You are on page 1of 15

Materials Science and Engineering A283 (2000) 274 – 288

www.elsevier.com/locate/msea

An experimental study of the recrystallization mechanism during


hot deformation of aluminium
S. Gourdet, F. Montheillet *
Ecole Nationale Supérieure des Mines de Saint-Etienne, Centre Science des Matériaux et des Structures, URA CNRS 1884, 158 Cours Fauriel,
42023 Saint-Etienne Cedex 2, France
Received 10 February 1999; received in revised form 22 October 1999

Abstract

Discontinuous dynamic recrystallization (involving nucleation and grain growth) is rarely observed in metals with high stacking
fault energies, such as aluminium. In this metal, two other types of recrystallization have been observed: continuous dynamic
recrystallization (CDRX, i.e. the transformation of subgrains into grains); and geometric dynamic recrystallization (due to the
evolution of the initial grains). The main purpose of this work was to bring clearly into evidence and to better characterize CDRX.
Uniaxial compression tests were carried out at 0.7 Tm and 10 − 2 s − 1 on three types of polycrystalline aluminium: a pure
aluminium (1199), a commercial purity aluminium (1200) and an Al-2.5wt.%Mg alloy (5052), and also on single crystals of pure
aluminium. In addition, 1200 aluminium specimens were strained in torsion. The deformed microstructures were investigated at
various strains using X-ray diffraction, optical microscopy, scanning electron microscopy and electron back-scattered diffraction.
Observations of the single crystalline samples confirm that subgrain boundaries can effectively transform into grain boundaries,
especially when the initial orientation is unstable. In the case of polycrystalline specimens, after separating the effects of the initial
and new grain boundaries, it turns out that CDRX operates faster in the 1200 aluminium compared to the two other grades.
Moreover, it appears that the strain path does not alter noticeably the CDRX kinetics. © 2000 Elsevier Science S.A. All rights
reserved.

Keywords: Hot deformation; Aluminium; Dynamic recrystallization; Single crystals; Subgrain boundaries; Grain boundaries; Misorientations

1. Introduction Discontinuous dynamic recrystallization, which is


commonly observed in low stacking fault energy
Aluminium and its alloys exhibit very high rates of metals, remains exceptional in aluminium and alu-
dynamic recovery, which is generally expected to com- minium alloys. Nevertheless, it seems to occur in two
pletely inhibit dynamic recrystallization. However, the specific cases, viz., in high purity aluminium and in
formation of new grains during hot deformation of aluminium alloys containing large particles:
aluminium has been frequently reported. Three types of 1. Single crystals and polycrystals of 99.999 wt.% alu-
dynamic recrystallization are likely to produce such a minium have been subjected to hot deformation in
microstructure: (i) discontinuous dynamic recrystalliza-
compression by Yamagata [1–5]. Stress-strain
tion (DDRX), i.e. the classical recrystallization, which
curves exhibit strong oscillations, typical of DDRX,
operates by nucleation and grain growth; (ii) continu-
although more irregular. The associated microstruc-
ous dynamic recrystallization (CDRX), which involves
tures generally display new grains without substruc-
the transformation of low angle boundaries into high
angle boundaries; and (iii) geometric dynamic recrystal- ture. It is therefore not excluded that these grains
lization (GDRX), generated by the fragmentation of have grown ‘after’ deformation, all the more as
the initial grains. aluminium of such high purity recrystallizes stati-
cally very rapidly, even at room temperature. How-
* Corresponding author. Tel.: + 33-4-77420026; fax: +33-4-
ever, one micrograph [3] clearly displays the
77420157. presence of several grains containing a substructure,
E-mail address: montheil@sms.emse.fr (F. Montheillet) in an initially monocrystalline sample. Moreover,

0921-5093/00/$ - see front matter © 2000 Elsevier Science S.A. All rights reserved.
PII: S 0 9 2 1 - 5 0 9 3 ( 0 0 ) 0 0 7 3 3 - 4
S. Gourdet, F. Montheillet / Materials Science and Engineering A283 (2000) 274–288 275

