XCT data of samples S1 to S5 were collected using a Zeiss Xradia VersaXRM 410 microscope operated at 100 kV and 125 μA (no beam filtration), and the reconstructed datasets have a spatial resolution of 3.8 μm/voxel (3D equivalent of a pixel). All samples were imaged before and after each experiment. XCT data of sample S6 was collected using a BP proprietary XCT system [double-helix scanning trajectory (Varslot et al., 2012)], at Sunbury-on-Thames (United Kingdom), operated at 80 kV and 120 μA. The sample was imaged with a spatial resolution of 5.57 μm/voxel (S6A) between each experimental stage, and higher resolution (1.7 μm/voxel, S6B and S6C1-C3) images collected from a smaller region. The data were reconstructed using an algorithm that corrects for geometric motion errors via passive auto-focus (Kingston et al., 2011) and iterative reprojection alignment (Latham et al., 2018). For more information on the XCT acquisition characteristics of the raw data and the analyzed subvolumes please refer to the Supplementary Material.

To aid the reader, XCT resolution is color-coded in all figures as green (5.57 μm/voxel), light blue (3.8 μm/voxel), and orange (1.7 μm/voxel). The image processing workflow was performed in Thermo Scientific Avizo^® (v 2019.2). All scans for each sample were spatially registered to enable direct comparison of the microstructure before and after each temperature treatment. An anisotropic diffusion filter was employed to denoise all low and medium resolution datasets (Weickert et al., 1998), but was not required for the highest resolution data. Cylindrical and cuboidal subvolumes were extracted from the samples (Figures 1A,B) and used for the XCT image analysis.

Sub-micron scale features that are challenging to quantify with high accuracy in the 3D image can be captured through direct correlation of volume image data with high-resolution 2D microscopy after image registration. After the hydrothermal experiments, sample S6 was mounted in a block of resin and polished to expose a planar surface that lies within the 3D tomograpic volume. Backscatter SEM imaging of the carbon-coated polished block was carried out uisng a Hitachi SU-70 Field Emission Gun (FEG) SEM scanning electron microscope with an energy-dispersive detector (EDS) at an acceleration voltage of 20 kV and measured beam current of 0.6 nA.

Automated quantitative mineralogy (AQM) was performed on a Zeiss Gemini SUPRA 40VP field emission scanning electron microscope coupled with a Bruker XFlash^® 60 EDS detector. Images were collected using a high-resolution raster, with a step size of 3.5 μm at a 255x magnification and a 20 kV acceleration voltage. Dwell time was adjusted to capture approximately 20,000 counts per pixel. The 2D section image was then registered to the corresponding slice of the 3D data to support the XCT segmentation (Figure 2).

The XCT results show that the initial porosity of the cylindrical subvolumes was ∼5–7% with the cubic subvolumes showing slightly higher values (∼10%). This validates the approach for subvolume selection and suggests that the large subvolumes (relative to sample size) are an appropriate REV. This difference is caused by the smaller volume of the analysis and the natural heterogeneity within the samples. Both sample sets show no change in porosity at 100°C, and only modest (<1% reduction, cubic sample only) at 150°C. Porosity decreases significantly at the experimental temperatures of 200, 225 and 250°C (Figure 3A).

At temperatures >200°C the bulk porosity change hides additional complexity. The S4 (cylindrical subvolume) shows a reduction in porosity from 7.1 to 6%, and in S5 porosity shows a very similar decrease from 6.3 to 5.1%. The ceasing of further porosity loss is associated with grain coating growth (Figures 3A,D,E). Local increases in porosity are caused by the development of dissolution pores (Figures 3B,C).

Grain coating volume increases from 15.1 to 16.5% for sample S4 and, from 15.5 to 16.8% for sample S5 (Figure 3A). The 3D observations indicate increase of both the coating thickness and coating continuity (Figures 3D,E). Previous microscopic analysis has identified this as precipitation of authigenic chlorite coatings (Charlaftis et al., 2021). The porosity and grain coating volume measured on the smaller cubic subvolumes show the same trend as the larger cylindrical subvolumes, but with small differences (mean 3.6% difference in porosity, 1.5% difference in grain coating volume) because of the heterogeneity of the local microstructure.

