Texture evolution of five wrought magnesium alloys during route …library.nmlindia.org/FullText/AM5305113135.pdf · Texture evolution of five wrought magnesium alloys during route
Post on 20-Mar-2020
12 Views
Preview:
Transcript
Acta Materialia 53 (2005) 3135–3146
www.actamat-journals.com
Texture evolution of five wrought magnesium alloys during routeA equal channel angular extrusion: Experiments and simulations
S.R. Agnew a,*, P. Mehrotra a, T.M. Lillo b, G.M. Stoica c, P.K. Liaw c
a Department of Materials Science and Engineering, University of Virginia, Charlottesville, VA 24590, USAb Environmental and Energy Sciences Division, Idaho National Engineering and Environmental Laboratory, Idaho Falls, ID 83415, USA
c Department of Materials Science and Engineering, University of Tennessee, Knoxville, TN 37996, USA
Received 20 January 2005; received in revised form 9 February 2005; accepted 10 February 2005
Abstract
Equal channel angular extrusion (ECAE) has been demonstrated to induce unusual deformation textures and resulting properties
in magnesium alloys, such as the remarkably enhanced room temperature ductility first reported by Mukai et al. [Mukai T, Yamanoi
M, Watanabe H, Higashi K. Scr Mater 2001;45:89]. This paper documents a wide range of textures which evolve during ECAE of
magnesium alloys. The fact that different alloys exhibit different texture evolutions is an indication of distinctions in the balance of
deformation mechanisms which operate within the different alloys. Polycrystal plasticity modeling is used to develop explanations
for these texture distinctions in terms of the relative activities of non-basal secondary slip modes, involving Æaæ and Æc + aæ type dis-locations. AZ alloys appear to exhibit balanced secondary slip of non-basal Æaæ and Æc + aæ dislocations, while ZK60 and WE43
appear to favor non-basal Æc + aæ slip. A binary Mg–Li alloy exhibits a radically distinct texture evolution, which is associated with
large-scale strain accommodation by non-basal Æaæ slip.� 2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: Magnesium; Texture; ECAE; ECAP; Crystal plasticity; Ductility
1. Introduction
Texture can result in strong anisotropy and asymme-
try, particularly in the case of the plastic behavior ofnon-cubic materials, such as hexagonal close packed
(hcp) magnesium and its alloys. One example is the ten-
sion/compression strength asymmetry exhibited by mag-
nesium extrusions, where the yield strength in
compression along the prior extrusion axis may be as lit-
tle as half that of the tensile yield strength along the
same direction (e.g., Refs. [1,2]).
If the flow stresses and hardening behaviors of theindividual deformation mechanisms are known, poly-
crystal-plasticity models provide a means of predicting
1359-6454/$30.00 � 2005 Acta Materialia Inc. Published by Elsevier Ltd. A
doi:10.1016/j.actamat.2005.02.019
* Corresponding author. Tel.: +1 434 924 0605; fax: +1 434 982 5660.
E-mail address: agnew@virginia.edu (S.R. Agnew).
the evolution of texture given boundary conditions
appropriate for the prescribed deformation [3]. Con-
versely, if the texture evolution and strain path are well
characterized, an investigation of the underlying defor-mation mechanisms is possible [4]. Both avenues are
examined in the context of equal channel angular extru-
sion of magnesium alloys in the present paper.
1.1. Equal channel angular extrusion
A novel metal forming process, which has come to be
known as equal channel angular extrusion (ECAE) (e.g.,
Ref. [5]) or pressing (ECAP) (e.g., Ref. [6]), can impart
enormous levels of strain without significantly changing
the overall dimensions of the sample. In this regard, it is
similar to torsion, which has also been used extensivelyin the study of strain hardening [7,8] and hot working
ll rights reserved.
3136 S.R. Agnew et al. / Acta Materialia 53 (2005) 3135–3146
[9,10]. Some researchers have also noted a similarity in
the geometry of the ECAE process and metal cutting
[11]. The main interest in such techniques of severe plas-
tic deformation (SPD) has been the potential to reduce
the grain-size of the material [12]. The concept of reduc-
ing the microstructure scale using SPD is not new (e.g.,Refs. [13,14]). In fact, it is a core tenet denoted ‘‘simili-
tude’’ in a leading theory of strain hardening [15]. The
interest in grain-size reduction is driven by the possibil-
ity to produce ultrahigh strength metals [16] and high
strain rate superplasticity (e.g., Refs. [17,18]). Other
methods of manufacturing nanocrystalline or ultrafine-
grained metals have been plagued by extrinsic defects
[19,20], such as cracks and pores, limiting their abilityto push the limits of grain-size strengthening or high rate
superplasticity. SPD has proven capable of increasing
both the strength and ductility of metals [21] as well as
consistently producing materials that exhibit high strain
rate superplasticity [17].
A final area of ECAE research has focused on
modeling the plastic deformation of metals and the cor-
responding texture evolution, in particular. Rather thanenforcing primarily pure shear characteristic of conven-
tional rolling or extrusion, ECAE essentially enforces
simple shear at the intersection of two channels
(Fig. 1). In addition to the unique deformation geome-
try, the work-piece may be passed through the die
multiple times for further deformation. A nomenclature
has been developed in the literature [6] to describe multi-
ple pass deformation routes, where route A denotes nosample rotations about the billet axis between passes,
route BA denotes 90� back-and-forth rotations, route
BC (simplified to route B in the current work) corre-
sponds to continuous 90� rotations, and route C denotes
180� rotations. It has been noted that these routes have
varying degrees of redundancy in the strain they impart,
with route A being the least redundant, and route C
essentially ‘‘undoing’’ the strain each second pass [22].Thus, the ECAE process offers a distinct possibility to
generate unique textures [23]. Comparisons of
experimental textures and those simulated using poly-
+
90˚
==
y'x' y
xzz'
(a) (b)
Fig. 1. Schematic of the ECAE processing showing (a) an idealized
case where a simple shear is localized within an infinitesimal band at
the intersection of the channels and (b) where the simple shear
accumulates along flow lines, which simultaneously rotate.
crystal-plasticity simulations have been used to test the
underlying assumptions about the strain path during
ECAE [24–27]. Researchers have also explored the use
of polycrystal models as a foundation for predicting
the microstructure refinement, which will take place dur-
ing deformation, and ECAE has been used as a test case[28]. The vast majority of ECAE texture studies have fo-
cused on cubic metals, such as Al (e.g., Ref. [27]), Cu
(e.g., Refs. [24,26]), Fe [25,29], and their alloys. There
have been comparatively few studies of ECAE textures
of non-cubic metals, such as Be [30], Mg [Error! Book-
mark not defined, 31, 32], Ti [33], and Zr [34].
