Element scavenging by recently formed travertine deposits in the alkaline springs from the Oman Semail Ophiolite

Ultramafic rocks, such as the Semail Ophiolite in the Sultanate of Oman, are considered to be a potential storage site for CO2. This type of rock is rich in divalent cations that can react with dissolved CO2 and form carbonate minerals, which remain stable over geological periods of time. Dissolution of the ophiolite mobilizes heavy metals, which can threaten the safety of surface and groundwater supplies but secondary phases, such as iron oxides, clays and carbonate minerals, can take up significant quantities of trace elements both in their structure and adsorbed on their surfaces. Hyperalkaline spring waters issuing from the Semail Ophiolites can have pH as high as 12. This water absorbs CO2 from air, forming carbonate mineral precipitates either as thin crusts on the surface of placid water pools or bottom precipitates in turbulent waters. We investigated the composition of the spring water and the precipitates to determine the extent of trace element uptake. We collected water and travertine samples from two alkaline springs of the Semail Ophiolite. Twenty seven elements were detected in the spring waters. The bulk of the precipitate was CaCO3 in aragonite, as needles, and rhombohedral calcite crystals. Traces of dypingite (Mg5(CO3)4(OH)2·5H2O) and antigorite ((Mg,Fe)3Si2O5(OH)4) were also detected. The bulk precipitate contained rare earth elements and toxic metals, such as As, Ba, Cd, Sr and Pb, which indicated scavenging by the carbonate minerals. Boron and mercury were detected in the spring water but not in the carbonate phases. The results provide confidence that many of the toxic metals released by ophiolite dissolution in an engineered CO2 injection project would be taken up by secondary phases, minimizing risk to water quality.


Introduction
CAPTURING and storing CO 2 in rock as carbonate minerals, formed during injection of CO 2 into geological formations, has great potential for reducing atmospheric concentrations of anthropogenic CO 2 . Some carbonate minerals form readily in nature but by increasing our understanding, the processes could be enhanced and used on an industrial scale. Improved under-standing of carbonate mineral formation would also contribute towards more robust models for predicting the global carbon balance (Mackenzie and Andersson, 2013). Field-scale projects, where CO 2 -charged waters are injected into the subsurface, have already begun, e.g. into basaltic formations in Iceland and northwestern USA (Alfredsson et al., 2008;Oelkers et al., 2008;Gislason et al., 2010;McGrail et al., 2011;Alfredsson et al., 2013;Gislason and Oelkers, 2014). Carbon dioxide-enriched waters are corrosive (at 1 atm CO 2 , pH = 3.6) and when injected into geological formations, they attack the host rock, increasing weathering rates. Mafic and ultramafic rock releases ions such as Ca, Mg and Fe, which can combine with the dissolved CO 2 and form stable carbonate phases. The overall reaction is energetically favourable (Lackner et al., 1995) but depends on fluid composition, ionic strength, reduction potential, pH, temperature and total pressure (Stumm, 1992;Krauskopf and Bird, 1994;Gislason et al., 2010;Olsson et al., 2012). Mafic and ultramafic rocks are well suited to CO 2 storage because of their large divalent cation:Si ratio, and thus greater capacity for carbon capture, and lower tendency to form clay minerals, which use up cations, thus decreasing carbonation potential. Reaction in these rock formations neutralizes the acidic waters, thus promoting carbonate mineral precipitation. However, injecting CO 2 -charged waters, and the ensuing dissolution of the host rock, can mobilize toxic heavy metals, such as As, Ba, Cd, Cr, F, Hg and Pb (Turekian and Wedepohl, 1961), which could pose a risk for surface and groundwater supplies (Aiuppa et al., 2000;Flaathen et al., 2009;Galeczka et al., 2013;Olsson et al., 2013;Galeczka et al., 2014;Olsson et al., 2014).
