The kinetics of thermal donor formation in Czochralski-silicon at ca. 450° C are explained by a simple model based on the work of Suezawa and Sumino which derives forward and reverse reaction rates for each electrically active species from the general features of the infrared electronic absorption spectra. The model, which is independent of the chemical nature of the thermal donor core, assumes that all thermal donors beyond the first donor species are chemically stable at the donor formation temperature, and approximates the reactions for species smaller than the first thermal donor as being in chemical equilibrium. The model is shown to be consistent with both sets of the available IR spectra of thermal donors (Oeder-Wagner and Suezawa-Sumino) when differences in the annealing temperature and initial oxygen concentration are taken into account.

For thermal donor and oxygen precipitation studies in Cz silicon wafers, extended 450°C thermal annealing (e.g., anneal at 450°C for 64 h) has frequently been used. But, our recent studies have shown that this kind of extended low temperature anealing leads to a low thermal donor (LTD) layer up to ∼ 8 um from the surface in the case of 450°C/64 h thermal annealing. In this study, effects of annealing ambients, protective layers, surface conditions, and bulk oxygen contents upon the LTD layer are discussed. Our data indicate that the LTD layer is a product of both surface- and bulk-related phenomena.

We report the observation of a new shallow donor center detected in CZ Silicon annealed at 600°C, by EPR and luminescence experiments. The binding energy varies between 30 meV and 17 meV with the annealing time, showing the presence of several centers. The creation rate of these new defects is favoured in carbon rich material. This series of defects is probably one of the components of the New Donors.

We report the results of an experiment of annealing Czochralski silicon at 450°C for up to 500 hrs. Concentrations of oxygen atoms (Ci) and thermal donors (TD) have both been monitored. Analyses of the oxygen concentration data yielded the apparent interpretation that the overall kinetics is dominated by the formation of small clusters (dimers and trimers). This cannot account for TD formation, since they are supposed to be larger clusters. On the other hand, analyses of existing TD kinetic models did not yield calculated Ci values to satisfactorily account for the present Ci data. We believe that a satisfactory TD model is not yet available.

Oxygen in silicon precipitates with second order kinetics at the normal diffusion rate at 450° C and there Is evidence for self-Interstitial generation. These findings are discussed In relation to thermal donors.

Ab-initio quantum chemical computations have been applied to a set of molecular clusters derived from Si5 H12 to model defects in crystalline silicon involving boron, carbon, nitrogen, oxygen, and hydrogen. In computations of defect structure, hydrogen atoms terminating silicon valencies are fixed at their computed positions in Si5H12, to represent forces from the lattice, while the position of other atoms are varied.

We have computed the stable bonding structures of boron, carbon, nitrogen and oxygen atoms to a vacancy, as well as interstitial oxygen, the silicon-oxygen ylid and two oxygen atoms bound to a vacancy. The structures of the dipositive ions of the oxygen bearing clusters have been computed as part of a search for candidates for the core of the 450° C oxygen thermal donor in silicon crystal. The computed cluster energies are employed to give an account of defect thermochemistry; the addition of the free atoms to a vacancy, the addition of interstitial oxygen atoms to a vacancy, the reaction of interstitial oxygen atoms to form a vacancy-oxygen complex with the emission of silicon monoxide, and the reaction of interstitial oxygen with the dipositive ion of substitutional oxygen to form the dipositive ion of two oxygen atoms bound to a vacancy.

The understanding of the precipitation phenomena of light non dopant Impurities has been recently improved thanks to high resolution electron microscopy and microanalysis. After a one-step annealing in Czochralski silicon long coesite (SIO2) ribbons are formed between 485° and 750°C; amorphous platelets (SIOx with x =1. 2 to 2) are formed between 650°C -1050°C. Silicon Interstitlals generated during the precipitation partly relax the strain energy associated with the volume change. These Interstltlals are also able to precipitate In various forms. After a two-step annealing both platelets and/or octahedra containing amorphous SIOx are formed. The role of carbon on oxygen precipitation Is important: It changes the nucleation parameters and gives a retardation phenomena In a two-step annealing treatment. Similar phenomena are observed in oxygen implanted silicon. The nucleation and growth process able to explain these observations is far from being well understood. The SIO2 polymorphism, the Important role of SI Interstitials and the mutual attraction between oxygen and carbon are some of the ingredients which explain this complexity.

Evidence for a new microcrystalline precipitate in CZ Si annealed for 256 hrs at 635°C is presented by electron microdiffraction. This may be a precursor phase for the formation of amorphous platelets [9]. Multiple scattering microdiffraction calculations which distinguish the symmetries of two models for the thermal donor are also given.

The method of electron energy loss spectroscopy in a transmission electron microscope is applied to the analysis of small particles in silicon. Elemental and chemical microanalysis is demonstrated on nitrogen and oxygen-associated defects in silicon, using silicon-based standards.

The precipitation of oxygen in silicon exhibits many interesting phenomena: the precipitates are often distributed in striated zones; the morphology of precipitates changes with annealing temperatures; the kinetics of precipitation in the bulk is affected by annealing ambients and other surface treatments. These phenomena are briefly reviewed. A model consistent with all these phenomena is presented. Then, the energetics and the kinetics of nucleation and precipitation are discussed. Finally, a growth law is derived in general for disc precipitates, and applied in particular to oxygen precipitates in silicon.

In this paper we first describe, in some detail, the nature of an exigentaccommodation- volume factor associated with oxygen (Oi) precipitation in Czochralski (CZ) Si. This factor is regarded as a causal or characteristic factor which influences many aspects of the SiO2 precipitate nucleation and growth phenomena. Employing this factor, we then describe the possible explanations of two outstanding features of Oi precipitation in CZ Si: the precipitation retardation/recovery phenomena and the nucleation incubation phenomena.

