an experimental study of basaltic glassâ€“h2oâ€“co2

Available online at www.sciencedirect.com

www.elsevier.com/locate/gca

ScienceDirect

Geochimica et Cosmochimica Acta 126 (2014) 123–145

An experimental study of basaltic glass–H2O–CO2 interactionat 22 and 50 �C: Implications for subsurface storage of CO2

Iwona Galeczka a,⇑, Domenik Wolff-Boenisch a,b, Eric H. Oelkers a,c,Sigurdur R. Gislason a

a Institute of Earth Sciences, University of Iceland, Sturlugata 7, 101 Reykjavik, Icelandb Department of Applied Geology, Curtin University, GPO Box U1987, Perth, 6845 Western Australia, Australia

c GET-UMR 5563 CNRS Universite Paul-Sabatier IRD, 14, Avenue Edouard Belin, 31400 Toulouse, France

Received 30 May 2013; accepted in revised form 29 October 2013; Available online 13 November 2013

Abstract

A novel high pressure column flow reactor was used to investigate the evolution of solute chemistry along a 2.3 m flow pathduring pure water- and CO2-charged water–basaltic glass interaction experiments at 22 and 50 �C and 10�5.7 to 22 bars partialpressure of CO2. Experimental results and geochemical modelling showed the pH of injected pure water evolved rapidly from6.7 to 9–9.5 and most of the iron released to the fluid phase was subsequently consumed by secondary minerals, similar tonatural meteoric water–basalt systems. In contrast to natural systems, however, the aqueous aluminium concentrationremained relatively high along the entire flow path. The aqueous fluid was undersaturated with respect to basaltic glassand carbonate minerals, but supersaturated with respect to zeolites, clays, and Fe hydroxides. As CO2-charged water replacedthe alkaline fluid within the column, the fluid briefly became supersaturated with respect to siderite. Basaltic glass dissolutionin the column reactor, however, was insufficient to overcome the pH buffer capacity of CO2-charged water. The pH of thisCO2-charged water rose from an initial 3.4 to only 4.5 in the column reactor. This acidic reactive fluid was undersaturatedwith respect to carbonate minerals but supersaturated with respect to clays and Fe hydroxides at 22 �C, and with respectto clays and Al hydroxides at 50 �C. Basaltic glass dissolution in the CO2-charged water was closer to stoichiometry thanin pure water. The mobility and aqueous concentration of several metals increased significantly with the addition of CO2

to the inlet fluid, and some metals, including Mn, Cr, Al, and As exceeded the allowable drinking water limits. Iron becamemobile and the aqueous Fe2+/Fe3+ ratio increased along the flow path. Although carbonate minerals did not precipitate in thecolumn reactor in response to CO2-charged water–basaltic glass interaction, once this fluid exited the reactor, carbonates pre-cipitated as the fluid degassed at the outlet. Substantial differences were found between the results of geochemical modellingcalculations and the observed chemical evolution of the fluids during the experiments. These differences underscore the need toimprove the models before they can be used to predict with confidence the fate and consequences of carbon dioxide injectedinto the subsurface.� 2013 Elsevier Ltd. All rights reserved.

1. INTRODUCTION

This study focuses on the improved understanding ofCO2-charged fluid–basalt interaction. This process is of cur-

0016-7037/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.gca.2013.10.044

⇑ Corresponding author. Tel.: +354 525 5414; fax: +354 5629767.

E-mail address: [email protected] (I. Galeczka).

rent interest due to its potential application to subsurfacecarbon storage efforts. Engineered in situ mineral carbon-ation attempts to combine injected CO2 with divalent cat-ions such as Ca2+, Mg2+, and Fe2+ to form stablecarbonate minerals (IPCC, 2005; Kelemen and Matter,2008; Oelkers et al., 2008; Gislason et al., 2010; Broecker,2012). Because basalts and ultramafic rocks are (1) rich indivalent cations, (2) abundant on the Earth surface, and


mailto:[email protected]


124 I. Galeczka et al. / Geochimica et Cosmochimica Acta 126 (2014) 123–145

(3) highly reactive – they are considered to be well suited forCO2 mineral storage (McGrail et al., 2006; Goldberg et al.,2008). This possibility is currently being tested as part of theCarbFix CO2 storage pilot project in Iceland(Gislason et al., 2010; Aradottir et al., 2012; Alfredssonet al., 2013).

The interaction of CO2-charged fluids with basalts alsoplays a major role in the global carbon cycle. Carbon diox-ide released from the crust and the mantle during continen-tal drift is balanced, on a geological time scale byweathering of Ca, Mg–silicates, formation of carbonatesand burial of organic carbon (Berner, 2004; Marini, 2007;Mackenzie and Andersson, 2013). Part of the CO2 releasedfrom volcanoes and crustal intrusions, however, neverreaches the atmosphere. It dissolves in groundwater andgeothermal waters, making the water corrosive and provok-ing carbonate mineral precipitation. This ‘short-cut’ in thecarbon cycle provides a natural analogue for industrialin situ CO2 storage (Brady and Gislason, 1997; Gislasonet al., 2002; Kelemen and Matter, 2008; Wiese et al.,2008; Flaathen et al., 2009; Beinlich and Austrheim, 2012;Kampman et al., 2012; Olsson et al., 2012a).

Numerous laboratory studies investigating the poten-tial efficiency of in situ CO2 storage in basalts and ultra-mafic rocks have focused on the secondary products ofCO2–H2O/CO2–H2S interaction and/or the effects of sec-ondary mineral coatings on primary mineral dissolutionrates (e.g. Giammar et al., 2005; Andreani et al., 2009;Daval et al., 2009, 2011; Schaef and McGrail, 2009;Schaef et al., 2010, 2011, 2013; Hovelmann et al., 2011;Stockmann et al., 2011, 2013; Gysi and Stefansson,2012; Munz et al., 2012; Olsson et al., 2012b; Milleret al., 2013; Rani et al., 2013). In contrast, this study fo-cuses on the solute chemistry evolution during water–basaltic glass and CO2-charged water–basaltic glass inter-action. Such results can be applied to engineered in situ

mineral carbonation as well as to volcanic systems whereCO2 is degassed from magma into groundwater (Federicoet al., 2002, 2004; Aiuppa et al., 2005; Fridriksson et al.,2006; Flaathen and Gislason, 2007; Flaathen et al., 2009;Olsson et al., 2012a). A major difference between thisstudy and previous studies of water–basalt interaction isthe experimental design; in this study we follow the fluidcompositional evolution within a 2.3 m long high pressurecolumn flow reactor (HPCFR) (Galeczka et al., 2013a)using sampling ports located along the column. This reac-tor allows reactive fluid sampling along the flow path atboth ambient and at elevated pressure and temperature,providing insight into the temporal and spatial evolutionof reactive fluid composition. This reactor, thereby, allowsfor the direct assessment of the ability of geochemicalmodelling codes to reproduce the solute chemistry alongthe flow path during CO2–water–rock interaction. Suchmodel validation is an essential step to assess our abilityto predict the fate of CO2 injected into the subsurfaceduring carbon storage efforts. The purpose of this paperis to present the results of this experimental study andto apply these to the improved understanding of carbonstorage in basalts, and to the fate of CO2 in natural vol-canic systems.

2. METHODS

2.1. Materials

The basaltic glass used in this study was collected fromthe Stapafell Mountain located in SW Iceland. The compo-sition of the material is similar to that of mid-ocean ridgebasalt (MORB) and its dissolution kinetics was reportedin previous studies (Oelkers and Gislason, 2001; Gislasonand Oelkers, 2003; Stockmann et al., 2011; Wolff-Boenischet al., 2011). Its chemical composition is consistent withSi1.0Ti0.024Al0.355Fe0.207Mg0.276Ca0.265Na0.073K0.007O3.381

(see Table 1 in Electronic supplement 1). The material wascrushed in a jaw crusher and dry sieved to obtain the 45–100 lm particle size fraction, which was subsequentlywashed by repeated gravitational settling to remove ultra-fine particles. The resulting powder was dried at 50 �C for2 days. The cleaned glass was poured into the column reac-tor. Scanning electron microscope (SEM) photomicro-graphs of the material inserted into the column before theexperiments can be seen in Fig. 1 in the Electronic supple-ment 1. As shown in the photomicrographs, the powderwas free of fine particles. The total mass of glass insidethe column reactor was �8.3 kg. The void fraction of thematerial, which is randomly and loosely packed, was esti-mated to be �0.4 (Weltje and Alberts, 2011). The powderedglass was inserted into the reactor as a slurry precluding thedirect measurement of void fraction. To ensure equal voiddistribution, the powdered glass was inserted into the col-umn gradually; the slurry addition was regularly stoppedto allow the solid material to settle. In total, it took a weekto fill the column. The reactor was also physically shaken attimes during slurry addition to further aid grain settling.

The BET specific surface area (ABET) of the basaltic glassbefore the experiment was 22,000 cm2/g, as measured by six-point N2 adsorption using a Quantachrome Gas Sorptionsystem. The specific geometric surface area (Ageo), calcu-lated assuming the glass powder to be identical cubes was290.4 cm2/g (Wolff-Boenisch et al., 2011), resulting in a sur-face roughness factor (ABET/Ageo) of 76. The total BET sur-face area in the column was thus �182,600,000 cm2, and thecorresponding geometric surface area was �2,410,000 cm2.Taking into account an estimated porosity of 40%, the totalfluid volume in the reactor was 1.84 L, yielding a surfacearea to fluid volume ratio of �105 cm�1.

2.2. Experimental design

Fluid–basaltic glass interaction experiments were per-formed in a high pressure column flow reactor (HPCFR).A detailed description of the column reactor is providedin Galeczka et al. (2013a). Due to the corrosive nature ofCO2-charged water, nearly the entire reactor was made oftitanium, Hastelloy, or PEEK (Polyether ether ketone).The titanium column measured 234 cm in length, 5.0 cmin inner diameter, and held a total volume of �4.6 L. Dur-ing the CO2-charged water experiments, liquid CO2 wasdelivered to the reactor via a supercritical fluid pump(SCF) and the degassed deionized H2O via a high pressureliquid chromatography pump (HPLC). The complete

I. Galeczka et al. / Geochimica et Cosmochimica Acta 126 (2014) 123–145 125

mixing and dissolution of CO2 into the water was assuredby a titanium mixing chamber installed before the columnreactor inlet. The initial in situ DIC (Dissolved InorganicCarbon) concentration during the CO2-charged waterexperiments was measured in the mixing chamber every2–3 days to verify the H2O/CO2 ratio. The HPCFR has se-ven lateral sampling ports, which were used to sample thefluid along the flow path. Fluid sampling from these portsallowed determination of the elemental concentrations aswell as DIC, pH, and Eh along the flow path within the col-umn. The column reactor was wrapped with a heating tapeto control temperature. Fluid sampling was performed bydiverting the fluid from the reactor outlet to a pressurizedsampling loop connected to an ‘expander’ for DIC analysisand to in-line pH and Eh electrodes. The ‘expander’ wasequipped with a pressure transducer that recorded the pres-sure resulting from CO2 expansion. Further samples werecollected for major and trace cation measurements. Detailsof this reactor are presented by Galeczka et al. (2013a).

2.3. Geochemical modelling

The standard state adopted in this study was that of unitactivity for pure minerals and H2O at any temperature andpressure. The standard state for aqueous species was that ofa hypothetical 1 molal solution referenced to infinite dilu-tion at any temperature and pressure. Aqueous speciation,mineral saturation states, and reactive transport modellingwere performed using the PHREEQC 2.17 geochemicalcode (Parkhurst and Appelo, 1999) and the standard phre-

eqc.dat database, which was updated with selected aqueousspeciation and mineral solubility constants from Gysi andStefansson (2011). The thermodynamic properties (equilib-rium constant and enthalpy) of the hydrated basaltic glasssurface were estimated from the stoichiometrically weightedsum of the hydrolysis reactions of amorphous SiO2 andamorphous Al(OH)3 (Bourcier et al., 1990; Wolff-Boenischet al., 2004). The equilibrium constant (logK) and enthalpy(DH) of individual hydrolysis reactions were taken from thephreeqc.dat database. The logarithm of the equilibrium con-stant at 25 �C calculated for the hydrated basaltic glass dis-solution reaction given by

Si1:0Al0:35O2 OHð Þ1:05 þ 1:05Hþ þ 0:95H2O

¼ 0:35Al3þ þH4SiO4 ð1Þ

is 1.07. The enthalpy of reaction is calculated to be�24.4 kJ/mol. For aqueous speciation and mineral satura-tion state calculations, the chemical composition of thesampled fluid together with measured in-line DIC andpH were used. In cases where DIC measurements weremissing, PHREEQC calculations were performed to esti-mate DIC concentrations using measured in situ pH assum-ing charge balance. Resulting DIC estimates were used forfurther calculations. Aqueous speciation and saturationstate calculations included measured Fe2+ and Fe3+ con-centrations as redox indicators for the CO2-charged waterexperiments, but were calculated from measured dissolvedO2 concentrations of the degassed inlet fluid for the purewater experiment.

