Spatially correlated electron microscopy and atom probe tomography: Current possibilities and future perspectives

SMM-11530; No of Pages 8 Scripta Materialia xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

Scripta Materialia journal homepage: www.elsevier.com/locate/scriptamat

Viewpoint article

Spatially correlated electron microscopy and atom probe tomography: Current possibilities and future perspectives M. Herbig Max-Planck-Institut für Eisenforschung GmbH, 40237 Düsseldorf, Germany

a r t i c l e

i n f o

Article history: Received 12 December 2016 Received in revised form 25 January 2017 Accepted 11 March 2017 Available online xxxx Keywords: SEM TEM ECCI TKD APT

a b s t r a c t Performing electron microscopy and atom probe tomography at the same location on the same specimen combines the strengths of electron microscopy, which is primarily the analysis of defects and crystal structures, with the strengths of atom probe tomography, which is primarily the robust, accurate and sensitive three dimensional compositional analysis. This viewpoint article provides a summary of the broad range of electron microscopy techniques that have been performed on atom probe specimens to date. It describes what technique is best suited to address a specific materials science question and finishes with an outlook on possible future developments in the field. © 2017 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

1. Introduction The strengths of atom probe tomography (APT) and electron microscopy (EM)1 are remarkably complementary and many materials science questions require their joint application on the same location in order to be answered. The major strength of APT is the three dimensional (3D) local, quantitative elemental analysis at the near-atomic scale. APT is inherently a three-dimensional technique, and it combines high mass resolution with high elemental sensitivity in the range of 10 parts-permillion, irrespective of the element [1]. This even enables the quantification of trace amounts of light elements in the bulk of a matrix composed of heavy elements. In an APT mass spectrum, ionic species can, in most cases, simply be identified according to their mass-to-charge ratio, which makes interpretation and quantification of absolute concentrations rather robust. In electron microscopes, the main techniques to quantify local chemical compositions are energy-dispersive X-ray spectroscopy (EDX) and electron energy loss spectroscopy (EELS). The elemental sensitivity of EDX depends on the atomic number of the element; the lighter the element, the higher the associated measurement error [2]. Therefore, the concentration of light elements is usually determined through other means, e.g. in transmission electron microscopes (TEM) EELS is typically used because it does not suffer from this shortcoming (although it often has difficulties in detecting heavy elements). The element-specific edges in EELS spectra are superimposed on a strong background caused by the tails of plasmons, single electron E-mail address: [email protected]. Here: both, scanning electron microscopy (SEM) and transmission electron microscopy (TEM). 1

excitations or by other element-specific edges, making the quantification of absolute concentrations challenging. For these reasons, quantitative chemical analysis at the near-atomic scale can be conducted in many cases more easily, and more accurately, by APT rather than by EM. However, electron microscopes are excellent for crystallographic/ structural analysis since electron diffraction and imaging can be performed at the (sub-)nanometer scale. Grain orientations can be measured, the structure of complex phases can be identified based on diffraction patterns, defects such as interfaces, dislocations, slip bands, or stacking faults can be imaged, even the structure of defects or strain gradients can be investigated with spatial resolutions capable of resolving individual atomic columns. Although crystallographic analysis is also possible by APT, this is mainly restricted to the analysis of grain orientations [3–7]. One of the factors that determine the spatial resolution of APT datasets is the regularity of the field-evaporation sequence. In the ideal case, the specimen maintains the shape of a half-sphere on top of a truncated cone throughout the experiment. This ideal behavior is disturbed by local compositional changes or at lattice defects due to the dependence of field-evaporation on grain orientations and on the local bonding state [8]. These effects lead to a reduction of the spatial resolution in the reconstructed 3D atom maps; particularly at locations where the highest spatial resolution is desired — at the defects. Therefore, EM is in most cases better suited for crystallographic analysis of defects than APT. EM and APT are thus complementary techniques, enabling the synergistic characterization of materials at the atomic level. For the investigation of materials science phenomena that involve both compositional and crystallographic changes at near atomic scale, such as the segregation to defects, the onset of nucleation of a new phase or

http://dx.doi.org/10.1016/j.scriptamat.2017.03.017 1359-6462/© 2017 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Please cite this article as: M. Herbig, Spatially correlated electron microscopy and atom probe tomography: Current possibilities and future perspectives, Scripta Materialia (2017), http://dx.doi.org/10.1016/j.scriptamat.2017.03.017

