On the deformation twinning mechanisms in twinning-induced plasticity steel

Abstract

Deformation twinning in coarse-grained fcc metals results from the highly coordinated glide of Shockley partial dislocations with the same Burgers vector on successive {111}-type twinning planes. The main issue of the formation mechanism of deformation twinning is how the arrangement of Shockley partials required for twinning evolves. Here, we present the deformation twinning mechanism that operates in Fe-17Mn-0.45C-1.5Al-1Si TWIP steel by means of in-situ TEM deformation tests and post-mortem TEM characterization. The in-situ TEM observation shows the formation of a thin twin by Shockley partial dislocations generated in a grain boundary, which indicates the nucleation of deformation twins at grain boundary defect sites. The observed twinning process is similar to deformation twinning in nano-crystalline materials through successive emission of partial dislocations from grain boundaries. In addition, a high density of Frank partial dislocations is observed inside deformations twins. These defects affect the growth of deformation twins and could contribute to the high work hardening of TWIP steel.

Full text

Full length article
On the deformation twinning mechanisms in twinning-induced
plasticity steel
Jin-Kyung Kim
*
, Min-Hyeok Kwon, Bruno C. De Cooman
Graduate Institute of Ferrous Technology, Pohang University of Science and Technology, 77 Cheongam-Ro, Nam-Gu, Pohang 37673, South Korea
article info
Article history:
Received 19 April 2017
Received in revised form
20 September 2017
Accepted 20 September 2017
Available online 21 September 2017
Keywords:
In-situ transmission electron microscopy
Deformation twinning
TWIP steel
Shockley partial dislocation
Frank partial dislocation
abstract
Deformation twinning in coarse-grained fcc metals results from the highly coordinated glide of Shockley
partial dislocations with the same Burgers vector on successive {111}-type twinning planes. The main
issue of the formation mechanism of deformation twinning is how the arrangement of Shockley partials
required for twinning evolves. Here, we present the deformation twinning mechanism that operates in
Fe-17Mn-0.45C-1.5Al-1Si TWIP steel by means of in-situ TEM deformation tests and post-mortem TEM
characterization. The in-situ TEM observation shows the formation of a thin twin by Shockley partial
dislocations generated in a grain boundary, which indicates the nucleation of deformation twins at grain
boundary defect sites. The observed twinning process is similar to deformation twinning in nano-
crystalline materials through successive emission of partial dislocations from grain boundaries. In
addition, a high density of Frank partial dislocations is observed inside deformations twins. These defects
affect the growth of deformation twins and could contribute to the high work hardening of TWIP steel.
©2017 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
1. Introduction
Deformation twinning is an important deformation mechanism
in fcc metals and alloys [1]. Generally, low stacking fault energy
(SFE) fcc metals and alloys show a high twinning activity while high
SFE fcc metals and alloys deform by slip. Deformation twinning in
coarse-grained fcc metals results from homogeneous shearing of
matrix by the highly coordinated glide of Shockley partial dislo-
cations with the same Burgers vector on successive {111}-type
twinning planes [1e3]. In contrast, there have been reports on the
occurrence of deformation twinning by the movement of Shockley
partial dislocations with different Burgers vectors in nano-
crystalline metals [3e5]. Deformation twinning has been associ-
ated with a high work hardening during deformation of fcc metals
[6].
Due to the importance of deformation twinning in fcc metals
and alloys, twin formation mechanisms have been a focus of
research. As Mahajan [7] pointed out, the main issue of the for-
mation mechanism of deformation twinning is how the arrange-
ment of Shockley partials required for twinning evolves. Most
models of deformation twinning assume that a twin is nucleated at
a heterogeneity such as a particular dislocation arrangement [8].
Venables [9] proposed that a prismatic glide source lying on the
primary slip plane could dissociate into a sessile a/3〈111 〉-type
Frank partial dislocation and a a/6〈112〉-type Shockley partial
twinning dislocation on the conjugate slip plane, under the action
of an externally applied stress. Other models are based on a devi-
ation process [10,11]. Cohen and Weertman [10] proposed that a
perfect a/2〈110 〉-type dislocation in front of a strong barrier such as
a Lomer-Cottrell lock could dissociate into a sessile a/3〈111〉-type
Frank partial dislocation and a a/6〈112〉-type Shockley partial
dislocation on a conjugate plane. In their model, the partial dislo-
cation can glide away from the Frank partial dislocation, trailing a
wide stacking fault. Fujita and Mori [11] proposed the stair-rod
cross-slip mechanism which requires the dissociation of a a/6
〈112 〉-type Shockley partial dislocation on the primary slip plane
into a sessile a/6〈110 〉-type stair rod dislocation and a glissile a/6
〈112 〉-type Shockley partial dislocation on a conjugate slip plane.
Mahajan and Chin [8] proposed a different type of model based on
the presence of an extrinsic stacking fault. They proposed that two
coplanar perfect a/2〈110 〉-type dislocations could form a set of
three Shockley partial dislocations on three consecutive {111}-type
slip planes, resulting in an extrinsic stacking fault configuration. In
the Miura-Takamura-Narita mechanism, extended dislocations on
the primary slip plane form a dislocation pileup at a Lomer-Cottrell
barrier and interact with a Lomer dislocation [12]. This interaction
*Corresponding author.
E-mail address: intobe@postech.ac.kr (J.-K. Kim).
Contents lists available at ScienceDirect
Acta Materialia
journal homepage: www.elsevier.com/locate/actamat
https://doi.org/10.1016/j.actamat.2017.09.043
1359-6454/©2017 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Acta Materialia 141 (2017) 444e455

