High-pressure torsion for enhanced atomic diffusion and promoting solid-state reactions in the aluminum–copper system

Abstract

This study reports that solid-state reactions occur by the application of high-pressure torsion (HPT) to the Al–Cu system even at low homologous temperature. A bulk form of disc consisting of two separate half-discs of pure Al and pure Cu are processed by HPT at ambient temperature under a pressure of 6 GPa. X-ray diffraction analysis and high-resolution transmission electron microscopy confirm the formation of different intermetallic phases such as Al2Cu, AlCu and Al4Cu9, as well as the dissolution and supersaturation of Al and Cu in each matrix. It is shown that the diffusion coefficient is enhanced by 1012–1022 times during the HPT processing in comparison with the lattice diffusion and becomes comparable to the surface diffusion. The enhanced diffusion is attributed to the presence of a high density of lattice defects such as vacancies, dislocations and grain boundaries produced by HPT processing.

Full text

1
Acta Materialia 61 (2013) 3482–3489
Received 22 January 2013; Accepted 24 February 2013; Available online 2 April 2013
High-pressure torsion for enhanced atomic
diffusion and promoting solid-state
reactions in the aluminum-copper system
Keiichiro Oh-ishi a, Kaveh Edalati b,c,*, Hyoung Seop Kim d, Kazuhiro Hono a,
Zenji Horita b,c
a National Institute for Materials Science, 1-2-1 Sengen, Tsukuba 305-0047, Japan
b Department of Materials Science and Engineering, Faculty of Engineering, Kyushu University,
Fukuoka 819-0395, Japan
c WPI, International Institute for Carbon-Neutral Energy Research (WPI-I2CNER), Kyushu
University, Fukuoka 819-0395, Japan
d Department of Materials Science and Engineering, Pohang University of Science and Technology,
Pohang 790-784, Republic of Korea
Abstract
This study reports that solid-state reactions occur by the application of high-pressure torsion (HPT)
to the Al–Cu system even at low homologous temperature. A bulk form of disc consisting of two
separate half-discs of pure Al and pure Cu are processed by HPT at ambient temperature under a
pressure of 6 GPa. X-ray diffraction analysis and high-resolution transmission electron microscopy
confirm the formation of different intermetallic phases such as Al2Cu, AlCu and Al4Cu9, as well as
the dissolution and supersaturation of Al and Cu in each matrix. It is shown that the diffusion
coefficient is enhanced by 1012–1022 times during the HPT processing in comparison with the lattice
diffusion and becomes comparable to the surface diffusion. The enhanced diffusion is attributed to
the presence of a high density of lattice defects such as vacancies, dislocations and grain boundaries
produced by HPT processing.
Keywords: Severe plastic deformation (SPD); Intermetallics; Ultrafine grains; Diffusion
coefficient; Phase transformation
* Corresponding author at: Department of Materials Science and Engineering, Faculty of
Engineering, Kyushu University, Fukuoka 819-0395, Japan. Tel./fax: +81 92 802 2992.
E-mail address: Kaveh.edalati@zaiko6.zaiko.kyushu-u.ac.jp (K. Edalati).

