J Mater Sci (2009) 44:4758–4762

DOI 10.1007/s10853-009-3725-z

LETTER

Principles of grain refinement in magnesium alloys processed

by equal-channel angular pressing
Roberto B. Figueiredo Æ Terence G. Langdon

Received: 6 June 2009 / Accepted: 2 July 2009 / Published online: 17 July 2009
Ó Springer Science+Business Media, LLC 2009

Processing through the application of severe plastic a two-step procedure in which the grain size is initially
deformation (SPD) provides a useful tool for introducing reduced by extrusion prior to ECAP [10] and this process
very significant grain refinement in bulk solids [1]. of extrusion and ECAP, termed EX-ECAP, was used suc-
Although several SPD techniques are now available, pro- cessfully for the production of ultrafine-grained structures
cessing by equal-channel angular pressing (ECAP) is in several Mg alloys [11–17]. Other processing procedures
especially attractive because it is easily scaled-up for use were also proposed for reducing the pressing temperature
with large samples [2]. Numerous experiments have shown in magnesium alloys including by using a back-pressure
that processing by ECAP produces ultrafine-grain struc- [18, 19], increasing the channel angle within the ECAP die
tures in a range of fcc metals, including Al and Cu, and this [20, 21], reducing the pressing speed [22] and using
led to the development of a model for grain refinement sequential reductions in the pressing temperature with
based on the formation of an elongated array of subgrains subsequent passes through the die [23, 24].
or cells in the first pass of ECAP, the further development The evidence available to date suggests there are very
of elongated arrays in subsequent passes and the evolution significant differences between the grain refinement pro-
of this structure into an array of equiaxed grains separated cess in magnesium alloys and in fcc metals such as alu-
by high-angle grain boundaries [3]. In practice, there is minum. Whereas the process of grain refinement is
excellent agreement between this model of grain refine- insensitive to the pressing speed in pure aluminum and an
ment for fcc metals [3], the shearing patterns predicted in Al–1% Mg solid solution alloy [25], experiments on the
ECAP processing [4] and experimental observations of the AZ31 magnesium alloy show the grain size is reduced by
microstructural evolution in Al single crystals [5, 6] and pressing at a faster speed [24]. Furthermore, there is evi-
polycrystalline Al [7, 8] processed by ECAP. dence in the AZ31 and ZK60 alloys for the formation by
Despite the success with fcc metals, early experiments ECAP of either a homogeneous distribution of equiaxed
on pure magnesium and a magnesium-based alloy revealed grains [13, 26] or a bimodal distribution where coarse
significant difficulties in achieving grain refinement by grains are surrounded by arrays of much smaller grains [19,
ECAP due to a necessity to press at high temperatures in 26]. The present investigation was initiated to investigate
order to avoid premature cracking [9]. Subsequently, it was these differences and specifically to develop a model for
shown that ultrafine grains may be achieved by introducing grain refinement when magnesium alloys are processed by
ECAP.
Experiments were conducted using two different com-
R. B. Figueiredo
T. G. Langdon (&)
mercial magnesium alloys, AZ31 (Mg–3% Al–1% Zr) and
Departments of Aerospace & Mechanical Engineering
and Materials Science, University of Southern California, ZK60 (Mg–5.5% Zn–0.5% Zr), where these two alloys
Los Angeles, CA 90089-1453, USA were selected because extensive information is available on
e-mail: langdon@[Link] the microstructures produced in these alloys both with and
without ECAP processing. The alloys were received in an
T. G. Langdon
Materials Research Group, School of Engineering Sciences, extruded condition with initial grain sizes of *9.4 and
University of Southampton, Southampton SO17 1BJ, UK *2.9 lm for AZ31 and ZK60, respectively. The alloys

