(122b) Understanding the Physics Underpinning Nanoparticle Dispersion in Polymers Via Statistical Interrogation

Polymer/nanoparticle (NP) mixtures can either have well-dispersed or agglomerated NPs depending on the used casting common solvent. Here, we use traditional statistical methods and machine learning (ML) classification models to elucidate the critical underpinning variables determining this behavior. Building on two databases, one sourced from over the existing literature and one created in-house, we conclude that the interaction strength between NPs and the solvent dielectric constant can reliably predict NP dispersion states. Physics based models then point to the key roles of charge stabilization of the NPs and the ability for a polymer to form a bound layer on the NP in this context. Finally, we also analyze the differences between our two datasets, and comment on the importance of using controlled datasets when training ML models or otherwise performing statistical analysis.