Field Reaction Experiments of Carbonate Minerals in Spring Waters : Natural Analogue of Geologic CO 2 Storage

To diminish the uncertainty of the mineral trapping rate during geologic CO2 storage, the growth rate of carbonate minerals was measured in CO2-containing spring waters, which can be regarded as a natural analogue of geologic CO2 storage. The authors’ approach, using nanoscale analysis of seed crystal surfaces after immersion into spring waters, enables rapid and accurate measurement of mineral reaction rates. The results show that calcite growth rates in spring waters were lower by 1–3 orders than the values given in a database of laboratory experiment results. We verified the traditional paradigm that Mg controls carbonate reaction kinetics. An increase of theMg/Ca ratio to around 5 by addingMg to spring waters markedly reduced the calcite growth rate. However, even if effects of Mg and flow rate are considered, we were unable to explain satisfactorily the difference of the calcite growth rates between those of spring waters and laboratory experiments. Therefore, other factors might also be related to the slow growth rate in nature. The present results, including the fact such that neither dolomite nor magnesite was formed even at the high Mg/Ca ratio, are expected to provide an important constraint to overestimation of the mineral trapping rate.


Introduction
As a countermeasure against global warming, geologic carbon dioxide (CO 2 ) storage (GCS) has been proposed, whereby the CO 2 captured from a main source such as a power plant is injected artificially into a deep saline aquifer (i.e., [1]).The implementation of GCS demands elucidation not only of the CO 2 migration process, but also of the geochemical interaction in the CO 2 -brine-rock system.Such an interaction proceeds along with the dissolution of CO 2 into formation waters (solubility trapping), the leaching of divalent cations from aquifer minerals into acidified waters, and carbonate mineralization from leached components and dissolved CO 2 (mineral trapping).Of these, mineral trapping is the most ideal trapping form of injected CO 2 from the viewpoint of safety.Nevertheless, it has a very long timescale: several hundred to more than several thousand years.For that reason, its magnitude and rate cannot be evaluated directly.Although we must depend on numerical simulations, such simulations include various assumptions that entail many uncertainties and numerous unknown parameters.Moreover, their output result is not necessarily reliable from the viewpoint of kinetics.
A mineral reaction rate is generally defined (e.g., [2]) as Rate =  exp −/ *  Π where  signifies the rate constant, A denotes the reactive surface area,  is the activation energy,  * is the gas constant, T represents absolute temperature,   stands for the activity of species i, and   is the reaction order for species .In that equation, (Δ  ) denotes the Gibbs free energy change of the system, which is the function of the saturation of the solution.
In addition, Δ  is defined as (2) using activity product  and equilibrium constant  eq as Δ  =  *  ln (   eq ) . ( Regarding (1), past numerical simulation studies have not devoted sufficient attention to     , except for cases in which  is a proton.This term denotes the promotional or inhibitory effects of reactions induced by an impurity.Actually, it is unrealistic to examine the effect of all chemical species dissolved in formation waters on each mineral.For example, however, the magnesium ion Mg 2+ is well known to inhibit the formation of calcium carbonate (CaCO 3 ): the most important CO 2 -fixing mineral [3][4][5][6][7][8][9][10][11][12][13].Considering that formation waters universally contain Mg 2+ , we should therefore incorporate such a fundamental effect into the above rate.
The initial setting of the mineral species also strongly affects the simulation result.Although the starting mineral can be set from the mineral composition of rock samples collected from the target site, the later-forming secondary mineral cannot be determined solely from the thermodynamic stability.In other words, analysis from a kinetic viewpoint requires knowledge of the reaction rate that is applicable to real environmental conditions, even for the stable phase, in addition to setting of the metastable phase as an intermediate product [14].In this context, geochemical simulations of GCS in sandstone aquifer often assume the CO 2 fixation as mineral forms of dolomite (Ca-Mg(CO 3 ) 2 ) or magnesite (MgCO 3 ).However, the production of these carbonates as a primary mineral is extremely restricted in nature.As a result, no knowledge exists of whether they actually precipitate in GCS conditions.
Generally, it has been pointed out that the reaction rate in nature is much lower than that obtained from laboratory experiments because of various factors such as reaction time, surface area, surface conditions (e.g., defects and clay coatings), brine compositions, material transfer, and biologic activities [15,16].The alternate natural analogue study allows one to obtain qualitative knowledge of geochemical reactions over a geologic timescale according to an equilibrium theory, but it fails to estimate the actual reaction rate because the reaction process remains unknown.Instead, if the mineral sample, which is characterized in advance, is exposed to the field and is then recovered after reaction, then analysis of this sample would enable the quantitative estimation of the reaction rate in nature.
This study, conducted with the intention of improving the accuracy of geochemical simulation on GCS, measured both the saturation and impurity effects on the growth rate  of carbonate minerals, and examined the formation conditions of various carbonate minerals.As a new approach to dissipate the discrepancy between laboratory and underground conditions, we used seed crystals at the CO 2 -containing spring to conduct a field reaction experiment, which can be regarded as a natural analogue of GCS.

