CV 08

Materials Transactions, Vol. 50, No. 1 (2009) p.
Special Issue on Severe Plastic Deformation for Production

of Ultrane Structures and Unusual Mechanical Properties:
Understanding Mechanisms
PREFACE
There is an ongoing national research project in Japan in the priority area of a Grant in Aid for Scientic
Research from the Ministry of Education, Culture, Sports, Science and Technology (so-called Monkasho
Grant), which started in September 2006 under the title of Giant Straining Process for Advanced
Materials Containing Ultra-High Density Lattice Defects. The present special issue includes many
contributions from the members of the project. It also includes several papers which will be presented in
The International Symposium on Giant Straining Process for Advanced Materials (GSAM2008) held in
Fukuoka from November 21 to 24, 2008 (http://www.congre.co.jp/gsam2008). Thus, this special issue
provides an opportunity to collect formal reports for the national research project in the last two years and
also to publish papers as a part of proceedings for the symposium of GSAM2008.
It is well known that signicant renement of the microstructure is achieved in metallic materials using
the processes of severe plastic deformation (SPD) such as equal-channel angular pressing (ECAP),
accumulative roll bonding (ARB), high-pressure torsion (HPT) and multi-directional forging (MDF). High
strength and high ductility may be attained simultaneously when the materials are processed by SPD. This
special issue focuses on understanding the mechanisms for the microstructural renement and the
simultaneous achievement of high strength and high ductility. Several papers of this issue are concerned
with theoretical approach including simulation for the microstructural renement. In particular, direct
comparison of experimental observation and analysis with theoretical calculation and simulation is raised
in this issue. We are sure that this issue is the rst attempt to edit such a comparison in the study of SPD.
We hope that this special issue serves to promote research activity in the led of materials science and
technology. Although current topics are mostly based on fundamental research, they will be more practical
issues in the near future as we need to develop materials for supporting our environment. The SPD process
should be a key technology for such a development.
Finally, we are very grateful to all authors who contributed to providing latest information from their
research work. We also thank very much reviewers for spending their valuable time for many useful
comments and suggestions.
November 14, 2008
Guest Editors
Tadanobu Inoue1
Yoshikazu Todaka2
Zenji Horita3
1
National Institute for Materials Science, Tsukuba 305-0047, Japan

Toyohashi University of Technology, Toyohashi 441-8580, Japan
3
Kyushu University, Fukuoka 819-0395, Japan
2
Materials Transactions, Vol. 50, No. 1 (2009) pp. 2 to 10

Special Issue on Severe Plastic Deformation for Production of Ultrane Structures and Unusual Mechanical Properties: Understanding Mechanisms
#2009 The Japan Institute of Metals
Eect of Extrinsic Grain Boundary Dislocations on Mechanical Properties

of Ultrane-Grained Metals by Molecular Dynamics Simulations
Tomotsugu Shimokawa1 , Tomoyuki Hiramoto2; * , Toshiyasu Kinari1 and Sukenori Shintaku1
1
Division of Innovative Technology and Science, Graduate School of Natural Science and Technology,
Kanazawa University, Kanazawa 920-1192, Japan
2
Division of Mechanical Science and Engineering, Graduate School of Natural Science and Technology,
The eect of extrinsic grain boundary dislocations (EGBDs) in nonequilibrium grain boundaries on the mechanical properties of ultranegrained metals is investigated by molecular dynamics simulations. Aluminum bicrystal models containing cracks and EGBDs impinged from the
crack tips are prepared. First, the dependence of the local grain boundary structure on the accommodation mechanism of EGBDs, and on its
stress eld is studied. Then, the shielding eect of EGBDs on the emissions of dislocations from crack tips is investigated, and the eect of
nonequilibrium grain boundaries on the intragranular deformation is discussed. Finally, to investigate the relationship between EGBDs and
intergranular deformations, shear loading is applied to the bicrystal models. It is found that extrinsic grain boundaries function as the
intergranular deformation source, and the Burgers vector components of the EGBDs lead to anisotropic grain boundary sliding.
[doi:10.2320/matertrans.MD200810]
(Received July 15, 2008; Accepted August 29, 2008; Published October 22, 2008)
Keywords: extrinsic grain boundary dislocation, grain boundary sliding, grain boundary migration, nonequilibrium grain boundary, shielding
eect, grain boundary structure, molecular dynamics
1.
Introduction
Grain renement is one of the techniques used to improve

the mechanical properties of polycrystalline metals. Recently, severe plastic deformation (SPD) processes1) such as
equal-channel angular pressing (ECAP),2) accumulative roll
bonding (ARB),3) and high-pressure torsion (HPT)4) processes, have attracted the attention of researchers because
grain sizes of several hundred nanometers can be obtained
using these processes. Ultrane-grained (UFG) metals
produced by SPD processes show high strength consistent
with the Hall-Petch relationship, and they also exhibit
unusual mechanical properties compared to coarse-grained
metals, for instance, a sudden decrease in uniform elongation,5,6) strain-rate sensitivity of ow stress,7,8) and improved
low-temperature toughness.9) However, the detailed mechanisms of these unusual mechanical properties have not been
fully understood yet.
In the case of coarse-grained metals, dislocation structures
are formed in the grains, and a large number of interactions
between the lattice dislocations take place. The mechanical
properties of coarse-grained metals are governed mainly by
these interactions. However, in the case of UFG metals,
which have grain sizes below several hundred nanometers,
dislocation structures in the grains do not form easily. Hence,
obstacles in the grains to mobile dislocations decrease with
the grain size. Comparatively, the proportion of grain
boundary regions in UFG metals is larger than that in
coarse-grained metals. Therefore, the interaction between
lattice dislocations and grain boundaries is a very important
phenomenon that has to be studied to elucidate the unusual
mechanical properties.10)
It has also been reported that the grain boundaries of
UFG metals produced by SPD processes consist of not
*Graduate
Student, Kanazawa University. Present address: Aisin Aw

Industries Co. Ltd., Echizen 915-8520, Japan
only geometrically necessary defect structures, i.e., intrinsic

grain boundary dislocations (IGBDs),11) but also many
EGBDs,1214) which do not contribute to the grain boundary
misorientation angle. These grain boundaries containing
EGBDs are called nonequilibrium grain boundaries, and they
can transform into an equilibrium grain boundary after
annealing. Interesting mechanical properties of hardening by
annealing and softening by deformation in nanostructured
metals have been reported.15) The result implies a longerrange stress eld of nonequilibrium grain boundaries due to
the EGBDs than equilibrium grain boundaries and the
Burgers vector components of the EGBDs in nonequilibrium
grain boundaries could aect the mechanical properties of
UFG metals.16) Therefore, to elucidate the unusual mechanical properties of UFG metals, it is important to investigate
the mechanical behavior of nonequilibrium grain boundaries.
In this study, we investigate the eects of nonequilibrium
grain boundaries containing EGBDs on the mechanical
properties of UFG metals. For this, we use molecular
dynamics simulations17) because they can directly express
the interactions between defect structures with atomic
resolution.18) Aluminum bicrystal models containing cracks
are developed, and the following phenomena are studied by
molecular dynamics simulations: the accommodation mechanism of impinged lattice dislocations at the grain boundaries, the shear-stress eld of nonequilibrium grain boundaries
and its shielding eect on the emissions of dislocations from
the crack tip, and the eect of EGBDs on intergranular
deformations. Finally, the eect of nonequilibrium grain
boundaries on the mechanical properties of UFG metals is
discussed on the basis of the obtained results.
2.
Computational Setup
2.1 Analysis model

Figure 1 shows the schematic of the analysis model used
in this study. The interatomic potential for aluminum can be
Eect of Extrinsic Grain Boundary Dislocations on Mechanical Properties of Ultrane-Grained Metals by Molecular Dynamics Simulations
grain
z [111]
[112]
UA
UB
y
x [110]
LI
Lz
Lc
grain
bL
n
grain
bL
sit
sit
sit
Lx
z
y
bL
Fig. 1 Analysis model of h112i5 bicrystal containing cracks. The grain boundary misorientation angle 101:5 . A grain boundary
consists of three periodic structural units: UA, UA, UB, UA, . The Burgers vector bL 1=21 10 of a lattice dislocation can be
resolved into two components bnL and btL , which are normal and parallel to the grain boundary, respectively. bnL 1=101 52 and
btL 1=52 01 .
obtained by using the embedded-atom method proposed by

Mishin et al.19) The analysis model consists of two grains,
grains I and II, and the x0 -, y0 -, and z0 -axes of grain I are along
11 0, 112 , and 111, respectively. The values of Lx , Ly ,
and Lz are approximately 91, 1.5, and 91 nm, respectively.
The total number of atoms is approximately 0.7 millions. One
crack of length Lc 20 nm is introduced into the center
region in grain I, as shown in Fig. 1. The periodic boundary
conditions are adopted in all directions; therefore, an innite
number of cracks align in grain I. Two dierent values of LI ,
45 and 73 nm, are adopted in order to control the distance
between the crack tip and the grain boundaries.
The interface structure between grain I and II is set to be
the h112i5 grain boundary. Therefore, is equal to
101.5 .20) The detailed atomic structure of the grain boundary
after relaxation is shown in the right image of Fig. 1. Six
dierent symbols are used to represent the {112} atomic
planes. The CSL unit cells are indicated by the solid squares.
The grain boundary consists of three periodic structural units:
UA, UA, UB, UA, UA, UB, . Hence, there are three
dierent grain boundary sites where lattice dislocations are
impinged.
The relaxation simulations of the analysis models are
carried out for 20 ps for normal components of stress, x and
z (0 GPa), using the Parrinello-Rahman algorithm.21) The
temperature of the analysis models is controlled by applying
the simple velocity-scaling algorithm. The time increment for
the numerical integration is set to 2 fs, and the velocity Verlet
algorithm is used.
2.2 Tensile loading-unloading simulations
To transform equilibrium grain boundaries into nonequilibrium grain boundaries, lattice dislocations are generated
from the crack tip under tensile loading in the z-direction at
100 K, and the dislocations are impinged on the h112i5
grain boundaries at dierent sites by controlling the crack
position. When the tensile strain "t reaches 0.02 or 0.025, one
or two pairs of dislocations are formed from the crack tips,

respectively. Then, unloading is carried out until zz becomes
zero. After unloading, using the Parrinello-Rahman algorithm, each model is relaxed for 10 ps at 100 K, maintaining
x and z at zero. The strain rates of loading and unloading,
"_z , are set to 8 108 s1 . To investigate the dissociation of
EGBDs, the model containing one pair of EGBDs at site 1 is
annealed at 800 K for 500 ps.
2.3 Shear loading-unloading simulations
In order to examine the eect of EGBDs on the grain
boundary motion, shear loading zx 800 MPa is applied
for 90 ps at 100 K to the analysis models with equilibrium
grain boundaries and nonequilibrium grain boundaries. Here,
nonequilibrium grain boundaries contain one or two pairs of
EGBDs. When the loading time reaches 90 ps, unloading is
carried out and the models are relaxed for 20 ps at 100 K. To
investigate the change in the grain boundary structure after
the grain boundary sliding, the model containing one pair of
EGBDs at site 1 is annealed at 800 K for 200 ps.
3.
3.1
Results
Dissociation of impinged lattice dislocations in grain

boundaries
The accommodation mechanism of the core structures of
the lattice dislocations impinged at a grain boundary could be
aected by the local grain boundary structure where the
dislocations are impinged.22) Hence, the energy per unit
length of a lattice dislocation could change when it enters a
grain boundary because the impinged dislocation dissociates
depending on the local grain boundary structure. To investigate the eect of the local grain boundary structures on the
energy change in the impinged dislocations, tensile loadingunloading simulations are carried out. In the simulations, one
dislocation is formed from each crack tip, and the dislocations are impinged at the grain boundaries at dierent sites.
T. Shimokawa, T. Hiramoto, T. Kinari and S. Shintaku
t,1
bL
b n,1
L
bL
zx [MPa]
500
z
x
500
(a) Impinged dislocation at site 1
(b) Structural change at site 1

t,2a
bL
t,2b
bL
b
e,2a
b 2L
b n,2
L
e,2b
t,2c
bL
(c) Impinged dislocation at site 2
(d) Structural change at site 2
t,3
bL
b n,3
L
(e) Impinged dislocation at site 3
b 3L
(f) Structural change at site 3
Fig. 2 Shear stress distributions of z0 x0 of nonequilibrium grain boundaries each containing one EGBD at dierent slip planes, and local
structural change due to impingement of lattice dislocations under tensile loading.
Figure 2 shows the shear stress distributions z0 x0 23) and

structural changes around the impinged dislocations at three
dierent sites. Here, x0 and z0 correspond to the directions
parallel and normal to the direction of the slip plane of the
impinged dislocations, respectively. It can be conrmed that
the lattice dislocations are trapped when they enter the

equilibrium h112i5 grain boundary at any site and thus do
not disappear. We can also conrm that the shear stress elds
far from the trapping sites precisely correspond to those at
isolated lattice edge dislocations. Therefore, when it is
Table 1 Energy EGB of dislocations impinged grain boundaries at dierent
sites. Energy of a lattice dislocation EL 1:41 nJ/m.
EGB /nJm1
site 1
site 2
site 3
1.07
1.46
1.31
dicult for the impinged dislocations to transform into

IGBDs in the case of low temperature, the shear stress elds
due to EGBDs may aect the intragranular deformation.16)
We compare the energy of a lattice dislocation, EL , and the
energy of the impinged lattice dislocation in a grain
boundary, EGB . First, EL is calculated using a single-crystal
model with one edge dislocation dipole under the periodic
boundary condition. The dislocation density in the singlecrystal model is the same as that in the h112i5 grain
boundary models containing one EGBD at each grain
boundary. EL is found to be equal to 1.41 nJ/m. Using the
linear dislocation theory,24) the energy of a lattice edge
dislocation not including the core contribution can be
estimated at 1.14 nJ/m;25) this energy is of the same order
as the energy obtained using the atomic model. Then, EGB is
estimated from the energy change in the grain boundary
model before and after the tensile loading-unloading simulations. Therefore, EGB can be calculated by the following
equation.
EGB
ENEGB EEGB
2Ly
Here, EEGB is the energy of the model with equilibrium grain

boundaries before loading, ENEGB is the energy of the model
with nonequilibrium grain boundaries after loading, and Ly is
the dimension of the model in the dislocation line direction.
Table 1 shows EGB for each grain boundary site. Even though
the dislocations exist at dierent sites, such as a crystal lattice
or a grain boundary, the dierences between EL and EGB are
not large. However, it can be conrmed that EGB shows
dierent values for dierent sites. In particular, at site 2, EGB
is higher than EL . The energy dierence could be due to the
dierence in the accommodation mechanism of impinged
dislocations.
Figures 2(b), 2(d), and 2(f) shows the local structural
change around the trapping sites. At sites 1 and 3, a step in the
vicinity of the trapping sites at the interface can be observed.
The steps are formed by the 1=52 01 DSC dislocation26)
indicated by the broken arrows in Fig. 1, which corresponds
to the component btL of the impinged lattice dislocation with
bL 1=21 10, as shown in Fig. 1. Therefore, the impinged
dislocations at site i, biL , should decompose into at least two
n;i
grain boundary dislocations, bt;i
L and bL , as shown by the blue
L
dislocations. Generally, E of the dislocation is proportional
to the square of the magnitude of the Burgers vector. Hence,
at sites 1 and 3, EGB is lower than EL . It should be noted that
the decrease in the energy of the dislocations is also caused
by the dierence in the shear moduli of the grain boundary
region and the crystal structure. On the other hand, at site 2,
double steps due to the grain boundary dislocations, bt,L 2a and
bt,L 2b indicated in blue, can be observed. Moreover, a grain
boundary dislocation dipole, be, 2a and be, 2b indicated in red,
is generated. Hence, at site 2, EGB becomes higher than EL
probably due to the extra components of the grain boundary
dislocations generated in the dipole. However, even though

the dissociation of the impinged dislocations occurs at
dierent sites, the long-range shear stress eld at each site is
not strongly aected by the accommodation mechanism,
which is also shown in the next section, because of the short
range of the dissociation area.
To investigate the eect of temperature on the dissociation
of an impinged dislocation, nonequilibrium grain boundaries
each containing one EGBD at site 1 are annealed at 800 K for
500 ps. Figures 3(a) and 3(b) shows the stress distributions of
the rst invariant of stress I1 x z of the nonequilibrium
grain boundaries before and after annealing, respectively.
When the dissociation of an impinged dislocation occurs, it
is not easy to identify their slip plane. Therefore, I1 , which
does not depend on the selection of rectangular coordinate
systems, is adopted instead for z0 x0 . From the stress
distribution shown in Fig. 3(a), it can be assumed that the
impinged dislocations do not decompose into some dislocations. Contrary to this assumption, the impinged dislocations
dissociate into at least two grain boundary dislocations in the
narrow region, as shown in Fig. 2(b). A dierent stress
distribution is observed after annealing, as shown in
Fig. 3(b). From the stress eld I1 , it is observed that the
grain boundary dislocation with btL parallel to the grain
boundary gets separated from the grain boundary dislocation
with bnL perpendicular to the grain boundary. The grain
boundary dislocations with btL slide along the grain boundary
during annealing. Therefore, we call these grain boundary
dislocations btL and bnL as glissile and sessile grain boundary
dislocations, respectively. In order to dissociate the sessile
grain boundary dislocations along the grain boundary, the
climb process is required, but it is dicult to simulate the
dissociation phenomenon within the treatable time scale in
molecular dynamics simulations.
We have also considered a case of asymmetrical grain
boundaries produced by rotating grain II around the h112i
axis; in this cases, is equal to 78.5 . Figure 3(c) shows I1
obtained after annealing the asymmetrical grain boundaries
at 800 K. Even though the asymmetrical grain boundaries
containing one impinged dislocation are subjected to
annealing, no remarkable dissociation similar to that in the
symmetrical grain boundaries is observed. This absence
could be due to the high resistance to the sliding of glissile
grain boundary dislocations in such asymmetrical structures.
Consequently, it can be conrmed that the symmetry of a
grain boundary structure could also aects the dissociation of
the impinged dislocation into glissile and sessile grain
boundary dislocations. The population of symmetrical grain
boundaries could be one of the important factors controlling
the mechanical properties of UFG metals.
3.2
Shielding eect of impinged dislocations on dislocation emissions from cracks

As discussed in previous sections, nonequilibrium grain
boundaries show longer-range stress elds than equilibrium
grain boundaries due to EGBDs, though the stress eld
becomes smaller when EGBDs transform into IGBDs.
Hence, these stress elds may aect the motion of mobile
lattice dislocations at the vicinity of grain boundaries.
Furthermore, the impinged dislocations generated from a
(a)
I1
(c)
(b)
[MPa]
1000
0
1000
Fig. 3 Stress distributions of rst invariant of stress I1 of nonequilibrium grain boundaries each containing one EGBD at site 1. (a) Before
and (b) after annealing at 800 K. Dissociation of EGBD is observed after annealing. (c) Asymmetrical grain boundaries with the same
condition as in (b). No remarkable dissociation of EGBD is observed.
Table 2 Average shear stress around crack tip. z0 x0 z0 x0 ;S1 z0 x0 ;S2 =2.
( indicates just before dislocation formation from the crack tip under
loading. indicates after annealing at 800 K for 500 ps.)
0, EGBD
LI /nm
45
Slip plane
z0 x0 /MPa
1, EGBD
45
site 1
26 (878)
138 (124)
45
2, EGBDs
45
73
site 2 site 3 site 1

140
138
60
45
site 1
249
dislocation source may exhibit a shielding eect on the

source due to their back stress.27) To investigate the eect of
EGBDs on the stress elds around the crack tips in our
analysis models, the average shear stresses z0 x0 in regions
within 5 nm from the crack tips S1 and S2 around the left and
right crack tips (Fig. 4) are calculated. Table 2 shows each
value of z0 x0 . Figure 4 shows the shear stress distributions
around the cracks. In the case of equilibrium grain boundaries, z0 x0 is equal to 26 MPa under no applied loading. Under
tensile loading, z0 x0 becomes equal to 878 MPa just before
the formation of the rst dislocation from the crack tip. In the
case of nonequilibrium grain boundaries, each contains one
impinged dislocation at dierent sites, z0 x0 is approximately
140 MPa in the relaxed state. Therefore, it can be conrmed
that the EGBDs produce a shielding eect on the crack tip. It
should be noted that the dependence of the grain boundary
structures where the dislocations are impinged, shown in
Figs. 2(b), 2(d), and 2(f), on the value of z0 x0 is small in this
simulation. In the case of nonequilibrium grain boundaries,
each containing two EGBDs at site 1, z0 x0 is equal to
249 MPa, the dislocation shielding eect is stronger than in
the previous case.
When the distance between the crack tips and the grain
boundaries becomes larger, z0 x0 dramatically decreases to
60 MPa, as shown in Fig. 4(c). The dependence of the EGBD
shielding eect on the distance between the crack tips and the
grain boundaries can be conrmed from Fig. 5, which shows
the stress-strain curves of the bicrystal models with LI 45
and 73 nm under tensile loading at "_t 1 108 s1 at 100 K.
The circles and triangles indicate the formation of lattice
dislocations from S1 and S2 , respectively. It is clearly
observed that after the formation of the rst dislocation, the
rate of increase in the tensile stress in the model with

LI 45 nm is higher than that in the model with LI 73 nm.
Therefore, the absolute back stress value of EGBDs around
the crack tips, which depend on the distance between the
crack tips and the grain boundaries, prevents the formation of
successive lattice dislocations at the same tensile stress at
which the previous dislocation has been generated.
Figure 4(d) shows the shear stress distribution of the
nonequilibrium grain boundaries each containing one EGBD
after annealing at 800 K for 500 ps. As shown in Fig. 3(b), the
impinged dislocation dissociates into two grain boundary
dislocations after the annealing; hence, z0 x0 reduces after
annealing. Consequently, the EGBD shielding eect on the
crack tips is strongly aected by the temperature and the
applied loading that controls the dissociation of the EGBDs.
More detailed discussions about the EGBD shielding eect
can be found elsewhere.28)
3.3
Eect of impinged dislocations on grain boundary

sliding
It is essential to study the intergranular deformation in
order to determine the mechanical properties of UFG metals,
especially nanocrystalline metals that contain a large number
of grain boundaries. To investigate the eect of EGBDs on
the intergranular deformation, a shear stress zx 800 MPa
is applied to bicrystal models containing 0, 1, and 2 EGBDs
in each grain boundary at 100 K. At this temperature, it is not
easy for atoms forming a grain boundary to diuse in the
interface by a thermal activation process. Therefore, interface
sliding is expected to take place by the motion of the glissile
dislocations at the grain boundaries. Figure 6 shows the
change in equivalent strain distributions at an interval of 5 ps
in the upper nonequilibrium grain boundaries each containing
one EGBD under the applied shear stress of 800 MPa. The
dashed line indicates the position where the lattice dislocation was impinged under tensile loading. It can be observed
that atoms with large equivalent strains are generated from
the site of the impinged dislocation, and the distributions can
be divided into two parts on the basis of the direction of the
grain boundary sliding under shear loading. Figure 7 shows
detailed atomic arrangement in the white boxes A and B
shown in Fig. 6. The atomic arrangement before shear
zx [MPa]
500
0
500
S1
10 nm
S2
(a) 1 EGBD
(b) 2 EGBDs
(c) 1 EGBD
(d) 1 EGBD after annealing
Fig. 4 Stress distributions of z0 x0 after tensile deformation at site 1, which shows shielding eect of EGBDs on the crack tip. (a), (b), and
(d) LI 45 nm and (c) LI 73 nm.
loading is shown in Fig. 2(b). The origin of grain boundary

sliding from the impinged dislocation can be observed in
Fig. 7(a), and it propagates in the D2 direction, as shown in
Fig. 7(b). The atomic arrangements after grain boundary
sliding shows us that the components of grain boundary
sliding correspond to the DSC lattice, as shown by the arrows
inside the ellipses in Fig. 1, and grain boundary migration
accompanies grain boundary sliding. Hence, a coupling
motion29) takes place in the grain boundary under shear
loading. In case of nonequilibrium grain boundaries each
containing two EGBDs, the intergranular deformation also
occurs from the site of the impinged dislocations. On the
other hand, in the case of equilibrium grain boundaries
containing no impinged dislocation, atoms with large
equivalent strain are generated from the dierent sites from
the nonequilibrium grain boundaries. Figure 8 shows the

generation of the coupling motion under shear loading of
800 MPa at 100 K. At 16.5 ps, two parts corresponding to the
coupling motion are generated, and each edge propagates in
the opposite direction, as shown in Fig. 8(b). The coupling
motion of the equilibrium grain boundary cannot be observed
under low shear loading. However, the intergranular deformation can be conrmed in nonequilibrium grain boundaries
under shear loading of up to 600 MPa. Comparing the atomic
arrangements in Figs. 7(b) and 8(b), it is found that the
coupling motions in the nonequilibrium and equilibrium
grain boundaries are the same. Therefore, the EGBDs
function as the source of intergranular deformation, and
reduce the energy barrier for the intergranular deformation.
This is the same tendency reported by Kurtz et al.30)
1.60
L I = 45 nm
1.55 L I = 73 nm
3 3
1.45
1.40
1
22
1.35
3
1.6
11
1.30
1.4
Stress, /GPa
Stress, /GPa
1.50
1.25
1.20
1.2
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
0.000
1.15
0.015
0.005
0.010
0.015
0.020
0.025
Strain
0.017
0.019
0.021
0.023
0.025
Strain
Fig. 5 Stress-strain curves of bicrystal models with dierent LI values under tensile loading. The circles and triangles indicate the tensile
strains when dislocations are generated from the left and right crack tips, respectively. The red and blue plots show the cases of LI 45
and 73 nm, respectively.
b Lt : glissile EGBD
b nL : sessile EGBD
b L : EGBD
grain
GB
(a)A
grain
510 ps
(b)B
D1
1015 ps
D2
1520 ps
2025 ps
Fig. 7 Grain boundary coupling motion in the D2 direction in the upper

nonequilibrium grain boundaries each containing one EGBD under shear
loading zx 800 MPa in white broken boxes A and B shown in Fig. 6. (a)
5 ps and (b) 20 ps. The atomic arrangement before shear loading is shown
in Fig. 2(b).
2530 ps
(a)
3035 ps
D1
0.00
0.06
Equivalent strain
(b)
D2
D3 D4
Fig. 6 Change in equivalent strain distributions of nonequilibrium grain

boundaries, each containing one EGBD under shear loading
zx 800 MPa at 100 K. The directions of grain boundary motions are
dened in the lower right panel.
Fig. 8 Generation of intergranular deformation in upper equilibrium grain

boundaries containing no EGBD under shear loading zx 800 MPa at
100 K. (a) 16.5 ps and (b) 17 ps at the same position.
Although the coupling motion is the same in both

equilibrium and nonequilibrium grain boundaries, it is
found that the propagation velocities are dierent. Table 3
shows the propagation velocities of the intergranular

deformation in each grain boundary. The symbol of
indicates the absence of intergranular deformation in the
I1
[MPa]
1500
0
1500
(a) just after shear loading

Fig. 9
Table 3
(b) after annealing at 800 K
Distributions of the rst invariant of stress I1 of nonequilibrium grain boundaries each containing one EGBD after shear loading.
Velocity (m/s) of grain boundary motion under shear stress.

zx 800 MPa
zx 800 MPa
direction
D1
D2
D3
D4
D1
D2
D3
D4
0 EGBD
1380
1600
970
690
1170
1040
1 EGBD
840
2210
2670
830
1600
1350
2 EGBDs
410
3680
2760
900
2120
900
1890
1620
simulations. The propagation directions are dened in the

lower right panel in Fig. 6. The propagation velocities in
equilibrium grain boundaries do not show strong dependence on the direction. However, in the case of
zx 800 MPa, the propagation velocities in the D2 and
D3 directions are higher than those in the D1 and D4
directions, respectively, and vice versa in the case of
zx 800 MPa, except for D3 and D4 for two EGBDs.
High propagation velocities can be observed in the direction
of the Peach-Koehler force acting on the glissile grain
boundary dislocations with btL decomposed from the
impinged dislocations with bL under shear loading. This
behavior is more pronounced when the number of EGBDs is
larger. As a result, the intergranular deformation depends
strongly on the Burgers vector component of the impinged
dislocations, and the mobility is remarkably anisotropic with
respect to the loading direction. Grain boundary sliding
related to the glissile grain boundary dislocations could take
place under high shear loading even at low temperature,
although no remarkable dierence in atomic arrangement
after the coupling motion is observed.
Figure 9(a) shows the distributions of the rst invariant of
stress I1 of nonequilibrium grain boundaries each containing
one impinged dislocation just after shear loading, and
Fig. 9(b) shows I1 after annealing the grain boundaries at
800 K for 200 ps. Shear loading is applied for 90 ps and the
boundary conditions are assumed to be periodic in all
directions. Hence, the edges of the coupling motion propagating in opposite directions meet each other at a point, and
another intergranular deformation occurs at a site dierent
from the impinged dislocation during the shear loading.
Therefore, as compared to the equilibrium grain boundaries
shown in Fig. 3(a), the shear stress eld around the grain
boundaries just after shear loading is disordered, as shown in
Fig. 9(a). After annealing, the grain boundaries far away
from the impinged dislocations show shorter-range stress
elds shown in Fig. 9(b) than that of nonequilibrium grain
boundaries after shear loading shown in Fig. 9(a); therefore,
the grain boundaries regain their equilibrium states, as shown
in Fig. 9(b). On the other hand, regions where lattice
dislocations are impinged show longer-range stress elds
than equilibrium grain boundaries, which correspond to the
vertical component of the impinged lattice dislocation; the
component of bnL exists even after the shear loading and
annealing processes within the limitation of treatable time
scale in the molecular dynamics simulations.
4.
Conclusions
To investigate the eects of nonequilibrium grain boundaries containing EGBDs on the mechanical properties of
UFG metals, aluminum bicrystal models with the h112i5
grain boundaries are used in molecular dynamics simulations. The following results were obtained.
(1) The stress eld of the EGBD in the nonequilibrium
grain boundary remains because the Burgers vector of
the impinged dislocation does not immediately disappear in the grain boundary at low temperature. Moreover, it was found that the stress eld of the EGBD does
not strongly depend on the local grain boundary
structure, although the accommodation mechanism of
the dislocation core structure depends on the local grain
boundary structure.
10
(2) The stress eld of the EGBDs can aect the formation
of dislocations from the crack tip. The shielding eect
is aected by the distance between the crack tip and the
grain boundary and by the number of EGBDs. Moreover, the EGBD shielding eect on the crack tips is
strongly aected by the temperature and the applied
loading that controls the dissociation of the EGBDs.
(3) The EGBDs can function as the intergranular deformation sources under shear loading even at low temperature by decomposing the Burgers vector into two
components bnL and btL , which are normal and parallel to
the grain boundary, respectively. Therefore, the intergranular deformation strongly depends on the Burgers
vector component of the impinged dislocations, and the
mobility is remarkably anisotropic with respect to the
loading direction.
It was found that the EGBDs in nonequilibrium grain
boundaries play important roles for both intragranular and
intergranular deformations, and the EGBDs can be transformed into other structures at the grain boundaries easily
compared to the lattice dislocations in a crystal structure. As
the grain size decreases to several hundred nanometers, the
proportion of grain boundaries increases dramatically. Therefore, the dependence of nonequilibrium grain boundaries on
temperature and strain rate could be one of causes of unusual
mechanical properties of UFG metals.
Acknowledgments
The authors would like to thank the support from a Grantin-Aid for Scientic Research from the Ministry of Education, Culture, Sports, Science and Technology (MEXT),
Japan on Priority Areas Giant Straining Process for
Advanced Materials Containing Ultra-High Density Lattice
Defects.
REFERENCES
1) R. Z. Valiev, R. K. Islamgaliev and I. V. Alexandrov: Prog. Mater. Sci.
45 (2000) 103189.
2) Y. Iwahashi, J. Wang, Z. Horita, M. Nemoto and T. G. Langdon: Scr.
Mater. 35 (1996) 143146.
3) Y. Saito, H. Utsunomiya, N. Tsuji and T. Sakai: Acta Mater. 47 (1999)

579583.
4) R. Z. Valiev, Yu. V. Ivanisenko, E. F. Rauch and B. Baudelet: Acta
Mater. 44 (1996) 47054712.
5) N. Tsuji, Y. Ito, Y. Saito and Y. Minamino: Scr. Mater. 47 (2002)
893899.
6) E. Ma: Scr. Mater. 49 (2003) 663668.
7) G. T. Gray, T. C. Lowe, C. M. Cady, R. Z. Valiev and I. V.
Aleksandrov: Nanostruc. Mater. 9 (1997) 477480.
8) Q. Wei, S. Cheng, K. T. Ramesh and E. Ma: Mater. Sci. Eng. A 381
(2004) 7179.
9) N. Tsuji, S. Okuno, Y. Koizumi and Y. Minamino: Mater. Trans. 45
(2004) 22722281.
10) T. Shimokawa, T. Kinari and S. Shintaku: Phys. Rev. B 75 (2007)
144108(1-11).
11) W. T. Read and W. Shockley: Phys. Rev. 78 (1950) 275289.
12) J. Wang, Y. Iwahashi, Z. Horita, M. Furukawa, M. Nemoto, R. Z.
Valiev and T. G. Langdon: Acta Mater. 44 (1996) 29732982.
13) J. Y. Huang, Y. T. Zhu, H. Jiang and T. C. Lowe: Acta Mater. 49 (2001)
14971505.
14) K. Ikeda, K. Yamada, N. Takata, F. Yoshida, H. Nakashima and
N. Tsuji: Mater. Trans. 49 (2008) 2430.
15) X. Huang, N. Hansen and N. Tsuji: Science 312 (2006) 249251.
16) A. A. Nazarov: Phil. Mag. A 69 (1994) 327340.
17) M. P. Allen and D. J. Tildesley: Computer Simulation of Liquids,
(Oxford University Press, New York, 1989).
18) T. Shimokawa, A. Nakatani and H. Kitagawa: Phys. Rev. B 71 (2005)
224110(18).
19) Y. Mishin, D. Farkas, M. J. Mehl and D. A. Papaconstantopoulos: Phys.
Rev. B 59 (1999) 33933407.
20) P. H. Pumphrey and K. M. Bowkett: Scr. Met. 5 (1971) 365370.
21) M. Parrinello and A. Rahman: J. Chem. Phys. 76 (1982) 26622666.
22) L. Priester: Interface Sci. 4 (1996) 205219.
23) Stress distributions at 0.1 K are shown in all cases in this paper in order
to eluminate the inuence of thermal vibration.
24) J. P. Hirth and J. Lithe: Theory of Dislocations, 2nd Ed., (McGraw-Hill,
New York, 1982).
25) The energy of a lattice dislocation is estimated using b2 =
, 3:3,
f41 g lnR=r0 .24) Here, 30 GPa, b 2:86 A
R 50 nm, and r0 1 nm. R is determined to obtain the dislocation
density, which is the same as density obtained from the atomic model.
26) A. P. Sutton and R. W. Ballu: Interfaces in crystalline materials,
(Claredon Press, Oxford, 1995).
27) R. Thomson: Solid State Phys. 39 (1986) 1129.
28) M. Tanaka, K. Higashida, T. Shimokawa and T. Morikawa: Mater.
Trans. in press. DOI: 10.2320/matertrans.MD200817.
29) J. W. Cahn, Y. Mishin and A. Suzuki: Acta Mater. 54 (2006) 4953
4975.
30) R. J. Kurtz, R. G. Hoagland and J. P. Hirth: Phil. Mag. A 79 (1999)
665681.

First-Principles Study of the Stability and Interfacial Bonding

of Tilt and Twist Grain Boundaries in Al and Cu
Ru-Zhi Wang1;2 , Masanori Kohyama1; * , Shingo Tanaka1 ,
Tomoyuki Tamura3 and Shoji Ishibashi3
1
Research Institute for Ubiquitous Energy Devices, National Institute of Advanced Industrial Science and Technology,
Ikeda 563-8577, Japan
2
College of Materials Science and Engineering, Beijing University of Technology, Beijing 100124, P.R. China
3
Research Institute for Computational Sciences, National Institute of Advanced Industrial Science and Technology,
Tsukuba 305-8568, Japan
Nature of grain boundaries (GBs) should aect the micro-structural evolution and mechanical properties of metallic micro-crystalline
formed by severe plastic deformation. The stability and interfacial bonding of coincidence tilt and twist GBs in Al and Cu have been examined
by using the projector-augmented wave method within the density-functional theory. For the {221} 9 tilt boundary, glide models are more
stable than mirror models for Al and Cu, and the {001} 5 twist boundary is more stable than the 9 tilt boundary for Al and Cu, due to
smaller structural distortions. There is a tendency that the boundary energies in Al are substantially smaller than those in Cu. This can be
explained by the electronic and atomic behavior of bond reconstruction at the interfaces in Al, due to the covalent nature of Al as observed in the
charge density distribution, in contrast to rather simple metallic bonding at Cu GBs. [doi:10.2320/matertrans.MD200820]
(Received August 11, 2008; Accepted September 18, 2008; Published November 6, 2008)
Keywords: coincidence boundary, aluminum, copper, rst-principles calculation, severe plastic deformation, interfacial bonding
1.
Introduction
Grain boundaries (GBs) have serious eects on the

mechanical properties of metallic materials as barriers of
dislocation transmission or as sources or sinks of dislocations. This feature is greatly enhanced in nano- or sub-micron
grained metals formed by the severe plastic deformation
(SPD) process, such as ECAP (equal-channel angular
pressing), ARB (accumulative roll bonding) or HPT (highpressure torsion), where we can see peculiar mechanical
behavior such as co-existence of hardness and ductility.1,2) In
such systems, GB regions dominate substantial volume ratio
and there are only few movable dislocations inside grains.
Thus the response or deformation of GB regions for applied
stresses and the dislocation nucleation at GBs directly
dominate the macroscopic mechanical behavior. On the
other hand, the nature of GBs and the interactions between
GBs and dislocations also dominate the micro-structural
evolution during the SPD process. In this way, we have to
understand the primary behavior of metallic GBs under
various tensile or shear stresses, and the primary interactions
between GBs and dislocations. For this purpose, we would
like to emphasize the importance of computer simulations at
various length scales eectively combined with experimental
studies.
Al and Cu are typical fcc metals which can be easily
deformed into micro-crystalline by the SPD, although the
features of micro-structural evolution and eects of impurity
concentrations are quite dierent for Al and Cu.3) This should
be mainly caused by quite dierent behavior of dislocations
due to dierent stacking-fault energies (SFEs) for Al (166
mJ/m2 ) and Cu (45 mJ/m2 ).4) Recently, the origin of the
dierent SFEs has been explained by the dierent bonding
*Corresponding
author, E-mail: m-kohyama@aist.go.jp
nature.5) In Al, sp electrons reveal rather covalent or

directional-bonding features in deformed or defective congurations6,7) in contrast to rather isotropic metallic bonding
of Cu with a closed-shell d-band. More distant inter-atomic
interactions in Al via such sp electrons generate a substantial
energy dierence between fcc and hcp structures, while the
structural dierence over the 2nd neighbors is not so
sensitive in Cu with short-ranged inter-atomic interactions,
resulting in much larger SFE for Al. On the other hand, it is
quite interesting to examine this aspect on the bonding nature
at GBs themselves. The bonding nature at GB regions should
have serious eects on the mechanical response for applied
stresses and on the interactions with dislocations and impurities.
Thus in this paper, we examine the stability and atomic
bonding of typical coincidence-site-lattice (CSL) GBs in
Al and Cu, by using rst-principles calculations based
on density-functional theory (DFT). We deal with the
{221} 9 tile and {001} 5 twist GBs in Al and Cu.
It is essential to understand the behavior of valence electrons
at various GBs in each metallic species.
2.
Theoretical Method
2.1 Ab initio scheme

We used the projector-augmented wave (PAW) method8,9)
based on the DFT within the generalized gradient approximation (GGA)10) via generalized package program QMAS
(Quantum MAterials Simulator).11) The PAW method is the
most improved scheme within the pseudo-potential framework with both excellent accuracy and eciency. Realistic
oscillating behavior of valence-wave functions near nuclei
is accurately reproduced by the projection for pseudo-wave
functions described by a plane-wave basis, dierently from
conventional pseudo-potential schemes. For Al, the cut-o
12
R.-Z. Wang, M. Kohyama, S. Tanaka, T. Tamura and S. Ishibashi
radii for the projectors of s and p channels are all 1.9 a.u., and
the atomic local pseudo-potential is constructed as a normconserving pseudo-potential for a d component. For Cu, the
cut-o radii for the projectors of s, p and d channels are all
2.2 a.u., and the atomic local pseudo-potential is given by
truncating the all-electron atomic potential. The accuracy of
the scattering properties of these two sets of local potentials
and projectors was examined by logarithmic derivatives.8)
In QMAS, the electronic ground-state is obtained iteratively by the block-Davidson scheme12,13) coupled with the
ecient Kerker-Pulay charge-mixing scheme14,15) using the
technique of MPI parallelization.16) The plane-wave cut-o
energy is 40 Ry for Al and Cu, and the k-point sampling is
performed by the Monkhorst-Pack scheme.17) The selfconsistence convergence criteria is 1:0 105 . In the
relaxation of GB models according to Hellman-Feynman
forces, the force convergence criteria is set as 1:0 104
Hatree (Ht)/a.u. For the fcc bulk crystals, we obtained the
for Al and 3.63 A
for Cu, which are
lattice constants of 4.03 A
and
in good agreement with previous experimental (4.05 A
and 3.64 A
in
in18)) and theoretical results (4.04 A
3.61 A
Ref. 5).
2.2 Atomic models and supercells for GBs
CSL GBs are constructed as follows.19) We imagine the
superposition of the two identical crystals. If one of the two
crystals is rotated around one common axis such as [001],
[110] and [111] by some specic angle, the superposition of
the two crystals contains sparse lattice points to coincide with
each other, which are CSL points existing periodically in the
original lattice. The inverse ratio of the number of the CSL
points to that of the original lattice points, in other words, the
ratio of the unit-cell volume of the CSL to that of the original
lattice, is called as value. Then we introduce a GB plane
along some lattice plane of the CSL, and we remove each half
of the two crystals in the superposition. A formed GB should
have a periodic conguration along the GB plane due to the
periodicity of the CSL points along the GB plane. In this way,
we can construct CSL GBs with periodic congurations.
There are additional degrees of freedom for the boundary
geometry. One is a rigid-body translation (RBT) between the
two grains, which is a pure shift without any rotational
component. The above argument of a CSL GB is only
involved in the periodicity of a GB conguration, although a
RBT seriously aects a detailed GB conguration. The other
is the place of the GB plane in the CSL. In the above
argument of a CSL GB, the direction of the GB plane is
already xed, although the detailed place of the GB plane in
the original superposition of the two crystals also aects a
detailed GB conguration, associated with a RBT.
The {221} 9 tilt GB in fcc metals is constructed by
rotating one grain with 38.94 along the [110] axis and
setting (2,2,1) as a GB plane. Here the GB plane and the
rotation axis are parallel to each other, and thus this is a tilt
GB. The resultant CSL cell is dened by 12 a0 [-114] and
1
2 a0 [110] on the GB plane, where a0 is the lattice constant.
This means that the interface of this CSL GB has twodimensional periodicity of 12 a0 [-114] and 12 a0 [110] as clearly
shown in Fig. 1(a)1(d). Next, the RBT parallel to the
interface seriously aects the symmetry of the interface
conguration, although the RBT normal to the interface,

namely simple dilatation or compression, does not aect the
symmetry. Two dierent RBTs parallel to the interface
generate the glide-plane symmetry model (Figs. 1(a) and
1(b)) and the mirror-plane symmetry model (Figs. 1(c) and
1(d)), respectively. These two models were experimentally
observed in Al and Cu by the high-resolution electron
microscopy (HRTEM).20,21)
Then we construct 42-atom and 36-atom supercells
containing two interfaces in each cell for the glide-plane
and mirror-plane models, respectively, as shown in
Figs. 1(a)1(d). For band-theoretical calculations, we have
to deal with supercells with three-dimensional periodicity,
and thus each cell has to contain two identical interfaces with
inversion symmetry to each other. For the 36-atom supercell
of the mirror-plane model, each atomic layer is initially
stacked with an equal distance of a0 /6 even at the interfaces.
Thus an initial cell-length normal to the interface is 6a0 . For
the 42-atom supercell of the glide-plane model, atomic layers
in the bulk regions are stacked with a distance of a0 /6
similarly, although the interface layers have a dilated
distance of a0 /3 in the initial conguration. Thus an initial
cell-length normal to the interface is (22/3)a0 .
The {001} 5 twist GB in fcc metals is constructed by
rotating one grain with 36.87 along the [001] axis and
setting (001) as a GB plane.22) Here the GB plane and the
rotation axis are perpendicular to each other, and thus this is a
twist GB. The resultant CSL cell is dened by 12 a0 [310] and
1
2 a0 [-130] on the GB plane, which means that the interface
has two-dimensional periodicity of 12 a0 [310] and 12 a0 [-130]
as shown in Figs. 1(e)1(f). On the RBT parallel to the
interface, we deal with one symmetric conguration formed
by rotation around the [001] axis passing the atomic sites
of both the crystals, similar to the CSL model in Ref. 22).
About the supercell shown in Figs. 1(e)1(f), two symmetric
boundary planes are introduced per every 5 atomic layers,
alternately, resulting in a 50-atom cell with an initial celllength normal to the interface of 5a0 .
2.3 Structural relaxation
For each system, the Brillouin zone (BZ) is constructed by
using translation vectors of the supercell structure. 4 12
2 and 6 6 2 k-point meshes of the full BZ are used for
the 9 tilt and 5 twist GB cells, respectively,
corresponding to an enough dense mesh in the bulk fcc
crystals. For each cell, we utilize the symmetry to reduce the
total number of k points dealt with practically, and the
relaxation is performed with keeping the symmetry.
In order to obtain the most stable congurations for the
above models, only the RBT normal to the interface is
examined. This is because the RBT parallel to the interface
breaks the symmetry of each model. For the 5 twist GB,
there are no reasons to prevent the RBT parallel to the
interface, although in the present study we only deal with one
conguration similar to the CSL model in Ref. 22) without
introducing any RBT parallel to the interface as explained
above.
For each model, we iteratively perform the relaxation
of the conguration with a dierent RBT normal to the
interface. The relaxation run is iterated by changing the
Stability and interfacial bonding of tilt and twist boundaries in Al and Cu
13
Fig. 1 Relaxed atomic congurations of the GBs in Al and Cu: the glide model of the {221} 9 tilt GB in Al (a) and in Cu (b), the
mirror model of the {221} 9 tilt GB in Al (c) and Cu (d), and the {001} 5 twist GB in Al (e) and Cu (f). For each tilt GB, atomic
positions inside the supercell are shown with the atomic number of each atom. The size of each supercell represents the real size of each
relaxed supercell. For each twist GB, atomic positions inside the CSL cell on the interface plane are shown as well as the side view of the
supercell.
cell-length normal to the interface, where initial atomic

positions in the cell are uniformly changed before the
relaxation. In each relaxation run, all the atomic positions are
relaxed according to the Hellman-Feynman forces within the
constraints of the symmetry. And then we nally obtain the
most stable structure with the optimum atomic conguration
and RBT. Note that the RBT is dened by the relation
between the bulk regions of the two grains. The RBT normal
to the interface, namely dilatation or compression, is not
necessarily localized at the interface layer, but may exist
spread combined with the atomic conguration around the
interface. The validity of the optimized RBT is nally judged
by the stress calculation normal to the interface.23,24)
The boundary energy is dened as the energy increase
against the total energy of the bulk crystal with the same
number of atoms. In order to evaluate the boundary energy
accurately, we perform ab initio calculations of the bulk
supercells with similar periodicity as the GB cells. For the

9 tilt and 5 twist GBs, the 36-atom bulk cell with
the translation vectors of 12 a0 [-114], 12 a0 [110] and 2a0 [2-21],
and the 50-atom bulk cell with the translation vectors of
1
1
2 a0 [310], 2 a0 [-130] and 5a0 [001] are constructed, respectively. The computational conditions such as k-point meshes
are the same as those for the GB cells.
3.
3.1
Calculated Results
Optimum RBT, boundary energy and structural

disorder
Figure 1 shows the relaxed congurations of the six
boundaries. As explained above, only the RBT normal to the
interface has been examined in each system as the optimum
cell-length normal to the interface. The relative change ratio
is dened as:
14

Table 1
Calculated results for the CSL GBs in Al.
Glide-plane model
of the 9 tilt GB
Mirror-plane model
of the 9 tilt GB
The 5 twist GB
0.5373 A
0.6381 A
0.3426 A
GB energy
453 mJ/m2
(0.689 eV/CSL)
656 mJ/m2
(0.997 eV/CSL)
351 mJ/m2
(0.890 eV/CSL)
Range of the
bond-length
distortions
6:4% to +4.8%
10:5% to +11.3%
7:5% to +5.2%
Boundary excess
volume
Table 2
Calculated results for the CSL GBs in Cu.
Glide-plane model
of the 9 tilt GB
Mirror-plane model
of the 9 tilt GB
The 5 twist GB
0.4538 A
0.5566 A
0.3812 A
GB energy
755 mJ/m2
(0.931 eV/CSL)
1152 mJ/m2
(1.421 eV/CSL)
698 mJ/m2
(1.435 eV/CSL)
Range of the
bond-length
distortions
7:0% to +5.6%
10:9% to +10.5%
7:0% to +4.6%
Boundary excess
volume
RBT
lRBT lz
100%
2 dlayer
where dlayer is the interlayer distance in the bulk crystal

(dlayer a0 =6 and dlayer a0 =2 for the tilt and twist GBs) and
lRBT and lz are the cell length with the optimized conguration and RBT and the initial cell length normal to the
interface, respectively. Two in the denominator indicates the
presence of two interfaces in the cell. The calculated values
of RBT are 20% and 25% for the glide model of the
9 tilt GB in Al and Cu, +95% and +92% for the mirror
model of the 9 tilt GB in Al and Cu, and +17% and
+21% for the 5 twist GB in Al and Cu, respectively. Of
course, the present values depend on the initial setting of the
interface distance. For the glide models, the initial distance
between the interfacial two atomic layers is set to be a0 /3 in
contrast to a0 /6 in the mirror model. Thus the values of RBT
for the mirror models can be regarded as 5% and 8% for
Al and Cu for the initial setting similar to the glide models.
Similarly the values of RBT for the glide models can be
regarded as +80% and +75% for Al and Cu for the initial
setting similar to the mirror models. In any case, we can
see the natural tendency that the interface layers expand as
compared with the initial interface distance of the bulk
interlayer distance, due to the structural disorder. From the
above argument, we can obtain the boundary excess
volumes25) as absolute dilatation values compared to the
bulk crystal as listed in Tables 1 and 2.
The obtained boundary energies are also listed in Tables 1
and 2. We can see the following tendencies: (1) the glide
model is more stable than the mirror mode in both Al and Cu,
(2) the twist GB is more stable than the tilt GB in both Al and
Cu, (3) the boundaries in Al have smaller energies than those
in Cu, and (4) there is positive correlation between the
boundary excess volume and the boundary energy in both Al
and Cu. The nal point is consistent with the results of
classical simulations.25)
Fig. 2 Distribution of bond-length changes in the GB supercells: the glide

model of the {221} 9 tilt GB in Al (a) and Cu (b), the mirror model of
the {221} 9 tilt GB in Al (c) and Cu (d), and the {001} 5 twist
GB in Al (e) and Cu (f).
For each interface model, the optimized structures of Al

and Cu are essentially similar. Figure 2 shows the bondlength changes in each interface. Tables 1 and 2 also list the
ranges of the bond-length changes. For the mirror models, the
degree of bond distortions and the density of distorted bonds
are the largest for Al and Cu. For the other GBs, the degree of
bond distortions and the density of distorted bonds seem to be
in a similar range.
On the coordination defects in Fig. 1, four atoms have ten
1st neighbors and four atoms have eleven 1st neighbors in
one unit cell of the interface of the glide mode, compared to
twelve 1st neighbors for the bulk atom. In the mirror model,
four atoms have ten 1st neighbors and two atoms have eleven
1st neighbors. For the twist GB, eight atoms within the ten
interface atoms in one unit cell have ten 1st neighbors. Each
15
Fig. 3 Variation of interlayer distances in the GB supercells: the glide model of the {221} 9 tilt GB in Al (a) and Cu (b), the mirror
model of the {221} 9 tilt GB in Al (c) and Cu (d), and the {001} 5 twist GB in Al (e) and Cu (f).
interface atom has eight 1st neighbors within each grain, and
thus the eight interface atoms have only two neighbors on the
other side of the interface, respectively.
There exists clear correlation between the boundary
energy and the degree of local structural disorder for each
set of the three GBs for Al and Cu, which should be the
origin of the correlation between the boundary energy and
the boundary excess volume. Figure 3 shows the variation
of the interlayer distances for each GB. This clearly
indicates that the mirror models in Al and Cu have so large
structural distortions at the interfaces as to be extended into
the bulk regions seriously, while the structural disorder is
almost limited at the interfaces for the twist GBs. This
should be the reason of the lowest GB energies for the
twist GBs. It should be noted that the present twist GBs
possibly become more stable by introducing RBTs parallel
to the interface.
3.2 Comparison between Al and Cu
The boundary energies in Al are much smaller than those
in Cu, which are 60%, 57% and 50% for the glide and mirror
models of the 9 tilt GB and for the 5 twist GB,
respectively. On the gross energy increase per CSL cell, the
energies in Al are 74%, 70% and 62% of those in Cu for
the three kinds of GBs, respectively.
This does not simply mean the bond distortions for Al GBs
are smaller, although this tendency can be seen. Note that
defect energies and surface energies are generally larger in
Cu than in Al. For example, the vacancy formation energy is
experimentally 1.03 eV,26) 1.19 eV,27) or 1.30 eV28) for Cu,
in contrast to 0.68 eV for Al (52{66% of Cu),29) which are
well reproduced by recent ab initio calculations.6,7,30) Recent
ab initio calculations also indicate that the surface energy of
Cu(110) is 0.901 eV per surface atom in contrast to 0.708 eV
(78%) for Al(110).31,32) Thus the present results of the

smaller boundary energies for Al than for Cu are consistent
with other types of defect energies.
However, here we have to consider the relation of the
cohesive energies for Al and Cu, which are 3.39 eV/atom and
3.49 eV/atom experimentally.18) The dierence is less than
3%. If we consider that the cohesive energy is proportional to
a necessary energy to break inter-atomic bonds, the defect
energies in Al seem to be too small as compared with those in
Cu. There emerges a hypothesis that the covalent nature
of valence electrons in Al5) may generally reduce defect
energies, namely stabilize defective structures via electronic
and atomic behavior of reconstruction as observed in ab initio
calculations of Al surfaces33) or point defects.6,7) This
problem will be discussed again later.
3.3 Comparison with other ab initio results
Here we present comparison of our results with previous
ab initio results of Al and Cu GBs. There are not so many
ab initio results of metallic GBs using LDA (local density
approximation) or GGA as compared with classical computations using the embedded-atom method.34,35) For Cu GBs,
the boundary energies are 1270 mJ/m2 (GGA) for the {210}/
[001] 5 tilt GB,36) 1070 mJ/m2 (LDA) for the {310}/
[001] 5 tilt GB37) and 1040 mJ/m2 (LDA) for the
{331}/[110] 19 tilt GB.38) Our values for Cu are in a
similar range, while our values are somewhat lower than
these previous values. For Al, the boundary energies obtained
by LDA are 202 mJ/m2 and 427 mJ/m2 for the {113}/[110]
11 tilt and {112}/[110] 3 tilt GBs,39) 458 mJ/m2
for the {310}/[001] 5 tilt GB,22) and 502 mJ/m2 for the
{210}/[001] 5 tilt GB.40) At least on the energy range,
the agreement between our results and these results is good.
Note that for both Cu and Al cases, LDA values have a
16
(c)
(d)
(e)
(a)
(b)
(f)
Fig. 4 Valence-charge density iso-surfaces for the GBs in Al and Cu: the {001} 5 twist GB in Al (a) and Cu (b), the glide model of the
{221} 9 tilt GB in Al (c) and Cu (d), and the mirror model of the {221} 9 tilt GB in Al (e) and Cu (f). Iso-surfaces of
0.017 electrons/a.u.3 and of 0.051 electrons/a.u.3 are shown for Al and Cu, respectively.
tendency of overestimation of defect or surface energies

compared to GGA.
For the same GBs as ours, Ogata et al.22) obtained the
LDA value for the {001} 5 twist GB in Al as 676 mJ/
m2 as compared with our GGA value of 351 mJ/m2 . The
conguration seems to be similar to each other. In addition
to the usage of GGA, the dierence could be caused by the
usage of the large bulk cell to evaluate the boundary energy
in the present computations which seems to have substantial
eects on the accuracy. For the {221} 9 tilt GB in Al,
Uesugi and co-workers41) performed detailed examination
using GGA. For the glide model, they examined two models,
one is similar to ours and the other is a model with a RBT of
1
21)
The boundary energy of the former model is
4 a0 [110].
426 mJ/m2 in good agreement with our value of 453 mJ/m2 .
But the latter model is more stable with the energy of
408 mJ/m2 . Our model is similar to that in Ref. 42), of
which the tensile strength with and without impurities was
examined.4244) At present we cannot conclude which one
really exists, because both congurations lead to similar
HRTEM images consistent with experimental observations.41) In any case, both congurations have essential
features common to other tilt GBs. For the mirror model,
they obtained the energy of 430 mJ/m2 , which is much
smaller than our value. This is because the mirror model in

Ref. 41) is constructed by removing atoms from our model,
resulting in substantial reduction of local bond distortions.
Thus if we aim to clarify the overall features of the
{221} 9 tilt GB for wide freedoms of RBTs and atomic
densities, we have to add computations for new models
similar to those in Ref. 41) in the future. In the present
study, we deal with the mirror model constructed by simple
stacking of the bulk layers as a model with large local bond
distortions.
3.4 Electron distribution
Figure 4 shows the charge-density iso-surfaces for Al and
Cu GBs. In each gure, the iso-surfaces reveal the regions of
relatively high valence-electron density. It is clear that the
features of valence charge distributions at the Al interfaces
are quite dierent from those of the Al bulk regions and of
the Cu GBs. At the Al interfaces, we can see remarkable
charge accumulations at the local open regions formed by
greatly distorted bonds or less-coordinated atoms.
In the 5 twist GB in Al (Fig. 4(a)), this feature is
strikingly remarkable. We can see the bond-like chargedensity accumulations at the interfacial 11-18, 12-17, 13-20
and 14-19 bonds as indicated in Fig. 1(e). These four bonds
are equivalent symmetrically. The interfacial atoms involved

in these bonds have ten 1st neighbors, and have only two
neighbors on the other side of the interface. Then the
relatively strong interfacial bond is formed with one of the
two. For example, the atom of No. 11 has two neighbors,
No. 16 and 18 on the other side of the interface, and the 11-18
bond is strengthened with the length compressed by 7.5%,
resulting in the bond-like charge accumulation. In contrast,
there is no charge accumulation in the same GB in Cu
(Fig. 4(b)). Only spherical d-electron distributions centered
on each atom are seen at the interface, although the bond
lengths of the 11-18, 12-17, 13-20 and 14-19 bonds are also
compressed by 7.0% in the Cu GB.
In the glide model of the 9 tilt GB in Al (Fig. 4(c)),
there exist balloon-like accumulations at the pentagon-like
interfacial region formed by the atoms of No. 8, 9, 12, 13 and
11 in Fig. 1(a) for example. The coordination numbers are
ten for No. 9, 12 and 11, and eleven for No. 8 and 13 as
compared with twelve in the bulk. The 9-12 bond is
compressed by 6.4% as compared with the bulk bond
length and the large charge density occurs at this bond
especially. In the mirror model of the 9 tilt GB in Al
(Fig. 4(e)), there also exist balloon-like charge accumulations at the pentagon-like interfacial region formed by the
atoms of No. 7, 8, 12, 13 and 10 in Fig. 1(c) for example.
The coordination numbers are ten for No. 8 and 12, eleven
for No. 7 and 13 and twelve for No. 10. The 9-11 bond is
compressed by 10.5%, and the charge accumulation also
occurs at this bond.
On the other hand for the tilt GBs in Cu, we can see only
spherical charge distributions of d electrons centered on each
atom even at the disordered interfacial congurations as
shown in Figs. 4(d) and 4(f), in spite of larger compression of
the interfacial bonds, 7.0% for the 9-12 bond in the glide
model and 10.9% for the 9-11 bond in the mirror model, for
example. There are no special charge re-distributions at the
interfaces, although there are slight overlaps between the
spheres at the short bonds of the mirror model of the 9
tilt GB in Cu (Fig. 4(f)).
The present contrast of the interface charge distribution
between Al and Cu is rather similar to that observed in ab
initio calculations of {111}/h11-2i shear deformation of Al
and Cu,5) where the valence-charge distribution reveals
directional-bonding features for greatly distorted Al in
contrast to simple spherical d-electron distribution centered
on each atom in deformed Cu, resulting in larger ideal shear
strength for Al.
The present features of signicant charge accumulations
for local regions formed by less-coordinated or greatly
distorted atoms are general features observed in distorted or
defective congurations in Al such as point defects6,7) and
surfaces.33) In a single vacancy in Al, the rst shell is
strengthened by charge accumulation among less-coordinated atoms, which stabilizes a single vacancy against a
divacancy. In the present study, we have found that this
kind of phenomena generally occurs at Al GBs. In previous
ab initio studies, Uesugi and co-workers observed similar
bond-like charge accumulation at the 5 tilt GB in Al,45)
although they did not make any remarks from the present
view point.
4.
17
Discussion and Summary
Generally in defective or distorted congurations of Al

such as GBs, surfaces and point defects, we can see the
covalent behavior of valence electrons of Al, such as
charge accumulations or directional-bond formation at lesscoordinated or distorted atoms. This nature can be also
called as screening as discussed in Ref. 46). Historically,
this nature of Al was recognized as fails of simple shortrange pair-wise inter-atomic potentials47) to describe
energies and structures of various crystals, defects or
surfaces of Al. All these features are quite dierent from
Cu.
The relation between the bonding nature and the SFE is
discussed in Ref. 48). The SFE can be approximated by the
energy dierence between fcc and hcp structures. For Cu
with a closed-shell d band and s electrons, the energy of
crystal structures or defect structures can be described by
rather short-range inter-atomic interactions, due to rapid
decrease of d-d orbital interactions in the tight-binding view
point.47) This results in a rather smaller hcp-fcc energy
dierence due to common short-range order for fcc and hcp,
leading to a smaller SFE. On the other hand, for Al with sp
electrons, the energy of crystal structures or defect structures
can be described by rather long-range inter-atomic interactions, due to slowly decreasing s-s, s-p and p-p orbital
interactions and higher valence-electron density,49) resulting
in a larger SFE.
The eects of the covalent nature of Al compared to Cu, in
other words, screening eects or more distant inter-atomic
interactions, on general defective structures such as GBs,
point defects and surfaces, are quite dierent from the SFE
problem. The covalent nature of Al leads to a larger SFE than
that of Cu, due to the sensitivity for the distant structure in Al
in contrast to Cu. On the other hand, for general defective
structures, the covalent nature of Al causes smaller defect
energies than those in Cu, due to atomic or electronic
reconstruction. As discussed in Sec. 3.2, the smaller defect
energies in Al than those in Cu in spite of similar cohesive
energies in Al and Cu can be explained by the present view
point.
In this way, it is probable that behavior of Al GBs is
quite dierent from Cu GBs due to its bonding nature. Of
course, the behavior of dislocations dominated by the SFE
is closely concerned with the behavior of GBs, and it is
not so easy or practical to consider the intrinsic behavior
of GBs separated by dislocations. However, at least the
present results indicate the possibility of lowered GB
energies and strengthened interfacial bonding for Al GBs as
compared with Cu GBs. The strengthening of GBs via
reconstructed interfacial bonds was indeed observed in the
glide model of the {221} 9 tilt GB in Al by ab initio
tensile tests.42) In any case, it is of great importance to
examine the intrinsic mechanical properties of GBs and
interactions with impurities and dislocations from the
present view point in the near future. This will contribute
to the development of designing technology of the mechanical properties associated with the micro-structures by
the SPD process for various metallic materials from the
electronic viewpoint.
18
Acknowledgement
The present study was supported by Grant-in-Aid for
Scientic Research (KAKENHI) on Priority Areas, Giant
Straining Process on Advanced Materials Containing UltraHigh Density Lattice Defects (No. 18062007). The authors
thank Prof. Zenji Horita, Prof. Guang-Hong Lu, Prof. Masato
Yoshiya, Dr. Tokuteru Uesugi and Dr. Yoshinori Shiihara for
fruitful discussion.
REFERENCES
1) Z. Horita, K. Ohashi, T. Fujita, K. Kaneko and T. G. Langdon: Adv.
Mater. 17 (2005) 1599.
2) X. Huang, N. Hansen and N. Tsuji: Science 312 (2006) 249.
3) K. Edalati, T. Fujioka and Z. Horita: Mater. Sci. Eng. A (2008) to be
published.
4) J. P. Hirth and J. Lothe: Theory of Dislocations (Wiley, New York, 2nd
ed., 1982).
5) S. Ogata, J. Li and S. Yip: Science 298 (2002) 807.
6) K. Carling, G. Wahnstrom, T. R. Mattson, A. E. Mattson, N. Sandberg
and G. Grinvall: Phys. Rev. Lett. 85 (2000) 3862.
7) T. Uesugi, M. Kohyama and K. Higashi: Phys. Rev. B 68 (2003)
184103.
8) P. E. Blochl: Phys. Rev. B 50 (1994) 17953.
9) G. Kresse and D. Joubert: Phys. Rev. B 59 (1999) 1758.
10) J. P. Perdew, K. Burke and M. Ernzerhof: Phys. Rev. Lett. 77 (1996)
3865.
11) S. Ishibashi, T. Tamura, S. Tanaka, M. Kohyama and K. Terakura:
unpublished, http://www.qmas.jp
12) D. Singh: Phys. Rev. B 40 (1989) 5428.
13) N. Kosugi: J. Comp. Phys. 55 (1984) 426.
14) G. P. Kerker: Phys. Rev. B 23 (1981) 3082.
15) P. Pulay: J. Comp. Chem. 3 (1982) 556.
16) T. Tamura, G.-H. Lu, R. Yamamoto, M. Kohyama, S. Tanaka and
Y. Tateizumi: Modell. Simul. Mater. Sci. Eng. 12 (2004) 945.
17) H. J. Monkhorst and J. D. Pack: Phys. Rev. B 13 (1976) 5188.
18) C. Kittel: Introduction to Solid State Physics (Wiley, New York, 8th
ed., 2005).
19) A. P. Sutton and R. W. Ballu: Interfaces in Crystalline Materials
(Oxford, 1995).
20) D. A. Muller and M. J. Mills: Mater. Sci. Eng. A 260 (1999) 12.
21) J. R. Hu, S. C. Chang, F. R. Chen and J. J. Kai: Mater. Chem. Phys. 74

(2002) 313.
22) S. Ogata, H. Kitagawa, Y. Maegawa and K. Saitoh: Comp. Mater. Sci. 7
(1997) 271.
23) O. H. Nielsen and R. M. Martin: Phys. Rev. B 32 (1985) 3780.
24) O. H. Nielsen and R. M. Martin: Phys. Rev. B 32 (1985) 3792.
25) D. Wolf: Scr. Metall. 23 (1989) 1913.
26) W. Schule: Z. Metallkd. 89 (1998) 672.
27) Th. Hehenkamp, W. Berger, J.-E. Kluin, Ch. Ludecke and J. Wol:
Phys. Rev. B 45 (1992) 1998.
28) R. W. Ballu: J. Nucl. Mater. 69/70 (1978) 240.
29) H.-E. Schaefer, R. Gugelmeier, M. Schmolz and A. Seeger: Mater. Sci.
Forum 1518 (1987) 111.
30) T. Korhonen, M. J. Puska and R. M. Nieminen: Phys. Rev. B 51 (1995)
9526.
31) J. L. F. Da Silva: Phys. Rev. B 71 (2005) 195416.
32) J. L. F. Da Silva, C. Barreteau, K. Schroeder and S. Blugel: Phys. Rev.
B 73 (2006) 125402.
33) P. J. Feibelman: Phys. Rev. Lett. 65 (1990) 729.
34) A. Suzuki and Y. Mishin: Interface Sci. 11 (2003) 131; 425.
35) F. Sansoz and J. F. Molinari: Acta Mater. 53 (2005) 1931.
36) P. Ballo, J. Degmova and V. Slugen: Phys. Rev. B 72 (2005) 064118.
37) A. Y. Lozovoi, A. T. Paxton and M. W. Finnis: Phys. Rev. B 74 (2006)
155416.
38) R. Schweinfest, A. T. Paxton and M. W. Finnis: Nature 432 (2004)
1008.
39) A. F. Wright and S. R. Atlas: Phys. Rev. B 50 (1994) 15248.
40) G. Lu and N. Kioussis: Phys. Rev. B 64 (2001) 024101.
41) Y. Inoue, T. Uesugi, Y. Takigawa and K. Higashi: Mater. Sci. Forum
561565 (2007) 1837.
42) G.-H. Lu, S. Deng, T. Wang, M. Kohyama and R. Yamamoto: Phys.
Rev. B 69 (2004) 134106.
43) G.-H. Lu, Y. Zhang, S. Deng, T. Wang, M. Kohyama, R. Yamamoto,
F. Liu, K. Horikawa and M. Kanno: Phys. Rev. B 73 (2006) 224115.
44) Y. Zhang, G.-H. Lu, S. Deng, T. Wang, H. Xu, M. Kohyama and
R. Yamamoto: Phys. Rev. B 75 (2007) 174101.
45) T. Uesugi, K. Tsuchiya, M. Kohyama and K. Higashi: Mater. Sci.
Forum 447448 (2004) 27.
46) T. Hoshino, M. Asato, R. Zeller and P. H. Dederichs: Phys. Rev. B 70
(2004) 094118.
47) D. Pettifor: Bonding and Structure of Molecules and Solids (Clarendon,
Oxford, 1995).
48) C. Brandl, P. M. Derlet and H. Van Swygenhoven: Phys. Rev. B 76
(2007) 054124.
49) A. K. McMahan and J. A. Moriarty: Phys. Rev. B 27 (1983) 3235.

SPH Analysis of ECAP Process by Using Grain Renement Model

Ken-ichi Saitoh and Yuuki Ohnishi*
Department of Mechanical Engineering, Faculty of Engineering, Kansai University, Suita 564-8680, Japan
Plastic working of metals using severe plastic deformation (SPD) recently attracts interest of researchers, because it is a method that will
improve the toughness as well as the strength just by applying enormously large strain on the material. In this study, deformation, stress and
strain of aluminum in ECAP (equal channel angular pressing) are analyzed computationally by using smoothed particle hydrodynamics (SPH)
method which is one of particle methods. In addition to elastic-plastic constitutive relation, a newly developed theoretical framework in which
grain renement process is involved is proposed here. We try to implement the framework into the SPH computation. The main idea to conduct
grain renement in the material is that applied energy by plastic working causes continuous change in the total area of grain boundaries and
averaged diameter in the material becomes smaller and smaller. The reasonable change can be observed by using such simple but theoretical
framework. The present paper is focused on implication, formulation and possibility of the grain renement model.
(Received July 22, 2008; Accepted September 9, 2008; Published October 22, 2008)
Keywords: severe plastic deformation, numerical analysis, smoothed particle hydrodynamics, grain renement, equal-channel angular
pressing (ECAP), plastic work, aluminum
1.
Introduction
Plastic working process of metallic material using severe

plastic deformation (SPD) has attracted interests of many
researchers. The advantage of SPD processing is that it is
capable of improving toughness of the metallic material as
well as strength, just by injecting heavy deformation.1)
One of typical methods in SPD process is ECAP (equalchannel angular pressing), where a material is extruded
through a die which has hollow space with angular shape.2)
Since the sectional dimensions of the ECAP die at both
entrance and exit are almost the same, the molded shape has
the same dimension even though it has been subjected to very
large strain history inside.
It is reported that experimentally ECAP processing is
applied to ductile metals such as aluminum and copper alloy,
which are strengthened and are toughened suciently.
General principle of SPD is that averaged grain size in usual
polycrystalline metals diminishes to very small regime by
giant strain. As a results, they results in nanocrystalline
or ultrane-grained (UFG) material.3) Metallurgically and
physically vital mechanisms for strengthening of the material
are known so far. For example, there are entanglement of
dislocation motion, trapping of moving dislocation mainly
due to grain boundaries (GBs), strengthening of GBs
themselves, and etc. Some studies show that there are not
only such well-known mechanisms related to dislocation
theory, but are also new mechanisms such as GB sliding or
bending and the appearance/annihilation of non-equilibrium
GBs.4,5) Therefore it is worth while clarifying the mechanism
of GBs responsible for SPD processing from the viewpoint
of microstructual evolution.
The viewpoint of grain renement (GR) behavior is truly
important, but is still not clear. It is expected that, by using
atomistic simulation, GR mechanism will provide insight
along with the detailed change of crystalline structure, where
*Graduate
limited
Student, Kansai University. Present address: Isuzu motors
coarse grains nally change to UFGs.6) However, at the

present, atomic simulations are limited to too small region to
obtain the whole aspect of GR mechanism. Therefore, it is
necessary to study the microstructural change in a size larger
than atomic scale. Thus, a method of numerical simulation
based on the macroscopic framework is adopted in the
present study.
Nowadays, computational method such as nite element
(FE) method or mesh-free methods (alternatively called
particle methods)7) is extensively applied to the dynamics
simulation of solid materials. Our choice is the smoothed
particle hydrodynamics (SPH) method, which is one of
particle methods using particle approximation. We are trying
to utilize it to analyze the behavior of metal forming under
SPD process.
Recently, there have been some studies using the FE
analysis on ECAP processing.8,9) The FE analysis has been
applied widely to the plastic working problem and nowadays
we have numerous sophisticated softwares. Nevertheless, the
FE analysis needs very complicated treatment and theory to
include contact condition, fracture criterion, or large deformation. The particle methods such as SPH method will be a
kind of breakthrough for such complexities. For example,
when the large distortion of the material appears, it is free
from troublesome re-meshing (it will be inevitable in usual
FE analysis). Generally, the particle methods are suitable for
the formulation of large deformation. Besides, they can quite
naturally integrate the contact condition.
Although the SPH method is using particles, it is still a
macroscopic (i.e. continuum) simulation. Therefore it requires parameters for the constitutive relation which expresses mechanical or physical response of the material.
However, an ordinary materials law does not contain the
GR behavior. To our knowledge, any constitutive model does
not have a grain size as an time-evolution variable in order
to express the GR behavior in SPD process. Therefore we
propose the theoretical framework of GR behavior in SPD
process and attempt to implement it numerically with SPH
modeling.
20
K. Saitoh and Y. Ohnishi
The basic framework of our model is as follows. In the

SPD process, just like in ordinary plastic working (metal
forming), the external applied energy (plastic work) alters
microstructure inside material. The enlargement of the area
of GBs can be focused on, because it is directly connected
to GR behavior. The excessive free energy required for
extension of GB region is supplied from a part of the plastic
work which has been done on the material. The famous HallPetch law which provides the relation between the averaged
grain size and yield stress tends to raise the yield stress of the
material. When there is also work hardening behavior, in a
certain part the change due to the plastic work will become
small eventually and the GR behavior will cease to proceed
where grains have a certain grain size.
In the present paper, it is shown that the results are strongly
aected by the existence of grain-size evolution. The purpose
of this paper is to show the implication of theoretical
formulation of GR mechanism and to discuss its possibility
for analyzing SPD behavior.
This paper is organized as follows. In the next section,
some basic concepts of SPH method needed for the analysis
of plastic deformation are shown. Then our proposal about
the formulation of GR mechanism is presented. Also the
detailed computational model of ECAP processing by using
SPH method is shown. In the subsequent section, results are
shown and discussed. Finally, concluding remarks are shown.
2.
Theory and Method
2.1 SPH analysis for elastic-plastic solid materials

In the present study, the smoothed particle hydrodynamics
(SPH),1012) which is one of particle methods (mesh-free
methods) for dynamics of materials, is used. The SPH
method was originally introduced for astrophysical problem
such as coalescence between galaxies or between stars. Then
it has been developed to be applied to various elds of
mechanics and dynamics.13,14) The problem called tensile
instability can be overcome by many methods15,16) and the
SPH method is now developing to a broad and main framework of particle methods.7) In the method, time-evolution of
particle systems, as though in molecular dynamics (MD)
atomic trajectories are solved through equations of motion,
are determined by interaction between particles. The discrepancy between SPH and MD is that particles of SPH mean
only non-existing lumps of the material, each of them having
local physical values. Therefore, their interaction is to be
determined within macroscopic (continuum) relation, in the
same way as in the FE formulation.
The advantage of particle methods is that they do not
require any mesh but particles. This fact enables the SPH
method to deal easily with large deformation, rupture and
fracture sometimes we see actually. Indispensable equations
to be solved in SPH analysis are the equations of motion. The
equation of motion of each particle is approximated and discretized by a so-called kernel function, by which superposition of interaction between neighbors is carried out. For solid
materials, equation of motion of the particle i is given by,
!

X
dvi
i j

mj
2 ij Wij; ;
1
dt
2i
j
j
where j particles are in the range of interaction around i

particle. Subscripts and indicate the Cartesian coordinate
components (i.e. ; 1; 2; 3 x; y; z) and vi ; mi ; i ; i
are the velocity, the mass, the density and the stress tensor of
the particle i, respectively. Wij is the kernel function between
i and j particles and in the present study a limited-range cubic
spline function is adopted. In the equation of motion, the
derivative of kernel function Wij; (the comma before the
coordinate index means the derivative) is used. ij is called
the articial viscosity which is required to avoid inadequate
overlapping of particles.
In order to reproduce reaction in the solid material, stress
value has to be updated at every time step. That is, in addition
to equations of motion, variation of stress (internal force)
should reect the deformation (strained) state at each
materials point. The incremental approach with the formulation of large deformation is applied to the present study.
Using elastic-plastic coecient matrix De p , incremental
stress is determined by
d De p d";
where d" is strain increment. Switching from elastic analysis

to plastic one is carried out particle to particle, by comparing
Von-Misess equivalent stress with a yield stress y .
Because the length of paper is limited, we have to omit the
detailed explanation concerning the formulation of De p . In
brief, however, the required material constants are yield
stress y , Youngs modulus E, Poissons ratio , plastic
constant C and strain hardening exponent n (they are listed in
describing the SPH model in later section). The strain tensor
of particle i, "
i , is estimated based on the kernel approximation17) as,
o
X 1 n

_"
vj vi Wij; vj vi Wij;
3
i
2
j
where vi ; vj are particles velocities.
The diculty of SPH for plasticity is that the original
method does produce compressibility. Therefore, SPH
method has been improved by Sakai et al. to comprise
incompressible condition by using the SMAC (simplied
marker and cell) algorithm,18) which is a famous technique in
uid dynamics analysis. The modied version is called
SMAC-SPH and the present study utilizes its framework.
In the SMAC-SPH, after particles move according to a
constitutive law and dynamics, the velocity and the position
of particles are modied so as to fulll incompressible
condition in all. This modifying stage adopts implicit
algorithm and so nally all positions and velocities become
consistent with a given constitutive relation.
2.2 A simple model of grain renement
In polycrystalline metal, change in grain size d is supposed
to be caused by numerous kinds of physical or mechanical
condition, e.g. temperature T, time t, stress tensor , plastic
strain "p , deviatric stress p , mean stress m . Indeed, in a
strict treatment, grain size d can be presented by a function
d dt; T; ; p ; m ; "p ; _ ; _p ; _m ; "_ p ; :
At the present, what is needed is the model which is

constructed on an essential and simple idea on grain
renement behavior. Therefore, rst, we assume that d

depends on plastic work density wp dened as,
Z "p
wp
d";
5
21
a grain
the refined grain

SPD
(Plastic work)
which is composed of equivalent stress and increment of

equivalent strain d " within plastic regime. That is, wp means
total energy (work) density injected into the material during
the processing, excluding elastic contribution which will
disappear after the processing.
The wp results in the evoluting microstructure inside the
material. The change would contain various mechanism, such
as dislocation motion, formation of dislocation cell, creation
and annihilation of voids or vacancies, GB migration,
nucleation of new GBs, and these combination. Besides,
most of the injected energy will be dissipated as heat and be
lost eventually. Thus, plastic work can be presented as
wp eGB edefects eheat eother ;
where eGB , edefects , eheat and eother are increase of GB energy,

increase of defects energy including dislocation motion, heat
dissipation, and other, respectively. In considering SPD
processing, the creation of GB should be focused on. Since
the creation of new GB is equivalent to increase of GB area
and hence it means small grains, it automatically reects the
GR mechanism. Thus, increase of GB energy eGB can be
related to total plastic work wp in linear form as,
eGB wp
where is an arbitrary factor less than unity which

determines the contribution. In practical use, will have to
be specied by comparison with experiment. Since there is
large amount of heat dissipation, it is reasonable that be
very small, such as 10% or less, even though we do not have
any reliable way to estimate it now. Anyway, by using a
certain small fraction number , we are capable of estimating
the amount of expansion of GB area and its energy.
Figure 1 shows a conceptual relationship between GB area
(A) and the grain size (d). It assumes that in the total volume
all the grains are regularized and is closely packed by
polyhedora which have the same volume as sphere of grains
shown there (since the seamless packing of spheres is
impossible in reality, we just assume that the volume of one
sphere is equivalent to one polyhedron of grain.). Reduction
of the averaged grain size corresponds to the increase of
the GB area. This formulation is possible in full threedimensions, of course, but two-dimensional formulation is
shown here suitable for the present two-dimensional SPH
model. That is to say, spheres are recognized as cylinders
under the plane strain condition.
Suppose that regular-sized grains with diameter d occupy
a volume V in the material. When GB energy is (per area),
GB energy density (per volume V) will be given using only
and d by,
2
1 A 1
d
2
d
:
8
4
2 V
2
d
The factor 1/2 is needed, because all GB has to be made by
two surfaces of spheres meaning grains. Then, when initial
grain size was d0 and changed to the current size d, assuming
d
total volume
V0
diameter
d0
area A
0
V=V0
no change
actual boundary (no void)

boundary using spheres (with void)
(This is not real.)
Magnified view of GB & triple junction
Fig. 1 Conceptual relationship between increase of grain boundary (GB)
area and reduction of grain size (d) in a regular polycrystalline structure.
V0 , A0 and d0 are initial values of V, A and d, respectively.
the GB energy is unchangeable, increase of GB energy

density is calculated by,
eGB
2 2

d
d0
Consequently, the grain size d can be expressed by the

function of plastic work density wp . The expression is,
d
2
wp
2
d0
10
It means that grain size d is simply determined by initial

diameter d0 , GB energy and plastic work density wp . In this
expression, larger accumulation of plastic work derives
smaller grain size.
Now, for grain diameter over 100 nm, yielding stress y is
related to d so that y increases with decrease of grain size.
This is realized by an empirical law known as the Hall-Petch
relation,
1
y y0 Kd 2 ;
11
where y0 and K are constant values. According to this

relation, when the material begins to accumulate larger
plastic strain, plastic work wp increases and d decreases.
After that, many SPH particles which has experienced work
hardening will be easily brought back to elastic regime. Then,
wp reaches a limited value, and grain size d also settles at a
certain small value. The point is that expression of grain size
d (eq. (10)) together with this Hall-Petch relation (eq. (11))
realizes plausible distribution of grain size, dmin < d < d0
(from the UFG to the coarse-grained). In the same way, the
strain hardening exponent n is also assumed to be determined
by grain size d as,
1
n n 0 Kn d 2 ;
12
where n0 and Kn are constant values again.

2.3 Computation model of SPH simulations
Referring to the ECAP process schematically shown in
22
(b)
(a)
w=10
magnified view (corner)
particle (steel die)
xz
adding
rotation
h(ram )=6
ram
ram
t= 3
material
l=46
h( mat. )=30
z
x
material
process zone
angle
die
C
process
zone
2D plane strain
approximation
SPH modeling of ECAP

Fig. 2
Computation model for SPH simulations.
Table 1 Material constants and parameters for grain renement (GR)

model.
Poissons ratio
Plastic coecient
Work hardening exponent
length unit= mm
l=46 mm
Outline of ECAP process
Initial yield stress

Density
channel
die
ECAP processing
Youngs modulus
particle
(aluminum)
die wall
die
parameter
additional
die
w=10
material
value
variable/unit
Table 2
Computation condition for SPH calculation.
property
(variable)/unit
values
Width of billet
/mm
10.0
Width of die
/mm
the number of particles
47.5
2303 0
2406 90
Billet
Die/Ram
E/GPa
y /MPa
71.0
265.0
206.0
400.0
/kg/m3
2700
2700
Velocity of pressing
V/m/s
0.30
10.0
Time increment
t/s
1:0 107
Total ram stroke
/mm
20:0{30:0
0.5
Initial distance between particles
Rx /mm
0.5
Kernel length
h 2Rx )/mm
1.0
/
C/
n/
0.34
7.37
0.27
Initial grain size

Constant (for Hall-Petch)
d0 /mm
p
K/MPa m
10.0
3:875 102
Constant (for Hall-Petch)
y0 /MPa
p
Kn = m
14.25
Constant (for exponent n)
3:949 104
Constant (for exponent n)
n0 /
0.145
Contribution factor
/
0.10
Grain boundary energy
/J/m2
0.324
Fig. 2(a), computational model for SPH analysis was constructed as Fig. 2(b). It is reported that in the ECAP method a
strong processing zone of SPD deformation (shown by area
ABC in the gure) is formed at the corner.19,20) The region is
called plastic-deformation zone (PDZ). The external curve
(curve AC) contours so-called dead metal region at the
outer corner, but the eect and size of PDZ is not well
known. Therefore, the angle for PDZ shown in Figs. 2(a),
2(b) is varied from 0 to 90 in our model. The results are
compared with regard to . Larger corresponds to larger
PDZ. Without a round corner, the outer corner have a dead
metal to make the material ow smoothly. Accordingly, the
ow in the round corner with larger resembles one with
larger dead metal region.
Table 1 shows material constants and parameters for our
model. The material is aluminum (Al) and the GB energy
0:324 J/m2 is adopted. This value of is usually
obtained for that of symmetrical low-energy GB ( 3
CSL). Other computational parameters are shown in Table 2.
The contribution factor 0:1 10% is used. Constants
1.0
needed for the Hall-Petch relation (eq. (11)) and strain

hardening exponent (eq. (12)) are roughly derived from an
experimental fact. The fact says that, when grain size 10 mm
reduces to 1 mm, yield stress y 26:5 MPa raises to almost
2y (twice). Likewise, we assumes that then n increases to 2n.
The constants, y0 , K, n0 , Kn used in eq. (11) and eq. (12) are
determined so as to fulll these values. The initial value n is
obtained from experimental value of an annealed aluminum21) and is somewhat dierent from that of practical 10 mm
grains. However, we would like to see the qualitative change
of d with the change of n.
Actual ECAP processing is composed of repeated passes
through the same die. In the present model, however, the
specimen is just once pressed onto downward (y direction
in Fig. 2) by moving a ram (a plunger) with constant velocity
V. This is for simplicity, so here we do not discuss about the
eect of multi-pass ECAP.
3.
3.1
Results and Discussion
Behavior of particles in ECAP processing: without

GR model
Figure 3 shows the material ow for the model of 0 ,
in which no GR model is comprised. In the upper gures,
spheres lightly shading are in the plastic analysis, distinguished from the elastic particles which are shown in darker
23
300
Equivalent stress /MPa
(a)
Elastic
Plastic
Equivalent stress
290
280
y =265 MPa (initial value)
270
an averaged area
260
0
ram
10
y=0.5 mm
Distance from the bottom /mm
plastic
plastic
plastic
78 degrees
0.8
0.6
Equivalent plastic strain

0.4
0.2
0
D=2.0 mm
D=10.0 mm
D=20.0 mm
10mm
materials front edge
Equivalent plastic strian
(D: ram stroke)
4
6
8
schematic: averaging method of

equivalent strain,stress or grain size
With distinction of plasic/elastic regions

(b)
0.000
10mm
10
Fig. 4 Distribution of equivalent plastic strain for 0 with no grain

renement model (the direction of inspection is perpendicular to the
longitudinal direction in the horizontal channel).
0.042
0.083
0.125
0.167
(a)
0.208
materials front
edge
0.250
Elastic
Plastic
D=2.0 mm
D=10.0 mm
D=20.0 mm
plastic
(D: ram stroke)
plastic
(D: ram stroke)

materials front edge
With equivalent plastic strain distribution

Fig. 3 Materials ow of the model with
renement model.
plastic
0 and without grain
D=2.0 mm
gray. On the other hand, lower gures superposes the equivalent plastic strain "p which is evaluated at the point of each
particle. After plastic region appears at the center of the
processing zone of PDZ as predicted in Fig. 2(a), it spreads
forward over the horizontal (2nd) channel. The front edge of
the material has an inclined angle, because the material at the
bottom is more pushed forward than in the upper material.
Higher value in the distribution of "p appears in a parallelogram shape emphasized in Fig. 3(b) (indicated in cases of
D 10 and 20 mm). On the other hand, region recovered
back to elastic is found just in the small region of front edge.
Figure 4 shows the distribution of "p and which are
averaged in the area within 10 mm forward from the inner
corner of the die. Each averaged region is a slab shape having
thickness of y 0:05 mm along y direction as shown inside
Fig. 4 (on the right). In very vicinity of the bottom wall, "p
raises to twice. Equivalent stress shows adequately the
range 280290 MPa, in which the bottom wall has larger
value.
Quantitatively, the value range of "p is much smaller than
that computationally predicted by FE analysis in usual
condition of ECAP processing where "p ranges 1{2.20) It is
guessed that this is because we do not include wall friction
model nor strict relation in the material of billet. By adjusting
these conditions to the real experiments, the results will come
to quantitatively plausible range of "p . In SPH simulation,
interaction between particles of die and those of the material
is capable of being adjusted to realistic friction,22) but we do
not include it to this SPH model. Hence, here, we will be able
to discuss our results only in qualitative sense.
elastic
D=10.0 mm
elastic
74 degrees
D=20.0 mm
With distinction of plastic/elastic regions

(b)
0.000
0.042
0.083
0.125
0.167
0.208
materials front
edge
D=2.0 mm
D=10.0 mm
D=20.0 mm
0.250
(D: ram stroke)
With equivalent plastic strain distribution

Fig. 5 Material ow with grain renement (GR) model for the model with
0 .
3.2 Results using GR model

Figure 5 shows the material ow of the model which is
applied grain renement (GR) model (in the case of 0 ),
with distinction of plastic particles or with the value of "p
(equivalent plastic strain). When comparing these with the
results without GR model shown in Fig. 3, it is found that
many particles recover back to elastic regime from plastic
regime. This is because plastic work given by pressing once
produces smaller d at the process zone, then accordingly
yield stress y increases due to those smaller d through HallPetch relation, and consequently particles need still larger
strain to remain in plastic analysis. Interestingly, ahead the
24
(a)
(a)
0.000
d=10.0
0.042
d=8.4
0.083
d=6.8
0.125
d=5.2
0.167
d=3.6
strong
grain refinement
0.208
d=2.0
materials front
edge
D=2.0 mm
D=10.0 mm
D=20.0 mm
0.250
d=0
(D: ram stroke)
=30 deg.
=60 deg.
Time evolution of grain size distribution
No grain refinement model
(b)
(b)
=90 deg.
0.000
Averaged grain size d / m
0.042
7
0.083
0.125
0.167
0.208
4
0.250
3
0
10
Averaged grain size along y direction shown in Fig.4
Fig. 6 Grain size distribution by using grain renement (GR) model for
0 .
process zone (PDZ) at the die corner, most particles recover

back to elastic regime.
As a result, the front edge of the material inclines more
than for the model without GR. The angle to the horizontal
line changes from 78 to 74 degrees as shown in Fig. 3 and
Fig. 5, respectively. This is partly because the eect that the
pressed material is pushed forward by spring-back behavior
beyond the process zone is larger in the bottom region.
Indeed, "p increases drastically at the entrance of process
zone, but, passing through the corner between channels,
particles with large "p begin to scatter and the distribution
loses obvious parallelogram meaning intensive area, dierent
from Fig. 3(b).
Figure 6(a) shows the grain size distribution. The smaller
grains start at the center of the process zone. The region in
which grains are made small corresponds to the region which
has larger "p . Ahead the corner of the die, the strong GR
behavior is found along the bottom wall line (the area is
emphasized by small broken lines).
Figure 6(b) shows the distribution of grain size along the
line perpendicular to the channel (with the same averaging
as Fig. 4). Grain size monotonically increases from the
bottom wall to the upper wall. Although the grain diameter
begins with 10 mm, the minimum diameter is found down
to 3 mm near the bottom wall. On the other hand, nearby
the upper wall the diameter remains 6{7 mm. That is,
the dierence in renement behavior is more than twice
between upper and lower dies. This result agrees well with
FE analysis.23)
4.
Eect of the Process Zone Angle

Figure 7 shows the dierence of materials ow for the
=30 deg.
=60 deg.
=90 deg.
Using grain refinement (GR) model
Fig. 7 Dependence of material ow on angle (shading indicates intensity

of "p ). The height of ram is dierent because its initial position is also
dierent so as that the head of the material start at the end of outer curve of
process zone.
variation of the angle

(30, 60 and 90 ), with GR and
without GR models. Basically, larger the
is, the more
smoothly the material ows. The position of top of the ram is
not at the same height because the pictures are tried to
compared at the same ram stroke (D 10 mm) from the
starting time at which the head of the material begins to
interfere with the PDZ. Actually, the ow seems smooth in
any , but the intensity of "p or grain reduction obviously
decreases with increase of the angle .
For GR model, the distribution of "p (equivalent plastic
strain) and the distribution of grain size d are shown in
Figs. 8(a) and 8(b), respectively. The averaging method is the
same as Fig. 4. Since in any case "p becomes steeply large in
regions very close to the two die walls, graphs here omit the
value in those wall regions.
When is small, i.e. the outer corner of the die is sharp,
"p comes to large in upper region of the channel. On the
other hand,
is large, i.e. the outer corner is round, "p
diminishes on the whole region. This is because the channel
with round corner makes the stream line of the material
smooth and makes the ow easy. When the outer corner is
sharp, the material ow tends to stagnate at the outer corner
region and forms dead metal zone there. Therefore, when
the die itself has the round corner, the outer curve of the die
works just alike an interface between the material and the
dead metal. In all cases, the center of channel has a minimum
in "p .
Figure 8(b) indicates that grain size distribution strongly
depends on . When 0, the grain size d is monotoni-
5.
(a)
Equivalent plastic strian
0.2
=0
= 0SPD
= 30SPD
= 60SPD
= 90SPD
=30
=60
0.15
=90
0.1
1
Equivalent plastic strain
(b)
Grain size parameter / m
8
7
=90
=60
=30
5
4
=0
= 0SPD
= 30SPD
= 60SPD
= 90SPD
3
1
5
7
Distribution of grain size
Fig. 8 Equivalent plastic strain "p and grain size d distribution for various
s, the angle of process zone at the corner. (for GR model)
cally increase from bottom die (the left in the gure) to

upper die (the right in the gure). In other words, the
intensity of GR is stronger near the bottom die than near the
upper die. However, when
is large, GR is hard to be
carried out nearby the bottom die. Especially when 90 ,
coarse grains (large d) still remain the bottom region in the
material. Thus, in the horizontal channel, the material ow
becomes smoother by using GR model. Hence in GR model,
grain renement tends to be conned in the center of the
material.
The intensity of GR becomes the largest in the case
of
0 which shows large "p and stagnating ow.
This means that there is a reverse relation between the
smoothness of the material ow and the grain renement
behavior.
In the actual experiments, the grain size dmin obtained in
SPD processing is typically discussed in sub-micrometer
regime.1) In our case, dmin stops at around 3 mm which is
larger than in such experiments. In our cases, parameter
setting for initial diameter d0 , yield stress y with Hall-Petch
relation, or strain hardening exponent n is so-to-speak
temporary. However, we can obtain qualitative evaluation
concerning the existence of GR model and clarify the eect
of variation in . So, we should expect that further study will
quantitatively survey the eect of these parameters referring
more to actual SPD processing, based on the framework
proposed in our model.
25
Conclusion
In the present study, we perform numerical simulations

using smoothed particle hydrodynamics (SPH), for the ECAP
(equal-channel angular pressing) processing which is one of
promising severe plastic deformation (SPD) methods. The
vital physical behavior common to SPD processes is that the
grain size reduces during heavy deformation (we call it grain
renement: GR). We consider a simple relation between
injection of plastic work and expansion of grain boundary
energy. Then we formulate a certain GR model for the
macroscopic framework of SPH simulation with large
deformation. The major ndings are as follows.
(1) Our formulation involving a simple GR relation works
well so as to present plausible evolution of grain size
occurring in ECAP processing.
(2) Strong plastic deformation is found in a region so-called
PDZ (process deformation zone). It starts at the
connecting region between the inner and outer corners.
The angle
which characterizes the size of process
zone aects the ow of the material as well as the
plastic strain inside. Larger produces smooth ow,
but it goes with weak GR behavior.
(3) When using GR model, equivalent plastic strain
becomes small in regions nearby wall. It leads to strong
reduction of grain size in the center of the material. The
reduction of grain size is saturated at some level (initial
diameter 10 mm changes to the nal 3 mm).
Acknowledgments
This work was supported by KAKENHI 18062004 (Grantin-Aid for Scientic research on Priority Areas). Part of this
work was supported by HRC, Kansai University and reserach
group of ORDIST, Kansai University. The authors also
acknowledge Prof. Sakai (Yokohama National University,
Japan) for the discussion about numerical implementation of
SMAC-SPH method. One of the authors (KS) also acknowledge Prof. Liu (Northwestern University, USA) for the
fruitful discussion and suggestion about using particle
method in solid mechanics problem.
REFERENCES
1) R. Z. Valiev, Y. Estrin, Z. Horita, T. G. Langdon, M. J. Zehetbauer and
Y. T. Zhu: JOM (2006) 3339.
Mater. 35 (1996) 143146.
3) Z. Horita (ed.): Nanomaterials by severe plastic deformation, (UetikonZurich, Switzerland: Trans. Tech. Publications, 2005).
4) A. Nazarov, A. E. Romanov and R. Z. Valiev: Acta Metall. Mater. 41
(1993) 10331040.
5) N. Q. Chinh, P. Szommer, T. Csanadi and T. G. Langton: Mat. Sci. Eng.
A 434 (2006) 326334.
6) R. Matsumoto, T. Hayashida and M. Nakagaki: J. Soc. Mater. Sci.
Japan 55 (2006) 693699.
7) S. Li and W. K. Liu: Appl. Mech. Rev. 55 (2002) 134.
8) J.-Y. Suh, H.-S. Kim, J.-W. Park and J.-Y. Chang: Scr. Mater. 44
(2001) 677681.
9) W. J. Zhao, H. Ding, Y. P. Ren, S. M. Hao, J. Wang and J. T. Wang:
Mater. Sci. Eng. A 410/411 (2005) 348352.
10) L. B. Lucy: Astron. J. 82 (1977) 10131024.
11) J. J. Monaghan: Annu. Rev. Astron. Astrophys. 30 (1992) 543574.
26
12) W. G. Hoover and H. A. Posch: Phys. Rev. E 54 (1996) 51425145.

13) L. D. Libersky, A. G. Petschek, T. C. Carney and F. A. Allahdadi:
J. Comp. Phys. 109 (1993) 6775.
14) P. W. Randles and L. D. Libersky: Comput. Methods Appl. Mech. Eng.
139 (1996) 375408.
15) J. J. Monaghan: J. Comp. Phys. 159 (2000) 290311.
16) P. W. Randles and L. D. Libersky: Int. J. Num. Methods in Eng. 48
(2000) 14451462.
17) K. Shintate and H. Sekine: Composites: Part A 35 (2004) 683692.
18) Y. Sakai and A. Yamashita: Trans. JSME., Ser. A 67 (2001) 1093
1102.
19) G. M. Stoica, D. E. Fielden, R. McDaniels, Y. Liu, B. Huang, P. K.

Liaw, C. Xu and T. G. Langdon: Mater. Sci. Eng. A 410/411 (2005)
239242.
20) W. Wei, A. V. Nagasekhar, G. Chen, Y. Tick-Hon and K. X. Wei: Scr.
Mater. 54 (2006) 18651869.
21) N. Noda and T. Nakamura: Kiso soseirikigaku, (1991), 19, Nissinsyuppan (Japan).
22) H. Fukumoto, K. Saitoh and N. Shinke: Proc. Computational Engineering Conference (JSCES 11 2005) 4952.
23) J.-H. Lee, I.-H. Son and Y.-T. Im: Mater. Trans. 45 (2004) 2165
2171.

Distributions of Hardness and Strain during Compression in Pure Aluminum

Processed with Equal-Channel Angular Pressing and Subsequent Annealing
Tadanobu Inoue1; * , Zenji Horita2 , Hidetoshi Somekawa3 and Fuxing Yin1
1
Innovative Materials Engineering Laboratory, National Institute for Materials Science, Tsukuba 305-0047, Japan
Department of Materials Science and Engineering, Faculty of Engineering, Kyusyu University, Fukuoka 819-0395, Japan
3
Structural Metals Center, National Institute for Materials Science, Tsukuba 305-0047, Japan
2
Single compression tests were conducted for two types of pure Al (99.99%): one processed using equal-channel angular pressing (ECAP)
through route Bc for 8 passes and the other with the same ECAP but successively annealed at 573 K for 1 h. The distributions of the equivalent
strain introduced by the compression were quantitatively examined by nite element analysis (FEA). Comparison was made between the ECAP
samples and the ECAP-annealed samples with regard to the inhomogeneity and magnitude of the strain developed during the compression.
Hardness variations with the equivalent strain were also compared with those obtained by earlier studies using ECAP and high-pressure torsion.
Moreover, the hardness variation throughout the cross section after compression was well predicted using the FEA for both ECAP sample and
ECAP-annealed sample. It is shown that the change in the strain path from ECAP to compression aects the hardness variation with respect to
the equivalent strain. [doi:10.2320/matertrans.MD200819]
(Received August 5, 2008; Accepted September 4, 2008; Published October 22, 2008)
Keywords: severe plastic deformation, nite element analysis, strain inhomogeneity, aluminum, hardness
1.
Introduction
Various processes of severe plastic deformation (SPD)

such as equal-channel angular pressing (ECAP), highpressure torsion (HPT), accumulative roll-bonding (ARB)
and multidirectional deformation (MDD) have been proposed as useful methods for introducing intense plastic strain
in metallic materials and thus for producing utrane-grained
(UFG) structures. Many experimental studies are available
for observations of microstructures as well as measurements
of mechanical properties in bulk ultra-ne grained (UFG)
materials produced by the SPD processes.17) Recently, nite
element analysis (FEA) is actively used to understand the
microstructural inhomogeneity in SPD-processed materials
and to determine an optimum condition in each SPD
method.814)
For the ECAP process, it is known that route Bc is
experimentally an optimum processing route for the grain
renement.15) This suggests that the change in strain path
aects microstructural evolution and its hardness variation. It
was reported that a homogeneous equiaxed grain structure
develops after processing through route Bc for 4 passes in
pure Al and the microstructure and hardness remain
essentially unchanged in the subsequent passes.1517) However, a question arises whether this may be the case if a strain
path dierent than ECAP is successively applied to the
ECAP-processed sample.
This study is thus initiated to conduct a conventional
axisymmetric compression tests for two types of the samples
processed with ECAP and with ECAP followed by annealing.
An FEA is performed for obtaining the strain distributions
created by the compression tests, and inhomogeneity and
magnitude of the strain are compared between the two
samples exhibiting dierent stress-strain relations. A relationship is thus constructed between the strain simulated by
*Corresponding
author, E-mail: INOUE.Tadanobu@nims.go.jp
FEA and the hardness measured after the compression test.

The results are discussed with reference to the hardness
variations obtained by ECAP and HPT.
2.
Experimental Procedure
Pure aluminum of 99.99% purity was used in this study.

