Mineralogical and geochemical analysis of Fe-phases in drill-cores from the Triassic Stuttgart Formation at Ketzin CO2 storage site before CO2 arrival

Reactive iron (Fe) oxides and sheet silicate-bound Fe in reservoir rocks may affect the subsurface storage of CO2 through several processes by changing the capacity to buffer the acidification by CO2 and the permeability of the reservoir rock: (1) the reduction of three-valent Fe in anoxic environments can lead to an increase in pH, (2) under sulphidic conditions, Fe may drive sulphur cycling and lead to the formation of pyrite, and (3) the leaching of Fe from sheet silicates may affect silicate diagenesis. In order to evaluate the importance of Fe-reduction on the CO2 reservoir, we analysed the Fe geochemistry in drill-cores from the Triassic Stuttgart Formation (Schilfsandstein) recovered from the monitoring well at the CO2 test injection site near Ketzin, Germany. The reservoir rock is a porous, poorly to moderately cohesive fluvial sandstone containing up to 2–4 wt% reactive Fe. Based on a sequential extraction, most Fe falls into the dithionite-extractable Fe-fraction and Fe bound to sheet silicates, whereby some Fe in the dithionite-extractable Fe-fraction may have been leached from illite and smectite. Illite and smectite were detected in core samples by X-ray diffraction and confirmed as the main Fe-containing mineral phases by X-ray absorption spectroscopy. Chlorite is also present, but likely does not contribute much to the high amount of Fe in the silicate-bound fraction. The organic carbon content of the reservoir rock is extremely low (<0.3 wt%), thus likely limiting microbial Fe-reduction or sulphate reduction despite relatively high concentrations of reactive Fe-mineral phases in the reservoir rock and sulphate in the reservoir fluid. Both processes could, however, be fuelled by organic matter that is mobilized by the flow of supercritical CO2 or introduced with the drilling fluid. Over long time periods, a potential way of liberating additional reactive Fe could occur through weathering of silicates due to acidification by CO2.


Introduction
As a mitigation strategy to reduce the emission of the greenhouse gas carbon dioxide (CO 2 ) produced during the combustion of fossil fuel, storage of CO 2 below the earth surface is considered as a potentially important technology (IPCC 2005;IEA 2013). While the feasibility and longterm effectiveness of this approach are still debated, several large-scale experiments have been conducted, and demonstration projects are active to better understand the behaviour of the rock reservoir during injection and longterm storage of CO 2 (IPCC 2005). Besides the physical properties, the geochemical changes in rocks and pore waters of the storage formation and the microbiology in the rock aquifer also need to be better understood. A largescale test injection of CO 2 has been conducted near the town of Ketzin, Germany, between 2008 and 2013 (e.g. Würdemann et al. 2010;Martens et al. 2012Martens et al. , 2013. The reservoir rock is a porous sandstone belonging to the Stuttgart Formation (Fm.; former Schilfsandstein) and occurs at a depth of approximately 630-650 m. The reservoir is exemplary for other storage sites, where typically porous siliciclastic rocks overlain by an impermeable cap rock are used for gas storage, such as at the Sleipner gas storage site in the North Sea, where the CO 2 is injected into sands of the Miocene-Pliocene Utsira Fm. (e.g. Lackner 2003;Zweigel et al. 2004) and others (IPCC 2005) or considered for storage, such as the Early Jurassic Navajo Sandstone (Colorado Plateau, western USA; Chan et al. 2000Chan et al. , 2005Parry et al. 2007).
In siliciclastic sediments and rocks, Fe is commonly the most abundant redox-active solid-phase element and plays an important role in biogeochemical cycling due to its function as electron donor or acceptor for microbial processes in the deep biosphere (e.g. Froelich et al. 1979;Lovley and Phillips 1986;Canfield et al. 1993). In the subsurface, abiotic reactions and microbial Fe-metabolism may lead to the dissolution or formation of various Fephases, such as Fe-oxides, hydroxides, sulphides or carbonates. In CO 2 storage reservoirs, reactions involving Fe may affect the geochemistry in several ways: (1) the reduction of Fe(III), both abiotic and microbially mediated, leads to an increase in the pH, buffering the acidification imposed by the dissociation of injected CO 2 (Coleman and Raiswell 1995;Curtis et al. 1986;Fisher et al. 1998). This process would then further support the sequestration of CO 2 in the form of dissolved bicarbonate or even induce the precipitation of solid-phase carbonate, hence permanently trapping the CO 2 . While carbonate precipitation is generally favourable for CO 2 trapping, it causes problems in proximity to the injection well as it may reduce porosity and thereby injectivity of CO 2 . (2) In combination with dissimilatory sulphate reduction, Fereduction may drive sulphur cycling via the formation of insoluble Fe-sulphide precipitates. Also, Fe-sulphides may form as a result of corrosion of the drill string or injection pipelines by oxidation of elemental Fe to Fe(II) coupled to sulphate reduction (Enning et al. 2012). Such dissolution-precipitation reactions would alter the porosity and permeability of the reservoir rock. (3) The microbially induced changes of valence states in silicate-bound Fe may have an impact on silicate weathering (Santelli et al. 2001), which could alter the pH and alkalinity over a long period of time.
In order to predict the geochemical changes associated with the oxidation and reduction of Fe after injection of CO 2 into the reservoir rocks, this study provides a semiquantitative assessment of the Fe-mineral phases occurring in the reservoir rock at the Ketzin injection site, in the monitoring well, before the arrival of CO 2 . Total Fe-content was analysed by X-ray fluorescence, and differently reactive Fe-mineral fractions were quantified by X-ray diffraction, as well as by using a sequential extraction procedure (Poulton and Canfield 2005). Reduced sulphidebound Fe-phases were extracted as acid-volatile sulphide and chromium-reducible Fe-sulphur fractions (AVS and CRS, respectively) and compared with the total organic carbon content available as a substrate for microbial Fe and SO 4 2reduction. Furthermore, the predominant structure of solid Fe-phases was analysed by synchrotron-based X-ray spectroscopy. Results provide better constraints on the microbial and abiotic Fe-oxidation/reduction and their potential effect on subsurface CO 2 storage reservoirs in the Stuttgart Fm. and other porous sandstones that could be suitable for CO 2 injection.