the occurrence of DDRX in high purity aluminium high angle boundaries effectively takes place when the
has been recently confirmed by Ponge et al. [6]. boundaries are pinned by small particles. This mecha-
High purity produces two opposite effects. On the nism has been used to promote superplasticity in Zr
one hand, it favors DDRX by increasing grain bearing or high Mg aluminium alloys (Al-6wt.%Cu-
boundary mobility. On the other hand, it can inhibit 0.4wt.%Zr [21,22], Al-0.25wt.% Zr-0.1wt.% Si [23,24],
DDRX since the high level of recovery prevents Al-10wt.% Mg-0.1wt.% Zr [25,26], Al-10wt.% Mg-
dislocation accumulation, thus reducing the driving 0.5wt.% Mn [27,28]). These alloys are generally cold or
force. Experimental results indicate that the first warm rolled, to increase their dislocation density. Un-
effect prevails over the second [7]. der these conditions, subgrains form very quickly dur-
2. There is some evidence that DDRX can occur dur- ing the subsequent hot tension testing. Since the
ing hot deformation of Al – Mg – Mn alloys, because boundaries are pinned by small particles (Al3Zr,
a high Mg solute addition raises the dislocation (Al8Mg5)b or Al6Mn, according to the alloy composi-
density and thus the driving force for DDRX, while tion) and continuously absorb dislocations, the sub-
large Al6Mn particles (\ 1 mm) stimulate nucle- grains transform into grains without growing. A fine
ation. The presence of small new grains adjacent to and equiaxed grain structure is thus obtained in the
large particles has been reported for instance after early stages of superplastic deformation.
plane strain compression of Al-1wt.% Mg-1wt.% Finally, geometric dynamic recrystallization has first
Mn [8] and extrusion of Al-5wt.% Mg-0.8wt.%Mn been described by McQueen et al. [29] in a commercial
[9]. However, the volume fractions of recrystallized purity aluminium. During deformation, the original
grains remained small and no effect on the shapes of grains flatten (compression) or elongate (tension, tor-
the stress-strain curves was detected. This means sion), and their boundaries become progressively ser-
that DDRX is a possible but limited restoration rated while subgrains form. Consequently, the grain
mechanism in aluminium alloys. boundary area per unit volume grows strongly and an
Continuous dynamic recrystallization occurs in turn increasing fraction of subgrain facets is made of those
by the progressive accumulation of dislocations in low
‘initial’ grain boundaries. Ultimately, when the original
angle boundaries, leading to the increase of their mis-
grain thickness is reduced to about two subgrain sizes,
orientation and the formation of large angle grain
the grain boundaries begin locally to come into contact
boundaries when their misorientation angles reach a
with each other, causing the grains to pinch-off [29–
critical value uc (uc :15°). This mechanism has been
32]. In the case of Al-5Mg alloys, the tendency to the
observed in several high stacking fault energy metals,
serration of boundaries is stronger and a secondary
such as aluminium and aluminium alloys [10–13], b
process of GDRX by pinching off of the serrations has
titanium alloys [14 – 16], and ferritic steels [17 –20]. The
been reported [33,34].
microstructure of a commercial purity (1050 grade)
The first purpose of this work was to bring forward
aluminium strained in torsion has been investigated by
Perdrix et al. [10], and Montheillet [11]. These authors further evidence for CDRX. The main objection to the
found that, at small and medium strains (o : 1), the results of Perdrix et al. [10] was that the high fractions
microstructure consists of the deformed initial grains of large angle boundaries did not necessarily result
containing subgrains, which is typical of a recovered from CDRX, but could also be due to the evolution of
state. By contrast, strongly strained samples (o :40) the initial grain boundaries. In order to exclude the
exhibit a completely different microstructure: it is no latter possibility, single crystalline samples were used.
longer possible to distinguish the initial grains, and the The second objective was to better characterize CDRX.
former subgrains now appear as ‘crystallites’ bounded The influence of the following parameters was therefore
partly by low and partly by high angle boundaries. investigated:
Furthermore, the misorientation angles, which display a “ crystalline orientation, by using single crystals of
bimodal distribution at small strains (with subgrain various orientations;
boundaries less than 15° and initial grain boundaries “ purity, by testing three grades of polycrystalline alu-
between 30 and 63°) become uniformly distributed be- minium, ranging from 99.99 to 97 wt.% Al. Impuri-
tween 0 and 63° at large strains. Perdrix et al. [10] have ties and alloying elements reduce the recovery
explained these results by assuming a progressive trans- capacity of the material, since they decrease its stack-
formation of subgrain boundaries into grain ing fault energy (although this effect seems to be
boundaries. However, this mechanism remains contro- relatively weak in aluminium alloys) and solutes
versial and some authors have suggested that the in- as well as precipitates reduce the dislocation mobil-
creased fraction of high angle boundaries could result ity;
from GDRX. “ strain path, by comparing the microstructures
Nevertheless, there is a general agreement to consider obtained from uniaxial compression and torsion
that the transformation of low angle boundaries into tests.
276 S. Gourdet, F. Montheillet / Materials Science and Engineering A283 (2000) 274–288

Table 1
Chemical compositions of the three aluminium grades

Mg Si Cr Mn Fe Cu Zn

1199 (mg/g) 7 12 1 1 10 45 4
1200 (mg/g) 9 1500 62 82 5700 59 200
5052 (wt.%) 2.49 0.11 0.2 0.08 0.3 0.01 –

2. Experimental procedure tron back-scattered diffraction (EBSD) and X-ray


diffraction on the electropolished specimens. POM dis-
The three selected grades of aluminium, a pure 1199 plays both the initial grains and the new crystallites
aluminium, a commercial purity 1200 aluminium and (subgrains or new grains), whereas only the crystallites
an Al–Mg 5052 alloy (Table 1), were provided by the are generally revealed by SEM. This is due to the fact
Centre de Recherches de Voreppe (Pechiney) in the that POM colors are related to the crystalline orienta-
form of hot rolled plates. The initial structures were tions, and the original grains are associated with re-
fully recrystallized, exhibiting equiaxed grains of 220 gions of similar colors. This is not the case in SEM,
mm in the 1199 aluminium, flattened grains of average where crystallites of similar orientations can display
size 200 mm in the 1200 aluminium, and again equiaxed quite different grey levels, and conversely. However,
grains of 80 mm in the 5052 alloy (Table 2). The POM is known to overestimate subgrain sizes
textures displayed a strong cube component, especially [31,36,37], and is therefore inadequate for quantitative
in the 1200 grade. Cylindrical compression specimens analyses of the substructure. For this purpose, SEM
were machined with their axes parallel to the rolling micrographs were therefore used. Moreover, the color
direction. In addition, torsion specimens were prepared contrast in POM is not precise enough to distinguish a
from the 1200 aluminium with the torsion axis parallel new grain from a subgrain and even less to estimate the
to the rolling direction. Finally, single crystals were
misorientation between two crystallites. The orientation
grown from the 1199 aluminium and sectioned to ob-
of each crystallite was thus determined using EBSD and
tain cubic specimens with a Ž001, Ž011 or Ž111
the misorientation associated with each boundary sub-
direction parallel to the compression axis (Table 3).
sequently calculated. On account of detection limitation
Hot compression tests were carried out at a constant
in hot worked structures, boundaries with a misorienta-
strain rate o; = 10 − 2 s − 1 and 0.7 Tm (i.e. 380, 368 and
tion angle of less than 1° were not taken into consider-
333°C for the 1199, 1200, and 5052 grades, respec-
tively). The specimens were lubricated with graphite to ation. This omission is likely to be of no consequence,
minimize strain inhomogeneities, and water quenched however, since the study is focused on the transition
within 1 s after deformation. The torsion specimens from low angle to high angle boundaries around 15°. In
were strained under the same strain rate and tempera- addition to these local texture measurements, global
ture conditions, and water quenched within 5 s. The textures were investigated by X-ray diffraction.
compression specimens were then sectioned parallel to
Table 2
the axis, ground and electropolished (5 ml HClO4, 95 Mean intercept lengths of the grains in the hot rolled plates along the
ml C2H5OH, 20 V, 0°C, 60 s); some of them were rolling (RD), transverse (TD), and normal (ND) directions
subsequently anodized (10 ml HBF4, 90 ml H2O, 30 V,
20°C, 120 s). Since deformation was not uniform, local RD TD ND
strains were estimated using a mechanical model [35].
1199 (mm) 226 204 227
In what follows, the strain values correspond to the 1200 (mm) 285 215 98
center part of each specimen, where all observations 5052 (mm) 78 86 59
were carried out. However, even at large strains, the
strain gradient was quite low, e.g. in the compression
direction, Do/Dz : 0.2 mm − 1 at o =1.5 [35]. The tor- Table 3
Crystallographic directions parallel to the compression axis (CA) and
sion cylinders were ground to obtain a flat surface, and perpendicular to the lateral faces (TD1 and TD2) of the cubic single
then prepared the same way. The strain and strain rate crystals
undergone by these samples were estimated to approxi-
mately 80% of their nominal (surface) values. CA TD1 TD2
The deformed microstructures and textures were in-
[001] [100] [010]
vestigated using polarized optical microscopy (POM) [011] [100] [011( ]
on the anodized specimens, scanning electron mi- [111] [11( 0] [112( ]
croscopy (SEM) in the channeling contrast mode, elec-
S. Gourdet, F. Montheillet / Materials Science and Engineering A283 (2000) 274–288 277

