Within the past decade, process simulation has become an essential part of new technology development in the silicon IC industry. The use of TCAD (technology computer aided design) tools has been driven by the enormous cost of purely experimental approaches to technology development. Yet the power of these tools and their predictive capability are still greatly limited by the models they use. TCAD models for doping processes are universally based today on point defects. These models have evolved considerably in the past decade to incorporate additional understanding. The state-of-the-art today includes concentration dependent diffusion through Fermi level effects on defect concentrations, full coupling between defects and dopants which allows prediction of non-local diffusion effects, basic models for the effects of ion implantation damage (the +1 model), surface and interface effects (through effective recombination velocities and segregation), and full 2D and 3D simulations.

As devices continue to shrink, better models will certainly be required. Challenges for the future include more detailed information about damage resulting from ion implantation, better understanding of point defect properties (equilibrium populations, diffusivities, transient response to temperatures changes), better models for point defect behavior at interfaces, and finally, development of accurate methods to actually measure 2D and 3D dopant profiles. This paper will attempt to describe where we are and where we need to be in the future.

The status of our knowledge on intrinsic point defects and diffusion mechanisms is reviewed. Special attention is given to the question of the possible role of carbon in influencing effective diffusivities of intrinsic point defects and the resulting consequences for the values of vacancy and self-interstitial thermal equilibrium concentrations and diffusivities. It is pointed out that we might have to deal with the unfortunate situation that the effective diffusivities of intrinsic point defects might be influenced already by a carbon concentration which is below the detection limit and therefore not amenable to measurement and control. Whereas dopant diffusion processes have been modeled and simulated for a long time the first attempts to quantitatively simulate various gettering processes have just started as will be described in the paper. Finally, the subject of microcrack formation in hydrogen implanted silicon will be dealt with as used for the so-called “smart-cut” process for fabricating silicon-on-insulator (SOI) substrates with thin and uniform silicon layers by wafer bonding. Presently no quantitative treatment of this fascinating hydrogen agglomeration phenomenon is available.

Diffusion of Au into dislocation-free and highly dislocated Si with high B-background doping levels has been investigated with the aid of neutron activation analysis in conjunction with mechanical sectioning. The high B-doping level causes extrinsic conditions, i.e., the hole concentration exceeds the intrinsic carrier concentration even at diffusion temperatures between 900°C and 1100°C. All profiles are accurately described on the basis of the kick-out diffusion model and a mechanism which takes into account segregation of Au at dislocations. Our analysis provides solubility data of Au in Si and effective diffusion coefficients related to interstitial Au and Si self-interstitials I. The dependence of these quantities on the B-background doping level is well described by the Fermi-level effect. This analysis supports singly positively charged states in p-type Si of Au on interstitial (Aui) and substitutional (Aus) sites and of Si self-interstitials. Successful fitting of additionally requires an acceptor level of Aus. The electrical properties deduced for Aui, Aus and I are summarized in Table 2. Au profiles in highly dislocated Si obtained especially after diffusion at 900° C give evidence of Au trapped at dislocations. From our preliminary experimental results we determine an enthalpy difference of 2.7 eV between Au on substitutional sites and Au captured at dislocations.

The thermodynamics of diffusion under hydrostatic pressure and nonhydrostatic stress is developed for single crystals free of extended defects and is applied to the case of boron diffusion in silicon. The thermodynamic relationships obtained permit the direct comparison of hydrostatic and biaxial stress experiments and of atomistic calculations under hydrostatic stress. Assuming various values for the anisotropy in the migration strain, a currently unknown parameter, comparison is made between various measurements under hydrostatic pressure and nonhydrostatic stress, and various atomistic calculations of the volumetrics of B and Si diffusion by an interstitial-based mechanism. An independent determination of the anisotropy of the migration strain would permit a parameter-free determination of the predominant diffusion mechanism and would permit the prediction of the ratio of the diffusivity normal to the free surface to the diffusivity parallel to the surface for biaxially strained films. Procedures for measuring and calculating the anisotropy in the migration strain are described.

