You are on page 1of 14

Available online at www.sciencedirect.

com

Acta Materialia 58 (2010) 16291642


www.elsevier.com/locate/actamat

Microtexture tracking in hot-deformed polycrystalline


aluminium: Experimental results
R. Quey *, D. Piot, J.H. Driver
Ecole des Mines de Saint-Etienne, Centre SMS, Laboratoire PECM CNRS UMR 5146, 158 cours Fauriel, 42023 Saint-Etienne, CEDEX 2, France
Received 9 September 2009; received in revised form 4 November 2009; accepted 5 November 2009
Available online 16 December 2009

Abstract
A split sample of Al0.1%Mn has been deformed by a series of compression tests in a channel-die at 400 C to a nal strain of 1.6. The
orientations of 176 grains in a 4  4 mm2 internal surface were followed by high-resolution electron backscatter diraction at four different strains to compare with crystal plasticity models. Typically 3000 orientations per grain were used to quantify the average lattice
rotations of each grain together with their orientation spreads (termed microtexture tracking). The average orientations tend towards the
standard b-bre plane-strain compression texture components, albeit with some variations. The in-grain orientation spreads develop
strongly at rst, then tend to saturate at high strains. Finally, the inuence of grain environment on lattice rotation is examined by means
of the rotation variability at constant orientation. On average and at the beginning of the deformation, two grains of the same initial
orientation, but dierent neighbours, would rotate by angles that vary by 25% and axes separated by 37; their orientations at e 1:2
would vary by 12.
2009 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: Microstructure; Plastic deformation; Deformation structure; Electron backscattering diraction (EBSD); Aluminium alloys

1. Introduction
Accurate deformation texture simulations require
advanced models of polycrystal deformation for which
there are now several new variations that go beyond the
standard Taylor model to incorporate grain interaction
eects [14]. These models, although based on the behaviour of the individual grains in an aggregate, are usually
evaluated by a comparison of the experimental and simulated macrotextures of a large number of grains. Clearly,
a comparison of the rotations of individual grains in their
environments would be much more instructive. However,
there are relatively few such studies, since they require difcult experimental procedures.

*
Corresponding author. Present address: Sibley School of Mechanical
and Aerospace Engineering, Cornell University, Ithaca, NY 14853, USA.
E-mail addresses: quey@emse.fr (R. Quey), piot@emse.fr (D. Piot),
driver@emse.fr (J.H. Driver).

A rst approach is to use two-dimensional (2-D) microstructures, as done by Skalli et al. [5] and later Fortunier
and Driver [6] using large-grained aluminium. A similar
method is to use a columnar grain structure, within the limitations of a h1 0 0i-bre texture [7]. These studies are of
interest, but may not be representative of what occurs in
a real polycrystal. Two strategies have been proposed to
follow grains in a 3-D polycrystal. In situ 3-D X-ray diffraction characterization of grain average rotations has
been used by Poulsen et al. [8], but up to now the technique
appears limited to a tensile strain less than 0.1. The other
method is to use a split sample, as rst done by Barrett
and Levenson [9] for uniaxial compression, then by Panchanadeeswaran et al. [10] for hot plane-strain compression
(PSC). In the latter case, an AA 1050 sample was split in
half perpendicular to the transverse direction and grain orientations were measured by electron backscatter diraction
(EBSD) on the internal surface before and after a strain
of 0.5. In these studies on 3-D polycrystals, the experimental rotations were compared to those predicted by the

1359-6454/$36.00 2009 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2009.11.007

1630

R. Quey et al. / Acta Materialia 58 (2010) 16291642

Taylor model. Their conclusions are contradictory: from 3D X-ray diraction microscopy, Winther et al. [11] found
the Taylor model to be successful for 75% of the grains,
whereas Panchanadeeswaran et al. [10] concluded that it
fails almost completely.
Although Panchanadeeswaran et al. [10] have demonstrated the potential of the split-sample technique to follow
grains in a polycrystal during large deformations, their
study is not really satisfactory because few orientation
measurements per grain were made and an accidental
macroscopic shear strain (as large as 1/3 of the imposed
PSC) occurred in the sample during the deformation.
In a recent study by Quey et al. [12], a split sample of
Al0.1 wt.%Mn was deformed in PSC at 400 C to a strain
of 0.4. After each of the two passes, the sample was
quenched to avoid recrystallization, and the deformation
microstructure was analysed by EBSD. In the present
study, this experiment has been repeated with another sample, up to a strain of 1.6. The same 176 grains were followed throughout the deformation with an average of
3000 measurements per grain. This paper rst discusses
the inuence of analysing grain rotations on the internal
surface of a split sample. Then, the rotation properties of
the 176 grains are described in detail, in terms of average
rotations and in-grain orientation spreads. Finally, a
method is proposed to investigate the possible eects of
grain interaction on these rotation properties.
A comparison between the experimental results and the
predictions of various polycrystal deformation model is
provided in a second article [13].
2. Experimental
The high-purity binary Al0.1 wt.%Mn alloy was provided by the Alcan Research Centre at Voreppe. After casting, it was cold-rolled 50% then recrystallized at 530 C for
5 min. The resulting microstructure showed near-equiaxed
grains of size about 300 lm and a weak crystallographic texture. A split sample of 8  7  10 mm (along the rolling
direction (RD), transverse direction (TD) and normal direction (ND), respectively) was used. It was made of two identical parts assembled along the transverse direction, see
Fig. 1. The sample surfaces were machined at, and both
internal surfaces were mechanically then electrolytically
polished. At the centre of one of them, a 4  4 mm2 observation zone was marked with microhardness indentations of
diameter about 30 lm, located 1 mm apart. The sample
was deformed by successive heating, compression and
quenching cycles, as illustrated in Fig. 2. In this way, the
microstructure could be analysed at successive (logarithmic)
strains of 0, 0.19, 0.42, 0.77, 1.2, and 1.6. After closing, the
two parts of the sample were held together through glued
points at the corner of the internal surface, and then wrapped
in Teon lms to reduce friction eects. The hot channel-die
compression was imposed at a strain rate of 1 s1 and a temperature of 400 C in the equipment described by Maurice
et al. [14]. Typical heating times to temperature stabilization

ND

TD
RD

Fig. 1. Split sample deformed in channel-die compression.

