Chemistry

AI predicts which chemicals dissolve in supercritical carbon dioxide

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.


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

Machine learning 118 articles Explore Concept → Supercritical fluid Concept coming soon Solubility Concept coming soon

Source: Symbolic and domain-generalized machine learning for interpretable solubility modeling in supercritical CO₂