Geochemical Reactions of Na-montmorillonite in Dissolved scCO 2 in Relevance of Modeling Caprock Behavior in CO 2 Geological Storage

One of the challenges of the present century is to limit the greenhouse gas emissions for the mitigation of climate change which is possible for example by a transitional technology, CO2 geological storage. Clay minerals are considered to be responsible for the low permeability and sealing capacity of caprocks sealing off stored CO2. However, their reactions are not well understood for complex simulations. This work aims to create a kinetic geochemical model of Na-montmorillonite standard SWy-2 supported by solution and mineral composition results from batch experiments. Such experimentally validated numerical models are scarce. Four 70-hours experiments have been carried out at atmospheric conditions, and with CO2 supercritical phase at 100 bar and 80 °C. Solution samples have been taken during and after experiments and their compositions were measured by ICP-OES. The treated solid phase has been analyzed by XRD and ATR-FTIR and compared to in-parallel measured references (dried SWy-2). Kinetic geochemical modelling of the experimental conditions has been performed by software PHREEQC. Experiments and models show fast reactions under the studied conditions and increased reactivity in presence of scCO2. Solution composition results cannot be described by the change of the uncertain reactive surface area of mineral phases. By considering the clay standard’s cation exchange capacity divided proportionally among interlayer cations of Na-montmorillonite, the measured variation can be described on an order of magnitude level. It is furthermore indicated that not only the interlayer cations take part in this process but a minor proportion of other, structural ions as well, differently in the reference and scCO2 environments.


Introduction
It is beyond doubt that human activity influences the climate system and that mitigating climate change requires reduction of greenhouse gas emissions [1].The Paris Agreement [2] aims to keep global temperature rise well below 2 °C and pursue efforts for a 1.5 °C limit.This "low-stabilization" scenario can only be reached by not only applying low, but also net negative emission technologies [3].In the near-term, the only approach capable of making a major contribution to these seems to be the transitional technology of Carbon Capture and Storage (CCS) [4] and its combination with bioenergy (BECCS) [3].
When CO 2 is to be stored in a potential deep geological reservoir, special attention must be paid to its low permeability seals, caprocks [5].These rocks represent the physical barriers that could hinder the escape of injected supercritical CO 2 (scCO 2 ).Currently, it is not completely clear what factors (i.e.mineralogy, grain-size, cementation) are responsible for their retention capacity.It is assumed, however, that clay minerals play an essential role in the low permeability of these lithologies.It is crucial, therefore, to observe, understand and predict (model) how clays behave in the presence of reactive CO 2 which is in contact with [6] and present in the caprock both in dissolved and free phase [7,8].
For the observation of mineral reactions induced by dissolved CO 2 , one can either study natural analogues or conduct laboratory batch experiments.Despite of the drawbacks (for instance, the usually short laboratory time scales, non-natural solution compositions, variations in water / mineral ratio, the potential formation of experimental artefacts), these treatments and study of experimental solutions and solids are efficient to derive conclusions on long-term processes in CCS systems.With the possibility of "on-the-go" fluid sampling [9], these batch experiments can better support theoretical geochemical models due to the higher sensitivity of detecting solution than mineral composition changes and excluding the effect of reactor depressurization and cooling artefact formation in these samples.However, model validation by such experimental solution data is scarce.Also, there were only a few attempts made to couple any batch experimental results of soaked CCS caprocks with modeling [7,10,11].Moreover, the number of experiments focusing especially on clay mineral reactions for the effect of dissolved CO 2 in pore water is very low since most clay studies have been conducted not specific to CCS [12] or in free phase scCO 2 [13][14][15][16][17][18][19][20][21][22].
Sendula et al. [9] studied illite, kaolinite and montmorillonite mixtures in scCO 2 -saturated water.Their results indicated rapid and total dissolution of any traces of carbonates.The clay mixtures, initially containing montmorillonite, were more reactive than the sample of illite-kaolinite mixture.However, further results are needed to understand the behavior of cations, especially to gain data to be incorporated into geochemical models.
In present work batch experiments of pure SWy2 Namontmorillonite clay standard in scCO 2 -saturated water with "on-the-go" solution sampling are carried out and their data are used to fit the standard's kinetic PHREEQC model.Solution samples taken before and after reactor depressurization and cooling also provide information about expected artefacts in the solid samples in this and similar experimental works.Since the high pressure and temperature CO 2 -experiments were prepared by mixing the clay with water at atmospheric conditions, additional reference experiments and modeling were performed in this ambient environment.The interlayer cations, and among them mostly the Na + and Ca 2+ , are expected to be mainly responsible for the high cation exchange capacity (CEC) of the clay which equals to 76.4 meq (milliequivalent) for 100 g weight [23].Based on the given stoichiometric equation the molar mass (M) is 549.07 g/mol and therefore, the CEC is calculated to be 0.42 in mol/mol units.Note that CEC of clays has been shown to decrease with decreasing pH [24,25].
The specific surface area (SSA) of Na-montmorillonite is large, the N 2 area is given to be 31.82± 0.22 m 2 ⁄g [23], but several even higher other data can be found in the literature.For example, SSA is mentioned to equal to 48 m 2 ⁄g [26] and, with taking into account the interlayer region, also to 750 m 2 ⁄g [27].The RES³T database [28] goes as high as 898 m 2 ⁄g SSA.
To use in experiments, pure, deionized water with less than 6.7 × 10 -2 μS⁄cm electric conductance was prepared in the laboratory of Mining and Geological Survey of Hungary by multiple distillation steps and filtering.For the acidification and therefore, stabilization of any solution samples taken, a 99:1 ratio mixture of the above deionized water and 68 % HNO 3 was used.In the CO 2 -experiments Linde ® CO 2 gas (> 99.5 %) was applied freshly distilled.