3. Experimental results

3.1. Pure aluminium single crystals

During the compression tests, the cross-sections of the


Ž001 specimens remain square, as expected since the
Ž001 crystallographic axis and the geometric axis
(square cross-section) are both fourfold, and the trans-
verse directions TD1 and TD2 are crystallographically
equivalent (Table 3). The Ž011 (twofold crystallo-
graphic axis) specimens lengthen in the Ž100 direction.
It has been shown [38] that this shape change can be
explained by the activation of the four octahedral slip
systems (111)[101( ], (111)[11( 0], (1( 11)[1( 01( ], and (1( 11)[1( 1( 0].
Finally, the cross-sections of the Ž111 specimens be-
Fig. 1. Stress-strain curves of the Ž001, Ž011, Ž111 single crystals come irregular, which is due to the combination of their
and polycrystals of 1199 aluminium.
threefold crystallographic axis and four-fold geometric
axis.
Fig. 1 shows that the flow stresses reach a steady state
value of about 10.5 MPa in the case of the Ž001 and
Ž111 specimens (note that the polycrystal flow stress
tends to the same value); by contrast, it is much higher
for the Ž011 specimens (about 14 MPa). It will be
shown below that these values are closely related to the
subgrain sizes. The flow stress drop associated with the
Ž111 crystals can be mainly attributed to the evolution
of the Taylor factor: indeed, at o= 0, M= 3
3/2 :3.67,
while at o =1.5 (Ž101 orientation, see below) M=

6 : 2.45.
Global texture measurements show that the Ž011
orientation is perfectly stable (Fig. 2(b)). The Ž001
orientation is also fairly stable, since its decomposition
only starts at o= 1.5 (Fig. 2(a)). On the other hand, the
Ž111 orientation is very unstable (Fig. 2(c)). At o=0.3,
the compression axis is roughly parallel to Ž112 and
eventually reaches the Ž011 stable orientation at o=
1.5. These results are in good general agreement with the
literature, although the rotation amplitudes are larger
than those observed by Mécif et al. [39], after uniaxial
compression of aluminium single crystals at the same
temperature. This difference can be attributed, to a large
extent, to the higher strain (o= 1.5 vs 0.35) and strain
rate (10 − 2 vs. 2× 10 − 4 s − 1) applied in the present work.
The aspect of the sections perpendicular to the com-
pression axis does not change significantly with increas-
ing strain for the Ž100 specimens observed by POM
(Fig. 3(a, b)). However, the band structure observed on
the lateral sections at o= 0.3 transforms into a subgrain
structure at larger strains. It should be noted that SEM
reveals the presence of many subgrains within these
bands. The Ž011 specimens exhibit symmetrical mosaic
patterns on their (011( ) lateral sections (Fig. 3(c, d)), with
dislocation walls parallel to the planes of the
Fig. 2. Global textures of the monocrystalline specimens strained to
o= 1.5. (a) Ž001; (b) Ž011; (c) Ž111. The horizontal and vertical activated slip systems. SEM also reveals the formation
axes are associated with the TD1 and TD2 directions, respectively of small subgrains inside the cells at o= 1.5. Such a
(Table 3). mosaic microstructure of the Ž011 specimens (which
278 S. Gourdet, F. Montheillet / Materials Science and Engineering A283 (2000) 274–288

elongate only in the [100] direction) has already been The evolutions of the misorientation distributions
observed in the case of single crystals of various orien- with increasing strain are compared for the three orien-
tations deformed by channel die compression [40]. This tations in Fig. 5. In the case of the Ž001 crystal, the
suggests that such a particular structure is due to the average misorientation strongly increases with strain.
geometry of the slip systems, which allows a unidirec- At o = 0.9, a significant fraction of misorientations al-
tional strain to take place. By contrast to the previous ready exceeds 15°, which clearly means that part of the
orientations, the Ž111 specimens exhibit an inhomoge- low angle boundaries have transformed into large angle
neous microstructure. Horizontal and tilted bands are boundaries. At o= 1.5, this trend is more pronounced,
displayed on the TD1 lateral sections at o = 0.3. How- since almost 20% of the interfaces are now grain
ever, this inhomogeneity tends to vanish at larger boundaries. However, the microstructural steady state
strains. is not yet attained, which suggests that a more recrys-
Misorientation maps were plotted from EBSD mea- tallized microstructure (with crystallites bounded
surements for each single crystal strained to o =1.5. In mainly by large angle boundaries) could form at larger
the Ž001 specimen (Fig. 4(a)), a large number of strains. By contrast, for the Ž011 orientation, all the
boundaries exhibit a misorientation greater than 15°, measured misorientations are lower than 15°, even at
which indicates that subgrain boundaries have trans- o= 1.5. Furthermore, no evolution is noticeable be-
formed into grain boundaries. By contrast, the tween o= 0.9 and 1.5, and thus, the formation of grain
boundaries in the Ž011 specimen (Fig. 4(b)) exhibit boundaries is unlikely, even at larger strains. The be-
low misorientations, generally smaller than 6°, so that havior of the Ž111 specimens is more complex. In
no high angle boundaries have formed. The behavior of particular, the specimen strained to o= 0.9 has devel-
the Ž111 specimen (Fig. 4(c)) is intermediate, since oped a large amount of very high angle boundaries
only a small fraction of the interfaces consists of high (30–60°). This is due to the splitting of the initial
angle boundaries. orientation into two components (Fig. 6), which was

Fig. 3. POM microstructures of the monocrystalline specimens. (a) Ž001, o= 0.3; (b) Ž001, o =1.5; (c) Ž011, o= 0.3; (d) Ž011, o= 1.5.
S. Gourdet, F. Montheillet / Materials Science and Engineering A283 (2000) 274–288 279

Fig. 4. SEM micrographs and associated misorientation maps of the monocrystalline specimens strained to o = 1.5. (a) Ž001; (b) Ž011; (c) Ž111.
The vertical and horizontal axes are associated with the CA and TD1 (Ž001 or Ž011) or TD2 (Ž111) directions, respectively (Table 3).

observed in only one specimen of this orientation. It obtained from the compression and torsion tests was
should also be pointed out that, in the case of this investigated (see Fig. 1 for the 1199 grade). In all cases,
unstable orientation, large angle boundaries are formed the flow stress seems to reach a plateau at o:0.3
in the early stages of the deformation: at o = 0.3, they (although a small decrease of the torsion stress is
already represent more than 8% of the boundaries. expected at very large strains [10,11]). The flow stresses
reach steady state levels of 10.5, 21 and 77 MPa for the
3.2. Polycrystalline specimens 1199, 1200 and 5052 grades, respectively. Global tex-
ture measurements were carried out on the compression
The mechanical behavior of the various specimens specimens. The pole figures of the three grades (Fig. 7
280 S. Gourdet, F. Montheillet / Materials Science and Engineering A283 (2000) 274–288

Fig. 5. Strain dependence of the misorientation distributions for the three single crystals (N: number of measurements, u( : average misorientation
angle).