The saturation state of the reactive fluid with respect tohydrated basaltic glass is given as the Gibbs free energy ofreaction, DGr, but the saturation state of the experimentalfluid with respect to secondary minerals is given as satura-tion index, SI. The relationship between these two functionsis expressed as

DGr ¼ RT 2:303 logðQ=KÞ ¼ RT 2:303 SI; ð2Þ

where R corresponds to the gas constant, T designates thetemperature, Q stands for the reaction quotient (also calledion activity product), and K denotes the equilibrium constantof the hydrolysis reaction at the temperature of interest. BothDGr and SI are zero at equilibrium and negative when thefluid is undersaturated with respect to the dissolving phase.

The dissolution rate of basaltic glass can be describedusing (Gislason and Oelkers, 2003)

rþ;geo ¼ AAexp�EARTð Þ a3

Hþ

aAl3þ

� �13

1� expDGr

rRT

� �� ; ð3Þ

where r+,geo signifies the geometric surface area normalizedsteady-state basaltic glass dissolution rate, AA designates apre-exponential factor (10�5.6 mol of Si/cm2/s), EA refers toa pH independent activation energy (25.5 kJ/mol), and r cor-responds to the Temkin parameter, equal to 1 for basalticglass when its formula is normalized to one Si atom (Dauxet al., 1997). The activation energy adopted in this study islower than that commonly reported in the literature becauseit includes provision for the identity of the Al-depleted acti-vated complex (Gislason and Oelkers, 2003). The DGr sym-bolizes the Gibbs free energy of the hydrated basaltic glassdissolution (Eq. (1)). The term 1� exp DGr

rRT

� �� reflects the sat-

uration state of the fluid with respect to the hydrated basalticglass in accord with reaction (1).

The basaltic glass dissolution rates along the flow pathwere calculated using Eq. (3). The activities of Al3+ andH+ were calculated from the measured fluid chemistry usingPHREEQC. Apparent basaltic glass dissolution rates dur-ing the experiments were also determined from measuredfluid Si concentration using:

rþ;Si ¼ðSi1 � Si0Þfr

Ageom; ð4Þ

where r+,Si designates the apparent basaltic glass dissolu-tion rate based on Si release, Si1 represents the measuredoutlet fluid Si concentration, Si0 refers to the initial Si con-centration, and fr, Ageo, and m stand for fluid flow rate, spe-cific geometric surface area, and the mass of basaltic glass,respectively. Note that the precipitation of Si bearing sec-ondary minerals within the column reactor will diminishapparent dissolution rates calculated with Eq. (4).

The fluid compositional evolution in the reactor wasmodelled using a one dimensional reactive transport simu-lation with the aid of PHREEQC. The fluid phase was al-lowed to react with basaltic glass in accord with Eq. (3),taking into account the fluid composition and its saturationstate with respect to the hydrated basaltic glass surface. Theinlet fluid chemistry used in the model was set equal to theinlet fluid chemistry used during the experiments (seebelow) and the basaltic glass was assumed to dissolve


stoichiometrically. The one dimensional reactive transportmodel consisted of seven cells. The flow of the fluid was di-rected from the first towards the seventh cell, and the timestep of the simulation, as required by the PHREEQC com-puter code, equalled the fluid residence time within each cell(the fluid residence time is equal to the total fluid volumeinside the cell divided by its flow rate). Diffusion was alsotaken into account using a diffusion coefficient of3 � 10�10 m2/s. Note that the effects of dispersion and fluidchannelling were not taken into account directly by thismodel. The basaltic glass surface area used in the modelwas the initial Ageo. In the first scenario, the basaltic glasswas allowed to dissolve according to the rate expression(Eq. (3)) but secondary minerals were not allowed to pre-cipitate. In the second scenario, basaltic glass dissolutionwas also calculated with Eq. (3), but the secondary mineralscommonly forming in basaltic groundwater systems wereassumed to form at local equilibrium. The local equilibriumassumption was used instead of distinct mineral precipita-tion rates since (1) such rates are generally unavailable,(2) there is no constraint on secondary mineral surface area,and (3) some precipitation rate expressions, predicted bytransition state theory, have been shown to fail (Saldiet al., 2012; Schott et al., 2012). Although the assumptionof partial equilibrium between fluid and secondary phaseshas been questioned (c.f. Zhu and Lu, 2009), it was usedin the present study for illustrative purpose to provide in-sight into the processes influencing the evolution of the fluidchemistry, and the potential application of geochemicalmodels to predict the evolution of CO2–rock interactionin laboratory and geo-engineered systems.

2.4. Experimental stages

The experiment carried out in this study was performedin 3 distinct stages. In the first stage deionized water (DIwater) was pumped through the column at a flow rate of5 ml/min, at 22 �C and ambient pressure for 2500 h. Thisexperiment will be referred to as the pure water experiment.The residence time of the fluid inside the column was �6 h.The second stage of the experiment started after completionof the pure-water experiment. The total/hydraulic pressurein the reactor was increased to 8 MPa and CO2-chargedwater was injected. This experiment will be referred to asthe 22 �C CO2-charged water experiment. The total pressureand the flow rates of water and liquid CO2 were set to ensurethat the CO2 was fully dissolved in the water before enteringthe column reactor. In this instance it was set at�25% of theCO2 solubility limit at this pressure and temperature. TheCO2 solubility was calculated based on the Duan et al.(2006) model. This CO2 concentration thus avoids degassingeven if there were some fluctuations in the total pressure dueto sampling. In the third stage of the experiment the temper-ature of the HPCFR was increased to 50 �C ± 0.5. Thisexperiment will be referred to as the 50 �C CO2-chargedwater experiment. These three experimental stages were cho-sen to mimic different conditions in basaltic aquifers. Thefirst stage (pure water–basaltic glass experiment) representsnatural basaltic groundwater systems, the second stage(22 �C CO2-charged water experiment) simulates natural

shallow magmatic CO2 fluxes into basaltic aquifers andengineered CO2 injection similar to the conditions at theCarbFix CO2 injection site (Alfredsson et al., 2013), whereasthe third stage mimics CO2 fluxes/injection at greater depthswhere the temperature is higher. The three experimentalstages were conducted sequentially, with secondary phasesformed in one experiment potentially influencing the follow-ing one. This mimics natural geological systems, where sec-ondary alteration products are commonly present withinpores. Similar, the CO2 injection during the CarbFix project(Gislason et al., 2010) has been proceeded into basalts con-taining some secondary product phases.

In both CO2-charged water experiments, the CO2 andwater flow rates were set to 0.13 and 3.5 ml/min, respectivelyresulting in an average fluid residence time of �8 h in thecolumn. The duration of the two CO2-charged water exper-iments were 1000 h each. The velocity of the fluid throughthe pores was �7 m/day which is 100 times faster than nat-ural groundwater flow at the CarbFix injection site (Aradot-tir et al., 2012). The initial measured DIC concentration andcalculated pH of the CO2-charged inlet fluid was�300 mmol/L and �3.4, respectively. The fluid phase pH,Eh, and DIC were measured in samples collected from sevenpositions along the column reactor. The precision of the pHmeasurements during the CO2-rich fluid experiments ob-tained using high P/T electrodes was estimated to be ±0.1pH unit but was ±0.05 pH unit for measurements made atambient pressure during the pure water experiment. Reactorsampling involved taking fluids from all seven samplingports during a single day, over the course of 8–9 h. Sampleswere taken a minimum of 6 residence times from each other(see Fig. 1). In total, 347 fluid samples were collected: 191during the pure water experiment, 79 during the 22 �CCO2-charged water experiment, and 77 during the 50 �CCO2-charged water experiment. The sampling affectedsomewhat the fluid flow in the column; the fluid above thesampling port was stagnant during sampling. To samplefresh fluids, sampling was always started from the upper-most sampling port and continued downwards. The outletfluid sampled for major and trace elemental compositionwas filtered using a 0.2 lm cellulose acetate filter, and thenacidified with concentrated supra-pure HNO3 (0.5 vol.%)prior to analysis by ICP-OES (Inductively Coupled PlasmaOptical Emission Spectrometer). Analytical uncertainties ofICP-OES analyses are on the order of 65%. During theCO2-rich experiments, the concentrations of Fe species(Fe2+ and Fe3+) were measured by ion chromatography(Dionex IC 3000). These samples were captured directlyfrom the filtered outflow into a syringe filled with 0.1 MHCl, thus avoiding direct contact of the fluid with the atmo-sphere. This method takes advantage of the slow oxidationof Fe2+ to Fe3+ at low pH to preserve the iron oxidationstate of the sample (Stumm and Morgan, 1996).

3. RESULTS

3.1. Pure water–basaltic glass interaction experiment

The HPCFR provides the opportunity to study the tem-poral fluid chemistry evolution along the flow path. Fig. 1

Fig. 1. Example of the data collected from the different outlet sampling ports during the pure water–basaltic glass interaction experiment. Theinlet fluid of pH 6.7 was pumped bottom-up through the column continuously for 2500 h (elapsed time) and it is shown on the horizontal axis.The calculated residence time from the inlet to each sampling outlet is displayed on the vertical axis. Seven fluid samples were collected duringeach sampling session, one from each outlet port.


shows an example of results obtained during the pure waterinjection at 22 �C. The inlet fluid percolated through thecolumn passing sequentially the first, second, third, fourth,fifth, sixth, and seventh sampling ports, respectively. Thevertical axis, labelled ‘residence time’ in Fig. 1, shows theaverage time it took the fluid to flow from the inlet to eachsampling port at a flow rate of 5 ml/min. The horizontalaxis labelled ‘elapsed time’ represents the elapsed time sincethe beginning of the experiment. In case of the pure water–basaltic glass experiment, deionized water was pumpedthrough the column continuously for 2500 h (104 days).The aqueous fluid concentrations attained steady-state firstin the first outlet and then later in the higher outlets, asshown for Mg in Fig. 1. Steady-state is defined as constantfluid composition within analytic uncertainty for at least 10residence times.

Major element concentrations and the pH evolutionalong the flow path are shown at elapsed times of 500and 2200 h, respectively in Fig. 2. These two elapsed timeswere chosen because after 500 h the concentrations of mostelements were increasing. After 2200 h, the dissolved major

element concentrations attained steady-state in all outlets.The initial pH of the degassed, deionized water injected intothe column was 6.7 at 22 �C. The fluid pH increased tomore than 9 during the first 30 min of water–rock interac-tion consistent with proton consumption by basaltic glassdissolution (see Fig. 2a). After 2–3 h, the pH slightly de-creased consistent with proton production by secondaryphase precipitation.

Along the flow path, the concentrations of most majorelements increased during the first 2 h of water–rock inter-action at both 500 and 2200 h, as indicated by samples col-lected from the first three outlets (see Fig. 2b and c). Majorelement concentrations continued to increase with furtherfluid–basaltic glass interaction but at a slower rate as thefluid continued through the column. After 3 h, the Mg con-centration in the reactor fluid decreased suggesting prefer-ential incorporation of Mg into secondary phases. Thestoichiometric coefficients (the ratio between measured ele-ment concentration to Na in the fluid normalized to that inthe dissolving glass) of Si, Al, Ca, Mg were close to 1. So-dium is considered to be a conservative element in basaltic

6.5

7.0

7.5

8.0

8.5

9.0

9.5

0 1 2 3 4 5 6pH

Residence time in the column [hours]

Elapsed time of 500 hours


0.00

0.01

0.02

0.03

0.04

0.05

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0 1 2 3 4 5 6

[mm

ol/k

g]

[mm

ol/k

g]


Si Al Fe Mg Ca Na

0.00

0.01

0.02

0.03

0.04

0.05

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0 1 2 3 4 5 6

[mm

ol/k

g]

[mm

ol/k

g]


Si Al Fe Mg Ca Na

(a)

(b)

(c)

Fig. 2. The pH (a) and concentration of major elements from the pure water–basaltic glass interaction experiment along the flow path atelapsed times of 500 h (b) and 2200 h (c), respectively. The silicon concentration is shown on the left axis whereas all other elementconcentrations are shown on the right axis. The increase in the pH from the first to third outlet sampling ports indicates proton consumptionby basalt dissolution. The subsequent pH decrease indicates proton release. The Fe concentration was below detection limit of analysis,revealing precipitation of Fe phases.


groundwater systems due to its negligible incorporationinto secondary phases. Despite these ratios being close tounity, the concentration curves levelled off along the flowpath as shown in Fig. 2b and c. The Al concentrationwas still comparatively high in the fluid phase after 6 h ofwater–basaltic glass interaction (approximately 40 lmol/kg) whereas in natural basaltic groundwater systems aque-ous Al ranges only up to few lmol/kg (Gislason et al., 1996;Alfredsson et al., 2013). The Fe concentration along theflow path was close to or below the 0.36 lmol/kg detectionlimit of the analytical method. It is common to observe upto 1 lmol/kg of Fe in natural basaltic groundwater systems;this low concentration is likely due to Fe2+ oxidation cou-pled to Fe hydroxide precipitation (Gislason and Eugster,1987; Alfredsson et al., 2013).

