Although alveolar macrophages are the primary defensive cells of the lung against invading bacteria and foreign particles, recent studies suggest that particle uptake and translocation is a role common to alveolar epithelium (reviewed by Churg). The purpose of this investigation is to define the phagocytic capabilities of human A549 cell line using 50 nm inert Titanium Dioxide, (TiO2), particles.

A549 cells at 0.5×l06cells/ml, in F12 Kaighn’s modification medium (F12K) with 10% heat-inactivated fetal bovine serum, l00U/ml penicillin and lOOng/ml streptomycin were cultured on alcar discs in 24-well plates, incubated at 37°C, 5% CO2 in a humidified incubator for 4 days. The cells were then exposed to TiO2 particles (final concentration 4Oμg/ml) in serum-free F12K media for 24 h. After 24h, the cells were fixed in 2.5% glutaraldehyde in 0.1 M potassium phosphate buffer overnight. The cells were post-fixed with 1% OSO4, dehydrated in series of ethanol and propylene oxide, infiltrated and embedded with araldite 502.

Recent work on biological electron energy-loss spectroscopy (EELS) in the analytical electron microscope has focused on measuring concentrations of the important elements, calcium and phosphorus. Analysis of the light elements, for which EELS was originally thought to be best suited, has so far not yielded much useful information from biological specimens. Whereas low-loss fine structure has been analyzed successfully to measure distributions of water and protein in frozen-hydrated cells, core-edge fine structure cannot be recorded at sufficiently low dose to obtain useful chemical information. However, the atomic ratios of the light elements (carbon, nitrogen and oxygen) obtained from EELS are less dependent on electron dose. Together with sulfur (in proteins) and phosphorus (in nucleic acids, phospholipids and phosphates) these ratios can help determine the types and proportions of compounds that are present at the subcellular level. Here we re-examine the potential for making these kinds of measurements.

Multi-dimensional data sets are now produced by many analytical instruments. They include the series of spectra, the series of images and spectrum-images, which can be considered as a series of spectra at different positions or series of images at different wavelengths.

The automatic (or semi-automatic) handling of such data sets requires that new multivariate analysis methods are made available. For instance, if we restrict ourselves to image sets, there is a need to deduce (from the multiple maps) a single map in which regions of the specimen with approximate homogeneous properties (composition ...) can be identified and quantified.

At the present time, only a limited number of software tools are available for this purpose: - the scatterplot allows the display of the correlations between two or three spectra or images, - Interactive Correlation Partitioning (ICP) allows the user to divide the scatterplot into several parts and to reconstitute images with one selected part, -Multivariate Statistical Analysis (MSA) allows us to analyze a data set composed of several images and to identify the different sources of information, and to filter out noise and experimental artefacts.

Recent developments in instrumentation and computing power have greatly improved the potential for quantitative imaging and analysis. For example, products are now commercially available that allow the practical acquisition of spectrum images, where an EELS or EDS spectrum can be acquired from a sequence of positions on the specimen. However, such data files typically contain megabytes of information and may be difficult to manipulate and analyze conveniently or systematically. A number of techniques are being explored for the purpose of analyzing these large data sets. Multivariate statistical analysis (MSA) provides a method for analyzing the raw data set as a whole. The basis of the MSA method has been outlined by Trebbia and Bonnet.

MSA has a number of strengths relative to other methods of analysis. First, it is broadly applicable to any series of spectra or images. Applications include characterization of grain boundary segregation (position-), of channeling-enhanced microanalysis (orientation-), or of beam damage (time-variation of spectra).

The natural ability of Artificial Neural Networks to perform pattern recognition tasks makes them a valuable tool in Electron Microscopy, especially when large data sets are involved. The application of Neural Pattern Recognition to HREM, although incipient, has already produced interesting results both for one dimensional spectra and 2D images.

In the case of ID spectra, e.g. a set of EELS spectra acquired during a line scan, given a “vigilance parameter” (which sets the threshold for the correlation between two spectra to be high enough to consider them as similar) an ART-like network can distribute the incoming spectra into classes of similarity, defining a standard representation for each class. In order to enhance the discrimination ability of the network, the standard representations are orthonormalized, allowing for subtle differences betwen spectra and peak overlapping to be resolved. The projection of the incoming vectors onto the basis vectors thus formed gives rise to a profile of the data set.

