Vacancy assisted arsenic diffusion and time dependent clustering effects in silicon

TL;DR

This study uses kinetic lattice Monte Carlo simulations to explore arsenic diffusion in silicon mediated by vacancies, revealing time-dependent diffusion behavior, the importance of long-range interactions, and how vacancy and arsenic concentrations influence clustering and diffusivity.

Contribution

It demonstrates the critical role of long-range interactions and vacancy concentrations in arsenic diffusion and clustering in silicon, with new insights into time-dependent diffusion models.

Findings

Time dependence in pair diffusion model applicability.

Long-range interactions are essential for accurate simulations.

Arsenic diffusivity varies with concentration, increasing up to 10^{19} cm^{-3} and decreasing thereafter.

Abstract

We present results of kinetic lattice Monte Carlo (KLMC) simulations of substitutional arsenic diffusion in silicon mediated by lattice vacancies. Large systems are considered, with 1000 dopant atoms and long range \textit{ab initio} interactions, to the 18th nearest lattice neighbor, and the diffusivity of each defect species over time is calculated. The concentration of vacancies is greater than equilibrium concentrations in order to simulate conditions shortly after ion implantation. A previously unreported time dependence in the applicability of the pair diffusion model, even at low temperatures, is demonstrated. Additionally, long range interactions are shown to be of critical importance in KLMC simulations; when shorter interaction ranges are considered only clusters composed entirely of vacancies form. An increase in arsenic diffusivity for arsenic concentrations up to $10^{19} \text{cm}^{-3}$ is observed, along with a decrease in arsenic diffusivity for higher arsenic concentrations, due to the formation of arsenic dominated clusters. Finally, the effect of vacancy concentration on diffusivity and clustering is studied, and increasing vacancy concentration is found to lead to a greater number of clusters, more defects per cluster, and a greater vacancy fraction within the clusters.