J. R. Bowen, Department of Energy Conversion and Storage, Technical University of Denmark, Risø Campus, Frederiksborgvej 399, Roskilde, Denmark 4000. Tel: +45 4677 4720; fax: +45 4677 5858; e-mail: email@example.com
Serial sectioning by focused ion beam milling for three-dimensional electron backscatter diffraction (3D-EBSD) can create surface damage and amorphization in certain materials and consequently reduce the EBSD signal quality. Poor EBSD signal causes longer data acquisition time due to signal averaging and/or poor 3D-EBSD data quality. In this work a low kV focused ion beam was successfully implemented to automatically polish surfaces during 3D-EBSD of La- and Nb-doped strontium titanate of volume 12.6 × 12.6 × 3.0 μm. The key to achieving this technique is the combination of a defocused low kV high current ion beam and line scan milling. The line scan was used to restrict polishing to the sample surface and the ion beam was defocused to ensure the beam contacted the complete sample surface. In this study 1 min polishing time per slice increases total acquisition time by approximately 3.3% of normal 3D-EBSD mapping compared to a significant increase of indexing percentage and pattern quality. The polishing performance in this investigation is discussed, and two potential methods for further improvement are presented.
Three-dimensional electron backscattered diffraction (3D-EBSD) by focused ion beam (FIB) serial sectioning enables the determination of true crystallographic orientation of grain morphology and grain boundary information. The technique has been used for 3D material analysis in, for example, recrystallization (Gholinia et al., 2010), grain boundary characterization (Bastos et al., 2008; Dillon & Rohrer, 2009; Khorashadizadeh et al., 2011), texture analysis (Jin et al., 2005; Konrad et al., 2006; Petrov et al., 2007; Zaafarani et al., 2006), grain growth (Bastos et al., 2008; Liu et al., 2008) and deformation (Lin et al., 2010). Even though the FIB is a powerful tool integrated in modern electron microscopes it is well known that FIB milling is inherently destructive to specimen surfaces, for example, crystal structure damage and amorphization of the milled surface (Pelaz et al., 2004; Rubanov & Munroe, 2005). Poor EBSD signal can yield a longer data acquisition time from signal averaging and/or poor quality 3D-EBSD data. From our previous work (Saowadee et al., 2012) to study the effect of FIB milling (30 kV) on Nb-doped strontium titanate and stabilized zirconia we have shown that electron backscatter diffraction pattern quality, in terms of band contrast and band slope,1 is decreased by approximately 60% relative to mechanical polishing on strontium titanate but does not have any significant effect on yttria stabilized zirconia. The damage from FIB milling can be reduced by low kV FIB polishing as it is used in transmission electron microscopy sample preparation (Michael & Giannuzzi, 2007; Michael & Kotula, 2008). Our work (Saowadee et al., 2012) also shows that on Nb-doped strontium titanate 5 kV polishing can lead to a 100% improvement of band contrast and band slope and approximately 20% improvement of indexing percentage relative to 30 kV FIB milling. In this work the low kV FIB polishing was included in the normal 3D-EBSD process to improve data quality and is described below. La- and Nb-doped strontium titanate was used in this study as it is known to suffer from Ga+ ion beam damage and reduced EBSD pattern quality (Saowadee et al., 2012).
FIB polishing for 3D-EBSD
Normally, 3D-EBSD data collection by FIB serial sectioning comprises of two main processes: FIB milling and EBSD data collection as illustrated in Figure 1(a). The sample is first moved to the FIB milling position and a designated area is milled to create a flat and smooth surface for EBSD data collection. The sample is then moved to the EBSD position to collect an EBSD map on the milling surface. Figure 1(b) shows a diagram of 3D-EBSD with low kV FIB polishing. The polishing process is performed at the milling position after the milling process is finished. The ion beam is switched to a low kV probe for polishing which for the present instrument involves an FIB gun shutdown between voltage changes.2 The selected milling shape for the low kV polishing is a line scan, as polishing should be restricted to the near surface and 2D scanning will increase the polishing time due to scanning of redundant areas. We observe large random beam shift (in the order of 1–2 μm) when switching to the low kV probe in our instrument. Thus a focused low kV polishing line often will not impinge on the EBSD mapping surface for each slice. Furthermore, as present 3D-EBSD systems are not designed to incorporate polishing there is no software capacity for additional drift correction. Therefore, in our method, the low kV probe is defocused to broaden the beam to ensure that the polishing beam will impinge on the working surface. Figure 2(a) shows the complete defocused line scan to the left of the alignment fiducial mark. Figure 2(b) shows a working area after performing 3D-EBSD with FIB polishing.
Sample preparation and experimental methods
Sr0.96La0.02Ti0.9Nb0.1O3 was used in this study. The Sr0.96La0.02Ti0.9Nb0.1O3 was synthesized into bulk pellets by isostatic pressing and was sintered at 1450°C in 9% H2/Ar. For a detailed sample preparation see Saowadee et al. (2012). Sr0.96La0.02Ti0.9Nb0.1O3 has a cubic lattice and is easily indexed using the crystal structure of SrTiO3 (Brous et al., 1953).
