Читать книгу Encyclopedia of Glass Science, Technology, History, and Culture - Группа авторов - Страница 214

As a complement to the wealth of information on micro‐ and nanostructure that can be obtained from parallel‐illumination‐based TEM observations, one can also perform chemical analysis with modern machines, for example, by means of EDXS. In order to do so, sample areas as small as possible should be assessed directly with a finely focused probe to achieve the best lateral resolution possible. That is where scanning transmission electron microscopy (STEM) comes into play. In STEM, the sample is not illuminated with a broad parallel electron beam. Instead, the condenser lens system is used to form a fine electron beam that can be moved laterally over the sample surface with the help of deflection coils, in close analogy to what is done in an SEM. The difference is that SE emitted from the sample upon electron impingement are not detected. The primary electron beam passing through the very thin specimen of course remains used for imaging, but it loses energy through interactions with the sample atoms by either elastic (diffraction) or inelastic scattering. Of special interest for imaging are the inelastically scattered electrons, since the resulting contrast is highly sensitive to the mass of the atoms in the specimen. The scattering cross section of that electron–sample interaction approximately varies with the atomic number Z of the screened sample atoms according to Z ^a , where a is in the 1.2–1.8 range [22]. Thus, the heavier or denser the analyzed sample position, the larger the scattering angle of the primary electrons.

Usually the setup for catching the scattered primary electrons consists of a ring‐shaped detector below the sample (Figure 5). In this (high‐angle) annular dark‐field [(HA)ADF] imaging mode, the larger the scattering angle of the primary electrons, the more signal will be recorded on the detector. Thus, heavier elements in the sample generally appear as brighter features in the image. The camera length is the distance between the sample and the detector plane. Because one can change it easily by adjusting the electromagnetic lenses of the TEM, it is possible either to induce only Z‐contrast in the image (short camera length) or to use a fraction of the elastically scattered (diffracted) electrons to gain signal on the detector (long camera length), thus incorporating also information concerning the crystallinity of the sample in the micrograph.

Figure 5 Working principle of the scanning transmission electron microscope.

The MAS glass‐ceramic sample may also serve to compare the images formed with TEM and STEM. In a TEM image in bright‐field mode (Figure 6a), the zirconia crystals appear dark in a bright glass matrix. This contrast is caused by both mass thickness effects (i.e. stronger inelastic scattering of primary electrons and absorption of primary electrons by zirconia) and diffraction effects (elastic scattering of primary electrons by the zirconia lattice planes). On the contrary, the STEM‐HAADF micrograph in Figure 6b shows bright zirconia crystals in a dark matrix because the signal collected by the annular HAADF detector from the electron beam scattered at large angles is much more intense for the heavy Zr atoms than for the relatively light Mg, Al, Si, and O atoms of the glass.

A big advantage of STEM analysis is of course that one can use the primary electron beam for other purposes than just imaging of the sample. With nanobeam diffraction, one can address areas as small as a few nanometers to generate diffractions patterns from which crystallographic structures can be analyzed in great detail, with a probe diameter lower than 5 nm, that can be moved over the sample to study specific areas with nm resolution. Just as in SEM, it is also possible to detect and analyze secondary signals emerging from the electron beam–specimen interaction, such as X‐ray quanta or visible light (CL). Especially EDXS in STEM is an extremely helpful tool for a quick determination of the distribution of chemical elements in a sample at a scale that can be truly nanometric because the sample is so thin that the excitation volume is not subjected to lateral resolution broadening: for a TEM sample thinner than 50–70 nm, the excitation volume is now just the very neck of the interaction volume shown in Figure 1. In modern (S)TEM machines, EDX detectors may cover a large solid angle (>1 steradian) above the sample, thus ensuring fast elemental analysis [23].

Figure 6 Comparisons of TEM and STEM images of ZrO₂ crystals in the glassy MAS matrix. (a) TEM bright‐field and (b) STEM‐HAADF micrographs.

