Near-Surface Material Phases and Microstructure of Scandate Cathodes

Scandate cathodes that were fabricated using the liquid-solid process and that exhibited excellent emission performance were characterized using complementary state-of-the-art electron microscopy techniques. Sub-micron BaAl2O4 particles were observed on the surfaces and edges of tungsten particles, as seen in cross-section samples extracted from the scandate cathode surface regions. Although several BaAl2O4 particles were observed to surround smaller Sc2O3 nanoparticles, no chemical mixing of the two oxides was detected, and in fact the distinct oxide phases were separately verified by chemical analysis and also by 3D elemental tomography. Nanobeam electron diffraction confirmed that the crystal structure throughout W grains is body-centered cubic, indicating that they are metallic W and did not experience noticeable changes, even near the grain surfaces, as a result of the numerous complex chemical reactions that occur during cathode impregnation and activation. 3D reconstruction further revealed that internal Sc/Sc2O3 particles tend to exhibit a degree of correlated arrangement within a given W particle, rather than being distributed uniformly throughout. Moreover, the formation of Sc/Sc2O3 particles within W grains may arise from W surface roughening that occurs during the liquid-solid synthesis process.

These persistent issues are due, at least to some extent, to knowledge gaps in the relationships between processing, microstructure, composition, and emission properties of scandate cathodes.

Materials and Methods
The scandate cathodes characterized in this paper were prepared with W precursor powder mixed with nanoscale scandia, and this powder mixture had been prepared using the liquid-solid (L-S) technique. Detailed procedures of the L-S approach can be found in an earlier paper by one of the authors [1]. The L-S process utilizes an aqueous nitrate solution to dissolve scandia, and then solid tungsten powder is added as a substrate for formation of scandia nanoparticles. The nitrate solution may etch or roughen the surfaces of tungsten particles. Multiple scandate cathodes were provided by e beam inc., where emission testing was also performed. Fabrication included the following steps: for each cathode, the W-scandia powder mixture was pressed into a pellet~2 mm diameter and~1 mm tall; then the pellet was sintered and subsequently impregnated with barium calcium aluminate (6BaO-1CaO-2Al 2 O 3 ); afterwards, the sample was washed with deionized water to remove residual impregnate material; finally, the cathodes were activated by heating at 1150 • C b for 1 h. Prior to microstructural characterization, two cathodes (designated cathode #1 and cathode #2) were emission tested in a close-spaced diode (CSD) configuration. These scandate cathodes exhibited excellent emission capabilities as indicated by knee temperatures of 820 • C b and 837 • C b (brightness temperature, with respect to W) at the end of extensive emission testing: 3024 and 9840 h, respectively.
Morphological characterization and location-specific chemical analysis of scandate cathode surfaces was performed in a dual-beam focused ion beam and scanning electron microscope (FIB-SEM, FEI Helios NanoLab 660, Thermo Fisher Scientific, Hillsboro, OR, USA) equipped with an X-ray energy dispersive spectrometer (EDS; Oxford X-Max 80 mm 2 detector, Oxford Instruments, Abingdon, UK). Two lift-out lamellae were extracted from cathode #1 using the FIB and were examined in a transmission electron microscope (TEM; FEI Talos F200X (Thermo Fisher Scientific, Hillsboro, OR, USA) and JEOL JEM-F200 microscopes (JEOL, Tokyo, Japan) were used in this study). Along with scanning transmission electron microscope (STEM) imaging for structural characterization and scanning nanobeam electron diffraction for crystal structure determination, high-resolution elemental mapping was performed on the TEM cross-section lamella using the Super-X EDS system (which includes 4 confocal EDS detectors that simultaneously collect X-rays from the same sample location) in the FEI Talos F200X TEM. Serial sectioning for 3D reconstruction was performed on cathode #2 using the Auto Slice and View technique on the FIB-SEM. Back scattered electron (BSE) micrographs were collected in automated mode by the FIB-SEM, and these were subsequently used for 3D reconstruction of the W crystal shape using Avizo visualization software (Avizo 9.3).

Surface Morphology of Scandate Cathodes
Images of the surfaces of two scandate cathodes are presented in Figure 1. It is seen in Figure 1a,c that the W particle size is~1 µm and the pores are open, i.e., not obstructed by impregnate material. Two types of particles, each with a size on the order of 100 nm, are distributed across the surfaces of W grains: Sc 2 O 3 and BaAl 2 O 4 , as determined from EDS point analysis of composition. Complementary and more detailed analysis is also discussed in a related article [37]. Additionally, a critical observation, apparent in Figure 1b,d, is the highly faceted surface morphology of W grains in both cathodes. Based on Wulff analysis and as discussed in a separate publication [37], these facets correspond to the {100}, {110} and {112} crystallographic planes of W.
