There is a critical need to analyze many material systems in three dimensions (3D), for example to understand the connectivity of phases, porous networks, and complex shapes. Fortunately, there are now several tools available for 3D characterization, for example, X-ray computed tomography (CT) , serial section SEM Tomography (SST) [2–4], transmission electron tomography , and atom probe tomography [5, 6], each covering different length scales (see Figure 1).
The emergence of focused ion beam scanning electron microscopy (FIB-SEM) using a gallium ion FIB has provided a means of accessing volumes of interest (VoI) using SST. The Ga+ FIB-SEM is also an important tool for the creation of transmission electron microscopy (TEM) samples. In practice, acquisition times limit the method to volumes about (50 μm)3 for site-specific 3D analysis using SST close to the surface with slice thicknesses down to ~10 nm. Despite its destructive nature, SST enables detailed 3D imaging of phases, grain structure (via electron backscatter diffraction (EBSD)), and chemistry (by energy dispersive X-ray spectroscopy (EDS)) . However there are many cases where it would be of interest to probe large VoIs that are submerged deeper within the sample. In an effort to respond to this challenge, the concept of correlative tomography, the 3D equivalent of correlative microscopy, has been proposed as a way of studying a VoI over multiple scales by coupling X-ray CT and SST to acquire multiple types of data (structural, crystallographic, chemical, etc.) that can be brought into registry for the same region .
Recently Xe+ plasma FIBs [10, 11] have emerged, demonstrating faster materials removal rates compared to Ga+ liquid metal ion source FIBs . In this article, we examine the capabilities of a Xe+ Plasma FIB (PFIB)-SEM dual-beam system for 3D analysis across a range of materials and applications. The FEI Helios PFIB used in this study has a maximum current of 1.3 μA compared to a maximum current of ~65 nA for many Ga+ FIB systems. In addition to a higher maximum current, which is useful for coarse milling duties, we can operate the PFIB at currents up to 180 nA for the preparation of high-quality cross sections for imaging. In addition, the Xe+ PFIB avoids potentially harmful chemical reactions, such as those that can result from Ga+ FIB milling . We are continuing to build our understanding of this technology to enable us to apply it to a large range of potential applications. Here we present four example applications where we have successfully applied the PFIB: describing the size of the volumes interrogated, the resolution, and the 3D reconstruction.
Three different serial sectioning results are presented: a stress corrosion crack tip in a 7000 series Al alloy, a Zr-based bulk metallic glass (BMG) with embedded dendrites, and a tungsten carbide hard metal with a cobalt binder phase (WC-11 wt% Co) analyzed by 3D EBSD. A fourth example shows the preparation of a site-specific pillar of human dentine for nanoscale X-ray CT.
All experiments were conducted using an FEI Helios PFIB operated at 30 kV. For the serial sectioning studies, a current of 59 nA was used within the automated procedure whilst higher currents were employed for the initial shaping of the block for serial sectioning. The FEI Auto Slice and ViewTM (ASNV) software was used to manage the slicing and imaging process with integrated rocking milling. This method has been developed to minimize curtaining that may arise during unidirectional FIB milling. It uses a stage move to create each alternate slice with the incident ion beam coming from +/- 5° with respect to vertical top-down 0° milling. Details of the setup can be found in . The Nanoscale X-ray CT measurements were made using a Zeiss Xradia Ultra 810 operating at 5.5 kV. All 3D data were aligned, post-processed, and visualized using FEI Avizo 9.0.0 software.
A stress corrosion crack in a 7000 series aluminum alloy is presented to show the ability to target site-specific locations using the PFIB-SEM. It is only possible to look at a part of this large crack, but it is of sufficient volume to appreciate the morphology at the 100 μm scale and relate it to the complete crack. A 50 nm slice thickness and a pixel size of 16 nm was used in the ASNV software, which was sufficient to observe many of the precipitates in the material and the particles decorating the grain boundaries. This made it possible to look for evidence of an interaction of the crack with these particles as it propagated along the grain boundaries.
The bulk metallic glass (BMG) sample was analyzed with a 100 nm slice thickness and a pixel size of 24 nm. This is a novel new material with a composition Zr58.5Ti14.3Nb5.2Cu6.1Ni4.9Be11.0 consisting of a bulk metallic glass matrix within which crystalline dendrites have grown. This material is being investigated as a high-strength, high-ductility material. The goal of the analysis was to understand the complex 3D microstructure of this material, in particular the size, morphology, and arrangement of the embedded dendrites. The size of the dendrites in this case was too large for effective observation using the limited volumes accessible using a Ga+ FIB-SEM.
