 revealing that the recrystallization rate is faster (t 0:5 ¼ 34 s)
tion (SRX). The second grain type had generally a larger subgrain at 0.4 strain compared to that at 0.2 strain (t 0:5 ¼ 109 s). However,
size and lower misorientation spread within the grain but the there is no accelerating effect between strains 0.4 and 0.6,
substructure was not able to form large subgrains suitable for indicating that no more deformation is stored beyond 0.4 strain
recrystallization nuclei. This type of mechanisms could explain at 1000 1C, which is in line with the observation of Akbari et al.
the slowing down of the recrystallization rate also observed here, [47] that in the ferrite region of an IF steel the strain had an
since the heterogeneity of substructure between grains was accelerating effect on the SRX kinetics but only to a certain point.
present in the current study as well (see Fig. 12). The amount of Glover and Sellars noted [45] that in a-iron, the SRX rate is
substructure and the size and location of subgrains varied independent of strain in the ‘‘steady-state’’ region of the flow
between the grains, which could be due to the different orienta- curve. However, at low strains, where work hardening occurs, the
tion of the grains regarding the direction of deformation. Because accelerating effect is very strong. Further, it has been proposed
the test specimens used in the present work were axisymmetric that the strain exponent is constant at low strain values, but
and hence no rolling direction existed in the deformation, the decreases when the strain approaches the critical strain for the
effect of grain orientation and texture on the deformation and initiation of DRX [48]. The lack of an accelerating effect of strain
restoration mechanisms could not be verified. However, it is likely could be because strong DRV prevents dislocation storage in the
that grain orientations play a significant role in explaining the microstructure such that increasing strain in the ‘‘steady-state’’
different substructures observed. Also it is known that the region does not increase the driving force for recrystallization.
CDRX kinetics is strongly dependent on the crystallographic The power of strain, p in Eq. (18), on t 0:5 between the strains of
orientation [4]. Hinton and Beynon [32] proposed the possibility 0.2 and 0.4 was calculated to be  1.7, which is significantly
of two different recrystallization kinetics for AISI 430 ferritic smaller than values previously proposed for austenitic steels;
 3.81 to  3.55 [49], 4 [48] and 2.2 [50].

3.7. Static microstructure evolution

Typical examples of the partially recrystallized microstruc-


tures are illustrated in Fig. 18, where the effect of strain on the
deformed and annealed microstructures at 1000 1C is shown
together with the recrystallization rate. The recrystallized grain
size depended on the deformation conditions but it was always
quite coarse ( 140–400 mm), i.e. roughly equal to or even larger
than the initial grain size (157 mm). As seen in Fig. 18, the
recrystallized grains are larger in the 0.2 strain specimen after
120 s of soaking than in the 0.4 strain specimen after 180 s of
soaking. In the first instance, some of the recrystallized grains
were several hundred microns in diameter, whereas after
0.6 strain at 1000 1C and complete SRX, the recrystallized grain
size was  140 mm. Also, in the partially recrystallized structures,
the number of recrystallized grains was higher in the 0.6 strain
specimens compared to the 0.2 specimens. Overall though, the
recrystallization resulted in only little or no grain refinement
Fig. 17. Determination of the apparent activation energy for the SRX. under the deformation conditions used.
S.V. Mehtonen et al. / Materials Science & Engineering A 571 (2013) 1–12 11

Fig. 18. The effect of strain on the SRX rate at 1 s  1 at 1000 1C, and the corresponding SEM-EBSD maps of the partly recrystallized microstructures: (a) 0.6 strain, 10 s
holding, (b) 0.6 strain, 30 s holding, (c) 0.4 strain, 30 s holding, (d) 0.4 strain, 180 s holding and (e) 0.2 strain, 120 s holding.