results in a sessile a/3〈111〉-type Frank partial dislocation and two
a/6〈112 〉-type twinning partial dislocations on the primary slip
plane. While various models for deformation twinning are based on
suitable dislocation reactions, the role of grain boundaries on the
initiation of deformation twins has not been explored in detail
despite the evidence for deformation twinning near grain bound-
aries in coarse-grained fcc materials [13e15]. Beyerlein and Tom
e
[16] proposed a probabilistic theory for the nucleation of defor-
mation twins in hcp metals based on the assumption that twin
nucleation relies on the dissociation of grain boundary defects
under stress into the required number of twinning partials to create
a twin nucleus.
High Mn Twinning-induced plasticity (T WIP) steel is a type of
structural steel, characterized by both high strength and superior
formability [17,18]. In TWIP steel, it is generally accepted that
deformation twinning results in an increased strain hardening rate
by the creation of twin boundaries which act as effective obstacles
to dislocation glide by a dynamic Hall-Petch effect [19]. There have
been earlier attempts to characterize deformation twinning
mechanisms in TWIP steels [20e24]. Karaman et al. [20] proposed a
modified version of the Miura-Takamura-Narita mechanism by
considering local pile-up stresses, SFE, the effect of an applied stress
on the separation of partial dislocations and the increase in the
friction stress due to a high solute concentration for Hadfield
Fe12Mn1C steel. Bracke et al. [21] argued that their Schmid factor
analysis supported the twinning model proposed by Mahajan and
Chin, although they did not carry out a thorough characterization of
the dislocations they observed. Idrissi et al. [22] characterized the
twinning mechanisms of Fe20Mn1.2C steel by transmission elec-
tron microscopy (TEM). Their results appear to support the Cohen
and Weertman mechanism or the Miura-Takamura-Narita mecha-
nism for twin nucleation. They also reported that the presence of a
high density of sessile a/3〈111〉-type Frank dislocations within
twins affects the growth and stability of twins. Based on ex-situ
TEM observations, Liu et al. [23] proposed a modified version of
the Fujita and Mori model, i.e. after the cross-slip of the first a/6
〈112 〉-type Shockley partial dislocation, the interactions between a
sessile a/6〈110 〉-type stair rod dislocation and the second a/6〈112 〉-
type Shockley partial dislocation form a sessile a/3〈111 〉-type Frank
partial dislocation. Recently, Mahato et al. [24] reported that the
Mahajan and Chin model was consistent with their observations of
the contrast from three-layer stacking faults.
Deformation twins and bundles of stacking faults were
frequently observed near grain boundaries in TWIP steel [14,15,25].
However, there has been no detailed research on deformation
twinning mechanisms focusing on the role of grain boundaries. The
present work aims at revealing deformation twinning mechanisms
of a TWIP steel by in-situ TEM straining tests and post-mortem TEM
characterizations, and discusses the role of grain boundaries on the
formation of deformation twins.
2. Experimental
The chemical composition of the TWIP steel used in the present
study is listed in Table 1. An as-cast ingot was machined to smaller
slabs, which were rough rolled to a thickness of 25 mm, and finish
rolled to a thickness of 2.5 mm. As-cast slabs were reheated at
118 0

C for 1 h to remove the segregation of the alloying elements.
Hot-rolling was finished at 900

C. The hot-rolled steels were given
a coiling simulation at 450

C for 1 h. The as-hot rolled steels were
not fully recrystallized. The hot-rolled materials were therefore
given an additional annealing treatment of 10 min at 1000

C,
cooled down to 300

C using a very low cooling rate of 1

C/min,
and cooled in air to room temperature.
ASTM E8 sub-size tensile test samples with a gauge length of
25 mm, a gauge width of 6 mm and a thickness of 2.5 mm were
machined from the hot-rolled sheets with their long axis aligned
along the rolling direction. Tensile tests were carried out at a strain
rate of 10
3
s
1
using a ZWICK Z100 universal tensile testing ma-
chine. Interrupted tensile tests were carried out to observe the
microstructure developed at predetermined strains. Microstructure
analysis was carried out in a TEM equipped with an in-situ defor-
mation holder and post-mortem analysis was done in a conven-
tional TEM. TEM samples for post-mortem TEM characterizations
were prepared as 3 mm diameter disks which were mechanically
polished to a thickness less than 100
m
m and thinned by the elec-
trolytic double jet technique at room temperature. A mixture of 5%
perchloric acid and 95% acetic acid was used as electrolyte.
TEM samples for in-situ deformation TEM tests were also pre-
pared as 3 mm diameter disks, similarly to the standard TEM
sample preparation method. The edges of the 3 mm diameter disks
were trimmed to obtain the samples elongated along the tensile
direction. Electrolytic polishing was carried out using the same
method as the one used for samples for post-mortem TEM char-
acterizations. The as-polished samples were glued on a Cu support.
The in-situ TEM observations were carried out in a JEOL JEM-
2100F TEM operating at 200 kV. To strain the samples in the
TEM, the edge-sliced disc for in-situ deformation tests was loaded
in a single-tilt straining TEM holder (Model 652TM, Gatan) which
elongated the sample by means of a micrometer screw driven by a
DC motor. The sample was strained at a strain rate of 0.1
m
m/s. The
in-situ TEM observations were recorded in real time with a CCD
camera (ORIUS 200D, Gatan) at 25 frames per second.
The post-mortem TEM observations were carried out in a JEOL
JEM-2100F TEM operating at 200 kV using a double-tilt specimen
holder. Bright-field (BF), weak-beam dark-field (WBDF) TEM im-
ages and related selected-area diffraction patterns (SADPs) were
recorded. In order to analyze dislocations, stacking faults and twins,
their TEM contrast for several two-beam conditions was analyzed
using the g$b¼0 invisibility criterion. For the measurements of the
stacking fault energy, long isolated dislocations were analyzed on
their {111}-type glide plane by means of WBDF imaging in (g, 3g)
diffraction conditions.
3. Results
3.1. Mechanical properties and stacking fault energy of Fe-17Mn-
0.45C-1.5Al-1Si TWIP steel
Fig. 1(a) shows the engineering stress-strain curve of the
investigated hot-rolled and annealed Fe-17Mn-0.45C-1.5Al-1Si
TWIP steel. The investigated TWIP steel shows a yield strength
(YS) of 327 MPa, an ultimate tensile strength (UTS) of 826 MPa, and
a total elongation of 63.2%. The tensile curve shows a smooth
elasto-plastic transition without a yield plateau. As a result of the
addition of Al, the number of serrations on the stress-strain curve is
very limited. Fig. 1(b) presents the true stress-strain curve and
strain hardening rate curve. The strain hardening curve is charac-
terized by three stages, i.e. stages A, B and C. In stage A, the strain
hardening continuously decreases with strain. In stage B, approxi-
mately from ε¼0.08 to 0.23, there is a slight increase of the strain
hardening with strain. Stage C shows a continuous decrease of the
strain hardening with strain.
Table 1
Chemical composition of the investigated TWIP steel.
Mn C Al Si Ti N Fe
16.6 0.45 1.57 1.00 0.062 0.007 Balance
J.-K. Kim et al. / Acta Materialia 141 (2017) 444e455 445

