Precipitation in CZ-silicon during post-growth two-stage heat treatment has been examined using the methods of high resolution analytical electron microscopy. Electron transparent specimens prepared from these specimens, exhibited a low density of plate type precipitates on {100} planes. Microdiffraction experiments showed the precipitates to be consistently non-crystalline. Electron energy loss spectra showed that the precipitates contained oxygen, but carbon was not detected. It was found that carbon artifact absorption edges could be induced in spectra by specimen contamination in the microscope. The use of a low temperature stage eliminated this problem. Complementary characteristic x-ray microanalysis showed that metallic impurities had not segregated to these precipitates in this particular case, although this has been observed elsewhere.

Small crystalline particles in the vicinity of the Si/SiO2 interface have been directly observed by high resolution transmission electron microscopy. These crystallites have typical diameters between 20 and 120 Å. Based on the observed interplanar spacings and angles in lattice images, the structure of these particles has been found to match those of cristobalite. Some orientation relationships also appear to exist between these particles and the silicon layer.

The physical properties of crystalline solids depend on the presence of point defects. The concentrations of these defects in turn depend on the conditions of preparation and the presence of dopants. Quantitative relations between these conditions (partial pressures of components, concentrations of dopants, temperature) and the defect concentrations is arrived at on the basis of defect chemistry. Examples of pure and doped binary compounds, alloys of binary compounds, and ternary compounds, are given. Whereas binary compounds have one composition variable, the alloy systems and the ternary compounds have two. The role of phase diagrams in preparing systems of required composition and properties is stressed.

A survey is given on deep defects in GaP, GaAs and InP identified by electron spin resonance (ESR). Defect structures to be discussed are (i) 3d transition metals, (ii) antisite defects and (iii) radiation induced centers. The relevance of such defects for the III–V materials technology will be illustrated by representative examples.

Self-diffusion studies are vital in the elucidation of atomic mechanisms of diffusion; as well as in the better understanding of device fabrication processes, such as the annealing of ion-implanted layers. This review outlines first the major reasons for interest in self-diffusion in III–V and II–VI compounds. It discusses the main differences with elemental semiconductors, including the wide variety of possible defects in the compounds, the role of departures from stoichiometry, and the value of tracer and interdiffusion studies. Self-diffusion studies in III–V compounds are next reviewed, including recent measurements in GaAs, where more information on diffusion mechanisms is becoming available. Interdiffusion between different III–V compounds is also discussed in the light of self-diffusion studies. Next, recent progress on self-diffusion in certain II–VI compounds is discussed, where interdiffusion studies have also provided a significant contribution. The review concludes by suggesting areas where research is urgently needed to clarify diffusion mechanisms.

Results of a detailed photoluminescence study of deep radiative transitions in InP crystals prepared by bulk and epitaxial techniques with a variety of dopants are reported. In order. to understand the origin of the photoluminescence spectra, bulk samples were subjected to isothermal anneals at different partial pressures of phosphorus. Similarly, the LPE wafers were grown with and without phosphorus in the gas stream. The electrical nature of some of the species responsible for the PL emission was inferred by a study of Cddiffused bulk samples.

Based on these experiments the following tentative assignments are proposed. The photoluminescence band at 0.99 eV, common to all samples, is due to emission from a donor-like level related to the P-vacancy-impurity complex. Bands at 1.21 eV and 1.15 eV appear to be due to emission to native acceptor levels associated with the In-vacancy. The 1.08 eV band is attributed to emission to a level involving the complex of the donor (0.99 eV) and acceptor (1.21 eV) species. The relationship between these bands and residual impurities is discussed. A comparison with the work of other investigators tends to support these assignments.

A study of the electronic states associated with divacancy defects and with the defect complexes involving an anion antisite with a group IV atom (Ac-IV) in GaAs and GaP is reported. The local densities of states have been determined using the large cluster recursion approach. The properties as well as the position of the gap states of the divacancy defect in GaAs are found to be consistent with the experimental results for the EL2 level. The change of the position of the defect levels of (Ac-IV) as a result of the change of bonding is analyzed. The effect of GaAs-A&As interface on the (Ac-IV) defect level is also examined.