Reaction System.
The Utoro hot spring in Shari, Hokkaido, was selected as the experiment site.The area geology consists of Neocene rocks such as andesite, rhyolite, and andesitic and rhyolitic tuffs, sandstones, and mudstones, and Quaternary andesitic rocks belonging to the volcanoes of Shiretoko.The spring water is categorized as Na-Cl-HCO 3 type.
We constructed the reaction system: spring water, pumped up from the spring source, was stored in a 60 L plastic tank; then the water was drained through a 40 m long PVC pipe.The drainage rate can be controlled arbitrarily by adjusting the drains attached to each upper and lower stream of the tank.Moreover, the addition of any chemical species into the tank water allows examination of the effects of such impurities specifically beyond the chemical composition of natural spring water.We set three measurement points: the tank interior (point I), the pipe midpoint (point II), and the pipe endpoint (point III).At each point, crystal samples were immersed into spring waters.Figure 1 presents a schematic diagram of the field experiment system.We used carbonate seed crystals of four kinds: calcite, aragonite, dolomite, and magnesite.These single crystals were crushed to make blocks of a few millimeters, which were adhered on a stainless plate by resin with the plane of cleavage oriented upward.As a reference surface, a part of each cleavage plane was pressed and covered by a silicone rubber film.For shorter time experiments (i.e., 1 h), part of the surface was covered with platinum-palladium by sputtering.
The maximum times to immerse crystal samples were, respectively 21, 22, and 23 h for Runs A, B, and C.These reaction times are long enough compared to previous studies of carbonate growth (e.g., [6,8,11,12]).For calcite and aragonite, three or four samples were set at all measurement points.They were taken out one-by-one after 1 and 6 h.Only point III also included the sample after 3 h.The numbers of dolomite and magnesite samples were decreased: two samples at each measurement point.They were taken out only after 6 h.For Runs D and E, only calcite and aragonite were set to points I and III and were taken out after 1 h.For sampling, the moisture adhered onto the crystal surface was removed quickly using compressed air.

Surface Analysis.
We observed the surface of reacted sample using differential interference microscopy (DIM).Then we analyzed its vertical configuration using a phaseshift interferometer (Fabulous; Olympus Co.) or a laser microscope (VK-8510; Keyence Co.) [18].To evaluate the growth rate, the difference in height level between a reference and reacted surface, Δℎ, was measured for the vertical cross-section profile between two points set arbitrarily on the crystal surface.The height of the reacted plane was ascertained from an average along the vertical cross-section profile.
The grown phase was identified using an imaging-plate Xray diffractometer (R-AXIS IV++; Rigaku Corp.), where an imaging plate enables production of the diffraction pattern in a short time, even for small samples.The method is the following: the solid material formed or adhered on the seed crystal was whittled away using a needle, mixed with nail polish, rolled into a sphere of several tens of micrometers, and fixed on a 10 m diameter needle tip.The diffraction pattern was measured by rotating the needle locked on the sample holder.This mineral identification was performed only for samples set at points I and III.

