Three-Dimensional Orientation Microscopy in a Focused Ion

Beam–Scanning Electron Microscope: A New Dimension

of Microstructure Characterization
S. ZAEFFERER, S.I. WRIGHT, and D. RAABE

In the present work, we report our recent progress in the development, optimization, and
application of a technique for the three-dimensional (3-D) high-resolution characterization of
crystalline microstructures. The technique is based on automated serial sectioning using a
focused ion beam (FIB) and characterization of the sections by orientation microscopy based on
electron backscatter diffraction (EBSD) in a combined FIB–scanning electron microscope
(SEM). On our system, consisting of a Zeiss–Crossbeam FIB-SEM and an EDAX-TSL EBSD
system, the technique currently reaches a spatial resolution of 100 · 100 · 100 nm3 as a stan-
dard, but a resolution of 50 · 50 · 50 nm3 seems to be a realistic optimum. The maximum
observable volume is on the order of 50 · 50 · 50 lm3. The technique extends all the powerful
features of two-dimensional (2-D) EBSD-based orientation microscopy into the third dimension
of space. This allows new parameters of the microstructure to be obtained—for example, the full
crystallographic characterization of all kinds of interfaces, including the morphology and the
crystallographic indices of the interface planes. The technique is illustrated by four examples,
including the characterization of pearlite colonies in a carbon steel, of twins in pseudonano-
crystalline NiCo thin films, the description of deformation patterns formed under nanoindents
in copper single crystals, and the characterization of fatigue cracks in an aluminum alloy. In
view of these examples, we discuss the possibilities and limits of the technique. Furthermore, we
give an extensive overview of parallel developments of 3-D orientation microscopy (with a focus
on the EBSD-based techniques) in other groups.

DOI: 10.1007/s11661-007-9418-9

The Minerals, Metals & Materials Society and ASM International 2008

I. INTRODUCTION tion that is completely missing in 2-D microstructure

observations is the full crystallographic characterization
A. Motivation: Why Perform Three-Dimensional of all kinds of interfaces. In 2-D orientation microscopy,
Investigation? the misorientation across a boundary can be deter-
METALLOGRAPHIC techniques for the character- mined; however, since the spatial orientation of the
ization of microstructures of materials are usually boundary plane is unknown, the orientation of the
applied to two-dimensional (2-D) planes cut through a boundary plane relative to the crystal lattices separated
sample. Statistical stereological techniques can be used by the plane cannot be determined. This information is
to gain insight into the three-dimensional (3-D) aspects of great importance for the investigation of phase
of microstructure.[1,2] However, there are also a large transformations, grain growth processes, and intergran-
number of cases in which a true 3-D characterization of ular fracture.
a sample volume is critical to a correct understanding of Principally, the 3-D characterization of microstruc-
microstructure formation mechanisms or particular tures can be performed by two different approaches,
material properties. Cases in which 3-D information is either by serial sectioning or by observation with some
of great importance include, for example, the true size sort of transmissive radiation. The latter techniques
and shape of grains for grain growth investigations, the obtain the 3-D information from bulk samples, either by
size and shape of strain fields for studies on deforma- reconstruction from a large number of extinction images
tion, and the distribution and position of nuclei for taken from different directions of the sample or by a ray-
recrystallization investigations. An example of informa- tracing technique. The ray-tracing techniques allow
crystallographic information to be obtained. They have
undergone intensive development in recent times in
S. ZAEFFERER, Head of the Microscopy and Diffraction Group,
various groups, as in, for example, References 3 through
and D. RAABE, Department Director, are with the Max Planck 6, and made impressive progress. The current spatial
Institute for Iron Research (MPIE), D-40235, Düsseldorf, Germany. resolution is on the order of the volumes of some cubic
Contact e-mail: s.zaeff[email protected] S.I. WRIGHT, Director of microns and the method has been shown to yield in-situ
Applications Science and Software, is with EDAX-TSL, Draper, observations of recrystallization and deformation pro-
UT 84020.
Manuscript submitted May 22, 2007. cesses of cubic materials, as in, for example, References
Article published online January 4, 2008 7 and 8.

374—VOLUME 39A, FEBRUARY 2008 METALLURGICAL AND MATERIALS TRANSACTIONS A

B. Tomography by Serial Sectioning marked and easily classified according to their crystal-
Significantly simpler than the bulk methods, yet lographic character. Each feature can be separated and
comparably powerful (and partly complementary) meth- displayed individually. Areas (or volumes in 3-D)
ods for the 3-D characterization of crystalline materials belonging to the same grain in different sections can
are the serial sectioning techniques, which are tomo- be easily assembled by their similar orientation. Orien-
graphic techniques, in the wordÕs original meaning. The tation microscopy is, therefore, the pre-eminent method
technique is simply composed of cutting material slices for 3-D microstructure characterization based on serial
with some cutting technique, recording the structure of sectioning.
these slices with an appropriate microscopic or macro- The combination of EBSD-based orientation micros-
scopic recording technique, and reconstructing the 3-D copy with serial sectioning is not completely new, but is
structure, e.g., by stacking the recorded images. Serial currently under strong development. The recent view-
sectioning is applicable to a very wide range of materials point set in Reference 19 documents examples. In
and material problems; the only serious disadvantage is addition, the combination of orientation microscopy
that the method is destructive. with serial sectioning by FIB has been described
For serial sectioning, numerous and different methods previously.[19–23]
can be imagined, e.g., mechanical cutting, grinding, or The aim of the present article is to give a compre-
polishing; chemical polishing; or laser or electrical hensive overview on the emerging technique of 3-D
discharge ablation. For observation of the serial sec- orientation microscopy based on serial sectioning, using
tions, all kinds of microscopic techniques are available, a FIB- and 2-D EBSD-based orientation microscopy.
e.g., light optical observations following some etching We illustrate the technique by means of our own
procedure or scanning electron microscope (SEM) developments, which include, in fact, a new geometrical
observations using contrast created by backscattered setup that has advantages over setups developed thus far
electrons (BSEs). The main challenges associated with by other groups. The potential of the technique for
many of the sectioning methods are controlling the materials science is demonstrated by some of our recent
sectioning depth, obtaining flat and parallel surfaces, experimental results. Other results obtained by a manual
and correctly redetecting and aligning the area of procedure have been documented already in earlier
observation. On the observation side, the main obstacle publications.[9,24,25]
is that most observation techniques do not allow certain
microstructural features, such as grain boundaries or
individual grains, to be separated, because the contrast II. PRINCIPLE OF 3-D CHARACTERIZATION IN
conditions are not well enough defined. Finally, most A FIB-SEM
serial sectioning methods are extremely laborious.
Examples of various serial sectioning methods have A. The Geometrical Setup of the System
been compiled in one of our previous publications.[9] The system described here has been developed on a
A technique that avoids, in an ideal way, all of these Zeiss–Crossbeam (Carl Zeiss SMT AG, Oberkochen,
problems with sectioning and imaging is the combina- Germany, CrossBeam (R)) XB 1540 FIB-SEM, which
tion of serial sectioning with an ion beam- and electron consists of a Gemini-type field emission gun (FEG)
backscatter diffraction (EBSD)-based orientation electron column and an Orsay Physics (Orsay Physics
microscopy in a combined focused ion beam (FIB)- FIB is a part of the CrossBeam instrument) ion beam
SEM. The FIB consists of Ga+ ions, accelerated to, for column mounted 54 deg from the vertical. For EBSD, a
example, 30 keV. The impact of the beam onto a sample Digiview camera of EDAX-TSL (EDAX/TSL, Draper,
leads to localized sputtering of the target material and UT) is mounted on a motorized slide; this allows the
can therefore be used to perform cuts into the material camera to be driven under the computer control to its
of few nanometers in width. Irradiating a material operating position in the chamber and withdrawn as
surface in grazing incidence allows the preparation of needed. An energy-dispersive X-ray spectroscopy
smooth surfaces that show little radiation damage.[10–12] (EDX) detector is mounted such that the EBSD and
This technique has been used extensively for the EDX signals can be simultaneously detected. The
preparation of transmission electron microscopy geometric arrangement of the system is schematically
(TEM) samples, as well as for the preparation of flat shown in Figure 1(a). In our system, in contrast to all
surfaces for metallographic investigations. Through the other systems already available, the EBSD camera and
sputtering of subsequent parallel surfaces, serial sections EDX detector are mounted opposite the ion column.
can be created that may be separated by precisely For EBSD, the investigated surface is usually tilted 70
defined distances of some 10 nm up to several microns. deg from the horizontal position. As illustrated in
A variety of articles describe such a technique for the Figure 1(a), this requires a sample tilt between grazing-
preparation of serial sections or TEM samples, e.g., incidence ion milling and EBSD of 34 deg. We therefore
References 12 through 15. call this setup the ‘‘tilt’’ setup, due to the angular
For observation of the microstructures of the serial limitations of the stage currently used and due to the
sections, orientation microscopy using the EBSD tech- fact that, in the EBSD position, the EBSD camera is
nique in the SEM is an excellent method.[16–18] Here, positioned very close to the sample (approximately
different phases and grains can be clearly recognized and 20 mm) and the sample must be mounted on a relatively
quantitatively described. Grain boundaries can be high pretilted stage. Therefore, the sampled area cannot

