A computational test for the removal of plural scattering from angle-resolved energy loss spectra

A recently developed method based on matrix analysis for the removal of plural scattering from angle-resolved energy loss spectra is tested. A single loss function, Lorentzian in the energy and Gaussian in the angular variable is assumed as input for the test. Multiple scattering probabilities are simulated by summing up n-fold self-convolutions of the input function according to the Poisson distribution for incoherent n-fold scattering. The simulated profile serves as input for the retrieval algorithm, the result of which is compared with the original single-loss probability. It is concluded that the method is feasible, but not likely to be suited for routine investigations.     

Dose-rate dependence of electron-induced mass loss from organic specimens

Electron energy-loss measurements on thin films of collodion at low temperatures (90 K) show that the characteristic electron dose D_e for mass loss first increases and then decreases with increasing dose rate (current density). This behaviour is explained in terms of the limited diffusion rates at low specimen temperature and the heating effect of the electron beam, and can be approximately modelled using a simple computer program. For a small-diameter electron probe, the increase in D_e can be several orders of magnitude, suggesting a substantial advantage of STEM (in comparison to fixed-beam TEM) for examining beam-sensitive specimens.     

Characterising the surface and interior chemistry of core-shell nanoparticles using scanning transmission electron microscopy

A method for extracting core and shell spectra from core-shell particles with varying core to shell volume fractions is described. The method extracts the information from a single EELS spectrum image of the particle. The distribution of O and N was correctly reproduced for a nanoparticle with a TiN core and Ti-oxide shell. In addition, the O distribution from a nanoparticle with a Cu core and a Cu-oxide shell was obtained, and the extracted Cu L_2,3-core and shell spectra showed the required change in EELS near edge fine structure. The extracted spectra can be used for multiple linear least squares fitting to the raw data in the spectrum image. The effect of certain approximations on numerical accuracy, such as treating the nanoparticle as a perfect sphere, as well as the intrinsic detection limits of the technique have also been explored. The technique is most suitable for qualitative, rather than quantitative, work.     

Fourier-ratio deconvolution techniques for electron energy-loss spectroscopy (EELS)

We discuss several ways of using Fourier-ratio deconvolution to process low-loss spectra. They include removal of the tail arising from the zero-loss peak, extraction of the spectrum of a particle from data recorded from the particle on a substrate, separation of the bulk and surface components in spectra recorded from samples of the same composition but different thickness, and investigation of interface energy-loss modes. We also demonstrate the use of a Bayesian-equivalent procedure based on the Richardson–Lucy algorithm.     

Image-EELS: Simultaneous recording of multiple electron energy-loss spectra from series of electron spectroscopic images

A new approach for element microanalysis with energy-filtering transmission electron microscopy (EFTEM) is presented which was accomplished with the CEM 902 electron microscope (Zeiss, Germany). This method is called Image-EELS, because it is a synthesis of electron energy-loss spectroscopy (EELS) and electron spectroscopic imaging (ESI). Series of energy-filtered images at increasing energy losses are recorded from one area with a TV camera. In a second step the intensity of selected regions in the image stack is measured with an image analysis system and plotted as a function of the energy loss. Thus many spectra from different objects can be calculated from one image series and compared with each other. The spatial resolution of EELS is considerably enhanced, the noise is decreased because many pixels from irregular objects are integrated, and the information from ESI can be analysed as a function of the energy loss.     

Diffraction effects in inner-shell ionization edges

The influences of elastic on inelastic scattering under systematic-row conditions are described in a simple way. A kinematic approach, superposing inelastic intensity-distribution profiles centred at different Bragg spots is shown to be qualitatively correct but quantitatively unsatisfying. A model considering Bragg reflection of the fast electron before and after the inelastic scattering process and thus introducing interference effects is in good agreement with experimental results. Experimentally, we recorded inelastic intensities in the diffraction pattern of an epitaxial copper foil using a PEELS spectrometer and observed energy filtered extinction contours of a copper crystal.     

Electron energy-loss near-edge structure – a tool for the investigation of electronic structure on the nanometre scale

Electron energy-loss near-edge structure (ELNES) is a technique that can be used to measure the electronic structure (i.e. bonding) in materials with subnanometre spatial resolution. This review covers the theoretical principles behind the technique, the experimental procedures necessary to acquire good ELNES spectra, including potential artefacts, and gives examples relevant to materials science.     