Fig. 6. Formation of deformation bands in the Ž111 specimen strained to o = 0.9.


S. Gourdet, F. Montheillet / Materials Science and Engineering A283 (2000) 274–288 281

All the microstructures are quite homogeneous. At


o=0.3, subgrains are well formed in the 1199 and 1200
grades (Fig. 8(a)), whereas they are not visible in the
5052 grade, because their formation is delayed by the
strong solute content of this alloy. Note also that at
o=1.5, the grain thickness is still much larger than the
subgrain size (Fig. 8(b)), which indicates that the initial
grains are far from decomposing into smaller grains by
a GDRX process. In Fig. 9, the mean intercept lengths
Fig. 7. Global texture of the 1199 polycrystalline specimen strained to D of the subgrains measured from the SEM micro-
o= 1.5. The horizontal and vertical axes are associated with the TD graphs are compared with a theoretical value calculated
and ND directions, respectively (Table 2). from the flow stress s, using the relationship s/G=ab/
D. The value of a= 28 was chosen here, since it is
commonly used for aluminium [34]. In the case of the
1199 and 1200 grades, the calculated subgrain sizes are
consistent with the measured ones, although slightly
larger. The coefficient 28 seems therefore overestimated,
and a better agreement between the two sets of data is
obtained by using a value of 24. However, according to
Blum et al. [34], this difference is due to the condensa-
tion of free dislocations into additional subgrain
boundaries between the end of deformation and
quenching, which causes a decrease of the average size
D. By contrast, the calculated subgrain size is much
smaller than the experimental value for the 5052 alloy.
This is due to the slow formation of the subgrains in
that case, since the presence of regions without well-
formed subgrains leads to an overestimation of the
mean intercept length.
Fig. 10 displays misorientation maps of the polycrys-
tals strained to o= 1.5. In the 1199 specimen (Fig.
10(a)), two initial grain boundaries can still be iden-
tified because they form a continuous chain of high
misorientation segments (\ 30°), whereas isolated high
angle boundaries with lower misorientations (15–30°)
are very probably new grain boundaries. In the case of
the 1200 aluminium strained in compression (Fig.
10(b)), the map shows the presence of a large fraction

Fig. 8. POM microstructures of the 1200 polycrystalline specimens.


(a) o= 0.3; (b) o =1.5. The compression axis is vertical.

illustrates the texture of the 1199 aluminium) show that


the texture consists mainly of a Ž011 fiber component,
which is especially strong for the 5052 alloy. Since a
Fig. 9. s/G versus b/D plot for the single crystals and polycrystals
strong cube component was observed in the initial strained to 1.5 (the shear moduli G = 20.6, 20.8, 21.2 GPa for the
state, this means that the crystallites have strongly 1199, 1200, 5052 grades, respectively, at 0.7 Tm, and the Burgers
rotated during compression. vector length b =2.86 ×10 − 10 m).
282 S. Gourdet, F. Montheillet / Materials Science and Engineering A283 (2000) 274–288

Fig. 10. SEM microstructures and associated misorientation maps of the polycrystalline specimens strained to o = 1.5. (a) 1199 specimen; (b) 1200
compression specimen; (c) 1200 torsion specimen. The compression axis is vertical.

of high angle boundaries, but the original grain butions of the 1199 aluminium are depicted in Fig. 11.
boundaries are no longer recognizable. It is likely that For this aluminium, as for the two other grades, there
subgrains have rotated more in the 1200 than in the is a progressive shift of the small misorientations to-
1199 grade, thus leading to a fragmentation of the wards the larger ones, which leads to an increase of the
initial grain boundaries. In the torsion specimen of Fig. average misorientation of about 8° between o =0.3 and
10(c), the fraction of high angle boundaries is very o = 1.5. The fraction of subgrain boundaries with small
similar, although some chains of segments exhibiting misorientations (B6°) is strongly reduced; the number
very large misorientations at the top of the map look of subgrain boundaries with larger misorientations in-
like initial grain boundaries. The misorientation distri- creases at first, and then decreases, whereas the fraction
S. Gourdet, F. Montheillet / Materials Science and Engineering A283 (2000) 274–288 283

