Combined Focused Ion Beam-Ultramicrotomy Method for TEM Specimen Preparation of Porous Fine-Grained Materials

Abstract

A new transmission electron microscopy (TEM) specimen preparation method that utilizes a combination of focused ion beam (FIB) methods and ultramicrotomy is demonstrated. This combined method retains the benefit of site-specific sampling by FIB but eliminates ion beam-induced damage except at specimen edges and allows recovery of many consecutive sections. It is best applied to porous and/or fine-grained materials that are amenable to ultramicrotomy but are located in bulk samples that are not. The method is ideal for unique samples from which every specimen is precious, and we demonstrate its utility on fine-grained material from the one-of-a-kind Paris meteorite. Compared with a specimen prepared by conventional FIB methods, the final sections are uniformly thin and free from re-deposition and curtaining artifacts common in FIB specimens prepared from porous, heterogeneous samples.

Introduction

Focused ion beam (FIB) specimen preparation is a powerful and versatile technique to extract a thin transmission electron microscopy (TEM) specimen (<100 nm), or a specimen with complicated geometry, from a bulk sample (e.g., Heaney et al., 2001; Engelmann et al., 2003; Lee et al., 2003, 2019; Zega et al., 2007). A major advantage is the capability to extract such samples from specific locations identified in the bulk using well-defined patterns of sputter-milling and gas-assisted deposition. Microscopic features of interest can be analyzed thoroughly in the context of the bulk sample using a scanning electron microscope (SEM) or a dual-beam FIB-SEM to carry out secondary electron (SE) and/or back-scattered electron (BSE) imaging, as well as energy-dispersive X-ray spectroscopy (EDX). In a dual (or single) beam FIB, those specific microscopic features of interest can be extracted and prepared for (scanning) transmission electron microscopy (TEM/STEM). Typical TEM specimen size can vary between ~5 and 30 μm in width and ~5 and 10 μm in height. Despite its power and versatility, FIB specimen preparation has several disadvantages. In order to extract a single specimen, a relatively large footprint of material is usually removed by sputtering, typically covering an area of ~20 × 40 μm with a depth of ~15–20 μm. [There are a few notable exceptions, including the examples of Graham et al. and Berger and Keller, where multiple thin section specimens were extracted by FIB from small volume samples (Graham et al., 2008; Berger & Keller, 2015).] The Ga⁺ ion beam used to sputter-thin the specimen to electron transparency also creates an amorphous layer of about ~1 nm per 1 keV (acceleration voltage) on the specimen surfaces (McCaffrey et al., 2001; Mayer et al., 2007; Schaffer et al., 2012). Gradually reducing incident ion beam energy (low-kV polishing) can reduce the ion-induced amorphous layer thickness to ~3–5% of the total section thickness, but it remains a potentially significant complication for samples with indigenous amorphous components. Re-deposition of milled materials is another source of amorphous artifacts and is most significant for porous samples. Pores readily fill with the re-deposited material, obscuring the original morphology, and creating amorphous regions that can be difficult to distinguish from indigenous amorphous materials without extensive elemental mapping (Wirth, 2004; Kakizaki et al., 2017). Moreover, samples with pores and/or multiple phases can result in linear artifacts in the final specimen called “curtaining” (Ishitani et al., 2004; Montoya et al., 2007), fluctuations in sample thickness, due to different materials milling at different rates and to enhanced milling at pores, that complicate data interpretation.