Three separate “anomalous” effects in the precipitation of oxygen in silicon may be explained if typical poorly-crystallized platelet oxygen precipitates begin as tiny crystalline clusters. The first anomaly, sometimes referred to as the induction effect, may be explained if one postulates the existence of kinetically stable precipitate embryos (seeds) containing no more than one or two oxygen atoms. We show here that such a postulate, coupled with observations, places rather specific constraints on binding energy as a function of size for such tiny clusters. The second and third anomalies, arising in precipitate shape and retrogrowth behavior dependences, respectively, may be explained if one postulates the existence of a relatively dense precursor phase which undergoes first order phase transition, following otherwise classical rules, to the final-stage amorphous oxide normally found. In this case, both precipitate shape and strain field can be interpreted as a barometer of the interstitial ambient during key periods in a precipitate's history.

Rapid thermal annealing and furnace annealing steps have been combined to investigate the effects of thermal donor generation and annihilation upon oxygen precipitation in low and high carbon content silicon wafers. Thermal donors were formed by furnace annealing at 450°C. Rapid thermal annealing was performed in 10 s periods at temperatures between 600° and 1000°C and was followed by two step furnace annealing at 700° and 950 °C. Rapid thermal annealing separates the annealing stage for thermal donor removal from that for removal of oxygen precipitate nuclei, and a marked dependence upon carbon is observed for nuclei stability under RTA. Implications of these observations for models of precipitate nuclei are considered.

The advantages of oxygen precipitation in terms of intrinsic gettering to reduce oxidation induced surface defect densities and improve minority carrier generation lifetimes (τ) are well documented in the literature (1–3). Various gettering cycles to establish oxygen precipitation in n type and p type nondegenerate substrates have been developed (4–6), based on homogeneous nucleation theory. For nondegenerate silicon no differences in oxygen precipitation kinetics related to the dopant type have been reported. However, with the increasing interest in both p/p+ and n/n+ epitaxial layers for CMOS devices (7), work on the precipitation of oxygen in both p+ and n+ degenerate silicon substrates has revealed a dependence of oxygen precipitation kinetics on dopant type (8–11). In the case of p+ silicon the differences in precipitation kinetics are small when compared to p- silicon (10,11). For n+ silicon it has been reported (8–11) that the dopant concentration plays a major role and that a significant retardation of oxygen precipitation is observed. There are currently two different explanations for the difficulty encountered in producing sufficient oxygen precipitation in n+ Si. These are the failure to incorporate sufficient oxygen into the crystal during crystal growth (12) and an interaction between intrinsic point defects and the n type dopant, which influences the oxygen precipitation kinetics (8,13).

We present a numerical model that simulates the evolution of precipitates and the diffusion of interstitial oxygen in Czochralski silicon. The growth and/or dissolution of each precipitate and the local concentration of interstitial oxygen with which the precipitates interact are followed as a function of time. We treat realistic densities of discrete, interacting precipitates and determine how the precipitate density influences the extent of the precipitation. The model also treats oxygen outdiffusion and the formation of precipitate-free or denuded zones. We apply the model to previous experimental data on the time dependence of precipitate growth and to the development of denuded zones during intrinsic gettering.

Czochralski silicon was implanted with oxygen at 0.4 and 3.5MeV to obtain concentrations near 1020 oxygen/cm3 in the implanted region. Following implantation the wafers were aged at about 1000°C for 7 hours, and the resulting precipitates were examined by HREM. A high density of octahedral SiOx precipitates (∼1015/cm3) was the dominant morphology. Plate type precipitates and dislocations were also present at lower density. The data indicate octahedra grow from the plates.

Interfaces between Si substrates and epitaxial Si buffer layers grown by Molecular Beam Epitaxy (MBE) are shown to contain a high density of SiOx pockets for certain sustrate preparation conditions. It is also shown that post-deposition thermal annealing of these structures grown upon Czochralski wafers can lead to a greatly increased defect density at the interface. The primary model proposed for this increase is trapping of background oxygen diffusing from the bulk of the Czochralski substrate wafers.

Computer modelling of the silicon single crystal grown by the Czochralski method has been carried out in order to calculate the temperature isotherms as well as to evaluate the thermal history of the crystal. The aim of the present work is to find out to which extent the thermal history of the crystal can be altered without structural modifications of the single crystal furnace. Furthermore, another aim is to study the influence of external heaters, reflectors and cooling elements to the temperature isotherms as well as to the form of the solid-melt interface. The model developed so far takes into account the crystal diameter and the pull rate as well as the radiation from the heater, the crucible and the melt. Furthermore, the inclusive study of the cooling effect of the inert gas flow is under progress.

Despite the apparent dissimilarities between different gettering methods, we show that many can be understood in terms of two basic mechanisms. The first involves the interaction of selfinterstitials (emitted, for example, by P in-diffusion and precipitation, 0 precipitation, or Ar implantation) with the impurities to be gettered. The second relies on a strain-field/point defect interaction (for example around a dislocation), which appears capable of enhancing the diffusivity of the impurities. In either case, gettering can be viewed as a particular and particularly useful instance of the wider class of defect/defect interactions.

Intrinsic gettering of iron in silicon has been investigated with a novel quantitative approach. Concentrations of electrically active, interstitial iron were determined by Electron Pararnagnetic Resonnance before, and after, various annealing cycles, using Cz silicon with oxygen precipitates of different well defined morphologies. Significant differences in the gettering rate in as-grown Cz and Cz silicon with high temperature oxygen precipitates were found. High temperature treatments after metal gettering resulted in the re-emission of the transition metal atoms for both as-grown and high temperature precipitated Cz silicon. This allowed the conclusion that gettering took place via FeSi2 formation in these cases and did not involve any thermally more stable phases.