Вы находитесь на странице: 1из 11

Strains, planes, and EBSD

in materials science
Electron back scatter diffraction (EBSD) has made an impressive impact
on the characterization of materials by directly linking microstructure and
crystallographic texture to provide very rich and quantitative datasets which
in many instances have forced us to rethink how microstructure should be
defined and analyzed. In this article we try to first give a very basic idea of
how an EBSD map is obtained and what the data produced is like. We then
give a brief history detailing some of the more major steps in developing the
technique to what it is today. Finally, we explore two advanced and exciting
technique areas of strain mapping and 3D microscopy and demonstrate
how the EBSD technique continues to evolve to tackle new applications and
bolster our materials characterization toolbox.
Angus J. Wilkinson* and T. Ben. Britton
Department of Materials, University of Oxford, Parks Road, Oxford, OX1 3PH, UK
*E-mail: angus.wilkinson@materials.ox.ac.uk

Electron backscatter diffraction (EBSD) allows crystallographic is generally highly tilted (60 to 80) toward the detector, as is
information to be obtained from small volumes of material in a depicted schematically in Fig. 1, to increase the quality of the pattern
scanning electron microscope (SEM) which provides versatility in obtained. Each pattern consists of many bands of raised intensity
mapping orientation, crystal type, and perfection over a wide range that span across the screen. The bands, termed Kikuchi bands, appear
of step sizes making it a powerful microstructural characterization at first glance to have parallel straight edges but in fact the edges are
tool. EBSD maps are formed by moving a focused probe of electrons slightly curved and are the hyperbola formed by the intersection of
point by point across a grid of positions on the surface of a bulk Kossel diffraction cones with the plane of the detector. An example
sample in a scanning electron microscope. At each point, some of pattern from a CCD based detector is shown and compared to
the electrons backscattered from the sample are collected by a examples from earlier technologies and simulations in Fig. 2. Most
detector comprising a scintillator screen coupled generally by a importantly the centers of the bands are straight lines and mark
lens, but sometimes by a fiber optic bundle, to a photon sensitive the projections onto the detector screen from the point-like volume
imaging detector, typically a charge coupled device (CCD) camera, illuminated by the incident electron beam of the diffracting planes
to form an electron backscatter diffraction pattern. The scintillator within the crystal.
is generally 30 to 40 mm in size and held close enough to the sample Patterns are transferred from camera to the computer for indexing
so as to subtend a relatively large angle (60 to 70). The sample and determination of crystal orientation, as detailed in Fig. 3. Briefly, the

3 66 SEPTEMBER 2012 | VOLUME 15 | NUMBER 9 ISSN:1369 7021 Elsevier Ltd 2012. Open access under CC BY-NC-ND license.
Strains, planes, and EBSD in materials science REVIEW

Fig. 1 Schematic diagram showing the experimental set-up for EBSD observations.

Hough transform algorithm is used to convert the nearly straight bands the axes of the microscope and EBSD detector. This may seem obvious
from lines to points which can be more easily located. With knowledge but it is sometimes overlooked or rather poorly performed in the rush to
of the experimental geometry, the peak locations can be converted to a get on to rather more interesting parts of the operation.
table of interplanar angles and compared with look up tables of expected In practice, there is no one way of correctly setting up an EBSD scan
angles for the phases present within the sample. as the parameters used depend on many competing factors including the
With modern systems, pattern capture and indexing can be performed nature of the sample and the information that is required. For instance, if
very rapidly (many 100s of patterns per second) and when combined the main aim is to measure the overall crystal texture with little regard
with precise scanning of the electron beam, a detailed mapping of the for the grain size and morphology, then it is sensible to use a relatively
sample crystallography (crystal type, crystal orientation, pattern quality coarse step size so as to quickly asses many grains within the sample. In
etc.) can be obtained. These maps provide rich measurements of the other situations such as an analysis of the path taken by a fatigue crack
sample microstructure. and assessing damage in the surrounding lattice, then it would be sensible
Fig. 4 outlines the major stages in the process of characterizing a to choose a smaller step size and a slower and more accurate use of the
sample using EBSD. No technique provides suitable information without pattern capture and peak detection algorithms (i.e., increased resolution
adequate specimen preparation. For EBSD, standard metallographic in the Hough transform and longer exposure times) to realize an increase
processes of sectioning, grinding and polishing are the starting point but sensitivity in the measurement of intragranular misorientation.
additional care must be taken to ensure that the final polishing steps For samples which are entirely unknown at the outset, it is generally
leave the sample surface free of any sectioning damage, as the EBSD useful to run a rather quick map with widely spaced measurement
pattern is generated from a very thin surface layer (~40 nm) of material. points, or a line scan, to get an estimate of the grain size and to visualize
Polishing with colloidal silica, chemical etching, electro-polishing, ion the microstructure morphology before deciding on settings for more
milling, and plasma etching are all methods that have been shown detailed work.
to work in various material systems 1-3. If the sample is etched an For all these applications, assessment of how representative a
assessment of microstructure uniformity using an optical microscope given dataset might be should be made. For example if the grain size
is often time well spent. Once the sample is in the SEM chamber, it is distribution is known to be unimodal and the average grain size is the only
sensible to have a quick manual interactive point and click survey of the requirement then we agree with Humphreys4 and Randles5 guidance
sample to assess the quality of patterns being obtained, as at this point that a minimum of 200 grains should be sampled, with at least ~10
if the patterns are of poor quality, the preparation steps may need to be points across a grain, i.e., 100 points per grain so that a relatively modest
revisited. Secondary and/or back scattered electron imaging can be used map of 20 000 points results, with the step size set at approximately
to guide selection of the initial area and step size to use. Furthermore, if a tenth of a first guess at the grain size. If the grain size distribution is
the measured crystal orientations need to be related to some external required, rather than just the average, then a similar but slightly more
macroscopic direction (e.g., extrusion direction, rolling direction, growth elaborate assessment of the size of the required dataset can and should
direction) then care needs to be taken to align the sample correctly with be made. Similar assessment of the size of datasets required to establish

SEPTEMBER 2012 | VOLUME 15 | NUMBER 9 367


REVIEW Strains, planes, and EBSD in materials science

(d) and micro texture (separating alpha and beta phases automatically), beta
(a) (b)
grain size, lath morphology and intergranular misorientations. Two very
useful reviews of strategies for EBSD analysis and quantitative analysis of
microstructure have been given by Randle5 and Humphreys4 respectively.

