The urgent need to mitigate atmospheric CO2 levels has sparked growing interest in developing efficient methods for post-combustion carbon capture. However, identifying suitable molecular systems for this purpose remains a formidable challenge due to the vastness of chemical space and the need for accurate quantum chemical calculations of CO2 binding properties. To address this challenge, we propose a novel procedure to find high-performing amine sorbents for post-combustion carbon capture through the combination of quantum chemistry, active learning and generative modeling. Specifically, we will leverage our previously published benchmarking framework for active learning in regression problems (10.1088/2632-2153/addf11) to systematically explore amine-based CO2 capture sorbents with as few quantum chemical calculations as necessary. Through this data-efficient process we accumulate a compact dataset of amine sorbents for subsequent generative modeling. With this procedure we aim to enable more sustainable and targeted design strategies for carbon capture.
Elizaveta Surzhikova