Sie sind auf Seite 1von 14

Microsc. Microanal.

17, 316–329, 2011


doi:10.1017/S1431927611000055 Microscopy AND

Microanalysis
© MICROSCOPY SOCIETY OF AMERICA 2011

A Review of Strain Analysis Using Electron


Backscatter Diffraction
Stuart I. Wright,1, * Matthew M. Nowell,1 and David P. Field 2
1
EDAX-TSL, 392 East 12300 South, Draper, UT 84020, USA
2
Washington State University, 239C Dana Hall, Pullman, WA 99164, USA

Abstract: Since the automation of the electron backscatter diffraction ~EBSD! technique, EBSD systems have
become commonplace in microscopy facilities within materials science and geology research laboratories
around the world. The acceptance of the technique is primarily due to the capability of EBSD to aid the research
scientist in understanding the crystallographic aspects of microstructure. There has been considerable interest
in using EBSD to quantify strain at the submicron scale. To apply EBSD to the characterization of strain, it is
important to understand what is practically possible and the underlying assumptions and limitations. This
work reviews the current state of technology in terms of strain analysis using EBSD. First, the effects of both
elastic and plastic strain on individual EBSD patterns will be considered. Second, the use of EBSD maps for
characterizing plastic strain will be explored. Both the potential of the technique and its limitations will be
discussed along with the sensitivity of various calculation and mapping parameters.
Key words: elastic strain, residual strain, electron backscatter diffraction ~EBSD!, orientation imaging micros-
copy ~OIM!

I NTR ODUCTION W HAT C AN B E O BSERVED IN I NDIVIDUAL


With the automation of the electron backscatter diffraction EBSD P ATTER NS ?
~EBSD! technique, EBSD systems have become common- The presence of strain in the crystal lattice produces observ-
place in microscopy facilities within materials science and able effects in EBSD patterns. The types of effects produced
geology research laboratories all around the world. The are different for elastic and plastic strains.
capabilities of EBSD to aid the scientist in understanding
the crystallographic aspects of microstructures have been Elastic Strain
well used in a wide variety of materials and applications.
Elastic strains distort the crystal lattice. If this strain is
However, another potential use of EBSD is in quantifying
uniaxial and along one of the principal directions of the
strain. Unfortunately, there is a considerable amount of
unit cell, then the strain produces a change in one of the cell
confusion as to what is practically possible to extract from
parameters as shown schematically in Figure 1 ~for more
EBSD in terms of strain analysis given the current state of
accurate pattern simulation, see Winkelmann, 2008!. This
EBSD technology. In EBSD studies of strain, it is important
distortion manifests itself in the patterns as a shift in some
to distinguish between elastic and plastic strain. Elastic
of the zone axes along with changes in the width of some of
strain and residual strain manifest themselves differently in
the diffraction bands. In the case of uniform dilation, the
EBSD results. This article reviews the current state of tech-
changes will only occur in the bandwidths.
nology in terms of strain analysis. First, the effects of strain
Figure 1 shows an exaggerated degree of strain; the
on the patterns themselves are reviewed, followed by indica-
elastic strains achieved in real materials are much smaller
tions of strains that can be observed in orientation maps,
and produce changes of just a pixel or two in a typical EBSD
and lastly statistical studies of strain. An effort has been
system with cameras with 1K ⫻ 1K resolution. It should be
made to include a wide selection of bibliographic refer-
noted that retracting the camera improves the pixel resolu-
ences. However, the references are not meant to be compre-
hensive but rather give a snapshot of current strain analyses
performed using EBSD.
It is important to recall that strain is not simply a scalar
value. While we can derive various scalar parameters to
describe strain ~i.e., dislocation density for describing plas-
tic strain!, strain is a second rank tensor ~Nye, 1957!. It
varies with direction in the material.

Figure 1. A crystal lattice “strained” 11% uniaxially in the hori-


Received September 7, 2010; accepted January 4, 2011 zonal direction and a schematic overlay of the patterns with ~gray!
*Corresponding author. E-mail: stuart.wright@ametek.com and without strain ~black!.
Strain and EBSD 317

Figure 2. An unstrained and a “bent” crystal and the resulting diffraction patterns.

tion so that the change may be as large as several pixels.


However, retracting the camera also reduces the signal inten-
sity that can make it difficult to get high quality patterns. If
the crystal is “bent,” as shown schematically in Figure 2,
then in addition to the zone axis shifts previously men-
tioned there would be a slight degradation in pattern qual-
ity as the planes within the diffraction volume are no longer
exactly parallel. As Keller et al. ~2004! have pointed out, this
leads to slight deviations in Bragg angle along the length of
the planes leading to blurring of the edges of the diffraction
bands.
As with uniaxial extension or compression or dilation
Figure 3. EBSD pattern from ~a! a well-prepared surface and from
effects, the effects from lattice bending are quite small
~b! a poorly prepared surface of zirconium.
resulting in shifts of only a pixel or two in the zone axis
positions and only slight degradation of pattern quality.
Thus, very careful image analysis of high-resolution images
is needed to make meaningful measurements of elastic
strain.

Plastic Strain
It is well known that plastic strain degrades the quality of
diffraction patterns as shown in Figure 3. This is one of the
reasons careful preparation of sample surfaces is critical to
successful EBSD work ~Katrakova & Mucklich, 2001!. The
pattern from the poorly prepared surface shown in Figure 3
is due to plastic strain introduced during polishing.
With plastic strain, the distortions in the crystal lattice
are relieved by the formation of dislocations. This can result
in two different effects on the diffraction patterns. Consider
Figure 4. Schematic of dislocations and subgrain boundary.
the diagram of dislocations in a material shown in Figure 4.
There are regions in the material with significant disloca-
tion density with a net Burgers vector of zero. These are
sometimes called “statistically stored” dislocations ~SSDs!.
There are also areas with net nonzero Burgers vectors across
which there is a change in crystallographic orientation or
lattice curvature. These dislocations are often termed “geo-
metrically necessary” dislocations ~GNDs!. Arrays of GNDs
can form subgrain boundaries.
If the diffraction volume is contained within a region
of high dislocation density but with a net Burgers vector of
zero ~i.e., contained entirely within a subgrain!, then the
resulting pattern is degraded due to local perturbations of
the diffracting lattice planes leading to incoherent scatter-
ing. This is shown schematically in Figure 5 for a single Figure 5. Diagram showing the disturbance in the crystal lattice
dislocation—the higher the dislocation density, the greater due to two opposing edge dislocations leading to a degraded EBSD
the degradation in pattern quality. pattern.
318 Stuart I. Wright et al.