1.2. Wrought magnesium alloys
Wrought magnesium alloys exhibit marginal cold
formability and the accepted explanation is the intrinsic
plastic anisotropy of the magnesium single-crystal [35].
However, recent modeling and transmission electron
microscopy (TEM) of deformed polycrystals has
stressed that magnesium grains may be less anisotropic
than single-crystal measurements have suggested[36,37]. It has also been reported that the effect of
grain-size upon the strength of magnesium is potent
due a high apparent Taylor factors [38,39], and the effect
upon ductility may also be dramatic [40]. The current
study of ECAE processing of magnesium alloys was ini-
tiated due to the potential to produce simultaneously
high-strength, high-ductility, and low-density alloys
through grain refinement [41–43].The range of textures observed in conventional mag-
nesium wrought products is limited and the potentially
positive impact on the mechanical behavior of magne-
sium alloys by developing a unique texture using ECAE
processing was first reported by Mukai et al. [75]. Alloy
AZ31 exhibited a 2–3 times improvement in the tensile
elongation after ECAE processing and annealing (in or-
der to restore the original grain-size of �15 lm). It wassuggested that the ductility improvement was due to the
randomization of the originally strong extrusion texture.
Subsequent work verified the incredible improvement in
ductility for both AZ61 [31] and AZ31 [32] and showed
that it was due to a change in the texture. However, the
resulting texture was shown to be stronger than the ori-
ginal extrusion texture, not weaker [32].
The objective of the current study is to determine thedeformation texture evolution which is enforced by
ECAE processing of five magnesium alloys (Table 1)
subjected to various processing routes. These texture
data could be useful in designing processing treatments
to explore the effect of texture on various behaviors
in magnesium (and some other hcp alloys as well).
The present article focuses on route A processing
and the possibility to make inferences regarding therelative activities of deformation mechanisms (slip
modes) within the different magnesium alloys using
Table 1
Alloys examined in this study, their nominal compositions, and typical applications
Alloy Al Zn RE Zr Mn Other Applications
AZ31 3 1 – – 0.3 Fe, Ni, Cu low Forgings, extrusions, photoengraving plate
AZ80 8 0.5 – – 0.3 High-strength forgings
ZK60 – 6 0.5 – High-strength forgings
WE43 – – 4a 0.5 – 3Y High-temperature castings, forgings
Mg4L – – – – – 4Li Experimental alloy
a RE mischmetal, rich in Nd in this case.
S.R. Agnew et al. / Acta Materialia 53 (2005) 3135–3146 3137
polycrystal-plasticity simulations [4]. A companion pa-
per [44] documents the texture evolutions of some of
the alloys subjected to route B and C processing.
2. Procedures
2.1. Material
The selected alloys are representative of the two most
common wrought alloy families, AZ (aluminum–zinc)
and ZK (zinc–zirconium); a new class of commercial al-
loys designed for use in high-temperature applications,
WE (yttrium-rare earth) [45]; and, finally, an experimen-
tal binary alloy with 4 weight percent (wt%) Li (near the
solid solubility limit of the a-hcp phase) that has beenthe focus of a number of prior studies [4,46–49]. The al-
loys used in this study were received from commercial
vendors listed in Table 2, with the exception of the
experimental lithium containing alloy, which was pro-
cessed at the Oak Ridge National Laboratory (ORNL).
For this alloy, the pure metals (99.99 wt% Mg, and
99.9% Li) were alloyed in a mild steel crucible within a
resistively heated (Pt-wire) furnace (�800 �C) andpoured into a heated (�150–200 �C) Cu mold (74 mm
diameter). The alloying and casting operations were
conducted within a glove-box, which was evacuated
and back-filled with a mixture of Ar, SF6 and O2 as a
protective cover gas. The ingot was, then, directly ex-
truded (4:1) into a 35 mm diameter bar. Conventional
metallography, including a picral-acetic etch commonly
used for magnesium alloys [50], revealed an essentiallyequiaxed grain structure in all of the as-received alloys
Table 2
Vendor and as-received condition of alloys examined
Alloy Vendor Initial condition
AZ31 Mark Metals Plate (25 mm)
AZ31 Timminco Ltd. Cast extrusion
AZ31 Mark Metals Extrusion (32 m
AZ80 Intercontinental Manuf. Extrusion (178
ZK60 Mark Metals Extrusion (76 m
WE43 Magnesium Elektron Extrusion (50 m
Mg4L n/a Extrusion (35 m
a O – fully annealed, F – as deformed.
with the exception ZK60, which had some larger elon-
gated grains dispersed among a matrix of fine equiaxed
grains [51].
2.2. Texture measurements and analyses
Crystallographic texture measurements were con-
ducted using X-ray diffraction in the reflection geometrywith a four circle goniometer and Cu Ka radiation.
Experimental f10�10g, (0002), and f10�11g, pole figureswere collected on a 5� · 5� grid for sample tilts,
a = 0�80�, and azimuthal rotations, u = 0�355�. Defo-
cusing corrections were made using experimentally
determined defocusing curves from random powder
samples. Complete orientation distributions and recal-
culated full pole figures were determined using theWIMV algorithm within the popLA (preferred orienta-
tion package of Los Alamos) code [52]. The texture sam-
ples were sections from the flow plane (or x–y plane in
Fig. 1). The X-ray spot (�2 mm) was focused on the
more uniformly deformed central portion of the billets.
2.3. ECAE processing
Equal channel angular extrusion was performed at
the Idaho National Engineering and Environmental
Laboratory (INEEL). The die utilized has an included
angle u = 90� between the square (22 mm) entrance
and exit channels (Fig. 1). There is no relief radius in
the die corner, rather a sliding bottom wall on the exit
channel after the patented design by Segal et al. [53].
