A single statistical point defect based model for the nucleation and evolution of dislocation loops during annealing in Si is developed. The model assumes that the radius and the density of dislocation loops follow a log normal distribution. The loop nucleation part of the model also assumes that all the loops come from {311} unfaulting. The model is verified with the experimental results obtained by studying the formation of dislocation loops and threading dislocation loops as a function of implant condition in boron implanted silicon by varying the dose from 1×1013 to 5×1014 cm−2 at an energy of 1.5 MeV. Due to the statistical nature of the model, the threading dislocation loop density is easily obtained from simulation results. The dramatic change in the threading dislocation loop density withthe increasing implant dose is also predicted by the simulations.

The effect of point defect injection on the diffusion of antimony and boron in silicon and silicon-germanium alloys has been studied by comparison of inert with injection diffusions. In this work, Sb and B in Si were used as control wafers to investigate Sb and B diffusion behavior in Si0.9Ge0.1. The point defect injection technique was carried out by rapid thermal annealing (RTA) Sb and B in Si and Si0.9Ge0.1 samples with the various surface coatings in either oxygen or ammonia atmospheres to inject either interstitial or vacancy defects. The diffusion profiles for as-grown and RTA annealed samples were measured by Secondary Ion Mass Spectrometry (SIMS). Diffusivities for B in Si and Si0.9Ge0.1 were obtained using computer simulations of the measured boron profiles for their annealed samples. Sb diffusion in Si and Si0.9Ge0.1 was found enhanced by vacancy injection and retarded by interstitial injection. The enhanced B diffusion in Si and Si0.9Ge0.1 was found by interstitial injection. These results confirm that Sb diffusion in Si0.9Ge0.1 is primarily dominated by vacancy-mediated mechanism, while B diffuses in Si0.9Ge0.1 by an interstitially mediated mechanism. The effect of the RTA diffusion time on the B diffusion in Si and Si0.9Ge0.1 has also been investigated. The diffusivity versus diffusion time of B in Si and Si0.9Ge0.1 for inert and injection samples is presented. It was found that the shorter annealing time had the faster diffusion. This suggested that it caused by transient diffusion effect arising from point defects.

Simultaneous conductance imaging and constant current mode STM imaging have been used to delineate Si pn junction arrays over a range of reverse bias conditions. Conductance has been obtained by adding a modulation signal to voltages applied in the p and n regions of a model device, and by measuring the modulation signal of the tunneling current with a lock-in amplifier. Both constant current and conductance imaging ofthe electrically different regions (n, p, and depletion zone) show a pronounced dependence on applied pn junction bias. The conductance contrast is mainly due to electrically different behaviors of metal-gap-semiconductor junction which are determined by the tip-induced band bending of the oxide-passivated silicon surface.

Cryo-implantation technology is proposed for reducing crystal defects in Si substrates. The substrate temperature was controlled to be below at -160°C during ion implantation. No dislocation was observed in the implanted layer after rapid thermal annealing. Pn junction leakage was successfully reduced by one order of magnitude as compared with room temperature implantation. Precise dose control is indispensable in channel region of high performance MOSFETs. In order to improve the precision of implanted dose, chip size implantation technology without photoresist mask was developed. In this technology, chip-by-chip implantation can be carried out by step-and-repeat wafer stage, and different implantation conditions are available in the same wafer independent of wafer size.

Previously, it has been reported that PMOS capacitors with heavily boron-doped polycrystalline SiGeC gates are less susceptible to boron penetration than those with poly Si gates [1]. Boron appears to accumulate in the poly SiGeC layers during anneals, reducing boron outdiffusion from the gate despite high boron levels in the poly SiGeC at the gate/oxide interface. In this abstract, we report clear evidence of strong boron segregation to polycrystalline SiGeC layers from poly Si, with boron concentration in poly SiGeC (Ge=25%, C=1.5%) increasing to four times that of adjacent poly Si layers. A separate experiment confirms that this result is not due to any SIMS artifacts. Electrical measurements of heavily in-situ doped single layer samples show that the conductivity of poly SiGeC is similar to poly Si and remains roughly constant with annealing at 800°C. However, in a two-layer sample where the poly SiGeC is initially lightly doped and subsequently heavily doped by diffusion by from an adjacent poly Si layer, conductivity appears lower than in poly Si.