Point defect properties, including diffusivities and equilibrium concentrations for both interstitials and vacancies, are commonly extracted from metal diffusion experiments, and these values are widely used in process simulation software. However, in many cases, these parameter values were extracted using oversimplified models which ignore interactions between interstitial and vacancy diffusion mechanisms. Questions about the accuracy of these parameters have come from ab-initio defect calculations which conclude that vacancies diffuse faster than interstitials, in contrast with published reports on metal diffusion which find vacancies diffuse much more slowly than interstitials. We have reanalyzed published data for zinc and platinum diffusion and find that it is possible to match all of the data using fast vacancy diffusivity. The most direct evidence for slow vacancy diffusion (and a high equilibrium concentration) comes from platinum diffusion experiments. However, we are able to reproduce these results with fast V diffusion and carbon/interstitial clustering, using carbon concentrations typical of Czochralski and float zone silicon (1016cm−3). We evaluate the effectiveness of metal diffusion experiments in determining point defect parameters, and find that it is not possible to reliably determine both diffusivities and equilibrium concentrations for both interstitials and vacancies from metal diffusion results.

Room temperature migration and clustering behaviour of implanted boron into silicon has been investigated by performing ion implantation of the 11B isotope into MBE-grown in-situ 10B-doped epitaxial Si-layers. We, for the first time, show that a fraction of the implanted boron migrates very deep into the bulk of the Si with substitutional 10B acting as trap centers for the migrating 11B.

Ultra low energy boron implants (0.2 to 3 keV) have been carried out on Si (100) at doses between 1×1014cm−2 and 1×1015cm−2 using xRLEAP. The samples were annealed at temperatures between 900°C and 1050°C. The atomic profiles of these samples was measured using SIMS. Monte Carlo and diffusion simulations were performed using the SSupreme code. Comparisons between the simulations and experimental measurements show interesting differences these are discussed.

The substitutional gold concentration introduced by a diffusion step between 850 and 1000°C was measured by Deep Level Transient Spectroscopy (DLTS) both in FZ and Cz silicon containing different dislocation densities introduced by cantilever bending. The comparison, in the same sample, of dislocated and undislocated regions allows first the self interstitial (Sii) effective diffusivity and then the efficiency of dislocations as sinks for self-interstitials γto be measured. In FZ silicon, γ is quite independent of temperature whereas in Cz Si a remarkable temperature dependence was observed, with an effective activation energy of leV, which can be attributed to the release of dislocations by a thermally stimulated climbing mechanism from obstacles (oxygen segregation or precipitation). Increasing the gold diffusion annealing times for a given temperature (850°C) underlines once more the role of the oxygen precipitation in the samples.

Some of our most important and quantitative insights on intrinsic point defects have been established as a result of the study of the influence of oxidation on their “excess” concentrations. From study of the growth and shrinkage of oxidation induced stacking faults (OSF) and from oxidation enhanced or retarded diffusion (OED-ORD) the fractional contribution of self-interstitials and vacancies can be assessed. The dominant influence on relatively thin dry or wet oxides is that of the injection of self-interstitials away from the oxidizing interface into the bulk crystal. These self-interstitials have been variously attributed to the incompleteness of oxidation or interracial stress and relaxation at the Si-SiO2 interface. We have studied the role of fluorine additives (ppm) to the dry oxidation process. This paper will discuss similar oxidation-related phenomena (OSF, OED, and ORD), confirming that under our experimental conditions with normally dry oxidation plus fluorine the active point defect injection consists mainly of “excess” vacancies. OSF anneal at lower temperatures than with nitridation processes and with a lower activation energy (1.7 eV). The study of P,B,As, Sb oxidation enhances-retarded diffusions validates the conclusion, with P,B yielding ORD instead of OED and Sb being enhanced rather than retarded as in normal oxidation. Oxidation-induced diffusion experiments not only explored the role of the specific impurity (P,B,As,Sb) but also utilized specimens with four adjacent active regions (bare Si, oxide covered Si, nitride covered Si, and the normal control oxide covered by nitride). The ratios of Ci/Ci* and Cv/Cv* for fluorine addition will be contrasted with those formed by dry oxidation.

