CX Plasticity

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Materials Science & Engineering A 561 (2013) 167–173

Contents lists available at SciVerse ScienceDirect

Materials Science & Engineering A


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

An experimental and numerical study of texture change and


twinning-induced hardening during tensile deformation
of an AZ31 magnesium alloy rolled plate
Adrien Chapuis n, Pei Liu, Qing Liu
College of Material Science and Engineering, Chongqing University, Chongqing 400044, China

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

Article history: The increase in flow stress due to work hardening, texture change and twinning has been examined
Received 14 June 2012 using experiments and a crystal plasticity model. Samples with five different orientations were cut
Received in revised form from a thick rolled plate of magnesium alloy AZ31 with a strong basal texture and subjected to tensile
1 November 2012
deformation. Samples elongated along the normal direction ND (01 samples) deformed mainly by
Accepted 5 November 2012
Available online 10 November 2012
tensile twinning, those deformed along the rolling direction RD (901 samples) deformed mainly by
prismatic slip, and samples deformed along different angles from ND to RD (301, 451 and 601 samples)
Keywords: deformed by both slip and tensile twinning. The mechanical properties and the observed microtextures
Magnesium alloy AZ31 were compared with the predictions of an original crystal plasticity model and to evaluate the
Twinning
strengthening induced by work hardening, texture rotation and twins. The analysis shows that the twin
Crystal plasticity
boundaries introduce additional Hall–Petch hardening that increases the CRSS of the hard slip systems
Hall–Petch
by 30%.
& 2012 Elsevier B.V. All rights reserved.

1. Introduction effect during compression parallel to the c axis of the lattice, but
in the present study pyramidal oc þa4 mode does not con-
Magnesium alloys are mainly used but seldom shaped at room tribute to the deformation.
temperature, because of their low ductility. The magnesium alloy The grain size dependency on twinning has already been
AZ31 (3% Al, 1% Zn, 0.6% Mn) is one of the most commonly used studied, and in compression perpendicular to the c axis, the yield
magnesium alloys at and below room temperature. Single crystal stress follows the Hall–Petch equation and varies from s0.002 ¼
studies [1] have been made to compute the critical resolved shear 32þ9.3d  0.5, with a slope k¼9.3 MPa mm1/2 [14,15], to sy ¼ 32þ
stresses (CRSSs) of pure magnesium at room temperature, but the 5.06  d  0.5 (grain size d in mm) [16].
CRSSs of pure magnesium are very different from the CRSSs of The main goal of this paper is to illustrate the influence of tensile
AZ31, as reviewed by Hutchinson and Barnett [2]. So it is twins and twin boundaries on the CRSSs and hardening of the AZ31
necessary to use models to evaluate the CRSSs of alloy AZ31 at alloy. In a planar simple shear test where prismatic slip is the
room temperature. Studies using experimental data and simula- expected deformation mode, Kang et al. [17] observed a high tensile
tions with EPSC or VPSC models [3,4] have determined the CRSSs twinning activity, and used the VPSC model to predict the texture
to input in these models. Models can be used to fit the stress change and the flow stress evolution. Stress was controlled more by
strain curves, the r-value and/or the texture evolution. The initial the RSS of prismatic slip than by any effect of twinning.
CRSSs values used in various models vary from 0.55 [5] to 25 MPa Knezevic et al. [9] used a crystal plasticity finite element
[6] for basal slip, and from 15 [7] to 55 MPa [8] for tensile model and studied a rolled sheet in compression along TD0 and
twinning. Thus, in some studies basal slip is easer than twinning RD0, implying a high activity of the tensile twinning. Their work
and in others it is the opposite. For prismatic slip the values concluded that the ‘main contribution (of the extension twins) to
obtained in tensile deformed samples vary from 1.5  CRSSbasal [9] hardening comes from texture hardening (rotation of grains into
to 90 MPa [10]. The initial CRSS of pyramidal o cþa 4 slip is hard orientations)’. Had they made a compression test along ND0
reported in the range 50 MPa [8] to 100 MPa [11–13]; Agnew to compare and fit the parameters of their program, they could
et al. [7] have emphasized the necessity of pyramidal slip and its have evaluated the contribution of the twin boundaries. In fact, if
the model considers a large amount of work hardening or a high
value of CRSS for the pyramidal ocþa 4 slip, the calculated
n
Corresponding author. Tel./fax: þ 86 23 65111295. stress–strain curves fit well with the experimental curve of a
E-mail address: adrienchapuis@orange.fr (A. Chapuis). sample compressed in a direction normal to ND0. On the other