In recent years quantitative high resolution electron microscopy (HREM) became a reliable tool to investigate physical processes at an atomic scale. Certainly, there is no unique approach to quantify the information content from lattice images. However, specific methods like Chemical Imaging^ and QUANTITEM were established that allow to investigate the atomic structure of crystalline solids quantitatively and almost routinely.

Nevertheless, there are limits to the application of these methods and they are of principle and of practical nature. Obviously, the resolution of quantitative HRTEM is limited by the point resolution of modern microscopes that has reached 0.1 mn. However, the quantification of the information from lattice images requires two important steps, namely, the application of a pattern recognition procedure and the extraction of changes of the electron scattering potential from the lattice images. The application of the pattern recognition procedure influences the lateral resolution of the method because a lattice image has to be broken up into unit cells of identical size (figure 1).

Multivariate Statistical Analysis (MSA) has been applied to spectroscopic data acquired by various methods but only recently to EDX spectra acquired in the transmission electron microscope where segregation sensitivity was improved by a factor of x2-3. Equilibrium segregation data were ideally suited for MSA analysis. Each spectrum was composed of a linear sum of contributions from the segregation layer and the two adjacent grains or phases, in proportions determined by the probe current distribution and location. MSA successfully identified an eigenvalue explicitly associated with segregation. Smaller eigenvalues related to self-absorption and coherent bremsstrahlung (CB) were also revealed. The use of an orthogonal MSA constrained the information sources to be independent. This constraint, which was reasonable for equilibrium segregation, has now been examined for a more complicated case: thermally and radiation-induced diffusion near an interface where profiles derived by conventional processing have varying features (Fig.l) from which it is difficult to extract real concentration profiles or even qualitative correlations between elements.

The sophisticated acquisition procedures now available in time or space resolved spectroscopies, also known as spectrum-imaging modes, produce large amounts of data which require specific developments for efficient processing and information extraction. For instance, in electron energy-loss spectroscopy (EELS), a line-spectrum consists of typically one hundred spectra recorded at regular intervals when scanning the subnanometer incident electron probe across the feature of interest: interfaces, multilayers or nanostructures of various shapes and dimensions. The useful information in any of these spectra depends on many factors such as the problem under investigation, the involved energy-loss range or the signal-to-noise ratio of the different features. However it is generally contained in the spectral changes, as well in energy channel as in position along the sequence.

To detect, measure and identify these variations, new methods have to be developed and the accompanying algorithms to be implemented. A first category encompasses all the routines which apply successively to all spectra in the sequence the well-known software which have been elaborated for processing individual spectra.

Polycrystals containing small quantities of impurity can exhibit dramatically different properties from that of the pure material. One example of this is the tenfold increase in the creep resistance of Y or La doped alumina relative to ultra-pure alumina. Such changes are correlated to grain boundary segregation, however the segregant’s potency to influence properties remains unexplained. Determining the driving force to segregation only partially addresses this issue. A full understanding must take account of the chemical bonding and structural changes occurring at the boundary after segregation. The purpose of this study is to observe the electronic structure associated with the segregating species using EELS spectrum-line profiling. This is accomplished by monitoring the segregation profile along with the characteristic near edge absorption fine structures which reflect changes in the localized density of unoccupied states along that profile. In these experiments a finely focused beam was sequentially stepped across the boundary of interest and a spectrum recorded at each step.

The local electronic structure of ceramic materials can be determined from valence electron energy loss (Veel) spectra via the dielectric function. The quality of the data is comparable to vacuum ultraviolet spectroscopy with the added benefit of high spatial resolution. We have built and implemented a system for spectrum imaging which not only allows automatic data acquisition but also analysis of the full data set. The system consists of a Gatan PEELS and a Gatan Digiscan fitted to a VG HB501 UX STEM and home-built hardware additions. The hardware extensions allow to acquire 1, 2, 3, 4 or 8 spectra for each pixel and in the case of 2 spectra/pixel to vary the exposure time of the specimen to the primary electron beam by controlling the beam blanker (Fig. 1). The first spectrum (2 to 20 ms) contains an unsaturated zero loss peak and the second spectrum (0.05 to 100 s) at the same position the plasmon peak near to but below saturation (Fig. 2).

