Machine learning‐assisted design of low‐energy, high‐efficiency nonaqueous amine solutions

Abstract

Conventional aqueous alkanolamines suffer from high regeneration energy, corrosion, and solvent degradation, limiting large‐scale CO 2 capture. Here, a machine‐learning‐assisted strategy was developed to screen nonaqueous absorbents from a limited experimental dataset by correlating molecular structure with CO 2 desorption performance. The identified 2‐(butylamino)ethanol (BEA)‐tetramethylethylenediamine (TMEDA) system achieved a desorption capacity of 0.668 mol CO 2 ·mol −1 amine and a maximum desorption rate of 0.0513 mol CO 2 ·mol −1 amine·min −1 , while exhibiting good cyclic stability, low‐temperature performance, and low corrosivity. Its regeneration energy requirement was only 1.38 GJ·t −1 CO 2 , 62.3% lower than that of aqueous monoethanolamine. Quantitative 13 C NMR spectroscopy and density functional theory calculations revealed that BEA governs CO 2 fixation, whereas TMEDA acts as a proton acceptor, facilitates proton transfer, destabilizes the intermediate products, and lowers the heat of reaction. This work establishes a data‐driven and mechanistically informed framework for the rational development of energy‐efficient nonaqueous absorbents for CO 2 capture.