The nanoscale particles with diameter on the order of 10 nm, which decorate W facets in a fine dispersion, are assumed to be either Ba or BaO, based on results reported in the literature [37,38]. It appears, however, that numerous Ba/BaO particles were distributed over the W surfaces for cathode #1, while relatively fewer particles were observed on the surface of cathode #2, as shown in Figure 1b,d. As discussed in a previous paper [37], these Ba/BaO particles may result from transformation of a metallic Ba monolayer that exists at high temperature but is unstable during cooling after testing; this transformation may occur during cathode cooling after operation, creating the nanoscale particles observed here, and these particles would likely oxidize upon exposure to oxygen. Considering the emission test times, it may be reasonable to observe relatively lower areal density of Ba/BaO particles on the W facets of cathode #2, which had undergone emission testing for 9,840 h. This is more than three times the duration of testing for cathode #1, which could have resulted in lower (sub-monolayer) Ba coverage on the W surfaces of cathode #2 due to Ba loss during operation, leading to lower density of the Ba/BaO particles. high-resolution elemental mapping was performed on the TEM cross-section lamella using the Super-X EDS system (which includes 4 confocal EDS detectors that simultaneously collect X-rays from the same sample location) in the FEI Talos F200X TEM. Serial sectioning for 3D reconstruction was performed on cathode #2 using the Auto Slice and View technique on the FIB-SEM. Back scattered electron (BSE) micrographs were collected in automated mode by the FIB-SEM, and these were subsequently used for 3D reconstruction of the W crystal shape using Avizo visualization software (Avizo 9.3).

Surface Morphology of Scandate Cathodes
Images of the surfaces of two scandate cathodes are presented in Figure 1. It is seen in Figure 1a,c that the W particle size is ~1 µm and the pores are open, i.e., not obstructed by impregnate material. Two types of particles, each with a size on the order of 100 nm, are distributed across the surfaces of W grains: Sc2O3 and BaAl2O4, as determined from EDS point analysis of composition. Complementary and more detailed analysis is also discussed in a related article [37]. Additionally, a critical observation, apparent in Figure 1b,d, is the highly faceted surface morphology of W grains in both cathodes. Based on Wulff analysis and as discussed in a separate publication [37], these facets correspond to the {100}, {110} and {112} crystallographic planes of W.
The nanoscale particles with diameter on the order of 10 nm, which decorate W facets in a fine dispersion, are assumed to be either Ba or BaO, based on results reported in the literature [37,38]. It appears, however, that numerous Ba/BaO particles were distributed over the W surfaces for cathode #1, while relatively fewer particles were observed on the surface of cathode #2, as shown in Figure 1b,d. As discussed in a previous paper [37], these Ba/BaO particles may result from transformation of a metallic Ba monolayer that exists at high temperature but is unstable during cooling after testing; this transformation may occur during cathode cooling after operation, creating the nanoscale particles observed here, and these particles would likely oxidize upon exposure to oxygen. Considering the emission test times, it may be reasonable to observe relatively lower areal density of Ba/BaO particles on the W facets of cathode #2, which had undergone emission testing for 9,840 h. This is more than three times the duration of testing for cathode #1, which could have resulted in lower (submonolayer) Ba coverage on the W surfaces of cathode #2 due to Ba loss during operation, leading to lower density of the Ba/BaO particles.

