Fluid–Rock Interactions in Geothermal Reservoirs, Germany: Thermal Autoclave Experiments Using Sandstones and Natural Hydrothermal Brines

As renewable energy, geothermal can contribute substantially to the energy transition. To generate electricity or to harvest heat, high-saline fluids are tapped by wells of a few kilometres and extracted from hydrothermal reservoirs. After the heat exchanger units have been passed by, these fluids are reinjected into the reservoir. Due to the pressure and temperature differences between the subsurface and the surface, as well as the cooling of the fluids in the power plant, unwanted chemical reactions can occur within the reservoir, in the borehole, and within the power plant itself. This can reduce the permeability of the reservoir as well as the output of the geothermal power plant. This study aims to simulate real subsurface reactions using batch and leaching experiments with sandstone or sandstone powder as solid phase, and deionised water or natural brine as liquid phase. It is demonstrated that fluid composition changes after only a few days. In particular, calcite, aragonite, clay minerals, and zinc phases precipitate from the natural brine. In contrast, in particular minerals containing potassium, arsenic, barium, and silica are dissolved. Due to the experimental set-up, these mineral reactions mainly took place on the surface of the samples, which is why no substantial changes in petrophysical properties could be observed. However, it is assumed that the observed reactions on the reservoir scale have a relevant influence on parameters such as permeability.


Introduction
With the Paris Agreement in 2015, the global community has agreed to limit global warming to well below 2 °C. To this end, Germany has drawn up a climate protection plan to comply with the Paris Agreement and achieve greenhouse gas neutrality by 2050. Since the energy industry is responsible for 40% of greenhouse gas emissions in Germany, a shift away from fossil fuels and an expansion of renewable energies is necessary (BMU 2019). A promising path to decarbonisation is the use of geothermal energy, which, unlike wind and solar energy, is produced permanently. Moreover, geothermal energy can not only generate electricity but also heat and cooling energy (Kulasekara and Seynulabdeen 2019). In Germany, the main hydrothermal reservoirs to produce geothermal energy are Jurassic limestones of the Molasse Basin (Mayrhofen et al. 2014;Stober 2014) and Permian and Triassic sandstones of the North German Basin (NGB; Lüders et al. 2010;Wolfgramm et al. 2011;Regenspurg et al. 2010;Franz et al. 2018) as well as within the Upper Rhine Graben (URG;Pauwels et al. 1993;Aquilina et al. 1997;Stober and Jodocy 2011;Stober and Bucher 2015;Sanjuan et al. 2016;Schäffer et al. 2021). While the Jurassic limestones are successfully exploited, especially in the region around Munich, there are hardly any geothermal power plants making use of the vast potential of Permian and Triassic sandstones (Agemar et al. 2014). Unfortunately, some commercial as well as research projects faced severe problems with respect to scaling, caused by the production of hot brines and/ or reservoir clogging, as reinjection of cooled fluids result in substantial permeability decreases (Boch et al. 2017;Wanner et al. 2017). Precipitates can consist of barite, calcite, chlorite, native copper, dolomite, galena, illite, kaolinite, laurionite, magnetite, pyrite, quartz, and amorphous phases (Regenspurg et al. 2015;Griffiths et al. 2016;Orywell et al. 2017;Schmidt et al. 2017;Tranter et al. 2021b). Pauwels et al. (1993) and Nollet et al. (2009) found that natural fracture fillings in sandstones consisted of barite, calcite, galena, pyrite, and quartz as well as anhydrite, calcite, halite, and quartz in the URG and the NGB, respectively.
Despite numerous studies on hydrothermal fluid-rock interactions in sandstones (see Appendix Table 7, e.g. Bertrand et al. 1994;Tenthorey et al. 1998;Milsch et al. 2009; 1 3 Rosenbrand et al. 2014;Orywell et al. 2017;Schmidt et al. 2017), it remains challenging to predict particular processes on-site and derive technical measures as a result of the complex interplay between geochemistry, hydrochemistry, thermodynamics, and kinetics. This study aims to contribute to a better understanding of processes taking place in hydrothermal sandstone reservoirs and their impact on permeability. The research questions are how natural brine interacts with reservoir rocks, which minerals dissolve or precipitate, as well as how these reactions influence the composition of the fluid.
Remlinger and Flechtinger sandstones were chosen as representatives for the URG and the NGB, respectively. Related studies used unfractured or fractured sandstone samples (Milsch et al. 2009;Griffiths et al. 2016;Orywell et al. 2017;Schmidt et al. 2017). For this study, it was decided to perform a first set of experiments with intact samples to investigate the pore space exclusively, and to do a second set of experiments dedicated to both naturally and artificially fractured samples. For comparison, identical experiments were also conducted with ground samples (cf. Gong et al. 2012). This approach, in which only few studies compare powdered or grained samples with cored samples, is rather unusual (Schepers and Milsch 2013a, b).
Another distinctive feature in this study is the choice of a natural fluid for the batch experiments as in Kühn et al. (1998), as well as analyses of a set of anions and trace elements. Normally, deionised water or synthetic solutions are used for such experiments (Milsch et al. 2009;Orywell et al. 2017). Reservoir fluids are brines with a total dissolved solids (TDS) typically within the range of 25-130 g/L in the URG (Stober and Jodocy 2011) and 50-360 g/L in the NGB, respectively (Wolfgramm et al. 2011), mostly with high concentrations of carbon dioxide and trace elements. As a good representative, the acidulous thermal brine of Bad Nauheim spa was chosen (Schäffer and Sass 2016). For comparison, identical experiments were realised with deionised water.

