This paper reviews the basis for phenomenological path for gettering: release, diffusionand capture in the context of old and new developments. More recent precipitation studies in silicon are now recognized to involve complex interactive effects. Controlled precipitation requires knowledge of and control of not only oxygen but carbon, native defects(related to crystal thermal history and/or process history), high doping, microfluctuations and interactions between “extrinsic” and “intrinsic” gettering.

Intrinsic gettering of iron in silicon has been studied with a novel approach that involved the combination of two complementary techniques. Electron Paramagnetic Resonance (EPR) was used to determine the concentration of interstitial iron atoms which are electrically active in the silicon matrix. A systematic comparison of the effect of various oxygen precipitates of well defined morphologies on the solubility of iron was performed. A surprising result was found for bulk samples containing oxygen precipitates obtained after low temperature heat treatments (700ºC): the solubility of iron was reduced byalmost two orders of magnitude from the values obtained in untreated Cz silicon samples or in samples containing oxygen precipitates obtained at high temperature (1150ºC). The practical importance of this new result was demonstrated by a series of iron concentration measurements after various gettering and high temperature re-heating treatments.

Transmission Electron Microscopy (TEM) was used in order to confirm the proposed mechanism of formation of a new phase, different from FeSi2 in the silicon matrix. This new phase is assumed to limit the possible electrically active metallic content by acting as a boundary phase in quasi-equilibrium with interstitial iron atoms. Evidence ofthe formation of Fe2SiO4 was found in wafers that had been oxygenimplanted and subsequently subjected to similar precipitation and diffusion treatments than those performed with the bulk samples.

These findings suggest the possibility of dramatic improvements in the efficiency and the stability of the gettering process.

We have recently found that a very effective intrinsic gettering of metal contaminants in silicon can be realized via interaction of native defects. This process makes it possible to extend intrinsic gettering of crystals with low or virtually zero oxygen concentration. Transmission electron microscopy and energy dispersive x-ray spectroscopy were used to identify the microscopic structure of the gettering centers in oxygen-free silicon. These centers exhibited butterfly-type shape and they consisted of the interstitial type extended dislocation loops of 0.1 to 0.7 μm diameter. Small precipitates (3-15 nm diameter) of metal (Cu, Ni, Fe) silicides were found inside the dislocation loops and/or on the dislocation lines. The origin of butterfly-type gettering centers will be discussed in terms of the intrinsic gettering model involving lattice supersaturation with silicon self-interstitials as a driving force for their complexing and subsequent formation of dislocation loops and precipitates which immobilize metal contaminants.

The new intrinsic gettering mechanism based on the interactions of silicon interstitialshas been studied in float zone silicon with virtually zero oxygen content. Silicon wafers were subjected to the high-low-medium temperature annealing cycle, with the high temperature step carried out subsequently in an inert ambient, followed by the dry 02, followed by an inert ambient. The annealing caused formation of a multilevel structure consisting of two defect free regions with a high lifetime value, separated by a layer with high defect density and a low lifetime value. The experimental data corresponds to the theoretical calculations describing silicon interstitial injection during dry 02 oxidation and their diffusion during annealing in an inert ambient.

The effectiveness of CVD thin film backside gettering on n-type CZ silicon wafers for CMOS technology has been investigated using optical techniques for bulk microdefect analyses and transmission electron microscopy for interfacial structure study. The deposition of LPCVD polysilicon (500-2000 nm), silicon nitride (150 nm), or poly/nitride films on the backside of Si wafers was found to enhance the bulk precipitation. Bulk microdefect density increased as the thickness of polysilicon increased. At the polysilicon/silicon interface, no extended line defects from polysilicon were observed. Based on the results of minority carrier lifetime and oxide breakdown measurements, the best gettering efficiency was given by 1300 nm polysilicon backside gettering scheme.

The effect of HCl in an oxidizing ambient on Si interstitial formation during oxidation has been studied by the buried marker diffusion technique. Adding HCl reduces the self-interstitial flux generated at the oxidation front but it does not completely eliminate it. A uniform blocking layer model predicts fairly well the Cl effect on self-interstitial generation during the thermal oxidation. By reducing the rigidity of the SiO2, Cl incorporation into the SiO2 is strongly proposed to alter the Si interstitial partition coefficient at the interface.

Synchrotron section topography is applied to the study of silicon wafers pulled out froma simulated advanced CMOS twin-tub process. Substrates are heavily doped with antimony and phosphorus. For comparison also high-resistivity samples are studied. Prior to epi deposition of arsenic doped layers some wafers are subjected to a three-step intrinsic gettering cycle.

