Materials Transactions, Vol. 45, No. 11 (2004) pp.

3114 to 3119
#2004 The Japan Institute of Metals

Grain Size Measurements in Mg-Al High Pressure Die Castings

Using Electron Back-Scattered Diraction (EBSD)
Amanda Bowles1;2; *1 , Kazuhiro Nogita2 , Matthew Dargusch1;2 ,
Cameron Davidson1;3 and John Griths1;3; *2
1

Cooperative Research Centre for Cast Metals Manufacturing (CAST), UDP 055, University of Queensland, QLD 4072, Australia
Division of Materials, School of Engineering, University of Queensland, Brisbane, QLD 4072, Australia
3
CSIRO Manufacturing and Infrastructure Technology, PO Box 883, Kenmore, QLD 4069, Australia
2

Optical metallographic techniques for grain-size measurement give unreliable results for high pressure diecast Mg-Al alloys and electron
back-scattered diraction mapping (EBSD) provides a good tool for improving the quality of these measurements. An application of EBSD
mapping to this question is described, and data for some castings are presented. Ion-beam milling was needed to prepare suitable samples, and
this technique is detailed. As is well-known for high pressure die castings, the grain size distribution comprises at least two populations. The
mean grain size of the ne-grained population was similar in both AZ91 and AM60 and in two casting thicknesses (2 mm and 5 mm) and,
contrary to previously published reports, it did not vary with depth below the surface.
(Received May 18, 2004; Accepted September 13, 2004)
Keywords: magnesium-aluminium alloys, electron backscattered diraction (EBSD), grain size

1.

Introduction

Interest in the grain size of magnesium high pressure

die castings is two-fold. First, magnesium benets from
potent grain size strengthening, deriving greater strength
increases from a reduction in grain size than is the case for
aluminium alloys. The Hall-Petch parameter, ky , depends on
alloy content and texture and varies from 0.2 to 1.25
MPa m
1=2 (see Ref. 1 for a review), in magnesium alloys
compared to 0:05 MPa m
1=2 in aluminium alloys.2) However, unlike many castings and wrought alloys the grain size
distribution in thin-walled high pressure die castings does not
consist of one population. Large dendritic grains form in the
shot sleeve and they are then carried through the gate (where
some are fragmented) into the die cavity while a second
population of small equiaxed grains is formed in the die
cavity and these comprise the majority of the structure.37)
The eect of this mixed distribution on the Hall-Petch
relationship is crucial to an understanding of the yield
strength and it is, therefore, important to measure the grain
size distribution. Second, many authors have noted that
magnesium high pressure die castings have a skin, and
some have claimed that it is characterised by an especially
small grain size.8) The skin is believed to increase the
strength and ductility of castings;810) measurements of the
grain size variation with depth below the surface are required
to test this claim and to assess the importance of such a negrained region on tensile properties.
Unfortunately, there are a number of diculties associated
with the measurement of the grain size in high pressure die
cast Mg-Al alloys. Principally, no etchant is available that
reliably reveals the grain boundaries in as-cast Mg-Al alloys
(this, for reasons that are not clear, is not true for solutiontreated alloys). An alternative to etching is available in high
aluminium alloys such as AZ91. In these alloys the as-cast
*1Present

address: Oak Ridge National Laboratory, Metals and Ceramics

Division, TN 37931, USA
*2Corresponding author, E-mail: john.griths@csiro.au

microstructure consists of primary
grains and an (
)