3.2. CO2-charged water–basaltic glass interaction

experiment at 22 �C

The injection of a CO2-charged water was startedafter 2500 h of the pure water–basaltic glass experiment.The pH decrease stemming from the injection of theCO2-charged water into alkaline fluid is shown in Fig. 3a.During the first hours following the start of the injectionof the CO2-charged water, it was difficult to measure in-linepH in all outlets because of the time required for sampling.A steady-state pH of �4.5 was attained after �200 h. ThepH increased from the initial inlet fluid value of 3.4 to 4.5in the samples obtained from the first outlet port; fluid sam-ples from this port experienced only 40 min of CO2-chargedwater–basaltic glass interaction (see Fig. 4a). Further

4.0

4.5

5.0

5.5

0 200 400 600 800 1000

pH

Elapsed time [hours]

column outletoutlet 7outlet 6outlet 5outlet 4outlet 3outlet 2outlet 1

0.0

0.2

0.4

0.6

0.8

1.0

0 200 400 600 800 1000

DIC

[m

ol/L

]


(b)

(a)

Fig. 3. The pH (a) and DIC (b) measured in-line in the reactivefluids collected from the sampling ports indicated in the legendduring the 22 �C CO2-charged water–basaltic glass experiment.Due to complexity of the sampling procedure, only the columnoutlet fluid pH was determined during the first 60 h of theexperiment when the alkaline solution inside the column wasreplaced by CO2-charged water. The dotted curve represents initialmeasured DIC concentration in the mixing chamber beforeentering the column. Fluctuations of the initial DIC concentrationwere related to the fluctuations in delivery of CO2 by thesupercritical fluid pump. Circles on the lines represent the samplesthat were singled out for further discussion in the text.


interaction between basaltic glass and CO2-charged wateralong the flow path was not sufficient to overcome the pHbuffering capacity of the fluid, resulting in a steady-state be-tween proton consumption and proton production at pH�4.5 (see Figs. 3a and 4a). The DIC concentration in thecolumn was similar to the inlet DIC concentration of�300 mmol/L and it remained constant along the flow path(Fig. 3b). The constant concentration of DIC along the flowpath confirms that there was no CO2 degassing or carbon-ate precipitation in the column. A decrease of DIC at anelapsed time of 500 h was caused by a fluctuation of liquidCO2 delivery by the injection pump system.

Dissolved element concentrations along the flow path490 and 875 h after the beginning of the CO2-charged waterinjection showed element release patterns (Fig. 4b and c)distinct from those of the pure water experiment (seeFig. 2b and c). We chose to show the results after 490 h,as at this time the initial DIC concentration had been stablefor almost 300 h (see Fig. 3b) and the element concentra-tions (Si, Ca, Mg, Na, K, Fe) in all of the sampling outletports were approaching steady-state. The concentrations

after 875 h are shown because at this time the pH had beenstable for �300 h in all fluid samples (Fig. 3a) and all ele-ment concentrations had achieved steady-state.

The observed abundance of major elements in the fluidphase agreed with those of the basaltic glass:Si > Al > Mg > Ca > Fe > Na. This observation contrastswith the results of the pure water experiment where theabundance of elements was Si > Al > Ca > Mg > Na > Fe.There was a slight non-linearity in Si, Ca, Mg, and Fe con-centrations versus flow distance, whereas the Al concentra-tion increased linearly with flow distance. At an elapsedtime of �875 h Si, Ca, Mg, and Fe showed preferentialretention in the solid phase at all sampled fluids (stoichiom-etric coefficient <1 compared with Na). The Al was prefer-entially retained in the solid compared to Na (stoichometriccoefficient was <1) at the first and second sampling outletport.

Outflow chemistry in this CO2-charged water experi-ment exhibited higher trace element concentrations com-pared to the pure water experiment. For example, theoutflow fluid in the latter experiment contained more than40 times Mn and 6 times Sr than the former (Fig. 5). Ele-ments which were close to or below detection limits duringthe pure water–basalt experiment were detectable duringthe CO2-charged water experiment. This observation is con-sistent with faster basaltic glass dissolution rates at anacidic pH of 4.5 compared to those in the basic fluids pres-ent in the pure water experiment. Concentrations of B, Mn,and Sr after 875 h showed a linear increase along the flowpath similar to the major elements. The B is considered asthe one of the most mobile elements in basaltic groundwa-ter systems since it is not incorporated into secondary min-erals (Arnorsson et al., 1982, 2002; Arnorsson andAndresdottir, 1995; Kaasalainen and Stefansson, 2012).

The CO2-charged water, when reacting with basalt, notonly liberates divalent metals which can form carbonateminerals, but also releases metals that can be harmful tobiota. Major metals of concern include Al, Cr, and As.The Fe and Mn can be both essential for life and toxicdepending on their concentrations. The degree to whichthey migrate depends on the chemistry and hydrology ofthe system. The mobilization of contaminants due to engi-neered CO2 injection was studied among others by Harveyet al. (2013) and Siirila et al. (2012). Since the experimentalconditions in this study were similar to natural groundwa-ter systems where magmatic CO2 dissolves in aquifers, theconcentrations of metals were compared to World HealthOrganisation (WHO) drinking water standards. Concentra-tions of some of the metals in the sampled fluids exceededWHO drinking water limits (WHO, 2008). Aluminium,iron, and chromium concentrations were always higherthan the WHO drinking water limits (>3.7 lmol/kg forAl, >36 lmol/kg for Fe, and >0.1 lmol/kg for Cr). Manga-nese exceeded the WHO drinking water limit (7.3 lmol/kg)during the first 40 h after the pure water inlet fluid was re-placed by the CO2-charged water. The As concentrationwas close to the WHO drinking water limit of 0.13 lmol/kg.The mobility of metals during short-term CO2 pulses inthe high pressure column reactor is described in more detailin Galeczka et al. (2013b).

3.0

3.5

4.0

4.5

5.0

0 2 4 6 8pH




0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 2 4 6 8

[mm

ol/k

g]

[mm

ol/k

g]


SiAlFeMgCaNa

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 2 4 6 8

[mm

ol/k

g]

[mm

ol/k

g]


SiAlFeMgCaNa

(a)

(c)

(b)

Fig. 4. The pH (a) and major element concentrations during the 22 �C CO2-charged water–basaltic glass experiment at elapsed times of 490 h(b) and 875 h (c), respectively. Silicon concentration is shown on the left axis whereas all other element concentrations are shown on the rightaxis.


The results of the first 90 h of the 22 �C CO2-chargedwater experiment are comparable to the results of theCO2-charged water–basaltic glass interaction experimentdescribed in Galeczka et al. (2013a). The element concen-trations showed similar trend along the flow path, and themeasured pH was similar even though the inlet CO2 con-centration was a factor of four higher than in this study.

3.3. CO2-charged water–basaltic glass interaction at 50 �C

After approximately 1000 h of CO2-charged water–basaltic glass interaction at 22 �C, deionized water was

pumped through the column at 3.5 ml/min for �900 h toneutralize the system. After this time, CO2 was again mixedwith DI water at the same ratio as during the 22 �CCO2-charged water experiment (H2O at 3.5 ml/min, CO2(l)

at 0.13 ml/min) and pumped through the column reactor.The reactor temperature was increased to 50 �C. The tem-poral pH and DIC evolution during this experiment isshown in Fig. 6. The pH was measured at ambient temper-ature and recalculated to 50 �C using PHREEQC. The pHdecreased gradually with time, possibly related to anunintentional increase in initial DIC concentration dueto fluid injection system variations. Even though the

0.0

0.1

0.2

0.3

0.4

0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

0 2 4 6 8

[µm

ol/k

g]

[µm

ol/k

g]


BMnSrBaCrAsTi

0.0

0.1

0.2

0.3

0.4

0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

0 2 4 6 8

[µm

ol/k

g]

[µm

ol/k

g]


BMnSrBaCrAsTi

(b)

(a)

Fig. 5. Trace element concentrations during the 22 �C CO2-charged water–basaltic glass experiment at elapsed times of 490 h (a) and 875 h(b), respectively. The B and Mn concentrations are shown on the left axis; other elements are shown on the right axis.


DIC concentration was homogeneous and close to theinitial DIC concentration throughout the reactor, pHvaried more between the fluids obtained from the variousoutlet sampling ports than during the 22 �C CO2-chargedwater experiment (Figs. 4a and 7a). After an elapsed timeof 840 h in the 50 �C CO2-charged water experiment, pHmaximized to 4.4 in outlet samples 2 and 3, and then itdecreased along the rest of the flow path to 4.2. The increasein temperature from 22 to 50 �C did not enhance basalticglass dissolution rates sufficiently to overcome the pHbuffer capacity of the CO2-charged water.

Solute concentrations along the fluid flow path atelapsed times of 360 and 840 h during the 50 �C CO2-charged water experiment were distinct from those ob-served during the 22 �C CO2-charged water experiment(compare Fig. 4b and c with Fig. 7b and c). During the first2–3 h of CO2-charged water–basaltic glass interaction, theAl concentration increased significantly, followed by its ra-pid decrease after �3–4 h. The dissolved Al concentrationat elapsed time of 360 h, and after 7.5 h of CO2-chargedwater–basaltic glass interaction, was lower than the corre-sponding Si, Mg, Ca, Fe, and Na concentrations (seeFig. 7b). Reactive fluid Si concentrations along the flowpath decreased from 360 to 840 h of elapsed time (Fig. 7band c), in contrast with Mg and Ca, which increased withelapsed time. The pattern of Ca and Mg enrichment alongthe flow path was similar to that observed during the 22 �CCO2-charged water experiment. The Si:Na ratio of the reac-tive fluid was greater than that of the basaltic glass atelapsed time of 840 h in all sampling outlet ports. Ca and

Mg were preferentially released to the fluid compared toNa up to the fourth sampling port, Al was preferentially re-leased to the fluid compared to Na up to the third samplingport and Fe was preferentially released to the fluid phasecompared to Na in just the first sampling port. Later alongthe flow path, these elements were preferentially retained bythe solid.

The Sr and Mn concentrations increased along the flowpath (see Fig. 8b), similar to that seen during the 22 �CCO2-charged water experiment (Fig. 5b). The Cr releasepattern was similar to that of Al along the flow path. Con-centrations of toxic metals in the sampled fluids exceededWHO drinking water limits for a number of elements(WHO, 2008). Aluminium and iron concentrations were al-ways higher than the WHO limits (>3.7 lmol/kg for Al,>36 lmol/kg for Fe). The Cr concentration consistently ex-ceeded WHO limits in the first and second sampling port,but its concentration decreased along the flow path anddid not exceed the WHO limits in the last sampling port.Manganese exceeded the WHO drinking water limit(7.3 lmol/kg) during the first 200 h of the experiment inall outlet samples. The As concentration was very close tothe WHO drinking water limit of 0.13 lmol/kg.

3.4. Iron chemistry and redox conditions

According to Oelkers and Gislason (2001), approxi-mately 90% of the total Fe in the basaltic glass used in thisstudy was Fe2+. The inlet fluids were originally saturatedwith atmospheric O2 at 22 �C, but much of this was

4.0

4.5

5.0

5.5

0 200 400 600 800 1000

pH


outlet 7outlet 6outlet 5outlet 4outlet 3outlet 2outlet 1

0.0

0.2

0.4

0.6

0.8

1.0

0 200 400 600 800 1000

DIC

[m

ol/L

]


(a)

(b)

Fig. 6. The pH (a) and DIC (b) measured in the fluid collectedfrom the sampling ports indicated in the legend during the 50 �CCO2-charged water–basaltic glass experiment. Measurement of pHwas performed at ambient temperature and recalculated to 50 �Cwith PHREEQC. The dotted line represents the initial measuredDIC concentration in the mixing chamber before entering thecolumn. Fluctuations of the initial DIC concentration were relatedto the variations of CO2 delivery by the supercritical pump. Circleson the lines represent the samples that were singled out for furtherdiscussion in the text.


degassed before the inlet fluid entered the column reactor(Galeczka et al., 2013a). The total dissolved Fe concentra-tion was low and below the 0.36 lmol/kg detection limitduring the pure water experiment at 22 �C in all samplingports (Fig. 2b and c and Table 1 in Electronic supplement2). The dissolved Fe concentration was sufficiently high todetermine both total dissolved Fe and Fe speciation inthe fluids sampled from both CO2-charged water experi-ments. To confirm the quality of the analysis, the sum ofFe2+ and Fe3+ measured with ion chromatography wascompared with Fetot measured with ICP-OES. The differ-ences between these concentrations were between 0% and15%, but most were less than 10%. Note that minor oxida-tion could have occurred during sampling. Hence, Fe spe-cies concentrations reported in this study may somewhatunderestimate Fe2+ and overestimate Fe3+.

