Microscopy and Microanalysis (2020), 1–11

doi:10.1017/S1431927620001671

Original Article

Comparison of Orientation Mapping in SEM and TEM

Joshua D. Sugar1*, Joseph T. McKeown2, Dhego Banga1 and Joseph R. Michael3
1
Sandia National Laboratories, Livermore, CA 94550, USA; 2Lawrence Livermore National Laboratory, Livermore, CA 94550, USA and 3Sandia National Laboratories,
Albuquerque, NM 84550, USA

Abstract
Multiple experimental configurations for performing nanoscale orientation mapping are compared to determine their fidelity to the true
microstructure of a sample. Transmission Kikuchi diffraction (TKD) experiments in a scanning electron microscope (SEM) and nanobeam
diffraction (NBD) experiments in a transmission electron microscope (TEM) were performed on thin electrodeposited hard Au films with
two different microstructures. The Au samples either had a grain size that is >50 or <20 nm. The same regions of the samples were mea-
sured with TKD apparatuses at 30 kV in an SEM with detectors in the horizontal and vertical configurations and in the TEM at 300 kV.
Under the proper conditions, we demonstrate that all three configurations can produce data of equivalent quality. Each method has
strengths and challenges associated with its application and representation of the true microstructure. The conditions needed to obtain
high-quality data for each acquisition method and the challenges associated with each are discussed.
Key words: materials science, electron diffraction, electron back-scattered diffraction (EBSD), transmission Kikuhi diffraction (TKD),
precession electron diffraction (PED)
(Received 11 February 2020; revised 15 May 2020; accepted 25 May 2020)

Introduction Currently, there are several commonly used approaches for

performing orientation mapping on crystalline features less than
Electron backscattered diffraction (EBSD) is a commonly used
10 nm with an electron beam. They differ by the accelerating volt-
electron microscopy technique in the scanning electron micro-
age, the geometry of the diffraction pattern generation and anal-
scope (SEM) that provides crystallographic information over rel-
ysis, and whether they are implemented in an SEM or a
atively large areas (μm to mm) (Dingley, 2004; Schwarzer et al.,
transmission electron microscope (TEM). Although all of these
2009; Goldstein et al., 2018). EBSD can provide information
approaches have successfully mapped microstructural informa-
about orientation, texture (Dingley & Randle, 1992; Prior et al.,
tion at fine length scales, it has not been demonstrated that
1999, Schwartz et al., 2009; Dingley, 2004), strain (Wilkinson
they are sensitive to the same microstructural features. To resolve
et al., 2006; Wright et al., 2011), grain boundary character
this issue, we measure crystallographic orientation with all of
(Humphreys, 2001), and phase (Goehner & Michael, 1996;
these approaches on the same specimens. We compare the results
Michael, 2000) and is a well-accepted technique for understand-
from these different possible experimental configurations to
ing materials microstructure. EBSD is limited, however, in lateral
understand the relationship between the measured map and the
spatial resolution to 50–200 nm depending on the material, the
actual microstructure of the materials.
type of electron emitter, and the incident electron energy
The orientation maps were performed with three different
(Isabell & Dravid, 1997; Ren et al., 1998; Zaefferer, 2007, 2011;
electron microscopes, each providing a different data acquisition
Chen et al., 2011). In thin, electron-transparent samples, it is pos-
configuration. These different methods are schematically drawn
sible to generate transmission Kikuchi diffraction (TKD) patterns
in Figure 1. Note that all three of these arrangements require
(Nishikawa & Kikuchi, 1928a, 1928b) that appear very similar to
that the sample be thin and transparent to the electron beam at
EBSD patterns, and they can be indexed in the same way with the
the given voltage. Figure 1a depicts the conventional arrangement
same automated routines (Schwarzer, 2003; Trimby, 2012; Abbasi
for TKD using a standard EBSD camera where the phosphor
et al., 2015; Borrajo-Pelaez & Hedström, 2018). It is, therefore,
screen is in a vertical position. This arrangement is consistent
possible to increase the lateral spatial resolution using TKD to
with the phosphor positioning for standard EBSD orientation
below 10 nm due to minimization of the interaction volume
mapping and is thus widely applicable. The less conventional
(Keller & Geiss, 2012; Rice et al., 2014; Bober et al., 2015;
TKD arrangement with a horizontally positioned phosphor
Sneddon et al., 2016; van Bremen et al., 2016).
screen is shown in Figure 1b. This is also referred to as an on-axis
geometry. Note that for TKD in either geometry, limitations on
SEM accelerating voltage require that these measurements are
*Author for correspondence: Joshua D. Sugar, E-mail: jdsugar@sandia.gov
Cite this article: Sugar JD, McKeown JT, Banga D, Michael JR (2020) Comparison
done at or around 30 kV. The geometry for precession electron
of Orientation Mapping in SEM and TEM. Microsc Microanal. doi:10.1017/ diffraction (PED) in the TEM is shown in Figure 1c. As in
S1431927620001671 TKD, a thin electron transparent sample is required, but unlike
© Microscopy Society of America 2020

Downloaded from https://www.cambridge.org/core. Carleton University Library, on 28 Jun 2020 at 09:13:28, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S1431927620001671
 2 Joshua D. Sugar et al.

Fig. 1. Three experimental configurations used to collect orientation maps in this study. (a) TKD at 30 kV in an SEM with conventional off-axis detector geometry,
(b) TKD in an SEM at 30 kV with a horizontal on-axis detector, and (c) transmission nanobeam electron diffraction (NBD) at 300 kV in a TEM.