The recrystallized grains appeared first at the original grain in flow stress curves because of extensive dynamic recovery
boundaries, which is the site where the dynamically formed new and continuous dynamic recrystallization. At high Zener–
grains were mainly detected (Figs. 10 and 13). Further, in order Hollomon parameter, work hardening continued up to max-
for the nucleus to be able to grow, it must have a high-angle grain imum strains of 0.4–0.6 used.
boundary, which in many cases was formed dynamically during 2. Under low Zener–Hollomon parameter deformation condi-
the deformation. Hence, it is likely that the dynamically formed tions, the formation of new grains by continuous dynamic
small grains grew statically during the holding period. recrystallization was detected but without a drop in the flow
Increasing the degree of deformation beyond the strain of stress level, presumably due to the small fraction of dynami-
0.4 is not likely to promote SRX kinetics because of the intense cally formed grains in the microstructure.
DRV as shown above. In order to promote SRX after hot deforma- 3. Under high Zener–Hollomon parameter deformation condi-
tion, the amount of stored energy in the microstructure should be tions, microstructures were dynamically recovered, but no
increased so that the driving force for SRX would be as high as dynamic formation of new grains occurred.
possible. This can be achieved by using deformation conditions 4. The parameters of the constitutive equation for deformation
not favorable for DRV, i.e. low deformation temperature and high varied as a function of strain and therefore the effect of strain
strain rate followed by annealing at a higher temperature. was incorporated in the constitutive equation. The strain
Increasing the degree of deformation would, however, result in modified constitutive equation was successfully used for pre-
smaller recrystallized grain size as was the case after 0.6 strain: dicting the flow stress.
the SRX rate was not accelerated from that of after 0.4 strain but 5. The average activation energy for deformation, Qdef, was
the higher strain resulted in smaller recrystallized grain size. determined to be 397 kJ/mol and the apparent activation
energy for static recrystallization, Qapp, 221 kJ/mol.
6. The rate of static recrystallization was increased by increasing
4. Summary strain up to 0.4, but beyond this value the rate of recrystalliza-
tion remained unchanged. Increasing the strain to 0.6 did,
The hot deformation behavior at different temperatures (950– however, lead to finer recrystallized grain size when hardly
1050 1C) and strain rates (0.01–1 s  1) of a stabilized 21Cr ferritic any refinement of grain size was to be achieved after straining
stainless steel has been studied. Because of the composition, the to 0.2 or 0.4.
investigated steel is in ferritic state in the hot working tempera- 7. In order to promote static recrystallization after hot deforma-
ture range. The results can be summarized as follows: tion, the stored energy in the microstructure, i.e. the driving
force for static recrystallization, should be increased. This can
be achieved by using deformation conditions, which are
1. Under low Zener–Hollomon parameter deformation condi- unfavorable for dynamic recovery, i.e. low deformation
tions, no work hardening was seen above the strain of  0.05 temperature and high strain rate. Carrying out the subsequent
12 S.V. Mehtonen et al. / Materials Science & Engineering A 571 (2013) 1–12