The evolution of the iron chemistry along the flow pathduring the 22 �C CO2-charged water experiment is shown inFig. 9a and b. During the first 2 h of water–basaltic glassinteraction, Fe was mostly present as Fe3+, but the Fe2+

concentration increased along the flow path and becamemore abundant in the upper part of the column. A similarpattern was observed during the 50 �C CO2-charged water

experiment but Fe2+ became the dominant species earlierthan in the 22 �C experiment (see Fig. 9c and d, and Table 1in Electronic supplement 2).

The measured fluid Eh prior to its injection into thereactor at 22 �C was �200–300 mV (Fig. 2 in Electronicsupplement 1). After an elapsed time of �200 h duringthe 22 �C CO2-charged water experiment, the Eh droppedto ��200 mV and remained constant thereafter. Duringthe 50 �C CO2-charged water experiment, the measuredEh increased continuously from ��200 to ��90 mV. TheEh measured in-line at ambient temperature differed fromthat calculated with PHREEQC based on the Fe2+/Fe3+ ra-tio. Measured Eh was about 800 and 600 mV lower thanthat calculated during the 22 and 50 �C CO2-charged waterexperiments, respectively. The significant difference betweenmeasured Eh and Eh calculated based on measuredFe2+/Fe3+ concentrations during both CO2 experiments(see Fig. 2 in Electronic supplement 1) may be due to a lackof equilibrium between different redox couples in the sam-pled fluids (Lindberg and Runnells, 1984) and the analyticaldifficulties in Eh measurements with a Pt-electrode (Appeloand Postma, 2005; Stefansson et al., 2005). Aqueousspeciation and mineral saturation state calculations withPHREEQC used measured Fe2+ and Fe3+ species to fix theredox conditions for the CO2-charged water experiments,but the calculated dissolved O2 for the pure waterexperiment. As the redox state of these systems was notbuffered, these assumptions lead to some uncertainties inthe geochemical modelling calculations.

3.5. Relative mobility of elements

The relative mobility of the elements was calculated bydividing the measured dissolved element concentrations inthe sampled fluids by the elemental concentrations in the dis-solving basaltic glass. Relative mobilities calculated for ma-jor and trace elements after elapsed time of 2200, 875, and840 h for the pure water experiment, the 22 �C CO2-chargedwater experiment, and the 50 �C CO2-charged water experi-ment, respectively are illustrated in Fig. 10. Fig. 10a, c, and eshows the relative mobility in fluids collected from the sev-enth outlet port, whereas Fig. 10b, d, and f displays the evo-lution of this relative mobility at the same elapsed times as afunction of distance along the fluid flow path.

The most mobile elements were As and B in all experi-ments. Although As concentration in basalts is low, it ishighly mobile in shallow basaltic groundwater systems attemperatures up to 90 �C (Arnorsson, 2003). The least mo-bile measured element was Ti, an element frequently con-sidered immobile in basaltic groundwater systems (e.g.Gislason et al., 1996; Alfredsson et al., 2013). The logmobility of the major elements Si, Ca, Mg, Na, and K dur-ing the pure water experiment was ��4.7 and during bothCO2-charged water experiments ��4.0, indicating that thepresence of CO2 in the fluid rather than a temperature in-crease from 22 �C to 50 �C was the major factor enhancingelement mobility. The highest mobility of Al was observedduring the 22 �C CO-charged water experiment. The logmobility of Fe increased from ��7.0 to ��4.0 when CO2

charged water was injected. Similarly, the mobility of

3.0

3.5

4.0

4.5

5.0

0 2 4 6 8pH


Elapsed time of 360 hoursElapsed time of 840 hours

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 2 4 6 8

[mm

ol/k

g]

[mm

ol/k

g]


SiAlFeMgCaNa

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 2 4 6 8

[mm

ol/k

g]

[mm

ol/k

g]


SiAlFeMgCaNa

(a)

(b)

(c)

Fig. 7. The pH (a) and major element concentrations during the 50 �C CO2-charged water–basaltic glass experiment at elapsed times of 360 h(b) and 840 h (c), respectively. Silicon concentration is shown on the left axis; other element concentrations are shown on the right axis. Thealuminium concentration profile suggests precipitation, with Al enrichment in the fluid at the beginning of the column and a decline furtheralong the flow path.


numerous trace elements increased with the injection ofCO2-charged water; the relative mobility of As, Sr, Cr,Mn, Ba, and V increased by factors of 5, 7, 20, 45, 156,and 160, respectively, by changing the inlet fluid from pureto CO2-charged water.

3.6. Saturation state of basaltic glass and secondary minerals

PHREEQC was used to calculate the in situ saturationstate of the sampled fluids with respect to basaltic glass(Fig. 11) and selected secondary minerals (see Fig. 3 in Elec-tronic supplement 1 and Electronic supplement 3). As de-

scribed in Section 2, the saturation state for basaltic glassis expressed as the Gibbs free energy of reaction, DGr.When the absolute value of DGr is low, basaltic glass disso-lution may slow as equilibrium is approached (c.f. Schottand Oelkers, 1995); based on transition state theory onewould expect a measurable decrease in basaltic glass disso-lution rates when DGr exceeds �10 kJ/mol at 25 �C and�11 kJ/mol at 50 �C. At lower DGr, the 1� exp DGr

rRT

� �� term

in Eq. (3) is >0.98 and therefore does not influence signifi-cantly the overall dissolution rate. The degree to whichbasaltic glass dissolution rates are slowed by the fluid satu-ration state is shown in Fig. 11.

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

0 2 4 6 8

[µm

ol/k

g]

[µm

ol/k

g]


BMnSrBaCrAsTi

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

0 2 4 6 8

[µm

ol/k

g]

[µm

ol/k

g]


BMnSrBaCrAsTi

(b)

(a)

Fig. 8. Trace element concentrations during the 50 �C CO2-charged water–basaltic glass experiment at elapsed times of 360 h (a) and 840 h(b), respectively. The B and Mn concentrations are shown on the left axis; other elements are shown on the right axis.


During the pure water experiment, the basaltic glass dis-solution rate was independent of the fluid saturation stateduring the first 30 min of water–rock interaction (e.g. onlyin samples collected from the first sampling outlet). Thefluid became less undersaturated with respect to basalticglass along the flow path, causing its dissolution rate tobe slowed down by up to 4% (Fig. 11a). During the 22 �CCO2-charged water experiment, the fluid saturation stateinfluenced the basaltic glass dissolution rate along thewhole flow path; the dissolution rate was calculated to haveslowed by up to 13% after �7.5 h of water–rock interaction(e.g. in fluid sampled from the last sampling port) as dis-played in Fig. 11b. During the 50 �C CO2-charged waterexperiment, this dissolution rate depended on the fluid sat-uration state along the whole flow path, and the dissolutionrate was calculated to have slowed by up to 48% after�7.5 h of water–basaltic glass interaction at the beginningof the experiment but by less than 16% at the latter stages(e.g. in fluid sampled from the last sampling port at elapsedtime from 600 h) – see Fig. 11c.

Example of activity-activity diagrams for the reactivefluid evolution of all experiments in the Mg–SiO2–H2O sys-tem, similar to those presented in Marini (2007), Choppingand Kaszuba (2012) and Lu et al. (2013), are provided inFig. 4 of Electronic supplement 1. The reactive fluid duringthe pure water experiment was basic (pH �9.3) and super-saturated with respect to the secondary minerals commonlyfound in natural basaltic systems such as zeolites (repre-sented here by heulandite), clay minerals (represented hereby Mg-clay, Ca-montmorillonite, gibbsite, imogolite, kao-linite), and Fe-phases such as goethite, amorphous goethite,

and Fe(OH)3(am) (see Fig. 3 in Electronic supplement 1 andElectronic supplement 3). Note that the thermodynamicdescription of amorphous goethite, reported as a secondaryproduct of basalt weathering (Stefansson and Gislason,2001), was retrieved by these authors based on measure-ments presented by Diakonov et al. (1994, 1999) andArnorsson and Andresdottir (1995). The fluid was under-saturated with respect to amorphous and cryptocrystallinephases such as Al(OH)3(am) and chalcedony, respectively.During the 22 �C CO2-charged water experiment, the fluidwas undersaturated with respect to zeolites, but supersatu-rated with respect to Ca-montmorillonite after 2–3 h ofwater–basaltic glass interaction. It was supersaturated withrespect to amorphous Fe(OH)3(am), goethite(am), chalce-dony, and undersaturated with respect to amorphousAl(OH)3(am). During the 50 �C CO2-charged water experi-ment, the fluid was again undersaturated with respect tozeolites, but supersaturated with respect to Ca-montmoril-lonite. The fluid was close to and/or in equilibrium withamorphous Al(OH)3(am), goethite(am) and chalcedony, butundersaturated with respect to amorphous Fe(OH)3(am).Supersaturation with respect to Al(OH)3(am), Fe(OH)3(am),

and chalcedony suggests that precipitation of these phasescontrolled the mobility of Al, Fe, and Si in the fluid phase.In natural systems, amorphous phases are the first to pre-cipitate (Gislason et al., 1997; Stefansson and Gislason,2001). Slight supersaturation with respect to siderite oc-curred only at the beginning of the CO2-charged waterexperiments, when the pH was �5. Also, PHREEQC simu-lations of the outlet fluid degassing indicate this fluid wassupersaturated with respect to siderite and ankerite.

0.0

20.0

40.0

60.0

80.0

100.0

120.0

0 2 4 6 8

[µm

ol/k

g]


22 °C CO2 experiment at elapsed time of 490 hours

Fe² Fe³

0.0

20.0

40.0

60.0

80.0

100.0

120.0

0 2 4 6 8

[µm

ol/k

g]



Fe² Fe³

0.0

20.0

40.0

60.0

80.0

100.0

120.0

0 2 4 6 8

[µm

ol/k

g]



Fe² Fe³

0.0

20.0

40.0

60.0

80.0

100.0

120.0

0 2 4 6 8

[µm

ol/k

g]



Fe² Fe³

(b)(a)

(d)(c)

Fig. 9. Distribution of Fe species along the flow path during the 22 �C (a and b) and 50 �C (c and d) CO2-charged water–basaltic glassexperiments. Predominance of Fe3+ in the lower part of the column could be a result of Fe oxidation due to remaining O2 in the inlet fluid.Later along the flow path, Fe2+ became the predominant species.


Carbonates were observed to have precipitated at the reac-tor outlet and their presence was confirmed by simple HCltest. According to model calculations, approximately 6 g ofcarbonate minerals (siderite, ankerite, and calcite) precipi-tated at the reactor outlet during the 1000 h of the 50 �CCO2-charged water experiment.

3.7. Basaltic glass apparent dissolution rates

Fig. 12 shows a comparison between basaltic glass disso-lution rates calculated using Eq. (3) and those obtainedexperimentally based on Eq. (4). Rates shown in Fig. 12were normalized to the total geometric surface area of thebasaltic glass from the bottom of the column to the specificoutlet port. According to the calculations, experimentallyobtained apparent basaltic glass dissolution rates were al-ways slower than those calculated using Eq. (3) (seeFig. 12). This difference increased along the flow path andit was the largest in the 50 �C CO2-charged water experi-ment. The minimum amount of basaltic glass dissolved in-side the column during the experiment was calculated basedon the concentration of Si in the fluid samples collectedfrom the highest sampling port and the amount of waterpumped through the reactor. Note this calculation excludesprovision for secondary Si-phase precipitation. The total

amount of dissolved glass was time dependent. Approxi-mately 40 g of basaltic glass was dissolved during 1000 hof the 50 �C CO2 experiment, �13 g dissolved during1000 h of the 22 �C CO2 experiment, and �6 g dissolvedduring the first 1000 h of the pure water experiment. The to-tal amount of dissolved basaltic glass during the experi-ments (2500 h of pure water experiment and 2000 h forthe two CO2 experiments) was �68 g which was �0.8% ofthe total basaltic glass in the column.