0921-5093/$ - see front matter & 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.msea.2012.11.018
168 A. Chapuis et al. / Materials Science & Engineering A 561 (2013) 167–173

hand, Barnett et al. [18] used both a semi-physical analytical


constitutive model and the VPSC model to study an extruded bar
in compression and tension along the extrusion direction, imply-
ing tensile twinning and prismatic slip as the main deformation
mechanisms, respectively. To model the flow stress they used a
factor z ¼1.35 in the twinned grains, where z reflects the hard-
ening due to twin boundaries and dislocation–twin interactions.
In another study Barnett et al. [19] used a Sachs model and an
additional hardening factor z ¼ 2 in the twins.
Also Hong et al. [20] studied tensile twinning, performing
compression test along RD and tension test along ND of a rolled
plate. They concluded that f1 0 1 2g twinning has 3 roles:
(i) accommodation of plastic deformation, causing a low flow
stress and strain hardening rate; (ii) Hall–Petch hardening by
twinning-induced grain size change; and (iii) twin texture
induced change in activities of slips. They evaluated the Hall–
 0.5  0.5
Petch hardening (DsH  P ¼k(deff  dini ), k¼290 MPa mm0.5)
contribution to be 18% of the flow stress at 6–8% deformation in
tension, or 30 MPa. However, this value is not reliable because
they did not use a crystal plasticity program but just an ideal Fig. 1. True stress–strain curves of samples oriented at 0, 30, 45, 60 and 901 of ND
to RD of the initial rolled plate, tensile test at RT, strain rate 0.0001 s  1.
orientation analysis. They did not use their experimental stress to
compute the coefficient k, and evaluated k when twinning was
still the main deformation mechanism whereas the Hall–Petch f1 0 1 1g twin boundaries are in blue and 381 double twin
hardening mainly affects the deformation by slip. Flow stress f1 0 1 1g f1 0 1 2g boundaries in purple.
hardening in tension along the c-axis is mainly due to the
nucleation of different twin variants in one grain, and so to the
creation of twin variants boundaries [21]. 3. Simulation results
In this paper we first present experimental results from the
tensile tests on sample stressed at several angles to the plate from A rate insensitive model, i.e., Taylor-Bishop and Hill, was used
ND to RD, then simulations by crystal plasticity, and finally to simulate the stress, deformation, and texture evolution of the
discuss the additional necessary work hardening in the grains 5 different initial orientations. The rate insensitive law is ideal for
that experience twinning. To study the anisotropy in the ND–RD low temperature and twinning. It is based on the Schmid law: if
plane allow to activate different amount of tensile twinning, basal ts o tsc , g_ s ¼ 0, and if ts ¼ tsc , g_ s Z0, where tsc is the CRSS of the slip
and prismatic slip, as confirmed by the simulations. or twinning system ‘s’, ts is the shear stress and g_ s is the shear
rate. The shear rates and shear stresses are linked to the macro-
scopic stress tensor sij and strain rate tensore_ ij by standard
2. Experimental results P P
n
equations ts ¼ sij M sij and e_ ij ¼ g_ s Msij ; where Msij ¼ 12 ðbsi nsj þ
ij s¼1
2.1. Experimental procedures s
bj nsi Þ is the Schmid factor matrix, ‘n’ is the normal to the slip or
The initial material was a 50 mm thick rolled sheet of magne- twinning plane and ‘b’ is the normalised burger vector or shear
sium AZ31, annealed 1 h at 400 1C. Five samples of gauge length direction of the twinning. The program uses the simplex method
20 mm and Section 1mm by 7 mm, of total length 45 mm, were to maximise the function W _ ¼ Psij e_ ij , summed over the imposed
ij
spark cut with the tensile direction at various orientations: ND (01
strain components [22], under the constraint of the CRSSs.
sample), at 301, 451 and 601 from ND to RD and along RD (901
According to some authors [23,24], twinning is also assumed to
samples). Tensile tests were carried out on an AGX-10kN machine
respect the Schmid law, and twinning is modelled as a unidirec-
at a constant strain rate of 0.0001 s  1. Stress strain curves are
tional slip system. A particular feature of the model in the case of
shown in Fig. 1. Some of the samples were deformed to 4 and 8%
twinning, is to break up the crystal into twinned parts and an
true strain for EBSD observation.
un-twinned parent part, whose volume fractions evolve with
Metallographic samples were manually polished to p4000 and
deformation; the twins and parent crystal are assumed to deform
electro-polished with ACII-1 solution for 1 min at 20 V. Observa-
with the same imposed strain ezz. The twin volume fraction of a
tions were made using an FEI Nova400 FEG SEM with an Oxford s
Instruments-HKL Technology Nordlys EBSD system, using a step grain is computed with the relation g_ s ¼ f_ :gs where gs is the
v
size of 2 mm. shear rate associated with each twin (0.1289 for f1 0 1 2g, 0.1377
for f1 0 1 1g). Our study used an original grain constraint model:
2.2. Results in uniaxial tension ezz is imposed, exx and eyy are free. To allow for
the interaction between grains and its evolution with deforma-
The initial material has a fairly homogeneous grain size of tion, we assume exy, exz and eyz are free if they are small at low
about 40 mm (Fig. 2a, observed on the ND–RD plane) with a strain; but at high strain exy, exz or eyz are imposed to low values if
common basal texture. Fig. 2b to f show the EBSD maps for all of the shear becomes large. In practice the shear components are
the samples after a strain e ¼ 8% (observed on the plane normal to first computed with a relaxed model, then compared, and, if
the loading direction). For clarity Fig. 2g to i show the maps with necessary, the deformation is re-computed with a constrained
only the twin boundaries. In the 01 sample the common 861 model (i.e., imposed shear component). It is not a problem to
tensile twin boundaries between the twins and the matrix are impose exy whereas exx and eyy are always free, because in most
shown in red and the 601 boundaries between 2 tensile twin samples the texture is not axi-symmetric around z. To fit with the
variants in green. In the 901 sample some 561 compression 5 experimental yield stresses and the 901 sample stress strain
A. Chapuis et al. / Materials Science & Engineering A 561 (2013) 167–173 169