Solution Analysis.
At each measurement point, the temperature (±0.1 ∘ C), pH (±0.01), and EH (±0.01 mS/cm) were monitored continuously and were recorded every 10 s using a data logger.Moreover, the spring water was collected during experiments.In this instance, the in situ measurement of electric conductivity and dissolved CO 2 concentration were done, where the concentration of dissolved CO 2 (±1 mg/l) and HCO 3 (±1 mg/l) was calculated from titration to both pH 4.8 and to pH 8.3.After bringing back the collected water, we applied ICP-AES (inductively coupled plasmaatomic emission spectroscopy) to analyze both cation and silica component, except for ion chromatography for Cl and SO 4 .The analytical precision of these measurements was within 3 to 5%. 2 presents the chemical composition of the spring water obtained at points I-III, along with the degree of saturation with respect to each carbonate mineral, log Ω (Ω = / eq ).The log Ω and CO 2 fugacity ( CO2 ) was calculated using geochemical software: The Geochemist's Workbench (Aqueous Solutions Llc).In this experiment, we expected that the log Ω increased as the water flowed down because the CO 2 originally contained in the water would be degassed during drainage process.In fact, however, point II showed the lowest log Ω, although the highest log Ω was found predictably at point III.

Spring Water Compositions. Table
In Run B, the spring water was not actually equilibrated with CO 2 at 1 bar, and the  CO2 remained to be 0.37 bar.This is because the tank was not enclosed completely and also the residence time of spring waters within a tank was short.
In Table 2, the solution at point II in Run B shows a negative value of log Ω ara , but indeed aragonite was grown as described herein.For this reason, presumably, the solution was supersaturated with respect to aragonite, but an error in solution analysis causes the mistaken log Ω ara value.
It is noteworthy that the Mg 2+ concentration at point I in Run C is comparable to that in Run A despite the addition of MgCl 2 .For this reason, the sampling point within the In contrast, neither dolomite nor magnesite seed crystals grew.At point III, block or needle-like crystals, which became larger over time, were formed on dolomite and magnesite surfaces.The size of these crystals in Run C increased compared to those in Run A.
Table 3 presents results of identification obtained from X-ray diffractometry.The phases grown, respectively, on calcite and aragonite seed crystals were definitely calcite and aragonite.Similarly, we confirmed that the euhedral crystals on dolomite and magnesite seed crystals were also calcite or aragonite.Specifically, additives on dolomite seed crystals at point I were not identifiable because their amounts were too small to detect using X-ray diffractometry.Moreover, the existence of amorphous phase was suggested on magnesite seed crystal at point I in Runs A and B (and partially on aragonite in Run B).However, it is not clear that the amorphous phase in Run A actually derived from mineral reactions because of its very little amount.

Growth Rate.
The  was analyzed for calcite and aragonite because only these two minerals showed growth of their own surfaces.Figures 3 and 4, respectively, portray the time course change of Δℎ of calcite and aragonite at each measurement point in Runs A-C.We measured different crystal samples at each measurement time, which caused data dispersion.Nevertheless, both minerals increased Δℎ with an almost constant rate.The value of Δℎ at point III was the largest at all reaction times, although Δℎ at points I and II did not change systematically.The comparison of each run highlighted that the calcite growth rate  cal at points I and II in Run B and at all points in Run C was decreased markedly compared to the results obtained from Run A. Aragonite in Runs A and B also exhibited a similar trend to those found for calcite.However, Run C showed a different aspect: aragonite growth rate  ara was reduced only at point III.