T ( C)

PSC to
0.19

PSC to
0.42

PSC to
0.77

PSC to
1.2

PSC to
1.6

400
Heating ( 60 s)
Deformation at 1 s
Quenching (

2 s)
20
EBSD

time
(EBSD)

EBSD

EBSD

EBSD

(EBSD)

Fig. 2. Cycles of hot plane-strain compression and EBSD analysis.


(EBSD) denotes analyses not reported in this paper.

were about 1 min. After deformation, the sample was water


quenched within about 2 s to retain the deformation microstructure (the sample does not recrystallize under these conditions due to the strong eect of Mn in solid solution
inhibiting grain boundary movement). After removing the
Teon lms the two parts of the sample adhered, but could
be easily separated for metallographic observation. Before
deformation and after each pass, the grain structure was analysed as hot deformed, without any additional polishing
(except at the strain of 0.42, where the two parts of the sample
moved slightly with respect to each other during quenching.
A small amount of electropolishing was conducted to clean
up the surface, removing no more than 35 lm of material.).
The analyses were done by EBSD in a JSM 6501F FEGSEM
equipped with an HKL EBSD system, using a scanning step
of 5 lm. The orientation maps were studied with Hermes, an
in-house software development built on Orilib [15]. The
grains were detected automatically at e 0 using a minimum
disorientation angle for the grain boundaries of 5 (this value
may seem small with respect to the usual 15, but was
adopted to obtain near single-crystalline grains). At higher
strains, it was not possible to use this automated procedure
due to the development of in-grain local disorientations, so
the grains were delineated by hand. Only the data at strains
of 0, 0.42, 0.77, and 1.2 are reported here, with the aim of
studying successive grain rotations for three similar strain
increments of about 0.4. At a strain of 1.6, EBSD analysis
became impossible in some regions due to a pronounced surface rumpling (see Section 3.2).

3. Validation of the method

(a)

The fact that grains are observed on the internal surfaceof the sample is the main limitation of the method
and could be open to criticism concerning its relevance to
the behaviour of grains in a real polycrystal. The purpose
of this section is to ensure, as far as possible, that this does
not signicantly aect grain rotations. This is achieved
partly by following the methodology introduced by Panchanadeeswaran et al. [10].

3.2. Aspect of the internal surface


After deformation, the internal surface appeared rumpled, as a result of strain heterogeneities at the grain scale.
This is illustrated on Fig. 4 by rugosity proles made along
RD and ND at e 0:42. The maximal out-of-plane displacements are of the order of 30 lm (independent of the
direction). The local slopes appear less pronounced along
RD than along ND, which is easily explained by the fact
that although the out-of-plane displacements are the same
in both directions, the distances over which they occur are
proportional to the grain size, and so are higher along RD
due to the sample elongation. At e 0:42, the slopes along
RD are less than 10 and, by grain elongation, were found

1 mm

30
20
10
0
10
20
30
7 6 5 4 3 2 1 0

altitude [m]

(b)

30
20
10
0
10
20
30
7 6 5 4 3 2 1 0

1 mm

distance [mm]
Fig. 4. Internal surface rumpling analysed by rugosimetry proles along
(a) ND and (b) RD. Measurements carried out on a sample deformed at
e 0:42. Note that there are two orders of magnitude between the
horizontal and vertical axes, and that the local slopes are more
pronounced along ND.

not to exceed this value at higher strains. On the contrary,


the slopes along ND are close to 20 at e 0:42 and were
found to attain values as high as 60 at the higher strains.
Usually, for the EBSD analysis of a at RDND surface,
the sample is tilted 70 about RD to avoid major defocusing and geometrical constraints in the scanning electron
microscope. In this conguration, the angle between the
electron beam and ND is equal to 20. In our case, the local
slopes along ND can be higher than 20, which would
cause parts of the surface to be hidden from the beam during the EBSD analysis (shadowing). To make the whole
surface visible, it was necessary to tilt the sample about
ND.
As found by Panchanadeeswaran et al. [10], no galling
or sliding marks were observed on the internal surface after
deformation indicating that the two halves of the sample
adhered during deformation. This is attributed to the local
friction stresses, and is almost certainly favoured by the
surface rumpling. This makes the internal surface

1 mm

(a)

1631

distance [mm]

3.1. Deformation mode


The deformation mode experienced at the scale of the
sample can be characterized by the array of microhardness
indentation marks on the internal surface. They are represented, at successive strains, on Fig. 3. It appears that the
sample undergoes a nearly pure PSC at the beginning of
the deformation e 00:4, then shows additional shears
caused by the sample/tool friction. At e 1:2, shears are
noticeable, with maximum values of 0.20.3 at the extremities of the zone of observation. The study of lattice rotations
presented in the following, as well as the comparison with
simulations provided separately [13], show that the shears
do not aect signicantly grain rotations, so they are
neglected in the presentation of the results.

altitude [m]

R. Quey et al. / Acta Materialia 58 (2010) 16291642

1 mm

(b)

(c)

1 mm

(d)

(e)

Fig. 3. Array of microhardness indentations at successive strains: (a) e 0, (b) e 0:19, (c) e 0:42, (d) e 0:77, and (e) e 1:2.