Section topographs show that in the lightly doped samples a denuded zone is formed by the CMOS process itself in contrast with the heavily doped ones in which the intrinsic gettering is needed.

Capacitance DLTS measurements have been performed on heavily precipitated n— and p—type silicon wafers. The results indicate heavy metal gettering with a mid-bandgap deep level (0.55eV) for n—type silicon. The results for p—type siliconshow a band of states present in the lower half of the bandgap. This band of states correlates well to the band of allowed energies found in heavily dislocated p—type silicon.

One of the principal difficulties in applying the Zerbst analysis to MOS capacitor transient recovery is the extremely long retention time of the MOS capacitor in high lifetimematerials. Typically the retention time is thousands of seconds at room temperature, so long times are required for a significant number of measurements. Furthermore, extrapolating the result of high temperature measurements back to room temperature can lead to large errors. A modified C—t transient response using a step from accumulation to depletion followed by light and then a small step from inversion to stronger inversion speeds up the generation lifetime measurement and separates the surface generation from the bulk generation. Pre— and post—epitaxial intrinsically gettered p/p+ wafers were subjected to high temperature CMOS process simulation and MOScapacitors with Al dots and guard rings were fabricated. Generation lifetime of about 5 msec were measured using the modified fast Zerbst method on properly intrinsically gettered epitaxial wafers, particularly the pre—epitaxial intrinsically gettered wafers.

In an attempt to decide the question whether enhanced oxygen diffusion is important for heat-treatments of silicon at ∼450ºC where thermal donors are formed we have conducted two types of experiments aimed at providing a measure of the “effective” oxygen diffusivity. First, we have extensively measured the temperature dependence of the thermal donor introduction rate for very short heat treatment times (20min). This measurement provides the thermal activation energy of TD formation. Since effects of long range diffusion and formation of large oxygen clusters are negligible for suchtimes and temperatures and, presumably, thermal donor formation at the lowest heat treatment temperatures is oxygen diffusion limited, it should be possible to interprete the obtained activation energy in terms of oxygen diffusivity. The change of the interstitialoxygen content is immeasureable for 20min heat treatment times. Therefore, the decay of the interstitial oxygen content was measured for longer heat treatments at 450ºC (up to 500hours). The two experiments are complementary in several ways: In one experiment the oxygen diffusion activation energy is extracted, while the other measurement provides the value of the diffusion coefficient at a given temperature. In one case thermal donors are monitored for short heat treatment times while in the other experiment the interstitial oxygen content is measured for long heat treatment times. The present measurements are different from other diffusion experiments in this temperature range where theatomic jump of isolated oxygen is monitored [1]. Here we attempt to extract an effective oxygen diffusivity under conditions of thermal donor formation since the thermal donor formation process itself might be the cause of an enhanced oxygen diffusivity.

Dislocations extending through channel regions of NMOSFETs with submicron gate lengths have been observed. Additionally, arrays of dislocations in the field regions of IC's have also been observed and found to be generating from the corners of the first n+ region during subsequent high temperature process.

The problems associated with two-dimensional diffusion of impurities coupled with point defects are outlined. The relative importance of the generation, diffusion and surface recombination rates are examined and the implications for 2-D modelling are discussed. Experimental results for the diffusion coefficient of interstitials indicate that the point defects can have orders of magnitude faster diffusivity than dopants. For accurate 2-Dprofiles, a large simulation space is needed to solve for the defects even though the impurity diffusion of interest occurs in the near surface region. The lateral extent of the diffusion profile is determined by the surface recombination rates of the defects. Results indicate that different technologically important interfaces, such as nitride and pad oxide have different recombination rates. The time dependence of the lateral decay isexperimentally investigated and indicates a steady state condition being reached. Parameters are extracted for the 2-D process simulator SUPREM-IV and the computer simulations are compared with experiment.

DLTS was applied to p—type polycrystalline silicon, grown by a casting technique to form ingots with a nominal doping level of ∼1016 acceptors/cmg. Both Schottky diodes and n+p mesa structures were used for the measurements. Very complex DLTS spectra were obtained from diodes that contained electrically active grain boundaries, whereas no traps were detected in areas that did not contain electrically active grain boundaries. Several electron and hole traps were resolved.

