The demands of the National Technology Roadmap for Semiconductors will necessitate measurement of dopant concentrations with greater spatial resolution than now possible. Current experimental and simulation experience indicate that Annular Dark Field (ADF) imaging in a Scanning Transmission Electron Microscope (STEM) should be able to determine dopant distributions with near atomic resolution. The ADF signal is derived from electrons diffusely scattered to high angles, resulting in contrast due to atomic number (Z-contrast) and defects in the crystal lattice. Thus, heavy atoms can be imaged by their Z-contrast and small atoms by the misfit strain induced in the silicon lattice. Atomic number scattering is proportional to Zn where n is between 1.5 and 1.9 depending upon the inner detector angle of the ADF detector.

The ability to map semiconductor doping distributions with high spatial resolution (∽10 nm) and high compositional sensitivity (of order 1 ppm) is of enormous importance to the microelectronics and optoelectronics industry. Although several methods (e.g. scanning capacitance microscopy, masked secondary ion mass spectroscopy, secondary electron microscopy, electron holography) are under development, no technique currently offers the combination of sufficient resolution, sensitivity and reproducibility to fully address characterization requirements. In this paper, we describe development of a technique which utilizes focused ion beam (FIB) sputtering and transmission electron microscopy (TEM) to enable direct imaging of dopant distributions in InP-based semiconductor devices, with spatial resolution of order 10 - 30 ran, and compositional sensitivity of order 1017 cm-3 (i.e. 5 ppm).

Laser diode samples are prepared for TEM imaging using a FEI 200 30 kV Ga+ focused ion beam system. A schematic of the relevant structure is shown in Figure 1(a).

The semiconductor industry demands elemental information from ever smaller regions. Two types of information in demand are two dimensional dopant profiles for the MOS transisitor and identification of particles as small as 30 nm diameter. The work of Levi-Setti and others resulted in liquid metal ion source (LMIS) instruments that provided secondary ion mass spectrometry (SIMS) images using Ga+ beams with 20 nm lateral resolution. It is now possible to purchase focused ion beam (FIB) systems with 5 nm beam capability and SIMS detection.

The application of LMIS SIMS to meet semiconductor demands has been pursued in our laboratory with a FEI-800 FIB. SIMS imaging of semiconductor patterns after etch has shown the ability to identify boron and carbon contamination. Figure 1 shows boron in a comb structure after a BC13 etch. The boron can be shown to be removed by a cleaning step.

A great deal can be learned about integrated circuits (ICs) and microelectronic structures simply by imaging them in a focused ion beam (FIB) system. FIB systems have evolved during the past decade from something of a curiosity to absolutely essential tools for microelectronics design verification and failure analysis. FIB system capabilities include localized material removal, localized deposition of conductors and insulators, and imaging. A major commercial driver for FIB systems is their usefulness in the design debugging cycle by (1) rewiring ICs quickly to test design changes and (2) making connection to deep conductors to facilitate electrical probing of complex ICs. FIB milling is also used for making precision cross sections and for TEM sample preparation of microelectronic structures for failure analysis and yield enhancement applications.

Recently, we developed the dedicated FIB system(FB-2000A) for the TEM specimen preparation. The system has ion optics with the maximum ion beam current density of 15 A/cm2 and the minimum ion probe size of l0nm. In the system, the side entry type specimen stage originally designed for high resolution TEM was employed, and the drift rate of the stage during FIB milling is less than lnm/min. The FIB-TEM compatible specimen holder has been developed it made milling of specimen in FIB system and observation of milled specimen in TEM without remounting the specimen possible.. The holder was designed to rotate the specimen around the axis of specimen holder so that both side of milled surfaces can be faced to incident electron beam in TEM.

To improve positional resolution in pin point TEM specimen preparation, it is important to know the inside structure of specimen before starting the milling.