1632

R. Quey et al. / Acta Materialia 58 (2010) 16291642

equivalent to a mechanical grain boundary, according to


the terminology proposed by Panchanadeeswaran et al.
[10].
3.3. Inuence on orientation measurements
The conditions are clearly non-standard for EBSD analyses, for which a perfectly at surface is usually used. In
this study, the surface rumpling causes defocusing, the
inuence of which on orientation measurements can be
quantied. On a standard (at) sample, the orientation of
a single-crystalline grain was measured in standard conditions (perfect focusing), and then with defocusing, which
was controlled by changing the working distance of the
microscope. The measurement error appeared nearly proportional to the defocusing, about 0:025 =100 lm over a
range of up to 2000 lm. In the deformed sample, the
maximum defocusing amplitude is about 200 lm and so
the angular error does not exceed 0.05. This error is very
small compared to the rotations caused by plastic deformation (see Section 4.2) and so can be neglected.
3.4. Inuence on grain rotations
The internal surface has been described above as a
mechanical grain boundary. The inuence of such conditions on grain rotations can be characterized by comparing
the orientations on the internal surface to reference orientations, which would not suer from such conditions.
Firstly, the macrotexture on the internal surface can be compared with the one measured on a cut surface of a standard
sample deformed in the same conditions. This is illustrated
by the {1 1 1} pole gures in Fig. 5. Within the statistical error
arising from the limited number of grains, the macrotextures
can be considered similar. Furthermore, the local orientations on the internal surface can be compared with those of
the same grains just below, in the subsurface regions. The latter can be obtained after removal of 3050 lm of material
from the internal surface by electropolishing. As this method
is destructive, it was not done on the principal sample followed in this study, but on another sample deformed under

(a)

the same conditions to e 0:5. The orientation maps are


illustrated in Fig. 6, where the seven studied grains are
marked. Their orientations in the surface and subsurface
can be compared quantitatively in terms of average orientation and in-grain orientation spread [16]. The latter is calculated as the average of the disorientation angles with respect
to the average orientation. The results are given in Table 1
and show that the average orientations dier only by 1
and the orientation spreads by 0.5 (relative dierence of
7%). These values are small compared to those due to plastic
deformation (see Sections 4.2 and 4.3). These two comparisons show that the rotations on the internal surface are close
to the ones in volume, so that the grains of this study are considered as grains of a real polycrystal.
4. Results
4.1. Microtextures and macrotextures
The microtextures obtained at the successive strains are
illustrated by Fig. 7. The 176 grains that were followed during the deformation are numbered and their contours are
marked. In practice, grains exhibit two dierent rotation
modes: either a unimodal rotation, composed of an average rotation and an orientation spread, or fragmentation.
From e 01.2, it was found that 90% of the grains
undergo a unimodal rotation, conrming the results of
the previous study by microtexture tracking [12]. Only
these grains are considered in this study. The specic case
of the grains undergoing fragmentation, usually high-symmetry orientations, will be reported separately. The example of a grain undergoing a unimodal rotation is illustrated
in Fig. 8. As a large number of orientations are known, it is
possible to accurately calculate an average orientation and
an in-grain orientation spread.
The corresponding macrotextures can be characterized
in terms of number of grains in the components of the bbre. The bre runs through the ideal components Copper
{1 1 2}h1 1 1i, S {1 2 3}h634i and Brass {110}h112i. These
are approximate positions, however; their exact orientations in our experiments are, in Euler angles (Bunge con-

(b)

Fig. 5. The inuence of the internal surface on the macrotexture. (a) Macrotexture measured on the internal surface of the split sample. (b) Macrotexture
measured on a cut surface of a standard sample. Both measured at e 0:77.

R. Quey et al. / Acta Materialia 58 (2010) 16291642

1633

(b)

(a)
1
4
5

2
4
5

100 micrometers, step = 5 micrometers

100 micrometers, step = 5 micrometers

Fig. 6. The inuence of the internal surface on the microtexture of a sample deformed at e 0:5. Rodrigues vector maps measured (a) on the internal
surface andp(b) in subsurface
p
(after removal of 3050 lm of material). (R being a Rodrigues vector, the RGB colour levels are
255  Ri 2  1= 2  2  1 with i 2 f1; 2; 3g, respectively; grey is for non-indexed points). The seven grains considered are marked.

Table 1
Inuence of the internal surface on the microtexture: dierence between the orientations of grains measured in surface and subsurface (after removal of
3050 lm of material), as illustrated in Fig. 6. (a) Average orientations and (b) average in-grain disorientations hd .
(a)
n

Surface orientation u1 ; /; u2 ()

Subsurface orientation u1 ; /; u2 ()

disorientation ()

1
2
3
4
5
6
7

(164.5, 33.2, 1.7)


(43.5, 54.3, 16.4)
(33.7, 78.1, 31.8)
(69.5, 52.4, 10.3)
(39.6, 72.2, 10.1)
(148.5, 34.5, 6.6)
(33.3, 69.5, 15.5)

(163.3, 33.2, 2.6)


(43.4, 55.4, 15.8)
(34.4, 77.7, 32.1)
(69.6, 51.9, 11.1)
(39.4, 71.6, 10.5)
(146.8, 34.4, 7.8)
(33.9, 68.4, 15.1)

0.7
1.3
0.9
1.0
0.8
1.0
1.2
Average = 1.0

(b)
n

Surface hd ()

Subsurface hd ()

Relative dierence (%)