Fig. 2 shows TEM micrographs of the investigated TWIP steel
deformed to a strain of 1%. Fig. 2(a) shows a two-beam WBDF image
of dissociated dislocations together with stacking faults formed on
the two inclined {111} type slip planes. Assuming the crystal
orientation as shown by the Thomson tetrahedron, several inclined
stacking faults indicated by the white arrows are formed on the
ð111Þplane, parallel to AC in the Thompson tetrahedron. In addi-
tion, a stacking fault on the ð111Þplane, parallel to AB in the
Thompson tetrahedron, is indicated by the white dashed arrow.
This observation indicates that multiple slip is activated in slightly
deformed Fe-17Mn-0.45C-1.5Al-1Si TWIP steel. Fig. 2(b) shows a
WBDF image of the dissociated dislocations taken in the same grain
as in Fig. 2(a). Among the three {220} diffraction conditions avail-
able in the 〈111 〉zone axis of austenite, only one {220} diffraction
condition makes both partial dislocations visible. Since both partial
dislocations are visible in the two-beam condition for g ¼ð
220Þ, the
Burgers vector of the perfect dislocation is parallel to the direction
of the reciprocal lattice vector in the two-beam condition. As
indicated by the white arrows, constrictions are present between
the regions where partial dislocations are straight and parallel. At
the constrictions, two partial dislocations recombine and locally
become a perfect dislocation by following dislocation reaction:
a
6h121iþa
6h211i/a
2h110i(1)
Partial dislocation separations were measured along the length
of the dislocation as shown in Fig. 2(b). The observed partial
dislocation separation, d
obs
, ranges from 7.4 nm to 13.8 nm while
the angle between the Burgers vector of the perfect dislocation and
the dislocation line,
q
, ranges from 13

to 46

.Fig. 2(c) shows the
true partial dislocation separation (d) as a function of
q
. The effect of
the dislocation core on the partial dislocation separation was
considered by following the method proposed by Williams and
Carter [26], thereby yielding the true partial dislocation separation
distance, d, derived from d
obs.
The SFE was then estimated by
comparing the experimental data points with the theoretically
predicted curves based on the following equation [27]:
d¼G
b
p


2
8
pg
$
2
y
1
y
12vcos 2
q
2
y
(2)
Here G is the shear modulus, b
p
is the magnitude of the Burgers
vector of partial dislocations and
y
is the Poisson's ratio. The value
of shear modulus G (70.1 GPa) and Poisson's ratio
y
(0.24) for Fe-
18Mn-0.6C-1.5Al TWIP steel was used for the analysis [28]. The
magnitude of the Burgers vector of partial dislocations b
p
(0.1476 nm) for the investigated TWIP steel was obtained from the
lattice parameter measured by X-ray diffraction. The comparison of
the experimentally measured data points with Eq. (2) yields the SFE
value of 12 ±5 mJ/m
2
. The estimated SFE value of Fe-17Mn-0.45C-
1.5Al-1Si TWIP steel is similar to that of Fe-18Mn-0.6C TWIP steel
(13 ±3 mJ/m
2
)[29]. This is due to the compensating effect of each
alloying element on the SFE of TWIP steel, i.e. C and Al increase the
SFE while Si decreases the SFE [28e30].
3.2. In-situ TEM observations of deformation twinning
Fig. 3 shows snapshots of an in-situ straining TEM test of the
investigated TWIP steel. In the initial stage of deformation, three
planar defects are present in the investigated grain as shown in
Fig. 3(a). The tensile axis is indicated by a white double arrow. The
three planar defects are parallel, indicating that the defects have
the same habit plane. In addition to the planar defects, some par-
ticles are also observed. Energy dispersive spectroscopy (EDS)
analysis was used to identify the particles as (Fe, Mn)C carbides. The
upper two planar defects show stacking fault fringes while the
bottommost planar defect has a homogeneous dark contrast.
Therefore, the upper two planar defects are stacking faults or multi-
layered stacking faults of twin nucleus, while the bottommost
planar defect is a growing deformation twin which was confirmed
by the presence of twin diffraction spots of 〈125〉type zone axis
diffraction. Hereafter, the uppermost fault is referred to as fault 1
while the fault in the middle is referred to as fault 2. In fault 1, four
dislocations are visible, indicated by the labels A, B, C and D. It
should be noted that the dislocation D is not clearly visible in
Fig. 3(a). At increasing load, a new dislocation generated on the
right side of Fig. 3(b) approaches dislocation A, changing the
stacking fault contrast from fringes to a homogeneous white
contrast (Supplementary movie 1). It is known that the stacking
fault contrast vanishes for every third stacking fault in overlapping
stacking faults [26]. Here, we assume that the newly generated
dislocation is a twinning dislocation (TD), and that the white fault is
a thin twin with the thickness of multiples of three atomic layers.
During further straining, the TD moves continuously towards the
position of the dislocation D (Fig. 3(c)e(e) and supplementary
movies 2e4). The distance between the TD and the dislocation D is
approximately 0.77
m
m for Fig. 3(c), 0.24
m
m for Fig. 3(d) and 0.1
m
m
for Fig. 3(e). As the TD approaches the dislocation D, the dislocation
C seems to move into the grain boundary region on the left side of
the image. Fig. 3(e) shows new two planar defects, fault 3 and fault
4, generated on the lower side of the three planar defects; the
defects were emitted from the grain boundary on the left side of the
image (Supplementary movie 4). Between the timeframe of
Fig. 1. (a) Engineering stress-strain curve and (b) true stress-strain curve and strain
hardening rate curve of the hot-rolled and annealed Fe-17Mn-0.45C-1.5Al-1Si T WIP
steel.
J.-K. Kim et al. / Acta Materialia 141 (2017) 444e455446

Fig. 3(e) and (f), the in-situ imaging was not able to capture the
changes taking place on the fault 1. However, after further straining,
the entire fault 1 is found to become homogeneously white as
shown in Fig. 3(f). In addition, the dislocation E is present although
its origin is not clear. Considering the fact that the entire region of
the fault 1 is white, the TD is expected to have propagated into the
grain boundary region on the left side of the image.
Supplementary video related to this article can be found at
https://doi.org/10.1016/j.actamat.2017.09.043
The deformation microstructure obtained after the in-situ
straining TEM test was characterized in detail. Fig. 4 presents a
two-beam BF micrograph showing the same grain analyzed in
Fig. 3. It should be noted that the fault 1 in Fig. 3 is present at the
bottomside in Fig. 4. Here, the grain on the right side of the
investigated grain is referred to as grain 2. The [110] zone axis
diffraction pattern and the Thompson tetrahedron indicate the
crystal orientation of the investigated grain. Both planar faults 1
and 2 are on the ð111Þplane.
In order to examine the nature of dislocations in the white fault,
a two-beam analysis using various diffraction conditions near the
[110] zone axis was carried out as shown in Fig. 5. The analysis
focused on three regions, i.e. (1) the region where the dislocations D
and E are present, (2) the region where the dislocations A and B are
present and (3) the region near the grain boundary where some
dislocations are localized. The g$b values for each perfect, Shockley
partial and Frank partial dislocation on the ð111Þplane associated
with the operating reflections (g-vector) are summarized in Table 2.
As shown in Fig. 5(a)e(h), the dislocations A, B, D and E are visible
for two-beam conditions with g¼(111Þwhile the same dislocations
are invisible for two-beam conditions with g¼(111Þ. For two-beam
conditions with g¼(020Þ, as shown in Fig. 5(j) and (k), the dislo-
cations are visible, but the contrast is lower as compared to the
two-beam conditions with g¼(111Þ. The dislocations show a strong
contrast for two-beam conditions with g¼(113Þ(Fig. 5(n) and (o)),
while the same dislocations show a faint contrast for two-beam
conditions with g¼(113Þ(Fig. 5(r) and (s)). For two-beam condi-
tions with g¼(220Þas shown in Fig. 5(v) and (w), the dislocations
show a faint contrast. Based on the g$b values shown in Table 2, the
Fig. 2. (a) TEM two-beam WBDF image of dissociated dislocations (dashed yellow arrow) together with stacking faults formed on the two inclined {111} type slip planes (white
arrows and dashed white arrow) (b) TEM two-beam WBDF image of a long dissociated dislocation segment taken in the same grain as (a). The white arrows indicate constrictions
present between the regions where the partial dislocations are straight and parallel. Partial dislocation separations (d
obs
) range from 7.4 nm to 13.8 nm while the angle between the
Burgers vector of the perfect dislocation and the dislocation line (
q
) ranges from 13to 46. (c) Comparison of the experimental data points for the partial dislocation separation (d)
as a function of
q
and the theoretically predicted d(
q
) curves for different values of the SFE using Eq. (2). (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)
J.-K. Kim et al. / Acta Materialia 141 (2017) 444e455 447

