Recent experimental investigations have shown that electrical deactivation of arsenic in silicon creates excess silicon interstitials. This study investigated the possibility of excess vacancy generation during arsenic activation. We used Sb doping superlattice structures containing six 10 nm wide Sb doped spikes separated by 100 nm. It was found that antimony diffusion was not enhanced as active arsenic concentration increased, indicating there is no observable vacancy injection out of the arsenic layer during the activation process. Plan-view transmission electron microscopy study of the samples revealed dislocation loops before the activation anneal. Although the loops completely dissolved during the activation anneal, they do not seem to be sufficient enough to absorb all the vacancies generated by the activated arsenic. When germanium was present at the surface instead of arsenic, antimony diffusion was enhanced.

The behavior of extrinsic dislocation loops in silicon was investigated by transmission electron microscopy. Loops were formed by an amorphizing implant and recrystallization anneal of Si wafers. Wafers were further annealed in either Ar or NH3. Wafers annealed in NH3 formed a thin (∼4 nm) SiNx film. The loops in samples in Ar showed a constant net number of interstitials bound by the loops, while those in samples annealed in NH3 showed a marked decrease. The results are explained by a supersaturation of vacancies recombining with the interstitials in loops. By integrating the measured difference between interstitials bound by the loops in samples annealed in Ar vs. NH3 over the distance from the surface to the loop layer, an estimate for the relative vacancy supersaturation is extracted. Comparison with estimates of vacancy supersaturations with nitridation from the change in Sb diffusivity show good agreement between the two methods.

It is shown that defects generated by ion implantation, remaining after annealing at low temperature, are deep-distributed in the bulk silicon and their amount is demonstrated to be function of the substrate type and the implanted ion species. The confirmation that defects penetrate deeply into the silicon is made by a new method that consists in damaging by ion implantation a previously formed pn junction that shows very low leakage current and has a deep junction. It is proposed that the dopants in the substrate act as nucleation centers for the formation of point defect clusters and that these clusters actually degrade the junction. It was found that point defects penetrate much more deeply in p+n junctions than in n+p junctions. It was also found that BF2+ introduces much more defects into the silicon than As+, owing to the presence of fluorine. The leakage currents at 5 V of n+p and p+n diodes made by implantation of P+ and B+, respectively, could be lowered by one to two orders of magnitude with respect to values obtained by implantation of As+ and BF2+ because the former ones produce less defects than the latter.

In this study, we investigate the gettering process of Fe in p-type Cz silicon after iron has been introduced at the solubility limit at 1000°C. Deep Level Transient Spectroscopy (DLTS) was used to measure [FeB], a fingerprint of [Fei], at the center of samples. The minority carrier diffusion length and lifetime were calculated from Electron Beam Induced Current (EBIC) measurements. The fact that [FeB] is proportional to the negative second power of the minority carrier diffusion length at the high [FeB] regime confirms that FeB donors are the dominant recombination centers limiting solar cell performance with high Fe contamination. By quenching after heat treatment, we can maintain and measure the kinetics and thermodynamics of gettering exclusively. The getter/silicon interface was studied by comparison of the gettering rates of molten Al at 620°C, 700°C, and 800°C, and iron silicide at 700°C. We model Fe gettering with respect to temperature, time, solubility and precipitate nuclei density. In the early stage of Fe gettering, the process is dominated by precipitate formation around oxygen precipitate nuclei. The precipitate density is estimated to be on the order of 5×108cm−3. In later stages, Fe outdiffusion contributes to the [Fei] reduction. The early stage precipitation limits [Fei] reduction after short time to the solubility at the gettering temperature.

Effective methods of gettering metallic impurities were proposed. To achieve effective gettering, an annealing process to induce gettering was modified taking fundamental gettering steps and the difference in the gettering mechanism into account. As the concentration of heavy metal is below solubility in a state-of-the-art clean room, a combination of segregation type gettering and slow cooling heat treatment is an effective technique to remove metal impurities from the device active region. Using this technique, DRAM device characteristics such as leakage current and data retention time can be improved.

A Si wafer is contaminated with 1.5 × 1013 Fe/cm3 and the Fe gettering ability of low dose I/I defects is compared with that of a Si substrate with/without a poly-silicon back seal. Polysilicon has higher gettering ability than I/I defects and prevents Fe from gettering at these defects. When there is no poly-silicon and temperature is as low as 700°C, however, I/I defects act as gettering sites for Fe, even if the dose is as low as 2 × 1013 cm−2. Fe gettering by I/I defects leads to a decrease in minority carrier diffusion length. DLTS measurement reveals that Fe getters at I/I defects in the interstitial atom state.