Geological setting
The Ketzin CO 2 storage site is located 25 km west of Berlin (Germany; Würdemann et al. 2010). The reservoir horizon is up to 20 m thick and occurs at approximately 630-650 m depth on the southern limb of an east-west striking anticline ( Fig. 1a; Norden et al. 2010). The reservoir rock is a poorly to moderately cohesive reddish sandstone deposited in a fluvial environment during a humid period in the otherwise arid Germanic Basin during the Late Triassic (the Carnian Pluvial Episode; Kozur and Bachmann 2010, and references therein). Under the arid conditions, large amounts of evaporite were deposited, which are preserved in underlying units and partially within the Stuttgart Fm. as gypsum and anhydrite cements which precipitated from supersaturated hypersaline and sulphate-rich brine during early diagenetic processes. Arid conditions as well as diagenetic mobilization and re-oxidation led to the coating of grains with Fe-oxide/hydroxides . Previous studies reported an overall high Fe-content in the sandstones (6-7 wt% Fe 2 O 3 tot) of the Stuttgart Fm., partly derived from volcanic rock fragments .
The porous sandstone of the Stuttgart Fm. (Fig. 1b) overlies impermeable mudstone of the Grabfeld Fm. (Förster et al. 2006;Norden and Frykman 2013) and is sealed off by 200 m of mudstone of the Upper Triassic Weser Fm., Arnstadt Fm. and Exter Fm. A shallower reservoir is present at 250-400 m, which during previous years was used for natural gas storage (Förster et al. 2006;Würdemann et al. 2010). This shallower reservoir is sealed by argillaceous sediments of Tertiary age. One injection well for CO 2 -injection (Ktzi 201) and two observation   (Prevedel et al. 2008;Schilling et al. 2009;Förster et al. 2010;Würdemann et al. 2010). In 2011, an additional (fourth) shallow well (P300) was drilled ca. 25 m north-west from the observation well Ktzi 202, to monitor hydraulic and geochemical impacts of CO 2 on the groundwater of the shallower aquifer overlying the reservoir rock of Stuttgart Fm. and the caprock. It reached about 450 m deep, into the Upper Triassic ( Fig. 1c; Pellizzari et al. 2017;Martens et al. 2014). The reservoir rock has a porosity of 13-26% and a permeability of 40-110 mD (Wiese et al. 2010). A temperature of 35°C was measured at the injection depths at 650 m. The chemical composition of the reservoir fluid is dominated by the presence of sodium (ca. 90 g/l), calcium (2 g/l) and chloride (ca. 135 g/ l). The sulphate (SO 4 2-) concentration was about 4 g/l and the Fe-concentration (Fe tot ) 5.5-7.4 mg/l. The total dissolved solid content (TDS) was 235 g/l and the pH was 6.5. For more details concerning the chemical characteristics of the reservoir fluids, see Würdemann et al. (2010).
The CO 2 injection started on 30 June 2008 and ended on 29 August 2013 with 67,271 t of supercritical CO 2 injected into the reservoir. The gas consisted of CO 2 (99.7-99.9% purity) with traces of N 2 , He and CH 4 (Martens et al. 2012). According to Würdemann et al. (2010), the migration of CO 2 was confirmed when the arrival of CO 2 at the first observation well (Ktzi 200) was detected after three weeks of injection of about 500 t of gas, and at the second observation well (Ktzi 202) nine months after the beginning of injection when ca. 11,000 t was injected. More details concerning the site operations can be found in Würdemann et al. (2010) and citations therein. Ivandic et al. (2015) monitored the CO 2 plume evolution.

Sample preparation
During coring of the injection (Ktzi 201) and two monitoring wells (Ktzi 200 and 202), a water-based CaCO 3 / bentonite/organic polymer drill mud, containing carboxymethylcellulose (CMC; Wandrey et al. 2010), was used to lubricate the drill bit, transport cuttings to the surface and stabilize and maintain the bottom-hole pressure (Grace 2007). CMC was used because it is a biodegradable organic polymer and does not pollute the subsurface environment. For the third deep observation well (Ktzi 203), a bentonite drill mud containing cellulose-based polymers [CMC and polyanionic cellulose (PAC)] and a natural polysaccharide-based polymer (Biolam) was used together with K 2 CO 3 (Pellizzari et al. 2013). For well P300 (shallow hydraulic and geochemical monitoring well) reaching the aquifer above the CO 2 storage formation (Exter Fm.), a K 2 CO 3 -based drill mud was used (Pellizzari et al. 2013).
For this study, aliquots from six core sections from the observation well Ktzi 202, sampled between 627 and 638 m depth before the arrival of the CO 2 , were investigated for their Fe-mineralogy (Table 1). After coring, the reservoir rock material was roughly cleaned using sterile synthetic formation fluid to remove the drill mud. Subsequently, rock core samples were wrapped into autoclaved aluminium foil and stored at 4°C until processing. Seven samples were immediately processed, whereby the outer 2 cm of rock material was removed using an autoclaved chisel to prevent penetration of drill mud into the rock core . Subsequently, the samples were shock-frozen in liquid nitrogen and stored at -20°C.
For SEM analyses, sub-samples were freeze-dried and ground to \10 lm. To specifically target the Fe-composition of the sand grain coatings, some of the poorly cohesive sandstone samples were slightly crushed to disintegrate the single sand grains, but not milled to a powder. Thin sections of two selected samples were analysed under a petrographic microscope.
The reservoir fluid retrieved during the hydraulic tests and downhole sampling was analysed, and physico-chemical parameters were determined. For more details, see Würdemann et al. (2010). The hydraulic pumping tests were carried out as open-hole tests with production rates held at the maximal achievable rate. The fluids were collected directly from the well head, filled into sterilized glass bottles, cooled and transferred to the laboratory for chemical and molecular biological analyses.

