PHILOSOPHICAL MAGAZINE A, 1999, VOL. 79, NO.

7, 1671± 1695

Study of the mechanical properties of Mg± 7.7at.% Al by

in-situ neutron di€ raction

M. A. Gharghouri² , G. C.Weatherly, J. D. Embury, and J. Root³

Department of Materials Science and Engineering, McMaster University,
1280 Main Street West, Hamilton, Ontario, Canada L8S 4L7

[Received 14 May 1998 and accepted 17 September 1998]

Abstract
Internal stresses in tension and compression in an extruded and arti® cially
aged Mg± 7.7at.% Al alloy have been determined by in-situ neutron di€ raction.
Measurements were made on grains having the c axis normal to, parallel to and at
62ë from the stress axis, and on the reinforcing phase. The results are consistent
with basal slip and {1012} twinning. The second-phase particles of Mg17Al12 bear
much higher stresses than the magnesium matrix does. A simple mean stress
model proposed by Brown and Clarke is shown to describe adequately the
strengthening due to the particles. Twinning manifests itself clearly through
variations in the integrated intensity of di€ raction peaks during loading. Most
of the observed variations in scattered peak intensity can be explained by referring
to the lattice reorientation produced by {1012} twinning. The calculated stress
tensors corresponding to these intensity variations have been used to show that a
critical resolved shear stress criterion is applicable for {1012} twinning, and that
the onset of twinning does not correlate with the stress normal to the twin plane.
Intensity variations, which cannot be explained by {1012} twinning, occurred in
cyclic tension tests. It is suggested that this is due to {1011} twinning, which yields
a contribution to the grain strain tensor consistent with the direction of straining.
The reason why such twinning has not been observed in polycrystalline samples
by previous workers is that it appears to be essentially elastic, in the sense that it
disappears almost entirely upon removal of the applied stress.

§1. Introduction
Polycrystalline magnesium and magnesium alloys show limited ductility, which
varies considerably with both texture and stress state. The most general requirement
for achieving an arbitrary plastic strain in a particular grain in a polycrystalline
specimen is that ® ve independent deformation modes be activated. The primary
deformation mode in magnesium and its alloys is slip on the (0001) basal plane
with a h 1120i Burgers vector. Slip on {1010} prism and {1011} pyramidal planes
with a h 1120i Burgers vector has been observed experimentally, although the critical
resolved shear stress is much higher than for basal slip (see the review by Partridge
(1967)). The combination of these slip systems provides only four independent defor-
² Present address: Department of Mining and Metallurgical Engineering, Dalhousie
University, Sexton Campus, Halifax, Nova Scotia, Canada B3J 2X4.
³ Present Address: National Research Council Canada, Steacie Institute for Molecular
Sciences, Neutron Program for Materials Research, Chalk River Laboratories, Chalk River,
Ontario, Canada K0J 1J0.

Philosophical Magazine A ISSN 0141± 8610 print/ISSN 1460-6992 online Ñ 1999 Taylor & Francis Ltd
http://www.tandf.co.uk/JNLS/phm.htm
http://www.taylorandfrancis.com/JNLS/phm.htm
1672 M. A. Gharghouri et al.

mation mechanisms, none of which can contribute to plastic strain normal to the
basal plane. Strain parallel to the c direction can be accommodated elastically or by
twinning. In pure magnesium, only {1012} twinning is common and this can produce
elongation but not compression in the c direction, owing to the polar nature of
twinning.
As a result of the anisotropic plastic behaviour of magnesium crystals, polycrys-
talline samples are plastically inhomogeneous, that is some grains are soft, while
others are hard. This results in a non-uniform internal stress ® eld on a length scale of
the order of the grain size. In this paper, we show how in-situ neutron di€ raction
experiments have been used to characterize the internal stresses and to investigate the
twinning and strengthening mechanisms during loading in an extruded and arti® -
cially aged binary Mg± 7.7 at.% Al alloy. The present paper presents an analysis of
both the internal stresses and the twinning behaviour in the material.

1.1. Neutron di€ raction for internal stress determination

Changes in the lattice spacings of di€ erent crystallographic planes as a function
of applied load can be measured by neutron di€ raction. The method has been
described in detail by Allen et al. (1992). For a monochromatic beam of neutrons
(constant wavelength ¸), Bragg’s di€ raction law yields the following relationship
between the crystal plane spacing and the measured di€ raction angle through
which the strain ( e ) can be determined:
d ­ d0 sin µs
e = s ­ 1. ( 1)
d0 = sin µ0
In equation (1), (d0, 2µ0) and (ds , 2µs ) are the plane spacing and scattering angle of a
given di€ raction peak in the unloaded and loaded (applied stress s ) conditions
respectively. Thus a shift in the scattering angle for a given re¯ ection is associated
with a lattice strain e perpendicular to the re¯ ecting planes. The lattice strain is
measured parallel to the scattering vector which is the bisector of the incident and
di€ racted beams. When the scattering vector does not move with respect to the
specimen geometry, di€ erent hkil planes correspond to di€ erent grain orientations.
In order to characterize the stress state completely in hcp metals, it is in general
necessary to measure the change in spacing for six hkil planes for each grain orienta-
tion in a polycrystalline sample. The appropriately transformed single-crystal elastic
constants can then be used to determine the stress tensor for each grain orientation
studied. In a polycrystal, each di€ raction peak captures an in® nite number of grain
orientations obtained by rotating the crystal lattice about the normal to the di€ ract-
ing plane. As a result it is impossible to obtain unambiguously six lattice strain
measurements corresponding to a single grain orientation in a polycrystal, although
this is possible in a single crystal. For in-situ experiments, there is also the practical
di culty of orienting an entire load frame in the neutron beam for each measure-
ment.
The peak shift in di€ raction experiments measures only the elastic lattice strain,
which is generally proportional to the elastic component of the macroscopic strain
after the yield point has been reached.
In-situ neutron di€ raction experiments have been used mainly for metal-matrix
composites (Allen et al. 1992, Prangnell et al. 1995). The most extensive neutron
di€ raction work on a hcp material was done by MacEwen and co-workers
(MacEwen et al. 1983, 1988, 1989, 1990, MacEwen and Tome 1987) on Zircaloy
 Mechanical properties of Mg± 7.7at.% Al 1673

rod characterized by a high concentration of basal poles perpendicular to the axis of

the rod. They measured the evolution of residual thermal strains during cooling and
the residual stresses left in the material on unloading after di€ erent amounts of cold
working. Only residual strains in the unloaded condition were measured by
MacEwen and co-workers. The current work is, to our knowledge, the ® rst in-situ
neutron di€ raction study on a hcp metal.

