High-throughput simulations and machine learning to predict miscibility in polymers

Abstract The miscibility of polymers in multicomponent systems plays a critical role in determining the performance of a wide range of materials, including polymer blends, composites, and solutions. However, predicting miscibility remains a significant challenge due to the complex interplay of molecular architecture, thermodynamic interactions, and processing conditions. Traditional approaches, such as Flory–Huggins theory and solubility parameters, provide valuable insight but are often limited in predictive accuracy and rely on costly experimental input. Recent advances in high-throughput simulation workflows have created new opportunities to explore large molecular design spaces and accelerate miscibility predictions. In parallel, machine learning (ML) approaches have emerged as powerful tools for capturing complex relationships between structure, composition, and phase behavior in polymer mixtures. Despite these advances, progress is limited by the high computational cost of generating reliable training data and the lack of standardized, automated pipelines for polymer miscibility data collection. Here, we review recent computational approaches to calculate miscibility in multicomponent polymer mixtures, and how high-throughput simulations and ML offer a robust predictive approach to quickly screen for the most promising molecules. Continued development of high-throughput simulation tools, combined with scalable data infrastructure, will be critical for enabling more complex and transferable ML models. Addressing these challenges will be key for the development of reliable data-driven approaches to polymer miscibility prediction.