We report on the use of lithium ion (Li+) drifting1 as a sensitive means to study Si self-interstitial (SiI) diffusion.2 Li+ properties in silicon are well known from extensive ion drift studies and Li+ interactions with dopants and point defects.3 We have used this low temperature (∼100°C) technique in combination with Si1 injection from oxides to delineate, identify and eliminate D defects4 in certain p-type floating zone (FZ) Si single crystals.5 Our results suggest Si1 diffusion occurs to a depth of at least 10 mm into the bulk during phosphorus (P) diffusion with oxidation (i.e., POCI3 process) at 950°C for 100 min. Process modeling of this lower bound SiI diffusion using SUPREM-IV9 results in a Sii diffusivity of 3.5×10−6 cm2/s at 950°C.

The continued decrease in critical dimensions and increasing integration levels in Si CMOS technology is imposing ever tighter constraints on quality control parameters for the IC manufacturing industry. One very important issue is the need to ensure a uniform, high quality Si substrate, i.e. minimise defect/dislocation densities and eliminate strain distributions in the starting wafer material. A comprehensive Synchrotron X-Ray Topography (SXRT) study was applied to commercially supplied 200mm diameter Si wafers. These wafers, which all included a surface Si epilayer growth were supplied from manufacturers from around the globe. The study revealed not only differences in the overall quality of the wafers, but also differences in the quality of the individual Silicon epilayers and substrates. In all wafers the substrate quality varied dramatically with position across the wafer, as measured by the distribution of oxygen precipitates and stacking faults in the wafer. This distribution also varied significantly from manufacturer to manufacturer. The strain fields induced by the growth of lightly doped Si epilayers were also observed to qualitatively vary with location on a wafer, together with (as expected) thickness of the epilayers. The results clearly indicate that optimal quality control for such commercial wafers has not yet been achieved.

Microdefects in multicrystalline silicon grown by directional solidification have been investigated by transmission electron microscopy. Their density (=106 cm−2) correlates with the density of shallow etch pits observed after chemomechanical polishing and selective etching. Different types of microdefects (size 10 – 100 nm) could be identified: i) spherical precipitates most likely amorphous silicon dioxide, ii) small plates lying on {111} planes and iii) groups of closely spaced stacking faults having the character of dipoles. It is argued that these defects are the result of agglomeration process of intrinsic point defects and impurities, where oxygen and carbon are the main candidates. A qualitative comparison to the point defects agglomeration observed in Cz material will be given.

The effect of oxygen gettering by buried defect layers at post-implantation annealing of hydrogen implanted Czochralski (Cz) grown silicon has been investigated. Hydrogen ions were implanted with an energy of 180 keV and doses of 2.7.1016cm−2 into p-type Cz and for comparison into p-type float zone (Fz) Si. The samples were annealed at temperatures between 400 °C and 1200 °C in a forming gas ambient and examined by secondary ion mass spectrometry (SIMS) in order to measure the hydrogen and oxygen concentration profiles. Spreading resistance probe (SRP) measurements were used to obtain depth resolved profiles of the resistivity. The observed changes of the resistivity after post-implantation annealing of hydrogen implanted Cz and Fz Si can be explained by hydrogen enhanced thermal donor formation processes (oxygen or hydrogen related) and charges at the SiOx precipitates. The effective oxygen gettering in hydrogen implanted Cz silicon is attributed to hydrogen enhanced diffusion of oxygen to buried defect layers.

Our experiments show a hydrogen plasma assisted creation of p-n junctions in p-type Cz silicon due to a hydrogen enhanced thermal donor (TD) formation at temperatures ≤450 °C. Applying DC or HF plasma treatments a conversion of p-type into n-type Cz silicon by TD formation occurs. One can distinguish one step processes (p-n junction formation appears just after the plasma exposure) and two step processes (p-n junction formation requires subsequent post hydrogenation annealing). The samples are studied by depth resolved spreading resistance probe (SRP), capacitance-voltage (CV) and Hall measurements. For the one step processes a kinetic model for hydrogen enhanced TD formation is presented.