4. DISCUSSION

4.1. Chemical trends during the experiments

The pure water–basaltic glass interaction experimentprovides insight into natural basaltic groundwater systems.The pH in the reactor was similar to that found in naturalbasaltic groundwaters (e.g. Gislason and Eugster, 1987;Gislason et al., 1996; Arnorsson et al., 2002; Alfredssonet al., 2013). Element mobility measured in the reactorwas similar to that in natural basaltic systems with theexception of Al (Fig. 10a). The low Al mobility commonlyobserved in natural environments indicates the precipita-tion of Al hydroxides at low pH and clays and zeolites athigh pH (e.g. Kristmannsdottir, 1979, 1982; Gislason and

-8

-7

-6

-5

-4

-3

-2

-1

0

As B Li Na Si V Ca Al K Mg Sr Mn Cr Ba Fe Ti

log

rela

tive

mob

ility

Pure water experiment

-8.0

-7.0

-6.0

-5.0

-4.0

0 1 2 3 4 5 6

log

rela

tive

mob

ility



Si Al Fe Ca Mg Na Sr Mn

-8

-7

-6

-5

-4

-3

-2

-1

0

As B V Ba Li Al Mn K Sr Na Mg Si Ca Fe Cr Ti

log

rela

tive

mob

ility

22 C CO2 experiment

-8.0

-7.0

-6.0

-5.0

-4.0

0 1 2 3 4 5 6 7 8

log

rela

tive

mob

ility


22 C CO2 experiment

Si Al Fe Ca Mg Na Sr Mn Cr

-8

-7

-6

-5

-4

-3

-2

-1

0

As B V Li Ba Mn Na Sr Mg Si Ca K Fe Al Cr Ti

log

real

tive

mob

ility

50 C CO2 experiment

-8.0

-7.0

-6.0

-5.0

-4.0

0 1 2 3 4 5 6 7 8

log

rela

tive

mob

ility


50 C CO2 experiment

Si Al Fe Ca Mg Na Sr Mn Cr

d)

(a)

(c)

(e)

(b)

(d)

(f)

Fig. 10. The relative mobility of elements in fluids sampled from the seventh outlet after an elapsed time of 2200 h for the pure water–basalticglass experiment (a), 875 h for the 22 �C CO2-charged water–basaltic glass interaction experiment (c) and 840 h for the 50 �C CO2-chargedwater–basaltic glass interaction experiment (e), respectively. Plots (b), (d), and (f) represent relative mobility of major elements along the flowpath during the pure water experiment, the 22 �C CO2 experiment, and the 50 �C CO2 experiment, respectively, at the same elapsed times as(a), (c), and (e). Circles focus on the mobility of Fe.


Eugster, 1987; Crovisier et al., 1992; Gislason et al., 1996;Stefansson and Gislason, 2001). The slowing of Si, Al,Ca, Mg, and Na release rates as shown in Fig. 2b and c isconsistent with either (1) a slowing of basaltic glass dissolu-tion rates along the flow path stemming potentially fromfluid channelling and/or passivation of the basaltic glasssurfaces by secondary phases or (2) the incorporation ofelements into secondary phases. A number of studies havefocused on the effect of secondary coatings on the dissolu-tion rates of primary phases (e.g. Cubillas et al., 2005;

Daval et al., 2009, 2011; Saldi et al., 2013; Stockmann et al.,2013) and some confirmed experimentally a passivation ef-fect of secondary Si-rich layers on dissolution rates (e.g.Daval et al., 2011; Saldi et al., 2013). It seems likely thatsome aluminosilicate minerals precipitated along the entireflow path since the fluids collected from the first samplingport were supersaturated with respect to some clays (gibb-site, imogolite, Mg-clay) and as the fluid travelled furtherit became supersaturated with respect to additional phasesincluding chalcedony, clays (Ca-montmorillonite) and

-12.0

-10.0

-8.0

-6.0

-4.0

-2.0

0.0

0 1000 2000 3000

ΔG

[kJ

/mol

]



outlet 7outlet 6outlet 5oultet 4oultet 3outlet 2outlet 1

-12.0

-10.0

-8.0

-6.0

-4.0

-2.0

0.0

0 200 400 600 800 1000

ΔG

[kJ

/mol

]


22 °C CO2 experiment

-12.0

-10.0

-8.0

-6.0

-4.0

-2.0

0.0

0 200 400 600 800 1000

ΔG

[kJ

/mol

]



4%

13%

48%

16%

(a)

(b)

(c)

Fig. 11. The saturation state of collected fluids with respect to thehydrated basaltic glass calculated with PHREEQC. Example ofthese calculations can be found in Wolff-Boenisch et al. (2004). Thedotted lines represent the percent slowdown of the basaltic glassdissolution rates due to saturation state according to Eq. (3).

-15.0

-14.5

-14.0

-13.5

-13.0

-12.5

-12.0

0 2 4 6 8

log

[mol

Si/c

m2 /

s]


calculated dissolution rate - pure water experimentcalculated dissolution rate - 22 °C CO experimentcalculated dissolution rate - 50 °C CO experimentapparent dissolution rate - pure water experimentapparent dissolution rate - 22 °C CO experimentapparent dissolution rate - 50 °C CO experiment

Fig. 12. Calculated and measured Si release rates along the flowpath at an elapsed time of 2200, 875, and 840 h for the pure water–basaltic glass interaction experiment, the 22 �C CO2-chargedwater–basaltic glass interaction experiment, and for the 50 �CCO2-charged water–basaltic glass interaction experiment, respec-tively. Open symbols represent the Si release rates calculated usingthe rate expression from Gislason and Oelkers (2003) and filledsymbols represent experimentally obtained Si release rates calcu-lated from sampled fluid compositions. These latter values werenormalized to the surface area of the basaltic glass from the bottomof the column to the specific sampling outlet.


zeolites (heulandite) (Fig. 3 in Electronic supplement 1). Theprecipitation of secondary Si-bearing minerals is also sug-gested by the apparent Si release rates, which were lowerthan those calculated using Eq. (3). Also, the apparent Sirelease rates decreased as the fluid flowed in the columnreactor (Fig. 12) consistent with secondary precipitation.Furthermore, the Na concentration also levelled off alongthe flow path indicating Na consumption by secondaryphases (e.g. zeolites). This latter observation is in contrastwith the assumption that Na is a conservative element innatural alkaline basaltic groundwater systems. Iron concen-trations measured in the column were close to or belowdetection limit. Released Fe2+ from the basaltic glass mostlikely oxidized with what O2 remained in the inlet solution,forming insoluble Fe hydroxide due to fast iron oxidationkinetics at elevated/alkaline pH (c.f. Stumm and Morgan,

1996). This assumption is corroborated by calculated fluidsupersaturation with respect to trivalent Fe-phases (Fig. 3in Electronic Supplement 1). The Fe3+ could be also incor-porated into secondary clays.

During the 22 �C CO2-charged water experiment, thepH decreased from alkaline to acidic, enhancing basalticglass dissolution (Figs. 3a and 6a). The inlet fluid pH was3.4 and its pH increased to 4.5 during the first 40 min ofwater–basaltic glass interaction. A further pH increase didnot occur because the pH buffer capacity of the fluid atpH �4.5 was approximately an order of magnitude higherthan that of the dissolving glass. Although the inlet fluidDIC concentration fluctuated somewhat during the firstpart of this experiment (Figs. 3b and 6b), the overall samplefluid pH was constant along the flow path giving the confi-dence in the measured fluid chemistry. Due to low pH, themobility of all elements increased (Fig. 10c and d) com-pared to the pure water experiment (Fig. 10a and b). Iron,manganese, and chromium became more mobile comparedto the pure water experiment. Also, other elements, whichwere close to the analytical detection limit during the purewater experiment, increased in concentration (e.g. Mn, B,Sr, Ba, As Ti) (see Figs. 5 and 8). Even though element re-lease along the flow path in this experiment was more con-stant compared to the pure water experiment, there was a


slight non-linearity in Si, Ca, Mg, Na, and Fe concentra-tions versus flow distance, suggesting either a slowing ofbasaltic glass dissolution or secondary Si-bearing phase pre-cipitation. The near to linear increase in aqueous Al con-centration along the flow path, however, suggests thatthere was no major slowing of basaltic glass dissolution,assuming that glass was the only dissolving phase. Notethat preferential release of Al could also stem from re-disso-lution of the secondary Al-bearing phases formed duringthe pure water experiment. The Si release rates obtainedfrom this experiment were also lower than those calculatedusing Eq. (3) (Fig. 12) consistent with secondary Si-bearingphase precipitation and the supersaturation of the fluidwith respect to chalcedony, quartz, and other Si-bearingphases such as clays (Fig. 3 in Electronic supplement 1).Note also the calculated stoichiometric coefficients for Si,Ca, Mg, and Fe were below 1 indicating their preferentialretention in the solid. Rogers et al. (2006) reported thatamorphous silica is the common silica phase formed duringthe low-temperature alteration of basalts by CO2-rich flu-ids. Iron chemistry showed that Fe3+ was the dominantFe aqueous species during the first 2 h of water–basalticglass interaction (Fig. 9a and b) suggesting that most ofthe Fe2+ released from basaltic glass was oxidized to Fe3+

during the initial stage of water–basaltic glass interaction,probably due to O2 remaining in the degassed inlet fluid.Later along the flow path Fe2+ became the major iron spe-cies. Furthermore, the Fe3+ concentration did not increasewith time but slowly decreased, consistent with Fe3+ precip-itation; this likelihood is consistent with the observed super-saturation of the fluid phase with respect to amorphousFe(OH)3 (Fig. 3 in Electronic supplement 1). Later alongthe flow path there was no O2 left to oxidize Fe2+, allowingthe Fe2+ concentration to increase gradually, while Fe3+

decreased. Although the iron oxidation kinetics is slow atpH <�6 as calculated based on rates reported by Stummand Morgan (1996), the measured Fe2+/Fe3+ concentra-tions at the first sampling port suggests that nearly 80%of the Fe2+ released from the basaltic glass would have beenoxidized to Fe3+during the first 40 min of CO2-chargedwater–basaltic glass interaction in the column reactor.

Element mobility in the CO2-charged water at 50 �C wassimilar to that measured at 22 �C (see Fig. 10c and e). Thistemperature increase did not enhance the basaltic glass dis-solution rates sufficiently to raise pH and it was stable at�4.2 along the flow path. After a significant initial rise dur-ing the first 2–3 h of water–basaltic glass interaction alongthe flow path, aluminium concentration decreased consis-tent with Al-bearing secondary phase precipitation(Fig. 7b and c), in accord with the supersaturation of thefirst two outlet fluids with respect to amorphous Al(OH)3

(see Fig. 3 in Electronic supplement 1). The discrepancy be-tween measured and calculated Si release rates (Fig. 12)suggests incorporation of Si into secondary phases, consis-tent with the supersaturation of the fluid with respect tochalcedony and clay minerals (see Fig. 3 in Electronic sup-plement 1). In addition, the stoichiometric coefficients formajor elements confirm their preferential retention in thesolid phase. The Cr concentration followed the Al concen-tration pattern suggesting Cr incorporation into an Al

hydroxide phase (Kaasalainen and Stefansson, 2012). Thespatial variation in reactive fluid Sr and Mn concentrationsalong the flow path might indicate incorporation of theseelements into secondary phases.

4.2. Reactive transport modelling

The results of geochemical reactive transport modellingare compared to the experimental results in Figs. 13 and 14.Modelled fluid concentrations assuming no secondary min-erals formed during the pure water–basaltic glass interac-tion experiment were higher by a factor of �8 from thosemeasured in samples collected from the seventh samplingport but by only a factor of �3 from those measured inthe first sampling outlet for Si, Ca, Mg, Na, and Al (seeFig. 13a). Corresponding modelled fluid concentrationsfor the 22 �C CO2-charged water experiment yielded con-centrations 3–4 times higher than their experimental coun-terparts for all major elements (see Fig. 13c). Modelled fluidconcentrations of the 50 �C CO2-charged water experimentwere �6 times higher than those measured in the first sam-pling outlet and �3 times higher than in the seventh sam-pling outlet for Si, Ca, Mg, Na, and Fe (see Fig. 13e).The saturation state of hydrated basaltic glass, based onthe log K of 1.07 for reaction (1) described above, did notaffect significantly the modelling results of the purewater–basaltic glass and the 22 �C CO2-charged waterexperiment. However, there was some effect of saturationstate on the modelled concentrations for the 50 �C experi-ment; the relatively high fluid saturation state slowed basal-tic glass dissolution rates somewhat. The saturation state ofhydrated basaltic glass decreased modelled fluid concentra-tions starting from the second sampling port consistentwith experimental results. The non-linearity of the modelledcurves during the 22 �C CO2-charged water experimentstemmed from the effect of increasing fluid Al concentrationalong the flow path which decreased basaltic glass dissolu-tion rate.