TKD, the accelerating voltage is not limited to ≤30 kV (voltages diffraction pattern is collected. Precession of the electron beam
available in a typical SEM), and PED can be collected at higher at each point may also be utilized to enhance the spot diffraction
voltages (≥120 kV) putting less emphasis on thin samples. patterns by removing some of the dynamical effects normally seen
As shown in Figures 1a and 1b, TKD can be accomplished in in stationary spot diffraction patterns (Vincent & Midgley, 1994;
two different geometries related to the placement of the phosphor Rauch & Veron, 2014). This type of diffraction experiment in a
screen that detects the TKD patterns. Traditional EBSD and TKD TEM, however, requires access to specialized hardware and soft-
have used standard EBSD cameras with the phosphor screen ware for automation and analysis. TKD in an SEM, on the
mounted in a vertical position (Keller & Geiss, 2012; Trimby, other hand, can be performed with the standard EBSD software
2012). More recently, the phosphor screen has been implemented and detectors often installed on existing analytical SEM instru-
in the horizontal position located underneath the thin sample. ments. Therefore, a pertinent question is whether the same crys-
This has also been termed the on-axis location and this terminol- tallographic data can be acquired from thin specimens at both
ogy will be used in the present paper (Fundenberger et al., 2015, 30 kV in an SEM, using the horizontal and the vertical phosphor
2016, Niessen et al., 2018). In both the vertical and horizontal screen arrangements, and at 300 kV in a TEM. Here, we compare
arrangement, a high energy (up to 30 kV in the SEM) electron data from the same electrodeposited Au thin films using various
beam is used to illuminate a thin sample that is positioned so detector and microscope configurations in the SEM and TEM,
that the electron beam is normal (or near normal) to the sample and we discuss the challenges and benefits of each.
surface. The advantage of this arrangement with respect to a tilted
sample is that no corrections for sample tilt nor dynamic focus
Experimental
correction need to be utilized (Trimby et al., 2014). The electrons
that are transmitted through the sample are diffracted and form The comparison of TKD experiments under various conditions
both Kikuchi and spot patterns that are collected on the phosphor required model samples. In this study, electroplated hard Au
screen. Comparison of the on-axis and the vertical phosphor thin films on Si substrates were used. Ni was first sputter depos-
screen orientations have shown that the on-axis arrangement ited on the Si substrates to make the surface conductive. Then, Au
has some advantages due to the higher intensity of the TKD pat- was electroplated using a hard-gold (Au(Co)) plating bath pre-
terns allowing a reduced beam current to be utilized, and thus, an pared according to the specifications indicated by the vendor
improved spatial resolution can be obtained (Niessen et al., 2018). (Electro-Spec, Inc.). This bath consisted of 15 mM potassium
The increased signal associated with on-axis detection also gold cyanide (KAu(CN)2), 89.86 g/L of proprietary OroTech
allowed higher-speed acquisition to be accomplished, and this 551 conducting organic acid salt from EngoTech, and 100 mL/L
minimizes the contribution of sample drift to the resulting data of a cobalt-based proprietary OroTech Replenisher C solution
(Niessen et al., 2018). In addition, the geometry minimizes the from EngoTech, LTD. An OroTech acid salt (EngoTech, LTD)
gnomonic distortion introduced by the use of a vertical phosphor was added to adjust the pH of the bath to 4.1 at room tempera-
screen, allowing the possibility of more accurate orientation deter- ture. During electrodeposition, the bath temperature was kept at
minations (Niessen et al., 2018). 43.3 °C. In one case, the Au was grown under conventional direct
Alternatively, in the TEM, instrumentation developments have current (DC) electroplating conditions (4.3 × 10−3 A/cm2, 15 min),
enabled orientation mapping and indexing of spot diffraction pat- which resulted in a fine-grained Au structure (<20 nm grain size).
terns at higher accelerating voltages (Wright & Dingley, 1998; Additionally, Au was grown under pulsed-current (PC) conditions
Zaefferer, 2000, 2002; Rauch & Dupuy, 2005; Rauch et al., 2010; that combined a low current density (1.07 × 10−3 A/cm2, 60 min),
Moeck et al., 2011; Rauch & Veron, 2014; Sneddon et al., short “on-time” (10 ms), and long “off-time” (90 ms) during plat-
2016). This technique is often referred to as nanobeam diffraction ing, which resulted in a larger grain structure (>50 nm). The sam-
(NBD). In general, the higher accelerating voltages used in the ples were thinned to electron transparency in an FEI Helios
TEM enable higher spatial resolution imaging and analysis. To Nanolab 660 with a final polishing step that utilized 5 kV Ga ions.
accomplish orientation mapping in the TEM, the electron beam TKD data were collected with a conventional off-axis detector
is stepped across the sample surface, and at each point, a geometry as shown in Figure 1a using an Oxford Nordlys Max 2

Downloaded from https://www.cambridge.org/core. Carleton University Library, on 28 Jun 2020 at 09:13:28, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S1431927620001671
 Microscopy and Microanalysis 3

Table 1. Summary of Orientation Map Acquisition Parameters

Parameter On-Axis (Horizontal) TKD Off-Axis (Vertical) TKD NBD

Accelerating voltage 30 kv 30 kV 300 kV

Beam current 3 nA 1.6 nA ∼200 pA
Diffraction pattern resolution 400 × 300 pixels 160 × 120 pixels 144 × 144 pixels
Frame time per pixel 11 ms, 2 frame integration 45 ms, 1 frame integration 40 ms
Pixels per second 46 22 25
Pixel step size 4 nm 4 nm 5 nm
Indexing success rate 97% (large grains), 80% (small grains) 90% (large grains), 80% (small grains) 100%*
Data “cleaning” method None None IQ < 15
Grain misorientation threshold 10° 10° 10°
*ASTAR provides an indexing solution for every pixel regardless of pattern quality, including those acquired from regions of the scan where no sample is present. Thus, a data cleaning
method was used to remove pixels with image quality, IQ, below a threshold value. This removed all data obtained from the vacuum. This cleaning was purely cosmetic and had minimal
effect on the data acquired from the specimen.