Such a design has been shown to reduce the strain inho-mogeneity discussed above. The die was operated within
Tempera Grain size (lm)
O 50 [51]
billet n/a Large (mm)
m B) F 6 [32]
mm B) F 140
m plate) F Bimodal � 10, 50 [51]
m B) F 6.5
m B) F 38
Table 3
Processing conditions explored during this investigation
Alloy/condition Temp. (�C) Passes/route
AZ31/plate 300 1, 2, 4, 8A
AZ31/cast 300, 200BPa 1A and 2, 8B
AZ31/extr. 200BP 1, 2A and 2, 8B
AZ80/extr. 200BP 1A and 2, 8B
ZK60/extr. 260 1, 2, 4, 8A and 4B and 4C
WE43/extr. 325 2, 4A
Mg4L/extr. 260 1, 2, 4A
BP, back-pressure.a Initial pass at 300 �C, subsequent passes at 200 �C.
3138 S.R. Agnew et al. / Acta Materialia 53 (2005) 3135–3146
a 445 kN universal testing machine. Billets were ma-
chined from the as-received materials with dimensions
of 22 · 22 · 100–150 mm for ECAE processing. It was
necessary to process all of these alloys at elevated-tem-
peratures ranging from 175 to 325 �C (see Table 3) in or-
der to avoid fracturing the billets. Depending upon the
alloy, a temperature between 200 and 300 �C was first
attempted in view of hot-workability data from the liter-ature [1]. The temperature adopted was the lowest at
which successful extrusions were possible, in order to
promote grain refinement (see Table 3). The whole
ECAE die is heated to the desired temperature using
built-in cartridge heaters. The billets were lubricated
with MoS2 paste, inserted into the die, held until they
reached the processing temperature (approximately
25 min), and, then, pressed at a rate of 25 mm/min (or12 mm/min in the case of alloy WE43). Samples sub-
jected to multi-pass processing were machined slightly
to allow them to fit back in the entrance channel.
Some of the ECAE runs were performed with a back-
pressure enforced by a second ram and hydraulic cylin-
der on the exit channel. The back-pressure superimposes
a hydrostatic compressive stress on top of the primarily
shear stress at the intersection of the channels in order toprevent shear failures. For example, an initial attempt to
process alloy AZ31 at 200 �C or below resulted in cata-
strophic shear failures (Fig. 2). However, processing the
alloy at 300 �C yielded only marginal grain refinement
Fig. 2. Shear failures exhibited by the alloy AZ31 ECAE processed at (a) 20
Use of a backpressure during 200 �C processing prevented such failures. (Th
[51]. Enforcing a back-pressure of 58–77 MPa (as com-
pared with the �300–350 MPa entrance pressure) pre-
vented the shear failures mentioned above and enabled
processing at a temperature of 200 �C, which resulted
in substantial grain refinement [32]. Attempts to process
at lower temperatures (175 �C) were unsuccessful evenwith a back-pressure (Fig. 2(b)).
2.4. Modeling texture evolution
It will be shown that ECAE processing results in dif-
ferent texture evolutions for the different classes of
alloys. The viscoplastic self-consistent (VPSC) polycrys-
tal modeling scheme originally proposed by Hill [54] andimplemented later by Hutchinson [55] has been shown to
be very effective for modeling the plastic response and
texture evolution of non-cubic metals [56]. The details
of the particular VPSC algorithm employed in the pres-
ent work are explained in the paper by Lebesohn and
Tome [56]. In brief, the texture of the polycrystalline
aggregate is represented by assigning a finite number
(�1000) of discrete crystallographic orientations a vol-ume fraction. The resulting discrete texture approxi-
mates the more continuous orientation distribution
measured using X-ray diffraction. Each orientation is as-
signed an anisotropic viscoplastic constitutive response
characteristic of the single-crystal, e.g. 57, where all
the slip and twinning mechanisms have been assigned
a high stress exponent, n = 20. This exponent is not in-
tended to explicitly reflect the overall rate-sensitivity ofthe material. Rather, it serves to round the vertices of
the single crystal �yield surface�, providing a mathemati-
cally convenient and unambiguous connection between
the stressing direction and the resulting strain rate.
The coupling of the grain-level response with the
aggregate is accomplished using a self-consistent
homogenization scheme based upon Eshelby�s inclusionformalism [58]. The interaction strength is assignedthrough the parameter neff = 10, which is an intermedi-
ate value between the compliant tangent modulus
0 �C without any back-pressure, and (b) 175 �C with a back-pressure.
e billets� widths and thicknesses are 22 mm.)
S.R. Agnew et al. / Acta Materialia 53 (2005) 3135–3146 3139
approach, neff = n = 20, and the more rigid secant mod-
ulus approach, neff = 1. Simulations are performed in an
incremental fashion and after each small straining step
(De � 0.02), the grains� shapes and individual orienta-
tions are updated to account for the crystallographic
rotations due to slip on the specified slip systems, andthe critical resolved shear stresses may be updated to ac-
count for a variety of effects ranging from strain harden-
ing, to latent hardening, Bauschinger-type effects.
Recent implementations have even linked these latter
phenomena to microstructure features, such as disloca-
tion substructures [59,60].
In order to explore the influence of different slip
systems on the texture evolution, different criticalresolved shear stresses (CRSSs) were assigned to
the various slip and twinning modes known to be ac-
tive within magnesium h1�210i(0002) or basal Æaæ,h1�210i f10�10g or prism Æaæ, h11�23if11�2�2g or pyra-
midal Æc + aæ, and h10�1�10if10�12g tensile twinning.
For the current simulations, all the individual defor-
mation mechanisms were assumed to behavior in a
perfectly plastic fashion, which is consistent with theobservation that metals may not significantly harden
during hot deformation. Twinning is treated distinctly
from dislocation slip in that the CRSS in the anti-
twinning direction is set to a very high (impossible)
stress level, and the crystallographic reorientation
associated with twinning is accounted for using a pre-
dominant twin reorientation scheme described else-
where [61]. It is noted that neither dynamic recoverynor recrystallization is explicitly modeled, although
their role in the deformation process of these alloys
at these temperatures is undeniable. Additionally,
changing the stress exponent, n, introduced above
would also affect secondary slip mode activity (lower
n values would promote more secondary slip). How-
ever, rather than allow multiple factors to influence
the activity of slip systems, the present investigationfocuses on the direct connection between a slip mode�sactivity and its CRSS value.