§ 2. Experimental procedure
2.1. Preparation and characterization of the materials
A Mg± 7.7 at.% Al alloy was prepared at the PeÂchiney Research Centre in
Voreppe, France. Pure magnesium (99.9%) and aluminium were melted under a
protective atmosphere of argon and sulphur hexa¯ uoride at 650ë C. The alloy was
then cast into cylindrical ingots of 64 mm diameter which were extruded at 250ë C
and a pressure of 25MPa, yielding cylindrical samples 15mm in diameter. The
samples were then annealed for 24 h at 400ë C. This as-received material was
machined into test samples and sealed under argon in Pyrex ampoules. The samples
were then solution treated for 8 h at 415ë C and quenched into cold water, being
careful to break the ampoules before quenching to ensure proper contact between
the water and the samples. The samples were then resealed in Pyrex ampoules and
arti® cially aged by holding them for 10 min at 330ë C, allowing them to cool to 100ë C
at a rate of 0.86ë C h­ 1 and then air cooling to room temperature. The arti® cially
aged material will be referred to as `the alloy’ in the current paper.
The ® nal grain size was approximately 60 m m, and the ageing treatment gave a
volume fraction of 10 1% of b precipitates dispersed within the grains. The volume
fraction of the precipitates was determined by neutron di€ raction as described in
appendix A. The error quoted above is the standard deviation of data obtained from
a total of 12 measurements (three measurements each on four specimens). The pre-
cipitates were plate shaped, with an average thickness of 165 nm, a length that varied
from 6 to 20 m m and a length to width ratio of between 2 and 6. An optical micro-
graph of the alloy is shown in ® gure 1. The stereographic basal pole ® gure in ® gure 2
shows that most grains have the basal pole oriented at greater than about 60ë to the
extrusion axis.
The b -intermetallic compound Mg17Al12 has a bcc structure with 58 atoms per
unit cell. The habit plane is most frequently the matrix basal plane and the orienta-
tion relationship is that of Burgers, that is
( 011) b // ( 0001) Mg,
( 2)
[111]b // [2110]Mg.
To avoid ambiguity, four-index Miller± Bravais notation will be used when discuss-
ing planes in the magnesium matrix, and three-index notation will be used for planes
in the precipitate. A single crystal of the intermetallic compound was prepared and
its mechanical properties determined from ultrasonic measurements and Vickers
indentations (Gharghouri et al. 1999). It was found to have an average Young’s
modulus of about 80 GPa (compared with 45GPa for magnesium) and a yield stress
of the order of 1 GPa.
Pure magnesium rod was extruded as described above and annealed for 8 h at
415ë C in order to obtain the same grain size and texture as the alloy.
1674 M. A. Gharghouri et al.

Figure 1. Optical micrograph showing the precipitate distribution in the alloy ( 457).

Figure 2. Stereographic basal pole ® gure for the alloy. The centre of the projection is parallel
to the extrusion axis.
 Mechanical properties of Mg± 7.7at.% Al 1675

Transmission electron microscopy (TEM) observations were made on thin foils

of the alloy. The foil preparation has been described in detail elsewhere (Gharghouri
et al. 1998).
The stress± strain characteristics of the pure magnesium and the alloy were
obtained using an MTS-810 load frame equipped with a 100 kN load cell. The strain
was measured using an MTS extensometer with a gauge length of 13.5 mm.

2.2. In situ neutron di€ raction

For the in-situ neutron di€ raction tests, an Applied Test Systems (Butler,
Pennsylvania) screw-driven load frame equipped with a load cell and hydraulic
grips was mounted in a neutron beam on the sample table of the L3 di€ ractometer
at Chalk River Laboratories. Both axial (parallel to the sample axis) and radial
(normal to the sample axis) lattice strains were measured. The two types of measure-
ment required di€ erent load frame set-ups which have been described in detail else-
where (Gharghouri 1997). The sample was mounted in the load frame, which was
carefully aligned such that the centre of the gauge length lay above the centre of
rotation of the sample table. This alignment was necessary to ensure that the gauge
volume did not move out of the incident beam when the sample table was rotated.
An MTS extensometer was used to measure the macroscopic strain at the same time
as the di€ raction data were collected. The extensometer was attached to the sample
with steel springs. Cadmium masks were used to cover the sections of the grips and
the extensometer body which were in the path of the neutron beam and would have
resulted in unwanted re¯ ections.
Consistent statistics for all measurements made on a given peak were achieved by
counting until a pre-set number of neutrons, referred to as the monitor, had emerged
from the beam tube.
For each test the load frame was programmed to step automatically through a
pre-set sequence of applied stresses. For each stress level, data were collected for
each peak of interest. For each peak, the detector moved to the appropriate Bragg
angle and the sample table rotated until the scattering vector was parallel to the
desired specimen direction. It was only necessary for the operator to verify that the
experiment was proceeding properly every few hours. When the applied stress fell
outside a pre-set error range during a measurement, the count stopped automatically
until the load was restored by a computer-controlled feedback loop.
A di€ raction peak was de® ned by a set of x± y coordinates, where x is the
detector position in degrees, and y is the number of neutrons which entered the
detector during the pre-set monitor. The data were ® tted to
2
1 2µ ­ 2µs
Y = b + m( 2µ ­ 2µ0) + A exp ­ , ( 3)
2 s

which represents a Gaussian function (amplitude A, mean 2µs and standard devia-
tion s) superimposed on a linear background (intercept b, and slope m), using a
Newton± Raphson least-squares ® tting algorithm. The di€ raction angle was then
equal to µs and the integrated intensity I was determined using

I = As ( 2p ) 1/ 2 ( 4)
1676 M. A. Gharghouri et al.

The rms error c given by

N 1/ 2
c = ( Y icalc ­ Y iexp 2
) ( N ­ 5) ( 5)
i=1
was used to calculate the uncertainties in the ® tted parameters from which the errors
in 2µs and I were determined.
A typical error of ( 1- 2) 10­ 4 was traced to the overall goodness of ® t of the
raw data by the model function (equation (3)). The error in the integrated intensity
was generally of the order of 3± 4%, although it could be as high as about 7% for the
precipitate peak which was weak compared with the matrix peaks.

