During the oxidation of silicon, interstitials are generated at the oxidizing surface and diffuse into the silicon. Boron diffusion was used to map the local interstitial super-saturation, the ratio of interstitial concentration to the equilibrium concentration of interstitials I/I*, versus depth above buried Si0.795Ge0.2C0.005 layers during oxidation. The average interstitial supersaturation at the silicon surface, extrapolated from the depth profiles, is measured as, ∼24 and ∼11.5 for 750°C and 850°C respectively. Using the measured interstitial concentration at the surface, the silicon interstitial injection into the silicon is calculated for oxidation at 750°C and 850°C. Finally, it is found that the surface boundary condition remains fixed over an interstitial injection rate ranging over 4 orders of magnitude.

In this paper, we report for the first time the effect of sacrificial oxide (sacox) on the boron diffusion in ultra-shallow P+/N junctions. It is shown that the boron diffusivity is enhanced when low energy implantations are performed through sacrificial oxide. The various experimental data lead to conclude that the Post-Oxidation Enhanced Diffusion (POED) is due to a « mirror effect » seen by the Si interstitials incoming into the sacox layer. POED occurs even for sacox as thin as 1.5 nm. From a simple model, the reflection coefficient is estimated to be about 100 % for a 2.5 nm-thick sacox.

We have developed a physically based modeling of TED of implanted boron in amorphised Si. The simulation starts with a supersaturation of Si free interstitials located below the amorphous/crystalline interface which, upon annealing, tend to diffuse out or to precipitate in the form of extended defects (clusters, {113}s, dislocation loops). The modeling of the nucleation and growth of these defects is divided into three distinct stages: the nucleation, the “pure growth” and the Ostwald ripening. This system can interact with a surface (characterized by a given recombination velocity for Si interstitials) only after the SPE regrowth is completed. Implementation of this model into a process simulator allows to describe the isothermal and isochronal evolutions of the sizes and of the densities of dislocation loops in agreement with TEM observations. Assuming that boron diffusion is caused by the concomitant time and space variations of the free interstitial supersaturation in the wafer, TED can be accurately predicted for a variety of experimental conditions.

In a grazing incidence x-ray diffuse scattering study of defects in boron implanted and annealed silicon we have discovered narrow rods of intensity along [111] directions. These diffuse rods of intensity arise from stacking faults formed in the early stages of annealing in the range around 1000°C. From the width of the stacking fault induced rods we can estimate their size, while the integrated intensity is a measure of the total stacking fault area in the implanted layer. Surprisingly we find that these faults initially grow in size and density and after reaching a maximum begin to dissolve and ultimately are totally annihilated. The intensity rods are distinct from the point defect or point defect cluster scattering in the tails of the Bragg peaks referred to as Huang Scattering. From the nature of the q dependence of the diffuse scattering in the Bragg tails we find unequivocal evidence for the presence of clusters in our annealed samples. The average effective size seems remarkably independent of annealing temperature. These observations will be discussed in the context of the enhanced diffusion of boron over its bulk value.

In this paper we show that dopant decoration of process-induced defects is responsible for a failure mechanism of memory devices. From the electrical point-of-view, the defect-related failure consists in a source-to-drain resistive path formed by junction piping. This mechanism is made active by the very close spacing which is typical of present device structures. A device-like test structure is used for defect detection. This structure proves to be a very effective tool for studying the impact of various process steps on defect generation, in that it allowes statistical data about the formation of these defects to be collected. TEM analyses are extensively used for studying the evolution of end-of-range defects during subsequent thermal treatments and for measuring the amorphous layer width under various implantation conditions.

The role of high dose implantations in the generation of this sort of defects is discussed. Even if the amorphous layer is completely recovered by a suitable recristallization annealing, residual defects grow and become dopant-decorated during post-implantation thermal treatments. Defect density is increased by oxidizing treatments. In this case point defect injection is active both in enhancing dopant diffusion and in growing defects.

Defect formation is suppressed if the amorphous layer is made very shallow (≤ 50 nm) by suitable choices of the screen oxide and of the implantation energy. A binary collision code is used in order to estimate the dependence on energy of the self-interstitial excess outside the amorphous region. The results of these calculations indicate that defect suppression can be tentatively explained by point defect annihilation at the silicon surface.

