Autonomous High-Throughput Characterization of Liquid-Liquid Phase Behavior

Self-driving labs and data-driven formulation have outpaced the characterization of liquid-liquid miscibility and phase behavior, despite the role of such characterization in determining the stability and efficacy of complex formulations across diverse applications. Traditional characterization methods rely on labor-intensive visual inspection or single-proxy measurements that lack chemical generality, limit throughput, and provide only partial insight into phase behavior. Here, we report an automated platform that enables continuous, high-throughput screening of liquid-liquid phase behavior across diverse fluid chemistries. The device integrates an asymmetric capacitance sensor that is sensitive to the emergence and motion of phase boundaries, along with multi-angle turbidimetry that quantifies cloudiness and emulsion stability, in a single flow-through chamber. We demonstrate the classification of chemically diverse binary mixtures, resolution of real-time phase separation kinetics, and identification of partial miscibility across compositions and temperatures. For multicomponent systems, we employ Gaussian-process-based active learning to autonomously map ternary phase diagrams in ∼2 h, together with a nonlinear programming framework that extracts tie lines in ∼5 min per line. By unifying miscibility classification, kinetic characterization, and thermodynamic mapping in a single automated workflow, the platform enables comprehensive phase behavior screening at the scale and throughput required for autonomous formulation discovery.