First, an ingot was swaged to rods of 10 mm in diameter. The
rods were cut to a length of 60 mm and then annealed for 1 h
at 773 K. ECAP was conducted at room temperature for 8
passes through route Bc (hereafter, designated as the ECAP
sample). Full details of the procedures for ECAP were given
earlier.16) Following ECAP, samples were annealed for 1 h at
573 K for comparison (hereafter, designated as the ECAPannealed sample). At this time, the Vickers hardness and the
average grain size were measured and they were used in this
study as the initial Vickers hardness, Hv0 , and initial average
grain size, d0 : thus, for the ECAP sample, Hv0 37 and
d0 2 mm and for the ECAP-annealed sample, Hv0 25
and d0 50 mm. An earlier report16) gave details of the
microstructural evolution occurring in ECAP samples during
annealing. All ECAP samples including the annealed
samples were machined to cylindrical specimens that were
12 mm in height and 8 mm in diameter for compression
testing.
The compression test was performed at a strain rate of
0.01 s1 in a single pass at room temperature (293 K) using
a compression simulator. This strain rate was selected to
avoid heat generation during compression testing. The load,
F, and the displacement, uzz , were recorded continuously as a
function of time to obtain Fuzz curves. The compression
tests were terminated at either a compression ratio of 49 or
73% for both the ECAP samples and the ECAP-annealed
samples. The specimens after the compression tests were cut
for the hardness measurement and microstructure observation in the direction parallel to the compression direction and
the sectional planes were carefully polished to a mirror-like
T. Inoue, Z. Horita, H. Somekawa and F. Yin
surface. The hardness measurements were undertaken on the

sectional planes along the two symmetry axes intersecting
parallel and perpendicular to the compression direction. For
each hardness measurement, a load of 25 g was applied for
15 s. Microstructures in the cross section were observed using
the electron back scattered diraction (EBSD) method in a
Schottky type scanning electron microscope (SEM) operated
at 15 kV. EBSD analysis was conducted using a Carl-Zeiss
LEO Gemini 1550 scanning electron microscope equipped
with a TSL-OIM analytical system.
(a)
35
ECAP
30
Load F, / kN
28
ECAP-annealed
25
Displacement uzz
20
z
r
15
3.
73%
Load F
10
Numerical Procedure
5
4.
4.1
Comparison of experimental measurements and

numerical results
Figure 1 shows the Fuzz curves and the true stress -true
strain " curves up to " 0:7 for the ECAP sample and the
ECAP-annealed sample. The curves were signicantly
dierent between the two types of the samples in the early
stages in deformation, and this dierence decreased with
increasing displacement of compression. The dierence in
the early stages depends on the initial microstructure, and the
conformance in the nal stages indicates that the hardness in
the compressed samples was, on average, almost the same
when a large strain was introduced. In the ECAP-annealed
sample, a yield stress of YS 32 MPa is observed, followed
by work hardening under plastic deformation. The ECAP
sample reaches a yield stress of YS 132 MPa, which is
four times higher than that of the ECAP-annealed sample.
However, strain softening immediately occurs, and the ow
49%
0
10
Displacement, uzz / mm
(b)
True stress, / MPa
FE simulation was carried out using the FE-code ABAQUS/Explicit ver. 6.3-5. A 1/4 model was used for the
analysis by utilizing the symmetry condition. An 8-node
linear element in an axisymmetric model was used for the
sample, and the anvil was regarded as the rigid body. The
nite element mesh in the sample included 2501 nodes and
2400 elements. As the analysis proceeded, the mesh became
too degenerate by heavy deformation, and remeshing was
inevitably required. A Youngs modulus of 70 GPa and a
Poissons ratio of 0.35 were used for the sample at 293 K. The
stress-strain relationships employed in the analysis were
obtained using the experimental data of compression tests at
a room temperature and a strain rate of 0.01 s1 . Experimental data up to a true strain of " 0:7 were used to avoid
excessive inhomogeneous deformation due to barreling.
To obtain the relationships beyond the true strain of 0.7,
extrapolation through Hollomons equation, 0 K"n 18)
was used for the ECAP-annealed sample. For the ECAP
sample, because the ow stress was saturated around the
strain of 0.5 and remained constant after that, a relation of
constant stress with strain was adopted beyond the strain of
0.5. The Coulomb condition was used as the frictional
condition between the anvil and the sample, p, where
denotes the shear stress, is the friction coecient and
p is the contact pressure. The value of was determined
using the experimental Fuzz curves including sample
congurations.
200
ECAP
ECAP-annealed
150
YS =132
100
50
YS =32
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
True strain,
Fig. 1 (a) Load-displacement curves and (b) true stress-true strain curves
obtained with axisymmetric compression test for ECAP sample subjected
to 8 passes through route Bc and for ECAP sample after annealing at 573 K
for 1 h.
stress levels o thereafter, i.e., strain hardening becomes

zero.
Figure 2 shows a comparison between the Fuzz curves
predicted by the FEA using a friction coecient of 0:35
and those experimentally measured. The curves showed good
agreement with the experimental data. It should be noted that,
although comparison was made with several values for , the
best agreement was achieved with 0:35. Furthermore,
the shapes of the compressed samples became the same as
those by the FEA for 0:35. Hence, this study used the
friction coecient of 0.35 in the FEA.
4.2 Dierence of strain distributions
As seen from the -" curves shown in Fig. 1(b), the
deformation behaviors are signicantly dierent between the
ECAP sample and the ECAP-annealed sample. This suggests
that the microstructure aects the -" curves and thus the
inhomogeneity of the strain imposed by the compression test
must be dierent even when the compression conditions are
Distributions of Hardness and Strain during Compression in Pure Aluminum Processed with Equal-Channel Angular Pressing
35
ECAP
ECAP-annealed
30
FE analysis
Load, F / kN
25
ECAP
ECAP-annealed
20
15
10
5
0
10
Displacement, uzz / mm
Fig. 2 Comparison of load-displacement curves calculated by nite
element analysis with friction coecients 0:35 and experimental
measurements for ECAP samples and ECAP-annealed samples.
the same. The contours of the equivalent strain "eq are

depicted in Figs. 3(a) and 3(b) for the ECAP sample and the
ECAP-annealed sample, respectively, after compression of
73%. The inset in Fig. 3 describes the cross section where
the contours are depicted. It is evident that both magnitude
and inhomogeneity of "eq depend critically on the initial
microstructural conditions. Three important features arise
29
from the strain distribution. The rst feature, which is

common to both the ECAP sample and the ECAP-annealed
sample, is that the strain accumulation is higher as the
position closer to the center of the compressed specimen
with additional maxima near the upper and lower surfaces
around the outer edges. This feature is consistent with the
general understanding that the deformation tends to concentrate at the specimen center due to the eect of friction in the
compression test.5,19) The second feature is that the strain
accumulation is more pronounced and spreading in the
ECAP-annealed sample than the ECAP sample. This feature
is further demonstrated in Fig. 4(a) where "eq is delineated
with the distance from the specimen center along the
compression axis z as illustrated in the inset. For comparison, Fig. 4(a) also includes the results of the compression
tests up to compression of 49%. It is apparent that the area of
higher strain is extended more away from the center in the
ECAP-annealed sample than in the ECAP sample and this
extension tends to be reduced with an increase in the
compression from 49 to 73%. The third feature is that the
strain at the center is always higher for the ECAP sample
than the ECAP-annealed sample regardless of the magnitude
of compression. This feature is further documented in
Fig. 4(b) where the change in "eq at the specimen center for
each type of the sample is shown as a function of the
compression ratio: R % 100 uzz mm=12. The second
and third features must be the consequences arising from the
dierence in the -" curves shown in Fig. 1(b). For the
ECAP sample, the stress is decreased in the early stage of
compression test due to strain softening and hence, the strain
tends to be more released throughout the compression
anvil
compression
z
r
anvil
Equivalent strain
(a) ECAP sample
anvil
(b) ECAP-annealed sample
anvil
Fig. 3
Contours of equivalent strain after compression by 73% for each sample.
30
Measured area
center
(a)
(a)
eq
Equivalent strain,
73%
2.5
2
1.5
49%
40
Vickers hardness, Hv
ECAP
ECAP-annealed
3.5
35
30
Hv = 37 3.0 eq + 0.49 eq 2
( eq 3.0)
Initial hardness of
25
20
ECAP sample
(8 passes, route Bc)
0.5
0
0.5
1.5
2.5
Distance from center, z / mm
(b)
Equivalent strain at center,
eq
4
ECAP
3.5
ECAP-annealed
2.5
(b)
Vickers hardness, Hv
Equivalent strain, eq
40
35
30
Hv = 37 3.0 eq + 0.49 eq 2
25
Initial hardness of sample

annealed at 573K for 1 h
after ECAP
20
1.5
Fig. 5 Hardness variations with equivalent strain calculated by FEA for (a)
ECAP sample and (b) ECAP-annealed sample.
1
0.5
For the ECAP-annealed sample

0
10
20
30
40
50
60
70
80
Compression ratio, R / %
Fig. 4 (a) Variations of equivalent strain "eq along compression axis, z, and
(b) variation of "eq with compression ratio, R, at center of compressing
specimen.
specimen. For the ECAP-annealed sample, the stress is

increased in the early stage by strain hardening and hence,
the strained region spreads to soft areas in the compression
specimen.
4.3 Hardness variations with strain
Figure 5 plots the hardness against "eq for (a) the ECAP
sample and (b) the ECAP-annealed sample. Here, the
hardness data were measured along the symmetric axes
corresponding to r and z dened in the inset for the 49- and
73%-compressed samples. The values of "eq were used from
those calculated from the FEA at the measurements of
hardness. Correlation between hardness and equivalent strain
is represented by the following equations;
For the ECAP sample
Hv 37 3:0"eq 0:49"eq 2
where "eq 3:0
Hv 25 21:7"eq 12:3"eq 2 2:0"eq 3

where "eq 2:4
Now, the hardness distribution throughout the cross

section is depicted in Fig. 6 based on the strain distributions
shown in Fig. 3 and using eqs. (1) and (2) derived above.
Here, Fig. 6(a) is for the ECAP sample and Fig. 6(b) for the
ECAP-annealed sample. To check the validity of this
hardness distribution, hardness measurement was attempted
throughout the cross section for both ECAP sample and
ECAP-annealed sample and the results are shown in Fig. 6(c)
and (d), respectively. It is apparent that the hardness
distributions predicted using eqs. (1) and (2) are in good
agreement with those measured after compression, although
both prediction and measurement appear to be somewhat
dierent for regions near the outer edges. It is likely that this
dierence is caused by an eect of the shear deformation
because a large shear strain is imposed near outer edges by
compression test and because microstructural evolution and
hardness variation are reported to depend on the shear
deformation or deformation mode.8,2022) It should be noted
that this comparison has used calculation and measurement
from a 1/4 area of the cross section, provided that the
axisymmetry is preserved for the cylindrical compression
sample. It is seen that the hardness distribution is dierent
Hardness (Hv)
37.4
36.7
36.4
36.1
35.8
35.5
35.2
34.9
34.6
34.3
34.0
32.0
Prediction by numerical simulation

(a) ECAP sample
31
Experimental measurements
(c) ECAP sample
z
r
37.4
35.9
34.8
33.7
32.6
31.5
30.4
29.3
28.2
27.1
26.0
25.0
(b) ECAP-annealed sample
(d) ECAP-annealed sample
Fig. 6 Hardness distribution measured after compression by 73% and predicted from strain distribution obtained by FEA for (a), (c)
ECAP sample and (b), (d) ECAP-annealed sample.
between the ECAP sample and ECAP-annealed sample. For

the ECAP sample shown in Fig. 6(a) and (c), the hardness
follows a distribution with the maximum near the surfaces on
the longitudinal axis and with the minimum in the center
region. However, for the ECAP-annealed sample shown in
Fig. 6(b) and (d), the hardness maximum appears at the inner
regions and the hardness minimum forms near the surfaces
around the longitudinal axis.
4.3.1 For ECAP sample
For the ECAP sample, the hardness decreases with
increasing "eq . This decrease is inconsistent with an earlier
observation17) that the hardness remains constant with
straining after processing for more than 4 passes. It should
be noted that the ECAP sample used in this study was already
pressed for 8 passes before compression testing. Nevertheless, the hardness decreases as plotted in Fig. 5(a) and it is
considered that this must be due to the change in the strain
path from ECAP to compression. It was reported that the
hardness level tends to be lower for the cyclic rotation than
for one-way rotation in the HPT of an aluminum alloy and a
low carbon steel even when the number of revolution and
applied pressure are the same.20,21) The sample subjected to
ECAP through route Bc is processed by multidirectional
deformation2,5,7,8,14,22) but the sample is never deformed by
compression from the longitudinal direction parallel to the
extrusion direction. The change in the stain path from ECAP
to compression thus alters the local slip systems and creates
additional active slip systems. This must be the reason why
the hardness level further decreases with straining despite
the fact that a steady state is already established by ECAP
for 8 passes through route Bc. It was shown in earlier
studies,1,6,13,15) the microstructures formed by SPD processes
consist of grains with high angles of boundary misorientations but at the same time they are deformed structures so that
there is an appreciable fraction of low angle boundaries.
Figure 7 shows the boundary maps near the surface ("eq 0)
and at the center ("eq 4:0) after compression by 73% for
the ECAP sample. It is apparent that the fraction of low-angle
grain boundaries (LAGB) below 5 decreases with increasing
"eq and the fraction of high-angle grain boundaries (HAGB)

above 15 increases. Therefore, the decrease in hardness for
the ECAP sample is attributed to the changes in substructures
related to the stain path from ECAP to compression rather
than the formation of ne grains surrounded by HAGB.
4.3.2 For ECAP-annealed sample
The hardness behavior with "eq is dierent for the ECAPannealed sample as shown in Fig. 5(b). The hardness
increases initially with increasing "eq , takes a maximum at
"eq 1:3 and decreases thereafter. This behavior is very
similar to the ones obtained by ECAP and HPT.17,23) In
ECAP, the hardness increases with increasing number of
ECAP passes, takes a maximum at two passes ("eq 2:0),
decreases beyond the two passes and levels o at four passes
("eq 4:0).17) In HPT, the behavior is also similar such that
the hardness exhibits a maximum at "eq 2:0 and a steadystate begins to occur at "eq 6:0.23) For the ECAP-annealed
sample in the present study, the hardness maximum appears
because the ECAP samples were annealed before compression and therefore the microstructure developed by ECAP
was restored by the annealing. It is then reasonable that the
hardness behavior with straining became similar to those
obtained by ECAP and HPT. The boundary maps are shown
in Fig. 8 together with the fractions of LAGBs and HAGBs
for the ECAP-annealed sample. The map in Fig. 8(a)
corresponds to the strain where the hardness increases,
Fig. 8(b) at the strain where the hardness takes the maximum,
and Fig. 8(c) corresponds to the strain where the hardness
decreases. The fraction of LAGBs increases initially with
increasing "eq , takes a maximum at "eq 1:3 where the
hardness exhibits the maximum, and decreases with further
increasing "eq . The fraction of HAGBs exhibits an opposite
trend so that it reaches a minimum at "eq 1:3. In particular,
in Fig. 8(c) the majority of grains are slightly elongated in the
direction perpendicular to the compressive direction with a
size of 2 mm being surrounded by HAGBs. The microstructure seems to approach the one at the center in the ECAP
sample shown in Fig. 7(b). Furthermore, because the hardness is homogeneous across the samples strained largely by
32
(b) center(eq 4.0)

Compressive direction
(a) Near surface(eq 0)
15m
15m
Fraction of LAGB:67%,
HAGB:33%
HAGB:60%
Fig. 7 Boundary maps near surface and at center after 73% compression for ECAP sample. Red lines represent misorientation of 15
(HAGB) and black lines represent 2 < 15 (LAGB).
(b) eq 1.3
(c) eq 2.4
Compressive direction
(a) eq 0.8
15m
HAGB:27%
15m
15m
HAGB:19%
HAGB:43%
Fig. 8 Boundary maps with equivalent strain of (a) "eq 0:8, (b) "eq 1:3, (c) "eq 2:4 after 73% compression for ECAP-annealed
sample. Red lines represent misorientation of 15 (HAGB) and black lines represent 2 < 15 (LAGB).
HPT,23,24) it is anticipated that the hardness shown in

Fig. 5(b) levels o if hardness data would be available at
higher strain. The increase in hardness with strain up to "eq
1:3 is attributed to dislocation accumulation and formation of
subgrain boundaries. The subsequent decrease in hardness
from "eq 1:3 to "eq 2:4 is because of a decrease in
dislocation density within (sub)grains. It is considered that
the decrease in dislocation density occurs because some
dislocations are annihilated at (sub)grain boundaries. Such
softening behavior induced by decrease in dislocation density
is observed in some literatures using transmission electron
microscope23,24) or X-ray diractometry25) for ECAP, HPT,
and ARB processed pure Al.
Close comparison reveals that the hardness maximum
appears at a lower strain than those reported: "eq 1:3 for
this study but "eq 2 for the earlier studies.23) It is
considered that this may be due to the dierence in the
annealing conditions before intense straining. The ECAP
sample was annealed at 573 K for 1 h in this study but, in the

earlier studies, the annealing was conducted at 773 K for 1 h
after swaging or cold rolling. This dierence resulted in the
dierence in grain size: 50 mm for the former but 700 mm
for the latter. Harai et al. reported23) that dislocation
accumulation occurs with straining and reach a maximum
density giving rise to a maximum hardness. At the same time,
subgrains develop and the misorientation angles increase
with straining, as shown in Fig. 8. However, the dislocation
density decreases within the subgrains beyond the strain
corresponding to the maximum hardness because of the
absorption of dislocations in the grain boundaries which
happens more easily when the misorientation angle is large. It
is considered therefore that, if the initial grain size is smaller,
there may be more chances that the generated dislocations
are blocked at the grain boundaries. Thus, the hardness
maximum appears at a lower value of "eq and subsequently
enters into a steady state.
5.
Conclusions
The following conclusions are drawn from the comparison

between FEA and hardness measurements after compression
of the ECAP samples and ECAP-annealed samples.
(1) Both magnitude and inhomogeneity of equivalent
strain, "eq , depend on the initial microstructural conditions. The "eq at the center of the compressed
specimen is always higher for the ECAP sample than
the ECAP-annealed sample regardless of the magnitude
of compression.
(2) The hardness maximum appears for the compression
of the ECAP-annealed sample when the hardness is
plotted against the equivalent strain calculated through
FEA. The appearance of the hardness maximum is
consistent with the earlier reports using ECAP and HPT
processes. However, the strain at the hardness maximum is lower in this study and this can be attributed to
the dierence in initial grain size: 50 mm for this study
but 700 mm for the earlier studies.
(3) The hardness variation throughout the cross section
after compression is well predicted using the FEA for
both ECAP sample and ECAP-annealed sample.
(4) The change in the strain path from ECAP to compression aects the hardness variation with "eq .
Acknowledgements
This study was supported in part by a Grant-in-Aid for
Scientic Research from the Ministry of Education, Culture,
Sports, Science and Technology (MEXT), Japan in Priority
Areas Giant Straining Process for Advanced Materials
Containing Ultra-High Density Lattice Defects and on
Grant No. 20360339, and in part by IKETANI Science and
Technology Foundation.
33
REFERENCES
45 (2000) 103189.
2) R. Z. Valiev and T. G. Langdon: Prog. Mater. Sci. 51 (2006) 881981.
3) R. Z. Valiev, Y. Estrin, Z. Horita, T. G. Lngdon, M. J. Zehetbauer and
Y. T. Zhu: JOM 58 (2006) 3339.
4) Z. Horita, M. Furakawa, M. Nemoto and T. G. Langdon: Mater. Sci.
Technol. 16 (2000) 12391245.
5) T. Inoue, S. Torizuka and K. Nagai: Mater. Sci. Technol. 17 (2001)
13291338.
6) N. Tsuji: Tetsu-to-Hagane 88 (2002) 359369.
7) A. Belyyakov, T. Sakai, H. Miura and R. Kaibyshev: ISIJ Int. 39 (1999)
592599.
8) T. Inoue, F. Yin and Y. Kimura: Mater. Trans. 48 (2007) 20282035.
9) J. H. Lee, I. H. Son and Y. T. Im: Mater. Trans. 45 (2004) 21652171.
10) A. V. Nagasekhar and Y. T. Hon: Comp. Mater. Sci. 30 (2004) 489
495.
11) A. Aour, F. Zairi, J. M. Gloaguen, M. N.-Abdelaziz and J. M. Lefebvre:
Comp. Mater. Sci. 37 (2006) 491506.
12) H. S. Kim: Mater. Proc. Technol. 113 (2001) 617621.
13) T. Inoue, F. Yin and Y. Kimura: Mater. Sci. Eng. A 466 (2007) 114
122.
14) T. Inoue, T. Ochiai, F. Yin and K. Nagai: Tetsu-to-Hagane 93 (2007)
693702.
15) Y. Iwahashi Y, Z. Horita, M. Nemoto and T. G. Langdon: Acta Mater.
46 (1998) 33173331.
16) H. Hasegawa, S. Komura, A. Utsunomiya, Z. Horita, M. Furukawa, M.
Nemoto and T. G. Langdon: Mater. Sci. Eng. A 265 (1999) 188196.
17) Z. Horita, K. Kishikawa, K. Kimura, K. Tatsumi and T. G. Langdon:
Mater. Sci. Forum 558559 (2007) 12731278.
18) J. J. Hollomon: Trans. AIME 162 (1945) 268.
19) R. W. Evans and P. J. Scharning: Mater. Sci. Technol. 17 (2001) 995
1004.
20) Z. Horita and T. G. Langdon: Mater. Sci. Eng. A 410411 (1999)
422425.
21) Y. Todaka, M. Umemoto, A. Yamazaki, J. Sasaki and K. Tsuchiya:
Mater. Trans. 49 (2008) 4753.
22) T. Inoue and F. Yin: Mater. Sci. Forum 561565 (2007) 881884.
23) Y. Harai, Y. Ito and Z. Horita: Scr. Mater. 58 (2008) 469472.
24) C. Xu, Z. Horita and T. G. Langdon: Acta Mater. 55 (2007) 203212.
25) Y. Murata, I. Nakaya and M. Morinaga: Mater. Trans. 49 (2008)
2023.

Eect of Strain on Microstructural Evolution under Warm Deformation

in an Ultra-Low Carbon Steel
Joo-Hee Kang* , Tadanobu Inoue and Shiro Torizuka
The microstructural evolution was investigated according to the equivalent strain in uniaxial compression using an ultra-low carbon steel.
The equivalent strain was quantitatively calculated by using nite element analysis. The microstructure was rapidly rened up to the equivalent
strain of 2.4 with the material ow to rolling direction (RD) with the remarkable development of high angle boundaries. At that time, texture
such as (001)[110] and (111)[110] signicantly developed. Inhomogeneous deformation behavior according to texture was observed;
h100i == ND was elongated without the development of subgrain even under large strain, but h111i == ND was divided into small grains isolated
with high angle boundaries. [doi:10.2320/matertrans.MD200816]
(Received July 24, 2008; Accepted October 16, 2008; Published December 3, 2008)
Keywords: warm deformation, strain, texture, microstructure, electron backscattered diraction, ultra-low carbon steel
1.
Introduction
A microstructure is dependent on the parameters of

deformation processing, such as the temperature, strain,
strain rate, and deformation mode. Among them, the strain is
an important factor that aects the microstructure during
deformation. Recently, severe plastic deformation (SPD)
techniques have been developed for obtaining ultrane
materials, such as equal channel angular extrusion,1) multiaxis deformation,2) accumulative roll-bonding,3) and highpressure torsion.4) Although a ne microstructure can be
achieved through SPD processes, microstructural change by
processing parameters has not been quantitatively evaluated
in each deformation path because the deformation history is
not clear. Therefore, it is dicult to compare microstructures
formed by each SPD technique. Irrespective of deformation
processing, the equivalent strain, "eq , can be considered to
be a more accurate general physical parameter than the
reduction ratios of thickness or area. A fundamental understanding of the strain eect could provide a clue to the
control of the microstructure and texture in various kinds of
deformation processes.5)
On the other hand, in an earlier study,6) the microstructural
change of ferrite was reported under large-strain single
deformation with high strain rates by warm deformation
above 923 K. A strong texture, such as h100i, h111i == ND
and h110i == RD, was developed, and ne ferrite was
obtained under the large strain of 4. However, the mechanism
of such texture evolution diered from that of typical static
recrystallization, and the deformation behavior was not
thoroughly understood. In order to understand the mechanism
of grain renement and texture development in regions under
large strain,3,68) the microstructure and texture variation
should be determined according to "eq . In this study, the
eect of "eq on the microstructural evolution was investigated under a plane strain condition using nite element
*Corresponding author, E-mail: joohee@kims.re.kr. Present address: Grain
structure control research group, Korea Institute of Materials Science, 66

Sangnam, Changwon, Gyeongnam, 641-831, Korea
analysis (FEA) and electron backscattered diraction

(EBSD) when an ultra-low carbon steel was compressed
uniaxially at 773 K.
2.
Experimental
A uniaxial compression technique involving a plane strain

condition was used for thermo-mechanical processing.9,10) A
test specimen with dimensions of 18 mm 15 mm 12 mm
(width) was machined from a warm rolled bar with the
composition 0.0014C-0.30Si-0.20Mn (mass%).11) The specimen was machined to make the compression axis (ND)
parallel to the elongated direction of the as-received rolled
bar. A thermo-mechanical treatment simulator, Gleeble
2000, was used for the uniaxial compressive deformation,
which was applied parallel to the direction of the width with a
reduction ratio of 66%. The specimen was heated at a rate of
5 K/s to a soaking temperature of 773 K, and compressive
deformation was applied with a nominal strain rate of 0.1 s1
after soaking for 5 s. It was then quenched to room temperature using a water jet to freeze the microstructure. An
explicit FEA was used to evaluate the strain distribution in
the specimen compressed by a reduction ratio of 66%.
Simulation by the explicit FE code ABAQUS/Explicit was
performed and has been explained in detail elsewhere.12,13)
Here, to avoid confusing the directional notation in the
deformed specimen, the ND and RD correspond to the
compressive and elongated directions after compression,
respectively. All directions will be expressed as ND, RD,
and TD. On the basis of the calculated strain distribution,
the microstructural evolution of the TD section was observed in the regions without shear strain along the ND.
In our specimen, the eect of the compressive strain
considered under a plane strain condition of the observed
area was restricted to the middle part of the RD and TD.12,14)
The microstructure was measured by the EBSD method.
For the EBSD measurement, the specimen was polished
using a colloidal silica suspension after routine mechanical
polishing.
RD
ND
ND
max:6.52
Plastic strain ii, Equivalent strain eq
Eect of Strain on Microstructural Evolution under Warm Deformation in an Ultra-Low Carbon Steel
35
4
xx
yy
zz
eq
1
0
-1
y(ND)
z(TD)
x(RD)
-2
-3
0.0
0.5
1.0
1.5
2.0
y distance from center, mm

RD
max:7.48
Fig. 1 Boundary map and inverse pole gures of the ND and RD in the
non-deformed region. The black and gray lines show the high angle
boundary of 15 and the low angle boundary of 2 < 15 ,
respectively.
3.
The boundary map and inverse pole gures of the RD and

ND in a non-deformed region after the thermo-mechanical
procedure are shown in Fig. 1. Here, the black and gray lines
indicate the high angle boundary of 15 and the low
angle boundary of 2 < 15 , respectively. In order to
clearly examine the materials ow according to the strain, a
specimen was machined with its ND parallel to the elongated
direction of the as-received rolled bar, which consists of a
fully deformed matrix without recrystallized grains.11) In
addition, the uniaxial compression at a lower temperature of
773 K was performed to prevent microstructural change by
recrystallization during soaking in order to clearly observe
the materials ow by compressive deformation. The microstructure remained fully deformed and elongated parallel to
the ND, as shown by the high angle boundaries marked in
black in Fig. 1. Low angle boundaries developed within a
grain, which is a typical characteristic of a deformed matrix.
Some orientation components were somewhat stronger, but
we focused primarily on the materials ow and microstructural change according to "eq under the plane strain
condition.
Fig. 2 Distributions of the equivalent strain and compressive strain in the

x(RD), y(ND), and z(TD) directions in x z 0.
The distributions of "eq and each strain component, "xx ,

"yy , and "zz , in the x (RD), y (ND), and z (TD) directions in
x z 0 quantied by FEA are shown in Fig. 2. The
illustration of a deformed specimen in Fig. 2 clearly explains
the directional notation. As mentioned above, the observed
area was limited to the center area perpendicular to the TD
with x 0, and the microstructure was observed along the
ND to exclude the eect of shear strain, as shown in Fig. 2,
because microstructural evolution is considered to depend on
the shear deformation.5,14) Although the strain in the TD is
not completely zero, it is negligibly small compared with all
other normal strain components, and the deformation
condition is almost a plane strain compression.12) As the
position deviates from the center, the "eq decreases rapidly
according to the y distance from the center.
On the basis of this calculation, the microstructure was
observed in the regions of "eq of 1.1, 1.5, 2.0, 2.4, 3.1, and
3.5. The ND orientation maps are shown in Fig. 3. A black
line displays the high angle boundaries over 15 . The
microstructure changes rapidly over 2.0, and that below 2.0
(Fig. 3(a) and (b)) is similar to the non-deformed microstructure in Fig. 1. Under strain larger than 2.4, vigorous
subdivision of the microstructure occurs. On the other hand,
regarding the texture evolution, the components of h100i and
h111i parallel to the ND are strengthened, and subdivision
surrounded by high angle boundaries is signicantly promoted over 2.4. The materials ow to the RD (horizontal
direction in Figs. 3 and 4) is known from the change in the
pancaked grain shape and the development of high angle
boundaries. In Fig. 4, the RD orientation characteristics show
that (110) components were mostly developed and the (112)
component was locally promoted under larger strain. It is
noteworthy that a strong texture, such as h100i, h111i == ND
and h110i == RD, evolved under large strain and was
consistent with the results of low carbon steel compressed
under warm working temperatures.7,15)
In order to examine the sudden microstructural change, the
orientation maps were measured by low magnication in the
equivalent strain range between 1.1 and 2.9, as shown in
Fig. 5. Abrupt microstructural alteration accompanied with
grain elongation and renement began above the strain of 2.0.
36
J.-H. Kang, T. Inoue and S. Torizuka
(a)
RD
(b)
(c)
(e)
(f)
ND
(d)
10m
Fig. 3 ND orientation maps in the deformed region according to the equivalent strain ("eq ). (a) 1.1; (b) 1.5; (c) 2.0; (d) 2.4; (e) 3.1; (f) 3.5.
The black line indicates the high angle boundaries ( 15 ); magnication is identical in all maps.
(a)
RD
(b)
(c)
(e)
(f)
ND
(d)
10m
Fig. 4 RD orientation maps in the deformed region according to the equivalent strain ("eq ). (a) 1.1; (b) 1.5; (c) 2.0; (d) 2.4; (e) 3.1; (f) 3.5.
The black line indicates the high angle boundaries ( 15 ); magnication is identical in all maps.
ND
37
=1.1
RD
=1.5
=2.0
=2.5
70m
=2.9
Fig. 5 Orientation maps of the ND and RD that show the microstructural change and materials ow from "eq 1:1 to 2.9. Various strain
levels are marked in the RD map.
From a textural viewpoint, it is interesting that the orientation

rotation and development of high angle boundaries are
suppressed in f100gh011i, while the grains of f111gh112i are
inuenced by the signicant orientational change occurring
with grain subdivision. Generally, it is well known that
h111i == ND has a larger Taylor factor than h100i == ND in
BCC metals, which means that many slip systems are
activated and grain fragmentation is easily promoted in
h111i == ND.16) Additionally, some researchers have reported
that in-grain shear bands, which contribute to orientation
rotation and the formation of subgrains, developed in the
region of a larger Taylor factor17) and the most prevalent
large Taylor factor orientations were those of the -ber
(h111i == ND).18) These results can account for inhomogeneous deformation according to the orientation during warm
deformation.
Figure 6 shows the inhomogeneous deformation behavior
using local misorientation, the kernel average misorientation
(KAM).19) Figs. 6(a) and 6(c) present the boundary maps
with high angle boundaries (black line, 15 ) and low
angle boundaries (red line, 2 < 15 ), respectively. In
"eq 1:5, no apparent deformation behavior was observed
from the boundary map of Fig. 6(a) and Fig. 3(b), but many
traces of deformation bands by slip were evident within a

grain in the KAM map, as indicated by the white arrows in
Fig. 6(b). It is noteworthy that the deformation bands seldom
advance in the h100i == ND grains. As the strain increases,
small grains (fragmentations by deformation) isolated by
high angle boundaries are generated by the collapse of the
boundaries (black arrows in Fig. 6(c)). This kind of recrystallization corresponds to continuous recrystallization
without recognizable nucleation.7,20) In addition, the remarkable nature of the inhomogeneous deformation behavior
between h100i and h111i in the ND, already as explained in
Fig. 5, is noteworthy. Figures 6(c) and 6(d) shows that the
deformation in the h111i orientation concentrates more
intensively than that in the h100i orientation even under the
larger strain of "eq 3:1. Therefore, a subdivision easily
occurs by the deformation concentration in the h111i
component. However, the h100i component seldom shows
orientation variation within a grain, even under large strain.
In other words, h100i grains turn to an elongated shape
adapted with compressive deformation without signicant
subdivision and rotation of orientation. Similar results have
been reported in a warm deformation behavior in low carbon
steel and IF steel.3,7) With respect to the inhomogeneous
38
J.-H. Kang, T. Inoue and S. Torizuka
(a)
(b)
100
10m
(c)
(d)
100
111
10m
100
Fig. 6 Boundary characteristics according to the equivalent strain, "eq , and distributions of local misorientation. (a), (b): "eq 1:5; (c),
(d): "eq 3:1. (a) and (c) are boundary maps with high angle boundaries ( 15 ) marked in black and low angle boundaries
(2 < 15 ) marked in red, respectively. (b) and (d) show the kernel average misorientation.
deformation by texture, Choi et al., using the crystal

plasticity nite element method, explained that h100i grains
parallel to ND have lower stored energy.21,22) Similarly, Tsuji
et al. reported that (001)[100] and (001)[110] grains in BCC
crystal are very stable without rotation during deformation
and that grains with other orientations easily rotated to
(001)[110] or (111)[110].23) These results are similar to those
obtained in previous research when the specimen was
deformed by the strain of 4 at 9231123 K.6) Such orientation
characteristics are evidence of continuous dynamic recrystallization unaccompanied by apparent nucleation. Nevertheless, a more in-depth quantitative examination on texture
evolution is needed during warm deformation, including
the strong texture development of (001)[110] at higher
temperature.6)
Regarding the mechanism of grain renement through
severe plastic deformation, some researchers have proposed
the microshear band mechanism, whereby local lattice

rotation takes places at the intersection of microshear
bands.17,24) In our specimen, deformation bands inclined
20 to the RD are actively generated in h111i == ND, as
shown in Figs. 3(c) and 3(d). It is also interesting that most
of the new grains (Fig. 6(c)) form near grain boundaries
or along the deformation bands. Dorner et al. reported that
microbands with periodic lattice rotation occurred only near
the boundary and the boundary plays an important role in the
formation of these bands by altering or restricting the active
slip systems in the vicinity of the grain boundaries.24)
Nevertheless, a precise observation of the microstructure
and textural evolution near deformation bands with higher
magnication is needed for better understanding of the
mechanism by grain renement.
Figure 7 displays the change of the boundary characteristics as a function of "eq imposed by compression. The
60
230
low angle boundary
medium angle boundary
high angle boundary
220
Hardness, HV(1N)
Fraction, %
50
40
30
20
210
200
190
180
170
160
10
0
1.5
Fig. 7 Change in the boundary characteristics according to the equivalent

strain. The low angle is 2 < 5 , and the medium angle and high angle
are 5 < 15 and 15 , respectively.
boundary is divided into three kinds of misorientation angles,

i.e., low angle (2 < 5 ), medium angle (5 < 15 ),
and high angle ( 15 ) boundaries. Under the strain of 2.4,
the boundary characteristics in the deformed region are
similar to those of the undeformed region. However, when
the "eq exceeds 2.4, the boundary distribution changes
drastically, high angle boundaries develop, and the fraction
of the low angle boundaries decreases. The fraction of the
medium angle boundaries increases under the strain of 3.1,
but it decreases under the strain of 3.5. It is considered that
most boundaries become high angle boundaries in an area
under large strain because grain fragmentation is accompanied by the continuous recrystallization. These results are
consistent with the microstructural evolution according to
"eq , as presented in Figs. 3 and 4.
On the other hand, the distribution of the Vickers hardness
was measured under a load of 1N with respect to "eq , as
shown in Fig. 8. The hardness increases as the "eq becomes
larger. The increment in the hardness can be understood
because of work hardening through the accumulation of the
dislocation and grain renement by the increase of the strain.
The deviation of hardness for each strain level decreases as
the "eq becomes larger, and it is considered that average
value of hardness is reected from a lot of ne grains.
Furthermore, this result corresponds to the microstructural
evolution in Figs. 35. From the distribution of the hardness,
we can determine that grain renement by fragmentation is
an eective method to increase the strength in warm
deformation.
Conclusion
The microstructural evolution was studied according to the

equivalent strain, "eq , when an ultra-low carbon steel was
deformed under uniaxial compression at 773 K. The microstructure was rened over the strain of 2.4, and a strong
texture, such as (001)[110] and (111)[110], developed at that
time. In addition, the inhomogeneous deformation behavior
was examined, in which grain renement occurs mainly in
h111i == ND due to the development of high angle boundaries. However, the (001) component was stable and elongated
2.0
2.5
3.0
3.5
4.
39
Fig. 8
Hardness variations with the equivalent strain.
to RD even under large strain. The microstructure and texture

evolution indicated the characteristics of continuous recrystallization unaccompanied with recognizable nucleation.
Furthermore, the high angle boundaries were remarkably
promoted at "eq above 2.4. The increase of the hardness by
grain renement was measured by the Vickers hardness and
was consistent with the microstructural evolution during
warm deformation.
REFERENCES
1) V. M. Segal: Mater. Sci. Eng. A 271 (1999) 322333.
2) A. Belyakov, T. Sakai, H. Miura and R. Kaibyshev: ISIJ Int. 39 (1999)
592599.
3) N. Kamikawa, T. Sakai and N. Tsuji: Acta Mater. 55 (2007) 5873
5888.
4) I. Saunders and J. Nutting: J. Metal Sci. 18 (1984) 571575.
5) T. Inoue, F. Yin and Y. Kimura: Mater. Trans. 48 (2007) 20282035.
6) J.-H. Kang and S. Torizuka: Scr. Mater. 57 (2007) 10481051.
7) R. Song, D. Ponge, D. Raabe and R. Kaspar: Acta Mater. 53 (2005)
845858.
8) S. V. S. Narayana Murty, S. Torizuka and K. Nagai: ISIJ Int. 45 (2005)
16511657.
9) A. Ohmori, S. Torizuka, K. Nagai, K. Yamada and Y. Kogo: Tetsu-toHagane 88 (2002) 857864.
10) S. V. S. Narayana Murty, S. Torizuka, K. Nagai, T. Kitai and Y. Kogo:
Scr. Mater. 53 (2005) 763768.
11) S. Torizuka, E. Muramatsu, S. V. S. Narayana Murty and K. Nagai: Scr.
Mater. 55 (2006) 751754.
12) T. Inoue, S. Torizuka, K. Nagai, K. Tsuzaki and T. Ohashi: Mater. Sci.
Tech. 17 (2001) 15801588.
13) T. Inoue, S. Torizuka and K. Nagai: Mater. Sci. Tech. 18 (2002) 1007
1015.
14) J.-Y. Cho, T. Inoue, F. Yin and K. Nagai: Mater. Trans. 45 (2004)
29661973.
15) S. V. S. Narayana Murty, S. Torizuka and K. Nagai: ISIJ Int. 45 (2005)
16511657.
16) G. Y. Chin and W. L. Mammel: Trans. AIME 239 (1967) 14001405.
17) M. R. Barnett: ISIJ Int. 38 (1998) 7885.
18) L. Kestens and J. J. Jonas: Metall. Trans. 27A (1996) 155164.
19) S. I. Wright, D. P. Field and D. J. Dingley: Electron Backscatter
Diraction in Materials Science, (Kluwer Academic, New York, 2000)
pp. 148149.
20) F. J. Humphreys and M. Hatherly: Recrystallization and related
annealing phenomena, (Elsevier, Oxford, 2004) pp. 415467.
21) S.-H. Choi: Acta Mater. 51 (2003) 17751788.
22) S.-H. Choi and J. H. Cho: Mater. Sci. Eng. A 405 (2005) 86101.
23) N. Tsuji, K. Tsuzaki and T. Maki: ISIJ Int. 32 (1992) 13191328.
24) D. Dorner, Y. Adachi and K. Tsuzaki: Scr. Mater. 57 (2008) 775778.

Coupled Analysis of Heat Transfer and Deformation

in Equal Channel Angular Pressing of Al and Steel
Pham Quang1 , A. Krishnaiah2 , Sun Ig Hong3 and Hyoung Seop Kim4; *
1
Faculty of Material Science and Technology, Hanoi University of Technology, Hanoi, Vietnam
Department of Mechanical Engineering, University College of Engineering (Autonomous),
Osmania University, Hyderabad-500 007, India
3
Department of Nano Materials Engineering, Chungnam National University, Daejeon 305-764, Korea
4
Department of Materials Science and Engineering, POSTECH, Pohang 790-784, Korea
2
In the present paper, the heat transfer and plastic deformation during the equal channel angular pressing of Al alloy and steel were
numerically analyzed. The nite element simulations were carried out at various pressing speeds 0:18, 1.8, and 18 mm/s, and the
temperature and plastic strain distributions were investigated. It is concluded that the pressing speed is the most important processing parameter
in rising temperature. Temperature rise in steel is higher than that in Al alloy due to the combined eect of high strength (heat generation eect)
and low thermal diusivity (heat release eect) in steel. [doi:10.2320/matertrans.MD200823]
(Received September 29, 2008; Accepted October 31, 2008; Published December 25, 2008)
Keywords: equal channel angular pressing, severe plastic deformation, nite element method, heat transfer analysis
1.
Introduction
Equal channel angular pressing (ECAP) is an attractive

severe plastic deformation (SPD) technique for the development of ultrane grained (UFG) microstructures in metals
and alloys by imposing large strain, without changing the
cross-sectional dimensions of a billet.14) The production of a
UFG microstructure using ECAP results from the intense
plastic strains introduced in the metallic material that drives
the recovery/recrystallization in the strained materials. A
schematic of the ECAP die setup is shown in Fig. 1. The
ECAP die consists of two (entry and exit) equal channels that
intersect at an angle usually between 90 and 135 . The main
advantage of the ECAP process is that the process can be
repeated several times using the same specimen to impose
large plastic strains leading to the production of UFG
microstructures because of its capability of maintaining the
net dimensions of the workpiece.
Numerous experimental and simulation works on the
deformation behavior of workpiece in ECAP have been
presented recently. Most of the studies on ECAP published to
date are concerned mainly with the ability to produce UFG
microstructures and the unique deformation geometry of the
workpiece in ECAP. The evolution of the microstructures
and physical properties of the SPD processed material is
related directly to the histories of plastic deformation and
temperature, and hence the understanding of the phenomenon
associated with strain and temperature developments is very
important for the successful manipulation of ECAP process.
There are many publications on the deformation behavior
during ECAP by using various numerical methods. However,
only a few activities on the analysis of temperature and heat
transfer have been reported, in comparison to those of
deformation and microstructural behavior. To the best of the
authors knowledge, nobody attempted the nite element
method (FEM) simulation of ECAP process to predict the
*Corresponding
author, E-mail: hskim@postech.ac.kr
Fig. 1
Die and workpiece geometry of ECAP.
temperature rise. Nishida et al.5) conducted an experimental

investigation on two Al alloys, one Mg alloy and a metal
matrix composite to measure the billet temperature rise
during ECAP at a constant pressing speed. Yamaguchi et al.6)
investigated the signicance of adiabatic heating in ECAP on
Al alloys. Kim made heat transfer analyses using a lumped
analytical method in ECAP7) and the nite volume method
(FVM) in equal channel multi-angular pressing (ECMAP)8)
to predict the temperature rises.
Practically, the workpiece temperature in ECAP is
controlled through the preheating of the workpiece. However, the exact workpiece temperature during an ECAP
operation may be signicantly above the preheating temper-
Coupled Analysis of Heat Transfer and Deformation in Equal Channel Angular Pressing of Al and Steel
2.
Finite Element Analyses
Figure 1 represents the schematic of ECAP process,

showing die, punch, workpiece, and important design angles
(channel angle and corner angle ). The points with
numbers show reference positions in the workpiece along
longitudinal direction for tracking thermal history. A coupled
deformation and heat transfer FEM simulation of ECAP was
carried out using the rigid-plastic commercial FEM code
DEFORM-3D ver. 6.1.9) It should be noted that a twodimensional simulation cannot correctly reproduce the
temperature responses because heat transfers along the
normal to the entry and exit channel directions as well as
parallel to the channel directions, although two-dimensional
deformation simulation is acceptable due to the plan strain
condition of deformation in ECAP. A typical value of friction
factor m 0:1 for cold metal forming was considered
between the inner surfaces of the die channels and the outer
surface of the workpiece. The simulations were carried out at
three dierent punch speeds; 0:18, 1.8, and 18 mm/s.
The numbers of initial elements and nodes of an 8-node
hexagonal element were 45,105 and 10,315, respectively.
This number of elements was found to be sucient to
express the local deformation and heat of the strain rate
insensitive materials through the calculations with varying
the number of elements in DEFORM-3D.
The simulations were carried out using a die with die
channel angle 90 and outer corner angle 20 for
single pass. Die and punch were modeled with rigid parts.
Cylindrical specimen with dimensions of diameter 10 mm
height 60 mm was used for all simulations. The stress-strain
curves of the materials (AISI1015 steel and Al 6063 alloy) at
various temperatures supported by the DEFORM database9)
are shown in Fig. 2, where strain hardening and temperature
dependent ow curves are clearly observed. In the present
study, the inuence of processing speeds during the ECAP of
Al alloy and steel in terms of temperature distribution was
also investigated. The mechanical and thermal constants of
Al alloy and steel used in ECAP simulation are shown in
Table 1.
3.
The simulated equivalent strain distribution fringes with

deformed geometries for Al alloy and steel are represented in
800
Steel (AISI1015)
20.C
300C
400C
600C
Al6063
20C
50C
150C
250C
700
Stress, / MPa
ature due to the heat generated by the plastic work. It is

therefore necessary to estimate the actual workpiece temperature during the ECAP for a precise control of recovery/
recrystallization, and hence of the grain size. The temperature
rise during the ECAP is considered to depend on both the
sample material and the deformation rate (pressing speed). In
the present study, nite element analyses were conducted to
analyze the temperature histories and distributions in the
workpiece during the ECAP of Al alloy and steel. In order to
better understand both temperature distribution and strain
characteristics, and in order to determine the pressing load
during the processing of Al alloy and steel, FEM simulations
were carried out for ECAP at room temperature under
dierent pressing speeds.
41
600
500
400
300
200
100
0.0
0.2
0.4
0.6
0.8
1.0
Strain
Fig. 2
Stress-strain curves of Al alloy and steel used in FEM simulations.
Table 1 Mechanical and thermal constants of Al alloy and steel used in

ECAP simulation.
Aluminum
Steel
Youngs modulus E, GPa
Properties
68.9
210
Poissons ratio
0.33
2700
0.3
7833.4
22 106
12 106
Density, kg/m3
Coecient of thermal, C1
Steel (AISI 1015, 201100 C)
Temperature,

C
Thermal conductivity,
W/(mK)
Specic heat, J/(kg C)
100
50.708
3.809
199
48.112
4.039
299
45.689
3.125
399
41.718
4.598
499
38.279
5.090
Aluminum (Al 6063, 20250 C)

20250
180.2
2.433
Al
Steel
Fig. 3 Equivalent strain distributions in Al alloy and steel pressed at speed
of 1:8 mm/sec.
Fig. 3. The strain distributions are almost the same for Al and
steel because the die geometries are the same, although the
material properties and thermal responses are dierent.
Strains are the same in all cases of ram speeds ( 0:18, 1.8
and 18 mm/s). The independence of strain and geometry with
pressing speed is due to the strain rate insensitivity of ow
behavior used for the simulations. The developed strain value
42
P. Quang, A. Krishnaiah, S. I. Hong and H. S. Kim

5
1.0x10
1.2
Steel (AISI 1015)
4
8.0x10
1.0
Load, F / N
Strain
0.8
0.6
Al
Steel
0.4
6.0x10
v = 0.18 mm/s
v = 1.8 mm/s
v = 18 mm/s
4.0x10
Aluminium(AI 6063)
2.0x10
0.2
0.0
0
Bottom
Fig. 4
Distance, d / mm
10
Top
Path plot in Al alloy and steel pressed at speed of 1:8 mm/sec.
Aluminium
10
20
30
40
50
60
Distance, d / mm
Fig. 5
Load-time plot of Al alloy and steel at various pressing speeds.
Steel
(a)
(b)
(c)
Fig. 6 Temperature distributions in Al alloy and steel pressed at speeds of (a) 0:18 mm/sec; (b) 1:8 mm/sec; and (c)
18 mm/sec.
( 1) in the middle region is in good agreement with results

of analytical solution value7) after ECAP by using 90
and angle 20 for one pass. Strain at the inner corner is
higher than that in the outer corner in the main shear
deforming zone. The low strain in the bottom region is due to
the eect of round corner and strain hardening characteristic
in the ow behavior of the material.10) The strain uniformity
can be observed in the middle (steady state region) of the
sample along the exit channel direction of the workpiece.
Comparing with the other regions, the equivalent plastic
strains are much more uniform and larger along the exit
channel direction in this steady deformation region. The
strain distribution along the radial (width) direction in the
steady region is shown in Fig. 4. Due to the strain hardening
eect,10) strain in the bottom region is smaller than the other
upper region.
Figure 5 shows the load-time history plots at various
pressing speeds. The steel exhibited a high pressing load
(80 kN) as compared to Al alloy (16 kN) in all cases of
pressing speed, because of the intrinsic harder properties of
steel than Al alloy. It should be noted that stress levels in steel
is about 5 times of those of Al at the same strain level as
shown in Fig. 2, which matches well with the ratio of

forming loads in Al and steel as depicted in Fig. 5.
The temperature contours of Al alloy and steel are plotted
in Fig. 6, for dierent pressing speeds. The temperature
increased with the increasing pressing speed, contrary to the
results of strain which are insensitive to the pressing speed.
The internal heat generation due to the plastic work within
the main deforming zone of the specimen is transferred to the
ECAP die and the remaining energy is stored. The internal
heat generation due to the plastic work is usually around
90% of the work of plastic deformation.11) According to the
authors analytical derivation,7) the temperature rise T of
the workpiece is
p A
0:9" 0:5m= 3u t
V
;
T
A
C ht
V
where , ", , C, V, A, t, m, u and h denote stress, strain, the
density of the material, its heat capacity, the volume of the
calculation domain (i.e. of the main deforming zone), the
outer surface area of the domain contacting the die, the dwell
Coupled Analysis of Heat Transfer and Deformation in Equal Channel Angular Pressing of Al and Steel
among the six points during the ECAP of Al alloy and steel,
but at higher pressing speeds ( 18 mm/s) there is a
notable temperature rise between the six points. The temperature of Al increased in the range of 11.7 C18 C for low
pressing speeds and an average of 40 C for higher pressing
speed. On the other hand, temperature rise in steel is high
(150 C) especially a high pressing speed, which is the
combined eect of high strength (heat generation eect) and
low thermal diusivity (heat release eect).
70
P.1
P.2
P.3
P.4
P.5
P.6
Temperature, T / C
60
v=18mm/s
50
40
v=1.8mm/s
30
43
4.
Conclusions
v=0.18mm/s
20
0
10
20
30
40
50
60
70
Distance, d / mm
Al
220
P.1
P.2
P.3
P.4
P.5
P.6
200
180
Temperature, T / C
160
v=18mm/s
140
120
100
80
v=1.8mm/s
60
40
v=0.18mm/s
20
0
10
20
30
40
50
60
The inuence of pressing speed on the deformation

temperature of Al alloy and steel during the equal channel
angular pressing was studied by the coupled heat-deformation 3D-nite element method. The homogeneous plastic
deformation is obtained in the middle of the sample for all
processing speeds. For the high speed condition ( 18 mm/
s) sample pressed during the ECAP of steel, higher temperature rise obtained is compared to the case of low pressing
speed. Due to the deformation induced heat release, the
temperature rise during the ECAP with high ram speed,
especially for high strength materials, can be as large as to
raise the workpiece temperature above the recovery temperature or the grain growth temperature. Temperature rise in
steel is higher than that in Al alloy due to the combined eect
of high strength (heat generation eect) and low thermal
diusivity (heat release eect) in steel. The heat transfer
analysis presented in this study can be used as an informative
method to control the ECAP process parameters.
70
Distance, d / mm
Acknowledgments
steel
Fig. 7
Temperature history in Al alloy and steel.
time of the domain within the deforming zone, friction factor,

punch velocity, and the heat transfer coecient between the
workpiece and the die, respectively. Hence, the temperature
increment in steel is higher than that in Al due to the higher
deformation heat, all other conditions being the same.
Figures 6 and 7 show the eect of the pressing velocity,
the most important factor governing the temperature rise.
With increasing pressing velocity, temperature rises in both
Al and steel, due to the lack of time for the heat release from
the workpiece to the die. Figure 7 also show overshooting
and stabilizing stages, which cannot be observed in the
analytical solution.7) This overshoot is smaller in steel having
lower thermal diusivity (thermal conductivity/density heat capacity) that in Al alloy. The variation of
temperature at six points has been monitored by the particle
tracking method. Figure 7 shows that for low pressing speeds
( 0:18 and 1.8 mm/s) there is no temperature dierence
This research was supported by a grant from the Center for

Advanced Materials Processing (CAMP) of the 21st Century
Frontier R&D Program funded by the Ministry of Science
and Technology, Republic of Korea.
REFERENCES
1)
2)
3)
4)
5)
6)
7)
8)
9)
10)
11)
V. M. Segal: Mater. Sci. Eng. A 197 (1995) 157.

R. Z. Valiev: Metall. Mater. Int. 7 (2001) 413.
I. V. Alexandrov: Metall. Mater. Int. 7 (2001) 565.
Y. Fukuda, K. Oh-ishi, Z. Horita and T. G. Langdon: Acta Mater. 50
(2002) 1359.
Y. Nishida, T. Ando, M. Nagase, S. Lim, I. Shigematsu and A. Watazu:
Scr. Mater. 46 (2002) 211.
D. Yamaguchi, Z. Horita, M. Nemoto and T. G. Langdon: Scr. Mater.
41 (1999) 791.
H. S. Kim: Mater. Trans. 42 (2001) 536538.
H. S. Kim: Mater. Sci. Eng. A (2008) in press.
SFTC, DEFORM-3D ver. 6.1.
H. S. Kim, M. H. Seo and S. I. Hong: Mater. Sci. Eng. 291A (2000) 86.
S. Kobayashi, S.-I. Oh and T. Altan: Metal Forming and the FiniteElement Method, (Oxford University Press, Oxford, 1989) pp. 228
229.

Evolution of Mechanical Properties and Microstructures with Equivalent Strain

in Pure Fe Processed by High Pressure Torsion
Kaveh Edalati* , Tadayoshi Fujioka and Zenji Horita
Department of Materials Science and Engineering, Faculty of Engineering, Kyushu University, Fukuoka 819-0395, Japan
Pure Fe (99.96%) was processed by high pressure torsion (HPT) using disc and ring samples. When the microhardness and tensile
properties are plotted against the equivalent strain, the individual properties fall well on unique single curves, level o at the equivalent strain of
40. At the saturated level, the tensile strength of 1050 MPa and the elongation to failure of 2% are attained. Transmission electron microscopy
showed that a subgrain structure containing dislocations develops at an initial stage of straining. More dislocations form within the grains and the
subgrain size decreases with further straining. At the saturation stage, the average grain size reaches 200 nm, the misorientation angle increases
and some grains which are free from dislocations appear. It is suggested that at the saturation stage, a steady state condition should be established
through a balance between hardening by dislocation generation and softening by recrystallization. [doi:10.2320/matertrans.MD200812]
(Received July 8, 2008; Accepted October 3, 2008; Published November 19, 2008)
Keywords: severe plastic deformation, high pressure torsion, equivalent strain, iron, microhardness, tensile strength
1.
Introduction
Processing via severe plastic deformation (SPD) has been

an attractive research topic in recent years. The grain size is
signicantly rened through the SPD and thus superior
mechanical properties are often observed in the SPDprocessed metallic materials. Several techniques are available for the SPD process but high pressure torsion (HPT) is
especially eective to introduce extremely high strain and
thus produce metallic materials having ultrane grain
sizes.13) The HPT process has advantages over other
techniques that it is applicable to hard and less ductile
materials4) and also it is possible to consolidate powders
without heating.5) A major drawback of the HPT process is
that the sample is in a form of disc and the size of the disc is
generally limited to 10 mm in diameter in order to keep a
high applied pressure. Because the strain is generated in
proportion to the distance from the disc center, an inhomogeneous distribution of microstructure is developed across
the diameter: coarser at the center because of lower strain but
ner at the outer edge due to intense strain.610) In order to
overcome this inhomogeniety of microstructure and resultant
inhomogeniety of mechanical properties, a ring-shaped
sample was introduced in the HPT process11) with a simple
modication of the earlier design.12,13) Using the ring HPT, it
is possible not only to eliminate a center part with a lessstrained and coarse-microstructure region but also to increase
the diameter of the sample by the amount corresponding to
the center part.
The ring HPT was applied to pure Al11) and pure Cu14) and
it was demonstrated that the variation of hardness has a
unique function of equivalent strain. It was also conrmed
that the ultrane-grained structures and mechanical properties are homogeneous within the ring having 3 mm width.
However, it was shown that the hardness variation with
respect to the equivalent strain is very dierent between Al
and Cu despite the fact that the purity level is the same and as
*Corresponding
ac.jp
author, E-mail: kaveh.edalati@zaiko6.zaiko.kyushu-u.
high as 99.99% for both metals. The dierence in microstructures and mechanical properties may be due to the
dierence in the melting point, the stacking fault energy
and/or the crystal structure. In this study, therefore, the
application of the ring HPT is extended to pure Fe having
dierent physical properties than Al and Cu and the microstructural evolution and mechanical behavior are examined
with respect to equivalent strain as investigated earlier.
There are several studies reported on Fe by SPD.1523)
Although the studies on HPT of Fe1520) are concerned with
pure Fe (so called Armco iron), little information was given
for the content of C except the studies of Todaka et al.16,20) It
appears that the C inclusion is sensitive to the strength and
microstructure. It also appears that the strength is dependent
on the total imposed strain and on the strain path for the grain
renement when Fe samples are processed by SPD. In fact,
Wetscher and Pippan18) reported that the hardness is dierent
depending on whether it is rotated in a forward direction or a
cyclic direction in HPT processing. In this study, the purity
level of 99.96% with the C content of 11 ppm was used to
minimize any possible eect of C. Furthermore, the hardness
as well as the tensile properties including microstructure is
examined in a wide range of equivalent strain with the single
forward rotation so that the present results are directly
compared with the earlier ones of pure Al and Cu.11,14,25,26)
2.
Experimental Materials and Procedures
High purity iron (99.96%) of chemical composition given

in Table 1 was received in a form of sheet with a thickness of
1 mm and an area of 220 mm 150 mm after rolling a billet
from 90 to 3 mm at 1373 K by 12 passes and further reducing
the thickness to 1 mm by one pass at room temperature. The
sheet was cut to square pieces having dimensions of 1 mm
40 mm 40 mm and annealed at 1273 K for one hour under
pure argon atmosphere. Discs with 10 mm diameter and rings
with inner and outer diameters of 24 and 30 mm were cut
from these plates using a wire-cutting electric discharge
machine and their thicknesses were further reduced to
0.8 mm by grinding.
Evolution of Mechanical Properties and Microstructures with Equivalent Strain in Pure Fe Processed by High Pressure Torsion
45
Table 1 Chemical compositions of Fe (all compositions are in ppm except for Al which is given in mass%).
Element
Si
Mn
Cr
Cu
Mo
Nb
Al
Ti
Amount
11
<30
<30
<20
<3
<30
<30
<30
<20
<30
0.03
<20
<2
14
Fig. 1 Schematic illustration of HPT facilities for (a) disc and (b) ring
samples.
Fig. 2
Fig. 3 Dimensions of (a) disc and (b) ring samples including positions for
microhardness measurements and locations for TEM discs and tensile
specimens.
Appearance of disc (left) and ring (right) samples after HPT for 1 revolution.
HPT was conducted using the facilities as schematically

illustrated in Fig. 1 for disc (left) and ring (right) samples.
The facilities consist of upper and lower anvils having a
shallow hole of 10 mm diameter and 0.25 mm depth for the
disc sample. For the ring sample, the upper and lower anvils
have a shallow circular groove with the inner and outer
diameters of 24 and 30 mm with the groove depth of
0.25 mm. The hole on the anvil for the disc and the groove on
the anvil for the ring were at bottom which were roughened
and subsequently nitried to attain a surface roughness of
30 mm. This surface roughness was conrmed by laser
scanning microscopy at three dierent locations on the
roughened surfaces of the anvils. Each sample was placed on
the hole or on the groove and the upper and lower anvils were
rotated with respect to each other at room temperature with a
rotation speed of 0.2 rpm under a selected pressure of 2 GPa.
The rotation was terminated after a revolution of either 1/8,
1/4, 1, 2, 4, 10 or 20. Figure 2 shows the appearance of disc
and ring samples after HPT for one revolution. The surfaces
of the samples after HPT were in almost the same roughness

as the roughened surface of the hole or the groove.
The disc and ring samples subjected to HPT were
evaluated in terms of Vickers microhardness, tensile properties and microstructures. First of all, both ring and disc
samples were polished to a mirror-like surface and the
Vickers microhardness was measured along the radii from
the center to periphery at 8 dierent radial directions with
0.5 mm increments as drawn by dotted lines in Fig. 3. The
disc and ring samples processed for one revolution were cut
to half and, after polishing, the hardness was measured across
the thickness at every 0.08 mm from the bottom surfaces of
the samples. For each hardness measurement, a load of 200 g
was applied for 15 seconds. Second, the disc and ring
samples were ground and polished to reduce the total
thickness to 0.3 mm. Thereafter, as illustrated in Fig. 3,
miniature tensile specimens having 1 mm gauge length and
1 mm width were cut from both types of the samples using
the wire-cutting electric discharge machine. Each tensile
46
K. Edalati, T. Fujioka and Z. Horita
specimen was mounted horizontally on grips and pulled to

failure using a tensile testing machine with an initial strain
rate of 3:3 103 s1 . The stress-strain curve was delineated
for each specimen and the ultimate tensile strength and
elongation to failure were measured from the curves. To
obtain the average, at least three tensile tests were repeated
on the samples processed at the same HPT conditions. Third,
discs with 3 mm in diameter were punched from the HPT
discs at 2 mm away from the center and from the center of the
ring width as also illustrated in Fig. 3. The 3 mm discs were
ground mechanically to a thickness of 0.1 mm and further
thinned for transmission electron microscopy (TEM) with a
twin-jet electro-chemical polisher using a solution of 10 vol%
HClO4 and 90 vol% CH3 COOH at 298 K. A transmission
electron microscope was operated at 200 kV for the observation of microstructures. Selected area electron diraction
(SAED) patterns were taken from a region with 6.3 mm
diameter to complement the TEM observations.
3.
Results and Discussions
Figure 4 plots microhardness as a function of distance

from the centers of disc and ring samples after revolutions
from 1/8 to 20. Microhardness increases with increasing
distance from the center for the disc samples but the hardness
is almost constant across the width for the ring samples. For
both disc and ring samples, the microhardness increases with
an increasing number of revolutions. The saturation of the
hardness level appears in the disc samples after two or more
revolutions and in particular the hardness values after 20
revolutions lie close to this saturated level.
All hardness values in Fig. 4 are plotted as a function of
equivalent strain in Fig. 5 as attempted in earlier papers.8,13,14) Here, the equivalent strain is calculated using
ZN
2r
p
dN
1
" 1 s
3tN
0
where r is the distance from the center of disc (or ring), N is
the number of revolutions, s is the fraction of sample slippage
(with s 0:17 in this study) and tN is the thickness of disc
(or ring) representing as a function of N. Using eq. (1), the
Fig. 4 Vickers microhardness plotted against distance from center for disc
and ring samples processed after various revolutions.
Fig. 5 Vickers microhardness plotted against equivalent strain for disc and
ring samples processed after various revolutions (all data points shown in
Fig. 4).
eects of thickness reduction and sample slippage are

considered. The importance of this equation is given in the
Appendix. It is found that for both disc and ring samples all
data points fall well on a single curve. The hardness increases
with an increase in the equivalent strain at an early stage of
straining but levels o and enters into a steady state where the
hardness remains unchanged with further straining. There are
three points to be noted regarding the plots in Fig. 5.
First, although there is a certain deviation in hardness
between the disc sample and the ring sample at low
equivalent strain, overall consistency is well held for the
two types of samples. Therefore, it is concluded that the ringshaped sample is used as an alternative for the disc samples in
HPT.
Second, despite the fact that the equivalent strain is
theoretically zero at the center of disc according to eq. (1),
close inspection of Fig. 4 reveals that the hardness at the disc
center increases with increasing N, especially for the
revolution greater than two. In practice, hardness measurement at the exact center is dicult but the hardness values
were measured within a circular area with 0.2 mm in
diameter. Furthermore, a small misalignment of the rotation
axes between the lower and upper anvils may result in
imposing an additional shear strain to the center of the disc,
although the alignment was adjusted well within 0:01 mm
in this study.
Third, the following important points arise from comparisons with the plots of pure Al13) and pure Cu:14) (1) the
shape of the curve for pure Fe is similar to that for Cu but
dierent than Al, (2) there is a hardness maximum in Al at
an equivalent strain of 2 but no maximum appears in Fe
and Cu, and (3) the saturation level at the steady state, where
no change in hardness takes place with respect to the
equivalent strain, begins at an equivalent strain of 40 for
Fe but at equivalent strains of 6 for Al and 15 for Cu,
suggesting that the critical strain for the onset of steady state
is higher for harder materials.
For the dierence in the hardness behavior given above,
the dierence in melting temperatures may be considered.
The melting temperatures are 1812, 1356 and 933 K for Fe,
Cu and Al, respectively, and thus, the room temperature
(298 K) corresponds to 0.16, 0.22 and 0.32 when normalized
Fig. 6 Vickers microhardness plotted against height from bottom surface

across thickness of disc and ring samples processed for one revolution at
distances of 0, 2, 4 and 13.5 mm from center.
47
Fig. 8 Nominal stress versus nominal strain curves for annealed sample
and some selected disc and ring samples processed to dierent equivalent
strains.
(a)
Fig. 7 Vickers microhardness plotted along periphery of ring samples at

13.5 mm from center of ring samples processed after various revolutions.
by each of the melting points. Compared with conventional

deformation methods, the density of lattice defects such as
dislocations are very high in the HPT deformed samples. It is
reasonable to consider that dislocation recovery and recrystalization begin at lower temperatures than that after conventional deformation. Recrystallization was observed in highly
deformed Cu after annealing at a homologous temperature of
0.3.27,28) Therefore, it seems that the dierence between Fe
and Al can be attributed to recovery and recrystalization in Al
samples after HPT processing because of its lower melting
temperature.
Figure 6 plots hardness variations across the thickness of
disc and ring samples after one revolution at the distances of
0, 2, 4 and 13.5 mm from the center. The hardness increases
with an increase in the distance from the center in consistence
with the plots in Fig. 4 and Fig. 5. However, it is almost
constant across the thickness. These results demonstrate that
the structure is homogenous across the specimen thickness.
Hardness variations are plotted in Fig. 7 with respect to the
angle along the periphery at the mid width of ring samples (at
r 13:5 mm) after revolutions of 1/8 to 4. The hardness
level increases as N increases. It is also found that the angular
variation is prominent for N < 1 but it becomes less as
N 1. This suggests that homogeneity is developed throughout the ring sample as the number of revolutions increases.
(b)
Fig. 9 Variation of (a) ultimate tensile strength and (b) elongation to

failure with equivalent strain for some selected disc and ring samples.
Some representative stress-strain curves are delineated in

Fig. 8 from tensile testing of the disc and ring samples
conducted at room temperature and at an initial strain rate of
3:3 103 s1 . The tensile strength increases but the total
elongation to failure decreases with an increase in the
equivalent strain. The variation with respect to the equivalent
strain of (a) ultimate tensile strength and (b) elongation to
failure are more clearly documented in Fig. 9 where the data
48
Fig. 10 TEM micrographs and SAED patterns for (a) disc sample after 1/4 revolutions with " 2:2, (b) ring sample after 1/4 revolutions
with " 15:3, (c) disc sample after 2 revolutions with " 18:1 and (d) ring sample after 2 revolutions with " 122:4.
points indicate the average of three repeated measurements

and the error bars represent the standard deviation of the
averages. Figure 9(a) shows that the tensile strength increases with increasing the equivalent strain, levels o at an
equivalent strain of 40 and remains constant at 1050 GPa
with a further increase in the equivalent strain. This trend is
very similar to the hardness variation with respect to the
equivalent strain as plotted in Fig. 5. In fact, all values of the
tensile strength fall well on the single curve determined by
the hardness measurement when the tensile strength () is
converted to the hardness (Hv) through the relation as
3:5 Hv where the factor, 3.5, is consistent with the factor,
3, accepted for the conversion. Figure 9(b) shows the
variation of the elongation to failure with respect to the
equivalent strain. The ductility decreases with an increase in
the equivalent strain and it reaches the constant level of 2%
at an equivalent strain of 40.
The present study has shown that the hardness as well as
the tensile strength at the saturated level shown in Fig. 5 and
Fig. 9(a) is less than the values reported in earlier works of
Fe.1520) This can be attributed to the dierence in carbon
contents in the materials employed in individual experiments.
Although the C content was not described explicitly except
the study of Todaka et al.,16) which is 0.03% C, it must have
been much higher than 11 ppm in the Fe used in this study.
Very recently, Todaka et al.20) reported extraordinarily high
tensile strength in the same high purity Fe after processing by
HPT. The enhancement results in almost twice when
compared with the present tensile strength despite the fact
that the hardness is in almost the same level as the preset
study. Todaka et al. considered that the reason for the
abnormal increase in the tensile strength may be attributed to
anisotropic strengthening due to single directional deforma-
tion. It is considered that such an anisotropy may arise from a

higher applied pressure as 5 GPa employed by Todaka et al.
This pressure is higher by 2.5 times than the present pressure
2 GPa and thus there can be a possibility that a strong
anisotropic texture might have developed in the sample.
Alternatively, the abnormal strengthening may be attributed
to the dierence in the tensile specimen geometry. While the
thickness in this study is 0.3 mm, the one for the study
employed by Todaka et al. is 0.6 mm. Although the gauge
length is the same in both experiments as 1 mm, the
dierence in the thickness might have caused an enhanced
strength. This reason may be probable because thicker
specimens tend to lead to higher strength and higher ductility.
More study is required for this clarication.
TEM observations are summarized in Fig. 10 where
bright-eld images and corresponding SAED patterns were
taken from four samples subjected to dierent magnitudes of
the equivalent strain: (a) " 2:2 after processing a disc
sample for 1/4 revolutions, (b) " 15:3 after processing a
ring sample for 1/4 revolutions, (c) " 18:1 after processing
a disc sample for 2 revolutions, and (d) " 122:4 after
processing a ring sample for 2 revolutions. The rst sample
corresponds to an initial stage of the hardness increase, the
second and third samples to a mid stage of the hardness
increase, and the forth sample to a stage of the steady state. It
is apparent from Fig. 10(a) that at " 2:2 the microstructure
consists of subgrains with the average subgrain size of 1 mm
and there are many dislocations within the subgrains.
Because of a net type of the SAED pattern, the boundary
misorientation should be very low. In Figs. 10(b) and 10(c)
" 15:3 and 18.1, respectively, the subgrains become
smaller to the size of 0:3 mm having many dislocations
within them and the boundary misorientation is increased
because the diractions spots are now diused around rings.

In Fig. 10(d) at " 122:4, the grains even become smaller
to 0:2 mm and some grain boundaries appear to be better
dened. There are also grains with a low density of
dislocations visible at the center of the micrograph as
indicated by an arrow. This suggests that recrystallization
may have taken place during or after HPT processing. The
SAED pattern exhibits a complete form of rings, indicating
that the microstructure consists of very small grains having
high angles of misorientations.
Based on the TEM observations shown in Fig. 10 along
with the variations of hardness and tensile properties given in
Figs. 5, 8 and 9, the grain renement of Fe may be deduced
as follows. In the initial stage of straining, the number
of dislocations increases as anticipated from the normal
deformation mode and thus dislocation coalescence and
subgrain formation occur. Consequently, pronounced strain
hardening takes place due to increasing the stored energy
through accumulation of dislocations. With further straining,
the subgrain size decreases, the population of dislocations
increases within the subgrains, the subgrain boundary width
becomes better dened and the misorientation angle increases. Han et al.19,21) reported similar results in high purity
Fe using the process of equal-channel angular pressing
(ECAP). They found that the hardness increases signicantly
at the initial stages of straining due to an increase in
dislocations and it increases slightly at subsequent stages
where grains are rened signicantly as a result of the
rearrangement of dislocations.
When the dislocation accumulation reaches a critical level,
the recrystallization is induced and the rate of strain
hardening decreases. The process of this grain renement is
very similar to the one illustrated earlier for pure Cu during
operation of HPT.14,25) At the saturation stage (steady state),
hardening occurs due to dislocation accumulation and grain
renement whereas softening occurs because of mutual annihilation of dislocations within grains and disappearance of
dislocation through recrystallization. Because of the balance
between the hardening and the softening, the hardness and
strength remain constant at the steady state. Some studies
report that the steady-state in SPD-processed materials is
reached by annihilation of dislocations by dynamic recovery,29,30) whereas some other studies support that steady-state
can be reached by dynamic recrystallization.18,31,32) The
present experiment conrms that the latter idea is more
appropriate for the mechanism of the steady state.
4.
49
strains higher than 40. This behavior is very similar to

the one observed in pure Cu.
(3) Ultimate tensile strength and elongation to failure can
also be expressed by single functions of the equivalent
strain.
(4) Dislocation density increases and the subgrain formation and its renement occurs at an early stage of
straining, thereby increasing the hardness level. However, at the saturation stage, a balance between the
accumulation of dislocations and recrystallization is
established so that a steady state appears where the
hardness and tensile properties remains unchanged with
straining.
Acknowledgements
This work was supported by a Grant-in-Aid for Scientic
Research from the Ministry of Education, Culture, Sports,
Science and Technology, Japan, in Priority Areas Giant
Straining Process for Advanced Materials Containing UltraHigh Density Lattice Defects. The pure Fe was supplied by
Nippon Steel Corporation for the study of Materials Science
of Ultrahigh Plastic Strain in Iron and Steel Institute of
Japan (ISIJ). One of the authors (KE) would like to thank
Islamic Development Bank for his scholarship.
Appendix
During operation of HPT, thickness reduction and sample
slippage may occur so that a rigorous evaluation is required
for both eects on the equivalent strain. The equivalent strain
is estimated through the following form,7) provided that
neither thickness reduction nor sample slippage occurs.
2rN
" p
3t
A:1
However, with the semi-constrained HPT system as illustrated in Fig. 1, some material ows out and thus the sample
thickness is reduced. Figure A1 plots the thickness change
with respect to N. The thickness decreases more signicantly
within the rst few revolutions and gradually thereafter. The
change may be approximated as a function of N in the
following form.
Conclusions
High pressure torsion (HPT) was conducted on pure Fe

(99.96%) with the carbon content of 11 ppm using disc and
ring samples. The following conclusions were reached:
(1) Ring samples can be used as an alternative for the disc
samples to obtain a uniform structure throughout the
samples.
(2) Data points of Vickers microhardness fall well on a
single curve when plotted against the equivalent strain
for both disc and ring samples. The hardness increases
with increasing the equivalent strain at initial stages of
straining, but saturates to a constant level at equivalent
Fig. A1
Thickness after HPT plotted against number of revolutions.
50
REFERENCES
Fig. A2 Ratio of equivalent strain after correction through eq. (1) to
equivalent strain through eq. (A1) without correction.
tN 0:5 0:3 exp0:7N 0:4
A:2
The value of the rst term, 0.5, arises because the thickness is
never decreased below 0.5 mm because the hole depth of
each upper and lower anvils is 0.25 mm in this study. Thus,
the second term represents the reduction in thickness with
sample revolution.
The slippage between the sample and anvil may also occur
during HPT. The measurement of the slippage was undertaken by using samples that two straight lines were drawn
along the diameters on both sides of the disc. After HPT
operation, the discrepancy between the lines from the initial
position was taken as the slippage. In this study, it turned out
that the fraction was 0.17 and this fraction was kept almost
constant with N.33)
Now, considering the thickness reduction and sample
slippage, eq. (A1) may be modied to the form given in
eq. (1). Thus, both eects were corrected in this study using
eq. (1). Figure A2 shows the ratio of the equivalent strain
after correction through eq. (1) ("corrected ) to the equivalent
strain without correction through eq. (A1) ("uncorrected ).
Because the thickness reduction and the sample slippage
act in the opposite sense, both eects seem canceled out. This
is demonstrated clearly in Fig. A2 and it turns out that the
equivalent strain calculated with the initial thickness without
slippage correction is reasonably close to the one corrected
through eq. (1).

45 (2000) 103189.
Y. T. Zhu: JOM 58 (2006) 3339.
3) Z. Horita, D. J. Smith, M. Furukawa, M. Nemoto, R. Z. Valiev and
T. G. Langdon: J. Mater. Res. 11 (1996) 18801890.
4) C. Rentenberger, T. Waitz and H. P. Karnthaler: Mater. Sci. Eng. A 462
(2007) 283288.
5) I. V. Alexandrov, K. Zhang, A. R. Kilmametov, K. Lu and R. Z. Valiev:
Mater. Sci. Eng. A 234236 (1997) 331334.
6) A. P. Zhilyaev, S. Lee, G. V. Nurislamova, R. Z. Valiev and T. G.
Langdon: Scr. Mater. 44 (2001) 27532758.
7) F. Wetscher, A. Vorhauer, R. Stock and R. Pippan: Mater. Sci. Eng. A
387389 (2004) 809816.
8) A. Vorhauer and R. Pippan: Scr. Mater. 51 (2004) 921925.
9) G. Sakai, Z. Horita and T. G. Langdon: Mater. Sci. Eng. A 393 (2005)
344351.
11) Y. Harai, Y. Ito and Z. Horita: Scr. Mater. 58 (2008) 469472.
12) S. Erbel: Met. Techno. 6 (1979) 482486.
13) I. Saunders and J. Nutting: Metal. Sci. 18 (1984) 571575.
14) K. Edalati, T. Fujioka and Z. Horita: Mat. Sci. Eng. A 497 (2008)
168173.
15) R. Z. Valiev, YU. V. Ivanisenko, E. F. Rauch and B. Baudelet: Acta.
Mater. 44 (1996) 47054712.
16) Y. Todaka, M. Umemoto, J. Yin, Z. Liu and K. Tsuchiya: Mater. Sci.
Eng. A 462 (2007) 264268.
17) F. Wetscher, A. Vorhauer and R. Pippan: Mater. Sci. Eng. A 410411
(2005) 213216.
18) F. Wetscher and R. Pippan: Philos. Mag. 86 (2006) 58675883.
19) B. Q. Han, E. J. Lavernia and F. A. Mohamed: Metall. Mater. Trans. A
34A (2003) 7183.
20) Y. Todaka, Y. Miki, M. Umemoto, C. Wang and K. Tsuchiya: Mater.
Sci. Forum 584586 (2008) 597602.
21) B. Q. Han, E. J. Lavernia and F. A. Mohamed: Metall. Mater. Trans. A
35A (2003) 13431350.
22) M. Sus-Ryszkowska, T. Wejrzanowski, Z. Pakiea and K. J.
Kurzydowski: Mater. Sci. Eng. A 369 (2004) 151156.
23) S. V. Dobatkin, V. I. Kopylov, R. Pippan and O. V. Vasileva: Mater.
Sci. Forum 467470 (2004) 12771282.
893899.
25) Y. Ito, Y. Harai, T. Fujioka, K. Edalati and Z. Horita: Mat. Sci. Forum
584586 (2008) 191196.
26) Y. Ito and Z. Horita: Mater. Sci. Eng. A (2008) in press.
27) E. Schaer and M. B. Kerber: Mater. Sci. Eng. A 462 (2007) 139143.
28) H. Jiang, Y. T. Zhu, D. P. Butt, I. V. Alexandrov and T. C. Lowe:
Mater. Sci. Eng. A 290 (2000) 128138.
29) H. P. Stuwe: Acta Metall. 13 (1965) 13371342.
30) M. Zehetbauer: Acta Metall. Mater. 41 (1993) 589599.
31) T. Hebesberger, H. P. Stuwe, A. Vorhauer, F. Wetscher and R. Pippan:
Acta Mater. 53 (2005) 393402.
32) N. Hansen: Metall. Mater. Trans. A 32A (2001) 29172935.
33) K. Edalati, Z. Horita and T. G. Langdon: Scr. Mater. 60 (2009) 912.

High Tensile Strength of Low-Carbon Ferritic Steel Subjected to Severe Drawing

Tetsuya Suzuki1 , Yo Tomota2 , Atsushi Moriai3 and Hitoshi Tashiro4; *
1
The Research Center for Superplasticity, Faculty of Engineering, Ibaraki University, Hitachi 316-8511, Japan
Graduate School of Science and Engineering, Ibaraki University, Hitachi 316-8511, Japan
3
Research Group for Neutron Residual Stress Analysis, Quantum Beam Science Directorate,
Japan Atomic Energy Agency, Tokai, Ibaraki 319-1196, Japan
4
Department of Communication and Information Science, Fukushima National College of Technology, Iwaki 970-8034, Japan
2
Severely drawn low-carbon ferritic steel wire was investigated to make clear the reason of its high tensile strength which is higher than that
predicted from the hardness. The microstructure of a specimen subjected to a true strain of 6.6 consists of ne elongated grains and subgrains
with high dislocation density. Neutron diraction showed the existence of (hkl) intergranular residual stress, suggesting a composite
strengthening mechanism. The reasons of higher tensile strength achieved by drawing in comparison with other severe deformation processes,
like ECAP, ARB and HPT, are estimated to be texture, residual intergranular stress and high dislocation density.
Keywords: low-carbon ferritic steel, IF steel, drawing, severe deformation, neutron diraction, residual stress, texture
1.
Introduction
Severe Plastic Deformation (SPD) processes to produce

ultrane-grained materials have intensively been studied.
The mechanisms of microstructural evolution and mechanical behavior have also been studied by many researchers.
The highest tensile strength of commercially available
materials has been recorded by drawn pearlite wire,1) often
called piano wire, which can have a tensile strength higher
than 5 GPa. Even in the case of low-carbon ferritic steel,
Tashiro has reported 3 GPa for a drawing strain (hereafter,
drawing strain means true strain given by the drawing) of
11.5.2) Since the mechanisms of high tensile strength in
drawn pearlite wire are still unclear, the characteristics of
drawn low-carbon ferritic steel need to be claried in order to
avoid the inuence of cementite.
In order to evaluate the drawing process, the tensile
strengths of low-carbon ferritic steel after newly developed
SPD processes such as equal-channel angular pressing
(ECAP), accumulative roll bonding (ARB) and high-pressure
torsion (HPT) are collected from existing references39) and
summarized in Fig. 1. The tensile strength of drawing
increases linearly with the drawing strain, however, that of
ARB and ECAP becomes nearly constant at approximately
0.8 GPa from the equivalent strain of 2 (in case of drawing,
the drawing strain is equal to the equivalent strain). The
tensile strength of a HPT specimen is 1.8 GPa at the
equivalent strain of 91. It is lower than that of the drawn
wire which has tensile strength of 3 GPa at the drawing strain
of 11.5.
The reason why such a high tensile strength is achieved by
drawing in comparison with the other severe deformation
processes is discussed in this paper.
2.
Experimental Procedures
2.1 Specimens
The chemical composition of a low-carbon ferritic steel
*Engineering
Consultant, Fukushima National College of Technology
Fig. 1 Tensile strength of low-carbon ferritic steels after ARB, ECAP,

HPT and drawing.
Table 1
Chemical composition of a low-carbon ferritic steel (mass ppm).
Si
Mn
Ni
Cr
80
10
1300
120
40
100
200
used in this study is shown in Table 1. A hot-rolled bar

5.4 mm in diameter was drawn to 1.9, 0.74, 0.38 and 0.2 mm
for the drawing strains of 2.0, 4.0, 5.3 and 6.6, respectively.
Hereafter, these specimens are called "2.0, "4.0, "5.3 and
"6.6, respectively. Additionally, "0 was used as a reference
specimen.
2.2 Microstructural observations
The microstructures on the cross section of the wires were
observed by scanning electron microscopy (SEM), and the
crystal orientation distributions were investigated with
Electron Back-Scatter Patterns (EBSP). The specimens were
prepared by mechanical polishing followed by chemical
polishing using colloidal silica.
2.3 Mechanical tests
A tensile test was performed using a gear-driven tester
with air chucking at room temperature (RT). A Vickers
52
T. Suzuki, Y. Tomota, A. Moriai and H. Tashiro
and (211) with rotation of the bundled sample in 5-degree

steps from 0 to 90 in texture measurements and 0, 22.5, 45,
67.5 and 90 degrees in lattice strain measurements with
respect to the drawing direction. The diraction prole
obtained was tted by the Gaussian function to determine the
peak position (i.e., lattice spacing). The lattice strain was
then calculated by employing the spacing for specimen "0 as
the reference. That is,
"hkl fdhkl dhlk0 g=dhkl0
Here, dhkl refers to the (hkl) lattice spacing, and subscript

zero indicates the reference spacing.
3.
Fig. 2
Schematic illustration of neutron diraction measurement.
hardness test was also conducted on the cross section of the

wires with an applied load of 200 g. Hardness was measured
on the transverse section plane across the drawing direction
and the longitudinal cross-section plane along the drawing
direction. Since there exists a hardness distribution from the
surface to the center in a drawn wire, the average value was
used for comparisons with the tensile strength. The hardness
at the center was approximately HV 0.4 GPa higher than that
near the surface.
2.4
Texture and lattice strain measurement by neutron

diraction
Neutron diraction measurements were conducted using
the neutron diractometer for residual stress analysis
(RESA) at the Japan Atomic Energy Agency (JAEA). The
drawn wires were bundled to approximately 5 mm in
diameter and 5 mm in length. This bundle of wires was
totally irradiated by the neutron incident beam, as illustrated
in Fig. 2. Diraction proles were obtained for (110), (200)
Fig. 3
3.1 Microstructural change by drawing

Figure 3 shows cross-sectional SEM images of the drawn
wires. The mean grain size of specimen "0 with equiaxial
grains in (a) is 17.7 mm. As the drawing strain is increased,
the grains are elongated in proportion, as observed for "2.0
(b) and "6.6 (c). The "6.6 was studied with transmission
electron microscopy (TEM) by Ohsaki and Hono10) and
bright (a) and dark (b) eld TEM images are presented in
Fig. 4. Here, many dislocations exist inside a grain, and the
grain width is approximately 64 nm after drawing. As may be
speculated from the diraction pattern and dark eld image,
some of the grain boundaries are low angle; such boundaries
must develop from dislocation cell boundaries.
The result of the EBSP analysis for "6.6 is shown in Fig. 5.
Because of the high density of dislocations and internal
stresses, the image quality is poor. Low angle boundaries
observed by TEM can not be identied clearly in Fig. 5.
From the orientation map, the formation of the h110i ber
texture is found. The grain shape is very elongated along the
drawing direction, and the width across the drawing direction
Cross-sectional SEM images of (a) "0, (b) "2.0 and (c) "6.6.
Fig. 4
TEM images of "6.6: (a) bright eld image and diraction pattern and (b) dark eld image.10)
Fig. 5 Result of the EBSP analysis for "6.6: (a) orientation map for the transverse direction and (b) orientation map for the drawing
direction.
is approximately 0.48 mm (0.280.69 mm). By using a simple

geometrical calculation with the assumption of constant
volume, a spherical grain with a 17.7 mm width becomes an
ellipsoid with the short axis of 0.66 mm, which is comparable
to the observation in Fig. 5. Consequently, the microstructure
of the drawn low-carbon ferritic steel consists of elongated
high angle grains with width of 0.48 mm, in which the
elongated subgrains with width of 64 nm.
Figure 6 shows the 110 orientation distributions of "6.6
and "4.0 measured by neutron diraction. Both "6.6 and "4.0
show strong h110i texture along the drawing direction and
this is consistent with the EBSP results.
3.2 Stress-strain curves and hardness
The stress-strain curves obtained for the drawn wires are
shown in Fig. 7. The tensile strength increases with the
Fig. 6 Texture of "4.0 and "6.6 measured by neutron diraction.
53
54
T. Suzuki, Y. Tomota, A. Moriai and H. Tashiro
Fig. 7 Nominal stress-strain curves.
of 3. However, the tensile strength estimated from the

hardness becomes lower with the increase of the drawing
strain. Such a discrepancy has also been found in drawn
pearlite steels.12) The drawing (i.e., one-directional plastic
deformation) is assumed to yield the anisotropic strength of a
material. Todaka et al. reported a similar discrepancy in lowcarbon ferritic steel after HPT.3) Heavily deformed lowcarbon ferritic steel with the equivalent strains of 45 and 91
shows higher tensile strength than the strength estimated
from the hardness. They suggested that this result is due to
both the single-directional deformation and the segregation
of impurity elements during HPT straining.
Here, two possible reasons for the high tensile strength
found in Fig. 8 are discussed: texture eect and intergranular
stress (Section 3.3). First, the strong h110i ber texture
should be noted. The Schmidt factor of h110i oriented grains
for the {110} h111i slip system is 0.408. A few grains with a
Schmidt factor larger than 0.41 may lead to higher tensile
yield strength of a polycrystalline material. In the case of the
hardness test, the stress distribution around the indenter is
complex and hence such texture hardening does not appear
apparently.
3.3
Fig. 8
Vickers hardness as a function of true strain given by drawing.
drawing strain, whereas the elongation decreases. In particular, uniform elongation disappears by drawing but large
local elongation remains, resulting in a dimple fracture. This
is a common feature of metals and alloys subjected to severe
deformation. Even in the tensile test performed for the "6.6
wire at 77 K in liquid nitrogen, a cup and cone with a dimple
fracture were observed.11)
Figure 8 shows the results of the Vickers hardness test.
Hardness measured on the transverse and longitudinal cross
sections increases with the drawing strain; this result is
similar to that of the tensile strength. The hardness measured
on the transverse section plane is approximately 0.2 GPa
higher than that on the longitudinal cross-section plane.
In most cases, it is known that Vickers hardness is three
times that of the 8% ow stress, at which point the yield
strength is often employed. In the present case, work
hardening is lost in the drawn specimens, so the tensile
strength is expected to be related to the hardness by a factor
Intergranular residual stress estimated from neutron diraction

It should be noted that residual lattice strain is dened as
the lattice strain measured after deformation in the unloaded
situation. Figure 9 shows the result of the residual lattice
strains of (110) and (200). A bundle of wire specimens was
rotated in steps of 22.5 degrees with respect to the drawing
direction (see Fig. 2). In the case of "0 (Fig. 9(a)), the
residual lattice strain is nearly zero and hardly changes with
the rotating angle. In contrast, specimen "6.6 shows strong
anisotropic behavior with respect to the rotating angle and
(hkl) plane measured. As can be seen, the residual lattice
strain from (110) shows a convex shape whereas (200) shows
a concave shape. Suzuki et al. reported on the residual lattice
strain of the ferrite phase in a drawn pearlite steel with a
drawing strain of 1.4.12) The features found for the convex
shape of (110) and the concave shape of (200) are consistent
with the ferrite phase of pearlite steel. Then it is estimated the
h110i oriented family grains with respect to the drawing
direction play softer grains and h200i harder ones. Hence, the
plastic mist strains produce internal stresses that remain as
intergranular residual stresses. Softer grains must show
compressive residual stresses and harder ones tensile along
the drawing direction. This is similar to the composite
strengthening like pre-stressed concrete which shows
anisotropic strength. In the present case, a few harder (hkl)
grains are speculated to play a role of ber in a composite.
3.4 Comparison of drawing and other SPD processes
Figure 10 shows the comparison of Vickers hardness in
low-carbon ferritic steel after drawing and after ARB, ECAP
and HPT. The hardness of low-carbon ferritic steels after
drawing increases with the increase of the drawing strain,
however, the hardness after ECAP and HPT becomes nearly
constant at 2 GPa for ECAP and 4 GPa for HPT, whereas the
hardness after ARB is 2.5 GPa at the equivalent strain of 4.0.
It was found that the drawing of low-carbon ferritic steel
55
Fig. 10 Vickers hardness of low-carbon ferritic steels after ARB, ECAP,

HPT and drawing.
4.
Conclusions
The following conclusions are obtained.

(1) The microstructure of low-carbon ferritic steel wire
drawn to a true strain of 6.6 shows elongated grains,
including elongated subgrains. The width of the subgrains
is approximately 64 nm and many dislocations are observed.
(2) The tensile strength and hardness of the drawn ferrite
steel wire is higher than those observed for specimens
subjected to the same equivalent strain by other severe
deformation processes.
(3) The observed tensile strength is higher than that estimated from the hardness. The higher tensile strength must be
due to the strong h110i ber texture, intergranular residual
stress and high dislocation density.
Acknowledgments
Fig. 9 Residual lattice strains measured by neutron diraction: (a) before
drawing and (b) true drawing strain of 6.6.
produces higher hardness compared with that of ECAP and

ARB processes.
The grain size (width) of low-carbon ferritic steel after
ARB at the equivalent strain of 4.0 is 220 nm.6) Considering
that the grain size is not so dierent from the case of drawing,
the extremely high hardness and tensile strength of the
present wire should be noted.
One reason for the above result may be high dislocation
density as observed in Fig. 4. Because the drawing was
performed in water in order to avoid heating, it can be
speculated that the temperature during deformation causes
the dierence between the drawing and other processes. For
example, the process of ARB is performed by the repetition
of rolling and heating at 773 K. Restoration during heating
or deformation heating must decrease hardness and tensile
strength in the other SPD processes.
Consequently, in comparison with the other SPD processes, it can be conrmed that drawing is an eective process
to obtain high hardness and high strength. Neutron diraction
has revealed the existence of residual intergranular stress and
strong h110i texture. The hardness test seems to show the
existence another factor for the high tensile strength of the
present drawn wire.
We wish to thank Ms. M. Uno for her experimental

assistance. We are also grateful to Dr. S. Ohsaki and Dr. K.
Hono for the TEM observation and Dr. Y. Adachi for the
EBSP analysis. This study was nancially supported by a
Grant-in-Aid for Scientic Research on Priority Areas
(#18062002).
REFERENCES
1) Y. Tomota, T. Suzuki, A. Kanie, Y. Shiota, M. Uno, A. Moriai, N.
Minakawa and Y. Morii: Acta. Mater. 53 (2005) 463467.
2) H. Tashiro: Ph.D thesis (1992) Tohoku University.
3) Y. Todaka, Y. Miki, M. Umemoto, C. Wang and K. Tsuchiya: Mater.
Sci. Forum 584586 (2008) 597602.
4) N. Tsuji, R. Ueji and Y. Saito: Materia Japan 39 (2000) 961967.
893899.
6) N. Kamikawa, N. Tsuji and Y. Saito: Tetsu-to-Hagane 89 (2003)
6370.
7) J. T. Wang, C. Xu, Z. Z. Du, G. Z. Qu and T. G. Langdon: Mater. Sci.
Eng. A410411 (2005) 312315.
8) K.-T. Park and D. H. Shin: Mater. Sci. Eng. A334 (2002) 7986.
9) Y. Fukuda, K. Oh-ishi, Z. Horita and T. G. Langdon: Acta Mater. 50
(2002) 13591368.
10) S. Ohsaki and K. Hono: unpublished work.
11) S. Sugaya, T. Shiratori, S. Morooka, Y. Shiota and Y. Tomota: CAMP
ISIJ 21 (2008) PS 20.
12) T. Suzuki, Y. Tomota, M. Isaka, A. Moriai, N. Minakawa and Y.
Morii: ISIJ Int. 44 (2004) 14261430.