Experimental procedures and samples 2.2.1 Reference SWy-2 samples
During the experimental work, 8 reference samples have been prepared for the same analysis at the same time as the treated samples.These aimed to represent the average mineral composition of the SWy-2 clay standard and to follow any time-sale changes of the analytical methods.These SWy2 samples were dried at 80 °C for 30 minutes and marked from SWy-2-0a to 0h.

Atmospheric batch experiments
Two repeats of atmospheric experiments were carried out in the laboratory of Mining and Geological Survey of Hungary.In a glass vessel, 1.6 g of SWy-2 was layered by the deionized water in multiple steps to reach approx.35 water / mineral (mass) ratio.This ratio is highly increased compared to the natural, approx.0.2 porosity / mineral ratio in caprocks however, necessary for homogenization and possibility to sample the solution.
The clay-water mix then was placed onto a magnetic stirrer for continuous mixing during the 70 hours (three days) experiments.The air temperature was 25 °C and the vessel was covered by a glass cap to decrease the evaporation of the water.At gradually increasing time intervals, fluid samples of 0.5 g were taken by a syringe from the top of the mixture.These samples were filtered through an attached membrane filter with pore diameter of 0.45 µm to remove solid particles and the pure solution drops with measured mass were added into a determined amount of stabilization solution.At the end of the experiments the solid material was separated from the rest of the solution on a glass filter dried at dust-free laboratory air for a week, then sent to analysis.The experimental conditions of these samples marked with names starting with SWy2-ATM are summarized in Table 1.

Dissolved scCO2 batch experiment
Three successful high pressure and temperature experiments were carried out in the laboratory of Budapest University of Technology and Economics where suitable reactor for dissolved scCO 2 -treatments is available.Similarly to the atmospheric experiments, but in smaller proportions, about 0.35 g SWy-2 was layered by the deionized water in multiple steps to reach the 35 water / mineral ratio in a special glass vessel.
The layered sample then was placed onto a magnetic stirrer for mixing until homogenization was reached.The vessel exactly fits into the high pressure and temperature resistant reactor with the possibility of "on-the-go" fluid sampling, which schematic representation was published by Sendula et al. [9].The reactor volume then was saturated by CO 2 and heated and pressurized up to reach 80 °C temperature and 100 bar pressure.This was a technically feasible pressure-temperature range close to the estimated conditions for potential local, Hungarian CCS caprocks in an earlier study [11].During the 70 hours (three days) treatments the magnetic stirring of the water-mineral mixture was continued except at the time of "on-the-go" and post-experimental solution sampling.These were all filtered by a syringe attached membrane filter (0.45 µm) and added into the stabilization solution.The solid material was separated from the rest of the solution by a glass filter and/or drying on laboratory air in dust-free environment.The experimental conditions of these SWy2-WCO2 samples (meaning treated in water and CO 2 ) are summarized in Table 1.