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 39A, FEBRUARY 2008—375

 Fig. 2—Microstructure of the investigated pearlitic steel sample after
milling (no image has been taken before milling). Image shows the
edge of the sample as viewed in EBSD position (investigated area at
20 deg to the primary beam). The cross on the right side of the mil-
led area is the alignment marker. The two deeper milled areas, left
and right, have been milled with a coarse beam, to avoid shadowing
of the EBSD camera.

is illustrated in Figure 1(c): the initial sample prepara-

tion must produce a fine polished surface. For milling on
Fig. 1—Schematics of the geometry of the 3-D procedure. (a) Cross the free surface, the sample is tilted such that the ion
section through the instrument chamber, showing the sample posi- beam enters the surface at an angle of 10 to 20 deg. As in
tions for milling and EBSD analysis in the tilt setup. To change be- the GIEM method, the sample is subsequently tilted by
tween the two positions, the sample need only be tilted and moved 34 deg, and EBSD is performed on the milled surface.
in the y direction. During milling, the EBSD camera is retracted to
the chamber wall. Working distance for milling is 8 mm; for EBSD,
The advantage of this method is that virtually any
it is 13 mm. The camera is positioned at a distance of 22 mm from volume close to the sample surface can be investigated.
the sample. (b) and (c) Schematics of different variations of the tilt The disadvantage is that the milling area needs to be
geometry: (b) the GIEM method and (c) the LISM method. For significantly larger than in the case of GIEM, in order to
both schemes, the black cross indicates the marker for image align- avoid shadowing of the EBSD screen.
ment that is milled into the sample before the process begins. (d)
The rotation geometry. Both the EBSD and milling positions are In an alternative geometrical setup that has been
shown together. A tilt-free setup requires a pretilted holder mounted adopted by other groups (e.g., Reference 21)—we call it
on a pretilted stage. the ‘‘rotation’’ setup—the EBSD camera is situated
below the FIB gun. The movement from milling to
EBSD position has then to be accomplished by a 180-deg
be brought into the eucentric tilt position of the stage. sample rotation and some x and y movements, if the
Consequently, each sample tilt must be combined with a observed position is not precisely on the rotation center.
y movement of the stage. If the sample is mounted on a pretilted holder and the
Using the tilt setup, two measurement strategies can stage is additionally tilted such that the rotation axis is
be realized. The first, called the grazing-incidence edge- positioned halfway between the EBSD and the milling
milling (GIEM) method, is illustrated in Figure 1(b). position, then it is possible to obtain a situation in which
The sample is prepared by grinding and polishing to no further sample tilt is required to change between the
produce a sharp rectangular edge. Milling is then two characteristic positions. This situation, illustrated in
performed under grazing incidence to one surface at Figure 2(c), may allow the achievement of a higher stage-
this sample edge. After milling has created a smooth positioning accuracy than does the tilt setup; however,
surface at the desired distance from the sample surface, this advantage is not really evident, as we will show by a
the sample is moved into EBSD position by a 34-deg tilt comparison of the two possible setups in Section IV.
and the appropriate y movement of the stage. The For both setups, the size and shape of the milled area
EBSD is performed on this surface. After EBSD has is determined, on the one hand, by the size of the
completed, the sample is returned to the milling position investigated surface. On the other hand, it has to be
and the next cut is performed. This method has the considered that the edge of the milled area projects a
advantage of being easy to set up and allowing the shadow onto the EBSD camera. The shadow increases
investigation of large volumes. In contrast, the micro- with increasing distance of the milled area from the free
structure to be investigated must be close to the edge, sample surface and decreasing distance between the
which is not always easily achieved or possible at all. In EBSD camera and the sample surface. For high-speed
this case, a second, alternative method may be applied, EBSD, a large acceptance angle of the diffraction
the low-incidence surface-milling (LISM) method, which pattern is beneficial. At a camera position of 20 mm,

376—VOLUME 39A, FEBRUARY 2008 METALLURGICAL AND MATERIALS TRANSACTIONS A

we obtain an acceptance angle of approximately 120 (9) Once milling is finished, the FIB gun is switched
deg. This means that the milled area must be bound by off, if requested, and the stage is moved back to
at least 60-deg angles on all three edges, in order to the predefined EBSD position.
allow shadow-free diffraction patterns (Figure 13). This
Automatic 3-D orientation microscopy puts high
additionally milled volume may significantly increase
demands on the measurement process: (1) the stage
milling time at large milling depths or when applying the
positioning has to be accurate enough to allow rede-
LISM method. A reasonable alternative is to remove the
tection of the marker in each position in the limits of
shadowing areas with a large ion beam current and mill
the beam shift controls, (2) the precision of the stage tilt
only the investigated surface with a small and, therefore,
must be high enough to avoid any visible distortion of
finely focused ion current. This is accomplished in our
the field of view, (3) stage drift after sample reposition-
system by setting up a milling list that holds a number of
ing must be minimized and the stage must reliably come
user-defined milling areas.
to a complete stop in the range of tens of seconds, (4)
For both measurement strategies, a position marker is
the position-correction algorithm must be precise
required that allows accurate positioning of the sample
enough to achieve the desired resolution, and (5) the
after movement from EBSD observation to milling or
SEM and the FIB must work stably over long mea-
vice versa. As a position marker, we use a cross that is
surement times. All of these demands are fulfilled on the
milled next to the measured area into the sample surface.
system presented here. We discuss the limits in more
For the tilt setup, a single marker is sufficient, because
detail in Section IV.
only translations are needed (no sample rotation is
applied). In contrast, for the rotation setup, two
markers are used. The accuracy of sample positioning C. Software for 3-D Analysis
will be discussed further in Section IV.
For the analysis of 3-D data, rudimentary software
has been developed thus far. It allows any position of
B. Automatic 3-D Orientation Microscopy the measured volume in space to be viewed through the
Fully running the 3-D milling and EBSD cycle sequential display of 2-D microstructure images. For
automatically requires a number of instruments and preparation of the individual slice images, the batch-
processes to be synchronized: for orientation micros- processing function of the EDAX-TSL OIM Analysis*
copy, the SEM, the EBSD program, the camera slide
controller, the camera controller, and the electron image
*OIM Analysis is a trade mark of EDAX/TSL, Draper, UT.
drift correction program need to cooperate. For sec-
tioning, the FIB and the FIB image drift correction
program are required. Finally, for changing between the software package is used. This function allows the
milling and the EBSD position, the sample stage needs application of similar operations to any number of
to be accurately controlled. All participating processes measured data sets—for example, the rotation of the
are controlled by one 3-D measurement program. Once reference system, cleanup, map construction, and saving
the software is configured, it runs the milling and EBSD of images to disk.
cycle until a predefined number of cycles has been Perhaps the most important advantage of 3-D orien-
reached. A cycle consists of the following steps. tation microscopy over 2-D systems is the possibility of
(1) The EBSD camera is moved into position in the determining the crystallography and morphology of
chamber. interfaces. Our 3-D software, therefore, contains a tool
(2) The position marker is detected by image cross that allows the manual selection of grain boundaries or
correlation and the sample brought to its reference other interfaces. Subsequently, the local grain-boundary
position using beam shift functions. normals in the sample and the crystal reference systems
(3) A reference image is taken in the EBSD position. are calculated.
(4) Orientation mapping is performed using the prede- An important precondition for the clear rendering of
fined camera control parameters. 3-D data is the precise alignment of the slices. Although
(5) After orientation mapping has been fully accom- this is achieved to a large extent during the data
plished, the camera is retracted, the stage is moved acquisition process, the data still show misalignment
to the predefined milling position, and the FIB gun between the slices. We, therefore, developed an align-
is switched on if it was off. ment tool that is based on the calculation of a
(6) The microscope is switched to FIB mode and the correlation function:
position marker is detected by cross correlation. A X
Cij ¼ jðu1; U; u2Þt;x;y
ðu1; U; u2Þw;xþi;yþj j
shift of the FIB beam brings the milling area into ij
reference position.
(7) A reference image is taken, using the FIB beam to which means that the values of the three Euler angles at
document the milling progress from the FIB posi- a position x,y in the template orientation map are
tion. subtracted one by one from those at the position x + i,
(8) The FIB switches to milling mode, steps the beam y + j in the working map. The position at which the
forward to the desired z-step size, and performs sum of all differences, Cij, is at a minimum produces
the milling list. the best alignment. Note that, before Cij is calculated,