1
2
3
4
5
6
7

6.4
5.2
5.8
9.9
11.0
7.7
5.8

6.0
5.7
6.1
10.6
10.6
6.8
6.2

6
10
5
7
4
12
7
Average = 7

vention): (90, 27.5, 45), (57.5, 30, 75) and (35, 35,
80), respectively. The grains are characterized in terms
of their average orientations, and are assigned to a component using the standard 15 criterion. The results at the successive strains are reported in Table 2. It appears that each
b-bre component systematically increases, while the number of grains outside the bre continuously decreases.
These results are consistent with previous hot-deformation
macrotexture studies (e.g. [17]).
4.2. Average rotations
The average rotations are expressed in the sample coordinate system as axis/angle pairs r; h. One can consider
either the rotations at the successive strains e measured
from the initial orientations, noted re0 ; he0 , or the rotations
for the successive strain increments, called incremental
rotations and noted rei ; hei . A study of the rotation axis
r requires particular precautions, especially when there
are uncertainties for the rotation angle h. Bate et al. [18]
have shown that the angular accuracy b of the axis r is
given by tan1 d=h, where h is the rotation angle and d

its absolute accuracy (both expressed in degrees). With a


typical value of d 1 attributed to alignment error in
the electron microscope and a maximum angular value of
the accuracy on the axis of b 20 , the rotation angle h
must be greater than 2.7. As a consequence, in the following only the grains whose incremental rotations are greater
than 2.7 are taken into consideration for the study of the
rotation axis.
The use of axis/angle pairs for the rotations is conventional, but implies direct, or linear, rotations from one orientation to the next. This is unlikely to be correct for large
rotations, e.g. those from the initial to the nal orientations
1:2
r1:2
0 ; h0 , but not far from real behaviour for the incremental rotations which are preferred here (noting that the
incremental rotations should not be too small because of
the above uncertainty problem for the axes).

4.2.1. Rotations with respect to the initial orientations


re0 ; he0
The frequency distributions of the rotation angles he0 at
successive strains are represented in Fig. 9a. As expected,

1634

R. Quey et al. / Acta Materialia 58 (2010) 16291642

(a)

035

034
169 026

063
007
099
189
056
085
087
173
016

149
158
073

120
101 068
191

021

010

170

024
090

137

096

076
138
041 168

075 143
118
185 197
188

178
124
047 205 200

032
042
135
179203
136 196 102 095
062
183193 043
036
119 150
164
019
088
125
003
054
013
072
162
097

015

038

172

037

018

057

081

070

094

017

011

050

152

161

121

199
159

006

020
116 069

089

083

051
201
167
130

105

139

100

059

080

030

049

053

065

151140
142
009

045

014

184
175 145
112 115
084
111

001

160

002

171
061
148
126
123
165 129
023
128
156
027
048
110
195
109
202 092
082
103
077
177
044
012
182
146
033
133
131
104 064 079
055 204 052 107 174 098
207
153
141
127 122
060
039
066 198
093 091
190
058
029
154
186
194
008
206 031
192
040
181
132 046 074
157 108
028
005
144
155
113
078
067
106
086

022

147

176
134 180

071
117 163114
004

187 025
166

500 micrometers; step = 5 micrometers; grid 871x806

(b)

035

105

034
116
063

089
159
083

099
189
056

020

026

169

007
085

016

069 199
057

161
017

018

011

173 087

081

038

021

120
101

191
024
090

006
094

015

088

068

070
172
121
042
179 203
193
183
043
119
150

072
010

170

054

162

045
080 059
184
139
137
100
175
115 145
076 138
030
112
168
041
050
084
152
053
065
151 140 049
111
075 143
158
061
171
009
142
118 185
148
073
123
165
197 188
129
023
128
126
156 110
103
082
177
202 092
077
195 027
109
146
182
012
133
174 098
052 107
204
055
033
044
079
207
104 064
131
122
198
127
141
066
190
153
091
093
029
039
060
058
186
192
206 031
154
040
194
181
157
008
108
074
005
144
132 046
155
106
067
028
113
086
078
149

096

062
164
003

051
201
037
167
178
130
047 205 124 200
032
136 196 102
095
036
019
125
013
097

014
160

001

002
048

147
117 163

022
134
114

176
180

071

004
187

025
166

500 micrometers; stepsize = 5 micrometers; grid 1310x546

Fig. 7. Microtextures at successive strains: (a) e 0, (b) e 0:42, (c) e 0:77 and (d) e 1:2 (Rodrigues vector maps). The 176 grains are numbered and
the black lines represent (a) the boundaries >5, (bd) the contours of the same grains at the subsequent strains.

they vary signicantly from one grain to the other, e.g. at


e 1:2, they go from 2 to 39. However, it appears that,
as deformation increases, the rotation angles increase less:
on average, they are 11 at e 0:42, 15 at e 0:77 and 18
at e 1:2.
The distributions of the rotation axes re0 are illustrated in
Fig. 9b by pole gures which are reduced to one-quarter
using the orthotropic symmetry. Each axis is represented
by a point and a density function is constructed by associating a Gaussian spread of half-width 7 to each axis. The
axes appear to be preferentially distributed about TD at the
rst increment (density = 5), and then about an axis
located between TD and ND, more precisely 0:64

TD 0:77  ND (density = 3). This transition accounts


for irregular rotation paths, which can be better characterized through the incremental rotations.
4.2.2. Incremental rotations rei ; hei
The frequency distribution of the incremental rotation
angles hei are represented in Fig. 10a. To facilitate a comparison between successive increments, the angles can be
normalized to a constant strain increment of 0.4. The normalized angles ^hei are proportional to the angles hei and are
dened by ^hei hei  0:4=De, where De is the increment
deformation (0.42, 0.770.42, or 1.20.77). They take
average values of 10 at e 0:42, 7 at e 0:77 and 5 at

R. Quey et al. / Acta Materialia 58 (2010) 16291642

(c)

1635

004

025

163 114

187

147
117

002

184
059
112 175
111
053
065
148
061
129
023
165
082
103
182
077
177
146
133
098207
174
052 107
204
122 190
127
066
029
206
031
192
157
108
040
181
106
067 144
080

005

055
091
093
155
086

104
039
194
132 046
078

044
153

028
131

113

008

060
033
154

500 micrometers; step = 5 micrometers; grid 2000x377

186

079

074

141
058
128
092
064
027
195

159
083

126
109

152
073

056

016

158 050
118

075
185 197

143

096

087

173
038
149
085

007

012

009

076
151
041

081

021
137
168

011
116
063
026
169
089

034

500 micrometers; step = 5 micrometers; grid 2893x261

010

090
100
138
049
140
171
142
123

024

068
101

069

035

057

020

017

120

015

161

139
030

170

018

094

088

072

119

054

070
042
203
179 183 193
150

162
045
115 145
084

043

048

014

003

062
164

006

166

180
001

022
134
071

201
167
130
032
036
125

051

136

013

047

196

176

037
178
205

124
102
019

097

095

(d)