avoid shifts arising from deflection of the beam between the


strained and unstrained material.
If the cross-correlation methods are employed, a pat-
tern can be analyzed to extract both the elastic and plastic
strain simultaneously. Cross correlations from at least four
different regions in a strained pattern are compared against
the same regions in an unstrained reference pattern to
measure a displacement gradient tensor of components
]u i /]x j , where u ⫽ ~u 1 , u 2 , u3 ! is the displacement at posi-
tion x ⫽ ~ x 1 , x 2 , x3 ! in the sample. The elastic strain ~eij ! is
the symmetric part of the displacement gradient tensor:

Figure 6. Schematic showing the effect of a subgrain boundary on


eij ⫽ 冉
1 ]u i
2 ]x j

]u j
]x i

the EBSD pattern. and the antisymmetric part are the rotations ~wij ! associ-
ated with the plastic strain:
If the diffraction volume contains GNDs, then the
pattern quality is degraded as the pattern is essentially a
superposition of the patterns from each individual subgrain
wij ⫽ 冉
1 ]u i
2 ]x j

]u j
]x i
冊 .

within the diffraction volume as shown schematically in Because a hydrostatic dilation does not generate a shift
Figure 6. As the rotations associated with subgrain bound- in the strained pattern, the individual diagonal terms of the
aries are small, the material within the diffraction volume displacement gradient tensor ~]u i /x i ! cannot be deter-
will no longer meet a specific Bragg condition but rather a mined. e11-e33 and e22-e33 can be determined but not all
range of near Bragg conditions in three dimensions. This three terms independently. However, if we assume the stress
results in a diffraction pattern with degraded contrast. The g33 normal to the free surface is zero and use the single
degree of degradation is dependent on the amount of crystal stiffness constants for the material being analyzed,
deformation within the interaction volume. then the three terms can be isolated. Using this methodol-
Differentiating the contribution of SSDs versus that ogy both the elastic strain tensor and the GND tensor can
from GNDs in a single pattern is practically impossible. be determined.
Since the size of the diffraction volume is a function of Recent work has shown that resolutions similar to
beam size, the degradation effect will differ in lower resolu- those achieved by Wilkinson et al. can be achieved using an
tion microscopes with tungsten filaments as opposed to the iterative process based on a simulated strain-free reference
higher resolution microscopes with field emission sources. pattern to extend the capabilities of this type of approach
In fact, depending on the amount of deformation and ~Kacher et al., 2009!. The limits of this kind of analysis have
material ~i.e., materials that form well organized dislocation been explored by Villert et al. ~2009!. It should be noted that
structures!, the contrast degradation in the patterns may be the measurement of the rotations due to plastic strain
minimal. requires a reference pattern for the calculation. However, it
need not be a strain-free pattern, rather a pattern from a
neighboring point of known position in the strained crystal.
Quantitative Analysis
Various approaches have been proposed ~Troost et al., 1993; W HAT C AN B E O BSERVED IN
Wilkinson, 1997; Tao, 2003; Bertness et al., 2004! for mak-
ing quantitative measurements of both elastic and plastic
O RIENTATION I MAGING
strain but have not seen widespread adoption yet within the M ICR OSCOPY M APS ?
EBSD community. Generally, these approaches employ cali- Automated EBSD is ideal for characterizing the spatial
bration curves based on measured shifts of pixels in pat- distribution of crystallographic orientation within polycrys-
terns in strained materials relative to measurements made talline microstructures. This technique is often termed ori-
in reference patterns from unstrained material of the same entation imaging microscopy ~OIM! ~Adams et al., 1993!.
crystallographic orientation. From these shifts it is possible Orientations maps constructed from OIM data provide a
to calculate the elastic strains as well as high-resolution visualization of the crystallographic orientations of the con-
rotations that are indicative of plastic strain. Recent work by stituent grains.
Wilkinson et al. ~2006a! using cross-correlation functions
for comparing the patterns from the strained and un- Elastic Strain
strained material have shown that most components of the To maximize speed during automated scans, the captured
strain tensor can be obtained using this methodology. They image size of the diffraction patterns can be quite small ~i.e.,
were able to obtain resolutions of strain measurements to 2 100 ⫻ 100 pixels!. Thus, the shifts in the zone axes and/or
parts in 10,000 using this method. Care must be taken to bandwidths are difficult if not impossible to observe in
Strain and EBSD 319

Figure 8. IQ map of a lenticular martensite grain and the surround-


ing austenite. The intensity hisogram has been modified from that
Figure 7. Line scans of the elastic strain tensor component ~e11 ! shown in Figure 7 so as to accentuate the strained regions around
in austenite surrounding a lenticular martensite grain ~max ⫽ the martensite grain. ~Data courtesy of G. Miyamoto.!
0.009!. ~Data courtesy of G. Miyamoto.!

