We here present the results of the first materials science analyses obtained with the prototype of a serial block-face sectioning and imaging tool, 3View™ of Gatan, Inc (Pleasanton, CA, U.S.A.). It is a specially designed ultramicrotome operating in situ within an environmental scanning electron microscope originally developed for life science research. The microtome removes thin slices from the sample and the environmental scanning electron microscope images each new block surface of the specimen (serial block-face scanning electron microscopy). The Schottky emitter (FEG) of the microscope delivers high spatial resolution and has the advantage of stable performance and high durability. The slice thickness can typically be selected between 50 and 100 nm. It is possible to cut hundreds of slices and simultaneously acquire images with Digital Micrograph™ Model 700 (Gatan, Inc.). This article outlines the set-up and describes the automated process. The preparation of specimens for in situ ultramicrotomy is explained and the parameters for good image quality are discussed. In addition, special operative and analytic features of the controlling software are presented. Three different technical materials and one botanical specimen were analyzed delivering first results of this method for materials science and for botany.
The two-dimensionality of the information obtained by microscopy has always been a limiting factor in materials science and life sciences. The recording of three-dimensional (3D) spatial structures of specimens is an important issue and can be necessary in understanding microscopic objects and their functions (Denk & Horstmann, 2004).
In electron microscopy, the spatial structure of soft materials is usually elucidated by serial sectioning with an ultramicrotome and imaging of the individual sections in the transmission electron microscope (TEM) (Rakic et al., 1974; Harris et al., 2006). This is very tedious and time-consuming work, which cannot be fully automated. Finally, all the images have to be aligned correctly to generate the 3D stack of images, which can be both cumbersome and a source for errors (e.g. distortion).
A promising new method to obtain this information at the EM level with much less effort is automated serial sectioning of specimens with an ultramicrotome integrated into an environmental scanning electron microscope (ESEM). This method, known as serial block-face scanning electron microscopy (SBFSEM), involves imaging the surface of the specimen after the removal of each slice. It was developed by W. Denk in 2004 (Denk & Horstmann, 2004) and subsequently further automated by Gatan, Inc. (3View™).
The main reason for the choice of an ESEM is to avoid specimen charging. Even a variable-pressure SEM (VPSEM) would fulfil this requirement (Goldstein et al., 2003).
A combination of a microtome and a conventional scanning electron microscope (SEM) was already developed by Leighton in 1981 (Leighton, 1981). Here the possibilities were limited since neither a VPSEM nor an ESEM was used and computer capabilities were in their infancy (Denk & Horstmann, 2004).
The principle of this new in situ method consists in cutting a specimen with a diamond knife (Diatome, Switzerland) and taking an image of the fresh block face. The sequence of slicing and imaging is automated and has the additional advantage that all the images are correctly aligned and are free of distortion. The slices themselves cannot be collected and are not used.
Contrary to a conventional SEM, the ESEM can also be operated at chamber pressures between 10 and 2700 Pa. At this pressure range, electrically insulating specimens can be imaged at electron energies typically up to 30 keV without a conductive coating (Goldstein et al., 2003; Stokes, 2006).
Until now, applications have only been described for neuroscience applications (Denk & Horstmann, 2004; Macke et al., 2008). In this paper, we describe the technique for materials science for the first time and present new applications.
This work was carried out on an ESEM Quanta 600 FEG equipped with a Schottky emitter (FEI, Eindhoven, The Netherlands). The 3View™in situ ultramicrotome from Gatan is located directly beneath the electron column and mounted on a specially designed microscope stage that is attached to the microscope door.
The specimen must first be pre-cut using a standard ultramicrotome and is subsequently glued onto a rivet that is mounted to a removable specimen holder of 10 mm diameter. This holder is then mounted on the specimen stage of the ultramicrotome. The specimen is moved vertically by a high-precision motor to define the slice thickness. The knife movement defines the imaging plane, which is about 11 mm (working distance) below the pole piece. The diamond knife cuts and retracts driven by a piezo-actuator. It is a specially designed diamond knife from Diatome (knife angle 45°, clearance angle 11°, cutting edge length 1.5 mm).
The specimen must be trimmed to a cuboid with a surface of 0.25 mm2 and a maximum thickness of 500 μm. The travel range of the vertical motor is 700 μm and determines the maximum number of slices that can be cut. Typically, the slice thickness can be set from 50 to 100 nm (even cuts with a thickness of 30 nm were performed with the present system); thus a stack can consist of hundreds of slices. The duration of one sectioning and imaging cycle is mainly determined by the image acquisition time (e.g. 30 μs per pixel at an image size of 1000 × 1000 would take 30 s). The cutting speed and the knife retraction speed can be set typically between 0.1 and 0.6 mm s−1.