of high angle boundaries continuously increases, reach- With regard to CDRX, the Ž001 single crystal (Fig.
ing more than 30% at o =1.5. This indicates that the 12(d)) and the polycrystal of same purity (and strong
very low angle boundaries ( B 6°) are continuously con- initial cube texture, Fig. 12(a)) display similar behav-
verted into medium angle boundaries (6 – 15°), which in iors: at o= 1.5, roughly 15–18% of their interfaces
turn transform into high angle boundaries. Such an consist of new high angle boundaries. When deforma-
evolution may be referred to as continuous tion bands form in the Ž111 crystal, the fraction of
recrystallization. new grain boundaries rises very quickly; if not, it
remains rather small, about 5–8%. Among the poly-
crystalline specimens, the highest fraction of new grain
3.3. Comparison of the geometric and continuous boundaries is observed in the 1200 specimens (35% in
dynamic recrystallization kinetics compression, 39% in torsion at o= 1.5), then in the
5052 alloy (about 24%), and last in the pure aluminium
In the polycrystalline specimens, the increase of the (only 15%). This can be explained by the very high
high angle boundary fraction is due not only to the recovery rate in pure aluminium, which lowers the
formation of new grain boundaries, but also to the accumulation rate of dislocations in the subgrain
expansion of the initial grain boundary area. In order boundaries. On the other hand, in the Al-Mg alloy, the
to get a quantitative estimation of the respective effects solute atmospheres impede dislocation movements,
of CDRX and GDRX, a geometrical model was used. thereby delaying the formation of subgrain boundaries
The evolutions of the lengths per unit area of initial [7]. It should be noted that the grain boundaries present
grain boundaries, new high angle boundaries, and low in the 1199 aluminium originate in equal parts from
angle boundaries were determined from a simple GDRX and CDRX, while those present in the 1200
derivation detailed in the Appendix A. and 5052 grades have mainly developed by CDRX. The
Fig. 12(a–e) illustrate the GDRX and CDRX kinet- present results also indicate that the compression and
ics. The strain dependence of the initial grain boundary torsion specimens display quantitatively the same be-
fraction is not monotonic since it grows due to the havior. Therefore, the strain path does not seem to
flattening of the initial grains, whereas the formation of modify the CDRX kinetics significantly.
subgrain boundaries makes it decrease. At o =1.5, the
initial grain boundaries represent about 8% of the
interfaces for the 1200 and 5052 grades. The 1199
aluminium differs by a much higher fraction, about 4. Discussion
17%, which is due to a smaller initial grain size/sub-
grain size ratio. By comparing the thickness h of the 4.1. New grain boundaries and deformation bands
initial grains at o = 1.5 (the latter were estimated by
calculation to 68, 80, and 33 mm for the 1199, 1200, and Experiments carried out on single crystals confirm
5052 grades, respectively) and the subgrain sizes D at that the subgrain boundary misorientations strongly
the same strain (14.3, 6.6 and 3.4 mm, respectively), it increase with strain: a maximum of at least 15° is
appears that the h/D ratios are about 5 for the 1199 reached for the three investigated orientations at o=
grade, and 10–12 for the 1200 and 5052 grades. This 0.9. Moreover, the transformation of low angle
confirms that GDRX is more developed in the pure boundaries into high angle boundaries is clearly demon-
aluminium. But this ratio still remains far from 2, strated for the two unstable orientations, as well as in
which means that the strain achieved by compression is the polycrystalline specimens where the fraction of high
not large enough to obtain a complete GDRX angle boundaries is much larger than expected for the
structure. deformed initial grain boundaries.

Fig. 11. Strain dependence of the misorientation distributions for the 1199 polycrystals (N: number of measurements, u( : average misorientation).
284 S. Gourdet, F. Montheillet / Materials Science and Engineering A283 (2000) 274–288

However, the misorientation increase of the subgrain


boundaries and their transformation into high angle
boundaries has been controversial for a long time,
principally on the basis of the work by Kassner and
McMahon [30]. These authors studied the microstruc-
tural evolution of high purity (99.999 wt.%) polycrys-
talline aluminium specimens strained in torsion (370°C,
5×10 − 4 s − 1). The observations were mainly carried
out by transmission electron microscopy, and the mis-
orientations were estimated from either dislocation
spacings (subgrain boundaries) or selected area diffrac-
tion patterns (grain boundaries). These authors noticed
only very limited increases in subgrain misorientations
with strain: the mean values varied from 0.5° at o=0.2
to a saturation value of 1.2° from o= 1.2 to 16. Most of
the discrepancies between this prior work and the
present investigation can be explained by differences in
the experimental procedures. First of all, the aluminium
used by the authors was purer, and the strain rate,
lower. This means that recovery was more efficient, and
thus the increase in misorientation was slowed down.
Furthermore, by contrast to the SEM-EBSD technique
used here, TEM allowed to account for boundaries
with very low misorientations (B 1°), therefore decreas-
ing the average misorientation value. However, some
boundaries with medium (about 10°), and large misori-
entations (\ 30°) were also observed by Kassner and
McMahon [30]. Since a limited number of measure-
ments (about 20 for each specimen) were carried out,
the authors’ misorientation distributions are not
‘smooth’ and discontinuous values are observed within
the range 5–15°, instead of the continuous distributions
displayed here (Figs. 5 and 11). This is a reason why
these medium angle boundaries, which were also
present in single crystalline specimens strained under
similar conditions [41], were interpreted as interfaces
between persistent deformation bands (although such
bands were not clearly identified), and not as former
subgrain boundaries transformed into new grain
boundaries.
It should be noted that the classification of the
various boundary types has been developed in the case
of cold deformation and its application to hot worked
structures in not obvious. Indeed, at temperatures be-
low 0.4 Tm, several kinds of interfaces are observed.
The initial grains decompose into deformation bands
separated by a thin (1–2 mm) transition band contain-
ing dislocation cells. Inside the deformation bands,
random low misorientation cells are grouped together
in cell blocks separated by dense dislocation walls of
higher misorientations [43]. However, as temperature
increases, dislocation walls become thinner and a much
Fig. 12. Strain dependence of the surface fractions pertaining to the more homogeneous microstructure develops. Generally,
various kinds of interfaces, viz. low angle boundaries (LAB), new above 0.6 Tm, only equiaxed subgrains are observable
high angle boundaries (nHAB) [including the deformation band
boundaries (DB)], and old ‘initial’ high angle boundaries (oHAB). (a)
inside the deformed initial grains. Even when a decom-
1199 polycrystals; (b) 1200; (c) 5052; (d) 1199 Ž001; (e) 1199 Ž111. position occurs, the deformation bands are not easily
S. Gourdet, F. Montheillet / Materials Science and Engineering A283 (2000) 274–288 285