It is well known that the high-temperature properties of polycrystalline Si3N4 ceramics are strongly influenced by the nanometer-scale glassy phase at the grain boundaries. We have recently analyzed the variation of the near-edge fine structure (ELNES) of the Si-L2,3 edges using a combination of TEM spectrum-line acquisition with an imaging filter and multivariate statistical analysis.

The glassy phase at the Si3N4 grain boundaries is easily damaged by the the fine probes usually used in scanning transmission electron microscopy to acquire ELNES data. Thus an alternative method using a Gatan imaging filter (GIF), called TEM spectrum-line analysis, was used. In this mode, energy-loss spectra are dispersed along one axis of the CCD detector, while the orthogonal axis displays the spatial variation across the interface. In this mode, a slotted washer is used as the GIF entrance aperture, with its minor axis parallel to the energy-dispersion direction of the spectrometer.

The ability of high resolution STEM instruments to provide EELS data at the ultimate atomic resolution offers significant new insights into interfacial phenomena. Not only can composition profiles now be collected plane-by-plane across an interface, but in addition, the EELS near edge fine structure can be analysed to determine the valence states of atoms at the interface, and hence the nature of the bonding across the interface plane. by applying the spatial difference technique, these effects can be studied quantitatively with high sensitivity.

Atomic resolution EELS played a critical role in explaining a completely unexpected interfacial structure found recently at CdTe/Si interfaces grown by a particular molecular beam epitaxy growth procedure. The high resolution Z-contrast image in Fig. 1 shows the CdTe film to be terminated by Te, but in addition, for several monolayers into the Si substrate, occasional columns show brighter than the surrounding Si, although remaining in the correct position.

There are now several commercially available methods for performing spatially resolved electron energy-loss spectroscopy in a transmission electron microscopy. In the energy-filtering electron microscope (EFEM), an Omega filter (Zeiss 912) or a post-column magnetic-prism filter (Gatan Imaging Filter) forms an electron image with electrons of a selected energy loss. A series of such images are acquired to provide spatially resolved information on sections of the energy-loss spectrum. These images can be processed to yield elemental maps from areas of interest in the material. In the spectrum imaging method (EMISPEC System), a small focused electron probe is rastered over the sample and a parallel energy-loss spectrum acquired from each point in the selected area. These spectra are then batch processed and information like elemental distributions extracted and displayed in the form of a map. Each of these two approaches has advantages and disadvantages and the appropriate choice is often dictated by factors such as time constraints, sample stability and the nature of the information of interest.

The spatially-resolved structure and composition of amorphous or semicrystalline polymer-polymer interfaces are poorly understood. The absence of atomic-level periodicity precludes many of the transmission electron-optical techniques that have been so fruitful in studying interfaces in polycrystalline inorganic materials. Similarly, there has been relatively little work done on spatially-resolved spectroscopic analysis of polymer-polymer interfaces to determine chemical widths, largely because of concerns over electron-beam damage. Digital microscope control and parallel spectral acquisition provide for low-dose exposure, quantification of dose, and efficient data collection which open new windows to study polymer interfaces.

This research is studying various multiphase polymer systems by focused-probe spectroscopic analysis to identify appropriate spectral features, data acquisition protocols, and background-modeling/data-processing algorithms in order to establish chemical widths characteristic of specific polymer-polymer interfaces (1). This research uses a 200keV Philips CM20 FEG TEM/STEM with a Gatan 666 PEELS spectrometer and an Emispec digital data acquisition/control system.

ELFS (energy loss fine structure), similar to XAFS (x-ray absorption fine structure) provides short-range order atomic structural information of crystalline and amorphous materials. Although known quite sometime, ELFS spectroscopy only until recently has gained importance, mainly due new developments in spectrometer and TEM designs, spectra acquisition and interpretation techniques. This paper focuses on some new developments in the acquisition, interpretation, and the use of the extended regime of the ELFS spectroscopy.