Usually the cut-off slices pile up on the top surface of the diamond knife. It rarely occurs that debris remains on the block face. In this case, the respective image can be reconstructed by an interpolation between the adjacent images using an easy-to-use feature of the program Digital Micrograph™ (centrepiece of Model 700. P2000.1, Gatan Microscopy Suite, Gatan, Inc.). In the present system, a technical mode to remove debris is not implemented.
Two types of detectors (among others) can operate in the low vacuum mode of the ESEM: the large field detector (LFD), which primarily detects secondary electrons, and the solid-state detector (SSD), which is sensitive to back-scattered electrons. Detection of the back-scattered electrons is a conventional imaging method for serial sectioning (Denk & Horstmann, 2004), since the use of stains (such as osmium tetroxide, ruthenium tetroxide) leads to high material contrast (Sawyer & Grubb, 1996; Michler et al., 2004) in biological specimens, for which this method has been developed.
The LFD (Fig. 1) represents a special architecture of the gaseous secondary electron detector (GSED) operating in the low vacuum mode (pressure range nominally around 10–300 Pa) (Goldstein et al., 2003; Stokes, 2006). Since this detector has no pressure-limiting aperture (as in the conventional GSED), the field of view is not restricted. This detector is advantageous in cases where topography contrast rather than phase contrast is of interest, for example, the size, orientation and distribution of pores or cracks in a material.
In addition, the SE mode can be useful for finding the area of interest on the sample and for alignment purposes. A primary electron energy of 7 keV was used in all investigations, with the exception of the paper specimen, which was imaged at an electron energy of 10 keV. Here, we used at least 7 keV to get a good signal-to-noise ratio (SNR) with the BSE-detector available at our microscope. In fact the resolution can be enhanced using smaller voltages as discussed in Denk and Horstmann (2004).
In all cases, imaging was performed in the low vacuum mode with water vapour as gas.
Digital Micrograph™ was used for microscope control, automated serial sectioning, data acquisition and two-dimensional visualization.
The DigiScan™ software is a plug-in module for Digital Micrograph™ that allows operation of the Model 788 DigiScan™ II controller (Gatan, Inc.) with most current SEM or scanning transmission electron microscopy (STEM) units. The DigiScan™ hardware and software system allows images to be captured, displayed and processed at a maximum picture size of 8192 × 8192 pixels, which results in a large amount of data (e.g. 8 GB for a stack of 1000 slices with a pixel resolution of 2000 × 2000 and a 16-bit intensity depth).
The microscope was controlled using the scripting interface xTLib from FEI, with the scripts integrated in the Digital Micrograph™ program. Thus, all functions for routine operation of the microscope can be controlled by the Digital Micrograph™ environment without the necessity of resorting to the user interface of the microscope. Digital Micrograph™ provides several additional imaging modes and tools for data analysis.
Specimen preparation and positioning
The bulk specimens for in situ ultramicrotomy were trimmed to a cuboid (Leica EM TRIM, Leica Mikrosysteme GmbH, Vienna, Austria) with a cross section of about 0.5 × 0.5 mm2 and subsequently cut to a cuboid of about 0.5 × 0.5 × 0.5 mm3 in size. This cuboid was then glued onto a rivet mounted on a removable specimen holder. The rivet was subsequently mounted on the in situ ultramicrotome. An optical microscope with a built-in CCD camera attached to the open door of the ESEM was used for specimen alignment. The position of the specimen relative to the diamond knife was adjusted manually, and the final vertical adjustment was done using the motor drive.
Generally the preparation conditions are the same as in the field of TEM investigations with the fundamental difference that SBFSEM images the block face of a specimen. Thus, SBFSEM does not allow contrast enhancement after cutting (Denk & Horstmann, 2004).
Polymer blend after tensile test
A tensile test was performed with a polymer blend (isotactic polypropylene/ethylene propylene rubber particles). Part of the fracture zone (Poelt et al., 2006; Zankel et al., 2007) was embedded in resin (Epofix™; Struers A/S, Ballerup, Denmark) and subsequently stained with ruthenium tetroxide. Finally, it was cut to a cuboid of the above size and fixed on a rivet.
