- © 2014 E. Schweizerbart'sche Verlagsbuchhandlung Science Publishers
In order to understand the groundwater flow at near-underground facilities such as waste repositories, we have studied the effects of flowing an alkaline solution leached from cementitious building materials through the fractures of low-porosity granitic rocks under laboratory conditions. The results show that silica released from the dissolution of sodium-rich plagioclase and quartz reacts with the calcium leached from cementitious buildings to form calcium silicate hydrates (C-S-H) phases in the form of hollow tubular structures. These tubular structures form selectively on the surface of plagioclase in a similar way to reverse silica gardens structures. It was found that the rate of precipitation of C-S-H phases is faster than the rate of dissolution of plagioclase. This self-triggered dissolution/precipitation phenomenon may be an important factor controlling groundwater permeation in natural alkaline underground systems.
- fluid-rock interaction
- flow-through experiment
- C-S-H formation
- alkaline waters
- granite alteration
- chemical garden
The impact of anthropological contribution is becoming increasingly important in the weathering of the Earth’s surface rocks, especially near industrial and urban areas. For example, cementitious buildings such as dams, tunnels and underground repositories for radioactive wastes, made of soluble alkaline materials (Na-K-Ca with silica), are responsible for the formation of highly alkaline ground-waters that may eventually have an impact on the surrounding rocks. Cementing takes place within the pores of ordinary portland cement (OPC) by precipitation of hydration products of portlandite (CH: Ca(OH)2), C-S-H (calcium silicate hydrate) and C-A-S-H (calcium aluminum silicate hydrate) that are produced from ‘alite’ (C3S: 3CaO · 3SiO2), ‘belite’ (C2S: 2CaO · SiO2) and aluminate (C3A: 3CaO · Al2O3) clinkers. Ordinary portland cement can produce highly alkaline solutions for very long periods of time (Atkinson, 1985), mainly due to the dissolution of readily available K and Na sulfates (Lothenbach & Winnefeld, 2006). However, because of the very small porosity in OPC, the strongly alkaline condition is limited to the cement porewater (pH 13.6; Longuet et al., 1973). The flow of highly alkaline solution is estimated to last around 10 kyr when the main leachates are K and Na, and subsequently maintained for approximately another 100 kyr at pH 12.5 due to portlandite dissolution. Therefore, it is most likely that highly alkaline groundwater circulating near cementitious buildings will provoke long-term alteration of the host rock. It appears necessary to understand the kinetics of chemical degradation of natural rocks by aqueous alkaline solutions generated from OPC buildings. Several experimental laboratory experiments have been performed to acquire information on this issue.
Satoh et al. (2007) studied anorthite (CaAl2Si2O8) dissolution rates at pH = 7.0–13.0 and 22 °C for a period of ~1000 hr with the aim of simulating the alkaline degradation of natural rocks. However, no alteration product could be found within this experimental timeframe, demonstrating the need for accelerated experiments at higher temperatures in order to measure the alteration kinetics. Mader et al. (2005) studied alkaline alteration by means of flowing groundwater through the fractures of granitic rocks. In this experiment, the flow was eventually blocked due to precipitation of the alkaline solution inside the fracture during early cement leaching (K-Na-Ca-OH, pH 13.4 at 15 °C). Soler & Mader (2007) investigated this alteration process by chemical modeling and found that the precipitates should be related to the so-called C-S-H (calcium silicate hydrate) phases. The C-S-H phases from fresh OPC are crystallographically ill-defined with typical Ca/Si ratio of 1.7 (Allen et al., 2007) and are the principal binding reaction product of cement hydration. In summary, all these previous studies have demonstrated that alkaline solutions flowing through granitic rock produce secondary mineral precipitation leading to a reduction of groundwater flow that may eventually occlude the fractures of the rocks. However, the type of texture and morphology of precipitates that may play an important role in the evolution of the water flow is yet unknown.
