1 X-Ray CT analyses , models and numerical simulations – a comparison with common analytical methods of an experimental CO 2 study

An essential part of the collaborative research project H2STORE (“hydrogen to store”), which is founded by the 10 German government, was a comparison of various analytical methods to characterize reservoir sandstones from different stratigraphic units. In this context Permian, Triassic and Tertiary reservoir sandstones were analysed. Rock core materials, provided by RWE Gasspeicher GmbH (Dortmund), GFD Suez E&P Deutschland GmbH (Lingen), E.ON Gas Storage GmbH (Essen) and RAG Rohöl-Aufsuchungs Aktiengesellschaft (Wien), was processed by different laboratory techniques; thin sections were prepared, rock fragments were crushed, cubes of 1 cm edge length and plugs of 5 cm in length were sawn from 15 macroscopic homogenous cores. With this prepared sample material, polarized light microscopy and scanning electron microscopy coupled with image analyses, specific surface area measurements (BET), He-porosity and N2-permeability measurements and high resolution micro-computer-tomography (μ-CT), which were used for numerical simulations were conducted. All these methods were applied to most of the same sample material, before and after static CO2 experiments under reservoir conditions. A major concern in comparing the results of these methods is an appraisal of the reliability of the 20 given porosity, permeability and mineral specific reactive (inner) surface areas data. The CO2 experiments are modifying the petrophysical as well the mineralogical/geochemical rock properties. These changes are detectable by all applied analytical methods. Nevertheless, a major outcome of the high resolution μ-CT analyses and proceeded numerical data simulations results in quite similar data sets and data interpretations maintained by the different standard methods; even regarding only CT-single scan of the rock samples. Moreover, this technique is not only time saving, but also none destructive. This is an 25 important point, if only minor sample material is available and a detailed comparison before and after the experimental tests on micro meter, pore scale of specific rock features is envisaged.


Introduction
The globally rising carbon dioxide emissions, increasing climate extremes and the "Energiewende" in Germany are keywords which are highly relevant and intensively discussed by governmental authorities, scientists, industrial representatives and the public society.As part of these debates the BMBF founded project -Hydrogen to Store-(H2STORE) one main objective of research was to characterize mineral-, synthetic formation fluid-, microbiological-and petrophysical interactions in reservoir sandstones, induced by hydrogen and carbon dioxide autoclave batch experiments (Pudlo et al., 2013, 2015and Henkel, 2016 submitted).Such reactions were investigated by autoclave experiments at reservoir temperatures and pressures, using none corrosive autoclaves and fluids, similar to reservoir specific formation fluids.
To satisfy the demands of reducing greenhouse gas emissions given in the Kyoto protocol (1997) and most recently strengthen by the UN climate conference in Paris in 2015, intentions to capture, inject and store CO 2 into the geological underground (CCS) are even more relevant.Also the utilisation of CO 2 combined with a H 2 injection into the geological underground for "green methane" generation (Šmigiáň et al., 1990), enforced by microorganism activity (Panfilov, 2010) are options which asks for further research regarding potential CO 2 -biotic-mineral-formation fluid-reactions.
Additionally, the integrity of wells, including the behaviour of e.g.bore hole casings in a highly corrosive environment of H 2 /CO 2 -gas mixtures is an essential point in the potential usage of the geological underground for storage options.
In this study we used rock samples provided by the industrial partners RWE Gasspeicher GmbH (Dortmund), GFD Suez E&P Deutschland GmbH (Lingen), E.ON Gas Storage GmbH (Essen) and RAG Rohöl-Aufsuchungs-Aktiengesellschaft (Wien).The high resolution computed tomography (μ-CT) data associated with numerical fluid flow simulations using the GeoDict -Math2Market ® and standard petrophysical analyses, which were conducted at the Technical University Clausthal before and after the static autoclave experiments, were applied.Thus to verifying potential reactions induced by the CO 2 autoclave experiments and to compare these data with the results, achieved by the other conventional analytical methods (e.g.Pudlo, 2014).

Material
The used sandstone samples originated from areas and stratigraphic formations, which comprise the main reservoir units in Germany and Austria.The sample material originates from three different stratigraphic units, the Permian in Northeast Germany, the Triassic in Northwest Germany and the Tertiary of South Germany and Northwest Austria (Fig. 1).Therefore, the source areas, the detrital content, the depositional environment and the diagenetic evolution of these sediments are diverse, implying that the samples experienced different burial pressures, temperatures and interactions with their site specific formation fluids.The current reservoir conditions of the sample sites of interest are summarized in table 1, showing the marked differences, especially in comparing the Permian and Tertiary sample sites.The current burial depth of the Permian samples is about 3500 m with a reservoir temperature of 120-125°C.This is in contrast to the recent Tertiary reservoir conditions with a burial depth of 1600 m and reservoir temperatures of about 40°C (Tab.1).Also    Thus a wide range of different rock compositions are available in conducting this study.Also the grain sizes of the different sandstone samples are varying from silt to coarse sand (Fig. 5) fraction after Wentworth (1922).
For the autoclave experiments and the petrophysical, mineralogical and chemical analyses the core material was prepared as given in Henkel et al. (2014).For reservoir characterisation mainly macroscopic homogeneous parts of the rocks (Fig. 4) were sampled and used for the different analytical methods and the CO 2 experiments.