A brief history of EBSD


In 1954 Alam, Pashley and Nicholson6 recorded onto EM film patterns
which we now call EBSD patterns (see Fig. 2a) and observed that the
pattern contrast was greatest when the sample was inclined so that
(c) (e) the beam was incident within ~20 30 of the surface plane. They also
noted that the part of the pattern at low take-off angles gave the highest
contrast which was at best 15 %. Todays EBSD experiments adopt
geometries that are in accord with these earliest observations with 20
between the sample surface and incident beam routinely used.
In the 1970s Venables and co-workers7-9 implemented the technique
on a VG STEM operated in conditions more usual for an SEM and recording
patterns on film either directly exposed to electrons within the chamber
or using an externally mounted film camera coupled to a scintillator
Fig. 2 Electron backscatter patterns: (a) High angle diffraction pattern from lead screen within the microscope chamber. Significantly they showed how
sulphide captured by Alam et al.6 on film, (b) picture of monitor showing pattern
the patterns could be used to measure the orientations of micro-crystals
obtained using SIT camera with BBC micro-computer graphics overlay indicating
indexing15, (c) EBSD obtained using a current day CCD-based EBSD detector, (d) selected with the scanned beam in the SEM and gave a detailed error
kinematic and (e) dynamical simulations of EBSD patterns from Si101. analysis estimating the orientation accuracy to be 0.5 to 1 very similar
to todays measurements8.
Dingley who had been working on micro-Kossel diffraction10-12 picked
sub-grain size, texture, phase fractions in multiphase materials can also up the EBSD technique and pushed it forward by using a low light level
guide quantitative measurement of microstructure with EBSD5, just as it ISIT camera to view the scintillator and overlaying the video output from
does with other techniques. this with graphics from a BBC micro-computer allowing a significant
An example of the information available from an EBSD dataset is increase in the rate of pattern indexing and orientation measurements13-15.
shown in Fig. 5, in this case from a Ti-15wt%Mo alloy. The first map An example pattern from this type of set up is shown in Fig. 2b. This is one
(Fig. 5a) distinguishes regions of the high temperature retained bcc of the very earliest attempts at integrating capture of video information
phase in blue, from the low temperature hcp phase in red. We find that at a microscope to enable quantitative on-line analysis. Pattern indexing
~20 % of the phase has transformed to . Some information concerning methods were improved so that minimal user input of only 3 zone axes
the crystal orientations is revealed in Fig. 5b in which the color of every positions were required to index and calculate orientations for patterns
point depends on the crystal pole along the surface normal direction, from cubic crystals14, non-cubic crystals16 and full automation of the
through the color-key provided for each phase. This makes it clear that analysis set as a clear goal very early on. The main materials science
there are several grains within the field of view and within a particular challenge EBSD was aimed at in these early days was in grain boundary
grain there are one or more colonies of lathes with similar orientations. characterization, a theme that continues to this day, and in particular
Here mapping of the orientation data makes it clear that changes in identifying any significant special properties associated with low CSL
lathe crystallography are linked to changes in the alignment of the boundaries, e.g.17-19. Other topics that were explored were possibilities for
long axis of the lathe (noting that we are looking at a 2D cross section phase identification including combining EBSD with local chemistry from
from a 3D microstructure). We can also see that phase has formed on EDX20. A further notable study by Baba-Kishi and Dingley21-23 attempted
many of the grain boundaries. Looking in more detail at the crystal to distinguish between different space and point groups from symmetry
orientations of the in relation to the parent grains we can see that the elements and systematic reflections in EBSD, again a topic that has
Burgers orientation relationship is obeyed: been returned to more recently with much greater image analysis and
computing power available now24.
The next big advance in the method came with automation of the
band detection step which finally allowed the whole process of pattern
collection and indexing to be undertaken without user intervention.
By adding in computer control of the SEM beam position it was then
This type of information can be obtained very quickly using EBSD and a relatively simple task to implement EBSD orientation mapping. The
commercial EBSD software packages provide many of the analysis tools Yale group of Adams, Wright, and Dingley (on sabbatical from Bristol)25,26
that are required. Repeating this process for a larger area with a finer step first implemented a Burns algorithm for band detection but moved to
size could be used to examine quantitative metrics such as sample macro the modified Hough transform which the Ris group (Krieger Lassen, Juul

368 SEPTEMBER 2012 | VOLUME 15 | NUMBER 9


Strains, planes, and EBSD in materials science REVIEW

Fig. 3 Overview of EBSD indexing procedure showing pattern capture through to determination of crystal orientation.

Jensen and Conradsen) had shown to be successful27-29. Both groups had relationships between layers on substrates in metal, semiconductor
successfully implemented automated EBSD mapping by the early 1990s. and superconductor systems, characterizing texture and its changes
Goehner and Micheal30 showed that phase discrimination is possible during deformation and annealing in metal alloys and geological
by comparing measured interplanar angles to those expected from the samples, establishing links between grain size and texture components
known crystallography for each of a wide range of possible phases held in during deformation and annealing, contributing to determination of
a database. The problem to be overcome here is in reducing the number grain boundary energy through thermal grooving, and ex situ and in
of possible phases to a manageable level either through knowledge of situ experiments on grain boundary mobility. Although long, this list
local chemistry or by the primitive lattice volume. is not exhaustive and so we have decided to explore in detail only the
following two aspects of strain measurement and three dimensional
Application areas characterization. We have selected these aspects as we feel they are
EBSD has now been used for many aspects of materials characterization currently undergoing considerable development (and of course are of
including characterization of grain boundary types, establishing epitaxial significant interest to the authors).