Composite material ABS/pine
Composite materials consisting of layers of wood and polymer were fabricated in a special injection moulding tool (Blaschke, 2005). Of special interest was the bonding at the interface of the two materials. The investigations involved cutting the composite material to the appropriate dimensions using a trimmer and an ultramicrotome. The interface between the polymer (ABS) and the wood (pine) was oriented orthogonally to the surface of the rivet and fixed with superglue.
To obtain information about the distribution of filler particles and fibres in a paper, a small portion was embedded in resin (Epofix™) and then cut to a cuboid in the manner described earlier and glued onto a rivet. To image the cross sections, the surface of the paper was oriented orthogonally to the surface of the rivet.
A young geranium leaf (P. zonale) was primarily fixed with glutaraldehyde (2.5%, pH 7.0) and post-fixed with 1% OsO4. After dehydration, the sample was embedded in resin (Agar 100 resin, Agar Scientific Ltd., Stansted, U.K.). Finally, it was cut to a cuboid of the appropriate size and placed on the ultramicrotome.
Polymer blend after tensile test
The 3D analysis of the fracture behaviour is an important issue in materials science (Park et al., 2004). For example, serial sectioning is a method of choice for revealing the 3D field of micro-cracks in the bulk of a polymer specimen after a tensile test (Zankel et al., 2007). Special filler particles are often added, especially to polymers, to stop the propagation of cracks forming under stress and thus prevent fracturing of the material by cracks propagating through the whole material (Michler, 1992). Serial sectioning might provide a much better insight into the underlying mechanisms.
A tensile test was performed on a Charpy V-notched bar of an iPP/EPR blend pre-cracked with a razor blade and stopped after reaching 25% of the yield stress. Subsequently, 160 sections with a thickness of 100 nm each were cut in part of the fractured region using SBFSEM.
Imaging was performed at a pressure of 40 Pa using a pixel dwell time of 8 μs. A pixel size of 50 × 50 nm2 was chosen.
Figure 2(a) shows the micrograph of the first slice of the stack. The EPR particles appear brighter than the matrix because the ruthenium of the ruthenium tetroxide stain is mainly incorporated in the rubber of the EPR particles.
The pre-crack with the adjacent field of cracks is visible at the top of the micrograph. Three different types of cracks can be discerned: a bundle of cracks through the matrix leading to the well-known phenomenon of crazing in (α) (Michler, 1992), cracked filler particles in (β) and the debonding of a filler particle from the matrix at the end of a bundle of cracks in (γ).
Digital Micrograph allows to define lines as regions of interest (ROIs). Within these ROIs, the depth of the stack can be analyzed (z projection).
The selected ROIs are outlined in Fig. 2(a). The depth profiles in Figs 2(b) and (c) provide information about the development of the cracks in the ROIs.
It was observed that the cutting quality was higher in the area that had been irradiated by electrons during imaging (Denk & Horstmann, 2004). Figure 3 shows a micrograph at a lower magnification following the cutting of the block face of the polymer sample after serial sectioning and imaging (160 sections). Two regions are outlined: region 1 was subjected to electron irradiation as a consequence of section imaging, whereas in region 2, no electron irradiation occurred. It should also be noted that the slice thickness (100 nm) is much smaller than the penetration depth of the electrons (1.6 μm in carbon at 7 keV). Thus, with the exception of the first few slices, every slice is subjected to electron irradiation during imaging of many layers lying on top of it. Since the amount of staining decreases with the depth of this particular specimen, the brittleness of the block faces also decreases. As a result, the cutting quality deteriorates with the depth of the specimen. This can be seen in region 2, which shows more blurred structures, probably because of smearing effects, although the amount of stain inside the scan field (region 1) is the same as outside. This points to a hardening effect caused by the electron irradiation (Kitching & Donald, 1998; Royall et al., 2001; Denk & Horstmann, 2004).
Composite material ABS/pine
The conventional imaging method in SBFSEM is to use a back-scattered electron detector to obtain material contrast of, for example, stained specimens (Denk & Horstmann, 2004). But in materials science, the contrast mechanism of an SE detector might be used as well. This has the advantage that much smaller probe currents can be used to minimize irradiation damage. To demonstrate this new detection method, we used a composite material consisting of ABS and pine. A stack of 25 cuts (slice thickness 200 nm) was successfully imaged using the secondary electron detector. Imaging was performed at a pressure of 50 Pa using a pixel dwell time of 5 μs (pixel size 120 × 120 nm2).
