(398d) A Computational Microscopy Study of Self-Interstitial Aggregation in Ion-Irradiated Silicon

The evolution of self-interstitials and their aggregates during the annealing of ion-implanted silicon has received a tremendous amount of attention because of their strong, non-linear effects on the diffusion of dopants. Ion-implantation is the most common way of introducing dopants into a silicon substrate, as required for microelectronic device fabrication. The lateral distribution of dopant atoms can be finely controlled using masks, while the depth can be controlled by the implant energy. However, the implantation process leads to extensive lattice damage, and at high enough energies, complete amorphization, which must be recrystallized by thermal annealing following the implantation process. Also generated by the implantation process is a large number of excess atoms (interstitials) which are kicked out of their lattice sites by the incoming dopant atoms. Silicon self-interstitials are the primary diffusion mediators for some dopants such as boron and lead to greatly enhanced diffusion during the annealing process. The enhanced diffusion is highly detrimental to device performance because it effectively “smears” the device junctions. As the self-interstitial population relaxes towards equilibrium during annealing, the diffusion enhancement varies strongly, a phenomenon that has been described as Transient Enhanced Diffusion, or TED.

A major obstacle to understanding and quantitatively predicting TED is the formation of a variety of self-interstitial aggregates, which range from small amorphous three-dimensional clusters, to planar stacking-faults with various crystallographic orientations. Experimental and theoretical studies have demonstrated that slightly different annealing conditions lead to substantially different TED behavior, which can be linked to differences in the aggregation and dissolution dynamics. While a global picture of TED is now available and the various cluster structures identified, there are still several outstanding issues related to the atomistic mechanisms of self-interstitial clustering that are not understood. In order to develop a fully predictive TED model, a quantitative, temperature dependent understanding of these processes is required.

In the present study, we use large-scale constant-stress MD simulations to dynamically simulate the evolution of an ensemble of highly supersaturated self-interstitials at various temperatures and pressures. We show that the interstitial clustering exhibits a complex thermodynamic-kinetic phase diagram that is highly consistent with numerous experimental observations. The role of certain “magic” cluster sizes (four-interstitials and eight-interstitials) is shown to be temperature dependent and leads to a bifurcation in the aggregation pathway of larger cluster sizes. At low temperatures the magic cluster sizes assume a well-known reconstructed structure that is highly stable and immobile. This structure serves as a constraining building block for one type of (non-equilibrium) planar defect oriented along the (113) crystallographic plane. On the other hand, at higher temperatures, configurational entropy destabilizes the magic cluster configuration and allows for the growth of larger amorphous three-dimensional clusters that spontaneously collapse into different types of planar defect structures that are oriented along the (111) planes. Both the simulated (113) and (111) structures are in excellent agreement with DFT results and experimentally derived structures