Rock Specimen
Two different sandstones have been well investigated and thus were chosen for the experimental study: Flechtinger sandstone (FS) from Lower Permian Rotliegend formation (Fischer et al. 2012;Monsees et al. 2021), and Remlinger sandstone (RS) from Middle Triassic Röt formation being part of the Upper Buntsandstein (Schwarzmeier 1978). As representatives for the URG and the NGB, they were collected as fresh, unweathered cuboids in quarries in Remlingen and Bebertal.
Cored samples (CS) with a diameter of 28.3 mm and a length of 35 and 20 mm, respectively, were prepared and then washed and cleaned twice in an ultrasonic bath in deionised water. The samples were then put in a compartment dryer at 105 °C for two days and stored under dry conditions. Additionally, ground samples (GS) were prepared by first manually crushing parts of the perforated cuboid to pieces of roughly 40 mm. Subsequently, this material was washed and dried like the CS. It was then ground first in a jaw crusher and afterwards in a tungsten carbide vibratory disc mill.
The original grain size of both sandstones was determined by ESEM. A total of 225 grains were measured for each FS and RS. For silicates with the exception of clay minerals, the average diameter was 281 µm for FS and 157 µm for RS. Clay minerals and other accessory minerals have an average size of < 1 µm. The largest components in RS are mica, which are often over 500 µm in size. Remaining silicate components tend to be within the range of the average grain size. In the case of FS, the largest measured grain is approximately 530 µm. Sizes between 400 and 500 µm are relatively common in this rock. This grain size analysis proved that most of the mineral grains were crushed in the grinding process.

Petrophysical Measurements
Grain density was measured on dry samples fitted into the analysis chamber and purged with helium at least ten times prior to the analysis (Table 2). For the density analysis the sample and analysis chamber were filled with pressurised helium and the sample's grain volume was analysed. Grain density is then calculated with the sample's weight and volume.
Bulk density was measured in a powder pycnometer applying a fine-grained, well-sorted powder (Dry Flo) as the displacement material ( Table 2). As for the grain density, the bulk  Ultraschall-Generator USG 40; Geotron-Elektronik Self-constructed air permeameter Accuracy/detection limit ± 0.06% of the analysed volume ± 1.1% of the obtained value not applicable Detection limit is stated at 1 × 10 −16 m 2 1 3 density measurement is based on the analysis of the bulk volume. Bulk volume is then calculated from the difference between a measurement of powder with a rock specimen and a calibration measurement only analysing the pure powder volume. The specimen's porosity is calculated from the difference between bulk and grain volume.
Compression and shear wave velocities were measured with a Geotron ultrasonic generator (Table 2). Both the ultrasonic emitter (USP-G) and receiver (USP-E) are pressed against the flat core surface at 2 bar pressure. To increase the contact and wave transmission between sensor and specimen, an ultrasonic testing couplant (Magnaflux 54-T04) was applied. Wave velocities for compressive and shear waves are picked up on an averaged signal of 16 single measurements. The analysis was performed at 80 kHz.
Apparent air permeability was analysed in a self-constructed column permeameter (Table 2) and analysed after Klinkenberg (1941). Herein compressed air is injected at constant and pre-defined differential pressures through the sample. Additionally, the sample's cylindrical surface is sealed with a latex membrane pressed against the surface applying compressed air. Table 3 shows the mineralogical analyses done. For the XRF analysis the samples were ground with a tungsten carbide mill and measured as pressed powder pellets. In order to investigate the cored samples with the ESEM, a fracture surface was created as a horizontal cut and then carbon coated. The fractured surface was examined in the centre of the plug as well as in the peripheral area (up to ~ 200 µm width of the rim) to identify differences. For TEM investigations, precipitates washed out from plugs were collected on filters. For sample preparation, they were scraped off the filter and mortared before being applied as a suspension with isopropanol to a carbon-coated copper grid.

Mineralogical Analyses
Phase determination was carried out using automated diffraction tomography (ADT) (Kolb 2012). In this case, the datasets were acquired using a FAST-ADT measurement routine (Plana-Ruiz et al. 2020). A tilt series of diffraction patterns of a crystal was recorded in the nano-beam diffraction mode (NBD) with a smallest aperture of 10 µm and spot size of 0.5 nm. In order to increase the number of reflections recorded in the images, a beam precession angle of 1° was applied. With the help of the program eADT, the set of two-dimensional patterns was merged to form a three-dimensional reciprocal space (Kolb 2012). From this, cell parameters can be determined with an accuracy of 1-2% in distance and around 1° in angle and the space group was analysed (Kolb 2012). Subsequently, an electron diffraction intensity file can be extracted, which can be used for structure solution in direct method programs. For this purpose, SIR2019 was used (Burla et al. 2015).

Autoclave Experiments
The presented set-up of batch experiments was defined after preliminary tests. It aims to study petrophysical and hydrochemical changes under varying conditions. A large number of single experiments resulted by the combination of different parameters shown in Table 4. Comparability is guaranteed by a constant rock-to-fluid ratio of 0.8 and 0.2 for CS and GS, respectively. For all ground samples, 15 g powder was used with 75 g fluid. Precisely weighted CS were dosed with the calculated volume of fluid.
In order to simulate reinjection of thermal fluids under thermodynamic disequilibrium into geothermal reservoirs, all sets of rock and fluid samples were mixed under room Deionised water (DW)/Natural brine (NB) temperature and then heated up in compartment dryers. It was neither possible to manipulate or measure pressure, nor to utilise stirrers within the autoclave/containers. Internal pressure was calculated by gas and vapour pressure to be approximately 1. 3, 4.4, and 11.0 hPa at 30, 90, and 150 °C, respectively (Baehr and Kabelac 2016). The containers were regularly inclined by hand to stimulate mixing. The experiment was executed in seven time steps of 1, 2, 3, 7, 14, 21, and 28 days, each performed in an extra vessel. Before sampling, the container was cooled down to room temperature and weighted to determine the fluid loss. Sampled fluid was centrifuged for 20 min at 2500 rpm. Solid residues were dried and stored. Experiments have also been carried out with acidified fluids based on Yongwang et al. (2009). A disturbance of anion analysis can be avoided by using organic acids. Here, deionised water and natural brine were acidified to pH 4 by adding acetic acid, but these test series did not lead to additional findings, and are therefore not further discussed here. Reasons might be that clay minerals only form in the absence or in very low concentrations of organic acids and that high salt loads suppress the effect of organic acids on dissolution rates (Blake and Walter 1999).
After finishing the test series with deionised water, the series was repeated as leaching experiments. Both sample types and sandstones were treated at 30, 90, and 150 °C. After 7, 14, and 21 days, fluid was extracted and always completely substituted by freshly deionised water. Final sampling took place after 28 days.

