A comprehensive model for B implantation, diffusion and clustering is presented. The model, implemented in a Monte Carlo atomistic simulator, successfully explains and predicts the behavior of B under a wide variety of implantation and annealing conditions by invoking the formation of immobile precursors of B clusters, prior to the onset of transient enhanced diffusion. The model also includes the usual mechanisms of Si self-interstitial diffusion and B kick-out. The immobile B cluster precursors, such as BI2 (a B atom with two Si self-interstitials) form during implantation or in the very early stages of the annealing, when the Si interstitial supersaturation is very high. They then act as nucleation centers for the formation of B-rich clusters during annealing. The B-rich clusters constitute the electrically inactive B component, so that the clustering process greatly affects both junction depth and doping level in high-dose implants.

A comprehensive model of the nucleation, growth, and dissolution of B clusters in Si is presented. We analyze the activation of B in implanted Si on the basis of detailed interactions between B and defects in Si. In the model, the nucleation of B clusters requires a high interstitial supersaturation, which occurs in the damaged region during implantation and at the early stages of the postimplant anneal. B clusters grow by adding interstitial B to preexisting B clusters, resulting in B complexes with a high interstitial content. As the annealing proceeds and the Si interstitial supersaturation decreases, the B clusters emit Si interstitials, leaving small stable B complexes with low interstitial content. The total dissolution of B clusters involves thermally generated Si interstitials, and it is only achieved at very high temperatures or long anneal times.

The evaporation of {311} self-interstitial clusters has recently been linked to the phenomenon of transient enhanced diffusion in silicon. A theory of cluster evaporation is described, based on first-order kinetic equations. It is shown to give a good account of the data over a range of temperatures. The theory simultaneously explains several of the unexpected features of transient enhanced diffusion, including the apparently steady level of the enhancement during its duration, and the dependence of the duration on implant energy and dose. The binding energy used to match the theory to data is in good agreement with molecular dynamics calculations of cluster stability in silicon.

The effect of substitutional C on interstitial-enhanced B diffusion in Si has been investigated. Substitutional C was incorporated into B doped Si superlattices using molecular-beam-epitaxial growth under a background of acetylene gas. Excess Si self-interstitials were generated by near-surface 5×1013/cm2, 40 keV Si implants and diffused at 790 °C. The interstitial-enhanced diffusion of the B marker layers is fully suppressed for C concentrations of 2×1019/cm3, thus demonstrating that substitutional C acts as a trap for interstitials in crystalline Si. Uniform C incorporation of 5×1018/cm2 significantly reduces the transient enhanced diffusion of a typical B junction implant without perturbing its electrical activity.

Ion implantation of Si (60 keV, 1×1014/cm2) has been used to introduce excess interstitials into silicon predoped with high background concentrations of B, which were varied between 1×1018 and 1×1019/cm3. Following post-implantation annealing at 740 °C for 15 min to allow agglomeration of the available interstitials into elongated {311} defects, the density of the agglomerated interstitials was determined by plan-view transmission electron microscopy observation of the defects. We report a significant reduction in the fraction of excess interstitials trapped in {311} defects as a function of boron concentration, up to nearly complete disappearance of the {311} defects at boron concentrations of 1×1019/cm3. The reduction of the excess interstitial concentration is interpreted in terms of boron-interstitial clustering, and implications for transient-enhanced diffusion of B at high concentrations are discussed.

Pulsed laser melting experiments were performed on GexSi1−x alloys (x≤0.10) with regrowth velocities ranging from 0.25 to 3.9 m/s. Analysis of post-solidification Ge concentration profiles, along with time-resolved melt depth measurements, allowed determination of the liquid-phase diffusivity Dl for Ge in Si and the dependence of the Ge partition coefficient k on interface velocity v. A Dl of 2.5×10−4 cm2/s was measured. The k vs v data were analyzed using various models for partitioning, including both the dilute and nondilute Continuous Growth Models (CGM). Extrapolating to zero velocity using the partitioning models, an equilibrium partition coefficient of approximately 0.45 was obtained. Best fitting of partitioning data to the nondilute CGM yields a diffusive speed of 2.5 m/s. These measurements quantify previous indications of partitioning observed in other studies of pulsed laser processed GexSi1−x alloys.

Atomistic process modeling, a kinetic Monte Carlo simulation technique, has the interest of being both conceptually simple and extremely powerful. Instead of reaction equations it is based on the definition of the interactions between individual atoms and defects. Those interactions can be derived either directly from molecular dynamics or first principles calculations, or from experiments. The limit to its use is set by the size dimensions it can handle, but the level of performance achieved by even workstations and PC's, together with the design of efficient simulation schemes, has revealed it as a good candidate for building the next generation of process simulators, as an extension of existing continuum modeling codes into the deep submicron size regime. Over the last few years it has provided a unique insight into the atomistic mechanisms of defect formation and dopant diffusion during ion implantation and annealing in silicon. Object-oriented programming can be very helpful in cutting software development time, but care has to be taken not to degrade performance in the critical inner calculation loops. We discuss these techniques and results with the help of a fast object-oriented atomistic simulator recently developed.

The reduction of transient enhanced diffusion (TED) with reduced implantation energy has been investigated and quantified. A fixed dose of 1×1014 cm−2 Si+ was implanted at energies ranging from 0.5 to 20 keV into boron doping superlattices and enhanced diffusion of the buried boron marker layers was measured for anneals at 810, 950, and 1050 °C. A linearly decreasing dependence of diffusivity enhancement on decreasing Si+ ion range is observed at all temperatures, extrapolating to ∼1 for 0 keV. This is consistent with our expectation that at zero implantation energy there would be no excess interstitials from the implantation and hence no TED. Monte Carlo modeling and continuum simulations are used to fit the experimental data. The results are consistent with a surface recombination length for interstitials of <10 nm. The data presented here demonstrate that in the range of annealing temperatures of interest for p-n junction formation, TED is reduced at smaller ion implantation energies and that this is due to increased interstitial annihilation at the surface.