Grain boundaries in fee metals with low stacking fault energy are known to undergo extended relaxations that can at times lead to a thin layer of a different structure. In Cu, for example, it has been found that ∑3﹛ 112﹜ boundaries relax into a 9R phase [1]. In this work, we have used high resolution electron microscopy to investigate the atomic structure of ∑3 grain boundaries in mazed bicrystal films of Au. Using ﹛111﹜ Ge surfaces as a template, Au bicrystals can be grown in two orientation variants, related to each other by a 60° rotation about the surface normal. As described previously, such films have a strong tendency to facet onto the coherent twin plane parallel to the substrate [2], also known as “double positioning” [3]. If films are made very thin, the likelihood for such in-plane boundaries to lie in the foil decreases, and it becomes possible to observe the atomic structure of edge-on interfaces along <111>.

Ag-Ge forms a simple binary eutectic with only two solid phases - a Ag-rich fee phase and a Ge-rich diamond-cubic phase, both with very limited solid solubility for the other element. Because both phases are cubic and there are no intermediate compounds, precipitation of Ge in an Al matrix might be expected to be crystallographically and morphologically simple. However, previous investigations of similar alloys (Al-Ge and Al-Si) have shown that precipitation in such apparently simple eutectic systems can be surprisingly complex [1,2]. For example, it was found that Ge precipitation in Al-Ge alloys depends sensitively on the availability of lattice vacancies and that at least five different orientation relationships and even more types of morphologies can form during aging after a quench. In the course of that work it was discovered that twinning is a key factor in the development of precipitate morphologies. To understand the role of twinning in precipitation this study investigates the characteristics of precipitation in Al-Ge alloys.

Spectrum imaging, where a full spectrum is acquired at each pixel in a two-dimensional array, provides a new paradigm for materials characterization that combines the advantages of traditional paradigms for imaging and analysis in electron optical characterization. Traditionally, electron optical imaging (e.g., secondary electron imaging of electron-opaque or bright-field imaging of electron-transparent specimens) emphasizes contrast production among distinct features of the microstructure. An excellent survey of the specimen microstructure is achieved - many pixels are acquired - but typically no quantitative analysis of image intensities is performed. In contrast, electron optical analysis techniques traditionally concentrate on quantitative elemental analysis (e.g., electron probe microanalysis), or at least comprehensive qualitative analysis through full spectrum acquisition (e.g., energy-dispersive X-ray or electron energy-loss spectrometry). However, these analyses are typically performed at only a few locations on the specimen, which are pre-selected on the basis of the image contrast generated as described above.

Many crucial measurements in the semiconductor industry involve determining the root cause of an electrical failure, often requiring the capability of extracting microstructural and chemical information with nanometer resolution [1]. The microanalysis is achieved by stepping the focused probe over the region of interest in STEM mode and recording an EDX and EELS spectrum at each pixel. Even for relatively modest image sizes, the resultant spectrum-image may consist of more than 10,000 spectra. With such large data sets, the prospect of manually inspecting and quantifying each spectrum in real-time is impossible and even automatic background modeling followed by peak area integration over selected spectral regions-of-interest can be excessively time consuming. Other methods to extract useful physical information include “fingerprinting” characteristic ELNES shapes to known compounds or bonding environments, multiple least squares regression to a relatively small set of suspected components or peak-fitting to model systematic changes in electronic density of states.

As more x-ray energy dispersive spectroscopy (EDS) manufacturers begin to offer spectrum imaging (a complete x-ray spectrum from each pixel in an image), there is a clear need for robust and automated methods for quickly extracting the relevant information from the large spectrum image data sets. A typical spectrum image may consist of 100 x 100 pixels (10000 spectra) each with 1000 channels which (when stored at double precision) is 80 Mbytes. It is clear that a large four-dimensional data set such as this cannot be viewed in its entirety and the time to analyze individual spectra by hand is prohibitive. Conventional analysis of spectrum images by mapping energy windows is useful as a first pass only for finding the elements present and only if at sufficient concentrations. Additional problems with mapping include systematic overlaps of other x-ray peaks, changes in the background shape and displaying the maps so they faithfully portray the actual signal intensity.

