Date of Award

Winter 12-2015

Embargo Period

8-3-2017

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Physics

Advisor(s)

Michael Widom

Abstract

With the help of density functional theory (DFT) and powerful computers, firstprinciples computation of solid state systems can be performed to accurately explain and predict nanoscale phenomena. This thesis focuses on our first-principles investigation of bismuth on the Ni(111) surface, at transition metal interfaces, and describes our study of boron carbide bulk thermodynamics combining DFT calculations, machine learning methods and Monte Carlo simulations. Our Bi on Ni(111) surface study confirms the stability of odd-layer Bi films, proposes specific stable atomic structures, and explains their stability with covalent chemical bonding. Our research of Bi at transition metal grain boundaries verifies the stability of bilayer films, explains the difference between transition metals, and proposes a model for bilayer stability on general grain boundaries. Although DFT calculations are accurate, they can be time consuming and scale badly with system size. Our DFT-based machine learning interaction models are used to capture certain non-linear effects associated with many-body interactions blue which reduce the error of prediction by 20 − 33% comparing to a linear model. We utilize these models to evaluate the thermodynamics of boron carbide in Monte Carlo simulations and identify three boron carbide phases.

Share

COinS