2
1. Introduction
There are several processing methods to achieve solid-state reactions in metallic systems such
as diffusion bonding, mechanical milling of elemental powders, rolling and folding of laminated
thin layers, drawing or extrusion of banded fibrous wires, and so on [1]. The solid-state reactions
can be promoted by increasing temperature and/or by imposing intense plastic strain, and thus both
attempts enable the reactions faster in bulk forms [1]. In particular, severe plastic deformation
(SPD) is gaining much attention in these days because it can attain not only significant grain
refinement of bulk metallic materials [2] but also consolidation and bonding of metallic powders [3],
composites [4] and machining chips [5] at ambient temperatures without the conventional sintering
process.
Typical processes of SPD include equal-channel angular pressing (ECAP), accumulative-roll
bonding (ARB) and high-pressure torsion (HPT) [6]. Solid-state alloying was then achieved with
such processes in different metallic systems: Al-Cu [7], Al-Zn-Mg-Cu [8] and Cu-Cr-Ag [9] by
ECAP; Cu-Ag [10] and Cu-Zr [10] by ARB, Al-Mg [11], Al-W [12,13], Cu-Ag [14], Cu-Ni [15],
Cu-Cr [16], Cu-Co [17], Cu-Fe [17], Cu-W [18], W-Ti [13] and W-Ni [13] and Ni-Al-Cr [19] by
HPT.
Among various SPD processes, HPT may be the most unique process [20]. It provides grain
refinement of hard-to-deform materials such as W [21] and intermetallics [22] and induces phase
transformation because of high pressure [23] or high strain [24]. Furthermore, solid-state reactions
can be achieved with HPT: amorphization in Cu-Zr [25], Cu-Ag [25], Cu-Zr-Ti [26] and Ni-Ti [27],
intermetallics formation in Al-Mg [11], Al-Ni [28] and Al-Ti [29], carbide formation in Cu-Nb-C
[30] and hydride formation in Hf-H [31].
In this study, HPT is applied to a bulk form of discs in the Al-Cu system at ambient
temperature and it is demonstrated that solid-state reactions well undergo during the HPT operation
because of unusually enhanced lattice diffusion.
2. Experimental procedures
Rods of 10 mm in diameter were prepared from high purity Al (99.99%) and Cu (99.96%).
They were annealed for 1 hour at 773 K for Al and at 873 K for Cu. Each rod was cut into two
halves along the longitudinal axis using a wire cutting electrical discharge machine. The half-rods
were sliced to the thicknesses of 0.8 mm. One half disc of Al and one half disc of Cu were placed
together in a circular shallow hole of the lower HPT anvil, as shown in Fig. 1(a). The lower anvil
was then raised to contact with the upper anvil having the same shallow hole at the center. While
applying a pressure of 6 GPa at room temperature, both anvils were rotated with respect to each
other at a rotation speed of 1 rpm and the rotation was terminated after either 1, 10 or 100 turns. The
appearance of a sample before HPT and after 100 turns was shown in Fig. 1(b). The temperature
during HPT operation was measured using a thermocouple placed 10 mm away from the bottom
surface of the upper shallow hole. The temperature reached ~333 K during HPT operation after 100
turns. Details concerning the temperatures rise during HPT was given in earlier papers [32,33]. The
HPT-processed discs were first examined by X-ray diffraction (XRD) analysis using the Co K

3
radiation at a scanning speed of 0.2-0.4 o/min and a step interval of 0.01o.
Figure 1. (a) Schematic illustration of HPT processing and (b) appearance of Al-Cu sample before
HPT and after HPT for 100 turns.
Each HPT-processed disc was cut into two halves along the radial axis using a wire cutting
electrical discharge machine. The cross section of discs was examined using scanning electron
microscopy (SEM) after mechanical polishing. For transmission electron microscopy (TEM) and
scanning-transmission microscopy (STEM), foils were cut from the cross section of the discs at
positions 3 mm from the center and were mechanically ground down to the thicknesses of ~30 m.
Note that the samples were intentionally prepared at 3 mm from the disc center to avoid a low
strained area at the disc center [6] and a dead metal zone at the disc edge [34]. TEM specimens
were further thinned by ion milling at an operating voltage of 4 kV. Microstructure analyses were
conducted using Philips CM200 and TECNAI G2 F30 TEMs.
Elemental mapping was also conducted by the Gatan Imaging Filter Tridium attached on the
TECNAI G2 F30 TEM and the jump ratio method was employed to obtain energy filtered maps.
High-angle annular dark-field (HAADF) observations as well as energy dispersive X-ray
spectroscopy (EDS) were carried out in the STEM mode with the condenser aperture of 100 m and
nanoprobe mode using a beam size of ~3 nm. The chemical compositions and the dissolution of Al
and Cu in each other were calculated by standard less method using a TIA software (TEM control
software, Imaging and Analysis).
3. Results
Figure 2 shows back scattered electron (BSE) SEM images of the cross-sectional views for
the samples after (a) 1 turn and (b-d) 100 turns. The brighter contrast corresponds to Cu-enriched