Scanning electron microscopy with energydispersive spectrometry (SEM-EDS)
Air-dried and disintegrated sandstone fragments were mounted on SEM stubs using conducting tape, coated with carbon and examined with an Ultra 55 Plus (Carl Zeiss SMT) scanning electron microscopy (SEM) operating at an accelerating voltage of 20 kV, using the secondary electron (SE) signal. Energy-dispersive X-ray (EDX) spectroscopy was used for quantitative elemental analyses. Identification of elements in spot analyses and their distribution using the option of automatic or manual search of elements were performed using the analytical software Noran Vantage NSS. Element abundances were determined from the EDX spectra by integrating peak areas and normalizing the results to 100%.

X-ray diffraction
The mineralogical content of sediment was analysed by a Philips XPERT pro X-ray diffractometer at the University of Bremen. CuKa radiation was used and the samples were scanned from 3°to 85°(2h). Relative abundances of different minerals were estimated from integrated peak areas.
In addition, the clay fraction of four selected samples with the highest clay mineral content was separated according to the procedure of Moore and Reynolds (1997). For these analyses, sandstone samples were disaggregated using a hydraulic press. Siltstones were placed in a plastic bag, gently squeezed by hand and dispersed with distilled water in an Atterberg cylinder. The 75-100 mg of tetrasodium pyrophosphate (Na 4 P 2 O 7 9 10H 2 O) was added to 500 ml of suspension to prevent coagulation. The fraction \2 lm was prepared as a suspension. A 1-1.5 ml of suspension was pipetted onto a porous ceramic tile made of corundum. The water was drained through the tile by means of a suction pump, which allowed the clay particles to settle with an orientation parallel to the surface. X-ray diffraction patterns of the separated clay fraction were acquired by a Bruker D8 (LynxEye) diffractometer. CuKa radiation was used and the samples were scanned from 3°t o 70°(2h). Each sample was prepared as oriented air-dried sample, as glycolized under ethylene glycol atmosphere for 12 h at 50°C and as tempered at 550°C for 1 h. Clay minerals were identified with the Powder Diffraction File (PDF Ò ) Database and the Crystallography Open Database (COD; Grazulis et al. 2009).

Total, organic and inorganic carbon
Total carbon (TC) and total sulphur (TS) contents were determined with a Carlo Erba NA-1500 CNS analyzer using in-house standard (DAN1). Total inorganic carbon (TIC) content was measured using a CM 5012 CO 2 Coulometer (UIC) after acidification with phosphoric acid (3 M). Precisions (2r) were 0.08 wt% for TC, 0.05 wt% for TIC and 0.04 wt% for TS. Total organic carbon (TOC) was calculated as the difference between TC and TIC.

X-ray fluorescence
For the elemental analysis, approximately 4 g of sediment was dried, finely ground, poured into sample cups and firmly pressed to remove air from the interstices. Samples were analysed using the compact benchtop energy-dispersive polarization X-ray fluorescence (EDPXRF) analysis system Spectro Xepos. Standard deviation of repeated measurements was B1%, and the detection limit corresponds to a signal three times the standard deviation (Wien et al. 2005).
Total dissolved Fe-concentrations of the extracts were measured with a Thermo iCE 3000 Series atomic absorption spectrometer (AAS) after ten-or hundred-fold dilution. The precision of the measurements was better than ±2% (Standard deviation); the reproducibility of the extraction method for triplicate measurements was 10%.

Acid-volatile sulphide and chromium-reducible sulphur extraction
On the same set of samples (frozen samples only), a sulphide extraction was performed following the standard methods of Canfield et al. (1986) and Fossing and Jørgensen (1989). The samples were covered with 50% ethanol, 16 ml of 6 M HCl was added, and samples were distilled under nitrogen atmosphere for 1 h. Hydrogen sulphide evolved from AVS was precipitated in 5% Znacetate traps as ZnS. Following AVS extraction, Zn-acetate traps were replaced and 16 ml of reduced 1 M chromium chloride (CrCl 2 ) solution was added to the reaction vessel. Samples were heated and distilled for 1.5 h. Hydrogen sulphide liberated from chromium-reducible sulphur (CRS; from pyrite and S 0 ) was precipitated as ZnS. Concentrations of ZnS suspended in both traps were analysed spectrophotometrically at 670 nm by the diamine complexation method using N,N-dimethyl-1,4-phenylenediamine-dihydrochloride (Cline 1969). Detection limit of the spectrophotometric analyses was 1 lM.
X-ray absorption near-edge structure (XANES) spectroscopy XANES spectra were collected at the A1 beamline of the Hamburger Synchrotronstrahlungslabor (HASYLAB, Hamburg, Germany). Acquisition parameters were described in Meister et al. (2014). XANES spectra were collected at the Fe K-edge from 6960 to 8000 eV with 5 eV steps up to 7082 and 0.25 eV between 7082 and 7152 eV. A reference foil of metallic Fe(0) was used for internal energy calibration of the monochromator (the first inflection point of the Fe K-edge was set at 7112.1 eV).
XANES spectra were processed and analysed using the Horae Athena free software (Newville 2001;Ravel and Newville 2005). Experimental spectra were normalized and fitted to a linear combination of standard spectra of Feminerals using a least-square minimization procedure. The pre-edge centroid was calculated using the fitting procedure of Wilke et al. (2001) and using the free program Fityk (Wojdyr 2010) to determine the redox state of Fe in the samples.