Brittle-Ductile Transition in Low Carbon Steel Deformed

by the Accumulative Roll Bonding Process
Masaki Tanaka1 , Kenji Higashida1 , Tomotsugu Shimokawa2 and Tatsuya Morikawa1
1
Department of Materials Science and Engineering, Kyushu University, Fukuoka 819-0395, Japan
Division of Mechanical Science and Engineering, Graduate School of Natural Science and Technology,
2
Brittle-ductile transition (BDT) behaviour was investigated in low carbon steel deformed by an accumulative roll-bonding (ARB) process.
The temperature dependence of its fracture toughness was measured by conducting four-point bending tests at various temperatures and strain
rates. The fracture toughness increased while the BDT temperature decreased in the specimens deformed by the ARB process. Arrhenius plots
between the BDT temperatures and the strain rates indicated that the activation energy for the BDT did not change due to the deformation with
the ARB process. It was deduced that the decrease in the BDT temperature by grain rening was not due to the increase in the dislocation
mobility controlled by short-range obstacles. Molecular dynamics simulations revealed that moving dislocations were impinged against grain
boundaries, creating a shielding eld, and were reemitted from there with increasing strain. Grain rening led to an increase in the fracture
toughness at low temperatures and a decrease in the BDT temperature. In the present paper, the roles of grain boundaries have been discussed in
order to explain the enhancement in the fracture toughness of ne-grained materials at low temperatures, and the decrease in the BDT
temperature. [doi:10.2320/matertrans.MD200817]
Keywords: grain boundary shielding, crack, severe plastic deformation, accumulative roll-bonding (ARB)
1.
Introduction
Solid materials are plastically deformed at high temperatures. In contrast, some materials fail without a substantial
plastic deformation at low temperatures; this is referred to as
brittle fracture. It is a prominent event that occasionally
occurs in bcc metals as well as semi-conductors and
ceramics,16) which is well-known as brittle-ductile transition
(BDT). Since a BDT event is critical, especially in structural
materials, some counter measures to decrease the BDT
temperature have been suggested. One of the methods is
grain rening.710) Tsuji et al.11) deformed interstitial free
(IF) steel by an accumulative roll-bonding (ARB) process
and measured the absorbed energy at various temperatures by
conducting Charpy impact tests. They demonstrated that the
IF steel gained both strength as well as ductility at low
temperatures. It implies that the BDT temperature itself is
decreased by the ARB process.
In single crystal materials wherein the Peierls stress is
high, fracture tests provide the relation between the strain
rate, "_, and the BDT temperature, TBDT , as follows:

Ea
"_ "o exp
;
1
kTBDT
where Ea is the activation energy, k is the Boltzmann
constant, and "o is a pre-factor. Many studies1215) have
conrmed that the values of Ea in eq. (1) were close to those
for dislocation glide. It was deduced that the BDT behaviours
in those materials are controlled by a thermal activation
process related to the dislocation glide. Michot et al.16)
pointed that the pre-factor in eq. (1) depends on the number
of dislocation sources along the crack front. It indicates that
the BDT temperature depends on dislocation mobility as a
thermal activated term as well as dislocation source density
around a crack tip as an athermal term in eq. (1).
Since dislocations have long-range stress elds, the stress

eld around a crack tip is required to be modied when
dislocations are introduced from the crack tip.17,18) The
concept of stress accommodation is termed as dislocation
shielding, which has been independently suggested by
Thomson and Weertman.19,20) It was experimentally conrmed by transmission election microscopy and photoelasticity in silicon single crystals.21,22)
In polycrystalline materials, the BDT temperature strongly
depends on the grain size of specimens. The BDT temperature decreases with the grain size, which could be explained
by employing a dislocation pile-up model.23) However, the
application of the shielding theory to a pile-up model deduces
that the BDT temperature in ne-grained materials should
increase because the dislocations cannot continue to emit
from the crack tip due to the back stress of the pile-up
dislocations around the grain boundaries.
In the present study, the BDT behaviours in ne-grained
materials have been discussed on the basis of the shielding
theory. Low carbon steel was plastically deformed by the
ARB process to obtain ne-grained specimens. The temperature dependence of its apparent fracture toughness was
measured at various strain rates. The activation energy
obtained from the BDT temperature in roll-bonded specimens was measured and was compared with that for fully
annealed specimens with a large grain size. Molecular
dynamics calculations were performed in order to demonstrate the roles of grain boundaries in a BDT event. The roles
of grain boundaries have been discussed in order to explain
the decrease in the BDT temperatures.
2.
Experimental
Low carbon steel (0.02 mass% C) was employed, the

chemical composition of which is presented in Table 1. Since
Brittle-Ductile Transition in Low Carbon Steel Deformed by the Accumulative Roll Bonding Process
Table 1
Chemical composition of the material used.
Si
Mn
0.02
<0:01
0.35
0.1
0.03
tends to be higher than that with a crack. For samples that did
not break during the bending tests, the yield stress was used
instead of frac in eq. (2), giving a lower-bound value for the
apparent fracture toughness.
Fe
bal.
mass%
3.
10mm
c
L=1 mm
1 mm
4 mm
TD
RD
ND
Fig. 1 Schematic diagram of a four-point bending specimen. The distances

between inner and outer beams are 4 and 10 mm, respectively. Rolling
direction (RD), transverse direction (TD) and normal direction (ND) are
indicated in the diagram.
low carbon steel demonstrates rather high ductility, a small

amount of phosphorus was added to reduce its ductility and to
demonstrate the remarkable eect of grain rening on its
fracture toughness at low temperatures. Cold-rolled 2-mm
thick sheets were annealed at 1023 K for 3.6 ks in order to
complete recrystallization. The mean grain size of the fully
annealed sheets was found to be 0.8 mm. The sheets were
cold-rolled by 50% reduction in thickness for a subsequent
ARB process. After degreasing and wire-brushing one
surface of each sheet, two pieces of a 1 mm thick sheet were
stacked, resulting in a thickness of 2 mm, and were spotwelded at their front and back ends. The stacked set of sheets
was held at 773 K for 600 s in a furnace and then roll-bonded
by 50% reduction in thickness. The ARB process was
repeated three times. The nal equivalent strain was found to
be 3.04. Miniature four-point bending samples were prepared
from the roll-bonded and fully annealed sheets. Figure 1
illustrates a schematic diagram of the four-point bending test
specimen, revealing the geometry of the sample. The inner
and outer beam distances were 4 mm and 10 mm, respectively. A notch with a depth of 300 mm was introduced
perpendicular to the transverse direction (TD) at the middle
of the tensile surface. The curvature radius at the notch tip
was about 0.1 mm. Four-point bending tests were performed
at strain rates of 4:46 104 , 2:23 103 and 1:12
102 s1 . The strain rate was represented by those at the
tensile surface where the tensile stress was the maximum in
the specimen beam. The temperature was controlled by using
liquid nitrogen at 77 K and cold nitrogen gas from 110 to
170 K. The temperature was controlled to be within 0:5 K.
The apparent fracture toughness, K , during the four-point
bending test was calculated using the following equation:24)
p
K f frac 2c; with c=L;
f 1:112 1:40 7:332 13:083 14:04 ;
57
where frac is the fracture stress, c is the notch depth and L is

the specimen thickness. A notch was introduced in the
specimen instead of a crack in the present study. It was
assumed that the apparent fracture toughness with a notch
Results
Figure 2 presents a TEM bright eld image observed along

the TD of the as roll-bonded specimen, revealing that some
grains included a high density of dislocations. The grains
were elongated along a direction nearly parallel to the rolling
direction (RD). It was found that the grain size perpendicular
to the RD was approximately 200 nm. The selected area
diraction pattern was ring-like, indicating that the boundaries between the adjacent grains had large misorientations.
Figures 3(a) and 3(b) illustrate the orientation image
mappings (OIM) for the as roll-bonded specimens, indicating the normal direction (ND) and RD; these were observed
along the TD at the centre of the surface by using electron
backscatter diraction (EBSD) with an accelerating voltage
of 20 kV. The EBSD patterns were obtained every 0.05 mm
(Hitachi H-4300SE). The grains were elongated along the
RD, which has also been observed in sheets deformed by
conventional rolling processes. Figure 3 roughly indicates
three layers with a width of approximately 10 mm along the
RD. The middle layer in the gure consists of ne grains
with a width of approximately 500 nm. The top and bottom
layers in the gure consist of large band-like structures,
inclining towards the RD. The orientation distribution of the
middle layer is relatively random as compared to that of the
top and bottom layers. Figure 3(a) indicates that the ND
directions of the majority of the grains were almost along
h111i in the top layer. Figure 3(b) indicates that the RD
directions of the grains were nearly parallel to h011i in the
top and bottom layers. These gures indicate that some
typical preferential orientations were developed due to the
ARB process, which are usually induced by a conventional
rolling process in bcc crystals.25) Figures 3(c) and 3(d)
illustrate {111} and {011} pole gures obtained from the
same regions in (a) and (b). They reveal the development
of textures and are consistent with the results of the OIM
images. The orientation scattering indicated that the trend of
the texture was not very strong, and hence, the eect of
those textures on the fracture event was negligible in the
present study.
Figures 4(a) and 4(b) indicate the fractured surfaces of
the fully annealed and the roll-bonded specimens at 77 K,
respectively. Figure 4(a) exhibits river patterns, indicating a
typical fracture surface in a brittle manner. Figure 4(b) shows
layers with a thickness of approximately 100200 mm parallel
to the RD, corresponding to the depth of the fully annealed
sheet before the ARB process. Micro-voids were observed at
some places on the fractured surface, indicating that a local
plastic deformation occurred during crack advance even at
77 K; the fully annealed samples demonstrated complete
brittle fracture at this temperature. In order to evaluate the
temperature dependence of the apparent fracture toughness,
the four-point bending tests were performed.
Figures 5(a), 5(b) and 5(c) illustrate the apparent fracture
toughness obtained for the roll-bonded samples as a function
58
M. Tanaka, K. Higashida, T. Shimokawa and T. Morikawa
RD
800 nm
Fig. 2 Bright eld TEM image from the as roll-bonded specimen. Grains are elongated along the direction parallel to RD. The grain size
parallel to the ND is found to be 200 nm. Selected area diraction pattern is shown at the right-top corner of the gure.
(c)
(a) ND
ND
RD
111
(b) RD
(d)
ND
RD
011
4m
111
RD
001
011
Fig. 3 EBSD orientation image mappings form the as roll-bonded specimen observed from TD; (a) normal direction, (b) rolling direction.
The accelerating voltage was set to be 20 kV and data were taken every 0.05 mm. {111} and {011} pole gures are shown in (c) and (d),
obtained from the same areas in (a) and (b).
of the temperature at the strain rates of 4:46 104 , 2:23

103 and 1:12 102 s1 , respectively. The solid symbols
indicate that the crack tip plastic zone of the fractured
specimen corresponded to small-scale yielding. The open

symbols indicate that the crack tip plastic zone of the
ruptured specimens corresponded to large-scale yielding. The
59
(b)
(a)
RD
50m
50m
ND
Fig. 4
TD
SEM images of fracture surfaces at 77 K; (a) fully annealed specimen, (b) roll-bonded specimen.
BDT temperature was dened as one at which the highest

fracture toughness was obtained although it is necessary to
conrm the correlation between the BDT temperature dened
in this study and that in Charpy impact tests. When fracture
toughness did not denitively show the highest value, the
BDT temperature was dened as the mean value of the
highest temperature in fracture and the lowest temperature in
ow. Figure 5 reveals that the BDT temperature increases
from 114 to 135 K with the strain rate, as observed in other
conventional bcc metals that were not roll-bonded.26) In order
to evaluate the eect of grain rening, the temperature
dependence of the apparent fracture toughness in the fully
annealed samples was also measured.
Figures 6(a), 6(b) and 6(c) illustrate the apparent fracture
toughness obtained for the fully annealed samples as a
function of temperature at the strain rates of 4:46 104 ,
2:23 103 and 1:12 102 s1 , respectively. The open
symbols indicate that the specimens reached 11% strain (the
test rigs limit) during the tests without fracture; this was
termed as ductile bending. In the case of these specimens,
the values of apparent fracture toughness shown in the gure
were calculated using the stress when the load is deviated
from the load-displacement curve, and thus the values of
those apparent fracture toughness were the lower limits. The
average value of fracture toughness at 77 K increased from
47 to 85 MPam1=2 by the ARB process. The increase in
the fracture toughness at 77 K implied that dislocation
activity occurred in the roll-bonded specimens, which
corresponded to the result that some dimples were observed
on the fractured surface of the roll-bonded specimens shown
in Fig. 4. However, no visible dislocation activity was
observed in the fully annealed samples. The BDT temperatures in the roll-bonded specimens were lower than those
in the fully annealed ones. The activation energy in eq. (1)
was obtained from the results of the roll-bonded and fully
annealed specimens in order to examine whether the
dislocation mobility controlled by short-range obstacles had
decreased.
Figure 7 illustrates the Arrhenius plots between the strain
rates and the BDT temperatures. The solid squares indicate
the data obtained for the fully annealed specimens, and the
solid triangles indicate those obtained for the roll-bonded

specimens. Regression lines were tted to each set of plots,
giving the activation energy, Ea . The Ea values of the rollbonded and fully annealed specimens were found to be 0.20
and 0:25 0:09 eV, respectively, wherein the dierence was
negligible, taking the experimental accuracy into account.
The error in the activation energy was due to the ambiguity in
the BDT temperature observed in the case of the fully
annealed samples in Fig. 6(a). The activation energy was not
changed by the ARB process, implying that the decrease in
the BDT temperature was not due to the decrease in the
dislocation mobility controlled by short-range obstacles.
Therefore, an alternative mechanism to decrease the BDT
temperature is required to be considered, which has been
discussed in the next section.
4.
Discussion
In this section, we will discuss the mechanism behind the

improvement in the fracture toughness at low temperatures,
i.e., the decrease in the BDT temperature with the grain size
by the ARB process. Hodge et al.27) revealed a liner relation
between ASTM grain-size numbers and reciprocal BDT
temperatures in Fe-Ni alloys, indicating that the BDT
temperature generally decreases with the grain size. We
focus on the eect of grain boundaries on the decrease in
the BDT temperature since the ARB process increases the
volume fraction of grain boundaries in the specimen. The
eect of dislocations introduced inside grains due to the ARB
process is neglected in this paper, which will be discussed in
a separated paper. We will not discuss the apparent fracture
toughness at high temperatures since the experiment conducted in the present study was invalid in the case of fractures
with large-scale yielding at the crack tip.
4.1
Dislocation pile-up against grain boundaries and

fracture toughness
In the case of a grain boundary playing the role of only an
obstacle against dislocation glide,23) the dislocations pile up
against the grain boundary. This implies that grain rening
should rather decrease the fracture toughness at low temper-
Apparent Fracture toughness, K / MPa m1/2
200
(a) = 2.23 10 -4 s-1
160
Apparent faracture toughness, K / MPa m1/2
60
TBDT =114K
Fracture
Flow
120
80
40
0
0
50
100
150
200
(a) =4.46 10-4 s-1
120
80
40
200
50
Temperture, T / K
Apparent faracture toughness, K / MPa m1/2
160
TBDT=124K
Fracture
Flow
120
80
40
0
0
50
100
150
200
Fracture
Ductile bending
160
40
0
0
50
Apparent fracture toughness, K / MPa m1/2
TBDT=135K
80
40
0
50
100
100
150
200
Temperature, T / K
120
TBDT=135K
80
200
Fracture
Flow
200
120
200
(c) = 1.12 10 -2 s-1
160
150
(b) = 2.23 10-3s-1
Temperture, T / K
200
100
Temperature, T/K
200
(b) = 2.23 10 -3 s-1
TBDT =122K
Fracture
Ductile bending
160
150
200
Temperture, T / K
(c) = 1.12 10-2 s-1

Fracture
Ductile bending
160
TBDT=145K
120
80
40
0
0
50
100
150
200
Temperature, T / K
Fig. 5 Temperature dependence of apparent fracture toughness from rollbonded specimens. The solid symbols indicate that the crack tip plastic
zone of the fractured specimen corresponded to small-scale yielding. The
open symbols indicate that the crack tip plastic zone of the ruptured
specimens corresponded to large-scale yielding.
Fig. 6 Temperature dependence of apparent fracture toughness from fully

annealed specimens. Open symbols indicate that the specimens reached
11% strain (the test rigs limit) during the tests without fracture; these are
classied as ductile bending. The arrows indicate that the values of the
actual fracture toughness are higher than those plotted.
atures, i.e., increase the BDT temperature, which contradicts

the fact that grain rening improves the fracture toughness at
low temperatures, as observed in the Charpy impact tests11)
and the present study.
Suppose (i 1) dislocations had been emitted from the
crack tip. Then, the critical stress intensity factor for the next
dislocation to be emitted is given by the following equation in
a two-dimensional approximation:28)

Xi1
b
1
;
3
kIe,i
g b
h
j1
rc
kr j rc k
crack core size and the distance between the crack tip and jth
dislocation, respectively. The rst term on the right-hand side
contributes to that from the image force, and the second term
does to that from dislocation-dislocation interaction. The
dislocation-dislocation interaction term increases with the
number of emitted dislocations; hence it becomes more
dicult for the subsequent dislocations to be continuously
emitted from the crack tip. In the presence of a grain
boundary near the crack tip, some dislocations pile up against
the grain boundary. Since the back stress of the pile-up
dislocations inhibits the emitted dislocations from moving
against the grain boundaries, it becomes dicult for the
subsequent dislocations to continue emitting from the crack
tip. This tendency becomes more signicant when the grain
size becomes smaller, i.e., materials tend to be more brittle as
where the subscript i indicates the parameters of the ith

dislocation emitted from a crack tip. b and represent the
Burgers vector and shear modulus, respectively. is the angle
between the crack plane and the slip plane. rc and rj are the
160
Temperature, T / K
140
130
120
150
110
roll-bonded
fully annealed
-4
10-2
ln(d/dt)
-6
10-3
Ea, FA=0.25
Strain rate, / s-1
Ea, ARB=0.20 eV
0.09 eV
-8
103T-1/ K-1
Fig. 7 Arrhenius plots between strain rates and BDT temperatures. Solid
squares and triangles indicate data obtained from fully annealed and rollbonded specimens, respectively. The slops of line give the activation
energy for the BDT. The error in the activation energy is due to the
uncertainness of the BDT temperature in full annealed sample in Fig. 6(a).
the grain size decreases. This contradicts the fact that grain
rening decreases the BDT temperature. Therefore, other
functions of the grain boundary are required to be discussed.
In the next subsection, two functions of the grain boundary
have been discussed: its role as a dislocation sink and source,
and its role in accommodating stress due to the impinged
dislocations via grain boundary shielding. Firstly, the role of
the grain boundary as a dislocation sink and source is
discussed.
(a)
61
4.2 The role of grain boundaries as sinks and sources

Figure 8 indicates the congurations of atoms around a
crack under the application of a tensile stress perpendicular to
the crack plane, as calculated by using molecular dynamics
(MD). 5 symmetric tilt boundaries were set to be lying
parallel to the crack plane. The embedded atom method for
aluminum developed by Mishin el al.29) was used. Although
the crystal structure was dierent from that of low carbon
steel, the MD simulations revealed the role of the grain
boundary the essence of which is considered to be independent of the crystal structures. Periodic conditions were applied
at each boundary of the calculation box. The number of total
atoms was set to be 731640. A tensile stress perpendicular
to the crack plane was applied at a strain rate of 8 108 s1
at 100 K. The details regarding the calculation condition
are presented elsewhere.30) Figures 8(a), 8(b), 8(c) and (d)
present images after the application of a tensile strain normal
to the crack plane, i.e. "zz of 0.016, 0.021, 0.032 and 0.038,
indicating that the dislocations were emitted from the crack
tip with increasing "zz . The rst dislocation emitted from the
crack tip that glided against the grain boundary and then was
impinged against the grain boundary is shown in (b) and (c).
Figure 8(e) illustrates an enlarged image of (c), indicating the
dislocations reemitted from the grain boundary (the dashed
arrows indicate the direction of the dislocation glide).
Figure 8(f) presents an enlarged image of the dashed square
in (d), indicating that the dislocation reemissions continued
as "zz increased. Six dislocations were emitted on reaching
"zz 0:038.
Thus, the MD calculations indicated that grain boundaries
act not only as barriers for dislocation motion but also
(b)
(c)
GB
GB
GB
2
3
(d)
(e)
(f)
GB
3 4
GB
GB
Fig. 8 The congurations of atoms around the crack tip calculated by MD under applying the tensile stress normal to the crack plane.
Tensile stress was applied perpendicular to the crack plane with a strain rate of 8 108 s1 at 100 K. (a), (b), (c) and (d) exhibit snap
shops when "zz is 0.016, 0.021, 0.032 and 0.038, respectively. Gray atoms are fcc structures, red atoms are hcp structures, and blue atoms
are defects. Two grains coloured blue is lying above and below of the crack. The dashed arrows are indicating directions along which the
dislocations are moving. (e) and (f) show an enlarged gure in (c) and (d), respectively.
62
(b)
(a)
(d)
(c)
-1.0
-500
0
0
1.0 [GPa]
500 [MPa]
Fig. 9 Colour maps of the stress eld (z0 x0 ) around a crack tip and a 5
symmetric tilt boundary, calculated by MD. The colour scale in (a) is 1:0
to 1.0 GPa and that in (b)(d) is 500 to 500 MPa. The stress eld which
induces dislocations from the crack tip is set to be negative. Snap shots of
"zz at; (a) 0.017, (b) 0.02, (c) 0.025 and (d) 0.03. The applied stress is
subtracted so that the stress eld due to impinged dislocations is indicated.
as dislocation sources, which implies that grain rening

increases the number of dislocation sources. A quasi-twodimensional desecrate dislocation dynamics simulation at the
positions of the dislocation sources indicated that when the
number of dislocation sources increased, the dislocations
shielded a crack tip more eectively, resulting in the decrease
in the BDT temperature.31)
4.3 Grain boundary shielding
The other role of the grain boundary is to accommodate the
stress intensity at the crack tip. Figure 9 presents colour maps
of the stress eld (z0 x0 ) around the crack tip. A tensile stress
was applied perpendicular to the crack plane and was
increased at a strain rate of 8 108 s1 at 100 K. The stress
was measured at 0.1 K in order to eliminate the eect of
thermal vibrations of atoms. Figure 9(a) indicates the stress
eld at "zz 0:017, where no dislocations were emitted from
the crack tip. The negative sign of z0 x0 is set to be one which
extracts dislocations from the crack tip. Figures 9(b), 9(c)
and (d) indicate the stress elds around the crack tip at
strains of 0.02, 0.025 and 0.03, respectively, when dislocations were emitted from the crack tip and impinged against
the grain boundary. The applied stress was subtracted from
the total stress so that only the stress elds due to the
impinged dislocations are indicated. The MD calculations of
the stress elds revealed that the impinged dislocations
induced stress intensity at the crack tip. The stress intensity
due to the impinged dislocations increased with the number
of dislocations impinged against the grain boundary, as

shown in (b), (c) and (d). The normal stress on the crack
plane, zz , derived from z0 x0 was found to be compressive,
indicating the formation of a shielding eld at the crack tip.
It is termed as grain boundary shielding,31) which is the
essential mechanism required to understand the mechanism
behind the decrease in the BDT temperature and enhancement in the fracture toughness of ne-grained materials at
low temperatures.
5.
Conclusion
The BDT behaviour was investigated in low carbon steel

deformed by an ARB process. The BDT temperature was
decreased by the ARB process. Arrhenius plots between the
BDT temperatures and the strain rates indicated that the
activation energy was not changed by the ARB process.
This indicated that the decrease in the BDT temperature by
grain rening was not due to the increase in the dislocation
mobility for short-range obstacles. MD simulations revealed
that dislocations emitted from a crack were impinged against
grain boundaries and were reemitted as the strain increased
as well as create a shielding eld. It is therefore concluded
that the decrease in the BDT temperature by grain-rening
is due to not the increase in dislocation mobility controlled
by short range obstacles but the increase in the number of
dislocation sources.
Acknowledgement
The authors appreciate Prof. N. Tsuji, Prof. D. Terada, and
Mr. S. Horiuchi for sample preparations. This work is partly
supported by Ministry of Education, Culture, Sports, Science
and Technology, Japan. Grant-in-Aid for Scientic Research
on Priority Areas (18062002) and Grand-in-Aid for Young
Scientists (Start-up: 19860059).
REFERENCES
1)
2)
3)
4)
5)
6)
7)
8)
9)
10)
11)
12)
13)
14)
15)
16)
N. J. Petch: Philos. Mag. 3 (1958) 10891097.

C. St. John: Philos. Mag. 32 (1975) 11931212.
F. C. Serbena and S. G. Roberts: Acta Metall. 42 (1994) 25052510.
M. Tanaka, A. J. Wilkinson and S. G. Roberts: J. Nucl. Mater. 378
(2008) 305311.
A. Giannattasio, M. Tanaka, T. D. Joseph and S. G. Roberts: Phys. Scr.
T 128 (2007) 8790.
T. D. Joseph, M. Tanaka, A. J. Wilkinson and S. G. Roberts: J. Nucl.
Mater. 367370 (2007) 637643.
V. M. Segal, V. I. Reznikov, A. E. Drobyshevskiy and V. I. Kopylov:
Russ. Matall. 1 (1981) 99105.
K. Lu and J. Lu: J. Mater. Sci. Technol. 15 (1999) 193197.
G. Sakai, Z. Horita and T. G. Langdon: Mater. Sci. Eng. A 393 (2005)
344351.
Y. Saito, N. Tsuji, H. Utsunomiya, T. Sakai and R. G. Hong: Scr.
Mater. 39 (1998) 12211227.
N. Tsuji, S. Okuno, Y. Koizumi and Y. Minamino: Mater. Trans. 45
(2004) 22722281.
J. Samuels and S. G. Roberts: Proc. R. Soc. Lond. A 421 (1989) 123.
H. S. Kim and S. Roberts: J. Am. Ceram. Soc. 77 (1994) 30993104.
S. Fujita, K. Maeda and S. Hyodo: Philos. Mag. A 65 (1992) 131147.
P. B. Hirsch, A. S. Booth, M. Ellis and S. G. Roberts: Scr. Mater. 27
(1992) 17231728.
G. Michot, M. A. L. de Oliveira and G. Champier: Mater. Sci. Eng. A
272 (1999) 8389.
17) I. H. Lin and R. Thomson: Acta Metall. 34 (1986) 187206.
18) N. Narita, K. Higashida and S. Kitano: Scr. Metall. 21 (1987) 1273
1278.
19) R. Thomson: J. Mater. Sci. 13 (1978) 128142.
20) J. Weertman: Acta Metall. 26 (1978) 17311738.
21) K. Higashida, M. Tanaka, A. Hartmaier and Y. Hoshino: Mater. Sci.
Eng. A 483484 (2008) 1318.
22) K. Higashida, T. Kawamura, T. Morikawa, Y. Miura, N. Narita and
R. Onodera: Mater. Sci. Eng. A 319321 (2001) 683686.
23) E. O. Hall: Proc. Phys. Soc. B64 (1951) 747753.
24) J. F. Brown, Jr. and J. E. Srawley: ASTM STP 410 (1966) 165.
25) M. Holscher, D. Raabe and K. Lucke: Steel Res. 62 (1991) 567575.
63
26) A. Giannattasio and S. G. Roberts: Philos. Mag. 87 (2007) 25892598.

27) J. M. Hodge, R. D. Manning and H. M. Reichhold: Trans. Am. Inst.
Min. Metall. Eng. 185 (1949) 233240.
28) R. Thomson: Solid State Physics, 39, ed. by F. Seitz and D. Turnbull
(Academic Press, INC., Orlando, San Diego, New York, Austin,
Boston, London, Sydney, Tokyo, Toronto, 1986), pp. 1129.
29) Y. Mishin, D. Farkas, M. J. Mehl and D. A. Papaconstantopoulos:
Phys. Rev. B 59 (1999) 33933407.
30) T. Shimokawa, T. Hiramoto, T. Kinari and S. Shintaku: Mater. Trans.,
doi: 10.2320/matertrans.MD200810.
31) M. Tanaka, N. Fujimoto and K. Higashida: Mater. Trans. 49 (2008)
5863.

Temperature and Strain Rate Dependence of Flow Stress

in Severely Deformed Copper by Accumulative Roll Bonding
Takahiro Kunimine1; * , Naoki Takata2 , Nobuhiro Tsuji3 ,
Toshiyuki Fujii4 , Masaharu Kato1 and Susumu Onaka1
1
Department
Department
3
Department
4
Department
2
of
of
of
of
Materials Science and Engineering, Tokyo Institute of Technology, Yokohama 226-8502, Japan
Metallurgy and Ceramics Science, Tokyo Institute of Technology, Tokyo 152-8552, Japan
Adaptive Machine Systems, Osaka University, Suita 565-0871, Japan
Innovative and Engineered Materials, Tokyo Institute of Technology, Yokohama 226-8502, Japan
Tensile tests and strain rate jump tests have been carried out at low temperatures (77 K room temperature (RT)) using pure Cu
specimens that were severely deformed by accumulative roll bonding (ARB). The dependence of the ow stress on the temperature and the strain
rate has been investigated and the strain rate sensitivity m and its variation caused by the change in the ARB cycle N are discussed. At RT, the
strain rate sensitivity for N 4 stays at about 0.005. However, for N 5, m increases with increasing N to become 0:018 when N 8. The
deformation mechanisms of the ARB processed Cu are discussed with the activation volume V . The temperature dependence of V and its
variation with increasing N are also discussed. [doi:10.2320/matertrans.MD200809]
(Received July 22, 2008; Accepted October 6, 2008; Published November 19, 2008)
Keywords: copper, ultra-ne grained materials, accumulative roll bonding, strain rate sensitivity, activation volume, deformation mechanism
1.
Introduction
The strain rate dependence of ow stress in ultra-ne

grained (UFG) and nanocrystalline (NC) materials has been
investigated extensively.14) The higher strain rate dependence of the ow stress of UFG and NC materials compared
with that of conventional materials has been reported at room
temperature (RT). However, little is known about the
temperature dependence of the ow stress in UFG materials
prepared by severe plastic deformation (SPD) processes.1)
Although models to describe characteristic mechanical
properties of UFG materials have been proposed,37) experimental studies are still necessary to discuss the deformation
mechanisms.
The accumulative roll bonding (ARB) is one of the SPD
processes.8) This process provides homogeneous nanostructures and large bulky metallic materials.9) In the present
study, tensile tests and strain rate jump tests at low temperatures are made for ARB processed pure Cu. The strain rate
sensitivity as well as activation volume obtained by the strain
rate jump tests are the indexes of the strain rate dependence
of the ow stress. The variation of these strain rate
parameters with the ARB cycles (i.e., the plastic strain
applied by ARB) together with the temperature dependence
of strength are discussed.
2.
Experimental
Polycrystalline copper of 99.99% purity was used as a

starting material. After annealing at 873 K for 2 h, copper
sheets with an initial grain size of about 100 mm were
subjected to the ARB process at room temperature (RT) with
lubrication.10) The ARB process was carried out up to N 8
where N is the number of ARB cycles. The equivalent strain
for N 8 is about 6.4 in terms of the logarithmic strain.11)
*Graduate
Student, Tokyo Institute of Technology
The ARB process initially induces microstructure with

elongated grains to the rolling direction. The microstructure
of copper at N 4 possesses the so-called lamellar boundary
structure.12,13) However, the microstructure for N 6 consists of a mixture of relatively equiaxed and slightly
elongated grains.12,13) At N 8, the spacing of high-angle
lamellar boundaries of elongated grains was about 370 nm in
the present study.
Using an electro-discharging machine, tensile specimens
with 10 mm gauge length and 3 mm gauge width were cut
from the ARB processed sheets with 1 mm in thickness.
Although the tensile axis is parallel to the ARB rolling
direction for the 6 and 8-cycle specimens, other ARB
processed specimens have the tensile axis parallel to the
transverse direction of the sheets. Tensile specimens cut from
the starting Cu polycrystal (the annealed Cu before ARB)
were also prepared. Tensile tests were made with an Instrontype testing machine at 77, 175 K (the melting point of
methanol), 273 and 298 K within the accuracy of 2 K.
Strain rate jump tests were conducted between the strain
rates of 8:3 105 s1 and 8:3 104 s1 . The strain rate
sensitivity m dened as
m
@ ln
1 @
@ ln "_
@ ln "_
is an index to evaluate the dependence of the ow stress on

the strain rate "_. When a strain rate jump from "_1 to "_2 causes
the change in the ow stress from 1 to 2 during tensile tests,
the strain rate sensitivity m is given by
m

2 1
:

ln "_ 1 2 =2 ln"_2 ="_1
For comparison, tensile tests of 99.99% copper single

crystals were also performed. The single crystals were grown
by the Bridgman method. Tensile specimens with 10 mm
gauge length and 3 mm width were cut from the single
crystals. The tensile axis of the specimens was chosen so that
400
True Stress,
6 Cycles
300
200
8 Cycles
100
0
0
ARB 5 ~ 8 Cycles
High
Annealed Cu before ARB ~ 4 Cycles

Single Crystal
ARB Cycles
RT
65
The number of
4 Cycles
500
Strain Rate Sensitivity, m
/ MPa
Temperature and Strain Rate Dependence of Flow Stress in Severely Deformed Copper by Accumulative Roll Bonding
Low
RT
0.01 0.02 0.03 0.04 0.05 0.06 0.07
True Shear Stress,
True Plastic Strain,
/ MPa
Fig. 1 The true stress-true strain (") curves of the specimens ARB
processed by 4, 6 and 8 cycles at RT. The steps on the " curves are the
changes in the ow stress caused by the strain rate jump tests between
8:3 105 s1 and 8:3 104 s1 .
Fig. 2 The relationship between the strain rate sensitivity m and the true
shear stress at RT for the conventional polycrystals (the annealed Cu
before ARB), the specimens ARB processed by various cycles and the
single crystal specimens.
the initial plastic deformation occurs by single slip. The strain

rate jump tests were carried out at 77, 175 and 298 K. The
change in the strain rate sensitivity m caused by work
hardening was measured during tensile deformation.
using the Schmid factor of 0.5 for (3) the single crystal
specimens.
In Fig. 2, the data points shown by circles are the results
for the single crystals, which indicate the variation of m due
to the increase in the true shear stress (work hardening
during tensile deformation). Although increases from about
10 to 100 MPa during the work hardening, the strain rate
sensitivity m does not change signicantly and stays between
0.003 and 0.005. These values are in agreement with
previously reported values for work hardened Cu single
crystals.17)
On the other hand, the data points shown by open and
closed diamonds in Fig. 2 are the results for the starting
polycrystals and ARB processed specimens. The m
relations for the starting polycrystals and ARB processed
specimens with N 4 are similar to that for the single
crystals. However, the higher-cycle ARB processes with N
5 markedly increase the m value as shown by closed
diamonds and m reaches about 0.018 when N 8. The
tensile strength of the ARB processed specimens with N 4
was nearly the same as shown in Fig. 1, i.e., the large
increase in m and the almost constant are found for the
higher-cycle ARB processed specimens.
The Cottrell-Stokes (CS) law states that the ratio (=) is
constant for various when ln "_ is xed.1719) Hence, from
eq. (2), the strain rate sensitivity m should be unchanged
during plastic deformation if the CS law is obeyed. The
validity of the CS law in f.c.c. single crystals during work
hardening has been discussed in previous studies.17,20,21) In
contrast, the ARB processed Cu at N 8 shows the value of
m reaching 0.018. Such high values of m about 0.02 have
been also reported for UFG Cu processed by ECAP.14) In
fact, it is known that nanocrystalline Cu and Ni show high
values of m such as 0.03622) and 0.019.6)
3.
Results
3.1 Stress-strain curves at room temperature

Figure 1 shows the true stress-true strain (") curves of
specimens ARB processed by 4, 6 and 8 cycles at RT, which
were simply converted from the nominal stress-nominal
strain curves. Since uniform deformation is assumed to
obtain true stress and true strain, the curves after macroscopic
necking are not correct, strictly speaking. Although the
tensile strength of the starting polycrystal before ARB was
about 220 MPa, it increased with increasing the number of
ARB cycles and nearly saturated (450 MPa) even at N 4.
This saturation stress is slightly higher than 400 MPa reported
in a previous study.13) Another change in mechanical
properties caused by ARB is the increase in ductility after
4 cycles: nearly twice larger at N 8 compared with that at
N 4. A similar increase in ductility with increasing the
number of ARB cycles or ECAP passes has also been found
in previous studies.1316)
The changes in the ow stress associated with the
sudden change of strain rate by a factor of ten amounted to
9 MPa at N 4 and 18 MPa at N 8. Generally speaking,
increased with increasing the number of ARB cycles.
3.2 Strain rate sensitivity at room temperature
The results of the strain rate jump tests for (1) the starting
polycrystals (the annealed Cu before ARB), (2) the ARB
processed specimens and (3) the single crystal specimens
are summarized in Fig. 2 as the relationship between the
strain rate sensitivity m and the ow stress that were
converted into the true shear stress . We have evaluated
the true shear stress from the true tensile stress by
using the Taylor factor MT 3:06 for (1) the starting
polycrystals and (2) the ARB processed specimens, or by
3.3
Stress-strain curves and strain rate sensitivity at low

temperatures
Figure 3 shows the temperature dependence of the true
stress-true strain (") curves of ARB processed specimens
T. Kunimine et al.
66
True Stress,
/ MPa
800
ARB 8 Cycles
77K
600
175K
400
273K
200
RT
0
0
0.05
0.1
ARB 8 Cycles
ARB 6 Cycles
ARB 4 Cycles
Annealed Cu before ARB
Single Crystal
0.15
Temperature, T / K
True Plastic Strain,
Fig. 3 The true stress-true strain (") curves of the 8-cycle ARB
specimens tested at 77, 175, 273 K and RT.
Change in m with increasing

the number of ARB cycles
RT
Single Crystal
273K
175K
RT
77K
175K
77K
Fig. 5 The relationship between m and T. The curves in this gure giving
the T dependence of m are obtained by considering m at certain strains, i.e.,
" 0:01 for the specimens ARB processed by 4, 6 and 8 cycles, " 0:15
for the annealed Cu before ARB and " 0:02 for the single crystals.
4.
Discussion
4.1 Activation volume at room temperature

The rate equation to describe thermally activated plastic
ow is given by
3
_ _0 expG =kT;
where _ is the shear strain rate, _0 the pre-exponential
constant, k the Boltzmann constant and G the activation
energy.24,25) The activation volume V is given by24)
V @G =@ kT@ ln _ =@;
True Shear Stress, / MPa

Fig. 4 The relationship between the strain rate sensitivity m and the true
shear stress for the usual polycrystals (the annealed Cu before ARB), the
specimens ARB processed by various cycles and the single crystal
specimens tensile-tested at 77, 175, 273 K and RT.
by 8 cycles. With decreasing the test temperature T from RT

to 77 K, both strength and ductility increase. However, with
decreasing T, caused by the strain rate jump test
decreases. As a result, the strain rate sensitivity m given by
eq. (2) decreases with decreasing T. Figure 4 shows the m
relations at various temperatures. As expected, the m curve
moves towards lower m and higher with decreasing T, as
observed earlier in Cu single crystals.17) The higher-cycle
ARB processes with N 5 increase m at 273 and 175 K.
However, this eect of N on m is less pronounced at 77 K.
Figure 5 shows the relationships between m and T. The
data in this gure were obtained by plotting m measured
at a certain strain: " 0:01 for the ARB processed specimens, " 0:15 for the annealed Cu before ARB and "
0:02 for the single crystals. We nd that m decreases with
decreasing T for all the specimens. We can see that the
dependence of the m value on the number N of the ARB
cycles decreases as temperature lowers. A similar decrease
in m with decreasing temperature between 273 and 523 K
has also been found for aluminium processed by severe
plastic deformation.16,23)
where is the applied shear stress. Then, from eqs. (1) and
(4), V is written as a function of m:
V MT
kT@ ln "_
kT
MT
:
@
m
Although the strain rate sensitivity m itself is a phenomenological parameter obtained by mechanical testing, m can be
related to the activation volume V by eq. (5).
The activation volume V is also a value connected with
the geometry of the activation process. Let us suppose that
many short-range obstacles to the motion of dislocations are
dispersed in materials and the activation process of plastic
deformation is the interaction between a dislocation and a
short-range obstacle. Then, the activation volume V is
written as
V Ld b;
Where L is the dislocation length which takes part in a local

activation event, d the average activation distance and b the
magnitude of the Burgers vector. The product Ld is the
activation area swept by a dislocation during the activation
event.
Figure 6 shows the relationships between the true shear
stress and the activation volume V (calculated from
eq. (5)) for the specimens tested at RT. Since m was almost
constant during work hardening of the single crystals, it is
natural that the calculated V of the single crystals is
inversely proportional to the true shear stress .26) Similar
67
Activation Volume, V* / b3
ECAP 1 and 2 Passes
ARB Cycles
Low
The number of
Activation Volume, V* / b 3
RT
ECAP 4, 8, 12 and 16 Passes
Single Crystal
Annealed Cu before ARB ~ 4 Cycles
High
ARB 5 ~ 8 Cycles
True Shear Stress,
/ MPa
Fig. 6 The activation volume V as a function of the true shear stress at

RT. The previous results given by Dalla Torre et al.3,4) for ECAPed Cu are
also shown by the hatched regions in this gure.
Annealed Cu before ARB

ARB 4 Cycles
ARB 6 Cycles
ARB 8 Cycles
Vapor Deposited Cu (d = 500nm)
Temperature, T / K
V relation is also found for the starting polycrystals and
the ARB processed specimens with N 4. As shown in the
log-log plot of Fig. 6, these data points lie on a single line
with a slope of 1.
To understand the V / 1= relation, let us consider the
Bailey-Hirsch equation
p
b ;
7
where is the density of forest dislocations, a constant
(0.30.6) and the shear modulus.27) When the forest
dislocations are the thermal obstacles to the motion of
primary dislocations, the spacing between the obstacles L can
be considered as the dislocation length assigned to one
activation event. Therefore, we have
p
L 1= :
8
From eqs. (6), (7) and (8) and assuming d b, we indeed
have the V / 1= relation of the form
V b3 =:
On the other hand, the obvious deviation from the V /

1= relation with increasing number of ARB cycles must be
explained in a dierent manner. Dalla Torre et al. have also
reported such a decrease in V for Cu severely processed by
ECAP.3,4) For comparison, their data are included in Fig. 6.
The anomalous decrease in V appears to be common to Cu
specimens that were severely deformed by ARB or by ECAP.
This point will be discussed below.
4.2
The temperature dependence of the activation

volume
Using the results shown in Fig. 5 and eq. (5), we have
obtained the temperature T dependence of the activation
volume V . Figure 7 shows the V T relations for various
specimens. For the starting polycrystals and ARB processed
specimens at N 4, V decreases with decreasing T, as in
the case of coarse-grained polycrystals.5) Generally speaking,
ow stress is lower and work hardening is less signicant at
Fig. 7 The activation volume V as a function of the temperature T. The

previous results for vapor deposited Cu (d 500 nm) shown in the
literature5) are also indicated.
higher temperatures. This means that some of the obstacles to

the motion of dislocations are not eective and L is larger at
higher temperatures. Moreover, the activation distance d is
usually an increasing function of temperature (decreasing
function of stress). Since both L and d become larger as
temperature becomes higher, the experimental result that V
is an increasing function of T can be understood naturally.
However, Fig. 7 shows that the T dependence of V starts
to change with increasing the number of ARB cycles N. For
the ARB processed specimens at N 6, V is almost
independent of T and, then, V certainly increases with
decreasing T in the ARB processed specimens at N 8. The
same T dependence of V also has been observed in vapor
deposited Cu (d 500 nm),5) as shown in Fig. 7. Judging
from the fact that Wang et al. reported the same temperature
dependence of V for NC Ni (d 30 nm),6) it appears that
the change in the temperature dependence of V with
decrease in the grain size is a universal phenomenon in
f.c.c. UFG metals. As shown in Fig. 7, the transition in the T
dependence of V occurs in the present ARB processed Cu
at N 6. This critical number N 6 of the ARB cycles
coincides reasonably with the onset of the anomalous
increase of m in Fig. 2 and with the start of the anomalous
decrease in V in Fig. 6.
To discuss this transition, it is necessary to consider
changes in microstructure with increasing the ARB cycle.
Grain renement occurs markedly by the initial ARB process
up to N 2.28) Afterward, the spacing of high-angle lamellar
boundaries slightly decreases with increasing N and approaches to a constant size, about 350 nm when N 4.28) In
addition to the formation of lamellar boundary structure, the
ARB process initially induces deformation microstructure
consisting of low-angle boundaries.28) However, the fraction
68
T. Kunimine et al.
be regarded as an attempt to nd the transition of the

deformation mechanism. Further experimental and theoretical studies will reveal whether or not the present discussion,
analysis and assumptions are indeed eective and reasonable.
(a) Low Temperature
Boundary Pinning Point
5.
LGB
Grain Boundary
(b) Room Temperature
LGB
Fig. 8 The schematic illustration showing the shapes of bowing-out
dislocations emitted from grain-boundary sources between pinning
obstacles (a) at low temperature, (b) at room temperature.
Conclusions
The most important ndings of this study on copper after

severe plastic deformation by ARB are:
(1) At RT, the values of the strain rate sensitivity m for the
single crystals and the polycrystals before ARB are about
0.005. The low-cycle ARB processes with N 4 do not
aect m. However, the high-cycle ARB processes with N 5
increase the values of m to be 0.018 when N 8.
(2) The strain rate sensitivity m monotonically decreases
with decreasing the temperature from room temperature.
Although the increase in m caused by the increase in the ARB
cycle is clear at RT when N 5, this eect of N on m is less
pronounced at 77 K.
(3) The ow stress and temperature dependence of the
activation volume V suggests that the transition of deformation mechanisms in the ARB processed specimens occurs
at the ARB cycle N 6.
Acknowledgements
fHAGB of high-angle boundaries in the microstructure

increases when N 2. The N dependence of fHAGB shown
in previous studies for Cu is fHAGB 30% at N 2,28)
fHAGB 55% at N 628) and fHAGB 60% at N 8.29)
When N 6, both the ultra-ne grains and the large fraction
of high-angle boundaries are developed in the ARB processed Cu.28,29) An explanation of the transition in the T
dependence of V is possible if we assume that the ratecontrolling deformation mechanism changes, as will be
discussed below, when such characteristic microstructures
develop in Cu.
Instead of the interaction between dislocations and the
intra-grain obstacles, Kato et al. have proposed a dislocation
bow-out model to explain mechanical properties of UFG
materials.7) In this model, a glide dislocation is considered to
be nucleated at grain-boundary sources and bow out between
two pinning points on the boundary. Plastic deformation
occurs when a dislocation takes an unstable shape under the
applied stress. The dislocation bow-out between pinning
points on a grain boundary is illustrated in Fig. 8. The
boundary pinning points are most probably short-range
obstacles for the detachment of dislocations from the
boundary, as pointed out by Wang et al.6) Then, a thermal
activation process can operate and a smaller dislocation bowout is sucient to thermally overcome the obstacles as
temperature becomes higher. As a result, as shown in Fig. 8,
the associated activation area (shaded area) becomes smaller
at a higher temperature. Therefore, when the thermally
activated escape of a dislocation from boundary pinning
points is rate-controlling, the anomalous V dependence on T
can be explained at least qualitatively.
It is now generally believed that for plastic deformation of
UFG and NC materials, grain-boundary dislocation sources
becomes more important than in-grain sources as grain size
becomes smaller.30,31) In this respect, the present study may
This research was supported by a Grant-in-Aid for

Scientic Research on Priority Areas Giant straining
process for advanced materials containing ultra-high density
lattice defects (18062002) by the Ministry of Education,
Culture, Sports, Science and Technology of Japan.
REFERENCES
1) Y. M. Wang and E. Ma: Appl. Phys. Lett. 83 (2003) 31653167.
2) Q. Wei, S. Cheng, K. T. Ramesh and E. Ma: Mater. Sci. Eng. A 381
(2004) 7179.
3) F. H. Dalla Torre, E. V. Pereloma and C. H. J. Davies: Scr. Mater. 51
(2004) 367371.
4) F. H. Dalla Torre, E. V. Pereloma and C. H. J. Davies: Acta Mater. 54
(2006) 11351146.
5) H. Conrad: Mater. Sci. Eng. A 341 (2003) 216228.
6) Y. M. Wang, A. V. Hamza and E. Ma: Acta Mater. 54 (2006) 2715
2726.
7) M. Kato, T. Fujii and S. Onaka: Mater. Trans. 49 (2008) 12781283.
8) Y. Saito, N. Tsuji, H. Utsunomiya, T. Sakai and H. G. Hong: Scr.
Mater. 39 (1998) 12211227.
9) B. L. Li, N. Tsuji and N. Kamikawa: Mater. Sci. Eng. A 423 (2006)
331342.
5888.
11) B. S. Altan, I. Miskioglu, G. Purcek, R. R. Mulyukov and R. Artan
(eds.): Severe Plastic Deformation, Towards Bulk Production of
Nanostructured Materials, (NOVA Science Publishers, New York,
2006) pp. 543564.
12) N. Takata, K. Yamada, K. Ikeda, F. Yoshida, H. Nakashima and N.
Tsuji: Mater. Trans. 48 (2007) 20432048.
13) N. Takata, S.-H. Lee, C.-Y. Lim, S.-S. Kim and N. Tsuji: J. Nanosci.
Nanotechno. 7 (2007) 15.
14) H.-W. Kim, S.-B. Kang, N. Tsuji and Y. Minamino: Acta Mater. 53
(2005) 17371749.
15) H. W. Hoppel, J. May and M. Goken: Adv. Eng. Mater. 6 (2004)
781784.
16) H. W. Hoppel, J. May, P. Eisenlohr and M. Goken: Z. Metallkd. 96
(2005) 566571.
17)
18)
19)
20)
21)
22)
W. Bochniak: Acta Metall. Mater. 43 (1995) 225233.

A. H. Cottrell and R. J. Stokes: Proc. Roy. Soc. A 233 (1955) 1734.
M. A. Adams and A. H. Cottrell: Philos. Mag. 46 (1955) 11871193.
W. Bochniak: Acta Metall. Mater. 41 (1993) 31333140.
S. Saimoto and H. Sang: Acta Metall. 31 (1983) 18731881.
L. Lu, R. Schwaiger, Z. W. Shan, M. Dao, K. Lu and S. Suresh: Acta
Mater. 53 (2005) 21692179.
23) J. May, H. W. Hoppel and M. Goken: Scr. Mater. 53 (2005) 189194.
24) G. B. Gibbs: Phys. Stat. Sol. 10 (1965) 507512.
25) B. de Meester, C. Yin, M. Doner and H. Conrad: Rate processes in
plastic deformation of materials, (ASM, Metals Park, Ohio, 1975)
69
pp. 175226.
26) K. Saitou and T. Hikage: J. Japan Inst. Metals 49 (1985) 344351.
27) H. Mughrabi: Constitutive Equations in Plasticity, (M. I. T. Press,
Cambridge, Mass., 1975) pp. 199250.
28) B. L. Li, N. Shigeiri, N. Tsuji and Y. Minamino: Mater. Sci. Forum
503504 (2006) 615620.
29) N. Shigeiri: Master Thesis, Osaka University (2006).
30) S. Cheng, J. A. Spencer and W. W. Milligan: Acta Mater. 51 (2003)
45054518.
31) Y. J. Li, X. H. Zeng and W. Blum: Acta Mater. 52 (2004) 5009
5018.

Cyclic Deformation Behavior of Ultra-Fine Grained Copper

Produced by Equal Channel Angular Pressing
Yoji Furukawa1; *1 , Toshiyuki Fujii2; *2 , Susumu Onaka1 and Masaharu Kato1
1
2
Department of Materials Science and Engineering, Tokyo Institute of Technology, Yokohama 226-8502, Japan
Department of Innovative and Engineered Materials, Tokyo Institute of Technology, Yokohama 226-8502, Japan
Cyclic deformation behavior of ultra-ne grained (UFG) Cu of 99.99% purity processed by equal-channel angular pressing was
investigated. In tension-compression fatigue tests under strain control, UFG Cu showed cyclic softening. Shear bands were formed along the
direction inclined by about 45 from the loading axis. Observations using an electron backscattering diraction technique and transmission
electron microscopy revealed that local grain growth took place in the shear bands and overall grains elongated along the shear direction. Cyclic
softening can be understood as a result of dynamic grain coarsening occurred intensively in the strain localized shear bands.
Keywords: ultra-ne grained copper, equal channel angular pressing, cyclic deformation, cyclic softening, shear bands, grain coarsening
1.
Introduction
In recent years, much attention has been given to the

mechanical properties of ultra-ne grained (UFG) materials
produced by severe plastic deformation (SPD) techniques
such as equal channel angular pressing (ECAP),14) accumulative roll-bonding (ARB)5,6) and high-pressure torsion
(HPT).7,8) In considering practical applications, the cyclic
deformation and fatigue properties of UFG materials have
been studied extensively. In particular, the following
characteristic features in cyclic deformation of UFG Cu
processed by ECAP have been sporadically reported: (1)
cyclic softening is frequently detected as a stress-strain
response,913) (2) macroscopic shear bands (SBs) are formed
inhomogeneously,9,10,1318) (3) coarse grains are locally
developed.9,12,14,15,1921) For example, Wu et al.14,15) observed no grain coarsening, while they found apparent SBs
by scanning electron microscopy (SEM) observation. Therefore, they concluded that there is no relationship between the
SB formation and the grain coarsening. Contrary to this
result, Mughrabi and Hoppel19) suggested that the grain
coarsening is closely related to the formation of the SBs and
proposed two possible SB formation mechanisms. As far as
the authors know, there has been no consistent explanation
to understand all the above features of the cyclic deformation of UFG Cu. To obtain a thorough understanding of the
microstructural changes during cyclic deformation of UFG
materials, it is essential to conduct precise observations with
the combined use of SEM and transmission electron
microscopy (TEM).
In this study, plastic-strain-controlled cyclic deformation
tests were performed using high purity UFG Cu produced by
ECAP. Based on the results of SEM and TEM observations,
the relationship among cyclic softening, SB formation and
local grain coarsening was discussed.
2.
UFG copper (99.99% purity) billets of 10 mm in diameter

and 60 mm in length were processed by the ECAP technique.
Each billet was subjected to pressing for 8 passes in such a
way that after each pressing the billet was rotated by 90 in
longitudinal direction, usually denoted as route Bc.3,4)
Specimens for fatigue tests with the gauge dimensions of
4 mm 6 mm 10 mm were taken from the central part of
the ECAP billets by spark erosion in the direction parallel to
the rod axis. The specimens were electrolytically polished to
avoid any inuence of mechanical pre-treatment. Fully reversed tension-compression low-cycle fatigue tests were carried out at room temperature (RT) under the constant plasticstrain amplitudes ranged from "pl 2 104 to "pl 2
103 using an electro-hydraulic testing machine (Shimadzu
Servopet). Strain was measured with an extensometer
mounted directly on the gauge section and a constant strain
rate of 1 103 s1 was employed using a triangular command signal. The stress response was monitored on a digital
oscilloscope. For microstructural observation, the fatigue
tests were interrupted when the stress amplitude decreased
and reached 85% of the maximum stress amplitude.
Surface observations were performed on both as-pressed
(as-ECAPed) and fatigued specimens by using a Hitachi S4300 scanning electron microscope equipped an electron
backscattering diraction (EBSD) system (Oxford INCA
crystal). Grain maps were constructed with an individual step
size of 40 nm and a lower cut-o misorientation angle of 5 .
The fatigued specimens were sliced into 3 mm discs and
ground down to 0.2 mm using silicon-carbide paper. Final
thin foils were prepared by electrolytic polishing, and then
TEM observations were carried out on a JEOL 2011
microscope at an accelerating voltage of 200 kV.
3.
*1Graduate
Student, Tokyo Institute of Technology. Present address:

Nippon Steel Corporation, Kimitsu 299-1141, Japan
*2Corresponding author, E-mail: fujii.t.af@m.titech.ac.jp
3.1 Cyclic stress-strain response

Cyclic hardening/softening curves of UFG Cu specimens
Cyclic Deformation Behavior of Ultra-Fine Grained Copper Produced by Equal Channel Angular Pressing
800
pl =2 10-3
pl =1 10-3
pl =5 10-4
pl =2 10-4
350
Stress amplitude, a / MPa
Stress amplitude, a / MPa
400
300
250
10
Cumulative plastic strain, cum

Fig. 1 Cyclic hardening/softening curves of specimens cyclically deformed under plastic-strain amplitudes ranged from "pl 2 104 to
"pl 2 103 .
Cyclic softening ratio
600
Present study
Hppel et al. (2002)
UFG Cu
400
200
100
80
60
200
0.1
71
10-4
Mughrabi and Wang (1981)

Polycrystalline Cu
10-3
10-2
Plastic strain amplitude, pl

Fig. 3 Pseudo-cyclic stress-strain curve of UFG Cu, comparison with
previously obtained data of UFG Cu12) and conventional coarse-grained
polycrystalline Cu.22)
0.2
by Hoppel et al.12) and those on conventional coarse-grained

polycrystalline Cu by Mughrabi and Wan22) are also shown in
Fig. 3. It is found that there is a good agreement between the
present result and that obtained by Hoppel et al.12) For the
cyclic stress-strain response of polycrystalline cubic metals,
the power-law relationship
0.1
n
a k"pl
0
10-4
10-3
10-2
Plastic strain amplitude, pl

Fig. 2 Plastic-strain-amplitude dependence of the amount of cyclic softening.
are shown in Fig. 1. Cyclic softening apparently occurs at

all plastic-strain amplitudes as previously reported in the
literature.912,19) To discuss the strain-amplitude dependence
of the amount of cyclic softening, a cyclic softening ratio was
dened as 1 "3 =Max . Here, "3 is the stress amplitude at
the late stage of "cum 3 and Max is the maximum stress
amplitude. A higher ratio indicates a higher degree of cyclic
softening. Figure 2 shows a plot of the cyclic softening ratio
against the plastic-strain amplitude. Although the cyclic
softening ratio at "pl 5 104 is exceptionally small, the
cyclic softening becomes more remarkable at the higher
plastic-strain amplitude. Therefore, it is reasonable to
conclude that the amount of cyclic softening increases with
increasing the plastic-strain amplitude.
Since the cyclic hardening/softening curves shown in
Fig. 1 did not exhibit stress saturation, the pseudo-cyclic
stress-strain curve was drawn in Fig. 3 by using the values of
stress amplitudes at the cumulative plastic strain of "cum 1.
The results obtained from strain-controlled tests on UFG Cu
has been established,23) where k is the cyclic strength

coecient and n is the cyclic strain hardening exponent.
Applying eq. (1) to the experimental results of the UFG Cu,
the values of k 1060 MPa and n 0:168 were received.
Similarly, the values of k 376 MPa and n 0:175 were
obtained in the case of the conventional coarse-grained
polycrystalline Cu. Although the stress levels are quite
dierent between the two kinds of specimens, the values of
n are almost the same. This implies that the multiple slip
deformation may occur even in the UFG Cu specimens as
expected in the coarse-grained polycrystalline Cu.
3.2 Microstructural characterization
Figure 4 shows a set of SEM images taken from specimens
cyclically deformed at various plastic-strain amplitudes. On
the surfaces of all deformed specimens, so-called SBs were
developed. The length of the SBs varied widely from 1 mm to
100 mm in each specimen. The morphology of the SBs was
quite similar to the extrusions frequently formed in persistent
slip bands of fatigued Cu single crystals.2426) The SBs were
also inclined by about 45 from the stress axis. This indicates
that the SBs are formed parallel to the planes of maximum
resolved shear stress. Since micro-crack initiation along the
SBs was detected, strain localization indeed occurs in the SBs
during cyclic deformation. These structural features of the
SBs characterized by the present study are consistent with
those reported in the literature.9,11,14,16,18)
Changes in grain size and grain morphology of the fatigued
specimens were observed using the EBSD method. Grain
maps obtained from an as-ECAPed material and a specimen
72
Y. Furukawa, T. Fujii, S. Onaka and M. Kato
(b)
(c)
(d)
stress axis
(a)
2m
Fig. 4 SEM images of shear bands formed during cyclic deformation at (a) "pl 2 104 , (b) "pl 5 104 , (c) "pl 1 103 and
(d) "pl 2 103 . Stress axis is in the vertical direction.
ion direction
sion
xtrus
extru
(a)
stress axis
(b)
2m
Fig. 5 EBSD grain maps obtained from (a) an as-ECAPed specimen and
(b) a specimen cyclically deformed to "cum 3:3 at "pl 2 104 .
fatigued at "pl 2 104 are shown in Figs. 5(a) and 5(b),

respectively. It can be seen from Fig. 5(a) that ne grains
with an average size of approximately 500 nm were formed
after 8 passes of ECAP. As shown in Fig. 5(b), coarse grains
with sizes of about 10 mm long were locally developed in the
cyclically deformed specimen. These grains were banded
along the direction almost parallel to the trace of SBs (see
Fig. 4(a)). Here, it is needed to pay attention to the actual
sizes of the coarsened grains since small-angle boundaries

with misorientations of up to 5 are indistinguishable in
our SEM observation. The results of TEM observations to
elucidate the grain interior in the fatigued specimens will be
shown later in this paper.
It also turns out from Fig. 5(b) that the other major grains
slightly grow up to 12 mm and elongate into the same
direction as that observed in the coarse grains. Although Wu
et al.,14) Kunz et al.,13) Lukas et al.17,27) and Xu et al.18)
conducted SEM observations in fatigued commercial-purity
UFG Cu (99.8%99.9%) produced by ECAP, they all
concluded that no detectable grain coarsening took place
during cyclic deformation. This result is obviously in
disagreement with the present result. On the other hand,
Hoppel et al.12,20,28) have reported that shear banding and
pronounced grain coarsening take place during cyclic
deformation of high-purity UFG Cu (99.99%), while they
have not observed the grain elongation as represented in the
present study. Therefore, the results of our study accord
qualitatively with those obtained by Hoppel et al.12,20,28)
From all these ndings, it is evident that the stability of UFG
structure during cyclic deformation strongly depends on the
purity of materials.
The relationship between the SB formation and the local
grain coarsening is one of the most interesting problems in
cyclic deformation of UFG Cu. In the previous studies
mentioned above,13,14,17,18,27) it has been concluded that there
is no relationship between the SB formation and the grain
coarsening since as-ECAPed grain structures are kept in
fatigued specimens with clear SBs. On the contrary,
Mughrabi and Hoppel,19) and Hoppel et al.20,28) have insisted
that the grain coarsening plays a dominant role in the
formation of the SBs. From Fig. 4(a) and Fig. 5(b), it is
reasonable to conclude in the present study that the grain
coarsening is closely related to the SB formation. The
following experimental results of TEM observations will
give strong evidence for this conclusion.
73
500nm
Fig. 6
TEM photograph taken from an as-ECAPed specimen.
1m
Fig. 7 TEM photograph taken from a specimen cyclically deformed to "cum 3:3 at "pl 2 104 . Stress axis is in the vertical
direction. Subgrain structure was developed heterogeneously in the specimen.
The grain morphology in a specimen after ECAP through 8

passes is shown in Fig. 6. Dense dislocations are found to be
tangled in the grain interiors. In contrast, it is obvious from
Fig. 7 that local grain growth occurred characteristically in a
specimen cyclically deformed at "pl 2 104 . It is noteworthy that the coarse grains having a low dislocation density
are aligned along the direction approximately parallel to the
SBs as represented in Fig. 5(b). The average size of the
coarse grains was measured as about 2 mm, and this value is
much smaller than that obtained from the grain map shown in
Fig. 5(b). It can be readily understood from Figs. 5(b) and 7
that each elongated grain about 10 mm in length actually
consists of subgrains whose size is in the range of 13 mm.
From our multi-scale observations of the SBs by SEM and
TEM, we can justify the conclusion that the plastic strain
localization during cyclic deformation of UFG Cu induces
the dynamic grain/subgrain growth and causes the development of the SBs.
74
Y. Furukawa, T. Fujii, S. Onaka and M. Kato
4.
Conclusions
Plastic-strain-controlled cyclic deformation of high purity

UFG Cu (99.99%) produced by ECAP was performed at RT.
The stress-strain response and the microstructural change
were investigated. The conclusions can be summarized as
follows.
(1) UFG materials show cyclic softening at all plasticstrain amplitudes.
(2) SBs are developed along the planes of maximum
resolved shear stress. Grain coarsening occurs intensively in the SBs and gives rise to subgrain formation.
(3) Cyclic softening can be understood as a strain localization phenomenon and is closely related to the SB
formation and the dynamic grain coarsening.
1m
Fig. 8 Dislocation microstructure developed in a locally coarsened grain of

a specimen cyclically deformed to "cum 7:7 at "pl 5 104 .
3.3
Correlation among cyclic softening, shear banding

and local grain coarsening
In the present study, isolated coarse grains with sizes of
13 mm were also observed frequently in specimens cyclically deformed at various strain amplitudes. A typical result
is shown in Fig. 8. Since the dislocation wall structure is
well developed in this isolated coarse grain as originally
reported by Agnew and Weertman,9) it may be expected
that the strain localization takes place even in the isolated
grain. This is consistent with the fact that the length of
the SBs varied widely from 1 mm to 100 mm. To explain
the formation mechanism of the SBs, Mughrabi and
Hoppel19) have proposed two possible scenarios. One is
that grain/subgrain coarsening is locally initiated and leads
to SBs in a large scale. The other is that a catastrophic
extended SB rstly happens and subsequently triggers the
formation of a new coarsened microstructure within the SB.
A reasonable implication of our observations is that the
former scenario is more likely to occur as the SB formation
mechanism.
As shown in Fig. 1, the cyclic softening was observed at all
plastic-strain amplitudes. The SB formation and the grain
growth were also detected by the SEM and TEM observations. These ndings inevitably raise the question of whether
the development of the SBs is related to the cyclic softening.
Xu et al.18) and Kunz et al.13) have concluded that there is not
a one-to-one correlation between the SB formation and the
cyclic softening since they both have found distinct SBs even
in the case of cyclic hardening. Moreover, they could provide
no microstructural evidence of the cyclic softening. On the
other hand, the present study gives strong evidence for the
direct relationship between the SB development and the local
grain coarsening. Therefore, it is quite reasonable to suppose
that the cyclic softening is closely related to the strain
localization and the resultant shear-banding phenomenon,
while the purity of materials strongly aects the stability of
the UFG structure.
Acknowledgments
The authors wish to express their gratitude to Professor
Zenji Horita, Kyushu University, Professors Hisashi Sato and
Yoshimi Watanabe, Nagoya Institute of Technology, for
performing the ECAP processing of the samples used in this
study. This research was supported by a Grant-in-Aid for
Scientic Research on Priority Areas (18062002) by the
Ministry of Education, Culture, Sports, Science and Technology of Japan.
REFERENCES
1) V. M. Segal, V. I. Reznikov, A. E. Drobyshevskiy and V. I. Kopylov:
Russi. Metall. 1 (1981) 99105.
2) R. Z. Valiev, N. A. Krasilnikov and N. K. Tsenev: Mater. Sci. Eng. A
137 (1991) 3540.
3) Y. Iwahashi, Z. Horita, M. Nemoto and T. G. Langdon: Acta Mater. 46
(1998) 33173331.
4) M. Furukawa, Y. Iwahashi, Z. Horita, M. Nemoto and T. G. Langdon:
Mater. Sci. Eng. A 257 (1998) 328332.
5) Y. Saito, H. Utsunomiya, N. Tsuji and T. Sakai: Acta Mater. 47 (1999)
579583.
6) N. Tsuji, Y. Saito, H. Utsunomiya and S. Tanigawa: Scr. Mater. 40
(1999) 795800.
7) N. A. Smirnova, V. I. Levit, V. I. Pilyugin, R. I. Kuznetsov, L. S.
Davydova and V. A. Sazonova: Fiz. Metal. Metalloved. 61 (1986)
11701177.
8) A. P. Zhilyaevb, G. V. Nurislamovac, B.-K. Kimd, M. D. Baroa, J. A.
Szpunard and T. G. Langdon: Acta Mater. 51 (2003) 753765.
9) S. R. Agnew and J. R. Weertman: Mater. Sci. Eng. A 244 (1998)
145153.
10) S. R. Agnew, A. Vinogradov, S. Hashimoto and J. R. Weertman:
J. Electro. Mater. 28 (1999) 10381044.
11) A. Vinogradov and S. Hashimoto: Mater. Trans. 42 (2001) 7484.
12) H. W. Hoppel, Z. M. Zhou, H. Mughrabi and R. Z. Valiev: Phil. Mag. A
82 (2002) 17811794.
13) L. Kunz, P. Lukas and M. Svoboda: Mater. Sci. Eng. A 424 (2006)
97104.
14) S. D. Wu, Z. G. Wang, C. B. Jiang, G. Y. Li, I. V. Alexandrov and
R. Z. Valiev: Scr. Mater. 48 (2003) 16051609.
15) S. D. Wu, Z. G. Wang, C. B. Jiang, G. Y. Li, I. V. Alexandrov and
R. Z. Valiev: Mater. Sci. Eng. A 387389 (2004) 560564.
16) M. Goto, S. Z. Han, T. Yakushiji, C. Y. Lim and S. S. Kim: Scr. Mater.
54 (2006) 21012106.
17) P. Lukas, L. Kunz and M. Svoboda: Metall. Mater. Trans. A 38A
(2007) 19101915.
18) C. Xu, Q. Wang, M. Zheng, J. Li, M. Huang, Q. Jia, J. Zhu, L. Kunz and
M. Buksa: Mater. Sci. Eng. A 475 (2008) 249256.
19) H. Mughrabi and H. W. Hoppel: Mat. Res. Sym. Proc. 634 (Mat. Res.
Soc., 2001) pp. B2.2.1B2.2.11.
20) H. W. Hoppel, M. Brunnbauer, H. Mughrabi, R. Z. Valiev and A. P.
Zhilyaev: Proc. Materials Week 2000, (Wiley-VCH, 2001) http://
www.materialsweek.org/proceedings.
21) S. D. Wu, Z. G. Wang, C. B. Jiang and G. Y. Li: Phil. Mag. Lett. 82
(2002) 559565.
22) H. Mughrabi and R. Wang: Proc. 2nd Ris Int. Sym. Metall. Mater.
Sci., ed. by N. Hansen et al., (Ris National Lab., 1981) pp. 8798.
23) S. Suresh: Fatigue of Materials, (Cambridge, Cambridge University
Press, 1998) p. 94.
24)
25)
26)
27)
75
J. M. Finney and C. Laird: Phil. Mag. 31 (1975) 339366.

Z. S. Basinski and S. J. Basinski: Scripta Metall. 18 (1984) 851856.
B.-T. Ma and C. Laird: Acta Metall. 37 (1989) 325336.
P. Lukas, L. Kunz, M. Svoboda, M. Buksa and Q. Wang: Proc. H.
Mughrabi Honorary Symposium, ed. by K. J. Hsia, M. Goken, T.
Pollock, P. D. Portella and N. R. Moody, (TMS, 2008) pp. 161166.
28) H. W. Hoppel, C. Xu, M. Kautz, N. Barta-Schreiber, T. G. Langdon and
H. Mughrabi: 2nd Int. Conf. on Nanomaterials by Severe Plastic
Deformation, ed. by M. J. Zehetbauer and R. Z. Valiev, (Vienna,
Wiley-VCH, 2002) pp. 677683.

Fabrication and Characterization of Supersaturated Al-Mg Alloys

by Severe Plastic Deformation and Their Mechanical Properties
Kenji Kaneko1;2; *1 , Tetsuro Hata1 , Tomoharu Tokunaga1; *2 and Zenji Horita1
1
2
Department of Materials Science and Engineering, Graduate School of Engineering, Kyushu University, Fukuoka 819-0395, Japan
JST, CREST, Fukuoka 819-0395, Japan
Excess amount of Mg was intentionally added to fabricate supersaturated ultra-ne grained Al-Mg alloy by high pressure torsion (HPT).
Their microstructures and mechanical properties were examined by X-ray diraction (XRD) analysis, transmission electron microscopy (TEM)
and Vickers microhardness test. It was conrmed that the minimum grain size reached 40 nm and maximum microhardness 292 Hv at the Mg
content with 30 mass%. Further increase of the amount of Mg resulted the formation of -Al3 Mg2 and decrease of microhardness.
(Received July 8, 2008; Accepted November 4, 2008; Published December 25, 2008)
Keywords: aluminum-magnesium alloy, severe plastic deformation (SPD), high pressure torsion (HPT), transmission electron microscopy
(TEM), X-ray diraction (XRD)
1.
Introduction
Severe plastic deformation is a promising technique to

achieve ultrane-grained (UFG) metallic materials with high
strength according to the Hall-Petch relationship.15) One of
the SPD processes, high pressure torsion (HPT) introduces
intense strain to the sample by way of two anvils with
rotation under a high pressure, as illustrated in Fig. 1.13) In
addition, HPT has a great potential to synthesize densely
packed materials from mixed powders without a sintering
process.69) Furthermore, Senkov et al. showed that it was
possible to synthesize the supersaturated solid solution of
Fe in the Al matrix using HPT from cast Al-Fe alloy.10)
Consequently, there are possibilities to prepare supersaturated Al-Mg alloys by mixing powders of Al and Mg.
Addition of Mg to Al is popular in Al industry because Mg
is a typical element to introduce signicant hardening of Al
through a solid solution eect. Although the equilibrium
solid solubility of Mg in Al is about 1 mass% at room
temperature,11,12) an extended solid solubility of Mg in Al can
be achieved by rapid solidication as well as by mechanical
alloying of elemental powders. For example, Luo et al.
reported that the solid solubility of Mg in Al can be
extended far beyond the equilibrium concentrations up to
34:6 mass% by rapid quenching.13) A mechanical alloying
method has also been employed for synthesis of supersaturated alloys including those in amorphous states. Calka et
al. showed that it was possible to extend the solid solubility
of Mg in Al up to 16.5 mass% in the case of Al70 Mg30 and
42.4 mass% for Al50 Mg50 by mechanical alloying.14) Similarly, Zhang et al. and Schoenitz and Dreizin observed the
formation of a solid solution containing 21.2 and 19.1 mass%
of Mg in Al from Al60 Mg40 by mechanically alloying,
respectively.15,16)
In this study, fabrication of supersaturated Al-Mg alloys
*1Corresponding
author, E-mail: kaneko@zaiko.kyushu-u.ac.jp

address: Department of Materials Science and Engineering and
Department of Quantum Engineering, Nagoya University, Nagoya 4648601, Japan
*2Present
Fig. 1
A schematic illustration of HPT.
with dierent contents of Mg is attempted using HPT. The

solubility of Mg and related microstructures are characterized by both X-ray diraction (XRD) analysis and transmission electron microscopy (TEM). Mechanical properties
are also examined by Vickers microhardness test.
2.
Experimental Methods
The materials used in this study were high-purity (99.99%)

Al powders with 75 mm in diameter and high-purity (99.99%)
Mg powders with 150 mm in diameter. The Al powders were
mixed with various amounts of the Mg powders. Approximately 0.3 g of the powder mixture was placed in a shallow
circular hole, 10 mm in diameter and 0.25 mm in depth,
located at the center of the lower anvil of the HPT facility.
The lower anvil was lifted until it made contact with the
upper anvil, which had the same dimensions as the lower
anvil, and then rotated with respect to the upper anvil at a
rotation speed of 1 rpm. This HPT operation was undertaken
at room temperature with an applied pressure of 2.5 GPa and
6.0 GPa for dierent numbers of revolutions as summarized
in Table 1. The rotation was initiated for 5 s after the load
application and the temperature rise during the HPT
operation was monitored using a thermocouple placed
10 mm above the sample. The thermocouple was guided
through a hole made vertically in the center axis of the upper
anvil. The measurement indicated that the temperature
Fabrication and Characterization of Supersaturated Al-Mg Alloys by Severe Plastic Deformation and Their Mechanical Properties
77
Table 1 Experimental condition, solid circles indicate the application of largest strain for the particular Mg content.
Mg content
(mass%Mg)
Applied
pressure
(GPa)
2.5
10
2.5
20
30
40
Number of rotation, N
1
10
15
gradually increased up to 353 K after 100 turns. The nal

sample after the HPT operation was in the shape of a disk
with dimensions of 10 mm in diameter and about 0.7 mm in
thickness. It should be noted that an extra amount of the
mixed powder was deliberately added to prevent the upper
and lower anvils from contacting each other directly so that
the damage and wear to both anvils would be avoided.6,7)
Analysis using XRD was carried out with a Rigaku RINT2100 using the Co K radiation with LaB6 as a reference
material. Regions of 4 mm diameter at the center of the HPT
disks were deliberately cut out to minimize the part of
specimen where less strain was introduced and thus, the
eect of strain was not correctly reected in the XRD
analysis. Then, 3 mm disks were cut from the HPT disks at
positions 3 mm away from the center using an electrical spark
discharge machine for TEM observations. The 3.0 mm disks
were further ground mechanically with abrasive papers to a
thickness of 0.15 mm and thinned in a solution of 20%
perchloric acid (HClO4 ), 10% glycerol (C3 H8 O3 ) and 70%
ethanol (C2 H5 OH) using a twin-jet electropolishing apparatus. Microstructures were observed using a TECNAI-20 (FEI,
Eindhoven, The Netherlands) and H-8100 (Hitachi, Japan)
operated at 200 kV. In particular, selected area electron
diraction (SAED) patterns were taken from areas of 2.5 mm
diameter for HPT-processed specimens. Mean grain sizes
were measured from 100 grains in dark eld TEM images.
The polished disk samples were also examined for Vickers
microhardness and the measurements were undertaken across
the diameters of the HPT disks using an Akashi MVKE3.
Loads of 50 g for 5 and 10 mass%Mg and 200 g for 20, 30 and
40 mass%Mg were applied for 20 s in such measurements.
3.
3.1 X-ray diraction analysis

The XRD patterns of Al-Mg alloys with dierent contents
of Mg are shown in Fig. 2. Except the peaks from LaB6 used
as the reference, only peaks from Al-based fcc crystal
structure are present for the alloys up to 30 mass% of Mg.
However, a trace amount of -Al3 Mg2 was detected from the
sample with 40 mass%Mg in addition to the fcc peaks. This
clearly indicates that the alloying was achieved from the
mixed powders and the supersaturation as well as the new
phase formation occurred by the process of HPT for the alloy
compositions much higher than the equilibrium composition.
Close examination of the XRD patterns reveals that there are
appreciable shifts of the fcc peak positions and this shifting
occurs invariably to lower angles with increasing addition of
20
30
30
Fig. 2
40
40
50
50
60
60
70
80
100
80 90 100
The XRD patterns of Al-Mg alloy with dierent contents of Mg.
Mg. Peak broadening is also apparent with increasing

addition of Mg. The peak shifting suggests that the lattice
parameter was increased due to the supersaturation of Mg in
Al and the peak broadening indicates that large strain was
generated in the HPT-processed alloys. Figure 3 plots the
lattice parameters as a function of equivalent strain for the
dierent compositions. Here, the lattice parameters were
measured using the standard Braggs law17) and the equivalent strain, ", was calculated through the following
equation18)
2rN
" p
3t
where r is the distance from the center of disk, N is the

number of revolutions and t is the thickness of disk. It should
be noted that r 4 was used in this study and this use
provides the lower limit for each disk because the central
region with 4 mm in diameter was cut out from each disk for
the XRD measurement.
78
K. Kaneko, T. Hata, T. Tokunaga and Z. Horita
, mass%
Fig. 3 The relationship between the lattice parameters and the equivalent
strain.
Table 2
Fig. 4 Lattice parameters, a0 , of the fcc-Al(Mg) solid solutions as a

function of Mg content for HPT Al-Mg alloys (solid circles), RQ AlMg
alloys13) (open triangles), for mechanically alloyed AlMg powders (solid
squares)16) and for MM AlMg prealloys (open squares).19)
Lattice parameters measured from X-ray peaks for each Mg content and numbers of rotation.
0 mass%
5 mass%
Mg (N 40)
10 mass%
Mg (N 50)
20 mass%
Mg (N 60)
30 mass%
Mg (N 80)
40 mass%
Mg (N 100)
Lattice
parameter
a0 (nm)
0.405
0.406
0.408
0.412
0.415
0.415
Mean grain
size (nm)
N/A
80
65
45
40
40
The lattice parameters increases with straining and

saturates to the levels determined by the added amount of
Mg. This indicates that Mg is solved in the Al matrix by
straining through HPT. The lattice parameters at the
saturation levels are documented in Table 2 and they are
also plotted in Fig. 4 against the Mg content together with
other results reported from rapidly quenched alloys and
mechanically milled alloys.13,16,19) At the lower content of
Mg, there was a steady expansion of the lattice parameter due
to the increased amount of Mg and the full dissolution of Mg
with straining. Thus, the lattice parameter increases linearly
with the Mg content as shown by the thick dotted line in
Fig. 4. It should be noted that there is no longer increase in
the lattice parameter for the addition of 40% Mg but the
saturation level of the lattice parameter is the same as that for
the 30% Mg. This is most likely to be due to the formation of
-Al3 Mg2 as detected by the XRD analysis. Comparison with
other reports suggests that the lattice parameters obtained in
this study are consistent reasonably well with those obtained
by rapid quenching and mechanical milling. This indicates
that HPT can be used as an alternative process for achieving
supersaturation of Mg in Al.
3.2 Transmission electron microscopy

Figures 5 and 6 show (a) bright-eld (BF)-TEM images,
(b) dark-eld (DF)-TEM images and (c) selected area
electron diraction (SAED) patterns for the powder mixtures
of Al-5 mass% Mg and Al-40 mass% Mg alloys after HPT
processing, respectively. The microstructures consist of
many grains having curved and ill-dened grain boundaries
and these microstructural features are similar to the earlier
observation of an HPT-processed Al-3%Mg alloy prepared
by ingot metallurgy.20) In the present study, TEM observations were also performed on the other compositions and it
was found that the microstructural features are the same as
the ones shown in Figs. 5 and 6 except that the grain size
decreases with an increase in the Mg content. As listed in
Table 3, the average grain sizes are 80, 65, 45, 40 and 40 nm
for the alloys with Mg contents of 5, 10, 20, 30 and
40 mass%, respectively. However, these grain sizes are
smaller by an order of magnitude than those reported earlier
on bulk rods of Al-1 mass%Mg and Al-3 mass%Mg alloys
processed by ECAP (Equal-Channel Angular Pressing).21)
The average grain sizes for the ECAP-processed Al1 mass%Mg and Al-3 mass%Mg alloys were 450 and
Fig. 5 Bright-eld (BF)-TEM image, dark-eld (DF)-TEM image and

selected area electron diraction pattern (SAED) pattern of the powdered
mixture of Al-5 mass% Mg alloy after the HPT deformation.
Table 3
79
Fig. 6 Bright-eld (BF)-TEM image, dark-eld (DF)-TEM image and

selected area electron diraction pattern (SAED) pattern of the powdered
mixture of Al-40 mass% Mg alloy after the HPT deformation.
The measured values of mean grain size, maximum Vickers hardness for each Mg content.
5 mass% Mg
(N 40)
10 mass% Mg
(N 50)
20 mass% Mg
(N 60)
30 mass% Mg
(N 80)
40 mass% Mg
(N 100)
Mg content
(mass%Mg)
2.52
6.89
15.4
22.0
34.1
Maximum
Vickers
microhardness
(Hv)
202
252
288
292
257
270 nm, respectively. Although the grain size tends to be

smaller as the addition of Mg increases, the dierence in the
Mg content is not totally responsible for the dierence in the
grain size by an order of magnitude. HPT processing for high
purity Al powders led to the grain size of 500 nm whereas
the grain size was of 1300 nm when HPT was conducted
with a bulk form of Al disks. The dierence in the grain size
is thus not only the presence of excess Mg but also the
presence of oxide dispersion arising from the surface layers
of powders. Both Mg and oxide dispersion can be good
obstacles for dislocation movement and thus, as discussed
earlier,7) they facilitate the accumulation of dislocations to

make grain boundary formation easier.
According to the XRD patterns, the precipitation of Al3 Mg2 took place when 40 mass% of Mg was added.
However it should be noted that the diracted beams
associated with precipitates were too weak to be detected
in the SAED pattern.
3.3 Vickers hardness test
The maximum Vickers microhardness, 4 mm away from
the center of the sample disk, is plotted in Fig. 7 with the Mg
80
K. Kaneko, T. Hata, T. Tokunaga and Z. Horita
d -1/2/ m-1/2
Fig. 7 The maximum Vickers microhardness, measured at 4 mm away
from the center, of the sample disk.
content. The increase of saturated hardness is obvious for the

higher content of Mg as shown in Fig. 7. The maximum
Vickers microhardness was 40 Hv for pure Al prepared by
HPT from Al powders,6) but it increased gradually, 202 Hv
(at 5 mass%Mg), 252 Hv (at 10 mass%Mg), 288 Hv (at
20 mass%Mg), and 292 Hv (at 30 mass%Mg). However, it
decreased to 257 Hv at 40 mass%Mg, possibly due to the
formation of -Al3 Mg2 .
Combination of solid solution hardening and Hall-Petch
hardening mechanisms are expected to take place as shown in
Fig. 8. The Al-Mg alloy increased its hardness up to 292 Hv
according to the Vickers microhardness test. The microhardness data are slightly better than the Hall-Petch relationship
of pure Al22) for specimens with Mg content up to 30 mass%,
due to the presence of solid-solution hardening in each
sample. Their grain sizes were decreased when the content
of Mg were increased because of the increase of the number
of rotations during the HPT process, namely the amount of
strain.
4.
Conclusions
(1) Mixed powders of varying additions of Mg in Al were

subjected to HPT processing. Supersaturation was attained
for up to the Mg content of 30 mass% but there is a formation
of -Al3 Mg2 for the addition of 40 mass%Mg. The lattice
parameter increases linearly with the Mg content up to 30%
but remains the same for the 40% Mg addition.
(2) The average grain size decreases from 80 nm for the
addition of 5%Mg to 40 nm for the Mg content of 30% or
more. These grain sizes are smaller by an order of magnitude
than those reported on bulk form of Al-1%Mg and Al-3%Mg
alloys.
Fig. 8 The relationship between the Vickers microhardness and the grain
size. The straight line is obtained by Tsuji et al.22)
(3) The maximum Vickers microhardness reached almost 7.5

times higher at 30 mass%, 292 Hv, than pure Al 40 Hv. It is
apparent that the eects of solid solution and grain renement contribute to the hardening of the Al-Mg alloys.
Acknowledgements
This work was supported in part by a Grant-in-Aid for
Scientic Research (No. 19360317) and in part by a Grantin-Aid for Scientic Research from the Ministry of Education, Culture, Sports, Science and Technology, Japan, on
Priority Areas Giant Straining Process for Advanced
Materials Containing Ultra-High Density Lattice Defects,
in part by Global COE program Science for Future
Molecular Systems of the Ministry of Education, Culture,
Sports, Science and Technology, Japan.
REFERENCES
1) R. Valiev: Nature Mater. 31 (2004) 511516.
2) N. A. Smirnova, V. I. Levit, V. I. Pilyugin, R. I. Kuznetsov, L. S.
Davydova and V. A. Sazonova: Phys. Met. Metall. 61 (1986) 127134.
45 (2000) 103189.
4) R. Valiev: Nature 419 (2002) 887889.
5) R. Z. Valiev, I. V. Alexandrov, Y. T. Zhu and T. C. Lowe: J. Mater.
Res. 17 (2002) 58.
6) T. Tokunaga, K. Kaneko, K. Sato and Z. Horita: Scr. Mater. 58 (2008)
735738.
7) T. Tokunaga, K. Kaneko and Z. Horita: Mater. Sci. Eng. A 490 (2008)
300304.
8) I. V. Alexandrov, K. Zhang, A. R. Kilmametov, K. Lu and R. Z. Valiev:
Mater. Sci. Eng. A 234236 (1997) 331334.
9) V. V. Stolyarov, Y. T. Zhu, T. C. Lowe, R. K. Islamgaliev and R. Z.
Valiev: Mater. Sci. Eng. A 282 (2000) 7885.
10) O. N. Senkov, F. H. Froes, V. V. Stolyarov, R. Z. Valiev and J. Liu:
Nano Structured Mater. 10 (1998) 691698.
11) T. B. Massalski: Binary alloy phase diagrams, (ASM International,
Materials Park, OH 1992).
12) N. Saunders: CALPHAD 14 (1990) 6170.
13) H. L. Luo, C. C. Chao and P. Duwez: Trans. Metall. Soc. AIME 230
(1964) 14881489.
14) A. Calka, W. Kaczmarek and J. S. Williams: J. Mater. Sci. 28 (1993)
1518.
15) D. L. Zhang, T. B. Massalski and M. R. Paruchuri: Metall. Mater.
Trans. A 25 (1994) 7379.
16) M. Schoenitz and E. L. Dreizin: J. Mater. Res. 18 (2003) 18271836.
81
17) T. Ungar: Scr. Mater. 51 (2004) 777781.

18) F. Wetscher, A. Vorhauer, R. Stock and R. Pippan: Mater. Sci. Eng. A
387389 (2004) 809816.
19) S. Scudino, S. Sperling, M. Sakaliyska, C. Thomas, M. Feuerbacher,
K. B. Kim, H. Ehrenberg and J. Eckert: Acta Mater. 56 (2008) 1136
1143.
20) M. Furukawa, Z. Horita, M. Nemoto, R. Z. Valiev and T. G. Langdon:
Acta Mater. 44 (1996) 46194629.
21) Y. Iwahashi, Z. Horita, M. Nemoto and T. G. Langdon: Metall. Mater.
Trans. A 29A (1998) 25032510.
893899.

The Evolution of Microstructure and Mechanical Properties of a 5052 Aluminium

Alloy by the Application of Cryogenic Rolling and Warm Rolling
Ui Gu Gang* , Sang Hun Lee* and Won Jong Nam
School of Advanced Material Engineering, Kookmin University, Seoul 136-702, Korea
The microstructural evolution and the corresponding mechanical properties of a 5052 Al alloy processed by cryogenic rolling followed by
warm rolling were investigated. The application of cryogenic rolling combined with warm rolling at 448 K showed a signicant improvement of
tensile strength up to 452 MPa. The kinetics and the microstructural evolution occurred during warm rolling were investigated using dierential
scanning calorimeter and transmission electron microscope. This notable increase of tensile strength was achieved by the formation of ne
precipitates during warm rolling. It was found that the cryogenic rolling combined with warm rolling would be eective in improving
mechanical properties. [doi:10.2320/matertrans.MD200801]
(Received March 31, 2008; Accepted July 10, 2008; Published December 25, 2008)
Keywords: cryogenic rolling, warm rolling, severe plastic deformation, precipitation
1.
Introduction
The recent development of ultra-ne grained (UFG)

aluminium alloys for structural applications has attracted
much attention, because they are expected to provide
extremely high strength at ambient temperature without
degradation of toughness. Earlier works have shown that
severe plastic deformation (SPD) processes, such as equal
channel angular pressing (ECAP),15) accumulative roll
bonding (ARB)68) and severe torsional straining (STS),9,10)
etc. are eective in improving strength of metallic materials
through the grain renement.
Among these SPD processes, an ARB process is known to
have a potential for continuous production of materials in
bulk. In the process, a repeated procedure of cutting, stacking
and roll-bonding of sheets at warm rolling temperatures
introduces high plastic strain without any geometrical
change. Thus, the microstructural evolution occurred during
an ARB process, including work hardening, the subdivision
of grains and dynamic recovery, has yielded ultrane grains
and high strength after several repetition cycles. However, an
ARB process also would not be appropriate for practical
application to produce bulk materials, since they require
extremely large amount of plastic deformation and special
procedures. Meanwhile, the deformation at cryogenic temperature is also recognized as an eective process to produce
UFG materials. The suppression of dynamic recovery during
deformation at extremely low temperatures is expected to
preserve a high density of defects generated by deformation.
Consequently, the larger driving force for recrystallization,
due to the larger density of accumulated dislocations, would
lower recrystallization temperature or accelerate recrystallization during annealing. Thus, the deformation at cryogenic
temperature would require less plastic deformation for
achieving ultrane grains than other SPD processes at
ambient or elevated temperatures.1117) Thus, it is anticipated
that the large stored energy accumulated during the deformation at cryogenic temperature would lead to the larger
driving force for the occurrence of dynamic recovery and,
*Graduate
Student, Kookmin University
then, possibly reduce the required number of cycles in an

ARB process to achieve ultrane grains and high strength.
Accordingly, the combination of cryogenic rolling with
warm rolling, whose temperature range is similar to that of
an ARB process, would become a good candidate for the
formation of ultrane grains with the less amount of the
required deformation.
In view of the aforementioned, the present work was
carried out to investigate the microstructural evolution and
the corresponding mechanical properties of a 5052 Al alloy
processed by the application of cryogenic rolling followed by
warm rolling.
2.
A commercial 5052 Al alloy used for structural applications was chosen for the present work. The chemical
compositions of Al alloy are 2.50 mass% Mg, 0.40 mass%
Fe, 0.25 mass% Si, 0.25 mass% Cr, 0.10 mass% Cu,
0.01 mass% Zn and 0.10 mass% Mn. The material was
annealed at 613 K for 2 hours and then quenched in water
resulting in a grain size of 65 mm. The combination of
cryogenic rolling with warm rolling consisted of two steps.
Cryogenic rolling with 55% reduction was followed by warm
rolling with 56% reduction (total reduction of 80% in
thickness). As the rst step, the plates, 8 mm in thickness,
were rolled with the reduction of 55% at cryogenic temperature. Cryogenic rolling was performed by dipping plates into
liquid nitrogen for at least 15 min before each rolling pass.
The second step was carried out in the temperature range
from cryogenic temperature to 573 K. Before each rolling
pass of warm rolling, the sheets were heated at the temperature range of 448573 K for 5 min to homogenize the
temperature.
For tensile tests, the sheets rolled at various temperatures
were machined into the ASTM subsize specimen of the
25 mm gauge length. Uniaxial tensile tests were conducted
with the initial strain rate of 3 103 /s on an INSTRON
machine operating at a constant crosshead speed. For a
detailed understanding of the microstructural evolution, a
transmission electron microscope (TEM) was used. Thin foils
The Evolution of Microstructure and Mechanical Properties of a 5052 Aluminium Alloy
parallel to the transverse cross section of the sheets were

prepared by utilizing a conventional jet polishing technique
in a mixture of 75% Methanol and 25% HNO3 at the
temperature of 30 C. Thermal analysis was performed
under a owing Ar gas in a NETZSC-DSC 200 F3 (dierential scanning calorimeter, DSC) with the temperature
calibrated using pure In, Zn, Bi and Sn standards. As the DSC
peak position depends on the heating rate,18) the measured
data of the peak positions for dierent heating rates of
1 32 K/min in a owing Ar atmosphere were used. When
obtaining the activation energy, both the Kissinger19) and
Chen and Spaepen20) methods were applied in this work.
3.
3.1 Microstructures and mechanical properties

Figure 1 shows the engineering stress-strain curves of
5052 Al alloy deformed at cryogenic temperature with 55%
reduction and subsequently deformed at various temperatures
with 56% reduction. The stress-strain curves, reecting the
characteristics of microstructures evolved during the deformation, show the dierent response to rolling temperatures
of the second step. The further deformation at cryogenic
500
Engineering Stress, MPa
Cryogenic and Warm rolled at 448K
400
Cryogenic rolled 80%
300
Cryogenic rolled 55%
200
100
0
0
10
Engineering strain, %
Fig. 1 The stress-strain curves of 5052 Al alloys deformed at cryogenic

temperature with 55% reduction and subsequently deformed at various
temperatures with 56% reduction (total reduction of 80% in thickness).
83
temperature led to the increase of tensile strength (TS) by

9.4% (30 MPa), compared with 320 MPa for cryo-rolling
with 55% reduction. This indicates that cross-slip or climb
of dislocations associated with dynamic recovery would be
eectively suppressed during cryo-rolling and therefore the
dislocation density remained high. The microstructure cryorolled with 55% reduction (Fig. 2(a)) consisted of parallel
bands of severely elongated substructures along the rolling
direction, 0:150:25 mm in width, while the narrower width
of elongated substructures, 0:050:15 mm, containing higher
density of dislocations, was observed in the sheet cryo-rolled
with 80% reduction (Fig. 2(b)). The stress-strain curves in
Fig. 1, except for a rolling temperature of 448 K, reects the
general trend that the increase of rolling temperatures in the
second step accelerates softening due to dynamic recovery
rather than work hardening. Meanwhile, it is interesting to
note that TS increases from 320 to 452 MPa (nearly 40%
increase) after warm rolling at 448 K. It is worth mentioning
that the cryogenic temperature rolling with the same
reduction results in TS of 350 MPa (9.4% increase), which
is lower than TS, 452 Mpa, of the sheet warm rolled at 448 K.
To understand this unusual increase of TS, the microstructure
of the sheet deformed at 448 K (Fig. 3(a), (b)) was examined,
compared with the microstructure deformed at cryogenic
temperature (Fig. 2(b)). The microstructure deformed at
448 K shows the similar characteristics to that deformed at
cryogenic temperature, consisting of parallel bands of
elongated substructures, 0:050:15 mm in width and containing a high density of dislocations. The only dierence
was the presence of ne precipitates, about 50100 nm in
diameter (indicated as arrows in Fig. 3(a), (b)). Thus, it can
be expected that the remarkable increase of TS during warm
rolling at 448 K would be closely related with the formation
of ne precipitates. To investigate the eect of warm rolling
on the formation of precipitates, hardness of samples, cryorolled and subsequently annealed for 5 min., is compared
with that cryo-rolled and warm-rolled in Fig. 4. Static
annealing at 448 K increases hardness from 103 Hv (cryorolled) to 111 Hv. This increase of hardness would be
attributed to the formation of precipitates. However, high
hardness of 127 Hv in samples cryo-rolled and warm rolled at
448 K cannot be explained with the contributions of the
precipitation during static annealing and work hardening
Fig. 2 TEM micrographs of a 5052 Al alloys, (a) deformed at cryogenic temperature with 55% reduction and (b) deformed at cryogenic
temperature with 80% reduction.
84
U. G. Gang, S. H. Lee and W. J. Nam
Fig. 3 TEM micrographs of 5052 Al alloys deformed at cryogenic temperature with 55% reduction and subsequently deformed with 56%
reduction at (a) 448 K (arrows indicate precipitates), (b) at 448 K and SAD pattern (arrows indicate precipitates), (c) 523 K and (d) 623 K.
Vickers Hardness, 200g
140
120
15Hv
100
80
60
cryo-rolled and annealed

cryo-rolled and warm-rolled
40
250
300
350
400
450
500
550
600
650
700
Temperature, K
Fig. 4 The variations of hardness with temperature in 5052 Al alloys,

received cryo-rolling and annealing, and cryo-rolling and warm-rolling.
during warm rolling. From the above discussion, it seems

most probable that this notable increase of tensile strength
would be caused by the ner precipitates produced during
warm rolling at 448 K than during static annealing.
Warm rolling at 523 K, leads to the formation of subgrains
and the rearrangement of dislocations including some loss of
dislocations, as recovery proceeds. Many of elongated
substructure boundaries recovered into subgrain boundaries
exhibiting the distinct contrast (Fig. 3(c)). Simultaneously,
the slightly increased subgrain width results in the reduction
of the subgrain aspect ratio. The ultrane subgrains with the
small aspect ratio would inhibit the formation of dislocation

cells during tensile deformation.21) Without the cell formation as well as the dramatic decrease of dislocation density,
the slight increase of dislocation mean free path associated
with the slight increase of the subgrain width does not cause
any signicant change in tensile properties in Fig. 1. That is,
the small decrease of TS from 350 MPa (cryo-rolling) to
318 MPa (warm rolling at 523 K) would be attributed to little
change of the dimension of dislocation substructures during
warm rolling. It is worth mentioning that the warm
deformation at 523 K produces the similarly shaped stressstrain curve to the deformation at cryogenic temperature with
55% reduction. This implies that during warm rolling at
523 K the degree of work hardening due to plastic deformation would be balanced with the degree of softening due to
dynamic recovery. As rolling temperature increases up to
623 K, nearly equiaxed grains, whose diameters range from
100200 nm, as well as elongated subgrains, become the
dominant microstructure. It is expected that the occurrence
of grain subdivision by deformation-induced boundaries
observed by Tsuji, et al.8) during an ARB process, would
contribute to the formation of nearly equiaxed grains during
the warm deformation at 623 K. Thus, the dominant eect of
dynamic recovery caused a drop in tensile strength to
258 MPa. The increased width of subgrains and the low
density of dislocations observed in Fig. 3(d) provide an
evidence of dynamic recovery.
Figure 5 shows the relationship between TS and total
elongation in severely plastic deformed 5 series Al alloys.
Tensile stress of nano or ultra-ne grain structured 5series
alloys is inversely proportional to total elongation. Such a
The Evolution of Microstructure and Mechanical Properties of a 5052 Aluminium Alloy
1.0
-1
2.13 eV
205KJ/mol
1.40 eV
135KJ/mol
1.18 eV
113KJ/mol
0.8
a=b/ Tp
K ln(), 10 eVK
Tensile stress, MPa
1.2
Commercial 5052 Al
Cryo-rolled 5083 Al
Commercial 5083 Al
Cryo- and Warm-rolled 5052 Al*
ECAPed 5083 Al
ARBed 5083 Al
Cryo-rolled 5052 Al*
ARBed 5052 Al
Cryo-rolled & annealed 5052 Al
600
85
400
5
4
4
0.6
a=b/ Tp
2.18 eV
210KJ/mol
0.4
1.44 eV
138KJ/mol
1.22 eV
117KJ/mol
200
0.2
0
10
20
30
1.5
1.6
1.7
1.8
1.9
2.0
2.1
2.2
2.3
Elongation, %
1000/ T, k
Fig. 5 The relationship between TS and tensile elongation for severe
plastic deformed 5 series Al alloys. Number next to symbols denotes the
pressing number. ( means work done in this study)
518.2K
Heat Flow, Arbitrary Unit
486.8K
628.8K
Heating Rate = 16K /min
598.2K
458.5K
489.2K
Heating Rate = 2K /min

400
500
600
700
Temperature, K
Fig. 6 DSC curves of 5052 Al alloys, deformed with 80% reduction at

cryogenic temperature, at the heating rates of 2 and 16 K/min.
trend of strengthening accompanied by a loss of ductility is

general for Al and other metals processed in various ways.
However, the combination of cryogenic rolling with warm
rolling moves from the region of tensile properties representing severely plastic deformed 5052 Al alloys to the region for
severely plastic deformed 5083 Al alloys in Fig. 5 (indicated
by an arrow). This implies that the combination of cryogenic
rolling with warm rolling would become a useful method to
achieve a superior combination of mechanical properties.
Another important thing to note is that this achievement is
obtained only rolling process without the application of any
other severe plastic deformation process.
3.2 Kinetics of the precipitation
To understand the kinetics and the microstructural evolution occurred during warm rolling, thermal properties of
cryo-rolled sheets with 80% reduction were examined using
DSC. The DSC curves in Fig. 6 reveal three distinct
exothermic peaks for the dierent heating rates of (a) 2 K/
min and (b) 16 K/min. These peaks correspond to the heat
evolved during the microstructural evolution, such as the
formation of precipitates, recovery and recrystallization. For
-1
Fig. 7 Calculated activation energies for the precipitation, recovery and

recrystallization of the cryo-deformed 5052 Al alloy with 80% reduction.
the heating rate of 2 K/min, the temperature of the rst broad

peak was found as 459 K and that of the second peak was
489 K and that of the third peak was 598 K. Meanwhile,
for the heating rate of 16 K/min, the temperatures of the
observed peaks were 487, 518 and 629 K, respectively. The
activation energies for the formation of precipitates, recovery
and recrystallization could be determined with the data from
the DSC measurements. The apparent activation energy of
the DSC peaks obtained by analyzing a Kissinger plot19) or a
Chen and Spaepen plot20) is useful for understanding the
mechanisms more precisely. According to the methods, the
heating rate (B), the apparent activation energy of the process
(Q), and the maximum temperature of the peak Tp are
related by the eq. (1);
ln
Q
C
kB T p
where kB is the Boltzmanns constant and C is an integration

constant. Figure 7 shows plots of ln (where B=T 2 p 19)
and B=Tp 20)) against 1=kB Tp and the activation energy
Q calculated from each slope. The two methods show a
reasonable agreement with each other, yielding Q values
of 1:181:22 eV (113117 kJ/mol) for the rst peak,
1:401:44 eV (135138 kJ/mol) for the second peak and
2:132:18 eV (205210 kJ/mol) for the third peak. The
apparent activation energy of the rst peak, 1:181:22 eV,
agrees with the measured activation energy for diusion of
Mg in Al in the dilute limit22) and is close to the calculated
activation energy for Mg diusion by the nearest neighbor
mechanism (1.228 eV).23) This conrms that the mechanism
for the rst peak is the diusion of Mg atoms in the matrix
Al. Thus, it is expected that the unusual increased TS of the
sheet deformed at 448 K would be caused by the formation of
ne precipitates, which is controlled by the diusion of Mg in
the matrix Al. The Q value of the second peak, 135138 kJ/
mol, seems close to the activation energy for lattice self
diusion of aluminum (= 142 kJ/mol24)), implying that the
recovery would be mainly controlled by the self diusion of
Al atom. However, the measured activation energy of the
third peak, 205210 kJ/mol, seems considerably higher than
86
U. G. Gang, S. H. Lee and W. J. Nam
the reported Q values available for Al-Mg alloys. There has

been no clear explanation regarding the mechanism of the
third peak. To clarify the mechanism of the third peak, we
compared the Q values of the third peak with the activation
energy for recrystallization in 1050 Al alloy, 162167 kJ/
mol. And then it is expected that Mg in 5052 Al alloy would
play an important role in increasing the activation energy
required for the occurrence of recrystallization. Although the
authors do not have physical insight at the moment, the
pinning eect of boundary movements by ne precipitates5,25,26) might become a likely candidate for the cause of
the increase in the activation energy for recrystallization in a
5052 Al alloy deformed at cryogenic temperature.
4.
Conclusions
The microstructural evolution and the corresponding

mechanical properties of a 5052 Al alloy processed by
cryogenic rolling followed by warm rolling, were investigated.
(1) The combination of cryogenic rolling with warm rolling
was found more eective than a single cryogenic rolling
process in improving mechanical properties of a 5052 Al
alloy.
(2) Most stress-strain curves showed that the increase of
rolling temperature eectively reduced tensile strength, due
to the contribution of dynamic recovery. However, warm
rolling at 448 K combined with cryogenic rolling was found
to enhance tensile properties. This notable increase of tensile
strength would be attributed to the formation of ne
precipitates during warm rolling at 448 K.
(3) The formation of ne precipitates during warm rolling
would be controlled by the diusion of Mg in the matrix Al.
Acknowledgements
This work was supported by the 2007 research fund of
Kookmin University in Korea.
REFERENCES
1) R. Z. Valiev, N. A. Krasilnikov and N. K. Tsenev: Mater. Sci. Eng. A
137 (1991) 3540.
2) R. Z. Valiev, E. V. Kozlov, Yu. F. Ivanov, J. Lian, A. A. Nazarov and
B. Baudelet: Acta Metall. Mater. 42 (1994) 24672475.
3) M. Furukawa, Z. Horita, M. Nemoto, R. Z. Valiev and T. G. Langdon:
Acta Mater. 44 (1996) 46194629.
4) K. Neishi, Z. Horita and T. G. Langdon: Mater. Sci. Eng. A 325 (2002)
5458.
5) S. Lee, A. Utsunomiya, H. Akamatsu, K. Neishi, M. Furukawa, Z.
Horita and T. G. Langdon: Acta Mater. 50 (2002) 553564.
6) Y. Saito, N. Tsuji, H. Utsunomiya, T. Sakai and R. G. Hong: Scr.
Mater. 39 (1998) 12211227.
7) N. Tsuji, Y. Saito, H. Utsunomiya and S. Tanigawa: Scr. Mater. 40
(1999) 795800.
8) N. Tsuji, Y. Ito, Y. Saito and Y. Minamino: Scr. Mater. 47 (2002) 893
899.
9) R. Z. Abdulov, R. Z. Valiev and N. A. Krasilnikov: J. Mater. Sci. Lett. 9
(1990) 14451447.
Mater. 44 (1996) 47054712.
11) Y. Wang, M. Chen, F. Zhou and E. Ma: Nature 419 (2002) 912915.
12) Y. B. Lee, D. H. Shin, K. T. Park and W. J. Nam: Scr. Mater. 51 (2004)
355359.
13) F. Zhou, X. Z. Liao, Y. T. Zhu, S. Dallek and E. J. Lavernia: Acta
Mater. 51 (2003) 27772791.
14) Nikhil Rangaraju, T. Raghuram, B. Vamsi Krishna, K. Prasad Rao and
P. Venugopal: Mater. Sci. Eng. A 398 (2005) 246251.
15) Y. M. Wang and E. Ma: Appl. Phys. Lett. 85 (2004) 27502752.
16) S. Cheng, Y. H. Zhao, Y. T. Zhu and E. Ma: Acta Mater. 55 (2007)
58225832.
17) U. G. Gang, Y. S. Lee, K. T. Park and W. J. Nam: Solid State
Phenomena 124126 (2007) 13931396.
18) F. Zhou, X. Z. Liao, Y. T. Zhu, S. Dallek and E. J. Lavernia: Acta
Mater. 51 (2003) 27772791.
19) H. E. Kissinger: Anal. Chem. 29 (1957) 17021706.
20) L. C. Chen and F. Spaepen: J. Appl. Phys. 69 (1991) 679.
21) K. T. Park and D. H. Shin: Metall. Mater. Trans. A 33 (2002) 705707.
22) S. I. Fujikawa and Y. Takada: Defect Di. Forum 143 (1997) 409.
23) R. C. Picu and D. Zhang: Acta Mater. 52 (2004) 161171.
24) H. J. Frost and M. F. Ashby: Deformation-Mechanism Maps,
(Pergamon, Oxford, UK, 1982).
25) Y. Iwahashi, Z. Horita, M. Nemoto and T. G. Langdon: Metall. Mater.
Trans. A 29 (1998) 25032510.
26) R. Z. Valiev: Nano Struct. Mater. 6 (1995) 7382.

