Autonomous Mapping of Ternary Polymer Solution Phase Diagrams

Ternary polymer-solvent phase diagrams are technologically important, enabling synthesis, self-assembly, formulation, and coating processes. Recently, machine learning (ML) models have demonstrated the ability to predict co-existence curves within 1-3 °C, both interpolating within, and extrapolating beyond, the polymer-solvent training data. Expanding the utility of such models however requires expansive formatted datasets based on FAIR data principles. The goal of this work is to develop an instrument that autonomously and efficiently maps polymer 1-polymer 2-solvent phase diagrams to generate such databases that can be augmented with literature to train an accurate ML model. The design, software, function, and algorithms of the autonomous instrument are described, and demonstrated on the polyacrylamide-poly(ethylene glycol)-water system.