eutectic which forms an almost-continuous network around
the grain boundaries and which is readily observed. However,
boundaries that lack the  (Mg17 Al12 ) phase will be missed
and the grain size will then be over-estimated. It can also be
dicult to dene the large dendritic grains formed outside
the die cavity since the eutectic forms between the dendrite
arms and the size of these grains can then be under-estimated.
This method is, of course, not easily applied to alloys such as
AM60 because the lower volume fraction of the eutectic
phases means that the cell boundaries are not completely
decorated. Another method, applicable to both AM60 and
AZ91, is to solution heat treat the castings, dissolving the 
phase and enabling etching to reveal the grain boundaries, but
careful selection of a heat treatment time is required to avoid
grain growth.
Electron back-scattered diraction (EBSD) mapping can
overcome the limitations of the conventional methods
described above. EBSD patterns are Kikuchi patterns and
these are determined by the orientations of the grains being
examined. The orientation at a particular point where the
electron beam is positioned can be determined by commercially available software and, by scanning the beam in a grid
pattern, a map of the grain orientations is produced. When a
discontinuous change in the orientation occurs, this signies
that a grain boundary has been crossed. In this way, the entire
grain structure of a selected region can be mapped.1113) The
technique is being increasingly used for grain size determination and, while it unlikely to supplant established optical
techniques due to the time-consuming nature of specimen
preparation and the limited size of the area analysed, it can be
invaluable in materials where standard optical techniques
prove dicult to apply or where details of misorientation are
required. It is dicult to apply optical analysis to Mg(Al)
alloys in the as-cast state and previous attempts to use EBSD
for these alloys have not been successful (but see Ref. 7) The
reasons for this are unclear but appear to revolve around
methods for surface preparation. Specically, preparing the
surface by mechanical polishing (e.g., with diamond polish-

Grain Size Measurements in Mg-Al High Pressure Die Castings Using Electron Back-Scattered Diraction (EBSD)

3115

Table 1 The grain size (equivalent circle diameter) of the ne-grained population as measured by EBSD at a number of depths below the
casting surface in AZ91 (2 mm thick), AZ91 (5 mm thick) and AM60 (5 mm thick). Maximum grain sizes are from a separate population.
Alloy

0.2% proof stress and

fracture strain

AZ91
(2 mm)

161 MPa
2.2%

AZ91
(5 mm)

142 MPa
5.9%

AM60

121 MPa

(5 mm)

11%

Depth below surface,

mm

Mean grain size,

mm

Maximum size,
mm

No. of grains
measured

40

5.7

20

127

70

5.1

23

157

500

7.7

29

131

1000

5.7

43

144

150

6.1

33

153

180

6.0

34

165

500

6.0

32

177

1000

6.7

34

181

2500

6.5

52

430

100

5.1

27

747

550

5.0

43

688

2000

7.8

68

311

ing agents) or by electropolishing does not produce a surface

nish good enough to allow EBSD mapping.
In this paper we describe the successful production of
EBSD maps using ion-beam milling to prepare the surface.
Data for three dierent sets of high pressure die castings are
presented.
2.

Experimental Procedure

Two magnesium alloys were used: AZ91D (9.0Al/0.8Zn/

0.2Mn) and AM60B (5.9Al/0.01Zn/0.3Mn). High pressure
die castings in the form of tensile specimens were produced
on a 250 tonne Toshiba cold chamber die casting machine
(see Ref. 14 for details of the casting conditions). The
specimens had a rectangular cross-section with a width of
10 mm, and the thickness, t, was either 5 mm or 2 mm for the
AZ91D specimens and was 5 mm for the AM60B specimens.
The gauge length was 50 mm.
The tensile properties were published in Refs. 5 and 14,
and are summarised in Table 1.
Sample preparation for EBSD involved specimens being
mechanically polished to a 1 mm nish and then cold-stage
argon-ion-beam milled. A gun tilt of 13 was used at 3.5 kV
and 0.8 mA and the milling time was 1 hour. After milling,
the specimens were stored under vacuum before being
transferred to the SEM. In contrast to experience with, for
example, aluminium alloys it was dicult to guarantee
satisfactory specimens in that the EBSD patterns were
sometimes unacceptably noisy. In such cases the specimens
had to be re-milled, and small adjustments were usually made
to the ion-beam current. A Philips XL30 scanning electron
microscope equipped with an Oxford Instruments EBSD
analyser was used. The system was operated at 20 kV with
the sample tilted to 70 . The smallest detectable grain was
1 mm, so that all grain size calculations/distributions were
made with a minimum equivalent circle diameter of 1 mm.
Orientation maps were produced to reveal the structure of the
material in which misorientations of greater than 15 were
recorded as being distinct grains. The choice of 15 is far
greater than the instrument resolution which is 3 but we