(Oxford Instruments, High Wycombe, UK) on an FEI Helios microscopes as it is difficult to mount the sample with the exact
Nanolab 660 operating at 30 kV with a beam current of 1.6 nA. same alignment with respect to the electron column and the
A second TKD configuration, with a horizontal on-axis detector EBSD or the TKD camera. Figure 3 shows how small tilts can
as shown in Figure 1b, used a Bruker e-Flash HR Optimus be added in the analysis software to bring the datasets into orien-
(Bruker, Billerica, MA, USA) on a Zeiss Supra 55 VP (Zeiss, tation alignment. The on-axis and off-axis TKD IPFZ results
Oeberkochen, Germany) operating at 30 kV and with a beam cur- show slightly different colorings for the grain orientation in
rent of 3 nA. Finally, orientation map data were collected in NBD Figure 2. The addition of a sample tilt in w (along stage tilt
mode using a Philips CM 300 FEG TEM operating at 300 kV and axis) of approximately ±5° resulted in orientation colorings that
equipped with a Nanomegas Astar/Topspin (NanoMEGAS SPRL, more closely matched for the two TKD configurations as shown
Brussels, Belgium) system, as shown in Figure 1c. The TEM probe in Figures 3a–3d. Misorientation profiles in Figures 3e and 3f con-
was 5 nm in diameter as measured on the camera at high magni- firm that the same features are indexed the same in the two con-
fication with precession on. A precession angle of 0.7° and a step figurations. The NBD misorientation profile results at 300 kV
size of 5 nm were used during data collection. Data were acquired from the TEM are also plotted for comparison to the TKD results
with both precession on and precession off to ensure that the pre- at 30 kV. There are subtle differences in the locations of grain
cession system was aligned with a consistent beam size at the boundaries, grain shapes, absence or presence of twins, and orien-
specimen plane in both conditions. This ensured no spatial tations when the maps shown in Figure 2 and profiles of Figure 3
resolution-limiting artifacts were introduced that could cause are directly compared. The misorientation profiles of Figures 3e
image blurring and incorrect indexing of diffraction patterns. and 3f are similar for a majority of the data, which shows that
The data presented here utilized precession as there were no evi- a simple rotation can bring orientations into alignment, particu-
dent differences between the two datasets. The same locations in larly for different TKD measurements. However, some differences
the samples were mapped in all three configurations for compar- between the indexing go beyond a simple rotation or translation
ison of the data acquired in the three configurations. A summary of the data. For example, in Figures 2f and 2g, there are some
of the relevant acquisition parameters is shown in Table 1. grains that are twinned (dashed circles). However, the mapping
data obtained using TEM in Figure 2h does not show these
same twins. In addition, the profiles in Figure 3f show variations
Results in the misorientation profiles that go beyond stretching or trans-
lation that would occur as the result of sample drift. The differ-
Comparison of the Three Experimental Configurations
ences between NBD and TKD and the resulting measured maps
A comparison of orientation maps for the three different config- will be addressed in more detail in the Discussion section.
urations shows that similar results are obtained for samples with a For finer-grained samples, obtaining the same quality data
grain size larger than approximately 50 nm, as shown in Figure 2 from all configurations is more difficult and requires special con-
where the images and inverse pole figure Z (IPFZ) maps for all siderations. Orientation maps for DC-plated Au samples are
three experimental configurations are presented. In this case, shown in Figure 4, where the grain size is significantly smaller
the Z direction is normal to the thin foil surface. Inspection of than in the PC-plated samples shown in Figure 2. When the
the images and maps reveals that similar results are obtained three detector configurations are compared for finer-grained sam-
from the same regions of the sample in all three configurations. ples, the conventional TKD detector geometry data, shown in
While there are slight differences in grain morphology between Figure 4f, has significant disadvantages to the other configura-
the two TKD configurations, where some grains appear more tions. There are a large number of unindexed pixels that appear
elongated in the map of Figure 2g, this is likely due to sample black in the data, as can be seen in Figure 4f when compared
drift. In addition, the coloring in some regions of the IPFZ to Figure 4g that shows the results from the on-axis configuration.
maps is slightly different, and this is likely due to small differences Comparison of Figure 4g from the on-axis configuration and the
in sample orientation relative to the detector in the different TEM mapping results shown in Figure 4h demonstrates that these

Downloaded from https://www.cambridge.org/core. Carleton University Library, on 28 Jun 2020 at 09:13:28, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S1431927620001671
 4 Joshua D. Sugar et al.

Fig. 2. Automated orientation mapping using different configurations for pulsed-plated Au films. (a) Secondary electron image at 30 kV in an SEM showing little
information about grain structure and morphology. (b) Virtual bright-field (VBF) image at 300 kV in a TEM. Corresponding band contrast and IPFZ map for a con-
ventional off-axis TKD geometry are shown in (c) and (f), respectively. Band contrast and IPFZ map for a horizontal on-axis TKD geometry are shown in (d) and (g),
respectively. Corresponding pattern quality map and IPFZ map for NBD in the TEM at 300 kV are shown in (e) and (h), respectively. The green box in (a) highlights
the mapped regions in (c), (d), (f), and (g), and this box is superimposed on (e) and (h) for comparison with the region mapped in the TEM (green box in (b)). The
circled grain demonstrates a region where TKD has identified twins within a grain, but NBD has not.

maps differ in both the apparent grain structures as well as the and ran the sample cleaning cycle on the Helios (5 min of air
orientation of the grains. The orientation differences may be plasma cleaning), in addition to using the cryo-clean vessel as a
related to the difficulty of exact specimen alignment between cold-finger to improve the chamber vacuum during the experi-
techniques, but the difference in the grain structure is more ment. It is clear from Figure 5 that the maps collected with the
troubling. two detector arrangements differ only by the sample drift that
Figure 5 shows the orientation maps collected with the hori- occurred during the slower acquisition with the vertical phosphor
zontal and the vertical phosphor screen arrangements from the screen. Proper optimization of the experiment is important in
fine-grained sample, as shown in Figure 4. The maps shown in these cases.
Figure 5 resulted after the thin sample was subject to plasma Figure 6 demonstrates the change in the maps that is observed
cleaning the sample first in Ar (200 mtorr, 50 Watts, 10 min), depending on which side of the sample faces the electron beam in
then in O (200 mtorr, 40 Watts, 5 min). In addition, we used the SEM when a vertical phosphor screen (Figs. 6a, 6b) is used for
the chamber cleaning cycle on the FEI Helios before loading TKD. Figure 6d shows the result obtained from the NBD experi-
the sample, which cleans the chamber with an air plasma in ment. Note that the TKD maps are not identical as shown in the
5 min increments (5 min pump cycles in between) for a total of circled region and that the NBD map (Fig. 6d) appears to show
15 min of plasma cleaning time. We then loaded the sample some convolution of the TKD maps shown in Figures 6b and

Downloaded from https://www.cambridge.org/core. Carleton University Library, on 28 Jun 2020 at 09:13:28, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S1431927620001671
 Microscopy and Microanalysis 5

Fig. 3. IPFZ map for large-grained sample in horizontal detector TKD configuration (a) and (b) with an additional post-processing tilt of w = +5° added. IPFZ map for
large-grained sample in vertical TKD configuration with an additional post-processing tilt of w = −7° added (c) and no additional tilt (d). Misorientation profiles for
profile 1 and profile 2 are shown in (e) and (f), respectively, and show good agreement despite some slight shifting of feature positions due to sample drift.