SEPTEMBER 2012 | VOLUME 15 | NUMBER 9 369


REVIEW Strains, planes, and EBSD in materials science

Strain measurement These pattern shift and their systematic variation across the screen can
Early work with EBSD concentrated on annealed structures however be related back to the magnitude and nature of lattice strain variation
there has always been interest in applying EBSD to deformation studies. and lattice rotation relative to the reference pattern51-55,60,67-69. The strain
Initial attempts followed earlier work with ECP patterns, e.g.31-34 in sensitivity of ~10-4 can be thought of as measurement of shifts to a few
using the blurring of the EBSD patterns to assess the level of cold work. hundredths of a pixel compared to sample to screen distances of a few
These methods are not as well established as for peak broadening of hundred pixels. Two of the most recent improvements in the method
XRD and/or neutron diffraction and results have not be quantified in have been firstly to increase the number of ROIs used in the analysis and
terms of dislocation densities but instead calibration samples are used move to an robust iterative fitting analysis to reduce the effects of outlier
to read across pattern blurring parameters to equivalent plastic strains. shift measurements on the resulting strain and rotation analysis67, and
Applications were to superalloys in creep35, Al36 and Al-SiC MMCs37, 38, secondly to use a remapping of the test pattern to remove the larger
and distinguishing between deformed and recrystallized grains29. Since lattice rotations which cause errors in the smaller elastic strains68,69.
automated EBSD mapping has been available pattern quality has not The most significant remaining challenge for strain mapping is the
been used so much for deformation studies in any quantitative sense. so-called reference pattern problem. In many circumstances there is a
Instead plastic deformation studies have concentrated on use of location on the sample from which a strain free reference pattern can be
intragranular misorientations which result from the residual dislocations obtained with the crystal in the required orientation (e.g., far field from
accumulated during plastic straining. Various empirical metrics have the deformed area of interest). However there are many more where this
been suggested and again in some cases calibration samples have been is not the case and so the measured strains represent variations from
used to read across from these parameters to equivalent plastic strain. an unknown strain state at the position at which the reference pattern
Methods exist to estimate a lower bound on the dislocation density either was obtained. One route to overcoming this is to use simulated patterns
by a physically motivated minimization of dislocation line energy or a for the reference as suggested by Kacher et al.57, although many of the
computationally convenient minimization of the square of the densities specifics in the approach used by Kacher et al. have been criticized mainly
of the different dislocation types39,40. Being physically based this approach due to the lack of fidelity in the simple simulations used and the lack of
is also being used by the X-ray synchrotron41,42 and extensively within the certainty over the detector geometry and aberrations70, 71.
micromechanical modeling43-45 communities. Grand challenges within the An alternative to cross-correlation analysis, that does not rely on a
materials science and engineering communities will benefit from the use of reference pattern, has been suggested by Maurice and Fortunier72 who
a common physically based framework such as the Nye tensor to compare use the 3D-Hough transform as a means of increasing the precision of
and combine results across these differing and complementary techniques. band center positions, and thus interplanar angles, compared to the
Accurate analysis of elastic strain, or lattice parameters, from standard Hough transform. Assessment of simulated patterns suggested
measurement of Kikuchi band edge positions though Braggs law has a sensitivity of 2 10-4, but factors such as noise and asymmetry of
largely proved unsuccessful, despite some success with ECP46, primarily the Kikuchi band profiles are likely to significantly increase errors for
because Kikuchi lines of sufficiently high order are not present47,48. Goehner measurements on real patterns which have yet to be reported.
and Micheal49 have explored the use of HOLZ rings for lattice parameter Applications of the cross-correlation analysis have been to both
determination but found limited sensitivity for strain measurement functional and structural materials. Early measurements were reported
and these rings are quickly removed by plastic deformation and so the for SiGe alloys grown as unrelaxed blanket films on Si substrates in plan
method would not be very general. view48, 51, 52 and in cross-section54, 55. Work on SiGe/Si systems patterned
The most successful form of elastic strain measurement to date has with mesas has provided good agreement between EBSD measurements
been performed by measuring changes in interplanar angles and therefore and elasticity theory60, 73, 74. Further validation of the EBSD method has
determination of the deviatoric components of the elastic strain tensor. been provided by the group at NIST who have compared EBSD results with
This can be performed using cross-correlation functions to determine micro-Raman65 and AFM66 measurements near wedge indents in Si. They
small shifts in the patterns compared to a reference pattern obtained assessed the stress sensitivity of the EBSD measurements to be ~10 MPa.
from the crystal in a known strain state. The approach can be found in the Good agreement between EBSD and micro-Raman has also been given
early work of Troost et al.50 and Wilkinson48,51,52. The method was improved by Tomita et al.64 for strained Si on insulator and SiN/Si systems. Ishido
notably by Wilkinson Meaden and Dingley53-56 with the geometry altered et al.62 have examined cross-sections of GaN structures including a 5 nm
from an initial cumbersome long camera length-low capture angle set- AlN/25 nm GaN multilayer for which they give a line scan within which
up to the more conventional large capture angle set-up while retaining the 50 repeats of the multilayer can be distinguished indicating that good
a sensitivity of ~10-4. Several other groups have now implemented this spatial resolution can be achieved. Speller et al.75 have recently widened
method or modifications of it57-61 or are using a commercial version62-66 *. the scope of cross-correlation EBSD studies by using it to map changes in
In this approach one pattern is used as a reference and small shifts c/a ratio with chemistry obtained from EDX in the iron containing FeySe1-
are measured for a number of regions of interest (ROI) dispersed across x
Tex superconductor. Applications to structural metallics began with single
the wide capture angle of the EBSD pattern. In common with various crystals including fatigue crack tips in superalloys54, and tensile cracks in
other image analysis applications the shifts are determined automatically tungsten76 or single grains of polycrystals including mapping strains and
using cross-correlation analysis and sensitivities of a few hundredths of dislocation densities near indents in Fe40 and Ti45, phase transformation
a pixel can be obtained, although this depends on the pattern quality. induced strains in martensitic steels58,77 and thermally and mechanically