Fluid and Hydrochemical Analyses
Deionised water was made from demineralised water with Synergy® UV Water Purification System (Merck Millipore) ( Table 10). As a representative for natural brine, the acidulous thermal brine of Bad Nauheim (Schäffer and Sass 2016) was chosen and sampled the wellhead of Sprudel VII. The carbon dioxide concentration of the brine is so high that a degassing is unavoidable until arriving in the lab and starting the experiments, as well as when storing the brine in fridges. Therefore, the brine was analysed at the beginning of every test series to check the initial physicochemical properties (Table 11). The aim was not to analyse the most accurate concentrations of the original fluid at the wellhead, but the most precise composition of the starting fluid for the experiments.
Experiments at 30 and 150 °C usually did not lose more than 0.5% of fluid volume due to opening, sampling and closing. In two experiments at 150 °C the vessels were opened to fast and a fluid volume of 14% and 18% was lost. At 90 °C, higher fluid losses, typically between 1 and 3%, were observed. Comparisons of measured and calculated electric conductivities, using the results of hydrochemical analysis and the method after Rossum (1975), led to concordant results for all vessel experiments.
After the autoclave experiments, a small fluid portion of every sample, approximately 8 ml, was used directly to measure pH, electrical conductivity (EC), oxidation-reduction potential (ORP), and temperature (HQ40D Multimeter by Hach). All three measurements were realised in a same small test tube to save fluid. Measurement errors are given in Appendix Table 9.
Furthermore, bicarbonate concentration was measured immediately with the HI775 Checker HC for Alkalinity by Hanna Instruments. All major cations and anions were measured with an ion exchange chromatography system (Metrohm 882 Compact IC plus). SiO 2 concentration was measured photometrically (Specord 200 Plus by AnalytikJena) with a calibration of nine standards up to 7 mg/L. Inductively coupled plasma mass spectrometry (ICP-MS, AnalytikJena Plasma Quant MS Elite®) was used for the analysis of 21 trace elements (Al, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Se, Ag, Cd, Sn, Sb, Ba, Tl, Pb, U). A volume of 7 ml fluid sample was filtered with a 0.45-µm Perfect Flow PVDF filter and acidified with ultrapure HNO 3 to a pH < 2. Optimisation of instrumental conditions was performed using a tuning solution, leading to sensitivities in 115 In and 232 Th of 1.4 × 10 6 c/s and 1 × 10 6 c/s, respectively, and an oxide ratio ( 140 Ce 16 O/ 140 Ce) and doubly charged ion ratio ( 138 Ba ++ / 138 Ba + ) < 2%. Helium was used as a collision gas in the integrated collision-reaction cell (iCRC) for minimisation of potential interferences. Four internal standards (Sc, Y, In, and Bi) were added online via a peristaltic pump to all samples and standards in order to compensate for drifts of the ICP-MS system. For the purpose of quality assurance, the certified standard Nist 1640a was analysed. Recovery rates of 21 elements were 101 ± 2% and the relative standard deviation (RSD) of the replicates was 1.6 ± 1.5%.

Hydrochemical Modelling
Changes of hydrochemical compositions due to fluid-rock interactions were calculated with PhreeqC interactive, versions 3.5.0 and 3.7.3 (Parkhurst and Appelo 2013). The database phreeqc.dat was used, as phreeqc.dat and wateq4f.dat are best suited for modelling calcite and CO 2 dissolution in NaCl solutions, as Haase et al. (2013) demonstrated when comparing different databases.
For inverse modelling, PhreeqC calculates all stoichiometrically possible model variants on the basis of the mineral phases defined in the input file, even if these hardly make sense from a hydrogeochemical point of view. It is therefore important to first decide which of the mineral phases can be primarily responsible for the measured changes in hydrochemistry, before starting modelling and second check the results carefully. Mineral phases that might contribute to the fluid composition by dissolution result from the XRD/FTIR analyses (Table 1). However, halite, gypsum, barite, and in RS also calcite, were below the XRD/FTIR detection limit. At least for the experiments with deionised water, it is important to consider these minerals, because even low mineral amounts will influence the fluid composition. Environmental conditions at deposition and diagenesis make it likely that both sandstones contain low amounts of these minerals (Plein 1995;Lepper and Röhling 2013). For the inverse modelling with PhreeqC a fraction of 0.5 wt% is assumed for halite, gypsum, barite, and calcite, respectively. With the experimental conditions chosen here, it seems possible that kaolinite (Fulignati 2020), smectite (Fulignati 2020), illite (Yasuhara et al. 2011), calcite (Yasuhara et al. 2011), and barite precipitate. In contrast, the formation of feldspars (MacKenzie 1957;Euler and Hellner 1961;Martin 1969;Hövelmann et al. 2010), quartz (Matsuhisa et al. 1978;Batkhishig et al. 2005), mica (Ferry 1979), or chlorite (Morrow et al. 2001) is unlikely. The following chemical formulas are used in PhreeqC to calculate molar masses and mass transfers: • KMg 3 AlSi 3 O 10 (OH) 2 for biotite (Pabst 1955), • Mg 5 Al 2 Si 3 O 10 (OH) 8 for chlorite (McMurchy 1934), • K 0.6 Mg 0.25 Al 2,3 Si 3.5 O 10 (OH) 2 for illite (Gaudette et al. 1964), • KAl 3 Si 3 O 10 (OH) 2 for muscovite or K-mica (Richardson and Richardson 1982), • Ca 0.165 Al 2.33 Si 3.67 O 10 (OH) 2 for smectite and montmorillonite.