2

M. Herbig / Scripta Materialia xxx (2017) xxx–xxx

mechanically-induced mixing, conducting only EM or APT alone is often not sufficient. Getting a full picture understanding of these phenomena often requires combining both techniques on the same location of the same specimen, i.e. performing spatially correlated microscopy. This manuscript focuses on work that has been performed with an experimental setup described in ref. [9], it details for which materials scientific question which EM technique is best suited and comments on potential future developments in this field. 2. Experimental setup Fundamental to the experimental setup is a modified commercial high-angle single-tilt TEM retainer of the type JEOL EM-21311HTR that carries up to four APT specimens on a TEM grid cut in half [9]. This setup is a further development of existing ones [10–13]; it combines high specimen throughput and exact control of the specimen orientation with a flexible and robust experimental procedure [14]. APT specimens are positioned onto the grid using standard focused ion beam (FIB) lift-out techniques [11,15]. Half of the grid-containing section of the retainer, hereafter referred to as the “grid holder”, was removed to give access to the specimens in the dual-beam scanning electron microscope (SEM)/FIB instrument and in the atom probe (here: FEI Helios Nanolab 600i and Cameca LEAP 3000HR, respectively). Compatibility between the grid holder, the dual-beam instrument and the atom probe was established by customized adapters that enable precise control about the specimen orientation [9]. Single-tilt TEM experiments were done directly using the grid holder in a JEOL JEM2200FS microscope. For double-tilt TEM experiments in this microscope, or for the measurement in a Philips CM20 and a FEI Titan Themis, the grid containing the specimens was transferred into the respective double-tilt TEM retainers using a vacuum pick-up tool. The SEM techniques of electron channeling contrast imaging (ECCI) and transmission Kikuchi diffraction (TKD) discussed in this manuscript were conducted using the above described experimental setup with corresponding adapters for the grid holder. This approach is not confined to the microscopes described here. With the help of customized adapters it can be conducted on other dual beam SEM/FIBs, atom probes and TEMs as well. There is a broad range of EM techniques available to be correlated with APT. The choice of the EM technique depends on the materials science question being answered and on the required spatial resolution. Not all questions have to be addressed by TEM. In many cases the spatial resolution of an SEM is sufficient, which is in general less time consuming than TEM experiments. The following section lists a selection of materials science questions that were answered by correlative EM/APT. The ensuing section details the specific advantages of the various EM techniques when being conducted on APT specimens. 3. Application examples 3.1. Grain boundary segregation One of the core applications of correlative EM/APT is the investigation of grain boundary segregation. For the complete crystallographic description of a grain boundary five parameters must be known — three that specify the misorientation between the adjacent grains and two that specify the orientation of the grain boundary plane [16]. Each of the five parameters has a strong influence on solute segregation, and thus none of them can be neglected for the investigation of this phenomenon. Segregation often involves light elements that may be only slightly enriched/depleted within a distance that is typically below 2 nm from the interface. EM is ideally suited to characterize the crystallography of grain boundaries and APT for characterizing the solutes segregated to them. In the case of coarse grained materials, four crystallographic grain boundary parameters can be detected on a metallographically prepared specimen by electron backscatter diffraction (EBSD). After site-specific preparation of the interface of interest in

the dual beam SEM/FIB, the missing fifth grain boundary parameter (the inclination angle of the interface in the depth of the bulk specimen) can be determined on the lift-out piece using the SEM [9]. When working with grain sizes of less than 50 μm, conducting site-specific lift-outs becomes challenging. In such cases four crystallographic grain boundary parameters can be characterized by conducting non-site-specific liftouts and then either performing EBSD directly on the lift-out [17,18] or TKD directly on the APT specimens [19,20]. In the case of very fine grain sizes, all five crystallographic parameters have to be determined by TEM. The combination of scanning nanobeam diffraction (NBD) to determine grain misorientations, and scanning transmission electron microscopy (STEM) imaging to determine the grain boundary plane orientation was shown to be an efficient and accurate approach to solving this challenging task. This is illustrated in Figs. 1 and 2 on the example of a cold-drawn pearlitic wire. NBD and STEM were performed in a JEOL JEM-2200FS TEM operated at 200 kV. The spot size, step size and exposure time used for NBD were 0.5 nm, 1.5 nm and 40 ms, respectively. APT was conducted in a Cameca LEAP 3000HR, operated in voltage mode at a set-point temperature of 70 K, 15% pulse fraction and 0.5% target evaporation rate. Using scanning NBD it is possible to investigate nanocrystalline materials with columnar grains with 30 nm diameter. With this grain size, each APT measurement usually contains more than 10 grain boundaries. Using this approach it is thus possible to accurately and efficiently quantify the solute excess and the five crystallographic parameters for a large number of interfaces [16]. However, the possibilities of this approach go beyond of what was demonstrated in [16]. One further complex materials science topic that can be investigated with this combination of techniques is heterogeneous nucleation. Understanding this phenomenon is required to further develop the concept of segregation engineering [21,22]. The local crystal structure and chemical composition of the nucleation site as well as the orientation relationship between precipitate and matrix and corresponding chemical gradients must be known to get the full picture. All this information is contained in the combined STEM image/scanning NBD/APT datasets as demonstrated for the case of position “X” in Fig. 2. The NBD pattern reveals that a cementite precipitate has formed at this grain boundary. From this diffraction pattern, and the ones of the adjacent ferrite grains, the orientation relationship between both grains and the precipitate can be extracted. Moreover, the NBD inverse pole figure map allows for extracting orientation gradients. This makes it possible to estimate the density of geometrically necessary dislocations in proximity to the interfaces. The grain boundary plane can be clearly seen in the STEM image (Fig. 2(b)) but not the cementite precipitate at the position “X” or its interfaces. However, APT gives access to this information. Using this technique, it is not only possible to quantify what kind of elements segregate at the ferrite-ferrite grain boundary, but also what is the position, size, morphology and composition of the precipitate, if there is a concentration gradient inside of the precipitate or what kind of solutes are enriched at its phase boundaries. The APT dataset further tells if there is a pile-up of elements or a depletion zone at the precipitate/ferrite interfaces and what is the composition of the surrounding ferrite matrix.