Sediment description
Samples taken from Ktzi 202 (sampled between 627 and 638 m depth) consist of brittle and poorly cohesive sandstone. Only two samples (sample B2-1 and B4-2) are strongly lithified. The colour shows different reddish and beige domains. Some of the samples are very dark and easily disintegrate to sand. A millimetre-scale lamination is common. Thin sections show a well-sorted fine-grained sandstone (Fig. 2a, b). The structure is densely packed, grains are poorly rounded, and angular clasts show preferential orientation in the direction of the lamination (Fig. 2b). Mineral content is dominated by quartz with plagioclase, lithic fragments, opaque and sporadic single 50-lm-scale fibres (inset in Fig. 2c), possibly zeolites or sheet silicates ( Fig. 2a-d). The matrix is microcrystalline, and its colour varies between light beige and dull with the lamination. Opaque domains are either rich in organic matter or opaque minerals (most likely Fe-oxides) (Fig. 2d).
In the SEM images ( Fig. 3a), quartz, potassium feldspar, plagioclase, clay minerals and Fe-oxides were identified from semi-quantitative element abundances from EDX analyses. Also authigenic anhydrite cement, barite and single celestine crystals, all with a characteristic cleavage, were detected and confirmed by EDX. The main mineral phases, such as quartz and feldspar, are partly idiomorphic, usually with visible signs of dissolution and/or formation of authigenic cements on partly dissolved grains. The surfaces of quartz (Fig. 3b) and feldspar are coated by Fe-oxides, but also by clay minerals as shown in Fig. 3c, d (cf. also Förster et al. 2010). Clay minerals also grow in pits formed during dissolution, alteration and/or secondary precipitation processes (Fig. 3c).

X-ray diffraction of the bulk sample
Relative abundances of minerals were calculated from the ratios of the major peak areas normalized to 100% (Table 1a). The sandstone predominantly consists of quartz and plagioclase and occasionally contains analcime, an igneous zeolite. Several samples contain significant amounts of anhydrite. In particular, the strongly cemented sample B4-2 almost entirely consists of anhydrite. In some of the non-frozen samples, small amounts of gypsum were detected based on the 020 (hkl) peak, while the 021 peak of gypsum interferes with the 100 peak of quartz. The 200 peak is also always present in these cases. However, gypsum in the non-frozen samples may be due to hydration of precursor anhydrite during sample storage. No carbonate minerals were detected. Several sheet silicates are present showing peaks at small 2h angles. Illite (or muscovite) and chlorite indicate the best match with the peak distribution. A shoulder on the 001 peak (towards lower 2h) of chlorite at 6.2°2h (d = 14.3 Å ) may indicate the presence of smectite. A small peak at 24.16°2h (d = 3.68 Å ) matches the 012 peak of haematite, as reported by Förster et al. (2010).

X-ray diffraction of the clay fraction
Clay mineral separation and analysis revealed chloritegroup and mixed-layer clay minerals composed of illite and smectite in each sample. The samples exhibit similar diffraction patterns (Fig. 4). Quartz and feldspar are present due to disaggregation of detrital particles during the separation process. Other phases, which are not present in every sample, were identified as anhydrite, analcime and haematite. The most common phase shows the major reflection at *10 Å . It is a mixture of illite with interlayering of minor amounts of expandable clay minerals. Best fitting patterns in the diffractograms revealed an illite-smectite mixed phase defined by the formula (K 0.66 Ca 0.33 Na 0.03 )(Al 1.78 Mg 0.22 Fe 0.01 )[(Si 3.43 Al 0.57 O 10 ) (OH) 2 ] (Gournis et al. 2008). Additionally, a montmorillonite (also smectite) is present. The interlayering type of smectite within the illite structure is mainly a sodium-and calcium-bearing component [montmorillonite (Na,Ca) 0.33 (Al,Mg) 2 (Si 4 O 10 )(OH) 2 * nH 2 O], but other smectite interlayers cannot be excluded. The 10 Å peak interferes with the major reflections of biotite; however, only in sample B4-3, a higher content of biotite was detected. Illite crystallinity varies from poor to well crystallized based on the Kübler index as full width at half maximum (FWHM; Table 1b).
The second most common phase is a chlorite-type mineral. The identified pattern fitting best represents the chemical composition of clinochlore (Mg 2.96 Fe 1.698-Al 1.275 )(Si 2.624 Al 1.376 O)(OH 8 ). Haematite is present in minority compared to clay minerals. It was identified by its major peak at *33.3°2h. Peak intensity of the haematite 110 peak (35.7°2h) increased after heating.

Total, organic and inorganic carbon
Total inorganic carbon content is near 0.1 wt% in several samples. The same samples show TOC around 0.2 wt% ( Table 2). All other samples show TIC and TOC contents that are clearly below the detection limit. S was measured by CNS analysis in three samples, which is due to the presence of anhydrite or gypsum. In the anhydrite cemented sample B4-2, up to 6 wt% S was measured (Table 2).

X-ray fluorescence
Total Fe-content analysed by XRF is in the range of 2-4 wt% ( Table 3). Only sample B4-2 cemented by anhydrite contains less Fe. Samples containing anhydrite show high concentrations of calcium. Fe-to-Al ratios (wt/wt) vary between 0.4 and 0.7.

