IAP-24-090

Exploring CRM Mineralisation Processes in Layered Intrusions with AI: A Corescale to Microscale Journey and Back Again

The transition to a low carbon future, relies heavily on an increased usage of metals. Examples include rare earth elements for the permanent magnets used in wind turbines platinum group elements leveraged for catalysis, and lithium for battery technologies. Furthermore, due to socioeconomic and geopolitical forces, the demand for these elements drives their status as critical to the Green Economy; hence their designation as Critical Resource Materials (CRM). Most CRMs occur in trace-to-minor concentrations within metal oxides and sulphides in orebodies mined for conventional metals such as iron, titanium, and copper, as well as other natural resources. The grand challenge is that the abundance of CRMs in these systems is often unknown and unaccounted for(Mudd et al., 2017; Nassar et al., 2015; Werner et al., 2017). Market adoption of automated, high-throughput technical scanning of drill core avails deep interrogation of CRM distributions and behaviours in ore systems throughout entire drill holes. Moreover, developments in machine-learning and artificial intelligence can help to uncover cryptic relationships in these large multivariate datasets, providing new means to develop and test hypotheses concerning CRM distributions in ores and the processes governing their formation. A considerable outstanding barrier to fully leveraging core scanning data to tackle CRM related problems is how to bridge these relatively coarse-resolution observations with detailed datasets of the micro and nano spatial scale; CRMs commonly occur in grains smaller than the commercial operating resolution of core scanning systems. The microscale population is often overlooked and understudied, yet provides petrological and paragenetic context to both coarser ( macro-scale/scanning scale)s and finer (nano-scale) observations. Linkage of the disparate spatial scales through this microscale bridge can maximize the depth, breadth, and value of information that can be extracted from core scanning systems, bringing crucial advancements, such as automatic mineral and lithology classification, that stand to improve the efficiency and effectiveness of resource estimation and mining activities. Herein lies an opportunity to develop protocols and methods to quantify CRMs in environmentally and economically important ore systems. Working in partnership with GeologicAI, the student will:
1) Address knowledge and data gaps in automated core logs of CRM- bearing materials
2) Develop multiscale protocols for linking core scale features to micro and nano scale assessment, and develop tools to connect these data sets into a unified digital twin
3) Develop and explore these concepts using publicly accessible cores from multiple ore-bearing layered intrusions as well as other ore systems.

The above questions will be explored by linking core scanning and microanalytical (SEM EDS, EBSD, FIB-nT and S/TEM microscopy) datasets from ore-bearing drill cores, combined with machine learning methods ((Bérubé et al., 2018; Khudhur et al., 2024; Radulescu et al., 2024; Tominaga et al., 2021) _ ). Initially, full core datasets combining Hyperspectral (VNIR/SWIR), Laser Profiling (LiDAR), RGB Imaging, XRF, and Magnetic Susceptibility will be classified into mineralogical and lithological types using GeologicAIs proprietary methods. The student will then develop rules for defining regions of high classification uncertainty and/or ambiguous/ undifferentiated regions. These metrics will then be used to select samples for thin sectionings for multiscale optical and electron microscopy measurements. These statistically determined regions of interest will be further characterized with large -area automated chemical (energy dispersive spectroscopy) and crystallographic (electron back scatter diffraction) mapping, combined with machine learning workflows, to identify micrometre to nanometre scale minerals including CRMs, which would be inaccessible through the too-coarse core scanning datasets. Analysis and visualisation of these micrometre phases will be used to refine our understanding of the economic and environmental potential for these CRM rich ore systems.

Year 1

• Getting started and understanding the project
• Literature review
• Identifying appropriate methods for to collection of relevant data.
• Training in working with the GeologicAI core scanner and its output data
• Basic training in electron microscopy and microanalysis
• Introduction to coding skills required for machine learning data analysis.
• Secondment to GeologicAI headquarters in Canada
• Laboratory analysis (Optical petrology and core scanning)

Year 2

• Processing and selecting subsections of core
• Sample preparation
• Secondment to GeologicAI headquarters in Canada
• Laboratory analysis (Modal Mineralogy, High-res SEM EDS)

Year 3

• Laboratory analysis
• Data processing and interpretation
• Development of conceptual model
• Secondment to GeologicAI headquarters in Canada

Year 3.5

• Any final data collection/processing
• Thesis and paper writing

This project will equip the student a range of analytical and transferable skills which are desirable for careers in research or industry.
Research Methods
Fieldwork at the case study sites will be conducted with the supervisory team. Full training will be given in all of the laboratory and machine learning techniques to be used in the project, mainly at the University of Glasgow but also in collaboration with some external facilities.
Researcher Development
• Technical & personal skills development will be undertaken with guidance from doctoral advisors and within the framework of the DTP Researcher Development Statement.
• Researcher developmental training will be provided by IAPETUS2 and supplemented by the University of Glasgow.
• The School of Geographical and Earth Sciences at the University of Glasgow (GES) has a large research student cohort (currently ~60 PhD students) that will provide peer-support throughout the research program.
• The scholar will participate in GES’s annual progression assessment and post-graduate research conference, providing an opportunity to present their research to postgraduates and staff within the School, and to also learn about the research conducted by their fellow postgraduate peers.
• Additionally, skills in NERC’s ‘most wanted’ list for PhD student training will be developed, including in multi-disciplinarity, data management, numeracy, and fieldwork, in addition to principles and practice of various other laboratory analytical techniques.
• Training and experience in national and international conference presentations, and preparation and submission of papers to international peer-reviewed journals will also be provided. Any final data collection/processing
Thesis and paper writing