3.2. Phase identification Displacive phase transformations such as the transformation of austenite into martensite go along with a change in structure but not in composition. This can make it difficult for APT to distinguish a phase that has formed by a displacive mechanism from its parent phase. Besides TKD, scanning NBD can be used to create phase maps of APT specimens [23]. Apart from being able to characterize grain orientations, this TEM technique can deliver the complementary crystallographic information needed to distinguish different phases with a lateral spatial resolution of 1–2 nm [9,24].

Please cite this article as: M. Herbig, Spatially correlated electron microscopy and atom probe tomography: Current possibilities and future perspectives, Scripta Materialia (2017), http://dx.doi.org/10.1016/j.scriptamat.2017.03.017

M. Herbig / Scripta Materialia xxx (2017) xxx–xxx

3

Fig. 1. Correlative TEM/APT characterization on a cold-drawn pearlitic wire [9,16]. (a) Inverse pole figure map measured by scanning nanobeam diffraction (NBD). (b) Bright field scanning transmission electron microscopy (BF-STEM) micrograph. The image was processed with a sharpening algorithm to enhance visibility of the interfaces. (c) 3D carbon atom map measured by atom probe tomography. A magnification of the area marked with a rectangle is depicted in Fig. 2.

3.3. Defect analysis The position of dislocations in APT specimens can be detected by TEM imaging. By the use of correlative TEM imaging and APT it can be univocally proved that hose-shaped regions in APT datasets with a different chemical composition than the matrix phase are solutes that are segregated to dislocations [25]. Also, for the site-specific preparation of shear bands into the apex of APT specimens, TEM imaging can be used. This can, for example, be applied for characterizing the redistribution of alloying elements along shear bands that intersect two different phases [26].

equivalent to the matrix when they have formed at room temperature (deformation twins) [9,27]. 4. What EM technique is best suited for each case The last section outlined the kind of materials science questions that can be answered by correlative EM/APT. Each available electron microscopy technique is particularly suited to a certain materials science aspect and has a certain magnification window where it is most efficiently applied. The following section aims at clarifying which EM technique is best suited in conjunction with APT to investigate a given materials science phenomenon. Figs. 3 and 4 give an overview of selected techniques that have proved useful to be combined with APT.

3.4. Proof of absence of segregation 4.1. SEM techniques The presence of defects such as dislocations in APT datasets is often only visible indirectly by the solutes that are enriched at the defect. Proving that there is no segregation of solutes to a given defect is therefore often difficult by APT alone as such defects are difficult to observe in APT reconstructions. Correlative EM/TEM can, in this case, be used to prove the absence of segregation to a given lattice defect. By using correlative TEM imaging and APT it is, for example, possible to show that Mn segregation only occurs on part of the dislocations in martensitic steel, while others remain free of solutes [25], or that Σ3 coherent twin boundaries in austenite are carbon-depleted when they have formed at 700 °C (annealing twins) while they are chemically

When conducting correlative EM/APT, one has to keep in mind that not all the specimens prepared will yield a successful experiment. Specimens can get damaged during transport between instruments and oxidation or absorption of embrittling atmospheric gasses such as hydrogen during specimen preparation and APT measurement as well as carbon contamination during EM experiments will reduce the yield in the subsequent APT experiments. For each of these reasons, one wants to minimize the amount of time between specimen preparation and APT measurement and keep the EM experiment as simple as possible. SEM techniques may have a lower lateral spatial resolution than

Please cite this article as: M. Herbig, Spatially correlated electron microscopy and atom probe tomography: Current possibilities and future perspectives, Scripta Materialia (2017), http://dx.doi.org/10.1016/j.scriptamat.2017.03.017

4

M. Herbig / Scripta Materialia xxx (2017) xxx–xxx

Fig. 2. Magnification of the area marked with a rectangle in Fig. 1. The correlative characterization gives access to complementary information at the near-atomic length scale. (a) Scanning NBD maps the sample with a convergent electron beam. The acquired local diffraction patterns give access to orientation maps, phase maps and orientation gradients. (b) (S)TEM imaging allows for accurate crystallographic and morphological analysis of lattice defects such as interfaces or dislocations. (c) APT provides 3D local compositional information and thereby information on 3D concentration gradients and on the 3D morphology of objects with different composition, such as precipitates.

TEM techniques, but they are available in the dual beam SEM/FIB instruments where the APT specimen preparation typically takes place. Performing correlative EM experiments in the SEM means that the specimens do not have to be transported to an additional instrument, that no additional waiting for instrument time is required, that precious TEM instrument time is saved and that the time between specimen preparation and APT measurement is minimized. Moreover, SEM measurements are usually easier to learn and require less effort than TEM measurements. For cases where the spatial resolution of SEM is sufficient, the correlative EM/APT should be conducted in the SEM. The SEM technique known as transmission Kikuchi diffraction (TKD, Fig. 3a, b) [19,28] represents an elegant and easy way to get an overview of the specimen's microstructure based on orientation and phase maps. TKD exploits the same detector system as used for EBSD but works in transmission mode on thin specimens. This means that the inelastically scattered electrons that generate the Kikuchi patterns come from a smaller region of the specimen, which improves the spatial resolution for dense materials from about 20 nm in the case of EBSD down to about 6–8 nm in the case of TKD [28]. In the case of equiaxed microstructures, as a rule of thumb, TKD can be applied on materials with grain sizes above 50 nm. TKD is particularly suited for the preparation of selected grain or phase boundaries into the apex of APT specimens, in order to ensure that they can be analyzed. Fig. 3a, b show an application example of TKD: with a step size of 10 nm a thin layer of austenite can be located in a martensitic APT specimen.