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 39A, FEBRUARY 2008—377

the Euler angles must be reduced to a single symmetric low structure factors. The occurrence of bending and of
variant. the related orientation change is an interesting hint of
Other groups have developed more advanced soft- the physical mechanism of phase transformation. A
ware. A well-known software package for 3-D rendering number of models have been proposed to explain the
is, for example, the program IMOD (IMOD package is pearlite formation, e.g., References 28 through 30. A
Copyright (c) 1994–2006 by the Boulder Laboratory for 3-D investigation of the arrangement of pearlite lamellae
3-Dimensional Electron Microscopy of Cells and the yields new and otherwise inaccessible information. First,
Regents of the University of Colorado), which is based the direction of maximum orientation gradients may be
on the construction of 3-D wire-frame models from 2-D determined and spatially correlated with the strength of
cuts.[26] We have used the software ourselves for the lamella curvature. Second, the ferrite-cementite phase
presentation of 3-D data;[9] however, a serious disad- boundary might be precisely characterized with respect
vantage is the enormous manual work required for to the ferrite grain-boundary normal.
model preparation. Others have developed software that For our 3-D investigation, a simple high-carbon steel
allows the selection of 3-D grains and the determination Fe 0.49 pct C sample has been selected. The sample was
of volume properties of these grains, e.g., Reference 21. heated to 950 C and then furnace cooled at a rate of
A very powerful 3-D representation of grain boundaries 0.1 C/s. Figure 2 shows the remaining microstructure of
has been presented by Rowenhorst et al.,[27] who use a the pearlite colonies after milling. The cementite lamella
coloring of grain boundaries according to the position (white) had a width of 100 to 200 nm and a distance
of the local crystallographic normal vector in a standard varying between 200 and 400 nm. The measurement
orientation triangle. system has been set up for a GIEM measurement. Fifty
slices of 30 · 20 lm2 were milled with a step size of
100 nm. Orientation mapping was performed using a
III. APPLICATION EXAMPLES lateral step size of 100 nm as well. Only the ferrite phase
was indexed, because cementite did not produce any
In this section, four selected examples of 3-D mate- indexable patterns, but appeared as simply a degradation
rials analysis are presented, each measured using the of the ferrite patterns. The time for one complete cycle
described system in tilt setup. The selection of the was approximately 35 minutes, including 16 minutes for
samples has been made such that the strengths and the milling and 15 minutes for EBSD mapping.
limits of the system can be well demonstrated. The Figure 3 displays the result of the measurement in
strengths lie clearly in the ease of operation and in the the form of a 3-D orientation map, in which the color
spatial resolution—in the precise definition of micro- of each voxel (volume pixel) is a combination of the
structure, which allows, for example, deformed or diffraction pattern quality, which defines the brightness
multiphase material to be characterized. The limits are of the voxel and the orientation of the measurement
confined by the relatively small investigable volume and plane normal this, in turn, defines the color, according
by possible material/microstructure alteration due to ion to the inverse pole figure (IPF) color triangle. Though
beam irradiation. cementite could not be indexed, the reduced ferrite
For all measurements presented, the ion beam column diffraction pattern quality indicates the position of the
was operated at an accelerating voltage of 30 kV, with lamellae. The figure shows the high alignment accuracy
beam currents between 100 and 2 nA. For EBSD, an of the measurement: at many positions, it is possible to
electron accelerating voltage of 15 kV was used. The
EBSD camera was always positioned approximately
20 mm away from the sample. All EBSD measurements
were performed at a high measurement rate of 65 to 70
patterns per second. This could only be accomplished by
minimizing the pattern indexing calculation time. To
this end, in multiphase materials, only the most simple
phase was indexed online. The complete indexing for all
phases was then performed offline from the recorded
Hough-peak data, using the full crystallographic infor-
mation.

A. Pearlite
Pearlite is a lamellar arrangement of ferrite (a-Fe with
a bcc lattice structure) and cementite (Fe3C); in 2-D
investigations, pearlite often exhibits the characteristic
bending of the lamella structure that is related to a
significant orientation change of the ferrite and likely
also of the cementite. It is not clear whether the Fig. 3—The 3-D microstructure of a pearlite block. The color is
composed of image quality (gray value) and a color code for the
information about the cementite is accurate, because crystal direction parallel to the x axis of the sample. The outer block
the crystal orientation of cementite is very difficult to indicates the full measured volume. A part has been cut off because
measure by electron diffraction techniques, due to its of low pattern quality in these areas.

378—VOLUME 39A, FEBRUARY 2008 METALLURGICAL AND MATERIALS TRANSACTIONS A

track the pearlite lamella through all three dimensions. The results of some first interface investigations are
Nevertheless, in areas in which the lamellae are close demonstrated in Figure 4: two colonies of pearlite
together, their exact direction is not clear. We discuss lamellae, 1 to 6 and 7 to 10, have been investigated.
problems of alignment in Section IV. The strong As indicated by the pole figures in Figure 4(b), in each
orientation change across the volume is displayed in of the lamellae groups, the ferrite crystals continuously
Figure 4(a), in which the color code denotes the rotate about a (111) plane. The rotation pole, however,
misorientation angle of a given voxel with respect to is different for both groups. The positions of the
a reference orientation. As reference, the orientation of lamellae normals relative to the sample reference frame
the lamella-free area in the upper back corner is are displayed in the pole figure in Figure 4(c) and,
selected. The maximum indicated misorientation is relative to the ferrite crystal orientation, in IPFs in
15 deg. Figure 4(d). The IPFs show that the ferrite phase
Figure 3 shows that the observed pearlite consists of boundaries are close to (111), albeit with a significant
two grains and several pearlite colonies. While strong deviation of up to 16 deg. It has been shown by TEM
continuous ferrite orientation gradients run parallel to investigations that the atomic habit plane of ferrite-
the pearlite lamellae, generally no continuous orienta- cementite lamellae may be (1 2 5), (1 1 2), or (0 1 1)
tion change occurs perpendicular to the lamellae; in (References 31 and 32, for example), depending on the
contrast, small-angle grain boundaries are found to lie orientation relationship between both phases. While the
mainly parallel to the lamellae. The convergence of the (111) plane determined here deviates considerably from
orientation gradient with the lamella position shows that these results, it has, in fact, been shown that there is not
the orientation gradient is directly related to the growth necessarily any correspondence between the atomic
process of the lamellae. habit plane and the macroscopic one.[33] Furthermore,