A new analytical method for characterising the bonding environment at rough interfaces in high-k gate stacks using electron energy loss spectroscopy

Determining the bonding environment at a rough interface, using for example the near-edge fine structure in electron energy loss spectroscopy (EELS), is problematic since the measurement contains information from the interface and surrounding matrix phase. Here we present a novel analytical method for determining the interfacial EELS difference spectrum (with respect to the matrix phase) from a rough interface of unknown geometry, which, unlike multiple linear least squares (MLLS) fitting, does not require the use of reference spectra from suitable standards. The method is based on analysing a series of EELS spectra with variable interface to matrix volume fraction and, as an example, is applied to a TiN/poly-Si interface containing oxygen in a HfO₂-based, high-k dielectric gate stack. A silicon oxynitride layer was detected at the interface consistent with previous results based on MLLS fitting.     

Intensity analysis of boron and nitrogen K-edge spectra for hexagonal boron nitride by EELS

Boron and nitrogen K-edge spectra for hexagonal boron nitride (h-BN) were measured by electron energy loss spectroscopy (EELS) at orientations of a momentum transfer q parallel and perpendicular to the c-axis in order to investigate anisotropic unoccupied states π* and σ*, separately. Differences between intensity profiles for B_K and N_K spectra are attributed to different transition probabilities at B_K and N_K edges from Is states to the π* and σ* states. The intensity of the first peak at q ∥ c in the B_K spectrum, which is ascribed to the transition from B-1s to the Q^—₂ state in the π* band, is 10 times as large as that in the N_K spectrum. Quantitative calculation of transition probabilities for B_K and N_K spectra at q ∥ c shows that the contribution of B-2p_z orbital to the Q^—₂ state is 1–7 times as large as that of N-2p_z. This value suggests that the enhanced peak in the B_K spectrum is explained by means of single electron excitations, disregarding the electron-electron interactions for the exciton reported by several workers. The contributions of 2p_x and 2p_y orbitals of B and N atoms to σ* states are also estimated by intensity analysis of B_K and N_K spectra at q⊥c .     

A method for the characterization of surface-modified asbestos fibres by electron energy-loss spectroscopy and electron spectroscopic imaging

A method for the characterization of surface-treated asbestos fibres with electron microscopy is presented. Electron spectroscopic imaging (ESI) of organosilane-treated chrysotile asbestos fibres has been carried out. Initially, the region below the carbon edge was inspected in ESI mode for its effectiveness as a background correction. Elemental mapping was performed on standard untreated fibres to take into account non-characteristic signals from extrapolation errors and camera artefacts. The highest resulting pixel value that results from non-characteristic signals was used as a threshold for further background correction in the net images. Samples for electron energy-loss spectroscopy were prepared in two different ways, either by gluing on grids, or by using perforated carbon foils. The results show that the use of a conducting carbon film is necessary for the analysis of such electrically insulating asbestos fibres. Focusing of the electron beam on the individual fibres results in a thermal effect promoting the evaporation of the organosilane reaction products.     

Nanoanalysis by a high-resolution energy filtering transmission electron microscope

An energy-filtering transmission electron microscope with 300 kV acceleration voltage was developed and the spatial resolution of elemental distribution images was improved. Observing oxygen monolayers in Al(11)O(3)N(9), it was shown that the actual resolution attained is up to 0.5 nm. Surface plasmon loss images of silver particles were taken with a resolution of better than 0.4 nm. Furthermore, the sensitivity is sufficiently high to distinguish indium content differences of 2.5 atomic percent in In(x)Al(1-x)As. This performance is good enough to analyze elemental distribution with atomic-level resolution. Furthermore, since analysis with the energy-filtering microscope is easy and practical, nanoanalysis may come into wide use not only in academic fields but also in industry.     

Quantitative high resolution transmission electron microscopy: the need for energy filtering and the advantages of energy-loss imaging

Evidence is presented that inelastically scattered electrons contribute significant detail at the atomic level to high resolution images, particularly in high voltage instruments. The implications for quantitative image interpretation are shown to be serious and a case is made for incorporating facilities for energy-filtered imaging in future high resolution electron microscopes.     

