You are on page 1of 6

Characterisation of ne-scale microstructures by electron

backscatter diraction (EBSD)


F.J. Humphreys
*
Manchester Materials Science Centre, Grosvenor Street, Manchester M1 7HS, UK
Accepted 6 May 2004
Available online 7 June 2004
Abstract
Recent developments in instrumentation and software now enable grain structures >0.1 lm to be quantitatively characterised by
EBSD in conjunction with a eld emission gun scanning electron microscope. The paper discusses the advantages and limitations of
the technique.
2004 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: EBSD; FEGSEM; Grain boundaries; ECAE; Cold working
1. Introduction
Electron backscattered diraction (EBSD) is based
on the acquisition of diraction patterns from bulk
samples in the scanning electron microscope. Such pat-
terns were rst obtained 50 years ago [1], but the use by
Dingley [2] of low light TV cameras for pattern acqui-
sition and on-line pattern solution, stimulated the
widespread interest in the technique which led to the
development of commercially available systems. Of
particular importance in the emergence of EBSD as a
metallographic technique was the development of rapid
automated pattern analysis, and this when used in
conjunction with control of the microscope beam en-
abled area scans of a sample surface to be obtained
rapidly and automatically. The term orientation
imaging microscopy or OIM has been used to describe
these area scans of a sample, although the terms ori-
entation map or EBSD map which are more com-
monly used, will be adopted in this article. A more
recent development has been the use of EBSD in con-
junction with Field Emission Gun Scanning Electron
Microscopes (FEGSEM) [3], and the consequent in-
crease in spatial resolution has further extended the
range of applications of EBSD. It is this last factor
which has made EBSD viable as a technique for cha-
racterising sub-micrometre grain structures, and only
high resolution EBSD carried out on such instruments
will be considered in this paper.
The EBSD acquisition hardware generally comprises
a sensitive CCD camera, and an image processing system
for pattern averaging and background subtraction. The
software controls the data acquisition by scanning the
beam over the sample in a raster, obtaining and solving
the diraction patterns at each point, and storing the
data. Further software is then required to analyse,
manipulate and display the data. EBSD is carried out on
a specimen which is tilted between 60 and 70 from the
horizontal. More detailed discussion of the technique
and its development can be found in the recent reviews
of Randle and Engler [4] and Humphreys [5].
2. The viability of EBSD for characterising ne-scale
microstructures
In this paper we consider the use of EBSD for cha-
racterising sub-micrometre grain structures, and in this
section we summarise the advantages and disadvantages
of EBSD relative to other metallographic methods. The
advantages of EBSD over optical microscopy for
quantitative metallography include improved spatial
resolution, more accurate data and more complete
microstructural characterisation. The advantages of
EBSD over the TEM include:
*
Tel.: +44-161-200-3554; fax: +44-161-200-8877.
E-mail address: john.humphreys@umist.ac.uk (F.J. Humphreys).
1359-6462/$ - see front matter 2004 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.scriptamat.2004.05.016
Scripta Materialia 51 (2004) 771776
www.actamat-journals.com
The use of bulk samples rather than thin foils, which
avoids the dicult and questionable practice of pre-
paring thin foils which are representative of the bulk
material.
Very large areas or regions selected from large sam-
ples are readily studied.
Data may be obtained from specic areas of interest,
e.g., near surfaces, welds etc.
Rapid and automated acquisition and analysis of the
diraction patterns, which is not routinely available
in the TEM, enables many thousands of grains/sub-
grains to be characterised from a single map.