Phosphorus diffusion has been studied in both pure epitaxially grown silicon and Cz silicon, with a substantial amount of impurities like oxygen and carbon. Anneals have been performed in different atmospheres, N2 and dry O2, as well as in vacuum, at temperatures between 810 – 1100°C. Diffusion coefficients extracted from these anneals show no difference for the P diffusion in the epitaxially grown or the Cz silicon. The diffusion coefficients follow an Arrhenius dependence with the activation energy Ea=2.74±0.07 eV and a prefactor D0 = (8±5)×10−4 cm2/s. These parameters differ considerably from the previously reported and widely accepted values (3.66 eV and 3.84 cm2/s, respectively). However, vacuum anneals of the same samplesresult in values close to this 3.6 eV diffusion mode. Furthermore, control anneals of boron doped samples, with similar design as the phosphorus samples, suggest the same trend for boron diffusion in silicon – lower versus higher values of activation energies for nitrogen and vacuum anneals, respectively. These results are discussed in terms of the concentration of Si self-interstitials mediating the diffusion of phosphorus and boron.

In this paper, we study the effect of the Ge+ preamorphisation dose on the thermal evolution of End of Range (EOR) defects upon annealing. Amorphisations were carried out by implanting Ge+ at 150 keV to doses ranging from 1×1015 ions/cm2 to 8×1015ions/cm2. Rapid Thermal Annealing (RTA) was performed for various time/temperature combinations in nitrogen ambient. Plan view transmission electron microscopy under specific imaging conditions was used to measure the size distributions and densities of the EOR defects. We found that for a fixed thermal budget, the increase in the Ge ion dose results in an increase in the defect density but has no effect on the defect size distribution. This invariance of the mean size of defects with respect to the initial supersaturation introduced in the matrix is an expected characteristic of a conservative Ostwald ripening mechanism. Moreover, the total number (Nb) of Si interstitial atoms bound to the EOR defects is a monotonically increasing function of the Ge ion dose. Furthermore, we found that Nb is directly proportional to the number of Si atoms in excess of the vacancies found below the a/c interface as calculated by MonteCarlo simulations. This is consistent with the “excess interstitial” model which explains the origin of the EOR defects.

The effect of annealing silicon capped pseudomorphic Si0.7865Ge0.21C0.0035 or Si0.998C0.002 layers on silicon substrates in nitrogen or oxygen at 850°C was examined using x-ray diffraction (XRD) and secondary ion mass spectrometry (SIMS). Most substitutional carbon is lost from the alloy layers due to carbon out-diffusion rather than from precipitation. The carbon is found to diffuse more rapidly out of the SiGeC layer than the SiC layer after nitrogen and the carbon is found to leave the sample entirely, an effect that is enhanced by oxidation and thin cap layers. All substitutional carbon can be removed from the sample in some cases implying negligible formation of silicon-carbon complexes. Furthermore, it is found that each injected silicon interstitial atom due to oxidation causes the removal of one additional carbon atom for the SiGeC layer.

The point defect injection from arsenic precipitation was studied using boron marker layers and antimony doped superlattices. Comparisons of arsenic and germanium amorphizing implants showed similar boron marker layer diffusion enhancements after spike annealing. The results indicate that the end of range damage caused by the implants was the source of the diffusion enhancement. Additional annealing cycles showed that there was retardation in the diffusion enhancement of the boron marker layers for precipitation range arsenic implants. Antimony marker layers showed no diffusion enhancement due to vacancy injection. The results of the experiments indicate that arsenic-interstitial complexes are the cause of the decrease flux of interstitials to the bulk.

Heavily-doped, ultra-shallow junctions in boron implanted silicon using pulsed laser annealing have been created. Laser energy in the nonmelt regime has been supplied to the silicon surface at a ramp rategreater than 1010°C/sec. This rapid ramp rate will help decrease dopant diffusion while supplying enough energy to the surface to produce dopant activation. High-dose, non-amorphizing 1 keV, 1e15 ions/cm2 boron is used. Four-point probe measurements (FPP) show a drop in sheet resistance withnonmelt laser annealing (NLA) alone. Transmission electron microscopy (TEM) shows the NLA dramatically affects the defect nucleation resulting in fewer defects with post annealing. Hall mobility and secondary ion mass spectroscopy (SIMS) results are also shown.