4
regions because all images were taken by BSE. For the sample after 1 turn, as shown in (a), an Al
layer is present between the two Cu layers and the boundaries between Al and Cu are clear. For the
sample after 100 turns, as in (b), a fine layered structure is visible throughout the disc with a
complex nature. The high magnification views show that the layered structure is curved irregularly
at the center as in (c), and it is heavily distorted at the edge as in (d).
Figure 2. SEM images of Al-Cu samples processed by HPT for (a) 1 turn and (b-d) 100 turns,
where (c) and (d) are magnified views of regions indicated by A and B in (b), respectively.
An XRD profile the sample after 100 turns is shown in Fig. 3. The analysis indicates the
presence of an Al2Cu phase as well as Al and Cu phases. The solid-state reaction and formation of
intermetallic phases by HPT was reported in the other Al-based systems such as Al-Mg, Al-Ni and
Al-Ti [11,28,29].
Figure 3. XRD profiles of Al-Cu sample processed by HPT for 100 turns.

5
Figure 4 shows TEM micrographs and corresponding selected-area electron diffraction
(SAED) patterns for the sample after 10 turns. The microstructure is inhomogeneous over a wide
area and Al-enriched regions, as in (a), and Cu-enriched regions, as in (b), are present separately.
Figure 4(a) shows that thin Cu layers with widths of less than ~100 nm are present in the thick fine
grained Al layers in the Al-enriched regions. It is found that the Al layers consist of grains with the
size of ~500 nm which is much finer than the grain sizes of 1-2 m reported in earlier experiments
using ECAP [35] and HPT [36]. Few dislocations are visible in most of the grains in the Al layers
with smooth and well defined grain boundaries. Figure 4(b) shows that microstructure in the
Cu-enriched region has an ultrafine-grained structure with an average grain size of ~300 nm with a
high density of dislocations within grains. This microstructural feature is similar to the one reported
in pure Cu subjected to ECAP [37] and HPT [38].
Figure 4. TEM micrographs and corresponding SAED patterns of Al-Cu sample processed by
HPT for 10 turns, showing (a) layered structure in Al-enriched region and (b) refine-grained
structure in Cu-enriched region.

6
A TEM micrograph and a corresponding SAED pattern after 100 turns are shown in Fig. 5. It
is apparent that a finer layered structure is developed when compared to that observed after 10 turns.
The SAED analysis exhibits a ring pattern, indicating that the small grains are separated by high
angles of misorientation without preferred crystallographic orientation.
Figure 5. TEM micrograph and corresponding SAED pattern of Al-Cu sample processed by HPT
for 100 turns.
In order to analyze the elemental distributions of Al and Cu, energy filtered images were
taken from the sample after 100 turns. A TEM bright-field image and the corresponding elemental
maps of Al and Cu are shown in Figs. 6(a-c) from a representative area and in Figs. 6(d-e) from
another representative area. Note that the sample is the same as the one used in Fig. 5, but the
images were taken from different areas. Bright contrasts in the Al and Cu maps indicate the
presence of Al-enriched and the Cu-enriched regions. Figure 6(a-c) clearly shows that there are
isolated Cu-enriched regions with the size of 10-30 nm in the Al layers. Some elongated grains are
also visible in the Al layer and it appears that the Cu-enriched particles lie along the grain
boundaries, as indicated by the arrows in Fig. 6(b).
Close inspection of the elemental maps in Fig. 6(e) reveals that there should be four
distinctive regions associated with the difference in contrast. Regions A and D should be based on
Al and Cu, respectively, because they exhibit the brightest and darkest contrasts. Regions B and C
can be intermetallic phases with Al-rich and Cu-rich compositions, respectively. The bright-field
image in Fig. 6(d) shows that all regions consist of ultrafine grains with the sizes of 50 to 500 nm,
where the grain size tends to be smaller with increasing the fraction of Cu. Micro-diffraction was
carried out by positioning a focused beam on regions B and C to identify the phases. The diffraction
patterns from regions B and C are shown in Fig. 7(a) and (b), respectively. It turns out that the
diffraction pattern from region B is consistent with the [131] zone axis pattern of Al2Cu phase and

