this method opens up a new way to discuss the local structure and material properties, giving a better understanding of phase transition or magnetic domain wall pinning in strongly correlated materials. (iv) The stability of probe position, where the mechanical vibration of the floor from any adjacent disturbance should be reduced, and the specimen drift during the acquisition should be measured under the ADF imaging mode, the collection of semi-angle for EELS was as large as 31 mrad to reduce the extent of delocalization. In terms of the ultimate goal to realize the quantitative chemical analysis of each atomic column, more improvement needs to overcome the dechanneling and delocalization, and the absorption in electron scattering should be evaluated for each atomic column.
One of the most practical applications for EELS is to study the adsorption and reaction of molecules on metal oxide surfaces, and it is also possible to characterize the activation of adsorbed species on defects sites, particularly for O vacancies. For instance, the nature of hydrogen adsorption on TiO2 (110) can be studied by EELS [54]. After exposing the TiO2 (110) surface to atomic hydrogen at high temperatures, the vibration mode of O–H disappears, while no H2O or H2 molecules were found to desorb from the surface, which demonstrates that the H atoms adsorbed on O-bridge diffused into the bulk rather than desorption. These findings have important consequences for chemical processes involving H atoms absorbed on the TiO2 surfaces. Besides, CO oxidation on RuO2 (110) has also been evaluated by EELS, where CO was bonded weakly to Ru sites while undergone either desorption or reaction with neighboring O upon heating [55]. Notably, the EELS data further reveal that oxygen-depleted at the surface after CO2 desorption. This can be restored at the O2 atmosphere and establishes a remarkable surface redox system. This study can help to understand the mechanism of two types of Ru atom sites, where one is twofold coordinated oxygen atoms (O-bridge) and the other is fivefold coordinated Ru atoms. Another discovery was that (0001) of ZnO, with the oxygen-terminated polar surface, can be the most active surface for methanol synthesis [56]. It is expected that EELS can provide more detailed information about the growth, the chemical reactivity, and the electronic structure of metal oxide surfaces. Especially for heterogeneous catalysis, this technique can better elucidate the microscopic reaction mechanisms under industrial conditions, by bridging the material to pressure gap thereby promoting more study in surface science.
Figure 2.3 (a) Crystal structure of layered perovskite manganite La1.2Sr1.8Mn2O7. Yellow–green spheres corresponding to A site (La and Sr), blue spheres to B sites (Mn), and red spheres to oxygen. There are two different crystallographic A sites in the perovskite block (yellow arrow p) and the rock salt layer (white arrow r). (b) ADF image of the specimen observed along the [010] direction. The areas for two-dimensional EELS and drift measurement are shown by rectangles. (c) EELS spectrum acquired from the rectangular area for the two-dimensional EELS. Source: Reproduced with permission from Kimoto et al. [53]. Copyright 2007, Nature Publishing Group.
Another study carried out by EELS, particularly with an aberration corrector, is to characterize single-atom impurities or defects. The concerns include, where those atomic species are located, how exactly these atoms are bonded to another, how structural difference influences their electronic configuration, and which one exhibits unique properties, at edges or point defects as a single atom, depending on their local bonding differences [57, 58]. Also, the fine structure of electron energy loss spectra can provide the answer to the valence state of single-dopant atoms [59]. Obviously, the developments of the EELS technique have been particularly beneficial for 2D materials such as graphene [60]. To produce very high signal-to-noise data, EELS spectra were recorded with optimized acquisition conditions, such as loading graphene with Si, 1 s of acquisition time per spectrum, and calibrating the spectrometer with acceptance semi-angle of 35 mrad. Then, two types of single Si atom substitutional defects can be observed, with threefold or fourfold coordinated Si defects (as shown in Figure 2.4) [61]. The result of EELS spectra suggested that the bonding network is quite disrupted around the fourfold Si atoms, which depicted the spatial distribution of all orbitals contributing to the electron density, while the threefold showed largely undisturbed, suggesting that the threefold Si can act as a good chemical replacement for a single carbon atom, which would then integrate by distorting the sp2 structure to make the Si atom and its neighbors pucker slightly out of the graphene plane to accommodate the longer Si–C bonds. It should be noticed that more EELS were conducted under a low voltage of 60 keV to avoid the irradiation damage to the specimen. Another example is the detection of K (Z = 19) and Ca (Z = 20), which is of prime importance in the field of biological molecules [62, 63]. While in bright-field or ADF imaging, the elastic scattering power of a Ca atom is very weak to be detected. For single-atom characterization to give better EELS counts, a small diameter is not always beneficial; instead, an appropriately larger probe might be more efficient, which is capable of covering the overall space occupied by the fullerene and avoiding the escape of those encapsulated atoms during the timescale of the spectrum acquisition.
To further gain insights into the spectral imaging of EELS, many spectra are acquired as the electron probe is rastered across the specimen, forming a 2D spectral map [64]. The number of scanned points and the signal-to-noise ratio (SNR) of EELS imaging quality are limited by the amount of signal, instrument stability, and user’s time. This is greatly improved with the advent of aberration-corrected electron microscopes, which allow a larger probe-forming apertures and improve collection optics [65, 66]. Most often, the interested area contained in the core loss energy edges, where they appear in the EELS spectrum with a shape and energy onset uniquely defined by a specimen’s excitation of core-level electrons to the available density of states in the conduction band and modified by the core–hole interaction, it has been shown that the background often follows an inverse power law [67]. The signal is usually obtained after the background has been modeled and substracted over the edge of interest. However, atomic resolution EELS maps are often oversampled with pixel dimensions smaller than the probes’ transfer limit. This can be well solved with local background averaging (LBA) to estimate the background signal. This approach provides an improved background modeling, where its position can be averaged with those from neighboring spectra to obtain an accurate background signal at every position, and the reduced noise could enable a more reliable background fit and extrapolation, showing a dramatic improvement in the image contrast and SNR. Meanwhile, if the spectrum is taken in very large energy windows, the error in each channel is not equal, especially for backgrounds in the first few hundred electron volts of a spectrum. This can be overcome by iterative weighted least-squares approaches to incorporate the change in variance over the background to combine with LBA [68]. In general, the detection limits and SNR of images extracted from spectroscopic mapping highly depend on the signal processing methods, where the pre-edge power law background modeling can greatly affect the accuracy and range of the extrapolated background. In addition, LBA works well when the background has been spatially oversampled, avoiding the distortions of EELS fine structure.