The increasing number of energy filtering transmission electron microscopes (EFTEMs) has given many microscopists the ability to apply the fast and very efficient tool of electron spectroscopic imaging (ESI) for analytical characterisation, rather than to record EELS spectra. In a more general framework, the new ESI methods and the standard PEELS technique both aim at exploring the same three-dimensional data space. This data space is represented by the spatial co-ordinates x, y and the energy loss ΔE. The resulting intensity distribution I(x,y, ΔE) is frequently referred to as a spectrum image and the experimental techniques to acquire the corresponding data are referred to as spectrum imaging. The actual analysis can either be based on recording PEELS spectra for a twodimensional array of probe positions, or on recording series of energy filtered images across inner shell loss edges or in the low loss region.

MacLispix, a public domain image processing system for the Macintosh*, has been applied to a variety of image processing problems, such as using Principal Component Analysis to explore correlated images. The tools provided by MacLispix are now available in Lispix, the updated version that runs on both the PC and the Macintosh.

I will illustrate the utility of Lispix by way of an example ‘data cube', a low voltage energy dispersive x-ray spectrum image provided by Ian Anderson. The ‘cube'is 200x150 pixels, each pixel having a spectrum of 512 two-byte channels The spectra were smoothed and reduced by adding adjacent channels to reduce them to 256 channels each. Since Lispix is image oriented rather than spectrum oriented, the reduced cube is stored, and represented internally as 256 images (one image for each channel in the spectrum), rather than as 200x150 spectra (one spectrum for each pixel in the image).

There has been intense interest in preparing single crystalline yttrium oxide (Y2O3) thin films for applications in ultra-large scale integration (ULSI) gate insulators, ULSI capacitors, and, by addition of suitable dopant species, for electroluminescent devices. Y2O3 has a C-type rare-earth sesquioxide structure, closely related to the fluorite structure with a cell parameter a= 1.060 nm and space group Th (Ia3). LaAIO3 (LAO) is a rhombohedral structure with lattice parameters a = 0.378 nm, θ ≤ 90.5°. The lattice mismatch with the <110> direction of the YO is therefore less than 0.8%, and so we would anticipate epitaxial growth of single crystalline YO thin films on the LAO (001) substrate to be feasible. Eu activated YO thin films were deposited by laser ablation on (001) LAO substrates. TEM bright field images and electron diffraction patterns were recorded in a Philips EM-400 electron microscope operated at lOOkV. Z-contrast imaging was conducted in a VG HB603 STEM at 300kV. In this presentation, we will report the epitaxial growth of YO thin films doped with ∼ wt 4% Eu on a LaAlO3 substrate, and the atomic structure of the interface.

In recent years there has been a high degree of interest in multilayer film (MLF) structures because of their applications as magnetoresistive sensors and as memory elements in magnetoresistive random access memory arrays. As each of the layers in a spin-valve-based MLF structure1 is only a few nanometers in thickness, the morphology of the layers is crucial in controlling the magnetic and transport properties of the devices. In general, the microstructural features that can influence the film properties include layer roughness, layer composition, interfacial chemical mixing, grain size, grain boundary morphology, and crystallographic orientation, with the most important microstructural parameter being the nature of the interfaces between adjacent layers. With so many internal interfaces, each of which can have a different morphology and degree of intermixing, it is extremely difficult to determine the nature of each individual interface unless a technique is used that can analyze them independently, and with high spatial resolution.

The equilibrium shape of liquid inclusions in a crystalline matrix is a direct measure of the orientation dependence, or anisotropy, of the liquid/solid interfacial energy. In this work we have used diffraction contrast imaging in TEM to observe the shape of nanoscale liquid Pb inclusions in a solid Al matrix as a function of temperature. While a previous investigation reported a highly anisotropy equilibrium shape [1], more recent work showed that the inclusion shape depends on size, temperature and thermal history [2-4]. Here we report on a new method to make accurate automated measurements of these particle shapes during in-situ heating above the melting point of the inclusions.

Fig.l shows a typical bright field image of an array of liquid Pb inclusions in a solid Al matrix recorded at temperature during in-situ heating to 423°C. Under these two-beam diffraction conditions, particles are visible by absorption contrast and appear dark on a light background.