6c. It is clear from the maps shown in Figures 6b and 6c that the differences in the techniques will always be present that result in
twins in the grain circled in Figures 2f and 2g penetrate through different representations of the sample microstructure in the data.
the thickness of the sample as they are visible on either side of the
sample. It is not clear why the TEM mapping result does not
show these same twins. Discussion
Figure 7 shows two locations within the same sample as in
Elimination of Beam-Induced Contamination at 30 kV
Figure 2 and compares their on-axis TKD and NBD diffraction
patterns. The orientation of these two grains is such that one is At 30 kV, the most significant discrepancy between the horizontal
close to a high-symmetry zone axis and one is not. This is evident and vertical TKD detector configurations is the large number of
by the large number of bright spots in the NBD pattern in unindexed pixels in the case of the vertical detector. This differ-
Figure 6c compared to Figure 7a. The TKD patterns, on the ence is largely caused by the thickness of the beam-induced con-
other hand, experience very little change in intensity with the tamination that develops during acquisition and no significant
same change in sample crystallographic orientation. differences between the electron optics or detector configurations.
Why do these techniques appear to yield different results from While it has been shown previously that the spatial resolution dif-
the same sample? We will show that it is possible to optimize the ference between off-axis (conventional) and on-axis (horizontal)
acquisition parameters to achieve similar quality results for the TKD can be as large as 10 nm (Shen et al., 2019), the unindexed
three experimental configurations, but some fundamental pixels here do not appear to be the result of a limited spatial

Downloaded from https://www.cambridge.org/core. Carleton University Library, on 28 Jun 2020 at 09:13:28, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S1431927620001671
 6 Joshua D. Sugar et al.

Fig. 4. Automated orientation mapping in different configurations for DC-plated Au films. (a) Secondary electron image at 30 kV in an SEM showing little informa-
tion about grain structure and morphology. (b) VBF image at 300 kV in a TEM. Corresponding band contrast and IPFZ map for conventional TKD geometry are
shown in (c) and (f), respectively. Band contrast and IPFZ map for horizontal TKD geometry are shown in (d) and (g), respectively. Corresponding pattern quality
map and IPFZ map for NBD in the TEM at 300 kV are shown in (e) and (h), respectively. The green boxes highlight the mapped regions in (d) and (g). The red dashed
boxes show the overlapping regions in the different datasets.

resolution. Some unindexed pixels may be a result of the extreme amorphous contamination layer results in weaker patterns on a
gnomonic projection distortion that is caused by the sample and more intense background, which reduces the likelihood that the
detector geometry used with a vertical phosphor screen. However, indexing software can converge on a solution. It is absolutely cru-
the use of a plasma cleaning system and a cold-finger to improve cial to reduce this contamination in a vertical TKD detector con-
the quality of the vacuum in the SEM chamber decreased the con- figuration to obtain high-quality results. One option is to tilt the
tamination rate and resulted in higher-quality results and an sample to increase θ as shown in Figure 5c. This results in a
increased index rate for the conventional/vertical detector config- smaller effective thickness of the carbon contamination. The
uration. This is demonstrated in Figure 5. best option is to eliminate or substantially reduce the contamina-
SEM chambers are not known for their cleanliness, and the tion rate to a more acceptable level through the use of careful
buildup of C contamination under the electron beam is a problem sample handling and plasma cleaning of the sample and the
that most operators face. Figure 5a demonstrates that it was pos- microscope chamber.
sible to grow a C contamination “rectangle” into a hole of the The resulting orientation map after all of these cleaning steps is
sample in our SEM chamber (outlined with a dashed line). The shown in Figure 5e. This map is much improved over the original
schematics in Figures 5b and 5c show that in the case of a vertical map shown in Figure 4f, and there are very few unindexed pixels.
TKD detector, the electron beam path length through the con- In addition, the similarity between the horizontal detector config-
tamination can be longer than in the horizontal case. This uration in Figure 5d and the vertical configuration in Figure 5e is

Downloaded from https://www.cambridge.org/core. Carleton University Library, on 28 Jun 2020 at 09:13:28, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S1431927620001671
 Microscopy and Microanalysis 7

Fig. 5. (a) 30 kV scanning transmission electron microscopic image of a sample showing the deposition of a rectangular-shaped region of C contamination from the
electron beam, outlined by the dashed line. Schematics showing the effective thickness of the beam-deposited contamination for a horizontal detector (b) and a
vertical detector (c). IPFZ orientation maps for a horizontal detector in (d) and vertical detector in (e) after plasma cleaning the chamber, plasma cleaning the
sample, and using a cold-finger to increase the vacuum quality in the SEM. The data in (d) and (e) are comparable and show similar grain orientations and mor-
phologies despite some differences from sample drift.