Ultramicrotomy involves slicing an epoxy-embedded sample with a diamond knife and is well suited for TEM specimen preparation of porous and/or fine-grained materials; however, only small samples can be prepared in this manner because the sample has to fit in an epoxy bullet whose widest dimension is about 8 mm in diameter (Melo et al., 2016). Sample preparation for ultramicrotomy involves placing the sample on top of an epoxy bullet, depositing additional epoxy on top to fully encase the specimen, and trimming excess epoxy around the sample to form a flat “mesa” with well-defined sidewalls, often in a trapezoidal shape. To reduce the stress induced by the blade, the shape of the mesa and cutting speed can be modified. A major advantage of ultramicrotomy is that it can produce many TEM specimens with nearly constant thickness from one sample with almost no material loss, and those specimens are free from ion beam damage, re-deposition, or curtaining. However, unlike FIB specimen preparation, specific sites on the sample cannot be selected for ultramicrotoming, or “sectioning”, and sample sizes are limited for embedding and sectioning. (The terms TEM “specimen” and “section” are used interchangeably here to refer to electron transparent thin samples.) The hardness of material plays a role in the size of samples that can be ultramicrotomed. For hard materials, in particular, if the sample is too large, pull-out of some material from the cutting face and/or loss of material from the centers of sections is more likely to occur. Severe damage to the diamond knife also becomes more likely.

For inhomogeneous bulk samples, it can be crucial to be able to select specific sites of interest, and often the components in the sample or its size are unsuited to ultramicrotomy. The Paris meteorite sample, like most extraterrestrial samples, presents this dilemma. Paris is a recently discovered meteorite that is believed to be the most pristine example of its type (CM, or a chondrite that is Mighei-like) (Hewins et al., 2014). A handful of petrographic sections have been cut from the meteorite and mounted on slides for scientific analyses. These samples contain large crystalline silicate, metal, and sulfide minerals of varying dimensions (as large as a millimeter) that are embedded in a matrix of very fine-grained, porous material. The fine-grained matrix is comprised of amorphous silicates and micro- and nanoscale crystals, and this assemblage is of particular interest because its chondritic (solar) composition is like that of the most primitive dust from which the solar system formed (Leroux et al., 2015). The Paris matrix is appropriate for ultramicrotomy, but the larger components in the meteorite are not. Due to its one-of-a-kind nature, the sample cannot be ground or further diced; it must be retained in its original geometry for future analyses.

To overcome these challenges, we have developed a new combined FIB-ultramicrotomy TEM specimen preparation method. Prior sample preparation methods have used FIB to pick up and move single, small grains for ultramicrotomy and TEM analyses (Xu et al., 2018). Here, a block of sample material is extracted using FIB and mounted for ultramicrotomy in order to generate many TEM specimens. In this report, we describe the details of this process and discuss the merits of this approach.

Materials and Methods

A polished thick-flat surface of the Paris meteorite sample was first surveyed by electron microscopy and EDX. The Paris meteorite is heterogeneous in the degree of alteration it has experienced due to water and heat. In more altered areas, metal has been completely oxidized. Areas of the amorphous silicate matrix that still retain metal as inclusions were targeted to investigate less altered components, and a region of porous fine-grained material was selected as the site for extraction of a conventional FIB specimen and for extraction of a block for combined FIB-ultramicrotomy specimen preparation for TEM. Below, we describe the instrumentation used, the conventional FIB-preparation method, and the general procedure for combined FIB-ultramicrotomy specimen preparation.

Instrumentation

All of the scanning electron microscopy and FIB procedures for this work were carried out using the Helios Nanolab 660 DualBeam FIB-SEM (FEI, now Thermo Fisher Scientific) at the Advanced Electron Microscopy Center (AEMC) at the University of Hawai’i at Mānoa (UH). The UH FEI Helios 660 FIB is a dual-beam instrument with the ion column at 52° from the electron column. In addition to several standard detectors for SE, BSE, and secondary ion (SI) imaging, the instrument is equipped with a retractable concentric BSE detector, which was used for surface surveys. FIB deposition capabilities are enabled by three gas injection systems for Pt, W, and C. An EasyLift^™ micromanipulator permits the transfer of samples extracted from the bulk to specialized FIB grids for TEM/STEM analyses. A large-area (80 mm²) silicon drift detector (Oxford Instruments X-maxN 80 with AZtec Energy Advanced Microanalysis System) allows EDX and mapping for elemental chemistry.