Grain Renement and Mechanical Behavior of the Al Alloy,

Subjected to the New SPD Technique
Ruslan Z. Valiev* , Maxim Yu. Murashkin, Elena V. Bobruk and Georgy I. Raab
Institute of Physics of Advanced Materials, Ufa State Aviation Technical University, 12 K.Marx str., Ufa 450000, Russia
The paper focuses on producing of ultrane-grained (UFG) structure in the Al 6061 alloy by a new severe plastic deformation (SPD)
technique, developed recently in our laboratory, namely equal channel angular pressing with parallel channels (ECAP-PC). The evolution of
microstructure at ECAP-PC was examined and was proved that the alloy becomes of a homogenous UFG structure after 4 passes. Such a
structure increases essentially the alloys mechanical properties, specically strength and ductility. The advantages of this new technique in
producing of UFG alloys over conventional ECAP are considered and discussed as well. [doi:10.2320/matertrans.MD200821]
(Received August 18, 2008; Accepted October 27, 2008; Published December 25, 2008)
Keywords: ultrane-grained structure, severe plastic deformation, equal channel angular pressing with parallel channels, strength and
ductility
1.
Introduction
Fabrication of ultrane-grained (UFG) metals and alloys

by severe plastic deformation (SPD) processing has become a
well-known direction in modern materials science.13) These
works started in the early 90-s with two techniques: equal
channel angular pressing (ECAP) and high pressure torsion
(HPT),4) have been developed intensively last decade in
connection with creation of principles of materials nanostructuring for advanced properties and elaboration of more
eective SPD techniques.
Conventional ECAP technique, widely used to produce
UFG structure in metals and alloys is currently applied
mainly as laboratory-based SPD method as it has some
limitations, which restrict its industrial application. These
restrictions are due to labor-intensity and low eciency of
the method, because a large number (812) of passes is
required for a UFG structure formation and only small
amount (less than 60%) of the material is subjected to
homogenous shear straining in the bullet.5) The new SPD
technique-equal channel angular pressing with parallel
channels (ECAP-PC),6,7) newly developed in our laboratory
is able to reduce a number of restrictions for the conventional
ECAP. Computer simulation and experiments applied to Cu
and Ti as model materials have shown, that UFG structure
is produced more eectively by ECAP-PC, due to more
advanced strain and homogeneity rates.5,7)
The given paper reports the rst successful application of
the new SPD technique to commercial Al alloy with the aim
of producing an UFG structure and improving its mechanical
properties. The experiment was conducted on 6061 Al alloy.
This alloy is one of the commercial aluminum alloys widely
used in aerospace and automobile industries. Advantages
of the new method versus conventional ECAP are being
discussed as well.
2.
A commercial 6061 aluminum alloy (0.80Mg-0.41Si0.12Cu-0.03Mn-0.02Ti-0.22Fe-balance Al (mass%)) was

*Corresponding
author, E-mail: RZValiev@mail.rb.ru
used in a cast condition. The alloy under investigation was

characterized by lower level of alloying elements (Si and
Mg), comparing to traditional Al 6061 alloy.8) Before
ECAP-PC treatment, initial billets were subjected to solid
solution treatment for 5 h at 530 C with the following water
quenching.
The principles of ECAP with PC are shown in Fig. 1.5,7) A
distinctive feature of the ECAP-PC is that during a single
processing pass, two distinct shearing events take place:
shear in two deformation zones subsequently corresponding
to two subsequent channels intersections in the die-set
(Fig. 1(a), (b)).
The value of the displacement between the two channels,
K, and the angle of intersection of the two channels , are the
main parameters of the die geometry, which inuence both
the ow pattern and the strain-stress state of the ECAP
process. It has been established by computer simulation, that
the optimal value of these parameters leading to the largest
strain homogeneity are 100 and K d, where d is the
channel diameter.5,7) Under these conditions, the accumulate
strain for one pass is approximately equal to 2. The
simulation results have been conrmed experimentally using
a grid method. Furthermore, the studies showed that the
deformation pattern realized in ECAP-PC die-set optimal
geometry makes the structure homogeneous along the length
of the bulk sample including up to the ends (the material
utilization ratio is higher than 90%).
Billets of 6061 Al alloy with the diameter of 18 mm and
100 mm length were subjected to ECAP-PC at 100 C, since
the temperature provides signicant grain renement, preserving the facility for its further strengthening by subsequent
aging at 160175 C.
Vickers hardness (HV) was measured using a Micromet5101 microidentation tester with a load of 200 g for 15 s. In
order to receive reliable results, each sample was measured
more than 10 times.
Tensile tests at room temperature were performed on
Instron 1185 at a strain rate of 5 104 s1 . Mechanical
properties were measured by at least 2 specimens with gage
length 15 mm and diameter of 3 mm. Tensile specimens were
cut of parallel to the longitudinal axis of the ECAP-processed
material.
88
R. Z. Valiev, M. Yu. Murashkin, E. V. Bobruk and G. I. Raab
Fig. 1 The principles of ECAP with parallel channels: (a) a view of the deformation zone obtained by two dimensional nite element
method (2-D FEM). (b) a schematic illustration where N is in the shear direction; d-diameter of the channel; K- is the displacement
between the two channels; - is the angle of intersection between the two parts of the channel and the internal shaded areas depict the
sharing as the sample traverses the shearing zone; (c) a general view of the experimental ECAP die-set.
Fig. 2 Microstructure of Al 6061 alloy after ECAP-PC through one pass: dark eld (a), SAED pattern (b).
The structural characterization by TEM was performed in

JEM-100B and JEOL-2000 EX electron microscopes using
dark and bright elds. Average grain size was estimated from
more than 250 grains measurement.
3.
Observations by TEM showed that mainly subgrain

structure of a lamellar type with clear orientation to shear
plane was formed after ECAP-PC thorough one pass at
100 C (Fig. 2(a)). The mean cross subgrain size comprised
400 nm and longitudinal one made up 1200 nm correspondingly. Electron diraction patterns prove subgrain structure
formation with mainly low-angle grain boundary misorientaions (Fig. 2(b)). After ECAP-PC through the second pass at
the same temperature the structure changed signicantly.
(Fig. 3). Fragmentation by means of cross-subboundary
formation can be seen in elongated subgrains formed after
ECAP-PC through 1 pass (Fig. 3(a), (b)). Separate areas
(with a volume fraction of around 30%) of equi-axed
grains of 500 nm are also observed in the structure along with

subgrain fragmentation (Fig. 3(c), (d)).
The electron diraction pattern (Fig. 3(f)) represents
numerous spots distributed uniformly over the circles. This
testies to formation of the UFG structure at the areas of a
granular type with high-angle grain boundaries at the areas.
The UFG structure formed after ECAP-PC through 4
passes is also of equiaxed (Fig. 4). Besides, secondary phase
dispersive precipitates of less than 10 nm in size were found
in the alloy UFG structure after ECAP-PC (Fig. 4, A and B
areas). According to the results9,10) dealing with 6061 Al
alloy processing by conventional ECAP, these dispersive
precipitates represent Mg2 Si particles. These precipitates
testify to the UFG structure formation being accompanied by
dynamic aging at ECAP-PC through 4 passes. As the
precipitates were not found after the ECAP-PC through 1
and 2 passes, we can suggest that they are absent in these
states, or their size was quite small and required investigations by high-resolution electron microscopy. It should be
also noted that there were more lattice dislocations in the
Grain Renement and Mechanical Behavior of the Al Alloy, Subjected to the New SPD Technique
89
Fig. 3 Microstructure of the alloy after ECAP-PC through 2 passes: subgrain fragmentation (a), (b), grain structure (d), (e)-bright and dark
elds) SAED patterns (c), (f).
structure after 1 and/or 2 passes (Fig. 2 and 3) when

compared to the structure formed in the material after 4
passes (Fig. 4).
The Vickers hardness (HV) of the initial billets after solid
solution treatment was 410 18 MPa. ECAP-PC processing
increases the rates up to 1155 24 MPa and to 1220
27 MPa after 1 and 2 passes correspondingly. However, the
hardness values decreased to 1100 21 MPa after ECAP-PC
through 4 passes. Mechanical properties of 6061 alloy after
ECAP-PC through 1, 2 and 4 passes are presented in Fig. 5
and Table 1. The data can be compared with that of the same
alloy after conventional T6 treatment (quenching at 535 C
and subsequent articial aging at 160 C for 12 hours).
The results correlate closely to the evolution of the
microstructure revealed (Fig. 24). The formed substructure
(Fig. 2) with a high density of lattice dislocations provides
signicant increase in strength after ECAP-PC through 1
pass, but its ductility is low enough and elongation to failure
comprised 6:5%. The formation of the UFG structure of a
combined type containing grains and subgrains (Fig. 3) in
the billets after ECAP-PC through 2 passes provides some
increase in both strength and ductility up to 8% (Table 1).
At the same time the homogeneous UFG structure (Fig. 4)

formed after ECAP-PC through 4 passes leads to insignicant
loss in strength. However, its ductility increases by two times
and its elongation reaches 20%. Herewith, the UFG alloy
demonstrates considerable uniform elongation which makes
up 9%. Such a high value, which comprises 40{45%
from total elongation, shows that the alloy with UFG
structure has a potential to display enhanced fatigue.5,11)
Thus, the analysis of structural changes in the alloy 6061
after ECAP-PC at 100 C proves that the ECAP-PC technique
makes it possible to produce a homogeneous UFG structure
in a cast alloy after 4 passes in comparison with 810 passes
in conventional ECAP,5,9) in other words, using of ECAP-PC
can decrease the number of passes in 23 times under nearly
the same processing conditions. Another advantage of the
pressing is signicant increase in the volume of UFG
structure in a billet processed: about 8590% compared to
5060% after conventional ECAP. This can be of special
interest for the process industrial application.
Regarding UFG structure formation in the alloy, an
important feature of ECAP-PC is producing of practically
equiaxed UFG structure already after 2 passes (Fig. 3),
90
R. Z. Valiev, M. Yu. Murashkin, E. V. Bobruk and G. I. Raab
A
100 nm
100 nm
Fig. 4 Microstructure of the alloy after ECAP-PC through 4 passes in the longitudinal section of the sample (a), (b), in the cross-section of
the sample (disperse particles of Mg2 Si phase precipitated during the processing-areas A and B) (d), SAED pattern (c).
Table 1 Mechanical properties of the 6061 Al alloy at room temperature

after ECAP-PC at 100 C.
UTS,
(MPa)
YS,
(MPa)
El.,
(%)
ECAP-PC at 100 C, 1 pass
370
335
6.5
ECAP-PC at 100 C, 2 passes
380
350
8.0
ECAP-PC at 100 C, 4 passes
345
305
20.0
T6
230
178
30.0
Engineering Stress (MPa)
500
Treatment
1 pass
2 passes
400
4 passes
300
T6
200
100
after ECAP in PC at 100C
0
0
10
15
20
25
30
35
40
Engineering Strain (%)

Fig. 5 The engineering stress-strain curves of the 6061 Al alloy at room
temperature after ECAP-PC processing and convention heat-treatment
(T6).
though the grains observed after the rst pass were of

elongated laminar structure (Fig. 2(a)). Gradual transformation of the grains to equiaxed ones is typical also of
conventional ECAP.5) At the same time, the present work
proves the connection of this transformation with the
formation of low-angle subgrain boundaries inside the
elongated grains (Fig. 3(a)) and their following arrangement
to high angle boundaries of the equiaxed UFG structure.
Similar evolution of nanostructure during SPD processing
has been discussed elsewhere.12)
Grain Renement and Mechanical Behavior of the Al Alloy, Subjected to the New SPD Technique
One more eect revealed in the alloy after 4 passes is the

formation of highly-disperse precipitations of the second
phase, obviously, the particle of Mg2 Si phase. The formation
of such particles evidences dynamic aging, which was
observed recently in 6061 alloy after conventional ECAP.9,10)
However, this time the aging occurs at a low temperature
(100 C) and aging time less than 10 minutes, which is
considerably lower, than 175 C, 8 h of the peak aging
condition (T6) for a 6061 alloy.9) Such an acceleration of
aging kinetics can result from high vacancy concentration
during processing. Recent in situ HPT experiments in
synchrotron beamline showed13) that excess vacancy concentration can reach 105 , which is near to equilibrium
concentration at melting temperature point. Besides, 3D atom
probe applied to the alloy showed the formation of alloying
elements segregations after HPT even at room temperature.14) Such segregations appear to be the second phase
precipitations at the subsequent deformation. Thus, aging
processes are more rapid at ECAP-PC, which can inuence
the alloys mechanical behavior.
It is well-known, that ECAP processing signicantly
increases the strength, but reduces ductility. The Al 6061
alloy after ECAP-PC through 1 and 2 passes behaved in
accordance with this rule. However, it showed considerable
increase in ductility with the small reduction of strength (to
3040 MPa) after ECAP-PC through 4 passes. Such extraordinary combination of high strength and high ductility of
UFG Al alloy was also observed recently.5,1517) This
attractive mechanical behavior can be caused by either
formation of UFG structure with high-angle grain boundaries
capable to grain boundary sliding,16) or diculty in strain
localization due to the presence of disperse particles.17) We
are planning to study the phenomenon in detail in our
ongoing research.
Conclusions:
(1) Using ECAP-PC we could successfully produce UFG
structure in Al 6061 alloy. This new technique has a
number of advantages over conventional ECAP for
grain renement. These advantages attribute to gain in
the processing performance and increase of the UFG
material amount in the rod.
(2) The UFG structure formation during the processing of
the alloy is associated with transformation of low-angle
boundaries to high-angle grain boundaries. The formation is also accompanied by acceleration of dynamic
aging processes.
91
(3) The 6061 Al alloy after ECAP-PC through 4 passes

showed combination of high strength and ductility,
which is of special interest for its advanced application.
Acknowledgements
The present paper was supported in part through grants of
the Ministry for Education and Science of the Russian
Federation and Russian Foundation for Basic Research as
well as by the NIS-IPP Program of DOE (USA). Cooperation
with co-authors mentioned in references is gratefully acknowledged as well.
REFERENCES
45 (2000) 103189.
2) M. J. Zehetbauer (Ed.): Adv. Eng. Mater. 5 (2003) 262378.
3) Z. Horita (Ed.): Proc. 3rd Int. Conf. on Nanomaterials by Severe Plastic
Deformation, (Trans. Tech. Publications LTD, Switzerland, 2006)
p. 1030.
4) R. Z. Valiev, Yu. Estrin, Z. Horita, T. G. Langdon, M. J. Zehetbauer
and Y. T. Zhu: JOM 58 (2006) 3339.
5) R. Z. Valiev and T. G. Langdon: Prog. Mater. Sci. 51 (2006) 881
981.
6) G. I. Raab, R. Z. Valiev, G. V. Kulyasov and V. A. Polozovsky:
Russian Patent No. 2181314 (2002).
7) G. I. Raab: Mater. Sci. Eng. A 410411 (2005) 230.
8) J. R. Davis (ed): ASM Special Handbook, (Aluminium and Aluminium
Alloys, 1993) P. 784.
9) W. J. Kim, J. K. Kim, T. Y. Park, S. I. Hong, D. I. Kim, Y. S. Kim and
J. D. Lee: Metall. Mater. Trans. 33A (2002) 31553164.
10) G. Nurislamova, X. Sauvage, M. Murashkin and R. Valiev: Ultrane
Grained Materials IV. ed. by Y. T. Zhu, T. G. Langdon, Z. Horita et al.
TMS 2006 Annual Meeting in San Antonio, Texas, USA March 1216,
(2006) pp. 4145.
11) A. Yu Vinogradov, V. V. Stolyarov, S. Hashimoto and R. Z. Valiev:
Mater. Sci. Eng. A 318 (2001) 163.
Mater. 44 (1996) 47054712.
13) A. R. Kilmametov, G. Vaughan, A. R. Yavari, A. LeMoulec, W. J.
Botta and R. Z. Valiev: Mater. Sci. Eng. (in press).
14) G. Nurislamova, X. Sauvage, M. Murashkin, R. Islamgaliev and R.
Valiev: Phil. Mag. Lett. 88 (2008) 459466.
15) H. J. Roven, H. Nesboe, J. C. Werenskiold and T. Seibert: Mater. Sci.
Eng. A 410411 (2005) 426429.
16) I. Sabirov, Y. Estrin, M. R. Barnett, I. Timokhina and P. D. Hodson:
Acta Mater. 56 (2008) 22232230.
17) Z. Horita, K. Ohashi, T. Fujita, K. Kaneko and T. G. Langdon: Adv.
Mater. 17 (2005) 15991602.

RAPID PUBLICATION
Scaling-Up of High Pressure Torsion Using Ring Shape

Kaveh Edalati* and Zenji Horita
A 100 mm-diameter ring sample of high purity Al (99.99%) was processed by high-pressure torsion (HPT). It was conrmed that the
results of hardness measurements were consistent with those obtained using disk samples and demonstrated that scaling-up the HPT process was
feasible using the ring-shaped sample. Evaluating the advantage of using ring sample, the potential for the scaling-up arises from the fact that the
ring diameter can be increased by the amount corresponding to the inner central area of disk sample. Furthermore, the applied load is entirely
used for the introduction of intense strain and thus for the development of a homogeneous microstrcuture throughout the ring sample.
Keywords: high-pressure torsion, severe plastic deformation, pure aluminum
1.
Introduction
It has been demonstrated that processing by high-pressure

torsion (HPT) provides important technology for materials
development. The grain sizes of various metallic materials is
rened to the submicrometer or nanometer ranges and
consequently the HPT processing leads to the enhancement
of strength and ductility.13) Second phase particles may be
rened signicantly and occasionally they are dissolved into
the matrix.4,5) It is also used as a consolidation of powders
without sintering process.610) The HPT is applicable to a
wide range of materials including relatively hard and low
ductile materials such as intermetallics and Mg alloys.1114)
Major limitation of the HPT process is that the sample size
is small typically as 10 mm in diameter with thicknesses less
than 1 mm, and the microstructure develops inhomogeneously from the center to edge of the disk because the strain is
generated in proportion to the distance from the center.15,16)
To overcome this limitation, it was shown that a ring sample
may be useful1719) because it uses only the outer areas of the
disk where homogeneous structures are developed throughout the periphery.1921) Furthermore, it is possible to increase
the diameter by the area corresponding to the inner hollow
area.19) Provided that the rotation angle is the same, the
increase in the ring diameter leads to introduction of more
strain, or if the designated strain is required, less rotation is
sucient.
Now, in this report, we demonstrate that scaling-up of the
HPT process is feasible by taking advantage of the ringshaped feature. Furthermore we show that a 100 mm ring
with 3 mm width is successfully processed. Advantage of
using ring sample is discussed in comparison with disk
sample.
2.
Experimental
High-purity Al (99.99%) was received in a form of coldrolled sheets with 1 mm thickness and an area of 220 mm
150 mm. The sheet was then cut to rings with dimensions
*Corresponding
ac.jp
author, E-mail: kaveh.edalati@zaiko6.zaiko.kyushu-u.
Fig. 1 Schematic illustration of ring sample and anvils for HPT.
of 94 mm and 100 mm for the inner and outer diameters,

respectively, having a width of 3 mm. Disks with 10 mm and
20 mm diameters with thicknesses of 0.8 mm were also
prepared from cold-extruded rods for comparison. The rings
and disks were subjected to annealing at 773 K for 1 h before
HPT. The grain sizes after the annealing were 700 mm for
the disks and 100 mm for the ring. According to a recent
study in the same 99.99% purity Al,22) the dierence in grain
size has little eect on the hardness behavior with intense
strain if the grain size is greater than 100 mm.
HPT was conducted at room temperature at a rotation
speed of 1 rpm for 1/4 revolutions under a pressure of
0.3 GPa. This study used a pair of upper and lower anvils with
grooves made axisymetric in 0.25 mm depth and 3 mm width.
The alignment was adjusted about the center axes of the
upper and lower anvils within 0:01 mm. Schematic illustration of the upper and lower anvils with a ring sample is
given in Fig. 1. For a comparison purpose, disks with 10 mm
and 20 mm diameters annealed at the same conditions were
placed in shallow holes made on the anvils with diameters of
10 and 20 mm, respectively, and a depth of 0.25 mm. Such
disk samples were subjected to HPT at room temperature for
2 and 1/4 revolutions, respectively under a pressure of 1 GPa
93
Fig. 2 Appearance of 100 mm ring sample and comparison with 10 mm and 20 mm disk samples.
with the rotation speed of 1 rpm. The appearance of the disks

and ring samples after HPT is shown in Fig. 2.
The disks and ring samples after HPT were polished to a
mirror-like surface and thus the Vickers microhardness was
measured along the radii from the center to edge at 8 dierent
radial directions for the disk samples and along the periphery
for the ring samples. The average was taken from the 8
measurements at the same distance from the center of disks or
rings. For each hardness measurement, a load of 50 g was
applied for 15 seconds.
3.
The results of hardness measurement are shown in Fig. 3

for disk and ring samples after HPT including the hardness
level after annealing but before HPT. All hardness values of
10 mm and 20 mm disk samples fall on a single curve when
they are plotted as a function of equivalent strain. Here, the
equivalent strain was calculated by the equation given as.16)
2rN
" p
3t
where r is the distance from the center, N is the number of

revolution and t is the thickness. A maximum appears at an
equivalent strain of 2 and a constant level of hardness
follows at an equivalent strain of more than 7. These
features are consistent with the earlier measurements
performed on high purity Al (99.99%).20,21) The average
hardness measured on the 100 mm ring lies well on the steady
state level established by the 10 mm and 20 mm disks. It is
thus conrmed that processing of the 100 mm ring by HPT is
feasible and it is demonstrated that the use of ring samples
can be a practical way for scaling up the HPT process.
Although the applied pressure is less in the ring sample than
in the disk samples for the present experiments, the steady
state level is little aected by the applied pressure. This is
also consistent with the reports that the eect of the applied
pressure is insignicant on the hardness level when the
imposed strain is large.23,24)
Fig. 3 Plots of Vickers microhardness against equivalent strain obtained

from 10 mm and 20 mm disks including results of 100 mm ring sample.
A major advantage of using the ring sample for the HPT

process is that it is possible to increase the diameter by the
area corresponding to the inner hollow area. Furthermore, the
use of ring samples eliminates microstructural inhomogeneity developed in the inner central areas of disk samples as
reported earlier.1921) Now, we estimate quantitatively how
the ring diameter can be enlarged with respect to the disk
diameter, provided that the same pressure is applied on both
ring and disk samples.
As illustrated in Fig. 4, the disk diameter is dened as DD
and the inner and outer diameters of ring as Di and Do ,
respectively, with the ring width as w. Under the condition of
equal area for both disk and ring (SD SR ), it follows that Do
can be given as
Do w DD 2 =4w
Figure 5 plots the relation of Do against DD for several

values of w. This gure is read such that the 20 mm diameter
for disk will be enlarged to 100 mm diameter for ring if the
ring width is 1 mm. Alternatively, it may be read such that a
100 mm ring with 3 mm width as in this study is equivalent
94
K. Edalati and Z. Horita
temperature increase was negligible in this study because of

only 1/4 revolutions, there is in fact a report that the hardness
decreased due to temperature increase at a faster rotation
speed.25)
It has been said that a major limitation of the HPT process
to practical application is the sample size which is small.
However, the present study shows that this can be overcome
using ring samples.
4.
Fig. 4 Denitions of disk diameter and ring inner and outer diameters
including ring width.
Summary and Conclusions
(1) It was demonstrated that scaling up of HPT is feasible

using ring sample. It was shown that the 100 mm ring
with 3 mm width was successfully processed by the
HPT facility. The hardness results were well consistent
with those obtained from the conventional disk sample.
(2) An advantage of using ring sample was discussed
quantitatively. First, the ring diameter is increased by
the amount corresponding to the inner hollow area.
Second, the whole area of ring sample can be usable
because of homogeneous introduction of strain and
thus of homogeneous development of microstructure
throughout the ring.
Acknowledgements
One of the authors (KE) thanks the Islamic Development
Bank for a scholarship. This work was supported in part by
the Light Metals Educational Foundation of Japan, in part by
a Grant-in-Aid for Scientic Research from the Ministry
of Education, Culture, Sports, Science and Technology of
Japan in the Priority Area Giant Straining Process for
Advanced Materials Containing Ultra-High Density Lattice
Defects and in part by Kyushu University Interdisciplinary
Programs in Education and Projects in Research Development (P&P).
REFERENCES
Fig. 5 Relation of ring outer diameter with disk diameter for various ring
widths.
to the use of 35 mm disk. It should be emphasized that there is

a signicant dierence between the use of ring samples and
that of disk samples. Intense strain is equally introduced
throughout the ring and a homogeneous microstructure is
developed for ring samples. However, this is not the case for
the disk sample because the strain introduced at the inner area
is small despite the fact that the applied pressure is the same
for both ring and disk samples. Another advantage of using
ring sample is that the same strain is introduced by less
rotation when compared with the disk sample. This is
because the diameter is enlarged for the ring sample. Note
that the equivalent strain is given by the product between the
distance from the center (r) and the rotation angle (2N) as
evident from eq. (1). It should be noted that, when the
diameter becomes larger, the deformation occurs at faster
rate even at the same rotation speed and consequently this
may introduce heat generation in the sample. Although the

45 (2000) 103189.
2) R. Z. Valiev, I. V. Alexandrov, Y. T. Zhu and T. C. Lowe: J. Mater.
Res. 17 (2002) 58.
3) A. P. Zhilyaev and T. G. Langdon: Prog. Mater. Sci. 53 (2008) 893
979.
4) M. V. Degtyarev, T. I. Chashchukhina and L. M. Voronova: Fiz. Met.
Metalloved. 77 (1994) 141146.
5) R. K. Islamgaliev, N. F. Yunusova, I. N. Sabirov, A. V. Sergueeva and
R. Z. Valiev: Mater. Sci. Eng. A 319321 (2001) 877881.
6) I. V. Alexandrov, Y. T. Zhu, T. C. Lowe, R. K. Islamgaliev and R. Z.
Valiev: Nanostruct. Mater. 10 (1998) 4554.
7) V. V. Stolyarov, Y. T. Zhu, T. C. Lowe, R. K. Islamgaliev and R. Z.
Valiev: Mater. Sci. Eng. A 282 (2000) 7885.
8) A. R. Yavari, W. J. Botta, C. A. D. Rodrigues, C. Cardoso and R. Z.
Valiev: Scr. Mater. 46 (2002) 711716.
9) T. Tokunaga, K. Kaneko, K. Sato and Z. Horita: Scr. Mater. 58 (2008)
735738.
10) T. Tokunaga, K. Kaneko and Z. Horita: Mater. Sci. Eng. A 490 (2008)
300304.
11) A. V. Sergueeva, C. Song, R. Z. Valiev and A. K. Mukherejee: Mater.
Sci. Eng. A 339 (2003) 159165.
12) C. Rentenberger, T. Waitz and H. P. Karnthaler: Mater. Sci. Eng. A 462
(2007) 283288.
13) J. Cizek, I. Prochazka, B. Smola, I. Stulkova, R. Kuzel, Z. Matej, V.
14)
15)
16)
17)
18)
19)
20)
Cherkaska, R. K. Islamgaliev and O. Kulyasova: Mater. Sci. Eng. A

462 (2007) 121126.
Y. Harai, M. Kai, K. Kaneko, Z. Horita and T. G. Langdon: Mater.
Trans. 49 (2008) 7683.
A. Vorhauer and R. Pippan: Scr. Mater. 51 (2004) 921925.
F. Wetscher, A. Vorhauer, R. Stock and R. Pippan: Mater. Sci. Eng. A
387389 (2004) 809816.
S. Erbel: Met. Technol. 6 (1979) 482486.
I. Saunders and J. Nutting: Metal. Sci. 18 (1984) 571575.
Y. Harai, Y. Ito and Z. Horita: Scr. Mater. 58 (2008) 469472.
Y. Ito, Y. Harai, T. Fujioka, K. Edalati and Z. Horita: Mater. Sci. Forum
95
584586 (2008) 191196.

21) K. Edalati, T. Fujioka and Z. Horita: Mat. Sci. Eng. A 498 (2008) 168
173.
22) T. Inoue, Z. Horita, H. Somekawa and K. Ogawa: Acta Mater. (2008)
in press.
24) K. Edalati, T. Fujioka and Z. Horita: Proc. 156th meeting of ISIJ and
JIM: Joint Session on Current Advances in Materials and Processes,
Kumamoto, Japan, 21(2) (2008) p. 516.
25) Y. Todaka, M. Umemoto, A. Yamazaki, J. Sasaki and K. Tsuchiya:
Mater. Trans. 49 (2008) 714.

Microstructure Evolution in Pure Al Processed with Twist Extrusion

Dmitry Orlov1 , Yan Beygelzimer2 , Sergey Synkov2 , Viktor Varyukhin2 ,
Nobuhiro Tsuji1 and Zenji Horita3
1
Department of Adaptive Machine Systems, Graduate School of Engineering, Osaka University, Suita 565-0871, Japan
Department of Physics of High Pressures and Advanced Technologies, Donetsk Institute for Physics & Engineering of
the National Academy of SciencesUkraine, 72 R. Luxembourg St., Donetsk, 83114, Ukraine
3
2
High purity Al (99.99%) was subjected to severe plastic deformation through twist extrusion at room temperature. Microstructures were
examined for 1 pass and 4 passes on the cross section perpendicular to the longitudinal axis of billets using optical microscopy and electron back
scatter diraction analysis. It was shown that a vortex-like material ow was observed on the cross section and this became more intense with
increasing number of the pressing. After one pass, subgrain structures with low angle grain boundaries were developed throughout the section
but after 4 passes, the microstructure consisted of grains surrounded by high angle boundaries with fraction of 70% in the edge parts. The
average grain size at the edge parts is rened to 1:6 mm. [doi:10.2320/matertrans.MD200802]
(Received April 18, 2008; Accepted July 28, 2008; Published September 10, 2008)
Keywords: aluminum, twist extrusion, severe plastic deformation, optical microscopy, electron backscattered diraction (EBSD)
1.
Introduction
Recently, scientists and engineers are very much attracted

to nano- and ultrane-grained (UFG) materials because of
their unique structures and properties.13) Special attention is
paid to the process of severe plastic deformation (SPD) since
it allows to produce UFG materials in bulk forms.2) Many
SPD processes are available now, such as high pressure
torsion (HPT),4) equal-channel angular pressing (ECAP),5,6)
multiple forging,7) repetitive corrugation and straightening,8)
accumulative-roll bonding (ARB),9,10) and twist extrusion
(TE).11,12)
The TE process was rst introduced by Beygelzimer
et al.1114) and related works have been published thereafter.1517) In TE, a billet is pressed through an extrusion die
whose cross section remains unchanged, while the billet is
twisted with an angle about its longitudinal axis as
illustrated in Fig. 1. The billet maintains its size and shape of
the cross section after each TE pass. Thus it is possible to
repeat the pressing to accumulate a large strain and to attain
grain renement as other SPD processes. The TE process can
be applicable for any samples except having circular cross
section. Main parameters aecting plastic ow (and, accordingly, accumulated strain value and distribution) in TE are
shape of the twist channel cross-section, the twisting angle ,
height h and slope angle of the twist part of the channel
(see Fig. 1). In more details features of the twist extrusion
techniques has been described in,16) and strain estimation
in TE is shown in the next section.
TE has been used for grain renement of pure Cu of a
commercial grade,16) titanium1820) and Al-based alloys.15,17)
Nevertheless, there seems no systematic experiment so far
for microstructure analysis in TE-processed billet. It may be
useful to understand macro ow patterns in TE using physical
and mathematical modeling. However, further study for
microstructure evolution is also needed for experimental
verication on real metallic systems. In this study, pure
polycrystalline Al is adopted to TE as it is the most
representative metallic system having fcc crystal structure.

TE processed pure Al has been well studied in terms of both
macro- and microstructure to obtain much information for
comparison with the data from other SPD processes.2124)
2.
Experimental
This study used a high purity 99.99% aluminum produced

by ingot casting. Edge parts and 2 mm surface layers of the
as-received ingot were machined out. The ingot was hydroextruded at room temperature to a billet with cross-sectional
dimensions of 18 28 mm2 and it was cut to length of
100 mm for TE. The billets were annealed at 500 C for 1 hour
and cooled in furnace.
TE was conducted at room temperature with the extrusion
speed of 3 mms1 using a twisting die having 60 . A
backpressure of 200 MPa was applied to the billets during
TE. According to Ref. 12), net strain in twist extrusion has
distribution within cross-section of the billet so that minimal
("min ) and maximal ("max ) strain accumulated in a billet after
one TE pass could be estimated by simplied equations:
1
"min 0:4 0:1 tan ;
2
2
"max p tan :
3
More details of strain distribution after 1 TE pass are found
elsewhere.12,15,25) For the tool parameters used in this study,
average true strain of "ave 1:2 was introduced for a single
pass.12) This level of strain is comparable to strain introduced
in a billet after 1 ECAP pass through 90 channel intersection
die or 70% reduction in rolling. The twist extrusion was
repeated up to 4 passes which come to a total average strain
of "ave 4:8.12) This level of strain may be sucient to attain
UFG structures by ECAP22) or ARB.21)
In our earlier study,25) TEM observations and microhardness data of this investigation have been reported. The
present report is mostly concentrated on analysis of structure
evolution through optical microscopy and Electron Back-
97
Table 1 Total accumulated strain calculated by equations from Ref. 12) related to fractions of HABs and (sub)grain sizes at positions A
and B after 1 and 4 TE passes.
Equivalent strain
Point A
1.2
Point B
4 passes
Point A
4.8
Point B
scattered Diraction (EBSD) analysis. Optical microscopy

was undertaken using Nikon ECLIPSE ME600L equipped
with a digital camera to observe material ow patterns in the
billets before and after TE. Electron backscatter diraction
(EBSD) analysis was also conducted using Hitachi S-4300SE
equipped with a eld emission gun at an accelerating voltage
of 30 kV. Imaging of crystal orientation using EBSD
involved automatic beam scanning with step sizes of 0.08
0.3 mm. Grain sizes and misorientation angles were determined along with the EBSD analysis. Data acquisition and
subsequent analysis were performed using a TSL orientation
image microscopy system. A cleaning-up procedure was
applied to all EBSD images to adjust points with condence
index lower than 0.1. It is noted that, for all the EBSD maps
scanned, average condence index was >0:3. Misorientations less than 2 were excluded from the analysis because
of the limitations of the angular resolution of the EBSD
technique.26) Thus, low angle boundaries (LABs) were
dened as those having misorientations from 2 to 15 and
high angle boundaries (HABs) as having misorientations
greater than 15 .
Specimens for optical microscopy were cut by an abrasive
saw and mechanically ground by abrasive papers. They were
anodized in a solution containing 5% HBF4 in H2 0 at room
temperature with a voltage of 20 V. For EBSD analysis, the
samples were electro-polished in a perchloric-based solution
at room temperature with a voltage of 12 V. Special
attention was paid to dene directions of the samples with
respect to the billet geometry as illustrated in Fig. 1, where
the extrusion direction (ED) is parallel to the longitudinal
axis of the billet and the normal direction (ND) and
transverse direction (TD) are, respectively, parallel to the
short and long axes on the cross-sectional plane perpendicular to ED of the billet. EBSD analysis was conducted at
points marked A and B in Fig. 1, which represents minimal
and maximal strain values, respectively (see Table 1).
3.
After annealing of the hydrostatically extruded billet, the

microstructure consists of dierent sizes of recrystallized
grains as shown in Fig. 2(a). It appears that the sizes of
individual grains vary over one order of magnitude. EBSD
analysis showed that the average grain size was 280 mm.
Figure 2(b) shows an optical micrograph after the rst pass
of TE ("ave 1:2). The microstructure is heterogeneous with
the peripheral areas more deformed than the central areas.
Grains in the peripheral areas are mostly elongated and
material ow is created due to the TE tool geometry. The
(sub)grain size,
[%]
[mm]
0.6
7.6
3.16
14.5
3.08
2.4
53
1.88
71
1.61
28
mm
A
B
ND
ED
TD
=60
h
1 pass
Fractions of HABs,
local
mm
average
18
Specimen state, position
=90
Fig. 1 Illustration of die for twist extrusion showing the channel
dimensions, denition of directions and points of observations.
central areas contain rather coarse grains, some of which

appear equiaxed.
An optical micrograph after 4 passes of TE ("ave 4:8) is
shown in Fig. 2(c). The microstructure becomes ner with
clear vortex ow of the material. The structure indicates that
the billet underwent very high strain at the outer edge but less
at the center. The microstructure appears as continuous
development of the structure formed under the rst TE pass.
The observation by optical microscopy reveals that the
billet processed by TE has axial symmetry in microstructure.
These observations of the macro-ow patterns are in
qualitative agreement with the previous work,15) although
materials structure-related specic features introduce minor
dierences. Indeed, investigation of microstructure in15) was
limited to representative, but separate areas, while this study
has allowed to understand whole the cross-sectional macroow patterns.
EBSD analysis was then conducted at the center area
(point A) and at the outer edge (point B) of the billets as
marked in Fig. 1. Figure 3(a) and (b) show orientation
images after one pass ("min 0:6) and 4 passes ("min 2:4)
of TE at point A (center), respectively. The images reveal
that a subgrain structure with low angle boundaries is well-
98
D. Orlov et al.
Number Fraction
0.8
0.6
99.99% Al
1 pass;
4 passes
0.4
0.2
0
10 20 30 40 50 60
Misorientation Angle, /degree

Fig. 4 Distribution of misorientation angle after one and four passes of TE
at position A (center).
Fig. 2 Optical micrographs of cross-sections (a) before TE, (b) after one
pass of TE and (c) after four passes of TE.
developed after one pass and this is essentially the same after
4 passes except that there are some grains with high angle
boundaries in Fig. 3(b). This is demonstrated more clearly in
Fig. 4 where the fraction of boundaries is plotted against
misorientation angle. The fraction of low angle grain
boundaries is large after one pass but it signicantly
decreases after 4 passes.
Orientation images after one pass ("max 2) and 4 passes
("max 8) of TE at point B (edge) are shown in Fig. 5(a) and
(b), respectively. The microstructure consists of subgrains
with low angle boundaries after one pass, but an alternate
banded structure is developed after 4 passes with clear
formation of ne grains surrounded by high angle boundaries. Figure 6 demonstrates more quantitatively the evolution
of grain boundary structures from one pass to 4 passes: a
signicant decrease in low angle boundaries but instead a
large increase in high angle boundaries is seen. When EBSD
results are compared between point A and point B, the
fraction of low angle boundaries is lowered and the overall
fraction of high angle boundaries is higher at point B, that
Fig. 3 Orientation images after one pass (a) and after four passes (b) of TE at point A (center). LABs (2 < 15 ) are shown in white
lines and HABs ( 15 ) are shown in black lines. Grayscale contrast on the images corresponds to crystallographic orientations.
99
Number Fraction
Fig. 5 Orientation images after one pass (a) and after four passes (b) of TE at point B (edge). LABs (2 < 15 ) are shown in white
lines and HABs ( 15 ) are shown in black lines. Grayscale contrast on the images corresponds to crystallographic orientations.
0.8
0.6
99.99% Al
1 pass;
4 passes
0.4
0.2
0
10 20 30 40 50 60
Misorientation Angle, /degree

Fig. 6 Distribution of misorientation angle after one pass and four passes
of TE at position B (edge).
rather well corresponds to the strain predicted by eqs. (1)

and (2).
Summary of the EBSD analysis including (sub)grain size
is given in Table 1. It is shown that the development of high
angle boundaries is not sucient after one pass but after 4
passes the fraction of high angle boundaries reaches more
than 50% at point A and more than 70% at point B. When
compared with quantitative results reported earlier for an Al0.13%Mg alloy,15) they are, again, in a good agreement
qualitatively, but signicantly dierent quantitatively.
Namely, the fractions of HAB were limited only to 27%
and 57% at the center and edge in the Al-Mg alloy even after
8 TE passes. This discrepancy should be attributed to
substantial dierence in dislocation structures between pure
Al and Al-Mg. In pure Al, enhanced recovery reduces the
accumulation of dislocations during the process but instead
increases the fraction of HABs in the microstructure, while
dynamic and static recovery is signicantly inhibited by the
dilute addition of Mg.
It is also shown that the grain size is reduced to 1:6 mm at
point B after 4 passes of TE. This grain size is slightly larger
than the ones obtained using the ECAP2224) and ARB21)

processes for the same grade of materials despite that all the
ECAP, ARB and TE were conducted at room temperature.
The dierence might be attributed to a reversal character of
straining under the TE processing. A similar eect was also
observed in ECAP processing by route C. It should be noted
that the process of route C in ECAP involves rotation of the
billet by 180 about the longitudinal axis after each pass and
thus reversal shear is introduced in the billet.27) It should also
be noted that route BC has been used in ECAP process to
obtain the ner grain sizes in previous reports,2224) where a
billet is rotated by 90 between each pressing in the same
sense so that dierent shear planes are activated to make the
grain size ner.27) These results show that structure evolution
in TE of 99.99%Al is similar to all the SPD techniques
previously reported and any of the SPD techniques is capable
of providing an UFG structure. Main dierence between
these techniques is technological implementation. Thus, the
TE technique has the following features and advantages:16)
size of distorted areas of the specimen, that is the head and
rear parts, of the billet, is smaller under TE, which is
especially important when doing repeated runs; (2) TE can
maintain the initial shape of billets including those with an
axial channel; (3) since TE does not change the direction
of a billets movement, it can be introduced into existing
industrial lines by replacing a standard extrusion die with a
twist extrusion die.
From the observations of this study, it is envisaged that,
under TE operation, there may be two processes of structure
evolution in 99.99%Al: rst, development of the deformation-induced grain boundaries and second, a continuous
increase in grain misorientations. Initially, the rst process
dominates, where the development of the deformationinduced boundaries occurs heterogeneously in the billet
while directions of all boundaries are strongly aected by the
tool geometry. The boundary spacing converges to a lower
limit which is 1:6 mm for this study. Thereafter, the
formation of new boundaries saturates and the second
100
D. Orlov et al.
process of a misorientation increase becomes dominant.

LABs form on the rst stages and continuously convert
into HABs: namely, subgrain structures with low angle
boundaries become grains with high angle boundaries. These
observations are in a reasonable agreement with earlier
results on microhardness and dislocation structure evolution
reported in.25)
4.
5)
6)
7)
8)
9)
10)
(1) High purity Al (99.99%) was processed by twist

extrusion and microstructure evolution was analyzed
using optical microscopy and electron backscattered
diraction analysis.
(2) After one pass of TE at average strain "ave 1:2, the
microstructure was heterogeneous with the peripheral
areas more deformed than the central areas. Subgrain
structures with low angle boundaries were developed
throughout the cross section perpendicular to the
longitudinal axis of the billet.
(3) After four passes of TE at strain level "ave 4:8, a
vortex-like material ow clearly appeared with axis
symmetry and the microstructure consisted of grains
with an increasing fraction of high angle boundaries:
more than 50% at the center parts and more than
70% at the outer edge parts. The average grain size
was reduced to 1:6 mm at the outer edge part after 4
passes of TE.
11)
12)
13)
14)
15)
16)
17)
Acknowledgements
18)
This work was supported in part by the Light Metals

Educational Foundation of Japan; in part by the 21st COE
program Functional Innovation of Molecular Informatics
and Global COE program Center of Excellence for
Advanced Structural and Functional Materials Design, and
in part by a Grant-in-Aid for Scientic Research in Priority
Areas Giant Straining Process for Advanced Materials
Containing Ultra-High Density Lattice Defects, all from
the Ministry of Education, Culture, Sports, Science and
Technology, Japan.
19)
20)
21)
22)
23)
REFERENCES
24)
1) T. M. Osman, D. E. Rardon, L. B. Friedman and L. F. Vega: The
Journal of Materials 58 (2006) 2124.
2) T. C. Lowe: The Journal of Materials 58 (2006) 2832.
Y. T. Zhu: The Journal of Materials 58 (2006) 3339.
4) P. W. Bridgman: Studies in Large Plastic Flow and Fracture: With
25)
26)
27)
Special Emphasis on the Eects of Hydrostatic Pressure, (McGrawHill, New York, 1952).
V. M. Segal, V. I. Reznikov, A. E. Drobyshevkiy and V. I. Kopylov:
Russian Metall (1981) 99.
V. M. Segal: Mater. Sci. Eng. A 197 (1995) 157164.
G. A. Salishchev, O. R. Valiahmetov, R. M. Galeev and S. P.
Malysheva: Izvestiya Akademii Nauk SSSR Metally (1996) 8691.
J. Y. Huang, Y. T. Zhu, H. Jiang and T. C. Lowe: Acta Mater. 49 (2001)
14971505.
N. Tsuji, Y. Saito, H. Utsunomiya and S. Tanigawa: Scr. Mater. 40
(1999) 795800.
Y. Saito, H. Utsunomiya, N. Tsuji and T. Sakai: Acta Mater. 47 (1999)
579583.
Y. Beygelzimer, D. Orlov and V. Varyukhin: 2002 TMS Annual
Meeting and Exhibition, ed. by Y. T. Zhu, T. G. Langdon, R. S. Mishra,
S. L. Semiatin, M. J. Saran, T. C. Lowe., (TMS, Seattle, Washington,
USA, 2002) pp. 297304.
Y. Beygelzimer, V. Varyukhin, D. Orlov and S. Synkov: Twist
extrusionprocess for strain accumulation, (TEAN, Donetsk, 2003)
p. 87.
Y. Y. Beygelzimer, V. N. Varyukhin, S. G. Synkov, A. N. Sapronov
and V. G. Synkov: Physics and Technology of High Pressures 9 (1999)
109110.
Y. Beygelzimer and D. Orlov: Defect and Diusion Forum 208209
(2002) 311314.
M. Berta, D. Orlov and P. B. Prangnell: Inter. J. Mater. Res. 98 (2007)
200204.
Y. Beygelzimer, D. Orlov, A. Korshunov, S. Synkov, V. Varyukhin,
I. Vedernikova, A. Reshetov, A. Synkov, L. Polyakov and I.
Korotchenkova: Solid State Phenomena 114 (2006) 6978.
D. Orlov, A. Reshetov, A. Synkov, V. Varyukhin, D. Lotsko, O. Sirko,
N. Zakharova, A. Sharovsky, V. Voropaiev, Y. Milman and S. Synkov:
Twist Extrusion as a Tool for Grain Renement in Al-Mg-Sc-Zr
Alloys, ed. by Y. T. Zhu, V. Varyukhin vol 212, (Springer Netherlands,
2006) pp. 7782.
V. V. Stolyarov, Y. E. Beigelzimer, D. V. Orlov and R. Z. Valiev: The
Physics of Metals and Metallography 99 (2005) 204211.
D. V. Orlov, V. V. Stolyarov, H. S. Salimgareyev, E. P. Soshnikova,
A. V. Reshetov, Y. Y. Beygelzimer, S. G. Synkov and V. N. Varyukhin:
2004 TMS Annual Meeting, ed. by Y. T. Zhu, T. G. Langdon, R. Z.
Valiev, S. L. Semiatin, D. H. Shin, T. C. Lowe, (TMS, Charlotte, North
Carolina, USA, 2004) pp. 457462.
Y. Beygelzimer, V. Varyukhin, D. Orlov, B. Efros, V. Stolyarov and
H. Salimgareyev: 2002 TMS Annual Meeting and Exhibition, ed. by
Y. T. Zhu, T. G. Langdon, R. S. Mishra, S. L. Semiatin, M. J. Saran and
T. C. Lowe, (TMS Seattle, Washington, USA, 2002) pp. 4346.
N. Kamikawa, N. Tsuji, X. Huang and N. Hansen: Acta Mater. 54
(2006) 30553066.
Y. Iwahashi, Z. Horita, M. Nemoto and T. G. Langdon: Acta Mater. 46
(1998) 33173331.
S. D. Terhune, D. L. Swisher, K. Oh-ishi, Z. Horita, T. G. Langdon and
T. R. McNelley: Metall. Mater. Trans. A 33A (2002) 21732184.
A. P. Zhilyaev, D. L. Swisher, K. Oh-ishi, T. G. Langdon and T. R.
McNelley: Mater. Sci. Eng. A 429 (2006) 137148.
D. Orlov, Y. Beygelzimer, S. Synkov, V. Varyukhin and Z. Horita:
Mater. Trans. 49 (2008) 26.
F. J. Humphreys: Scr. Mater. 51 (2004) 771776.
M. Furukawa, Y. Iwahashi, Z. Horita, M. Nemoto and T. G. Langdon:
Mater. Sci. Eng. A 257 (1998) 328332.

Partial Grain Renement in Al-3%Cu Alloy during ECAP

at Elevated Temperatures
Inna Mazurina1 , Taku Sakai1 , Hiromi Miura1 , Oleg Sitdikov1;2 and Rustam Kaibyshev3
1
Department of Mechanical Engineering and Intelligent Systems, UEC Tokyo (The University of Electro-Communications),
Tokyo 182-8585, Japan
2
Institute for Metals Superplasticity Problems, Khalturina 39, Ufa 450001, Russia
3
Belgorod State University, Belgorod 308034, Russia
Microstructural changes during equal channel angular pressing (ECAP) in a temperature interval from 523 to 748 K (0:6{0:8 Tm ) were
studied in a coarse-grained binary aluminum alloy Al-3%Cu. Hot ECAP results in grain renement taking place at all temperatures investigated,
leading to a non-uniform development of deformation-induced ne grains and the remnant original grains containing subgrains with low-angle
boundaries. Fine-grained structure is developed along microshear bands formed in grain interiors as well as along initial grain boundaries. The
number and the average misorientation angle of the boundaries of microshear bands start to increase at above a critical strain of about 2, nally
leading to development of new ne-grained structures. The volume fraction and the average misorientation of deformation-induced boundaries
are reduced with rising temperature. The thermal stability of the evolved deformation microstructures and several factors controlling grain
renement during hot ECAP are discussed in details. [doi:10.2320/matertrans.MD200807]
(Received June 2, 2008; Accepted August 1, 2008; Published September 18, 2008)
Keywords: grain renement, temperature eect, microshear band, equal channel angular pressing (ECAP), aluminum alloy
1.
Introduction
In recent years, there has been an increasing interest in the

production of ultrane-grained (UFG) materials by means of
severe plastic deformation (SPD) for both commercial and
investigative purposes.1) Several SPD techniques such as
equal channel angular pressing (ECAP),120) torsion under
high pressure1,3,21) and multi-directional forging (MDF)2228)
are now available for attaining of the UFG structures in
numerous metals and alloys. There are several reports to date
dealt with the investigation of the ne-grained structure
evolution in Al alloys at dierent temperatures during
SPD.24,6,8,9,1114,16,1822,2426) Under severe plastic deformation conditions, fragmentation of original grains takes place
at medium strains followed by in-situ formation of ne
crystallites at high strains, which is sometimes called as continuous dynamic recrystallization (CDRX).9,16,1820,22,24,25)
Such structural changes in Al alloys during SPD are mostly
investigated at ambient temperature and only few studies
address high-temperature SPD and UFG formation at
elevated temperatures.6,810,12,13,16,1820,2426)
The present authors have investigated the microstructural
evolution in Al alloys during hot SPD in the recent
works.9,12,16,1820,24,25) Multi-component commercial-based
Al alloys were used for these investigations, because both
thermally stable dispersion particles and soluble precipitations coexist in a wide temperature range. It has been shown
that the deformation microstructures having high density of
dislocations can be easily developed and kept in such alloys
due to a pinning eect of the particles. The latter may strongly inhibit the rate of dynamic recovery and provide a high
thermal stability of deformation microstructures.9,12,16,24,25)
As a result, rather ne-grained structures with the moderate
dislocation density can be formed at large strains even at
high temperatures. On the other hand, such microstructural
changes may be dierent from those in more pure Al alloys
without any dispersion particles, because the microstructures
evolved may be characterized by a low thermal stability at

high temperatures.8) The main features of the microstructural
evolution in such alloys during hot SPD, however, have been
not still studied in sucient details.
The aim of the present work was to study the microstructural changes taking place during hot ECAP of a coarsegrained binary alloy Al-3%Cu. The microstructural evolution
and mechanisms of new grain formation were investigated
mainly at a temperature of 523 K. The misorientation characteristics of deformation-induced boundaries and the other
parameters of the evolved microstructures were studied in
details at various strains up to 12. An inuence of increasing
the deformation temperature in the interval from 523 to
748 K (0:6{0:8 Tm ) on grain renement process is also
analyzed and discussed.
2.
A binary aluminum alloy Al-3%Cu (in mass%) used in the

present work was produced by direct chill casting. The alloy
was homogenized at 793 K for 4 h as a solution treatment,
followed by cooling with a rate of about 50 K/h leading to
formation of relatively coarse particles of Al2 Cu (-phase)
and Al-0.1%Cu matrix at room temperature.29) Note that such
preliminary heat treatment regime was especially selected for
this alloy to avoid any eects of deformation temperature on
precipitation and analyze only the microstructural evolution
during ECAP. The initial microstructure was composed of
roughly equiaxed coarse grains with the grain size varied
from 100 to 350 mm (Fig. 1(a)). The precipitates of -phase
showed round platelet shape and the size ranging from 0.5 to
1 mm (Fig. 1(b)). Samples for ECAP were machined into
rods with a diameter of 20 mm and a length of 100 mm.
ECAP was conducted at temperatures from 523 to 748 K
using an isothermal die with a circular internal cross-section
with a diameter of 20 mm. The channel had an L-shaped
conguration with an angle of intersection of 90 and that at
102
I. Mazurina, T. Sakai, H. Miura, O. Sitdikov and R. Kaibyshev
(a)
100 m
Al2Cu
(b)
500 nm
Fig. 1 Initial microstructure of Al-3%Cu alloy before ECAP. (a) Optical
and (b) TEM micrographs.
the outer arc curvature of 0 . Deformation through this

channel produced a strain of about " 1 upon each pass
through the die.2) The samples were pressed repeatedly to
strains of about 12 using route A, i.e. without rotation of the
billet in each pass. This route, being the most eective one
for formation of ne-grained structure with high angle
boundaries (HABs),12) has been used in the authors previous
works9,12,16,1820) and applied here to provide some compatible data.
The pressed samples were quenched in water after each
ECAP pass and nally cut parallel to the pressing direction in
longitudinal section for microstructural observation. Metallographic analysis was carried out using an optical microscope (OM) after etching by the standard Dics-Keller
reagent. The orientation imaging microscopy (OIM) with
automated indexing of electron back scattering diraction
(EBSD) patterns was performed in LEO-1530 scanning
electron microscope (SEM) with OIM analysis software
provided by HKL Technology, Inc. The misorientation
distributions of deformation-induced boundaries were derived from EBSD technique. The boundaries with misorientation less than 2 were not taken into account. On the
other hand, thin discs with a diameter of 3 mm for transmission electron microscopy (TEM) examination were
mechanically thinned and then electropolished in a solution
of 30% HNO3 and 70% CH3 OH at about 30 C using a
Tenupol-3 twin-jet polishing unit. Foils were examined
using a JEOL-2000EX TEM.
3.
Results
3.1 Microstructural evolution at 523 K

3.1.1 Deformation microstructure
Typical OM microstructures developed in Al-3%Cu alloy
during ECAP at 523 K are shown in Fig. 2. It is remarkable to
see that the deformed microstructures are highly inhomogeneous even at a nal strain of 12. Microstructures developed
after the rst passage are mainly characterized by elongated
initial grains aligned roughly parallel to the pressing direction
(PD), in interiors of which deformation bands (DBs) are
formed frequently and heterogeneously (Fig. 2(a) and 2(b)).
These bands evolved in aluminum alloy are found to be
mainly microshear bands (MSBs), the characteristics of
which are described elsewhere.19,20,25,30) The longitudinal
grain size and the apparent width of these MSBs are over
400 mm and varied from 6 to 40 mm, respectively. Deformation in the strain interval from 4 to 8 leads to an increase in
number of MSBs (Fig. 2(c) and 2(d)), nally leading to a
roughly full development of a ne granular structure in the
regions with high density of MSBs. It is clearly seen in
Figs. 2(d) and 2(e) that such deformed microstructures are
inhomogeneously distributed in original grain interiors even
at high strains.
Typical orientation imaging microscopy (OIM) maps and
the corresponding misorientation proles are represented in
Figs. 3 and 4, respectively. Figure 4 shows typical distributions of point-to-point () and cumulative (point-to-origin,
) misorientations developed along test lines indicated in
Fig. 3. The values of and dene the misorientations evolved by 1 mm step relative to the previous point and
to the rst point, respectively. It can be seen from Fig. 3(a)
that new ne grains are inhomogeneously formed in grain
interiors. These grains with boundaries of moderate-to-high
misorientation angles are evolved in colony in coarse grain
interiors in the left hand side of Fig. 3(a) (see along the line
A1 ), while, as denoted by the line A2 , long-distance straight
LABs with low misorientation angles are developed in the
right hand side. The misorientation changes along the line
A2 show that low-to-moderate angle boundaries are inhomogeneously developed across grain interiors over a distance
of about 30 mm (Fig. 4(b)). These low angle boundaries
ranging from 5 to 10 may correspond to those for dislocation
subboundaries and embryos of MSBs,19,20) which will be
discussed later.
With further deformation to high strains, the boundaries of
such embryos of MSBs may transform to macroscopic and
long-distance MSBs, the boundaries of which have high
angle misorientations of more than 15 . The number and
misorientation of these boundaries increase with deformation
and the boundaries are pinched o in the length of ne
grains, leading to new ne crystallites developed along
these boundaries (Fig. 3(b)). As a result, the microstructures
developed at " 8 consist of the regions of new ne grains
with HABs and parts of original grains containing subgrains
with LABs as well as the boundaries of MSBs with moderate
angles (Fig. 4(c)). Further deformation to " 12 leads to an
increase in the volume fraction of ne grains with HABs
developed in places with high density of MSBs, while some
parts of original grains containing mainly subgrains with
Partial Grain Renement in Al-3%Cu Alloy during ECAP at Elevated Temperatures
103
(a)
300 m
(b)
(c)
(d)
(e)
PD
50 m
Fig. 2 Optical micrographs of Al-3%Cu alloy deformed by ECAP at 523 K up to strains: (a) " 1 (b) " 2, (c) " 4, (d) " 8,
(e) " 12.
LABs still exist in the right hand side in Fig. 3(c) (see also
Fig. 4(d)). It is concluded that, even at " 12, ne grains
with HABs are not fully evolved in Al-3%Cu alloy ECAPed
at 523 K.
Strain dependence of the average grain sizes in the regions
of newly developed grains is represented in Fig. 5. The
minimum width or spacing of MSBs is also plotted against
strain in this gure. The average sizes of crystallites were
measured in both the transverse and longitudinal directions of
developed microstructures, i.e. in the direction of elongated
grains (see Fig. 3). It is seen that the longitudinal grain size
and the width of MSBs tend to decrease rapidly at low strains
below " 4 and then gradually with increasing strain, while
the transverse grain size does not change signicantly. It is
interesting to note here that the transverse size of newly
developed grains is about two times less than the width of
MSBs formed in strain interval from 4 to 8. These three
parameters approach to a similar value of about 5 mm at
" 12 within experimental scatters. Such a good coincidence of the crystallite size and MSBs spacing suggests that
dynamic formation of these new grains may be closely
connected with the development of MSBs.
3.1.2 Misorientation characteristics of microstructure
Changes in the misorientation distribution of deformation-
induced boundaries are represented in Fig. 6. The distributions developed at " 1 and 2 are characterized by a high
fraction of LABs, so that the histograms plotted have a
single peak at around 4 . Some HABs start to develop at
" 3 due to the evolution of boundaries of MSBs and
formation of new ne grains, as conrmed in Fig. 3(a). As a
result, the distributions of deformation-induced boundaries
are shifted to the regions of HABs and show two peaks
corresponded to LABs and HABs at moderate and high
strains. Note that such bimodal distribution with some
fraction of LABs that exist even at high strains is a very
common feature of deformation-induced microstructure
misorientation.31)
Strain dependence of the average misorientation angle
(av ), the fraction of HABs with misorientation angle above
15 (FHAB ) and the volume fraction of new grains (VUFG ) are
shown in Fig. 7. The average misorientation does not exceed
5 and remains constant at early deformation from " 1 to 2,
and then starts to increase rapidly at moderate strains of 3 and
4. With further straining, the av continues to increase
gradually to 25 at " 12 (Fig. 7(a)). On the other hand,
both the fraction of HABs and the volume fraction of new
grains start to rise at " 2 and approach a saturation value
of about 0.5 at 8 " 12 (Fig. 7(b)). This suggests that the
104
30
(a)
Al - 3%Cu T = 523 K
A1
A2
10 m
(b)
Crystallite size, d / m
25
Minimum spacing of DBs
Transverse
Longitudinal
20
15
10
5
PD
25 m
(c)
25 m
Fig. 3 Typical OIM micropgraphs of Al-3%Cu alloy deformed at 523 K to

various strains: (a) " 3, (b) " 8, (c) " 12. Gray and black lines
correspond to the boundaries with misorientation from 5 to 15 and >15 ,
respectively.
6
8
Strain,
10
12
Fig. 5 Changes in crystallite size with strain for Al-3%Cu. Transverse and
longitudinal sizes were measured in perpendicular and parallel directions
of developed microstructure.
parameters av , FHAB and VUFG have some close connection

between each other and strain-induced grain formation starts
to take place at strains beyond a critical strain of about 2. In
other words, the gradual increase in the av , FHAB and VUFG
at large strain can be resulted from high density of MSBs
developed in original grain interiors during SPD.
3.1.3 TEM microstructures
Typical microstructures developed at various strains are
represented in Fig. 8. It can be seen in Fig. 8(a) that ECAP to
" 2 leads to development of parallel bands of subgrains
having low misorientation angles less than 5 , as suggested
Fig. 4 Point-to-point () and cumulative () misorientations of strain-induced boundaries developed along lines marked in Fig. 3:
(a) and (b) " 3, (c) " 8, (d) " 12.
0.8
(a)
=1
av = 4.1
0.6
0.2
Frequency, f
av = 14.04
0
0.8
(d)
=4
0.4
0.4
105
(e)
=8
(b) 0.4
=2
0.3
av = 4.3
0.2
0.4
av = 19.7
0.1
0
0
0.6
0.4
(c) 0.4
=3
0.3
av = 9.6
0.2
0.2
(f)
= 12
av = 25.4
0.1
0
0
0 10 20 30 40 50 60
0 10 20 30 40 50 60
Misorientation angle, / deg.
Fig. 6 Changes in the misorientation distribution of deformation-induced boundaries developed in Al-3%Cu alloy deformed by ECAP
at 523 K.
Average misorientation
angle, av / deg.
40
30
(a) Al-3%Cu
T = 523 K
20
10
Fraction of HABs and new

grains, FHAB and VUFG
0
(b)
0.6
0.4
FHAB
VUFG
0.2
0.0
4
6
8
Strain,
10
12
Fig. 7 Changes in (a) the average misorientation angle av and (b) the
fraction of strain-induced HABs, FHAB as well as the volume fraction new
grains VUFG with repeated ECAP at 523 K for Al-3%Cu alloy.
by single spot electron diraction (SAED) pattern obtained

in 6 mm diameter. Further straining to " 8 provides a
formation of roughly equiaxed (sub)grained structure, which
can be resulted from gradual breaking-up of the bands
described above (Fig. 8(b)). The SAED pattern suggests that
most of these boundaries have high misorientation angles.
After 12 passes of ECAP, roughly equiaxial (sub)grains
surrounded by HABs are formed (Fig. 8(c)). There are rather
coarse precipitates of -phase distributed roughly homogeneously in elongated bands and (sub)grain interiors and also
along the boundaries developed in Fig. 8. It is interesting
to note here that the particle size in Fig. 8(a) is remarkably
smaller than that of coarse precipitates present in the original
structure (Fig. 1(b)). This may suggest that the initial coarse
Al2 Cu precipitates are fragmented and/or dissolved and reprecipitated on developed dislocation substructures at earlier
stages of ECAP.15,17)
3.2 Temperature eect on microstructural development
3.2.1 Microstructures developed at high strain
Typical OIM maps developed at a strain of 12 at various
temperatures from 573 to 748 K are represented in Fig. 9.
The microstructure formed at 573 K looks like roughly
equiaxed ne (sub)grains with HABs in a whole volume.
106
(a)
(a)
25 m
0.5 m
(b)
(b)
PD
50 m
(c)
0.5 m
(c)
100 m
1 m
Fig. 8 Transmission electron micrographs of Al-3%Cu alloy deformed by
ECAP at 523 K up to strains: (a) " 2 (b) " 8, (c) " 12.
Subgrains with a rather coarse size and LABs are, however,

developed in some elongated grain interiors between them, as
indicated by arrows in Fig. 9. The volume fraction of such
subgrain structures increases with increasing temperature.
Further increasing temperature to 748 K results in the
formation of very coarse grains having almost no dislocation
substructures and a few regions with ne grains surrounded
by HABs (Fig. 9(c)). The former may be developed not only
by dynamic recovery, but also by interpass static annealing
taking place during ECAP.18) The average interpass time
under the present ECAP conditions was 45 min at each
temperature. On the other hand, the temperature at solubility
limit (solvus) line for Al-3%Cu alloy is around 730 3 K,
as it can be obtained from the equilibrium diagram.29) This
heat treatment can result in dissolution of Al2 Cu particles and
so, decrease the thermal stability of deformation-induced
Fig. 9 Typical optical micrographs of Al-3%Cu alloy deformed up to

strain 12 at various temperatures: (a) 573 K, (b) 673 K, (c) 748 K. Gray and
black lines correspond to the boundaries with misorientation from 5 to 15
and >15 , respectively.
substructures, leading to static recrystallization during ECAP

at 748 K.
3.2.2 Quantitative analysis of deformed microstructures
The misorientation distributions for deformation-induced
boundaries, measured only in the ne-grained regions developed at a strain of 12, are represented in Fig. 10. It can be
seen that the shape of the distributions is bimodal at each
temperature and does not change with increasing temperature
even at 748 K. The percentage of HABs with misorientations
over 15 as well as the average misorientation angle is
gradually decreased with increasing temperature.
Figure 11 shows the temperature dependencies of (a) the
grain size, dfine , and (b) the average misorientation angle,
av , measured in the regions of ne grains. The data for a
commercial 2219 Al alloy obtained at the same ECAP
conditions20) are plotted in Fig. 11 for reference. It can be
seen that the size of deformation-induced grains in Al-3%Cu
alloy tends to grow with increasing deformation temperature
and, in contrast, their boundary misorientation gradually
0.4
(a)
523 K
0.4
0.2
0.2
Frequency, f
av = 25.4
0
0.4
(b)
573 K
107
(c)
673 K
av = 18.3
0
0.6
(d)
748 K
0.4
0.2
av = 24.8
0
0 10 20 30 40 50 60
0.2
av = 14.1
0
0 10 20 30 40 50 60
Misorienation angle, / deg.

Fig. 10 Misorientation distributions for deformation-induced boundaries developed at " 12 in Al-3%Cu alloy after ECAP at various
temperatures.
1.0
Misorientation angle, av / deg.
(a)
30 = 12
Al-3%Cu
20
AA 2219
10
Fraction of fine grains, VUFG
Grain size, dfine / m
40
ECAP
0.8 =12
0.6
0.4
Al-3%Cu
0.2
( )
0.0
400
0
40
(b)
AA 2219
30
AA 2219
500
600
700
Temperature, T / K
800
Fig. 12 Temperature dependence of the volume fraction of new ne grains

VUFG developed at " 12 in Al-3%Cu alloy. The data for AA 2219
deformed at the same conditions are shown for reference 20).
20
10
0
Al-3%Cu
500
600
700
800
Temperature, T / K
Fig. 11 Changes in (a) average grain size, dfine , and (b) average
misorientation angle, av , in the regions of deformation-induced grain
structure developed in Al-3%Cu alloy at various temperatures after ECAP
at " 12. The data for AA 2219 deformed at the same conditions are
shown for reference.20)
decreases. It is interesting to note that the smaller dfine

evolved in the 2219 Al alloy increases at a remarkably slower
rate and the av of deformation-induced boundaries does not
change even at high temperatures, comparing with those for
the Al-3%Cu alloy.
The temperature dependencies of the volume fraction of
new ne grains, VUFG , developed at " 12 in these alloys are
represented in Fig. 12. The VUFG for Al-3%Cu alloy at 748 K
is shown with a parenthesis because it may be aected by
recrystallization (see Fig. 9(c)). The fractions of ne grains
developed in the Al-3%Cu and 2219 Al alloys approach
maximum values 0.5 and 0.85 at 523 K, respectively, and
decrease rapidly with increasing deformation temperature.
108
It is concluded, therefore, that strain-induced ne grains are

not developed completely during warm-to-hot ECAP and
there may be a limit of strain-induced grain formation in both
the alloys at high temperatures. Note also that the VUFG for
the 2219 Al alloy remains signicantly larger at all temperatures investigated. Such inuence of temperature on the
microstructural parameters developed during ECAP will be
discussed in details in the next sections.
4.
Discussion
4.1 Strain-induced grain renement process

The experimental results described above show that grain
renement takes place in the Al-3%Cu alloy during ECAP at
523748 K. It can be suggested in the results of Figs. 2, 3, 5
and 7 that the mechanism of ne grain formation may be
directly associated with MSBs evolved during ECAP. Note
that these results may be similar to the characteristics of
substructural development and grain renement taking place,
for instance, during cold rolling of Al-0.1%Mg alloy,30)
room-temperature ECAP of Al-0.13%Mg alloy14) and during
multi-directional forging at low, intermediate and high
temperatures in pure copper,27) Fe-20%Cr alloy28) and 7475
Al alloy,24,25) respectively. It has been pointed out in these
works that grain renement process taking place during SPD
may be related to evolution of MSBs, which are inhomogeneously formed by dynamic relaxation of strain gradients
developed in grain interiors, irrespectively of deformation
mode and temperature.9,24,25,28)
The present authors investigated the process of new grain
formation due to development of MSBs in a commercial Al
alloy 2219 during ECAP at 523748 K.19,20) The main results
are briey summarized here. The microstructural changes
during ECAP can be subdivided in three stages: i.e. an
incubation period for new grain evolution in stage 1; grain
fragmentation by frequent development of MSBs in stage 2
and continuous development of new grains in stage 3. In
stage 1, below a critical strain, "c 2, conventional dislocation substructures with LABs are homogeneously developed accompanied by numerous boundaries with low-tomoderate misorientation angles appearing as embryos of
MSBs. These contain elongated substructures with misorientation angles of 5 15 , while conventional equiaxed subgrains with LABs of 5 are developed in the
exteriors of MSBs. In stage 2, such embryos can transform
into macroscopic MSBs. The number and the average
misorientation of the boundaries of MSBs rapidly grow with
increasing the deformation, leading to fragmentation of
original grain into dierent misoriented regions. In stage 3,
new grains surrounded by HABs are fully developed in a
whole volume.
The similar sequence of the structural changes can be
observed in the current Al-3%Cu alloy (see Figs. 27).
Namely, the boundaries of MSBs with low-to-moderate
misorientations, introduced by early deformation, start to
transform into HABs at above a critical strain of "c 2 with
repeated ECAP. The value of "c 2 does not depend on not
only temperature, but also the chemical composition of Al
alloys.19,20) The crystal orientation of each band is frequently
alternated and changed discontinuously at the boundaries
of MSBs, as seen in Figs. 4(a) and 4(b). This is reasonably

consistent with a rigid local lattice rotation akin to kinking.30,31) Such bands may be roughly categorized as those
with transition boundaries having persistent nature. Hence,
they are introduced by mechanical working and thus, are
mainly controlled by athermal processes.20,27,28) It has been
pointed out20) that the development of ne-grains regions
accompanied by such MSBs can be scarcely inuenced by
pressing temperature in stages 1 and 2.
On the other hand, the boundaries of MSBs introduced at
stages 1 and 2 are rather non-equilibrium and diuse
interfaces1,14) (see Fig. 8). Dynamic recovery that occurs
with higher rate at elevated temperatures can assist the
transformation of such strain-induced non-equilibrium boundaries to more equilibrium ones, leading to a more rapid
formation of new grains. Repeated deformation to large
strains can give enough time for dislocation rearrangement
in strain-induced boundaries, leading to both an increase
of grain boundary misorientation and a decrease of dislocation density in (sub)grain interiors.20,27,28,31) Thus, the straininduced grain formation may be controlled by a dynamic
formation of non-equilibrium moderate-to-high angle boundaries in stages 1 and 2, and subsequently by frequent
operation of dynamic recovery in these boundaries at large
strains (i.e. in stage 3). Microstructure changes in stage 3 are,
therefore, considered to be controlled by thermally activated
rate processes.27,28,31) It is important to note in Figs. 57 that
the values of VUFG , FHABs and av are gradually increased
during grain renement, while the average crystallite size
developed in ne grained regions retains essentially stable
at moderate and high strains. Such dependencies may be
considered to be specic features of straininduced grain
structures, typical of cDRX.22,31) It can be concluded,
therefore, that new ne grains evolved during warm-to-hot
ECAP of the present Al alloy are resulted from a series of
strain-induced continuous reactions; this is essentially similar
to cDRX.
4.2
Controlling factors of strain-induced grain renement

It can be seen in Figs. 912 that (i) there is a limit of grain
renement in a dilute Al alloy at elevated temperatures, i.e.
grain renement occurs incompletely even after SPD to large
strains, leading to non-uniform development of deformationinduced ne grains with HABs and the remnant original
grains containing subgrains with LABs; (ii) the volume
fraction and the average misorientation angle of deformationinduced grains are reduced with increasing temperature.
These results are roughly similar to those for commercial
2219 Al alloy at the similar SPD conditions. The VUFG for the
2219 Al alloy, however, shows a maximum of about 0.85 at
523 K and decreases with increasing temperature, approaching about 0.5 at 748 K. These values are much larger than
those for the Al-3% Cu alloy, as can be seen in Fig. 12. This
suggests that dynamic evolution of new grains would be
strongly controlled by some material/deformation conditions. The latter will be discussed in the following sections.
4.2.1 Plastic constrain during severe deformation
It is found in the present study that any deformation
heterogeneities that result in development of MSBs are
expected to play an important role in grain renement by

severe deformation. It is known that inhomogeneous characteristics may lead to formation of MSBs due to relaxation
of strain gradients introduced by ECAP.4,9,12,16,18) It should be
noted that MSBs may be developed in grain interiors due to
various kinds of plastic constrains, which may be resulted
from deformation/material conditions, for instance, ECAP
channel and strain compatibility requirements by neighboring grains, etc. During cold-to-warm deformation, the
formation of MSBs occurs because it is easier for a grain to
deform if it is split into MSBs, where the number of slip
systems required for constrained deformation are fewer than
5.14,31) The MSBs may also be formed by grain subdivision as
a result of microshear banding that is caused by an intrinsic
structure instability.3,30,31) Some external constrains may also
appear on macroscopic level, as an inevitable consequence of
interaction between sample and tool, e.g. due to friction in
metalworking process. It is obvious that when a sample is
deformed complying with any external constrains, dierent
macroscopic strains should be developed in various parts;
this may lead to a strong inhomogeneous deformation in
a whole volume as well as grain interiors and thus, to
development of the microstructural inhomogeneity followed
by formation of MSBs.
4.2.2 Deformation temperature
It is known that the structural heterogeneity introduced by
plastic deformation generally decreases with increase in
deformation temperature. First, an increase in number of
operating slip systems gives rise to more homogeneous
deformation. Second, grain boundary sliding (GBS) can
result in a high rate of relaxation processes operating in
matrices with heterogeneous microstructures. Third, precipitation particles in the aluminum alloys are dissolved in
matrix, although dispersion particles can exist stably. This
will be discussed in more details in the next section. Fourth,
the deformed microstructure becomes rather unstable at high
temperatures due to frequent operation of dynamic restoration, such as dynamic recovery and even recrystallization. As
the results, the number of grains having large-scale MSBs
can be gradually decreased at elevated temperatures32,33) and
then ne-grained microstructures with HABs are hardly
developed in the original coarse grains containing subgrains
with LABs (see Fig. 9).
4.2.3 Second phase particles
It is seen that coarse second phase precipitates of Al2 Cu
phase are distributed in grain interiors in the present Al3%Cu alloy after overageing treatment (Fig. 1) and numerous MSBs are formed close to the regions containing rich
second phase particles11,12) (Fig. 2). These precipitates are
fragmented after ECAP and developed homogeneously in
elongated subgrains and along the subboundaries (Fig. 8).
These particles may reprecipitate along strain-induced
boundaries and can restrict rearrangement of lattice dislocation in short and long range, leading to stabilization of
developed dislocation substructures and then rapid formation
of MSBs.12) This may more eectively work at high
temperatures (T > 0:5 Tm ) despite operation of dynamic
recovery, leading to dislocation annihilation within (sub)grains. It can be concluded, therefore that such high-density
ne particles can retard or prevent any relaxation of strain
109
gradients,9,11,12,24,25) resulting in rapid increase in misorientation of strain-induced boundaries, followed by formation

of new grains with HABs.
In contrast to the dilute Al-3%Cu alloy, the 2219 Al alloy
contains various alloying elements with moderate concentrations and several secondary phase particles, as well as
some dispersion particles (Al3 Cr, Al6 Mn, Al3 Zr).19,20) Thus,
grain renement process in this alloy may occur more readily
because several second phase particles including dispersoids
in the 2219 Al alloy can eectively stabilize high-density
dislocation microstructures introduced by ECAP and so,
deformation-induced microstructures in such alloys can be
more stable for a long period of processing time during
hot ECAP. This should result in a smaller grain size and a
larger average misorientation and a larger volume fraction of
ne grains developed at the same deformation temperature
(Figs. 11 and 12).
4.2.4 Static restoration during ECAP
An important feature of the ECAP technique at elevated
temperatures is that the ECAPed samples inevitably undergo
static annealing during repeated ECAP. The developed
microstructures should be aected by not only severe
deformation, but also static annealing, taking place due to
the following reasons.13,18) First, it is not possible to quench
the sample immediately after its transition through the
deformation zone. The shear deformation occurs by ECAP
only in the intersection plane of channels, while the material
remains under static annealing conditions inside heated die
at deformation temperature. Second, static annealing also
occurs at intervals of time between ECAP passes, since
deformation temperature should be maintained or periodically restored during the process. This is inherent to the
cyclic deformation mode of ECAP. The present data in
Fig. 9(c) imply that such static annealing, resulting in static
recovery and, sometimes, static recrystallization, may exercise the signicant inuence on the deformation microstructure of Al-3%Cu alloy, especially comparing with that
of the AA2219 during ECAP (see Figs. 11 and 12). Static
restoration processes may eectively decrease the strain
gradients and/or strain heterogeneities accumulated in each
ECAP pass at elevated temperatures31) and thus, provide
incomplete grain renement even at large strains achieved
by ECAP.
5.
Conclusions
The microstructure evolution, taking place during ECAP at

elevated temperatures, was investigated in a binary Al-3%Cu
alloy. The inuence of increasing deformation temperature
on formation of strain-induced grains was studied in the
temperature interval from 523 to 748 K. The main results
can be summarized as following.
(1) ECAP of Al-3%Cu alloy results in partial grain renement at all the pressing temperatures, leading to nonuniform development of deformation-induced ne
grains with HABs and the remnant original grains
containing subgrains with LABs.
(2) Fine-grained structure is developed by initial grain
subdivision due to formation of microshear bands
having moderate angle misorientation. The number
110
and the average misorientation angle of these boundaries increase with deformation, nally leading to
development of new ne grains along microshear
bands.
(3) The volume fraction and the average misorientation
angle of deformation-induced grains are reduced with
rising temperature, since inhomogeneous microstructures are hardly developed during hot deformation due
to dissolution of precipitates as well as frequent
operation of dynamic recovery.
(4) Second phase particles in Al alloys can serve as very
eective stabilizers for deformation-induced microstructures, which are developed during and after cyclic
ECAP at elevated temperatures.
Acknowledgment
The authors acknowledge with gratitude the nancial
supports received from the Ministry of Education, Culture,
Sports, Science and Technology on Priority Areas Giant
Straining Process for Advanced Materials Containing UltraHigh Density Lattice Defects and the Light Metals
Education Foundation in Japan. One of the authors (I.M.)
wishes to thank DAAD for the possibility to perform part of
this work in IMM-RWTH Aachen, Germany, and the
Japanese government for providing the scholarship. The
other one (O.S.) would like to thank the UEC Tokyo for
providing a scientic fellowship.
REFERENCES
1) R. Z. Valiev, R. K. Islamgaliev and I. V. Alexandrov: Progr. Mater. Sci.
45 (2000) 103189.
2) Y. Iwahashi, Z. Horita, M. Nemoto and T. G. Langdon: Acta Mater. 45
(1997) 47334741.
3) F. J. Humphreys, P. B. Prangnell, J. R. Bowen, A. Gholinia and
C. Harris: Phil. Trans. R. Soc. A 357 (1999) 16631681.
4) J. P. Bowen, P. B. Prangnell and F. J. Humphreys: Proc. 20th Ris
International Symposium, ed. by J. B. Bilde-Srensen, J. V. Carstensen,
N. Hansen, D. Juul Jensen, T. Leers, W. Pantleon, O. B. Pedersen and
G. Winther, (Ris Lab., Roskilde, Denmark, 1999), p. 269.
5) A. Yamashita, D. Yamaguchi, Z. Horita and T. G. Langdon: Mat. Sci.
Eng. A 287 (2000) 100106.
6) C. Pithan, T. Hashimoto, M. Kawazoe, J. Nagahora and K. Higashi:
Mat. Sci. Eng. A 280 (2000) 6268.
7) J. Y. Chang, J. S. Yoon and G. H. Kim: Scr. Mater. 45 (2001) 347354.

8) U. Chakkingal and P. F. Thomson: J. Mater. Proc. Tech. 117 (2001)
169177.
9) A. Goloborodko, O. Sitdikov, T. Sakai, R. Kaibyshev and H. Miura:
Mater. Trans. 44 (2003) 766774.
10) Y. C. Chen, Y. Y. Huang, C. P. Chang and P. W. Kao: Acta Mater. 51
(2003) 20052015.
11) P. J. Apps, J. R. Bowen and P. B. Prangnell: Acta Mater. 51 (2003)
28112822.
12) A. Goloborodko, O. Sitdikov, R. Kaibyshev, H. Miura and T. Sakai:
Mat. Sci. Eng. A 381 (2004) 121128.
13) M. Popovic and B. Verlinden: Mater. Sci. Technol. 21 (2005) 606612.
14) P. J. Apps, M. Berta and P. B. Pragnell: Acta Mater. 53 (2005)
499511.
15) Z. Zhang, S. Hosoda, I.-S. Kim and Y. Watanabe: Mater. Sci. Eng. A
425 (2006) 5563.
16) O. Sitdikov, T. Sakai, E. Avtokratova, R. Kaibyshev, Y. Kimura and
K. Tsuzaki: Mat. Sci. Eng. A 444 (2007) 1830.
17) J. Gubicza, I. Schiller, N. Q. Chinh, J. Illy, Z. Horita and T. G.
Langdon: Mat. Sci. Eng. A 460461 (2007) 7785.
18) O. Sitdikov, T. Sakai, E. Avtokratova, R. Kaibyshev, K. Tsuzaki and
Y. Watanabe: Acta Mater. 56 (2008) 821834.
19) I. Mazurina, T. Sakai, H. Miura, O. Sitdikov and R. Kaibyshev: Mat.
Sci. Eng. A 473 (2008) 297305.
20) I. Mazurina, T. Sakai, H. Miura, O. Sitdikov and R. Kaibyshev: Mat.
Sci. Eng. A 486 (2008) 662671.
21) R. Z. Valiev, Y. V. Ivanisenko, E. F. Rauch and B. Baudelet: Acta
Mater. 44 (1996) 47054712.
22) A. Belyakov, T. Sakai, H. Miura and K. Tsuzaki: Phil. Mag. A 81
(2001) 26292643.
23) R. M. Imaev, G. A. Salishchev, O. N. Senkov, V. M. Imayev, M. R.
Shagiev, N. K. Gabdullin, A. V. Kuznetsov and F. H. Froes: Mater. Sci.
Eng. A 300 (2001) 263277.
24) O. Sitdikov, T. Sakai, A. Goloborodko and H. Miura: Scripta Mat. 51
(2004) 175179.
25) O. Sitdikov, T. Sakai, A. Goloborodko, H. Miura and R. Kaibyshev:
Phil. Mag. 85 (2005) 11591175.
26) S. Ringeval, D. Piot, C. Desrayaud and J. H. Driver: Acta Mater. 54
(2006) 30953105.
27) C. Kobayashi, T. Sakai, A. Belyakov and H. Miura: Phil. Mag. Lett. 87
(2007) 751766.
28) T. Sakai, A. Belyakov and H. Miura: Met. Mat. Trans. 39A (2008)
22062214. doi: 10.1007/s11661-008-9556-8.
29) Metals Handbook: Metallography, Structures and Phase Diagrams, ed.
by T. Lyman (American Society of Metals, 8, 1973) pp. 163168.
30) P. J. Hurley and F. J. Humphreys: Acta Mater. 51 (2003) 10871102.
31) F. J. Humphreys and M. Hatherly: Recrystallization and Related
Annealing Phenomena, 2nd Edition, (Elsevier, 2004) pp. 451467.
32) E. Ball and F. J. Humphreys: Thermomechanical Processing (TMP2 ),
ed. by W. B. Hutchinson, et al., (ASM, Ohio, 1996) pp. 184.
33) A. Duckham, R. D. Knutsen and O. Engler: Acta Mater. 49 (2001)
27392749.

OVERVIEW
Achieving Microstructural Renement in Magnesium Alloys

through Severe Plastic Deformation
Roberto B. Figueiredo1 and Terence G. Langdon1;2; *
1
Departments of Aerospace & Mechanical Engineering and Materials Science, University of Southern California,
Los Angeles, CA 90089-1453, U.S.A.
2
Materials Research Group, School of Engineering Sciences, University of Southampton, Southampton SO17 1BJ, U.K.
Equal-channel angular pressing (ECAP) is an excellent processing tool for achieving exceptional grain renement, typically to the
submicrometer level, in a range of pure metals and metallic alloys. Although processing by ECAP is relatively straight-forward when using soft
metals, the processing becomes more dicult when using magnesium-based alloys and other similar dicult-to-work materials. This overview
examines the procedures that must be adopted for the successful processing of these materials and then describes some of the advanced
properties achieved after the successful processing of two magnesium alloys. [doi:10.2320/matertrans.MD200818]
Keywords: damage, equal-channel angular pressing, nite element modeling, magnesium alloys, superplasticity
1.
Introduction
The processing of metals through the application of severe

plastic deformation (SPD) is now well established for the
production of ultrane-grained materials having exceptional
mechanical and physical properties.1) Typically, the materials produced in SPD have submicrometer or nanometer grain
sizes, they exhibit exceptionally high strength and, if the
microstructures are reasonably stable at elevated temperatures, they exhibit a potential for use in superplastic forming
operations. Although several SPD processing techniques are
now available, the two most promising procedures are EqualChannel Angular Pressing (ECAP)2) and High-Pressure
Torsion (HPT).3) Furthermore, processing by ECAP is
especially attractive because it has the capability of producing relatively large bulk samples.
Many of the early studies of SPD processing were limited
to soft pure metals or solid solution alloys where processing
by ECAP is relatively easy. However, more recent attention
has focused on the processing of dicult-to-work materials
such as magnesium-based alloys. Magnesium alloys are now
attracting much attention because of their low density, lightweight and high potential for recycling.46) These alloys are
excellent candidate materials for use in a wide range of
applications in the automotive and other transportation
industries and in the manufacturing of commercial products.
They also exhibit the optimum damping capacities of any
known metal so that the alloys are useful for applications
where external vibrations must be dampened: examples
include in helicopters and satellites. This wide range of
applications has led to the prediction that we are now on the
verge of entering a new age of magnesium7) where
magnesium alloys will be widely used for the fabrication of
numerous parts and components.
The processing of magnesium alloys by ECAP is challenging. Early experiments showed that ultrane grain sizes
were not produced when pure magnesium and an Mg-9% Al
alloy were processed by ECAP8) and an approach was
*Corresponding
author, E-mail: langdon@usc.edu
developed whereby a preliminary extrusion step was introduced prior to ECAP so that a majority of the basal (0001)
planes became oriented parallel to the extrusion direction.9)
This procedure of extrusion plus ECAP was designated
EX-ECAP10) and experiments have shown it works well in
producing signicant grain renement in many magnesium
alloys.1113) However, despite this success there remains a
problem that many magnesium alloys fail during processing
through the development of substantial cracking or segmentation. This overview examines this problem in section 2 and
in section 3 there is a description of some of the advanced
properties that may be attained when magnesium alloys are
successfully processed by ECAP.
2.
The Development of Procedures for Successfully

Processing Magnesium Alloys
To obtain a detailed understanding of the characteristics of

the deformation occurring during ECAP when using dicultto-work materials such as magnesium-based alloys, it is
appropriate to make use of nite element modeling (FEM) as
described in an earlier report.14) Following conventional
practice, the FEM simulations incorporated the following
relationship to describe the material behavior:
A"n "_m
where is the eective stress, " is the accumulated eective

strain, "_ is the eective strain rate, the exponents n and m are
the strain hardening coecient and the strain rate sensitivity,
respectively, and A is a constant. For a direct evaluation of
the validity of the FEM predictions, parallel experiments
were also conducted using a commercial extruded ZK60
alloy with a composition, in mass%, of Mg-5.5% Zn-0.5%
Zr.
In practice the local strain distributions occurring within
the billets during ECAP vary as a function of the strain rate
sensitivity of the material. This variation is illustrated in
Fig. 1 where the angle between the two parts of the channel
in the ECAP die is xed at 90 and the strain levels are
plotted pictorially for strain rate sensitivities of (a) 0, (b) 0.2
112
R. B. Figueiredo and T. G. Langdon
relationship in eq. (2) is generally modied into a normalized

form in which the maximum principal tensile stress is
normalized by the equivalent stress to give:16)
Z "f
T
d "
3
CN

0
Fig. 1 Predicted local strain distributions within the billets for strain rate
sensitivities of (a) 0, (b) 0.2 and (c) 0.4.14)
and (c) 0.4. This modeling shows directly that the billet with
the highest strain rate sensitivity of m 0:4 exhibits the
highest strain level of 1.6 and this occurs in a relatively
narrow region near the bottom surface of the billet. By
contrast, the distribution in the local strain is reasonably
uniform across the width of the billet when m 0 although
the strain level is lower near the bottom surface, increases
rapidly to 1:0 within a very short region and thereafter
remains constant within the central region and the top region
of the billet. Inspection shows that the shapes of the strain
level contours are dierent in the front regions of the billets
for materials with high and low strain rate sensitivities. Thus,
whereas the simulation with m 0:4 in Fig. 1(c) gives strain
contours perpendicular to the billet axis in the front section
of the billet with the contours corresponding to strain levels
of 0.8 and 1.0 displaying a marked curvature towards the
central part of the billet, the simulation with m 0 shows
contours lying at approximately 45 to the billet axis in the
front section and with a small pocket having the highest strain
of 1.12 near the front face.
An important factor that governs the ability to process
dicult-to-work alloys is the occurrence of cracking within
the billet during the pressing operation. This may be
evaluated by employing a macroscopic damage criterion to
evaluate the formability of materials during the FEM
simulations. To understand this approach, it is convenient
to make use of the traditional Cockcroft and Latham15)
damage model which is based on the following relationship:
Z "f
T d "
2
C
0
where C is a constant corresponding to a critical condition

associated with fracture, " is the eective strain, T is the
maximum component of the tensile stress at any point within
the billet and the integral is evaluated from zero strain to the
nal eective strain, "f . The integral shown in eq. (2)
describes the tensile energy per unit volume and this
relationship demonstrates that fracture will occur when the
value of C reaches a critical value. For convenience, the
where is the eective stress.

The approach with the Cockcroft-Latham damage criterion
employs the concept that shearing occurs within the billet
when it passes through the theoretical shear plane at the point
of intersection of the two parts of the channel. This is
demonstrated in Fig. 2 where the three columns relate to
channel angles, , of 90 , 110 and 135 and the three rows
correspond to strain rate sensitivities of (a) 0, (b) 0.2 and (c)
0.4. Within each set of illustrations, the contour lines
delineate consecutive points experiencing constant levels of
strains rate where the predicted strain rates are marked on the
diagrams in s1 . An important conclusion from Fig. 2 is that
the deformation zones and the strain rates depend both upon
the channel angle within the die and the value of the strain
rate sensitivity. Thus, the deformation zones are wide when
pressing a material with a high strain rate sensitivity and
these widths are reduced when the value of m is decreased.
Furthermore, the strain rates at all angles are lowest in the
billets with m 0:4 but become higher as the strain rate
sensitivity is reduced.
The normalized Cockcroft-Latham damage distributions
are shown in Fig. 3 based on simulations involving strain
rate sensitivities of (a) 0, (b) 0.2 and (c) 0.4 for the three
dierent die angles of 90 , 110 and 135 . The contour lines
in Fig. 3 correspond to the values predicted for CN using
eq. (3) and they represent the local level of damage occurring
in each billet. It is readily apparent from Fig. 3 that the strain
rate sensitivity aects not only the precise level of damage
occurring within the billet but also the distribution of this
damage. For a material having m 0 as shown in Fig. 3(a),
there is a constant region of damage along the billet and the
highest levels of damage, with values of CN 0:28{0:32,
occur between the central regions and the lower edges of the
billets. Figures 3(b) and (c) show there are higher damage
levels when the strain rate sensitivity is increased and also the
region of maximum damage is displaced from the middle of
the billet when m 0 to the top surface when m 0:4. The
damage levels for the billet with m 0:4 are high and up to a
maximum of 0:49 for the die with a channel angle of
90 . Similar damage levels are also visible on the top
surface for the die when the channel angle is 110
although the damage is signicantly lower when 135 .
It follows from these and other similar calculations that the
overall level of damage is reduced when the channel angle is
increased. This trend is clearly shown by plotting the
maximum value of damage, quantied by the values of CN
along the middle cross-sections of the billets, as a function of
the channel angle within the die as given in Fig. 4.
A direct experimental verication of the predictions of the
FEM simulation is shown in Fig. 5 where two billets of the
ZK60 alloy were annealed to give an initial strain rate
sensitivity of m 0:2 and then processed by ECAP. The side
proles of the billets are shown in Fig. 5 where both billets
were pressed through one pass at a temperature of 473 K
Achieving Microstructural Renement in Magnesium Alloys through Severe Plastic Deformation
113
Fig. 2 Predicted strain rates within the deformation zones for channel angles of 90 , 110 and 135 using strain rate sensitivities of (a) 0,
(b) 0.2 and (c) 0.4.14)
using dies having channel angles of either 90 for the upper
billet or 110 for the lower billet. It is apparent that the upper
billet exhibits very regular segmentation which occurs
throughout the length of the billet and was initiated at the
upper surface of the billet. By contrast, the lower billet
demonstrates a successful pressing without the introduction
of any signicant damage when using a die with a channel
angle of 110 . These results are therefore consistent
with the FEM predictions and they provide strong support for
using this type of approach in evaluating the optimum
conditions for achieving successful processing and signicant grain renement by ECAP.
3.
Examples of Advanced Properties Achieved in Magnesium Alloys Processed by ECAP
Experiments were conducted on a Mg-0.8 mass% Li twophase alloy to evaluate the potential for achieving superplastic properties after ECAP.17) Initial experiments showed
that the alloy cracked after a single pass when pressing at
room temperature using a die with a channel angle of
90 . Accordingly, the channel angle was increased to

135 where it was possible to press through ten passes at
room temperature without experiencing any visible cracking.
For an ECAP die with a channel angle of 135 and an
arc of curvature of 20 , it can be shown that the imposed
strain is equal to 0:5 on each separate pass through the
ECAP die18) where this is one-half of the strain generally
imposed in a single pass when using a conventional die with
a channel angle of 90 .
An example of the experimental results with the Mg-8% Li
alloy are shown in Fig. 6 where the measured elongation to
failure is plotted against the initial strain rate for tests
conducted under conditions of a constant rate of cross-head
displacement. These results demonstrate the very signicant
dierences that may be recorded in the same alloy when
processing in dierent ways. Thus, the cast alloy having an
average phase width of 60{70 mm exhibits only low
elongations up to a maximum of 200%. By contrast, the
alloy which was cast and extruded using an extrusion
temperature, Tex , of 373 K and an extrusion speed, rex , of
1 mm s1 contained a microstructure with an average phase
114
Fig. 3 Predicted distributions of the damage factor, CN , for channel angles of 90 , 110 and 135 using strain rate sensitivities of (a) 0, (b)
0.2 and (c) 0.4.14)
Fig. 4 Peak values of CN across the centers of the billets as a function of

the channel angle for strain rate sensitivities of 0 and 0.4.14)
size of 3{5 mm and in this condition the maximum

elongation was 630% at the lowest strain rate of
1:5 104 s1 . Finally, the cast alloy which was extruded
and then processed by ECAP through four passes using an
ECAP die with a channel angle of 135 , corresponding to an
imposed strain of 2, contained an average phase size of
1{3 mm and in this condition a maximum elongation of
1780% was achieved at the slowest experimental strain
rate. These results demonstrate the inuence of additional
grain renement in promoting higher elongations at the
Fig. 5 Massive billet segmentation (upper) and successful pressing of a

billet (lower) using a ZK60 alloy in an annealed condition with m 0:2
pressed through one pass at 473 K using dies with channel angles of 90
and 110 for the upper and lower billets, respectively.14)
lowest strain rates. Nevertheless, it is also apparent from

Fig. 6 that all processing conditions lead to similar low
elongations when testing to failure at strain rates above
102 s1 , where this corresponds to the conventional
transition to a ow region where an intragranular dislocation
creep mechanism is dominant.19,20)
The results documented in Fig. 6 show that exceptional
superplastic properties may be achieved at elevated temperatures even in a magnesium-based alloy. Furthermore, the
Achieving Microstructural Renement in Magnesium Alloys through Severe Plastic Deformation
Fig. 6 Variation of the elongation to failure with the initial strain rate at a
testing temperature of 473 K for samples of a Mg-8% Li alloy in the Cast,
Cast + Extrude and Cast + Extrude + ECAP conditions: the extrusion
was conducted at 373 K using an extrusion speed of 1 mm s1 .17)
Fig. 7 Appearance of specimens of the ZK60 alloy tested to failure at

473 K with an initial strain rate of 1:0 104 s1 after processing by
ECAP for dierent numbers of passes: the upper specimen is untested.22)
recorded maximum elongation in Fig. 6 is very high even by

comparison with the many other reports of superplasticity in
aluminum and other alloys processed by ECAP: a complete
tabulation of these various results is given elsewhere.21)
An example of an exceptionally high superplastic ductility
is shown in Fig. 7 for a ZK60 alloy containing, in mass%,
Mg-5.5% Zn-0.5% Zr.22) This alloy was received in an
extruded condition with an initial grain size of 2:9 mm
and it was then processed by ECAP at 473 K using a die
with a channel angle of 90 and an arc of curvature of
20 using processing route BC in which the billets are
rotated by 90 in the same sense between each consecutive
pass.23) The tensile tests were conducted at the same temperature of 473 K using an initial strain rate, "_, of 1:0
104 s1 .
In Fig. 7 the upper specimen is untested and the other
specimens were processed through dierent numbers of
passes from 1 to 6 passes and then pulled to failure. It is
readily apparent that all of these specimens exhibit elongations larger than 500% which corresponds to the onset of
true superplasticity. Furthermore, the specimen processed
through 2 passes exhibits an exceptionally high superplastic
115
elongation of 3050%. This elongation is the highest recorded

in any magnesium alloy processed under any conditions with
or without ECAP. In addition, a comparison with the detailed
tabulation of superplastic data presented elsewhere21) shows
this ductility is also the highest recorded in any of the many
alloys processed by ECAP to date including the highly
ductile aluminum-based alloys.
An apparently unusual result in Fig. 7 is that the maximum
superplastic elongation is achieved in this alloy after
processing through only 2 passes. After 3 and 4 passes the
ductilities remain unusually high but they are lower than in
the specimen processed through only 2 passes. This result
tends to be unexpected because earlier tests on an aluminum
alloy demonstrated that higher ductilities are generally
achieved after larger numbers of passes through the ECAP
die.24) It is appropriate, therefore, to examine the separate
ow characteristics of the samples taken through 2 and 6
passes to give elongations of 3050% and 930%, respectively.
The results may be understood by examining the role of
strain hardening which controls the overall elongations to
failure achieved at this relatively low testing temperature.
Strain hardening serves a dual purpose under these experimental conditions because it both balances the reduction in
the cross-sectional area and stabilizes the load during the
tensile ow. Figure 8 shows the plots of true stress versus
true strain for the two specimens taken through N 2 and 6
passes, where N represents the number of passes in ECAP.
Following the early analysis by Hart,25) it is assumed that
plastic instability occurs at the point at which the rate of
variation of the strain rate at the smallest cross-section of the
specimen rst starts to increase. This condition may be
written in explicit form as:

d=d"
1 m
where and " are the true stress and true strain, respectively.
In Fig. 8, the parameter on the right of the Hart relationship in eq. (4) is plotted on the right-hand axis using the
experimentally determined values for the strain rate sensitivity of m 0:44 after 2 passes and m 0:40 after 6 passes.
The point of instability is then given by the point where these
Fig. 8 True stress and the rate of strain hardening normalized by (1 m) as

a function of the true strain for two samples of the ZK60 alloy processed
by ECAP through 2 and 6 passes: the intersection of the two curves marks
the onset of plastic instability.22)
116
two curves intersect and these points are delineated by the

two vertical lines. It is apparent that the higher ow stress and
the lower value of m leads to a relatively low strain for the
onset of plastic instability in the specimen processed through
6 passes. By contrast, the higher strain rate sensitivity and
lower ow stress in the specimen processed through 2 passes
leads to a higher strain for plastic instability in this specimen.
After the onset of plastic instability, there is also a larger
accrued strain in the specimen processed through 2 passes
and this is a direct consequence of the higher strain rate
sensitivity in this specimen. Thus, the onset of plastic
instability is consistent with the Hart criterion25) given by
eq. (4).
optimized by using nite element modeling to delineate the

strain distributions and the internal damage introduced during
the processing operation. Using this approach, it is possible to
achieve excellent properties after successful processing by
ECAP.
Acknowledgements
One of the authors (RBF) was supported by a CAPES/
Fulbright Scholarship. This work was supported by the U.S.
Army Research Oce under Grant No. W911NF-08-1-0201.
REFERENCES
4.
Discussion
The successful processing of magnesium alloys by ECAP

requires a careful consideration of the experimental conditions. A step of extrusion prior to ECAP is benecial in
reducing the grain size to a level that is more amenable to
easy renement to produce a true ultrane-grained microstructure. But even for extruded magnesium alloys, the
occurrence of cracking and segmentation during the processing operation may hinder the ability to achieve successful
results. These problems may be overcome by using nite
element modeling to evaluate the local strain distributions
and the inherent strain rates involved in the processing
operation. The incorporation of an appropriate damage factor
such as the Cockcroft-Latham criterion15) permits the development of criteria that may be used to successfully process
many dicult-to-work alloys. For example, although the
present paper deals exclusively with magnesium-based
alloys, an increase in the channel angle within the ECAP
die has been used to successfully press other hard materials
such as tungsten26) and titanium.27)
When magnesium-based alloys are processed successfully
by ECAP, mechanical testing may produce evidence for
exceptional properties that are not usually achieved in
magnesium alloys. Two examples are shown in Figs. 6 and
7 where extremely good tensile ductilities are evident in tests
conducted at elevated temperatures. Furthermore, it is known
that improved results are often achieved by ECAP when
using an experimental facility which imposes a back-pressure
during the processing operation,28) but the present results,
including the record tensile elongation shown in Fig. 7, were
attained without the imposition of a back-pressure.
5.
(1) There are unusual diculties associated with the

processing of magnesium alloys by equal-channel angular
pressing (ECAP) because of the relatively easy development
of cracking and/or massive segmentation of the billets.
(2) It is shown that the processing procedure may be

45 (2000) 103189.
2) R. Z. Valiev and T. G. Langdon: Prog. Mater. Sci. 51, (2006) 881981.
3) A. P. Zhilyaev and T. G. Langdon: Prog. Mater. Sci. 53 (2008) 893
979.
4) R. F. Decker: Adv. Mater. Proc. 154 (9) (1998) 3133.
5) F. H. Froes, D. Eliezer and E. Aghion: JOM 50 (9) (1998) 3034.
6) K. Johnson: Adv. Mater. Proc. 160 (6) (2002) 6265.
7) H. Friedrich and S. Schumann: J. Mater. Process. Tech. 117 (2001)
276281.
8) A. Yamashita, Z. Horita and T. G. Langdon: Mater. Sci. Eng. A 300
(2001) 142147.
9) Z. Horita, K. Matsubara, K. Makii and T. G. Langdon: Scr. Mater. 47
(2002) 255260.
10) K. Matsubara, Y. Miyahara, Z. Horita and T. G. Langdon: Acta Mater.
51 (2003) 30733084.
11) K. Matsubara, Y. Miyahara, Z. Horita and T. G. Langdon: Metall.
Mater. Trans. 35A (2004) 17351744.
12) H. K. Lin, J. C. Huang and T. G. Langdon: Mater. Sci. Eng. A 402
(2005) 250257.
13) Y. Miyahara, Z. Horita and T. G. Langdon: Mater. Sci. Eng. A 420
(2006) 240244.
14) R. B. Figueiredo, P. R. Cetlin and T. G. Langdon: Acta Mater. 55
(2007) 47694779.
15) M. G. Cockcroft and D. J. Latham: J. Inst. Metals 96 (1968) 3339.
16) S. I. Oh, C. C. Chen and S. Kobayashi: Trans. ASME J. Eng. Industry
101 (1979) 3644.
17) M. Furui, H. Kitamura, H. Anada and T. G. Langdon: Acta Mater.
55 (2007) 10831091.
Mater. 35 (1996) 143146.
19) T. G. Langdon: Metall. Mater. Trans. 33A (2002) 249259.
20) T. G. Langdon: Mater. Trans. 46 (2005) 19511956.
21) M. Kawasaki and T. G. Langdon: J. Mater. Sci. 42 (2007) 17821796.
22) R. B. Figueiredo and T. G. Langdon: Adv. Eng. Mater. 10 (2008)
3740.
23) M. Furukawa, Y. Iwahashi, Z. Horita, M. Nemoto and T. G. Langdon:
Mater. Sci. Eng. A 257 (1998) 328332.
24) S. Komura, M. Furukawa, Z. Horita, M. Nemoto and T. G. Langdon:
Mater. Sci. Eng. A 297 (2001) 111118.
25) E. W. Hart: Acta Metall. 15 (1967) 351355.
26) I. V. Aleksandrov, G. I. Raab, L. O. Shestakova, A. R. Kilmametov
and R. Z. Valiev: Phys. Metals Metallog. 93 (2002) 493500.
27) X. Zhao, W. Fu, X. Yang and T. G. Langdon: Scr. Mater. 59 (2008)
542545.
28) R. Lapovok, Y. Estrin, M. V. Popov and T. G. Langdon: Adv. Eng.
Mater. 10 (2008) 429433.