Saturation State Dependence of 𝑅.
We calculated  by dividing the slope of linear approximation to Δℎ in Figures 3  and 4 by the molar volume.The saturation dependence of  for calcite and aragonite is presented, respectively, in Figure 5.In Run B, R of both minerals was decreased relative to Run A because the CO 2 injection lowered the degree of saturation at each measurement point.Such a reduction of  was noticeable at points I and II, where the initial log Ω was low.Here, the pH change can also affect R, but the pH decrease remained at 0.6, even at point II with the highest pH reduction (Table 2).Therefore, considering the fact that the pH provides no impact to carbonates' R in the neutral range, only the degree of saturation is definitely responsible for  reduction.Regarding the effect of CO 2 injection on R, no distinct difference was found between calcite and aragonite.
In Run C, the enhanced Mg/Ca ratio, that is, ten times that of original waters, caused a significant decrease of  cal at all measurement points, keeping log Ω almost constant (Figure 5(a)).Specifically,  cal at point II was decreased to just 4% of  cal in Run A. This result confirms that Mg 2+ inhibits calcite growth [5,6,8,11,12].For aragonite, an increase of the Mg/Ca ratio provided little impact to  ara (Figure 5(b)).These results are consistent with the findings reported by Berner [5] that the Mg 2+ controls only  cal and that it does not affect  ara .
Here, it is noteworthy that  cal in Runs A and B can be shown on a unique curve, but that  cal in Run C deviates from this curve (Figure 5(a)).In other words, addition of Mg 2+  not only changes  cal but also changes the calcite growth mechanism itself.

Deviation of 𝑅 cal between Nature and Laboratory.
The reaction rate database compiled by U. S. Geological Survey (USGS) was applied to compare the  found in nature and in the laboratory.The database presents the respective reaction rates of typical minerals based on many past references, particularly addressing laboratory experiments [19].Figure 6 presents the comparison between  cal found in this study and  cal calculated by assuming surface-controlled and diffusioncontrolled growth mechanisms.Actually, T and pH were more or less varied depending on the experimental conditions and measurement points.Therefore, the measured  was corrected to the value at 40 ∘ C and pH 7 using the functions of T (activation energy E = 23.5 kJ mol −1 ) and pH dependence from the USGS database [19].As described above, in fact, the pH change effect was negligible because the pH of spring waters corresponds to the neutral range.
Here, it is noteworthy that  cal obtained in this study is significantly lower than the values calculated for both growth mechanisms.The trend of Runs A and B is lower by one order of magnitude with respect to the diffusion-controlled model and by 1-3 orders by magnitude depending on log Ω with respect to the surface-controlled model.Such a discrepancy can be attributed to the difference of the Mg 2+ concentration.The spring waters indeed include 35 mg/L of Mg 2+ (Table 2).However, it is highly unlikely that an increase of the Mg/Ca ratio from 0 to 0.5 reduced  cal more than one order of magnitude [12].Moreover, the difference of  cal is enhanced with increased log Ω (Figure 6), but this trend is contrary to that of Mg 2+ effect in Figure 5(a).
Comparison of Run A with Runs D and E revealed no distinct positive correlation between flow rate and R (Figure 7).In other words, under the present experimental condition, calcium carbonates grow according to the surfacecontrolled mechanism.The flow rate does not affect .Only the calcite at point III seems likely to follow diffusioncontrolled growth because  cal in this point increases slightly with the flow rate.However, even  cal when the flow rate is 36 L/min, which is apparently the almost maximum value, is lower by one order than the calculated result obtained using the diffusion-controlled model.Flow rate (l As a result, the discrepancy of  cal between the values found in nature and in the laboratory cannot be explained solely by the effects of Mg 2+ and the flow rate.Other factors such as organic inhibitory species or biological activities might be related singly or compositely.Quantitative evaluation of each factor is beyond the scope of this work, but the result highlighted that the actual value  cal can be much lower than our expectation.