Ultramicrotomy was performed using a Leica EM UC7 ultramicrotome at AEMC. Epoxy-embedded specimens were trimmed using glass knives followed by a diamond trimming knife (ultrathin, Diatome) for final trimming, and the final section cutting was performed using a diamond knife with a boat (ultra 35°, 4 mm, Diatome). Sections were floated onto deionized water in the knife boat and transferred to Cu grids using a Perfect Loop (Ted Pella, Inc.) for ultramicrotomy.

TEM bright-field and dark-field imaging were performed on an FEI Titan3 G2 60–300 aberration-corrected (scanning) TEM, referred to as a TEM/STEM, at 300 keV in order to assess the electron transparent specimens prepared.

Conventional FIB-Prepared TEM Specimen for Comparison

For comparison to the TEM specimens prepared by the combined FIB-ultramicrotomy method (described below), we studied another TEM specimen from a similar region in Paris prepared by conventional FIB methods (e.g., Engelmann et al., 2003; Lee et al., 2019). Because these methods are commonly used in TEM specimen preparation, we provide only a brief summary here: A Pt strap (~12 μm in length) is deposited, first by electron beam deposition (e-Pt) and then by ion beam deposition (i-Pt), to protect the region of interest from ion beam damage. Trenches are milled on either side of the Pt strap, and the surfaces are cleaned up using gradually reduced beam currents until the sample is about 1.5 μm in thickness. For samples with porosity, like the Paris meteorite matrix, the section is extracted from the bulk while still relatively thick in order to limit possible re-deposition in the interior via porosity. A J-cut is made using the ion beam at 0° tilt to undercut the section and mostly free it from the bulk. The micromanipulator needle is “welded” to the section using i-Pt, and the section is cut completely free from the bulk by ion beam sputtering. The section is transferred to and “welded” to a post on a Cu TEM half-grid. Ion beam thinning is carried out using long dwell times and gradually decreasing beam currents to minimize curtaining. For the specimen of Paris meteorite matrix, the final section thickness was limited to ~100–120 nm by the onset of warping. Final “polishing” of the FIB-prepared surface is carried out at 5 keV and then 2 keV to reduce the thickness of the ion beam-generated amorphous surface layer on each side of the TEM specimen.

Preparation of Combined FIB-Ultramicrotomy TEM Specimens

Here, we describe the general procedure for the combined FIB-ultramicrotomy specimen preparation. Details of electron and ion beam energies and currents, dimensions of deposition and milling steps used in preparing the sections of the Paris meteorite, and accompanying images are given in the Supplementary Material.

The first part of the procedure involves preparing a block of sample material for lift-out. Using the electron beam, 200 nm of Pt is first deposited on top of the site of interest in a rectangular pattern (on the order of 10 μm on a side) to protect the surface from Ga⁺ ion beam damage during subsequent sample preparation steps. For the fragile and porous amorphous surface of our Paris meteorite matrix, the thickness of the protective electron beam deposited Pt was chosen to be greater than that often chosen for conventional FIB section extraction in order to ensure that the surface was well protected. (Thinner e-Pt protective layers may be sufficient.) The protected area is much wider than is typical for conventional FIB sections in order to allow extraction of a block instead of the typical cross-sectional slab. Additional Pt, to a total thickness of ~4 μm, is then deposited using the ion beam with the sample tilted, so its surface is perpendicular to the ion beam (52° tilt in the Helios 660 FIB) (Fig. 1a). The first ~500 nm of i-Pt may be deposited using relatively low ion beam current to further reduce the likelihood of any ion beam damage to the specimen surface. The dimensions of the block are selected, so that the final extracted block can be observed under the optical microscope of the microtome instrument (~10 μm × 10 μm × 10 μm). With the sample tilted, so its surface is perpendicular to the ion beam, two trenches are sputter-milled using the ion beam on either side of the Pt-protected area, the same as in conventional FIB preparation of a TEM specimen (Fig. 1b). With the sample surface still perpendicular to the ion beam, a C-shaped pattern (C-cut) is milled into the sample with the Pt-protected region in its center. The C-cut is generated by overlapping rectangle patterns, leaving a 3 μm opening on one side where the specimen will remain attached to the bulk until lift-out. It is important to ensure the sides are milled all the way down to the target depth in this step of the process to avoid reconnecting the block to the bulk during the later undercut step. When the C-cut reaches the target depth, the sides of the block are thoroughly cleaned (i.e., made more vertical by additional material removal at lower ion beam current) with the ion beam at an extra degree of tilt (Fig. 1c). Depending on the materials being sputtered, re-deposition can be significant and the width of the C-cut has to account for the expected re-deposition during cleaning. At this stage, the opening of the C-cut is reduced down to 1 μm. Again, the re-deposition rate is higher at the corner of the opening due to the more crowded geometry. When all of the sides are cleaned up, the sample is tilted back to 0°, so that the surface is perpendicular to the electron beam, and a J-cut pattern is milled from both sides to undercut the block leaving a single connection to the bulk (Fig. 1d). The completion of the two J-cuts can be verified by observing from the opposite side of the C-cut opening (Fig. 1e). The bottom of the block thus formed has a V-shape when the cuts are completed. The final size of the block extracted from the Paris meteorite was 10 μm wide × 7 μm deep × 10 μm tall (at the point of the V-shape).