FIB-SEM and EBSD for this work was performed on a Zeiss CrossBeam 1540XB™ (Oberkochen, Germany) equipped with an Oxford Instruments Nordlys S™ EBSD detector (Hobro, Denmark). Oxford Instruments’ software HKL Fast Acquisition 1.3 and CHANNEL 5 were used for data collection and analysis. A 30 kV 2 nA FIB probe was used for milling material and a 5 kV 2.5 nA probe was used for polishing. EBSD mapping was performed with a scanning electron microscope voltage of 20 kV using the 60 μm aperture and high current mode that yields the electron beam current of approximately 7.2 nA. The EBSD camera was set to the same parameters as used in Saowadee et al. (2012). To compare the effect of signal averaging and FIB polishing on electron backscatter diffraction pattern quality improvement four EBSD maps of 12.6 × 12.6 μm with step size 0.075 μm were performed on the same FIB milling surface with different signal averaging and polishing conditions. The first map was collected with electron backscatter diffraction pattern frame average 1 where the second map was collected with frame average 2. Subsequently the surface was polished with the 5 kV probe. The third map and the fourth map were collected on the polished surface with frame average of 1 and 2, respectively. Two 3D-EBSD data sets of volume 12.6 × 12.6 × 3.0 μm were collected with EBSD step size 0.075 μm and slice thickness 0.1 μm. The first data set was collected by the normal 3D-EBSD process with frame average 1 and the second data set was collected by 3D-EBSD with FIB polishing and frame average 2. These data sets represent data collection for maximum speed and optimum data quality, respectively. The polishing time for each slice was 1 min.
Results and discussion
Table 1 shows indexing rate, average band contrast and average band slope of the four EBSD maps and the two 3D-EBSD maps. Relative to the milled surface, increasing the number of averaged frames from 1 to 2 increases indexing rate by 23% and improves average band contrast and band slope by approximately 8% and 14%, respectively. Whereas 1 min of low kV FIB polishing can increase the indexing rate by approximately 41% and improves the average band contrast and band slope by approximately 42% and 48%, respectively, relative to the unpolished equivalent. Increasing the frame averaging from 1 to 2 on the polished surface further improves the indexing rate by 8%, band contrast by 8% and band slope by 12%. Inverse pole figure (x-axis) plots of the four EBSD maps are shown in Figure 3. As indicated by white circles in Figures 3(a) and (c), some small grains in the unpolished map are missing but can be observed in the polished map. In Figure 3(c) zero solutions (black dots) are significantly more prevalent in the upper area of the map compared to the lower area; note that the ion beam mills from the bottom of the image. This indicates that polishing performance decreases as the beam mills further down the working surface, that is, from the bottom to the top of the maps in the direction of the arrow in Figure 3. Thus the heterogeneous polishing is likely to be caused by a more broadened ion beam at areas far from milling edge. This can possibly be solved by increasing the polishing time. Although we cannot eliminate the effects of re-deposition from the available results, we believe that beam broadening is a more likely cause of the loss of polishing performance. 3D-EBSD maps of the Sr0.96La0.02Ti0.9Nb0.1O3 sample with and without FIB polishing are shown in Figures 4(a) and (b), respectively. The acquisition time per slice of the polishing data set is approximately 30 min: FIB milling time 6 min, FIB polishing time 1 min and EBSD mapping time 23 min (frame averaging 2). For the additional polishing step the acquisition time increases modestly by 3.3% compared to the significant improvement of the indexing rate and pattern quality.
Table 1. Indexing rate, average band contrast and average band slope of the four electron backscatter diffraction (EBSD) maps and two 3D-EBSD maps. The acquisition time for each map/slice is 11:29 and 22:57 (min:s) for averaging of 1 and 2 frames, respectively.
Band contrast (0–255)
Band slope (0–255)
During 3D-EBSD we observed that the polishing beam removes relatively more material where the ion beam first meets the sample yielding a curved edge. The depth of the curved edge was observed to increase from the first slice to the last slice. The curved edge depth at the last slice is approximately 5 μm. In this experiment EBSD data were acquired at a safe distance, that is, approximately 10 μm from the edge. However, we performed an additional EBSD map on the curved surface of the last slice and found that the mapping quality at the curved edge is better than the flat part of the polished surface as illustrated in Figure 5(a). This means the polishing performance may be further improved by creating a third new sample position particularly for polishing. This would be achieved by slightly increasing the sample tilt angle, thus the defocused ion beam will impinge on the sample surface at a small angle as illustrated in Figure 5(b). By this method the damaged or amorphous layer can be removed quicker than milling with a parallel beam. Further experiments are though required to identify the optimum tilt angle and polishing time. If the beam is exposed to the surface for too long or the beam angle is too large this may cause further damage to the surface. In this manner it is expected that the curved edge problem may be reduced and polishing times could be significantly reduced.
Another potential method to improve the polishing performance is to use the focused polishing beam parallel to the sample surface. However this method requires precise position realignment (i.e. drift correction) after switching the FIB gun to low kV. For our instrument, the focused low kV ion beam diameter of a high current probe can be in the order of a micron. If the position realignment error is less than a half micron, low kV FIB polishing using a line scan of a high current beam should be achievable. Apart from the curved edge no further artefacts, such as curtaining, were observed on the surface whilst using high current low kV FIB polishing.
Automatic low kV FIB polishing is successfully applied for 3D-EBSD of La- and Nb-doped strontium titanate. The selected milling shape for the low kV polishing is a line scan as polishing should be restricted to the near surface and 2D scanning will increase the polishing time due to scanning of redundant areas. The low kV beam is defocused to ensure that the polishing beam is always in contact with the working surface. The polishing time per slice used in this study (1 min) yields a modest total acquisition time increase of 3.3% relative to normal 3D-EBSD data acquisition. Furthermore, polishing leads to a significant improvement of the index rate and pattern quality. Lastly, two potential methods to further improve FIB polishing during automatic 3D-EBSD are proposed.
Band slope and contrast are EBSP pattern quality parameters specific to Oxford Instruments EBSD systems.
We have not observed adverse effects of repeated gun shut downs and voltage change on gallium source lifetime.