ML-Accelerated Design of Amine-Grafted LDHs for Direct Air Capture

This project develops a machine-learning–accelerated framework to design high-performance amine-grafted layered double hydroxides (LDHs) for direct air capture (DAC) of CO₂. An expanded density functional theory (DFT) dataset systematically varying amine chemistry, grafting density, LDH composition, and humidity effects will be used to train an optimized random-forest model through rigorous descriptor selection and active learning. The trained model will enable large-scale virtual screening of thousands of amine/LDH combinations, dramatically reducing reliance on costly DFT calculations. Ultimately, the project will identify and prioritize a small set of optimized sorbent configurations with clear design guidelines, directly supporting the advancement of DAC and broader CCUS technologies.