apparent. The grain orientations and morphologies are the same. However, close inspection of the results at 300 kV with NBD
There is some distortion of the grains, and this is most likely due reveals differences in the orientation maps. It can be observed
to sample drift. Sample drift is a challenge for these types of mea- from Figure 2 and the profiles in Figures 3e and 3f that the
surements. It is possible to use automated drift correction rou- TKD maps and the NBD maps are different in the details of
tines, but because of the typical pixel times (∼1 ms), this the grain shapes, the presence of twins within the grains, and
usually results in an image with jitter or discontinuities every the measured misorientation of some of the microstructural fea-
time the drift correction routine applies a beam shift. The quality tures. These differences are mainly due to the volume of material
of images is better without drift correction, but the grain morphol- sampled by each technique and where the detected diffraction
ogy may not be accurate. The specific goals of any experiment will patterns originate. When the grain size is smaller than the sample
define whether or not such a routine should be used and whether a thickness NBD produces spot diffraction patterns that contain
visually appealing image with continuous boundaries or more accu- information from all the grains sampled through thickness
rate grain morphology with discontinuous boundaries is useful. In (Viladot et al., 2013; Rauch & Veron, 2014; Valery et al., 2017),
our case, the drift is a constant mechanical drift, for which longer which can be useful for 3D measurements (Liu et al., 2011).
settle times reduce the drift considerably. In cases where the sample TKD measurements originate from a thinner region of the sample
is charging and drift is caused by the electron beam, other strategies and are more representative of a single plane (Rice et al., 2014).
must be used to reduce and prevent any sample charging (better It is well known that the diffraction patterns in TKD are pro-
sample grounding). The best option to reduce drift, in most cases duced by the material closest to the exit surface of the electron
when charging is not an issue, is to allow the sample significant beam in the sample (Rice et al., 2014; Liu et al., 2019). Recent
time to settle in the microscope prior to data acquisition so that work has shown that for higher atomic number targets like Pt,
the inherent drift is small (Trimby et al., 2014). Given the high- the depth from which the TKD patterns arise is between 5 and
throughput environments of most SEM labs, however, this may 12 nm (Sneddon et al., 2017; Liu et al., 2019). This is schemati-
not always be possible. The operator must, therefore, choose cally shown in Figure 6a. At 30 kV, material at the uppermost sur-
between image quality and accurate morphology. face of the sample scatters the electron beam (dark blue path), but
it is then scattered again (light blue path) as it continues to prop-
agate through the sample. These multiple scattering events con-
Additional Challenges at 300 kV with NBD
tribute to the background in the Kikuchi pattern. Eventually,
The data provided in Figures 2–5 show that it is possible to obtain only the material near the exit surface (the red path) contributes
orientation map results with equivalent quality from both the to the measured indexable pattern because the multiply scattered
horizontal and vertical TKD detector configuration at 30 kV. blue paths contribute to the background. At 300 kV, however, the

Downloaded from https://www.cambridge.org/core. Carleton University Library, on 28 Jun 2020 at 09:13:28, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S1431927620001671
 8 Joshua D. Sugar et al.

Fig. 6. (a) Schematic showing an example of a TKD pattern at 30 kV that has contributions only from the grain at the exit surface. Multiple scattering from the grains
closer to the entry surface of the sample contribute to the background signal. At 300 kV, the NBD “spot” pattern is a superposition of all of the grains that diffract
through the sample thickness. The IPFZ maps in (b) and (c) were acquired using a vertical TKD detector at 30 kV, where the sample has been flipped so that the
opposite surface is at the beam exit. The IPFZ map in (d) was acquired from a 300 kV NBD experiment. Regions circled with the dashed lines show where the 300 kV
map is a superposition of grains that are singularly visible in (b) and (c).

Fig. 7. VBF image from the NBD experiment at 300 kV. Diffraction patterns from selected grains are shown in (a) and (c). The TKD patterns are recorded at 30 kV
with a horizontal detector configuration, the NBD patterns at 300 kV. In (a), the grain is not near a zone axis orientation. The TKD pattern still has sufficient data for
a proper Hough indexing, but the NBD pattern only has two bright spots that must be compared to a library of patterns to find the best fit. In (c), a grain is selected
closer to a zone axis orientation. The TKD pattern still shows sufficient bands for a Hough indexing, and the NBD patterns have a more symmetrically bright spot
pattern that enables easier indexing.

electron mean free path is much longer, so the recorded diffrac- thickness of the sample. This is problematic from an indexing
tion pattern is a superposition of the patterns produced by the perspective because the software can only choose one orientation
various grains that scatter the electron beam through the for every pixel in the orientation map. At 30 kV, for very fine

Downloaded from https://www.cambridge.org/core. Carleton University Library, on 28 Jun 2020 at 09:13:28, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S1431927620001671
 Microscopy and Microanalysis 9