Sequential iron extraction
The total concentration of reactive (extractable) Fe (sum of the five extraction steps without Fe-sulphides) is close to the total Fe-content for all samples and ranges from 2 to 100 mg/g ( Table 4). Most of the extractable Fe is in the dithionite fraction (fraction III) and, in some of the samples, also in the boiling HCl fraction (fraction V). Concentration of fraction III Fe varies between 0.17 and 6.7 wt%, while sheet silicate-bound Fe in most samples is around 0.2 wt%. In samples B3-3a and B3-1, silicatebound Fe is strongly enriched (2.7 and 5.1 wt%). These samples also show the highest concentrations of total extractable Fe of 9.9 to 11.9 wt%. This enrichment in Fe is not observed in XRF measurements and is probably due to inhomogeneities in crushed but not ground samples.
Results from the standard minerals (Table 4) reveal that most minerals were extracted as predicted by Poulton and Canfield (2005). We highlight that besides the unreactive sheet silicates extracted by boiling HCl, some sheet silicates are more reactive, in particular the smectite clays. Otherwise, results are unclear, such as for illite. These minerals also largely leach with the dithionite fraction, such that this fraction cannot be exclusively ascribed to goethite and haematite.
The total dissolved Fe-content in aerobically stored reservoir fluid was 0.0136 g/l. AAS measurements also showed 0.736 g/l Ca and 1.095 g/l Mg in this water sample.

Acid-volatile sulphide and chromium-reducible sulphur
AVS concentrations in all samples are below detection (Table 4). The samples contain between 6 and 15 ppm (weight) CRS (Table 4; reported in ppm due to small values), which stoichiometrically represents 5-13 ppm (weight) of pyrite-Fe. The highest pyrite content of 32.5 ppm was measured in the strongly lithified sample B4-2.

XANES spectroscopy
Linear combination fitting analyses of XANES spectra with a number of standard minerals (akaganéite, beidellite, chamosite, chlorite, illite, nontronite, goethite, haematite, saponite and zinnwaldite) show that illite and smectite are the most abundant Fe-containing phases in the analysed rock samples (Fig. 5; Table 5). Results also show that chlorite and haematite are minor fractions in the sediment although chlorite is present in high abundance in X-ray diffractograms. Thus, the Fe-content in chlorite is very low. Redox states of Fe in the samples were calculated using the pre-edge of the XANES spectra and were shown to range between 2.6 and 2.8.

Fe-mineralogy of Stuttgart Formation
Before addressing possible Fe-related processes, we evaluate the different results presented above with respect to the dominating Fe-phase in the host rock. Both XRD and XANES spectroscopy clearly show that most of the Fe in the sandstones from Ketzin is bound to sheet silicate minerals. This outcome is consistent with the sequential extraction, taking into account that reactive sheet silicatebound Fe may also partially leach from the dithionite fraction (e.g. nontronite). Illite is the most abundant Fecontaining phase indicated by the abundance of illite-Fe from linear combination fits of the XANES spectra, suggesting that illite contributes more than half of the total Fe in the samples. Illite-Fe from XANES spectroscopy analysis correlates with the illite content determined by XRD (Fig. 6a). Illite-Fe does not positively correlate with the Fe/ Al ratio (Fig. 6b), which could suggest that Fe is bound to another phase not containing Al. However, the poor or even anti-correlation could also be explained by Fe 3? substituting for Al(III) in the illite lattice (e.g. Seabaugh et al. 2006). A redox state of 2.6-2.8 was determined by analysing pre-edge peaks in the XANES spectra. XRD analysis of the clay fraction allowed us to identify mixed-layered structures, in particular expandable layers of smectite within the illite structure and smectite and/or vermiculite within the chlorite structure. The smectite interlayers are confirmed by the XANES spectra, indicating up to 40% nontronite-bound Fe. Nontronitebound Fe also explains the high Fe-content in the dithionite fraction.
While the smectite layers in illite are clearly identified from the peak shift during saturation with ethylene glycol and subsequent heating to 550°C, the mixed layers in the chlorite structure may consist of vermiculite rather than smectite. A clear identification of the mixed layers in the chlorite structure is hampered by an atypical collapse of the 002, 003 and 004 hkl peaks during temperature treatment. Collapsing peaks have been described by Humphreys et al. (1989) for detrital and authigenic chlorite in late Triassic sandstones after heating the samples to 600°C, but the  cause is not clearly known. The diffraction pattern of chlorite also depends on the Fe-content and the position of the Fe atoms in the silicate layer or the hydroxide sheet (Moore and Reynolds 1997). Usually, the relatively stable peak positions and some varying peak intensities for the 001, 002, 003 and 004 reflections during temperature treatment can be used as an indicator for the estimation of the Fe-content in chlorite (Moore and Reynolds 1997). However, due to the atypical reaction of the chlorite in our samples at 550°C, these peaks are destroyed. Based on the untreated diffraction patterns, the identified clinochlore fits best to the diffraction patterns exhibiting a Fe-content of 15.5 wt% and Mg content of 11.85 wt%. XANES spectra only indicate subordinate levels of chlorite-bound Fe (max. 16.7%). In contrast to smectites, chlorite mainly leaches in the boiling HCl fraction (Raiswell et al. 1994, and this study) and could contribute to parts of the illite-bound Fe to the high content of Fe in this fraction. Chlorite is indicated by an increasing intensity and a slight peak shift upon heating from 6.226°2h to 6.405°2h, which corresponds to a d-spacing reduction from 14.185 to 13.787 Å . This reaction of the 001 reflection of chlorite is typical and known from several examples (e.g. Hillier and Velde 1992;Humphreys et al. 1989;Moore and Reynolds 1997). It is explained by the dehydroxylation of the hydroxide sheet and attendant decrease in the d-spacing.
Chlorite in this study may be of both detrital or authigenic origins. Authigenic chlorite is commonly formed by the alteration of mafic igneous minerals during burial diagenesis (e.g. biotite; Meunier 2005). As detrital chlorite may form during grinding of the sample from minerals in the larger grain fractions, some chlorite in our samples may also be an artefact. This could explain the absence of chlorite from XANES spectrometry, which was treated separately from the clay mineral preparation.
A small peak at 3.68 Å in the diffractograms suggests haematite in several of the samples. Also Förster et al. (2010) report the presence of minor amounts of haematite in the Stuttgart Fm. Fe-oxyhydroxides is often difficult to detect by XRD due to the extremely small crystal size and low refraction intensity. Also most concentrations in the sequential extraction are far below the detection limit of 5% for XRD. Fitting analysis of XANES spectra of the Ketzin samples confirms the low contribution of haematite to the total Fe-content. The large concentration of Fe in the dithionite fraction is not indicative of a high haematite content, since it co-elutes with smectite and possibly illite. Despite its low content, haematite may possibly be important as it forms coatings on sand grains (Fig. 3b), together with illite or smectite clays. Accordingly, haematite provides one of the most reactive Fe-phases in the host rock. Other minerals, such as feldspar, may only contain traces of Fe, while quartz anhydrite, gypsum and halite do not contribute any Fe to the reservoir rock. Although analcime itself does not contain Fe, zeolites are known for their ability to bind Fe (Pirngruber et al. 2004).  Origin of the iron and past Fe-cycling The content of Fe of up to 4 wt% in the fluvial sandstone of the Stuttgart Fm. is relatively high. This high Fe-content in Triassic fluvial sandstone is commonly explained by weathering under temporarily arid climate, causing oxidizing conditions during vadose zone diagenesis (e.g. Kozur and Bachmann 2010;Förster et al. 2010). The humid intervals during the Carnian (Kozur and Bachmann 2010) may have provided temporarily phreatic zones in which anoxic conditions could have prevailed leading to a mobilization and redistribution of Fe-phases . Under anoxic conditions, Fe-reducing bacteria would have used Fe(III) from minerals as electron acceptor to degrade relatively fresh and reactive organic matter. Iron became mobilized as Fe(II) and was re-oxidized as haematite coatings on the sand grains at redox boundaries within the laminated sediment (Busigny and Dauphas 2007;Förster et al. 2010). This explanation is in line with relatively low organic carbon contents, allowing for temporarily oxic conditions. Where the sediment is more finegrained and adjacent to coal seams, reductive conditions were maintained. These zones are recognized by beige or green reduction spots within the otherwise red sandstone. A similar explanation for early Fe mobilization has been proposed for aeolian Navajo Sandstone on the Colorado Plateau (Utah, Colorado, Arizona and New Mexico; Chan et al. 2000Chan et al. , 2005. Possibly acidic pore water in forest soil or in freshwater swamps, as they prevailed temporarily during the Carnian Pluvial Episode (Kozur and Bachmann 2010), may have contributed to leaching Fe. Humic substances available surrounding these zones could have acted as chelators for the otherwise insoluble Fe(III) (cf. Lipson et al. 2010).
Overall, the haematite coating is a minor phase in the samples. More quantitatively important are the illitesmectite components. Illite is a weathering product under predominantly arid conditions, whereas smectites or illite-smectite mixed layers are likely the product of seasonally humid conditions during the Carnian (see, e.g. Robert and Kennett 1994;and references therein). It is known that Fe may change its redox state in situ (Kostka and Luther III 1994) such that also silicate-bound Fe may have undergone redox changes in the past. A predominantly three-valent state supports that depositional and early diagenetic conditions were largely oxic, but perhaps, temporarily anoxic, supporting episodic pluvial conditions.
Besides, illitization also could have occurred later during burial. Maximum burial temperatures of 85-135°C (Norden and Frykman 2013) were reached, while the maximum burial depth for the Upper Keuper sediments was about 1200 m (Förster et al. 2006; the Stuttgart Fm. is Middle Keuper; thus, max. burial depth was a bit higher). However, illite crystallinity (Table 1b) is too variable to give a clear indication of burial overprint and timing of illite formation.