typical OIM scans. Even when utilizing large patterns dur-


ing a scan, the shifts are difficult to measure. In principle,
the image quality ~IQ! of the diffraction patterns can be
used to map out the elastic strain ~Keller et al., 2004!.
However, these results are only qualitative, simply giving a
relative indication of elastic strain. Any plastic strain in the
material will dominate the IQ map masking any degrada-
tion in the diffraction patterns due to elastic strain: quanti-
tative results obtained using the cross-correlation methods
introduced by Wilkinson et al. ~2006a! suggest that maps of
components of the elastic strain tensor are possible. An
example is shown in Figure 7 for line scans made across an
austenite region surrounding a lenticular martensite grain
in work by Miyamoto et al. ~2009!. Figure 9. IQ map showing the many factors influencing the IQ
value. In this sample of a reaction zone between copper and
aluminum, IQ contrast resulting from scratches during sample
Plastic Strain preparation, different phases, pores, cracks, and grain boundaries
There are essentially two approaches to analyzing plastic can all be observed.
strain: the first based on the degradation of the diffraction
patterns in strained materials ~Krieger Lassen et al., 1994! linear relationship between the IQ value and the local
and the second based on local misorientation ~Field, 1995!. strain.
As discussed previously, the degradation is a function of the Unfortunately, IQ is not solely dependent on strain;
scanning electron microscopy ~SEM! spot size. Thus, the other factors affect image quality as well. Such factors
pattern quality approach may be better suited to low- include grain boundaries, surface topology, second phases,
resolution microscopes ~i.e., tungsten filament SEMs! where beam conditions, sample preparation, and camera settings,
the diffraction volumes will be larger and the local misori- some of which can be observed in Figure 9. When the beam
entation approach to high-resolution microscopes ~i.e., field is located on a grain boundary the diffraction volume will
emission gun SEMs!. include crystal lattices in two different orientations result-
ing in a combined pattern with contributions from both
Image Quality Approach lattices. This leads to a lower IQ leading to an incorrect
At each measurement point in an OIM scan, a parameter characterization of every boundary in the material being a
quantifying the quality of the corresponding diffraction site for high strain concentration. Surface topology will also
pattern is recorded. Maps can be generated based on this IQ affect IQ as the surface will deviate from the ideal tilt
parameter as shown in Figure 8. In such maps, strained condition for EBSD. ~This can actually be helpful in some
areas appear darker than unstrained regions of the micro- strain studies, as slip traces can be observed in OIM maps.!
structure. As already expressed, the image quality is affected If the beam is unstable during an OIM scan, there may be a
by residual strain in the diffracting volume. Thus, an indica- loss of intensity during the scan resulting in reduced IQ
tion of the distribution of strain in the material can be values as the scan proceeds. The relative grain-to-grain
observed in an IQ map ~Wardle et al., 1994!. For a very large values will be useful, but the absolute values cannot be
scan area on a bulk sample, if the average IQ value is correlated to strain in this case. Video conditions may also
assumed to correspond to the overall strain measured me- affect the IQ. Imagine a case where the gain is controlled
chanically, then the strain can be quantified by assuming a automatically during a scan. In this case the strain effects
320 Stuart I. Wright et al.

As was mentioned previously, the size of the diffraction


volume will affect the results. Thus, the SEM used for
making the measurements may have an impact on the
resulting maps. For a lower resolution microscope with a
larger diameter beam, we expect a larger diffraction volume.
Thus, we would expect the IQ effect to be more marked as
the larger diffraction volume will be more likely to contain
lattice imperfections. For example, in a deformed material
with a well organized subgrain structure, the larger diffrac-
tion volumes obtained in a low-resolution microscope are
likely to contain multiple subgrains leading to diffuse pat-
Figure 10. IQ map showing low IQ values at grain boundaries
terns, whereas in a high-resolution microscope, it may be
and a grain averaged IQ map ~94.7% indexing success rate and
16.0% cleanup!. possible to sample each subgrain separately leading to good
quality patterns despite the presence of deformation in the
material. An example is shown for partially recrystallized
could not be tracked using the IQ parameter as the camera steel in Figure 11. The impact of the larger diffraction
system would attempt to compensate for low IQ values by volume associated with the tungsten filament SEM relative
changing the video conditions. These factors make it diffi- to the field emission gun SEM is clearly more pronounced
cult to correlate absolute values of IQ with strain. Some in the case of the image quality map than in the orientation
work has been done to mitigate the camera configuration map. The data in these maps have been processed using the
effects by using a control sample for normalization, but the same dilation “clean-up” routine ~Wright, 2006!. It is inter-
results have not been promising ~R.A. Witt & M.J. Merwyn, esting to note than 1.2% of the data points were modified in
2008, personal communication!. the tungsten case and only 0.1% were modified in the field
Some of these effects can be mitigated through various emission case.
means. For example, the grain boundary effects can be
reduced by averaging IQs within the individual grains as Local Misorientation Approach
shown in Figure 10. The variation in strain within a grain As dislocations form in the material, the residual strain is
would be lost, but the ability to make grain-to-grain com- manifest as local variations in lattice orientation. This is
parisons of strain would be improved and boundaries would evident in maps on grain boundary maps of strained mate-
not be assumed to be sites of strain concentration. The rials. OIM measurements allow a user to define which kind
indexing success rate stated in the caption for Figure 10 is of boundaries should be drawn in maps created in OIM
defined as the percentage of points with confidence index data. For example, Figure 12 shows low angle grain bound-
values ~Field, 1997! greater than 0.1 after standardizing the aries ~2–158! in red and high angle ~.158! boundaries in
confidence index to the point in each grain with the highest blue. Regions with high concentrations of low angle bound-
confidence index ~Nowell & Wright, 2005!. The percentage aries are indicative of areas of concentrated GND density.
of points “cleaned up” are those points modified using a Some authors ~Wheeler et al., 2003! have used the
grain dilation post-processing routine ~Wright, 2006! di- boundary density ~boundary line length over area! to char-
vided by the total number of points in the scan. For the acterize deformation. Figure 13 shows the change in bound-
dilation routine, a grain tolerance angle of 58 and a mini- ary density in successive scans of a heavily deformed copper
mum grain size of 3 pixels were used. The grain was also sample held at temperature ~1558C! ~Wright et al., 2005! to
required to extend across at least two rows in the scan. study the recrystallization process in situ. The copper was
These same definitions are used in the captions throughout deformed by equal-angle channel extrusion. The scan times
the article for each figure where maps are shown. were approximately 1.6 min each.
A subtle factor affecting image quality is the orienta- As with the IQ, local misorientations provide an indica-
tion of the crystal lattice itself. Some planes in the crystal tion of the strain distribution in the material. Several useful
lattice produce bands of higher intensity than others. If the methods for characterizing local misorientations have been
lattice is in an orientation such that the pattern contains proposed. The first two methods are “grain” based. This
many of these high-intensity reflections, then the pattern means that every measurement contained within a grain is
will have a higher IQ value than a pattern from an orienta- assigned the same local misorientation value, but the values
tion with less high-intensity reflectors. Thus, there will be differ from grain to grain. The second three are based on
some grain-to-grain variation in IQ simply from this orien- individual measurement points. In this case, each point in
tation effect ~Wright & Nowell, 2006!. the scan has its own individual value of local misorienta-
Other measures of IQ ~Tao & Eades, 2005! have been tion. The last one is a hybrid of the two approaches. Each
proposed to reduce such effects, but these provide only point in the scan has its own individual value of local
minimal improvement ~Wright & Nowell, 2006!. It should misorientation, but the calculation of this value is based on
be emphasized that such effects are generally much smaller the grain to which each individual scan point belongs. Most
than those due to strain. of the maps shown in the following figures were obtained
Strain and EBSD 321