make two points. First, using a gure of 3 would demand

very high quality surface preparation and, as noted above,
this was dicult to achieve. Second, we estimate that the
fraction of grains missed at 15 is negligible (1%) and,
indeed, preliminary runs using 5 , 10 and 15 produced
similar numbers. A qualitative examination was made of the
inuence of misorientation angle on the delineation of grain
boundaries in two samples. These were compared to grain
boundaries determined from the same elds by a manual
estimation from orientation images. This showed that using
5 discrimination and, to a lesser extent, 10 , gave a false
identication of tiny grains. On the other hand, 5 discrimination would occasionally split large grains at low-angle
grain boundaries, while 10 and 15 would sometimes result
in counting of apparently dierent adjacent grains as a single
larger one. These would bias the distributions somewhat,
although the numbers involved were few enough to result in
only minor dierences in the mean grain size. It was
concluded that 15 was the optimum discrimination, giving
the most self-consistent measure of grain size.
A nal source of error, that is not accounted for in this
study, is that a few large grains in the centre of castings
exceeded the size of an individual eld of view for EBSD.
Such heterogeneity would require far more elds in order to
characterise it fully, as well as a modied acquisition or
analysis technique. The elds of view for this work were
deliberately chosen to avoid the large grains.
3.

Results and Discussion

The microstructures of high pressure die castings of AZ91

and AM60 have been described by numerous authors.8,15,16)
Figure 1(a) is a typical micrograph and shows the presence of
the large
grains that formed outside the die cavity and that
are segregated towards the centre of the casting. Figure 1(b)
shows the primary
(Mg rich) grains with a divorced eutectic
of  (Mg17 Al12 ) and
(Al rich). The dark-etched regions
around the  phase seen in Fig. 1(b) are commonly referred
to as eutectic
, (in this case labelled
0 ) and are enriched in
aluminium.

3116

A. Bowles, K. Nogita, M. Dargusch, C. Davidson and J. Griths

(a)

(b)

Fig. 1 Optical micrographs of AZ91. (a) the large white grains are the primary
phase and the surface of the casting is at the top of the
micrograph. Image (b) shows grain boundary  phase precipitates, the primary
phase and the aluminium rich
0 . Etching conditions:
1 s in Nital (2% concentrated nitric acid in ethanol).

Fig. 2 Images of AZ91 (2 mm thick). Top pair show a 15 EBSD grain orientation map and the associated secondary electron (SE) image
obtained during acquisition of EBSD data. Bottom pair show the optical and SEM secondary electron images of the same region after
removal of the carbon contamination followed by etching. All images are to the same scale shown by the 30 mm bar in the centre.

Figure 2 is a series of images from the EBSD analysis

together with an optical micrograph of the same area. The
area used for the ESBD mapping was visible under the
optical microscope because of carbon contamination on the
surface, and microhardness indents were made at the corners

of this area to act as markers for subsequent optical

metallography. The electron image shows damage produced
from the ion beam milling (long rows of dimples running
diagonally across the eld of view). The optical and
secondary electron images show approximately the same

Grain Size Measurements in Mg-Al High Pressure Die Castings Using Electron Back-Scattered Diraction (EBSD)

3117

Fig. 3 EBSD grain maps of AZ91 (2 mm thick) at depths of 40 mm, 70 mm, 500 mm and 1000 mm below the casting surface in a transverse
section. All maps are to the same scale, shown by the 20 mm bar in the centre.

