It has been known for some time that the velocity of dislocations in semiconductors depends strongly on the concentration of electrically active impurities. Various explanations based on different models of the electronic structure of the dislocations have been proposed; they involve the dependence of the formation energy and/or activation energy of migration of dislocation kinks on the Fermi level. A review will be presented of these theories and of recent structural models of dislocation cores and kinks.

Transient junction capacitance measurements on deformed silicon reveal a variety of states after deformation. Upon annealing or following more homogeneous deformation the defect spectrum shows a single broad feature. A state at E(0.38) has been tentatively assigned to defect sites along the dislocation. A broad band of states at H(0.35) is postulated to represent states due to the reconstructed dislocation core.

It is not presently possible to resolve the individual atoms in any semiconductor by high resolution electron microscopy (HREM). However symmetry arguments may be used to allow near-atomic resolution lattice images to be interpreted in rare favorable cases. This method is applied to the problem of distinguishing shuffle and glide set partial dislocations in silicon. It is also proposed that two dimensional characteristic loss energy selected diffraction patterns be used to reveal the local symmetry about selected substitutional species implanted in semiconductor lattices.

The structure of the silicon-sapphire interface of CVD silicon on a (1102) sapphire substrate has been studied in crøss section by high resolution transmission electron microscopy. Multibeam images of the interface region have been obtained where both the silicon and sapphire lattices are directly resolved. The interface is observed to be planar and abrupt to the instrument resolution limit of 3 Å. No interfacial phase is evident. Defects are inhomogeneously distributed at the interface: relatively defect-free regions are observed in the silicon layer in addition to regions with high concentration of defects.

The crystallography of the epitaxial Pd2 Si-silicon interface has been analyzed with special reference to the character of interfacial dislocations. It is proposed that the Pd2 Si which forms at the interface does so by the diffusion stimulated glide of transformation dislocations.

The electrical properties and crystallographic nature of linear and planar defects in EFG silicon ribbons were studied. A direct correlation between electrical and structural properties was obtained by imaging the same areas first with EBIC and then with HVEM. Coherent twin boundaries were found to be electrically inactive, but higher order twins and other grain boundaries generally enhanced minority carrier recombination. Partial dislocations confined to coherent twin boundaries were usually electrically active, but in certain instances partial dislocations were observed which had no apparent EBIC contrast.

EBIC contrast of structural defects in as-received and hydrogen passivated polysilicon ribbon is studied using aluminum Schottky barrier diodes. Enhanced charge collection after passivation is demonstrated by comparing the EBIC signal from two sections of a split ribbon, one half of which was passivated. Electrical activity of specific linear defects before and after passivation is examined by recording charge collection profiles across these defects under standardized conditions of beam voltage and current. Such measurements show that passivation does reduce the electrical activity of selected defects. Possible reasons for this behavior will be discussed.

During melt-growth of macroscopically dislocation free bulk silicon crystals (floating-zone and Czochralski technique) microdefects can form due to the condensation of thermal point defects (self-interstitials, vacancies). The formation of these imperfections, generally referred to as “swirl defects”, is strongly affected by the growth conditions (e.g. the crystal pulling rate) and crystal purity. The various reported defect formation models will be discussed. Special attention will be paid to the effect of doping on swirl defect formation.

A technique for the direct viewing of topographic images was developed based on a television system using an x-ray sensing PbO-vidicon camera tube (resolution 25 μm). Melting and growth processes of plate-shaped crystals were observed in an argon flow by this technique. The observation showed that dislocations have very high mobilities by superheating and the equilibrium dislocation density of zero is achieved just before melting. In melting of dislocation-free crystals, liquid drops (locally molten regions) were observed inside the crystals. While, crystallites appeared in the supercooled melt near the growth interface. It was found that both the drops and crystallites are not formed in 5% hydrogen/95% argon environment. Melting behavior of crystals covered with oxide films indicates that the solid-liquid interfacial free energy is greatly lowered by oxygen impurity. This oxygen effect may be responsible for both the drop and crystallite formation. Origin of swirl defects in dislocation-free crystals is discussed in the basis of these observations. Both composite and common a/2<110> dislocations were observed near the growth interface: The former are composed of three a/2<110> dislocations and are present at the interface so that, as the crystal grows, they propagate into the new crystal. They influence the interface morphology, when their Burgers vectors have the component perpendicular to the interface. The common dislocations cannot exist stably at the interface and are driven into the newly grown dislocation-free region by thermal stresses.

Some salient features of the phenomena of oxidation stacking faults (OSF) and oxidation enhanced diffusion(OED) are summarized, and some theories of OSF growth are critiqued. Then a new theory is developed which shows the growth to be reaction controlled, and the retrogrowth to be naturally a regime in the entire OSF growth process. The theory has provided an estimated bound on the activation energy of self-diffusion in silicon. It is shown that the observed power law growth kinetics can be explained quite naturally by a bimolecular annihilation process of the excess self-interstitials. A possible physical model of such an annihilation process is discussed. A number of phenomena related to the nucleation of oxygen precipitates in silicon are reported. Then a model of nucleation is presented which is shown to be consistent with all observed phenomena.In this model, the homogeneous nucleation proceeds not merely via the agglomeration of oxygen atoms, but in combination with small vacancy clusters. Evidence from the OSF and OED phenomena, and the nucleation of oxygen precipitates thus necessitates the dualism of vacancies and interstitials as thermal defects in silicon at high temperatures.

Enhanced diffusion of dopants and the formation of defects during thermal oxidation of silicon has been investigated using electron microscopy, Rutherford backscattering, and secondary ion mass spectrometry techniques. Enhanced diffusion of boron was clearly demonstrated in laser annealed specimens in which secondary defects were not present. In the presence of secondary defects, such as precipitates, enhanced diffusion of boron was not observed. The absence of enhanced diffusion during thermal oxidation was also observed for arsenic in silicon. The mechanisms associated with thermal–oxidation enhanced diffusion are discussed briefly.