Figure 13. Boundary density for successive scans in an in situ


heating experiment of heavily deformed copper.

from a partially recrystallized steel sample at two different


magnifications.

1. The grain orientation spread ~GOS! is the average devia-


tion in orientation between each point in a grain and the
average orientation of the grain. This approach leads to
assigning the same value to every scan point contained
Figure 11. IQ and inverse pole figure ~IPF! maps from a partially within the grain. The average orientation of the grain is
recrystallized steel sample using a thermionic field emission gun calculated using the methodology outlined by Kunze
SEM: ~a,b! 99.9% indexing success rate and 0.1% cleanup, and et al. ~1993!. The GOS can also be calculated by determin-
using a tungsten filament SEM, ~c,d! 98.7% indexing success rate
ing the average misorientation between each point in the
and 1.2% cleanup.
grain with every other point in the grain ~Wright, 1999!.
An example GOS map is shown in Figure 14.
2. A similar measure is the grain average misorientation
~GAM!, which is the average misorientation between
each neighboring pair of measurement points within the
grain ~Wright, 1999!. This approach is more sensitive to
the step size of the measurement grid than the GOS
approach. In general, as the step size decreases, the mis-
orientation between neighboring points on the scan grid
will decrease as well. An example GAM map is shown in
Figure 14.
3. The kernel orientation spread ~KOS! is similar to GOS but
done within a kernel. A kernel is a set of points of
prescribed size surrounding the scan point of interest.
The size of the kernel is generally prescribed to the nth
nearest-neighbor. Examples of 3rd nearest-neighbor ker-
nels are shown in Figure 15. The term window is also
used to describe the set of points but is generally re-
served for calculations on a square grid. The calculations
performed on this kernel are performed with the kernel
centered at each point in the scan and value obtained by
the calculations assigned to the center point. If more
than one grain is contained within the kernel, then
multiple average orientations are calculated for the ker-
nel. The value calculated for the kernel is assigned to the
scan point at the center of the kernel.
4. The kernel average neighbor misorientation ~KANM! is
similar to GAM but calculated within a kernel instead of
Figure 12. Grain boundaries in 85% recrystallized low carbon a grain. That is, misorientations between all neighboring
steel: red, 2–158; blue, .158 ~96.6% indexing success rate and 0% points within the kernel are averaged. If the kernel con-
cleanup!. tains a grain boundary, then some of the misorientations
322 Stuart I. Wright et al.

Figure 14. OIM maps from a partially recrystallized steel sample. ~a! Color code orientation map, ~b! GAM map, and
~c! GOS map. Grains are outlined with black boundaries and were constructed assuming a 58 tolerance angle ~99.5%
indexing success rate and 0.7% cleanup!.

2002! shows the deviation in orientation of a measure-


ment point from the average orientation of the grain to
which the point belongs. Various approaches have been
considered for specifying the reference orientation. Two
approaches that have been reported are the use of the
average orientation for a grain as the reference orienta-
tion ~Wright, 1993! and the point in the grain with the
lowest kernel average misorientation ~Brewer et al., 2002!.
More complex characterizations of the orientation devi-
Figure 15. 3rd nearest-neighbor kernels for a hexagonal and a ation have been mapped using Rodrigues vectors ~Field
square grid. et al., 2005!.
Local Misorientation Parametric Sensitivities
would be quite large. Thus, to focus on local small
rotations, the grain boundary effect is excluded by only With each of the approaches for characterizing local misori-
including misorientations less than a specified tolerance entation described above, various associated parameters will
value from the averaging calculation ~Lehockey et al., affect the results. For example, as the scan step size in-
2000!. creases, the misorientations between neighboring points on
5. The kernel average center misorientation ~KACM! is the the scan grid tend to increase as well. These parameteric
average misorientation between a point on the measure- sensitivities and the various map types that they affect are
ment grid and its neighbors. For a hexagonal measure- described below.
ment grid, this is the average misorientation between the Grain Tolerance Angle (GOS, GAM, ROD)
orientation of a given measurement point and the orien- Grains in OIM are constructed in a very specific way.
tations of the six equidistant neighbors. As with KANM, The grain grouping algorithm simply enforces that any two
the grain boundary effect is removed by only including neighboring points on the scan grid, which differ in orien-
points misoriented relative to the point at the center of tation less than a specified tolerance angle, belong to the
the kernel within some prescribed value included in the same grain group or simply “grain.” Thus, in a deformed
averaging calculation. If the kernel is larger than the first grain the point-to-point misorientation will all be less than
nearest-neighbors, then KACM can be calculated in two the tolerance angle, but the misorientation between a data
different ways: ~a! the misorientation from each point in point at one end of a grain with respect to a point at the
the kernel with respect to the center point of the kernel is other end may exceed the grain tolerance angle by a consid-
calculated and averaged or ~b! only the points at the erable amount. Thus, the choice of tolerance angle has a
perimeter of the kernel are used in the calculation. If significant effect on any grain-based calculations of local
only the nearest-neighbors are used, then the variants are misorientation. Figure 17 shows maps from a partially re-
identical. Because this method focuses on the center crystallized steel sample. At large tolerance angles, recrystal-
point of the kernel, the boundary edges are well pre- lized grains are grouped together with adjacent deformed
served as can be observed in Figure 16. grains, whereas at small tolerance angles they are separated.
6. The reference orientation deviation ~ROD! map ~Wright, Also at the fine step sizes used in this example ~750 nm!,
1993; Brewer et al., 2002; Kysar & Briant, 2002; Li et al., some of the deformed grains are broken down into individ-
Strain and EBSD 323

