AI Insight
Researchers developed a machine learning approach combining symbolic regression with domain generalization to predict the solubility of various compounds in supercritical carbon dioxide (CO₂). The model generates interpretable mathematical equations rather than black-box predictions, allowing scientists to understand the physical relationships governing solubility behavior. This method demonstrated improved accuracy and generalizability across different chemical compounds compared to traditional empirical correlations.
Why it matters
Supercritical CO₂ is widely used in pharmaceutical manufacturing, food processing, and green chemistry applications, where accurate solubility predictions are essential for process design. This interpretable modeling approach could accelerate development of CO₂-based extraction and separation processes while reducing the need for extensive experimental testing.
Understand the Science