observed on micrographs (Fig. 6). The reason is that leading to the formation of high angle grain
they appear bounded by true grain boundaries, instead boundaries. A question still remains open: which
of the straight above-mentioned transition bands. Dif- mechanism leads to the increase in misorientation?
ferentiation between several types of low and high angle If one considers that dislocations of opposite signs
boundaries is therefore difficult in hot worked struc- are created in equal densities during straining, each
tures, and the relevance of such a classification even dislocation wall will absorb dislocations of both
questionable. signs, keeping its misorientation at a low level. A
misorientation increase can only occur if a subgrain
4.2. Influence of crystalline orientation boundary absorbs an excess of dislocations of one
sign, which supposes that the various types of dislo-
Results obtained on single crystals confirm that the cation are not uniformly distributed in the material.
CDRX kinetics strongly depend on the crystallographic This assumption does not seem unrealistic, however,
orientation. The fully stable Ž011 orientation enables all the more as misorientations between adjacent
only a limited increase of the misorientations. No high subgrains will increase such inhomogeneities. In-
angle boundary creation was observed, although some deed, slip system activities can be affected by small
misorientations are close to 15° at o =0.9. In the case of orientation changes. Furthermore, among the vari-
the quasi-stable Ž001 orientation, the formation of ous orientations introduced by strain, those belong-
new high angle boundaries, with 15 – 30° misorienta- ing to the Ž011 fiber are fully stable and will
tions, was observed in the specimens strained to o= 0.9 probably not disappear, thus leading to increased
and 1.5. For the Ž111 orientation, which is fully and permanent misorientations. A slight trend to-
unstable, new grain boundaries in the same misorienta- wards fiber formation in the pole figure of the Ž011
tion range were observed, starting even at o =0.3. Such crystal strained to o =1.5 can be noticed (Fig. 2(b)).
results are in total agreement with previous work on If compression specimens could be deformed to very
aluminium single crystals. Theyssier et al. [40] observed, large strains, a Ž011 fiber texture would certainly
after channel die compression of single crystals of the be observed.
same purity, that some orientations (e.g., {112}Ž111 2. Unstable orientations. New interfaces can be intro-
or {421}Ž112) led to the formation of high angle duced by both strain hardening and lattice rota-
boundaries, whereas other ones (e.g. {110}Ž112) did tions. Two cases must be distinguished, according to
not. More recently, Ponge et al. [6] investigated the whether deformation bands occur or not. Let us first
misorientations developed during hot uniaxial compres- consider the case in which the whole crystal rotates
sion (260°C, 4× 10 − 4 s − 1) of a very high purity alu- towards the same orientation. In the Ž111 single
minium (99.9995 wt.%), which recrystallized by DDRX. crystal strained to o = 0.3, the EBSD local pole
In the unrecrystallized matrix, these authors observed figure clearly shows that some orientations still re-
the occurrence of smaller misorientations in the Ž011 main close to the initial one, whereas others are
single crystal (B9°), than in the Ž112 specimen (up to already located near the final one. That is why high
26°). The misorientation ranges were thus very close to angle boundaries are observed at such a low strain.
those obtained in the present work. However, at o= 1.5, all the crystallites have reached
their final Ž011 orientation, which explains the
4.3. Continuous dynamic recrystallization and grain lower fraction of high angle boundaries. The subse-
rotations quent behavior has been described in (i). In the
second case, when the initial orientation splits into
In order to clarify the origin of the new high angle symmetrical components, very large and permanent
boundaries, it is interesting to look more closely at the misorientations are rapidly built up. Deformation
relationship between global texture and local misorien- bands are expected to occur only for specific initial
tations. In the case of uniaxial compression, the follow- orientations, e.g. Ž001, Ž111, or the intermediate
ing remarks can be made: Žuuw orientations, and they are more likely to
1. Stable orientation. During straining, the global tex- occur in single crystals than in polycrystals (except
ture of the Ž011 crystals remains unchanged, ex- for very large initial grains).
cept that it becomes slightly less sharp. This is due Lyttle and Wert [23] have formulated three models
to strain hardening and dynamic recovery: disloca- based on dislocation glide, boundary sliding, and neigh-
tions accumulate in the subgrain boundaries, thus bor switching to account for the increased misorienta-
altering the initial orientation. Misorientations up to tions during straining of superplastic alloys. They
15° have been reported in pure aluminium single concluded that combination of the boundary sliding
crystals. The results obtained on polycrystals indi- model and the neighbor switching model most closely
cate that in a less pure aluminium, such as the 1200 reproduced the misorientations measured experimen-
grade, the misorientations are probably larger, thus tally. In the case of the dislocation glide model, the
286 S. Gourdet, F. Montheillet / Materials Science and Engineering A283 (2000) 274–288

misorientations tended to decrease rather than increase, parameter of the flow stress. The assumption of Mc-
therefore reflecting texture development. However, it Queen and Blum was based mostly on creep data from
can be inferred from the above discussion that the specimens deformed to low or moderate strains, which
convergence of the crystallite orientations and the cre- can explain why these authors did not observe any
ation of large misorientations are not incompatible. A misorientation increase of the subgrain boundaries.
reason is that strain hardening tends to move the
crystallites away from their ideal orientations. This 4.5. Elementary mechanisms of the continuous dynamic
effect was probably very pronounced in this case be- recrystallization
cause of the low recovery level associated with high
alloying element content. Moreover, it was experimen- Continuous dynamic recrystallization has sometimes
tally found that, for some reason, the lattice rotation been labeled extended dynamic recovery, either because
rate varied from one crystallite to another, thus intro- some authors restrict the term recrystallization to the
ducing high misorientations during the transient. All classical discontinuous recrystallization, or because the
these disturbing factors were not taken into account in microstructures generally consist of crystallites only
the model. The increase in subgrain boundary misorien- partially bounded by high angle boundaries. This termi-
tation by accumulation of dislocations was probably nology is somewhat misleading, however, since dynamic
predominant in the early steps of straining. However, recovery can have detrimental as well as beneficial
since the crystallite size was very small, grain boundary effects on CDRX. Indeed, the above experimental ob-
sliding and grain switching probably prevailed as soon servations indicate that the CDRX process results from
as large enough misorientations were built up. the combination of three elementary mechanisms:
1. The formation of subgrain boundaries. These
4.4. Kinetics of the continuous dynamic recrystallization boundaries are created with a very low misorienta-
tion angle (about 1°), as a result of dynamic
It is worth noting that the various mechanical and recovery.
structural parameters tend to their steady state values 2. The transformation of subgrain boundaries into
at different rates. The flow stress remains approxi- grain boundaries. The increase in misorientation of
mately constant from o =0.3, except for the Ž111 the subgrain boundaries is more or less rapid, ac-
crystal. In this case, the crystallographic rotation delays cording to the material and the experimental condi-
the steady state up to o =0.9. The subgrain size is tions. It was shown that it is accelerated by medium
generally well established at o = 0.3. However, a sub- recovery levels and when the initial orientation is
grain refinement is still observed in the 5052 alloy at unstable. If such favorable conditions are brought
o= 1.5. By contrast, the misorientations are not yet together, the subgrain boundaries can be gradually
stabilized at o=1.5, with the only exception of the transformed into grain boundaries.
Ž011 crystal. This is due to the combined effects of the 3. The elimination of subgrain and grain boundaries.
increasing misorientations of the subgrains and newly Measurements carried out on a b titanium alloy
formed grain boundaries on the one hand, and the [13,47] and some aluminium alloys have recently
increasing surfaces of the original grain boundaries on shown that the grain boundaries migrate, even in
the other hand. Apart from some minor cases, the flow the absence of classical DDRX, although at much
stress and the average subgrain size thus reach their lower rates. All the interfaces present in the volume
steady state values quite rapidly (o = 0.3), while the swept by the migrating boundaries disappear, which
average misorientation is still increasing at o = 1.5. certainly plays a major role in the establishment of
These experimental results are thus in contradiction the steady state, for both GDRX and CDRX. It has
with the similitude principle, which stipulates that the been shown for instance that, in compression, the
various microstructural spacings are inversely propor- initial grains can reach a quasi-steady state thick-
tional to the flow stress. For instance, McQueen and ness, which is an increasing function of the
Blum [42,44] proposed that there is a unique relation- boundary velocity [13,47]. Moreover, the elimina-
ship between the average dislocation spacing s in the tion of interfaces allows the subgrain size and the
subgrain boundaries, (or, equivalently, the average sub- misorientation distribution to stabilize after a tran-
grain boundary misorientation u: b/s) and the flow sient period.
stress. The fact that the average subgrain boundary From the previous experimental observations, a
misorientation increases without affecting the flow model of CDRX has recently been proposed, in which
stress significantly can be explained if one considers, as the above three mechanisms are combined in order to
established by several authors [45,46], that the strength- predict the microstructural evolutions [38,48]. The main
ening associated with dislocations inside the subgrains features are well reproduced: during the transient, the
is larger than that due to dislocations in the subgrain subgrain size decreases while the misorientation grows,
boundaries. The former is thus the main controlling and the subgrain boundaries are gradually converted
S. Gourdet, F. Montheillet / Materials Science and Engineering A283 (2000) 274–288 287