This material is used in a number of applications where a very hard-wearing surface is required, which includes faces for various cutting tools. The WC-11 wt% Co sample is very challenging to prepare effectively using the Ga+ FIB, possibly because of some chemical interaction. Therefore images and EBSD maps were recorded from the PFIB-prepared cross sections of this material to assess the quality of results on what is considered a material that is difficult to FIB machine.
Lastly a biological sample of human dentine was prepared for examination with a Nanoscale X-ray CT (NanoCT). This insulating material illustrates a different perspective on the type of work that can be effectively carried out with the PFIB. The time needed to produce a pillar of ~50 μm diameter and ~100 μm height would be practically prohibitive in the Ga+ FIB. The bulk sample was gold-coated to alleviate charging. Coarse milling was conducted at 30 kV, and a series of milling currents of decreasing magnitude, from 1300 nA to 180 nA, was employed followed by a final shape-defining milling at 59 nA to give a smooth finish. A layer of Pt was deposited on top of the pillar to protect it during machining. Platinum was also used to attach the pillar to the EasyliftTM probe for lifting the pillar out of the bulk sample and placing it on the pin sample holder.
The results of the ASNV of the stress corrosion crack in the 7000 series aluminum sample are shown in Figure 2. The total volume imaged was 100 × 110 × 12 μm3, acquired with a total of 241 slices each 50 nm thick. The image pixel size was (16 nm)2 as recorded using the through-the-lens electron detector (TLD). The total time to collect 1 slice was 3 minutes including 40 s slicing time. Figure 2b shows a magnified region that has been post-processed with an adaptive histogram equalization and median filters. The grains and precipitates are clearly visible. Figure 2c shows a 3D rendering of the precipitates in a small volume of the material. The arrangement of these precipitate particles along the grain boundaries is visible. Many of the smaller intragranular precipitates are visible in the slice images but are typically smaller than the slice thickness of 50 nm, and so they are difficult to label separately. This range of scales (100 µm × 100 μm cross section and the 18 nm pixel size) allows understanding of the meaningful scales of the stress corrosion crack such that the results here can be linked to light microscopy and microscale X-ray CT studies going up in scale to the entire crack. Further details of the microstructure could also be obtained by linking to higher-resolution analysis using fine-scale SEM or possibly TEM analysis. The nature of stress corrosion cracks and the interaction they have with the microstructure demands a 3D understanding.
The bulk metallic glass sample is shown in Figure 3. The total volume imaged was 120 × 100 × 68 μm3, a total of 678 separate 100 nm thick slices. The image pixel size was (24 nm)2 as recorded using the TLD. The total time to collect 1 slice, including slicing and imaging was 120 s. The resulting microstructure is a complex array of dendrites that grow out from a central nucleation point, which is difficult to understand from a 2D section. As shown in Figures 3a and 3b, SST quickly reveals the 3D arrangement of the dendrites within the glass matrix and forms the basis for understanding the microstructure. Again the combination of large volumes (to capture multiple dendrites) with high resolution (to capture the dendrite arms), as well as the strong contrast, necessitates the use of the PFIB-SEM and its capabilities.
Figure 4 shows the microstructure of the WC-11 wt% Co specimen. Two volumes were captured: (1) an ASNV volume 100 × 100 × 80 μm3 with a total of 800 individual 100 nm thick slices. (2) A 3D EBSD volume 100 × 100 × 3 μm3, capturing a total of 33 slices at 100 nm thick. The EBSD conditions used were 20 kV and 22 nA, with a step size of 200 nm and a map size of 515 × 530 pixels. The total time to acquire the data for one slice was about 51 minutes: 51 s slicing time and 50 minutes to collect the EBSD map from each slice.
Figure 4a shows a 3D image of the structure. A large number of grains have been captured making these results statistically significant when trying to understand the size and shape distribution of the WC grains. A practical volume of (100 μm)3 is possible in the PFIB-SEM compared to a typical volume of (50 μm)3 captured in the Ga+ FIB-SEM. Based on an average grain size in this material of 11.5 μm, determined from the PFIB-SEM dataset, it is possible to capture ~660 grains compared to ~80 grains possible in the Ga+ FIB-SEM.