dislocations A, B, D and E are either
a
2
½011perfect dislocations or
a
6
½121Shockley partial dislocations. Considering the fact that there
is no stacking fault contrast associated with the dislocations, the
dislocations A, B, D and E are perfect dislocations. Their Burgers
vector is
a
2
½011.
In addition, Fig. 5(h) shows the stacking fault contrast without
the contrast of dislocations near the grain boundary. However, for
other two-beam conditions, no clear stacking fault contrast is
observed in the same region. This indicates that the localized dis-
locations are most likely composed of perfect dislocations and
some Shockley partial dislocations.
While the contrast of the fault 1 is always white, independent of
the operating reflection, the stacking fault contrast of the fault 2
varies depending on the operating reflection. Since the fault 2 is
formed on the ð111Þplane, the translation vector R at the fault is
±
a
3
½111. The g$R values associated with the operating reflections
are summarized in Table 3. The g$R values are ±
1
3
for {111}-type
reflections, ±
2
3
for (020) reflection, ±1 for {131}-type reflections,
and 0 for (220) reflection. According to the g$R¼0 invisibility
criterion, the stacking faults for which the g$R dot product is an
integer such as 0 or 1 will be invisible, while the stacking faults for
which the g$R dot product is not an integer will be visible. The fault
2 is visible for two-beam conditions with g¼ð111Þ,ð111Þ, and
ð020Þ, while it is invisible for two-beam conditions with g¼(113Þ,
(113Þand (220Þ. This indicates that the fault 2 is a stacking fault
with the thickness of either 3n-1 or 3n-2 (n ¼1, 2, 3 …). On the
other hand, the stacking fault contrast of the fault 1 is always
invisible for all the operating reflections. The fault 1 is therefore a
stacking fault with the thickness of multiples of three atomic layers.
As shown in Fig. 3(a), the dislocations A, B, C and D are present at
the initial stage of deformation. Since the dislocations A, B and D are
perfect dislocations, they do not affect the stacking sequence of the
white fault. The in-situ imaging also reveals that the perfect dis-
locations A, B and D do not move and only the Shockley partial TD
moves during deformation. In addition, the dislocation C is most
likely a Shockley partial dislocationwith the same Burgers vector as
the TD, considering its glissile character. The fact that the move-
ment of the TD is not affected by the presence of the perfect dis-
locations suggests that the perfect dislocations and the TD are on
different glide planes, and the interaction between the TD and the
Fig. 3. (a) Initial snapshot of an in-situ straining TEM test of the Fe-17Mn-0.45C-1.5Al-1Si TWIP steel showing three planar defects, i.e. fault 1, fault 2 and a twin. In fault 1, four
dislocations are indicated by A, B, C and D. The tensile axis is indicated by the white double arrow. (b) A twinning dislocation (TD) generated on the right hand side of the image
approaches dislocation A, replacing the stacking fault fringes contrast by a homogeneous white contrast. (cee) The TD moves continuously towards the position of dislocation D. (f)
The entire fault 1 has a white contrast, and the dislocation E is present.
J.-K. Kim et al. / Acta Materialia 141 (2017) 444e455448

perfect dislocations is weak. This suggests that the white fault is
most likely a thin twin with the thickness more than six atomic
layers, rather than a three-layer twin. Therefore, the fault 1 in
Fig. 3(a) is expected to be bound by several Shockley partial dis-
locations, and it is most likely a stacking fault with the thickness of
3n-1 (n ¼2, 3 …). Considering the fact that the stacking fault
contrast changes from fringes to a homogeneous white contrast
after the passage of a single twinning partial TD, the Burgers vector
of the TD is same as for the other Shockley partial dislocations
which were initially present in fault 1. Therefore, the original fault
with the thickness of 3n-1 (n ¼2, 3 …)inFig. 3(a) changes to a twin
with the thickness of 3n (n ¼2, 3 …) by the movement of a single
twinning partial TD. The TD is assumed to be generated in the grain
boundary region between grain 1 and grain 2. The present obser-
vation provides evidence of the nucleation of a thin twin at a grain
boundary by the emission of suitable Shockley partial dislocations
from grain boundary region.
Fig. 6 shows a TEM two-beam BF micrograph obtained after
another in-situ straining TEM test. This provides an additional
example that the stacking fault contrast vanishes for every third
stacking fault in overlapping stacking faults. The bottommost
stacking fault shows the variation of the fault contrast by the
movement of Shockley partial dislocations trailing overlapping
stacking faults. The red brackets indicate that the contrast of
stacking fault vanishes for three or multiples of three of over-
lapping stacking faults. This contrast is due to the activation of
Shockley partial dislocations with the same Burgers vector on three
successive {111}-type twinning planes. In the vicinity of the
bottommost stacking fault, a red arrow indicates a narrow stacking
fault contrast trailing a white planar defect. It should be however
noted that the white planar defect is different from the fault 1 in
Figs. 3e5. The narrow stacking fault contrast originates from an
extended dislocation composed of a leading and a trailing Shockley
partial dislocation. The white planar defects A and B in Fig. 6 are
therefore slip-traces formed at the TEM specimen surface by the
movement of the extended dislocations.
3.3. Post-mortem TEM characterizations of deformation twinning
Further insights on deformation twinning mechanisms of TWIP
steel were gained by post-mortem TEM characterizations of the
TWIP steel deformed to an engineering strain of 3%. Fig. 7(a) shows
a planar fault indicated by the black arrow. The twinning diffraction
spots shown in the DP in Fig. 7(a) indicate that the planar fault is a
deformation twin. The [101] zone axis diffraction pattern and the
Thompson tetrahedron indicate the crystal orientation of the
investigated grain. The twin is formed on the ð111Þplane, which is
parallel to AC in the Thompson tetrahedron. In order to examine the
nature of the twin, a two-beam analysis using various diffraction
conditions near the [101] zone axis was carried out for the red
square region in Fig. 7(a), as shown in Fig. 7(b)e(g). Since the twin
is formed on the ð111Þplane, the translation vector R at the fault is
±
a
3
½111. The g$R values associated with the operating reflections
are summarized in Table 4. The g$R value for the ð111Þreflection is
an integer while the g$R values for the other reflections are not
integers. The twin does therefore not show a fault contrast for the
two-beam conditions with g ¼ð
111Þ(Fig. 7(b)) while it shows a
certain amount of a fault contrast for other reflections as shown in
Fig. 7(c) and (d). In Fig. 7(b), the contrast from the twin is therefore
not due to fault contrast, but it arises from the dislocations asso-
ciated with the twin. The g$b values associated with the operating
reflections are summarized in Table 5. As shown in Table 5, only
Frank partial dislocations are expected to be visible in the two-
beam condition for g ¼ð
111Þ. The contrast from the twin in
Fig. 7(b) is therefore due to closely spaced Frank partial
dislocations.
However, since the twin was observed in the edge-on condition
in Fig. 7, the dislocations could not be characterized in detail.
Therefore, the same region was characterized by tilting to the [112]
zone axis in order to observe an inclined view of the twin plane.
Fig. 8(a) presents a two-beam TEM BF image showing the same
region investigated in Fig. 7. It should be noted that the sample
orientation was changed during reloading of the TEM specimen.
The [112] zone axis diffraction pattern and the Thompson
Fig. 4. TEM two-beam BF micrograph showing the grain analyzed in Fig. 3 after the in-situ straining test. The [110] zone axis diffraction pattern (DP) and the Thompson tetrahedron
indicate the crystal orientation of the investigated grain.
J.-K. Kim et al. / Acta Materialia 141 (2017) 444e455 449