TKD allows for the analysis of orientation relationships between adjacent grains. However, a trace analysis of the grain/phase boundary plane is only possible for long and straight boundaries and the inclination of the grain/phase boundary plane in the depth of the specimen remains completely inaccessible due to the two dimensional nature of the TKD technique. Electron channeling contrast imaging (ECCI) [29] is an SEM imaging technique where the contrast is based on high-angle backscattered electrons. Grains oriented in channeling condition appear dark, others bright. Crystal defects fulfilling the Bragg condition within grains in channeling condition appear bright on a dark background. The strong orientation sensitivity of this technique, combined with a spatial resolution in the range of 10 nm [29], makes it possible to resolve microstructures with nano-sized grains. The successful implementation of ECCI on APT specimens is demonstrated in Fig. 3c. This technique can thus be used to guide the target preparation of interfaces into the apex of the APT specimens. This is usually not possible by secondary electron SEM imaging as the contrast of the interface normally fades throughout the final steps of the preparation in the dual beam SEM/FIB. The sitespecific preparation of interfaces into the apex of an APT specimen can also be carried out using TKD. However, ECCI is simpler and faster to use than TKD. The ability to visualize and analyze defects such as dislocations, stacking faults or the inclination angle of interfaces by ECCI in APT specimens, as is possible in bulk SEM specimens [29], is currently being investigated.

Please cite this article as: M. Herbig, Spatially correlated electron microscopy and atom probe tomography: Current possibilities and future perspectives, Scripta Materialia (2017), http://dx.doi.org/10.1016/j.scriptamat.2017.03.017

M. Herbig / Scripta Materialia xxx (2017) xxx–xxx

5

Fig. 3. SEM micrographs of APT specimens. Orientation maps (inverse pole figure maps, a) and phase maps (b) can be generated by transmission Kikuchi diffraction. The pronounced orientation sensitivity of the technique electron channeling contrast imaging enables imaging of the microstructure within the APT specimens (c).

4.2. TEM techniques TEMs are usually operated at much higher acceleration voltages than SEMs and can consequently also yield a higher spatial resolution. Thus modern SEMs are able to detect grain orientations, phases and possibly even analyze lattice defects within APT specimens, but TEMs can do the same at higher spatial resolution. What this means in practice can be seen when comparing Fig. 3a and b with Fig. 4a. These figures depict the same specimen. Fig. 3a and b shows the result of a TKD measurement at 30 kV acceleration voltage with a step size of 10 nm, which is close to the maximum spatial resolution in this instrument. Fig. 4a shows a bright field scanning transmission electron microscopy (BFSTEM) image acquired in a JEOL 2200 FS TEM at 200 kV acceleration voltage with a step size of 1.5 nm which is by far not the limit of the spatial resolution in this instrument. TKD gives only a rough overview of the position of grains and phases in the APT specimen while the BFSTEM micrograph reveals detailed microstructural information on dislocations, grain shapes and interface shapes. By using the scanning mode in a TEM, different microstructural features can easily be highlighted by varying camera length and/or detector type; thereby emphasizing Bragg or elemental Z contrast. In BFSTEM, electrons that are scattered in the direct scattering cone are detected when scanning over the specimen and this imaging mode is dominated by diffraction (Bragg) contrast. Annular dark field-STEM (ADF-STEM) records forward scattered electrons at intermediate scattering angles in the range of 20–55 mrad on an annular detector. This configuration is well-suited to characterize lattice defects, as shown in Fig. 4b, and to identify correlations between dislocation character and segregation. For visualizing subtle changes in composition within APT specimen, high angle annular dark field-STEM (HAADF-STEM) is the method of choice. The techniques records forward scattered electrons on an annular detector at high scattering angles of greater than 55 mrad. One scenario where this is required is the investigation of the deformation mechanism of metallic glasses (Fig. 4c). HAADF-STEM measurements that were performed on conventional TEM specimens indicated that there is a redistribution of elements into high (bright)