Methods
Analytical methods were applied to macroscopically homogeneous sandstones from the three stratigraphic formations on identical hand specimen to compare the outcomes achieved by the different analytical methods.Additionally, the Permian samples were used for the CO 2 experiments at sample specific reservoir conditions and were analysed before and after these experiments at the same samples and marked sample positions, respectively.Therefore, a pin point comparison of different mineral reactions due to the CO 2 experiments is possible.The 4 to 7 week lasting CO 2 autoclave batch experiments were conducted at the Clausthal University of Technology, in rock-formation brine systems, in high pressure, high temperature (HPHT) and corrosion resistant autoclaves using selected thin sections, rock fragments and plugs (Pudlo et al., 2015 andHenkel et al., 2015).Helium porosity and nitrogen permeability measurements before and after the tests were performed on sandstoneplugs at the TU Clausthal.For porosity analysis a set-up after Torsaeter and Abtahi (2003) was used and permeability determinations were conducted with a hassler cell in an up-and downstream regime and calculated after Darcys Law equations.
The fluid samples of the formation fluids used in the autoclave experiments (5-10 ml) were taken before, during and after the tests at the TU Clausthal and chemically analysed in regard of their major, minor and trace element contents at the Friedrich Schiller University Jena.For the analyses, with an error of about 1 %, an inductively coupled plasma spectroscope with optical emission (Varian 725 ES) and a mass spectroscope (Thermo Fischer scientific X Series II) were used (ICP-OES/MS).
Also the physico-chemical characterisation of the formation fluid was conducted in the laboratories of the Institute of Geoscience in Jena.With the software PHREEQC 3.3 (Parkhurst & Appelo, 2013) and the database for highly saline fluids (Pitzer et al., 1984) the saturation indices of the fluids for the pore filling minerals (carbonate and anhydrite) were calculated.
Polarized light microscopy and field emission scanning electron microscopy (FE-SEM) were used at the FSU Jena for the mineralogical investigations.Thereby a Zeiss Axioplan II petrological microscope, equipped with an ocular of 10x magnification and objectives of 2.5x, 5x, 10x, 20x, 40x magnifications were applied, which was combined with a mounted Math2Market ® were used (Wiegmann, 2007 andWiegmann et al., 2013).The Navier Stokes-Brinkmann flow solver in the module FlowDict with the density properties of water at 20 °C were used for the fluid flow simulation and permeability calculations.
The specific surface area (BET) measurements on crushed rock fragments were determined after Brunauer et al. (1938) at the Technical University Munich before and after the experimental runs at exactly the same sample material.

Results
The initial state of the statistical porosity, permeability and specific surface area data determined by the different methods from all the analysed sample material of this study is shown in figure 5.The measured specific surface area by the BET-method after Brunauer et al. (1938) on crushed rock fragments of the Permian samples reveal values of 0.64-2.34m²/g (mean: 1.18 m²/g) and for the Triassic samples values ranging from 0.45-1.72 m²/g (mean: 1.05 m²/g).The calculated surface areas from the CT-data sets for the Permian sandstones range from 0.0018-0.0033m² (mean: 0.0024 m²) and from 0.0006-0.0038m² (mean: 0.0023 m²) in the Triassic samples.For the determination of the specific surface area in these 16 Permian samples the BET method was used and revealed values of 0.51-2.75m²/g (mean: 1.31 m²/g).In contrast calculated surface areas by the µ-CT-data from the two sample cubes are lower with 0.17 cm² and 0.24 cm², respectively.