A comparison of the pre- and post-reaction pore network and permeability models of each sample show the effect of these alterations, in terms of pore connectivity, pore and throat size and, absolute permeability (Figure 4). In agreement with the total porosity, up to temperatures of 150°C we see negligible changes in the volume of the connected pore network (Figure 4B), or in the mean pore diameter, although there is still some pore isolation and minor breaks in some areas of the pore network. While there is some heterogeneity in the initial permeability, controlled by the sample structure, no significant change in permeability is seen in samples subjected to temperatures up to 150°C.

As the temperature increases, we see volume reductions of the connected porosity of up to 30% (reduction from 7.2 to 4.9% at 250°C), accompanied by reductions in both mean pore diameter (9% reduction, from 126 to 115 μm at 250°C) and throat radii (15% reduction, from 27 to 23 μm at 250°C), and mean co-ordination number (from 3.3 to 2.5 at 250°C). Consequently, the observed pore network changes start to reduce permeability (∼2.5 mD reduction at 200°C; ∼ 4 mD reduction at 225 and 250°C) (Figure 4A). Visualisation of the fluid velocity magnitude maps before and after treatment at 250°C shows changes to the location of the main flow pathways (Figure 4B).

Compared to the single-stage thermal treatment, the multi-stage thermal treatment allows us to apply our new understanding of temperature-dependent behavior to a scenario that simulates progressive burial.

Figure 5A presents the porosity, grain coating, and microporous material evolution with increasing temperature for each of the extracted subvolumes (e.g., cuboids and 500³ voxel cubes). As with the single-stage experiments, negligible change was observed between 0 and 100°C and only temperatures ≥100°C are shown here. Porosity reduction initiates at 100°C, with the main reduction window occurring between 150 and 200°C, irrespective of subvolume size and scan resolution. The porosity in sample S6A (5.57 μm resolution) reduces from 8.6% at 100°C to 8% at 150°C to 6.5% at 200°C. With higher spatial resolution and a smaller scan volume (S6B, 1.7 μm resolution), we see higher absolute values (a result of the higher resolution) and a reduction from 14.2 to 13.2%–9.8% over the same temperatures; a 25-30% reduction at both image resolutions. The smaller cubic subvolumes (1.7 μm resolution, S6C1-S6C3, taken from areas of low, medium and high initial porosity) show the same trend in porosity reduction, but have greater variability in both the initial value and the reductions because of the greater heterogeneity preserved between these smaller volumes.

Above 200°C porosity reduction is retarded (as for the single-stage experiments), with reductions from 6.5 to 5.5% (5.57 μm resolution) and from 9.8 to 9% (1.7 μm resolution), and two the cubic subvolumes (S6C1 and S6C2) show no reduction at all. This porosity loss retardation can also be seen in the along-axis evolution of the connected porosity of sample S6B (Figure 5B). In all cases, the reduction of the connected pore volume occurs between 150 and 200°C.

Grain coating volumes show little change below 150°C, and increases at higher temperatures are smaller than those seen in the single-stage experiments, from 12.5 to 13.7% (S6A, 5.57 μm resolution) and from 12.8 to 13.3% (S6B, 1.7 μm resolution). As in the single-stage experiments, the coating volume increase is associated with the development of authigenic chlorite coatings (Figure 5C), and little net change in porosity. The SEM analysis revealed a progressive loss of the rhombohedral morphology of siderite crystals leading to replacement by chlorite, as well as precipitation on clean host-grain surfaces. Pre-experimental X-ray diffraction (XRD) and AQM analysis have shown an initial chlorite content of less than 2% by weight in all samples (Charlaftis et al., 2021), increasing to ∼14% by weight chlorite after the experimental treatment (Figure 2C).

Microporous material is diminished by ∼ 50% in all the analyzed volumes. This is due to the gradual coarsening and morphological evolution of the kaolinite crystals from an initial booklet morphology with pseudohexagonal habit to a more euhedral blocky habit at 200°C and a completely cemented phase with no visible microporosity at 250°C (Figure 5D).

Based on XCT and SEM observations, albite dissolution generating secondary pore space is apparent from 150°C onwards. Minor albitization of K-feldspar detrital grains and the formation of albite overgrowths were also observed. Albitization is generally incomplete and occurs preferentially along cleavage surfaces or microcracks presenting narrow irregular networks. Albite overgrowths form euhedral outlines on parts of the detrital feldspar grains that are not occupied by chlorite or siderite coatings. Finally, authigenic quartz overgrowth precipitation was not observed in the analyzed samples.

