Knowledge about compositional profiles on an atomic scale is important for semiconductor multilayers. In this paper, we attempt to quantify the Ti atomic fraction in a TixAll–xN multilayer and the total As concentrations in As δ-doped layers using energy-filtered imaging. These two materials represent materials where the characteristic energy loss edges are located in widely different energy losses with the L edge of Ti being above 450eV and that of As around 1350eV. The accuracy of the Ti atomic fraction in TixAll–xN is found to be around 10at% for specimens of uniform thickness made by focused ion beam milling, whereas the resolution and As concentration for the As containing δ-layer is found to be dominated by the signal to noise ratio.

The development of energy filtered imaging systems for the TEM has opened new approaches to analyzing structures with very small dimensions. One of the benefits of such systems is the ability to form an EELS spectrum image containing energy information on one axis and spatial information on the other axis. By dissecting such an image in the spatial dimension, it is possible to generate a line profile across a layered structure at high spatial dimension without the need for a small probe. We have applied this approach to two structures typical of semiconductors, the Si/SiO2/Si interfaces in a gate oxide and the Ti/SiO2 interface for Al barrier metallization, which illustrate the spatial resolution possible with this technique. To analyze the spectrum images, the element specific near edge structure (ELNES) data are processed by conventional EELS background routines and multivariate statistical techniques using MATLAB software to extract both profiles of principal bonding components and composition.

The latest transmission electron microscopes with field emission guns and imaging filters now provide much of the microanalysis and imaging necessary in applications such as ULSI device development. The installation and operating environment of the instruments are critical to their successful operation. Information from two such installations is presented here, one in a purpose built facility and the other in an existing building. Ground vibration, acoustic noise, stray electromagnetic fields, air flow and temperature variation are considered, and the measures implemented to achieve desirable levels of each parameter are discussed. The physical layout of an installation is also shown.

The use of SEM-based defect review tools has increased dramatically over the past five years as the semiconductor industry moved from 0.7 micron to 0.25 micron design rules. During this period, a dramatic inflection occurred at the 0.5 micron node; optical microscopy lacked sufficient resolution to determine even if a simple etch step was properly performed. Accordingly, many “inspection SEMs” were introduced into the wafer fabrication facility. With ever increasing focus on yield improvement, defect review SEM's proliferated in the fab in an effort to drive down both baseline defects and process excursions. In order for such an effort to be successful, a clear improvement in the low voltage image resolution performance of the scanning electron microscope was required. Commercial vendors have responded with impressive tools achieving image resolutions of 2.5–4nm @1 kV. At this level of performance, routine observation of semiconductor wafers is possible at 100,000X magnification. This dramatic improvement in resolution is the result of many factors, by far the biggest of which is the use of “immersion lens” designs which employ a strongly excited objective lens operated at short working distances (∼2mm). This electron optic design was first introduced into the laboratory SEM market, with instruments capable of producing approximately 1.2nm resolution at 20kV. These high performance lenses in which the magnetic field extends below the bottom of the polepiece, were fitted onto 200mm platforms allowing whole wafer inspection/defect review. The features of such tools and their application to the IC industry is the subject of this paper. An example of the superb imaging performance of such a 200mm tool is illustrated in figure 1.

Microstructures of Barium Strontium Titanate (Ba, Sr)TiO3 [BST] thin film play an important role in determining the electrical properties of BST. In particular, it is found that the grain size distribution as a function of deposition conditions correlates with the dielectric constant of BST film. Traditionally, Transmission Electron Microscopy (TEM) provides an accurate method for determining microstructures such as interface structure between BST and electrodes and BST grain size distribution. However, TEM analysis relies heavily upon successful sample preparation, and film adhesion for BST proves to be a difficult problem to overcome for successful sample preparation. With the state of the art Scanning Electron Microscope (SEM) and Atomic Force Microscopy (AFM), useful information can be obtained for BST and electrode microstructures requiring little or no sample preparation. A good correlation among TEM, SEM and AFM techniques is achieved which allows useful information of BST grain size distribution to he obtained via SEM and AFM analyses. Power spectral density (PSD) analysis of contrast enhanced AFM images proves to be an efficient method for estimating BST grain size distribution.