The ongoing reduction of scale of semiconductor device structures places increasing demands on the sample preparation methods used for transmission electron microscopy (TEM). Much of the semiconductor industry's failure analysis and new process development effort requires specific transistor, metal or dielectric structures to be analyzed using TEM techniques. Focused ion beam (FIB) milling has emerged as a valuable technique for site-specific TEM sample preparation. FIB milling, typically with 25-50kV Ga+ ions, enables thin TEM samples to be prepared with submicron precision. However, Ga+ ion milling significantly modifies the surfaces of TEM samples by implantation and amorphization. Previous work using 90° milling angles has shown that Ga+ ion milling of Si produces a surface damage layer that is 280Å thick. This damage is problematical since the current generation of semiconductor devices requires TEM samples in the 500-1000Å thickness range.

Heteroepitaxial growth of Group IV elements on Si is attracting increased attention because of the possibility of strain compensation in addition to bandgap engineering. The incorporation of the smaller C atom into Si1-xGex binary alloys to compensate for the larger size of the Ge atom offers the prospect of lattice matching and hence strain-free growth. In our early work, ternary SiGeC alloy films with up to ∽ 2% C were epitaxial with excellent crystallinity and very few interfacial defects. However, with increased C content, the films developed considerable disorder away from the substrate and finally became amorphous. Moreover, even at low C contents (∽2-3%), it appears that substantial C is being incorporated interstitially rather than substitutionally into the covalent lattice.2 Our recent work has therefore been aimed at establishing growth conditions that enable the amount of substitutional C to be maximized while still maintaining high crystal quality.

Wide gap nitride semiconductors have attracted significant attention recently due to their promising performance as short-wavelength light emitting diodes (LEDs) and blue lasers. One interesting issue concerning GaN is that the material is relatively insensitive to the presence of a density of dislocations which is six orders of magnitude higher than that for III-V arsenide and phosphide based LEDs. Although it is well known that these dislocations originate at the film-substrate interface during film growth, thread through the whole epilayer with line direction along <0001> and are perfect dislocations with Burgers vectors of a, c, or c+a, the reason why they have such a small effect on the properties of GaN is unclear.

To develop a fundamental understanding of the properties of these dislocations, the core structures are studied here by high resolution Z-contrast imaging in a 300kV VG HB603 scanning transmission electron microscope (STEM) with a resolution of 0.13nm.

Strained epitaxial films are often unstable to morphological transformations. The classic example of this phenomenon is the SiGe system. A nonplanar morphology results in non-uniform strains in the system (Figure 1). These strain fields can modify the local chemical potential of a uniform alloy driving Ge enrichment in the regions of low compression.

Recent work has shown that annealed films of SiGe on Si{001} exhibit both vertical and lateral segregation during the transition from a metastable planar film to an array of ﹛100﹜ directed island. Vertical segregation is evidenced by an enrichment of Ge in the last few monolayers of the film while lateral segregation is seen as an enhancement of this surface layer in the region near the island peaks. This segregation is due to surface diffusion as the bulk diffusion lengths are too short. The case of annealed, metastable films can be compared to growing films.

Low-temperature GaAs (LT-GaAs) of a high resistivity and a short carrier lifetime (0.2-20 ps) is promising for device applications. High temperature annealing of LT-GaAs often leads to formation of twinned As precipitates in the GaAs matrix. Here high resolution electron microscopy has been applied to study the structure of As precipitates formed in MBE grown LT-GaAs layers during rapid thermal annealing at 850 and 950 °C. Twinning in the rhombohedral As was a special focus of this study.

Fig. 1 shows a HREM image of a typical As precipitate formed in a LT-GaAs layer after annealing at 850 °C. The precipitate has a rhombohedral structure with a=0.376 nm and c= 1.055 nm and a polyhedral shape with distinct facets. The long facets bounded by the {111}A planes of the GaAs are planar while the others have a high density of steps.

