Hierarchical simulation of microcrystalline PECVD silicon film growth and structure

Abstract

We have designed and implemented a hierarchical simulation methodology capable of addressing the growth rate and microstructural features of thin silicon films deposited through PECVD (Plasma Enhanced Chemical Vapor Deposition). Our main objective is to elucidate the microscopic mechanisms as well as the interplay between atomic level and macroscopic design parameters associated with the development of nano- or micro-scale crystalline regions in the grown film. The ultimate goal is to use multi-scale modeling as a design tool for tackling the issue of local crystallization and its dependence on operating variables. At the heart of our simulation approach is a very efficient, large-scale kinetic Monte Carlo (kMC) algorithm which allows generating samples of representative Si films based on a validated chemistry model. In a second step, the generated film is subjected to an atomistic simulation study which restores the molecular details lost or ignored in the kMC model. The atomistic simulations are computationally very demanding; they are, however, an important ingredient of our work: we use it to back-map the coarse grained model employed in the kMC simulations to an all-atom model which is further relaxed through detailed NPT molecular dynamics (MD) or Monte Carlo simulations. This tunes local structure thus also important morphological details associated with the presence of crystalline and amorphous regions (and the intervening interfacial domains) in the grown film.

The kMC algorithm is based on a carefully chosen set of reacting or active radicals (species) in the gas phase impinging the film and a detailed set of surface reactions. Inputs for species fluxes are taken from a well-tuned plasma fluid model that includes a detailed gas phase chemistry reaction scheme. The growth mechanism consists of various surface kinetic events including radical-surface and adsorbed radical-radical interactions, radical-surface diffusion, and surface dissociation reactions. The very fast surface diffusion is decoupled from the rest of the kMC events and is treated deterministically in our work. For a three-dimensional Si(001)-(2x1):H crystalline lattice, our kMC algorithm allows us to simulate film growth over several seconds, resulting in thickness on the order of tens of nanometers. In the following pages we provide more details about the implementation of our kMC algorithm along with validation results.