As semiconductor device dimensions continue to decrease, the main challenge in the area of junction formation involves decreasing the junction depth while simultaneously increasing the active dopant concentration. Laser annealing is being investigated as an alternative to rapid thermal annealing (RTA) to repair the damage from ion implantation and to activate the dopants. With this technique, uniform, box-shaped profiles are obtained, with dopant concentrations that can exceed equilibrium solubility limits. Unfortunately, these super-saturated dopant concentrations exist in a metastable state and deactivate upon further thermal processing. In this work, a comprehensive study of the deactivation kinetics of common dopants (P, B, and Sb) was performed across a range of annealing conditions. For comparison, As deactivation data from the work of Rousseau et al.1 is also presented. Each dopant exhibits different deactivation behavior, however, As and P can be classified as unstable species while B and Sb are stable against deactivation until higher temperatures of 700-800°C. In addition, a means of maintaining these metastably doped layers is being investigated with the goal of meeting the International Technology Roadmap for Semiconductors (ITRS) requirements for ultrashallow junctions.

Hydrogen is known to cause the passivation of boron acceptors after such processing steps as wet etching, reactive ion etching, sputter deposition of metal contacts, and Ar ion beam etching. Previous studies of this effect employed CV profiling, spreading resistance profiling, and SIMS measurements on samples diffused with deuterium. These methods are either destructive to the Si surface or require deposition of metal contact. In the present study we used a non-contact small signal ac-surface photovoltage technique, currentlyavailable in commercial diagnostic tools. Simultaneous measurements of the semiconductor surface barrier, Vsb, and the capacitance of the surface depletion layer, CD, give the concentration of boron acceptors in a submicron distance from the Si surface or Si/SiO2 interface. The technique has proven very successful in monitoring low dose implants and also near surface doping in oxidized wafers. Inbare silicon wafers the method occasionally indicated surface boron concentration noticeably below the bulk value. We found such behavior in wafers after the chemical cleaning, used to prepare a hydrogen terminated surface. Thermal annealing at temperatures from 150°C to 200°C reactivates the boron dopant. We will discuss the effect of various cleaning and annealing conditions on passivation and reactivation of boron acceptors in the near surface region. The results obtained with the non-contact SPV technique show excellent agreement with previous studies. They also provide a basis for reliable measurement of the boron concentration free of interference from hydrogen passivation.

The effect of arsenic on {311} defect formation was determined for temperatures ranging from 700°C to 800°C. Arsenic well structures were formed at arsenic concentrations of 3×1017, 3×1018, and 3×1019 cm−3. A 40 keV 1×1014 cm−2 silicon implant, that is known to form {311} defects, was then incorporated into the structures. Extended defect evolution and dissolution was then studied after furnace annealing at 700°C, 750°C and 800°C for various times. It was determined that arsenic has a strong affect on the nucleation of extended defects. However, once the defects were formed, the dissolution time constant was the same for all concentrations considered. The activation energy for defect dissolution was found to be 3.4eV and was also independent of arsenic concentration. Using a newly developed {311} model in the FLOOPS process simulation software, the effect of the arsenic on {311} formation and dissolution was simulated. It was found that by using a pair model with an arsenic-interstitial binding energy of 0.95eV, the experimental results were able to be simulated.

We discuss atomistic simulations of ion implantation and annealing of Si over a wide range of ion dose and substrate temperatures. The DADOS Monte Carlo model has been extended to include the formation of amorphous regions, and this allows simulations of dopant diffusion at high doses. As the dose of ions increases, a continuous amorphous layer may be formed. In that case, most of the excess interstitials generated by the implantation may be swept to the surface as the amorphous layer regrows, instead of diffusing through the crystalline region. This process reduces the amount of transient enhanced diffusion during annealing. This model also reproduces the dynamic annealing during high temperature implants.

