Computational studies of silicon interfaces and amorphous silica

Abstract

Computational simulations are a new branch of science where one can perform computational experiments by simulating a model system. The model contains an approximation of a real system with simplified physical laws and reduced size in order to make it computationally tractable. Many physical systems cannot be analytically solved and also experimentally not everything can be measured. Computational simulations fill the gaps and allow new kinds of problems to be addressed.

Two main issues are studied in this thesis with computational simulations. The first issue concerns interfaces in silicon and is studied with deterministic molecular dynamics simulations. The second issue concerns the amorphous state of silicon and silica created with a stochastic algorithm.

As for the first issue the first study is a preliminary investigation of the heat conductivity over an amorphous-crystalline interface in silicon. Second, we present a methodology for calculating the life-time of large amorphous clusters embedded in a crystalline matrix by simulating much smaller clusters. We employ this methodology to study amorphous silicon clusters and find that the activation energy of the boundary movement is temperature dependent with a change in behavior at 1150 K. The last problem concerns the structure of twist grain boundaries at 0 K. We present detailed simulations where we explore many possible grain boundary structures by allowing the number of atoms at the interface to change. This is a degree of freedom that has not been previously considered in computational studies and proved to be of utmost importance. We find that the twist grain boundaries have ordered structures at 0 K for atomic densities at the interface not previously considered. We also find that the structural unit model is valid for twist grain boundaries and present the structural units forming the boundary.

The second issue concerns the amorphous state of silicon and silica. With a stochastic Monte Carlo simulation method that operates on the bond network of silicon and silica, we have created both amorphous silicon and silica. We find that we can improve the structural description of amorphous silica by adjusting the potential model. In addition, we developed optimizations for the simulation method which made it tractable for larger systems.