grains, it is also possible to have a superposition of patterns and mathematical routines based on a Hough transform that index
an interface contained within one pixel, but it is less likely because conventional EBSD patterns are used to index the TKD patterns
the near-exit surface region that contributes to the diffraction pat- (Lassen, 1998; Tao & Eades, 2005). The patterns may have
tern is small, on the order of tens of nanometers. In the case some additional distortion in the TKD configuration, but these
where a superposition and an interface are both present, the soft- routines are robust enough that the same software can index
ware typically does not index the pixel, but this is much less likely these patterns without modification. Indexing “spot” patterns
at 30 kV. At 300 kV, it is much more likely to have multiple pat- like those that are collected at 300 kV is more mathematically
terns in every pixel (for a fine-grained sample) because of the long demanding. Typically, a “dictionary” approach is used, where a
electron scattering length, and it will be more often necessary to large number of simulated patterns of different orientations are
choose the best fit for a map. generated and compared to the experimental patterns to find
A demonstration of this effect is shown in Figures 6b–6d. In the best fit (Rauch & Dupuy, 2005; Chen et al., 2015;
Figures 6b and 6c, the maps are collected with a vertical off-axis Marquardt et al., 2017; Ram et al., 2017; Friedrich et al., 2018;
TKD detector configuration at 30 kV. In Figure 6c, the sample was Jha et al., 2018; Ram & De Graef, 2018; Charpagne et al., 2019;
flipped 180° so that the opposite surface from Figure 6b was at the Foden et al., 2019; Jackson et al., 2019; Zhu et al., 2019;
electron beam exit. The dashed and circled regions show the areas Kaufmann et al., 2020). In Figure 7, an example is shown for
where the grain structures were different at the exit surfaces. One two grains. The virtual bright-field (VBF) image in Figure 7b
example is a twinned region, that is thin and visible in Figure 6b, shows the grain structure of the sample. In Figure 7a, a 30 kV
but totally invisible in Figure 6c. This twin is inclined relative to TKD diffraction pattern and a 300 kV NBD pattern are shown
the exit surface, and only some of it is mapped in Figure 6b. In for a grain that is not near a zone axis orientation. In Figure 7c,
Figure 6d, at 300 kV, however, the twin is mapped and much the same comparison is made for a grain that is near a zone
thicker, because at 300 kV the entire projected thickness of the axis orientation with high symmetry. In the case of the 30 kV
twin contributes to the diffraction pattern being indexed. The TKD patterns, it is clear that regardless of the orientation and
region on the right side of the images also shows differences in proximity to a high-symmetry orientation, there are always clear
the grain structure between the two exit surfaces in Figures 6b Kikuchi bands in the pattern that are successfully indexed using
and 6c. In this case, the TEM indexing software has chosen a bet- a Hough transform. The sample is a high-quality sample, and pat-
ter fit that more closely resembles the structure in Figure 6c. This terns of sufficient quality are collected such that they can be
could be because the samples are oriented with the same exit sur- indexed. At 300 kV, however, when the grain is not near a zone
face in Figures 6b and 6c so the same single-scattering diffraction axis, the diffraction pattern is low symmetry and only has 2–3
events from the exit surface are detected in those two cases. bright Bragg reflections. This adds complexity to the indexing
Another example of this is the dashed and circled regions routine and is more likely to result in multiple solutions that pro-
shown in Figures 2f–2h. The TKD data shows green-colored vide a good fit due to the sparseness of the data. In Figure 6c,
twins within the pink/orange-colored grain. However, the NBD when the grain is close to a high-symmetry zone axis orientation,
data do not show these twins. For the TEM mapping case, it is the spot pattern is more easily indexed because there are several
necessary to make assumptions about the sample and the inten- bright Bragg reflections that can be used for pattern matching.
sities in the diffraction patterns from overlapping patterns to The additional mathematical difficulties in indexing spot patterns
attempt to deconvolve the patterns and present orientation with sparse data create more opportunities for differences between
maps that represent a plane within the thin sample (Valery the data at 30 and 300 kV. Another demonstration of this concept
et al., 2017). It is clear that from Figure 6 that when the sample is shown in Figure 2 where twins are identified in Figures 2(f) and
is flipped so that the TKD patterns come from either surface 2(g) that are not shown in Figure 2(h) because there are not
that these twins exist and are through thickness. However, the enough spots from the twin in this orientation for the software
TEM result appears to be insensitive to these microstructural to properly index it. Proper indexing may require more interfer-
details. Perhaps the intensities from the twin are too weak to be ence from the operator to properly index patterns. However, it
detectable in the NBD patterns, the software is not able to delin- is clear from the data presented here that despite these challenges,
eate these twin reflections from the bulk grain in the particular it is still possible to index these patterns and have orientation
orientation of the sample, or the orientation of the twin does maps that are representative of the sample.
not contain enough spots to uniquely identify it (as discussed
in Fig. 7). For TKD, it is important to recognize that the ability
Conclusions
to identify fine-scale features will depend largely on the pixel
size used and what feature is present at the exit surface of the sam- The ability to measure grain orientation and structure at the
ple. If the feature is much smaller than the pixel size, the diffrac- nanoscale is a critical piece of an electron microscopist’s toolbox.
tion signal may be overwhelmed by the surrounding signal and Typically, when analyzing grains <50 nm, one would likely
the feature may not be properly indexed. For NBD, the overlap assume that it is necessary to use a TEM because of the resolution
of several features through the thickness of the sample can create limits of SEM. We have shown here, however, that at 30 kV mul-
difficulties in indexing and mapping the structure in a 2D projec- tiple experimental configurations can achieve resolutions
tion and explains the differences observed on the right side of approaching 5 nm or smaller on the same sample as demon-
Figures 6b–6d. Regardless of these differences, one needs to be strated by the feature sizes visible in TKD orientation maps.
aware of how the interpretation of the microstructural features Data of the same quality can be obtained using either a horizontal
may be affected by these differences between SEM- and (on-axis) TKD detector configuration at 30 kV, a vertical detector
TEM-based techniques. (off-axis) configuration at 30 kV, or with NBD at 300 kV. The dif-
Additional challenges are present at 300 kV simply because of ferences between the data visualization can be reasonably
the difficulty associated with indexing parallel beam “spot” pat- explained with simple scattering length arguments. If it is possible
terns compared to Kikuchi “line” patterns. At 30 kV, the same to clean the SEM chamber to prevent contamination, then TKD at

Downloaded from https://www.cambridge.org/core. Carleton University Library, on 28 Jun 2020 at 09:13:28, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S1431927620001671
 10 Joshua D. Sugar et al.