Fig. 4—Ferrite-cementite interface analysis: (a) microstructure with some cementite lamellae marked; (b) (111) pole figure of the area of lamellae
7 to 10 (left) and 1 to 6 (right), indicating that each pearlite colony is characterized by one common (111) pole as rotation axis, with the com-
mon pole marked by a circle; (c) stereographic projection of the cementite lamellae plane normals; and (d) IPF of the crystallographic ferrite
plane normals of the cementite lamellae. For both colonies, ferrite plane normals close to (111) are found, although with a significant deviation.
For (c) and (d), each lamella position has been measured several times, as indicated by similar symbols connected by lines.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 39A, FEBRUARY 2008—379

from the present results, it seems that there is also no before milling. The 3-D orientation microscopy was
clear relationship between the crystal orientation and performed using the GIEM method. A step size of
the macroscopic habit plane. 100 nm in all dimensions was selected and 60 maps
covering a volume of 83 · 16 · 6 lm3 were measured
within less than two days. The results are shown in
B. ‘‘Nanocrystalline’’ NiCo Deposits
Figure 6(a), in the form of a material block colored
Electrodeposition from aqueous solution has been according to the IPF in a direction perpendicular to the
intensively studied as a method to create metallic films growth direction. The usual large-angle grain bound-
with a nanocrystalline microstructure.[34–39] The deposi- aries are colored black, while all possible twin bound-
tion parameters, such as solution composition, current aries are displayed as white lines. In the center of the
density, and bath movement, strongly influence the block, a part has been cut out to display a typical twin
developing microstructure. The exact mechanisms of triple. In Figure 6(b), the block is cut along the plane
microstructure formation are still largely unknown, in that contains the twin triple indicated in Figure 7(a).
part because a precise characterization of the micro- The triple is not growing straight, but is visibly quite
structure has rarely been undertaken. In a recently heavily curved. Together with this curvature, the
published work,[40] we applied EBSD-based high-reso- orientation of the triple changes, as shown in
lution orientation mapping to obtain a deeper knowl- Figure 6(c), which displays the deviation

 of the macro-
edge of the microstructure of electrodeposited thin films scopic growth direction from the 1120 crystal direc-
of NiCo. Most of the investigated films consist of a large tion. The crystal orientation is continuously deflected by
volume fraction of columnar crystals with cross sections the regular occurrence of lattice defects of some sort.
in the submicrometer range but lengths on the order

of The nature of these lattice defects, visible in form of
more than 30 lm. For all columnar grains, the 1120 inclined stripes in Figure 6(c), is not yet clear. In the
direction is close to the growth direction of the crystal center of the measured block, the
twin triple consists of
column. Only in niches beside these crystals do areas of three grains with two 56-deg 1120 twin relations

truly nanocrystalline material exist. We found, further- between them and an approximately 66-deg 11 20
more, that the elongated crystals are, in most cases, relationship on the remaining boundary (Figure 7).
growing in triples of three parallel crystals related to From our earlier 2-D investigations,[40] we proposed
each other by a twin orientation relationship. The the following formation mechanism for the triple: a
existence of these triples can be seen on the free grown primary crystal (1) twins independently with the same
surface of the electrodeposited material in the form of twin relationship into two twin
variants,
 (2) and (3) such
threefold pyramids. We suspect that the existence of that all three crystals show a 1120 direction parallel to
twins stabilizes the growth of columnar grains by the the macroscopic growth direction optimum for quick
formation of low-energy boundaries. growth. The boundary between the crystals
(2) and (3) is
Analysis using 3-D orientation microscopy was com- then automatically fixed to a 66-deg 1120 relation-
pleted on a full 100-lm wide cross section of an ship. In the light of the new 3-D data, this explanation
electrodeposited material. The sample was manually no longer seems totally correct. Following the triple in
ground and polished on the cross sections to produce a the growth direction, we first notice that the grains
(1)
sharp rectangular edge. Figure 5 shows the sample and (2) exist as twins with a sharp 57-deg 11 20
relation already when they first enter into the measure-
ment area (marked as point A in Figure 6(b)) (note that
grain (2) is not visible in the figure, as it is on top of
grain (1)). At point B, grain (3) nucleates and grows for
the first approximately 10 lm as a very thin tube
parallel to grains (1) and (2). At the position at which
the growth direction), grain (3) also expands and grows
larger (point C). At this position, we applied the 3-D
analysis program to determine the grain-boundary
characteristics. The results are displayed in Figure
 7:
between grains (1) and (2), we find the 1011 twin
boundary, typical for these twins.[41] Grain (3), although
it shows a precise twin orientation relation with grain
(2), does not have a twin boundary with the latter, i.e.,
the plane indices on both sides of the grain boundary
are different, the index of grain (2) being very precisely
the (0001) basal plane and the other index being some
quite highly indexed one. The same is true for the
boundary with grain (1), although this grain boundary
Fig. 5—Secondary electron image of the cross section of the NiCo even does not show any typical twin orientation
thin film ready for 3-D orientation microscopy. The investigated vol- relationship. We conclude that grain (3) is a twin of
ume is indicated by a black line frame; the growth direction is indi-
cated by an arrow. The rectangular edge has been prepared by grain (2), but its boundary is incoherent and therefore
mechanical grinding and polishing. Note that, due to grinding, the of a quite different quality than that between (1) and
typical growth surface has been erased. (2). We suppose, furthermore, that the (0001) basal

380—VOLUME 39A, FEBRUARY 2008 METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 6—3-D orientation maps from a NiCo electrodeposited film. Measurement of voxel size is 100 · 100 · 100 nm3. (a) Full measured orienta-
tion map colored according to the IPF of the x direction. General large-angle grain boundaries (w > 15 deg) are displayed in black; all types of
twin boundaries are in white. A part is cut out in the middle, to display the position of one twin triple, marked by a circle. (b) Orientation map
as in (a), cut along the plane marked in (a). The markers A, B, and C indicate particular positions in the formation of the twin triple. Note that
the scattered white pixels on top of the crystal marked A occur due to the twin boundary being almost parallel to the cutting plane. (c) Orienta-
tion map colored according to the deviation of the crystal h11 to 20i direction to the macroscopic growth direction Y. Areas in gray indicate
crystals with fcc crystal structure.

planes that appear in the triple grain arrangement form

low-energy boundaries both in the case of the incoher-
ent twin ((2)–(3)) as well as in case of the less specific
large-angle grain boundary ((1)–(3)).

C. Deformation and Crystallographic Orientation

Patterns below Nanoindents in Cu Single Crystals
As another example of the application of the new 3-D
orientation imaging method, we conducted a 3-D study
at submicrometer-scale resolution on the microstructure
and crystallographic texture formed during nanoinden-
tation in single crystals. Nanoindentation is a method
increasingly used to probe the local mechanical proper-
Fig. 7—Model of the twin triples frequently observed in NiCo ties of materials. However, the complex deformation
deposits, as determined by 3-D orientation microscopy. The mea- mechanisms taking place during this process are only
surements allow determination of the crystal orientations, and mis- poorly understood; it is, therefore, not clear how, for
orientations and the crystallographic indices of all grain boundaries. example, the measured data can be related to macro-
Crystals (1) and (2) form a coherent 57-deg h11 to 20i compression-
type twin. Crystal (3) is in a noncoherent twin relation to crystal (2), scopic properties of a material.
but it has no special relationship with (1). Note the occurrence of In a set of recent articles, the group of Larson
basal planes as boundary planes between (2) and (3) and (1) and (3). et al.[42–45] presented the results of the microstructure