Titanium disilicide (C54-TiSi2) is a low resistivity silicide (15 - 20 μΩ-cm) and is widely used in the device industry. It is formed at about 750-850 °C when thin layers (∽30- lOOnm) of Ti on poly- or single-crystal Si substrates are subjected to rapid thermal annealing (3 °C/sec) in a controlled atmosphere (N2). During the anneal, other Ti silicides such as Ti5Si3, Ti5Si4 ,TiSi and C49-TiSi2 may form prior to the desirable C54-TiSi2.

Some attempts have been made to promote low-temperature C54-TiSi2 formation. Depositing a Mo (l-2nm) interlayer between Ti and Si has been reported to decrease the C54 formation temperature by 100 °C.2 Codepositing Ti with Ta, Nb or Mo has successfully decreased the formation temperature by about 150 °C.3 These findings have been interpreted in terms of a template mechanism which facilitates formation of C54 by advantageous lattice matching between similar planes in C54 and a hexagonal ternary (Ti- X-Si, X=Ta, Nb, Mo) C40 precursor phase.

GaN is a semiconductor with a direct band-gap of 3.4 eV which has a potential application in optoelectronic devices such as UV-emitting lasers and blue light emitting diodes. Extensive efforts have been made to develop low-resistance, thermally stable and uniform ohmic contacts on GaN. In this study, n-GaN films (Si-doped) with an intermediate A1N buffer layer were grown on 6H-SiC substrates via MOVPE. The contacts were formed by depositing Ti and then Au films on GaN by e-beam evaporation at room temperature. Current-voltage measurements as a function of temperature showed that the ohmicity of the Au/Ti contact improved upon annealing, corresponding to changes in the microstructure of the contact interfaces involving a redistribution of nitrogen and gold.

Fig. 1 shows the microstructure of the as-deposited Au/Ti/GaN films. The nanocrystalline Ti film was about 50 nm thick. Au film was polycrystalline, containing many growth twins. The thickness of Au film varied from 260 to 350 nm.

Over the last few years, considerable efforts have been made in developing fabrication processes for thick superconducting YBa2Cu3O7-δ film on a metallic substrate for electric power applications. Since these applications require the composite tape to be longer than 4-500m, any methods chosen need to be compatible with such length requirement. One possible approach is the so-called BaF2 process which involves deposition of a unreacted mixture of a BaF2-Y-Cu-O layer on a substrate by an electron beam evaporation method, followed by heat-treatment of the layer in a mixed-gas atmosphere of a low partial pressure O2, saturated H2O, and N2. Although it was shown previously that c-axis aligned films up to ∼ I m long can be produced under these conditions it is uncertain how thick the films can be made and this is of crucial importance as most applications require film thickness to be greater than ∼5μm to carry a large, dissipationless current.

In recent years, many efforts have been made to fabricate high-Tc superconductor/normal-metal/ superconductor (SNS) Josephson junctions, but no breakthrough has been made. The major difficulties are the poor reproducibility of the junction parameters due to the structural complexity and defects in the barrier layer. Some junctions, such as the one with Co-doped YBCO barrier, have relatively small spreads in properties. However, the value of their normal state resistance Rn is low, which limits their use for many applications. One approach to solve this “low-Rn” problem is to use materials of high resistivity for the barrier layer. PBCO is an insulator near the metal-insulator transition. Doping with Co drives it further into the insulating regime. The lattice constants of PBCO are close to that of YBCO, even closer when doped with Co. Therefore, Co-PBCO may form a high-resistivity barrier which matches with YBCO very well and hence is a potential candidate barrier material.

(Bi,Pb)2Sr2Ca2Cu3Ox/Ag composite tapes are currently the most promising high-temperature superconductors for large-scale applications. A major issue in further improving the supercurrent-carrying-capability and optimizing fabrication procedures of these tapes is to understand the formation kinetics of the Bi(2223) from precursor powders of Bi(2212), CaO, and CuO. Since the microstructural transformation during the complicated phase evolution is not well known yet, identification of minor phases at different transformation stage is desirable. Here, we report our crystallographic analysis of several frequently observed intermediate phases during the phase evolution.