The overall boundary conditions of the ECAE pro-
cess are approximated by a simple shear within an infin-
itesimal layer aligned with the plane of intersection
between the two channels, as suggested by the inventor
[23,53,62]. For the present die with an included angle,
u = 90�, the deformation gradient, F 0, and the corre-sponding velocity gradient, L 0, are:
F0 ¼1 �2 0
0 1 0
0 0 1
0B@
1CA and L0 ¼
0 �2 _c 0
0 0 0
0 0 0
0B@
1CA ð1Þ
expressed within the rotated frame labeled x 0 and y 0 in
Fig. 1 (and _c is the nominal strain rate). A transforma-
tion of basis allows expressing these within the labora-
tory coordinate frame, x and y,
F ¼2 �1 0
1 0 0
0 0 1
0B@
1CA and L ¼
_c � _c 0
_c � _c 0
0 0 0
0B@
1CA: ð2Þ
This approach ignores the fact that the actual deforma-
tion takes place over a finite width, and that there maybe a finite radius at the intersection of the dies by design
or by virtue of a ‘‘dead metal zone’’ in the corner of the
die [63]. Some researchers have painstakingly deter-
mined the specifics of the strain path experienced by
the work-piece as it passes through the die either exper-
imentally [5,64,65] or theoretically [11,66–69]. It is gen-
erally agreed that the overall deformation is well
described by Eqs. (1) and (2). However, some research-ers have concluded that it is necessary to include details
of the strain path, while others have obtained good re-
sults assuming the simplified view. In order to model
multi-pass ECAE, it is emphasized that the material
must undergo a right hand rotation of 180� � u( = 90�for the present die) about the z-axis (shown in Fig. 1)
prior to re-insertion in the die, regardless of the process-
ing route
RA ¼0 1 0
�1 0 0
0 0 1
0B@
1CA: ð3Þ
For route A, this is sufficient and additional rotations
will apply for the other routes. This is a critical issue that
many researchers appear to have over-looked and led
them to suggest that route A processing is nothing morethan sequential simple shear of the same sense in the
same direction. As documented previously [24,70], route
A processing involves shearing on an intersecting planes
and may be modeled by enforcing the following defor-
mation for even numbered passes (2,4,6,8, . . .)
F00 ¼0 �1 0
1 2 0
0 0 1
0B@
1CA and L00 ¼
� _c � _c 0
_c _c 0
0 0 0
0B@
1CA
ð4Þand that expressed by Eq. (2) used for odd numbered
passes (1,3,5,7, . . .). For other routes, it may be more
convenient to rotate the material frame of reference
(e.g., for describing the material�s texture), rather thanthat of the deformation, in between each subsequent
pass.
3. Results
3.1. Initial textures
The initial textures of pre-extruded samples (Table 2)
are typical of magnesium extrusions (Fig. 3) having the
Fig. 3. Inverse pole figures showing the dominant h10�10i fiber texture in all the extruded magnesium alloys. ZK60 and MgLi are most strongly
textured, WE43 is most weakly textured, and the large AZ80 forging billet shows a distinct secondary Æ0001æ fiber.
3140 S.R. Agnew et al. / Acta Materialia 53 (2005) 3135–3146
basal planes aligned with the extrusion axis and a ten-
dency for the h10�10i directions to align with the extru-
sion axis. Alloys ZK60 and Mg–Li have the strongest
initial textures, and WE43 is the most weakly textured.
The texture of the large AZ80 extrusion is distinct, in
that there is also a secondary Æ0001æ fiber aligned with
extrusion axis. The effect of initial texture on the texture
evolution during ECAE was examined using two otherinitial conditions of alloy AZ31: a plate with an axis-
symmetric Æ0001æ fiber aligned with the plate normal
with an intensity of four multiples of a random distribu-
tion (MRD) and a weakly textured direct-chill cast
billet.
3.2. The effect of alloying
The initial textures (labeled 0) and the texture ob-
served after 1, 2 and 4 passes by route A are shown in
Fig. 4. All the texture results are presented in terms of
recalculated (0002) and ð10�10Þ pole figures from the
flow plane (i.e., the plane which contains the entrance
and exit channels) with the exit direction (the x-axis in
Fig. 1) to the right. Due to the 2-fold rotational symme-
try of the process (about the flow plane normal, thez-axis in Fig. 1), and the symmetry of the experimental
pole figures, the orientation distributions and full pole
figures were calculated assuming monoclinic symmetry.
The pole figures are, thus, completely represented by
their upper half in order to save space.
After a single pass, alloy AZ31 essentially exhibits a
Æ0001æ fiber texture (Fig. 4(a)) with the dominant c-axis
fiber is oriented approximately 20� aft of vertical (they-axis in Fig. 1). Subsequent passes serve to strengthen
this texture, although it is noted that there are enormous
rigid body rotations (Eqs. (2) and (3)) and plastic rota-
tions inherent to the process. Therefore, this result is
not evidence of texture stabilization, as might occur dur-
ing monotonic straining, such as multi-pass rolling.
Rather, it demonstrates reorientation to a very similar
texture after each subsequent pass. After subsequent
passes, there is also a notable strengthening of the
h10�10i parallel to (or i) z. Finally, there is a trend for
the dominant Æ0001æ texture component to split into
two components, one closer to the z-axis. Alloy AZ80
processed under the same conditions adopted an essen-
tially identical texture, which is not shown in the interest
of space.
The texture evolution of alloy ZK60 (Fig. 4(b)) showsthree major distinctions from that observed for the AZ
alloys after a single pass. (i) The major Æ0001æ fiber is
oriented closer to the y-axis; (ii) there is a secondary fi-
ber Æ0001æ iz; and (iii) h11�20ik z, placing maxima in the
ð10�10Þ pole figure 30� away from the z-axis. During
subsequent passes, the secondary texture component
Æ0001æ iz weakens, and the dominant fiber �5� aft of
the y-axis. The texture after 8 passes by route A is verysimilar to that shown after 4 passes, albeit with higher
texture strength, with peak intensities in the basal pole
figure of �8 multiples of a random distribution
(MRD), in comparison with �6 MRD for 4 passes. Al-
loy WE43 is designed for high temperature strength and,
thus, required the highest processing temperature
(325 �C) of all the alloys examined. Although it exhibits
slightly weaker textures throughout (maximum peakintensities of �4 in the basal pole figure), the texture
evolution of alloy WE43 (Fig. 4(c)) is most similar to
that of the other fine-grained Zr-containing alloy in this
study, ZK60.