Much less expensive equipment is required than for
TEM.
There are also a number of important microstructural
parameters which are now routinely available from
EBSD analysis but which are not obtainable from con-
ventional methods of grain characterisation. These in-
clude local relationships between microstructure and
grain orientations, variations of stored energy in the
microstructure and boundary character [5].
Disadvantages of EBSD relative to TEM might
include:
worse spatial resolution, as discussed later,
inability to image individual dislocations and defects,
longer time for data acquisition.
However, as the technique of EBSD matures, it is
being found capable of undertaking much of the
microstructural analysis at the subgrain/grain level,
which was previously carried out by TEM.
3. Current and future EBSD performance
The data required for quantitative microstructural
analysis using EBSD is usually obtained in the form of
an orientation map. A large amount of data needs to be
collected in the shortest possible time and the critical
parameters are the speed of data acquisition and the
spatial and angular resolutions. These depend on a
number of factors including the specimen, the equip-
ment and the method of operation. The results pre-
sented in this paper were obtained on three Field
Emission Gun SEMs (Philips XL30, CAMSCAN
Maxim 2040SF and FEI Sirion) in the Manchester
Materials Science Centre. All instruments were equipped
with CCD cameras capable of on-chip integration or
slow scan acquisition, and the CHANNEL EBSD sys-
tem (HKL Technology, Denmark). Subsequent data
analysis and presentation were carried out using VMAP,
an in-house software development [6] and ICE (HKL
Technology). It should be noted that the microscopes
and EBSD acquisition systems are all standard com-
mercially available equipment with no signicant mod-
ications.
Typical values of spatial and angular resolution for
EBSD in a number of metals are shown in Table 1.
There are a number of factors which must be taken into
account when deciding if EBSD can be successfully used
for a particular investigation.
3.1. The specimen
The backscattered electron signal increases with the
atomic number (z) of the material. The quality of the
diraction pattern increases with z and the spatial res-
olution may also improve with increasing z [3]. The re-
sults cited in this paper were mainly obtained from
aluminium alloys and somewhat better results may be
obtained from steels or nickel alloys. The diraction
pattern comes from the surface layer and therefore good
sample preparation is required. In order to obtain an
analysable diraction pattern the region of the specimen
from which the pattern is obtained must have a single
crystallographic orientation (see below). Thus the
smallest grain or subgrain size which can be measured is
related to the spatial resolution of the technique. Defects
such as dislocations may cause the pattern to lose
sharpness, but unless this is severe, the pattern will still
be analysable.
3.2. The speed of data acquisition
The time to acquire a data point during a scan de-
pends on the slowest of two operations:
(a) The time required to obtain an analysable dirac-
tion pattern. This depends primarily on the material
and microscope operating conditions and is typically
101000 ms.
(b) The time required to analyse the pattern. This de-
pends on the processing speed of the computer,
the speed of the pattern-solving algorithm and the
Table 1
Summary of typical EBSD performance for various metals in W-la-
ment and FEG microscopes
Sample and
microscope type
Eective spatial
resolution K
A
(nm)
Angular
resolution ()
Aluminium
W 60 1
FEG 20 1
Brass
W 25 1
FEG 9 1
a-Iron
W 30 1
FEG 10 1
772 F.J. Humphreys / Scripta Materialia 51 (2004) 771776
number of lines in the pattern required for a solu-
tion, and is typically 1020 ms.
3.3. The spatial resolution
If the area of the sample contributing to a diraction
pattern contains more than one crystallographic orien-
