Search

Fresnel fringes are a common feature in CTEM defocused images of specimen edges. The detailed contrast is actually governed by the exact geometry of the termination of the edge. Calculations by Fert and Laffite (1) and by Fukushima et al (2) have shown that the intensity profile is very sensitive to the existence of a region of thickness gradient. Joy et al (3) have clearly interpreted these contrast variations as a consequence of the refraction of the electrons in this zone.In order to determine how far these effects can be used to evaluate physical parameters of an edge, we have computed images of thin electrostatic or magnetic edges to fit them with experimental observations. An optical model with a pure phase or a mixed phase^and amplitude object is used. The structure of the wave function a on the exit surface of the specimen forms the input signal for the subsequent electron optical imaging system which involves spherical and defocusing aberrations.

Energy filtered imaging of thick biological specimens was analysed using a dedicated STEM fitted with an energy loss spectrometer and interfaced with a sophisticated data collection setup. All images were digital, thus permitting a quantitative analysis of the data. We also present a mathematical explanation of the data, which is useful in pradicting the quality of thick specimen images formed with energy filtered electrons. It is known that increasing specimen thickness leads to a decrease of the zero energy loss intensity and an increase in higher (multiply scattered) energy loss electrons. We show that contrast decreases gradually with increased energy loss but, most important, the signal to noise ratio is maximal at an energy loss position slightly below the intensity maximum. This is the optimal position for imaging thick specimens. Moreover our studies confirm that the following parameters have similar effects on the energy loss spectra: (1) increased thickness (t); (2) higher average Z number elements (or lower mean free path Ai); and (3) lower primary voltage (V0).

In analytical problems, EELS is a powerful technique to locate, identify and measure the atoms contained within volumes of very small size. The present study is an experimental investigation of its resolution limits on a test specimen made of small metallic Tb and U clusters down to the atomic size and lying on a thin amorphous carbon layer. Several parameters have to be determined : the spatial resolution power (d), the minimum detectable concentration (Co) and the ultimate number of detectable atoms N = Co d2 tn within a specimen of local thickness t. The used instrument is a dedicated FEG-STEM VG HB 501 equipped with a Gatan spectrometer and governed by a home made digital acquisition system. This equipment offers unique capabilities for complete data quantification : the simultaneous acquisition for each pixel of two information channels (the annular dark field (adf) and the bright field one after the spectrometer) allows to discriminate the primary probe contribution from the inelastic effect. Sequences of up to 10 energy filtered images are acquired for different loss values covering the EELS range of interest from below the U-O45 edge (at 105 eV) to above the Tb-N4B edge (at 155 eV).

Contrast mechanisms, which are responsible of the various types of image formation, are generally thickness dependant. In the following, two imaging modes in the 100 kV CTEM are described : they are highly sensitive to thickness variations and can be used for quantitative estimations of step heights.Detailed calculations (1) of the bright-field intensity have been carried out in the 3 (or 2N+l)-beam symmetric case. They show that in given conditions, the two important symmetric Bloch waves interfere most strongly at a critical thickness for which they have equal emergent amplitudes (the more excited wave at the entrance surface is also the more absorbed). The transmitted intensity I for a Nd2O3 specimen has been calculated as a function of thickness t. The capacity of the method to detect a step and measure its height can be more clearly deduced from a plot of dl/Idt as shown in fig. 1.