Result of the modelling including the assumption thatcommon secondary minerals precipitated at local equilib-rium yielded lower estimates of fluid concentrations forsome elements compared to those results obtained assum-ing no secondary minerals precipitated (see Fig. 13b, d,and f). During the pure water–basaltic glass interactionexperiment, many phases were supersaturated in the fluidphase including various clays, zeolites, and Fe and Alhydroxides. In the simulation imogolite, goethite, heulan-dite, and Mg-saponite were chosen to precipitate at localequilibrium as these are common secondary minerals foundin natural basaltic systems (e.g. Alfredsson et al., 2013) andthe thermodynamic properties of these phases are available(Gysi and Stefansson, 2011). Results of the modellingshowed that the fluid attained equilibrium with imogolite,goethite, and Mg-saponite causing a decrease in Si, Mg,Al, and Fe concentrations compared to model calculationsperformed assuming that secondary minerals did not pre-cipitate (Fig. 13b). As basaltic glass dissolution rates in-crease with decreasing aqueous Al concentration, theaqueous concentrations of some metals, including Ca andNa, increased when secondary mineral precipitation was

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

0 1 2 3 4 5 6

[mm

ol/k

g]


exp Si model Siexp Al model Alexp Fe model Feexp Mg model Mgexp Ca model Caexp Na model Na

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

0 1 2 3 4 5 6

[mm

ol/k

g]

[mm

ol/k

g]


0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0.0

0.5

1.0

1.5

2.0

2.5

0 2 4 6 8

[mm

ol/k

g]



0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0.0

0.5

1.0

1.5

2.0

2.5

0 2 4 6 8

[mm

ol/k

g]

[mm

ol/k

g]


0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0.0

0.5

1.0

1.5

2.0

2.5

0 2 4 6 8

[mm

ol/k

g]


0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0.0

0.5

1.0

1.5

2.0

2.5

0 2 4 6 8[m

mol

/kg]

[mm

ol/k

g]


(b)(a)

(d)(c)

(f)(e)

Fig. 13. Comparison of the results of reactive transport modelling calculations with measured reactive fluid concentrations for the purewater–basaltic glass interaction experiment at an elapsed time of 2200 h (a, b), for the 22 �C CO2-charged water–basaltic glass interactionexperiment at an elapsed time of 875 h (c and d), and for the 50 �C CO2-charged water–basaltic glass interaction experiment at an elapsed timeof 840 h (e and f). The symbols correspond to the experimental results. The Si concentration is shown on the left axis and the rest of theelements are displayed on the right axis. Plots (a), (c), and (e) represent reactive transport modelling assuming basaltic glass dissolvedaccording to the rate expression given by Gislason and Oelkers (2003). The logarithm of equilibrium constant for the hydrated basaltic glasswas 1.07. Plots (b), (d), and (f) represent reactive transport modelling under the same conditions but additionally assuming local equilibriumwith imogolite, goethite, heulandite, and Mg-saponite for the pure water experiment, and goethite(am), chalcedony, and gibbsite for both CO2

experiments, respectively. The right y axis scale in (d) and (f) was decreased to give better comparison between modelled concentration curvesand thus the curves are cut.


considered. This latter result stems from the fact that theprecipitation of secondary minerals decreases fluid Al con-centration leading to accelerated basaltic glass dissolutionrates as calculated using Eq. (3).

A comparison between modelled and measured fluid pHis provided in Fig. 14. Although calculated pH was similarto corresponding measured pH for the case of the 22 �CCO2-charged water experiment, differences of more than a

pH unit were evident between the modelled and measuredpH for the other two experiments.

The relative simplicity of the pure water–basaltic glassinteraction experiment allows some assessment of the sensi-tivity of the calculations to the input parameters. In the ab-sence of secondary mineral precipitation, calculated fluidcompositions are a function of three parameters: (1) thebasaltic glass dissolution rate constant, (2) the basaltic glass

6.0

6.5

7.0

7.5

8.0

8.5

9.0

9.5

10.0

10.5

11.0

0 1 2 3 4 5 6

pH



exp pHmodel dissolutionmodel dissol/precip

3.0

3.5

4.0

4.5

5.0

5.5

0 2 4 6 8

pH




3.0

3.5

4.0

4.5

5.0

5.5

0 2 4 6 8

pH




(c)(b)(a)

Fig. 14. Comparison of the reactive fluid pH obtained from reactive transport modelling calculations with those measured for the pure water–basaltic glass interaction experiment at an elapsed time of 2200 h (a), during the 22 �C CO2-charged water–basaltic glass interactionexperiment at an elapsed time of 875 h (b), and during the 50 �C CO2-charged water–basaltic glass interaction experiment at an elapsed time of840 h (c). The symbols correspond to the experimental results. The calculations were performed either assuming basaltic glass dissolution ratesthat followed the expression given by Gislason and Oelkers (2003) (‘model dissolution’) or dissolution consistent with this rate expression andassuming local equilibrium with imogolite, goethite, heulandite, and Mg-saponite for the pure water experiment, and goethite(am), chalcedony,and gibbsite for both CO2 experiments, respectively (‘model dissol/precip’).


surface–fluid interaction area, and (3) the equilibrium con-stant (logK) for the basaltic glass surface hydrolysis reac-tion. Of these parameters, the equilibrium constant islikely the poorest constrained – the equilibrium constantfor the surface layer of this glass has not been determinedexperimentally. In this study logK was estimated from thesum of amorphous oxides solubilities assuming the surfacecontains only Si and Al. Ferric ion was not included in thebasaltic glass leached layer due to its low concentration inthe glass structure (Wolff-Boenisch et al., 2004). A numberof recent studies, however, suggests that the presence of aminor quantity of Fe in a silicate structure can slow signif-icantly its dissolution rates (e.g. Daval et al., 2013; Saldiet al., 2013). The sensitivity of model results to logK varia-tion can be seen in Fig. 15. Decreasing the log K value inEq. (1) from 1.07 of �0.85 yielded a close agreement be-tween the modelled and measured element concentrationsduring pure water–basaltic glass interaction experiment.Nevertheless, if secondary mineral precipitation was consid-ered, the model calculations again failed significantly (seeFig. 15b). Using this same logK to describe the two CO2-charged water experiments also yielded a closer match be-tween modelled and measured element concentrations;however, the model results failed to reproduce the variationof fluid concentrations as a function of distance along theflow path (e.g. the strong increases in the fluid concentra-tions of numerous elements along the flow path – seeFig. 15c, d, e, and f). The description of these fluid concen-tration variations is critical to modelling the distributionand quantity of precipitated secondary minerals includingcarbonates along the flow path.

Overall there is a poor match between geochemical mod-elling and experimental results. This poor description of themeasured fluid compositions underscores the challenges ofpredicting the fate and consequences of CO2 injected intothe subsurface during carbon storage efforts. There arelikely a number of sources of these observed discrepancies.First is the poor choice of secondary minerals. The selection

of secondary phases attaining local equilibrium with thefluid controls to a large extent the mobility of modelled ele-ment concentrations and the dissolution rate of the basalticglass. A second factor influencing the modelling results isthe poor understanding of the thermodynamic propertiesof minerals such as clays and zeolites (c.f. Oelkers et al.,2009). Gysi and Stefansson (2011) attempted to constructa thermodynamic database for the clays and zeolites com-monly observed in natural basaltic systems. The diversityof the structures and compositions of clays and zeolites,however, leads to significant uncertainties in these results.Simulations performed using a distinct logK for the hy-drated basaltic glass surface layer (see Fig. 15) illustrateshow variations in thermodynamic properties can alter mod-el results. Due to lack of experimental data on the thermo-dynamic properties of the hydrated basaltic glass itsdissolution rates can be overestimated. In addition, passiv-ating layers, which can slow mineral and glass dissolutionrates are not included in the geochemical modelling calcula-tions. A third factor is the lack of precipitation rate expres-sions for secondary minerals; such are a prerequisite for theaccurate modelling of the temporal and spatial compositionof reactive fluids. Even if the fluid is supersaturated with re-spect to a secondary phase, its precipitation rates may besluggish (e.g. Saldi et al., 2009, Schott et al., 2012). Alsoas reported by Schott et al. (2012) rates close to equilibriumare dependent on the availability of reactive surface sites forthe nucleation which is closely related to the history of themineral surface which is often unknown. The limitations ofcurrent models of nucleation and precipitation kinetics arediscussed in Zhu and Lu (2009), Pham et al. (2011), andHellevang et al. (2013). Finally, errors in the estimationof glass–fluid interfacial surface area, due in part to fluidchannelling, can contribute to the discrepancies betweenmodelled and experimental outlet concentrations. Fluidchannelling would tend to lower this surface area as somesurfaces may not be exposed to fresh reactive fluid. As men-tioned by many authors, mineral–fluid interfacial surface

0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0 1 2 3 4 5 6

[mm

ol/k

g]


0.00

0.05

0.10

0.15

0.20

0.25

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0 1 2 3 4 5 6

[mm

ol/k

g]

[mm

ol/k

g]



0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.0

0.2

0.4

0.6

0.8

0 2 4 6 8

[mm

ol/k

g]


0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.0

0.2

0.4

0.6

0.8

0 2 4 6 8

[mm

ol/k

g]

[mm

ol/k

g]



0.0

0.1

0.2

0.3

0.0

0.2

0.4

0.6

0.8

1.0

0 2 4 6 8

[mm

ol/k

g]


0.0

0.1

0.2

0.3

0.0

0.2

0.4

0.6

0.8

1.0

0 2 4 6 8[m

mol

/kg]

[mm

ol/k

g]


(a) (b)

(c) (d)

(f)(e)

Fig. 15. Comparison of reactive transport modelling calculation results with measured reactive fluid concentrations for the pure water–basaltic glass interaction experiment at an elapsed time of 2200 h (a, b), for the 22 �C CO2-charged water–basaltic glass experiment at anelapsed time of 875 h (c and d), and for the 50 �C CO2-charged water–basaltic glass experiment at an elapsed time of 840 h (e and f). Thesymbols correspond to the experimental results. The Si concentration is shown on the left axis and all other concentrations are presented onthe right axis. The model calculations were performed using a logarithm of the equilibrium constant for hydrated basaltic glass surface layer of�0.85, generated by a best fit of the data shown in graph (a). Plots (a), (c), and (e) represent reactive transport modelling results obtainedassuming basaltic glass dissolved according to the rate expression given by Gislason and Oelkers (2003). Plots (b), (d), and (f) representreactive transport modelling results taking into account both this basaltic glass dissolution expression and assuming local equilibrium withimogolite, goethite, heulandite, and Mg-saponite for the pure water experiment, and goethite(am), chalcedony and gibbsite for both CO2

experiments, respectively.


area is crucial to predict not only rates measured in the lab-oratory but also the rates of natural processes (e.g. Anbeeket al., 1994; White and Brantley, 2003; Zhu et al., 2006;Maher et al., 2009; Navarre-Sitchler et al., 2011, 2013; Gysiand Stefansson, 2011; Aradottir et al., 2012; Paukert et al.,2012; Trautz et al., 2012; Zheng et al., 2012).

4.3. Implications for CO2 mineral storage

The results presented above provide insight into in situ

mineral storage efforts in basaltic rocks. As the experimentalresults and natural observations reveal, CO2 will not befixed as carbonates during the first hours of water–basalt


interaction in flow systems. In contrast, this region isdominated by primary rock dissolution. This observationfavours the industrial scale injection of CO2 into subsurfacebasalts as it will tend to increase porosity and permeabilitynear the well outlet. Nevertheless, some Si–Al bearing phaseprecipitation cannot be excluded. A potential risk associatedwith such secondary Si–Al bearing phase formation isthe passivation of the primary rock, slowing its furtherdissolution (e.g. Daval et al., 2011; Schaef et al., 2013). Theonly potential for carbonate precipitation near the injectionwell is at the beginning of CO2-charged water injection,where mixing of CO2 with natural alkaline groundwatercould lead to carbonate precipitation. This was evidencedby the reactive fluid composition during the first hours ofthe 22 �C CO2-charged water experiment, which was the onlytime during the CO2-charged experiments that the fluidbecame supersaturated with respect to carbonate minerals.The potential for carbonate formation via dilution andmixing of CO2-charged waters with alkaline groundwaterwas discussed in detail by Wolff-Boenisch (2011).

The column reactor is a natural analogue for the migra-tion of magmatic CO2 rich groundwater in the vicinityof volcanoes or geothermal systems (e.g. Flaathen andGislason, 2007; Olsson et al., 2012a). When the CO2-chargedgroundwater emerges at the surface, some of the CO2 deg-asses leading to the supersaturation of the fluid with respectto carbonate minerals provoking their precipitation (e.g.travertine deposits). During the two CO2-charged water–basalt interaction experiments �10 grams of carbonates werecalculated to have precipitated as a result of CO2 degassingat the outlet (�4 g during the 22 �C and �6 g during the50 �C experiment) suggesting that divalent cation harvestingfrom basalts using CO2-charged waters could be used forCO2 storage. Carbonates precipitation can also scavengesome metals such as Al, Fe, Cd, Cu, Mn, and Sr (e.g. Olssonet al., 2012a). Such processes could be enhanced by decreas-ing the fluid flow rate, decreasing the basaltic glass grainsize, and by elevating the temperature.

5. CONCLUSIONS

This experimental study of CO2-charged water basalticglass interaction provides a number of insights relevant tomineral carbonation efforts and natural CO2-rich systems:

1. The pH increase from 3.4 to 4.5 of the CO2-rich inletfluid does not immobilize toxic elements at ambient tem-perature but immobilizes Al and Cr at 50 �C. This indi-cates that further neutralization of CO2-charged water isrequired for decreased toxic element mobility.