The heat treatment of the tapes consisted of ramping the temperature to 840°C in air and of holding it for a prescribed period and then oil-quenching from various temperatures.

Grain boundaries (GBs) in laser deposited YB2Cu3O7-δ/MgO(001) thin films have been investigated by high-resolution transmission electron microscopy (TEM) and scanning TEM (STEM). We report both statistics and atomic structure of low-angle and high-angle [001] tilt grain boundaries resulting from almost perfect c-axis textured YBCO films.

Atomic structure of low-angle GBs was analyzed using a dislocation model. These boundaries have been found to be aligned primarily along (100) and (110) interface planes. For (100) boundary plane, the GB consists of a periodic array of [100] dislocations (Fig.l). For (110) boundary plane, the array is also periodic but every [110] dislocation is split by ∼ 1.5 nm into two [100] and [010] dislocations (Fig.2). We have calculated energy of various configurations and shown that the energy of the (110) boundary with dissociated dislocations is comparable to that of (100) boundary, which explains the coexistence of (100) and (110) interface facets along the boundary.

Sometimes oxides are friendly; x-ray or neutron diffraction solves the structure and (at most) simple HREM does a double check. Sometimes they are not. Problems can range from phase purity (identifiable using EDX) to more subtle issues about the true unit cell and/or atom locations. The latter are often particularaly hard for x-ray or neutron to handle when single crystal data is not available and, for instance, the samples are polyphase or textured.

One common problem is that the unit cell is rather larger than originally thought due to a superstructure. While superstructure reflections may be obvious in the microscope due to the very large dynamic range of TED patterns, they may not show in powder patterns. The issue then is to determine the true structure using (often) a combination of HREM and TED plus multislice simulations to confirm the structure.

A key towards understanding the properties of high-temperature superconductors is knowledge about the distribution of the valence electrons and holes. Despite their complicated crystal structure, electronic structure of many cuprates has been calculated. The most successful calculation method is based on the theorem that the ground state energy is a unique functional of the electron density. Thus, experimental techniques that give information about the valence electron density are essential to test the calculations and to gain further insight into these important materials.

To move beyond locating the atoms in the crystal unit cell towards addressing the details of the electron distribution requires in general extremely accurate diffraction experiments. Highly perfect crystals with small unit cells have been approached. It has been known for several decades that in electron diffraction the amplitudes of those reflections at small scattering angles that are present in crystals with large unit cells, are strongly influenced by the valence electron distribution.

Cu2O and CuO are of great interest after the discovery of the high Tc superconductors. The superconductivity is believed to be dependent on charge states in CuO2 layers whose structure is similar to that of CuO. These transition-metal compounds are also important because of the different copper coordination numbers and different copper-oxygen distances, therefore these reflect different electron density distributions and bonding mechanisms. The electronic structure of cuprite has been studied extensively both theoretically and experimentally because of its symmetrical crystallographic structure (space group ) while that of CuO (space group C2/c) which has a monoclinic structure has been studied only recently. Previous experimental and theoretical results show that the valence electrons of copper in cuprite form hybrid states which is proposed by Orgel instead of spherical d10 state as in ionic copper.

A knowledge of grain boundaries (GBs) and interfaces is essential in establishing the relationship between the structural details of the boundaries and the strength of superconducting coupling and critical current densities at the boundaries in textured Bi cuprates. As a step towards such understanding, we studied GB crystallography and lattice mismatch using the constrained Coincidence-Site-Lattice (cCSL) theory. We note that to characterize arbitrary GBs in a complex system such as Bi2212, it would be desirable to find some simple analytical solutions for the cCSL approach.

We report our new concept of a “two-mirrors” model (2M-model), which allow us, in principle, to find sufficient conditions to generate corresponding analytical solutions for CSL (or cCSL), when it is applied to crystals with cubic, hexagonal, or orthorhombic symmetry. The 2M-model is based on the fact that any boundary can be described by rotating one crystal relative to the other along a common axis.