Figure 3. Orientation of the hexagonal prism in grains A, B and C, and of the precipitate
unit cell with respect to the stress axis (vertical).
 Mechanical properties of Mg± 7.7at.% Al 1677

Axial lattice strains (parallel to the extrusion direction) were measured for the
(0002), {1010} and {1011} planes. The corresponding grains will henceforth be
referred to as the A, B and C grains respectively. The c axis is parallel to the
stress axis in the A grains and normal to it in the B grains. The C grains
represent an intermediate orientation with the c axis approximately 62ë from
the stress axis. The orientations of the hexagonal prism with respect to the stress
axis are shown for the A, B and C grains in ® gures 3 (a), (b) and (c) respectively
(stress axis vertical).
Radial strains (perpendicular to the stress axis) were measured for the (0002),
{1010}, {1011} and {1012} planes. Only the {411} lattice strains were measured in the
precipitate (axial and radial) as these were the only planes to give rise to a strong
peak which was within the range of the detector for the wavelength used. The
orientation of the cubic lattice with respect to the stress axis is shown for the pre-
cipitate in ® gure 3 (d).
In what follows, axial and radial lattice strains are denoted by the superscripts A
and R respectively. Thus the axial and radial lattice strains normal to the {1010}
planes are denoted by e A R
f1010g and e f1010g respectively.

§ 3. Experimental results
Uniaxial tension and compression stress± strain curves for pure magnesium and
the alloy are shown in ® gure 4, in which the compression curves have been rotated
into the ® rst quadrant. The ds / de against e curves obtained from the tensile tests are
plotted in ® gure 5 for the pure magnesium and the alloy together with the di€ erence
between the curves. This graph illustrates the reinforcement provided by the second-
phase particles in the alloy.
The results from in-situ neutron di€ raction experiments are presented in ® gures 6
(tension) and 7 (compression). The axial strains are presented in ® gures 6(a) and
7 (a) while the radial strains are shown in ® gures 6 (b) and 7 (b). Two graphs arranged
horizontally are shown for each crystallographic plane for which measurements were

Figure 4. Stress± strain curves for monotonic tension and compression for the alloy and pure
magnesium.
1678 M. A. Gharghouri et al.

Figure 5. ds / de against e curves for the alloy and the magnesium deformed in tension. The
di€ erence between the two curves is also plotted.

made. The planes corresponding to given rows are identi® ed in the left-hand graphs.
The applied stress is plotted against lattice strain in the left-hand graphs and against
integrated intensity in the right-hand graphs. The graphs are aligned vertically for
easy comparison between planes. Data points connected by a solid line correspond
to measurements made while the absolute value of the applied load was being
increased (loading) while a broken line is used to connect points for measurements
made during unloading.
Neutron di€ raction results for the (1010) peak (B grains) from a cyclic tensile
test are shown in ® gure 8. The maximum applied stress was increased from one
cycle to the next. Graphs of applied stress against axial lattice strain for each
cycle are arranged in a row in ® gure 8 (a). As in ® gures 6 and 7 the loading
curves are solid lines while the unloading curves are broken lines. The graphs all
share the same y axis which is therefore only labelled for the leftmost graph. The
intensity variations of the peak during the test are shown in ® gure 8(b). Rather
than superposing loading (solid) and unloading (broken) curves, the sequence of
applied stresses is plotted along the x axis in the sequence in which they were
applied during the test. The graphs in ® gure 8 (b) are drawn such that the portion
pertaining to a given cycle aligns vertically with the corresponding applied stress±
lattice strain graph in ® gure 8(a). The results in ® gure 8 are important for the
analysis of twinning in § 4.4.
TEM revealed very straight c -type dislocations normal to g = ( 0002) in the
undeformed material (® gure 9). The density and distribution of these dislocations
did not change with deformation. Dislocations with a -type Burgers vectors were
commonly found in foils obtained from deformed samples which had lines par-
allel to the trace of {1010}, {1011} and {1012} planes (® gure 10). A much higher
density of a -type dislocations was observed in the basal plane. Only {1012} twins
were found in the thin foils examined by TEM. The interaction of these twins
with the precipitates has been discussed in detail elsewhere (Gharghouri et al.
1998).
 Mechanical properties of Mg± 7.7at.% Al 1679

Figure 6. Neutron di€ raction results for monotonic tension: (a) axial strains; (b) radial
strains.

§ 4. Discussion
The general shape of the stress± strain curves will ® rst be discussed with reference
to the known deformation mechanisms in magnesium. The internal stresses in the
three families of grains will then be derived from the measured lattice strains and
discussed in terms of the deformation modes. The hardening mechanism related to
the second-phase particles will be considered, followed by a ® nal section with an
analysis of twinning.

4.1. Stress± strain behaviour

Basal slip in most grains is not favoured in either tension or compression; the
orientation (Schmid) factor for basal slip is very low in most grains since the stress
axis is within or very close to the basal plane. However, the critical resolved shear
stress (CRSS) for basal slip is very low (approximately 0.51 MPa in pure magne-
sium), and it should thus be easy to activate even in unfavourably oriented grains.
1680 M. A. Gharghouri et al.

Figure 7. Neutron di€ raction results for monotonic compression: (a) axial strains; (b) radial
strains.

The texture is such that {1012} twinning is favored in compression for a majority
of the grains (B grains). Relatively few grains are favourably oriented for {1012}
twinning in tension (A grains).
In tension, most of the deformation is produced by basal slip, with some twin-
ning and non-basal slip at regions of stress concentration and in grains oriented with
the c axis close to the stress axis (A grains). The stress± strain curves show essentially
no linear elastic regime in either material since basal slip is so easy to activate. Basal
slip occurs ® rst in those grains most favourably oriented with respect to the stress
axis. The number of grains which undergo basal slip at a given applied stress level
depends on the texture of the material. Eventually, most of the grains undergo basal
slip, at which point the rate of strain hardening is very low. There is a region of very
high apparent hardening rate prior to general yielding in the alloy stress± strain curve
which does not occur for the pure magnesium. The second-phase particles must be
the origin of the observed reinforcement, which produces a much higher general yield
stress in the alloy than in the pure magnesium. However, the work-hardening rates in
the fully plastic regime are essentially the same in the two materials (® gure 4).
 Mechanical properties of Mg± 7.7at.% Al 1681

Figure 8. Neutron di€ raction results for the (1010) peak for cyclic tension: (a) applied stress±
lattice strain graphs; (b) intensity as a function of applied stress.

General yielding occurs at a much lower stress in compression than in tension for
both materials. Twinning has been observed to propagate through the gauge section
by a LuÈders process (i.e. propagation of a plastic front) in pure polycrystalline
magnesium having a similar texture (Reed-Hill 1973). It is likely that the same
process occurs in the materials studied here.

4.2. Internal stresses determined by neutron di€ raction

Lattice strains and the derived internal stresses are determined on the assumption
that there are no internal stresses present in the material prior to deformation.
Measurements of the lattice parameters of pure magnesium after heat treatment
show that they are identical with those of pure single-crystal magnesium. Thus we
can assume that there are no appreciable grain-to-grain interaction stresses. This is
1682 M. A. Gharghouri et al.

Figure 9. Bright-® eld TEM image for undeformed material showing straight c -component
dislocations, g = (0002).