Fig. 5. TEM two-beam analysis of the grain analyzed in Fig. 3 using various diffraction conditions near the [110] zone axis. The analysis focuses on three regions, i.e. (1) the region
where the dislocations D and E are present, (2) the region where the dislocations A and B are present and (3) the grain boundary region, where some dislocations are localized. The
dislocations A, B, D and E are determined to be perfect dislocations with the Burges vector of a
2½011.
J.-K. Kim et al. / Acta Materialia 141 (2017) 444e455450

tetrahedron indicate the crystal orientation of the investigated
grain. The investigated twin in Fig. 7 is indicated by a dashed orange
line. The twin is found to be more inclined as compared to the
images in Fig. 7. In order to examine the nature of the dislocations
associated with the twin, a two-beam analysis using various
diffraction conditions near [112] zone axis was carried out as shown
in Fig. 8. The g$b values associated with the operating reflections
are summarized in Table 6. For the two-beam condition with
g¼ð111Þ, only the Shockley partial dislocation with the Burgers
vector
a
6
h121iis visible and the WBDF in Fig. 8(b) does therefore
not show a clear dislocation contrast. However, for other two-beam
conditions, the dislocations are visible in the WBDF images shown
in Fig. 8(c)e(e). The dislocation contrast in Fig. 8(e) is very strong.
This is in agreement with the large g$b value for a Frank partial
dislocation for the two-beam condition with g¼ð311Þ. The analysis
shown in Figs. 7 and 8 supports the fact that the straight disloca-
tions are Frank partial dislocations. The Frank partial dislocations
seem to be connected to several lattice dislocations as indicated by
the yellow arrows in Fig. 7(a).
4. Discussion
4.1. Strain hardening behavior of TWIP steel
The investigated hot-rolled and annealed Fe-17Mn-0.45C-1.5Al-
Table 2
The g$b values for each perfect, Shockley partial and Frank partial dislocation on the ð111Þplane associated with the operating reflection (g-vector) of Fig. 5.
Operating reflections Perfect Shockley partial Frank partial
a
2
½011
a
2
½110
a
2
½101
a
6
½112
a
6
½211
a
6
½121
a
3
½111
ð111Þ±1±10 ±
1
3
±
1
3
±
2
3
±
1
3
ð111Þ0±1±1±
1
3
±
2
3
±
1
3
±
1
3
ð020Þ±1±10 ±
1
3
±
1
3
±
2
3
±
2
3
ð113Þ±2±1±1±10 ±1±1
ð113Þ±1±1±2±1±10 ±1
ð220Þ±1±2±10 ±1±10
Table 3
The g$R values for stacking faults associated with the operating reflection (g-vector)
of Fig. 5.
R Operating reflections
ð111Þð111Þð020Þð113Þð113Þð220Þ
±
1
3
(111) ±
1
3
±
1
3
±
2
3
±1±10
Fig. 6. TEM two-beam BF micrograph after an in-situ straining TEM test. The bottommost stacking fault shows the fault contrast due to multiple overlapping stacking faults. The red
brackets indicate that the contrast of stacking fault vanishes for three or multiples of three overlapping stacking faults. In the vicinity of the bottommost stacking fault, there are two
white planar defects labelled A and B. The red arrow indicates the narrow stacking fault contrast trailing a white planar defect, originates from an extended dislocation. (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
J.-K. Kim et al. / Acta Materialia 141 (2017) 444e455 451

1Si TWIP steel shows a characteristic three stage strain hardening
behavior (Fig. 1(b)). This is consistent with previous reports on the
strain hardening behavior of Fe-Mn-C-Al [31], Fe-Mn-C [32] and Fe-
Mn-Al-Si [24] TWIP steels. Stage A (approximately from ε¼0to
0.08) is characterized by a steep decrease of strain hardening with
strain. In the initial deformation stage, the increasing rate of
dislocation annihilation due to dynamic recovery results in a steep
Fig. 7. (a) TEM two-beam BF micrograph showing a deformation twin near the grain boundary, indicated by the red arrow. In the [101] zone axis DP, the twin reflections are
indicated by the red arrows. The Thompson tetrahedron indicates the crystal orientation of the investigated grain. (beg) TEM two-beam analysis of the twin in (a) using various
two-beam diffraction conditions near the [101] zone axis. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Table 4
The g$R values for the twin associated with the operating reflection (g-vector) of
Fig. 7.
R Operating reflections
ð111Þð111Þð020Þð1 31Þð131Þð202Þ
±
1
3
ð111Þ±1±
1
3
±
2
3
±
1
3
±
5
3
±
4
3
Table 5
The g$b values for each perfect, Shockley partial and Frank partial dislocation on the (111) plane associated with the operating reflection (g-vector) of Fig. 7.
Operating reflections Perfect Shockley partial Frank partial
a
2
½011
a
2
½110
a
2
½101
a
6
½112
a
6
½121
a
6
½211
a
3
½111
ð111Þ000000 ±1
ð111Þ±1±10 ±
1
3
±
2
3
±
1
3
±
1
3
ð020Þ±1±10 ±
1
3
±
2
3
±
1
3
±
2
3
ð131Þ±2±20 ±
2
3
±
4
3
±
2
3
±
1
3
ð131Þ±1±10 ±
1
3
±
2
3
±
1
3
±
5
3
ð202Þ±1±10 ±
1
3
±
2
3
±
1
3
±
4
3
J.-K. Kim et al. / Acta Materialia 141 (2017) 444e455452