In this review, dislocations introduced by prismatic punching at SiO2 precipitate sites in Czochralski silicon are shown to act as metal adsorption centers. Conditions necessary to localize SiO2 precipitate and dislocation complexes in wafer regions remote from semiconductor devices are discussed. This localization achieves an intrinsic gettering effect. The application of the intrinsic gettering mechanism to bipolar and MOS device technologies is shown to improve device performance and leakage limited yields.

Nucleation and growth of stacking faults formed in CZ silicon during oxygen precipitation have been studied using x-ray topography, TEM and FTIR. Samples were annealed in argon for various times at 550°C and 750°C followed by a 16 hour anneal in dry oxygen at 1000°C. In samples annealed at 550°C, the stacking faults were several layers thick with colonies of precipitates at their centers. The faults in samples annealed at 750°C contained only one particle and were single in nature. It is proposed that the faults are formed by thin oxygen precipitate platelets and that the different morphologies are due to different oxygen precipitation rates. The platelets are probably a modified cristobalite, as determined by micro-diffraction results.

Researchers in both the U. S. and Japan have reported that carbon impurities can significantly alter the precipitation of oxygen in CZ silicon. We have employed FTIR, x-ray topography and TEM to study this phenomenon in silicon containing < 2 × 1016 C cm−3 or 1.3 × 1017 C cm−3. A single step anneal of 16 hours at 1000°C will cause oxygen precipitation to occur in the high carbon material, while a 32 hour pre-anneal at 600°C is necessary for precipitation in the low carbon material. If a single anneal of 120 hours at 750°C is used, light precipitation occurs in both types of material. A pre-anneal of 15 minutes at 1000°C followed by 120 hours at 750°C reduces precipitation in the high carbon material only slightly, but completely eliminates precipitation in the low carbon material. It can be concluded that carbon causes heterogeneous nucleation of oxygen precipitation in CZ silicon and it is proposed that it does so by lowering the interfacial energy of the precipitates.

The pulse processing techniques that have assumed prominence over the past few years offer various important advantages for device fabrication technology. However, the usefulness of each individual method depends substantially upon the specific annealing mechanism involved. This article demonstrates the role of electron microscopy in elucidating such mechanisms and in analysing annealed semiconductor structures of importance to both research workers and semiconductor technologists. The range of laser and electron beam pulse annealing methods is covered and defect structure transitions observed are related to the solid and liquid phase processes occurring. Characteristic impurity trapping and segregation phenomena are described.

Annealing of displacement damage, the dissolution of boron precipitates, the formation of constitution supercooling cells, and the broadening of dopant profiles have been studied in laser annealed silicon. These samples were irradiated with a dye laser (λ = 0.485 µm, τ = 9 ns, E = 0.7–1.2 J cm−2) and an Excimer laser (λ = 0.193 µm, τ= 9 ns, E = 0.5–0.7 J cm−2) pulses. These results can be consistently interpreted by invoking melting during pulsed laser irradiation. Thus these results provide convincing evidence for the melting phenomenon.

The nature and depth distributions of residual damage in ion implanted and pulsed ruby laser annealed GaAs have been studied using both plan-view and cross-section transmission electron microscopy (TEM) specimens for high dose (1.0 × 1015 cm−2) Zn+, Se+ and Mg+ implants. It was found that laser energy densities above 0.36 J/cm2 were required to remove the implantation damage, this threshold energy density giving good agreement with that indicated by electrical activation measurements. Laser induced surface degradation of the GaAs was present even for energy densities as low as 0.25 J/cm2, and more severe damage, with the introduction of dislocations near the surface, was present for energy densities above 0.8 J/cm2. The use of thin SiO2 layers for encapsulation during laser annealing was found to substantially reduce this surface degradation.

The formation of convection and constitutional supercooling induced cells has been studied in indium implanted, laser annealed silicon using plan-view and cross-section electron microscopy. The convection cells were associated with the spatial inhomogeneity in the laser pulse, which leads to large temperature gradients in the lateral direction. The average size of constitutional supercooling cells decreases with increasing velocity of solidification, and it also decreases from (111) orientation to (100). The results of a perturbation theory will be discussed, which predicts the cell size and limiting concentration of instability as a function of velocity of solidification.

The crystallite morphology and defect structure of laser-crystallized Ge films have been characterized by transmission (TEM) and scanning (SEM) electron microscopy. The elongated crystallites present in these films have their long axis in the [001] direction. In regions where these crystallites are well aligned with each other, competitive lateral growth and (110) texture are observed. The defects most frequently found are [111] microtwins. A growth model that accounts for the texture and defects is proposed.

Raman scattering is used for the characterization of defects in Si. Damage is produced in single crystal silicon by ion-implantation of As and Si. The phonon structure of the damaged layer is that of the typical amorphous Si. After irradiation by pulsed laser(10ns,532nm) at energy density of approximately 0.1J/cm2, a Raman peak appears at a frequency between 508 cm−1 and 517 cm−1 depending on implant dosage. The higher the implant dosage, the lower is the frequency. We explain this in terms of the residual bonding defects caused by the presence of extraneous atoms such as oxygen. On the other hand, irradiation at an energy density in excess of 0.5 J/cm2, a Raman peak appears at a frequency close to that of the single crystal except for small shifts due to Fano-shift. For implant dosage in excess of 4×1016 As/cm2 , we have found additional peaks at 222 cm−1 and 267 cm−1 which are close to the metallic arsenic modes indicating the presence of arsenic clusters.