This material was specifically chosen for this 3D EBSD study because it has presented significant problems when investigated at a smaller scale within the Ga+ FIB-SEM. The WC phase is resistant to FIB machining and is difficult to prepare damage-free. Figures 4b to 4d show a series of maps collected from a single slice prepared during an automated EBSD procedure. It has been possible to prepare a large cross section of this material using PFIB at 30 kV and 59 nA with a time of only 51 s to create each new slice. The EBSD results show the indexing rate is over 80%. An issue facing this large-volume EBSD analysis is the time taken to collect the EBSD maps. In this study acquisition of 33 slices took ~30 hours, although it should be noted that much faster acquisition rates would be possible on many other materials such as ferrite, austenite, nickel, copper, etc. with mapping times of less than 10 minutes, depending on the map size and resolution required. The speed and quality of the PFIB cut opens up volumes of real significance, but time has to be spent optimizing the EBSD mapping for maximum speed. Despite the current level of optimization we have achieved, we need further hardware and software developments to alleviate this bottleneck in acquisition of large-volume, high-resolution 3D EBSD datasets.
Lastly, by connecting the non-invasive X-ray techniques to the destructive 3D imaging techniques within a correlative framework, there is a need to locate, excavate, and analyze sub-volumes in further detail. As well as providing high-resolution images of volumes, the PFIB promises to be able to excavate regions of interest, for example, creating samples for nanoscale X-ray CT. This is especially challenging for materials science specimens as an ideal sample size (in this case determined by the X-ray energy and their penetration) with a pillar that is 20–100 μm in diameter and approximately 100 μm tall. To achieve this, milling to a depth of over 100 μm is required, as well as a milled radius of the hole equal to the pillar height. These dimensions put this task beyond the practical means of the Ga+ FIB-SEM. The ability to create these samples from a site-specific location and then create a sample of the desired dimensions makes this an important capability, combining the precision of the cutting with the fast material-removal rate.
Figure 5a shows a pillar of human dentine prepared by the PFIB. This study had two aims: one was the desire to understand better the 3D morphology of the dentine tubules, and the second was to determine the accurate orientation of the tubules with respect to the subsequently prepared TEM lamellae. Understanding the detailed geometry of the dentine tubules at the sub-micron level is exceptionally difficult to appreciate from 2D cross sections. The quality of the NanoCT results was ensured by predicting the X-ray absorption of the sample and then specifically machining a pillar of the required dimensions using the PFIB. Figures 5b and 5c show the results of the nanoscale X-ray CT scan providing a (120 nm)3 voxel size and the ability to see the tubules and the increased density along the walls of the tubules.
These results give an overview of some applications where we have successfully applied the PFIB-SEM. The ability to prepare high-quality cross sections at high currents has meant that large volumes can be quickly captured with nanoscale resolution. This is an exciting area for development because many important engineering materials have grain structures many tens of micrometers and can only be characterized effectively over volumes hundreds of micrometers in size. In addition to the results presented here, we have also examined several other materials, including steel, aluminum, zirconium, titanium, alumina, various coatings, and an olivine rock sample. In each case, the parameters had to be optimized. With each new material comes a learning curve to optimize the conditions used. In addition we are continuing to push the range of materials and operating parameters used to evaluate the true limits of this technology. To date, single cross sections up to 1 mm wide have been produced, and serial-sectioning cross sections up to 300 μm across and 200 μm deep have been fabricated with slice thicknesses 50–500 nm thick.
More work still needs to be done to better understand material interactions with the Xe+ Plasma FIB. However, in comparison to traditional metallographic preparations of similar materials, PFIB-prepared samples have improved surface quality, yielding EBSD maps and high-resolution images that are essentially artifact-free.
3D data from serial sectioning using the PFIB-SEM across a range of materials suggest that the Xe+ PFIB can routinely and quickly provide 3D serial section tomographs over dimensions of many hundreds of micrometers whilst retaining nanoscale resolution. This tomographic technique has the ability to capture the rich detail of microstructures, as well as provide analysis by EBSD and EDS. The range of scales available helps bridge the gap between conventional Ga+ FIB and X-ray tomography for multiscale, multimodal 3D analysis.
We acknowledge the EPSRC for grants EP/J021229/1 and EP/M010619/1 and BIS Capital Funding that established the Multidisciplinary Characterization facility. Further authors gratefully acknowledge N. J. Henry Holroyd for the provision for the Aluminum sample, Peter Liaw from University of Tennessee for the BMG sample, and Mark Gee from NPL for the WC-Co sample. DWM would like thank Dr. J. S. Earl, GlaxoSmithKline Consumer Healthcare, for financial support. Our thanks also go to Rene Dobbe, Doug Hahn, and Trisha Rice at FEI Company for their support during this work.