With these improvements, most of the problems faced with SEM for the chemical analyses of the MAS glass ceramics are readily overcome. First, there is no need to worry about excitation volume and lateral resolution of the EDX analysis since the sample is very thin. Second, the generally higher resolution of TEM systems enables finer details to be seen and analyzed. Third, the increased energy of the primary electron beam (here, a 300 kV acceleration voltage) allows high‐energy spectral regions far beyond 10 keV to be accessed so that the Y‐K_α (at ≈14.9 keV) and Zr‐K_α (at ≈15.7 keV) lines can now be unambiguously separated.

One might wonder why 300 kV acceleration voltage in TEM does not cause the same charging problems as 30 kV in SEM. The reason is that in a thinner sample the interaction cross sections of a very‐high‐energy electron beam with matter are getting small enough that a majority of electrons can pass through without any interaction. However, this does not mean that radiation damage is not at all an issue in TEM. Although the process has a low probability, an electron may directly hit an atom or ion and displace it in what is called a knock‐on damage. Besides, one finds that glasses and glass ceramics are prone to damages caused by the electron beam, such as motion or even evaporation of Na, Li, or other volatile elements [24], amorphization, or, conversely, crystallization, to name just a few. The experimental conditions to be used (acceleration voltage, maximum beam dose, etc.) always depend on the particular sample investigated and often demand a great deal of expertise and experience. Whereas MAS glasses and glass ceramics are quite robust, lithium aluminosilicate (LAS) materials, in contrast, survive a TEM experiment only if irradiated with electrons of reduced energies (typically ≈80 kV) and then might still deteriorate readily if exposed even to moderate electron beam doses [25]. To find suitable conditions, it is thus important that large sample areas be available for testing and that settings can be adjusted and saved in the microscope computer.

An example for the use of STEM‐EDXS with high lateral resolution is shown for the MAS glass ceramics in Figure 7. The distributions of selected elements are indicated in false color as evaluated from the intensity of the relevant X‐ray K‐lines in the pixelwise EDX spectra recorded when the beam was rastered over the sample. Even at comparably low magnification (upper panel of Figure 7), these elemental maps clearly show that spinel (MgAl₂O₄) is growing on dendritic zirconia as secondary phase (Al can be seen as a “marker” for MgAl₂O₄ here). Furthermore, it is obvious that Y is pushed away from the growing spinels into the residual glassy areas. But other features are also apparent, such as the existence of a silicon‐enriched shell between the spinels and the residual glass. At higher magnification (lower panel), additional stunning details are seen, e.g. the dendritic structures formed by the growing spinels when they expand into the surrounding matrix.

Figure 7 Micrographs of element distribution as determined for the areas shown on the left panels by STEM with a 300 kV acceleration voltage at low (upper panel) and high magnification (lower panel) from the indicated emission lines. The Zr, Al, and Y distributions map the zirconia, spinel, and glass phases, respectively. A dendritic spinel is shown in the lower panels, embedded into a silicon‐rich matrix that isolates it from the actual yttrium‐rich glass matrix.

Table 1 Composition of the MAS glass (at %),^a where “pristine” denotes the homogeneous, non‐annealed sample and “annealed” denotes the residual glass areas in the glass ceramics after annealing.

	Mg	Al	Si	Zr	Y
Nominal	16.5	33.9	41.4	4.2	4.0
EDXS pristine	16.8	33.7	41.1	4.5	3.9
EDXS annealed	11.8	17.2	43.1	2.7	25.2

^a Oxygen not considered because of self‐absorption of the low‐energy O‐K X‐rays.

In view of the small excitation volume in TEM experiments, it is rather straightforward to compute the relative intensity ratios of the element‐specific peaks in an EDX spectra and thus to get a quantitative element distribution of tiny areas within a sample once the system is properly calibrated. In this way, crystallization‐induced changes in the composition of the residual glass can, for instance, be investigated, even if the lateral dimensions of the residual glassy areas are a few nm³ only. As shown by Table 1, the STEM‐EDX calculated compositions of the pristine green MAS glass are in excellent agreement with the nominal data, whereas important changes are observed after heat treatment for Zr and especially for Y.