In this work we have carried out a new experimental test by flowing alkaline solution through granitic rocks from the Grimsel Test Site. (Mader et al., 2005), with the aim of obtaining quantitative experimental information about the kinetics of alteration. In the course of the investigation we have identified C-S-H phases forming on the fractures of the rocks by chemical analyses using an electron microscope equipped with X-ray spectroscopy. Noticeably, we have observed that these phases appear as tubular hollow structures of millimeter size in length. This type of morphology may have an important influence on the porosity of the rocks, and therefore on the flow kinetics and hydraulic dynamics. We have identified these remarkable structures as reverse silica gardens, i.e. self-organized structures which until now had only been synthetized in the laboratory (Glaab et al., 2012). This is actually the first proof of the formation of silica garden structures as a result of the alteration process of a natural rock.
We conducted the alteration experiment with alkaline solution at 80 °C using granitic rock from the Grimsel Test Site, Switzerland (Mader et al., 2005). The experiments were performed in a flow-through system consisting of a PTFE-SUS316 cell (23 ml) connected to a HPLC pump with 1/16” (1.5875 mm O.D.) PEEK capillary tubes (Fig. 1). The cell was slightly pressurized to ~0.9 MPa by pumping with a controlled relief valve that prevents the solution from interacting with atmospheric CO2.
The rock sample (labeled ID: NF-2) has a volume of approximately 1.5 cm3 (1.5 × 2.0 × 0.5 cm3). It was cut from a larger core sample collected from the Grimsel Test Site. The rock was mounted in the cell with PTFE spacers, in order to reduce the solution volume to ~12 ml. During the flow-through experiments, the solution enters the inlet capillary at the top of the cell, then it goes down passing over the rock surface and finally goes out through the outlet capillary at the bottom. The volumes of effluent solutions were monitored to test flow rates between 12.8 and 20.0 μl/min. The residence time of the solution is calculated around 600–938 min. The water/rock volume ratio in the reactor was 12/1.5 = 8.
The original granitic rock from which the sample was cut had many natural cracks which were filled with fine-grained materials (fault gouge) during the tectonic history of the geological site. This rock sample has a natural fracture surface covered with fault gouge. The surface of the fracture was covered by plagioclase, K-feldspar, quartz, biotite and an unknown secondary phase, which could be the result of modern weathering. All the rock analyses were made both on this fault gouge material and on a whole piece of rock with the fractured surface. The starting solution of our experiment is a model solution of cement-leachate (Sasaki et al., 2005) and was prepared from He-bubbled pure water and reagents of KOH, NaOH and CaCl2 at concentrations of 0.3 M, 0.22 M, and 0.22 mM, respectively (pH 13.44 at 25 °C). The chemical compositions of the starting granitic rock and minerals are shown in Table 1, while the chemical composition of the starting solution is listed in Table 2. Chemical analyses of the rocks and solutions were performed by X-ray fluorescence (XRF, JEOL JSX-3100RII) using rock standards and by inductively coupled plasma atomic emission spectrometry (ICP-AES, SII SPS3100). Mineral analyses were made on a polished section of the same rock sample using an electron microprobe analyzer (EMPA, JEOL JXA-8200) based on oxide-ZAF correction with the use of silicate and oxide standards. Modal composition of minerals in this granitic rock was evaluated using backscattered-electron imaging over a polished section of fault gouge which is likely to be representative of the composition of the rock sample. This analysis showed feldspar-rich and quartz-poor features as 1.79, 66.57, 22.48, 9.10, and 0.06 vol.% of quartz, plagioclase, K-feldspar, biotite and sphene, respectively.
We observed the surface of the NF-2 rock before and after the experimental runs with a binocular microscope. For further topographic evaluation of the evolution of the rock surface we performed ex situ interferometric measurements with a vertical scanning interferometer (VSI, RSI MM5500) equipped with a Mirau objective lens (20 × , wave mode) and a monochromatic filter (λ = 530 nm). For the purpose of VSI measurements, we made three subsequent runs using the same rock sample. We added crystalline gold particles on the initial rock surface before the first alteration run (Run1), as reference for detecting topographic changes (Van Driessche et al., 2011). After each run, the residual solution was pipetted out of the cell and the rock specimen was carefully transferred into ethanol using plastic tweezers. We repeatedly made the VSI measurements between each run to confirm the continuity of the reaction. The 3-D data thus obtained were batch processed with SPIP-EX software (Satoh et al., 2013) to calculate relative height changes in each area of interest (AOI). Finally, the alteration product that appeared on the rock specimens was recovered manually using titanium tweezers. Then it was fixed on a carbon tape and spatter-coated with Pt for microscopic observation/analysis with a field-emission scanning electron microscope (FESEM-EDS, JEOL JSM-6700F) at an acceleration potential of 15.0 kV. The EDS analysis was made with semi-quantitative oxide-ZAF correction without standards. The estimated error of this correction is less than 1 wt.% for major elements.