region as that analysed (the EBSD map is produced when the

sample is tilted 70 so it is dicult to match the images
exactly even though a tilt correction was applied). The large
grain labelled A in the EBSD map is visible in the top left
corner of all of the images.
The optical and scanning electron images illustrate the
diculties in using the eutectic grain boundary decoration to
measure grain size. EBSD grain maps were obtained at a
number of depths below the casting surface. An example of a
series from AZ91 (2 mm thick) is shown in Fig. 3. At depths
less than 100 mm the microstructure consists primarily of
equiaxed ne
grains while at 1000 mm, several large

grains can be seen (see also Fig. 1).
The grain size data are presented as histograms in Fig. 4.
The eect of a few large grains on the distributions is seen in
the area plots, it being less obvious in the number plots.
There have been many studies of the grain size distribution
in annealed alloys1719) although not, it appears, of chill-cast
structures. Most authors take, as a starting point, the
lognormal distribution in which the data form a straight line
on a logarithmic probability plot. Figure 5 shows the results
of tests to investigate whether this is true of the data in Fig. 4.
It can be seen that the log(grain diameter) data t a normal
distribution only moderately well, with the 40 mm, 70 mm and
500 mm data showing a positive deviation from the straight
line at larger grain sizes (13 to 20 mm). All data show a slight

negative deviation at grain sizes below 3 to 4 mm, with the

eect being slightly more pronounced in the 1000 mm data.
Similar plots were made for the other two sets of castings and
these showed the same general features. There are several
possible explanations for such behaviour. First, and fundamentally, the size distribution in chill castings cannot be
exactly lognormal since the impingement of growing grains
during solidication sets an upper limit to their size, even
though the distribution may approximate lognormal at
intermediate sizes. Such an upper size limit will appear as
a positive deviation at that end of the distribution. However,
we do not at present have a large enough dataset to examine
whether this explanation is valid. A second possibility is
related to the previous one. Schuckher18) shows that such
deviations are characteristic of log-probability plots for linear
measurements of grain size even when the volume distribution is lognormal. However, we believe this explanation may
be invalid in the present case because the deviations in Fig. 5
appear to be too severe. A third explanation is that the
population is mixed (it is readily shown that adding data
points at one end, for example, of a lognormal distribution
has this eect) and the optical micrographs certainly suggest
this is the case. A useful rule-of-thumb20) is that if the area of
the largest grain diers from the mean by more than a factor
of 8 then the distribution is likely to be mixed, and this
criterion is easily met for the present data, see Table 1.

3118

A. Bowles, K. Nogita, M. Dargusch, C. Davidson and J. Griths

60

1800
1600

50

40 m

1400

40

Area, A / m

Counts

1000

30
20

600
200

10

20
30
40
Equiv Circle Dia, ECD/m

50

80

10

20
30
40
Equiv Circle Dia, ECD/m

50

1400

70

1200

70 m

60

70 m

1000

Counts

Area, A / m

50
40
30
20

800
600
400
200

10
0

800
400

10
0

40 m

1200

10

20
30
40
Equiv Circle Dia, ECD/m

50

40

10

20
30
40
Equiv Circle Dia, ECD/m

50

1600
1400

500 m

500 m

1200

Counts

Area, A / m

30

1000

20

10

800
600
400
200

10

20
30
40
Equiv Circle Dia, ECD/m

50

10

20
30
40
Equiv Circle Dia, ECD/m

50

70
1400

60

1000 m

50

Area, A / m

Counts

1000

40
30
20
10
0

1000 m

1200

800
600
400
200

10

20
30
40
Equiv Circle Dia, ECD/m

50

10

20
30
40
Equiv Circle Dia, ECD/m

50

Fig. 4 Histograms showing the number of grains (counts) and the grain
area as a function of equivalent circle diameter (ECD) from the EBSD
maps in Fig. 3, at 40 mm, 70 mm, 500 mm and 1000 mm below the casting
surface.