Fig. 2. EBSD map and (0 0 0 1) and {11–20} pole figures (a) initial texture, (b) 01 sample deformed 8%, (c) 301 sample deformed 8%, (d) 451 sample deformed 8%, (e) 601
sample deformed 8%, (f) 901 sample deformed 8%, (g) 01 sample deformed 8%, (h) 451 sample deformed 8%, (i) 901 sample deformed 8%. (For interpretation of the references
to color in this figure legend, the reader is referred to the web version of this article.)
170 A. Chapuis et al. / Materials Science & Engineering A 561 (2013) 167–173

curves, the initial CRSSs used in the program are 16 MPa for basal subset of grains with the c axis nearly perpendicular to the
slip, 82 MPa for prismatic slip, 98 MPa for pyramidal ocþ a4 loading direction; at this strain the EBSD indexing rate is still
slip, 33 MPa for tensile twinning and 163 MPa for compressive high and according to the stress strain curves most of the tensile
f1 0 1 1g twinning. To simplify the work hardening, all the slip twinning has occurred. The 01 sample at fracture (e ¼ 15%) still has
modes follow the same Voce saturating law: ts ¼ tsc0 þ 39 a 10% volume fraction of parent material (indexed in EBSD). The
ð1exp ð15  GÞÞ, withts the current RSS of slip mode s, tsc0 the number and fraction of grains with twins was counted manually.
P For the simulated results the values at 20% strain are presented
initial CRSS and G ¼ gs the total amount of shear. There is no
s because the model under-estimates the twin volume fraction
work hardening for the CRSS of twinning, neither positive nor during deformation and this study is focused on the flow stress
negative as used by Proust et al. [8]. The total stress is a weighted after twinning. We did not try to model accurately what hap-
average of the stresses of the grains and their parts. To obtain the pened during twinning because the model split the grains into
S shape curve characteristic of tensile twinning, the softest part of parent and twinned parts. While the simulated twin volume
the grain (i.e., parent part) have more weight than the hard parts fractions and percentages of grains with twins seem to be less
(i.e., tensile twin). The simulated stress–strain curves are shown than the experimental ones, the differences observed are too
in Fig. 3. A few f1 0 1 1g compression twins appear in the small to explain the difference between the experimental and the
simulations at about 10% strain in the 901 sample, in agreement simulated flow stress.
with the observations (Fig. 2f and i). The increase of the flow Fig. 4a to f compare the experimental (e ¼8%) and the simu-
stress is due to work hardening, texture rotation by slip, texture lated (e ¼10%) inverse pole figures for the 0, 45 and 901 sample.
change due to twinning, and the effect of the constraints on exy, exz The texture change due to twinning and to slip is more pro-
and eyz. nounced in the simulations.
The flow stress tendency is in relatively good agreement with In order to validate our model, we compared some of our
the experimental results, but regardless of the parameters chosen results with simulations made with EVPSC for similar tilted
in the model, the 01 sample flow stress is always much too low orientations [13]. The microtexture of the initial rolled sheet is
after twinning (at e 410%); similarly in the samples between 30 not exactly the same, in particular the intensity of the basal
and 601 the hardening is too low. Also we cannot have both texture, so the results cannot be expected to be identical. Also the
significant twinning (observed in the 01 sample, and, to a lesser initial CRSSs used in the EVPSC are different, with lower CRSS for
degree, in the 301 and 451 samples), and a low flow stress for the basal slip and harder twinning, (respectively 9 and 47 MPa). Fig. 5
451 sample (which need a low initial CRSS for basal slip). So the shows the activity of deformation modes, basal, prismatic and
selected initial CRSS and the work hardening rates are the result pyramidal oc þa4 slip, and tensile twinning; compared to the
of a compromise. simulations of Wang et al. [13]: basal and prismatic activity are
A useful way to validate the model is to compare the simulated very close for the 45, 60 and 901 samples, despite the lower CRSS
and experimental twin volume fractions. Table 1 summarizes the in the VPSC. Although twinning is more active at 10% strain than
tensile twin volume fraction and the percentage of grains with 1% because of the twinning feature of our model, Wang et al. also
twins for experiments and simulations. The experimental simulated a similar amount of twinning. Note the figures report
twinned volume fraction at e ¼8% has been measured from a the true amount of slip g for a deformation step of 1%. The
r-values computed by Wang et al. for a 45, 60 and 901 samples are
0.21, 0.37 and 2.22, respectively, our model calculated r ¼ eTD/eX
(X ¼RD for the 01 sample, X ¼ND for the 901 sample) at 10% strain
for the 0, 30, 45, 60 and 901 samples: simulated r values are 1.02,
0.64, 0.46, 0.58 and 2.33, respectively.