and low density regions (dark) within these shear bands [30,31]. However, the shear bands are only 5–10 nm thick and have a complex morphology what makes their chemical quantification in the TEM difficult — a clear case for APT. On the one hand, the use of correlative HAADFSTEM and APT is needed due to the sparse distribution of the shear bands in the material that requires their target preparation into the apex of the specimens and on the other hand to clearly be able to link bright and dark regions of the shear bands to regions that have according to APT measurement a distinct composition. Fig. 4d depicts a DF-TEM image of sheared precipitates in an austenitic matrix of a Fe-Mn-Al-C steel [32,33] that was exposed to 15% plastic deformation. The precipitates that are imaged bright are intersected by an aggregation of coplanar slip bands that show up as parallel black lines. The materials science question here is in how far elements of matrix and precipitates are redistributed by the shearing process. TEM and APT must be correlated in this case as the shear bands are not detectable by APT and the subtle redistribution of elements that mainly involve carbon is difficult to detect by TEM. In the case of this material the small compositional difference and the cube-on-cube orientation relationship between matrix and precipitates don't provide enough contrast to clearly distinguish the phases with the above mentioned STEM techniques. The superlattice DF-TEM imaging employed here enables to select a g vector (a diffraction spot) of the precipitate phase and to transform its intensity into an image where all regions that give rise to this g vector appear bright. At the same time the slip bands appear dark in the DF-TEM images and therefore both features can be visualized at the same time what is crucial for this correlative investigation. Performing STEM-EDX (Fig. 4e) on APT specimens, at first glance, might seem to be unnecessary as the compositional information obtainable with this technique is redundant with the one obtainable from APT. However, this approach can become useful, e.g. in the case of multiphase materials for the target preparation of a certain phase into the apex of the APT specimen. In such cases, it can become difficult to distinguish individual phases based on imaging and diffraction alone. EDX can provide the additional complementary chemical information that helps to identify the phase of interest.

Please cite this article as: M. Herbig, Spatially correlated electron microscopy and atom probe tomography: Current possibilities and future perspectives, Scripta Materialia (2017), http://dx.doi.org/10.1016/j.scriptamat.2017.03.017

6

M. Herbig / Scripta Materialia xxx (2017) xxx–xxx

Fig. 4. TEM micrographs of APT specimens. (a) Bright field-STEM of a martensitic/austenitic steel. (b) Annular dark field-STEM of a network of dislocations (bright lines) in martensitic steel. (c) High angle annular dark field-STEM of shear bands in a plastically deformed metallic glass. (d) Dark field-TEM of sheared kappa carbides in austenitic steel. (e) STEM-energydispersive X-ray spectroscopy of martensitic steel with Mn-rich precipitates. (f) High resolution high angle annular dark field-STEM of a kappa carbide in austenite.

Conducting selected area electron diffraction (SAED) on APT specimens is, in practice, mostly required for orienting grains to fulfill particular diffraction conditions before other TEM techniques such as DF-TEM or HRTEM can be performed. In case of materials with grain sizes above 50 nm the analysis of orientation relationships between grains and the identification of phases is usually conducted more easily by TKD in the SEM, as mentioned above. In cases of grain sizes below 50 nm, where TKD reaches its limits, SAED can neither be applied as its spatial resolution is limited by the size of the smallest SAED aperture available, which is usually around 50 nm. For identifying grain orientations and phases in microstructures with smaller grain sizes nanobeam diffraction (NBD [34]) has proven to be useful. In contrast to SAED that works with a parallel beam, NBD makes use of a slightly convergent beam with a convergence angle in the order of 1.5–3.5 mrad which causes the diffraction spots to become disc-shaped. However, as the center positions and

the symmetry of the diffraction pattern of SAED and NBD are the same, a NBD pattern can be indexed and interpreted just like a SAED pattern. The spatial resolution of NBD is basically only limited by the combination of spot size, convergence angle and specimen thickness and was shown to reach down to 1–2 nm which is at least a factor of three higher as compared to TKD [9,24]. NBD can be conducted manually on individual positions of the specimen, e.g. in STEM spot mode. The obtained NBD diffraction patterns can be indexed e.g. using the TOCA software [35] that not only identifies the grain orientation but also the phase at a given location. However, NBD can also be used as a scanning technique that gives access to a whole range of further possibilities. Orientation and phase maps, as known from EBSD and TKD but here with 1–2 nm spatial resolution, provide a precise overview of the specimen (see Figs. 1 and 2). Orientation gradients and dislocation densities can also be analyzed, even post-mortem

Please cite this article as: M. Herbig, Spatially correlated electron microscopy and atom probe tomography: Current possibilities and future perspectives, Scripta Materialia (2017), http://dx.doi.org/10.1016/j.scriptamat.2017.03.017

M. Herbig / Scripta Materialia xxx (2017) xxx–xxx

after the destructive APT experiment a local diffraction pattern for each position of the specimen is available. So too can grain misorientations as well as grain boundary planes and virtual dark and bright field images can be generated [24]. Fig. 4f gives an impression of what can be currently achieved in term of high resolution (HR) imaging on APT specimens [14,36,37] using an aberration corrected STEM on a 20 nm thin part at the apex of the APT specimen. The figure depicts a HR-HAADF-STEM image of a coherent precipitate of the kappa phase in an austenitic matrix in an Fe-Mn-AlC steel [32,33]. Fig. 4f was produced from an image series of 15 single images, each acquired with a pixel dwell time of 10 μs. The single images were aligned by rigid registration, based on cross-correlation, and were subsequently summed and Fourier-filtered to reduce scan noise. Atomic columns can easily be resolved and their intensity is proportional to the average Z number of the respective column (here: dark columns are rich in the light alloying element Al, while bright columns contain higher amounts of Fe and Mn). For example, such correlative HR-STEM/APT investigations are required when the correlation between atomic-scale defect structures and segregation of light elements or the correlation between strain and chemical gradients close to lattice defects need to be quantified. The examples given in Figs. 3 and 4 demonstrate that APT specimens can be measured by EM just like conventional SEM/TEM specimens and that their needle-shaped geometry does not compromise the characterization quality. Local specimen thickness, damage-free specimen preparation, oxide layers and carbon contamination are decisive for the quality of the obtained micrographs, just as it is the case for conventional TEM specimens. APT specimens can easily be prepared by FIB milling in such way that the first few 100 nm have a thickness between 20 and 100 nm — a thickness range where most TEM techniques can be applied with reasonable quality. Also, the preparation of specimens with tip radii of 10 nm is possible and the FIB preparation can be conducted with so little beam damage that well-interpretable NBD diffraction patterns can be obtained from the top 10 nm of the apex [9] — ideal conditions also for HR-STEM (Fig. 4f). However, in the case of sparsely distributed lattice defects or precipitates that need to be investigated by correlative HR-STEM/APT, target preparation into the top part of the specimen apex that is thin enough to conduct high quality HRSTEM can be challenging.