Microstructural Evolution of Ti-Mo-Ni-C Powder by Mechanical Alloying

Hiroyuki Hosokawa, Kiyotaka Kato, Koji Shimojima and Akihiro Matsumoto
Materials Research Institute for Sustainable Development, National Institute of Advanced Industrial Science and Technology,
Nagoya 463-8560, Japan
The microstructural evolution of (Ti,Mo)C-Ni powder by mechanical alloying of pure titanium, nickel, carbon and molybdenum as starting
powder with the composition of TiC-20 Mo2 C-20 Ni in mass% were investigated by X-ray diraction (XRD) using CuK radiation, eld
emission scanning electron microscopy (FE-SEM) with energy-dispersive X-ray spectroscopy (EDS), and transmission electron microscopy
(TEM) with EDS. The powders were mixed gradually and the size of the mechanically alloyed particles was of sub-micrometer order. The TiC
structure, which indicated smaller lattice constant than TiC, clearly appeared after 100 h milling time. The particles consisted of ne grains of
about 5 nm size. The nickel was transformed from crystallized phase into an amorphous phase by mechanical alloying. The molybdenum was
dissolved into TiC structure and Mo/Ti ratio in (Ti,Mo)C was about 3/7. [doi:10.2320/matertrans.MRA2008280]
Keywords: TiC, mechanical alloying, molybdenum, microstructure
Introduction
TiC based cermets have attracted a great deal of attention

because of their high strength, high hardness, and their good
chemical and thermal stability. Finer and modied hard
phases as well as modied binder phases have been main
focus of attention among the researchers, in order to improve
toughness and wear resistance of their materials.111) As far
as renement of hard phase is concerned, mechanical
alloying is one of the eective methods to produce the ner
TiC grain size than the conventional method.11,12)
It is well known that one of the main disadvantage of TiC
based cermets is wettability between TiC and binder
materials. Molybdenum can be added in order to improve
this, because molybdenum increases the wetting of TiC by Ni
binder and ner hard phase.13) There have been reported
about TiC based cermet containing Mo and Mo2 C.13,58,10,12)
Focusing on microstructures in TiC based cermet containing Mo and Mo2 C, in general, they have a core-rim structure,
that is, a (Ti,Mo)C rim surrounding a TiC core.13) However,
Kim et al. reported that nano-particle (Ti,Mo)C-Ni cermets
without a core-rim structure have been synthesized by
mechanical alloying, having high hardness of 92 HRA .12)
To date, to the authors knowledge, there is no detailed
information about the microstructural evolution of Ti-MoNi-C into (Ti,Mo)C-Ni powder by ball milling. The purpose
of this work is to investigate the change of morphologies,
chemical components, and microstructures for Ti-Mo-Ni-C
powder into (Ti,Mo)C-Ni in order to better understand it.
2.
The powders used in this work were 99.9%Ti with 45 mm,

99.9% C with 20 mm, >99:9% Ni with 5 mm, and >99:9% Mo
with 3 mm. A powder mixture having a composition of TiC20 Mo2 C-20 Ni in mass% was prepared. The powders were
dry mixed together with WC/Co balls into a 500 ml hardened
steel pot with an O-ring seal under 667 KPa of Ar atmosphere. A planetary ball mill (Fritsch Pulverisette 5) was
used for durations of up to 700 h. The ball-to-powder weight
ratio (BPR) was xed at 20 : 1.
The powders were milled for various times, and were then
analyzed to determine the change of the phase compositions
and morphologies by means of X-ray diraction using
CuK radiation, eld emission scanning electron microscopy (FE-SEM) with energy-dispersive X-ray spectroscopy
(EDS), and transmission electron microscopy (TEM) with
EDS. The crystallite size of the TiC structure for powder after
700 h milling was estimated, based on the half-width of the
X-ray diraction pattern of the specimen using Scherrers
formula. The powders for FE-SEM with EDS observation
were embedded in polymer and Cu powders for conductivity
and then were mechanically polished. The lattice constants
for the powder and the sintered compact were estimated by
means of selected area diraction pattern (SADP).
The as-synthesized powders after 700 h milling time were
pressed into pellets under 100 MPa to pellets. These green
pellets were then sintered at 1723 K for 1 h under vacuum.
The sintered compact was also analyzed by X-ray diraction,
FE-SEM and TEM-EDS.
3.
The XRD patterns of the milled powders for dierent

times are shown in Fig. 1. Peaks of crystalline titanium,
Ti
C
Ni
Mo
TiC
1h
Intensity (a.u.)
1.
8h
24 h
48 h
100 h
200 h
700 h
20
30
40
50
60
70
80
90
2 (degree)
Fig. 1 XRD patterns for the milled Ti-Mo-Ni-C powders with dierent
milling times.
118
H. Hosokawa, K. Kato, K. Shimojima and A. Matsumoto
(a)
(d)
100 m
20 m
(b)
(e)
50 m
20 m
(f)
(c)
20 m
20 m
Fig. 2 SEM photographs of the powders for (a) 1 h milling time, (b) 8 h milling time, (c) 48 h milling time, (d) 100 h milling time,
(e) 200 h milling time and (f) 700 h milling time. The framed parts by the dashed lines are the regions for EDS mapping images in Fig. 4.
carbon, molybdenum and nickel clearly existed and were

sharp after 1 h milling. However, the peaks gradually
became broader with increasing milling time, indicating
that the grains became ner with an increase in milling time.
The carbon peak and the nickel peaks disappeared after 8 h
and 100 h milling time, respectively. After 200 h milling,
a new crystalline phase having a TiC structure clearly
appeared. It is known that the milling of Ti and C
synthesizes TiC through ball milling.11) Therefore, it is
considered that these powders were reacted as a TiC
structure. The peaks of the TiC structure in this work
showed a trend toward higher angles, compared with TiC
peaks. This trend is the same as (Ti, Mo)C reported by Kim
et al.12) After 700 h milling time, the Ti and Mo peaks nally
disappeared. The peaks of the TiC structure were broad after
700 h milling time, indicating that a ne crystallite size for
TiC structure is achieved by ball milling. The grain size for
it was estimated to be about 4 nm, applying Scherrers
formula to half-width.
SEM photographs of the powders for various milling times

are shown in Figs. 2 and 3. While powders agglutinated after
8 h milling time, the boundaries between the starting powders
were clear. After 48 h milling time or longer, some parts of
the aggregates had insucient milled regions, and other parts
had sub-micrometer particles (see Fig. 3(a)). The ratio of the
regions consisting of particles with sub-micrometers increased along with an increase in milling time. After 200 h milling,
almost all regions consisted of sub-micrometer particles, with
a few white regions having particles of several micrometers
(see Fig. 3(b)). After 700 h milling time, almost all particles
were sub-micrometers (see Fig. 3(c)), despite the 4 nm or so
as calculated by Scherrers formula. It is speculated that the
particle consists of nm-scale grains.
EDS mapping images of the powders for various milling
times are shown in Fig. 4. There were clearly the regions for
each element in aggregate after 1 h milling time. In particular,
the Ti region was larger due to a larger starting powder
diameter. The high concentrated regions of each element
(a)
10 m
(b)
5 m
(c)
2 m
Fig. 3 SEM photograph of the powders for (a) 100 h milling time, (b) 200 h
milling time and (c) 700 h milling time.
decreased with an increase in milling time. Those for

titanium, molybdenum, nickel and carbon disappeared after
700, 700, 100, and 8 h milling time, respectively.
The results of a TEM-EDS analysis of the powder after
700 h milling time are shown in Fig. 5. It is noted that ne
grains, of about 5 nm, were in the particles (from Fig. 5(a)).
This size is in close agreement with that as calculated by
Scherrers formula. The diraction pattern of the particle
119
was a typical fcc Debye ring of polycrystalline structure

(Fig. 5(b)). As a consequence of the diraction pattern, the
lattice constant was estimated to be 0.4315 nm. This value is
smaller than that for TiC (0.43274 nm), but close. Therefore,
the particles denitely had a TiC structure. It is conrmed
from the result of EDS (Fig. 5(c)) that the compositional ratio
for the particles was approximately the same as that for the
starting powder. Although the particles contained molybdenum and nickel, there was no diraction pattern, except for a
TiC structure. It may be considered that these elements
dissolved into a TiC structure and/or transformed into an
amorphous phase. It has been reported that amorphization has
been found in Ni-Ti systems by mechanical alloying.14) It is
also reported that a Ni solid solution has been transformed
from a crystalline phase into an amorphous phase in Ti-C-Ni
by ball milling.11) As mentioned above, there were no peaks
for Ni by XRD. It may be concluded that the Ni solid solution
transformed the crystalline phase into an amorphous phase
in Ti-Mo-Ni-C during ball milling.
The SEM photograph of the sintered compact is shown in
Fig. 6. A narrow grain size distribution was obtained and the
average grain size is about 500 nm. It may be considered that
these are attributed to nano-scale grain size of sucient
mixed powder. It is known that the core-rim structure of the
hard phase shows a contrast between core and rim in SEM
observation,13) however, there was no contrast for the hard
phase. It may be indicated that it consisted of a single phase.
The XRD patterns of the sintered compact and the milled
powders are shown in Fig. 7. The peaks of nickel and TiC
structure existed for the sintered compact. This may indicate
that the nickel based amorphous phase was transformed into
crystallization during the sintering process. The angles for the
nickel peaks trended toward a lower angle. It is known that a
Ni binder in TiC system cermets containing Mo has some
amount of Ti and Mo.15) Therefore, it is presumed that some
Mo and Ti caused the shift of the nickel peak in this work.
On the other hand, it is noted that the peak positions of
TiC structure for the powder is almost the same as that for
sintered compact, presuming that the chemical compositions
of them are almost the same.
The results of TEM/EDS analysis of the sintered compact
are shown in Fig. 8. As a consequence of the diraction
pattern, the lattice constant is 0.4315 nm. This value is almost
the same as that of powder, but smaller than the lattice
constant of TiC (0.43274 nm). From Fig. 8(c), the Mo
content is about 33 mass% for the center and the edge of
the TiC structure grain. In addition, it is shown from Fig. 7
that the peak positions of TiC structure were the same for the
powder and the sintered compact. Therefore, it is noted that
the solution of molybdenum caused the dierence of lattice
for a TiC structure lower than that for TiC, and the solution
of molybdenum in TiC structure was caused by ball milling
in powder, and the Mo/Ti ratio is about 3/7 in this work.
4.
Conclusions
The microstructural evolution of (Ti,Mo)C-Ni powder by

mechanical alloying pure titanium, nickel, carbon and
molybdenum as a starting powder with a composition of
TiC-20 Mo2 C-20 Ni in mass% were investigated by X-ray
120
Mo
Ti
hr
Ni
1
20 m
20 m
20 m
20 m
50 m
50 m
50 m
50 m
20 m
20 m
20 m
20 m
50 m
50 m
50 m
50 m
20 m
20 m
20 m
20 m
20 m
20 m
20 m
20 m
48
100
200
700
Fig. 4
EDS mapping images of the powders for various milling times.
diraction (XRD) using CuK radiation, eld emission

scanning electron microscopy (FE-SEM) with energy-dispersive X-ray spectroscopy (EDS) and transmission electron
microscopy (TEM) with EDS in detail. The results were as
follows.
(1) The titanium, molybdenum, nickel and carbon peaks
gradually became weaker with an increase in milling
time, and disappeared after 200, 700, 100, and 8 h
milling time, respectively. A TiC structure, which

indicated a lower lattice constant than that for TiC,
appeared after 100 h milling time and there were only
the peaks for the TiC structure after 700 h milling time.
(2) Sub-micrometers particles were produced by ball milling, and their ratio increased with an increase in milling
time. The size of almost all particles was sub-micrometers after 700 h milling time.
121
(a)
Intensity (a.u.)
Ni
TiC
Sintered compact
Powder
30
40
50
2 (degree)
Fig. 7
XRD patterns for the sintered compact and the 700 h milled powder.
(a)
(b)
(c)
Ti
Mo
Ni
53.1 23.1 23.8

1
Fig. 5 TEM/EDS analysis of the powders for 700 h milling time. (a)
bright-eld image, (b) the selected area diraction pattern and (c) EDS
analysis (in weight). The circle with broken line indicate the area of (b)
and (c).
(b)
(c)
Ti
1
2
2 m
Fig. 6 SEM photograph of the sintered compact at 1723 K for 1 h by 700 h
milling time.
(3) The high concentrated regions of each element in

aggregate analyzed by EDS mapping decreased with
increased milling time, and dissapeared after 700 h
milling time for Ti, 700 h milling time for Mo, 100 h
milling time for Ni, and 8 h milling time for C.
(4) The particles consisted of ne grains with a size of
about 5 nm. There was only TiC structure in SADP,
Mo
66.4 33.6
67.1 32.9
Fig. 8 TEM/EDS analysis of the sintered body. (a) bright-eld image, (b)
the selected area diraction pattern for (001) plane and (c) EDS analysis
(in weight). The circles with broken line and the solid line indicate the
areas of (b) and (c), respectively.
although the chemical composition of the powder is

approximately the same as the starting chemical
component.
(5) When the lattice constant and chemical composition for
the sintered compact is compared with those for the
powder, the molybdenum is dissolved into the TiC
structure and the Mo/Ti ratio in (Ti,Mo)C was about
3/7.
122
Acknowledgement
The authors gratefully acknowledge the nancial support
by Project (P08023) of the New Energy and Industrial
Technology Development Organization (NEDO).
REFERENCES
1) M. Humenik Jr and N. M. Parikh: J. Am. Ceram. Soc. 39 (1956) 6063.
2) N. M. Parikh and M. Humenik Jr.: J. Am. Ceram. Soc. 40 (1957) 315
320.
3) N. M. Parikh: J. Am. Ceram. Soc. 40 (1957) 335339.
4) P. Ettmayer, H. Kolaska, W. Lengauer and K. Dreyer: Int. J. Refractory
Metal & Hard Materials 13 (1995) 343351.
5) D. Mari, S. Bolognini, G. Feusier, T. Cutard, C. Verdon, T. Viatte and
W. Benoit: Int. J. Refractory Metals & Hard Materials 21 (2003) 3746.
6) D. Mari, S. Bolognini, G. Feusier, T. Cutard, T. Viatte and W. Benoit:

Int. J. Refractory Metals & Hard Materials 21 (2003) 4753.
7) N. Liu, Y. Xu, Z. Li, M. Chen, G.i Li and L. Zhang: Ceram. Int. 29
(2003) 919925.
8) P. Feng, W. Xiong, L. Yu, Y. Zheng and Y. Xia: Int. J. Refractory
Metals & Hard Materials 22 (2004) 133138.
9) J. Jung and S. Kang: Acta Mater. 52 (2004) 13791386.
10) J. C. LaSalvia, D. K. Kim and M. A. Meyers: Mater. Sci. Eng. A 206
(1996) 7180.
11) C.-J. Choi: J. Mater. Proc. Technol. 104 (2000) 127132.
12) Y. K. Kim, J.-H. Shim, Y. W. Cho, H.-S. Yang and J.-K. Park: Int. J.
Refractory Metals & Hard Materials 22 (2004) 193196.
13) H. Suzuki, K. Hayashi and H. Matsubara: J. Japan Inst. Metals 22
(1983) 312319 (in Japanese).
14) M. X. Quan, K. Y. Wang, T. D. Shen and J. T. Wang: J. Alloy. Compd.
194 (1993) 325330.
15) H. Suzuki, K. Hayashi, H. Matsubara and K. Tokumoto: J. Jpn. Soc.
Powder Powder Metall. 30 (1983) 106. (in Japanese)

Kinetics and Morphology of Isothermal Transformations

at Intermediate Temperature in 15CrMnMoV Steel
Mokuang Kang1 , Ming-Xing Zhang2; * , Feng Liu1 and Ming Zhu1
1
2
School of Materials Science, Northwestern Polytechnical University, Xian, Shaanxi, 710072, P.R.China
Division of Materials, School of Engineering, The University of Queensland, St Lucia, Brisbane, Qld 4072, Australia
Upper bainite (BU ), lower bainite (BL ) and granular structure (GS) are three distinguishable microstructures formed through isothermal
treatment at temperature range from 350 to 520 C in 15CrMnMoV steel. Although long time isothermal holding leads to the mixed
microstructures, single microstructure can be obtained in the initial stage of isothermal transformation when the total transformation fraction is
less than 20%. The Arrhenius equations were used to determine the overall activation energies of the transformations in the steel with accuracy
of 8 kJ/mole. To ensure the determined overall activation energy is the true value for single microstructure rather than for a mixture of two or
more microstructures, the present determination of the kinetic data of isothermal transformations was carried out in terms of the isothermally
transformed fractions of 1 and 5% at various temperatures. Experimental results show that the overall activation energy for single microstructure
is signicantly dierent from that for mixed microstructure, which was measured in terms of 50% transformation. Upon the determined kinetic
data and the morphology of the microstructures, together with consideration of the reported atom diusion activation energies, following
conclusions are achieved. In steels, the BU transformation is controlled by carbon diusion in the austenite; while, the formation of BL is only
partially governed by carbon diusion in austenite. As a dierent microstructure from other two, the GS transformation is controlled by carbon
diusion and has the common features of both equiaxed ferrite and massive ferrite transformations. These results, can be used to further
understand the formation mechanism of BU , BL and GS in the 15CrMnMoV steel. [doi:10.2320/matertrans.MRA2008194]
(Received June 23, 2008; Accepted October 20, 2008; Published December 25, 2008)
Keywords: bainite, 15CrMnMoV steel, arrhenius equations, activation energy
1.
Introduction
Since Bain1) rst reported bainite microstructure in the

1930s, its properties, morphology and transformation mechanism have been considerably studied in the past 70 years.
One of the milestones in the history of bainite study was the
establishment of the shear model of bainitic transformation in
the 1950s by Ko and co-workers.2) As typical kinetic data of
phase transformations in solid, the overall activation energy
(OAE) has been widely used to understand the characteristics
of bainitic transformations. However, in most cases, the
reported experimental results are inconsistent with the
calculated overall activation energy. The major reason for
this discrepancy is that the previously published OAE data
was experimentally determined for a mixture of two or more
dierent microstructures due to the diculty to separat them,
rather than for single microstructure. Considerable experimental work3,4) has shown that isothermal treatment of low
carbon low alloy steels at intermediate temperature ranging
from 350 to 550 C produces upper bainite (BU ), lower bainite
(BL ), granular structure (GS) or their mixture. These three
microstructures are not only morphologically dierent but
also can be characterized by their distinguishable C-curves
on the TTT diagram,3,4) even though the GS, in most
publications, is regarded as granular bainite or BU . Normally,
in initial stage of the transformation at specied temperatures, single microstructure can be obtained when the totally
transformed fraction is less than 20%, whereas long time
isothermal holding leads to the formation of mixed microstructure. Hence, the aims of the present work are: (1) to
experimentally determine the OAE for single BU , BL and GS
in 15CrMnMoV steel in terms of the total isothermally
*Corresponding
author, E-mail: Mingxing.Zhang@uq.edu.au
transformed fractions of 1 and 5% (only single microstructure is obtained) using the Arrhenius equations; and (2)
to compare these OAE values determined with those
measured for 50% transformation, at which mixed microstructure is produced. The OAE data of phase transformations, together with the morphologies of microstructures and
the atomic diusion activation energies in steels, can be used
to further understand the formation mechanism of BU , BL and
GS in low carbon low alloy steels.
1.1 Arrhenius equations
The Arrhenius equation has two versions. One describes
the transformation OAE as a function of time and temperature; another relates the OAE to transformation rate and
temperature. The two versions of Arrhenius equation are as
follows:
Version 1OAE is a function of time and temperature:
2
3
@ ln t
1
QX1 R6 7
4 1 5
@
T
X
1
Version 2relationship between OAE and transformation rate

and temperature:
2 3
@X
6 @ ln @t 7
6
7
2
QX1 R6 7
4
5
1
@
T
X1
Where QX1 is the OAE, R the gas constant, t the isothermal
transformation time, T the transformation temperature and X
the transformed fraction. For a specied fraction of transformation, the slope of the straight line that graphically
124
M. Kang, M.-X. Zhang, F. Liu and M. Zhu

Table 1
The overall activation energy of bainite transformation in steels.

Marked
change
temperature
Steels
QBU
QBL
Equation,
Fraction
(K) from
kJ/mol
Comments
Ref.
used
Bu to BL
Steel No
mass%C
mass%Mn
0.77 mass%C steel
623
77
35
eq. (1),
1.13 mass%C steel
613
127
52
X 8%
80121
5063
113
87
60
105
eq. (1),
X 50%
83125
5967
eq. (2),
Steel No. 1, 2, 3
Steel No. 4
0.81
1.00
1.09
1.00
1.10
1.49
1.42
573603
2.37
60
130
X 50%
80
X 1%
Low carbon Cr-Mo-V steel
140
40
0.4 mass%C, 1.73 mass%Cr, 1.03 mass%Mn,

1.21 mass%Si, 0.50 mass%Mo, 0.1 mass%V
describes the relationships between ln t and 1=T or between

ln@X=@t and 1=T is proportional to QX1 . The OAE at the
transformed fraction X can be calculated by multiplying the
gas constant. Although Hillert5) suggested that QX1 from
eq. (2) is more accurate, eq. (1) is usually used due to its
simplicity.
Overall activation energy of bainitic transformation
in steels
Previous work has shown that the OAE of bainitic
transformation is associated with the carbon content of
steels. In the late of 1950s, using eq. (1) Vasudevan et al.6)
determined the OAE, QBU and QBL (as listed in Table 1) for
BU and BL , respectively, based on the transformed fraction
of 50% in a plain carbon steel containing 0.97 mass%C. The
result indicates that QBU is greater than QBL . Because QBU is
very close to the value of diusion activation energy of
carbon in -Fe (130 kJ/mol), Vasudevan considered that the
transformation of upper bainite is controlled by carbon
diusion in -Fe. Similarly, as the determined QBL value is
close to the diusion activation energy of carbon in -Fe
(70 kJ/mol), it was considered that the lower bainite transformation relies on carbon diusion in -Fe. However, this
hypothesis could not explain the subsequent experimental
results that show much greater OAE values than the carbon
diusion activation energy in both -Fe and -Fe. Radclie
and co-workers7) determined the OAE for both upper and
lower bainite in steels with dierent carbon contents using
the same method. It was found that the OAE of transformations increases with the decrease of carbon content in
the steels, and that QBU is always bigger than QBL . The same
conclusion was reported by Yu et al.8) who measured the
74
Steel No. 4
Continuous
cooling
eq. (1),
X 5%
7), 8), 9)
10)
Continuous
cooling
175 25
126
Steel No. 1, 2, 3
eq. (1),
X 10%
26 steels
HSLA-100 steel
1.2
eq. (1),
X 1075%
6)
Single structure
10)
12)
3)
OAE based on the transformed fraction of 25{30%. All the

reported OAE data for various steels is listed in Table 1, in
which 613{623 K is used as the transition temperature from
upper bainite to lower bainite.
In 1966, Barford determined the OAE for BU and BL in
terms of transformed fraction of 1075% (or 60%) in high
carbon Mn-containing steels9) using both eqs. (1) and (2).
Although the results are consistent with those previously
reported for the steels with carbon content less than
1.1 mass%, conict results were obtained in the steel
containing 1.42 mass%C. The OAE obtained from both
equations shows QBU > QBL at the transformed fraction of
10%, but the opposite (QBL < QBL ) at transformed fraction of
50%, as bolded in Table 1. The latter is abnormal. Although
Barford did not explain the abnormal results, the reported
data demonstrates that the experimentally determined OAE
not only related to the carbon content of steels, but also
depends on the transformed fraction, on which the determination is based. Hillert et al.10) reviewed the TTT diagrams of
more than 20 steels, including plain carbon steels, alloy steels
and Mn-containing steels. It was found that the QBL value is
constantly 80 kJ/mol for the lower bainite transformation in
all steels when the determination was in terms of transformed
fraction of 1%. This is consistent with Barfords results
determined in terms of the transformed fraction of 10% in the
1.42 mass%C steel (see Table 1). In addition, the OAE of
bainite transformation in continuous cooling conditions was
also investigated.11,12) The published data is listed in Table 1.
It can be seen that the QBU is always bigger than the QBL for
all steels.
However, all previous researchers did not pay enough
attention to the eect of the fraction of transformation on the
Kinetics and Morphology of Isothermal Transformations at Intermediate Temperature in 15CrMnMoV Steel

Table 2
125
Chemical compositions, pre-heat treatment processes and MS temperature of the 15CRMnMoV steel.
Chemical compositions (mass%)
Cr
Mn
Si
Mo
Bal
0.156
1.36
0.89
0.13
1.03
0.34
Fe
Homogenization
annealing
T (K) t (h)
1470 72
Austenization
Ms
T (K) t (h)
(K)
1243 10
623
morphology of nal microstructures as well as on the

experimentally determined OAE data. In the early 1990s
Kang et al.3) noticed this inherent aw and suggested that
more accurate QBU and QBL can only be determined from
eq. (2) in terms of the transformed fraction less than 5%,
because low transformed fraction can guarantee the formation of single microstructure (BU or BL ). The OAE data
previously reported by Kang and co-workers is also listed in
Table 1.
2.
The chemical composition of 15CrMnMoV steel used is

listed in Table 2. The as received bars (20 mm in diameter)
were homogenization annealed at 1470 K for 72 hours, then
3 mm diameter and 30 mm long thermo-magnetic specimens
were machined. After austenization in NaCl salt bath at
1243 K for 10 hours, these specimens were then isothermally
treated at dierent temperatures for various times in either
liquid lead or liquid lead-tin baths, followed by water
quenching. A thermo-magnetic apparatus tted with an X-Y
recorder, which has a sensitivity of 0.2%,3) was used to
determine the transformed fraction with according to the
following equation. The transformed fraction, X:
X
0
1 0
where , 1 and 0 are the magnetic property related

parameters to be read directly from the X-Y recorder. 1 is
taken at the annealed state, 0 is the reading at austenization
temperature and is the reading at a specied isothermal
holding temperature and time before water quenching.
Therefore, X only associates with the isothermal transformation fractions without eect of the martensite formed
in the subsequent water quenching.
TEM specimens were prepared in the process as follows:
0.81.0 mm thick discs were cut o from the thermomagnetic specimens using a spark cutter. After mechanically
grinding to 0.080.1 mm thick, the discs were then jet
polished in a GIKOCE-10 machine with voltage of 35 V at
15 C. The electrolyte used is 710% perchloric acid and 90
93% glacial acetic acid. All specimens were examined in H800 TEM at an operating voltage of 200 keV.
3.
Fig. 1 TTT-Curve within intermediate temperature transformation range

in 15CrMnMoV steel.
3.1 TTT diagram and microstructure

The TTT diagram of 15CrMnMoV steel constructed
based on the data from the thermo-magnetic apparatus is
shown in Fig. 1. There are three independent 1% transformation C curves, corresponding to the lower bainite (BL ),
the upper bainite (BU ) and the granular structure (GS),
Fig. 2
Schematically illustration of GS (a), BMU (b), Bg (c) and BL (d).
respectively. As shown by the dashed lines, longer isothermal holding leads to the overlap of these C curves. This
implies that at beginning of the transformation at a specied
temperature single microstructure can be obtained and as the
increase in holding time mixtures of two or three types of
microstructure form.
The morphology of these three microstructures is distinguishable.3,4) GS consists of massive ferrite matrix and
randomly distributed martensite-austenite (M-A) islands in
the matrix, as schematically illustrated in Fig. 2(a). The M-A
islands are normally in irregular shape and are made of either
martensite-austenite mixture or sole martensite or single
retained austenite (AR ).3) BU includes meta-upper bainite
126
Fig. 3 Microstructure in 15 CrMnMoV steel austempered at 643 K after austenization: (a) Lower bainite (BL ) isothermally held for 3 s,
(b) TEM image showing typical lower bainite (BL ) isothermally held for 7 s, (c) (Lower bainite+upper bainite) after isothermal holding
for 10 s, and (d) TEM micrograph of (c).
Fig. 4 Microstructure in 15 CrMnMoV steel austempered at 703 K after austenization: (a) Upper bainite (BU ) after isothermal holding for
10 s, (b) TEM micrograph of typical upper bainite (Bu ) showing in (a).
(BMU ) and granular bainite (Bg).3,4) It consists of lath bainitic

ferrite and either AR thin lms (BMU ) or M-A islands (Bg) in
between the ferrite laths, as schematically illustrated in
Figs. 2(b) and 2(c). BL in 15CrMnMoV steel is comprised
of plate ferrite and systematically distributed AR thin lms
within the ferrite plates.3,4) The optical morphology of BL is
needle-like and in TEM AR thin lms can be observed as
schematically shown in Fig. 2(d). The AR and M-A islands in
both BU and BL will decompose into carbides and ferrite in
the consequent long time isothermal holding or tempering,
thus typical BU and BL form.3,4)
Isothermal holding of the 15CrMnMoV steel after
austenization at 643 K for 3 seconds produces needle-like
BL as shown in Fig. 3(a) is the optical micrograph.
Figure 3(b) is the TEM micrograph illustrates the AR thin

lms systematically distributed within the bainitic ferrite
plate. Longer time holding (10 seconds) leads to the
formation of BU beside the existing BL plates as shown in
Fig. 3(c) (optical micrpgraph) and (d) (TEM micrograph).
Similarly, single BU can only be obtained at the initial stage
of the transformation within the BU transformation zone at
higher temperatures. Isothermal holding at 703 K for 10
seconds produces Bg as shown in Fig. 4(a) and BMU
exhibiting in Fig. 4(b). Both Bg and BMU can be regarded
as a meta-stable or special type of BU and both share the
same C-curve in TTT diagram. Similarly, if the isothermal
holding time is long enough, BL forms at the same
temperature as indicated in Fig. 1. Single GS forms as a
127
Fig. 5 Optical microstructure in 15CrMnMoV steel austempered at 773 K after austenization: (a) Granular structure (GS) after isothermal
holding for 150 s, (b) GS+Bg mixture after isothermal holding for 3000 s.
consequence of isothermal holding at 773 K for 200 seconds

as shown in Fig. 5(a). It can be distinguished from Bg based
on the shape of the ferrite matrix and the distribution of M-A
islands on it. In GS, M-A islands are randomly distributed in
the massive ferrite matrix. In contrast, the M-A the islands in
Bg distributes in between the ferrite laths.3,13) Generally, in
low carbon steels the GS transformation obeys an independent C-curve on the TTT diagram. Like the transformation at other temperatures, if the isothermal holding
time is long enough, Bg forms after GS as shown in
Fig. 5(b), where arrow A indicates Bg, and arrow B points to
the GS.
As the stated above, it can be concluded that, (1) single
microstructure such as GS, Bg, BU and BL , can only be
obtained at the very early stage of transformation; (2) long
time isothermal holding at a specied temperature leads to
the formation of mixed microstructures; (3) in terms of the
starting C-curve on the TTT diagram, the transition from BU
to BL occurs at a critical temperature, 673 K for 15CrMnMoV
steel. However, BL can also form after BU at temperatures
above 673 K, BU can form as a consequence of BL below
673 K if the isothermal holding time is long enough in
15CrMnMoV steel.
In addition, a similar phenomenon is also observed in
30CrMnSiNi2 steel.3,4,14) There is no GS and Bg in medium
carbon steels. Carbide-free (meta) bainite consisting of
bainitic ferrite (BF) and retained austenite (AR ) is the
common microstructure in Si-containing steels.3,4,14)
3.2
The overall activation energy of bainitic transformation

The eect of temperature on transformation rate is well
described by the linear relationship between ln@X=@t and
1=T for a given transformation fraction (X). The overall
activation energy can be calculated based on the slope of the
straight line using eq. (2). Figure 6 shows such a relationship in terms of X 1% and 5% for GS, BU and BL
respectively for the 15CrMnMoV steel. For comparison the
ln@X=@t vs. 1=T relationship at X 50% is also included
in Fig. 6 because mixed microstructure of BU and BL
formed at this transformed fraction. Although the lines
associated with X 1% and X 5% for a specied
microstructure are parallel, the slops of lines for dierent
Fig. 6 Relationship between ln@x=@t and 1=T at transformation fraction

X 1, 5 and 50% within medial temperature transformation region in
15CrMnMoV steel.
microstructures, including the mixed one vary. This indicates that the overall activation energies of GS, BU and BL
are dierent. The parallelism of the 1% and 5% lines
implies, at least when the total transformed fraction is below
5%, single microstructure is obtained. From the slope of the
lines and using eq. (2) the OAEs of GS, BU and BL are
calculated to be QGS 170 kJ/mol, QBU 132 kJ/mol and
QBL 52 kJ/mol, with accuracy of 8%. Because these
OAE values were measured based on the onset of the
transformation and on single microstructure, they are
independent from the transformed fraction. In addition,
although the nucleation sites also aect the overall activation energy, such as austenite grains and dislocations, the
present determinations were based on the same austenization
temperature and used the same cooling agent (NaCl salt
128
bath). Therefore, it is reasonable to ignore these inuences.

Because the determined value of QBU is very close to the
reported diusion activation energy of carbon in -Fe
(QC 131 kJ/mol),8,15) it is reasonable to consider that the
transformation of upper bainite is controlled by the diusion
of carbon in -Fe. The present QBL value agrees well with
Hillerts data.10) Although it is close to the value of the
diusion activation energy of carbon in -Fe (QC 70 kJ/
mol),8) in the initial stages of transformation, lower bainite
cannot form through decomposition of martensite, which
might form from austenite, within the lower bainite forming
temperature range.25 This is because carbon diusion in Fe is more dicult at such low temperatures. In addition,
since the determined QBL is much lower than the QC , hence,
the rate of the lower bainite transformation may be partially
controlled by carbon diusion in -Fe. The determined OAE
of GS, QGS , is in between the OAE value of equiaxed
transformation, (Q! 193{198 kJ/mol)17) and the OAE
value of the massive transformation, (QM 144 kJ/
mol).18,19) Hence, GS transformation has the common
characteristics of both equiaxed and massive transformations. It can be regarded as diusion-less (for Fe atoms)
ferrite/austenite interface immigration process controlled by
carbon diusion. The GS is a product of independent
process that diers from both upper and lower bainite
transformations.
The determined overall activation energy at X 50% is
dierent from all the three OAE values for QU , QL and GS
determined at X 1% and 5%. In Fig. 6, there are two nonparallel straight lines for X 50%. One was determined
within lower bainite temperature range and another within
upper bainite temperature range (T < 673 K). From the slop
of the line obtained at lower temperatures, the OAE for the
mixed microstructure was calculated to be Q0 BL 127 kJ/
mol. Because the transformation occurred within the lower
bainite zone, the Q0 BL value should be regarded as the OAE
of lower bainite. However, Q0 BL is much too higher than the
QBL . Similarly, the calculated overall activation energy,
Q0 BU 118 kJ/mole, in terms of the slope of the X 50%
line in Fig. 6, which was determined within the upper
bainite region (T > 673 K). Although it is determined within
upper bainite zone, the value is lower than the QUB
determined for single BU . More importantly, one can notice
that Q0 BL is greater than Q0 BU , which is abnormal, even
though a similar result was reported by Barford.9) This
inconsistence is attributed to the mixing of BU and BL at
X 50%. Although single upper (or lower) bainite can be
obtained at initial stage of the isothermal transformation
within upper (or lower) bainite temperature range, as
consequence, lower (or upper) bainite forms at 50% of the
total transformed fraction. Thus, both Q0 BU and Q0 BL are
actually the OAEs of the formation of mixture of BU and BL .
Neither of them is true OAE for BU or BL . These two values
are considered as false OAE and meaningless. They cannot
be used to study of kinetics of the transformations. Therefore, accurate determination of OAE should be in terms of
single microstructure, which can be obtained at the early
stage of the transformation.
The presently determined values QBL (52 kJ/mol) and
QBU (132 kJ/mol) with single microstructure in
15CrMnMoV steel are not only far below the self-diusion

activation energy of Fe in -Fe (Q 250 kJ/mol),4) and in
-Fe (Q 285 kJ/mol),15) but also smaller than the OAE
of equiaxed ferrite formation (Q! 193{198 kJ/mol)17)
and massive ferrite formation (QM 144 kJ/mol).18,19)
These experimental data suggests that bainitic ferrite,
including upper and lower bainite, cannot form from
austenite through either conventional diusion controlled
or massive diusion-less transformation. In order to retain
the lath morphology, both the BU and the BL transformations
can only occur by diusion-less martensite-like shear that is
partially controlled by carbon diusion in austenite. However, as the QGS is in between the Q! and QM , the
granular structure is regarded as a product of massive type
diusion-less transformation.
4.
Conclusions
(1) Upper bainite (BU ), lower bainite (BL ) and granular

structure (GS) in low carbon alloy steels have distinguishable independent C-curves in TTT diagrams,
although the C-Curves may overlap. Single microstructure can only be obtained at initial stage of
isothermal transformation with low total transformed
fraction. Long time isothermal holding leads to the
increase of total transformed fraction accompanied with
the formation of mixed microstructures.
(2) The overall activation energy for a phase transformation can only be accurately determined in terms of
single microstructure, which normally appears when
the total transformed fraction is less than 20%.
(3) In steels, from the similarity of QBU to QC , the BU
transformation is controlled by carbon diusion in the
austenite; however, the formation of BL is only partially
governed by carbon diusion in austenite as QBL is
much lower than QC . As an independent microstructure, the GS transformation is controlled by carbon
diusion and has the common features of both equiaxed
ferrite and massive ferrite transformations, because the
QGS is in between Q! and QM .
Acknowledgements
The authors would like to thank Liu Donghui, Yang
Yanqing, Meng Xiangkang, Lu Yangsheng and Sun Xiaoyan
for the experimental work.
Liu Feng is grateful to the National Natural Science
Foundation of China 50501020 for funding support.
REFERENCES
1) E. S. Davenport and E. C. Bain: Trans. AIME 3 (1930) 117.
2) T. Ko and S. A. Cottrel: JISI 172 (1952) 307.
3) M. K. Kang, D. M. Chen, S. P. Yang and G. L. Hu: Metall. Trans. A 30
(1990) 1992.
4) M. K. Kang, J. L. Sun and Q. M. Yang: Metall. Trans. 21A (1990)
853.
5) M. Hillert: Acta Metall. 7 (1957) 653.
6) P. Vasudevan, L. W. Graham and H. J. Axon: JISI 190 (1958) 386.
7) S. V. Radclie and E. C. Rollason: JISI 191 (1959) 56.
8) D. G. Yu and D. K. Shi: Proc. the First Conf. on Heat Treatment,

(Institution of EMES, Wuhan, China Machine Press, 1996) p. 105.
9) J. Barford: JISI 204 (1966) 609.
10) A. Borgenstam and M. Hillert: Acta Mater. 45 (1997) 651.
11) C. Gupta, G. K. Dey, J. K. Chakravatty, D. Srivastav and S. Banerjee:
Scr. Mater. 53 (2005) 559.
12) M. Venkatraman, O. N. Mohanty and N. Ghosh: Scand. J. Metall. 30
(2001) 8.
13) N. K. Xu, D. H. Guan and M. K. Kang: Phys. Chem. Exam.Physical
Party 19 (1983) 2.
14) M. K. Kang, M. Zhou and M.-X. Zhang: J. Mater. Sci. Technol. 21
(2005) 437.
129
15) R. S. Wmanski, V. Y. Finkelstan, M. E. Blanger, S. T. Kistin, N. S.

Fastov and S. S. Golerik: Physical Base on Metallography (Physical
Metallurgy), translated by Metal Institute of China Academy of
Science, (Science Press, 1958) p. 188.
16) M. K. Kang and M. Zhu: Trans. Mater. Heat Treat. 28 (2007) 42.
17) Z. S. Hou, X. G. Zhao and W. Y. Hou: Trans. Mater. Heat Treat. 26
(2005) 6.
18) J. J. Wits, T. A. Kop, Y. V. Leeuwen, J. Seitsma and S. V. D. Zwaag:
Mater. Sci. Eng. A 283 (2000) 234.
19) D. A. Porter and K. E. Easterling: Phase Transformation in Metals and
Alloys, (Van Nostrand Reinhold Company, New York, 1981) p. 158.

Solid-State Reactive Diusion between Sn and Electroless NiP at 473 K

Masafumi Yamakami1 and Masanori Kajihara2; *
1
2
Graduate School, Tokyo Institute of Technology, Yokohama 226-8502, Japan

Department of Materials Science and Engineering, Tokyo Institute of Technology, Yokohama 226-8502, Japan
Nickelphosphorus alloys are electrolessly deposited onto Cu-base conductors to suppress formation of CuSn compounds during
soldering using Sn-base solders. However, a NiSn compound is produced during soldering, and continuously grows during energization heating
at solid-state temperatures. To examine inuence of P on the growth behavior of the NiSn compound during energization heating, the kinetics
of the solid-state reactive diusion between Sn and electroless NiP alloys was experimentally determined at 473 K in the present study. For the
experiment, pure Cu sheets were electrolessly deposited with NiP alloys containing 4.6 at%, 18.5 at% and 20.4 at% of P, and then sandwiched
between pure Sn plates. Such Sn/(NiP)/Cu/(NiP)/Sn diusion couples were isothermally annealed at 473 K for various periods up to 1307 h.
During annealing, a layer of Ni3 Sn4 is formed along the Sn/(NiP) interface in the diusion couple. The annealing time dependence of the mean
thickness of the Ni3 Sn4 layer is expressed by a parabolic relationship. The parabolic coecient slightly increases with increasing P
concentration in the NiP. Thus, P in the NiP lightly accelerates the growth of Ni3 Sn4 at the interconnection between the NiP and the Sn-base
solder. Using the experimentally determined values of the parabolic coecient, the interdiusion coecient in Ni3 Sn4 was analytically
evaluated by a mathematical model. The acceleration eect of P on the growth of Ni3 Sn4 is quantitatively explained by the dependence of the
interdiusion coecient on the P concentration in the NiP. [doi:10.2320/matertrans.MRA2008266]
Keywords: diusion bonding, reactive diusion, intermetallic compounds, bulk diusion, solder, conductor, electroless deposition
1.
Introduction
In the electronics industry, Cu-base alloys are widely used

as conductor materials. However, Cu is not completely
corrosion resistant. Hence, the Cu-base alloy is usually plated
with a Au layer to improve corrosion resistance. If the Au/Cu
conductor material is interconnected using a Sn-base solder,
the Au layer may quickly dissolve into the molten Sn-base
solder during soldering unless the Au layer is suciently
thick. When the Au layer dissolves, CuSn compounds are
formed at the interconnection due to the reactive diusion
between the Cu-base alloy and the molten Sn-base solder.
Since the CuSn compounds are brittle and possess high
electrical resistivities, their formation deteriorates the electrical and mechanical properties of the interconnection. The
formation behavior of the CuSn compounds has been
experimentally studied by many investigators.117) To suppress CuSn compounds formation during soldering, the
Cu-base alloy is generally plated with a Ni layer prior to Au
layer deposition. In such a multilayer Au/Ni/Cu conductor
material, the Ni layer actually acts as an eective barrier
against the reactive diusion between the Sn-base solder and
the Cu-base alloy.1820)
The Ni layer electrolessly deposited onto the Cu-base alloy
usually contains 520 at% of P. The reactive diusion
between the electroless NiP layer and the Sn-base solder
has been experimentally studied by many researchers.2039)
However, a soldering technique has been used in most of
these studies. In this technique, a diusion couple is prepared
from a molten Sn-base solder and a (NiP)/Cu conductor
material at soldering temperatures, and then isothermally
annealed at solid-state temperatures. Compound layers with
certain thicknesses are formed during soldering, and continously grow during annealing. Therefore, it is not so easy
to distinguish the growth rate during annealing from the
formation rate during soldering. Consequently, the soldering
*Corresponding
author, E-mail: kajihara@materia.titech.ac.jp
technique may not be adequate to observe the growth

behavior of the compound layers due to annealing at solidstate temperatures.
On the other hand, a diusion bonding technique has been
used to examine experimentally the solid-state reactive
diusion in many alloy systems in previous studies.4061) In
this technique, a diusion couple is prepared from dierent
metals or alloys by isothermal heating at the same temperature as isothermal annealing. Hence, the growth behavior of
compound layers during heating and annealing at a constant
temperature can be readily observed by the diusion bonding
technique. Although Ni is an eective barrier to suppress
the CuSn compounds formation, addition of P to Ni may
degrade the eect of Ni.20) However, there is hardly reliable
information for such inuence of P. To examine inuence
of P on the eect of Ni, the solid-state reactive diusion
between Sn and electroless NiP alloys with dierent compositions was experimentally determined by the diusion
bonding technique in the present study. The mathematical
model reported in a previous study62) was used to analyze
quantitatively the experimentally observed kinetics of the
reactive diusion.
2.
Experimental
Polycrystalline plate specimens of pure Sn with a size of

12 mm 5 mm 2 mm were prepared by cold rolling and
spark erosion from a commercial 1 kg rectangular ingot of
pure Sn with purity of 99.99%. The cold-rolled plate
specimens were separately annealed in evacuated silica
capsules at 473 K for 2 h, followed by air cooling without
breaking the capsules. The annealed Sn plate specimens were
chemically polished in an etchant consisting of 20 vol% of
nitric acid, 20 vol% of hydrochloric acid and 60 vol% of
distilled water. The two surfaces with an area of 12 mm
5 mm of each Sn plate specimen were mechanically polished
on 800 emery paper. One of the two polished surfaces was
again mechanically polished on 10004000 emery papers

Table 1
Chemical composition, initial thickness and crystal structure of

NiP layer electrolessly deposited onto Cu sheet.
NiP layer
Mol fraction
P, y
Mean of
y
A1
A2
0.045
0.045
0.046
A3
0.047
B1
0.181
B2
0.190
C1
C2
0.204
0.204
C3
0.205
0.185
0.204
Initial thickness
(a)
Crystal structure
13 mm
12 mm
fcc
13 mm
fcc
11 mm
amorphous
11 mm
amorphous
10 mm
13 mm
amorphous
amorphous
13 mm
amorphous
fcc
until a depth of 100 mm and then nished using diamond with

a diameter of 1 mm.
Polycrystalline Cu sheet specimens with a thickness of
50 mm and purity of 99.99% were electrolessly deposited
with NiP layers with various compositions. Such (NiP)/
Cu/(NiP) sheet specimens were supplied by Japan Kanigen
Co. Ltd. The chemical composition and the initial thickness
of the electroless NiP layer are listed in Table 1. In this
table, the composition is shown as the mol fraction y of P.
The NiP layers with y 0:045{0:047, y 0:181{0:190 and
y 0:204{0:205 are denominated layers A1A3, B1B2 and
C1C3, respectively, and the mean of y is equal to 0.046,
0.185 and 0.204 for layers A1A3, B1B2 and C1C3,
respectively. Sheets with an area of 12 mm 5 mm were cut
from the (NiP)/Cu/(NiP) specimen, and then degreased
with acetone in an ultrasonic cleaning equipment. The crystal
structure of the NiP layer was determined by an X-ray
diraction (XRD) technique using Cu K radiation.
After degreasing, a (NiP)/Cu/(NiP) sheet specimen was
immediately sandwiched between the nished surfaces of
two freshly prepared Sn plate specimens in ethanol by a
technique used in a previous study.19) The Sn/(NiP)/Cu/
(NiP)/Sn couples were completely dried and then heat
treated for diusion bonding in an oil bath with silicone oil at
473 K for 24 h. After the heat treatment, the diusion couples
were isothermally annealed at 473 K for various times up to
1283 h. The summation of the heat-treating and annealing
times is hereafter merely called the annealing time t. Crosssections of the annealed diusion couples were mechanically
polished using diamond with diameters of 15, 3 and 1 mm and
then nished with an OP-S liquid by Struers Ltd. The
microstructure of the cross-section was observed with an
optical microscope (OM). The composition of each phase
was measured on the cross-section by electron probe microanalysis (EPMA).
3.
131
(b)
(c)
3.1 X-ray diraction measurement

Typical XRD patters are shown in Fig. 1. Figure 1(a), (b)
and (c) indicates the results of layers A3, B1 and C3,
respectively. As can be seen, a sharp diraction peak is
observed at 2 44:74 for layer A3, and broad diraction
peaks are recognized at 2 44:86 and 44.90 for layers B1
and C3, respectively. Since the lattice parameter of Ni is
aNi 0:35238 nm and the wavelength of Cu K X-ray is
Fig. 1
X-ray diraction patters of (a) layer A3, (b) layer B1 and (c) layer
C3.
132
M. Yamakami and M. Kajihara
Fig. 3
Fig. 2
The lattice spacing d of the (111) or closed-pack plane versus the

mol fraction y of P for layers A3, B1 and C3 shown as open circles.
The bold straight line indicates Vegards law.
0:154178 nm,63) the diraction peak for the (111) plane

of Ni will appear at 2 44:533 according to the following
relationship.
2d sin
Here, the lattice spacing for the (111) plane of Ni is

dNi 0:20345 nm. As a consequence, layer A3 is a crystalline phase with the face-centered cubic (fcc) structure, but
layers B1 and C3 are amorphous. The XRD measurements of
all the NiP layers are summarized at the fth column in
Table 1. The crystal structure of electroless NiP was
experimentally observed in the composition range of y
0:055{0:236 by Kumar and Nair,64) Gao et al.,65) and
Palaniappa and Seshadri.66) According to their observations,
the electroless NiP is the fcc phase for y 0:055{0:079,
a mixture of the fcc and amorphous phases for y
0:079{0:159, and amorphous for y 0:159{0:236. The
result in Table 1 is consistent with their observations.
Using the values of 2 for the sharp and broad diraction
peaks in Fig. 1, the lattice spacing d of the (111) plane was
evaluated from eq. (1). Unlike layer A3, however, layers B1
and C3 are amorphous. In the case of the amorphous layer, d
corresponds to the mean lattice spacing of the closed-pack
plane. The evaluated values of d are plotted against the mol
fraction y of P as open circles in Fig. 2. In this gure, an open
square shows the corresponding value of Ni. As can be seen,
d is smaller for layers A3, B1 and C3 than for Ni. Thus,
addition of P to Ni decreases the interatomic distance.
According to Vegards law, d may be expressed as a linear
function of y by the equation
d 1 y dNi y dP ;
where dP is the lattice spacing of the (111) plane for a

hypothetical fcc-phase of P. Combining dNi 0:20345 nm
with the atomic radii rNi 0:125 nm and rP 0:109 nm of
Ni and P, respectively,63) we obtain dP 0:17741 nm. Using
these values of dNi and dP , the dependence of d on y was
calculated from eq. (2). The result is shown as a bold straight
Optical micrograph of cross-section for diusion couple A1

annealed at 473 K for 309 h.
line in Fig. 2. On the other hand, vertical thin dashed-anddotted lines indicate the critical compositions y 0:079 and
0.159. As can be seen, the open circle of layer A3 with y
0:047 lies well on the bold straight line. According to a recent
phase diagram in the binary NiP system,67) the two-phase
microstructure consisting of the fcc Ni phase and Ni3 P is
stable for y 0:00045{0:25 at room temperature. Furthermore, the solubility of P in the Ni phase is merely y 0:0032
even at 1143 K. Nevertheless, any diraction peaks of Ni3 P
were not observed in all the XRD measurements. As
mentioned above, the open circle of layer A3 almost obeys
eq. (2). Thus, it is concluded that the electroless NiP with
y 0:05 is the fcc Ni phase supersaturated with P and the Ni
atoms in the fcc lattice are substitutionally replaced with P
atoms. In contrast, the open circles of layers B1 and C3 are
located on the upper side of the bold straight line. Hence,
eq. (2) is not valid for the amorphous phase in the binary
NiP system.
3.2 Microstructure
A typical micrograph of OM for the cross-section of the
annealed diusion couple is shown in Fig. 3. This gure
indicates the micrograph for the diusion couple with layer
A1 annealed at 473 K for 309 h (1.11 Ms). In Fig. 3, the
regions on the upper and lower sides are the Sn specimen, the
horizontal band with a thickness of about 50 mm is the Cu
specimen, and the gray layers with a uniform thickness
adjacent to the Cu are the electroless NiP. As can be seen, a
rather wavy compound layer is formed at each interface
between the NiP and Sn. This compound layer is Ni3 Sn4 .
Concentration proles of Ni, Sn and P across the Ni3 Sn4 and
NiP layers were determined by EPMA. The results of the
diusion couples with layers A3, B1 and C1 annealed at
473 K for 1307 h (4.71 Ms) are shown as open symbols in
Fig. 4(a), (b) and (c), respectively. Hereafter, the diusion
couples with layers A1A3, B1B2 and C1C3 are merely
called diusion couples A1A3, B1B2 and C1C3, respectively. In Fig. 4, the abscissa indicates the distance x, and the
ordinate shows the mol fraction yi of component i (i Ni,
Sn, P). As can be seen, the mol fraction yP of P is negligible in
the Sn, Cu and Ni3 Sn4 . On the other hand, in the NiP layer,
133
(a)
Fig. 5
Diusion paths corresponding to the concentration proles in

Fig. 4.
(b)
(c)
Fig. 4
Concentration proles of Ni, Sn and P across the Ni3 Sn4 and NiP
layers after annealing at 473 K for 1307 h: (a) diusion couple A3,
(b) diusion couple B1 and (c) diusion couple C1.
yP is greater after annealing than before annealing. Hereafter,

the values of yP in the NiP layer before and after annealing
are denoted by y and yNiP
P , respectively. The concentration
proles in Fig. 4 are plotted as diusion paths in Fig. 5. In
this gure, the ordinate and the abscissa show ySn and yP ,
respectively, and open triangles, rhombuses and circles
indicate the results of diusion couples A3, B1 and C1,
respectively. In contrast, open squares show y. Since yP in
Ni3 Sn4 is negligible, P atoms are excluded from Ni3 Sn4 to
the NiP layer due to growth of Ni3 Sn4 . Consequently, yNiP
P
gradually increases with increasing annealing time. As can be
at an annealing
seen in Fig. 4, the maximum value of yNiP
P
time of t 1307 h (4.71 Ms) is 0.064, 0.243 and 0.253 for
diusion couples A3, B1 and C1, respectively. The mol
of P in Ni3 P is equal to 0.25. The values yNiP
fraction yNi3P
P
P
0:25.
0:243 and 0.253 are close to yNi3P
P
In the binary NiSn system,68) Ni3 Sn2 and Ni3 Sn as well as
Ni3 Sn4 appear as stable compounds at 473 K. On the other
hand, P3 Sn, P4 Sn3 and P3 Sn4 are the stable compounds at
473 K in the binary PSn system.69) Nevertheless, only
Ni3 Sn4 was observed as a Sn-base compound under the
present annealing conditions. The solid-state reactive diusion between Sn and an electroless NiP layer with y 0:174
was experimentally studied by Tomlinson and Rhodes.20) In
their experiment, a NiP layer with a thickness of 20 mm was
deposited onto a mild steel substrate, and then a pure Sn layer
with a thickness of 20 mm was deposited onto the NiP layer.
Although the thickness of the NiP layer is not so dierent
between their study and the present study, the thickness of the
Sn layer is two orders of magnitude smaller in their study
than in the present study. Such multilayer Sn/(NiP)/steel
diusion couples were isothermally annealed in the temperature range between 453 and 493 K for various times up to
508 h. Also in their experiment, a Ni3 Sn4 layer was observed
at the Sn/(NiP) interface in the annealed diusion couple.
Furthermore, the enrichment of P was observed in the NiP
layer after annealing. Unfortunately, however, the concentration change of P in the NiP layer was not quantitatively
determined in their experiment.
134
3.3 Growth behavior of Ni3 Sn4

From cross-sectional OM micrographs like Fig. 3, the
mean thickness l of the Ni3 Sn4 layer was evaluated by the
equation
l A=w;
(a)
where A and w are the area and the length of the Ni3 Sn4 layer,
respectively, on the cross-section. Hereafter, the Ni3 Sn4 layer
is merely called the intermetallic layer. At each annealing
time, l was calculated from eq. (3) using the total values of A
and w for various cross-sections. The results of diusion
couples A1A3, B1B2 and C1C3 are shown as open
symbols in Fig. 6(a), (b) and (c), respectively. In this gure,
the ordinate indicates the thickness l, and the abscissa shows
the square root of the annealing time t. Although the open
symbols are slightly scattered, we may expect that l is
proportional to the square root of t. Such a relationship is
called the parabolic relationship. The parabolic relationship
is expressed by the following equation:
l2 Kt:
(b)
Here, K is the parabolic coecient. From the open symbols

in Fig. 6, K was evaluated by the least-squares method. In the
evaluation, a common value of K was adopted for each mean
of y. The mean of y is listed at the third column in Table 1.
Using the evaluated values of K, l was calculated as a
function of t from eq. (4). The results are shown as solid lines
in Fig. 6. On the other hand, K is plotted against the mean of
y as open circles with error bars in Fig. 7. In a similar
manner, K was evaluated for y 0 using the experimental
data reported in a previous study.19) The result is shown as an
open square with an error bar in Fig. 7. As can be seen, K
slightly increases with increasing value of y. This means that
the growth of Ni3 Sn4 is lightly accelerated by P in the
electroless NiP.
The growth behavior of Ni3 Sn4 during the solid-state
reactive diusion between Sn and NiP alloys with y 0 and
0.174 was experimentally studied at 453493 K by Tomlinson and Rhodes.20) Their results for y 0 and 0.174 at 473 K
are shown as open circles and squares, respectively, in Fig. 8.
In this gure, like Fig. 6, the ordinate and the abscissa
indicate the thickness l of the intermetallic layer and the
square root of the annealing time t, respectively. As can be
seen, the parabolic relationship holds between l and t, though
the open symbols are slightly scattered. From the open
symbols in Fig. 8, K was evaluated by the least-squares
method. The evaluated values of K are plotted against y as
open rhombuses with error bars in Fig. 7. At y 0, K is
similar in the open rhombus and square. On the other hand, K
is lightly smaller for the open rhombus at y 0:174 than for
the open circles at y 0:185 and 0.204. Nevertheless, the
open rhombuses also indicate that K slightly increases with
increasing value of y.
As previously mentioned, the reactive diusion between
the Sn-base solder and the electroless NiP has been
experimentally studied by many investigators.2039) According to these studies, the enrichment of P in the NiP occurs
owing to growth of Ni3 Sn4 and causes formation of Ni3 P in
the NiP adjacent to Ni3 Sn4 . In the case of the experiment by
Li et al.,37) diusion couples composed of electroless NiP
(c)
Fig. 6
The mean thickness l of the Ni3 Sn4 layer versus the annealing time t
at 473 K: (a) diusion couples A1A3, (b) diusion couples B1B2
and (c) diusion couples C1C3.
135
l 9:1 mm at t 4:71 Ms (1307 h). On the other hand, as

shown in Fig. 4, the thickness of the NiP layer at t 1307 h
is smaller than 9.1 mm in all the diusion couples. Furthermore, the mol fraction of P in the NiP layer is close to 0.25
for diusion couples B1B2 and C1C3 at t 1307 h.
Consequently, it is concluded that the electroless NiP with
y 0:181{0:205 is completely transformed into Ni3 P in the
late stages of annealing at 473 K due to the growth of Ni3 Sn4 .
Fig. 7
The parabolic coecient K versus the mean of y at 473 K shown as

open circles with error bars. The corresponding result for y 0 in a
previous study19) is indicated as an open square with an error bar,
and those for y 0 and 0.174 by Tomlinson and Rhodes20) are
shown as open rhombuses with error bars.
3.4 Interdiusion coecient of Ni3 Sn4

The growth behavior of a compound layer during solidstate reactive diusion was theoretically analyzed for a
hypothetical binary AB system consisting of one compound
() phase and two primary solid solution ( and ) phases in a
previous study.62) If a semi-innite diusion couple composed of the and phases is isothermally annealed at an
appropriate temperature, a layer of the phase with a
uniform thickness will be formed at the at / interface and
then grows continuously. In such a case, the interdiusion of
components A and B takes place along the direction
perpendicular to the / and / interfaces. This direction
is called the diusional direction. For the growth of the
phase controlled by volume diusion, the distances z and
z of the / and / interfaces are expressed as functions
of the annealing time t by
p
5a
z K 4Dt
and
p
z K 4Dt;
5b
respectively. Here, D is the interdiusion coecient of the

phase, K and K are dimensionless coecients, and z
and z are measured from the initial position of the /
interface along the diusional direction. Since the thickness l
of the phase is equal to the dierence between z and z ,
the following relationship holds between l and t.
l2 z z 2 4DK K 2 t
Comparing eq. (6) with eq. (4), we obtain the relationship

K 4DK K 2 fD:
Fig. 8
The mean thickness l of the Ni3 Sn4 layer versus the annealing time t
by Tomlinson and Rhodes20) at 473 K for y 0 and 0.174 shown as
open circles and squares, respectively.
alloys with P concentrations of 1518 at% and a SnBi solder

with a Bi concentration of 58 at% were prepared by soldering
at 493 K for 300 s, and then isothermally annealed at 473 K
for various periods up to 48 h. During soldering, a Ni3 Sn4
layer with a thickness of 0.93 mm is formed at the (NiP)/
(SnBi) interface, and a Ni3 P layer with a thickness of
0.29 mm is produced at the (NiP)/Ni3 Sn4 interface. The
thickness l of the Ni3 P layer increases to 1.25, 1.72 and
1.96 mm due to annealing for 16, 24 and 48 h, respectively.
Thus, the Ni3 P layer grows by l 0:96, 1.43 and 1.67 mm
owing to annealing for t 57:6, 86.4 and 172.8 ks (16, 24
and 48 h), respectively. Using these values of l and t, the
parabolic coecient for the growth of the Ni3 P layer at 473 K
is evaluated to be K 1:75 1017 m2 /s by the leastsquares method. The value K 1:75 1017 m2 /s results in
Here, the dimensionless factor f is dened as

f 4K K 2 :
As reported in a previous study,62) f is a function of the initial

compositions of the and phases, the compositions of the
/ and / interfaces, and the interdiusion coecients of
the , and phases. If all the independent variables of f
except the interdiusion coecient D of the phase are
known, D is evaluated from eq. (7) using an experimental
value of K. Such evaluation was conducted using the values
of K for y 0, 0.046, 0.185 and 0.204 by the analytical
method proposed in previous studies62,7076) according to the
following assumptions: (1) local equilibrium is realized at
each interface; (2) the interdiusion coecient of each phase
is independent of the composition; and (3) the molar volume
is constant and equivalent among all the phases. In the
evaluation of D for y 0, the value K 3:41 1017 m2 /s
was combined with the following parameters63,68) of the
binary NiSn system at 473 K:
136
y0
Sn 0;
9a
y
Sn
y
Sn
y
Sn
y
Sn
y0
Sn
0:0048;
9b
0:56;
9c
0:572;
9d
0:999;
9e
1;
D 1:66 10
9f
33
m /s
9g
D 1:42 1012 m2 /s:
9h
and
Here, and stand for the Ni and Sn solid-solution phases,

respectively, is Ni3 Sn4 , y0
Sn is the initial value of the mol

fraction of Sn in the phase, y
Sn and ySn are the mol fractions
of Sn on the and sides, respectively, of the / interface,
D is the interdiusion coecient of the phase (, ,
, ). The evaluation provides f 4:26 102 and thus
D 8:01 1016 m2 /s. As to diusion couples A1A3, B1
B2 and C1C3, the concentrations of P in the and phases
are negligible as shown in Figs. 4 and 5. Hence, the and
phases are considered as the binary NiSn phases. Furthermore, the solubility y
Sn of Sn in the NiP ( phase) is very
62)
small. For such small values of y
Sn , f is insensitive to ySn .
Consequently, eq. (9a)(9f) and (9h) is applicable to the
evaluation of D for y 0:046{0:204. In contrast, the NiP is
transformed into Ni3 P due to exclusion of P atoms from
Ni3 Sn4 to the NiP in diusion couples B1B2 and C1C3 in
the late stages of annealing. Since D is dissimilar in Ni, Ni3 P
and the NiP, eq. (9g) is not valid for the phase in diusion
couples A1A3, B1B2 and C1C3. Even in such a case, D
may be much smaller than D. For such small values of D , f
is insensitive to D .62) As a result, we can use f 4:26
102 in the evaluation of D also for y 0:046{0:204. The
evaluated values of D are plotted against y as open circles in
Fig. 9. As can be seen, D slightly increases with increasing
value of y.
In a previous study,19) the growth behavior of the Ni3 Sn4
layer was originally described by the following equation but
not by eq. (4).
l kt=t0
10
Here, k is the proportionality coecient possessing the same

dimension as l, n is the exponent, and t0 is unit time, 1 s.
According to the result in a previous study,19) n is equal to
0.459 and 0.405 at 453 K and 473 K, respectively, and thus
slightly smaller than 0.5. The values of n smaller than 0.5
indicate that both boundary and volume diusion contributes
to the layer growth of Ni3 Sn4 and grain growth occurs in
Ni3 Sn4 . If the layer growth of Ni3 Sn4 is controlled by
boundary and volume diusion, eq. (4) is no longer valid.
However, the open symbols in Fig. 6 are rather scattered.
Hence, unlike a previous study,19) the exponent n could not be
reliably determined by the least-squares method using the
open symbols in Fig. 6. Considering the value n 0:405 for
y 0, we anticipate that n is slightly smaller than 0.5 also for
y 0:046{0:204. Thus, for y 0{0:204, it is plausible that
boundary and volume diusion contributes to the layer
growth of Ni3 Sn4 and the grain growth of Ni3 Sn4 takes place
Fig. 9
The interdiusion coecient D in Ni3 Sn4 versus the mean of y at

473 K shown as open circles.
at certain rates. In a previous study,77) the growth behavior

of a polycrystalline compound layer during the solid-state
reactive diusion controlled by boundary and volume
diusion was numerically analyzed for the same hypothetical
binary AB system as the theoretical analysis reported in a
previous study.62) According to the numerical analysis, the
eective diusion coecient of the compound layer increases with decreasing grain size of the compound layer.
Since the concentration of P in Ni3 Sn4 is negligible as
indicated in Figs. 4 and 5, the dependence of D on y is not
attributed to the inuence of P on the volume diusion in
Ni3 Sn4 . Therefore, D may contain information of boundary
and volume diusion. The dependence of D on y in Fig. 9
implies that the grain size of Ni3 Sn4 at each annealing time
gradually decreases with increasing value of y. To conrm
such dependence of the grain size on y, the chemical etching
of Ni3 Sn4 was conducted using various etchants. However,
each grain in Ni3 Sn4 could not be distinguished by the
chemical etching. Consequently, the dependence of the grain
size of Ni3 Sn4 on y was not conrmed experimentally.
4.
Conclusions
The solid-state reactive diusion in the ternary Sn/(NiP)

system was experimentally observed using the sandwich Sn/
(NiP)/Cu/(NiP)/Sn diusion couples. The diusion couples with mean P concentrations of 4.6 at%, 18.5 at% and
20.4 at% were prepared by the diusion bonding technique,
and then isothermally annealed at 473 K for various times
up to 1307 h in the oil bath with silicone oil. Under such
annealing conditions, the Ni3 Sn4 layer is formed along the
Sn/(NiP) interface in the diusion couple. The square of the
mean thickness of the Ni3 Sn4 layer is proportional to the
annealing time. This relationship is called the parabolic
relationship. The parabolic coecient lightly increases with
increasing P concentration of the NiP. This means that P in
the NiP slightly accelerates the growth of Ni3 Sn4 . Using the
mathematical model reported in a previous study,62) the
interdiusion coecient in Ni3 Sn4 was analytically evaluated from the parabolic coecient. The dependence of the
interdiusion coecient on the P concentration of the NiP
quantitatively accounts for the acceleration eect of P on the

growth of Ni3 Sn4 .
Acknowledgements
The authors are grateful to Dr. S. Watanabe, Mr. Y. Saiki
and Mr. R. Kojima at Japan Kanigen Co. Ltd., Japan for
stimulating discussions. The present study was supported by
Japan Kanigen Co. Ltd., Japan. The study was also partially
supported by a Grant-in-Aid for Scientic Research from the
Ministry of Education, Culture, Sports, Science and Technology of Japan.
REFERENCES
1)
2)
3)
4)
5)
6)
7)
8)
9)
10)
11)
12)
13)
14)
15)
16)
17)
18)
19)
20)
21)
22)
23)
24)
25)
26)
27)
28)
29)
30)
31)
32)
33)
34)
L. Zakraysek: Welding Res. Suppl. Nov. (1972) 536541.

K. N. Tu: Acta Metall. 21 (1973) 347354.
M. Onishi and H. Fujibuchi: Trans. Jpn. Inst. Met. 16 (1975) 539548.
H. N. Keller: IEEE Trans. Components Hybrids Manuf. Tech. CHMT2 (1979) 180195.
H. N. Keller and J. M. Morabito: Surf. Interface Anal. 3 (1981) 1622.
J. O. G. Parent, D. D. L. Chung and I. M. Bernstein: J. Mater. Sci. 23
(1988) 25642572.
A. J. Sunwoo, J. W. Morris, Jr. and G. K. Lucey, Jr.: Metall. Trans. A
23A (1992) 13231332.
P. T. Vianco, P. F. Hlava and A. L. Kilgo: J. Electron. Mater. 23 (1994)
583594.
D. R. Frear and P. T. Vianco: Metall. Trans. A 25A (1994) 15091523.
S. Choi, T. R. Bieler, J. P. Lucas and K. N. Subramanian: J. Electron.
Mater. 28 (1999) 12091215.
P. G. Kim and K. N. Tu: J. Appl. Phys. 80 (1996) 38223827.
A. M. Minor and J. W. Morris, Jr.: Metall. Mater. Trans. A 31A (2000)
798800.
J. H. Lee, J. H. Park, Y. H. Lee and Y. S. Kim: J. Mater. Res. 16 (2001)
12491251.
H. G. Song, J. P. Ahn, A. M. Minor and J. W. Morris, Jr.: J. Electron.
Mater. 30 (2001) 409414.
M. O. Alam and Y. C. Chan: Chem. Mater. 15 (2003) 43404342.
M. O. Alam and Y. C. Chan: J. Mater. Res. 19 (2004) 13031306.
T. Takenaka, S. Kano, M. Kajihara, N. Kurokawa and K. Sakamoto:
Mater. Sci. Eng. A 396 (2005) 115123.
C.-M. Chen and S.-W. Chen: Acta Mater. 50 (2002) 24612469.
M. Mita, M. Kajihara, N. Kurokawa and K. Sakamoto: Mater. Sci. Eng.
A 403 (2005) 269275.
W. J. Tomlinson and H. G. Rhodes: J. Mater. Sci. 22 (1987) 1769
1772.
C.-Y. Lee and K.-L. Lin: Thin Solid Films 249 (1994) 201206.
J. W. Jang, P. G. Kim, K. N. Tu, D. R. Frear and P. Thompson: J. Appl.
Phys. 85 (1999) 84568463.
M. O. Alam, Y. C. Chan and K. C. Hung: J. Electron. Mater. 31 (2002)
11171121.
M. O. Alam, Y. C. Chan and K. C. Hung: Microelectronics Reliability
42 (2002) 10651073.
J.-W. Yoon, C.-B. Lee and S.-B. Jung: J. Electron. Mater. 32 (2003)
11951202.
S. J. Wang and C. Y. Liu: Scr. Mater. 49 (2003) 813818.
M. N. Islam, Y. C. Chan, A. Sharif and M. O. Alam: Microelectronics
Reliability 43 (2003) 20312037.
M. He, Z. Chen and G. Qi: Acta Mater. 52 (2004) 20472056.
J.-W. Yoon and S.-B. Jung: J. Alloy. Compd. 376 (2004) 105110.
M. Date, K. N. Tu, T. Shoji, M. Fujiyoshi and K. Sato: J. Mater. Res. 19
(2004) 28872896.
A. Sharif, Y. C. Chan, M. N. Islam and M. J. Rizvi: J. Alloy. Compd.
388 (2005) 7582.
A. Sharif and Y. C. Chan: J. Alloy. Compd. 393 (2005) 135140.
D.-G. Kim, J.-W. Kim, J.-G. Lee, H. Mori, D. J. Quesnel and S.-B.
Jung: J. Alloy. Compd. 395 (2005) 8087.
Y.-C. Lin and J.-G. Duh: Scr. Mater. 54 (2006) 16611665.
137
35) A. Kumar and Z. Chen: Mater. Sci. Eng. A 423 (2006) 175179.
36) H.-S. Chun, J.-W. Yoon and S.-B. Jung: J. Alloy. Compd. 439 (2007)
9196.
37) J. F. Li, S. H. Mannan, M. P. Clode, K. Chen, D. C. Whalley, C. Liu and
D. A. Hutt: Acta Mater. 55 (2007) 737752.
38) A. Sharif and Y. C. Chan: J. Alloy. Compd. 440 (2007) 117121.
39) R. Labie, W. Ruythooren and J. V. Humbeeck: Intermetallics 15 (2007)
396403.
40) M. Kajihara, T. Yamada, K. Miura, N. Kurokawa and K. Sakamoto:
Netsushori 43 (2003) 297298.
41) T. Yamada, K. Miura, M. Kajihara, N. Kurokawa and K. Sakamoto:
J. Mater. Sci. 39 (2004) 23272334.
42) T. Yamada, K. Miura, M. Kajihara, N. Kurokawa and K. Sakamoto:
Mater. Sci. Eng. A 390 (2005) 118126.
43) T. Takenaka, S. Kano, M. Kajihara, N. Kurokawa and K. Sakamoto:
Mater. Trans. 46 (2005) 18251832.
44) K. Suzuki, S. Kano, M. Kajihara, N. Kurokawa and K. Sakamoto:
Mater. Trans. 46 (2005) 969973.
45) T. Takenaka, M. Kajihara, N. Kurokawa and K. Sakamoto: Mater. Sci.
Eng. A 406 (2005) 134141.
46) Y. Muranishi and M. Kajihara: Mater. Sci. Eng. A 404 (2005) 3341.
47) T. Takenaka and M. Kajihara: Mater. Trans. 47 (2006) 822828.
48) M. Mita, K. Miura, T. Takenaka, M. Kajihara, N. Kurokawa and K.
Sakamoto: Mater. Sci. Eng. B 126 (2006) 3743.
49) Y. Yato and M. Kajihara: Mater. Trans. 47 (2006) 22772284.
50) T. Takenaka, M. Kajihara, N. Kurokawa and K. Sakamoto: Mater. Sci.
Eng. A 427 (2006) 210222.
51) Y. Yato and M. Kajihara: Mater. Sci. Eng. A 428 (2006) 276283.
52) T. Hayase and M. Kajihara: Mater. Sci. Eng. A 433 (2006) 8389.
53) Y. Tanaka, M. Kajihara and Y. Watanabe: Mater. Sci. Eng. A 445446
(2006) 355363.
54) A. Furuto and M. Kajihara: Mater. Sci. Eng. A 445446 (2006) 604
610.
55) D. Naoi and M. Kajihara: Mater. Sci. Eng. A 459 (2007) 375382.
56) S. Sasaki and M. Kajihara: Mater. Trans. 48 (2007) 26422649.
57) K. Mikami and M. Kajihara: J. Mater. Sci. 42 (2007) 81788188.
58) M. Kajihara and T. Takenaka: Mater. Sci. Forum 539543 (2007)
24732478.
59) M. Kajihara and T. Sakama: Proc. 13th Symp. Microjoining Assembly
Tech. Electrn., Yokohama, Japan, Feb. 12, 2007, (Microjoining
Comm., Tokyo, 2007) 187192.
60) T. Sakama and M. Kajihara: J. Alloy. Compd. (2008), doi: 10.1016/
j.jallcom.2008.07.113, in press.
61) M. Kajihara and S. Tanaka: Proc. 14th Symp. Microjoining Assembly
Tech. Electrn., Yokohama, Japan, Feb. 56, 2008, (Microjoining
Comm., Tokyo, 2008) 95100.
62) M. Kajihara: Acta Mater. 52 (2004) 11931200.
63) Metals Data Book, ed. Japan Institute of Metals (Maruzen, Tokyo,
1993) pp. 23317.
64) P. S. Kumar and P. K. Nair: J. Mater. Proc. Tech. 56 (1996) 511520.
65) Y. Gao, Z. J. Zheng, M. Zhu and C. P. Luo: Mater. Sci. Eng. A 381
(2004) 98103.
66) M. Palaniappa and S. K. Seshadri: Mater. Sci. Eng. A 460461 (2007)
638644.
67) T. B. Massalski, H. Okamoto, P. R. Subramanian and L. Kacprzak:
Binary Alloy Phase Diagrams, vol. 3 (ASM International, Materials
Park, OH, 1990) p. 2835.
Park, OH, 1990) p. 2864.
Park, OH, 1990) p. 2985.
70) M. Kajihara: Mater. Sci. Eng. A 403 (2005) 234240.
71) M. Kajihara: Mater. Trans. 46 (2005) 21422149.
72) M. Kajihara: Def. Di. Forum 249 (2006) 9196.
74) Y. Tanaka and M. Kajihara: Mater. Trans. 47 (2006) 24802488.
75) M. Kajihara and T. Yamashina: J. Mater. Sci. 42 (2007) 24322442.
77) A. Furuto and M. Kajihara: Mater. Trans. 49 (2008) 294303.

#2009 The Japan Society for Technology of Plasticity
Determination of True StressStrain Curves of Sheet Metals

in Post-Uniform Elongation Range*1
Kenichi Hasegawa*2 , Zhong-Chun Chen*3 , Keiou Nishimura*4 and Keisuke Ikeda
Department of Metallurgy, School of Engineering, Tohoku University, Sendai 980-8579, Japan
A novel approach to determination of the true stressstrain relation in a post-uniform elongation range by nite element modeling has been
proposed, on the basis of the variation of strain-hardening exponent (n) with strain during uniaxial tensile tests. The strain dependence of n-value
can be expressed by a linear or simple quadratic function. The proposed method has been applied to several sheet metals with large dierences in
strain-hardening behavior. The nite element analytical results, obtained by using constitutive equations derived from a variation of n-value with
strain, are in good agreement with the measured loadelongation curves. As the number of parameters is sucient by two at the most, the
optimization by an inverse analysis becomes very easy. It has been veried that the proposed method is eective to estimate the stressstrain
curves of sheet metals in the post-uniform elongation range. [doi:10.2320/matertrans.P-MRA2008838]
(Received April 22, 2008; Accepted September 29, 2008; Published November 19, 2008)
Keywords: tensile test, post-uniform elongation, strain dependence of n-value, constitutive equations, nite element modeling
1.
Introduction
In nite element (FE) simulations of sheet metal forming

with large strain deformation, such as press forming and deep
drawing, it is necessary to establish appropriate constitutive
equations that are applicable to both uniform and postuniform elongation ranges. A stressstrain relation can be
easily obtained by uniaxial tensile tests. A correct true stress
strain curve, however, is achieved only in the uniform
elongation range. It is dicult to directly obtain a true stress
strain relation in the post-uniform elongation range from the
results of tensile tests, due to occurrence of plastic instability
and ow localization1) beyond the onset of necking after a
maximum load.
So far, a lot of research29) has concerned with the necking
phenomenon during tensile tests. The well known Bridgman
equation2,3) can be used for necking correction for round
tensile specimens, which is based on a dimensional change
in the necking section after diuse necking starts. This
correction method, however, requires continuous measurements of the diameter and curvature radius of the neck during
the tests. Needleman6) and Le Roy et al.7) investigated the
necking of axisymmetric specimens by FE analysis. Recently, Zhang et al.,10,11) Scheider et al.,12) and Bacha et al.13)
proposed several methods to determine true stressstrain
curves from tensile specimens with a rectangular cross
section. Nevertheless, practically, a simple extrapolation
method is often used to describe the stressstrain relation in
the post-uniform elongation range, where the stressstrain
curve prior to necking is expressed by the power-hardening
law or multinomial. Obviously, the resultant stressstrain
curve is not guaranteed to be correct, and there may be a large
discrepancy between the experimental and analytical results.
Furthermore, in some case, it is dicult to estimate the
*1This
Paper was Originally Published in Japanese in J. JSTP 49 (2008)

143147.
*2Graduate Student, Tohoku University. Present address: Dentsu Inc.,
Tokyo, Japan
*3Corresponding author, E-mail: czc@material.tohoku.ac.jp
*4Present address: Visual Technology, Inc., Tokyo, Japan
stressstrain relation by the extrapolation method. For

example, two tempered Cu alloys, Cu-Ni-Si and Cu-Sn-P,
show highly dierent deformation behavior14) during tensile
tests. The former exhibits large uniform and small postuniform elongations, while the latter (phosphor bronze) has a
relatively large post-uniform elongation. Although both have
almost the same total elongation values, in the case of Cu-SnP alloy, the majority of the total elongation consists of the
post-uniform elongation. This dierence in deformation behavior cannot be estimated through the extrapolation by the
power-hardening law. To improve the accuracy in analyzing
sheet forming, it is essential to correctly describe a true
stressstrain relation in the post-uniform elongation range.
With the help of FEM, it is possible to determine
constitutive equations through an inverse analysis which
makes a loadelongation or loadnominal strain curve to
agree with measured data. Some researchers15,16) have
attempted to give a unied expression for stressstrain
relations of various materials under dierent loading conditions. It is not easy to reect the inuence of material
factors, even in the case of uniaxial tensile tests at room
temperature. In order to express constitutive equations
accurately, one has to increase the number of parameters
and, of course, take a long time for optimization. From the
viewpoint of practical use, it is expected to establish
constitutive equations with less optimization parameters.
For annealed metals, the strain-hardening behavior is
commonly approximated by a power-law relationship with a
constant strain-hardening exponent (n):
K"n
where is the true stress, " is the true strain, and K is the
strength coecient. If n-value is considered as a variable
depending on strain, various stressstrain relations may be
described. In fact, the strain-hardening exponent of a metal is
of strain dependence to some extent. A lot of metals show
continuously varying values of strain-hardening exponent
during plastic deformation.1719) It has been found that some
metals exhibit monotonous variations of n-value with strain,
such as linear or parabolic shape.20) The variation of n-value
Determination of True StressStrain Curves of Sheet Metals in Post-Uniform Elongation Range
can be easily calculated from measured tensile load

elongation curves. In addition, a relation of " n is valid
at the maximum load, even if n-value varies with deformation. In this way, uniform and post-uniform elongation ranges
can be clearly distinguished from each other. Accordingly, if
the variation of n-value with strain is approximated by simple
functions, various constitutive equations may be expressed
with less parameters and, hence, the optimization becomes
easy.
In the current work, we propose a method to determine the
true stressstrain relations in the post-uniform elongation
range through an inverse analysis by FEM, on the basis of
measured change of n-value with strain in the uniform
elongation range. Besides, the validity of the proposed
approach is veried by applying it to several representative
sheet metals.
2.
Strain Dependence of n-Value and Its Relation with

Constitutive Equations
Assuming that n-value varies with strain, the relation

between strain-hardening coecient (d=d") and n-value
can be expressed as
1 d n"
d"
"
Equation (2) is originally derived from the power-hardening

law (eq. (1)) in which n-value remains unchanged during the
deformation. The variation of n-value with strain is generally
calculated by considering both strength coecient K and
n-value as constants in an innitesimal strain range. Hence,
the same relation is eective even in the case of a variation
of n-value.
If n" varies with strain in a simple proportional relation
given by
n" "h al " "h
3a
where "h is the uniform strain and al is a parameter.

Substituting eq. (3a) into eq. (2) and then integrating it, the
stressstrain relation can be written as
1al "h
"
3b
h expal " "h
"h
where h is the stress at the uniform strain, i.e., stress
corresponding to the maximum load. If al 0, the above
equations are equivalent to the power-hardening law, whereas if al 1, n" " continues, no necking growth occurs
and the deformation takes place under a constant tensile load.
Similarly, supposing that n" varies in a gradual parabolic
mode, n" and can be expressed as
n"
ap " "h "h

1 "h "
h 1 "h "
ap "h
1"h
4a
1ap "h
" 1"h

"h
4b
where ap is a parameter. Equations (4a) and (4b) are

consistent with the power-hardening law when ap "h . If
ap 1, linear hardening behavior is shown, that is, the stress
is proportional to the strain.
139
In fact, n" does not always follow eq. (3a) or eq. (4a)
which contains only one parameter. If n" is expressed as a
quadratic equation of the strain, the following equations can
be written.
n" "2
h exp2" "h

12"h "h

"
exp "h 2 "2
2
"h
5a
5b
In eq. (5a), there are three parameters, , , and . Since

"h n"h at the maximum load, in eq. (5a) can be
removed. Thus, two parameters, and , remain, which can
be used to represent various variations of n" with strain.
If n" in a uniform elongation range is known, to a rst
approximation, it is supposed that the same variation trend
continues in the subsequent post-uniform elongation range.
As a consequence, the initial value of the parameter al or ap
in eq. (3) or eq. (4) can be determined immediately. Because
n" does not always keep the same relation with strain during
the deformation up to fracture, it is expected that the accuracy
of a loadelongation curve, predicted by FEM, may turn
worse from some point within the post-uniform elongation
range. In that case, the parameters can be optimized by
replacing eq. (3) or eq. (4) by eq. (5) completely, or changing to eq. (5) around the strain at which the accuracy starts to
become worse. Moreover, there exist some materials whose
uniform strain is extremely small and the variation tendency
of n" is unknown. Some examples are given in Section 4.
3.
Experimental and Analytical Methods
Four kinds of sheet metals with large dierences in strainhardening behavior, annealed high-tensile strength steel
(HTSS), pre-strained HTSS, annealed -brass (C2600), and
tempered phosphor bronze (C5191), were used in this
investigation. The dimensions and tensile elongation values
of the specimens used in the current work are summarized in
Table 1. A 4% elongation pre-strain was added to some
annealed HTSS sheets along the rolling direction (RD) by
uniaxial tensile tests. The tensile tests were conducted with a
constant crosshead speed of 10 mm/min. An extensometer
with a gage length of 50 mm was used to measure the
displacement during tensile tests.
The stressstrain curves were derived from loadelongation curves which were corrected to take into account the
elastic deformation of the testing machine. The strainhardening exponent, n, was calculated by using the strain,
stress, and strain-hardening coecient at each strain via
eq. (2). The strain-hardening coecient was obtained by
approximating the practical stressstrain curves in the uniform elongation range using a multinomial expression and
then making a least-squares treatment.
A commercially available FE code MARC was used for
analysis in the current work. It was known from the
preliminary analysis that there was almost no dierence in
solutions between solid and plane elements. Consequently,
plane elements were used for saving calculation time, and
two-dimensional rigid-plastic analysis was carried out.
Considering the symmetry of sheet specimens, 1/4 of
140
K. Hasegawa, Z.-C. Chen, K. Nishimura and K. Ikeda

Table 1 Dimensions and tensile elongation values of specimens.
Annealed
HTSS
Pre-strained
HTSS
Annealed
-brass
Tempered
Dimensions
L W T (mm)
Tensile
direction
Uniform
elongation (%)
Post-uniform
elongation (%)
Total
elongation (%)
120 12:5 1:6
TD
19.6
10.0
29.6
120 12:5 1:6
TD
14.5
11.8
26.3
190 25 0:29
RD
36.3
8.1
44.4
RD
5.7
9.4
15.1
TD
2.2
11.1
13.3
190 25 0:2
phosphor bronze
900
12.4
800
Stress, / MPa
Load, P / kN
12
11.6
Experimental
n: parabolic (ap=0.2), Eq.(4)
Power-hardening law
n: quadratic (=2.51, =0.22), Eq.(5)
n=
11.2
10.8
0.2
0.225
0.25
0.275
0.5
700
0.4
n=
600
0.3
500
0.2
n-value
400
300
0
0.3
Nominal strain, e
Fig. 1 Tensile loadnominal strain curves for annealed high-tensile
strength steel sheets.
parallel region was used as an analysis model. In addition, an

initial imperfection was given by making the width in the
central section to be 0.5% smaller than that in the parallel
region to ensure occurrence of necking at the central section.
The parameter values estimated from the variation of n" in
the uniform elongation range were regarded as the initial
values. The optimal constitutive equations were determined
by modifying the parameter values until the dierences
between the FE analytical results and measured load
nominal strain data were less than 0.5%.
4.
0.6
Power-hardening law
Estimated by Eq.(5)
(=2.51, =0.22)
n-value
Specimen
The representative results by applying the proposed

method to annealed and pre-strained HTSS, annealed brass, and tempered phosphor-bronze sheets are described as
follows.
4.1 Annealed HTSS sheets
The steel sheet used in this work was one of the HTSS
sheets with a uniform elongation of 19.6% and total
elongation of 29.6% (Table 1). The n" in the uniform
elongation range showed a slight increase trend except for the
initial period of deformation, and it was well approximated
using a parabolic mode with ap 0:20 (eq. (4a)). Since the
change of n" was very small, the power-hardening law with
n 0:18 exhibited a similar stressstrain curve. Figure 1
0.1
0.1
0.2
0.3
Strain,
0.4
0
0.5
Fig. 2 Stressstrain curves and variation of n-value with strain for

annealed high-tensile strength steel sheets. The arrow indicates the
position of the maximum load.
shows a comparison of the tensile loadnominal strain curves

between the FE analytical results and experimental data. The
result based on the power-hardening law is also given in the
gure. The beginning of the curve by eq. (4) was in good
agreement with the experimental curve, but the analytical
values were higher than the measured ones from the position
where the load rapidly decreased. By using eq. (5) and
making an optimization treatment, the analytical and experimental results were extremely consistent with each other
within the whole strain range as shown in Fig. 1. The values
of the optimized parameters were 2:51 and 0:22.
Figure 2 shows the true stressstrain curves and variation
of n" with strain. The values in the uniform elongation
range, plotted in the gure, were achieved directly from the
measured loadnominal strain data. After the diuse necking
starts, at rst almost no change of n" was observed.
Subsequently, because of the decrease in n" with increasing
the strain, the stressstrain curve, obtained from the powerhardening law, was overestimated. However, the dierence
was not so large, and a quite good approximation could be
achieved even using the power-hardening law. In addition to
HTSS sheets, other annealed metals with large n-values also
exhibit good approximations of stressstrain curves when the
power-hardening law is used (see Section 4.3). On the other
hand, it was observed from the stressstrain curves shown in
Fig. 2 that the stress increase due to strain-hardening became
gentle in the post-uniform elongation range, although a large
0.7
900
800
11.6
11.2
10.4
0.15
Experimental
n: quadratic ( =2.04, =0.295), Eq.(5)
n=
0.2
Stress, / MPa
Load, P / kN
12
Power-hardening law
Estimated by Eq.(5)
(=2.04, =0.295)
0.6
700
0.5
n=
0.4
600
0.3
500
0.2
400
n-value
0.1
300
200
0
n-value
12.4
10.8
141
0.1
0.2
0.25
0.3
Strain,
0.4
0.5
0.6
Nominal strain, e
Fig. 3 Tensile loadnominal strain curves for pre-strained (4%) hightensile strength steel sheets.
Fig. 4 Stressstrain curve and variation of n-value with strain for prestrained high-tensile strength steel sheets. The arrow indicates the position
of the maximum load.
4.2 Pre-strained HTSS sheets

The annealed HTSS sheets described above were uniaxially pulled in rolling direction (RD) to incorporate a 4% prestrain, followed by tensile test in transverse direction (TD).
Because of the incorporation of the pre-strain, the uniform
and total elongation values were reduced to 14.5 and 26.3%,
respectively, but the post-uniform elongation was somewhat
increased (Table 1). The n" showed a minimum value in the
initial period of deformation, and then increased with strain
parallel to the straight line of n " (see Fig. 4). Nevertheless, similar to the annealed HTSS sheets, the pre-strained
specimens can be approximated by a parabolic change of n"
with ap 0:65, except for the early period of deformation.
Figure 3 shows the tensile loadnominal strain curves
obtained from FE analysis and experiments. The analytical
results by a parabolic approximation (eq. (4)) were overestimated in the latter part of the curve. When the variation of
n" was expressed by a quadratic equation (eq. (5)), the
analytical results were in good agreement with the experimental data.
Figure 4 illustrates the stressstrain curves and variation of
n", when modied constitutive equations with a good
approximation were used. The n" increased with increasing
the strain and gradually closed to the value of the annealed
specimen. After passing through a peak, the n" showed a
variation trend similar to the annealed specimen (Fig. 2).
Corresponding to the change of n", the stress rose linearly
as the strain increased, and gradually closed to the stress
value of the annealed specimen. The strain at which n"
reached a maximum value (" 0:295) was larger than that of
the annealed specimen, thus leading to a larger post-uniform
elongation. Due to a large change of n", it is clear that the
power-hardening law is not applicable to express the stress
strain relation of pre-strained HTSS sheets in the postuniform elongation range. If the n-value in the uniform
elongation range is used, the ow stress is necessarily
underestimated, as shown in the dotted curve in Fig. 4.
Load, P / kN
strain-hardening feature was shown in the uniform elongation

range. This is attributed to the decrease of n-value with strain
after necking.
2.95
2.9
Experimental
Power-hardening law
n: linear (al =-0.45), Eq.(3)
n: quadratic (=5.12, =0.27), Eq.(5)
n=
0.375
0.4
0.425
0.45
Nominal strain, e
Fig. 5 Tensile loadnominal strain curves for annealed -brass (C2600)
sheets. Tensile test was carried out in rolling direction.
4.3 Annealed -brass sheets

As another example with large n-values, annealed -brass
sheets were examined. With regard to the materials with large
n-values, it is necessary to describe the variation slope of n"
up to the maximum load. -brass showed serrations in the
stressstrain curve until the maximum load, and thus there
existed large scatters of n" values (see Fig. 6). On the
whole, however, it seemed that n" showed a decrease trend
(represented by a thin line in Fig. 6) up to the maximum load
as the strain increased. The constitutive equations can be
derived by using the variation slope of n", and the FE
analytical results are shown in Fig. 5. When al 0:45 in
eq. (3), the analytical results coincided with the experimental
values up to around 2/3 of the post-uniform elongation.
Subsequently, the dierence between the analytical and
experimental results became large with increasing the strain.
Similarly, no good agreement was found under the conditions
of ap 0:45 by eq. (4). To improve the analysis accuracy,
it is assumed that n" varies in a quadratic curve. When the
resulting constitutive equations were used, the analytical
results were in good agreement with the measured values as
shown in Fig. 5. In this case, the corresponding values of
142
700
1000
n=
400
300
0.6
0.4
200
n-value
100
0
0
0.1
0.2
0.3
0.4
0.5
Stress, / MPa
0.8
500
n-value
Stress, / MPa
600
Power-hardening law
Estimated by Eq.(5)
0.2
0
0.6
4.4 Tempered phosphor-bronze alloy sheets

Tempered phosphor-bronze alloy sheets have a feature of
small uniform elongation and large post-uniform elongation
as mentioned previously. The values of the uniform and total
elongations in RD were 5.7 and 15.1%, respectively, while
those in TD were 2.2 and 13.3%, respectively (Table 1).
During the tensile tests in TD, the n" was rapidly reduced to
an extremely low value in the uniform elongation range (see
Fig. 9). Due to a large post-uniform elongation, it is expected
that n" turns to increase beyond the onset of necking, but
the variation trend of n" is unclear. As a consequence, the
variation of n" of tempered phosphor-bronze alloy sheets
after the maximum load cannot be described by the variation
slope of n" in the uniform elongation range.
On the basis of a strain distribution in the tensile direction
when fractured, the authors21) have derived simplied
equations for estimating the true stress and true strain in a
post-uniform elongation range as follows:

1
Ei

6a
i P
Ah A0 1 exp

A0
Ei
"i ln
6b
Ah 1 exp
700
0.1
0.2
0.3
0.4
0.5
Strain,
Fig. 7 Stressstrain curves for tempered phosphor-bronze alloy (C5191)
sheets.
3.4
3.2
Load, P / kN
and in eq. (5) were 5.12 and 0.27, respectively. Besides, the
tensile loadnominal strain curve with a constant n-value was
also shown in the gure, and the load values were overestimated.
Figure 6 shows the stressstrain curves and variation of
n" of annealed -brass sheets, when modied constitutive
equations showing a good approximation of loadnominal
strain curves were used. The n" values in the uniform
elongation range were relatively large. It is believed that a
smaller post-uniform elongation in annealed -brass sheets is
the result of a large reduction in n" with strain. In
comparison with two stressstrain curves obtained from the
power-hardening law and modied equation using eq. (5),
the dierence was very small. In other words, the powerhardening law can give a moderate approximation for those
materials with large n-values. This is similar to the results of
the annealed HTSS sheets described in Section 4.1.
800
600
Strain,
annealed -brass (C2600) sheets. Tensile test was carried out in rolling
direction. The arrow indicates the position of the maximum load.
900
Estimated by Eq.(6)
n: linear (al=0.55), Eq.(3)
Power-hardening law
n=
3
2.8
2.6
2.4
2.2
0.025
Experimental
Power-hardening law
Estimated by Eq.(6)
n: linear (al=0.55), Eq.(3)
n=
0.05
0.075
0.1
0.125
Nominal strain, e
Fig. 8 Tensile loadnominal strain curves for tempered phosphor-bronze
alloy (C5191) sheets. Tensile test was carried out in transverse direction.
where i is the true stress, "i the true strain, P the tensile load,
A0 the initial cross-sectional area of a specimen, Ah the
minimum cross-sectional area in a uniform elongation range,
i.e., the area corresponding to the maximum load, Ei the
nominal post-uniform strain, and the parameter representing the degree of strain localization. Equations (6a) and (6b)
are obtained by calculating the minimum cross-sectional area
after necking. Because the shape eect of a necking section is
not taken into consideration, eq. (6) causes an overestimation
of the stress. Nevertheless, a reasonable stressstrain curve
can be achieved in a short period after the necking starts.
To a rst approximation, the parameter al in eq. (3) can be
determined to make the results obtained by eq. (6) to agree
with eq. (3). If the estimated results are not consistent with
the measured loadnominal strain data, they can be modied
by eq. (4) or eq. (5). A comparison of some stressstrain
curves is given in Fig. 7. The dotted line in Fig. 7 is the
stressstrain curve predicted by eq. (6), whereas the solid
line corresponds to the results from eq. (3), and al 0:55
gives a good approximation. It is seen from Fig. 8 that the
tensile loadnominal strain data obtained by FEM are in good
accordance with the experimental results. The determined
stressstrain curve and variation of n" are shown in Fig. 9.
900
800
0.6
700
0.4
n=
600
n-value
500
400
0
0.1
0.2
0.3
0.4
0.2
0
0.5
Stress, / MPa
0.8
n-value
Stress, / MPa
900
1000
Power-hardening law
Estimated (al =0.55) by Eq.(3)
0.8
Power-hardening law
Estimated (ap=0.60) by Eq.(4)
0.6
800
700
0.4
n=
n-value
1000
143
600
n-value
500
400
0
0.1
0.2
0.3
0.4
0.5
0.2
0
0.6
Strain,
Strain,

tempered phosphor-bronze alloy (C5191) sheets. Tensile test was carried
out in transverse direction. The arrow indicates the position of the
maximum load.
Fig. 11 Stressstrain curve and variation of n-value with strain for

tempered phosphor-bronze alloy (C5191) sheets. Tensile test was carried
out in rolling direction. The arrow indicates the position of the maximum
load.
Load, P / kN
3.2
3.1
2.9
2.8
Experimental
Power-hardening law
Estimated by Eq.(6)
n=
0.075
0.1
0.125
0.15
Nominal strain, e
Fig. 10 Tensile loadnominal strain curves for tempered phosphor-bronze
alloy (C5191) sheets. Tensile test was carried out in rolling direction.
As a result of the increase in n" immediately after the

uniform elongation, the stress increased with increasing the
strain almost linearly. The tensile stress values, estimated by
the power-hardening law using the n-value at the maximum
load, are much lower than the experimental ones. This is one
of the typical examples indicating that the stressstrain
relation in a post-uniform elongation range cannot be
approximated by the power-hardening law.
With regard to the tensile tests in RD, the loadnominal
strain and stressstrain curves are shown in Figs. 10 and 11,
respectively. Although the uniform elongation value was
small, the n" revealed a change tendency from decrease to
increase with strain immediately over a straight line of n ".
The loadnominal strain curve, predicted from a parabolic
approximation with ap 0:60, was consistent with the
experimental curve. In comparison with Fig. 8 and Fig. 10,
the latter looks like somewhat larger dierences between the
predicted and experimental results. This is because the load
axis in Fig. 10 was magnied with respect to that in Fig. 8.
The error was actually less than 0.3%. Moreover, in the case
of the tensile tests in RD, the variation behavior of n-value
was similar to that in TD, except that the uniform elongation
in RD was slightly larger because of the reversal of n-value

over the straight line of n ".
When the tempered phosphor-bronze alloy sheets were
pulled in RD and TD, although there existed some dierences
in variation of n" (either parabolic or straight line), a similar
change trend of n" from decrease to increase has been
observed. However, the strain amount where n" reached the
minimum value (nmin ) was dierent, nmin > " in RD and
nmin < " in TD. This may result in the dierence in the
uniform elongation. It is well known22,23) that the postuniform elongation also depends on strain-rate sensitivity
(m) in addition to n-value. The m-value of the tempered
phosphor-bronze alloy sheets at room temperature is so small
that it can be ignored. In fact, no dierence in necking growth
has been found even though the inuence of m-value was
taken into consideration in the FE analysis. Large values of
the post-uniform elongation in both RD and TD are likely to
result from a continuous increase of n", which restrains the
growth of necking.
From the above results, if a sheet specimen has a larger
uniform elongation, just as annealed -brass, annealed and
pre-strained HTSS specimens as mentioned above, the
constitutive equations with an initial parameter value of al
or ap in eq. (3) or eq. (4) can be obtained by using the
variation slope of n" in the uniform elongation range. If
necessary, the analysis accuracy can be further improved by
using a quadratic mode (eq. (5)). As for the materials with a
small uniform strain or their variation of n" is unknown, for
example, tempered phosphor-bronze alloy sheets described
above, the variation of n" immediately after the start of
necking can be estimated from the stressstrain relation
calculated by eq. (6). In this way, the constitutive equations
are determined to make them to agree with the estimated
variation of n". Because almost no necking growth occurs
immediately after the maximum load, the estimated slope
from calculated stressstrain relation is believed to be
reliable. The subsequent optimization can be done in a way
similar to the materials with a larger uniform elongation.
These suggest that the proposed approach in the current
work can be applied to various materials with dierent
strain-hardening behavior.
144
The above results indicate that the dierences in tensile

deformation behavior in annealed, tempered, and prestrained materials are reected by variations of n" distinctly. It should be pointed out that the eects of m and r
(a plastic anisotropy parameter) values on a post-uniform
elongation are not considered in this FE analysis. The
inuences of m-value, r-value, and even microstructure of
materials, however, if there exist, must be reected in
measured tensile loadnominal strain curves. Accordingly, it
is reasonable to consider that the eects of various factors on
a stressstrain relation are included in an optimized n". It is
important to quantitatively describe the eects of abovementioned factors on both magnitude of ow stress and
dierence in strain-hardening behavior in the post-uniform
elongation range.
5.
Conclusions
An approach to estimation of the true stressstrain relation

in a post-uniform elongation range through an inverse
analysis by FEM has been proposed. The analysis is based
on the dependence of n-value on strain during uniaxial tensile
tests. The proposed method has been applied to several sheet
metals with dierent strain-hardening behavior. The results
show that the strain dependence of n-value can be expressed
by a linear or simple quadratic function for various materials
(such as annealed, tempered, and pre-strained metals) with
large dierences in strain-hardening behavior. If there is a
larger uniform elongation, the constitutive equations with a
good approximation can be obtained by using the variation
slope of n" in the uniform elongation range. With regard to
the materials with a small uniform strain or their variation of
n" is unknown, the variation of n" immediately after the
start of necking can be estimated from the stressstrain
relation calculated by the derived equations.
FE solutions can be obtained by using constitutive
equations derived from a variation of n-value with strain. It
has been shown that the FE analytical results are in good
agreement with the measured loadnominal strain curves.
The proposed method is simple with only one or two
parameters, and can be applied to various metals with
dierent strain-hardening behavior. Moreover, the powerhardening law with a constant n-value gives a moderate
approximation for annealed metals with large n-values, but
large errors occur when it is used to deal with pre-strained
and tempered metals.
REFERENCES
1) S. L. Semiatin and J. J. Jonas: Formability and Workability of Metals,
(ASM, Ohio, 1984).
2) P. W. Bridgman: Trans. Am. Soc. Metals 32 (1944) 553573.
3) P. W. Bridgman: Studies in Large Plastic Flow and Fracture,
(McGraw-Hill, New York, 1952).
4) N. N. Davidenkov and N. I. Spiridonova: Proc. ASTM 46 (1946) 1147
1158.
5) E. R. Marshall and M. C. Shaw: Trans. Am. Soc. Metals 44 (1952)
705721.
6) A. Needleman: J. Mech. Phys. Solids 20 (1972) 111127.
7) G. Le Roy, J. D. Embury, G. Edwards and M. F. Ashby: Acta Metall. 29
(1981) 15091522.
8) V. Tvergaard and A. Needleman: Acta Metall. 32 (1984) 157169.
9) C. Garca-Garino, F. Gabaldon and J. M. Goicolea: Finite Elem. Anal.
Des. 42 (2006) 11871197.
10) Z. L. Zhang, M. Hauge, J. degard and C. Thaulow: Int. J. Solids
Struct. 36 (1999) 34973516.
11) Z. L. Zhang, J. degard, O. P. Svik and C. Thaulow: Int. J. Solids
Struct. 38 (2001) 44894505.
12) I. Scheider, W. Brocks and A. Cornec: J. Eng. Mater. Technol. 126
(2004) 7076.
13) A. Bacha, D. Daniel and H. Klocker: J. Mater. Process. Technol. 184
(2007) 272287.
14) K. Hatakeyama: Ph.D. thesis (Tohoku University, 2003).
15) M. Gotoh: J. Japan Soc. Technol. Plast. 43 (2002) 270277.
16) M. Kuroda and K. Shizawa: J. Japan Soc. Technol. Plast. 43 (2002)
299309.
17) K. Saka, M. J. Painter and R. Pearce: J. Mech. Work. Technol. 3 (1979)
1730.
18) M. De Meyer, D. Vanderschueren and B. C. De Cooman: ISIJ Int. 39
(1999) 813822.
19) L. Gu, A. Tanaka, K. Tanaka and K. Nakajima: Tetsu-to-Hagane 88
(2002) 8894.
20) K. Hatakeyama, Y. Yamashita and K. Ikeda: J. JRICu 41 (2002)
241245.
21) K. Hasegawa, K. Nishimura, Z. Chen and K. Ikeda: Proc. 55th Jpn.
Conf. Technol. Plast. (2004) 275276.
22) E. W. Hart: Acta Met. 15 (1967) 351355.
23) A. K. Ghosh: J. Eng. Mater. Technol. 99 (1977) 264274.

Weibull Statistics for Evaluating Failure Behaviors and Joining

Reliability of Friction Stir Spot Welded 5052 Aluminum Alloy
Chung-Wei Yang1 , Fei-Yi Hung2 , Truan-Sheng Lui1; * , Li-Hui Chen1 and Jiun-Yu Juo1
1
Department of Materials Science and Engineering, National Cheng Kung University,

No. 1, University Road, Tainan 701, Taiwan, R.O.C.
2
Institute of Nanotechnology and Microsystems Engineering, Center for Micro/Nano Science and Technology,
National Cheng Kung University, No. 1, University Road, Tainan 701, Taiwan, R.O.C.
Friction stir spot welding (FSSW) was applied to make lap-joints of 5052 rolled (5052-R) aluminum alloys. The resulting microstructural
observation, microhardness, tensile shear failure load and failure morphologies are reported, including a brief look into failure behaviors in the
present study. The metallurgical bonded TMAZ region with a plastic metal ow is obviously created around the probe, and the microhardness is
signicantly increased at the TMAZ region for FSSW-joined 5052 aluminum alloys. The FSSW lap-joints with an obvious metallurgical bonded
TMAZ region generally display a higher failure load and ductile failure morphologies with dimples fracture. In addition, the failure load of
FSSW lap-joints was increased with increasing the probe penetration depth and the welding time. Based on the data uctuation of tensile shear
failure load, the Weibull model provided a statistical analysis method for assessing the minimum failure, the failure mechanism and the joining
reliability for the FSSW lap-joints. Through the statistical analysis of the Weibull distribution function, FSSW-joined aluminum alloys with a
wear-out failure model are recognized as reliable lap-joints for further engineering application. [doi:10.2320/matertrans.MRA2008341]
Keywords: friction stir spot welding, aluminum alloy, mechanical properties, failure mechanism, Weibull statistics
1.
Introduction
Lightweight aluminum alloys are widely used in automotives for the purpose of reducing vehicle weight and fuel
consumption. Resistance spot welding (RSW) is generally
applied for joining steels in the automotive industries of
nowadays. However, RSW process is not suitable for joining
aluminum alloys1) because of its high electrical current
requirements and the inconsistent quality of the nal
welds.2,3) Recently, a new solid-state joining technology of
the friction stir spot welding (FSSW), which is a derivative
process of the friction stir welding (FSW),4) has been
developed as a widespread technique and successfully
applied for producing lap-joints of aluminum alloys.57)
Compared with the traditional RSW process, FSSW can
avoid severe heating and cooling cycles induced during
welding process. Furthermore, the heat aected zone (HAZ)
and residual stresses associated with the welds are relatively
small.8) Signicant performance advantages of the solid-state
approach have made FSSW the preferred method for joining
aluminum alloys.9) Therefore, FSSW as well as FSW is now a
simple and indispensable process to acquire better joining
strength and vibration fracture resistance for the lightweight
structural metals.6,7,10,11)
Since the failure of FSSW-joined structural components
depends on the stress to approach a critical weakest link of
the welding spot, the variability of failure strength is fairly
correlated with the welding qualities of joints, including
microstructural homogeneity, bond formation and bond
width.1214) However, few current reports focus on assessing
the failure mechanism of FSSW aluminum alloys under
mechanical loading for various probe penetration depth.
Furthermore, the eect of microstructural feature on failure
probability and joining reliability of FSSW lap-joints still
*Corresponding
author, E-mail: Rei@mse.ncku.edu.tw
needs further investigation through a statistical examination.

