Subsurface CO₂ mineral trapping

Carbon dioxide (CO₂) sequestration in the subsurface is important for achieving atmospheric decarbonisation. CO₂ can be stored in the pore spaces and through mineral trapping in clastic reservoirs (e.g. Equinor Sleipner project, offshore Norway) or through mineral trapping in mafic rocks (e.g. Carbfix Orca project, Iceland).

The geological community is at an early stage of understanding the carbon-binding mineral reactions that will permanently lock CO₂ in subsurface rocks. Measurement, modelling and verification of carbon-binding mineral reactions through analysis of QEMSCAN (automated SEM-EDS) data provides an opportunity to rapidly quantify and visualise these processes. Where QEMSCAN data are available across a basin (such as the Norwegian Oil & Gas Released Wells Initiative dataset), these data will inform regional screening of storage locations and support optimal design of subsurface CCS operations.

This talk will present results from modelled CO₂ sequestration reactions of mafic volcanic, volcaniclastic and clastic units, based on cuttings samples from eight Faroe Islands offshore exploration wells. This talk will also present results from CO₂ sequestration experiments we have carried out on cuttings and core samples from the Faroe Islands offshore exploration wells. Mafic rocks are of great interest for CO₂ mineral trapping due to fast reaction rates of the phases present in these rocks.

We acknowledge that the model results presented here do not account for a range of subsurface processes that will no doubt be key for adequate prediction of subsurface CO₂ carbon-binding reactions. These processes include but are not limited to: incomplete reactions due to grain size, permeability, fracture density and chemical variations, reaction inhibition by grain coating and pore plugging by produced phases and volumetric changes due to lower density of produced phases.

Through an iterative work programme of experimentation, diagenetic modelling and incorporation of published mineralisation results we are confident that model sophistication can rapidly be improved.