30 kV is a good compliment or possible alternative to analysis that Fundenberger JJ, Bouzy E, Goran D, Guyon J, Morawiec A & Yuan H
would traditionally be obtained in a TEM at higher accelerating (2015). Transmission Kikuchi diffraction (TKD) via a horizontally posi-
voltage. The true microstructure of materials is three dimensional, tioned detector. Microsc Microanal 21(S3), 1101–1102.
Fundenberger JJ, Bouzy E, Goran D, Guyon J, Yuan H & Morawiec A
however, and both TKD and NBD orientation maps offer two-
(2016). Orientation mapping by transmission-SEM with an on-axis detec-
dimensional projections of those microstructures where the spe-
tor. Ultramicroscopy 161, 17–22.
cifics depend on the technique and specimen. Microscopists typ- Goehner RP & Michael JR (1996). Phase identification in a scanning electron
ically interpret orientation maps as representing a single specimen microscope using backscattered electron Kikuchi patterns. J Res Natl Inst
plane. This is reasonably true with either a TKD approach, but it Stand Technol 101(3), 301–308.
is clearly not the case for NBD orientation mapping in the TEM Goldstein JI, Newbury DE, Michael JR, Ritchie NW, Scott JHJ & Joy DC
where it is unclear as to the actual plane that is sampled and rep- (2018). Scanning Electron Microscopy and X-Ray Microanalysis.
resented in orientation maps. The goal of any specific measure- New York: Springer.
ment or experiment will define which technique will be most Groeber MA, Haley BK, Uchic MD, Dimiduk DM & Ghosh S (2006). 3D
appropriate based on the desired sampling volume. When true reconstruction and characterization of polycrystalline microstructures
using a FIB-SEM system. Mater Charact 57(4–5), 259–273.
3D representations are needed, more complex and slower 3D
Humphreys FJ (2001). Review – Grain and subgrain characterisation by elec-
serial sectioning or tomography techniques may be required
tron backscatter diffraction. J Mater Sci 36(16), 3833–3854.
(Groeber et al., 2006; Liu et al., 2011; McDonald et al., 2015). Isabell TC & Dravid VP (1997). Resolution and sensitivity of electron back-
When TKD or NBD techniques are combined with EBSD, it is scattered diffraction in a cold field emission gun SEM. Ultramicroscopy
possible to access crystallographic information from the nanoscale 67(1–4), 59–68.
up to the millimeter scale to understand the crystallographic Jackson MA, Pascal E & De Graef M (2019). Dictionary indexing of electron
parameters that define a materials’ microstructure at all practical back-scatter diffraction patterns: A hands-on tutorial. Integr Mater Manuf
length scales. Innov 8(2), 226–246.
Jha D, Singh S, Al-Bahrani R, Liao WK, Choudhary A, De Graef M &
Acknowledgments. Special thanks to Warren York for preparing these Agrawal A (2018). Extracting grain orientations from EBSD patterns of
electron-transparent samples. Sandia National Laboratories is a multimission polycrystalline materials using convolutional neural networks. Microsc
laboratory managed and operated by the National Technology & Microanal 24(5), 497–502.
Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Kaufmann K, Zhu CY, Rosengarten AS, Maryanovsky D, Harrington TJ,
Honeywell International Inc., for the U.S. Department of Energy’s National Marin E & Vecchio KS (2020). Crystal symmetry determination in electron
Nuclear Security Administration under contract DE-NA0003525. Work at diffraction using machine learning. Science 367(6477), 564.
Lawrence Livermore National Laboratory (LLNL) was performed under the Keller RR & Geiss RH (2012). Transmission EBSD from 10 nm domains in a
auspices of the U.S. Department of Energy by LLNL under Contract scanning electron microscope. J Microsc 245(3), 245–251.
DE-AC52-07NA27344. This paper describes objective technical results and Lassen NCK (1998). Automatic high-precision measurements of the location
analysis. Any subjective views or opinions that might be expressed in the and width of Kikuchi bands in electron backscatter diffraction patterns. J
paper do not necessarily represent the views of the U.S. Department of Microsc (Oxford) 190, 375–391.
Energy or the United States Government. Liu HH, Schmidt S, Poulsen HF, Godfrey A, Liu ZQ, Sharon JA & Huang X
(2011). Three-dimensional orientation mapping in the transmission elec-
tron microscope. Science 332(6031), 833–834.
Liu JL, Lozano-Perez S, Wilkinson AJ & Grovenor CRM (2019). On the
References
depth resolution of transmission Kikuchi diffraction (TKD) analysis.
Abbasi M, Kim DI, Guim HU, Hosseini M, Danesh-Manesh H & Abbasi M Ultramicroscopy 205, 5–12.
(2015). Application of transmitted Kikuchi diffraction in studying nano- Marquardt K, De Graef M, Singh S, Marquardt H, Rosenthal A & Koizuimi
oxide and ultrafine metallic grains. ACS Nano 9(11), 10991–11002. S (2017). Quantitative electron backscatter diffraction (EBSD) data analyses
Bober DB, Kumar M & Rupert TJ (2015). Nanocrystalline grain boundary using the dictionary indexing (DI) approach: Overcoming indexing difficul-
engineering: Increasing Sigma 3 boundary fraction in pure Ni with thermo- ties on geological materials. Am Miner 102(9), 1843–1855.
mechanical treatments. Acta Mater 86, 43–54. McDonald SA, Reischig P, Holzner C, Lauridsen EM, Withers PJ, Merkle
Borrajo-Pelaez R & Hedström P (2018). Recent developments of crystallo- AP & Feser M (2015). Non-destructive mapping of grain orientations in
graphic analysis methods in the scanning electron microscope for applica- 3D by laboratory X-ray microscopy. Sci Rep 5, 14665.
tions in metallurgy. Crit Rev Solid State Mater Sci 43(6), 455–474. Michael JR (2000). Phase identification using electron backscatter diffraction
Charpagne MA, Strub F & Pollock TM (2019). Accurate reconstruction of in the scanning electron microscope. In Electron Backscatter Diffraction in
EBSD datasets by a multimodal data approach using an evolutionary algo- Materials Science, Schwartz AJ, Kumar M & Adams BL (Eds.), pp. 75–89.
rithm. Mater Charact 150, 184–198. Boston, MA: Springer US.
Chen D, Kuo JC & Wu WT (2011). Effect of microscopic parameters on Moeck P, Rouvimov S, Rauch EF, Veron M, Kirmse H, Hausler I, Neumann
EBSD spatial resolution. Ultramicroscopy 111(9–10), 1488–1494. W, Bultreys D, Maniette Y & Nicolopoulos S (2011). High spatial resolu-
Chen YH, Park SU, Wei D, Newstadt G, Jackson MA, Simmons JP, De tion semi-automatic crystallite orientation and phase mapping of nanocrys-
Graef M & Hero AO (2015). A dictionary approach to electron backscatter tals in transmission electron microscopes. Cryst Res Technol 46(6), 589–606.
diffraction indexing. Microsc Microanal 21(3), 739–752. Niessen F, Burrows A & Fanta ABD (2018). A systematic comparison of
Dingley D (2004). Progressive steps in the development of electron backscatter on-axis and off-axis transmission Kikuchi diffraction. Ultramicroscopy
diffraction and orientation imaging microscopy. J Microsc (Oxford) 213, 186, 158–170.
214–224. Nishikawa S & Kikuchi S (1928a). Diffraction of cathode rays by calcite.
Dingley DJ & Randle V (1992). Microtexture determination by electron back- Nature 122, 726–726.
scatter diffraction. J Mater Sci 27(17), 4545–4566. Nishikawa S & Kikuchi S (1928b). Diffraction of cathode rays by mica. Nature
Foden A, Collins DM, Wilkinson AJ & Britton TB (2019). Indexing electron 121, 1019–1020.
backscatter diffraction patterns with a refined template matching approach. Prior DJ, Boyle AP, Brenker F, Cheadle MC, Day A, Lopez G, Peruzzo L,
Ultramicroscopy 207, 112845. Potts GJ, Reddy S, Spiess R, Timms NE, Trimby P, Wheeler J &
Friedrich T, Bochmann A, Dinger J & Teichert S (2018). Application of the Zetterstrom L (1999). The application of electron backscatter diffraction
pattern matching approach for EBSD calibration and orientation mapping, and orientation contrast imaging in the SEM to textural problems in
utilising dynamical EBSP simulations. Ultramicroscopy 184, 44–51. rocks. Am Miner 84(11–12), 1741–1759.