Since the Weibull statistics15) of survival analysis has been
developed as an engineering design method for metals,
metallic glasses and ceramic coatings,1620) therefore, the aim
of present study is to evaluate the weldability and failure load
of FSSW-joined aluminum alloy under tensile shear tests.
The joining reliability and failure mechanism will be
examined by the Weibull statistics.
2.
Materials and Methods
The base-metal used in this study was a 2 mm-thick 5052H34 aluminum rolled sheets (5052-R) with chemical composition of 2.58 Mg, 0.26 Cr, 0.15 Si, 0.05 Mn, and 0.07 Zn
(mass%). The sheets were machined into welding samples
with dimensions of 72 mm (l) 30 mm (w). Figure 1(a)
schematically shows a lap-shear specimen used to measure
the strength of FSSW-joined 5052-R sheets under a tensile
shear loading condition. During FSSW process, a cylindrical
rotating tool with a protruding probe plunged into the
overlapping 5052-R aluminum sheets. The downward push
force and the rotational speed were maintained for an
appropriate time to generate frictional heat. The welding tool
rotational speed was 1040 rpm, and the downward push force
was controlled at about 2.6 kN. The shoulder and probe
diameters of the tool were 10 and 3 mm, respectively. Then
the generated frictional heat softened the aluminum basemetal and the rotating probe caused material plastic ow in
both circumferential and axial directions. The probe penetration depth was controlled for 1.0 and 2.5 mm by changing
various probe length, and the welding time was maintained
under 90, 105 and 135 s.
After welding, the lap-joined specimens were crosssectioned parallel to the welding direction for the metallographic examination. The cross-sections of the metallographic specimens were polished with diamond polishing
146
C.-W. Yang, F.-Y. Hung, T.-S. Lui, L.-H. Chen and J.-Y. Juo
rotation
(a)
shoulder
probe
30 mm
spacer
25 mm
2 mm
72 mm
15 mm
(b)
15 mm
Fig. 1 (a) A schematic plan of sheets dimension and a lap-joint specimen by the friction stir spot welding for the tensile shear test. (b) A
FSSW lap-joined specimen for measuring the failure load with the tensile shear test.
3.
3.1
Statistical evaluation of failure probability for

FSSW lap-joints with the Weibull model
Figure 2 shows the failure load data distribution (represents 20 valid results for each joining condition) and the
average value of various FSSW lap-joints measured by the
tensile shear tests. The failure load of FSSW-joined specimens with the probe penetration depth of 2.5 mm (P2.5series) and 1.0 mm (P1.0-series) is increased with increasing
the welding time. Generally, the FSSW lap-joints with larger
Average failure load
3.0
Failure load, x / kN
agent. Then the specimens were etched in a Poultons

solution (30 mL HCl + 15 mL HNO3 + 2.5 mL HF +
2.5 mL H2 O).21) The microstructural morphologies of lapjoints were observed by the optical microscopy. To clarify
the relationship between failure locations and microhardness
variations of the FSSW-joined specimens, microhardness
data across the cross-sections of specimens were obtained
using a Vickers indenter with a 980 mN load for 10 s dwell
time. Each measured microhardness data was an average of
three tests.
The failure load of FSSW 5052-R lap-joints was evaluated
by tensile shear tests at a crosshead speed of 1.0 mm/s until
failure. Figure 1(b) shows a FSSW lap-joined specimen for
the tensile shear test. The setting of two spacers, which as
represented in light gray regions in Fig. 1(a) with dimensions
of 30 mm (l) 15 mm (w) at the both ends of the lap-shear
specimens, were to induce a pure shear to the interfacial
plane of the nugget for the joined sheets. For the statistical
signicance of following Weibull analysis, twenty lap-joined
specimens of each joining condition were performed for
failure load measurements. The as-failed samples were then
examined by a SEM (Philips XL-40 FEG) to observe the
fracture surfaces of FSSW lap-joints.
2.671
2.761
2.5
2.325
2.0
1.765
1.5
P2.5-90s
P2.5-105s
P1.0-105s
P1.0-135s
Various friction stir spot welding conditions

Fig. 2 Failure load data uctuation and the average value of FSSW lapjoints for various welding conditions.
probe penetration depth (P2.5-90s and P2.5-105s) show a

higher joining failure load than the other conditions, but
there is little signicant dierence (p > 0:05) for the data
distribution between two P2.5-series joining conditions
according to the one-way ANOVA statistical examination.
However, the failure load for P1.0-series lap-joints (P1.0135s) is signicantly increased through increasing the
welding time (p < 0:05) compared with the P2.5-series
conditions. Furthermore, the failure load is decreased with
decreasing the probe penetration depth for the same welding
dwell time, and P1.0-105s lap-joints have the lowest average
failure load of all.
Previous reports represented that the Weibull statistics15,22)
can be successfully adopted to evaluate the failure probability
and the engineering reliability of materials.1720) Therefore,
the Weibull distribution function of eq. (1), which describes
Probability density function, f (x)
Weibull Statistics for Evaluating Failure Behaviors and Joining Reliability of Friction Stir Spot Welded 5052 Aluminum Alloy
(a)
P2.5-90s
P2.5-105s
P1.0-105s
P1.0-135s
2x10-3
1x10-3
900
1800
2700
3600
Failure load, x / Newton
(b)
lnln[1/1F(xi)]
1
0
-1
ln(xi xo)
R = 0.99
m = 3.3
-2
R = 0.99
m = 4.5
2
-3
R = 0.99
m = 10.5
8
2
R = 0.99
m = 8.1
P2.5-90s
P2.5-105s
P1.0-105s
P1.0-135s
Fig. 3 (a) The failure probability density function f x curves, and (b) the
Weibull distribution plots of various FSSW-joined specimens. Fxi is the
cumulative failure probability at the corresponding failure load (xi ) and the
slope represents the Weibull modulus (m).
the cumulative failure probability Fxi for each corresponding failure load data, can help to statistical clarify the joining
reliability, failure probability and the failure mechanism of
FSSW lap-joints.

Z xxi
xi xo m
f xdx 1 exp
1
Fxi
xc
x0
According to the denition of eq. (1), the failure behavior
of materials is determined by three parameters m, xc and xo .
The symbol x represents the measured failure load value in
this study, the parameter m represents the Weibull modulus
(or called the shape parameter), the characteristics failure
load xc corresponds to a value at which the cumulative
probability of failure is 0.632, and xo denotes the minimum
failure load which means the failure probability of FSSW lapjoints lower than this value is zero and the joining reliability
is 1.0. The minimum strength can be recognized as the safety
value for the FSSW lap-joints herein.
Fitting the failure load data into eq. (1), the failure
probability density function f x curves of the FSSW 5052R lap-joints are plotted in Fig. 3(a). The cumulative failure
probability Fxi was estimated using the Benards median
rank of eq. (2).22) It is a very close approximated solution of a
statistical function,22,23) where n is the total number of tested
samples and i is the sample rank in ascending order of failure
load. In addition, the joining reliability function Rxi with a
relation of Rxi 1 Fxi is dened as the survival
probability19) of lap-joints for the purpose of following
discussion. Figure 3(b) shows the double natural logarithmic
(ln) graphs (Weibull distribution plots) for the cumulative
failure probability at each corresponding failure load xi
(i 1{20), so as to graphically evaluate the Weibull
modulus from the slope of a least-squares tting method of
eq. (3) at a maximum coecient of determination (R2 ). As
can be seen in Fig. 3(b), a good linear relationship is
observed, suggesting that the experimental data can be
reasonably described by the Weibull model. The results of
Weibull statistical analysis are listed in Table 1, and the
detail of Weibull statistics on the failures and the mechanical
properties of the FSSW lap-joints will be discussed in the
next paragraph.
i 0:3
2
Fxi
n 0:4

1
m lnxi xo m ln xc
3
ln ln
1 Fxi
3.2 Joining microstructures of the FSSW lap-joints
Figure 4 shows the cross-sectional features of the FSSW
P2.5-90s and P2.5-105s lap-joined specimens with the heat
aected zone (HAZ) and thermo-mechanically aected zone
(TMAZ) created around the probe. With the severe plastic
deformation during FSSW process, evident plastic metal ow
is observed close to the HAZ and ne grains are formed
within TMAZ. The average grain size within TMAZ for
P2.5-90s and P2.5-105s is about 3:8 0:2 mm and 3:9
0:4 mm, respectively. This eect should be resulted from the
dynamic recrystallization at the stir zone during the friction
stirring process.24) In the present study, the bond width of
P2.5-series lap-joints (as shown in Fig. 4(a) and (b)) is
recognized as a region of grain rening with fully metallurgical bond according to the denition of previous study,13)
but the obvious lap-sheets interface represents that there is
almost no contact away from the bond region. Besides, an
apparent microcrack is observed close to the HAZ/TMAZ
interface as shown in Fig. 4(c) and (d), which are denoted by
triangle arrows in Fig. 4(a) and (b), respectively. This area
Table 1 Results of the failure load measurements for FSSW 5052 aluminum alloy lap-joints with various joining conditions.
Failure load,
x (kN)
Minimum failure load,

xo (kN)
Characteristics failure
load, xc (kN)
Weibull modulus,
m
P2.5-90s
2.671
0.416
2.812
8.1
P2.5-105s
2.761
1.927
2.858
4.5
P1.0-105s
1.765
1.121
1.855
3.3
P1.0-135s
2.325
0.796
2.399
10.5
Each value was the average of twenty tests (n 20).

Data were calculated from ln ln1=1 Fxi vs. lnxi xo plots using eq. (3).
147
148
(a)
(a)
HAZ
0.4 mm
bond width
TMAZ
kissing bond
0.6 mm
(b)
(c)
TMAZ
(d)
metallurgical bond
kissing bond
(b)
0.4 mm
50 m
50 m
HAZ
(c)
TMAZ
(d)
bond width
100 m
0.6 mm
Fig. 4 The cross-sectional feature of FSSW lap-joints for (a) P2.5-90s, (b)
P2.5-105s conditions (TMAZ: thermo-mechanical aected zone, HAZ:
heat aected zone). The HAZ/TMAZ interface of P2.5-90s and P2.5-105s
conditions are shown in images (c) and (d), respectively.
may become the crack propagation path as the failure

occurred during the tensile shear tests.
Figure 5(a) and (b) show the cross-sectional features of
P1.0-105s and P1.0-135s conditions, which are dierent
from P2.5-series lap-joints. For the P1.0-105s lap-joints
(Fig. 5(a)), the bonded area can be recognized just underneath the probe, and the mechanical kissing bond is observed
as shown in the high magnication image of Fig. 5(c).
Compared with the P1.0-105s lap-joints, however, the P1.0135s lap-joints display a dierent microstructural feature,
which represents a darkling TMAZ region with an indistinct
HAZ/TMAZ interface. The metallurgical bonded region can
be recognized at the TMAZ region also created around the
probe. The presence of TMAZ region for the P1.0-135s lapjoints should be resulted from a longer welding dwell time,
which provides enough friction heat to induce the metal
plastic ow and the dynamic recrystallization during FSSW
process. The average grain size within the TMAZ region of
P1.0-135s lap-joints is about 4:0 0:3 mm. Away from the
TMAZ, the kissing bond is formed as shown in Fig. 5(b) and
the high magnication image of Fig. 5(d). In addition, the
metal plastic ow occurred under this welded condition is
also distinctly observed from the concave curvature of
kissing bonded line. Since the variability of failure load
should probably be aected with the microstructural features
for dierent welding conditions, the results of the Weibull
100 m
Fig. 5 The cross-sectional feature of FSSW lap-joints for (a) P1.0-105s

and (b) P1.0-135s. The kissing bond area of P1.0-105s and P1.0-135s
conditions are shown in images (c) and (d), respectively.
statistical analysis of variance made it possible to assess the

role of the microstructure in determining the failure behavior
of FSSW lap-joints.
3.3
Microstructural variations aect the failure load

data uctuation and failure behaviors
Figure 6 displays the microhardness (Hv) proles along
the cross-sections of various FSSW-joined specimens. The
dash line represents the Vickers indenter testing area, which
is located above the lap-sheets interface. The results
represents that the variability of microhardness generally
displays a W-type data distribution throughout the crosssections for these joining conditions. The microhardness
within TMAZ region is increased and signicantly higher
than the average level of the base-metal. This phenomenon
can be resulted from the grain rening eect within the
TMAZ region as shown in Figs. 4 and 5(b). But the obviously
lower microhardness at the TMAZ region of P1.0-105s lapjoints can be recognized as slight metallurgical bond within
TMAZ than the other conditions. However, the microhardness decreased markedly at the HAZ region, which is a soften
region and shows the lowest microhardness within the crosssections of lap-joined specimens.
Considering the above-mentioned microstructural examinations, a fairly amount of data uctuation for these
specimens shown in Fig. 2 can be probably recognized as a
result of the signicant dierences of microhardness and the
microstructural inhomogeneity with various probe penetration depth and welding dwell time. Because the Weibull
P1.0
Hv
(a)
Hardness (Hv)
HAZ
TMAZ
Failure rate, (x)
P2.5
100
149
HAZ
P2.5-90s
P2.5-105s
P1.0-105s
P1.0-135s
90
Base metal
P2.5-90s
P2.5-105s
P1.0-105s
P1.0-135s
2x10-2
1x10-2
80
0
70
900
1800
2700
3600

1.0
(b)
-6
-3
0
3
Position, L / mm
Fig. 6 Microhardness (Hv) of various FSSW lap-joints. The indentations

were made with a spacing of 0.25 mm along the parallel dash line.
distribution function has been adopted to model the reliability

and the failure mechanism of materials,2527) a failure rate
function xi shown in eq. (4) at each corresponding failure
load is further dened for the assessment of failure behaviors.
Figure 7(a) and (b) show the failure rate function x and
reliability function Rx curves of FSSW lap-joints, respectively.
f xi
m
m xi xo m1
xi
Rxi xc
Reliability, R(x)
-9
0.632
0.5
P2.5-90s
P2.5-105s
P1.0-105s
P1.0-135s
0.0
900
1800
2700
3600

Fig. 7 (a) Failure rate function x curves, and (b) reliability function Rx
curves of various FSSW-joined specimens.
Referring to the analysis results listed Table 1, the

examination of the Weibull modulus (m) and failure rate
function represent that FSSW-joined 5052 aluminum alloys
are reliable lap-joints with a right-shift wear-out failure
model (m > 3) of increasing failure rate (IFR).22,28,29)
Furthermore, P2.5-90s and P1.0-135s FSSW lap-joints with
a larger Weibull modulus is recognized as the failure
occurred in a relatively concentrated data distribution range
(Fig. 7(a)) and less reliability decrease (Fig. 7(b)) while the
loading exceeds the minimum strength. Meanwhile, knowledge of the Weibull modulus can be used to determine which
material has higher reliability.30) Since the Weibull modulus
is a measure of the variability of material mechanical
properties,31) two or more materials with similar values of
mechanical properties (i.e., an average value or the extent of
data uctuation) may be dierentiated and compared through
their respective Weibull modulus. Giving the same welding
dwell time for the probe penetration depth of 2.5 mm and
1.0 mm, the P2.5-105s lap-joints not only resist a higher
extent of failure load but also show more reliable than the
P1.0-105s lap-joints because it has a larger Weibull modulus.
According to this analysis, the FSSW lap-joints with a
metallurgical bonded TMAZ region (Fig. 4(b)) show a better
joining reliability than just kissing bonded (Fig. 5(a)) for the
short probe penetration depth.
However, it should be noted that a larger value of Weibull
modulus does not necessarily imply a higher mechanical
property. The Weibull modulus is a dimensionless parameter,
which indicates the degree of the uniformity of a mechanical

property throughout the specimen, not its absolute value. For
example, the average failure load of P2.5-90s lap-joints is
larger than P1.0-135s lap-joined specimens, whereas the
P2.5-90s lap-joints show a smaller Weibull modulus. In
addition, the signicant variation in the Weibull modulus for
P2.5-90s (m 8:1) and P2.5-105s (m 4:5) lap-joined
specimens, which display a similar failure load data uctuation, can reect the dierence in the failure behaviors.
Although both of P2.5-90s and P2.5-105s lap-joints show
similar FSSW-joined microstructural feature with metallurgical bonded TMAZ region and microhardness, but the
cracks between the HAZ/TMAZ interface indicated by
triangle arrows in Fig. 4(b) and (d) for the P2.5-105s
condition are extended to approach the lap-sheets interface
compared with the P2.5-90s condition. This phenomenon
makes cracks further propagate along this defect when the
failure occurred during the tensile shear loading. Therefore,
longer welding dwell time is not available for the FSSW
process with a probe penetration depth passes the lap-sheets
interface. For the P1.0-series conditions, a large Weibull
modulus for P1.0-135s lap-joints (m 10:5) corresponds to a
metallurgical bonded microstructure with a uniform failure
load, whereas just a kissing bonded microstructure with a
broad distribution of failure load is associated with a
relatively low Weibull modulus for the P1.0-105s lap-joints
(m 3:3). The reason can be recognized as the metallurgical
bonded TMAZ region formed with plastic ow for the longer
150
(a)
T.D.
upper sheet
lower sheet
0.5 mm
(b)
(d)
upper sheet
T.D.
lower sheet
0.5 mm
(e)
20 m
20 m
(f)
(c)
tear ridge
20 m
20 m
Fig. 8 (a) Typical failure morphologies of P2.5-series FSSW lap-joints. SEM fracture surface of (b) P2.5-90s and (c) P2.5-105s lap-joined
specimens. (d) Typical failure morphologies of P1.0-series FSSW lap-joints. SEM fracture surface of (e) P1.0-105s and (f) P1.0-135s lapjoined specimens. (T.D.: tensile direction).
welding dwell time of the P1.0-135s (Fig. 5(b)) than the

P1.0-105s.
Properly welded joints should have high failure load,
display ductility and failed by tear-out of a large section of
the joined interface,13) whereas poorly welded joints with low
failure load will fail in a brittle manner at the welding region.
To distinct the fracture modes observed from the tensile shear
testing of FSSW lap-joined specimens, Fig. 8 shows the
failure morphologies and SEM fracture surfaces of these
FSSW lap-joints. Figure 8(a) represents typical failure
morphology for the P2.5-serious FSSW lap-joints, which
generally display a hole at the upper sheet and a fracture ring
observed at the TMAZ around the probe hole after tensile
shear tests. Figure 8(b) and (c) show the SEM fracture
surfaces taken from circled area indicated by the arrow in
Fig. 8(a). Both P2.5-90s and P2.5-105s lap-joints exhibit
ductile failure morphologies with dimples fracture, but a
partial fracture surface of the P2.5 105s lap-joints (Fig. 8(c))
shows a at, brittle-like failure area. This phenomenon may
be an instable cause for the decrease of the Weibull modulus.
Figure 8(d) shows the general failure morphology for the
P1.0-serious FSSW lap-joints without any probe hole at the
upper sheet. The failure morphologies of P1.0-series FSSW
lap-joints are quite dierent from the P2.5-series. Figure 8(e)
shows a brittle fracture for the P1.0-105s lap-joints, which
have relatively lower failure load than other conditions. It
should be resulted from the non-metallurgical kissing bond as
displayed in Fig. 5(a) and (c). However, the fracture surface

of P1.0-135s lap-joints as shown in Fig. 8(f) displays a
microstructural feature of the ductile failure with obvious tear
ridges. Based on the above observations, it is recognized that
the ductile fracture of FSSW lap-joints is generally occurred
with increasing the tensile shear failure load.
4.
Conclusions
The microstructural features, tensile shear failure load and

failure mechanism with Weibull statistical analysis for the
FSSW lap-joints of 5052-H34 aluminum alloy have been
characterized. The results can be summarized as follows:
(1) The TMAZ region with a metallurgical bond is created
around the probe, and the microhardness is signicantly
increased at the TMAZ region for FSSW-joined 5052
aluminum alloys. But the microhardness is sharply
decreased within the HAZ region.
(2) A higher failure load is obtained for the FSSW lapjoints with an obvious metallurgical bonded TMAZ
region. Besides, the failure load as well as the Weibull
modulus is decreased with decreasing the probe
penetration depth.
(3) The failure load and the Weibull modulus are signicantly increased with performing a longer welding
dwell time for the FSSW lap-joints welded with a short
probe penetration depth.
(4) Through the powerful statistical analysis of the Weibull

model, FSSW-joined 5052 rolled aluminum alloys with
a right-shift wear-out model (m > 3) can be recognized
as reliable lap-joints for further engineering application.
(5) A larger value of the Weibull modulus should not
necessarily imply a higher failure load or a similar data
uctuation range. It represents the degree of the
uniformity of a failure pertaining to the microstructural
features throughout the material.
Acknowledgments
This study was nancially supported by the National
Science Council of Taiwan (Contract No. NSC 96-2628-E006-008-MY3) for which we are grateful.
REFERENCES
1) S. Fukumoto, I. Lum, E. Biro, D. R. Boomer and Y. Zhou: Weld. J. 82
(2003) 307s312s.
2) P. Thornton, A. Krause and R. Davies: Weld. J. 75 (1996) 101s108s.
3) A. Gean, S. A. Westgate, J. C. Kucza and J. C. Ehrstorm: Weld. J. 78
(1999) 80s86s.
4) W. M. Thomas, E. D. Nicholas, J. C. Needham, M. G. Murch, P.
Templesmith and C. J. Dawes: GB Patent Application No. 9125978.8,
December 1991.
5) Y. S. Sato and H. Kokawa: Metall. Mater. Trans. A 32A (2001) 3023
3031.
6) S. Lathabai, M. J. Painter, G. M. D. Cantin and V. K. Tyagi: Scr. Mater.
55 (2006) 899902.
7) Y. C. Chen and K. Nakata: Scr. Mater. 58 (2008) 433436.
8) P. C. Lin, S. H. Lin, J. Pan, T. Pan, J. M. Nicholson and M. A. Garman:
SAE Technical Paper No. 2004-01-1330, (Society of Automotive
Engineering, Warrendale, PA, 2004).
9) R. Hancock: Weld. J. 83 (2004) 40.
10) K. T. Huang, T. S. Lui and L. H. Chen: Mater. Trans. 45 (2004) 3216
3222.
151
11) Y. S. Sato, H. Kokawa, K. Ikeda, M. Enomoto, S. Jogan and T.

Hashimoto: Metall. Mater. Trans. A 32A (2001) 941948.
12) Y. Hovanski, M. L. Santella and G. J. Grant: Scr. Mater. 57 (2007)
873876.
13) D. Mitlin, V. Radmilovic, T. Pan, J. Chen, Z. Feng and M. L. Santella:
Mater. Sci. Eng. A 441 (2006) 7996.
14) A. Gerlich, M. Yamamoto and T. H. North: Sci. Technol. Weld. Joining
13 (2008) 254264.
15) W. Weibull: J. Appl. Mech. 18 (1951) 293297.
16) R. Tsujino, K. Morikawa, Y. Ohue, H. Yamaguchi, K. Ogawa and
H. Ochi: J. Jpn. Inst. Light Met. 50 (2000) 281286.
17) L. Esposito, A. Tucci and D. Naldi: J. Eur. Ceram. Soc. 25 (2005)
14871498.
18) Y. Y. Zhao, E. Ma and J. Xu: Scr. Mater. 58 (2008) 496499.
19) C. W. Yang and T. S. Lui: J. Eur. Ceram. Soc. 28 (2008) 21512159.
20) S. S. Scherrer, I. L. Denry, H. W. Anselm Wiskott and U. C. Belser:
Dent. Mater. 17 (2001) 367371.
21) R. T. Kiepura and B. R. Sanders (Eds.): Metals Handbook Volume 9,
Metallography and Microstructures, ninth ed., (American Society for
Metals, Metals Park, Ohio, 1985) pp. 352354.
22) R. B. Abernethy: The New Weibull Handbook: Reliability and
Statistical Analysis for Predicting Life, Safety, Survivability, Risk,
Cost and Warranty Claims, 4th Ed. (North Palm Beach, FL, 2000).
23) B. Faucher and W. R. Tyson: J Mater. Sci. Lett. 7 (1988) 11991203.
24) S. Y. Chang, J. G. Lee, K. T. Park and D. H. Shin: Mater. Trans. 42
(2001) 10741080.
25) K. Morikawa, Y. Ohue, K. Oquwa and H. Nakayama: Proc. Int.
Oshore and Polar Engineering Conf. 4 (2000) 145149.
26) N. Biery, M. de Greaf, J. Beuth, R. Raban, A. Elliott, C. Austin and
T. M. Pollock: Metall. Mater. Trans. A 33 (2002) 31273136.
27) M. Sivapraqash, P. R. Lakshminarayanan, R. Karthikeyan, K.
Raqhukandan and M. Hanumantha: Mater. Design 29 (2008) 1549
1553.
28) R. S. Lima and B. R. Marple: J. Therm. Spray Technol. 12 (2003)
240249.
29) T. E. Wondmaqeqnehu: Nav. Res. Log. 51 (2004) 491500.
30) R. S. Lima and B. R. Marple: J. Therm. Spray Technol. 12 (2003)
360369.
31) T. H. Courtney: Mechanical Behavior of Materials, (McGraw-Hill,
New York, 1990).

Geometry Constrained Plasticity of Bulk Metallic Glass

K. Mondal* and K. Hono
Distinction between the geometry constrained plasticity and the unconstrained plasticity in compression stress-strain behaviors of bulk
metallic glasses (BMG) has been attempted. Unusually large plasticity of BMG exceeding 20% is possible due to the geometry constrained
eect in compression tests even if sample aspect ratios of 1.52 are used. Care must be taken because apparent large plasticity can arise if the
unconstrained plastic deformation by the uniform formation of ne shear bands allows the deformed parts to touch the crosshead.
[doi:10.2320/matertrans.MRA2008342]
Keywords: metallic glass, shear band, plastic deformation
1.
Introduction
Bulk metallic glasses (BMGs) possess very high yield

strength (>1:5 GPa) and high elastic strain limit (2%).
However, BMGs show little plastic deformation at room
temperature because of the lack of dislocations and highly
localized deformation along shear bands, which nally lead
to catastrophic failure under unconstrained conditions.
Therefore, the plasticity enhancement of BMGs is possible
to a certain extent if highly localized shear process along a
single shear band can be avoided by branching it to many ne
homogeneously distributed shear bands. There has been a
continuing eort to improve the plasticity of metallic glasses
based on this concept. Introduction of crystal/glass composite by dispersing micron sized dendritically grown crystals
during solidication,1) nanoscale compositional heterogeneity due to glass phase separation,24) dispersion of embedded
nanocrystals,57) high Poissons ratio8) or presence of hard
and soft zones,9) all of which can cause pinning, nucleation or
multiplication of shear bands. Very recently, it has been
shown that the total plasticity of BMG is a function of both
free volume and nanocrystal present.10)
In fact, recent studies have shown that a metallic glass with
very low unconstrained plasticity can show large plasticity
under geometry constrained conditions. Under hydrostatic
pressure, metallic glasses can exhibit large plasticity as a
result of geometry constraint of shear bands.1113) Several
investigators1417) have shown the eect of lower aspect
ratios (AR) of length/diameter of test pieces on the large
enhancement of macroscopic plasticity of BMGs. Bei et al.16)
showed that about 50% or more plastic strain has been
observed if AR is below 0.9. More recently, Wu et al.17)
reported that large deformation of brittle metallic glass under
unconstrained condition was possible at very low AR below
0.9 due to the geometrical constraint to the ow of shear
bands by crossheads, resulting in multiplication and nearly
homogeneous distribution of shear bands. In the present
work, we have chosen two dierent alloys with dierent
unconstrained plasticity to demonstrate that the apparent
plasticity of BMGs is largely inuenced by the constrained
*Present
address: Department of Materials and Metallurgical Engineering,

Indian Institute of Technology (IIT) Kanpur 208016, India
eect even with the sample aspect ratio to be larger than 1.

Based on this experiment, we have shown that the extraordinary plasticity of BMGs of more that 50% is mostly due
to the geometry constrained eect. Such compressive plastic
behavior under geometrically constrained conditions as
observed in the present case should be distinguished from
the inherent ductility of BMGs.
2.
Experimental
Zr58 Cu22 Co4 Fe4 Al12 and Zr55 Cu30 Ni5 Al12 alloy ingots
were prepared by arc melting high purity Zr (99.9%), Cu
(99.99%), Co (99.9%), Fe (99.9%), Ni (99.9%) and Al
(99.99%) under an Ar atmosphere. The former alloy shows
relatively large unconstrained plasticity, i.e., inherently
plastic BMG, while the latter was selected as a typical alloy
that rupture without plastic deformation, i.e., inherently
brittle BMG. 2 mm diameter rods of 50 mm length metallic
glass samples were prepared using copper mold casting from
the ingots in an Ar-atmosphere. The casting unit was
equipped with a radiation temperature recorder. Melt temperatures before casting were varied from 1253 to 1323 K
for the Zr58 Cu22 Co4 Fe4 Al12 alloy to provide two dierent
levels of unconstrained plasticity.18) The Zr55 Cu30 Ni5 Al12
alloy was cast from 1253 K. The macroscopic amorphous
nature of the cast alloys was conrmed using x-ray
diraction (XRD) before use. Compression tests were
performed at a quasi-static strain rate of 1 102 s1 in an
Instron machine using rod samples of 2 mm diameter at room
temperature. Both the contact surfaces of the sample with jig
are lubricated with silicon grease to minimize the friction.
Clip gauge was used during mechanical testing. The minimum length up to which the clip gauge can work was 1.6 mm.
The length of the samples has been varied in such a way to
get AR from 2 to 1.2.
3.
The XRD patterns and the DSC traces of the

Zr58 Cu22 Co4 Fe4 Al12 alloys that were cast from two dierent
temperatures18) and the as-cast Zr55 Cu30 Ni5 Al12 alloy18)
match with the previously reported results (data not shown),
conrming that all the samples used in this work were
153
Fig. 1 Nominal stress strain curves of the Zr58 Cu22 Co4 Fe4 Al12 alloy cast
from 1253 K at dierent aspect ratios at a strain rate of 1 102 s1 .
Fig. 3 Nominal stress strain curves of the Zr55 Cu30 Ni5 Al12 alloy cast from
1253 K at dierent aspect ratios at a strain rate of 1 102 s1 .
amorphous. Figure 1 shows the eect of dierent sample

aspect ratios (AR) ranging from 1.2 to 2 on compression
behaviors of the Zr58 Cu22 Co4 Fe4 Al12 sample cast from
1253 K. As AR decreases below 1.5, the curves tend to show
an apparent softening before complete failure. There is little
increase in strain as AR decreases. The strain value of the
sample with AR 2 is of 67%, which can be dened as the
unconstrained strain of the sample as reported in our previous
work.18)
Figure 2 shows the eect of dierent AR ranging from 1.2
to 2 on the mechanical properties of the Zr58 Cu22 Co4 Fe4 Al12
sample cast from 1323 K. All curves of the samples with AR
1.2 to 1.8 shows very large plastic deformation exceeding
40% strain without failure. The sample with AR 2 shows
around 15% strain before failure, which is the unconstrained
strain of the sample reported in our previous work.18) As the
machine was stopped before it reaches the critical distance
between the crosshead of the test machine (the minimum
distance up to which the clip gauge works 1:6 mm), the
compression tests more than the strain values shown in
Fig. 2(a) and (b) could not be carried out.
Figure 3 shows the eect of three dierent ARs (1.2, 1.5
and 2) on the stress-strain behaviors of the Zr55 Cu30 Ni5 Al12
sample cast from 1253 K. The sample with AR larger than

2 does not show any unconstrained plasticity as reported in
the previous report.19) As the AR decreases, the plasticity
values gradually increases. The sample with AR 1.2 did not
fail till the end of the compression test. The behavior of the
samples of all the three alloys at dierent AR is quite similar
at a strain rate of 1 104 s1 suggesting little eect of
dierent strain rates in the range of strain rate of 102
104 s1 . Figure 4 shows the eect of dierent AR ranging
from 1.1 to 2 on the mechanical properties of the
Zr58 Cu22 Co4 Fe4 Al12 sample cast from 1323 K at a strain
rate of 104 s1 for comparison purpose. There are reports
that the dierence in strain rates would change the
deformation behavior of the BMG,2023) though the eect
of strain rate within 102 104 s1 is minimal. Therefore,
only the data at higher strain rate have been presented in the
present case.
Figures 5(a), (b), and (c) show the SEM micrographs of
the Zr55 Cu30 Ni5 Al10 sample, and the Zr58 Cu22 Co4 Fe4 Al12
samples cast from 1253 K and 1323 K with AR 1.5. Three
samples show total strain of 10%, 10% and 35% with
AR 1.5, respectively. The surface of the Zr58 Cu22 Co4 Fe4 Al12
alloy cast from 1323 K (Fig. 5(c)) shows dense intersecting
154
K. Mondal and K. Hono
Fig. 6 SEM micrographs of the Zr58 Cu22 Co4 Fe4 Al12 alloy cast from
1323 K (a) the top surface at aspect ratio 1.5, (b) side view at aspect ratio
1.5 and (c) side view at aspect ratio 1.8.
Fig. 5 SEM micrographs of (a) the Zr55 Cu30 Ni5 Al12 alloy cast from
1253 K (b) the Zr58 Cu22 Co4 Fe4 Al12 alloy cast from 1253 K and (c) the
Zr58 Cu22 Co4 Fe4 Al12 alloy cast from 1323 K at the aspect ratio 1.5.
shear bands near the contact with the crosshead, whereas the
shear band density of the other two samples are less,
reecting the amount of plasticity of the samples. However,
careful observation shows that there are long shear bands
aligned preferably in one direction. In Fig. 5(c), short shear
bands are conned within two long shear bands. The maximum angles formed by the long shear band with the loading
axis of the three samples are 40 , 42 and 50 as shown on
Fig. 5(a), (b) and (c), respectively. It has been observed from
previous reports2427) that several degree deviations from the
maximum shear-stress direction are possible for BMGs. This
indicates that the fracture behavior under compression does
not follow the von Mises criterion.27) Such deviations have
been attributed to a combined eect of the normal stress and
shear stresses on the fracture plane.27)
Figure 6(a) and (b) show the SEM micrographs of the

top surface of mechanically tested sample of the
Zr58 Cu22 Co4 Fe4 Al12 alloy cast from 1323 K and side view
of the same sample with AR 1.5, respectively. The inset
of Fig. 6(a) shows the overall top view of the surface. Due to
the large deformation of more than about 35% strain (Fig. 2),
the deformation along the primary shear band allows the
deformed parts to touch both the crossheads of the machine.
Figure 6(c) shows the side view of the same alloy sample
with AR 1.8. In this case, the deformed part touches only the
top crosshead. Careful observation of the magnied part of
the section indicated by arrowhead in the inset micrograph of
Fig. 6(a) shows several white wavy patters concave inward
the center of the top surface of the sample. Step-like patterns
are present at the boundary between two deformed parts as
shown by the arrows and these patterns are formed by shear
band formation during progressive deformation of the two
sections (Fig. 6(a)). White wavy patterns are formed when
these deformed steps meet the top surface. Hence, it is clear

that the shear band interaction with the top surface always
makes a new edge that increases the overall surface area,
thereby increasing the frictional force which prevents the
material to ow outward.
The large plastic deformation of BMGs with AR below 1.0
has been explained by the termination of the shear bands at
the crossheads that leads to the multiplication of the shear
bands as shown in recent reports.1417) During compression,
besides the uniaxial stress, there is often a lateral stress
induced by the friction with the crossheads of the testing
machine.14) The uniaxial and lateral frictional stresses will
enhance a hydrostatic pressure in the specimen close to the
platen. The lower is the AR, the more uniform will be the
distribution of lateral forces over the length side. However,
the present results have shown that an extraordinary strain
level can be recorded under the geometrically constrained
conditions depending on the unconstrained plasticity even if
AR is in the range of 1.5 to 2, which is the recommended
values by the ASTM standard for compression test (Designation: E9-89a (Reapproved 2000)).
The downward trend of the stress strain curves is
pronounced below AR of 1.5 for the Zr58 Cu22 Co4 Fe4 Al12
sample cast from 1253 K (Fig. 1(a)). This sample exhibits
unconstrained plastic strain of 67% as reported previously,18) but the plastic strain observed from the sample
increases as AR decreases. In our previous work,28) we have
explained that the inection in the plastic regime of
the stress strain curve indicates the crack formation along
the primary shear band and downward movement in stress
strain curve is due to the mechanical interlocking of the
cracks and their sliding along the primary shear band. The
pronounced downward movement of the stress strain curve
is due to the pronounced frictional eect arising from the
shear bands interaction with the crosshead, which is clear
from Fig. 5(b). The pronounced frictional eect would lead
to increased lateral forces along the length side which do
not allow the formed crack to grow rapidly; hence the
sample keeps deforming till the unbroken ligament becomes
unable to withstand the stress, which is the point of complete
failure.
The Zr58 Cu22 Co4 Fe4 Al12 samples cast from 1323 K with
AR in the range of 1.8 to 1.2 show very large deformation
without complete failure (Fig. 2). As can been seen from
the sample with AR 2, the unconstrained plasticity of this
sample is about 15%. So the enhanced plasticity observed
from the samples with AR smaller than 2 are all due to the
geometrical constraint from the cross heads or the result of
the mechanical interlocking of the crack and shear bands.28)
However, the unconstrained plasticity also plays a major
role here. Because of large unconstrained plasticity (15%),
the number of shear bands interacting with the crossheads
would be more than the case of the Zr58 Cu22 Co4 Fe4 Al12
samples cast from 1253 K. This will result in higher
frictional coecient, which does not allow the formed
crack to grow rapidly. If the sample shape is observed
carefully (Fig. 6(b)), the deformed parts along the primary
shear band as shown by the arrows touch the top and bottom
crossheads. If the size of the unbroken ligaments are large
enough to withstand the stress till the deformed parts touch
155
the crosshead, stress strain curve will take an upward trend.

This is mainly because of the anvil eect that the shear
bands in the two deformed parts along the primary shear
band are totally conned and terminated by two crossheads.
In this condition, the BMG sample will not break and it will
keep deforming. In addition, if the crack is more strongly
locked by the formation of jog in its path or by crossover of
another shear band as shown elsewhere,28) the stress-strain
curves show an upward trend as shown in Fig. 2. This
behavior appears even if either of the deformed parts along
the primary shear band does not touch the crossheads as
shown by the arrows in the Fig. 6(c).
In the case of the Zr55 Cu30 Ni5 Al10 sample where the
unconstrained plasticity is nearly 0%, the enhancement of the
apparent plasticity can be seen even if AR is larger than 1 in
the range of 1.5 to 1.2. In the previous reports,1417)
geometrically constrained plasticity was reported only for
AR < 1. The AR value that causes geometrical constraint
appears to be related to the angle formed by the shear bands
with the uniaxial stress as shown in Fig. 5. Most of the shear
bands lie at an angle 40 with the uniaxial stress in the case
of the Zr55 Cu30 Ni5 Al10 sample (Fig. 5(a)), whereas in the
other two cases, most of the shear bands have an angle of 45
and 50 with the axial stress (Fig. 5(b) and (c)). Earlier
reports1417) showed that the intersecting long shear bands
morphology stretching from top to bottom surface are the
reason for large deformation if AR is less than 0.9. But even
if the deformed parts along the primary shear band touch the
two crossheads of the machine, intersecting shear bands
morphology stretching from top to bottom surface are not
present (Fig. 6(b) and (c)). This suggests that the eect of
crosshead constrain is more signicant on the long shear
bands than the short ones. Because of the lower angle
between the long shear bands and the uniaxial stress, there
would be a larger numbers of long shear bands interacting
with the crossheads in the case of the Zr55 Cu30 Ni5 Al10
sample. This denitely increases the frictional component
along the interface between crossheads and round surface.
The increase in lateral forces allows the Zr55 Cu30 Ni5 Al10
sample to deform more with decreasing AR. A similar
situation is true for the Zr58 Cu22 Co4 Fe4 Al12 samples cast
from 1253 K as the strain is also increasing with decreasing
AR. However, the angle at which shear bands are aligned will
be a strong function of the composition or the Poissons ratio
of BMGs.
From the above discussion, the large plasticity (>20%)
observed in the compression test using the samples with AR
ranging from 2 to 1 can be explained from explanation
provided by previous investigators.14,16) Once the glass starts
to deform by forming shear bands after reaching the yield
stress, the frictional force allows the lateral force to act
along the length side. In the course of deformation, the
major deformation along the primary shear band forms steps
or cracks. Increasing shear band interactions with the
crossheads with progressive deformation leads to increasing
lateral forces along the length side, thereby preventing the
crack to grow catastrophically. If the crack is locked by
forming either jog or by another intersecting shear band, the
sample will not fail and it will keep deforming by forming
secondary shear bands conned within the two major shear
156
K. Mondal and K. Hono
are, glasses with dierent unconstrained plasticity will

behave dierently under the presence of geometry constraints. Even at AR 2, there is a possibility of prolong
deformation as shown in Fig. 2, which has already been
shown in our recent work.28) Although the Zr55 Cu30 Ni5 Al10
sample shows nearly zero unconstrained plasticity, long
shear bands make an angle of 40 with the axial stress.
Hence, the critical AR is 1.3, much larger than 1. The
distribution of lateral force will be throughout the length
side at lower AR. That is why the Zr55 Cu30 Ni5 Al10 sample
at lower AR showed larger plastic strain although the sample
is inherently brittle (Fig. 3).
4.
Fig. 7 Schematic shows the eect of crosshead as the deformation along
the primary shear band (dotted line) allows the sample to touch (a) one
crosshead and (b) both the crossheads.
bands as mentioned in our previous report.28) Therefore,

mechanical interlocking and lateral force act as complementary process. The two deformed parts along the primary
shear band like Fig. 6 will eventually touch the crossheads
of the machine. In this condition, if the unbroken ligaments
between cracks along the primary shear bands are large
enough to withstand the stress, it will not fail anymore and
large deformation even more than 100% can be observed.
This condition will be met even if the AR is larger than 1,
which will be determined by the angle formed by the
primary shear band with the axial stress. For example, if the
primary shear band makes an angle with the axial stress of
about 45 , the deformed parts should touch the crossheads at
AR equal or less than 1. If it makes an angle of about 40 ,
the AR should be equal to or less than 1.3. However, this is
for an ideal case. The major deformation can be possible at
any point along the length direction. If the major deformation does not start at the half of the length, the condition
may arise that only one deformed part touches either of the
crossheads as shown in Fig. 6(c). Figure 7(a) and (b)
schematically show the eect discussed above. The deformation along the primary shear band (dotted line) can touch
the crosshead on one side (Fig. 7(a)) or both sides
(Fig. 7(b)). If mechanical interlocking intervenes, the
unbroken ligaments of the sample will not break anymore.
However, the situation as explained in Fig. 7(a) is more
likely to happen in the sample with AR > 1 and situation as
presented in Fig. 7(b) will more probable in the sample with
AR < 1 provided the long shear bands make an angle of
45 with the axis. Therefore, once the deformed parts touch
the crosshead, further deformation must be regarded as
geometrically constrained plasticity, which should be distinguished from the unconstrained deformation.
The above results on the eect of aspect ratio on the
compressive plasticity of three alloys with dierent inherent
plasticity is able to distinguish the geometry-constrained and
unconstrained plasticity of BMG. Most of the geometryconstrained strains arise after the two deformed parts along
the primary shear band touch the crossheads. Depending on
how strong mechanical locking of crack and lateral force
Conclusions
Geometry constrained shear band interaction, lateral force

and mechanical interlocking of crack can lead to very large
plasticity of BMGs even if aspect ratio is in the range of 1.5 to
2, which are commonly used value for the compression test.
The wavy pattern in the stress strain curve of BMG is
generally artifact. The present results suggest that sucient
care must be taken to get rid of the constrained plasticity to
understand the inherent ductility of the BMG.
Acknowledgement
This work was supported by the grant-in-aid of scientic
research, Priority Area Science of Metallic Glass of
MEXT, Japan. The assistance of Prof. Y. Shibutani of the
Department of Mechanical Engineering of Osaka University
of Japan in connection with the preparation of this manuscript
is also much acknowledged.
REFERENCES
1) C. C. Hays, C. P. Kim and W. L. Johnson: Phys. Rev. Lett. 84 (2000)
29012904.
2) K. F. Yao, F. Ruan, Y. Q. Yang and N. Chen: Appl. Phys. Lett. 88
(2006) 12210611221063.
3) J. C. Oh, T. Ohkubo, Y. C. Kim, E. Fleury and K. Hono: Scr. Mater. 53
(2005) 165169.
4) E. S. Park and D. H. Kim: Acta Mater. 54 (2006) 25972604.
5) H. Kato, K. Yubuta, D. V. Louzguine, A. Inoue and H. S. Kim: Scr.
Mater. 51 (2004) 577581.
6) Y. C. Kim, D. H. Kim and J. C. Lee: Mater. Trans. 44 (2003) 2224
2227.
7) J. C. Lee, Y. C. Kim, J. P. Ahn, H. S. Kim, S. H. Lee and B. J. Lee: Acta
Mater. 52 (2004) 15251533.
8) J. Schroers and W. L. Johnson: Phys. Rev. Lett. 93 (2004) 2555061
2555064.
9) Y. H. Liu, G. Wang, R. J. Wang, D. Q. Zhao, M. X. Pan and W. H.
Wang: Science 315 (2007) 13851388.
10) K. Mondal, T. Ohkubo, T. Toyama, Y. Nagai, M. Hasegawa and K.
Hono: Acta Mater. 56 (2008) 53295339.
11) L. A. Davis and S. Kavesh: J. Mater. Sci. 10 (1975) 453459.
12) J. J. Lewandowski and P. Lowhaphandu: Phil. Mag. A 82 (2002) 3427
3441.
13) J. Lu and G. Ravichandran: J. Mater. Res. 18 (2003) 20392049.
14) Z. F. Zhang, H. Zhang, X. F. Pan, J. Das and J. Eckert: Phil. Mag. Lett.
85 (2005) 513521.
15) W. H. Jiang, G. J. Fan, H. Choo and P. K. Liaw: Mater. Lett. 60 (2006)
35373540.
16) H. Bei, S. Xie and E. P. George: Phys. Rev. Lett. 96 (2006) 1055031
1055034.

17) F. F. Wu, Z. F. Zhang and S. X. Mao: J. Mater. Res. 22 (2007) 501507.
18) K. Mondal, T. Ohkubo, T. Mukai and K. Hono: Mater. Trans. 48 (2007)
13221326.
19) H. Kato, T. Hirano, A. Matsuo, Y. Kawamura and A. Inoue: Scr. Mater.
43 (2000) 503507.
20) T. Mukai, T. G. Nieh, Y. Kawamura, A. Inoue and K. Higashi:
Intermetallics 10 (2002) 10711077.
21) A. V. Sergueeva, N. A. Mara, D. J. Branagan and A. K. Mukherjee: Scr.
Mater. 50 (2004) 13031307.
22) F. H. D. Torre, A. Dubach, M. E. Siegrist and J. F. Loer: Appl. Phys.
Lett. 89 (2006) 09191810919183.
157
23) J. Zhang, J. M. Park, D. H. Kim and H. S. Kim: Mater. Sci. Eng. A 449
451 (2007) 290294.
24) P. E. Donovan: Mater. Sci. Eng. 98 (1988) 487490.
25) P. Lowhaphandu, S. L. Montgomery and J. J. Lewandowski: Scr.
Mater. 41 (1999) 1924.
26) W. J. Wright, R. Saha and W. D. Nix: Mater. Trans. 42 (2001) 642649.
27) Z. F. Zhang, J. Eckert and L. Schultz: Acta Mater. 51 (2003) 1167
1179.
28) K. Mondal, G. Kumar, T. Ohkubo, K. Oishi, T. Mukai and K. Hono:
Phil. Mag. Lett. 87 (2007) 625635.

Residual Stress Relaxation in Shot Peened Surface Layer

on TiB2 /Al Composite under Applied Loading
Weizhi Luan1 , Chuanhai Jiang1; * and Vincent Ji2
1
School of Materials Science and Engineering, Shanghai Jiao Tong University,

800 Dongchuan Road, Shanghai 200240, P.R. China
2
LEMHE/ICMMO, UMR 8182, Universite Paris-Sud 11, 91405 Orsay, France
The residual stress relaxations in the shot peened surface layer on the TiB2 /6351Al composite under static and cyclic loading were
investigated. The results show that the compressive residual stresses were relaxed under applied tensile stresses. The relaxation of the residual
stresses in the longitudinal direction is more important than the one in the transverse direction. Under cyclic loading, the fast relaxation of
residual stresses occurred in the rst few circles. The main reason of the residual stresses relaxations is the inhomogeneous local plastic
deformation caused by the large number of the reinforcement particles in the matrix during applied loadings.
Keywords: residual stress relaxation, applied loading, shot peening, metal matrix composite
1.
Introduction
Metal matrix composites, MMCs, are found to exhibit

excellent physical and mechanical properties such as specic
high modulus, strength as well as thermostability etc., and
have been widely concerned.1) But the large dierence in the
coecient of thermal expansion between the matrix and the
reinforcements gives rise to residual stresses (r ) during
manufacturing and subsequent heat treatment. Since these
stresses are tensile in the matrix, they may deteriorate the
fatigue properties.2) It is therefore of interest to apply a
mechanical surface treatment to enhance the fatigue properties of the MMCs. Shot peening (SP), an eective method
widely used in industry, can considerably improve fatigue
strength and fatigue life of cyclically loaded metallic
components.3) These improvements are achieved by inducing
a compressive residual stress (CRS) eld in the surface
deformation layers. Unfortunately, these residual stresses
could relax signicantly due to subsequent mechanical
loading which could be detrimental in fatigue properties.4)
It was believed that the improvement in the fatigue properties
depends mainly on the stability of the residual stresses on the
tension-loaded state.4,5) And the major residual stress
relaxation takes place during the early cyclic stage of cyclic
loading.6) The short report deals with the relaxation behaviors
of the residual stresses, in the shot peened surface layer on an
in-situ TiB2 /6351Al composite, under applied tensile loadings. The results are also discussed.
2.
Experimental
The TiB2 /6351Al composite (10 vol% TiB2 ) used in this

study was synthesized in-situ according to Refs. 1, 7) with a
nominal chemical composition of 1.0 Si, 0.6 Mg, 0.6 Mn, 6.7
B, 3.0 Ti, and balance Al (in mass%). During the synthesis,
mixed salts of potassium hexauorotitanate (K2 TiF6 ) and
potassium tetrauororate (KBF4 ) are introduced into a stirred
*Corresponding
author, E-mail: wzluan@sjtu.edu.cn
flake
rod-like
1 m
Fig. 1
SEM image of the reinforcements (TiB2 ) distributed in the matrix.
aluminum melt with an atomic ratio in accordance with Ti/

2B. Exothermal reactions between the salts and the molten
aluminum take place according the following sequences,
3K2 TiF6 13Al ! 3TiAl3 3KAlF4 K3 AlF6
2KBF4 3Al ! AlB2 2KAlF4
AlB2 TiAl3 ! TiB2 4Al
The size of the reinforcements is between 50 nm and 500 nm
with rod-like or ake shapes as shown in Fig. 1. The
synthesized composites were hot extruded to a rod with a
diameter of 35 mm at a temperature of 450 C using extrusion
ratio of 10 : 1. Before SP, heat treatments were conducted,
solution treatment at 530 C for 110 min, then quenched into
the water, and nally aging at 170 C for 6 h. This was
regarded as the initial state of the composite. Dogbone
samples were cut from the center of the rod alone the
extrusion direction (longitudinal direction) with the crosssection dimensions of 5 2 mm2 .
The SP treatments were carried out according to the
conditions: 0.3 MPa jet pressure, 1 min time, 0.25 mm
average diameter of Al2 O3 ceramic beads, 100 mm distance
Residual Stress Relaxation in Shot Peened Surface Layer on TiB2 /Al Composite under Applied Loading
159
100
Residual Stresses, MPa
Applied stress, MPa
50
a 150 MPa
5 min
0
-50
-100
Transverse
Longitudinal
-150
-200
0
Time, t / min
100
200
300
400
500
Depth, D / m
Fig. 2 Applied stress vs time t during cyclic loading.

Fig. 4 Depth distributions of the longitudinal and transverse residual
stresses in the deformation layer, D is the depth from the top surface.
400
-20
E = 80 GPa
e= 210 MPa
0.2= 300 MPa
200
100
0
0.0
Residual stresses, MPa
Stress, MPa
300
0.5
1.0
1.5
2.0
2.5
3.0
Stain in %
Fig. 3
Monotonic tensile stress-strain curve for the initial state composite.
between nozzle and specimen, and 0.24 mm A Almen

peening intensity. For tensile test, a SANS EMT1503
micro-tensile tester was built on the specimen that was
mounted under a ST X350A X-ray stress analyzer. Static
tensile loadings of up to 350 MPa and cyclic tensile loading
of a 150 MPa were applied along the longitudinal
direction. The relationship between applied stress and time
during cyclic loading is shown in Fig. 2. In the case of static
loading, the magnitude of applied stress increasing step by
step. At each step, the residual stresses in the longitudinal
direction and transverse direction were determined by the Xray stress analyzer after unloading. Cr-K radiation was
used, and the shifts of Al(311) diraction proles were
detected.
3.
The mechanical properties of the initial state composite

were estimated from the monotonic tensile stress-strain
curve, as shown in Fig. 3. The Youngs modulus E and proof
stress 0:2 of the composite are 80 GPa and 300 MPa,
respectively. The corresponding value of E and 0:2 of the
unreinforced 6351 alloy is 69 GPa and 235 MPa. The elastic
limit e of the composite was also estimated to be 210 MPa.
The increments of the mechanical properties of the composite
are attributed to the TiB2 ceramic particles exist in the
matrix.
Longitudinal
Transverse
-40
-60
-80
-100
-120
-140
-160
50
100
150
200
250
300
350
Applied Stress, a / MPa

Fig. 5 Relationships between residual stresses r after unloading and the
applied stresses under static loadings.
The depth distributions of the residual stresses in the

deformation layer along the longitudinal and transverse
directions are presented in Fig. 4. The results reveal that the
values of CRS in the both directions increase with the depth
to the peak value, then decrease and nally become identical
at the depth about 400 mm. At the same depth, the value of
CRS in the longitudinal direction is larger than the one in the
transverse direction, which is attributed to the extrusion
eect during composite preparation.
Relaxations of the residual stresses were observed under
external tensile loadings. Figure 5 shows the relaxation
behaviors of the longitudinal and transverse residual stresses
under static loadings. It can be seen that the CRS relaxation
took place even for low applied tensile stress. The value of
residual stresses decreased with the increasing applied tensile
stress. And more signicant relaxation of the longitudinal
stresses was observed. The relaxation behaviors of the
residual stresses under cyclic loading of 150 MPa were
depicted in Fig. 6. It can be seen that the residual stresses
relaxed with the number of the loading cycles N. Fast
relaxation of the residual stresses can be observed in the rst
few cycles, as indicated by the arrows in Fig. 6. The cycle of
the fast relaxation stage is 10 and 2, respectively, in the
160
W. Luan, C. Jiang and V. Ji
Residual Stresses, MPa
-100
-110
-120
-130
-140
-150
Longitudinal
Transverse
0
10
15
20
Number of Loading Cycles, N

Fig. 6 Relationships between residual stresses r after unloading and the
number of cycles N under cyclic loading, a 150 MPa.
directions but mainly in the loading direction (longitudinal

direction) which leads to the easy relaxation of residual
stresses. It was in the rst few circles that the fast reduction of
CRS occurred, dierent from the results of Hanus5) and
Wyman6) that only the rst cycle played a predominant role
in the stress relaxation. It is reasonable to consider that this
phenomenon was caused by the nature of the reinforcement
size. The size of the ceramic particles in the cases Hanus and
Wyman were much larger (9.3 mm). In the present study, the
size of reinforcement particles is less than 500 nm. Thus there
are more reinforcement particles in the matrix though their
volume fraction is relatively lower. The more reinforcement
particles distributed in the matrix could cause more inhomogeneous local plastic deformation under applied static and
cyclic tensile stresses, which resulted in the more relaxation
of the CRS.
4.
longitudinal and transverse directions. In the transverse

direction, the relaxation is less important for the shot peened
composite. This result is in accordance with the result of
Hanus.5) The tensile cold working increases the tensile yield
strength. The greater the tensile cold working, the higher the
tensile yield strength. This phenomenon is known as the
Bauschinger eect. Due to the improved tensile yield
strength, the deformation decreased in further cycles. Then
after the rst few cycles, the residual stresses became stable
gradually.
The residual stress relaxation varied linearly with the
logarithm of the fatigue cycles and can be described by the
equation proposed by Kodama8)
NRS A m log N
where NRS is the surface residual stresses after N cycles. A

and m are denoted as the material constants depending on the
applied stress. Using the eq. (1) and the datum on the right
hand of the arrow in Fig. 6, the relaxation behaviors of the
residual stresses were described. For the longitudinal residual
stress relaxation, NRS L 126 20 log N, and for the
transverse stress relaxation, NRS T 114 6:6 log N. The
subscript L and T stand for the longitudinal and transverse
residual stresses. The results show that the relaxation of the
longitudinal stresses is faster than that of the transverse
stresses under the same applied cyclic stress.
For a certain peened sample, the relaxation of the residual
stresses due to cyclic loading is aected mainly by the
applied stress amplitude and the number of cycles. In
principle, residual stress relaxation due to mechanical
loading occurs when the superposition of applied stress and
residual stress reaches the yield point of the material. For
composites, due to the existence of reinforcements, inhomogeneous local plastic deformation in the matrix could be
anticipated under the applied stress.9) In the present study,
the applied cyclic loading is 150 MPa, lower than the elastic
limit e of the composite. In this stress level, only local
plastic ow occurred due to the presence of local stress peaks
close to the reinforcement particles which result in the
reduction of the CRS. The local plastic ow occurred in both
Conclusions
The residual stress relaxations in the shot peened surface

layer on the TiB2 /Al composite under cyclic and static
loading were investigated. The results show that the value of
the longitudinal residual stresses is lager than the one of
transverse residual stresses in the shot peened layer. The
compressive residual stresses were relaxed under applied
tensile stresses. The relaxation of residual stresses in the
longitudinal direction is more important than the one in the
transverse direction. The fast relaxation of residual stresses
occurred in the rst few circles which are 10 and 2 cycles,
respectively, for the longitudinal and transverse residual
stress relaxation. The inhomogeneous local plastic deformation caused by the large number of the reinforcement
particles in the matrix during applied stress is the main
reason for the compressive residual stress relaxation.
Acknowledgements
This research was supported by Magnolia Science &
Technology Talent Fund under the project no. 2007B071.
The authors are indebted to Prof. Haowei Wang for the
composite preparation.
REFERENCES
1) S. C. Tjong and Z. Y. Ma: Mater. Sci. Eng. R 29 (2000) 49113.
2) M. E. Fitzpatrick, P. J. Withers, A. Baczmanski, M. T. Hutchings, R.
Levy, M. Ceretti and A. Lodini: Acta Mater. 50 (2002) 10311040.
3) V. Schulze: Proc. 8th Int. Conf. on Shot Peening, ICSP-8 in GarmischPartenkirchen, DGM, Wiley-VCH, Weinheim, Germany, (2003)
pp. 145160.
4) S. A. Meguid, G. Shagal, J. C. Stranart, K. M. Liew and L. S. Ong:
J. Eng. Mater. Technol. 127 (2005) 170178.
5) E. Hanus and T. Ericsson: Mater. Sci. Eng. A 194 (1995) 147156.
6) W. Z. Zhuang and G. R. Halford: Int. J. Fatigue 23 (2001) 3137.
7) H. Z. Yi, N. H. Ma, X. F. Li, Y. J. Zhang and H. W. Wang: Mater. Sci.
Eng. A 419 (2006) 1217.
8) S. Kodama: Proc. Int. Conf. on Mechanical Behavior of Metals II.
Kyoto: Society of Material Science, 2 (1972) pp. 111118.
9) T. S. Srivatsan, M. Al-Hajri, M. Petraroli, B. Hotton and P. C. Lam:
Mater. Sci. Eng. A 325 (2002) 202214.

#2009 The Japan Society for Technology of Plasticity
Finite Element Simulation with Coupled Thermo-Mechanical Analysis

of Superplastic Dieless Tube Drawing Considering Strain Rate Sensitivity*
Tsuyoshi Furushima and Ken-ichi Manabe
Department of Mechanical Engineering, Tokyo Metropolitan University, Tokyo 192-0397, Japan
The deformation behavior in superplastic dieless tube drawing is studied numerically by the nite element method (FEM). The FEM with
coupled thermo-mechanical analysis is conducted considering strain rate sensitivity to clarify the eect of dieless tube drawing conditions such
as tensile speed and material properties on the deformation behavior of the tube. In the calculation, the material properties were dealt with as a
function of temperature in a special subroutine, whose constitutive equation considering strain rate sensitivity and strain hardening was used, and
was linked to the solver. FEM results for both heat transfer and the deformation prole are in good agreement with experimental results.
Therefore, the validity of FE modeling of superplastic dieless drawing is demonstrated. In addition, the eect of material properties m and n
values on the deformation prole is demonstrated numerically. As a result, a higher m value constrains the local instability deformation. In cases
where the m value is high, the n value has little eect on the deformation prole. Consequently, it is found that analysis considering strain rate
sensitivity is important for the prediction of the deformation prole in the dieless drawing process. [doi:10.2320/matertrans.P-MRA2008837]
(Received January 25, 2008; Accepted September 29, 2008; Published December 3, 2008)
Keywords: dieless tube drawing, nite element method, strain rate sensitivity, heat transfer
1.
Introduction
Recently, the manufacture of ultrane and high-precision

processed components has led to the enhancement of the
functions of medical and biological equipment, and equipment used for communication and measurement, and has
contributed to developments in the elds related to microelectro-mechanical systems (MEMS). In particular, ultrane
microtubes are expected to be applicable in various devices,
such as mechanical/structural components in cooling micronozzles, electrode tubes used in electric discharge machining,
contact probes, and painless needles.1)
We proposed superplastic dieless drawing for fabricating
microtubes and succeeded in fabricating a microtube with an
outer diameter of D 343 mm and an inner diameter of
d 161 mm without using ne dies, which otherwise cause
various problems at micro-scale, while maintaining the ratio
of the inner to outer diameters. Thus, the eectiveness of
superplastic dieless drawing was demonstrated.2,3) Moreover,
the eects of the drawing conditions on the deformation
proles were examined.
Investigating the eects of various factors independently,
such as processing conditions and material properties, by
means of experiments is, however, limited. In addition, it is
very dicult to experimentally observe in detail the
deformation behavior of tubes during dieless drawing at
micro-scale. Therefore, no detailed experimental study was
presented in the previous report,2,3) indicating the necessity of
a numerical investigation.
In a numerical study on dieless drawing, Kobatake et al.
calculated temperature distribution by the nite element
method (FEM) considering heat transfer obtained by the
deformation proles during dieless drawing experimentally.4)
The temperature distribution in the axial direction obtained
by the analysis showed good agreement with experimental
results, indicating the eectiveness of the analysis. However,
*This Paper was Originally Published in Japanese in J. JSTP 48 (2007) 51
55.
only heat conduction was considered and deformation

behavior was not included in the analysis; therefore, the
eects of each factor on the deformation behavior were not
investigated. Wang et al. calculated the dieless drawing force
in a steady state on the basis of the energy principle and
studied the eects of various factors on the drawing force.5)
However, the eect of heat transfer and the deformation
behavior in an unsteady state were not considered, resulting
in an insucient analysis. Moreover, we previously carried
out the FE simulation coupled with the thermo-mechanical
analysis of dieless drawing, taking heat conduction inside a
tube into consideration.6) However, in the conventional
model, the temperature distribution on the tube surface had to
be determined beforehand by presetting the heating and
cooling temperatures and the distance between the heater and
the cooler. In fact, the temperature distribution on the tube
surface was unknown because the temperature distribution
should be determined by the amount of heat supplied from
the heating coil and the heat transfer between the tube surface
cooling air and atmosphere. Therefore, the conventional
model was ineective for studying the eects of the
conditions, such as the amount of heat supplied from the
heating coil and the eciency of the cooling coil, on the
temperature distribution.
In this study, we made a new heat transfer model in which
we consider the amount of heat supplied from the heating coil
and the heat transfer between the tube and the cooling coil
and that between the tube and the atmosphere. A material
with strain rate sensitivity and strain hardening dependence,
as well as their temperature dependences, were considered
was modeled as a tube. The FE simulation coupled with the
thermo-mechanical analysis of superplastic dieless drawing,
in which the evaluation of the heat transfer on a tube surface
and the heat conduction inside the tube was coupled with the
deformation analysis, was carried out. More concretely, we
analyzed the dieless drawing of the Al-78Zn superplastic
alloy, which was also used in the previous study, compared
the results with the previous experimental results,2,3) and
demonstrated the eectiveness of the present analysis. The
162
T. Furushima and K. Manabe
Cooling coil
A2
V1
Superplastic tube
V1
50mm
FE model of superplastic dieless drawing process.
Schematic illustration of superplastic dieless drawing process.
eects of each factor on the drawing behavior were then

investigated. In particular, for the eects of drawing
conditions and material properties on the deformation prole,
the aspects that cannot be elucidated in detail by experiments
were numerically studied.
2.
Tube
Fig. 2
Heat transfer Heat flux

q
coefficient h
Fig. 1
qs
0
hz
ha
0
r 1 A2 =A1 V1 =V1 V2 :
It is possible to obtain a large r using a superplastic material

with a high ductility for the original tube.
FE Analysis of Superplastic Dieless Drawing
3.1 FE model for superplastic dieless drawing

MSC.Marc/Mentat ver. 2001 was used as the commercial
FEM code to analyze superplastic dieless drawing. The
coupled thermo-mechanical rigid-plastic FE simulation was
carried out. As shown in Fig. 2, an axisymmetric model was
used in the FE model. The dimensions of the tube were used
in the previous experiment:2,3) outer diameter D 2 mm
and inner diameter d 1 mm. The length of the tube was
l 50 mm. The number of elements was 500 in the axial
direction and 4 in the tube thickness direction, and a 4-node
axisymmetric element was used. The left end of the tube was
xed and tensile deformation was applied to the right end of
the tube at tensile speed V1 .
3.2 Heat transfer model
The temperature distribution of the tube during dieless
drawing was changed in accordance with the heat supplied
from the heating coil, the heat conduction inside the tube, and
the heat transfer from the tube to the cooling air from the
cooling coil and that from the tube to the atmosphere. The
heat conduction equation of the axisymmetric model in an
unsteady state considering the above factors is expressed as

1 @
@T
@2 T
@T
r
2 c
Q_ 0:
2

r @r
@r
@z
@t
Here, is the thermal conductivity, c is the specic heat, is
ce
cz
Superplastic Dieless Drawing
Superplastic dieless drawing is a method of fabricating a

thin tube by the following procedure: x one end of a
superplastic alloy tubular material, as shown in Fig. 1, heat
part of the material and then draw the other end of the
material at speed V1 . Simultaneously move the heated
deformation region at speed V2 . The cross-sectional areas
of the tube before and after deformation are A1 and A2 ,
respectively. The reduction in area r is obtained as7)
hw
hc
Tube
3.
Cooling coil
Heating coil
V2
D= 2mm
d = 1mm
A1
t = 0.5mm
V2
Heating coil
Heating coil
V2
Heat flux
cw
z
Cooling coil
Cooling air
Fig. 3 Heat transfer model on tube surface in FE analysis of superplastic

dieless drawing process.
the mass density, and Q_ is the heat generated per unit time
and unit volume.
The tube is heated by the heat supplied from the heating
coil and cooled by the cooling air from the cooling coil and
the atmosphere. In this analysis, the boundary conditions for
the heat transfer at the tube boundaries are as shown in Fig. 3.
The heat is transferred through these boundaries, where the
following boundary conditions are satised.
3.2.1 Heat transfer inside heating coil
Heat is transferred on the tube surface by the heating coil,
and the following boundary condition is satised on the tube
surface.
@T
q
@n
The heat ux q outside the heating coil (z < hz and

hz hw z) is dened as
q0
and that inside the heating coil (hz z < hz hw ) is dened

as
q qs :
Here, hz is the position coordinate of the heating coil on the z

axis, hw is the width of the heating coil, and qs is the amount
of heat ux supplied.
Because the heating coil is moved at speed V2 in time t, hz
is expressed as
hz hz0 V2 t:
Here, hz0 is the initial position coordinate of the heating coil

on the z axis.
3.2.2 Heat transfer between tube and cooling coil and
that between tube and atmosphere
Heat transfer occurs between the tube surface and the
Thermo-mechanical FE Simulation of Dieless Drawing with Strain Rate Sensitivity
@T
hT Tc
@n
Here, h is the heat transfer coecient and Tc is the

atmospheric temperature. In the analysis, the air owing
from the inside to the outside of the cooling coil was
modeled assuming that h gradually changes. Namely, h
outside the cooling coil (z < cz ce and cz cw ce z)
is dened as
h ha :
500
0.4
400
0.3
300
0.2
200
0.1
K value
11 -3.6701
K=6 10 T
100
0
273
h in the region outside the cooling coil where the eect of

cooling air is observed (cz ce z < cz and cz cw z <
cz cw ce ) is dened as
hc ha
fz cz ce g ha
ce
cz ce 5 z < cz ;
hc ha
h
fcz cw zg hc
ce
cz cw 5 z < cz cw ce ;
m=2 10 T -0.0003 T +0.0266

m value
0.5
m value
600
0.6
Fig. 4
K value
cooling air from the cooling coil and between the tube surface
and the atmosphere, and the following boundary condition is
satised.
163
323
373 423 473 523

Temperature T /K
573
Material properties of Al-78Zn alloy used in FEM analysis.
10
and h under the cooling coil (cz z < cz cw ) is dened as

h hc :
11
Here, cz is the position coordinate of the cooling coil on the z

axis, ce is the distance which is aected by the cooling air, cw
is the width of the cooling coil, ha is the heat transfer
coecient between the tube surface and the atmosphere, and
hc is the heat transfer coecient between the tube surface
and the cooling air from the cooling coil.
Because the cooling coil is moved at speed V2 in time t,
similarly to the heating coil, cz is expressed as
cz cz0 V2 t:
12
Here, cz0 is the initial position coordinate of the cooling coil

on the z axis.
The thermal properties of the material are a specic heat c
of 383 J kg1 K1 , a mass density of 7,130 kgm3 , and a
thermal conductivity of 113 Wm1 K1 . The heat transfer
coecient between the material and the atmosphere ha is
30 W m2 K1 , and the atmospheric temperature and the
temperature of the cooling air are both 293 K.
In this analysis, the amount of heat (qs ) and the heat
transfer coecient (ha , hc ) are assumed to be constant values.
In the case of the analysis with higher precision, it is better to
set the amount of heat and the heat transfer coecient on the
basis of the actual heating and cooling sources. For example,
when modeling the high-frequency-induced heating, a more
precise analysis can be realized by adding the coupled
electromagnetic-thermal analysis and calculating the amount
of heat generated to this model. In this case, however,
it is necessary to prepare a numerical model in which all
of electromagnetic-thermo-mechanical factors are coupled,
which is very complicated. In this report, only the coupled
thermo-mechanical analysis was taken into consideration for
simplicity.
3.3 Material model

In dieless drawing, a tube is fabricated by locally
increasing the temperature. In general, the ow stress of a
material is greatly aected by the strain rate at a high
temperature. Particularly in the case of the superplastic
material used in this study, the strain rate sensitivity of the
ow stress is very high. Now, we apply the following ow
stress function considering the strain hardening dependence
and the strain rate sensitivity into consideration to the
material model dealt with in this analysis.
K"n "_m
13
Here, K is the strength coecient, n is the drawing hardening

index, and m is the strain rate sensitivity index. Moreover, K,
m, and n are assumed to approximately be functions of
temperature.
A1-78Zn superplastic alloy, which was used in the
experiment, was used as the material in this analysis.
Because this material rarely hardens even at room temperature, n 0 was assumed in every temperature range. The
other material properties K and m were obtained by the
jump tensile test in the temperature region between room
temperature and 523 K in the range of strain rate between
4:17 104 and 102 s1 , and were dened to be functions
of temperature on the basis of the test result (Fig. 4).
3.4 User subroutine
The boundary condition of heat transfer and the material
model dened in the previous section were incorporated into
MSC.Marc employing user subroutines to carry out the FE
simulation coupled with the thermo-mechanical analysis,
considering the strain rate sensitivity, as shown in Fig. 5.
4.
Analytical Results and Discussion
4.1 Validity of heat transfer analysis

To demonstrate the validity of the heat transfer model, the
temperature distribution on the tube was experimentally
measured by moving the heating and cooling coils at
V2 3:6 mm/min without applying tensile deformation.
The apparatus and method of dieless drawing were the same
164
= K
n
K(T ), n(T ), m(T )
Main Solver
MARC
Mechanical
analys is
Heat flux
Thermal
analys is
q(z, t)
Heat transfer coefficient
h(z, t)
593
Power :60%
q s =1500kW m
-2
-2
q s =1270kW m
493
-2
q s =1000kW m
Power :45%
393
Material properties
Initial Data
Exp. (0.2MPa)
-2 -1
FEM (h c =2000W m K )
V2=3.6mm/min
Special subroutine
Temperature at thermocouple T /K
693
Start
Thermocouple
-2
q s =400kW m
293
Increment
(t)
-5
Fig. 7
Output Data
0
5
10
Coil position from thermocouple l /mm
Eect of heat ux under heating coil on temperature history.
End
(a)
Heating coil
Fig. 5 Flow chart of coupled thermo-mechanical analysis and subroutine

program.
(b)
573
-2
-1
-2
-1
h c = 0Wm K
No cooling
473
-2
h c = 500W m K
-1
h c = 1000W m K
-2
-1
h c = 2000W m K
373
Exp. (Power 45%)

-2
FEM (q s =400kW m )
V2=3.6mm/min
Temperature at thermocouple T / K
673
(c)
Thermocouple
Instable deformation
Cooling (0.2MPa)
273
-5
0
5
10
Coil position from thermocouple l /mm
Fig. 6 Eect of heat transfer coecient under cooling coil on temperature

history.
(d)
5mm
as those used in the previous study.2,3) The experiments were

carried out under condition of the high-frequency powers of
45 and 60% without cooling and with the cooling air pressure
of 0.2 MPa.
Figure 6 shows the eect of the heat transfer coecient
under the cooling coil on the temperature distribution of the
tube. FE analysis conditions were as follows: amount of heat
supplied qs 400 kWm2 and heat transfer coecient under
the cooling coil hc 0{2;000 Wm2 K1 . With increasing hc ,
the temperature decreases in general. Also in the experiment,
the temperature in the case of applying 0.2 MPa cooling air
generally decreases more than that in the case without
cooling, indicating that the FE analysis well simulates the
experiment. Figure 7 shows the eect of the amount of
supplied heat on the temperature distribution of the tube. FE
analysis conditions were as follows: qs 400{1;500 kWm2
and hc 2;000 Wm2 k1 . With increasing qs , the temperature increases in general, which is in good agreement with
the experimental results. By this comparison of the analysis
and experimental results, the validity of the heat transfer
model used in this analysis was veried.
Fig. 8 Eect of tensile speed, V1 on deformation prole in superplastic

dieless drawing process (a) V1 3:6 mm/min, V2 3:6 mm/min,
r 50% (b) V1 7:2 mm/min, V2 3:6 mm/min, r 66:7% (c)
V1 10:8 mm/min, V2 3:6 mm/min, r 75% (d) V1 14:4 mm/
min, V2 3:6 mm/min, r 80%.
4.2 Eects of tensile speed on deformation prole

The deformation prole was observed after superplastic
dieless drawing while maintaining the moving speed of the
heating coil V2 at 3.6 mm/min and changing the tensile speed
V1 in the range of 3.614.4 mm/min. The analysis was
carried out by setting qs 1;100 kWm2 such that the
temperature of the heated part of the tube was approximately
523 K. Figure 8 shows the deformation proles under
conditions of the experiment and analysis. With increasing
V1 , the reduction in area r increases and the outer diameter
of the tube D decreases. Moreover, unstable deformation in
the initial stage of dieless drawing can be observed also in
the FE analysis results. The degree of unstable deformation
is dened as the unstable deformation rate D=Dave .
35
165
(b)
(a)
FEM
Exp.
30
0.8
0.5
523K
0.7
0.4
0.6
D ave
20
D
Dave
15
0.5
D min
Dave Dmin
Dave
0.4
0.3
0
293
Exp.
0
0
20
40
60
80
100
0.2
0.1
0.1
10
0.3
0.2
0.2
FEM
n value
25
m value
Instability deformation ratio D/Dave /%
Thermo-mechanical FE Simulation of Dieless Drawing with Strain Rate Sensitivity
393
493
Temperature T /K
523K
0
593
293
393
493
Temperature T /K
593
Fig. 10 Temperature dependency of m and n values used in FEM analysis

(a) m value (b) n value.
Reduction in area r /%
Fig. 9 Relationship between ratio of instability deformation, D=Dave and
reduction in area, r.
(a)
Heating coil
Localized thinning
(b)
Here, Dave is the average outer diameter in the steady state

and Dmin is the minimum outer diameter of the unstably
deformed part.
Figure 9 shows the relationship between D=Dave and r.
D=Dave increases with increasing r. The higher the
D=Dave , the higher the risk of fracture in an unsteady state
in the initial stage of drawing. In the experiment, drawing
with r of more than 80% failed and fracture occurred. Here,
we judged that D=Dave can be used as an index for
evaluating the risk of fracture. The analysis result shows
good agreement with the experimental result, indicating that
the behavior during dieless drawing in the unsteady state until
reaching the steady state is well simulated.
4.3
14
Eects of material characteristics on deformation

prole
In the previous study,2,3) the eect of the temperature of
the heated part on the deformation prole was examined, and
it was found that the unstable deformation can be suppressed
at an appropriate temperature. We considered this to be due
to the eect of n value being temperature dependent.
However, some other parameters also depend on the temperature, and it is impossible to individually determine each
eect of various material characteristics, including m value,
experimentally. In the present analysis, we focused on the
individual eect of m and n values on the deformation prole.
Figure 10 shows the temperature dependences of m and n
values. The relationship between the material characteristics
and the temperature from room temperature to 523 K was
approximated to be linear. The representative m and n values
were the values at 523 K and room temperature, respectively.
K value was set to be the same as that of the Al-78Zn alloy
shown in Fig. 4. The analysis was carried out under the
conditions of V1 3:6 mm/min, V2 3:6 mm/min, and
r 50%.
(c)
Fig. 11 Eect of m value on deformation prole in dieless drawing process

(a) m 0:1, n 0:2 (b) m 0:2, n 0:2 (c) m 0:5, n 0:2.
D/Dave /%
20-25
Instable deformation ratio

D/Dave /%
D
Dave Dmin
Dave
Dave
25
15-20
10-15
20
5-10
15
0-5
10
5
0.4
0
0.2
0.3
0.3
m value
0.2 n value
0.5
0
0.7
Fig. 12 Eect of m and n values on instable deformation ratio D=Dave in

dieless drawing process.
Figure 11 shows the eect of m value on the deformation

prole. When m value is 0.1 or less, unstable deformation
locally occurs, resulting in necking. The unstable deformation is suppressed with increasing m value.
Figure 12 shows the eects of m and n values on D=Dave .
As described above, the magnitude of m value greatly aects
D=Dave . Generally, D=Dave is suppressed with high m
value regardless of n value, suggesting a low probability of
166
breakage in the unsteady state. On the other hand, n value

aects D=Dave greatly in the case of low m value but
negligibly in the case of high m.
These results reveal that m value under high-temperature
conditions aects the deformation prole more greatly than
n value at room temperature during dieless drawing. Here,
the deformation behavior in the unsteady state is considered
to be the same as that upon the occurrence of necking
during tensile testing. Namely, in the case of a material with
high m value, when the strain rate of the necking part
increases, the ow stress increases and further deformation
is suppressed. This behavior also occurs in the unsteady
state during dieless drawing and the unstable deformation
behavior is suppressed. Most materials exhibit strain rate
sensitivity at high temperatures, which also illustrates the
importance of taking into consideration the strain rate
sensitivity index m value in the analysis. Moreover, to
achieve successful drawing without unstable deformation, it
is necessary to select a material with high m, such as
superplastic materials, regardless of the magnitude of n
value at room temperature. The analytical results reveal the
eectiveness of superplastic dieless drawing proposed in
this study. Furthermore, for a material with low m under
high-temperature conditions, selecting a material with high
n value at room temperature is eective for achieving
successful drawing.
5.
Conclusions
In this study, we developed an FE model with coupled

analysis for superplastic dieless drawing, in which a material
with strain rate sensitivity and strain hardening sensitivity,
the heat transfer, and the deformation were taken into
consideration. The validity of the model and the eects of
each condition on the deformation prole were studied, and

the following conclusions were obtained.
(1) The validity of the numerical model was demonstrated
from the aspects of heat transfer and deformation. The
results of FE analysis agreed well with the temperature
distribution and the deformation prole obtained in the
experiments, indicating the validity of the FE simulation coupled with thermo-mechanical analysis.
(2) The eects of m and n, which are material characteristics with temperature dependence, on the deformation
prole were investigated. As a result, it was found that
m greatly aects the deformation prole, whereas n
value aects the deformation prole only in the case of
a material with low m value. The importance of taking
into consideration the strain rate sensitivity during
dieless drawing was veried.
(3) It was found that the higher the m value of a material,
the more the unstable deformation is suppressed. This
demonstrates the eectiveness of superplastic dieless
drawing numerically.
REFERENCES
1) E. Nakamachi: J. Jpn. Soc. Mech. Eng. 107 (2004) 45. (in Japanease)
2) T. Furushima, K. Manabe and T. Sakai: J. Jpn. Soc. Technol. Plast. 47
(2006) 870874. (in Japanease)
3) T. Furushima and K. Manabe: J. Mater. Process. Technol. 187188
(2007) 236240.
4) K. Kobatake, H. Sekiguchi, K. Osakada and K. Yoshikawa: J. Jpn. Soc.
Technol. Plast. 21 (1980) 5258. (in Japanease)
5) Z. T. Wang, G. F. Luan and G. R. Bai: J. Mater. Process. Technol. 94
(1994) 7377.
6) T. Furushima, T. Sakai and K. Manabe: Proc. NUMIFORM04 (2004)
pp. 522527.
7) K. Kobatake, H. Sekiguchi, K. Osakada and K. Yoshikawa: J. Jpn. Soc.
Technol. Plast. 20 (1979) 814819. (in Japanease)

Fine Measurement of Thermal Conductivity for (Bi0:5 Sb1:5 )Te3 Compounds

Takashi Hamachiyo, Maki Ashida and Kazuhiro Hasezaki
Department of Materials Science, Shimane University, Matsue 690-8504, Japan
A ne measurement system was constructed for thermal conductivity based on the static comparison method. The thermal conductivity of
the sample was 1.397 W/mK. The system has an accuracy of less than 1% for 1.411 W/mK of the reference quartz. The (Bi0:5 Sb1:5 )Te3 samples
were prepared by mechanical alloying followed by hot pressing. Grain sizes were controlled in a range from 1 to 10 mm by controlling the
sintering temperature in a range from 623 to 773 K. The thermal conductivity was 0.89 W/mK for the sample sintered at 623 K, where the grain
size of 1.75 mm was measured by scanning electron microscopy (SEM). The thermal conductivity increased on the sample sintered at 673 K
because of grain growth and decreased on those sintered at the temperatures from 673 to 773 K because the increase of pore size caused to
decrease thermal conductivity. The increase of thermal conductivity for the samples sintered at temperatures above 773 K was aected by the
increase of carrier concentration. [doi:10.2320/matertrans.MRA2008241]
Keywords: thermoelectric material, bismuth telluride, thermal conductivity
1.
Introduction
Bi2 Te3 -based compounds are known to be the highest

thermoelectric gure of merit at room temperature and are
widely used as thermoelectric materials for electronic cooling devices. Crystal grains in thermoelectric materials have
been controlled to attain a low thermal conductivity leading
to an enhanced gure of merit. However, the relationship
between grain size and thermal conductivity is not always
claried for thermoelectric materials. Powder metallurgical
processes such as mechanical alloying (MA) have been used
in the study of Bi2 Te3 -based compounds, which has been
adopted to obtain homogeneous materials with ne grain
sizes.1,2) The ne grain size materials had lower thermal
conductivities than single crystals of the same materials
because phonons can be scattered at grain boundaries. In the
Bi-Te based alloys, however, detailed research of the thermal
conductivity dependence on grain size is still limited. Since
the thermal conductivity of thermoelectric materials is small,
the measurement is very dicult. The laser ash method used
for thermal conductivity measurements is known to have an
accuracy of approximately 10%.36)
In the present study, a ne static comparative thermal
conductivity measurement system was constructed. Bi2 Te3 based compounds prepared by MA and hot pressing (HP)
were investigated for their grain size dependence on thermal
conductivity.
2.
Experimental
The MA samples were p-type (Bi0:5 Sb1:5 )Te3 +

0.07 mass% Te. The constituent elements Bi(5N), Sb(6N),
and Te(5N) were weighed as target compositions, put in
stainless steel vessels and milled with silicon nitride balls.
These powders were introduced into the vessels under an
argon atmosphere and then sealed. The MA process was
carried out in a planetary ball mill for 30 h. The MA powder
was set in a carbon mold and sintered by hot pressing at a
sintering temperature Ts from 623 to 823 K under mechanical
pressure of 39 MPa for 10 h.
Fig. 1 The schematic diagram of the constructed thermal conductivity

measurement system based on static comparison method.
The microstructures were observed using optical microscopy (OM) and scanning electron microscopy (SEM).
Dierential thermal analysis (DTA) and thermogravimetric
analysis (TG) were performed at heating rates of 20 K/min in
a nitrogen gas ow. The grain size was measured by the
linear intercept technique.
Thermal conductivity () was measured using the constructed static comparison method system. Figure 1 shows
the schematic diagram of the system constructed in order to
measure thermal conductivity at room temperature. The
measurement system consisted of a reference quartz and
measurement sample. Quartz ( 1:411 W/mK) was used
as a reference sample. The Bi2 Te3 sample and quartz were
sandwiched each other between copper blocks of 10 mm in
diameter. A lm heater of thickness 0.8 mm was used as a
heat source in order to prevent the heat radiation toward the
side wall. The water-cooled sink was used for the low
temperature side. The thermal conductivity for the static
comparison method is given by eq. (1):
s ref fAref Tref Lref =As Ts Ls g;
where s and ref are the thermal conductivity (W/mK) of the

sample and reference sample, As and Aref are the crosssectional area (m2 ) of the sample and reference sample, Ts
168
T. Hamachiyo, M. Ashida and K. Hasezaki
and Tref are the temperature dierence (K) measured for the
sample and reference sample, Ls and Lref are the length (m) of
the sample and reference sample, respectively. The Seebeck
coecient was estimated from the linear relationship
between the thermoelectromotive force (E) and temperature
dierence (T) of 3 K at room temperature by E=T.
3.
Figure 2 shows the plots of input power P against

temperature dierence T for the same quartz sample used
as both the reference and measurement samples. R2 is the
authenticity coecient, which was estimated to be 1. The
slope of the line of Fig. 2 was substituted for eq. (1) and the
thermal conductivity was calculated. The thermal conductivity of the sample was 1.397 W/mK. The system has an
accuracy of less than 1% for 1.411 W/mK of the reference
quartz.
Figure 3 shows (1) optical micrographs of polished surface
and (2) SEM photographs of etched surfaces for the samples
sintered at temperatures from 623 to 823 K, respectively.
Fig. 2 Plots of input heating power P against temperature deference T

measured by static comparison method. R2 is authenticity coecient given
by regression analysis.
(a-1)
(b-1)
(c-1)
(a-2)
(b-2)
(c-2)
(d-1)
(e-1)
(d-2)
(e-2)
Fig. 3 (1) Optical micrographs of polished surface and (2) SEM photographs of etched surfaces for the samples sintered at (a) 623 K,
(b) 673 K, (c) 723 K, (d) 773 K and (e) 823 K, respectively.
Fine Measurement of Thermal Conductivity for (Bi0:5 Sb1:5 )Te3 Compounds
Fig. 4 DTA and TG curves for (Bi0:5 Sb1:5 )Te3 + 0.07 mass%Te powder
prepared by MA.
169
Fig. 6 Plots of thermal conductivity at room temperature against

sintering temperature Ts of (Bi0:5 Sb1:5 )Te3 + 0.07 mass%Te.
Figure 6 shows plots of thermal conductivity against the

sintering temperature Ts for MA samples. The value for the
sample of Ts 623 K was 0.89 W/mK at the grain size of
1.75 mm. The thermal conductivity increased to 1.01 W/mK
for the sintering temperature of Ts 673 K, which was
attributable to the grain growth from 1.75 to 2.83 mm. The
decrease of thermal conductivity for the samples of Ts from
623 to 773 K occurred due to the presence of pores caused
from the evaporation of Te in Fig. 3. The thermal conductivity was then increased in the Ts range between 773 and
823 K. In general, the thermal conductivity of the thermoelectric materials is given eq. (3):
Fig. 5 Plots of Seebeck coecient at room temperature against sintering
temperature Ts of (Bi0:5 Sb1:5 )Te3 + 0.07 mass%Te.
Many pores were observed in MA samples at the sintering

temperature over 723 K as shown in Figs. 3(c)(e). The grain
size of (Bi0:5 Sb1:5 )Te3 alloys was found to increase from 1.75
to 10.5 mm with an increase of sintering temperature.
Figure 4 shows DTA and TG curves on the heating of MA
powder. The endothermic peak around 873 K corresponds to
the melting point of (Bi0:5 Sb1:5 )Te3 .7) The TG curves began
to decrease at 723 K, which was due to the evaporation of
Te.8) From Fig. 3 and Fig. 4, pores observed over 723 K was
considered to have formed as a result of the evaporation of
Te.
Figure 5 shows plots of the Seebeck coecient against
sintering temperature Ts for the MA samples at room
temperature. The value gradually decreased with an increase
of Ts . The Seebeck coecient is dependent on the carrier
concentration n dened as follows:9)
(
3)
k
5
22m kT 2
;
2
log

nh3
e
2
where k is Boltzmanns constant, e is the electron charge, is
the scattering factor, m is eective mass and h is Planck
constant.
It is known that the carrier concentration of Bi2 Te3 -based
compounds increases as the amount of Te decreases. These
results are found to be consistent with the available data
that evaporation of Te caused the carrier concentration to
increase.10,11)
ph el ;
where ph is the lattice thermal conductivity and el is the

carrier thermal conductivity. According to the WiedemannFrantz law, el can be expressed as:
el LT LTne;
where L is the Lorentz number, is the electric conductivity

and is the carrier mobility.
Since the decrease of the Te content by evaporation was
responsible for the increase of carrier (hole) concentration,
and the thermal conductivity is given by the sum of the
contribution of electrons and phonons, the increased thermal
conductivity was attributable to the carrier concentration
increased by the evaporation of Te.
4.
Conclusions
The measurement system for thermal conductivity was

constructed and applied for the (Bi0:5 Sb1:5 )Te3 sample
prepared by the MA method. The eects of grain size on
the thermal conductivity were investigated for the samples
sintered at temperature Ts between 623 and 823 K.
(1) A ne measurement system was constructed for thermal
conductivity based on the static comparison method. The
thermal conductivity of the sample was 1.397 W/mK. The
system has an accuracy of less than 1% for 1.411 W/mK of
the reference quartz.
(2) Sintering at temperatures ranging from 623 to 673 K
caused to increase the thermal conductivity from 0.89 to
1.01 W/mK, resulting from the grain growth.
170
T. Hamachiyo, M. Ashida and K. Hasezaki
(3) Sintering at temperatures ranging from 673 to 723 K

caused the evaporation of Te and the formation of pores in
the samples. The thermal conductivity decreased from 1.01 to
0.82 W/mK.
(4) The thermal conductivity of the sample sintered over
773 K was increased because of the increase of the carrier
concentration due to Te evaporation.
Acknowledgement
The authors thank Dr. Isoda of the National Institute for
Materials Science, Japan for technical advice to measure the
thermal conductivity of the reference quartz.
REFERENCES
1) K. Hasezaki, M. Nishimura, M. Umata, H. Tsukuda and M. Araoka:
Mater. Trans. JIM 35 (1994) 28432.

2) H. Kohri, X. Dauphin, K. Hasezaki, I. A. Nishida and I. Shiota: Proc.
20th Int. Conf. on Thermoelectrics, Beijing, (2001) pp. 121124.
3) T. M. Tritt, Ed. by D. Rowe: Thermoelectrics handbook: macro to
nano, ch23, (CRC Press. New York, 2006) pp. 1115.
4) T. M. Tritt: Thermal Conductivity, (Kluwer Academic/Plenum Publishers, 2004) pp. 187200.
5) V. V. Mirkovich: J. Am. Ceram. Soc. 48 (1965) 387.
6) T. Caillat, J.-P. Fleurial and A. Borshchevsky: J. Phys. Chem. Solids
58 (1997) 11191125.
7) H. Scherrer and S. Scherrer, Ed. by D. Rowe: Thermoelectrics
handbook: macro to nano, ch27, (CRC Press. New York, 2006) pp. 1
18.
8) Ed. by Y. Suga: Thermoelectric semiconductor, (Makisyoten, Tokyo,
1966) p. 317.
9) Ed. by Y. Suga: Thermoelectric semiconductor, (Makisyoten, Tokyo,
1966) pp. 100103.
10) C. B. Sattethwaite and R. W. Ure, Jr.: Phys. Rev. 108 (1957) 1164
1170.
11) K. Hasezaki: J. Jpn. Soc. Powder Powder Metall. 54 (2007) 543546.

#2009 The Mining and Materials Processing Institute of Japan
Removal of Trace Impurity from Limestone Using Flotation Techniques

Masami Tsunekawa1; * , Yugo Honma1 , Kyoungkeun Yoo2 , Naoki Hiroyoshi1 and Mayumi Ito1
1
2
Graduate School of Engineering, Hokkaido University, Sapporo 060-8628, Japan

Korea Institute of Geoscience & Mineral Resources, Daejeon 305-350, Korea
High grade limestone was obtained by removing trace impurities through reverse and carrier otation techniques. A simple method for
measuring the impurity content of limestone was developed by correlating the amount of impurities and the turbidity of suspensions of residue
from limestone dissolved in 20% acetic acid.
Impurity removal more than 50% with limestone recovery better than 85% was obtained under suitable conditions by both reverse otation
and carrier otation using sodium oleate (NaOl) and dodecylammonium acetate (DAA) as collectors. Limestone containing more than
0.15 mass% impurity was used as the carrier. These results suggest that limestone treated by otation can be used as high quality limestone in
paper manufacturing. [doi:10.2320/matertrans.M-MRA2008839]
(Received January 23, 2008; Accepted October 7, 2008; Published December 3, 2008)
Keywords: trace impurity, limestone, reverse otation, carrier otation
1.
Introduction
A large amount of limestone (CaCO3 ) is used widely in

various industries such as cement, plastic, and paper, due to
its low cost.13) However, high grade limestone deposits are
being depleted and low grade deposits need to remove their
impurities before they can be utilized.3)
Limestone to be used as ller for paper requires strict
quality control in properties such as particle size and
whiteness.2) Limestone containing more than 0.1 mass% of
trace impurities is dicult to use, because the impurities
decrease the whiteness of paper and the durability of the
equipment used in paper manufacturing. Here, the impurities
are dened as materials not dissolved by acetic acid through a
method developed in this study. Due to this, limestone with
more than 0.1 mass% impurities is only used for cement. The
establishment of a method to remove trace impurities from
limestone with impurity contents above 0.1 mass% is
important to be able to utilize such resources as ller for
paper manufacturing.
Flotation is the most important and versatile mineral
processing technique for ne particle processing and upgrading of ore. Both uses and applications of otation are being
expanded to treat greater tonnages and to cover new areas.4)
There are various otation researches related to limestone:3,510) (1) calcite or limestone is removed as a gangue
to recover uorite,5,6) apatite,7) or phosphate ore;9,10) (2)
limestone is beneciated to use for cement or paper
manufacturing;1114) and (3) silica is removed from limestone
using an emulsion otation technique.12) However, there
have been no papers reporting the removal of trace impurities
from limestone using otation, obtaining high grade limestone as ller in paper manufacturing.
The present study characterized trace impurities in a
limestone by X-ray diractometry and X-ray uorescence
analysis, then determined the turbidity using 20% acetic acid
and glycerol, and the trace impurities were removed by a
modied otation technique.
*Corresponding
author, E-mail: tunekawa@eng.hokudai.ac.jp
2.
Materials and Experimental Procedures
2.1 Materials
Two samples of limestone from the Kawara mine in
Fukuoka, Japan were used and the impurity contents were
determined to be 0.13 mass% and 0.15 mass% by the method
described below in 2.2. The sample with 0.13 mass%
impurity was ground with a ball mill (UBM-2L, Masuda
Co. Ltd., Japan) and the ground product was used as feed for
the otation tests. The median diameter of the ground product
was 9.45 mm and the 90% of cumulative undersize (D90 )
particle diameter was 40.54 mm. The limestone with
0.15 mass% impurity was used as carrier in the otation
tests, because aside from it can adsorb the impurities, it can
be used for cement and has little eect on the grade of the
concentrate.
Impurity particles in the limestone were collected by
dissolving limestone in 20% acetic acid (Wako Pure Chemical Industries, Ltd.) at 5 mass% pulp density and centrifuging
the suspension at 17600 g centrifugal acceleration for
10 min in a RS-18III high speed refrigerated centrifuge
(TOMY SEIKO Co. Ltd., Japan). The precipitate was dried
at 65 C overnight and characterized by the method described
in 2.3.
2.2
Quantitative analysis of impurity in the limestone by

turbidity measurement
The impurity particles in limestone collected from the
centrifuge tube by the method described in 2.1 (residue
obtained by dissolving limestone with 20% acetic acid) were
added to a mixture of 25 cm3 of glycerol and 25 cm3 of
distilled-deionized water (DDW). After agitation, the suspension was introduced into a glass cell, and the turbidity of
the suspension in the glass cell was measured with a
turbidimeter (SEP-PT-201, Nihon Seimitsu Kagaku Co.,
Ltd., Japan).
2.3 Characterization of the impurities
The impurity particles obtained using the method described in 2.2 were analyzed with an X-ray diractometer (XRD,
172
M. Tsunekawa, Y. Honma, K. Yoo, N. Hiroyoshi and M. Ito
JEOL JDX-3500), an energy diusive type X-ray uorescence analyzer (XRF, JEOL JSX-3210A), a CHN-corder
(Yanako, Japan), a microscope, and a particle size analyzer
(Microtrac MT3000, Brookhaven Instruments Co.). The
XRD data were obtained with a monochromator under the
following conditions: radiation, Cu K, 30 kV, 200 mA; step
scanning; time constant 0.5 seconds; and angle range 060
degree/2. The XRF was measured under the following
conditions: Rh, 30 kV; time 300 seconds. Before analyzing
the particle size, the impurity particles were conditioned to
disperse fully by ultrasonic treatment for 30 seconds.
The zeta potentials of the impurities were analyzed with a
zeta potential analyzer (Zeta-Plus, Brookhaven Instruments
Co., USA).
removal ratio
2.4 Flotation
Reverse otation and carrier otation were conducted to
remove the impurities from the limestone. The limestone,
ground as in section 2.1, with 0.13 mass% impurity and
9.45 mm median diameter was used as feed for the otation
tests. All otation experiments were performed using a
otation tester (FT-1000, Heiko Seisakusho Ltd.): otation
cell 8 cm long, 8 cm wide, and 12 cm high, and with a 5 cm
diameter impeller blade. The pH of the suspensions was
adjusted with 1 kmolm3 of NaOH or H2 SO4 . All reagents
were of analytical grade.
The removal ratio was calculated by determining the
dierences between the amount of impurities in 3 g of
limestone before and after the otation tests.
impurity amount before otationimpurity amount after otation

100
impurity amount before otation
The recovery ratio was determined by the following equation:

amount of limestone in non-froth product
100
limestone amount of feed
2.4.1 Reverse otation

A 40 g of limestone feed was added to 340 cm3 of distilleddeionized water (DDW). With a magnetic stirrer, the
suspension was agitated at 600 rpm for 5 min, then
0.015 cm3 methyl isobutyl carbinol (MIBC, Nippon Kasei
Chemical Co., Ltd.) as frother and 0.0150.03 cm3 kerosene
(Wako Pure Chemical Industries, Ltd.) as collector were
added. The suspension was agitated for conditioning at
600 rpm for 5 min. After conditioning, the suspension was
introduced into the otation cell, and 60 cm3 of DDW was
added to adjust the pulp density to 10 mass%, and the pH of
the suspension was adjusted to a desired value. Flotation
was conducted under the following conditions: air ow
rate, 1,200 cm3 min1 ; otation time, 5 to 30 min; impeller
speed, 1200 rpm. The impurities were removed as froth, and
the froth and the non-froth product were dried at 105 C for
one day, then weighed and their impurity contents were
measured.
2.4.2 Carrier otation combined with reverse otation
The limestone with 0.15 mass% impurity was used as the
carrier. Carrier otation was conducted using sodium oleate
(NaOl) or dodecylammonium acetate (DAA) as collectors
under the following conditions: for the limestone, suspension
30 g of limestone feed was added to 270 cm3 of DDW; for the
carrier suspension, 10 g of carrier was added to 90 cm3 of
DDW. Each suspension was agitated at 600 rpm for 10 min
with a magnetic stirrer and then for another 3 min after
adding a desired amount of collector. The two suspensions
were mixed and conditioned in a otation cell at 1700 rpm
for 5 min after adding 0.015 cm3 of MIBC, 0.015 cm3 of
kerosene (or isooctane), and 40 cm3 of DDW, and adjusting
the pH. Flotation was conducted for 10 min under the same
conditions as in the reverse otation.
3.