Figure 16. ~a! KOS, ~b! KANM, and ~c! KACM maps from a partially recrystallized steel sample—6th nearest-neighbor
kernels ~99.9% indexing success rate and 0.1% cleanup!.

Figure 17. IPF maps for a partially recrystallized steel sample: ~a! full scan field, ~b! close up of highlighted area where
grains are formed using a 7.58 grain tolerance angle ~grains delineated by black lines!, and ~c! with a 2.58 grain tolerance
angle ~99.5% indexing success rate and 0.7% cleanup!.

Figure 18. GOS, min grain size ⫽ 5 pixels, 2.58, 58, and 108 grain tolerance angles overlaid with “grain” boundaries
~99.5% indexing success rate and 0.7% cleanup!.

ual subgrains. This is evident in the figure where the black Grain Size Effects (GOS, GAM, ROD)
lines delineate grains. Even with a small 28 grain tolerance In the grain grouping algorithm, a grain must contain a
angle, there appears to be a recrystallized grain attached to minimum number of scan points to be identified as a grain.
the deformed grain. This demonstrates why it is difficult to This minimum size may also affect the results. As the
use grain-based local misorientation approaches to estimate minimum grain size is increased, small grains ~or subgrains!
volume fractions of deformed versus recrystallized material may be excluded from the calculations. Figure 19 shows the
in partially recrystallized samples. impact of the choice of the minimum grain size on the GOS
The choice of tolerance angle can affect both the GAM results focused on the deformed grains ~grains with orienta-
and GOS results. The effect on the GOS results is shown in tion spreads less than 28 were excluded from the analysis!.
Figure 18. As the tolerance angle gets larger, the subgrains As the minimum grain size is increased, we exclude small
are encompassed within the larger deformed grains, and the subgrains contained within the larger deformed grains.
spread of orientation in these larger grains increases. Within a subgrain, the local variation in orientation may be
324 Stuart I. Wright et al.

Figure 19. GOS distribution for scans over the same area at differ-
ent minimum grain sizes.

Figure 21. KACM maps, 58 exclusion angle, 2nd nearest-neighbors


at step sizes of 200, 400, 800, and 1600 nm ~99.9% indexing success
rate and 0.1% cleanup!.
Figure 20. GOS and GAM distributions for the partially recrystal-
lized steel at step sizes of 200, 400, 800, and 1600 nm. The grain tion ~KAM! maps for the same partially recrystallized steel
tolerance angle is 58, and and the minimum grain size is 2 points. sample, albeit scanned at higher magnification. The exis-
tence of subgrains is evident in the smallest scan step
quite small, and thus as the minimum grain size is increased size, whereas the subgrain structure is lost at the larger step
the curves shift toward a higher fraction of grains with large sizes.
orientation spreads. Some of these effects can be mitigated by other param-
eters in the software such as the minimum grain size or
Scan Step Size (All) kernel size depending on the approach being used. For
Local misorientation measurements are very sensitive example, Figure 22 shows KAM maps for three different
to the step size at which an OIM scan is collected. This can step size scans, but the size of the KAM is adjusted so that
be observed for the GOS and GAM results for the same the kernel is physically the same size in each map. It should
partially recrystallized steel sample used previously. Fig- be noted that the results are very similar ~yet not identical!
ure 20 shows the average GOS and GAM values for the for each step size and appropriate choice of kernel size
deformed grains for the map shown in Figure 17 with the 58 indicating that the two parameters are indeed coupled.
grain tolerance angle as a function of step size. The GAM The orientation spread approaches ~GOS and KOS!
results increase until a step size of about 3 mm and then tend should be less sensitive to step size than the approaches that
to level out. This suggests that the subgrain size in this mate- focus on misorientations between adjacent points on the
rial is about 3 mm. The GOS results are more difficult to measurements grid ~GAM, KANM, and KACM!. However,
interpret. In this case, the average orientation spread in each the results in Figure 23 do not seem to bear this out. In the
grain decreases with decreasing step size. This may simply be scan data used in this example, the KACM distributions are
due to the decreasing statistics for the spread calculation in less sensitive to step size than the KOS distributions. How-
each grain as the number of points in each grain decreases ever, as expected the KANM distributions are very sensitive
with decreasing step size. A wide range of grain sizes is seen to step size in a predictable way. As the step size increases,
in this scan area, and an attempt was made to see if these the KANM values increase as well.
results varied by grain size. This was simply done by calculat-
ing area averages in addition to the number averages shown Kernel Size (KOS, KANM, KACM)
in Figure 20. It is evident that the trends are consistent with The kernel size effect essentially mimics the step size
the trends observed for the number averages. effect as shown in Figure 24. As the kernel size increases, the
The step size sensitivity is even greater in the kernel- small substructure details are smeared out. The smearing
based calculations. This is particularly evident in Figure 21 observed is similar to that observed when applying smooth-
showing 2nd nearest-neighbor kernel average misorienta- ing operations in image processing.
Strain and EBSD 325

Figure 22. KACM maps at step sizes of 400, 800, and 1600 nm with 8th, 4th, and 2nd nearest-neighbor kernels ~99.9%
indexing success rate and 0.1% cleanup!.

