Computational Methods to Quantify Nanoparticle Interactions

Abstract

Nanoparticles have emerged as a promising class of materials with potential applications in a variety of diverse fields. Central to the study of nanoparticles is understanding how these structures interact with their surrounding environment as these interactions can govern both how nanoparticles form and their specific functions. Computational methods show promise as a means to characterize these interactions on an atomic level as well as efficiently quantify a large number of potential nano-interactions. Still, many challenges exist in applying computational methods to the study of nanoscale interactions including the lack of available datasets, complex nanoscale chemistry, and heterogeneous nanoparticle environments. In this thesis, I show how multiple computational methods can be applied together to most effectively overcome these challenges and quantify nanoparticle interactions while maintaining a high level of chemical interpretability. Atomistic simulation provides a physically grounded means to produce nanoparticle interaction data, numerical descriptors provide a chemically relevant method to interpret atomistic simulations, while machine learning offers a computationally efficient tool to relate chemical descriptors to complex nanoscale interactions. I apply these methods to answer two broad questions: First, how do nanoscale interactions drive nanoparticle growth and the resulting properties of these nanoparticles. Second, how do the properties and chemistries of nanoparticles contribute to their function through the interactions in which they participate. In the first application, I focus on the chemical interactions leading to the formation of polycyclic aromatic compounds (PACs), a key class of structures in the creation of soot aggregates and the synthesis of gas-phase carbon nanoparticles. I show how kinetic Monte Carlo simulations of these interactions can reproduce the diverse PAC chemical space in complex flame environments, while numerical descriptors and machine learning can help us better understand these processes. In the second application, I demonstrate how computational techniques can explain the physical interactions of these PAC nanostructures that lead to their aggregation into larger nanoparticles. Finally, I introduce a versatile nanoscale interaction prediction tool that uses machine learning to accurately predict interaction sites between nanostructures, showing how it can help understand the interactions and functions of liquid-phase biological nanoparticles. The wide variety of nanoparticle systems studied in this work underscores that these computational methods are not confined to a single class of nanostructure or type of interaction but rather provide a robust framework that can be applied to computationally quantify nanoparticle interactions in a diverse range of applications.