7

species from air such as hydrogen might penetrate deeply into the thin APT specimens and carbon contamination will occur during the EM experiments. These effects often lead to partial ruptures or early failure in the APT experiments. Cleaning the specimens prior to the APT measurement by annular milling using low acceleration voltage in the FIB improves yield, but by far not to a point comparable to specimens that are measured by APT directly after preparation. In addition, fieldinduced low temperature oxidation [40] causes problems, as this destroys the top part of the specimen which is best suited for HRTEM measurements. All efforts invested in specimen preparation and EM during a correlative EM/APT experiment are lost when the specimen subsequently does not yield a successful APT experiment. With the current specimen yield, it is for reasons of efficiency that it is challenging to conduct time consuming complementary EM, such as a Burgers vector analyses. Improving the yield of correlative EM/APT is thus the key to efficiently conduct also more complex EM experimental techniques on APT specimens. Promising measures to improve yield are (a) consistent grounding of all parts that come in contact with the specimens as this can prevent field-induced low temperature oxidation, (b) plasma cleaning of grid and grid holder prior to APT specimen preparation as this reduces the amount of carbon contamination and (c) the removal of oxide layers before and after EM, homogenously and from all sides of the specimen, by low energy sputtering. The applicability and efficiency of these measures to improve yield is currently being explored at our institute. 5.2. The potential of electron channeling contrast imaging The technique cECCI (ECCI under controlled geometrical conditions) [29] has been shown to be capable of crystallographically analyzing lattice defects within bulk specimens in the SEM for low defect densities. If this was also possible within APT specimens, part of the correlative crystallographic analysis on lattice defects could be done in the dual beam SEM/FIB so that for cases of low defect densities the characterization in the TEM could be entirely avoided. This would significantly reduce the experimental efforts and would give access to such investigations also to groups without TEM. Therefore the limits of conducting ECCI on APT specimens should be further explored. 5.3. The need for 3D TEM orientation mapping techniques

5. Potential future developments in the field of correlative EM/APT Figs. 3 and 4 demonstrate that not only in principle but also in practice a large variety of EM techniques including HR-STEM can be performed on APT specimens [9]. This list of EM techniques that are useful to be combined with APT certainly exceeds the ones discussed here. It can be expected, that in the future, even more methods will be applicable to APT specimens as there is no physical reason why not all EM techniques should be applicable also on APT specimens. The applicability of the above described experimental setup that employs the “grid holder” as a core part and uses adapters to create compatibility with various instruments and techniques [9] is not limited to the examples shown here: As the grid holder handles standard three millimeter TEM grids, the setup also allows for the FIB preparation of TEM lamellae, as well as for the insertion of electropolished TEM specimens, and the investigation of both of them in the SEM and TEM [38]. The setup further allows for the direct preparation of APT specimens from TEM foils/lamellae as described by [39]. The versatile nature of the experimental concept opens the door for further innovation in this field. 5.1. Improving the yield of correlative EM/APT experiments However, the bottleneck currently is not the possibility to conduct EM on APT specimens but specimen yield in the APT after having conducted EM. During specimen manipulation and transport between the instruments, oxide layers will form on the surface, embrittling gaseous

The correlative EM/APT community would profit from the development of a robust, user-friendly and fast 3D electron microscopy technique that is able to map grain shapes, orientations and phases of materials with arbitrary grain shapes and sizes. A spatial resolution of about one nm should be sufficient for most cases. Such a technique could potentially be developed on the basis of scanning NBD, being performed at different tilt-angles. The currently commercially available 2D scanning NBD technique already gives access to orientation and phase maps at a lateral spatial resolution as low as 1 nm [24]. Orientation gradients, densities of geometrically necessary dislocations and interface types can be extracted in an automated way from these datasets. Virtual bright and dark field images with arbitrary aperture size, shape and position can be created and the visibility of defects can be enhanced using a correlation coefficient algorithm [41]. The technique can be easily combined with precession electron diffraction (PED) by the use of the obtained quasi-kinematic diffraction patterns that simplify the determination of the crystal structure [42,43]. However, the current 2D scanning NBD method cannot handle grain overlaps so that its application on thicker parts of nanocrystalline APT specimens is limited to columnar grain shapes [16]. By conducting a tilt-series of NBD scans, this limitation can be overcome, as recently demonstrated by Eggeman et al. [44] who reconstructed a matrix containing two different types of coherent precipitates. Considering the results of the synchrotron community, where more complex scenarios with many grains stacked on top of each other in the direction of the incident beam have been successfully