The effect of nitrogen implants on boron transient enhanced diffusion was studied for nitrogen-only, boron-only, and boron plus nitrogen implants. A boron buried layer was used as a detector for interstitial supersaturation in the samples. Boron dose ranged from 1×1014 to 1×1015 cm−2 and N2+ dose from 5×1013 and 5×1014 cm−2. The energies were chosen such that the location of the nitrogen and boron peaks matched. After the implants, RTA and low temperature furnace anneals were carried out. The diffusivity enhancements were extracted from the buried layer profiles by simulation. Nitrogen-only implants were found to cause significant enhanced diffusion on the buried boron layer. For lower doses, the enhancement of the nitrogen implant is about half as that of boron whereas the enhancements are equal at higher doses. Nitrogen coimplant with boron increases the transient enhanced diffusion of boron at low boron doses, which implies that nitrogen does not act as a strong sink for excess interstitials unlike carbon. At high boron doses, nitrogen co-implant does not significantly change boron diffusion. Sheet resistance measurements indicate that low nitrogen doses do not affect the activation of boron whereas high nitrogen doses either reduce the activation of boron or the mobility of the holes.

This paper discusses problems inherent to scanning spreading resistance microscopy (SSRM) and ways to correct them to increase the resolution of two-dimensional dopant profiling. Specifically this paper looks into issues related to the probe-silicon contact and the damaged surface layer created by the sample preparation technique. Degradation of the measured dopant profile was observed when the probe scanned over the nitride spacers. Attempts to reduce the required contact pressures to increase the lifetime and effectiveness of the probes are addressed. The force required for SSRM was successfully reduced after the damage layer was partially removed by isotropic etching.

We found an anomalous behavior of dopant diffusion in Si substrate during the oxynitride process. SIMS measurements showed a notably enhanced diffusion during the NO and N2O oxynitride process. The considerably enhanced diffusion was also observed in re-oxidation of oxynitride film grown by the NO annealing or NH3 nitridation of a SiO2 film. In order to simulate the enhanced diffusion, an enhancement coefficient was introduced, showing that the simulation results are in reasonable agreement with the experimental ones. We applied the diffusion model to the simulations of MOSFETs fabricated under various conditions of the oxynitride process. The device characteristics of MOSFETs were successfully reproduced by adopting a suitable dependence of nitrogen concentration CN on the surface recombination rate of interstitial Si at the oxynitride/Si substrate interface.

A new method for non-destructive, small area characterization of ultra-shallow junctions, called Carrier Illumination™ (CI) has recently been developed. This work validates the CI measurement in full CMOS and NMOS process flows, with the aim of demonstrating its capability to provide in-line characterization of junction depth and uniformity on product wafers. Measurements have been carried out on both unpatterned and patterned wafers at various steps in the SEMATECH standard process flow. CI was used to characterize annealed 800 eV B11 PLDD implants into n-wells and annealed As75 NLDD source/drain (S/D) implants into p-wells. This work demonstrates correlation to dose, junction depth as measured with SIMS and SRP, and electrical properties of test structures.

(1) Proceedings of the Fifth International Workshop on Measurement, Characterization and Modeling of Ultra-Shallow Doping Profiles in Semiconductors, usj-99, Research Triangle Park, NC, March 1999, pp. 314-18.

GeB− Cluster ions have been used to effectively produce 0.65-2keV boron for low energy ion implantation. We have generated the GeB− cluster ions using the SNICS ion source (source of negative ion by cesium sputtering). Shallow junctions have been made by the GeB− cluster ions implanting into Si substrates at 15keV, 1×1015/cm2 and 5keV, 5×1014/cm2. The junction depth as small as 37nm has been achieved by rapid thermal annealing of the 5 keV sample at 1000°C for 1 second. A two-step annealing was also performed to study the diffusion of B in the GeB− ion cluster implanted Si by annealing the 15 keV implanted sample at 550°C/300sec+1000°C/10sec. We found that the junction depth of the two-step annealed sample was only half of the one-step annealed sample. TEM (transmission electron microscopy) showed clear recrystallization of the amorphized layer with no observable residual defects. We briefly discussed the role of Ge in regards to reduction of the junction depth.