123
J Mater Sci (2009) 44:4758–4762 4759

were processed by ECAP at a temperature of 473 K using a In Fig. 1a there is an initial relatively coarse structure
die with an angle of 110° between the two parts of the where d [ dc and processing by ECAP leads to the for-
channel and an outer arc of curvature of *20°: these mation of new grains along the initial grain boundaries in a
angles produce a strain of *0.8 on every pass through the necklace-like arrangement. However, the initial grain size
die [27]. The pressing was conducted at a speed of is now sufficiently large that, as illustrated in Fig. 1b, a
*2 mm s-1 with the billets rotated by 90° in the same bimodal structure is produced wherein the new grains
sense between each pass using processing route Bc [28]. occupy only a fraction of the entire volume of the material
Following pressing, small samples were cut from the bil- and there are areas in the centers of the larger grains which
lets, mounted, ground on abrasive papers, polished with are not consumed by the formation of these smaller grains.
0.05 lm alumina powder and then etched in a solution of Conversely, Fig. 1c shows an initial fine structure where
picric acid, acetic acid, distilled water and ethanol to reveal d \ dc and in this situation processing by ECAP leads
the grain boundaries. Optical microscopy was used to again to the formation of new grains along the original
examine the microstructures and the average grain sizes grain boundaries but this now produces a homogeneous
were measured using the mean linear intercept method. microstructure of fine grains which, as shown in Fig. 1d,
An examination of many samples pressed under differ- occupies the entire volume of the sample.
ent conditions led to the conclusion that a critical grain Figure 2 shows an example for the extruded AZ31 alloy
size, dc, is needed in order to achieve a homogeneous where (a) is the initial grain structure and (b) gives the
distribution of equiaxed grains following processing by microstructure after one pass. The initial structure has an
ECAP. Accordingly, the principles of grain refinement in average grain size of *9.4 lm but it is evident that some
magnesium alloys may be represented by the schematic relatively coarser grains are also present. After one pass the
illustrations in Fig. 1 where the left column denotes the grain size distribution is bimodal such that there are large
initial condition for grain sizes larger and smaller than areas of new fine grains but there are also discrete and
dc and the right column shows the resultant microstructures well-defined large grains that appear to correspond to the
after processing through one pass by ECAP. inner core areas of some of the larger initial grains: the

Fig. 1 A model for grain

refinement in Mg alloys
processed by ECAP from a
coarse initial structure with an
average grain size larger than
the critical value dc (upper row)
and a fine initial structure with
an average grain size smaller
than the critical value dc (lower
row): the left column shows the
initial microstructures prior to
ECAP and the right column
shows the microstructures after
one pass

123
4760 J Mater Sci (2009) 44:4758–4762

Fig. 3 Representative grain structures in the ZK60 alloy in a the

extruded condition prior to ECAP and b after one pass of ECAP

reported in magnesium alloys processed by ECAP are

transitional in nature and they may evolve into regular
arrays of fine grains if the processing is continued through
a sufficiently large number of passes.
A second example is given in Fig. 3 for the ZK60 alloy
where (a) shows the initial extruded condition where the
average grain size is *2.9 lm and (b) shows the formation
after a single pass of a homogeneous distribution of
ultrafine grains with an average size of *0.8 lm: these
microstructures correspond to Fig. 1c, d in the schematic
illustration. Thus, for the extruded ZK60 alloy the initial
Fig. 2 Representative grain structures in the AZ31 alloy in a the
extruded condition prior to ECAP, b after one pass of ECAP and
grain size produced by extrusion is sufficiently small that it
c after six passes of ECAP is possible to form a homogeneous distribution of grains in
only one pass.
Grain refinement in magnesium alloys is therefore
microstructures in Fig. 2a, b correspond to Fig. 1a, b, characterized by the nucleation of fine grains along the pre-
respectively, in the schematic illustration. An important existing grain boundaries, where this nucleation is attrib-
additional observation is given in Fig. 2c which shows the uted to the development of stress concentrations at the
microstructure of the alloy after six passes where the boundaries and the subsequent activation of both basal and
bimodal condition has now evolved to ultimately form a non-basal slip processes [29]. This is consistent with the
homogeneous distribution of fine grains with a measured conclusion that the production of homogeneous three-
average size of *1.6 lm. These photomicrographs dem- dimensional microstructures in magnesium alloys requires
onstrate that the bimodal grain distributions frequently the activation of both non-basal and basal slip and with