Batch Experiments with Deionised Water
Fluid concentrations increased considerably within the first two or three days after reactions with both sandstones and were more or less constant thereafter (Fig. 1a, b). However, in the GS at 90 °C and 150 °C, concentrations decreased slightly until the end of the 28 day duration. In general, it can be observed that the higher the temperature, the higher the TDS. Moreover, GS experiments show twice to four times the TDS than the equivalent CS experiments (Fig. 1a, b). With the exception of CS at 30 °C, and 90 °C featuring comparable concentrations, TDS in FS are higher concentrated than in RS by a factor of 1.2 to 1.4. Dominant dissolved species are K, Na, HCO 3 , SO 4 as well as SiO 2 . K concentrations increase during the first three days and stay stable thereafter (Fig. 1c, d). Exemptions are both GS experiments at 150 °C, where K concentrations decrease up to 30% from the third day until day 28. A roughly ten times higher K dissolution occurs from GS in comparison to CS. Na concentrations develop comparably to K concentrations ( Fig. 1e, f), but are always considerably lower and do not decrease after three days in the experiments with GS at 150 °C. Noticeable is the FS-CS experiment at 150 °C ( Fig. 1e), resulting in relatively high Na concentration.
HCO 3 behaves comparable to K. Concentrations already obtained after one or two days are more or less constant thereafter (Fig. 2a, b). Exemptions are the GS experiments at 90 °C and 150 °C where HCO 3 concentrations decrease about 40% in comparison to the maxima reached after a few days. SO 4 concentrations increase strongly within the first two or three days and are thereafter more or less constant or only increase slightly (Fig. 2c, d). HCO 3 and SO 4 dissolve much stronger from GS than from CS, as well as from FS than from RS. In contrast, SiO 2 dissolution is controlled strongly by the temperature and hardly by the sample or sandstone type (Fig. 2e, f). The higher the temperature, the higher the Si concentration. Si concentrations increase rapidly during the first three days and afterwards increase only moderately or stay constant. Exemptions are the FS experiments at 150 °C. Fluids reacted with CS and GS showed a strong increase, also after seven days or a continuous SiO 2 reduction, respectively.
Temperature and sample type have a decisive influence on TDS. However, ion ratios are mainly influenced by the sample type, being almost independent from temperature or reaction time. Deionised water leaching GS leads independently from the sandstone type to quite similar K-HCO 3 -SO 4 -waters (Fig. 3). Whereas deionised water leaching FS-and RS-CS result in Ca-K-HCO 3 -waters and Ca-HCO 3 -waters, respectively. Waters reacted with FS-CS show a substantially higher K concentration at 150 °C than at 30 °C or 90 °C (Fig. 3). All these synthetic waters feature similar anion ratios compared to natural groundwaters of the respective formation, but show clearly different cation ratios. K has higher concentrations, replacing Mg and Na.
Trace elements with concentrations above 10 µg/L are Al, As, Ba, Cu, V, and partly Fe. Al had the highest concentrations in the GS-experiments for both sandstones at 90 °C, followed by GS at 30 °C and GS at 150 °C (Fig. 4a, b). Minerals containing Al seem to dissolve quickly within a few days and then precipitate slowly. The Al concentration is mainly dependent on the sample type-ground or cored-but not on the sandstone type or the temperature. The striking Al peak of 3,674 µg/L (Fig. 4b) was confirmed by a repeated measurement but is ignored in the further interpretation.
In general, the released arsenic from RS is roughly one order of magnitude higher than from FS (observe the different scale in Fig. 4c, d). Interestingly, the highest As concentration results from the CS-experiments at 150 °C, followed by CS at 90 °C. The dissolution increases with temperature, with exemption of RS-GS 150 °C and the FS-GS series. In most experiments, the As concentration increases continuously towards the end of the experiment.
Higher temperatures tend to result in higher Ba concentrations, and more Ba is released from CS than from GS (Fig. 4e, f). Minerals containing Ba dissolve quickly within the first three days and precipitate slowly thereafter in most experiments. V concentrations are higher in the experiments with GS than with CS and are obviously positively correlated with the temperature (Fig. 4g, h). Solid phases containing V also dissolve quickly within the first three days, followed by a moderate concentration increase and in some experiments a slight reduction within the last week.