The experimental Mg–Li alloy exhibits the most dis-
tinct texture evolution (Fig. 4(d)), which is not surpris-
ing in view of the unique properties and deformation
mechanisms already reported for this alloy (e.g., Refs.[46–49]. The initial extrusion texture is similar to the
other alloys due to the high symmetry of the extrusion
process itself. The lower symmetry of the ECAE process
reveals a strong distinction between Mg–4wt%Li and
the other alloys. The dominant texture component in
all the other alloys (Æ0001æ iy) becomes secondary, while
the Æ0001æ iz dominates the texture by the 4th pass
through the die. This appears consistent with the prior
Fig. 4. (0002) and ð10�10Þ pole figures (upper half) show initial extrusion textures (labeled 0) and ECAE textures after 1, 2, and 4 passes by route A.
The distinctions suggest that different deformation mechanisms operate in the different alloys. The extrusion axis is to the right, and the intensity scale
is the same as in Fig. 3 in this figure and all that follow.
S.R. Agnew et al. / Acta Materialia 53 (2005) 3135–3146 3141
observation of enhanced non-basal slip in this alloy, in
particular, the observation that prismatic slip dominated
the single-crystal deformation of Mg at elevated
temperatures [47].
3.3. The effect of initial texture
The hot-rolled AZ31 plate has a very distinct initial
texture compared to the extrusions discussed above,
however, the resulting ECAE textures (Fig. 5) are simi-
lar. The primary distinction is the complete absence of
any Æ0001æ iz, which most likely results from the factthat the initial texture places essentially all the grains
with their c-axes far from this orientation. A second dis-
tinction is that the dominant Æ0001æ fiber component is
closer to the z-axis after ECAE. Taking a most simplistic
view, the ECAE textures shown in Fig. 5 are similar to
the initial plate texture. However, the initial plate tex-
ture exhibits an orthotropic symmetry characteristic of
the rolling process, while the ECAE textures exhibit alower (monoclinic) symmetry characteristic of ECAE
[25]. Proceeding to 8 passes by route A only succeeds
in strengthening the texture. Beginning with a very
weakly textured, but coarse-grained, as-cast AZ31 mate-
rial resulted in an ECAE texture evolution almost iden-
tical to that shown for the extruded AZ31.
3.4. Modeling the texture evolution during ECAE, route A
In all cases, it has been assumed that the basal Æaæ slipmechanism is the easy slip mechanism and, therefore, it
is shown to accommodate most of the strain and leads to
the overall similarity between the textures of the various
alloys (see Figs. 3–5). In most of the simulations, the
three slip modes explored were basal Æaæ, prismatic Æaæ,and pyramidal Æc + aæ slip. In a few cases, the pyramidalÆaæ slip and f10�12g tension twinning modes were
included.
3.4.1. Modeling the effect of the non-basal slip mode
activity
The activities of the non-basal slip modes are var-
ied by assuming different critical resolved shear stress
(CRSS) values, relative to the CRSS for basal slip(sbasal = 1). A range of CRSS values from 2 to 8 is
explored for both prismatic and Æc + aæ slip modes.
Fig. 6 shows the simulated textures after a single
ECAE pass, where the initial texture was assumed
Fig. 5. (0002) and ð10�10Þ pole figures of AZ31, from initially (a) hot-rolled plate (labeled 0) or (b) cast billet and ECAE textures after 1, 2, and 4
passes by route A.
Fig. 6. (0002) and ð10�10Þ pole figures of simulated single-pass ECAE textures beginning with an initially random texture. Labels indicate relative
CRSS values of the basal, prismatic, and pyramidal Æc + aæ slip modes used in the simulations.
3142 S.R. Agnew et al. / Acta Materialia 53 (2005) 3135–3146
to be random. The contour plots presented have been
smoothed over 10� in keeping with the fact that the
initial discrete textures were determined for a
10� · 10� · 10� grid in Euler space. In all cases, basal
slip dominates the strain accommodation and causes a
dominant basal fiber texture to evolve. The clearest
case of this is labeled 1-8-8, indicating that there isa relatively high CRSS value for both of the non-ba-
sal slip modes sprismatic = sÆc + aæ = 8. The result is a
simple Æ0001æ fiber component tilted �20� aft of
the y-axis. Non-basal slip mechanisms accommodate
no more than �20% of the strain individually, with
the prismatic Æaæ slip mode being about twice as ac-
tive as the Æc + aæ.Increasing the activity of the Æc + aæ slip system up to
�20% by lowering its CRSS value to sÆc + aæ = 4 leads to
a splitting of the dominant Æ0001æ fiber into two compo-
nents, with one closer to the y-axis. There is a corre-
sponding weakening of the intensity in the ð10�10Þpole figure, with some tendency for h10�10ik z-axis. On
the other hand, if the prismatic slip mode becomes more
active (accommodating up to �40% of the strain for the
cases where sprismatic 6 4), orientations close toÆ0001æ iz are stabilized and the band in the ð10�10Þ polefigure splits into peaks corresponding to h11�20ik z.
Simulations incorporating the pyramidal Æaæ slip
mode yield very similar results to those involving
prismatic slip, since a combination of prismatic and
basal slip gives rise to the same shear strains and
crystallographic rotations as pyramidal Æaæ slip.
Therefore, to keep the simulation parameters to a
minimum, the non-basal slip of Æaæ dislocations is
considered collectively and only prismatic slip is mod-
eled in practice [71]. Additionally, while f10�12g ten-sion twinning is known to be a very important
deformation mechanism at low temperatures, it is less
active at elevated temperatures [2]. A few simulations
including tension twinning were performed, however,
the texture components which resulted are not ob-
served experimentally.
3.4.2. Modeling the effect of the initial texture
Simulations beginning with an initial texture, which
models the AZ31 extrusion or plate texture, exhibit
more non-basal slip mode activity than an initially ran-
dom texture. Even with the relative CRSS values set
quite high, sprismatic = sÆc + aæ = 8, the prismatic slip sys-
tem accommodates �25% of the strain for the case of
the initial extrusion texture. This trend results in a stron-
ger Æ0001æ iz fiber component (Fig. 7) than observed forthe initially random texture. Similarly, decreasing the
CRSS for the Æc + aæ slip mode to sÆc + aæ = 4 results in
�40% of the strain being accommodated by that mode
throughout the deformation.
Fig. 7. Pole figures showing textures simulated using an initial extrusion texture. Again, the labels indicate the relative CRSS values of the basal,
prismatic, and pyramidal Æc + aæ slip modes.