Gifkins21) has earlier proposed a similar criterion but with a

factor of 4. Finally, we have already commented that the size
resolution for the technique is 1 mm so that the data at that
end of the graph are less reliable than elsewhere.
Unfortunately, the sample size here is too small to allow
quantitative estimates of the means of possible constituent
populations and much further work would be needed to
improve the statistics. The populations cover a diameter
range in excess of two orders of magnitude, and analysing
sucient elds of view would require an unjustiably long
amount of machine time. As a rst approximation to
determining the relative amounts of the two populations,
we note that the fraction solidied prior to injection through
the gate is reported to be in the range of 1020%4,6) and our
own earlier analysis of these present die castings found that
the volume fraction of large
grains (dened as any with an
area greater than 110 mm2 ) is 4%, 4% and 10% in the AZ91
(5 mm thick), AZ91 (2 mm thick) and AM60 (5 mm thick)
castings, respectively.5)
Despite these uncertainties, the means for the four datasets,
assuming straight lines in Fig. 5 in the interval 16% < p <
84% (i.e.  one standard deviation), are given in Table 1
together with data for the other two types of castings. The
maximum grain sizes in the eld of view are also given

although they are not part of the recorded means; these grains
are possibly part of the separate populations. The logarithms
of the standard deviations are given by the slope of the lines
in Fig. 5 and are, typically, 0:4. This is only slightly larger
than the value of 0:3 found for several annealed metals18)
and cannot of itself be used to assert that the population is
mixed although as pointed out in the previous paragraph, the
maximum grain sizes measured are a strong indication of
that.
Table 1 shows that the mean grain size of the ne-grained
population measured by EBSD ranges from 5 to 8 mm and
does not show a systematic variation with the casting
thickness, alloy composition or, especially, depth below the
surface. This last nding conicts with three earlier reports of
a signicant variation in the grain size of the ne-grained
population with depth below the surface. Sequeira et al.8)
report that this grain size increased from approximately
0.54 mm at the surface to 10 mm in the centre of AZ91
castings (1, 2 and 6 mm thick). Similarly, Carlson22) found an
increase from 13 mm at the edge to 26 mm at the centre of an
AM60 casting. Finally, Rodrigo et al.23) found that the grain
size of samples from 2 mm thick AM60 castings was 7 mm at
the edge and 17 mm in the centre. In all these cases standard
optical metallography was used to measure the grain size and,
as pointed out in the Introduction, there are several
diculties with this technique. It is possible, therefore, that
the optical methods have unique observational errors that do
not apply to EBSD. Interestingly, Weiss and Davies24)
compared EBSD with a conventional metallographic method
for measuring the grain size of an aluminium alloy. They
found that the optical technique overestimated the grain size
by more than 40% and considered the EBSD data to be more
reliable. Clearly, further work is needed to make direct
comparisons of optical and EBSD methods for Mg-Al alloys
in the same way as done by Weiss and Davies. A nal
comment regarding our calculations of grain size (and those
of the other authors cited here) is that the numbers refer to the
apparent grain size as measured on a polished cross-section.
That is, no corrections have been made to transform the data
to the true three-dimensional grain size.
4.

Conclusions

The following conclusions can be drawn from this work:

(1) If appropriate surface preparation methods are used,
EBSD can be used to measure the true as-cast grain size
and grain size distribution in high pressure die cast
Mg-Al alloys.
(2) The grain size distribution in the castings examined is
mixed although the sample size is too small to make
good estimates of the means of the constituent
populations.
(3) The grain size of the ne-grained population in the three
groups of castings examined in this work does not
depend on alloy content, casting section size nor,
contrary to previously published data, with distance
below the cast surface.

