Machine learning-aided optimisation of adsorption materials and processes for carbon capture applications

Abstract

Carbon capture and sequestration (CCS) is needed to mitigate anthropogenic CO2 emissions. Adsorption using solid porous materials, in particular metal-organic frameworks (MOFs), stands out as a promising technology option for this task. By pairing the right adsorbent with an optimal adsorption process cycle, one can potentially separate CO2 from any carbon source. However, given the near-infinite design space of MOFs, this task is extremely challenging. Computational models help us identify promising adsorbents for commercial scale-up by predicting their performance in the industrial separation process. These predictions are based on material property data obtained from either molecular simulations or experimental methods, and are sometimes applied to thousands of materials in high-throughput screening (HTS) workflows. However, to-date, none of the technologies identified by these approaches have been deployed commercially. This gap between computational predictions and practical application can be attributed to several factors, including uncertainties in experimental systems, inaccuracies in multiscale models, and a lack of industrially viable materials for large-scale CCS, among other challenges. In this thesis, we present three peer-reviewed published articles (in chapters 2 through 4) and one pre-print article (in chapter 5). Chapter 2 explores uncertainty in experimental syntheses and measurements of adsorbents, and how these uncertainties impact process-level modelling outcomes. Chapters 3 and 4 seek to understand whether purely computational, multiscale HTS workflows can consistently identify small subsets of high-performing materials from hundreds-to-thousands of possible candidates, given the lack of a community standard for modelling adsorption equilibrium using molecular simulation. Finally, in chapter 5 we develop a novel computational approach to generate materials in silico that satisfy targeted property constraints, with the objective of driving research towards identifying industrially-viable adsorbents for CCS. Overall, this thesis aims to establish a more robust framework for the discovery and deployment of adsorbents, ultimately contributing to the development of more effective and scalable CCS technologies.

Date of Award	6 Jan 2025
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Flor Siperstein (Supervisor) & Lev Sarkisov (Supervisor)