Computational Workflows to Predict Ensemble Property Distributions from Structures Obtained by Single-Molecule Imaging: Application to Steam-Cracked Tar, Coal-Derived Asphaltene, and Petroleum Pitch

Petroleum crude oil, coal tars, bitumen, petroleum pitch, and related materials are extraordinarily complex chemical mixtures. Over the past decade, single-molecule imaging has been used to measure chemical structures in many of these organic mixtures. These structures represent a rich data set for refining hypotheses about structure/property relationships. To facilitate this, we present a workflow which uses computational chemistry to predict ensemble property distributions from ensembles of chemical structures. We apply this workflow to three databases of chemical structures taken from single-molecule imaging of a steam-cracked tar, a coal-derived asphaltene, and M-50 petroleum pitch. We use our workflow to predict multiple properties of each ensemble: chemical diversity, octanol/water partitioning, “druglikeness” as defined by Lipinski’s rules, carbon X-ray photoelectron spectroscopy (XPS) spectra, UV/vis spectra, and near-infrared absorbance. Our results confirm that the imaged structures are chemically diverse, with only 35% of the structures present in the PubChem database. The predicted ensemble distribution of octanol/water partitioning is quite broad, and highlights the presence of “druglike” molecules satisfying Lipinski’s rules. The predicted carbon XPS spectra provide a proof-of-principle for future experimental studies combining single-molecule imaging and ensemble XPS measurements of the same sample. The predicted UV/vis absorption spectra are consistent with previous work. Most importantly, the predicted near-infrared absorbance provides a novel structural hypothesis for the near-IR absorption tail seen in crude oil, asphaltenes, and related materials. We hypothesize that near-IR absorption comes from cata-condensed polycyclic aromatic diradicaloids, three of which are visible in the single-molecule images. We suggest that computational chemistry workflows can be a useful tool for refining ensemble structure/property hypotheses in complex mixtures.