2024 AIChE Annual Meeting
(169cu) Leveraging Grand Canonical Monte Carlo and Machine Learning for Classification of Solvent Affinity in Functionalized Polymer Membrane Materials
Authors
Quach, C., Vanderbilt University
Jennings, G. K., Vanderbilt University
McCabe, C., Vanderbilt University
Cummings, P., Vanderbilt University
Understanding the affinity of polymers towards different solvents is crucial for the development of polymeric membranes for separation processes. In this study, we utilized Grand Canonical Monte Carlo (GCMC) simulations to assess the absorption behavior of water and ethanol on a target polymer. Leveraging the scripting capabilities of the MosDeF software suite, an extensive screening of functionalized dicyclopentadiene norbornene diacyl chloride (NBDAC) monomers was conducted. From the GCMC simulation results, the relative affinity of each functionalized NBDAC monomer towards water/ethanol can be quantified and the data used to train a classification machine learning model that can predict the polymers’ preference for each solvent species. The model is trained on molecular descriptors derived from the cheminformatics library RDKits and simulation system details, capturing intricate intermolecular interactions governing solvent adsorption on the functionalized polymer membrane material. Through rigorous validation and evaluation, the efficacy of the machine learning approach in accurately categorizing polymer materials based on their solvent preference is demonstrated. This interdisciplinary methodology bridges the gap between molecular simulation techniques and data-driven approaches, providing insights into the solvent affinity of functionalized polymer membrane materials with implications for membrane design and separation processes. Our findings underscore the potential of combining computational simulations with machine learning for efficient characterization of membrane material properties.