370 SEPTEMBER 2012 | VOLUME 15 | NUMBER 9


Strains, planes, and EBSD in materials science REVIEW

Fig. 4 Flow chart indicating significant steps in a typical EBSD observation.

induced strain localization near hard inclusions in a superalloy78,79. The near the indent measured by conventional Hough transform analysis
method has also been applied to polycrystalline Ti-6Al-4V deformed in along with a wireframe showing the crystal orientation. The indent was
uniaxial tension at room temperature80 and by hot rolling81 in studies made with a load of 50 gF, which was large enough to generate radial
aimed at determining dislocation densities using the Nye tensor. Ojima cracks which radiate out from the center along the indent diagonals
et al.63 have recently reported the first in situ deformation of a polycrystal which are aligned with the [110] and [1-10] direction corresponding to
coupled with the cross-correlation EBSD method which successfully horizontal x1 and vertical x2 axes. EBSD measurements were made on a
demonstrates that grains with <100> along the tensile axis tend to JEOL JSM 6500F FEG-SEM at a beam energy of 15 keV, a beam current of
continue increasing in elastic strain beyond the applied stress level at ~16 nA and a sample tilt of 70. EBSD patterns were recorded at the full
which <110> aligned grains tend to show a saturated elastic strain. These ~1k 1k pixel 12 bit deep resolution of the peltier cooled CCD camera
generalities are in agreement with volume averaged behavior seen for the and saved to hard disk for batch-wise off-line analysis using CrossCourt3
same alloy using neutron diffraction, but the EBSD shows significant grain software. The data acquisition rate was approximately 1 pattern/sec with
to grain variation which presumably depends on the local neighborhood. the camera gain kept low so as to achieve good signal to noise and the
To illustrate the data available from the cross-correlation based analysis 50 50 m area was mapped at a step size of 1 m in approximately
we will examine the elastic strains and rotations measured near a Vickers half an hour.
hardness indent in (001) Si. Fig. 6 (upper left) shows the misorientations Two data quality parameters are used to assess the dataset (lower left

SEPTEMBER 2012 | VOLUME 15 | NUMBER 9 371


REVIEW Strains, planes, and EBSD in materials science

(a)

(b)

Fig. 5 Example EBSD data obtained from Ti-15%Mo. Inserts show raw diffraction patterns which are successfully indexed to determine crystal orientation and phase (a).
The resultant data set can be separated into maps showing alpha and beta phase orientations (b) shown here with respect to the sample normal. In this alloy, the Burgers
orientation relationship is seen easily, as shown for the colony/parent beta grain insert, which relates the basal plane (0001) and an <11-20> type lattice direction in the
hexagonal phase, with a {110} plane and a <111> type lattice direction in the beta phase.

of Fig. 6). The geometric mean of the normalized cross-correlation peak are to the left and right while the tensile strains occur above and below
heights shows low values for regions within the indent due to shadowing the indent. This is in accord with expected compressive radial and tensile
effects from the topography of the sample and pattern blurring from the hoop strains that should develop. The in plane 12 shear strains are also
extensive deformation. The mean angular error describes the difference consistent with this and show clear positive and negative lobes at 45
between measured pattern shifts and back calculations of pattern shifts to the horizontal. The lattice rotation 23 about the horizontal axis shows
based upon the best fit solution for the strain and rotation tensor. There is positive and negative rotations above and below the indent that result
generally strong correlation between regions with low cross-correlation from the uplift of material surrounding the indent, to account for the
peak height and large mean angular errors. Thresholds can be used to materials volume displaced from the indent impression itself. A similar
remove the poorer quality data from subsequent analysis. In this instance effect is seen for the 31 rotation about the vertical x2 axis. Abrupt changes
patterns generating a cross-correlation peak height of less than 0.3 or a in lattice rotation are accommodated by opening displacements caused
mean angular error of greater than 10-3 rads were removed from other by the radial cracks; this is most clearly demonstrated in 12 rotations
maps. All components of the strain and lattice rotations are determined, (about the surface normal) which are generally of smaller magnitude.
and these are shown in pictorial tensor layout in Fig. 6. The lower parts This is expected for mode I cracks loaded in tension and has also been
of the tensors are not shown as the strain tensor is symmetric (i.e., seen for cracks in single crystal tungsten76. This relatively simple test case
Ij
= ji mirror across lead diagonal) and the rotation tensor is antisymmetric gives an indication of the completeness of the quantitative data available
(i.e., ij= - ji mirror across lead diagonal and swap sign). from EBSD strain mapping.
The 11 strain along the horizontal axis is seen to be compressive above
and below the indent and tensile to the left and right with the magnitude 3D microscopy
of the strains falling significantly with distance from the center of the The development of automated data collection in three dimensions using
indent. For the 22 strain this is swapped so that the compressive regions EBSD has opened up an exciting field of materials characterization. Two

372 SEPTEMBER 2012 | VOLUME 15 | NUMBER 9


Strains, planes, and EBSD in materials science REVIEW

(a)