Fig. 1.

SE images of steps in the process of extracting and transferring a block of material from the bulk sample to a Cu TEM half-grid. (a) Ion beam Pt deposition, (b) trench milling, (c) C-cut (with cleaned up sides), (d) J-cut from both sides, (e) side view after J-cut, and (f) block attached on a Cu grid center post. The scale bar in each image is 10 μm.

The second part of the procedure, lift-out and weld-to-grid for a block, is similar to the conventional FIB procedure for TEM specimen preparation, but there are some key differences. The in situ micromanipulator is used to perform the lift-out of the block from the bulk sample. A corner of the block is attached to the tungsten micromanipulator needle using i-Pt, and then the block is detached from the bulk by ion beam sputter-milling. A Cu TEM half-grid is used to transfer the block to the microtome instrument. In conventional FIB preparation of TEM specimens, a FIB-extracted cross-sectional slab is attached on the side or in a “V” shape of a finger on the Cu TEM half-grid in order to protect it from being accidentally broken off. In the combined FIB-ultramicrotomy method, however, the block needs to be exposed with minimal Cu nearby to allow ultramicrotoming of slices of the specimen that do not contain high-Z Cu that might interfere with TEM analyses. As a result, the block is best placed on top of the center post. A shallow slot can be milled on the top of the Cu grid finger to better fit the block prior to attaching the block with i-Pt; this provides a more secure attachment because the bottom of the block has a V-shape that matches the slope produced by the ion beam when milling the slot (with the grid at 0° tilt). The block is secured to the Cu grid using i-Pt. After detaching the micromanipulator needle by ion beam milling, the sides of the block are milled until any re-deposited layer is removed and the specimen block is exposed (Fig. 1f).

