Understanding CO2 at aqueous interfaces using atomistic machine-learned potentials

In today’s scientific landscape, no molecule is more synonymous with risk and catastrophe than carbon dioxide (CO2). Over the past several decades, billions of tons of CO2 have been pumped into the atmosphere, giving rise to drastic increases in global temperatures alongside a myriad of other detrimental effects. Understanding the way in which CO2 interacts with its environment is crucial for being able to mitigate rising CO2 levels and some of its more harmful effects. Many of these environments are aqueous in nature; accordingly, it is vital that we can describe the way in which CO2 and H2O interact under various conditions. This PhD constitutes new insights into the way these two molecules interact with one another. Utilising machine-learned interatomic potentials (MLIPs), we provide a new understanding of how CO2 behaves at the air-water, liquid-water, and solid-water interfaces. First, we demonstrate the existence of a new type of reaction mechanism affecting gaseous CO2 molecules adsorbed at the air-water interface. This surface mediated mechanism involves the partial dissolution of the reaction site at the topmost water layer, imparting bulk-like thermodynamic properties on an inherently interfacial process. Second, we show the efficacy of MLIPs for estimating interfacial tensions and identify the build-up of a liquid-like CO2 monolayer at the water interface. Finally, we show that CO2 uptake in solvent saturated nanoporous carbon environments occurs due to favourable solute-wall interactions which out compete those between the solvent and the pore wall. This thesis represents a step forward in understanding the behaviour of CO2 at aqueous interfaces, leading to the uncovering of new fundamental physicochemical insights as well as providing clarity on experimental measurements and observations.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2025

Author dc:creator

• Brookes, Sam

Advisors dc:contributor.advisor

• Michaelides, Angelos
• Schran, Christoph

• chemistry
• machine learning
• molecular dynamics
• physics

Licence

• https://www.repository.cam.ac.uk/bitstreams/d597ea3f-f662-4846-b617-3cb4ec5e7b95/download
• https://creativecommons.org/licenses/by/4.0/

Language dc:language

eng

Author Identifier

0000-0003-2821-7255

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/397673