Figure 10. Bright-® eld TEM image of a sample deformed in tension using a [2110] zone axis
(several re¯ ections operating).

to be expected because magnesium is essentially isotropic with respect to thermal

expansion.
The situation is more complicated in the aged alloy. As for the pure magnesium,
it is unlikely that there will be thermal residual grain interaction stresses.
Furthermore, the thermal expansion coe cients of magnesium and of the b inter-
metallic compound are very close (a Mg = 27 10­ 6ë C­ 1; a ppt = 28.4 10­ 6ë C­ 1) .
In the absence of plastic relaxation, the maximum stress on cooling from 330ë C
would not exceed about 21 MPa ( = ( 28.4- 27) 10­ 6 300 50 10­ 3) in the
matrix. This stress is likely relaxed by the straight c -type dislocations parallel to
the long direction of the precipitates observed by TEM (® gure 9) and/or by slip
with a h 1120i Burgers vector on the basal plane. {1011} pyramidal slip with a
 Mechanical properties of Mg± 7.7at.% Al 1683

h 1120i Burgers vector occurs easily at temperatures above approximately 225ë C and
must also contribute to the relaxation of stresses due to the thermal mismatch
between phases. It is thus not unreasonable to assume that there are no residual
stresses present in the aged alloy prior to deformation.
In order to determine the stress tensor in a grain, it is necessary to evaluate the
following equation

[s ] = [C][e ] ( 6)
where matrix notation has been used and [C] is the elastic sti€ ness matrix, trans-
formed appropriately according to the grain orientation. In this study, six measure-
ments of lattice strain could not be made for each set of grains in ® gure 3 from which
to calculate [s ] rigorously. In ® gure 3, the crystallographic direction parallel to the
stress axis (the x3 axis) and two orthogonal directions normal to the stress axis (the
x1 and x2 axes) are shown for the A, B and C grains. As shown in table 1, e 1, e 2 and
e 3 have been determined or can be estimated for the three grain orientations. Note
that we require e R R
f2110g and e f022 7. 04g (parallel to [0112]) which have not been mea-
sured. The only radial strains measured in a direction within the basal plane is
e R
f1010g . The grains giving rise to this signal include A grains, grains rotated 30
ë
about [0001] in ® gure 3 (b), grains rotated approximately 30ë about [0112] in ® gure
3 (c), as well as a multitude of other grain orientations. It is thus clear that an error
will be introduced in the calculation. However, the elastic properties of materials
with the hexagonal structure are isotropic in the basal plane, as can be shown by
applying an arbitrary rotation about the c axis to the tensor of elastic constants. In
addition, the radial strains measured are very small. We therefore assume that the
measured value of e R f1010g is a reasonable approximation to the radial strains required
for the stress calculations, as shown in table 1.
Finally, since the {0 2 27.04} and {0112} planes are only about 15ë apart, the
strains normal to the two sets of planes (and therefore in the corresponding two sets
of grains) are likely to be quite close in magnitude and thus we substitute the radial
strains normal to the {0112} planes for the radial strains normal to the {0 2 2 7.04}
planes in equation (6).
The matrix equations for the stress tensors for the three sets of grains are shown
in equation (7). It can be seen from the sti€ ness matrices that s 1, s 2 and s 3 do not
depend at all on e 4, e 5 and e 6 in the A and B grains and are only very weakly
dependent on them for the C grains. It should thus be possible to determine these
stresses quite accurately from the measured values of e 1, e 2 and e 3.
Similarly, s 4, s 5 and s 6 do not depend at all on e 1, e 2 and e 3 for the A and B
grains and are only very weakly dependent on them for the C grains. Moreover, since
the applied stress is uniaxial it is reasonable to expect that e 4, e 5 and e 6 will be small

Table 1. Lattice strains corresponding to the x1, x2 and x3 axes in ® gure 3.

e 1 e 2 e 3
R
A grains e f2110g ( e Rf1010g) e R
f1010g e A
( 0002)
R
B grains e f2110g ( e Rf1010g) e R
( 0002) e A
f1010g
R
C grains e f2110g ( e R )
1010g e R
f0227.04g ( e R
f0112g ) e A
f0111g
1684 M. A. Gharghouri et al.

in magnitude. These two observations suggest that s 4, s 5 and s 6 are small compared
with s 1, s 2 and s 3 and they shall be considered to be zero:
For A grains,

R
s 1 59.75 23. 24 21. 7 0 0 0 e 1 =e ( 1010)
R
s 2 23.24 59. 74 21. 7 0 0 0 e 2 =e ( 1010)
s 3 21.7 21. 7 61. 7 0 0 0 A
=
e 3 = e (0002) ; ( 7 a)
s 4 0 0 0 16. 39 0 0 e 4=0
s 5 0 0 0 0 16.39 0 e 5=0
s 6 0 0 0 0 0 18. 25 e 6=0

For B grains,

R
s 1 59.74 21.27 23.24 0 0 0 e 1 =e ( 1010)
s 2 21.7 61.7 21. 7 0 0 0 e 2 =e
R
( 0002)
s 3 23.24 21.7 59.74 0 0 0 e A
= e (1010)
= 3 ; ( 7 b)
s 4 0 0 0 16.39 0 0 e 4 =0
s 5 0 0 0 0 18.25 0 e 5 =0
s 6 0 0 0 0 0 16.39 e 6 =0

For C grains,

R
s 1 59.12 22.7 23.86 0 1.94 0 e 1 =e ( 0112)
s 2 22. 70 59.74 25.23 0 ­ 1.85 0 e 2 =e
R
( 1010)
s 23.86 25.23 59.02 0 ­ 0.94 0 R
3
=
e 3 = e (0111) ;
s 4 0 0 0 16.67 0 ­ 0.1235 e 4=0
s 5 1.94 ­ 1. 85 ­ 0.94 0 18.54 0 e 5=0
s 6 0 0 0 ­ 0. 1235 0 16. 47 e 6=0
( 7 c)

The precipitate stress tensor is somewhat more di cult to determine as the lattice
strain in only one crystallographic direction was measured parallel and normal to the
stress axis. However, axes can be set up as shown in ® gure 3 (d) and, since the angle
between [141] and [140] is only about 19ë , we can reasonably assume that
e R R
( 140) e ( 141) . In order to perform the calculation, it is also necessary to assume
that the strain tensor is symmetric with respect to the x3 axis and therefore that
e R R
( 4117) e ( 141) . We then obtain the expression
 Mechanical properties of Mg± 7.7at.% Al 1685
R
s 1 85. 21 32.69 34.20 0.39 ­ 3.09 ­ 0.75 e 1 =e ( 411)
R
s 2 32. 69 85.29 34.12 ­ 3.14 0.17 0.04 e 2 =e ( 411)
s 3 34.20 34.12 83.78 2. 75 2.92 0.71 e =e
A
( 411)
= 3 ; ( 8)
s 4 0. 39 ­ 3. 14 2.75 21. 02 0.71 0.17 0
s 5 ­ 3.09 0.17 2. 92 0. 71 21.10 0.39 0
s 6 ­ 0.75 0.04 0. 71 0. 17 0.39 19.59 0

where once again the sti€ ness constants have been transformed from the standard
orientation to that shown in ® gure 3 (d).
Based on equations (7) and (8) the stress tensors for the three types of grain and
for the precipitates have been calculated and are presented in graphical form in
® gures 11 and 12. For each test, s 1, s 2 and s 3 are plotted against applied stress.
Rather than superimposing the loading and unloading curves, the sequence of stres-
ses is plotted along the x axis in the sequence in which they were applied during the
tests. The applied stress is plotted as a broken line for easy comparison with the
stresses that were determined from di€ raction measurements.