Applied analytical techniques 2.3.1 Mineral composition determination
The reference samples and the air dry treated samples were all analyzed by X-Ray Diffraction (XRD) and Attenuated Total Reflectance Fourier Transform Infrared Spectroscopy (ATR-FTIR) in the Mining and Geological Survey of Hungary.For XRD measurements a Philips ® PW 1730 instrument and for ATR-FTIR analysis a Bruker ® Vertex 70 with Bruker ® Platinum single reflection diamond Table 1 Treated SWy-2 samples, their experimental conditions and times of "on-the-go" fluid sampling.Water / mineral ratio is given as at the beginning of the experiments.In the case of scCO 2 -experiments, solution samples have been taken also after depressurization and cooling, "post-exp.".ATR module and mercury-cadmium-telluride detector were used.Find more detailed description of these methods in Sendula et al. [9].

Solution composition determination
The Ca-, Mg-, K-, Na-, Fe-, Mn-, dissolved Si-and Al-concentrations (uncertainty 2-20 % for high and low concentrations) of experimental solution samples were analyzed by Inductively Coupled Plasma-Optical Emission Spectrometry (ICP-OES).For the measurements, a HORIBA Jobin Yvon ® ULTIMA 2C instrument at the Mining and Geological Survey of Hungary was used.
Further details can be found in Sendula et al. [9].

Kinetic geochemical modeling
Kinetic geochemical modeling was carried out in the PHREEQC 3.0 software [29] to describe the experimental results, among them, the solution compositions.The below described kinetic geochemical modeling methodology is a result of numerous test runs and following tuning of the PHREEQC model / code.The initial solution is pure water and the pressure and temperature is given as in experiments, 1 bar and 25 °C for atmospheric and 100 bar and 80 °C for scCO 2 -experiments.

Gas phases
For all models a gas phase has been defined with either of the above given pressures and temperatures.The composition of the gas phase in atmospheric experiment modeling was specified to be 0.2095 bar O 2 and 0.0003 bar CO 2 based on these gases' partial pressures in air.The unreactive N 2 was not detailed among the constituents.

Mineral phases and their thermodynamic data
For all simulations the PHREEQC.DAT thermodynamic database was used.Thermodynamic data of Na-montmorillonite, however, is not included in this database.Therefore, the phase definition is taken from the LLNL.DAT with a few modifications.
To the original formula (Na 0.33 Mg 0.33 Al 1.67 Si 4 O 10 (OH) 2 ) several other elements had to be added to be able to simulate all of the measured ionic concentrations (Ca-, Mg-, K-, Na-, Fe-, Mn-, Si-and Al).The new definition is based on the structure given for Na-montmorillonite in the SWy-2 clay standard [23], however, it excludes the minor amount of non-measured Ti.Furthermore, SiO (1) Note that the exact equilibrium constant (logK) of the given clay / reaction is not known and the models assume it to remain as in LLNL.DAT even though elemental composition changed.The latter modification (SiO 2 conversion to H 4 SiO 4 ) is possible to take into account in the value of logK however; it does not seem to influence modeling results.
Besides Na-montmorillonite, the carbonate mineral dawsonite was also taken from LLNL.DAT.This is a potentially precipitating mineral in CO 2 geological storage environments [30,31].

Mineral kinetic rates and specific surface areas
Kinetic dissolution and precipitation equations of mineral phases have been defined similarly to earlier works of the authors [11,32], which modeling methodology is primarily based on a USGS report [33] and follows Hellevang et al. [34].
The equations in present study cover OH --and HCO 3-catalyzed mechanisms beside the neutral and H + -promoted reactions, if rate constants listed for 25 °C were available [33].Calculation of rate constants to experimental temperature also follows above works [11,[32][33][34].
Kinetic rate definition code blocks were defined for all minerals in the standard (Na-montmorillonite, quartz, calcite and K-feldspar) and potential secondary minerals precipitating in experiments, such as Ca-montmorillonite, kaolinite, gibbsite, dawsonite, illite, dolomite, chlorite and anhydrite.For Na-montmorillonite the rate constants listed for montmorillonite, for Ca-montmorillonite listed for smectite and for illite listed for muscovite were used.
For all minerals, the largest SSA found in the RES³T database [28] has been used.For K-feldspar the orthoclase and for dawsonite the dolomite data were taken into account.In the definition of rates, SSA is converted into reactive SSA (RSA), which might be only a small portion of the total area.Several RSA scenarios were tested to fit the model on experimental data.Most important input parameters are at Mendeley Data [35].