7
the diffraction pattern from region C is consistent with the [113] zone axis pattern of Al4Cu9 phase.
The calculated diffraction patterns for the Al2Cu and Al4Cu9 phases are shown in Figs. 7(c) and (d),
respectively, for comparison.
Figure 6. (a,d) Bright-field images and corresponding energy filtered elemental maps of (b,e) Al
and (c,f) Cu in Al-Cu sample processed by HPT for 100 turns.
Figure 8 shows an HAADF image of the sample processed by 100 turns, where brighter
contrasts correspond to regions containing more Cu. EDS analysis was then carried out to determine
local chemical compositions by positioning a focused beam on the locations marked 1 to 7 in Fig. 8.
Table 1 documents the compositions obtained by the EDS analysis. Position 1 is the Cu phase
containing ~6 at.% of Al in the form of solid solution. Position 2 must be the Al4Cu9 phase which
corresponds to region C in Fig. 7. Compositions of positions 3, 5 and 7 can be well matched with
the one derived from the Al2Cu phase and therefore, corresponds to region B in Fig. 7. Position 4 is
an supersaturated Al phase where Cu is dissolved by 2 at.%. The composition at position 6 neither
matches with the Al2Cu phase nor with the Al4Cu9 phase. Based on the Al-Cu equilibrium phase
diagram, position 6 is considered to correspond to the AlCu phase, although such a presence was
not detected within the sensitivity limit of XRD and SAED analyses. The changes in the
concentration of Cu from the Cu-enriched region 1 to the Al-enriched region 4 in Fig. 8 clearly
confirm that the reactions are controlled by atomic diffusion.

8
Figure 7. Micro-diffraction patterns obtained from grains (a) B and (b) C in Fig. 6, and simulated
diffraction patterns of (c) Al2Cu in [131] direction and (d) Al4Cu9 in [113] direction.
Figure 8. HAADF image for Al-Cu sample processed by HPT for 100 turns, including positions
where the EDS analysis was conducted (EDS results are given in Table 1).

9
4. Discussion
A question arises from the current investigation why the solid-state reaction and the formation
of intermetallic phases are achieved by the HPT processing. It is reasonable that the atomic reaction
may be enhanced by the following three factors during the HPT operation: (1) an increase in
temperature due to plastic deformation and shear friction, (2) a reduction in atomic diffusion
distance by microstructural refinement as discussed in an earlier report [28] and (3) an increase in
the density of lattice defects such as vacancies, dislocations and grain boundaries. The relation
between the diffusion distance, x, and diffusion coefficient, D, is generally described for a given
period of time, t, as [39]
Dtx
(1)
First, temperature was monitored during HPT operation using a thermocouple located at 10
mm above the bottom surface of the shallow hole on the upper anvil. The temperature gradually
increased and reached 333 K after 20 turns and leveled off during further turns. This measurement
indicates that the temperature of the sample does not exceed 340 K, as discussed in details in Refs.
[32,33]. Furthermore, differential scanning calorimetry measurements confirmed that there were no
peaks appearing at least below the temperature of 473 K by thermal reactions. It is considered that
340 K should not be high enough to promote the solid state reaction as this temperature corresponds
to ~0.4Tm where Tm is the melting temperature taken from the Al- 33 wt% Cu eutectic composition
(the lowest melting point in the Al-Cu system). Therefore, although more quantitative evaluation
will be given below, the solid-state reactions and the formation of the ordered phases cannot be
attributed to the temperature rise during the HPT processing.
Second, as discussed in an earlier report [28], the grains are significantly elongated so that the
widths of the elongated grains are considerably reduced by shear strain introduced by HPT
processing. Considering the diffusion paths for solid-state reaction be equal to the widths of
elongated grains and the time for diffusion be 6000 s corresponding to 100 turns in the HPT
operation, the diffusion coefficients are estimated using Eq.(1) to be D = 10-19-10-17 m2/s as
documented in Table 1 together with the average path lengths for regions 1-7 in Fig. 8. For
comparison, Table 1 also includes diffusion coefficients calculated from the following equation [39]