6H-SiC is a wide band-gap semiconductor. In recent years, a variety of high-power, -temperature, -speed and optoelectronic devices have been produced in SiC films. The search for metals which can form thermally stable, uniform ohmic contacts with SiC with low resistivity is still ongoing. In this study, Cr and CrBx (1<x<2) films were deposited by electron beam evaporation onto p-type, vicinal Si-terminated (0001) 6H-SiC surface. Both contacts exhibited rectifying behavior in the as-deposited condition. Ohmic-like behavior was observed for Cr/SiC system after annealing at 1000 °C for 240 seconds in a rapid thermal anneal furnace in an Ar atmosphere. It was also reported that ohmic behavior was observed for CrB2 /SiC system after annealing at 1100 °C for 2 minutes at a pressure of 5x10“7 Torr.2 The microstructure and chemistry of these two contact systems in both as-deposited and annealed conditions were investigated by transmission electron microscopy (TEM).

BaRuO3 chemically belongs to ARuO3 family. It has unique hexagonal structure rather than an orthorhombic GdFeO3-type structure of other members, such as SrRuO3 and CaRuO3. The structure of ARuO3 compounds can be described on the basis of the close-packed stacking of AO3 layers in which the center of oxygen octahedra is occupied by ruthenium. A number of polymorph structures exist by mixing the cubic and hexagonal close-packed stacking of AO3 layers.

Up to date, only three polymorph structures have been observed in BaRuO3 bulk materials, which are denoted by 4H, 6H and 9R structure. The 9R structure (Fig. 1 (a)) is most stable at 1100 °C at atmosphere pressure. As pressure is increased, the 9R structure will transform through metastable 4H structure (Fig. 1(b)) to metastable 6H structure. The physical properties of BaRu03 polymorphs differ significantly with the variation of the structure. Hence, it is desirable to form phase pure polymorphs of BaRuO3 to examine the structure-property relationship.

The solid/liquid interface is a junction between two condensed phases with completely different atomic arrangements. At the interface between the periodically ordered solid and the amorphous liquid, the atoms adopt a structure that minimizes the excess energy due to the abrupt change between the surrounding phases. Faceted and diffuse interfaces describe two extremes in morphology of a solid/liquid interface. In a faceted interface, the change from solid to liquid occurs over one atomic layer, however periodic order extends into the first few liquid layers adjacent to the crystalline solid, as predicted by numerous models.1 The faceted interface advances by nucleation and growth of ledges on the interface. A diffuse interface has a structure in which the change from solid to liquid occurs over several atomic layers. This interface contains many ledges to which liquid atoms may attach continuously as the interface advances.

Exquisite control over the morphology of inorganic materials is well demonstrated in biological mineralization. An elegant example is the mulluscan nacre, in which aragonite (a polymorph of calcium carbonate) forms as thin films of about 0.5|im thick between organic matrices as a result of an interplay between templating and inhibition (Figure 1). Not surprising, biomineralization has inspired many recent research efforts in biomimetic materials synthesis, especially the synthesis of inorganic thin films. The majority of these efforts have exclusively focused on exploring the promoting effect on mineral formation by templates. A major drawback of this approach is the lack of control over the mineral growth in the direction normal to the template, which often leads to the formation of discrete patches instead of a true film. In this report, we describe a strategy which takes advantage of the interplay between templating and inhibiting, as utilized by organisms, to synthesize macroscopic and continuous CaCO3 thin films.

Calcium carbonate is one of the most abundant biominerals. Organisms have developed sophisticated controls over its polymorph, morphology and orientation through protein matrices. However, there are many unknowns as to how the matrices affect the crystallization process, in particular, the initial nucleation. A prevailing view is that crystals form via epitaxy from the templates. Alternatively, a crystalline phase could form through a multistep phase transformation, initiated by the formation of an amorphous phase. Since the amorphous phase is more soluble than the stable crystalline phase, it should be the first solid phase formed during the crystallization and therefore could be prevalent in biomineralization. In fact, amorphous phase minerals have been identified in many organisms and more recent discoveries have suggested that amorphous calcium carbonate is more widespread than commonly supposed in biology, but has been overlooked due to the difficulty of identifying an amorphous phase in the presence of a crystalline phase of the same composition.