Table 2 lists the time evolution of the effluent solution during the flow-through experimental runs. These data, as plotted in Fig. 2, show that the concentration of Si of the effluent solution increases until it reaches a value ~0.5 mM, while the initial concentration of Ca in the model solution slightly decreases from 0.2 to ~0.18 mM. Aluminum increases near ~0.05 mM at the beginning and later steadily decreases to ~0.01 mM. During runs, the flow rates varied depending on fluctuations of the relief valve conditions (12.8–20.0 μl/min for the main reaction period). Several positive spikes of Si during flow could be attributed to flow-rate fluctuations (slower flow). As the flow rates varied, the concentration of Si correlated negatively with Ca. This indicates that during the alteration process of the granitic rock, Si is released slowly by dissolution of silica-rich minerals while Ca supplied by the solution flow is consumed by fast precipitation of a Ca-rich phase.
After each experimental run, the rock specimen was removed from the cell for chemical, mineral and morphological characterization. It was immediately soaked in ethanol to avoid any further chemical alteration. After drying, the surface of the rock was prepared for topography measurements using interferometric microscopy. We noticed different clear signs of the alteration of the initial rock surface. Figure 3 shows topographic features on the surface of plagioclase before and after the runs. The starting plagioclase crystals display larger roughness consisting of many cleavage steps with fine gouge particles. After the experimental run, the initial alteration phase of gouge material grew while the cleaved surface of plagioclase changed to be slightly lowered and roughed as seen from VSI observation (Fig. 3c, roughness parameters, Sa and Sq decreased, Sdq and Spk increased). These observations suggest that the plagioclase grains are dissolved by the alkaline fluid becoming the main source for silica enrichment of the solution. After normalizing the height on the basis of a stable gold crystal initially fixed on the surface, the relative height changes against a standard gold (z = 0 μm), Δz (μm) for either alteration mineral phase or plagioclase were quantified using the equation:(1)
where Δv is relative volume change for each area of interest (AOI). Figure 3c shows the averaged relative height, Δz in representative AOIs (42.6 × 39.0 μm2), demonstrating the trends of precipitation of the newly forming phase and the dissolution on plagioclase. The average normal velocity, Vn of AOI can be calculated using the equation:(2)
where Δt is reaction period. The velocities of the alteration phase and the plagioclase were calculated from linear regressions to be 8.54 ± 2.57E-3 μm/hr (=2.37 ± 0.71E-3 nm/s, SE-based) and −3.21 ± 1.70E-3 μm/hr (=−0.89 ± 0.47E-3 nm/s), respectively (Fig. 3c). This value of the dissolution velocity can be converted to the rate, R as molar flux assuming a molar volume, MV by following equation:(3)
Since the plagioclase composition of this rock is Ab98 (0.98 mole fraction of albite), the dissolution rate of plagioclase was calculated as 8.87E-9 mol/m2/s, assuming albite composition (NaAlSi3O8: MV = 100.47 cm3/mol).
Quartz surface on the rock specimen displayed different topographic features. As shown in Fig. 4, the initial and final surfaces of quartz do not show either significant dissolution or precipitation signals. Selected surfaces of quartz (AOI1 and 2) showed dissolution velocities of Vn = −4.90 ± 1.61E-4 and −8.14 ± 5.87E-4 μm/hr (−1.36 ± 0.45E-4 and −2.26 ± 1.63E-4 nm/s), of around one order of magnitude slower than that of plagioclase. Their dissolution rates can be converted to be R = −5.95 ± 1.95E-9 and −9.88 ± 7.13E-9 mol/m2/s assuming a molar volume of quartz (SiO2: 22.88 cm3/mol).