The diffusion of B in Si and SiGe under the influence of point defect injection by Rapid Thermal Anneal (RTA) and conventional furnace anneal is studied in this work. B-doped regions in SiGe and Si were grown by LPCVD, and point defects were injected by RTA or furnace annealing bare, Si3N4 or SiO2 + Si3N4 covered samples in an oxygen atmosphere. Self-interstitial defects will be injected into bare Si while vacancy defects will be injected into Si3N4 covered samples, and inert annealing will occur in SiO2 + Si3N4 covered samples. The annealed and asgrown profiles were determined using SIMS analysis, and the diffusivities extracted by direct comparison of the profiles. Both interstitials and vacancies were injected during furnace annealing of SiGe, as demonstrated by the respective enhancement and retardation of the B diffusion. Enhanced B diffusion in SiGe was observed even for 5 s RTA at 1000°C, with an enhancement factor of ∼2.5. The B in Si diffusivity enhancement for interstitial injection by RTA oxidation was found to be a factor of ∼3 compared to inert anneals, close to the factor for SiGe.

State-of-the-art semiconductor devices require an accurate control of the complete twodimensional dopant distribution. The routine use of process simulators to predict the envisaged distributions and their resulting accuracy, is strongly linked to the physical models contained in these programs as well as their calibration. Whereas SIMS and SRP have been used extensively for the calibration of 1D-profiles, calibration of 2D-profiles has been very limited.

In this work, we report some results obtained with the 2D-profiling techniques SSRM (Scanning Spreading Resistance Microscopy) and SCM (Scanning Capacitance Microscopy ) for the study of two-dimensional effects on diffusion. In particular, we discuss the role of the nitride spacer on the lateral diffusion of arsenic and boron. Using series of transistors with different nitride spacers with or without TEOS-liners, a strong dependence between the lateral diffusion and the nitride spacer thickness can be observed using SSRM and SCM. The process flow eliminates the possible contribution of Transient enhanced diffusion (TED) as a dominant mechanism. At the same time an enhancement of the lateral stress underneath the spacers has been observed with CBED and Raman, suggesting a correlation between the lateral diffusion and the nitride spacers. The enhanced diffusion of As and B is strongly linked to the spacer size whereby the differences in enhancement suggest that the proximity of the dopants to the stress field field region is an important parameter.

A new model for carbon diffusion in silicon that explains carbon diffusion during annealing at 850°C and 900°C in superlattice carbon structures grown by MBE is implemented using the Monte Carlo atomistic simulator DADOS. Carbon concentrations in the delta layers are 2×1020 cm−3, exceeding by far the solid solubility. The simple kick-out mechanism which incorporates the well established values of the product of diffusivity and equilibrium concentrations of intrinsic point defects and in-diffusion experiments of carbon in silicon does not explain the observed C diffusion profiles. A more detailed analysis of the experiments shows that, in order to fit them, a more unstable Ci is required. Therefore, we include the formation of clusters in the simulations. The formation of carbon/Si self-interstitial clusters promotes the premature break-up of Ci and the increase of the Si self-interstitial concentration in the carbon rich regions and, consequently, provides a better fit to the experiments. The low solubility of carbon in silicon at the annealing temperatures explains why these clusters are formed, even under conditions where the self-interstitial concentration is below the equilibrium value.

Carbon is a native impurity in Si which is known to trap self-interstitials and decrease their diffusivity. Carbon has also been observed to decrease B transient enhanced diffusion (TED) in Si through these interstitial interactions. Recently it has been proposed that vacancies must also be considered when accounting for the reduction of B TED. We have incorporated both the kick-out mechanism and the Frank-Turnbull (F-T) mechanism in simulations of interstitial diffusion and carbon diffusion, as well as experiments involving B diffusion in B doped superlattices (DSLs) with varying C concentration regions. We have used the binding energy between a carbon atom and a self-interstitial as a basis for the reaction rates for both mechanisms, and have found that an single energy of 2.25 eV best reproduces the results from several experiments, assuming equilibrium initial conditions for both mechanisms and ab-initio equilibrium values for all point defects.