into grain boundaries; ultimately, the crystallite size Acknowledgements


reaches a steady state value, as well as the misorienta-
tion distribution, although at larger strains. The authors are indebted to Professor J.J. Jonas,
McGill University, Montreal, for providing access to
the torsion facility. They are also grateful to Professor
H.J. McQueen, Concordia University, Montreal, for
5. Conclusions many fruitful discussions. The work of S. Gourdet was
supported in part by the Région Rhône-Alpes, France,
Various aluminium specimens were submitted to uni- through a scientific fellowship (‘Avenir’ program).
axial compression and torsion testing at 0.7 Tm and
10 − 2 s − 1, up to o =1.5. The main results obtained after
investigation of the hot worked structures are the Appendix A
following.
(1) Experiments carried out on Ž001, Ž011 and The evolution of the fraction of the various interface
Ž111 single crystals confirm that the subgrain types (i.e. low angle boundaries, initial and new high
boundary misorientations strongly increase with strain. angle boundaries) is addressed here in the case of
For the three investigated orientations, the largest val- compression. Similar equations apply for torsion (see
ues are close to or beyond 15°. In the case of the Ref. [38] for more details). Since experimental measure-
stable Ž011 orientation, the increase in misorien- ments provide interface lengths per unit area, a two-di-
tation is not sufficient to transform low angle into mensional approach was chosen. Initial grains and
high angle boundaries. Among the two other orienta- subgrains are approximated by ellipses with semiaxes
tions, the conversion occurs earlier for the most un- a = p/4 DCA and b= p/4 DTD, where DCA and DTD
stable one: it starts at o = 0.3 in the unstable Ž111 denote the measured intercept lengths parallel to the
crystal, but only at o =0.9 in the metastable Ž001 compression axis and the transverse direction,
crystal. respectively.
(2) Most of the new high angle boundaries have The mean intercept lengths of the initial grains can
15–30° misorientations. They originate either from dif- be accurately measured only at o= 0. Their evolutions
ferences in the rotation rate towards the final orienta- with strain are evaluated by taking into account the
tion, or from fluctuations around the average flattening of the grains and the migration of the initial
orientation introduced by strain hardening and dy- boundaries, which leads to an increase of the average
namic recovery. In some cases, the initial orientation thickness [13,38]. Along the direction parallel to the
splits into symmetrical components, causing the occur- compression axis:
rence of deformation bands. In this case, very high
D: CA = − DCAo; + 26 (1)
misorientations (\ 30°) are rapidly built up.
(3) In the case of polycrystals, the high angle where the migration rate 6 =0.1 mm/s at o; =0.01 s − 1
boundary fractions strongly increase with strain. Since [13,38]. This yields after integration:
both continuous and geometric dynamic recrystalliza-
DCA = (D0 − DS) exp(−o)+ DS (2)
tion are likely to occur, calculations were carried
out in order to estimate the surface fraction of the where D0 is the initial length and DS = 26/o; , or, for the
initial grain boundaries. It turns out that the high angle semiaxis:
boundary fraction is much larger than could be ac-
ag = (a0 − aS) exp(−o)+ aS (3)
counted for by the expanded initial boundaries, thus
confirming the presence of many new high angle where aS = p6/2o; . Moreover, perpendicular to the com-
boundaries. pression axis:
(4) Continuous dynamic recrystallization is more effi-
bg = b0 exp(o/2) (4)
cient in the commercial purity aluminium than in the
pure aluminium and the Al-Mg alloy. This (in the above equations, a0 and b0 are the semiaxes
suggests that the transformation of low angle lengths at o =0).
boundaries into high angle boundaries is faster when In addition, the initial grain boundary length is af-
the recovery level is neither too high (the accumulation fected by the presence of serrations. Measurements
rate of dislocations in the subgrain boundaries de- from POM micrographs showed that the length ratio k
creases), nor too low (the formation of subgrains is very between a serrated and a straight boundary first in-
slow or they do not form at all). Comparisons between creases with strain (as subgrains form) and then
compression and torsion specimens also indicate that remains approximately constant. This evolution can be
the strain path does not alter CDRX kinetics notice- accurately described by the equation k=1+
ably. 0.25 [1− exp(− 4o)] . The current length per unit
288 S. Gourdet, F. Montheillet / Materials Science and Engineering A283 (2000) 274–288

