Images of Ketton Limestone after drainage and imbibition. (a) A 3D rendering of the long composite core after imbibition. Each unique disconnected ganglion of residual scCO 2 is coloured in a different colour. The large 

Context in source publication

Context 1

... microtomography has been recognised since its outset as a potentially transformative technique in the analysis of flow in geological systems, allowing for the “study [of] contained systems under conditions of temperature, pressure and environment representative of process conditions” 1 . Progress has, however, been hampered by limits on sample size, computing capacity, image resolution and material technology. We have developed a technique which allows for the imaging of fluid flow and distribution in rock-cores, at the pore scale, at conditions (pressures and temperatures) representative of subsurface flow. We use this imaging technique for two specific applications; the examination of capillary trapping and the measurement of contact angle. When carbon dioxide is stored in the subsurface for Carbon Capture and Storage (CCS) applications four principal mechanisms are available to limit its spread in the subsurface; stratigraphic trapping, mineral trapping, solubility trapping and residual trapping. The only rapid mechanism is residual trapping, occurring when water invades the pore-space after the injection of super critical CO 2 , which is of great interest when designing carbon storage schemes. We replicate these conditions and use micro-CT scanning to compare the size distribution of residual ganglia to predictions made from percolation theory. The most important macroscopic parameters for multiphase flow in porous media, such as capillary pressure and relative permeability, are fundamentally controlled by the pore-scale topology, interfacial tension and contact angle. The measurement of contact angle has, however, traditionally been limited to pore mineral surfaces or core-scale indirect measurements, such as the use of the Amott wettability index 2 . As surface roughness can have a large impact on contact angle in real systems 3 the application of these measurements to the range of angles found in reservoir and aquifer rocks with heterogeneous surface roughness, mineralogical composition and pore topography remains unclear. We present a new method for the measurement of contact angle in real systems at representative subsurface conditions using non-invasive micro-CT scanning. A small diameter (4-6.5 mm) rock core was placed within high pressure high temperature carbon fibre micro-CT coreholder which was attached to high precision syringe pumps by flexible flow lines. Flexible flow lines were used as they provide very little lateral load to the coreholder during image acquisition, which would degrade image quality. Temperature within the flow cell was maintained using an external Kapton insulated flexible heater and monitored using a thermocouple sitting in the confining annulus of the cell. The temperature was set at 50 o C and seen to be constant throughout the experiment. Potassium Iodide was used as an ionic salt as it has a high atomic weight, causing a high x-ray attenuation coefficient, and the salinity chosen (7 wt%) was representative of aquifer salinities. The fluids were vigorously mixed in a high pressure reactor prior to fluid injection in order to maintain chemical equilibrium and prevent chemical reaction within the core during the experiment. Further description of the experimental methods used in this paper can be found in Andrew et al. 4 , and further description of the image analysis techniques used to measure contact angle can be found in Andrew et al. 5 . Prior to CO 2 injection the core was saturated with brine and pressurised to 10 MPa to ensure a completely saturated initial state. Tomographies of the partially saturated core were acquired after drainage (CO 2 injection) and imbibition (brine injection) then reconstruction using proprietary software on the Versa system from a set of 400 radiographs. The reconstructed tomography would consist of a volume of roughly 1000 3 voxels, with and individual voxel size of around 6.6 μ m. In order to increase the examined volume in the capillary trapping study, composite volumes were created by scanning multiple points along the core sequentially then stitching together five overlapping sections into one large cuboidal volume of around 900 × 900 × 3,300 voxels. Each composite volume took around 90 minutes to acquire. After acquisition the images were filtered using a non-local means edge preserving filter . They were then corrected for any beam hardening or softening artefacts due to the use of a polychromatic x-ray source and segmented using a variant of the watershed algorithm acting on the grayscale gradient of each of the images, seeded using a 2D histogram 7 , eliminating much of the arbitrary voxel misidentification associated with simple threshold segmentation 8,9 . In this segmentation step the brine and rock were treated as one phase, and CO 2 the other. The segmented image was then analysed in 3D in order to identify and measure the volume of each unique disconnected ganglion, which was then labelled (figure 1). The techniques used in the acquisition of images for contact angle measurement were similar to those described above, with some small changes. A smaller voxel size of 2.013 μ m was used so that the CO 2 -brine interface and the three phase contact point were better resolved. In order to increase the signal to noise ratio at these high resolutions, more projections were used during the scanning process, with a single volume reconstructed from a set of 1,600 projections. Only a single volume was used, so the total analysed volume was around 1000 3 voxels, representing a volume of around 2mm × 2mm × 2mm, as opposed to the 6mm × 6mm × 22mm analysed in the capillary trapping study. Significant additional image analysis was required for the measurement of contact angle. First the sample was filtered and globally segmented as described above for the capillary trapping study and each disconnected ganglion identified using 3D analysis. A subvolume was then extracted around each unique ganglion and the filtered data was resegmented into three phases using the same 2D histogram based method detailed above, this time treating each phase separately. The edges of each phase were found on this new segmented image using a 3D sobel filter 10 . The intersection of the edges of all three phases was labelled as the contact line which could be traced in3D. The data was the resampled onto a plane perpendicular to the contact line at a particular point. Finally a 3D bilinear filter was applied to the resampled slice to eliminate possible angular quantisation due to the voxelized nature of the image. The contact was then measured manually on the resampled data using a 3D angle measurement tool according to the best interpretation of the tangential direction of the relevant surface at the contact line (figure 2) . No effort to “smooth” the surface was made. Measurements were performed at 300 points randomly selected along the scCO 2 -brine-rock contact lines of different ganglia. Contact angle was not measured on the segmentation as it was highly dependent on the detail of segmentation close to the contact line, where we would expect the segmentation to be least accurate. The segmented partially saturated images were analyzed by counting the number of voxels of residually trapped scCO 2 to find the proportion of the rock volume occupied by trapped scCO 2 – the capillary trapping capacity. This is related to residual saturation by dividing the capillary trapping capacity by the porosity. Significant proportions of scCO 2 were trapped as a residual phase with a residual saturation of 0.203 ± 0.012 4 . The ingress of brine into a scCO 2 saturated core is a percolation like process 11 , and predictions can be made about the size distribution of the residual ganglia, specifically that the number n of clusters of volume s (as measured in voxels) should scale as , where τ is the Fisher exponent 12 . Network modelling on regular cubic lattices found the value of the fisher exponent is around 2.189 13 . The Fisher exponent can be found from real data by plotting a binned ...