Machine Learning-Boosted Design of Ionic Liquids for CO₂ Absorption and Experimental Verification

Area of Science:

• Materials Science
• Chemical Engineering
• Computational Chemistry

Background:

• Efficient carbon dioxide (CO2) capture is vital for a carbon-neutral society.
• Ionic liquids (ILs) show promise for CO2 absorption, but optimizing their cation-anion combinations is challenging due to their complex nature.
• The vast chemical space of ILs makes empirical screening for optimal CO2 absorbents impractical.

Purpose of the Study:

• To explore a large dataset of IL candidates for high CO2 solubility using computational methods.
• To identify key molecular features governing CO2 absorption in ILs.
• To synthesize and experimentally validate a novel IL with enhanced CO2 capture capabilities.

Main Methods:

• Applied electronic structure informatics and machine learning to analyze 402,114 IL candidates.
• Utilized quantum chemistry calculations on isolated ion fragments to determine feature variables.
• Employed the wrapper method to identify important molecular geometries and electronic states.
• Synthesized the top-performing IL candidate and experimentally measured its CO2 solubility using a magnetic suspension balance.

Main Results:

• Electronic states of ionic species were identified as critical factors for CO2 physisorption capacity in ILs.
• A machine learning model successfully narrowed down the vast IL candidate pool.
• The synthesized IL, trihexyl(tetradecyl)phosphonium perfluorooctanesulfonate, demonstrated superior CO2 solubility compared to the previous benchmark IL.

Conclusions:

• Electronic structure informatics is an effective approach for designing ILs with high CO2 absorption.
• The identified electronic properties are essential for optimizing ILs for carbon capture applications.
• The newly synthesized IL represents a significant advancement in CO2 capture technology.