Mineral composition and water / mineral ratio
PHREEQC models simulating rock-fluid interactions need the mineral composition of the rock as input data given in mol/kgW (mol/kgH 2 O) units.The original mineral composition of the SWy-2 standard (83 % Na-montmorillonite, 7 % quartz, 4 % calcite and 6 % K-feldspar, weight%) is determined based on the XRD measurements of the 8 reference samples.The weight% to mol/kgW calculation has previously been published [32] and it needs the porosity of the system.In the models of this work, the porosity was taken to be 97 % (volume%), which approximates the porosity in the experiments estimated from the 35 water / mineral ratio.

Cation exchange of Na-montmorillonite
The CEC of Na-montmorillonite in SWy-2 clay standard is 76.4 meq / 100 g [23] as described above.This was calculated to be 0.42 in mol/mol units (Section 2.1).However, the stoichiometric equation used in the model has been modified (Eq.( 1)) and in this the M of Na-montmorillonite is 372.865instead of 549.07 g/mol.Like this, the total CEC to be used in the model is calculated to be 0.285 mol/ mol Na-montmorillonite.
Several thermodynamic ion exchange scenarios have been tested to fit the model on experimental data.The division was done primarily among interlayer cations proportionally to their amounts in the clay mineral structure.However, their charge had to be considered (for example Ca values were divided by two).Other scenarios, with including other ions, have also been tested.

Measured "on-the-go" solution composition changes
During batch experiments, solution compositions sensitively changed due to standard-fluid interaction already starting at atmospheric conditions.These runs (SWy2-ATM) show K concentrations comparable to those of scCO 2 experiments (SWy-2-WCO2) and Na concentrations in the same order of magnitude (Table 2).These are both cations in the interlayer space.All other ions (including Ca), though, show an increased reactivity at 80 °C, 100 bar and in the presence of scCO 2 .Therefore, the preparation time on air before the reactor pressurization does not significantly affect these results.

Expected artefacts in the dissolved scCO2-treated sample
At the end of high pressure and temperature rock-fluid experiments thermodynamically favored reactions change due to reactor depressurization and cooling and therefore, minerals may precipitate (or dissolve) by the time of solution sampling at atmospheric conditions.Comparison of solution compositions before and after stopping dissolved scCO 2 -treatments allows identifying which minerals might be experimental artefacts, and not the results of the contact with CO 2 .In this work, solid samples were not informative (see below), but artefacts might be detected in similar experiments of other authors.Two pairs of last "on-the-go" and post-experimental solution samples were available and used for calculating the % change of ion concentrations presented on Fig. 1.Considering the systematically low % change of Al (around -90 %) and Si (around -50 %, Fig. 1) and their comparably high absolute ion concentrations presented in details later, the chemical composition of the experimental artefact will include primarily these two elements.These artifacts could be for example Al-oxides / hydroxides and amorphous silica.Additionally, minor amounts of Mg, Fe and K are expected to mineralize.
These findings partially overlap with previous experiences of the authors [29] reporting similar Al concentration drop and precipitation of boehmite (AlO(OH)) in some of the cooling experimental solution samples.The authors also note that saturation index (SI) changes in an equilibrium PHREEQC model, not detailed in present paper, also support Al-and Fe-oxides / hydroxide and amorphous silica precipitation as artefacts.It also indicates that kaolinite might form.This model uses the composition of an "on-the-go" solution sample at pressure and temperature conditions before and after stopping the reactor (conditions summarized in Table 1).

Mineral composition / structure changes
The analysis of XRD data, since new phases were not detected, focused on 3° to 11° 2Theta angles, which is the corresponding region to montmorillonite.These results are presented on Fig. 2. A systematic shift into two new peak positions can be observed.These analytical results of solid samples (Fig. 2), however, are determined by the post-experimental treatment, i.e., drying time (Table 1) and not the type of the batch experiment, atmospheric or scCO 2 .Samples with short drying time (~ one week) display significant shift from the reference value (6° corresponding to d 001 = 14.9 Å; 7. ATR-FTIR peak positions / widths show similar changes to ATR-FTIR peak areas summarized in Table 3.A continuous change for the reference samples (from SWy-2-0A to 0H) was detected what connects to the winter to summer sample transportation.Otherwise, the same effect of post-experimental treatment (short or long drying time, Table 1) can be only observed as on XRD data (Fig. 2).