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 39A, FEBRUARY 2008—381

evolution below microindents, using a nondestructive investigations of the deformation and texture patterns
3-D synchrotron diffraction method. In their work, the below indents were not reproduced in full 3-D maps, as
authors observed a systematic deformation-induced ori- was the case in this study, these concentric reorientation
entation pattern below indents in Cu single crystals with a rings were not yet reported in the literature. We interpret
(111) indentation plane normal. The experimentally the constant diameter of these rings as being related to
observed patterns were characterized by outward rota- the direction of the radial vector on the conical surface
tions at the rim of the indent (tangent zone of the indent) of the indenter, creating at every surface radius a
and inward rotations directly below the indent close to different deformation direction relative to the crystal
the indenter axis. The results were discussed in terms of and, therefore, activating different sets of dislocations. A
KrönerÕs concept of geometrically necessary dislocations more detailed analysis will be given in a subsequent
(GNDs). Zaafarani et al.[25] investigated orientation article.[49]
patterns below nanoindents in (111) Cu crystals using a
manual 3-D EBSD method. These experiments revealed
D. Fatigue Crack in Aluminum
a pronounced deformation-induced 3-D lattice rotation
field below the indent. In the (112) cross-sectional planes The formation of fatigue cracks is an important
below the indenter tip, the observed deformation- problem in aluminum alloys used in high-cycle fatigue
induced rotation pattern was characterized by an outer conditions, e.g., in aircraft construction elements. In
tangent zone with large absolute values of the rotations order to optimize alloys against fatigue, the mechanisms
and an inner zone closer to the indenter axis with small of damage have to be well understood; this is particu-
rotations. Additional similar studies on the microstruc- larly the case with regard to microstructure and crys-
ture evolution below small-scale indents were also tallography. Crack formation in the aerospace alloy Al
published by other authors.[46–48] 7075-T651 has been intensively studied by Weiland
In this work, we report on a more systematic study et al.[50] They found that cracks start at particles close to
along these lines. The conical nanoindents were placed the surface and grow into the bulk material. Since a
into a (111) Cu single crystal using different forces, crack is a 3-D feature, 3-D orientation microscopy is an
namely, 4, 6, 8, and 10 mN. ideal tool for studying its microstructure and crystal-
Figure 8 shows a 3-D view of the volume around the lography. To a certain extent, the strain distribution
four different (111) indents. A sequence of sections for along the crack path may also be revealed from the
the 3-D orientation maps were taken perpendicular to small orientation changes created by dislocations emit-
the indenter axis. Figure 8(a) shows the EBSD pattern ted from the crack tip during crack growth. For the
quality in the vicinity of the indents, in terms of a gray observation of cracks nucleated at particles, the GIEM
scale. The pattern quality is a general measure for the procedure is inadequate for the following reasons: (1) it
local defect density. Figure 8(b) shows the kernel is difficult to prepare a sample such that an appropriate
average misorientation (KAM) as a measure for crack is positioned close to a sample edge, (2) cracks
strongly localized orientation gradients; it also indicates developing at edges may behave very differently from
the density of GNDs. It is calculated as the average of those developing on the free surface, and (3) cracks at
all the misorientation angles of the pixels from a ‘‘ring’’ edges are frequently not observed in full. The ideal
surrounding a center pixel and this pixel at the center of method is, therefore, LISM. Figure 9 shows the shape of
the kernel. The KAM maps indicate the occurrence of the investigated sample and the measurement geometry
local crystal rotations. Figure 8(c) reveals the orienta- that was used.
tion difference relative to a common reference orienta- For the present study, 38 slices were cut and observed
tion. Hence, the latter image reveals the absolute at a location near the center of the trench. The milling
magnitude of the deformation-induced lattice rotations, area was 34-lm wide and extended over a height of more
which amounts to up to 15 deg. Figure 8(d) shows in than 50 lm for the last slices. Each slice had a thickness
more detail the lattice rotations directly under the of 200 nm, milled at 15 deg into the surface. Orientation
indenter axis in a (112) plane perpendicular to the microscopy was carried out with a lateral step size of
indented (111) plane: the sense of rotation changes 200 nm. In summary, a volume of 34 · 50 · 7.6 lm3
several times, depending on the position. In accordance (x y z) has been investigated with a 3-D step size of
with the mirror symmetry parallel to the (011) face, 200 · 200 · 200 nm3.
lattice rotations are mirrored on opposite sites of the Figure 10(a) shows a cracked particle on the surface
indenter. Following a tangential vector along the dashed of the material, recorded with a backscatter detector.
circle in Figure 8(c), indent 3, the magnitude of the The particle has a surface diameter of approximately
lattice rotation is highest in the ±[011] directions and 10 lm. Figure 10(b) is a micrograph taken in EBSD
slowly fades away while approaching the ±[112] direc- position after eight cuts with a secondary electron
tions. The sense of rotation remains constant if one detector. On the far right side, a portion of the
follows the tangential vector. alignment marker is visible. To the left and right sides
A particularly interesting feature is visible in of the observation area, small stripes are milled with a
Figure 8(b), which shows the existence of two concentric large ion current, in order to avoid shadowing. The
rings of high KAM values, indicating a high density of particle is still visible in the image, but it is about to
GNDs. The rings have the same diameter, independent disappear. Figure 10(c) shows the area after 21 cuts: the
of the indentation load and depth, but they appear in particle has completely disappeared, but the crack,
different depths under the surface. Since all previous visible in Figure 10(a) as a fine white line, continues

382—VOLUME 39A, FEBRUARY 2008 METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 8—3-D orientation maps of 4 nanoindents produced with a spherical indenter under loads of 4, 6, 8, and 10 mN (indents 1 through 4). (a)
Diffraction pattern quality map that serves as a rough indication of the lattice defect contents. (b) KAM map, calculated up to third neighbors,
with a maximum allowed misorientation of 5 deg. This map indicates local rotations due to GNDs. (c) Reference orientation deviation map. As
reference, an orientation close to one side of indent 3 is selected, in order to demonstrate the asymmetry of rotations. The cube in the lower left
corner indicates the orientation of the undeformed matrix (i.e., not the reference orientation). (d) Pole figures and detailed 2-D orientation map
of the center part of indent 3, as observed in h112i direction. The colors are selected such that a unique correlation between a position in the
pole figure and a position in the orientation map can be made. Arrows in the map indicate the sense of local lattice rotation.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 39A, FEBRUARY 2008—383

into the matrix material. In Figure 10(d), the appropri- For analysis, the measured orientations were rotated
ate orientation map (as described in the following such that they correspond to the reference system
paragraph) is shown. It indicates that the crack extends indicated in Figure 9. From the orientation data,
over several grains. microstructure maps were created: for the display of
crystal orientations, IPF maps for the normal direction
(ND) and rolling direction (RD) were selected. The IPF
maps are overlaid with a gray value map indicating the
diffraction pattern quality. Figure 11 shows the RD-IPF
map of the entire characterized volume. Due to its
narrowness, the crack is difficult to reveal in the
orientation maps. Nevertheless, the thin area of lower
image quality, highlighted by the black arrows in the
3-D map displayed in Figure 12(a), corresponds to the
position of the crack. It is found that the crack continues
below the particle, mainly inside of the grain that also
contains the particle. It does not follow any obvious
grain or subgrain boundary. The crack extends into a
number of differently oriented grains, as indicated by the
Fig. 9—Schematic of the investigated fatigue sample and position of white arrows in the map in Figure 12(a).
the reference coordinate system prescribed by the cold-rolling pro- In order to determine local misorientation fields, 2-D
cess of the sheet. The figure is drawn from the viewpoint of the
EBSD camera. The indicated ion beam mills into the surface under KAM maps have been calculated from the orientation
an angle of 15 deg; Dr indicates the fatigue strain direction. data. The KAM value may be created by a regular

Fig. 10—Surface view of different steps of the milling process along the crack in aluminum. (a) Backscattered electrons (BSE) micrograph of the
particle on the original sample surface. (b) Milling area after eight cuts, observed with secondary electrons (SE) in EBSD position (milled area
tilted 70 deg). Note that the original surface is under an angle of 15 deg, with respect to the milled area. The particle with crack is still visible.
On the far right side, a part of the alignment marker is visible. The area for orientation microscopy is marked with a rectangle. (c) Detail of the
milled area after 21 cuts. The area used for orientation microscopy is marked by a rectangle. The crack is visible as a very thin white line. The
start and the end of the visible crack are marked by arrows. (d) IPF map for the ND of area (c). (Map is compressed in vertical direction to fit
to the area observed under 70-deg tilt).

384—VOLUME 39A, FEBRUARY 2008 METALLURGICAL AND MATERIALS TRANSACTIONS A

 through C, we will assess the possibilities of the
technique, initially by comparison of the two possible
geometrical setups described in Section II. Subsequently,
we discuss the accuracy obtained by our own system;
finally, we describe the material cases in which the
technique cannot be applied.