(b)

(c)

Fig. 6 Information obtained from a high angular resolution EBSD map of a 50gF micro hardness indent in single crystal silicon. Upper left shows the crystal misorientations
with an overlay indicating the how the crystal is oriented. Lower left gives the two data quality parameters: mean angular error (lower is better) and cross-correlation
peak height (normalized to between 0 and 1 with 1 best). To the right are shown the elastic strain (top) and lattice rotation (bottom) tensors. These are displayed as
pictorial tensors with a map shown for each non-zero component in the upper part of the tensor with the symmetry operator needed to populate the lower part also
indicated.

dimensional information obtained from surface sections of materials, major steps: data acquisition; model generation; and quantitative
can be descriptive for an elegantly designed experiment. However, analysis.
in many cases additional information regarding the structure and Data acquisition is performed by successively removing layers from
connectivity of microstructural units in the third dimension would add a block of material and observation of the newly revealed surface. Any
significant value by reducing our reliance upon assumptions required preparation route that produces a surface suitable for the successful
to extend our observations to describe microstructures completely. generation of high quality diffraction patterns can be used. Mechanical
Data obtained with 3D EBSD can be used to populate models with polishing and focused ion beam machining are the two leading
realistic microstructural information, for example with finite element routes.
or fast Fourier transform based crystal plasticity modeling82-84. Mechanical routes utilize readily available equipment but require
For these data sets, the advantage of EBSD over other methods of many man hours to observe multiple slices. Material removal rates are
observation (e.g., optical or secondary electron image acquisition) is high, which limits the minimum slice thickness (>1 m) but potentially
that crystalline materials can easily be segmented and characterized allows sampling of large volumes of material. In addition, precise
using orientation information. In many cases this provides greater calibration of the thickness of the layer removed is difficult and can
contrast than simply relying on electron (or ion beam) channeling introduce uncertainty into the data reconstruction step. Time can be
contrast or the use of etchants to reveal microstructural features. In saved if orientation information is not required for every slice, instead
addition, orientation information combined with simple rules regarding recording optical micrographs of an etched surface for every slice and
microstructural connectivity can be utilized to improve the fidelity of orientation maps for every nth slice can be used86. Uchic and co-workers
the 3D reconstruction. Furthermore, the EBSD analysis can easily be are currently developing a system which integrates mechanical
extended using energy dispersive spectroscopy85. polishing, SEM operation and EBSD acquisition with robotic interchange
Production and successful exploitation of 3D EBSD involves three to aid automation87. Such a system should make data acquisition using

SEPTEMBER 2012 | VOLUME 15 | NUMBER 9 373


REVIEW Strains, planes, and EBSD in materials science

mechanical routes significantly less demanding and greatly increase frequency is necessary, thus sampling a larger volume of material, which
the quality and size of the available 3D datasets. may ensure that a representative volume element is sampled.
In contrast, dual beam SEM-focused ion beam (FIB) instruments Once several slices have been obtained, reconstruction of the 3D
are expensive but recent developments in automation and instrument model must be performed. Firstly, the data that has been acquired will
stability have made data acquisition much simpler. Data is acquired not be a perfect representation of the sampled material: successive
by moving the sample into a position suitable for EBSD observation, removal will not consistently reproduce rectangular sections (due to a
scanning the top surface, moving the sample into a position suitable misalignment during the mill step); there will be optical distortions in
for milling and removing a thin layer. This is repeated for as many slices the electron microscope; each slice will be misaligned with respect to its
as the user requires (or time allows). Slices as thin as ~50 nm can be neighbors; there will be mis-indexed points; and finally, we do not have a
successively removed, making the analysis of submicron grain sizes continuous description of the microstructure (i.e., approximating curved
possible88. lines on grain boundaries as a staircase).
However, the sampling volume attainable by this route is significantly Careful data acquisition will reduce the magnitude of the sample
smaller due to limited material removal rates. Furthermore, as the EBSD mis-alignment and optical distortions (yet they will always be present).
camera capture angle is relatively large, the volume of material that Generally the use of a suitable fiducial marker of a known shape (i.e.,
must be removed for each slice typically must be larger than the area a milled circle or cross for FIB polishing or a series of hardness indents
scanned (or else edges of the pattern can become shadowed). A specimen for mechanical polishing) can be combined with image correlation
geometry used is a mechanically polished sharp right angle. This can be to reposition the sample well between each slice. In addition, these
used either to cross section a bulk material or removal of material from a features could be imaged to produce a distortion free micrograph (e.g.,
standard cross section. For example, Zaafarani and co-workers used this using a travelling optical microscope or normal incidence SEM) to assess
geometry to sample the deformation flow fields beneath nanoindention the presence of optical distortions91. Subsequently a suitable correction
impressions in copper revealing the full 3D character of the deformation could be applied, e.g. using an affine transformation86, to ensure that the
field, which compared well with their finite element prediction 89. appropriate data cube is reconstructed.
Alternatively, if significantly large volumes are required, it may be more Physical alignment of the fiducial markers results in close alignment
efficient to manufacture a protruding finger-like specimen, which can of each successive slice. Typically, alignment of the slices prior
be gradually shortened to remove each slice87, 90. This geometry reduces to reconstruction of the data cube is still required. A high quality
EBSD pattern shadowing, waterfall and re-deposition artifacts as each reconstruction can be performed by aligning not only microstructure
free surface remains as a protrusion. features, such as grain boundary triple points, but also local crystal
Knowledge of the microstructure and the eventual use for the misorientations between voxels in neighboring slices. One strategy
data obtained is essential to make decisions regarding slice thickness suggested by Groeber and colleagues, is to allow iterative realignment
and EBSD step size. For both of these acquisition routes, the choice of the slices in the form of translations in x and y (but in theory it could
of slice thickness is determined at the lower end by the limits of the also include any potential misalignment such as sample tilt) in order
sectioning technique and at the upper end by the speed (and accuracy) to reduce the total misorientation in the cube86. Clearly, care must be
of the material removal process. The choice of EBSD step size is largely taken in dealing with poorly indexed points and grain boundary regions.
dependent on the data acquisition speed available; here the use of a Correction of poorly indexed points can be performed in a similar fashion
fast camera capturing several hundred patterns per second can be to convention 2D clean up routines, but now may include information
desirable, either enabling larger areas to be scanned or a finer step size gained from voxels in neighboring slices.
to be chosen. At this point we have a data stack that consists of voxels which are
The combination of step size and slice thickness defines the size and well alignment spatially with respect to each other but they may not
shape of the unit voxel (volume pixel) in the final map. The area scanned yet represent the microstructure as fully as our interrogation requires.
combined with the thickness of each slice and the number of slices For example if the morphology of the individual grain boundary
viewed defines the volume of material sampled. planes is important, then the apparent stepped nature of the interface
Choice of voxel size and shape and the volume of material sampled inherited from the discrete sampling strategy must be addressed. For
is dependent on the nature of the experiment performed. In general, this application, neighboring voxels of significantly different orientation
this is a combination of both the amount of instrument time available, could be segmented from the data stack. Subsequently for each neighbor
the fidelity of the data required, the nature of the problem studied and voxel a marching cubes algorithm can be utilized to transform the
the amount of effort required. For example, in order to analyze the rectangular faceted voxel into a series of triangular facetted polygons92.
nature of grain boundary planes within a material it is necessary to Finally additional smoothing of the microstructure can be performed as
obtain many slices within each grain in order to accurately trace the required.
grain boundary plane region (as a general rule of thumb Groeber et al. One example of 3D-EBSD obtained from FIB-SEM tomography is
suggest at least 10 slices per grain86 which is similar to the suggestion of shown in Fig. 7, courtesy of Groeber and co-workers (described in detail
Humphreys et al. for 2D EBSD noted previously4). Yet, if the distribution elsewhere87, 90, 93-95). Here a polycrystalline nickel superalloy sample has been
of grain orientations in a 3D volume or the generation of a mesh for examined with FIB-SEM tomography and 3D EBSD with voxel size of 250
finite element modeling is required, then a significantly lower sampling 250 250 nm3. Fig. 7a shows a reconstruction of the 3D data cube after