Ongoing iron and sulphur cycling
At present, in the buried Stuttgart Fm., Fe-reducing activity is low, despite the fact that there is enough reactive Fe present in the rock. XANES spectra indicate a predominantly oxidized state of the Fe. Also low concentrations of pyrite suggest that abiotic or biotic Fe-reduction is not an important process in the Stuttgart Fm. and neither was it in the past. Since the pore fluid is sulphate-rich brine, formation of sulphide by sulphate reduction is also most likely limited by the low availability of organic matter as electron donor. The ternary diagram of relative abundances of total Fe, TOC and CRS (Fig. 7) shows that Fe-reduction is limited by the availability of organic carbon and sulphide. With an organic carbon and energy source available, microbial Fe-reduction (Eq. 1) could be a dominant anaerobic process under the current geochemical conditions in the Stuttgart Fm. as it is one of the most energyefficient terminal electron-accepting pathways (Lovley and Phillips 1987).
Despite the delivery of ''fresh'' organic matter, this process would be kinetically limited by the defined reactivity of Fe-mineral surfaces (Afonso and Stumm 1992), such that, for kinetic reasons, sulphate reduction (Eq. 2) could likewise be a dominating pathway in sulphate-rich brine (Hansel et al. 2015): Reactive Fe-oxides would then undergo reductive dissolution coupled to sulphide oxidation: The precipitation of Fe-sulphide may cause problems by reducing permeability of the reservoir rock Pellizzari et al. 2017) during injection.
Together with the consumption of one mole of protons, this reaction results in an increase in alkalinity and may therefore also induce carbonate precipitation (e.g. Wehrmann et al. 2009). In addition, following Wächtershäuser's (1988) reaction, a loss of ferrous Fe from FeS with the release of H 2 may lead to alteration of FeS to pyrite (FeS 2 ; Wilkin and Barnes 1996): This reaction can drive a cryptic sulphur cycle (Holmkvist et al. 2011) and, at the same time, also consumes protons, which contributes to an increase in alkalinity.
As shown by Canfield et al. (1993) and a recent molecular study from marine sediments by Reyes et al. (2016), microbially driven Fe-reduction may outcompete sulphur cycling at relatively low substrate levels (presumably low enough that the reactivity of Fe-mineral surfaces is not kinetically limiting) resulting in an excess production of Fe(II) (up to 760 lM dissolved Fe in Ketzin pore water) over sulphide.
Our finding of low Fe-reducing activity is supported by results from a microbial community study in the same sandstones using the 16S rDNA fingerprinting method (Wandrey et al. 2011a(Wandrey et al. , 2011b. According to these studies, the microbial abundance in the rock material is very low and only small amounts of DNA could be extracted. Sequences, which belong to chemoheterotrophs related to Burkholderia fungorum (97% similarity), Agrobacterium tumefaciens (95% similarity) as well as facultative chemolithoautotrophs related to Hydrogenophaga (100% similarity), were identified. All groups are able to oxidize either organic molecules or hydrogen to gain energy (Wandrey et al. 2011a).