The Semail Ophiolite of the Sultanate of Oman is a potential site for CO 2 -charged water injection for carbon sequestration (Kelemen and Matter, 2008;Matter and Kelemen, 2009;Paukert et al., 2012). It is one of the largest and best preserved ophiolites in the world and covers~30,000 km 2 , with an average thickness of~5 km (Neal and Stanger, 1984b;Nicolas et al., 2000;Chavagnac et al., 2013a). About 30% of this volume is the ultramafic rock, peridotite, which contains mostly olivine and pyroxene and <45% silica (Nicolas et al., 2000;Kelemen and Matter, 2008 (1) releasing cations and increasing pH. At low partial CO 2 pressure, intermediate products associated with the incomplete dissolution of peridotite are typically serpentine (Mg,Fe) 3 Si 2 O 5 (OH) 4 ) and talc (Mg 3 Si 4 O 10 (OH) 2 ). Spring waters discharging from ultramafic rocks are extremely alkaline, with a pH as high as 12 and CO 2 partial pressure <10 À8 bars (Barnes et al., 1978;Neal and Stanger, 1984a;Paukert et al., 2012;Chavagnac et al., 2013b). As the hyperalkaline water emerges at the Earth's surface, it equilibrates with atmospheric CO 2 , carbonate minerals become supersaturated at the airÀwater interface and precipitate ( Fig. 1) FIG. 1. Precipitation of carbonate minerals at the locations where Sample OM01 (a) and Sample OM08 (b) were collected. In calm hyperalkaline waters, the precipitation rate is controlled by the diffusion of CO 2 from the atmosphere through the airÀwater interface. Carbonate minerals form as a thin crust that looks like ice (a). In fastmoving spring waters (white waters), where the airÀwater surface film is broken, the nucleation and precipitation is homogenous within the water body (b).
As protons are released, pH decreases. The minerals observed are calcite and aragonite (both CaCO 3 ), as well as brucite (Mg(OH) 2 ) and antigorite ((Mg,Fe) 3 Si 2 O 5 (OH) 4 ) (Neal and Stanger, 1984a;Chavagnac et al., 2013a). Spring waters in natural systems can be considered as an analogue for CO 2 injection into ophiolites in general (Kelemen and Matter, 2008;Beinlich et al., 2010Beinlich et al., , 2012Paukert et al., 2012) and by studying the uptake of trace metals into the secondary phases, we can gain information about their mobility as a result of CO 2 injection. The aim of this study was to analyse ultramafic spring waters and the natural precipitated products to investigate the extent of trace element uptake into the solids and loss from the solution phase.

Sampling and analyses
The samples were collected on 13 and 14 January 2011, a total of eight liquid and eight solid samples from two alkaline springs in Oman. The springs were located near the village of Falaji (Spring #1) and Qafifah (Spring #2) discharging from partly serpentinized peridotite and gabbroic rocks. Samples from the two springs and the surroundings have been described elsewhere (Nicolas et al., 2000;Kelemen and Matter, 2008;Matter and Kelemen, 2009;Paukert et al., 2012;Chavagnac et al., 2013a,b). pH and alkalinity were measured in the field. The pH was measured using a plastic body, double junction electrode, connected to a CyberScan 310 pH meter, calibrated against certified buffers (pH 4.01, 7.00, 10.01 and 12.46) from Oakton. Electrode and buffers were at the same temperature as the spring water. Alkalinity was determined by Gran titration, using HCl (Gran, 1952;Stumm and Morgan, 1981). Samples for ion chromatography (IC) and inductively coupled plasma optical emission spectrometry (ICP-OES) were filtered in the field through 0.2 mm cellulose acetate membranes. They were collected in plastic vials that had been flushed three times with the sample. Samples for ICP-OES were acidified with concentrated, suprapure HNO 3 (1:100) onsite to prevent precipitation of solids during transport. An overview of the fluid samples and locations is found in Table 1.
Solid samples were collected in plastic bags and freeze dried for 24 h in the laboratory to minimize alteration. The solids were characterized by means of standard X-ray powder diffraction (XRD) and scanning electron microscopy (SEM), using instrument conditions and procedures described by Olsson et al. (2014). Residual water in the pores could not be removed by freeze drying and is responsible for precipitation of halite (NaCl) observed in the XRD patterns. The solid samples were flash dissolved in a 1% HNO 3 solution which dissolves the carbonate mineral phases while minimizing the dissolution of possible silicate mineral grains originating from the ophiolite. Selected spring water and flash-dissolved samples were analysed at Analytica-SGAB, Luleå, Sweden, using ICP-OES and inductively coupled plasma sector field mass spectroscopy (ICP-SFMS). Mercury was measured with atomic fluorescence spectroscopy (AFS). The elements measured by IC were Cl À , F À and SO 4 2À and by ICP-OES were Al, As, B, Ba, Br, Ca, Cl, Cr, Fe, K, Li, Na, Mg, Mn, Mo, P, S, Sb, Si, Sr, Ti, V and W. In-house, multielement standards were used for calibration and these are regularly tested against the Standard Canadian River Waters (SLRS-4) and single element international standards (SPEX). The uncertainties of the IC and ICP-OES data were estimated from analytical measurements. The reproducibility of duplicate samples for both methods is within 5%.