Kinetic aspects of thermal donor (TD) formation in Czochralski silicon are shown to be consistent with the evolution of small oxygen clusters, as described within the classical theory of nucleation. Predictions for TD generation and interstitial oxygen loss are presented. Favorable agreement with experimental data requires that the rate constants describing cluster evolution be increased over those expected for a oliffusion-limited flux based on a normal diffusion coefficient for oxygen in silicon. This may signal an anomalously high diffusion rate for temperatures less than 500°C, as has been suggested by others. However, it may instead signal an enhanced concentration of free oxygen near clusters smaller than the critical size for nucleation. This is expected when the interfacial attachment rates become comparable with the rates at which oxygen atoms arrive in the vicinity of the sub-critical clusters. The link between thermal donor generation and oxygen precipitation processes demonstrated here provides a consistent framework for better understanding and controlling oxygen precipitation in silicon. Further, the kinetic TD generation and oxygen loss data provide a new window into the dynamical processes for small clusters, which underlie all nucleation phenomena.

We study the mechanisms of damage production during ion irradiation using molecular dynamics simulations of 400 eV -10 keV collision cascades in four different materials. The materials Al, Si, Cu and Ge are contrasted to each other with respect to the mass, melting temperature and crystal structure. The results show that the crystal structure clearly has the strongest effect on the nature of the damage produced, and elucidate how the open crystal structure affects the nature of defects produced in silicon.

Thermal cycles in advanced CMOS processing can nucleate an annular ring of oxygen precipitate-induced stacking faults (OSF-ring) via activation of bulk nuclei grown-in during the crystal pulling process. Because the OSF-ring can adversely affect device characteristics, it is important that substrates with OSF-ring characteristics be detected early in the process. Results are presented in this paper from a typical DRAM device which show that the ring can act either in a beneficial gettering mode or as a device-degrading zone, depending on the depth distribution of the OSF-ring defects and the background iron impurity concentration.

Silicon-On-Insulator (SOI) materials have emerged as a very promising technology for the fabrication of high performance integrated circuits since they offer significant improvement to device performance. Thin silicon layers of good crystalline quality are now widely available on buried oxide layers of various thicknesses with good insulating properties. However, the SOI structure is quite different from that of bulk silicon. This paper will discuss a study of point-defect diffusion and recombination in thin silicon layers during high temperature annealing treatment through the investigation of stacking-fault growth kinetics. The use of capping layers such as nitride, thin thermal oxide and thick deposited oxide outlines the diffusion mechanisms of interstitials in the SOI structure. It also shows that the buried oxide layer is a very good barrier to the diffusion of point defects and that excess silicon interstitials may be reincorporated at the top interface with the thermal oxide through the formation of SiO species. Finally, from the experimental values of the activation energies for the growth and the shrinkage of stacking-faults, the energy of interstitial creation is evaluated to be 2.6 eV, the energy for interstitial migration to be 1.8 eV and the energy of interstitial generation during oxidation to be 0.2 eV.

The influence of N2O oxidation of silicon on the kinetics of point defects at high temperatures is investigated. Oxidation Stacking Faults (OSF) are used to monitor the interstitials that are generated during the oxidation process. We show that at high temperatures (1050°-1150°C) the supersaturation of self-interstitials in the silicon substrate is enhanced when oxidation is performed in an N2O ambient compared to 100% dry oxidation. This behavior is attributed to the presence of nitrogen at the oxidizing interface. However, at lower temperatures this phenomenon is reversed and oxidation in N2O ambient leads to reduced supersaturation ratios.

A review is given of what has been learned from EPR studies over the last ∼35 years concerning vacancies and interstitials in silicon. Lattice vacancies are well understood, their diffusional migration energies determined vs charge state and electronic excitation, and their interactions with most of the common impurities established. The isolated interstitial has not been observed by EPR but a great deal has been learned concerning it from studies of its interactions with impurities, and more recently from theory. The properties of these two intrinsic defects and their progeny will be analyzed to help clarify their role in both the normal thermally activated diffusion processes and the transient-enhanced ones, which are of particular current concern.