The final portion of the procedure involves embedding and ultramicrotomy of the block-on-grid. Once the sides are cleaned of re-deposition, the grid is removed from the FIB instrument and embedded in epoxy (EmBed 812, Mixture B, Electron Microscopy Sciences). The grid is placed, using tweezers, in a narrow slot in the top of an epoxy bullet previously cured at 60°C under vacuum (25 inHg) (Micromoulds, Ted Pella, Inc.) that is made using a jeweler’s fret saw (400-μm blade width, Beadalon). Additional epoxy is placed on the grid to envelop the block-on-grid and cured (60°C at 25 inHg). Multiple applications of epoxy may be necessary to completely embed the block and grid. Cured epoxy around the block is trimmed in the ultramicrotome with a glass knife to remove the Cu “wings” on the sides of the grid as well as the posts to the left and right of the post on which the block sample is mounted (Fig. 2a). Depending on the design of the Cu TEM half-grid, some posts may be taller than the central post on which the sample is mounted. For the Cu grids used in this example, the epoxy mesa is trimmed to about 370 μm in width. Then, the mesa is trimmed on all sides using a diamond trimming knife, so the block is visible through the side (Fig. 2b) and so that the surface has a trapezoidal shape (acute angle 80° in this example). The mesa surface was then trimmed until the block is exposed at the surface. The distance from the epoxy mesa surface to the block surface can be roughly estimated by using the side view of the mesa (Fig. 2c). When the sectioning surface reaches the i-Pt layer on the sample block, a discontinuity on the surface of the sections is clearly visible. While one can estimate the progressing distance toward the actual sample block according to the thickness of the i-Pt, the presence of Pt in the sections prior to reaching the sample block surface can also be confirmed by optical microscopy. Figure 2d shows a microtomed section of Pt protective layer on a TEM grid: the Pt curls up leaving a window in the epoxy in the thin section. The epoxy bullet is rotated as needed in order to section the block, so that its shortest dimension intersects the cutting surface of the knife to minimize stress applied on the block. Final sections are made with a diamond sectioning knife with its boat filled with deionized water. The section thickness was initially set to 80 nm, and several sections were generated. Continuing ultramicrotomy at this thickness would result in >100 TEM specimens from the 10-micron total height of the sample block. Sections were transferred onto TEM grids with a carbon substrate film. (Not all specimens were retrieved from the boat due to section folding.)

Fig. 2.

Shape of the mesa. (a) Schematic of the epoxy mount before trimming (side view); (b) top view and side view of the ideal mesa shape after trimming; (c) schematic of the mesa before sectioning. Excess epoxy is present above the sample block and (d) optical reflected light image of the sectioned protective Pt layer embedded in epoxy. Pt curls up when sectioned into a thin section and leaves a window in the epoxy.

Results and Discussion

In comparing the conventional FIB and combined FIB-ultramicrotomy methods of TEM specimen preparation, the latter method provides the advantages of both FIB and ultramicrotomy. These include the FIB-based capability to (1) extract TEM specimens from specific sites in the bulk sample and the ultramicrotome-based capabilities to (2) create many TEM specimens from a small volume of sample, (3) make thinner TEM specimens of near-constant thickness, and (4) generate TEM specimens that are free from ion beam damage, re-deposition, or curtaining. In addition to the far greater number of TEM specimens from the combined FIB-ultramicrotomy method, it is important to note that they are obtained from approximately the same FIB trench milling footprint typical of a single section lift-out. In both the conventional FIB and combined FIB-ultramicrotomy methods, the ion milled area of the sample was about 20 μm × 40 μm. Because FIB is necessarily a sample greedy technique, in terms of material consumption, extracting many sections from the same volume greatly improves the “return on investment”, particularly for one-of-a-kind samples.

It is important to note that this procedure is only appropriate for those materials that are amenable to ultramicrotomy. These include many fine-grained and porous materials. Attempting to ultramicrotome ultrahard materials can damage the diamond knife used for generating ultramicrotomed sections, and large grains can cause sample pluck-out from the epoxy bullet and loss of the majority of the material to blade chatter. For such materials, or locations containing them in a heterogenous bulk sample, conventional FIB specimen preparation would be advisable.

At first glance, it appears that the combined FIB-ultramicrotomy method of TEM specimen preparation requires much longer time to complete than a conventional FIB section preparation. However, many tens of thin sections are produced by the combined method, whereas the conventional FIB method yields only a single section so that the specimen preparation time per TEM specimen can actually be shorter for the combined FIB-ultramicrotomy method depending on the block size and on specimen thickness. We also note that achieving high quality and quantity of TEM specimens is more important than minimizing specimen preparation time when working with one-of-a-kind samples. The total specimen preparation time required for the conventional FIB method was about 4 h, including careful 2 kV polishing. The combined FIB-ultramicrotomy method required about 2.5 h for lift-out and attaching to a Cu grid. Then, because multiple layers of epoxy had to be added and cured sequentially in order to fully cover the block-on-grid, three days (with overnight cures) were required for epoxy mounting, and about 12 h were spent in ultramicrotomy. Deeper blocks would provide more samples in approximately the same total sample preparation time, but the FIB trench milling footprint would also grow larger.