decrease of strain hardening with strain [33]. It should be noted
that deformation twinning is initiated in stage A. Our post-mortem
TEM analysis shows that deformation twins were not frequently
observed in the 3% deformed steel as shown in Fig. 7, while there
was no deformation twin for the 1% deformed steel. This is
consistent with the previous reports of the critical strain for the
onset of deformation twinning which ranges from ε¼0.02 to 0.04
[14,22,34]. In stage A, the volume fraction of deformation twins is
however still very small, and the twins have a limited influence on
the strain hardening behavior of the TWIP steel [32]. Stage B
(approximately from ε¼0.08 to 0.23) is characterized by an in-
crease of the strain hardening with strain. Although the origin of
this increase is still a matter of debate, stage B most likely corre-
sponds to an increased twinning activity rather than the initiation
of deformation twinning [32,35] considering the fact that initiation
of deformation twinning already takes place in stage A for the TWIP
Fig. 8. (a) TEM two-beam BF micrograph taken near the [112] zone axis. The [112] zone axis DP and the Thompson tetrahedron indicate the crystal orientation of the investigated
grain. The position of the twin analyzed in Fig. 6 is indicated by the dashed line. (bee) TEM two-beam analysis showing WBDF images of the twin. The straight dislocations
associated with the twin are most likely Frank partial dislocations.
Table 6
The g$b values for each perfect, Shockley partial and Frank partial dislocation on the (111) plane associated with the operating reflection (g-vector) of Fig. 8.
Operating reflections Perfect Shockley partial Frank partial
a
2
½011
a
2
½110
a
2
½101
a
6
½112
a
6
½121
a
6
½211
a
3
½111
ð111Þ±1±10 ±
1
3
±
2
3
±
1
3
±
1
3
ð131Þ±2±1±1±1±10 ±1
(220) ±10 ±1±
2
3
±
1
3
±
1
3
±
4
3
(311) 0±1±1±
1
3
±
1
3
±
2
3
±
5
3
J.-K. Kim et al. / Acta Materialia 141 (2017) 444e455 453

steel investigated in the present work. Stage C shows a continuous
decrease of strain hardening with strain. This is most likely due to
the gradual saturation of deformation twinning [32]. The saturation
of deformation twinning at a relatively low volume fraction of
deformation twins [14] suggests that the number of available twin
nucleation sites at grain boundaries is limited. If the twins were
nucleated at sites created by specific dislocation-dislocation in-
teractions, a continuous increase in twin volume fraction with
strain is expected.
4.2. Twin formation mechanism of TWIP steel
The present work provides the evidence of nucleation of a
deformation twin from grain boundaries based on the in-situ TEM
observation of the stacking fault with the thickness of multiples of
three atomic layers as shown in Figs. 3e5. The deformation twin-
ning observed in the present work has the following characteris-
tics: (1) only a single primary deformation system is observed, for
which stacking faults and twins are activated, (2) the dislocations
involved in the nucleation of deformation twin are perfect dislo-
cations and Shockley partial dislocations. These characteristics are
also features of the Mahajan and Chin model [8] since the model
assumes that twinning occurs on the primary slip system and that
perfect dislocations and Shockley partial dislocations are involved
in the process. In the Mahajan and Chin model, two coplanar per-
fect a/2〈110 〉-type dislocations can form a set of three Shockley
partial dislocations on three consecutive {111}-type slip planes,
resulting in a three-layer twin. However, the observed grain
boundary-related process is, strictly speaking, different from the
original Mahajan and Chin model. In the present case it is assumed
that grain boundaries provide the twin nuclei dislocation configu-
rations. The Shockley partial twinning dislocations are therefore
most likely directly emitted from grain boundaries. Therefore, the
original stacking fault with the thickness of 3n-1 (n ¼1, 2, 3 …)in
Fig. 3(a) could change into a twin with the thickness of multiples of
three atomic layers by the movement of a single twinning partial
TD emitted from the grain boundary. This process is similar to
deformation twinning in nano-crystalline materials through the
successive emission of partial dislocations from grain boundaries
[40,43,45].
Partial dislocation emission from grain boundaries has been
predicted in molecular dynamics (MD) simulations [36e39] and
experimentally observed in nano-crystalline metals [40]. Van
Swygenhoven et al. [39] demonstrated in their MD simulation that
grain boundaries containing grain boundary dislocations could
emit partial dislocations during deformation by local atomic shuf-
fling and stress-assisted free volume migration. The sources of
partial dislocations at grain boundaries include dissociated dislo-
cations on the grain boundaries [40], grain boundary ledges [41]
and triple junctions [42]. Wu and Zhu [40] reported the occur-
rence of deformation twinning from non-equilibrium grain
boundaries with a high density of extrinsic dislocations. They
proposed that the leading partial dislocation from dissociated dis-
locations on the nonequilibrium grain boundaries could slip into
the grain interior under an externally applied stress. Zhu et al. [43]
proposed that twinning partial dislocations could be multiplied at
grain boundaries by specific dislocation reactions. Beyerlein and
Tom
e[16] mentioned that grain boundary dislocations could either
dissociate or react with other defects to produce one or more
twinning partial dislocations. According to them, achieving the
critical number of partial dislocations in order to form a stable twin
nucleus depends on the sizes of the grain boundary dislocations
involved in the reactions, the spacing of the grain boundary dislo-
cations and the local stress.
It is therefore plausible that a partial dislocation-mediated
process at grain boundaries and the subsequent formation of
deformation twins could occur in the coarse-grained high Mn TWP
steel investigated in the present work. This is also consistent with
the work of Gutierrez-Urrutia and Raabe [25] who assumed that
grain boundaries were the main sites of twin nucleation in Fe-
22Mn-0.6C TWIP steel. Although an exact grain boundary mecha-
nism could not be determined, grain boundary dislocations and/or
grain boundary ledges are expected to play a role for the formation
of deformation twins under a highly localized stress. As shown in
Fig. 5(i), (p) and (x), dislocations are also observed in grain 2. The
dislocations in grain 2 are expected to be initiated at the same grain
boundary source that generated the white fault, i.e. the twin with
the thickness of multiples of three atomic layers. During the process
of deformation twinning from grain boundaries, the source of
deformation twinning is expected to generate other dislocations on
the opposite side of the grain. This process can act as a stress-
relaxation mechanism for the highly localized stress at the grain
boundary source.
4.3. Twin growth mechanism of TWIP steel
Deformation twins thicken by the propagation of Shockley
partial dislocations on twin boundaries. Recently, Casillas et al. [44]
observed both extrinsic and intrinsic stacking faults on twin
boundaries of Fe-24Mn-3Al-2Si-1Ni-0.06C TWIP steel. Figs. 7 and 8
show a high density of Frank partial dislocations inside the defor-
mation twin. This is similar to the observations of Idrissi et al.
[22,46] who also reported a high density of Frank partial disloca-
tions inside deformation twins of Fe-20Mn-1.2C steel. They pro-
posed that the additional Frank partial dislocations were due to the
reaction of twinning Shockley partial dislocations with perfect
dislocations in the matrix by following dislocation reaction:
a
6h211iþa
2½011/a
3h111i(3)
The Frank partial dislocations seem to be connected to several
lattice dislocations as indicated by the yellow arrows in Fig. 7(a),
indicating the possible occurrence of dislocation interaction of Eq.
(3). Shockley partial twinning dislocations that move further in the
twinning plane must overcome the Frank partial dislocations since
the Frank partial dislocations are located at the twin-matrix inter-
face [46]. Thus, the presence of Frank partial dislocations in the
twinning plane could hinder the movement of Shockley partial
dislocations, thereby contributing to the high work hardening
observed for TWIP steel. The presence of Frank partial dislocations
inside deformation twins also explains both very fine twin thick-
ness and stability of twins in C-alloyed TWIP steels [46].
5. Conclusions
In this study we investigated the deformation twinning mech-
anism that operates in Fe-17Mn-0.45C-1.5Al-1Si TWIP steel by
means of in-situ TEM deformation tests and post-mortem TEM
characterizations. We draw the following conclusions:
1. The Fe-17Mn-0.45C-1.5Al-1Si TWIP steel shows a characteristic
three stage strain hardening behavior. The SFE of the investi-
gated TWIP steel was measured to be 12 ±5 mJ/m
2
using WBDF
TEM.
2. The in-situ TEM observation shows the formation of a stacking
fault with the thickness of multiples of three atomic layers by
Shockley partial dislocations generated in a grain boundary. This
indicates that deformation twins can be nucleated in the grain
boundary region. During the process of deformation twinning,
J.-K. Kim et al. / Acta Materialia 141 (2017) 444e455454

