(394h) Active Learning for Polymer Blend Miscibility with Flory-Huggins Theory

Polymer blends are increasingly being used for engineered solutions in materials development. Miscibility of polymer blends imparts an equal distribution of individual components throughout the volume of the blend, leading to a more uniform distribution in the material after formulation, processing, or cure. Given that three or more polymers can exist in such a blend, the search for compositions that are miscible demands numerous experiments. Active learning to guide experiments is done using Flory-Huggins theory as a surrogate model to map miscible and immiscible regions. Experiments are chosen using surrogate model uncertainty. The method is shown to greatly decrease the number of experiments needed to generate ternary phase diagrams for blends of homopolymers relative to space-filling methods.