Figure 23. ~a! KOS, ~b! KACM, and ~c! KANM distributions for the partially recrystallized steel data shown in Figure 17
~1,000⫻ data!. The distributions are for step sizes of 400, 800, and 1600 nm with 8th, 4th, and 2nd nearest-neighbor
kernels, respectively.

points varying more than a specified angle are excluded


from the analysis. If a larger exclusion angle is used, then
the calculations actually focus on grain boundaries as op-
posed to the local orientation perturbations as shown in
Figure 25.

All Points Versus Perimeter Points (KACM)


In the case of the KACM calculations, the option to
look at the orientation of each point in the kernel with
respect to the orientation of the point at the center of the
kernel can be used. Alternatively, only those points at the
perimeter of the kernel may be used. If only the perimeter
Figure 24. All points KACM, 58 exclusion angle, 1st and 10th points are used, the average misorientation calculated for a
nearest-neighbors ~99.9% indexing success rate and 0.1% cleanup!. kernel tends to be larger because the average distance be-
tween the points used in the calculation from the center
Kernel Exclusion Angle (KOS, KANM, KACM) point are farther away, thus essentially mimicking the step
To distinguish grain boundary effects from local orien- size effect discussed previously. This effect is evident in
tation variations due to strain, misorientations between Figure 26.
326 Stuart I. Wright et al.

Figure 25. All points KACM, 2nd nearest-neighbors, 5, 15, and 1808 exclusion angles ~99.9% indexing success rate and
0.1% cleanup!.

Figure 26. ~a! All points and ~b! permiter points only KACM, Figure 27. ~a! Reference orientation deviation maps using the av-
10th nearest-neighbors, 58 exclusion angles ~99.9% indexing suc- erage orientation for each grain as the reference and ~b! the orien-
cess rate and 0.1% cleanup!. tation in the grain with the smallest 1st nearest-neighbor KACM as
the reference ~99.9% indexing success rate and 0.1% cleanup!.
Average Orientation Versus Minimum Kernel-Based
Reference Orientations (ROD)
In the ROD map, various methods can be employed for
selecting the reference orientation. The results for two po-
tential choices of reference orientation are shown in Fig-
ure 27. One approach is to use the average orientation
calculated from all scan points contained in the grain. It
should be noted in this case that a point with this orienta-
tion may not actually exist in the grain. However, when
using the point in the grain with the lowest kernel average
misorientation, a point with this orientation must necessar-
ily exist within the grain. It should be pointed out that any
point colored blue in Figure 27 is very near the reference Figure 28. Distribution of average misorientation between the
orientation for the grain to which the point belongs. kernel center points and the diagonal and off-diagonal neighbors
in a first-order kernel.
Grid Type (All)
For each type of kernel-based map, there will be an This is essentially the same effect as the step size effect as the
effect that will arise from the definition of a kernel on a diagonal neighbors are farther away than the off-diagonal
square map versus that defined for a hexagonal. This is neighbors.
most notable in a first nearest-neighbor KACM map. Fig-
ure 28 shows the average misorientation between the center Tensor Analysis
of the kernel and the off-diagonal first nearest-neighbor These local misorientation measurement approaches essen-
points and between the center of the kernel and the diago- tially assume plastic strain is a scalar value, i.e., they focus
nal first nearest-neighbors for square grid data obtained simply on GND density. More advanced analyses of OIM
from a nickel superalloy. As expected, the diagonal elements data can be performed to extract the GND tensor providing
are shifted slightly toward greater misorientation values. a more complete description of the strain state within the
Strain and EBSD 327

Figure 29. Lines scans of the rotation tensor component w23


overlaid on a 10th nearest-neighbor KACM map ~white . 28!.
~Data courtesy of G. Miyomoto.!

Figure 31. 5th nearest-neighbor KAM distribution and ~100!, ~110!,


and ~111! pole figures for textures calculated for two subsets of the
full dataset partitioned based on the KAM distribution for par-
tially recrystallized steel.

that the strain distribution in the material is inherently


three-dimensional ~3D! while the local orientation varia-
tions from a single OIM map are obtained on a two-
dimensional section plane. The variations in orientation
can vary considerably from those measured in a single
section plane. This is a particular problem with grain-based
measures of misorientation as in one plane a grain may
appear strained, whereas in another plane it may appear
unstrained. This can be observed in the results shown in
Figure 30. EBSD data were collected on serial sections
through a Ni-based superalloy sample sectioned by in situ
Figure 30. Grain orientation spread calculated for two grains for a focused ion beam milling ~Uchic et al., 2004!. The GOSs of
series of two-dimensional slices in the x ~ yz plane!, y ~xz plane!, two grains in the 3D data cube were analyzed along each of
and z ~xy plane! directions for a 3D OIM dataset. ~Data courtesy
the three principal directions. The measured values clearly
of M. Uchic.!
differ depending on which cross sections are examined.

material ~Sun et al., 2000; El-Dasher et al., 2003; Field et al.,


2005; Wilkinson et al., 2006b; Pantleon, 2008!. Figure 29
S TATISTICAL C ORR ELATION OF
shows the w12 rotation component of the strain tensor in EBSD S TRAIN I NDICATORS WITH
austenite surrounding a lenticular martensite grain. C RYSTAL O RIENTATION
As described previously, various metrics such as IQ or local
Scaling to Bulk Strain misorientation can be correlated to calibration curves to
If the average values from such maps are assumed to corre- provide a somewhat quantitative measure of strain within
spond to the measured overall strain, then the local strain the scan area. The general idea is to correlate the strain
can be quantified by assuming a linear relationship between behavior of polycrystals with crystallographic orientation.
the local misorientation metric and the strain. However, it Two approaches can be taken. The first is to partition the
should be remembered that the misorientation metrics are data using one of the various measures of “strain,” while the
based solely on GNDs and provide no information on second is a more continuous approach termed “scalar”
SSDs. One severe limitation to such an approach is the fact texture.
328 Stuart I. Wright et al.