Formation Phase of Carbonate
Minerals.These experiments revealed that only calcite and aragonite grew in all conditions, and that neither dolomite nor magnesite showed growth of their seed crystals.
Generally, dolomite never forms in laboratory experiments under near-surface conditions (e.g., [20]).Alternatively, a high Mg/Ca ratio, high salinity, and low sulfate ion (SO 4 2− ) concentration are required as factors for dolomite formation [17].In contrast, Sánchez-Román et al. [21] reported that dolomite can precipitate even at surface temperature and irrespective of the presence or absence of SO 4 2− under a condition assisted by microbes.In either case, the present spring water indicates low SO 4 2− concentration.In this study, the spring water was forced to transfer within the dolomite stability field by adding Mg 2+ artificially (Figure 8).Furthermore, seed crystals were used to overcome the barrier for nucleation.Nevertheless, dolomite did not grow, or other carbonates (i.e., calcite and aragonite) were formed on the seed crystal surface.Therefore, even if the conditions for dolomite formation, such as the Mg/Ca ratio, salinity, and SO 4 2− concentration, were obtained, authigenic dolomite never forms.In fact, results show that although some dolomite precipitates from seawater with high salinity under the evaporation circumstances, most dolomite is formed from calcareous sediments by alternation through diagenesis [17].Therefore, our results support this presumption.Similarly, strong hydration energy of Mg 2+ inhibits magnesite formation at low temperature [22].Factors that might overcome this constraint include high temperature and high CO 2 pressure [23][24][25].Actually, authigenic magnesite precipitation was observed at 10 MPa and 90-150 ∘ C in the presence of supercritical CO 2 [26].In addition, Saldi et al. [27] measured the step velocity on magnesite crystal surface using atomic force microscopy under temperature conditions of 80-105 ∘ C despite a lack of high CO 2 pressure.In contrast, below 60 ∘ C, hydrated magnesium carbonate, that is, hydromagnesite ((MgCO 3 ) 4 ⋅Mg(OH) 2 ⋅4H 2 O) or nesquehonite (MgCO 3 ⋅3H 2 O), has been reported to form instead of stability phase of magnesite [28,29].Nevertheless, these metastable phases were not identified in this study.The reason is expected to be examined further, but it is readily apparent that Ca-carbonate is assigned priority over Mg-carbonate at least under low CO 2 pressure conditions.A small amount of amorphous phase identified at point I in Run B can be regarded as the reprecipitation of components leached from dissolving magnesite because the solutions in these cases were undersaturated with respect to magnesite, as shown in Table 2.

Implication to GCS.
The present results can provide an important constraint against overestimation of the mineral trapping rate on geochemical simulations of GCS.In other words, both the magnitude of  cal and its Gibbs free energy change dependence differ between those found in nature and in the laboratory.
Moreover, geochemical simulations often assume the precipitation of dolomite and magnesite.However, formation waters generally correspond to values that are out of the stability field of these carbonates (i.e., the Mg/Ca ratio of formation waters is generally not high).In addition, the formation of authigenic dolomite is at least restricted to special conditions as described above.Therefore, only calcite or aragonite is formed on GCS.Dolomite (and also magnesite in some cases) would not contribute to mineral trapping.Although the accurate judgment of magnesite formation requires further experiment under higher CO 2 pressure condition, we can conclude that the magnesite growth would be inhibited at least under low CO 2 pressure condition, such as shallow under the ground and within a reservoir far from a CO 2 plume.
If dolomite or magnesite is not precipitated, even under high Mg/Ca conditions, then Mg 2+ remains in solution without coprecipitation, which would act as an inhibitor of  cal .Consequently, R of total carbonates depends on the ratio between forming calcite and aragonite because  ara is unaffected by Mg 2+ .Hereinafter, the Mg 2+ effect on the mineral trapping rate, including whether Mg 2+ remains into solution or not, should be considered quantitatively and comprehensively.Misunderstanding of the formation phase of Ca-Mg carbonates is expected to affect the subsequent estimate of carbonates' R.