tation, e.g. a grain boundary region, a single crystal dif-
fraction pattern is not obtained, the automated pattern
solving routines may fail and the pattern will not be in-
dexed. This is clearly seen in the small EBSD map of Fig.
1a, where the grey levels represent the orientations of the
pixels and the individual grains are clearly revealed. Non-
indexed points, which are black, are seen to occur at some
points on the grain boundaries. These points may be re-
moved during post-acquisition analysis by replacement,
on a probability basis, with the orientations of adjacent
analysed pixels. Such pixel replacement introduces no
new orientations, although such data processing must be
used with care. Fig. 1b shows the eect of replacing the
non-indexed pixels in the map of Fig. 1a.
The area fromwhich an EBSDpattern is acquired with
an electron beam focused on a 70 tilted sample is
approximately elliptical, with the major axis, which is
perpendicular to the tilt axis, being some three times that
of the minor axis. It is a function of material, beam
accelerating voltage, specimen tilt and probe size, and the
resolution parallel to the tilt axis (K
A
) for FEGSEM is
typically 50 nm for aluminium. The resolution in the
orthogonal direction (K
P
) is about three times larger [3].
However, when analysing a sample with small grains or
subgrains the achievable or eective spatial resolution is
often smaller than this because, when patterns from two
grains overlap, the acquisition software can, if there is a
signicant dierence in intensity of the patterns, suc-
cessfully analyse the stronger pattern. Fig. 1c shows the
band contrast (a measure of the pattern quality), in the
Fig. 1. Part of an EBSD map of an aluminium alloy in which a ne grain size (0.4 lm) has been produced by ECAE and subsequent annealing: (a)
raw map data, showing non-indexed points (black) at the boundaries, (b) map with the non-indexed points replaced, and high angle (black) and low
angle boundaries (white) marked, (c) pattern quality (band contrast) map of the sample showing the lower pattern quality at the boundaries, (d)
misorientation distribution and (e) grain size distribution. (The data for (d) and (e) were obtained from the whole map which contained 3000
grains.)
F.J. Humphreys / Scripta Materialia 51 (2004) 771776 773
map of Fig. 1, and it is seen that this is lowest at the
boundaries, where the overlapping patterns are obtained.
Fig. 2 is a schematic diagram showing two grains in a
sample. The intensity of the diraction pattern from
grain 1 is shown by the line ABCD, and that from grain
2 as EFGH. The size of the area from which a diraction
pattern originates, depends on the spatial resolution (R)
which is indicated on the diagram. When the beam is
positioned within a distance R of the boundary (e.g. to
the right of point B), both grains are sampled, two
overlapping patterns are obtained. The software can
solve the stronger diraction pattern if the dierence in
intensity (D) between the two patterns is suciently
large (>D
c
). However, in reality, the patterns, which
arise from electron scattering events, are noisy as shown
in Fig. 2, and the Rose criterion, used for determining
the visibility of features in images, states that a signal is
readily detectable above its background noise (N) only if
it reaches a level of 5 N. Thus the value of D
c
required
for pattern indexing increases as the noise (N) increases,
and as seen from Fig. 2, the eective spatial resolution
(R
eff
) then increases. We therefore see that R
eff
is a
function not only of the resolution of the instrument,
but also of the pattern quality.
The eective resolution, which is the most important
parameter in determining whether or not a ne-grained
material can be analysed by EBSD, may be conveniently
determined by measuring the fraction of patterns (N
S
)
which are solved during a raster scan of the sample, and
for equiaxed grains of mean size D, K
A
is given
approximately [7] by
K
A