Elemental Analysis of Phases in Cross-Section Samples
High-resolution elemental mapping was performed on a TEM lamella extracted from cathode #1, using the multi-spectrometer EDS system of the FEI Talos F200X TEM. These results are presented in Figure 2, where the images and elemental maps show a cross-section of the cathode that was created during preparation of the TEM lamella. This specimen therefore reveals the structure and composition of the near-surface cathode region, including what is inside the W particles as well as what is on their surfaces and in the pores. The high-angle annular dark field (HAADF) image in Figure 2a shows the equiaxed shape of W particles, with flat edges corresponding to different surface facets of a given particle. Pt was deposited on the top cathode surface for protection during subsequent ion milling, as indicated by the white arrow in this image. Figure 2b displays a composite map of composition across the lamella, with different colors corresponding to the major elements of interest, and Figure 2c-h display elemental maps for W, Sc, Ga, Ba, Al and O, respectively. Comparing Figure 2d,h, specifically the area marked by the white arrow, it is seen that some compounds attached to the edges of tungsten particles show consistent Sc and O signals during EDS mapping, while certain other Sc-containing compounds located inside tungsten particles (e.g., the row of 4 particles inside the W grain, below the white arrow in Figure 2d,h) do not exhibit a noticeable corresponding O signal. In addition to the Sc-containing particles, it was observed in the cross-section maps of Figure 2f-h that other particles exhibit highly correlated signals of Ba, Al and O. These particles are interpreted to have a mixed Ba-Al oxide phase. The majority of these Ba-Al oxides exhibit only Ba, Al and O signals, e.g., as indicated by the black arrows in Figure 2a. However, several of the Ba-Al oxides also exhibit simultaneous Sc and O signals. In these cases, the Sc-O appears to be entrapped within the Ba-Al oxides, as indicated by red arrows in Figure 2b, as well as Figure 2d,f,g. Note that the Sc-O particles are distinct from the Ba-Al oxide particles. In addition, a number of Ba-Al oxide particles are observed to be immediately adjacent to other particles that show Sc and O signals, as marked by white arrows in Figure 2b, as well as Figure 2d,f,g. It is noted that a shell of W encases some of the particles seen in the cross-section lift-out specimen, for instance as indicated by the yellow arrow in Figure 2c, but this is attributed to redeposition of sputtered W from the ion-milling process during fabrication of the TEM lamella. This is also consistent with the Ga map in Figure 2e, since Ga is present in the redeposited layers. The composition of a Ba-Al oxide particle near the cathode surface (visible as a dark particle at the upper left of Figure 2b), was investigated using EDS coupled with EELS. The results are presented in Figure 3. The HAADF image in Figure 3a shows the Ba-Al oxide particle located between three W particles and a Pt cap layer (from the FIB lift-out process). The EDS measurement of composition was taken from the region marked by the blue box in Figure 3a, and indicated the following composition: Ba (12.2 at.%), Al (20.7 at.%), O (63.4 at.%), W (0.5 at.%), and C (3.2 at.%). No Sc signal was detected here, i.e., no measurable peak was seen at 4.09 keV on the EDS spectrum acquired in the region outlined by the green box that encompasses the oxide particle. This measured composition provides a reasonable match to the compound BaAl 2 O 4 , albeit with excess O. Moreover, as shown in Figure 3b, a selected area electron diffraction pattern was obtained from the oxide particle, and it matched the known hexagonal crystal structure of BaAl 2 O 4 . Indexing of this diffraction pattern indicated that it corresponds to a [100] zone axis orientation, and it matches the theoretical pattern expected for BaAl 2 O 4 in this orientation. A separate EDS measurement from a neighboring grain, in the region marked by the yellow box of Figure 3a, yielded a composition of W (78.5 at.%), Ba (0.1 at.%), Al (0.8 at.%), O (7.3 at.%) and C (13.3 at.%), confirming that the neighboring grain is W. The measured oxygen content of~7 at.% and carbon content of~13 at.% in the W area are assumed to arise from surface contamination, which is typical in such samples; it is noted that this would imply the EDS measurement in the blue box region more closely matches the expected stoichiometric composition of BaAl 2 O 4 . Figure 3c-f present the elemental maps acquired by EELS in the region outlined by the green box in Figure 3a. These maps indicate highly correlated signals of O, Al and Ba (but with an absence of W), in agreement with the EDS analysis. The composition of a Ba-Al oxide particle near the cathode surface (visible as a dark particle at the upper left of Figure 2b), was investigated using EDS coupled with EELS. The results are presented in Figure 3. The HAADF image in Figure 3a shows the Ba-Al oxide particle located between three W particles and a Pt cap layer (from the FIB lift-out process). The EDS measurement of composition was taken from the region marked by the blue box in Figure 3a, and indicated the following composition: Ba (12.2 at.%), Al (20.7 at.%), O (63.4 at.%), W (0.5 at.%), and C (3.2 at.%). No Sc signal was detected here, i.e. no measurable peak was seen at 4.09 keV on the EDS spectrum acquired in the region outlined by the green box that encompasses the oxide particle. This measured composition provides a reasonable match to the compound BaAl2O4, albeit with excess O. Moreover, as shown in Figure 3b, a selected area electron diffraction pattern was obtained from the oxide particle, and it matched the known hexagonal crystal structure of BaAl2O4. Indexing of this diffraction pattern indicated that it corresponds to a [100] zone axis orientation, and it matches the theoretical pattern expected for BaAl2O4 in this orientation. A separate EDS measurement from a neighboring grain, in the region marked by the yellow box of Figure 3a, yielded a composition of W (78.5 at.