Leaching Experiments with Deionised Water
The dissolution of components of FS and RS follow different trends. Overall, the dissolution is stronger for GS than for CS and increases with the experimental temperature ( Fig. 5a, b). After the first 7-day-cycle, remarkable TDS values of 102 to 533 mg/L and 110 to 390 mg/L are observed for FS and RS, respectively. At the end of the second cycle after 14 days, the TDS value reaches only about half of this concentration. In the CS experiments, slightly lower concentrations are found in the deionised water in the following two cycles, ranging from 40 to 105 mg/L and 24 to 94 mg/L for FS and RS, respectively. In the GS experiments, the concentration also decreases continuously in the third and fourth cycle, but at the end of the fourth cycle, it is clearly still above the concentration of the CSexperiments (Fig. 5a, b). The only exception is the FS-GS-experiment at 150 °C, where the TDS increased in the fourth cycle in relation to the third cycle.
Regardless of the sandstone type, a K-HCO 3 -SO 4 water develops from the initially deionised water in the GS-experiments and a Ca-HCO 3 water in the CS-experiments. In the FS-CS-experiment at 150 °C, Na is a dominant cation in addition to Ca. As in the autoclave experiments with deionised water (Fig. 1c, d), there are remarkably high K concentrations in the fluids reacting with GS, whereas these are rather low in the fluids reacting with CS ( Fig. 5c, d). Overall, the development of the K concentrations is similar to the TDS. This also applies to Na, although Na concentrations are substantially lower than K concentrations (FS up to 29.5 mg/L, RS up to 22.4 mg/L). Ca and Mg behave differently from the alkali metals. Higher temperatures cause higher concentrations, but Ca concentrations are higher in the CS-experiments than in the GS-experiments (Fig. 5e, f). The Ca concentrations only changed slightly in the third and fourth cycle, compared to the second cycle, and decreased in most experiments. Mg concentrations are rather low, up to 2.2 mg/L and 1.9 mg/L in the experiments with FS and RS, respectively.
The development of HCO 3 and SO 4 concentrations is similar to those of TDS and K ( Fig. 6a-d). A clearly lower solution of HCO 3 reacting with RS-GS at 150 °C is noticeable at the third and fourth cycle. Cl, however, only dissolved in appreciable concentrations in the first cycle (Fig. 6e, f) and its concentrations were mostly similar in the experiments with GS and CS. The dissolution of SiO 2 is also temperature dependent; the higher the temperature, the stronger the solution (Fig. 6g, h). It does not make much difference whether the fluid reacted with CS or GS. Apart from the FS-GS-experiments at 150 °C and both RS at 150 °C, the solution of SiO 2 was more or less constant over the cycles.
Trace elements with concentrations above 10 µg/L are Al, As, Ba, Cu, Fe, and V, as in the autoclave experiments with deionised water (Fig. 4). Al concentrations are higher in the experiments with GS than with CS (Fig. 7a, b). The latter also shows an influence of the temperature; the higher the temperature, the higher the solution of Al-phases. A reduction of the Al dissolution over the cycles is, however, not visible. Like in the autoclave experiments with deionised water, As released from RS is roughly one order of magnitude higher than from FS (Fig. 7c, d). An effect of temperature or the sample type on the As concentration cannot be recognised. The As concentration tends to decrease with advancing cycle numbers. In the CS-experiments, the dissolution of Ba-minerals is positively correlated with the temperature (Fig. 7e, f), whereas in the GS-experiments, nearly the same amount of Ba dissolved independently from the temperature. Ba concentrations tend to increase with advancing cycles. Like in the autoclave experiments with deionised water, V concentrations are higher in the experiments with GS than with CS and are obviously positively correlated with the temperature (Fig. 7g, h). Less V dissolved with advancing cycles, a behaviour that is comparable to TDS, K, HCO 3 , or SO 4 .

Batch Experiments with Natural Brine
During the experiments it turned out that the sampled brine volume was not sufficient to realise the test series at 150 °C and is therefore missing in the following. The TDS decreases unevenly from about 28.5 to 27.0 g/L (Fig. 8a, b), which corresponds to a decrease of about 5%. The behaviour of both major and trace elements varies. Some concentrations decrease, some remain constant, and others decrease. The decrease of TDS is almost exclusively caused by Ca and HCO 3 . Overall, Ca and HCO 3 concentrations decrease by 200 to 500 mg/L and by 250 to 1000 mg/L, respectively. Ca precipitates within the first three days and within the first week for RS-CS and stagnates thereafter (Fig. 8e, f).
The reduction in dissolved Ca is greater in both sandstones at 90 °C than at 30 °C and greater in GS than in CS. For HCO 3 , the precipitation behaviour is also similar in both sandstones, hardly dependent on the sample type, but rather on the temperature (Fig. 9a, b). At 90 °C, precipitation is strong within the first three days and then weaker, but steady. At 30 °C, the HCO 3 concentration initially increases after three days and then decreases steadily until the end of the experiment. The concentrations of Na, Mg, SO 4 , and Cl remain more or less constant over the test period of 28 days, shown exemplarily for Cl in Fig. 9c, d. The concentrations of these ions in the natural brine are so high that possible changes caused by fluid-rock interactions are not noticeable or are subject to measurement uncertainties.
The evolution of the K concentration depends strongly on the sample type, a little on the sandstone, and hardly on the temperature (Fig. 8c, d). In the GS experiments, a strong K solution already occurs after three days. Overall, K increases by almost 100% in FS and by 50% in RS. In the CS experiments, however, the K concentration remains almost constant in FS and slightly decreases in RS. Si also shows a differentiated behaviour (Fig. 9e, f). In the experiments at 30 °C, the concentration remains approximately constant regardless of the sandstone or sample type. In the experiments at 90 °C, the concentrations increase substantially. For GS, the increase occurs within the first three days and is about twice as large as for CS, where a gradual increase in concentration is observable.
For As, there is a temporary mobilisation limited to the first week of the experiment (Fig. 10a, b). The already considerable As concentrations of the natural brine are exceeded many times during this period and are higher at CS than at GS and tend to be higher at 30 °C than at 90 °C. After the first week, As concentrations fall back to about the original level. Only in the CS experiments at 90 °C does the As concentration increase in both sandstones as a result. With the exception of FS-CS, the Ba concentration increases strongly after three days and reaches a multiple of the initial concentration at the end of the experimental series (Fig. 10c, d). Apparently, more Ba dissolves at 90 °C than at 30 °C. The value at RS-CS after 21 days is considered an outlier. V concentrations are actually unremarkable with one exception (Fig. 10e, f): in the GS experiments at 90 °C, the concentration increases substantially for both sandstones. Zn, in turn, precipitates rapidly, and is almost completely irrespective of the sandstone, the sample type, and the experimental temperature (Fig. 10g, h). Also, the bulk of Cu in solution precipitates within the first days, independently of the sample type and temperature.

Batch Experiments with Deionised Water
The hydrochemical data (Sect. 3.2.1) provide clear indication that solution and precipitation have taken place. Thus, ESEM and EDS measurements on unaltered and altered sandstones have been taken. Figure 11 shows exemplarily an ESEM image of FS where K-feldspar, illite, and albite are visible. In all samples the dissolution features are most frequently seen in feldspars, in which they preferably take place along the two cleavages and are visible as flat surface areas. This effect was also observed in experiments of Morrow et al. (2001) and Tenthorey et al. (1998). The remaining minerals are not visibly affected. In general, FS and RS surfaces appear similar.