The melt kinetics of shallow junction formation by laser thermal processes has been studied using transient conductance measurements. The melt and solidification dynamics of 20 nm amorphous layers were measured and shown to follow behaviors predicated by deeper melts, including explosive crystallization and interface bounce back. The effects of surface barrier oxides and metal absorber layers, required for CMOS process integration, were examined and shown to be nearly negligible. Quantitative evaluation of a device process window by these measurements was in good agreement with sheet resistance results. Finally, the effect of the buried oxide in SOI structures was investigated. Solidification velocities in such structures were reduced by a factor of three as compared with bulk silicon.

Understanding the diffusion and activation of arsenic is critical for the formation of low resistance ultra-shallow junctions as required for nanoscale MOS devices. In this work, we use results of ab-initio calcultions in order to gain insight into the fundamental process involved in arsenic activation/deactivation. Utilizing continuum modeling, we find it is possible to account for both the very rapid initial deactivation of arsenic as well as the strongly superline independence of interstitial supersaturation on doping level which accompanies deactivation. The critical process is the rearrangement of A s-atoms via in testitial mediated diffusion leading to ejection of silicon atoms from arsenic complexes and formation of arsenic-vacancy clusters.

This paper presents a model for {311} defects based on in-situ experiments. The model fits the 311 dependence on silicon implant energy and doses. The surface dependence of the model is described in detail, and compared to previous literature data. New data is presented on the surface effect on {311} dissolution and the model is compared to that data. In addition, the model is also used to explain the effects of doping on {311} defect behavior. Doping does not influence the dissolution of {311} defects, but only influences their nucleation behavior.

The work described herein focuses on examining the effect of solid phase epitaxial regrowth (SPER) on boron implanted silicon. It is shown that boron levels within the silicon can greatly enhance or reduce the regrowth rate of the silicon. Electrical measurements show optimum sheet resistances for 5 keV, 2×1015 cm−2 implant conditions yielding sheet resistance values of ∼140 Ω/sq at 500 °C annealing to ∼120 Ω/sq at 650 °C. Results using Hall effect and four-point probe show lower doses of boron will become fully active but levels will drop significantly as dose is increased. Lastly, maximum active concentrations of boron appear to reach values of ∼3-4×1020 cm−3 for a boron dose of 1×1015 cm−2 after SPER. Lower SPER anneal temperatures or higher doses tend to activate less boron.

In this paper we report measurement of the silicon diffusion coefficient in silicon dioxide films using isotopically enriched 28Si silicon dioxide layers. 30Si atoms are introduced in excess in a stoichiometric isotopically pure silicon dioxide layer either by ion implantation or by a predeposition technique and the time evolution of the 30Si concentration profile under various thermal conditions is monitored using SIMS. The estimated diffusivity values are significantly higher than previouslyreported values for Si diffusion within a stoichiometric oxide and closer to reported values for excess Si diffusion within an oxide.

To meet the market demands for higher performance LSIs, traditional scaling has been aggressively pursued and has enjoyed great success over the 0.1-μm generations. To maintain continued growth of CMOS performance beyond the 0.1-μm generation, key issues originating from traditional scaling are addressed from the viewpoint of the doping processes (channel engineering, high-activation, and gate- electrode structure) in this paper.

To meet the acceleration in gate-length miniaturization, short-channel effects must be suppressed at a low threshold voltage by using aggressive channel engineering. A channel-impurity profile must be optimized two-dimensionally, not uniformly or one-dimensionally. Channel engineering using tilted-channel implantation (TCI) with Indium is demonstrated.

Traditional scaling results in large variations of threshold voltage due to the statistical-impurity variation in a channel region. We studied the effect of the above channel engineering on threshold voltage fluctuation caused by a statistical- dopant variation by measurement and simulation. It is reported that the two-dimensionally optimized channel profile enhances threshold-voltage fluctuation even if the implantation process variation is negligible.

As CMOS device scales, reduction of parasitic resistance becomes very important for a high performance operation. Resistance at extension egde and contact resistance at silicide-Si interface are dominant factors.The traditional approach is to use a higher RTA temperature and a shorter RTA time. The ultimate RTA is laser annealing. We will demonstrate the laser annealing process with an ultra-low contact resistance of 4 × 10−8 Ω-cm2.

By integrating the above technologies, ultra-thin gate insulators, and reduction in gate to source/drain overlap length, we can establish front-end process for sub-0.1 μm generations.