%), Ba (0.1 at.%), Al (0.8 at.%), O (7.3 at.%) and C (13.3 at.%), confirming that the neighboring grain is W. The measured oxygen content of ~7 at.% and carbon content of ~13 at.% in the W area are assumed to arise from surface contamination, which is typical in such samples; it is noted that this would imply the EDS measurement in the blue box region more closely matches the expected stoichiometric composition of BaAl2O4. Figure 3c-f present the elemental maps acquired by EELS in the region outlined by the green box in Figure 3a. These maps indicate highly correlated signals of O, Al and Ba (but with an absence of W), in agreement with the EDS analysis. To explore the relationship between Ba-Al oxide and embedded Sc-oxide, the region denoted by red arrows in Figure 2 was studied in more detail and these results are presented in Figure 4. The HAADF images, where Figure 4b corresponds to the red box in Figure 4a, shows the morphology of the Ba-Al oxide region, where it is seen that Sc-oxide is surrounded by Ba-Al oxide.  To explore the relationship between Ba-Al oxide and embedded Sc-oxide, the region denoted by red arrows in Figure 2 was studied in more detail and these results are presented in Figure 4. The HAADF images, where Figure 4b corresponds to the red box in Figure 4a, shows the morphology of the Ba-Al oxide region, where it is seen that Sc-oxide is surrounded by Ba-Al oxide. In order to further investigate the relation between BaAl2O4 and Sc2O3, including possible physical mixing or chemical interactions between the oxides, a small W particle with attached oxides was selected for detailed visualization via a 3D EDS tomogram; this region is indicated by the red box in Figure 5a. It is noted that this experiment was conducted with a separate TEM lamella than the one imaged in Figures 1-4, but both lamellae were cut from cathode #1. The higher magnification image in Figure 5b shows a BaAl2O4 particle atop the larger W particle. 2D slices from the 3D EDS tomogram are presented in Figure 5c-e, showing progressively more elements in each EDS map. This figure sequence conveys the relative distribution of Ba, Sc and W, and confirms that Sc is distinct from both the W particle and the BaAl2O4 particle that encases it, i.e., Sc does not appear to be physically or chemically mixed with W or BaAl2O4. Supplementary Movie 1 shows the 3D tomogram in imaging mode and in EDS mapping mode, and also rotated about the vertical axis, is available in the online Supplemental Files. The EDS composite elemental mapping in this 3D tomography movie corroborates the observations described above and, along with the observations portrayed in Figures 1-4, indicates that Sc exists in distinct Sc2O3 oxide particles that decorate the surfaces of W grains.
The Ba-Al oxide particle marked by the leftmost white arrow of Figure 2 was characterized at high-magnification, since it was adjacent to a Sc-oxide particle and offered an opportune location for more detailed study of these distinct but adjacent oxides. Figure 6 presents high spatial resolution images and elemental maps of these particles; as seen by comparison of Figure 2a and 6a, the TEM lamella has been imaged from the opposite orientation in Figure 6, i.e., the features are reversed leftto-right. The HAADF image of Figure 6b, corresponding to the red box in Figure 6a, shows a wide darker gray region that indicates lower average atomic number and therefore suggests the presence of oxide particle(s). The elemental maps of Figure 6c-g agree with other observations of distinct BaAl2O4 and Sc2O3 particles and, moreover, they show a sharp boundary between Sc and Al. These high spatial resolution EDS maps also revealed a Ca signal, albeit a weak one. Ca may be expected to exist in the tested cathode, as it is intentionally added as part of the impregnate mix during cathode fabrication; however, Ca is not typically detected at significant levels during EDS analysis. However, when it is detected in the near-surface region of an activated and tested scandate cathode, the Ca signal is usually coupled with a Sc signal, as is the case in Figure 6. This may be due to the presence of a mixed Sc-Ca-oxide at certain locations, or remnant CaO (from the impregnate material) situated above/below the Sc2O3; however, the low number of such observed particles precludes a definitive statement at this point. In order to further investigate the relation between BaAl 2 O 4 and Sc 2 O 3 , including possible physical mixing or chemical interactions between the oxides, a small W particle with attached oxides was selected for detailed visualization via a 3D EDS tomogram; this region is indicated by the red box in Figure 5a. It is noted that this experiment was conducted with a separate TEM lamella than the one imaged in Figures 1-4, but both lamellae were cut from cathode #1. The higher magnification image in Figure 5b shows a BaAl 2 O 4 particle atop the larger W particle. 2D slices from the 3D EDS tomogram are presented in Figure 5c-e, showing progressively more elements in each EDS map. This figure sequence conveys the relative distribution of Ba, Sc and W, and confirms that Sc is distinct from both the W particle and the BaAl 2 O 4 particle that encases it, i.e., Sc does not appear to be physically or chemically mixed with W or BaAl 2 O 4 . Video Supplementary Video S1 shows the 3D tomogram in imaging mode and in EDS mapping mode, and also rotated about the vertical axis, is available in the online Supplemental Files. The EDS composite elemental mapping in this 3D tomography movie corroborates the observations described above and, along with the observations portrayed in Figures 1-4, indicates that Sc exists in distinct Sc 2 O 3 oxide particles that decorate the surfaces of W grains.