GaAs Has Been Grown By Molecular-Beam Epitaxy (MBE) With Large Concentrations (∼1018CM−2) Of Sn, Si, Ge, And Mn As Dopants. The Heavily-Doped N-Type Material Has Been Found To Contain Regions Of A Very High Dislocation Density. An Analysis Of The Less Complex Defect Areas Shows That The Dislocations Originate In The MBE-Grown Layer. These Observations And Others On More Complex Defect Clusters Are Compared With Recent Studies Of Defects In Material Grown By Liquid Phase Epitaxy (LPE). The More Heavily Doped P-Type Material Contains Discs Of Mn-Rich Material At The Surface Of The MBEgrown Epilayer. Both The Structure And Composition Of These Regions Have Been Examined.

In the liquid phase epitaxy on indium phosphide, the substrate,just prior to the growth of the first epitaxial layer, is commonly etched back with an indium melt to remove any thermallydegraded surface and to ensure uniform and consistent wetting.This procedure, however, often produces defects which degrade both planar and buried heterostructure devices. For planar edgeemitters and lasers, the resulting rippled surface morphology degrades device performance by scattering light. For buried heterostructures, the meltback in the regrowth step leads to indium-rich inclusions. The effects of meltback on material quality are presented, and a new multiple meltback procedure which maintains flat surface morphology and eliminates inclusions is described.

High temperature annealing of Si 3 N4 and SiO2 capped high purity LPE GaAs is shown to result in a reduction in the surface carrier concentration by about an order of magnitude. Au Schottky contacts made on the annealed samples were found to have severely degraded breakdown characteristics. Using deep level transient spectroscopy, deep levels at EC–.58eV, EC–.785eV were detected in the SiO2, capped samples and EC–.62eV, EC–.728eV in the Si3N4 capped Samples while none was detected in the unannealed samples.The electrical degradations are explained in terms of compensation mechanisns and depletion layer recombination-generation currents due to the deep levels.

Thermal degradation of InP single crystal substrates prior to LPE growth has been virtually eliminated by using an improved protection technique. The phosphorus partial pressure provided by a Sn-In-P solution localized inside an external chamber surrounding the InP substrate prior to growth prevents thermal damage to the surface. Nomarski contrast photomicrographs,photoluminescence and X-ray diffractometric measurements indicate that InP substrates protected by this method suffer negligible deterioration, in contrast to the results of the more commonly used cover wafer method.

A conventional 120 keV high resolution TEM equipped with a TV camera has been used to make lattice resolution video recordings of cadmium telluride in the <110> orientation.The motion of dislocations and the migration of atomic species at the edge of the crystal have been studied. These processes are thought to be induced in the crystal by localized heating of the specimen by the electron beam. The observations are made on a TV monitor at a total magnification of 15 million times.

CID imaging arrays were fabricated on Hg0.7 Cd0.3Te produced by the solid state recrystallization technique. It was found that the most serious source of dark current was sub-grain boundaries. SEM studies of the microstructure revealed by etching showed that boundaries with a high denisty of dislocations were detectable sources of dark current, while those boundaries with a low density of dislocations, as well as individual dislocations were not. TEM showed that all dislocations were free of precipitates, and most were not dissociated. The sub-grain boundaries were found to arise from misorientation between dendrites which form during the solidification from the melt.