C ONCLUSION
Various research teams around the world are exploring ways
to improve the strain measurement capabilities of EBSD for
both elastic and residual strain. However, it is important to
remember that most of the proposed methods for character-
izing plastic or residual strain give only an indication of
strain; extrapolating actual strain values remains difficult
and should be done with great care. It is also important to
be very careful of sample preparation as EBSD is very much
Figure 32. ~100!, ~110!, and ~100! pole figures showing the aver- a surface sensitive technique. Any deformation introduced
age KAM value as a function of pole orientation for a partially
during preparation of the sample can mask the true defor-
recrystallized steel.
mation structure ~Field et al., 2010!.
EBSD is very much capable of identifying areas of
Partitioning concentrated strain within a microstructure but not well
Because of the discrete nature of OIM data, it is possible to suited to identifying the actual magnitude of strain in those
correlate the strain measurements with specific orienta- areas. However, some of the current research shows promise
tions. One approach is to divide a dataset into two subsets for more quantitative measurement of both elastic and
or partitions. Figure 31 shows an example for a large plastic strain using EBSD.
combined beam and stage scan obtained on the partially It should be noted that most of the measurements
recrystallized steel sample already described. This measure- described here are highly sensitive to sample preparation.
ment contains nearly 3.4 million individual orientation Most sample preparation induces some deformation into
measurements. The KAM distribution was calculated for the surface of a sample. Absolute values of plastic strain will
these measurements, and the data were then partitioned always be difficult to obtain on prepared surfaces. When
into two subsets: one containing all of the measurements comparing results from different samples, it is very impor-
with KAM values less than 1.58 and another subset con- tant to maintain the same sample preparation routines and
taining all of the orientation measurements with KAM setting on the microscope and EBSD hardware as well as
values greater than 1.58. The texture calculated for the low any applied post processing.
angle KAM partition thus represents the texture of the
recrystallized material, whereas the texture of the high angle A CKNOWLEDGMENTS
KAM partition represents the texture of the deformed
The authors are indebted to Goro Miyamoto of the Institute
material.
of Materials Research at Tohoku University and Michael
Uchic of the Air Force Research Lab at Wright-Patterson Air
Scalar Textures Force Base for providing data used in constructing some of
The correlation between any scalar parameter and crystal- the figures in the article. We also acknowledge Rouman
lographic orientation can be characterized using an ap- Petrov of Ghent University for providing the partially recrys-
proach termed scalar textures ~Wright et al., 1993; Nowell & tallized steel samples used in the various local misorienta-
Wright, 2005!. In a scalar texture, orientation space is tion maps.
divided up into many bins. For each orientation measure-
ment in the OIM scan, the bin corresponding to the R EFER ENCES
measured orientation is incremented by the value of the Adams, B.L., Wright, S.I. & Kunze, K. ~1993!. Orientation imag-
scalar parameter in question. A second value is tracked for ing: The emergence of a new microscopy. Metall Trans 24A,
the bin, that is, the number of points contributing to the 819–831.
bin. Thus, an average value for the given scalar parameter Bertness, K.A., Geiss, R.H., Keller, R.R., Quinn, T.P. & Roshko,
can be calculated for each bin. This enables a function like A. ~2004!. EBSD measurement of strains in GaAs due to oxida-
the orientation distribution function to be calculated. In- tion of buried AlGaAs layers. Microelectron Eng 75~1!, 96–102.
stead of the function providing a value of times random, it Brewer, L.N., Othon, M.A., Young, L.M. & Angeliu, T.M. ~2002!.
provides the average scalar value. “Pole figures” and “inverse Misorientation mapping for visualization of plastic strain via
pole figures” can be constructed as well showing the varia- electron back-scattered diffraction. In Microscopy and Micro-
tion in the scalar parameter average value with crystallo- analysis 2002, Voelkel, E., Piston, D., Gauvin, R., Lockley, A.J.,
Bailey, G.W. & McKernan, S. ~Eds.!, pp. 684CD–685CD. Québec
graphic orientation. This technique works best when there
City, Québec, Canada: Cambridge University Press.
are many random orientations. Using this approach any of El-Dasher, B.S., Adams, B.L. & Rollett, A.D. ~2003!. Viewpoint:
the parameters described previously such as the IQ or the Experimental recovery of geometrically necessary dislocation
measures of local misorientation can be determined as a density in polycrystals. Scripta Mater 48~2!, 141–145.
function of orientation. Figure 32 shows the KAM scalar Field, D.P. ~1995!. Quantification of partially recrystallized poly-
texture for the large dataset described in the previous crystals using electron backscatter diffraction. Mater Sci Eng A
section. 190, 241–246.
Strain and EBSD 329