The Ba-Al oxide particle marked by the leftmost white arrow of Figure 2 was characterized at high-magnification, since it was adjacent to a Sc-oxide particle and offered an opportune location for more detailed study of these distinct but adjacent oxides. Figure 6 presents high spatial resolution images and elemental maps of these particles; as seen by comparison of Figures 2a and 6a, the TEM lamella has been imaged from the opposite orientation in Figure 6, i.e., the features are reversed left-to-right. The HAADF image of Figure 6b, corresponding to the red box in Figure 6a, shows a wide darker gray region that indicates lower average atomic number and therefore suggests the presence of oxide particle(s). The elemental maps of Figure 6c-g agree with other observations of distinct BaAl 2 O 4 and Sc 2 O 3 particles and, moreover, they show a sharp boundary between Sc and Al. These high spatial resolution EDS maps also revealed a Ca signal, albeit a weak one. Ca may be expected to exist in the tested cathode, as it is intentionally added as part of the impregnate mix during cathode fabrication; however, Ca is not typically detected at significant levels during EDS analysis. However, when it is detected in the near-surface region of an activated and tested scandate cathode, the Ca signal is usually coupled with a Sc signal, as is the case in Figure 6. This may be due to the presence of a mixed Sc-Ca-oxide at certain locations, or remnant CaO (from the impregnate material) situated above/below the Sc 2 O 3 ; however, the low number of such observed particles precludes a definitive statement at this point. Figure 5. (a) HAADF image of near-surface cathode region where structure and elemental distribution were investigated by tomography. (b-e) HAADF image and EDS elemental maps, shown as 2D slices from the 3D tomogram, of a Ba-Al oxide particle that encases a smaller Sc-containing particle and sits on a W grain. Sc is distinct from both the W particle and the Ba-Al oxide. Quantitative EDS analysis was performed on the Ba-containing and Sc-containing oxide particles in Figure 6, in the regions marked by blue and yellow boxes in Figure 6b, respectively. These results are presented in Table 1 1). Again, this agrees with observations described above, except for the presence of Ca. Given the additional presence of a low but not insignificant amount of Ba in the yellow box region, it is proposed that these signals may arise from remnant amounts of CaO and BaO that were added in the impregnate mix. Therefore, the primary oxide particle in the yellow box region would be Sc 2 O 3 , along with low amounts of CaO and BaO. Table 1. Quantified measurement of particle composition, obtained from EDS point analysis of the regions marked in Figure 6.

EDS Site Measured Composition (at.%) Compound(s) Present
Blue box Ba (9. In addition to the oxide particles that were observed at the surfaces and edges of W grains, the Sc-containing nanoparticles embedded in the interior regions of larger W particles were investigated using a combination of HAADF imaging and high spatial resolution EDS mapping, as shown in Figure 7. The red box in Figure 7a indicates the Sc particle of interest, with a corresponding magnified image in Figure 7b. The white spots dispersed throughout the Sc particle, as seen in the blue box of Figure 7b, appear to be spherical in shape, with a diameter in the range of 1-10 nm. The inset image and elemental map in Figure 7b reveal that the nanoscale particles dispersed throughout the larger (~100 nm) Sc-containing particle are W. A possible explanation for this curious result is that W may be etched during the L-S fabrication process, and W surfaces may become rough or pitted, allowing Sc 2 O 3 to form in these pitted regions. This is discussed in more detail below.

Confirmation of Crystal Structure of W Grains
α-W is the most common phase of tungsten, and it exists in the body centered cubic (BCC) crystal structure. The structure of the W particles that form the cathode body were studied in more detail here. It would be convenient if the structure of W in the cathode body were the same as ubiquitous α-W, but given the complex chemical reactions that take place during preparation and emission testing of a cathode, this may not be true, and therefore the crystal structure was characterized. Scanning nanobeam diffraction was performed on a TEM lamella extracted from cathode #1. In this diffraction experiment, the beam was shifted along a line-scan direction with a step size of 20 nm, from point A to point C in Figure 8a, in order to determine the crystal structure at regular intervals from the grain interior to its surface. Diffraction patterns at three points of this line scan are displayed in Figure 8b-d, which correspond to sites A, B, and C, respectively. In all cases, the diffraction pattern corresponds to a typical BCC structure oriented along a <111> zone axis. The diffraction patterns obtained at sites B and C are identical. The indexed diffraction pattern displayed in the lower right corner of Figure 8d is representative of all locations characterized in this investigation, confirming that the W grains exist in the BCC crystal structure. In other words, the W matrix has not undergone significant changes due to activation or emission testing. Note that Supplementary Video S2 portrays the scanning nanobeam diffraction experiment and includes the diffraction patterns recorded at each 20 nm step position.