2. The CO2-charged water injection enhances the mobility ofredox sensitive Fe2+ significantly making it available forthe storage of injected carbon as iron carbonate minerals.

3. The precipitation of aluminosilicates likely occurred at apH of 4.2–4.5 in CO2-charged waters. These secondaryphases can (1) fill the available pore space and thereforeclog the host rock in the vicinity of the injection well,and (2) incorporate some divalent cations limiting theiravailability for carbon storage.

4. The inability of simple reactive transport models todescribe accurately the fluid evolution in this well con-strained one dimensional flow system suggests that signif-icant improvements need to be made to such modelsbefore we can predict with confidence the fate and conse-quences of injecting carbon dioxide into the subsurface.

5. Column reactors such as that used in this study could beused to facilitate ex situ carbon mineral storage. Carbon-ate precipitation at the outlet of the reactor suggests thatthe harvesting of divalent metals from rocks using CO2-charged waters could potentially be upscaled to anindustrial carbonation process.

ACKNOWLEDGEMENTS

This manuscript benefited from insightful and constructive re-views provided by Damien Daval and two anonymous reviewers.We are grateful to Chen Zhu for his comments and editorial han-dling. This study is part of the CarbFix project (www.carbfix.com)in Iceland and was funded by the European Union through theEuropean Marie Curie network Delta-Min (Mechanisms of Min-eral Replacement Reactions; Grant PITN-GA-2008-215360) andSP1-Cooperation (FP7-ENERGY-2011-1; Grant 283148), Rey-kjavık Energy, University of Iceland and RANNIS, Icelandic Fundfor Research Equipment; Grant 10/0293 and 121071-0061). Theauthors would like to thank all colleagues and co-workers at Uni-versity of Iceland and Reykjavik Energy, in particular ÞorsteinnJonsson, Eydis Eiriksdottir, Kiflom Gebrehiwot, SnorriGudbrandsson, Helgi Alfredsson, Nicole Keller, Niels Oskarson.

APPENDIX A. SUPPLEMENTARY DATA

Supplementary data associated with this article can befound, in the online version, at http://dx.doi.org/10.1016/j.gca.2013.10.044.

REFERENCES

Aiuppa A., Federico C., Allard P., Gurrieri S. and Valenza M.(2005) Trace metal modeling of groundwater–gas–rock inter-actions in a volcanic aquifer: Mount Vesuvius, Southern Italy.Chem. Geol. 216, 289–311.

Alfredsson H. A., Oelkers E. H., Hardarsson B. S., Franzson H.,Gunnlaugsson E. and Gislason S. R. (2013) The geology andwater chemistry of the Hellisheidi, SW-Iceland carbon storagesite. Int. J. Greenhouse Gas Control 12, 399–418.

Anbeek C., Van Breemen N., Meijer E. L. and Van Der Plas L.(1994) The dissolution of naturally weathered feldspar andquartz. Geochim. Cosmochim. Acta 58, 4601–4613.

Andreani M., Luquot L., Gouze P., Godard M., Hoise E.and Gibert B. (2009) Experimental study of carbonsequestration reactions controlled by the percolation ofCO2-rich brine through peridotites. Environ. Sci. Technol.

43, 1226–1231.Appelo C. A. J. and Postma D. (2005) Geochemistry, Groundwater

and Pollution, second ed. CRC Press, Amsterdam.Aradottir E. S. P., Sonnenthal E. L., Bjornsson G. and Jonsson H.

(2012) Multidimensional reactive transport modeling of CO2

mineral sequestration in basalts at the Hellisheidi geothermalfield, Iceland. Int. J. Greenhouse Gas Control 9, 24–40.



http://refhub.elsevier.com/S0016-7037(13)00616-9/h0005
























Arnorsson S. (2003) Arsenic in surface- and up to 90 �C groundwaters in a basalt area, N-Iceland: processes controlling itsmobility. Appl. Geochem. 18, 1297–1312.

Arnorsson S., Sigurdsson S. and Svavarsson H. (1982) Thechemistry of geothermal waters in Iceland. I. Calculation ofaqueous speciation from 0 to 370 �C. Geochim. Cosmochim.

Acta 46, 1513–1532.Arnorsson S. and Andresdottir A. (1995) Processes controlling the

distribution of B and Cl in natural waters in Iceland. Geochim.

Cosmochim. Acta 59, 4125–4146.Arnorsson S., Gunnarsson I., Stefansson A., Andresdottir A. and

Sveinbjornsdottir A. E. (2002) Major element chemistry ofsurface- and ground waters in basaltic terrain, N-Iceland.: I.primary mineral saturation. Geochim. Cosmochim. Acta 66,4015–4046.

Beinlich A. and Austrheim H. (2012) In situ sequestration ofatmospheric CO2 at low temperature and surface cracking ofserpentinized peridotite in mine shafts. Chem. Geol. 332–333,32–44.

Berner R. A. (2004) The Phanerozoic Carbon Cycle: CO2 and O2.Oxford University Press.

Bourcier W. L., Peiffer D. W., Knauss K. G., McKeegan K. D. andSmith D. K. (1990) A kinetic model for borosilicate glassdissolution based on the dissolution affinity of a surfacealteration layer. Mat. Res. Soc. Symp. Proc. 176, 209–216.

Brady P. V. and Gislason S. R. (1997) Seafloor weathering controlson atmospheric CO2 and global climate. Geochim. Cosmochim.

Acta 61, 965–973.Broecker W. S. (2012) The carbon cycle and climate change: memoirs

of my 60 years in science. Geochem. Perspect. 1, 221–351.Chopping C. and Kaszuba J. P. (2012) Supercritical carbon

dioxide–brine–rock reactions in the Madison Limestone ofSouthwest Wyoming: an experimental investigation of a sulfur-rich natural carbon dioxide reservoir. Chem. Geol. 322–323,223–236.

Crovisier J. L., Honnorez J., Fritz B. and Petit J. C. (1992)Dissolution of subglacial volcanic glasses from Iceland: labo-ratory study and modelling. Appl. Geochem. 7(Suppl. 1), 55–81.

Cubillas P., Kohler S., Prieto M., Causserand C. and Oelkers E. H.(2005) How do mineral coatings affect dissolution rates? Anexperimental study of coupled CaCO3 dissolution – CdCO3

precipitation. Geochim. Cosmochim. Acta 69, 5459–5476.Daux V., Guy C., Advocat T., Crovisier J.-L. and Stille P. (1997)

Kinetic aspects of basaltic glass dissolution at 90 �C: role ofaqueous silicon and aluminium. Chem. Geol. 142, 109–126.

Daval D., Martinez I., Corvisier J., Findling N., Goffe B. andGuyot F. (2009) Carbonation of Ca-bearing silicates, the caseof wollastonite: experimental investigations and kinetic model-ing. Chem. Geol. 265, 63–78.

Daval D., Sissmann O., Menguy N., Saldi G. D., Guyot F.,Martinez I., Corvisier J., Garcia B., Machouk I., Knauss K. G.and Hellmann R. (2011) Influence of amorphous silica layerformation on the dissolution rate of olivine at 90 �C andelevated pCO2. Chem. Geol. 284, 193–209.

Daval D., Hellmann R., Saldi G. D., Wirth R. and Knauss K. G.(2013) Linking nm-scale measurements of the anisotropy ofsilicate surface reactivity to macroscopic dissolution rate laws:new insights based on diopside. Geochim. Cosmochim. 107, 121–134.

Diakonov I., Khodakovsky I., Schott J. and Sergeva E. (1994)Thermodynamic properties of iron oxides and hydroxides, I.Surface and bulk thermodynamic properties of goethite (aFeO-OH) up to 500 K. Eur. J. Mineral. 6, 967–983.

Diakonov I., Schott J., Martin F., Harrichourry J. Cl. and EscalierJ. (1999) Iron (III) solubility and speciation in aqeous solution.Experimental study and modelling. Part 1. Hematite solubility

from 60 to 300 C in NaOH–NaCl solutions and thermodynamicproperties of Fe(OH)4-(aq). Geochim. Cosmochim. Acta 63,2247–2261.

Duan Z., Sun R., Zhu C. and Chou I. M. (2006) An improvedmodel for the calculation of CO2 solubility in aqueous solutionscontaining Na+, K+, Ca2+, Mg2+, Cl�, and SO4

2�. Mar. Chem.

98, 131–139.Federico C., Aiuppa A., Allard P., Bellomo S., Jean-Baptiste P.,

Parello F. and Valenza M. (2002) Magma-derived gas influxand water–rock interactions in the volcanic aquifer of Mt.Vesuvius, Italy. Geochim. Cosmochim. Acta 66, 963–981.

Federico C., Aiuppa A., Favara R., Gurrieri S. and Valenza M.(2004) Geochemical monitoring of groundwaters (1998–2001)at Vesuvius volcano (Italy). J. Volcanol. Geotherm. Res. 133,81–104.

Flaathen T. K. and Gislason S. R. (2007) The effect of volcaniceruptions on the chemistry of surface waters: the 1991 and 2000eruptions of Mt. Hekla, Iceland. J. Volcanol. Geotherm. Res.

164, 293–316.Flaathen T. K., Gislason S. R., Oelkers E. H. and Sveinbjornsdot-

tir A. E. (2009) Chemical evolution of the Mt. Hekla, Iceland,groundwaters: a natural analogue for CO2 sequestration inbasaltic rocks. Appl. Geochem. 24, 463–474.

Fridriksson T., Kristjansson B. R., Armannsson H., Margretardot-tir E., Olafsdottir S. and Chiodini G. (2006) CO2 emissions andheat flow through soil, fumaroles, and steam heated mud poolsat the Reykjanes geothermal area, SW Iceland. Appl. Geochem.

21, 1551–1569.Galeczka I., Wolff-Boenisch D., Jonsson T., Sigfusson B., Stefans-

son A. and Gislason S. R. (2013a) A novel high pressurecolumn flow reactor for experimental studies of CO2 mineralstorage. Appl. Geochem. 30, 91–104.

Galeczka I., Wolff-Boenisch D. and Gislason S. R. (2013b)Experimental studies of basalt–H2O–CO2 interaction with ahigh pressure column flow reactor: the mobility of metals.Energy Procedia 37, 5823–5833.

Giammar D. E., Bruant, Jr, R. G. and Peters C. A. (2005)Forsterite dissolution and magnesite precipitation at conditionsrelevant for deep saline aquifer storage and sequestration ofcarbon dioxide. Chem. Geol. 217, 257–276.

Gislason S. R. and Eugster H. P. (1987) Meteoric water–basaltinteractions. II: a field study in N.E. Iceland. Geochim.

Cosmochim. Acta 51(10), 2841–2855.Gislason S. R., Arnorsson S. and Armannsson H. (1996) Chemical

weathering of basalt in Southwest Iceland; effects of runoff, ageof rocks and vegetative/glacial cover. Am. J. Sci. 296, 837–907.

Gislason S. R., Heaney P. J., Oelkers E. H. and Schott J. (1997)Kinetic and thermodynamic properties of moganite, a novelsilica polymorph. Geochim. Cosmochim. Acta 61, 1193–1204.

Gislason S. R., Snorrason A., Kristmannsdottir H. K.,Sveinbjornsdottir A. E., Torsander P., Olafsson J., Castet S.and Dupre B. (2002) Effects of volcanic eruptions on the CO2

content of the atmosphere and the oceans: the 1996 eruptionand flood within the Vatnajokull Glacier, Iceland. Chem. Geol.

190, 181–205.Gislason S. R. and Oelkers E. H. (2003) Mechanism, rates, and

consequences of basaltic glass dissolution: II. An experimentalstudy of the dissolution rates of basaltic glass as a function ofpH and temperature. Geochim. Cosmochim. Acta 67, 3817–3832.

Gislason S. R., Wolff-Boenisch D., Stefansson A., Oelkers E. H.,Gunnlaugsson E., Sigurdardottir H., Sigfusson B., Broecker W.S., Matter J. M., Stute M., Axelsson G. and Fridriksson T.(2010) Mineral sequestration of carbon dioxide in basalt: a pre-injection overview of the CarbFix project. Int. J. Greenhouse

Gas Control 4, 537–545.
































































































































































Goldberg D. S., Takahashi T. and Slagle A. L. (2008) Carbondioxide sequestration in deep-sea basalt. Proc. Natl. Acad. Sci.

105, 9920–9925.Gysi A. P. and Stefansson A. (2011) CO2–water–basalt interaction.

Numerical simulation of low temperature CO2 sequestrationinto basalts. Geochim. Cosmochim. Acta 75, 4728–4751.

Gysi A. P. and Stefansson A. (2012) Experiments and geochemicalmodeling of CO2 sequestration during hydrothermal basaltalteration. Chem. Geol. 306–307, 10–28.

Harvey O. R., Qafoku N. P., Cantrell K. J., Lee G., Amonette J. E.and Brown C. F. (2013) Geochemical implications of gasleakage associated with geologic CO2 storage-A qualitativereview. Environ. Sci. Technol. 47, 23–36.