Fig. 8. Pole figures showing textures simulated using an initial plate
texture. Again, the labels indicate the relative CRSS values of the
basal, prismatic, and pyramidal Æc + aæ slip modes.
S.R. Agnew et al. / Acta Materialia 53 (2005) 3135–3146 3143
When the CRSS for the prismatic slip mode is quite
low, sprismatic = 2, the Æ0001æ iz component becomes
the dominant feature in the texture labeled 1-2-8. Corre-spondingly, there are strong texture components within
the flow plane shown in the ð10�10Þ pole figure. The caseinvolving extensive Æc + aæ slip (�35% of the strain
accommodation) labeled 1-8-4 is similar to that simu-
lated from a random texture.
Since only alloy AZ31 was processed from plate mate-
rial, only CRSS conditions appropriate for simulating
the AZ31 texture evolution were explored. Fig. 8 showsthe impact of increasing the contribution of the Æc + aæslip mode. Again, increasing the activity of Æc + aæ causesthe dominant Æ0001æ fiber to split into two components
and with the stronger of the two rotated close to the
y-axis.
3.4.3. Modeling multi-pass ECAE texture evolution
The metallographic investigation of some ECAE pro-cessed magnesium alloy sample [51] reveals that the
grain shapes after ECAE are equiaxed. This observation
combined with the assumption of no slip system harden-
Fig. 9. Pole figures showing textures simulated after a second route A pass
simulations performed with the same CRSS ratios.
ing reduces the problem of multi-pass simulation to an
issue of initial texture. As for the case of the plate tex-
ture above, the focus here will be on simulating the tex-
ture evolution of alloy AZ31, thus the simulations wereall performed with the CRSS combination of 1-8-6 (i.e.,
sprismatic = 8 and sÆc+ aæ = 6).
The four cases in Fig. 9 show the texture evolution
during a second ECAE pass by route A, beginning with
simulated single pass textures with (a) random (Fig. 6),
(b) plate (Fig. 8) and (c) extrusion (Fig. 7) initial tex-
tures. Each of these cases appear rather similar, how-
ever, beginning with the random texture has theultimate result of promoting significant activity of the
prismatic slip mode during the second pass, which re-
sults in components not observed experimentally.
Finally, second pass textures beginning with the experi-
mental textures after a single pass from the extrusion
(Fig. 4) and plate (Fig. 5) textures were simulated. This
allows exploring the possibility that simulating multiple
passes results in compounding errors. The result fromthe experimental plate texture was very similar to that
of the simulated (labeled Plate), while the subtle distinc-
tion between the simulated extrusion (labeled Extrusion)
and the experimental (labeled ECAE) are noteworthy.
4. Discussion
While all five of the examined alloys develop qualita-
tively similar h10�10i fiber texture during conventional
axis symmetric extrusion (see Fig. 3), the lower symme-
try of the ECAE process results in the significant quali-
tative differences between the ECAE textures of the five
alloys shown in Fig. 4.
with different initial textures as labeled and described in the text. All
3144 S.R. Agnew et al. / Acta Materialia 53 (2005) 3135–3146
4.1. Comparisons of experimental and simulated textures
Comparisons of the experimental and simulated tex-
tures yield a number of insights regarding the activities
of the different non-basal slip systems within the various
magnesium alloys. Ultimately, the mechanistic hypothe-ses developed through this sort of inverse approach are
best examined through direct observation (e.g., using
transmission electron microscopy).
The texture evolution of alloy AZ31 does not show
any of the signatures of the extensive non-basal slip of
Æaæ dislocations, such as Æ0001æ and h11�20ik z-axis,which are evidenced by all the other alloys (Fig. 4)
and in the simulations involving significant non-basalÆaæ slip (labeled 1-2-8 and 1-8-8 in Figs. 6 and 7). Rather,
the AZ31 textures are similar to those simulated with
conditions of balanced non-basal Æaæ and Æc + aæ slip (la-
beled 1-8-6 in Figs. 6 and 7). Finally, the texture evolu-
tion of the initially cast (randomly textured) AZ31 is
similar to the pre-extruded material. This is particularly
significant because these samples spanned a wide range
of initial grain-sizes, and the similarities between theECAE textures of cast and extruded suggest grain-size
does not significantly influence the texture evolution.
The texture evolution of the pre-extruded alloy AZ80
is very similar to that of extruded AZ31 (not shown here
in the interest of space). The similarity between the two
alloy�s textures also holds for route B processing [44].
This not surprising since the relative strengths of the dif-
ferent slip modes are most strongly affected by the solidsolution alloying. Alloy AZ31 already contains more
than the equilibrium solid solubility of aluminum. Thus,
alloy AZ80 has a similar solid solution Al content, be-
cause the additional Al is contained in second phase
particles.
The experimental texture evolution of alloy ZK60 is
not precisely modeled by any of the simulated textures
shown in Figs. 6 and 7. Nevertheless, the case labeled1-8-4 in Fig. 7 shows the most similarity. The experi-
mental texture after 1 pass exhibits some of the signa-
tures of the prismatic slip mode�s activity, such as
Æ0001æ and h11�20i components iz-axis). However, sub-
sequent passes show more tendency for the splitting of
the Æ0001æ fiber close to the vertical axis, which is a sig-
nature of extensive Æc + aæ slip, and an absence of any
h10�10i components within the flow plane, which is asignatures of extensive prismatic slip. Alloy WE43
exhibits a qualitatively similar texture evolution, how-
ever, it is weaker at every stage. This trend may be
due to a randomizing effect related to particle stimulated
nucleation during recrystallization [72].
The final alloy examined in this study, Mg–4wt%Li
shows the most similarity to the case labeled 1-2-8 in
Fig. 7, in particular the sharp texture componentsshown within the flow plane of the ð10�10Þ pole figure.
Thus, it appears that extensive prismatic Æaæ slip occurs
in this alloy, as suggested in a number of previous stud-
ies, specifically in the study of elevated temperature
deformation of Mg–Li single-crystals which showed
only non-basal slip Æaæ above a certain temperature
[47], as opposed to the dominant role played by the ba-
sal slip mode in all other alloy and temperaturecombinations.