D Tomographic Reconstruction
Standard characterization techniques such as SEM, TEM and AFM typically reveal planar, projected or surface information, and therefore may not be ideal for revealing internal microstructure. As such, 3D reconstruction can provide insights into the microstructure of a complex 3D object that is not attainable with 2D characterization [39]. In the current study, a 3D tomogram was constructed for the first time on a scandate cathode, by employing the dual-beam FIB-SEM microscope. This technique involves layer-by-layer milling/removal of thin slices of the sample, using the ion beam, with SEM images recorded after each slice. 10 nm thick slices were milled sequentially, culminating in 137 backscattered electron (BSE) images stitched together to form a 3D tomogram. A schematic of the serial sectioning technique is presented in Figure 9. In order to perform FIB milling, the sample stage is tilted to 52° and the region of interest is situated at the eucentric height, which is where the sample region of interest will not change its vertical position during tilting (it is also where the ion beam and the electron beam converge). Prior to slicing, the sample was coated with a layer of Pt to protect the underlying region of interest. The x-y plane is perpendicular to the ion beam, and the x-z plane (parallel to the ion beam) is imaged after each slice. The red lines in Figure 9 represent the serial sectioning that is performed at regular intervals along the y-axis of the sample.

3D Tomographic Reconstruction
Standard characterization techniques such as SEM, TEM and AFM typically reveal planar, projected or surface information, and therefore may not be ideal for revealing internal microstructure. As such, 3D reconstruction can provide insights into the microstructure of a complex 3D object that is not attainable with 2D characterization [39]. In the current study, a 3D tomogram was constructed for the first time on a scandate cathode, by employing the dual-beam FIB-SEM microscope. This technique involves layer-by-layer milling/removal of thin slices of the sample, using the ion beam, with SEM images recorded after each slice. 10 nm thick slices were milled sequentially, culminating in 137 backscattered electron (BSE) images stitched together to form a 3D tomogram. A schematic of the serial sectioning technique is presented in Figure 9. In order to perform FIB milling, the sample stage is tilted to 52 • and the region of interest is situated at the eucentric height, which is where the sample region of interest will not change its vertical position during tilting (it is also where the ion beam and the electron beam converge). Prior to slicing, the sample was coated with a layer of Pt to protect the underlying region of interest. The x-y plane is perpendicular to the ion beam, and the x-z plane (parallel to the ion beam) is imaged after each slice. The red lines in Figure 9 represent the serial sectioning that is performed at regular intervals along the y-axis of the sample. stage is tilted to 52° and the region of interest is situated at the eucentric height, which is where the sample region of interest will not change its vertical position during tilting (it is also where the ion beam and the electron beam converge). Prior to slicing, the sample was coated with a layer of Pt to protect the underlying region of interest. The x-y plane is perpendicular to the ion beam, and the x-z plane (parallel to the ion beam) is imaged after each slice. The red lines in Figure 9 represent the serial sectioning that is performed at regular intervals along the y-axis of the sample.  The 3D reconstructed shape of a particular W particle was generated using Avizo software, and involved 137 BSE micrographs obtained after each step of automated serial sectioning. This W particle is shown in Figure 10a and exhibits surface facets. A noise reduction filter was applied to each micrograph prior to reconstruction. Each image was aligned with respect to a fiducial mark, resulting in the reconstructed tomogram displayed in Figure 10b. The SEM image of the selected W particle in Figure 10a and the 3D reconstruction of that particle in Figure 10b agree well with each other, and the surface facets are consistent and clear in both figures. The linear surface steps in Figure 10b result from the serial slicing process, specifically from slight misalignments between image positions after each slice, i.e., these surface steps are an artifact of reconstruction. Since BSE imaging was used to generate the tomogram, it was possible to optimize atomic number contrast for setting thresholds in the reconstruction. The metallic W regions exhibited consistent brightness levels, while the oxide particles were darker gray and clearly distinguishable from the main W particle. Leveraging this contrast, the (darker) oxide particles were differentiated, and Figure 10c reveals Sc/Sc 2 O 3 particles trapped inside the larger W particle. As shown in Movie Supplementary Movie S3, where the tomogram of Figure 10c is rotated and viewed from multiple directions, the smaller oxide particles can be seen to lie inside the W particle. These observations are consistent with the results described earlier, namely that Sc/Sc 2 O 3 particles are contained inside larger W grains. Figure 10. 3D reconstruction of a W particle from scandate cathode #2, generated by FIB serial sectioning and imaging in the SEM. (a) Secondary electron SEM micrograph of the selected W particle, (b) reconstructed tomogram, and (c) spatial distribution of Sc/Sc 2 O 3 inside the W particle. Note that the viewing direction of image (c) is rotated with respect to that of images (a,b).