Fig. 7 (continued)

e 1:2, conrming that the rotation rates tend to decrease


as deformation increases.
The equivalent distributions of the incremental rotation
axes are represented in Fig. 10b. During the deformation,
these axes tend at rst to lie preferentially along TD then
change closer to NDRD, more precisely 0:9  ND
0:4  RD.
4.2.3. The relations to texture development
The evolution of the incremental rotation angles and
axes can be explained in relation to the texture development. The tendency of the rotation angles to decrease with

strain is expected since grains, as they rotate, attain more


stable orientations and so have smaller rotations.
Concerning the rotation axes, relating their distributions
to texture development can be carried out by partitioning
the set of grains in two subsets: the grains that converge
into the b-bre (i.e. which did not belong to the bre at
e 0, but do at e 1:2) and the others. Orientations are
considered to belong to the bre if they belong to one of
its components as described in Section 4.1. The distributions of the rotation axes are given in Fig. 11. The grains
that converge into the b-bre appear to have the same
properties as those of all grains: rotations about TD at

1636

R. Quey et al. / Acta Materialia 58 (2010) 16291642

(a)

(b)

{111}
stereo. proj.

{111}
stereo. proj.

TD

TD

RD

(c)

RD

(d)

{111}
stereo. proj.

1/2-width: 2
levels: 1, 5, 10, 20, 50, 100

{111}
stereo. proj.

TD

1/2-width: 2
levels: 1, 5, 10, 20, 50, 100

RD

TD

RD

1/2-width: 2
levels: 1, 5, 10, 20, 50, 100

Fig. 8. Example of the unimodal rotation of a grain (number 062). (a) e 0, (b) e 0:42, (c) e 0:77, and (d) e 1:2. Note the average rotation and the
orientation spread.
Table 2
Macrotexture evolution through strain, in terms of number of grains
within 15 of the b-bre components (total number of grains = 157).
Component

e0
e 0:42
e 0:77
e 1:2

Copper

Brass

Others

3
11
14
15

7
20
35
44

25
40
42
42

122
86
66
56

the beginning of the deformation, then rotations about


0:9  ND 0:4  RD. On the other hand, the grains which
do not converge into the b-bre show distributions of rotation axes that are fairly uniform (except, perhaps, at the
second increment). This indicates that the preferential locations of distributions of rotation axes are related to the texture development. It can be explained by the fact that the
b-bre, running from Copper through S to Brass, has an
axis that can be expressed in the form of: a  RD
b  ND, with a 0:97; b 0:24 between Copper and S,
and a 0:25; b 0:96 between S and Brass. Thus, TD
is a preferential direction for convergence towards the
b-bre. This is particularly true at the beginning of the
deformation. Subsequently, the grains, which are closer
to the bre, rotate more parallel to it in order to reach a
particular component.

4.3. In-grain orientation spreads


The in-grain orientation spreads are quantied by the
average disorientation with respect to the average orientation [16], noted hd . Before deformation, the grains have
very similar orientation spreads of 0.3 on average, which
is typically the error associated with EBSD measurements.
Their distributions at successive strains are represented in
Fig. 12. It appears that the orientation spreads develop
strongly at the beginning of the deformation, and then
tend to stabilize: their successive average values are 5.1,
6.4 then 7.0. These observations conrm those of Glez
and Driver [19], made on single crystals of stable orientations (Brass, S and U) deformed under the same
conditions.
4.4. Rotation variability at constant orientation
It is often considered that, in a polycrystal, the active slip
systems and the lattice rotations of a grain do not only
depend on its orientation, but also on the interaction with
its neighbours. This means that two grains of initially the
same orientation can rotate dierently according to their
neighbours, known in the widest sense as grain interactions.
Here we shall investigate this eect by what we call the rota-

R. Quey et al. / Acta Materialia 58 (2010) 16291642

1637

(a)

(b)

TD

ND

RD

equal-area. proj.
1/2-width: 7
levels: 1,2,3,4,5,6,7

TD

ND

RD

equal-area. proj.
1/2-width: 7
levels: 1,2,3,4,5,6,7

r00.42

r00.77

TD

ND

RD

equal-area. proj.
1/2-width: 7
levels: 1,2,3,4,5,6,7

r01.2

Fig. 9. Rotations with respect to the initial orientations, at the successive strains: (a) rotation angle h0 and (b) rotation axis r0 . Each dot represents a grain;
the contour plots highlight their distributions.

tion variability at constant orientation (VCO). To our


knowledge, such an eect of individual grain interaction on
grain rotations has never been quantied (at least
experimentally).
4.4.1. Average rotations
Let r1 ; h1 and r2 ; h2 be the rotations of two grains of
the same orientation, but dierent neighbours. The dierences in rotation can be quantied by the following parameters for the rotation angles (Dr h) and the rotation axes
(angle a):


 h1  h2 
r

; a acos r1  r2
1
D h 2
h1 h2 
the values of which are zero if the rotations are equal.
Here, we consider the rotations during the rst strain
0:42
increment r0:42
0 ; h0 . Ideally, one would wish to quantify
the VCO for every orientation. By denition, this could be
done by comparing the rotations of several grains of the
same orientation, but dierent neighbours. However, in
our sample (and in practice for most other samples), every
pair of grains is mutually disoriented by at least a few
degrees, so that it is impossible to compare the rotations of
grains having exactly the same orientation. From a theoretical point of view, an innite number of grains would be
required.