374 SEPTEMBER 2012 | VOLUME 15 | NUMBER 9


Strains, planes, and EBSD in materials science REVIEW

destructive and thus in addition to sampling larger volumes of material, in


situ dynamical information can be obtained.

Summary and outlook


EBSD has come from being a very specialist technique being used in a small
number of physics laboratories to a mainstream materials characterization
tool. Potential of the technique to solve real materials problems was
recognized near the outset and these end goals have always driven its
Fig. 7 Three dimensional reconstructions of microstructure obtained from a nickel development. In this field, technique development had been strongly
superalloy (IN100): (left) after alignment and application of marching cubes; associated with two aims: (i) looking for general solutions so that the tool
(middle) after grain boundary smoothing; (right) selected grains from the 3D
can be rolled out easily from a specific initial problem to wide range of
model. The material volume analyzed measured 50 m by 50 m by 50 m, with
EBSD maps acquired after 250 nm FIB slices. Figure courtesy of Groeber, Rollett related ones and (ii) automation so the statistically significant datasets
and colleagues86, 90. can be obtained for quantitative assessment of material microstructure.
This has led to the powerful tool we have today with a vast range of
applications which continues to expand.
We have picked strain mapping and 3D microscopy as two areas which
initial alignment of consecutive slices, allowing for small shifts in the EBSD demonstrate that technique development is currently very active and to
maps for each slice to minimize total misorientation between neighbors, demonstrate the pace at which substantial advances have been achieved.
and then application of a marching cubes algorithm to render triangular In addition, we note that the work of Winkelmann on development of
facets. In order to analyze the nature of grain boundary planes, the grain dynamical diffraction simulations99-101 and detailed analysis of the energy
boundaries have been smoothed to reduce alias induced roughening distributions contributing to Kikuchi bands within EBSD patterns102 is another
caused by discreet nature of voxel sampling. Qualitative analysis via on-going major contribution. Work by Geiss et al.103 on transmission-EBSD
rendering of particular grains in the volume showing Fig. 7c hints at some also contributes to a better understanding of the physics of generating
of the quantitative metrics that can be explored with in such a data cube, EBSD patterns. We hope that these advances may develop into methods
such as grain volumes, misorientation and other metrics which have two of crystal structure determination direct from EBSD patterns.
dimensional counterparts90. Data of this type can be used to measure five We may also see some developments in detector technology over
parameter grain boundary plane distributions, which includes not only the next decade. We already see two competing trends in EBSD either
four parameters from standard 2D EBSD, grain boundary misorientation attempting to go faster which has driven down pattern resolution, or to
(axis 2, angle 1) and surface trace (1) but in addition the subsurface obtain as much detail in the pattern as possible which drives down speed.
grain boundary plane inclination94. More sensitive detectors would be of benefit in both cases. It seems very
While FIB-SEM or mechanical section routes of 3D EBSD provides a likely new detector strategies being developed for X-ray and TEM may also
wealth of data on volumes of material at an intermediate length scale, lead to improved EBSD detectors.
resolution of the technique is limited by the probe size (~20 nm) and EBSD has come a long way but the journey is far from over.
slicing and reconstruction ability at smaller length scales. Analysis of truly
nanocrystalline materials of very fine detail in microstructures is therefore Acknowledgements
difficult and TEM experiments on these materials may be more suitable96 We gratefully acknowledge funding from EPSRC (EP/E044778/1, EP/
and are certainly complementary. G004676/1, and EP/H018921/1). We have enjoyed continuing discussions
At larger length scales the total analyzed volume of material and the of the EBSD technique with Prof David J Dingley (University of Bristol) and Dr
sampling resolution is limited primarily by experiment time and instrument Graham Meaden (BLG Productions Ltd). We are grateful to Prof Tony Rollett
stability. Recent advancements in 3D X-ray diffraction (XRD) microscopy (Carnegie-Mellon University) and Dr Michael Groeber (Wright-Patterson
provide similar information non-destructively with limited resolution97,98. Air Force Base) for sharing and discussing 3D EBSD data with us. We also
One advantage with these techniques is that although 3D-XRD thank Dr Aimo Winkellman for his insight into the physics of EBSD pattern
experiments are difficult to perform, the technique is fundamentally non- formation and simulation code used for Fig. 2e.