Grain Size Measurements in Mg-Al High Pressure Die Castings Using Electron Back-Scattered Diraction (EBSD)

Cumulative frequency (%)

99.5
99
98

40
70
500
1000

95
90

3119

AZ91 (2mm thick casting)

80
70
60
50
40
30
20
10
5
2
1
1

10

Equivalent Circle Dia, ECD/m

Fig. 5 A logarithmic probability plot to test for lognormality of the small grain-size data from Fig. 4 (AZ91: 2 mm thick casting).

Acknowledgments
We acknowledge the nancial support of the Cooperative
Research Centre for Cast Metals Manufacturing, CAST,
which is established and part-funded by the Australian
Governments Cooperative Research Centres Scheme.
REFERENCES
1) C. H. Caceres, C. J. Davidson, J. R. Griths and C. L. Newton: Mater.
Sci. Eng. 325A (2002) 344355.
2) N. Hansen: Acta Metall. 25 (1977) 863869.
3) M. T. Murray and J. R. Griths: Metall. Mater. Trans. 27B (1996)
11511.
4) W. P. Sequeira: The microstructure and mechanical properties of high
pressure diecast magnesium alloy AZ91D; PhD thesis (2001) The
University of Queensland, Australia.
5) A. L. Bowles, J. R. Griths and C. J. Davidson: in Magnesium
Technology 2001, (The Minerals, Metals and Materials Society, 2001)
161168.
6) B. Niedermann and L. Iten: in Proceedings of 20th International Die
Casting Congress, World of Die Casting, (North American Die Casting
Association, 1999) 195203.
7) H. I. Lauki, O. Lohne, S. Sannes, H. Gjestland and L. Arnberg: Int. J.
Cast Metals Research 16 (2003) 515521.
8) W. P. Sequeira, G. L. Dunlop and M. T. Murray: in Proceedings of the
3rd International Magnesium Conference 1996, (Institute of Materials,
1997) 6373.
9) Y. B. Unigovski and E. M. Gutman: Appl. Surf. Sci. 153 (1999) 4752.

10) S. Sannes and H. Westengen: in Proceedings of International Conference Magnesium Alloys and Their Applications, (Wolfsburg,
Germany: Werksto-Informations GmbH, 1998) 223228.
11) A. J. Wilkinson and P. B. Hirsch: Micron 28 (1997) 279308.
12) F. J. Humphreys: J. Microscopy 195 (1999) 170185.
13) D. J. Dingley and D. P. Field: Mater. Sci. Technol. 13 (1997) 6978.
14) A. L. Bowles: The eect of moderate temperature ageing in high
pressure die cast Mg-Al alloys, PhD thesis (2003) The University of
Queensland, Australia.
15) D. J. Sakkinen: SAE Transactions: J. Mater. Manufacturing 103 (1994)
558569.
16) E. F. Emley: Principles of magnesium technology, (Pergamon Press,
1966).
17) H. E. Exner: Qualitative and quantitative surface microscopy, in
Physical metallurgy, R. W. Cahn, P. Haasen, editors (Elsevier Science,
1996).
18) F. Schuckher: Grain size, in Quantitative microscopy, R. T. DeHo,
F. N Rhines, editors (McGraw-Hill, 1968).
19) E. E. Underwood: Quantitative stereology, (Addison-Wesley 1970).
20) American Society for Testing and Materials: Test method E1181,
Methods for characterizing duplex grain sizes, (2002).
21) R. C. Gifkins: Optical microscopy of metals, (Pittman, 1970).
22) B. E. Carlson: Society of Automotive Engineers Transactions: Journal
of Materials and Manufacturing 104 (1995) 343351.
23) D. Rodrigo, M. Murray, M. Mao, J. Brevick, C. Mobley and R. Esdaile:
in Proceedings of 20th International Die Casting Congress, World of
Die Casting, (Cleveland, USA: North American Die Casting Association, 1999) 219225.
24) T. Weiss and P. Davies: Trans. Inst. Metal Finishing (UK) 78 (2000)
6566.