To further investigate the source of the Sc 2 O 3 particles that were observed inside the W grains of the cathode pellet, the starting materials were characterized. This included pure W powder, as well as Sc 2 O 3 -doped W powder that had been processed from the same W powder using the L-S technique. Figure 11 presents SEM micrographs of each powder set, showing particle morphology and distribution. As seen in Figure 11a, the W particles have clean, flat surface facets prior to the L-S processing step. However, after the L-S stage, W surfaces are decorated with Sc 2 O 3 particles and W exhibits microfacets at the intersections of neighboring facets. These microfacets may result from etching of W during L-S processing. Additionally, there are numerous cavities that appear to have been etched into the W grains, following by backfilling of these cavities with Sc 2 O 3 particles. An example of this is shown in the central region of Figure 11b. Thus, the L-S processing step may allow Sc 2 O 3 to penetrate some distance into W grains, which later form seemingly regular, faceted W grains during high-temperature activation and emission testing stages. In this way, the Sc 2 O 3 particles may become trapped within W, resulting in the microstructural features that were described above.

Discussion
Since the primary materials difference between scandate cathodes and conventional impregnated tungsten cathodes is the addition of scandia to the W matrix, it is reasonable to assume that scandia (or scandium) is responsible for the significant improvement in thermionic emission and the low work function of scandate cathodes. However, the exact role of Sc/Sc2O3 in cathode processing and/or performance remains unclear. Uda et al. [40] suggested it is impossible for Sc2O3 to migrate from interior regions of the porous W matrix to the cathode surface during activation, due to the low diffusion coefficient (1.6 × 10 −18 cm 2 /s at 1370 K). Therefore, in order for Sc to diffuse or migrate to the cathode surface within a reasonable timeframe, Sc2O3 would need to be transformed into metallic Sc, as proposed by Wang [41] and Liu [32]. However, the reaction(s) required to generate free Sc and the mechanism of Sc migration have not been experimentally verified, despite these research efforts. In this section, chemical reactions that may enable the creation of metallic Sc are discussed, as are Sccontaining compounds that may be relevant to the role of Sc in scandate cathodes.
Sc has a stronger affinity for O than does Ba, which means that Ba cannot reduce Sc2O3 to Sc directly [9]. In a cathode study by Yamamoto et al. [35], it was proposed that metallic Sc can be produced via a chemical reaction between Sc2W3O12 and Ba, as described in Equation (1), which was based on their work on (W-Sc2O3)-coated [42] and (W-Sc2W3O12)-coated impregnated cathodes [35].
Sc2W3O12 + 3Ba = 3BaWO4 + 2Sc (1) In that reaction scenario, metallic Ba was created by impregnate materials reacting with W at pore walls, similar to the case for regular impregnated tungsten cathodes. Multi-step chemical reactions that generate Ba were described by Jones [18] as: 2Ba3Al2O6 (s) + W (s) = BaWO4 (s) + 2BaAl2O4 (s) + 3Ba (g) According to these reactions, the production of free metallic Ba is accompanied by generation of the compound BaAl2O4. As demonstrated by Rittner [43], this compound can be regarded as an inert material, i.e. it is thermally stable and does not participate in thermionic electron emission from the cathode surface. Since Equations (2) and (3) take place at W pore walls, i.e., the surfaces of W grains, Figure 11. SEM micrographs of (a) tungsten powder and (b) scandia-doped tungsten powder. Both images were recorded in secondary electron imaging mode.