Effect of borefluid
Organic matter introduced by the drilling mud ) could trigger microbial processes in the nearwell area. Pellizzari et al. (2017) demonstrated that injectivity loss at the CO 2 storage site in Ketzin was related to microbially mediated processes. Exposure to organics (drill mud components and biodegradation products) caused changes in autochthonous microbial community and acceleration in activity of some microbial groups. A total of 10 6 cells ml -1 (similar to the cell abundance in marine sediments with low TOC or at greater burial depth, [100 m; e.g. D' Hondt et al. 2004;Parkes et al. 2005) was detected in well fluids from a depth of 647 m that were influenced by organic drilling mud (Morozova et al. 2011). Fluorescent in situ hybridization analyses revealed a high abundance of sulphate-reducing bacteria in these fluids (Pellizzari et al. 2016). Furthermore, fermentative halophilic bacteria, which were related to species of Halanaerobium, and sulphate-reducing bacteria distantly related to species of Desulfohalobium were detected by genetic fingerprinting. In this case, it can be clearly demonstrated that the substrate is no longer limited, while Fe(III) becomes limiting as electron acceptor, such that sulphate reduction becomes the dominant terminal electron-accepting process. As a result, injectivity loss due to FeS precipitation was recorded .
Furthermore, methane was detected in a fluid sample from a depth of 647 m after performing a N 2 injection (Zimmer et al. 2011). The N 2 injections (N 2 lifts) were performed to remove organic-rich drill mud and restore the injectivity  in order to prepare the well for CO 2 injection. The formation water affected by the bore fluid contained 64 ml of dissolved gas/l of fluid being composed of 90% N 2 , 4% CO 2 , 0.44% He, 0.22% CO, 0.18% H 2 and 5.7% CH 4 (Zimmer et al. 2011). The fluid also contained acetate, with both H 2 and acetate providing intermediates for methanogenesis. Indeed, methanogenic archaea were detected in high numbers by fluorescence in situ hybridization (FISH) in the well fluids (Morozova et al. 2011). During CO 2 injection, methane was transported with the gas plume towards the monitoring wells. Zimmer et al. (2011) showed that a considerable amount of methane was still present in the N 2 -rich gas from the N 2 lift dissolved in the formation water even after a travel time of nine months within the reservoir (before the arrival of the CO 2 ). This suggests that methane was not readily consumed by anaerobic methane oxidation (AOM) neither through sulphate reduction (Boetius et al. 2000; Eq. 5) nor through Feoxide reduction (Wankel et al. 2012;Eq. 6). Iron reduction coupled to AOM is energetically more favourable than sulphate reduction coupled to AOM (Riedinger et al. 2014), and thus, the latter process would likely prevail under low substrate availability.
AOM may be inhibited due to high salinity of the brine, an effect that is known from highly saline or alkaline environments (Kulp et al. 2007;Boetius and Joye 2009). Accordingly, we could expect that freshening of the pore fluid near the injection well by lower-salinity drilling fluids ) and the subsequent reduction of salinity in the rock pores may stimulate microbial activity ). Despite its low energy yield and long adaptation times (Nauhaus et al. 2007), AOM acts as an important sink of sulphate and methane in many marine sediments and could likewise establish a zone of AOM within CO 2 storage reservoirs. AOM coupled to sulphate or Fe-reduction would efficiently contribute to an increase in alkalinity (Eqs. 5 and 6). AOM zones are well known for the formation of hard lithified layers of carbonate, in particular dolomite (Meister et al. 2007;Meister 2015).

Possible effects of CO 2 injection on Fe-cycling
Effects of CO 2 -injection on the mineralogical composition of the reservoir rock were reported by Bock et al. (2013) in a study of drill-cores recovered from the Ketzin site before and four years after CO 2 injection. Even though the changes in the bulk-rock composition were only slight, small-scale studies focused on the mineral surface reactions revealed alteration of Fe-rich grain coatings after the exposure to supercritical CO 2 . A transformation of haematite to goethite coatings in the CO 2 -penetrated rocks was noticed (Bock et al. 2013). SEM and EMPA analyses also revealed authigenic poikilitic dolomite not observed before CO 2 -injection. Small aggregates are composed of a core of dolomite-ankerite solid solution and mantled by siderite and calcite, consistent with the high Fe-content measured in the Stuttgart Fm. . Siderite may form upon increasing availability of Fe(II) due to Fe-reduction. Siderite would replace the precursor dolomite due to its much lower solubility, even under acidic conditions (Bock et al. 2013). The presence of siderite is also an indicator for non-sulphidic conditions (cf. Rodriguez et al., 2000;see compilation in Meister 2015).
While Fe(III) is more soluble under acidic conditions, CO 2 -injection per se would not drive Fe-reduction. Even in a dissolved state, Fe-reduction would not occur under electron donor limitation. Potential electron donors could be added as organic matter injected together with the CO 2 . Also, bio-available organic compounds could be leached from unreactive solid-phase organic matter by supercritical CO 2 . For example, Scherf et al. (2011) showed a mobilization of up to 39% of sedimentary organic matter by the flow of supercritical CO 2 in laboratory flow-through experiments with undisturbed inner cores of Ketzin sandstone samples, obtained after the removal of drill mud contaminations. In addition, low molecular weight organic acids such as formate, acetate and propionate as well as butanoic, pentanoic, lactic, pyruvic, glycolic and gluconic acid were extracted to a sum of up to 538 lg/kg reservoir rock due to exposure to supercritical CO 2 . The mobilized organic substrate could serve as additional feedstock for the microorganisms and thus induce their growth (Scherf et al. 2011) for a certain time, even if the CO 2 loses its solvent properties due to a pressure decrease or dilution effects. Scherf et al. (2011) also detected intact polar lipid fatty acids, indicating a bacterial origin. Pellizzari et al. (this volume) observed that an increased availability of organic carbon led to a changed autochthonous microbial community of the rock samples exposed to 50 bar CO 2 . Facultative anaerobic organisms affiliated to the genera Ralstonia, Burkholderia and Variovorax, which are capable of nitrate reduction (King 2006;Tiemeyer et al. 2007;Im et al. 2010), became dominant. Even though the detected DNA sequences might represent uncultured species of these genera, known representatives are adapted to a CO 2 atmosphere, and for example, Ralstonia can change to an organotrophic metabolism when organic substrates are available (Park et al. 2011).
These studies show that the amount of organic matter derived from remobilization of TOC could drive significant microbial activity presumably resulting in the reduction of parts of the reactive Fe-coatings and, besides, reduce injectivity through biofilm formation. Even though the amount of organic carbon present in the rock would not be sufficient to reduce significant amounts of the total Fecontent of the sandstone, these studies imply that supply of a significant amount of organic matter with the CO 2 could stimulate microbial growth, thereby inducing diagenetic processes in the reservoir.