Before the final VSI measurements, the surfaces of the granitic rock were observed by classical optical microscopy at the end of Run3. Unexpectedly, we found tubular hollow structures. As these tubular structures were not observed during the VSI measurements after Run1 and Run2, they must have precipitated during Run3. The tubes have the end either open (Fig. 5a, b and d) or closed (Fig. 5c). The length of the tubes varies from 0.5 to 6 mm and the diameter from 0.1 to 0.5 mm. For the longer type of tubes, the thickness of the wall decreases from bottom to top (Fig. 5a and b), whereas the shorter tubes are translucent and seem to have stable wall thickness. We also found a rather thin film covering the rock mineral grains within the rock in physical contact with these tubes (see Fig 5a). Figure 6 shows the biggest tube (~0.5 mm diameter, ~6 mm long), which grew on one side of the sample (white arrow in Fig. 6b).
One of these tubes was collected from the sample for further FESEM observation (Fig. 7) to characterize its composition. The surface of the tube exhibits many cracks caused by drying, which indicates that the material was made of a hydrated phase. The inner and outer surfaces of the tube show clear differences in texture. The outer surface shows a relatively fine and dense surface with numerous pores smaller than 1 μm in diameter, whilst the inner surface has a slightly rough texture similar to as-grown natural smectite. When the cross-section of the tube is observed, it is clear that the inner and outer surface are connected by pores that allow the transport of fluid across the wall. We have performed a chemical analysis on the sampled tube cut in half (Fig. 8). In the interior of the cross-section of tube wall (Fig. 8b), EDS data showed a Si-rich composition with a C/S (Ca/Si) = 0.52. In contrast, the outer part of the tube showed slightly Ca-rich composition with C/S = 0.74. As shown in FESEM images (Fig. 8a), the inner wall is covered with many fine particles of submicrometric sizes and rich in Ca.
The experimental results shown in Fig. 2 and Table 2 demonstrate that the silica released by dissolution of the granitic rock at high pH reacts with calcium to form the alteration product. The characterization studies of the altered rock’s surface demonstrate that the dissolution of plagioclase (8.87E-9 mol/m2/s) and quartz (5.95–9.88E-9 mol/m2/s) are the main contributors to the release of silica. Based on the modal proportions of plagioclase (66.57 vol.%) and quartz (1.79 vol.%) in the rock sample, the release rate of Si from both minerals were calculated as 1.74E-8 and 1.07–1.77E-10 mol/m2/s, respectively. Thus, the contribution of plagioclase to the supply of Si is two orders of magnitude higher than quartz. Note that in addition to Si, plagioclase also supplies a small amount of Ca (Ca/Si = 0.02/2.94 = 0.007). Therefore, it is clear that when the starting model solution interacts with the granitic rock, silica is dissolved and converted into silicate at pH higher than 12. Then, the silicate reacts with the calcium-rich waters provoking the precipitation of the highly insoluble calcium silicates, plausibly in the form of hydrated phases, as previously suggested by Mader et al. (2005). Our experiments allow the identification and characterization of the precipitated phases as thin layers of transparent gel-like material. We have also found the formation of several bizarre tubular structures, millimeters in length, which we have identified as silica gardens, i.e. self-assembled structures forming when soluble metal salts are brought into contact with alkaline solutions of silicate. Upon reaction of the metal and the silicate, a metal silicate hydrate colloidal membrane forms around the dissolving crystals of the soluble salt, triggering a morphogenetic mechanism based on osmosis, buoyancy and chemical reaction which creates peculiar tubular plant-like structures (e.g. Coatman et al., 1980; Cartwright et al., 2002; Kellermeier et al., 2013). This explanation is also based on the typical textural gradient we have measured across the wall of the tubes, a pattern that has also been found in silica gardens as reported by Glaab et al. (2012); namely, a dense and Si-rich layer at the exterior wall and a metal-rich, porous thin layer in the inner part of the tube. While in classical silica garden laboratory experiments the tubes are grown directly from metal chloride salts in Na2SiO3 solution, in our experiments these tubular structures are formed in a way similar to the method described by Pagano et al. (2007). These authors successfully created tubular structures by injection of a metal salt into an alkaline silica-rich solution. Although the growth direction of the classical silica garden is controlled by fast buoyancy from the dissolving metals, our observed tubes show irregular or a little downward orientation (Fig. 5 and 6). This is because the amount of Ca transported from Ca-poor plagioclase is limited by its much slower dissolution rate in comparison to the salts. Consequently, we suggest that our observed CS-H tubes have formed as a result of a reverse chemical garden process, a process already proposed by Coatman et al. (1980) to explain the formation of Portland cement. As shown in Fig. 8e, the inverse C/S gradient across the tube wall compared to the classical chemical gardens is consistent with the products of reverse chemical gardens (Pratama et al., 2011).