4. Discussion of Hall–Petch hardening

We observed that the stress strain curves of all the twinned


samples were lower than the experimental stress strain curves.
In this section we interpret this phenomenon as a Hall–Petch
effect due to the twin boundaries.
The main hypothesis is that the grain size changes the CRSSs
which must therefore increased in the grains with twins to obtain
flow stresses close to those of the experiments. Table 2 shows a
correlation between the percentage of grains with twins and the
Fig. 3. Simulated stress–strain curves of 0, 30, 45, 60 and 901 samples, the same difference between the experimental and the previously simu-
parameters are used in grains and twins. lated flow stresses. We consider the twin boundaries are obstacles
for the dislocations, in consequence we must increase the CRSS of
Table 1
the slip modes, and so the stress in the twinned grains increases.
Experimental (e ¼ 8%) and simulated (e ¼20%) {10–12} twin volume fraction and Grains without twins need one set of CRSSs (in practice the CRSSs
fraction of grains with {10–12} twins (Gwt). used in the previous simulation in Fig. 3), and twinned grains
require another set of CRSSs, i.e., an additional Hall–Petch work
Sample Exp Ttw vol frac Exp frac Gwt Sim Ttw vol frac Sim frac Gwt
hardening, taken proportional to the first set.
(e ¼ 8%) (e ¼ 8%) (e ¼ 20%) (e ¼20%)
Jain et al. [25] simulated the CRSS dependency with grain size,
0 59 98 79.3 94.8 and observed that the Hall–Petch strength dependence is mainly
30 46.1 85 51.9 78.9 determined the strength of the prismatic slip systems. To test this
45 30 75 26 54.9 observation, different simulations have been made with an
60 28 45 12.7 31.5
90 – 10 0.48 4.7
additional hardening in the grains with twins to increase the
CRSSs of either all the slip modes or only those of prismatic and
A. Chapuis et al. / Materials Science & Engineering A 561 (2013) 167–173 171

Fig. 4. Z (Tensile direction) inverse pole figures, (a), (c), and (e) experimental at e ¼ 8% and (b), (d), and (f) simulated at e ¼ 10% for the ((a), (b)) 0, ((c), (d)) 45 and ((e), (f))
901 samples.