Dark current spectroscopy (DCS) is used to identify the signature of metals that generate dark or leakage current in silicon image sensors. Individual metal atoms or defects are detected by DCS on a pixel-by-pixel basis. DCS is applied here to show how the number of electrically active iron atoms in a pixel changes with light and with low-temperature anneals. The measurements explore the dissociation and association of iron-boron pairs and the diffusion of iron near room temperature.

We have demonstrated the influence of surface microroughness on the electrical characteristics of MOS devices and investigated the influence of wafer's manufacturing methods, such as Czochralski(Cz), floating-zone(FZ), and epitaxial(Epi) silicon wafers, on the susceptibility to the surface microroughness when some chemical treatment was performed. As a result, it was found that Cz and FZ wafers are very susceptible to the surface microroughness and the amount of the vacancy of Epi wafer is much smaller than that of another wafers. It was also demonstrated that the electrical characteristics of very thin gate oxide films are strongly influenced by the silicon substrate quality. Epi wafer is a strong candidate for fablication of highly-reliable devices on 300mm wafers.

Micro-Raman spectroscopy and chemical etching were applied to determine the depth of subsurface damage in silicon wafers undergoing different machining processes: cutting, grinding, polishing and lapping. In comparison with the Raman spectrum of perfect single crystal silicon, both the shape and intensity at the shoulder (500 cm−1) and the subpeak (300 cm−1) spectral regions were changed in all the machined wafers. The intensities at shoulder and subpeak gradually decreased and finally resumed to normal, as the depth of the investigated layer increased. According to the chemical etch rate, the depth of the subsurface damage was thus evaluated for the different wafers. TEM observations further confirmed the obtained results.

Formation of shallow donors has been studied in 100-keV hydrogen-implanted n-type silicon with a dose of 1×1015 cm−2. Annealing experiments are made in the temperature range 100 to 800°C. Gold Schottky contacts are fabricated on these samples to obtain carrier concentration profiles. The hydrogen profiles are measured by secondary ion mass spectroscopy. In the annealing temperature range 350 to 450°C, an increase of the carrier concentration is observed near the surface region with almost flat depth profiles. The well- known hydrogen-related donors (1st shallow donors) disappear around 400°C, while shallow donors (2nd shallow donors)observed here decrease above 500 °C. Hydrogen profiles show almost no change in the 2nd donor formation temperatures, although hydrogen starts to outdiffuse at 250°C. This indicates that the 2nd shallow donors are related to hydrogen atoms and to implantation-induced damage. The implantation-induced defects trap diffusing hydrogen atoms, which leads to the formation of the 2nd shallow donors.

With ab initio calculations we show that the experimentally observed large segregation energies of As at Si grain boundaries can be explained by the formation of isolated dimers or ordered chains of dimers of threefold-coordinated As along the cores of grain boundary dislocations. We also find the intriguing possibility that As segregation may drive structural transformation of certain grain boundaries. Recently we have obtained the first atomic-resolution STEM images of As in a Si grain boundary, consistent with the formation of As dimers. Segregation energy of As dimers was found to be significantly higher in isolated dislocation cores, where larger site-variation in strain than in grain boundaries lead to further lowering of the electronic levels of As deep into the bandgap.

An overview is presented on mechanisms of electronically induced (enhanced) defect reactions in semiconductors, which are classified into the local heating, the structural instability, and the recombination enhanced. A mechanism for the annihilation of a hydrogen-carbon complex in silicon is given as an example of the second one. The last two mechanisms can be treated in a unified scheme by using the correct configuration coordinate diagram, which enable us to treat correctly the correlation in successive captures of an electron and a hole. The energy conversion mechanism during the reaction is precisely discussed paying attention to the relation between the lattice relaxation mode and the symmetry breaking reaction coordinate.

The interaction between self-interstitials (Sii), impurities, and dopants in Si leads the formation of undesirable point defects which affect device operation. Electron beam irradiation has been used to generate Sii and initiate defect reactions, and the hierarchy and competition of interstitial defect reactions involving O, C, B, and P in Si have been explored by DLTS measurements. We describe the interstitial defect reactions as a three-step process: (i) displacement reaction for the generation of Sii, (ii) Watkins replacement reaction for the generation of C and B interstitials (Ci and Bi), and (iii) diffusion limited reaction for the formation of pairs. Within the framework of reaction kinetics, for the first time, we have successfully set up a nonlinear system model to simulate the reaction processes. The interstitial migration enthalpy and the pair formation capture radius are two parameters used in the model to describe long range migration and near neighbor interaction. The good agreement between the model and experiments not only supports the defect assignments by DLTS, but also provides an initial glimpse into the interaction of point defects in Si.