Conclusions
The given comparison of the porosity data deduced high resolution computer tomography scans and measurements by helium porosity confirm that both methods achieve most similar results.Nevertheless, grain sizes of sandstones (cp.fig. 5 A-D) and therefore the distinct required resolution rate of the µ-CT Scans will influence the quality of the calculated porosity data.Results from the fine and very fine grained sandstones samples from the Tertiary and Triassic imply that the data of helium porosity measurements are more precisely due to the helium intrusion even into micro pores (< 7.6 µm resolution of µ-CT) and thus the capability of helium migration even through very narrow pore throats.In contrast, because of the limited resolution of µ-CT scans (7.6-8.6 µm/voxel), such micro pores are not detectable and calculated porosities are less than revealed by helium porosity measurements.Most likely these differences in regarding micro pores led to distinct porosity data in fine to very fine grained rocks, whereas porosity determinations on coarse and medium grained sandstone samples are most similar.By comparing the results of both methods from identical samples after the CO 2 experiments, both methods confirm the same increase in porosity.The mean helium porosity measurements reveal an increase of total porosity by 16.0 % and the calculated µ-CT data one of 12.6 % due to the CO 2 experiments.Thus, an effect on porosity characteristic after the 4-7 weeks lasting CO 2 experiments under reservoir specific conditions are verified by both methods.
Also, the calculation of the surface areas by µ-CT data is limited due to the restricted resolution of the µ-CT scans.However, the software GeoDict calculations can offer reference values.However, the results of the specific surface area measurements after Brunauer et al. (1938) are more realistic, if compared with sandstone data from the literature (see e.g.Pudlo et al., 2015).Most likely, by including additional parameters like e.g.rock/mineral densities, the volume and the compositional as well as the morphological features of the pore space exposed mineral phases to the GeoDict approach, more reliable data are achieved.An important outcome in comparing data sets compiled before and after the CO 2 experiments is that both methods confirm an increase in surface areas; by µ-CT investigations and calculations this enhancement in the surface area is about 14.3 %, which is most similar to the 10.4 % maintained by the BET-method.The measured nitrogen permeability data and the results from the µ-CT calculations by Geodict are in the same order.However again, the quality of calculated permeabilities from the µ-CT data sets of the segmented pore space (pixel resolution: 7.6-8.6µm) depends on the grain size of the samples.Whereas for coarse and medium grained sandstones the results are most similar to the nitrogen permeability measurements, these resemblances are minimized with decreasing grain sizes.Nevertheless, by comparing the permeability data after the CO 2 experiments with the results before the experiments, both methods confirm a marked increase in permeability induced by the tests.This increase is about 1.75 times higher in the calculations than in the nitrogen permeability measurements.
This study confirms that high resolution computer tomography is a suitable method to verify the alteration of various mineral phases of sandstone samples induced during CO 2 batch experiments.The dissolution of carbonates and anhydrite, present as pore filling cements before the tests, is increasing rock porosity and the surface areas.an enhanced exposure of small grain rimming clay minerals to the pore space, which were covered by the pore filling cements before the experiments.These results from µ-CT analyses and numerical simulations are in accordance to helium porosity and nitrogen permeability measurements, the determination of specific surfaces by BET after Brunauer et al. (1938) and findings by scanning electron microscopy as well as hydro-, mineral-and geochemical analyses on similar sandstones (e.g.Pudlo et al., 2012, 2013, 2015and Henkel, 2016 submitted) confirming marked modifications in rock composition induced by CO 2 autoclave experiments, which are most relevant in evaluating the suitability and quality of potential underground reservoirs.
In the future, the discussed deviations of data reliability related to grain size with very good in coarse and medium grained and minimized accordance in fine to very fined grained sandstones can be precluded by applying higher scan resolutions and/or smaller sample sizes, following the approach of Nordahl and Ringrose (2008) in determining an appropriate

Figure 1 .
Figure 1.Location map of the study areas with depth information of the reservoirs depth and company symbols, which provided 5 the formation fluid salinities are varying from 35.22 % salinity in the Permian reservoirs to only 1.81 % salinity in the Tertiary formation fluids.Solid Earth Discuss., doi:10.5194/se-2016-36,2016 Manuscript under review for journal Solid Earth Published: 8 March 2016 c Author(s) 2016.CC-BY 3.0 License.

Figure 2 .
Figure 2. The petrographic sandstone classification after McBride (1963) for samples of this study with the stratigraphic classification is shown.This petrographic classification was conducted by identifying and analysing 300 minerals in thin sections of 5

Figure 3 .
Figure 3. Thin section images with a 50x magnification given different diagenetic features (pore filling cements) of the sandstone samples.In A a transmitted light image of a sample with free pore space is shown (blue colour = open pore).In B an image of a thin section with crossed nicols is shown, highlighting pore filling anhydrite cement (rainbow colours).In C a transmitted light image of a thin section with pore filling carbonate cement is shown (arrows).