Our explanation is schematically shown in Fig. 9. The calcium-rich alkaline solution simulating OPC-leachates (enriched in K and Na; Atkinson, 1985) reaches the granitic rock and dissolves silica. As described above, we found that mineral alteration occurs mainly on plagioclase crystals. The dissolution of plagioclase provokes not only the release of silica (~3.0 atoms per plagioclase formula unit: apfu) but also a little amount of calcium (only 0.02 apfu). This provokes a local increase of silica concentration that triggers the formation of C-S-H films and tubes. It can be inferred that even a little release of Ca can contribute to the local saturation of C-S-H near plagioclase. The preferential appearance of C-SH on plagioclase rather than on quartz and the other minerals supports this hypothesis. According to Novella (2000), precipitation takes place in the form of films at low silica concentrations, whereas tubes are the predominant morphology at high silica concentration. This may suggest that slower dissolving minerals such as quartz can form C-S-H films and prevent contact with the high-pH solution, whilst faster dissolving plagioclase can form films and sometimes develop from films to tubes by releasing more Si and Ca. However, it is not clear whether a small release of Ca may affect the local saturation of C-S-H near plagioclase. Therefore, we carried out three further experimental tests. One was performed on the surface of the natural fracture and the other two were performed on the polished surface at the same chemical condition. We confirmed the formation of tubes (about ten tubes per square centimeter of rock) on both the natural and the polished surfaces.
The chemical analysis of the tubes shows an average C/S ratio of 0.67 with variations between 0.52–0.74. This C/S range corresponds to gyrolite (NaCa16Si23AlO60(OH)8 -16H2O; C/S = 0.70; Merlino, 1988) or okenite (Ca3Si6O15-6H2O; C/S = 0.50, Merlino, 1983). Only the inner wall of the tube showed a very high value of C/S due to co-precipitation of CH (portlandite). Thus, the mineral phases forming the tube can be identified as low-C/S members of C-S-H, which is partly associated with CH.
Previously, Mader et al. (2005) investigated hydraulic behavior during infiltration of a hyperalkaline solution (0.13 M KOH + 0.07 M NaOH + 2.00 mM Ca(OH)2) into a fracture rock of Grimsel granite using an infiltration apparatus under constant differential pressure. In their experiment, it was found that after 9 months of flowing into the rock fracture, there was a gradual decrease of the flow rate by a factor of 25 (0.47 to 0.018 g/h, Mader et al. 2005) as a result of interaction between rock andcement leachate. Thus, it is known that C-S-H precipitation may play an important role in the flow of fluid through the fractures of rocks. Because of the larger molar volume of C-S-H with respect to the initial minerals, the precipitation of C-S-H during alkaline alteration may significantly reduce the flow rate and finally clog the fracture. Table 3 lists a number of possible secondary Ca-phases and their molar volumes. Since these C-S-H candidates have a variety of chemical formulae, the molar volume (MV) per Si atom in the molecules (MVSi = MV/Si) was calculated for comparison purposes. The higher C/S ratio shows the larger MVsi. The largest MVSi = 102.24 cm3/mol corresponds to C-S-H with C/S = 1.75 (which is very common in ordinary portlandite cement), which is nearly double of that for anorthite plagioclase (50.38 cm3/mol). The ability of the C-S-H to clog fractures is related to its texture and therefore to its precipitation conditions. If the C-S-H precipitates as tubular structures, the precipitation rate should increase significantly due to a larger apparent volume, so that the rock fractures can be expected to clog quickly. However, these C-S-H tubes may be considered as macropores. This, along with the nanoporosity of the walls of the tubes, is the reason why ionic mobility can be detected after clogging the fractures. According to Glaab et al. (2012), the tube wall of silica gardens behaves like a permeable membrane. As a result, clogging can occur in the granitic fracture, but perfect clogging can be hardly achieved because the solution flow may still continue through macro and nanopores of the altered materials, as reported in Mader et al. (2005). This is in agreement with our experimental results. The observed rate of C-S-H precipitation obtained from the slope of Fig. 3 was 2.37E-3 nm/s as dry state. Since C-S-H can increase in volume when it is wet, the observed rate should be actually slower. The actual precipitation rate of C-S-H in solution may be twice as fast as the observed rate under dry conditions due to its large molar volume (Table 3). Thus, our experimental results demonstrate that the precipitation velocity is two orders of magnitude faster than the dissolution velocity of plagioclase. If the difference in molar volumes between wet and dry C-S-H can be assumed to be a factor of 1.51 (Brunauer & Greenberg, 1960; Taylor, 1997), the VSI measured precipitation velocity in dry conditions can be re-calculated as 3.58E-3 nm/s. It is likely that this is the maximum velocity that could be achieved as C-S-H in film form in solution.
