Date of Award


Degree Type


Degree Name

Doctor of Philosophy (PhD)

Graduate Group

Chemical and Biomolecular Engineering

First Advisor

Talid Sinno


High-quality Ge substrates have numerous applications, including high-efficiency III-V multijunction solar cells and photodetectors. But the high cost of single-crystalline Ge makes the use of Ge-on-Si virtual substrates more practical for device fabrication. However, the lattice mismatch between Ge and Si leads to a highly strained Ge layer when grown directly on the Si lattice. The high mismatch strain unavoidably leads to defects, primarily dislocations, that degrade the Ge film quality. Several approaches for mitigating these defects have been proposed, including selective epitaxial growth (SEG), in which one employs an amorphous layer (most often SiO2) as a mask to reduce the epitaxial contact between the Ge and Si lattices to lower the mismatch strain. SEG has been demonstrated to successfully produce high-quality Ge films on Si, although defects are not fully eliminated. Further improvements will require quantitative understanding of the underlying atomic-scale mechanisms.

In this work, we present a computational framework to atomistically model the components of the SEG system (Si/SiO2/Ge). The model is validated by comparing predictions to experimental observations and ab initio calculations of various properties related to crystalline Si and Ge and amorphous SiO2, as well as combinations of these materials. The framework is then applied to study in detail the deposition of Ge on amorphous SiO2. It is shown that the simulations are able to access experimentally meaningful deposition conditions and reproduce several quantities related to the island size distribution. We then extend our simulation framework for deposition to include coarse projective integration (CPI). CPI is a multiscale modeling technique well-suited for situations, like atomic deposition, in which a system exhibits fast, stochastic processes, superposed onto slowly-evolving dynamics. In particular, we demonstrate an approach for generating atomistic configurations from limited knowledge of an island size distribution, which represents one of the key challenges in applying CPI to atomistic deposition. The results generated here should be easily adaptable to other deposition systems.

Files over 3MB may be slow to open. For best results, right-click and select "save as..."