annealing at a higher temperature than the deformation would [19] P. Gay, P.B. Hirsch, A. Kelly, Acta Metall. 1 (1953) 315–319.
further promote static recrystallization. [20] J. Hodowany, G. Ravichandran, A.J. Rosakis, P. Rosakis, Exp. Mech. 40 (2000)
113–123.
[21] A. Oudin, M.R. Barnett, P.D. Hodgson, Mater. Sci. Eng. A 367 (2004) 282–294.
[22] C. Huang, E.B. Hawbolt, X. Chen, T.R. Meadowcroft, D.K. Matlock, Acta Mater.
Acknowledgments 49 (2001) 1445–1452.
[23] C. Zhang, Z. Liu, G. Wang, J. Mater, Process. Tech. 211 (2011) 1051–1059.
[24] B. Roebuck, J.D. Lord, M. Brooks, M.S. Loveday, C.M Sellars, R.W. Evans, Mater.
The financial support from the Finnish Funding Agency for
High Temp. 23 (2006) 59–83.
Technology and Innovation (Tekes) in Project CSP1 of the [25] S. Gourdet, F. Montheillet, Acta Mater. 51 (2003) 2685–2699.
Demanding Applications program of the Finnish Metals and [26] C. Zener, J.H. Hollomon, J. Appl. Phys. 15 (1944) 22–32.
Engineering Competence Cluster (FIMECC Ltd.) is gratefully [27] M. Abolghasemzadeh, H.S.S. Pour, F. Berto, Y. Alizadeh, Mater. Sci. Eng. A 534
(2012) 329–338.
acknowledged. The authors would also like to thank Outokumpu [28] C.M. Sellars, W.J. McTegart, Acta Metall. 14 (1966) 1136–1138.
Oyj for providing experimental materials and for supporting the [29] H.J. McQueen, N.D. Ryan, Mater. Sci. Eng. A 322 (2002) 43–63.
research. S.M. also express her gratitude for the support provided [30] A. Belyakov, R. Kaibyshev, T. Sakai, Metall. Mater. Trans. A 29 (1998)
161–167.
by the Academy of Finland through the Graduate School on
[31] F. Gao, Z.-Y. Liu, G.-D. Wang, Dongbei Daxue Xuebao, J. Northeast. Univ. 32
Advanced Materials and Processes. (2011) 1406–1409.
[32] J.S. Hinton, J.H. Beynon, ISIJ Int. 47 (2007) 1465–1474.
[33] G. Glover, C.M. Sellars, Metall. Trans. 4 (1973) 765–775.
References
[34] D. Sun, M. Li, Y. Zou, R. Yang, F. Li, Chin. Sci. Bull. 42 (1997) 1211–1215.
[35] P. Shewmon, Diffusion in Solids, second ed., TMS, Warrendale, 1989.
[1] M.-Y. Huh, O. Engler, Mater. Sci. Eng. A 308 (2001) 74–78. [36] F.A. Slooff, J. Zhou, J. Duszczyk, L. Katgerman, Scr. Mater. 57 (2007) 759–762.
[2] S.H. Park, K.Y. Kim, Y.D. Lee, C.G. Park, ISIJ Int. 42 (2002) 100–105. [37] Y.-C. Lin, M.-S. Chen, J. Zhang, Mater. Sci. Eng. A 499 (2009) 88–92.
[3] F.J. Humphreys, M. Hatherly, Recrystallization and Related Annealing Phe- [38] D. Samantaray, C. Phaniraj, S. Mandal, A.K. Bhaduri, Mater. Sci. Eng. A 528
nomena, first ed., Pergamon Press, Oxford, 1995. (2011) 1071–1077.
[4] S. Gourdet, F. Montheillet, Mater. Sci. Eng. A 283 (2000) 274–288. [39] S. Mandal, V. Rakesh, P.V. Sivaprasad, S. Venugopal, K.V. Kasiviswanathan,
[5] H. Yamagata, Y. Ohuchida, N. Saito, M. Otsuka, Scr. Mater. 45 (2001) Mater. Sci. Eng. A 500 (2009) 114–121.
1055–1061. [40] H. Mirzadeh, J.M. Cabrera, A. Najafizadeh, Metall. Mater. Trans. A 43 (2012)
[6] M.E. Kassner, J. Pollard, E. Evangelista, E. Cerri, Acta Metall. Mater. 42 (1994) 108–123.
3223–3230. [41] J.K. Solberg, H.J. McQueen, N. Ryum, E. Nes, Philos. Mag. A 60 (1989)
[7] M.E. Kassner, S.R. Barrabes, Mater. Sci. Eng. A 410–411 (2005) 152–155. 447–471.
[8] S.-I. Kim, Y.-C. Yoo, Met. Mater. Int. 8 (2002) 7–13. [42] M. Avrami, J. Chem. Phys. 7 (1939) 1103–1112.
[9] J.Y. An, S.M. Han, Y. Kwon, Y.C. Yoo, Mat. Sci. Forum 475–479 (2005) 145–148. [43] C.M. Sellars, J.A. Whiteman, Met. Sci. 13 (1978) 187–194.
[10] T.R. de Oliveira, F. Montheillet, Steel Res. Int. 79 (2008) 497–504. [44] M.C. Somani, L.P. Karjalainen, Mater. Sci. Forum 550 (2007) 583–588.
[11] T.R. Oliveira, F. Montheillet, Mater. Sci. Forum 467–470 (2004) 1229–1236. [45] G. Glover, C.M. Sellars, Metall. Trans. 3 (1972) 2271–2280.
[12] F. Gao, B. Song, Y. Xu, K. Xia, Metall. Mater. Trans. A 31 (2000) 21–27. [46] C.W. Sinclair, J.-D. Mithieux, J.-H. Schmitt, Y. Bréchet, Metall. Trans. A 36
[13] A. Belyakov, Y. Kimura, K. Tsuzaki, Mater. Sci. Eng. A 403 (2005) 249–259. (2005) 3205–3215.
[14] E.A. Jägle, E.J. Mittemeijer, Metall. Mater. Trans. A 43 (2012) 1117–1131. [47] G.H. Akbari, C.M. Sellars, J.A. Whiteman, Mater. Sci. Tech. 18 (2002) 885–891.
[15] S.S. Hazra, A.A. Gazder, E.V. Pereloma, Mater. Sci. Eng. A 524 (2009) 158–167. [48] D.R. Barraclough, C.M. Sellars, Met. Sci. 13 (1978) 257–267.
[16] M. Taheri, H. Weiland, A. Rollett, Metall. Mater. Trans A 37 (2006) 19–25. [49] A. Laasraoui, J.J. Jonas, Metall. Mater. Trans. A 22 (1991) 151–160.
[17] A. Godfrey, W.Q. Cao, N. Hansen, Q. Liu, Metall. Mater. Trans. A 36 (2005) [50] L.P. Karjalainen, J.A. Koskiniemi, X.D. Liu, in: Proceedings of the 37th MWSP
2371–2378. Conference, Hamilton, Canada, ISS, Warrendale, 1995, pp. 861–869.
[18] M.R. Barnett, J.J. Jonas, ISIJ Int. 37 (1997) 697–705.

You might also like