Besides the different contrast of wood and polymer, the higher surface roughness of the sectioned wood may also contribute to secondary electron contrast.
The interface between the polymer and the wood can be seen in Fig. 4(a). The marked area shows a defined ROI, for which a 3D visualization was created using the software AMIRA 3.1™ (Fig. 4(b)). The polymer and the wood seem to bind tightly together; no cracks are visible. One can also perceive that the wood in the region close to the interface has been damaged by the manufacturing process.
In-depth knowledge of the spatial micro-structure of papers provides the basis for fundamental research on the interrelations between paper quality and the paper-making process. Many physical properties of paper are strongly influenced by the spatial distribution of its raw materials, such as fibres, fibre fragments or filler particles. Other important parameters are the homogeneity and thickness of the coating layer (Donoser et al., 2005). SBFSEM can provide insight into the internal structure of paper specimens and deliver all the necessary information.
A series of 100 slices, each with an area of 60 × 60 μm2 and a thickness of 200 nm, was cut from a paper specimen. Imaging was performed at a pressure of 65 Pa using a pixel dwell time of 20 μs (pixel size 60 × 60 nm2).
Imaging was carried out using the electron back-scatter detector. Several phases of the specimen can be distinguished in Fig. 5. The brightest phase (1) is formed by the filler particles (generally calcium carbonate, kaolin, quartz) both in the coat and the paper itself, whereas the medium-grey region represents the cellulose fibres (2). The darkest region is the resin used for embedding the specimen (3).
These intensity ranges were used to separate the three phases by setting appropriate thresholds in the 3D visualization software AMIRA 3.1™. The 3D structures of the phases were then calculated by triangulation. The results are shown in Fig. 6. They reveal a strong variation in the thickness of the coating and also the presence of cavities in the coat. Although this is already indicated in the image of a single layer in Fig. 5, the 3D structures in Fig. 6 provide a much better and more reliable assessment not only of the thickness variations and the homogeneity of the coating but also of the distribution of the filler particles in general.
The specific volume and surface fraction of each phase can also be calculated from the 3D structures (Table 1). The error of the calculated values, which is mainly determined by inaccuracies in the thresholding, is estimated to be approximately 3%.
Table 1. Volume and surface fraction of each phase of Fig. 7 calculated with AMIRA 3.1™.
The much higher lateral resolution provided by the ESEM (by contrast to light microscopy) allows the detailed investigation of the distribution of filler particles, which are generally of a sub-micrometre size.
Different types of glandular and non-glandular trichomes can be found on the epidermis of geranium leaves. This investigation was performed on the head cell of a long-stalked glandular trichome.
The head cell was imaged with the back-scattered electron detector. A stack of 547 slices with a slice thickness of 50 nm was produced.
Imaging was performed at a pressure of 50 Pa using a pixel dwell time of 20 μs (pixel size 50 × 50 nm2).
The 3D visualization was carried out using the program AMIRA 3.1™ (Fig. 7). A small part of a cell forming stalk can be seen on the right side of the image. This cell is characterized by a large central vacuole, and it is separated from the head cell by a clearly visible cell wall. The view into the head cell shows a large cell nucleus with the nucleolus in the centre. A large amount of organelles and small vacuoles are embedded in the cytoplasm of the metabolically very active cell.
3D reconstruction based on serial sectioning by ultramicrotomy in the ESEM is much easier and far less time-consuming than using a standard microtome and a TEM for imaging. But there is one profound difference between the two methods: TEM analyses image the slice itself and the information in the image thus originates from this slice only. SBFSEM removes a slice and then images the surface of the remaining bulk material. As the information depth of the electrons is generally greater than the slice thickness (e.g. 200 nm at an electron energy of 4 keV and a density of 1000 kg m−3, as is typical for polymers), one image contains a convolution of the information of several slices depending on the primary electron energy. Therefore, using slice thicknesses smaller than 50 nm would not necessarily improve the vertical resolution. The resolution can thus differ substantially in the block face and in the direction perpendicular to it. An increase in vertical resolution can be realized by a decrease in electron energy (Denk & Horstmann, 2004), but an ESEM requires a minimum electron energy to obtain a good SNR depending on the pressure of the imaging gas and the working distance used. By the nature of the technique, the minimum spatial resolution is specimen-specific. Thus, a minimum lateral resolution of about 10–30 nm can be assumed, and the slice thickness determines the z resolution. With this system, a minimum thickness of 30 nm can be achieved.