the source of deformation twins generates other dislocations on
the opposite side of the grain. This process can act as a stress-
relaxation mechanism for highly localized stresses at the grain
boundary.
3. The nucleation of deformation twin has the following charac-
teristics: (1) only a single primary deformation system is
observed, for which stacking faults and twins are activated, (2)
the dislocations involved in the nucleation of deformation twin
are perfect dislocations and Shockley partial dislocations. The
observed twinning process is similar to deformation twinning in
nano-crystalline materials through the successive emission of
partial dislocations from grain boundaries.
4. A high density of Frank partial dislocations was observed in the
deformation twin. The presence of Frank partial dislocations on
the twinning plane hinders the movement of Shockley partial
dislocation and thereby contributes to the high work hardening
of TWIP steel.
Acknowledgements
The authors gratefully acknowledge the support of POSCO
Technical Research Laboratories, Gwangyang, Korea. The authors
also wish to thank Mr. Hojun Gwon (GIFT, POSTECH) for his tech-
nical support and expertise.
References
[1] J.W. Christian, S. Mahajan, Deformation twinning, Prog. Mater. Sci. 39 (1995)
1e157.
[2] M. Niewczas, Dislocations and twinning in face centred cubic crystals, in:
F. Nabarro, J. Hirth (Eds.), Dislocations in Solids, Elsevier Science, 2007,
pp. 263e364.
[3] Y.T. Zhu, X.Z. Liao, X.L. Wu, Deformation twinning in nanocrystalline mate-
rials, Prog. Mater. Sci. 57 (2012) 1e62.
[4] X.L. Wu, X.Z. Liao, S.G. Srinivasan, F. Zhou, E.J. Lavernia, R.Z. Valiev, Y.T. Zhu,
New deformation twinning mechanism generates zero macroscopic strain in
nanocrystalline metals, Phys. Rev. Lett. 100 (2008) 095701.
[5] B.Q. Li, B. Li, Y.B. Wang, M.L. Sui, E. Ma, Twinning mechanism via synchronized
activation of partial dislocations in face-centered-cubic materials, Scr. Mater.
64 (2011) 852e855.
[6] I.J. Beyerlein, X. Zhang, A. Misra, Growth twins and deformation twins in
metals, Annu. Rev. Mater. Res. 44 (2014) 329e363.
[7] S. Mahajan, Critique of mechanisms of formation of deformation, annealing
and growth twins: face-centered cubic metals and alloys, Scr. Mater. 68
(2013) 95e99.
[8] S. Mahajan, G.Y. Chin, Formation of deformation twins in f.c.c crystals, Acta
Metall. 21 (1973) 1353e1363.
[9] J.A. Venables, Deformation twinning in face-centered cubic metals, Philos.
Mag. 6 (1961) 379e396.
[10] J.B. Cohen, J. Weertman, A dislocation model for twinning in f.c.c. metals, Acta
Metall. 11 (1963) 996e998.
[11] H. Fujita, T. Mori, A formation mechanism of mechanical twins in F.C.C. metals,
Scr. Metall. 9 (1975) 631e636.
[12] S. Miura, J. Takamura, N. Narita, Trans. Jpn. Inst. Met. 9 (1968) 555.
[13] G.T. Gray III, Deformation twinning in Al-4.8 wt% Mg, Acta Metall. 36 (1988)
1745e1754.
[14] J. Kim, Y. Estrin, B.C. De Cooman, Application of a dislocation density-based
constitutive model to Al-alloyed TWIP steel, Metall. Mater. Trans. A 44
(2013) 4168e4182.
[15] J.-K. Kim, B.C. De Cooman, Stacking fault energy and deformation mechanisms
in Fe-xMn-0.6C-yAl TWIP steel, Mater. Sci. Eng. A 676 (2016) 216e231.
[16] I.J. Beyerlein, C.N. Tome, A probabilistic twin nucleation model for HCP
polycrystalline metals, Proc. R. Soc. A 466 (2010) 2517e2544.
[17] B.C. De Cooman, K.G. Chin, J. Kim, High Mn TWIP steels for automotive ap-
plications, in: M. Chiaberge (Ed.), New Trends and Developments in Auto-
motive System Engineering, InTech, Rijeka, 2011, pp. 101e128.
[18] O. Bouaziz, S. Allain, C.P. Scott, P. Cugy, D. Barbier, High manganese austenitic
twinning induced plasticity steels: a review of the microstructure properties
relationships, Curr. Opin. Solid State Mater. Sci. 15 (2011) 141e168.
[19] O. Bouaziz, N. Guelton, Modeling of TWIP effect on work-hardening, Mater.
Sci. Eng. A 319e321 (2001) 246e249.
[20] I. Karaman, H. Sehitoglu, K. Gall, Y.I. Chumlyakov, H.J. Maier, Deformation of
single crystal hadfield steel by twinning and slip, Acta Mater. 48 (2000)
1345e1359.
[21] L. Bracke, L. Kestens, J. Penning, Direct observation of the twinning mecha-
nism in an austenitic FeeMneC steel, Scr. Mater. 61 (2009) 220e222.
[22] H. Idrissi, K. Renard, L. Ryelandt, D. Schryvers, P.J. Jacques, On the mechanism
