Computational Studies on the Non-Covalent Interactions of Asphaltene Model Compounds and Related Systems

Abstract

In this thesis, the non-covalent interactions of asphaltene model compounds are explored computationally using a combination of semi-empirical methods and density functional theory. New methods and workflows are developed with which to make more accurate predictions for reduced computational expense. In Chapter 2, a new synthetic procedure for pyrene-4,5-dione is disclosed. In developing this procedure, unusual purification difficulties were encountered, which led to the computational study in Chapter 3. Here, we used the Grimme group’s conformer-rotamer ensemble sampling tool (CREST) to generate ensembles of the low-energy homodimers of pyrene, pyrene-4,5-dione, and pyrene-4,5,9,10-tetraone, and of the heterodimer of pyrene with pyrene-4,5,9,10-tetraone. These ensembles were then further refined using density functional theory (DFT) to give high-quality geometries and energies. We found that the difficulty in purification in Chapter 2 likely originated in the relatively strong interactions between pyrene and pyrene-4,5,9,10-tetraone, which could cause the formation of a cocrystal and interfere with the purification of pyrene-4,5-dione. When the same computational methodology was extended to realistic model asphaltene systems, it became apparent that CREST was providing insufficient sampling of conformational space for those systems, so Chapter 4 details the development and implementation of a new algorithm for non-covalent complexes of flexible monomers, including those involving microhydration. This new algorithm is still based on GFN2-xTB metadynamics, like CREST, but it uses modified settings, with a somewhat weaker biasing, intended to preserve important directed non-covalent interactions for longer, and starts each cycle from a diversity of structures found in the previous cycle, rather than just from the single lowest-energy structure that CREST uses. Thus, we have termed the new algorithm a Low-Energy Diversity-Enhanced variant on CREST (LEDE-CREST). Chapter 5 applies LEDE-CREST to a realistic system of asphaltene model compounds for which there is experimental data available (Org. Biomol. Chem., 2015, 13, 6984-6991). Various different stoichiometries of clusters of model compounds were tested. LEDE-CREST was used to explore the possible geometries for each stoichiometry, and the lowest-energy structure found by LEDE-CREST was then reoptimized using DFT. Unfortunately, due to computational cost of using DFT for systems of this size, we were unable to reoptimize even a portion of the ensembles. The results give insight into motifs in model compound aggregation, the relative importance of different interaction modes (π-π stacking, hydrogen bonding), and shed some doubt on the importance of microhydration. Chapter 6 discusses the results of the entire thesis, and presents directions for future work.