Hellevang H., Pham V. T. H. and Aagaard P. (2013) Kineticmodelling of CO2–water–rock interactions. Int. J. Greenhouse

Gas Control 15, 3–15.Hovelmann J., Austrheim H., Beinlich A. and Anne Munz I. (2011)

Experimental study of the carbonation of partially serpenti-nized and weathered peridotites. Geochim. Cosmochim. Acta 75,6760–6779.

IPCC, 2005. IPCC special report on carbon dioxide capture andstorage, prepared by Working Group III of the Intergovern-mental Panel on Climate Change. In (eds. B. Metz, O.Davidson, H. de Coninck, C. M. Loos and L. A. Meyer).Cambridge University Press.

Kaasalainen H. and Stefansson A. (2012) The chemistry of traceelements in surface geothermal waters and steam, Iceland.Chem. Geol. 330–331, 60–85.

Kampman N., Burnside N. M., Shipton Z. K., Chapman H. J.,Nicholl J. A., Ellam R. M. and Bickle M. J. (2012) Pulses ofcarbon dioxide emissions from intracrustal faults followingclimatic warming. Nat. Geosci. 5, 352–358.

Kelemen P. B. and Matter J. (2008) In situ carbonation ofperidotite for CO2 storage. Proc. Natl. Acad. Sci. U.S.A. 105,17295–17300.

Kristmannsdottir H. (1979) Alteration of basaltic rocks byhydrothermal-activity at 100–300 �C. Dev. Sedimentol. 27,359–367.

Kristmannsdottir H. (1982) Alteration in the IRDP dill holecompared with other drill holes n Iceland. J. Geophys. Res. 87,6525–6531.

Lindberg R. D. and Runnells D. D. (1984) Ground water redoxreactions: An analysis of equilibrium state applied to Ehmeasurements and geochemical modeling. Science 225(4665),925–927.

Lu P., Fu Q., Seyfried, Jr, W. E., Hedges S. W., Soong Y., Jones K.and Zhu C. (2013) Coupled alkali feldspar dissolution andsecondary mineral precipitation in batch systems – 2: newexperiments with supercritical CO2 and implications for carbonsequestration. Appl. Geochem. 30, 75–90.

Mackenzie F. T. and Andersson A. J. (2013) The marine carboncycle and ocean acidification during phanerozoic time. Geo-

chem. Perspect. 2, 1–227.Maher K., Steefel C. I., White A. F. and Stonestrom D. A. (2009)

The role of reaction affinity and secondary minerals inregulating chemical weathering rates at the Santa Cruz SoilChronosequence, California. Geochim. Cosmochim. Acta 73,2804–2831.

Marini L. (2007) Geological Sequestration of Carbon Dioxide.

Thermodynamics, Kinetics and Reaction Path Modeling. Devel-

opments in Geochemistry, first ed. Elsevier, Amsterdam.McGrail B. P., Schaef H. T., Ho A. M., Chien Y.-J., Dooley J. J.

and Davidson C. L. (2006) Potential for carbon dioxidesequestration in flood basalts. J. Geophys. Res. 111, B12201.

Miller Q. R. S., Thompson C. J., Loring J. S., Windisch C. F.,Bowden M. E., Hoyt D. W., Hu J. Z., Arey B. W., Rosso K. M.

and Schaef H. T. (2013) Insights into silicate carbonationprocesses in water-bearing supercritical CO2 fluids. Int. J.

Greenhouse Gas Control 15, 104–118.Munz I. A., Brandvoll Ø., Haug T. A., Iden K., Smeets R., Kihle J.

and Johansen H. (2012) Mechanisms and rates of plagioclasecarbonation reactions. Geochim. Cosmochim. Acta 77, 27–51.

Navarre-Sitchler A., Steefel C. I., Sak P. B. and Brantley S. L.(2011) A reactive-transport model for weathering rind forma-tion on basalt. Geochim. Cosmochim. Acta 75, 7644–7667.

Navarre-Sitchler A. K., Cole D. R., Rother G., Jin L., Buss H. L.and Brantley S. L. (2013) Porosity and surface area evolutionduring weathering of two igneous rocks. Geochim. Cosmochim.

Acta 109, 400–413.Oelkers E. H. and Gislason S. R. (2001) The mechanism, rates and

consequences of basaltic glass dissolution: I. An experimentalstudy of the dissolution rates of basaltic glass as a function ofaqueous Al, Si and oxalic acid concentration at 25 C andpH = 3 and 11. Geochim. Cosmochim. Acta 65, 3671–3681.

Oelkers E. H., Gislason S. R. and Matter J. (2008) Mineralcarbonation of CO2. Elements 4, 333–337.

Oelkers E. H., Benezeth P. and Pokrovski G. S. (2009) Thermo-dynamic databases for water–rock interaction in reviews inmineralogy & geochemistry: thermodynamics and kinetics ofwater–rock interaction. Mineral. Soc. Am. 70, 1–46.

Olsson J., Stipp S. and Gislason S. R. (2012a) Calcium carbonateformation at the Eyjafjallajokull volcano: a carbon capture andstorage analogue. 30th Nordic Geological Winter Meeting,Reykjavik, Iceland 9–12 January 2012, p. 174.

Olsson J., Bovet N., Makovicky E., Bechgaard K., Balogh Z. andStipp S. L. S. (2012b) Olivine reactivity with CO2 and H2O on amicroscale: implications for carbon sequestration. Geochim.

Cosmochim. Acta 77, 86–97.Parkhurst D. L. and Appelo C. A. J. (1999) User’s guide to

PHREEQC (Version 2) – a computer program for speciation,batch-reaction, one-dimensional transport, and inverse geo-chemical calculations: U.S. In Geological Survey Water-

Resources Investigations Report 99-4259. p. 312.Paukert A. N., Matter J. M., Kelemen P. B., Shock E. L. and

Havig J. R. (2012) Reaction path modeling of enhanced in situCO2 mineralization for carbon sequestration in the peridotiteof the Samail Ophiolite, Sultanate of Oman. Chem. Geol. 330–

331, 86–100.Pham V. T. H., Lu P., Aagaard P., Zhu C. and Hellevang H. (2011)

On the potential of CO2–water–rock interactions for CO2

storage using a modified kinetic model. Int. J. Greenhouse Gas

Control 5, 1002–1015.Rani N., Pathak V. and Shrivastava J. P. (2013) CO2 mineral

trapping: an experimental study on the carbonation of basaltsfrom the eastern deccan volcanic province, India. Proc. Earth

Planet. Sci. 7, 806–809.Rogers K. L., Neuhoff P. S., Pedersen A. K. and Bird D. K. (2006)

CO2 metasomatism in a basalt-hosted petroleum reservoir,Nuussuaq, West Greenland. Lithos 92, 55–82.

Saldi G. D., Jordan G., Schott J. and Oelkers E. H. (2009)Magnesite growth rates as a function of temperature andsaturation state. Geochim. Cosmochim. Acta 73, 5646–5657.

Saldi G. D., Schott J., Pokrovsky O. S., Gautier Q. and Oelkers E.H. (2012) An experimental study of magnesite precipitationrates at neutral to alkaline conditions and 100–200 �C as afunction of pH, aqueous solution composition and chemicalaffinity. Geochim. Cosmochim. Acta 83, 93–109.

Saldi G. D., Daval D., Morvan G. and Knauss K. G. (2013) Therole of Fe and redox conditions in olivine carbonation rates: anexperimental study of the rate limiting reactions at 90 and150 �C in open and closed systems. Geochim. Cosmochim. Acta

118, 157–183.













































































































































Schaef H. T. and McGrail B. P. (2009) Dissolution ofColumbia River Basalt under mildly acidic conditions as afunction of temperature: experimental results relevant to thegeological sequestration of carbon dioxide. Appl. Geochem.

24, 980–987.Schaef H. T., McGrail B. P. and Owen A. T. (2010) Carbonate

mineralization of volcanic province basalts. Int. J. Greenhouse

Gas Control 4, 249–261.Schaef H. T., McGrail B. P. and Owen A. T. (2011) Basalt

reactivity variability with reservoir depth in supercritical CO2

and aqueous phases. Energy Procedia 4, 4977–4984.Schaef H. T., McGrail B. P., Owen A. T. and Arey B. W. (2013)

Mineralization of basalts in the CO2–H2O–H2S system. Int. J.

Greenhouse Gas Control 16, 187–196.Schott J. and Oelkers E. H. (1995) Dissolution and crystallization

rates of silicate minerals as a function of chemical affinity. Pure

Appl. Chem. 67, 903–910.Schott J., Oelkers E. H., Benezeth P., Godderis Y. and Franc�ois L.

(2012) Can accurate kinetic laws be created to describe chemicalweathering? CR Geosci. 344, 568–585.

Siirila E. R., Navarre-Sitchler A. K., Maxwell R. M. and McCrayJ. E. (2012) A quantitative methodology to assess the risks tohuman health from CO2 leakage into groundwater. Adv. Water

Resour. 36, 146–164.Stefansson A. and Gislason S. R. (2001) Chemical weathering of

basalts, Southwest Iceland: effect of rock crystallinity andsecondary minerals on chemical fluxes to the ocean. Am. J. Sci.

301, 513–556.Stefansson A., Arnorsson S. and Sveinbjornsdottir A. E. (2005)

Redox reactions and potentials in natural waters at disequilib-rium. Chem. Geol. 221, 289–311.

Stockmann G. J., Wolff-Boenisch D., Gislason S. R. and Oelkers E.H. (2011) Do carbonate precipitates affect dissolution kinetics?1: basaltic glass. Chem. Geol. 284, 306–316.

Stockmann G. J., Wolff-Boenisch D., Gislason S. R. and Oelkers E.H. (2013) Do carbonate precipitates affect dissolution kinetics?:2: Diopside. Chem. Geol. 337–338, 56–66.

Stumm W. and Morgan J. J. (1996) Aquatic Chemistry. Chemical

Equilibria and Rates in Natural Waters, third ed. John Wiley &Sons, New York, p. 683.

Trautz R. C., Pugh J. D., Varadharajan C., Zheng L., Bianchi M.,Nico P. S., Spycher N. F., Newell D. L., Esposito R. A., Wu Y.,Dafflon B., Hubbard S. S. and Birkholzer J. T. (2012) Effect of

dissolved CO2 on a shallow groundwater system: a controlledrelease field experiment. Environ. Sci. Technol. 47, 298–305.

Weltje G. J. and Alberts L. J. H. (2011) Packing states of idealreservoir sands: Insights from simulation of porosity reductionby grain rearrangement. Sed. Geol. 242, 52–64.

White A. F. and Brantley S. L. (2003) The effect of time on theweathering of silicate minerals: why do weathering rates differin the laboratory and field? Chem. Geol. 202, 479–506.

WHO (2008) World Health Organisation Guidelines for Drinking-water Quality. Third edition incorporating the first and secondagenda. Vol. 1, Recommendations. Geneva.

Wiese F., Fridriksson Th. and Armannsson H. (2008) CO2 fixationby calcite in high-temperature geothermal systems in Iceland.Report for Icelandic GeoSurvey ISOR, Reykjavik, ISOR-2008/003.

Wolff-Boenisch D. (2011) On the buffer capacity of CO2-chargedseawater used for carbonation and subsequent mineral seques-tration. Energy Procedia 4, 3738–3745.

Wolff-Boenisch D., Gislason S. R., Oelkers E. H. and Putnis C. V.(2004) The dissolution rates of natural glasses as a function oftheir composition at pH 4 and 10.6, and temperatures from 25to 74 �C. Geochim. Cosmochim. Acta 68, 4843–4858.

Wolff-Boenisch D., Wenau S., Gislason S. R. and Oelkers E. H.(2011) Dissolution of basalts and peridotite in seawater, in thepresence of ligands, and CO2: implications for mineral seques-tration of carbon dioxide. Geochim. Cosmochim. Acta 75, 5510–5525.

Zheng L., Apps J. A., Spycher N., Birkholzer J. T., Kharaka Y. K.,Thordsen J., Beers S. R., Herkelrath W. N., Kakouros E. andTrautz R. C. (2012) Geochemical modeling of changes in shallowgroundwater chemistry observed during the MSU-ZERT CO2injection experiment. Int. J. Greenhouse Gas Control 7, 202–217.

Zhu C., Veblen D. R., Blum A. E. and Chipera S. J. (2006)Naturally weathered feldspar surfaces in the Navajo Sandstoneaquifer, Black Mesa, Arizona: electron microscopic character-ization. Geochim. Cosmochim. Acta 70, 4600–4616.

Zhu C. and Lu P. (2009) Alkali feldspar dissolution and secondarymineral precipitation in batch systems: 3. Saturation states ofproduct minerals and reaction paths. Geochim. Cosmochim.

Acta 73, 3171–3200.

Associate editor: Chen Zhu






















































































an experimental study of basaltic glassâ€“h2oâ€“co2

Documents