Composition graded buffer layers of group Ila fluorides allow the heteroepitaxial growthof device quality narrow gap lead chalcogenides onto Si. Mechanical stresses in the layers are almost completely relaxed at room temperature despite large thermal expansion mismatches. Photovoltaic infrared sensors with up to about 9.5 um cut—off wavelengths and which operate at or near the 300K background noise limit have been fabricated in such PbTe and (Pb,Sn)Se on Si structures.

Furthermore, epitaxial graded fluoride buffers seem to be suited to connect other semiconductors with even large lattice mismatches. Initial heteroepitaxial growth of CdTe on fluoride/Si(lll) substrates (mismatch 20%) supports such more general applications.

The growth of epitaxial insulators on semiconductors in general and on Si in particular is an area of research which has developed quickly over the last five years. Most of thecurrent interest is in films of materials which can be grown by molecular beam epitaxy. A significant fraction of all of the work in this area has concerned the CaF2/Si system. This review will cover that body of work.

SiGe/Si superlattices were grown using limited reaction processing in a chamber which allows both W-halogen and Hg arc wafer illumination. Each multilayer structure was fabricated in-situ by changing the gas composition between high temperature cycles. Commensurate SiGe alloy layers as thin as 15 nm were reproducibly deposited and were examined using transmission electron microscopy, sputtering Auger electron spectroscopy,and Rutherford backscattering. Preliminary results are presented on UV/ozone cleaning of LRP substrates to remove residual carbon contamination in-situ prior to film deposition.

Sheet resistance mapping has become an indispensable tool in characterizing ion implanters for both integrated circuit manufacturers and equipment manufacturers. The sheet resistance mapping technique is now being extended into additional applications such as the characterization of metal deposition, CVD, and epitaxial silicon growth. This technique has become especially necessary with the advent of 150mm and 200mm wafers, where 5 or 9 site measurements cannot provide sufficient data essential for process control.

In order to optimize the performance of an epi reactor it is necessary to control and characterize the gas flows and temperature distributions inside the reactor. The control of these variables is essential for thickness and resistivity uniformity in epi layers. This paper describes the use of sheet resistance profiles and contour maps to study the resistivity and thickness uniformity variations in an epi reactor. The sheet resistance maps allow for control of the epi process without requiring data from other test sources.This ensures real time process control for production, as well as very rapid feedback for maintenance while doing equipment repair.

A new technique for silicon epitaxial growth has been developed using mercury—sensitized photochemical vapor deposition (photo—CVD). Epitaxial thin layers were grown on (100) Si substrates at 100–300ºC from a gas mixture of Si2H6+AiH2F2+H2 by irradiation of a low pressure mercury lamp (1849A, 2537A). Refiective high energy electron diffraction (RHEED) and Raman scattering measurements showed that the epitaxial layers had good crystallinities.The epitaxial layers were characterized by secondary ion mass spectroscopy (SIMS) and the van der Pauw Hall2 measurements. The undoped Si layer showed the electron mobility of 520cm2/Vs with a carrier concentration of 3.2xl04cm

CMOS test circuit chips have been fabricated for the purpose of evaluating Si film recrystallized on quartz wafers by using a graphite-strip heater oven equipped with a micro-computer feedback system. Two new methods have been applied to the zone-melting recrystallization technique to produce a highly uniform and grain boundary (GB) free recrystallized Si film. The slant-scanning method was used to control the GB's location in recrystallized Si film of 180μm wide stripes separated by 20μm. The intersticebridging method was used to reduce the (111) texture generation to less than 1%. The average electron mobility and the average threshold voltage for MOSFET's were 960cm2/V sec, with a standard deviation of 43cm2/V sec, and 1.38V with a standard deviation of 0.25V, respectively. For the ring oscillator, at a supply voltage of 12V, the propagation delay time was 1.7nsec per stage. The maximum response frequency was higher than IMHz for the CMOS inverters.

By low pressure vapour phase epitaxy (LPVPE) epitaxial growth down to 760ºC has been achieved in the SiC12H2/H2 system, as revealed by RBS and channeling measurements ( min ∼4.0%). High and relative homogeneous doping was obtained with boron resp. phosphorus in the range 1x1018 -5x1019 cm-3 for T: 800º - 900ºC. Sharp transitions in the boron doping profile to the substrate, of the order of 15mn, result only for temperatures much lower than 850ºC. Schottky barrier enhanced diodes with promising performances were formed with Ti evaporated on p+n layers grown selectively on n+ substrates.