Irradiation damage is a fundamental topic in the investigation of soft materials like textile fibres and polymers in an SEM (Sawyer & Grubb, 1996). In the ESEM additionally parameters like the type of imaging gas, the gas pressure and the production of free radicals, especially in water-containing specimens, have to be taken into account (Kitching & Donald, 1998; Royall et al., 2001).
However, the problem can generally be solved by using the appropriate sample preparation method (e.g. staining with ruthenium tetroxide) and by adjusting the imaging parameters. It was observed that although beam damage may occur for the first few slices (e.g. 10), the cutting process leads to stable material properties with minimum beam damage because the surface is renewed for each slice. It was also observed that the influence of the beam can enhance the cut properties of the material (Denk & Horstmann, 2004). Investigations of the changes of the chemical structure of the specimens caused by electron irradiation using FTIR (Kitching & Donald, 1998; Zankel et al., 2008) are pending.
One of the advantages of the presented method is its capability of investigating larger volumes than is possible using other 3D techniques of similar lateral resolution. Focussed ion beam tomography, for example, can analyze volumes of typically up to 50 × 50 × 50 μm3 (Möbus & Inkson, 2007; Xu et al., 2007), whereas the SBFSEM method enables 3D reconstruction of volumes up to 500 × 500 × 500 μm3. It should be noted, however, that the 3D FIB techniques are somewhat complementary to the SBFSEM with regard to materials that can be easily sectioned. An ultramicrotome could be especially suitable for soft materials or biological specimens that pose difficulties in FIB sectioning. By contrast, hard and brittle materials, which are easy to cut with an FIB, still cause problems for microtome cutting. However, methods to overcome these problems are currently under development. Another difference to the FIB techniques is that SEM image acquisition is directly performed with a beam perpendicular to the freshly produced block face, whereas FIB imaging either involves additional sample tilting or is done with an electron beam under a certain angle with respect to the surface normal.
By contrast, the high resolution and high magnification capabilities of SBFSEM give it a special significance compared with light microscopy methods. A comparable method has been described for the analysis of paper samples (Donoser et al., 2005), where larger surfaces (millimetre-range) can be investigated. This method is limited to the resolution limits of a light microscope. However, these two methods can be regarded as complementary depending on the required scope of analysis.
The presented technique of in situ ultramicrotomy in an ESEM is a unique method combining back-scattered electron and secondary electron imaging. Secondary electrons can be used when the internal structure of the sectioned specimen delivers sufficient topography contrast as exemplified by the wood and polymer composite sample presented in this paper. Secondary electron imaging provides the opportunity to enhance the resolution, since these electrons are generated in a smaller layer of the sample. Thus, several properties of a modern ESEM can be used for advanced specimen analysis. This is supported by the user-friendly software Digital Micrograph™, which offers different imaging modes (e.g. signal mixing) and image-processing tools (e.g. FFT, drift correction).
The serial block-face ultramicrotome 3View™ of Gatan, Inc., operated in an environmental scanning electron microscope, proved to be a very useful tool for the investigation of the internal structure of both technical and biological materials.
Up to now, the method has been used exclusively for investigating soft technical materials because both the preparation method and the problems are similar to those of biological samples, for which this ultramicrotome was initially developed. The investigation of hard materials is in progress.
Several imaging modes were employed, and imaging with secondary electrons also proved useful in some cases. The back-scattered electron micrographs were taken with a solid-state detector that provides optimal performance at high electron energies, long frame times and a working distance of 8.5 mm (FEI Company, 2003). A new BSE detector from Gatan, Inc., that has a higher signal-to-noise ratio even at fast scanning rates will be tested in future experiments. This detector ensures good imaging quality even at low voltages and short frame times, thus both reducing the information depth of the signals and minimizing possible irradiation damage. In addition, we are interested in developing preparation methods that eliminate the need for gluing the specimens.
The authors are deeply grateful to Prof. Dr. Ferdinand Hofer for enabling the cooperation between our institute and the Gatan Company.
We thank Dr. Markus Gahleitner from the Borealis Group for providing the polymer specimen and Dr. Edith Stabentheiner from the Institute of Botany of the University of Graz for providing the botanical specimen. We also thank M. Wallner and G. Stoiser for graphical support and Manuel Paller for modifying the schematic diagram of the instrument provided by Gatan, Inc.