The geochemical speciation program PHREEQC (Parkhurst and Appelo, 1999), with the MINTEQ database (Allison et al., 1991), was used to determine the charge balance of the spring water samples, to estimate the dissolved inorganic carbon (DIC) and the saturation state of selected solid phases. The saturation state of lansfordite (MgCO 3 ·5H 2 O) was determined in a parallel PHREEQC operation run using the LLNL database (Wolery, 1992). The charge balance of all of the spring water samples was within 2% and ionic strength was <20 mM.

Spring-water chemistry
Five water samples were collected from Spring #1 and its runoff and three from Spring #2 and its runoff, at a series of sites downstream from the springs. At Spring #1, the pH was 11.6 and total alkalinity was 5.9 meq/l and these values remained constant downstream (Table 1). A summary of the dissolved components can be found in Tables 2 and 3. With the exception of the DIC, Mg, Ni and Zn, concentrations remained constant with distance downstream for Spring #1 (Fig. 2). The DIC increased as expected because of the influx of CO 2 from the atmosphere. Mg concentration was very low,~6.3 mM, probably because it was controlled by the formation of serpentine, brucite or hydrotalcite (Neal and Stanger, 1984b;Paukert et al., 2012;Chavagnac et al., 2013b). The water in the pool near this spring was covered by a thin crust of solid CaCO 3 , which inhibited influx of atmospheric CO 2 . The constant composition of the spring water suggests that relatively little of the dissolved elements formed precipitates. The reasons are: (1) the short period during which precipitates could form during the time the water emerged from the spring outlet to 20 to 30 m downstream, where the last sample was collected; and (2) carbonate minerals could not precipitate because the concentration of dissolved carbonate was hindered by the thin crust of CaCO 3 at the water surface (Fig. 1a). The influx of CO 2 depends on water turbulence, i.e. dissolution is faster when the gas forms foam and bubbles during mixing, such as at waterfalls, breaking waves or rapid transport (Jähne et al., 1987;Gislason et al., 1996). The water of Spring #1 experienced only slight turbulence, therefore slow CO 2 influx. At Spring #2, pH decreased significantly from 11.5 to 10.9 downstream and alkalinity, from 3.2 to 1.9 meq/l (Table 1). Ca, K, Mg, Si, SO 4 and Sr decreased slightly downstream; Ba, Br, Cl and Na increased slightly. In the last sample, F increased by a factor 5 (Fig. 3). The complicated changes in the chemical composition suggest multiple dissolution/precipitation reactions and possible input from a second spring. The DIC increased from 0.22 mM to 0.83 mM over the first 15 m but decreased again to 0.65 mM at 30 m. Precipitation of carbonate minerals consumes DIC. It is likely that magnesium carbonates formed because the Mg concentration in Spring #2 decreased with distance downstream and was more than 2 orders higher than the Mg concentration in Spring #1. Spring #2 water fell over a series of short drops so the turbulence increased the CO 2 influx much more than at the calm waterÀair interface at site OM01 (Fig. 1). CO 2 uptake decreased the pH. Rapid influx promoted fast carbonate mineral precipitation, where the rate was limited by nucleation and growth within the water body rather than by CO 2 absorption only. Alternatively, the presence of a second spring could discharge water with lower pH and higher DIC into the stream bed, which would promote precipitation.
The water from both Springs #1 and #2 contains several elements which, at high concentration, pose a threat to water quality (Table 2). However, the concentrations of all of these elements are under the WHO threshold limits for clean drinking water (WHO, 2008). Many of the potentially toxic elements in these waters are likely to be scavenged by the precipitating travertine phases (Olsson et al., 2014).