Figure 11. Stresses in the A, B and C grains and precipitate for the tension test. The broken
lines have a slope of unity and are equal to the applied stress.
1686 M. A. Gharghouri et al.

Figure 12. Stresses in the A, B and C grains and precipitate for the compression test. The
broken lines have a slope of unity and are equal to the applied stress.

A simple way to verify that the calculated internal stresses are reasonable is to
check whether the following force balance is satis® ed
s applied = s A VA + s B VB + s C VC + s ppt V ppt , ( 9)
where s A, s B, s C and s ppt are the stresses (parallel to the stress axis) in the three
families of grains and in the precipitate respectively, and V A, V B, V C and V ppt are the
corresponding volume fractions. The volume fraction of the second phase is
10 1% . The volume fractions of the three families of grains were determined
from neutron di€ raction experiments (see appendix B) and are given in table 2.
The average internal stress is plotted against the applied stress for a tensile test in
® gure 13. The broken lines have a slope of one and correspond to a state of balanced
forces. The force balance is clearly satis® ed for this tensile test, which gives us some
con® dence that the assumptions to determine the internal stresses in the grains are
reasonable.

Table 2. Volume fractions of di€ erent components of the

microstructure
Component Volume fraction
A grains 0.22
B grains 0.62
C grains 0.062
Precipitates 0.1
 Mechanical properties of Mg± 7.7at.% Al 1687

Figure 13. Graph showing that force balance (equation (9)) is veri® ed.

It is now possible to analyse the calculated stresses with reference to the known
deformation mechanisms in magnesium. The potential contribution of each slip and
twinning system to the strain tensor in the A, B and C grains can be determined using
(Groves and Kelly 1963)
g
e ij = 2 ( ni bj + nj bi ) , ( 10)
where n is the unit vector normal to the slip or shear plane, b is the unit vector
parallel to the shear direction, and g is the magnitude of the shear. n and b must be
expressed in a Cartesian coordinate system related to the specimen axes. For twin-
ning, the value of g is ® xed; it is uniquely de® ned for a given twinning mode.
In the C grains, basal, prism and pyramidal slip can produce appreciable strain
parallel to the stress axis for a given shear. Two of the {1012} twinning systems can
produce compressive strain (3± 4%), the others contributing at most only 0.34%
tensile strain. The favourable orientation of these grains for basal slip suggests
that they should be soft in both compression and tension. This is indeed observed;
® gures 11 (c) and 12 (c) show clearly that the stress parallel to the stress axis in the C
grains is generally lower than the applied stress.
In the B grains, only prism and pyramidal slip can provide deformation parallel
to the stress axis. All the {1012} twinning systems can give compression parallel to
the stress axis and extension normal to it. These grains are thus expected to be hard
in tension since the CRSS for prism and pyramidal slip is relatively high, and softer
in compression when {1012} twinning can contribute to deformation. This is clearly
observed in ® gures 11(b) and 12 (b).
None of the slip systems observed by TEM can produce strain parallel to the
stress axis in A grains. {1012} twinning produces c -axis extension and can thus
contribute to deformation in tension but not in compression. Thus, in the absence
of other deformation mechanisms, the A grains should be soft in tension but rela-
tively hard in compression. This is indeed what is observed in the neutron di€ raction
tests. For the tensile test (® gure 11(a)), the stress in the A grains follows the applied
stress up to about 125 MPa, at which point a plateau is reached. The stress in the A
1688 M. A. Gharghouri et al.

grains is lower than the applied stress for the remainder of the test. The stress in the
A grains follows the applied stress during loading in compression, and they are left
with a small compressive stress at the end of the test (® gure 12 (a)).
It is clear from the above discussion that the stresses determined from the
neutron di€ raction experiments for monotonic tensile and compressive tests are
consistent with the previous literature concerning the deformation mechanisms in
magnesium.
The stress in the precipitate phase is always considerably higher than the applied
stress, indicating that some load is transferred from the matrix to the particles
(® gures 11 (d) and 12 (d)). The strengthening due to the particles is considered in
the following section.

4.3. Strengthening from second-phase particles

The alloy is signi® cantly stronger than the polycrystalline magnesium in both
tension and compression. The stress at general yielding (i.e. 0.2% plastic strain) is
higher in both cases. In addition, the slope of the stress strain curve after yielding is
higher for the alloy in compression, although this is not the case in tension. Various
strengthening mechanisms must be considered in order to explain these di€ erences
between the two materials.
Solid solution strengthening can contribute to the yield strength of the alloy by
raising the critical resolved shear stress for slip. Given the limited solid solubility of
aluminium in magnesium at room temperature (about 0.5 at.% ) and the small dif-
ference in atomic diameter between aluminium and magnesium (about 3%), it is
unlikely that this mechanism contributes signi® cantly to the strength of the alloy.
Indeed, experimental tensile stress± strain curves obtained for the solution-treated
alloy (7.7 at.% Al in magnesium) were essentially identical with those of pure mag-
nesium.
Orowan hardening can also contribute to the alloy yield stress. In this case,
hardening arises because a dislocation must curve around the precipitates before it
can bypass them. The closer the particles, the smaller is the radius of curvature of the
dislocation, and therefore the higher is the shear stress required. In the present case,
the distance to consider is that between particles in the basal plane since basal slip is
by far the most important slip mode. This distance is of the order of 1 m m for this
system. The maximum possible increase in the yield stress determined from
¹b
¿( Orowan) = , ( 11)
l
where ¹ is the shear modulus, b the Burgers vector and l is the interparticle spacing
with ¹ = 16.4 GPa, b = 3.2 A and l = 1 m m, is 5MPa.
To determine whether the Orowan mechanism can account for the increase in
yield strength of the alloy with respect to pure magnesium, it is ® rst necessary to
determine how best to de® ne this value. The 0.2% o€ set stress which is usually used
does not seem suited to the present case. In both materials, this stress is much higher
than the stress at which substantial slip has already occurred, as demonstrated by the
departure from linear behaviour at very low applied stresses. Thus this stress is not a
good measure of the stress at which dislocations bypass the precipitates.
It would seem best to consider the stress at the smallest o€ set possible given the
sensitivity of the extensometer used. The minimum measurable strain is about
8 10­ 5 (i.e. 1 m m in 13.5 mm). An o€ set of 10­ 4 can easily be measured on the
 Mechanical properties of Mg± 7.7at.% Al 1689

