Process level optimization of materials for CO2 capture using multi-scale and machine learning accelerated simulations
\(^{1}\) University of Ottawa
\(^{2}\) University of Alberta
\(^{3}\) Svante
CO2 capture technologies are being explored to reduce greenhouse gas emissions. Solid adsorbent-based processes using porous materials are emerging as an alternative to current energy intensive solvent-based processes. The challenge is to find the right material that can minimize the energy and cost of CO2 capture. Here, metal organic framework (MOF) materials have attracted the most attention due to their chemical diversity and tunability. Researchers have computationally screened hundreds of thousands of MOFs and put forward promising candidates based solely on their equilibrium adsorption properties without knowing how these properties relate to the performance of the adsorbent at the process scale. In order to determine this, one must perform detailed, macroscopic process simulations on a specific gas separation process. In this work, multi-scale modeling that combines atomistic simulations and detailed process simulations was used to optimize materials for CO2 capture in a four-stage pressure swing adsorption system. Machine learning models were developed to accelerate both the atomistic simulations and the process modelling to achieve a practical scheme to optimize the composition and structure of a MOF that minimizes the energetic cost of CO2 capture (or other performance indicators). A previous screening of experimentally characterized MOFs that took over a year to complete identified a MOF with an energy of CO2 capture as low as 217 kWh/MT CO2. Using our new accelerated workflow, hundreds of MOFs have been identified with energies of less than 200 kWh/MT CO2 within a few days.