The visualizations in Figure 6A demonstrate a clear difference between the initial and final pore network model of the hydrothermally treated sandstone sample. The post-experiment network is less dense, owing to the porosity loss, with the number of the extracted pore bodies and throats decreasing from 1,188 to 750 and from 4,758 to 1822, respectively. Pore connectivity has been drastically decreased with the mean coordination number being dropped from 8 (0°C) to 4.9 (250°C). Equally, pore and throat radius distributions reveal a shift towards smaller pore and throat sizes with the mean pore radius decreasing from ∼75 to ∼73 μm and the mean throat radius from ∼17 to ∼15 μm (Figures 6B–D).

Again, like the single-stage experiments, the APES results show a decrease in permeability with increasing temperature (Figure 7A) for all samples, with the main phase of permeability reduction occurring between 150 and 200°C. These samples show greater variability in both the initial permeability and the reduction, driven by microstructural differences between the sample locations within S6B. Above 200°C the APES shows only small changes (both increase and decrease) in permeability; consistent with the negligible changes in the porosity and microporous volumes at this temperature.

Figures 7B,C display the APES models of sample S6C2 at 0 and 250°C. The results indicate a reduction from an initial value of 29 to 17 mD. S6C1 which is located in a highly porous region within the larger cuboid volume (S6B), has an initial permeability value of 160 mD and presents the largest permeability change. Although the preferential fluid flow pathways are preserved as in sample S6C2, permeability at the end of the experiment is reduced to 65 mD (Figures 7D,E).

The analysis of the single-stage samples (S1-S5, 3.8 μm/voxel) shows a grain coating volumetric increase of ∼9% at 250°C whereas the multi-step sample (S6B, 1.7 μm/voxel) presents an increase of ∼4%. This difference is partly attributed to the applied multi-scale analysis. To test this, we took a randomly positioned high-resolution subvolume from the pre-treatment S6 data (∼14.1% coating) and down-sampled the data to match the voxel resolution of the S1-S5 datasets. Reapplying the segmentation workflow (see Supplementary Material) on the down-sampled data yields a grain coating volume of ∼18.5%, caused purely by the difference in resolution between the two images. This is further validated by the equal drop of porosity (from ∼13.3 to ∼8.5%) after down-sampling. Considering this difference when calculating the grain coating volume increase from 0 to 250°C we infer a calculated value of ∼5.2% for the down-sampled data, 31% higher than the ∼4% calculated for the high-resolution data. However, this is still lower than the ∼9% volumetric increase recorded for the single-stage sample implying an additional influencing parameter derived from the experimental treatment each sample has experienced (discussed in Diagenetic Evolution and Subsequent Influence on Reservoir Quality).

The accuracy of any image-based analysis of pore geometries or subsequent flow simulation is influenced by the pore segmentation (Iassonov et al., 2009; Peng et al., 2014). Higher resolution images will always allow more precision and better capture of fine porosity but will also mean a smaller area or volume being analyzed. The down-sampled data of the 1.7 μm/voxel dataset yielded a porosity comparable to that of the 3.8 μm/voxel dataset (∼8.5%) suggesting that the 3.8 μm/voxel data are unable to resolve the 50% of the porosity that is in the smallest pores.

When compared with helium pycnometry measurements low-resolution tomography-based porosities are significantly lower. The initial XCT-based image porosity (mean 6.6%, 3.8 μm/voxel) of the single-stage samples is lower than the helium pycnometry value (∼28%) (Charlaftis et al., 2021) for the same sample volume (the cylindrical volume). In accordance with previous studies, significant differences between the two techniques are to be expected, given the spatial scales of the measurements (Callow et al., 2020). Here, the samples being analyzed are the same, and the difference is caused by the spatial resolution of the XCT data. The missing porosity is mainly credited to the unresolved microporous regions (e.g., below the maximum voxel resolution) that contribute to the total porosity (Saxena et al., 2019), and to a lesser extent to partial volume effects occurring between high-contrast edges (e.g., grain coatings and void space).

An appraisal of those parameters was performed on sample S6B (1.7 μm/voxel) which presents an initial XCT-derived porosity and microporous volume of ∼14.5 and ∼10%, respectively (Figure 5A). Pore space within the clay mineral fraction is dominated by disconnected intragranular pores (Milliken and Curtis, 2016). Diagenetic kaolinite, depending on its texture, has microporosity up to 61% (Hurst and Nadeau, 1995), with blocky compact kaolinite minerals presenting an average microporosity of 20% (Alansari et al., 2019). By adding the estimated microporous volume, the XCT-derived total porosity increases to ∼20.5%, compared to a helium-derived value of ∼28%. Although sample S6B is not volumetrically equal to samples S1-S5, the higher resolution imaging increases the pore volume quantified from the XCT data.