The travertine deposits
The travertine deposits in the alkaline springs were analysed using XRD and SEM/EDXS to identify the mineral composition and the morphology of the samples collected. The precipitates of the springs consist mainly of aragonite and calcite. The XRD patterns of the samples from Spring #1 also showed traces of dypingite (Mg 5 (CO 3 ) 4 (OH) 2 ·5H 2 O), antigorite ((Mg,Fe) 3 Si 2 O 5 (OH) 4 ) and halite (NaCl). We were not able to detect amorphous material. With the exception of halite, these precipitates are typical of the hyperalkaline springs of the Oman ophiolites (Neal and Stanger, 1984a;Chavagnac et al., 2013a). The presence of halite can be explained by evaporation of residual water when the samples were freeze dried. The SEM images of the thinsurface precipitate show rhombohedral calcite in a network of aragonite crystals (Figs 1a, 4a), whereas the precipitate from the bottom of Spring #1 consisted of dypingite rosettes with bouquets of aragonite (Fig. 4b).
The XRD patterns showed that the travertine samples from Spring #2 were predominantly of aragonite. The SEM images showed small networks of rosettes growing on aragonite needles (Fig. 4c,d). The rosettes were rich in magnesium, indicating either dypingite or brucite (Power et al., 2007;Chavagnac et al., 2013a). The chemical composition of the carbonate minerals in the travertine samples was quantified by flash dissolving the samples and analysing the chemical components released to solution. The average composition for travertine from both springs was determined and normalized to one mole of dissolved CaCO 3 , assuming all Ca ions originated from pure CaCO 3 (Table 2 and 3). The travertine contains a wide range of elements, including rare earth elements (REE) and the toxic elements: As, Ba, Cd, Cr, Cu, Mn, Mo, Ni and Pb. Some elements were below detection limits in the spring water but were found in the travertine which indicates that these trace elements have, to a significant extent, been taken up by the precipitates. Elements observed in the spring waters were also found in the travertine, except for B and Hg. Thus, most toxic elements released from an engineered CO 2 injection project would be sequestered by secondary minerals but some, namely B and Hg, would not.

Mineral saturation estimated using PHREEQC
Numerous carbonate minerals can form in the alkaline spring waters of Oman. We investigated which phases are supersaturated and evaluated if they would be likely to precipitate while the spring water equilibrated with atmospheric CO 2 . Knowing which phases are present in the travertine is important because trace element coprecipitation and adsorption are controlled by the composition and structure of the sorbing phase. We used the PHREEQC geochemical speciation model to determine the saturation states. In the model, CO 2 was added to the spring water in small steps and the saturation states were calculated until equilibrium with atmospheric CO 2 was reached (SI = À3.4). No minerals were allowed to precipitate. For the model, we used data from Sample OM01 (water from Spring #1) and OM06 (from Spring #2), including temperature, chemical composition, pH and alkalinity. The saturation index (SI) maximum at pH 10.6, for the pure carbonate phases, corresponds to the point at which the CO 3 2À concentration peaks. At lower pH, the HCO 3 À ion dominates. PHREEQC predicted that the supersaturated phases in Spring #1 would be calcite (hexagonal CaCO 3 ), aragonite (orthorhombic CaCO 3 ) and dolomite (CaMg(CO 3 ) 2 ) and in Spring #2, artinite (Mg 2 (CO 3 )(OH) 2 ·3H 2 O), dolomite, huntite (Mg 3 Ca(CO 3 ) 4 ), magnesite (MgCO 3 ), hydromagnesite (Mg 5 (CO 3 ) 4 (OH) 2 ·4H 2 O), calcite and aragonite (Fig. 5). Lansfordite (MgCO 3 ·5H 2 O, not shown) and nesquehonite (MgCO 3 ·3H 2 O) were predicted to be undersaturated. Inorganic synthesis of dolomite, magnesite and huntite at atmospheric pressure and temperature below 40ºC has never been reported (Deelman, 1999;Saldi et al., 2009;Deelman, 2011;dos Anjos et al., 2011) so they are not expected to precipitate in the Oman spring waters. Hydrous magnesium carbonate phases, such as artinite, hydromagnesite and dypingite (Mg 5 (CO 3 ) 4 (OH) 2 ·5H 2 O), are far more likely to form (Hsu, 1967;Botha and Strydom, 2001;Hopkinson et al., 2008;Hänchen et al., 2008). No thermodynamic data are available for predicting the saturation state of dypingite but precipitation of both dypingite and hydromagnesite can be mediated by microorganisms (Ming and Franklin, 1985;Braithwaite and Zedef, 1996;Power et al., 2007;Cheng and Li, 2009;Power et al., 2009). The hydrous magnesium carbonate minerals are relatively stable and form more readily than magnesite at temperatures of <80ºC and they have therefore been suggested as alternative reaction products for carbonate storage (Ballirano et al., 2013). The magnesium concentration in the outlet spring waters is very ELEMENT SCAVENGING BY RECENTLY FORMED TRAVERTINE low as a result of serpentinization. In an engineered CO 2 injection scenario, serpentinization is expected to be minimal and therefore the relative concentration of Mg would be much higher, i.e.~50 times higher than the Ca concentration (Boudier and Coleman, 1981). In this scenario, magnesium phases would probably dominate the secondary phases and these could complement the scavenging of contaminants above that observed in this study. More research is needed to confirm this assessment.