123
J Mater Sci (2009) 44:4758–4762 4761

experimental observations showing that non-basal slip is dc, is not defined uniquely for any selected alloy but rather
activated more easily in the interiors of the grains of it is dependent upon the pressing conditions such that dc
magnesium alloys when the grain size is reduced [30]. becomes larger, and homogeneous structures are attained
The model for grain refinement depicted schematically more easily, when the pressing temperature is increased to
in Fig. 1 is consistent with a number of experimental aid in the development of the non-basal slip processes.
observations recorded from the processing of magnesium Fourth, although earlier experiments on pure Al and an
alloys by ECAP, hot compression and torsion. Al–1% Mg alloy demonstrated there was no influence of
First, in order to successfully process magnesium alloys the pressing speed on the equilibrium grain size produced
by ECAP under conditions leading to homogeneous ultra- by ECAP [25], recent experiments on the AZ31 alloy
fine-grain structures, it has been established that it is nec- showed that smaller grains are produced when the speed of
essary to introduce a step of intermediate grain refinement processing by ECAP is increased [24]. Again, this is con-
[10] where this is needed to produce conditions that avoid sistent with the model because higher processing speeds
billet failure and lead to the accommodation of deformation impose larger stresses and thereby facilitate the occurrence
throughout the volume of the material. The model depicted of non-basal slip and the easier development of a more
in Fig. 1 explains the excellent homogeneous grain refine- refined microstructure.
ment achieved with the EX-ECAP process when extrusion is Finally, it should be noted that a very recent report
used to produce an average grain size that is smaller than the described a model for grain refinement in Mg alloys pro-
critical size required for homogeneous refinement by ECAP cessed by ECAP where shear bands and dislocation
at the processing strain rate and temperature. The model also entanglements form within the grains and lead directly to a
explains the good results observed when pressing Mg alloys refined grain structure [36]. However, this model fails to
by ECAP with an initial high temperature and subsequent account for the formation of bimodal grain structures, it
lower temperatures because the initial processing at high does not satisfactorily address the experimental observa-
temperatures produces an intermediate grain refinement tions of grain refinement as recorded in Fig. 2 and it is not
which is then suitable for processing at lower temperatures. consistent with the observations of a necklace-like forma-
Second, the model is consistent with the reports of new tion of new grains along the original grain boundaries.
grains forming along the original grain boundaries in the In summary, there is a significant difference in the nature
hot compression testing of the AZ31 [31] and Mg–0.8% Al of grain refinement by ECAP in fcc metals and magnesium
[32] alloys and with reports of a necklace-like appearance alloys. A model is presented to describe the mechanism of
of new grains in several magnesium-based alloys tested in grain refinement in magnesium alloys based on the forma-
hot compression [33, 34] and hot torsion [35]. For the tion of new grains along the original grain boundaries in a
experiments in torsion [35], it was concluded specifically necklace-like manner. There is a critical grain size, dc, which
that, as in the present model, a homogeneous grain struc- delineates whether the pressed microstructure is homoge-
ture is attained only if the initial structure is sufficiently neous or bimodal. In practice, bimodal grain distributions
fine that the newly formed grains are able to impinge on the are transitional in nature and they may evolve into homo-
new grains developing on the opposite side of the grain. geneous grain distributions with increasing numbers of
Third, there are reports of the development of either passes. The proposed mechanism is consistent with experi-
bimodal or homogeneous grain structures when using mental results on the pressing of two magnesium alloys and
ECAP with magnesium alloys but these differences arise with extensive data in the literature on the processing of
because of variations in the initial grain sizes and/or the magnesium alloys in hot compression and hot torsion.
pressing conditions. For example, an AZ31 alloy with an
extruded grain size of 2.5 lm was successfully processed Acknowledgements One of the authors (RBF) was supported by a
CAPES/Fulbright Scholarship. This work was supported by the U.S.
by ECAP at 473 K using a die with a channel angle of 110° Army Research Office under Grant No. W911NF-08-1-0201.
and no back-pressure and this produced a homogeneous
grain size of *0.7 lm [13]. By contrast, a cast AZ31 alloy
with an initial grain size of 640 lm was pressed at 423 K
References
using a die with a channel angle of 90° and with an
imposed back-pressure to produce a bimodal array of 1. Valiev RZ, Islamgaliev RK, Alexandrov IV (2000) Prog Mater
coarse (16.5 lm) and fine (0.5 lm) grains. Thus, these two Sci 45:103
conditions correspond directly to the models shown in 2. Valiev RZ, Langdon TG (2006) Prog Mater Sci 51:881
Fig. 1. Furthermore, the cast alloy yielded a homogeneous 3. Langdon TG (2007) Mater Sci Eng A 462:3
4. Furukawa M, Horita Z, Langdon TG (2002) Mater Sci Eng A
microstructure of fine grains, with an average size of 332:97
*2.5 lm, when the temperature of ECAP was increased to 5. Fukuda Y, Oh-ishi K, Furukawa M, Horita Z, Langdon TG
473 K. This result demonstrates that the critical grain size, (2004) Acta Mater 52:1387