Downloaded from https://www.cambridge.org/core. Carleton University Library, on 28 Jun 2020 at 09:13:28, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S1431927620001671
 Microscopy and Microanalysis 11

Ram F & De Graef M (2018). Phase differentiation by electron backscatter dif- Trimby PW (2012). Orientation mapping of nanostructured materials using
fraction using the dictionary indexing approach. Acta Mater 144, 352–364. transmission Kikuchi diffraction in the scanning electron microscope.
Ram F, Wright S, Singh S & De Graef M (2017). Error analysis of the crystal Ultramicroscopy 120, 16–24.
orientations obtained by the dictionary approach to EBSD indexing. Trimby PW, Cao Y, Chen ZB, Han S, Hemker KJ, Lian JS, Liao XZ,
Ultramicroscopy 181, 17–26. Rottmann P, Samudrala S, Sun JL, Wang JT, Wheeler J & Cairney JM
Rauch EF & Dupuy L (2005). Rapid spot diffraction patterns identification (2014). Characterizing deformed ultrafine-grained and nanocrystalline
through template matching. Arch Metall Mater 50(1), 87–99. materials using transmission Kikuchi diffraction in a scanning electron
Rauch EF, Portillo J, Nicolopoulos S, Bultreys D, Rouvimov S & Moeck P microscope. Acta Mater 62, 69–80.
(2010). Automated nanocrystal orientation and phase mapping in the trans- Valery A, Rauch EF, Clement L & Lorut F (2017). Retrieving overlapping
mission electron microscope on the basis of precession electron diffraction. crystals information from TEM nano-beam electron diffraction patterns. J
Z Kristallogr Cryst Mater 225(2–3), 103–109. Microsc 268(2), 208–218.
Rauch EF & Veron M (2014). Automated crystal orientation and phase map- van Bremen R, Gomes DR, de Jeer LTH, Ocelik V & De Hosson JTM
ping in TEM. Mater Charact 98, 1–9. (2016). On the optimum resolution of transmission-electron backscattered
Ren SX, Kenik EA, Alexander KB & Goyal A (1998). Exploring spatial res- diffraction (t-EBSD). Ultramicroscopy 160, 256–264.
olution in electron back-scattered diffraction experiments via Monte Viladot D, Veron M, Gemmi M, Peiro F, Portillo J, Estrade S, Mendoza J,
Carlo simulation. Microsc Microanal 4(1), 15–22. Llorca-Isern N & Nicolopoulos S (2013). Orientation and phase mapping
Rice KP, Keller RR & Stoykovich MP (2014). Specimen-thickness effects on in the transmission electron microscope using precession-assisted diffrac-
transmission Kikuchi patterns in the scanning electron microscope. J tion spot recognition: State-of-the-art results. J Microsc 252(1), 23–34.
Microsc 254(3), 129–136. Vincent R & Midgley PA (1994). Double conical beam-rocking system for
Schwartz AJ, Kumar M, Adams BL & Field DP (2009). Electron Backscatter measurement of integrated electron-diffraction intensities. Ultramicroscopy
Diffraction in Materials Science. New York, NY: Springer. 53(3), 271–282.
Schwarzer RA (2003). Automated grain orientation measurement by backscat- Wilkinson AJ, Meaden G & Dingley DJ (2006). High-resolution elastic strain
ter Kikuchi diffraction. Phys Met Metallogr+ 96, S104–S115. measurement from electron backscatter diffraction patterns: New levels of
Schwarzer RA, Field DP, Adams BL, Kumar M & Schwartz AJ (2009). sensitivity. Ultramicroscopy 106(4–5), 307–313.
Present state of electron backscatter diffraction and prospective developments. Wright SI & Dingley DJ (1998). Orientation imaging in the transmission elec-
In Electron Backscatter Diffraction in Materials Science, Schwartz AJ, Kumar tron microscope. Mater Sci Forum 273–275, 209–214.
M, Adams BL & Field DP (Eds.), pp. 1–20. Boston, MA: Springer US. Wright SI, Nowell MM & Field DP (2011). A review of strain analysis using
Shen YT, Xu JC, Zhang YS, Wang YZ, Zhang JM, Yu BJ, Zeng Y & Miao H electron backscatter diffraction. Microsc Microanal 17(3), 316–329.
(2019). Spatial resolutions of on-axis and off-axis transmission Kikuchi dif- Zaefferer S (2000). New developments of computer-aided crystallographic
fraction methods. Appl Sci (Basel) 9, 21. analysis in transmission electron microscopy. J Appl Crystallogr 33, 10–25.
Sneddon G, Trimby P & Cairney J (2017). The influence of microscope and Zaefferer S (2002). Computer-aided crystallographic analysis in the TEM. Adv
specimen parameters on the spatial resolution of transmission Kikuchi dif- Imag Elect Phys 125, 355–415.
fraction. Microsc Microanal 23(S1), 532–533. Zaefferer S (2007). On the formation mechanisms, spatial resolution and inten-
Sneddon GC, Trimby PW & Cairney JM (2016). Transmission Kikuchi diffrac- sity of backscatter Kikuchi patterns. Ultramicroscopy 107(2–3), 254–266.
tion in a scanning electron microscope: A review. Mat Sci Eng R 110, 1–12. Zaefferer S (2011). A critical review of orientation microscopy in SEM and
Tao XD & Eades A (2005). Measurement and mapping of small changes of TEM. Cryst Res Technol 46(6), 607–628.
crystal orientation by electron backscattering diffraction. Microsc Microanal Zhu CY, Kaufmann K & Vecchio K (2019). Automated reconstruction of
11(4), 341–353. spherical Kikuchi maps. Microsc Microanal 25(4), 912–923.

Downloaded from https://www.cambridge.org/core. Carleton University Library, on 28 Jun 2020 at 09:13:28, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S1431927620001671