A. Comparison of Geometrical Setups

Each of the two possible geometrical setups, the tilt
and the rotation setup, has advantages and disadvan-
tages. The most serious disadvantage of the tilt setup,
which was developed by us, is the requirement to tilt the
relatively heavy stage up or down against gravity. This
leads to some inaccuracy for the tilt angle, which
actually cannot be directly controlled, because it has
virtually no visible effect. There are, however, easy ways
to measure and correct this inaccuracy: the measure-
ment of tilt inaccuracy is accomplished by measurement
of the misorientation between neighboring points in
successive slices. On samples with largely defect-free
crystals, few measurement points will directly yield the
inaccuracy of the tilt angle. On samples with defected
crystals, the required value may be determined to be the
average misorientation between all neighboring points
in successive slices, excluding those that exceed a certain
threshold misorientation. The correction of the tilt
inaccuracy is easy, because the related distortion of the
measurement field is purely linear: a wrong tilt angle
leads to the expansion or compression of the measure-
Fig. 11—3-D view of the full investigated volume. Colors according ment field in the direction perpendicular to the tilt axis.
to the IPF for the rolling direction. Figure 10(d) provides the corre- If the angle of tilt is a (=70 deg for EBSD) and the tilt
sponding color key. The arrow indicates the position of the particle.
inaccuracy is Da, the linear distortion of the measure-
ment area is
arrangement of dislocations—for example, a dislocation ðl0 þ DlÞ=l0 ¼ sin a=sinða
DaÞ
pileup at (sub)grain boundaries—and therefore may be
interpreted in terms of local fields of plastic strain, as where l0 denominates the true length of the measure-
well. A 3-D view of 2-D KAM maps is shown in ment area, and Dl is its distortion due to misalignment.
Figure 12(c). Areas with high KAM values are mainly A correction of this distortion may be easily applied to
small-angle grain boundaries, but the crack positions the final orientation map. As described in Section B, the
also turn out to be areas of higher local misorientations tilt inaccuracy is less than 1 deg on the present
(partly marked with arrows). This may indicate that instrument. This means that the distortion is on the
crack progression leads to the formation of regular order of 1 pct. On a 10-lm-long orientation map, the
dislocations that remain in the material. extreme end of the measurement field may therefore be
displaced by 100 nm, a value which can, in most cases,
be disregarded.
IV. DISCUSSION: ACCURACY AND For the rotation setup, the most serious inaccuracy is
APPLICATION LIMITS that due to rotation. Although this inaccuracy can be
measured directly by observation of two markers (it is
The examples discussed here have shown that 3-D not known to the authors whether this is actually done
orientation microscopy based on serial sectioning with a on these systems), the correction of this inaccuracy is
FIB and observation of these sections by EBSD-based less evident: a correction via beam rotation is not
orientation microscopy fully preserve the powerful possible, because this leads to serious nonlinear distor-
functions of the 2-D orientation microscopy technique tions of the measurement field on the steeply inclined
and extend them into the third dimension. At each surface (cf. the article of Nolze[51]). Also, a software-
volume pixel in space, a multidimensional vector of data based solution is difficult, due to the nonlinear correc-
is obtained that includes the crystallographic phase, the tions required. A second disadvantage of the rotation
crystal orientation, and an indication of the lattice setup is its inflexibility: different pretilted sample holders
defect density via EBSD pattern quality. If determined are required for the GIEM and for every LISM setup. In
also, the elemental composition as measured by simul- contrast, with the tilt setup, any tilt angle for the GIEM
taneous EDS analysis may be included. In Sections A or LISM method can be set freely.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 39A, FEBRUARY 2008—385

Fig. 12—(a) 3-D view of the part of the investigated volume close to the crack path. Colors according to the IPF for the rolling direction.
Figure 10(d) provides the corresponding color key. (b) Pole figure of the front slice of the present view. The orientation of the main grain con-
taining the crack path (indicated by black arrows) is indicated by color. (c) KAM map. Arrows indicate those areas of high KAM values that
are due to cracks. Other areas of high KAM values are due to small-angle grain boundaries.

B. Measurement Accuracy
directions and 1 deg for the tilt (point (2)), which is
Important parameters for assessing the performance sufficient.
of the system are as follows: (1) the spatial resolution, The depth resolution, i.e., the information depth of
(2) the angular resolution, and (3) the maximum size of EBSD, was found to be less than 10 nm.[52] This value,
the observable volumes with regard to the long-term therefore, defines the lower limit of the achievable 3-D
stability of the instrument. We will discuss each of these depth resolution, as crystallographic information from
parameters separately. slices thinner than 10 nm will overlap significantly. This
(1) The spatial resolution of the system results from value is inconsequential, because the slice cutting is less
the lateral resolution achieved on the 2-D sections, the precise than this. This is well known from the prepara-
accuracy of alignment of the sections, and the depth tion of thin foils for TEM via FIB milling, for which it is
resolution. The lateral resolution of EBSD-based 2-D difficult to obtain slices thinner than 50 nm. As for the
orientation microscopy has been shown to be on the alignment of EBSD slices, the accuracy of cutting in the
order of a few 10 nm on a FEG SEM, depending on the automated system is defined mainly by the FIB image
material investigated and the microscope parameters resolution and the accuracy of the automatic image
used.[18] Alignment of the serial sections is achieved by alignment via beam shift. Also, for the FIB image, the
the application of image recognition software that pattern recognition precision is, at best, as good as one
detects the position of a marker in the image and then pixel in the digital image, provided that image resolution
applies a beam shift to align this marker with a is high enough. For a 50-pA beam, which is usually used
previously stored position. Image recognition and align- for image alignment, image resolution is better then
ment is possible with an accuracy of better than 1 or 2 20 nm. Therefore, for a sufficiently high magnification, it
pixels in the digital image. On the present system, the is theoretically possible to reach a depth resolution of
pixel size on the viewing screen is 0.27 mm, which approximately 20 nm. In reality, we found that with the
corresponds to 27 nm at 10,000 times onscreen magni- currently used algorithm the image recognition is usually
fication. This value is very close to the lateral resolution less precise than 1 pixel, i.e., on the order of 3 or 4 pixels.
of the EBSD technique. At 5000 times magnification, a Thus, a FIB image magnification of 20,000 times is
resolution of 50 to 100 nm is achievable. A prerequisite required to reliably obtain a resolution of 50 nm. We
for achieving this resolution is that the alignment believe that this value represents the highest resolution
marker lies within the visible image after each stage currently achievable. Most of the studies performed so
repositioning. The stage must be accurate enough to far have been carried out with 100-nm step sizes in all
guarantee this. On the present system, the stage posi- directions; this gives satisfactory section alignment. It
tioning accuracy is approximately 2 lm in the x and y should, however, be noted that not all image alignment