Finally, it is well known that these alloys do undergo
dynamic recrystallization (DRX) at the observed tem-
peratures and strain rates employed in these studies
[72,73]. However, the effect of DRX on the texture evo-
lution of magnesium is not definitively known. In fact,
the study of static recrystallization textures in magne-
sium alloys is only in its infancy [74]. Nevertheless, com-parisons of the textures induced by cold-rolling and by
hot-rolling suggest that there is a randomizing effect
associated with hot deformation. Even in cases where
the texture was qualitatively well-modeled by the poly-
crystal simulation, the magnitude of the texture strength
was exaggerated. Additionally, there were cases where
the simulated texture after a single ECAE pass exhibited
characteristics shown experimentally only in the multi-pass results. This suggests that the texture evolution is
slowed by unaccounted for processes, such as recovery
and DRX.
5. Conclusions
Equal channel angular extrusion offers the potentialto induce previously unobserved crystallographic tex-
tures, which may be of significant interest for modifying
the properties. This paper documents some of the vari-
ety of textures which are generated by route A ECAE
processing, for a number of magnesium alloys, with a
range of initial textures and grain-sizes. For a given al-
loy, route A processing appears to result in similar tex-
tures from one pass to the next. However, it isemphasized that the material undergoes large rigid-body
rotations during the course of loading into the die and
during the process itself, so it is not really the same ori-
entations persisting, rather each pass through the die in-
volves a texture evolution that results in a final texture
very similar in appearance to the previous pass. Initial
texture is shown to have a pronounced effect, particu-
larly in dictating the presence or absence of certain tex-ture components which arise due to their rotational
stability.
The lower symmetry of the ECAE process as com-
pared with conventional deformation processes reveals
distinctions in the texture evolutions of the major classes
of wrought magnesium alloys, despite the similarity of
their conventional extrusion textures. In turn, these tex-
ture evolutions are used to extract information aboutthe relative activities of the deformation mechanisms,
which accommodate the plastic deformation, using an
S.R. Agnew et al. / Acta Materialia 53 (2005) 3135–3146 3145
inverse approach. Alloys AZ31 and AZ80 demonstrate
basal slip dominated deformation with a small balanced
contribution of the non-basal Æaæ and Æc + aæ slip modes.
Alloys ZK60 and WE43 also show similarities with each
other, and a significant contribution from the non-basal
Æc + aæ slip mechanism (in addition to basal slip) appearsto explain most of the observed texture components. Fi-
nally, the unique texture evolution of the Mg–Li solid
solution exhibits a number of characteristics of simu-
lated textures resulting from extensive activity of the
prismatic Æaæ slip mode. Although the simulated textures
mentioned above qualitatively reproduce many of the
experimental observations, there is a tendency to over-
predict the strength of the textures. A more detailedunderstanding of texture evolution during the hot defor-
mation of magnesium alloys awaits a better understand-
ing of the effect of dynamic recovery and
recrystallization processes on the texture evolution.
Acknowledgments
This research was supported by a grant from the US
Department of Energy, Office of FreedomCAR at
ORNL and INEEL. The encouragement of our pro-
gram managers, P.S. Sklad and S. Diamond, is grate-
fully acknowledged. National Science Foundation
Grants EEC-0203415 and DMR-0231320 to the Univer-
sity of Tennessee are also acknowledged. Thanks to
C.A. Carmichael and K. Blakeley of ORNL, for theirhelp with casting and extrusion of experimental alloys,
and to B. MacDonald (Timminco), P. Chaudhury (for-
merly with Intercontinental Manuf.), and K. Clark
(Reade Manuf.) for providing the commercial alloys
examined in this study. M. Olson is acknowledged for
performing some of the texture measurements.
References
[1] Avedesian MM, Baker H, editors. Magnesium and magnesium
alloys ASM specialty handbook. Metals Park (OH): ASM
International; 1999.
[2] Barnett MR, Keshavarz Z, Beer AG, Atwell D. Acta Mater
2004;52:5093–103.
[3] Kocks UF. In: Kocks UF, Tome CN, Wenk H-R, editors,
Texture and anisotropy. Cambridge; 1998.
[4] Agnew SR, Yoo MH, Tome CN. Acta Mater 2001;49:4277–89.
[5] Ferasse S, Segal VM, Hartwig KT, Goforth RE. Metall Mater
Trans A 1997;28:1047–57.
[6] Furukawa M, Iwahashi Y, Horita Z, Nemoto M, Langdon TG.
Mater Sci Eng A 1995;197:328.
[7] Gil Sevillano J, van Houtte P, Aernoudt E. Prog Mater Sci
1981;25:69–411.
[8] Hughes DA, Nix WD. Metall Trans A 1988;19:3013–24.
[9] MacCagno TM, Yues S, Jonas JJ, Dyck K. Metall Trans A
1993;24:1589–96.
[10] McQueen HJ, Ryan ND. Mater Sci Eng A 2002;322:43–63.
[11] Semiatin SL, Delo DP, Shell EB. Acta Mater 2000;48:1841–51.
[12] Valiev RZ, Kornikov AV, Mulyokov RR. Mater Sci Eng A
1993;168:141–8.
[13] Langford G, Cohen M. Trans ASM 1969;62:623–38.
[14] Langford G, Cohen M. Metall Trans A 1975;6:901–10.
[15] Kuhlmann-Wilsdorf D, Hansen N. Metall Trans A 1989;20:
2393–7.
[16] Valiev RZ, Islamgaliev RK, Alexandrov IV. Prog Mater Sci
2000;45:103–89.
[17] Lee S, Berbon PB, Furukawa M, Horita Z, Nemoto M, Tsenev
NK, Valiev RZ, Langdon TG. Mater Sci Eng A 1999;272:63–72.
[18] Horita Z, Furukawa M, Nemoto M, Barnes AJ, Langdon TG.
Acta Mater 2000;48:3633–40.
[19] Legros M, Elliott BR, Rittner MN, Weertman JR, Hemker KJ.
Philos Mag A 2000;80:1017–26.
[20] Malow TR, Koch CC. Acta Mater 1998;46:6459–73.
[21] Zhu YTT, Liao XZ. Nat Mater 2004;3:351–2.
[22] Gholinia A, Prangnell PB, Markushev MK. Acta Mater
2000;48:1115–30.
[23] Segal VM. US patent # 5513512; 1996.
[24] Agnew SR, Kocks UF, Hartwig KT, Weertman JR. In: Cartensen
JV, editor. Proceedings of the 19th Risoe international sympo-
sium. Roskilde, Denmark: Risoe National Lab; 1998. p. 201–6.