The dimensions of one combined FIB-ultramicrotomy TEM specimen (Fig. 3) were measured by TEM bright-field imaging to be ~18% smaller along the dimension parallel to the cutting direction than the original block surface dimensions measured by FIB-SEM. This is believed to be due primarily to folds in the thin section on the carbon substrate of the TEM grid. Other than the folds, the specimen showed uniform thickness across the entire section. Compression of the block face during ultramicrotomy also results in blade chatter, or horizontal cracks, that are aligned perpendicular to the sectioning direction, and chatter was observed throughout this section. Adjusting the orientation of the trapezoidal trimmed mesa of the epoxy bullet to present a corner (a smaller initial length to cut) of the epoxy—and also of the embedded block—to the diamond cutting blade can help to minimize blade chatter as well as using slower cutting speed.

Fig. 3.

Bright-field STEM images of TEM specimens prepared by (a) the combined FIB-ultramicrotomy method and (b) the conventional FIB method. The dark lines in the center (a) are due to overlapping material in wrinkles and the bright horizontal lines are due to chattering. The reduced contrast in the bottom center region of the FIB section (b) is due to over-thinning there. High magnification high angle annular dark-field (HAADF) images of the specimen prepared by (c) combined FIB-ultramicrotomy method and (d) the conventional FIB method. The difference in the clarity is evident at a higher magnification.

Bright-field STEM images were also obtained from a TEM specimen prepared by the conventional FIB method. This specimen shows curtaining, the effects of more rapid milling that occurs at pores and the interfaces of different phases present due to enhanced sputter rates (Fig. 3b). Curtaining is observed even though longer-than-typical scan dwell time was applied during 30 keV specimen thinning in order to reduce the curtaining effect (Fig. 3d). Re-deposition on the FIB thinned specimen was not significant on the body of the section, as evidenced by a lack of significant Pt detected there, but re-deposition (including Cu from the TEM half-grid) was observed at the bottom of the specimen (Fig. 3b). The thickness of the specimen is non-uniform, and the curtaining effect is more apparent as the specimen becomes thinner. The curtaining effect can be reduced by changing the direction of the ion beam, but it requires a special instrumental setting or manual rotation of the Cu grid (Ishitani et al., 2004; Montoya et al., 2007). Curtaining and sample bending are often the limiting factor in creating very thin (<80 nm) TEM specimens by conventional FIB methods, particularly for specimens prepared from heterogeneous samples that do not respond to the FIB ion beam uniformly. Ga was detected on the conventional FIB specimen from Ga ion implantation during ion sputter-milling and is especially prevalent in “soft” phases like organic carbon and glasses. We note that ion damage on the surfaces of conventional FIB-prepared TEM specimens is unavoidable, even if it does not show up in images. One advantage of the conventional FIB method over the combined FIB-ultramicrotomy method is that it can preserve the petrography, the position of every component, within a section, whereas microtomy can cause chattering such that portions of the section can be lost or displaced.

Conclusions

With approximately the same FIB milling footprint on the bulk sample as the conventional FIB method, the combined FIB-ultramicrotomy method produces many, uniform thickness TEM specimens that are free from ion beam-generated damage from specific sites in the bulk sample. In particular, curtaining, ion beam surface amorphization, re-deposition, and Ga ion implantation are eliminated with the combined FIB-ultramicrotomy method. This approach is ideal for TEM specimen preparation for fine-grained and porous materials, including those that contain amorphous materials. We demonstrate the benefits of the combined FIB-ultramicrotomy method over the conventional FIB method on the fine-grained matrix of the Paris meteorite, where it is especially advantageous since any unnecessary loss of material from this unique sample is to be avoided. Many steps in the procedure we describe are customizable for specific materials and applications (e.g., block size and shape), and this report should be considered a guideline that may be modified, as needed, to meet requirements for specific materials.