The electron energy-loss near-edge structure (ELNES) on the N K-edges from the transition metal mononitrides with the rock-salt structure and its comparison with that on the C K-edges from the corresponding transition metal monocarbides

This paper presents the shapes of the electron energy-loss near-edges structure (ELNES) on the N K-edge of the group IV_A (Ti, Zr, Hf) and group V_A (V, Nb, Ta) transition metal mononitrides close to stoichiometry. With the exceptions of NbN and TaN, these compounds have the rock-salt (B1) structure when close to stoichiometry. NbN exists with both the rock-salt structure and a hexagonal structure. Two distinct ELNES shapes were observed from it, one of which corresponds closely with previously published data from the rock-salt structure. Under normal conditions, TaN is considered to exist only in the hexagonal form, the rock-salt form being a high-temperature/high-pressure phase although it has been reported as the result of plasma jet heating of the hexagonal form. Again two distinct ELNES shapes were observed, one of which appeared to fit into the pattern of the shapes from the other compounds with the rock-salt structure. The systematic changes of shape observed are very similar to those observed in the equivalent carbides and qualitatively follow the behaviour expected from theoretical band structures. The change in the chemical shift of the N K-edge on going from a group IV_A nitride to a group V_A nitride is ～-0·8 eV while that on going from a group IV_A carbide to a group V_A carbide is ～+0·8 eV. This difference in behaviour is explained as the result of differences in the densities of states at the Fermi levels of the compounds. The position of the first peak in the ELNES also shows a systematic change in its energy relative to the core state as the number of valence electrons in the compound increases and also as the transition series of the metal species changes. The energies, E_r, of the peaks in the ELNES relative to the threshold follow a relationship similar to that predicted by Natoli, i.e. (E_r - V)a = const. where V is the ‘muffin tin’ potential and a is the lattice parameter. The first peak gives a negative constant in the relationship. The value of constant increases for each subsequent peak up to the sixth becoming positive for the fourth and higher peaks but drops slightly on going from the sixth to the seventh peak. Each peak gives a different value of V in the relationship. The data sets for the carbides and the nitrides are systematically different in a similar way for each peak and there are deviations from linearity within each set. The systematic difference is minimized and the linearity significantly improved if the difference in the energies of two prominent peaks is used instead of E_r. This systematic variation of peak energy with lattice parameter can be used to predict the lattice parameter. If both the nitride and the carbide data for the energy of a prominent peak relative to the threshold are used, this results in a maximum deviation of 4% (or ～0·02 nm). However, if the differences in the energies of two prominent peaks are used and the data for the carbides and the nitrides are treated independently, the maximum deviation drops to 0·4% (or ～0·002 nm). At this level, uncertainties in the lattice parameters themselves come into play and better characterized materials are required to set true limits to the accuracy of the predictions. Finally some applications in the microanalysis of materials are outlined briefly.     

Mass thickness determination by electron energy loss for quantitative X-ray microanalysis in biology

As is well known, electron energy loss spectroscopy can be used to determine the relative sample thickness in the electron microscope. This paper considers how such measurements can be applied to biological samples in order to obtain the mass thickness for quantitative X-ray microanalysis. The important quantity in estimating the mass thickness from an unknown sample is the total inelastic cross section per unit mass. Models for the cross section suggest that this quantity is constant to within ±20% for most biological compounds. This is comparable with the approximation made in the continuum method for measuring mass thickness. The linearity of the energy loss technique is established by some measurements on evaporated films and quantitation is demonstrated by measurements on thin calcium standards. A significant advantage of the method is that the energy loss spectrum can be recorded at very low dose, so that mass thickness determination can be made before even the most sensitive samples suffer damage resulting in mass loss. The energy loss measurements avoid the necessity to correct the continuum measurement for stray radiation produced in the vicinity of the sample holder. Unlike the continuum method the energy loss technique requires uniform mass thickness across the probe area, but this is not usually a problem when small probes (<100 nm diameter) are used.     

