Data-Driven Representative Models of Bio-Oils for Atomistic Simulations of Biobased Asphalt Rejuvenators

A large variety of waste can be transformed into bio-oils and biocrude oils via processing techniques such as pyrolysis and hydrothermal liquefaction (HTL), respectively.¹ These techniques provide a waste valorization route with applications such as the production of biofuel precursors, asphalt additives and other carbon-based materials, which can be derived from plastics, lignocellulosic biomass, chitin, etc. However, there are challenges in the fundamental understanding of the structure-property relationships within these bio-based oils, which are exasperated with difficulty in experimental characterisation of these complex bitumen-like materials. To address this lack of understanding, the development of molecular models for density functional theory and molecular dynamics (MD) studies is key and has potential to identify applications for oils produced by the pyrolysis or HTL of plastics and bio-based polymers (i.e shopping bags, food utensils, packaging) at the end of their life cycle. Computational studies have been carried out to elucidate the mechanisms occurring during production of bio-based oils during hydrothermal liquefaction²^,³ but there are few of these studies investigating the properties and behaviours of the produced bio-based oils with only one comprehensive atomistic model having currently been developed for algae-derived biocrude⁴. This model is based on existing models for crude oil and asphaltenes⁵^,⁶, which are based off years of development and research but are tailored to a specific material. In this work, we present a methodology to generate data-driven representative models from pyrolysis gas-chromotography (py-GC-MS) experimental characterization of any complex organic fluid, e.g., bio-oil, biocrude oil, or asphalt samples by generating a selection of statistically representative models that simplify the complexity of the fluid into a limited group of molecules. A variety of different molecular models have generated using a range of different approaches to evaluate the best method for capturing the key chemical features a complex organic fluid that are required for further computational chemistry calculations and large scale molecular dynamic simulations. These atomistic models are key to understanding and implementing waste valorization routes to sustainable, added-value products with potential to accelerate material development from waste. A computational framework has been developed to automate model creation from py-GC-MS data, which could enable future high-throughput MD simulations; ultimately helping to shape future waste-to-materials strategies.<br/><br/>1 A. Bachs-Herrera, D. York, T. Stephens-Jones, I. Mabbett, J. Yeo and F. J. Martin-Martinez, <i>iScience</i>, 2023, <b>26</b>, 106549.<br/>2 S. Yan, D. Xia, X. Zhang and X. Liu, <i>Energy</i>, 2022, <b>255</b>, 124561.<br/>3 D. C. Hietala and P. E. Savage, <i>Chemical Engineering Journal</i>, 2021, <b>407</b>, 127007.<br/>4 D. López Barreiro, F. J. Martin-Martinez, C. Torri, W. Prins and M. J. Buehler, <i>Algal Research</i>, 2018, <b>35</b>, 262–273.<br/>5 O. C. Mullins, <i>Energy Fuels</i>, 2010, <b>24</b>, 2179–2207.<br/>6 O. C. Mullins, H. Sabbah, J. Eyssautier, A. E. Pomerantz, L. Barré, A. B. Andrews, Y. Ruiz-Morales, F. Mostowfi, R. McFarlane, L. Goual, R. Lepkowicz, T. Cooper, J. Orbulescu, R. M. Leblanc, J. Edwards and R. N. Zare, <i>Energy Fuels</i>, 2012, <b>26</b>, 3986–4003.