Computational and experimental design of solvents for CO2 capture

Abstract

At current rates of emission, the carbon budget for 1.5°C is running out in less than 10 years. Existing energy generating infrastructure, if operated at historical levels, will use up most of that budget. We thus need to decommission these power plants early or capture the carbon that they emit. Further, pathways to reach net zero targets by 2050 include negative emissions and carbon capture for hard to decarbonize industries. According to the IEA, carbon capture technologies still face challenges in optimizing their stability, fouling, emissions, regeneration energy requirements, selectivity, and CO2 binding capacity. Discovery of new materials can take on the order of 10 years or more, therefore we must accelerate the discovery process in order to create viable carbon capture technologies to meet critical CO2 emissions targets and mitigate climate change. Here we focus on creating an AI-based pipeline for accelerating discovery of new liquid sorbents for CO2 capture. Using cloud-based containerized computational chemistry workflows, we are building a predictive model of CO2 capture properties, ranking candidate materials, and filtering these materials using a parallelized CO2-capture assay platform. The adaptive filtering approach identifies and tests both single and blended sorbents, generating CO2 binding capacity, kinetics, and other physical chemical parameters. These data are delivered back to the model to improve its predictive capabilities with the end goal of generating new materials and solvent blends with improved performance for carbon capture.