From Atoms to Asphalt: Structure-Property Relationships in Bituminous Hydrocarbons

Abstract

Heavy petroleum streams such as heavy crudes, residues, and bitumen support modern energy and infrastructure, yet their properties vary strongly with composition, temperature, and chemical history. Because these materials consist of chemically diverse and evolving molecular populations rather than well-defined compounds, conventional structure–property correlations often lack robustness.

This dissertation develops a composition-based structure–property framework that combines Molecular Dynamics simulations with neural-network surrogate models to predict thermophysical and mechanical properties directly from molecular descriptors and temperature.

By enabling near-instantaneous property screening, comparison, and optimization of candidate formulations, the resulting draw-and-predict workflow enhances engineering design, accelerating the development of advanced hydrocarbon materials for next-generation energy and infrastructure systems.