The alternative approach that we propose is to determine the average of the VCO over all orientations. This
is done by a comparison of the rotations of grains of dierent orientations taking all pairs of grains into consideration. This method involves three steps, which lead to
Fig. 13a and c:
1. For every pair of grains, one calculates (i) the disorientation between their initial orientations, allowing for
both the crystal and sample symmetries, and (ii) the differences between their rotations Dr h and a. On the gures, for every pair of grains, Dr h and a are
represented as a function of the disorientation, which
is limited to 20.
2. The average variabilities by disorientation levels are calculated using intervals of 4 . This value has been chosen
to obtain a smooth evolution of the average tendency.
The averages at zero disorientation obviously cannot
be calculated this way because of the absence of data.
3. The values at zero disorientation are calculated by
extrapolating the average tendency from higher disorientations. By denition, the corresponding value of
Dr h or a is the dierence between the rotations of
grains having the same orientation, i.e. the VCO.
The values obtained in the present study are:
Dr h 0:25 (25%) for the rotation angle and a 37
for the rotation axis.

1638

R. Quey et al. / Acta Materialia 58 (2010) 16291642

(a)

(b)

TD

ND

RD

equal-area. proj.
1/2-width: 7
levels: 1,2,3,4,5,6,7

TD

ND

RD

equal-area. proj.
1/2-width: 7
levels: 1,2,3,4,5,6,7

ri0.42

ri0.77

TD

ND

RD

equal-area. proj.
1/2-width: 7
levels: 1,2,3,4,5,6,7

ri1.2

Fig. 10. Incremental rotations. (a) Rotation angle ^hi and (b) rotation axis ri . Each dot represents a grain; the contour plots highlight their distributions.

Another quantity of interest is the rotation relative variability at constant orientation (RVCO). It is dened by
the VCO divided by the overall variability, the latter
being obtained by averaging over all pairs of grains. By
denition, the RVCO takes values between 0 and 1, and
the higher the value, the lower the eect of the orientation
and the stronger the eect of grain interaction. For the
rotation angle and axis, the overall variabilities are
Dr h 53% and a 83 , respectively, and so the RVCO
values are 0.50 and 0.45, respectively.
A non-zero (R)VCO is incompatible with the Taylor
model, for which rotations depend only on orientation.
To validate the method, it has been applied to the rotations provided by the Taylor model for the same set of
176 grainssee Fig. 13b and d. The VCO obtained in
this way are 9%/8, i.e. are not strictly zero, which is
attributed to the limited number of data. For conrmation, the same study was repeated with a more representative set of 2000 random orientations, leading to
variabilities very close to zero. The dierences between
the two can be associated with the statistical error arising
from the limited number of data, and which can therefore be considered to represent the possible error of
the experimental values. One can note, however, that
these errors are small when compared to the experimental VCO, which ensures that the latter can be considered
as valid measurements.

The rotation axis/angle VCO acting through deformation leads to a nal-orientation VCO. The nal orientations are compared by the disorientation angle between
them, and the results obtained are illustrated in
Fig. 14. The nal-orientation VCO is 12 (with an error
of 2.4).
4.4.2. In-grain orientation spreads
The same characterization can be applied to the in-grain
orientation spreads at e 1:2. The dierence in spreads of
two grains can be quantied as for the rotation angles,


 hd  hd 

2
2
Dr hd 2 1d

h1 hd2 
One can study the relation between the in-grain orientation spreads and the initial orientations (as before) or the
current orientations. The results are illustrated in Fig. 15.
The VCO obtained are Dr hd 31% for the initial orientation and Dr hd 24% for the current orientation. The orientation spread is therefore more dependent on the current
orientation than on the initial orientation (lower VCO).
As for average orientations, these values are to be compared to the overall variability, whose value is
Dr hd 34%. For the current orientation, the RVCO is
equal to 0.71. Such a high value means that the dependence
of the in-grain orientation spreads on the grain orientation
is small.

R. Quey et al. / Acta Materialia 58 (2010) 16291642

(a) TD

ND

RD

equal-area proj.
1/2-width: 7
levels: 1,2,3,4,5,6,7

TD

ND

RD

equal-area proj.
1/2-width: 7
levels: 1,2,3,4,5,6,7

TD

ND

RD

equal-area proj.
1/2-width: 7
levels: 1,2,3,4,5,6,7

ri0.77

ri0.42

(b) TD

ND

RD

equal-area proj.
1/2-width: 7
levels: 1,2,3,4,5,6,7

ri1.2

TD

ND

RD

equal-area proj.
1/2-width: 7
levels: 1,2,3,4,5,6,7

TD

ND

RD

equal-area proj.
1/2-width: 7
levels: 1,2,3,4,5,6,7

ri0.77

ri0.42

1639

ri1.2

Fig. 11. Relation of the incremental rotation axes to the texture development. (a) Case of the grains that converge into the b-bre. (b) Case of the other
grains. Note that grains in (a) show the same distribution as all grains (see Fig. 10).

0.28

= 0.42
= 0.77
= 1.20

0.24

Frequency

0.2
0.16
0.12
0.08
0.04
0

11

13

15

17

19

21

Disorientation [degrees]
Fig. 12. In-grain orientation spreads through strain. hd is the average
disorientation with respect to the mean grain orientation.