The impurity content of the limestone is extremely low. A
20
R2 = 0.9758
Turbidity / -
recovery ratio
15
10
0
0
The amount of impurities / mg

Fig. 1 Turbidities (arbitrary) of suspensions containing various amount of
impurities, where the impurities were collected as residue by dissolving
limestone with 20% acetic acid.
simple and accurate method was developed to determine

trace amounts of impurities by dissolving limestone in 20%
acetic acid solution. Figure 1 shows the turbidity of suspensions containing known amounts of impurities. There was a
good correlation between the turbidity and the amount of
impurities (R2 0:9758). The trace amount of impurities
was determined by this method using the calibration curve
shown in Fig. 1.
Figure 2 shows the XRD pattern of the impurities with
peaks for quartz and calcite observed; some calcite remains
after dissolving the limestone with 20% acetic acid. As can
be seen in Table 1, the results of the XRF analysis indicate
that the major components of the impurities are Si and Ca,
agreeing with the results of XRD. The impurities were
analyzed with a CHN-corder to measure the carbon content
100
Cumulative / %
Intensity/counts
3000
Q Ca
2000
1000
Ca
173
Q
Ca
80
60
40
20
10 20 30 40 50 60
2 / Deg.
0
1
Fig. 2 XRD pattern of impurities collected from the limestone dissolved

with 20% acetic acid. Q, quartz; Ca, calcite.
Element
Ca
Si
Fe
Al
Mg
Zn
Cr
w/w%
39.8
34.8
10.2
6.8
4.0
2.6
1.3
0.5
Table 2 Amounts of C, H, and N in the impurities determined by CHNcorder, where the impurities were collected as residue by dissolving
Element
w/w%
10.5
0.7
Not detected
100
1000
Particle size / m
Fig. 4 Cumulative undersize distribution of impurities collected from the
dissolved limestone.
Recovery and removal ratio / %
Table 1 Composition of the impurities analyzed with XRF, where the

impurities were collected as residue by dissolving limestone with 20%
acetic acid.
10
100
80
60
40
Recovery ratio
20
Removal ratio
0
0
10
20
30
40
Flotation time / min

Fig. 5 Limestone recovery and impurity removal ratios with increasing
otation time in reverse otation. Conditions: suspension vol. 400 cm3 ;
pulp concentration 5%; MIBC 0.0015 cm3 ; kerosene 0.0015 cm3 .
50
Fig. 3 Microscopic observation of impurities collected from the dissolved

limestone.
and the result is shown in Table 2. The carbon content other

than CaCO3 is 10.5 mass% in the impurities, showing that
quartz is a main component of the impurities. As shown in
Fig. 3, microscopic observation indicated that the silicate
particles are partly coated with carbon components.
Figure 4 shows the particle size distribution of the impurities. The cumulative undersize of 10, 50, and 90% were
obtained as 8.39 mm, 44.35 mm, and 158.70 mm, respectively.
Because 90% of the impurities were coarser than 9 mm, about
90% of the impurities could be liberated by grinding
limestone to ner than 10 mm.
Because the impurities are coated partly with carbon, it is

expected that they can be removed by otation using
kerosene as a collector for carbon. This otation technique
was conducted to obtain high grade limestone as the nonfroth product, therefore this is termed as reverse otation.
Figure 5 shows the removal ratio of the impurities and the
recovery ratio of the limestone as a function of otation time.
The removal ratio increases from 21.0% at 5 min to 65.7%
at 30 min, while the recovery ratio is almost constant and
higher than 97%. To obtain high removal ratio, longer
otation time is needed due to low collision probability of air
bubbles and few amount of impurities. Figure 6 shows the
relation between the removal ratio and the recovery ratio to
kerosene addition with otation time kept at 10 min in the
reverse otation. The removal ratio increases from 34.2 to
58.8% with increase in the amount of kerosene and the
100
Zeta potential / mV
174
80
60
40
Recovery ratio
20
Removal ratio
0.01
0.02
0.03
Amount of collector added /
0.04
cm3
Fig. 6 Limestone recovery and impurity removal ratios with increasing

amounts of collector added in the reverse otation: pulp concentration 5%;
MIBC 0.0015 cm3 ; otation time 10 min.
recovery ratio is always higher than 95%. Generally,

kerosene adsorbs on surfaces of hydrophobic particles such
as carbon and it accelerates adhesion of particles on the
bubbles and improves otation performance. This would be
the reason why the removal ratio of impurities increases with
increasing amounts of kerosene.
As stated above, the reverse otation using kerosene as a
collector is an eective method to remove impurities in
limestone, but the removal ratio of impurities is not sucient
when kerosene amounts are low and otation times are short:
for example the removal ratio is only 21.0% with 5 min of
otation time and 0.015 cm3 kerosene addition.
There are two reasons for the low removal ratio of
impurities in the reverse otation: (1) when the surface
coverage of silica particles with carbon is small due to low
carbon content, the particles behave as hydrophilic and this
causes the low oatability of particles. In this case, addition
of collector, which will change the silica surface to hydrophobic, may be eective to improve the removal of silicate
impurities; (2) because the inertia of very ne impurities is
very small, they move together with the uid and the
probability of collision between particles and bubbles is low.
To overcome this, carrier otation combined with reverse
otation was applied, where carrier particles of sizes
suciently large to collide with bubbles were introduced in
the otation pulp. The carrier attaches to the ne particles,
then adsorbs on bubbles and oats. It has been widely
reported that carrier otation is eective to remove ne
particles in small amounts15) with coal,16) anatase,17) and
gold.18) In the carrier otation, a low-grade limestone with
0.15 mass% impurities and of 106149 mm particle size was
used as the carrier, because the low-grade limestone is easily
obtained in situ and the carrier/impurity mixture recovered
by the otation can be expected to nd usage in cement
production.
The quartz surface is negatively charged over a wide pH
range and with a cationic surfactant (DAA) it can be expected
to adsorb on silica surfaces by electrical attraction19) and act
140
120
100
80
60
40
20
0
-20
-40
-60
-80
-100
Silica with DAA
Impurity with DAA
Impurity without DAA
Silica without DAA
10
11
12
pH
Fig. 7 Zeta potential of the impurities collected from the dissolved
limestone (circle) and silica (rhombus) with (open) and without (solid)
0.0001 moldm3 DAA.
as a collector. For simplicity, the carrier otation using DAA

is termed as DAA-carrier otation in the succeeding
paragraphs.
Prior to the DAA-carrier otation experiments, the zeta
potentials of the impurities and silica with and without
0.0001 moldm3 DAA were determined to conrm if the
potential of the impurities is similar to that of silica. The
result is shown in Fig. 7. Here, silica was used as the
substitute for the main component of the impurities, i.e.
quartz. The zeta potentials of the impurities without DAA
agree well with the reported values20,21) for quartz and the
curve shapes of the impurities and silica were similar. Both
the zeta potentials of the impurities and silica with DAA
changed from negative to positive and are close to each other.
This suggests that the surfaces of the impurities behave like
the silica surface and that DAA can be adsorbed.
Carrier otation using 0.001 moldm3 DAA as a collector
and 0.015 cm3 isooctane as an accelerator was performed.
Figure 8 shows that the removal and the recovery ratios
decrease with the increase in pH from 9 to 11. The
dissociation constant of dodecylammonium ions is 10.63,22)
and above pH 10.6 the capability of DAA as a collector
would decrease. The highest removal ratio of 58.3% was
obtained at pH 9.
The eect of DAA concentration on the recovery and the
removal ratios was investigated at pH around 10. As shown
in Fig. 9, with increasing DAA addition, the removal ratio
increases and the recovery ratios decreases sharply. The
cationic collector DAA adsorbs on the negatively charged
surface of impurities and carrier, and both collide and
adhere to each other mainly due to hydrophobic bond, then
oat. Because DAA acts both as a collector and a frother, the
increase in DAA addition produces an abundant froth
volume, causing the decrease in the recovery of limestone
as non-froth product.
Sodium oleate (NaOl) acts as a collector for quartz23,24) in
the presence of Ca2 ions. The preliminary experiment
results of bubble pick-up tests for the impurity showed that
80
60
40
Recovery ratio
20
Removal ratio
0
9
10
11
100
80
Recovery ratio
60
Removal ratio
40
20
0
149-210
Removal (%)
60
40
297-590
590-850
Fig. 10 The eect of carrier size on limestone recovery and impurity

removal ratios in the NaOl-carrier otation. Conditions: suspension vol.
400 cm3 ; pulp concentration 5%; MIBC 0.015 cm3 ; kerosene 0.015 cm3 ;
NaOl 0.0004 moldm3 ; otation time 10 min; pH 8.08.4.
100
80
210-297
Carrier size / m
pH
Fig. 8 The eect of pH on limestone recovery and impurity removal ratios
in the DAA-carrier otation. Conditions: suspension vol. 400 cm3 ; pulp
concentration 5%; MIBC 0.0015 cm3 ; isooctane 0.0015 cm3 ; DAA
0.0001 moldm3 ; otation time 10 min.
175
100
100
80
80
56.3%
60
60
40
40
20
20
Carrier Size:-210 +150
Product Recovery (%)
100
-4
20
NaOl: 3.74 10 M
3 4 5 6 7 8 9 10 11 12 13
0
0.0004
0.0008
0.0012
collector concentration / moldm-3

Fig. 9 The eect of DAA concentration added on limestone recovery and
impurity removal ratios in the DAA-carrier otation. Conditions:
suspension vol. 400 cm3 ; pulp concentration 5%; MIBC 0.015 cm3 ;
isooctane 0.015 cm3 ; otation time 10 min; pH 10.0. The symbols are the
same as in Fig. 9.
the impurity particles became hydrophobic with NaOl, thus

carrier otation using NaOl was conducted and is termed as
NaOl-carrier otation.
The eect of particle size of the carrier in NaOl-carrier
otation was investigated with these carrier size fractions:
149210 mm, 210297 mm, 297590 mm, and 590850 mm.
Figure 10 shows that the removal ratio decreases with
increasing carrier size, while the recovery ratio remains
almost constant. This could be explained from the fact that
smaller sized carriers have larger specic surface areas and
adsorb more impurity particles such as quartz and carbon.
The 149210 mm carrier was used in the succeeding experiments.
Removal (%)
pH
Product Recovery (%)
Fig. 11 The eect of pH on limestone recovery and impurity removal

ratios in the NaOl-carrier otation. Conditions: suspension vol. 400 cm3 ;
pulp concentration 5%; MIBC 0.015 cm3 ; kerosene 0.015 cm3 ; NaOl
0.0004 molm3 ; otation time 10 min.
Figure 11 shows the eect of pH on the removal ratio in

the NaOl-carrier otation. The highest ratio (56.3%) with
85.2% recovery was obtained in the pH region between 10
and 11, where activation of the quartz by calcium species
dissolved from limestone is expected to occur.23,24) This
activation is explained by the formation of Ca-oleate salts in
solution or in the vicinity of the quartz surface followed by
adsorption. This adsorption mechanism requires the presence
of Ca(OH) ions in solution, which only becomes signicant
near pH 11.23) At pH values lower than 11, concentrations of
Ca(OH) decrease, and this would cause a decrease in the
adsorption amount of the collector on the quartz surface and
a decrease in its oatability.
As discussed above, the impurity removal ratio of more
than 50% with limestone recovery higher than 85% was
176
obtained under suitable conditions by both reverse otation

and carrier otation. As the whiteness index of limestone
with less than 0.06 mass% impurity is higher than 90 (Hunter
whiteness degree measured according to JIS-P8123), the
limestone treated by the otation methods can be used as a
high quality limestone in paper manufacturing.25)
REFERENCES
1)
2)
3)
4)
5)
4.
Conclusions
Trace impurities from limestone were removed by reverse

otation and carrier otation combined with reverse otation.
A simple method of measuring impurity content of limestone
was developed by correlating the amount of impurity and the
turbidity of residue suspensions prepared by dissolving
In the reverse otation with kerosene as collector, the
removal ratios increased from 21.0% to 58.8% and 65.7%
with the increase in the amount of collector and otation
time, respectively. In the carrier otation combined with
reverse otation, where limestone with 0.15 mass% impurities was used as the carrier, the removal ratios were 56.3%
with NaOl as collectors at pH 1011 and 58.3% with DAA at
pH 9, respectively. These results suggest that more than 50%
of trace impurity can be removed by otation techniques, and
the non-froth product can be used as a high quality limestone
in paper manufacturing.
Acknowledgements
The authors gratefully acknowledge the support provided
by Taiheiyo Cement Corporation, and would like to thank
Mr. S. Okawa and Mr. S. Okada for their help.
6)
7)
8)
9)
10)
11)
12)
13)
14)
15)
16)
17)
18)
19)
20)
21)
22)
23)
24)
25)
K. Huruyama: Limestone, No. 255 (1992) 4053.

T. Ikegami: Eng. Mater. 44 (1996) 3637.
K. V. Rao and P. N. Pathak: Met. Mater. Process. 14 (2002) 3540.
B. A. Wills: Mineral Processing Technology (6th ed.) (ButterworthHeinemann, Oxford, 1997) pp. 267344.
V. V. Morozov, H. Baldauf and H. Schubert: Int. J. Miner. Proc. 35
(1992) 177189.
M. Tsunekawa and T. Takamori: J. MMIJ 97 (1981) 11871192.
J. Hanna and I. Anazia: Miner. Metall. Proc. 7 (1990) 8489.
Y. Hu and Z. Xu: Int. J. Miner. Proc. 72 (2003) 8794.
N. A. Abdel-Khalek: Miner. Eng. 13 (2000) 789793.
R. C. C. Ribeiro, J. C. G. Correia, M. B. M. Monte, P. R. Seidl, C. G.
Mothe and C. A. Lima: Miner. Eng. 16 (2003) 873875.
H. Peter: Pit and Quarry 77 (1984) 3640.
O. Ofor: J. Min. Geology 33 (1997) 3538.
F. J. Wellenkamp: Miner. Proc. 41 (2000) 522528.
S. Rao, B. V. Narayana and S. Subramanian: Indian J. Eng. Mater. S. 4
(1997) 139144.
T. V. Subrahmanyam and K. S. Eric Forssberg: Int. J. Miner. Proc. 30
(1990) 265286.
G. Atesok, F. Boylu and M. S. Cel
k: Miner. Eng. 14 (2001) 661670.
S. Mathur: J. Colloid Interface Sci. 256 (2002) 153158.
L. Valderrama and J. Rubio: Int. J. Miner. Proc. 52 (1998) 273278.
M. C. Fuerstenau, Oxide, Silicate Flotation: Flotation Science and
Engineering (K. A. Matis, Eds., Marcel Dekker Inc., New York, 1995)
p. 91.
K. P. Ananthpadmanabhan, P. Somasundaran and T. W. Healy: Trans.
SME/AIME 266 (1979) 20032009.
J. S. Laskowski and J. Ralston: Colloid Chemistry in Mineral
Processing (Elsevier, Amsterdam, 1992) p. 252.
M. Tsunekawa, E. Ogata, H. Habata, T. Takamori and T. Hirajima:
J. Soc. Powder Technol. (Jpn.) 27 (1990) 603608.
S. C. Brown, Y. I. Rabinovich and B. M. Moudgil: Miner. Metall. Proc.
21 (2004) 164168.
F. Rashchi, Z. Xu and J. A. Finch: Colloid Surface A 132 (1998)
159171.
M. Tsunekawa, M. Ito, N. Hiroyoshi, Y. Honma, S. Okada and S.
Okawa: Toku Kai 2006-256893 (2006).

Eects of Water Absorption on Impact Value of Aluminum Dispersed

Composite Nylon6
Masae Kanda* and Yoshitake Nishi
Department of Metallurgical Engineering, Graduate School of Engineering, Tokai University, Hiratsuka 259-1292, Japan
Composite polymers, which are nylon6 materials with aluminum powders dispersed homogeneously, are prepared. Inuences of water
absorption on Charpy impact value of composite polymers have been investigated. Although the aluminum dispersion from zero to 40 vol%Al
decreases the impact value, the water absorbing treatment for 105 s at boiling point largely increases the impact values of all composite
polymers. It is explained that water molecules in nylon6 mainly relaxes the impact force. If the water absorption also enhances the interfacial
energy to form the crack in nylon6 and its composites, the high impact value with smooth undulating surface can be explained.
(Received May 28, 2008; Accepted October 30, 2008; Published December 17, 2008)
Keywords: aluminum powders, nylon6, composite polymer, water absorption, charpy impact test
Introduction
Industrial waste is one of the serious environmental

problems, currently. As a solution, the injection process of
composite materials with high volume fraction of dispersed
particles of industrial waste has been suggested. When I
utilize building scrap by this new composite material, it is
eective to solve the environmental problem.
Polymer materials are useful substances with lightweight,
ductility, low cost and high productivity.13) The cheap
composite polymer dispersed by metal powders often feels
like heavy and cold such as metals.4) Namely, the mixing
metal powders in polymer often induce an attractive image
to make practical use of articles for new markets.
The most important point is how to increase the volume
of scrap in polymer. To increase the volume fraction of scrap
in polymer, a new injection molding with homogeneous
distribution of particles has been developed.4) The most
attractive point of the injection-molding machine is the
preparation method for the metal particles dispersed composite polymer, because the volume ratio of metal-particles in
nylon6 is more than 80 mass%. Thus, it is possible to make
practical uses of various articles with dierent combinations
of the composite polymers. The composite polymer prepared
by the injection machine has been applied to exterior articles,
which is often existed outside of buildings. The mechanical
property, especially impact brittleness, is also a serious
problem to make a practical use of the exterior articles.
However, the water absorption softens the polymers. In order
to evaluate the brittleness and to clear the serious problems
of many demands from various markets with high quality
controlling, the fundamental data related to impact value has
been expected.
Inuences of water absorption on Charpy impact value for
aluminum (Al) powders dispersed composite polymers have
been investigated.
*Graduate
Student, Tokai University
10mm
1.
8mm
80mm
Fig. 1
2.
4mm
Charpy specimen with sizes.
2.1 Sample
The composite polymer was homogeneously mixed with
nylon6 powders (UNITIKA, Ltd., A1030JR) and Al powder.
Thereafter, drying was performed for 10 hours at 363 K. The
specimen, which shape was JIS K7111, was molded at 533 K
after melting. The molten polymer was injected into the die,
which temperature was 323 K by injection molding machine
equipped with mixing screw (TKK Co., Ltd., mold locking
force 980.7 kN). The dispersed volume ratios of Al powders
in nylon6 were 10, 20, 30 and 40 vol%Al. Average diameter
of Al particle size was 28.8 mm. Figure 1 shows Charpy
specimen.
The sample size was 10 mm in width, 80 mm in length and
4 mm in thickness. The V notch in sample was precisely
formed. The size of the V notch is 2 mm.
2.2 Mass measurement for water absorption
Water absorption into composite nylon6 was performed
for 105 s at 373 K. Mass measurement time was 0 s, 102 s,
2 102 s, 5 102 s, 103 s, 2 103 s, 5 103 s, 104 s, 2
104 s, 5 104 s, and 105 s. The water absorption ratio was
expressed by the following eq. (1).
C mt m0 =m0 100
Here, C, m0 and mt were water absorption ratio (%), mass

values before and after water absorption, respectively.
2.3 Charpy impact test
To evaluate the impact fracture toughness,5,6) the Charpy
impact values of the composite polymer were measured using
M. Kanda and Y. Nishi
a standard impact fracture energy measurement system (JIS

K7111). The Charpy impact value was expressed by the
following eq. (2). By this measurement, a simple revision
method for energy is used at the time of the Charpy impact.
Since this method corrects an energy loss, the absorption
energy is directly calculated by using the simple blank test
angle 0 .
E WRcos cos
0
cos cos =
Here, E, W, R, , and were impact fracture energy (kJ),

hammer weight (N), length (m) of hammer weight point
from rolling center, start angle before impact, the maximum
angle after impact and the maximum angle of the blank
test, respectively. The Charpy impact value (kJm2 ) was
expressed by the following eq. (3). Charpy specimen
auc E=bN t 103
2.4 Change in fracture morphology

The fracture surface morphology of composite polymers
(0, 10 and 40 vol%Al powders in nylon6) before and after
water absorption was observed by using the scanning electron
microscope (SEM: S-3200N, HITACHI). To observe the
fracture surface, the samples have been coated for 60 s by
Au-Pd alloy.
3.
Results
Water absorption ratio (mass %)
3.1 Mass increase by water absorption

Figure 2 shows changes in water absorption ratio against
water absorption time in composite nylon6 dispersed at each
6
4
2
0
0
10
20
30
40
Volume of aluminum (vol%)
10
10
10
0
10
20
30
40
Fig. 4 Changes in Charpy impact value of composite polymers before and

water absorption for 105 s against volume change from 0 to 40 vol% of
dispersed Al powders in nylon6. ( ; water absorbed samples, ; dry
samples)
volume% of Al powders. As a result, all samples conrmed

water absorption and became to be approximately saturated
at 105 s of water absorption time.
Figure 3 shows changes in water absorption ratio at 105 s
of treated time against dispersed volume from 0 to 40 vol%
of dispersed Al powders in nylon6. It is conrmed that
the water absorption ratio decrease when powder mass
increases.
4
2
0
Fig. 3 Changes in water absorption ratio at 105 s of treated time against

dispersed volume from 0 to 40 vol% of dispersed Al powders in nylon6.
m-2
Here, E, bN ( 8 mm) and t ( 4 mm) were impact fracture

energy (J), sample width (mm) and span distance (sample
thickness, mm), respectively. The distance between supporting points was 60 mm.
Charpy impact value, auc /kJ
Water absorption ratio (mass %)
178
0.2
0.4
0.6
0.8
1
5
[10 ]
Water absorption time , t / s
Fig. 2 Changes in water absorption ratio against water absorption time in
nylon6 dispersed at each volume% of Al powders. ( ; nylon6, ;
10 vol%Al, ; 20 vol%Al, ; 30 vol%Al and ; 40 vol%Al)
3.2 Impact value improvement by water absorption

Figure 4 shows changes in Charpy impact value of
composite polymers before and after water absorption for
105 s against volume change from 0 to 40 vol% of dispersed
Al powders in nylon6. Inuences of water absorption on
Charpy impact value for composite polymers have been
investigated. Although the Al dispersion from 0 to 40 vol%Al
decreases the impact value, the water absorbing treatment
Eects of Water Absorption on Impact Value of Aluminum Dispersed Composite Nylon6
Dry sample
179
Water absorbed sample

Nylon6
2mm
Nylon6 10vol%Al
2mm
Nylon6 40vol%Al
2mm
Water
absorbed
10vol%Al
dispersed
nylon6
50 m
2mm
2mm
2mm
Water
absorbed
40vol%Al
dispersed
nylon6
50 m
Fig. 5 SEM photographs of water absorbed composite polymers (0, 10 and 40 vol%Al powders in nylon6), together with the composite
polymers before water absorption.
for 105 s at boiling point largely increases the impact values

of all composite polymers.
3.3 Change in fracture morphology by water absorption
The fracture surface of the sample has been observed by
SEM. Figure 5 shows SEM photographs of water absorbed
composite polymers (0, 10 and 40 vol%Al powders in
nylon6), together with the composite polymers before water
absorption. The peculiar undulating surface photographs are
observed on the impact fractured surface of nylon6 samples
with and without water absorption.
Furthermore, the water absorption conrmed that macroscopic local deformation with smooth undulating surface
could be observed in the nylon6.
Although the water absorbed composite samples also

exhibit smooth undulating surface rather than the dry
samples, the Al particles addition generally shows the ne
morphology on the fractured surface. Furthermore, the ultra
ne patterns with homogeneous roughness are observed on
the fractured surface of composite polymer dispersed with
40 vol%Al powders with and without water absorption.
4.
Discussion
4.1 Reinforcement induced by water absorption

The result of Fig. 3 conrms that the water absorption
ratio decreases, when the powder mass increases. Therefore,
the amount of nlyon6 roughly corresponds to amount of
M. Kanda and Y. Nishi
M(H20)/M(Nylon6+Al+H20)
180
that water molecules in nylon6 eectively relaxes the

impact force.
0.3
4.2
Inuence of water absorption on impact value of

composite nylon6
The interfacial energy () to form crack is strongly
correlated to the critical stress intensity parameter (Kc ) to
propagate the crack, as expressed by a following equation.8)
0.2
1 2 Kc 2 =2E
0.1
0
20 40 60 80 100
Fig. 6 Changes in experimental and estimated mole fraction (M(H2 O)/

M(Nylon6 + Al + H2 O)) of absorbed water in composite polymers
against dispersed volume of Al powders in nylon6. The water treatment
time is for 105 s at 373 K. ( ; experimental value, ; estimated value)
water absorption. In addition, the water absorption substantially increases the impact value, although volume fraction
of the nylon6 is small from the Fig. 4. The impact value
of water-absorbed sample of 40 vol%Al dispersed nylon6
corresponds to that of dry sample of 10 vol%Al dispersed
nylon6. Therefore, substantial increases of impact value by
water absorption and its dependence were discussed.
If the free molecules of residual water essentially cause the
impact force relaxation, the results can be explained by the
following discussion. Figure 6 shows changes in experimental and estimated mole fraction (M(H2 O)/M(6PA + Al +
H2 O)) of absorbed water in composite polymers against
dispersed volume of Al powders in nylon6, where the water
treatment time is for 105 s at 373 K. Based on the absorbed
water in pure nylon6 and simple volume fraction of Al and
nylon6, the estimated mole fraction of absorbed water in Al
dispersed composite polymer is dened as an estimated
value. Based on this results, the experimental fractions of
absorbed water of Al dispersed composite polymer is higher
than that of estimated values. The water molecules probably
diuse into the nylon6 through the inter-molecular sites. The
weak attractive inter-molecular bonding between water
molecule and nylon6 monomer probably occurs. When the
free molecules of absorbed water cause the impact force
relaxation, the results are explained.
Although the Al dispersion from 0 to 40 vol%Al decreases
the impact value, the water absorbing treatment for 105 s at
boiling point largely increases the impact values of all
composite polymers. Since the water absorption softens the
nylon,7) it is easy to deform the nylon6 sample. Thus, it can
be explained that water molecules in nylon6 mainly
decreases the impact force.
Al addition decreases the volume fraction of nylon6, as
shown in Fig. 6. Since the amount of absorbed water mainly
is contributed by the volume fraction of nylon6, Al addition
decreases the water absorption in the composites. Although
the amount of water absorption is not so large, it is enough
Here, and E are poison ratio and Young modulus,

respectively. The temperature (T=K) dependence of critical
stress intensity parameter (Kc ) related to fracture toughness
is applied to the thermal activation process and is often
expressed by a following equation.8)
Kc Ko expTo =T
Here, To is constant. As shown in Fig. 5, the macroscopic

local deformation photographs with smooth and peculiar
undulating surface are observed on the impact fractured
surface of composites with and without water absorption.
Although the Al particles addition generally nes the
morphology on the fractured surface, the water absorption
coarsens the morphology of nylon6. As shown in Fig. 4, the
water absorbed nylon6 with smooth undulating surface
exhibit the high impact values rather than that of the dry
nylon6 with peculiar undulating surface.
When the crack with stress intensity tip in nylon6
propagates at room temperature and/or with low strain rate,
the crack propagation time is often long enough to select
the fractured passes. As a result, the undulating surface in
the nylon6 exhibits the peculiar. Namely, since the crack
gradually propagates the nylon6 matrix through the fractured
passes at dierent directions, the peculiar undulation surface
can be gotten, as shown in Fig. 5.
On the other hand, if fracture with the stress intensive tip
occurs at low temperature and/or with high strain rate, the
crack propagation time is probably short to select the
fractured passes at dierent directions. Namely, since the
impact fracture is dominated by the thermal activation
process, it is not easy to turn the crack directions at low
temperature and/or with high strain rate. Thus, the fractured
area of smooth undulating surface becomes to be small (see
Fig. 5).
The larger impact value is required to propagate the crack
of water absorbed composites nylon6, as shown in Fig. 4.
Namely, the smooth undulating surface of water absorbed
composites nylon6 corresponds to their high impact value.
The water absorption probably generates the relaxation sites
induced by both free and intermolecular bonded water
molecules in nylon6. If the water absorption also enhances
the interfacial energy to form the crack in nylon6 and its
composites, the high impact value with smooth undulating
surface can be explained.
5.
Conclusion
In summary, composite polymers, which are nylon6

matrix with Al powders dispersed homogeneously, are
prepared. To evaluate the mechanical properties of nylon6
Eects of Water Absorption on Impact Value of Aluminum Dispersed Composite Nylon6
polymer with Al powders homogeneously dispersed, the

impact value has been obtained. Inuences of water
absorption on Charpy impact value for composite polymers
have been investigated.
(1) As a result, all samples conrmed water absorption and
became to be approximately saturated at 105 s of water
absorption time. It was conrmed that the water absorption
ratio decrease when powder mass increases.
(2) Although the Al powders dispersion from 0 to 40 vol%Al
decreases the impact value, the water absorbing treatment
for 105 s at boiling point largely increases the impact values
of all composite polymers.
(3) It is explained that water molecules in nylon6 mainly
relaxes the impact force.
(4) If the water absorption also enhances the interfacial
energy to form the crack in nylon6 and its composites, the
high impact value with smooth undulating surface can be
explained.
181
Acknowledgement
Authors would like to thank Mr. Keisuke Itou and TKK
Co. Ltd. for the sample preparation and his useful help.
REFERENCES
1) T. Takahashi, T. Morishita and Y. Nishi: J. Jpn. Inst. Metals 69 (2005)
759762.
2) K. Oguri, T. Takahashi, A. Kadowaki, A. Tonegawa and Y. Nishi: J. Jpn.
Inst. Metals 68 (2004) 537539.
3) M. Kanda, K. Itou and Y. Nishi: J. Jpn. Inst. Metals 71 (2007) 716720.
4) K. Itou, M. Kanda and Y. Nishi: J. Jpn. Inst. Metals 70 (2006) 940944.
5) K. Inoue, K. Iwata, T. Morishita, A. Tonegawa, M. Salvia and Y. Nishi:
J. Jpn. Inst. Metals 70 (2006) 461466.
6) K. Iwata, K. Yamada, A. Kadowaki, N. Yamaguchi, A. Tonegawa and
Y. Nishi: J. Jpn. Inst. Metals 68 (2004) 534536.
7) N. Jia, A. H. Raenkel and A. V. Kagan: J. Reinf. Plastics Comp. 23
(2004) 792737.
8) Y. Kosiga: Bull. Jpn. Inst. Metals 5 (1966) 451460.

EXPRESS REGULAR ARTICLE
Direct Dry Syntheses and Thermal Analyses of a Series

of Aluminum Complex Hydrides
T. Sato1 , K. Ikeda1 , H.-W. Li1 , H. Yukawa2 , M. Morinaga2 and S. Orimo1; *
1
Institute for Materials Research, Tohoku University, Sendai 980-8577, Japan

Department of Materials Science and Engineering, Graduate School of Engineering,
Nagoya University, Nagoya 464-8603, Japan
2
Synthesis of a series of aluminum complex hydrides, namely, LiAlH4 , Li3 AlH6 , NaAlH4 , Na3 AlH6 , Mg(AlH4 )2 , MgAlH5 , Ca(AlH4 )2 , and
CaAlH5 , was attempted by the mechanochemical milling of AlH3 and the other elemental hydrides. Aluminum complex hydrides, except
MgAlH5 , were synthesized, and their dehydriding (decomposition) properties were systematically investigated by thermogravimetry. The
dehydridng temperatures possibly showed a correlation with the geometrical distances in space between M 0 (M 00 ) and H in the crystal structures
of the aluminum complex hydrides. [doi:10.2320/matertrans.MER2008251]
Keywords: aluminum, mechanical milling, thermogravimetric analysis, X-ray diraction, hydrides
1.
Introduction
Aluminum complex hydrides M(AlH4 )n (M: typically

alkali or alkaline-earth metals; n: valence of M) have been
intensively studied as candidates for hydrogen storage
materials because of their high gravimetric hydrogen
densities, reversible reactions under moderate conditions
(temperatures and hydrogen pressures), and also their release
of hydrogen without impurity gases.111) It has been reported
that M 0 AlH4 (M 0 : alkali metals) decomposes into M 0 3 AlH6
upon heating and that one molar of hydrogen is released:11)
M 0 AlH4 ! 1/3M 0 3 AlH6 2/3Al H2 "
For example, LiAlH4 decomposes into Li3 AlH6 , which has

isolated AlH6 -units in the crystal structure.1113) On the other
hand, M 00 (AlH4 )2 (M 00 : alkaline-earth metals) has been
reported to decompose into the intermediate phase, M 00 AlH5 ,
as follows:10,11,1416)
M 00 (AlH4 )2 ! M 00 AlH5 Al 3/2H2 "
Ca(AlH4 )2 decomposes into CaAlH5 , which has cornershared AlH6 -units in the crystal structure.14,15) Corner-shared
AlH6 -units are probably a characteristic of aluminum complex hydrides containing alkaline-earth metals (Ex. BaAlH5 17)), although the structure of MgAlH5 is still unclear.1820)
Systematic studies on aluminum complex hydrides are
quite important to obtain further insights on various complex
hydrides as candidates for hydrogen storage materials. For
example, on the basis of theoretical studies, Lvvik et al.
predicted that an interaction between M 0 and H can aect the
thermal stability of M 0 AlH4 and M 0 3 AlH6 .21) These interactions were explained given the existence of a geometrical
space between M 0 and H along with the results of crystal
orbital overlap population (COOP) calculation. Experimental
data on crystal structures and thermal analyses of aluminum
complex hydrides are indispensable for conrming the results
of theoretical studies.
For performing systematic experimental studies, the synthesis of solvent-free high-quality samples of aluminum
*Corresponding
author, E-mail: orimo@imr.tohoku.ac.jp
complex hydrides are required. There are two methods for the
synthesis of hydrides: wet synthesis (metathesis and precipitation reactions in solutions)22) and dry synthesis (solid-solid
and/or solid-gas reactions).11) In the latter case, for example,
LiAlH4 has been synthesized from LiH, Al and TiCl3 ,23)
Na3 AlH6 has been synthesized from NaH, Al and TiF3 ,24,25)
and Ca(AlH4 )2 has been synthesized from CaCl2 and
NaAlH4 .26)
In the present study, for synthesizing aluminum complex
hydrides, we focused on the direct dry synthesis method,
namely, the mechanochemical milling of elemental hydrides
AlH3 and M 0 H or M 00 H2 (M 0 : Li, Na; M 00 : Mg, Ca).27,28) The
possible advantages of this simple but reliable synthesis
method are as follows: (1) elemental hydrides including AlH3
are inductile materials (in contrast to Al) that promote
mechanochemical milling and (2) AlH3 exhibits an unstable
property, the observed formation enthalpy of which is 9
12 kJ/mol AlH3 ,2931) and also promotes the formation of
more stable complex hydrides.11,32)
In the present study, rstly, a series of aluminum complex
hydrides was synthesized by the direct dry synthesis method
from elemental hydrides. Secondly, the samples were studied
to elucidate the thermal properties of the complex hydrides.
Finally, the obtained data were used to show a possible
correlation between the dehydriding (decomposition) temperatures, and the geometrical distances in space between M 0
(M 00 ) and H in the crystal structures for a series of aluminum
complex hydrides.
2.
The starting elemental hydride -AlH3 was prepared in

ether according to the chemical reactions of LiAlH4 and
AlCl3 as described in Ref. 33), and X-ray diraction peak
positions were identied as -AlH3 (Fig. 1).34) Powders of
elemental hydrides, namely, LiH (CERAC 99.99%), NaH
(Aldrich 95%), MgH2 (Alfa Aesar 98%), CaH2 (Aldrich
99.99%), and AlH3 , were mixed with chemical compositions
of LiAlH4 , Li3 AlH6 , NaAlH4 , Na3 AlH6 , Mg(AlH4 )2 ,
MgAlH5 , Ca(AlH4 )2 , and CaAlH5 . A mixture of AlH3 and
the other elemental hydrides (weighing a total of 200 mg) was
Direct Dry Syntheses and Thermal Analyses of a Series of Aluminum Complex Hydrides
183
samples were handled in an Ar- or He-gas-lled glove box

with a dew point below 183 K and with less than 1 ppm of O2
in order to avoid (hydro-)oxidation.
3.
Fig. 1 X-ray diraction patterns of observed (top) and simulated (bottom)

-AlH3 .33)
mechanochemically milled by using Fritsch P7 at 400 rpm at

a hydrogen gas pressure of 0.3 MPa for up to 20 h, milling
times of 5 min being alternated with 5 min of rest.
All the samples were investigated by using an X-ray
diractometer (PANalytical XPERT, with Cu-K radiation). Simulated X-ray diraction patterns were generated by
PowderCell 1.0.35) To investigate the dehydriding (decomposition) temperatures and the amounts of released hydrogen,
the samples were then studied by thermogravimetry (TG,
Rigaku TG-8210, with an Al sample holder) and heated up to
673 K at 5 K/min in a He gas ow of 150 ml/min. All the
Mechanochemical milling of only -AlH3 accelerates the

dehydriding (decomposition) reaction and forms elemental
Al in less than 1 h, especially at the beginning stage of the
milling.30) Interestingly, the successful milling of -AlH3
together with the other elemental hydrides, however, results
in the synthesis of aluminum complex hydrides, as explained
below. It is assumed that the other elemental hydrides that are
attached to the surface of -AlH3 prevent the dehydriding
(decomposition) reaction of -AlH3 , although the detailed
mechanism is still under investigation.
LiAlH4 and Li3 AlH6 were synthesized by the mechanochemical milling of LiH and -AlH3 for 5 h. The X-ray
diraction patterns in Fig. 2 are in good agreement with the
simulated patterns. By the same synthesis procedures (the
milling time was optimized for each hydride), NaAlH4
(milling time: 5 h), Na3 AlH6 (10 h), Mg(AlH4 )2 (10 h),
Ca(AlH4 )2 (10 h), and CaAlH5 (5.5 h) were also identied
from X-ray diraction patterns (Fig. 2). Although MgAlH5
has been reported to be formed from Mg(AlH4 )2 as an
intermediate phase,18,19) no evidence of its formation was
obtained under the present synthesis conditions. The X-ray
diraction patterns of the samples (5 and 20 h) corresponded
to Mg(AlH4 )2 , MgH2 , -AlH3 , and Al peaks (Fig. 2).
Fig. 2 Observed X-ray diraction patterns of LiAlH4 , Li3 AlH6 , NaAlH4 , Na3 AlH6 , Mg(AlH4 )2 , Ca(AlH4 )2 and CaAlH5 (MgAlH5 was
not actually formed under the present syntheses condition). Simulated X-ray diraction patterns of LiAlH4 ,36) Li3 AlH6 ,13) NaAlH4 ,37)
Na3 AlH6 ,38) Mg(AlH4 )2 ,39,40) MgAlH5 ,20) Ca(AlH4 )2 41) and CaAlH5 15) locate under the each observed patterns.
184
T. Sato et al.
Table 1 Calculated unit cell parameters. The reported unit cell parameters were obtained from deuterided (D)/hydrided (H) samples with
structural determination (Rietveld renement) results by powder neutron (N)/X-ray (X) diraction. Ca(AlH4 )2 has not been examined by
Rietveld renement, but calculated unit cell parameters by computational study are given in italics in parentheses. Unit cell parameters of
Na3 AlH6 could not be obtained because of the small number of observed X-ray diraction peaks.
References
Present work
LiAlH4
a
b
0.4830(6) nm
0.7839(7) nm
0.48254(1) nm
0.78040(1) nm
0.7933(6) nm
0.78968(1) nm

Li3 AlH6
NaAlH4
Na3 AlH6
Mg(AlH4 )2
Ca(AlH4 )2
CaAlH5
Unit cell parameters
112.2(1)
0.8113(5) nm
0.807117(10) nm
0.957(1) nm
0.95130(2) nm
0.5023(1) nm
0.50119(1) nm
1.1365(3) nm
1.13147(4) nm
0.5390(2) nm
0.5514(2) nm
0.7725(3) nm
0.5196(2) nm
0.51949(2) nm
0.5845(4) nm
0.58537(2) nm
1.351(4) nm
1.3486 nm (1.337 nm)
0.956(4) nm
0.9532 nm (0.928 nm)
0.907(3) nm
0.9053 nm (0.9053 nm)
0.839(1) nm
0.83797(9) nm
0.690(2) nm
0.69293(8) nm
0.981(2) nm
0.98138(11) nm
(5:3 mass%)
(2:7 mass%)
(2:7 mass%)
3a
3b
3c
As shown in Fig. 3(a), LiAlH4 exhibited two-step mass

changes from hydrogen release with amounts of 4.7 mass%
(onset temperature of 426 K) and 2.3 mass% (459 K) that
correspond to eqs. (3a) and (3b), respectively. Furthermore,
Li3 AlH6 is also assumed to obey the following reactions:
(5:6 mass%) 3d
(5:6 mass%) 3e
In fact, we can conrm that 4.9 mass% of hydrogen was

released from Li3 AlH6 (476 K), corresponding to eq. (3d)
(also shown in Fig. 3(a)).
Next, NaAlH4 has been reported to decompose and release
hydrogen according to the following reactions:13,57,11,32)
NaAlH4
! 1/3Na3 AlH6 2/3Al H2 "
N/X
295
[36]
N/X
295
[13]
295
[37]
295
[38]
N/X
295
[39,40]
298
[15,41]
298
[15]
93.78(1)
The unit cell parameters of all the aluminum complex

hydrides, except MgAlH5 , were calculated by TREOR42) and
PIRUM43) and are summarized in Table 1.
Dehydriding properties of thus synthesized M 0 AlH4 ,
0
M 3 AlH6 , M 00 (AlH4 )2 and M 00 AlH5 were systematically
investigated by thermogravimetry (TG), as shown in Fig. 3.
Upon heating, LiAlH4 has been reported to decompose and
release hydrogen according to the following reactions:11,12)
Li3 AlH6 ! 3LiH Al 3/2H2 "

! Li-Al alloy 3/2H2 "
Temp. (K)
89.86(3)
LiAlH4
! 1/3Li3 AlH6 2/3Al H2 "
! LiH Al 1/2H2 "
! LiAl 1/2H2 "
N/X
112.268(1)
94.2(2)
Measured
D/H
(3:7 mass%)
4a
! NaH Al 1/2H2 "
(1:9 mass%)
4b
! Na Al 1/2H2 "
(1:9 mass%)
4c
In the case of Na3 AlH6 , the reactions are as follows:

Na3 AlH6 ! 3NaH Al 3/2H2 "
! 3Na Al 3/2H2 "
(3:0 mass%) 4d
(3:0 mass%) 4e
Na3 AlH6 exhibited two-step hydrogen release with amounts

of 2.7 mass% (507 K) and 2.9 mass% (610 K) that correspond
to eqs. (4d) and (4e), respectively (Fig. 3(b)).
Furthermore, Mg(AlH4 )2 has been reported to decompose
and release hydrogen according to the following reactions:10,11,16)
Mg(AlH4 )2 ! MgH2 2Al 3H2 " (7:0 mass%) 5a
(2:3 mass%) 5b
! Mg 2Al H2 "
As shown in Fig. 3(c), Mg(AlH4 )2 exhibited two-step hydrogen releases with amounts of 5.5 mass% (410 K) and
1.9 mass% (534 K) that correspond to eqs. (5a) and (5b),
respectively.
Finally, the reactions for Ca(AlH4 )2 are suggested to be as
follows:14,15)
Ca(AlH4 )2
! CaAlH5 Al 3/2H2 " (3:0 mass%)
6a
(3:0 mass%) 6b
! CaH2 2Al 3/2H2 "
(2:0 mass%)
6c
! Ca-Al alloy H2 "
According to Fig. 3(d), Ca(AlH4 )2 released 2.2 mass%
(415 K) and 2.2 mass% (499 K) of hydrogen by a two-step
Direct Dry Syntheses and Thermal Analyses of a Series of Aluminum Complex Hydrides
185
Fig. 3 Thermogravimetry (TG) curve of (a) LiAlH4 (red) and Li3 AlH6 (blue), (b) Na3 AlH6 , (c) Mg(AlH4 )2 , and (d) Ca(AlH4 )2 (red) and
CaAlH5 (blue). Numbers indicate the mass reductions that the theoretical mass reductions are given in parentheses. In order to avoid any
reactions between the samples and the sample holder, the measurements of LiAlH4 and Li3 AlH6 were terminated at 573 K. NaAlH4 could
not be measured by TG.
reaction, which correspond to eqs. (6a) and (6b). On the other

hand, compared to the reported one-step reaction of (6c),
hydrogen was released by a two-step reaction, i.e., 0.4 mass%
(561 K) and 0.2 mass% (618 K). A similar phenomenon also
appears in the dehydriding reaction of the synthesized
CaAlH5 , although it is predicted to release hydrogen
according to the following reactions:
CaAlH5 ! CaH2 Al 3/2H2 " (4:2 mass%)
(2:8 mass%)
! Ca-Al alloy H2 "
6d
6e
The dehydriding temperatures and the amounts of hydrogen release for all the samples are basically comparable to
the previously reported dehydriding reactions. The results
indicate that the direct solid reaction of MHn and AlH3 by
using mechanochemical milling is a new and reliable
synthesis process of the aluminum complex hydrides. The
optimization of the synthesis condition is under investigation
to obtain higher quality aluminum complex hydrides with
dierent thermodynamic stabilities.
It is generally known that in the aluminum complex
hydrides, hydrogen covalently bonds to Al in complex
anions, [AlH4 ] or [AlH6 ]3 , which form an ionic bonding
with M n . On the other hand, a recent theoretical study
proposed that there is a possible weak M-H bonding,
the strength of which depends on the distance in space
between M and H.21) In other words, a shorter M-H distance
weakens the Al-H bonding and leads to a lower dehydriding

temperature. Therefore, the onset temperatures of the
dehydriding reaction of aluminum complex hydrides are
plotted as a function of the distances in space between M 0
(M 00 ) and H.13,15,3641) In this study, in order to estimate the
distances, averaged M 0 (M 00 )-H distances were subtracted
from total distances of the ionic radius of H (0.110 nm21)) and
that of M 0 (M 00 ). The result, shown in Fig. 4, roughly
indicates that the shorter the distance in space between
M 0 (M 00 ) and H, the stronger the interaction between them.
This induces breakage of the Al-H bonds and lowers the
dehydriding (decomposition) temperatures of the aluminum
complex hydrides.
Detailed studies on the dehydriding reactions of the
aluminum complex hydrides including Ca(AlH4 )2 and
CaAlH5 , and on the correlation between dehydriding (decomposition) temperature and distance are required to obtain
further insights on various complex hydrides for potential use
as hydrogen storage materials.
4.
Conclusions
The direct dry synthesis method was employed to obtain

the aluminum complex hydrides from elemental hydrides.
The hydrides, except MgAlH5 , were successfully synthesized
and characterized by using a powder X-ray diractometer.
186
T. Sato et al.
Fig. 4 Correlation between dehydrding temperature compared with the

distances in space between M 0 (M 00 ) and H. A six-coordinated ionic radius
of Li was used for LiAlH4 because ve-coordinated ionic radius has not
been reported. The dehydriding temperature of NaAlH4 (483 K) was
obtained from Ref. 44). Na3 AlH6 has two atomic positions for Na where
one Na atom has eight-coordinations (0.219 nm on distance in space
between Na and H) and the other has six-coordinations (0.133 nm).
Li3 AlH6 and CaAlH5 each have two plots that are Li3 AlH6 (459 K)/
CaAlH5 (499 K) decomposed from LiAlH4 /Ca(AlH4 )2 , and directly
synthesized Li3 AlH6 (476 K)/CaAlH5 (530 K). BaAlH5 was plotted on
the basis of information provided in Ref. 45).
Moreover, the dehydriding (decomposition) temperatures

and the amounts of released hydrogen upon heating were
investigated by TG. By the comparison study, the possible
existence of a correlation between the dehydriding temperatures and geometrical distances in space between M 0 (M 00 )
and H in the crystal structures of a series of aluminum
complex hydrides was indicated.
Acknowledgements
This work was partially supported by Grant-Aid for
Science Research (17106008) from the Ministry of Education, Culture, Sports, Science and Technology of Japan, by
the Japan Society for the Promotion of Science; and by the
Global COE Program Materials Integration, Tohoku University, MEXT, Japan. Synthesis of AlH3 was done in
collaboration with Dr. K. Hashi, Dr. H. Ito, and Dr. T.
Kabutomori at Japan Steel Works, Ltd., and Prof. C.M.
Jensen at the University of Hawaii.
REFERENCES
1) B. Bogdanovic and M. Schwickardi: J. Alloy. Compd. 253254 (1997)
19.
2) C. M. Jensen and K. J. Gross: Appl. Phys. A 72 (2001) 213219.
3) B. Bogdanovic and M. Schwickardi: Appl. Phys. A 72 (2001) 221223.
4) J. Chen, N. Kuriyama, Q. Xu, H. T. Takeshita and T. Sakai: J. Phys.
Chem. B 105 (2001) 1121411220.
5) D. Sun, T. Kiyobayashi, H. T. Takeshita, N. Kuriyama and C. M.
Jensen: J. Alloy. Compd. 337 (2002) L811.
6) S. S. Srinivasan, H. W. Brinks, B. C. Hauback, D. Sun and C. M.
Jensen: J. Alloy. Compd. 377 (2004) 283289.
7) B. Bogdanovic, U. Eberle, M. Felderho and F. Schuth: Scr. Mater. 56
(2007) 813816.
8) J. Graetz and J. Reilly: Scr. Mater. 56 (2007) 835839.
9) N. Eigen, C. Keller, M. Dornheim, T. Klassen and R. Bormann: Scr.

Mater. 56 (2007) 847851.
10) K. Komiya, N. Morisaku, Y. Shinzato, K. Ikeda, S. Orimo, Y. Ohki, K.
Tatsumi, H. Yukawa and M. Morinaga: J. Alloy. Compd. 446447
(2007) 237241.
11) S. Orimo, Y. Nakamori, J. R. Eliseo, A. Zuttel and C. M. Jensen: Chem.
Rev. 107 (2007) 41114132.
12) H. W. Brinks, B. C. Hauback, P. Norby and H. Fjellvag: J. Alloy.
Compd. 351 (2003) 222227.
13) H. W. Brinks and B. C. Hauback: J. Alloy. Compd. 354 (2003) 143
147.
14) M. Mamatha, C. Weidenthaler, A. Pommerin, M. Felderho and F.
Schuth: J. Alloy. Compd. 416 (2006) 303314.
15) C. Weidenthaler, T. J. Frankcombe and M. Felderho: Inorg. Chem. 45
(2006) 38493851.
16) M. Fichtner, O. Fuhr and O. Kircher: J. Alloy. Compd. 356357 (2003)
418422.
17) Q. A. Zhang, Y. Nakamura, K. Oikawa, T. Kamiyama and E. Akiba:
Inorg. Chem. 41 (2002) 69416943.
18) P. Claudy, B. Bonnetot and J. M. Letoe: J. Thermal. Anal. 15 (1979)
119128.
19) T. N. Dymova, N. N. Maltseva, V. N. Konoplev, A. I. Golovanova,
D. P. Aleksandrov and A. S. Sizareva: Russ. J. Coord. Chem. 29 (2003)
385389.
20) A. Klaveness, P. Vajeeston, P. Ravindran, H. Fjellvag and A. Kjekshus:
Phys. Rev. B 73 (2006) 094112.
21) O. M. Lvvik, O. Swang and S. M. Opalka: J. Mater. Res. 20 (2005)
31993213.
22) R. H. Pawelke, M. Felderho, C. Weidenthaler and F. Schuth: Scr.
Mater. 59 (2008) 515517.
23) Y. Kojima, Y. Kawai, T. Haga, M. Matsumoto and A. Koiwai: J. Alloy.
Compd. 441 (2007) 189191.
24) P. Wang and C. M. Jensen: J. Phys. Chem. B 108 (2004) 1582715829.
25) P. Wang, D. X. Kang and H. M. Cheng: Appl. Phys. Lett. 87 (2005)
071911.
26) M. Mamatha, B. Bogdanovic, M. Felderho, A. Pommerin, W.
Schmidt, F. Schuth and C. Weidenthaler: J. Alloy. Compd. 407 (2006)
7886.
27) S. Chaudhuri, J. Graetz, A. Ignatov, J. J. Reilly and J. T. Mucherman:
J. Am. Chem. Soc. 128 (2006) 1140411415.
28) H. Kabbour, C. C. Ahn, S. J. Hwang, R. C. Bowman Jr. and J. Greatz:
J. Alloy. Compd. 446447 (2007) 264266.
29) G. C. Sinke, L. C. Walker, F. L. Oetting and D. R. Stull: J. Chem. Phys.
47 (1967) 27592761.
30) S. Orimo, Y. Nakamori, T. Kato, C. Brown and C. M. Jensen: Appl.
Phys. A 83 (2006) 58.
31) J. Graetz and J. Reilly: J. Alloy. Compd. 424 (2006) 262265.
32) W. Grochala and P. P. Edwards: Chem. Rev. 104 (2004) 12831316.
33) F. M. Brower, N. E. Matzek, P. F. Reigler, H. W. Rinn, C. B. Roberts,
D. L. Schmidt, J. A. Snover and K. Terada: J. Am. Chem. Soc. 98
(1976) 24502453.
34) J. W. Turley and H. W. Rinn: Inorg. Chem. 8 (1969) 1822.
35) W. Kraus and G. Nolze: J. Appl. Cryst. 29 (1996) 301303.
36) B. C. Hauback, H. W. Brinks and H. Fjellvag: J. Alloy. Compd. 346
(2002) 184189.
37) B. C. Hauback, H. W. Brinks, C. M. Jensen, K. Murphy and A. J.
Maeland: J. Alloy. Compd. 358 (2003) 142145.
38) E. Ronnebro, D. Noreus, K. Kadir, A. Reiser and B. Bogdanovic:
J. Alloy. Compd. 299 (2000) 101106.
39) A. Fossdal, H. W. Brinks, M. Fichtner and B. C. Hauback: J. Alloy.
Compd. 387 (2005) 4751.
40) M. Fichtner, J. Engel, O. Fuhr, A. Gloss, O. Rubner and R. Ahlrichs:
Inorg. Chem. 42 (2003) 70607066.
41) O. M. Lvvik: Phys. Rev. B 71 (2005) 144111.
42) P. E. Werner, L. Eriksson and W. Westdahl: J. Appl. Cryst. 18 (1985)
367370.
43) P. E. Werner: Ark. Kemi. 31 (1969) 513516.
44) M. P. Balogh, G. G. Tibbetts, F. E. Pinkerton, G. P. Meisner and C. H.
Olk: J. Alloy. Compd. 350 (2003) 136144.
45) Q. A. Zhang, Y. Nakamura, K. Oikawa, T. Kamiyama and E. Akiba:
J. Alloy. Compd. 361 (2003) 180186.

Synthesis of Ultra Fine Titanium-Tungsten Carbide Powder

from Titanium Dioxide and Ammonium Metatungstate
Gil-Geun Lee1 and Gook-Hyun Ha2
1
Division of Materials Science and Engineering, Pukyong National University,

San 100, Yongdang-Dong, Nam-Gu, 608-739, Korea
2
Korea Institute of Materials Science, 531 Changwondaero, Changwon, Kyungnam, 641-831, Korea
In the present study, the focus is on the synthesis of ultra ne titanium-tungsten carbide powder by the carbothermal reduction process. The
starting powder was prepared by the combination of drying and calcination methods using titanium dioxide powder and an aqueous solution of
ammonium metatungstate to obtain a target composition of 60 mass%TiC-40 mass%WC, i.e., (Ti0:83 W0:17 )C. The synthesized oxide powder was
mixed with carbon black, and this mixture was then heat-treated under a owing argon atmosphere. The changes in the phase structure, thermal
gravity and particle size of the mixture during heat treatment were analyzed using XRD, TG-DTA and SEM. The synthesized oxide powder has a
mixed phase structure of anatase-TiO2 and WO3 phases. This composite oxide powder was carbothermally reduced to titanium-tungsten carbide
by solid carbon through three steps with increasing temperature; the reduction of WO3 , the reduction of TiO2 and formation of tungsten carbides,
and the formation of titanium-tungsten carbide. The synthesized (Ti1x Wx )Cy powder at 1673 K for 10.8 ks has an average particle size of about
200 nm. [doi:10.2320/matertrans.MER2008271]
Keywords: titanium-tungsten carbide, hard metals, carbothermal reduction, ultra ne powder
1.
Introduction
Hard metal based on WC/Co has been widely used for

wear-resistant machine parts or tool materials because it has
superior mechanical properties, such as wear resistance and
high temperature strength, etc.1) In order to increase the
mechanical properties and resistance to cratering at
elevated temperatures, TiC is added to WC/Co hard metal.
WC/TiC/Co system hard metals were devised for machining
steels and other ferrous alloys, the purpose of the TiC content
being to resist the high-temperature diusive attack that
causes chemical breakdown and cratering. WC/TiC/Co
system hard metals have two distinct carbide phases, angular
crystals of almost pure WC and rounded TiC/WC mixed
crystals.24) Because of dierential diusion eects, the
latter often show a cored structure. The third phase is the
cobalt binder. A solid solution or mixed crystal of WC in
TiC (TiC is almost insoluble in WC) retains the anti-cratering
property to a great extent.14)
The mechanical properties of the hard metals depend not
only on the chemical composition but also on their microstructure.1) The hardness, fracture toughness and wear
properties of the hard metals were strongly inuenced by
the size of the carbide particles, the degree of homogeneity of
the microstructure and purity of the initial powders. These
mechanical properties increased simultaneously with decreasing particle size and increasing homogeneity. To
manufacture high performance hard metal with an ultra ne
microstructure, raw powder materials with an ultra ne
particle size should be used. Recently, several methods have
been proposed for the synthesis of ultra ne particles.58) In
the case of WC and TiC, it was possible to synthesize the
WC/Co and TiC/Co composite powders with an average
particle size under 100 nm and 200 nm, respectively.8,9)
These composite powders were manufactured by the combination of spray conversion and carburization processes. For
the other carbides (TaC, NbC and (Ti,W)C, etc.), however,

no commercial process has been established for the synthesis
of such ultra ne particles.
Titanium-tungsten carbide, (Ti1x Wx )Cy , has been used as
a dispersed particle in the WC/TiC/Co system hard metals
for increasing high temperature hardness and decreasing
the reaction with steel alloy during machining. Generally,
titanium-tungsten carbide was commercially manufactured
by carbothermal reduction using metallic tungsten powder,
titanium dioxide powder and carbon black as the raw
materials.1) This process requires a temperature range of
20732473 K for the commercial production of the titaniumtungsten carbide with a solid solution or mixed-crystal,
because of the diculty of reduction of titanium dioxide.
Therefore, it is dicult to produce ne particles using this
process due to the agglomeration of the particles at the high
synthesis temperature. Recently, a new mechano-chemical
process was proposed for the synthesis of ultra ne WC/TiC/
Co composite powder.10) This process could be used to
synthesize ultra ne WC/TiC/Co composite powder at about
1673 K by the combination of spray conversion and direct
carburization processes using a metallic salts solution as the
raw material. However, this process has an environmental
pollution problem due to the chemical processing based on
the spray conversion of the metallic salts, especially titanium
chloride. The titanium chloride, TiCl3 , dissolved into
titanium and chloride ions in distilled water during preparation of the starting solution for spray conversion. The
chlorine ions in the solution were transformed into chlorine
gas during spray conversion. It become generally known that
this chlorine gas causes serious environmental pollution
problems. Recently, we proposed a new carbothermal
reduction process for the synthesis of ultra ne TiC and
TiC/Co powders.9,11,12) Because this new process used ultra
ne titanium dioxide powder instead of chloride as the
titanium source, it could possibly decrease the environmental
188
G.-G. Lee and G.-H. Ha
pollution problem. Also, it decreased the reduction temperature of the titanium dioxide and titanium carbide formation
temperature compared to the conventional processes.
In the present study, the focus is the synthesis of titaniumtungsten carbide powder by the carbothermal reduction
process to investigate the synthesis possibilities of ultra ne
titanium-tungsten carbide using titanium dioxide powder as
the titanium source without the chemical processing based on
the spray conversion of the chloride solution.
2.
The starting slurry was prepared by suspending the

titanium dioxide powder (an average particle size of 50 nm
and an anatase structure) in an aqueous solution of
ammonium metatungstate (AMT, (NH4 )6 (H2 W12 O40 )4H2 O),
to obtain a target composition of 60 mass%TiC-40
mass%WC, i.e., (Ti0:83 W0:17 )C. Drying was performed using
an ultrasonic generator with a frequency of 20 kHz at about
358 K. The precursor powder obtained by the drying was
decomposed at 1073 K for 7.2 ks in air to remove the salt
components and form a titanium-tungsten-oxygen-based
composite oxide powder. The calcined composite oxide
powder was mixed with carbon black (mean particle size:
0.5 mm) for 86.4 ks using a tumbler-ball mill with a ball-topowder mass ratio of 10 : 1. The mixture was placed in a
graphite crucible and then heated in the temperature range
from 823 K to 1673 K for 1.8 ks in a tube furnace under a
owing stream of argon atmosphere. After heat treatment, the
samples were analyzed by X-ray diraction using Cu-K
radiation and observed by SEM. The changes in the thermal
gravity of this powder mixture during heat treatment from
room temperature to 1673 K in owing argon was recorded
by TG-DTA at a heating rate of 8:3 102 K/s.
3.
Figure 1 shows the X-ray diraction patterns of the dried

precursor powder and calcined powder. The diraction peak
of the precursor powder prepared by the drying method
shows only the anatase-TiO2 phase. There are no visible X-
Intensity (arb.units)
TiO2 (anatase)
WO3
Fig. 2 SEM micrograph of ball-milled mixture of calcined composite

oxide powder and carbon black.
ray diraction peaks related to tungsten. This means that the

dried powder from the starting slurry has a homogeneously
mixed state of the titanium dioxide, tungsten-based salt and
moisture. However, the calcined powder shows diraction
peaks of anatase-TiO2 and WO3 . By heating the precursor
powder in air atmosphere, the volatile components (such as
NHx , NOx and H2 O) were evaporated by decomposition of
the tungsten-based salt. The anatase-TiO2 phase was less
stable than the rutile-phase-structured one.13) In general,
titanium dioxide of the metastable anatase phase transforms
into the stable rutile phase during heating. In this study, the
phase transformation of the titanium dioxide was not found
during calcination at 1073 K for 7.2 ks. No chemical reaction
occurred between the titanium dioxide and the tungsten
component during calcination.
Figure 2 shows an SEM micrograph of the ball-milled
mixture of the calcined composite oxide powder and carbon
black. The synthesized starting composite oxide powder from
the drying had an irregular morphology with an average
particle size of about 2 mm. After milling, the particle
morphology changed to a globular shape, and the average
particle size decreased to below about 0.5 mm. The composite
oxide particles were homogeneously mixed with carbon
black, and there were no visible dierences in the X-ray
diraction patterns of the powder before and after the ballmilling.
Figure 3 shows the mass change in the mixture of the
calcined composite oxide powder and carbon black as a
function of the temperature during heating in the owing
argon gas. The theoretical quantities of the carbon for the
carbothermal reduction of the calcined composite oxide
powder are estimated by:
WO3s TiO2s 6Cs ! (Ti,W)Cs 5COg
dried precursor powder
calcined powder
20
30
40
50
60
70
80
2
Fig. 1 X-ray diraction patterns of the dried and calcined powders.
In this study, 10% excess carbon was used for the reaction.
The theoretical mass fraction of the reaction was 58.83%
after complete carbothermal conversion of the titaniumtungsten-oxygen-based composite oxide powder of 60
mass%TiC-40 mass%WC target composition by the reaction.
As shown in Fig. 3, the value of the mass fraction of the
mixture at 1673 K was higher than 58.83%, which means that
the present titanium-tungsten-oxygen-based oxide in the
mixture could not be completely converted to titaniumtungsten carbide during the analysis by TG-DTA. The mass
Synthesis of Ultra Fine Titanium-Tungsten Carbide Powder from Titanium Dioxide and Ammonium Metatungstate
(102)
80
70
50
300
500
700
900
1100
1300
1500
1700
Temperature, T / K
Fig. 3 Change in the mass fraction of the mixture of the calcined
composite oxide powder and carbon black versus temperature.
of the mixture powder started to decrease around 1123 K and

showed a rst remarkable decrease in the mass fraction at
about 1203 K. A second signicant mass loss occurred at
about 1273 K, and the mass fraction continuously decreased
with increasing temperature until the heat-treatment temperature reached about 1513 K. These mass losses mean that the
carbothermal reactions remarkably occur at these temperatures.
Figure 4 shows the X-ray diraction patterns of the
mixture of the calcined composite oxide powder and carbon
black heat-treated in the tube furnace under a owing stream
of argon at a specied temperature for 1.8 ks. The heattreatment temperature was determined from the TG curve
(Fig. 3). Below 1073 K, the X-ray diraction patterns
showed anantase-TiO2 and WO3 phases as the calcined
composite oxide powder. The diraction pattern changed to
the mixed state of the anatase-TiO2 , rutile-TiO2 and W18 O49
phases at 1123 K. At 1223 K, there were two kinds of phases
in the heat-treated powder; i.e., the rutile-TiO2 and W. This
means that the titanium dioxide had undergone phase
transformation from the metastable anatase phase structure
to the stable rutile phase structure above 1073 K. The WO3 in
the calcined powder was preferentially reduced to tungsten
below 1223 K. Therefore, this indicates that the rst
remarkable decrease in the mass fraction at about 1203 K is
the preferential carbothermal reduction of the WO3 in the
calcined composite oxide powder by the solid carbon. The Xray diraction pattern of the heat-treated powder at 1273 K
showed the coexistence of W2 C, W and magneli phases. The
magneli phase is the lower oxide in the Ti-O system, which is
written as Tin O2n1 (n > 3).4) This means that the carburization of the tungsten and reduction of the titanium dioxide
simultaneously occurred at about 1273 K. Therefore, the
second signicant decrease in the mass fraction at about
1273 K indicates the reduction of the titanium dioxide in the
calcined composite oxide powder. At 1473 K, there were four
kinds of phases in the diraction pattern: Ti3 O5 , W2 C, W and
Ti2 O3 . The X-ray diraction pattern of the heat-treated
powder at 1573 K showed W2 C, W, Ti2 O3 and (Ti1x Wx )Cy .
The titanium-tungsten carbide, (Ti1x Wx )Cy , is a solid
solution of TiC and WC with a maximum solubility of about
72 mass%WC.14) In the case of the carbothermal reduction of
pure titanium dioxide powder by solid carbon, the titanium
dioxide was carbothermally reduced to TiC through three
1673K
Intensity (arb. units)
1513
60
(222)
(111)
1273
TiO2 (anatase)
TiO2(rutile)
magneli Ti2O3
Ti3O5
W
(Ti1-xWx)Cy
(311)
90
Mass Fraction (%)
WO3
W18O49
W2C
1123
(220)
1203
(200)
100
189
1573K
1473K
1273K
1223K
1123K
1073K
20
30
40
50
60
70
80
2
Fig. 4 X-ray diraction patterns of the mixture of the calcined composite
oxide powder and carbon black heat treated in the tube furnace under a
owing stream of argon at a specied temperature for 1.8 ks.
steps; i.e., reduction of TiO2 to form Tin O2n1 (n > 3)

phases, formation of titanium oxycarbide and deoxidization
of the titanium oxycarbide to titanium carbide.12,15) The
titanium oxycarbide, TiCx Oy , is a solid solution of TiC and
TiO.16,17) Generally, the initially formed carbide from the
titanium dioxide by the solid carbon is not pure titanium
carbide. Thermodynamic calculation predicts that the initial
composition of TiCx Oy is dependent on the temperature and
partial pressure of CO and will be between TiC0:6 O0:4 and
TiC0:7 O0:3 .17) However, no diraction peaks related to the
titanium oxycarbide are found in the Fig. 4. At 1673 K, there
were two kinds of phases in the heat-treated powder; i.e., the
(Ti1x Wx )Cy and a small amount of W2 C. No dierent oxide
phases were found in the cases of heat-treated powders below
1573 K. The diraction peaks related only to tungsten almost
disappeared in the diraction pattern compared to other
X-ray diraction patterns below 1573 K. However, the
diraction peaks related to titanium-tungsten carbide more
clearly appeared in the diraction pattern. This means that
the tungsten carbide and/or tungsten dissolved into the
carburized titanium phase.
No chemical reaction occurred between the titanium and
tungsten components during reduction of the oxide phases in
the calcined composite oxide powder. It was considered that
the reduction reactions of the titanium oxides and tungsten
oxides in the titanium-tungsten-oxygen-based composite
oxide powder by the solid carbon occurred independently.
G.-G. Lee and G.-H. Ha
Relative Intensity of XRD (W2C / (Ti1-xWx )Cy )
190
1
1573K
1673K
0.1
0.01
0.001
0
0
10
15
20
25
Heat Treatment Time, t / ks
Fig. 5 Change in the relative intensity of the XRD peaks of tungsten

carbide to (Ti1x Wx )Cy versus heat-treatment time at 1573 K and 1673 K.
The carburization reaction of the tungsten component

occurred at lower temperature than that of the titanium
component. The dissolving reaction of the tungsten carbide in
the titanium component simultaneously occurred with the
carburization reaction of the titanium component. Contrary to
the titanium-oxygen-based oxide powder, no titanium oxycarbide, TiCx Oy , was found. It was considered that the
carburization of the titanium component and the formation of
(Ti1x Wx )Cy simultaneously occurred above 1473 K. For
verication of the reaction mechanism during heat treatment,
however, fundamental research is needed concerning the
crystal structure and chemical composition in the future.
Figure 5 shows the change in the relative intensity of the
XRD peaks of the W2 C to that of (Ti1x Wx )Cy in the samples
heated at 1573 K and 1673 K. The peak intensities for the
W2 C and (Ti1x Wx )Cy corresponding to the (102) and (111)
reections, respectively, are used. As shown in Fig. 5, the
amount of the retained W2 C decreased with the increasing
heat treatment time. In spite of 21.6 ks heat treatment, a small
amount of W2 C still remained in the case of the 1573 K
sample. At 1673 K, however, no diraction peak related to
the W2 C phase was found above the heat-treatment time of
10.8 ks. The heat-treatment time needed for complete
dissolution of the retained W2 C to (Ti1x Wx )Cy for the nal
(Ti0:83 W0:17 )C composition was at least 10.8 ks.
Figure 6 shows SEM micrographs of synthesized
(Ti1x Wx )Cy powders at 1673 K for 10.8 ks and 21.6 ks. No
dierence was found in particle size, morphology and degree
of agglomeration with the heat-treatment time. The mean
particle size was less than 200 nm and agglomeration degree
was very low. It was known that ultra ne (Ti1x Wx )Cy
powder can be synthesized by the carbothermal reduction
process using titanium dioxide powder and an ammonium
metatungstate. For verication of the crystal structure of
the synthesized particles, however, fundamental research is
needed concerning the lattice constant and chemical composition in the future.
4.
Conclusion
The present study focused on the synthesis of titaniumtungsten carbide powder by a carbothermal reduction process
Fig. 6 SEM micrographs of synthesized (Ti1x Wx )Cy powder at 1673 K

for (a) 10.8 ks and (b) 21.6 ks by the carbothermal reduction using
titanium-tungsten-oxygen-based composite oxide powder.
using a titanium-tungsten-oxygen-based composite oxide

powder. This composite oxide powder with a target composition of 60 mass%TiC-40 mass%WC (or (Ti0:83 W0:17 )C) can
be synthesized by the combination of drying and calcination
processes using titanium dioxide powder and an aqueous
solution of ammonium metatungstate. The synthesized
composite oxide powder has a mixed phase structure of
anatase-TiO2 and WO3 phases with an average particle size
of about 0.5 mm. This composite oxide powder is carbothermally reduced to the titanium-tungsten carbide through
three steps with increasing temperature; i.e., the reduction
of WO3 , the reduction of TiO2 and formation of tungsten
carbides, and the formation of titanium-tungsten carbide.
The reduction reactions of the titanium and tungsten
components in the composite oxide powder by the solid
carbon occurred independently. The carburization of the
tungsten component occurred at a lower temperature than
that of the titanium component. The formation of the
titanium-tungsten carbide occurred by the simultaneous
reaction of the carburization of the titanium component and
the dissolution of tungsten carbide. The synthesized
(Ti1x Wx )Cy powder at 1673 K for 10.8 ks has an average
particle size of about 200 nm.
REFERENCES
1) K. J. A. Brookes: Word Directory and Handbook of Hardmetals and
Hard Materials, 6th ed., (International Carbide Data, Marsh Barton,
1996) pp. 9102.
2) ASM International Handbook Committee: ASM Handbook, Vol. 7,
Powder Metal Technologies and Applications ed. by ASM International Handbook Committee, (ASM International, Materials Park,
Synthesis of Ultra Fine Titanium-Tungsten Carbide Powder from Titanium Dioxide and Ammonium Metatungstate
1998) pp. 932940.
3) M. Sherif El-Eskandarany: J. Alloy. Compd. 305 (2000) 225238.
4) H. Presis, L. M. Berger and D. Schultze: J. Europ. Ceram. Soc. 19
(1999) 195206.
5) G. G. Lee, G. H. Ha and B. K. Kim: Powder Metall. 43 (2000) 7982.
6) G. G. Lee and W. Y. Kim: Metals and Materials International 11 (2005)
177181.
7) K. E. Gonsalves, S. P. Rangarajan and J. Wang: Handbook of
Nanostructured Materials and Nanotechnology ed. by H. S. Nalwa,
(Academic Press, London, 2000) pp. 156.
8) B. K. Kim, G. H. Ha, D. W. Lee and G. G. Lee: Adv. Performance
Mater. 5 (1998) 341352.
9) G. G. Lee, C. M. Moon and B. K. Kim: J. Korean Powder Metall. Inst.
10 (2003) 228234.
10) G. H. Ha, G. G. Lee, M. C. Yang and B. K. Kim: Proc. EURO PM 2006
11)
12)
13)
14)
15)
16)
17)
191
1 Hard Materials, (European Powder Metallurgy Association, Shrewsbury, 2006) pp. 97102.
G. G. Lee and B. K. Kim: Mater. Trans. 44 (2003) 21452150.
G. G. Lee and G. H. Ha: Mater. Trans. 47 (2006) 30073011.
S. Morooka, A. Kobat and K. Kusakabe: Ceramic Transaction 22,
Ceramic Powder Science IV, ed. by S. I. Hirano, G. L. Messing and H.
Hausner, (The American Ceramic Society, Inc., Ohio, 1991) pp. 109
114.
P. Villars, A. Prince and H. Okamoto: Handbook of Ternary Alloy
Phase Diagrams Vol. 6, (ASM International, New York, 1995)
pp. 74237434.
L. M. Berger, W. Gruner, E. Langholf and S. Stolle: Int. J. Refract. Met.
Hard Mater. 17 (1999) 235243.
R. Koc: J. Mater. Sci. 33 (1998) 10491055.
R. Shaviv: Mater. Sci. Eng. A 209 (1996) 345352.

Surface Modication of Porous Metals Using Friction Phenomena

Yong-Jai Kwon1 , Ichinori Shigematsu2 and Naobumi Saito2
1
School of Materials Science and Engineering, University of Ulsan, Ulsan 680-749, Korea
Materials Research Institute for Sustainable Development, National Institute of Advanced Industrial Science
and Technology (AIST), Nagoya 463-8560, Japan
2
The surface region of the aluminum foams (trade name ALPORAS) was modied through the friction stir processing (FSP) which was
performed by using friction phenomena with a high-speed rotating tool. The tool was rotated at speeds ranging from 820 to 2400 rpm, and
plunged from the top surface of the aluminum foam, and then traversed at speeds ranging from 50 to 300 mm/min. The surface-modied zone
(SMZ) had considerably smoother surface in comparison to the unprocessed zone (UZ). Especially for 1390 rpm and 150 mm/min, the
smoothest surface was obtained, which was attributed to the smaller amount of pores in the SMZ. In addition, a very dense layer was formed near
the surface of the SMZ through the localized collapse and densication of the cell structure near the surface region, which was attributed to the
friction phenomena with the high-speed rotating tool. The mechanical properties of the aluminum foams were signicantly improved through
the FSP. Especially for 1390 rpm and 150 mm/min, the SMZ exhibited the highest average maximum indentation strength and energy absorption
ability, which were equivalent to about 2.2 times the values of the UZ. The tool rotation speed and the tool traverse speed were very important
parameters not only in controlling the surface morphology, but also in improving the mechanical properties of the aluminum foams. The FSP
was a very eective technology for the remarkable improvement in the mechanical properties through the cell structure control of the surface
region of the aluminum foams, without any dense skin materials. [doi:10.2320/matertrans.MER2008288]
Keywords: friction stir processing, porous metal, friction phenomena, surface modication, cell structure, aluminum foam, surface
morphology, mechanical properties, indentation test, energy absorption
1.
Introduction
In the recent past, there has been a remarkable increase in

research on porous metals with extremely low relative
densities which are attributed to their cellular structure with a
great number of pores.17) This is because these materials
exhibit a variety of benecial properties, such as low weight,
high energy-absorbing capability and good acoustic damping
capacity. In addition, in comparison with polymer foams,
they have excellent recycling eciency, high specic stiness, good thermal conductivity and high melting point.
Many previous researches have been carried out on development and establishment of manufacturing processes in order
to produce the porous metals with consistent properties,813)
and also on investigation of their various characteristics for
high degree of freedom in engineering design.1418) These
eorts have generated a more diversied range of useful and
cost-eective manufacturing technologies, with enhancing a
fundamental understanding of the porous metals. Although
there are several problems to be solved for their practical
applications, the porous metals are potential materials for
further improvement in the various characteristics through
control of the cell structure and the cell wall material, as can
be expected from cell structures in nature. That is, nature also
creates and uses the cell structures, e.g. bones, plants, etc. In
general, however, they are more complex and more inhomogeneous than man-made ones. Namely, many natural cell
structures have porosity and/or pore size gradients or dense
structured layer near the surface for constructional and/or
functional purposes. This fact suggests that the porous metals
with suitably controlled inhomogeneous structures are
required to eectively capitalize on their benecial properties in various industrial applications. The processing
technologies should, hence, also be developed and established in connection with surface morphology control,
porosity and/or cell size distribution control, mechanical

properties improvement, etc.
In this context, we focused our attention on the fact that
porous metals can be surface-modied by the well-known
friction stir processing (FSP),19,20) as shown in Fig. 1. In this
process, a high-speed rotating tool is plunged to a designed
depth from the top surface of a xed porous metal, and then
traversed horizontally to the top surface. At this time, heat is
generated by the friction between the rotating tool and the
porous metal. This heat induces the decrease in the
deformation resistance of the cell wall material with
increasing in its temperature. The softened cell wall is
Fig. 1 Schematic illustration showing the basic principle of the friction stir
processing (FSP) in this study. The surface modied zone and the
unprocessed zone are labeled SMZ and UZ, respectively. And then, X, Z
and Y represent the tool traverse direction, the tool rotation axis direction,
and the width direction of porous metal, respectively.
plastically deformed near the surface region by the rotating

tool. And then, the surface-modied zone (SMZ) is formed
near the surface region of the porous metal by the local
collapse and densication of the cell structure. Through this
process, the mechanical properties of the porous metals are
enhanced by controlling the surface region structure, e.g.
surface morphology, porosity distribution, cell morphology,
etc.
In this study, the surface region of aluminum foams was
modied by the FSP. And then, the inuences of the tool
rotation speed and the tool traverse speed on surface
morphology, cell structure and mechanical properties of the
friction-stir-processed (FSPed) aluminum foams were experimentally investigated.
2.
Fig. 2 Optical macrograph showing a typical cell structure of the

aluminum foam used as the starting material in this study.
A commercially available aluminum foam (trade name

ALPORAS, Shinko Wire, Japan) with closed pores was used
as the starting material in this study. ALPORAS is produced
by adding calcium and titanium hydride (TiH2 ) powders to
molten aluminum. The calcium acts to increase the viscosity
of the melt. The bubbles are formed by gaseous hydrogen,
which is sourced from the titanium hydride. Further details
on the various characteristics, cell structures, manufacturing
processes, and on the suppliers of these foams are given by
Miyoshi et al.21) The dimensions of the foam were 100 mm
long 50 mm wide 30 mm thick, as shown in Fig. 1. A
at-bottomed circular cylindrical tool was made of quenchhardened SKD 61 (JIS), and its diameter was equal to 15 mm.
The tool was rotated in the clockwise direction at speeds
ranging from 820 to 2400 rpm, and plunged to a depth of
2.6 mm from the top surface of the aluminum foam, and then
traversed at speeds ranging from 50 to 300 mm/min to the
length direction (X-axis direction). At this time, the tool
rotation axis (Z-axis) was tilted by 10 degrees to the tool
traverse direction (X-axis direction). All experiments were
carried out at room temperature.
For cell structure observations, the unprocessed and FSPed
aluminum foams were cross-sectioned with the aid of electric
discharge machining in order to minimize any possible
machining damage. The FSPed aluminum foam was crosssectioned perpendicularly to the tool traverse direction. And
then, the cross-sections were mechanically polished before
examining the cell structures. Axisymmetric indentation tests
were performed by using a steel ball with a diameter of
10 mm. The steel ball was indented to a depth of about 5 mm
into the aluminum foam from the top surface at a constant
speed of 1.0 mm/min. At least three tests were conducted for
each aluminum foam surface-modied under the above
mentioned conditions.
3.
193
3.1 Cell structure of aluminum foam

Figure 2 shows the optical macrograph showing a typical
cell structure of the aluminum foam which was used the
starting materials in this study. The aluminum foam had the
porosity of 91%, and the apparent density of about 0.24 g/
cm3 which was equal to the relative density of about 8.9%. A
great number of closed pores were ranged from about 1 to

4 mm in diameter, which were nearly equivalent to the
relative frequency of about 70%. The average pore diameter
was equal to about 2.4 mm. The cell wall thickness was
mainly distributed in the range of 100 to 250 mm, which the
average value corresponded approximately to 190 mm.
3.2 Surface appearance
Figure 3 shows the surface appearances of the aluminum
foams surface-modied at the tool rotation speeds of (a) 820,
(b) 1390, and (c) 2400 rpm under the constant tool traverse
speed of 150 mm/min. The unprocessed zone (UZ) represents the cell and surface structures of the starting material. In
contrast, a new surface structure was formed in the FSPed
aluminum foams for all the tool rotation speeds during the
FSP, which was labeled SMZ and had the appearance of
shiny metallic luster. Namely, the surface of the SMZ was
considerably smoother than that of the UZ, although traces of
the cell structure were slightly observed in the SMZ.
Especially for 1390 rpm, the smoothest surface was obtained,
which was attributed to the smaller amount of pores in the
SMZ. This result shows that the rotation speed is a very
important parameter for controlling in the surface morphology of the porous metals through the FSP.
Figure 4 shows the surface appearances of the aluminum
foams surface-modied at the tool traverse speeds of (a) 50
and (b) 300 mm/min under the constant tool rotation speed of
1390 rpm. Note the case of 150 mm/min was showed in
Fig. 2(b). Also in these cases, a new surface structure with a
shiny metallic luster was formed during the FSP as well as the
above-mentioned cases. Especially for 150 mm/min, as
shown in Fig. 2(a), the smoothest surface was obtained. This
result suggests that there is some optimum tool traverse speed
to obtain porous metals with a smooth surface for the FSP.
These experimental results demonstrate that the FSP is one
of the eective technologies for the surface morphology
control of the porous metals, and the tool rotation speed and
the tool traverse speed are very important parameters in this
process.
3.3 Macrostructure
Figure 5 shows a typical optical macrograph of the crosssections perpendicular to the tool traverse direction (X-axis
direction) of the FSPed aluminum foams. The surface region
194
Y.-J. Kwon, I. Shigematsu and N. Saito
Fig. 4 Surface appearances of the aluminum foams surface-modied at the

tool traverse speeds of (a) 50 and (b) 300 mm/min under the constant tool
rotation speed of 1390 rpm.
Fig. 3 Surface appearances of the aluminum foams surface-modied at the

tool rotation speeds of (a) 820, (b) 1390 and (c) 2400 rpm under the
constant tool traverse speed of 150 mm/min.
of the UZ was composed of the open pores, although the

closed pores were observed in the inner region. In contrast, a
very dense layer was formed near the surface of the SMZ.
From this result, it can be known that some cell wall material
was plastically deformed near the surface region by the highspeed rotating tool during the FSP. The buckling of the cell
wall was observed beneath the dense surface layer of the
SMZ. These results show that the SMZ was formed by the
localized collapse and densication of the cell structure
near the surface region, which was attributed to friction
phenomena with the rotating tool. In addition, these results
demonstrate that the FSP is a very eective technology in
forming the dense layer near the surface of the porous metals,
without any dense skin materials.
3.4 Mechanical properties
Figure 6(a) shows the load-displacement proles during
the indentation test of (1) the UZ (i.e., starting material), and
the SMZs of the aluminum foams which were produced at the
Fig. 5 Typical optical macrograph of the cross-sections perpendicular to

the tool traverse direction of the surface-modied aluminum foam.
tool rotation speeds of (2) 820, (3) 1390 and (4) 2400 rpm
under the constant tool traverse speed of 150 mm/min. And
then, Fig. 6(b) shows the indentation load-displacement
curves of the SMZs of the aluminum foams which were
produced at the tool traverse speeds of (5) 50 and (6)
300 mm/min under the constant tool rotation speed of
1390 rpm. In all cases, the indentation load-displacement
curves were qualitatively similar tendency. Namely, the
indentation load was gradually increased with the increase
Fig. 6 Indentation Load-displacement curves of (1) the unprocessed zone

(i.e., starting material), and the surface-modied zones of the aluminum
foams produced at the tool rotation speeds of (2) 820, (3) 1390 and (4)
2400 rpm under the constant tool traverse speed of 150 mm/min, and at the
tool traverse speeds of (5) 50 and (6) 300 mm/min under the constant tool
rotation speed of 1390 rpm, respectively.
in the displacement until reaching to the plateau region,

without the remarkable yield phenomena.
Figure 7 shows the inuences of (a) the tool rotation speed
and (b) the tool traverse speed on the maximum indentation
load of the SMZ of the aluminum foams which were
produced at the constant tool traverse speed of 150 mm/min
and tool rotation speed of 1390 rpm, respectively. The
maximum indentation load means the peak load in each
load-displacement curve. In all cases, the SMZ exhibited
higher average indentation strength than the UZ, i.e. starting
material, although there was not a remarkable tendency for
the variation of the tool rotation speed and the tool traverse
speed. However, it is noteworthy that, for 1390 rpm and
150 mm/min, the highest average indentation strength was
obtained in this study, which was about 2.2 times the value of
the UZ. This result suggests that there is some optimum tool
rotation speed and tool traverse speed to obtain porous metals
with the highest indentation strength through the FSP.
Figure 8 shows the inuences of (a) the tool rotation speed
and (b) the tool traverse speed on the energy absorption
ability of the SMZ of the aluminum foams which were
produced at the constant tool traverse speed of 150 mm/min
and tool rotation speed of 1390 rpm, respectively. The energy
absorption ability was calculated by integrating each loaddisplacement curve. There was not a noticeable trend for the
change of the tool rotation speed and the tool traverse speed.
In all cases, however, the SMZ exhibited higher average
energy absorption ability than the UZ, i.e. starting material.
Especially for 1390 rpm and 150 mm/min, the SMZ exhib-
195
Fig. 7 Inuences of (a) the tool rotation speed and (b) the tool traverse
speed on the maximum indentation load of the surface-modied zone of
the aluminum foams produced at the constant tool traverse speed of
150 mm/min and tool rotation speed of 1390 rpm, respectively. The
positive and negative error bars indicate the maximum and minimum
values, respectively.
Fig. 8 Inuences of (a) the tool rotation speed and (b) the tool traverse
speed on the energy absorption ability of the surface-modied zone of the
aluminum foams produced at the constant tool traverse speed of 150 mm/
min and tool rotation speed of 1390 rpm, respectively. The positive and
negative error bars indicate the maximum and minimum values,
respectively.
196
Y.-J. Kwon, I. Shigematsu and N. Saito
ited the highest average energy absorption ability in this

study, which was equivalent to about 2.2 times the value of
the UZ. This result demonstrates that there is some optimum
tool rotation speed and tool traverse speed to obtain porous
metals with the most excellent energy absorption ability
through the FSP.
It is noteworthy that the smoothest surface was also
obtained for 1390 rpm and 150 mm/min, as shown in
Fig. 3(b). This result shows that the tool rotation speed and
the tool traverse speed are very important parameters not
only in controlling the surface morphology but also in
improving the mechanical properties of the porous metals. In
addition, from the above-mentioned results, it was claried
that the FSP is a very eective technology for the remarkable
improvement in the mechanical properties through the cell
structure control of the surface region of the porous metals,
without any dense skin materials.
4.
Conclusions
In this study, the surface region of the aluminum foams

was modied by the FSP. And then, the inuences of the tool
rotation speed and the tool traverse speed on the surface
morphology, the cell structure and the mechanical properties
of the FSPed aluminum foams were experimentally investigated. The following results were obtained.
(1) The surface of the SMZ was considerably smoother than
that of the UZ. Especially for 1390 rpm and 150 mm/min, the
smoothest surface was obtained, which was attributed to the
smaller amount of pores in the SMZ.
(2) A very dense layer was formed near the surface of the
SMZ through the localized collapse and densication of the
cell structure in the surface region, which was attributed to
friction phenomena with the high-speed rotating tool.
(3) The mechanical properties of the aluminum foams were
signicantly improved through the FSP. Especially for
1390 rpm and 150 mm/min, the SMZ exhibited the highest
average maximum indentation strength and energy absorption ability in this study, which were equivalent to about
2.2 times the values of the UZ.
(4) In the FSP, the tool rotation speed and the tool traverse
speed are very important parameters not only in controlling
the surface morphology but also in improving the mechanical
properties of the aluminum foams.
(5) The FSP was a very eective technology for the
remarkable improvement in the mechanical properties

through the cell structure control of the surface region of
the aluminum foams, without any dense skin materials.
Acknowledgement
This work was supported by the 2008 Research Fund of
University of Ulsan.
REFERENCES
1) L. J. Gibson and M. F. Ashby: Cellular Solids: Structures and
Properties, 2nd ed., (Cambridge University Press, Cambridge, 1997).
2) M. Ashby, N. Fleck, H. Wadley, J. Hutchinson and L. Gibson: Metal
Foams: A Design Guide, (Butterworh-Heinemann, Boston, 2000).
3) J. Banhart: Prog. Mater. Sci. 46 (2001) 559632.
4) F. Yi, Z. Zhu, F. Zu, S. Hu and P. Yi: Mater. Charact. 47 (2001) 417
422.
5) C. M. Moran, J. A. Ross, C. Cunningham, M. Butler, T. Anderson,
D. Newby, K. A. A. Fox and W. N. McDicken: Ultrasound Med. Biol.
32 (2006) 421428.
6) Z. Xie, T. Ikeda, Y. Okuda and H. Nakajima: Mater. Sci. Eng. A 386
(2004) 390395.
7) W. Jiejun, L. Chenggong, W. Dianbin and G. Manchang: Comp. Sci.
Technol. 63 (2003) 569574.
8) T. Miyoshi, M. Itoh, S. Akiyama and A. Kitahara: Porous and Cellular
Materials for Structural Applications, ed. D. S. Schwartz, D. S. Shih,
A. G. Evans, H. N. G. Wadley, (MRS, San Francisco, 1998).
9) L. Ma and Z. Song: Scr. Mater. 39 (1998) 15231528.
10) A. E. Simone and L. J. Gibson: Acta Mater. 46 (1998) 31093123.
11) V. Gergely and T. W. Clyne: Adv. Eng. Mater. 2 (2000) 175178.
12) V. Gergely, H. P. Degischer and T. W. Clyne: Comprehensive
Composite Materials, Vol. 3: Metal Matrix Composites, ed. T. W.
Clyne, (Elsevier, Amsterdam, 2000).
13) O. Prakash, H. Sang and J. D. Embury: Mater. Sci. Eng. A 199 (1995)
195203.
14) Y. Sugimura, J. Meyer, M. Y. He, H. Bart-Smith, J. Grenestedt and
A. G. Evans: Acta Mater. 45 (1997) 52455259.
15) J. T. Beals and M. S. Thompson: J. Mater. Sci. 32 (1997) 35953600.
16) W. Sanders and L. J. Gibson: Porous and Cellular Materials for
Structural Applications, ed. D. S. Schwartz, D. S. Shih, A. G. Evans,
H. N. G. Wadley, (MRS, San Francisco, 1998).
17) E. Andrews, W. Sanders and L. J. Gibson: Mater. Sci. Eng. A 270
(1999) 113124.
18) H. Fusheng and Z. Zhengang: J. Mater. Sci. 34 (1999) 291299.
19) Y. J. Kwon, I. Shigematsu and N. Saito: J. Japan Inst. Metals 66 (2002)
13251332.
20) Y. J. Kwon, I. Shigematsu and N. Saito: J. Japan Inst. Metals 67 (2003)
547554.
21) T. Miyoshi, M. Itoh, S. Akiyama and A. Kitahara: Adv. Eng. Mater. 2
(2000) 179183.

Dissimilar Friction Stir Welding for Tailor-Welded Blanks

of Aluminum and Magnesium Alloys
Ichinori Shigematsu1 , Yong-Jai Kwon2; * and Naobumi Saito1
1
Materials Research Institute for Sustainable Development, National Institute of Advanced Industrial Science
and Technology (AIST), Nagoya 463-8560, Japan
2
School of Materials Science and Engineering, University of Ulsan, Ulsan 680-749, Korea
In this study, tailor-welded blanks (TWBs) composed of A5052P-O aluminum and AZ31B-O magnesium alloys were produced by the
friction stir welding process. The sound surfaces without large defects were successfully obtained at the tool rotation speeds ranging from 1000
to 1400 rpm under the constant tool traverse speed of 300 mm/min, and the surface roughness was decreased in the friction-stir-welded zone
(SZ) with the increase in the tool rotation speed. Also for 100, 300 and 500 mm/min, the sound surfaces without large defects were obtained
under the constant tool rotation speed of 1400 rpm. The increase of the tool traverse speed led to the decrease in the surface roughness of the SZ.
In the SZ, the bonded interface was clearly evident with plastic ow pattern between the aluminum and magnesium alloys, although an onion
ring pattern was not formed. During tensile testing, the TWBs were fractured in the SZ at the early stage of the plastic deformation, i.e. strain
hardening region, without remarkable area reduction near the fracture region. The TWBs exhibited the similar average tensile strength, showing
that there were not noteworthy changes in the tensile strength as a function of the tool rotation speed. The average tensile strength, however, was
slightly decreased with the increase in the tool traverse speed. It is noticeable that, in all cases, the joint eciency of the TWBs exceeded 62%,
and the maximum average tensile strength of about 143 MPa was obtained at 1400 rpm under the tool traverse speed of 100 mm/min, which was
nearly equivalent to the joint eciency of about 72%. The average total elongation of the TWBs was about 2% or less considerably lower than
those of the base metals, without signicant changes as functions of the tool traverse speed and the tool traverse speed.
[doi:10.2320/matertrans.MER2008326]
(Received September 5, 2008; Accepted November 4, 2008; Published December 17, 2008)
Keywords: tailor-welded blank, aluminum alloy, magnesium alloy, friction stir welding, surface roughness, macrostructure, tensile properties,
joint eciency
1.
Introduction
Recently, there is increasing interest in aluminum and

magnesium alloys in many industrial elds, because of a
wide variety of unique properties, especially lightweight
properties. For their practical applications, bonding and
welding technologies should also be established aside from
alloy design, microstructure control, plastic forming, casting
and surface treatment etc.
Friction stir welding (FSW) process is one of the attractive
technologies for welding of metallic materials. Especially for
aluminum18) and magnesium912) alloys, the FSW process
has been actively studied as a new solid state welding
technology, since it was invented by The Welding Institute
(TWI) in 1991.13) In addition, the FSW has recently gotten a
lot of attention also in order to produce tailor-welded blanks
(TWBs) composed of dissimilar materials.1420) However,
there were not a lot of researches on dissimilar FSW between
the aluminum and magnesium alloys.1822) Hirano et al.18)
have carried out a dissimilar FSW between 1050 aluminum
and AZ31 magnesium alloys, and reported that an intermediate layer was formed near their bonded interface, which
was composed of an intermetallic compound Al12 Mg17 and
an undened compound. Sato et al.19) also have studied on
the same system as Hirano et al.,18) although the plate
thickness was dierent. According to their research, a
eutectic microstructure was formed in the friction-stirwelded zone (SZ), which was composed of some pores,
magnesium solid solution, and a large volume of intermetallic compound Al12 Mg17 . Yan et al.20) have found that
*Corresponding
author, E-mail: yongjaikwon@ulsan.ac.kr
intermetallic compounds, i.e. Al3 Mg2 and Al12 Mg17 , were

formed in the SZ, when the rotating probe was traversed
along the centerline of the SZ in a dissimilar FSW between
1060 aluminum and AZ31 magnesium alloys. The abovementioned previous researches show that the FSW is
applicable to the dissimilar bonding between the aluminum
and magnesium alloys. The previous researches, however,
were mainly performed on the macro- and micro-structure
characteristics, and/or hardness properties of the TWBs.
Hence, it is dicult to understand fully the inuences of
principal processing parameters, e.g. tool rotation speed and
tool traverse speed etc., on the various mechanical properties
of the TWBs produced by the dissimilar FSW between the
aluminum and magnesium alloys.
In this context, TWBs composed of A5052P-O aluminum
and AZ31B-O magnesium alloys were produced by the FSW
process in this study. And then, the inuences of the tool
rotation speed and tool traverse speed on surface appearance,
macrostructure and tensile properties of the TWBs were
experimentally investigated.
2.
Figure 1 shows a schematic representation of the FSW

process for the production of the TWBs composed of the
A5052P-O aluminum and AZ31B-O magnesium alloys in
this research. Note that, in this paper, the friction-stir-welded
zone (SZ) implies a region changed in the microstructure by
mechanical stirring action of the rotating tool during the FSW
process. The dimensions of the base metal plates were 30 mm
wide 160 mm long 2 mm thick. In all FSW experiments, the aluminum and magnesium alloy plates were
198
I. Shigematsu, Y.-J. Kwon and N. Saito
Fig. 2
Fig. 1 Schematic representation of the friction stir welding (FSW) process
in this research. The friction-stir-welded zone, the advancing side and the
retreating side are labeled SZ, AS and RS, respectively. And then, X, Z and
Y represent the tool traverse direction, the tool rotation axis direction and
the rolling direction of the base metals, respectively.
located on the advancing side (AS) and the retreating side

(RS), respectively. This was because the arrangement of the
base metals may aect the characteristics of the friction-stirwelded (FSWed) plates. The shoulder diameter of the tool
was equal to 10 mm. The diameter and height of the probe
were 4 and 1.7 mm, respectively. The tool was rotated in a
clockwise direction at speeds ranging from 800 to 1600 rpm,
and then the probe was inserted into the materials. The
rotating tool was traversed at speeds ranging from 100 to
500 mm/min along the weld line perpendicular to the rolling
direction of the base metal plates. The tool rotation axis was
tilted by 3 to the tool traverse direction during the FSW
process.
Macrostructural observations were performed on XY and
YZ planes which were the cross-sections parallel and
perpendicular to the tool traverse direction, respectively. In
the FSW, processing parameters will exert an eect on
surface morphology of the SZ, which aect various characteristics of the FSWed plates. Unfortunately, there is little
research on the inuences of the various processing param-
Geometry and dimensions of the tensile test specimen.
eters on the surface morphology. Therefore, in this study, the

inuences of the tool rotation speed and tool traverse speed
on the surface roughness prole of the SZ were investigated.
The surface roughness proles were measured at intervals of
100 mm to the tool traverse direction along the centerline on
the surface of the SZ by a laser displacement meter (trade
name LC-2450, Keyence, Japan). Figure 2 show the geometry and dimensions of the tensile test specimen. For the
tensile tests, the TWBs were cut perpendicularly to the tool
traverse direction, and then machined into the tensile test
specimens with a parallel portion of 4 mm wide 24 mm
long 1 mm thick and a llet radius of 5 mm. The tensile
tests were carried out at a constant crosshead speed of 1 mm/
min under room temperature.
3.
3.1 Inuences of tool rotation speed

3.1.1 Surface appearance and morphology
Figure 3 shows the surface appearances of the TWBs
produced at the tool rotation speeds of (a) 800, (b) 1000, (c)
1200, (d) 1400 and (e) 1600 rpm under the constant tool
traverse speed of 300 mm/min. For (a) 800 rpm, large defects
with valley-like structures were discontinuously formed with
a large amount of burr to the tool traverse direction (X-axis
Fig. 3 Surface appearances of the tailor-welded blanks produced at the tool rotation speeds of (a) 800, (b) 1000, (c) 1200, (d) 1400 and (e)
1600 rpm under the constant tool traverse speed of 300 mm/min.
Dissimilar Friction Stir Welding for Tailor-Welded Blanks of Aluminum and Magnesium Alloys
199
Fig. 4 Surface roughness proles along the centerlines to the tool traverse
direction of the friction-stir-welded zone in the tailor-welded blanks
produced at the tool rotation speeds of (a) 800, (b) 1000, (c) 1200, (d) 1400
and (e) 1600 rpm under the constant tool traverse speed of 300 mm/min.
Note that the scale of vertical axis is dierent in each prole.
direction) in the SZ with a very rough surface. In contrast,

sound surfaces without large defects were successfully
obtained at the increased tool rotation speeds of (b) 1000,
(c) 1200 and (d) 1400 rpm. In addition, the surface morphology of the SZ became smoother with the increase in the tool
rotation speed. For (e) 1600 rpm, however, large defects
presenting as cracks were again formed with a large amount
of burr along the tool traverse direction near the center region
of the SZ. These results show that there are optimum tool
rotation speeds in order to obtain sound surfaces for the
dissimilar FSW between the A5052P-O aluminum and
AZ31B-O magnesium alloys.
Figure 4 show the surface roughness proles along the
centerlines to the tool traverse direction of the SZ in the
TWBs produced at the tool rotation speeds of (a) 800, (b)
1000, (c) 1200, (d) 1400 and (e) 1600 rpm under the constant
tool traverse speed of 300 mm/min. Note that the scale of
vertical axis is dierent in each prole. For (a) 800 rpm, the
SZ had the very rough surface with the roughness distribution
range of 4.64 mm, which was associated with the large
macroscopic defects foamed in the SZ. The relatively smooth
surfaces were obtained at (b) 1000, (c) 1200 and (d)
1400 rpm, where the sound surfaces without large defects
were successfully obtained. In these cases, the roughness
distribution ranges were (b) 0.18, (c) 0.12 and (d) 0.08 mm,
showing that the surface roughness was decreased with
the increase in the tool rotation speed. However, for (e)
1600 rpm, the surface roughness was again increased, and
distributed at a range of 5.87 mm, which was attributed to
the large macroscopic defect. These results show that there
are optimum tool rotation speeds in order to obtain the
sound SZ with the smooth surface for the dissimilar FSW
between the A5052P-O aluminum and AZ31B-O magnesium
alloys.
3.1.2 Macrostructure
Figure 5 show the typical optical macrographs of the
cross-sections of the friction-stir-welded (FSWed) TWB. (a)
YZ plane was the cross-section perpendicular to the tool
traverse direction. And, (b) XY plane 1 and (c) 2 were the
cross-sections near the top surface and the central region of
the SZ, respectively, which were parallel to the tool traverse
Fig. 5 Typical optical macrographs of the cross-sections of the tailorwelded blank: (a) YZ plane, (b) XY plane 1 (near the top surface) and (c)
XY plane 2 (near the central region of the friction-stir-welded zone).
direction. Yan et al.20) have reported that cracks were formed

in the SZ when the rotating probe was traversed along the
center of weld line. In this study, however, no bonding
defects were formed near and in the SZ for the tool rotation
speeds ranging from 1000 to 1400 rpm, even though the
rotating probe was traversed along the center of the weld line.
This result shows that there are some optimum conditions for
each dissimilar FSW. In all the cross-sections, the bonded
interface was clearly evident between the aluminum and
magnesium alloys. Plastic ow pattern was also formed in the
SZ, which was more clearly observed on the aluminum alloy
region. As shown in Fig. 5(a), however, an onion ring pattern
was not formed, which was sometimes observed in the SZ
between similar materials.68) Although a precise conclusion
is not yet possible, it is likely that no formation of the onion
ring pattern results from dierences in plastic ow behavior
alloys. As mentioned above, Sato et al.19) have reported that a
eutectic microstructure was formed in the SZ by constitutional liquation and solidication for a dissimilar FSW
between aluminum and magnesium alloys. In this study,
however, such solidication microstructure was not formed
suggesting that the microstructure of the SZ is dependent on
FSW conditions for each system.
3.1.3 Tensile properties
Figure 6 shows the appearances of the tensile tested
specimens of (a) the A5052P-O aluminum alloy and (b)
AZ31B-O magnesium alloy plates, and the TWBs produced
at the tool rotation speeds of (c) 1000, (d) 1200 and (e)
1400 rpm under the constant tool traverse speed of 300 mm/
min. The (a) A5052P-O aluminum alloy and (b) AZ31B-O
magnesium alloy plates were fractured in the parallel portion.
And, in all cases, the FSWed TWBs were fractured near
the central region of the parallel portion, suggesting that
the fracture occurred in the SZ without remarkable area
reduction near the fracture region.
Figure 7 shows the typical nominal stress-strain curves for
the tensile deformation of the A5052P-O aluminum alloy and
200
Fig. 6 Appearances of the tensile tested specimens of (a) the A5052P-O

aluminum alloy and (b) AZ31B-O magnesium alloy plates, and the tailorwelded blanks produced at the tool rotation speeds of (c) 1000, (d) 1200
and (e) 1400 rpm under the constant tool traverse speed of 300 mm/min.
Fig. 7 Typical nominal stress-strain curves for the tensile deformation of

the A5052P-O aluminum alloy and AZ31B-O magnesium alloy plates.
AZ31B-O magnesium alloy plates. In both cases, the curves

exhibited a tendency similar to a typical stress-strain curve of
other aluminum and magnesium alloys. Namely, the stress
was abruptly increased with elastic strain at an early stage,
and then the ow curve exhibited a continuous type of
transition behavior from elastic to plastic deformation which
produces a yield point in the typical stress-strain curve of
aluminum and magnesium alloys. In the plastic deformation
region, the stress increased with plastic strain, and reached
the maximum value because of strain strengthening, and then
dropped continuously with further strain until fracture.
Serration phenomenon was clearly observed only in the
plastic deformation region of the aluminum alloy. The
magnesium alloy had higher tensile and yield strengths than
the aluminum alloy. The elongation of the aluminum alloy,
however, was larger than that of the magnesium alloy.
Figure 8 shows the typical nominal stress-strain curves for
the tensile deformation of the TWBs produced at the tool
rotation speeds of 1000, 1200 and 1400 rpm under the
constant tool traverse speed of 300 mm/min. For comparison,
the enlarged curves for the A5052P-O aluminum and

the tailor-welded blanks produced at the various tool rotation speeds under
the constant tool traverse speed of 300 mm/min. The enlarged curves for
the A5052P-O aluminum and AZ31B-O magnesium alloys were also
shown for comparison.
AZ31B-O magnesium alloys were also shown. In all the

TWBs, the curves had a tendency similar to those of the base
metals. The TWBs, however, were fractured at the early stage
of the plastic deformation, i.e. strain hardening region, after
the curves exhibited a continuous type of transition behavior
from elastic to plastic deformation. As a result, the TWBs
exhibited lower tensile strength and elongation than those of
the base metals, although their yield strengths were similar to
that of the aluminum alloy.
Figure 9 shows the inuence of the tool rotation speed on
the (a) average ultimate tensile strength and total elongation,
and (b) joint eciency of the TWBs, which are produced
under the constant tool traverse speed of 300 mm/min. The
broken and solid lines show the average ultimate tensile
strength and elongation of the A5052P-O aluminum and
AZ31B-O magnesium alloys, respectively. And then, the
positive and negative error bars represent the maximum and
minimum values, respectively. The joint eciency was
calculated by dividing the tensile strengths of the TWBs by
that of the aluminum alloy. The magnesium alloy exhibited
the average tensile strength of about 256 MPa, which was
higher than that of the aluminum alloy (about 200 MPa). In
all the tool rotation speeds, the TWBs had the similar tensile
strength, showing that there were not noteworthy changes in
the tensile strength as a function of the tool rotation speed. It
is noticeable that, in all cases, the average joint eciency of
the TWBs exceeded 66%, and the maximum average tensile
strength of about 137 MPa was obtained at 1000 rpm, which
was nearly equivalent to the joint eciency of about 69%.
Unfortunately, however, the average elongation of the
TWBs was about 2% or less considerably lower than those
of the base metals, without signicant change as a function
of the tool rotation speed. The previous researches have
reported that fracture of FSWed specimens occurred near
intermetallic compounds such as Al12 Mg17 and Al3 Mg2
which were formed in the SZ during dissimilar FSW.18,22)
Hence, it is believed that formation and/or distribution of the
intermetallic compounds should be controlled for further
improvement of the mechanical properties by various
processing parameters.
201
Fig. 10 Surface appearances of the tailor-welded blanks produced at the

tool traverse speeds of (a) 100 and (b) 500 mm/min under the constant tool
rotation speed of 1400 rpm.
Fig. 9 Inuence of the tool rotation speed on the (a) average ultimate
tensile strength and total elongation, and (b) joint eciency of the tailorwelded blanks produced under the constant tool traverse speed of 300 mm/
min. The broken and solid lines show the average ultimate tensile strength
and elongation of the A5052P-O aluminum alloy and AZ31B-O
magnesium alloy plates, respectively. And then, the positive and negative
error bars represent the maximum and minimum values, respectively.
3.2 Inuences of tool traverse speed

3.2.1 Surface appearance and morphology
Figure 10 shows the surface appearances of the TWBs
produced at the tool traverse speeds of (a) 100 and (b)
1400 rpm. Sound surfaces without large defects were successfully obtained in all the tool traverse speeds including
300 mm/min shown in Fig. 3(d). For (a) 100 mm/min,
however, the continuous burr with a wave pattern was
formed to the tool traverse direction (X-axis direction) on
the aluminum alloy side near the SZ with a relatively
rough surface. These results show that the sound surfaces
without large defects can be obtain in a relatively wide
range of the tool traverse speeds for the dissimilar FSW
alloys.
Figure 11 shows the surface roughness proles along the
centerlines to the tool traverse direction of the SZ in the
TWBs produced at the tool traverse speeds of (a) 100 and (b)
1400 rpm. For (a) 100 mm/min, the SZ had the relatively
rough surface with the roughness distribution range of
0.17 mm. The surface roughness, however, was decreased
with the increase in the tool traverse speed, showing the
tool traverse speed is an important parameter in controlling
the surface roughness of the SZ.
Fig. 11 Surface roughness proles along the centerlines to the tool traverse
direction of the friction-stir-welded zone in the tailor-welded blanks
produced at the tool traverse speeds of (a) 100 and (b) 500 mm/min under
the constant tool rotation speed of 1400 rpm. Note that the scale of vertical
axis is dierent in each prole.
Fig. 12 Appearances of the tensile tested specimens of the tailor-welded

blanks produced at the tool traverse speeds of (a) 100 and (b) 500 mm/min
under the constant tool rotation speed of 1400 rpm.
3.2.2 Tensile properties

Figure 12 shows the appearances of the tensile tested
specimens of the TWBs produced at the tool traverse speeds
of (a) 100 and (b) 500 mm/min under the constant tool
rotation speed of 1400 rpm. The TWBs were fractured near
the central region of the parallel portion for (a) 100 and (b)
500 mm/min as well as for 300 mm/min shown in Fig. 6(d),
suggesting that the fracture occurred in the SZ without
remarkable area reduction near the fracture region, similarly
to the above-mentioned ones.
202

the tailor-welded blanks produced at the various tool traverse speeds under
the constant tool rotation speed of 1400 rpm. The enlarged curves for the
A5052P-O aluminum and AZ31B-O magnesium alloys were also shown
for comparison.
Figure 13 shows the typical nominal stress-strain curves

for the tensile deformation of the TWBs produced at the
various tool traverse speeds under the constant tool rotation
speed of 1400 rpm. For comparison, the enlarged curves for
the A5052P-O aluminum and AZ31B-O magnesium alloys
were also shown. In all the TWBs, the curves had a tendency
similar to the above-mentioned ones. Namely, the TWBs
were fractured at the early stage of the plastic deformation,
i.e. strain hardening region, after the curves exhibited a
continuous type of transition behavior from elastic to plastic
deformation. As a result, the TWBs exhibited lower tensile
strength and elongation than those of the base metals,
although their yield strengths were similar to that of the
aluminum alloy.
Figure 14 shows the inuence of the tool traverse speed on
the (a) average ultimate tensile strength and total elongation,
and (b) joint eciency of the TWBs produced under the
constant tool rotation speed of 1400 rpm. The broken and
solid lines show the average ultimate tensile strength and
total elongation of the A5052P-O aluminum alloy and
AZ31B-O magnesium alloy plates, respectively. And then,
the positive and negative error bars represent the maximum
and minimum values, respectively. In all the tool traverse
speeds, the TWBs exhibited lower average tensile strength
than the base metals. It is noticeable that, in all cases, the
average joint eciency of the TWBs exceeded 62%, and
the maximum average tensile strength of about 143 MPa
was obtained at 100 mm/min, which was nearly equivalent
to the joint eciency of about 72%. The average tensile
strength was slightly decreased with the increase in the tool
traverse speed. The average elongation of the TWBs was
about 2% or less considerably lower than those of the base
metals, without signicant changes as a function of the tool
traverse speed.
4.
Conclusions
In this study, the TWBs of A5052P-O aluminum and

AZ31B-O magnesium alloys were produced by the FSW
process. And then, the inuences of the tool rotation speed
and tool traverse speed on surface appearance, macrostruc-
Fig. 14 Inuence of the tool traverse speed on the (a) average ultimate
tensile strength and total elongation, and (b) joint eciency of the tailorwelded blanks produced under the constant tool rotation speed of
1400 rpm. The broken and solid lines show the average ultimate tensile
strength and total elongation of the A5052P-O aluminum alloy and
AZ31B-O magnesium alloy plates, respectively. And then, the positive
and negative error bars represent the maximum and minimum values,
respectively.
ture and tensile properties of the TWBs were experimentally

investigated. The following results were obtained.
(1) The sound surfaces without large defects were successfully obtained at 1000, 1200 and 1400 rpm under the constant
tool traverse speed of 300 mm/min, although, for 800 and
1600 rpm, the large macroscopic defects were formed in the
SZ with the rough surface. In addition, also for 100, 300 and
500 mm/min, the sound surfaces without large defects were
obtained under the constant tool rotation speed of 1400 rpm.
(2) In the range of 1000 to 1400 rpm, the surface roughness
was decreased in the SZ with the increase in the tool rotation
speed under the constant tool traverse speed of 300 mm/min.
In addition, the increase of the tool traverse speed led to the
decrease in the surface roughness under the constant tool
rotation speed of 1400 rpm. These results show that the tool
rotation speed and tool traverse speed are very important
parameters in controlling the surface morphology of the SZ.
(3) In the SZ, the bonded interface was clearly evident with
plastic ow pattern between the aluminum and magnesium
alloys, although an onion ring pattern was not formed.
(4) During tensile testing, the FSWed TWBs were fractured
in the SZ at the early stage of the plastic deformation, i.e.
strain hardening region, without remarkable area reduction
near the fracture region.
(5) The TWBs exhibited the similar average tensile strength,
showing that there were not noteworthy changes in the tensile
strength as a function of the tool rotation speed. The average

tensile strength, however, was slightly decreased with the
increase in the tool traverse speed. It is noticeable that, in all
cases, the joint eciency of the TWBs exceeded 62%, and
the maximum average tensile strength of about 143 MPa
was obtained at 1400 rpm under the tool traverse speed of
100 mm/min, which was nearly equivalent to the joint
eciency of about 72%.
(6) The average total elongation of the TWBs was about 2%
or less considerably lower than those of the base metals,
without signicant changes as functions of the tool traverse
speed and the tool traverse speed.
REFERENCES
1) W. M. Thomas and E. D. Nicholas: Mater. Des. 18 (1997) 269273.
2) C. J. Dawes and W. M. Thomas: Weld. J. 75 (1996) 4145.
3) A. Sanderson, C. S. Punshon and J. D. Russell: Fusion Eng. Des. 4950
(2000) 7787.
4) I. Shigematsu, Y. J. Kwon, K. Suzuki, T. Imai and N. Saito: J. Mater.
Sci. Lett. 22 (2003) 353356.
5) I. Shigematsu, K. Suzuki, T. Imai, Y. J. Kwon and N. Saito: J. Mater.
Sci. 40 (2005) 29712974.
6) K. N. Krishnan: Mater. Sci. Eng. A 327 (2002) 246251.
7) W. B. Lee, Y. M. Yeon and S. B. Jung: Scr. Mater. 49 (2003) 423428.
203
8) Y. Chen, H. Liu and J. Feng: Mater. Sci. Eng. A 420 (2006) 2125.
9) M. B. Kannan, W. Dietzel, R. Zeng, R. Zettler and J. F. D. Santos:
Mater. Sci. Eng. A 460461 (2007) 243250.
10) B. M. Darras, M. K. Khraisheh, F. K. Abu-Farha and M. A. Omar:
J. Mater. Proc. Tech. 191 (2007) 7781.
11) R. P. Dobriyal, B. K. Dhindaw, S. Muthukumaran and S. K. Mukherjee:
Mater. Sci. Eng. A 477 (2008) 243249.
12) N. Afrin, D. L. Chen, X. Cao and M. Jahazi: Mater. Sci. Eng. A 472
(2008) 179186.
13) W. M. Thomas, D. E. Nicholas, J. C. Needham, M. G. Murch, P.
Templesmith and C. J. Dawes: GB Patent Application No. 9125978.8,
Dec. 1991, US Patent No. 5460317, Oct. 1995.
14) A. Steuwer, M. J. Peel and P. J. Withers: Mater. Sci. Eng. A 441 (2006)
187196.
15) T. Watanabe, H. Takayama and A. Yanagisawa: J. Mater. Proc. Tech.
178 (2006) 342349.
16) A. Scialpi, M. De Giorgi, L. A. C. De Filippis, R. Nobile and F. W.
Panella: Mater. Des. 29 (2008) 928936.
17) A. Abdollah-Zadeh, T. Saeid and B. Sazgari: J. Alloy. Compd. 460
(2008) 535538.
18) S. Hirano, K. Okamoto, M. Doi, H. Okamura, M. Inagaki and Y. Aono:
Quart. J. Jpn. Weld. Soc. 21 (2003) 539545.
19) Y. S. Sato, S. H. C. Park, M. Michiuchi and H. Kokawa: Scr. Mater. 50
(2004) 12331236.
20) J. Yan, Z. Xu, Z. Li, L. Li and S. Yang: Scr. Mater. 53 (2005) 585589.
21) S. A. Khodir and T. Shibayanagi: Mater. Trans. 48 (2007) 25012505.
22) T. Morishige, A. Kawaguchi, M. Tsujikawa, M. Hino, T. Hirata and
K. Higashi: Mater. Trans. 49 (2008) 11291131.