Batch Experiments with Natural Brine
ESEM and EDS measurements from FS and RS reacted with brine from Bad Nauheim were taken. The samples exhibit etching of feldspar along its cleavages. Additionally, in both samples several precipitated phases could be found at the rim of the plug, approximately 600-700 µm towards the centre. Figure 12 shows these precipitates exemplarily for RS at 90 °C.
The largest precipitates are acicular and dodecahedral calcium carbonate crystals (Fig. 12a, b). The acicular habit is pure calcium carbonate, while the dodecahedral crystals can also contain ~ 1 at% Mg. Sizes of the calcium carbonate crystals vary with the duration of the experiment and reach a maximum size of around 60 µm as needles and 25 µm as dodecahedra. Crystals in RS are generally smaller than those in FS. In addition, precipitation of NaCl was found in three out of four samples. These accumulate in layers on the surrounding grains or gather in grooves and creases (Fig. 12c). Additionally, an Fe-containing phase was precipitated (Fig. 12d) that occurs in aggregates of a few µm, whereby the individual crystals rarely exceed a size of 200 nm. To determine the calcium carbonate phases -with and without Mg -TEM measurements were taken. First, EDS was used to identify the different calcium carbonates by their Mg content. Then, FAST-ADT measurements were conducted. A tilt series from − 50° to + 50° was recorded in 1° steps with an exposure time of 0.5 s per diffraction pattern and a beam size of 150 nm. From this, a three-dimensional reciprocal space was reconstructed and the cell parameters were determined. The cell parameters indicated the pure calcium carbonate to be aragonite (experimental cell parameters: a = 5.057 Å, b = 8.012 Å, c = 5.540 Å, α = 89.2°, β = 90.4°, γ = 89.4°) and the phase with Mg to be calcite (experimental cell parameters: a = 5.036 Å, b = 4.953 Å, c = 17.064 Å, α = 90.2°, β = 89.8°, γ = 119.4°). To validate these results, a structure solution with. hkl data was carried out using SIR2019. These two different structures found agreed with structures from literature (De Villiers 1971 for aragonite, Effenberger et al. 1981 for calcite). The TEM investigation showed that the Fe-rich phase had an amorphous diffraction pattern and could therefore not be further characterised. An insight in the precipitation mechanisms of poorly crystalline Fe-phases can be found in Grundl and Delwiche (1993). The existence of calcite, aragonite, and halite was confirmed by an XRD measurement directly on the filtered washed out material, as shown in the powder diffractogram in Fig. 19 in Appendix.

Petrophysics
Petrophysical properties were analysed on the dataset of deionised water as a proof of concept for further planned experiments. Nonetheless, the data generated are included in this study as they provide insights into the applicability of four parameters to quantify effects  (2010) of alterations. To characterise the selected sandstones, petrophysical properties were measured on a set of samples not dedicated for batch experiments. Nonetheless, the samples were prepared from the same sandstone blocks for comparability. Petrophysical property  analysis comprises, bulk density, porosity, compressional wave velocity (Fig. 13), and intrinsic permeability (Figs. 14 and 15). Averaged values are summarised in Table 5. After every alteration in the autoclave experiments with deionised water, petrophysical properties were measured to identify any alteration of petrophysical properties of the rock samples. Analysing petrophysical properties prior to alteration tests was only applicable for air permeability. Measurements of grain and bulk density as well as compressional wave velocity require the application of additives such as Dry Flo (fie-grained powder; density measurement) or couplant (wave velocity) which can contaminate samples. Therefore, these parameters are compared to average data provided in Table 5.
Evidently, due to the experimental time of the batch experiments, only a single sample of each temperature and test duration could be analysed for petrophysical properties. Therefore, natural variabilities within the test set cannot be eradicated with statistical measures. Variability is most visible in density measurements (Fig. 13). Nonetheless, the data shown above indicate changes due to alteration during the batch experiments. Altering petrophysical properties are also recognisable in measured apparent air permeability (Figs. 14 and 15) which was measured before and after batch experiments on each sample.

