this will, for instance, apply to the Ti Kα1 and Ba Lα1 at about 4.51 and 4.47 keV, respectively. This problem is less serious with WDXS (with a spectral resolution of ≈10 eV), where Bragg’s law of selective reflection forms the basis of detection in a proportional counter. Whereas EDXS is a parallel‐registering technique, WDXS (which is very similar to electron microprobe analysis [EMPA]) is a serial technique, making it more time consuming and less commonly used than EDXS.
For a sufficient yield of X‐rays of a certain energy, as a rule of thumb, the primary electrons should have a kinetic energy of at least twice that amount. As an example, Si‐Kα X‐rays possess an energy of 1.74 keV, which is similar to the difference between the Si‐K and the Si‐L2,3 energy levels in the electron shell of the Si atom. Thus, one needs to use at least 3.5 kV acceleration voltage to be able to detect Si in a glassy sample by means of EDXS or WDXS, let alone any heavier elements. Low‐voltage excitation in SEM might thus be advisable for imaging purposes, but not for chemical analysis.
2.3 Application to Glass Ceramics
To illustrate the information depth one can gain with that technique, the aforementioned MAS glass ceramics will be selected. In this sample, precipitation of zirconia and spinel is known from X‐ray diffraction (XRD) experiments, but this technique is inappropriate for investigating the micro‐ and nanostructure because it gives information about the sample as a whole, with a low spatial resolution defined by the X‐ray spot size. Techniques such as SEM or TEM must be used instead, and they can in addition yield the composition of the residual glass and the structure of the crystals – in this case whether zirconia is the tetragonal, cubic, or monoclinic polymorph.
For a piece of this MAS sample, polished and rendered conductive by deposition of a ≈3–5 nm carbon film on its surface, an SEM micrograph (SE) taken with an acceleration voltage of 10 kV clearly shows bright, elongated structures and areas with brighter and darker appearance (Figure 2a). A comparison with an SEM micrograph acquired at a 30 kV acceleration voltage (Figure 2b) illustrates the loss of both resolution and contrast when the energy of the primary electrons increases and results in deeper electron penetration in the bulk. In turn, the effect makes the origin of the image‐forming SE less predictable, so fine details are smeared out at the surface of the sample. At a higher magnification (Figure 2c), the microstructure observed with a 10 kV acceleration voltage from backscattered primary electrons (BSE) shows the dendritic appearance of the brightest, elongated features that indicate their crystalline nature. Resulting from strong scattering of the primary electrons, the marked contrast of these features with their surrounding suggests the presence of heavy elements (and associated high backscattering yield). Given that zirconia is present in the sample and that the other heavy element besides zirconium, i.e. Y, is not expected to enter another crystalline phase in this material, one can conclude that the bright dendritic microstructure is likely made up of zirconia. Around the dendrites, the brighter and darker areas should then be the images of the spinel and the residual glass, respectively.
Figure 2 Scanning electron micrographs of the MAS glass‐ceramic sample: resolution differences between 10 kV (a) and 30 kV (b) acceleration voltages and BSE micrograph taken at 10 kV at higher magnification (c) showing the bright (1) and dark (2) areas for which the EDX spectra of (d) were acquired.
To check this conclusion, EDXS analyses were made in the two areas indicated in Figure 2c. The spectra are presented in Figure 2d. In both, a small carbon peak is an artifact of the sample preparation. Besides the most intense oxygen and silicon peaks, the other peaks are readily assigned to the other elements, but a clear separation of the Y‐Lβ and Zr‐Lα peaks is not possible owing to the limited energy resolution of the EDX detector system. The better separated Y‐K (at ≈14.9 keV) and Zr‐K lines (at ≈15.7 keV) would be more suitable, but a higher acceleration voltage would be needed to access that higher‐energy part of the EDX spectrum with its ensuing loss of image resolution.
Nevertheless, these EDX spectra suggest that the brighter areas likely are silica rich, most probably depicting the residual Y‐bearing glass. Conversely, the darker areas should consist of spinel, which is Si‐ and Y‐free, as confirmed by analyses made in the “darker” area of the sample where O, Al, and Mg are found with a higher Al/Si ratio than in the other spectrum. The reason why certain amounts of Si and also of Zr (but not of Y) are nonetheless detected is that a certain interaction volume is also excited below the surface within the sample (cf. Figure 1). As a result, the lateral resolution is worsened, and the probed area does not include only the “dark” region 1 that is visible in Figure 2c, yet also parts of the sample bulk below. In conclusion, like any technique, SEM and EDXS have their own limitations. Other methods having a higher resolving power must be used to overcome them.
3 Transmission Electron Microscopy
3.1 Conventional Observations
Analytical (scanning) transmission electron microscopy [(S)TEM], currently is the best method for both imaging and determining chemical compositions [18] whenever the relatively large excitation volume of SEM becomes problematic. The price to be paid is the necessity to prepare samples as thin as a few 10 nm, without preparation artifacts, to make them transparent to electrons. The task is difficult, but over the years a wide range of techniques have been developed for this purpose, including mechanical grinding, ion beam etching, preparation of replica, ultramicrotomy, and FIB machining of lift‐out TEM lamellae [19]. In this respect one should distinguish between samples taken from the bulk and cross‐sectional samples cut out from specifically targeted areas (e.g. if a specific source of devitrification in a glass has to be assessed for further analysis) for which FIB machining is particularly well suited.
In its principle, a transmission electron microscope is amazingly similar to a light microscope: the electron source corresponds to the lamp; beams are formed by electromagnetic lenses (which are true zoom lenses as one can finely change their focal length by altering the lens current); image and object planes are defined; and after converting the electron distribution into light with a scintillator, an image is formed and registered by a CCD or CMOS camera. Of course, however, TEM offers a number of tremendous advantages over light microscopy that result from the much lower wavelengths of electrons compared with those of visible light. First and foremost, it has a much higher spatial resolution, as the best commercially available instruments can now resolve lattice planes as close as 50 pm apart [20]. Second, unlike for optical lenses, the focal lengths of electron lenses can be changed during operation, allowing for switching between parallel illumination and formation of a focus spot of less than 100 pm in diameter for dedicated analyses of very small sample regions. Third, one can switch between imaging and diffraction modes, which makes it possible to identify crystalline phases from the recorded electron diffraction patterns. Fourth, the higher energy of electrons (as compared with visible light) allows electronic transitions to be excited, laying the ground not only for analyzing chemical compositions (EDXS) but also for assessing the coordination and valence of