D1 N
S

4
1
There is a range of beam currents over which the
EBSD spatial resolution is optimised. If the current is
too small, the pattern quality is degraded, thereby
worsening the spatial resolution (R
eff
) as discussed
above. However, too large a beam current increases the
beam spread (R) and therefore degrades the resolution.
It is signicant that the resolution of the FEGSEM is
much less sensitive to the probe current than the W-l-
ament SEM [5]. This arises because the beam size in a
W-lament microscope is a much stronger function of
probe current than for a eld emission gun [8]. There are
a number of other factors which aect the EBSD per-
formance. As the length of time (or number of frames)
over which the pattern is averaged increases, the quality
of the pattern and thus the eective spatial resolution
improve [3], although the data acquisition time is
lengthened. The microscope accelerating voltage also
has some eect. The beam spread in the sample increases
with accelerating voltage and should therefore be kept
as low as possible. The latest FEGSEM instruments
which provide suciently large beam currents at lower
voltages (510 keV) to allow good diraction patterns to
be obtained, are enabling the EBSD resolution to be
further improved.
3.4. Angular resolution
The relative orientation between adjacent data points
is related to the precision with which the orientations of
data points within the same crystallite can be measured.
If diraction patterns are obtained from a small area of
a single crystal or a single grain within a large-grained
polycrystal, then although their analysis should result in
identical orientations, this is not usually the case, and a
range of measured orientations results. The resulting
orientation noise is typically 12, but depends
strongly on the pattern quality [5] and improves with
larger probe currents and longer pattern acquisition
times. Averaging of the orientation of data points (pix-
els) within a grain or subgrain is a simple and eective
method of improving the angular resolution [9]. Further
signicant improvement in angular resolution, to <0.1,
may be obtained if the misorientations are measured by
directly comparing diraction patterns from adjacent
regions [10]. However such a method is slow, and re-
quires patterns to be stored for later analysis. Although
the angular resolution may be problematic for the
determination of boundaries of very low misorientation
[5], the availability of excellent statistics for boundary
character distribution, such as shown in Fig. 1d, is one
of the strengths of EBSD.
4. Measuring small grains by EBSD
The sizes and shapes of grains or subgrains in an
EBSD map may be obtained by either linear intercept or
grain reconstruction methods [5]. A prior decision as to
what misorientation constitutes a high angle boundary
must be taken and 15 is often used. The ability to
precisely dene the nature of the boundaries constitutes
a signicant advantage over imaging techniques where
Fig. 2. Schematic diagram showing the variation of pattern intensity
at a boundary, and its eect on the eective spatial resolution.
774 F.J. Humphreys / Scripta Materialia 51 (2004) 771776
the visibility of a boundary is a function of the technique
and where all visible boundaries must be measured.
Thus in Fig. 1b, we are able to dierentiate between the
boundaries designated as high angle (>15) shown in
black, and the low angle boundaries, shown in white.
Fig. 1e shows the size distribution of the grains in the
sample of Fig. 1. When measuring ne-grained materi-
als, there are a number of factors which aect the
accuracy of the data.
4.1. The EBSD step size
If grain or subgrain sizes are to be determined from
an EBSD map then consideration must be given to the
pixel step size (d) in relation to the grain size (D) The
greatest accuracy will clearly be obtained if d is small,
and as d is increased, there is an increasing chance of
missing grains or grain intercepts and the measured
grain size will be larger than the true grain size. It has
been shown that to obtain an accuracy of 10% at least 5
pixels per grain are required, and for an accuracy of 5%,
a minimum of 8 pixels per grain are required [5].
4.2. The eects of non-indexed points
Although the methods of characterising grains or
subgrains by EBSD are simple in principle, problems
will arise when the size approaches the limit of spatial
resolution for EBSD.
Eq. (1) shows the relationship between the eective
spatial resolution, the grain size, and the fraction of
diraction patterns which can be solved. If we consider
an EBSD linescan, in a direction parallel to the specimen
tilt axis, of a material with an equiaxed grain structure
of mean linear intercept

L, then if K
P


L, all bound-
aries on the line will be detected and the measured grain
size (

L
M
) will be correct. However, as

L approaches K
P
,
the amount of non-indexing of patterns at boundaries
will cause some small boundary segments to be missed
and thus the measured linear intercept grain size in-
creases. This eect has been modelled [5], and the vari-
ation of

L
M
with K
P
, for K
P
=

L < 0:5, found to be given


by the empirical equation:
L
M
L
1
K
P
L
2
K
P
L

2
2
This indicates that the error in determining the grain
size is less than 10% if