We studied the Al redistribution in the vicinity of a well-defined EOR band formed by Ge pre-amorphization. Aluminum was implanted in Si at 3 MeV to a low dose before the preamorphization step in order to localize the as-implanted Al peak away from the EOR band. Rapid thermal annealings were performed in the temperature range [900-1000°C] for times varying from 20 s up to 200 s. The results of this study evidence a clear accumulation of the dopant on the extended defects, indicating a direct trapping mechanism of the dopant by the EOR.

The interaction between boron and silicon interstitials caused by ion implant damage is a physical process which hinders the formation of ultra-shallow, low resistivity junctions. The possibility of mitigating the effective interstitial point defect population via introduction of nonmetallic impurities in ion implanted silicon has been investigated. Amorphization of a n-type Czochralski wafer was achieved using a series of Si+ implants of 40 keV and 150 keV, each at a dose of 1×1015/cm2. The Si+ implants produced a 2800Å deep amorphous layer, which was then implanted with 8 keV 1×1014/cm2 B+. The samples were then implanted with high doses of either carbon, oxygen, sulfur, chlorine, selenium, or bromine. The implant energies of the impurities were chosen such that the damage and ion profiles of the impurity were contained within the amorphous layer. This allowed for the chemical species effect to be studied independent of the implant damage caused by the impurity implant. Post-implantation anneals were performed in a tube furnace at 750° C. Secondary ion mass spectrometry was used to monitor the dopant diffusion after annealing. Hall effect measurements were used to study the dopant activation. Transmission electron microscopy (TEM) was used to study the end-of-range defect evolution. The addition of carbon and chlorine appear to reduce the boron diffusion enhancement compared to the boron control. Carbon and chlorine also appear to prevent boron out-diffusion during annealing compared to the control, which exhibited 20% dose loss following annealing.

Transient enhanced diffusion of implanted arsenic in silicon with/without additional self-implantation has been investigated. The experimental results show the suppression of As diffusion with Si self-implantation during initial stage of annealing in contrast to the prediction of conventional models. The results suggest that the arsenic and self-interstitial atoms might form immobile clusters during Si implantation or initial stage of annealing. After the clusters dissolve for further annealing, the transient enhanced diffusion of As increases with silicon implantation dose as expected from the “+1” model. These results clarify that interstitials/ As-I clusters play a major role in transient arsenic diffusion.

This paper describes the application of a novel optical metrology tool to in-line monitoring of shallow junction processes in a production setting. Recent results demonstrate the sensitivity of junction depth to implant dose and energy, as well as to annealing temperature and time. Results are also presented on monitoring of amorphizing implants and low dose implants. The in-line measurement has been performed on both bare wafers and product wafers. Measurements are compared with results obtained by destructive techniques such as SIMS and SRP. These results demonstrate that the optical method has the capability to provide in-line control of shallow junction processes.

In this paper we propose structural models of self-interstitial clusters in silicon using the empirical potential method. Novel, fully co-ordinated compact clusters based on the hexagonal interstitial up to size four have been proposed. The energetics of their formation indicates that energy minimisation occurs due to elimination of dangling bonds. While the conventional dangling bond model yields an exponential decrease in the formation energy with size, we observe stable defects for sizes in multiples of four. The energy barrier required for a transition from compact model to the dangling bond chain model can possibly explain the experimentally observed energy barrier for sizes greater than eight.

We have studied dose loss for B11 and BF2 implants with energies ranging from 10 keV to 1keV for B11 and 45 keV to 2keV for BF2. We found that B11 implants during a 1050C-10s RTA anneal segregate mainly to the bulk of the oxide. High dose BF2 implants on the other hand, show significantly larger amounts of dose loss after stripping the oxide, due to a pile-up of boron at the Si/SiO2 interface. In order to simulate B11 diffusion we have introduced a simple average cluster size model to simulate Boron-Interstitial-Cluster (BIC) evolution. For BF2 implants, we have simulated the effect of F by reducing the damage and we have used interface traps to account for the dose loss observed experimentally. Using these models, we have been able to fit the SIMS profiles across the whole matrix of implant conditions for both B11 and BF2 implants.