386—VOLUME 39A, FEBRUARY 2008 METALLURGICAL AND MATERIALS TRANSACTIONS A

problems have been resolved yet, as can be seen in (2) The angular resolution of the classical EBSD
Figure 3: the lamella arrangement of pearlite requires a technique depends mainly on the size and quality of the
particular precise alignment of the slices; otherwise, due recorded EBSD patterns. For high-speed measurements,
to the narrow spacing, different lamella may be grouped it is on the order of 0.5 to 1 deg. The 3-D EBSD may
together by mistake. The continuity of the lamellae is, show a larger value, due to the inaccuracy of the stage
therefore, a particularly critical measure for the precision tilt. By measuring the orientation variation from slice to
of alignment. There are areas in which the lamellae slice inside of a well-recrystallized grain of constant
appear perfectly aligned, particularly on the upper part orientation, we found an average imprecision of tilt of
of the measured block. However, other areas, at the 0.5 deg. For typical orientation microscopy applica-
lower part of the milled area in particular, show lamellae tions, therefore, the angular resolution of 3-D EBSD is
that cannot be uniquely grouped together. Three reasons as good as that of the 2-D technique. Since the angular
for this misalignment have been identified. Two are misalignment between two slices affects all pixels in a
related to specimen drift during the relatively slow EBSD map in a similar way, it can be approximately deter-
measurements—first, due to charging of the phosphor mined—and thus corrected—by the calculation of the
screen of the EBSD camera and second, due to changes average misorientation of all neighboring points in
in environment (mainly room-temperature variations). neighboring slices, excluding all points that have a
These problems have been resolved nowadays by the misorientation larger than a certain threshold value.
exchange of the camera, which is now equipped with a (3) The maximum size of the observable volume
thinner phosphor layer (Hikari camera, EDAX/TSL, depends mainly on the material and on the time
Draper, UT) and by a better control of the air-condi- investment one wants to make. While the width of the
tioning system. The third reason is due to imperfectly milled surface is virtually unlimited, the height of the
flatmilled surfaces. In fact, FIB milling never leads to observable surface depends on the sensitivity of the
absolutely flat surfaces; rather, it leads to slightly curved material to curtaining. As indicated in Figure 13, from a
surfaces, which also sometimes show ‘‘curtain’’-like certain distance to the top edge of the milled area,
structures or small horizontal steps. Figure 13 displays intense curtaining prevents the formation of good EBSD
a sketch of a milled surface. A slight variation in patterns and renders the depth position very imprecise.
curvature, for example, due to a small variation in The distance at which curtaining starts depends strongly
sample tilt, may lead to a misalignment of slices, as has on the material, its microstructure and crystal orienta-
been described in Section IV-A. Optimization of the tion, the quality of the top surface, and the milling
milling process (optimum beam alignment and focusing, current. To date, the largest curtain-free height obtained
optimization of the milling time and the milled area) is 30 lm on a large-grained Fe3 pct Si alloy milled with
leads to an effective reduction in these problems. a 500-pA beam current. The depth of milling, like the
width, is, in principle, unlimited. However, the further
one cuts into a sample, the more material has to be
additionally removed, in order to prevent shadows on
the EBSD screen and the serious redeposition of milled
material on the sides of the milled volume. The volume-
limiting parameter is, therefore, the milling time: a
curtain-free area of 20 · 20 lm may be milled 1 lm into
a low-alloyed steel sample in approximately 15 minutes,
with a beam current of 500 pA. The sides of the volume
may be milled with a larger beam current at a higher
rate, since they do not require any surface quality. We
believe that the largest volume investigable by the
method presented here is on the order of w · h · d =
50 · 50 · 50 lm3. Other groups have shown that, by
using a particular sample preparation (e.g., a ‘‘finger’’
prepared by mechanical machining), still larger volumes
can be obtained.[22]
Large volumes or high-resolution measurements
require long measurement times and, therefore, the
long-term stability of the instrument, in particular of the
FIB gun.** With the Zeiss XB 1540, the stability of the
Fig. 13—Schematic drawing of the milled area indicating some geo-
metrical features and potential sources of inaccuracy. In order to
avoid shadowing the EBSD camera, the milled area has to be boun- **Note, for example, that decreasing the step size from 1003 to
ded by side walls with angles ‡60 deg. The large side walls may be 503 nm3 while keeping the investigated volume constant increases the
milled with a high-beam current, while the investigated area is milled measurement time by a factor of 8, i.e., for example, from 1 day to
with a small-beam current. A small-beam current leads to a sharper more than a week.
beam profile and, therefore, to a smaller curvature of the milled sur-
face. Inaccuracy of sample tilt (here indicated as a change of beam
incidence) may lead to a significant misalignment of subsequent sli-
ces, in particular because the ‘‘natural’’ curvature of the surface is
FIB gun is high enough to perform measurements
changed. Curtaining may also lead to local misalignment of slices. lasting even as long as 10 days, providing the FIB gun is

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 39A, FEBRUARY 2008—387

switched off during the EBSD measurement. After inaccuracies can be easily corrected for by software,
switching the FIB gun on, usually only minutes are because they create only linear image distortions. The
required to obtain stable emission behavior. The stabil- image distortions created by the rotation setup, in
ity of the gun is controlled by the automatic 3-D contrast, are nonlinear and, therefore, are significantly
software, and milling is performed only after a stable more difficult to correct. Another clear advantage of the
emission current has been obtained. tilt method is given in terms of its larger measurement
flexibility: we devised two different milling strategies
that allow 3-D microstructures close to the edge of a
C. Materials Issues
sample or anywhere on a free surface to be investigated.
For many metals and intermetallics investigated so far We have shown in the present article that a pixel
(various ferritic, austenitic, and martensitic steels, nickel, resolution of 100 · 100 · 100 nm3 is obtained as a
copper, aluminum, FeAl, and CuZnAl) excellent diffrac- standard, but 50 · 50 · 50 nm3 seems to be reachable
tion patterns were obtained from milled surfaces, and the (we consider this to be a more realistic estimation than
microstructures observed on these were similar to those the smaller values claimed by other groups). The largest
prepared by other methods. Nevertheless, various prob- volumes observable are on the order of 50 · 50 ·
lems exist. One is the amorphization of certain materials. 50 lm3.
We have observed that Nb-containing Laves-phase inclu- The various examples presented in the work give an
sions in Fe3Al, which give clear diffraction patterns after overview of the field of application of the technique:
mechanical preparation, do not give any patterns when conductive materials with grain sizes on the order of
prepared by milling.[9] Similar effects have been observed 100 nm to 5 lm are very accessible via the technique. In
on certain carbides. It is known from silicon that Ga+ ion particular, heavily deformed structures can also be
beam milling with 30 kV ions creates an amorphous layer investigated. Furthermore, we consider the technique to
approximately 20 nm in thickness,[53] which is too thick be well suited for the study of all problems in which local
for EBSD. On TEM sample preparation, low-kV ion interface properties are of importance. In contrast, we
polishing subsequently reduces the thickness of the have had so far no success with the application of the
amorphous layer significantly. It is possible, but not yet technique to nonconductive materials, which is due to the
tested, that this would also help with 3-D orientation fact that the charging of the sample leads to important
microscopy. A second, serious problem has been observed drift problems. Furthermore, there are a number of
when milling steels with metastable residual austen- metallic or intermetallic materials where beam-induced
ite—for example, low-alloyed TRIP steels. Here, all material changes make the application of the technique
austenite transforms under the influence of the ion beam difficult or impossible. Finally, it should be mentioned
into ferrite. Up until now, this transformation could not that the technique is destructive and, therefore, not suited
be avoided, either by small-beam currents or by low-kV for in-situ experiments, for example. Here, the newly
milling or a milling with Ar+ ions. Finally, we have developed synchrotron-radiation-based 3-D techniques
observed that the microstructure of heavily deformed are of great interest. In terms of spatial resolution and
aluminum may be destroyed by intense ion beam irradi- observable volume, the synchrotron- and electron-based
ation, which can probably be attributed to the formation techniques appear to be largely complementary.
of a low-temperature melting eutectic.

V. CONCLUSIONS
ACKNOWLEDGMENTS
In recent times, a number of groups have developed
systems for 3-D orientation microscopy based on The authors express their thanks to the people of the
automated serial sectioning via ion beam milling and Zeiss hardware and software development group—in
EBSD-based orientation microscopy in a FIB-SEM. particular, to P. Hoffrogge, P. Gnauck, D. Hubbart, and
The technique is capable of conserving all-powerful H. Walaart—for their support in building up the 3-D sys-
features of the well-known 2-D EBSD-based orientation tem. The authors are also grateful to H. Weiland, Alcoa,
microscopy technique and extends them into the third for initiating the investigations on crack formation and
dimension of space. It yields a multidimensional data delivering the appropriate sample. The authors also
vector for each voxel of the measured volume, including thank T. Takahashi, for the preparation of the pearlite
the crystal orientation, the crystallographic phase, a sample, and N. Zaafarani, MPIE, for the preparation of
value for the lattice defect density, and, if measured by the nanoindents. Finally, the authors thank J. Konrad,
simultaneous EDS analysis, the elemental composition. A. Bastos, and M. Nellessen, co-workers at MPIE, and
Most systems developed so far are based on a rotation P. Scutts, D. Dingley, and other co-workers, EDAX-
setup in which the sample has to be rotated by 180 deg TSL, for their help in developing the 3-D system.
from the EBSD to the milling position. In contrast, we
have developed a system based on a tilt setup that
requires the sample to be tilted, in order to change REFERENCES
between the two positions. We found that the tilt 1. B.L. Adams: Metall. Trans. A, 1986, vol. 17A, pp. 2199–207.
setup may be related to slightly larger sample position- 2. B.L. Adams, P.R. Morris, T.T. Wang, K.S. Willden, and
ing inaccuracies than the rotation setup, but the S.I. Wright: Acta Metall., 1987, vol. 35, pp. 2935–46.