A review is presented of the work on the influence of impurities on the dynamic behavior of dislocations in silicon crystals and also on the resulting effects in the mechanical strength performed by the author's group. Special emphasis is laid on the effects of light element impurities such as oxygen, nitrogen and carbon. Although all of these impurities do not affect the velocities of dislocations moving under relatively high stresses, oxygen and nitrogen atoms are found to lock slowly moving dislocations and dislocations at rest very effectively. Such locking of dislocations results in the decreases in the activities of generation and multiplication centers for dislocations, leading to the strengthening of the crystals. The high strength of silicon crystals brought about by the impurities is lost when the crystals are subjected to the heat treatments which allow the precipitation of the impurities on a macroscopic scale. This softening is shown to be caused by dislocations punched out from precipitates and by the exhaustion of the impurities dissolved in the matrix crystal.

Simple models have been suggested to predict electronic properties of lattice defects in semiconductor crystals: dislocations ought to act via the acceptor character of dangling bonds, and small-angle grain boundaries ought to consist of regular arrays of dislocations. The actual situation in most semiconductors is, however, much more complicated. The observed electrical effects of dislocations do not confirm the dangling-bond concept, they are affected by dissociation and reconstruction. There appear to be differences between straight and kinked dislocations. Dislocations owe much of their electronic behavior to clouds and precipitates of impurities; oxygen in silicon plays a significant role. This review summarizes the present status of experimental methods and results, including luminescence and capacitance spectroscopy as well as mapping and imaging techniques using electron-microscopes.

Despite the fact that lattice imaging studies have shown that grain boundaries in group IV semiconductors often have structures which are complicated and inhomogeneous on the scale of tens-tohundreds of angstroms, simple theories assuming uniform double depletion layers have recently been shown to successfully predict many of the majority carrier transport properties of these defects. On the other hand our knowledge of the interaction of grain boundaries with minority carriers is in a considerably more primitive state. I will describe recent attempts to understand the effects of illumination on grain boundary potential barrier heights and the influence of these defects on the optically generated minority carrier population. Quantifying this latter interaction is particularly important in estimating the performance of polycrystalline solar cells. Simple but elegant scanned excitation measurements for measuring s, the minority carrier recombination velocity at grain boundaries, will be reviewed. I will discuss recent measurements of s as a function of temperature and illumination intensity and show how these data can be correlated with zero-bias impedance measurements.

Normal grain growth in polysilicon material proceeds by a twinning mechanism such that the individual grain orientations are related by single or multiple twinning steps. The twinning relationship between adjacent grains can result in the alignment of their boundary interfaces along certain planar segments which have a very low electrical activity. The varying degree of electrical activity of these boundaries, and boundary positions, is attributed to the smaller dislocation portion of the larger change in orientation across the grains.

Electronic states at a 20° symmetrical(100) tilt boundary in p-type silicon were studied using deep level transient spectroscopy (DLTS) and other electrical measurements. The data can be explained with a model in which the local barrier height at the grain boundary varies on a scale much smaller than the boundary plane (∼I mm2) under study. Based on a relationship between the carrier capture cross section and energy level deduced from the experimental data, we have been able to calculate the distribution of the density of states in the energy bandgap at the boundary, which contains two groups of continuously distributed states; a major one whose density of states increases monotonically with the position of the state from the valance band, and a minor narrow one whose density of states is centered at Ev,+0.20 eV.

The dislocation structure of a near coincidence Σ=5 [001]twist boundary in silicon was studied using transmission electron microscopy. Secondary dislocations, with localized cores, were observed in the boundary accommodating a small deviation (<.5°) from perfect coincidence. The O-lattice theory for general low angle boundaries was extended to calculate the expected dislocation content of near coincidence boundaries. Comparison between predictions and observations was used to deduce information on the primary dislocation structure of the boundary.

The electronic properties of tilt boundaries with misorientation angles ranging from 0 to 30° in n-type GaAs bicrystal layers have been investigated. The current-voltage and capacitance-voltage characteristics are consistent with a double-depletion-region model. The height of the grain boundary potential barrier remains constant while the density of grain boundary states varies with misorientation angle. Deep level transient spectroscopy has revealed the presence of two bands of grain boundary states at approximately 0.65 and 0.9 eV below the conduction band. These states are attributed to bond reconstruction at the grain boundary.