A digital rock physics workflow for crystalline reservoirs: Developing digital twins through a geologically driven workflow

Reliable characterization of crystalline geothermal reservoirs requires linking rock microstructure to effective physical properties across scales, from pore to reservoir level. Digital rock physics (DRP) provides a promising framework by combining three-dimensional imaging with numerical simulation. However, established DRP workflows for sedimentary rocks are often insufficient when applied to crystalline lithologies. Low porosity, complex mineral intergrowths, fine inclusions, and alteration textures complicate phase identification and introduce biases in predicted elastic, permeability, and thermal properties, limiting established DRP workflows for low-porosity crystalline rocks. This study presents a geologically driven workflow for a granitoid rock sample from the Frontier Observatory for Research in Geothermal Energy (FORGE) site in Utah, USA. High-resolution X-ray computed tomography (XRCT) of cylindrical core plugs (10 mm diameter, 40 mm length) at 6.9 µm voxel resolution provides the basis for digital pore-scale analysis. Multiphase segmentation, i.e., assigning gray-scale intensities in XRCT volume to specific mineral phases, was performed by integrating grayscale-based thresholding techniques with geological constraints derived from thin-section petrography and scanning electron microscopy (SEM). This integrated workflow reduces misclassification caused by overlapping gray-scale intensities, partial-volume effects at phase boundaries, and unresolved microporosity. The resulting segmentation distinguishes pore space, quartz, feldspar, ferromagnesian minerals (amphibole, biotite), titanite, and accessory phases (zircon, opaque oxides, apatite). Initial digital twin analysis shows results that deviate from laboratory measurements for porosity and the determined P- and S-wave velocities. We suspect that assigning completely intact single-crystal properties to the segmented phases may be incorrect, as the microstructure provides clear information about mechanical stresses, e.g., undulatory extinction or mineral alignment. Additionally, the analyzed subvolume (400³ with an edge length of 2.76 mm) does not yet constitute a representative volume element (RVE) relative to the coarse feldspar grain size (1-3 mm). This results in the following challenges for a DRP workflow in relation to crystalline rocks compared to established sedimentary rocks: (1) XRCT scans of larger field of views to encounter the larger minerals within the granitoid sample, (2) assigning reduced elastic properties to the individual segmented mineral phases due to microcracks and fluid inclusions, (3) Preserving high-resolution imaging to resolve the small volumes of porosity (~1.2 %). We present a refined DRP workflow that addresses these challenges through multi-scale imaging strategies and microstructure-informed elastic property assignments, validated against laboratory measurements on FORGE crystalline samples.