£950k Awarded to Advance Carbon Mineralisation Research
£950k awarded to advance carbon mineralisation research, supporting climate solutions and sustainable manufacturing.
Funded through the NERC Pushing the Frontiers of Environmental Research programme, this new three-year project will explore how to better predict and control the formation of carbon-negative carbonate materials from CO₂, with applications in sectors such as construction, polymers and cosmetics.
The project includes Oxford EARTH Co-Investigator Associate Professor Julie Cosmidis, alongside collaborators including Prof Ros Rickaby and Dr Luca Stigliano (Earth Sciences), and Prof Garrett Morris (Department of Statistics).
By combining high-throughput mineralogy with machine learning, the research aims to better understand how environmental variables shape carbonate formation. This could help reduce reliance on resource-intensive materials such as cement and concrete, lowering the environmental impact of mineral extraction.
The project is due to begin in September.
Project Details
Title: "Decoding carbonates: towards predictive control of carbon mineralisation through high-throughput mineralogy and machine learning"
Project summary
Carbonates are climate-controlling minerals, acting as long-term sinks in the biogeochemical carbon cycle, and serving as stable carbon stores in many carbon capture, utilisation and storage (CCUS) technologies. Carbonates are also increasingly used as carbon-negative materials in applications such as fillers for construction (cements), polymers or cosmetics, potentially enhancing the economic profitability of CCUS. The design of effective CCUS and material applications requires the ability to predict and control both the rate at which carbonates form and the mineralogical properties (e.g., crystal structure, size, morphology) of the precipitated particles. These factors can significantly impact the efficiency of carbonate mineralisation and long-term stability of its products, as well as industrial usability.
However, we still lack a comprehensive understanding of how multiple environmental variables control carbon mineralisation. Indeed, the crystallisation of carbonates is influenced by many physicochemical variables such as temperature, pH, supersaturation and salinity, as well as a wide range of inorganic and organic species (called additives) that can act as inhibitors or promoters of nucleation and growth. While the impacts of these factors have been studied individually, their combined effects remain poorly understood. Indeed, our current knowledge is largely based on empirical studies that investigate these factors in isolation, producing findings that are difficult to extrapolate to different conditions, and failing to capture the complexity of natural and engineered systems where multiple variables interact.
This is a typical example of a complex, multivariable system where artificial intelligence and Machine Learning (ML) approaches could be transformative. In the context of carbonate mineralisation, ML has the potential to uncover hidden patterns in experimental data, identify key controlling parameters, and predict mineralogical outcomes across a wide range of geochemical conditions. However, the application of ML in this field is currently limited by the scarcity of large, high-quality mineralogical datasets derived from well-controlled and comparable carbonate precipitation experiments. To fully leverage ML’s potential, we must first overcome this data bottleneck by generating comprehensive and consistent empirical mineralogical data at scale.
We will address this challenge by deploying newly developed high-throughput methodologies that allow us to rapidly perform and characterise thousands of carbonate mineralisation experiments. These experiments will span a wide multi-dimensional space of physicochemical variables and include hundreds of organic additives. Using high-throughput microscopy and Raman spectromicroscopy, we will analyse carbonates forming in situ within these experiments, measuring particle size, morphology, mineralogical composition and structure, as well as growth rates. The resulting datasets will be used to train and test ML models capable of predicting mineralogical outcomes from any given set of variables and additives, and identifying key parameters critically controlling carbonate mineralisation under different geochemical environments.
Additionally, we will gain atomic-level mechanistic insights into how inorganic and molecular factors influence carbonate crystallisation by employing high-resolution imaging and kinetic Monte Carlo modelling of particle growth for a subset of mineralisation experiments. Altogether, these approaches will provide a systematic and mechanistic understanding of carbonate mineralisation, enabling predictive control over its outcomes.
The proposed research establishes a new empirical paradigm for the study of multivariable mineralisation systems. By generating the first large-scale, systematic dataset on carbonate crystallisation under complex geochemical conditions, the project will lay the foundation for predictive, mechanism-informed control of carbon mineralisation processes, strengthening our fundamental understanding of carbonate geochemistry in the environment, and supporting the implementation of applied climate solutions and sustainable manufacturing.
