(709g) Discovery of New Surfactants Used in Firefighting Foams By Active Learning

Aqueous Film Forming Foams (AFFF) used to suppress liquid fuel fires contain fluorinated surfactants that are being banned internationally due to their persistence and toxicity. New surfactant solutions for replacing AFFF are being sought, and fuel transport at the water-fuel interface has been identified as a potential key mechanism affecting fire extinguishment performance. Traditional design approaches, however, are limited in exploring the vast chemical space of surfactants. In this work, we initially collected a dataset containing 201 samples quantifying octane transport resistance of surfactants accumulated water-fuel interface generated by coarse-grained molecular dynamics simulations. The dataset encompasses a variety of surfactant types, including cationic, anionic, zwitterionic, non-ionic, sugar-based, and Gemini surfactants. For design purposes, we focused on hydrocarbon surfactants with identical tails, varying only the headgroups, as our previous work indicated that the surfactant tail has negligible impact on fuel transport. Subsequently, we employed an active learning protocol that combines a Graph Convolutional Network (GCN) model with Bayesian optimization to predict interfacial resistance and identify top-performing candidates efficiently. Notably, surfactants featuring beta-D-glucuronic acid heads show excellent performance in blocking octane transport; the resistance of their monolayer at interfaces is equivalent to that of ~ 3800 nm-thick water layer. In summary, this study not only enhances our understanding of fuel transport across the water-fuel interface but also identifies promising new surfactant solutions for fire suppression. Furthermore, the generalizable design framework we established can be extended to other surfactants, such as siloxane surfactants.