Dependence of the donor center concentration, the Er activation coefficient and the nlayer thickness on the implantation and subsequent annealing conditions was studied. A model of formation of donor centers has been developed. The model is based on the interaction of Er atoms with self-interstitials. It describes the dependence of electrical parameters of Er-doped layers on temperature and time of postimplantation annealing.

Consideration has been given to production processes of shallow donor centers formed in silicon after implantation of erbium ions or co-implantation of erbium and oxygen ions followed by annealing at 700° and 900°C. Analysis of the experimental data obtained in this work made it possible to put forward some suggestions concerning the nature of these defects.

A systematic study of defects in the Er/O implanted silicon was conducted using TEM, HRTEM and SIMS. Defect-free material was obtained after the annealing of 400 keV Er+ implanted (100)Si. In sharp contrast, several forms of secondary defects consisting of dislocations, dislocation loops and precipitates were induced upon annealing at different temperatures in the 4.5 MeV implanted (100)Si sample. The isothermal evolution of the defects and reactions between dopants and defects were studied. Oxygen tends to segregate into the dislocation loop zones, where platelet precipitates with habit planes of {111} were found. Following dissociation of oxygen and erbium, plate-like Er precipitates were generated, which are most likely ErSi2 with a habit plane of {111}

A new stable structure and the electronic structure of boron clusters in silicon crystal have been calculated using the first-principles local density functional approach for Si54B12H60 clusters. According to our calculation, the cubo-octahedral B12 cluster was found to be more stable than the icosahedral one proposed previously. The total energy difference was about 4.6 eV. The analysis of the partial density of states showed that the cubo-octahedral B12 cluster should act as a double acceptor.

Interactions between structural defects and metallic impurities were studied in multicrystalline silicon for solar cell applications. The objective was to gain insight into the relationship between solar cell processing, metallic impurity behavior and the resultant effect on material/device performance. With an intense synchrotron x-ray source, high sensitivity x-ray fluorescence measurements were utilized to determine impurity distributions with a spatial resolution of ≈ lμm. Diffusion length mapping and final solar cell characteristics gauged material/device performance. The materials were tested in both the as-grown state and after full solar cell processing. Iron and nickel metal impurities were located at structural defects in asgrown material, while after solar cell processing, both impurities were still observed in low performance regions. These results indicate that multicrystalline silicon solar cell performance is directly related to metal impurities which are not completely removed during typical processing treatments. A discussion of possible mechanisms for this incomplete removal is presented.

Hydrogen molecules have been formed in crystalline silicon at various temperatures by a hydrogen-atom remote treatment. The Raman spectrum of the vibrational lines of hydrogen molecules in crystalline silicon is detected for silicon samples treated at temperatures between 250 and 500° C. The maximum production is obtained at 400° C. The Raman spectrum of hydrogen molecules in silicon observed at room temperature exhibits a frequency shift of around 4158 cm−1 and a very broad half-width of approximately 34 cm−1. Isotope shift also can be observed at around 2990 cm−1 in silicon treated with deuterium atoms at 400° C. The frequency shifts of the observed lines are in close agreement with those reported for molecular hydrogen and deuterium in gas, liquid, and solid phases. We discuss a model for the hydrogen molecule configuration and rule out the possibility of high-pressure hydrogen molecular gas in microvoids in crystalline silicon. These results indicate that hydrogen molecules exist at the tetrahedral interstitial sites in crystalline silicon.

Several vibrational bands have been observed in the region 1750–1900 cm−1 for n-type silicon into which gold and hydrogen had been introduced by high temperature diffusion. This paper focuses primarily on a defect which gives rise to bands at 1813.3 and 1827.1 cm−1. This center contains a single hydrogen atom and has two charge states in the gap with trigonal symmetry. We propose a tentative model in which this complex contains a substitutional Au atom and a single hydrogen atom at one of two possible antibonding sites, bonded either to Au or to Si. A defect observed previously in Si samples doped with Pt and H which gives rise to hydrogen stretching bands at 1880.7 and 1897.2 cm−1 has similar properties to the Au-H complex discussed here, suggesting that these defects have the same structure. Other hydrogenstretching modes observed in Si samples that contain Au and H are also discussed.