Chemically, C-S-H precipitation can be evaluated as flux. As long as C-S-H is precipitating, the concentration of Ca in the solution is lowered toward a C-S-H equilibrium. We carried out a preliminary calculation for the final composition of effluent using the PHREEQC program (Parkhurst & Appelo, 1999). Our calculation shows a moderate saturation index (= log Q/K, where Q is the ionic activity product and K is the solubility product) of 0.42 with respect to tobermorite-14Å (C/S = 0.83), which supports the possibility of the precipitation. If we consider a difference of Ca concentration between initial and final solutions (ΔCCa = 0.036 mM), a total solution volume flowing in the reaction cell (V = 763.41 mL) and a duration (t = 2653560 s), then the total precipitation rate, R of C-S-H as tobermorite (stoichiometric factor, NCa = 5) on rectangular-simplified rock surface (S = 0.00095 m2 for volume of 1.5 × 2.0 × 0.5 cm3) can be simply calculated as a steady-state rate using the following equation:(4)
The calculated rate, R = 2.20E-9 mol/m2/s can be converted again into a normal velocity of tobermorite-14Å as Vn = 7.23E-4 nm/s (Table 4). This rate estimated from solution chemistry agrees with VSI measurements for the alteration phase in the order approximately 1E-8 mol/m2/s. A discrepancy between the VSI observed (2.37E-3 nm/s) and this average velocity (7.23E-4 nm/s) may reflect a heterogeneous distribution of the precipitation. This fact confirms that C-S-H precipitates locally and may favor various forms with different densities, from film to tube with different aggregations.
Consequently, it is likely that the dissolution of plagioclase derived from the interaction of the cement leachate fluid with the granitic rock will contribute to enrichment in Si (1.74E-8 mol/m2/s for whole fracture surface) and, in turn, will provoke the precipitation on the surface of the fractures thus decreasing the flow of underground water.
We have demonstrated that the alteration of granitic rocks by model underground waters leached from cementitious building structures provokes the precipitation of C-S-H phases. These phases form tubular structures which consist of permeable mineral membranes chemically and texturally identical to reverse silica gardens. The formation of these tubular structures occurs selectively on calcium-bearing plagioclase after dissolution of quartz and sodium-rich plagioclase that increases the local concentration in silica and calcium. Mineral dissolution and C-S-H precipitation on the surface of the granitic rock was followed quantitatively by interferometry. It was found that the rate of precipitation of C-S-H phases is faster than the rate of dissolution of plagioclase. et al. Our results also demonstrate the plausibility of the formation of chemicals gardens phenomenon within geochemically environments.
Experimental support by A. Ozawa and Y. Odashima of Mitsubishi Materials Techno Co. for reaction runs is gratefully acknowledged. Financial support from the Spanish Consolider-Ingenio 2010 project “Factoría de Cristalización” CSD2006-00015 and the MICINN project CGL2010-16882 (J.M.G.R.) is greatly appreciated. This study was financially supported by JAEA, Japan, but does not represent JAEA’s opinion.
- Received 15 May 2013.
- Modified version received 17 December 2013.
- Accepted 17 January 2014.