References
* CrossCourt 3, BLG Productions Ltd, Bristol, UK (www.blgproductions .co.uk) 3. Wynick, G. L. and Boehlert, C. J., Materials Characterization (2005) 55, 190-202.
1. Koll, L., et al., Journal of Microscopy (2011) 243, 206-219. 4. Humphreys, F. J., Journal of Materials Science (2001) 36, 3833-3854.
2. Schwarzer, R. A., Proceedings of the 10th International Conference on Textures of 5. Randle, V., Materials Characterization (2009) 60, 913-922.
Materials 157, 201-206. 6. Alam M. N., et al., Proceedings of the Royal Society A (1954) 221, 224-242.

SEPTEMBER 2012 | VOLUME 15 | NUMBER 9 375


REVIEW Strains, planes, and EBSD in materials science

7. Venables J. A. and Harland C. J., Philosophical Magazine (1973) 27, 1193-1200. 54. Wilkinson, A. J., et al., Materials Science and Technology (2006) 22, 1271-1278.
8. Venables J. A. and bin-Jaya R, Philosophical Magazine (1977) 35, 1317-1332. 55. Wilkinson, A. J., et al., Ultramicroscopy (2006) 106, 307-313.
9. Harland C. J., et al., Journal of Physics E: Scientific Instruments (1981) 14, 175-181. 56. Wilkinson, A. J., et al., Electron Backscatter Diffraction in Materials Science,
10. Biggin, S. and Dingley, D. J., Journal of Applied Crystallography (1977) 10, 376-385. (2009), eds. Adams, B. L., et al., Kluwer Academic/Plenum Publishers.
11. Dingley, D. J., Scanning Electron Microscopy (1981) 273-286. 57. Kacher, J., et al., Ultramicroscopy (2009) 109, 1148-1156.
12. Dingley, D. J. and Ferran, G., Micron (1977) 8, 145-149. 58. Miyamoto, G., et al., Acta Materialia (2009) 57, 1120-1131.
13. Dingley D. J., Scanning Electron Microscopy (1984) 2, 569-575. 59. Tao, X. and Eades, A., Microscopy and Microanalysis (2005) 11, 341-353.
14. Dingley D. J., Inst Phys Conf Series (1989) 98, 473-476. 60. Villert, S., et al., Journal of Microscopy-Oxford (2009) 233, 290-301.
15. Dingley D. J., et al., Scanning Electron Microscopy (1987) 2, 451-456. 61. Krause, M., et al., AIP Conference Proceedings (2010),1300, 139-144.
16. Dingley, D. J., et al., Textures and Microstructures (1991) 14, 91-96. 62. Ishido, T., et al., Ieice Electronics Express (2007) 4, 775-781.
17. Randle V and Brown A Philosophical Magazine A (1989) 59, 1075-1089. 63. Ojima, M., et al., Acta Materialia (2011) 59, 4177-4185.
18. Randle V, The measurement of grain boundary geometry (1993), Inst. of Physics. 64. Tomita, M., et al., Japanese Journal of Applied Physics 50, 010111.
19. Randle V and Dingley D. J., Scripta Metall (1989) 23, 1565-1569. 65. Vaudin, M. D., et al., Applied Physics Letters (2008) 93, 193116.
20. Dingley, D. J., et al., Institute of Physics Conference Series (1990) 451-454. 66. Vaudin, M. D., et al., Ultramicroscopy (2011) 111, 1206-1213.
21. Baba-Kishi K. Z. and Dingley D. J., Scanning (1989) 11, 305-312. 67. Britton, T. B. and Wilkinson, A. J., Ultramicroscopy (2011) 111, 1395-1404.
22. Baba-Kishi K. Z. and Dingley D. J., Journal of Applied Crystallography (1989) 22, 68. Britton, T. and Wilkinson, A., Ultramicrscopy (2012), 114, 82-95
189-200. 69. Maurice, C., et al., Ultramicroscopy (2012) 113, 171-181.
23. Dingley D. J., et al., Atlas of backscattering Kikuchi diffraction patterns (1995), 70. Maurice, C., et al., Ultramicroscopy (2010) 110, 758-759.
Inst of Phys.
71. Britton, T. B., et al., Ultramicroscopy (2010) 110, 1443-1453.
24. Dingley, D. J. and Wright, S. I., Electron Backscatter Diffraction in Materials
Science, (2009), 97-107, eds. Schwartz, A. J., et al., Springer. 72. Maurice, C. and Fortunier, R., Journl of Microscopy (2008) 230, 520-529.

25. Wright S. I. and Adams B. L., Metallurgical Transactions A (1992) 23A, 759-767. 73. Wilkinson, A. J., Applied Physics Letters (2006) 89, 241910.