5. Discussion
5.1. The experiments
This study provides the rst statistically sound set of
rotations of grains in a polycrystal undergoing a large plastic deformation. The rotations of 176 grains were measured
during hot PSC applied in several passes up to a strain of
1.2. The advantage of high-temperature deformation is
twofold: rst the ease of EBSD measurements up to high

strains, and secondly the better deformation homogeneity


at the grain scale. This type of experiment could be performed at room temperature, but the EBSD indexation
rate would be lower and greater internal surface rumpling
(due to the higher strain heterogeneities) would cause more
problems. The use of a split sample under the present conditions enables one to follow the grains on the internal surface. It is shown that EBSD analyses can be carried out
directly on such a surface without signicant measurement
errors. Moreover, the internal surface does not appear to
signicantly aect the grain rotations. As a result, the
grains are considered to be true internal grains of a polycrystal. The main dierences with the experiments of Panchanadeeswaran et al. [10] carried out in 1996 are that (i)
the sample did not experience any accidental macroscopic
shear strain; (ii) the deformation was imposed in several
passes up to a strain of 1.2; and (iii) the microtexture was
analysed much more accurately, with typically 3000 measurements per grain for each deformation. This enables
one to obtain representative average grain rotations as well
as in-grain orientation spreads, and to study their evolutions in a quantitative way during the deformation.
5.2. Average rotations
The average grain rotations were studied in terms of
rotation angles and axes. They can be calculated from

1640

R. Quey et al. / Acta Materialia 58 (2010) 16291642

(a)

(c)

180

pair of grains
average

160
140

140

[degrees]

r [%]

100
80

V = 53

40
20
0

120
100

V = 83

80
60
40
20

VCO = 25
0

pair of grains
average

160

120

60

180

12

16

20

VCO = 37
0

pair of grains
average
2000 grains

160
140

r [%]

120

16

20

140

80
60

VCO = 9

40

pair of grains
average
2000 grains

160

100

120
100
80
60

VCO = 8

40
20

20
0

12

(d) 180

180

[degrees]

(b)

Disorientation [degrees]

Disorientation [degrees]

12

16

20

Disorientation [degrees]

12

16

20

Disorientation [degrees]

0:42
Fig. 13. Variability at constant orientation (VCO) of the grain rotations obtained during the rst increment r0:42
0 ; h0 . (a and b) Rotation angle in the
experiments and for the Taylor model, respectively. (c and d) Rotation axis in the experiments and for the Taylor model, respectively. V stands for the
overall variability, which is calculated with no limitation of disorientation (x-axis). For (b) and (d), the VCO does not fall to zero due to the limited
number of grains in use, but does with a more representative set of 2000 grains.

the initial orientations or for the successive increments,


which leads to the incremental rotations. The latter
method appears to be more interesting because it emphasizes the evolution of the rotation paths. The incremental
rotation angles tend to decrease as deformation increases.
It is shown for the rst time that the incremental rotation
axes are initially preferably distributed about TD, then
tend to rearrange along a direction situated between RD
and ND. These properties are related to the development
of the standard b-bre texture. The decrease in rotation
angles is due to the obvious fact that, as grain orientations
approach stable orientations, their rotation rates decrease.
Concerning the rotation axes, their initial distribution
about TD can be explained by the fact that it is the direction of preferential convergence to the b-bre, which is a
direction a  RD b  ND (a 0:97; b 0:24 between
Copper and S, and a 0:25; b 0:96 between S and
Brass). Then, when the orientations approach the b-bre,
they tend to rotate along the bre to reach a particular
component and the rotation axes approach a direction
0:9  RD 0:4  ND. An important experimental validation is that this evolution is opposite to the one that would
have been caused by a major friction eect: the slight sample barreling that can be seen in Fig. 3 would cause rota-

tions about TD that should increase with the


deformation (and not decrease). One can therefore conclude that the eventual friction-induced rotations can only
be of second order with respect to the rotations associated
with plane-strain compression.
These properties, and particularly those of the rotation
axes, as features clearly linked to the texture development,
can be used for a comparison between experimental and
simulated rotations at the level of the individual grains.
This can be done not only in a qualitative way, by comparing the location of the preferential rotation axes, but also in
a quantitative way, through the intensities of their distribution functions.
5.3. In-grain orientation spreads
The in-grain orientation spreads strongly develop at the
beginning of the deformation: while they are 0 at e 0,
they reach on average 5.1 at e 0:42, then tend to stabilize, with values of 6.4 at e 0:77 and 7.0 at e 1:2. This
conrms the results by Glez and Driver [19] obtained on
single crystals of Al1 wt.%Mn with stable orientations
(S, Brass and Copper) deformed under the same conditions. The orientation spreads develop through the forma-

R. Quey et al. / Acta Materialia 58 (2010) 16291642

(a) 140

pair of grains
average

pair of grains
average

120
100
80

60
56
52
48
44
40
36
32
28
24
20
16
12
8
4
0

r [%]

Final disorientation [degrees]

(a)

60

V = 34

40
20

VCO = 12
0

12

16

20

VCO = 31
0

Disorientation [degrees]

12

16

20

(b) 140

pair of grains
average
2000 grains

pair of grains
average

120
100
80

60
56
52
48
44
40
36
32
28
24
20
16
12
8
4
0

Disorientation [degrees]

r [%]

Final disorientation [degrees]

(b)

1641

60
40

VCO = 2.4

V = 34

20
0

12

16

20

Disorientation [degrees]
Fig. 14. Variability at constant orientation (VCO) of the nal orientation
e 1:2: (a) experimental results and (b) Taylor model.

tion of a dislocation sub-boundary structure which has


been examined by some EBSD maps over small areas using
steps of 0:5 lm. This work is not reported here, but one can
note that the sub-structures are very similar to those
reported by Glez and Driver [16] and more recently by
Humphreys and Bate [20].
5.4. Rotation variability at constant orientation (VCO)
In a polycrystal, two grains of the same orientation, but
possessing dierent neighbours, can rotate dierently;
herein termed rotation VCO. This variability is obviously
expected to be smaller than the one between grains of different orientations, denoted the overall variability. The
RVCO is dened as the ratio between the two. This parameter gives an idea of the relative inuence of the local grain
environment on its lattice rotation. The RVCO would be 0
if the rotations were only dependent on the orientation,
and 1 if there was no such a dependency. A method was
proposed in order to determine the VCO and RVCO. It
is based on a comparison of the rotations of grains of different orientations, and the resulting values are the averages of the variabilities over all orientations.
First, this method has been applied to the average grain
rotations for the rst strain increment, in terms of rotation
angle and axis. The rotation angle/axis VCO pair is 25%/

VCO = 24
0

12

16

20

Disorientation [degrees]
Fig. 15. Variability at constant orientation (VCO) of the orientation
spread hd at e 1:2. V stands for the overall variability, which is
calculated with no limitation of disorientation (x-axis). (a) Initial
orientation and (b) nal orientation.