Although fine structure spectroscopy in EELS and its similarity to XAFS were recognized sometime ago, ELFS has only seen limited use mainly because of insufficient spectral statistics, various systematic errors and drifts, and the absence of sophisticated software that faithfully uses the XAFS software and incorporates the latest advances in modeling theory. We have systematically improved the data acquisition technique and developed ELFS data analysis procedure which utilizes the sophisticated UWXAFS data analysis software package8 after correction of the differences between ELFS and XAFS data.

Colossal magnetoresistive (CMR) has been observed in a class of oxides, La1-xAxMnO3 (A = Ca, Sr, or Ba). Transition and rare earth metal elements with mixed valences are usually present in these materials for stimulating structural and chemical evolutions, leading to specific functionality. The partial substitution of trivalent La3+ by divalent element A2+ is balanced by the conversion of Mn valence states between Mn3+ and Mn4+ and the creation of oxygen vacancies as well, since the ionic structure of La1-xAxMnO3-y is proposed to be

in which the valence conversion is the key for determining the material's properties. In practice, quantifying of oxygen vacancies is a challenge to existing microscopy techniques particularly for thin film specimens because of the strong effect from the defects at the substrate-film interface and the surface disordering.

Sharp peaks at threshold are a prominent feature of the L23 electron energy loss edges of both first and second row transition elements. Their intensity decreases monotonically as the atomic number increases across the period. It would therefore seem likely that the number of d electrons at a transition metal atom site and any variation with alloying could be measured from the L23 electron energy loss spectrum. Pearson measured the white line intensities for a series of both 3d and 4d transition metals. He normalised the white line intensity to the intensity in a continuum region 50eV wide starting 50eV above threshold. When this normalised intensity was plotted against the number of d electrons assumed for each elements he obtained a monotonie but non linear variation. The energy loss spectrum is given by

which is a product of p<,the density of d states, and the matrix elements for transitions between 2p and d states.

The analysis of near edge structure on inner shell ionisation edges in electron energy loss spectroscopy (EELS) can lead to new insights on the nature of bonding on an atomic scale. To fully understand the origins of spectral features it is necessary to calculate the near edge structure from a suitable theoretical model . Many of the previously published theories are based on multiple scattering of the ejected electron in a cluster of atoms surrounding the site of the excitation. The techniques used for the selection of scattering paths account for most of the differences between the various theories. In the XANES method the atoms in the vicinity of the excited atom are assigned to coordination shells and a separation is made between scattering within a given shell (intrashell) and scattering between shells (intershell). The FEFF method selects paths up to a given maximum length according to the number of scatterings and estimated amplitude.

Electron energy-loss spectroscopy (EELS) has been shown to be a powerful tool to determine not only the chemistry, but also the electronic and atomic structure at grain boundaries by analysing the energy-loss near-edge structure (ELNES). However, as the experimental ability to obtain quality spectra from interfaces with high-spatial resolution is relatively new, the interpretation of ELNES has been mostly qualitative. Here we discuss the ability of ab-initio density of state calculations to perform detailed quantitative analysis at interfaces.

In this methodology, ignoring any multiparticle effects such as excitons, the ELNES can be described as the symmetry projected density of states (DOS) and a matrix element varying slowly with energy. The calculated DOS, which best reproduce the experimental ELNES, are obtained using density functional theory (DFT) in the local density approximation (LDA) (i.e.: Weng et al.). Traditionally, because a high number of atoms (> 60) is needed to reproduce the periodicity of interface structures, more approximate methods such as the real-space multiple scattering method3 and its equivalent in reciprocal space (KKR-method) were used.

For the last few years analytical developments in intermediate voltage AEM's have remained relatively quiescent. We have instead seen a number of improvements in the basic microscope mainly along the lines of microprocessor controls, automation, and of course improvements in image resolution and electron sources. Since the spectroscopic technologies (XEDS, EELS) have seemed to reach a mid-term plateau it is appropriate to consider whether or not we have achieved the expectations in these methodologies that we predicted nearly a decade ago.

In XEDS, there are three topics to consider: signal, sensitivity and quantification. In AEM's employing conventional electron sources we expected that as a function of accelerating voltage (E0) we should realize a continuous increase in intensity even though the ionization cross-section decreases with kV, and this has been realized (figure 1). It occurs simply due to the fact that the increase in probe current with E0 is more rapid than the more slowly decreasing ionization cross-section.