388—VOLUME 39A, FEBRUARY 2008 METALLURGICAL AND MATERIALS TRANSACTIONS A

 3. H.F. Poulsen, S.F. Nielsen, E.M. Lauridsen, S. Schmidt, 28. F.A. Khalid and D.V. Edmonds: Acta Mater., 1993, vol. 41,
R.M. Suter, U. Lienert, L. Margulies, T. Lorentzen, and D. Juul pp. 3421–34.
Jensen: J. Appl. Crystallogr., 2001, vol. 34, pp. 751–56. 29. M. Durrand-Charre: Microstructures of Steels and Cast Irons,
4. B.C. Larson, W. Yang, G.E. Ice, J.D. Budai, and J.Z. Tischler: Springer, New York, NY, 2004.
Nature, 2002, vol. 415, pp. 887–90. 30. G.J. Shiflet: Mater. Sci. Eng., 1986, vol. 81, pp. 61–100.
5. W. Yang, B.C. Larson, J.Z. Tischler, G.E. Ice, J.D. Budai, and 31. D.S. Zhou and G.J. Shiflet: Metall. Trans. A, 1992, vol. 23A,
W. Liu: Micron, 2004, vol. 35, pp. 431–39. pp. 1259–69.
6. A. Preusser, H. Klein, and H.J. Bunge: Solid State Phenom., 2005, 32. G. Spanos and H.I. Aaronson: Acta Metall. Mater., 1990, vol. 38,
vol. 105, pp. 3–13. pp. 2721–32.
7. E.M. Lauridsen, S. Schmidt, S.F. Nielsen, L. Margulies, 33. S.A. Hackney and G.J. Shiflet: Scripta Metall., 1986, vol. 20,
H.F. Poulsen, and D. Juul-Jensen: Scripta Mater., 2006, vol. 55, pp. 389–94.
pp. 51–56. 34. A. Vicenzo and P.L. Cavallotti: Electrochim. Acta, 2004, vol. 49,
8. L. Margulies, G. Winther, and H.F. Poulsen: Science, 2001, pp. 4079–89.
vol. 291, pp. 2392–94. 35. D. Golodnitsky, Y. Rosenberg, and A. Ulus: Electrochim. Acta,
9. J. Konrad, S. Zaefferer, and D. Raabe: Acta Mater., 2006, vol. 54, 2002, vol. 47, pp. 2707–14.
pp. 1369–80. 36. E. Gómez, J. Ramirez, and E. Vallés: J. Appl. Electrochem., 1998,
10. S.V. Prasad, J.R. Michael, and T.R. Christenson: Scripta Mater., vol. 28, pp. 71–79.
2003, vol. 48, pp. 255–60. 37. K. Daheum, D.-Y. Park, B.Y. Yoo, P.T.A. Sumodjo, and
11. J.R. Michael, J. Schischka, and F. Altmann: HKL Technology N.V. Myung: Electrochim. Acta, 2003, vol. 48, pp. 819–30.
EBSD Application Catalogue 2003, HKL Technology, Hobro, 38. A.M. El-Sherik and U. Erb: J. Mater. Sci., 1995, vol. 30,
Denmark, 2003. pp. 5743–49.
12. T.L. Matteson, S.W. Schwartz, E.C. Houge, B.W. Kempshall, and 39. B.Y.C. Wu, P.J. Ferreira, and C.A. Schuh: Metall. Mater. Trans.
L.A. Gianuzzi: J. Electron. Mater., 2002, vol. 31, pp. 33–39. A, 2005, vol. 36A, pp. 1927–36.
13. T.T. Sheng, G.P. Goh, C.H. Tung, and L.F. Wang: J. Vac. Sci. 40. A. Bastos, S. Zaefferer, D. Raabe, and C. Schuh: Acta Mater,
Technol., B, 1997, vol. 15, pp. 610–13. 2006, vol. 54, pp. 2451–62.
14. Z. Cheng, T. Sakamoto, M. Takahashi, Y. Kuramoto, M. Owari, 41. P.G. Patridge: Metall. Rev., 1967, vol. 12, pp. 169–94.
and Y. Nihei: J. Vac. Sci. Technol., B, 1998, vol. 16, pp. 2473–78. 42. W. Yang, B.C. Larson, G.M. Pharr, G.E. Ice, and J.Z. Tischler:
15. H. Jin, P.D. Wu, M.D. Ball, and D.J. Lloyd: Mater. Sci. Technol., MRS Symp. Proc., 2003, vol. 779, p. W5.34.1.
2005, vol. 21, pp. 419–28. 43. B.C. Larson, W. Yang, G.E. Ice, J.D. Budai, and J.Z. Tischler:
16. Electron Backscatter Diffraction in Material Science, A.J. Schwartz, Nature, 2002, vol. 415, pp. 887–90.
M. Kumar, and B.L. Adams, eds., Kluwer Academic/Plenum 44. R.I. Barabash, G.E. Ice, B.C. Larson, G.M. Pharr, K.S. Chung,
Publishers, New York, NY, 2000. and W. Yang: Appl. Phys. Lett., 2001, vol. 79, pp. 749–51.
17. F.J. Humphreys: J. Mater. Sci., 2001, vol. 36, pp. 3833–54. 45. G.E. Ice and B.C. Larson: Adv. Eng. Mater., 2000, vol. 2,
18. D.J. Dingley: J. Microsc., 2004, vol. 213, pp. 214–24. pp. 643–46.
19. G. Spanos: Scripta Mater., 2006, vol. 55, p. 3. 46. Y. Liu, B. Wang, M. Yoshino, S. Roy, H. Lu, and R. Komandri:
20. M.D. Uchic, M. Groeber, R. Wheeler IV, F. Scheltens, and J. Mech. Phys. Solids, 2005, vol. 53, pp. 2718–41.
D.M. Dimiduk: Microsc. Microanal., 2004, vol. 10, pp. 1136–37. 47. K.K. McLaughlin, N.A. Stelmashenko, S.J. Lloyd, L.J. Vandeperre,
21. J.J.L. Mulders and A.P. Day: Mater. Sci. Forum, 2005, and W.J. Clegg: Mater. Res. Soc. Symp. Proc., 2005, vol. 841,
vols. 495–497, pp. 237–43. pp. 3–4.
22. M.D. Uchic, M.A. Groeber, D.M. Dimiduk, and J.P. Simmons: 48. D. Kiener, R. Pippan, C. Motz, and H. Kreuzer: Acta Mater.,
Scripta Mater., 2006, vol. 55, p. 23. 2006, vol. 54, pp. 2801–11.
23. M.A. Groeber, B.K. Haley, M.D. Uchic, D.M. Dimiduk, and 49. N. Zaafarani, D. Raabe, F. Roters, and S. Zaefferer: Acta Mater,
S. Gosh: Mater. Charact., 2006, vol. 57, pp. 259–73. 2008, vol. 56, pp. 31–42.
24. S. Zaefferer, J. Konrad, and D. Raabe: Microscopy Conf. 2005, 50. H. Weiland, S. Cheong, J. Nardiello, S. Zaefferer, J. Papazian, and
Davos, 2005, p. 63. D. Raabe: submitted to J. Fatigue.
25. N. Zaafarani, D. Raabe, R.N. Singh, F. Roters, and S. Zaefferer: 51. G. Nolze: Ultramicroscopy, 2007, vol. 107, pp. 172–83.
Acta Mater., 2006, vol. 54, pp. 1863–76. 52. S. Zaefferer: Ultramicroscopy, 2007, vol. 107, pp. 254–66.
26. J.R. Kremer, D.N. Mastronarde, and J.R. McIntosh: J. Struct. 53. N.I. Kato, Y. Kohno, and H. Saka: J. Vac. Sci. Technol., A, 1999,
Biol., 1996, vol. 116, pp. 71–76. vol. 17, pp. 1201–04.
27. D.J. Rowenhorst, A. Gupta, C.R. Feng, and G. Spanos: Scripta
Mater., 2006, vol. 55, pp. 11–16.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 39A, FEBRUARY 2008—389