Please cite this article as: M. Herbig, Spatially correlated electron microscopy and atom probe tomography: Current possibilities and future perspectives, Scripta Materialia (2017), http://dx.doi.org/10.1016/j.scriptamat.2017.03.017

8

M. Herbig / Scripta Materialia xxx (2017) xxx–xxx

reconstructed [45,46], the TEM community should also be able to handle more complex cases in future with the appropriate software development. In combination with PED, a scanning NBD tilt-series would enable the identification of unknown complex phases based on automated diffraction tomography [47] as such a dataset would contain for each grain a 3D map of the reciprocal space. Also, the 3D crystallographic analysis of lattice defects would be possible and the inherent problem with 180° ambiguities of the 2D scanning NBD [48] could be solved. To this already rich set of 3D information, APT could contribute information on the compositions of grains and phases, on the segregation to defects [49], on impurity concentrations, on clustering of atoms, on concentration gradients inside individual grains and on concentration profiles close to interfaces amongst many others. Precise information on the 3D microstructure acquired by a tilt-series of NBD scans on APT specimens could further help understanding the field-evaporation mechanisms of APT. This would help optimize APT reconstruction algorithms and thereby improve the spatial resolution of this technique. 5.4. Alternative approaches for combined crystallographic and compositional analysis There are more methods than correlative EM/APT that give access to combined 3D chemical and crystallographic information at the nm scale that should be further developed because most likely each of them will have a particular niche where it is found to be more performant than the others. APT crystallography [3–7,50,51] might not be able to achieve the spatial and angular resolution of EM but for many cases it will be sufficient. The technique further has a high potential for automation which is less the case for correlative EM/APT. Analytical TEM might have trouble to quantify trace amounts of light elements in a heavy matrix or to quantify the composition of objects with complex 3D shapes. However, in the case of objects that can be projected in a linear way, not only accurate chemical analysis is possible by TEM but also with a higher spatial resolution as compared to APT. Conducting only EM or APT individually takes less effort than using these techniques in a correlative way. Apart from that, from a practical viewpoint, APT and TEM are not always available in the same location. The development of an instrument that combines EM and APT in a single machine would allow the measurement of characteristic features within APT specimens that can guide reconstruction [52], it would allow measuring the development of the specimen apex during the APT experiment [53] and it would also enable the characterization of the charge density distribution on the specimen using electron holography [54]. Such measurements would probably have to be conducted in an “interrupted in-situ” manner as a strong influence of the electron beam on the ion trajectories can be expected. However, such measurements would provide better understanding of the field-evaporation mechanisms, would make the development of better reconstruction algorithms possible and potentially enhance the spatial resolution achievable with APT. Acknowledgements The images depicted in this manuscript were acquired in collaboration with several colleagues. The author gratefully acknowledges here the work of M. Yao, A. Kostka, C. Scheu, M. Kuzmina and M. Köhler and in particular thanks E. Welsch and C. Liebscher who acquired Fig. 4d and f, respectively. The author further thanks S. Zaefferer and