Kinetic geochemical model fit
Measured solution compositions (Table 2) provide a sensitive experimental control of the geochemical models of   Na-montmorillonite in air-and scCO 2 -saturated water.
After transforming mg/l data (Table 2) to mol/kgW, the goodness of model fit was possible to test by visualization in R. Several RSA and CEC scenarios have been run, their solution evolution is presented on Fig. 3 in comparison with measured data.The dissolution is always considered stoichiometric despite of the findings of Marty et al. [12], who reported the preferential release of Mg at 25 °C.The interlayer cation (Ca 2+ , Na + and K + , Section 2.1) concentrations in the atmospheric, SWy2-ATM experiments (Table 1) cannot be described solely by the change of RSA in the model and several orders of magnitude different values are predicted (ATM-noExch models, Fig. 3).By considering the clay standard's cation exchange capacity divided proportionally among interlayer cations, the measured variation can be reproduced on an order of magnitude level (ATM-(NaKCa)Exch models, Fig. 3).To gain the measured curve shapes for Na and K, cation exchange should be possibly defined also as a kinetic process, just like mineral dissolution and precipitation, however, suitable parameters are not known.
For the dissolved scCO 2 -environment (SWy2-WCO2), the reproduction of processes gets more complicated.Measured vs. modeled data indicates that not only the interlayer cations (Fig. 3) take part in ion exchange but a minor proportion of other, structural ions as well (WCO2-(NaCaK0.4-4AlMgFeMn)Exchmodels, Fig. 3).This is mostly shown by Fe and Mn (Fig. 3).Taking into account CEC decrease with decreasing pH [24,25], a minor improvement can be reached (WCO2-0.75Exchmodels, Fig. 3).Measured Si concentrations are higher than predicted by WCO2 models (Fig. 3) probably due to its dissolution from the glass vessel in the reactor.

Conclusions
Batch experiments and kinetic geochemical models of this study primarily aimed to simulate the behavior of a selected clay mineral in CO 2 geological storage caprocks.Besides, a few technical conclusions were derived.One of them is that the detection of rock-fluid interactions is more sensitive by analyzing solution compositions than solid samples; XRD and ATR-FTIR data were primarily affected by the drying time of solid samples.Another drawback of solid samples is their potential contamination by experimental artefacts due to reactor depressurization and cooling.It was shown that these artefacts are expected to be mainly Al (e.g.boehmite) and Si containing phases.
Na-montmorillonite (SWy-2 clay standard) starts to react very fast already under atmospheric conditions but its reactivity highly increases when placed into CCS relevant conditions.The measured solution compositions could not be modeled by changing any of the dissolution parameters (e.g.RSA), but by considering the clay standard's cation exchange.With the combined approach, it was possible to simulate measured values on an order of magnitude level.In the atmospheric experiments dividing CEC proportionally among interlayer cations worked well in the models.However, when scCO 2 was present, data indicate that not only the main interlayer cations take part in cation exchange but a minor proportion of other ions as well.To better fit curve shapes probably the kinetic definition of ion exchange would be necessary.
The relevance of this experimentally validated definition of Na-montmorillonite was already shown by the more complex models of Szabó et al. [32].Incorporation of CEC, based on this study, in their reactive transport simulations resulted in a very different and more representative prediction than earlier models.
7° corresponding to d 001 = 11.47Å), whereas those with longer drying time (~ 34 weeks) fall in an intermediate position (7.2° corresponding to d 001 = 12.27 Å) between reference values and those of short drying time.The measured d 001 values for Na-montmorillonite are in agreement with the values reported as a function of integral hydration states, similar to the results of Sendula et al. [9].

Fig. 2
Fig. 2 Effect of post-experimental drying time (one week vs. 3-4 weeks) on XRD results of Namontmorillonite.The data of eight references are presented by their average and 95 % confidence interval.
Namontmorillonite clay standard originates from Newcastle formation, County of Crook, State of Wyoming, USA.Na-montmorillonite is a swelling clay mineral belonging to the Smectite group.The standard's chemical composition and other basic information can be found online [23].Montmorillonite structure is given as (Ca 0.12 Na 0.32 K 0.05 ) [Al 3.01 Fe(III) 0.41 Mn 0.01 Mg 0.54 Ti 0.02 ][Si 7.98 Al 0.02 ]O 20 (OH) 4 , Ca 2+ , Na + and K + are interlayer cations.
2 is represented as H 4 SiO 4 for compatibility with PHREEQC.DAT.The reaction is therefore,

Table 2
Measured "on-the-go" solution compositions in the experiments (Table1).Concertation is given in mg/l and time (t) in hh:mm.

Table 3
ATR-FTIR band areas corresponding to H 2 O, H 2 O, carbonates, silicates and quartz, respectively.From light to dark the values increase.