stress± strain curve and yields stress values of 100 and 40 MPa for the alloy and the
magnesium respectively. Thus the Orowan mechanism can contribute at most about
12% of the observed increase in yield stress.
The above two mechanisms do not seem capable of accounting for the di€ erent
stress± strain behaviour of the two materials. It is thus necessary to consider con-
tinuum e€ ects due to constraints imposed by the particles. During deformation, the
elastic and plastic incompatibilities between the precipitates and the magnesium
matrix result in a mean stress in the matrix which opposes the applied stress but
aids reverse ¯ ow. Brown and Clarke (1960) calculated this mean back stress in a two-
phase material for a number of particle geometries using Eshelby’s equivalent inclu-
sion method. They showed that, provided that there is no plastic relaxation around
the particles, the contribution of the mean back stress to the slope of the stress± strain
curve is given by

ds
de = Eppt V ppt ( 12)
ppt

for ribbons, discs, and ® bres parallel to the stress axis. In this equation, Eppt and V ppt
are Young’s modulus and the volume fraction of the second phase respectively.
In the present case, the precipitates can be considered as discs which have been
elongated in one direction. Moreover, since most grains have the basal plane at a
small angle from the stress axis, and the habit plane of the precipitates is the basal
plane, they can be considered parallel to the stress axis. Thus equation (12) can be
used to determine the contribution of the mean stress to the slope of the stress± strain
curve. With Eppt = 80 GPa, equation (12) yields a back stress contribution of 7.2±
8.8 GPa for volume fractions ranging from 9% to 11%. This range is in very good
agreement with the constant di€ erence in the slopes of the alloy and magnesium
stress± strain curves up to a true strain of about 40 10­ 4 (® gure 5).
Equation (12) implies that the particles support the applied strain elastically. The
precipitate lattice strain for the second phase at a given applied stress (® gure 6) does
compare favourably with the corresponding plastic strain measured from the alloy
stress± strain curve in ® gure 4.
The slope of the alloy stress± strain curve in compression beyond about 0.2%
strain is higher than that for the pure magnesium. In both cases (up to the strains
tested in the present study), deformation probably proceeds by the propagation of
{1012} twinning through the gauge section (Reed-Hill 1973). In pure magnesium,
ds / de is essentially zero, indicating that both intragranular propagation and inter-
granular propagation are very easy. In the alloy, intragranular propagation is more
di cult because twins must negotiate the precipitates. These mechanisms require
that additional energy be expended compared with the pure magnesium, resulting
in an increase in the shear stress for propagation. The large precipitates which occur
at grain boundaries limit the size of twins which can be transmitted from one grain to
the next and may also make it more di cult for an existing twin in one grain to
nucleate a twin in a neighbouring grain.
In tension, ds / de for the alloy falls rapidly to the same level as for the pure
magnesium beyond about 0.6% strain. The particles have thus ceased to contribute
signi® cantly to hardening, indicating that relaxation processes are active. Secondary
slip processes induced by stress concentrations near particle± matrix boundaries can
1690 M. A. Gharghouri et al.

limit the extent of dislocation pile-ups at precipitates. Slip with a -type Burgers
vectors have been observed on non-basal planes in deformed samples (® gure 10).

4.4. Analysis of twinning

It is generally accepted that slip is activated when a CRSS is reached on the slip
plane in the slip direction, and that the CRSS is essentially independent of the other
stress components and of the hydrostatic pressure. Although the resolved shear
stress in the twinning plane in the shear direction appears to be an important factor
in the activation of twinning, experimental evidence seems to indicate that the stress
normal to the twin plane can also play an important role. Thus Cahn (1954) con-
cluded from compression experiments on rutile, dypside and baride that twinning is
favoured by the superposition of hydrostatic stress. On the basis of experiments on
rolled coarse-grained silicon steel samples, Priestner and Louat (1963) concluded
that the resolved shear stress required for twinning is dependent on the crystal
orientation with respect to the stress axis. MacEwen et al. (1988) evaluated the
internal stresses during mechanical loading of Zircaloy bar. They observed a marked
relaxation of the stress normal to the basal planes at a stage of deformation which
corresponds to mechanical twinning, from which they deduced that twinning may be
either enhanced or inhibited by superposing a normal stress to the shear component.
Lebensohn and Tome (1993) modelled the twin lamella as a ¯ at inclusion of
elliptic section embedded in an elastically anisotropic medium acted on by an exter-
nally applied stress in order to derive the stress state associated with the activation of
twinning. They found that a CRSS criterion can be used for twinning, provided that
the normal stress components are not much larger than the CRSS. The in¯ uence of
normal stresses is zero in an elastically isotropic medium and becomes more impor-
tant as the anisotropy increases. However, these workers indicated that their model
can only take elastic e€ ects into account. Twin growth requires the motion of twin
dislocations, which shift atoms within the twinned region to the appropriate lattice
positions in the twinned crystal. For single lattice structures it can be shown (Jaswon
and Dove 1956, 1957, 1960) that a simple shear carries all atoms into their correct
® nal positions. In double lattice structures such as the hcp structure, some atomic
shuƒ ing in addition to the simple shear is necessary except for certain special modes.
For {1012} twinning, atomic shuƒ es normal to the twin plane are required, and the
facility with which this shuƒ ing occurs may be a€ ected by the stress normal to the
twinning plane (Hosford, 1993).
Twinning results in an abrupt reorientation of the crystal lattice. It is thus quite
easy to detect by neutron di€ raction since, provided that it occurs in a large enough
volume of the sampled material, it results in a change in peak intensity. The stress
state in the grains when such an intensity change occurs can then be used to deter-
mine an activation criterion for twinning. When a grain twins, the intensity of the
corresponding peak decreases since part of the grain has been rotated such that the
conditions for Bragg re¯ ection are no longer respected. When an increase in intensity
occurs, some other component of the texture has twinned and in so doing has rotated
the lattice so as to increase the volume of material in which the crystallographic
plane giving rise to the peak is oriented appropriately for Bragg re¯ ection. In this
case, the stress tensor for the grain under study does not represent the state of stress
at the onset of twinning.
To determine whether a CRSS criterion is suitable for twinning, it is necessary to
determine the resolved shear stress at the moment of twinning (decrease in peak
 Mechanical properties of Mg± 7.7at.% Al 1691