Hydrochemistry
The leaching experiments show that the soluble amount of K, Na, Ca, Mg, HCO 3 , SO 4 , Cl, As, and V phases decrease with every leaching cycle. Therefore, the solution of these species is controlled by the amount of mineral or sample surface available and not by saturation indices. In contrast, concentrations of dissolved SiO 2 and Al remain constant over time and concentrations of dissolved Ba even increase slightly. For these species, reaction velocities and saturation indices are relevant rather than their occurrence in the samples. Lüders et al. (2020) also report somehow unexpected continuously increasing Ba concentrations and suppose that Ba is released by prolonged mineral dissolution.
Within the experiment with deionised water, barite, calcite, chlorite, gypsum, halite, illite, and quartz seem to dissolve or react (Table 6, cf. Ludwig et al. 2011). At the end of the experiments, barite, calcite, and quartz are in the range of saturation (Fig. 16), whereas gypsum and halite are still clearly undersaturated. Na and Cl in solution may originate from halite, Ca and HCO 3 from calcite, SO 4 from gypsum, and Ba from barite, witherite, celsian or other not detected Ba-minerals. Ba also may be released from impurities of other minerals like biotite, muscovite or K-feldspar (Deer et al. 2013).
With increasing temperatures As is more mobile and may originate from not detected sulphides like arsenopyrite or iron oxides (Smedley and Kinniburgh 2002;Bonte et al. 2013). The latter could also be a source for V (Bonte et al. 2013). Muscovite (Lee et al. 1997) or fluorapatite (Härter et al. 2020) could also be geogenic sources for the noteworthy V concentrations. A contamination of V caused by the tungsten carbide vibratory disc mill cannot be excluded, but seems rather unlikely, as noticeable concentration trends are also observed in the experiments with cored samples (Figs. 4 and 7). These results are in accordance with Lüders et al. (2020), who report that As and V, among other elements, are released as a pulse in column tests with increasing temperature. Due to plots in E H -pH-diagrams and the PhreeqC calculations, dissolved Al, As and V form Al(OH) 4 − , HAsO 4 2− and H 2 AsO 4 − as well as VO 2 (OH) 2 − and VO 3 OH 2− oxyanions. Bonte et al. (2013) andLüders et al. (2020) also documented a preferential release of oxyanions, which correlates to the weaker sorption of anions like arsenate or arsenite at increased temperatures (Kersten and Vlasova 2009). The observed decrease of HCO 3 − (90 and 150 °C, FS-and RS-GS) can be explained by the re-precipitation of calcite. Due to the inverse modelling, illite seems to react to smectite and at 150 °C also to kaolinite (Table7), releasing K, Al, and SiO 2 (Ludwig et al. 2011). Illite, smectite, and kaolinite are close to saturation at the end of the experiments at 30 °C, whereas all three minerals are substantially oversaturated in the experiments at 150 °C (Fig. 16). A decrease of K (150 °C, FS-and RS-GS), Ba (150 °C, RS-CS and RS-GS), and Al concentrations (all temperatures, FS-GS) may be the results of smectite or kaolinite formation (Table 6), or ion exchange.
Substantially increased concentrations from the experiment with ground samples compared to cored samples are caused by the larger reactive surface area and ion exchange. Ion exchange reactions can be recognised by the fact that the higher-valent ions such as Ca 2+ , Mg 2+ , Ba 2+ , and Sr 2+ show different curves than the TDS (Figs. 1, 4). Monovalent cations such as Li + , Na + , and K + show similar curves. Lüders et al. (2020) argue that a preferential sorption of bivalent ions in exchange for monovalent ion takes places with increasing temperatures. Ion exchange takes place especially at the phyllosilicates, where cations are bound according to their binding strength (Udo 1978). However, the large surface area of the phyllosilicates is only available for fluid-rock interactions if the fluid has contact with this surface. In CS only a small fraction of the fluid reaches the silicate surface. The process is thus almost completely diffusion-controlled. Especially Ca 2+ , Ba 2+ , and HAsO 4 2− are  more concentrated in fluids reacted with CS than with GS, because cation exchange hardly takes place in CS, whereas a lot of Ca 2+ , Ba 2+ , Mg 2+ , and Sr 2+ are bound in GS. Other ions like K + , Na + , HCO 3 − , or SO 4 2− are much stronger mobilised from GS than from CS due to the larger available reaction surface. From the analysed cations, Ba is most strongly bound to phyllosilicates and can also substitute Sr in different minerals like strontianite or celestite (Tranter et al. 2021a). In any case, the different concentrations of Ba and SO 4 indicate that there must be other sources and sinks for Ba besides barite.
In the experiments with natural brine, possible reactions of most species are masked by their high natural concentration. The results of the inverse modelling (Table 6) suggest that barite or other Ba-minerals, k-feldspar, and-only in the experiments at 90 °C-quartz are dissolved. This would at least explain the observed increase of K (Fig. 8c, d), SiO 2 (Fig. 9e, f), and Ba (Fig. 10c, d) in solution. Also Bonte et al. (2013) observed elevated K and SiO 2 concentrations and suggested as explanation congruent, partly incongruent dissolution of K-feldspar. At temperatures about 100 °C, K-feldspar weathering can cause elevated K concentrations (Holm et al. 1987), whereas the release of hydrated silica from K-feldspar is controlled by reactions with Al(OH) 3 at room temperature (Wollast 1967). Zn precipitates most probably as carbonate hydroxides, like hydrozincite, or in combination with Cu as rosasite, including smithsonite or aurichalcite as minor product.
Barite and quartz are slightly oversaturated at the end of the experiments (Fig. 17). Due to the inverse modelling, calcite, gypsum, kaolinite, and in most experiments also illite may have precipitated (Table 6). In the case of calcite, this fits well with the observed decrease of dissolved Ca (Fig. 8e, f) and HCO 3 (Fig. 9a, b) and the calculated saturation indices (Fig. 17). In contrast, a precipitation of large amounts of gypsum appears not to be very likely due to constant SO 4 concentrations in solutions and slightly negative indices (Fig. 17). Illite, kaolinite, and smectite seem to be rather oversaturated (Fig. 17), but a possible formation of these minerals cannot be attributed to hydrochemical changes (Figs. 8, 9 and 10).

Mineralogy
Electron microscopy on FS and RS with non-altered, altered by deionised water, or altered by brine revealed etched feldspars in all samples and on the whole fractured surface. Therefore, these features were assumed not to be produced by the experiment. Since in both sandstones the cementation consists to a large extent of clay minerals, accumulations of those often occur, especially in pores. It is assumed that only the cementation is affected by the fluids, not the grains themselves. There, fibrous illite (Guven et al. 1980) connects the grains via a meshwork (Fig. 18). For FS this was also observed by Fischer et al. (2012) and Jacob et al. (2021). These clay mineral networks provide a large contact surface for solution processes.
Comparing Fig. 11 with Fig. 18, illite fibres are found to be larger in the altered sample than in the unaltered sandstones. They even entangle crystals of several tens of microns in size.
For sandstones altered with deionised water no substantial amount of other precipitates have been detected, thus it proved to be difficult to detect solution processes in the rock using electron microscopy. In the case of brine altered sandstone, even larger precipitated particles were detectable. The fact that these precipitates could be found only in the rim of the plug indicates that the fluid reaches only around 600-700 µm into the plug. The Na and Cl required for the formation of halite originate from the brine, as no halite is present in the experiments with deionised water. The question that arises here is whether the halite precipitates during the test period or only when the rock is drying. Due to the bulky shape, it is assumed that this only happens during the drying process and therefore does not play a role in the experiment. The best indicator for the brine interaction is the precipitation of large calcite and aragonite crystals. Here, it is likely that the required ions originate from the brine, as the crystals are only formed in connection with it. Calcite is the stable phase of CaCO 3 at ambient pressure (Davis and Adams 1965); however, at a high level of calcite supersaturation, aragonite nucleates as well regardless of the low pressure. As opposed to aragonite, calcite can incorporate Mg in its structure. This is in agreement with Sun et al. (2015). In this paper it is also described that calcite growth is inhibited by the incorporation of Mg, while it does not affect the growth of aragonite. This could be the reason why there is more aragonite present in the precipitates than calcite.