of twin formation in FeeMneC TWIP steels, Acta Mater. 58 (2010)
2464e2476.
[23] J.B. Liu, X.H. Liu, W. Liu, Y.W. Zeng, K.Y. Shu, Transmission electron microscopy
observation of a deformation twin in TWIP steel by anex situtensile test,
Philos. Mag. 91 (2011) 4033e4044.
[24] B. Mahato, S.K. Shee, T. Sahu, S. Ghosh Chowdhury, P. Sahu, D.A. Porter,
L.P. Karjalainen, An effective stacking fault energy viewpoint on the formation
of extended defects and their contribution to strain hardening in a
FeeMneSieAl twinning-induced plasticity steel, Acta Mater. 86 (2015)
69e79.
[25] I. Gutierrez-Urrutia, D. Raabe, Grain size effect on strain hardening in
twinning-induced plasticity steels, Scr. Mater. 66 (2012) 992e996.
[26] D.B. Williams, C.B. Carter, Transmission Electron Microscopy, Plenum Press,
New York, 1996.
[27] E. Aerts, P. Delavignette, R. Siems, S. Amelinckx, Stacking fault energy in sil-
icon, J. Appl. Phys. 33 (1962) 3078e3080.
[28] J. Kim, S.-J. Lee, B.C. De Cooman, Effect of Al on the stacking fault energy of
Fee18Mne0.6C twinning-induced plasticity, Scr. Mater. 65 (2011) 363e366.
[29] K. Jeong, J.-E. Jin, Y.-S. Jung, S. Kang, Y.-K. Lee, The effects of Si on the me-
chanical twinning and strain hardening of Fee18Mne0.6C twinning-induced
plasticity steel, Acta Mater. 61 (2013) 3399e3410.
[30] D.T. Pierce, J.A. Jim
enez, J. Bentley, D. Raabe, C. Oskay, J.E. Wittig, The influence
of manganese content on the stacking fault and austenite/ε-martensite
interfacial energies in FeeMne(AleSi) steels investigated by experiment and
theory, Acta Mater. 68 (2014) 238e253.
[31] J.E. Jin, Y.K. Lee, Effects of Al on microstructure and tensile properties of C-
bearing high Mn TWIP steel, Acta Mater. 60 (2012) 1680e1688.
[32] K. Renard, P.J. Jacques, On the relationship between work hardening and
twinning rate in TWIP steels, Mater. Sci. Eng. A 542 (2012) 8e14.
[33] U.F. Kocks, H. Mecking, Physics and phenomenology of strain hardening : the
FCC case, Prog. Mater. Sci. 48 (2003) 171e273.
[34] J. Kim, Y. Estrin, H. Beladi, I. Timokhina, K.G. Chin, S.K. Kim, B.C. De Cooman,
Constitutive modeling of the tensile behavior of Al-TWIP steel, Metall. Mater.
Trans. A 43 (2012) 479e490.
[35] K.M. Rahman, V.A. Vorontsov, D. Dye, The effect of grain size on the twin
initiation stress in a TWIP steel, Acta Mater. 89 (2015) 247e257.
[36] J. Shiotz, F.D. Di Tolla, K.W. Jacobsen, Softening of nanocrystalline metals at
very small grain sizes, Nature 391 (1998) 561e563.
[37] V. Yamakov, D. Wolf, S.R. Phillpot, A.K. Mukherjee, H. Gleiter, Dislocation
processes in the deformation of nanocrystalline aluminium by molecular-
dynamics simulation, Nat. Mater. 1 (2002) 45e48.
[38] H. Van Swygenhoven, P.M. Derlet, A.G. Froseth, Stacking fault energies and
slip in nanocrystalline metals, Nat. Mater. 3 (2004) 399e403.
[39] H. Van Swygenhoven, P.M. Derlet, A. Hasnaoui, Atomic mechanism for
dislocation emission from nanosized grain boundaries, Phys. Rev. B 66 (2002).
[40] X.L. Wu, Y.T. Zhu, Partial-dislocation-mediated processes in nanocrystalline Ni
with nonequilibrium grain boundaries, Appl. Phys. Lett. 89 (2006) 031922.
[41] A. Hunter, I.J. Beyerlein, Stacking fault emission from grain boundaries: ma-
terial dependencies and grain size effects, Mater. Sci. Eng. A 600 (2014)
200e210.
[42] M.Y. Gutkin, I.A. Ovid'ko, N.V. Skiba, Emission of partial dislocations from
triple junctions of grain boundaries in nanocrystalline materials, J. Phys. D.
Appl. Phys. 38 (2005) 3921e3925.
[43] Y.T. Zhu, X.L. Wu, X.Z. Liao, J. Narayan, S.N. Mathaudhu, L.J. Kecsk
es, Twinning
partial multiplication at grain boundary in nanocrystalline fcc metals, Appl.
Phys. Lett. 95 (2009) 031909.
[44] G. Casillas, A.A. Gazder, E.V. Pereloma, A.A. Saleh, Evidencing extrinsic stack-
ing faults in twinning-induced plasticity steel, Mater. Charact. 123 (2017)
275e281.
[45] I.A. Ovid'ko, N.V. Skiba, Generation of nanoscale deformation twins at locally
distorted grain boundaries in nanomaterials, Int. J. Plast. 62 (2014) 50e71.
[46] H. Idrissi, K. Renard, D. Schryvers, P.J. Jacques, On the relationship between the
twin internal structure and the work-hardening rate of TWIP steels, Scr.
Mater. 63 (2010) 961e964.
J.-K. Kim et al. / Acta Materialia 141 (2017) 444e455 455