intensity) for at least two di€ erent loading conditions in which the internal stress
tensors are appreciably di€ erent. {1012) twinning results in a reorientation of 86ë of
the crystal lattice and thus converts a B grain into an orientation very close to that of
an A grain (about 4ë o€ ) and vice versa. In tension, the A grains are expected to
twin, and we observe a decrease in intensity of the (0002) peak (® gure 6 (a)). This is
accompanied by a strain plateau for the (0002) lattice strain during which the strain
does not increase with increasing load. There is also, as expected, a slight increase in
the intensity of the {1010} peak (® gure 6 (a)) which is demonstrated by the fact that
the unloading (broken) curve in the intensity graph (® gure 6 (a)) for this peak is to
the right (higher intensity level) of the loading (solid) curve. The reverse intensity
variations are observed, as expected, in compression (® gure 7 (a)). Any increase in
(0002) axial peak intensity in compression is due to {1012} twinning in the B grains.
Since the intensity of the (0002) axial peak is much smaller than that of the {1010}
axial peak because of the texture, it is likely to be a much more sensitive indicator of
{1012} twinning in B grains.
The resolved shear stress and the normal stress for each {1012} twin system in a
given family of grains can be obtained for each value of the applied stress by suitable
transformation of the internal stress tensors derived in § 4.2. The resolved shear
stress on the most highly stressed system just before and just after a signi® cant change
in intensity then provides a range for the CRSS for twinning. The results of such an
analysis are presented in table 3. Results are presented for monotonic tension and
compression, tension-® rst Bauschinger and cyclic tension. For each case, the applied
stress, the resolved shear stress, and the normal stress before and after a signi® cant
change in peak intensity are shown.
Table 3 shows that twinning occurs when the resolved shear stress is between 65
and 75 MPa. This is not a signi® cant variation given the error in the stress determi-
nations (approximately 10MPa) and suggests that a CRSS criterion is reasonable
for {1012} twinning. This is all the more convincing as the CRSS does not seem to
depend at all on the sign of the normal stress, which varies from ­ 70 to +70 MPa.
It is also possible to estimate the shear stress when the grains `untwin’ upon
changing the loading direction. The principle is the same as for twinning, except
that the changes in intensity are reversed. The results of such an analysis are shown
in table 4.
The scatter in the shear stress for untwinning is much greater than for twinning
and it is not likely that a well de® ned CRSS exists for this situation. However, the
results do show clearly that untwinning occurs before the sign of the resolved shear
stress changes during unloading.

Table 3. Stress states in grains at the applied stress at which an appreciable change I in
intensity of the (0002) axial peak is observed (twinning).
Applied Shear Normal Twinned
Applied stress state I stress stress stress grains
Tension, no pre-strain # 125 ! 150 73 ! 73 61 ! 42 A
Compression, no pre-strain " ­ 115 ! ­ 130 62 ! 64 ­ 46 ! ­ 58 B
Compression after tension " ­ 115 ! ­ 130 64 ! 66 ­ 71 ! ­ 66 B
Third cycle of cyclic tension # 175 ! 200 63 ! 65 69 ! 65 A
Fourth cycle of cyclic tension # 200 ! 230 61 ! 74 62 ! 81 A
1692 M. A. Gharghouri et al.
Table 4. Stress states in grains at the applied stress at which an appreciable change I in
intensity of the (0002) axial peak is observed (untwinning).
Applied stress state I Applied stress Shear stress Twinned grains
Unloading from compression # ­ 50 ! ­ 25 28 ! 12 B
(compression-® rst Bauschinger)
Unloading from ® rst cycle of # ­ 50 ! ­ 25 16 ! 0 B
cyclic compression
Unloading from second cycle # ­ 100 ! ­ 75 30 ! 21 B
of cyclic compression
Unloading from third cycle of " 150 ! 100 47 ! 24 A
cyclic tension
Unloading from fourth cycle " 125 ! 75 28 ! 9 A
of cyclic tension

Twins ® rst nucleate and then grow. Smaller grains inhibit twinning by limiting
the size of the twin, thereby increasing the surface to volume ratio and consequently
the shear stress for nucleation (Hosford 1993). The precipitates in the alloy are not
sheared by twins, so their presence has the same e€ ect on the CRSS for twinning as a
decrease in grain size.
Most of the intensity variations observed in the neutron di€ raction experiments
can be easily rationalized in terms of {1012} twinning which appears to be the only
bulk deformation mechanism capable of producing plastic strain parallel to the c
axis. However, the intensity variations obtained for the {1010} axial peak (i.e. B
grains) in cyclic tension cannot be explained in this way (® gure 8(b)). In the second,
third and fourth cycles, the intensity of this peak decreases during loading and
increases again during unloading, which means that these grains are twinning during
loading and untwinning during unloading. If these grains were to twin on {1012}
during loading, they would contract parallel to the tensile axis. Thus another
mechanism must exist to account for these observations.
The most likely mechanism is {1011} twinning, which has been observed in pure
magnesium samples deformed under highly constrained conditions (Wonsiewicz and
Backofen 1967). TEM analysis in the present study did not reveal the presence of
such twins. This may be because the twinned volume in the unloaded condition is
extremely small, making it di cult to observe. That this is so can be con® rmed by
comparing the overall increase in the B grain intensity with the overall decrease in
the A grain intensity at the end of the test. All {1012} twinning in A grains (or rather
in grains very close to the A orientation) must cause an increase in the B grain
intensity. Equation (B1) can be used to correlate the changes in intensity of
the two orientations as shown in the table 5. Since the magnitudes of Thkil

Table 5. Analysis showing that the change in intensity of the A and B grains can be correlated
for cyclic tension.
Peak Change in intensity ( Ihkil ) Thkil / Ihkil / Phkil
{1010} + 15 + 673
(0002) ­ 18 ­ 684
 Mechanical properties of Mg± 7.7at.% Al 1693

(proportional to the volume of twinned material) are essentially the same for the two
orientations, it is clear that {1012} twinning can account for essentially all the
increase in {1010} axial peak intensity at the end of the test, and therefore that
the amount of {1011} twinning left in the unloaded condition is very small. This
analysis is consistent with the lack of twinning observed in B grains by optical
microscopy.
If1010g increases immediately upon unloading from the maximum load for the
third and fourth cycles (® gure 8 (b)). {1011} twinning therefore appears to behave
more or less elastically, in the sense that it is reversed immediately upon reducing the
applied load. Another observation which points to elastic behaviour is the remark-
able linearity of the applied stress± lattice strain curve for the {1010} peak. This is in
contrast with the (0002) axial curve in tension which is strongly a€ ected by the
occurrence of {1012} twinning (® gure 6 (a)).