[25] Agnew SR. In: Szpunar JA, editor. Proceedings of the ICOTOM-
12. Ottawa, Canada: NRC Reseach Press; 1999. p. 575–80.
[26] Toth LS, Zehetbauer M, Kopacz I, Alexandrov IV. In: Proceed-
ings of the THERMEC �2000, December 4–8, Las Vegas, USA
[CD, ISBN 0080440266].
[27] Gholinia A, Bate P, Prangnell PB. Acta Mater 2002;50:2121–36.
[28] Beyerlein IJ, Lebensohn RA, Tome CN. Mater Sci Eng A
2003;345:122–38.
[29] Gibbs MA, Hartwig KT, Cornwell LR, Goforth RE, Payzant EA.
Scr Mater 1998;39:1699–704.
[30] Field RD, Hartwig KT, Necker CT, Bingert JF, Agnew SR.
Metall Mater Trans A 2002;23:965–72.
[31] Kim WJ, An CW, Kim YS, Hong SI. Scr Mater 2002;27:39.
[32] Agnew SR, Horton JA, Lillo TM, Brown DW. Scr Mater
2004;50:377–81.
[33] DeLo DP, Bieler TR, Semiatin SL. In: Mishra RS, et al, editors.
Ultrafine grainedmaterials. Warrendale (PA): TMS; 2000. p. 257.
[34] Kim HS, Joo DH, Kim MH, Hwang SK, Kwun SI, Chae SW.
Mater Sci Tech 2003;19:403–5.
[35] Roberts CS. Magnesium and its alloys. New York: Wiley; 1960.
p. 84.
[36] Agnew SR, Duygulu O. Int J Plasticity 2005;21:1161.
[37] Koike J, Kobayashi T, Mukai T, Watanabe H, Suzuki M,
Maruyama K, et al. Acta Mater 2003;51:2055–65.
[38] Thompson RW, Codd I, Douthwaite RM, Petch NJ. Philos Mag
1962;7:45.
[39] Meyers MA. In: Meyers MA, Armstrong RW, Kirchner HOK,
editors. Mechanics and materials. New York: Wiley; 1999. p.
377–81.
[40] Chapman JA, Wilson DV. Bull J Inst Met 1962/3;91:39.
[41] Mabuchi M, Ameyama K, Iwasaki H, Higashi K. Acta Mater
1999;47:2047.
[42] Kamado S, Ashie T, Yamada H, Sanbun K, Kojima Y. Mater Sci
Forum 2000;350/351:65.
[43] Agnew SR, Stoica GM, Chen LJ, Lillo TM, Macheret J, Liaw
PK. In: Zhu YT, et al, editors. Ultrafine grained materials
II. Warrendale (PA): TMS; 2002. p. 643–52.
[44] Agnew SR, Mehrotra P, Lillo TM, Stoica GM, Liaw PK, Mater
Sci Eng A [submitted].
[45] Unsworth W, King JF, Bradshaw SL. US Patent # 4401621;
1983.
[46] Hauser FE, Landon PR, Dorn JE. Trans ASM 1958;50:856–83.
[47] Ward Flynn P, Mote J, Dorn JE. Trans TMS-AIME
1961;221:1148–54.
[48] Ando S, Tonda H. Mater Trans JIM 2000;41:1188–91.
3146 S.R. Agnew et al. / Acta Materialia 53 (2005) 3135–3146
[49] Agnew SR, Horton JA, Yoo MH. Metall Mater Trans A
2002;23:813–22.
[50] Mill K, editor. Metallography and microstructures. ASM hand-
book, vol. 9. Materials Park (OH): ASM; 1978.
[51] Agnew SR, Lillo TM, Stoica GM, Liaw PK. Magne-
sium technology 2001. Warrendale (PA): TMS; 2001. p.
243–8.
[52] Kallend JS, Kocks UF, Rollet AD, Wenk H-R. Mater Sci Eng A
1991;132:1–11.
[53] Segal VM, Goforth RE, Hartwig KT. US Patent #5,400,633;
1995.
[54] Hill R. J Mech Phys Solids 1965;13:89–101.
[55] Hutchinson JW. Proc Roy Soc A 1976;348:101–27.
[56] Lebensohn RA, Tome CN. Acta Metall 1993;41:2611–24.
[57] Asaro RJ, Needleman A. Acta Metall 1985;6:923–53.
[58] Eshelby JD. Proc Roy Soc Lond A 1957;241:376.
[59] Peeters B, Teodosiu C, Kalidindi SR, Van Houtte P, Aernoudt E.
J Phys IV France 2001;11:119–26.
[60] Peeters B, Van Houtte P, Kalidindi SR, Aernoudt E. J Phys IV
France 2001;11:365–72.
[61] Tome CN, Lebensohn RA, Kocks UF. Acta Metall Mater
1991;39:2667–80.
[62] Segal VM. Mater Sci Eng A 1995;197:157–64.
[63] Iwahashi Y, Wang J, Horita Z, Nemoto M, Langdon TG. Scr
Mater 1996;35:143–6.
[64] Wu Y, Baker I. Scr Mater 1997;37:437–42.
[65] Bowen JR, Gholinia A, Roberts SM, Prangnell PB. Mater Sci
Eng A 2000;287:87–99.
[66] Prangnell PB, Harris C, Roberts SM. Scr Mater 1997;37:983–9.
[67] Stoica GM, Liaw PK. JOM 2001;53:36–40.
[68] Toth LS, Massion RA, Germain L, Baik SC, Suwas S. Acta
Mater 2004;52:1885–98.
[69] Beyerlein IJ, Tome CN. Mater Sci Eng A 2004:171–90.
[70] Furukawa M, Horita Z, Langdon TG. Mater Sci Eng A
2002;332:97–109.
[71] Agnew SR, Tome CN, Brown DW, Holden TM, Vogel SC. Scr
Mater 2003;48:1003–8.
[72] Mackenzie LWF, Lorimer GW, Humphreys FJ, Wilks T. Mater
Sci Forum 2004;467–470:477–82.
[73] Galiyev A, Kaibyshev R, Gottstein G. Acta Mater 2001;49:
1199–207.
[74] Nadella RK, Samajdar I, Gottstein G. In: Kainer KU, editor.
Proceedings of the magnesium alloys and their applica-
tions. Weinheim: Wiley–VCH; 2003. p. 1052–7.
[75] Mukai T, Yamanoi M, Watanabe H, Higashi K. Scr Mater
2001;45:89.
top related