Figure 4 .
Figure 4. Different examples of lithotypes of the analysed Permian sandstones are outlined.In A a brownish, plane-horizontal stratified medium grained sandstone type is shown.In B a red, massive, medium coarse grained sandstone type is visualized and in C a greyish, obliquely bedded fine-medium grained sandstone type is presented.
Solid Earth Discuss., doi:10.5194/se-2016-36,2016 Manuscript under review for journal Solid Earth Published: 8 March 2016 c Author(s) 2016.CC-BY 3.0 License.Hitachi HV-C20 digital camera for the light microscope.For the field emission scanning electron microscopy (FE-SEM) and the energy dispersive X-ray spectroscopy (EDX) a SMT ULTRA plus field emission scanning electron microscope of Carl Zeiss enterprises, coupled with an EDX-detector and the analytical software of Bruker AXS Microanalysis GmbH was used for mineral documentation and identification.For the high resolution computer tomography, a Procon CT Alpha 160 device at the Johannes Gutenberg University Mainz was used.The detector resolution was about 2048x2048 pixels.The used parameters for the measurements are 100 kV of voltage and a current of 80 µA.Scans were performed at 0.45° steps over a total rotation range of 360°.The achieved resolution for the different sample material is between 7.6 µm and 8.6 µm per voxel.The reconstruction of the scans was performed with the software Octopus 7.0 ® .The µ-CT image adaption with anisotropic diffusion filtering and pore space segmentation was performed with the software Avizo ® fire 7.1.Intense distinct beam hardening was corrected with the software MATLAB ® and the approaches afterJovanovic et al. (2013).For the calculations of porosity, surface area, fluid flow and deduced permeability the corresponding modules of the software package GeoDict from the company

Figure 5 .
Figure 5.In A the porosity results of the same eleven µ-CT and He-porosity analysed samples are shown for three different 5

Figure 6 .
Figure 6.In A the µ-CT porosity and µ-CT permeability data for two Permian samples before and after the CO 2 experiment are shown.In B the helium porosity and nitrogen permeability measurement results of 16 Permian samples before and after CO 2 experiments are presented (B, data from Pudlo et al., 2012).In C the mean measured BET surface areas and the mean calculated µ-CT surface areas before and after the CO 2 experiment are shown.

Figure 7 .B
Figure 7.In A and B the details of the same Permian sandstone sample before (A) and after (B) the seven weeks lasting CO 2 experiments under reservoir conditions is shown.Note the reduction of the blocky cements (light grey) and the fluid flow field with increased fluid migration pathways after the experiments (blue-red colours).

Figure 8 .
Figure 8.In A and B the identical Permian sandstone sample is presented before (A) and after (B) CO 2 experiments under reservoir conditions.The images are enhanced section from fig. 7 and verifying the dissolution of the blocky pore filling cements 5

Figure 9 .
Figure 9. Secondary electron FE-SEM images of the same section of a Permian sandstone sample before (A + C) and after (B + D) CO 2 experiments under reservoir specific conditions for seven weeks.The dissolution of pore filling calcite cement (star in B) and the exposure of grain rimming clay minerals (arrow in B) leading to the increasing porosity outcomes after the CO 2 experiments and the increase of the specific surface area (modiefied after Henkel et al., 2014).In C and D the dissolution of pore filling anhydrite cement (star in D) is shown at exactly the same thin section position before (C) and after (D) the CO 2 experiment also leading to increasing porosity values after the experiments.

Table 1 . Overview of the recent reservoir conditions of sampled locations.
In comparing the detrital rock composition due to their quartz, feldspar and rock fragment content the different locations a wide range of sandstone types is present (Fig.2-after McBride, 1963).
The measured helium porosity and nitrogen permeability values for the Permian plug samples ranges from 5.64-14.73% in porosity (mean:10.96%)and 2.61-161.32mD(mean: 96.27 mD) in permeability.For the measured Triassic samples, the helium porosity data are between 11.02-29.95%(mean: 21.01 %) and for nitrogen permeability data between 251.93 and 1609.43 mD (mean: 521.26 mD).The measured Tertiary plug samples represent porosity values of 19.14-26.62% (mean: 25.67 %) and permeability values from 25.80-388.00mD (mean: 188.82 mD).The corresponding parameters calculated from the µ-CT data with the Software GeoDict and the implemented Navier Stokes-Brinkmann solver for the permeability results in the Permian samples in connectivity in the samples (no detectably connected pore networks) modelling of permeability from CT-data was not possible.Solid Earth Discuss., doi:10.5194/se-2016-36,2016 Manuscript under review for journal Solid Earth Published: 8 March 2016 c Author(s) 2016.CC-BY 3.0 License.

Table 3 . PHREEQC modelling, showing saturation results of the interaction of the synthetic formation fluid used before, during and after the CO 2 experiments exposed to the Permian sandstones (n.a. = not analysed, SI = saturation index, pE = negative logarithm of electron concentration -directly proportional to the redox potential).
The gain in surface areas is caused by Solid Earth Discuss., doi:10.5194/se-2016-36,2016 Manuscript under review for journal Solid Earth Published: 8 March 2016 c Author(s) 2016.CC-BY 3.0 License.