Long-term evolution and reactivity of silicates
Based on the discussion above, the major portion of the Fe occurs in the sheet silicate fraction. The silica-bound Fe can be redox active, but it is still a matter of ongoing discussion whether and how Fe is incongruently leached from silicate phases. In smectite and illite, these reactions are estimated to occur on 1-to 10-million-year timescales (Canfield 1989;Canfield et al. 1992). Fe release may be accelerated in the presence of sulphide leading to the precipitation of pyrite. Also more Fe could be leached through silicate alteration due to acidification by CO 2 (Lichtschlag et al. 2015). In situ reduction of Fe in clay minerals under reducing conditions has been demonstrated (Kostka and Luther III 1994;Lee et al. 2006;Ribeiro et al. 2009;Stucki and Kostka 2006), and this process can be catalysed by microbes. Santelli et al. (2001) observed accelerated leaching due to oxidative leaching of Fe from silicates in which Fe occurs in the reduced state. However, this would rather be under oxic conditions and would cause a lowering of the pH. ''Longer''-term experiments over a duration of up to 21 months with rocks of the Stuttgart Fm. at Ketzin exposed to synthetic reservoir brine with a CO 2 pressure of 55 bar were performed by Fischer et al. (2010Fischer et al. ( , 2011Fischer et al. ( , 2013. Already after this (relatively short) time period, signs of alteration were observed, such as anhydrite dissolution and corrosion textures on feldspar. Despite corrosion features, also neo-formed albite was observed.
Independent of Fe-oxidation or reduction, silicate alteration itself (e.g. alteration of plagioclase, volcanic glass or mafic minerals) could significantly contribute to pH buffering. Even though silicates react slowly, they may over long time periods be the dominating pH-neutralizing process. Such a long-term behaviour is difficult to predict. Silicate alteration is likely to be a strong pH-buffering process as it can be observed in naturally occurring CO 2 reservoirs, such as within the methanogenic zones of deepsea sediments (cf., Wallmann et al. 2008;Meister et al. 2011;Scholz et al. 2013;Wehrmann et al. 2016) or in basalt (e.g. Shishkina et al. 2010;Shilobreeva et al. 2011). Also model calculations could be used to predict such longterm behaviour (e.g. Chan et al. 2007). Perhaps, selective leaching of Fe may also enhance the silicate dissolution or alteration rates (cf. Santelli et al. 2001). Long-term monitoring studies and the study of natural analogues (e.g. Bickle and Kampman 2013) will be necessary in order to trace such slow proceeding changes of sediment composition, mineralogy and fluid chemistry within the CO 2 reservoir.

Conclusions
The rock samples taken during deep drilling at the Ketzin pilot test site show a relatively high total Fe-content of up to 4 wt%. Most of the Fe is present in the sheet silicate fraction, such as smectite-illite mixed-layer clay minerals and possibly chlorite, and only minor amounts occur as haematite coatings. The redox state of the Fe is 2.7-2.8 and therefore rather high, but still not entirely oxidized. This composition largely reflects the mixed redox state of the Fe partitioned in the clay mineral phases, consistent with the arid to seasonally humid conditions in the depositional environment in the Triassic (Kozur and Bachmann, 2010) and perhaps minor late diagenetic alteration. The reduced, sulphide-bound Fe-fractions (acid-volatile sulphide and chromium-reducible sulphur fractions) are small. Likewise the total organic carbon content is small (less than 0.3 wt%), suggesting that Fe-cycling in the reservoir rock is carbon limited. Carbon limitation is also reflected in a low microbial abundance.
Leaching of Fe by acidification due to CO 2 -injection may also not significantly stimulate microbially driven Fecycling. However, Fe-redox cycling may be stimulated by sedimentary organic matter mobilized by supercritical CO 2 . These results suggest that microbial activity can be induced by supply of organic substrate in combination with the CO 2 . Perhaps both the addition of hydrocarbons and freshening of the highly saline brine by the drilling fluid might stimulate Fe-cycling in the near-well area. For longterm interactions of the CO 2 with the host rock (over thousands to millions of years), the alteration of clay minerals requires a more detailed examination, as this process may significantly buffer the acidification caused by the CO 2 .
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://crea tivecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.