An Integrated Framework: Inverse Design for Optimal Amine Solvent using Reinforcement Learning and Enhanced CO₂ Chemical Absorption Processes

This study proposes an integrated framework that combines inverse material design via reinforcement learning (RL) and process optimization to determine the optimal solvent for CO₂ chemical absorption process (Figure 1). The framework addresses the challenges posed by Computer-Aided Molecular and Process Design (CAMPD) and efficiently explores the design space characterized by infeasible subregions and a highly nonlinear relationship between process and molecular structure. The RL model, when combined with combinatorial chemistry, inversely designs amine solvent based on target properties such as CO₂ absorption capacity and solubility. Then, conductor-like screening model for real solvents (COSMO-RS) predicts the phase behavior through thermodynamic analysis and evaluates whether the solvent can achieve the desired CO₂ removal. This novel approach provides an efficient and systematic way to design an effective solvent and improve the CO₂ absorption process.