FeSiBPCu Nanocrystalline Soft Magnetic Alloys with High Bs of 1.9 Tesla

Produced by Crystallizing Hetero-Amorphous Phase
Akihiro Makino, He Men* , Takeshi Kubota, Kunio Yubuta and Akihisa Inoue
Technically important nanocrystalline soft magnetic alloys and their derivatives always include metal elements such as Nb, Zr, Mo, etc.
and/or Cu to realize the nanostructure, which results in a remarkable decrease of saturation magnetic ux density (Bs ) and a signicant increase
in material cost. With the aim to solve the serious problem, we successfully developed new FeSiBPCu nanocrystalline soft magnetic alloys. The
melt-spun Fe83:3{84:3 Si4 B8 P3{4 Cu0:7 (at%) alloys have heterogeneous amorphous structures including a large amount of -Fe clusters, 23 nm in
size, due to the unusual eect of the simultaneous addition of the proper amounts of P and Cu. The hetero-amorphous alloys exhibit higher Bs of
about 1.67 T than the representative amorphous and the nanocrystalline alloys, and the low coercivity (H c ) of 510 Am1 . A homogeneous
nanocrystalline structure composed of small -Fe grains with a size of about 10 nm can be realized by crystallizing the hetero-amorphous alloys.
The nanocrystalline alloys show extremely high Bs of 1.881.94 T almost comparable to the commercial Fe-3.5 mass%Si crystalline soft
magnetic alloys, and low H c of 710 Am1 due to the simultaneous realization of the homogeneous nanocrystalline structure and small
magnetostriction of 2{3 106 . In addition, the alloys have a large economical advantage of lower material cost and better productivity than the
ordinary soft magnetic nanocrystalline alloys now in practical use. [doi:10.2320/matertrans.MER2008306]
(Received September 2, 2008; Accepted November 5, 2008; Published December 17, 2008)
Keywords: soft magnetic material, high saturation magnetic ux density, amorphous alloy, nanocrystalline alloy
1.
Introductions
Nanocrystalline soft magnetic alloys13) produced by

crystallizing amorphous phases attracted great attention due
to their excellent soft magnetic properties and rather high
magnetizations. The developed nanocrystalline soft magnetic
alloys and their derivatives always include a large amount of
metal elements such as Nb, Zr, Mo, etc. and/or Cu to realize
the nanocrystallized structure.48) It is well-known that the
uniform nanocrystalline structure with excellent magnetic
softness can be obtained by an amorphous precursor including some chemical or topological heterogeneities induced by
the above-mentioned metal elements.9,10) During heating, an
apparent short-medium range ordering11) or a nanoscale
phase separation12) in the amorphous matrix raises from the
heterogeneity, which acts as the nucleation site for the
primary crystal and leads to the renement of the grains.
Thus, the precursor for nanocrystallized structure is needed to
be an amorphous phase essentially including the metal
elements, however, which cause a remarkable decrease of the
saturation magnetic ux density (Bs )13) and a signicant
increase of the material cost. Considering current energy
problems, higher Bs accompanied with excellent magnetic
softness is strongly required for the magnetic materials used
in electrical power supplies.
We developed new Fe83:3{84:3 Si4 B8 P3{4 Cu0:7 (at%) nanocrystalline soft magnetic alloys with much higher Bs of
around 1.9 T than those of the representative amorphous, the
ordinary13,5,8) and the recently reported14,15) nanocrystalline
alloys, almost comparable to Fe-3.5 mass%Si crystalline
alloys16) now in practical use.
2.
B (99.5 mass%), Cu (99.99 mass%) and premelted Fe-P

(99.9 mass%) in a high purity Argon atmosphere. Fe78 Si9 B13
alloy ingot was also prepared by the same method for
comparison. A single-roller melt-spinning method in air was
used to produce the rapidly solidied tapes with about 1
3 mm in width and approximately 20 mm in thickness. The
alloy compositions are nominally expressed since the dierence between nominal and chemical analyzed composition
was negligibly small. The structure was identied by X-ray
diractometry (XRD) and transmission electron microscopy
(TEM). The mean grain size of the -Fe phase was estimated
by using Scherrers equation from the full width at half
maximum of the bcc(110) reection peak from for the
specimens. Thermal property of melt-spun alloys was
evaluated with a dierential scanning calorimeter (DSC) at
a heating rate of 0.67 C/s under an Argon ow. The
melt-spun specimens were subjected to annealing for 600
seconds at various temperatures in a vacuum atmosphere
under the non-magnetic eld. Saturation magnetization (J s )
and coercivity (H c ) under a maximum applied eld of
800 kAm1 and 2 kAm1 were measured by a vibrating
sample magnetometer (VSM) and a dc B-H loop tracer,
respectively. Eective permeability (e ) at 1 kHz and core
loss (W) at 50 Hz for were measured with a vector impedance
analyzer under a eld of 0.4 Am1 and an ac B-H analyzer
operated under sinusoidal input voltage, respectively, by
using the wounded toroidal cores with a diameter of about
20 mm. The W and e should be inferior to those measured
by using a single sheet and a ring sample. Magnetostriction
was measured in an applied eld up to 800 kAm1 by using
a strain gauge. Density was measured by the Archimedean
method with n-tridecane.
3.
FeSiBPCu alloy ingots were prepared by inductionmelting mixtures of Fe (99.98 mass%), Si (99.998 mass%),
*Graduate
Student, Tohoku University
3.1 Microstructure of as-quenched alloys

Figure 1 shows the dependence of -Fe grain size on the
substitutional amounts of P for B and Cu for Fe in an as-
FeSiBPCu Nanocrystalline Soft magnetic Alloys with High Bs of 1.9 Tesla Produced by Crystallizing Hetero-Amorphous Phase
Fig. 1 Dependence of -Fe grain size on substitutional amounts of P for B

and Cu for Fe in Fe82y Si9 B9x Px Cuy in an as-quenched state.
205
Fig. 2 XRD proles for as-quenched (a) Fe82 Si9 B9 and (b) Fe81:7 Si9 B7 P2 Cu0:3 alloys.
Fig. 3 (a) HRTEM image and (b) SAED pattern for as-quenched Fe81:7 Si9 B7 P2 Cu0:3 alloy.
quenched Fe82 Si9 B9 alloy with higher Fe content than the

limit of about 80% for the formation of a single amorphous
structure. The grain size is estimated from a sharp diraction
peak at about 45 corresponding to -Fe phase in the XRD
proles or the high resolution TEM (HRTEM) images. The
alloy without P and Cu has the largest -Fe grains with the
size of 93 nm estimated from XRD prole, shown in
Fig. 2(a). The grain size decreases with the P and/or Cu
addition, and reaches to the minimum values of around 3 nm
smaller for the alloys simultaneously containing 24 at%P
and 0.3%Cu. The grain size for the alloys with the proper
amounts of P and Cu is much smaller than that for the
recently reported the FeCuB(Si) alloys.14) The XRD prole,
the HRTEM image and the selected area electron diraction
(SAED) pattern of the as-quenched Fe81:7 Si9 B7 P2 Cu0:3 alloy
are shown in Fig. 2(b), Fig. 3(a) and (b), respectively. The
SAED pattern shows diused diraction rings and no spots
corresponding to any crystalline phase in agreement with the
XRD results shown in Fig. 2(b). However, the HRTEM
image reveals that the as-quenched structure of the alloy is
not fully amorphous in disagreement with the XRD and the

SAED patterns, but consists of an extremely small crystalline-like phase with about 3 nm or smaller in diameter,
randomly dispersed within the amorphous matrix. Judging
from the lattice fringes as indicated by white bars in Fig. 3(a),
corresponding to the bcc (110) plane distance of -Fe
crystalline, the phase is completely revealed to be -Fe
crystal. It is understandable that the XRD and the SAED
patterns are not able to reveal the existence of the nanocrystals due to their very small size and low volume
fraction.17) Therefore, we can notice that the coarse -Fe
grains formed within the as-quenched amorphous matrix of
Fe82 Si9 B9 alloy change to the nano-sized -Fe phase with
3 nm or smaller in size by the simultaneous substitution of
24%P for B and 0.3%Cu for Fe. With deviation from these
compositions with the P and the Cu contents, the grain size
rapidly increases.
We discuss on the origin of the remarkable eect of the
simultaneous addition of P and Cu on the renement of the Fe crystals formed in the amorphous matrix of the alloys with
206
A. Makino, H. Men, T. Kubota, K. Yubuta and A. Inoue
Fig. 4 Compositional dependence of H c on P and Cu contents for asquenched Fe82y Si9 B9x Px Cuy (x: 06, y: 00.4) alloys.
higher Fe content than the limit for formation of a single

amorphous phase. Taking account of the result that the eect
is observed only for the simultaneous addition of the proper
amounts of P and Cu, the mixing enthalpy (H)18) between
the constituent elements is considered. It is noted that H is
positive (+13 kJ/mol) between Fe and Cu and negative
(9 kJ/mol) between Cu and P, suggesting that there is a
repulsive and an attractive interactions existing between Fe
and Cu, and Cu and P atoms, respectively. Therefore, during
melt-spinning, an extremely small region including enriched
Cu and P elements could separate from the Fe-Si-B-(P)
amorphous phase and possibly act as the nucleation sites for
-Fe clusters, which should result in the renement of the
grains. The number and the size of the Cu- and P-rich region
possibly change with the amounts of P and Cu. The
simultaneous addition of 24%P and around 0.3%Cu might
realize the smallest size and the largest number of the region,
which is considered to be most suitable condition for
decreasing the -Fe precipitates.
Figure 4 shows the compositional dependence of H c on P
and Cu contents for the as-quenched Fe82y Si9 B9x Px Cuy
(x: 06, y: 00.4) alloys. The dependence coincides with the
change in the -Fe grain size with P and Cu addition shown in
Fig. 1. The alloy without the addition of P and Cu has the
largest H c of about 45 Am1 responsible to the coarse -Fe
grains. The H c decreases with decreasing the grain size, and
reaches the minimum value of 78 Am1 for the heteroamorphous alloys with 24%P and around 0.3%Cu.
The similar eect of the simultaneous addition of P and
Cu was observed for the alloys with higher Fe content
up to 85.7%. Figure 5 shows the XRD proles for
Fe83:3 Si4 B8 P4 Cu0:7 (A alloy), Fe84:3 Si4 B8 P3 Cu0:7 (B
alloy) and Fe85:7 Si4 B9 P1 Cu0:3 (C alloy), indicating an
amorphous structure with small H c of 7, 8 and 11 Am1 ,
respectively. Whereas, the HRTEM observation revealed that
the alloys have an as-quenched hetero-amorphous structure
with a large amount of -Fe clusters, as shown in Fig. 6 for
A alloy as an example. The hetero-amorphous alloys have
higher Bs of about 1.67 T than the FeSiB amorphous alloy
now in practical use. Moreover, all the alloys shown in
Fig. 5 XRD proles for as-quenched Fe83:3 Si4 B8 P4 Cu0:7 (A), Fe84:3 Si4 B8 P3 Cu0:7 (B) and Fe85:7 Si4 B9 P1 Cu0:3 (C) alloys.
Fig. 6
HRTEM image of as-quenched Fe83:3 Si4 B8 P4 Cu0:7 alloy.
Fig. 1 and hetero-amorphous A, B and C alloys are

conrmed to be completely ductile in three-point bending
test, which is important for applications.
3.2
Microstructure and magnetic properties of crystallized alloys from hetero-amorphous phase

Crystallization behavior was examined by DSC for the
hetero-amorphous A and B alloys, shown in Fig. 7. The
alloys crystallize through two exothermic peaks and large
temperature intervals of 168 C and 135 C between each one,
respectively. It was conrmed by XRD that the rst and the
second peaks on the curves result from the phase transition
from the amorphous to -Fe and then to some compounds,
respectively. It has been already reported in the representative nanocrystalline soft magnetic alloys Nanoperm2,19)
that the amorphous precursor with a large temperature
Fig. 7 DSC curves for as-quenched Fe83:3 Si4 B8 P4 Cu0:7

Fe84:3 Si4 B8 P3 Cu0:7 (B) alloys.
Fig. 9
(A) and
207
Fig. 8 XRD proles for Fe83:3 Si4 B8 P4 Cu0:7 (A) and Fe84:3 Si4 B8 P3 Cu0:7
(B) alloys.
Bright eld image and SAED patterns of nanocrystalline Fe83:3 Si4 B8 P4 Cu0:7 alloy.
interval can change to a uniform nanocrystalline structure

without any compounds with large magnetocrystalline
anisotropy by annealing within the temperature range
between the two peaks.
Figure 8 shows the XRD proles for A and B alloys
crystallized at 475 C and 450 C, respectively, between the
temperatures of the two peaks on the DSC curves. The
structures were found to consist of -Fe phase, whose grain
sizes were estimated to be 11 nm and 17 nm for A and B
alloys, respectively. The SAED prole and the TEM image
of nanocrystalline A alloy are shown in Fig. 9. A uniform
nanocrystalline structure with the grain size of approximately
10 nm, which is in almost agreement with the estimated value
from the XRD prole, is successfully formed presumably
because the cluster formed at an as-quenched state could act
as the nucleation site for the -Fe grains during the
crystallization. The structure is very similar to the previously

developed representative nanocrystallized alloys with excellent magnetic softness1,2,20,21) in spite of not-containing a
large amount of the metal elements.
Figure 10 shows the hysteresis loop the nanocrystalline
A alloy. The alloy shows the extremely high Bs of 1.88 T
due to the high Fe content and almost not-containing any
metal elements except for Fe. The Bs is higher than those of
the amorphous and the previously reported nanocrystalline
alloys,13,13,15) and slightly lower than the Fe-3.5 mass%Si
crystalline soft magnetic alloys.16) A alloy also exhibits
the low H c of 7 Am1 and the high e of 25,000 due to the
simultaneous realization of the homogeneous nanocrystalline
structure and the much smaller magnetostriction () of 2
106 than that of 27 106 of Fe78 Si9 B13 amorphous
alloy.
208
A. Makino, H. Men, T. Kubota, K. Yubuta and A. Inoue
These nanocrystalline alloys should make a contribution to

energy saving though their low core losses and improve the
conversation of earth resources. In addition, the alloys have a
large economical advantage of low material cost and good
productivity because of not-containing expensive and easily
oxidized metal elements which has been considered to be
essential for the nanocrystalline soft magnetic alloys.
4.
Fig. 10 Hysteresis loop for the nanocrystalline Fe83:3 Si4 B8 P4 Cu0:7 alloy
measured by a VSM. Insert shows enlarge part of near zero eld measured
by a dc B-H tracer.
Fig. 11 Core loss (W) of nanocrystalline Fe83:3 Si4 B8 P4 Cu0:7 alloy as a

function of maximum magnetic ux density (Bm ). The data of Fe78 Si9 B13
amorphous alloy annealed at 380 C for 120 min are also shown for
comparison.
Core loss (W) is important for the materials in power

devices. Figure 11 shows the W at 50 Hz of the nanocrystalline A alloy as a function of maximum magnetic ux
density (Bm ), in comparison with the data of the optimal
annealed Fe78 Si9 B13 amorphous alloy. Comparing to the
commercial non-oriented and the oriented Fe-3 mass%Si, W
at Bm of 1 T for the nanocrystalline alloy is much smaller
than those measured by using punched ring samples for the
Fe-Si alloys.22) Although the W of the nanocrystalline and the
amorphous alloys increase with increasing Bm , the nanocrystalline alloy exhibits the superior W to the amorphous
alloy over the whole Bm range. The rapid increase of W takes
place at higher Bm of 1.7 T for the nanocrystalline alloy than
that of 1.4 T for the amorphous alloy. Here, the very low W in
the Bm range up to 1.7 T of the nanocrystalline alloy should
be an outstanding feature never seen in the other soft
magnetic materials.
The nanocrystalline B alloy with higher Fe content have
a high Bs of 1.94 T comparable to Fe-3.5 mass%Si alloy, and
accompanied with low H c of 10 Am1 due to small of
3 106 .
Conclusions
Structure and magnetic properties of as-quenched and

crystallized Fe-rich FeSiBPCu alloys were studied. The
uniform nanocrystalline structure composed of -Fe grains is
found to be realized by annealing the hetero-amorphous
structure with -Fe clusters formed by an unusual eect of
the simultaneous addition of P and Cu. The conclusions
obtained are summarized as follows.
(1) The simultaneous substitution of 24%P for B and
0.3%Cu for Fe signicantly and drastically decreases
the size of -Fe grains within the as-quenched amorphous matrix of an as-quenched Fe82 Si9 B9 alloy from
93 nm to 3 nm or smaller. That renement eect was
also observed for the alloys with higher Fe content up to
85.7%.
(2) The melt-spun Fe83:3{84:3 Si4 B8 P3{4 Cu0:7 alloys with
heterogeneous amorphous structures including a large
number of -Fe clusters exhibit higher Bs of about
1.67 T than the representative Fe78 Si9 B13 amorphous
alloy, and the low H c of 510 Am1 .
(3) The uniform nanocrystallized structures from the
hetero-amorphous alloys consist of small -Fe grains
with about 10 nm evaluated from the TEM image and
17 nm estimated from the XRD prole in size for
Fe83:3 Si4 B8 P4 Cu0:7 and Fe84:3 Si4 B8 P3 Cu0:7 , respectively. With the structural change from the amorphous to
the nanocrystalline structure, the Bs rapidly increases
from 1.67 T to 1.881.94 T.
(4) The Bs of the nanocrystalline alloys is considerably
higher than those of any soft magnetic amorphous and
nanocrystalline alloys previously reported.
(5) The Fe83:3 Si4 B8 P4 Cu0:7 nanocrystalline alloy exhibits
excellent magnetic softness, H c of 7 Am1 and e of
25,000 at 1 kHz, due to the simultaneous realization of
the uniform nanostructure composed by small -Fe
grains with about 10 nm in size and a small magnetostriction of 2 106 .
(6) The Fe83:3 Si4 B8 P4 Cu0:7 nanocrystalline alloy exhibits
the extremely lower W over the whole Bm range than
the Fe78 Si9 B13 amorphous alloy. The rapid increase of
W takes place at higher Bm of 1.7 T for the nanocrystalline alloy than that of 1.4 T for the amorphous
alloy.
REFERENCES
1) Y. Yoshizawa, S. Oguma and K. Yamauchi: J. Appl. Phys. 64 (1988)
60446046.
2) K. Suzuki, A. Makino, N. Kataoka, A. Inoue and T. Masumoto: Mater.
Trans. JIM 32 (1991) 93102.
3) M. A. Willard, D. E. Laughlin, M. E. McHenry, D. Thoma, K. Sickafus,
J. O. Cross and V. G. Harris: J. Appl. Phys. 84 (1998) 67736777.
4) Y. Yoshizawa and K. Yamauchi: Mater. Trans. JIM 31 (1990) 307
314.
5) A. Makino, T. Hatanai, Y. Naitoh, T. Bitoh, A. Inoue and T. Masumoto:
IEEE Trans. Magn. 33 (1997) 37973798.
6) A. Makino, A. Inoue and T. Masumoto: Mater. Trans. JIM 36 (1995)
924938.
7) K. Suzuki, N. Kataoka, A. Inoue and A. Makino: Mater. Trans. JIM 31
(1990) 743746.
8) K. Suzuki, A. Makino, A. Inoue and T. Masumoto: Sci. Rep. RITU A39
(1994) 133140.
9) J. D. Ayers, V. G. Harris, J. A. Sprague, W. T. Elam and H. N. Jones:
Acta Mater. 46 (1998) 18611874.
10) Y. Zhang, K. Hono, A. Inoue, A. Makino and T. Sakurai: Acta Mater.
44 (1996) 14971510.
11) M. Nakamura, Y. Hirotsu, K. Anazawa, A. Makino and A. Inoue:
Mater. Sci. Eng. A 179/180 (1994) 487490.
12) M. Ohnuma, K. Hono, H. Onodera, J. S. Pedersen and S. Linderoth:
NanoStructured Mater. 11 (1999) 693696.
209
13) M. E. McHenry, M. A. Willard and D. E. Laughlin: Prog. Mater. Sci. 44

(1999) 291433.
14) M. Ohta and Y. Yoshizawa: Jpn. J. Appl. Phys. 46 (2007) L477L479.
15) M. Ohta and Y. Yoshizawa: Appl. Phys. Lett. 91 (2007) 062517
(3pages).
16) A. Makino, K. Suzuki, A. Inoue and T. Masumoto: Mater. Trans. JIM
32 (1991) 551556.
17) K. Hono, D. H. Ping, M. Ohnuma and H. Onodera: Acta Mater. 47
(1999) 9971006.
18) A. Takeuchi and A. Inoue: Mater. Trans. 46 (2005) 28172829.
19) K. Suzuki, M. Kikuchi, A. Makino, A. Inoue and T. Masumoto: Mater.
Trans. JIM 32 (1991) 961968.
20) A. Makino, T. Bitoh, A. Inoue and T. Masumoto: Scr. Mater. 48 (2003)
869874.
21) A. Makino, M. Bingo, T. Teruo, K. Yubuta and A. Inoue: J. Appl. Phys.
101 (2007) 09N117 (3pages).
22) Y. Takada, M. Abe, S. Masuda and J. Inagaki: J. Appl. Phys. 64 (1988)
53675369.

Microstructure Renement and Texture Evolution of Titanium

by Friction Roll Surface Processing
Meiqin Shi1; *1 , Yoshimasa Takayama1; *2 , Tomohiro Umetsu1; *1 ,
Hajime Kato1 , Hideo Watanabe1 and Hirofumi Inoue2
1
Department of Mechanical and Intelligent Engineering, Graduate School of Engineering,
Utsunomiya University, Utsunomiya 321-8585, Japan
2
Department of Materials Science, Graduate School of Engineering, Osaka Prefecture University, Sakai 599-8531, Japan
Friction roll surface processing (FRSP), as a novel severe plastic deformation technique, has been put forward and experimentally
investigated. Commercial pure titanium sheets were subjected to FRSP and subsequent annealing. An extremely large strain was imposed on the
surface layer of the sheet. Consequently, a uniform ultrane-grained microstructure with a grain size of approximately 80 nm and a deformation
texture were obtained in the surface layer. The FRSP texture was found to be strongly related to the processing direction. The evolution of
texture during annealing was also investigated. The texture analysis revealed that new preferred orientations could be developed during FRSP.
[doi:10.2320/matertrans.MER2008294]
(Received August 28, 2008; Accepted November 6, 2008; Published December 17, 2008)
Keywords: texture, titanium, severe plastic deformation, electron backscatter diraction, friction roll surface processing
1.
Introduction
Over the last decade, severe plastic deformation (SPD) has

been successfully applied to fabricate ultrane-grained
(UFG) materials.1) Extensive investigations showed the
prospect of variety of SPD techniques on the point of
developing nanostructured materials.2) In addition, many
mechanical and physical properties of crystals are anisotropic, and therefore the properties of a polycrystalline
aggregate will depend on whether the individual grains or
subgrains, which comprise the sample, are randomly oriented
or tend to have some preferred crystallographic orientation.3)
Moreover, the preferred orientation often appeared as a
consequence of deformation. This implies that preferred
orientation is supposed to be formed on the metal by SPD.
However, attention to this possibility of SPD technique,
which is controlled by crystallographic orientation, has not
been well directed. On the other hand, some of material
properties, such as corrosion, formability and wear, are
strongly related to the microstructure of surface layer. From
this point of view, there has recently been more attention paid
to the microstructure control on the surface layer of the
material.
This study presents a new SPD technique named friction
roll surface processing (FRSP). It has been found that FRSP
and subsequent annealing have an excellent possibility for
controlling both microstructure and texture in the surface
layer of the materials. During FRSP, regarded as surface
friction processing, a large amount of plastic strain is
imposed into samples by the mechanical rolling action of a
high speed rotating tool. As a result, not only grain
renement, but also formation of a crystallographic texture
can be achieved.
*1Graduate
Student, Utsunomiya University

author, E-mail: takayama@cc.utsunomiya-u.ac.jp
*2Corresponding
2.
Commercial purity (CP) titanium sheet (grade 1) with an

average grain size of 30 mm, containing impurities (mass%)
of 0.041 O, 0.002 N, 0.0024 H, 0.031 Fe and 0.009 C, was
used as the starting material. Three types of samples, marked
as RD, TD and RD+45 with size of 60 mm 20 mm
1 mm, were machined from the titanium sheet. FRSP was
conducted with the processing direction parallel to the rolling
direction (RD), transverse direction (TD) and the direction
rotated by 45 degrees from RD to TD, respectively. The rolls
rotation direction, samples feeding direction, FRSP direction and sample orientations are sketched in Fig. 1. A xed
roll with a dimension of 70 10 mm and made of tool
steel SK3, was rotated at 48 rpm. The indentation, dened as
the depth at which the rotating roll was pressed down into the
sample, was selected as 0.2 mm. The feeding speed of sample
was set as 14 mm/min. FRSP was performed as follows: rst,
the interface where the roll and the sample just touched was
located, and this was regarded as the reference plane at which
the indentation was zero. Then, the roll was placed within the
given indentation without contacting with the sample. After
rotating the roll, the sample was moved forward until the
Roll
(b) RD
ND
(c)RD+45
ND
(d) TD
ND
TD
RD
TD
RD
TD RD
(a)
Fig. 1 Schematic illustration of Friction Roll Surface Processing: (a)

Indication of roll, feeding direction and FRSP direction; Conguration and
name of the samples, (b) RD sample, (c) RD+45 sample and (d) TD
sample.
Microstructure Renement and Texture Evolution of Titanium by Friction Roll Surface Processing
entire surface was treated. This procedure resulted in making

the working plane lower than the reference plane and led to a
great amount of friction. After FRSP, the samples were
annealed at 823 K for 1 h in Ar gas and air cooled to study the
texture evolution.
The imposed strain was examined on the ND-RD plane of
RD sample (i.e. normal direction (ND)-rolling direction (RD)
plane). The microstructure of FRSPed sample was observed
by using optical microscopy and transmission electron
microscopy (TEM). Global textures were measured as
incomplete pole gures for (0002) and (101 0) diraction
planes using CuK radiation X-ray diraction (XRD) to
reveal the textures of deformed materials. Evolutions of
crystallographic texture of the titanium samples, after FRSP
and annealing, were then evaluated using a scanning electron
microscope/electron back scatter diraction pattern technique (SEM/EBSD; HITACHI S-3500H, TSL Orientation
Image Microscopy system) after electropolishing. In order to
depict the gradient of microstructure and texture through the
thickness, the cross sections (i.e. ND-TD plane of RD sample
or ND-RD plane of TD sample) were scanned continuously
every 200 mm from the processed surface to center part
using a step size of 1 mm.
3.
3.1 Strain distribution

For the purpose of evaluating the strain distribution
through the thickness of the FRSPed samples, shear strain
measurements were carried out using the embedded-pin
method.4) A cylinder-shaped pin of 2 mm in diameter and
1 mm in thickness and made of the same material, was
embedded at the center of the sample. The pin was supposed
to be deformed to the same extent as the whole sample. After
FRSP, the longitudinal plane (i.e. ND-RD plane of RD
sample) of the pin was observed by optical microscopy and
strain was calculated using the bent line of pin-matrix
interface according to the denition of shear strain. An
example of the result is displayed in Fig. 2 and listed in
Table 1. This optical micrograph of pin-matrix interface
reveals that FRSP generates large plastic strain that can reach
about 100 mm beneath the surface. These strains fall o
rapidly through the thickness and give a strain gradient
dependence based on depth. It was very dicult to measure
RD
Feeding direction
ND
100m
Fig. 2 Optical micrograph of pin-matrix interface in the titanium sheet

processed after FRSP.
211
Table 1 Shear strain through the thickness.

Distance from
surface
0
tan
50 mm
100 mm
9.35
1.12
0.95
200 mm
0.30
TD
Feeding direction
ND
50m
Fig. 3 Optical micrograph showing material ow of TD sample after
FRSP.
the strain just on the surface because the pin-matrix interface

could not be distinguished in the vicinity of the surface. The
strain of 9.35 measured on the surface should be regarded as
the distinguishable maximum value, that is, an underestimated one. Therefore, the strain imposed by FRSP is
probably much higher than a strain of about 1 by one pass
of the other SPD techniques for bulk materials such as
ECAP5) or ARB.6)
3.2 Rened microstructure
The optical micrograph on the ND-TD plane of the TD
sample after FRSP is shown in Fig. 3. Material ow
corresponding to the strain gradient appears evident in the
vicinity of the surface. Dierent deformation modes7,8) are
observed depending on the distance from the processed
surface. In the surface layer, to a depth of about 50 mm,
lamellar microstructures with an inclination about 2025 to
TD are found to be proof of the material ow. In the second
layer, from a depth of 50 mm to 200 mm from the surface,
there are many deformation twins are formed, which implies
a low strain of deformation. In rolled titanium deformation
twinning was found in all grains at a very early stage.9) The
absence of additional twinning during the large deformation
of titanium has been discussed as being the result of the
reduced grain size introduced by prior twinning.7) For the
center layer, equiaxed grains without twinning are observed.
This heterogeneous microstructure through thickness of
sample is consistent with the strain gradient formed after
FRSP displayed in Fig. 2. Similar result was found during
the general cold rolling. Twin was observed in some grains
of the titanium sheet after 30% cold rolling. Additional coldrolling led to the fragmentation of lamellae as a result of slip
212
M. Shi et al.
grain subdivision continued to rene with the increase of

strain. This process led to the formation of new ultra-ned
grains.
TD
ND
100 nm
Fig. 4
TEM bright-eld micrograph of RD sample after FRSP.
activity and multiple twinning.10,11) TEM observation taken

in the surface layer is provided in Fig. 4 to illustrate the
elongated nanostructures with 110 nm in TD and 50 nm in
ND. The microstructure in this zone is fragmented down to
the sub-micrometer scale due to the severe strain. Grain
boundaries are well discerned and a high density of
dislocation can be observed within individual grains. These
lamellar structures and accumulated dislocations were
observed in other heavily deformed materials processed by
dierent SPD technique such as ECAP and ARB. However,
the formation mechanism of the UFG materials during SPD
is still an issue under discussion.12) On the other hand,
sliding friction under a large load is especially eective for
the formation of surface nanocrystallization, since it can
cause a large shear stress, which is a necessary condition for
the generation of severe plastic deformation. The nanocrystalline material was generated by processes similar to
those involved in the early stages of mechanical alloying,
commonly done in a high energy ball mill.13) It can be
suggested that during FRSP without lubrication a large
amount of shear strain, owing to large friction between the
roll and the sample, was introduced in the surface region.
The grains became sheared resulting in an oblate grain
structure with the long axis of grain inclined to the FRSP
direction. Moreover, grain fragmentation set in14) and the
(a) As received sample
3.3 Texture evolution

3.3.1 Deformation texture
The normalized pole gures of the as received sample,
RD and TD samples after FRSP measured by XRD are
presented in Fig. 5. As displayed in Fig. 5(a), (0001) pole
gure of the as received one indicates that the basal poles are
tilted 3040 away from the ND toward the TD, thereby
exhibiting a TD-split texture (f1 21 4gh101 0i or f0225 gh21 1 0i
(frolling planeghrolling directioni)) that is commonly observed in rolled titanium sheets after annealing. After FRSP,
the (0001) pole gure of the RD sample shows that the
intensity of (0001) pole spreads out between ND and TD in
Fig. 5(b). The initial TD-split texture is extended from the
basal texture component to higher tilted angle. In addition,
the peak is also concentrated with a spread angle about TD
ranging from 13 to +13 for the as received one, while for
the RD sample after FRSP, a spread angle ranging from 9
to +9 about TD. It is also worth mentioning that not h101 0i,
but near h112 0i, is aligned along RD in the (101 0) pole gure
of the RD sample (Fig. 5(b)). This is dierent from the
texture with h101 0i parallel to RD, which is well known for
titanium subjected to cold-rolling.10) Thus, FRSP has a
dierent deformation mechanism from general cold-rolling
and this is possibly due to the severe shear strain imposed
during FRSP. Figure 5(c) describes the main textures in the
TD sample after FRSP. It is interesting to note that the
resulting texture in the TD sample has the same feature as
that of the RD sample except that it was rotated 90 about ND
from RD to TD. Therefore, the characteristic texture was
formed in the sample by FRSP, and the new FRSP texture
was strongly related to the processing direction.
3.3.2 Annealing texture
Figure 6 shows (0001), (101 0) and (21 1 0) pole gures
obtained from EBSD data of RD, RD+45 and TD samples
annealed at 823 K for 1 h after FRSP. The as received
sample is also presented for comparison. Analyzed ranges of
RD, RD+45 and the TD samples were set in the processed
(b) RD sample
(c) TD sample
(0001)
(1010)
Fig. 5 Deformed texture of RD and TD sample after FRSP. (a) As received sample; (b) RD sample; (c) TD sample. Contour levels of
(0001) pole gures: 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0, 4.5, 5.0; Contour levels of (101 0) pole gures: 0.8, 1.0, 1.2, 1.4, 1.6, 1.8.
Microstructure Renement and Texture Evolution of Titanium by Friction Roll Surface Processing
(a)
(b)
(c)
213
(d)
(b)
(c)
(d)
(e) Indication of FRSP
directions in (b), (c) and
(d) samples.
Max= 5.83
Max= 7.50
Max= 5.21
Max= 6.45
Fig. 6 Pole gures of (a) As received sample (analyzed range: 01000 mm); and (b) RD sample; (c) RD+45 sample; (d) TD sample after
annealing, (sample (b), (c) and (d), analyzed range: 050 mm from the processed surface).
surface regions with 50 mm thick in ND while the range of as

received sample was a whole thickness of 1000 mm in ND.
As displayed in Fig. 6(a), (0001) pole gure of the as
received one shows a typical TD-split texture, which is
consistent with the XRD analysis (Fig. 5(a)). Pole gures of
RD, RD+45 and TD samples present distinct textures after
FRSP and following annealing. Figure 6(b) displays a
similar texture to the as received one in the RD sample.
However, the basal pole density that was observed around
ND in deformed state (Fig. 5(b)) disappears after annealing.
In the TD sample shown in Fig. 6(d), the original TD-split
texture in the as received one is completely replaced by RDsplit texture rotated 90 about ND and the basal poles are
tilted 2030 away from the ND toward the RD. This
emerging texture had a similar maximum intensity to that of
the TD-split in the as received one. Dependence of textural
evolution on FRSP direction was conrmed by the result in
the case of RD+45 sample (Fig. 6(c)). A resultant texture
was rotated 45 about the normal axis compared with RD
sample. Compared with the as received one, the peaks
shifted along FRSP direction are found in (101 0) pole
gures of RD, RD+45 and TD samples. This might suggest
that FRSP is a unidirectional process and the annealing
texture corresponds with deformation texture (Fig. 5). These
results indicate that FRSP and subsequent annealing can
form a split texture reoriented on the sides of the FRSP
direction. The formation of preferred orientations was
already reported in various metals after sliding.1518)
Considerable plastic deformation was produced by friction
forces in the surface layers of metallic objects. This caused
a preferred orientation of the crystallites in a particular
direction with a large amount of strain. However, the
stability and component of textures were dierent according
to the metals, and the mechanism of texture formation was
not given or clearly explained. Heilmann et al.18) explained
that the special orientation was the result of the rotation of
the lattice. This rotation is about the transverse axis normal

to the sliding direction and parallel to the sliding interface.
After FRSP and following annealing, the distinguished
changes of textures in the surface layer of RD+45 and TD
samples suggest that FRSP produces the deformation texture
along the FRSP direction in the surface layer of samples
irrespective of initial texture. This can be possibly attributed
to three slip systems in titanium, which are f101 0gh112 0i,
f0001gh112 0i and f11 01gh112 0i. During FRSP, shear stress
between the roll and sample is the main cause of slips on the
planes. The slip direction h112 0i was always parallel to
the shear direction, i.e. FRSP direction, in RD, RD+45 and
TD samples, as displayed in (21 1 0) pole gures shown in
Fig. 6(b), (c) and (d). It is well known that in titanium, the
critical resolved shear stress for prismatic slip is lower than
those for basal and pyramidal slip at room temperature. In
the case of deformation by rolling, not only prismatic, but
also basal and pyramidal slip should be active.19) Therefore,
FRSP, which gives an extremely large strain, probably
develops a h112 0i shear texture due to the activation of all
the three slip systems in the progress.
In order to explore the eect of the strain gradient on the
texture evolution through the thickness, pole gures in each
200 mm layer of TD sample were also analyzed as displayed
in Fig. 7. Textures in TD samples vary remarkably depending on the distance from the processed surface. In the
surface layer, the symmetric peaks of RD-split texture were
conspicuously observed, whereas the center region without
deformation by FRSP showed a high intensity of TD-split
texture. The layer from 100 mm to 400 mm could be regarded
as a transient region where a newly formed mixed texture
coexisted with the initial one. The severe strain gradually
decreased from the surface to the center of the sample, and
dierent deformation modes were adopted to accommodate
to the dierent level of strain. The experimental texture
gradient is well consistent with these ndings.
214
M. Shi et al.
(d)
(c)
(b)
(a)
ND
FRSP direction
400 m
n
tio
ec
ss
os
TD
Cr
600 m
ed
SP ce
FR rfa
su
200 m
(e)
RD
Max= 6.03
Max= 7.47
Max= 8.84
Max= 13.4
Fig. 7 Pole gures of TD samples after annealing along the distance from processed surface. (a) 100 mm (b) 100200 mm (c)
200400 mm (d) 400600 mm; (e) Schematic illustration of cross section.
4.
Conclusions
In summary, during FRSP, a large plastic strain was

imposed in the surface layer and caused grain renement to
form the nanostructured material. The deformation mode was
dierent depending on the amount of induced strain and
developed a corresponding texture gradient. Characteristic
textures were formed in the surface of sample during FRSP
and subsequent annealing. The FRSP texture was strongly
related to the processing direction. After annealing the
preferred orientation remained distinct. These preliminary
studies have shown that FRSP was a process that allowed
ecient control of not only microstructure, but also texture in
the surface layer. This study provided a deeper understanding
of the evolution of structure and texture during severe plastic
straining.
REFERENCES
1) R. Z. Valiev: Nature Mater. 3 (2004) 511516.
2) Y. T. Zhu, T. C. Lowe and T. G. Langdon: Scr. Mater. 51 (2004) 825
830.
Annealing Phenomena, (Pergamon Press, Oxford, 2004) pp. 6779.
5888.

Mater. 35 (1996) 143.
6) Y. Saito, N. Tsuji, H. Utsunomiya and S. Tanigawa: Scr. Mater. 39
(1998) 1221.
7) Y. B. Chun, S. H. Yu, S. L. Semiatin and S. K. Hwang: Mater. Sci. Eng.
A 398 (2005) 209219.
8) Akhtar: Metall. Tran. A 6 (1975) 11051113.
Annealing Phenomena, (Pergamon Press, Oxford, 2004) pp. 3940.
10) N. Bozzolo, N. Dewobroto, H. R. Wenk and F. Wagner: J. Mater. Sci.
42 (2007) 24052416.
11) N. Bozzolo and F. Wagner: Proc. ICOTOM15 Ed. By A. D. Rollett,
(The American Ceramic Society, Pittsburgh, 2008) Symposium 9:
Hexagonal Metals on CD.
12) R. Z. Valiev, M. J. Zehetbauer, Y. Estrin, H. W. Hoppel, Y. Ivanisenko,
H. Hahn, G. Wilde, H. J. Roven, X. Sauvage and T. G. Langdon: Adv.
Eng. Mater. 9 (2007) 527533.
13) D. A. Rigney, X. Y. Fu, J. E. Hammerberg, B. L. Holian and M. L. Falk:
Scr. Mater. 49 (2003) 977983.
14) I. J. Beyerlein, R. A. Lebensohn and C. N. Tome: Mater. Sci. Eng. A
345 (2003) 122138.
15) H. Krause and A. H. Demirci: Wear 61 (1980) 325332.
16) H. Krause and E. Ocalan: Wear 108 (1986) 337342.
17) Z. N. Farhat: Wear 250 (2001) 401408.
18) P. Heilmann, W. A. T. Clark and D. A. Rigney: Acta Metall. 31 (1983)
12931305.
19) H. Inoue and N. Inakazu: Proc. ICOTOM8 Ed. By J. S. Kallend and
G. Gottstein, (The Metallurgical Society, Warrendale, 1988) pp. 997
1004.

RAPID PUBLICATION
Deoxidation Thermodynamics of -Titanium by Ca-CaCl2 and Ca-CaF2 Fluxes

Seung-Min Han1; * , Young-Seok Lee2 , Joo-Hyun Park3 , Good-Sun Choi4 and Dong-Joon Min1
1
Department of Metallurgical Engineering, Yonsei University, Seoul 120-749, Korea

Technical Research Laboratories, POSCO, Pohang 790-785, Korea
3
School of Materials Science & Engineering, University of Ulsan, Ulsan 680-749, Korea
4
Materials Development Group, Korea Institute of Geoscience and Mineral Resources, Daejeon 305-350, Korea
2
Deoxidation of -titanium was investigated by Ca-CaCl2 and Ca-CaF2 uxes at 900 to 1400 C. The content of oxygen in Ti was strongly
dependent not only on the Ca potential but on the stability of CaO in the uxes. The present experimental results indicate that deoxidation of Ti
by Ca-calcium halide uxes could be controlled by the activity ratio of CaO to Ca. The ratio of aCaO =aCa 0:060:1 was strongly
recommended for obtaining oxygen content of 40160 ppm. [doi:10.2320/matertrans.MRP2008284]
(Received August 19, 2008; Accepted October 9, 2008; Published November 27, 2008)
Keywords: titanium, deoxidation, calcium, halide ux, activity
Titanium has been used in novel applications because of its

inherent specic strength and corrosion resistance, which are
dependent on the concentration of oxygen in Ti. However,
oxygen in Ti has been known to be a controlling element
for the mechanical properties of Ti due to its strong anity
with Ti.1) However, one of the diculties with the oxygen
concentration in Ti is the lack of metallurgical options
available for deoxidation due to the large solubility and
strong anity of oxygen in Ti. Two concepts for oxygen
controlling for purication of Ti have been applied up to
now:2) electrochemical method311) and molten ux method.1218) Since the electrochemical method38) has been
applied to the production of sponge Ti, several studies have
reported the possibility for deoxidation of Ti by electrochemistry with molten halide salt.911)
On the other hand, recent studies have usually focused on
deoxidation of solid Ti by applying chemical rening
technology with molten ux in view of mass producing
process, energy eciency, etc.1218) However, these studies
for chemical rening technology also found that the content
of oxygen in Ti equilibrated with pure metallic calcium is
higher than the expected value in view of thermodynamics
because of local precipitation of CaO on the surface of Ti.
Based on those experimental results, Okabe et al.1216) and
Suzuki et al.17,18) proposed that deoxidizing eciency could
be increased by using molten uxes (Ca-CaO, Ca-CaCl2 ,
and Ca-BaCl2 ). These previous results concluded that the
removal of precipitated CaO on Ti surface plays an important
role in improving the eciency of deoxidation of Ti by using
Ca bearing salt. Even though these recent studies proposed
the possibility for deoxidation of Ti with Ca-calcium halide
ux, together with the deoxidation mechanism, the optimum
condition for the maximization of oxygen removal from Ti is
still not clearly understood.
Therefore, the present study focused on the eects of the
calcium halides (CaCl2 and CaF2 ) that might prove suitable
as a solvent for dissolution of CaO and temperature on
deoxidation of Ti. Furthermore, the eect of activity of
Ca and CaO in calcium halide uxes on the degree of
deoxidation was analyzed to propose the optimum condition
*Graduate
Student, Yonsei University
for minimizing the oxygen content in -titanium using the

thermochemical equilibration technique.
A SiC electric resistance furnace was used for equilibrating Ti located on a molybdenum basket with the Ca-CaOCaX2 (X Cl and F) ux charged in a capped iron crucible
(>99:9%, 27 mm outer diameter 20 mm inner diameter 37 mm height). The crucible assembly was maintained
in an Al2 O3 holder (52 mm outer diameter 48 mm inner
diameter 70 mm height) lled with alumina powder
(>99:9%) and then placed in the furnace for 12 h. Equilibration time had been considered 12 h after preliminary test
using 17 mol% Ca-CaCl2 ux at 900 C. The temperature was
controlled within 2 C using an R-type (Pt-13% Rh/Pt)
thermocouple and a proportional-integral-dierential controller. In order to guarantee the uniform initial content of
oxygen and nitrogen in solid Ti, the LECO standard
samples (Part No. 502201, 0.2670 (70) mass% O and
0.0170 (30) mass% N) were used. Calcium granules
(>99:9%), and dehydrated CaCl2 and CaF2 powders
(>99:9%), which were kept in vacuum desiccators to protect
them from air and moisture, were employed in the present
study. The CaO powder was calcined from CaCO3 of reagent
grade at 1000 C for 4 h. The experimental apparatus is shown
in detail in a previous article.19) After the experiment, the iron
crucible was quenched by an Argon gas stream. To remove
the residual chloride ux adhered to the Ti sample, the
samples were treated in diluted HCl (1:50) solutions at 70 C
for 1060 min. In the case of the uoride ux, the diluted HCl
(1:20) solutions were used at 70 C for 510 min. The
contents of oxygen and nitrogen in Ti were analyzed by an
inert gas fusion-infrared absorption method using LECO
TC-300. It was conrmed that the nitrogen content was less
than about 50 ppm.
The equilibrium concentration of oxygen in -titanium
equilibrated with calcium can be estimated as follows;
Ca(l) Oin Ti CaO(s)

aCaO
1

G1 2:303RT log
aCa fOin Ti mass%Oin Ti
583;000 88:5T [J/mol] (at 900{1100 C)12
562;000 91:6T [J/mol] (at 1050{1300 C)20
1
2
216
S.-M. Han, Y.-S. Lee, J.-H. Park, G.-S. Choi and D.-J. Min
Temperature (K)
-1.0
1773
-1.2
aCaO/aCa
-0.6
-1.6
-1.1
-1.6
-2.2
Ca-CaCl 2(-CaO) flux

at 900C
-2.8
1473
1373
1273
1173
(a)
1.0
0.8
0.6
0.4
Ca-CaCl 2 flux system

Ca-CaO equilibration
[12]
at 1173 to 1673 K
T.H.Okabe et al.
[22]
Sano and Tsukihashi

[12,22]
Ca-CaO Equil.
0.1
0.06
-2.0
-2.6
1573
0.2
-1.8
log[mass%O]in Ti
log [mass%O]in Ti
-1.4
-2.4
1673
-0.1
Ca-CaF2(-CaO) flux
at 1100C
at 1000C
at 1100C
at 1200C
at 1200C
at 1300C
at 1300C
at 1400C
at 1400C
-3.0
-1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0
0.2
0.4
-2.1
Casatd in Ca-CaCl2
[23]
after Staffansson and Sichen
-2.6
-0.1
aCaO/aCa
-0.6
-1.1
(b)
Ca-CaF2 flux system

at 1373 to 1673 K
Ca-CaO equilibration
[12]
T.H.Okabe et al.
1.0
0.8
0.6
0.4
[22]
Sano and Tsukihashi

[12,22]
Ca-CaO Equil.
0.2
-1.6
0.6
log(aCaO/aCa)
Fig. 1 Dependence of oxygen content in Ti on aCaO =aCa in the CaCaX2 (X Cl, F) ux system.
where aCaO and aCa are the activity of CaO and Ca,
respectively. fOin Ti is the Henrian activity coecient of
oxygen in -titanium, which can be assumed to be unity by
regarding the standard state as that of 1 mass% oxygen. Thus,
eq. (2) is rearranged as follows;

aCaO
G1
3
logmass%Oin Ti log
aCa
2:303RT
Equation (3) indicates that the content of oxygen in Ti has a
linear relationship with temperature and the ratio of aCaO to
aCa in the ux in logarithmic scale. In the present study, the
activity of CaO in ux was estimated from the activity of Ca
in Ca-CaX2 , which was calculated with equation of activity
coecient after Fishbach,21) and the concentration of oxygen
in Ti at a given temperature.
Deliberating that the phase diagram of Ca-CaX2 and CaOCaX2 binary systems proposed the maximum chemical
potentials of Ca and the minimum of CaO in each ux, the
optimization of chemical potentials between CaO and Ca in
the ux would be an important factor to determine the
eciency of deoxidation of Ti. If the chemical potential of
CaO in the ux was higher than the solubility, the
precipitated solid CaO lm on the surface of Ti would act
as a reaction barrier to achieve equilibrium.18) Therefore, the
optimization of chemical potential of Ca and CaO in the ux
plays an important role in progress the deoxidizing reaction
on the surface of Ti and thus in achieving the equilibrium
level of oxygen in Ti.
Figure 1 shows the relationship between the content of
oxygen in Ti and the activity ratio of CaO to Ca in Ca-CaCl2
and Ca-CaF2 ux systems, respectively. The content of
oxygen in Ti has a good linear relation with the activity ratio
of CaO to Ca in logarithmic scale at a xed temperature
(eq. 3) irrespective of ux type. Also, deoxidation of Ti
equilibrated with the Ca-CaCl2 ux is more eective than
that with the Ca-CaF2 ux system, and the pronounced
condition for the minimum concentration of oxygen in Ti by
0.1
-2.1
Ca satd in Ca-CaCl2
[23]
after Staffansson and Sichen
aCa = 0.5 0.8

(in Ca-CaF2) Assuming aCa = 1
-2.6
5.5
6.0
6.5
7.0
7.5
8.0
8.5
9.0
1/ T 10 (1/ K)
4
Fig. 2 Eect of CaX2 (X Cl, F) ux on deoxidation of Ti at 900 to

1400 C, (a) in case of Ca-CaCl2 ux, (b) in case of Ca-CaF2 ux.
CaCl2 ux system can be observed at 10001100 C. It may

have originated not only because the deoxidation of Ti by Ca
bearing ux is exothermic as shown in eq. (2), but also due to
the fact that the melting point of chloride ux system is lower
than uoride ux. The lowest levels of oxygen in Ti in CaCaCl2 and Ca-CaF2 ux were obtained near 40 ppm
(aCaO =aCa is about 0.06) at 1100 C and 160 ppm (aCaO =aCa
is about 0.15) at 1300 C, respectively.
Figure 2 shows the eect of temperature on the critical
content of oxygen in Ti equilibrated with Ca-CaX2 systems
and compared with the Ca-CaO system reported by Sano and
Tsukihashi22) and Okabe et al.,12) where the activity ratio of
CaO to Ca was xed to unity. The content of oxygen in Ti
equilibrated with the Ca-CaO system is dramatically decreased by introducing CaX2 at temperatures higher than 900 C.
The pronounced condition for the critical content of oxygen
in Ti in case of Ca-CaCl2 and Ca-CaF2 ux can be obtained at
aCaO =aCa 0:06 and 1.0, respectively. In Fig. 2(a), a dashed
line indicates a Ca saturated composition in the Ca-CaCl2
ux system, which was measured by Staansson and
Sichen,23) and this is in good accordance with the lowest concentration of oxygen in Ti at xed temperature. On the other
hand, a chain line for the Ca-CaF2 system in Fig. 2(b) is not
a Ca saturated composition in the Ca-CaF2 ux. The activity
of Ca on a chain line changes from about 0.8 to 0.5 with
increasing temperature. If the activity of Ca in the Ca-CaF2
ux should be increased to unity, the chain line for the CaCaF2 system would be lowered to a chain double-dashed line.
These complex behaviors of equilibrium oxygen in Ti can
be understood by taking the phase diagram into account as
follows. The phase diagram of the Ca-CaCl2 binary system
that had been determined by Eastmann et al.24,25) indicates
that the solubility of Ca exhibits a minimum at about 1000 C
when temperature increases to greater than 825 C. This
Deoxidation Thermodynamics of -Titanium by Ca-CaCl2 and Ca-CaF2 Fluxes
means that the activity coecient of Ca in the ux is maximized at 10001100 C. However, this phase diagram has a
discrepancy in the solubility from the result of Staansson
and Sichen. In the present study, the solubility of Ca in the
Ca-CaCl2 binary system is in good agreement with the phase
diagram of Stanansson and Sichen. On the other hand, the
solubility of CaO in the CaO-CaCl2 binary system26) steadily
increases with increasing temperature. This means that the
activity of CaO shows a positive deviation and decreases with
increasing temperature at a xed composition. Therefore, the
activity of Ca in the ux is a dominant factor aecting the
deoxidation limit of Ca-CaCl2 ux at 10001100 C.
In the case of a Ca-CaF2 ux system, the pronounced
condition of deoxidation is obtained at 12001300 C, which
corresponds to a maximum solubility of Ca from the phase
diagram of a Ca-CaF2 binary system.27) At this temperature
region, the CaO+-CaF2 solid solution could be formed from
the CaO-CaF2 binary phase diagram.28) Since the thermodynamic behavior of Ca and CaO in the Ca-CaO-CaF2 ternary
system has not been experimentally proved yet, further
studies will be required.
Figure 3 shows the eect of CaO concentration on
deoxidation eciency of Ti at 1000 C. The equilibrium
100
1.1
[23]
[26]
CaO saturated
1.0
0.9
10
Activity of Ca or CaO
0.8
0.7
0.6
0.5
At 1273 K
2.5-20 mol% Ca-CaCl2(-CaO)
5mol% Ca-CaCl2-CaO
0.1
0.4
0.3
[18]
Ca-CaCl2, Suzuki et al.
0.01
0.2
aCa in CaCl2, Present study

[23]
aCa in CaCl2, Staffansson and Sichen
0.1
[18]
aCaO in CaCl2, Suzuki et al.
0.05
0.10
0.15
0.20
0.25
0.30
0.0
0.9 1.0
Mole Fraction of Ca or CaO

Fig. 3 Eect of CaO in Ca-CaCl2 ux on deoxidation of Ti and the activity
of CaO at 1000 C.
Acknowledgements
The present study has been supported by the 21C Frontier
R&D Program and the Second Stage of Brain Korea 21
Project. Thereupon, authors gratefully acknowledge the
Resources Recycling R&D Center under the support of the
Ministry of Science & Technology and the Ministry of
Environment, Korea and the Division of Humantronics
Information Materials in the Department of Metallurgical
CaO
0.1
eta
no
tio
tio
rac
rac
Ca
3.59%
0.2
of
aO
fC
le f
0.2
on
no
Mo
4.847%
cti
tio
0.3
0.1
fra
rac
no
le f
fM
eta
Mo
0.3
le
Mo
llic
llic
Ca
Ca
0.0
fM
At 1273 K
Ca-CaCl 2(-CaO) system
Ca-CaO-CaCl 2 system
0.0 0.3
le f
1E-3
0.00
concentration of oxygen in Ti gradually increases by

increasing the content of CaO in the 5 mol% Ca-CaCl2 ux
up to about 10 mol% CaO, followed by a constant value of
5 mass% O. This tendency is in qualitatively good agreement
with the results measured by Suzuki et al.18) The activity of
Ca in the Ca-CaCl2 23) and that of CaO in the CaO-CaCl2
system17) are also shown. The maximum dierence in the
activity between CaO and Ca can be obtained at about
7 mol% Ca or CaO, viz. 93 mol% CaCl2 , where aCaO =aCa
0:34. If Ca content is higher than the saturated composition,
the dierence in the activity between CaO and Ca decreases
and thus the deoxidation eciency diminishes sharply.
Therefore, it is suggested that the activity ratio of CaO to
Ca can be considered as a deoxidation index of Ca-calcium
halide ux.
The iso-oxygen content lines in Ti, which is equilibrated
with the CaCl2 rich Ca-CaO-CaCl2 ux at 1000 C, were
shown in Fig. 4. The iso-oxygen lines lean to the Ca-CaCl2
binary side and increase from 60 to 110 ppm O in Ti with
increasing CaO content. This tendency can be understood by
the activity change of Ca and CaO in the ux.
We can summarize that deoxidation of -titanium with Ca
is strongly aected by the stability of byproduct (CaO) and
can be controlled by calcium halide uxes, such as CaCl2 and
CaF2 . And there exists a critical point of deoxidation of Ti
which combined the lowest content of oxygen in Ti and
corresponded with the minimum aCaO =aCa at specic temperature. In calcium halide uxes such as CaCl2 or CaF2 , the
lowest content of oxygen in Ti could be attained by the
optimization of the activity ratio of CaO to Ca as 0.06 at
1100 C and 0.1 at 1300 C, respectively. The activity ratio of
byproduct to reactant can be suggested for the deoxidation
index of -titanium with calcium halide ux.
Mo
Content of Oxygen in Ti [mass %]
Ca saturated
217
370 ppm
110 ppm
0.0
Ca
0.7
Mole fraction of CaCl2
Fig. 4
1.0
CaCl2
0.3
0.7
90 ppm
80 ppm
0.8 60 ppm
0.9
0.0
1.0 CaCl
Mole fraction of CaCl2
Iso-oxygen contents in Ti which equilibrated with Ca-CaO-CaCl2 ux system at 1000 C.
218
S.-M. Han, Y.-S. Lee, J.-H. Park, G.-S. Choi and D.-J. Min
Engineering, Yonsei University, Korea. Also, authors are

gratefully appreciated for the reviewers considered advices.
The English has been revised through the native speaker, who
is working in professional correction of technical writing in
the College of Engineering, Yonsei University, Korea.
REFERENCES
1) A. D. McQuillan and M. K. McQuillan: Titanium (Butterworths
Scientic Publications, London, 1956) pp. 337343.
2) Y. Waseda and M. Isshiki: Purication Process and Characterization of
Ultra High Purity Metals (Springer 2001) p. 31.
3) T. H. Okabe and Y. Waseda: JOM 49 (1997) 28.
4) G. Z. Chen, D. J. Fray and T. W. Farthing: Nature 407 (2000) 361.
5) D. J. Fray: JOM 53 (2001) 26.
6) R. O. Suzuki and S. Fukui: Mater. Trans. 45 (2004) 16651671.
7) R. O. Suzuki, K. Teranuma and K. Ono: Metall. Trans. B 34B (2003)
287.
8) R. L. Centeno-Sanchez, D. J. Fray and G. Z. Chen: J. Mater. Sci. 42
(2007) 7494.
9) T. H. Okabe, M. Nakamura, T. Oishi and K. Ono: Metall. Trans. B 24B
(1993) 449.
10) G. Z. Chen, D. J. Fray and T. W. Farthing: Metall. Mater. Trans. B 32B
(2001) 1041.
11) P. K. Tripathy, M. Gauthier and D. J. Fray: Metall. Mater. Trans. B 38B
(2007) 893.
12) T. H. Okabe, R. O. Suzuki, T. Oishi and K. Ono: Mater. Trans. JIM 32
(1991) 485.
13) T. H. Okabe, R. O. Suzuki, T. Oishi and K. Ono: Titanium Prod. Appl.,
Proc. Tech. Program Int. Conf., (1990) pp. 822829.
14) T. H. Okabe, R. O. Suzuki, T. Oishi and K. Ono: Tetsu-to-Hagane 77
(1991) 93.
15) T. H. Okabe, T. Oishi and K. Ono: J. Alloy. Compd. 184 (1992) 43.
16) T. H. Okabe, K. Fujiwara, T. Oishi and K. Ono: Titanium92 Sci.
Technol., (1993) pp. 12111218.
17) R. O. Suzuki, M. Aizawa and K. Ono: J. Alloy. Compd. 288 (1999)
173.
18) R. O. Suzuki and S. Inoue: Metall. Trans. B 34B (2003) 277.
19) M. H. Song, S. M. Han, D. J. Min, G. S. Choi and J. H. Park: Scr. Mater.
59 (2008) 623.
20) H. Niiyama, Y. Tajima, F. Tsukihashi and N. Sano: J. Less-Common
Met. 169 (1991) 209.
21) H. Fischbach: Steel Res. 56 (1985) 365.
22) N. Sano and F. Tsukiahshi: Report No. 1, 69th Committee JSPS, (1989)
p. 31.
23) L. I. Staansson and D. Sichen: Scand. J. Metall. 21 (1992) 165.
24) E. D. Eastman, D. D. Cubicciotti and C. D. Thurmond: Chemistry and
Metallugy of Miscellaneous Materials (McGraw-Hill Book Co., New
York 1950) p. 2.
25) E. D. Eastman, D. D. Cubicciotti and C. D. Thurmond: Chemistry and
Metallugy of Miscellaneous Materials (McGraw-Hill Book Co., New
York 1950) p. 10.
26) D. A. Wnz, I. Johnson and R. D. Wolson: J. Chem. Eng. Data 14 (1969)
252.
27) B. D. Lichter and M. A. Bredig: J. Electrochem. Soc. 112 (1965) 508.
28) J. Mukerji: J. Am. Ceram. Soc. 48 (1965) 212.

RAPID PUBLICATION
The Synthesis of Graded Thermal Barrier Coatings on Nickel Substrates

by Laser Induced Thermite Reactions
T. M. Yue, H. O. Yang, T. Li and K. J. Huang
The Advanced Manufacturing Technology Research Centre, Department of Industrial and Systems Engineering,
The Hong Kong Polytechnic University, Hung Hom, Hong Kong, P.R.China
Al2 O3 and ZrO2 -Al2 O3 bearing coatings were synthesized onto nickel substrates by laser induced thermite reactions. The oxide coatings
were produced via thermite reactions between NiO and Al powders with and without the addition of Y2 O3 -stabilised ZrO2 (YSZ), respectively.
In the case where no YSZ was added, a graded coating with a top layer of Al2 O3 was produced. Whereas, when YSZ was added, the coating
essentially comprised ZrO2 dendrites with the interdendritic regions lled with the ZrO2 -Al2 O3 eutectic phase. The formation process of the
coatings is described. [doi:10.2320/matertrans.MRP2008304]
Keywords: laser processing, coating, thermal barrier, thermite reaction
1.
Introduction
The improved thermal barrier coatings (TBCs), which are

now being developed, are playing an increasingly signicant
role in todays aero engines, in engines for automobiles, as
well as in power generators. In this connection, functionally
graded TBCs (FG-TBCs) are being seriously explored.
Currently, the two most commonly used coating techniques
for commercial applications of FG-TBCs are plasma spraying and electron beam physical vapour deposition (EB-PVD).
Although plasma spraying provides a exible and economical means for producing FG-TBCs, especially in forming
thick coatings, there is evidence that EB-PVD formed
coatings are superior to those produced by plasma spraying.1)
On the other hand, although EB-PVD is capable of producing
graded TBCs, it can be dicult to adjust the chamber
conditions in such a way as to achieve the desired
evaporation behaviour.2)
In considering other alternative methods for fabricating
FG-TBCs, laser surface treatment is becoming a strong
candidate because of its relatively simple operating principles. Indeed, it has been demonstrated that laser treatment is
potentially a method for producing TBCs.3,4) With this
background in mind, this research proposes a self-synthesizing laser irradiation method that is based on thermite
reactions, to produce FG-TBCs on Ni substrates.
In the present research, the reaction between NiO and Al
is of particular interest because the reaction products of
aluminium oxide and Ni-Al intermetallics are ideal for
forming FG-TBCs. Up to now, only very few publications
concerning ceramic coatings formed by laser treatment
through thermite reactions can be found in the literature,57)
and to the authors knowledge, no published work using such
an approach to produce FG-TBCs on nickel substrates can be
found in the literature.
2.
Experimental Method
FG-TBCs were fabricated on pure nickel (99.9 mass%)

substrates by using laser deposition. In this study, Al and NiO
powders were used to synthesize the TBC. Under laser
irradiation, a reduction-oxidation reaction between NiO and
Al occurs. The Al2 O3 phase that is formed acts as an excellent oxidation-resistant barrier. Nickel oxide was employed
because aluminium oxide has a lower level of Gibbs free
energy than nickel oxide, right up to very high temperatures.
In addition, the intention was to form Ni-Al intermetallics by
the exothermic reactions between nickel and aluminium as
the intermetallics possess high melting points and good oxidation resistance. The stoichiometry to produce Al2 O3 and
some common nickel aluminides is given by the reactions:
3NiO 5Al ! Al2 O3 3NiAl (+heat)
1
3NiO 2Al ! Al2 O3 3Ni (+heat)
3NiO 3Al ! Al2 O3 Ni3 Al (+heat)
It was envisaged that the heat generated by the thermite

reactions between NiO and Al would raise the temperature of
the product, aluminium oxide, to its melting temperature.
Owing to the low density of the Al2 O3 phase, it would form at
the top surface of the reaction coating, while the Ni-Al
intermetallics that are formed would likely be embedded in
the Ni-based matrix. Such a graded system is ideal for
resisting high temperature oxidation, but its thermal conductivity is relatively high. In order to reduce thermal
conductivity, Y2 O3 -stabilised ZrO2 (YSZ) powder (7 mass%
Y2 O3 ), an excellent thermal insulation material, was used in
some of the experiments. Mixed powders of YSZ, NiO and
Al powders were used in the experiment to form a composite
coating. The blended powders comprised 30 mass% or
60 mass% YSZ power with balanced NiO and Al powders
(in a molar ratio of 3 : 5).
The mixed powders of Al-NiO with and without YSZ were
delivered onto the surface of the nickel substrate through a
powder feeder. Laser irradiation was performed using a 5 kW
CO2 continuous wave laser. The laser processing parameters
were: laser power between 1100 W and 2100 W, scanning
velocity between 800 and 1000 mm/min. To determine the
thermite reaction temperature, a DSC experiment was
performed using the mixed powders of Al and NiO (molar
ratio 3 : 5) in argon gas at a heating rate of 20 K/min.
3.

Figure 1 shows a cross-section of the coating, without
220
T. M. Yue, H. O. Yang, T. Li and K. J. Huang
Fig. 3
Fig. 1
A cross-section of the coating formed without the addition of YSZ.
Fig. 2
X-ray diraction pattern of the coating without YSZ.
the addition of YSZ, which consists of two distinct layers.

The results of an XRD analysis show that Al2 O3 , NiAl and
Ni3 Al are present in the reaction coating (Fig. 2), moreover
the X-ray distribution maps of elements show that the top
layer basically consists of Al and O. The results thus show
that the top layer is composed essentially of the -Al2 O3
phase with a trace of -Al2 O3 . While a relatively dense layer
of Al2 O3 was obtained, cracks were found in the coating, this
was likely due to thermal stresses developed during rapid
solidication. It is obvious that these nickel aluminides are
by-products of the thermite reactions.
The formation process of the top surface layer of Al2 O3 is
considered to be as follows. At the start of laser irradiation, a
shallow molten pool was created in the substrate and
subsequently Al and NiO powders were fed into the pool,
while the workpiece was moving at a constant speed. Since
Al has a lower melting point than NiO, it started melting once
it reached the melting point which is about 933 K. When the
temperature was raised to a certain temperature, the thermite
reaction between NiO and Al started. According to the results
of the DSC curve for the mixed powders of Al and NiO, an
endothermic peak and an exothermic peak occurred at the
temperatures 935 K and 1350 K respectively (Fig. 3). The
endothermic peak corresponds to the melting of Al while the
exothermic peak represents the thermite reaction. These
results suggest that when the temperature of the molten pool
reached 1350 K, the exothermic chemical reaction between
DSC curve of the mixed powders of Al and NiO.
molten Al and NiO started and the reactants, Al2 O3 and NiAl
were produced. Some variation in composition is expected in
the melt. In fact not only NiAl was produced, other nickel
aluminides, such as those according to reaction (3), were
formed.
Since all the three reactions (1), (2) and (3) are exothermic,
the temperature of the entire reaction layer could be increased
and exceeds the melting point of alumina and forms liquid
Al2 O3 . Assuming that the temperature of the reaction layer
was at 1350 K when the reactions (1)(3) occurred, the
exothermic heat generated by the three reactions can be
calculated as 1402 kJ, 987 kJ, 1162 kJ, respectively. It can be
calculated that heat energies of 636 kJ, 422 kJ, and 388 kJ are
sucient to raise the temperature of the reaction products of
reactions (1)(3) respectively from 1173 to 2327 K, i.e. the
melting point of alumina. It is therefore considered that the
exothermic heat generated by the reactions was high enough
to cause the melting of the Al2 O3 phase. Comparing the
density of liquid alumina and those of the other reaction
products, the former was lighter, therefore it oated to the top
of the reaction layer, whilst the rest of the other phases sank
and mixed with the already molten substrate and formed the
graded reaction coating as shown in (Fig. 1). The presence of
a small amount of -Al2 O3 phase indicated by the two small
peaks at 2 of 51.001 and 74.548 in Fig. 2, which were not
marked in the gure, is believed to be due to the rapid cooling
rate of the laser processing condition (103 104 K/s). It is
known that -Al2 O3 is a transitional phase that usually forms
by rapid cooling below 1200 K.6) Cao8) reported that lasermelted -Al2 O3 re-solidied as a transitional -Al2 O3 phase
after excimer laser irradiation. He suggested that the high
cooling rate during solidication after the pulsed laser
irradiation had prevented the ordering of aluminium atoms
required for the formation of the stable -Al2 O3 . The Al2 O3 can form by changing the stacking sequence of the
closed-packed oxygen sub-lattice from hcp to fcc and the
occupying of aluminium cations from two-thirds of the total
octahedral sites available in -Al2 O3 to both octahedral and
tetrahedral sites in -Al2 O3 .
With regard to the coating produced with the addition of
YSZ (Fig. 4), the microstructure is rather dierent from that
of the coating without YSZ (Fig. 1). In the case with YSZ, a
graded coating with a top layer of Al2 O3 was not obtained.
The XRD results show that beside Al2 O3 and the various NiAl intermetallics, t-ZrO2 phase and a small amount of m-
The Synthesis of Graded Thermal Barrier Coatings on Nickel Substrates by Laser Induced Thermite Reactions
Fig. 4 A cross-section of the coating formed with the addition of 60 mass%

YSZ.
ication process, it is believed that the high melting point

ZrO2 phase will solidify rst and form dendrites. As the
results of the EDX analysis showed that the constituents of
the dendrite were primarily Zr and O with a few percent Al
present, it is therefore believed that Al2 O3 exists in solid
solution form in the ZrO2 dendrite. This is in agreement with
the ZrO2 -Al2 O3 phase diagram.9) As the temperature continuously decreased, the eutectic structure of ZrO2 -Al2 O3
nally formed in the interdendritic regions together with
some nickel aluminides. Under the experimental conditions
where 30 mass% or 60 mass% YSZ was employed, a graded
coating was not formed. It is considered that if the amount of
YSZ is much reduced, a graded coating can be produced and
further studies on this will follow shortly.
It is worth noting that unlike the coating formed without
YSZ, which contained horizontal cracks at the interface
between the top Al2 O3 layer and the underneath reaction
coating, such cracks were not observed in the coating when
YSZ was added. This suggests that the introduction of the
ZrO2 phase into the TBC helps to alleviate stresses arising
from the thermal expansion mismatch between the Al2 O3
phase the metal phase. Actually, the thermal expansion
coecient of ZrO2 is between that of the Al2 O3 phase and
nickel.
4.
Fig. 5 TEM micrograph and the SAED pattern of an interdendritic region;

Al2 O3 is labelled (A), ZrO2 is labelled (B).
ZrO2 phase were present in the coatings. Moreover, as YSZ

increased from 30 to 60%, the amount of ZrO2 increased. A
TEM study was conducted to examine the microstructure of
the coating. The results revealed the presence of a welldened lamellar eutectic structure of Al2 O3 -ZrO2 in the
interdendritic regions (Fig. 5).
In the case where the YSZ powder was added to NiO and
Al powders, the main reaction is considered to have occurred
between NiO and Al, since a large free energy of reaction
(G < 0) is obtained from the reactions (1)(3). Therefore
spontaneous reactions are expected to occur between NiO
and Al. On the other hand, the free energy of the reaction
between ZrO2 and Al, i.e. 4Al 3ZrO2 3Zr 2Al2 O3 , at
elevated temperatures is positive (G > 0), this indicates
that it is a non-spontaneous reaction. Assuming this is the
case, then a simple calculation shows that a heat energy of
1072 kJ will be sucient to raise the temperature of the
reaction products of NiO and Al (1 moL Al2 O3 and 3 moL
NiAl) and the extra 1 moL ZrO2 from 1350 to 2950 K, i.e., the
melting point of ZrO2 . Given that the exothermic heat
generated by reaction (1) is 1402 kJ which is higher than the
calculated value of 1072 kJ, therefore Al2 O3 , NiAl and ZrO2
phases would be in liquid form. In considering the solid-
221
Conclusions
The study showed that laser treatment is a promising

method for coating nickel with an Al2 O3 or a ZrO2 -Al2 O3
bearing thermal barrier coating. Without the addition of YSZ,
a top layer of Al2 O3 was obtained and a graded coating was
produced, while when YSZ was added, the entire coating was
composed essentially of ZrO2 dendrites with the ZrO2 -Al2 O3
eutectic structure. Further work is required to optimise the
parameters of the process. This, may include adjusting the
composition of the mixed powders, in order to obtain
coatings of good integrity.
Acknowledgements
The work described in this paper was fully supported by a
grant from the Hong Kong Polytechnic University (Project
No. G-U283).
REFERENCES
1) N. P. Padture, M. Gell and E. H. Jordan: Science 296 (2002) 280284.
2) Functionally Graded Materials: Design, Processing and Application,
ed. by Y. Miyamoto, W. A. Kaysser, B. H. Rabin, A. Kawasaki and
R. G. Ford (Kluwer Academic Publishers, 1999) pp. 194196.
3) Y. T. Pei, J. H. Ouyang and T. C. Lei: Surf. Coat. Technol. 81 (1996)
131135.
4) G. F. Chen, Z. C. Feng and Y. Liang: Trans. Nonferrous Met. Soc. 10
(2000) 9293.
5) X. B. Zhou and J. Th M. De Hosson: Scr. Metall. Mater. 28 (1993)
219224.
6) J. A. Vreeling, Y. T. Pei, B. Wind, V. Ocelk and J. Th M. De Hosson:
Scr. Mater. 44 (2001) 643649.
7) Y. X. Li, J. K. Yao and Y. Liu: Surf. Coat. Technol. 172 (2003) 5764.
8) S. Cao, A. J. Pedraza, D. H. Lowndes and L. F. Allard: Appl. Phys. Lett.
65 (1994) 29402942.
9) S. N. Lakiza and L. M. Lopato: J. Am. Ceram. Soc. 80 (1997) 893902.

RAPID PUBLICATION
Stabilization of Stacking Faults and a Long Period Stacking Phase Dispersed

in -Mg Crystalline Grains of Mg-0.7 at%Zn-1.4 at%Y Alloy
Jongbeom Lee, Kazuhisa Sato, Toyohiko J. Konno and Kenji Hiraga
Mg-0.7 at%Zn-1.4 at%Y alloys annealed at low temperatures after quenching in water from 520 C were studied by high-resolution
transmission electron microscopy (HRTEM) and high-angle annular detector dark-eld scanning transmission electron microscopy (HAADFSTEM). Stacking faults, thin bands of a 14H-type long period stacking (LPS) phase and relatively thick bands of LPS were precipitated in -Mg
crystalline grains by annealing at 300 C, 400 C and 500 C, respectively. The precipitation of stacking faults, LPS phase and a supersaturated
solid solution without any precipitates were reversibly transformed by annealing at low temperatures. It can be concluded that the stacking faults
and LPS phase are stabilized by the segregation of Zn and Y from a supersaturated solid solution. [doi:10.2320/matertrans.MRP2008335]
Keywords: magnesium-zinc-yttrium alloy, stacking fault, long period stacking phase, supersaturated solid solution, precipitate
1.
Introduction
Since an Mg97 Zn1 Y2 alloy prepared by a rapid solidication method was found to display a yield strength of
610 MPa and elongation of 5% at room temperature,1) MgZn-RE (RE: rare earth metals) alloys have received much
attention. The excellent mechanical properties of Mg-Zn-Y
are considered to result from the coexistence of ne -Mg
crystals and dispersed LPS phases,2) and a similar microstructure consisting of ne -Mg crystals and LPS phases
obtained by the hot-extrusion of normally solidied Mg-ZnY cast alloys were found to exhibit a high yield strength of
about 350 MPa and elongation of 5%.3) Microstructures
composed of -Mg crystals and LPS phases have been found
for many Mg-Zn-RE alloys, which can be classied into two
types.4) In type I Mg-Zn-RE (RE Y, Dy, Ho, Er, Er, Tm)
alloys, the LPS phases are formed during solidication, and
their morphology cannot be controlled. On the other hand, in
type II Mg-Zn-RE (RE Gd, Tb) alloys, the LPS phases are
formed by annealing at high temperatures. Therefore, as for
an Mg97 Zn1 Gd2 alloy, various microstructures with precipitates of not only the LPS phase, but also of stacking faults, 0
precipitates and GP-zones are obtained by annealing at
various temperatures.
The purpose of the present work was to examine the
possibility of microstructure control for Mg-Zn-Y alloys,
which belong to the type I alloy system. A supersaturated
solid solution with a composition of Mg-0.6 at%Zn-1.1 at%Y
was formed by quenching from 520 C, and various morphologies of precipitates, such as dispersed stacking faults
and bands of an LPS phase, in -Mg crystals were obtained
by the segregation of Zn and Y from the supersaturated solid
solution by annealing at low temperatures.
2.
Transmission electron microscopy (TEM) specimens were

cut from the annealed alloys, and thinned by mechanical
polishing and nally by ion-milling. HRTEM observations
were performed by a 400 kV electron microscope (JEM4000EX) with a resolution of 0.17 nm, and HAADF-STEM
images were taken by a 300 kV electron microscopy
(TITAN) equipped with a eld emission gun in the scanning
transmission electron microscope mode. Almost all of
HRTEM and HAADF-STEM images were taken using a
systematic condition, namely, with one-dimensional diraction patterns along the c -axis of a hexagonal Mg-matrix, in
order to obtain sharp contrasts of stacking faults. SEM
observations and electron probe microanalysis (EPMA) were
done with an EPMA machine (JXA-8621MX).
3.
Experimental Results
Figure 1 shows an SEM image of the alloy quenched in

water after annealing at 520 C for 5 h. In the image, thin
bright regions along grain boundaries and some bright
particles inside the matrix grains are Zn- and Y-rich second
phases, almost all of which are an LPS phase. The
composition of the gray matrix grains was estimated to be
An alloy with a nominal composition of Mg-0.7 at%Zn1.4 at%Y was prepared by melting Mg (99.99%), Y (99.9%)
and Zn (99.9%) with a high frequency induction heater under
Ar gas in a carbon crucible. The alloy was quenched in water
after annealing at 520 C for 5 h, and then annealed at various
temperatures of 300500 C.
Fig. 1 SEM image of the Mg97:9 Zn0:7 Y1:4 alloy quenched in water after
annealing at 520 C for 5 h.
Stabilization of Stacking Faults and a Long Period Stacking Phase Dispersed in -Mg Crystalline Grains of Mg-0.7 at%Zn-1.4 at%Y Alloy
Fig. 2 HRTEM lattice image of the Mg97:9 Zn0:7 Y1:4 alloy annealed at
300 C for 5 h after quenching from 520 C.
223
Fig. 3 HRTEM structure image of stacking faults in the Mg97:9 Zn0:7 Y1:4
alloy annealed at 300 C for 5 h, taken with the incident beam parallel to
the [110] direction of the hexagonal Mg-lattice. The arrows referred as I
and E show intrinsic and extrinsic stacking faults, respectively.
Fig. 4 HAADF-STEM images of low (a) and high (b) magnications of stacking faults in the Mg97:9 Zn0:7 Y1:4 alloy annealed at 300 C for
5 h. The arrows in (b) indicate stacking fault planes.
Mg-0.6 at%Zn-1.1 at%Y by EPMA. Since no precipitates

except for the second phases were observed in -Mg
crystalline grains by TEM, it can be concluded that a
supersaturated solid solution with a composition of Mg0.6 at%Zn-1.1 at%Y was formed by quenching from 520 C.
Figure 2 shows an HRTEM lattice image of the alloy
annealed at 300 C for 5 h after quenching from 520 C. Many
stacking faults are observed as sharp line contrasts parallel to
the close-packed planes of the hexagonal Mg-matrix. Since
close examination of the image shows the existence of
stacking faults with both terminations and one termination in
addition to extended stacking faults without terminations, it
can be concluded that the stacking faults were formed and
grew inside the -Mg crystalline grains.
In order to examine the property of stacking faults,
HRTEM structure images were taken with the incident
beam parallel to the [110] direction of the hexagonal Mgmatrix, as shown in Fig. 3. Most all of stacking faults are
intrinsic ones with stacking sequences of ABAB" CACA and
ABABA" CBCB (": a stacking fault plane), as can be seen
in the stacking faults referred as I in Fig. 3, and extrinsic
stacking faults with an ABAB" C" ABAB sequence, which
can be seen in the stacking fault referred as E in Fig. 3, were
occasionally observed.
The stacking faults are represented as bright lines in
HAADF-STEM observations, as shown in Fig. 4(a), and

consequently, the segregation of heavy atoms, such as Zn
and Y, occurs around the stacking faults. The atomic scaled
HAADF-STEM image of Fig. 4(b) clearly shows the
existence of two Zn- and Y-rich atomic planes, which are
represented as bright fringes in the gure, sandwiching the
stacking fault plane indicated by arrows. This feature has
already been observed in an LPS phase of Mg-Zn-Y2) and
stacking faults of Mg-Zn-Gd alloys.5)
annealed at 400 C for 5 h after quenching from 520 C. In the
image, thin bands with periodic arrays of stacking faults,
namely, thin bands of an LPS phase, can be seen in -Mg
crystalline grains. Diraction patterns and HRTEM observations show that the LPS phase is a 14H-type one consisting
of a periodic array of stacking faults with an interval of 7c0
(c0 : a lattice constant of the hexagonal unit cell of the Mgmatrix).6) It should be noted here that most LPS bands consist
of arrays of stacking faults with an interval of 7c0 , even if
they are formed with two or three stacking faults, as can be
seen in Figs. 4(b) and 5(b).
annealed at 500 C for 5 h after quenching from 520 C. In the
image, relatively thick bands of the LPS phase can be seen.
Figure 7 shows distributions of the number of stacking faults
224
J. B. Lee, K. Sato, T. J. Konno and K. Hiraga
Fig. 5
HRTEM lattice image (a) of the Mg97:9 Zn0:7 Y1:4 alloy annealed at 400 C for 5 h and enlarged image (b).
Fig. 6 HRTEM lattice image of the Mg97:9 Zn0:7 Y1:4 alloy annealed at
500 C for 5 h.
Fig. 7 Distributions of the number of staking faults arrayed periodically

with an interval of 7c0 at annealing temperatures.
arrayed with an interval of 7c0 . In the alloy annealed at

300 C, almost all of the stacking faults are isolated, but a
few are paired. Thin LPS bands with arrays of 2-7 stacking
faults and relatively thick LPS bands with 4-18 stacking
faults are observed in alloys annealed 400 C and 500 C,
respectively.
In order to examine the stability of stacking faults and LPS
bands precipitated in -Mg crystalline grains, re-annealed
Fig. 8 Number of stacking faults in a length of 1 mm along the c-axis at

annealing temperatures.
alloys were investigated. The stacking faults in the alloy

annealed at 300 C for 5 h were transformed into the bandshaped LPS phase, which is nearly the same as the image of
Fig. 6, by re-annealing at 500 C for 5 h. The band-shaped
LPS phase in the alloy annealed at 500 C for 5 h were
transformed into dispersed stacking faults, which is nearly
the same as the image of Fig. 2, by re-annealing at 300 C for
5 h. The results show that the stacking faults and band-shaped
LPS phase are stable morphologies of precipitates at each
temperature. Also, the stacking faults and LPS bands
disappeared by re-annealing at 520 C for 5 h. That is, Zn
and Y segregated by annealing at low temperatures redissolved in -Mg crystals by re-annealing at 520 C, and
consequently the stacking faults and LPS bands cannot be
stabilized. This result shows that the stacking faults and LPS
phase are formed by the segregation of Zn and Y in a
supersaturated solid solution. This conclusion is conrmed
by the distribution density of stacking faults depending on
annealing temperature, as shown in Fig. 8. In Fig. 8, the total
number of stacking faults, both isolated and in the form of
LPS bands, in a length of 1 mm along the c-axis is shown in
relation to annealing temperatures. The distribution density
of stacking faults decreases with increasing solubility limit,
namely, with decreasing amounts of segregated Zn and Y at
elevated temperatures.
Stabilization of Stacking Faults and a Long Period Stacking Phase Dispersed in -Mg Crystalline Grains of Mg-0.7 at%Zn-1.4 at%Y Alloy
4.
Summary
A supersaturated solid solution of -Mg with a composition of Mg-0.6 at%Zn-1.1 at%Y was obtained by quenching
in water from 520 C for the Mg-0.7 at%Zn-1.4 at%Y alloy.
Stacking faults, thin bands of a 14H-type LPS phase and
relatively thick bands of LPS were precipitated in -Mg
crystalline grains by annealing the supersaturated solid
solution at 300 C, 400 C and 500 C, respectively. By reannealing the alloy, it was found that these precipitates and
the supersaturated solid solution with no precipitates are
reversibly transformed by changing annealing temperatures.
The result suggests that the stacking faults and LPS phase are
stabilized by the segregation of Zn and Y from the supersaturated solid solution by annealing at low temperatures.
The present investigation, therefore, demonstrates that one
can control the microstructure of Mg-Zn-Y alloy by selecting
annealing temperature.
225
Acknowledgements
The authors are grateful to Mr. T. Takahashi and Y.
Murakami, who helped us for the sample preparation and
EPMA analysis, respectively, Institute for Materials Research, Tohoku University.
REFERENCES
1) Y. Kawamura, K. Hayashi, A. Inoue and T. Masumoto: Mater. Trans. 42
(2001) 11721176.
2) E. Abe, Y. Kawamura, K. Hayashi and A. Inoue: Acta Mater. 50 (2002)
38453857.
3) S. Yoshimoto, M. Yamasaki and Y. Kawamura: Mater. Trans. 47 (2006)
959965.
4) Y. Kawamura and Y. Yamasaki: Mater. Trans. 48 (2007) 29862992.
5) M. Yamasaki, M. Sasaki, M. Nishijima, K. Hiraga and Y. Kawamura:
Acta Mater. 55 (2007) 67986805.
6) T. Itoi, T. Seimiya, Y. Kawamura and M. Hirohashi: Scr. Mater. 51
(2004) 107111.

RAPID PUBLICATION
Direct Bonding to Aluminum with Silver-Oxide Microparticles

Toshiaki Morita1 , Yusuke Yasuda1 , Eiichi Ide1 and Akio Hirose2
1
2
Materials Research Laboratory, Hitachi Ltd., Hitachi 319-1292, Japan

Division of Materials and Manufacturing Science, Osaka University, Suita 565-0871, Japan
This study demonstrated that direct bonding between electrode and aluminum by using silver oxide particles which added acetic acid
silver, followed by heating in air at 350 C under a pressure of 2.5 MPa. Direct bonding with aluminum by using solder materials containing tin,
lead, or the like has been impossible in the past. We think this technique creates strong bonding between silver and the natural oxide lm formed
on an aluminum surface by the combustion heat generated during oxidation of acetic acid (which is generated by decomposition of the silver
acetate as it is heated). [doi:10.2320/matertrans.MRP2008277]
Keywords: silver oxide, reduction, bonding technique, aluminum
1.
Introduction
Other than noble metals, nickel and copper are often used
in the mountings of semiconductor chips. Aluminum is also
often used. However, since a thermally stable natural-oxide
lm forms on aluminum surfaces, bonding it with a solder
material containing tin or lead has been impossible in the
past. For that reason, a lm of nickel, which reacts readily
with tin, is usually applied to the aluminum surfaces.
Accordingly, if a technique enabling direct bonding with
aluminum were realised, it would reduce the bonding costs in
semiconductor mountings. In addition, it would expand the
range of electrodes that can be used for bonding, thus giving
the freedom to use mounting techniques that have been
impossible to apply up to now.
The authors found that adding myristyl alcohol to silveroxide particles can be reduced by heating in air without
the use of a reducing atmosphere.1) It can bond to gold or
silver1) like the previously mentioned silver nanoparticles.26)
If the heat generated by the oxidation of the reduction
material (alcohol-based material) gives a suciently high
temperature, bonding between a ceramic (such as Alumina,
Silicon nitride, Aluminum nitride) and silver would be
possible.
We therefore attempted to create a direct bond between
silver and aluminum by making use of the heat generated
during the reduction of silver oxide (to which a reduction
material had been added), and to cause bonding between the
silver produced by the reduction and the aluminum.
2.
2.1 Particle materials used in bonds

The silver-oxide particles used in our investigation as
shown in Fig. 1, had a diameter of 2 to 3 mm. Ethyleneglycol-based and alcohol-based substances were conventionally considered as materials for the silver-oxide reduction;7)
however, in this investigation, we used silver acetate, which
has a lower outgassing volume in comparison with those
other substances. The amount of silver acetate added to the
silver-oxide particles was 20 mass% (called bond material
hereafter).
3m
Fig. 1
Appearance of silver-oxide (Ag2 O) particles used in tests.
2.2 Discussion of bonding structure

2.2.1 Bond-strength
We rst investigated bonding capability of silver-oxide
bond material with aluminum. Bonding capability was
evaluated as bond strength under shear. Perspective views
of the shearing test piece are shown in Fig. 2 (JIS Z3198-5).
The test piece was composed of aluminum. The bond
material was painted onto each test piece in the regions
shown in gray in Fig. 2. The paint thickness was approximately 100 mm. Bonding was then accomplished by simultaneous application of heat and pressure in air.
We used an SS-100KP bond tester (Seishin Trading,
maximum load: 100 kg) for the shear tests. The rate of shear
was 30 mm/min, and the maximum load during fracture was
measured by applying a shearing force to a small aluminum
disk (shown in Fig. 2) to fracture the bond. The shear
strength was taken as this maximum load divided by the bond
surface area (equivalent to the surface area of the disk).
2.2.2 Observation of bond microstructure
The cross-sectional samples used for SEM observation
were rst nished by bung with diamond paste (particle
diameter: 0.25 mm) using special polishing oil and then
Direct Bonding to Aluminum with Silver-Oxide Microparticles
Pressures
Small Al disk
5 mm diameter
2 mm thick
Shear tool
227
(a)
Heating
in air
Bond material
(100m thick)
Ag sintering layer
Large Al disk
10 mm diameter
5 mm thick
(1)
Al
Fig. 2 Specication of shear-test sample and schematic of shear-test

method.
50m
Shear strength (MPa)
10
8
(b)
bond material
silver-oxide particles
Pb-5mass%Sn
Ag sintering layer
6
4
2
0
200
250
300
350
Bonding temperature (C)
400
Fig. 3 Results of bond-strength evaluation with respect to bonding

temperature.
subjected to argon-ion milling (at an accelerating voltage of

15 kV) for 180 s.
To obtain more detailed images of the bond between silver
and aluminum, we observed the interface by transmission
electron microscope (TEM). The samples used for observing
the cross-sections of bonds by TEM were milled using
focused-ion-beam (FIB) milling and observation device
(Hitachi FB-2000A). The milled thin-lm samples were then
mounted on a cutout mesh and observed by TEM (Hitachi
H9000; acceleration voltage: 300 kV). If necessary, ion
milling was also used to make the observation samples even
thinner.
3.
3.1 Evaluation of bond-strength

The results of the bond-strength evaluation with respect to
bonding temperature are shown in Fig. 3. Bonding was done
in air at a pressure of 2.5 MPa and a predetermined
temperature held for 2.5 minutes. For comparison, we also
evaluated silver-oxide particles alone, without the addition of
silver acetate. Furthermore, the evaluation result of the bondstrength of Pb-5 mass%Sn solder (Liquidus Temperature:
298 C) was also shown in Fig. 3. Soldering was performed
in hydrogen.
As for the silver-oxide particles with no added silver
acetate, the shear strength remained unchanged (at a low
Al
1m
Fig. 4 SEM images of bond cross-sections for bonding temperature of
350 C: (a) low-magnication image and (b) enlargement of area (1) in
Fig. 4(a).
level) even when bonding temperature was increased, so

bonding was not possible. This is thought to be because the
silver oxide did not reduce in this case. On the other hand,
as for the silver-oxide bond material to which silver acetate
had been added, bonds exhibited a low bond strength of
3 MPa at 250 C, which increased to approximately 5 MPa at
300 C and an average of 9 MPa at 350 C and above, showing
that bonding can be facilitated by raising the bonding
temperature. This is thought to be because raising the
temperature promotes sintering of the silver, so bond strength
increases. In addition, in Sn-5Pb solder, 300, 350 and 400 C
of bond was not completed. The solder which melted did
not spread on aluminum. It is thought that this reason is
because the oxide layer forms on the surface of aluminum.
The oxidization lm on aluminum was not reduced at
400 C.
3.2 Observation of detailed bond structure
Figure 4 shows SEM images of cross-sections through
bonds formed at a bonding temperature of 350 C. Figure 4(a)
is a low-magnication image of the interface, and Fig. 4(b) is
a high-magnication image of the area marked (1) in
Fig. 4(a). A few minute voids are scattered within the silver
layer that has been sintered, but the silver-sintering layer is
composed of grains with diameters in the order of 500 nm and
is densely sintered.
228
T. Morita, Y. Yasuda, E. Ide and A. Hirose
TEM images of the bond interfaces of the specimen

shown in Fig. 4 are shown in Fig. 5. Figure 5(a) is a lowmagnication image of the interface, whereas Fig. 5(b)
shows the lattice structure of the interface. Figure 5(a) shows
that there is an interlayer in the interface between the silver
and the aluminum. Figure 5(b) shows that interlayer is
bonded to both the silver and the aluminum. Furthermore, the
interlayer has observed like an amorphous layer. This
interlayer is thought to consist primarily of oxide layer that
was formed on the aluminum surface.
(a)
Ag
Interface
4.
Al
100nm
(b)
Ag
Conclusions
From the results described above, the following conclusions were obtained.
(1) Bonding with aluminum is possible at 350 C if silveroxide particles with an average particle diameter of 2 to 3 mm
are mixed with silver acetate and heated in air.
(2) The silver produced by the reduction process from the
silver-oxide particles bonds to an oxide lm that is formed
on the surface of the aluminum.
REFERENCES
Interlayer
Al
5nm
Fig. 5 TEM images of bond cross-sections for bonding temperature of
350 C: (a) low-magnication image and (b) high-resolution image.
1) T. Morita, Y. Yasuda, E. Ide, Y. Akada and A. Hirose: Mater. Trans. 49

(2008) 28752880.
2) A. Hirose, E. Ide and K. F. Kobayashi: Jpn. Inst. Electron. Packag. 7
(2004) 511515.
3) E. Ide, S. Angata, A. Hirose and K. F. Kobayashi: Acta Mater. 53 (2005)
23852393.
4) E. Ide, A. Hirose and K. F. Kobayashi: Mater. Trans. 47 (2006) 211217.
5) T. Morita, E. Ide, Y. Yasuda, A. Hirose and K. Kobayashi: Jpn. J. Appl.
Phys. 47 (2008) 66156622.
6) Y. Akada, H. Tatsumi, T. Yamaguchi, A. Hirose, T. Morita and E. Ide:
Mater. Trans. 49 (2008) 15371545.
7) Y. Yasuda, T. Morita, E. Ide, H. Hozoji and T. Ishii: Japan Patent
178911 (2008).

CV 08

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

CV 08

Hochgeladen von

Copyright:

Verfügbare Formate

Materials Transactions, Vol. 50, No. 1 (2009) p.

Special Issue on Severe Plastic Deformation for Production

National Institute for Materials Science, Tsukuba 305-0047, Japan

Materials Transactions, Vol. 50, No. 1 (2009) pp. 2 to 10

Eect of Extrinsic Grain Boundary Dislocations on Mechanical Properties

Grain renement is one of the techniques used to improve

Student, Kanazawa University. Present address: Aisin Aw

only geometrically necessary defect structures, i.e., intrinsic

2.1 Analysis model

obtained by using the embedded-atom method proposed by

or two pairs of dislocations are formed from the crack tips,

Dissociation of impinged lattice dislocations in grain

T. Shimokawa, T. Hiramoto, T. Kinari and S. Shintaku

(a) Impinged dislocation at site 1

(b) Structural change at site 1

(c) Impinged dislocation at site 2

(d) Structural change at site 2

(e) Impinged dislocation at site 3

(f) Structural change at site 3

Figure 2 shows the shear stress distributions z0 x0 23) and

the lattice dislocations are trapped when they enter the

dicult for the impinged dislocations to transform into

Here, EEGB is the energy of the model with equilibrium grain

dislocations generated in the dipole. However, even though

Shielding eect of impinged dislocations on dislocation emissions from cracks

T. Shimokawa, T. Hiramoto, T. Kinari and S. Shintaku

site 2 site 3 site 1

dislocation source may exhibit a shielding eect on the

rate of increase in the tensile stress in the model with

Eect of impinged dislocations on grain boundary

(d) 1 EGBD after annealing

loading is shown in Fig. 2(b). The origin of grain boundary

the nonequilibrium grain boundaries. Figure 8 shows the

T. Shimokawa, T. Hiramoto, T. Kinari and S. Shintaku

Fig. 7 Grain boundary coupling motion in the D2 direction in the upper

Fig. 6 Change in equivalent strain distributions of nonequilibrium grain

Fig. 8 Generation of intergranular deformation in upper equilibrium grain

Although the coupling motion is the same in both

shows the propagation velocities of the intergranular

(a) just after shear loading

(b) after annealing at 800 K

Velocity (m/s) of grain boundary motion under shear stress.

zx 800 MPa

simulations. The propagation directions are dened in the

T. Shimokawa, T. Hiramoto, T. Kinari and S. Shintaku

3) Y. Saito, H. Utsunomiya, N. Tsuji and T. Sakai: Acta Mater. 47 (1999)

Materials Transactions, Vol. 50, No. 1 (2009) pp. 11 to 18

First-Principles Study of the Stability and Interfacial Bonding

Grain boundaries (GBs) have serious eects on the

author, E-mail: m-kohyama@aist.go.jp

nature.5) In Al, sp electrons reveal rather covalent or

2.1 Ab initio scheme

R.-Z. Wang, M. Kohyama, S. Tanaka, T. Tamura and S. Ishibashi

conguration, although the RBT normal to the interface,

Stability and interfacial bonding of tilt and twist boundaries in Al and Cu

cell-length normal to the interface, where initial atomic

supercells with similar periodicity as the GB cells. For the

Optimum RBT, boundary energy and structural

R.-Z. Wang, M. Kohyama, S. Tanaka, T. Tamura and S. Ishibashi

Calculated results for the CSL GBs in Al.

Calculated results for the CSL GBs in Cu.

where dlayer is the interlayer distance in the bulk crystal

Fig. 2 Distribution of bond-length changes in the GB supercells: the glide

zx 800 MPa

where is an arbitrary factor less than unity which