The role of fluorine (F) in the break-up of the native oxide and the regrowth of As doped poly-Si layers on unpatterned Si wafers and on patterned regions of Si device wafers at temperatures of 900°C to 1000°C are investigated by TEM and by the fabrication of npn poly-Si emitter bipolar devices. Results for unpatterned wafers with F show i) a 950°C dopant drive-in anneal causes oxide break-up and regrowth after a time suitable for the fabrication of devices, ii) a pre-anneal, before the As implant, further enhances the break-up and regrowth and iii) there is an optimum F dose of 5×1015/cm2. Based on these results poly-Si emitter bipolar devices were fabricated using F=5×1015/cm2, a pre-anneal and a 900°C As drive-in anneal. The results establish quantitatively the relationship between the interface structures and the specific emitter resistance, i.e. with no F there is no break-up or regrowth and the emitter resistance is high (114Ωμm2) while with F there is break-up and regrowth and the emitter resistance is low (17Ωμm2).

In-situ transmission electron microscopy (TEM) has been utilized in conjunction with conventional ex-situ Rutherford backscattering spectrometry and channeling (RBS/C), in-situ time resolved reflectivity (TRR) and ex-situ TEM to study the influence of substrate orientation on the solid-phase epitaxial growth (SPEG) of amorphised GaAs. A thin amorphous layer was produced on semi-insulating (100), (110) and (111) GaAs substrates by ion implantation of 190 and 200 keV Ga and As ions, respectively, to a total dose of 1e14/cm2. During implantation, substrates were maintained at liquid nitrogen temperature. In-situ annealing at ∼260°C was performed in the electron microscope and the data obtained was quantitatively analysed. It has been demonstrated that the non-planarity of the crystalline-amorphous (c/a)-interface was greatest for the (111) substrate orientation and least for the (110) substrate orientation. The roughness was measured in terms of the length of the a/c-interface in given window as a function of depth on a frame captured from the recorded video of the in-situ TEM experiments. The roughness of the c/a-interface was determined by the size of the angle subtended by the microtwins with respect to the interface on ex-situ TEM cross-sectional micrographs. The angle was both calculated and measured and was the largest in the case of (111) plane. The twinned fraction as a function of orientation, was calculated in terms of the disorder measured from the RBS/C and it was greatest for the (111) orientation.

The characterisation of semiconductor thin films and device structures increasingly requires the use of a variety of complementary electron microscope-based techniques as feature sizes decrease. We illustrate how layer electrical and structural properties can be correlated: firstly averaged over the bulk and then on the individual defect scale, e.g. scanning transmission electron beam induced conductivity can be used to image the recombination activity of orthogonal <110> misfit dislocations within relaxed MBE grown Si/Si1-xGex/Si(001) heterostructures on the sub-micrometre scale. There is also need for improved understanding of sample preparation procedures and imaging conditions such that materials issues relevant to ULSI development can be addressed without hindrance from artefact structures. Hence, we consider how point defects interact under the imaging electron beam and the relative merits of argon ion milling, reactive ion beam etching, focused ion beam milling and plasma cleaning when used for TEM sample preparation. Advances in sample preparation procedures must also respect inherent problems such as thin foil surface relaxation effects, e.g. cleaved wedge geometries are more appropriate than conventional cross-sections for the quantitative characterisation of δ-doped layers. Choice of the right imaging technique for the problem to be addressed is illustrated through consideration of polySi/Si emitter interfaces within bipolar transistor structures. The development of microscopies for the rapid analysis of electronic materials requires wider consideration of non-destructive techniques of assessment, e.g. reflection high energy electron diffraction in a modified TEM is briefly described.

The crystalline structure of metamorphic heterostructures grown by epitaxy has been studied by digital processing of High Resolution Electron Microscopy (HREM) images. Two systems have been investigated: the GaSb/(001)GaAs, known to be fully relaxed by a perfect Lomer dislocation network and the GaAs/(001)InP relaxed by partial and 60° dislocations randomly distributed. A transition zone can be defined between the perfect substrate and the relaxed epitaxial layer: its thickness is less than 20Å in GaSb/GaAs and more than 80Å in GaAs/InP. These results indicate that the misfit dislocations are only one of the elements involved in the relaxation of misfit stress.