Conclusions and outlook
The spring waters from the Semail Ophiolite are highly alkaline, with pH between 11.6 and 10.9. The most prominent cations are Ca, K, Na, Mg and Si; anions are Br, Cl, F and SO 4 . Within uncertainty, the concentrations remain constant downstream for Spring #1 but complicated changes in the chemical composition for Spring #2 suggest dissolution and precipitation reactions and probably input from a second spring. Travertine samples from the alkaline springs consist of aragonite often mixed with calcite and traces of dypingite and antigorite. Other phases that were supersaturated in the spring waters, and predicted to precipitate at atmospheric conditions, were artinite and hydromagnesite. The most common alkali, alkali earth and transition metals were detected in the solid travertine samples along with REE and toxic elements such as As, Ba, Cd, Cr, Cu, Mn, Mo, Ni and Pb. This suggests active and effective scavenging by precipitation of the carbonate minerals, thus limited mobility of contaminants. However, boron and mercury were not removed from the water. If they were present at increased concentrations they would pose a health risk.
Our study shows that several carbonate minerals precipitate readily in groundwater from the Semail Ophiolite when CO 2 dissolves in it. Aragonite and calcite dominate the precipitates and these phases are well known to incorporate trace elements. Thus we can conclude that toxic metals that are released during dissolution of the ophiolite are taken up by the travertine. From this we can draw analogies with likely scenarios for geological CO 2 sequestration. Although the reaction conditions in an engineered CO 2 injection system would be different from the conditions in our study, in both cases, Ca, Mg, Fe, CO 2 and trace metals would be present and where carbonate minerals such as calcite reach supersaturation and precipitate, we would expect the precipitates to scavenge the trace elements. Thus we can conclude that if toxic trace metals are released by dissolution of mafic or ultramafic rocks during CO 2 injection, they would probably be sequestered by the carbonate minerals or other secondary phases (such as Fe oxides and clays) that form as the waters approach equilibrium.
Thus, this ophiolite system serves as a natural analogue, providing insight into what could be expected from engineered injection of CO 2 into porous peridotite and offers an example scenario of what might happen if CO 2 leaks from an injection site. Such studies of natural analogues are useful for providing information about natural processes that laboratory studies of ideal systems cannot. Studying natural systems gives information about how elements or products are favoured or disfavoured in complicated competitive reactions.
FIG. 5. The saturation state of selected carbonate minerals as a function of pH (addition of CO 2 ) with respect to the water from (a) Spring #1 at Sampling site OM01, and (b) Spring #2 at Sampling site OM06. Secondary minerals were not allowed to precipitate in the model. The reaction progress is from right to left in the figures. Before the addition of CO 2 , the pH was 11.6 and temperature was 30.6ºC for Spring #1, and 11.5 and 20.3ºC for Spring #2. The logarithm of the initial partial pressure of CO 2 was À9.3 bar for Spring #1 water and À8.3 bar in Spring #2. The logarithm of the final CO 2 pressure after the solution reached equilibrium with the Earth's atmosphere was À3.4 bar or 395 ppm CO 2 .