area of the initial grain boundaries is then given by [20] P. Cizek, B.P. Wynne, Mater. Sci. Eng. A230 (1997) 88–94.
[21] R.H. Bricknell, J.W. Edington, Metall. Trans. 10A (1979) 1257–
kPg/2pagbg, where the ellipse perimeter Pg is calculated
1263.
numerically from the lengths of the semiaxes ag and bg. [22] R.H. Bricknell, J.W. Edington, Acta Metall. 27 (1979) 1303–
Since the subgrain semiaxes asg and bsg can be mea- 1311.
sured at various strains, the total interface length per [23] M.T. Lyttle, J.A. Wert, J. Mater. Sci. 29 (1994) 3342–3350.
unit area is simply given by Psg/2pasgbsg. The total [24] H. Gudmundson, D. Brooks, J.A. Wert, Acta Metall. Mater. 39
(1991) 19 – 35.
(old+ new) high angle grain boundary length is then
[25] E.-W. Lee, T.R. McNelley, Mater. Sci. Eng. 93 (1987) 45–55.
estimated by multiplying the total interface length by [26] S.J. Hales, T.R. McNelley, Acta Metall. 36 (1988) 1229–1239.
the fraction of high angle boundaries measured by [27] T.R. McNelley, E.-W. Lee, M.E. Mills, Metall. Trans. 17A
EBSD (see the misorientation distributions). Finally, (1986) 1035 – 1041.
the length per unit area of the new high angle [28] E.-W. Lee, T.R. McNelley, A.F. Stengel, Metall. Trans. 17A
(1986) 1043 – 1050.
boundaries is obtained by difference.
[29] H.J. McQueen, O. Knustad, N. Ryum, J.K. Solberg, Scripta
Metall. 19 (1985) 73 – 78.
[30] M.E. Kassner, M.E. McMahon, Metall. Trans. 18A (1987) 835–
846.
References [31] J.K. Solberg, H.J. McQueen, N. Ryum, E. Nes, Phil. Mag. 60A
(1989) 447 – 471.
[1] H. Yamagata, Scripta Metall. Mater. 27 (1992) 201– 203. [32] G.A. Henshall, M.E. Kassner, H.J. McQueen, Metall. Trans.
[2] H. Yamagata, Scripta Metall. Mater. 27 (1992) 727– 732. 23A (1992) 881 – 889.
[3] H. Yamagata, Scripta Metall. Mater. 27 (1992) 1157– 1160. [33] W. Blum, Q. Zhu, R. Merkel, H.J. McQueen, Mater. Sci. Eng.
[4] H. Yamagata, Scripta Metall. Mater. 30 (1992) 411– 416. 205A (1996) 23 – 30.
[5] H. Yamagata, Acta Metall. Mater. 43 (1992) 723–729. [34] W. Blum, Q. Zhu, R. Merkel, H.J. McQueen, Z. Metallkde 87
[6] D. Ponge, M. Bredehöft, G. Gottstein, Scripta Mater. 37 (1997) (1996) 14 – 23.
1769 – 1775. [35] B. Dumanowski, Thesis, 1997, Ecole des Mines de Saint-Etienne,
[7] C. Chovet, S. Gourdet, F. Montheillet, Int. Conf. UHPM 98, France.
France, pp. 67 – 74. [36] C. Perdrix, F. Montheillet, G. Wyon, F. Weill, Metallography 13
[8] F.R. Castro-Fernandez, C.M. Sellars, Mat. Sci. Tech. 4 (1988) (1980) 289 – 297.
621 – 627. [37] Q. Zhu, H.J. McQueen, W. Blum, High Temp. Mat. Proc. 17
[9] T. Sheppard, M.G. Tutcher, Met. Sci. 14 (1980) 579– 589. (1998) 289 – 297.
[10] C. Perdrix, M.Y. Perrin, F. Montheillet, Mém. Sci. Rev. Metall. [38] S. Gourdet, Thesis, Ecole des Mines de Saint-Etienne, France,
78 (1981) 309 – 320. 1997.
[11] F. Montheillet, in: P. Costa, et al. (Eds.), 24ème Colloque de [39] A. Mécif, B. Bacroix, P. Franciosi, Acta Mater. 45 (1997)
Métallurgie, INSTN, Saclay, 1981, pp. 57–70. 871 – 881.
[12] S. Gourdet, E.V. Konopleva, H.J. McQueen, F. Montheillet, [40] M.-C. Theyssier, B. Chenal, J.H. Driver, N. Hansen, Phys. Stat.
Mat. Sci. Forum 217 (1996) 441–446. Sol. (a) 149 (1995) 367 – 378.
[13] S. Gourdet, A. Girinon, F. Montheillet, in: T. Chandra, T. Sakai [41] M.E. Kassner, Metall. Trans. 20A (1989) 2182 – 2185.
(Eds.), Int. Conf. Thermec ‘97, The Minerals, Metals and Mate- [42] H.J. McQueen, W. Blum, ICAA6, Japan Inst. Metals, pp. 99–
rials Society, Wollongong, Australia, 1997, pp. 2117– 2124. 112.
[14] F. Chaussy, J.H. Driver, in: A. Vassel, D. Eylon, Y. Combres [43] J.H. Driver, 16th Risø Int. Symp., Danemark, 1995, pp. 25–34.
(Eds.), Les Alliages de Titane b, vol. 8, Editions de la Revue de [44] H.J. McQueen (section 9, R.D. Doherty, D.A. Hughes, F.J.
Métallurgie, Paris, 1994, pp. 57–64. Humphreys, J.J. Jonas, D. Juul Jensen, M.E. Kassner, W.E.
[15] A.-M. Chaze, F. Montheillet, in: A. Vassel, D. Eylon, Y. Com- King, T.R. McNelley, H.J. McQueen, A.D. Rollett), Mat. Sci.
bres (Eds.), Les Alliages de Titane b, vol. 8, Editions de la Revue Eng. A238 (1998) 219 – 274.
de Métallurgie, Paris, 1994, pp. 41–48. [45] M.E. Kassner, A.K. Miller, O.D. Sherby, Metall. Trans. 13A
[16] F. Chaussy, J.H. Driver, Rev. Métall.-CIT/Sci. Génie Matér. 93 (1982) 1977 – 1986.
(1996) 1057 – 1066. [46] F.R. Castro-Fernandez, C.M. Sellars, Phil. Mag. 60A (1989)
[17] L. Lombry, C. Rossard, B.J. Thomas, Rev. Métall./CIT 78 487 – 506.
(1981) 975 – 988. [47] A. Girinon, Mémoire CNAM, 1999, Conservatoire National des
[18] C.G. Schmied, C.M. Young, B. Walser, R.H. Klundt, O.D. Arts et Métiers, Saint-Etienne, France.
Sherby, Metall. Trans. 13A (1982) 447–456. [48] S. Gourdet, C. Chovet, F. Montheillet, in: T. Sakai, H.G. Suzuki
[19] A. Belyakov, R. Kaibyshev, T. Sakai, Metall. Trans. 29A (1998) (Eds.), 4th Int. Conf. on Recrystallization and Related Phenom-
161 – 167. ena, The Japan Institute of Metals, Tsukuba, 1999, pp. 259–264.

You might also like