Using Transmission Electron Microscopy, we studied the misfit and threading dislocations in InAs epilayers. All the samples, with thickness around 0.5μm, were grown on GaAs(001) substrates by molecular beam epitaxy under As-rich or In-rich conditions. The As-rich growth undergoes 2D-3D mode transition process, which was inhibited under In-rich surface. High step formation energy under As-deficient reconstruction inhibits the formation of 3D islands and leads to 2D growth. The mechanism of misfit dislocations formation was different under different growth condition which caused the variation of threading dislocation density in the epilayers.

A Transmission Electron Microscopy (TEM), Photoluminescence (PL) and Photoluminescence Excitation Spectroscopy (PLE) investigation has been conducted on Ga0.52In0.48P epilayers grown on (001) GaAs substrates by Gas-Source Molecular Beam Epitaxy.For Ga0.52In0.48P epilayers grown on exact (001) GaAs substrates, increasing the growth temperature from 480°C to 535°C increased the antiphase domain plate thickness, t, from 7.3±0.4 to 17.4±0.9 Å, and decreased the long range order parameter, n, from 0.32 to 0.18±0.1. For epilayers grown at 530°C, on GaAs(001) substrates off-cut 0°, 7°, 10° and 15° towards [111]A, increasing the substrate misorientation from 0° to 15° decreased the antiphase domain plate thickness, from 12.3±0.6 to 6.0±0.3 Å. The long range order parameter also decreased from 0.19 to 0.10±0.01.

The band gap energies of these samples, grown by GS-MBE, were close to those reported for fully disordered Ga0.52In0.48P epilayers grown by MOCVD at ∼760°C. This shows that GSMBE is also a good technique to grow GaInP for high band gap optical data storage applications and at lower growth temperatures. The optimum growth conditions in this study were at a growth temperature of 530°C on (001) GaAs substrate with 15° off-cut towards [111]A.

In this paper, the ordering of ternary semiconductor compounds is briefly reviewed by means of a coordination polyhedron model. Long-range ordering of chalcopyrite and CuAu-type structures can be represented as an array of repeating A2B2 tetrahedra. A CuAu-type ordered phase in a chalcopyrite AIBIIICVI2 compound is surrounded by an A3B+ AB3 boundary, whereas a CuPt-type ordered phase in a zinc-blende (A, B)IIICV compound is surrounded mainly by A2B2 type tetrahedra and thus restricted in size. Following the description of the ordered structure model, the detection of the asymmetry in ordering directions in (A, B)IIICV compounds is discussed. Some examples that employ transmission electron microscopy are presented.

The microstructural evolution of the GaSe/GaAs(100) thin film system was characterized by high resolution transmission electron microscopy (HRTEM). The relationship of crystallographic orientation between the GaSe thin film and the GaAs substrate is [011]GaAs ‖ [1100]GaSe/(100)GaAs ‖ (0001)GaSe, with the dominant polytype for the GaSe thin film being a γ-type which has a 3R-rhombohedral structure with R3m space group. An intermediate layer was observed between the GaAs substrate and the GaSe thin film, with a structure distinct from that of either GaAs or GaSe. The result of a fast Fourier transform (FFT) of the lattice image corresponding to the intermediate layer indicated that its crystal structure was associated with that of α-Ga2Se3. A new vacancy ordered structural model of β-Ga2Se3 was suggested.

The two sister techniques, Electron Backscatter Diffraction and Orientation Imaging Microscopy which operate in a scanning electron microscope, are well established tools for the characterization of polycrystalline materials. Experiment has shown that the limiting resolution for mapping is the order of 0.1 microns. The basic techniques have been extended to include multiphase mapping. Whereas it has been possible to distinguish between phases of different crystal systems easily, it has not been possible to distinguish between phases that differ in lattice parameter by less than 5 %.

An equivalent transmission electron microscope procedure has been developed. The technique couples standard hollow cone microscopy procedures with dark field microscopy. All possible dark field images that can be produced by tilting the electron beam are scanned to detect under what settings each crystal is brought into a diffracting condition. Subsequent analysis permits determination of both crystal phase and orientation.