 RT
PVQ
DD F
exp
0
(2)
where D0 is the frequency factor, Q is the activation energy for diffusion, P is the applied pressure,
VF is the activation volume, R is the gas constant and T is the absolute temperature. The values for
D0 and Q were taken from Ref. [40], and P and T were used as 6 GPa and 340 K, respectively. To
examine the effects of pressure and temperature rise during the HPT operation, calculation was also
attempted without application of pressure and at temperatures of 300 and 340 K. For all selected
compositions, VF was assumed to be 6.1x10-6 m3/mol, which is an average of VF for Al (6.5x10-6
m3/mol) and Cu (5.7x10-6 m3/mol) [41]. Referring to Table 1, comparison shows that the estimated
diffusion coefficients using Eq. (1) are 1012-1022 times higher than those calculated using Eq. (2) for
the lattice diffusion in the Al-Cu system under a pressure of 6 GPa and at a temperature of 340 K. It

10
is noted that the temperature rise from 300 K to 340 K during HPT operation increases the diffusion
coefficient by ~102 -104 times but this is insufficient to account for the high diffusion coefficients
obtained using Eq. (1). The application of the pressure, 6 GPa, for the HPT operation rather reduces
the diffusion coefficient by ~5x106 times.
Table 1. Chemical analyses for Al and Cu obtained by EDS analysis from selected regions in Fig. 9,
diffusion path (x) or average width of selected regions, and diffusion coefficients (D) for selected
regions calculated using Eqs. 1 and 2.
Regions in Fig. 8
1
2
3
4
5
6
7
Al (at.%)
6
30
64
98
63
47
60
Cu (at.%)
94
70
36
2
37
53
40
x (nm)
150
600
300
100
450
480
430
D1 (m2/s)
4X10-18
6X10-17
1X10-17
2X10-18
3X10-17
4X10-17
3X10-17
D2 (m2/s)
1X10-39
3X10-31
3X10-30
2X10-30
3X10-30
7X10-33
3X10-30
D1 obtained using Eq. (1).
D2 calculated using Eq. (2) with P = 6 GPa, T = 340 K.
Thus, the third factor, which is related to the population of lattice defects, must be very
important to explain the difference. Earlier papers reported that the diffusivity can strongly be
enhanced by SPD processing because of the presence of large fractions of high-angle grain
boundaries formed during SPD [42-45]. It was also reported that the phase transformations are
accelerated because of enhanced diffusivity in the SPD-processed materials [23,24,28,24]. Eq. (2)
may be described in more detail with the form as [39,40]
)exp()exp( RT
H
R
S
ACDMM
V

(3)
where A is a constant depending on the attempted frequency, lattice parameter and crystallographic
structure, SM and HM are the vacancy migration entropy and enthalpy, respectively, and CV is the
vacancy concentration given by the following equation [39,40]
)exp()exp()exp()exp( RT
PVE
R
S
RT
H
R
S
CFFFFF
V



(4)
where SF, HF and EF are the vacancy formation entropy, enthalpy and energy, respectively. Thus, D0
and Q in Eq. (2) are represented as
)exp( RSS
AD M
F
0

(5)
FM EHQ 
(6)
It is reasonable that EF and HM should be affected significantly by the presence of lattice defects

11
whereas the effect may be minimal on SF and SM. Because many vacancies are expected to be
present in the sample after HPT processing, EF should be lower than that in annealed state [46]. It is
also considered that HM should be small [47-49] because a high density of lattice defects are
available during the HPT processing [50]. It is well known that both dislocations and grain
boundaries play roles as rapid diffusion paths. A precise estimation is difficult for EF and HM but,
given QL the activation energy for lattice diffusion, Q may be selected as 1/2-2/3 of QL for
dislocation-core diffusion (pipe diffusion) and grain boundary diffusion, and as 1/4-1/3 of QL for
surface diffusion. It should be noted that the strain introduced during the HPT processing may
influence the diffusivity via the changes of solute-vacancy binding energy [51], although this effect
was not considered in this study.
Figure 9 plots the diffusion coefficients obtained using Eq. (1) for the corresponding Cu
concentrations in regions 1-7 and compares with those calculated using different magnitudes of Q
through Eq. (2). It is shown that they are well comparable with those for surface diffusion. It is then
concluded that the solid-state reaction and the formation of intermetallic phases are controlled by
rapid diffusion due to the presence of many vacancies, dislocations and grain boundaries.
Figure 9. Estimated diffusion coefficients during HPT processing plotted against Cu content in
comparison with the reference data calculated using three different activation energies such as
lattice diffusion (Q = QL), grain boundary diffusion (Q = (1/2-2/3)QL) and surface diffusion (Q =
(1/4-1/3)QL). Diffusion coefficients obtained from Eq.(1) are labeled 1-7 corresponding to regions
1-7 in Fig.8.

