(172b) Machine-Learning Assisted Development of PFAS and PFAS-Free Omniphobic Coatings

PFAS (Per- and polyfluoroalkyl substances) are widely used as coating materials because of their omniphobicity, repellency to both oil and water. However, concerns about their health and environmental impact have increased the demand for alternatives that offer similar beneficial properties. To that end, we use molecular dynamics (MD) to study the omniphobicity of PFAS and potential PFAS alternatives for textile materials. The MD derived contact angles are used to represent their repellency to oil and water, with a higher contact angle indicating greater repellency. The MD results are incorporated into a database to support a machine learning study which aims to predict new and effective PFAS alternatives.

Our MD simulation models a system consisting of a water or oil droplet, a CH₃-(CF₂)_n-1-CF₃ coating on a Nylon 6,6 (110) slab substrate. Various CF_x chain lengths (n = 2, 4, 6, 8) and a broad range of packing densities (D_p = 1.28, 2.55, 5.10 nm⁻²) are considered (Figure 1). The lowest contact angle for water droplet (~95°) is observed for the shortest PFAS and lowest D_p, indicating the least hydrophobicity. A similarly low repellency is observed for oil droplet at the same PFAS chain length and D_p. At this D_p, the contact angle for oil droplets increased by ~20° as the chain length increases from 2 to 8, and the contact angle for water droplet also increases by ~20° over the same chain length range. As D_P increases, the effect of chain length on repellency becomes less prominent for both oil and water. The best repellency is observed for the highest considered D_p (5.10 nm⁻²) and chain lengths of 6 and 8. The omniphobicity of PFAS alternatives including biobased materials will also be simulated. Finally, new PFAS alternatives along with chain length and optimal packing density will be predicted using machine learning.