Conclusions
To diminish the uncertainty of the mineral trapping rate on GCS, R of carbonate minerals was measured at the CO 2containing spring, which can be regarded as a natural analogue of GCS.We immersed seed crystals in flowing spring waters and analyzed nanoscale aspects of their surfaces.The authors' approach enables measurement of the reaction rates of minerals, including minerals that are not present originally, in a short time and with high accuracy.
This study specifically addressed carbonate's reaction kinetics in a Ca-Mg system.For this purpose,  and formation phase of carbonate minerals were evaluated in CO 2 or Mg 2+ added spring water and in pure spring waters.In particular, an application of the USGS database on both  and pH dependencies highlighted the log Ω dependence of  under the GCS conditions, in addition to the Mg 2+ effect on .The addition of CO 2 decreased log Ω, keeping the pH almost constant, which caused reduction of carbonate's .This effect was enhanced in spring water with low log Ω.The present experiment did not engender the unsaturated condition with respect to carbonates, but CO 2 injection at higher pressures also allows the measurement of the dissolution rate.For the Mg 2+ effect, an increase of the Mg/Ca ratio to around 5 decreased  cal considerably.This effect also became noticeable with a decrease of log Ω.  cal decreased to just 4% of the original value under conditions of the lowest log Ω.In contrast, we found no marked impact of Mg 2+ to  ara .This result is consistent with observations by Berner [5] that Mg 2+ inhibits calcite formation and promotes aragonite precipitation, at least from a viewpoint of different Mg 2+ effects between these minerals.
Neither dolomite nor magnesite was formed even at such a high Mg/Ca ratio.In connection with the behavior of Mg 2+ remaining in solutions, this might affect our knowledge related to the Ca-Mg system carbonate kinetics.
However, even if all effects above and flow rate effects are considered, we cannot explain satisfactorily the difference between  cal in spring waters and the values in databases of laboratory experiment results.Therefore, a problem to be examined in future studies is the verification of other factors such as organic inhibitory species and microbial activities, plus the comprehensive impacts of these factors.The present results are expected to provide important constraints on overestimation of the mineral trapping rate, as often seen in geochemical simulations of GCS.
of carbonate minerals is important also from a viewpoint of an operation of hot spring facilities.The data are useful not only for preestimate of the formation rate of hot spring deposits (i.e., scale) within a water pipe, but also for the validation of the potential of CO 2 and Mg 2+ as additives to suppress the scale formation.

Figure 1 :
Figure 1: Schematic diagram of the field experiment system.

Figure 2 :Figure 3 :
Figure 2: DIM images of sample surfaces: (a) after 21 h in Run A, (b) after 22 h in Run B, and (c) after 23 h in Run C. Part of the original surface covered by a silicone rubber film during experiments is denoted as "OS."

Figure 4 :
Figure 4: Time course change of Δℎ of aragonite in each run: (a) Run A, (b) Run B, and (c) Run C. Each line corresponds to a linear approximation passing through the origin.

Figure 5 :
Figure 5: The log Ω dependence of R: (a) calcite and (b) aragonite.Roman numerals near respective dots denote measurement points.Each curve was drawn arbitrarily to clarify the trend of the relation between R and log Ω.

Figure 6 :
Figure 6: Comparison of  cal between spring water and laboratory experiments.

Table 2 :
Chemical compositions of spring waters at each measurement point in three runs (mg/L).

Table 3 :
Formation phase on each seed crystal.was distant from an outlet of MgCl 2 solution.The tank water was not necessarily well mixed.Such mixing would have caused sampling of the water for which the composition was similar to that of original water.The crystal sample was exposed to high Mg 2+ water because it was located near an outlet from the tank.Therefore, we assumed that the Mg 2+ concentration at point I was equal to that at point II for Run C.
Cal, calcite; Ara, aragonite; Am, amorphous phase; DL, below detection limit.tank experiments.Moreover, expansion of the hillock was readily apparent on calcite surfaces.Such a tendency was more noticeable at point III.Results also show that the amount of growth in Runs B and C was less than that in Run A. Specifically, in Run C, the growing part of calcite at point III presented the fine crystalline form.