The two most important characteristics of any surface considered for wafer bonding are cleanliness, surface smoothness and macroscopic flatness. Silicon wafers in the as-received condition have a native oxide on the surface several nanometers thick [1], Figure la shows that they also have a layer of hydrocarbons. While they are not clean, they are smooth. Our wafers were plasma or ion cleaned, chemically treated, and ultra high vacuum (UHV) thermal desorption annealed in different combinations to find the best method for providing smooth, contamination free substrates that will produce an atomically flat, chemically clean Si/Si bonded interface.

The first approach was a single step process to remove the contaminants and then bond the clean wafers. Cleaning was accomplished by ion bombardment of the surface in an UHV chamber with base pressure 1x109 Torr. This ion cleaning chamber is connected between the UHV (2x10-10 ) bonding chamber and UHV (1x10-10) analysis chamber, allowing wafers to be cleaned, analyzed, and bonded without breaking vacuum [2].

It is now widely accepted that one of the major issues for the future deep sub-micron integrated circuit (IC) technologies regarding yield, device performance and stability and device and product reliability, is the mechanical stress built up in the layers and substrate. It is therefore important to give a quantitative account of these stresses and this can only be achieved by disposing of reliable, performant (as to spatial resolution) and quantitative techniques for the local stress determination in the substrate, of adequate and dedicated process simulation tools and of sensitive methods to analyse the stress effects on device performance.

The only experimental technique presently available which allows the strain field distribution in sub-micron CMOS devices is the convergent beam electron diffraction technique of the transmission electron microscopy. Its sensitivity is of the order of 10-4 and its spatial resolution is of 1 nm if a TEM/FEG is employed.

High-resolution transmission electron microscopy (HRTEM) is used extensively in the semiconductor industry for device characterization, and has become one of the highly favored techniques for characterizing the latest generation of ultra-thin gate oxides in MOSFET devices. However, relatively little is understood (either quantitatively or experimentally) about the limitations of HRTEM in detecting structural defects in gate oxides that could affect device performance. To investigate model defects experimentally, it would be necessary to construct “perfect” gate oxides, introduce defects with size and morphology known perfectly a priori, successfully make thin specimens that capture the defects, and then perform imaging experiments in the HRTEM. Since that task is virtually impossible, we have performed HRTEM image simulations to assess the visibility of various structural defects in gate oxides. The gate oxide was modeled as an amorphous silicon oxide 16.3Å-thick, sandwiched between a gate and substrate. The substrate was (100) silicon viewed along the [110] direction.

High resolution transmission electron microscopy (HRTEM) has found extensive use in the semiconductor industry for performing device metrology and characterization. However, shrinking device dimensions (gate oxides are rapidly approaching 10Å) present challenges to the use of HRTEM for many applications, including gate oxide metrology. In this study, we performed HRTEM image simulations of a MOSFET device to examine the accuracy of HRTEM in measuring gate oxide thickness. Length measurements extracted from simulated images were compared to actual dimensions in the model structure to assess TEM accuracy. The effects of specimen tilt, specimen thickness, objective lens defocus and coefficient of spherical aberration (CS) on measurement accuracy were explored for nominal 10Å and 16Å gate oxide thicknesses.

The gate oxide was modeled as an amorphous silicon oxide situated between a gate electrode and substrate, both modeled as single crystal Si(100). Image simulations of the sandwich structure were performed in cross-section (with Si[110] parallel to beam direction) using the multislice approximation for a 200 kV microscope with Cs=0.5mm.

With the need to introduce new materials and processes to the semiconductor fabrication process, there is a great need to understand the diffusion of a large number of elements in silicon. To this end, a systematic SIMS study was done for implanted elements with subsequent annealing. There is a surprising lack of earlier work on many of the elements of the periodic table and the work that was done did not involve the additional effects of ion implantation.

The recent SIMS results allow the diffusion of the many elements to be categorized into several groups. These effects range from elements that do not appear to move substantially upon annealing, rapid diffusion to the surface, and formation of various peaks in the concentration profile.

Based on the SIMS results, we have used TEM to study the microstructure and composition of subset of the implants. The figure below shows the SIMS and TEM results for a Chromium implant.