D. Raabe for many years of fruitful discussions and C. Liebscher, C. Scheu and B. Gault for the helpful comments regarding this manuscript. The author is grateful to Uwe Tezins and Andreas Sturm for their support to the FIB and APT facilities at MPIE. This work was funded by the German Research Foundation [Deutsche Forschungsgemeinschaft (DFG)] through SFB 761 “Steel Ab Initio” and by the Federal Ministry of Education and Research [Bundesministerium für Bildung und Forschung (BMBF)] through grant 03SF0535. References [1] T.F. Kelly, D.J. Larson, Annu. Rev. Mater. Res. 42 (2012) 1. [2] D.B. Williams, C.B. Carter, Transmission Electron Microscopy — A Textbook for Materials Science, Springer, 2009. [3] V.J. Araullo-Peters, B. Gault, S.L. Shrestha, et al., Scr. Mater. 66 (2012) 907. [4] B. Gault, M.P. Moody, J.M. Cairney, et al., Mater. Today 15 (2012) 378. [5] M.P. Moody, F.Z. Tang, B. Gault, et al., Ultramicroscopy 111 (2011) 493. [6] L. Yao, S.P. Ringer, J.M. Cairney, et al., Scr. Mater. 69 (2013) 622. [7] P.V. Liddicoat, X.Z. Liao, Y.H. Zhao, et al., Nat. Commun. 1 (2010). [8] C. Oberdorfer, S.M. Eich, M. Lutkemeyer, et al., Ultramicroscopy 159 (2015) 184. [9] M. Herbig, M. Kuzmina, C. Haase, et al., Acta Mater. 83 (2015) 37. [10] B.P. Gorman, D.R. Diercks, N. Salmon, et al., Microsc. Today 16 (2008) 42. [11] P.J. Felfer, T. Alam, S.P. Ringer, et al., Microsc. Res. Tech. 75 (2012) 484. [12] B.W. Krakauer, J.G. Hu, S.M. Kuo, et al., Rev. Sci. Instrum. 61 (1990) 3390. [13] H. Norden, K.M. Bowkett, J. Sci. Instrum. 44 (1967) 238. [14] M. Herbig, P. Choi, D. Raabe, Ultramicroscopy 153 (2015) 32. [15] K. Thompson, D. Lawrence, D.J. Larson, et al., Ultramicroscopy 107 (2007) 131. [16] M. Herbig, D. Raabe, Y.J. Li, et al., Phys. Rev. Lett. 112 (2014). [17] O. Cojocaru-Miredin, T. Schwarz, P.P. Choi, J. Vis. Exp. (2013). [18] A. Stoffers, O. Cojocaru-Miredin, O. Breitenstein, et al., 2014 IEEE 40th Photovoltaic Specialist Conference (PVSC), 42, 2014. [19] K. Babinsky, R. De Kloe, H. Clemens, et al., Ultramicroscopy 144 (2014) 9. [20] K. Rice, Y. Chen, T. Prosa, et al., Microsc. Microanal. 22 (2016) 583. [21] D. Raabe, M. Herbig, S. Sandlobes, et al., Curr. Opin. Solid State Mater. Sci. 18 (2014) 253. [22] D. Raabe, S. Sandlobes, J. Millan, et al., Acta Mater. 61 (2013) 6132. [23] Y. Toji, H. Matsuda, M. Herbig, et al., Acta Mater. 65 (2014) 215. [24] E. Rauch, M. Véron, Mater. Charact. 98 (2014) 1. [25] M. Kuzmina, M. Herbig, D. Ponge, et al., Science 349 (2015) 1080. [26] W. Guo, E.A. Jagle, P.P. Choi, et al., Phys. Rev. Lett. 113 (2014). [27] R.K.W. Marceau, I. Gutierrez-Urrutia, M. Herbig, et al., Microsc. Microanal. 19 (2013) 1581. [28] P.W. Trimby, Ultramicroscopy 120 (2012) 16. [29] S. Zaefferer, N.N. Elhami, Acta Mater. 75 (2014) 20. [30] V. Schmidt, H. Rosner, M. Peterlechner, et al., Phys. Rev. Lett. 115 (2015). [31] H. Rosner, M. Peterlechner, C. Kubel, et al., Ultramicroscopy 142 (2014) 1. [32] D.P.E. Welsch, S.M. Hafez Haghighat, S. Sandlöbes, P. Choi, M. Herbig, S. Zaefferer, D. Raabe, Acta Mater. 116 (2016) 188. [33] M.J. Yao, P. Dey, J.-B. Seol, et al., Acta Mater. 106 (2016) 228. [34] S. Zaefferer, U. Glatzel, Textures of MaterialsPts 1 and 2 408-4 2002 1531. [35] S. Zaefferer, J. Appl. Crystallogr. 33 (2000) 10. [36] L. Rigutti, I. Blum, D. Shinde, et al., Nano Lett. 14 (2014) 107. [37] W. Lefebvre, D. Hernandez-Maldonado, F. Moyon, et al., Ultramicroscopy 159 (2015) 403. [38] A. Stoffers, B. Ziebarth, J. Barthel, et al., Phys. Rev. Lett. 115 (2015). [39] M.I. Hartshorne, D. Isheim, D.N. Seidman, et al., Ultramicroscopy 147 (2014) 25. [40] C. Nowak, G. Schmitz, R. Kirchheim, Surf. Sci. 604 (2010) 641. [41] A.K. Kiss, E.F. Rauch, J.L. Labar, Ultramicroscopy 163 (2016) 31. [42] R. Vincent, P.A. Midgley, Ultramicroscopy 53 (1994) 271. [43] A.S. Eggeman, R. Krakow, P.A. Midgley, Nat. Commun. 6 (2015). [44] P.A. Midgley, A.S. Eggeman, IUCrJ 2 (2015) 126. [45] W. Ludwig, P. Reischig, A. King, et al., Rev. Sci. Instrum. 80 (2009). [46] D.J. Jensen, E.M. Lauridsen, L. Margulies, et al., Mater. Today 9 (2006) 18. [47] U. Kolb, T. Gorelik, M.T. Otten, Ultramicroscopy 108 (2008) 763. [48] E.F. Rauch, L. Dupuy, J. Appl. Crystallogr. 39 (2006) 104. [49] P. Felfer, B. Scherrer, J. Demeulemeester, et al., Ultramicroscopy 159 (2015) 438. [50] A.K. de Silva, G. Leyson, M. Kuzmina, et al., Acta Mater. 124 (2017) 305. [51] A.J. Breen, K. Babinsky, A.C. Day, et al., Microsc. Microanal. (2017) [in print]. [52] I. Arslan, E.A. Marquis, M. Homer, et al., Ultramicroscopy 108 (2008) 1579. [53] D. Haley, T. Petersen, S.P. Ringer, et al., J. Microsc. (Oxford) 244 (2011) 170. [54] V. Migunov, A. London, M. Farle, et al., J. Appl. Phys. 117 (2015).

Please cite this article as: M. Herbig, Spatially correlated electron microscopy and atom probe tomography: Current possibilities and future perspectives, Scripta Materialia (2017), http://dx.doi.org/10.1016/j.scriptamat.2017.03.017