12
5. Summary and conclusions
HPT was applied to a bulk form of disc consisting of two separate half-discs of pure Al and
pure Cu. The following conclusions were obtained.
1. Alternating Al-Cu layered structures with well-defined Al/Cu interfaces are formed with a
stacking sequence along the disc normal at an early stage of straining. With increasing the
imposed strain (the number of turns), distorted lamellar Al-Cu structures are extended over the
disc.
2. Processing by HPT promotes solid-state reaction of Al and Cu so that the formation of Al2Cu,
AlCu and Al4Cu9 intermetallic phases occurs as well as the dissolution of Al and Cu in each
matrix.
3. The diffusion coefficients estimated from the formation of the ultrafine structures during HPT
processing appear to be 1012-1022 times higher than lattice diffusion and be comparable to
surface diffusion.
4. The solid-state reaction and the formation of intermetallic phases are controlled by intense rapid
diffusion due to the presence of many vacancies, dislocations and grain boundaries.
Acknowledgments
This work was supported in part by the Light Metals Educational Foundation of Japan, in part
by a Grant-in-Aid for Scientific Research from the MEXT, Japan, in Innovative Areas "Bulk
Nanostructured Metals" and in part by Kyushu University Interdisciplinary Programs in Education
and Projects in Research Development (P&P).
References
[1] Westbrook JH, Fleischer RL. Intermetallic Compounds, Principles and Practice, Vol. 1,
Cichester: John Wiley & Sons; 1995.
[2] Valiev RZ, Islamgaliev RK, Alexandrov IV. Prog Mater Sci 45;2000:103-189.
[3] Yoon EY, Lee DJ, Ahn DH, Lee ES, Kim HS. J Mater Sci 2012;47:7770-7776.
[4] Alexandrov IV, Zhu YT, Lowe TC, Islamgaliev RK, Valiev RZ. Metall Mater Trans A
1998;29:2253-2260.
[5] Zhilyaev AP, Gimazov AA, Raab GI, Langdon TG. Mater Sci Eng A 2008;486:123-128.
[6] Valiev RZ, Estrin Y, Horita Z, Langdon TG, Zehetbauer MJ, Zhu YT. JOM 2006;58(4):33-39.
[7] Murayam M, Horita Z, Hono K. Acta Mater 2001;49:21-29.
[8] Sha G, Wang YB, Lia XZ, Duan ZC, Ringer SP, Langdon TG. Acta Mater 2009;57:3123-3132.
[9] Bera S, Zuberova Z, Hellmig RJ, Estrin Y, Manna I. Phil Mag 2010;90:1465-1483.
[10] Ohsaki S, Kato S, Tsuji N, Ohkubo T, Hono K. Acta Mater 2007;55:2885-2895.
[11] Kaneko K, Hata T, Tokunaga T, Horita Z. Mater Trans 2009;50:76-81.
[12] Rajulapati KV, Scattergood RO, Murty KL, Horita Z, Langdon TG, Koch CC. Metall Mater
Trans A 2008;39:2528-2534.
[13] Edalati K, Toh S, Iwaoka H, Horita Z. Acta Mater 2012;60:3885-3893.
[14] Tian YZ, Zhang JJ, Wu SD, Zhanng ZF, Kawasaki M, Langdon TG. Acta Mater
2012;60:269-281.
[15] Sauvage X, Wetscher F, Pareige P. Acta Mater 2005;53:2127-2135.
[16] Sauvage X, Jessner P, Vurpillot F, Pippan R, Scripta Mater 2008;58:1125-1128.
[17] Nishihata S, Suehiro K, Arita M, Masuda M, Horita Z. Adv Eng Mater 2010;12:793-797.