Field, D.P. ~1997!. Recent advances in the application of orienta- Tao, X. & Eades, A. ~2005!. Errors, artifacts and improvements in
tion imaging. Ultramicroscopy 67~1–4!, 1–9. EBSD processing and mapping. Microsc Microanal 11, 79–87.
Field, D.P., Kumar, M., Trivedi, P.B. & Wright, S.I. ~2005!. Troost, K.Z., Vandersluis, P. & Gravesteijn, D.J. ~1993!. Micro-
Analysis of local orientation gradients in deformed single crys- scale elastic-strain determination by backscatter Kikuchi diffrac-
tals. Ultramicroscopy 103~1!, 33–39. tion in the scanning electron-microscope. Appl Phys Lett 62~10!,
Field, D.P., Magid, K.R., Mastorakos, I.N., Florando, J.N., 1110–1112.
Lassila, D.H. & Morris, J.W., Jr. ~2010!. Mesoscale strain Uchic, M.D., Groeber, M., IV, Wheeler, R., Scheltens, F. &
measurement in deformed crystals: A comparison of X-ray Dimiduk, D.M. ~2004!. Augmenting the 3D characterization
microdiffraction with electron backscatter diffraction. Philos capability of the dual beam FIB-SEM. Microsc Microanal 10~S2!,
Mag 90, 1451–1464. 1136–1137 ~CD-ROM!.
Kacher, J., Landon, C., Adams, B.L. & Fullwood, D. ~2009!. Villert, S., Maurice, C., Wyon, C. & Fortunier, R. ~2009!.
Bragg’s law diffraction simulations for electron backscatter Accuracy assessment of elastic strain measurement by EBSD. J
diffraction analysis. Ultramicroscopy 109~9!, 1148–1156. Microsc 233~2!, 290–301.
Katrakova, D. & Mucklich, F. ~2001!. Specimen preparation Wardle, S.T., Lin, L.S., Cetel, A.D. & Adams, B.L. ~1994!. Orien-
and electron backscatter diffraction—Part I: Metals. Prac Met- tation imaging microscopy: Monitoring residual stress profiles
allog 38~10!, 547–565. in single crystals using and imaging quality parameter, IQ. In
Keller, R.R., Roshko, A., Geiss, R.H., Bertness, K.A. & Quinn, Proceedings of the 52nd Annual Meeting of the Microscopy Soci-
T.P. ~2004!. EBSD measurement of strains in GaAs due to ety of America, Bailey, G.W. & Garratt-Reed, A.J. ~Eds.!, pp. 680–
oxidation of buried AlGaAs layers. Microelec Eng 75~1!, 96–102. 681. San Francisco, CA: San Francisco Press.
Krieger Lassen, N.C., Juul Jensen, D. & Conradsen, K. ~1994!. Wheeler, J., Jiang, Z., Prior, D.J., Tullis, J., Drury, M.R. &
Automatic recognition of deformed and recrystallized regions Trimby, P.W. ~2003!. From geometry to dynamics of microstruc-
in partly recrystallized samples using electron backscattering ture: Using boundary lengths to quantify boundary misorienta-
patterns. Mater Sci Forum 157–162, 149–158. tions and anisotropy. Tectonophysics 376, 19–35.
Kunze, K., Wright, S.I., Adams, B.L. & Dingley, D.J. ~1993!. Wilkinson, A.J. ~1997!. Methods for determining elastic strains
Advances in automatic EBSP single orientation measurements. from electron backscatter diffraction and electron channeling
Text Microstruc 20~1–4!, 41–54. patterns. Mater Sci Tech 13~1!, 79–84.
Kysar, J.W. & Briant, C.L. ~2002!. Crack tip deformation fields in Wilkinson, A.J., Meaden, G. & Dingley, D.J. ~2006a!. High-
ductile single crystals. Acta Mater 50~9!, 2367–2380. resolution elastic strain measurement from electron backscatter
Lehockey, E.M., Lin, Y.-P. & Lepik, O.E. ~2000!. Mapping residual diffraction patterns: New levels of sensitivity. Ultramicroscopy
plastic strain in materials using electron backscatter diffraction. 106~4–5!, 307–313.
In Electron Backscatter Diffraction in Materials Science, Schwartz, Wilkinson, A.J., Meaden, G. & Dingley, D.J. ~2006b!. High
A.J., Kumar, M. & Adams, B.L. ~Eds.!, pp. 247–264. New York: resolution mapping of strains and rotations using electron
Kluwer Academic/Plenum Publishers. backscatter diffraction. Mater Sci Tech 22~11!, 1271–1278.
Li, B.L., Godfrey, A. & Liu, Q. ~2002!. Investigation of macro- Winkelmann, A. ~2008!. Dynamical effects of anisotropic inelastic
scopic grain sub-division of an IF-steel during cold-rolling. scattering in electron backscatter diffraction. Ultramicroscopy
Mater Sci Forum 408–412, 1185–1190. 108, 1546–1550.
Miyamoto, G., Shibata, A., Maki, T. & Furuhara, T. ~2009!. Wright, S.I. ~1993!. A review of automated orientation imaging
Precise measurement of strain accommodation in austenite microscopy ~OIM!. J Comput Assist Microsc 5, 207–221.
matrix surrounding martensite in ferrous alloys by electron Wright, S.I. ~1999!. Quantification of recrystallized fraction from
backscatter diffraction analysis. Acta Mater 57~4!, 1120–1131. orientation imaging scans. In Proceedings of the Twelfth Inter-
Nowell, M.M. & Wright, S.I. ~2005!. Orientation effects on national Conference on Textures of Materials, Szpunar, J.A. ~Ed.!,
indexing of electron backscatter diffraction patterns. Ultramicros- pp. 104–109. Ottawa, Ontario, Canada: NRC Research Press.
copy 103~1!, 41–58. Wright, S.I. ~2006!. Random thoughts on non-random misorien-
Nye, J.F. ~1957!. Physical Properties of Crystals. Their Representation tation distributions. Mater Sci Tech 22~11!, 1287–1296.
by Tensors and Matrices. London: Oxford. Wright, S.I., Adams, B.L. & Kunze, K. ~1993!. Application of new
Pantleon, W. ~2008!. Resolving the geometrically necessary dislo- automatic lattice orientation measurement technique to poly-
cation content by conventional electron backscattering diffrac- crystalline aluminum. Mater Sci Eng A 160, 229–240.
tion. Scripta Mater 58, 994–997. Wright, S.I., Field, D.P. & Nowell, M.M. ~2005!. Impact of local
Sun, S., Adams, B.L. & King, W.E. ~2000!. Observations of lattice texture on recrystallization and grain growth via in-situ EBSD.
curvature near the interface of a deformed aluminum bicrystal. In Textures of Materials—ICOTOM 14, Van Houtte, P. & Kes-
Philos Mag 80, 9–25. tens, L. ~Eds.!, pp. 1121–1130. Leuven, Belgium: Trans Tech
Tao, X. ~2003!. An EBSD study on mapping of small orientation Publications.
differences in lattice mismatched heterostructures. PhD Thesis. Wright, S.I. & Nowell, M.M. ~2006!. EBSD image quality map-
Bethlehem, PA: Lehigh University. ping. Microsc Microanal 12, 72–84.

Das könnte Ihnen auch gefallen