Discussion
Since the primary materials difference between scandate cathodes and conventional impregnated tungsten cathodes is the addition of scandia to the W matrix, it is reasonable to assume that scandia (or scandium) is responsible for the significant improvement in thermionic emission and the low work function of scandate cathodes. However, the exact role of Sc/Sc 2 O 3 in cathode processing and/or performance remains unclear. Uda et al. [40] suggested it is impossible for Sc 2 O 3 to migrate from interior regions of the porous W matrix to the cathode surface during activation, due to the low diffusion coefficient (1.6 × 10 −18 cm 2 /s at 1370 K). Therefore, in order for Sc to diffuse or migrate to the cathode surface within a reasonable timeframe, Sc 2 O 3 would need to be transformed into metallic Sc, as proposed by Wang [41] and Liu [32]. However, the reaction(s) required to generate free Sc and the mechanism of Sc migration have not been experimentally verified, despite these research efforts. In this section, chemical reactions that may enable the creation of metallic Sc are discussed, as are Sc-containing compounds that may be relevant to the role of Sc in scandate cathodes.
Sc has a stronger affinity for O than does Ba, which means that Ba cannot reduce Sc 2 O 3 to Sc directly [9]. In a cathode study by Yamamoto et al. [35], it was proposed that metallic Sc can be produced via a chemical reaction between Sc 2 W 3 O 12 and Ba, as described in Equation (1), which was based on their work on (W-Sc 2 O 3 )-coated [42] and (W-Sc 2 W 3 O 12 )-coated impregnated cathodes [35]. Sc 2 W 3 O 12 + 3Ba = 3BaWO 4 + 2Sc (1) In that reaction scenario, metallic Ba was created by impregnate materials reacting with W at pore walls, similar to the case for regular impregnated tungsten cathodes. Multi-step chemical reactions that generate Ba were described by Jones [18] According to these reactions, the production of free metallic Ba is accompanied by generation of the compound BaAl 2 O 4 . As demonstrated by Rittner [43], this compound can be regarded as an inert material, i.e. it is thermally stable and does not participate in thermionic electron emission from the cathode surface. Since Equations (2) and (3) take place at W pore walls, i.e., the surfaces of W grains, it is reasonable to observe BaAl 2 O 4 particles at W surfaces and junctions in cathode cross-section samples, as was shown for example in Figure 3.
As described by Yamamoto [35], the compound Sc 2 W 3 O 12 , which can result from oxidation during heat treatment of W-Sc 2 O 3 , is the most likely source of metallic Sc and would be produced according to Equation (1) above. However, Sc 2 W 3 O 12 is not observed in the cathode samples described in the current paper, and this may be due to complete reaction of the compound Sc 2 W 3 O 12 . By virtue of their nanoscale size and uniform distribution across W surfaces, the Sc 2 O 3 nanoparticles fabricated during L-S processing in the current study may react more readily with W, and thereby enhance the formation of Sc 2 W 3 O 12 . If these mixed-oxide particles are also sufficiently small and uniformly distributed over W pore walls, this would facilitate the creation of free metallic Sc throughout the porous network.
Additionally, according to Slooten [44], the rate of Sc supply is significant only in the initial hours of cathode operation. After this transient stage, the generation of metallic Sc decreases to an undetectable level due to the formation of stable compounds such as Ba 3 Sc 4 O 9 or Ba 2 ScAlO 5 , which inhibit electron emission. However, these phases were not observed during the characterization work in the current study. Future efforts will include a more detailed study of Sc-containing compounds in scandate cathodes.
The exact role of Sc in scandate cathode processing and performance remains elusive, although there is now an improved understanding of the various phases in the near-surface regions of scandate cathodes. Other studies in the scientific literature ascribe the enhanced emission capability and low work function of scandate cathodes to a Ba-Sc-O monolayer [29,45] or a relatively thick (~100 nm) Ba-Sc-O layer [7,8,32] that covers W grains at the cathode surface. However, as was described in a recent paper from the current authors' group [37,46], no Ba-Sc-O layer with thickness on the order of 10-100 nm was observed, despite using several pertinent and complementary techniques in the characterization of multiple scandate cathodes that had yielded good emission test results. Finally, as suggested by Hasker [29] and Zhou [46], one possible function of Sc is to regulate the chemical potential of oxygen on the emitting surface of a scandate cathode, since the affinity of O for Sc is greater than it is for Ba.

Conclusions
Multiple scandate cathodes were prepared using the liquid-solid fabrication method and were subjected to CSD emission testing. Subsequently, they were characterized using multiple advanced microscopy techniques, including (S)TEM imaging, high spatial resolution EDS mapping, EELS analysis, scanning nanobeam diffraction, and 3D reconstruction/tomography. The experimental observations and discussion above are summarized with these concluding remarks: • 3D tomographic reconstruction revealed the interior microstructure of W grains in a scandate cathode. Interestingly, nanoscale Sc/Sc 2 O 3 particles tend to cluster or align in certain regions of a given W grain, rather than being distributed uniformly throughout the W grain.