13
[18] Sabirov I, Kolednik O, Pippan R. Metall Mater Trans A 2005;36:2861-2870.
[19] Oh-ishi K, Horita Z, Smith DJ, Valiev RZ, Nemoto M, Langdon TG. J Mater Res
1999;14:4200-4207.
[20] Zhilyaev AP, Langdon TG. Prog Mater Sci 2008;53:893-979.
[21] Wei Q, Zhang HT, Schuster BE, Ramesh KT, Valiev RZ, Kecskes LJ, Dowding RJ, Magness
L, Cho K. Acta Mater 2006;54:4079-4089.
[22] Ciuca O, Tsuchiya K, Yokohama Y, Todaka Y, Ummemoto M. Mater Trans 2010;51:14-22.
[23] Perez-Prado MT, Gimazov AA, Ruano OA, Kassner ME, Zhilyaev AP. Scripta Mater
2008;58:219-222.
[24] Straumal BB, Mazilkin AA, Baretzky B, Schutz G, Rabkin E, Valiev RZ. Mater Trans
2012;53:63-71.
[25] Sun YF, Fuji H, Nakamura T, Tsuji N, Todaka D, Umemoto M. Scripta Mater
2011;65:489-492.
[26] Revesz A, Hober S, Labar JL, Zhilyaev AP, Kovacs Z. J Appl Phys 2006;100:103522.
[27] Huang JY, Zhu YT, Liao XZ, Valiev RZ. Phil Mag Lett 2004;84:183-190.
[28] Edalati K, Toh S, Watanabe M, Horita Z. Scripta Mater 2012;66:386-389.
[29] Edalati K, Toh S, Iwaoka H, Watanabe M, Horita Z, Kashioka D, Kishida K, Inui H. Scripta
Mater 2012;67:814-817.
[30] Long BD, Umemoto M, Todaka Y, Othman R, Zuhailawati H. Mater Sci Eng A
2011;528:150-1756.
[31] Edalati K, Horita Z, Mine Y. Mater Sci Eng A 2010;527:2136-2141.
[32] Edalati K, Miresmaeili R, Horita Z, Kanayama H, Pippan R. Mater Sci Eng A
2011;528:7301-7305.
[33] Figueiredo RB, Pereira PHR, Aguilar MTP, Celtin PR, Langdon TG. Acta Mater
2012;60:3190-3198.
[34] Lee DJ, Yoon EY, Park LJ, Kim HS. Scripta Mater 2012;67:384-387.
[35] Iwahashi Y, Horita Z, Nemoto M, Langdon TG. Acta Mater 1998;46:3317-3331.
[36] Xu C, Horita Z, Langdon TG. Acta Mater 2007;55:203-212.
[37] Komura S, Horita Z, Nemoto M, Langdon TG. J Mater Res 1999;14:4044-4050.
[38] Edalati K, Fujioka T, Horita Z. Mater Sci Eng A 2008;497:168-173.
[39] Shewmon P. Diffusion in Solids, second ed. Pennsylvania: The Minerals, Metals & Materials
Society; 1989.
[40] Mehrer H. Numerical Data and Functional Relationships in Science and Technology,
Diffusion in Solid Metals and Alloys, Vol. 26, Berlin: Springer-Verleg; 1990.
[41] Kraftmakher Y. Phys Rep 1998;299:79-188.
[42] Fujita T, Horita Z, Langdon TG. Phil Mag A 2002;82:2249-2262.
[43] Fujita T, Horita Z, Langdon TG. Mater Sci Eng A 2004;241:241-250.
[44] Divinski SV, Ribbe J, Baither D, Schmitz G, Reglitz G, Rosner H, Sato K, Estrin Y, Wilde G.
Acta Mater 2009;57:5706-5717.
[45] Divinski SV, Reglitz G, Rosner H, Estrin Y, Wilde G. Acta Mater 2011;59:1974-1985.
[46] Gavini V. Proc Roy Soc A 2009;465:3239-3266.
[47] Koehler JS. Phys Rev 1969;181:1015-1019.
[49] Sato K, Yoshiie T, Satoh Y, Xu Q. Mater Trans 2004;45:833-838.
[50] Setman D, Schafler E, Korznikova E, Zehetbauer MJ. Mater Sci Eng A 2008;493:116-122.
[51] Wurschum R, Oberdorfer B, Steyskal EM, Sprengel W, Puff W, Pikart P, Hugenschmidt C,
Pippan R. Physica B 2012;407:2670-2675.
[52] Wolverton C. Acta Mater 2007;55:5867-5872.