L > 10K
P
but then increases
rapidly as K
P
approaches

L. Comparison with Eq. (1)
shows that the criterion for an error of less than 10% is
equivalent to a pattern indexed fraction (N
S
) of >0.85.
This suggests that the smallest grain or subgrain size
that can be accurately measured by EBSD by linear
intercept analysis in aluminium is 0.5 lm for a FEG-
SEM. However, if non-indexed points are replaced as
discussed above, then it can be shown that if more than
50% of the grains have been indexed, the error is less
than 10%, giving a lower grain size limit for aluminium
of 0.12 lm in a FEGSEM [5].
For a material of higher atomic number, Table 1
suggests that these gures may be substantially reduced.
For materials with smaller grain sizes the apparent grain
size measured by linear intercept could in principle be
corrected by the use of Eq. (2).
4.3. Other errors in grain size measurement
There are a number of other factors which must be
taken into account if accurate measurements of grain
size are to be obtained.
Instrument calibration: Care must be taken that the
scan distances are accurately calibrated and that the x
and y scans are orthogonal. International standards for
scanning electron microscopy are not yet available, al-
though standards for the calibration of the magnica-
tion of the SEM are being developed (ISO 16700) [11].
Specimen tilt axis: The specimen tilt axis must be
accurately aligned with one of the scan directions
otherwise a distorted scan raster will result.
Sample alignment: The surface of the sample must be
planar and accurately parallel to the xy plane of the
stage (i.e. normal to the electron beam. If this is not
achieved, then the inaccuracies in the scan raster on a
70 tilted sample may be substantial and render quan-
titative metallography impossible. The most signicant
errors occur if the untilted sample surface is misaligned
from the horizontal about the tilt axis by an angle (a).
On tilting for EBSD, the real specimen tilt is now
70 a and the raster scan in the y-direction which is
inversely proportional to the tangent of this angle is
substantially altered. For misalignments (a) of 1, 2 and
5 the errors in the magnitude of the y-scan are 5%, 13%
and 35% respectively [5].
5. Summary
Recent developments in instrumentation and data
analysis enable grains of 0.1 lm to be measured by
EBSD. Advantages over TEM include sample prepara-
tion and fully automated data acquisition and analysis,
which enable statistically signicant data to be readily
obtained from many thousands of grains.
Because of these advantages, EBSD is now being used
extensively to characterise sub-micrometre microstruc-
tures in deformed alloys e.g. [12] and in sub-micrometre-
grained materials produced by severe deformation
processes such as equal channel angular extrusion
(ECAE) [13,14]. The ability to select specic areas of a
large sample for analysis are particularly useful in
F.J. Humphreys / Scripta Materialia 51 (2004) 771776 775
materials with heterogeneous microstructures, such as
those produced by friction stir welding [15].
Although EBSD is now a mature technique, steady
improvements in instrumentation and software will
continue to increase its use as a tool for quantitative
metallography in ne-grained materials.
References
[1] Alam MN, Blackman M, Pashley DW. Proc Roy Soc Lond
1954;A221:224.
[2] Dingley DJ, Randle V. J Mater Sci 1992;27:4545.
[3] Humphreys FJ, Brough IJ. Microscopy 1999;195:6.
[4] Randle V, Engler O. An Introduction to Texture Analysis.
Amsterdam: Gordon and Breach; 2000.
[5] Humphreys FJ. J Mater Sci 2001;36:3833.
[6] Humphreys FJ. VMAP is a suite of programmes developed for
quantitative analysis of the EBSD data generated by the HKL
Channel acquisition system. It can be made available on request,
2000.
[7] Humphreys FJ, Huang Y, Brough I, Harris C. J Microsc
1999;195:212.
[8] Goldstein JL, Joy DC, Romig AD, Lyman CE, Fiori C, Lifshin E.
Scanning Electron Microscopy and X-ray Microanalysis. New
York: Plenum; 1992.
[9] Humphreys FJ, Bate PS, Hurley PJ. J Microsc 2001;201:50.
[10] Wilkinson AJ. Scr Mater 2001;44:2379.
[11] Dyson D. Proc Roy Mic Soc 2000;35:147.
[12] Hurley PJ, Humphreys FJ. Acta Mater 2003;51:1087.
[13] Prangnell PB, Bowen JR, Gholinia A. In: Dineson et al., editors.
Proceedings of the 22nd Riso International Symposium, Riso,
2001. p. 105.
[14] Apps PJ, Bowen JR, Prangnell PB. Acta Mater 2003;51:2811.
[15] Hassan KA, Norman AF, Prangnell PB. Mater Sci Forum
2002;396402:1549.
776 F.J. Humphreys / Scripta Materialia 51 (2004) 771776

You might also like