123
4762 J Mater Sci (2009) 44:4758–4762

6. Fukuda Y, Oh-ishi K, Furukawa M, Horita Z, Langdon TG 20. Furui M, Kitamura H, Anada H, Langdon TG (2007) Acta Mater
(2006) Mater Sci Eng A 420:79 55:1083
7. Iwahashi Y, Horita Z, Nemoto M, Langdon TG (1997) Acta 21. Figueiredo RB, Cetlin PR, Langdon TG (2007) Acta Mater
Mater 45:4733 55:4769
8. Iwahashi Y, Horita Z, Nemoto M, Langdon TG (1998) Acta 22. Kang F, Wang JT, Peng Y (2008) Mater Sci Eng A 487:68
Mater 46:3317 23. Mussi A, Blandin JJ, Salvo L, Rauch EF (2006) Acta Mater
9. Yamashita A, Horita Z, Langdon TG (2001) Mater Sci Eng A 54:3801
300:142 24. Ding SX, Chang CP, Kao PW (2009) Metall Mater Trans A
10. Horita Z, Matusbara K, Makii K, Langdon TG (2002) Scr Mater 40:415
47:255 25. Berbon PB, Furukawa M, Horita Z, Nemoto M, Langdon TG
11. Matsubara K, Miyahara Y, Horita Z, Langdon TG (2003) Acta (1999) Metall Mater Trans A 30:1989
Mater 51:3073–3084 26. Figueiredo RB, Langdon TG (2009) Mater Sci Eng A 501:105
12. Matsubara K, Miyahara Y, Horita Z, Langdon TG (2004) Metall 27. Iwahashi Y, Wang J, Horita Z, Nemoto M, Langdon TG (1996)
Mater Trans A 35:1735 Scr Mater 35:143
13. Lin HK, Huang JC, Langdon TG (2005) Mater Sci Eng A 28. Furukawa M, Iwahashi Y, Horita Z, Nemoto M, Langdon TG
402:250 (1998) Mater Sci Eng A 257:328
14. Furui M, Xu C, Aida T, Inoue M, Anada H, Langdon TG (2005) 29. Galiyev A, Kaibyshev R, Gottstein G (2001) Acta Mater 49:1199
Mater Sci Eng A 410–411:439 30. Koike J (2003) Mater Sci Forum 419–422:189
15. Miyahara Y, Matsubara K, Horita Z, Langdon TG (2005) Metall 31. Beer AG, Barnett MR (2006) Mater Sci Eng A 423:292
Mater Trans A 36:1705 32. Ion SE, Humphreys FJ, White SH (1982) Acta Metall 30:1909
16. Miyahara Y, Horita Z, Langdon TG (2006) Mater Sci Eng A 33. Barnett MR (2003) Mater Trans 44:571
420:240 34. Hakamada M, Watazu A, Saito N, Iwasaki H (2009) Mater Trans
17. Figueiredo RB, Langdon TG (2006) Mater Sci Eng A 430:151 50:711
18. Xia K, Wang JT, Wu X, Chen G, Gurvan M (2005) Mater Sci 35. Spigarelli S, El Mehtedi M, Cabibbo M, Evangelista E, Kaneko J,
Eng A 410–411:324 Jäger A, Gartnerova V (2007) Mater Sci Eng A 462:197
19. Lapovok R, Estrin Y, Popov MV, Langdon TG (2008) Adv Eng 36. Su CW, Lu L, Lai MO (2006) Mater Sci Eng A 434:227
Mater 10:429

123