§ 5. Conclusions
In-situ neutron di€ raction experiments have been used to determine the stresses
for three grain orientations and the reinforcing phase in a precipitation-strengthened
Mg± 7.7 at.% Al alloy. The weighted average of the internal stresses compares very
favourably with the applied stress for tension and compression tests. The stresses
calculated from the neutron di€ raction measurements are consistent with the known
slip and twinnning mechanisms in magnesium. A simple Brown± Clarke mean-stress
model has been shown to account for the strengthening due to particles.
Twinning has been shown to result in strain plateaux and di€ racted peak inten-
sity variations during loading. Most intensity variations were explained by referring
to the lattice reorientation produced by {1012} twinning. The calculated stress ten-
sors corresponding to these intensity variations have been used to show that a CRSS
criterion is applicable for {1012} twinning in the Mg± 7.7 at.% Al alloy, and that the
onset of twinning does not correlate with the stress normal to the twin plane.
Intensity variations which cannot be explained by {1012} twinning occurred in
cyclic tension, the only in-situ test in which the material was strained well beyond the
elastic± plastic transition in tension. It is suggested that this is due to {1011} twinning,
which yields a contribution to the grain strain tensor consistent with the direction of
straining. The reason that such twinning has not been observed in polycrystalline
samples by previous workers is that it appears to be essentially elastic, in the sense
that it disappears almost entirely upon removal of the applied stress.

ACKNOWLEDGEMENTS
The authors are grateful to PeÂchiney for supplying the materials used in this
work, to Atomic Energy of Canada for use of the neutron di€ raction facilities, to the
National Sciences and Engineering Research Council of Canada for their ® nancial
support of this work, and to Dr Z. S. Basinski (McMaster University) and Dr C.
Tome (Los Alamos National Laboratory) for valuable discussions in the course of
this work. This project was initiated while M. G. was a visiting graduate student at
Grenoble and the support and guidance of Dr Y. BreÂchet in the ® rst stages of the
project are much appreciated.
1694 M. A. Gharghouri et al.

A PPE N D I X A

Precipitate volume fractiondeterminations fromneutron diffraction

The precipitate volume fraction in the aged alloy was determined based on the
following equation (Bacon 1975)
I / VPT, ( A 1)
where I is the integrated intensity of the di€ raction peak, V is the volume of material
giving rise to the peak and P is the product of the structure factor and the Lorentz
factor for the di€ raction peak. T is a parameter that is determined from texture
measurements, it is a correction of the structure factor, which is computed assuming
a completely random texture. In the present case, since the texture is essentially
symmetric about the specimen axis, T was determined from tilt scans as follows;
I I sin a
T=
h Ii
where h Ii =
sin a
, ( A 2)
where s denotes the angle between the extrusion axis and the normal to the hkil
planes. Measurements were made for a from 0 to 90ë in 5ë intervals. If the intensity
Imat of a matrix peak and that of a precipitate peak Ippt are measured, then volume
fraction of the prcipitate is given by equation (A 3)
Ippt
Pppt Tppt
Vf =
Ippt Imat
( A 3)
+
Pppt Tppt Pmat Tmat

A PPE N D I X B

Volume fractiondeterminations of eachgrain orientation fromneutron

diffraction
The volume fraction of each family of grains was determined, based on the
following equation (Bacon 1975):
Ihkil / VmatPhkil Thkil ( B1)
where Ihkil , Phkil and Thkil are as de® ned in Appendix A. V mat is the volume of the
magnesium matrix. As V mat is the same for all of the hkil planes, it can be omitted
from equation (B1). The volume fraction of each family of grains in the microstruc-
ture can then be determined using equation (B2) where V matrix is the volume fraction
of matrix material (90% in the present case)
T
Vhkil = hkil V matrix. ( B2)
Thkil

References
Allen, A. J., Bourke, M. A. M., Dawes, S., Hutchings, M. T., and Withers, P. J., 1992,
Acta metall. mater., 40, 2361.
Bacon, G. E., 1975, Neutron Di€ raction (London: Oxford University Press).
 Mechanical properties of Mg± 7.7at.% Al 1695
Brown, L. M., and Clarke, D. R., 1975, Acta metall., 23, 821.
Cahn, R. W., 1954, Adv. Phys., 3, 363.
Gharghouri, M. A., 1997, PhD Thesis, McMaster University, Ontario.
Gharghouri, M. A., Embury, J. D., Weatherly, G. C., and Garrett, J., 1999 (sub-
mitted).
Gharghouri, M. A., Weatherly, G. C., Embury, J. D., 1998, Phil. Mag. A, 78, 1137.
Groves, G. W., and Kelly, A., 1963, Phil. Mag., 3, 877.
Hosford, W. F., 1993, The Mechanics of Crystals and Textured Polycrystals, ® rst edition
(New York: Oxford University Press).
Jaswon, M. A., and Dove, D. B., 1956, Acta crystallogr., 9, 621; 1957, Acta cryst., 10, 14;
1960, ibid. 13, 232.
Lebensohn, R. A., and Tomeí, C. N., 1993, Phil. Mag. A, 67, 187.
MacEwen, S. R., Christodoulou, N., and Salinas-Rodriguez , A., 1990, Metall. Trans.,
A, 21, 1083.
MacEwen, S. R., Christodoulou, N., Tomeí, C. N., Jackman, J., Holden, T. M., Faber,
J., Jr., and Hitterman, R. L., 1988, Proceedings of the Eighth International
Conference on the Textures of Materials, edited by J. H. Kallend and G. Gottstein
(Warrendale, Pennsylvania: The Metallurgical Society), p. 825.
MacEwen, S. R., Faber, J. Jr., Turner, A. P. L., 1983, Acta metall., 31, 657.
MacEwen, S. R., and Tomeí, C. N., 1987, Zirconium in the Nuclear Industry: Seventh Inter-
national Symposium, ASTM Special Technical Publication (edited by R. B. Adamson,
and L. F. P. Van Swam) (Philadelphia, Pennsylvania: American Society for testing and
Materials), p. 49.
MacEwen, S. R., Tomeí, C. N., and Faber, J. Jr., 1989, Acta metall., 37, 979.
Partridge, P. G., 1967, Metals Rev., 12, 169.
Prangnell, P. B., Downes, T., Withers, P. J., and Lorentzen, T., 1995, Mater. Sci.
Engng, 197 , 215.
Priestner, R., and Louat, N., 1963, Acta metall., 11, 195.
Reed-Hill, R. E., 1973, The Inhomogeneity of Plastic Deformation, (Metals Park, Ohio:
American Society for Metals), p. 285.
Wonsiewicz, B. C., and Backofen, W. A., 1967, Trans. AIME, 239 , 1422.