Petrophysics
As shown in Figs. 13, 14 and 15, some variations of petrophysical properties due to alteration can be observed. Also, a dependency to test duration and temperature could be observed. Nonetheless, alteration in bulk density cannot be measured accurately. Variations in each sample set are within a range of 0.08 g/cm 3 (FS) and 0.04 g/cm 3 (RS) which is likely within the methodological measuring error. Measured porosity also varies within a range of approximately 2% but is in good accordance with density measurements.
Beside density and porosity, compressional wave velocity proved to be the most sensitive parameter to identify alterations. Compared to data gathered on fresh rock samples  Whitney and Evans (2010) 1 3 of the same outcrops (test duration at 0 d along 14,15), compressional wave velocity decreases by approximately 300-450 m/s after the batch experiments. Figure 15 also indicates a negative trend with increased test duration, as well as decreasing compressional wave velocities at increasing test temperatures. Such trends are expected, considering the increased TDS at elevated temperatures. Dedicated peaks at 3 d 14,15) test duration (as seen in the hydrochemical analysis) could not be observed in petrophysical properties.
Besides compressional wave velocity, the intrinsic permeability is also affected by the batch experiments. Although no dependency between intrinsic permeability and test duration or temperature could be found, the permeability increases in almost every batch experiment (Figs. 14 and 15).
Since no flow-through experiments were performed, it can be assumed that only the surface of the sample is affected by fluid-rock interactions. While no direct dissolution of complete minerals is observable, changes in petrophysical properties are most likely due to alterations in either filling of microfractures or inter-grain contacts, such as alteration in cementation. Resulting from the PhreeqC models, dissolution affects minerals such as calcite, halite or gypsum as well as quartz or illite. Since the cementation of both sandstones is mostly clayey and siliceous (FS also calcitic) it is likely that, for example, the ultrasound transmission is decreased due to alterations in cementation. Furthermore, changes might be caused by washing out fine-grained particles such as clay minerals at the surface of the rock cylinder, causing, for example, higher permeability at the very edge of the samples.
It also needs to be mentioned that petrophysical properties were only measured on a single specimen of each temperature and duration set-up. Hence, occurring variabilities in the used rock specimens cannot be eradicated.

Conclusion
The chosen approach allowed to perform a large test series in rather short time due to the simultaneous run of experiments under different settings. Thus, it was possible to compare the behaviour of two different sandstones, cored and ground samples as well as deionised water and natural brine at different temperatures. Up to our knowledge, natural brine is rarely used in such experiments. The experiments provided sufficient fluid for a comprehensive hydrochemical measuring program, including EC, pH, ORP, as well as main and trace components analysed by IC, ICP-MS and photometry. It is rather uncommon to analyse anions and trace elements in such detail due to the high TDS of natural or synthetic brines.
In most experiments a substantial change of the hydrochemical composition was already observed within the first three days. The development of TDS depends not only on time, but some correlation of particular species to the sample type or experiment temperature. The two most surprising findings in the experiments with deionised water were that potassium becomes the main cation which is completely unusual for natural waters penetrating sandstone and that smectite or other clay minerals probably precipitate due to an oversaturation in the meantime (Ludwig et al. 2011). However, clay minerals may have been affected by drying the samples at 105 °C even before the tests, which is why drying at a maximum of 65 °C would have been more appropriate.
In the experiments with natural brine, an overall precipitation of calcite and aragonite could be found, whereas the concentrations of dissolved potassium and silica increased. Sources and sinks of some main and trace elements could not be clarified beyond doubt. Nevertheless, a mobilisation of arsenic, barium, and partly vanadium was observed (cf. Himmelsbach and Wendland 1999;Bonte et al. 2013;Lüders et al. 2020), whereas zincphases precipitate.
In mineralogical investigations with ESEM, TEM, and ADT, both RS and FS show comparable phenomena. Dissolved minerals can be seen in both sandstones, but similar features are also present in the unaltered plugs, so that it cannot be precisely differentiated whether they originate from geological processes, sample preparation, or the experiment. The lamellar or fibrous habit of clay minerals can lead to a partial clogging of the pores, reducing the permeability of the rock (Griffiths et al. 2016;Jacob et al. 2021;Tenthorey et al. 1998). In addition, loose particles can get caught on such structures, as well as on acicular aragonite, and thus further reduce the permeability due to filtering effects.
By the means of ADT, the two calcium carbonate phases aragonite and calcite were determined in the precipitates of the experiments with natural brine. The higher proportion of aragonite compared to calcite confirms the statement of Sun et al. (2015), according to which the incorporation of magnesium into the crystal lattice of calcite inhibits its formation, while that of aragonite remains unhindered. Halite crystals are formed as a residue during drying and do not precipitate during the experiment.
Despite the porosity of the sandstone, the experimental set-up has the weakness that the fluid does not penetrate the samples due to a lack of pressure and a hydraulic gradient, but reactions just occur close to the surface. Thus, mineral dissolution or precipitation did not take place in the interior of the sample and hence did not cause substantial changes in the investigated petrophysical properties. To observe changes in permeability, thermotriaxial devices (e.g. Pei et al. 2014) or special reactors (e.g. Frank et al. 2021) are necessary, but the experiment series presented here could by far not be realised in such devices in a handy time frame. Experiments on deionised water showed that ultrasonic wave velocity is the most sensitive parameter to identify alterations caused by batch experiments. To improve statistical reliance, further experiments including larger dataset at each time and temperature level are needed. Further, compressional wave velocity is measured before and after the alteration experiment on each sample. In the case of thermo-triaxial or other flowthrough experiments, artificial alterations of density and porosity should be improved.
However, experiments in which fractures would also be investigated instead of the rock matrix, in order to test for influences on the permeability, are also interesting (Frank et al. 2020a, b). Therefore, as a next step, we are planning similar experiments with RS with natural fractures partially filled with calcite. More studies are listed in Baudracco and Aoubouazza (1995) and Rosenbrand et al. (2014)