26. Adams B. L., et al., Metallurgical Transactions A (1992) 24A, 819-831. 74. Wilkinson, A. J., et al., Superlattices and Microstructures (2009) 45, 285-294.
27. Juul Jensen D and Schmidt NH, Proc. Recrystallisation 90 219-224. 75. Speller, S. C., et al., Applied Physics Letters (2011) 99, 192504.
28. Krieger Lassen N. C., et al., Scanning Microscopy (1992) 6, 115-121. 76. Murphy, J. D., et al., IOP Conf Series: Materials Science and Engineering (2009) 3,
012015.
29. Krieger Lassen N. C., et al., Materials Science Forum (1994) 157-162, 149-158.
77. Wilkinson, A. J., et al., Journal of Strain Analysis for Engineering Design (2010) 45,
30. Goehner RP and Michael JR, J. Res. National Institute of Standards and Technology 365-376.
(1996) 101, 301-308.
78. Dingley, D. J., et al., Journal of Electron Microscopy (2010) 59, S155-S163.
31. Davidson, D., International Metals Review (1984) 29, 75.
79. Karamched, P. S. and Wilkinson, A. J., Acta Materialia (2011) 59, 263-272.
32. Ruff, A., Wear (1976) 40, 59.
80. Littlewood, P., et al., Acta Materialia (2011) 59, 6489-6500.
33. Davidson, D., Journal of Material Science Letters (1982) 1, 236.
81. Britton, T. B., et al., Scripta Materialia (2010) 62, 639-642.
34. Crompton, J. and Martin, J., Metallography (1980) 13, 225.
82. Bhandari, Y., et al., Computational Materials Science (2007) 41, 222-235.
35. Quested P. N., et al., Acta Metallurgica (1988) 36, 2743-2752.
83. Prakash, A. and Lebensohn, R. A., Modelling and Simulation in Materials Science
36. Wilkinson, A. and Dingley, D., Acta Metallurgica Et Materialia (1991) 39, 3047-
and Engineering (2009) 17, .
3055.
84. Raabe, D., et al., Acta Materialia (2001) 49, 3433-3441.
37. Wilkinson, A., et al., Journal of Microscopy-Oxford (1993) 169, 255-261.
85. West, G. D. and Thomson, R. C., Journal of Microscopy-Oxford (2009) 233, 442-
38. Wilkinson A. J. and Dingley D. J., Acta Metallurgica Et Materialia (1992) 40, 3357-
450.
3368.
86. Groeber, M., et al., Electron Backscatter Diffraction in Materials Science, (2010),
39. Arsenlis, A. and Parks, D. M., Acta Materialia (1999) 47, 1597-1611.
eds. Schwartz, A. J., et al., Springer.
40. Wilkinson, A. J. and Randman, D., Philosophical Magazine (2010) 90, 1159-1177.
87. Uchic, M., et al., Microscopy and Microanalysis (2011) 17, 988-989.
41. Field, D. P., et al., Philosophical Magazine (2010) 90, 1451-1464.
88. Zaefferer, S., et al., Metallurgical and Materials Transactions A-Physical Metallurgy
42. Wang, L., et al., Metallurgical and Materials Transactions A-Physical Metallurgy and and Materials Science (2008) 39A, 374-389.
Materials Science (2011) 42A, 626-635.
89. Zaafarani, N., et al., Acta Materialia (2006) 54, 1863-1876.
43. Nix, W. and Gao, H., Journal of the Mechanics and Physics of Solids (1998) 46,
411-425. 90. Groeber, M. A., et al., Materials Characterization (2006) 57, 259-273.

44. Kysar, J. and et al., Journal of the Mechanics and Physics of Solids (2007) 55, 1554- 91. Nolze, G., Ultramicroscopy (2007) 107, 172-183.
1573. 92. Khorashadizadeh, A., et al., Advanced Engineering Materials (2011) 13, 237-244.
45. Britton, T. B., et al., Proceedings of the Royal Society A-Mathematical Physical and 93. Bhandari, Y., et al., Computation Materials Science (2007) 41, 222-235.
Engineering Sciences (2010) 466, 695-719. 94. Rohrer, G. S., et al., Materials Science and Technology (2010) 26, 661-669.
46. Kozubowski, J., et al., Journal of Appllied Crystallography (1991) 24, 102-107. 95. Uchic, M., et al., Scripta Materialia et Materialia (2006) 55, 23-28.
47. Keller, R., et al., Microelectronics Engineering (2004) 75, 96-102. 96. Liu, H. H., et al., Science (2011) 332, 833-834.
48. Wilkinson, A., Materials Science and Technology (1997) 13, 79-84. 97. Jakobsen, B., et al., Science (2006) 312, 889-892.
49. Michael, J. R. and Goehner, R. P., Advances in Materials Problem Solving with the 98. Suter, R. M., et al., Review of Scientific Instruments (2006) 77, 123905.
Electron Microscope, (2001), 39-49, eds. Bentley, J., et al., Materials Research
Society. 99. Winkelmann, A., Ultramicroscopy (2008) 108, 1546-1550.
50. Troost K. Z., et al., Applied Physics Letters (1993) 62, 1110-1112. 100. Winkelmann, A., Journal of Microscopy (2010) 239, 32-45.
51. Wilkinson, A., Ultramicroscopy (1996) 62, 237-247. 101. Winkelmann, A., et al., Ultramicroscopy (2007) 107, 414-421.
52. Wilkinson, A. J., Journal of Electron Microscopy (2000) 49, 299-310. 102. Winkelmann, A. and Vos, M., Physical Review Letters (2011) 106, 085503.
53. Wilkinson, A. J., Scripta Materialia (2001) 44, 2379-2385. 103. Geiss, R. H., et al., Microscopy and Microanalysis (2011) 17, 386-387.

376 SEPTEMBER 2012 | VOLUME 15 | NUMBER 9

Вам также может понравиться