The difference in the diagenetic history between the single- and multi-stage samples is reflected in the evolution of the grain coating material.

In both scenarios the grain coating volumetric increase is attributed to the development of chlorite coatings (Figure 5C) which can be formed by siderite replacement and neoformation on precursor-free substrate surfaces. The progressive temperature increase adopted in the multi-stage treatment causes a gradual dissolution of the siderite crystals. In contrast, the single-stage treatment instigates an abrupt decomposition associated with dissolution voids at the central parts of the siderite crystals (Charlaftis et al., 2021). The rapid siderite decomposition promotes higher concentration of Fe ions, thus increased chlorite precipitation rates given sufficient temperatures. Based on our observations, precipitation rates are temperature-dependent and not time-dependent. Irrespective of experimental duration, both treatments suggest a synchronous, temperature-wise, initiation of chlorite development. This validates the widely held view that temperature is the principal control on chlorite coating authigenesis (Hillier, 1994; Ajdukiewicz and Larese, 2012).

The hydrothermal experiments were all performed in a silica supersaturated system, yet the main mineralogical change causing porosity reduction is the evolution of the pore-filling kaolinite (Figure 5D), not the development of quartz cement overgrowths. Temperature increase coupled with silica availability and the observed dissolution of feldspars, assist the transformation of kaolinite to dickite. The initiation of this replacement occurs at ∼ 150°C, comparable to natural systems where it is reported to occur in the temperature range of 110–165°C (Ehrenberg et al., 1993; Beaufort et al., 1998).

The experimental results suggest that the authigenic chlorite coatings develop in preference to quartz cement at temperatures >150°C. The overall grain coating volumetric increase is crucial for the effectiveness of quartz cement inhibition through offering fewer nucleation sites for quartz to precipitate. It also provides a mechanism for preserving effective storage space and connected pathways for unrestricted fluid flow at the pore scale. Figure 8 summarizes the main diagenetic processes observed during the hydrothermal treatment.

The evolution of the pore networks indicates that the magnitude of diagenesis governs the pore and throat sizes and pore connectivity. The decrease in size of pore and throat radii is attributed to the progressive kaolinite cement evolution and, to a lesser extent, to albite overgrowth development. Kaolinite progressively chokes primary pores and many of the throats have been completely sealed resulting in fluid flow restriction. The permeability models indicate that, between 150 and 200°C, initial areas of reduced velocity magnitudes are converted to preferential fluid flow pathways whereas others cease to exist. This suggests that microporosity plays an important role not only on the ability of rocks to store fluids but also to permit fluid flow. The dissolution of unstable grains, such as feldspars, and the pore filling kaolinite cement drive increased flow to parts of the network that initially had low flow velocity magnitudes. At temperatures higher than 200°C permeability is stabilized. Dissolution pore space becomes interconnected with intergranular pores, thereby maintaining an open pore system with improved effectiveness.

Natural diagenetic conditions cannot be fully reproduced in the laboratory due to the short experimental time compared to a geological timeframe and the simplification and uncertainties associated with the chosen aqueous chemistry. However, the experimental approach presented here, reveals the changing pore-throat characteristics and identifies the critical threshold temperature above which the increase in grain coating volume prevents further porosity reduction. Targets for future experiments would be prolonged experimental times and an X-ray compatible hydrothermal setup to allow observation during precipitation. This would permit better constraint of the temporal changes in porosity and the associated changes in permeability. From the technical perspective, synchrotron-based imaging with monochromatic X-rays would also help to capture phase distribution (e.g., chlorite and siderite) that cannot be easily resolved by lab scanners. A future target would also be to collect more high-resolution SEM data and permeability measurements at the micrometer scale to enable more advanced numerical modeling and understanding the effective permeability of the pore filling phases.

Sandstone reservoirs are primary targets for hydrocarbon, carbon capture and storage, hydrogen storage, and geothermal exploration operations. This research provides valuable insights into the extent, in terms of volume, of grain coating development and subsequent porosity and permeability evolution with increasing temperature. The ability to better constrain the temperature ranges and the associated volumetric alterations under which grain coating authigenesis occurs will result in more reliable reservoir quality predictions, essential as we undergo a low carbon energy transition.