Chemical space analysis and property prediction for carbon capture solvent molecules

Abstract

We present a new chemical representation (the CCS fingerprint) and data set (ccs-98) for carbon capture solvents. We then assess the chemical space, data availability and utility of common machine learning algorithms for high throughput virtual screening in the carbon capture solvents field. This is an area of growing importance, as carbon capture and storage is part of the road map towards net zero for many countries around the world. A major class of commercial carbon capture technology involves using solvents, which are commonly blends of amines and N-heterocyclic molecules in water. Whilst these blends have proved valuable, there is an increasing need to identify new candidate molecules which are more efficient and improve performance. We found that the CCS fingerprint can out-perform other common chemical representations when combined with standard machine learning approaches for classifying molecules based on absorption capacity. We demonstrate models achieving classification accuracy for absorption capacity of over 80%.