Fig. 5. Simulated deformation modes activities, between (a) e ¼ 0–1% and (b) e ¼ 9–10%.

pyramidal oc þa4 slip. If we added 25 or 30% to the CRSSs of only on the prismatic and pyramidal ocþa 4 slip modes
basal, prismatic and pyramidal ocþa 4 slip, the stress strain (Table 2). The 301 sample is always harder than in the experiment.
curves are improved. If 30% twin-induced hardening is added on Fig. 6 shows the simulated stress–strain curves with the CRSS of
the basal slip and hard slip modes (Hall–Petch factor¼1.3), the prismatic and pyramidal oc þa4 slip increased by 30% in the
simulated flow stresses (at e ¼ 16% of the 30, 45 and 601 samples) twinned grains. With such parameters the simulated stresses
are 10–15 MPa more than those with twin induced hardening have an acceptable difference. There are no appreciable changes
172 A. Chapuis et al. / Materials Science & Engineering A 561 (2013) 167–173

Table 2
Experimental stress (e ¼0.16 for 0, 30, 45 and 601 samples, e ¼ 0.12 for 901 sample)), simulated stress (e ¼ 16% for all samples)
without Hall–Petch hardening, simulated fraction of grains with twins, difference between experimental and simulated stress,
simulated stress (e ¼ 16%, in MPa) with Hall–Petch hardening ¼ 1.3 on basal, prismatic and pyramidal cþ a slip, and simulated stress
(e ¼16%, in MPa) with Hall–Petch hardening ¼1.3 only on prismatic and pyramidal o c þa4 slip.

Sample Exp sexp (MPa) Sim s(1)a


zz (MPa) % Gwt (e ¼0.2) sexp–szz (MPa) s(2)b
zz HP bas þpri s(3)c
zz HP pri

0 322 254 95 68 327 319


30 255 220 79 35 273 260
45 245 214 55 31 250 234
60 259 234 31 25 255 245
90 (e ¼ 0.12) 273 276 5 – 277 277

a
in all grains RSSbasal ¼ 16 þ39(1  exp(  15G)), RSSprismatic ¼ 82þ 39(1  exp(  15G)).
b
in twinned grains, RSSbasal ¼21 þ50(1  exp(  15G)), RSSprismatic ¼ 106 þ50(1  exp(  15G)).
c
in twinned grains, RSSbasal ¼16 þ39(1  exp(  15G)), RSSprismatic ¼ 106 þ50(1  exp(  15G)).

rolled plate with a basal texture, so that they deformed mainly by


tensile twinning, basal slip or prismatic slip. The CRSSs values
used in our model were consistent with those used in the VPSC
model. Three main conclusions were made from our study:

Experimentally the 901 sample and the 01 sample after twin-


ning (after ezz ¼10%) have very similar textures, with the c axis
of the grains or twins are perpendicular to the tensile direc-
tion. But the 01 sample after twinning exhibits a higher stress
than the 901 sample, so twinning leads to an extra hardening
(independent of the reorientation).
In the simulations all the samples with tensile twinning needed
an additional hardening component. This hardening is propor-
tional to the number of grains that deform by twinning. With the
chosen parameters this twin boundary-induced hardening
Fig. 6. Simulated stress–strain curves of 0, 30, 45, 60 and 901 samples, in the
increases the stress and CRSSs between 20 and 30%.
grains with twins the CRSS of prismatic and pyramidal o cþ a4 slip are 1.3 times We modeled this hardening as a Hall–Petch relationship due to
the CRSS in the grains without twins. the twin boundaries that reduced the grain size by a factor of
4. For prismatic slip in highly deformed samples, the Hall–
in the twin volume fraction, for either tension or compression Petch slope value, k, is 7.25. As in the study of Jain et al. [25],
twinning. we cannot conclude if there is a Hall–Petch hardening on the
We estimated the correlation between the grain size and the basal mode.
final CRSSs, using an initial grain size of 40 mm (0.04 mm), and, in
the twinned grains, a spacing between boundaries of about 10 mm
(Fig. 2a to i). We made the hypothesis that the CRSSs also follow Acknowledgments
the Hall–Petch law: CRSS(10 mm) ¼CRSS(N)þ k/O0.01 ¼1.3 
CRSS(40 mm) and CRSS(40 mm)¼ CRSS(N) þk/O0.04, where k is This project is co-supported by the National Natural Science
the Hall–Petch slope value for a given slip system. For prismatic Foundation of China (51101175 and 50890172).
slip: CRSSprismatic(40 mm) ¼121 MPa, CRSSprismatic(10 mm) ¼156 The first author carried out the theoretical study and thanks Pr.
MPa and k(prismatic)¼7.25. We estimate that the twin bound- J.H. Driver for insights on crystal plasticity. The second author has
aries increase the CRSSs of the prismatic and pyramidal oc þa4 performed the tensile tests and EBSD experiments.
slip system by 30%.
To compare the Hall–Petch slope with those of other studies,
References
for a given flow stress, we take the simulated flow stress at
e ¼20% in grains with twins in the 01 sample; the flow stresses are
[1] A. Chapuis, J.H. Driver, Acta Mater. 59 (2011) 1986–1994.
264 MPa for the first simulation (grain size ¼0.04 mm), and [2] W.B. Hutchinson, M.R. Barnett, Scr. Mater. 63 (2010) 737–740.
332 MPa with a Hall–Petch factor 1.3 (grain size¼0.01 mm), so [3] P.A. Turner, C.N. Tomé, Acta Metall. Mater. 42 (1994) 4143–4153.
s20% ¼198 þ13.4/Od. This value can be compared to the value of [4] R.A. Lebensohn, C.N. Tomé, Acta Metall. Mater. 41 (1993) 2611–2624.
[5] A. Staroselsky, L. Anand, Int. J. Plast. 19 (2003) 1843–1864.
k¼7.3 in tension along RD in the study of Jain et al. [25] obtained [6] A. Jain, S.R. Agnew, Mater. Sci. Eng., A 462 (2007) 29–36.
at e ¼2%. Their measures showed that the Hall–Petch hardening [7] S.R. Agnew, M.H. Yoo, C.N. Tomé, Acta Mater. 49 (2001) 4277–4289.
affects the initial CRSS value of prismatic slip but not the work [8] G. Proust, C.N. Tomé, A. Jain, S.R. Agnew, Int. J. Plast. 25 (2009) 861–880.
[9] M. Knezevic, A. Levinson, R. Harris, R.K. Mishra, R.D. Doherty, S.R. Kalidindi,
hardening behavior. Their simulations also showed similar results Acta Mater. 58 (2010) 6230–6242.
with or without Hall–Petch hardening on the CRSS of basal slip. [10] S.R. Agnew, D.W. Brown, C.N. Tomé, Acta Mater. 54 (2006) 4841–4852.
[11] H. Wang, B. Raeisinia, P.D. Wu, S.R. Agnew, C.N. Tomé, Int. J. Solids Struct. 47
(2010) 2905–2917.
[12] B. Clausen, C.N. Tomé, D.W. Brown, S.R. Agnew, Acta Mater. 56 (2008)
5. Conclusions 2456–2468.
[13] H. Wang, P.D. Wu, M.A. Gharghouri, Mater. Sci. Eng., A 527 (2010)
In this study we have developed a model to explain the flow 3588–3594.
[14] A. Ghaderi, M.R. Barnett, Acta Mater. 59 (2011) 7824–7839.
behavior of polycrystalline AZ31 Mg alloy in tension at room [15] M.R. Barnett, Z. Keshavarz, A.G. Beer, D. Atwell, Acta Mater. 52 (2004)
temperature. Samples of different orientations were cut from a 5093–5103.
A. Chapuis et al. / Materials Science & Engineering A 561 (2013) 167–173 173

[16] S.H. Park, S.-G. Hong, J.H. Lee, C.S. Lee, Mater. Sci. Eng. A 532 (2012) 401–406. [22] M. Renouard, M. Wintenberger, C.R. Acad, Sci. Paris 283B (1976) 237–240.
[17] J.-Y. Kang, B. Bacroix, R. Brenner, Scr. Mater. 66 (2012) 654–657. [23] D.R. Thornberg, H. Piehler, Metall. Trans. 6A (1975) 1511–1523.
[18] M.R. Barnett, C.H.J. Davies, X. Ma, Scr. Mater. 52 (2005) 627–632. [24] M.A. Gharghouri, G.C. Weatherly, J.D. Embury, Root J. Phil. Mag. 79A (1999)
[19] M.R. Barnett, Z. Keshavarz, X. Ma, Metall. Mater. Trans. 37A (2006) 1671.
2283–2293. [25] A. Jain, O. Duygulu, D.W. Brown, C.N. Tomé, S.R. Agnew, Mater. Sci. Eng., A
[20] S.-G. Hong, S.H. Park, C.S. Lee, Acta Mater. 58 (2010) 5873–5885. 486 (2008) 545–555.
[21] S.-G. Hong, S.H. Park, C.S. Lee, Scr. Mater. 64 (2011) 145–148.

You might also like