37. The rotation angle/axis RVCO pair is 0.50/0.45.


Therefore, rotation angle and axis show nearly the same
relative variabilities. These dierences in rotation can lead
to grains of the same initial orientation developing dierent
nal orientations. This can be accounted for by the nalorientation VCO, which, in terms of disorientation angle,
is 12. It should be noted that this value is close to the distance between the b-bre components (typically 1520). In
other words, grain interactions could lead grains of the
same initial orientation having dierent nal texture components. Consequently, from a qualitative point of view,
models for which grain rotations depend only on grain orientation cannot exactly predict grain rotations, and they
may even not provide the right nal texture component.
A quantitative comparison is presented separately [13].
The in-grain orientation spreads were found not to
depend greatly on orientation (high RVCO). This is also
important information for texture simulations since the
macrotexture would be unchanged by permuting the
spreads of the dierent grains or, equivalently, applying
the same average spread to all of them. For these singlephase face-centred cubic metals, it is not necessary to know
exactly what the spread of a particular grain is. As a result,
the in-grain orientation spread during high-temperature
deformation does not appear to be an important parameter

1642

R. Quey et al. / Acta Materialia 58 (2010) 16291642

to evaluate, and hence models based only on the evolution


of the average orientation, e.g. the Taylor model, do not
suer from this approximation. This is relevant information since at the moment only high-resolution nite-element simulations can provide in-grain orientation spreads.

The microtexture data are available from http://tel.archives-ouvertes.fr/tel-00414120/en/ (le microtexturetracking-data.tgz).

6. Conclusions

Acknowledgments

The rst statistically sound set of experimental rotations


of grains in a face-centred cubic polycrystal at large strain
is provided. A split sample was used, and the orientations
were followed on its internal surface by EBSD throughout
the deformation; we call this method microtexture tracking. 176 grains were analysed at successive strains of 0,
0.42, 0.77, and 1.2. Typically 3000 orientation measurements per grain were obtained, giving access not only to
the average rotations, but also to the in-grain orientation
spreads. The results can be used to evaluate models for
polycrystalline deformation. The main results are that:

One of the authors (R.Q.) thanks Severine Girard-Insardi


for assistance with the PSC tests and Paul Jourey for advice
and assistance with the SEM-EBSD work. The authors
gratefully acknowledge the provision of the alloy by JeanMarie Feppon of Alcan Research Centre at Voreppe.

1. Following grains by EBSD on the internal surface of a


split sample does not aect signicantly the orientation
measurements, or the rotations of the individual grains.
2. 90% of the grains exhibit a unimodal rotation, composed of an average rotation and an orientation spread.
3. The average rotations are studied in terms of angle and
axis. The evolution of the incremental rotations is of
major interest. The average rotation angles appear to
decrease as deformation increases: they are 10, 7,
and 5 at the three successive increments of about 0.4.
The rotation axes are initially preferentially distributed
about TD, then about 0:9  ND 0:4  RD. These
properties can be related to the convergence of the orientations into the b-bre.
4. The in-grain orientation spreads develop strongly at the
beginning of the deformation, and then stabilize: their
average values are 5.1, 6.4, and 7.0, respectively, for
the three strains.
5. A method is proposed to quantify the rotation VCO
caused by grain interactions. It was found that, for the
rst strain increment, two grains of the same orientation,
but dierent neighbours, have rotation angles that dier
by 25% and rotation axes that dier by 37. At e 1:2,
their nal orientations dier by 12. Such variabilities
cannot be accounted for by the standard Taylor model.

Supplementary data

References
[1] Sarma GB, Dawson PR. Int J Plasticity 1996;12:1023.
[2] Crumbach M, Pomana G, Wagner P, Gottstein G. In: Gottstein G,
Molodov DA, editors. Proceedings of the 1st joint international
conference on recrystallization and grain growth, vol. 2. SpringerVerlag; 2001. p. 1053.
[3] Van Houtte P, Delannay L, Kalidindi SR. Int J Plasticity
2002;18:359.
[4] Quey R, Ringeval S, Piot D, Driver J. Mater Sci Forum
2007;539:3371.
[5] Skalli A, Fortunier R, Fillit R, Driver JH. Acta Metall 1985;33:997.
[6] Fortunier R, Driver JH. Acta Metall 1987;35:1355.
[7] Kalidindi S, Bhattacharyya A, Doherty R. Proc Roy Soc A
2004;460:1935.
[8] Poulsen HF, Margulies L, Schmidt S, Winther G. Acta Mater
2003;51:3821.
[9] Barrett CS, Levenson LH. AIME 1939;137:112.
[10] Panchanadeeswaran S, Doherty RD, Becker R. Acta Mater
1996;44:1233.
[11] Winther G, Margulies L, Schmidt S, Poulsen HF. Acta Mater
2004;52:2863.
[12] Quey R, Piot D, Driver J. Ceram Trans 2008;210:205.
[13] Quey R, Piot D, Driver JH. Acta Mater, accepted for publication.
[14] Maurice C, Piot D, Klocker H, Driver J. Metall Mater Trans A
2005;36:1039.
[15] Orilib. <http://orilib.sourceforge.net>.
[16] Glez J-C, Driver JH. J Appl Crystallogr 2001;34:280.
[17] Maurice C, Driver JH. Acta Mater 1997;45:4627.
[18] Bate PS, Knutsen RD, Brough I, Humphreys FJ. J Microsc
2005;220:36.
[19] Glez J-C, Driver JH. Acta Mater 2003;51:2989.
[20] Humphreys FJ, Bate PS. Acta Mater 2007;55:5630.

You might also like