Successful application of spatial difference technique to electron energy-loss spectroscopy studies of Mo/SrTiO3 interfaces

The electron energy‐loss near‐edge structure (ELNES) of Mo/SrTiO₃ interfaces has been studied using high spatial resolution electron energy‐loss spectroscopy (EELS) in a dedicated scanning transmission electron microscope. Thin films of Mo with a thickness of 50 nm were grown on (001)‐orientated SrTiO₃ surfaces by molecular beam epitaxy at 600 °C. High‐resolution transmission electron microscopy revealed that the interfaces were atomically abrupt with the (110)_Mo plane parallel to the substrate surface. Ti‐L_2,3 (～460 eV), O‐K (～530 eV), Sr‐L_2,3 (～1950 eV) and Mo‐L_2,3 (～2500 eV) absorption edges were acquired by using the Gatan Enfina parallel EELS system with a CCD detector. The interface‐specific components of the ELNES were extracted by employing the spatial difference method. The interfacial Ti‐L_2,3 edge shifted to lower energy values and the splitting due to crystal field became less pronounced compared to bulk SrTiO₃, which indicated that the Ti atoms at the interface were in a reduced oxidation state and that the symmetry of the TiO₆ octahedra was disturbed. No interfacial Sr‐L_2,3 edge was observed, which may demonstrate that Sr atoms do not participate in the interfacial bonding. An evident interface‐specific O‐K edge was found, which differs from that of the bulk in both position (0.8 ± 0.2 eV positive shift) and shape. In addition, a positive shift (0.9 ± 0.3 eV) occurred for the interfacial Mo‐L_2,3, revealing an oxidized state of Mo at the interface. Our results indicated that at the interface SrTiO₃ was terminated with TiO₂. The validity of the spatial difference technique is discussed and examined by introducing subchannel drift intentionally.     

Subcellular localization of boron in cultured melanoma cells by electron energy-loss spectroscopy of freeze-dried cryosections

Boron neutron capture therapy (BNCT) is based on the ability of the non‐radioactive isotope ¹⁰B to capture thermal neutrons and to disintegrate instantaneously. This reaction opens a way to selectively destroy tumour cells after specific uptake of ¹⁰B. In this paper, a method based on electron energy‐loss spectroscopy is presented for detecting and quantifying boron in freeze‐dried cryosections of human melanoma cells. A practical detection limit of around 6 mmol kg^?1 in 0.1‐µm² areas is estimated using specimens prepared from standard boron solutions. Preliminary results of boron mapping in the spectrum‐imaging acquisition mode reveal boron penetration and probably spot‐like accumulation within melanoma cells when exposed to culture medium containing sodium borocaptate.     

Development of a high energy resolution electron energy-loss spectroscopy microscope

We have developed a high energy resolution electron energy-loss spectroscopy (EELS) microscope, which can take spectra from specified small specimen areas and specified small reciprocal space areas to investigate detailed electronic structures. The EELS microscope is equipped with retarding Wien filters as the monochromator and the analyser. The filters are designed to achieve a stigmatic focus. The energy resolutions are 12 meV and 25 meV for cases without and with a specimen, respectively. Spatial and momentum resolutions are 30–110 nm in diameter and 1.1 nm⁻¹ in angular diameter, respectively. EELS spectra are presented to show the performance of this instrument.     

Application of recording and processing of energy-filtered image sequences for the elemental mapping of biological specimens: Imaging-Spectrum

Computerized energy-filtered transmission electron microscope (EFTEM) permits the recording and the processing of energy-filtered images, allowing a part of an electron energy-loss spectrum for each picture element to be obtained. This method, called ‘Imaging-Spectrum’, uses a Zeiss CEM902 coupled to several image analysis systems. The actual configuration records sequences of 48 images, 256 × 256 pixels, in steps of the energy loss, ΔE. Processing these sequences results in part of a core-loss EELS-spectrum for each pixel. This approach produces elemental maps with a short processing time. We have implemented three kinds of background calculation for the image subtraction. The influence of the irradiation dose and of the energy selecting slit width on the quality of the spectra is investigated. The method is applied to the analysis of some biological specimens (pericellular coat behaviour during adhesion between macrophages and red blood cells and location of calcite microcrystals in dental pulp cells). The Imaging-Spectrum method appears to be suitable for the analysis of large areas.